Datasets:

nkandpa2
/

arxiv-papers

Dataset card Data Studio Files Files and versions Community

Split (1)

train · ~317k rows (showing the first 75.7k)

id stringlengths 9 10	text stringlengths 1 18.1M	source stringclasses 1 value	created timestamp[s]	added stringlengths 26 26	metadata dict
0807.3455	11institutetext: Leah B. Shaw Department of Applied Science, College of William and Mary, Williamsburg, VA 23187 22institutetext: Ira B. Schwartz US Naval Research Laboratory, Code 6792, Nonlinear Systems Dynamics Section, Plasma Physics Division, Washington, DC 20375 # Noise induced dynamics in adaptive networks with applications to epidemiology Leah B. Shaw and Ira B. Schwartz ###### Abstract Recent work in modeling the coupling between disease dynamics and dynamic social network geometry has led to the examination of how human interactions force a rewiring of connections in a population. Rewiring of the network may be considered an adaptive response to social forces due to disease spread, which in turn feeds back to the disease dynamics. Such epidemic models, called adaptive networks, have led to new dynamical instabilities along with the creation of multiple attracting states. The co-existence of several attractors is sensitive to internal and external fluctuations, which lead to enhanced stochastic oscillatory outbreaks and disease extinction. The aim of this paper is to explore the bifurcations of adaptive network models in the presence of fluctuations and to review some of the new fluctuation phenomena induced in adaptive networks. ## 0.1 Introduction In recent years, researchers have used a network approach in studying many systems, from networks of social contacts to the US power grid to the world wide web Barabasi99 , and a wide variety of mathematical tools have been developed to analyze static networks Newman03 ; CostaRTB07 . However, many natural systems are more complex than static network models. Both the properties of individuals (e.g., neurons, humans) and the connections between them change over time. Examples of networks where the links evolve dynamically occur in simple two state models , such as two player game theory PachecoTN06 ; SkyrmsP00 and opinion dynamics ZanetteG06 ; EhrhardtMV06 , as we will describe below. Static network models fail to capture systems in which dynamical properties are important. A new class of models, adaptive networks, has been introduced recently to address more fully the complexity of many physical systems Gross07 . In an adaptive network, the network geometry changes dynamically in response to the node characteristics, and these changes in geometry then alter the subsequent dynamics of the nodes. Many studies of adaptive networks have focused on steady state behavior, and rich new phenomena have been discovered in that context. However, the key aspect of an adaptive network is the interplay of node dynamics and network topology, which generally means that the nodes and links are evolving in time even if a steady state is reached. (Exceptions are cases where the network is evolved to a frozen state, as in, for example, Gil06 ; Benczik07 .) In this chapter, we consider the role of fluctuations in adaptive network models. One property that frequently occurs in adaptive networks is self organized criticality, a topic that is discussed in more detail in the chapters of Rohlf and Bornholdt and of Caldarelli and Garlaschelli. Because each node in the network receives dynamical information that depends on the connectivity of the entire network, this can provide global information to individual nodes and cause the system to organize itself. As a result, criticality and scale free behavior are often observed. In the first adaptive network model, that of Christensen et al., adding and removing links to match a node’s degree to the local average connectivity led to the network self organizing to a critical average connectivity, with a power law distribution of cluster sizes Christensen98 . In a model by Bornholdt and Röhl motivated by neural networks, the network again organized to a critical average connectivity, balancing the addition of new links when nodes are correlated and the removal of links when nodes are uncorrelated Bornholdt03 . Fan and Chen studied a growing network of chaotic maps, in which new nodes were linked to the most active previous nodes, and obtained scale free degree distributions Fan04 . Zhou and Kurths also obtained scale free distributions of connection weights for a network of chaotic oscillators in which the weights were evolved to increase synchrony Zhou06 . Extensions to the latter two models are described in the chapter by Chen and Kurths. None of the above-mentioned studies looked at fluctuations, but scale free effects on fluctuations have been observed previously. Bornholdt and Sneppen studied an evolving Boolean network, representing a genetic network, in which neutral mutations of the couplings, those that do not affect the network attractor, accumulate over time Bornholdt98 . The average connectivity of the network was monitored, and long periods of stasis in connectivity were interrupted by bursts of connectivity change. The stasis times followed a scale free distribution. In a model for an evolving network of chemical species, in which species that do not multiply as quickly are replaced by new random species, Jain and Krishna observed similar fluctuations in the number of populated species Jain01 . In the long time limit, all species are usually populated, but this value is punctuated by drop-out times due to the disruption of autocatalytic sets. Jain and Krishna did not do a statistical analysis of this effect, but their time series are reminiscent of punctuated equilibria. In evolutionary game theory models with perturbations, scale free distributions have been observed for the sizes of avalanches (number of nodes involved) as the system moves between stationary states (e.g., Ebel02 ; Scholz07 ). A model by Holme and Ghoshal for nodes that rewire the network to maximize their social influence and change their rewiring strategies adaptively also displayed avalanches in strategy changes, but statistics were not collected on the scaling of the avalanche size Holme06 . The degrees and cluster sizes in this model also fluctuated significantly. This model is discussed in detail in the chapter by Holme and Ghoshal. Another effect that is observed in adaptive network models for opinion formation in social networks is the existence of metastable states. Ehrhardt and Marsili studied a model in which new links were generated preferentially between nodes with a similar “opinion” or property, and the opinions were influenced by neighbor nodes Ehrhardt06 . In the limit where the nodes were at zero temperature (but links were added and removed stochastically), the system eventually approached one large connected component with uniform opinion, but it spent time in metastable states with multiple components, each with different opinion. The lifetimes of the metastable states could be understood analytically through an exact solution and stability analysis of the zero temperature case. Holme and Newman developed a model for opinion dynamics in which a parameter governed the relative frequencies of rewiring to new neighbors with identical opinions versus convincing one’s neighbors to share one’s opinion HolmeN06 . The system evolved to a frozen state containing one or more communities, where the number of communities depended on whether the parameter favored convincing or rewiring. The parameter controlled a continuous phase transition, and at the critical value the system exhibited a power law distribution in community sizes and large fluctuations in the time required to converge to the final state. The nature of the phase transition was further explained analytically by Vazquez et al. for a simpler model with only two opinion states Vazquez08 . A variety of models of opinion formation are described in detail in the chapter by Do and Gross. Other adaptive network models have considered synchrony of a network of coupled oscillators while adjusting the network connections adaptively Ito02 ; Ito03 ; Gong04 ; Gleiser06 ; Zhou06 . Gong and Leeuwen studied networks of coupled chaotic maps and added connections between correlated oscillators Gong04 . They found that the system formed a small world network with intermittent switching in the number of coherent clusters. Ito and Kaneko studied weighted networks of coupled maps and also strengthened the coupling between correlated oscillators Ito02 ; Ito03 . Here they considered in detail the role of the feedback mechanism between node and network dynamics. The model is also discussed in the chapter by Ito and Kaneko. They computed average weight matrices to determine whether there were stable structures in the network and found several phases, including a phase that was desynchronized but had a temporarily stable network structure, and a desynchronized phase with a disordered, rapidly changing network structure Ito03 . In the disordered network structure, the degree of individual nodes changed almost randomly. However, in the networks with temporarily stable structure, the nodes separated into two groups, controllers with high outdegree and others with low outdegree. The group in which a given node resided remained relatively stable over time. The node dynamics in the desynchronized phase was characterized by hopping between two groups whose dynamics were a half cycle out of phase from each other. In a feedback loop between the network geometry and node dynamics, nodes that hopped more slowly between groups tended to accumulate more connections, which led to further slowing of the hopping rate. This feedback loop was responsible for the splitting of the nodes into high outdegree and low outdegree groups. Effects such as self organized criticality, metastable states, and fluctuations in synchrony have been observed in adaptive networks. Thus far, these effects have mainly been quantified through simple network metrics such as time series of the average connectivity or clustering coefficient, or the average node properties may be tracked over time as will be discussed later in this chapter. When the network forms distinct clusters, tracking the number of clusters is also an option. However, higher order network structures are often difficult to track in a time-varying network, and it is not yet clear what are the key network properties to measure for an adaptive network. Also, in many cases the effects that have been observed have not yet been explained analytically. Further study of the fluctuations in adaptive networks is needed. In this chapter, we focus on fluctuations in a model for an epidemic spreading on an adaptive network. Epidemics have been briefly mentioned in the previous chapter. Some of the results in the present chapter have been published elsewhere Shaw08 . The layout of the paper is as follows: We describe the model and summarize key aspects of its bifurcation structure in Sections 0.2 and 0.3. Many properties can be predicted from a much lower dimensional mean field model. In Section 0.4, we focus particularly on fluctuations in the number of infection cases in the system, which is a physically important quantity. We present some additional results for phase relationships between node and link variables and for scaling of the epidemic lifetime in Sections 0.5 and 0.6. The dynamic network structure is more difficult to capture, but in Section 0.7 we discuss fluctuations in the degree of individual nodes in the system. ## 0.2 Model Gross et al. have introduced a susceptible-infected-susceptible (SIS) model on an adaptive network GrossDB06 , and Zanette and Gusmán have also studied an SIS model on an adaptive network Zanette07 . We have extended this work to a susceptible-infected-recovered-susceptible (SIRS) model Shaw08 . Although we have not chosen parameters corresponding to a particular real disease, tuning the average time a node spends in the recovered class allows us to adjust the average number of infections at the endemic steady state. Although some diseases in the past, such as plague, have eliminated as much as 50 per cent of a worldwide population, many infectious viral diseases, such as measles, mumps, and rubella, infect only 10% or less of a population at a given time Anderson91 , depending on epidemiological and social factors. Noise effects are expected to be especially prominent when the infection occurs at low levels. However, in this particular study, we restrict our attention to cases where the minimum endemic steady states are on the order of 10-40% of the population. Our model is constructed as an extension to that of Gross et al. GrossDB06 but includes the addition of a recovered class. The rate for a susceptible node to become infected is $pN_{I,\textrm{nbr}}$, where $N_{I,\textrm{nbr}}$ is the number of infected neighbors the node has. The recovery rate for an infected node is $r$. We fix $r=0.002$ throughout this chapter. A recovered node becomes susceptible again with rate $q$, the resusceptibility rate. Since the mean time spent in the recovered state is $1/q$, there is a natural limiting case for the SIRS model. Using the recovery rate, $r$, as a natural time scale, as the ratio $q/r$ becomes sufficiently large, nodes spend less time in the recovered state. In the limit $q/r\rightarrow\infty$, the model thus approaches the SIS model. The study of the stochastic dynamics of the SIRS model may then be examined with respect to changes in $q$. Rewiring of the network occurs as the epidemic spreads. If a link connects an infected node to a non-infected node, the link is rewired with rate $w$. The connection to the infected node is broken, and the original non-infected node is now connected to another non-infected node which is randomly selected out of all candidates in the network (excluding self links and multiple links between nodes). This rewiring rule is that of Gross et al. GrossDB06 (we treat the recovered nodes in the same manner as susceptibles for rewiring purposes), in contrast to the rewiring scheme of Zanette and Gusmán Zanette07 , which allows susceptible nodes to connect to infectives. We performed Monte Carlo simulations of this model for a system with $N$ nodes and $K$ links, where $K/N$ was fixed at 10. Details of the simulation procedure can be found in Shaw08 . Random sequential updating was used, and each node and eligible link had an opportunity to transition on average once per Monte Carlo step (MCS). As in GrossDB06 , we developed a corresponding mean field model using a moment closure approximation to track the dynamics of nodes and links. $P_{A}$ denotes the probability for a node to be in state $A$, and $P_{AB}$ denotes the probability for a link to connect a node in state $A$ to a node in state $B$. For higher order correlations, we assume $P_{ABC}\approx P_{AB}P_{BC}/P_{B}$. The time evolution of the node states is described by: $\displaystyle\dot{P}_{S}$ $\displaystyle=$ $\displaystyle qP_{R}-p\textstyle{\frac{K}{N}}P_{SI}$ (1) $\displaystyle\dot{P}_{I}$ $\displaystyle=$ $\displaystyle p\textstyle{\frac{K}{N}}P_{SI}-rP_{I}$ (2) $\displaystyle\dot{P}_{R}$ $\displaystyle=$ $\displaystyle rP_{I}-qP_{R}$ (3) The time evolution of the links is described by: $\displaystyle\dot{P}_{SS}$ $\displaystyle=$ $\displaystyle qP_{SR}+w\textstyle{\frac{P_{S}}{P_{S}+P_{R}}}P_{SI}-2p\textstyle{\frac{K}{N}}\frac{P_{SS}P_{SI}}{P_{S}}$ (4) $\displaystyle\dot{P}_{SI}$ $\displaystyle=$ $\displaystyle 2p\textstyle{\frac{K}{N}}\frac{P_{SS}P_{SI}}{P_{S}}+qP_{IR}-rP_{SI}-wP_{SI}$ (5) $\displaystyle-p\left(P_{SI}+\textstyle{\frac{K}{N}}\frac{P_{SI}^{2}}{P_{S}}\right)$ $\displaystyle\dot{P}_{II}$ $\displaystyle=$ $\displaystyle p\left(P_{SI}+\textstyle{\frac{K}{N}}\frac{P_{SI}^{2}}{P_{S}}\right)-2rP_{II}$ (6) $\displaystyle\dot{P}_{SR}$ $\displaystyle=$ $\displaystyle rP_{SI}+w\textstyle{\frac{P_{R}}{P_{S}+P_{R}}}P_{SI}+2qP_{RR}-qP_{SR}$ (7) $\displaystyle-p\textstyle{\frac{K}{N}}\frac{P_{SI}P_{SR}}{P_{S}}+w\frac{P_{S}}{P_{S}+P_{R}}P_{IR}$ $\displaystyle\dot{P}_{IR}$ $\displaystyle=$ $\displaystyle 2rP_{II}+p\textstyle{\frac{K}{N}}\frac{P_{SI}P_{SR}}{P_{S}}-qP_{IR}-rP_{IR}$ (8) $\displaystyle-wP_{IR}$ $\displaystyle\dot{P}_{RR}$ $\displaystyle=$ $\displaystyle rP_{IR}-2qP_{RR}+w\textstyle{\frac{P_{R}}{P_{S}+P_{R}}}P_{IR}$ (9) We integrated the mean field equations numerically and tracked their steady states using a continuation package auto . We have also considered a stochastic mean field system with internal fluctuations, modeled by multiplicative noise, or with external fluctuations, modeled by additive noise. The stochastic mean field system was studied using a fourth order Runge-Kutta solver. ## 0.3 Bifurcation structure In Shaw08 , we considered the bifurcation structure of the mean field for $q$ small and mapped the regions of stability for infectives as a functions of the parameters $w$ and $p$. We discovered and reported that there were regions of $q$ in which different bifurcation scenarios existed, as well as regions of bistability. In the first case, the value of $q=0.0064$, the disease free equilibrium became unstable as the infection rate $p$ was increased, and through a transcritical bifurcation was connected to the unstable branch of endemic states. A stable branch of endemic states was then connected to the unstable branch via a saddle-node bifurcation. However, in the second case, that of $q=0.0016$, there is no saddle-node bifurcation. Instead there exists a saddle-saddle connection. In this case, the unstable endemic state emanating from the disease free state has a one-dimensional unstable manifold. This branch is connected via a turning point to an unstable endemic branch having a two-dimensional unstable manifold, which in turn becomes stable through a Hopf bifurcation. The bifurcation diagram for this case is reproduced here for clarity of discussion in Fig. 1. For the saddle-saddle case, the lower (upper) branch has a one (two) dimensional unstable manifold. The upper branch then undergoes a reverse Hopf bifurcation. The connecting branch of periodic orbits (not shown) is unstable and sub-critical. These orbits have very long periods and large swings in amplitudes of infectives. Recalling that $q$ controls the resuscepibility rate and $w$ the rewiring rate, we examine the structure of the bifurcation onset of attracting endemic states while holding the other parameters fixed. The onset of Hopf bifurcation points in two parameters was computed, and is shown in Fig. 2, for both the mean field model and the full system. Bifurcation points in the full system were estimated as the largest $w$ value for which a single run started near the probable endemic steady state remained near that state for $10^{5}$ MCS without dying out. The region below the curve contains stable endemic branches, while the region above contains unstable endemic states and/or stable disease free equilibria. The same trends are observed in both the mean field and the full system. Notice that for $q$ greater than approximately $0.3$, the value of $w$ for the Hopf bifurcation does not change much. In addition, the infective fraction is also approximately independent of $q$ for $q$ sufficiently large, signifying that the model is approaching the SIS model. One would typically expect that as the resusceptibility rate $q$ increases, the number of nodes that are in the recovered state and thus protected from infection will decrease, and the infection will spread more easily. Therefore, a faster rewiring rate (larger $w$) will be needed to suppress the infection. Indeed, this is the trend observed in Fig. 2 for small and large $q$. However, the $w$ value for the bifurcation decreases with increasing $q$ between about 0.1 and 0.2. This nonmonotonic shift in the bifurcation point in Fig. 2 is a nonlinear effect which has not yet been explained. In later sections we will explore the fluctuations of the SIRS model for values of $q$ between the values 0 and 1. Effects on the fluctuations of the nonmonotonic bifurcation curve in Fig. 2 have not been observed. We remark that a direct comparison in Shaw08 of the infective fraction between the mean field model and Monte Carlo simulation of the full system showed excellent agreement along the attracting branches. The discrepancies occurred near the bifurcation point where the endemic state loses stability, partly because the actual location of the instability in the full stochastic system was difficult to detect accurately in Monte Carlo simulations and partly due to inaccuracies in the mean field approximation. However, the scaling results near the bifurcation, which we present in the next section, are generally consistent between the mean field model and the full system. Figure 1: A a bifurcation diagram of the infective fraction as a function of $p$, with $w=0.04$, $r=0.002$, $q=0.0016$. The squares denote the saddle- saddle point and transcritical point. Dashed lines are unstable branches. As $p$ is decreased, the endemic state loses stability in a Hopf bifurcation. Figure 2: A two parameter diagram of the Hopf bifurcation points as a function of $q$ and $w$. Solid curve: mean field model; points and dashed curve: full system. Parameters used are $r=0.002,p=0.004.$ ## 0.4 Effect of recovered class on fluctuations We have observed that the amplitude of fluctuations in the number of infectives is generally larger in the SIRS model than the SIS model. In the SIS system, links between two infectives are not broken because rewiring operates only on SI links. Thus when an infective becomes susceptible again, the newly formed susceptible may be connected to other infectives that it retained as neighbors while previously infected and can immediately become reinfected. This situation tends to suppress fluctuations, because small decreases in the total number of infectives correspond exactly to increases in the number of susceptibles, and rapid reinfection of the susceptibles can occur, preventing the number of infectives from dropping significantly. In the SIRS model, on the other hand, the recovered compartment introduces an effective time delay from recovery to possible reinfection and allows infective levels to fluctuate more. Figure 3 compares the scaling of fluctuations near the bifurcation point for two different values of the resusceptibility rate $q$. In the top panels $q=0.0016$, the rate used in Shaw08 . In the bottom panels $q=1$, effectively approximating the SIS case, since individuals spend very little time in the recovered class and much less than 1% of the population is in the recovered class at a given time. Fluctuations in the infectives (measured as the standard deviation divided by mean for long Monte Carlo simulations) are plotted as a function of $p$, the infection rate, as $p$ is swept towards the bifurcation point. Results were computed from $5\times 10^{5}$ MCS time series sampled every 10 MCS. The magnitude of the fluctuations is greater for the SIRS case (Fig. 3a) than for the SIS case (Fig. 3c). Notice that the increase in fluctuations is almost an order of magnitude. Figure 3: Fluctuations in infectives (standard deviation divided by mean) vs. infection rate $p$ near the bifurcation point, from Monte Carlo simulations: (a) $q=0.0016$ (SIRS case), (c) $q=1$ (approximately SIS case). Curves are to guide the eye. Log-log plots (data points with best fit lines) show power law scaling for both $q=0.0016$ (b) and $q=1$ (d). Other parameters: $w=0.04,r=0.002$. Parts (a) and (b) are reprinted from Shaw08 . The fluctuations exhibit power law scaling, shown in the log-log plots in Fig. 3b and 3d. On the horizontal axis, we plot $\ln(p-p_{c})$, where $p_{c}$ is the critical point at which the endemic state loses stability. The bifurcation points are not known exactly, so we approximate $p_{c}$ by the value that produces the most linear plot in each case. For the data of Figure 3, the scaling exponents are similar (-0.59 versus -0.51 for $q=0.0016$ and 1 respectively), so it is not clear that the resusceptibility rate has a significant effect on how the fluctuations scale with $p$. The power law scaling of the fluctuations can be understood by considering the scaling near a generic bifurcation point. From our mean field analysis, we expect the bifurcation point where the endemic steady state loses stability to be either a saddle-node bifurcation point or a Hopf point. A generic saddle- node bifurcation exhibits power law scaling of fluctuations near the bifurcation point, as we show in Fig. 4 and explain in the discussion below. A Hopf bifurcation can also appear locally to have power law scaling of fluctuations, although the scaling may be over smaller range of parameters. For a given standard deviation, the probability density function near a Hopf bifurcation is given by arnold : $p_{hb}(\beta,r,\sigma,R)=N{r}^{\,{\frac{\beta}{{\sigma}^{2}}}}{e^{-{\frac{R{r}^{2}}{{\sigma}^{2}}}}}{\sigma}^{-2}\left[\Gamma\left(1+1/2\,{\frac{\beta}{{\sigma}^{2}}}\right)\right]^{-1}$ (10) where the state space variable is a radial coordinate $r$, $\beta$ is the distance from the Hopf bifurcation point, parameter $R=0.5$ is fixed, $\Gamma$ is the gamma function, and $N$ is a normalization constant. An example of the fluctuations using Eq. 10, where we have computed the first and second order moments to find the ratio of the standard deviation to the mean, is shown in Fig. 4b. The data for the fluctuation size deviates from a power law scaling, as exhibited in the figure. However, there is a monotonic relationship in the fluctuations as measured by $\sigma/\mu$ as a function of the distance to the Hopf bifurcation. Therefore, we expect the fluctuation characteristics of the SIS and SIRS models to hold near bifurcation points regardless of whether the statistics are measured with respect to a Hopf or saddle-node bifurcation. Since the power law scaling is observed near either a saddle-node or Hopf bifurcation point, it may be understood by considering the local dynamics. For example, near the saddle-node, a center manifold reduction would reduce the study of the vector field to a system with a one-dimensional unstable manifold. Therefore, the power law scaling of the fluctuations can be motivated by considering the following simple stochastic differential equation $dx_{t}=(a-x_{t}^{2})dt+\sigmadW_{t}$ (11) for a one dimensional saddle-node bifurcation, where $a$ is the bifurcation parameter, $dW/dt$ is a white noise term, and $dW$ is a Brownian increment. In general, noise can cause a shift in the location of the saddle-node bifurcation, so we assume that the noise is sufficiently small that the location is fixed. By assuming we are always near the attracting branch of the saddle-node or Hopf bifurcation, we are in a near equilibrium setting driven by noise. Such an assumption allows us to examine the stationary probability density function (PDF) of the stochastic dynamics by employing the Fokker-Planck equation near steady state. For the stochastic differential equation, Eq. 11, the PDF is well known Horsthemke83 and is given by $p(a,x,\sigma)=Ne^{2(ax-x^{3}/3)/\sigma^{2}}.$ (12) Here $N$ is a normalization constant. From Eq. 12, we compute the first and second order moments to find the ratio of the standard deviation to the mean. We examine the fluctuations in the neighborhood of $a=0$, which is the location of the saddle-node point. The results display power law scaling, as depicted in Fig. 4a. Figure 4: (a) Fluctuation size of a generic saddle-node bifurcation as a function of bifurcation parameter $a$ near the bifurcation point using the probability density function in Eq. 12. Noise amplitude is $\sigma=0.005$. (b) Fluctuation size of a generic Hopf bifurcation as a function of bifurcation parameter $\beta$ near the bifurcation point using the PDF in Eq. 10 (squares). The line is a best linear fit. Noise amplitude is $\sigma=0.05$. To further examine the differences in fluctuations between the SIS and SIRS adaptive network models, we perform the following experiment. As discussed above, we can examine the fluctuation sizes as a function of resusceptibility rate $q$ to see how the fluctuation sizes compare between the two model classes. The other variable which controls the recovered, as well as the infected, populations is the rewiring rate, $w$. It has a significant effect on the fluctuations, since the degree of infectives is dramatically reduced by the rewiring. We examine the interplay between $q$ and $w$ and their effect on fluctuation sizes. Here we turn to a stochastic version of the mean field model. We have shown previously that the scaling behavior of the mean field model is typically similar to that of the full network system Shaw08 . We use additive noise to model fluctuations near the endemic equilibrium state. Details may be found in Shaw08 . The stochastic mean field model has the following form: $\bf{X}^{\prime}=\bf{F(X)}+\epsilon\bf{\eta}(t),$ (13) where $\bf{F(X)}$ is the mean field system in Eqs. 1-9, $\eta(\bf{t})$ is a noise term with $\langle\bf{\eta}(t)\bf{\eta}(t^{\prime})\rangle=\delta(t-t^{\prime})$, and $\epsilon$ is the noise amplitude. Fluctuations are computed by averaging the standard deviation over mean results for time series starting from 10 random initial conditions near steady state. The runs were computed for $5\times 10^{7}$ steps using a step length of 0.001, and transients were removed after $10^{6}$ steps. A typical example of the fluctuations as $w$ is varied is shown in Fig. 5b. Similar linear log-log behavior is observed in other stochastic simulations for other values of $q$. In Fig. 5b, $w_{h}$ denotes the location of the Hopf bifurcation branch. The Hopf bifurcation occurs for all values of $q$ considered here. A typical bifurcation plot is shown in Fig. 5a for $q=0.1$. Attracting states are solid curves, while unstable states are dashed and dotted curves. The Hopf bifurcation point is on the upper branch separating the stable and unstable steady states. Figure 5: (a) A bifurcation plot of the mean field model without noise. Plotted is the fraction infected as a function of $w$. Parameters are $r=0.002,p=0.004,q=0.1$. (b) Plot of the fluctuation sizes as a function of rewiring rate, $w$. The fixed parameters used are $r=0.002,\,p=0.004,\>K/N=10,\>q=0.01,\epsilon=0.0001$. Because of the power law scaling of the fluctuations, as in Figure 5b, we expect a functional relationship of the form $\sigma(q)/\mu(q)\propto\left[w_{h}(q)-w\right]^{m(q)}$ (14) where $m(q)$ is the average slope of the log-log plots. We can now examine how the average rate of change of fluctuations varies as a function of $q.$ The results are shown in Fig. 6. At smaller $q$ values, the fluctuations increase more quickly with $w$ than they do in the large $q$ limit. Therefore, the fluctuations are more sensitive with respect to $w$ in the SIRS model than in the SIS model. We attempted to confirm this mean field result for the slopes using the full model, but in the case of the full model, the exact locations of bifurcation points are unknown. It is difficult to estimate where the endemic state loses stability from time series because one cannot always distinguish a metastable state from a stable state in the presence of fluctuations. We can approximate the bifurcation point by the value that gives the most linear plot (largest $R$ value) for fluctuations vs. the bifurcation parameter as in Figure 3, but this approach is unreliable if the scaling deviates from a power law, as occurs in Figure 4b for a generic Hopf bifurcation. The best fit slope depends sensitively on the estimate for the bifurcation point, so we were not able to obtain robust results for the full system. Figure 6: A plot of the slope $m(q)$ as a function of $q.$ See text for details. Parameters used are $r=0.002,p=0.004,\epsilon=0.0001$. ## 0.5 Delayed outbreaks We have also considered phase relationships between the fluctuating node and link variables. At each time point in our simulations, we tracked the number of infected nodes as well as the number of non-infected neighbors of infected nodes (which corresponds to the number of SI and IR links). The rewiring causes the fluctuations in the number of infectives to lag behind fluctuations in the number of infective neighbors, as shown in Figure 7a. This effect was observed both in the mean field model (data not shown, see Shaw08 ) and in Monte Carlo simulations of the full system. We studied the dependence of this phase lag on both the rewiring rate $w$ and the resusceptibility rate $q$ when the system was fluctuating around the endemic steady state. Monte Carlo simulations were sampled every 1 MCS for $3\times 10^{4}$ MCS after discarding transients. We computed cross correlations between the infectives and the infective neighbors for varying phase shifts between the two time series and identified the lag maximizing the cross correlation. Figure 7b shows results for three different $q$ values. (Note: Curves for smaller $q$ terminate at lower $w$ values because the endemic steady state becomes unstable, as in Figure 2.) In each case, the lag increases with increasing rewiring rate. This effect does not have a simple explanation, but since it is also observed in mean field simulations, it depends on node and link dynamics primarily, rather than higher order geometries. Figure 7: Delayed outbreaks due to rewiring. (a) Monte Carlo time series. Black: infectives; gray: neighbors of infectives. Curves are scaled in arbitrary units for comparison of peak times. $p=0.0065,w=0.09,q=0.0016,r=0.002$. Reprinted from Shaw08 . (b) Time in MCS by which infectives lag behind infective neighbors vs. rewiring rate. Solid black: $q=0.0016$; gray: $q=0.001$; dashed black: $q=0.0005$. Other parameters: $p=0.0065,r=0.002$. Further, the lag time increases as the resusceptibility rate $q$ decreases. This occurs because the recovered class introduces an effective delay in the system. When a node becomes at risk because its neighbor is infected, it cannot itself become infected until it is susceptible. As $q$ is lowered, the fraction of infective neighbors that are recovered and have to wait to become susceptible again increases, and the average wait time also increases, so it is expected that the infective fluctuations will lag further behind. It should be noted that when $q=1$ and the system approximates the SIS model, the number of infectives and non-infected neighbors of infectives (i.e., SI links) are poorly correlated for any shift between time series. Therefore, the lags discussed here are not observed in the SIS model. ## 0.6 Lifetime of the endemic steady state Another effect we consider, which depends on fluctuations in the system, is the lifetime of the endemic steady state. Because the system is stochastic and the disease free state is absorbing, the disease will die out in the infinite time limit for any set of parameters. For a generic saddle-node bifurcation in one dimension, the scaling of the lifetime is expected to obey $\ln{T}\propto(p-p_{0})^{3/2},$ (15) where $T$ is the mean dwell time or lifetime of the steady state, $p$ is the bifurcation parameter, and $p_{0}$ is the location of the bifurcation point Dykman1980 ; Graham1987a . Using the computational methods in Shaw08 , we show preliminary results for the dependence of the lifetime on the infection rate $p$ in Figure 8. The bifurcation point $p_{0}$ was estimated by the value that gave the most linear plot for $\ln T$ vs. $(p-p_{0})^{3/2}$. The scaling results appear consistent with expectations, but further study is needed. In contrast to the mean field model, because the exact location of bifurcation points is not known for the full system, details such as slopes and scaling exponents can be very much dependent on estimates for the bifurcation point. Figure 8: Dependence of endemic state average lifetimes $T$ on infection rate $p$. Points: Monte Carlo simulations; line: best fit line. $q=0.0016$, $r=0.002$, $w=0.04$, $N=4\times 10^{4}$, $K=4\times 10^{5}$. (Reprinted from Shaw08 .) ## 0.7 Network geometry Defining appropriate statistics to capture a fluctuating network geometry is difficult. Here, the network does not display community structure nor is governed by an underlying spatial structure. Links are rewired to any acceptable target nodes, regardless of distance away. Because the mean field theory for nodes and links captures the dynamics of the system fairly well, higher order correlations involving three or more nodes do not have a large impact on the dynamics. The network may be fairly unstructured at the higher levels. To demonstrate the role of fluctuations in the network geometry on the scale of individual nodes and links, we show in Figure 9 the time-varying degree of a single arbitrarily chosen node. When the node becomes infected, its non- infected neighbors quickly rewire away from it (dashed gray curves). Because infected neighbors do not rewire away, the degree may not drop all the way to zero before pausing. If the infected node remains infected for sufficiently long, its neighbors will recover and then rewire away, further decreasing the degree. Once the node recovers, other S and R nodes in the system may rewire to it, and its degree begins to climb (black curves). When the node becomes susceptible, the degree continues increasing (solid light gray curves) until the node again becomes infected, and the cycle repeats. Figure 9: Degree of a single node versus time. Curves indicate the node’s disease status: black: recovered; light gray: susceptible; dashed medium gray: infected. Parameters: $p=0.002$, $q=0.0016$, $r=0.002$, $w=0.04$. It is not yet known how to predict the degree distributions from first principles Shaw08 , but if one assumes that the degree distributions are already known for each node class, the fluctuations in the degree of a single node can be easily understood. Figure 10 shows the statistics of the local maxima and minima of the degree time series for a single node. Results are computed for a $4\times 10^{6}$ MCS time series, which contains approximately 2000 SIRS transition cycles (and thus approximately 2000 maxima and minima). Frequency distributions for the maxima (Fig. 10a) and minima (Fig. 10b) are shown. Figure 10: Statistics for degree time series. Black curves: observed distributions; dashed gray curves: expected distributions. (a) Distribution of local maximum degrees. The degree distribution for susceptibles is shown in light gray for reference. (b) Distribution of local minimum degrees. Parameters: $p=0.002$, $q=0.0016$, $r=0.002$, $w=0.04$. The minima can be most easily understood. Minima occur when an infective recovers. Recovery is governed by the rate $r$ and is equally likely to occur for any infective, regardless of the degree. Therefore, the distribution for the degree minima is the same as the degree distribution for infectives. Figure 10b shows good agreement between the observed distribution of minima and that expected from the infective degree distribution (which was found from Monte Carlo simulations in Shaw08 ). Degree maxima occur immediately before a susceptible becomes infected. The infection rate depends on the number of infected neighbors a susceptible node has, which is almost directly proportional to its degree. (See Shaw08 for details.) Letting $d$ represent the degree of a susceptible and $P_{d}$ the degree distribution for susceptibles, we thus expect the infection rate to be proportional to $d$ and the distribution of degree maxima to be proportional to $dP_{d}$. This expected distribution is shown in Figure 10a, and there is again good agreement with the observed distribution of maxima and the expected values. The degree distribution for susceptibles is shown for reference. The distribution of maxima is skewed to higher degrees because of the dependence of the infection process on node degree. To develop a complete understanding of these processes, a theory to predict the degree distributions from first principles is needed. Such a theory must account for correlations between the infection status of an infective and its neighbors, as explained in Shaw08 . ## 0.8 Conclusions and discussion In this chapter, we considered a model of an adaptive network and its fluctuations. We introduced a model based on an SIRS epidemic structure which included transition probabilities between node states as well as link dynamics. In this model, the link dynamics are a function of the state variables, and since the state variables depend on the links, it forms a closed feedback system between nodes and links. The model is an extension of, and contains in the limit of large resusceptibility rate $q$, the SIS model studied in GrossDB06 . The fluctuations of the model were simulated in two ways: The full system was studied via Monte Carlo simulation on a finite population. In addition, a low dimensional approximation was studied using a Langevin simulation with an additive noise term to the mean field equations. Quantifying where the system is most sensitive to fluctuations required an examination of the bifurcation structure of the deterministic mean field equations. For the steady states of the mean field equations, we examined the locations of both Hopf bifurcations and saddle-node points. We saw that as the resusceptibility rate $q$ changes, the type of bifurcation changes. In general, for large $q$, we have a generic saddle-node bifurcation, while for small $q$ we have a saddle-saddle bifurcation giving rise to a Hopf bifurcation. Fluctuations were examined with respect to these bifurcation, and particular attention was paid to the large and small $q$ cases. This led us to examine the specific role of the recovered class in the SIRS model as compared with the limiting case of the SIS model. In the SIS case, we found that without the recovery class, newly created susceptibles may be still connected to another infective, and thus may become reinfected immediately. This mechanism led to a reduction in the fluctuations of infectives. On the other hand, the inclusion of the recovery class introduced a mean delay time prior to potential reinfection, thereby increasing infective fluctuations. By examining the fluctuation sizes near the bifurcation points, we found the existence of scaling laws in both the mean field model and the full system. Comparing the stochastic dynamics of SIRS and SIS cases, we examined the effect of the rewiring rate $w$ and resusceptibility rate $q$ on the fluctuations. For a large range of $q$ values, we showed the existence of a scaling law near the Hopf bifurcation, which includes the fluctuations for the limiting SIS case. We found that for small $q$ values, the fluctuations change more rapidly with rewiring rate $w$ than they do for the SIS model, making fluctuations more sensitive with respect to parameters in the SIRS case. Other effects, such as latency, or delay, between infective nodes and non-infected neighbors occurs in the SIRS model but not the SIS model. The degree fluctuations are still difficult to predict, although some of these phenomena can be understood in certain cases. However, much work still is required to understand the fluctuations of the network geometry. Current tools for the analysis of static networks are insufficient to make predictions about adaptive networks. A complete understanding of the dynamics, fluctuations, and geometry in the future requires tools which incorporate topology, stochastic dynamics, and time dependent graph theory. Because adaptive networks based on epidemiology contain many of the features of adaptive networks in general, we expect them to continue to reveal new and interesting dynamic phenomena as they are extended for more detailed modeling of social situations, including and beyond those of infectious disease spread. ## 0.9 Acknowledgments This work was supported by the Office of Naval Research and the Armed Forces Medical Intelligence Center. LBS was supported by the Jeffress Memorial Trust. ## References [1] R. M. Anderson and R. M. May. Infectious Diseases of Humans. Oxford University Press, 1991. * [2] L. Arnold. Random Dynamical Systems. Springer, New York, 2001. * [3] A. Barabási and R. Albert. Emergence of scaling in random networks. Science, 286(5439):509–512, 1999. * [4] I. J. Benczik, S. Z. Benczik, B. Schmittmann, and R. K. P. Zia. Lack of consensus in social systems. http://arxiv.org/abs/0709.4042, 2007. * [5] S. Bornholdt and T. Röhl. Self-organized critical neural networks. Physical Review E, 67(6):066118, 2003. * [6] S. Bornholdt and K. Sneppen. Neutral mutations and punctuated equilibrium in evolving genetic networks. Physical Review Letters, 81(1):236–239, 1998. * [7] K. Christensen, R. Donangelo, B. Koiller, and K. Sneppen. Evolution of random networks. Physical Review Letters, 81(11):2380, 1998. * [8] L. D. Costa, F. A. Rodrigues, G. Travieso, and P. R. V. Boas. Characterization of complex networks: A survey of measurements. Advances In Physics, 56:167–242, 2007. * [9] E. J. Doedel, R. Paffenroth, A. Champnets, T. Fairgrieve, Y. A. Kuznetsov, B. Sandstede, and X. Wang. AUTO: Software for continuation and bifurcation for ordinary differential equations, 2001. * [10] M. I. Dykman and M. A. Krivoglaz. Fluctuations in non-linear systems near bifurcations corresponding to the appearance of new stable states. Physica A, 104(3):480–494, 1980. * [11] H. Ebel and S. Bornholdt. Coevolutionary games on networks. Phys. Rev. E, 66(5):056118, 2002. * [12] G. C. M. A. Ehrhardt, M. Marsili, and F. V. Redondo. Phenomenological models of socioeconomic network dynamics. Physical Review E (Statistical, Nonlinear, and Soft Matter Physics), 74(3):036106, 2006. * [13] G. C. M. A. Ehrhardt, M. Marsili, and F. Vega-Redondo. Phenomenological models of socioeconomic network dynamics. Physical Review E, 74:036106, 2006. * [14] Z. Fan and G. Chen. Evolving networks driven by node dynamics. Internationl. Journal of Modern Physics B, 18:2540–2546, 2004. * [15] S. Gil and D. H. Zanette. Coevolution of agents and networks: Opinion spreading and community disconnection. Physics Letters A, 356(2):89–94, 2006. * [16] P. M. Gleiser and D. H. Zanette. Synchronization and structure in an adaptive oscillator network. European Physics Journal B, 53:233–238, 2006. * [17] P. Gong and C. van Leeuwen. Evolution to a small-world network with chaotic units. Europhys. Lett., 67:328–333, 2004. * [18] R. Graham and T. Tél. Nonequilibrium potentials for local codimension-2 bifurcations of dissipative flows. Physical Review A, 35(3):1328–1349, 1987. * [19] T. Gross and B. Blasius. Adaptive coevolutionary networks: a review. Journal of the Royal Society. Interface, 2007. DOI: 10.1098/rsif.2007.1229. * [20] T. Gross, C. J. D. D’Lima, and B. Blasius. Epidemic dynamics on an adaptive network. Physical Review Letters, 96:208701, 2006. * [21] P. Holme and G. Ghoshal. Dynamics of networking agents competing for high centrality and low degree. Physical Review Letters, 96(9):098701, 2006. * [22] P. Holme and M. E. J. Newman. Nonequilibrium phase transition in the coevolution of networks and opinions. Physical Review E, 74(5):056108, 2006. * [23] W. Horsthemke and R. Lefever. Noise-Induced Transitions: Theory and Applications in Physics, Chemistry, and Biology. Springer Series in Synergetics , Vol. 15, 1983. * [24] J. Ito and K. Kaneko. Spontaneous structure formation in a network of chaotic units with variable connection strengths. Physical Review Letters, 88(2):028701, 2002. * [25] J. Ito and K. Kaneko. Spontaneous structure formation in a network of dynamic elements. Physical Review E, 67(4):046226, 2003. * [26] S. Jain and S. Krishna. A model for the emergence of cooperation, interdependence, and structure in evolving networks. Proceedings of the National Academy of Science, 98:543–547, 2001\. * [27] M. E. J. Newman. The structure and function of complex networks. SIAM Review, 45(2):167–256, 2003. * [28] J. M. Pacheco, A. Traulsen, and M. A. Nowak. Coevolution of strategy and structure in complex networks with dynamical linking. Physical Review Letters, 97:258103, 2006. * [29] J. C. Scholz and M. O. W. Greiner. Topology control with ipd network creation games. New Journal of Physics, 8:185–199, 2007. * [30] L. B. Shaw and I. B. Schwartz. Fluctuating epidemics on adaptive networks. Physical Review E, 77:066101, 2008. * [31] B. Skyrms and R. Pemantle. A dynamic model of social network formation. Proceedings of the National Academy of Sciences, 97:9340–9346, 2000\. * [32] F. Vazquez, V. M. Eguíluz, and M. San Miguel. Generic absorbing transition in coevolution dynamics. Physical Review Letters, 100(10):108702, 2008. * [33] D. H. Zanette and S. Gil. Opinion spreading and agent segregation on evolving networks. Physica D-Nonlinear Phenomena, 224:156–165, 2006. * [34] D. H. Zanette and S. R. Gusman. Infection spreading in a population with evolving contacts. http://arxiv.org/abs/0711.0874, 2007. * [35] C. Zhou and J. Kurths. Dynamical weights and enhanced synchronization in adaptive complex networks. Physical Review Letters, 96(16):164102, 2006.	arxiv-papers	2008-07-22T11:37:53	2024-09-04T02:48:56.936412	{ "license": "Public Domain", "authors": "Leah B. Shaw and Ira B. Schwartz", "submitter": "Ira Schwartz", "url": "https://arxiv.org/abs/0807.3455" }
0807.3473	# Race for the Kerr field G. Dautcourt111 Max-Planck-Institut für Gravitationsphysik, Albert-Einstein- Institut Am Mühlenberg 1, D-14476 Golm, Germany, Email: daut@aei.mpg.de ###### Abstract Roy P. Kerr has discovered his celebrated metric 45 years ago, yet the problem to find a generalization of the Schwarzschild metric for a rotating mass was faced much earlier. Lense and Thirring, Bach, Andress, Akeley, Lewis, van Stockum and others have tried to solve it or to find an approximative solution at least. In particular Achilles Papapetrou, from 1952 to 1961 in Berlin, was interested in an exact solution. He directed the author in the late autumn of 1959 to work on the problem. Why did these pre-Kerr attempts fail? Comments based on personal reminiscences and old notes. ###### pacs: 01.65.+g, 04.20.Jb, 04.20.-q ## 1 Introduction The old Prussian Academy of Sciences in Berlin was certainly a good place after the Second World War to continue the research of its most prominent former member, Albert Einstein. Here Einstein had worked nineteen years and created his beautiful theory of gravitation. Here several attempts had been undertaken to test the theory: Einstein’s young coworker, the astronomer Erwin Finley Freundlich, had tried to see (with little success) gravitational redshift effects in astronomical objects like the Sun. Freundlich was also active in several campaigns to observe Solar eclipses with the aim to verify the predicted displacement of stars seen near the Sun. And here Karl Schwarzschild, the director of the Astrophysical Observatory at Potsdam, a town at the outskirts of Berlin, had published in 1916 the first exact solution of Einstein’s field equations, describing the exterior gravitational field of a nonrotating spherical mass [43]. In view of the extreme nonlinearity of the equations, it appeared almost as a miracle that exact solutions exist at all. The administrative buildings of the Academy (in the summer of 1946 reopened as ”German Academy of Sciences”) and of the Berlin university happened to be situated in the Eastern part of the divided city of Berlin, thus the Academy worked under the influence of the Soviet Military Administration and later the East German government. Other scientific institutions such as institutes of the former Kaiser-Wilhelm society in Berlin-Dahlem and the 1948 founded Free University, also there, did belong to West Berlin, the sphere of influence of the Western Allies. The borders between the two parts of Berlin were open until 1961, allowing at least some personal contact between scientists of both sides. The director of the Academy since November 1946 was Josef Naas, a mathematician and member of the communist party, who was sent to a concentration camp in the time of the Nazi regime. He and other officials of the Academy were interested in a continuation of research on Einstein’s path at this traditional place. In 1952 Achilles Papapetrou, a Greek scientist at the Physics Department of the University of Manchester, was invited as Senior Researcher to the Academy’s Research Institute for Mathematics, headed by Naas. Papapetrou had started his academic career in solid state theory in the German town Stuttgart (thus he had a fluent knowledge of German), but he was known in the scientific community for his excellent work in relativity as well. The astrophysicists of the Academy planned further tests of Einstein’s theory by astronomical means. Walter Grotrian, who now headed the Academy’s Astrophysical Observatory at Potsdam, organized in collaboration with Finley- Freundlich (at St Andrews, Scotland) and with Papapetrou’s and even Einstein’s222An exchange of letters between Naas and Einstein from November 1951 deals with the scientific chances of such a campaign [18]. advice a campaign to observe the total Solar eclipse on June 30, 1954 from the Svedish island Öland. As for many earlier expeditions, cloudy sky prevented any observation. Apparently, these efforts stimulated Papapetrou’s research interests. Already his latest papers in Manchester were concerned with the derivation of equations of motion for spinning test particles from the conservation law ${\cal T}^{\mu\nu}_{;\nu}=0$ [36], [37]. Removing the restriction to test bodies leads to the question how the gravitational field of a single spinning mass would look. For the nonrotating spherical point mass Schwarzschild had given the answer, but the spinning counterpart was still an open question, at least as far as an exact solution was concerned. Solving this problem was of principle interest for tests of General Relativity in the Solar system with the rotating Sun and planets, no matter how small the effects of rotation would turn out finally. Problems of this type had already attracted several theoreticians. As early as 1918 J. Lense and H. Thirring in Vienna had calculated the exterior gravitational field of a rotating sphere, describing the influence of rotation as linear perturbation to the Schwarzschild metric [34]. R. Bach [5] continued the Lense-Thirring calculation by adding terms which are quadratic in the rotation velocity. In 1924 K. Lanczos published a simple exact solution of the matter field equations for uniformly rotating dust [32]. The matter density in his model has a minimum on the rotation axis and increases exponentially with the coordinate distance from the axis, thus compensating the increasing centrifugal forces by an increased gravitational attraction. Later papers by W.R. Andress [3] and E.S. Akeley [1, 2] were mainly concerned with approximation methods for axisymmetric stationary fields. The first exact solutions of the vacuum field equations within this class of fields were found by T. Lewis and published in an important paper in 1932 [35]. A few years later W.J. Van Stockum [45] derived the gravitational field of an infinite rotating cylinder of dust particles and used one of the Lewis solutions to fit this interior field to an exterior vacuum field. A different class of exact vacuum solution of the Lewis equations was given by Papapetrou soon after he arrived in Berlin [38]. Exact solutions including time-independent gravitational fields were also systematically studied by Pascual Jordan’s research group at Hamburg university, mainly by J. Ehlers, W. Kundt and E. Schücking [26],[15]. However, no exact solution discovered so far could be considered as the gravitational field of a rotating nearly spherical mass. All of them had either singularities on the axis interpreted as violation of the vacuum equations and presence of a line distribution of rotating matter or, as in Papapetrou’s 1953 solutions [38], had angular momentum but zero total mass, as measured by the corresponding terms in an asymptotic expansion of the metric. From a geometrical point of view, these vacuum solutions did belong to a class of metrics, which were later shown by Papapetrou [40] as invariantly characterized by the existence of two commuting Killing fields $\xi^{\rho}$ (timelike) and $\eta^{\rho}$ (spacelike with closed orbits), which admit 2-spaces orthogonal to the group orbits. Usually this class of (vacuum) solutions is called the Lewis-Papapetrou class - also the Kerr metric belongs to this class. The author, who had a background in astrophysics from Schwarzschild’s Potsdam Observatory, entered Papapetrou’s small research group in the mathematics institute in May 1959. I enjoyed the stimulating atmosphere with regular guests from East German universities and international visitors like Marie- Antoinette Tonnelat from Paris and Felix Pirani from London. My first duty was to solve a problem in Einstein’s field theory with the asymmetric metric tensor. Having stood this test [13], not without help by Papapetrou, he considered me as being able to treat a more complicated problem. ”Find the gravitational field of a rotating point mass as a suitable generalization of the Schwarzschild metric” was his suggestion in the late autumn of 1959. Unfortunately for the project, Papapetrou was invited to visit the Institute Henri Poincaré in Paris for one year. He left Berlin early in 1960, thus his valuable advice and encouragement was missed. After all, communication in that time without electronic mail was mainly confined to sending yellow post letters occasionally. In the next sections this old attempt to tackle the problem is described. In the final section I return to the further history of the Kerr field. ## 2 Lewis equations and their generalization The immense literature on stationary axisymmetric gravitational fields known today [44] did not exist in 1959, apart from the three basic papers by Lewis, van Stockum and Papapetrou. To these papers one should have added Jürgen Ehlers’ 1957 thesis on exact solutions [16], but I was only later aware of this work. At some time I had access to the useful Jordan report [26], which summarized the research of the Hamburg group and included a chapter on stationary gravitational fields. Apart from these papers, the whole field was unexplored territory. The most influential paper was that of T. Lewis [35]. He wrote the line element without further explanation essentially as (the notation is taken from Papapetrou [38]) $ds^{2}=e^{\mu}(dx_{1}^{2}+dx_{2}^{2})+ld\phi^{2}+2md\phi dt-fdt^{2},$ (1) where all functions $\mu,l,m$ and $f$ depend only on the two coordinates $x_{1},x_{2}$. As shown by Lewis, with his metric the vacuum field equations $R_{\mu\nu}=0$ have a clear structure and admit a straightforward integration procedure: One obtains a set of three coupled nonlinear partial differential equations in two dimensions involving only the three functions $f,l,m$. Further equations allow to determine the remaining function $\mu$ by simple integration, provided a solution $f,l,m$ is given. Lewis and Papapetrou had found special classes of solutions, but not yet one which was singular only along a single worldline and had a specific asymptotic behaviour, tending to Minkowski spacetime at spatial infinity. I began to treat the problem in a systematic way. As it turned out, this was not the best method. The first question was: Is the form of the metric tensor assumed by Lewis already general enough to describe that axisymmetric vacuum field, which we wanted to find? Actually Lewis served well, but this could not be known beforehand. We had always assumed the metric field to admit two commuting Killing vectors $\xi^{\rho}$ (timelike) and $\eta^{\rho}$ (spacelike, at least near the axis), thus $\xi^{\rho}_{;\sigma}\eta^{\sigma}-\eta^{\rho}_{;\sigma}\xi^{\sigma}=0$. One should also have asked if this assumption of an _A_ belian isometry group $G_{2}$ is perhaps a restriction for the problem. I do not remember our arguments for adopting commutativity (apart from Ockham’s razor). Ten years later Brandon Carter [11] proved that the commutativity assumption means no loss of generality: Axisymmetric stationary fields which become asymptotically flat have commuting Killing fields. More recently Alan Barnes [6] noted that the Abelian character of $G_{2}$ follows in a simple manner from the fact that the orbits of $\eta^{\rho}$ should be topologically circles. In any case, writing down normal forms for the metric in the case of a non-Abelian $G_{2}$ would have convinced us that the running coordinate along the orbits of $\eta^{\rho}$ could not be a cyclic one. Then, asssuming an Abelian $G_{2}$, it was easy to see that one can introduce new coordinates by requiring $\xi^{\rho}=\delta^{\rho}_{0},\eta^{\rho}=\delta^{\rho}_{3}$, such that the metric depends only on the coordinates $x^{1},x^{2}$. This special coordinate form for the Killing fields is left invariant by coordinate transformations of the type $\bar{x}^{1}=\bar{x}^{1}(x^{1},x^{2}),~{}\bar{x}^{2}=\bar{x}^{2}(x^{1},x^{2}),$ (2) $\bar{x}^{3}=x^{3}+p(x^{1},x^{2}),~{}\bar{x}^{0}=x^{0}+q(x^{1},x^{2}),$ (3) apart from linear transformations of $x^{3},x^{0}$; $\bar{x}^{1},\bar{x}^{2}$ are invertible and $p,q$ arbitrary functions of $x^{1},x^{2}$. The question was now: Is it possible to reduce the general axisymmetric stationary vacuum field to the Lewis form by means of these transformations? To my surprise, the answer turned out to be no, not in general. I tried several ways to simplify the metric. Successful was a sort of covariant reduction, where the field equations are written as three-dimensional and in a second step as two-dimensional covariant relations. Methods of this type are described in the Jordan report [26] and also in the Landau-Lifschitz volume ”Field Theory” [33], which I just translated at that time from Russian into German. In the first step the metric tensor $g_{\mu\nu}$ was split into $g_{00}=-V^{2},~{}g_{0i}=-\gamma_{i}V^{2},~{}g_{ik}=\gamma_{ik}-\gamma_{i}\gamma_{k}V^{2}$ (4) ($i,k=1,2,3$). The three field equations $R^{i}_{0}=0$ then led to $(V^{3}\kappa^{ik})_{\|k}=0,$ (5) where $\kappa_{ik}\equiv\frac{1}{2}(\gamma_{i,k}-\gamma_{k,i})$, the stroke denotes the covariant derivative with repect to the 3-metric $\gamma_{ik}$ and indices are muved using $\gamma_{ik}$. If (5) holds, the quantity $E_{ikl}\kappa^{ik}V^{3}$, constructed with the three-dimensional totally antisymmetric Levi-Civita tensor $E_{ikl}$, must be a gradient $\psi_{,l}$. Solving for $\kappa_{ik}$ resulted in $\kappa_{ik}=\epsilon_{ikl}\gamma^{lm}\psi_{,m}/(2V^{3}).$ (6) The condition that $\kappa_{ik}$ is a rotation, requires that $\psi$ satisfies the field equation $\gamma^{kl}\psi_{,k\|l}-3V_{,k}\psi_{,l}\gamma^{kl}/V=0.$ (7) The other field equations $R^{0}_{0}=0$ and $R^{i}_{k}=0$ became $V^{3}\gamma^{kl}V_{,k\|l}+2\psi_{,k}\psi_{,l}\gamma^{kl}=0,$ (8) $R^{(3)i}_{k}-\frac{1}{V}\gamma^{il}V_{,k\|l}+\frac{2}{V^{4}}\delta^{i}_{k}\psi_{,l}\psi_{,m}\gamma^{lm}-\frac{2}{V^{4}}\psi_{,k}\psi_{,l}\gamma^{il}=0.$ (9) In a second step, the 3-metric $\gamma_{ik}$ was split into two-dimensional covariant quantities (capital indices always take values 1,2 in this article): $\gamma_{33}=W^{2},~{}\gamma_{3A}=\epsilon_{A}W^{2},~{}\gamma_{AB}=\epsilon_{AB}+\epsilon{{}_{A}}\epsilon{{}_{B}}W^{2}.$ (10) The 3-tensor equation (9) split into a scalar, vector and tensor equation in two dimensions: $\qquad\epsilon^{AB}W_{,A\parallel B}/W-W^{2}k^{AB}k_{AB}-2\psi_{,A}\psi_{,B}\epsilon^{AB}/V^{4}+W_{,A}V_{,B}\epsilon^{AB}/(VW)=0,$ (11) $\qquad(W^{3}k^{AB})_{\parallel B}+W^{3}k^{AB}V_{,B}/V=0,$ (12) $\qquad R^{(2)A}_{B}-\epsilon^{AC}W_{,B\parallel C}/W-2W^{2}k^{AC}k_{BC}-\epsilon^{AC}V_{,B\parallel C}/V$ $\qquad+2\delta^{A}_{B}\psi_{,C}\psi_{,D}\epsilon^{CD}/V^{4}-2\psi_{,B}\psi{,_{C}}\epsilon^{AC}/V^{4}=0.$ (13) Here $k_{AB}=\frac{1}{2}(\epsilon_{A,B}-\epsilon_{B,A})$ (14) and the double stroke denotes the covariant derivative with respect to the Christoffel affinity formed with the 2-metric $\epsilon_{AB}$ (indices are moved using $\epsilon_{AB})$. Using the coordinate transformations (2), I assumed that the 2-metric can be transformed into a conformally flat metric: $\epsilon_{AB}=\delta_{AB}e^{\mu}.$ (15) The equations (12) now became explicitly $k_{AB,B}+3k_{AB}W_{,B}/W+k_{AB}V_{,B}/V-k_{AB}\mu_{,B}=0.$ (16) The integration gave $k_{12}=\frac{ke^{\mu}}{VW^{3}}$ (17) with $k$ as integration constant. The relations (15) and (17) simplified the field equations considerably. The two equations (7) and (8) became $\qquad\Delta\psi-[\ln{\frac{V^{3}}{W}},\psi]=0,$ (18) $\qquad\Delta V+[V,W]/W+2[\psi,\psi]/V^{3}=0,$ (19) where the differential operator $\Delta$ is the Laplacian in two dimensions, $\Delta=\frac{\partial^{2}}{\partial x_{1}^{2}}+\frac{\partial^{2}}{\partial x_{2}^{2}}$, and the Lewis bracket is defined as $[A,B]\equiv\frac{\partial A}{\partial x^{1}}\frac{\partial B}{\partial x^{1}}+\frac{\partial A}{\partial x^{2}}\frac{\partial B}{\partial x^{2}}.$ (20) Similarly, (11) is a differential equation for $W$: $\qquad\frac{\Delta W}{W}+\frac{[V,W]}{VW}-2\frac{[\psi,\psi]}{V^{4}}=\frac{2k^{2}e^{\mu}}{V^{2}W^{4}}.$ (21) The remaining equations (13) were $\qquad\Delta\mu+\frac{\Delta W}{W}+\frac{\Delta V}{V}-2\frac{[\psi,\psi]}{V^{4}}=-\frac{4k^{2}e^{\mu}}{V^{2}W^{4}},$ (22) $\qquad\mu_{,1}(\frac{V_{,1}}{V}+\frac{W_{,1}}{W})-\mu_{,2}(\frac{V_{,2}}{V}+\frac{W_{,2}}{W})=2\frac{\psi_{,1}^{2}}{V^{4}}-2\frac{\psi_{,2}^{2}}{V^{4}}+\frac{W_{,11}}{W}{}-\frac{W_{,22}}{W}{}+\frac{V_{,11}}{V}{}-\frac{V_{,22}}{V},$ (23) $\qquad\mu_{,1}(\frac{V_{,2}}{V}+\frac{W_{,2}}{W})+\mu_{,2}(\frac{V_{,1}}{V}+\frac{W_{,1}}{W})=2\frac{V_{,12}}{V}+2\frac{W_{,12}}{W}+4\frac{\psi_{,1}\psi_{,2}}{V^{4}}.$ (24) For the function $R=VW$ one derives the simple relation $R^{3}\Delta R=2k^{2}V^{2}e^{\mu}.$ (25) Apparently, Lewis was not general enough. The field equations (18,19,21-25) differ from the Lewis equations through the occurence of an integration constant $k$, complicating Lewis’ integration scheme (note however, $\mu_{,1}$ and $\mu_{,2}$ as calculated from (23,24) still satisfy $\mu_{,1,2}=\mu_{,2,1}$ as well as (22) in virtue of the other equations, even if $k\neq 0$). The complication was not the only problem. I had also carried out the reduction process inversely, splitting the metric first with regard to $\eta^{\mu}$ and then to $\xi^{\mu}$. This introduced a different set of equations with a new constant $\bar{k}$. But neither set could represent the full system of vacuum equations in the Abelian $G_{2}$ case, since both constants $k,\bar{k}$ are expected to occur. The reason why the derivation given above missed $\bar{k}$ is an implicit assumption made in the calculation of (7)-(9), that the potential $\psi$ like all other functions does not depend on $x^{3}$, i.e. ${\cal L}_{\eta}\psi=0$. Yet the condition ${\cal L}_{\eta}\psi=-2\bar{k}\neq 0$ is compatible with the Killing symmetries for the metric and leads to the complete system of vacuum field equations in the presence of two commuting Killing fields. Today one recognizes that the constants $k$ and $\bar{k}$ are essentially the twist scalars associated with the two Killing vectors $\eta^{\mu},\xi^{\mu}$: $2k=E_{\mu\nu\rho\sigma}\xi^{\mu}\eta^{\nu}\eta^{\rho;\sigma},~{}2\bar{k}=E_{\mu\nu\rho\sigma}\eta^{\mu}\xi^{\nu}\xi^{\rho;\sigma}.$ (26) We know that $k$ and $\bar{k}$ are constants if the vacuum equations hold, more generally, they are constants if and only if the conditions $E^{\mu\nu\rho\sigma}\xi^{\lambda}R_{\lambda\mu}\xi_{\nu}\eta_{\rho}=E^{\mu\nu\rho\sigma}\eta^{\lambda}R_{\lambda\mu}\eta_{\nu}\xi_{\rho}=0$ (27) are satisfied [44],[47]. Little is known about the existence of solutions of the generalized Lewis equations. Apparently even today no vacuum solution is known (cf. [44], see however [20]). R. Geroch has shown that for $k,\bar{k}\neq 0$ his method of generating new vacuum solutions from given ones breaks down in the sense that the presence of Killing fields is not preserved [23], [22]. I believe, Papapetrou was not happy with the extension of the Lewis equations. Indeed, his intuition turned out to be correct. He showed in a remarkable paper six years later [41], that the twist scalars must vanish, restoring the Lewis equations for our problem. The geometrical background of this result became clear in papers by W. Kundt and M. Trümper [31] and by B. Carter [10]: While the orbits of the two Killing vectors $\xi^{\mu},\eta^{\mu}$ are always two-surface forming, the two-surface elements orthogonal to the group orbits do not fit to finite surfaces for non-vanishing twist scalars. The Lewis block diagonal form of the metric is just equivalent to ”orthogonal transitivity”, to the existence of two-surfaces orthogonal to the group surfaces. Papapetrou’s 1966 result could have been found already in 1960, had the boundary conditions on the symmetry axis been analyzed: We had fairly precise ideas for the behaviour of the metric at spatial infinity, but did not consider the axis, since here unknown singularities were expected. However, for a rotating localized mass the metric on part of the axis outside the body must be regular. The existence of a (at least partly) regular axis means that the cyclic Killing vector $\eta^{\mu}$ vanishes there (for a recent careful discussion of the axis conditions in axisymmetric spacetimes see, e.g., [42]). Since our coordinates were restricted such that $\eta^{\mu}=\delta^{\mu}_{0}$ everywhere, they must be singular on the axis. A look at (26) with the rhs now written in regular coordinates shows immediately that on the axis (and, since $k,\bar{k}$ are constants, everywhere) $k=\bar{k}=0$. If vacuum solutions with $k,\bar{k}\neq 0$ exist, they would carry singularities on the whole symmetry axis and could not represent the exterior gravitational field of a compact body. But this was not recognized in 1960. ## 3 Lewis-Papapetrou class of vacuum fields After wasting some time with a fruitless study of the extended equations, I returned to Lewis. Letting $k=0$ in (25) gives $\Delta R=0$, thus $R$ is a harmonic function. Provided $R$ is not a constant, this allowed to introduce canonical coordinates $\rho=R$ and eliminated $W=\rho/V$ in all equations. The basic system consisted now of only two coupled nonlinear partial differential equations for the two potentials $\psi(\rho,z)$ and $V(\rho,z)$, depending on the two cylindrical coordinates $x^{1}=\rho,x^{2}=z$: $\displaystyle\Delta\psi+\frac{1}{\rho}\frac{\partial\psi}{\partial\rho}$ $\displaystyle=$ $\displaystyle\frac{4}{V}[V,\psi],$ (28) $\displaystyle\Delta V+\frac{1}{\rho}\frac{\partial V}{\partial\rho}$ $\displaystyle=$ $\displaystyle-\frac{2}{V^{3}}[\psi,\psi]+\frac{1}{V}[V,V].$ (29) One was faced with the problem to find exact solutions of this system with a prescribed behaviour at spatial infinity. Lewis and Papapetrou had derived equivalent systems of equations for a different set of field quantities. No systematic integration theory was known for either system. The interesting mathematical properties of (28,29) as a completely integrable system were unknown at that time. To find solutions at all, Lewis and Papapetrou had to make special ad-hoc assumptions for their potentials. Could similar assumptions be tried for the system (28,29)? Some suggestions came from an article by B. Kent Harrison, just published in the December 1959 issue of Physical Review [24]. He presented many exact solutions of the vacuum field equations, obtained with heuristic methods such as separation of variables. The hope was that some of these techniques, perhaps in combination, would work also here. To have greater flexibility, I first transformed the system (28),(29) into a more general form, by substituting $\psi=\psi(X^{1},X^{2}),~{}V=V(X^{1},X^{2}),$ (30) assuming the $\psi,V$ are at least twice differentiable functions of $X^{A}$, with nonvanishing functional determinant $D\equiv\left\|\begin{array}[]{cc}\frac{\partial\psi}{\partial X^{1}}&\frac{\partial\psi}{\partial X^{2}}\\\ \frac{\partial V}{\partial X^{1}}&\frac{\partial V}{\partial X^{2}}\end{array}\right\|\neq 0.$ (31) Then the inverse functions $X^{A}=X^{A}(\psi,V)$ exist. Introducing (30) into (28),(29), one obtains equations of the type (summation convention for repeated indices) $\Delta X^{A}+\frac{1}{\rho}\frac{\partial X^{A}}{\partial\rho}+\lambda^{A}_{BC}[X^{B},X^{C}]=0,$ (32) where the six quantities $\lambda^{A}_{BC}=\lambda^{A}_{CB}$ are functions of the new independent field quantities $X^{A}$. I observed that equation (32) is invariant with respect to substitutions $\bar{X}^{A}=\bar{X}^{A}(X^{B}),$ (33) provided the $\lambda^{A}_{BC}$ transform as an affine connection (considering the $X^{A}$ as independent variables): $\bar{\lambda}^{A}_{BC}=\frac{\partial\bar{X}^{A}}{\partial X^{I}}(\frac{\partial^{2}X^{I}}{\partial\bar{X}^{B}\partial\bar{X}^{C}}+\lambda^{I}_{DE}\frac{\partial X^{D}}{\partial\bar{X}^{B}}\frac{\partial X^{E}}{\partial\bar{X}^{C}}).$ (34) Was there any hope to reduce (28,29) to linear equations using a suitable nonlinear transformation (33)? A necessary condition is the vanishing of the Ricci tensor ${\cal R}_{AB}=\lambda^{C}_{AB,C}-\lambda^{C}_{CA,B}+\lambda^{C}_{CD}\lambda^{D}_{AB}-\lambda^{C}_{DA}\lambda^{D}_{CB}$, formed with the $\lambda$-connection in the two-dimensional space of potentials $(X^{1},X^{2})$. A short calculation had shown that already ${\cal R}_{11}=-3/V^{2}$ is nonzero, the nonlinearity could not be removed. The simple potential space formalism allowed to anwer also other questions. The integration idea was to try heuristic methods for other potentials, if the original set ($V,\psi$) failed. At least for Papapetrou’s basic assumption in [38] this hope had to be given up. In my notation his condition is $V_{,1}\psi_{,2}-V_{,2}\psi_{,1}=0.$ (35) This relation is invariant with respect to arbitrary coordinate transformations $x^{\prime A}=x^{\prime A}(x^{B})$, but it is also invariant with respect to arbitrary substitutions (33) of the potentials. Thus choosing other potentials does not increase the chance to find solutions beyond the special Papapetrou class. Another heuristic assumption, introduced by Lewis for a pair of his variables, was $X^{1}=p(X^{2})$, a functional relationship between two potentials. It is seen that (35) is satisfied in this case, hence also this restriction leads only to solutions within the special Papapetrou class. An important restriction for the solutions is the proper behaviour at spatial infinity, assumed as ($r=\sqrt{\rho^{2}+z^{2}}$) $f\rightarrow 1,~{}l\rightarrow\rho^{2},~{}m\rightarrow\lambda\frac{\rho^{2}}{r^{3}}$ (36) for the Papapetrou functions $f,l,m$, where $\lambda$ is proportional to the angular momentum (”strong boundary condition”). For the potential $\psi$ this transforms to $\psi\rightarrow\lambda z/(2r^{3})$. The strong boundary condition ensures that the metric tends to the Minkowski spacetime at spatial infinity $r\rightarrow\infty$, it also provides finite values for the total angular momentum [14]. Unfortunately, metrics of the special Papapetrou class which satisfy the strong boundary condition have zero total mass or energy, this follows immediately from an $1/r$ expansion of (28,29) and (35). ## 4 Sample solutions Vacuum metrics with $VW=const$ did not allow canonical coordinates, but satisfy simple equations. One obtains from (18,19,21-25) with $k=0$ the compatible set $\Delta\psi=0,~{}~{}\Delta\mu=0,~{}~{}(V^{2})_{,1}=2\epsilon\psi_{,2},~{}(V^{2})_{,2}=-2\epsilon\psi_{,1}$ (37) ($\epsilon^{2}=1$). $\psi$ and $\mu$ are harmonic functions, their singularities had to be considered as resulting from a singular matter distribution. It seems obvious here to interpret $x^{1},x^{2}$ as quasi-Cartesian coordinates in a plane orthogonal to the $x^{3}$-axis. The Killing vector $\eta^{\mu}$ then represents a translational symmetry along the $x^{3}$-axis. The simplest solutions have a singularity at the origin of $x^{1},x^{2}$ and were believed to describe the exterior gravitational field of an infinite rotating cylinder along the $x^{3}$-axis. But such fields had nothing to do with the gravitational field of a rotating point mass. Later R.B. Hoffman discussed this class of stationary fields [25]. Harrison’s separation technique as applied to (32) was my main working tool. Similar methods are still used today [21]. In principle, this technique can be applied not only to transformed potentials but also in the case of transformed coordinates. Thus I filled many sheets of paper with formulae of that type, too often stopping the calculation once it became clear that the required strong boundary conditions could not be satisfied. For example, in the case of quasi-spherical coordinates $\rho=r\sin{\theta},~{}z=r\cos{\theta}$ one has with $f=V^{2}$ $\psi_{,rr}+\frac{2}{r}\psi_{,r}+\frac{\psi_{,\theta\theta}}{r^{2}}+\cot{\theta}\frac{\psi_{,\theta}}{r^{2}}=2\frac{f_{,r}}{f}\psi_{,r}+\frac{2}{r^{2}}\frac{f_{,\theta}}{f}\psi_{,\theta},$ (38) $f_{,rr}+\frac{2}{r}f_{,r}-\frac{1}{f}f_{,r}^{2}+\frac{1}{r^{2}}f_{,\theta\theta}+\cot{\theta}\frac{f_{,\theta}}{r^{2}}-\frac{1}{r^{2}}\frac{f_{,\theta}^{2}}{f}=-\frac{4}{f}\psi_{,r}^{2}-\frac{4}{r^{2}f}\psi_{,\theta}^{2}.$ (39) Separation assumptions led to a number of subcases. In the case where both $\psi$ and $f$ depend only on the radial coordinate $r$, (38) gives $\psi_{,r}=cf^{2}/r^{2}$, $c=const$. Introducing this into (39) leads to an equation for $f$ alone: $f_{,rr}+\frac{2}{r}f_{,r}-\frac{1}{f}f_{,r}^{2}+4\frac{c^{2}}{r^{4}}f^{3}=0.$ (40) The solutions are $-g_{00}=V^{2}=\frac{\alpha}{\mathfrak{Cos}(\gamma+\beta/r)},~{}~{}\psi=-\frac{\alpha}{2}{\mathfrak{Tan}}(\gamma+\beta/r).$ (41) The three integration constants $\alpha,\beta,\gamma$ ($c=\beta/(2\alpha)$) are not independent, requiring $lim~{}g_{00}\rightarrow-1for~{}r\rightarrow\infty$ gives $\alpha={\mathfrak{Cos}}{\gamma}$. Also the equations for $\mu$ can be integrated, the further non-vanishing components of the metric are $\displaystyle g_{rr}$ $\displaystyle=$ $\displaystyle\frac{1}{\alpha}\mathfrak{Cos}(\gamma+\beta/r)e^{-\beta^{2}\sin^{2}\theta/(4r^{2})},$ (42) $\displaystyle g_{\theta\theta}$ $\displaystyle=$ $\displaystyle r^{2}g_{rr},$ (43) $\displaystyle g_{\phi\phi}$ $\displaystyle=$ $\displaystyle({\mathfrak{Cos}}(\gamma+\beta/r)/\alpha-\frac{\beta^{2}\cot^{2}\theta}{\alpha r^{2}\mathfrak{Cos}(\beta/r+\gamma)})r^{2}\sin^{2}{\theta},$ (44) $\displaystyle g_{\phi t}$ $\displaystyle=$ $\displaystyle\frac{\beta\cos\theta}{2\mathfrak{Cos}(\gamma+\beta/r)}.$ (45) Expansion of $g_{00}$ in reciprocal powers of $r$ shows that $2M=\beta~{}\mathfrak{Tan}{\gamma}$ is the coefficient of the $1/r$-term. Several arguments suggest that $M$ is the total energy of the field. In 1960 I used pseudotensors derived from the Lagrange density of the gravitational field according to Noether’s procedure [14], but one obtains the same result with the ADM mass formula [4]. Again however, this was clearly not the solution we had looked for, since the boundary conditions at spatial infinity are not satisfied for the nondiagonal term, $\psi$ does not vanish but tends to the constant $-\alpha\mathfrak{Tan}{\gamma}/2$ at spatial infinity. Thus the total angular momentum [14] diverges. Since $\psi,V$ depend on $r$ only, a short look shows that (35) is satisfied, thus the solution belongs to the special Papapetrou class. The metric has nonzero total energy only because the strong boundary conditions are violated. It was time consuming and unsatisfactory to search for solutions with the rather simple trial-and-error methods at hand. According to my 1960 notebook, I was impressed by Buchdahl’s procedure to obtain new stationary solutions from a given static or stationary vacuum field. Later many successful recipes were developed to realize this idea of solution generation [27],[44], starting from the pioneering papers by Buchdahl [8], [9] and by Ehlers [17]. This has finally opened the door to the solution space. ## 5 Any chance in 1960? But still I did not give up. On March 6, 1960, I sent Papapetrou a letter with a short summary of results obtained so far (translated from German, signature of metric changed): > “[…] After two weeks of skiing with best snow conditions I’m back to Berlin. > The state of my work is roughly as follows. > > 1) It was guessed that forming a normal form for $g_{\mu\nu}$ already before > the field equations are reduced is preferable, since more transformation > freedom is available. However, Petrov’s choice > ($g_{11}=1,g_{12}=g_{13}=g_{10}=0$) offers no advantage to my previous > approach for a concrete solution of the field equations. It stands for a > certain choice of coordinates in the $x_{1}$-$x_{2}$-space, but the field > equations which must be solved next are not simplified. > > 2) The reverse reduction method (first $x_{3}$ then $x_{0}$) gives very > complicated equations as in the previous case, if $\bar{k}\neq 0$. > ($\bar{k}=0$ now means $g_{0I}=g_{I3}g_{03}/g_{33}$; for the original > reduction sequence the analogous condition $k=0$ was > $g_{3I}=g_{0I}g_{03}/g_{00}$). But if we assume $\bar{k}=0$ and require > Minkowskian boundary conditions, the equations reduce to those of Lewis. > > 3) A class of solutions ($k=0$), which are presumably uninteresting > physically, obviously describe the gravitational field of a rotating > cylinder with a multipole matter source. These fields are independent of > $z$, but they lost rotational symmetry, depending on the polar coordinates > in a plane orthogonal to the cylinder axis. A closer inspection seems not to > be worthwhile. > > 4) My earlier solution $\displaystyle ds^{2}=-\frac{\alpha}{\mathfrak{Cos}(\beta/r+\gamma)}dt^{2}+\frac{\mathfrak{Cos}(\beta/r+\gamma)}{\alpha}e^{-\beta^{2}\sin^{2}\theta/(4r^{2})}(dr^{2}+r^{2}d\theta^{2})$ $\displaystyle+(\frac{\mathfrak{Cos}(\beta/r+\gamma)}{\alpha}-\frac{\beta^{2}\cot^{2}\theta}{\alpha r^{2}\mathfrak{Cos}(\beta/r+\gamma)})r^{2}\sin^{2}{\theta}d\phi^{2}+\frac{\beta\cos\theta}{\mathfrak{Cos}(\beta/r+\gamma)}d\phi dt$ > has some interesting properties which might render it acceptable in spite of > the missing boundary condition for $g_{03}$: > > (a) The total energy (according to Møller) is finite and depends as in the > Schwarzschild case on the $1/r$ term in the expansion of $g_{00}$. > > (b) The total momentum vanishes. > > (c) For $\alpha\rightarrow\infty$ and neglecting the terms with $1/r^{2}$ > the solution tends asymptotically to the Schwarzschild solution. > > (d) For $r\rightarrow\infty$ the space becomes homogeneous, but is no longer > isotropic. Since there exists a distinguished direction, this appears to be > reasonable. > > Mr. Treder and I agree that this type of solution (the given one is only the > simplest) should be considered as physically reasonable. I would like to ask > you for your opinion. > > 5) Presently I am trying to transform the field equations with $k=0$ using > suitable coordinate conditions in the $x_{1}$-$x_{2}$-space to give them a > convenient structure. I hope that within one of these coordinate systems the > solution which we are looking for takes a fairly simple form, and can > therefore be found relatively easily. But one must be lucky! > > In spite of the small success so far I still believe that one can find a > stationary solution which satisfies all requirements. ” It would have been interesting to know Papapetrou’s reaction, but I do not remember having obtained a response. Needless to say, I had no luck with the recipe proposed under item 5). But the recipe itself - looking for suitable coordinates in the $x^{1}$-$x^{2}$-space - was indeed a route to the Holy Grail, the rotating Schwarzschild field: Transforming the cylindrical coordinates $\rho,z$ into some kind of radial and angular coordinates $r,\theta$ by means of $\rho=\sqrt{r^{2}+a^{2}-2rM\sin{\theta}},~{}z=(r-M)\cos{\theta},$ (46) as done by F.J. Ernst [19] after Kerr’s discovery, leads to the Kerr solution with the potentials $V^{2}=1-\frac{2rM}{r^{2}+a^{2}\cos^{2}{\theta}},~{}~{}\psi=-\frac{aM\cos{\theta}}{r^{2}+a^{2}\cos^{2}{\theta}}.$ (47) This looks simple indeed. But how could one have figured out the coordinate transformation (46) in 1960? There exist other coordinates in which the Kerr functions appear fairly simple, e.g. spheroidal prolate coordinates [7]. To find them by trial-and-error would not have been easy, but it was not impossible, given sufficient diligence and persistence. In 1968 F.J. Ernst also found that the complex combination ${\cal{E}}=V^{2}+2i\psi$ satisfies an elegant differential equation [19], easily derivable from (28,29), which has dominated the research on stationary axisymmetric gravitational fields since that time. The Kerr solution of the complex Ernst equation has the simple form ${\cal{E}}=1-2M/(r-ia\cos{\theta})$ in the coordinates $r,\theta$ given by (46). The related function $\xi=(1+{\cal{E}})/(1-\cal{E})$ (essentially a potential transformation as discussed above) satisfies a similar differential equation. Kerr is then represented by $\xi=r/M-ia\cos{\theta}/1-1$. Evidently, this is now even an almost trivial solution of the differential equation for $\xi$. A solution could hardly be simpler. It is satisfying that both concepts, coordinate transformation in $x^{1}$-$x^{2}$-space and potential transformation, were finally so successful, thanks to the efforts by F.J. Ernst. ## 6 The winner is Kerr The slow progress as well as Papapetrou’s absence from Berlin rapidly diminished the amount of time I spent in 1960 for the rotating Schwarzschild problem. During the year I became interested in many other questions of gravitation. Not being guided by my boss, I worked on boundary conditions, surface layers, shock waves and other problems of gravitational radiation. In particular, the characteristic initial value problem for the Einstein field equations was a very interesting topic, since here one could handle the two intrinsic degrees of freedom of the gravitational field rather directly. When Papapetrou returned from Paris in the beginning of 1961, I confronted him with new ideas about these perhaps more actual problems. There was a rumour that Bondi and his group in London were working on similar questions. The problem of stationary fields was forgotten for the present, at least for me. In the meantime the political situation had changed for the worse. There was the danger that Papapetrou’s position as prominent scientist in a communist country could lead to problems for his relatives in Greece. The Cold War was present everywhere, particularly in the divided city of Berlin, whose borders were still open. A steadily increasing number of East Germans escaped to the West. Suddenly, on August 13, 1961, the borders were closed by the communist authorities. Quite unexpectedly, the Berlin Wall was built, cutting off the free Western part of Berlin from the surrounding East German territory. For Papapetrou and his wife the year in Paris was so pleasing compared with the difficult situation in East Berlin, that they decided \- weeks before the borders were closed - to stay permanently in Paris. Not being citizens of the East German state, they were allowed to go. This was unpleasent news. I had to finish my PhD thesis on wave solutions and the characteristic initial value problem sufficiently early in 1961 that Papapetrou could act as adviser, before he finally left East Berlin at the end of the year. Apparently, no time was left to discuss rotating metrics in depth. While I did not seriously return to the spinning body, Papapetrou never forgot the problem. In at least two papers [39], [40] he further explored the properties of axisymmetric metrics and found a new subclass of solutions. However, also these metrics did not satisfy the required boundary conditions at infinity. Also other groups had no luck. In their well-known survey on exact solutions published in the legendary Witten volume in 1962, J. Ehlers and W. Kundt had to admit that “the old problem of constructing rigorously the field of a finite rotating body is as yet unsolved, even as to its exterior part” [15]. The solution of the long-standing problem came 1963 from the New Zealander Roy P. Kerr in a different way. Kerr received his PhD 1959 from Cambridge University (MA), worked later at Syracuse University and with the US Air Force relativity group under Joshua Goldberg at Wright-Patterson Field in Ohio, before he came to the University of Texas at Austin in the academic year 1962/1963. Here, in a newly founded Center of Relativity, organized by Alfred Schild, a circle of relativists had gathered, including besides Schild and Kerr temporarily also Roger Penrose, Ray Sachs, Engelbert Schücking and other excellent scientists. In a recent article Kerr gave a detailed description of his discovery in this stimulating environment [30]. He used a kind of null tetrad formalism, assuming from the beginning an algebraically special spacetime. Both the Schwarzschild metric as well as the Kerr metric are of Petrov type D, thus this restriction was crucial. In rather complicated calculations (more complicated than I ever tried) he further restricted the fields to satisfy stationary and then axisymmetric symmetries, so he finally found the famous solution bearing his name. His two-page paper ”Gravitational Field of a Spinning Mass as an Example of Algebraically Special Metrics” was published in the September 1, 1963 issue of Physical Review Letters. Kerr presented his solution at the First Texas symposium on Relativistic Astrophysics held in Dallas in December 1963 [29]. Papapetrou was possibly not aware of Kerr’s article when he came to the Texas symposium, since the note [40] was presented by Louis de Broglie on December 4 (Séance du 4 décembre), apparently before his departure to the States: therein is no reference to Kerr’s paper. Kip Thorne has given a vivid description of the Texas meeting in his book [46]: > “To foster dialogue between the relativists and the astronomers and > astrophysicists, and to catalyze progress in the study of quasars, a > conference of three hundred scientists was held on 16-18 December 1963, in > Dallas, Texas […] Lectures went on almost continuously from 8:30 in the > morning until 6 in the evening with an hour out for lunch, plus 6 P.M. until > typically 2 A.M. for informal discussions and arguments. Slipped in among > the lectures was a short, ten-minute presentation by a young New Zealander > mathematician, Roy Kerr, who was unknown to the other participiants. Kerr > had just discovered his solution of the Einstein field equation - the > solution which, one decade later, would turn out to describe all properties > of spinning black holes, including their storage and release of rotational > energy […] ; the solution which […] would ultimately become a foundation for > explaining the quasars’ energy. However, in 1963 Kerr’s solution seemed to > most scientists only a mathematical curiosity; nobody even knew it described > a black hole – though Kerr speculated it might somehow give insight into the > implosion of rotating stars. > > The astronomers and astrophysicists had come to Dallas to discuss quasars; > they were not at all interested in Kerr’s esoteric mathematical topic. So, > as Kerr got up to speak, many slipped out of the lecture hall and into the > foyer to argue with each other about their favorite theories of quasars. > Others, less polite, remained seated in the hall and argued in whispers. > Many of the rest catnapped in a fruitless effort to remedy their sleep > deficits from the late-night science. Only a handful of relativists > listened, with rapt attention. > > This was more than Achilles Papapetrou, one of the world’s leading > relativists, could stand. As Kerr finished, Papapetrou demanded the floor, > stood up, and with deep feeling explained the importance of Kerr’s feat. He, > Papapetrou, had been trying for thirty years to find such a solution of > Einstein’s equation, and had failed, as had many other relativists. The > astronomers and astrophysicists nodded politely, and then, as the next > speaker began to hold forth on a theory of quasars, they refocused their > attention, and the meeting picked up pace. ” Kerr’s paper of 1963 is a masterpiece of clarity and conciseness and, as Chandrasekhar noted, is ”surprisingly complete in enumerating the essential features of the solution” [12]. It has only one deficiency, it gives few hints how the solution was derived. When Wolfgang Kundt came to East Berlin in the spring of 1964 for a visit, we discussed Kerr’s metric, but found it hard to verify that it is indeed a solution of the vacuum equations. Many of our colleagues had the same problem. Summarizing, one can shortly answer why the pre-Kerr approaches failed: We had the adequate differential equations, essentially identical to the Ernst equation split into real and imaginary parts. Yet the group-theoretical properties of the solutions were not recognized, and thus no proper key to the unexpectedly large solution space was found. Kerr’s new way circumvented the problem by a restriction to algebraically special metrics from the beginning, a condition which could not easily be expressed in the formalism we had used. In spite of more complicated equations he finally brilliantly succeeded, not least because of his persistence. ## Acknowledgements The author is grateful to R. Kerr, W. Kundt, A. Rendall and G. Wallis for reading the manuscript and for comments and suggestions. ## References ## References * [1] Edward S. Akeley, _The Axially Symmetric Stationary Gravitational Field_ , Phil. Mag. (London) 11 (1931) 322–330. * [2] Edward S. Akeley, _The Rotating Fluid in the Relativity Theory_ , Phil. Mag. (London) 11 (1931) 330–344. * [3] W.R. Andress, _Some Solutions of Einstein’s Gravitational Equations for Systems with Axial Symmetry_ , Proc. Roy. Soc. (London) A 126 (1933) 592–602. * [4] R. Arnowitt, S. Deser and C.W. Misner, _Coordinate Invariance and Energy Expressions in General Relativity_ Phys. Rev. 167 (1961) 997–1006. * [5] R. Bach, _Neue Lösungen der Einsteinschen Gravitationsgleichungen. A. Das Feld in der Umgebung eines langsam rotierenden kugelähnlichen Körpers von beliebiger Masse in 1. und 2. Annäherung_ , Mathematische Zeitschrift 13 (1922) 119–133. * [6] Alan Barnes, _Some Restrictions on the Symmetry Groups of Axially Symmmetric Spacetimes_ , Class. Quant. Grav. 18 (2001) 5511–5520. * [7] Roberto Bergamini and Stefano Viaggiu, _A novel derivation for Kerr metric in Papapetrou gauge_ , Class. Quant. Grav. 21 (2004) 4567–4573. * [8] H.A. Buchdahl, _Reciprocal Static Solutions of the Equations of the Gravitational Field_ , Austral. J. Phys. 9 (1956) 13-18. * [9] H.A. Buchdahl, _Reciprocal Static Metrics and Scalar Fields in the General Theory of Relativity_ , Phys. Rev. 115 (1959) 1325-1328. * [10] Brandon Carter, _Killing Horizons and Orthogonally Transitive Groups in Space-Time_ , J. Math. Phys. 10 (1969) 70–81. * [11] Brandon Carter, _The Commutation Property of a Stationary, Axisymmetric System_ , Commun. math. Phys. 17 (1970) 233–238. * [12] S. Chandrasekhar, _The Mathematical Theory of Black Holes_ , (Oxford University Press, New York 1983) * [13] G. Dautcourt, _Sur la solution de l’équation d’Einstein $g_{\mu_{+}\nu_{-};\rho}=0$_, C.R. Acad. Sc. Paris 249, (1959) 2159-2161. * [14] G. Dautcourt, _Energie, Impuls und Drehimpuls in der allgemeinen Relativitätstheorie_ , Ann. Phys. (Leipzig) 9 (1961) 309-317. * [15] Jürgen Ehlers and Wolfgang Kundt, _Exact Solutions of the Gravitational Field Equations_ , in: Louis Witten (ed.), Gravitation: an introduction to current research, (John Wiley & Sons,Inc.: New York, London 1962), 49-101. * [16] J. Ehlers, _Konstruktionen und Charakterisierung von Lösungen der Einsteinschen Gravitationsfeldgleichungen_ , Dissertation Hamburg 1957 * [17] J. Ehlers, _Transformations of static exterior solutions of Einstein’s gravitational field equations into different solutions by means of conformal mappings_ , Colloques Int. C.N.R.S. (Les théories relativistes de la gravitation) 91 (1961) 275. * [18] 11/05/1951 Naas to Einstein (11-194.00), 11/26/1951 Einstein to Naas (11-195.00), Courtesy of the Albert Einstein Archives Jewish National & University Library Jerusalem * [19] F. J. Ernst, _New Formulation of the Axially Symmetric Gravitational Field Problem_ , Phys. Rev. 167 (1968) 1175–1178. * [20] B. Gaffet _The Einstein equations with two commuting Killing vectors_ , Class.Quant.Grav. 7 (1990) 2017–2044. * [21] J. Gariel, G. Marcilhacy and N.O. Santos, _Stationary axisymmetric solutions involving a third order equation irreducible to Painlevé transcendents_ , J. Math. Phys. 49 (2008) 022501. * [22] Robert Geroch, _A Method for Generating Solutions of Einstein’s Equations_ , J. Math. Phys. 12 (1970) 918–924. * [23] Robert Geroch, _A Method for Generating Solutions of Einstein’s Equation.II_ , J. Math. Phys. 13 (1970) 394–404. * [24] B. Kent Harrison, _Exact Three-Variable Solutions of the Field Equations of General Relativity_ , Phys. Rev. 116 (1959) 1285-1296. * [25] R.B. Hoffman, _Stationary ”non-canonical” solutions of the Einstein vacuum field equations_ , J. Math. Phys. 10 (1969) 953–956. * [26] Pasucal Jordan, _Research on the Theory of General Relativity_ , mimeographed notes, AIR RESEARCH AND DEVELOPMENT COMMAND, European Office, Air Force Contract AF 61(514) -1233 (1958). * [27] P. Kordas, _Aspects of Solution-generatic Techniques for Space-times with Two Commuting Killing Vectors_ , Class.Quant.Grav. 31 (1998) 1941–1984. * [28] Roy P. Kerr, _Gravitational field of a spinning mass as an example of algebraically special metrics_ , Phys. Rev. Lett. 11 (1963) 237–238. * [29] Roy P. Kerr, _Gravitational Collapse and Rotation_ , in _Quasi-Stellar Sources and Gravitational Collapse_ , 99-102, ed. I. Robinson, A. Schild and E.L. Schucking, (University of Chicago Press: 1965). * [30] Roy Patrick Kerr, _Discovering the Kerr und Kerr-Schild metrics_ , arXiv0706.1109v1 [gr-qc] * [31] W. Kundt and M.Trümper, _Orthogonal Decomposition of Axi-symmetric Stationary Spacetimes_ , Z. Phys. 192 (1966) 419–422. * [32] Kornel Lanczos _Über eine stationäre Kosmologie im Sinne der Einsteinschen Gravitationstheorie_ , Z. Phys. 21 (1924) 73–110. * [33] L.D. Landau and E.M. Lifschitz _Klassische Feldtheorie_ , (Akademie-Verlag: Berlin 1963). * [34] J. Lense and H. Thirring, _Über den Einfluß der Eigenrotation auf die Bewegung der Planeten und Monde nach der Einsteinschen Gravitationstheorie_ , Phys. Zeitschrift 19 (1918) 156–163. * [35] T. Lewis, _Some Special Solutions of the Equations of Axially Symmetric Gravitational Fields_ , Proc. Roy. Soc. London A 136 (1932) 176–192. * [36] A. Papapetrou, _Spinning test particles in general relativity. I_ , Proc. R. Soc. A 64 (1952) 248–258. * [37] E. Corinaldesi and A. Papapetrou, _Spinning test particles in general relativity. II_ , Proc. R. Soc. A 64 (1952) 259–268. * [38] A. Papapetrou, _Eine rotationssymmetrische Lösung in der allgemeinen Relativitätstheorie_ , Ann. Phys. (Leipzig) 12 (1953) 309–315. * [39] A. Papapetrou, _Quelques remarques sur les champs gravitationnels stationnaires_ , C.R. Acad. Sc. Paris 257 (1963) 2797–2800. * [40] A. Papapetrou, _Champs gravitationnels stationnaires à symétrie axiale_ , C.R. Acad. Sc. Paris 285 (1964) 90–93. * [41] A. Papapetrou, _Champs gravitationnels stationnaires à symétrie axiale_ , Ann. Inst. Henri Poincare IV (1966) 83–105. * [42] O. Rinne and J. M. Stewart, _A strongly hyperbolic and regular reduction of Einstein’s equations for axisymmetric spacetimes_ , Class.Quant.Grav. 22 (2005) 1143-1166. * [43] K. Schwarzschild, _Über das Gravitationsfeld eines Massenpunktes nach der Einsteinschen Theorie_ , Sitz. Preuss. Akad. Wiss. (1916) 189. * [44] H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers, E. Herlt, _Exact Solutions of Einstein’s Field Equations_ (Second Edition), (Cambridge University Press, 2003) * [45] W.J. van Stockum, _The Gravitational Field of a Distribution of Particles Rotating about an Axis of Symmetry_ , Proc. Roy. Soc. Edinb. 57 (1937)135–154. * [46] Kip S. Thorne, _Black Holes & Time Warps_, Foreword by Stephen Hawking (W.W.Norton & Company: New York, London 2004), 341-342. * [47] Robert M. Wald, _General Relativity_ , The University of Chicago Press : Chicago and London 1984.	arxiv-papers	2008-07-22T13:04:48	2024-09-04T02:48:56.942806	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "G. Dautcourt", "submitter": "G. Dautcourt", "url": "https://arxiv.org/abs/0807.3473" }
0807.3544	# Topological obstructions to embedding of a matrix algebra bundle into a trivial one A.V. Ershov ershov.andrei@gmail.com ###### Abstract. In the present paper we describe topological obstructions to embedding of a (complex) matrix algebra bundle into a trivial one under some additional arithmetic condition on their dimensions. We explain a relation between this problem and some principal bundles with structure groupoid. Finally, we briefly discuss a relation of our results to the twisted K-theory. ## Introduction The starting point of the present work was the following question. Let $X$ be (say) a compact manifold, (1) $A_{k}\stackrel{{\scriptstyle p_{k}}}{{\rightarrow}}X$ a locally trivial bundle with fibre a complex matrix algebra $M_{k}(\mathbb{C})$ (so its “natural” structure group is $\mathop{\rm Aut}\nolimits(M_{k}(\mathbb{C}))\cong\mathop{\rm PGL}\nolimits_{k}(\mathbb{C})$). Then is (1) a subbundle of a (finite dimensional) trivial bundle $X\times M_{n}(\mathbb{C})$, i.e. is there a fiberwise map (in fact embedding) (2) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 8.29584pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&\\\&&\crcr}}}\ignorespaces{\hbox{\kern-8.29584pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{A_{k}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 8.29584pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 41.18484pt\raise 5.1875pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.8264pt\hbox{$\scriptstyle{\mu\quad}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 71.36526pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 5.1231pt\raise-4.94444pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 8.39127pt\raise-24.40971pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.8125pt\hbox{$\scriptstyle{p_{k}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 32.29584pt\raise-31.1542pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern 36.83055pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 71.36526pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{X\times M_{n}(\mathbb{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 91.03139pt\raise-5.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 67.05083pt\raise-24.39581pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.8264pt\hbox{$\scriptstyle{pr_{1}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 47.36526pt\raise-33.57503pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern-3.0pt\raise-38.41666pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 32.29584pt\raise-38.41666pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{X}$}}}}}}}{\hbox{\kern 96.58754pt\raise-38.41666pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ such that $\forall x\in X$ its restriction $\mu\mid_{x}$ embeds the fiber $(A_{k})_{x}\cong M_{k}(\mathbb{C})$ into $M_{n}(\mathbb{C})$ as a unital subalgebra? It is natural to compare this question with the well-known fact that any finite dimensional vector bundle $\xi$ over a compact base $X$ is a subbundle of the product bundle $X\times\mathbb{C}^{n}$ for large enough $n$. A unital homomorphism $M_{k}(\mathbb{C})\rightarrow M_{n}(\mathbb{C})$ exists if and only if $n=kl$ for some $l\in\mathbb{N}.$ As in the case of vector bundles $n$ should be large enough relative to $\dim(X)$. Thus, the initial question can be refined as follows: are there “stable” (i.e. non-vanishing when $l$ tends to infinity) obstructions to the existence of embedding (2)? It turns out that there are no stable obstructions if we do not impose any additional condition on $l$ (for example, for any bundle (1) there is an embedding (2) with $l=k^{m},$ where $m$ is large enough). But if we assume, say, $l$ to be relatively prime to $k$, then stable obstructions arise. Moreover, those obstructions do not depend on the particular choice of $l$ provided that $(l,\,k)=1,$ i.e. they are characteristic classes of the $M_{k}(\mathbb{C})$-bundle (1) itself. In low dimensions this obstructions can be expressed in terms of known characteristic classes. For instance, the first obstruction to the existence of embedding (2) is the obstruction to the reduction of the structure group of bundle (1) from ${\rm PGL}_{k}(\mathbb{C})$ to ${\rm SL}_{k}(\mathbb{C})$. If (1) has the form $\mathop{\rm End}\nolimits(\xi_{k})$ for some vector $\mathbb{C}^{k}$-bundle $\xi_{k}$ then it is just $c_{1}(\xi_{k})\,\mathop{\rm mod}\nolimits\,k\in H^{2}(X,\,\mathbb{Z}/k\mathbb{Z}),$ where $c_{1}$ is the first Chern class. Note that for bundle (1) of the form $\mathop{\rm End}\nolimits(\xi_{k})\rightarrow X$ the class $c_{1}(\xi_{k})\,\mathop{\rm mod}\nolimits\,k$ is well-defined: indeed, $\mathop{\rm End}\nolimits(\xi_{k})\cong\mathop{\rm End}\nolimits(\xi_{k}^{\prime})\,\Leftrightarrow\,\xi_{k}^{\prime}=\xi_{k}\otimes\zeta,$ where $\zeta\rightarrow X$ is a complex line bundle, but $c_{1}(\xi_{k}\otimes\zeta)=c_{1}(\xi_{k})+kc_{1}(\zeta).$ If the just described obstruction is equal to $0$, then the next one belongs to $H^{4}(X,\,\mathbb{Z}/k\mathbb{Z})$ and can be described as follows. We can assume that (1) has the form $\mathop{\rm End}\nolimits(\xi_{k})\rightarrow X$ for some vector $\mathbb{C}^{k}$-bundle $\xi_{k}$ with the structure group ${\rm SL}_{k}(\mathbb{C})$. Then the mentioned obstruction is $c_{2}(\xi_{k})\,\mathop{\rm mod}\nolimits\,k\in H^{4}(X,\,\mathbb{Z}/k\mathbb{Z}).$ We define the sequence of obstructions $\kappa_{n}\in H^{2n}(X,\,\mathbb{Z}/k\mathbb{Z})$ inductively, where $\kappa_{n}$ is defined on the kernel of $\kappa_{n-1}$, and prove that they form in some sense the full set of obstructions. Bundles (1) that admit embeddings (2) with $(k,\,l)=1$ form a proper subclass in the class of Morita-trivial bundles of the form $\mathop{\rm End}\nolimits(\xi_{k})$ which is closed under the tensor product. In particular, the Dixmier-Douady class of any such bundle is equal to $0$, but this condition is not sufficient. Moreover, every such bundle is of the form $\mathop{\rm End}\nolimits(\xi_{k})$, where $\xi_{k}$ is a vector bundle with the structure group $\mathop{\rm SU}\nolimits(k)$, and in addition its second Chern class is equal to $0$ modulo $k$, etc. Moreover, one can give the following characterization of such bundles: (1) admits an embedding (2) with $n=kl,\;(k,\,l)=1$ if and only if the bundle $A_{k}\otimes M_{l^{\infty}}(\mathbb{C})$ is trivial as a bundle with the structure monoid $\mathop{\rm End}\nolimits(M_{kl^{\infty}}(\mathbb{C}))$ [13]. Note that in purely algebraic situation the analog of bundles that admit embedding (2) provided that $n=kl,\;(k,\,l)=1$ is trivial. More precisely, recall [7] that for a field $F$ there is the Brauer group $Br(F)$ consisting of Morita-equivalence classes of central simple algebras with respect to the operation induced by the tensor product of such algebras. It easily follows from the theory of central simple algebras over a field $F$ that any such algebra $A$ of linear dimension $k^{2}$ over $F$ that admits a unital embedding into a matrix algebra $M_{kl}(F),\;(k,\,l)=1$ is isomorphic to the matrix algebra over $F,\;A\cong M_{k}(F).$ Indeed, it directly follows from Artin-Wedderburn’s theorem that $A\cong M_{m}(D),$ where $D$ is a division ring over $F$, and $D$ is unique up to isomorphism. If there exists a unital homomorphism $A\rightarrow M_{n}(F),$ then the centralizer to the image of $A$ is a central simple algebra $B$ such that $A\otimes B\cong M_{n}(F).$ Then it follows from the theory of the Brauer group that $B\cong M_{r}(D^{o}),$ where $D^{o}$ is the so-called opposite ring (i.e. $D$ with the product of elements in reverse order). Clearly, this is only possible when $D$ is one-dimensional over $F$, i.e. $D=F$. This paper is organized as follows. In Subsection 1.1 we reduce the embedding problem (2) to a lifting problem for a suitable fibration ${\rm H}_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm BPU}\nolimits(k)$, see Proposition 1. We consider the general case of $M_{k}(\mathbb{C})$-bundles and the case of $M_{k}(\mathbb{C})$-bundles of the form $\mathop{\rm End}\nolimits(\xi_{k})$, where $\xi_{k}$ is a vector bundle. Moreover, in this section we define a homotopy equivalence between the total space ${\rm H}_{k,\,l}(A_{k}^{univ})$ of the constructed fibration and the so-called matrix Grassmannian (see Propositions 2 and 3). In Subsection 1.2 we describe the first obstruction to the existence of a section of the fibration ${\rm H}_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm BPU}\nolimits(k)$ which turns out to be the obstruction to the reduction of the structure group of $M_{k}(\mathbb{C})$-bundle from $\mathop{\rm PGL}\nolimits_{k}(\mathbb{C})$ to ${\rm SL}_{k}(\mathbb{C})$, see Theorem 5 (in case of bundles of the form $\mathop{\rm End}\nolimits(\xi_{k})$ it is just the characteristic class $c_{1}(\xi_{k})\,\mathop{\rm mod}\nolimits\,k$, see Theorem 6). In Subsection 1.3 we define the next obstruction provided that the previous one is equal to $0$. It follows from the last condition that our bundle has the form $\mathop{\rm End}\nolimits(\xi_{k})$, where $\xi_{k}$ is a vector bundle with the structure group $\mathop{\rm SU}\nolimits(k),$ and in this case it is the characteristic class $c_{2}(\xi_{k})\,\mathop{\rm mod}\nolimits\,k$, see Theorem 8. In Subsection 1.4 we give the general construction of obstructions $\kappa_{r}\in H^{2r}(X,\,\mathbb{Z}/k\mathbb{Z})$ in all dimensions $r$ by means of Postnikov’s tower. In Subsection 1.5 we show that our “higher” obstructions $\kappa_{r},\;r>2$ can be expressed by Chern classes $\widetilde{c}_{r}$ of connective covers $\mathop{\rm BU}\nolimits(k)\langle 2r\rangle$ of $\mathop{\rm BU}\nolimits(k)$ in the same way as in dimensions $1$ and $2$ (when $\kappa_{1}=c_{1}\,\mathop{\rm mod}\nolimits\,k,\;\kappa_{2}=c_{2}\,\mathop{\rm mod}\nolimits\,k$), see Theorem 12. It turns out that the fibration ${\rm H}_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm BPU}\nolimits(k)$ appeared in connection with the embedding problem admits a natural interpretation in terms of some groupoid ${\mathfrak{G}}_{k,\,l}$ related to the set of unital subalgebras $\cong M_{k}(\mathbb{C})$ in the fixed matrix algebra $M_{kl}(\mathbb{C})$. More precisely, ${\rm H}_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm BPU}\nolimits(k)$ is the universal principal ${\mathfrak{G}}_{k,\,l}$-bundle (see Theorem 18 and its corollary). We study the relation between such bundles and groupoids in Section 2. In Section 3 we continue to discuss some relations between groupoids ${\mathfrak{G}}_{k,\,l}$ and a geometry of related bundles. Section 4 is an attempt to apply our approach to the twisted $K$-theory. Important: the latest version of this approach is given in preprint [13]. In what follows we replace the groups ${\rm PGL}_{n}(\mathbb{C})$ by compact ones ${\rm PU}(n)$ considering only $$-homomorphisms instead of all unital homomorphisms of matrix algebras. Since ${\rm PU}(n)$ is a deformation retract of ${\rm PGL}_{n}(\mathbb{C})$ this does not have any effect on the homotopy theory. Acknowledgments: I am grateful to A.S. Mishchenko and E.V. Troitsky for all- round support and very helpful discussions. I would like to express my deep gratitude to Thomas Schick for hospitality and very helpful discussions during my visit to Göttingen. ## 1\. A homotopic description of obstructions ### 1.1. The reduction to the lifting problem The embedding problem (2) can be reduced to the lifting problem for a suitable fibration. The next construction can be regarded as a version of a “bijection” ${\rm Mor}(X\times Y,\,Z)\rightarrow{\rm Mor}(X,\,{\rm Mor}(Y,\,Z))$ adapted to the case of fibration (“$\mathop{\rm Mor}\nolimits$” means “morphisms”). So, let ${\rm Hom}_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))$ be the set of all unital $$-homomorphisms $M_{k}(\mathbb{C})\rightarrow M_{kl}(\mathbb{C})$. It follows from Noether-Skolem’s theorem [7] that there is the representation (3) ${\rm Hom}_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))\cong{\rm PU}(kl)/(E_{k}\otimes{\rm PU}(l))$ (here and below $E_{k}$ denotes the unit $k\times k$-matrix and the tensor product symbol “$\otimes$” denotes the Kronecker product of matrices) in the form of a homogeneous space of the group ${\rm PU}(kl).$ We denote this space by $\mathop{\rm Fr}\nolimits_{k,\,l}$ for short (“$\mathop{\rm Fr}\nolimits$” refers to “frame”). It follows from this representation together with Bott periodicity that the stable (i.e. low dimensional) homotopy groups of this space are as follows: (4) $\pi_{r}(\mathop{\rm Fr}\nolimits_{k,\,l})\cong\mathbb{Z}/k\mathbb{Z}\hbox{\; for \,}r\hbox{\, odd \; and \;}\pi_{r}(\mathop{\rm Fr}\nolimits_{k,\,l})=0\hbox{\; for \,}r\hbox{\, even}.$ Let $A_{k}^{univ}\rightarrow{\rm BPU}(k)$ be the universal $M_{k}(\mathbb{C})$-bundle. Applying the functor ${\rm Hom}_{alg}(\ldots,\,M_{kl}(\mathbb{C}))$ (taking values in the category of topological spaces) to $A_{k}^{univ}$ fiberwisely, we obtain the fibration (5) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 12.10304pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&\\\&\\\\}}}\ignorespaces{\hbox{\kern-12.10304pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 12.10306pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 38.02017pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 38.02017pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 57.83572pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 57.83572pt\raise-20.42831pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.42361pt\hbox{$\scriptstyle{p_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 57.83572pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern-3.0pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 36.10304pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm BPU}(k).}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ It is easy to see that there exists the canonical embedding $\widetilde{\mu}_{k,\,l}$ of $M_{k}(\mathbb{C})$-bundle $p^{}_{k,\,l}(A_{k}^{univ})\rightarrow{\rm H}_{k,\,l}(A_{k}^{univ})$ into the product bundle ${\rm H}_{k,\,l}(A_{k}^{univ})\times M_{kl}(\mathbb{C})$. More precisely, $\widetilde{\mu}_{k,\,l}(h,\,T)=(h,\,h(T))$ for $h\in p_{k,\,l}^{-1}(x),\>T\in(A_{k}^{univ})_{x}$, where $h$ is regarded as a homomorphism $h\colon(A_{k}^{univ})_{x}\rightarrow M_{kl}(\mathbb{C}).$ Let (6) $\bar{f}\colon X\rightarrow\mathop{\rm BPU}\nolimits(k)$ be a classifying map for $M_{k}(\mathbb{C})$-bundle (1), i.e. $A_{k}=\bar{f}^{}(A_{k}^{univ})$. It is easy to see that embedding (2) with $n=kl$ is the same thing as a lift $\widetilde{f}$ of the classifying map $\bar{f}$, $\widetilde{f}\colon X\rightarrow{\rm H}_{k,\,l}(A_{k}^{univ}),\quad p_{k,\,l}\circ\widetilde{f}=\bar{f}$ in fibration (5), and vice versa such a lift defines an embedding. Thus we have the following proposition. ###### Proposition 1. There is a natural one-to-one correspondence between embeddings (2) of $A_{k}=\bar{f}^{}(A_{k}^{univ})$ and lifts $\widetilde{f}$ of its classifying map $\bar{f}$ in fibration (5). The lift of $\bar{f}$ corresponding to an embedding $\mu$ we denote by $\widetilde{f}_{\mu}.$ Clearly, we also have the one-to-one correspondence between homotopy classes of embeddings and (fiberwise) homotopy classes of lifts given by $[\mu]\mapsto[\widetilde{f}_{\mu}].$ There is also a modification of the given construction to the case of bundles $A_{k}$ of the form $\mathop{\rm End}\nolimits(\xi_{k})$. More precisely, let $\xi_{k}^{univ}\rightarrow\mathop{\rm BU}\nolimits(k)$ be the universal $\mathbb{C}^{k}$-bundle. Applying the functor $\mathop{\rm Hom}\nolimits_{alg}(\ldots,\,M_{kl}(\mathbb{C}))$ to the $M_{k}(\mathbb{C})$-bundle $\mathop{\rm End}\nolimits(\xi_{k}^{univ})\rightarrow\mathop{\rm BU}\nolimits(k)$ fiberwisely, we obtain the fibration (cf. (5)): (7) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 12.10304pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&\\\&\crcr}}}\ignorespaces{\hbox{\kern-12.10304pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 12.10306pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 36.10304pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 36.10304pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm H}\nolimits_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 67.20335pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 67.20335pt\raise-20.4561pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.52779pt\hbox{$\scriptstyle{\widehat{p}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 67.20335pt\raise-30.41222pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern-3.0pt\raise-40.91222pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 48.87346pt\raise-40.91222pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BU}\nolimits(k).}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ Note that there is the canonical embedding $\widehat{p}^{}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))\hookrightarrow\mathop{\rm H}\nolimits_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))\times M_{kl}(\mathbb{C})$. Now it is easy to see that an embedding ${\rm End}(\xi_{k})\hookrightarrow X\times M_{kl}(\mathbb{C})$ is the same thing as a lift in (7) of the classifying map $f\colon X\rightarrow\mathop{\rm BU}\nolimits(k)$ for $\xi_{k}.$ It turns out that the total space ${\rm H}_{k,\,l}(A_{k}^{univ})$ of fibration (5) is homotopy equivalent to the so-called matrix Grassmannian ${\rm Gr}_{k,\,l}$ which is the homogeneous space parameterizing the set of unital $$-subalgebras isomorphic to $M_{k}(\mathbb{C})$ (“$k$-subalgebras”) in the fixed algebra $M_{kl}(\mathbb{C})$. Note that according to Noether-Skolem’s theorem it can be represented as the homogeneous space (8) ${\rm Gr}_{k,\,l}\cong{\rm PU}(kl)/({\rm PU}(k)\otimes{\rm PU}(l))$ of the group $\mathop{\rm PU}\nolimits(kl).$ ###### Proposition 2. The space ${\rm H}_{k,\,l}(A_{k}^{univ})$ is homotopy equivalent to the matrix Grassmannian ${\rm Gr}_{k,\,l}$. Proof. Consider the map (9) $\tau_{k,\,l}\colon{\rm H}_{k,\,l}(A_{k}^{univ})\stackrel{{\scriptstyle\simeq}}{{\rightarrow}}{\rm Gr}_{k,\,l}$ defined as follows: it takes a point $h\in{\rm H}_{k,\,l}(A_{k}^{univ})$ such that $p_{k,\,l}(h)=x\in{\rm BPU}(k)$ to the $k$-subalgebra $h((A_{k}^{univ})_{x})\subset M_{kl}(\mathbb{C})$ (here we identify points in ${\rm Gr}_{k,\,l}$ with corresponding $k$-subalgebras in $M_{kl}(\mathbb{C})$). It is easy to verify that $\tau_{k,\,l}$ is a fibration with contractible fibers ${\rm H}_{k,\,1}(A_{k}^{univ})\cong\mathop{\rm EPU}\nolimits(k)$ (where $\mathop{\rm EPU}\nolimits(k)$ is the total space of the universal principal $\mathop{\rm PU}\nolimits(k)$-bundle), whence $\tau_{k,\,l}$ is a homotopy equivalence.$\quad\square$ The tautological $M_{k}(\mathbb{C})$-bundle ${\mathcal{A}}_{k,\,l}\rightarrow{\rm Gr}_{k,\,l}$ can be defined as a subbundle of the product bundle ${\rm Gr}_{k,\,l}\times M_{kl}(\mathbb{C})$ consisting of all pairs $\\{(x,\,T)\mid x\in{\rm Gr}_{k,\,l},\>T\in M_{k,\,x}\subset M_{kl}(\mathbb{C})\\},$ where $M_{k,\,x}$ denotes the $k$-subalgebra corresponding to $x\in{\rm Gr}_{k,\,l}$. Let $\lambda_{k,\,l}\colon\mathop{\rm Gr}\nolimits_{k,\,l}\rightarrow\mathop{\rm BPU}\nolimits(k)$ be the classifying map for the principal $\mathop{\rm PU}\nolimits(k)$-bundle $\mathop{\rm PU}\nolimits(k)\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ (note that the tautological $M_{k}(\mathbb{C})$-bundle ${\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ is associated with it). Then the diagram --- $\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{p_{k,\,l}}$$\scriptstyle{\quad\tau_{k,\,l}}$$\textstyle{\mathop{\rm Gr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{k,\,l}}$$\textstyle{\mathop{\rm BPU}\nolimits(k)}$ is commutative. In particular, $p^{}_{k,\,l}(A_{k}^{univ})=\tau^{}_{k,\,l}({\mathcal{A}}_{k,\,l})$ and the diagram $\textstyle{p^{}_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\widetilde{\mu}_{k,\,l}\qquad}$$\scriptstyle{\widetilde{\tau}_{k,\,l}}$$\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\times M_{kl}(\mathbb{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\tau_{k,\,l}\times\mathop{\rm id}\nolimits}$$\textstyle{{\mathcal{A}}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mu_{k,\,l}\qquad}$$\textstyle{{\rm Gr}_{k,\,l}\times M_{kl}(\mathbb{C}),}$ is commutative, where $\mu_{k,\,l}\colon{\mathcal{A}}_{k,\,l}\hookrightarrow{\rm Gr}_{k,\,l}\times M_{kl}(\mathbb{C})$ is the tautological embedding and $\widetilde{\tau}_{k,\,l}$ covers $\tau_{k,\,l}$. Suppose that $(k,\,l)=1.$ Note that in this case the matrix Grassmannian $\mathop{\rm Gr}\nolimits_{k,\,l}$ can be represented in the form (10) ${\rm Gr}_{k,\,l}\cong{\rm SU}(kl)/({\rm SU}(k)\otimes{\rm SU}(l)).$ (cf. (8)). Indeed, if $k$ and $l$ are relatively prime then the center of the group ${\rm SU}(kl)$ (which is the group $\rho_{kl}$ of $kl$th roots of unity) is the product $\rho_{k}\times\rho_{l}$ of the centers of ${\rm SU}(k)$ and ${\rm SU}(l)$. In particular, the structure group of the tautological bundle ${\mathcal{A}}_{k,\,l}\rightarrow{\rm Gr}_{k,\,l}$ is $\mathop{\rm SU}\nolimits(k)$ (because it is associated with the principal bundle $\mathop{\rm SU}\nolimits(k)\rightarrow\mathop{\rm SU}\nolimits(kl)/(E_{k}\otimes\mathop{\rm SU}\nolimits(l))\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$). Note that the space $\mathop{\rm Gr}\nolimits_{k,\,l}$ provided that $(k,\,l)=1$ has the following homotopy groups in stable dimensions: $\pi_{r}(\mathop{\rm Gr}\nolimits_{k,\,l})\cong\mathbb{Z}\;\hbox{for $r\geq 4$ even and}\;0\;\hbox{otherwise.}$ The simplest way to compute them is to use the homotopy sequence of the fibration $\mathop{\rm SU}\nolimits(k)\times\mathop{\rm SU}\nolimits(l)\rightarrow\mathop{\rm SU}\nolimits(kl)\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ for $(k,\,l)=1$, where the first map is the homomorphism given by the Kronecker product of matrices. ###### Proposition 3. The space ${\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))$ is homotopy equivalent to ${\rm H}_{k,\,l}(A_{k}^{univ})\times\mathbb{C}P^{\infty}.$ Proof. Let $\chi_{k}\colon\mathop{\rm U}\nolimits(k)\rightarrow\mathop{\rm PU}\nolimits(k)$ be the group epimorphism (factorization by the center $\mathop{\rm U}\nolimits(1)\subset\mathop{\rm U}\nolimits(k)$). It is easy to see that the classifying map $\mathop{\rm BU}\nolimits(k)\rightarrow\mathop{\rm BPU}\nolimits(k)$ for the bundle $\mathop{\rm End}\nolimits(\xi_{k}^{univ})\rightarrow\mathop{\rm BU}\nolimits(k)$ as a bundle with the structure group $\mathop{\rm PU}\nolimits(k)$ is $\mathop{\rm B}\nolimits\chi_{k}.$ Consider the Cartesian square (11) $\begin{array}[]{ccc}\mathop{\rm H}\nolimits_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))&\stackrel{{\scriptstyle\widetilde{\mathop{\rm B}\nolimits\chi}_{k}}}{{\longrightarrow}}&\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\\\ {\scriptstyle\widehat{p}_{k,\,l}}\downarrow&&\downarrow{\scriptstyle p_{k,\,l}}\\\ \mathop{\rm BU}\nolimits(k)&\stackrel{{\scriptstyle\mathop{\rm B}\nolimits\chi_{k}}}{{\longrightarrow}}&\mathop{\rm BPU}\nolimits(k),\\\ \end{array}$ where $\mathop{\rm B}\nolimits\chi_{k}$ is a $\mathbb{C}P^{\infty}$-fibration classified by the map $\mathop{\rm BPU}\nolimits(k)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Z},\,3)$. Thus, $\widetilde{\mathop{\rm B}\nolimits\chi}_{k}$ is also a $\mathbb{C}P^{\infty}$-fibration which is the pullback of $\mathop{\rm B}\nolimits\chi_{k}$ by $p_{k,\,l}$. It follows from the above given homotopy groups of $\mathop{\rm Gr}\nolimits_{k,\,l}$ provided that $(k,\,l)=1$ that the first nontrivial homotopy group is $\pi_{4}(\mathop{\rm Gr}\nolimits_{k,\,l})\cong\mathbb{Z},$ in particular $H^{3}(\mathop{\rm Gr}\nolimits_{k,\,l},\,\mathbb{Z})=0,$ therefore $\mathbb{C}P^{\infty}$-fibration $\widetilde{\mathop{\rm B}\nolimits\chi}_{k}$ is trivial$.\quad\square$ ###### Remark 4. The matrix Grassmannians $\mathop{\rm Gr}\nolimits_{k,\,l}$ classify equivalence classes of pairs $(A_{k},\,\mu)$ over finite $CW$-complexes $X$, where $A_{k}\rightarrow X$ is a locally trivial $M_{k}(\mathbb{C})$-bundle over $X$ and $\mu$ is an embedding $A_{k}\rightarrow X\times M_{kl}(\mathbb{C})$ (see (2)). Two such pairs $(A_{k},\,\mu),\;(A_{k}^{\prime},\,\mu^{\prime})$ are equivalent if $A_{k}\cong A_{k}^{\prime}$ and $\mu$ is homotopic to $\mu^{\prime}$. ### 1.2. The first obstruction Now we give the promised description of the obstructions to the lifting in fibration (5). First, consider the first obstruction. According to the topological obstruction theory (and taking into account the homotopy groups (4)) it is a characteristic class $A_{k}\mapsto\bar{\kappa}_{1}(A_{k})=\bar{f}^{}(\bar{\kappa}_{1})\in H^{2}(X,\,\mathbb{Z}/k\mathbb{Z}),$ where $\bar{\kappa}_{1}:=\bar{\kappa}_{1}(A_{k}^{univ})\in H^{2}(\mathop{\rm BPU}\nolimits(k),\,\mathbb{Z}/k\mathbb{Z})$. ###### Theorem 5. The first obstruction $\bar{\kappa}_{1}(A_{k})$ is the obstruction to the reduction (or lift) of the structure group ${\rm PU}(k)$ of bundle (1) to ${\rm SU}(k)$ (here we mean the exact sequence of groups $1\rightarrow\rho_{k}\rightarrow{\rm SU}(k)\stackrel{{\scriptstyle\vartheta_{k}}}{{\rightarrow}}{\rm PU}(k)\rightarrow 1,$ where $\rho_{k}$ is the group of $k$th roots of unity). Proof. Note that it follows from (10) provided that $(k,\,l)=1$ that the structure group of $M_{k}(\mathbb{C})$-bundles ${\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ and $p^{}_{k,\,l}(A_{k}^{univ})\rightarrow{\rm H}_{k,\,l}(A_{k}^{univ})$ is $\mathop{\rm SU}\nolimits(k).$ From the other hand, if the structure group of $A_{k}$ can be reduced to $\mathop{\rm SU}\nolimits(k)$, then $\bar{\kappa}_{1}(A_{k})=0$ because the space $\mathop{\rm BSU}\nolimits(k)$ is $3$-connected.$\quad\square$ Clearly that $\bar{\kappa}_{1}$ is a generator of $H^{2}(\mathop{\rm BPU}\nolimits(k),\,\mathbb{Z}/k\mathbb{Z})\cong\mathbb{Z}/k\mathbb{Z}$. Now assume that $A_{k}$ has the form ${\rm End}(\xi_{k})$, where $\xi_{k}\rightarrow X$ is a vector $\mathbb{C}^{k}$-bundle. Note that not every $M_{k}(\mathbb{C})$-bundle can be represented in such a form: the obstruction is the class $\delta(\kappa_{1}(A_{k}))\in Br(X):=H^{3}_{tors}(X,\,\mathbb{Z}),$ where $\delta\colon H^{2}(X,\,\mathbb{Z}/k\mathbb{Z})\rightarrow H^{3}(X,\,\mathbb{Z})$ is the coboundary homomorphism corresponding to the coefficient sequence $0\rightarrow\mathbb{Z}\stackrel{{\scriptstyle\cdot k}}{{\rightarrow}}\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z}\rightarrow 0.$ ###### Theorem 6. For $A_{k}={\rm End}(\xi_{k})$ the first obstruction is $\kappa_{1}(A_{k}):=c_{1}(\xi_{k})\mathop{\rm mod}\nolimits\,k\in H^{2}(X,\,\mathbb{Z}/k\mathbb{Z})$, where $c_{1}$ is the first Chern class. Proof. Note that $\kappa_{1}=(\mathop{\rm B}\nolimits\chi_{k})^{}(\bar{\kappa}_{1}).$ It follows from (11) that the first obstruction for (7) coincides with the first obstruction for the fibration $\mathop{\rm PU}\nolimits(k)\rightarrow\mathbb{C}P^{\infty}\rightarrow\mathop{\rm BU}\nolimits(k)$ obtained by the extension of the exact sequence of groups $\mathop{\rm U}\nolimits(1)\rightarrow\mathop{\rm U}\nolimits(k)\stackrel{{\scriptstyle\chi_{k}}}{{\rightarrow}}\mathop{\rm PU}\nolimits(k)$ to the right. Consider the morphism of bundles --- $\textstyle{\mathop{\rm U}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm U}\nolimits(1)}$$\textstyle{\mathop{\rm EU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm U}\nolimits(1)}$$\textstyle{\mathop{\rm PU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathbb{C}P^{\infty}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm BU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{\mathop{\rm BU}\nolimits(k).}$ According to the definition by means of the obstruction theory, the first Chern class $c_{1}\in H^{2}(\mathop{\rm BU}\nolimits(k),\,\mathbb{Z})$ is the first obstruction to the existence of a section of the universal principal $\mathop{\rm U}\nolimits(k)$-bundle on the above diagram. Note that the left slanting arrow induces the map $\pi_{1}(\mathop{\rm U}\nolimits(k))\rightarrow\pi_{1}(\mathop{\rm PU}\nolimits(k)),\;\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z},\;\alpha\mapsto\alpha\,\mathop{\rm mod}\nolimits\,k$ of the fundamental groups. Now the required assertion follows from the functoriality of obstruction classes.$\quad\square$ Note that since $(\mathop{\rm B}\nolimits\chi_{k})^{}$ induces an isomorphism $H^{2}(\mathop{\rm BPU}\nolimits(k),\,\mathbb{Z}/k\mathbb{Z})\rightarrow H^{2}(\mathop{\rm BU}\nolimits(k),\,\mathbb{Z}/k\mathbb{Z}),$ we see that the equality $(\mathop{\rm B}\nolimits\chi_{k})^{}(\bar{\kappa}_{1})=c_{1}\,\mathop{\rm mod}\nolimits\,k$ uniquely determines the class $\bar{\kappa}_{1}$. The fact that the structure group of a bundle $\mathop{\rm End}\nolimits(\xi_{k})$ can be reduced to $\mathop{\rm SU}\nolimits(k)$ (see Proposition 5) if and only if $c_{1}(\xi_{k})\equiv 0\mathop{\rm mod}\nolimits k$ can be explained as follows. If $c_{1}(\xi_{k})\equiv 0\mathop{\rm mod}\nolimits k$, then $c_{1}(\xi_{k})=k\alpha,\;\alpha\in H^{2}(X,\,\mathbb{Z}).$ There exists a line bundle $\zeta^{\prime}\rightarrow X$ (unique up to isomorphism) such that $c_{1}(\zeta^{\prime})=-\alpha.$ Then $c_{1}(\xi_{k}\otimes\zeta^{\prime})=c_{1}(\xi_{k})+kc_{1}(\zeta^{\prime})=0,$ i.e. $\xi_{k}\otimes\zeta^{\prime}$ is an $\mathop{\rm SU}\nolimits(k)$-bundle. From the other hand, $\mathop{\rm End}\nolimits(\xi_{k})=\mathop{\rm End}\nolimits(\xi_{k}\otimes\zeta^{\prime}).$ Conversely, the classifying map $\mathop{\rm B}\nolimits\chi_{k}\colon\mathop{\rm BU}\nolimits(k)\rightarrow\mathop{\rm BPU}\nolimits(k)$ for $\mathop{\rm End}\nolimits(\xi_{k}^{univ})$ as a $\mathop{\rm PU}\nolimits(k)$-bundle has the fiber $\mathbb{C}P^{\infty}.$ Then the topological obstruction theory implies that $\mathop{\rm End}\nolimits(\xi_{k})\cong\mathop{\rm End}\nolimits(\xi_{k}^{\prime})$ as $M_{k}(\mathbb{C})$-bundles if and only if $\xi_{k}^{\prime}=\xi_{k}\otimes\zeta^{\prime}$ for some line bundle $\zeta^{\prime}\rightarrow X$. Clearly, $c_{1}(\xi_{k}^{\prime})\equiv c_{1}(\xi_{k})\mathop{\rm mod}\nolimits k$ and $\xi_{k}^{\prime}$ is an $\mathop{\rm SU}\nolimits(k)$-bundle $\Leftrightarrow$ $c_{1}(\xi_{k}^{\prime})=0\;\Rightarrow\;c_{1}(\xi_{k})\equiv 0\mathop{\rm mod}\nolimits\,k.$ ###### Remark 7. Let us describe the relation between two versions (“$\mathop{\rm PU}\nolimits$” and “$\mathop{\rm U}\nolimits$”) of the first obstructions and the Brauer group $Br(X)=H^{3}_{tors}(X,\,\mathbb{Z})$. Consider the exact coefficient sequence $0\rightarrow\mathbb{Z}\stackrel{{\scriptstyle\cdot k}}{{\rightarrow}}\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z}\rightarrow 0$ and the piece of the corresponding cohomology sequence: $H^{2}(X,\,\mathbb{Z})\stackrel{{\scriptstyle\lambda}}{{\rightarrow}}H^{2}(X,\,\mathbb{Z}/k\mathbb{Z})\stackrel{{\scriptstyle\delta}}{{\rightarrow}}H^{3}(X,\,\mathbb{Z}).$ Then $\delta(\kappa_{1}(A_{k}))=0\Leftrightarrow A_{k}$ has the form $\mathop{\rm End}\nolimits(\xi_{k})$ for some vector $\mathop{\rm U}\nolimits(k)$-bundle $\xi_{k}$ (note that $\delta(\kappa_{1}(A_{k}))\in H^{3}(X,\,\mathbb{Z})$ is exactly the class of $A_{k}$ in the Brauer group). If $\delta(\kappa_{1}(A_{k}))=0$, then $\kappa_{1}(A_{k})=\lambda(c_{1}(\xi_{k}))$, where $\lambda$ is the reduction modulo $k$. But the bundle $\xi_{k}$ such that $\mathop{\rm End}\nolimits(\xi_{k})=A_{k}$ is not unique: $\xi_{k}^{\prime}=\xi_{k}\otimes\zeta^{\prime}$ also suits. Clearly, $c_{1}(\xi_{k}^{\prime})\equiv c_{1}(\xi_{k})\mathop{\rm mod}\nolimits k$ and $c_{1}(\xi_{k})\equiv 0\mathop{\rm mod}\nolimits k\Leftrightarrow$ $\xi_{k}^{\prime}=\xi_{k}\otimes\zeta^{\prime}$ is an $\mathop{\rm SU}\nolimits(k)$-bundle for some line bundle $\zeta^{\prime}.$ ### 1.3. The second obstruction Now assume that for a bundle $A_{k}\stackrel{{\scriptstyle p_{k}}}{{\rightarrow}}X$ (1) the first obstruction is equal to $0$. We have shown that such a bundle has the form $\mathop{\rm End}\nolimits(\widetilde{\xi}_{k})$ for some vector $\mathbb{C}^{k}$-bundle ${\widetilde{\xi}_{k}}$ with the structure group $\mathop{\rm SU}\nolimits(k)$ (note that such a bundle ${\widetilde{\xi}_{k}}$ is unique up to isomorphism). Equivalently, the classifying map $\bar{f}\colon X\rightarrow\mathop{\rm BPU}\nolimits(k)$ (6) can be lifted to $f\colon X\rightarrow\mathop{\rm BSU}\nolimits(k).$ It follows from standard facts of topological obstruction theory and given above (see (4)) stable homotopy groups of the space $\mathop{\rm Fr}\nolimits_{k,\,l}={\rm Hom}_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))$ that the next obstruction belongs to $H^{4}(X,\,\mathbb{Z}/k\mathbb{Z})$. ###### Theorem 8. The second obstruction is $\kappa_{2}(A_{k}):=c_{2}(\widetilde{\xi}_{k})\mathop{\rm mod}\nolimits\,k$, where $c_{2}$ is the second Chern class. Proof. First note that the space $\mathop{\rm Fr}\nolimits_{k,\,l}$ has the universal covering (12) $\rho_{k}\rightarrow\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}.$ Hence $\pi_{r}(\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l})=\pi_{r}(\mathop{\rm Fr}\nolimits_{k,\,l})$ for $r\geq 2$ and $\pi_{1}(\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l})=0$ (while $\pi_{1}(\mathop{\rm Fr}\nolimits_{k,\,l})=\mathbb{Z}/k\mathbb{Z}$). Obviously, $\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\cong\mathop{\rm SU}\nolimits(kl)/(E_{k}\otimes\mathop{\rm SU}\nolimits(l))$ (cf. (3)). Now consider the following diagram: (13) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 9.6586pt\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 59.06213pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 83.26823pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 132.67175pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 132.67175pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm EPU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm PU}\nolimits(k)}}\mathop{\rm Fr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 173.59778pt\raise-6.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 173.59778pt\raise-20.5pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.42361pt\hbox{$\scriptstyle{p_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 173.59778pt\raise-30.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern-9.6586pt\raise-41.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 9.6586pt\raise-35.435pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\kern 60.69524pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 9.6586pt\raise-41.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 33.6586pt\raise-41.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 33.6586pt\raise-41.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 71.16518pt\raise-47.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 71.16518pt\raise-61.25pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.52779pt\hbox{$\scriptstyle{\widetilde{p}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 71.16518pt\raise-71.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 97.39467pt\raise-30.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 114.36812pt\raise-15.87688pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.62312pt\hbox{$\scriptstyle{\simeq}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 157.36047pt\raise-6.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern 153.25398pt\raise-41.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BPU}\nolimits(k)}$}}}}}}}{\hbox{\kern-3.0pt\raise-81.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 50.0575pt\raise-81.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BSU}\nolimits(k),\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 92.27287pt\raise-73.15175pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\kern 159.68665pt\raise-46.5pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces}}}}}\end{array}$ where $p_{k,\,l}$ is fibration (5). (Indeed, it is associated with the universal principal $\mathop{\rm PU}\nolimits(k)$-bundle with respect to the natural action of the group $\mathop{\rm PU}\nolimits(k)=\mathop{\rm Aut}\nolimits(M_{k}(\mathbb{C}))$ on the space $\mathop{\rm Fr}\nolimits_{k,\,l}=\mathop{\rm Hom}\nolimits_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))$). Note that the homotopy equivalence $\widetilde{\tau}_{k,\,l}\colon\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\simeq\mathop{\rm Gr}\nolimits_{k,\,l}$ (cf. (9)) can easily be obtained from representation (10). $\pi_{3}(\widetilde{\rm Fr}_{k,\,l})=\mathbb{Z}/k\mathbb{Z}\;\Rightarrow$ the “universal” obstruction is a characteristic class $\kappa_{2}\in H^{4}({\rm BSU}(k),\,\mathbb{Z}/k\mathbb{Z}).$ Let ${\widetilde{\xi}}_{k}^{univ}\rightarrow{\rm BSU}(k)$ be the universal ${\rm SU}(k)$-bundle. Since $c_{2}({\widetilde{\xi}}_{k}^{univ})\mathop{\rm mod}\nolimits\,k$ is a generator of $H^{4}(\mathop{\rm BSU}\nolimits(k),\,\mathbb{Z}/k\mathbb{Z})\cong\mathbb{Z}/k\mathbb{Z},$ we see that (14) $\kappa_{2}=\alpha c_{2}({\widetilde{\xi}}_{k}^{univ})\mathop{\rm mod}\nolimits\,k\in H^{4}({\rm BSU}(k),\,\mathbb{Z}/k\mathbb{Z}),\;\alpha\in\mathbb{Z}.$ We have the commutative diagram --- $\textstyle{\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\qquad\widetilde{\tau}_{k,\,l}}$$\scriptstyle{\widetilde{p}_{k,\,l}}$$\textstyle{\mathop{\rm Gr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{k,\,l}}$$\textstyle{\mathop{\rm BSU}\nolimits(k),}$ where $\lambda_{k,\,l}$ is the classifying map for ${\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ as an $\mathop{\rm SU}\nolimits(k)$-bundle. Thus, the piece of the homotopy sequence for the “$\mathop{\rm SU}\nolimits$”-part of (13) $\pi_{4}(\widetilde{\rm Fr}_{k,\,l})\rightarrow\pi_{4}(\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l})\rightarrow\pi_{4}({\rm BSU}(k))\rightarrow\pi_{3}(\widetilde{\rm Fr}_{k,\,l})\rightarrow\pi_{3}(\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l})$ is $0\rightarrow\mathbb{Z}\rightarrow\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z}\rightarrow 0$ $\Rightarrow$ the image $\pi_{4}(\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l})\hookrightarrow\pi_{4}({\rm BSU}(k))$ is the subgroup of index $k$ in $\pi_{4}({\rm BSU}(k))\cong\mathbb{Z}$. Take $X=S^{4}$ and consider the group homomorphism $\pi_{4}(\mathop{\rm BSU}\nolimits(k))\rightarrow H^{4}(S^{4},\,\mathbb{Z}/k\mathbb{Z}),\;[g]\mapsto g^{}\kappa_{2},$ where $g\colon S^{4}\rightarrow\mathop{\rm BSU}\nolimits(k)$ and $[g]\in\pi_{4}(\mathop{\rm BSU}\nolimits(k))$ is the corresponding homotopy class. If $k\nmid[g]$ in $\pi_{4}(\mathop{\rm BSU}\nolimits(k))\cong\mathbb{Z}$, then $g^{}\kappa_{2}\neq 0$ because $g^{}\kappa_{2}$ is the unique obstruction to the embedding in this case. Hence, $\alpha$ in (14) is invertible modulo $k$. In order to prove that $\alpha=1$ consider the morphism of the fibrations $\textstyle{\mathop{\rm SU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm ESU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm ESU}\nolimits(k){\mathop{\times}\limits_{\mathop{\rm SU}\nolimits(k)}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{\mathop{\rm BSU}\nolimits(k).}$ $c_{2}$ is the first obstruction to the existence of a section of the universal principal $\mathop{\rm SU}\nolimits(k)$-bundle. Note that the map $\mathop{\rm SU}\nolimits(k)\rightarrow\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}$ induces the homomorphism $\pi_{3}(\mathop{\rm SU}\nolimits(k))\rightarrow\pi_{3}(\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}),\;\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z},\;\beta\mapsto\beta\,\mathop{\rm mod}\nolimits\,k$ of the third homotopy groups. Now it follows from the functoriality of the first obstruction class that $\kappa_{2}=c_{2}\,\mathop{\rm mod}\nolimits\,k,$ as required.$\quad\square$ Note that the obstructions are stable in the sense that they do not vanish under the taking of the direct limit over pairs $\\{k,\,l\\}$ provided that $(k,\,l)=1$. Moreover, they do “behave well” with respect to the tensor product of bundles (cf. (25)). ### 1.4. Higher obstructions In this section we shall define obstructions to the existence of embedding (2) of a bundle of the form $A_{k}=\mathop{\rm End}\nolimits(\xi_{k})\rightarrow X$ by induction with the help of Postnikov’s tower. Of course, in dimensions $2$ and $4$ we shall obtain the above described obstructions. According to the obstruction theory (applied to (4)) the first obstruction $\kappa_{1}$ to the existence of a section in fibration (7) is in $H^{2}(\mathop{\rm BU}\nolimits(k),\,\pi_{1}(\mathop{\rm Fr}\nolimits_{k,\,l}))=H^{2}(\mathop{\rm BU}\nolimits(k),\,\mathbb{Z}/k\mathbb{Z}).$ The cohomology class $\kappa_{1}$ defines the map $\mathop{\rm BU}\nolimits(k)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,2)$ (which is unique up to homotopy), which we shall denote by the same symbol $\kappa_{1}.$ Let $\mathop{\rm F}\nolimits(\kappa_{1})$ be the fiber of the map $\kappa_{1}.$ We have the diagram (15) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 14.4336pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&\\\&&\\\&&\\\\}}}\ignorespaces{\hbox{\kern-3.0pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 57.43086pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 69.5339pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 69.5339pt\raise-30.10443pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 143.66129pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits^{[1]}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 143.66129pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 81.63696pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 143.66129pt\raise-5.49065pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\kern 93.54216pt\raise-30.10443pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern-3.0pt\raise-41.89554pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 38.4336pt\raise-41.89554pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 59.28197pt\raise-47.92886pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 23.84148pt\raise-56.01137pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.17361pt\hbox{$\scriptstyle{p^{[1]}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 14.4336pt\raise-74.30829pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 69.5339pt\raise-47.92886pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 69.5339pt\raise-62.35164pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.52779pt\hbox{$\scriptstyle{\widehat{p}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 69.5339pt\raise-72.30775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 151.83907pt\raise-41.89554pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern-14.4336pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm F}\nolimits(\kappa_{1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 14.4336pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 29.56116pt\raise-76.85498pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.6639pt\hbox{$\scriptstyle{i_{1}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 52.5929pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 52.5929pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 86.47491pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 106.16992pt\raise-77.65637pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.8625pt\hbox{$\scriptstyle{\kappa_{1}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 124.63422pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 124.63422pt\raise-82.80775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,2),}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ where the vertical arrows form fibration (7). The existence of $p^{[1]}_{k,\,l}$ follows from the universal property of a fiber (applied to the horizontal fibration) together with the fact that the composition $\kappa_{1}\circ\widehat{p}_{k,\,l}$ is homotopy trivial, and $\mathop{\rm Fr}\nolimits^{[1]}_{k,\,l}$ is the fiber of the map $p^{[1]}_{k,\,l}$. ###### Proposition 9. The fiber $\mathop{\rm F}\nolimits(\kappa_{1})$ is homotopy equivalent to $\mathbb{C}P^{\infty}\times\mathop{\rm BSU}\nolimits(k).$ Proof. Consider the diagram $\textstyle{\mathbb{C}P^{\infty}\times\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\alpha}$$\scriptstyle{\lambda}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{i_{1}}$$\textstyle{\mathop{\rm BU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\kappa_{1}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,2),}$ where $\lambda$ is the classifying map for $\zeta^{univ}\otimes\widetilde{\xi}_{k}^{univ},$ where $\widetilde{\xi}_{k}^{univ}\rightarrow\mathop{\rm BSU}\nolimits(k)$ is the universal vector $\mathbb{C}^{k}$-bundle with the structure group $\mathop{\rm SU}\nolimits(k)$ and $\zeta^{univ}\rightarrow\mathbb{C}P^{\infty}$ is the universal line bundle. The existence of the map $\alpha$ follows from the universal property of a fiber because the composition $\kappa_{1}\circ\lambda$ obviously homotopy trivial. Now an easy calculation with homotopy groups shows that $\alpha$ is in fact a homotopy equivalence.$\quad\square$ Note that if we take $\bar{\kappa}_{1}\colon\mathop{\rm BPU}\nolimits(k)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,2)$ in place of $\kappa_{1}$ then we obtain the fiber $\mathop{\rm F}\nolimits(\bar{\kappa}_{1})\simeq\mathop{\rm BSU}\nolimits(k).$ Clearly that the first factor $\mathbb{C}P^{\infty}$ in $\mathop{\rm F}\nolimits(\kappa_{1})$ has no effect on the next obstructions, therefore abusing notation in what follows we consider only the second factor $\mathop{\rm BSU}\nolimits(k)$ in $\mathop{\rm F}\nolimits(\kappa_{1})$ (and replace ${\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))$ by ${\rm H}_{k,\,l}(A_{k}^{univ})$ in diagrams analogous to (15), cf. Proposition 3). Now assume that for the bundle $A_{k}\stackrel{{\scriptstyle p_{k}}}{{\rightarrow}}X$ the first obstruction $\kappa_{1}(A_{k})$ is equal to $0$. Taking into account the previous remark we can assume that such bundle has the form $\mathop{\rm End}\nolimits(\widetilde{\xi}_{k})$ for some vector $\mathbb{C}^{k}$-bundle ${\widetilde{\xi}_{k}}$ with the structure group $\mathop{\rm SU}\nolimits(k)$ (note that its isomorphism class is uniquely determined). Equivalently, we consider a lift of the classifying map $f\colon X\rightarrow\mathop{\rm BU}\nolimits(k)$ to a map $\widetilde{f}\colon X\rightarrow\mathop{\rm BSU}\nolimits(k).$ It follows from the topological obstruction theory (applied to homotopy groups (4)) that the next obstruction $\kappa_{2}$ belongs to $H^{4}(X,\,\mathbb{Z}/k\mathbb{Z})$. The diagram for the definition of $\kappa_{2}$ has the form (16) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 14.4336pt\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 47.07137pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits^{[1]}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 58.24915pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 58.24915pt\raise-30.15999pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 119.70287pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 119.70287pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 69.42694pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 120.40482pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\kern 78.06471pt\raise-30.4228pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern-3.0pt\raise-41.83997pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 38.4336pt\raise-41.83997pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 49.64696pt\raise-47.87329pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 17.43405pt\raise-55.92801pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.17361pt\hbox{$\scriptstyle{p^{[2]}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 14.4336pt\raise-72.57619pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 58.24915pt\raise-47.87329pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 58.24915pt\raise-62.26828pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.17361pt\hbox{$\scriptstyle{p^{[1]}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 58.24915pt\raise-72.19661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 127.88065pt\raise-41.83997pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern-14.4336pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm F}\nolimits(\kappa_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 14.4336pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 38.53035pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 38.53035pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 77.96794pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 88.54834pt\raise-77.54523pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.8625pt\hbox{$\scriptstyle{\kappa_{2}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 102.0647pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 102.0647pt\raise-82.69661pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4)}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ (recall that we take $\mathop{\rm F}\nolimits(\bar{\kappa}_{1})=\mathop{\rm BSU}\nolimits(k)$ in place of $\mathop{\rm F}\nolimits(\kappa_{1})$ rejecting the factor $\mathbb{C}P^{\infty}$), where $\kappa_{2}\colon\mathop{\rm BSU}\nolimits(k)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4)$ is the map representing the class $\kappa_{2}$, and $\mathop{\rm F}\nolimits(\kappa_{2})$ is its homotopy fiber. The map $p^{[2]}_{k,\,l}$ exists because of $\kappa_{2}\circ p^{[1]}_{k,\,l}\simeq$ and $\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l}$ is the homotopy fiber of $p^{[2]}_{k,\,l}$. Note also that $\mathop{\rm Fr}\nolimits^{[1]}_{k,\,l}=\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}$ is the universal cover for $\mathop{\rm Fr}\nolimits_{k,\,l}.$ As we have already known, the second obstruction $\kappa_{2}$ is $c_{2}(\widetilde{\xi}_{k})\mathop{\rm mod}\nolimits\,k$, where $c_{2}$ is the second Chern class. Clearly that the homomorphism $(\kappa_{2})_{}\colon\pi_{4}(\mathop{\rm BSU}\nolimits(k))(\cong\mathbb{Z})\rightarrow\pi_{4}(\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4))(\cong\mathbb{Z}/k\mathbb{Z})$ is onto and it follows from the homotopy sequence of the fibration (17) $\mathop{\rm F}\nolimits(\kappa_{2})\to\mathop{\rm BSU}\nolimits(k)\stackrel{{\scriptstyle\kappa_{2}}}{{\to}}\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4)$ that $\pi_{4}(\mathop{\rm F}\nolimits(\kappa_{2}))\cong\mathbb{Z}$ and the homomorphism $\pi_{4}(\mathop{\rm F}\nolimits(\kappa_{2}))\to\pi_{4}(\mathop{\rm BSU}\nolimits(k))$ is $\mathbb{Z}\to\mathbb{Z},\,1\mapsto k.$ Moreover, the fiber inclusion $\mathop{\rm F}\nolimits(\kappa_{2})\to\mathop{\rm BSU}\nolimits(k)$ induces the isomorphisms $\pi_{r}(\mathop{\rm F}\nolimits(\kappa_{2}))\cong\pi_{r}(\mathop{\rm BSU}\nolimits(k))$ for $r\neq 4.$ Let us return to diagram (16). Consider the following piece of the morphism of the homotopy sequences: $\textstyle{\pi_{5}(\mathop{\rm F}\nolimits(\kappa_{2}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{4}(\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{4}({\rm H}_{k,\,l}(A_{k}^{univ}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{\pi_{5}(\mathop{\rm BSU}\nolimits(k))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{4}(\mathop{\rm Fr}\nolimits^{[1]}_{k,\,l})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{4}({\rm H}_{k,\,l}(A_{k}^{univ}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$ $\textstyle{\pi_{4}(\mathop{\rm F}\nolimits(\kappa_{2}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{3}(\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{3}({\rm H}_{k,\,l}(A_{k}^{univ}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{\pi_{4}(\mathop{\rm BSU}\nolimits(k))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{3}(\mathop{\rm Fr}\nolimits^{[1]}_{k,\,l})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\pi_{3}({\rm H}_{k,\,l}(A_{k}^{univ})),}$ i.e. $\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{\pi_{4}(\mathop{\rm Fr}\nolimits_{k,\,l}^{[2]})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathbb{Z}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\scriptstyle{=}$$\textstyle{\mathbb{Z}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\cdot k}$$\textstyle{\pi_{3}(\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathbb{Z}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\cdot k}$$\textstyle{\mathbb{Z}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathbb{Z}/k\mathbb{Z}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{0,}$ whence $\pi_{3}(\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l})=0=\pi_{4}(\mathop{\rm Fr}\nolimits^{[2]}_{k,\,l}),\;p_{k,\,l}^{[2]}\colon\pi_{4}({\rm H}_{k,\,l}(A_{k}^{univ}))\stackrel{{\scriptstyle\cong}}{{\rightarrow}}\pi_{4}(\mathop{\rm F}\nolimits(\kappa_{2})).$ It is easy to see that the map $\mathop{\rm Fr}\nolimits_{k,\,l}^{[2]}\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}^{[1]}$ (see (16)) induces isomorphisms $\pi_{r}(\mathop{\rm Fr}\nolimits_{k,\,l}^{[2]})\cong\pi_{r}(\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l})$ for $r\neq 3.$ In particular, $\pi_{2r+1}(\mathop{\rm Fr}\nolimits_{k,\,l}^{[2]})\cong\mathbb{Z}/k\mathbb{Z}$ for $r\geq 2$ and $0$ otherwise. Now assume that for the $\mathop{\rm SU}\nolimits(k)$-bundle $\xi_{k}\rightarrow X$ the second obstruction $\kappa_{2}(\xi_{k})=0.$ Assume that we have chosen a lift $f^{[2]}_{\xi}\colon X\rightarrow\mathop{\rm F}\nolimits(\kappa_{2})$ of its classifying map $f_{\xi}\colon X\rightarrow\mathop{\rm BSU}\nolimits(k).$ According to the obstruction theory, such a lift is not unique, more precisely, the set of homotopy classes of lifts are in one-to-one correspondence with the set $H^{3}(X,\,\mathbb{Z}/k\mathbb{Z})$ (because the fiber of the map $\mathop{\rm F}\nolimits(\kappa_{2})\rightarrow\mathop{\rm BSU}\nolimits(k)$ is $\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,3),$ see diagram (16)). The next obstruction is a characteristic class $\kappa_{3}\in H^{6}(\mathop{\rm F}\nolimits(\kappa_{2}),\,\pi_{5}(\mathop{\rm Fr}\nolimits_{k,\,l}^{[2]}))=H^{6}(\mathop{\rm F}\nolimits(\kappa_{2}),\,\mathbb{Z}/k\mathbb{Z}).$ Consider the following fibration (cf. (17)): $\mathop{\rm F}\nolimits(\kappa_{3})\to\mathop{\rm F}\nolimits(\kappa_{2})\stackrel{{\scriptstyle\kappa_{3}}}{{\to}}\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,6)$ and the corresponding diagram (cf. (16)): $\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}^{[2]}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm Fr}\nolimits^{[3]}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{p^{[3]}_{k,\,l}}$$\scriptstyle{p^{[2]}_{k,\,l}}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{3})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\kappa_{3}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,6).}$ We claim that the homomorphism $(\kappa_{3})_{}\colon\pi_{6}(\mathop{\rm F}\nolimits(\kappa_{2}))(\cong\mathbb{Z})\rightarrow\pi_{6}(\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,6))(\cong\mathbb{Z}/k\mathbb{Z})$ is onto. Indeed, take $\varphi\colon S^{6}\rightarrow\mathop{\rm F}\nolimits(\kappa_{2})$ such that $k\nmid[\varphi]$ in $\pi_{6}(\mathop{\rm F}\nolimits(\kappa_{2}))\cong\mathbb{Z}\;\Rightarrow\;\varphi$ can not be lifted to $\widetilde{\varphi}\colon S^{6}\rightarrow{\rm H}_{k,\,l}(A_{k}^{univ})$ (because $(p_{k,\,l}^{[2]})_{}\colon\pi_{6}({\rm H}_{k,\,l}(A_{k}^{univ}))\rightarrow\pi_{6}(\mathop{\rm F}\nolimits(\kappa_{2}))$ is the homomorphism $\mathbb{Z}\rightarrow\mathbb{Z},\;1\mapsto k)\;\Rightarrow\;\varphi^{}(\kappa_{3})\neq 0\in H^{6}(S^{6},\,\mathbb{Z}/k\mathbb{Z})$ (because this is the unique obstruction in this case). In particular, $\kappa_{3}$ is an element of order $k$ in the group $H^{6}(\mathop{\rm F}\nolimits(\kappa_{2}),\,\mathbb{Z}/k\mathbb{Z}).$ This implies that the inclusion $\mathop{\rm F}\nolimits(\kappa_{3})\to\mathop{\rm F}\nolimits(\kappa_{2})$ of the homotopy fiber induces the homomorphism $\pi_{6}(\mathop{\rm F}\nolimits(\kappa_{3}))\to\pi_{6}(\mathop{\rm F}\nolimits(\kappa_{2})),\;\mathbb{Z}\to\mathbb{Z},\,1\mapsto k$ in dimension $6$ and isomorphisms $\pi_{r}(\mathop{\rm F}\nolimits(\kappa_{3}))\stackrel{{\scriptstyle\cong}}{{\to}}\pi_{r}(\mathop{\rm F}\nolimits(\kappa_{2}))$ in other dimensions $r\neq 6.$ It is easy to see that $\pi_{2r+1}(\mathop{\rm Fr}\nolimits_{k,\,l}^{[3]})\cong\mathbb{Z}/k\mathbb{Z}$ for $r\geq 3$ and $0$ otherwise. In particular, $p_{k,\,l}^{[3]}$ induces a homotopy equivalence between ${\rm H}_{k,\,l}(A_{k}^{univ})$ and $\mathop{\rm F}\nolimits(\kappa_{3})$ up to dimension $7$. Using this pattern, by induction one can define characteristic classes $\kappa_{4}\in H^{8}(\mathop{\rm F}\nolimits(\kappa_{3}),\,\mathbb{Z}/k\mathbb{Z}),\,\kappa_{5}\in H^{10}(\mathop{\rm F}\nolimits(\kappa_{4}),\,\mathbb{Z}/k\mathbb{Z})$, etc., each of which is defined on the kernel of the predecessor. Note that $(p^{[i]}_{k,\,l})_{}\colon\pi_{r}({\rm H}_{k,\,l}(A_{k}^{univ}))\rightarrow\pi_{r}(\mathop{\rm F}\nolimits(\kappa_{i}))$ is an isomorphism for $r\leq 2i+1$. Let $\xi_{k}\rightarrow X$ be a $\mathop{\rm U}\nolimits(k)$-bundle with a classifying map $f_{\xi}\colon X\rightarrow\mathop{\rm BU}\nolimits(k)$, $\dim X<2\min\\{k,\,l\\}.$ A lift $\widehat{f}_{\xi}\colon X\rightarrow{\rm H}_{k,\,l}(A_{k}^{univ})$ of the classifying map $f_{\xi}$ can be constructed by induction as a sequence of maps $f_{\xi}^{[i-1]},\;2\leq i\leq(\dim X)/2$ that make the diagram ($i\geq 2$) (18) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 6.75pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&\\\&&&&\\\&&\\\\}}}\ignorespaces{\hbox{\kern-3.0pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 45.326pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 89.90201pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 100.6499pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 63.83333pt\raise-14.08804pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.17361pt\hbox{$\scriptstyle{p^{[i+1]}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 64.10677pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 109.71756pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 96.59541pt\raise-20.42831pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.17361pt\hbox{$\scriptstyle{p^{[i]}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 109.71756pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 118.59079pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 137.15567pt\raise-14.08804pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.17361pt\hbox{$\scriptstyle{p^{[i-1]}_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 154.42229pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}{\hbox{\kern-6.75pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\ldots\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 6.75pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 30.75pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 30.75pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm F}\nolimits(\kappa_{i+1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 65.90202pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 51.82298pt\raise-34.35664pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.75pt\hbox{$\scriptstyle{\quad\;\;\mathop{\rm K}\nolimits(\mathbb{Z}_{k},\,2i+1)}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 95.71933pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 95.71933pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm F}\nolimits(\kappa_{i})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 123.71579pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 116.9256pt\raise-34.35664pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.75pt\hbox{$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z}_{k},\,2i-1)\quad}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 153.53311pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 153.53311pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm F}\nolimits(\kappa_{i-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 186.19623pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 210.19623pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 210.19623pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\ldots}$}}}}}}}{\hbox{\kern-3.0pt\raise-79.2733pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 45.326pt\raise-79.2733pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 102.18285pt\raise-79.2733pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{X\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 102.18285pt\raise-74.56412pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 63.65363pt\raise-67.32884pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.09723pt\hbox{$\scriptstyle{f^{[i+1]}_{\xi}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 57.10774pt\raise-46.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 109.71756pt\raise-69.43999pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 109.71756pt\raise-60.06497pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.09723pt\hbox{$\scriptstyle{f^{[i]}_{\xi}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 109.71756pt\raise-46.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 117.25227pt\raise-74.4611pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 136.6992pt\raise-67.32884pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.09723pt\hbox{$\scriptstyle{f^{[i-1]}_{\xi}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 161.2548pt\raise-46.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}\ignorespaces\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces{\hbox{\lx@xy@drawline@}}\ignorespaces\ignorespaces}}}}}\end{array}$ commutative ($\mathop{\rm F}\nolimits(\kappa_{1}):=\mathop{\rm BSU}\nolimits(k),\;p^{[1]}_{k,\,l}=\widetilde{p}_{k,\,l},\;f_{\xi}^{[1]}=f_{\xi}$). The extension $f_{\xi}^{[i]}$ of $f_{\xi}^{[i-1]}$ exists only if $(f_{\xi}^{[i-1]})^{}(\kappa_{i})=0,\;f_{\xi}^{[0]}:=f_{\xi},\;1\leq i\leq(\dim X)/2,$ where $\kappa_{i}\in H^{2i}(\mathop{\rm F}\nolimits(\kappa_{i-1}),\,\mathbb{Z}/k\mathbb{Z})$ are the above defined characteristic classes. (Note that by abusing notation we take $\mathop{\rm F}\nolimits(\kappa_{1}):=\mathop{\rm BSU}\nolimits(k)$, but it is more natural to put $\kappa_{1}:=c_{1}\mathop{\rm mod}\nolimits k$ and $\mathop{\rm F}\nolimits(\kappa_{1})$ to be the homotopy fiber of $\mathop{\rm BU}\nolimits(k)\stackrel{{\scriptstyle\kappa_{1}}}{{\rightarrow}}\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,2)$ which, recall, is $\mathop{\rm BSU}\nolimits(k)\times\mathbb{C}P^{\infty}$. In this case every $\mathop{\rm F}\nolimits(\kappa_{i})$ should be replaced by $\mathop{\rm F}\nolimits(\kappa_{i})\times\mathbb{C}P^{\infty}$). In other words, a lift $\widehat{f}_{\xi}\colon X\rightarrow{\rm H}_{k,\,l}(A_{k}^{univ})$ is constructed step by step, and the obstruction on the $i$th step is $(f_{\xi}^{[i-1]})^{}(\kappa_{i})\in H^{2i}(X,\,\mathbb{Z}/k\mathbb{Z})$. The following theorem summarizes the obtained results. ###### Theorem 10. There is a lift $\widehat{f}_{\xi}\colon X\rightarrow{\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))$ of $f_{\xi}$ iff there is a sequence of maps $f_{\xi}^{[i-1]},\;2\leq i\leq(\dim X)/2$ making diagram (18) commutative and such that $(f_{\xi}^{[i-1]})^{}(\kappa_{i})=0,\;1\leq i\leq(\dim X)/2,$ where $\kappa_{i}\in H^{2i}(\mathop{\rm F}\nolimits(\kappa_{i-1}),\,\mathbb{Z}/k\mathbb{Z})$ are the above defined characteristic classes. Note that on every step the choice of a lift is not unique in general. Moreover, the defined obstructions are functorial and are “well-behaved” with respect to the tensor product of bundles (cf. (25)). In particular, the inclusion $\mathop{\rm BPU}\nolimits(k^{m})\rightarrow\mathop{\rm BPU}\nolimits(k^{m+1})$ (induced by the tensor product by the trivial $M_{k}(\mathbb{C})$-bundle) induces the embedding of obstructions $\kappa_{r}^{(k^{m})}\in H^{2r}(X,\,\mathbb{Z}/k^{m}\mathbb{Z})\rightarrow H^{2r}(X,\,\mathbb{Z}/k^{m+1}\mathbb{Z})\ni\kappa_{r}^{(k^{m+1})}$ corresponding to the embedding of the coefficient groups $\mathbb{Z}/k^{m}\mathbb{Z}\rightarrow\mathbb{Z}/k^{m+1}\mathbb{Z}.$ ### 1.5. Relation to the Chern classes of connective covers of $\mathop{\rm BU}\nolimits$ In general, “higher” obstructions (in stable dimensions) are in $H^{2r}(X,\,\mathbb{Z}/k\mathbb{Z}),\>r\in\mathbb{N}$. But for $r>2$ they do not coincide with the Chern classes reduced modulo $k$. To see this, take $X=S^{8}$ and consider a $6$-dimensional complex vector bundle $\xi_{6}\rightarrow S^{8}.$ It is well-known [4] that for $S^{2r}$ the Chern classes of complex vector bundles form the subgroup of index $(r-1)!$ in $H^{2r}(S^{2r},\,\mathbb{Z})\cong\mathbb{Z}$. In particular, in our case $r=4,\,k=6$ we have $c_{4}(\xi_{6})\equiv 0\,\rm{(mod\;6)}$, but it follows from the homotopy sequence of fibration (5) (or (13)) that not every such a bundle has a lift. In order to go further, one can use the modification of Chern classes for connective covers of $\mathop{\rm BU}\nolimits$. More precisely, let $\iota_{r}\colon\mathop{\rm BU}\nolimits\langle 2r\rangle\rightarrow\mathop{\rm BU}\nolimits$ be the connective cover of $\mathop{\rm BU}\nolimits$ whose first non-zero homotopy is in degree $2r$ (thus $\mathop{\rm BU}\nolimits\langle 2\rangle=\mathop{\rm BU}\nolimits,\;\mathop{\rm BU}\nolimits\langle 4\rangle=\mathop{\rm BSU}\nolimits,\ldots$). Then the image of the $r$’th Chern class under the pullback $\iota_{r}^{}\colon H^{}(\mathop{\rm BU}\nolimits,\,\mathbb{Z})\to H^{}(\mathop{\rm BU}\nolimits\langle 2r\rangle,\,\mathbb{Z})$ is divisible by $(r-1)!$111I am grateful to Professor Thomas Schick for bringing this result to my attention. [6]. Put $\widetilde{c}_{r}:=\frac{\iota_{r}^{}(c_{r})}{(r-1)!}.$ The following theorem generalizes Theorems 6 and 8. ###### Theorem 11. For bundles classified by the space $\mathop{\rm BU}\nolimits\langle 2r\rangle$ the first obstruction to the considered lifting problem is $\widetilde{c}_{r}\,\mathop{\rm mod}\nolimits\,k$. Proof. For the connective cover $\iota_{r,\,k}\colon\mathop{\rm BU}\nolimits(k)\langle 2r\rangle\rightarrow\mathop{\rm BU}\nolimits(k),\>k>r$ consider the $\mathop{\rm Fr}\nolimits_{k,\,l}$-fibration (19) $\iota_{r,\,k}^{}({\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ})))\rightarrow\mathop{\rm BU}\nolimits(k)\langle 2r\rangle$ induced from (7). Clearly, the first obstruction to lifting in this fibration is a characteristic class $\omega_{r}\in H^{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\pi_{2r-1}(\mathop{\rm Fr}\nolimits_{k,\,l}))=H^{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\mathbb{Z}/k\mathbb{Z})\cong\mathbb{Z}/k\mathbb{Z}.$ It follows from the homotopy sequence of fibration (19) that $\pi_{2r}(\iota_{r,\,k}^{}({\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))))\cong\mathbb{Z}$, the homomorphism $\pi_{2r}(\iota_{r,\,k}^{}({\rm H}_{k,\,l}(\mathop{\rm End}\nolimits(\xi_{k}^{univ}))))\rightarrow\pi_{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle)$ is injective and its image is the subgroup of index $k$ in $\pi_{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle)\cong\mathbb{Z}.$ Now using the same argument as in the proof of Theorem 8 with $S^{2r}$ in place of $S^{4}$, we see that for the bundle $\xi_{k}\rightarrow S^{2r}$ corresponding to the generator $1\in\pi_{2r}(\mathop{\rm BU}\nolimits(k))\cong\mathbb{Z}$ the class $\omega_{r}$ is a generator of $H^{2r}(S^{2r},\,\mathbb{Z}/k\mathbb{Z})\cong\mathbb{Z}/k\mathbb{Z},$ i.e. $\omega_{r}=\alpha\widetilde{c}_{r}\,\mathop{\rm mod}\nolimits\,k$, where $\alpha$ is invertible $\mathop{\rm mod}\nolimits\,k$. In order to prove that $\alpha=1$ we may use the argument analogous the conclusions of the proofs of Theorems 6 and 8. More precisely, it follows form Hurewicz’s theorem and the universal coefficients formula that $H^{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\mathbb{Z})\cong\mathbb{Z}.$ Furthermore, since $\widetilde{c}_{r}(\xi_{k})$ is a generator of $H^{2r}(S^{2r},\,\mathbb{Z})$ for some vector bundle $\xi_{k}\rightarrow S^{2r}$, we see that $\widetilde{c}_{r}(\xi_{k})\in H^{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\mathbb{Z})$ is a generator too. Let (20) $\iota_{r,\,k}^{}(\mathop{\rm EU}\nolimits(k))\rightarrow\mathop{\rm BU}\nolimits(k)\langle 2r\rangle$ be a $\mathop{\rm U}\nolimits(k)$-bundle induced from the universal $\mathop{\rm U}\nolimits(k)$-bundle $\mathop{\rm EU}\nolimits(k)\rightarrow\mathop{\rm BU}\nolimits(k)$ by the map $\iota_{r,\,k}$. Let $h\colon S^{2r}\rightarrow\mathop{\rm BU}\nolimits(k)\langle 2r\rangle$ be a generator in $H_{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\mathbb{Z})\cong\mathbb{Z}$. Since the homomorphism $\pi_{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle)\rightarrow\pi_{2r-1}(\mathop{\rm U}\nolimits(k))$ in the homotopy sequence of fibration (20) is an isomorphism, we see that the first obstruction to the lift in fibration (20) is a generator in $H^{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\mathbb{Z})=\mathop{\rm Hom}\nolimits(H_{2r}(\mathop{\rm BU}\nolimits(k)\langle 2r\rangle,\,\mathbb{Z}),\,\mathbb{Z})$ (i.e. it is the fundamental class $\mathop{\rm id}\nolimits\colon\mathbb{Z}\rightarrow\mathbb{Z}$), hence it is $\widetilde{c}_{r}$. Note that the map $\mathop{\rm SU}\nolimits(k)\rightarrow\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}$ induces the following homomorphism of homotopy groups: $\pi_{2r-1}(\mathop{\rm SU}\nolimits(k))\rightarrow\pi_{2r-1}(\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}),\;\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z},\;\forall r\geq 2.$ The proof concludes the same argument as in the conclusion of the proof of Theorem 8.$\quad\square$ The obtained result can also be reformulated as follows. Let $\xi_{k}\rightarrow X$ be a vector $\mathbb{C}^{k}$-bundle such that $c_{1}(\xi_{k})=0,\,c_{2}(\xi_{k})=0$. The its classifying map $f_{\xi}\colon X\rightarrow\mathop{\rm BSU}\nolimits(k)$ can be lifted to $\mathop{\rm BU}\nolimits(k)\langle 6\rangle:$ $\textstyle{\ldots\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 8\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,5)}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 6\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,3)\qquad}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 4\rangle=\mathop{\rm BSU}\nolimits(k)}$$\textstyle{X\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{f^{(4)}_{\xi}}$$\scriptstyle{f^{(3)}_{\xi}}$$\scriptstyle{f_{\xi}=f^{(2)}_{\xi}}$ (the upper row in the diagram is the Whitehead tower for $\mathop{\rm BSU}\nolimits(k)$). In fact, the space $\mathop{\rm BU}\nolimits(k)\langle 6\rangle$ represents some refined (because the choice of a lift $f^{(3)}_{\xi}$ is not unique) theory of bundles with $c_{2}=0$ and we can regard the lift $f^{(3)}_{\xi}$ as a classifying map for some bundle $\xi^{(3)}_{k}$ of this type. Thus, we have the characteristic class $\widetilde{c}_{3}(\xi_{k}^{(3)}):=f^{(3)}_{\xi}(\widetilde{c}_{3})\in H^{6}(X,\,\mathbb{Z})$. If $\widetilde{c}_{3}(\xi_{k}^{(3)})=0,$ then we choose a lift $f^{(4)}_{\xi}$ (see the above diagram) corresponding to some bundle $\xi_{k}^{(4)}$ (of even more “subtle” type of bundles with $\widetilde{c}_{2}=0,\,\widetilde{c}_{3}=0$) with the characteristic class $\widetilde{c}_{4}(\xi^{(4)}_{k})\in H^{8}(X,\,\mathbb{Z}),$ etc. Suppose that starting with the bundle $\xi_{k}\rightarrow X$ we obtain a sequence of bundles $\xi_{k}^{(i)},\;i\leq r,\;\widetilde{c}_{i}(\xi^{(i)}_{k})=0$ for $i<r$. Then the first obstruction for embedding $\mu\colon\mathop{\rm End}\nolimits(\xi_{k}^{(r)})\hookrightarrow X\times M_{kl}(\mathbb{C}),\;(k,\,l)=1$ is $\widetilde{c}_{r}(\xi_{k}^{(r)})\mathop{\rm mod}\nolimits k$. In order to establish a relation between two outlined approaches to definition of obstructions, consider the diagram: $\textstyle{\mathop{\rm BU}\nolimits(k)\langle 6\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{3}}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{c_{2}}$$\scriptstyle{\kappa_{2}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4)}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,4)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$ whose low right triangle is commutative and the map $\lambda_{3}$ is defined by the universal property of a fiber (and the right vertical arrow $\mathop{\rm K}\nolimits(\mathbb{Z},\,4)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4)$ is induced by the surjective homomorphism $\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z}$). Because of the functoriality of the first obstruction we have: (21) $\lambda_{3}^{}(\kappa_{3})=\widetilde{c}_{3}\,\mathop{\rm mod}\nolimits\,k.$ Now consider the diagram $\textstyle{\mathop{\rm BU}\nolimits(k)\langle 8\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{4}}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 6\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\widetilde{c}_{3}}$$\scriptstyle{\lambda_{3}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,6)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{3})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\kappa_{3}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,6),}$ where the right square is commutative because of (21) (note that the right vertical arrow is induced by the surjective homomorphism $\mathbb{Z}\rightarrow\mathbb{Z}/k\mathbb{Z}$). Again using the functoriality of the first obstruction, we have: $\lambda_{4}^{}(\kappa_{4})=\widetilde{c}_{4}\,\mathop{\rm mod}\nolimits\,k.$ Repeating this argument by induction, we obtain the following commutative diagram: $\textstyle{\ldots\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,7)\quad}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 8\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{4}}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,5)}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 6\rangle\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,3)\qquad}$$\scriptstyle{\lambda_{3}}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 4\rangle=\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{=}$$\textstyle{\ldots\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,7)\quad}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{3})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,5)}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,3)\qquad}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{1})=\mathop{\rm BSU}\nolimits(k),}$ where the symbols over arrows indicate the fibers. The following theorem summarizes the obtained result. ###### Theorem 12. For any $r\geq 2\;\lambda^{}_{r}(\kappa_{r})=\widetilde{c}_{r}\,\mathop{\rm mod}\nolimits\,k,$ where $\lambda_{r}\colon\mathop{\rm BU}\nolimits(k)\langle 2r\rangle\rightarrow\mathop{\rm F}\nolimits(\kappa_{r-1})$ are the above defined maps (note that $\lambda_{2}=\mathop{\rm id}\nolimits\colon\mathop{\rm BSU}\nolimits(k)\rightarrow\mathop{\rm BSU}\nolimits(k)$ in accordance with $\kappa_{2}=c_{2}\,\mathop{\rm mod}\nolimits\,k$). Note that this theorem generalizes the relation between the obstructions $\kappa_{1},\,\kappa_{2}$ and the corresponding Chern classes $c_{1},\,c_{2}$ (see Theorems 6 and 8). (More precisely, this theorem shows that we should consider the Chern classes $\widetilde{c}_{r}$ of the connective covers of $\mathop{\rm BU}\nolimits.$) This follows from the diagram $\textstyle{\mathop{\rm BU}\nolimits(k)\langle 4\rangle=\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{2}}$$\textstyle{\mathop{\rm BU}\nolimits(k)\langle 2\rangle=\mathop{\rm BU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\lambda_{1}}$$\scriptstyle{=}$$\scriptstyle{\qquad c_{1}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z},\,2)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm F}\nolimits(\kappa_{1})=\mathbb{C}P^{\infty}\times\mathop{\rm BSU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm BU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\kappa_{1}}$$\textstyle{\mathop{\rm K}\nolimits(\mathbb{Z}/k\mathbb{Z},\,4).}$ ## 2\. An approach via groupoids It turns out that above considered spaces and bundles (like $\mathop{\rm Gr}\nolimits_{k,\,l},\;{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l}),\;{\rm H}_{k,\,l}(A_{k}^{univ})$ etc.) can naturally be interpreted in terms of some groupoid $\mathfrak{G}_{k,\,l}$ of matrix subalgebras in the fixed matrix algebra $M_{kl}(\mathbb{C})$. ### 2.1. Groupoids $\mathfrak{G}_{k,\,l}$ Let $M_{kl}(\mathbb{C})$ be the complex matrix algebra. Recall that unital $$-subalgebras in $M_{kl}(\mathbb{C})$ isomorphic to $M_{k}(\mathbb{C})$ we call $k$-subalgebras. Define the following category $C_{k,\,l}.$ Its objects $\mathop{\rm Ob}\nolimits(C_{k,\,l})$ are $k$-subalgebras in the fixed $M_{kl}(\mathbb{C})$, i.e. actually points of the matrix grassmannian $\mathop{\rm Gr}\nolimits_{k,\,l}$. For two objects $M_{k,\,\alpha},\,M_{k,\,\beta}\in Ob(C_{k,\,l})$ the set of morphisms $\mathop{\rm Mor}\nolimits_{C_{k,\,l}}(M_{k,\,\alpha},\,M_{k,\,\beta})$ is just the space $\mathop{\rm Hom}\nolimits_{alg}(M_{k,\,\alpha},\,M_{k,\,\beta})$ of all unital $$-homomorphisms of matrix algebras (i.e. actually isometric isomorphisms). Put $\mathfrak{G}_{k,\,l}^{0}:=\mathop{\rm Ob}\nolimits(C_{k,\,l}),\quad\mathfrak{G}_{k,\,l}:=\bigcup\limits_{\alpha,\,\beta\in\mathop{\rm Ob}\nolimits(C_{k,\,l})}\mathop{\rm Mor}\nolimits_{C_{k,\,l}}(M_{k,\,\alpha},\,M_{k,\,\beta}).$ Clearly, $\mathfrak{G}_{k,\,l}$ is a topological groupoid (in fact, even a Lie groupoid). ###### Remark 13. Note that we do not fix an extension of a homomorphism from $\mathop{\rm Hom}\nolimits_{alg}(M_{k,\,\alpha},\,M_{k,\,\beta})$ to an automorphism of the whole algebra $M_{kl}(\mathbb{C})$, so it is not the action groupoid corresponding to the action of $\mathop{\rm PU}\nolimits(kl)$ on $\mathop{\rm Ob}\nolimits(C_{k,\,l})$. It is interesting to note that if $\mathfrak{G}_{k,\,l}$ would be an action groupoid for some topological group $\mathop{\rm H}\nolimits$ acting on $\mathfrak{G}_{k,\,l}^{0},$ then $\mathop{\rm H}\nolimits\simeq\mathop{\rm Fr}\nolimits_{k,\,l}.$ This result follows from the homotopy equivalence $\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}\simeq\mathop{\rm BPU}\nolimits(k)$ (see below) and the fact that for action groupoid $\mathfrak{G}:=X\rtimes\mathop{\rm H}\nolimits$ corresponding to an action of $\mathop{\rm H}\nolimits$ on $X$ the classifying space $\mathop{\rm B}\nolimits\mathfrak{G}$ is homotopy equivalent to $X{\mathop{\times}\limits_{\mathop{\rm H}\nolimits}}\mathop{\rm EH}\nolimits$ [3]. As a topological space $\mathfrak{G}_{k,\,l}$ can be represented as follows. Applying fiberwisely the functor $\mathop{\rm Hom}\nolimits_{alg}(\ldots,\,M_{kl}(\mathbb{C}))$ (see Subsection 1.1) to the tautological $M_{k}(\mathbb{C})$-bundle ${\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ we obtain the space ${\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})$ which is exactly $\mathfrak{G}_{k,\,l}$. Being a groupoid, $\mathfrak{G}_{k,\,l}$ has canonical morphisms: source and target $s,\,t\colon\mathfrak{G}_{k,\,l}\rightrightarrows\mathfrak{G}_{k,\,l}^{0},$ composition $m\colon\mathfrak{G}_{k,\,l}{\mathop{\times}\limits_{{}^{s}\;\mathfrak{G}_{k,\,l}^{0\quad t}}}\mathfrak{G}_{k,\,l}\rightarrow\mathfrak{G}_{k,\,l},$ identity $e\colon\mathfrak{G}_{k,\,l}^{0}\rightarrow\mathfrak{G}_{k,\,l}$ and inversion $i\colon\mathfrak{G}_{k,\,l}\rightarrow\mathfrak{G}_{k,\,l}$. Let us describe first two of them in terms of topological spaces $\mathop{\rm Gr}\nolimits_{k,\,l}\sim\mathfrak{G}_{k,\,l}^{0}$ and ${\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})\sim\mathfrak{G}_{k,\,l}.$ The source morphism $s\colon{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ is just the bundle projection (recall that ${\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})$ is obtained from the bundle ${\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ by the fiberwise application of the functor $\mathop{\rm Hom}\nolimits_{alg}(\ldots,\,M_{kl}(\mathbb{C}))$). The target morphism $t\colon{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ is the map $h\mapsto h(({\mathcal{A}}_{k,\,l})_{\alpha})$, where $h\in{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l}),\;s(h)=\alpha\in\mathop{\rm Gr}\nolimits_{k,\,l}$ and as usual we identify the $k$-subalgebra $h(({\mathcal{A}}_{k,\,l})_{\alpha})\subset M_{kl}(\mathbb{C})$ with the corresponding point in $\mathop{\rm Gr}\nolimits_{k,\,l}$. There are also analogous descriptions of maps $e\colon\mathop{\rm Gr}\nolimits_{k,\,l}\rightarrow{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l}),\;i\colon{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})\rightarrow{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})$ and (22) $m\colon{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l}){\mathop{\times}\limits_{{}^{s}\;\mathop{\rm Gr}\nolimits_{k,\,l}^{\quad t}}}{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})\rightarrow{\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l}).$ Note that there are bifunctors $C_{k,\,l}\times C_{m,\,n}\rightarrow C_{km,\,ln}$ induced by the tensor product of matrix algebras and therefore the corresponding morphisms of topological groupoids (23) $\mathfrak{G}_{k,\,l}\times\mathfrak{G}_{m,\,n}\rightarrow\mathfrak{G}_{km,\,ln}.$ They cover the maps $\mathop{\rm Gr}\nolimits_{k,\,l}\times\mathop{\rm Gr}\nolimits_{m,\,n}\rightarrow\mathop{\rm Gr}\nolimits_{km,\,ln}$ [12]. ###### Remark 14. Note that one can define an “$\mathop{\rm SU}\nolimits$”-analog of the groupoid $\mathfrak{G}_{k,\,l}$ replacing $\mathop{\rm PU}\nolimits(k)$ by $\mathop{\rm SU}\nolimits(k)$. This is a $k$-fold covering of $\mathfrak{G}_{k,\,l}$ (cf. Subsection 1.3). Note that for any $\alpha\in\mathop{\rm Ob}\nolimits(C_{k,\,l})$ we have the (full) subcategory with one object $\alpha.$ The corresponding groupoid morphism $\mathop{\rm PU}\nolimits(k)\rightarrow\mathfrak{G}_{k,\,l}$ is a Morita morphism, i.e. the diagram $\textstyle{\mathop{\rm PU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathfrak{G}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{s\times t}$$\textstyle{\alpha\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm Gr}\nolimits_{k,\,l}\times\mathop{\rm Gr}\nolimits_{k,\,l}}$ is a Cartesian square. It turns out (see the next subsection) that this Morita morphism induces a homotopy equivalence of the classifying spaces $\mathop{\rm BPU}\nolimits(k)\simeq\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}.$ ### 2.2. Groupoids $\widehat{\mathfrak{G}}_{k,\,l}$ Define a new category $\widehat{C}_{k,\,l}$ whose objects $\mathop{\rm Ob}\nolimits(\widehat{C}_{k,\,l})=\mathop{\rm Ob}\nolimits(C_{k,\,l})$ but morphism from $\alpha\in\mathop{\rm Ob}\nolimits(\widehat{C}_{k,\,l})$ to $\beta\in\mathop{\rm Ob}\nolimits(\widehat{C}_{k,\,l})$ is the set of all pairs $(\lambda,\,\mu),$ where $\lambda\colon M_{k,\,\alpha}\rightarrow M_{k,\,\beta}$ and $\mu\colon M_{l,\,\alpha}\rightarrow M_{l,\,\beta}$ are $$-isomorphisms, where $M_{l,\,\alpha}\cong M_{l}(\mathbb{C}),\;M_{l,\,\beta}\cong M_{l}(\mathbb{C})$ are centralizers (in $M_{kl}(\mathbb{C})$) of $M_{k,\,\alpha}$ and $M_{k,\,\beta}$ respectively. Let $\widehat{\mathfrak{G}}_{k,\,l}$ be the set of all morphisms in $\widehat{C}_{k,\,l}$. Clearly, it is again a topological (even a Lie) groupoid. As a topological space it can also be described as the total space of some $\mathop{\rm PU}\nolimits(k)\times\mathop{\rm PU}\nolimits(l)$-bundle over $\mathop{\rm Gr}\nolimits_{k,\,l}\times\mathop{\rm Gr}\nolimits_{k,\,l}$ (the projection is given by $s\times t\colon\widehat{\mathfrak{G}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}\times\mathop{\rm Gr}\nolimits_{k,\,l}$). We also have the map $\widehat{\vartheta}\colon\widehat{\mathfrak{G}}_{k,\,l}\rightarrow\mathop{\rm PU}\nolimits(kl),\;(\lambda,\,\mu)\mapsto\widehat{\vartheta}(\lambda,\,\mu),$ where $\widehat{\vartheta}(\lambda,\,\mu)\colon M_{kl}(\mathbb{C})\rightarrow M_{kl}(\mathbb{C})$ is the unique automorphism induced by $(\lambda,\,\mu)$. ###### Remark 15. In fact, $\widehat{\mathfrak{G}}_{k,\,l}$ is an action groupoid $\mathop{\rm Gr}\nolimits_{k,\,l}\rtimes\mathop{\rm PU}\nolimits(kl)$ related to the action of $\mathop{\rm PU}\nolimits(kl)$ on $\mathop{\rm Gr}\nolimits_{k,\,l}.$ We have the natural groupoid morphism $\pi\colon\widehat{\mathfrak{G}}_{k,\,l}\rightarrow\mathfrak{G}_{k,\,l},\;(\lambda,\,\mu)\mapsto\lambda.$ The fiber of $\pi$ is clearly $\mathop{\rm PU}\nolimits(l).$ Thus, we have the groupoid extension (24) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 15.63197pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&\\\\}}}\ignorespaces{\hbox{\kern-15.63197pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm PU}\nolimits(l)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 15.63199pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 39.63197pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 39.63197pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\widehat{\mathfrak{G}}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 58.94917pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 67.22144pt\raise 4.50694pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.50694pt\hbox{$\scriptstyle{\pi}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 82.94917pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 82.94917pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathfrak{G}_{k,\,l}.}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ ###### Remark 16. Note that $\mathfrak{G}_{k,\,l}$ can also be regarded as an extension of the pair groupoid $\mathop{\rm Gr}\nolimits_{k,\,l}\times\mathop{\rm Gr}\nolimits_{k,\,l}$ by $\mathop{\rm PU}\nolimits(k).$ ### 2.3. Universal principal groupoid $\mathfrak{G}_{k,\,l}$-bundle In this subsection we shall show that our previous construction (see Subsection 1.1) which to an $M_{k}(\mathbb{C})$-bundle $A_{k}\rightarrow X$ associates $\mathop{\rm Fr}\nolimits_{k,\,l}$-bundle ${\rm H}_{k,\,l}(A_{k})\rightarrow X$ is nothing but the extension functor from the structure group $\mathop{\rm PU}\nolimits(k)$ to the structure groupoid $\mathfrak{G}_{k,\,l}.$ Moreover, it turns out that ${\rm H}_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm BPU}\nolimits(k)$ is the universal principal $\mathfrak{G}_{k,\,l}$-bundle, in particular, the classifying spaces $\mathop{\rm BPU}\nolimits(k)$ and $\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}$ are homotopy equivalent. Consequently, every $\mathfrak{G}_{k,\,l}$-bundle can be obtained from some $M_{k}(\mathbb{C})$-bundle in this way. ###### Remark 17. Note that $\mathop{\rm B}\nolimits\widehat{\mathfrak{G}}_{k,\,l}\simeq\mathop{\rm BPU}\nolimits(k)\times\mathop{\rm BPU}\nolimits(l)$ because $\widehat{\mathfrak{G}}_{k,\,l}$ is an action groupoid (cf. Remarks 13 and 15). In Subsection 1.1 (see (9)) we defined the map $\tau_{k,\,l}\colon\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l},\;h\mapsto h((A_{k}^{univ})_{x})\subset M_{kl}(\mathbb{C}),$ where $x\in\mathop{\rm BPU}\nolimits(k)$ and $h\in p_{k,\,l}^{-1}(x)$ which is a fibration with contractible fibres; in particular, it is a homotopy equivalence. There is the free and proper action $\varphi\colon\mathfrak{G}_{k,\,l}{\mathop{\times}\limits_{{}^{s}\;\mathop{\rm Gr}\nolimits_{k,\,l}^{\;\tau}}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})$ ($\tau:=\tau_{k,\,l}$) defined by the compositions of algebra homomorphisms. More precisely, for $g\in\mathfrak{G}_{k,\,l},\>h\in p_{k,\,l}^{-1}(x),\>x\in\mathop{\rm BPU}\nolimits(k)$ such that $s(g)=\tau_{k,\,l}(h)$ we put $\varphi(g,\,h):=g(h((A_{k}^{univ})_{x}))\subset M_{kl}(\mathbb{C})$ (in particular, $\tau_{k,\,l}(\varphi(g,\,h))=t(g)$). ###### Theorem 18. The base space of the principal groupoid $\mathfrak{G}_{k,\,l}$-bundle $(\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ}),\,\mathfrak{G}_{k,\,l},\,\varphi)$ is $\mathop{\rm BPU}\nolimits(k)$ (see (5)). Proof. It is easy to see that the map $\mathfrak{G}_{k,\,l}{\mathop{\times}\limits_{{}^{s}\;\mathfrak{G}_{k,\,l}^{0\quad\tau}}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ}){\mathop{\times}\limits_{\mathop{\rm BPU}\nolimits(k)}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ}),\;(g,\,p)\mapsto(gp,\,p)$ is a homeomorphism$.\quad\square$ Thus, the action $\varphi$ turns the fibration (5) into a principal groupoid $\mathfrak{G}_{k,\,l}$-bundle. Moreover, it is the universal $\mathfrak{G}_{k,\,l}$-bundle because (as we have already noticed) $\tau_{k,\,l}\colon\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ has contractible fibers. Therefore there is a homotopy equivalence $\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}\simeq\mathop{\rm BPU}\nolimits(k).$ ###### Remark 19. The last result (in particular, that the homotopy type of $\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}$ does not depend on $l$) can be explained using the notion of Morita equivalence for groupoids (see [5]). Take a positive integer $m$ and define $\mathfrak{G}_{k,\,l}-\mathfrak{G}_{k,\,m}$-bimodule $\mathfrak{M}_{k,l;\,k,m}$ as follows. $\mathfrak{M}_{k,l;\,k,m}$ consists of all unital $$-homomorphisms from $k$-subalgebras in $M_{km}(\mathbb{C})$ to $k$-subalgebras in $M_{kl}(\mathbb{C})$. Clearly, $\mathfrak{M}_{k,l;\,k,m}$ is an equivalence bimodule [5]. If we take $m=1$ we obtain the homotopy equivalence $\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}\simeq\mathop{\rm BPU}\nolimits(k)$ directly. ###### Remark 20. It is easy to see that for the $\mathop{\rm SU}\nolimits$-analog of the groupoid $\mathfrak{G}_{k,\,l}$ (see Remark 14) the classifying space is homotopy equivalent to $\mathop{\rm BSU}\nolimits(k)$ (cf. (13)). Note that the groupoid $\mathfrak{G}_{k,\,l}$ itself is (the total space of) a principal $\mathfrak{G}_{k,\,l}$-bundle with the base space $\mathop{\rm Gr}\nolimits_{k,\,l}=\mathfrak{G}_{k,\,l}^{0}.$ This bundle is called unit [8]. A principal groupoid $\mathfrak{G}_{k,\,l}$-bundle ${\rm H}_{k,\,l}(A_{k})\rightarrow X$ (we have already noticed that every principal $\mathfrak{G}_{k,\,l}$-bundle is of this form) is called trivial w.r.t. a map $f\colon X\rightarrow\mathfrak{G}_{k,\,l}^{0}$ if it is the pullback of the unit bundle via this map [8]. In particular, the unit bundle is trivial with respect to the identity map $\mathop{\rm id}\nolimits\colon\mathfrak{G}_{k,\,l}^{0}\rightarrow\mathfrak{G}_{k,\,l}^{0}.$ (Thus, in general, there are non isomorphic trivial bundles over the same base space). Note that a $\mathfrak{G}_{k,\,l}$-bundle ${\rm H}_{k,\,l}(A_{k})\rightarrow X$ is trivial iff it has a section, i.e. there is an embedding (2) (with $n=kl$) iff ${\rm H}_{k,\,l}(A_{k})\rightarrow X$ is a trivial principal groupoid $\mathfrak{G}_{k,\,l}$-bundle. ###### Remark 21. Let us return to the functor $(A_{k},\,\mu)\mapsto A_{k}$ (see Remark 4) corresponding to the map of classifying spaces $\mathop{\rm Gr}\nolimits_{k,\,l}\rightarrow\mathop{\rm BPU}\nolimits(k).$ Now we see that it can be interpreted as the factorization by the action of the groupoid $\mathfrak{G}_{k,\,l}$ (cf. Subsection 3.2 below). ### 2.4. A remark about stabilization Note that maps (23) induce maps of classifying spaces (25) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 45.3575pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&\\\&\\\\}}}\ignorespaces{\hbox{\kern-45.3575pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm H}_{k,\,l}(A_{k}^{univ})\times{\rm H}_{m,\,n}(A_{m}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 0.0pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 0.0pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 45.3575pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 69.3575pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 69.3575pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{{\rm H}_{km,\,ln}(A_{km}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 95.77058pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 95.77058pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern-44.31721pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BPU}\nolimits(k)\times\mathop{\rm BPU}\nolimits(m)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 44.31723pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 71.03673pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 71.03673pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BPU}\nolimits(km)}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ (we should restrict ourself to the case $(km,\,ln)=1$), cf. [12]. In the direct limit we obtain the $H$-space homomorphism (26) $\mathop{\rm Gr}\nolimits\rightarrow\lim\limits_{\longrightarrow\atop{k}}\mathop{\rm BPU}\nolimits(k),$ where $\mathop{\rm Gr}\nolimits:=\lim\limits_{\longrightarrow\atop{(k,\,l)=1}}\mathop{\rm Gr}\nolimits_{k,\,l}$ [12], maps in the direct limits are induced by the tensor product and we use the homotopy equivalences ${\rm H}_{k,\,l}(A_{k}^{univ})\simeq\mathop{\rm Gr}\nolimits_{k,\,l}.$ Since there is an $H$-space isomorphism $\mathop{\rm Gr}\nolimits\cong\mathop{\rm BSU}\nolimits_{\otimes}$ [12], we see that (26) is the composition of the localization map $\mathop{\rm BSU}\nolimits_{\otimes}\rightarrow\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)$ and the natural inclusion $\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)\hookrightarrow\mathop{\rm K}\nolimits(\mathbb{Q}/\mathbb{Z},\,2)\times\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)\simeq\lim\limits_{\longrightarrow\atop{k}}\mathop{\rm BPU}\nolimits(k).$ Consider the abelian group (27) $\mathop{\rm coker}\nolimits\\{[X,\,\mathop{\rm Gr}\nolimits]\rightarrow[X,\,\lim\limits_{\longrightarrow\atop{k}}\mathop{\rm BPU}\nolimits(k)]\\},$ where the homomorphism of the groups of homotopy classes is induced by (26). It admits the following “geometric” description. We call an $M_{k}(\mathbb{C})$-bundle embeddable if there is an embedding $\mu\colon A_{k}\hookrightarrow X\times M_{kl}(\mathbb{C})$ as above for some $l,\,(k,\,l)=1.$ We say that $M_{k}(\mathbb{C})$ and $M_{m}(\mathbb{C})$-bundles $C_{k},\,D_{m}$ over $X$ are equivalent if there are embeddable bundles $A_{l},\,B_{n}$ such that $C_{k}\otimes A_{l}\cong D_{m}\otimes B_{n}.$ The set of such equivalence classes over the given base space $X$ is a group with respect to the operation induced by the tensor product. Clearly, this group is the cokernel (27). In particular, for every even-dimensional sphere $S^{2n}$ it is $\mathbb{Q}/\mathbb{Z}$ (and $0$ for every odd-dimensional one). ###### Remark 22. Since $\mathop{\rm BSU}\nolimits_{\otimes}$ is an infinite loop space [9], this invariant can be interpreted in terms of the coefficient sequence for the corresponding cohomology theory. ## 3\. Some constructions ### 3.1. Partial isomorphisms Let $A_{k}\rightarrow X$ be an $M_{k}(\mathbb{C})$-bundle over $X$ and $\mu\colon A_{k}\hookrightarrow X\times M_{kl}(\mathbb{C})\;((k,\,l)=1)$ a bundle map which is a unital $$-algebra homomorphism on each fiber as above. So every fiber $(A_{k})_{x},\;x\in X$ can be identified with the corresponding $k$-subalgebra $\mu\|_{x}((A_{k})_{x})\subset M_{kl}(\mathbb{C})$ and we have the triple $(A_{k},\,\mu,\,X\times M_{kl}(\mathbb{C})).$ Let $(A_{k}^{\prime},\,\mu^{\prime},\,X\times M_{kl}(\mathbb{C}))$ be another triple of such a kind. Assume that the bundles $A_{k}$ and $A_{k}^{\prime}$ are isomorphic and choose some $$-isomorphism $\vartheta\colon A_{k}\cong A_{k}^{\prime}$. Note that embeddings $\mu,\,\mu^{\prime}$ define the corresponding maps to the matrix Grassmannian $f_{\mu},\,f_{\mu^{\prime}}\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ and, moreover $\vartheta,\,\mu$ and $\mu^{\prime}$ define a map $\nu\colon X\rightarrow\mathfrak{G}_{k,\,l}$ such that $s\circ\nu=f_{\mu},\;t\circ\nu=f_{\mu^{\prime}}$ and $\nu\|_{x}=\mu^{\prime}\circ\vartheta\|_{x}\circ\mu^{-1}\colon\mu((A_{k})_{x})\rightarrow\mu^{\prime}((A_{k}^{\prime})_{x}).$ Conversely, a map $\nu\colon X\rightarrow\mathfrak{G}_{k,\,l}$ gives us some maps $f_{\mu}:=s\circ\nu$ and $f_{\mu^{\prime}}:=t\circ\nu\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ that come from some triples $(A_{k},\,\mu,\,X\times M_{kl}(\mathbb{C})),\;(A_{k}^{\prime},\,\mu^{\prime},\,X\times M_{kl}(\mathbb{C})),$ and an isomorphism $\vartheta\colon A_{k}\cong A_{k}^{\prime}.$ Such a $\nu$ will be called a partial isomorphism from $(A_{k},\,\mu,\,X\times M_{kl}(\mathbb{C}))$ to $(A_{k}^{\prime},\,\mu^{\prime},\,X\times M_{kl}(\mathbb{C}))$ or just a partial automorphism of the trivial bundle $X\times M_{kl}(\mathbb{C})$. Partial isomorphisms that can be lifted to “genuine” automorphisms of the trivial bundle $X\times M_{kl}(\mathbb{C})$ (i.e. to genuine bundle maps $\widetilde{\vartheta}\colon X\times M_{kl}(\mathbb{C})\rightarrow X\times M_{kl}(\mathbb{C})$ such that the diagram $\textstyle{A_{k}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\vartheta}$$\scriptstyle{\mu}$$\textstyle{A_{k}^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mu^{\prime}}$$\textstyle{X\times M_{kl}(\mathbb{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\widetilde{\vartheta}}$$\textstyle{X\times M_{kl}(\mathbb{C})}$ commutes) are just called isomorphisms. ###### Remark 23. An extension of a partial isomorphism $\nu\colon X\rightarrow\mathfrak{G}_{k,\,l}$ to a genuine isomorphism is equivalent to the choice of a lift $\widetilde{\nu}\colon X\rightarrow\widehat{\mathfrak{G}}_{k,\,l}$ of $\nu$ in (24) (to show this one can use the map $\widehat{\vartheta}\colon\widehat{\mathfrak{G}}_{k,\,l}\rightarrow\mathop{\rm PU}\nolimits(kl)$ introduced in Subsection 2.2). Now we claim that there are partial isomorphisms that are not isomorphisms. To show this, take $X=\mathop{\rm Fr}\nolimits_{k,\,l}.$ The map $\nu\colon\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathfrak{G}_{k,\,l}$ is defined as follows. Fix $\alpha\in\mathop{\rm Gr}\nolimits_{k,\,l}$ and consider all $$-isomorphisms from $M_{k,\,\alpha}\subset M_{kl}(\mathbb{C})$ to $M_{k,\,\beta}\subset M_{kl}(\mathbb{C})$, where $\beta$ runs over all $k$-subalgebras in $M_{kl}(\mathbb{C})$. Clearly, this defines a subspace in $\mathfrak{G}_{k,\,l}$ homeomorphic to $\mathop{\rm Fr}\nolimits_{k,\,l}$. In our case $A_{k}\cong A_{k}^{\prime}=\mathop{\rm Fr}\nolimits_{k,\,l}\times M_{k}(\mathbb{C}),$ but $\mu$ and $\mu^{\prime}$ are different. In order to show this, define the $M_{l}(\mathbb{C})$-bundle ${\mathcal{B}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ as the centralizer of the tautological subbundle ${\mathcal{A}}_{k,\,l}\hookrightarrow\mathop{\rm Gr}\nolimits_{k,\,l}\times M_{kl}(\mathbb{C})$ (for more details see e.g. [10]). Clearly, $f_{\mu}\colon\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ is the map to the point $\alpha\in\mathop{\rm Gr}\nolimits_{k,\,l},$ while $f_{\mu^{\prime}}\colon\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ is (the projection of) the principal $\mathop{\rm PU}\nolimits(k)$-bundle $\mathop{\rm PU}\nolimits(k)\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}.$ Clearly, both bundles $A_{k}=f_{\mu}^{}({\mathcal{A}}_{k,\,l})$ and $A_{k}^{\prime}=f_{\mu^{\prime}}^{}({\mathcal{A}}_{k,\,l})$ are trivial, as we have already asserted (note that $A_{k}^{\prime}=f_{\mu^{\prime}}^{}({\mathcal{A}}_{k,\,l})$ is trivial because $f_{\mu^{\prime}}\colon\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ is the frame bundle for ${\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$). The bundle $f_{\mu}^{}({\mathcal{B}}_{k,\,l})$ is also trivial, while $f_{\mu^{\prime}}^{}({\mathcal{B}}_{k,\,l})$ is nontrivial (because it is associated with the principal bundle $\mathop{\rm PU}\nolimits(l)\rightarrow\mathop{\rm PU}\nolimits(kl)\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$). This shows that for chosen $\nu\colon\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathfrak{G}_{k,\,l}\;\vartheta$ can not be extended to an automorphism of $\mathop{\rm Fr}\nolimits_{k,\,l}\times M_{kl}(\mathbb{C})$ (because such an automorphism induces an isomorphism not only between the subbundles $A_{k},\;A_{k}^{\prime}$, but also between their centralizers). In particular, we see that the analog of Noether-Skolem’s theorem is not true for matrix algebras $\Gamma(X,\;X\times M_{kl}(\mathbb{C}))=M_{kl}(C(X))$ over $C(X)$. ### 3.2. An action on fibers of a forgetful functor Consider the forgetful functor given by the assignment $(A_{k},\,\mu,\,X\times M_{kl}(\mathbb{C}))\mapsto A_{k}$ corresponding to the map of representing spaces $\mathop{\rm Gr}\nolimits_{k,\,l}\rightarrow\mathop{\rm BPU}\nolimits(k)$ (whose homotopy fiber is $\mathop{\rm Fr}\nolimits_{k,\,l}$). We claim that our previous construction can be regarded as an action of the groupoid on its fibres. First, let us recall some of the previous results. Applying fiberwisely the functor $\mathop{\rm Hom}\nolimits_{alg}(\ldots,\,M_{kl}(\mathbb{C}))$ to the universal $M_{k}(\mathbb{C})$-bundle $A_{k}^{univ}\rightarrow\mathop{\rm BPU}\nolimits(k)$ we obtain the fibration (28) $\begin{array}[]{c}\vbox{\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 12.10304pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&\\\&\\\\}}}\ignorespaces{\hbox{\kern-12.10304pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 12.10306pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 36.63129pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 36.63129pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 56.44684pt\raise-6.03333pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{}}$}}}}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 56.44684pt\raise-20.42831pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-0.42361pt\hbox{$\scriptstyle{p_{k,\,l}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 56.44684pt\raise-30.35664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{$\textstyle{\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}$}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern-3.0pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 36.10304pt\raise-40.85664pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathop{\rm BPU}\nolimits(k)}$}}}}}}}\ignorespaces\ignorespaces}}}}}\end{array}$ with fiber $\mathop{\rm Fr}\nolimits_{k,\,l}:=\mathop{\rm Hom}\nolimits_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))$. We have the map $\tau_{k,\,l}\colon\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l},\;h\mapsto h((A_{k}^{univ})_{x})\subset M_{kl}(\mathbb{C}),$ where $x\in\mathop{\rm BPU}\nolimits(k)$ and $h\in p_{k,\,l}^{-1}(x)$, which is a fibration with contractible fibres, i.e. a homotopy equivalence. Moreover, there is the free and proper action $\varphi\colon\mathfrak{G}_{k,\,l}{\mathop{\times}\limits_{{}^{s}\;\mathop{\rm Gr}\nolimits_{k,\,l}^{\;\tau}}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})$ which turns the fibration (28) into the universal principal groupoid $\mathfrak{G}_{k,\,l}$-bundle. We have also shown that for a map $\bar{f}\colon X\rightarrow\mathop{\rm BPU}\nolimits(k)$ the choice of its lift $\widetilde{f}\colon X\rightarrow\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})$ (if it exists) is equivalent to the choice of an embedding $\mu\colon\bar{f}^{}(A_{k}^{univ})\rightarrow X\times M_{kl}(\mathbb{C}).$ Such a lift we denoted by $\widetilde{f}_{\mu}.$ Given $\nu\colon X\rightarrow\mathfrak{G}_{k,\,l}$ such that $s\circ\nu=\tau_{k,\,l}\circ\widetilde{f}_{\mu}=f_{\mu},\,t\circ\nu=f_{\mu^{\prime}}\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ we define the composite map $\widetilde{f}_{\mu^{\prime}}\colon$ $X\stackrel{{\scriptstyle diag}}{{\rightarrow}}X\times X\stackrel{{\scriptstyle\nu\times\widetilde{f}_{\mu}}}{{\longrightarrow}}\mathfrak{G}_{k,\,l}{\mathop{\times}\limits_{{}^{s}\;\mathop{\rm Gr}\nolimits_{k,\,l}^{\;\tau}}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\stackrel{{\scriptstyle\varphi}}{{\rightarrow}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})$ which is (in general) another lift of $\bar{f}$ ($p_{k,\,l}\circ\widetilde{f}_{\mu}=\bar{f}=p_{k,\,l}\circ\widetilde{f}_{\mu^{\prime}}$), i.e. it corresponds to another (homotopy nonequivalent in general) embedding $\mu^{\prime}\colon\bar{f}^{}(A_{k}^{univ})\rightarrow X\times M_{kl}(\mathbb{C}),\quad\hbox{i.e.}\;f_{\mu^{\prime}}=\tau_{k,\,l}\circ\widetilde{f}_{\mu^{\prime}}\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}.$ Clearly, this action is transitive on homotopy classes of such embeddings. ###### Remark 24. It seems natural to consider the action $\varphi$ modulo the action $\widehat{\varphi}\colon\widehat{\mathfrak{G}}_{k,\,l}{\mathop{\times}\limits_{{}^{s}\;\mathop{\rm Gr}\nolimits_{k,\,l}^{\;\tau}}}\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})\rightarrow\mathop{\rm H}\nolimits_{k,\,l}(A_{k}^{univ})$ of the groupoid $\widehat{\mathfrak{G}}_{k,\,l}$ and the corresponding equivalence relation on lifts $\widetilde{f}$ (cf. Remark 23). ### 3.3. A remark about groupoid cocycles In this subsection we sketch an approach to groupoid bundles via local trivializing data and 1-cocycles. The reader can find the general results in [8], but we hope that our groupoids provide an instructive illustration of the general theory. In Subsection 2.3 we have already seen that a trivial $\mathfrak{G}_{k,\,l}$-bundle ${\rm H}_{k,\,l}(A_{k})\rightarrow X$ is the pullback of the unit bundle ${\rm H}_{k,\,l}({\mathcal{A}}_{k,\,l})\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ via some map $f\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$. Moreover, such a map $f$ is nothing but a trivialization of ${\rm H}_{k,\,l}(A_{k})\rightarrow X$. Such a trivialization can also be thought of as a triple $(A_{k},\,\mu,\,X\times M_{kl}(\mathbb{C}))$ (see Subsection 3.1), where $\mu\colon A_{k}\rightarrow X\times M_{kl}(\mathbb{C})$ is a fiberwise embedding as above, because $f=f_{\mu}$ is its classifying map. For a topological group $G$ the group of automorphisms of a trivial $G$-bundle over $X$ can be identified with the group of continuous maps $X\rightarrow G$ which take one trivialization to another. The analogous maps $\nu\colon X\rightarrow\mathfrak{G}_{k,\,l}$ to the groupoid $\mathfrak{G}_{k,\,l}$ were called partial isomorphisms in Subsection 3.1. Recall that such $\nu$ defines two compositions $s\circ\nu$ and $t\circ\nu\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ which give rise to some triples as above and therefore to some trivializations. Let $X$ be a compact manifold, ${\mathcal{U}}:=\\{U_{\alpha}\\}_{\alpha\in A}$ its open covering. A $\mathfrak{G}_{k,\,l}$ 1-cocycle can be defined as a groupoid homomorphism (more precisely, as a functor) from the Čech groupoid to $\mathfrak{G}_{k,\,l}$. So we get the following unfolded form of this definition. ###### Definition 25. A groupoid $\mathfrak{G}_{k,\,l}$ 1-cocycle $\\{g_{\alpha\beta}\\}_{\alpha,\,\beta\in A}$ is a collection of continuous maps $g_{\alpha\beta}\colon U_{\alpha}\cap U_{\beta}\rightarrow\mathfrak{G}_{k,\,l}$ such that * 1) $g_{\alpha\beta}$ and $g_{\beta\gamma}$ are composable on $U_{\alpha}\cap U_{\beta}\cap U_{\gamma},$ i.e. $\forall x\in U_{\alpha}\cap U_{\beta}\cap U_{\gamma}\;t(g_{\alpha\beta}(x))=s(g_{\beta\gamma}(x)),$ where $s$ and $t$ are the source and target maps for $\mathfrak{G}_{k,\,l};$ * 2) $g_{\alpha\beta}g_{\beta\gamma}=g_{\alpha\gamma}$ on $U_{\alpha}\cap U_{\beta}\cap U_{\gamma}$ (in particular, $g_{\alpha\alpha}\in e,\;g_{\beta\alpha}=i(g_{\alpha\beta})$, where $e$ and $i$ are the identity and the inversion for the groupoid $\mathfrak{G}_{k,\,l},$ see Subsection 2.1). In the same way one can define a groupoid $\widehat{\mathfrak{G}}_{k,\,l}$ 1-cocycle $\\{\widehat{g}_{\alpha\beta}\\}_{\alpha,\,\beta\in A}$. ###### Remark 26. Note that a trivial bundle over $X$ of our kind with a trivialization $(A_{k},\,\mu,\,X\times M_{kl}(\mathbb{C}))$ corresponds to the trivial $\mathfrak{G}_{k,\,l}$ 1-cocycle. Indeed, two maps $f_{\mu},\,f_{\mu^{\prime}}=f_{\mu}\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ give rise to the identity partial isomorphism (see Subsection 3.1) $\nu\colon X\rightarrow\mathfrak{G}_{k,\,l},\;\nu\|_{x}=\mathop{\rm id}\nolimits_{\mu((A_{k})_{x})}$ which is the trivial groupoid $\mathfrak{G}_{k,\,l}$ 1-cocycle as claimed. Now the gluing of a groupoid bundle using local data can be described as follows. So we start with an open covering ${\mathcal{U}}=\\{U_{\alpha}\\}_{\alpha\in A}$ and trivial groupoid bundles $(A_{k,\,\alpha},\,\mu_{\alpha},\,U_{\alpha}\times M_{kl}(\mathbb{C}))$ over $U_{\alpha},\;\alpha\in A.$ Suppose we are given a groupoid $\mathfrak{G}_{k,\,l}$ 1-cocycle $\\{g_{\alpha\beta}\\}_{\alpha,\,\beta\in A}$ (over the same open covering ${\mathcal{U}}$) such that $s(g_{\alpha\beta})\equiv f_{\mu_{\alpha}}\|_{U_{\alpha}\cap U_{\beta}}$ and $t(g_{\alpha\beta})\equiv f_{\mu_{\beta}}\|_{U_{\alpha}\cap U_{\beta}}\;\forall\alpha,\,\beta\in A.$ (In our previous notation it is natural to denote it by $\\{\nu_{\alpha\beta}\\}$). The groupoid $\mathfrak{G}_{k,\,l}$ 1-cocycle $\\{g_{\alpha\beta}\\}_{\alpha,\,\beta\in A}$ defines partial isomorphisms (see Subsection 3.1) from $(A_{k,\,\alpha},\,\mu_{\alpha},\,U_{\alpha}\times M_{kl}(\mathbb{C}))\|_{U_{\alpha}\cap U_{\beta}}$ to $(A_{k,\,\beta},\,\mu_{\beta},\,U_{\beta}\times M_{kl}(\mathbb{C}))\|_{U_{\alpha}\cap U_{\beta}}$ for all $\alpha,\,\beta\in A$ which agree on triple intersections. We can summarize the previous results as follows. Put $Y:=\coprod\limits_{\alpha}U_{\alpha},\quad Y^{[2]}:=Y{\mathop{\times}\limits_{X}}Y=\coprod\limits_{\alpha,\,\beta}U_{\alpha}\cap U_{\beta}.$ For every pair $\alpha,\,\beta\in A$ a 1-cocycle $\\{g_{\alpha\beta}\\}_{\alpha,\,\beta\in A}$ defines maps $\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\mathfrak{G}_{k,\,l}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{s}$$\scriptstyle{t}$$\textstyle{\mathop{\rm Gr}\nolimits_{k,\,l}}$$\textstyle{\mathop{\rm Gr}\nolimits_{k,\,l}}$$\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces U_{\alpha}\cap U_{\beta}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{i_{\alpha}}$$\scriptstyle{i_{\beta}}$$\scriptstyle{g_{\alpha\beta}}$$\scriptstyle{=\nu_{\alpha\beta}}$$\textstyle{U_{\alpha}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{f_{\alpha}}$$\textstyle{U_{\beta}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{f_{\beta}}$ satisfying the cocycle conditions on triple intersections. The idea is to regard the map $f_{\alpha}\colon U_{\alpha}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ (corresponding to the “identity” $g_{\alpha\alpha}$) as a local trivialization and $\nu_{\alpha\beta}\colon U_{\alpha}\cap U_{\beta}\rightarrow\mathfrak{G}_{k,\,l}$ for $\alpha\neq\beta$ as a gluing of different trivializations over the double intersection $U_{\alpha}\cap U_{\beta}$. Thus, we have maps $f\colon Y\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l},\quad g\colon Y{\mathop{\times}\limits_{X}}Y\rightarrow\mathfrak{G}_{k,\,l}$ such that $s\circ g=f\circ\pi_{1}\|_{Y{\mathop{\times}\limits_{X}}Y},\>t\circ g=f\circ\pi_{2}\|_{Y{\mathop{\times}\limits_{X}}Y},$ where $\pi_{i}\colon Y{\mathop{\times}\limits_{X}}Y\rightarrow Y$ are the projections onto $i$th factor. There is a natural equivalence relation on the set of groupoid 1-cocycles generalizing the equivalence relation on group 1-cocycles. As in the case of usual bundles constructed by means of group $G$ 1-cocycles, we have: * 1) equivalence of 1-cocycles over the same open covering $\mathcal{U}$; * 2) equivalence of 1-cocycles related to the refinement of the open covering. The case 1) concerns to the different choices of trivializations over open subsets $U_{\alpha}.$ We have already noticed that such a trivialization is actually a map $f_{\mu_{\alpha}}\colon U_{\alpha}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ and two such trivializations $f_{\mu_{\alpha}},\>f_{\mu_{\alpha}^{\prime}}$ are related by the map $\nu_{\alpha}\colon U_{\alpha}\rightarrow\mathfrak{G}_{k,\,l}$ (such that $s\circ\nu_{\alpha}=f_{\mu_{\alpha}},\>t\circ\nu_{\alpha}=f_{\mu_{\alpha}^{\prime}}$). ###### Remark 27. Note that using groupoid $\mathfrak{G}_{k,\,l}$ $1$-cocycle one can glue some global $M_{k}(\mathbb{C})$-bundle $A_{k}\rightarrow X$ such that $A_{k}\|{U_{\alpha}}=A_{k,\,\alpha}.$ It agrees with the proved above homotopy equivalence $\mathop{\rm B}\nolimits\mathfrak{G}_{k,\,l}\simeq\mathop{\rm BPU}\nolimits(k)$. In other words, the groupoid bundle glued by the $1$-cocycle is ${\rm H}_{k,\,l}(A_{k})\rightarrow X$. Note that local embeddings $\mu_{\alpha},\>\alpha\in A$ do not give rise to some global object (like local trivializations in the case of “usual” bundles). The case of the groupoid $\widehat{\mathfrak{G}}_{k,\,l}$ can be described in the similar way. In this case a map $f\colon X\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ (a “trivialization”) can be regarded as the product bundle $X\times M_{kl}(\mathbb{C})$ together with a chosen decomposition into the tensor product $A_{k}\otimes B_{l}$ of its $M_{k}(\mathbb{C})$ and $M_{l}(\mathbb{C})$-subbundles $A_{k}\rightarrow X$ and $B_{l}\rightarrow X$ respectively. Note that using a $\widehat{\mathfrak{G}}_{k,\,l}$-cocycle $\\{\widehat{g}_{\alpha\beta}\\}_{\alpha,\,\beta\in A}$ as above one can glue some $M_{k}(\mathbb{C})$ and $M_{l}(\mathbb{C})$-bundles $A_{k},\,B_{l}$ over $X$. It relates to the existence of a homotopy equivalence $\mathop{\rm B}\nolimits\widehat{\mathfrak{G}}_{k,\,l}\simeq\mathop{\rm BPU}\nolimits(k)\times\mathop{\rm BPU}\nolimits(l)$ (cf. Remarks 13 and 15). The relation between $\widehat{\mathfrak{G}}_{k,\,l}$ and $\mathfrak{G}_{k,\,l}$-groupoid bundles follows from the exact sequence (24). ## 4\. On the $K$-theory automorphisms ### 4.1. The case of line bundles First, consider the case of line bundles. The classifying space of $K$-theory can be taken to be $\mathop{\rm Fred}\nolimits({\mathcal{H}}),$ the space of Fredholm operators on Hilbert space ${\mathcal{H}}$. It is known [2] that for a compact space $X$ the action of the Picard group $Pic(X)$ on $K(X)$ is induced by the conjugation action $\gamma\colon\mathop{\rm PU}\nolimits({\mathcal{H}})\times\mathop{\rm Fred}\nolimits({\mathcal{H}})\rightarrow\mathop{\rm Fred}\nolimits({\mathcal{H}}),\;\gamma(g,\,T)=gTg^{-1}$ of $\mathop{\rm PU}\nolimits({\mathcal{H}})$ on $\mathop{\rm Fred}\nolimits({\mathcal{H}}).$ The more precise statement is given in the following theorem (recall that $\mathop{\rm PU}\nolimits({\mathcal{H}})\simeq\mathbb{C}P^{\infty}\simeq\mathop{\rm K}\nolimits(\mathbb{Z},\,2)$). ###### Theorem 28. If $f_{\xi}\colon X\rightarrow\mathop{\rm Fred}\nolimits({\mathcal{H}})$ and $\varphi_{\zeta}\colon X\rightarrow\mathop{\rm PU}\nolimits({\mathcal{H}})$ represent $\xi\in K(X)$ and $\zeta\in Pic(X)$ respectively, then the composite map (29) $X\stackrel{{\scriptstyle\mathop{\rm diag}\nolimits}}{{\longrightarrow}}X\times X\stackrel{{\scriptstyle\varphi_{\zeta}\times f_{\xi}}}{{\longrightarrow}}\mathop{\rm PU}\nolimits({\mathcal{H}})\times\mathop{\rm Fred}\nolimits({\mathcal{H}})\stackrel{{\scriptstyle\gamma}}{{\rightarrow}}\mathop{\rm Fred}\nolimits({\mathcal{H}})$ represents $\zeta\otimes\xi\in K(X)$. Proof see [2].$\quad\square$ If we want to restrict ourself to the action of line bundles corresponding to elements of finite order in the group $Pic(X)$ we have to consider the subgroups $\mathop{\rm PU}\nolimits(k)\subset\mathop{\rm PU}\nolimits({\mathcal{H}})$. Let us describe this inclusion. Let ${\mathcal{B}}({\mathcal{H}})$ be the algebra of bounded operators on the separable Hilbert space ${\mathcal{H}}$, $M_{k}({\mathcal{B}}({\mathcal{H}})):=M_{k}(\mathbb{C}){\mathop{\otimes}\limits_{\mathbb{C}}}{\mathcal{B}}({\mathcal{H}})$ the matrix algebra over ${\mathcal{B}}({\mathcal{H}})$ (of course, it is isomorphic to ${\mathcal{B}}({\mathcal{H}})$). Let $\mathop{\rm U}\nolimits_{k}({\mathcal{H}})\subset M_{k}({\mathcal{B}}({\mathcal{H}}))$ be the corresponding unitary group (isomorphic to $\mathop{\rm U}\nolimits({\mathcal{H}})$). It acts on $M_{k}({\mathcal{B}}({\mathcal{H}}))$ by conjugations (which are $$-algebra automorphisms). Moreover, the kernel of the action is the center, i.e. the subgroup of scalar matrices $\mathop{\rm U}\nolimits(1).$ The corresponding quotient group we denote by $\mathop{\rm PU}\nolimits_{k}({\mathcal{H}})$ (it is isomorphic to $\mathop{\rm PU}\nolimits({\mathcal{H}})$). $M_{k}(\mathbb{C})\otimes\mathop{\rm Id}\nolimits_{{\mathcal{B}}({\mathcal{H}})}$ is a $k$-subalgebra (i.e. a unital $$-subalgebra isomorphic to $M_{k}(\mathbb{C})$) in $M_{k}({\mathcal{B}}({\mathcal{H}}))$. Then $\mathop{\rm PU}\nolimits(k)\subset\mathop{\rm PU}\nolimits_{k}({\mathcal{H}})$ is the subgroup of automorphisms of the mentioned $k$-subalgebra. Thus we have defined the injective group homomorphism $\Psi_{k}\colon\mathop{\rm PU}\nolimits(k)\hookrightarrow\mathop{\rm PU}\nolimits_{k}({\mathcal{H}}),$ induced by the injective homomorphism $\mathop{\rm U}\nolimits(k)\hookrightarrow\mathop{\rm U}\nolimits_{k}({\mathcal{H}}),\>g\mapsto g\otimes\mathop{\rm Id}\nolimits_{{\mathcal{B}}({\mathcal{H}})}$. ###### Proposition 29. For a line bundle $\zeta\rightarrow X$ satisfying the condition (30) $\zeta^{\oplus k}\cong X\times\mathbb{C}^{k}$ the classifying map $\varphi_{\zeta}\colon X\rightarrow\mathop{\rm PU}\nolimits_{k}({\mathcal{H}})\cong\mathop{\rm PU}\nolimits({\mathcal{H}})$ can be lifted to a map $\widetilde{\varphi}_{\zeta}\colon X\rightarrow\mathop{\rm PU}\nolimits(k)$ such that $\Psi_{k}\circ\widetilde{\varphi}_{\zeta}\simeq\varphi_{\zeta}$. Proof. Consider the exact sequence of groups (31) $1\rightarrow\mathop{\rm U}\nolimits(1)\rightarrow\mathop{\rm U}\nolimits(k)\stackrel{{\scriptstyle\chi_{k}}}{{\rightarrow}}\mathop{\rm PU}\nolimits(k)\rightarrow 1$ and the fibration (32) $\mathop{\rm PU}\nolimits(k)\stackrel{{\scriptstyle\psi_{k}}}{{\rightarrow}}\mathop{\rm BU}\nolimits(1)\stackrel{{\scriptstyle\omega_{k}}}{{\rightarrow}}\mathop{\rm BU}\nolimits(k)$ obtained by its extension to the right. In particular, $\psi_{k}\colon\mathop{\rm PU}\nolimits(k)\rightarrow\mathop{\rm BU}\nolimits(1)\simeq\mathbb{C}P^{\infty}$ is a classifying map for the $\mathop{\rm U}\nolimits(1)$-bundle (31). Clearly, $\Psi_{k}\simeq\psi_{k}$ under the homotopy equivalence $\mathop{\rm PU}\nolimits({\mathcal{H}})\simeq\mathop{\rm BU}\nolimits(1)$. (Indeed, it follows from the map of $\mathop{\rm U}\nolimits(1)$-bundles $\textstyle{\mathop{\rm U}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm U}\nolimits_{k}({\mathcal{H}})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\simeq\mathop{\rm EU}\nolimits(1)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm PU}\nolimits(k)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm PU}\nolimits_{k}({\mathcal{H}})\simeq\mathop{\rm BU}\nolimits(1)).}$ It is easy to see from fibration (32) that for an arbitrary line bundle $\zeta\rightarrow X$ satisfying the condition (30) the classifying map $\varphi_{\zeta}\colon X\rightarrow\mathbb{C}P^{\infty}$ can be lifted to $\widetilde{\varphi}_{\zeta}\colon X\rightarrow\mathop{\rm PU}\nolimits(k)$ (because for such a bundle we have $\omega_{k}\circ\varphi_{\zeta}\simeq$ and $\omega_{k}$ is induced by $\zeta\mapsto\zeta^{\oplus k}$).$\quad\square$ The choice of a lift $\widetilde{\varphi}_{\zeta}$ corresponds to the choice of a trivialization (30): two lifts differ by a map $X\rightarrow\mathop{\rm U}\nolimits(k)$. Therefore the subgroup in $Pic(X)$ formed by line bundles satisfying condition (30) is isomorphic to $\mathop{\rm im}\nolimits\\{\psi_{k}\colon[X,\,\mathop{\rm PU}\nolimits(k)]\rightarrow[X,\,\mathbb{C}P^{\infty}]\\}=[X,\,\mathop{\rm PU}\nolimits(k)]/(\chi_{k}[X,\,\mathop{\rm U}\nolimits(k)])$ (cf. the definition of the “finite” Brauer group). ### 4.2. The general case In [1] M. Atiyah and G. Segal wrote: “The group $\mathop{\rm Fred}\nolimits_{1}$ is a product $\mathop{\rm Fred}\nolimits_{1}\simeq\mathbb{P}_{\mathbb{C}}^{\infty}\times{\rm SFred}_{1},$ where ${\rm SFred}_{1}$ is the fibre of the determinant map ${\rm Fred}_{1}\cong BU\rightarrow B\mathbb{T}\cong\mathbb{P}_{\mathbb{C}}^{\infty},$ and the twistings of this paper are those coming from $(\pm 1)\times\mathbb{P}_{\mathbb{C}}^{\infty}$. We do not know any equally geometrical approach to the more general ones.” In what follows we are going to describe the action of the group of (isomorphism classes of) $\mathop{\rm SU}\nolimits$-bundles of finite order on $K(X)$. It corresponds to the torsion subgroup in ${\rm SFred}_{1}$ (our notation differs from the one in [1]). We hope that this construction would provide a geometric approach to more general twistings in $K$-theory. As we have seen in the previous subsection, the group $\mathop{\rm PU}\nolimits({\mathcal{H}})$ from one hand acts on the representing space $\mathop{\rm Fred}\nolimits({\mathcal{H}})$ of the $K$-theory and from the other hand it is the base space of the universal line bundle. These two facts lead to the result that the action of $\mathop{\rm PU}\nolimits({\mathcal{H}})$ on $K(X)$ corresponds to the tensor product by elements of the Picard group $Pic(X)$ (i.e. by line bundles). This action can be restricted to the subgroups $\mathop{\rm PU}\nolimits(k)\subset\mathop{\rm PU}\nolimits({\mathcal{H}})$ which classify the elements of finite order $k,\,k\in\mathbb{N}.$ In what follows the role of subgroups $\mathop{\rm PU}\nolimits(k)$ will play the spaces $\mathop{\rm Fr}\nolimits_{k,\,l}$. From one hand, they in some sense “act” on the $K$-theory (more precisely, we have the action of their direct limit which is an $H$-space with respect to the natural operation), from the other hand they are bases of some $l$-dimensional bundles whose classes have order $k$. We shall show that the “action” of $\mathop{\rm Fr}\nolimits_{k,\,l}$ on $K(X)$ corresponds to the tensor product by those $l$-dimensional bundles (see Theorem 33). ###### Proposition 30. A map $X\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ is the same thing as an embedding (33) $\mu\colon X\times M_{k}(\mathbb{C})\hookrightarrow X\times M_{kl}(\mathbb{C})$ whose restriction to every fiber is a unital $$-algebra homomorphism. Proof follows directly from the bijection $\mathop{\rm Mor}\nolimits(X\times M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))\rightarrow\mathop{\rm Mor}\nolimits(X,\,\mathop{\rm Mor}\nolimits(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))).\quad\square$ For the $\mathop{\rm PU}\nolimits(k)$-bundle projection $\pi_{k,\,l}\colon\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}$ ($\pi_{k,\,l}=f_{\mu^{\prime}}$ in the notation of Subsection 3.1) the pull-back $\pi_{k,\,l}^{}({\mathcal{A}}_{k,\,l})$ has the canonical trivialization (because $\pi_{k,\,l}$ is the frame bundle for ${\mathcal{A}}_{k,\,l}$). In general $\mu$ in (33) is a nontrivial embedding, i.e. not equivalent to the choice of a constant $k$-subalgebra in $X\times M_{kl}(\mathbb{C})$ (equivalently, the homotopy class of $X\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ is nontrivial). In particular, the subbundle of centralizers $B_{l}$ for $\mu(X\times M_{k}(\mathbb{C}))\subset X\times M_{kl}(\mathbb{C})$ can be nontrivial as an $M_{l}(\mathbb{C})$-bundle. The fibration (34) $\mathop{\rm PU}\nolimits(l)\stackrel{{\scriptstyle E_{k}\otimes\ldots}}{{\longrightarrow}}\mathop{\rm PU}\nolimits(kl)\stackrel{{\scriptstyle\chi_{k}^{\prime}}}{{\longrightarrow}}\mathop{\rm Fr}\nolimits_{k,\,l}$ (cf. (3)) can be extended to the right (35) $\mathop{\rm Fr}\nolimits_{k,\,l}\stackrel{{\scriptstyle\psi_{k}^{\prime}}}{{\longrightarrow}}\mathop{\rm BPU}\nolimits(l)\stackrel{{\scriptstyle\omega_{k}^{\prime}}}{{\longrightarrow}}\mathop{\rm BPU}\nolimits(kl),$ where $\psi_{k}^{\prime}$ is the classifying map for the $M_{l}(\mathbb{C})$-bundle $\widetilde{\mathcal{B}}_{k,\,l}:=\pi_{k,\,l}^{}({\mathcal{B}}_{k,\,l})\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ (see the end of Subsection 3.1) which is associated with principal $\mathop{\rm PU}\nolimits(l)$-bundle (34). ###### Proposition 31. (Cf. Proposition 29) For $M_{l}(\mathbb{C})$-bundle $B_{l}\rightarrow X$ such that (36) $[M_{k}]\otimes B_{l}\cong X\times M_{kl}(\mathbb{C})$ (cf. (30)), where $[M_{k}]$ is the trivial $M_{k}(\mathbb{C})$-bundle over $X$, a classifying map $\varphi_{B_{l}}\colon X\rightarrow\mathop{\rm BPU}\nolimits(l)$ can be lifted to a map $\widetilde{\varphi}_{B_{l}}\colon X\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ (i.e. $\psi_{k}^{\prime}\circ\widetilde{\varphi}_{B_{l}}=\varphi_{B_{l}}$ or $B_{l}=\widetilde{\varphi}_{B_{l}}^{}(\widetilde{\mathcal{B}}_{k,\,l})$). Proof follows from fibration (35).$\quad\square$ Moreover, the choice of such a lift corresponds to the choice of trivialization (36) and we return to the interpretation of a map $X\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ given in Proposition 30. We stress that a map $X\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ is not just an $M_{l}(\mathbb{C})$-bundle but an $M_{l}(\mathbb{C})$-bundle together with a particular choice of trivialization (36). It follows from our previous results that the bundle $B_{l}\rightarrow X$ has the form $\mathop{\rm End}\nolimits(\eta_{l})$ for some (unique up to isomorphism) $\mathbb{C}^{l}$-bundle $\eta_{l}\rightarrow X$ with the structure group $\mathop{\rm SU}\nolimits(l)$ (that’s because we assumed that $(k,\,l)=1$). Let $\widetilde{\zeta}\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ be the line bundle associated with universal covering (12) and $\zeta^{\prime}\rightarrow X$ its pullback via $\widetilde{\varphi}_{B_{l}}$. Put $\eta_{l}^{\prime}=\eta_{l}\otimes\zeta^{\prime}.$ Let $\mathop{\rm Fred}\nolimits_{n}({\mathcal{H}})$ be the space of Fredholm operators in $M_{n}({\mathcal{B}}({\mathcal{H}}))$. Clearly, $\mathop{\rm Fred}\nolimits_{n}({\mathcal{H}})\cong\mathop{\rm Fred}\nolimits({\mathcal{H}})$. The canonical evaluation map (37) $\mathop{\rm Fr}\nolimits_{k,\,l}\times M_{k}(\mathbb{C})\rightarrow M_{kl}(\mathbb{C}),\quad(h,\,T)\mapsto h(T)$ (recall that $\mathop{\rm Fr}\nolimits_{k,\,l}=\mathop{\rm Hom}\nolimits_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C}))$) induces the map (38) $\gamma_{k,\,l}^{\prime}\colon\mathop{\rm Fr}\nolimits_{k,\,l}\times\mathop{\rm Fred}\nolimits_{k}({\mathcal{H}})\rightarrow\mathop{\rm Fred}\nolimits_{kl}({\mathcal{H}}).$ ###### Remark 32. Note that map (37) can be decomposed into the composition $\mathop{\rm Fr}\nolimits_{k,\,l}\times M_{k}(\mathbb{C})\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}{\mathop{\times}\limits_{\mathop{\rm PU}\nolimits(k)}}M_{k}(\mathbb{C})={\mathcal{A}}_{k,\,l}\rightarrow M_{kl}(\mathbb{C}),$ where the last map is the tautological embedding $\mu\colon{\mathcal{A}}_{k,\,l}\rightarrow\mathop{\rm Gr}\nolimits_{k,\,l}\times M_{kl}(\mathbb{C})$ followed by the projection onto the second factor. Now suppose $f_{\xi}\colon X\rightarrow\mathop{\rm Fred}\nolimits_{k}({\mathcal{H}})$ represents some element $\xi\in K(X).$ ###### Theorem 33. (Cf. Theorem 28). In the above notation the composite map (cf. (29)) $X\stackrel{{\scriptstyle\mathop{\rm diag}\nolimits}}{{\longrightarrow}}X\times X\stackrel{{\scriptstyle\widetilde{\varphi}_{B_{l}}\times f_{\xi}}}{{\longrightarrow}}\mathop{\rm Fr}\nolimits_{k,\,l}\times\mathop{\rm Fred}\nolimits_{k}({\mathcal{H}})\stackrel{{\scriptstyle\gamma_{k,\,l}^{\prime}}}{{\longrightarrow}}\mathop{\rm Fred}\nolimits_{kl}({\mathcal{H}})$ represents $\eta_{l}^{\prime}\otimes\xi\in K(X).$ Proof (cf. [2], Proposition 2.1). By assumption the element $\xi\in K(X)$ is represented by a family $F=\\{F_{x}\\}$ of Fredholm operators in a Hilbert space ${\mathcal{H}}^{k}$. Then $\eta_{l}^{\prime}\otimes\xi\in K(X)$ is represented by the family $\\{\mathop{\rm Id}\nolimits_{(B_{l})_{x}}\otimes\,F_{x}\\}$ of Fredholm operators in the Hilbert bundle $\eta_{l}^{\prime}\otimes({\mathcal{H}}^{k})$ (recall that $\mathop{\rm End}\nolimits(\eta_{l})=B_{l}\,\Rightarrow\,\mathop{\rm End}\nolimits(\eta_{l}^{\prime})=B_{l}$). A trivialization $\eta_{l}^{\prime}\otimes({\mathcal{H}}^{k})\cong{\mathcal{H}}^{kl}$ is the same thing as a lift $\widetilde{\varphi}_{B_{l}}\colon X\rightarrow\mathop{\rm Fr}\nolimits_{k,\,l}$ of the classifying map $\varphi_{B_{l}}\colon X\rightarrow\mathop{\rm BPU}\nolimits(l)$ for $B_{l}$ (see (35)).$\quad\square$ ###### Remark 34. In order to separate the “$\mathop{\rm SU}\nolimits$”-part of the “action” $\gamma_{k,\,l}^{\prime}$ from its “line” part, one can consider the analogous “action” of the space $\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}$ in place of $\mathop{\rm Fr}\nolimits_{k,\,l}$. Then one would have the representing map for $\eta_{l}\otimes\xi\in K(X)$ in the statement of Theorem 33. The commutative diagram $\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\times\mathop{\rm Fr}\nolimits_{k,\,l}\times M_{k^{2}}(\mathbb{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm Fr}\nolimits_{k^{2},\,l^{2}}\times M_{k^{2}}(\mathbb{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{(h_{2},\,h_{1};\,x_{1}\otimes x_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{(h_{2}\otimes h_{1};\,x_{1}\otimes x_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathop{\rm Fr}\nolimits_{k,\,l}\times M_{k^{2}l}(\mathbb{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{M_{k^{2}l^{2}}(\mathbb{C})}$$\textstyle{(h_{2};\,h_{1}(x_{1})\otimes x_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{(h_{1}(x_{1})\otimes h_{2}(x_{2}))}$ (where $h_{i}\in\mathop{\rm Fr}\nolimits_{k,\,l}=\mathop{\rm Hom}\nolimits_{alg}(M_{k}(\mathbb{C}),\,M_{kl}(\mathbb{C})),\>x_{i}\in M_{k}(\mathbb{C}),\>M_{k^{2}}(\mathbb{C})=M_{k}(\mathbb{C})\otimes M_{k}(\mathbb{C})$) gives rise to the “associativity” condition for the “action” $\gamma_{k,\,l}^{\prime}$. Note that the map $\mathop{\rm Fr}\nolimits_{k,\,l}\times\mathop{\rm Fr}\nolimits_{k,\,l}\rightarrow\mathop{\rm Fr}\nolimits_{k^{2},\,l^{2}}$ corresponds to the tensor product of corresponding $M_{l}(\mathbb{C})$-bundles. In fact, the maps $\mathop{\rm Fr}\nolimits_{k^{m},\,l^{m}}\times\mathop{\rm Fr}\nolimits_{k^{n},\,l^{n}}\rightarrow\mathop{\rm Fr}\nolimits_{k^{m+n},\,l^{m+n}}$ (induced by the tensor product of matrix algebras) define the structure of an $H$-space on $\mathop{\rm Fr}\nolimits_{k^{\infty}}:=\lim\limits_{\longrightarrow\atop{n}}\mathop{\rm Fr}\nolimits_{k^{n},\,l^{n}}$. Moreover, $\mathop{\rm Fr}\nolimits_{k^{\infty}}$ is an infinite loop space, because it is the fiber of the localization $\mathop{\rm BU}\nolimits_{\otimes}\rightarrow\mathop{\rm BU}\nolimits_{\otimes}[\frac{1}{k}]$ (and $\mathop{\rm BU}\nolimits_{\otimes}$ is an infinite loop space according [9]). Note that the “action” (38) is not invertible because we take the tensor product of $K(X)$ by some $l$-dimensional bundle. Therefore it makes sense to consider the localization $\mathop{\rm Fred}\nolimits({\mathcal{H}})_{(l)}:=\lim\limits_{\longrightarrow\atop{n}}\mathop{\rm Fred}\nolimits_{l^{n}}({\mathcal{H}}),$ where the direct limit is taken over the maps induced by the tensor product, so $l$ becomes invertible and the index takes values in $\mathbb{Z}[\frac{1}{l}].$ (In fact, our construction is independent of the choice of $l,\,(k,\,l)=1$, so we can consider a pair of such numbers in order to avoid the localization.) The idea is to associate a $\mathop{\rm Fred}\nolimits({\mathcal{H}})_{(l)}$-bundle with the universal $\mathop{\rm Fr}\nolimits_{k^{\infty},\,l^{\infty}}$-bundle $\mathop{\rm Fr}\nolimits_{k^{\infty},\,l^{\infty}}\rightarrow{\rm EFr}_{k^{\infty},\,l^{\infty}}\rightarrow{\rm BFr}_{k^{\infty},\,l^{\infty}}$ using the action $\lim\limits_{\longrightarrow\atop{n}}\gamma^{\prime}_{k^{n},\,l^{n}}\colon\mathop{\rm Fr}\nolimits_{k^{\infty},\,l^{\infty}}\times\mathop{\rm Fred}\nolimits({\mathcal{H}})_{(l)}\rightarrow\mathop{\rm Fred}\nolimits({\mathcal{H}})_{(l)}.$ This $\mathop{\rm Fred}\nolimits({\mathcal{H}})_{(l)}$-bundle in our version of the twisted $K$-theory should play the same role as the $\mathop{\rm Fred}\nolimits({\mathcal{H}})$-bundle associated (by the action $\gamma$) with the universal $\mathop{\rm PU}\nolimits({\mathcal{H}})$-bundle in the “usual” version of the twisted $K$-theory. ### 4.3. Some speculations We have already noticed (see Remark 13) that $\mathfrak{G}_{k,\,l}$ is not an action groupoid related to an action of some Lie group on $\mathop{\rm Gr}\nolimits_{k,\,l}.$ But in the direct limit it is an “action groupoid”. More precisely, consider $\mathfrak{G}:=\lim\limits_{\longrightarrow\atop{(k,\,l)=1}}\mathfrak{G}_{k,\,l}$ (the maps are induced by the tensor product). Since $\mathop{\rm Fr}\nolimits:=\lim\limits_{\longrightarrow\atop{(k,\,l)=1}}\mathop{\rm Fr}\nolimits_{k,\,l}$ is an $H$-space (even an infinite loop space [9]), we see that $\mathfrak{G}$ corresponds to the action of $\mathop{\rm Fr}\nolimits$ on $\mathop{\rm Gr}\nolimits$ (see Subsection 2.4). Moreover, in this situation the map (26) can be extended to the fibration (39) $\mathop{\rm Gr}\nolimits\rightarrow\lim\limits_{\longrightarrow\atop{k}}\mathop{\rm BPU}\nolimits(k)\rightarrow\mathop{\rm BFr}\nolimits\quad\hbox{i.e.}\quad\mathop{\rm BSU}\nolimits_{\otimes}\rightarrow\mathop{\rm K}\nolimits(\mathbb{Q}/\mathbb{Z},\,2)\times\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)\rightarrow\mathop{\rm BFr}\nolimits$ (cf. Subsection 2.4). Note that we can also define $\widetilde{\mathop{\rm Fr}\nolimits}:=\lim\limits_{\longrightarrow\atop{(k,\,l)=1}}\widetilde{\mathop{\rm Fr}\nolimits}_{k,\,l}$ (see (12)) and consider the corresponding fibration (40) $\mathop{\rm BSU}\nolimits_{\otimes}\rightarrow\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)\rightarrow\mathop{\rm B}\nolimits\widetilde{\mathop{\rm Fr}\nolimits}.$ In fact, $\mathop{\rm BFr}\nolimits=\mathop{\rm K}\nolimits(\mathbb{Q}/\mathbb{Z},\,2)\times\mathop{\rm B}\nolimits\widetilde{\mathop{\rm Fr}\nolimits}$. We also have the “unitary” version (41) $\mathop{\rm BU}\nolimits_{\otimes}\rightarrow\prod_{n\geq 1}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)\rightarrow\mathop{\rm BFr}\nolimits,$ where recall $\mathop{\rm BU}\nolimits_{\otimes}\cong\mathbb{C}P^{\infty}\times\mathop{\rm BSU}\nolimits_{\otimes}$ and therefore it splits as follows: $\mathbb{C}P^{\infty}\times\mathop{\rm BSU}\nolimits_{\otimes}\rightarrow\mathop{\rm K}\nolimits(\mathbb{Q},\,2)\times\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Q}/\mathbb{Z},\,2)\times\mathop{\rm B}\nolimits\widetilde{\mathop{\rm Fr}\nolimits}.$ The part $\mathbb{C}P^{\infty}\rightarrow\mathop{\rm K}\nolimits(\mathbb{Q},\,2)\rightarrow\mathop{\rm K}\nolimits(\mathbb{Q}/\mathbb{Z},\,2)$ corresponds to the “usual” finite Brauer group $H^{3}_{tors}(X,\,\mathbb{Z})$ ($=\mathop{\rm coker}\nolimits\\{H^{2}(X,\,\mathbb{Q})\rightarrow H^{2}(X,\,\mathbb{Q}/\mathbb{Z})\\}=\mathop{\rm im}\nolimits\delta\colon\\{H^{2}(X,\,\mathbb{Q}/\mathbb{Z})\rightarrow H^{3}(X,\,\mathbb{Z})\\}$, cf. Remark 7). Therefore using the fibration (40) one can define a “noncommutative” analog of the Brauer group of $X$ as $\mathop{\rm coker}\nolimits\\{[X,\,\prod_{n\geq 2}\mathop{\rm K}\nolimits(\mathbb{Q},\,2n)]\rightarrow[X,\,\mathop{\rm B}\nolimits\widetilde{\mathop{\rm Fr}\nolimits}]\\}.$ In this connection note that $\mathop{\rm BSU}\nolimits_{\otimes}$ represents the group of virtual $\mathop{\rm SU}\nolimits$-bundles of virtual dimension 1 with respect to the tensor product while $\mathbb{C}P^{\infty}$ represents the Picard group, i.e. the group of line bundles with respect to the tensor product too. The Picard group acts on $K(X)$ by group homomorphisms [2] and this leads to the “usual” twisted $K$-theory. Comparing two previous subsections we see that the “action” of $\mathop{\rm Fr}\nolimits_{k^{\infty}}$ on $K(X)$ (strictly speaking, on the localization $K(X)[\frac{1}{l}]$) is an analog of the action of $\mathop{\rm PU}\nolimits(k^{\infty}):=\lim\limits_{\longrightarrow\atop{n}}\mathop{\rm PU}\nolimits(k^{n})$ on $K(X)$ which leads to the $k$-primary component of $Br(X).$ So the idea is to show that $\gamma^{\prime}$ (see the previous subsection) gives rise to the action of the $H$-space $\mathop{\rm Fr}\nolimits$ on the $K$-theory spectrum and using this action to associate the corresponding $\mathop{\rm Fred}\nolimits({\mathcal{H}})$-bundle with the universal $\mathop{\rm Fr}\nolimits$-bundle over $\mathop{\rm BFr}\nolimits$ (as in the case of the “usual” twisted $K$-theory one associates a $\mathop{\rm Fred}\nolimits({\mathcal{H}})$-bundle with the universal $\mathop{\rm PU}\nolimits({\mathcal{H}})$-bundle over $\mathop{\rm BPU}\nolimits({\mathcal{H}})$ using the action $\gamma$ (see Subsection 4.1)). ## References [1] M. Atiyah, G. Segal Twisted K-theory // arXiv:math/0407054v2 [math.KT] * [2] M. Atiyah, G. Segal Twisted K-theory and cohomology // arXiv:math/0510674v1 [math.KT] * [3] R. Brown: From Groups to Groupoids: A Brief Survey. Bull. London Math. Soc. 19, 113-134, 1987. * [4] M. Karoubi: K-theory. An Introduction. Springer Verlag, 1978. * [5] R. Meyer: Morita Equivalence In Algebra And Geometry. http://citeseer.ist.psu.edu/meyer97morita.html * [6] F.P. Peterson: Some remarks on Chern classes, Ann. Math. 69 (1959) 414-420. * [7] R.S. Pierce: Associative Algebras. Springer Verlag, 1982. * [8] C.A. Rossi: Principal bundles with groupoid structure: local vs. global theory and nonabelian čech cohomology // arXiv:math/0404449v1 [math.DG] * [9] G.B. Segal: Categories and cohomology theories. Topology 13 (1974). * [10] A.V. Ershov: Theories of bundles with additional structures. Fundamentalnaya i prikladnaya matematika, vol. 13 (2007), no. 8, pp. 77 98. * [11] A.V. Ershov: Theories of bundles with additional homotopy conditions // arXiv:0804.1119v3 [math.KT] * [12] A.V. Ershov: A generalization of the topological Brauer group // Journal of K-theory: K-theory and its Applications to Algebra, Geometry, and Topology , Volume 2, Special Issue 03, December 2008, pp 407-444 * [13] A.V. Ershov: A model of the twisted $K$-theory // arXiv:1005.3807v5 [math.KT]	arxiv-papers	2008-07-22T19:50:30	2024-09-04T02:48:56.952554	{ "license": "Public Domain", "authors": "A.V. Ershov", "submitter": "Andrey V. Ershov", "url": "https://arxiv.org/abs/0807.3544" }
0807.3908	11institutetext: T-7, Center for Non-Linear Studies Los Alamos National Laboratory marko@lanl.gov # A Distributed Process Infrastructure for a Distributed Data Structure Marko A. Rodriguez ###### Abstract The Resource Description Framework (RDF) is continuing to grow outside the bounds of its initial function as a metadata framework and into the domain of general-purpose data modeling. This expansion has been facilitated by the continued increase in the capacity and speed of RDF database repositories known as triple-stores. High-end RDF triple-stores can hold and process on the order of 10 billion triples. In an effort to provide a seamless integration of the data contained in RDF repositories, the Linked Data community is providing specifications for linking RDF data sets into a universal distributed graph that can be traversed by both man and machine. While the seamless integration of RDF data sets is important, at the scale of the data sets that currently exist and will ultimately grow to become, the “download and index” philosophy of the World Wide Web will not so easily map over to the Semantic Web. This essay discusses the importance of adding a distributed RDF process infrastructure to the current distributed RDF data structure. ## 1 Introduction The Semantic Web community has introduced a set of standards and protocols for representing, querying, and manipulating a distributed, directed labeled graph of web resources. The Uniform Resource Identifier (URI), as the foundation standard of the World Wide Web, provides the distributed address space for web resources. The Resource Description Framework (RDF), as the foundation standard of the Semantic Web, provides a data model for graphing web resources. In a local environment, URIs can be minted and related to one another in a relatively straightforward manner. What becomes difficult is making the link from the URIs of one environment to the URIs of another. That is, the same thing can be given two unique URIs and the cross repository linking of such URIs is a difficult problem. With the World Wide Web, the author of an HTML document is responsible for the linking of web resources and this can be managed in a straightforward manner as HTML documents maintain a relatively small set of links. However, on the Semantic Web, there is no human consumable free-text per se. Instead, every minutia of information is encoded as a statement (or triple). The amount of resources and links can grow fast and thus is only maintainable, for the most part, by machines. The Linked Data community expends much effort ensuring that RDF repositories are linked appropriately such that what emerges is a global web of data that is rich in content and able to be traversed by both man and machine berners:ldata2006 . However, traversing the Semantic Web is not quite the same as traversing the World Wide Web. For the human, it is reasonable to traverse from repository to repository exploring the Semantic Web in a manner similar to how the World Wide Web is traversed. However, for a machine, it is a different story. There will be orders of magnitude more resources and links on the Semantic Web. While a machine can crawl and pull the data to its local environment for processing, this becomes inefficient when the process’ data requirements must span large parts of the Semantic Web graph. For distributed graph computations, a more efficient mechanism would be to migrate the process between RDF repositories so that information is not pulled back to the local environment, but instead, processed where the data is maintained. In other words, an efficient mechanism for processing the Semantic Web graph would be to move the process to the data, not the data to the process. While the Semantic Web provides an infrastructure to support the distributed representation of a graph of resources, there currently does not exist a distributed process infrastructure for analyzing and manipulating this graph. For the World Wide Web, the search engine philosophy of “download and index” has made it possible for end users to find information on the World Wide Web in a more efficient manner than by simply surfing and bookmarking. With modern commercial triple-stores scaling to the order of 10 billion triples, centralized indexing repositories will have to contend with not only the volume of data, but also the computational complexities of analyzing such a graph in a sophisticated way. The Semantic Web provides a much richer machine processable data structure than what is provided by HTML and thus, antiquated keyword search simply does not take significant advantage of what the Semantic Web is providing. Labeled graph query languages such as SPARQL are one method of retrieving data from the Semantic Web, but like keyword search, this is not the end story. The future of the Semantic Web will be rife with algorithms from many schools of thoughts. Many of these algorithms will compute across various repositories of the Semantic Web and will require a distributed Turing complete infrastructure to do so. ## 2 The RDF Virtual Machine Model The notion of process migration for distributed computing is not new. For instance, the standards and technologies associated with Grid computing are aimed at facilitating the distributed and shared use of computing resources foster:grid2001 . Furthermore, the idea of a Semantic Grid has been proposed as an extension to the current Grid infrastructure to support the semantic annotation of computing resources to aid in discovery, reuse, etc. roure:semgrid2005 . > “The Semantic Grid refers to an approach to Grid computing in which > information, computing resources and services are described using the > semantic data model.”111Semantic Grid article is available at > http://en.wikipedia.org/wiki/Semantic_grid The distributed process infrastructure discussed in this essay is aimed primarily at supporting the exploitation of the Semantic Web data structure directly as a data structure, not to augment an existing distributed processing infrastructure with the standards of the Semantic Web. The Web has provided a distributed address space that is infinite in size. It is possible to represent processes such they compute within this address space agnostic to the physical machines that support their execution. The idea of an RDF virtual machine (RVM) was introduced to facilitate this level of abstraction rodriguez:rvm2008 . With an RVM, the computing machine, the computing machine’s software, and the data being processed are all represented in RDF and within the URI address space (as well as the blank and literal space). Some of the major consequences of the this computing model are * • an RVM has no reference to the underlying physical (or virtual) machine executing it * • an RVM is constrained to the URI address space * • an RVM only executes RDF encoded software * • an RVM only processes RDF encoded data. A primary tenet of the RVM model is that RDF is the lowest level of representation and in being the lowest level of representation, RDF is used to model all aspects of computing. Thus, RDF is used to define those data structures that are common in computing such as the lists, the sets, the arrays, the maps, the vectors, the stacks, the program counters, the heaps, the trees, the objects, the instructions—in general, the general-purpose modeling construct: the graph. Thus, RDF is used not just for representing metadata, RDF is used for representing data. When all information is placed in the URI address space, computing is at a new level of abstraction. This level of abstraction is a shared memory space that is built on the existing Web infrastructure. As more data sets are linked to the growing Linked Data web, there will exist algorithms that will be burdened by the “over the wire” speeds of the Internet. With the RVM model, the RVM can compute regardless of the physical machine that is supporting its execution. In this way, an RVM can traverse the Linked Data set not by pulling data to a local environment, but by actually moving between machines and more specifically, moving to those machines that are maintaining the subgraph of the Semantic Web that is of interest to the algorithm at particular points in time. For every architecture, there are drawbacks. In the RVM model of computing, the issues that are obvious are the issues of security, speed, and adoption. With respect to security, the migration of process requires an infrastructure to protect the Semantic Web and the physical machines that support its representation. It is important to ensure that poorly or maliciously written RDF code does not destroy the integrity of an RDF data set, does not abuse the computational resources of a publicly available physical machine, and only accesses those aspects of an RDF data set that it has permission to access. With respect to speed, the RVM model of distributed computing on the Semantic Web requires fast execution times. When multiple levels of computing are represented in the URI address space there does not exist a direct physical memory correlate and as such, computing is slower. With respect to the adoption of the RVM model, the model requires RDF data providers to support RVM “farms” (or open computing spaces) for foreign processes to use for their execution. Furthermore, the model relies on programming languages that are engineered for a pure URI address space and thus, requires the development of algorithms in these languages. Initial solutions to the aforementioned issues have been previously discussed in rodriguez:rvm2008 . The next section provides a brief overview of the various RDF programming languages that exist. ## 3 RDF Programming Languages An RDF programming language is a language that is not only compiled into RDF, but also aimed at directly manipulating an RDF graph. RDF programming languages include: * • FABL: an object-oriented RDF programming language fabl:bureau2001 * • Adenosine: an RDF programming language * • Ripple: a functional, stack-based RDF programming language ripple:shinavier2007 * • Neno/Fhat: an object-oriented RDF programming language with RDF-encoded virtual machine. rodriguez:gpsemnet2007 These languages were designed to take explicit advantage of RDF as a data model. Unlike RDF APIs in other languages such as Java, C, etc. these languages do not require the developer to work with two different data models. That is, RDF APIs in other languages require the developer to write in the constructs of the programming environment as well as in the constructs of RDF. With RDF programming languages, the constructs of the language are simply URIs and RDF triples. There is no disjoint experience for the developer fabl:bureau2001 . RDF programming languages compile down to RDF and thus, can be accessed like any piece of data on the Semantic Web. Furthermore, unique situations emerge when RDF code is represented across different physical server machines. Because all RDF software is in the same URI address space, there is nothing that prevents the software, much like the data, to by physically distributed. With an RDF virtual machine executing compiled RDF code, it is possible for the virtual machine and the compiled code to be relocated by simply downloading the RDF subgraph to another environment. Thus, instead of migrating large amounts of data to a local environment for processing, the RDF virtual machine and code can be migrated to the remote environment. In this way, the process is moved to the data, not the data to the process. ## 4 Conclusion The Semantic Web provides an infrastructure that supports an instantiation of a distributed graph of web resources. However, it lacks a distributed infrastructure for processing that graph. For the World Wide Web, the solution to the issue of processing the vast amount of information has been to literally download the entire Web and index and process it at a single local environment. While the content on the World Wide Web is distributed, the means by which the information on the Web is analyzed is not. The Semantic Web need not fall into this same model. With the nearly limitless ways in which a directed labeled graph (an RDF graph) can be searched and manipulated, it would be a disappointment if the information on the Semantic Web was left to centralized repositories to store, index, and provide query functionality. ## Acknowledgements Joshua Shinavier read draft versions of this essay and after doing so, provided many useful comments. ## References * (1) Tim Berners-Lee. Linked data. Technical report, World Wide Web Consortium, 2006. * (2) David De Roure, Nicholas R. Jennings, and Nigel R. Shadbolt. The semantic grid: Past, present and future. In Proceedings of the IEEE, volume 93, pages 669–681, March 2005\. * (3) Ian Foster, Carl Kesselman, and Steven Tuecke. The anatomy of the grid: Enabling scalable virtual organizations. International Jounral of Supercomputer Applications, 15(3), 2001\. * (4) Chris Goad. Describing computation within RDF. In Proceedings of the International Semantic Web Working Symposium, 2004. * (5) Marko A. Rodriguez. General-purpose computing on a semantic network substrate. Technical Report LA-UR-07-2885, Los Alamos National Laboratory, 2007. * (6) Marko A. Rodriguez and Joshua Shinavier. The RDF virtual machine. Technical report, Los Alamos National Laboratory Technical Report, 2008\. * (7) Joshua Shinavier. Functional programs as linked data. In 3rd Workshop on Scripting for the Semantic Web, Innsbruck, Austria, 2007.	arxiv-papers	2008-07-24T15:16:16	2024-09-04T02:48:56.965545	{ "license": "Public Domain", "authors": "Marko A. Rodriguez", "submitter": "Marko A. Rodriguez", "url": "https://arxiv.org/abs/0807.3908" }
0807.3932	# Resonant Spin Excitation in the High Temperature Superconductor Ba0.6K0.4Fe2As2 A. D. Christianson Neutron Scattering Science Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA 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 OX11 0QX, United Kingdom R. Osborn Materials Science Division, Argonne National Laboratory, Argonne, IL 60439-4845, USA ROsborn@anl.gov S. Rosenkranz Materials Science Division, Argonne National Laboratory, Argonne, IL 60439-4845, USA M. D. Lumsden Neutron Scattering Science Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA C. D. Malliakas Materials Science Division, Argonne National Laboratory, Argonne, IL 60439-4845, USA Department of Chemistry, Northwestern University, Evanston, IL 60208-3113 l. S. Todorov H. Claus D. Y. Chung Materials Science Division, Argonne National Laboratory, Argonne, IL 60439-4845, USA M. G. Kanatzidis Materials Science Division, Argonne National Laboratory, Argonne, IL 60439-4845, USA Department of Chemistry, Northwestern University, Evanston, IL 60208-3113 R. I. Bewley T. Guidi ISIS Pulsed Neutron and Muon Facility, Rutherford Appleton Laboratory, Chilton, Didcot OX11 0QX, United Kingdom ###### Abstract The recent observations of superconductivity at temperatures up to 55K in compounds containing layers of iron arsenide Kamihara:2008p7994 ; Takahashi:2008p7382 ; Ren:2008p8174 ; Rotter:2008p8535 have revealed a new class of high temperature superconductors that show striking similarities to the more familiar cuprates. In both series of compounds, the onset of superconductivity is associated with the suppression of magnetic order by doping holes and/or electrons into the band DeLaCruz:2008p8095 leading to theories in which magnetic fluctuations are either responsible for or strongly coupled to the superconducting order parameter Norman:2007p2671 . In the cuprates, theories of magnetic pairing have been invoked to explain the observation of a resonant magnetic excitation that scales in energy with the superconducting energy gap and is suppressed above the superconducting transition temperature, Tc. Such resonant excitations have been shown by inelastic neutron scattering to be a universal feature of the cuprate superconductors Hufner:2008p9494 , and have even been observed in heavy fermion superconductors with much lower transition temperatures Sato:2001p11225 ; Stock:2008p6655 ; Stockert:2008p11663 . In this paper, we show neutron scattering evidence of a resonant excitation in Ba0.6K0.4Fe2As2, which is a superconductor below 38 K Rotter:2008p8535 , at the momentum transfer associated with magnetic order in the undoped compound, BaFe2As2, and at an energy transfer that is consistent with scaling in other strongly correlated electron superconductors. As in the cuprates, the peak disappears at Tc providing the first experimental confirmation of a strong coupling of the magnetic fluctuation spectrum to the superconducting order parameter in the new iron arsenide superconductors. Unconventional superconductivity has been the subject of considerable theoretical and experimental interest since the discovery of superconductivity in CeCu2Si2 and other heavy fermion compounds Steglich:1979p11524 , an interest that was only intensified by the discovery of cuprate superconductors with transition temperatures in excess of 100 K Norman:2007p2671 . Although significant progress has been made, the origin of unconventional superconductivity is still not understood. The observation of a magnetic resonance in the spin excitation spectrum which appears concurrently with the onset of superconductivity in both the high Tc cuprates RossatMignod:1991p10509 ; Mook:1993p10706 ; Fong:1999p11512 ; Dai:2000p10619 ; He:2002p11430 and the heavy fermion superconductors Sato:2001p11225 ; Stock:2008p6655 ; Stockert:2008p11663 offers the tantalizing possibility of a unifying theme for unconventional superconductivity that spans a diverse range of superconducting materials. Recently, a new family of superconductors containing layers of Fe2As2 has been discovered with Tcs in excess of 50 K stimulating considerable experimental and theoretical activity Kamihara:2008p7994 ; Takahashi:2008p7382 ; Ren:2008p8174 . Although there is mounting evidence that the superconductivity in this new family is also unconventional Mazin:2008p8170 , there is as yet no consensus concerning the mechanism giving rise to superconductivity or even the superconducting pairing symmetry. In this letter, we describe neutron scattering data that confirm for the first time the existence of a resonant spin excitation below Tc in the iron arsenide materials at an energy that shows similar scaling to other heavy fermion and high-Tc superconductors. Figure 1: The crystal structure of Ba0.6K0.4Fe2As2 (Fe: blue spheres, As: yellow spheres, Ba/K: red spheres). The unit cell contains two layers of Fe2As2 tetrahedra separated by planes of barium atoms. The blue arrows show the ordering of the iron spins observed in the undoped parent compound BaFe2As2 Huang:2008p8145 . The atomic distance of 2.77 Å that characterizes both the antiferromagnetic modulation and the newly observed resonant excitation is shown by the red arrows. Figure 2: Inelastic neutron scattering from Ba0.6K0.4Fe2As2 measured using incident neutron energies of 60 meV (a,b) and 15 meV (c,d), at temperatures of 7 K (a,c) and 50 K (b,d). Above Tc (b,d), the magnetic response consists of a column of excitations centred at 1.15 Å-1 and a dispersive ferromagnetic excitation. Below Tc, low-energy intensity in the column is transferred to the resonant excitation at 15 meV. The strong scattering at low energy transfers in each plot arises from the tail of strong elastic nuclear scattering, while the scattering increases strongly at higher Q due to inelastic phonon scattering. The colour scale is in units of mbarns/sr/meV/mol. Although the first iron arsenide superconductors were based on doped variants of RFeAsO, where R is a rare earth element, there has been considerable interest in a new series of tetragonal compounds based on AFe2As2 (A = Ba, Sr, Ca), in which superconductivity is induced either by doping the A-site with potassium or sodium Rotter:2008p8535 ; Sasmal:2008p8092 or by applying pressure Park:2008p8883 . These contain the same tetrahedrally-coordinated Fe2As2 planes as the LaFeAsO compounds (Fig. 1), separated by planes of the doped A-site, which acts as a charge reservoir. So far, the maximum Tc is 38 K Rotter:2008p8535 seen in Ba0.6K0.4Fe2As2, which is the compound we are investigating in this letter. The antiferromagnetic structure of the undoped parent compound, BaFe2As2 is illustrated in Figure 1 Huang:2008p8145 . Polycrystalline samples of Ba0.6K0.4Fe2As2 were prepared by solid state synthesis techniques described in the supplemental information. From XRD and SEM/EDS measurements, the estimated phase purity of the samples is $\sim$90% and a sharp superconducting transition was observed by magnetic susceptibility at the previously reported temperature of 38 K Rotter:2008p8535 . The inelastic neutron scattering experiments were performed on the recently commissioned time-of-flight MERLIN spectrometer at the ISIS Pulsed Neutron and Muon Facility Bewley:2006p11347 , using incident energies of 15, 30, 60, and 100 meV. The data were placed on an absolute intensity scale by normalization to a vanadium standard. Figure 2 shows colour plots of the measured inelastic neutron scattering intensity as a function of momentum transfer, $Q$, and energy transfer, $\omega$, at two incident neutron energies, 15 and 60 meV, below (T = 7 K) and above (T = 50 K) the superconducting transition temperature. The range of data is limited by the kinematics of the scattering process at these low scattering angles so it is necessary to combine data from multiple incident energies in order to access data over a reasonable range of energy transfers at such low $Q$s. The most striking difference between the scattering above and below Tc is seen at Q$\sim$1.15 Å-1 and an energy transfer of $\sim$14 meV. At 7 K, there is clearly a peak that is well-defined in both $Q$ and $\omega$ which is not present at 50 K (Fig. 2a and 2b). Conversely, measurements at lower incident energy show that above Tc, there is a column of scattering intensity as a function of energy transfer, centred at $Q=1.15$ Å-1 (Fig. 2d), which is no longer visible below Tc (Fig. 2c). The $Q$ characterizing these contributions to the magnetic response corresponds to the periodicity of the antiferromagnetic order within each plane of iron spins observed in the undoped parent compound, BaFe2As2 Huang:2008p8145 (see also Fig. 1), so this scattering is consistent with the persistence of antiferromagnetic fluctuations in the absence of long-range spin order. Our measurements indicate that these antiferromagnetic fluctuations condense into a sharp resonant excitation, localized in both Q and $\omega$, as previously seen in the cuprate and heavy fermion superconductors. Figure 3: Inelastic neutron scattering from Ba0.6K0.4Fe2As2 integrated over a Q-range of 1.0 to 1.3 Å-1 (a) at 7 K measured using incident neutron energies of 15 meV (yellow circles), 30 meV (blue circles) and 60 meV (green circles), and (b) at 7 K (green circles) and 50 K (red circles) using an incident neutron energy of 60 meV. The error bars are derived from the square root of the raw detector counts. The data show a clear resonant peak at 7 K and the transfer of spectral weight from this peak to lower energies at 50 K, i.e., above the superconducting transition. Before discussing this resonant excitation in more detail, we should mention that there is also clear evidence for ferromagnetic fluctuations shown by the additional band of excitations in Fig. 2c and 2d that disperses in Q and $\omega$. The (Q,$\omega$)-dependence is consistent with the dispersion of ferromagnetic spin waves emerging from Q = 0, flattening at the ferromagnetic Brillouin zone boundary (which is equivalent to the antiferromagnetic zone centre) at $\omega\sim 6$ meV. The existence of ferromagnetic fluctuations in the undoped parent compound has been predicted theoretically Mazin:2008p6770 , although it was argued that they would be suppressed by doping before the onset of superconductivity. We have also observed this mode in BaFe2As2 but it is clear that they are still present in the superconducting phase although there is some redistribution of intensity on crossing Tc. The strong intensity of this mode, coupled with the intensity changes at Tc, indicate that it is intrinsic to the bulk superconducting phase. We will discuss this aspect of our data further in a future publication. To elucidate the evolution of the magnetic response, we combine measurements below Tc at three incident energies using, in each case, a low-energy cutoff that excludes the tail of strong elastic nuclear scattering. The resulting data are shown in Fig. 3a where the resonant excitation is seen to peak sharply at $\omega_{0}\approx 14$ meV. We cannot rule out that there could be a small phononic contribution to the scattering within this energy window, but the strong temperature dependence indicates that it is predominantly magnetic. A comparison of the data above and below Tc measured with an incident energy of 60 meV shows that the spectral weight of the column of scattering evident in Fig. 2d has transferred into the resonant peak in the superconducting phase. Figure 4: Inelastic neutron scattering from Ba0.6K0.4Fe2As2 as a function of temperature integrated over a $Q$ range of 1.0 to 1.3 Å-1 and an energy transfer range of 12.5 to 17.5 meV. The integration range corresponds to the region of maximum intensity of the resonant excitation observed below Tc (see Fig. 2). The error bars are derived from the square root of the raw detector counts. The dashed line is a guide to the eye below Tc and shows the average value of the integrals above Tc. We performed a series of shorter measurements in order to determine the temperature dependence of this resonant excitation. Fig. 4 shows data integrated over the (Q,$\omega$)-region of maximum intensity in the resonant excitation. As also observed in the cuprates, the intensity of the resonance falls to zero at Tc confirming the strong coupling of this excitation to the superconducting order parameter. The existence of similar resonant excitations in other strongly correlated superconductors, such as the high-Tc cuprates and the heavy fermion superconductors, is commonly taken as evidence of an unconventional symmetry of the superconducting order parameter Chang:2007p11651 . The dynamic magnetic susceptibility is predicted to be enhanced at certain values of Q in the superconducting phase by a coherence factor provided that the energy gap symmetry has the following form: $\Delta_{\textbf{k}+\textbf{Q}}=-\Delta_{\textbf{k}}$ where k and $\textbf{k}+\textbf{Q}$ are wavevectors on different parts of the Fermi surface. In the cuprates and heavy fermion superconductors, this is realized by d${}_{x^{2}-y^{2}}$-symmetry, which has nodes in the energy gap within a single Fermi surface. In these cases, Q spans sections of the same Fermi surface that are gapped with opposite phase, such as $\textbf{Q}=(\pi,\pi)$ in the cuprates. However, such a scenario seems to be ruled out by photoemission results on Ba0.6K0.4Fe2As2 that show no evidence of any anisotropy of the energy gap Ding:2008p8816 . According to band structure calculation, the Fermi surfaces of the iron arsenide superconductors are predominantly derived from the iron d-electrons, and comprise two small hole pockets centred at the centre of the Brillouin zone and two small electron pockets at the zone boundary Liu:2008p8202 ; Yang:2008p8157 . ARPES sees isotropic gaps around each of the measured surfaces apparently ruling out a d-wave gap symmetry Ding:2008p8816 . A resolution of this apparent discrepancy has been provided by theoretical predictions that the gap symmetry is not d-wave, but rather extended s±-wave Mazin:2008p6770 . Mazin et al have postulated that the gaps in each hole and electron pocket are isotropic, but that the gaps on the hole and electron pockets have opposite phase. This means that magnetic fluctuations are amplified by the coherence factor at values of Q that couple the hole and electron pockets, as has been confirmed by explicit calculations of the neutron scattering intensities Maier:2008p9552 . This is precisely where we have observed the resonant excitation, so our measurements, combined with the ARPES data, provide strong experimental support for the validity of extended s±-wave gap models. In conclusion, we have presented the first experimental evidence in the new class of iron arsenide superconductors for the existence of a resonant excitation in the dynamic magnetic susceptibility that disappears above the superconducting transition temperature. The energy of this resonant excitation is at $\omega_{0}\sim 14$ meV, or 4.3 Tc, just under the canonical value of 5 Tc seen in the cuprate superconductors Hufner:2008p9494 . However, Stock et al have argued that it is more appropriate to scale $\omega_{0}/2\Delta_{0}$, where $\Delta_{0}$ is the maximum value of the gap, and they estimate that this ratio ranges from 0.62 to 0.74 in a wide range of materials Stock:2008p6655 . From ARPES data on Ba0.6K0.4Fe2, $\Delta_{0}\sim 12$ meV Ding:2008p8816 , giving a ratio of $\omega_{0}/2\Delta_{0}\sim 0.58$. It is remarkable that materials with such a divergent range of Tcs (over two orders of magnitude) could be unified by such a simple scaling relation. ###### Acknowledgements. We acknowledge helpful scientific discussions with Christopher Stock. This work was supported by the Division of Materials Sciences and Engineering Division and the Scientific User Facilities Division of the Office of Basic Energy Sciences, U.S. Department of Energy Office of Science, under Contract Nos. DE-AC02-06CH11357 and DE-AC05-00OR22725. Correspondence and requests for materials should be addressed to R. O. ## References * (1) Kamihara, Y., Watanabe, T., Hirano, M. & Hosono, H. Iron-based layered superconductor La[O1-xFx]FeAs (x=0.05-0.12) with Tc=26 K. _J Am Chem Soc_ 130, 3296–3297 (2008). * (2) Takahashi, H. _et al._ Superconductivity at 43 K in an iron-based layered compound LaO1-xFxFeAs. _Nature_ 453, 376–378 (2008). * (3) Ren, Z.-A. _et al._ Superconductivity and phase diagram in iron-based arsenic-oxides (RE=rare-earth metal) without fluorine doping. _Europhys. Lett._ 83, 17002 (2008). * (4) Rotter, M., Tegel, M. & Johrendt, D. Superconductivity at 38 K in the iron arsenide (Ba1-xKx)Fe2As2. _arXiv_ 0805.4630 (2008). * (5) Cruz, C. D. L. _et al._ Magnetic order close to superconductivity in the iron-based layered LaO1-xFxFeAs systems. _Nature_ 453, 899–902 (2008). * (6) Norman, M. R. High Temperature Superconductivity - Magnetic Mechanisms, vol. 5 p. 2671 of _Handbook of Magnetism and Advanced Magnetic Materials_ (Wiley, New York, 2007). * (7) Hüfner, S., Hossain, M. A., Damascelli, A. & Sawatzky, G. A. Two gaps make a high-temperature superconductor? _Rep. Prog. Phys._ 71, 062501 (2008). * (8) Stock, C., Broholm, C., Hudis, J., Kang, H. J. & Petrovic, C. Spin Resonance in the d-Wave Superconductor CeCoIn5. _Phys. Rev. Lett._ 100, 087001 (2008). * (9) Stockert, O. _et al._ Magnetism and superconductivity in the heavy-fermion compound CeCu2Si2 studied by neutron scattering. _Physica B_ 403, 973–976 (2008). * (10) Sato, N. K. _et al._ Strong coupling between local moments and superconducting ‘heavy’ electrons in UPd2Al3. _Nature_ 410, 340–343 (2001). * (11) Steglich, F. _et al._ Superconductivity in the Presence of Strong Pauli Paramagnetism: CeCu2Si2. _Phys. Rev. Lett._ 43, 1892–1896 (1979). * (12) Rossat-Mignod, J. _et al._ Neutron scattering study of the YBa2Cu3O6+x system. _Physica C_ 185, 86–92 (1991). * (13) Mook, H. A., Yethiraj, M., Aeppli, G., Mason, T. E. & Armstrong, T. Polarized neutron determination of the magnetic excitations in YBa2Cu3O7. _Phys Rev Lett_ 70, 3490–3493 (1993). * (14) Fong, H. F. _et al._ Neutron scattering from magnetic excitations in Bi2Sr2CaCu2O8+δ. _Nature_ 398, 588–591 (1999). * (15) Dai, P., Mook, H. A., Aeppli, G., Hayden, S. M. & Doğan, F. Resonance as a measure of pairing correlations in the high-Tc superconductor YBa2Cu3O6.6. _Nature_ 406, 965–968 (2000). * (16) He, H. _et al._ Magnetic Resonant Mode in the Single-Layer High-Temperature Superconductor Tl2Ba2CuO6+δ. _Science_ 295, 1045–1047 (2002). * (17) Mazin, I. I., Johannes, M. D., Boeri, L., Koepernik, K. & Singh, D. J. Challenge of unravelling magnetic properties of LaFeAsO. _arXiv_ 0806.1869 (2008). * (18) Huang, Q. _et al._ Magnetic order in BaFe2As2, the parent compound of the FeAs based superconductors in a new structural family. _arXiv_ 0806.2776 (2008). * (19) Sasmal, K. _et al._ Superconductivity up to 37 K in (A1-xSrx)Fe2As2 with A=K and Cs. _arXiv_ 0806.1301 (2008). * (20) Park, T. _et al._ Pressure-induced superconductivity in CaFe2As2. _J. Phys.-Cond. Matt._ 20, 322204 (2008). * (21) Bewley, R. _et al._ MERLIN, a new high count rate spectrometer at ISIS. _Physica B_ 385-386, 1029–1031 (2006). * (22) Mazin, I. I., Singh, D. J., Johannes, M. D. & Du, M. H. Unconventional sign-reversing superconductivity in LaFeAsO1-xFx. _arXiv_ 0803.2740 (2008). * (23) Chang, J., Eremin, I., Thalmeier, P. & Fulde, P. Theory of magnetic excitons in the heavy-fermion superconductor UPd2Al3. _Phys. Rev. B_ 75, 24503 (2007). * (24) Ding, H. _et al._ Fermi-surface–dependent nodeless superconducting gaps in Ba0.6K0.4Fe2As2. _Europhys. Lett._ 83, 47001 (2008). * (25) Liu, C. _et al._ The Fermi surface of Ba1-xKxFe2As2 and its evolution with doping. _arXiv_ 0806.3453 (2008). * (26) Yang, L. X. _et al._ Band structure and electronic signature of the spin density waves in bilayer iron-oxypnictide BaFe2As2. _arXiv_ 0806.2627 (2008). * (27) Maier, T. A. & Scalapino, D. J. Neutron scattering as a probe of the Fe-pnictide superconducting gap. _arXiv_ 0805.0316 (2008).	arxiv-papers	2008-07-24T17:40:13	2024-09-04T02:48:56.969160	{ "license": "Public Domain", "authors": "A. D. Christianson, E. A. Goremychkin, R. Osborn, S. Rosenkranz, M. D.\n Lumsden, C. D. Malliakas, l. S. Todorov, H. Claus, D. Y. Chung, M. G.\n Kanatzidis, R. I. Bewley, and T. Guidi", "submitter": "Ray Osborn", "url": "https://arxiv.org/abs/0807.3932" }
0807.3935	BABAR-PUB-08/23 SLAC-PUB-13317 ††thanks: Deceased††thanks: Deceased The BABAR Collaboration # Searches for $B$ meson decays to $\phi\phi$, $\phi\rho$, $\phi f_{0}(980)$, and $f_{0}(980)f_{0}(980)$ final states B. Aubert M. Bona Y. Karyotakis J. P. Lees V. Poireau E. Prencipe X. Prudent V. Tisserand Laboratoire de Physique des Particules, IN2P3/CNRS et Université de Savoie, F-74941 Annecy-Le-Vieux, France J. Garra Tico E. Grauges Universitat de Barcelona, Facultat de Fisica, Departament ECM, E-08028 Barcelona, Spain L. Lopezab A. Palanoab M. Pappagalloab INFN Sezione di Baria; Dipartmento di Fisica, Università di Barib, I-70126 Bari, Italy G. Eigen B. Stugu L. Sun University of Bergen, Institute of Physics, N-5007 Bergen, Norway G. S. Abrams M. Battaglia D. N. Brown R. N. Cahn R. G. Jacobsen L. T. Kerth Yu. G. Kolomensky G. Lynch I. L. Osipenkov M. T. Ronan K. Tackmann T. Tanabe Lawrence Berkeley National Laboratory and University of California, Berkeley, California 94720, USA C. M. Hawkes N. Soni A. T. Watson University of Birmingham, Birmingham, B15 2TT, United Kingdom H. Koch T. Schroeder Ruhr Universität Bochum, Institut für Experimentalphysik 1, D-44780 Bochum, Germany D. Walker University of Bristol, Bristol BS8 1TL, United Kingdom D. J. Asgeirsson B. G. Fulsom C. Hearty T. S. Mattison J. A. McKenna University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z1 M. Barrett A. Khan Brunel University, Uxbridge, Middlesex UB8 3PH, United Kingdom V. E. Blinov A. D. Bukin A. R. Buzykaev V. P. Druzhinin V. B. Golubev A. P. Onuchin S. I. Serednyakov Yu. I. Skovpen E. P. Solodov K. Yu. Todyshev Budker Institute of Nuclear Physics, Novosibirsk 630090, Russia M. Bondioli S. Curry I. Eschrich D. Kirkby A. J. Lankford P. Lund M. Mandelkern E. C. Martin D. P. Stoker University of California at Irvine, Irvine, California 92697, USA S. Abachi C. Buchanan University of California at Los Angeles, Los Angeles, California 90024, USA J. W. Gary F. Liu O. Long B. C. Shen G. M. Vitug Z. Yasin L. Zhang University of California at Riverside, Riverside, California 92521, USA V. Sharma University of California at San Diego, La Jolla, California 92093, USA C. Campagnari T. M. Hong D. Kovalskyi M. A. Mazur J. D. Richman University of California at Santa Barbara, Santa Barbara, California 93106, USA T. W. Beck A. M. Eisner C. J. Flacco C. A. Heusch J. Kroseberg W. S. Lockman T. Schalk B. A. Schumm A. Seiden L. Wang M. G. Wilson L. O. Winstrom University of California at Santa Cruz, Institute for Particle Physics, Santa Cruz, California 95064, USA C. H. Cheng D. A. Doll B. Echenard F. Fang D. G. Hitlin I. Narsky T. Piatenko F. C. Porter California Institute of Technology, Pasadena, California 91125, USA R. Andreassen G. Mancinelli B. T. Meadows K. Mishra M. D. Sokoloff University of Cincinnati, Cincinnati, Ohio 45221, USA P. C. Bloom W. T. Ford A. Gaz J. F. Hirschauer M. Nagel U. Nauenberg J. G. Smith K. A. Ulmer S. R. Wagner University of Colorado, Boulder, Colorado 80309, USA R. Ayad Now at Temple University, Philadelphia, Pennsylvania 19122, USA A. Soffer Now at Tel Aviv University, Tel Aviv, 69978, Israel W. H. Toki R. J. Wilson Colorado State University, Fort Collins, Colorado 80523, USA D. D. Altenburg E. Feltresi A. Hauke H. Jasper M. Karbach J. Merkel A. Petzold B. Spaan K. Wacker Technische Universität Dortmund, Fakultät Physik, D-44221 Dortmund, Germany M. J. Kobel W. F. Mader R. Nogowski K. R. Schubert R. Schwierz J. E. Sundermann A. Volk Technische Universität Dresden, Institut für Kern- und Teilchenphysik, D-01062 Dresden, Germany D. Bernard G. R. Bonneaud E. Latour Ch. Thiebaux M. Verderi Laboratoire Leprince-Ringuet, CNRS/IN2P3, Ecole Polytechnique, F-91128 Palaiseau, France P. J. Clark W. Gradl S. Playfer J. E. Watson University of Edinburgh, Edinburgh EH9 3JZ, United Kingdom M. Andreottiab D. Bettonia C. Bozzia R. Calabreseab A. Cecchiab G. Cibinettoab P. Franchiniab E. Luppiab M. Negriniab A. Petrellaab L. Piemontesea V. Santoroab INFN Sezione di Ferraraa; Dipartimento di Fisica, Università di Ferrarab, I-44100 Ferrara, Italy R. Baldini-Ferroli A. Calcaterra R. de Sangro G. Finocchiaro S. Pacetti P. Patteri I. M. Peruzzi Also with Università di Perugia, Dipartimento di Fisica, Perugia, Italy M. Piccolo M. Rama A. Zallo INFN Laboratori Nazionali di Frascati, I-00044 Frascati, Italy A. Buzzoa R. Contriab M. Lo Vetereab M. M. Macria M. R. Mongeab S. Passaggioa C. Patrignaniab E. Robuttia A. Santroniab S. Tosiab INFN Sezione di Genovaa; Dipartimento di Fisica, Università di Genovab, I-16146 Genova, Italy K. S. Chaisanguanthum M. Morii Harvard University, Cambridge, Massachusetts 02138, USA J. Marks S. Schenk U. Uwer Universität Heidelberg, Physikalisches Institut, Philosophenweg 12, D-69120 Heidelberg, Germany V. Klose H. M. Lacker Humboldt-Universität zu Berlin, Institut für Physik, Newtonstr. 15, D-12489 Berlin, Germany D. J. Bard P. D. Dauncey J. A. Nash W. Panduro Vazquez M. Tibbetts Imperial College London, London, SW7 2AZ, United Kingdom P. K. Behera X. Chai M. J. Charles U. Mallik University of Iowa, Iowa City, Iowa 52242, USA J. Cochran H. B. Crawley L. Dong W. T. Meyer S. Prell E. I. Rosenberg A. E. Rubin Iowa State University, Ames, Iowa 50011-3160, USA Y. Y. Gao A. V. Gritsan Z. J. Guo C. K. Lae Johns Hopkins University, Baltimore, Maryland 21218, USA A. G. Denig M. Fritsch G. Schott Universität Karlsruhe, Institut für Experimentelle Kernphysik, D-76021 Karlsruhe, Germany N. Arnaud J. Béquilleux A. D’Orazio M. Davier J. Firmino da Costa G. Grosdidier A. Höcker V. Lepeltier F. Le Diberder A. M. Lutz S. Pruvot P. Roudeau M. H. Schune J. Serrano V. Sordini Also with Università di Roma La Sapienza, I-00185 Roma, Italy A. Stocchi G. Wormser Laboratoire de l’Accélérateur Linéaire, IN2P3/CNRS et Université Paris-Sud 11, Centre Scientifique d’Orsay, B. P. 34, F-91898 Orsay Cedex, France D. J. Lange D. M. Wright Lawrence Livermore National Laboratory, Livermore, California 94550, USA I. Bingham J. P. Burke C. A. Chavez J. R. Fry E. Gabathuler R. Gamet D. E. Hutchcroft D. J. Payne C. Touramanis University of Liverpool, Liverpool L69 7ZE, United Kingdom A. J. Bevan C. K. Clarke K. A. George F. Di Lodovico R. Sacco M. Sigamani Queen Mary, University of London, London, E1 4NS, United Kingdom G. Cowan H. U. Flaecher D. A. Hopkins S. Paramesvaran F. Salvatore A. C. Wren University of London, Royal Holloway and Bedford New College, Egham, Surrey TW20 0EX, United Kingdom D. N. Brown C. L. Davis University of Louisville, Louisville, Kentucky 40292, USA K. E. Alwyn D. Bailey R. J. Barlow Y. M. Chia C. L. Edgar G. Jackson G. D. Lafferty T. J. West J. I. Yi University of Manchester, Manchester M13 9PL, United Kingdom J. Anderson C. Chen A. Jawahery D. A. Roberts G. Simi J. M. Tuggle University of Maryland, College Park, Maryland 20742, USA C. Dallapiccola X. Li E. Salvati S. Saremi University of Massachusetts, Amherst, Massachusetts 01003, USA R. Cowan D. Dujmic P. H. Fisher K. Koeneke G. Sciolla M. Spitznagel F. Taylor R. K. Yamamoto M. Zhao Massachusetts Institute of Technology, Laboratory for Nuclear Science, Cambridge, Massachusetts 02139, USA P. M. Patel S. H. Robertson McGill University, Montréal, Québec, Canada H3A 2T8 A. Lazzaroab V. Lombardoa F. Palomboab INFN Sezione di Milanoa; Dipartimento di Fisica, Università di Milanob, I-20133 Milano, Italy J. M. Bauer L. Cremaldi V. Eschenburg R. Godang Now at University of South Alabama, Mobile, Alabama 36688, USA R. Kroeger D. A. Sanders D. J. Summers H. W. Zhao University of Mississippi, University, Mississippi 38677, USA M. Simard P. Taras F. B. Viaud Université de Montréal, Physique des Particules, Montréal, Québec, Canada H3C 3J7 H. Nicholson Mount Holyoke College, South Hadley, Massachusetts 01075, USA G. De Nardoab L. Listaa D. Monorchioab G. Onoratoab C. Sciaccaab INFN Sezione di Napolia; Dipartimento di Scienze Fisiche, Università di Napoli Federico IIb, I-80126 Napoli, Italy G. Raven H. L. Snoek NIKHEF, National Institute for Nuclear Physics and High Energy Physics, NL-1009 DB Amsterdam, The Netherlands C. P. Jessop K. J. Knoepfel J. M. LoSecco W. F. Wang University of Notre Dame, Notre Dame, Indiana 46556, USA G. Benelli L. A. Corwin K. Honscheid H. Kagan R. Kass J. P. Morris A. M. Rahimi J. J. Regensburger S. J. Sekula Q. K. Wong Ohio State University, Columbus, Ohio 43210, USA N. L. Blount J. Brau R. Frey O. Igonkina J. A. Kolb M. Lu R. Rahmat N. B. Sinev D. Strom J. Strube E. Torrence University of Oregon, Eugene, Oregon 97403, USA G. Castelliab N. Gagliardiab M. Margoniab M. Morandina M. Posoccoa M. Rotondoa F. Simonettoab R. Stroiliab C. Vociab INFN Sezione di Padovaa; Dipartimento di Fisica, Università di Padovab, I-35131 Padova, Italy P. del Amo Sanchez E. Ben-Haim H. Briand G. Calderini J. Chauveau P. David L. Del Buono O. Hamon Ph. Leruste J. Ocariz A. Perez J. Prendki S. Sitt Laboratoire de Physique Nucléaire et de Hautes Energies, IN2P3/CNRS, Université Pierre et Marie Curie-Paris6, Université Denis Diderot- Paris7, F-75252 Paris, France L. Gladney University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA M. Biasiniab R. Covarelliab E. Manoniab INFN Sezione di Perugiaa; Dipartimento di Fisica, Università di Perugiab, I-06100 Perugia, Italy C. Angeliniab G. Batignaniab S. Bettariniab M. Carpinelliab Also with Università di Sassari, Sassari, Italy A. Cervelliab F. Fortiab M. A. Giorgiab A. Lusianiac G. Marchioriab M. Morgantiab N. Neriab E. Paoloniab G. Rizzoab J. J. Walsha INFN Sezione di Pisaa; Dipartimento di Fisica, Università di Pisab; Scuola Normale Superiore di Pisac, I-56127 Pisa, Italy D. Lopes Pegna C. Lu J. Olsen A. J. S. Smith A. V. Telnov Princeton University, Princeton, New Jersey 08544, USA F. Anullia E. Baracchiniab G. Cavotoa D. del Reab E. Di Marcoab R. Facciniab F. Ferrarottoa F. Ferroniab M. Gasperoab P. D. Jacksona L. Li Gioia M. A. Mazzonia S. Morgantia G. Pireddaa F. Polciab F. Rengaab C. Voenaa INFN Sezione di Romaa; Dipartimento di Fisica, Università di Roma La Sapienzab, I-00185 Roma, Italy M. Ebert T. Hartmann H. Schröder R. Waldi Universität Rostock, D-18051 Rostock, Germany T. Adye B. Franek E. O. Olaiya F. F. Wilson Rutherford Appleton Laboratory, Chilton, Didcot, Oxon, OX11 0QX, United Kingdom S. Emery M. Escalier L. Esteve S. F. Ganzhur G. Hamel de Monchenault W. Kozanecki G. Vasseur Ch. Yèche M. Zito DSM/Irfu, CEA/Saclay, F-91191 Gif-sur-Yvette Cedex, France X. R. Chen H. Liu W. Park M. V. Purohit R. M. White J. R. Wilson University of South Carolina, Columbia, South Carolina 29208, USA M. T. Allen D. Aston R. Bartoldus P. Bechtle J. F. Benitez R. Cenci J. P. Coleman M. R. Convery J. C. Dingfelder J. Dorfan G. P. Dubois-Felsmann W. Dunwoodie R. C. Field A. M. Gabareen S. J. Gowdy M. T. Graham P. Grenier C. Hast W. R. Innes J. Kaminski M. H. Kelsey H. Kim P. Kim M. L. Kocian D. W. G. S. Leith S. Li B. Lindquist S. Luitz V. Luth H. L. Lynch D. B. MacFarlane H. Marsiske R. Messner D. R. Muller H. Neal S. Nelson C. P. O’Grady I. Ofte A. Perazzo M. Perl B. N. Ratcliff A. Roodman A. A. Salnikov R. H. Schindler J. Schwiening A. Snyder D. Su M. K. Sullivan K. Suzuki S. K. Swain J. M. Thompson J. Va’vra A. P. Wagner M. Weaver C. A. West W. J. Wisniewski M. Wittgen D. H. Wright H. W. Wulsin A. K. Yarritu K. Yi C. C. Young V. Ziegler Stanford Linear Accelerator Center, Stanford, California 94309, USA P. R. Burchat A. J. Edwards S. A. Majewski T. S. Miyashita B. A. Petersen L. Wilden Stanford University, Stanford, California 94305-4060, USA S. Ahmed M. S. Alam J. A. Ernst B. Pan M. A. Saeed S. B. Zain State University of New York, Albany, New York 12222, USA S. M. Spanier B. J. Wogsland University of Tennessee, Knoxville, Tennessee 37996, USA R. Eckmann J. L. Ritchie A. M. Ruland C. J. Schilling R. F. Schwitters University of Texas at Austin, Austin, Texas 78712, USA B. W. Drummond J. M. Izen X. C. Lou University of Texas at Dallas, Richardson, Texas 75083, USA F. Bianchiab D. Gambaab M. Pelliccioniab INFN Sezione di Torinoa; Dipartimento di Fisica Sperimentale, Università di Torinob, I-10125 Torino, Italy M. Bombenab L. Bosisioab C. Cartaroab G. Della Riccaab L. Lanceriab L. Vitaleab INFN Sezione di Triestea; Dipartimento di Fisica, Università di Triesteb, I-34127 Trieste, Italy V. Azzolini N. Lopez-March F. Martinez-Vidal D. A. Milanes A. Oyanguren IFIC, Universitat de Valencia- CSIC, E-46071 Valencia, Spain J. Albert Sw. Banerjee B. Bhuyan H. H. F. Choi K. Hamano R. Kowalewski M. J. Lewczuk I. M. Nugent J. M. Roney R. J. Sobie University of Victoria, Victoria, British Columbia, Canada V8W 3P6 T. J. Gershon P. F. Harrison J. Ilic T. E. Latham G. B. Mohanty Department of Physics, University of Warwick, Coventry CV4 7AL, United Kingdom H. R. Band X. Chen S. Dasu K. T. Flood Y. Pan M. Pierini R. Prepost C. O. Vuosalo S. L. Wu University of Wisconsin, Madison, Wisconsin 53706, USA ###### Abstract We present the results of searches for $B$ decays to charmless final states involving $\phi$, $f_{0}(980)$, and charged or neutral $\rho$ mesons. The data sample corresponds to $384{\times 10^{{6}}}$ $B\kern 1.79993pt\overline{\kern-1.79993ptB}{}$ pairs collected with the BABAR detector operating at the PEP-II asymmetric-energy $e^{+}e^{-}$ collider at SLAC. We find no significant signals and determine the following 90% confidence level upper limits on the branching fractions, including systematic uncertainties: ${\cal{B}}(B^{0}\rightarrow\phi\phi)<2.0{\times 10^{{-7}}}$, ${\cal{B}}(B^{+}\rightarrow\phi\rho^{+})<30{\times 10^{{-7}}}$, ${\cal{B}}(B^{0}\rightarrow\phi\rho^{0})<3.3{\times 10^{{-7}}}$, ${\cal{B}}[B^{0}\rightarrow\phi f_{0}(980)]\times{\cal{B}}[f_{0}(980)\rightarrow\pi^{+}\pi^{-}]<3.8{\times 10^{{-7}}}$, and ${\cal{B}}[B^{0}\rightarrow f_{0}(980)f_{0}(980)]\times{\cal{B}}[f_{0}(980)\rightarrow\pi^{+}\pi^{-}]\times{\cal{B}}[f_{0}(980)\rightarrow K^{+}K^{-}]<2.3{\times 10^{{-7}}}$. ###### pacs: 13.25.Hw, 12.15.Hh, 11.30.Er ††preprint: BABAR-PUB-08/23††preprint: SLAC-PUB-13317 We report the results of searches for the decays $B^{0}\rightarrow\phi\phi$, $\phi\rho^{0}$, $\phi f_{0}(980)$, $f_{0}(980)f_{0}(980)$, and $B^{\pm}\rightarrow\phi\rho^{\pm}$ fz980 using data collected with the BABAR detector. The $B^{0}\rightarrow\phi\phi$ decay is an OZI suppressed process with an expected branching fraction in the range (0.1 to 3) ${\times 10^{{-8}}}$ in the Standard Model (SM) shalom ; lu2 ; beneke . The decays $B^{0}\rightarrow\phi\rho^{0}$ and $B^{+}\rightarrow\phi\rho^{+}$ are pure $b\rightarrow d$ loop processes; the expected branching fractions for these modes range from ($2$ to $7$)$\times 10^{-8}$ zou ; lu ; li ; bao ; gronau . The presence of new physics (NP) would give rise to additional amplitudes that could enhance the branching fractions for these decay modes relative to the SM predictions shalom ; lu ; lu2 . The branching fraction for $B^{0}\rightarrow\phi\phi$ could be enhanced to $10^{-7}$ shalom , and the branching fractions for $B\rightarrow\phi\rho$ decays could be enhanced by 20% bao in the presence of NP. We are not aware of branching fraction predictions for $B^{0}\rightarrow\phi f_{0}$ and $B^{0}\rightarrow f_{0}f_{0}$. The $B$ decays to $\phi\phi$ and $\phi\rho$ are complicated by the presence of one amplitude with longitudinal polarization and two amplitudes with transverse polarization. The fraction of longitudinally polarized events is denoted by $f_{L}$. Integrating over the angle between the vector meson decay planes, the angular distribution $(1/\Gamma)d^{2}\Gamma/d\cos\theta_{1}d\cos\theta_{2}$ is $\displaystyle\frac{9}{4}\left[f_{L}\cos^{2}\theta_{1}\cos^{2}\theta_{2}+\frac{1}{4}(1-f_{L})\sin^{2}\theta_{1}\sin^{2}\theta_{2}\right],$ (1) where the indices $1,2$ label the two vector mesons in the final state, and the helicity angles $\theta_{1,2}$ are the angles between the direction opposite to that of the $B^{0}$ ($B^{+}$) and the $K^{+}$ or $\pi^{+}$ ($\pi^{0}$) momentum in the $\phi$ or $\rho^{0}$ ($\rho^{+}$) rest frame. We define the angles $\theta_{1,2}$ for $f_{0}$ mesons in an analogous way. The expected values of $f_{L}$ range from $0.6$ to 0.8 li ; lu ; lu2 ; beneke for $B^{0}\rightarrow\phi\phi$, $\phi\rho^{0}$, and $B^{\pm}\rightarrow\phi\rho^{\pm}$. The presence of NP could lead to enhancements of the transverse polarization amplitudes shalom ; lu ; lu2 . The current upper limit on the $B^{0}\rightarrow\phi\phi$ branching fraction, obtained from a data sample of 82 $\mbox{\,fb}^{-1}$, is $1.5{\times 10^{{-6}}}$ babarphiphi . The upper limits on $B^{0}\rightarrow\phi\rho^{0}$ and $B^{+}\rightarrow\phi\rho^{+}$, determined using 3.1 $\mbox{\,fb}^{-1}$ of data, are $1.3{\times 10^{{-5}}}$ and $1.6{\times 10^{{-5}}}$ cleovv , respectively. Using a data sample of 349 $\mbox{\,fb}^{-1}$, BABAR recently reported an upper limit of $1.6{\times 10^{{-7}}}$ for $B^{0}\rightarrow f_{0}f_{0}$ babarrho0rho0 . This last result relies on the assumption that the $f_{0}\rightarrow\pi^{+}\pi^{-}$ branching fraction is 100%. In this analysis, we make the complimentary assumption that one $f_{0}$ decays to $\pi^{+}\pi^{-}$ and the other to $K^{+}K^{-}$ and search for $B^{0}\rightarrow f_{0}f_{0}$ in a cleaner final state than Ref. babarrho0rho0 . All these limits correspond to a confidence level (C.L.) of 90%. The results presented here are based on an integrated luminosity of 349$\mathrm{fb}^{-1}$, corresponding to $(384\pm 4)$ million $B\kern 1.79993pt\overline{\kern-1.79993ptB}{}$ pairs. These data were recorded at the $\mathchar 28935\relax{(4S)}$ resonance with a center-of-mass (CM) energy $\sqrt{s}$ = 10.58 $\mathrm{\,Ge\kern-1.00006ptV}$. The BABAR detector is described in detail elsewhere babarnim , and is situated at the interaction region of the PEP-II asymmetric energy $e^{+}e^{-}$ collider located at the Stanford Linear Accelerator Center (SLAC). We use Monte Carlo (MC) simulated events generated using the GEANT4 based ref:geant BABAR simulation. Photons are reconstructed from localized deposits of energy greater than 50$\mathrm{\,Me\kern-1.00006ptV}$ in the electromagnetic calorimeter that are not associated with a charged track. We require $\gamma$ candidates to have a lateral shower profile ref:lat that is consistent with the expectation for photons. $\pi^{0}$ candidates are reconstructed from two $\gamma$ candidates with invariant mass $0.10<m_{\gamma\gamma}<0.16$ ${\mathrm{\,Ge\kern-1.00006ptV\\!/}c^{2}}$. We use information from the vertex detector, drift chamber and detector of internally reflected Cherenkov light to select charged tracks that are consistent with kaon or pion signatures in the detector s2bprd . We reconstruct $\phi$ ($\rho^{0}$) candidates from pairs of oppositely charged kaon (pion) candidates with invariant mass $0.99<m_{KK}<1.05{\mathrm{\,Ge\kern-1.00006ptV\\!/}c^{2}}$ ($0.55<m_{\pi\pi}<1.05{\mathrm{\,Ge\kern-1.00006ptV\\!/}c^{2}}$). For $\rho^{0}$ candidates we require the helicity angles to satisfy $\|\cos\theta_{i}\|<0.98$ since signal efficiency falls off near $\|\cos\theta_{i}\|=1$. Charged $\rho$ candidates are reconstructed from a charged track consistent with the pion signature and a $\pi^{0}$ candidate. The invariant mass $m_{\pi\pi^{0}}$ of the $\rho^{+}$ candidate is required to lie between 0.5 and 1.0 ${\mathrm{\,Ge\kern-1.00006ptV\\!/}c^{2}}$. We also require that the helicity angles satisfy $-0.8<\cos\theta_{i}<0.98$ as signal efficiency is asymmetric because of the $\pi^{0}$ meson, and falls off near $\cos\theta_{i}=\pm 1$, and background peaks near $-1$. We select $f_{0}$ candidates from two charged tracks that are both either consistent with the kaon or the pion signature in the detector. We apply the same selection criteria to $f_{0}\rightarrow\pi^{+}\pi^{-}$ candidates as for $\rho^{0}$ mesons. Similarly we apply the same selection criteria to $f_{0}\rightarrow K^{+}K^{-}$ candidates as for $\phi$ mesons as the minimum $m_{KK}$ we can reconstruct in the detector is 0.99 ${\mathrm{\,Ge\kern-1.00006ptV\\!/}c^{2}}$. We reconstruct signal $B$ candidates ($B_{\mathrm{rec}}$) from combinations of two $\phi$ mesons, one $\phi$ and one $\rho$ or $f_{0}$, and two $f_{0}$ mesons. The $f_{0}f_{0}$ mode is required to have one $f_{0}$ decaying into $\pi^{+}\pi^{-}$, and the other decaying into $K^{+}K^{-}$. We require the $f_{0}$ in $\phi f_{0}$ to decay into $\pi^{+}\pi^{-}$. We use two kinematic variables, $m_{\mathrm{ES}}$ and $\Delta E$, in order to isolate the signal: $m_{\mathrm{ES}}=\sqrt{(s/2+{\mathbf{p}}_{i}\cdot{\mathbf{p}}_{B})^{2}/E_{i}^{2}-{\mathbf{p}}_{B}^{2}}$ is the beam-energy substituted mass and $\mbox{$\Delta E$}=E_{B}^{}-\sqrt{s}/2$ is the difference between the $B$ candidate energy and the beam energy in the $e^{+}e^{-}$ CM frame. Here the $B_{\mathrm{rec}}$ momentum ${\mathbf{p}_{B}}$ and four-momentum of the initial state $(E_{i},{\mathbf{p}_{i}})$ are defined in the laboratory frame, and $E_{B}^{}$ is the $B_{\mathrm{rec}}$ energy in the $e^{+}e^{-}$ CM frame. The distribution of $m_{\mathrm{ES}}$ ($\Delta E$) peaks at the $B$ mass (near zero) for signal events and does not peak for background. We require $m_{\mathrm{ES}}>5.25{\mathrm{\,Ge\kern-1.00006ptV\\!/}c^{2}}$. For the $\phi\phi$ final state we require $\|\mbox{$\Delta E$}\|<0.15\mathrm{\,Ge\kern-1.00006ptV}$. To reduce background from non-signal $B$ meson decays we apply the more stringent cut of $-0.07<\mbox{$\Delta E$}<0.15$ $\mathrm{\,Ge\kern-1.00006ptV}$ for all other modes. The angle in CM frame between the thrust axis of the rest of the event (ROE) and that of the $B$ candidate is required to satisfy $\|\cos(\theta_{T\\!B,T\\!R})\|<0.8$ in order to reduce background from $e^{+}e^{-}$ $\rightarrow$ $q\overline{q}$ ($q=u,d,s,c$) continuum events. The variable $\|\cos(\theta_{T\\!B,T\\!R})\|$ is strongly peaked near 1 for $q\overline{q}$ events, whereas $B\overline{B}$ events are more isotropic because the $B$ mesons are produced close to the kinematic threshold. Additional separation between signal and continuum events is obtained by combining several kinematic and topological variables into a Fisher discriminant $\cal F$, which we use in the maximum-likelihood fit described below. The variables $\|\cos(\theta_{T\\!B,T\\!R})\|$, $\|{\rm\Delta}t\|/\sigma({\rm\Delta}t)$, $\|\cos(\theta_{B,Z})\|$, $\|\cos(\theta_{T\\!B,Z})\|$, and the output of a multivariate tagging algorithm tag are used as inputs to $\cal F$. The time interval $\Delta t$ is calculated from the measured separation distance $\Delta z$ between the decay vertices of $B_{\mathrm{rec}}$ and the other $B$ in the event ($B_{\mathrm{ROE}}$) along the beam axis ($z$). The vertex of $B_{\mathrm{rec}}$ is reconstructed from the tracks that come from the signal candidate; the vertex of $B_{\mathrm{ROE}}$ is reconstructed from tracks in the ROE, with constraints from the beam spot location and the $B_{\mathrm{rec}}$ momentum. The uncertainty on the measured value of $\Delta t$ is $\sigma({\rm\Delta}t)$. The variable $\theta_{B,Z}$ is the angle between the direction of $B_{\mathrm{rec}}$ and the $z$ axis in the CM frame. This variable follows a sine squared distribution for $B\overline{B}$ events, whereas it is almost uniform for $q\overline{q}$. The variable $\theta_{T\\!B,Z}$ is the angle between the $B$ thrust direction and the $z$ axis in the laboratory frame. The decay modes studied are classified into three groups according to the final state particles: (i) $B^{0}\rightarrow\phi\phi$, (ii) $B^{+}\rightarrow\phi\rho^{+}$, and (iii) $B^{0}\rightarrow\phi\rho^{0}$, $B^{0}\rightarrow\phi f_{0}$, and $B^{0}\rightarrow f_{0}f_{0}$. We find that 6% of events for the mode in group (ii) and 3% of events for the modes in group (iii) have more than one candidate that passes our selection criteria. For such events we retain the candidate with the smallest $\chi^{2}$ for the $B_{\mathrm{rec}}$ vertex for use in the fits described below. The numbers of selected candidates are given in Table 1. The dominant background for all modes comes from continuum events. The yield of this background component is determined from the fit to data. The dominant $B$ backgrounds for group (i) are $B^{0}\rightarrow\phi K^{0}$ and $f_{0}K^{0}$, which are estimated to contribute 1.4 and 0.6 events to the data, respectively. The $B$ backgrounds for group (ii) are events from $B$ decays to final states including charm and $B^{+}\rightarrow\phi K^{+}$. These are estimated to contribute 107 and 5.5 events to the data. The $B$ backgrounds for group (iii) are events from $B$ decays to final states including charm, $B^{0}$ decays to $\phi K^{0}$, $f_{0}K^{0}$, $\phi K^{0}_{2}(1430)$, and $B^{+}$ decays to $\phi K^{+}$ and $\phi K^{+}$ estimated to contribute 249, 25.9, 9.1, 2.3, 4.7, and 1.8 events to the data. The branching fractions for the $B$ backgrounds are taken from Ref. PDG2006 , except for $B^{0}\rightarrow f_{0}K^{0}$, which has not yet been measured, and $\phi\rho^{+}$ where we use the results obtained here. The current upper limit on the $B^{0}\rightarrow f_{0}K^{0}$ branching fraction is $4.3{\times 10^{{-6}}}$ and we assume a branching fraction of $(2\pm 2){\times 10^{{-6}}}$. We obtain yields for each mode from extended unbinned maximum likelihood (ML) fits with the input observables $m_{\mathrm{ES}}$, $\Delta E$, and $\cos\theta_{1,2}$. In addition, for all modes except $\phi\phi$, we include $m_{1,2}$ and $\cal F$ in the likelihood, where $m_{1,2}$ is $m_{\pi\pi}$ or $m_{KK}$ for the $\phi$, $\rho$ or $f_{0}$ candidates. A total of three fits are performed, one for each group of signal modes. We include event hypotheses for signal events and the aforementioned backgrounds in each of the fits. For each event $i$ and hypothesis $j$, the likelihood function is ${\cal L}=\frac{e^{-\left(\sum n_{j}\right)}}{N!}\prod_{i=1}^{N}\left[\sum_{j=1}^{N_{j}}n_{j}{\cal P}_{j}({\bf x}_{i})\right]\ ,$ where $N$ is the number of input events, $N_{j}$ is the number of hypotheses, $n_{j}$ is the number of events for hypothesis $j$ and ${\cal P}_{j}({\bf x}_{i})$ is the corresponding probability density function (PDF) evaluated for the observables ${\bf x}_{i}$ of the $i^{th}$ event. The correlations between input observables are small and are assumed to be negligible. Possible biases due to residual correlations are evaluated as described below. We compute the combined PDFs ${\cal P}_{j}({\bf x}_{i})$ as the product of PDFs for each of the input observables. These combined PDFs are used in the fit to the data. Table 1: Number of events $N$ in the data sample, signal yield $\cal{Y_{S}}$ (corrected for fit bias), fit bias, detection efficiency $\epsilon$, daughter branching fraction product ($\prod{\cal{B}}_{i}$), significance $\sigma$ (including additive systematic uncertainties, taken to be zero if the fitted yield is negative), measured branching fraction where the first error is statistical, and the second systematic (see text), and the 90% C.L. upper limit on this branching fraction (including systematic uncertainties). For $B$ decays to $\phi\phi$ and $\phi\rho$, two efficiencies are reported, one for longitudinally and one for transversely polarized events. The reported branching fractions for $\phi f_{0}$ and $f_{0}f_{0}$ are product branching fractions that are not corrected for the probability of $f_{0}$ decaying into $\pi^{+}\pi^{-}$ or $K^{+}K^{-}$. Group \| $N$ \| Mode \| $\cal{Y_{S}}$ \| Bias \| $\epsilon$(%) \| $\prod{\cal{B}}_{i}$(%) \| $\sigma$ \| ${\cal{B}}$( $\times 10^{{-7}}$) \| UL($\times 10^{-7}$) ---\|---\|---\|---\|---\|---\|---\|---\|---\|--- (i) \| 209 \| $\phi\phi$ \| $-1.5^{+3.7}_{-2.9}$ \| $-0.4\pm 0.2$ \| 40.4 [28.7] \| $24.3\pm 1.2$ \| $0.0$ \| $-0.4^{+1.2}_{-0.9}\pm 0.3$ \| $<$$2.0$ (ii) \| 3175 \| $\phi\rho^{+}$ \| $22.5^{+11.3}_{-9.7}$ \| $+2.3\pm 1.1$ \| 5.7 [9.8] \| $49.3\pm 0.6$ \| $2.2$ \| $15^{+7}_{-6}\pm 9$ \| $<$$30$ (iii) \| 3949 \| $\phi\rho^{0}$ \| $3.9^{+6.3}_{-4.4}$ \| $+0.8\pm 0.4$ \| 24.1 [26.5] \| $49.3\pm 0.6$ \| $1.0$ \| $0.9^{+1.3}_{-0.9}\pm 0.9$ \| $<$$3.3$ \| \| $\phi f_{0}$ \| $0.8^{+2.4}_{-1.4}$ \| $-1.7\pm 0.5$ \| 22.1 \| $\ldots$ \| $0.0$ \| $0.2^{+0.6}_{-0.3}\pm 0.3$ \| $<$$3.8$ \| \| $f_{0}f_{0}$ \| $-13.6^{+4.8}_{-3.5}$ \| $-1.8\pm 0.5$ \| 25.5 \| $\ldots$ \| $0.0$ \| $-1.4^{+0.5}_{-0.4}\pm 1.5$ \| $<$$2.3$ For $B$ decays to $\phi\phi$ and $\phi\rho$, the $m_{\mathrm{ES}}$ distribution is parametrized with the sum of a Gaussian and a Gaussian with a low-side exponential component. The $\Delta E$ distribution is described by the sum of two Gaussian distributions, and the $\cos\theta_{1,2}$ distributions are described by Eq. (1) multiplied by an acceptance function. The acceptance function is a polynomial for all $\cos\theta_{1,2}$, with the exception of the $\rho^{+}$ helicity angle distribution for longitudinally polarized $\phi\rho^{+}$, which uses a polynomial multiplied by the sigmoid function $1/(1+\exp[\alpha(\cos\theta_{1,2}+\beta)])$, where the parameters $\alpha$ and $\beta$ are determined from MC simulated data. For the $\phi\rho$ final states we use a Gaussian to describe the $\cal F$ distribution, and the sum of a relativistic Breit-Wigner (BW) with two Gaussians for $m_{1,2}$. The continuum background $m_{\mathrm{ES}}$ distribution is described by an ARGUS function ref:argus . We parameterize the continuum $\Delta E$ distribution using a second-order polynomial and use polynomials to describe $\cos\theta_{1,2}$. Where appropriate, we parameterize the $\cal F$ distributions for the continuum background using a Gaussian, and we parameterize the $m_{1,2}$ distributions using the sum of a BW and a polynomial. We use smoothed histograms of MC simulated data as the PDFs for all other signal and background modes. We generate $B^{0}\rightarrow\phi f_{0}$ assuming that the $\phi$ is longitudinally polarised, and we use phase space distributions for $B^{0}\rightarrow f_{0}f_{0}$. Before fitting the data, we validate the fitting procedure using the methods described in Ref. rhorhoprd . We determine a bias correction on our ability to correctly determine the signal yield using ensembles of simulated experiments generated from samples of MC simulated data for the signal and exclusive backgrounds and from the PDFs for the other backgrounds. Our results are summarized in Table 1 where we show the measured yield, fit bias, efficiency, and the product of daughter branching fractions for each decay mode. We compute the branching fractions from the fitted signal event yields corrected for the fit bias, reconstruction efficiency, daughter branching fractions, and the number of produced $B$ mesons, assuming equal production rates of charged and neutral $B$ pairs. As we do not know the value of $f_{L}$ for the $\phi\phi$ and $\phi\rho$ modes, we fit the data for different physically allowed values of $f_{L}$ in steps of 0.1. We find no evidence for any of the signal modes and calculate 90% C.L. branching fraction upper limits $x_{UL}$ such that $\int_{0}^{x_{UL}}L({\cal{Y_{S}}},f_{L})d{\cal{Y_{S}}}/\int_{0}^{+\infty}L({\cal{Y_{S}}},f_{L})d{\cal{Y_{S}}}=0.9$, where $L({\cal{Y_{S}}},f_{L})$ is the likelihood as a function of signal yield ${\cal{Y_{S}}}$ and $f_{L}$ multiplied by a uniform prior. We report the most conservative (largest) upper limits for each mode, for which $f_{L}=0.5$, 0.7, and 0.2 for groups (i), (ii), and (iii), respectively. The central values of the branching fractions given in Table 1 correspond to these values of $f_{L}$. Figure 1 shows the $m_{\mathrm{ES}}$ distributions in subsamples of the data where $\|\Delta E\|<0.05$ $\mathrm{\,Ge\kern-1.00006ptV}$ for $B^{+}\rightarrow\phi\rho^{+}$, and $\|\Delta E\|<0.025$ $\mathrm{\,Ge\kern-1.00006ptV}$ for all other modes. Figure 1: (color online) Signal-enhanced distributions of $m_{\mathrm{ES}}$ in data, with a projection of the fitted likelihood for (top) $B^{0}\rightarrow\phi\phi$, (middle) $B^{+}\rightarrow\phi\rho^{+}$, and (bottom) $B^{0}\rightarrow\phi\rho^{0}$, $B^{0}\rightarrow\phi f_{0}$, and $B^{0}\rightarrow f_{0}f_{0}$. The solid line represents the total PDF, the dotted line represents signal, and the dashed line represents the sum of continuum and $B$ backgrounds. We estimate the systematic uncertainty related to the parameterization of the PDF by varying each parameter by its estimated uncertainty, and by substituting smoothed histograms by un-smoothed ones. The total contribution of all variations in signal yields, when added in quadrature, gives an error between $0.2$ and $5.6$ events, depending on the mode. We account for possible differences between data and MC events from studies of a control sample of $B\rightarrow D\pi$ events, yielding an uncertainty of $0.1$ to $12.2$ events depending on the mode. The uncertainty from fit bias is taken to be half the correction listed in Table 1. Incorporating the statistical uncertainty of the bias has a negligible effect. The uncertainty on $B$-daughter branching fractions is in the range ($1.2$ to $4.9$)% PDG2006 . The modes in group (iii), $\phi\rho^{0}$, $\phi f_{0}$, and $f_{0}f_{0}$ have systematic uncertainties from the $f_{0}$ lineshape fzsyst of 0.2, 3.1, and 15.9 events, respectively. The mode $B^{+}\rightarrow\phi\rho^{+}$ has a fractional systematic uncertainty of 3.0% from the reconstruction efficiency of $\pi^{0}$ mesons. Other sources of systematic errors are track reconstruction efficiency [($2.4-3.2$)%], uncertainty on the number of $B$ meson pairs (1.1%), particle identification efficiency (3.5%), and differences between data and MC efficiencies related to the cut on the vertex $\chi^{2}$ (0.6%). Assuming isospin is conserved in $f_{0}\rightarrow hh$ decays, where $h=\pi,K$, we correct for factors of ${\cal{B}}(f_{0}\rightarrow hh)/{\cal{B}}(f_{0}\rightarrow h^{+}h^{-})$ , to obtain the product branching fraction upper limits of ${\cal{B}}(B^{0}\rightarrow\phi f_{0})\times{\cal{B}}(f_{0}\rightarrow\pi\pi)<5.7{\times 10^{{-7}}}$, and ${\cal{B}}(B^{0}\rightarrow f_{0}f_{0})\times{\cal{B}}(f_{0}\rightarrow\pi\pi)\times{\cal{B}}(f_{0}\rightarrow KK)<6.9{\times 10^{{-7}}}$ at 90% C.L. In summary we have performed searches for the decays $B^{0}\rightarrow\phi\phi$, $\phi\rho^{0}$, $\phi f_{0}$, $f_{0}f_{0}$, and $B^{\pm}\rightarrow\phi\rho^{\pm}$ and place upper limits on these modes. The upper limit on $B^{0}\rightarrow\phi\phi$ reported here can be used to constrain possible NP enhancements suggested in Ref. shalom . We are grateful for the excellent luminosity and machine conditions provided by our PEP-II colleagues, and for the substantial dedicated effort from the computing organizations that support BABAR. The collaborating institutions wish to thank SLAC for its support and kind hospitality. This work is supported by DOE and NSF (USA), NSERC (Canada), CEA and CNRS-IN2P3 (France), BMBF and DFG (Germany), INFN (Italy), FOM (The Netherlands), NFR (Norway), MES (Russia), MEC (Spain), and STFC (United Kingdom). Individuals have received support from the Marie Curie EIF (European Union) and the A. P. Sloan Foundation. ## References (1) Throughout this paper, charge conjugation states are implied, and when we refer to $f_{0}$, we mean specifically $f_{0}(980)$. * (2) S. Bar-Shalom, G. Eilam, and Y. D. Yang, Phys. Rev. D 67, 014007 (2003). * (3) C. D. Lu et al., Eur. Phys. Jour. C 41, 311 (2005). * (4) M. Beneke, J. Rohrer, and D. Yang, Nucl. Phys. B 774, 64 (2007). * (5) W. Zou and Z. Xiao, Phys. Rev. D 72, 094026 (2005). * (6) C. D. Lu et al., Chin. Phys. Lett. 23, 2684 (2006). * (7) J. Li et al., hep-ph/0607249. * (8) S. Bao, F. S. Su, Y.-L. Wu, and C. Zhuang, Phys. Rev. D 77, 095004 (2008). * (9) M. Gronau and J. Rosner, arXiv:0806.3584. * (10) BABAR Collaboration, B. Aubert et al., Phys. Rev. Lett. 93, 181806 (2004). * (11) CLEO Collaboration, T. Bergfeld et al., Phys. Rev. Lett. 81, 272 (1998). * (12) BABAR Collaboration, B. Aubert et al., Phys. Rev. Lett. 98, 111801 (2007). * (13) BABAR Collaboration, B. Aubert et al., Nucl. Instrum. Methods Phys. Res., Sect. A 479, 1 (2002). * (14) GEANT4 Collaboration, S. Agostinelli et al., Nucl. Instrum. Methods Phys. Res., Sect. A 506, 250 (2003). * (15) A. Drescher et al., Nucl. Instrum. Methods Phys. Res., Sect. A 237, 464 (1985). * (16) BABAR Collaboration, B. Aubert et al., Phys. Rev. D 66, 032003 (2002). * (17) BABAR Collaboration, B. Aubert et al., Phys. Rev. Lett. 99, 171803 (2007); Phys. Rev. Lett. 94, 161803 (2005). * (18) Particle Data Group, Y.-M. Yao et al., J. Phys. G33, 1 (2006) and web-based 2007 partial update for the 2008 edition. * (19) ARGUS Collaboration, H. Albrecht et al., Phys. Lett. 254, 288 (1991). * (20) BABAR Collaboration, B. Aubert et al., Phys. Rev. D 76, 052007 (2007). * (21) Belle Collaboration, K. Abe et al., Phys. Rev. D 75, 051101 (2007).	arxiv-papers	2008-07-24T17:58:06	2024-09-04T02:48:56.973057	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "The BaBar Collaboration: B. Aubert, et al", "submitter": "Adrian Bevan", "url": "https://arxiv.org/abs/0807.3935" }
0807.4031	# Half period theorem of binary black holes Masaru Siino111E-mail:msiino@th.phys.titech.ac.jp and Daisuke Ida ###### Abstract Merging event horizons of the binary black holes is investigated. While recent development of the numerical study of the binary black hole coalescence has shown that their apparent horizons can orbit for many periods, we study the orbital motion of the event horizon. We discuss how many periods their event horizons orbit before their coalescence. Then, we find that they soon merge into one and the black holes cannot orbit for a half period while the apparent horizons can orbit many times. Department of Physics, Tokyo Institute of Technology, Oh-Okayama 1-12-1, Megro-ku, Tokyo 152-8550, Japan Department of Physics, Gakushuin University, Tokyo 171-8588, Japan ## 1 Introduction Recently the black hole binary system is much paid attention to as a candidate for gravitational wave source. Indeed numerical studies of the binary black hole coalescence advances so that it predicts the time profile of the gravitational radiation. Usually in numerical simulation, the black hole formation is examined by determining an apparent horizon. This is because the existence of the apparent horizon strongly suggests that of the event horizon. The event horizon is meaningful for an asymptotic observer, since he can just observe the outside of the event horizon, but not the apparent horizon. Moreover the event horizon is gauge invariant concept, while apparent horizon, even its existence depends on timeslicing. Hence, we would like to discuss the event horizon for the binary black hole system. Some authors have made sure that two apparent horizons can orbit around each other in a case of the binary black hole coalescence[1][2]. Then we attempt to discuss whether this picture also holds for the event horizon. Reason why we suspect the behavior of event horizon differs from that of the apparent horizon is that, in some studies of black hole formation, it is insisted that the event horizon soon settles to a single sphere[5][4]. In other words, we speculate that there does not remain sufficient time for the two black holes to orbit around each other. It is known that the event horizon is generated by null geodesics without a future endpoint [3]. In particular, the event horizon is a null hypersurface which may not be smooth at the past endpoints of the null geodesic generators. The set of the past endpoints of the null geodesic generators of the event horizon will be called the crease set [6][4]. Since the crease set is a subset of the null hypersurface, it is an achronal set(any two points are not connected by a timelike curve) as illustrated in Fig.1 [8][4]. That the crease set is achronal roughly implies two event horizons coalesce in superluminal speed. Figure 1: An event horizon and its crease set. The crease set is achronal and roughly implies two event horizon coalesce in superluminal speed. As a rough estimation, let two black holes be located with separation $2R$. They will coalesce within the time-interval of order of $\Delta t=R/c$ while we cannot expect any circular motion with this time duration. Hence we expect there is any upper bound of a rotation angle of the binary black holes before their coalescence. In the present article, we assert the existence of the upper bound of the rotation angle and demonstrate a half period theorem that binary black holes cannot orbit for a half period in terms of their event horizons assuming the reflection symmetry with respect to the orbital surface. This assumption is required for technical reason to formulate the notion of the half period of the binary black holes without ambiguities. Clearly, it is meaningless to say that the binary black holes go around or half around in terms of a coordinate system. We should formulate the half period of the binary black holes without referring to a specific coordinate system. The outline of our argument is as follows. We first consider the foliation of the spacetime by a family of timelike curves and choose a reference timelike curve as a center of the binary motion. Next we determine the opposite direction, which we call the light ray opposite, of each black hole with respect to the center. Then, whether the black hole orbits for a half period from the initial configuration or not can be determined in terms of the light ray opposite. Finally, we show that each black hole cannot orbit for a half period irrespective of the choice of the foliation and the center. This gives a possible way to measure the amount of the binary motion in a coordinate-free way. We first explain more details of our scheme in the next section. In the third section we give several definitions and set up the situation. The main theorem is stated in the fourth section. In the last section, we give some remarks and implications. ## 2 schematic picture Now, we try schematic discussion based on the topological notion of event horizon to illustrate that a half period of the binary rotation typically bounds duration of binary rotating era. From Newtonian picture, one may think that the coalescence of binary black hole event horizons occurs after two black hole event horizons orbit around each other for several periods. A simple picture of the orbiting event horizons, however, makes us suspect them of merging before the several periods have elapsed. Here we simply consider two black holes with identical mass fated to coalesce. They are on a binary orbit shrinking by the energy loss caused by gravitational radiation. In a coordinate system $(t,x^{i})$ such that each $x^{i}={\rm const.}$ line is timelike, every comoving observer inside a black hole at some moment will stay within the black hole region, while the black hole move on the binary orbit. This implies that a Newtonian picture of binary motion as in Fig.2 is incorrect. We expect two possible scenarios for binary black hole coalescence, which we call the quasi-stationary scenario and the rapid coalescence scenario, as follows. The quasi-stationary scenario is what we expect to occur in the quasi-stationary binary motion of black holes. In the quasi-stationary scenario, the event horizons form a toroidal event horizon along the binary orbit in a half period of binary motion (See Fig.2). The rapid coalescence scenario is expected to occur when the binary black holes don’t have orbital angular momentum enough. In this case, two black holes coalesce without forming a toroidal horizon (See Fig.2). In both scenarios, binary black holes do not orbit for a half period before coalescence. In the above argument, the term “a half period” is loosely used. It is in general difficult to determine whether the binary black holes go around without specific coordinate system. The present work is our attempt to seek for a rigorous statement corresponding to the above schematic discussion. For this purpose, we should study it in a generally covariant manner incorporating a technique based on the causal studies of general relativity. Figure 2: The figures in the left side illustrate a Newtonian picture of binary motion. The others are two possible scenarios for binary black hole coalescence. The center is the quasi-stationary scenario and the right is the rapid coalescence scenario. ## 3 set up First of all, we have to consider the amount of orbital rotation of the binary black holes. This will be determined by introducing a global angular coordinate function. However, it is in general difficult due to the lack of the standard coordinate system. For this reason, we give up considering one radian to be an upper bound of orbital rotation angle, and attempt to put a mark corresponding to a half period in order to check orbiting of black holes. We ought to be able at least to discuss a half period by introducing a ‘straight’ curve passing the antipodal point, even without the angular coordinate. Nevertheless, each black hole does not always pass the opposite side of the straight curve in general cases. For simplicity in the present work, we assume reflection symmetry such that the plane of symmetry, which we call the orbital surface, intersects binary black holes. With the reflection symmetry, it would be enough to discuss the motion of the object only on the orbital surface. We define a light ray opposite as the mark of a half of the orbital period. Suppose a spacetime $(M,g)$ is globally hyperbolic. Then, $(M,g)$ admits timeslicing $\\{\Sigma(t)\\}$ in terms of a global time function $t:M\to{\bf R}$ and there will be a timelike vector field $T$ without zero points. Let the vector field $T$ be normalized such that $\langle T,dt\rangle=1$ holds. Let us consider a cylindrical region ${\mathscr{U}}$ generated by the vector field $T$ with a timelike side-boundary $B_{{\mathscr{U}}}$ as illustrated in figure 3. Then the timelike vector field $T$ determines a natural projection $\pi_{t}:{\mathscr{U}}\mapsto{\mathscr{U}}\cap\Sigma(t)$ of the closed subset ${\mathscr{U}}$ of $M$ into each timeslice $\Sigma(t)$ along the integrated curve of $T$. ###### Definition 1 (Comoving ball with the origin). Let $(M,g)$ be a globally hyperbolic space-time. Let $t:M\to{\bf R}$ be a global time function in $M$. The sequence $\\{\Sigma(t)\\}$ of the $t={\rm const.}$ hypersurface $\Sigma(t)$, $t\in[t_{i},t_{f}]$ will be called the timeslicing. The space-time $M$ will admit a future directed timelike vector field $T$ such that $\langle T,dt\rangle=1$ holds. Let $U\subset\Sigma(\overline{t})$ be a topological 3-ball embedded in $\Sigma(\overline{t})$, $\overline{t}\in(t_{i},t_{f})$. Let ${\mathscr{U}}$ be the closed subset of $M$ generated by $T$ such that ${\mathscr{U}}\cap\Sigma(\overline{t})=U$ holds. Let $B_{{\mathscr{U}}}\subset\dot{\mathscr{U}}$ be the closed subset of $\dot{\mathscr{U}}$, which is the product set $\dot{U}\times[t_{i},t_{f}]$ generated by $T$, and let $o:[t_{i},t_{f}]\to M;t\mapsto o(t)$ be the integral curve of $T$ which pass through an interior point $o(\overline{t})$ of $U$. The 4-tuple $(\\{\Sigma(t)\\},{\mathscr{U}},T,o)$ will be called the comoving ball with the origin in $(M,g)$. ∎ Figure 3: A cylinder defined by the 4-tuple $(\Sigma(t),{\mathscr{U}},T,o(t))$ is the comoving ball with the origin. It is difficult to give a precise notion of orbital plane of binary black holes or period of binary motion in general settings. A possible way to overcome this difficulty is to impose a reflection symmetry of $(M,g)$. Although this restriction might be rather stringent, we could still consider a large class of space-times describing binary black holes. In the rest of the present paper, we will assume that the space-time $(M,g)$ admits a reflection symmetry with respect to a timelike hypersurface ${\mathscr{O}}$ in $M$. The fixed point set ${\mathscr{O}}$ under this isometry will be called the orbital surface and the orbital surface at the time $t$ will be denoted as ${\mathscr{O}}_{t}={\mathscr{O}}\cap\Sigma(t)$. All the settings including the global time function $t$ and the comoving ball with the origin $(\\{\Sigma(t)\\},{\mathscr{U}},T,o)$ are taken as respecting the reflection symmetry. Here we introduce a notion of the opposite side of a point $p$ beyond the origin $o$, which we call the light ray opposite (abbreviated to LRO) of $p$, in terms of a null geodesic generator of $J^{+}(p)$ (see Fig. 4). ###### Definition 2 (The light ray opposite). Let $(M,g)$ be a reflection symmetric spacetime with respect to the timelike hypersurface ${\mathscr{O}}$. Let $(\\{\Sigma(t)\\},{\mathscr{U}},T,o)$ be a comoving ball with the origin in $M$ respecting the reflection symmetry. Let us call ${\mathscr{O}}_{t}={\mathscr{O}}\cap\Sigma(t)$ the orbital surface at the time $t$. For a point $p\in{\mathscr{O}}_{t_{1}}\cap{\mathscr{U}}$, $t_{1}\in(t_{i},t_{f})$, $\gamma_{p}$ is defined to be the geodesic generator of $\dot{J}^{+}(p)$ from $p$ which pass through $o$ at the time $t=t_{c}\in(t_{1},t_{f})$, if there is exactly one null geodesic generator of $\dot{J}^{+}(p)$ from $p$ through $o$. The null geodesic $\gamma_{p}$ will be within ${\mathscr{O}}$ due to the reflection symmetry. Then, for a point $p$ and a time $t_{2}\in(t_{1},t_{f})$, the light ray opposite (LRO) $\lambda(p,t_{2})$ of $p$ at the time $t=t_{2}$ is defined by $\displaystyle\lambda(p,t_{2})=\pi_{t_{2}}\left[\gamma_{p}\cap\bigcup_{t\in[t_{c},t_{f})}\Sigma(t)\right]$ where $\pi_{t_{2}}$ denotes the projection ${\mathscr{U}}\to\Sigma(t_{2})$ naturally defined by the timelike vector field $T$, and the LRO of $p$ is defined to be the empty set if $\gamma_{p}$ is not defined. ∎ ###### Remark 1. The LRO is the empty set when the origin is far from the reference point $p$ such that $o~{}\cap J^{+}(p)=\emptyset$. Besides there might possibly occur the situation where $o~{}\cap J^{+}(p)\neq\emptyset$ holds but $o~{}\cap\dot{J}^{+}(p)=\emptyset$, when the congruence of the light rays from $p$ to $o$ have caustics due to the local gravitational effects. We have just precluded such a possibility for simplicity. Figure 4: A light ray opposite $\lambda$ on $\Sigma(t_{2})$ with respect to $p$ is given by $T$-projection of $\gamma(t)$. Next, we introduce the notion of coalescing binary black holes as follows (Fig. 5). ###### Definition 3 (The binary black hole coalescence system). Let $(M,g)$ be a black hole spacetime and let $(\\{\Sigma(t)\\},{\mathscr{U}},T,o(t))$ be a comoving ball with the origin in $M$. A binary black hole coalescence system $(H,\\{\Sigma(t)\\})$ is defined to be the pair consiting of the event horizon $H$ in $M$ and a timeslicing $\\{\Sigma(t)\\}$, $t\in[t_{i},t_{f}]$, such that there is a coalescence time $t^{\prime}\in(t_{i},t_{f})$ decomposing $H\cap{\mathscr{U}}$ into a pre- coalescence part $H_{pr}=H\cap{\mathscr{U}}\cap\left[\bigcup_{t\in[t_{i},t^{\prime})}\Sigma(t)\right]$ which has a pair of connected components, and a post-coalescence part $H_{po}=H\cap{\mathscr{U}}\cap\left[\bigcup_{t\in(t^{\prime},t_{f}]}\Sigma(t)\right]$ which is connected, by the spatial hypersurface $\Sigma(t^{\prime})$. ∎ Figure 5: This is a binary black hole coalescence system which is composed of a disconnected pre-coalescence part and a connected post-coalescence part. It should be noted that the concept of the black hole coalescence depends on the choice of the timeslicing. In a different timeslicing the black hole coalescence system can always be regarded as the formation of a single black hole [4, 10]. Here we consider the typical case $S^{2}\sqcup S^{2}\to S^{2}$ for the transition of the horizon topology. Since the reflection symmetry of $M$ is imposed, it is enough to consider the causal structure of the orbital surface ${\mathscr{O}}$. In other words, we will concentrate on the section of the event horizon by ${\mathscr{O}}$. In the following, we assume that $H_{pr}\cap{\mathscr{O}}_{t}$ consist of a pair of circles $S^{1}\sqcup S^{1}$ and $H_{po}\cap{\mathscr{O}}_{t}$ consists of a single circle $S^{1}$ in the binary black hole coalescence system. We want to define the half cycle of the binary black hole system by saying that that a black hole go around a half cycle if all the infinitesimal area elements of the black hole go around a half. For this purpose, we need to specify each trajectory of a point on the black hole. At first sight, the null geodesic generators of the event horizon seems to naturally determine each orbit. However, this is not appropriate, since new null geodesic generators emerge incessantly. Instead, we consider arbitrarily chosen one parameter family of homeomorphisms, $\phi_{t}:S^{1}\to S^{1}$ between the $H_{pr}\cap{\mathscr{O}}_{t_{i}}$ and $H_{pr}\cap{\mathscr{O}}_{t}$, $(t\in[t_{i},t^{\prime}))$, continuous with respect to $t$. Such one parameter family of homeomorphisms $\phi_{t}$ determines the motion of the infinitesimal area element of the black hole. Note that each orbit determined by $\phi_{t}$ necessarily exceeds or equals the speed of light, for it lies on the null hypersurface. The following is the definition of a half period of binary black holes in an orbital surface, where each black hole event horizon moves as shown in figure6. ###### Definition 4 (A half period of the binary coalescence system). Let $(H,\\{\Sigma(t)\\})$ be a binary black hole coalescence system with the reflection symmetry with respect to the orbital surface ${\mathscr{O}}$ with the coalescence time $t^{\prime}\in(t_{i},t_{f})$. Let the pre-coalescence part $H_{pr}\cup{\mathscr{O}}$ in ${\mathscr{O}}$ consists of a disconnected sum $H_{pr}\cup{\mathscr{O}}=H_{\rm I}\sqcup H_{\rm II}$ such that each of $H_{\rm I}\cup{\mathscr{O}}_{t}$ and $H_{\rm II}\cup{\mathscr{O}}_{t}$, $t\in[t_{i},t^{\prime})$ is a circle. Let $\phi_{t}^{A}:H_{A}\cap{\mathscr{O}}_{t_{1}}\to H_{A}\cap{\mathscr{O}}_{t}$ ($A={\rm I}$, ${\rm II}$) be a one parameter family of homeomorphisms such that $\phi_{t_{1}}^{A}$ is the identity map of $H_{A}$ and $\phi_{t}^{A}$ is continuous with respect to $t$. For $t_{i}<t_{1}<t_{2}<t_{f}$, we say that a half period of the binary coalescence system has elapsed during $(t_{1},t_{2})$, if the following statement hold both for $A={\rm I}$ and ${\rm II}$; There is a point $p$ on $H_{A}\cap O_{t_{1}}$ such that every orbit of a point on $H_{A}\cup{\mathscr{O}}_{t_{1}}$ intersects the LRO $\lambda(p,t)$ of $p$ during $t_{1}<t<t_{2}$. ∎ Figure 6: The definition of a half period of the binary coalescence system. Note that whether the half period has elapsed or not does not depend on the choice of the correspondences $\\{\phi^{\rm I}_{t},\phi^{\rm II}_{t}\\}$. ## 4 The half period theorem Now we show that a half period of the binary coalescence system does not elapse before the coalescence. Firstly, it is easily seen that two black holes before the coalescence are causally separated with each other as illustrated in figure 7. ###### Proposition 1. Let ${\mathscr{B}}$ be the black hole region in the binary black hole system with reflection symmetry, and let ${\mathscr{O}}_{pr}$ be the part of ${\mathscr{O}}\cap{\mathscr{U}}$ before the coalescence time defined by ${\mathscr{O}}_{pr}=\bigcup_{t\in[t_{i},t^{\prime})}{\mathscr{O}}(t)\cap{\mathscr{U}}$, so that the black hole region $B={\mathscr{B}}\cap{\mathscr{O}}_{pr}$ in the orbital surface is composed of two black hole regions $B_{\rm I}\simeq D^{2}\times[t_{i},t^{\prime})$ and $B_{\rm II}\simeq D^{2}\times[t_{i},t^{\prime})$ without intersection, where $D^{2}$ denotes the closed 2-disk. Then, $\displaystyle\forall p_{\rm I}\in B_{\rm I},\forall p_{\rm II}\in B_{\rm II},\quad{\mathscr{O}}_{pr}\cap J^{+}(p_{\rm I})\cap J^{+}(p_{\rm II})=\emptyset$ holds. ###### Proof. Since $p_{A}$ ($A={\rm I}$, ${\rm II}$) is a point in $B_{A}$, and $B_{A}$ is a future set in ${\mathscr{O}}_{pr}$, $J^{+}(p_{A})\cap{\mathscr{O}}_{pr}\subset B_{A}$ holds. Then, $B_{\rm I}\cap B_{\rm II}=\emptyset$ implies $(J^{+}(p_{\rm I})\cap{\mathscr{O}}_{pr})\cap(J^{+}(p_{\rm II})\cap{\mathscr{O}}_{pr})=\emptyset$. ∎ Figure 7: Two black holes are causally separated. This proposition just reflects the fact that a black hole region is a future set. The following corollary immediately follows. ###### Corollary 1. A timelike curve $o$ in ${\mathscr{O}}$ does not intersect both with $J^{+}(p_{\rm I})\cap{\mathscr{O}}_{pr}$ and $J^{+}(p_{\rm II})\cap{\mathscr{O}}_{pr}$ for any $p_{\rm I}\in B_{\rm I}$ and $p_{\rm II}\in B_{\rm II}$. Next, we show the following lemma. ###### Lemma 1. For every pair of points $p_{\rm I}\in B_{\rm I}\cap{\mathscr{O}}_{t_{1}}$ and $p_{\rm II}\in B_{\rm II}\cap{\mathscr{O}}_{t_{1}}$, either of the following statements holds. 1. 1. The LRO $\lambda(p_{\rm I},t_{2})$ of $p_{\rm I}$ does not intersect with $J^{+}(p_{\rm I})$ for any $t_{2}\in(t_{1},t^{\prime})$. 2. 2. The LRO $\lambda(p_{\rm II},t_{2})$ of $p_{\rm II}$ does not intersect with $J^{+}(p_{\rm II})$ for any $t_{2}\in(t_{1},t^{\prime})$. ###### Proof. It follows from the Cor. 1 that, for $p_{\rm I}\in B_{\rm I}\cap{\mathscr{O}}_{t_{1}}$ and $p_{\rm II}\in B_{\rm II}\cap{\mathscr{O}}_{t_{1}}$, $o$ does not intersect both with $J^{+}(p_{\rm I})\cap{\mathscr{O}}_{t_{1}}$ and $J^{+}(p_{\rm II})\cap{\mathscr{O}}_{t_{1}}$. Assume $J^{+}(p_{\rm I})\cap{\mathscr{O}}_{t_{1}}$ does not intersect $o$. Let the LRO $\lambda(p_{\rm I},t_{2})$ have an intersection with $J^{+}(p_{\rm I})\cap{\mathscr{O}}_{pr}$, then $\lambda(p_{\rm I},t_{2})$ starts from $o(t_{2})$ and extends to the point $q$ on $\dot{J}^{+}(p_{\rm I})\cap{\mathscr{O}}_{pr}$. It follows from the definition of the LRO that $q$ is on the past directed timelike curve generated by $T$ starting from the point $p$ on $\dot{J}^{+}(p_{\rm I})$. By slightly deforming the causal curve from $p_{\rm I}$ to $p$ obtained by joining the null geodesic generator from $p_{\rm I}$ to $q$ and the timelike curve from $q$ to $p$, one can construct the timelike curve from $p_{\rm I}$ to $p$. It follows that there is an open neighborhood $U$ of $p$, such that $U$ is contained in the chronological future $I^{+}(p_{\rm I})$ of $p_{\rm I}$. This contradicts the fact that the neighborhood $U$ of the boundary point $p$ of $J^{+}(p_{\rm I})$ necessarily contains an exterior point of $J^{+}(p_{\rm I})$. ∎ Figure 8: The LRO of a point $p$, will not belong to the causal future of $p$ itself. We are now at the position to state the main theorem. ###### Theorem 1 (The half period theorem). If the half period of the binary coalescence system elapses during $(t_{1},t_{2})$, two black holes merge into a single black hole at a time $t^{\prime}\in(t_{1},t_{2})$. ###### Proof. First note that the restriction $H\cap{\mathscr{O}}$ of $H$ on ${\mathscr{O}}$ is a null hypersurface in ${\mathscr{O}}$ generated by null geodesics, each without a future end point. By the definition of the half period of the binary coalescence system, there is a point $p$ on $H_{\rm I}\cap{\mathscr{O}}_{t_{1}}$ such that the trajectory of its LRO $\bigcup_{t\in(t_{1},t_{2})}\lambda(p,t)$ intersects with the orbit $\bigcup_{t\in(t_{1},t_{2})}\phi_{t}^{I}(p^{\prime})$ of every point $p^{\prime}$ on $H_{\rm I}$. This implies that a null geodesic generator of $J^{+}(p)$, which is also a null geodesic generator of $H_{\rm I}\cap{\mathscr{O}}$ through $p$, intersects with its own LRO $\lambda(p,t_{p})$ at a time $t_{p}\in(t_{1},t_{2})$. In the same way, there is a point $q$ on $H_{\rm II}\cap{\mathscr{O}}_{t_{1}}$ such that a null geodesic generator of $J^{+}(q)$ intersects with the LRO $\lambda(q,t_{q})$ of $q$ at a time $t_{q}\in(t_{1},t_{2})$. That is impossible is an immediate consequence of the Lemma 1. ∎ ## 5 Discussion First, we comment on the general covariance of our result. The set up of the problem here might seem to depend on a specific coordinate system. In fact, we prepare a time function and a timelike vector field to define the LRO. This obviously corresponds to a specific choice of the time coordinate and $x^{i}={\rm const.}$ $(i=1,2,3)$ lines. Then, the LRO of a point $p$ is regarded as the line $x^{i}=x^{i}(s)$ on $t={\rm const.}$ surface obtained by projecting points on the future directed null geodesic $x^{\mu}=x^{\mu}(s)$ $(\mu=0,1,2,3)$ starting from $p$ into the $t={\rm const.}$ surface along the $x^{i}={\rm const.}$ lines. In this sense, the definition of the LRO, hence that of the half period of the binary black hole system, depends on the coordinate system chosen. Nevertheless, the half period theorem is formulated in a covariant manner in the sense that it holds for arbitrary choice of such ordinary coordinate system, where ordinary coordinate we mean is such that a $t={\rm const.}$ surface is a spacelike hypersurface and an $x^{i}={\rm const.}$ line is a timelike curve. There are at least a couple of shortcomings in the definition of the LRO. One is that the LRO of a point $p$ can be empty set when the congruence of the light rays from $p$ to the orbit $o$ of the origin has caustics before reaching $o$. Another is that the LRO might not be a sufficiently long curve, so that it cannot be used as a goal line of the halfway around. The statement of the half period theorem might sound extraordinary, since it seems to contradict an existence of the quasi-stationary phase of binary black hole system. However, we would like to emphasis that it states an ordinary thing that a comoving observer of a black hole does not exceed a speed of the light, and that it does not contradict the quasi-stationary motion of the binary black holes. Recently, there have appeared numerical results [1, 2] showing that binary black holes go around many times, which apparently contradicts our theorem. In these numerical computations, apparent horizons, not the event horizons, are searched in the numerical space-time. There are two possibilities explaining this apparent contradiction. The first one is the following. We expect that each apparent horizon is surrounded by the event horizon. Then, there is a possibility that a pair of apparent horizons going around each other is already enclosed by a single event horizon. In other words, while apparent horizons go around many times, their event horizons quickly merge into one. The second possibility is that the coordinate system used in the numerical simulation is superluminal, where $x^{i}={\rm const.}$ lines become space-like. We expect that at least the second possibility is correct. First, the apparent horizon, when regarded as a dynamical surface, become a spacelike hypersurface in the space-time. This means that the apparent horizon has a confinement property. Therefore, each comoving observer, once entered the trapped region, never exit unless the apparent horizon disappears. Hence, if the apparent horizons seem to go around many times, the coordinate system describing this will be a superluminal one. Finally, we speculate on implications of our theorem to astrophysical observation. We cannot directly observe the event horizon. What are observed are light rays which miss the event horizon. Let us consider the situation where we can directly observe shadows of a binary system due to the existence of bright background light. This will be possible at least in principle. If the binary system consists of ordinary dark stars other than black holes and we observe it in the orbital surface (or with the maximal inclination angle), we will see two shadows intersect many times each other. While if it consist of black holes, we just observe two shadows merging into one exactly once. ## References * [1] F. Pretorius, Phys. Rev. Lett. 95, 121101 (2005); M. Campanelli, C. O. Loousto, P. Marronetti and Y. Zlochower, Phys. Rev. Lett. 96 111101 (2006); J. G. Baker, J. Centrella, D. I. Choi, M. Koppitz and J. van Meter, Phys. Rev. Lett. 96 111102 (2006). * [2] F. Petrorius, arXiv:0710.1338 [gr-qc] * [3] S. W. Hawking and F. F. R. Ellis, The large scale structure of space-time (Cambridge university press, Cambridge, 1973) * [4] M. Siino Phys. Rev. D58 104016 (1998) * [5] T. Jacobson and S. Venkataramani, Class. Quantum Grav. 12 (1995) 1055 * [6] J. K. Beem and A. Królak, J. Math. Phys. 39 (1998) 6001–6010 * [7] S. A. Hughes, C. R. Keeton, P. Walker, K. Walsh, S. L. Shapiro and S. A. Teukolsky, Phys. Rev. D49 (1994) 4004, A. M. Abrahams, G. B. Cook, S. L. Shapiro and S. A. Teukolsk Phys. Rev. D49 (1994) 5153, P. Anninos, D. Bernstein, S, Brandt, J. Libson, J. Massó, E. Seidel, L. Smarr, W. Suen, and P. Walker, Phys. Rev. Lett. 74 (1995) 630 * [8] S. L. Shapiro, S. A. Teukolsky and J. Winicour Phys. Rev. D52 (1995) 6982, S. Husa and J. Winicour,Phys.Rev.D60 084019 (1999) * [9] R. M. Wald, General Relativity (The University of Chicago Press, Chicago and London, 1984) * [10] M. Siino Phys. Rev. D59 064006 (1999); M. Siino and T. Koike gr-qc/0405056; D. Ida and M. Siino, Prog.Theor.Phys.118:715-727,2007	arxiv-papers	2008-07-25T08:51:02	2024-09-04T02:48:56.979644	{ "license": "Public Domain", "authors": "Masaru Siino and Daisuke Ida", "submitter": "Masaru Siino", "url": "https://arxiv.org/abs/0807.4031" }
0807.4043	# Scaling Relation between Sunyaev-Zel’dovich Effect and X-ray Luminosity and Scale-Free Evolution of Cosmic Baryon Field Qiang Yuan yuanq@mail.ihep.ac.cn Hao-Yi Wan Tong-Jie Zhang tjzhang@bnu.edu.cn Corresponding author. Ji-Ren Liu Long-Long Feng Li-Zhi Fang fanglz@physics.arizona.edu Department of Astronomy, Beijing Normal University, Beijing, 100875, P.R.China Key Laboratory of Particle Astrophysics, Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049, P.R.China Kavli Institute for Theoretical Physics China, Institute of Theoretical Physics, Chinese Academy of Sciences (KITPC/ITP-CAS), P.O.Box 2735, Beijing 100080, P.R. China Department of Physics, University of Arizona, Tucson, AZ 85721 Purple Mountain Observatory, Nanjing 210008, P.R. China National Astronomical Observatories, Chinese Academy of Science, Chao- Yang District, Beijing, 100012, P.R. China ###### Abstract It has been revealed recently that, in the scale free range, i.e. from the scale of the onset of nonlinear evolution to the scale of dissipation, the velocity and mass density fields of cosmic baryon fluid are extremely well described by the self-similar log-Poisson hierarchy. As a consequence of this evolution, the relations among various physical quantities of cosmic baryon fluid should be scale invariant, if the physical quantities are measured in cells on scales larger than the dissipation scale, regardless the baryon fluid is in virialized dark halo, or in pre-virialized state. We examine this property with the relation between the Compton parameter of the thermal Sunyaev-Zel’dovich effect, $y(r)$, and X-ray luminosity, $L_{\rm x}(r)$, where $r$ being the scale of regions in which $y$ and $L_{\rm x}$ are measured. According to the self-similar hierarchical scenario of nonlinear evolution, one should expect that 1.) in the $y(r)$-$L_{x}(r)$ relation, $y(r)=10^{A(r)}[L_{\rm x}(r)]^{\alpha(r)}$, the coefficients $A(r)$ and $\alpha(r)$ are scale-invariant; 2.) The relation $y(r)=10^{A(r)}[L_{\rm x}(r)]^{\alpha(r)}$ given by cells containing collapsed objects is also available for cells without collapsed objects, only if $r$ is larger than the dissipation scale. These two predictions are well established with a scale decomposition analysis of observed data, and a comparison of observed $y(r)$-$L_{x}(r)$ relation with hydrodynamic simulation samples. The implication of this result on the characteristic scales of non-gravitational heating is also addressed. ###### keywords: cosmology: theory , large-scale structure of universe , X-rays: galaxies: clusters , hydrodynamics , methods: numerical ###### PACS: 95.30.Jx , 07.05.Tp , 98.80.-k ## 1 Introduction Scaling relation of dimensional quantities is very powerful to reveal the dynamical feature of various physical systems. There has been a considerable effort devoting to study the correlations and scaling laws of various observable quantities of galaxy clusters. Since virialized self-gravitational system is characterized by one parameter, mass or virial temperature, one can find a set of scaling relations among mass, size, X-ray luminosity, temperature, and Compton parameter of Sunyaev-Zel’dovich (SZ) effect if the velocity and mass density fields of baryon fluid in clusters are assumed to be similar to the virialized dark matter halos (Kaiser, 1986). Observed data of galaxy clusters did yield scaling relations (Edge & Stewart, 1991; David et al., 1993; Wu et al., 1999; Helsdon & Ponman, 2000; Xue & Wu, 2000; Croston et al., 2005). However, observed scaling relations generally do not support the predictions given by the baryon-dark matter similarity of virialized dark halos (Helsdon & Ponman, 2000; Lloyd-Davies et al., 2000). Since Newtonian gravity is scale-free, the self gravitational system of collisionless dark matter shows scaling behavior if the power spectrum of initial density perturbations is scale-free. These scaling is regardless of whether the underlying gravitational field is virialized (Peebles, 1980). Thus, if the velocity and mass density fields of cosmic baryon matter are given by a similar mapping of the fields of dark matter, one may expect the scaling relations of clusters. However, the similar mapping assumption is correct only in linear regime (Bi et al., 1992), but is baseless in nonlinear regime (Shandarin & Zeldovich, 1989). The nonlinear evolution of cosmic baryon fluid leads to statistically decouple of the fluid from dark matter. The statistical properties of the velocity and mass density fields of baryon fluid do show deviation from the underlying dark matter field (Pando et al., 2004; He et al., 2005; Kim et al., 2005). Nevertheless, it has been pointed out by Shandarin & Zeldovich (1989): the dynamics of cosmic baryon fluid in the expanding universe is scale-free, i.e. no preferred special scales can be identified in the range from the onset of nonlinear evolution down to the length scale of dissipation. It likes fully developed turbulence in inertial range. This idea recently received substantial developments. With the hydrodynamic simulation sample of the concordance $\Lambda$CDM model, the velocity field of cosmic baryon fluid is found to be extremely well described by She-Leveque’s (SL) scaling formula (She & Leveque, 1994) in the “inertial range” (He et al., 2006). The SL formula is considered to be the basic statistical features of the scale-free evolution of fully developed turbulence. Moreover, the SL formula comes from self-similar log-Poisson hierarchy, which is related to the hidden symmetry of the Navier-Stokes equations (Dubrulle, 1994; She & Waymire, 1995). Very recently, it has been shown that the clustering of the mass density field of the cosmic baryon fluid can indeed be well described by a log-Poisson hierarchical cascade (Liu & Fang, 2008). All the scaling relations and non- Gaussian features predicted from the log-Poisson hierarchy are in very good agreement with the hydrodynamic simulation samples. These results indicate that, in the scale-free range, the nonlinear evolution of cosmic baryon fluid reaches a statistically quasi-steady state similar to a fully developed turbulence. For turbulence of incompressible fluid, the fluid undergoes a self-similar hierarchical evolution from largest to the smallest eddies and finally dissipates into thermal motion. For cosmic baryon fluid, the clustering on different scales can also be described by a self-similar hierarchy, and the fluid finally falls and dissipates into thermal motion. This scenario motives us to investigate the scaling relations of clusters from the self-similar hierarchy of cosmic baryon fluid. If the observed scaling relations come from the self-similar hierarchy, one can expect that 1.) the relations of dimensional quantities should be scale-free, i.e. all the scale- dependent coefficients of the scaling relations are scale-invariant; 2.) the relations should be held only if the scales of considered regions are larger than Jeans length, regardless of whether the underlying gravitational field is virialized, i.e. the relations given by cells containing collapsed objects is also available for cells without collapsed objects, only if the scale of cells is larger than the dissipation scale. Other relevant motivation comes from the non-gravitational heating of baryon gas of clusters. In order to solve the deviation from the similarity of virialized dark halos, various models of non-gravitational heating and cooling of baryonic gas have been proposed (e.g. Valageas & Silk, 1999; Tozzi & Norman, 2001; Voit et al., 2002; Zhang & Pen, 2003; Xue & Wu, 2003; Nagai et al., 2007). Since these cooling and heating may introduce characteristic scales, the self-similar hierarchy will no longer work on these characteristic scales. Therefore, it would be worth to detect the scale on which the above- mentioned two predictions to be broken. We study these properties with the relation between the Compton parameter $y$ of SZ effect and X-ray luminosity $L_{\rm x}$. The thermal SZ effect is due to the inverse Compton scattering of cosmic microwave background (CMB) photons by hot electrons of baryon fluid. The Compton parameter $y$ depends on the pressure of electron gas (Zeldovich & Sunyaev, 1969; Sunyaev & Zeldovich, 1980). There are many works on the $y$-$L_{x}$ relation (e.g. da Silva et al., 2004; Maughan, 2007; Bonamente et al., 2007). We will, however, focus on the above-mentioned two points, which have not yet been addressed right now. The outline of this paper is as follows. §2 presents the scaling relations $y=10^{A(r)}L_{x}^{\alpha(r)}$ with observed samples, and shows that $A(r)$ and $\alpha(r)$ are scale-invariant. §3 describes the hydrodynamic cosmological simulation samples. The comparison of the scaling relations of simulation samples with observed results is presented in §4. The conclusions and discussion are given in §5. ## 2 $y$-$L_{\rm x}$ Scaling Relations from Observed Samples ### 2.1 Data To study the scale free properties, we should find the $y(r)$ \- $L_{\rm x}(r)$ relations, where $y(r)$ and $L_{\rm x}(r)$ are, respectively, the Compton parameter of SZ effect and X-ray luminosity measured from regions with spatial scale $r$. The data of X-ray luminosity of these clusters are taken from McCarthy et al. (2003) (Xray1) and Morandi et al. (2007) (Xray2). The X-ray luminosity from area on comoving scale $r$ is calculated by $L_{\rm x}(r)=\int_{0}^{\theta_{\rm r}}L_{x}(\theta)\theta{\rm d}\theta,$ (1) where $\theta_{\rm r}=r/[(1+z)d_{A}(z)]$ is the angular radius corresponding to the comoving scale $r$, and $d_{A}(z)$ is angular diameter distance. $L_{x}(\theta)$ is proportional to the X-ray surface brightness $S_{x}(\theta)$, which can be well fitted by $\beta$-model $S_{x}(\theta)=S_{X0}[1+(\theta/\theta_{\rm c})^{2}]^{(1-6\beta)/2}$ up to $\theta\sim 10$ arcmin. Similarly, The mean of $y$ within a region on comoving scale $r$ is given by $y(r)=\frac{2}{\theta_{\rm r}^{2}}\int_{0}^{\theta_{\rm r}}y(\theta)\theta{\rm d}\theta.$ (2) We will use the SZ effect data from Reese et al. (2002) (SZ1) and Bonamente et al. (2006) (SZ2). The former compiled SZ effects of 18 clusters of galaxies spanning the redshift range of $0.14<z<0.78$, and the later includes 38 clusters in the same redshift range. These data are on angular scales up to $\sim 2$ arcmin, of which the corresponded $r$ is on about the same scale as, or larger than, the Jeans length on redshift $\sim 0.5$. Moreover, the $\theta$-dependencies of $y(\theta)$ are well fitted by $\beta$-model $y(\theta)=y_{0}[1+(\theta/\theta_{\rm c})^{2}]^{(1-3\beta)/2}$. Therefore, it would be reasonable to use the $\beta$ model fitted $y(r)$ to study the $y(r)$ \- $L_{\rm x}(r)$ relation. We will check this point below. ### 2.2 Result Figure 1 plots the relation of $y(r)$ vs. $L_{\rm x}(r)$ on scales $r$=0.1, 0.2, 0.39, 0.78, 1.56 and 3.12$h^{-1}$ Mpc respectively. In this figure the SZ and X-ray data are taken from SZ1 and Xray1, respectively. The cluster A370 is excluded as it shows a 3-$\sigma$ discrepancy with the distance-redshift relation (Reese et al., 2002). Three clusters, Cl0016, A611 and A697, are also excluded due to lacking the data of X-ray luminosity, and after all, there are totally 14 clusters used in Figure 1. Figure 1: $y$-$L_{\rm x}$ relation of observational samples SZ1+Xray1 (see Table 1). The Compton parameter $y$ are given by average over areas on the comving sizes 0.10, 0.20, 0.39, 0.78, 1.56, and 3.12 $h^{-1}$ Mpc, respectively. The solid lines indicate the best-fitting for all observational samples. Table 1. $r$-dependence of $\alpha(r)$ and $A(r)$ --- $r$($h^{-1})$ Mpc \| $\alpha$ \| $A$ 0.10 \| 0.32$\pm$0.06 \| -4.21$\pm$0.11 0.20 \| 0.36$\pm$0.08 \| -4.50$\pm$0.17 0.39 \| 0.44$\pm$0.10 \| -4.92$\pm$0.23 0.78 \| 0.51$\pm$0.11 \| -5.35$\pm$0.27 1.56 \| 0.56$\pm$0.12 \| -5.74$\pm$0.30 3.12 \| 0.61$\pm$0.13 \| -6.12$\pm$0.33 We make a best-fitting of the $y(r)$-$L_{\rm x}(r)$ relation with a power law $y=10^{A(r)}L_{\rm x}^{\alpha(r)}$ for various scales $r$ as displayed in Figure 1. The coefficients $\alpha(r)$ and $A(r)$ are listed in Table 1. Both $\alpha(r)$ and $A(r)$ are significantly dependent on the scale $r$. The exponent $\alpha(r)$ increases with scale $r$, while amplitude $A(r)$ decreases with $r$. If the system is given by a similar mapping of virialized halos, the scaling relation should be $y\propto L_{x}^{3/4}$ (Cole & Kaiser, 1988), which means that $\alpha$ is scale-independent and equal to 0.75. Table 1 shows that the values of $\alpha$ on all scales are less than 0.75. Accordingly, the baryon fluid on those scales should dynamically deviate from a similar mapping of underlying virialized gravitational field of dark matter halos. To check the effect of the angular scales of $\sim 2$ arcmin (§2.1), we re-calculate the $y(r)$-$L_{\rm x}(r)$ relation with clusters having $d_{A}>$1000 Mpc, which with $r>0.5$ Mpc for angular scales 2 arcmin. We find that the coefficients $\alpha(r)$ and $A(r)$ are consistent with Table 1 within 1-$\sigma$ range. Figure 2: The scale-dependence of the coefficients $\alpha(L)$ and $A(r)$ for various observational samples: SZ1+Xray1 (solid); SZ1+Xray2 (dashed); SZ2+Xray1 (dotted); and SZ2+Xray2 (dot-dashed). Figure 2 shows $\alpha(r)$ and $A(r)$ as functions of $r$. Note, both $\alpha(r)$ and $A(r)$ can be well fitted, respectively by $a_{\alpha}+b_{\alpha}\log r$ and $a_{A}+b_{A}\log r$, and therefore, both $\alpha(r)$ and $A(r)$ are scale-invariant. The amplitude $10^{A(r)}$ actually is a power law of $r$. This result is consistent with the dynamics of self- similar hierarchy. To further test this result, we used other data sets of SZ effect and X-ray luminosity. The results are also shown in Figure 2. Although these data sets are generally different from each other, all results of $\alpha(r)$ and $A(r)$ can well be fitted by the straight line of $\log r$. The fitting parameters are listed in Table 2. They are the same within 1-$\sigma$ errors. It strongly supports the scenario of scale-free dynamics. Table 2. Fitting Results of $\alpha(r)$ and $A(r)$. --- sample11footnotemark: 1 \| $a_{\alpha}$ \| $b_{\alpha}$ \| $a_{A}$ \| $b_{A}$ SZ1+Xray1 \| 0.52$\pm$0.05 \| 0.20$\pm$0.07 \| -5.46$\pm$0.13 \| -1.27$\pm$0.17 SZ1+Xray2 \| 0.47$\pm$0.06 \| 0.21$\pm$0.08 \| -5.33$\pm$0.13 \| -1.24$\pm$0.17 SZ2+Xray1 \| 0.49$\pm$0.05 \| 0.20$\pm$0.07 \| -5.33$\pm$0.11 \| -1.19$\pm$0.15 SZ2+Xray2 \| 0.50$\pm$0.04 \| 0.22$\pm$0.06 \| -5.34$\pm$0.10 \| -1.22$\pm$0.13 111SZ1 and SZ2 from Reese et al. (2002) and Bonamente et al. (2006), Xray1 and Xrya2 from McCarthy et al. (2003) (Xray1) and Morandi et al. (2007), respectively . It should also be pointed out that the scales of $r\leq 0.39$ $h^{-1}$ Mpc are actually less than the Jeans length of baryon fluid. However, the coefficients $\alpha(r)$ and $A(r)$ are still following the self-similar straight lines. In addition, $\alpha(r)$ is seen to be less than the value 0.75 from virialized halos. It implies that the dynamics of baryon fluid on these scales seems still to be scale free. ## 3 SZ effect samples of cosmological hydrodynamical simulations ### 3.1 Simulation Before embarking on the numerical calculations, we give a brief summary on the baryon fluid when it is in the turbulence-like or self-similar hierarchical clustering. The dynamics of growth modes of clustering in an expanding universe is sketched by a stochastic force driven Burgers equation (Berera & Fang, 1994; Jones, 1999; Matarrese & Mohayaee, 2002). The turbulence-like behavior is due to Burgers’ turbulence, which will be developed when the Burgers Reynolds number is large (Polyakov, 1995; Boldyrev et al., 2004; Kraichnan, 1968; Lässig, 2000). A turbulent flow in incompressible fluid consists of vortexes, while the clustering of cosmic matter is irrotational, because the modes with vorticity of the perturbed mass density field do not grow. The Burgers’ turbulence made the initially random field to result in a collection of shocks and a smooth variation of the field between the shocks. For cosmic baryon fluid, the Burgers Reynolds number generally is larger in nonlinear regime (He et al., 2004). Therefore, the baryon fluid in nonlinear regime consists of a collection of shocks with various strengths in both high, moderate and even low density areas. The kinetic energy of fluid is dissipated due to the shocks on various scales. The kinetic energy of fluid is effectively converted into thermal energy (He et al., 2006). In this context, it is clear that the method of hydrodynamical simulations should be capable of capturing discontinuities, like shocks, and their effects on energy conversion precisely. We will use the Weightly Essentially NonOscillatory (WENO) algorithm, which is effective to capture shocks and other discontinuities of the baryon fluid, and to give precise value of the fluid field between the discontinuities (Feng et al., 2004). This algorithm has been tested with 1.) Shock tube; 2.) the Sedov- Taylor self-similar blast wave solution, or the Bertschinger’s similarity solution; 3.) Zeldovich pancake. It is effectively to produce the baryon mass density contour, baryon temperature contour around massive halos (Feng et al., 2004; He et al., 2004, 2005). Some other properties of this simulation algorithm can also be found in Feng et al. (2004) and He et al. (2004). The simulations are performed in a cubic box 1003 $h^{-3}$ Mpc3 with a 10243 grid, and the number of dark matter particles is 5123. The mass of the dark matter particle is $\sim 10^{9}$M⊙, which corresponds to a density resolution about $0.01$ times of the mean density of intergalactic medium (IGM). The grid size is then $100/1024\sim 0.10$ $h^{-1}$ Mpc. This scale is smaller than the Jeans length at $z\leq 1$ (Bi et al., 2003). Therefore, the sample is suitable to describe the baryon fluid from the dissipation scale to a few ten $h^{-1}$ Mpc. We use the concordance $\Lambda$CDM cosmology model with parameters $\Omega_{\rm m}$=0.27, $\Omega_{\rm b}$=0.044, $\Omega_{\Lambda}$=0.73, $h$=0.71, $\sigma_{8}$=0.84, and spectral index $n=1$ and the ratio of specific heats is $\gamma=5/3$. The transfer function is calculated using CMBFAST (Seljak & Zaldarriaga, 1996). The atomic processes including ionization, cooling and heating are modeled as the method in Theuns et al. (1998). We take a primordial composition of H and He ($X$=0.76, $Y$=0.24) and use an ionizing background model of Haardt & Madau (2001). The simulations start at the redshift $z=99$, and output the density, velocity and temperature fields at redshifts $z$=2, 1, 0.5, and 0. It is easy for hydrodynamic simulation with Eulerian variable to reach low-density regions. They are suitable for a uniform analysis of weakly as well as strongly clustered fields of baryon fluid. These samples have been successfully applied to reveal the self-similar hierarchical behavior of cosmic baryon fluid (Kim et al., 2005; He et al., 2006; Liu & Fang, 2008), to explain the transmitted flux of HI and HeII Ly$\alpha$ absorption of quasars (Liu et al., 2006), and to study the relations between X-ray luminosity and temperature of groups of galaxies (Zhang et al., 2006). As mentioned in §1, the self-similar hierarchical scenario predicts that the relation $y(r)=10^{A(r)}[L_{\rm x}(r)]^{\alpha(r)}$ given by areas containing collapsed objects should also be available for areas without collapsed objects, only if $r$ is larger than the dissipation scale. On the other hand, either the effects of heating by injecting hot gas or metal cooling are localized in massive halos. Their characteristic scales are less than the Jeans length. Therefore, one can expect that the $y-L$ relations given by observed samples should be the same as that given by simulation samples either with or without considering the localized heating and cooling processes. To consider the effect of the metal abundance, we also use a sample with metal cooling of Zhang et al. (2006). For this sample, The metal cooling and metal line emission is calculated by the phenomenological method: 1) assuming an uniform evolving metallicity $Z=0.3Z_{\odot}(t/t_{0})$, $t_{0}$ being the present universe age; 2) computing the cooling function using the table of Sutherland & Dopita (1993). ### 3.2 Samples When relativistic corrections is negligible, the Compton parameter $y$ of the thermal SZ effect along a line of sight, $l$, with an angular distance $\theta$ from the center of a cluster in the plane of the sky is given by $y(\theta)=\frac{k_{\rm B}\sigma_{\rm T}}{m_{\rm e}c^{2}}\int n_{\rm e}(l,\theta)T_{\rm e}(l,\theta){\rm d}l,$ (3) where $\sigma_{\rm T}$ is the cross section of the Thomson scattering; $n_{\rm e}$ and $T_{\rm e}$ are, respectively, the number density and temperature of hot electrons. Since H and He atoms are almost fully ionized, we take the electron density $n_{\rm e}=\rho/\mu_{\rm e}m_{\rm p}$ where $\rho$ is the density of baryon gas, $\mu_{\rm e}=2/(1+X)$ with a hydrogen abundance of $X=0.76$. Using simulated density and temperature fields, we calculate the parameter $y(\theta)$ with Eq.(3). The mean of parameter $y(r)$ on various scales can be obtained using the scaling function of the discrete wavelet transform (DWT) $y_{\bf j,l}=\frac{1}{\int\phi_{\bf j,l}({\bf x}){\rm d}{\bf x}}\int y({\bf x})\phi_{\bf j,l}({\bf x}){\rm d}{\bf x}.$ (4) where $\phi_{\bf j,l}({\bf x})$ is the scaling functions related to cell $({\bf j,l})$ with comoving size $100/2^{j}$ $h^{-1}$ Mpc and at position ${\bf l}=(l_{1},l_{2},l_{3})$. The details of the DWT analysis can be found in Fang & Thews (1998). We take the comoving size $100/2^{j}$ $h^{-1}$ Mpc with $j=$10, 9, 8, 7, 6 and 5, corresponding to scales 0.10, 0.20, 0.39, 0.78, 1.56 and 3.12 h-1 Mpc, which are the same as that used in the analysis of the observed samples in last section. The total X-ray luminosity of thermal bremsstrahlung emission from a cell $({\bf j,l})$ is calculated by the same way as Zhang et al. (2006) $(L_{\rm x})_{\bf j,l}=V_{j}\frac{1}{\int\phi_{j,l}({\bf x}){\rm d}{\bf x}}\int\epsilon^{ff}({\bf x})\phi_{j,l}({\bf x}){\rm d}{\bf x},$ (5) where $\epsilon^{ff}({\bf x})$ is the map of X-ray emissivity, and $V_{j}$ is the volume of cell $({\bf j,l})$. The DWT decomposition has following advantages. First, the set of scaling functions are orthogonal and complete, and therefore, the decomposition of Eqs.(3) and (4) does not give rise to false correlation. Second, the DWT cell $({\bf j,l})$ with high mass density can directly be used to identify clumpy structures. The DWT-identified cells on scale 1.5 $h^{-1}$ Mpc is statistically the same as clusters identified by traditional method, such as the friend-of-friend algorithm (Xu et al., 1998). Therefore, the DWT variables $y_{\bf j,l}$ and $(L_{\rm x})_{\bf j,l}$ provides a uniform description of the Compton parameter and X-ray luminosity of the whole field, regardless of the dynamical details of all cells $({\bf j,l})$. Third, the DWT variables can be applied directly to non-Gaussian behavior (He et al., 2006; Liu & Fang, 2008) We will use the Harr wavelet (Daubechies 2) to do the calculation below. We also repeat the calculations with wavelet Daubechies 4. The statistical features given by Daubechies 4 are basically the same as Haar wavelet. ## 4 Scaling Relations between the SZ Effect and X-ray Luminosity ### 4.1 $y$-$\rho_{\rm igm}$ and $L_{x}$-$\rho_{\rm igm}$ relations Figure 3 shows the relations between the mean Compton parameter $y(r)$ and mean mass density $\rho_{\rm igm}(r)$ of cells on scale $r$ for the simulation samples at redshift $z=0.5$. The comoving scales $r$ are taken to be 0.39, 0.78, 1.56 and 3.12 $h^{-1}$ Mpc respectively. Figure 3 has a dark area as a bottom envelop of the $y$-$\rho_{\rm igm}$ distribution. It gives a tight correlation between $y$ and $\rho_{\rm igm}$ and can be described approximately by a power law of $y\propto\rho_{\rm igm}^{1.8}$ for all scales $r$. This relation is basically consistent with the so-called adiabatic ‘equation of state’ $T\propto\rho^{2/3}$ if considering $y\propto\rho_{\rm igm}T$. Figure 3: Compton parameter $y$ vs. baryon density $\rho_{\rm igm}$ of simulation samples at redshift $z=0.5$, in which $y$ and $\rho_{\rm igm}$ are the mean over cells with comoving scales $r=0.39$, 0.78, 1.56, and 3.12 $h^{-1}$ Mpc, respectively. $\rho_{\rm igm}$ is in unit of mean mass density $\bar{\rho}_{\rm igm}$ of the field. In Figure 4, the similar analysis has been performed for the relations between the total X-ray luminosity $L_{x}(r)$ and mean mass density $\rho_{\rm igm}(r)$. Clearly, there is also a dark area as the bottom envelop of the $L_{X}$-$\rho_{\rm igm}$ distribution. The tight correlation of the dark area yields $L_{X}\propto\rho_{\rm igm}^{2.4}$ for all scales. This is expected if considering $L_{X}\propto\rho^{2}_{\rm igm}T^{1/2}$. Figure 4: X-ray luminosity $L_{x}$ vs. baryon density $\rho_{\rm igm}$ of simulation samples at redshift $z=0.5$. The $L_{x}$ is the total X-ray luminosity from cell with comoving scale $r=0.39$, 0.78, 1.56, and 3.12 $h^{-1}$ Mpc. $\rho_{\rm igm}$ are the mean of density within the cell. $\rho_{\rm igm}$ is in unit of mean mass density $\bar{\rho}_{\rm igm}$ of the whole field. Therefore, $y$ and $L_{X}$ given by the tight correlations of bottom envelops in Figures 3 and 4 imply $y\propto L^{0.75}$, which is the same as that given by a virialized underlying gravitational field. If we adopt the adiabatic ‘equation of state’ $T\propto\rho^{2/3}$, the $y$-$L_{\rm x}$ scaling relation should be $y\propto L_{x}^{0.71}$. The observed coefficient $\alpha$ are less than 0.75 on all scales. The reason for $\alpha<0.75$ is clearly shown by Figures 3 and 4. For a given $\rho_{\rm igm}$, the distribution of $y$ is significantly scattered from the bottom envelop. For sample at $z\simeq 0.5$, the data points of $y$ corresponding to $\rho_{\rm igm}\simeq 1$ scatter in the range from $\sim 10^{-11}$ to $10^{-8}$. The scattering is due to the heating of Burgers’ shocks, which leads to the baryon fluid to be multiphasic, i.e. the relation between temperature and mass density can not be described by one polytropic equation. The points of $y$ higher than the bottom envelop correspond to state with temperature higher than those given by the ‘equation of state’ $T\propto\rho^{2/3}$ (He et al., 2004; Kim et al., 2005). The multiphases of baryon fluid are seen for all scales. The scattering in the $y$-$\rho_{\rm igm}$ distribution is substantial in high mass density areas ($\rho_{\rm igm}>1$) as well as low mass density areas ($\rho_{\rm igm}<1$). The $L_{x}$-$\rho_{\rm igm}$ distribution (Figure 4) shows similar scattering. It can be understood that the Burgers’ shock heating is not only working in high density regions, but also in low density areas. It leads to the deviation of $\alpha$ from 0.75. ### 4.2 $y$-$L_{\rm x}$ scaling relation We now study the scaling relation between $y(r)$ and $L_{\rm x}(r)$. Figure 5 plots the distribution of the Compton parameter $y(r)$ against the X-ray luminosity $L_{\rm x}(r)$ for the simulation samples at $z\leq 0.5$, in which the comoving scales of the cell are taken to be 0.78, 1.56 and 3.12 $h^{-1}$ Mpc respectively. The observed points of $y(r)$-$L_{\rm x}(r)$ are taken from the samples SZ1+Xray1. Figure 5: The scaling relation between the Compton parameter $y(r)$ and X-ray luminosity $L_{\rm X}(r)$. The scales $r$ are 0.78, 1.56, and 3.12 $h^{-1}$ Mpc. The observed data points (squares with errorbars) are from SZ1+Xray1 and the simulation data points (dotted) are taken from sample of redshift $z\leq 0.5$. On the scale of 3.12 $h^{-1}$ Mpc, the simulation data can be fitted by observations very well. Using the simulation data with $\log[L_{X}(r)/10^{43}{\rm erg/s]}\geq 1$, we found that the best fitting scaling relation $y=10^{A}L_{x}^{\alpha}$ of simulation sample gives $\alpha=0.65\pm 0.03$ and $A=-6.31\pm 0.05$, which are the same as observed samples shown in Table 1. For the scale of 1.56 $h^{-1}$ Mpc, Figure 5 also shows the consistency between observed and simulated samples. Since $\alpha$ is less than 0.75, the scaling relation shown in Figure 5 is dominated by data points located above the bottom envelop of Figure 3. In other word, the $y$-$L_{\rm x}$ relations are mainly determined by the structures which are involved in the evolution of the Burgers turbulence. On scale 0.78 $h^{-1}$ Mpc, the simulation sample still shows the same trend as observed data, but there are fewer points with $\log[L_{X}(r)/10^{43}{\rm erg/s}]\geq 2$. This phenomenon is more serious on scale 0.39 $h^{-1}$ Mpc. In this case, the best fitting scaling relation yields $\alpha=0.65\pm 0.01$ and $A=-5.47\pm 0.01$, which deviates from the observed result $\alpha=0.44\pm 0.10$ and $A=-4.92\pm 0.23$ (Table 1). The deviation on scales $\leq 0.78$h-1Mpc is expected. Since the scale 0.78 $h^{-1}$ Mpc is typical of the so-called $R_{2500}$ of clusters, and it is less than the virialization radius of clusters. Within this scale range, dissipation and non-gravitational processes are involved. It is beyond the regime of scale-free evolution. Because the observed data contains clusters with redshifts higher than 0.5, we also made a comparison between observed data and simulation samples of $z\leq 1$. The results are displayed in Figure 6, which yields almost the same results as Figure 5. We also analyzed the simulation samples including metal cooling of Zhang et al. (2006). The result is given in Figure 7. It shows the effect of metal cooling does not change the feature of Figure 5. Although metal cooling may have a big effect on $L_{\rm x}$ for groups. But it will have the similar effect on $y$. The $y$-$L{\rm x}$ scaling relation still keeps self-similar in the inertia range. Figure 6: The scaling relation between the Compton parameter $y(r)$ and X-ray luminosity $L_{\rm X}(r)$. The scale $r$ are 0.78, 1.56, and 3.12 $h^{-1}$ Mpc respectively. The observed data points ( squares with errorbars) are from SZ1+Xray1. The simulation data points (dots) are taken from sample of redshift $z\leq 1$. Figure 7: The same as Figure 5, but with simulation sample in which the the metal cooling and metal line emission are considered. ### 4.3 Available range of the $y$-$L_{\rm x}$ scaling relation As emphasized above, the scaling relation between physical quantities of cosmic baryon fluid due to the Burgers turbulence should hold for the entire field. Those relations inferred from statistical analysis made in regions containing collapsed structures, like clusters, should also be applicable in regions without that structures. Actually, Figures 5 - 7 have already shown that the $y$-$L_{x}$ scaling relation works well for all regions containing strong X-ray emission $\log[L_{X}(r)/10^{43}{\rm erg/s}]\geq 1$. Figure 8 presents the $y$-$L_{x}$ scaling relation covering a much wider range of the X-ray luminosity $-3\leq\log[L_{X}(r)/10^{43}{\rm erg/s}]\leq 3$, from rich clusters to weakly clustered areas. It shows clearly that the scaling relation of Figure 1 is available in the weakly clustered areas. Figure 8: The scaling relation between the Compton parameter $y$ and X-ray luminosity $L_{\rm x}$ of simulation data at redshift $z=0$ (top), 0.5 (middle), and 1 (bottom) respectively. The scale $r$ are 0.78 (left), 1.56 (middle) and 3.12 (right) $h^{-1}$ Mpc respectively. The solid line is given by the fitting of simulation data. Table 3. Fitting Coefficients $\alpha$ and $A$ for Simulation Data --- range of$\log[L_{X}(r)]$ \| $\alpha$ \| $A$ 1 $<\log[L_{X}(r)/10^{43}{\rm erg/s}]<$ 3 \| 0.65$\pm$0.03 \| -6.31$\pm$0.05 0 $<\log[L_{X}(r)/10^{43}{\rm erg/s}]<$ 3 \| 0.63$\pm$0.01 \| -6.27$\pm$0.02 -1 $<\log[L_{X}(r)/10^{43}{\rm erg/s}]<$ 3 \| 0.65$\pm$0.01 \| -6.29$\pm$0.01 -2 $<\log[L_{X}(r)/10^{43}{\rm erg/s}]<$ 3 \| 0.65$\pm$0.01 \| -6.30$\pm$0.01 -3 $<\log[L_{X}(r)/10^{43}{\rm erg/s}]<$ 3 \| 0.59$\pm$0.00 \| -6.32$\pm$0.01 We did a fitting of $y$-$L_{x}$ scaling relation in the simulation sample on the scale of $r=3.12$ $h^{-1}$ Mpc and $z\leq 0.5$ with different ranges of $\log L_{X}(r)$. The best-fitting values of the coefficients $\alpha$ and $A$ for various ranges of $\log L_{X}(r)$ are listed in Table 3. Obviously, the coefficients $\alpha$ and $A$ are almost independent of the range of $\log L_{X}(r)$. The scaling relation, $y(r)=10^{A(r)}[L_{X}(r)]^{\alpha(r)}$ is very stable within $10^{39}<L_{x}<10^{46}$ erg s-1. If $y$ and $L_{x}$ are measured from regions on scales larger than dissipation scales, the $y$-$L_{x}$ scaling relation still holds regardless of the dynamical details on the dissipation scales. The wide range of $L_{\rm x}$ is consistent with the scenario of self-similar hierarchical evolution, which gives a unified description of the dynamics of clustering on various level in the scale-free range. Similar analysis was made in simulation samples on scale of $r=1.56h^{-1}$ Mpc and 0.78 $h^{-1}$ Mpc at redshifts $z=0$, 0.5 and 1 respectively. These scaling relations are also valid within the entire range of $10^{39}<L_{x}<10^{46}$ erg s-1 as well. Moreover, from Figure 8, we can see that the scaling relations basically are redshift-independent within the range $z\leq 1$. Along with the decreasing redshift, there are more data points with larger $\log L_{X}$. The $y$-$L_{x}$ correlation shown in Figure 8 can be seen as a tree with root at left-bottom corner, and tip at right-top corner. The formation of clustered objects leads to the growth of the tip of the tree along the direction given by the scaling relation $y(r)=10^{A(r)}[L_{x}(r)]^{\alpha(r)}$. ## 5 Discussion and Conclusions The scaling studied in this paper comes from the scale-free dynamics of cosmic baryon fluid in the expanding universe. That is, there is no preferred special scales can be identified in the range from the onset of nonlinear evolution down to the characteristic length of dissipation. The clustering evolution is described by self-similar hierarchy of baryon fluid. This self-similar hierarchy is different from the gravitational self-similarity of virialzed system, which is characterized by the mass of the system. On the other hand, the scaling from the scale-free dynamics is characterized by the scale range, on which physical quantities of the mass and velocity field of cosmic baryon fluid are measured. That is, the dimensional quantities measured in cells with size larger than the dissipation characteristic length should satisfy the same scaling relations, regardless whether the cells contains massive collapsed objects. The scaling will be broken if the scale is less than the characteristic length of dissipation. We demonstrated this scaling with the $y(r)$-$L_{\rm x}(r)$ relation of baryon field. The observed $y(r)$-$L_{\rm x}(r)$ relation can be well reproduced with the hydrodynamical simulation. The important point is that the fitting of $y(r)$-$L_{\rm x}(r)$ scaling is not based on the identified clusters in simulation samples, but using all cells on a given scale $r$, regardless whether the cell contains rich clusters with strong X-ray emission. In other words, the scaling relations $y(r)=10^{A(r)}[L_{\rm x}(r)]^{\alpha(r)}$ can be used to describe cells containing strong X-ray emission, $L_{\rm x}>10^{43}$ erg s-1, as well as very weak X-ray emission regions $L_{\rm x}\sim 10^{40}$ erg s-1. This result supports the self-similar hierarchical scenario of the clustering of baryon fluid. The observed $y(r)$-$L_{\rm x}(r)$ relations starts to show a deviation from simulation results on scales smaller than 0.78 h-1 Mpc, which would be the characteristic scale of the dissipation. We calculated the Jeans lengths for each objects used in our statistics. All of the Jeans lengths are found to be less than 0.8 h-1 Mpc. It is very well consistent with our result. Moreover, this result is also consistent with previous studies on the non-gravitational heating of clusters. The non-gravitational heating generally is considered to be due to the injection of hot gas and energy from SN and AGN. However, the significant effect of injecting hot gas and energy by SN is mostly arising from dwarf galaxies, i.e. on scales much smaller than clusters. The Ly$\alpha$ observations of protoclusters shows that the feedback of AGN is not strong enough to heat the gas of clusters within comoving size 0.5 h-1 Mpc (Adelberger et al., 2003). Therefore, the effect of the SN and AGN heating would be dramatic only on scales smaller than about 0.5 h-1 Mpc, but probably cannot heat gas within cells on scales $\geq 1$ h-1 Mpc to amount of the order of 1 keV per nucleon (Pen, 1999; Wu et al., 1999). Actually we find that scaling relations on scales $>0.78$h-1 Mpc are still held after subtracting the contributions from the central part ($<0.2$ h-1 Mpc) of clusters to both $L_{\rm x}$ and $y$, as done in Markevitch (1998); Maughan (2007). As an application, the scaling relations of $y(r)$-$L_{\rm x}(r)$ would be useful for estimating the contamination of SZ effect on CMB, which is important, especially, on small scales (Cao et al., 2006). Recent simulation has shown that the Planck project would be capable of probing $y$ on the order of $y=10^{-7}-10^{-8}$ (Dolag et al., 2005). It might give a direct test on the universal scaling relations of $y(r)$-$L_{\rm x}(r)$ given by the self- similar hierarchical evolution. Our simulation of the baryon fluid is within the Eulerian framework. One can also study the nature of intermittency of fluid with Lagrangian point of view. In this approach, the hierarchical clustering can be tracked with Lagrangian trajectories. It has been found that the intermittent scaling is related to the long time correlations in the particle acceleration, namely, the random forces driving the particle motion is long range correlated (Mordant et al., 2002). It would be interesting to investigate the scaling in the Lagrangian scheme. Acknowledgments. T.-J.Z. is supported by the Fellowship of the World Laboratory. H.-Y.W., T.-J.Z. and L.-L.F. acknowledge support from the National Science Foundation of China (grants 10473002, 10573036 and 10545002). T.-J.Z thanks Xiang-Ping Wu for his valuable discussion. This work was also partially supported by US NSF AST 05-07340. ## References * Adelberger et al. (2003) Adelberger, K. L., Steidel, C. C., Shapley, A. E., & Pettini, M. 2003, ApJ, 584, 45 * Berera & Fang (1994) Berera, A. & Fang, L.-Z. 1994, Physical Review Letters, 72, 458 * Bi et al. (2003) Bi, H., Fang, L.-Z., Feng, L., & Jing, Y. 2003, ApJ, 598, 1 * Bi et al. (1992) Bi, H. G., Boerner, G., & Chu, Y. 1992, A&A, 266, 1 * Boldyrev et al. (2004) Boldyrev, S., Linde, T., & Polyakov, A. 2004, Physical Review Letters, 93, 184503 * Boldyrev et al. (2002) Boldyrev, S., Nordlund, Å., & Padoan, P. 2002, Physical Review Letters, 89, 031102 * Bonamente et al. (2007) Bonamente, M., Joy, M., LaRoque, S., Carlstrom, J., Nagai, D., & Marrone, D. 2007, ArXiv e-prints, 708 * Bonamente et al. (2006) Bonamente, M., Joy, M. K., LaRoque, S. J., Carlstrom, J. E., Reese, E. D., & Dawson, K. S. 2006, ApJ, 647, 25 * Cao et al. (2006) Cao, L., Chu, Y.-Q., & Fang, L.-Z. 2006, MNRAS, 369, 645 * Cole & Kaiser (1988) Cole, S. & Kaiser, N. 1988, MNRAS, 233, 637 * Croston et al. (2005) Croston, J. H., Hardcastle, M. J., & Birkinshaw, M. 2005, MNRAS, 357, 279 * da Silva et al. (2004) da Silva, A. C., Kay, S. T., Liddle, A. R., & Thomas, P. A. 2004, MNRAS, 348, 1401 * David et al. (1993) David, L. P., Slyz, A., Jones, C., Forman, W., Vrtilek, S. D., & Arnaud, K. A. 1993, ApJ, 412, 479 * Dolag et al. (2005) Dolag, K., Hansen, F. K., Roncarelli, M., & Moscardini, L. 2005, MNRAS, 363, 29 * Dubrulle (1994) Dubrulle, B. 1994, Physical Review Letters, 73, 959 * Edge & Stewart (1991) Edge, A. C. & Stewart, G. C. 1991, MNRAS, 252, 428 * Fang & Thews (1998) Fang, L. Z. & Thews, R. 1998, Wavelet in Physics (World Scientific, Singapore) * Feng et al. (2003) Feng, L.-L., Pando, J., & Fang, L.-Z. 2003, ApJ, 587, 487 * Feng et al. (2004) Feng, L.-L., Shu, C.-W., & Zhang, M. 2004, ApJ, 612, 1 * Haardt & Madau (2001) Haardt, F. & Madau, P. 2001, in Clusters of Galaxies and the High Redshift Universe Observed in X-rays, ed. D. M. Neumann & J. T. V. Tran * He et al. (2004) He, P., Feng, L.-L., & Fang, L.-Z. 2004, ApJ, 612, 14 * He et al. (2005) —. 2005, ApJ, 623, 601 * He et al. (2006) He, P., Liu, J., Feng, L.-L., Shu, C.-W., & Fang, L.-Z. 2006, Physical Review Letters, 96, 051302 * Helsdon & Ponman (2000) Helsdon, S. F. & Ponman, T. J. 2000, MNRAS, 315, 356 * Jamkhedkar et al. (2005) Jamkhedkar, P., Feng, L.-L., Zheng, W., & Fang, L.-Z. 2005, ApJ, 633, 52 * Jamkhedkar et al. (2003) Jamkhedkar, P., Feng, L.-L., Zheng, W., Kirkman, D., Tytler, D., & Fang, L.-Z. 2003, MNRAS, 343, 1110 * Jamkhedkar et al. (2000) Jamkhedkar, P., Zhan, H., & Fang, L.-Z. 2000, ApJ, 543, L1 * Jones (1999) Jones, B. J. T. 1999, MNRAS, 307, 376 * Kaiser (1986) Kaiser, N. 1986, MNRAS, 222, 323 * Kim et al. (2005) Kim, B., He, P., Pando, J., Feng, L.-L., & Fang, L.-Z. 2005, ApJ, 625, 599 * Kraichnan (1968) Kraichnan, R. H. 1968, Physics of Fluids, 11, 265 * Lässig (2000) Lässig, M. 2000, Physical Review Letters, 84, 2618 * Liu et al. (2006) Liu, J., Jamkhedkar, P., Zheng, W., Feng, L.-L., & Fang, L.-Z. 2006, ApJ, 645, 861 * Liu & Fang (2008) Liu, J.-R. & Fang, L.-Z. 2008, ApJ, 672, 11 * Lloyd-Davies et al. (2000) Lloyd-Davies, E. J. and Ponman, T. J. & Cannon, D. B. 2000, MNRAS, 315, 689 * Markevitch (1998) Markevitch, M. 1998, ApJ, 504, 27 * Matarrese & Mohayaee (2002) Matarrese, S. & Mohayaee, R. 2002, MNRAS, 329, 37 * Maughan (2007) Maughan, B. J. 2007, ApJ, 668, 772 * McCarthy et al. (2003) McCarthy, I. G., Holder, G. P., Babul, A., & Balogh, M. L. 2003, ApJ, 591, 526 * Morandi et al. (2007) Morandi, A., Ettori, S., & Moscardini, L. 2007, MNRAS, 379, 518 * Mordant et al. (2002) Mordant, N., Delour, J., Léveque, E., Arnéodo, A., & Pinton, J.-F. 2002, Physical Review Letters, 89, 254502 * Nagai et al. (2007) Nagai, D., Kravtsov, A. V., & Vikhlinin, A. 2007, ApJ, 668, 1 * Padoan et al. (2003) Padoan, P., Boldyrev, S., Langer, W., & Nordlund, Å. 2003, ApJ, 583, 308 * Pando et al. (2004) Pando, J., Feng, L.-L., & Fang, L.-Z. 2004, ApJS, 154, 475 * Pando et al. (2002) Pando, J., Feng, L.-L., Jamkhedkar, P., Zheng, W., Kirkman, D., Tytler, D., & Fang, L.-Z. 2002, ApJ, 574, 575 * Peebles (1980) Peebles, P. J. E. 1980, The large-scale structure of the universe (Research supported by the National Science Foundation. Princeton, N.J., Princeton University Press, 1980. 435 p.) * Pen (1999) Pen, U.-L. 1999, ApJ, 510, L1 * Polyakov (1995) Polyakov, A. M. 1995, Physical Review E, 52, 6183 * Reese et al. (2002) Reese, E. D., Carlstrom, J. E., Joy, M., Mohr, J. J., Grego, L., & Holzapfel, W. L. 2002, ApJ, 581, 53 * Seljak & Zaldarriaga (1996) Seljak, U. & Zaldarriaga, M. 1996, ApJ, 469, 437 * Shandarin & Zeldovich (1989) Shandarin, S. F. & Zeldovich, Y. B. 1989, Reviews of Modern Physics, 61, 185 * She & Leveque (1994) She, Z.-S. & Leveque, E. 1994, Physical Review Letters, 72, 336 * She & Waymire (1995) She, Z.-S. & Waymire, E. C. 1995, Physical Review Letters, 74, 262 * Sunyaev & Zeldovich (1980) Sunyaev, R. A. & Zeldovich, I. B. 1980, ARA&A, 18, 537 * Sutherland & Dopita (1993) Sutherland, R. S. & Dopita, M. A. 1993, ApJS, 88, 253 * Theuns et al. (1998) Theuns, T., Leonard, A., Efstathiou, G., Pearce, F. R., & Thomas, P. A. 1998, MNRAS, 301, 478 * Tozzi & Norman (2001) Tozzi, P. & Norman, C. 2001, ApJ, 546, 63 * Valageas & Silk (1999) Valageas, P. & Silk, J. 1999, A&A, 347, 1 * Voit et al. (2002) Voit, G. M., Bryan, G. L., Balogh, M. L., & Bower, R. G. 2002, ApJ, 576, 601 * Wu et al. (1999) Wu, X.-P., Xue, Y.-J., & Fang, L.-Z. 1999, ApJ, 524, 22 * Xu et al. (1998) Xu, W., Fang, L.-Z., & Wu, X.-P. 1998, ApJ, 508, 472 * Xue & Wu (2000) Xue, Y.-J. & Wu, X.-P. 2000, ApJ, 538, 65 * Xue & Wu (2003) —. 2003, ApJ, 584, 34 * Zeldovich & Sunyaev (1969) Zeldovich, Y. B. & Sunyaev, R. A. 1969, Ap&SS, 4, 301 * Zhang & Pen (2003) Zhang, P. & Pen, U.-L. 2003, ApJ, 588, 704 * Zhang et al. (2006) Zhang, T.-J., Liu, J., Feng, L.-l., He, P., & Fang, L.-Z. 2006, ApJ, 642, 625	arxiv-papers	2008-07-25T09:51:47	2024-09-04T02:48:56.985760	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Qiang Yuan (1,2), Hao-Yi Wan (1), Tong-Jie Zhang (1,3,4), Ji-Ren Liu\n (4), Long-Long Feng (5,6), Li-Zhi Fang (4),((1) Department of Astronomy,\n Beijing Normal University, (2) Key Laboratory of Particle Astrophysics,\n Institute of High Energy Physics, Chinese Academy of Sciences, (3) Kavli\n Institute for Theoretical Physics China, Institute of Theoretical Physics,\n Chinese Academy of Sciences (KITPC/ITP-CAS), (4) Department of Physics,\n University of Arizona, Tucson, (5) Purple Mountain Observatory, (6) National\n Astronomical Observatories, Chinese Academy of Science)", "submitter": "Qiang Yuan", "url": "https://arxiv.org/abs/0807.4043" }
0807.4059	# Bound states of two-dimensional Schrödinger-Newton equations Joachim Stubbe EPFL, IMB-FSB, Station 8, CH-1015 Lausanne, Switzerland Joachim.Stubbe@epfl.ch (Date: 06th February 2008) ###### Abstract. We prove an existence and uniqueness result for ground states and for purely angular excitations of two-dimensional Schrödinger-Newton equations. From the minimization problem for ground states we obtain a sharp version of a logarithmic Hardy-Littlewood-Sobolev type inequality. ###### Key words and phrases: Schrödinger-Newton equations, nonlinear Schrödinger equation, rearrangement inequality, logarithmic Hardy-Littelwood-Sobolev inequality ###### 1991 Mathematics Subject Classification: 35Q55, 35Q40,47J10 I am are indebted to Marc Vuffray for providing us his numerical studies of the model. I also wish to thank Philippe Choquard for many helpful discussins during the preparation of this paper. ## 1\. Introduction We consider the Schrödinger-Newton system (1.1) $iu_{t}+\Delta u-\gamma Vu=0,\quad\Delta V=\|u\|^{2}$ in two space dimensions which is equivalent to the nonlinear Schrödinger equation (1.2) $iu_{t}+\Delta u-\frac{\gamma}{2\pi}(\ln(\|x\|)\|u\|^{2})u=0$ with nonlocal nonlinear potential $V(x)=\frac{1}{2\pi}(\ln\|x\|\|u\|^{2})(x,t)=\frac{1}{2\pi}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\,\|u(t,y)\|^{2}\;dy.$ We are interested in the existence of nonlinear bound states of the form (1.3) $u(t,x)=\phi_{\omega}(x)e^{-i\omega t}.$ The Schrödinger-Newton system (1.1) in three space dimensions has a long standing history. With $\gamma$ designating appropriate positive coupling constants it appeared first in 1954, then in 1976 and lastly in 1996 for describing the quantum mechanics of a Polaron at rest by S. J. Pekar [1], of an electron trapped in its own hole by Ph. Choquard [2] and of selfgravitating matter by R. Penrose [3]. The two-dimensional model is studied numerically in [4]. For the bound state problem there are rigorous results only for the three dimensional model. In [2] the existence of a unique ground state of the form (1.3) is shown by solving an appropriate minimization problem. This ground state solution $\phi_{\omega}(x),\omega<0$ is a positive spherically symmetric strictly decreasing function. In [5] the existence of infinitely many distinct spherically symmetric solutions is proven and, in [6], a proof for the existence of anisotropic bound states is claimed. So far, there are no results for the two-dimensional model using variational methods. One mathematical difficulty of the two-dimensional problem is that the Coulomb potential in two space dimensions is neither bounded from above nor from below and hence does not define a positive definite quadratic form. However, recently in [7] the existence of a unique positive spherically symmetric stationary solution $(u,V)$ of (1.1) such that $V(0)=0$ has been proven by applying a shooting method to the corresponding system of ordinary differential equations. In the present paper, we are mainly interested in the ground states of the model (1.4) $u(t,x)=\phi_{\omega}(x)e^{-i\omega t},\quad\phi_{\omega}(x)>0.$ We prove the existence of ground states by solving an appropriate minimization problem for the energy functional $E(u)$. By use of strict rearrangement inequalities we shall prove that there is a unique minimizer (up to translations and a phase factor) which is a positive spherically symmetric decreasing function (theorems 3.2 and theorem 3.3). The existence and uniqueness of solutions of the form (1.4) for given $\omega$ ,however, depends on the frequency $\omega$. It will turn out that for any $\omega\leq 0$ there is a unique ground state of the form (1.4). In addition, we shall prove there is a positive number $\omega^{}>0$ independent of the coupling $\gamma$ such that for any positive $\omega<\omega^{}$ there are two ground states of the form (1.4) with different $L^{2}$-norm. In the limit case $\omega=\omega^{}$ there is again a unique ground state (theorem 3.4). We then apply our existence result to obtain a sharp Hardy-Littlewood-Sobolev inequality for the logarithmic kernel (theorem 4.1). Finally, we prove the existence of non radial solutions of the Schrödinger- Newton system (1.1) which in polar coordinates $(r,\theta)$ are of the form (1.5) $\psi(t,r,\theta)=\phi_{m,\omega}(r)e^{im\theta-i\omega t},\quad\phi_{m,\omega}(r)\geq 0$ for positive integers $m$. These are eigenfunctions of the angular momentum operator $L=-i\partial_{\theta}$. Again, we prove the existence of such $\phi_{m,\omega}(r)$ by solving an appropriate minimization problem for the corresponding energy functional reduced to functions of the form (1.5) for any given $m$ (theorem 5.1). The minimizers can be interpreted as purely angular excitations and we also prove their uniqueness in this class of functions (theorem 5.2). ## 2\. Mathematical Framework ### 2.1. Functional Setting The natural function space $X$ for the quasi-stationary problem is given by (2.1) $X=\\{u:\mathbb{R}^{2}\to\mathbb{C}:\;\int_{\mathbb{R}^{2}}\|\nabla u\|^{2}+\|u\|^{2}+\ln(1+\|x\|)\,\|u\|^{2}\;dx<\infty\\}.$ The space $X$ is a Hilbert space and by Rellich’s criterion (see, e.g. theorem XIII.65 of [8]) the embedding $X\hookrightarrow L^{2}$ is compact. We note $X_{r}$ the space of radial functions in $X$. Formally, the energy $E$ associated to (1.2) is given by (2.2) $\begin{split}E(u)&=\int_{\mathbb{R}^{2}}\|\nabla u(x)\|^{2}\;dx+\frac{\gamma}{4\pi}\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\,\|u(x)\|^{2}\|u(y)\|^{2}\;dxdy\\\ &=T(u)+\frac{\gamma}{2}V(u)\\\ \end{split}$ In order to prove that the energy is indeed well defined on $X$ we decompose the potential energy $V(u)$ into two parts applying the identity (2.3) $\ln(r)=-\ln(1+\frac{1}{r})+\ln(1+r)$ for all $r>0$. We then define the corresponding functionals (2.4) $\begin{split}&V_{1}(u)=-\frac{1}{2\pi}\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}\ln(1+\frac{1}{\|x-y\|})\,\|u(x)\|^{2}\|u(y)\|^{2}\;dxdy\\\ &V_{2}(u)=\frac{1}{2\pi}\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}\ln(1+{\|x-y\|})\,\|u(x)\|^{2}\|u(y)\|^{2}\;dxdy\\\ \end{split}$ ###### Lemma 2.1. The energy functional $E:X\longrightarrow\mathbb{R}_{0}^{+}$ is well defined on $X$ and of class $C^{1}$. ###### Proof. Since $0\leq\ln(1+\frac{1}{r})\leq\frac{1}{r}$ for any $r>0$ we have by the Hardy-Littlewood-Sobolev inequality (see e.g. [9]) and Sobolev interpolation estimates (see e.g. [10])that $\|V_{1}(u)\|\leq C_{1}\|\|u\|\|^{4}_{8/3}\leq C_{2}\|\|\nabla u\|\|_{2}\|\|u\|\|_{2}^{3}$ for some constants $C_{1},C_{2}>0$. To bound the second term of the potential energy we note that $\ln(1+\|x-y\|)\leq\ln(1+\|x\|+\|y\|)\leq\ln(1+\|x\|)+\ln(1+\|y\|)$ and therefore $\|V_{2}(u)\|\leq\frac{1}{\pi}\int_{\mathbb{R}^{2}}\ln(1+\|x\|)\,u(x)\|^{2}\int_{\mathbb{R}^{2}}\|u(y)\|^{2}\;dxdy.\\\ $ The regularity properties of $E(u)$ are obvious. ∎ Finally, we consider the particle number (or charge) defined by (2.5) $N(u)=\int_{\mathbb{R}^{2}}\|u(x)\|^{2}\;dx,$ which is also a well-defined quantity on $X$. ### 2.2. Scaling properties If $\phi_{\omega}(x)$ is a solution of the stationary equation (2.6) $-\Delta\phi_{\omega}(x)+\frac{\gamma}{2\pi}\bigg{(}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\;\|\phi_{\omega}(y)\|^{2}\;dy\bigg{)}\;\phi_{\omega}(x)=\omega\phi_{\omega}(x),$ with finite particle number $N_{\omega}=N(\phi_{\omega})$ then by the virial theorem, (2.7) $T(\phi_{\omega})=\frac{\gamma}{8\pi}N^{2}(\phi_{\omega}).$ For any $\sigma>0$, the scaled function $\phi_{\omega,\sigma}(x)=\sigma^{-2}\gamma^{1/2}\phi_{\omega}({x}/{\sigma})$ solves (2.8) $\begin{split}&-\Delta\phi_{\omega,\sigma}(x)+\frac{1}{2\pi}\bigg{(}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\;\|\phi_{\omega,\sigma}(y)\|^{2}\;dy\bigg{)}\;\phi_{\omega,\sigma}(x)\\\ &=\\\ &\sigma^{-2}\big{(}\omega+\frac{\gamma N_{\omega}\ln\sigma}{2\pi}\big{)}\phi_{\omega,\sigma}(x).\\\ \end{split}$ If we choose $\sigma=\sigma_{\omega}$ such that $\omega+\frac{\gamma N_{\omega}\ln\sigma_{\omega}}{2\pi}=0$, then $\phi:=\phi_{\omega,\sigma_{\omega}}$ is independent of $\omega,\gamma$ and satisfies the Schrödinger equation (2.9) $-\Delta\phi(x)+\frac{1}{2\pi}\bigg{(}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\;\|\phi(y)\|^{2}\;dy\bigg{)}\;\phi(x)=0.$ Define (2.10) $\Lambda_{0}:=N(\phi).$ Then the virial theorem (2.7) reads (2.11) $T(\phi)=\frac{\Lambda_{0}^{2}}{8\pi}.$ From (2.9) we get $V(\phi)=-T(\phi)$, hence (2.12) $E(\phi)=T(\phi)+\frac{1}{2}V(\phi)=\frac{\Lambda_{0}^{2}}{16\pi}.$ The scaling of the particle number $\Lambda_{0}=\sigma_{\omega}^{-2}\gamma N_{\omega}$ yields a relation between frequency $\omega$ and the particle number $N_{\omega}$ which we shall discuss in detail in the following section for ground state solutions. ## 3\. Ground states ### 3.1. Existence of ground states We consider the following minimization problem: (3.1) $e_{0}(\lambda)=\inf\\{E(u),u\in X,N(u)=\lambda\\}.$ We note that the functional $u\to E(u)$ is not convex since the quadratic form $f\to\int_{\mathbb{R}}\int_{\mathbb{R}}\|x-y\|f(x)\bar{f}(y)\;dxdy$ is not positive so that standard convex minimization does not apply. We shall prove the following theorem: ###### Theorem 3.1. For any $\lambda>0$ there is a spherically symmetric decreasing $u_{\lambda}\in X$ such that $e_{0}(\lambda)=E(u_{\lambda})$ and $N(u_{\lambda})=\lambda$. ###### Proof. Let $(u_{n})_{n}$ be a minimizing sequence for $e_{0}(\lambda)$, that is $N(u_{n})=\lambda$ and $\underset{n\longrightarrow\infty}{\lim}E(u_{n})=e_{0}(\lambda)$. We also may assume that $E(u_{n})$ is uniformly bounded above. Denoting $u^{}$ the spherically symmetric-decreasing rearrangement of $u$ we have (see e.g. lemma 7.17 in [11]) $T(u)\geq T(u^{}),\quad N(u^{})=N(u).$ Applying the decomposition $V(u)=V_{1}(u)+V_{2}(u)$ defined in (2.4) we may apply the strict version of Riesz’s rearrangement inequality (see e.g. theorem 3.9 in [11]) to $V_{1}(u)$ since $\ln(1+1/\|x\|)$ is positive and strictly symmetric-decreasing. Therefore $V_{1}(u)\geq V_{1}(u^{})$ with equality only if $u(x)=u^{}(x-x_{0})$ for some $x_{0}\in\mathbb{R}^{2}$. For the second term $V_{2}(u)$ we apply the following rearrangement inequality: ###### Lemma 3.2. Let $f,g$ be two nonnegative functions on $\mathbb{R}$, vanishing at infinity with spherically symmetric-decreasing rearrangement $f^{},g^{}$ , respectively. Let $v$ be a nonnegative spherically symmetric increasing function. Then (3.2) $\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}f(x)v(x-y)g(y)\;dxdy\geq\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}f^{}(x)v(x-y)g^{}(y)\;dxdy$ ###### Proof. The proof follows the same lines as in [12], lemma 3.2, where we proved the corresponding lemma in one space dimension. We give it here for the sake of completeness. If $v$ is bounded, $v\leq C$, then $(C-v)^{}=C-v$ and by Riesz’s rearrangement inequality (lemma 3.6 in [11]) we have $\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}f(x)(C-v(x-y))g(y)\;dxdy\leq\int_{\mathbb{R}^{2}}\int_{\mathbb{R}^{2}}f^{}(x)(C-v(x-y))g^{}(y)\;dxdy.$ Since $\int_{\mathbb{R}^{2}}f(x)\;dx\int_{\mathbb{R}^{2}}g(y)\;dy=\int_{\mathbb{R}^{2}}f^{}(x)\;dx\int_{\mathbb{R}^{2}}g^{}(y)\;dy$ the claim follows. If $v$ is unbounded we define a truncation by $v_{n}(x)=\sup{(v(x),n)}$ and apply the monotone convergence theorem. ∎ By the preceding lemma we have $V_{2}(u)\geq V_{2}(u^{})$ and consequently $V(u)\geq V(u^{})$ with equality only if $u(x)=e^{i\theta}u^{}(x-x_{0})$ for some $\theta\in\mathbb{R}$ and $x_{0}\in\mathbb{R}^{2}$. Therefore we may suppose that $u_{n}=u^{}_{n}$. We claim that $u^{}_{n}\in X$. Indeed, by Newton’s theorem (see e.g. theorem 9.7 in [11]) we have (3.3) $\begin{split}V(u^{}_{n})&=\frac{1}{2\pi}\int_{\mathbb{R}^{2}}{u^{}_{n}(x)}^{2}(\ln\|x\|)\;\bigg{(}\int_{B_{\|x\|}}{u^{}_{n}(y)}^{2}\;dy\bigg{)}\;dx\\\ &+\frac{1}{2\pi}\int_{\mathbb{R}^{2}}{u^{}_{n}(x)}^{2}\bigg{(}\int_{\mathbb{R}^{2}\setminus B_{\|x\|}}(\ln\|y\|)\;{u^{}_{n}(y)}^{2}\;dy\bigg{)}\;dx\\\ \end{split}$ where $B_{\|x\|}$ denotes the disc of radius $\|x\|$ centered at the origin. Since $\ln\|y\|\geq\ln\|x\|$ for all $y\in\mathbb{R}^{2}\setminus B_{\|x\|}$ we get $V(u^{}_{n})\geq\frac{\lambda}{2\pi}\int_{\mathbb{R}^{2}}(\ln\|x\|)\;{u^{}_{n}(x)}^{2}\;dx.$ Using $\ln\|x\|\geq\ln(1+\|x\|)-\frac{1}{\|x\|}$ and the sharp Sobolev inequality between the linear operators $-\Delta$ and $\frac{1}{\|x\|}$ (3.4) $\int_{\mathbb{R}^{2}}\frac{1}{\|x\|}\;\|u(x)\|^{2}\;dx\leq 2\|\|u\|\|_{2}\|\|\nabla u\|\|_{2}$ we finally get $V(u^{}_{n})\geq\frac{\lambda}{2\pi}\int_{\mathbb{R}^{2}}\ln(1+\|x\|)\;{u^{}_{n}(x)}^{2}\;dx-\frac{\lambda^{3/2}}{\pi}\|\|\nabla u^{}_{n}\|\|_{2}.$ Hence $E(u^{}_{n})\geq\|\|\nabla u^{}_{n}\|\|_{2}^{2}+\frac{\gamma\lambda}{4\pi}\int_{\mathbb{R}^{2}}\ln(1+\|x\|)\;{u^{}_{n}(x)}^{2}\;dx-\frac{\gamma\lambda^{3/2}}{2\pi}\|\|\nabla u^{}_{n}\|\|_{2}$ proving our claim. We may extract a subsequence which we denote again by $(u^{}_{n})_{n}$ such that $u^{}_{n}\to u^{}$ weakly in $X$, strongly in $L^{2}$ and a.e. where $u^{}\in X$ is a nonnegative spherically symmetric decreasing function. Note that $u^{}\neq 0$ since $N(u^{})=\lambda$. We want to show that $E(u^{})\leq\underset{n\longrightarrow\infty}{\lim\inf}\;E(u^{}_{n})$. Since $T(u^{})\leq\underset{n\longrightarrow\infty}{\lim\inf}\;T(u^{}_{n})$ it remains to analyze the functional $V(u)$. Let $\eta(x)=\int_{B_{\|x\|}}\|u^{}(y)\|^{2}\;dy,\quad\eta_{n}(x)=\int_{B_{\|x\|}}\|u^{}_{n}(y)\|^{2}\;dy.$ Then $\eta_{n}(x)\to\eta(x)$ uniformly since $\|\|\eta_{n}(x)-\eta(x)\|\|_{\infty}\leq\|\|u^{}_{n}-u^{}\|\|_{2}\;\|\|u^{}_{n}+u^{}\|\|_{2}\leq 2\sqrt{\lambda}\|\|u^{}_{n}-u^{}\|\|_{2}.$ We note that for any spherically symmetric density $\|u(x)\|^{2}$ with $u\in X$ we may simplify (3.3) to $V(u^{}_{n})=\frac{1}{\pi}\int_{\mathbb{R}^{2}}{u^{}_{n}(x)}^{2}(\ln\|x\|)\;\bigg{(}\int_{B_{\|x\|}}{u^{}_{n}(y)}^{2}\;dy\bigg{)}\;dx.$ Therefore, by the definition of $\eta_{n},\eta$, we have (3.5) $\begin{split}V(u^{}_{n})-V(u^{})&=\frac{1}{\pi}\int_{\mathbb{R}^{2}}(\ln\|x\|)\|u^{}_{n}(x)\|^{2}\big{(}\eta_{n}(x)-\eta(x)\big{)}\;dx\\\ &\quad+\frac{1}{\pi}\int_{\mathbb{R}^{2}}(\ln\|x\|)\eta(x)\big{(}\|u^{}_{n}(x)\|^{2}-\|u^{}(x)\|^{2}\big{)}\;dx\\\ \end{split}$ As $n\to\infty$ the first integral in (3.5) will tend to zero. In order to analyze the second integral we again decompose $\ln\|x\|$ according to (2.3). Then $\frac{1}{\pi}\int_{\mathbb{R}^{2}}\ln(1+\|x\|)\eta(x)\big{(}\|u^{}_{n}(x)\|^{2}-\|u^{}(x)\|^{2}\big{)}\;dx$ will remain nonnegative since the continuous functional $\phi\to\int_{\mathbb{R}^{2}}\ln(1+\|x\|)\eta(x)\|\phi(x)\|^{2}\;dx$ is positive while $-\frac{1}{\pi}\int_{\mathbb{R}^{2}}\ln(1+\frac{1}{\|x\|})\eta(x)\big{(}\|u^{}_{n}(x)\|^{2}-\|u^{}(x)\|^{2}\big{)}\;dx$ converges to zero since by Hölder’s inequality and inequality (3.4) we have the estimate $\begin{split}&\frac{1}{\pi}\int_{\mathbb{R}^{2}}\ln(1+\frac{1}{\|x\|})\eta(x)\big{(}\|u^{}_{n}(x)\|^{2}-\|u^{}(x)\|^{2}\big{)}\;dx\\\ &\leq\frac{2}{\pi}\|\|\eta\|\|_{\infty}\|\|\nabla(u^{}_{n}+u^{})\|\|_{2}^{1/2}\|\|u^{}_{n}+u^{}\|\|_{2}^{1/2}\|\|\nabla(u^{}_{n}-u^{})\|\|_{2}^{1/2}\|\|u^{}_{n}-u^{}\|\|_{2}^{1/2}\\\ &\leq C\|\|u^{}_{n}-u^{}\|\|_{2}^{1/2}\\\ \end{split}$ for a positive constant $C$. Hence $V(u^{})\leq\underset{n\longrightarrow\infty}{\lim\inf}\;V(u^{}_{n})$ proving the theorem. ∎ ### 3.2. Uniqueness of ground states Let $u_{\lambda}$ denote the solution of the minimization problem (5.3) found in theorem . ###### Theorem 3.3. For any $\lambda>0$ the solution $u_{\lambda}$ of the minimization problem (5.3) is unique in the following sense: If $v_{\lambda}\in X$ is such that $e_{0}(\lambda)=E(v_{\lambda})$ and $N(v_{\lambda})=\lambda$, then $v_{\lambda}\in\\{e^{i\theta}u^{}(x-x_{0}),\theta\in\mathbb{R},x_{0}\in\mathbb{R}^{2}\\}$. ###### Proof. Since by the strict rearrangement inequality $E(u)>E(u^{})$ for all $u\notin\\{e^{i\theta}u^{}(x-x_{0}),\theta\in\mathbb{R},x_{0}\in\mathbb{R}^{2}\\}$ established in the proof of theorem 3.2 it is sufficient to prove uniqueness in the class of spherically symmetric (decreasing) nonnegative functions. To do so we consider the Euler-Lagrange equation for $u_{\lambda}$. By theorem 3.2 for any $\lambda>0$ there is a Lagrange multiplier $\omega$ such that, at least in a weak sense (3.6) $-\Delta u_{\lambda}(x)+\frac{\gamma}{2\pi}\bigg{(}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\;\|u_{\lambda}(y)\|^{2}\;dy\bigg{)}\;u_{\lambda}(x)=\omega u_{\lambda}(x).$ We define the quantities $I_{\lambda}:=\frac{1}{2\pi}\int_{\mathbb{R}^{2}}\ln(\|y\|)\;\|u_{\lambda}(y)\|^{2}\;dy,\quad E_{\lambda}:=\omega-\gamma I_{\lambda}$ and the potential $W_{\lambda}(x):=\frac{1}{2\pi}\bigg{(}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\;\|u_{\lambda}(y)\|^{2}\;dy\bigg{)}-I_{\lambda}.$ Then $W_{\lambda}(x)\geq 0$ for all $x\in\mathbb{R}^{2}$, $W_{\lambda}(0)=0$ and $\Delta W_{\lambda}=\|u_{\lambda}\|^{2}$. The function $u_{\lambda}$ is the ground state of the Schrödinger operator $-\Delta+\gamma W_{\lambda}$ with eigenvalue $E_{\lambda}$ and therefore $E_{\lambda}>0$. The rescaled functions (3.7) $u(\|x\|):=\frac{\sqrt{\gamma}}{E_{\lambda}}\;u_{\lambda}(x/\sqrt{E_{\lambda}}),\quad W(\|x\|):=\frac{{\gamma}}{E_{\lambda}}\;W_{\lambda}(x/\sqrt{E_{\lambda}})$ then $u,W$ satisfy the universal equations (3.8) $\Delta u=(W-1)u,\quad\Delta W=\|u\|^{2}.$ In [7] it was shown that (3.9) admits a unique spherically symmetric solution $(u,W)$ such that $u$ is a positive decreasing function vanishing at infinity. See also theorem 5.2 of the present paper where we consider a more general system. ∎ ### 3.3. Dependence on parameters and admissible frequencies We determine explicitly $e_{0}(\lambda)$ as a function of $\lambda$ and the range of admissible frequencies $\omega$ for ground states of the form (1.4). We consider the unique spherically symmetric solution $(u,W)$, $u>0$ and $u$ vanishing at infinity , of the universal system (3.9). Let (3.9) $N:=\int_{\mathbb{R}^{2}}\|u(\|x\|)\|^{2}\;dx,\quad I:=\frac{1}{2\pi}\int_{\mathbb{R}^{2}}\ln(\|x\|)\;\|u(\|x\|)\|^{2}\;dx.$ These quantities are finite since, for any $p>0$, $u$ decays faster than $\exp(-px)$ at infinity and in particular $u\in X$. The corresponding numerical values are easily computed from the solution of and we give them in the appendix. By the scaling (3.7) we obtain the following identity for the parameters $\gamma,\lambda$ and the Lagrange multiplier $\omega$: (3.10) $\omega=\frac{\gamma\lambda}{N}\big{(}1+I-\frac{N}{4\pi}\ln\frac{\gamma\lambda}{N}\big{)}.$ Multiplying the variational equation (5.4) by $u_{\lambda}$ and integrating we obtain $2E(u_{\lambda})-T(u_{\lambda})=T(u_{\lambda})+\gamma V(u_{\lambda})=\omega\lambda.$ Applying the virial relation (2.7) and using (3.10) we finally have (3.11) $e_{0}(\lambda)=\frac{\gamma\lambda^{2}}{16\pi}\Big{(}1+8\pi\frac{1+I}{N}-2\ln\big{(}\frac{\gamma\lambda}{N}\big{)}\Big{)}$ with $N,I$ given in (3.9). Note that (3.11) can be also obtained by integrating $\omega=\omega(\lambda)$ in (3.10) with respect to $\lambda$ since, at least formally, $\frac{d\,e_{0}(\lambda)}{d\lambda}=\omega.$ Taking $\phi$ in (2.9) as the unique ground state solution the relation (3.10) between $\lambda,\omega$ and $\gamma$ simplifies to (3.12) $\omega=-\frac{\gamma\lambda}{4\pi}\ln\frac{\gamma\lambda}{\Lambda_{0}}$ with $\Lambda_{0}=N(\phi)$ as defined in (2.10). The energy is then given by (3.13) $e_{0}(\lambda)=\frac{\gamma\lambda^{2}}{16\pi}\big{(}1-2\ln\frac{\gamma\lambda}{\Lambda_{0}}\big{)}.$ The relations (3.10), respectively (3.12), yield the following result on admissible frequencies: ###### Theorem 3.4. (1) For any $\omega\leq 0$ the Schrödinger-Newton system (1.1) admits a unique spherically ground state of the form (1.4). In particular, for the solution $\phi$ of the stationary equation (2.9) the energy $e_{0}(\lambda)$ attains its maximum. (2) For any $0<\omega<\omega^{}$ with $\omega^{}=\frac{N}{4\pi e}\exp\big{(}4\pi\;\frac{1+I}{N}\big{)}=\frac{\Lambda_{0}}{4\pi e}$ the Schrödinger-Newton system (1.1) admits two spherically ground states of the form (1.4) with different particle numbers. (3) For $\omega=\omega^{}$ there is a unique ground state solution of the form (1.4). ## 4\. A logarithmic Hardy-Littlewood-Sobolev inequality In [13], E. Carlen and M. Loss proved the following logarithmic Hardy- Littlewood-Sobolev inequality: (4.1) $-\int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{n}}\ln(\|x-y\|)\,f(x)f(y)\;dxdy\leq\frac{1}{n}\int_{\mathbb{R}^{n}}f(x)\ln f(x)\;dx+C_{0}$ for all nonnegative real-valued $f$ with $\int_{\mathbb{R}^{n}}f(x)\;dx=1,\int_{\mathbb{R}^{n}}f(x)\ln(1+\|x\|)\;dx,\int_{\mathbb{R}^{n}}f(x)\ln f(x)\;dx<\infty$ and sharp constant $C_{0}$. We focus on the case $n=2$. Replacing $f$ by $\|u\|^{2}$ with $N(u)=\lambda$ this reads as (4.2) $-2\pi V(u)\leq\frac{\lambda}{2}\int_{\mathbb{R}^{n}}\|u(x)\|^{2}\ln\|u(x)\|^{2}\;dx+C_{0}\lambda^{2}-\frac{\lambda^{2}}{2}\ln\lambda.$ Applying the logarithmic Sobolev inequality ( [14], see also [11], theorem 8.14) (4.3) $\int_{\mathbb{R}^{2}}\|u(x)\|^{2}\ln\|u(x)\|^{2}\;dx\leq\lambda\ln(T(u))-\lambda(1+\ln\pi)$ to the integral on the r.h.s. of (4.2) and using $2C_{0}=1+\ln\pi$ when $n=2$ we get $-V(u)\leq\frac{\lambda^{2}}{4\pi}\ln\frac{T(u)}{\lambda}.$ However, this inequality is not sharp. Here we give the sharp version: ###### Theorem 4.1. For any $u\in X$ the following inequality holds: (4.4) $-V(u)\leq\frac{\lambda^{2}}{4\pi}\ln\frac{8\pi T(u)}{N\lambda}-\big{(}\frac{I+1}{N}-\frac{1}{8\pi}\big{)}\lambda^{2}$ where $\lambda=\int_{\mathbb{R}^{2}}\|u\|^{2}\;dx$ and the constants $I$ and $N$ are given in (3.9). ###### Proof. By theorem 3.2 we have for any $u\in X$ with $\lambda=\int_{\mathbb{R}^{2}}\|u\|^{2}\;dx$ and for any coupling constant $\gamma>0$ the sharp inequality $T(u)+\frac{\gamma}{2}V(u)\geq e_{0}(\lambda)$ with $e_{0}(\lambda)$ given in (3.11). We optimize with respect to $\gamma$. The optimal $\gamma$ is given by $\frac{1}{4\pi}\ln\frac{\gamma\lambda}{N}=-\lambda^{-2}V(u)+\frac{I+1}{N}-\frac{1}{8\pi}$ which yields the desired inequality. ∎ ## 5\. Purely angular excitations In this section we prove the existence of non radial solutions of the Schrödinger-Newton system (1.1) which in polar coordinates $(r,\theta)$ are of the form $\psi(t,r,\theta)=\phi_{\omega}(r)e^{im\theta-i\omega t},\quad\phi_{\omega}(r)\geq 0$ for positive integers $m$. These are eigenfunctions of the angular momentum operator $L=-i\partial_{\theta}$. We define the function space $X_{m}^{(s)}$ of spherically symmetric functions $u:\mathbb{R}^{2}\to\mathbb{C}$ by (5.1) $X_{m}^{(s)}=\\{u=u(\|x\|):\int_{\mathbb{R}^{2}}\|\partial_{r}u\|^{2}+\frac{m^{2}}{r^{2}}\|\|u\|^{2}+\|u\|^{2}+\ln(1+r)\,\|u\|^{2}\;dx<\infty\\}.$ We define the energy functional for purely angular excitations by (5.2) $E_{m}(u)=T_{m}(u)+\frac{\gamma}{2}V(u),\quad T_{m}(u)=\int_{\mathbb{R}^{2}}\|\partial_{r}u\|^{2}+\frac{m^{2}}{r^{2}}\|\|u\|^{2}\;dx,$ We consider the minimization problem (5.3) $e_{m}(\lambda)=\inf\\{E_{m}(u),u\in X_{m}^{(s)},N(u)=\lambda\\}.$ The following theorem holds: ###### Theorem 5.1. For any $\lambda>0$ there is a nonnegative $u_{\lambda}\in X_{m}^{(s)}$ such that $e_{m}(\lambda)=E_{m}(u_{\lambda})$ and $N(u_{\lambda})=\lambda$. ###### Proof. Consider any minimizing sequence $(u_{n})_{n}$ for $e_{m}(\lambda)$. Since $E_{m}(u)$ is decreasing under the replacement $u\mapsto\|u\|$ we may suppose that $u_{n}\geq 0$. Now we mimic the proof of theorem 3.2 starting from Newton’s theorem (3.3) replacing the energy $E$ by $E_{m}$ since only the spherical symmetry of the functions are required in this part of the proof. ∎ The minimizing $u_{\lambda}$ satisfies the Euler-Lagrange equation (5.4) $-u_{\lambda}^{\prime\prime}-\frac{1}{r}u^{\prime}_{\lambda}+\frac{m^{2}}{r^{2}}u_{\lambda}+\frac{\gamma}{2\pi}\bigg{(}\int_{\mathbb{R}^{2}}\ln(\|x-y\|)\;\|u_{\lambda}(\|y\|)\|^{2}\;dy\bigg{)}\;u_{\lambda}=\omega u_{\lambda}.$ for $r\geq 0$ and some multiplier $\omega$. By the same rescaling as in the proof of theorem 3.3 we obtain a system of universal equations given by (5.5) $u^{\prime\prime}+\frac{1}{r}u^{\prime}-\frac{m^{2}}{r^{2}}u=(W-1)u,\quad W^{\prime\prime}+\frac{1}{r}W^{\prime}=\|u\|^{2}$ such that $W(0)=W^{\prime}(0)=0$ and $u\geq 0$. We note that $f(r):=r^{-m}u(r)$ then satisfies $f(0)>0$, $f^{\prime}(0)=0$ and the differential equation $f^{\prime\prime}+\frac{2m+1}{r}f^{\prime}=(W-1)f$ from which we easily deduce $f>0$ and $f^{\prime}<0$ using the facts $f\geq 0$ and $f\to 0$ as $r\to\infty$. Also note that for any $p>0$, $u$ decays faster than $\exp(-pr)$ at infinity. Uniqueness of the solution $u_{\lambda}\in X_{m}^{(s)}$ follows again from the uniqueness of the solutions of the universal system (5.5) which we prove in the following theorem: ###### Theorem 5.2. For any $m\geq 0$ there is a unique solution $(u,W)$ of (5.5) such that $u>0$ for $r>0$ and $u\to 0$ as $r\to\infty$. ###### Proof. Suppose there are two distinct solutions $(u_{1},W_{1})$, $(u_{2},W_{2})$ having the required properties. We may suppose $u_{2}(r)>u_{1}(r)$ for $r\in]0,\bar{r}[$. We consider the Wronskian $w(r)=u_{2}^{\prime}(r)u_{1}(r)-u_{1}^{\prime}(r)u_{2}(r).$ Note that $w(0)=0$ and $rw(r)\to 0$ as $r\to\infty$. It satisfies the differential equation $(rw)^{\prime}=r(W_{2}-W_{1})u_{1}u_{2}.$ Suppose $u_{2}(r)>u_{1}(r)$ for all $r>0$. Then $W_{2}(r)>W_{1}(r)$ for all $r>0$ since $r(W_{2}-W_{1})^{\prime}=\int_{0}^{r}(u_{2}^{2}-u_{1}^{2})s\;ds>0$ and hence $(rw)^{\prime}>0$ for all $r>0$ which is impossible. Hence there exists $\bar{r}>0$ such that $\delta(r)=u_{2}(r)-u_{1}(r)>0$ for $r\in[0,\bar{r}[$, $\delta(\bar{r})=0$ and $\delta^{\prime}(\bar{r})<0$. However, then $w(\bar{r})=\delta^{\prime}(\bar{r})u_{1}(\bar{r})<0$, but $(rw)^{\prime}(r)>0$ for all $r<\bar{r}$, that is $w(r)>0$ for all $r<\bar{r}$ which is again impossible. ∎ ## Appendix A Numerical values Let $(u,W)$ be the unique spherically symmetric solution of the universal equations (3.9) such that $W(0)=W^{\prime}(0)=0$, $u>0$ vanishing at infinity, i.e. (A.1) $(ru^{\prime})^{\prime}=r(W-1)u,\quad(rW^{\prime})^{\prime}=r\|u\|^{2}.$ By integrating the second equation we get (A.2) $N=2\pi\int_{0}^{\infty}u^{2}(s)s\;ds=2\pi\underset{r\longrightarrow\infty}{\lim}rW^{\prime}(r).$ Multiplying the second equation by $\ln r$ and using $(rW^{\prime})^{\prime}\ln r=(rW^{\prime}\ln r-W)^{\prime}$ we obtain after integration (A.3) $I=\int_{0}^{\infty}u^{2}(s)s\ln s\;ds=\underset{r\longrightarrow\infty}{\lim}rW^{\prime}(r)\ln r-W(r).$ Solving (A.1) numerically we then find (A.4) $N=2\pi\cdot 1.64145=10.3135,\quad I=0.2276$ and therefore (A.5) $\Lambda_{0}=46.03.$ With these numerical values the logarithmic Hardy-Littlewood-Sobolev inequality (4.4) reads as follows: (A.6) $-V(u)\leq\frac{\lambda^{2}}{4\pi}\ln\frac{T(u)}{\lambda}-0.0084\lambda^{2}.$ ## References * [1] Pekar, S.I., Untersuchungen über die Elektronentheorie der Kristalle, §6, p. 29 -34, Akademie Verlag Berlin, 1954. * [2] Lieb, E.H., Existence and uniqueness of the minimizing solution of Choquard’s nonlinear equation, Studies in Applied Mathematics 57, 93-105 (1977). * [3] Penrose, R., On gravity’s role in quantum state reduction, Gen. Rel. Grav. 28, 581-600 (1996). * [4] Harrison, R., Moroz, I. Tod, K.P., A numerical study of the Schrödinger-Newton equation, Nonlinearity 16, 101-122 (2003). * [5] Lions, P.L., The Choquard equation and related questions, Nonlinear Analysis T.M.A. 4, 1063-1073 (1980). * [6] Lions, P.L., Solutions complexes d’équations elliptiques semi-linéaires dans $\mathbb{R}^{n}$, C.R. Acad. Sc. Paris t.302, Série 1 ,no. 19, 673-676 (1986). * [7] Choquard, Ph., Stubbe, J., Vuffray, M., Stationary solutions of the Schrödinger-Newton model - An ODE approach, Adv. Diff. Eq. to appear (2008). * [8] Reed, M. and Simon, B., Methods of modern mathematical physics, Vol. IV, Analysis of operators, Academic Press 1978. * [9] Lieb, E.H., Sharp constants in Hardy-Littlewood-Sobolev and related inequalities , Ann. Math. 118, 349 - 374 (1983). * [10] Weinstein, M.I., Nonlinear Schrödinger equations and sharp interpolation estimates , Comm. Math. Phys. 87, 567 - 576 (1983). * [11] Lieb, E.H. and Loss, M., Analysis, 2nd edition, Graduate Studies in Mathematics, vol. 14, AMS 2001. * [12] Choquard, Ph., Stubbe, J. The one-dimensional Schrödinger-Newton equations, Lett. Math. Phys. 81, 177-184 (2007). * [13] Carlen, E., Loss, M. Competing symmetries, the logarithmic HLS inequality and Onofri’s inequality on $S^{n}$, Geom. Funct. Anal. 2, 90-103 (1992). * [14] Weissler, F., Logarithmic Sobolev Inequalities for the Heat-Diffusion Semigroup, Trans. AMS 237, 255-269 (1978).	arxiv-papers	2008-07-25T11:07:50	2024-09-04T02:48:56.991888	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Joachim Stubbe", "submitter": "Stubbe Joachim", "url": "https://arxiv.org/abs/0807.4059" }
0807.4230	Inhomogeneous Bulk Viscous Fluid Universe with Electromagnetic Field and Variable $\Lambda$-Term Anirudh Pradhan 111Corresponding author, Vandana Rai 2, Kanti Jotania3 1 Department of Mathematics, Hindu Post-graduate College, Zamania-232 331, Ghazipur, India E-mail : pradhan@iucaa.ernet.in 2Department of Mathematics, Post-graduate College, Ghazipur-233 001, India E-mail : vandana_rai005@yahoo.co.in 3Department of Physics, Faculty of Science, The M. S. University of Baroda, Vadodara-390 002, India E-mail : kanti@iucaa.ernet.in ###### Abstract Cylindrically symmetric inhomogeneous cosmological model for bulk viscous fluid distribution with electromagnetic field is obtained. The source of the magnetic field is due to an electric current produced along the z-axis. $F_{12}$ is the non-vanishing component of electromagnetic field tensor. To get the deterministic solution, it has been assumed that the expansion $\theta$ in the model is proportional to the shear $\sigma$. The values of cosmological constant for these models are found to be small and positive at late time which are consistent with the results from recent supernovae Ia observations. Physical and geometric aspects of the models are also discussed in presence and absence of magnetic field. Keywords : cosmology, variable cosmological term, electromagnetic field, inhomogeneous universe PACS number: 98.80.Jk, 98.80.-k ## 1 Introduction Inhomogeneous cosmological models play an important role in understanding some essential features of the universe such as the formation of galaxies during the early stages of evolution and process of homogenization. The early attempts at the construction of such models have been done by Tolman [1] and Bondi [2] who considered spherically symmetric models. Inhomogeneous plane- symmetric models were considered by Taub [3, 4] and later by Tomimura [5], Szekeres [6], Collins and Szafron [7], Szafron and Collins [8]. Recently, Senovilla [9] obtained a new class of exact solutions of Einstein’s equations without big bang singularity, representing a cylindrically symmetric, inhomogeneous cosmological model filled with perfect fluid which is smooth and regular everywhere satisfying energy and causality conditions. Later, Ruiz and Senovilla [10] have examined a fairly large class of singularity free models through a comprehensive study of general cylindrically symmetric metric with separable function of $r$ and $t$ as metric coefficients. Dadhich et al. [11] have established a link between the FRW model and the singularity free family by deducing the latter through a natural and simple in-homogenization and anisotropization of the former. Also, Patel et al. [12] presented a general class of inhomogeneous cosmological models filled with non-thermalized perfect fluid by assuming that the background space-time admits two space-like commuting Killing vectors and has separable metric coefficients. Singh, Mehta and Gupta [13] obtained inhomogeneous cosmological models of perfect fluid distribution with electro-magnetic field. Recently, Pradhan et al. [14] have investigated plane-symmetric inhomogeneous cosmological models in various contexts. Cylindrically symmetric space-time play an important role in the study of the universe on a scale in which anisotropy and inhomogeneity are not ignored. Roy and Singh [15], Bali and Tyagi [16, 17], Chakrabarty et al. [18] and Pradhan et al. [19] have investigated cylindrically symmetric inhomogeneous cosmological models in presence of electromagnetic field. The occurrence of magnetic field on galactic scale is well-established fact today, and their importance for a variety of astrophysical phenomena is generally acknowledged as pointed out by Zeldovich et al. [20]. Also Harrison [21] has suggested that magnetic field could have a cosmological origin. As a natural consequences, we should include magnetic fields in the energy-momentum tensor of the early universe. The choice of anisotropic cosmological models in Einstein system of field equations leads to the cosmological models more general than Robertson-Walker model [22]. The presence of primordial magnetic field in the early stages of the evolution of the universe has been discussed by several authors [23]$-$[32]. Strong magnetic field can be created due to adiabatic compression in clusters of galaxies. Large-scale magnetic field gives rise to anisotropies in the universe. The anisotropic pressure created by the magnetic fields dominates the evolution of the shear anisotropy and it decays slower than the case when the pressure was isotropic [33, 34]. Such fields can be generated at the end of an inflationary epoch [35]$-$[39]. Anisotropic magnetic field models have significant contribution in the evolution of galaxies and stellar objects. Bali and Ali [40] obtained a magnetized cylindrically symmetric universe with an electrically neutral perfect fluid as the source of matter. Chakrabarty et al. [18] and Pradhan et al. [41] have investigated magnetized viscous fluid cosmological models in various contexts. There are significant observational evidence for the detection of Einstein’s cosmological constant, $\Lambda$ or a component of material content of the universe that varies slowly with time and space to act like $\Lambda$. Some of the recent discussions on the cosmological constant “problem” and on cosmology with a time-varying cosmological constant by Ratra and Peebles [42], and Sahni and Starobinsky [43] point out that in the absence of any interaction with matter or radiation, the cosmological constant remains a “constant”. However, in the presence of interactions with matter or radiation, a solution of Einstein equations and the assumed equation of covariant conservation of stress-energy with a time-varying $\Lambda$ can be found. This entails that energy has to be conserved by a decrease in the energy density of the vacuum component followed by a corresponding increase in the energy density of matter or radiation (see also Carroll, Press and Turner [44], Peebles [45], Padmanabhan [46]). There is a plethora of astrophysical evidence today, from supernovae measurements (Perlmutter et al. [47], Riess et al. [48], Garnavich et al. [49], Schmidt et al. [50], Blakeslee et al. [51], Astier et al. [52]), the spectrum of fluctuations in the Cosmic Microwave Background (CMB) [53], baryon oscillations [54] and other astrophysical data, indicating that the expansion of the universe is currently accelerating. The energy budget of the universe seems to be dominated at the present epoch by a mysterious dark energy component, but the precise nature of this energy is still unknown. Many theoretical models provide possible explanations for the dark energy, ranging from a cosmological term [55] to super-horizon perturbations [56] and time- varying quintessence scenarios [57]. These recent observations strongly favour a significant and a positive value of $\Lambda$ with magnitude $\Lambda(G\hbar/c^{3})\approx 10^{-123}$. In Ref. [48], Riess et al. have recently presented an analysis of 156 SNe including a few at $z>1.3$ from the Hubble Space Telescope (HST) “GOOD ACS” Treasury survey. They conclude to the evidence for present acceleration $q_{0}<0$ $(q_{0}\approx-0.7)$. Observations (Knop et al. [58]; Riess et al., [48]) of Type Ia Supernovae (SNe) allow us to probe the expansion history of the universe leading to the conclusion that the expansion of the universe is accelerating. Most studies in cosmology involve a perfect fluid. Large entropy per baryon and the remarkable degree of isotropy of the cosmic microwave background radiation, suggest that we should analyze dissipative effects in cosmology. Further, there are several processes which are expected to give rise to viscous effect. These are the decoupling of neutrinos during the radiation era and the recombination era [59], decay of massive super string modes into massless modes [60], gravitational string production [61, 62] and particle creation effect in grand unification era [63]. It is known that the introduction of bulk viscosity can avoid the big bang singularity. Thus, we should consider the presence of a material distribution other than a perfect fluid to have realistic cosmological models (see Grøn [64] for a review on cosmological models with bulk viscosity). A uniform cosmological model filled with fluid which possesses pressure and second (bulk) viscosity was developed by Murphy [65]. The solutions that he found exhibit an interesting feature that the big bang type singularity appears in the infinite past. Recently, Pradhan et al. [66] have obtained inhomogeneous perfect fluid universe with electromagnetic field. Motivated by the situation discussed above, in this paper, we have obtained a new cylindrically symmetric inhomogeneous cosmological model for bulk viscous fluid distribution in presence and absence of electromagnetic field. The coefficient of bulk viscosity is assumed to be a power function of mass density. This work generalize the previous work of Pradhan et al. [66]. Some physical and geometric behaviour of the models in presence and absence of magnetic field are also discussed. This paper is organized as follows. The metric and the field equations are laid down in Section 2. In Section 3, we deal with the solutions of the field equations in presence of bulk viscous fluid with electromagnetic field and variable cosmological term. We have also described the physical and geometric aspects of the models. In Section 4, we obtain the solutions of the field equations in absence of the magnetic field. Finally in Section 5 concluding remarks are given. ## 2 The Metric and Field Equations We consider the metric in the form $ds^{2}=A^{2}(dx^{2}-dt^{2})+B^{2}dy^{2}+C^{2}dz^{2},$ (1) where $A$ is the function of $t$ alone and $B$ and $C$ are functions of $x$ and $t$. The energy momentum tensor is taken as has the form $T^{j}_{i}=(\rho+\bar{p})u_{i}u^{j}+\bar{p}g^{j}_{i}+E^{j}_{i},$ (2) where $E^{j}_{i}$ is the electromagnetic field given by Lichnerowicz [67] $E^{j}_{i}=\bar{\mu}\left[h_{l}h^{l}\left(u_{i}u^{j}+\frac{1}{2}g^{j}_{i}\right)-h_{i}h^{j}\right],$ (3) and $\bar{p}=p-\xi u^{i}_{;i}.$ (4) Here $\rho$, $p$, $\bar{p}$ and $\xi$ are the energy density, isotropic pressure, effective pressure and bulk viscous coefficient respectively and $u^{i}$ is the fluid four-velocity vector satisfying the condition $g_{ij}u^{i}u^{j}=-1.$ (5) $\bar{\mu}$ is the magnetic permeability and $h_{i}$, the magnetic flux vector defined by $h_{i}=\frac{1}{\bar{\mu}}\,{{}^{}}F_{ji}u^{j},$ (6) where the dual electromagnetic field tensor ${}^{}F_{ij}$ is defined by Synge [68] $^{}F_{ij}=\frac{\sqrt{-g}}{2}\epsilon_{ijkl}F^{kl}.$ (7) Here $F_{ij}$ is the electromagnetic field tensor and $\epsilon_{ijkl}$ is the Levi-Civita tensor density. The co-ordinates are considered to be comoving so that $u^{1}$ = $u^{2}$ = $u^{3}$ = $0$ and $u^{4}=\frac{1}{A}$. If we consider that the current flows along the $z$-axis, then $F_{12}$ is the only non-vanishing component of $F_{ij}$. The Maxwell’s equations $F_{ij;k}+F_{jk;i}+F_{ki;j}=0,$ (8) $\left[\frac{1}{\bar{\mu}}F^{ij}\right]_{;j}=4\pi J^{i}$ (9) require that $F_{12}$ is the function of $x$-alone. We assume that the magnetic permeability is the functions of $x$ and $t$ both. Here the semicolon represents a covariant differentiation. The Einstein’s field equations (in gravitational units $c=1,G=1$) read $R^{j}_{i}-\frac{1}{2}Rg^{j}_{i}+\Lambda g^{j}_{i}=-8\pi T^{j}_{i},$ (10) for the line-element (1) lead to the following system of equations: $\frac{1}{A^{2}}\left[-\frac{B_{44}}{B}-\frac{C_{44}}{C}+\frac{A_{4}}{A}\left(\frac{B_{4}}{B}+\frac{C_{4}}{C}\right)-\frac{B_{4}C_{4}}{BC}+\frac{B_{1}C_{1}}{BC}\right]-\Lambda$ $=8\pi\left(\bar{p}+\frac{F^{2}_{12}}{2\bar{\mu}A^{2}B^{2}}\right),$ (11) $\frac{1}{A^{2}}\left(\frac{A^{2}_{4}}{A^{2}}-\frac{A_{44}}{A}-\frac{C_{44}}{C}+\frac{C_{11}}{C}\right)-\Lambda=8\pi\left(\bar{p}+\frac{F^{2}_{12}}{2\bar{\mu}A^{2}B^{2}}\right),$ (12) $\frac{1}{A^{2}}\left(\frac{A^{2}_{4}}{A^{2}}-\frac{A_{44}}{A}-\frac{B_{44}}{B}+\frac{B_{11}}{B}\right)-\Lambda=8\pi\left(\bar{p}-\frac{F^{2}_{12}}{2\bar{\mu}A^{2}B^{2}}\right),$ (13) $\frac{1}{A^{2}}\left[-\frac{B_{11}}{B}-\frac{C_{11}}{C}+\frac{A_{4}}{A}\left(\frac{B_{4}}{B}+\frac{C_{4}}{C}\right)-\frac{B_{1}C_{1}}{BC}+\frac{B_{4}C_{4}}{BC}\right]+\Lambda$ $=8\pi\left(\rho+\frac{F^{2}_{12}}{2\bar{\mu}A^{2}B^{2}}\right),$ (14) $\frac{B_{14}}{B}+\frac{C_{14}}{C}-\frac{A_{4}}{A}\left(\frac{B_{1}}{B}+\frac{C_{1}}{C}\right)=0,$ (15) where the sub indices $1$ and $4$ in A, B, C and elsewhere denote ordinary differentiation with respect to $x$ and $t$ respectively. ## 3 Solution of the Field Equations Equations (11)-(15) are five independent equations in six unknowns $A$, $B$, $C$, $\rho$, $p$ and $F_{12}$. For the complete determinacy of the system, we need one extra condition. The research on exact solutions is based on some physically reasonable restrictions used to simplify the Einstein equations. To get determinate solution we assume that the expansion $\theta$ in the model is proportional to the shear $\sigma$. This condition leads to $A=\left(\frac{B}{C}\right)^{n},$ (16) where $n$ is a constant. From Eqs. (11)-(13), we have $\frac{A_{44}}{A}-\frac{A^{2}_{4}}{A^{2}}+\frac{A_{4}B_{4}}{AB}+\frac{A_{4}C_{4}}{AC}-\frac{B_{44}}{B}-\frac{B_{4}C_{4}}{BC}=\frac{C_{11}}{C}-\frac{B_{1}C_{1}}{BC}=\mbox{K (constant)}$ (17) and $\frac{8\pi F^{2}_{12}}{\bar{\mu}B^{2}}=-\frac{C_{44}}{C}+\frac{C_{11}}{C}+\frac{B_{44}}{B}-\frac{B_{11}}{B}.$ (18) We also assume that $B=f(x)g(t)$ $C=f(x)k(t).$ (19) Using Eqs. (16) and (17) in (15) and (17) lead to $\frac{k_{4}}{k}=\frac{(2n-1)}{(2n+1)}\frac{g_{4}}{g},$ (20) $(n-1)\frac{g_{44}}{g}-n\frac{k_{44}}{k}-\frac{g_{4}}{g}\frac{k_{4}}{k}=K,$ (21) $ff_{11}-f^{2}_{1}=Kf^{2}.$ (22) Equation (20) leads to $k=cg^{\alpha},$ (23) where $\alpha=\frac{2n-1}{2n+1}$ and $c$ is the constant of integration. From Eqs. (21) and (23), we have $\frac{g_{44}}{g}+\beta\frac{g^{2}_{4}}{g^{2}}=N,$ (24) where $\beta=\frac{n\alpha(\alpha-1)+\alpha}{n(\alpha-1)+1},\,\,N=\frac{K}{n(1-\alpha)-1}.$ Equation (22) leads to $f=\exp{\left(\frac{1}{2}K(x+x_{0})^{2}\right)},$ (25) where $x_{0}$ is an integrating constant. Equation (24) leads to $g=\left(c_{1}e^{bt}+c_{2}e^{-bt}\right)^{\frac{1}{(\beta+1)}},$ (26) where $b=\sqrt{(\beta+1)N}$ and $c_{1}$, $c_{2}$ are integrating constants. Hence from (23) and (26), we have $k=c\left(c_{1}e^{bt}+c_{2}e^{-bt}\right)^{\frac{\alpha}{(\beta+1)}}.$ (27) Therefore we obtain $B=\exp{\left(\frac{1}{2}K(x+x_{0})^{2}\right)}\left(c_{1}e^{bt}+c_{2}e^{-bt}\right)^{\frac{1}{(\beta+1)}},$ (28) $C=\exp{\left(\frac{1}{2}K(x+x_{0})^{2}\right)}c\left(c_{1}e^{bt}+c_{2}e^{-bt}\right)^{\frac{\alpha}{(\beta+1)}},$ (29) $A=a\left(c_{1}e^{bt}+c_{2}e^{-bt}\right)^{\frac{n(1-\alpha)}{(\beta+1)}},$ (30) where $a=\frac{c_{3}}{c}$, $c_{3}$ being a constant of integration. After using suitable transformation of the co-ordinates, the model (1) reduces to the form $ds^{2}=a^{2}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2n(1-\alpha)}{(\beta+1)}}(dX^{2}-dT^{2})+e^{KX^{2}}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2}{(\beta+1)}}dY^{2}$ $+e^{KX^{2}}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2\alpha}{(\beta+1)}}dZ^{2},$ (31) where $x+x_{0}=X$, $t=T$, $y=Y$, $cz=Z$. The expressions for effective pressure $\bar{p}$ and density $\rho$ for the model (31) are given by $8\pi\bar{p}=\frac{1}{a^{2}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2n(1-\alpha)}{(\beta+1)}}}\Biggl{[}\frac{b^{2}\\{2n(1-\alpha^{2})+2\beta+2\alpha(\beta-\alpha)(1-\alpha)\\}}{2(\beta+1)^{2}}\times$ $\frac{(c_{1}e^{bT}-c_{2}e^{-bT})^{2}}{(c_{1}e^{bT}+c_{2}e^{-bT})^{2}}-\frac{b^{2}(3\alpha+1)}{2(\beta+1)}+K^{2}X^{2}\Biggr{]}-\Lambda,$ (32) $8\pi\rho=\frac{1}{a^{2}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2n(1-\alpha)}{(\beta+1)}}}\Biggl{[}\frac{b^{2}\\{2n(1-\alpha^{2})+2\alpha+(\beta-\alpha)(1-\alpha)\\}}{2(\beta+1)^{2}}\times$ $\frac{(c_{1}e^{bT}-c_{2}e^{-bT})^{2}}{(c_{1}e^{bT}+c_{2}e^{-bT})^{2}}-\frac{b^{2}(1-\alpha)}{2(\beta+1)}-K(2+3KX^{2})\Biggr{]}+\Lambda.$ (33) For the specification of $\xi$, we assume that the fluid obeys an equation of state of the form $p=\gamma\rho,$ (34) where $\gamma(0\leq\gamma\leq 1)$ is a constant. Thus, given $\xi(t)$ we can solve for the cosmological parameters. In most of the investigation involving bulk viscosity it is assumed to be a simple power function of the energy density [69]$-$[73] $\xi(t)=\xi_{0}\rho^{m},$ (35) where $\xi_{0}$ and $m$ are constants. For small density, $m$ may even be equal to unity as used in Murphy’s work [65] for simplicity. If $m=1$, (35) may correspond to a radiative fluid [73]. Near a big bang, $0\leq m\leq\frac{1}{2}$ is a more appropriate assumption [74] to obtain realistic models. $8\pi(p-\xi_{0}\rho^{m}\theta)=\frac{1}{a^{2}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2n(1-\alpha)}{(\beta+1)}}}\Biggl{[}\frac{b^{2}\\{2n(1-\alpha^{2})+2\beta+2\alpha(\beta-\alpha)(1-\alpha)\\}}{2(\beta+1)^{2}}\times$ $\frac{(c_{1}e^{bT}-c_{2}e^{-bT})^{2}}{(c_{1}e^{bT}+c_{2}e^{-bT})^{2}}-\frac{b^{2}(3\alpha+1)}{2(\beta+1)}+K^{2}X^{2}\Biggr{]}-\Lambda,$ (36) where $\theta$ is the scalar of expansion calculated for the flow vector $u^{i}$ and is given by $\theta=\frac{K_{2}}{(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{n(1-\alpha)}{(\beta+1)}}}\frac{(c_{1}e^{bT}-c_{2}e^{-bT})}{(c_{1}e^{bT}+c_{2}e^{-bT})},$ (37) where $K_{2}=\frac{b\\{n(1-\alpha)+(1+\alpha)\\}}{(\beta+1)a}$ (38) For simplicity and realistic models of physical importance, we consider the following two cases $(m=0,1)$. On using (35) in (32), we obtain ### 3.1 Model I: Solution when $m=0$ When $m=0$, Eq. (35) reduces to $\xi=\xi_{0}$. With the use of Eqs. (33), (34) and (37), Eq. (36) reduces to $8\pi(1+\gamma)\rho=\frac{1}{a^{2}T_{1}^{\frac{2n(1-\alpha)}{(1+\beta)}}}\Biggl{[}\frac{b^{2}}{2(\beta+1)^{2}}\Big{\\{}4n(1-\alpha^{2})+2(\alpha+\beta)\,+$ $(\beta-\alpha)(1-\alpha)(2\alpha+1)\Big{\\}}\left(\frac{T_{2}}{T_{1}}\right)^{2}-\,\frac{b^{2}(\alpha+1)}{(\beta+1)}-2K(1+KX^{2})\Biggr{]}+\,\frac{8\pi\xi_{0}K_{2}}{T_{1}^{\frac{n(1-\alpha)}{(1+\beta)}}}\left(\frac{T_{2}}{T_{1}}\right).$ (39) Eliminating $\rho(t)$ between (33) and (39), we get $(1+\gamma)\Lambda=\frac{1}{a^{2}T_{1}^{\frac{2n(1-\alpha)}{(1+\beta)}}}\Biggl{[}\frac{b^{2}}{2(\beta+1)^{2}}\Big{\\{}2n(1-\alpha^{2})(1-\gamma)+2(\beta-\alpha\gamma)\,+$ $(\beta-\alpha)(1-\alpha)(2\alpha-\gamma)\Big{\\}}\left(\frac{T_{2}}{T_{1}}\right)^{2}+\frac{b^{2}\\{(\gamma-1)-\alpha(\gamma+3)\\}}{2(\beta+1)}$ $+\,K\\{KX^{2}+\gamma(2+3KX^{2})\\}\Biggr{]}+\frac{8\pi\xi_{0}K_{2}}{T_{1}^{\frac{n(1-\alpha)}{(1+\beta)}}}\left(\frac{T_{2}}{T_{1}}\right),$ (40) where $T_{1}=(c_{1}e^{bT}+c_{2}e^{-bT}),$ $T_{2}=(c_{1}e^{bT}-c_{2}e^{-bT}).$ Figure 1: The plot of energy density $\rho(T)$ Vs. T Figure 2: The plot of cosmological term $\Lambda(T)$ Vs. T ### 3.2 Model II: Solution when $m=1$ When $m=1$, Eq. (35) reduces to $\xi=\xi_{0}\rho$. With the use of Eqs. (33), (34) and (37), Eq. (36) reduces to $8\pi\rho=\frac{1}{a^{2}T_{1}^{\frac{2n(1-\alpha)}{(1+\beta)}}\left\\{1+\gamma-\frac{\xi_{0}K_{2}}{T_{1}^{\frac{n(1-\alpha)}{(1+\beta)}}}\left(\frac{T_{2}}{T_{1}}\right)\right\\}}\Biggl{[}\frac{b^{2}}{2(\beta+1)^{2}}\Big{\\{}4n(1-\alpha^{2})+2(\alpha+\beta)\,+$ $(\beta-\alpha)(1-\alpha)(2\alpha+1)\Big{\\}}\left(\frac{T_{2}}{T_{1}}\right)^{2}-\,\frac{b^{2}(\alpha+1)}{(\beta+1)}-2K(1+KX^{2})\Biggr{]}.$ (41) Eliminating $\rho(t)$ between (33) and (41), we get $\Lambda=\frac{1}{a^{2}T_{1}^{\frac{2n(1-\alpha)}{(1+\beta)}}\left\\{1+\gamma-\frac{\xi_{0}K_{2}}{T_{1}^{\frac{n(1-\alpha)}{(1+\beta)}}}\left(\frac{T_{2}}{T_{1}}\right)\right\\}}\Biggl{[}\frac{b^{2}}{2(\beta+1)^{2}}\Big{\\{}4n(1-\alpha^{2})+2(\alpha+\beta)\,+$ $(\beta-\alpha)(1-\alpha)(2\alpha+1)\Big{\\}}\left(\frac{T_{2}}{T_{1}}\right)^{2}-\,\frac{b^{2}(\alpha+1)}{(\beta+1)}-2K(1+KX^{2})\Biggr{]}.$ $-\,\frac{1}{a^{2}T_{1}^{\frac{2n(1-\alpha)}{(1+\beta)}}}\Biggl{[}\frac{b^{2}}{2(\beta+1)^{2}}\Big{\\{}2n(1-\alpha^{2})+2\alpha+$ $(\beta-\alpha)(1-\alpha)\Big{\\}}\left(\frac{T_{2}}{T_{1}}\right)^{2}-\frac{b^{2}(1-\alpha)}{2(\beta+1)}-K(2+3KX^{2})\Biggr{]}.$ (42) Figure 3: The plot of energy density $\rho(T)$ Vs. T Figure 4: The plot of cosmological term $\Lambda(T)$ Vs. T From Eqs. (39) and (41) , we note that $\rho(t)$ is a decreasing function of time and $\rho>0$ for all times. This behaviour is clearly depicted in Figures $1$ and $3$ as a representative case with appropriate choice of constants of integration and other physical parameters using reasonably well known situations. Figures $1$ and $3$ show this physical behaviours of energy density as a decreasing functions of coordinate time $T$ of Model I and II. Here the coordinate time $T$ is related to normal time as shown in Ref.[19]. This also follows for rest part of the paper. In spite of homogeneity at large scale our universe is inhomogeneous at small scales, so physical quantities being position dependent are more natural in our observable universe if we do not go to super high scale. This result shows this kind of physical importance. In recent time the $\Lambda$-term has interested theoreticians and observers for various reasons. The nontrivial role of the vacuum in the early universe generate a $\Lambda$-term that leads to inflationary phase. Observationally, this term provides an additional parameter to accommodate conflicting data on the values of the Hubble constant, the deceleration parameter, the density parameter and the age of the universe (for example, see the references [75, 76]). Assuming that $\Lambda$ owes its origin to vacuum interactions, as suggested in particular by Sakharov [77], it follows that it would in general be a function of space and time coordinates, rather than a strict constant. In a homogeneous universe $\Lambda$ will be at most time dependent [78]. In our case this approach can generate $\Lambda$ that varies both with space and time. In considering the nature of local massive objects, however, the space dependence of $\Lambda$ cannot be ignored. For details discussion, the readers are advised to see the references (Narlikar, Pecker and Vigier [79], Ray and Ray [80], Tiwari, Ray and Bhadra [81]). The behaviour of the universe in this model will be determined by the cosmological term $\Lambda$ ; this term has the same effect as a uniform mass density $\rho_{eff}=-\Lambda/4\pi G$, which is constant in space and time. A positive value of $\Lambda$ corresponds to a negative effective mass density (repulsion). Hence, we expect that in the universe with a positive value of $\Lambda$, the expansion will tend to accelerate; whereas in the universe with negative value of $\Lambda$, the expansion will slow down, stop and reverse. From Eqs. (40) and (42), we see that the cosmological term $\Lambda$ is a decreasing function of time and it approaches a small positive value at late time. From Figures 2 and 4, we note this behaviour of cosmological term $\Lambda$ in both models I and II. Recent cosmological observations suggest the existence of a positive cosmological constant $\Lambda$ with the magnitude $\Lambda(G\hbar/c^{3})\approx 10^{-123}$. These observations on magnitude and red-shift of type Ia supernova suggest that our universe may be an accelerating one with induced cosmological density through the cosmological $\Lambda$-term. Thus, our model is consistent with the results of recent observations. Some Physical and Geometric Features : The non-vanishing component $F_{12}$ of electromagnetic field tensor is obtained as $F^{2}_{12}=\frac{\bar{\mu}}{8\pi}\frac{b^{2}(1-\alpha)}{(\beta+1)}e^{KX^{2}}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2}{(\beta+1)}}\times$ $\Biggl{[}1-\frac{(\beta-\alpha)}{(\beta+1)}\frac{(c_{1}e^{bT}-c_{2}e^{-bT})^{2}}{(c_{1}e^{bT}+c_{2}e^{-bT})^{2}}\Biggr{]}.$ (43) The expressions for the shear scalar $\sigma^{2}$, acceleration vector $\dot{u}_{i}$ and proper volume $V^{3}$ for model (31) are given by $\sigma^{2}=\frac{b^{2}\left[\\{n(1-\alpha)+(1+\alpha)\\}^{2}-3n(1-\alpha)(1+\alpha)-3\alpha\right]}{3(\beta+1)^{2}a^{2}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2n(1-\alpha)}{(\beta+1)}}}\frac{(c_{1}e^{bT}-c_{2}e^{-bT})^{2}}{(c_{1}e^{bT}+c_{2}e^{-bT})^{2}},$ (44) $\dot{u}_{i}=(0,0,0,0),$ (45) $V^{3}=\sqrt{-g}=a^{2}(c_{1}e^{bT}+c_{2}e^{-bT})^{\frac{2n(1-\alpha)+(1+\alpha)}{(\beta+1)}},$ (46) From Eqs. (44) and (37), we have $\frac{\sigma^{2}}{\theta^{2}}=\frac{\left[\\{n(1-\alpha)+(1+\alpha)\\}^{2}-3n(1-\alpha^{2})-3\alpha\right]}{3\\{n(1-\alpha)+(1+\alpha)\\}^{2}}=\mbox{ constant}.$ (47) The rotation $\omega$ is identically zero. From set of equations (37), (43) - (47), the model brings out the following features: The model starts expanding at $T>0$ and goes on expanding indefinitely when $\frac{n(1-\alpha)}{(\beta+1)}<0$. The model (31) represents an expanding, shearing and non-rotating universe in which the flow vector is geodetic. Since $\frac{\sigma}{\theta}$ = constant, the model does not approach isotropy. As $T$ increases the proper volume also increases. The model is non-accelerating. The physical quantities $p$ and $\rho$ decrease as $F_{12}$ increases. However, if $\frac{n(1-\alpha)}{(\beta+1)}>0$, the process of contraction starts at $T>0$ and at $T=\infty$ the expansion stops. The electromagnetic field tensor does not vanish when $b\neq 0$, and $\alpha\neq 1$. ## 4 Solution in Absence of Magnetic Field In absence of magnetic field the Einstein field equations for metric (1) read as $\frac{1}{A^{2}}\left[-\frac{B_{44}}{B}-\frac{C_{44}}{C}+\frac{A_{4}}{A}\left(\frac{B_{4}}{B}+\frac{C_{4}}{C}\right)-\frac{B_{4}C_{4}}{BC}+\frac{B_{1}C_{1}}{BC}\right]=8\pi p+\Lambda,$ (48) $\frac{1}{A^{2}}\left(\frac{A^{2}_{4}}{A^{2}}-\frac{A_{44}}{A}-\frac{C_{44}}{C}+\frac{C_{11}}{C}\right)=8\pi p+\Lambda,$ (49) $\frac{1}{A^{2}}\left(\frac{A^{2}_{4}}{A^{2}}-\frac{A_{44}}{A}-\frac{B_{44}}{B}+\frac{B_{11}}{B}\right)=8\pi p+\Lambda,$ (50) $\frac{1}{A^{2}}\left[-\frac{B_{11}}{B}-\frac{C_{11}}{C}+\frac{A_{4}}{A}\left(\frac{B_{4}}{B}+\frac{C_{4}}{C}\right)-\frac{B_{1}C_{1}}{BC}+\frac{B_{4}C_{4}}{BC}\right]=8\pi\rho-\Lambda,$ (51) $\frac{B_{14}}{B}+\frac{C_{14}}{C}-\frac{A_{4}}{A}\left(\frac{B_{1}}{B}+\frac{C_{1}}{C}\right)=0,$ (52) Eqs. (49) and (50) lead to $\frac{B_{44}}{B}-\frac{B_{11}}{B}-\frac{C_{44}}{C}+\frac{C_{11}}{C}=0.$ (53) Eqs. (19) and (53) lead to $\frac{g_{44}}{g}-\frac{k_{44}}{k}=0.$ (54) Eqs. (23) and (54) lead to $\frac{g_{44}}{g}+\alpha\frac{g^{2}_{4}}{g^{2}}=0,$ (55) which on integration gives $g=(c_{4}t+c_{5})^{\frac{1}{(\alpha+1)}},$ (56) where $c_{4}$ and $c_{5}$ are constants of integration. Hence from (23) and (56), we have $k=c(c_{4}t+c_{5})^{\frac{\alpha}{(\alpha+1)}}.$ (57) In this case (17) leads to $f=\exp{\left(\frac{1}{2}K(x+x_{0})^{2}\right)}.$ (58) Therefore, we have $B=\exp{\left(\frac{1}{2}K(x+x_{0})^{2}\right)}(c_{4}t+c_{5})^{\frac{1}{(\alpha+1)}},$ (59) $C=\exp{\left(\frac{1}{2}K(x+x_{0})^{2}\right)}c(c_{4}t+c_{5})^{\frac{\alpha}{(\alpha+1)}},$ (60) $A=a(c_{4}t+c_{5})^{\frac{n(1-\alpha)}{(1+\alpha)}},$ (61) where $a$ is already defined in previous section. After using suitable transformation of the co-ordinates, the model (1) reduces to the form $ds^{2}=a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}(dX^{2}-dT^{2})+e^{KX^{2}}(c_{4}T)^{\frac{2}{(\alpha+1)}}dY^{2}$ $+e^{KX^{2}}(c_{4}T)^{\frac{2\alpha}{(\alpha+1)}}dZ^{2},$ (62) where $x+x_{0}=X$, $y=Y$, $cz=Z$, $t+\frac{c_{5}}{c_{4}}=T$. The expressions for effective pressure $\bar{p}$ and density $\rho$ for the model (62) are given by $8\pi\bar{p}=\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}}\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}+K^{2}X^{2}\right]-\Lambda,$ (63) $8\pi\rho=\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}}\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}-K(2+3KX^{2})\right]+\Lambda.$ (64) On using (35) in (63), we obtain $8\pi(p-\xi_{0}\rho^{m}\theta)=\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}}\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}+K^{2}X^{2}\right]-\Lambda,$ (65) where $\theta$, in this case, is calculated for the flow vector $u^{i}$ and is given by $\theta=\frac{K_{3}}{T^{\frac{n(1-\alpha)+(1+\alpha)}{(1+\alpha)}}},$ (66) where $K_{3}=\frac{n(1-\alpha)+(1+\alpha)}{a(1+\alpha)c_{4}^{\frac{n(1-\alpha)}{(1+\alpha)}}}.$ (67) ### 4.1 Model I: Solution when $m=0$ When $m=0$, Eq. (35) reduces to $\xi=\xi_{0}$. With the use of Eqs. (64), (34) and (66), Eq. (65) reduces to $4\pi(1+\gamma)\rho=\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}}\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}-K(1+KX^{2})\right]+\frac{4\pi\xi_{0}K_{3}}{T^{\frac{n(1-\alpha)+(1+\alpha)}{(1+\alpha)}}}.$ (68) Figure 5: The plot of energy density $\rho(T)$ Vs. T Figure 6: The plot of cosmological term $\Lambda(T)$ Vs. T Eliminating $\rho(t)$ between (64) and (68), we get $(1+\gamma)\Lambda=\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}}\left[\frac{\\{n(1-\alpha^{2})+\alpha\\}}{(\alpha+1)^{2}}\frac{(1-\gamma)}{T^{2}}+K\\{KX^{2}(1+3\gamma)+2\gamma\\}\right]$ $+\,\frac{8\pi\xi_{0}K_{3}}{T^{\frac{n(1-\alpha)+(1+\alpha)}{(1+\alpha)}}}.$ (69) From Eq. (68), we see that $\rho(t)$ is a decreasing function of time and $\rho>0$ for all times. Figure $5$ shows this behaviour of energy density in Model I. From Eq. (69), we observe that the cosmological term $\Lambda$ is a decreasing function of time and it approaches a small positive value at late time. From Figure $6$, we note this behaviour of cosmological term $\Lambda$ in both model I. ### 4.2 Model II: Solution when $m=1$ When $m=1$, Eq. (35) reduces to $\xi=\xi_{0}\rho$. With the use of Eqs. (64), (34) and (66), Eq. (65) reduces to $4\pi\rho=\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}\left\\{1+\gamma-\xi_{0}\frac{K_{3}}{T^{\frac{n(1-\alpha)+(1+\alpha)}{(1+\alpha)}}}\right\\}}\times$ $\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}-K(1+KX^{2})\right].$ (70) Figure 7: The plot of (a) $\rho$ as a function of T, (b) $\Lambda$ as a function of T Eliminating $\rho(t)$ between (64) and (70), we get $\Lambda=\frac{2}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}\left\\{1+\gamma-\xi_{0}\frac{K_{3}}{T^{\frac{n(1-\alpha)+(1+\alpha)}{(1+\alpha)}}}\right\\}}\times$ $\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}-K(1+KX^{2})\right]$ $-\,\frac{1}{a^{2}(c_{4}T)^{\frac{2n(1-\alpha)}{(1+\alpha)}}}\left[\frac{n(1-\alpha^{2})+\alpha}{(\alpha+1)^{2}}\frac{1}{T^{2}}-K(2+3KX^{2})\right]$ (71) From Eq. (70) and also from Figure $7(a)$, it seems that in very early stage of universe the energy density could be negative and may be link with some of the early universe physics which is the domain of quantum cosmology or early quantum mechanics. So we do not infer any things about energy density. But note that once energy density is negative in the initial stage even after oscillations it returns back to small negative value at the later stage of the evolution. Hence the details of macro physics in early universe decides its fate. So we cannot infer from this model. Similar behaviours is also reflected about the cosmological constant $\Lambda$. We do not know for wide range of parameters this behaviour holds. It may be feasible that Figures $7(a,b)$ may coincide with Figures $3$ and $4$. But it does not seem to be vary generic behaviour of generalized model of section $3.2$. So we do not go to any conclusion about its applicability in general. But it seems that magnetic field and bulk viscosity prevent this kind of unusual behaviours as it is clear from Figures $1$ to $4$. Some Physical and Geometric Features: The expressions for the shear scalar $\sigma^{2}$, acceleration vector $\dot{u}_{i}$ and proper volume $V^{3}$ for model (62) are given by $\sigma^{2}=\frac{\\{n(1-\alpha)+(1+\alpha)\\}^{2}-3n(1-\alpha^{2})-3\alpha}{3(1+\alpha)^{2}a^{2}c_{4}^{\frac{2n(1-\alpha)}{(1+\alpha)}}T^{\frac{2n(1-\alpha)+2(1+\alpha)}{(1+\alpha)}}},$ (72) $\dot{u}_{i}=(0,0,0,0),$ (73) $V^{3}=\sqrt{-g}=a^{2}e^{KX^{2}}(c_{4}T)^{\frac{2n(1-\alpha)+(1+\alpha)}{(1+\alpha)}}.$ (74) From (72) and (66), we obtain $\frac{\sigma^{2}}{\theta^{2}}=\frac{\\{n(1-\alpha)+(1+\alpha)\\}^{2}-3n(1-\alpha^{2})-3\alpha}{3\\{n(1-\alpha)+(1+\alpha)\\}^{2}}.$ (75) The rotation $\omega$ is identically zero. The model in absence of magnetic field starts expanding with a big bang at $T=0$ and it stops expanding at $T=\infty$. In absence of magnetic field, the model in general represents an expanding, shearing and non-rotating in which the flow vector is geodetic. Since $\frac{\sigma}{\theta}$ = constant, the model does not approach isotropy. As $T$ increases the proper volume also increases. The model is non-accelerating. ## 5 Concluding Remarks We have obtained a new cylindrically symmetric inhomogeneous cosmological model of electro-magnetic bulk viscous fluid as the source of matter. Generally the model represents expanding, shearing and non- rotating universe in which the flow vector is geodetic. It is worth mention here that in presence of magnetic field the model (31) is expanding whereas in absence of magnetic field the model (62) starts with a big bang singularity. In both models $\frac{\sigma}{\theta}=$ constant and hence they do not approach isotropy. The models are non-accelerating. In these solutions all physical quantities depend on at most one space co-ordinate and time. It is important to note here that both the models (31) and (62) in presence and absence of magnetic field reduce to homogeneous universe when $K=0$. This shows that for $K=0$, inhomogeneity dies out. The effect of bulk viscosity is to produce a change in perfect fluid and hence exhibit essential influence on the character of the solution. The effect is clearly visible on the $p$ effective (see details in previous sections). In Section $3$, we have shown regular well behaviour of energy density, cosmological constant ($\Lambda$) and the expansion of the universe with parameter $T$. The section $4$ is a toy investigation to see that the effect of bulk viscosity plays dynamic role in the evolution equations. We also observe that Murphy’s conclusion [65] about the absence of a big bang type singularity in the infinite past in models with bulk viscous fluid, in general, is not true. The results obtained by Myung and Cho [60] also show that, it is, in general, not valid, since for some cases big bang singularity occurs in finite past. In presence and absence of magnetic field, the cosmological terms in models are decreasing function of time and approach a small value at late time (with exception $m=1$ in the absence of magnetic field). The values of cosmological “constant” for the models are found to be small and positive, as obtained in recent results from the supernovae observations recently obtained by the High-Z supernovae Ia Team and Supernovae Cosmological Project. Our solutions generalize the solutions recently obtained by Pradhan et al. [66]. ## Acknowledgments Authors would like to thank the Inter-University Centre for Astronomy and Astrophysics (IUCAA), Pune, India for providing facility and support where this work was carried out. A. P. and K. J. are visiting associates of IUCAA. ## References [1] R. C. Tolman, Proc. Nat. Acad. Sci. 20 (1934) 169. * [2] H. Bondi, Mon. Not. R. Astro. Soc. 107 (1947) 410. * [3] A. H. Taub, Ann. Math. 53 (1951) 472. * [4] A. H. Taub, Phy. Rev. 103 (1956) 454. * [5] N. Tomimura, II Nuovo Cimento B 44 (1978) 372. * [6] P. Szekeres, Commun. Math. Phys. 41 (1975) 55. * [7] C. B. Collins and D. A. Szafron, J. Math. Phy. 20 (1979) 2347; J. Math. Phy. 20 (1979b) 2362. * [8] D. A. Szafron and C. B. Collins, J. Math. Phy. 20 (1979) 2354. * [9] J. M. M. Senovilla, Phy. Rev. Lett. 64 (1990) 2219. * [10] E. Ruiz and J. M. M. Senovilla, Phy. Rev. D 45 (1990) 1995. * [11] N. Dadhich, R. Tikekar and L. K. Patel, Curr. Sci. 65 (1993) 694. * [12] L. K. Patel, R. Tikekar and N. Dadhich, Pramana-J. Phys. 49 (1993) 213. * [13] G. Singh, P. Mehta and S. Gupta, Astrophys. Space Sci. 281 (1990) 677. * [14] A. Pradhan, V. K. Yadav and N. N. Saste, Int. J. Mod. Phys. D 11 (2002) 857; A. Pradhan and H. R. Pandey, Int. J. Mod. Phys. D 12 (2003) 941; A. Pradhan and P. Pandey, Czech. J. Phys. 55 (2005) 749; A. Pradhan, K. D. Thengane and J. K. Jumale, Spacetime & Substance 4(29) (2005) 145; A. Pradhan, P. Pandey and S. K. Singh, Int. J. Theor. Phys. 46 (2007) 1584; A. Pradhan, P. Pandey, K. Jotania and M. K. Yadav, Int. J. Theor. Phys. 46 (2007) 2774. * [15] S. R. Roy and P. N. Singh, J. Phys. A - Mathematical and General, 9 (1986) 255. * [16] R. Bali and A. Tyagi, Gen. Rel. Grav. 21 (1989) 797. * [17] R. Bali and A. Tyagi, Astrophys. Space Sci, 138 (1987) 173. * [18] I. Chakrabarty, A. Pradhan and N. N. Saste, Int. J. Mod. Phys. D 10 (2001) 741. * [19] A. Pradhan, P. K. Singh and K. Jotania, Czech. J. Phys. 56 (2006) 641. * [20] Ya. B. Zeldovich, A. A. Ruzmainkin and D. D. Sokoloff, Magnetic field in Astrophysics, Gordon and Breach, New York (1983). * [21] E. R. Harrison, Phys. Rev. Lett. 30 (1973) 188. * [22] H. P. Robertson and A. G. Walker, Proc. London Math. Soc. 42 (1936) 90. * [23] C. W. Misner, K. S. Thorne and J. A. Wheeler, Gravitation, W. H. Freeman, New York (1973). * [24] E. Asseo and H. Sol, Phys. Rep. 6 (1987) 148. * [25] M. A. Melvin, Ann. New York Acad. Sci. 262 (1975) 253. * [26] R. Pudritz and J. Silk, Astrophys. J. 342 (1989) 650. * [27] K. T. Kim, P. G. Tribble and P. P. Kronberg, Astrophys. J. 379 (1991) 80. * [28] R. Perley and G. Taylor, Astrophys. J. 101 (1991) 1623. * [29] P. P. Kronberg, J. J. Perry and E. L. Zukowski, Astrophys. J. 387 (1991) 528. * [30] A. M. Wolfe, K. Lanzetta and A. L. Oren: , Astrophys. J. 388 (1992) 17. * [31] R. Kulsrud, R. Cen, J. P. Ostriker and D. Ryu, Astrophys. J. 380 (1997) 481. * [32] E. G. Zweibel and C. Heiles, Nature 385 (1997) 131. * [33] J. D. Barrow, Phys. Rev. D 55 (1997) 7451. * [34] Ya. B. Zeldovich, Sov. Astron. 13 (1970) 608. * [35] M. S. Turner and L. M. Widrow, Phys. Rev. D 30 (1988) 2743. * [36] J. Quashnock, A. Loeb and D. N. Spergel, Astrophys. J. 344, (1989) L49. * [37] B. Ratra, Astrophys. J. 391 (1992) L1. * [38] A. D. Dolgov and J. Silk, Phys. Rev. D 47 (1993) 3144. * [39] A. D. Dolgov, Phys. Rev. D 48 (1993) 2499. * [40] R. Bali and M. Ali, Pramana-J. Phys. 47 (1996) 25. * [41] A. Pradhan and O. P. Pandey, Int. J. Mod. Phys. D 7 (2003) 1299; A. Pradhan, S. K. Srivastav and K. R. Jotania, Czech. J. Phys. 54 (2004) 255; A. Pradhan and S. K. Singh, Int. J. Mod. Phys. D 13 (2004) 503; A. Pradhan, P. Pandey and K. K. Rai, Czech. J. Phys. 56 (2006) 303. * [42] B. Ratra and P. J. E. Peebles, Phys. Rev. D 37 (1988) 3406. * [43] V. Sahni and A. Starobinsky, Int. J. Mod. Phys. D 9 (2000) 373. * [44] S. M. Carroll, W. H. Press and E. L. Turner, Ann. Rev. Astron. Astrophys. 30 (1992) 499. * [45] P. J. E. Peebles, Rev. Mod. Phys. 75 (2003) 559. * [46] T. Padmanabhan, Phys. Rep. 380 (2003) 235. T. Padmanabhan, Dark Energy and Gravity, gr-qc/0705.2533 (2007). * [47] S. Perlmutter et al., Astrophys. J. 483 (1997) 565; S. Perlmutter et al., Nature 391 (1998) 51; S. Perlmutter et al., Astrophys. J. 517 (1999) 565. * [48] A. G. Riess et al., Astron. J. 116 (1998) 1009; A. G. Riess et al., Astrophys. J. 560 (2001) 49; A. G. Riess et al., Astrophys. J. 607 (2004) 665; A. G. Riess et al., Astrophys. J. 659 (2007) 98. * [49] P. M. Garnavich et al., Astrophys. J. 493 (1998) L53; P. M. Garnavich et al., Astrophys. J. 509 (1998) 74. * [50] B. P. Schmidt et al., Astrophys. J. 507 (1998) 46. * [51] J. P. Blakeslee et al., Astrophys. J. 589 (2003) 693. * [52] P. Astier et al., Astron. Astrophys. 447 (2006) 31. * [53] D. N. Spergel et al., Astrophys. J. Suppl. 170 (2007) 377. * [54] D. J. Eisentein et al., Astrophys. J. 633 (2005) 560. * [55] S. M. Carroll, eConf C0307282, TTH09 (2003) [AIP Conf. Proc. 743 (2005) 16]. * [56] E. W. Kolb, S. Matarrese and A. Riotto, New J. Phys. 8 (2006) 322; E. W. Kolb, S. Matarrese, A. Notari and A. Riotto, arXiv:hep-th/0503117. * [57] A. Upadhye, M. Ishak and P. J. Steinhardt, Phys. Rev. D 72 (2005) 063501. * [58] R. K. Knop et al., Astrophys. J. 598 (2003) 102. * [59] E. W. Kolb and M. S. Turner, The Early Universe, Addison - Wesley, U S A (1990). * [60] S. Myung and B. M. Cho, Mod. Phys. Lett. A 1 (1986) 37. * [61] N. Turok, Phys. Rev. Lett. 60 (1988) 549. * [62] J. D. Barrow, Nucl. Phys. B 310 (1988) 243. * [63] C. Wolf, S.-Afr. Tydskr. 14 (1991) 68. * [64] Ø. Grøn, Astrophys. Space Sci. 173 (1990) 191. * [65] G. L. Murphy, Phys. Rev. D 8 (1973) 4231. * [66] A. Pradhan, V. Yadav, L. Yadav and A. K. Yadav, Astrophys. Space Sci. 312 (2007) 267. * [67] A. Lichnerowicz, Relativistic Hydrodynamics and Magnetohydrodynamics, W. A. Benzamin. Inc. New York, Amsterdam (1967) p. 93. * [68] J. L. Synge, Relativity: The General Theory, North-Holland Publ., Amsterdam (1960) p. 356. * [69] R. Maartens, Class. Quant. Grav. 12 (1995) 1455. * [70] D. Pavon, J. Bafaluy and D. Jou, Class. Quant. Grav. 8 (1991) 357; D. Pavon, J. Bafaluy and D. Jou, in Proc. Hanno Rund Conf. on Relativity and Thermodynamics, ed. S. D. Maharaj, University of Natal, Durban (1996) p. 21. * [71] N. O. Santos, R. S. Dias and A. Benerjee, J. Math. Phys. 26 (1985) 878. * [72] W. Zimdahl, Phys. Rev. D 53 (1996) 5483. * [73] S. Weinberg, Graviatation and Cosmology, Wiley, New York, (1972), p. 57. * [74] U. A. Belinskii and I. M. Khalatnikov, Sov. Phys. JETP 42 (1976) 205. * [75] J. Gunn and B. M. Tinsley, Nature, 257 (1975) 454. * [76] E. J. Wampler and W. L. Burke, in New Ideas in Astronomy, eds. F. Bertola, J. W. sulentic, and B. F. Madore, Cambridge University Press (1988) p. 317. * [77] A. D. Sakharov, Doklady Akad. Nauk. SSSR 177 (1968) 70 (translation, Soviet Phys. Doklady, 12 (1968)) 1040. * [78] P. J. E. Peebles and B. Ratra, Astrophys. J. 325 (1988) L17. * [79] J. V. Narlikar, J. -C. Pecker and J. -P. Vigier, J. Astrophys. Astr. 12 (1991) 7. * [80] S. Ray and D. Ray, Astrophys. Space Sci. 203 (1993) 211. * [81] R. N. Tiwari, S. Ray and S. Bhadra, Indian Pure Appl. Math. 31 (2000) 1017.	arxiv-papers	2008-07-26T11:08:18	2024-09-04T02:48:56.999869	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Anirudh Pradhan, Vandana Rai and Kanti Jotania", "submitter": "Dr. Anirudh Pradhan", "url": "https://arxiv.org/abs/0807.4230" }
0807.4240	# Lepton Flavor Violating $l\to l^{\prime}\gamma$ and $Z\to l\bar{l}^{\prime}$ Decays Induced by Scalar Leptoquarks Rachid Benbrik1,2111Email: rbenbrik@phys.cycu.edu.tw Address after Aug. 2008: Department of Physics, National Cheng Kung University, Tainan 701, Taiwan. Chun-Khiang Chua1222Email: ckchua@phys.cycu.edu.tw 1Department of Physics, Chung Yuan Christian University Chung-Li, Taiwan 320, Republic of China 2LPHEA, Département de Physique,Faculté des Sciences-Semlalia B.P 2390 Marrakech, Morocco. ###### Abstract Motivated by the recent muon $g-2$ data, we study the lepton flavor violating $l\to l^{\prime}\gamma$ and $Z\to l\bar{l}^{\prime}$ ($l,l^{\prime}=e,\mu,\tau$) decays with $l\neq l^{\prime}$ in a scalar leptoquark model. Leptoquarks can produce sizable LFV $l\to l^{\prime}\gamma$ decay rates that can be easily reached by present or near future experiments. Leptoquark masses and couplings are constrained by the muon $g-2$ data and the current $l\to l^{\prime}\gamma$ bounds. We predict $Br(Z\to\tau^{\mp}e^{\pm})$ reaching the present limit ($10^{-5}$) and $Br(Z\to\mu^{\mp}\tau^{\pm})$ reaching $2\times 10^{-8}$, which will be accessible by future linear colliders, whereas, the current bounds on LFV impose very strong constraints on the $Br(Z\to\mu^{\mp}e^{\pm})$ and the ratio is too low to be observed in the near future. ###### pacs: 14.80.-j, 13.40.Em, 13.35.-r, 13.38.Dg ## I Introduction The excess value of the anomalous magnetic moment of muon was reported by the E821 collaboration at BNL [1] $a^{{\rm exp}}_{\mu}=116\,592\,080(63)\times 10^{-11}.$ (1) The Standard Model prediction for $a^{SM}_{\mu}$ with QED, hadronic and electroweak contributions is [2, 3] $a^{{\rm SM}}_{\mu}=116\,591\,785(61)\times 10^{-11}.$ (2) with the experimental value of $(g-2)/2$, the comparison gives $\Delta a_{\mu}\equiv a^{{\rm exp}}_{\mu}-a^{{\rm SM}}_{\mu}=(295\pm 87.7)\times 10^{-11},\qquad(3.4\sigma)$ (3) The 3.4 standard deviation difference between the two, may be a hint of new physics contribution. It has been shown that contributions from leptoquark (LQ) exchanges are capable to resolve the above deviation [4, 5, 6]. Leptoquarks are vector or scalar particles carrying both lepton and baryon numbers. LQs can be quite naturally introduced in the low-energy theory as a relic of a more fundamental theory at some high-energy scale, such as grand unified theories (GUT) [7, 8]. In some models, it is possible to have leptoquarks at TeV scale [9]. The low- energy LQ phenomenology has received considerable attention. Possible LQ manifestations in various processes have been extensively investigated [9]-[25]. Various constraints on LQ masses and couplings have been deduced from existing experimental data and prospects for the forthcoming experiments have been estimated. Direct searches of LQs as s-channel resonances in deep inelastic ep-scattering and pair production in hadron colliders placed lower limits on their mass $M_{LQ}\geq 73-298$GeV [18] depending on the LQ types and couplings. The interest on leptoquarks has been renewed during the last few years since ongoing collider experiments have good prospects for searching these particles [27]. For a recent review of leptoquarks, one is referred to [28]. Lepton Flavor Violation (LFV) are powerful tools to search for new physics. The present experimental limits give [18]: $\displaystyle{\rm Br}(\mu\to e\gamma)$ $\displaystyle<$ $\displaystyle 1.2\times 10^{-11},$ (4) $\displaystyle{\rm Br}(\tau\to e\gamma)$ $\displaystyle<$ $\displaystyle 1.1\times 10^{-7},$ (5) $\displaystyle{\rm Br}(\tau\to\mu\gamma)$ $\displaystyle<$ $\displaystyle 6.8\times 10^{-8}.$ (6) Since effects of leptoquark interactions can manifest in $a_{\mu}$, it is very likely that they can also give interesting contributions to these $l\to l^{\prime}\gamma$ processes [5, 6]. There are considerable efforts on experiments that aim at pushing the sensitivity of ${\rm Br}(\mu\to e\gamma)$ down by two order of magnitudes [29]. B factories and the upgraded super B factory can probe the $\tau\to e\gamma,\,\mu\gamma$ decays at better sensitivities. The $Z\to\ell\bar{\ell}^{\prime}$ decays are among the LFV interactions and the theoretical predictions of their branching ratios in the framework of the SM are extremely small [30, 31, 32]. These results are far from the experimental limits obtained at LEP1 [18]: $\displaystyle Br(Z\rightarrow e^{\pm}\mu^{\mp})$ $\displaystyle<$ $\displaystyle 1.7\times 10^{-6}\,\,\,\,,$ (7) $\displaystyle Br(Z\rightarrow e^{\pm}\tau^{\mp})$ $\displaystyle<$ $\displaystyle 9.8\times 10^{-6}\,\,\,\,,$ (8) $\displaystyle Br(Z\rightarrow\mu^{\pm}\tau^{\mp})$ $\displaystyle<$ $\displaystyle 1.2\times 10^{-5}\,\,.$ (9) Better sensitivities are expected from the Giga-Z modes at future colliders, such as International Linear Collider (ILC), to have [33, 34, 35]: $\displaystyle Br(Z\rightarrow e^{\pm}\mu^{\mp})$ $\displaystyle<$ $\displaystyle 2\times 10^{-9}\,,$ (10) $\displaystyle Br(Z\rightarrow e^{\pm}\tau^{\mp})$ $\displaystyle<$ $\displaystyle\kappa\times 6.5\times 10^{-8}\,,$ (11) $\displaystyle Br(Z\rightarrow\mu^{\pm}\tau^{\mp})$ $\displaystyle<$ $\displaystyle\kappa\times 2.2\times 10^{-8},$ (12) with $\kappa\simeq 0.2-1.0$. It will be interesting to study the leptoquark contributions to the $Z\to l\bar{l}^{\prime}$ processes. The aim of the present paper is to study the leptoquark effects in various LFV processes including $l\to l^{\prime}\gamma$ and $Z\to l\bar{l}^{\prime}$ decays, while considering leptoquark contribution to $a_{\mu}$ as a solution to the muon anomalous moment discrepancy. The layout of the present paper is as follows: In Sec. II we introduce the formalism. We then use it in Sec. III to study the leptoquark contributions to $a_{\mu}$ and LFV processes including $l\to l^{\prime}\gamma$ and $Z\to l\bar{l}^{\prime}$ decays. Sec. IV contains our conclusions. Some formulas and low energy constraints are given in Appendices. ## II Formalism ### II.1 Scalar Leptoquark Interactions In this section we list the relevant parts of the scalar leptoquark Lagrangian. We consider isosinglet scalar leptoquarks. The effective Lagrangian describing the leptoquark interactions in the mass basis is given by [24, 10]: $\displaystyle{\mathcal{L}}_{LQ}$ $\displaystyle=$ $\displaystyle\overline{u^{c}_{a}}\bigg{(}h^{{}^{\prime}}_{ai}\Gamma_{k,S_{R}}P_{L}+h_{ai}\Gamma_{k,S_{L}}P_{R}\bigg{)}e_{i}S^{}_{k}+\overline{e_{j}}\bigg{(}h^{{}^{\prime}}_{aj}\Gamma^{\dagger}_{S_{R},k}P_{R}+h^{}_{aj}\Gamma^{\dagger}_{S_{L},k}P_{L}\bigg{)}u^{c}_{a}S_{k}$ $\displaystyle-$ $\displaystyle eQ_{(u^{c})}A_{\mu}\overline{u^{c}_{a}}\gamma^{\mu}u^{c}_{a}-ieQ_{S}A_{\mu}S^{}_{k}\\!\stackrel{{\scriptstyle\leftrightarrow}}{{\partial^{\mu}}}\\!S_{k}+ieQ_{S}\tan\theta_{W}Z_{\mu}S^{}_{k}\\!\stackrel{{\scriptstyle\leftrightarrow}}{{\partial^{\mu}}}\\!S_{k}$ $\displaystyle-$ $\displaystyle\frac{e}{s_{W}c_{W}}Z_{\mu}\overline{u^{c}_{a}}\gamma^{\mu}\bigg{(}(T_{3(u^{c})}-Q_{(u^{c})}s^{2}_{W})P_{R}-Q_{(u^{c})}s^{2}_{W}P_{L}\bigg{)}u^{c}_{a},$ where $k=1,2$ are the indices of leptoquark, $T_{3}=-1/2$, $Q_{u^{c}}=-2/3$ are quark’s isospin and electric charge, $Q_{S}=-1/3$ is the electric charge of scalar leptoquarks $S_{k}$, $a$ and $i,j$ are quarks and leptons flavor indices and we use $c_{W}=\cos\theta_{W}$ and $s_{W}=\sin\theta_{W}$. The $\Gamma_{k,S_{L,R}}$ are elements of leptoquark mixing matrix that brings $S_{L,K}$ to the mass basis $S_{k}$: $\displaystyle S_{L}=\Gamma^{\dagger}_{S_{L},k}S_{k},\qquad S^{}_{R}=\Gamma_{k,S_{R}}S^{}_{k},$ (14) where the $S_{L(R)}$ is the field that associates with the $\overline{e_{j}}P_{L(R)}u^{c}_{a}$ terms in ${\mathcal{L}}_{LQ}$ [24]. Note that in the no-mixing case ($\Gamma=1$), $S_{1,2}$ reduce to $S_{L,R}$, which are called chiral leptoquarks, as they only couple to quarks and leptons in certain chirality structures. Finally, the couplings $h$ and $h^{\prime}$ are 3 by 3 matrices, which give rise to various LFV processes and must be subject to experimental constraints. In this work we do not aim at a comprehensive study of the effects of all possible leptoquark interactions. Instead, we try to demonstrate that a simple scalar leptoquark model can provide rich and interesting LFV phenomenons. ### II.2 Muon anomalous magnetic moment $(g-2)_{\mu}$ Figure 1: Feynman diagrams contributing to $\ell\to\ell^{\prime}\gamma$ and $Z\to\ell\bar{\ell}^{\prime}$, ${S}_{k}$ are the scalar leptoquark $k=1,2$, $u^{c}_{a}$ are quark up with $a=1,2,3$. The LQ interaction is capable to generate muon anomalous magnetic moment and resolve the discrepancy between theoretical and experimental results. The one- loop diagrams are shown in Fig. 1(a)-(b) with $l=l^{\prime}=\mu$. The extra contribution to $a_{\mu}$ arises from the LQ model due to quark and scalar leptoquark one-loop contribution is given by $\displaystyle a^{LQ}_{\mu}$ $\displaystyle=$ $\displaystyle-\frac{N_{c}m^{2}_{\mu}}{8\pi^{2}}\sum_{q=1}^{3}\sum_{k=1}^{2}\frac{1}{M^{2}_{S_{k}}}\bigg{[}\big{(}\|h_{q\mu}\Gamma_{k,S_{L}}\|^{2}+\|h^{\prime}_{q\mu}\Gamma_{k,S_{R}}\|^{2}\big{)}\big{(}Q_{(u^{c})}F_{2}(x)-Q_{S}F_{1}(x)\big{)}$ (15) $\displaystyle-\frac{m_{(u^{c}_{a})}}{m_{\mu}}{\rm Re}\big{(}h^{\prime}_{q\mu}h^{}_{q\mu}\Gamma^{+}_{S_{R},k}\Gamma_{k,S_{L}}\big{)}\big{(}Q_{(u^{c})}F_{3}(x_{ka})-Q_{S}F_{4}(x_{ka})\big{)}\bigg{]},$ In the above expression, $N_{c}=3$, $Q_{S}=-1/3$, $Q_{u^{c}}=-2/3$. Our expression agrees with that in [15, 6]. The kinematic loop functions $F_{i}$ $(i=1,...,4)$ depend on the variable $x=m^{2}_{(u^{c}_{a})}/m^{2}_{S_{k}}$ are given in the appendix A. Using leptoquark contribution to saturate the deviation given in Eq.(3), the leptoquark masses $M_{S_{1,2}}$, mixing angle $\theta_{LQ}$ and couplings $h^{(\prime)}_{q\mu}$, will be constrained. ### II.3 $\ell\to\ell^{\prime}\gamma$ In this subsection we give the amplitude of $\ell\to\ell^{\prime}\gamma$ from leptoquark exchange. According to the gauge invariance, the amplitude can be written as: $\displaystyle i{\mathcal{M}}^{\gamma}$ $\displaystyle=$ $\displaystyle ie\bar{u}(p_{2})\bigg{(}F^{\gamma}_{2RL}P_{L}+F^{\gamma}_{2LR}P_{R}\bigg{)}(i\sigma_{\mu\nu}q^{\nu})u(p_{1})\varepsilon^{\mu}_{\gamma},$ (16) where $\varepsilon_{\gamma}$ is the polarization vector and $q=p_{1}-p_{2}$ is the momentum transfer. For the amplitude of leptoquark exchange at one-loop level as depicted in Figure. 1, we have $\displaystyle F^{\gamma}_{2LR}$ $\displaystyle=$ $\displaystyle\frac{N_{c}}{16\pi^{2}}\sum_{q=1}^{3}\sum_{k=1}^{2}\frac{1}{M^{2}_{S_{k}}}\Bigg{[}\big{(}m_{l}h^{\prime}_{q\ell}h^{{}^{\prime}}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{R},k}\Gamma_{k,S_{R}}+m_{l^{\prime}}h_{q\ell}h^{}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{L},k}\Gamma_{k,S_{L}}\big{)}\big{(}Q_{(u^{c})}F_{2}(x)-Q_{S}F_{1}(x)\big{)}$ (17) $\displaystyle- m_{(u^{c}_{a})}\big{(}h_{q\ell}h^{{}^{\prime}}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{R},k}\Gamma_{k,S_{L}}\big{)}\big{(}Q_{(u^{c})}F_{3}(x)-Q_{S}F_{4}(x)\big{)}\Bigg{]},$ $\displaystyle F^{\gamma}_{2RL}$ $\displaystyle=$ $\displaystyle F^{\gamma}_{2LR}(h\leftrightarrow h^{\prime},R\leftrightarrow L),$ (18) with $x=m^{2}_{(u^{c}_{a})}/m^{2}_{S_{k}}$. The branching ratio of $\ell\to\ell^{\prime}\gamma$ is: $\displaystyle{\rm Br}(\ell\to\ell^{\prime}\gamma)$ $\displaystyle=$ $\displaystyle\frac{\alpha_{em}}{4\Gamma(\ell)}\frac{(m^{2}_{\ell}-m^{2}_{\ell^{\prime}})^{3}}{m^{3}_{\ell}}\bigg{(}\|F^{\gamma}_{2LR}\|^{2}+\|F^{\gamma}_{2RL}\|^{2}\bigg{)},$ (19) In our numerical calculations we analyze the Brs of the decays under consideration by using the total decay widths of the decaying leptons $\Gamma(\ell)$. ### II.4 $Z\to\ell\bar{\ell}^{\prime}$ The Feynman diagrams of LFV $Z$ decay process are shown in Fig. 1. The total contribution of all diagrams (c) and (d) can be written as $\displaystyle i{\mathcal{M}}^{Z}_{\mu}$ $\displaystyle=$ $\displaystyle iem^{2}_{Z}\bar{u}(p_{2})\bigg{[}\bigg{(}F^{Z}_{1L}P_{R}+F^{Z}_{1R}P_{L}\bigg{)}\bigg{(}-g_{\mu\nu}+\frac{q_{\mu}q_{\nu}}{m^{2}_{Z}}\bigg{)}\gamma^{\nu}$ $\displaystyle+\frac{1}{m^{2}_{Z}}\bigg{(}F^{Z}_{2RL}P_{L}+F^{Z}_{2LR}P_{R}\bigg{)}(i\sigma_{\mu\nu}q^{\nu})\bigg{]}u(p_{1})\varepsilon^{Z}_{\mu}(q)$ where $q_{\mu}$ is the $Z$ four-momentum. The decay rates involve both $F^{Z}_{1L(R)}$ and $F^{Z}_{2LR(RL)}$: $\displaystyle{\rm Br}(Z\to\ell\bar{\ell}^{\prime})$ $\displaystyle=$ $\displaystyle\frac{\alpha_{em}}{6}\frac{m_{Z}}{\Gamma_{Z}}\bigg{[}\bigg{(}\|F^{Z}_{1L}\|^{2}+\|F^{Z}_{1R}\|^{2}\big{)}$ $\displaystyle+\frac{1}{2m^{2}_{Z}}\bigg{(}\|F_{2LR}(Z)\|^{2}+\|F_{2RL}(Z)\|^{2}\bigg{)}\bigg{]},$ where the form factors $F^{Z}_{1L(R)}$ and $F^{Z}_{2LR(RL)}$ are given by $\displaystyle F^{Z}_{1L}$ $\displaystyle=$ $\displaystyle\frac{N_{c}}{16\pi^{2}}\frac{1}{M^{2}_{S_{k}}}\bigg{[}h^{\prime}_{q\ell}h^{\prime}_{q\ell^{\prime}}\Gamma^{+}_{S_{R},k}\Gamma_{k,S_{R}}\big{(}g_{S}G_{1}(x)+g_{R}G_{2}(x)\big{)}$ $\displaystyle-$ $\displaystyle\frac{m_{u_{a}}}{m^{2}_{\ell}-m^{2}_{\ell^{\prime}}}\big{(}g_{L}-g_{R}\big{)}\big{(}h_{q\ell}h^{\prime}_{q\ell^{\prime}}\Gamma^{+}_{S_{R},k}\Gamma_{k,S_{L}}m_{\ell}-h^{\prime}_{q\ell}h^{}_{q\ell^{\prime}}\Gamma^{+}_{S_{L},k}\Gamma_{k,S_{R}}m_{\ell^{\prime}}\big{)}G_{3}(x)\bigg{]},$ $\displaystyle F^{Z}_{1R}$ $\displaystyle=$ $\displaystyle F^{Z}_{1L}(h\leftrightarrow h^{\prime},L\leftrightarrow R),$ (23) and $\displaystyle F^{Z}_{2LR}$ $\displaystyle=$ $\displaystyle\frac{N_{c}}{16\pi^{2}}\frac{1}{M^{2}_{S_{k}}}\bigg{[}h_{q\ell}h^{\prime}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{R},k}\Gamma_{k,S_{L}}m_{u_{a}}(g_{R}+g_{L})G_{3}(x)$ $\displaystyle+$ $\displaystyle\big{(}g_{R}h^{\prime}_{q\ell}h^{\prime}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{R},k}\Gamma_{k,S_{R}}m_{\ell}+g_{L}h_{q\ell}h^{}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{L},k}\Gamma_{k,S_{L}}m_{\ell^{\prime}}\big{)}G_{4}(x)$ $\displaystyle-$ $\displaystyle g_{S}\bigg{(}(h^{\prime}_{q\ell}h^{\prime}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{R},k}\Gamma_{k,S_{R}}m_{\ell}+h_{q\ell}h^{}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{L},k}\Gamma_{k,S_{L}}m_{\ell^{\prime}})G_{5}(x)+m_{u_{a}}h_{q\ell}h^{\prime}_{q\ell^{\prime}}\Gamma^{\dagger}_{S_{R},k}\Gamma_{k,S_{L}}G_{6}(x)\bigg{)}\bigg{]},$ $\displaystyle F^{Z}_{2RL}$ $\displaystyle=$ $\displaystyle F^{Z}_{2LR}(h\leftrightarrow h^{\prime},L\leftrightarrow R),$ (25) where we have $x=m^{2}_{u_{a}}/m^{2}_{S_{k}}$ and the couplings $g_{R,L}$ and $g_{S}$ are given by $\displaystyle g_{R}$ $\displaystyle=$ $\displaystyle-\frac{2}{\sin\theta_{W}\cos\theta_{W}}\big{(}T_{3(u^{c})}-Q_{(u^{c})}\sin^{2}\theta_{W}\big{)},$ (26) $\displaystyle g_{L}$ $\displaystyle=$ $\displaystyle Q_{(u^{c})}\tan\theta_{W},\qquad g_{S}=Q_{S}\tan\theta_{W}.$ (27) In the above expressions of $F^{Z}_{1L(R)}$, we keep only the leading term in $m_{Z}^{2}/m_{S_{k}}^{2}$. The explicit expressions of one loop functions $G_{n}$ $(n=1,..6)$ can be found in the appendix A. ## III Numerical results and discussion We are now ready to give some numerical results. The quark mass are evaluated at the scale of the $\mu=300$ GeV [26], which is the typical leptoquark mass used in this work, $\displaystyle m_{t}=161.4\,{\rm GeV},\quad m_{c}=0.55\,{\rm GeV},\quad m_{u}=11.4\times 10^{-3}\,{\rm GeV},$ (28) and for the following quantities we use [18] $\displaystyle\alpha_{em}=1/137.0359,\quad M_{W}=80.45\,{\rm GeV},\quad M_{Z}=91.1875\,{\rm GeV}.$ (29) For simplicity, we assume that the couplings $h$ and $h^{\prime}$ are real and equal to each other, i.e. $\displaystyle h=h^{\prime}=h^{}.$ (30) We use leptoquark mass splitting $\Delta=500$ GeV in our analysis, where $\Delta$ is defined as $\sqrt{M^{2}_{S_{2}}-M^{2}_{S_{1}}}$. Consequently, the remaining parameters in the leptoquark model are the mass of the light scalar leptoquark $M_{S_{1}}$, the mixing angle $\theta_{LQ}$, and the couplings $h_{q\ell}$. ### III.1 Muon Anomalous Magnetic Moment $a_{\mu}$ \| ---\|--- Figure 2: Scatter plot in the plane ($M_{S_{1}}-\|h_{q\mu}\|^{2}$) in the left panel,($M_{S_{1}}-\sin 2\theta_{LQ}$) in the right panel. These are allowed regions in the parameter space that give $a^{LQ}_{\mu}=\Delta a_{\mu}=(295\pm 87.7)\times 10^{-11}$. In this section we discuss a few phenomenological aspects of the leptoquark contributions to $a_{\mu}$. In the left panel of Fig. 2, we present an scatter plot in the $(M_{S_{1}}-\|h_{q\mu}\|^{2})$ plane for top quark contribution (red) and charm quark contribution (green), which are allowed by $a^{LQ}_{\mu}=\Delta a_{\mu}=(295\pm 87.7)\times 10^{-11}$ [see Eq. (3)] within the 1$\sigma$ range of data. We note that it is not possible to use the up quark loop contribution alone for the $a_{\mu}^{LQ}=\Delta a_{\mu}$, since the mixing angle and couplings $h_{u\mu}$ are strongly constrained by the $\pi$ leptonic decays (see Appendix B). In order to see the impact of the mixing angle, we present in the right panel of Fig.2 the allowed regions $a^{LQ}_{\mu}=\Delta a_{\mu}$ in the $(M_{S_{1}}-\sin 2\theta_{LQ})$ plane. We use $\alpha_{em}\leq h^{2}_{q\mu}\leq 1$. The contribution dominates around $\sin 2\theta_{LQ}\sim 0.7$ both for top and charm quark contributions. We see that the constraint from $a_{\mu}$ confines the allowed range of $M_{S_{1}}$ to $M_{S_{1}}\lesssim 950$ GeV for top quark contribution and to $M_{S_{1}}\lesssim 350$ GeV for charm quark contribution at the 1$\sigma$ level. These parameter space will be used for later study of LFV processes. The light leptoquark mass should be below 1 TeV, if leptoquarks with couplings of electromagnetic strength are responsible to the deviation $\Delta a_{\mu}$. It is interesting that LHC may have good chance to observe these particles [27]. ### III.2 Lepton Flavor Violating $l\to l^{\prime}\gamma$ and $Z\to l\bar{l}^{\prime}$ Decays In this section, we investigate the LFV decay processes generated by the same leptoquark scalar interactions. We consider only parameter space that corresponds to $a^{LQ}_{\mu}=\Delta a_{\mu}$ when it is appropriate. We discuss $\mu\to e\gamma$ and $\tau\to e\gamma,\,\mu\gamma$ decays first. In Fig. 3 and Fig. 4 we show scatter plots of the allowed parameters in $(M_{S_{1}},h_{q\ell}h_{q\ell^{\prime}})$ planes from bounds of $\tau\to\mu\gamma$, $\tau\to e\gamma$ and $\mu\to e\gamma$ rates. Note that in the plots we use $\displaystyle 1.5\times 10^{-13}$ $\displaystyle\leq$ $\displaystyle{\rm Br}(\mu\to e\gamma)<1.2\times 10^{-11},$ $\displaystyle 1\times 10^{-9}$ $\displaystyle\leq$ $\displaystyle{\rm Br}(\tau\to e\gamma)<1.1\times 10^{-7},$ $\displaystyle 1\times 10^{-9}$ $\displaystyle\leq$ $\displaystyle{\rm Br}(\tau\to\mu\gamma)<6.8\times 10^{-8}.$ (31) where the upper bounds are from the current limits: Eqs. (4)-(6), while the lower bound for ${\rm Br}(\mu\to e\gamma)$ is from [29] and the lower bounds for $\tau\to l\gamma$ are for illustration. For the $\tau\to\mu\gamma$ and $\mu\to e\gamma$ cases the $(g-2)_{\mu}$ constraint is taken into account. For different quark contribution the couplings are bounded in the following ranges: $10^{-4}\lesssim h_{q\tau}h_{q\mu}\lesssim 10^{-2}$, $10^{-3}\lesssim h_{c\tau}h_{ce}\lesssim 1$, $10^{-4}\lesssim h_{t\tau}h_{te}\lesssim 1$ and $10^{-7}\lesssim h_{q\mu}h_{qe}\lesssim 10^{-6}$. For the $\tau\to\mu\gamma$ and $\mu\to e\gamma$ cases, the allowed leptoquark masses are $m_{S_{1}}\lesssim 250-300$ GeV and 1 TeV for $c$-quark and $t-$ quark loop contribution, respectively. These region are determined from the bounds and the muon $g-2$ constraint (see also Fig. 2) at the same time. On the other hand the couplings governing $\tau\to e\gamma$ decay and those generating muon $g-2$ contribution are decoupled, the parameters corresponding to the former bounds are free from the latter constraint. The resulting allowed regions are larger in this cases. The parameters in these allowed regions will be used to predict $Z\to l\bar{l}^{\prime}$ decays. To have an idea of the size the allowed couplings, we give that upper bound on $h_{q\ell}h_{q\ell^{\prime}}$ obtained form the present $l\to l^{\prime}\gamma$ limits in Table 1. We see that the $\mu\to e\gamma$ constraint is more effective in restricting the sizes of $h_{q\ell}h_{q\ell^{\prime}}$. Decay mode \| $h_{c\ell}h_{c\ell^{\prime}}$ \| $h_{t\ell}h_{t\ell^{\prime}}$ ---\|---\|--- $\tau\to\mu\gamma$ \| $<5.29\times 10^{-3}$ \| $<9.11\times 10^{-3}$ $\tau\to e\gamma$ \| $<0.81$ \| $<0.82$ $\mu\to e\gamma$ \| $<1.45\times 10^{-6}$ \| $<1.92\times 10^{-6}$ Table 1: Constraints on the parameters $h_{q\ell}h_{q\ell^{\prime}}$ $(q=t,c)$ coming from radiative FCNC processes induced by the scalar leptoquark using the present experimental bounds. \| ---\|--- Figure 3: Scatter plots of leptoquark parameters in $(M_{S_{1}},h_{q\ell}h_{q\ell^{\prime}})$ planes from $(\ell\to\ell^{\prime}\gamma)$ bounds given in Eq. (31). The left (right) figure is for the $\tau\to\mu\gamma$ ($\tau\to e\gamma$) case with top and charm quark contributions. Figure 4: Same as Fig. 3 except for the $\mu\to e\gamma$ case. In Fig. 5 and 6, we give the predicted $Z\to l\bar{l}^{\prime}$ rates in correlation with Br$(l\to l^{\prime}\gamma)$. We see that $Br(Z\to\tau^{\mp}e^{\pm})$ can reach $1.95\times 10^{-5}$, which is comparable with the present bound, and $Br(Z\to\mu^{\mp}\tau^{\pm})$ can reach $2.34\times 10^{-8}$, which will be accessible by future linear colliders. On the contrary, the current bound on the $\mu\to e\gamma$ decay imposes very strong constraints on the related couplings as shown in Table I. Hence the predicted $Br(Z\to\mu^{\mp}e^{\pm})$ is rather small and is too low to be observed in the near future. In Fig. 5, we see that the $Z\to l\bar{l}^{\prime}$ rates are roughly positively correlating with the $l\to l^{\prime}\gamma$ rates and the top quark loop contributions are larger than the charm quark’s ones. To have observable $Z\to\tau^{\mp}\mu^{\pm}$ and $Z\to\tau^{\mp}e^{\pm}$, the $\tau\to\mu\gamma,\,e\gamma$ rates are predicted to be close to the present bounds. \| ---\|--- Figure 5: The correlation between Br$(\tau\to\ell^{\prime}\gamma)$ and Br$(Z\to\tau\ell^{\prime})$ where $\tau=e,\mu$. Figure 6: The correlation between Br$(\mu\to e\gamma)$ and Br$(Z\to\mu e)$. In this work the analysis has been performed for the scalar leptoquark case. It is possible that vector leptoquarks may also contribute to $(g-2)_{\mu}$ and LFV processes. As shown in Ref. [4, 20], quite often $(g-2)_{\mu}$ and LFV processes provide more stringent constraints on vector leptoquark couplings and masses than on scalar leptoquark ones. For example, using the measured $m_{t}$ and the formula given in [4], the present $\Delta a_{\mu}$ leads to a very large mass scale $\Lambda\simeq 500$ TeV in the vector leptoquark case, where $\Lambda$ was defined from the relation: $4\pi/\Lambda^{2}\equiv g^{2}_{LQ}/m^{2}_{LQ}$. The mass scale is much larger than the corresponding mass scale exhibited in Fig. 2, which is found to be $\Lambda\simeq$ few – ${\mathcal{O}}(10)$ TeV. Similarly, in $l\to l^{\prime}\gamma$ processes, the constraints on vector leptoquark parameters are usually more severe [20]. ## IV Conclusion Motivated by the reported discrepancy of the muon $g-2$ results, we studied the lepton flavor violating $\ell\to\ell^{\prime}\gamma$ and $Z\to\ell\bar{\ell}^{\prime}$ decays in the LQ model. We showed that the $g-2$ anomaly favors LQ masses in rather low-energy regime, e.g. $<1$ TeV, which is within the reach of the forthcoming Large Hadron Collider. We found that leptoquarks can generate sizable LFV $l\to l^{\prime}\gamma$ decays. The present experimental limits are used to confined the leptoquark parameter space. On the other hand, it is interesting to search for these LFV effects in experiments, such as MEG, B factories and the super B factory. We predict $Br(Z\to\tau^{\mp}e^{\pm})$ reaching $10^{-5}$ and $Br(Z\to\mu^{\mp}\tau^{\pm})$ reaching $2\times 10^{-8}$, which can be accessible by present experiments and future linear colliders, such as ILC. On the contrary, the current bounds on LFV impose very strong constraints on the $Br(Z\to\mu^{\mp}e^{\pm})$ and the ratio is too low to be observed in the near future. In this case, it is much useful to search for the LFV effects in $\mu\to e\gamma$ decay. ## V Acknowledgment The authors would like to thank Dr. Stefan Ritt for useful discussions. This work is supported in part by the National Science Council of R.O.C under grant number: NSC96-2811-M-033-005 and NSC-95-2112-M-033-MY2. ## Appendix A One loop functions The loop functions $F_{i}$ and $G_{i}$ used in Sec. II are given by $\displaystyle F_{1}(x)$ $\displaystyle=$ $\displaystyle\frac{\big{[}2+3x-6x^{2}+x^{3}+6x\log(x)\big{]}}{12(1-x)^{4}},$ (32) $\displaystyle F_{2}(x)$ $\displaystyle=$ $\displaystyle\frac{\big{[}1-6x+3x^{2}+2x^{3}-6x^{2}\log(x)\big{]}}{12(1-x)^{4}},$ (33) $\displaystyle F_{3}(x)$ $\displaystyle=$ $\displaystyle\frac{-1}{2(1-x)^{3}}\big{[}3-4x+x^{2}+2\log(x)\big{]},$ (34) $\displaystyle F_{4}(x)$ $\displaystyle=$ $\displaystyle\frac{1}{2(1-x)^{3}}\big{[}1-x^{2}+2x\log(x)\big{]},$ (35) and $\displaystyle G_{1}(x)$ $\displaystyle=$ $\displaystyle\frac{\big{[}-2+9x^{2}-18x^{4}+11x^{6}-12x^{6}\log(x)\big{]}}{36(x^{2}-1)^{4}},$ (36) $\displaystyle G_{2}(x)$ $\displaystyle=$ $\displaystyle\frac{1}{36(x^{2}-1)^{4}}$ $\displaystyle\times$ $\displaystyle\big{[}16-45x^{2}+36x^{4}-7x^{6}+12(-2+3x^{2})\log(x)\big{]}$ $\displaystyle G_{3}(x)$ $\displaystyle=$ $\displaystyle\frac{3-4x^{2}+x^{4}+4\log(x)}{4(x^{2}-1)^{3}},$ (38) $\displaystyle G_{4}(x)$ $\displaystyle=$ $\displaystyle\frac{2+9x^{2}-6x^{4}+x^{6}+12x^{2}\log(x)}{12(x^{2}-1)^{4}},$ (39) $\displaystyle G_{5}(x)$ $\displaystyle=$ $\displaystyle\frac{1-6x^{2}+3x^{4}+2x^{6}-12x^{4}\log(x)}{12(x^{2}-1)^{4}},$ (40) $\displaystyle G_{6}(x)$ $\displaystyle=$ $\displaystyle\frac{1}{2(x^{2}-1)^{3}}\big{[}-1+x^{4}-4x^{2}\log(x)\big{]}.$ (41) ## Appendix B Constraint form $\pi\to e\nu_{e}$ and $\pi\to\mu\nu_{\mu}$ decays We follow [11, 20] to constrain leptoquark parameters using pion decay data. Form the interactions given in Eq. (II.1), we obtain the effective four-Fermi interaction $\displaystyle{\mathcal{L}}_{eff}$ $\displaystyle=$ $\displaystyle-\frac{h^{\prime}_{ai}h^{{}^{\prime}}_{bj}\Gamma^{+}_{R,k}\Gamma_{k,R}}{M^{2}_{S_{k}}}(\bar{e}^{c}_{i}P_{L}u_{a})(\bar{d}_{b}P_{R}\nu^{c}_{j})$ $\displaystyle-$ $\displaystyle\frac{h_{ai}h^{{}^{\prime}}_{bj}\Gamma^{\dagger}_{R,k}\Gamma_{k,L}}{M^{2}_{S_{k}}}(\bar{e}^{c}_{i}P_{R}u_{a})(\bar{d}_{b}P_{R}\nu^{c}_{j})$ By using the Fierz transformation, we can rewite the Eq.(B) as $\displaystyle{\mathcal{L}}_{eff}$ $\displaystyle=$ $\displaystyle-\frac{1}{2M^{2}_{S_{k}}}h^{\prime}_{ai}h^{{}^{\prime}}_{bj}\Gamma^{\dagger}_{R,k}\Gamma_{k,R}(\bar{d}_{L,b}\gamma_{\mu}u_{L,a})(\bar{\nu}_{L,j}\gamma^{\mu}e_{L,i})$ (43) $\displaystyle+$ $\displaystyle\frac{1}{2M^{2}_{S_{k}}}h_{ai}h^{{}^{\prime}}_{bj}\Gamma^{\dagger}_{R,k}\Gamma_{k,L}(\bar{d}_{L,b}u_{R,a})(\bar{\nu}_{L,j}e_{R,i})$ On the other hand, the conventional interation for the $\pi\to l\nu_{l}$ decay in the SM is given by $\displaystyle{\mathcal{L}}_{eff}$ $\displaystyle=$ $\displaystyle-\frac{G_{F}V_{ud}}{\sqrt{2}}[\bar{\nu}\gamma_{\mu}(1-\gamma_{5})l][\bar{d}\gamma^{\mu}(1-\gamma_{5})u]+{\rm h.c}$ here $\|V_{ud}\|$ is the Cabibbo-Kobayashi-Maskawa (CKM) matrix elements between the constituent of the pion meson, $G_{F}$ is the Fermi couplings constant. The ratio $R_{th}$ of the electronic and muonic decay modes is [36] $\displaystyle R_{th}$ $\displaystyle=$ $\displaystyle\frac{\Gamma_{SM}(\pi^{+}\to\bar{e}\nu_{e})}{\Gamma_{SM}(\pi^{+}\to\bar{\mu}\nu_{\mu})}$ (44) $\displaystyle=$ $\displaystyle\bigg{(}\frac{m^{2}_{e}}{m^{2}_{\mu}}\bigg{)}\bigg{(}\frac{m^{2}_{\pi}-m^{2}_{e}}{m^{2}_{\pi}-m^{2}_{\mu}}\bigg{)}^{2}\bigg{(}1+\delta\bigg{)}$ $\displaystyle=$ $\displaystyle(1.2352\pm 0.0001)\times 10^{-4}$ where $\delta$ is the radiative corrections, Thus the ratio $R_{th}$ is very sensitive to non standard model effects (such as multi-Higges, non-chiral leptoquarks). The experimental ratio is [18] $\displaystyle R_{exp}=(1.2302\pm 0.004)\times 10^{-4}$ (45) The interference between the standard model and LQ model can be expressed by $\displaystyle R_{SM-LQ}$ $\displaystyle=$ $\displaystyle R_{th}+R_{th}\,\frac{m^{2}_{\pi^{+}}}{m_{u}+m_{d}}\bigg{(}\frac{1}{\sqrt{2}}\frac{{\rm Re}(h_{ue}h^{{}^{\prime}}_{ue})}{G_{F}V_{ud}M^{2}_{S_{k}}}\frac{1}{m_{e}}-\frac{1}{\sqrt{2}}\frac{{\rm Re}(h_{u\mu}h^{{}^{\prime}}_{u\mu})}{G_{F}V_{ud}M^{2}_{S_{k}}}\frac{1}{m_{\mu}}\bigg{)}\Gamma^{\dagger}_{R,k}\Gamma_{k,L}$ (46) At 2$\sigma$ level, we get $\displaystyle R_{min}<\sum^{2}_{k=1}\bigg{(}\frac{m_{\pi}}{m_{e}}\frac{{\rm Re}(h_{ue}h^{{}^{\prime}}_{ue})}{M^{2}_{S_{k}}}-\frac{m_{\pi}}{m_{\mu}}\frac{{\rm Re}(h_{u\mu}h^{{}^{\prime}}_{u\mu})}{M^{2}_{S_{k}}}\bigg{)}\Gamma^{\dagger}_{R,k}\Gamma_{k,L}<R_{max}$ (47) where, $\displaystyle R_{min}$ $\displaystyle=$ $\displaystyle-1.06\times 10^{-8}{\rm GeV}^{-2},$ (48) $\displaystyle R_{max}$ $\displaystyle=$ $\displaystyle 2.45\times 10^{-9}{\rm GeV}^{-2}.$ (49) The total contribution to $R_{SM-LQ}$ must be smaller than the differences between SM and experiment within the error limits allow. ## References * [1] G. W. Bennett et al. [Muon g-2 Collaboration], Phys. Rev. Lett. 89, 101804 (2002) [Erratum-ibid. 89, 129903 (2002)] [arXiv:hep-ex/0208001]; Phys. Rev. Lett. 92, 161802 (2004) [arXiv:hep-ex/0401008]. * [2] J. P. Miller, E. de Rafael and B. L. Roberts, Rept. Prog. Phys. 70, 795 (2007) [arXiv:hep-ph/0703049]. * [3] T. Aoyama, M. Hayakawa, T. Kinoshita and M. Nio, Phys. Rev. D 77, 053012 (2008); T. Kinoshita and M. Nio, Phys. Rev. 73, 013003 (2006); T. Kinoshita and M. Nio, Phys. Rev. 70, 113003 (2004). * [4] I. I. Y. Bigi, G. Kopp and P. M. Zerwas, Phys. Lett. B 166, 238 (1986). * [5] U. Mahanta, Eur. Phys. J. C 21, 171 (2001) [Phys. Lett. B 515, 111 (2001)] [arXiv:hep-ph/0102176]. * [6] K. m. Cheung, Phys. Rev. D 64, 033001 (2001) [arXiv:hep-ph/0102238]. * [7] J. C. Pati and A. Salam, Phys. Rev. D 10, 275 (1974) [Erratum-ibid. D 11, 703 (1975)]. * [8] H. Georgi and S. L. Glashow, Phys. Rev. Lett. 32, 438 (1974); H. Georgi, AIP Conf. Proc. 23, 575 (1975); H. Fritzsch and P. Minkowski, Annals Phys. 93, 193 (1975). * [9] E. Farhi and L. Susskind, Phys. Rept. 74, 277 (1981); K. D. Lane and M. V. Ramana, Phys. Rev. D 44, 2678 (1991); B. Schrempp and F. Schrempp, Phys. Lett. B 153, 101 (1985). * [10] W. Buchmuller, R. Ruckl and D. Wyler, Phys. Lett. B 191, 442 (1987) [Erratum-ibid. B 448, 320 (1999)]. * [11] O. U. Shanker, Nucl. Phys. B 204, 375 (1982). * [12] O. J. P. Eboli and A. V. Olinto, Phys. Rev. D 38, 3461 (1988); J. L. Hewett and S. Pakvasa, ibid. D 37, 3165 (1988); J. Ohnemus, S. Rudaz, T. F. Walsh and P. M. Zerwas, Phys. Lett. B 334, 203 (1994) [arXiv:hep-ph/9406235]. * [13] J. Wudka, Phys.Lett., B167, 337 (1986); M.A. Doncheski and J.L. Hewett, Z.Phys. C56, 209 (1992). * [14] J.L Hewett and T.G. Rizzo, Phys.Rev. D36, 3367 (1987); J.L Hewett and S. Pakvasa, Phys.Lett. B227, 178 (1987); J.E. Cieza and O.J.P. Éboli, Phys.Rev. D47, 837 (1993). * [15] A. Djouadi, T. Kohler, M. Spira and J. Tutas, Z. Phys. C 46, 679 (1990). * [16] J. Blümlein and R. Rückl, Phys.Lett. B304, 337 (1993). * [17] O.J.P. Éboli et al., Phys.Lett. B311, 147 (1993); H. Nadeau and D. London, Phys.Rev. D47, 3742 (1993). * [18] W. M. Yao et al Particle Data Group, J. Phys. G30, 1 (2006) and 2007 partial update for the 2008 edition. * [19] G. Bélanger, D. London and H. Nadeau, Phys.Rev. D49, 3140 (1993). * [20] S. Davidson, D. Bailey and A. Campbell, Z.Phys. C61, 613 (1994). * [21] M. Leurer, Phys.Rev.Lett.71 (1993) 1324; Phys.Rev. D49, 333 (1994); Phys.Rev. D50, 536 (1994). * [22] W. Buchmüller and D. Wyler, Phys.Lett. B177, 377 (1986). * [23] C. K. Chua and W. Y. Hwang, Phys. Rev. D 60, 073002 (1999) [arXiv:hep-ph/9811232]. * [24] C. K. Chua, X. G. He and W. Y. Hwang, Phys. Lett. B 479, 224 (2000) [arXiv:hep-ph/9905340]. * [25] R. Benbrik and C. H. Chen, arXiv:0807.2373 [hep-ph]. * [26] N. Gray, D. J. Broadhurst, W. Grafe and K. Schilcher, Z. Phys. C48(1990) 673; K. G. Chetyrkin and M. Steinhauser, Phys. Rev. Lett 83 (1999) 4001; ibid. Nucl. Phys. B573 (2000) 617; S. G. Gorishny, A. L. Kataev, S .A .Larin and L. R. Surguladze, Mod. Phys. Lett. A5(1990)2703,ibid. Phys. Rev. D43 (1991) 1633. * [27] V. A. Mitsou, N. C. Benekos, I. Panagoulias and T. D. Papadopoulou, Czech. J. Phys. 55, B659 (2005) [arXiv:hep-ph/0411189]; S. Abdullin and F. Charles, Phys. Lett. B 464, 223 (1999) [arXiv:hep-ph/9905396]. * [28] S. Rolli and M. Tanabashi, in [18]. * [29] S. Ritt [MEG Collaboration], Nucl. Phys. Proc. Suppl. 162 (2006) 279; http://meg.web.psi.ch/. * [30] T. Riemann and G. Mann, in Proc. of the Int. Conf. Neutrino’82, 14-19 June 1982, Balatonfüred, Hungary (A. Frenkel and E. Jenik, eds.), vol. II, pp. 58-, Budapest, 1982, scanned copy at http://www.ifh.de/$\sim$riemann, G. Mann and T. Riemann, Annalen Phys. 40, 334 (1984). * [31] V. Ganapathi, T. Weiler, E. Laermann, I. Schmitt, and P. Zerwas, Phys. Rev. D27 (1983) 579; M. Clements, C. Footman, A. Kronfeld, S. Narasimhan, and D. Photiadis, Phys. Rev. D27 (1983) 570; M. A .Perez, G. T. Velasco, J. J. Toscano, Int.J.Mod.Phys. A19 (2004) 159; A. Flores-Tlalpa, J.M. Hernandez, G. Tavares-Velasco, J.J. Toscano, Phys. Rev. D65, 073010 (2002). * [32] J. I. Illana, M. Jack and T. Riemann, hep-ph/0001273, (2000); J. I. Illana, and T. Riemann, Phys. Rev. D63 , 053004 (2001). * [33] J. A. Aguilar-Saavedra et al. [ECFA/DESY LC Physics Working Group], arXiv:hep-ph/0106315. * [34] S. Heinemeyer, W. Hollik, A. M. Weber and G. Weiglein, arXiv:0711.0456 [hep-ph]. * [35] J. Erler and P. Langacker, arXiv:0807.3023 [hep-ph]. * [36] V. Cirigliano and I. Rosell, Phys. Rev. Lett. 99, 231801 (2007) [arXiv:0707.3439 [hep-ph]].	arxiv-papers	2008-07-26T14:23:40	2024-09-04T02:48:57.005002	{ "license": "Public Domain", "authors": "Rachid Benbrik (CYCU & UCAM), Chun-Khiang Chua (CYCU)", "submitter": "Rachid Benbrik", "url": "https://arxiv.org/abs/0807.4240" }
0807.4554	# Gravitational waves from fragmentation of a primordial scalar condensate into Q-balls Alexander Kusenko Department of Physics and Astronomy, University of California, Los Angeles, CA 90095-1547, USA Anupam Mazumdar Physics Department, Lancaster University, Lancaster, LA1 4YB, UK Niels Bohr Institute, Blegdamsvej-17, Copenhagen, DK-2100, Denmark ###### Abstract A generic consequence of supersymmetry is formation of a scalar condensate along the flat directions of the potential at the end of cosmological inflation. This condensate is usually unstable, and it can fragment into non- topological solitons, Q-balls. The gravitational waves produced by the fragmentation can be detected by Laser Interferometer Space Antenna (LISA), Advanced Laser Interferometer Gravitational-Wave Observatory (LIGO), and Big Bang Observer (BBO), which can offer an important window on the early universe and the physics at some very high energy scales. ###### pacs: 04.30.-w,04.30.Tv,11.30.Pb,12.60.Jv,98.80.-k ††preprint: UCLA/08/TEP/25 Supersymmetry is widely regarded as a likely candidate for physics beyond the Standard Model. While many variants of supersymmetry have been considered, all of them have scalar potentials with some flat directions lifted only by the supersymmetry-breaking terms. At the end of cosmological inflation, the formation of a scalar condensate along the flat directions can have a number of important consequences reviews . In particular, it can be responsible for generation of the matter-antimatter asymmetry via Affleck–Dine (AD) mechanism AD , and, in some models, dark matter can be produced in the same process KS ; dark_matter . Some flat directions could be responsible for the primordial inflation Randall:1995dj ; AEGM . The formation of AD condensate is a generic phenomenon, relying only on the assumptions of inflation and supersymmetry. In general, this condensate is unstable: an initially homogeneous condensate can break up into lumps of the scalar field, called Q-balls Coleman , under some very generic conditions KS . All phenomenologically acceptable supersymmetric generalizations of the Standard Model admit Q-balls Kusenko:1997zq , which can be stable, or can decay into fermions Kusenko:1997zq ; DKS . The formation of Q-balls is accompanied by a coherent motion of the scalar condensate, which creates the source of gravity waves. We will show that fragmentation of the scalar condensate into Q-balls can produce gravitational waves detectable by LISA LISA , LIGO III LIGOIII , and BBO BBO . The physics of AD condensate fragmentation has been studied both analytically KS ; EMc ; EJM ; Pawl ; Johnson:2008se and numerically Minos ; KK ; MV ; Volkov ; Palti ; Broadhead ; Campanelli . At the end of inflation (assuming that the inflation occurs in a hidden sector at high scales) the condensate has a uniform density, with small perturbations of the order of $10^{-5}$ reviews . Under some rather generic conditions, the instabilities develop and lead, eventually, to the formation of Q-balls, which can either decay or remain as stable relics KS ; EMc . Although the final product of such evolution, Q-balls in the ground state, are spherically symmetric, the coherent motions associated with the condensate fragmentation and re- arrangement are not spherically symmetric. Moreover, the newly formed Q-balls first appear in their excited states and oscillate until they settle in the spherically symmetric ground states KK ; MV . The lack of spherical symmetry in the process of fragmentation is essential for generating the gravity waves. Following the general picture developed in Refs. KS ; EJM , the scalar condensate undergoing fragmentation can be approximated, in the linear regime, as $\phi(x,t)=\phi(t)\equiv R(t)e^{i\Omega(t)}$, plus a perturbation $\delta R,\delta\Omega\propto e^{S(t)-i\vec{k}\vec{x}}$. One finds that the homogeneous solution is unstable due to some exponentially growing modes, ${\rm Re}\,\alpha>0$, where $\alpha=dS/dt$ KS ; EJM ; Johnson:2008se . The mass density of the condensate undergoing fragmentation can be written as $\rho(x,t)=\rho_{0}+\rho_{1}(x,t)\,,$ (1) where $\rho_{1}(x,t)=\epsilon\rho_{0}\int d^{3}k\,e^{\alpha_{k}t}\cos(\omega t-\vec{k}\cdot\vec{x})\,.$ (2) The instability develops when there is a band of growing modes with positive and large enough $\alpha_{k}$ KS . The linear approximation breaks down when $\epsilon\exp(\alpha_{k}t)\sim 1$, but we will use this representation, up to its limit of applicability, to get the estimates of the gravity waves produced. The quadrupole moment that generates gravity waves is given by Weinberg $D_{ij}=\int d^{3}x\ x_{i}x_{j}\,T^{00}(x,t)\,,$ (3) where the energy-momentum tensor $T^{00}(x,t)\approx\rho(x,t)$. The space integration is over some arbitrary volume. The power emitted in gravity waves in one frequency mode is given by: $P(\omega)=\frac{2}{5}G\omega^{6}\left(D^{}_{ij}(\omega)D_{ij}(\omega)-\frac{1}{3}\|D_{ij}(\omega)\|^{2}\right)\,,$ (4) and the total energy emitted in gravitational waves, in all frequencies, is given by: $E\sim\left(2\pi\int p(\omega)d\omega\right)\times\Delta t\,,$ (5) where $\Delta t$ is the duration of the fragmentation. Based on the analytical and numerical calculations of the condensate fragmentation KS ; KK ; Kasuya2 , we take the typical parameters of the fastest-growing mode: $k\sim\xi_{k}\times 10^{2}H_{\ast},\ \omega_{k}\sim vk\sim\xi_{k}\times 10^{2}\,vH_{\ast},$ (6) where $H_{\ast}$ is the Hubble constant at the time of the condensate fragmentation, and $v$ is the typical group velocity of the wave front in the evolution of the condensate, and we expect that the dimensionless factor $\xi_{k}\sim 1$, based on the results of Refs. KS ; KK ; Kasuya2 . Since no cancellations are expected in the absence of spherical symmetry, we replace the $x_{i}x_{j}$ by $(f_{k}\times 10^{2}H_{\ast})^{-2}$ in the space integration, take the volume to be $V\sim H^{-3}_{\ast}$, and assume that $\epsilon\exp(\alpha_{k}t)\sim 1$. Then, for the leading mode, $D_{ij}(t)\sim H_{\ast}^{-3}\left(10^{2}H_{\ast}\right)^{-2}\rho_{0}\cos(\omega_{k}t-kx)\,,$ (7) and, in frequency space, $D_{ij}(\omega)\sim 10^{-4}\xi_{k}^{-2}\frac{\rho_{0}}{H_{\ast}^{5}}\,.$ (8) For $\omega\sim 10^{2}vH_{\ast}$, we estimate the power in gravitational waves in a Hubble volume: $P\sim 10^{4}\xi_{k}^{-2}\,G\frac{\rho_{0}^{2}v^{6}}{H^{4}_{\ast}}\,.$ (9) To estimate the velocity of the wavefront in the process of fragmentation, we note that, for the mode $\phi(x,t)\approx R(t)\exp\\{\alpha_{k}t\\}\cos(\omega_{k}t-kx)$, where $R(t)$ is a slowly changing function of time,111The adiabatic limit $\dot{R}/R\rightarrow 0$ is amenable to perturbation theory KS . This case corresponds to a large global charge density. For a rapidly varying $R(t)$, the numerical calculations give similar results regarding the fragmentation time and length scales KK . $v\sim\xi_{v}\left\|\dot{\phi}/\phi^{\prime}_{x}\right\|\sim\xi_{v}\ \alpha_{k}/k,\ \ \xi_{v}\sim 1,$ (10) where the uncertainty factor $\xi_{v}\sim 1$ will be retained to keep track of the uncertainty in the final answer. The relation between $\alpha_{k}$ and $k$ is given by a dispersion relation KS , which takes a simple form, $(\alpha_{k}^{2}+k^{2})\left(\alpha_{k}^{2}-\left(\dot{\Omega}^{2}-V^{\prime\prime}(R)\right)\right)+4\dot{\Omega}^{2}\alpha_{k}^{2}=0,$ (11) under the following assumptions: $H\ll k\sim\alpha_{k}\ll\sqrt{\dot{\Omega}^{2}-V^{\prime\prime}(R)}\sim\dot{\Omega}\sim m_{\phi}$ valid in the case of the fastest-growing mode in gravity mediated supersymmetry breaking models (the latter is essential for the $\dot{\Omega}\sim m_{\phi}$ condition reviews ). This equation has an approximate solution: $\alpha_{k}\approx k/\sqrt{3}$, and $v^{6}\sim\xi_{v}^{6}(\alpha_{k}/k)^{6}\sim\xi_{v}^{6}(1/\sqrt{3})^{6}\sim 10^{-2}\xi_{v}^{6}\,.$ (12) The fragmentation takes place on the time scale of the order of $\Delta t\sim\alpha_{k}^{-1}\sim\xi_{k}^{-1}10^{-2}H^{-1}_{\ast}$. (Here we neglect the possible contributions from the collisions and oscillations of Q-balls, which can take place on a much longer time scale KS ; MV1 .) The total energy in gravity waves generated in the Hubble volume is $E\sim P\Delta t\sim G\frac{\rho_{0}^{2}}{H_{\ast}^{5}}\,\xi_{k}^{-3}\xi_{v}^{6}\,.$ (13) This corresponds to the energy density in gravitational waves at the time of production $\rho_{GW\ast}\sim 10^{-3}\xi_{k}^{-3}\xi_{v}^{6}\,\frac{\rho_{0}^{2}}{H^{2}_{\ast}M_{\rm Pl}^{2}}\,,$ (14) where $M_{\rm Pl}=1/\sqrt{8\pi G}$ is the reduced Planck mass. Hence, the fraction of the energy density in gravitational waves at the time of production is $\Omega_{GW}\sim 10^{-3}\xi_{k}^{-3}\xi_{v}^{6}\,\frac{\rho_{0}^{2}}{(H_{\ast}M_{\rm Pl})^{4}}\,.$ (15) If the energy density of the condensate is comparable to the total energy density, or if the condensate energy dominates the energy in the universe, then $\rho_{0}\sim 3H_{\ast}^{2}M_{\rm Pl}^{2}$, and $\Omega_{GW}\sim 10^{-3}$. The energy density in the condensate depends on the model, and, foremost, on the type of supersymmetry breaking terms that lift the flat direction. This is because the potential along the flat direction depends on supersymmetry breaking (it vanishes in the limit of exact supersymmetry), and there are many ways to break supersymmetry. In gauge-mediated supersymmetry breaking scenarios the potential can have the form KS $V(\phi)\approx M_{S}^{4}\log\left(1+\frac{\|\phi\|^{2}}{M_{S}^{2}}\right)\,.$ (16) Here $M_{S}$ is the scale of supersymmetry breaking, which is of the order of ${\cal O}(1)$ TeV. In gravity mediated scenarios, the flat directions are lifted by mass terms that persist all the way to the Planck scale EMc : $V(\varphi)\approx m_{\phi}^{2}\left(1+K\log\left(\frac{\|\phi\|^{2}}{M_{Pl}^{2}}\right)\right)\|\phi\|^{2}\,,$ (17) where $K\sim 0.05$ (for squark directions) describes the running of the mass term EMc . Since $m_{\phi}\sim M_{S}\sim 1-10$ TeV in typical models, both potentials are phenomenologically acceptable near the minimum. However, in AD condensate and inside the Q-balls that form in its fragmentation, the vacuum expectation value (VEV) can be very large. The main difference between gauge and gravity mediated cases for us is the mass per baryon number stored in the AD condensate and in the Q-balls that form eventually as a result of the fragmentation. In the gravity mediated scenarios, the mass density is $\rho_{0}\sim m_{\phi}^{2}\phi^{2}$, the global charge density is $n_{Q}\sim m_{\phi}\phi^{2}$, and the mass per unit global charge is of the order of $m_{\phi}$, independent of the VEV $\phi_{0}$. In gauge mediated scenarios, the mass density is $\rho_{0}\sim m_{\phi}^{4}$, the global charge density is $n_{Q}\sim m_{\phi}\phi^{2}$, and the mass per unit global charge is $\rho_{0}/n_{Q}\sim m_{\phi}^{2}/\phi$ KS ; DKS . The flat directions that carry a non-zero global charge $Q=(B-L)$ contribute to the generation of the baryon asymmetry via AD process AD . The requirement that $\eta_{B}=n_{B}/n_{\gamma}\sim 10^{-10}$ implies that the total mass density of such a condensate cannot be of the order of the total density of the universe in generic models. This is true in both gauge-mediated and gravity-mediated supersymmetry breaking models. The gravity waves from the fragmentation of such a condensate are well below the capabilities of the current and planned detectors. On the other hand, there are flat directions whose baryon number $B$ and lepton number $L$ are equal to each other DRT ; GKM ; reviews . While the flat directions with $B\neq L$ contribute to baryon asymmetry of the universe, those with $B=L$, have zero $(B-L)$ density. Electroweak sphalerons destroy any primordial $(B+L)$ asymmetry, and so the corresponding $\eta_{B+L}=n_{B+L}/n_{\gamma}$ is not constrained. It is possible that, at the time of the fragmentation, $\eta_{B+L}\gg\eta_{B}$. For $B=L$ flat directions, there is no reason why $\rho_{0}$ cannot be of the order of the total energy density. The fragmentation of such flat directions can produce a detectable level of gravitational waves. There are various examples in the literature of the flat directions that can dominate the energy density of the universe, while they do not contribute to (and are not constrained by) the observed baryon asymmetry of the universe. This is the case, for example, when the effective mass for the phase direction is large during inflation, which results in the initial condition with a very small $\dot{\Omega}$ for the scalar condensate $\phi(x,t)=\phi(t)\equiv R(t)e^{i\Omega(t)}$. This is also the case when the inflation is driven by a flat direction, $udd$, $LLe$ or $NH_{u}L$ Randall:1995dj ; AEGM ; AEGJM . Another well-know example is a flat direction that acts as a curvaton and dominates the energy density of the universe at the time of oscillations and decay Curvaton . In all of these cases, the net global charge of the condensate is negligible, but the fragmentation can still occur and produce Q-balls and anti-Q-balls Kasuya1 ; Kasuya2 . Once the gravitational waves are created, they are decoupled from the rest of the plasma. We can estimate the peak frequency of the gravitational radiation observed today, $f_{\ast}=\omega_{k}/2\pi$: $\displaystyle f$ $\displaystyle=$ $\displaystyle f_{\ast}\frac{a_{\ast}}{a_{0}}=f_{\ast}\left(\frac{a_{\ast}}{a_{\rm rh}}\right)\left(\frac{g_{s,0}}{g_{s,{\rm rh}}}\right)^{1/3}\left(\frac{T_{0}}{T_{\rm rh}}\right)\,$ $\displaystyle\approx$ $\displaystyle 0.6\,{\rm mHz}\ \xi_{k}\xi_{v}\,\left(\frac{g_{s,{\rm rh}}}{100}\right)^{1/6}\left(\frac{T_{\rm rh}}{1~{}{\rm TeV}}\right)\left(\frac{f_{\ast}}{10H_{\ast}}\right)\,,$ where we have assumed that $a_{\ast}\approx a_{\rm rh}$, which also means that during the oscillations of the AD condensate the effect of the Hubble expansion is negligible. The values of relativistic degrees of freedom are $g_{s,{\rm rh}}\sim 300$, $g_{s,0}\sim 3.36$. The subscript “rh” denotes the epoch of reheating and thermalization, while the subscript “0” refers to the present time. As we discussed, the typical frequency of the oscillations of the AD condensate is $\omega_{k}\sim 10^{2}H_{\ast}$ KS ; KK ; Kasuya2 . Then for $T_{\rm rh}\sim 1$ TeV (such a value of the reheat temperature is natural when the flat direction is responsible for reheating the universe Rouzbeh ; AEGJM ), the frequency is of the order of mHz, which is in the right frequency range for LISA LISA . A higher temperature $T_{\rm rh}\sim 100$ TeV corresponds to the LIGO III frequency range, $10-100$ Hz LIGOIII . Signals in both of these ranges will be accessible to BBO BBO . Since the supersymmetry breaking scale is related to the energy in the condensate, as well as the reheating temperature, LIGOIII and BBO could be in the position to probe supersymmetry broken above 100 TeV, beyond the reach of Large Hadron Collider (LHC). The fraction of the critical energy density $\rho_{c}$ stored in the gravity waves today is $\displaystyle\Omega_{GW}$ $\displaystyle=$ $\displaystyle\Omega_{GW^{\ast}}\left(\frac{a_{\ast}}{a_{0}}\right)^{4}\left(\frac{H_{\ast}}{H_{0}}\right)^{2}$ $\displaystyle\approx$ $\displaystyle\frac{1.67\times 10^{-5}}{h^{2}}\left(\frac{100}{g_{s,\ast}}\right)^{1/3}\Omega_{GW^{\ast}}\approx 10^{-8}\,\xi_{k}^{-3}\xi_{v}^{6}\,h^{-2}$ where $a_{0}$ and $H_{0}$ are the present values of the scale factor and the Hubble expansion rate. LISA can detect the gravitational waves down to $\Omega_{GW}h^{2}\sim 10^{-11}$ at mHz frequencies, while LIGO III is sensitive to $\Omega_{GW}h^{2}\sim(10^{-5}-10^{-11})$ in the $(5-10^{3})$ Hz frequency band. Therefore, the gravitational waves with $\Omega_{GW}\sim 10^{-3}$ and $\Omega_{GW}h^{2}\sim 10^{-8}$ from the fragmentation of an AD condensate can be detected. The first results of our numerical simulations (work in progress) appear to produce the gravitational wave signal that is somewhat weaker. We attribute the difference to the value of the uncertainty factor $\xi_{k}^{-3}\xi_{v}^{6}$, which is especially sensitive to the average wavefront velocity. As one can see from eq. (14), the power generated in frequency $\omega$ is proportional to $\xi_{k}^{-3}\sim\left(\frac{\omega}{10^{2}H_{}}\right)^{-3}.$ Hence, the spectrum is strongly peaked near the longest wavelength, of the order of the Q-ball size KS ; KK ; Kasuya2 . The relatively narrow spectral width will help distinguish this signal from the gravity waves generated by inflation inflation , which are expected to have an approximately scale- invariant spectrum (and a smaller amplitude). Future numerical simulations will help refine the prediction for the signal from Q-ball formation, which can help distinguish this source from a phase transition in the early universe Kamionkowski:1993fg . LISA and LIGO III will be able to discriminate the gravity waves due to fragmentation from those of point sources, such as merging black holes and neutron stars, which have specific “chirp” properties Owen:1998dk . Furthermore, the signal discussed here will not create a significant background for the cosmic microwave polarization experiments, such as B-Pol B-pol , which can detect the gravity waves with extremely long wavelength. Some additional gravitational waves can be generated by collisions and oscillations of Q-balls KS ; MV1 . We leave the discussion of the magnitude of this additional contribution to future studies. To summarize, the fragmentation of a scalar condensate into Q-balls, which is a generic consequence of supersymmetry and inflation, can produce a detectable level of gravitational waves, up to $\Omega_{GW}h^{2}\sim 10^{-8}$, near the peak frequency of BOO and either LISA or LIGO, depending on the reheating temperature. Detection of the gravitational waves form this process can shed light on the earliest post-inflationary epoch in the history of the universe, can probe supersymmetry even if it is broken at a scale above 100 TeV, and can provide information about new physics at some very high energy scales associated with the flat directions. A.K. thanks S. Phinney for helpful discussions. The work of A.K. was supported in part by the DOE grant DE-FG03-91ER40662 and by the NASA ATFP grant NNX08AL48G. Research of A.M. is supported in part by the grant MRTN- CT-2006-035863. A.K. thanks Aspen Center for Physics for hospitality. ## References (1) K. Enqvist and A. Mazumdar, Phys. Rept. 380, 99 (2003); M. Dine and A. Kusenko, Rev. Mod. Phys. 76, 1 (2004). * (2) I. Affleck and M. Dine, Nucl. Phys. B 249, 361 (1985). * (3) A. Kusenko and M. E. Shaposhnikov, Phys. Lett. B 418, 46 (1998). * (4) A. Kusenko, V. Kuzmin, M. E. Shaposhnikov and P. G. Tinyakov, Phys. Rev. Lett. 80, 3185 (1998); M. Laine and M. E. Shaposhnikov, Nucl. Phys. B 532, 376 (1998); A. Kusenko, L. Loveridge and M. Shaposhnikov, Phys. Rev. D 72, 025015 (2005); JCAP 0508, 011 (2005). * (5) L. Randall, M. Soljacic and A. H. Guth, Nucl. Phys. B 472, 377 (1996). * (6) R. Allahverdi, K. Enqvist, J. Garcia-Bellido and A. Mazumdar, Phys. Rev. Lett. 97, 191304 (2006). R. Allahverdi, A. Kusenko and A. Mazumdar, JCAP 0707, 018 (2007). * (7) G. Rosen, J. Math. Phys. 9 (1968) 996; 9 (1968) 999; R. Friedberg, T. D. Lee and A. Sirlin, Phys. Rev. D 13, 2739 (1976); S. R. Coleman, Nucl. Phys. B 262, 263 (1985) [Erratum-ibid. B 269, 744 (1986)]; A. Kusenko, Phys. Lett. B 404, 285 (1997). * (8) A. Kusenko, Phys. Lett. B 405, 108 (1997); * (9) G. R. Dvali, A. Kusenko and M. E. Shaposhnikov, Phys. Lett. B 417, 99 (1998). * (10) http://lisa.nasa.gov/ * (11) http://www.ligo.caltech.edu/ * (12) G. M. Harry, et al., Class. Quant. Grav. 23, 4887 (2006) [Erratum-ibid. 23, 7361 (2006)]; V. Corbin and N. J. Cornish, Class. Quant. Grav. 23, 2435 (2006); http://universe.nasa.gov/new/program/bbo.html * (13) K. Enqvist and J. McDonald, Phys. Lett. B 425, 309 (1998). K. Enqvist and J. McDonald, Nucl. Phys. B 538, 321 (1999). * (14) A. Pawl, Nucl. Phys. B 679, 231 (2004). * (15) K. Enqvist, A. Jokinen and J. McDonald, Phys. Lett. B 483, 191 (2000). * (16) M. C. Johnson and M. Kamionkowski, arXiv:0805.1748 [astro-ph]. * (17) M. Axenides, S. Komineas, L. Perivolaropoulos and M. Floratos, Phys. Rev. D 61, 085006 (2000). * (18) S. Kasuya and M. Kawasaki, Phys. Rev. D 61, 041301 (2000); S. Kasuya and M. Kawasaki, Phys. Rev. D 62, 023512 (2000). S. Kasuya and M. Kawasaki, Phys. Rev. Lett. 85, 2677 (2000). S. Kasuya and M. Kawasaki, Phys. Rev. D 64, 123515 (2001). * (19) T. Multamaki and I. Vilja, Nucl. Phys. B 574, 130 (2000). K. Enqvist, A. Jokinen, T. Multamaki and I. Vilja, Phys. Rev. D 63, 083501 (2001). T. Multamaki and I. Vilja, Phys. Lett. B 535, 170 (2002). T. Multamaki, Phys. Lett. B 511, 92 (2001). * (20) M. S. Volkov and E. Wohnert, Phys. Rev. D 66, 085003 (2002). * (21) E. Palti, P. M. Saffin and E. J. Copeland, Phys. Rev. D 70, 083520 (2004). M. I. Tsumagari, E. J. Copeland and P. M. Saffin, arXiv:0805.3233 [hep-th]. * (22) M. Broadhead and J. McDonald, Phys. Rev. D 72, 043519 (2005). M. Broadhead and J. McDonald, Phys. Rev. D 69, 063510 (2004). * (23) L. Campanelli and M. Ruggieri, Phys. Rev. D 77, 043504 (2008). * (24) S. Weinberg, Gravitation and Cosmology, John Wiley $\&$ Sons. 1972. * (25) K. Enqvist, S. Kasuya and A. Mazumdar, Phys. Rev. D 66, 043505 (2002). * (26) T. Multamaki and I. Vilja, Phys. Lett. B 482, 161 (2000). T. Multamaki and I. Vilja, Phys. Lett. B 484, 283 (2000). * (27) K. Enqvist, S. Kasuya and A. Mazumdar, Phys. Rev. Lett. 89, 091301 (2002). * (28) M. Dine, L. Randall and S. D. Thomas, Nucl. Phys. B 458, 291 (1996). * (29) T. Gherghetta, C. F. Kolda and S. P. Martin, Nucl. Phys. B 468, 37 (1996). * (30) R. Allahverdi, K. Enqvist, J. Garcia-Bellido, A. Jokinen and A. Mazumdar, JCAP 0706, 019 (2007). * (31) K. Enqvist, S. Kasuya and A. Mazumdar, Phys. Rev. Lett. 90, 091302 (2003). M. Postma, Phys. Rev. D 67, 063518 (2003). K. Enqvist, A. Jokinen, S. Kasuya and A. Mazumdar, Phys. Rev. D 68, 103507 (2003). R. Allahverdi, K. Enqvist, A. Jokinen and A. Mazumdar, JCAP 0610, 007 (2006). * (32) R. Allahverdi and A. Mazumdar, JCAP 0610, 008 (2006). M. Berkooz, D. J. H. Chung and T. Volansky, Phys. Rev. Lett. 96, 031303 (2006). R. Allahverdi and A. Mazumdar, arXiv:0802.4430 [hep-ph]. * (33) A. A. Starobinsky, JETP Lett. 30, 682 (1979) [Pisma Zh. Eksp. Teor. Fiz. 30, 719 (1979)]. S. Dodelson, W. H. Kinney and E. W. Kolb, Phys. Rev. D 56, 3207 (1997); M. Maggiore, Phys. Rept. 331, 283 (2000); T. L. Smith, M. Kamionkowski and A. Cooray, Phys. Rev. D 73, 023504 (2006); arXiv:0802.1530 [astro-ph]. * (34) M. Kamionkowski, A. Kosowsky and M. S. Turner, Phys. Rev. D 49, 2837 (1994); C. Grojean and G. Servant, Phys. Rev. D 75, 043507 (2007). * (35) B. J. Owen and B. S. Sathyaprakash, Phys. Rev. D 60, 022002 (1999). * (36) P. de Bernardis, M. Bucher, C. Burigana, L. Piccirillo and f. P. Collaboration, arXiv:0808.1881 [astro-ph].	arxiv-papers	2008-07-29T16:01:03	2024-09-04T02:48:57.014576	{ "license": "Public Domain", "authors": "Alexander Kusenko and Anupam Mazumdar", "submitter": "Alexander Kusenko", "url": "https://arxiv.org/abs/0807.4554" }
0807.4715	# Eventually Expanding Maps Peyman Eslami and Pawel Gora ###### Abstract. In this paper we show that the piecewise linear map $f(x)=px\chi_{[0,1/p]}(x)+(sx-s/p)\chi_{(1/p,1]}(x)$, $p>1$, $0<s<1$ which has an expanding, onto branch and a contracting branch is eventually piecewise expanding and exact. ## 1\. Introduction Let $f$ be a piecewise linear function on the unit interval $[0,1]$ with two increasing branches, one of which has slope greater than one and the other less than one. In this paper we will show that $f$ is piecewise expanding under certain conditions (first branch onto, and second branch touching the $x$-axis). In his paper [2], Md. Shafiqul Islam proved similar results for functions that are linearly conjugate to $f$ where the first branch is not necessarily onto, but has slope $1<p\leq 2$, and the second branch does not necessarily hit the $x$-axis and has slope $0<s\leq 1$; also, the point of discontinuity is fixed at $x=1/2$. ## 2\. Preliminaries ###### Definition 1 (Eventually piecewise expanding function). A piecewise linear function $f$ is said to be piecewise expanding if all of its branches have slope $>1$. If this is true for some iterate $f^{N}$ of $f$, then we say $f$ is eventually piecewise expanding. ###### Definition 2 (Exactness). Let $(X,\mathcal{B},\mu)$ be a normalized measure space and let $f\colon X\to X$ be a measure preserving map such that $f(A)\in\mathcal{B}$ for every $A\in\mathcal{B}$. $f$ is said to be exact if for every $A\in\mathcal{B}$ with $\mu(A)>0$ we have, $\lim_{n\rightarrow\infty}\mu(f^{n}(A))=1\text{.}$ Denote by $F$ the class of functions $f\colon[0,1]\to[0,1]$ defined by $f(x)=(px+a)\chi_{[0,d]}(x)+(sx+b)\chi_{(d,1]}(x)\text{,}$ where $p>1$, $0<s<1$, and $a$,$b$, and $d$(the point of discontinuity of $f$) are such that $f([0,1])\subset[0,1]$. In the next section we prove results about piecewise expansion and exactness of functions that belong to $F$. ## 3\. Main Theorem In this section we will consider the case where $f\in F$, $a=0$, $b=-s/p$ and $d=1/p$. That is, $f(x)=px\chi_{[0,1/p]}(x)+(sx-s/p)\chi_{(1/p,1]}(x)\text{.}$ See figure below: We will show the following: ###### Theorem 1. There exists $N\in{\mathbb{N}}$ such that for every interval $I\subset[0,1]$ $f^{N}$ is expanding on $I$. ###### Proof. Suppose BWOC that for every $N\in{\mathbb{N}}$ there exists an interval $I\subset[0,1]$ such that $f^{N}$ is not expanding on $I$. Then there exists an interval $J\subset I$ such that $f^{N}$ is contracting(having slop $\leq$ 1) on $J$. WLOG assume $f^{N}$ has a constant slope on $J$. Let $m=m(J)=\text{ \\# of expansions of J under f during $N$ iterations of $f$,}$ $n=n(J)=\text{ \\# of contractions of J under f during $N$ iterations of $f$.}$ Note that $0\leq n,m\leq N$ and $m+n=N$. Since $f^{N}$ is contracting on $J$, we must have: $p^{m}s^{n}\leq 1\Leftrightarrow m\leq-\frac{\ln{s}}{\ln{p}}n=-\frac{\ln{s}}{\ln{p}}(N-m)\Leftrightarrow m\leq\frac{-\ln s}{\ln p-\ln s}N\text{.}$ $J$ expands $m$ times and contracts $n$ times during $N$ iterations, hence $J$ must contract consecutively $\lceil\frac{n}{m+1}\rceil$ times during $N$ iterations of $f$. The minimum value of $\lceil\frac{n}{m+1}\rceil$ (minimum number of consecutive contractions) is achieved when $m=\lfloor\frac{-\ln s}{\ln p-\ln s}N\rfloor$ and it is: $\left\lceil\frac{n}{\lfloor\frac{-\ln s}{\ln p-\ln s}N\rfloor+1}\right\rceil\text{.}$ Since $0<c=\frac{-\ln s}{\ln p-\ln s}<1$, $\left\lceil\frac{n}{\lfloor\frac{-\ln s}{\ln p-\ln s}N\rfloor+1}\right\rceil\rightarrow\left\lceil\frac{1}{c}\right\rceil=\left\lceil 1-\frac{\ln p}{\ln s}\right\rceil\text{, as $N\rightarrow\infty$.}$ So this gives us a lower bound on the number of contractions of $J$. On the other hand, we can find an upper bound for the number of consecutive contractions of $J$ by looking at the forward orbit of $1$ since intervals close to $1$ can have the most number of consecutive contractions . Assuming that $f^{i}(x)\in(\frac{1}{p},1]$ for $i=0,1,2,\ldots,j$ we get: $f^{j}(x)=s^{j}x-\frac{s}{p}(\frac{1-s^{j}}{1-s})\text{.}$ We are interested in the maximum value $j$ such that $f^{j}(1)>\frac{1}{p}$. $\displaystyle f^{j}(1)$ $\displaystyle=$ $\displaystyle s^{j}-\frac{s}{p}(\frac{1-s^{j}}{1-s})>\frac{1}{p}$ $\displaystyle\Rightarrow$ $\displaystyle p(1-s)s^{j}-s(1-s^{j})>1-s$ $\displaystyle\Rightarrow$ $\displaystyle p(1-s)s^{j}-s+s^{j+1}>1-s$ $\displaystyle\Rightarrow$ $\displaystyle s^{j}(p(1-s)+s)>1,\ \ p(1-s)+s>0$ $\displaystyle\Rightarrow$ $\displaystyle s^{j}>\frac{1}{p(1-s)+s}$ $\displaystyle\Rightarrow$ $\displaystyle j<\frac{\ln(\frac{1}{p(1-s)+s})}{\ln s}=-\frac{\ln(p(1-s)+s)}{\ln s}$ So, the maximum possible value for $j$ such that $f^{j}(1)>\frac{1}{p}$ is: $\left\lfloor-\frac{\ln(p(1-s)+s)}{\ln s}\right\rfloor$ Therefore, an interval contained in $(\frac{1}{p},1]$ can contract consecutively at most $\left\lfloor-\frac{\ln(p(1-s)+s)}{\ln s}\right\rfloor+1\text{ times.}$ This gives us an upper bound on the number of contractions of $J$. In conclusion, we must have: $\left\lceil 1-\frac{\ln p}{\ln s}\right\rceil\leq\left\lfloor-\frac{\ln(p(1-s)+s)}{\ln s}\right\rfloor+1\text{.}$ But, since $0<s<1$ and $p>1$, $\displaystyle 0>s(1-p)$ $\displaystyle\Leftrightarrow$ $\displaystyle 0>s-ps$ $\displaystyle\Leftrightarrow$ $\displaystyle p>p(1-s)+s$ $\displaystyle\Leftrightarrow$ $\displaystyle\ln p>\ln(p(1-s)+s)$ $\displaystyle\Leftrightarrow$ $\displaystyle-\frac{\ln p}{\ln s}>-\frac{\ln(p(1-s)+s)}{\ln s}$ $\displaystyle\Rightarrow$ $\displaystyle\left\lceil-\frac{\ln p}{\ln s}\right\rceil>\left\lfloor-\frac{\ln(p(1-s)+s)}{\ln s}\right\rfloor$ $\displaystyle\Leftrightarrow$ $\displaystyle\left\lceil 1-\frac{\ln p}{\ln s}\right\rceil>\left\lfloor-\frac{\ln(p(1-s)+s)}{\ln s}\right\rfloor+1\text{,}$ a contradiction. Therefore, $f(x)=px\chi_{[0,1/p]}(x)+(sx-s/p)\chi_{(1/p,1](x)}$ is eventually piecewise expanding. ∎ ###### Theorem 2. $f$ as defined above is exact. ###### Proof. By Theorem 1, there exist $n\in{\mathbb{N}}$ such that $f^{n}$ is piecewise expanding. It is known that the composition of two piecewise linear maps whose branches touch zero has the same property. Therefore, the branches of $f^{k}$ touch zero for any $k\geq 1$. Now, if $A$ is any interval in $[0,1]$, since $f^{n}$ is piecewise expanding and all branches of $f^{n}$ touch $0$, $A$ will eventually contain an interval $B$ containing $0$. Since the first branch of $f$ is onto and expanding and includes the fixed point $0$, this interval $B$ eventually covers $[0,1]$. $f^{nk}$ is piecewise expanding, and by Theorem $7.2.1$ [1], all of its ergodic components are exact. By Theorem $8.2.1$ [1] each exact component contains a neighborhood of an inner partition point and images of these neighborhoods touch $0$. Therefore, since any interval eventually covers $[0,1]$ under $f$, $f^{nk}$ has only one exact component. Since $f^{nk}$ is exact, $f$ is exact. ∎ ## References * [1] Boyarsky, Abraham; Góra, Paweł, Laws of chaos. Invariant measures and dynamical systems in one dimension, Probability and its Applications, Birkhäuser Boston, Inc., Boston, MA, 1997, MR1461536 (99a:58102). * [2] Islam, Shafiqul, Absolutely continuous invariant measures of linear interval maps, Int. J. Pure Appl. Math. 27 (2006), no. 4, 449–464, MR2223985 (2006k:37100).	arxiv-papers	2008-07-29T17:43:59	2024-09-04T02:48:57.019581	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Peyman Eslami, Pawel Gora", "submitter": "Peyman Eslami", "url": "https://arxiv.org/abs/0807.4715" }
0807.4721	Anatomy and Phenomenology of the Lepton Flavor Universality in SUSY Theories A. Masieroa, P. Paradisib, and R. Petronzioc a Dip. di Fisica ‘G. Galilei’, Univ. di Padova and INFN, Sezione di Padova, Via Marzolo 8, I-35131, Padua, Italy bPhysik-Department, Technische Universität München, D-85748 Garching, Germany cDip. di Fisica, Università di Roma “Tor Vergata” and INFN, Sezione di Roma, “Tor Vergata”, Via della Ricerca Scientifica 1, I-00133, Rome, Italy Abstract High precision electroweak tests, such as deviations from the Standard Model expectations of the Lepton Flavor Universality breaking in $K\rightarrow\ell\nu_{\ell}$ (with $l=e$ or $\mu$), represent a powerful tool to test the Standard Model and, hence, to constrain or obtain indirect hints of New Physics beyond it. We explore such a possibility within Supersymmetric theories. Interestingly enough, a process that in itself does not need lepton flavor violation to occur, i.e. the violation of $\mu-e$ non-universality in $K_{\ell 2}$, proves to be quite effective in constraining not only relevant regions of SUSY models where lepton flavor is conserved, but even those where specific lepton flavor violating contributions arise. Indeed, a comparison with analogous bounds coming from $\tau$ lepton flavor violating decays shows the relevance of the measurement of $R_{K}$ to probe Lepton Flavor Violation in SUSY. We outline the role and the interplay of the direct New Physics searches at the LHC with the indirect searches performed by LFU tests. ## 1 Introduction The study of Lepton Flavor Universality (LFU) represents a powerful tool to test the Standard Model (SM) and, hence, to constrain or obtain indirect hints of new physics beyond it. Kaon and pion physics are obvious grounds where to perform such tests, for instance in the $\pi\rightarrow\ell\nu_{\ell}$ and $K\rightarrow\ell\nu_{\ell}$ decays, where $\ell=e$ or $\mu$. In particular, defining $(R_{P}^{e/\mu})_{SM}=\Gamma(P\to e\nu_{e})_{SM}/\Gamma(P\to\mu\nu_{e})_{SM}$ and $(R_{P}^{e/\mu})_{exp.}=\Gamma(P\to e\nu)_{exp.}/\Gamma(P\to\mu\nu)_{exp.}$, the difference of the ratio $R_{P}^{e/\mu}=\frac{(R_{P}^{e/\mu})_{exp.}}{(R_{P}^{e/\mu})_{SM}}=1+\Delta r_{P}^{e/\mu}$ (1) from unit signals the presence of LFU violating New Physics (NP). Given that $(R_{P}^{e/\mu})_{SM}$ is accurately predicted, both for $P=\pi$ (0.02% accuracy [1]) and $P=K$ (0.04% accuracy [1]), it turns out that the determination of $(R_{P}^{e/\mu})$ constitutes a major precision test of the SM. These precision tests are equally interesting and fully complementary to the flavor-conserving electroweak precision tests and to the FCNC tests performed in hadronic and leptonic physics (rare kaon, charm and B physics, lepton Flavor Violation (LFV)): the smallness of NP effects is more than compensated in terms of NP sensitivity by the excellent experimental resolution and the good theoretical control. The limiting factor in the determination $R_{K}^{e/\mu}$ is the $K\to e\nu$ rate, whose experimental knowledge has been quite poor so far. The current world average $(R_{K}^{e/\mu})_{exp.}=(2.45\pm 0.11)\times 10^{-5}$ [2] will be soon improved thanks to a series of preliminary results by NA48/2 and KLOE (see Fig. 1). The two results by NA48/2, being based on different data sets (2003 [3] and 2004 [3], respectively) with different running conditions, should be regarded as completely independent. Combining these new results with the PDG value yields [3] $(R_{K}^{e/\mu})_{exp.}=(2.457\pm 0.032)\times 10^{-5}~{}.$ (2) This result is in good agreement with the SM expectation and has a relative error ($\sim 1.3\%$) three times smaller compared to the previous world average. Further improvements in the knowledge of $(R_{K}^{e/\mu})_{exp.}$ would be more than welcome. Moreover, also the KLOE collaboration will reach an error down to the 1% level on $R_{K}^{e/\mu}$, once the remaining statistics will be added and the reconstruction efficiency improved [3]. Last but not least, an error on $(R_{K}^{e/\mu})_{exp.}$ of about $0.3\%$ is the ambitious goal of the 2007 dedicated run of the CERN-P326 collaboration (the successor of NA48) [3]. If these expectations will be fulfilled, in a short term the error on the world average of $R_{K}^{e/\mu}$ will decrease by an additional factor of four. In the following, we consider low-energy minimal SUSY extensions of the SM (MSSM) with R parity as the source of NP to be tested by $R_{K}^{e/\mu}$ [4]. As discussed in [4], it is indeed possible for regions of the MSSM to obtain $\Delta r^{e-\mu}_{\\!NP}$ of $\mathcal{O}(10^{-2})$ and, such large contributions to $K_{\ell 2}$, do not arise from SUSY lepton flavor conserving (LFC) effects, but, rather, from lepton flavor violating (LFV) ones. $(R_{K}^{e/\mu})_{exp.}$ $[10^{-5}]$ PDG 2006 [2] $2.45\pm 0.11$ NA48/2 ’03 prel. $2.416\pm 0.043\pm 0.024$ NA48/2 ’04 prel. $2.455\pm 0.045\pm 0.041$ KLOE prel. $2.55\pm 0.05\pm 0.05$ SM prediction $2.472\pm 0.001$ Figure 1: Current experimental data on $R_{K}^{e/\mu}$ from [3]. The main reason is that, whenever new physics acts in $K_{\ell 2}$ to create a departure from the strict SM $\mu-e$ universality, these new contributions will typically be proportional to the lepton masses. Hence, what occurs in the SUSY case is that LFC contributions are suppressed with respect to the LFV ones by higher powers of the first two generations lepton masses (it turns out that the first contributions to $\Delta r^{e-\mu}_{\\!NP}$ from LFC terms arise at the cubic order in $m_{\ell}$, with $\ell=e,\mu$). Instead, for the LFV contributions to $R_{K}^{e/\mu}$ one can select those which involve flavor changes from the first two lepton generations to the third one with the possibility of picking up terms proportional to the tau-Yukawa coupling which can be large in the large $\rm{\tan\beta}$ regime (the parameter $\rm{\tan\beta}$ denotes the ratio of Higgs vacuum expectation values responsible for the up- and down- quark masses, respectively). Moreover, the relevant one-loop induced LFV Yukawa interactions are known [5] to acquire an additional $\rm{\tan\beta}$ factor with respect to the tree level LFC Yukawa terms. Thus, the loop suppression factor can be (partially) compensated in the large $\rm{\tan\beta}$ regime. In this paper, we analise the domain of $\Delta r_{K}^{e/\mu}$ between $10^{-2}$ and $10^{-3}$. We show that: i) if $\Delta r_{K}^{e/\mu}$ is found to be $\Delta r_{K}^{e/\mu}\geq 5\times 10^{-3}$, then the signal unambiguously indicates the presence of LFV sources. ii) if $\Delta r_{K}^{e/\mu}\leq 5\times 10^{-3}$, then both the LFC and LFV sources can account for the effect. iii) a value of $\Delta r_{K}^{e/\mu}$ between $5\times 10^{-3}$ and $10^{-3}$ severely constrains the parameter space in the $M_{H}-\tan\beta$ plane. iv) if a signal exists at a such a level, the LHC results become the crucial tool to discriminate between the LFC and LFV sources of LFU breaking. v) there exists a strong correlation between large LFU violation and LFV in lepton decays (mainly $\tau$ decays); another interesting relation concerns the regions of SUSY parameter space where the deviation from the SM expectation for the muon anomalous magnetic moment finds a SUSY explanation and that allowing for a sizeable LFU violation. The paper is organized as follow: in Section 2, we outline general considerations about LFU in $P_{l2}$. In section Section 3, we specialize to the LFV case while in Section 4, we discuss the additional possibility of LFC contributions. In Section 5, we list the constraints we have imposed on the SUSY parameter space before starting the analysis of the LFU breaking effects. In Section 6, we discuss the correlation between LFU violation and LFV in lepton decays and their possible connection with a SUSY explanation for the anomalous magnetic moment of the muon. In Section 7, we present the quantitative analysis of our results incorporating the constraints of the above sections. In Section 8, we extend the analysis of LFU breaking effects to a generic two Higgs Doublet Model with tree level flavor changing interactions between the Higgs bosons and the fermions. Finally, in Section 9 we summarize the main results of the present analysis. ## 2 Lepton Flavor Universality in $P_{\ell 2}$ Within the SM, possible departures from the LFU are predicted to be $\|\Delta r^{\ell_{1}/\ell_{2}}_{\rm SM}\|={\cal O}[(\alpha/4\pi)\times(m^{2}_{\ell_{1(2)}}/M^{2}_{W})]\,,$ (3) and thus completely negligible. This explains why the study of LFU breaking represents a very useful tool to look for NP effects. On general grounds, violations of LFU in charged current interactions can be classified into two classes: i) corrections to the strength of the effective $(V-A)\times(V-A)$ four-fermion interaction, ii) four-fermion interactions with new Lorentz structures. As an example of the first class, we mention the $W\ell\nu_{\ell}$ vertex correction through a loop of new particles: the induced effect is of order $(\alpha/4\pi)\times(M^{2}_{W}/M^{2}_{NP})$, hence unobservably small. Second class is definitely more promising: the typical example is the scalar current induced by tree level Higgs exchange, with mass-dependent coupling ($H\ell\nu\sim m_{\ell}\tan\beta$). In the following, we will analyze LFU breaking effects arising from this latter class occurring in $P_{\ell 2}$. Due to the V-A structure of the weak interactions, the SM contributions to $P_{\ell 2}$ are helicity suppressed; hence, these processes are very sensitive to non-SM effects (such as multi-Higgs effects) which might induce an effective pseudoscalar hadronic weak current. In particular, charged Higgs bosons ($H^{\pm}$) appearing in any model with two Higgs doublets (including the SUSY case) can contribute at tree level to the above processes. The relevant four-Fermi interaction for the decay of charged mesons induced by $W^{\pm}$ and $H^{\pm}$ has the following form: $\frac{4G_{F}}{\sqrt{2}}V_{ud}\left[(\,\overline{u}\gamma_{\mu}P_{L}d\,)(\,\overline{l}\gamma^{\mu}P_{L}\nu_{l}\,)+\Delta^{ij}\,t_{\beta}^{2}\left(\frac{m_{d}m_{l_{i}}}{m^{2}_{H^{\pm}}}\right)(\,\overline{u}P_{R}d\,)(\,\overline{l}_{i}P_{L}\nu_{j}\,)\right]\,,$ (4) where $P_{R,L}=(1\pm\gamma_{5})/2$ and we kept only the $t_{\beta}$ (with $t_{\beta}=\tan\beta$) enhanced part of the $H^{\pm}ud$ coupling, namely the $m_{d}t_{\beta}$ term. The quantity $\Delta^{ij}=\Delta^{ij}(\delta^{ij},\tan\beta,m_{l_{j}},\tilde{m})$ may depend, in general, on the mixing angle $\delta^{ij}$ regulating the flavor transition $ij$, on the $\tan\beta$ parameter, on the masses of all charged lepton generations and, finally, on all the possible susy masses $\tilde{m}$ determining the effective vertex. The decays $P\rightarrow\ell\nu$ ($P=K,\pi$) proceed via the axial-vector part of the $W^{\pm}$ coupling and via the pseudoscalar part of the $H^{\pm}$ coupling. Then, once we implement the PCAC’s $<0\|\overline{u}\gamma_{\mu}\gamma_{5}d\|M^{-}>=if_{M}p^{\mu}_{M}\,\,\,\,\,\,,\,\,\,\,\,<0\|\overline{u}\gamma_{5}d\|M^{-}>=-if_{M}\frac{m^{2}_{M}}{m_{d}+m_{u}}\,,$ (5) it is found that $R_{P\ell_{i}\nu}=\bigg{[}1-\Delta^{ii}\bigg{(}\frac{m_{d_{P}}}{m_{d_{P}}\\!+\\!m_{u_{P}}}\bigg{)}\frac{m^{2}_{P}}{M^{2}_{H^{+}}}t_{\beta}^{2}\bigg{]}^{2}+\Sigma_{j\neq i}\|\Delta^{ij}\|^{2}\bigg{(}\frac{m_{d_{P}}}{m_{d_{P}}\\!+\\!m_{u_{P}}}\bigg{)}^{2}\frac{m^{4}_{P}}{M^{4}_{H^{+}}}t_{\beta}^{4}$ (6) The tree level charged Higgs exchange leads to a contribution with $i=j$ and $\Delta^{ii}=1$. However, the introduction of a charged scalar current (induced by a $H^{+}$) does not introduce any deviation from the SM expectation of the LFU breaking in $R_{P}^{e/\mu}$. Indeed, we observe that the SM amplitude is proportional to $m_{\ell}$ because of the helicity suppression while the charged Higgs one carries the $m_{\ell}$ dependence through the Yukawa coupling. As a result, the first SUSY contributions violating the $\mu-e$ universality in $P\to\ell\nu$ decays arise at the one-loop level with various diagrams involving exchanges of (charged and neutral) Higgs scalars, charginos, neutralinos and sleptons. For our purpose, it is relevant to divide all such contributions into two classes: i) LFC contributions, where the charged meson M decays without FCNC in the leptonic sector, i.e. $P\to\ell\nu_{\ell}$; ii) LFV contributions $P\to\ell_{i}\nu_{k}$, with $i$ and $k$ referring to different generations (in particular, the interesting case will be for $i=e,\mu$, and $k=\tau$). In the following sections we address separately the case of LFC and LFV contributions. ## 3 The lepton flavor violating case Within SUSY theories, there exist two different classes of LFV interactions: 1. i) Gauge-mediated LFV interactions , 2. ii) Higgs-mediated LFV interactions . As regards the class $i)$, LFV effects are induced by the exchange of gauginos and sleptons; these contributions decouple with the heaviest mass $m_{SUSY}$ circulating in the slepton/gaugino loops. Concerning the case $ii)$, we remind that models containing at least two Higgs doublets generally allow flavor violating couplings of the Higgs bosons with the fermions [6]. However in the MSSM such LFV couplings are absent at tree level since we have one higgs doublet coupling uniquely to the up-sector, while the other higgs doublet couples only to the down-sector. However, once non holomorphic terms are generated by loop effects (so called HRS corrections [7]) and given a source of LFV among the sleptons, Higgs-mediated $H\overline{\ell}_{i}\ell_{j}$ LFV couplings are unavoidable [5]. These effects decouple with the heavy Higgs mass scale $m_{H}$ but they do not decouple with the mass scale of the sleptons/gauginos circulating in the loop, given that the effective LFV Yukawa couplings arise from dimension four operators. As it is well known, higgs mediated effects to rare decays start being competitive with the gaugino mediated ones when $m_{SUSY}$ is roughly one order of magnitude heavier then $m_{H}$ and for $\tan\beta\sim\mathcal{O}(50)$ [8]. On general ground, there is no reason to assume that $m_{H}\simeq m_{SUSY}$, unless specific models of SUSY breaking are assumed. We stress that the quantity which is determined experimentally and accounts for the deviation from the $\mu-e$ universality is $(R_{P}^{e/\mu})_{exp.}=\frac{\sum_{i}\Gamma(P\rightarrow e\nu_{i})}{\sum_{i}\Gamma(P\rightarrow\mu\nu_{i})}\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,i=e,\mu,\tau.$ (7) with the sum extended over all (anti)neutrino flavors. In fact, experimentally, it is possible to measure only the charged lepton flavor in the decay products. The dominant SUSY contributions to $R_{P}^{e/\mu}=(R_{P}^{e/\mu})_{exp.}/(R_{P}^{e/\mu})_{SM}$ arise from the charged Higgs exchange. One could naively think that the SUSY effects in the LFV channels $P\to\ell_{i}\nu_{k}$ are further suppressed with respect to the LFC ones. On the contrary, charged Higgs mediated LFV contributions, in particular in the kaon decays into an electron or a muon and a tau neutrino, can be strongly enhanced. In particular, the expressions for the effective couplings $\Delta^{ij}$ in Eq. 4 read $\displaystyle\Delta^{\ell\ell}$ $\displaystyle=$ $\displaystyle\frac{1}{(1+\epsilon t_{\beta})(1+\epsilon_{\ell}t_{\beta})}+\frac{m_{\tau}}{m_{\ell}}\frac{\Delta^{\ell\ell}_{RL}t_{\beta}}{(1+\epsilon t_{\beta})(1+\epsilon_{\tau}t_{\beta})^{2}}$ (8) $\displaystyle\Delta^{\ell\tau}$ $\displaystyle=$ $\displaystyle\frac{m_{\tau}}{m_{\ell}}\frac{\Delta^{3l}_{R}t_{\beta}}{(1+\epsilon t_{\beta})(1+\epsilon_{\tau}t_{\beta})^{2}}\qquad\qquad\qquad\qquad\qquad\qquad\qquad l=e,\mu\,.$ (9) The first term in Eq. 8 refers to a tree level charged Higgs exchange while the second one stems from a double source of LFV that, as a final result, preserves the flavor. On the contrary, the contributions of Eq. 9 refer to LFV channels. Notice that the (loop induced) contributions arising from LFV sources is enhanced by the factor $m_{\tau}/m_{\ell}$ (compared to the contributions from a tree level charged Higgs exchange) when the electron or muon in $(R_{P}^{e/\mu})_{exp.}$ are accompanied by a tau neutrino. In the above expressions, we have also included the threshold corrections (proportional to $\epsilon,\epsilon_{\ell}$, with $\epsilon\sim\alpha_{s}/4\pi$ and $\epsilon_{\ell}\sim\alpha_{2}/4\pi$)) for the quark and lepton yukawas appearing when we integrate out heavy degrees of freedom from the low energy effective theory [7]. In particular, a relevant observation for the following analysis is that the one-loop induced $\epsilon_{\ell}=\epsilon_{\ell}(m^{2}_{\tilde{\ell}},M^{2}_{\tilde{\chi}})$ resummation factors carry a lepton flavor dependence through the slepton masses. Thus, as we will see in the next section, if the slepton generations have different masses, the $\epsilon_{\ell}$ factors will generate a breaking of the LFU in low-energy observables. The $\Delta^{3\ell}_{R,RL}$ terms are induced at one loop level by the exchange of Bino or Bino-Higgsino and sleptons. Since the Yukawa operator is of dimension four, the quantities $\Delta^{3\ell}_{R}$ depend only on ratios of SUSY masses, hence avoiding SUSY decoupling. In the so called mass insertion (MI) approximation, the expressions of $\Delta^{3\ell}_{R,RL}$ are given by: $\Delta^{3\ell}_{R}\\!\simeq\\!\frac{\alpha_{Y}}{8\pi}\mu M_{1}m^{2}_{R}\delta^{3\ell}_{RR}\left[I^{{}^{\prime}}\\!(M^{2}_{1},\mu^{2},m^{2}_{R})\\!-\\!(\mu\\!\leftrightarrow\\!m_{L})\right]$ (10) $\Delta^{\ell\ell}_{RL}\\!\simeq\\!-\frac{\alpha_{Y}}{16\pi}\mu M_{1}m^{2}_{L}m^{2}_{R}\,\delta^{\ell 3}_{RR}\delta^{3\ell}_{LL}\,I^{{}^{\prime\prime}}\\!(M^{2}_{1},m^{2}_{L},m^{2}_{R})\,,$ (11) where $\mu$ is the the Higgs mixing parameter, $M_{1}$ is the Bino ($\tilde{B}$) mass and $m^{2}_{L(R)}$ stands for the left-left (right-right) slepton mass matrix entry. The LFV MIs, i.e. $\delta^{3\ell}_{XX}\\!=\\!({\tilde{m}}^{2}_{\ell})^{3\ell}_{XX}/m^{2}_{X}$ $(X=L,R)$, are the off-diagonal flavor changing entries of the slepton mass matrix. The loop function $I^{{}^{\prime}}(x,y,z)$ is such that $I^{{}^{\prime}}(x,y,z)=dI(x,y,z)/dz$, where $I(x,y,z)$ refers to the standard three point one-loop integral which has mass dimension -2; morever, $I^{{}^{\prime\prime}}(x,y,z)=d^{2}I(x,y,z)/dydz$. As it is clearly shown by Eq. 10, $\Delta^{3\ell}_{R}$ vanishes for $\mu=m_{L}$. On the other hand, both $\Delta^{3\ell}_{R}$ and $\Delta^{3\ell}_{RL}$ assume the their maximum values when $\mu\gg M_{1},m_{L},m_{R}$; this is easily understood reminding that Higgs mediated effects originate from non holomorphic corrections that are driven by the $\mu H_{1}H_{2}$ term in the superpotential. In particular, in the limit where $\mu\gg\tilde{m}\\!=\\!M_{1}\\!=\\!m_{L}\\!=\\!m_{R}$, it turns out that $\Delta^{3\ell}_{R}\simeq\alpha_{Y}/16\pi\times\mu/\tilde{m}\times\delta^{3\ell}_{RR}$ and $\Delta^{\ell\ell}_{RL}\simeq\alpha_{Y}/32\pi\times\mu/\tilde{m}\times\delta^{\ell 3}_{RR}\delta^{3\ell}_{LL}$ 111Im($\delta^{13}_{RR}\delta^{31}_{LL}$) is strongly constrained by the electron electric dipole moment [10]. However, sizable contributions to $R^{LFV}_{K}$ can still be induced by Re($\delta^{13}_{RR}\delta^{31}_{LL}$).. Making use of the effective couplings of Eq. (9), it turns out that the dominant contribution to $\Delta r^{e-\mu}_{NP}$ reads $\displaystyle R^{e/\mu}_{K}\simeq\left\|1-\frac{m^{2}_{K}}{M^{2}_{H}}\frac{m_{\tau}}{m_{e}}\frac{\Delta^{11}_{RL}t_{\beta}^{3}}{(1+\epsilon t_{\beta})(1+\epsilon_{\tau}t_{\beta})^{2}}\right\|^{2}+\bigg{(}\frac{m^{4}_{K}}{M^{4}_{H}}\bigg{)}\bigg{(}\frac{m^{2}_{\tau}}{m^{2}_{e}}\bigg{)}\frac{\|\Delta^{31}_{R}\|^{2}t_{\beta}^{6}}{(1+\epsilon t_{\beta})^{2}(1+\epsilon_{\tau}t_{\beta})^{4}}.$ (12) In the above expression, we have included the interference between SM and SUSY LFC terms (arising from a double LFV source). In Eq. (12) terms proportional to $\Delta^{32}_{R}$ are neglected given that they are suppressed by a factor $m^{2}_{e}/m^{2}_{\mu}$ with respect to the term proportional to $\Delta^{31}_{R}$. Taking $\Delta^{31}_{R}\\!\simeq\\!5\cdot 10^{-4}$ (by means of a numerical analysis, it turns out that $\Delta^{3\ell}_{R}\leq 10^{-3}$ [9]), $\tan\beta\\!=\\!40$ and $M_{H}\\!=\\!500GeV$ we end up with $\Delta r^{e-\mu}_{\\!K\,SUSY}\simeq 10^{-2}$. We see that in the large (but not extreme) $\rm\tan\beta$ regime and with a relatively heavy $H^{\pm}$, it is possible to reach contributions to $\Delta r^{e-\mu}_{\\!K\,SUSY}$ at the percent level thanks to the possible LFV enhancements arising in SUSY models. Turning to pion physics, one could wonder whether the analogous quantity $\Delta r^{e-\mu}_{\\!\pi\,SUSY}$ is able to constrain SUSY LFV. However, the correlation between $\Delta r^{e-\mu}_{\\!\pi\,SUSY}$ and $\Delta r^{e-\mu}_{\\!K\,SUSY}$: $\Delta r^{e-\mu}_{\pi\,SUSY}\simeq\left(\frac{m_{d}}{m_{u}+m_{d}}\right)^{2}\left(\frac{m^{4}_{\pi}}{m^{4}_{k}}\right)\Delta r^{e-\mu}_{\\!K\,SUSY}$ (13) clearly shows that the constraints on $\Delta r^{e-\mu}_{\\!K\,susy}$ force $\Delta r^{e-\mu}_{\pi\,susy}$ to be much below its current experimental upper bound. ## 4 The lepton flavor conserving case We now reconsider Eq. 4 in the $i=j$ case, i.e. the lepton flavor conserving channels. In absence of LFV interactions, $\Delta_{ii}$ reads $\Delta_{ii}=1/[(1+\epsilon t_{\beta})(1+\epsilon_{\ell}t_{\beta})]$. This leads to $\frac{\Gamma(P\to\ell\nu)}{\Gamma(P\to\ell\nu)_{SM}}=\bigg{[}1-\bigg{(}\frac{m_{d_{P}}}{m_{d_{P}}\\!+\\!m_{u_{P}}}\bigg{)}\frac{m^{2}_{P}}{M^{2}_{H^{+}}}\frac{t_{\beta}^{2}}{(1+\epsilon t_{\beta})(1+\epsilon_{\ell}t_{\beta})}\bigg{]}^{2}$ (14) As discussed in the previous section, tree level $H^{+}$ contributions do not introduce any breaking of the LU in $R_{P}^{e/\mu}$. However, this is strictly true only if $\epsilon_{e}=\epsilon_{\mu}$, as clearly shown by Eq. 14. In particular, for non universal slepton masses, it turns out that $\epsilon_{e}\neq\epsilon_{\mu}$ (remind that $\epsilon_{\ell}=\epsilon_{\ell}(m^{2}_{\tilde{\ell}},M^{2}_{\tilde{\chi}})$), and LFU breaking effects are generated. By means of Eq.14, we find that $\Delta r_{P}^{e/\mu}$ is well approximated by the following expression $\Delta r_{P}^{e/\mu}\simeq-2\,\bigg{(}\frac{m_{d_{P}}}{m_{d_{P}}\\!+\\!m_{u_{P}}}\bigg{)}\frac{m^{2}_{P}}{M^{2}_{H^{+}}}\frac{t_{\beta}^{3}}{(1+\epsilon t_{\beta})}(\epsilon_{e}-\epsilon_{\mu})\,,$ (15) where we have assumed that $m^{2}_{P}/M^{2}_{H^{+}}\ll 1$ and $\epsilon_{\ell}t_{\beta}\ll 1$. We observe that the NP sensitivity to the above effects of $K\to\ell\nu$ is higher than that of $\pi\to\ell\nu$ by a factor of $\Delta r_{K}^{e/\mu}/\Delta r_{\pi}^{e/\mu}\sim m^{2}_{K}/m^{2}_{\pi}$. The current experimental resolutions on these modes imply that $K\to\ell\nu$ is the best probe of the above scenario. Moreover, LFC contributions to $R_{\pi}$ and $R_{K}$ can be also induced at the loop level by box, wave function renormalization and vertex contributions from SUSY particle exchange [4]. The complete calculation of the $\mu$ decay in the MSSM [11, 12] can be easily applied to the meson decays. The dominant contributions to $\Delta r^{e-\mu}_{SUSY}$ arise from the charginos/neutralinos sleptons ($\tilde{l}_{e,\mu}$) exchange and it has the form [4] $\Delta r^{e-\mu}_{SUSY}\sim\frac{\alpha_{2}}{4\pi}\left(\frac{\tilde{m}^{2}_{\mu}-\tilde{m}^{2}_{e}}{\tilde{m}^{2}_{\mu}+\tilde{m}^{2}_{e}}\right)\frac{m^{2}_{W}}{M^{2}_{SUSY}},$ (16) thus, $\Delta r^{e-\mu}_{SUSY}$ can be of order $\Delta r^{e-\mu}_{SUSY}\leq 10^{-3}$, provided there exists a large mass splitting among sleptons ($\tilde{m}^{2}_{\mu}\ll\tilde{m}^{2}_{e}$ or $\tilde{m}^{2}_{\mu}\gg\tilde{m}^{2}_{e}$) and a SUSY mass scale $M_{SUSY}$ not much above the EW scale, i.e. $M_{SUSY}\sim m_{W}$. So, it turns out that all these LFC contributions yield values of $\Delta r^{e-\mu}_{\\!K\,SUSY}$ which are smaller than the current and expected future experimental sensitivities in kaon physics. On the other hand, given that the NP sensitivity to the above effects of $\Delta r_{K}^{e/\mu}$ and $\Delta r_{\pi}^{e/\mu}$ is the same and since the experimental resolution is better in the pion system, for this flavor conserving SUSY contribution it is the decay $\pi\to\ell\nu$ to represent the best place where to look for LFU violation. In particular, $R^{exp.}_{\pi}=(1.230\pm 0.004)\cdot 10^{-4}\,\,\,\,\,\,\,\rm{PDG}$ (17) and by making a comparison with the SM prediction $R^{SM}_{\pi}=(1.2354\pm 0.0002)\cdot 10^{-4}$ (18) one obtains (at the $2\sigma$ level) $-0.0107\leq\Delta r^{e-\mu}_{\\!NP}\leq 0.0022.$ (19) Comparing this interval for $\Delta r^{e-\mu}_{\\!NP}$ with the above value of $\Delta r^{e-\mu}_{SUSY}$, it turns out that only under rather particular conditions (very large mass splitting of sleptons of different generation, relatively light SUSY scale) can one obtain visible LFC SUSY contributions to the LFU violation in pion decays [13]. ## 5 Constraints In this section, we list the constraints we have imposed on the SUSY parameter space before starting the analysis of the LU breaking effects. ### 5.1 Direct SUSY search The framework in which we work is a low-energy R-parity conserving susy model with generic LFV soft breaking terms. We perform a scan up to a mass scale of $5\rm{TeV}$ of the following low energy parameters: the gaugino masses $M_{i}$ ($i=1,3$), the $\mu$ term, the left-left and right-right sfermion mass terms for the first two and the third generations $M_{\tilde{f}}$, the trilinear coupling in the stop sector $A_{t}$; moreover $\tan\beta<60$. At the low scale, we impose the following constraints on each point: * • Lower bound on the light and pseudo–scalar Higgs masses [14]; * • The LEP constraints on the lightest chargino and sfermion masses [15]; * • The LEP and Tevatron constrains on squarks, gluino and charged Higgs masses [15] * • Absence of charge and/or colour breaking minima [16]. * • The lightest susy particle (LSP) is neutral. * • Electroweak Precision Observables (EWPO) constraints [17]. For future convenience, it is useful to recall which are the necessary conditions under which the lightest Higgs mass bounds are satisfied. First of all, we remind that the LEPII bound $(m_{h})_{\rm SM}\,\raisebox{-3.44444pt}{$\stackrel{{\scriptstyle\textstyle>}}{{\sim}}$ }114$ GeV, applies also to SUSY theories, irrespective of the $\tan\beta$ values, provided we assume the decoupling regime, roughly implying that $M_{H,A}\geq 200GeV$. Indeed, this will represent the case under study in the present work. Within the MSSM, the lightest Higgs mass is bounded from above. In particular, we can write $m_{h}^{2}=m_{h}^{2({\rm tree})}+m_{h}^{2({\rm loop})}$ where, for large $\tan\beta$, $m_{h}^{2({\rm tree})}\sim m_{Z}^{2}-4m_{Z}^{2}m_{A}^{2}/(m_{A}^{2}-m_{Z}^{2})\cot^{2}\beta$. The most significant loop contribution is given by $\displaystyle m_{h}^{2({\rm loop})}$ $\displaystyle=$ $\displaystyle\frac{3m_{t}^{4}}{4\pi^{2}v^{2}}\left[\ln\left(\frac{m_{\tilde{t}_{1}}m_{\tilde{t}_{2}}}{m_{t}^{2}}\right)+\frac{\|X_{t}\|^{2}}{m_{\tilde{t}_{1}}^{2}-m_{\tilde{t}_{2}}^{2}}\ln\left(\frac{m_{\tilde{t}_{1}}^{2}}{m_{\tilde{t}_{2}}^{2}}\right)\right.$ (20) $\displaystyle\left.+\frac{1}{2}\left(\frac{\|X_{t}\|^{2}}{m_{\tilde{t}_{1}}^{2}-m_{\tilde{t}_{2}}^{2}}\right)^{2}\left(2-\frac{m_{\tilde{t}_{1}}^{2}+m_{\tilde{t}_{2}}^{2}}{m_{\tilde{t}_{1}}^{2}-m_{\tilde{t}_{2}}^{2}}\ln\left(\frac{m_{\tilde{t}_{1}}^{2}}{m_{\tilde{t}_{2}}^{2}}\right)\right)\right],$ where $X_{t}=A_{t}-\mu^{}\cot\beta$. Thus, the tree level contribution, that is maximum for moderate to large $\tan\beta$, has to be supplemented by sizable loop corrections. In particular, if the stop mixing is small, $\|X_{t}/m_{\tilde{t}_{1,2}}\|^{2}\ll 1$, the correction depends only on the logarithm of the stop masses, so these must be rather heavy. If, however, the stop mixing is large, much lighter stops can still yield large loop corrections. However, as we will see, this last possibility is disfavored by the $b\to s\gamma$ constraints, specially in the large $\tan\beta$ regime. ### 5.2 B-physics observables #### 5.2.1 $\mathcal{B}(B\rightarrow X_{s}\gamma)$ As it is well known, $\mathcal{B}(B\rightarrow X_{s}\gamma)$ is a particularly sensitive observable to possible non-standard contributions and it provides a non-trivial constraint on the SUSY mass spectrum given its precise experimental determination and the very accurate SM calculation at the NNLO [18]. According to the recent NNLO analysis of Ref. [18], the SM prediction is ${\cal B}(B\to X_{s}\gamma;E_{\gamma}>1.6~{}{\rm GeV})^{\rm SM}=(3.15\pm 0.23)\times 10^{-4}$. Combining this result with the experimental average [19, 20, 21] ${\cal B}(B\to X_{s}\gamma;E_{\gamma}>1.6~{}{\rm GeV}))^{\rm exp}=(3.55\pm 0.24)\times 10^{-4}$ we obtain $\displaystyle R_{Bs\gamma}=\frac{{\cal B}^{\rm exp}(B\to X_{s}\gamma)}{{\cal B}^{\rm SM}(B\to X_{s}\gamma)}=1.13\pm 0.12\,.$ (21) In our numerical analysis we impose the above constraint at the $2\sigma$ C.L. Within Minimal Flavor Violating (MFV) frameworks [22], the dominant SUSY contributions to $\mathcal{B}(B\rightarrow X_{s}\gamma)$ arise from the one- loop charged-Higgs and chargino-squark amplitudes. Charged-Higgs effects unambiguously increase the rate compared to the SM expectation, while the chargino-squark ones can have both signs depending on the sign of $sign(\mu A_{\tilde{t}})$. In this work we choose $\mu>0$ (that is also preferred by the $(g-2)_{\mu}$ constraints) and $sign(A_{\tilde{t}})<0$, which implies destructive interference between chargino and charged Higgs contributions. A simple expression accounting for NP contributions in $B\to X_{s}\gamma$ is provided by [18, 23] $R_{Bs\gamma}\simeq 1-2.54\,C^{NP}_{7}(M_{W})-0.60\,C^{NP}_{8}(M_{W})$ (22) where $C^{NP}_{7,8}\\!=\\!C^{\tilde{\chi}^{\pm}}_{7,8}+C^{H^{\pm}}_{7,8}$ are the relevant Wilson coefficients for the New Physics contributions evaluated at the $M_{W}$ scale. In particular, starting from the full expressions for $C^{NP}_{7,8}$of Ref. [24], one can derive the following approximate expressions $\displaystyle C^{H^{\pm}}_{7}$ $\displaystyle\simeq$ $\displaystyle\left(\frac{1-\epsilon\,t_{\beta}}{1+\epsilon\,t_{\beta}}\right)\,\frac{m^{2}_{t}}{M^{2}_{H^{\pm}}}\,F^{7}_{H^{\pm}}\left(\frac{m^{2}_{t}}{M^{2}_{H^{\pm}}}\right)\,,$ $\displaystyle C^{\tilde{\chi}^{\pm}}_{7}$ $\displaystyle\simeq$ $\displaystyle-\frac{A_{\tilde{t}}}{\mu}\,\frac{m^{2}_{t}}{\mu^{2}}\,\frac{t_{\beta}}{1+\epsilon\,t_{\beta}}\,F^{7}_{\tilde{\chi}^{\pm}}\left(\frac{m_{\tilde{q}}^{2}}{\mu^{2}}\right)\,,$ (23) where $\epsilon\sim 10^{-2}$ for a degenerate SUSY spectrum and $F^{7}_{\tilde{\chi}^{\pm}}(1)\simeq 0.07$, $F^{7}_{\tilde{\chi}^{\pm}}(x\gg 1)\simeq(13/12-1/2\log(x))/x^{2}$ and $F^{7}_{\tilde{\chi}^{\pm}}(x\ll 1)\simeq 7/12+2/3\log(x)$ while $F^{7}_{H^{\pm}}(x_{tH})\simeq 1/4+1/3\log(x_{tH})$ for $x_{tH}=m^{2}_{t}/M^{2}_{H^{\pm}}\ll 1$ and $F^{7}_{H^{\pm}}(1)\simeq-0.2$. We observe that, when $m_{\tilde{q}}/\mu\ll 1$, the lower bound on $m_{\tilde{q}}$ is set by the experimental limits on the lightest stop mass $m^{2}_{\tilde{t}_{1}}\simeq m^{2}_{\tilde{q}}-m_{t}\|A_{\tilde{t}}\|$ and on the sbottom mass $m^{2}_{\tilde{b}_{1}}\simeq m^{2}_{\tilde{q}}-m_{b}\|\mu\|t_{\beta}$. Similar expressions for the subleading contributions proportional to $C^{NP}_{8}$ are not shown, although included in our numerical analysis. Taking $C^{H^{\pm}}_{7}$ and $C^{\tilde{\chi}^{\pm}}_{7}$ separately, the following observations follow: i) the lower bound on $m_{H^{\pm}}\geq 295\rm{GeV}$, holding at the $2\sigma$ level within a 2HDM framework ( where it is assumed $\epsilon=0$ and where $C^{\tilde{\chi}^{\pm}}_{7}=0$), can be significantly relaxed within SUSY scenarios thanks to a reduction of $C^{H^{\pm}}_{7}$ driven by the threshold corrections $\epsilon$; in particular if $\tan\beta\sim 50$ it turns out that $m_{H^{\pm}}\geq 200\rm{GeV}$ ii) for a natural scenario where all the SUSY masses have comparable size, in particular for $A_{\tilde{t}}/(\mu,m_{\tilde{q}})\sim 1$, the regime $\tan\beta\sim 50$ necessarily implies that $\mu$ and/or $m_{\tilde{q}}$ lie in the $\geq 1\rm{TeV}$ scale iii) the simultaneous requirement of large values for $\tan\beta$ and relatively light $m_{\tilde{q}},\mu$ ( below the $1\rm{TeV}$ scale) necessarily implies either large cancellations between $C^{H^{\pm}}_{7}$ and $C^{\tilde{\chi}^{\pm}}_{7}$ and/or $A_{\tilde{t}}/(\mu,m_{\tilde{q}})$ significantly less than 1. However, as we have seen before, the scenario with relatively light $m_{\tilde{q}}$ and $A_{\tilde{t}}/m_{\tilde{q}}\ll 1$ is constrained by the lower bound on the lightest Higgs mass $m_{h}$. #### 5.2.2 $B\to\tau\nu$ The recent Belle[25] and BaBar[26] results for $B\to\ell\nu$ leads to the average ${\cal B}({B\to\tau\nu})^{\rm exp}=(1.42\pm 0.43)\times 10^{-4}~{}$. This should be compared with the SM expectation ${\cal B}({B\to\tau\nu})^{\rm SM}=G_{F}^{2}m_{B}m_{\tau}^{2}f_{B}^{2}\|V_{ub}\|^{2}(1-m_{\tau}^{2}/m_{B}^{2})^{2}\tau_{B}/8\pi$, whose numerical value suffers from sizable parametrical uncertainties induced by $f_{B}$ and $V_{ub}$. Taking $\tau_{B}=(1.643\pm 0.010)\rm{ps}$, $V_{ub}=(4.00\pm 0.26)\times 10^{-3}$ and $f_{B}=0.216\pm 0.022\rm{GeV}$ [27], the best estimate is ${\cal B}({B\to\tau\nu})^{\rm SM}=(1.33\pm 0.23)\times 10^{-4}~{}$, which implies $\displaystyle R_{B\tau\nu}^{\rm exp}=\frac{\mathcal{B}^{\rm exp}({B\to\tau\nu})}{\mathcal{B}^{\rm SM}({B\to\tau\nu})}~{}=~{}1.07\pm 0.37~{}.$ (24) From the theoretical side, the $B\to\ell\nu$ process is one of the cleanest probes of the large $\tan\beta$ scenario due to its enhanced sensitivity to tree-level charged-Higgs exchange [28, 29]. In particular, a scalar charged current induced by NP theories with extended Higgs sectors, provides the following effects: $R_{B\ell\nu}=\bigg{[}1-\frac{m^{2}_{B}}{M^{2}_{H^{+}}}\frac{t_{\beta}^{2}}{(1+\epsilon t_{\beta})(1+\epsilon_{\ell}t_{\beta})}\bigg{]}^{2}$ (25) where we have included corrections both for the quark and lepton yukawas arising within SUSY theories. The new physics effect on $R_{K\mu\nu}=\Gamma^{\rm SUSY}(K\to\mu\nu)/\Gamma^{\rm SM}(K\to\mu\nu)$ is obtained from Eq. 25 with the replacement $m_{B}^{2}\to m_{K}^{2}$ [29]. Although the charged Higgs contributions are now suppressed by a factor $m_{K}^{2}/m_{B}^{2}\simeq 1/100$, this is well compensated by the excellent experimental resolution [3] and the good theoretical control. However, given that these new physics effects are, in the most favorable cases, at the % level, we would need a theoretical prediction for the SM contribution at the same level to use this decay as an effective constraint. We would then need an independent determination both of $f_{K}$ (possibly from lattice QCD) and $V_{us}$ with such a level of accuracy. The best strategy to fully exploit the New Physics sensitivity of $K_{l2}$ systems is to consider the ratio $R^{{}^{\prime}}=R_{K\mu\nu}/R_{\pi\mu\nu}$ [29, 3]. In fact, on the one side $R^{{}^{\prime}}$ and $R_{K\mu\nu}$ have the same New Physics content (being $R_{\pi\mu\nu}$ not sizably affected by charged current interactions) and, on the other side, $R^{{}^{\prime}}$ depends on $(f_{K}/f_{\pi})^{2}$ instead of $f_{K}^{2}$ with $f_{K}/f_{\pi}$ being much better under control compared to $f_{K}$ by means of lattice QCD. However, at present, unquenched lattice calculations of $f_{K}/f_{\pi}$ are still not well established. Therefore, although it may play a relevant role in the future, we do not include the constraints from $K\to l\nu$ in the present analysis. The above argument for $K\to l\nu$ does not apply to $B\to\ell\nu$. In fact, even if the hadronic uncertainties related to $f_{B}$ and $V_{ub}$ are much larger that those for $f_{K}$ and $V_{us}$, they cannot hide in any way the huge NP effects in $B\to\ell\nu$ that can arise in our scenario. #### 5.2.3 $B_{s}\rightarrow\mu^{+}\mu^{-}$ The SM prediction for ${\mathcal{B}}(B_{s}\rightarrow\mu^{+}\mu^{-})$ is ${\mathcal{B}}(B_{s}\rightarrow\mu^{+}\mu^{-})_{{\rm SM}}=(3.37\pm 0.31)\times 10^{-9}$. This value should be compared to the present 95% C.L. upper bound from CDF, ${\rm BR}(B_{s}\rightarrow\mu^{+}\mu^{-})_{\rm exp}<5.8\times 10^{-8}$ that still leaves a large room for NP contributions. In particular, the MSSM with large $\tan\beta$ allows, in a natural way, large differences between SM and SUSY expectations for ${\mathcal{B}}(B_{s}\rightarrow\mu^{+}\mu^{-})$ [30]. The SUSY contributions can be summarized by the approximate formula ${\rm BR}(B_{s}\to\mu^{+}\mu^{-})\simeq\frac{5\times 10^{-8}}{\left[1+0.5\times\frac{\tan\beta}{50}\right]^{4}}\Bigg{[}\frac{\tan\beta}{50}\Bigg{]}^{6}\left(\frac{500\rm{GeV}}{M_{A}}\right)^{4}\left(\frac{\epsilon_{Y}}{3\times 10^{-3}}\right)^{2}$ (26) where $\epsilon_{Y}\simeq-1/16\pi^{2}\times A_{t}/\mu\times H_{2}(y_{u_{R}},y_{u_{L}})$ with $y_{q_{R,L}}=M^{2}_{\tilde{q}_{L,R}}/\|\mu\|^{2}$, $H_{2}(1,1)=-1/2$, $H_{2}(x\gg 1,y=x)\simeq-1/x$ and $H_{2}(x\ll 1,y=x)\simeq 1+\log x$; thus, $\epsilon^{\tilde{\chi}^{-}}_{Y}\sim 3\times 10^{-3}$ holds in the limit of all the SUSY masses and $A_{t}$ equal. As we can see, the present CDF upper bound on ${\mathcal{B}}(B_{s}\to\mu^{+}\mu^{-})$ already provides constraints in some regions of the SUSY parameter space. Moreover, we remind that, although also $\Delta M_{s}$ is a New Physics sensitive observable in the scenario we are considering, it doesn’t provide any further constraints in addition to those inferred by the B-physics observables that we have already discussed. ### 5.3 $(g-2)_{\mu}$ The possibility that the anomalous magnetic moment of the muon [$a_{\mu}=(g-2)_{\mu}/2$], which has been measured very precisely in the last few years [31], provides a first hint of physics beyond the SM has been widely discussed in the recent literature. Despite substantial progress both on the experimental and on the theoretical sides, the situation is not completely clear yet (see Ref. [32] for an updated discussion). Most recent analyses converge towards a $3\sigma$ discrepancy in the $10^{-9}$ range [32]: $\Delta a_{\mu}=a_{\mu}^{\rm exp}-a_{\mu}^{\rm SM}\approx(3\pm 1)\times 10^{-9}~{}.$ (27) Recently, Passera et al. [33] have considered the possibility that the present discrepancy between experiment and the Standard Model (SM) prediction for $(g-2)_{\mu}$ may arise from errors in the determination of the hadronic leading-order contribution to the latter. If this is the case, the authors of Ref. [33] find a decrease on the electroweak upper bound on the SM Higgs boson mass. By means of a detailed analysis they conclude that this solution of the muon $(g-2)_{\mu}$ discrepancy is unlikely in view of current experimental error estimates. The main SUSY contribution to $a^{\rm MSSM}_{\mu}$ is usually provided by the loop exchange of charginos and sneutrinos. The basic features of the supersymmetric contribution to $a_{\mu}$ are correctly reproduced by the following approximate expression: $\frac{a^{\rm MSSM}_{\mu}}{1\times 10^{-9}}\approx 1.5\left(\frac{\tan\beta}{10}\right)\left(\frac{300~{}\rm GeV}{m_{\tilde{\nu}}}\right)^{2}\left(\frac{\mu M_{2}}{m^{2}_{\tilde{\nu}}}\right)~{},$ (28) which provides a good approximation to the full one-loop result [34]. The most relevant feature of Eqs. (28) is that the sign of $a^{\rm MSSM}_{\mu}$ is fixed by the sign of the $\mu$ term so that the $\mu>0$ region is strongly favored. ## 6 LFU vs LFV and the $(g-2)_{\mu}$ anomaly As we have previously seen, sizable LFU breaking effects can be generated in SUSY through LFV interactions which involve the third generation. Hence, a legitimate worry is whether the bounds on LFV tau decays, like $\tau\rightarrow eX$ (with $X=\gamma,\eta,\mu\mu$), are respected in the region of SUSY parameter space leading to a strong enhancement of the LFU violation [8]. The present and projected bounds (to be achieved at a super B factory) on some of these processes are summarized in Table 1. \| Process \| Present Bounds \| Expected Future Bounds ---\|---\|---\|--- (1) \| BR($\tau\to e,\gamma$) \| $9.4~{}\times~{}10^{-8}$ \| $\mathcal{O}(10^{-8})$ (2) \| BR($\tau\to e,e,e$) \| $2.0~{}\times~{}10^{-7}$ \| $\mathcal{O}(10^{-8})$ (3) \| BR($\tau\to e,\mu,\mu$) \| $2.0~{}\times~{}10^{-7}$ \| $\mathcal{O}(10^{-8})$ (4) \| BR($\tau\to e,\eta$) \| $4.5~{}\times~{}10^{-8}$ \| $\mathcal{O}(10^{-8})$ Table 1: Present and Upcoming experimental limits on various leptonic processes at 90% C.L. The most sensitive probe of Higgs mediated effects is generally provided by $\tau\rightarrow\ell_{j}\eta$ [35]; the corresponding branching ratio is given by [36, 37] $\frac{Br(\tau\rightarrow l_{j}\eta)}{Br(\tau\rightarrow l_{j}\bar{\nu_{j}}\nu_{\tau})}\simeq 18\pi^{2}\\!\left(\frac{f^{8}_{\eta}m^{2}_{\eta}}{m_{\tau}}\right)^{\\!2}\\!\\!\left(1\\!-\\!\frac{m^{2}_{\eta}}{m^{2}_{\tau}}\right)^{\\!2}\left(\frac{\|\Delta^{3j}\|^{2}t_{\beta}^{6}}{m^{4}_{A}}\right)$ (29) where $m^{2}_{\eta}/m^{2}_{\tau}\simeq 9.5\times 10^{-2}$ and the relevant decay constant is $f^{8}_{\eta}\sim 110{\rm MeV}$. Moreover, $\|\Delta^{3j}\|^{2}\\!=\\!\|\Delta^{3j}_{L}\|^{2}+\|\Delta^{3j}_{R}\|^{2}$, where $\Delta^{3j}_{L}$ has a similar expression to $\Delta^{3j}_{R}$ [36] and it is such that $\Delta^{3j}_{L}\sim\delta^{3j}_{LL}$. We note that, in order to generate a non-vanishing $\Delta r^{e-\mu}_{\\!K\,Susy}$, RR-type flavor structures are unavoidable; on the contrary, $Br(\tau\rightarrow e\eta)$ can be generated by both LL and/or RR-type mixing angles, being $Br(\tau\rightarrow e\eta)\sim\|\Delta^{31}_{L}\|^{2}+\|\Delta^{31}_{R}\|^{2}$. Given that $\Delta r^{e-\mu}_{\\!K\,Susy}$ and $Br(\tau\rightarrow e\eta)$ have the same SUSY dependence, the upper bound on $Br(\tau\rightarrow eX)$ is automatically found once we saturate the allowed range (at the % level) for New Physics contributions in $\Delta r^{e-\mu}_{\\!K\,Susy}$. We find that $Br(\tau\rightarrow e\eta)\simeq 10^{-2}\left(\frac{\|\Delta^{31}\|^{2}t_{\beta}^{6}}{m^{4}_{A}}\right)\simeq 10^{-8}\times\Delta r^{e-\mu}_{\\!K\,Susy}\,,$ (30) where the last equality holds when $\Delta^{31}_{L}=0$. So, employing the constraints for $\Delta r^{e-\mu}_{\\!K\,Susy}$ at the $\%$ level, we obtain $Br(\tau\rightarrow e\eta)\leq 10^{-10}$. We conclude that, the present and expected experimental upper bounds on $Br(\tau\rightarrow e\eta)$ (see Table 1) still allow large effects in $\Delta r^{e-\mu}_{\\!K\,Susy}$. On the other hand, $\tau\rightarrow\ell_{j}\gamma$ is the most sensitive probe of LFU violation induced by SUSY gauge mediated effects. In contrast to the $Br(\tau\rightarrow e\eta)$ case, it is not possible to link $Br(\tau\rightarrow e\gamma)$ and $\Delta r^{e-\mu}_{\\!K\,Susy}$ in a way that is independent of the specific choice for the susy breaking sector. In particular, as discussed before, the New Physics contributions to $\Delta r^{e-\mu}_{\\!K\,Susy}$ decouple with the heavy Higgs mass $m_{H}$, while $Br(\tau\rightarrow e\gamma)$ decouples with the heaviest SUSY particle mass $\tilde{m}$ circulating in the gaugino/slepton loop. In the following, to get a feeling of where we stand, we will evaluate $Br(\tau\rightarrow e\gamma)$ in the region of the parameter space where large LFU breaking effects in $\Delta r^{e-\mu}_{\\!K\,Susy}$ can be generated. In particular, a necessary ingredient in order to get large $\Delta r^{e-\mu}_{\\!K\,Susy}$ values is to maximize the size of the effective LFV coupling $\Delta^{31}_{R}$ (remember that $\Delta r^{e-\mu}_{\\!K\,Susy}\sim t^{6}_{\beta}/M^{4}_{H}\times\Delta^{31}_{R}$) and this happens when $\mu\gg\tilde{m}$. In this latter case, starting from the full expressions of Ref. [38], we find the following expression $\frac{BR(\tau\rightarrow\ell_{j}\gamma)}{BR(\tau\rightarrow\ell_{j}\nu_{\tau}\bar{\nu_{j}})}\simeq\frac{\pi\alpha_{el}}{3G_{F}^{2}}\,\left(\frac{\alpha_{Y}}{4}\right)^{2}\left(\left\|\delta_{LL}^{3j}\right\|^{2}+\left\|\delta_{RR}^{3j}\right\|^{2}\right)\,\frac{\mu^{2}}{\tilde{m}^{2}}\,\frac{t^{2}_{\beta}}{\tilde{m}^{4}}\,.$ (31) From Eq. 31 we can get $BR(\tau\rightarrow\ell_{j}\gamma)\approx 5\times 10^{-8}\left(\,\left\|\frac{\delta_{RR}^{3j}}{0.5}\right\|^{2}+\left\|\frac{\delta_{LL}^{3j}}{0.5}\right\|^{2}\,\right)\,\left(\frac{t_{\beta}}{50}\right)^{2}\,\left(\frac{1\,\rm{TeV}}{\tilde{m}}\right)^{4}\,\frac{\mu^{2}}{\tilde{m}^{2}}\,,$ (32) showing that large (order one) mixing angles for $\delta_{LL,RR}^{3j}$ are phenomenologically allowed, provided there exists a rather heavy spectrum for the soft sector. For instance, for $\mu/\tilde{m}=4$, it turns out that $\tilde{m}\geq 2\rm{TeV}$. Obviously, such a lower bound on $\tilde{m}$ can be relaxed for smaller values of $\delta_{LL,RR}^{3j}$ and/or $\tan\beta$. Moreover, one can easily find the following approximate expression $\Delta r^{e-\mu}_{\\!K\,Susy}\leq 10^{-1}\times\frac{BR(\tau\rightarrow e\gamma)}{10^{-7}}\left(\frac{\tilde{m}/M_{H}}{4}\right)^{4}\left(\frac{t_{\beta}}{50}\right)^{4}\,,$ (33) showing that, in the large $\tan\beta$ regime and for heavy Higgs masses lighter than those of the soft breaking terms, experimentally visible LFU breaking effects in $K\to\ell\nu$ can be naturally obtained. In particular, large LFU breaking effects even above the $10\%$ level (already excluded experimentally), can be always compatible with the experimental constraints on $BR(\tau\rightarrow e\gamma)$ for slepton/gaugino masses at the $\rm{TeV}$ scale. However, we stress again that it is not possible to correlate $BR(\tau\to e\gamma)$ to $\Delta r^{e-\mu}_{\\!K\,Susy}$, unless specific SUSY breaking mechanisms (relating $\tilde{m}$ and $M_{H}$) are assumed. On the contrary, the processes $\ell_{i}\rightarrow\ell_{j}\gamma$ are intimately linked to the muon anomalous magnetic moment $(g-2)_{\mu}$, as they both arise from dipole transitions [39]. Thus, in the following, we address the interesting question of whether it is possible, within SUSY theories, to account for the current $(g-2)_{\mu}$ anomaly, while generating, at the same time, LFU breaking effects in $\Delta r^{e-\mu}_{\\!K\,Susy}$ at the $\%$ level. As we will show, the answer is positive and this will lead to set a lower bound on $BR(\tau\rightarrow e\gamma)$. To see this point explicitly, let us derive the correlation between $BR(\ell_{i}\rightarrow\ell_{j}\gamma)$ and $(g-2)_{\mu}$ for the relevant case where $\mu/\tilde{m}\gg 1$; in this case, $\Delta a_{\mu}$ is well approximated by the expression $\displaystyle\Delta a_{\mu}$ $\displaystyle\simeq$ $\displaystyle\frac{\alpha_{Y}}{24\pi}\,\frac{\mu}{\tilde{m}}\,\frac{m_{\mu}^{2}}{\tilde{m}^{2}}\,t_{\beta}$ (34) $\displaystyle\simeq$ $\displaystyle 3\times 10^{-9}\,\left(\frac{\mu/\tilde{m}}{5}\right)\,\left(\frac{400\rm{GeV}}{\tilde{m}}\right)^{2}\,\left(\frac{t_{\beta}}{50}\right)$ and thus we find that $\displaystyle BR(\tau\rightarrow\ell_{j}\gamma)$ $\displaystyle\simeq$ $\displaystyle\frac{12\pi^{3}}{m^{4}_{\mu}}\,\left(\frac{\alpha_{el}}{G^{2}_{f}}\right)\left(\Delta a_{\mu}\right)^{2}\left(\left\|\delta_{RR}^{3j}\right\|^{2}+\left\|\delta_{LL}^{3j}\right\|^{2}\right)BR(\tau\rightarrow\ell_{j}\nu_{\tau}\bar{\nu}_{j})$ (35) $\displaystyle\simeq$ $\displaystyle 3\times 10^{-9}\,\left(\frac{\Delta a_{\mu}}{1\times 10^{-9}}\right)^{2}\,\left(\,\left\|\frac{\delta_{RR}^{3j}}{0.01}\right\|^{2}+\left\|\frac{\delta_{LL}^{3j}}{0.01}\right\|^{2}\,\right)\,.$ From Eqs. 33, 34, 35 we conclude that, LFU breaking effects at the $\%$ level, typically implying $\delta_{RR}^{31}\geq 0.01$, are compatible with the current experimental bounds on $BR(\tau\rightarrow e\gamma)$; moreover, if we additionally require that SUSY effects explain the current discrepancy for the muon anomalous magnetic moment, i.e. $\Delta a_{\mu}\geq 1\times 10^{-9}$ at the $2\sigma$ level [32], LFU breaking effects at the $\%$ level unavoidably imply large effects in $BR(\tau\rightarrow e\gamma)\geq 3\times 10^{-9}$, well within the expected reach of a superB factory. As we have seen, sizable LFU breaking effects originating from LFV interactions require a flavor mixing in the $13$ sector. Thus, from a phenomenological point of view, $\Delta r^{e-\mu}_{\\!K\,Susy}$ is naturally related to $\tau-e$ transitions as, for instance, $\tau\rightarrow e\eta$, $\tau\rightarrow e\gamma$ etc. However, a legitimate question that can be addressed is what one would expect for $\tau-\mu$ and $\mu-e$ transitions when sizable sources of LFV in the $\tau-e$ sector are assumed. In particular, from a model building point of view, it seems hard to generate large effects for $\tau-e$ transitions while keeping the effects for $\tau-\mu$ transitions small. Moreover, once $\tau-\mu$ transitions are induced, an effective $(\mu-e)_{eff.}$ transition of the type $(\mu-e)_{eff.}=(\mu-\tau)\times(\tau-e)$ is also induced and processes like $\mu\to e\gamma$ are unavoidable. In the following, we will address the above issue more quantitatively. In particular, the analogue expression of Eq. 31 for the $\mu\rightarrow e\gamma$ case reads $\displaystyle\frac{BR(\mu\rightarrow e\gamma)}{BR(\mu\rightarrow e\nu_{\mu}\bar{\nu_{e}})}$ $\displaystyle=$ $\displaystyle\frac{\pi\alpha_{el}}{3G_{F}^{2}}\,\frac{t^{2}_{\beta}}{\tilde{m}^{4}}\left(\frac{\alpha_{Y}}{10}\right)^{2}\left\|\delta_{RR}^{23}\delta_{RR}^{31}+\frac{m_{\tau}}{m_{\mu}}\delta_{LL}^{23}\delta_{RR}^{31}\right\|^{2}\left(\frac{\mu}{\tilde{m}}\right)^{2}\,,$ (36) where, besides the combination of LFV sources relevant for the present discussion, i.e. $RR$-type LFV sources, we have also kept the $\delta_{LL}^{23}\delta_{RR}^{31}$ contribution. In fact, this last contribution is enhanced, at the amplitude level, by the ratio $m_{\tau}/m_{\mu}$ compared to the $\delta_{RR}^{23}\delta_{RR}^{31}$ contribution and thus, potentially large even when $\delta_{LL}^{23}\\!<\\!\delta_{RR}^{32}$. Finally, it turns out that $BR(\mu\rightarrow e\gamma)\simeq 10^{-11}\left\|\frac{\delta_{RR}^{23}\delta_{RR}^{31}}{10^{-2}}+\frac{m_{\tau}}{m_{\mu}}\frac{\delta_{LL}^{23}\delta_{RR}^{31}}{10^{-2}}\right\|^{2}\,\left(\frac{t_{\beta}}{50}\right)^{2}\,\left(\frac{1\,\rm{TeV}}{\tilde{m}}\right)^{4}\left(\frac{\mu}{\tilde{m}}\right)^{2}\,.$ (37) Eq. 37 shows that it is not possible ( in the large $\tan\beta$ regime, at least) to have simultaneously order one MIs $\delta_{RR}^{23}$ and $\delta_{RR}^{31}$, unless we push $\tilde{m}$ in the multi-TeV regime. In particular, if $\delta_{RR}^{23}=\delta_{RR}^{31}=0.1$, $\delta_{LL}^{23}=0$ and $\mu/\tilde{m}=4$, it turns out that $\tilde{m}\geq 2\rm{TeV}$. Clearly, such a scenario is not compatible with an explanation of the $(g-2)_{\mu}$ anomaly. Moreover, as we have discussed in the previous sections, there is also the possibility to obtain negative values for $\Delta r^{e-\mu}_{\\!K\,Susy}$ (see Eq. 12) when both RR and LL-type of flavor violating sources for the $1-3$ transition are present. However, this possibility can be constrained, in some cases, by the experimental upper bounds on $BR(\mu\rightarrow e\gamma)$. In fact, combining Eq. 12 with Eq. 36, it turns out that $\|\Delta r^{e-\mu}_{\\!K\,Susy}\|\leq 3\times 10^{-3}\times\sqrt{\frac{BR(\mu\rightarrow e\gamma)}{10^{-11}}}\left(\frac{\tilde{m}/M_{H}}{10}\right)^{2}\left(\frac{t_{\beta}}{50}\right)^{2}\left\|\frac{\delta^{31}_{LL}}{\delta^{32}_{LL}}\right\|\,,$ (38) thus, unless we assume $\delta^{32}_{LL}/\delta^{31}_{LL}\\!<<\\!1$ (that is typically unnatural from a model building point of view), we are lead with large effects in $BR(\mu\to e\gamma)$ even for an heavy soft sector at the $\rm{TeV}$ scale. However, we stress that from a pure phenomenological perspective, $BR(\mu\rightarrow e\gamma)$ doesn’t impose a direct bound on $\Delta r^{e-\mu}_{\\!K\,Susy}$ given that different parameters enter the two quantities. Let us finally point out that, when $\delta_{RR}>>\delta_{LL}$, the following upper bound on $\Delta r^{e-\mu}_{\\!K\,Susy}$ holds $\Delta r^{e-\mu}_{\\!K\,Susy}\leq\frac{Br(\tau\rightarrow\mu\eta)}{10^{-8}}\times\frac{BR(\tau\rightarrow e\gamma)}{BR(\tau\rightarrow\mu\gamma)}\,.$ (39) Clearly, only the discovery of LFV signals in some of the above observables, by means of improved experimental sensitivities, would shed light on the scenarios outlined before. We conclude this section pointing out that the same New Physics effect observable in the Kaon system through $\Delta r^{e-\mu}_{\\!K\,Susy}$ is also observable, in principle, in B physics systems by means of purely leptonic decays of charged $\rm{B}$ meson. In particular, it is found that $\frac{BR(B\to e\nu)}{BR(B\to e\nu)_{SM}}=\left[1+\frac{m^{4}_{B}}{m^{4}_{K}}\,\Delta r^{e-\mu}_{\\!K}\right]\simeq\left[1+10^{2}\times\left(\frac{\Delta r^{e-\mu}_{\\!K}}{10^{-2}}\,\right)\right]\,.$ (40) This means that, a LFU breaking effect at the $\%$ level in the $K\ell 2$ systems, implies an enhancement of two orders of magnitude in $BR(B\to e\nu)$ compared to its SM expectation. However, given that $BR(B\to e\nu)_{SM}\approx 10^{-11}$, an experimental sensitivity at the level of $BR(B\to e\nu)_{exp.}\leq 10^{-9}$ would be necessary. ## 7 Numerical Analysis In the following, we will analyze the allowed size for the LFU breaking effects in $R^{e/\mu}_{P}$ both in the lepton flavor conserving and violating cases. In the former case, LFU breaking effects arise from mass splittings between sleptons of the first and second families $(m_{L1(L2)}$, $m_{R1(R2)}$), as discussed in Sec. 4. In Fig. 2, we perform a numerical analysis of the allowed values for $\Delta r_{K}^{e/\mu}$ (see Eq.15) through a scan over the following SUSY parameter space: $(m_{L1(L2)}$, $m_{R1(R2)}$, $m_{\tilde{Q}}$, $m_{\tilde{g}}$, $m_{\tilde{W}}$, $m_{\tilde{B}}$, $M_{H})<2.5\rm{TeV}$, $\mu\\!<\\!5$TeV and $\tan\beta<60$. In particular, we allow different entries for the left-left (LL) and the right-right (RR) blocks in the slepton mass matrix for the first two generations, i.e. for $m_{L1(L2)}$ and $m_{R1(R2)}$ respectively. Moreover, we also impose all the constraints discussed in Sec. 5. In Fig 2, on the left, we show $\Delta r_{K}^{e/\mu}$ as a function of the (left-handed) mass splitting between the second and first slepton generations. Black dots refer to the points satisfying the $(g-2)_{\mu}$ discrepancy at the $95\%$ C.L., i.e. $1\times 10^{-9}<\Delta a_{\mu}<5\times 10^{-9}$. As we can see, the maximum LFU breaking effects are reached for maximum mass splitting between sleptons. However, when $m_{L1}\\!=\\!m_{L2}$, we would expect LFU breaking effects going to zero, in contrast to what is shown by Fig 2. This happens because mass splittings for right-handed sleptons $m_{R1}\\!\neq\\!m_{R2}$ (not explicitly visible in Fig 2), can still generate LFU breaking effects even in the case where $m_{L1}\\!=\\!m_{L2}$. We see that values for $\|\Delta r_{K}^{e/\mu}\|$ as large as $5\times 10^{-3}$ are possible for slepton masses splitted by a factor $10$. However, potentially visible values for $\|\Delta r_{K}^{e/\mu}\|$ of order of $\sim 2\times 10^{-3}$ are obtained even for smaller mass splittings, i.e. for $m_{L1,L2}/m_{L2,L1}\sim 2$. Interestingly enough, the sign of these LFU breaking effects depends on the ratio between the slepton masses. In particular, if the left-handed smuons are heavier then the selectrons $\Delta r_{K}^{e/\mu}>0$, while $\Delta r_{K}^{e/\mu}<0$ if the smuons are lighter then the selectrons. The opposite situation happens for mass splittings of right-handed smuons. In Fig 2, on the right, we show the regions of the parameter space in the $\tan\beta-M_{H}$ plane where $0.001\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.003$ (red dots), $0.003\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.005$ (black dots) and $\|\Delta r_{K}^{e/\mu}\|\\!>\\!0.005$ (yellow dots). We observe that the narrow region where $m_{H}\leq 200$GeV corresponds to the points where the $B\to\tau\nu$ constraints are not effective. This does not occur not because the new physics contributions to $B\to\tau\nu$ are small; quite the contrary, the reason is that they are quite large ($\sim 2\times$ SM ones) and destructively interfere with the SM contribution (see Eq. 25). On the other hand, the region of the $\tan\beta-M_{H}$ plane between the two allowed areas is excluded by the $B\to\tau\nu$ constraint. Let us now discuss LFU breaking effects in $\Delta r_{K}^{e/\mu}$ as generated by LFV contributions stemming, in particular, from RR-type flavor violating sources only. In Fig. 3, on the left, we report $\Delta r_{K}^{e/\mu}$ as a function of $\mathcal{B}(\tau\\!\to\\!e\gamma)$ and $\mathcal{B}(\tau\\!\to\\!e\eta)$ while, on the right, we report $\Delta r_{K}^{e/\mu}$ as a function of $M_{H}$. The plots have been obtained by means of a scan over the following parameter space: $(m_{L,R},m_{\tilde{Q}},m_{\tilde{g}},m_{\tilde{W}},m_{\tilde{B}},M_{H})<2.5\rm{TeV}$, $\mu\\!<\\!5$TeV, $\|\delta_{RR}\|\\!<\\!0.5$, $\|\delta_{LL}\|\\!=\\!0$ and $\tan\beta<60$ and imposing all the constraints discussed in Sec.5. Black dots refer to the points satisfying the $(g-2)_{\mu}$ anomaly at the $95\%$ C.L., i.e. $1\times 10^{-9}<\Delta a_{\mu}<5\times 10^{-9}$. Fig. 3 clearly shows that there are quite a lot of points in the interesting region where $0.001\\!<\\!\Delta r_{K}^{e/\mu}\\!<\\!0.01$ accounting for the $(g-2)_{\mu}$ anomaly and that are compatible with the experimental constraints of $\mathcal{B}(\tau\\!\to\\!e\gamma)$ and $\mathcal{B}(\tau\\!\to\\!e\eta)$. As discussed in the previous section and as it is shown in Fig. 3, $\Delta r_{K}^{e/\mu}$ and $\mathcal{B}(\tau\\!\to\\!e\eta)$ are closely related, at least in the limiting case where $\Delta_{L}=0$. On the contrary, an analogue correlation between $\Delta r_{K}^{e/\mu}$ and $\mathcal{B}(\tau\\!\to\\!e\gamma)$ is absent, due to their different dependence on the SUSY mass spectrum. We also emphasize that experimentally visible effects in $\Delta r_{K}^{e/\mu}$ (at the $0.1\%$ level) can be reached up to charged Higgs masses at the TeV scale, as shown in Fig. 3 on the right. Moreover, we also stress that the present experimental bounds on $\Delta r_{K}^{e/\mu}$ at the $\%$ level already set constraints on the SUSY parameter space. Figure 2: SUSY Lepton Flavor Conserving contributions to $\Delta r_{K}^{e/\mu}$. Left: $\Delta r_{K}^{e/\mu}$ as a function of the (left- handed) mass ratio between the second and the first slepton generations. Black dots refer to the points satisfying $1\times 10^{-9}<\Delta a_{\mu}<5\times 10^{-9}$. Right: regions of the parameter space in the $\tan\beta-M_{H}$ plane where $0.001\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.003$ (red dots), $0.003\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.005$ (black dots) and $\|\Delta r_{K}^{e/\mu}\|\\!>\\!0.005$ (yellow dots). The plot has been obtained by means of a scan over the following parameter space: $(m_{L_{1,2}},m_{R_{1,2}},m_{\tilde{Q}},m_{\tilde{g}},m_{\tilde{W}},m_{\tilde{B}},M_{H})<2.5\rm{TeV}$, $\mu\\!<\\!5$TeV and $\tan\beta<60$. All the dots present in these and subsequent figures satisfy all the constraints discussed in Sec. 5. Figure 3: Left: $\Delta r_{K}^{e/\mu}$ as a function of $\mathcal{B}(\tau\\!\to\\!e\gamma)$ and $\mathcal{B}(\tau\\!\to\\!e\eta)$. Right: $\Delta r_{K}^{e/\mu}$ as a function of $M_{H}$. Both plots have been obtained by means of a scan over the following parameter space: $(m_{L,R},m_{\tilde{Q}},m_{\tilde{g}},m_{\tilde{W}},m_{\tilde{B}},M_{H})<2.5\rm{TeV}$, $\mu\\!<\\!5$TeV, $\|\delta_{RR}\|\\!<\\!0.5$, $\|\delta_{LL}\|\\!=\\!0$ and $\tan\beta<60$. Black dots refer to the points satisfying $1\times 10^{-9}<(g-2)_{\mu}<5\times 10^{-9}$. In Fig. 4, we show the SUSY parameter space in the $\tan\beta-M_{H}$ plane probed by an experimental resolution on $\Delta r_{K}^{e/\mu}$ up to the $0.1\%$ level. In particular, red dots refer to the points satisfying $0.001\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.003$, black dots refer to the points where $0.003\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.005$ and, finally, yellow dots are relative to the points where $\|\Delta r_{K}^{e/\mu}\|\\!>\\!0.005$. Figure 4: Regions of the parameter space in the $\tan\beta-M_{H}$ plane where $0.001\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.003$ (red dots), $0.003\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.005$ (black dots) and $\|\Delta r_{K}^{e/\mu}\|\\!>\\!0.005$ (yellow dots) as obtained by means of the same scan performed in Fig. 3. As discussed in Sec. 3, it is also possible to generate LFU breaking effects in $\Delta r_{K}^{e/\mu}$ by means of a double source of LFV that, as a final result, preserve the lepton flavor (see Eq. 12). This is the case when both LL and RR flavor violating sources are simultaneously non vanishing. The major novelty arising from this last possibility is that now the new physics contributions can interfere with the SM ones; thus, we can get both positive and negative values for $\Delta r_{K}^{e/\mu}$. This is clearly shown by Fig. 5 that is the analog of Fig. 3 but in the presence of non vanishing $\delta_{LL}$ LFV terms. We see that $\Delta r_{K}^{e/\mu}$ can lie in the experimentally interesting region while satisfying all the current constraints. We observe that also in this case, the requirement of large LFU breaking effects in $\Delta r_{K}^{e/\mu}$ at the level of $0.001\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.01$, can be compatible with an explanation for the $(g-2)_{\mu}$ anomaly while satisfying the constraints from $\mathcal{B}(\tau\\!\to\\!e\gamma)$ and $\mathcal{B}(\tau\\!\to\\!e\eta)$. Figure 5: Left: $\Delta r_{K}^{e/\mu}$ as a function of $\mathcal{B}(\tau\\!\to\\!e\gamma)$ and $\mathcal{B}(\tau\\!\to\\!e\eta)$. Right: $\Delta r_{K}^{e/\mu}$ as a function of $M_{H}$. Both plots have been obtained by means of a scan over the following parameter space: $(m_{L,R},m_{\tilde{Q}},m_{\tilde{g}},m_{\tilde{W}},m_{\tilde{B}},M_{H})<2.5\rm{TeV}$, $\mu\\!<\\!5$TeV, $\|\delta_{RR,LL}\|\\!<\\!0.5$ and $\tan\beta<60$. Black dots refer to the points satisfying $1\times 10^{-9}<(g-2)_{\mu}<5\times 10^{-9}$. Finally, Fig. 6 shows the parameter space in the $\tan\beta-M_{H}$ plane probed by an experimental resolution on $\Delta r_{K}^{e/\mu}$ up to the $0.1\%$ level in analogy to Fig. 4. Figure 6: Regions of the parameter space in the $\tan\beta-M_{H}$ plane where $0.001\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.003$ (red dots), $0.003\\!<\\!\|\Delta r_{K}^{e/\mu}\|\\!<\\!0.005$ (black dots) and $\|\Delta r_{K}^{e/\mu}\|\\!>\\!0.005$ (yellow dots) as obtained by means of the same scan performed in Fig. 5. ## 8 2HDM framework Theories with only one Higgs doublet, like the Standard Model (SM), do not contain flavor violating interactions of the fermions with the Higgs bosons. In particular, it is always possible to simultaneously diagonalize the fermion mass matrices and the Higgs-fermion couplings. In general, this is no longer true in models with several Higgs doublets. In fact, up and down-type fermions can couple, at the same time, to more than a single scalar doublet and this naturally leads to FCNC effects at the tree level. To suppress tree level FCNC in the theory so as not to be in conflict with known experimental limits, an ad hoc discrete symmetry is typically invoked. For instance, in the 2HDM, the up-type and the down-type quarks couple either to the same Higgs doublet (this is known as the 2HDM-I) [6], or to different doublets (2HDM-II) [6]. On the other hand, in the most general case, the so-called 2HDM-III [40], no discrete symmetries are assumed and FCNC phenomena naturally appear. The Lagrangian for the LFV Yukawa couplings of the 2HDM type III reads [40] $-{\cal L}=\eta_{ij}{\bar{\ell}}_{iL}H_{1}\ell_{jR}+\xi_{ij}{\bar{\ell}}_{iL}H_{2}\ell_{jR}+{\mathrm{h.c.}}$ (41) where $H_{1,2}$ are the Higgs doublets defined by $H_{1}=(\phi_{1}^{+}\,\,\phi_{1}^{0})$ and $H_{2}=(\phi_{2}^{+}\,\,\phi_{2}^{0})$ while $\eta_{ij}$ and $\xi_{ij}$ are off-diagonal $3\times 3$ matrices in the flavor space and $i$, $j$ are family indices. Passing to the basis where the leptons are in mass eigenstates and expressing the leptonic Lagrangian in terms of $\xi_{ij}$, one find that $\displaystyle{\cal L}$ $\displaystyle=$ $\displaystyle-\frac{m_{i}}{vc_{\beta}}{\bar{\ell}}_{i}\ell_{i}(s_{\alpha}h^{0}-c_{\alpha}H^{0})+i\frac{m_{i}t_{\beta}}{v}{\bar{\ell}}_{i}\gamma_{5}\ell_{i}A^{0}+\frac{m_{i}t_{\beta}}{\sqrt{2}v}{\bar{\nu}}_{i}(1+\gamma_{5})\ell_{i}H^{+}$ (42) $\displaystyle-$ $\displaystyle\frac{m_{i}}{vc_{\beta}}{\bar{\ell}}_{i}\xi_{ij}\ell_{j}\left[c_{\alpha-\beta}h^{0}+s_{\alpha-\beta}H^{0}\right]-\frac{im_{i}}{vc_{\beta}}{\bar{\ell}}_{i}\xi_{ij}\gamma_{5}\ell_{j}A^{0}-\frac{m_{i}}{\sqrt{2}vc_{\beta}}{\bar{\nu}}_{i}\xi_{ij}(1+\gamma_{5})\ell_{j}H^{+}+{\mathrm{h.c.}}$ (43) where $v=\sqrt{v_{1}^{2}+v_{2}^{2}}=(\sqrt{2}G_{F})^{-1/2}=246$ GeV. Note that the Lagrangian (43), expressed in terms of the $\xi_{ij}$ matrices, has the same lepton flavor conserving interactions of the to 2HDM-II. To be consistent with experimental data on FCNC processes, Cheng and Sher (CS), inspired by the hierarchy in the fermion masses, have proposed the following famous ansatz for the couplings $\xi_{ij}$ [41]: $\xi_{ij}=\lambda_{ij}\sqrt{\frac{m_{j}}{m_{i}}},$ (44) where the residual arbitrariness of flavor changing couplings is expressed by the parameters $\lambda_{ij}$ which is constrained by experimental bounds on LFV processes. By making use of the effective Lagrangion of Eq. 43, it is straightforward to compute the expression for the quantity $\Delta r^{e/\mu}_{K}$. It turns out that $\displaystyle\Delta r^{e/\mu}_{K}$ $\displaystyle\simeq$ $\displaystyle\bigg{(}\frac{m^{4}_{K}}{M^{4}_{H}}\bigg{)}\bigg{(}\frac{m_{\tau}}{m_{e}}\bigg{)}^{2}\|\xi_{31}\|^{2}t_{\beta}^{4}$ (45) $\displaystyle\simeq$ $\displaystyle 10^{-2}\times\bigg{(}\frac{500\rm{GeV}}{M_{H}}\bigg{)}^{4}\bigg{(}\frac{t_{\beta}}{40}\bigg{)}^{4}\|\lambda_{31}\|^{2}\,,$ (46) where in Eq. 46 we made use of the the CS ansatz; Eq. 46 clearly shows that a 2HDM of type III, with the addition of the CS ansatz, can naturally predict a LFU breaking in the $K\to\ell\nu$ systems at a visible level for natural values of $M_{H}$ and $t_{\beta}$. We also remind that the $\lambda_{ij}$ parameters should be typically of order one [41]. However, once we assume the CS ansatz, we are naturally lead with stringent correlations among all the LFV transitions, as for example $\mu\to e\gamma$, $\mu+N\to e+N$, $\tau\to e\gamma$, $\tau\to\mu\gamma$ and so on. We have explicitly checked that the precision test provided by LFU breaking effects in $\Delta r^{e/\mu}_{K}$ represents the most powerful probe of the CS ansatz, at least in the decoupling regime (where $M_{H}\simeq M_{A}$ and where the lightest higgs boson doesn’t have LFV couplings with fermions) and assuming non vanishing LFV interactions only for $3j$ transitions. Irrespective of the specific model one can assume, we wish to emphasize that LFU breaking effects in $\Delta r^{e/\mu}_{K}$ represent the best probe for $31$ LFV transitions in a generic 2HDM with tree level LFV couplings (see Eq. 43). To see this point explicitly, it is natural to compare the New Physics sensitivity in LFU violation to that achievable in $\tau\rightarrow e\eta$. Indeed, this latter decay channel represents the most sensitive channel to New Physics among all the rare $\tau$ decays (this is strictly true in the decoupling limit). It turns out that $Br(\tau\rightarrow e\eta)\simeq 10^{-8}\times\Delta r^{e-\mu}_{\\!K}$ (47) showing that, within a 2HDM-III framework, $\Delta r^{e-\mu}_{\\!K}$ sets tight constraints on the observation of $\tau\rightarrow e\eta$ at the level of $Br(\tau\rightarrow e\eta)\leq 10^{-10}$. ## 9 Conclusions High precision electroweak tests, such as deviations from the Standard Model expectations of the Lepton Universality breaking, represent a powerful tool to test the Standard Model and, hence, to constrain or obtain indirect hints of new physics beyond it. Kaon and pion physics are obvious grounds where to perform such tests, for instance in the $K\rightarrow\ell\nu_{\ell}$ decays, where $l=e$ or $\mu$. In this paper, we have analyzed the domain of $\Delta r_{K}^{e/\mu}$ between $10^{-3}<\Delta r_{K}^{e/\mu}<10^{-2}$. An evidence of LFU violation at the level of $\Delta r_{K}^{e/\mu}$ larger than $5\times 10^{-3}$ unambiguously points towards the presence of LFV sources. On the other hand, if our increased experimental sensitivity allows us to observe an LFU violation with values of $\Delta r_{K}^{e/\mu}$ smaller than $5\times 10^{-3}$, then both the flavor conserving and the flavor changing sources of LFU violation can be at play. In any case, the observation of a non-vanishing $\Delta r_{K}^{e/\mu}$ in the range $10^{-3}<\Delta r_{K}^{e/\mu}<5\times 10^{-3}$ would severely limit values in the $M_{H}-\tan\beta$ plane. If a signal exists at a such a level, the LHC results become the crucial tool to disentangle flavor conserving and flavor changing sources of LFU violation. Interestingly enough, a process that in itself does not need lepton flavor violation to occur, i.e. the violation of $\mu-e$ non-universality in $K_{\ell 2}$, proves to be quite effective in constraining not only relevant regions of SUSY models where lepton flavor is conserved, but even those where specific lepton flavor violating contributions arise. Indeed, a comparison with analogous bounds coming from $\tau$ Lepton Flavor Violation decays shows the relevance of the measurement of $R_{K}$ to probe Lepton Flavor Violation in SUSY. Acknowledgments: This work has been supported in part by the Cluster of Excellence “Origin and Structure of the Universe” and by the German Bundesministerium für Bildung und Forschung under contract 05HT6WOA. One of us (A. M.) acknowledges Research Grants from the Istituto Nazionale di Fisica Nucleare (INFN), project FA51 and from the Ministero dell’ Istruzione, dell’Universita’ e della Ricerca (MIUR) and Univ. of Padova, astroparticle research project $2006023491-001$. This research was supported in part by the European Community Research Training Networks under contracts MRTN- CT-2004-503369 and MRTN-CT-2006-035505. One of us (P. P.) acknowledges Andrzej J. Buras for reading the paper and for useful discussions. ## References [1] W. J. Marciano and A. Sirlin, Phys. Rev. Lett. 71 3629 (1993); M. Finkemeier, Phys. Lett. B 387 391 (1996); V. Cirigliano and I. Rosell, Phys. Rev. Lett. 99 (2007) 231801. * [2] W. M. Yao et al. [Particle Data Group], J. Phys. G 33 (2006) 1. * [3] M. Antonelli et al. [FlaviaNet Working Group on Kaon Decays], arXiv:0801.1817 [hep-ph]; Flavianet kaon WG, http://www.lnf.infn.it/wg/vus/ * [4] A. Masiero, P. Paradisi, and R. Petronzio Phys. Rev. D 74, 011701 (2006) [arXiv:hep-ph/0511289]. * [5] K. S. Babu and C. Kolda, Phys. Rev. Lett. 89 (2002) 241802 [hep-ph/0206310]. * [6] J. F. Gunion, H. E. Haber, G. L. Kane and S. Dawson, The Higgs hunter Guide (Addison-Wesley, Reading, MA, 1990). * [7] L. J. Hall, R. Rattazzi and U. Sarid, Phys. Rev. D 50 (1994) 7048 [arXiv:hep-ph/9306309]. * [8] P. Paradisi, JHEP 0510, 006 (2005) [arXiv:hep-ph/0505046]; JHEP 0602, 050 (2006) [arXiv:hep-ph/0508054]; JHEP 0608, 047 (2006) [arXiv:hep-ph/0601100]. * [9] A. Brignole and A. Rossi, Phys. Lett. B 566, 217 (2003) [arXiv:hep-ph/0304081]; E. Arganda, A. M. Curiel, M. J. Herrero and D. Temes, Phys. Rev. D 71, 035011 (2005) [arXiv:hep-ph/0407302]. * [10] I. Masina and C. A. Savoy, Nucl. Phys. B 661 (2003) 365 [arXiv:hep-ph/0211283]. * [11] P. H. Chankowski, A. Dabelstein, W. Hollik, W. M. Mosle, S. Pokorski and J. Rosiek, Nucl. Phys. B 417 (1994) 101. * [12] P. Krawczyk and S. Pokorski, Phys. Rev. Lett. 60 (1988) 182; P. H. Chankowski, R. Hempfling and S. Pokorski, Phys. Lett. B 333 (1994) 403 [arXiv:hep-ph/9405281]; Y. Grossman, Nucl. Phys. B 426 (1994) 355 [arXiv:hep-ph/9401311]. * [13] M. J. Ramsey-Musolf, S. Su and S. Tulin, Phys. Rev. D 76, 095017 (2007) [arXiv:0705.0028 [hep-ph]]. * [14] S. Schael et al. [ALEPH Collaboration], Eur. Phys. J. C 47 (2006) 547 [arXiv:hep-ex/0602042]. * [15] S. Eidelman et al. [Particle Data Group], Phys. Lett. B 592, 1 (2004). * [16] J.A. Casas and S. Dimopoulos, Phys. Lett. B 387 (1996) 107 [hep-ph/9606237]. * [17] S. Heinemeyer, W. Hollik and G. Weiglein, Phys. Rept. 425 (2006) 265 [arXiv:hep-ph/0412214]. * [18] M. Misiak et al., Phys. Rev. Lett. 98, 022002 (2007) [arXiv:hep-ph/0609232]. * [19] E. Barberio et al. [Heavy Flavor Averaging Group (HFAG) Collaboration], arXiv:0704.3575 [hep-ex] and online update at http://www.slac.stanford.edu/xorg/hfag. * [20] P. Koppenburg et al. [Belle Collab.], Phys. Rev. Lett. 93 (2004) 061803 [hep-ex/0403004]. * [21] B. Aubert et al. [BaBar Collab.], Phys. Rev. Lett. 97 (2006) 171803 [hep-ex/0607071]. * [22] R. S. Chivukula and H. Georgi, Phys. Lett. B188 (1987) 99; L. J. Hall and L. Randall, Phys. Rev. Lett. 65 (1990) 2939–2942; A. J. Buras, P. Gambino, M. Gorbahn, S. Jager, and L. Silvestrini, Phys. Lett. B500 (2001) 161–167; A. J. Buras, Acta Phys. Polon. B34 (2003) 5615–5668; G. D’Ambrosio, G. F. Giudice, G. Isidori, and A. Strumia, Nucl. Phys. B645 (2002) 155–187. * [23] M. Misiak and M. Steinhauser, Nucl. Phys. B 764, 62 (2007); M. Misiak and M. Steinhauser, private communication. * [24] G. Degrassi, P. Gambino and G. F. Giudice, JHEP 0012 (2000) 009 [hep-ph/0009337]; M. Carena, D. Garcia, U. Nierste and C. E. Wagner, Phys. Lett. B B499 (2001) 141 [hep-ph/0010003]. * [25] K. Ikado et al. [Belle Collab.], hep-ex/0604018. * [26] B. Aubert et al. [BaBaR Collab.], hep-ex/0608019. * [27] A. Gray et al. [HPQCD Collaboration], Phys. Rev. Lett. 95 (2005) 212001 [arXiv:hep-lat/0507015]. * [28] W. S. Hou, Phys. Rev. D 48 (1993) 2342. * [29] G. Isidori and P. Paradisi, Phys. Lett. B 639 (2006) 499 [hep-ph/0605012]. * [30] K. S. Babu and C. F. Kolda, Phys. Rev. Lett. 84 (2000) 228 [arXiv:hep-ph/9909476]. * [31] H. N. Brown et al. [Muon g-2 Collaboration], Phys. Rev. D 62 (2000) 091101 [hep-ex/0009029]; Phys. Rev. Lett. 8620012227 [hep-ex/0102017]; Phys. Rev. Lett. 89 (2002) 101804 [hep-ex/0208001]; Phys. Rev. Lett. 92 (2004) 161802 [hep-ex/0401008]. * [32] M. Passera, J. Phys. G 31 (2005) R75 [hep-ph/0411168]; Nucl. Phys. Proc. Suppl. 155 (2006) 365 [hep-ph/0509372]. * [33] M. Passera, W. J. Marciano and A. Sirlin, arXiv:0804.1142 [hep-ph]. * [34] T. Moroi, Phys. Rev. D 53 (1996) 6565 [hep-ph/9512396]. * [35] M. Sher, Phys. Rev. D 66, 057301 (2002) [arXiv:hep-ph/0207136]. * [36] A. Brignole and A. Rossi, Nucl. Phys. B 701, 3 (2004) [arXiv:hep-ph/0404211]. * [37] E. Arganda, M. J. Herrero and J. Portoles, JHEP 0806, 079 (2008) [arXiv:0803.2039 [hep-ph]]. * [38] J. Hisano, T. Moroi, K. Tobe and M. Yamaguchi, Phys. Rev. D 53, 2442 (1996) [arXiv:hep-ph/9510309]. * [39] J. Hisano and K. Tobe, Phys. Lett. B 510, 197 (2001) [arXiv:hep-ph/0102315]; G. Isidori, F. Mescia, P. Paradisi and D. Temes, Phys. Rev. D 75, 115019 (2007) [arXiv:hep-ph/0703035]. * [40] D. Atwood, L. Reina and A. Soni, Phys. Rev. D 55 (1997) 3156 [arXiv:hep-ph/9609279]. * [41] T. P. Cheng and M. Sher, Phys. Rev. D 35 (1987) 3484.	arxiv-papers	2008-07-29T19:35:23	2024-09-04T02:48:57.023298	{ "license": "Public Domain", "authors": "A.Masiero, P.Paradisi, and R.Petronzio", "submitter": "Paride Paradisi", "url": "https://arxiv.org/abs/0807.4721" }
0807.4817	# Singular cotangent model Carlos Currás-Bosch 111partially supported by DGICYT: MTM 2006-04353. ###### Abstract Any singular level of a completely integrable system (c.i.s.) with non- degenerate singularities has a singular affine structure. We shall show how to construct a simple c.i.s. around the level, having the above affine structure. The cotangent bundle of the desingularised level is used to perform the construction, and the c.i.s. obtained looks like the simplest one associated to the affine structure. This method of construction is used to provide several examples of c.i.s. with different kinds of non-degenerate singularities. Subject Classification: 53D05, 37J35,37G05 Keywords: Symplectic manifold; Singular Lagrangian foliation; Integrable Hamiltonian system; Affine structure on a level; Normal forms ## 1 Introduction Let $(M^{2n},\omega,F)$ be a non-degenerate integrable system. This means that $(M^{2n},\omega)$ is a symplectic manifold and $F=(f_{1},...,f_{n})$ is a proper moment map, which is non-singular almost everywhere, and its singularities are of Morse-Bott type. The $\mathbb{R}^{n}$-action generated by the Hamiltonian vector fields $H_{f_{1}},...,H_{f_{n}}$ gives a singular Lagrangian foliation on $(M^{2n},\omega)$. Any leaf is an orbit of this actions. At the same time, any connected component of $F^{-1}(c)$ is a level. We know that the regular levels are $n$-dimensional tori and the singular levels are finite union of several leaves. A semilocal classification of such integrable systems is still open. It consists in finding a complete system of invariants describing symplectically a neighborhood of a level. Some approaches to solve this question has been currently made, see [4], [13], [15], [16]. As the local description of non- degenerate singularities is given in terms of products of elliptic, hyperbolic and focus-focus components, the number of elliptic, hyperbolic and focus-focus components at each point of the level will play an essential role in the classification problem. As in the regular case, the Hamiltonian vector fields $H_{f_{1}},...,H_{f_{n}}$ endow any leaf and any level with an affine structure with singularities. In studying the semi-local classification, we have seen that this affine structure gives strong conditions on the set of invariants found. Our proposal in this paper is to prove that this affine structure allows to the construction of a completely integrable system around a given level $L_{0}$ of $(M^{2n},\omega,F)$, such that the affine structure on $L_{0}$ is the given one. This construction looks like the simplest one with the given affine structure on $L_{0}$. As the “1-jet ”of the former completely integrable system (c.i.s. from now on) and the one constructed in this way coincide, it takes sense to denote this c.i.s. as the linearized c.i.s. of the initial one. The process of construction of the linearized c.i.s. indicates us a way to construct c.i.s. (in fact thew will be linearized completely integrable systems) with prescribed non-degenerate singularities along a given singular level. Some constructions are given. I would like to express my gratitude to Pierre Molino for interesting and fruitful conversations on this subject. ## 2 Definition and basic properties. ### 2.1 Local expressions Let $(M^{2n},\omega,f_{1},...,f_{n})$ be an integrable system, i.e., $(M^{2n},\omega)$ is a symplectic manifold. The functions $f_{1},...,f_{n}$ are Poisson commuting (first integrals of a given Hamiltonian system), such that $df_{1}\wedge\cdot\cdot\cdot\wedge df_{n}\neq 0$ on a dense subset of $M^{2n}$, and the moment map $F:M^{2n}\longrightarrow\mathbb{R}^{n}$, $F=(f_{1},...,f_{n})$, is proper. Such an integrable system is said to be non- degenerate if, in a neighborhood of each point $p_{0}\in M^{2n}$, there exist canonical coordinates $(x_{1},y_{1},...,x_{n},y_{n})$, and $n$ local functions $h_{1},...,h_{n}$, which have one of the following expressions: $h_{i}=y_{i}\qquad\text{(regular terms)}$ $h_{i}=(x_{i})^{2}+(y_{i})^{2}\qquad\text{(elliptic terms)}$ $h_{i}=x_{i}y_{i}\qquad\text{(hyperbolic terms)}$ $\begin{cases}h_{i}=x_{i}y_{i}+x_{i+1}y_{i+1}\\\ h_{i+1}=x_{i}y_{i+1}-y_{i}x_{i+1}\qquad\text{(focus-focus terms)}\end{cases}$ such that: 1. 1. $f_{1},...,f_{n}$ Poisson commute with $h_{1},...,h_{n}$. 2. 2. $\\{j_{p_{0}}^{2}f_{1},...,j_{p_{0}}^{2}f_{n}\\}$ and $\\{j_{p_{0}}^{2}h_{1},...,j_{p_{0}}^{2}h_{n}\\}$ generate the same space of $2$-jets at $p_{0}$. This notion of non-degeneracy implies obvious conditions on the space of $2$-jets generated at each point by $f_{1},...,f_{n}$. Conversely, assuming these infinitesimal conditions, the existence of such adapted coordinates is an important result, due to H.Eliasson [9]. The demonstration has been completed by E.Miranda [13] and E.Miranda & V.N.San [14] . Adapted coordinates will be referred to as Eliasson coordinates, or simply E-coordinates. Let $\cal{L}$ be the Lagrangian singular foliation associated to this c.i.s., i.e. the leaves of $\cal{L}$ are the orbits of the $\mathbb{R}^{n}$-action generated by $H_{f_{1}},...,H_{f_{n}}$. From a practical point of view, one can locally regard at $(M^{2n},\omega,\cal{L})$ as $(\mathbb{R}^{2n},\omega_{0},{\cal{L}}_{0})$, where $\omega_{0}$ is the standard symplectic two-form on $\mathbb{R}^{2n}$ and ${\cal{L}}_{0}$ is given by $dh_{i}=0$, $i=1,...,n.$ Let $L_{0}$ be a singular level, and $p_{0}$ a point of $L_{0}$. By taking E-coordinates around $p_{0}$ we have the following characteristic numbers of the point $p_{0}$: the numbers $k_{e}$, $k_{h}$ and $k_{f}$ correspond to the number of elliptic, hyperbolic and focus-focus terms of the set $(h_{1},...,h_{n})$. The leaf through $p_{0}$ is $\mathbb{T}^{c}\times\mathbb{R}^{o}$, and the numbers $c$ and $o$ are called the degrees of closedness and openness of the leaf, respectively. Following Zung (see [20]), the $5$-tuple $(k_{e},k_{h},k_{f},c,o)$ is called the leaf-type and $(k_{e},k_{h},k_{f})$ the Williamson type of $p_{0}$. In general one has $k_{e}+k_{h}+2k_{f}+c+o=n$. In [20] it is proved that the three numbers $k_{e},k_{f}+c,k_{h}+k_{k}+o$, are invariants of each level. These numbers are known as the degrees of ellipticity, closedness and openness of the level. Summarizing, one can say that a singular level $L_{0}$ is a compact $(n-k_{e})$-manifold, with self-intersections provided with a manifold structure of dimension less than $(n-k_{e})$. This level is endowed with a non-degenerate $\mathbb{R}^{n-k_{e}}$-action. Non-degenerate action means that the isotropy of this action at each point is linearizable. A we have proved (see [5]) that a completely integrable system, wit $k_{e}\neq 0$, is equivalent to a product of a standard elliptic model (with degree of ellipticity equal to $k_{e}$) with a c.i.s. with zero ellipticity degree, we will restrict our attention along this paper to the case $k_{e}=0$. ### 2.2 Desingularized level We assume that on the singular level $L_{0}$, with singular affine structure $\nabla_{0}$, the degree of ellipticity vanishes. In order to give a construction of a “standard ”c.i.s., such that $L_{0}$ is a singular level, and the induced singular affine structure on it coincides with $\nabla_{0}$, we start by giving the construction of the so called desingularized level: As $k_{e}=0$ on $L_{0}$, $\text{dim.}L_{0}=n$. Let $\psi$ be a differentiable embedding of $M^{2n}$ in an euclidean space $\mathbb{R}^{l}$. Let $L_{0}^{r}\subset L_{0}$ be the subset of regular points of $L_{0}$, i.e., a point $p\in L_{0}$ lies in $L_{0}^{r}$ if and only if the rank of $dF$ is equal to $n$ at $p$. Note that $L_{0}^{r}$ is not, in general, a connected submanifold. We consider the following embedding of $L_{0}^{r}$ $\varphi:L_{0}^{r}\hookrightarrow M^{2n}\times G_{n}(\mathbb{R}^{l})\quad,\quad p\mapsto(p,[T_{\psi(p)}(\psi(L_{0}))]),$ where $G_{n}(\mathbb{R}^{l})$ is the $n$ dimensional Grassmann manifold of $\mathbb{R}^{l}$. We define the desingularized level, ${\hat{L}}_{0}$ as the closure of $\varphi(L_{0}^{r})$. ###### Claim 2.1 ${\hat{L}}_{0}$ is a $n$-dimensional submanifold of $M^{2n}\times G_{n}(\mathbb{R}^{l})$ Proof: Let $q$ be a point of $\varphi(L_{0}^{r})$. As $\varphi$ is an embedding, we have a $n$-dimensional natural chart defined around $q$. Let $q=(p,[V])$ be a point in $\text{Cl}(\varphi(L_{0}^{r}))\setminus\varphi(L_{0}^{r})$. Obviously $p\in L_{0}\setminus L_{0}^{r}$. We know that there is a chart $U$ in $M^{2n}$, with coordinates $(x_{1},y_{1},...,x_{n},y_{n})$, centered at $p$, and $L_{0}\cap U$ is given by $h_{i}=0$, $i=1,...,n$, where the functions $h_{i}$ are in the form of section 2.1. Let us have a look at the tangent space to $L_{0}$ at a point near to $p$: from the expressions of $h_{i}$, we see $L_{0}\cap U=\prod_{j=1}^{s}(L^{j}_{0}\cap U^{j}),$ where each $L_{0}^{j}\cap U^{j}$ is of the form a) $U^{j}=D_{2}$, with coordinates $(x_{j},y_{j})$, $h_{j}=x_{j}$, and $L_{0}^{j}\cap U^{j}$ is given by $h_{j}=0$ (regular term). b) $U^{j}=D_{2}$, with coordinates $(x_{j},y_{j})$, $h_{j}=x_{j}y_{j}$, and $L_{0}^{j}\cap U^{j}$ is given by $h_{j}=0$ (hyperbolic term). c) $U_{j}=D_{2}\times D_{2}$, with coordinates $(x_{j},y_{j},x_{j+1},y_{j+1})$, $h_{j}=x_{j}y_{j}+x_{j+1}y_{j+1}$, $h_{j+1}=x_{j}y_{j+1}-x_{j+1}y_{j}$, and $L_{0}^{j}\cap U^{j}$ is given by $h_{j}=h_{j+1}=0$ (focus-focus term). So, the tangent space at a point near to $p$ will be: In the case a) $[T_{(0,y_{j})}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial y_{j}}>],$ so $[T_{(0,0)}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial y_{j}}>]$ in the case b) $[T_{(0,y_{j})}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial y_{j}}>]$ or $[T_{(x_{j},0)}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial x_{j}}>],$ so $[T_{(0,0)}(L_{0}^{j}\cap U^{j})]$ is either $[<\frac{\partial}{\partial y_{j}}>]\quad\text{or }\quad[<\frac{\partial}{\partial x_{j}}>]$ . and in the case c) $[T_{(0,y_{j},0,y_{j+1})}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial y_{j}},\frac{\partial}{\partial y_{j+1}}>],$ $\text{or}\quad[T_{(x_{j},0,x_{j+1},0)}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial x_{j}},\frac{\partial}{\partial x_{j+1}}>],$ so $[T_{(0,0,0,0)}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial y_{j}},\frac{\partial}{\partial y_{j+1}}>]\quad\text{or}$ $[T_{(0,0,0,0)}(L_{0}^{j}\cap U^{j})]=[<\frac{\partial}{\partial x_{j}},\frac{\partial}{\partial x_{j+1}}>].$ So, in fact,there are $2^{h+f}$ possibilities for the class of the grassmannian $[V]$ at $p\in L_{0}\setminus L_{0}^{r}$. As the structure of the chart around any one of these points will be a product structure, it will be sufficient to show the differentiability in the hyperbolic and in the focus-focus case. Let us do it, for instance, in the focus-focus cases: we can consider that $p=(0,0,0,0)$ and $[V]=[<\frac{\partial}{\partial x_{j}},\frac{\partial}{\partial x_{j+1}}>]$, then the coordinates we take around $(p,[V])$ are $(x_{j},x_{j+1})$. One can check easily that with these charts, and the previous charts around the regular points, we have a structure of $C^{\infty}$-manifold on ${\hat{L_{0}}}$. $\square$ The map $j:{\hat{L}}_{0}\longrightarrow M^{2n}\quad,\quad(p,[V])\longmapsto p$ is differentiable, $j({\hat{L}}_{0})=L_{0}$, and the preimage of any singular point on $L_{0}$ consists of $2^{h+f}$ points on ${\hat{L}}_{0}$. This map is a Lagrangian immersion in $(M^{2n},\omega)$ ## 3 The singular cotangent model ### 3.1 Affine structure on the desingularized level The Poisson action of the local $F$-basic functions defines a singular affine structure $\nabla_{0}$ on $L_{0}$. On each leaf of the level, the affine structure is defined by considering the infinitesimal generators of the $\mathbb{R}^{n}$-action as parallel vector fields. This $\mathbb{R}^{n}$-action can be lifted in a natural way to ${\hat{L}}_{0}$. Note that this lift is possible because the map $j:{\hat{L}}_{0}\longrightarrow M^{2n}$ is an immersion. As the affine structure on $L_{0}$ is provided by the Hamiltonian vector fields $H_{f_{1}}\|_{L_{0}}=X_{1},...,H_{f_{n}}\|_{L_{0}}=X_{n}$ it seems natural to write as ${\hat{X_{i}}}$, $i=1,...,n$, the vector fields on ${\hat{L}}_{0}$ induced through the immersion $j$, and ${\hat{\nabla}}_{0}$ this affine structure. By using the affine structure on ${\hat{L}}_{0}$ we are going to construct a completely integrable system on $(T^{}{\hat{L}}_{0},{\hat{\omega}}_{0})$, where ${\hat{\omega}}_{0}$ is the standard symplectic two-form on $T^{}{\hat{L}}_{0}$, and such that the affine structure on ${\hat{L}}_{0}$, induced by this c.i.s. is ${\hat{\nabla}}_{0}$. As a last step, by a standard process of gluing, by using natural local identifications, we will obtain a c.i.s., such that $L_{0}$ is one level, and the affine structure on $L_{0}$ will be $\nabla_{0}$. ### 3.2 A completely integrable system on $T^{}{\hat{L}}_{0}$ We define $n$ differentiable functions $g_{1}$,…,$g_{n}$ on $T^{}{\hat{L}}_{0}$ by $g_{i}((p,[V]),w):=<{\hat{X}}_{i}(p,[V]),w>$ Our proposal is to see that $(T^{}{\hat{L}}_{0},{\hat{\omega}}_{0},(g_{1},...,g_{n}))$ is completely integrable. To do it, we need only to prove that $\\{g_{i},g_{j}\\}\|_{0}=0$. By a continuity argument, it will be sufficient to prove it at the points $((p,[V]),-)\in{\hat{L}}_{0}$, where $p$ is a regular point. We can take Eliasson coordinates $(x_{1},y_{1},...,x_{n},y_{n})$ around the point $p$, and as any basic function only depends on $(y_{1},...,y_{n})$ in a neighborhood of $p$, the expression of $H_{f_{i}}$ in this neighborhood will be of the form $\sum_{j=1}^{n}{\frac{\partial f_{i}}{\partial y_{j}}}(0,...,0)\frac{\partial}{\partial x_{j}}$, i.e. is a vector fields with constant coefficients (which is obvious because the vector field is affine parallel). Let $\alpha$ be the Liouville form on $T^{}{\hat{L}}_{0}$, in this neighborhood $\alpha=\sum_{i=1}^{n}y_{i}dx_{i}$, $g_{i}=\sum_{k=1}^{n}y_{k}{\frac{\partial f_{i}}{\partial y_{k}}}(0,...,0)$, so $dg_{i}=\sum_{k=1}^{n}{\frac{\partial f_{i}}{\partial y_{k}}}(0,....,0)dy_{k}$, and obviously $\Lambda^{0}(dg_{i},dg_{j})=0.$ Once we know that $(T^{}{\hat{L}}_{0},{\hat{\omega}}_{0},(g_{1},...,g_{n}))$ is completely integrable, we remark that, in general, is not proper: let us assume, for instance that $\text{dim.}L_{0}=1$, and let $p$ be an hyperbolic point, we can write $\omega_{0}=dx\wedge dy$, and $h=xy$, then $x=0$ is a leaf. In order to have a proper completely integrable system around $L_{0}$we have to define a gluing between points on ${\hat{L}}_{0}$ which are projected, via $j$, to the same singular point on $L_{0}$. As the singular composition of each point is, in fact, a product of hyperbolic and focus-focus components, we need show how this gluing is done in hyperbolic and focus-focus cases. Hyperbolic gluing: Let $(p,[V_{1}]),(p,[V_{2}])\in{\hat{L}}_{0}$, where $p\in L_{0}$ is a purely hyperbolic point (degree of hyperbolicity equal to one). Then in Eliasson coordinates $(x,y)$ around the point $p$, we may assume $[V_{1}]=[<\frac{\partial}{\partial x}>]$ and $[V_{2}]=[<\frac{\partial}{\partial y}>]$. We take canonical coordinates in $T^{}{\hat{L}}_{0}$ in two neighborhoods $U_{1}$ of $(p,[V_{1}])$, $U_{2}$ of $(p,[V_{2}])$, and we denote them by $(x,y)$ and $(X,Y)$ respectively. Any basic function in these neighborhoods depends on $xy$ and $XY$ respectively, and the canonical symplectic two form is $dx\wedge dy$ and $dX\wedge dY$ respectively. The symplectomorphism from $U_{1}$ onto $U_{2}$, we are searching for, is expressed in these coordinates by $X=-y,Y=x$. Focus-focus gluing:We can proceed in a similar way for two points $(p,[V_{1}])$, $(p,[V_{2}])$, in the preimage of a focus-focus point $p\in L_{0}$. Regarding to Claim 2.1 one can consider $[V_{1}]=[<\frac{\partial}{\partial x_{1}},\frac{\partial}{\partial x_{2}}>]\quad,\quad[V_{2}]=[<\frac{\partial}{\partial y_{1}},\frac{\partial}{\partial y_{2}}>].$ We take canonical coordinates in $T^{}{\hat{L}}_{0}$ in two neighborhoods $U_{1}$ of $(p,[V_{1}])$ and $U_{2}$ of $(p,[V_{2}])$, and we denote them by $(x_{1},x_{2},y_{1},y_{2})$ and $(X_{1},X_{2},Y_{1},Y_{2})$ respectively. Any basic function in these neighborhoods depends on $x_{1}y_{1}+x_{2}y_{2},-y_{1}x_{2}+x_{1}y_{2}$ and $X_{1}Y_{1}+X_{2}Y_{2},-Y_{1}X_{2}+X_{1}Y_{2}$, respectively. The symplectomorphism from $U_{1}$ to $U_{2}$ we need is given by: $X_{1}=-y_{1}$, $X_{2}=-Y_{2}$, $Y_{1}=X_{1}$, $Y_{2}=-X_{2}$. ### 3.3 The singular cotangent model Once we have shown how to identify the neighborhoods of the points in the same preimage we get a germ of $2n$-dimensional symplectic manifold $(N,\omega_{0})$ containing $L_{0}$ as a singular Lagrangian submanifold. Now we check that the functions $g_{i}$, $i=1,...,n$, can be projected to $N$. To do it, we must see that the functions $g_{i}$ remain invariant under the above defined identifications in the hyperbolic and the focus-focus cases. In the hyperbolic case, we see that in the above defined neighborhood $U_{1}$, any function $g_{i}$ is of the form $y\frac{\partial h_{i}}{\partial y}(0)$, and a similar expression in $U_{2}$. Having in mind the symplectomorphism $Y=-x,X=y$, one sees that $h_{i}$ is preserved. For the focus-focus identification the proof is similar. It is quite obvious from the given construction of $(N,\omega_{0},(g_{1},...,g_{n}))$that the singular affine structure on $L_{0}$ is the previous one. ### 3.4 Construction of c.i.s. with prescribed singularities As a sort of application of the above considerations, let us point out how to give some c.i.s. with prescribed singularities, around a singular level. The intrinsic geometry of the level, i.e. the number and kind of singular points, is obviously related with the kind of singularities along its singular points. We show how to obtain a c.i.s. around a $2$-dimensional singular level with a focus-focus point and one circle of hyperbolic points: Let us consider the $2$-sphere $S^{2}$. This manifold $S^{2}$ will play the role of $\hat{L}_{0}$ of the above sections. We shall consider $T^{}S^{2}$ and two vector fields, with singularities, which will be used to define two functions on $T^{}S^{2}$. These functions have singularities at several points, and by furnishing the gluings at the corresponding singular points, we will get the c.i.s. around the singular level. Let $\theta$ (longitude) and $\varphi$ (latitude) be polar coordinates on $S^{2}$.The vector fields to consider are: $X=\frac{\partial}{\partial\theta}$ and $Y=h(\varphi)\frac{\partial}{\partial\varphi}$. We have to give a “good”expression for $h(\varphi)$, obviously we take $h(-\frac{\pi}{2})=0$, $h(\frac{\pi}{2})=0$, and $h(-\frac{\pi}{4})=h(\frac{\pi}{4})=0$. We consider four open subsets of $S^{2}$ $U_{1}=\\{(\theta,\varphi)\quad\|\quad-\frac{\pi}{2}\leq\varphi<-\frac{\pi}{2}+\varepsilon\\}$ $V_{1}=\\{(\theta,\varphi)\quad\|\quad-\frac{\pi}{4}-\varepsilon<\varphi<\frac{\pi}{4}+\varepsilon\\}$ $V_{2}=\\{(\theta,\varphi)\quad\|\quad\frac{\pi}{4}-\varepsilon<\varphi<\frac{\pi}{4}+\varepsilon\\}$ $U_{2}=\\{(\theta,\varphi)\quad\|\quad\frac{\pi}{2}-\varepsilon<\varphi\leq\frac{\pi}{2}\\}$ where $\varepsilon<\frac{\pi}{8}$. On $U_{1}$ and $U_{2}$we can take as coordinates the first two cartesian coordinates $(x_{1},x_{2})$, and consider the vector fields: on $U_{1}$, $x_{1}\frac{\partial}{\partial x_{1}}+x_{2}\frac{\partial}{\partial x_{2}}=-\frac{\cos\varphi}{\sin\varphi}\cdot\frac{\partial}{\partial\varphi}$ and $-x_{2}\frac{\partial}{\partial x_{1}}+x_{1}\frac{\partial}{\partial x_{2}}=\frac{\partial}{\partial\theta}$. on $U_{2}$, $-x_{1}\frac{\partial}{\partial x_{1}}-x_{2}\frac{\partial}{\partial x_{2}}=\frac{\cos\varphi}{\sin\varphi}\cdot\frac{\partial}{\partial\varphi}$ and $-x_{2}\frac{\partial}{\partial x_{1}}+x_{1}\frac{\partial}{\partial x_{2}}=\frac{\partial}{\partial\theta}$. On $V_{1}$ and $V_{2}$ we can take $(\theta,\varphi)$ as coordinates and the following vector fields: on $V_{1}$, $-(\varphi+\frac{\pi}{4})\frac{\partial}{\partial\varphi}$ and $\frac{\partial}{\partial\theta}$. on $V_{2}$, $(\varphi-\frac{\pi}{4})\frac{\partial}{\partial\varphi}$ and $\frac{\partial}{\partial\theta}$. Now we see that the function $h(\varphi)$ we are searching for can be a differentiable function of $\varphi$, such that its values in $U_{1},V_{1},U_{2},V_{2}$ are the above prescribed and not vanishing on $S^{2}\setminus(U_{1}\cup V_{1}\cup U_{2}\cup V_{2})$. Finally, we consider on $T^{}S^{2}$ the pair of functions $f,g$ associated to the vector fields $X,Y$, defined as follows: for any point $(z,w)\in T^{}S^{2}$, $f(z,w):=<X(z),w>$, $g(z,w):=<Y(z),w>$. The c.i.s. we are searching for is obtained by a process of gluing from $(T^{}S^{2},\omega_{0},(f,g))$. This gluing can be easily established by defining two symplectomorphisms: one of them is a local symplectomorphism between $(T^{}U_{1},(x_{1}=0,x_{2}=0,0,0))$ and $(T^{}U_{2},(X_{1}=0,X_{2}=0,0,0))$. We recall (see hyperbolic gluing in section3.2) that the mapping is given by $(x_{1},x_{2},y_{1},y_{2})\longmapsto(X_{1}=-y_{1},X_{2}=-y_{2},Y_{1}=x_{1},Y_{2}=x_{2}).$ The other symplectomorphism we need to conclude the gluing is a semi-local symplectomorphism from $(T^{}V_{1},(\theta,\varphi=-\frac{\pi}{4},0,0))$ in $T^{}V_{2},(\theta,\varphi=\frac{\pi}{4},0,0))$, described as follows: on $T^{}V_{1}$ we take $(\theta,\varphi+\frac{\pi}{4},\Theta,\Phi)$ as canonical coordinates. In the same form, we take $({\bar{\theta}}=\theta,{\bar{\varphi}}-\frac{\pi}{4}=\varphi-\frac{\pi}{4},{\bar{\Theta}}=\Theta,{\bar{\Phi}})$ as canonical coordinates in $T^{}V_{2}$. The symplectomorphism we need to define the gluing is $(\theta,\varphi+\frac{\pi}{4},\Theta,\Phi)\longmapsto({\bar{\theta}}=\theta,{\bar{\varphi}}-\frac{\pi}{4}=\Phi,{\bar{\Theta}}=\Theta,{\bar{\Phi}}=-(\varphi+\frac{\pi}{4})).$ Thus, the quotient of a germ of neighborhood of $S^{2}$ in $T^{}S^{2}$, by using these identifications provides us a germ of completely integrable system around a level having one singular point of focus-focus type and one circle of hyperbolic points. One sees from this construction that the same arguments can serve to give c.i.s. with several points of focus-focus type; it should be necessary to use different copies of $S^{2}$ and define a gluing by using the poles alternatively. The above construction suggests different ways of having circles of hyperbolic points in the level. ## References [1] Arnold,V.I.,Mathematical methods of classical mechanics, Graduate Texts in Math., 60. * [2] Audin,M.& Cannas da Siva,A. &Lerman,E.,Symplectic geometry of integrable systems, Advanced Courses in Math., CRM Barcelona, Birkhäuser (2003). * [3] Bolsinov,A.V. & Fomenko,A.T.,Integrable Hamiltonian systems, geometry, topological classification, Chapman& Hall/CRC (2004). * [4] Boucetta,M.& Molino,P.,Géométrie globale des systèmes Hamiltoniennes complétement intégrables,C.R.Acad.Sci.Paris, I, 308, 13(1989), 421-424. * [5] Currás-Bosch,C.,Decomposition of integrable and non-degenerate Hamiltonian systems with non-zero ellipticity degree, preprint. * [6] Currás-Bosch,C.&Molino,P.,Holonomie, suspensions et classifications pour les feuilletages Lagrangiens, C.R.Acad.Sci.Paris, I, 326, 11, (1989), 1317-1320. * [7] Delzant,T.,Hamiltoniens periodiques et images convexes de l’application moment, Bull.Soc.Math.France, 116, 3 (1988), 315-339. * [8] Duistermaat,J.J.,On global action-angle coordinates, Coomm.Pure Appl.Math., 33, 6(1980), 687-706. * [9] Eliasson,L.H.,Normal forms for Hamiltonian systems with Poisson commuting integrals, Ph.D. Thesis (1984). * [10] Fomenko,A.T.,Topological classification of integrable systems, Adv.Soviet Math., 6, AMS, Providence R.I. (1991). * [11] Lerman,E.& Umanskii,Y., Classification of 4-dimensional integrable Hamiltonian systems, Methods of qualitative theory of bifurcations, Iz.Garkov Univ., Gorki (1988), 67-76. * [12] Mineur,H.,Réduction des systèmes mécaniques à n degrès de liberté, Jour.Math.Pure Appl., IX, 15 (1936), 385-389. * [13] Miranda,E.,On symplectic linearization of singular Lagrangian foliations , Tesi Univ. de Barcelona (2003). * [14] Miranda,E.&San,V.N.,A singular Poincaré Lemma, Int. Math. Res. Notices, N 1 (2005), 27-45. * [15] Miranda,E.&Zung,N.T.,Equivariant normal forms for non-degenerate singular orbits of integrable Hamiltonian systems, Ann.Sci.Ec.Norm.Sup., serie 4, 37 (2004), 819-839. * [16] Molino, P.,Action-angle with singularities and non-degenerate integrable systems, (preprint). * [17] Toulet,A.,Classification des systèmes intégrables en dimension 2, Thèse Univ. Montpellier II (1996). * [18] San,V.N.,On semi-global invariants for focus-focus singularities, Topology 42 (2003), 365-380. * [19] Zung,N.T.,Symplectic topology of integrable Hamiltonian systems, Thèse Univ. Strasbourg (1994). * [20] Zung,N.T.,Symplectic topology of integrable Hamiltonian systems I,Compos.Math., 101 (1996), 179-219. * [21] Zung,N.T.,Symplectic topology of integrable Hamiltonian systems II, Compos.Math. 138 (2003), n 2, 125-156. * [22] Weinstein,A.Lectures on symplectic manifolds, Regional Conf.Series in Math., AMS (1976). * [23] Williamson,J.,On the algebraic problem concerning the normal form of linear dynamical systems, Amer.Jour. of Math., 58-1 (1936), 141-163. Departament d’Àlgebra i Geometria, Universitat de Barcelona (Spain) E-mail address:carloscurrasbosch@ub.edu	arxiv-papers	2008-07-30T17:32:56	2024-09-04T02:48:57.030760	{ "license": "Public Domain", "authors": "Carlos Curr\\'as-Bosch", "submitter": "Carlos Curr\\'as-Bosch", "url": "https://arxiv.org/abs/0807.4817" }
0807.4880	# Is the Universe More Transparent to Gamma Rays than Previously Thought? Floyd W. Stecker Astrophysics Science Division, NASA Goddard Space Flight Center Greenbelt, MD 20771, U.S.A. Floyd.W.Stecker@nasa.gov Sean T. Scully Department of Physics, James Madison University, Harrisonburg, VA 22807, U.S.A. scullyst@jmu.edu ###### Abstract The MAGIC collaboration has recently reported the detection of the strong $\gamma$-ray blazar 3C279 during a 1-2 day flare. They have used their spectral observations to draw conclusions regarding upper limits on the opacity of the Universe to high energy $\gamma$-rays and, by implication, upper limits on the extragalactic mid-infrared background radiation. In this paper we examine the effect of $\gamma$-ray absorption by the extragalactic infrared radiation on intrinsic spectra for this blazar and compare our results with the observational data on 3C279. We find agreement with our previous results, contrary to the recent assertion of the MAGIC group that the Universe is more transparent to $\gamma$-rays than our calculations indicate. Our analysis indicates that in the energy range between $\sim$ 80 and $\sim$ 500 GeV, 3C279 has a best-fit intrinsic spectrum with a spectral index $\sim$ 1.78 using our fast evolution model and $\sim$ 2.19 using our baseline model. However, we also find that spectral indices in the range of 1.0 to 3.0 are almost as equally acceptable as the best fit spectral indices. Assuming the same intrinsic spectral index for this flare as for the 1991 flare from 3C279 observed by EGRET, viz., 2.02, which lies between our best fit indeces, we estimate that the MAGIC flare was $\sim$3 times brighter than the EGRET flare observed 15 years earlier. gamma-rays: theory, galaxies ; 3C279: cosmology: diffuse radiation ††slugcomment: Submitted to The Astrophysical Journal Letters ## 1 INTRODUCTION It has long been recognized that by studying the spectra of strong extragalactic $\gamma$-ray sources one can obtain information about the density and energy spectra of intergalactic photon fields. The luminous blazar 3C279 was discovered by the EGRET detector aboard the Compton Gamma Ray Observatory to be a strong flaring $\gamma$-ray source (Hartman et al. 1992). Shortly after this discovery, (Stecker, de Jager & Salamon 1992) (hereafter SDS) proposed that the study of the TeV spectra of such sources could be used to probe the intergalactic infrared radiation. SDS proposed to look for the energy dependent features indicating the mutual annihilation of very high energy $\gamma$-rays and low energy photons of galactic origin via the process of electron-positron pair production $\gamma+\gamma\rightarrow e^{+}+e^{-}$. The cross section for this process is exactly determined; it can be calculated using quantum electrodynamics (Breit & Wheeler 1934). Thus, in principle, if one knows the emission spectrum of an extragalactic source at a given redshift, one can determine the column density of photons between the source and the Earth as a function of redshift. Since the EGRET discovery of 3C279, the infrared background at wavelengths not totally dominated by galactic or zodiacal emission has been measured by the Cosmic Background Explorer (COBE). In addition, there have been extensive observations of IR emission from galaxies themselves, whose total emission is thought to make up the cosmic IR background (see review by Hauser & Dwek 2001). The latest extensive observations have been made by the Spitzer satellite. It is thus appropriate to use a synoptic approach combining the very high energy $\gamma$-ray observations with the extragalactic IR observations, in order to best explore both the $\gamma$-ray emission from blazars and the diffuse extragalactic IR radiation. Using this approach, one must take account of both the high energy $\gamma$-ray observations and the data from the many galaxy observations presently available. The MAGIC collaboration has recently observed the spectrum of the blazar 3C279 during a flare which occurred on 22–23 February 2006 (Albert et al. 2008). The highly luminous blazar 3C279 lies at a redshift of 0.536 (Marzioni et al.1996). To date, it is the most distant $\gamma$-ray source observed in the sub-TeV energy range. Thus, as noted by SDS, this source is potentially highly significant for probing the intergalactic background radiation. Albert et al. (2008) used their observational data to draw conclusions regarding the maximum opacity of the Universe to $\gamma$-rays in the sub-TeV energy range. Their conclusions regarding the extragalactic background radiation would appear to disfavor the results of the extensive semi-empirical calculations of the extragalactic IR background spectrum given by Stecker, Malkan & Scully (2006, 2007) (hereafter SMS). In particular, the limit shown in their Figure 2 appears to be inconsistent with one of the SMS models that was based primarily on galaxy studies by the Spitzer infrared satellite telescope. In this paper, we will reexamine both the analysis assumptions and the conclusions presented in the paper of Albert et al. Using a different analysis technique that we show to be superior to that of Albert et al. , we find that the observations of 3C279 are fully consistent with both diffuse IR background models obtained by SMS. We then discuss the implications of these results regarding both the intrinsic energy spectrum and luminosity of the 3C279 flare and the opacity of the Universe to $\gamma$-rays. ## 2 The diffuse extragalactic IR background Various calculations of the extragalactic IR background have been made (Stecker, Puget & Fazio 1977; Malkan & Stecker (1998, 2001); Totani & Takeuchi (2002); Kneiske et al. (2004); Primack et al. (2005); SMS). Of these models, the most empirically based are those of Malkan & Stecker (1998, 2001), Totani & Takeuchi (2002) and SMS. Since the largest uncertainty in these calculations arises from the uncertainty in the temporal evolution of the star formation rate in galaxies, SMS assumed two different evolution models, viz., a “baseline” model and a “fast evolution” model. These models produced similar wavelength dependences for the spectral energy distribution of the extragalactic IR background, but gave a difference of roughly 30-40% in overall intensity. The empirically based calculations mentioned above include the observationally based contributions of warm dust and emission bands from polycyclic aromatic hydrocarbon (PAH) molecules and silicates, which have been observed to contribute significantly to galaxy emission in the mid-IR (e.g., Lagache et al. 2004). These components of galactic IR emission have the effect of partially filling in the “valley” in the mid-IR spectral energy distribution between the peak from starlight emission and that from cold dust emission. The model of Primack et al. (2005), which was based on strictly theoretical galaxy spectra, does not take the warm dust, PAH, and silicate emission components of mid-IR galaxy spectra into account and therefore exhibits a steep mid-IR valley that is in direct conflict with solid lower limits obtained from galaxy counts obtained from observations of galaxies at mid-IR wavelengths (Altieri et al. 1999; (Elbaz et al2002). This is clearly shown in Figure 2 of the supplemental online material of Albert et al. (2008). However, since Albert et al. (2008) considered it to be a “lower limit” model, we will discuss the Primack et al. model in our analysis. ## 3 The observed spectrum of 3C279 and its derived intrinsic spectrum According to the MAGIC analysis, the 2006 flare on 3C279 had an observed spectral index of 4.11 $\pm$ 0.68 in the energy range between $\sim 80$ and $\sim 500$ GeV. In their analysis, the MAGIC group chose to multiply their data points by $e^{\tau(E_{\gamma})}$, where the optical depth, $\tau(E_{\gamma}))$ is chosen by using the results of various optical depth calculations. They then fit simple power-law spectra to the resulting fluxes. Using estimated optical depths for only the fast evolution model of SMS, they gave a best-fit power-law spectral index for the intrinsic source spectrum of $\Gamma_{s}$ of 0.49 $\pm$ 1.19 for the which they imply is ruled out by invoking the questionable criterion $\Gamma_{s}\geq 1.5$ (However, see, e.g., (Katarzyńsky et al. 2006; Stecker, Baring & Summerlin 2007; Resmi & Bhattacharya 2008). In this paper, we will adopt a different method for analyzing the intrinsic spectrum of 3C279, based on the one we first used for the analysis of the H.E.S.S. observations of the source 1ES0229 + 200 (Stecker & Scully 2006). The method is superior to the approach used by the MAGIC collaboration where one chooses to force a power-law fit to the implied deabsorbed data points. Instead we assume an intrinsic power-law spectrum emitted by the source over the limited observed energy range that covers less than a decade in energy. In order to compare with the observations, we multiply this power law by an absorption factor $e^{-\tau(E_{\gamma},z=0.536)}$, where the optical depth, $\tau$, is calculated for a redshift z = 0.536. We then employ a nonlinear least squares fit of our two parameter model to the observational data. Our method has several advantages over that chosen by the MAGIC collaboration. To begin with, the redshift of the source and the energy range of the observations implies that even the counts in the lowest MAGIC energy bin have been affected by at least some amount of intergalactic absorption. As a result, multiplying the data by $e^{\tau}$ and then fitting an arbitrary power-law as done by Albert et al. (2008) does not allow a proper fit to the normalization of the spectrum. Both the true blazar luminosity and the true form of the intrinsic spectrum are masked. Furthermore, the actual intrinsic spectrum cannot be assumed to have a power-law form after multiplication by an exponential that is nonlinear in energy at this particular redshift (see, however, Stecker & Scully 2006). Also, in order to properly account for the effect of the optical depth, one should directly include it in the unfolding method used to produce the fluxes from the raw photon counts. Since the MAGIC data points represent a mean for an energy bin, it is not sufficient to simply multiply that mean point by $e^{\tau}$. The function $e^{\tau(E_{\gamma})}$ is a rapidly changing function of energy. This fact must be taken into account when computing both the spectrum and error bars. The implied weighting towards a lower energy within a bin indicates that the appropriate value for $\tau(E)$ should be lower than that assumed for the mean energy in the bin. Thus, the values for $\tau$ used in the MAGIC analysis are too high for all of the EBL models they used. ## 4 Analysis We begin our analysis program by calculating the optical depths in the energy range of the observations for z = 0.536 for both the SMS fast evolution model and baseline model. For comparison, we also consider the optical depths taken from Primack et al. (2005) and two of the models of Kneiske et al. (2004) , viz., their best-fit model and their low-IR model. We find that all five of the optical depth models considered are well fit by third order polynomials in the energy range of interest of the form shown in equation (1). $\log{\tau}=a_{1}\log^{3}E_{\gamma}+a_{2}\log^{2}E_{\gamma}+a_{3}\log E_{\gamma}+a_{4}$ (1) Table 1 summarizes the fit parameters to the five models considered here. Figure 1 shows the excellent agreement of the third order polynomial fits to optical depths for the five models. Table 1: Parameters Used in Equation (1) for Opacity Models Model \| $a_{1}$ \| $a_{2}$ \| $a_{3}$ \| $a_{4}$ ---\|---\|---\|---\|--- Fast Evolution \| 0.0999257 \| -1.26786 \| 5.36642 \| -6.21713 Baseline \| 0.103179 \| -1.28947 \| 5.41409 \| -6.36993 Kneiske Best Fit \| -0.477755 \| 2.84965 \| -3.77753 \| -0.353419 Kneiske Low IR \| -0.334462 \| 1.75885 \| -1.19612 \| -2.30134 Primack \| -0.549224 \| 3.43815 \| -5.63807 \| 1.57098 Figure 1: Polynomial fits to the five optical depth models discussed in the text. The solid line and dotted line are fits to the SMS fast evolution (triangles) and baseline (filled boxes) models respectively. The short-dashed and long-dashed lines are polynomial fits to the best-fit Kneiske et al. model (filled circles) and low-IR Kneiske et al. model (diamonds) respectively. The dot-dashed line is the polynomial fit to the Primack et al. optical depth model (stars). Figure 2: Plot of the best-fit curves with $\Gamma_{s}$ and $K$ as free parameters for the SMS fast evolution (solid), SMS baseline (dotted), Kneiske et al. best-fit (short-dashed), Kneiske et al. low-IR (long-dashed), and Primack et al. (dash-dotted) models shown along with the observational data from Albert et al. (2008). The shaded region represents the combined systematic and statistical error for the observational data. The gray curve shows the EGRET flare power-law fit extended to higher energies multiplied by $e^{-\tau(E)}$. The respective $\chi^{2}$ values are given in the third number column in Table 2. Because of the nonlinear nature of the energy dependence of the optical depth, we do not expect that the observed spectrum will have a power-law form. Thus, we fit to spectra that are assumed to be power-law intrinsic source spectra ($K{E_{\gamma}}^{-\Gamma}$) multiplied by $e^{-\tau}$ as prescribed by equation (1) using the various optical depth models indicated in Table 1 to take account of intergalactic absorption. We then employ the Levenberg- Marquardt nonlinear least squares method to fit the resulting nonlinear spectra to the observed data. We choose to fit to the data using only statistical errors. We first do a $\chi^{2}$ a fit with two free parameters, viz., the normalization coefficient, $K$ and the intrinsic spectral index of the source $\Gamma_{s}$. The best-fit spectral curves for the various EBL models considered are shown in Figure 2. The corresponding best-fit values $K$ and $\Gamma_{s}$ are given in Table 2. In this table, in the first column, the $K$ values are given for $E$ in GeV. The second column gives the best-fit spectral indeces for all of the models. The third column gives the $\chi^{2}$ values for the two-parameter best fit. It can be seen from this column that although the best-fit spectral indeces are significantly different for the various EBL models, the best fit $\chi^{2}$ values are comparable for all of the models. Table 2: Best-fit Spectral Parameters for EBL Models with $\chi^{2}$ Values Model \| $K$ \| $\Gamma_{s}$ \| $\chi^{2}$ \| $\chi^{2}$ for $\Gamma_{s}=2.02$ ---\|---\|---\|---\|--- Fast Evolution \| 2.15 \| 1.78 \| 1.59 \| 1.20 Baseline \| 8.75 \| 2.19 \| 1.42 \| 1.07 Kneiske Low IR \| 128 \| 3.46 \| 1.59 \| 1.28 Kneiske Best Fit \| 283 \| 3.62 \| 1.55 \| 1.42 Primack \| 141 \| 3.49 \| 1.54 \| 1.25 The Levenberg-Marquardt method yields substantially different best-fit spectral indices for our modeled optical depths and those of Kneiske et al. (2004) and Primack et al. (2005). However, these fits are not particularly unique in their goodness of fit. We performed a $\chi^{2}$ analysis on all of the models, allowing both $\Gamma_{s}$ and $K$ to be free parameters in the fit. As can be seen from Table 2, the form of the intrinsic spectrum of the 3C279 flare is undetermined by the five data bins obtained by the MAGIC collaboration, even when one only considers the large statistical errors involved and neglects the significant systematic errors. This is because the lower energy points with the smaller error bars are weighted more highly than the highest energy point. Since the minimum $\chi{{}^{2}}$ is almost independent of choice of spectral index, the two-parameter fitting routine mainly tries to move the curve up and down (i.e. adjusts the normalization) to best fit the observational data. This feature is lacking in the MAGIC analysis of the spectrum of 3C279 since they have factored in the absorption effect prior to making their fits. Figure 3: Plot of the best fit curves obtained using the SMS baseline model for fixed spectral indices of 1.5, 2.0, 2.5, and 3.0. The shaded region represents the combined systematic and statistical error for the observational data. To further illustrate this point of ambiguity in the spectral index of the source, Figure 3 shows the best fit obtained for the SMS baseline model to the observational data for the range of spectral indices indicated. Spectra with the higher values of $\Gamma_{s}$ provide a good fit to the three lower energy data points which have the smallest error bars. However, they miss the higher energy data points. Those with lower values of $\Gamma_{s}$ miss the lower energy data point but pass through the four higher energy data points. all of these fits have almost identical reduced $\chi{{}^{2}}$ values of $\sim$ 1.1. Choosing a range of $\Gamma_{s}$ between 1.5 and 3 as shown in Figure 3, and varying the normalization constant, $K$ to find the best fit in each case, it is found that all of these spectral fits have almost identical reduced $\chi{{}^{2}}$ values of $\sim$ 1.1. As a result, all of them are equally probable. Other EBL models exhibit the same degeneracy in $\chi^{2}$ over a range of spectral indeces. ## 5 Hypothetical Consideration of the Luminosity of the 3C279 Flare Observed by MAGIC As discussed above, our statistical analysis shows that the spectral index of the flare is not well determined by the MAGIC data. Given this uncertainty, we can conditionally compare the luminosity of the MAGIC flare with that of an earlier flare observed by EGRET by assuming the same intrinsic spectral index as the EGRET flare. The 1991 EGRET flare had an observed spectral index of $\Gamma_{s}=2.02$ at energies between 50 MeV and 10 GeV (Hartman et al. 1992) where EBL absorption is negligible. We note that this choice of spectral index gives good $\chi^{2}$ fits for all the EBL models considered here. Assuming the value for $\Gamma_{s}$ of 2.02, we obtain the best fits to the MAGIC data using the Levenberg-Marquardt method for the single free parameter, $K$. The $\chi^{2}$ values obtained for these best fits are shown in the last column of Table 2. The resulting curves for the fits of all five models are illustrated in figure 2. If we then choose the form of $\tau(E)$ given by the SMS fast evolution model as input to equation (1), by extending the $\Gamma_{s}=2.02$ power-law fit from Hartman et al. to an energy of $\sim$ 500 GeV, we find a value for $K$ that is $\sim$ 3 times higher for the MAGIC flare than for the flare observed by EGRET 15 years earlier. Thus, we obtain a reasonable estimate for the luminosity of the MAGIC flare that is comparable to that of the earlier flare observed by EGRET. Figure 4: Plot of the best fit curves with $\Gamma_{s}=2.02$ for the SMS fast evolution (solid), SMS baseline (dotted), Kneiske et al. best-fit (short- dashed), Kneiske et al. low-IR (long-dashed), and Primack et al. (dash-dotted) models for a fixed spectral index of $\Gamma_{s}=2.02$ shown along with the observational data from Albert et al. (2008). The shaded region represents the combined systematic and statistical error for the observational data. The gray curve shows the EGRET flare power-law fit extended to higher energies multiplied by $e^{-\tau(E)}$. The respective $\chi^{2}$ values are given in the last column of Table 2. ## 6 Conclusions and Discussion We conclude that the observational data of 3C279 from the MAGIC collaboration do not significantly constrain the intergalactic low energy photon spectra, nor do they indicate that the Universe is more transparent to high energy $\gamma$-rays than previously thought or obtained by the SMS models. Including the effects of the systematic errors, which only significantly affect the two highest energy data points, would only strengthen our conclusions. We show here, as well as in our analysis of 1ES0229 (Stecker & Scully 2008), that the SMS models produce reasonable fits to the observational data. We further demonstrate that other models in the literature that give lower transparencies do not fit the 3C279 data better than the SMS models do. This is because the magnitude and shape of the intrinsic spectrum of the flare are not well determined by the MAGIC data, as we have shown. The spectral energy distribution of the intergalactic IR radiation is well constrained by astronomical data, as discussed in detail in SMS. However, the form of the intrinsic source spectrum of the $\gamma$-ray flare is not well constrained. If one should wish to speculate on producing a fit to the MAGIC data to go through all of the points, one would require an intrinsic flare spectrum which flattens at higher energies. Such a spectrum may arise either from relativistic shock acceleration (Stecker, Baring & Summerlin 2007; Resmi & Bhattacharya 2008) or from intrinsic source absorption (Liu & Bai 2006; Liu, Bai & Ma 2008; Sitarek & Bednarek 2008). However, given both the statistical and the systematic errors of the MAGIC data, particularly those for the highest energy bin, the invocation of such processes is unnecessary. If we apply our analysis technique to the MAGIC data and assume an intrinsic power-law spectrum with index 2.02 as observed for the earlier EGRET flare at lower energies where absorption is negligible, we find that the luminosity of the MAGIC flare was similar to that of the earlier EGRET flare. In fact, the MAGIC flare was $\sim$3 times brighter. This luminosity estimate could not be obtained using the analysis adopted by the MAGIC group, since all of the data points that they started from were affected by intergalactic absorption. One would need to have additional information concerning the intrinsic source spectrum such as additional observations at lower $\gamma$-ray energies that are unaffected by pair-production absorption in order to better constrain the intergalactic IR radiation through an analysis of the resulting optical depth at the higher energies where absorption is more significant. We note that the ideal situation for exploring the exact amount of intergalactic absorption would be to have simultaneously and with overlapping observational energy ranges (1) observations of a strong flare with large photon counts at lower energies unaffected by intergalactic absorption, as can be obtained by FGST (Fermi Gamma-ray Space Telescope, nee GLAST), and (2) ground based sub-TeV observations of such a strong flare with larger photon counts. We wish to thank Rudolf Bock for sending us a list of data on the spectrum of 3C279 observed by MAGIC. STS gratefully acknowledges partial support from the Thomas F. & Kate Miller Jeffress Memorial Trust grant no. J-805. ## References * Albert et al. (2008) Albert, J. et al. 2008, Science 320, 1752 * Altieri et al. (1999) Altieri, B. et al. 1999, A&A 343, L65 * Breit & Wheeler (1934) Breit, G. & Wheeler, J. A. 1934, Phys. Rev. 46, 1087 * Elbaz et al. (2002) Elbaz, D. et al. 2002, A&A 384, 848 * Hartman et al. (1992) Hartman, R. C. et al. 1992, ApJ 385, L1 * Hauser & Dwek (2001) Hauser, M.G. & Dwek, E. 2001 ARA&A 39, 249 * Katarzyńsky et al. (2006) Katarzyńsky, K., Ghisellini, G., Svensson, R., Gracia, J. & Maraschi, L. 2006, MNRAS 368, L52 * Kneiske et al. (2004) Kneiske, T. M. et al. 2004, A&A 413, 807 * Lagache et al. (2004) Lagache,G. et al. 2004, ApJS 154, 112 * Liu & Bai (2006) Liu, H. T. and Bai, J. M. 2006, ApJ 653, 1089 * Liu, Bai, & Ma (2008) Liu, H. T., Bai, J. M. & Ma, L. 2008, arXiv:0807.3133 * Malkan & Stecker (1998) Malkan, M. A. & Stecker, F. W. 1998, ApJ 496, 13 * Malkan & Stecker (2001) Malkan, M. A. & Stecker, F. W. 2001, ApJ 555, 641 * Marziani et al. (1996) Marziani, P. et al. 1996, ApJS 104, 37 * Primack et al. (2005) Primack, J. R., Bullock, J. S. & Somerville, R. S. 2005, AIP Conf. Proc. 745, 23 * (16) Resmi, L. & Bhattacharya, D. 2008, MNRAS 388, 144 * Sitarek & Bednarek (2008) Sitarek, J & Bednarek, W. 2008, arXiv:0807.4228 * Stecker, Baring & Summerlin (2007) Stecker, F. W., Baring, M.G., & Summerlin, E.J. 2007, ApJ 667, L29 * Stecker, de Jager & Salamon (1992) Stecker, F. W., de Jager, O. C. & Salamon, M. H. 1992, ApJ 390, L49 * Stecker, Malkan, & Scully (2006) Stecker, F. W., Malkan, M.A., & Scully S.T. 2006, ApJ 648, 774 * Stecker, Malkan, & Scully (2007) Stecker, F. W., Malkan, M. A. & Scully, S. T. 2007, ApJ 658, 1392 * Stecker, Puget, & Fazio (1977) Stecker, F. W., Puget, J.-L. & Fazio, G. G. 1977, ApJ 214, L51 * Stecker & Scully (2006) Stecker, F. W. & Scully, S. T. 2006, ApJ 652, L9 * Stecker & Scully (2008) Stecker, F. W. & Scully, S. T. 2008, A&A 478, L1 * Totani & Takeuchi (2002) Totani, T. & Takeuchi, T. T. 2002, ApJ 570, 470	arxiv-papers	2008-07-30T15:04:56	2024-09-04T02:48:57.035319	{ "license": "Public Domain", "authors": "Floyd W. Stecker, Sean T. Scully", "submitter": "Floyd Stecker", "url": "https://arxiv.org/abs/0807.4880" }
0807.4892	pchung@mail.chem.sunysb.edu danielewicz@nscl.msu.edu # Three-Dimensional Two-Pion Emission Source at SPS: Extraction of Source Breakup Time and Emission Duration P. Chung Dept of Chemistry, SUNY Stony Brook, Stony Brook, NY 11794, USA P. Danielewicz National Superconducting Cyclotron Laboratory, MSU, East Lansing, MI, USA, Dept of Physics & Astronomy, Michigan State University, East Lansing, MI, USA. the NA49 Collaboration ###### Abstract A model-independent, three-dimensional source function for pion pairs has been extracted from Pb+Pb collisions at $\sqrt{s}_{NN}=17.3$ AGeV. The extracted source exhibits long-range non-Gaussian tails in the directions of the pion- pair net transverse-momentum and of the beam. Comparison with the Therminator model allows for an extraction of the pion source proper breakup time and of emission duration in the collisions. ## 1 Introduction A deconfined phase of nuclear matter is expected to be formed at the high energy densities created in relativistic heavy ion collisions qgp06 . It is expected that the signatures of such a phase are reflected in the space-time extent and shape of particle emission source-functions. Recently, one-dimensional (1D) source imaging techniques brown97 ; brown98 ; brown01 have revealed a non-trivial long-range structure in the two-pion emission source at RHIC chung05 ; chung06 . The origins of this structure are still investigated. The nature of such structures and the potentially useful information they can provide could be revealed, on one hand, by examining the possible presence of such a structure in heavy-ion collisions at intermediate SPS energies and, on the other, by examining the dependence of a structure on the direction within the source. The NA49 Collaboration has investigated Pb+Pb collisions over a wide range of bombarding energies at the CERN SPS during the last decade alt05 , accumulating pion data. The rich data set provides a unique opportunity to search for non-trivial structures at the SPS, investigate their dependence on energy and direction and extract any useful information those structures may provide. In this paper, the first three-dimensional (3D) emission source image for low $p_{T}$ pion pairs produced in central Pb+Pb collisions at $\sqrt{s}_{NN}=17.3$ GeV is presented. The result is compared with calculations from the Therminator model in order to extract the source proper breakup time and emission duration. ## 2 Setup and Analysis ### 2.1 Experimental Setup The data presented here were taken by the NA49 Collaboration during the years 1996-2002. The incident beams of 158 AGeV were provided by the CERN SPS accelerator. The NA49 Large Acceptance Hadron Detector afa99 achieves a large-acceptance precision-tracking ($\delta p/p^{2}\approx(0.3-7).10^{-4}$ (GeV/c$)^{-1}$) and particle identification using time projection chambers (TPC’s). Charged particles are detected by the tracks left in the TPC’s and identified by the energy deposited in the TPC gas. Mid-rapidity particle identification is further enhanced by a time-of-flight wall (resolution of 60 ps). Event centrality is determined by a forward calorimeter which measures the energy of spectator matter. ### 2.2 Three-Dimensional Correlation Function The 3D correlation function, C($\mathbf{q}$), was calculated as the ratio of pair to uncorrelated reference distributions in relative momentum $\mathbf{q}$ for $\pi^{+}\pi^{+}$ and $\pi^{-}\pi^{-}$ pairs. Here, $\mathbf{q}=\frac{(\mathbf{p_{1}}-\mathbf{p_{2}})}{2}$ is half of the momentum difference between the two particles in the pair Center-of-Mass System (PCMS) frame. The pair distribution was obtained using pairs of particles from the same event and the uncorrelated distribution was obtained by pairing particles from different events. The Lorentz transformation of $\mathbf{q}$ from the laboratory frame to the PCMS was done by a Lorentz transformation to the pair Locally Co-Moving System (LCMS) frame along the beam direction followed by a Lorentz transformation to the PCMS along the direction of the transverse momentum of the pair. Track merging and splitting effects were suppressed by appropriate cuts on both the pair and uncorrelated distributions. The pair cuts require the two particles in each pair to be separated by at least 2.2 cm over 50 pad rows in the vertex TPC’s. A 20$\%$ increase in this minimum separation results in the correlation data points fluctuating within the statistical errors. Hence, the systematic uncertainty associated with the pair cuts is deemed smaller than the statistical uncertainty. The effects of track momentum resolution were assessed by jittering the momentum of the tracks in the data by the maximum momentum resolution, $\delta p/p^{2}\approx 7.10^{-4}$ (GeV/c$)^{-1}$, and by re-computing the 3D correlation function. The resulting correlation function incorporates then twice the effect of the momentum resolution and was found to be very close to the raw un-smeared correlation function. This is not surprising, considering that the mean momentum of the tracks used in this analysis is 1.4 GeV/c, resulting in a momentum resolution of $\delta p/p\approx 0.1\%$. The above procedure gives a very small smear to the correlation peak in the region $q<10$ MeV. In practice, this tends to reduce the imaged source intensity at large separation. As a consequence, the results for the source function presented here may slightly underestimate the actual source function at large $r$. ### 2.3 Angular Moments In the Cartesian surface-spherical harmonic decomposition technique daniel05 ; chung05 , the 3D correlation function is expressed as $C(\mathbf{q})-1=R(\mathbf{q})=\sum_{l}\sum_{\alpha_{1}\ldots\alpha_{l}}R^{l}_{\alpha_{1}\ldots\alpha_{l}}(q)\,A^{l}_{\alpha_{1}\ldots\alpha_{l}}(\Omega_{\mathbf{q}})$ (1) where $l=0,1,2,\ldots$, $\alpha_{i}=x,y\mbox{ or }z$, $A^{l}_{\alpha_{1}\ldots\alpha_{l}}(\Omega_{\mathbf{q}})$ are Cartesian harmonic basis elements ($\Omega_{\mathbf{q}}$ is solid angle in $\mathbf{q}$ space) and $R^{l}_{\alpha_{1}\ldots\alpha_{l}}(q)$ are Cartesian correlation moments given by $R^{l}_{\alpha_{1}\ldots\alpha_{l}}(q)=\frac{(2l+1)!!}{l!}\int\frac{d\Omega_{\mathbf{q}}}{4\pi}A^{l}_{\alpha_{1}\ldots\alpha_{l}}(\Omega_{\mathbf{q}})\,R(\mathbf{q})$ (2) where $q$ is the modulus of the 4-vector $\mathbf{q}$. The coordinate axes are oriented so that $z$ is parallel to the beam (long) direction, $x$ points in the direction of the total momentum of the pair in the locally co-moving system (LCMS) (out), and $y$ is perpendicular to the first two axes (side). The correlation moments, for each order $l$, may be calculated from the measured 3D correlation function using Eq. (2), but can be then vulnerable to directional inefficiencies. Alternatively, the moments may be fitted using Eq. (1), avoiding regions of poor efficiency or accounting for those in the fit weights. In following the latter type of analysis, Eq. (1) has been truncated at $l=4$ and expressed in terms of independent moments only. The higher order moments have been found to be negligible. Up to order 4, there are 6 independent moments: $R^{0}$, $R^{2}_{x2}$, $R^{2}_{y2}$, $R^{4}_{x4}$, $R^{4}_{y4}$ and $R^{4}_{x2y2}$ where $R^{2}_{x2}$ is shorthand for $R^{2}_{xx}$ etc. These independent moments were extracted as a function of $q$, by fitting the truncated series to the measured 3D correlation function with the moments as the parameters of the fit. ### 2.4 Source Reconstruction Each independent correlation moment can be imaged using the 1D source imaging code of Brown and Danielewicz brown97 ; brown98 ; brown01 to obtain the corresponding source moment within each order $l$. Both the effects of Bose- Einstein symmetrisation and of Coulomb interaction, the sources of the observed correlations, are accounted for in the source imaging code. Thereafter, the total source function, $S(\mathbf{r})$, can be reconstructed by combining the source moments for each $l$: $S(\mathbf{r})=\sum_{l}\sum_{\alpha_{1}\ldots\alpha_{l}}S^{l}_{\alpha_{1}\ldots\alpha_{l}}(r)\,A^{l}_{\alpha_{1}\ldots\alpha_{l}}(\Omega_{\mathbf{r}})$ (3) Another method of construction of the 3D source function from the 3D correlation function is by fitting the latter directly upon assuming the shape for the source function. Since the 3D correlation function is represented by the Cartesian moments in the Cartesian harmonic decomposition, this amounts to fitting the six independent moments, dependent on $q$, with a trial source function. Two trial functional shapes were considered in this analysis: the 3D Gaussian, frequently termed an ellipsoid shape, and the Hump function, with six adjustable parameters, given by $S(x,y,z)=\lambda\exp[-f_{s}(\frac{r^{2}}{4r_{s}^{2}})-f_{l}(\frac{x^{2}}{4r_{xl}^{2}}+\frac{y^{2}}{4r_{yl}^{2}}+\frac{z^{2}}{4r_{zl}^{2}})]$ (4) where $f_{s}=1/[1+(r/r_{0})^{2})]$ and $f_{l}=1-f_{s}$. ## 3 Results Figure 1 (a) shows a comparison between the 1D correlation function C($q$) and the $l=0$ moment $1+R(q)$ for mid-rapidity low-$p_{T}$ pion-pairs from central Pb+Pb collisions. The functions are in very good agreement with each other as expected in the absence of significant detection inefficiencies with dependence on angle in $\mathbf{q}$-space. The figure also shows that the Hump function (solid line) fits the data very well while the ellipsoid fit (dotted line) underestimates the data at $q<13$ MeV/c. This difference in correlation fits at low $q$ corresponds to a difference in the source function at large $r\gtrsim 15$ fm, as evident in Fig. 1 (b). The Hump shape (solid line) is in good agreement with the source image (squares) whereas the ellipsoid shape (dotted line) underestimates the image. Given that the discrepancy occurs at large $r$, that discrepancy becomes even more pronounced for the radial density in Fig. 1(c). Figure 2 shows the different $l=2$ and $l=4$ moments (open circles) as a function of pion separation. The $l=4$ moments are significantly smaller in magnitude compared to the $l=2$ moments: this justifies the truncation of the series Eq. (1) at $l=4$. The Hump fit (solid line), with $\chi^{2}/\text{ndf}=1.3$, is in close agreement with the data whereas the ellipsoid function (dotted line) gives a poor fit to the data ($\chi^{2}/\text{ndf}=6.8$), as is visually evident. Figure 3 (a)-(c) shows the 3D correlation function profiles in the $x$, $y$ and $z$ directions while the corresponding source function profiles are shown in panels (d)-(f). The source image (squares) is in good agreement with the Hump function fit (solid line), but is underestimated by the ellipsoid fit (dotted line) in all 3 directions. These non-Gaussian tails in the Hump source function profiles have corresponding manifestations in the correlation function profiles at low $q$, relative to the ellipsoid fit. ## 4 Therminator Model ### 4.1 Model Assumptions The Therminator model kis05 can shed more light on the source breakup and emission dynamics. The Therminator model incorporates the following: (1) the Bjorken assumption of longitudinal boost invariance, (2) blast-wave expansion in the transverse direction with transverse velocity profile semilinear in transverse radius $\rho$ kis07 , (3) thermal emission of particles from a cylinder of infinite longitudinal size and finite transverse radius $\rho_{max}$, and (4) all known resonance decays. Assumptions (1) and (2) of the model imply that the expanding source consists of fluid elements shaped like onion rings with their axes aligned with the beam axis. Each fluid element expands transversely with a transverse velocity semi-linear in $\rho$, as well as translates longitudinally with a velocity profile linear in $z$. Each fluid element breaks up after a proper breakup time $\tau$ in its own rest frame, with subsequent particle emission. Bjorken assumption of longitudinal boost invariance implies that the breakup (emission) time, $t$, of a source element at position $z$ in the laboratory frame is given by $t^{2}=\tau^{2}+z^{2}$ . At the point of source breakup, particles within the Therminator model leave each source element defined by their longitudinal and transverse positions: $z$ and $\rho$. All particle emissions are externally viewed as occuring from a freeze-out hypersurface defined in $\rho$-$\tau$ plane as $\tau=\tau_{0}+a\rho$, where $a$ is a parameter which controls the space-time correlation and $\tau_{0}$ is the proper breakup time of the source element located at $\rho=0$, i.e. on the beam axis, and $\tau$ is proper breakup time of the source element located at finite $\rho$. Hence, particles which are emitted from a generic fluid element with coordinates ($z$,$\rho$) will have emission time $t$ in the laboratory frame given by $t^{2}=(\tau_{0}+a\rho)^{2}+z^{2}$ . In this analysis, Therminator is used in the blast-wave mode and $a$ is set to the negative value of $-0.5$ kis06 . Hence, $\tau$ decreases linearly with $\rho$ down to a minimum value at $\rho=\rho_{max}$. The negative value of $a$ implies a negative space-time correlation, i.e. particles at large $\rho$ freeze-out earlier. Hence, the source emits or burns from outside in. Since each source element is defined by only one value of the proper breakup time $\tau$, all particle emissions from this fluid element happen instantaneously in the rest frame of the source element and the proper emission duration in the above scenario is 0. On the other hand, one can allow for a finite non-zero proper emission duration in each source element, if the single proper breakup time $\tau$ of each source element is replaced with a distribution of breakup times. This corresponds physically to the source element breaking up over a finite time interval rather than instantaneously. One such parametrization is an exponential distribution of breakup times $\tau^{\prime}$ with a width $\Delta\tau$ given by $dN/d\tau^{\prime}=\frac{\Theta(\tau^{\prime}-\tau)}{\Delta\tau}\,\text{exp}[-(\tau^{\prime}-\tau)/\Delta\tau]$ . The minimum value of $\tau^{\prime}$ is the initial single breakup time $\tau$ for that source element. With such a distribution of breakup times, each source element emits particles from a family of hypersurfaces, each of which is defined by a $\tau^{\prime}$ value which is sampled according to the exponential distribution. In such a scenario, particle emission from each source element takes place over a finite time duration defined by $\Delta\tau$ rather than instantaneously. In this parametrization, $\Delta\tau$ represents the proper emission duration in the rest frame of the fluid element. ### 4.2 Comparison to Data Figure 4 shows a comparison of the source image (squares) with the calculated source function (circles) from the Therminator model in its blast-wave mode. A good match is obtained with values of $\tau_{0}=7.3$ fm/c and $\Delta\tau=3.7$ fm/ c, when all resonance decays are turned on. The good agreement between extracted source image and calculated source function indicates that the data are consistent with the basic ingredients of Bjorken longitudinal expansion and with the blast-wave dynamics for transverse expansion which are incorporated in the Therminator model. It also indicates that a finite non- zero pion emission duration and outside-in burning of the source are needed to describe the data. ## 5 Conclusions The model-independent 3D source imaging technique, involving decomposition of the 3D correlation function into Cartesian surface-spherical harmonics, has been applied to pion pairs from Pb+Pb collisions at $\sqrt{s}_{NN}=17.3$ GeV. Prominent non-Gaussian tails have been observed in the outward direction of the pion pair transverse momentum and in the longitudinal direction of the beam. The extracted source image is well described by the Therminator model which incorporates Bjorken longitudinal expansion and blast-wave transverse- flow dynamics. The data is consistent with a proper emission duration of 3.7 fm/c for pions at $\sqrt{s}_{NN}=17.3$ GeV. ## References * (1) Ma Y-G, Wang E-K, Cai X, Huang H-Z, Wang X-N and Zhu Z-Y (ed) 2007 J. Phys. G: Nucl. Part. Phys. 34 issue 8 (Proc. QUARK MATTER 2007). * (2) D. A. Brown and P. Danielewicz, Phys. Lett. B 398, 252 (1997). * (3) D. A. Brown and P. Danielewicz, Phys. Rev. C 57, 2474 (1998). * (4) D. A. Brown and P. Danielewicz, Phys. Rev. C 64, 14902 (2001). * (5) P. Chung et al., Nucl. Phys. A 749, 275c (2005). * (6) S. S. Adler et al. (PHENIX Collaboration), Phys. Rev. Lett. 98, 132301 (2007) * (7) C. Alt et al. (NA49 Collaboration), CERN-SPSC-2005-041, CERN-SPSC-P-264-ADD-12, Nov 2005. * (8) S. Afanasiev et al., Nucl. Instrum. Meth. A430 210 (1999). * (9) P. Danielewicz and S. Pratt, Phys. Lett. B 618 60 (2005). * (10) A. Kisiel et al., Comput. Phys. Commun. 174, 669, 2006. * (11) A. Kisiel, Braz. J. Phys. 37, 3A, 917 (2007). * (12) A. Kisiel et al., Phys. Rev. C 73, 064902 (2006). 5mm Figure 1: (color on line) Angle averaged correlation function (top panel) and source function (middle) and radial probability density (bottom). Filled circles show correlations from a direct angle-averaging of the data. Open circles represent correlations from fitting the angular decomposition to the data. Squares show the imaged source and correlations corresponding to the imaged source. The dotted and solid lines represent, respectively, the fitted Gaussian and Hump sources and their corresponding correlation functions (see text). 5mm Figure 2: (color online) Correlation moments of multi-polarity $l=2$ (left panels), and $l=4$ (right panels) for $\pi^{+}\pi^{+}$ and $\pi^{-}\pi^{-}$ pairs. Results are represented in the same way as in Fig. 1, with error bars indicating statistical errors for the data. Systematic errors are smaller than the data points. 5mm Figure 3: (color online) Correlation $C(q)$ (left panels) and source $S(r)$ (right panels) function profiles for $\pi^{+}\pi^{+}$ and $\pi^{-}\pi^{-}$ pairs in the outward $x$ (top panels), sideward $y$ (middle) and longitudinal $z$ (bottom) directions. The use of symbols is analogous to that in Fig. 1. Error bars indicate statistical errors. Systematic errors do not exceed the symbol size. Here, $l=4$ moments make negligible contributions. 5mm Figure 4: Source function profiles, $S(r_{i})$, along $x$ (a), $y$ (b) and $z$ (c) directions, from imaged correlation function (squares) and from the Therminator model (circles).	arxiv-papers	2008-07-30T17:02:02	2024-09-04T02:48:57.039716	{ "license": "Public Domain", "authors": "P. Chung and P. Danielewicz", "submitter": "Paul Chung", "url": "https://arxiv.org/abs/0807.4892" }
0807.4934	# Modeling Galaxy-Galaxy Weak Lensing with SDSS Groups Ran Li1,2, H.J. Mo2, Zuhui Fan1, Marcello Cacciato3, Frank C. van den Bosch3, Xiaohu Yang4, Surhud More3 1Department of Astronomy, Peking University, Beijing 100871, China 2Department of Astronomy, University of Massachusetts, Amherst MA 01003, USA 3Max-Planck Institute for Astronomy, Königstuhl 17, D-69117 Heidelberg, Germany 4Shanghai Astronomical Observatory, the Partner Group of MPA, Nandan Road 80, Shanghai 200030, China E-mail:ranl@astro.umass.edu ###### Abstract We use galaxy groups selected from the Sloan Digital Sky Survey (SDSS) together with mass models for individual groups to study the galaxy-galaxy lensing signals expected from galaxies of different luminosities and morphological types. We compare our model predictions with the observational results obtained from the SDSS by Mandelbaum et al. (2006) for the same samples of galaxies. The observational results are well reproduced in a $\Lambda$CDM model based on the WMAP 3-year data, but a $\Lambda$CDM model with higher $\sigma_{8}$, such as the one based on the WMAP 1-year data, significantly over-predicts the galaxy-galaxy lensing signal. We model, separately, the contributions to the galaxy-galaxy lensing signals from different galaxies: central versus satellite, early-type versus late-type, and galaxies in haloes of different masses. We also examine how the predicted galaxy-galaxy lensing signal depends on the shape, density profile, and the location of the central galaxy with respect to its host halo. ###### keywords: dark matter - large-scale structure of the universe - galaxies: haloes - methods: statistical ††pagerange: Modeling Galaxy-Galaxy Weak Lensing with SDSS Groups–LABEL:lastpage††pubyear: 2008 ## 1 Introduction According to the current paradigm of structure formation, galaxies form and reside inside extended cold dark haloes. While the formation and evolution of dark matter haloes in the cosmic density field is mainly determined by gravitational processes, the formation and evolution of galaxies involves much more complicated, and poorly understood processes, such as radiative cooling, star formation, and all kinds of feedback. One important step in understanding how galaxies form and evolve in the cosmic density field is therefore to understand how the galaxies of different physical properties occupy dark matter haloes of different masses. Theoretically, the connection between galaxies and dark matter haloes can be studied using numerical simulations (e.g., Katz, Weinberg & Hernquist 1996; Pearce et al. 2000; Springel 2005; Springel et al. 2005) or semi-analytical models (e.g. White & Frenk 1991; Kauffmann et al. 1993, 2004; Somerville & Primack 1999; Cole et al. 2000; van den Bosch 2002; Kang et al. 2005; Croton et al. 2006). These approaches try to model the process of galaxy formation from first principles. However, since our understanding of the relevant processes is still poor, the predicted connection between the properties of galaxies and dark matter haloes needs to be tested against observations. More recently, the halo occupation model has opened another avenue to probe the galaxy-dark matter halo connection (e.g. Jing, Mo & Börner 1998; Peacock & Smith 2000; Berlind & Weinberg 2002; Cooray & Sheth 2002; Scranton 2003; Yang, Mo & van den Bosch 2003; van den Bosch, Yang & Mo 2003; Yan, Madgwick & White 2003; Tinker et al. 2005; Zheng et al. 2005; Cooray 2006; Vale & Ostriker 2006; van den Bosch et al. 2007). This technique uses the observed galaxy luminosity function and clustering properties to constrain the average number of galaxies of given properties that occupy a dark matter halo of given mass. Although the method has the advantage that it can yield much better fits to the data than the semi- analytical models or numerical simulations, one typically needs to assume a somewhat ad-hoc functional form to describe the halo occupation model. A more direct way of studying the galaxy-halo connection is to use galaxy groups111In this paper, we refer to a system of galaxies as a group regardless of its richness, including isolated galaxies (i.e., groups with a single member) and rich clusters of galaxies., provided that they are defined as sets of galaxies that reside in the same dark matter halo. Recently, Yang et al. (2005; 2007) have developed a halo-based group finder that is optimized for grouping galaxies that reside in the same dark matter halo. Using mock galaxy redshift surveys constructed from the conditional luminosity function model (e.g. Yang et al. 2003) and a semi-analytical model (Kang et al. 2005), it is found that this group finder is very successful in associating galaxies with their common dark matter haloes (see Yang et al. 2007; hereafter Y07). The group finder also performs reliably for poor systems, including isolated galaxies in small mass haloes, making it ideally suited for the study of the relationship between galaxies and dark matter haloes over a wide range of halo masses. However, in order to interpret the properties of the galaxy systems in terms of dark matter haloes, one needs to know the halo mass associated with each of the groups. One approach commonly adopted is to use some halo mass indicator (such as the total stellar mass or luminosity contained in member galaxies) to rank the groups. With the assumption that the corresponding halo masses have the same ranking and that the mass function of the haloes associated with groups is the same as that given by a model of structure formation, one can assign a halo mass to each of the observed groups. This approach was adopted by Y07 for the group catalogue used in this paper. There are three potential problems with this approach. First, the approach is model- dependent, in the sense that the assumption of a different model of structure formation will lead to a different halo mass function, and hence assign different masses to the groups. Second, even if the assumed model of structure formation is correct, it is still not guaranteed that the mass assignment based on the ranking of group stellar mass (or luminosity) is valid. Finally, even if all groups are assigned with accurate halo masses, the question how dark matter is distributed within the galaxy groups remains open. Clearly, it is important to have independent mass measurements of the haloes associated with galaxy groups to test the validity of the mass estimates based on the stellar mass (luminosity) ranking. Gravitational lensing observations, which measure the image distortions of background galaxies caused by the gravitational field of the matter distribution in the foreground, provide a promising tool to probe the dark matter distribution directly. In particular, galaxy-galaxy weak lensing, which focuses on the image distortions around lensing galaxies, can be used to probe the distribution of dark matter around galaxies, hence their dark matter haloes. The galaxy-galaxy lensing signal produced by individual galaxies is usually very weak, and so one has to stack the signal from many lens galaxies to have a statistical measurement. The first attempt to detect such galaxy- galaxy lensing signal was reported by Tyson et al. (1984). More recently, with the advent of wide and deep surveys, galaxy-galaxy lensing can be studied for lens galaxies of different luminosities, stellar masses, colors and morphological types (e.g. Brainerd et al. 1996; Hudson et al. 1998; McKay et al. 2001; Hoekstra et al. 2003; Hoekstra 2004; Sheldon et al. 2004; Mandelbaum et al. 2005, 2006; Sheldon et al. 2007a; Johnston et al. 2007; Sheldon et al. 2007b; Mandelbaum et al. 2008). Given that galaxies reside in dark matter haloes, these results provide important constraints on the mass distribution associated with galaxies in a statistical way. In this paper, we use the galaxy groups of Y07 selected from the Sloan Digital Sky Survey (SDSS), together with mass models for individual groups, to predict the galaxy-galaxy lensing signal expected from SDSS galaxies. We compare our model predictions with the observational results obtained by Mandelbaum et al. (2006) for the same galaxies. Our goal is threefold. First, we want to test whether the method of halo-mass assignment to groups adopted by Y07 is reliable. Since the method provides a potentially powerful way to obtain the halo masses associated with the galaxy groups, the test results have general implications for the study of the relationship between galaxies and dark matter haloes. Second, we want to examine in detail the contributions to the galaxy-galaxy lensing signal from different systems, such as central versus satellite galaxies, early-type versus late-type galaxies, and groups of different masses. Such analysis can help us interpreting the observational results. Finally, we would like to study how the predicted galaxy-galaxy lensing signal depends on model assumptions, such as the cosmological model and the density profiles of dark matter haloes. In a companion paper (Cacciato et al. 2008, hereafter C08), we use the relationship between galaxies and dark matter haloes obtained from the conditional luminosity function (CLF) modeling (Yang et al. 2003; van den Bosch et al. 2007) to predict the galaxy-galaxy cross correlation and to calculate the lensing signal, while here we directly use the observed galaxy groups and their galaxy memberships. This paper is organized as follows. In Section 2 we define the statistical measure that characterizes the galaxy-galaxy lensing effect expected from the mass distribution associated with the galaxy groups. We provide a brief description of the galaxy group catalogue and the models of the mass distribution associated with galaxy groups in Section 3. We present our results in Section 4 and conclude in Section 5. Unless specified otherwise, we adopt a $\Lambda$CDM cosmology with parameters given by the WMAP 3-year data (Spergel et al. 2007, hereafter WMAP3 cosmology) in our analysis: $\Omega_{\rm m}=0.238$, $\Omega_{\Lambda}=0.762$, and $h\equiv H_{0}/(100\rm{km\,s^{-1}\,Mpc^{-1}})=0.73$, $\sigma_{8}=0.75$ . ## 2 Galaxy-Galaxy Lensing Galaxy-galaxy lensing provides a statistical measure of the profile of the tangential shear, $\gamma_{t}(R)$, averaged over a thin annulus at the projected radius $R$ around the lens galaxies. This quantity is related to the excess surface density (hereafter ESD) around the lens galaxy, $\Delta\Sigma$, as $\Delta\Sigma(R)=\gamma_{t}(R)\Sigma_{\rm crit}=\bar{\Sigma}(<R)-\Sigma(R)\,,$ (1) where $\bar{\Sigma}(<R)$ is the average surface mass density within $R$, and $\Sigma(R)$ is the azimuthally averaged surface density at $R$. Note that, according to this relation, $\Delta\Sigma(R)$ is independent of a uniform background. In the above equation, $\Sigma_{\rm crit}=\frac{c^{2}}{4\pi G}\frac{D_{s}}{D_{l}D_{ls}(1+z_{l})^{2}}$ (2) is the critical surface density in comoving coordinates, with $D_{s}$ and $D_{l}$ the angular distances of the lens and source, $D_{ls}$ the angular distance between the source and the lens, and $z_{l}$ the redshift of the lens. By definition, the surface density, $\Sigma(R)$, is related to the projection of the galaxy-matter cross-correlation function, $\xi_{\rm g,m}(r)$, along the line-of-sight. In the distant observer approximation $\Sigma(R)=\bar{\rho}\int^{\infty}_{-\infty}\left[1+\xi_{\rm g,m}(\sqrt{R^{2}+\chi^{2}})\right]\,d\chi\,,$ (3) where $\bar{\rho}$ is the mean density of the universe and $\chi$ is the line- of-sight distance from the lens. The cross-correlation between galaxies and dark matter can, in general, be divided into a 1-halo term and a 2-halo term. The 1-halo term measures the cross-correlation between galaxies and dark matter particles in their own host haloes, while the 2-halo term measures the cross-correlation between galaxies and dark matter particles in other haloes. In the present work, we are interested in the lensing signals on scales $R\leq 2h^{-1}{\rm Mpc}$ where the observational measurements are the most accurate. As we will show in § 4, on such scales the signal is mainly dominated by the 1-halo term. Nevertheless, our model also takes the contribution of the 2-halo term into account. More importantly, since central galaxies (those residing at the center of a dark matter halo) and satellite galaxies (those orbiting around a central galaxy) contribute very different lensing signals (e.g. Natarajan, Kneib & Smail 2002; Yang et al. 2006; Limousin et al. 2007), it is important to model the contributions from central and satellite galaxies separately. As an illustration, in Fig. 1 we show the ESDs expected from a single galaxy in a host halo of mass $10^{14}h^{-1}{\rm M}_{\odot}$. The solid line represents the lensing signal expected for the central galaxy of the halo. While the dotted and dashed lines show the lensing signal of a satellite galaxy residing in a sub-halo of $10^{11}h^{-1}{\rm M}_{\odot}$ with a projected halo-centric distance $r_{p}=0.2\,h^{-1}{\rm Mpc}$ and $r_{p}=0.4\,h^{-1}{\rm Mpc}$ from the center of the host halo, respectively. In the calculation, the dark matter mass distribution in the host halo is assumed to follow the Navarro, Frank & White (1997) profile and that in the sub-haloes is assumed to follow the Hayashi et al. (2003) model. These models are described in detail in § 3.3. In order to estimate the ESD, we sample these profiles with mass particles and project the positions of all particles to a plane perpendicular to the line of sight. The $\Sigma(R)$ is then estimated by counting the number of dark matter particles in a annulus with radius $R$ centred on the selected galaxies. Fig. 1 shows clearly that the lensing signals of the central and the satellite are quite different. The ESD of the central galaxy follows the mass distribution of the host halo, decreasing monotonically with $R$. The ESD of a satellite, on the other hand, consists of two parts: one from the subhalo associated the satellite, which contributes to the inner part, and the other from the host halo, which dominates at larger $R$. This simple model demonstrates clearly that, in order to model the galaxy-galaxy lensing signal produced by a population of lens galaxies, one needs to model carefully the distribution of matter around both host haloes and subhaloes. To do this, we need not only to identify the haloes in which each lens galaxy resides, but also to model the mass and density profile of each host halo and subhalo. In addition we also need to model the distribution of dark matter relative to galaxies. In the following section, we describe our modeling with the use of observed galaxy groups. Figure 1: The ESD expected for a single galaxy. Here the host halo mass is assumed to be $10^{14}h^{-1}M_{\odot}$. The solid line represents the lensing signal for the central galaxy in such a halo. The dotted line represents the lensing signal of a satellite galaxy residing in a sub-halo of mass $10^{11}h^{-1}M_{\odot}$ which has a projected distance $r_{\rm p}=0.2h^{-1}{\rm Mpc}$ from the center of the host halo. The dashed line is the same as the dotted line, except that the subhalo’s projected distance is $0.4h^{-1}{\rm Mpc}$ from the center of the host halo. ## 3 Modeling the Mass Distribution Associated with the SDSS Groups ### 3.1 The SDSS Group Catalogue Our analysis is based on the SDSS galaxy group catalogue constructed by Y07. The groups are selected with the adaptive halo-based group finder developed by Yang et al. (2005), from the New York University Value Added Galaxy Catalog (NYU-VAGC; Blanton et al. 2005) which is based on the SDSS Data Release 4 (Adelman-McCarthy et al. 2006). Only galaxies with redshifts in the range $0.01\leq z\leq 0.2$, and with redshift completeness $\mathcal{C}>0.7$, are used in the group identification. The magnitudes and colors of all galaxies are based on the standard SDSS Petrosian technique (Petrosian 1976; Strauss et al. 2002), and have been corrected for galactic extinction (Schlegel, Finkbeiner & Davis 1998). All magnitudes have been K-corrected and evolution- corrected to $z=0.1$ following the method described in Blanton et al. (2003). In Y07, three group samples were constructed using galaxy samples of different sources of galaxy redshifts. Our analysis is based on Sample II, which includes 362,356 galaxies with redshifts from the SDSS and 7091 galaxies with redshifts taken from alternative surveys: 2dFGRS (Colless et al. 2001), PSOz (Saunders et al.2000) or from the RC3 (de Vaucouleurs et al. 1991). There are in total 301,237 groups, including those with only one member galaxy. The group finder has been applied to mock catalogue to test the completeness and purity of the groups in Y07. About $90\%$ of the groups have a completeness $f_{c}>0.6$ and $80\%$ groups with $f_{c}>0.8$, where $f_{c}$ is defined as the ratio between the number of true members that are selected as the members of the group and the number of the total true members of the group. ### 3.2 Halo Mass Assignment Figure 2: The halo mass $M_{S}$ (estimated using stellar mass), versus $M_{\rm\ast}$ (lower panels) and $L$ (upper panels) of the galaxies in the haloes. The left panels are for central galaxies and the right panels are for satellite galaxies. Figure 3: The distribution of the host halo masses for the central and satellite galaxies in different luminosity bins, as indicated by the $r$-band absolute-magnitude range in each panel. An important aspect of the group catalog construction is the determination of the halo mass, $M_{\rm vir}$, of each group. In Y07, two estimators are adopted. The first, $M_{L}$, is estimated using the ranking of the characteristic luminosity of a group, which is the total luminosity of all member galaxies in the group with $M_{r}-5\log h\leq-19.5$ (hereafter referred to as $L_{19.5}$). The second, $M_{S}$, is estimated using the ranking of the characteristic stellar mass, $M_{\rm stellar}$ which is defined to be the total stellar mass of group members with $M_{r}-5\log h\leq-19.5$. For each galaxy the stellar mass is estimated from its absolute magnitude and color using the fitting formula given by Bell et al. (2003). The basic assumption of the ranking method is that there is a one-to-one relation between $M_{\rm stellar}$ (or $L_{19.5}$) and the group mass. Using the dark matter halo mass function predicted by a model of structure formation, one can assign a halo mass to each group according to its $M_{\rm stellar}$ \- ranking (or $L_{19.5}$ \- ranking). In this paper, we use the mass function obtained by Warren et al. (2006). Note that this one-to-one mapping is applicable only when the group sample is complete. In Y07, three complete samples are constructed in three redshift ranges. Only groups in the complete samples are used in the ranking. The mass of other groups are estimated by a linear interpolation based on the $M_{\rm stellar}$-$M_{\rm vir}$ relation (or the $L_{19.5}$ \- $M_{\rm vir}$) obtained from the complete sample. Detailed tests using mock galaxy redshift samples have shown that the 1-$\sigma$ error of the estimated halo mass is $\sim 0.3$ dex (Y07). In addition, the two mass estimators, $M_{L}$ and $M_{S}$, agree remarkably well with each other, with a scatter that decreases from about 0.1 dex at the low- mass end to about 0.05 dex at the high-mass end. Since the correlation between $M_{\rm stellar}$ and halo mass is somewhat tighter than that between $L_{19.5}$ and halo mass, we adopt $M_{S}$ as our fiducial halo mass throughout. As we demonstrate in § 4.3, using $M_{L}$ instead yields results that are fairly similar. Fig. 2 shows the relation between the host halo mass, $M_{S}$, and the galaxy stellar mass $M_{\ast}$ (the lower two panels) or the galaxy luminosity $L$ (the upper two panels). Results are shown separately for central galaxies (left panels) and satellite galaxies (right panels). As one can see, the stellar mass (luminosity) of central galaxies is quite tightly correlated with their host halo masses. However, for satellite galaxies of a given stellar mass (or luminosity), their host halo mass covers a very large range, reflecting the fact that many low-mass galaxies are satellites in massive haloes. The distributions of host halo masses, $M_{S}$, for central or satellite galaxies in different luminosity bins are shown in Fig. 3. On average, brighter central galaxies reside in more massive haloes. For faint galaxies, the halo-mass distribution is broader, again because many faint galaxies are satellites in massive systems. In the group catalogue, the mass assignment described above is used only for groups where the brightest galaxy is brighter than $M_{r}-5\log h=-19.5$. This is because the mass ranking used in the group catalog is based on the total stellar mass (or total luminosity) of the member galaxies that are brighter than $M_{r}-5\log h=-19.5$. The groups with no galaxies brighter than this magnitude thus have no assigned rank. As described in Y07, the reason for choosing this maginitude is a compromise between having a complete sample in a relatively large volume and having more groups that are represented by a number of member galaxies. For groups in which all member galaxies have $M_{r}-5\log h>-19.5$, a different method has to be adopted. In modeling the luminosity function and stellar mass function of the central galaxies based on the same SDSS group catalogue as used here, Yang et al. (2008) obtain an average relation between the luminosity (or stellar mass) of the central galaxy and the halo mass down to $M_{r}-5\log h\sim-17$. We adopt this relation to assign halo masses to all groups (including those containing only one isolated galaxy) represented by centrals with $M_{r}-5\log h>-19.5$. For convenience, the halo masses obtained in this way are also referred to as $M_{S}$ (based on the stellar mass of central galaxies) and $M_{L}$ (based on the $r$-band luminosity of the central galaxies), respectively. ### 3.3 Mass Distribution in Haloes and Subhaloes With the group catalogue described above, we can model the dark matter distribution by convolving the halo distribution with the density profiles of individual haloes. In our modeling of the density profiles, the host halo of a group is assumed to be centered on the central galaxy. There are two ways to define a central galaxy: one is to define the central in a group to be the galaxy with the highest stellar mass, and the other is to define the central to be the brightest member. For most groups these two definitions give the same results, but there are very few cases (less than $\sim 2\%$) where different central galaxies are defined. In our fiducial model, we define the most massive galaxies (in term of stellar mass) to be the central galaxies. In a hierarchical model, a dark matter halo forms through a series of merger events. During the assembly of a halo, most of the mass in the merging progenitors is expected to be stripped. However, some of them may survive as subhaloes, although the total mass contained in subhaloes is small, typically $\sim 10\%$ (van den Bosch, Tormen & Giocoli 2005). Some of the subhaloes are associated with ‘satellite galaxies’ in a halo. In our modeling of the galaxy- galaxy weak lensing, we only take into account subhaloes associated with satellite galaxies, treating other subhaloes as part of the host halo. Giocoli, Tormen & van den Bosch (2008) provide a fitting function of the average mass function for subhaloes at the time of their accretion into the parent halo of a given mass. Using this mass function, we first sample a set of masses for each group mass. We then set the mass originally associated with a satellite galaxy according to the stellar mass ranking of the satellites in the group. Here we implicitly assume that the initial subhalo mass function is the same as the mass function of the subhalos that host satellites. This assumption is not proved by any observations, and we have to live with it since a more realistic model is not currently available. Fortunately, subhalos only contribute a small fraction to the total lensing signal on small scales. The uncertainty here will not have a significant impact on any of our conclusions. To obtain the final mass in the subhalo at the present time, the evolution of the subhaloes needs to be taken into account. In other words, we need to know the fraction of the mass that is stripped and how the structure of a subhalo changes after the stripping. Here it is convenient to introduce a parameter $f_{m}$ which is the retained mass fraction of the subhalo. Gao et al. (2004) studied the radial dependence of the retained mass fraction $f_{m}$ from a large sample of subhaloes in a large cosmological simulation. In their work, $f_{m}$ is considered as a function of $r_{s}/r_{\rm vir,h}$, where $r_{s}$ is the distance of the subhalo from the center of the host halo and $r_{\rm vir,h}$ is the virial radius of the host halo. The simulation of Gao et al. gives $f_{m}=0.65(r_{s}/r_{\rm vir,h})^{2/3}\,.$ (4) We will adopt this in our modeling of the masses associated with subhaloes. However, in the group catalogue, only the projected distance, $r_{p}$, from the group center is available. The 3D-distance, $r_{s}$, is obtained by randomly sampling the NFW profile of the host halo with the given projected radius $r_{p}$. Thus, the mass assigned to a subhalo is determined by the following three factors: (1) the stellar mass of the satellite galaxy; (2) the host halo mass; (3) the distance between the satellite and the center of the host. Here the host halo mass comes into our calculation in two ways. It not only determines the subhalo mass function, but also affects the parameter $f_{m}$ in Eqs.4. Note that the accretion history of the host halo may also affect the value of $f_{m}$. We have to neglect such effect because it is unclear how to model the accretion histories for individual groups. For host haloes, we use the following NFW profile (Navarro, Frank & White 1997) to model the mass distribution: $\rho(r)=\frac{\delta_{0}\bar{\rho}}{(r/r_{c})(1+r/r_{c})^{2}}\,.$ (5) where $\bar{\rho}$ is the mean density of the universe, $r_{c}$ is a scale radius, related to virial radius $r_{\rm vir}$ by the concentration, $c=r_{\rm vir}/r_{c}$, and $\delta_{0}$ is a characteristic over-density related to the average over-density of a virialized halo, $\Delta_{\rm vir}$, by $\delta_{0}=\frac{\Delta_{\rm vir}}{3}\frac{c^{3}}{\ln(1+c)-c/(1+c)}\,.$ (6) We adopt the value of $\Delta_{\rm vir}$ given by the spherical collapse model (see Nakamura & Suto 1997; Henry 2000). Numerical simulations show that halo concentrations are correlated with halo mass, and we use the relations given by Macciò et al. (2007), converted to our definition of halo mass. Note that here we use $r_{c}$, instead of the conventional notation $r_{s}$, to denote the scale radius of the NFW profile, as $r_{s}$ has been used to denote the distance of a subhalo from the center of its host. For sub-haloes, we model their density profiles using the results obtained by Hayashi et al. (2003), who found that the density profiles of stripped sub- haloes can be approximated as $\rho_{s}(r)=\frac{f_{t}}{1+(r/r_{t,\rm eff})^{3}}\rho(r)\,,$ (7) where $f_{t}$ is a dimensionless factor describing the reduction in the central density, and $r_{t,\rm eff}$ is a cut-off radius imposed by the tidal force of the host halo. For $f_{t}=1$ and $r_{t,\rm eff}\gg r_{c}$, $\rho_{s}(r)$ reduces to the standard NFW profile $\rho(r)$. Here $\rho(r)$ is calculated using the mass of the subhalo at the time of its accretion into the host halo. Both $f_{t}$ and $r_{t,\rm eff}$ depend on the mass fraction of the sub-halo that remains bound, $f_{m}$. Based on $N$-body simulations, Hayashi et al. obtained the following fitting formulae relating $f_{t}$ and $r_{t,\rm eff}$ to $f_{m}$: $\log(r_{t,\rm eff}/r_{c})=1.02+1.38\log f_{m}+0.37(\log f_{m})^{2}\,;$ (8) $\log(f_{t})=-0.007+0.35\log f_{m}+0.39(\log f_{m})^{2}+0.23(\log f_{m})^{3}\,.$ (9) It should be pointed out, though, that there are substantial uncertainties in modeling the mass distribution around individual satellite galaxies. In particular, many of the results about subhaloes are obtained from $N$-body simulations, and it is unclear how significant the effect of including baryonic matter is. Fortunately, the total mass associated with satellite galaxies is small (see e.g. Weinberg et al. 2008). Furthermore, the contribution of the subhaloes associated with the satellite galaxies to the galaxy-galaxy lensing signal is confined to small scales. We therefore expect that these uncertainties will not change our results significantly. With the mass distributions described above, we use a Monte-Carlo method to sample each of the profiles with a random set of mass particles. Note that the halo mass assigned to a group in the SDSS Group Catalog is $M_{180}$, which is the mass enclosed in the radius, $r_{180}$, defined such that $M_{180}={4\pi}r_{180}^{3}(180\bar{\rho})/3$. We therefore sample the particle distribution within $r_{180}$. After all the particles in each halo are sampled, we project the positions of all the particles to a plane and calculate $\Delta\Sigma(R)$ by stacking galaxies in each of the luminosity bin. Since the mass distribution is isotropic, an arbitrary direction can be chosen for the stacking. Thus, the projection effect is naturally included in our calculation. Each of the particles has a mass of $10^{10}h^{-1}{\rm M}_{\odot}$. Our test using particles of lower masses shows that the mass resolution adopted here is sufficient for our purpose. Using 2 times more particles leads to a difference of about $5\%$ at $R\sim 0.02\,h^{-1}{\rm Mpc}$, and almost no difference at $R\geq 0.1\,h^{-1}{\rm Mpc}$. Figure 4: Comparison of the lensing signal predicted by the fiducial model with the observational results. Here the ESD is plotted as the function of the transverse distance $R$ for lensing galaxies in different luminosity bins. Data points with error bars are the observational results of Mandelbaum et al. , while the lines are the model predictions. The dotted, dashed and dot-dashed lines represent the contributions of the ‘1-halo term’ of central galaxies, the ‘1-halo term’ of satellite galaxies, and the ‘2-halo term’ (of both centrals and satellites), respectively. The solid lines show the predicted total ESD. The $r$-band magnitude range for each case can be found in Table 1. Figure 5: The contribution to the ESD plotted separately for dark matter haloes of different masses. In each panel, the dotted line shows the contribution from haloes with $M_{S}\geq 10^{14}h^{-1}M_{\odot}$. The dashed line shows the contribution from haloes with $10^{13}h^{-1}M_{\odot}\leq M_{S}<10^{14}h^{-1}M_{\odot}$, and the dot-dashed line shows the contribution from haloes with $M_{S}<10^{13}h^{-1}M_{\odot}$. The solid line shows the total lensing signal predicted by the fiducial model. For comparison, the observational data are included as data points with error-bars. Figure 6: The right panels show the ESD of early galaxies in different luminosity bins, while the left panels show the results for late galaxies. The data points with error-bars show the observational results. The model predictions of the ESD using stellar mass as halo mass indicator are shown as the solid lines. For comparison, the dashed lines show the corresponding model predictions using $M_{L}$ as the halo masses. Figure 7: The dependence of the predicted $\Delta\Sigma(R)$ on various model parameters. For comparison, the observational data are included as data points with error-bars. The left column shows the dependence on halo concentration: the halo concentrations in CL (dotted line) and CH (dashed line) are assumed to be 1/2 and 2 times those in the fiducial model (solid line), respectively. The middle column of shows the effects of halo center offset. The dotted line and the dashed line show the results of models Dev1 and Dev2 model, respectively, while the solid line is the fiducial model. The right column shows the effect of assuming triaxial halo density profile. The dotted line shows the result of model TRI, while the solid line again shows the fiducial model. Both the observation and model predicted $\Delta\Sigma(R)$ are normalized by the fiducial prediction. Figure 8: The model prediction of ESD assuming different cosmological models. The solid and dotted lines show the fiducial model and model using WMAP1 parameters, respectively . The observational data are plotted as data points with error-bars. ## 4 Results ### 4.1 SDSS Lensing Data Before presenting our model predictions, we first describe the observational results that we will use for comparison. The observational results to be used were obtained by Mandelbaum et al. (2006), who analyzed the galaxy-galaxy lensing effects using galaxies in a sample constructed from the SDSS DR4 spectroscopic sample. Their sample of lensing galaxies is similar to the galaxy sample used in Y07 to construct the group catalogue used here. The only difference is that Mandelbaum et al’s sample includes all galaxies with redshifts in the range $0.02<z<0.35$, while the galaxies in Y07’s group catalogue are in $0.01\leq z\leq 0.2$. Since, as to be described below, we are interested in the lensing signals around galaxies of given luminosity and morphological type, this difference in redshift range is not expected to have a significant impact on our results. For the faint luminosity bins, our galaxy samples should be almost identical to that of Mandelbaum et al. (2006), because all faint galaxies are at $z\leq 0.2$ in the SDSS catalog. For bright galaxies we also expect the statistic properties of the two samples to be similar. Both Mandelbaum et al. (2006) and Y07 applied similar evolution correction and K correction, so that the evolution in the galaxy population has been taken into account, albeit in a simple way. As we will see, even for the the two brightest bins, the lensing signal is dominated by halos with masses $\sim 10^{14}\,h^{-1}{\rm M}_{\odot}$, and the change in the halo mass function around this mass is small between $z=0.2$ and $0.35$. Following Mandelbaum et al. (2006), we split the galaxy sample into 7 subsamples according to galaxy luminosity. Table 1 shows the properties of these subsamples: the luminosity range covered by each subsample, the mean redshift, the mean luminosity, and the fraction of late-type galaxies. As expected the mean redshifts of brightest bins are different from the corresponding redshifts in Mandelbaum et al. Table 1: The properties of galaxy samples. In each case, the absolute-magnitude range, the mean redshift, the mean luminosity, and the fraction of late-types are listed. Note that $L_{}=1.2\times 10^{10}h^{-2}L_{\odot}$ Sample \| $M_{r}$ \| $\langle z\rangle$ \| $\langle L/L_{}\rangle$ \| $f_{\rm late}$ ---\|---\|---\|---\|--- L1 \| $-18<M_{r}<-17$ \| 0.031 \| 0.075 \| 0.81 L2 \| $-19<M_{r}<-18$ \| 0.048 \| 0.191 \| 0.70 L3 \| $-20<M_{r}<-19$ \| 0.074 \| 0.465 \| 0.54 L4 \| $-21<M_{r}<-20$ \| 0.111 \| 1.13 \| 0.35 L5f \| $-21.5<M_{r}<-21$ \| 0.145 \| 2.09 \| 0\. 22 L5b \| $-21.5<M_{r}<-22$ \| 0.150 \| 3.22 \| 0.12 L6f \| $-22<M_{r}<-22.5$ \| 0.152 \| 5.01 \| 0.04 Also following Mandelbaum et al. (2006), we split each galaxy subsample into two according to galaxy morphology. The separation is made according to the parameter $frac\\_{\rm dev}$ generated by the PHOTO pipeline. The value of $frac\\_{\rm dev}$ is obtained by fitting the galaxy profile, in a given band, to a model profile given by $frac\\_{\rm dev}\times F_{\rm deV}+(1-frac\\_{\rm dev})\times F_{\rm exp}$, where $F_{\rm deV}$ and $F_{\rm exp}$ are the de Vaucouleurs and exponential profiles, respectively. As in Mandelbaum et al., we use the average of $frac\\_{\rm dev}$ in the $g$, $r$ and $i$ bands. Galaxies with $frac\\_{\rm dev}\geq 0.5$ are classified as early-type, while those with $frac\\_{\rm dev}<0.5$ as late-type. It should be pointed out that we did not carry a ray-tracing simulation to predict the galaxy-galaxy lensing results. Instead, we directly calculate the excess surface density around SDSS galaxies. Thus, our calculation doesn’t deal with source galaxies. On the other side, in Mandelbaum et al. (2006) the source galaxies are carefully weighted, and lensing signals are calibrated, to reduce any bias in the observational measurements (see Mandelbaum et al. 2005 for details). Thus, we assume the observational results are unbiased, and compare them directly with our model predictions. The observation data used here was kindly provided by R. Mandelbaum. The data used in Fig. 6 has been published in Mandelbaum et al. (2006) where lensing signal is calculated for early and late type galaxies separately. R. Mandelbaum also provided the lensing data combining early and late type galaxies for us to make the comparisons presented in all other figures. Note that here we only show the comparison of the lensing signal for galaxies divided according to luminosity. The errorbars on the observational points are $1\sigma$ statistical error. The systematic error of the galaxy-galaxy lensing has been discussed in detail in Mandelbaum et al. (2005). The test has been carried out for three source samples: $r<21$, $r>21$, and high redshift LRGs. The overall systematic error is found to be comparable to or slightly larger than the statistical error shown here. ### 4.2 The Fiducial Model In Fig. 4 we show the lensing signal around galaxies in different luminosity bins obtained from our fiducial model, which has model parameters as described in the last section and assumes the WMAP3 cosmology. Here the ESD is plotted as a function of the projected distance $R$ from galaxies. The solid line shows the averaged ESD of all galaxies in the corresponding luminosity bin. The amplitude of the predicted ESD increases with galaxy luminosity, reflecting the fact that brighter galaxies on average reside in more massive haloes, as shown in Figs. 2 and 3. These results are to be compared with the data points which show the observational results obtained by Mandelbaum et al. (2006) for the same luminosity bins. Overall, our fiducial model reproduces the observational data reasonably well, especially for bright galaxy bins where the observational results are the most reliable. The reduced $\chi^{2}$ is $3.2$ combining all the luminosity bins. The best match is for L5f, with a reduced $\chi^{2}$ of $0.9$. Given that we do not adjust any model parameters, the $\chi^{2}$ indicates a good agreement. For the three low-luminosity bins, the predicted ESD is lower than the corresponding observational result. For L1 and L2, the observational data are very uncertain. For L3, if we take the observational data points at face-value, the discrepancy with the model prediction is significant. As described in § 3.2, for groups which do not contain any member galaxies with $M_{r}-5\log h<-19.5$, their halo masses are not obtained from the ranking of $M_{\rm stellar}$, but from the average stellar mass-halo mass relation of central galaxies that is required to match the observed stellar mass function of central galaxies. While all the galaxies in the bright luminosity bins have their host halo mass assigned by ranking method, the fraction of the galaxies in halos that have masses assigned according to the mass-halo mass relation is about $30\%$ in L3 and about $70\%$ in L2 and L1. It is possible that this relation underestimates the halo mass. In order to see the effect caused by such uncertainties, we have used a set of parameters from Yang et al. (2008) that are still allowed by the observed stellar mass function but give larger halo masses to the hosts of faint central galaxies. This increases the predicted ESD for L3 by $\sim 20\%$, not sufficient to explain the discrepancy. Indeed, this discrepancy is not easy to fix. In the observational data, the amplitudes of the ESD for L3 at $R\sim 0.3$ \- $0.9h^{-1}{\rm Mpc}$ are actually slightly higher than for the brighter sample L4, while in our model the ESD for L3 is always lower than that for L4. There is an effect that may help to reduce the discrepancy between our model prediction and the observational results for low-mass galaxies. Since the group catalogue is only complete down to certain halo mass limit at different redshift (see Y07), additional assumptions have to be made in order to model the distribution of the haloes below the mass limit. In the model described above, we have assumed that the haloes below the mass limit have a random distribution, so that they do not contribute to the ESD. However, in reality these low-mass haloes are correlated with the more massive ones. As a result, our assumption will underestimate the 2-halo term of the ESD. In order to understand how the predicted ESD is produced, we also show separately the contributions from different sources. The dotted lines show the ESD contributed by the 1-halo term of central galaxies. For all the luminosity bins, this term dominates the ESD at small $R$. For the brightest two bins, this term dominates the ESD over the entire range of $R$ studied. This reflects the fact that almost all the galaxies in these two bins are central galaxies and the haloes in which they reside are more extended. The dashed lines show the ESD contributed by the 1-halo term of satellite galaxies. This term first decreases with $R$ and then increases to a peak value before declining at large $R$. This owes to the fact that, in the inner part, the lensing signal produced by satellite galaxies is dominated by the subhaloes associated with them, while at larger $R$ the lensing signal produced by satellite galaxies is dominated by their host halos. The value of $R$ at which the ESD reaches the minimum corresponds roughly to the average halo-centric distance of the subhaloes in the luminosity bin. Note that the contribution of satellites to the total ESD is only important at large $R$ in the low- luminosity bins. This reflects the fact that a significant fraction of the low-luminosity satellites reside in massive haloes. Note that although the 1-halo satellite term dominates the lensing signal at $R\geq 0.3$$h^{-1}{\rm Mpc}$ for faint luminosity bins, the discrepancy between the observation and our prediction cannot be simply solved by boosting up the satellite contribution. The reason is that the increase of the 1-halo satellite term requires the increase of the host halo mass of the satellites, which will make the 1-halo central term of the bright bins increase as well, causing significant discrepancy for the bright bins. Finally, the dash-dotted lines represent the contribution of the 2-halo term. As expected, this term is relevant only on relatively large scales. In our model this term never dominates at $R\leq 2\,h^{-1}{\rm Mpc}$. However, as discussed before this term may be underestimated here. In C08, the 2-halo term is found to be comparable to the 1-halo satellite term even at $R\sim 0.3h^{-1}{\rm Mpc}$. Unfortunately the 2-halo term in C08 may be overestimated because halo- exclusion effect is not properly included. The large fluctuations seen in L6f are due to the small number of galaxies in this luminosity bin. Since the halo mass of each lensing galaxy is known in our model, we can also examine the contributions to the total ESD in terms of the halo mass. Fig. 5 shows the results where the host haloes are split into three bins of $M_{S}$: $M_{S}\geq 10^{14}h^{-1}{\rm M}_{\odot}$ (dotted lines); $10^{13}h^{-1}{\rm M}_{\odot}\leq M_{S}<10^{14}h^{-1}{\rm M}_{\odot}$ (dashed lines); and $M_{S}<10^{13}h^{-1}{\rm M}_{\odot}$ (dashed-dotted lines). As one can see, the lensing signals in brighter bins are dominated by more massive haloes. Very massive haloes with $M_{S}\geq 10^{14}h^{-1}{\rm M}_{\odot}$ are not the dominant contributor, even for the brightest luminosity bin considered here, because the total number of galaxies hosted by such haloes are relatively small. For the low-luminosity bins, relatively massive haloes dominate the ESD at large $R$, because a significant fraction of the low-luminosity satellites are hosted by massive haloes (see Figs. 2 and 3). ### 4.3 Dependence on Galaxy Type In Fig. 6 we present the results separately for early-type and late-type galaxies. For a given luminosity bin, the predicted ESD has a higher amplitude for early-type galaxies, clearly due to the fact that early-type galaxies are more likely to reside in massive haloes (see e.g. van den Bosch, Yang & Mo 2003). For the faint samples, L1 and L2, the behavior of the predicted ESD for early-type galaxies resembles that of satellite galaxies in these luminosity bins, while the predicted ESD for the late-type galaxies looks like of the central galaxies in the corresponding luminosity bins. This, again, reflects the fact that faint early-type galaxies are mostly satellites in massive haloes, while the faint late-type population is dominated by the central galaxies in low-mass haloes. The dashes lines in Fig. 6 show the results obtained using $M_{L}$ as halo mass, rather than $M_{S}$ (see § 3.2). For early-type galaxies, the results based on this halo mass estimate are very similar to those based on $M_{S}$. However, for the late-type samples, the ESDs obtained using $M_{L}$ are significantly higher than those obtained using $M_{S}$, especially for the brighter samples. This is mainly due to the fact that late-type galaxies contain significant amounts of young stars, so that their stellar mass-to- light ratios are relatively low. Consequently, they are assigned a larger halo mass based on their luminosity than based on their stellar mass. The model predictions are compared with the observational results of Mandelbaum et al. (2006). Here again, the model prediction based on $M_{S}$ matches the observational data for the four bright samples. For the three faint bins, the model predictions are again lower than the observational results. As shown in Fig. 6, it seems that our model prediction agrees better with the observation for the late type galaxies. For example, for L3 our model prediction matches the observation reasonably well for the late-type subsample (with the reduced $\chi^{2}$ equal to $1.6$), while the discrepancy is quite large for the early-type subsample (with reduced $\chi^{2}$ equal to $5.3$). Since faint, early-type galaxies are preferentially satellite galaxies in relatively massive halos while faint, late-type galaxies are mostly central galaxies in relatively low mass halos, the above results seem to indicate that the discrepancy is due to the underestimate of the 1-halo satellite term in the model. Unfortunately, there is no simple modification of the model that can fix the discrepancy. Since the spatial distribution of galaxies is fixed by the observation, the only change that can be made is in halo properties. As we will see in the following subsection, increasing the halo concentration even by a factor of two can only increase the predicted ESD by about 20% for the L3 sample, insufficient to explain the discrepancy. An increase in the halo masses assigned to galaxy groups can reduce the discrepancy for faint galaxies, but it would also significantly over-predict the ESD for bright galaxies. The other possibility is that the halo masses of isolated, faint early-type galaxies are significantly underestimated. For example, isolated early-type galaxies may reside in much more massive halos than that given by their stellar masses. In order to explain the discrepancy, the halo masses for these galaxies need to be larger by a factor of at least 3. This will not affect significantly the prediction for bright galaxies, but can boost the prediction for L3 by 50% at $R\sim 0.2h^{-1}{\rm Mpc}$. Unfortunately, it is still unclear if the required halo mass increase is feasible in current models of galaxy formation. ### 4.4 Dependence on other Model Parameters The ESD signal predicted by the model outlined in § 4.2 is based on several assumptions. Beside the underlying cosmology and the halo mass assignment which are the most crucial model ingredients (see § 4.5 and § 4.6), the model requires a concentration-halo mass relation and it assumes that central galaxies reside at rest at the centre of a spherical dark matter halo. In this subsection, we test how our results are affected by these assumptions. The concentration parameter is a measure of the amount of dark matter in the central regions of the haloes. Accordingly, different models for the concentration-halo mass relations are expected to result in different predictions for the ESD signal (at least on small scales). The fiducial model described in § 4.2 uses the model of Macciò et al. (2007). However, other models are also available in the literature (e.g. Bullock et al. 2001; Eke, Navarro & Steinmetz 2001), which predict concentration-mass relations that are slightly different (see C08 for an assessment of the impact on galaxy-galaxy lensing). Furthermore, the presence of a (central) galaxy in a dark matter halo may have an impact on its concentration via, for example, adiabatic contraction (e.g. Blumenthal et al. 1986), which is not accounted for in the concentration-mass relations obtained from pure $N$-body simulations. Finally, attempts to measure halo concentrations observationally have thus far given conflicting results (e.g. van den Bosch & Swaters 2001; Comerford & Natarajan 2007, and referenecs therein). To examine how the lensing predictions depend on changes in halo concentation we consider two models: model CH, in which the concentrations are 2 times as high as in our fiducial model, and model CL, in which the concentrations are 2 times smaller. The left column of Fig. 7 shows the predictions of models CH (dashed lines) and CL (dotted lines) compared to our fiducial model (solid lines). Results are only shown for four luminosity bins, as indicated. Note that the model with higher (lower) halo concentrations predicts ESDs that are higher (lower). The effect is stronger on scales where the 1-halo central term dominates (see Fig 4). Accordingly, in the case of the brightest sample (L6f), models CL and CH differ from the fiducial model on scales up to $R\simeq 1\,h^{-1}{\rm Mpc}$, while in sample L4 the differences are only appreciable out to $R\simeq 0.5\,h^{-1}{\rm Mpc}$. We conclude that the predicted ESD depends quite strongly on the assumed halo concentrations, indicating that galaxy-galaxy lensing has the potential to constrain the density profiles of dark matter haloes (see Mandelbaum et al. 2008). In this paper, we have assumed that the concentration of a halo depends only on its mass, we have ignored the possible halo age-dependence of the concentration-mass relation (e.g. Wechsler et al. 2002; Zhao et al. 2003; Lu et al. 2006). If for a given mass older halos have higher concentration, and if the formation of a galaxy in a halo depends strongly on the formation history of the halo, then the age-dependence of the halo concentration must be taken into account. Unfortunately, it is unclear how to connect the halo age (hence halo concentration) with the properties of galaxies. For a given halo mass, the dispersion in the concentration is about 0.12 dex (e.g. Jing 2000), which corresponds to a change of about 20% in the predicted ESD on $R\sim 0.1\,h^{-1}{\rm Mpc}$. In our fiducial model, central galaxies are assumed to reside at the center of their dark matter haloes. However, as shown in van den Bosch et al. (2005), in haloes with masses $M_{h}>10^{13}h^{-1}{\rm M}_{\odot}$ there is evidence to suggest that central galaxies are offset from their halo centers by $\sim 3$ percent of the virial radius. A similar result was obtained by Berlind et al. (2003) using SPH simulations of galaxy formation. To examine how such an offset impacts on the ESDs, we consider two additional models. Following van den Bosch et al. (2005) and Yang et al. (2006), we assume that central galaxies in haloes with $M_{S}>10^{13}h^{-1}{\rm M}_{\odot}$ are offset from their halo centers by an amount that is drawn from a Gaussian distribution with zero mean. We use two different values for the dispersion of the distribution: $3\%$ of the virial radius (model ‘Dev1’) and $6\%$ of the virial radius (model ‘Dev2’). The corresponding lensing predictions are shown in the middle column of Fig. 7 as the dotted (Dev1) and dashed (Dev2) lines. Note that the offsets only affect the lensing signal for the brightest sample (L6f), where a larger offset results in a stronger suppression of the ESD on small scales ($R<0.1\,h^{-1}{\rm Mpc}$). For fainter samples, no (significant) differences with respect to the fiducial model are apparent, which owes to the fact that fainter centrals typically reside in haloes with $M_{h}<10^{13}h^{-1}{\rm M}_{\odot}$ which do not have an offset (at least in our models). As a final test, we examine the impact of halo shapes on the galaxy-galaxy lensing signal. In our fiducial model, dark matter haloes are assumed to be spherically symmetric. However, $N$-body simulations show that, in general, they are triaxial rather than spherical. Jing & Suto (1998) proposed a fitting formula for triaxial dark matter haloes, which has been applied to both strong and weak lensing analyses (e.g. Oguri, Lee & Suto 2003; Oguri & Keeton 2004; Tang & Fan 2005). In order to examine the impact of our assumption of halo sphericity, we consider an alternative model (model ‘TRI’), in which we assume that the dark matter density distribution is given by $\rho_{\rm TRI}(R)$, where $R$ specifies an ellipsoidal surface: $R^{2}=\left(\frac{x^{2}}{a^{2}}+\frac{y^{2}}{b^{2}}+\frac{z^{2}}{c^{2}}\right)c^{2}\,.$ (10) Here $a\leq b\leq c$ are the three principal semi-axes of the ellipsoid. We set $\rho_{\rm TRI}(R)=\rho(R^{\prime})$, where $\rho(R^{\prime})$ is the NFW profile, so that the total mass within a sphere of radius $R^{\prime}$ in $\rho(R^{\prime})$ is equal to the mass within the elliptical shell at $R$. For the axis ratios we adopt the distribution function given by Jing & Suto (2002): $\displaystyle p(a/c)$ $\displaystyle=$ $\displaystyle\frac{1}{\sqrt{2\pi}\times 0.113}\left(\frac{M_{\rm vir}}{M_{NL}}\right)^{0.07[\Omega(z)]^{0.7}}$ $\displaystyle\times$ $\displaystyle\exp\left\\{-\frac{\left[(a/c)(M_{\rm vir}/M_{NL})^{0.07[\Omega(z)]^{0.7}}-0.54\right]^{2}}{2(0.113)^{2}}\right\\}$ and $\displaystyle p(a/b\|a/c)$ $\displaystyle=$ $\displaystyle\frac{3}{2(1-\max(a/c,0.5))}$ (12) $\displaystyle\times$ $\displaystyle\left[1-\left(\frac{2a/b-1-\max(a/c,0.5)}{1-\max(a/b,0.5)}\right)^{2}\right]\,,$ where $M_{NL}$ is the characteristic mass scale, on which the rms of the top- hat smoothed over-density is equal to $1.68$. In practice, we proceed as follows: For each dark matter halo we first draw the axis ratio $a/c$ and $a/b$ using Eqs. (4.4) and (12), respectively. Next we draw a random 3D orientation of the principal axes, and project the dark matter particles along the (fiducial) line-of-sight. Next, for each halo, we determine the major axis of the projected distribution which we align with the major axis of the central galaxy. This assumption is motivated by observational claims that the major axis of a central galaxy tends to aligned with that of its host halo (e.g. Yang et al. 2006; Faltenbacher et al. 2008; Wang et al. 2008). As shown in the right-hand panels of Fig. 7, changing from spherical haloes (solid lines) to triaxial haloes (dotted lines) has almost no impact on the predicted ESDs. This should not come as an entire surprise, since the ESDs are azimuthally averaged over many haloes, which have random orientations on the sky. Note that here we assume that the halo is perfectly aligned with the central galaxy. However, the observational results mentioned above actually suggest a misalignment. We have tried a model in which the orientation of the host halo is uncorrelated with that of the central galaxy. The change in the results is very small compared with the model assuming perfect alignment. ### 4.5 Dependence on Cosmology Another very important assumption in our model prediction is the cosmological model in the calculations of the halo mass function and the geometrical properties of spacetime. The redshifts of galaxies considered here are restricted to $z\leq 0.2$, and the impact of changing cosmological parameters on the spacetime geometry is quite small in our analysis. On the other hand, changing cosmological parameters can change the halo masses assigned to individual groups, which may have significant impact on the expected lensing signal. In our fiducial model, we adopt the cosmology parameters obtained from the WMAP 3-year data, with $\Omega_{\rm m}=0.238$, $\Omega_{\Lambda}=0.762$, $n=0.951$, and $\sigma_{8}=0.75$ (Spergel et al. 2007). As comparison, we will show some results obtained assuming another set of cosmological parameters with $\Omega_{\rm m}=0.3$, $\Omega_{\Lambda}=0.7$, $n=1.0$, and $\sigma_{8}=0.9$, which is strongly supported by the first year data release of the WMAP mission (see Spergel et al. 2003) and has been considered in many previous studies. In what follows we will refer to this second set of parameters as the WMAP1 cosmology. Note that the cosmological parameters given by the recent WMAP 5-year data (Komatsu et al. 2008) are in between those of WMAP1 and WMAP3. Fig. 8 compares the ESD predicted by the fiducial model using the WMAP3 cosmology (solid lines) and that predicted by the WMAP1 cosmology (dotted lines). As one can see, the ESD predicted by WMAP1 is significantly higher than that predicted by WMAP3, especially for bright galaxies. Most of this increase is due to changes in the halo mass function, which causes (massive) groups to be assigned a larger halo mass. The changes in the halo concentrations and the spacetime geometry play only a minor role. A comparison with the SDSS data clearly favors the WMAP3 cosmology over the WMAP1 cosmology, especially for the brighter luminosity bins. The reduced $\chi^{2}$ for the WMAP1 cosmology is $21.3$, much larger than $3.2$ for the WMAP3 cosmology. This result is in good agreement with that obtained in C08 using the CLF model. Thus, we conclude that the galaxy-galaxy lensing data either prefer the WMAP3 cosmology, which has a relatively low $\sigma_{8}$, or our halo mass assignment is in serious error. In the following subsection we show that the uncertainties in our halo mass assignments are unlikely to change our results significantly. We therefore conclude that the galaxy-galaxy lensing data prefer a $\Lambda$CDM model with a relatively low $\sigma_{8}$. If we use WMAP5 parameters, the model prediction is in between WMAP1 and WMAP3, which is still too high to matched the observed ESD of bright galaxies. ### 4.6 Uncertainties in Halo-Mass Assignment In our model, the masses of groups are assigned according to the stellar-mass ranking and the halo mass function predicted by the adopted cosmology. The underlying assumption is that the mass function of the host haloes of groups is the same as that predicted by the cosmological model. However, even if the cosmological model adopted is a good approximation to the real universe, the observed halo density may be different from the model prediction because of cosmic variance introduced by the finite observational volume. The effect of such variance is expected to be most important for massive haloes, because the number density of such systems is small. If, for example, the number density of massive haloes in the observational sample is, due to cosmic variance, smaller than the model prediction, the mass assignment with the use of the theoretical halo mass function would assign a higher halo mass to groups. Consequently, the ESD of bright galaxies, which are biased toward massive haloes, would be overestimated. Here we test the importance of such effects by considering the uncertainties due to Poisson fluctuations. We use the halo mass function predicted with the WMAP3 cosmology to generate a set of random halo samples, each of which contains the same number of groups as the observational sample. The halo masses in each of these samples is then ranked in descending order and the halo mass of a given rank is then assigned to the group with the same rank in stellar mass. We find that the scatter in the ESD obtained in this way is negligibly small, even for the brightest sample. The reason is that, even for the brightest sample, the ESD is dominated by haloes with intermediate masses, $10^{13}h^{-1}{\rm M}_{\odot}<M_{S}<10^{14}h^{-1}{\rm M}_{\odot}$ (see Fig. 5), and the total number of such groups is quite large. Another uncertainty in the mass assignment may arise from fiber collisions. In the SDSS survey, no two fibers on the same SDSS plate can be closer than 55 arcsec. Although this fiber collision constraint is partially alleviated by the fact that neighboring plates have overlap regions, $\sim 7$ percent of all galaxies eligible for spectroscopy do not have a measured redshift. Our analysis here is based on the group catalog constructed from galaxy Sample II (see Y07), in which many of the galaxies missed due to fiber collisions are not included. Consequently, the total stellar mass (or the total luminosity) of some of the groups may be underestimated, thus introducing a bias in the ranking. This bias is expected to be stronger in richer systems because they have higher projected galaxy number density and are more likely to suffer from fiber collisions. However, this effect is not likely to have a big impact on our results, because it only changes the relative ranking of the groups that have similar stellar mass (luminosity) and because the galaxy-galaxy lensing signals are averaged in relatively broad bins of galaxy luminosity. In order to quantify this effect we carry out a similar analysis using the group catalog constructed from Sample III, where a galaxy affected by fiber collisions is assigned the redshift of its nearest neighbor (see Y07 for details). In this case, the situation is the opposite to that in Sample II, because here some galaxies may be wrongly assigned to groups in the foreground or the background due to the wrong redshifts assigned to some of the fiber- collision galaxies. The stellar mass and the luminosity of some of the groups will therefore be overestimated. In our test, we use Sample III to construct the group and to set mass to groups according to their ranking in Sample III, but we calculate the ESD only around galaxies that are in Sample II. The results obtained in this way are very similar to those based on Sample II alone, suggesting that the effect of missing fiber-collision galaxies in the mass assignment is not important. However, if the fiber-collision galaxies are themselves included in the calculation of the ESD, i.e. if we use galaxies in Sample III to calculate the ESD, the amplitude of the ESD is significantly larger than that obtained from the galaxies in Sample II. The reason for this is that the galaxies affected by fiber collisions are preferentially located in high density regions, so that they are more likely associated with a massive halo. The observational results of Mandelbaum et al. (2006) are based on galaxies that have spectroscopic redshifts, and so a fair comparison between the observational results and our model predictions can only be made with the use of galaxies in Sample II. Our test above therefore demonstrates that our conclusions about the comparison between the observational data and the model predictions are robust against the uncertainties due to fiber collisions. The scatter in the relation between halo mass and stellar mass (or luminosity) may also produce some uncertainties in our model prediction. As shown in Mandelbaum & Seljak (2007), the scatter is in partial degeneracy with cosmology model. In our investigation, we have fixed cosmological parameters and allowed no dispersion in the halo mass-total stellar mass relation. If, for example, we assume a log normal distribution with a dispersion of 0.3 index in halo mass for a given stellar mass, the predicted ESD would be a few percent larger than that predicted by the fiducial model. ## 5 Summary In this paper we model the galaxy-galaxy lensing signal expected for SDSS galaxies, using the galaxy groups selected from the SDSS to represent the dark matter haloes within which the galaxies reside. We use the properties of the dark halo population, such as mass function, density profiles and shapes, expected from the current $\Lambda$CDM cosmogony to model the dark matter distribution in each of the groups identified in the SDSS volume. The use of the real galaxy groups allows us to predict the galaxy-galaxy lensing signals separately for galaxies of different luminosity, morphological types, and in different environments (e.g. central versus satellite galaxies). We check the robustness of our model predictions by changing the assumptions about the dark matter distribution in individual groups (such as the shape, density profile, and center offset of dark matter haloes), as well as the cosmological model used to predict the properties of the halo population. We compare our model predictions with the observational data of Mandelbaum et al. (2006) for similar samples of lens galaxies. Although there is some discrepancy for lens galaxies in the low-luminosity bins, the overall observational results can be well understood in the current $\Lambda$CDM cosmogony. In particular, the observed results can be well reproduced in a $\Lambda$CDM model with parameters based on the WMAP3 data, but a $\Lambda$CDM model with a significantly higher $\sigma_{8}$, such as the one based on the WMAP1 data, significantly over-predicts the galaxy-galaxy lensing signal. Our results also suggest that, once a correct model of structure formation is adopted, the halo masses assigned to galaxy groups based on ranking their stellar masses with the halo mass function, are statistically reliable. The results obtained imply that galaxy-galaxy lensing is a powerful tool to constrain both the mass distribution associated with galaxies and cosmological models. In the future, when deep imaging surveys provide more sources with high image quality in the SDSS sky coverage, the galaxy-galaxy lensing signals produced by the SDSS galaxies can be estimated to much higher accuracy. The same analysis as presented here is expected to provide stringent constraints on the properties of the dark matter haloes associated with galaxies and galaxy systems, as well as on cosmological parameters. ## Acknowledgments We thank Rachel Mandelbaum for providing the SDSS lensing data in electronic format. Part of the computation was carried out on the SGI Altix 330 system at the Department of Astronomy, Peking University. Li Ran is supported by the National Scholarship from China Scholarship Council. XY is supported by the One Hundred Talents project of the Chinese Academy of Sciences and grants from NSFC (Nos. 10533030, 10673023). HJM would like to acknowledge the support of NSF AST-0607535, NASA AISR-126270 and NSF IIS-0611948. Zuhui Fan would like to acknowledge the supports from NSFC under grants. 10373001, 10533010, and 10773001, and 973 Program (No. 2007CB815401). ## References * Adelman-McCarthy (2006) Adelman-McCarthy J. K., et al., 2006, ApJS, 162, 38 * Bell (2003) Bell E. F. , McIntosh D. H., Katz N., Weinberg M. D., 2003, ApJS, 149, 289 * Berlind (2002) Berlind, A. A., Weinberg D. H., 2002, ApJ, 575, 587 * Berlind (2003) Berlind A. A., et al., 2003, ApJ, 593, 1 * Blanton (2003) Blanton M. R., et al. , 2003, AJ, 125, 2348 * Blanton (2005) Blanton M. R., et al., 2005, AJ, 129, 2562 * Blumenthal (1986) Blumenthal G. R., Faber S. M., Flores R., Primack J. R., 1986, ApJ, 301, 27 * Brainerd (1996) Brainerd T. G, Blandford R. D., Smail I., 1996, ApJ, 466, 623 * Bullock (2001) Bullock J. S., Kolatt T. S., Sigad Y., Somerville R. S., Kravtsov A. V., Klypin A. A., Primack J. R., Dekel A., 2001 MNRAS, 321, 559 * Cacciato (2008) Cacciato M., van den Bosch F. C., More S., Li R., Mo H. J, Yang X., 2008, submitted to MNRAS (C08) * cole (2000) Cole S., Lacey C.G., Baugh C.M., Frenk C.S. , 2000, MNRAS, 319, 168 * coll (2001) Colless M., et al., 2001, MNRAS, 328, 1039 * Comerford (2007) Comerford J. M., Natarajan P., 2007, MNRAS, 379, 190 * Cooray (2002) Cooray A., Sheth R., 2002, PhR, 372, 1 * Cooray (2006) Cooray A., 2006, MNRAS, 365, 842 * Croton (2006) Croton D. J., et al., 2006, MNRAS, 367, 864 * deVau (deVau) de Vaucouleurs G., de Vaucouleurs A., Corwin H.G., Buta R.J., Paturel G., Fouque P., 1991, Third Reference Catalogue of Bright Galaxies, (Springel-Verlag Heidelber) * Eke (2001) Eke V.R., Navarro J.F., Steinmetz M., 2001, ApJ, 554, 114 * Faltenbacher (2007) Faltenbacher A. , Li C., Mao S., van den Bosch F. C., Yang X., Jing Y. P., Pasquali A., Mo H. J. , 2007, ApJ, 662, 71 * Giocoli (2008) Giocoli C., Tormen G., van den Bosch F.C., 2008, MNRAS, 386, 2135 * Gao (2004) Gao L., White S. D. M., Jenkins A., Stoehr F., Springel V., 2004, MNRAS, 355, 819 * Hayashi (2003) Hayashi E., Navarro J. F., Taylor J. E., Stadel J., Quinn T., 2003, ApJ, 584, 541 * Henry (2000) Henry, J. P., 2000, ApJ, 534, 565 * Hoekstra (2003) Hoekstra H., Franx M., Kuijken K., Carlberg R. G., Yee H. K. C., 2003, MNRAS, 340, 609 * Hoekstra (2004) Hoekstra H., 2004, MNRAS 347, 1337 * Hudson (1998) Hudson M. J., Gwyn S. D. J., Dahle H., Kaiser N., 1998, ApJ, 503, 531 * Jing (1998) Jing Y. P., Suto, Y., 1998, ApJ, 494L, 5 * Jing (00) Jing Y. P., 2000, ApJ, 535, 30 * Jing (2002) Jing Y. P., Suto Y., 2002, ApJ, 574, 538 * JMB (1998) Jing Y. P., Mo H. J., Börner G., 1998, ApJ, 494, 1 * Johnston (2007) Johnston D. E., et al., 2007, preprint (arXiv: astro-ph/0709.1159) * Kang (2005) Kang X., Jing Y. P., Mo H. J., Börner G., 2005, ApJ, 631, 21 * (33) Kauffmann G., White S. D. M., Guiderdoni B., 1993, MNRAS, 264, 201 * (34) Kauffmann G. , White S. D. M. , Heckman T. M. , Menard B. , Brinchmann J. , Charlot S. , Tremonti C. , Brinkmann J. 2004, MNRAS, 353, 713 * (35) Katz N., Weinberg D. H., Hernquist L., 1996, ApJS, 105, 19 * (36) Komatsu E. , Dunkley J. , 2008, preprint (arXiv: astro-ph/0803.0547) * Lin (2007) Limousin M, Kneib J. P., Bardeau S., Natarajan P., Czoske O., Smail I., Ebeling H., Smith G.P., 2007, A&A, 461, 881 * Lu (2006) Lu Y., Mo H. J., Katz N., Weinberg M. D., 2006 , MNRAS, 368 1931 * (39) Macciò A. V., Dutton A. A., van den Bosch F. C., Moore B., Potter D., Stadel J., 2007, MNRAS, 378, 55. * (40) Mandelbaum R., et al., 2005, MNRAS, 361, 1287 * (41) Mandelbaum R., Seljak U., Kauffmann G., Hirata C. M., Brinkmann J., 2006 MNRAS, 368, 715 * (42) Mandelbaum R., Seljak, U., 2007, JACP, 06, 024 * (43) Mandelbaum R., Seljak, U., Hirata C. M., 2008, preprint (arXiv: astro-ph/0805.2552) * (44) McKay T. A., Sheldon E. S., Racusin J., et al. 2001, preprint (arXiv: astro-ph/0108013) * (45) McKay T. A., et al., 2002, ApJ, 571, L85 * (46) Nakamura T. T., Suto Y., 1997, PThPh, 97, 49 * Natarajan (2002) Natarajan P., Kneib J.P., Smail I., 2002, ApJ, 580, L11 * (48) Navarro J.F., Frenk C.S., White S.D.M., 1997, ApJ, 490, 493 * (49) Oguri M., Lee J., Suto Y., 2003, ApJ, 599, 7 * (50) Oguri M., Keeton C., 2004, ApJ, 610, 663 * (51) Peacock J. A., Smith R. E., 2000, MNRAS, 318, 1144 * (52) Pearce F. R., Thomas P. A., Couchman H. M. P., Edge A. C., 2000, MNRAS, 317, 1029 * (53) Petrosian V., 1976, ApJ, 209, L1 * (54) Saunders W., et al., 2000, MNRAS, 317, 55 * (55) Scranton R., 2003, MNRAS 339, 410 * (56) Schlegel D.J., Finkbeiner D.P., Davis M., 1998, ApJ, 500, 525 * (57) Sheldon E. S., et al., 2004, AJ, 127, 2544 * (58) Sheldon E. S., et al., 2007a, preprint (arXiv:astro-ph/0709.1153) * (59) Sheldon E. S., et al., 2007b, preprint (arXiv: astro-ph/0709.1162) * (60) Somerville R. S., Primack J. R., 1999, MNRAS, 310, 1087 * (61) Spergel D. N., et al., 2003, ApJS, 148, 175 * (62) Spergel D. N., et.al., 2007, ApJS, 170, 377 * (63) Springel V., 2005, MNRAS, 364, 1105 * (64) Springel V. , et al. , 2005, Nature, 435, 629 * (65) Strauss M.A., et al., 2002, AJ, 124, 1810 * (66) Tang J. Y. , Fan Z. H., 2005, ApJ, 635, 60 * (67) Tinker J.L., Weinberg D.H., Zheng Z., Zehavi I., 2005, ApJ, 631, 41 * (68) Tyson J. A., Valdes F., Jarvis J. F., Mills A. P. Jr., 1984, ApJ, 281L, 59 * (69) Vale A., Ostriker J. P., 2006, MNRAS, 371, 1173 * (70) van den Bosch F. C. , Swaters R. A., 2001, MNRAS, 325, 1017 * (71) van den Bosch F. C., 2002, MNRAS, 332, 456 * (72) van den Bosch F. C., Yang X. , Mo H. J. , 2003, MNRAS, 340, 771 * (73) van den Bosch F. C., Norberg P., Mo H. J., Yang X., 2004, MNRAS, 352, 1302 * (74) van den Bosch F. C., Weinmann S. M., Yang X., Mo H. J., Li C., Jing Y.P., 2005, MNRAS, 361, 1203 * (75) van den Bosch F. C., Tormen G., Giocoli C., 2005, MNRAS, 359, 1029 * (76) van den Bosch F. C., et al., 2007, MNRAS, 376, 841 * (77) Wang Y., Yang X., Mo H. J., Li C., van den Bosch F. C., Fan Z. H., Chen X., 2008, MNRAS, 385, 1511 * (78) Warren M. S., Abazajian K., Holz D. E., Teodoro L., 2006, ApJ, 646, 881 * (79) Wechsler R. H., Bullock J. S., Primack J. R., Kravtsov A. V., Dekel A., 2002, ApJ, 568, 52 * (80) Weinberg D. H., Colombi S., Davè R., Katz N., 2008, ApJ, 678, 6 * (81) White S. D. M., Frenk C., 1991, ApJ, 379, 52 * (82) Yan R., Madgwick D. S., White M. , ApJ, 2003, 598, 848 * (83) Yang X., Mo H. J., van den Bosch F. C., 2003, MNRAS, 339, 1057 * (84) Yang X., Mo H. J., van den Bosch F. C., Jing Y. P., 2005, MNRAS, 356, 1293 * (85) Yang X., Mo H. J., van den Bosch F. C., Jing Y. P., Weinmann S. M., Meneghetti M., 2006, MNRAS, 373, 1159 * (86) Yang X., Mo H. J., van den Bosch F. C., Pasquali A., Li C., Barden M., 2007, ApJ, 671, 153 (Y07) * (87) Yang X., Mo H. J., van den Bosch F. C., 2008, preprint, (arXiv: astro-ph/0808.0539) * (88) Zhao D. H., Jing Y. P., Mo H. J., Börner G., 2003, ApJ, 597, 9 * (89) Zheng Z. , et al. , 2005, ApJ, 633, 791 * (90) Zheng Z., Weinberg D. H., 2007, ApJ, 659, 1	arxiv-papers	2008-07-30T20:54:16	2024-09-04T02:48:57.045568	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Ran Li, H.J. Mo, Zuhui Fan, Marcello Cacciato, Frank C. van den Bosch,\n Xiaohu Yang, Surhud More", "submitter": "Ran Li", "url": "https://arxiv.org/abs/0807.4934" }
0807.4952	# Persistence of laminations Pierre Berger Institute for Mathematical Sciences, Stony Brook University, Stony Brook NY 11794-3660, USA berger@phare.normalesup.org ###### Abstract We present a modern proof of some extensions of the celebrated Hirsch-Pugh- Shub theorem on persistence of normally hyperbolic compact laminations. Our extensions consist of allowing the dynamics to be an endomorphism, of considering the complex analytic case and of allowing the laminations to be non compact. To study the analytic case, we use the formalism of deformations of complex structures. We present various persistent complex laminations which appear in dynamics of several complex variables: Hénon maps, fibered holomorphic maps… In order to proof the persistence theorems, we construct a laminar structure on the stable and unstable of the normally hyperbolic laminations. ###### Contents 1. 1 Geometry of lamination 2. 2 Persistence of laminations 3. 3 Main theorem on differentiable persistence 4. 4 Space of preorbits 5. 5 Persistence of complex laminations 6. 6 Proof of differentiable persistences of normally contracted laminations 7. 7 Proof of differentiable persistences of normally expanded laminations 8. 8 Proof of persistences of complex laminations by deformation 9. 9 Proof of persistences of normally hyperbolic laminations ## Introduction In 1977, M. Hirsch, C. Pugh and M. Shub [HPS77] developed a theory which has been very useful for hyperbolic dynamical systems. The central point of their work was to prove the $C^{r}$-persistence of manifolds, foliations, or more generally laminations which are $r$-normally hyperbolic and plaque-expansive, for all $r\geq 1$. We recall that a lamination is _$r$ -normally hyperbolic_, if the dynamics preserves the lamination (each leaf is sent into a leaf) and if the normal space of the leaves splits into two $Tf$-invariant subspaces, that $Tf$ contracts (or expands) $r$-times more sharply than the tangent space to the leaves. Hence, $0$-normal expansion (resp. contraction) means that the action of the dynamics on the normal bundle is expanding (resp. contracting), and so does not require the existence of a dominated splitting. _Plaque- expansiveness_ is a generalization of expansiveness to the concept of laminations. The _$C^{r}$ -persistence_ of such a lamination means that for any $C^{r}$-perturbation of the dynamics, there exists a lamination, $C^{r}$-close to the first, which is preserved by the new dynamics, and such that the dynamics induced on the space of the leaves remains the same. A direct application of this theory was the construction of an example of a robustly transitive diffeomorphism (every close diffeomorphism has a dense orbit) but not Anosov. Then their work was used for example by C. Robinson [Rob76] to achieve the proof of the Palis-Smale conjecture: the $C^{1}$-diffeomorphisms that satisfy Axiom A and the strong transversality condition are structurally stable. Nowadays, this theory remains very useful in several mathematical fields such as generic dynamical systems, differentiable dynamics, foliations theory or Lie group theory. A first result is an analogue of the HPS theorem to the endomorphism case: We allow $f$ to be an endomorphism that is to be possibly non-bijective and with singularities, but we suppose $f$ to be normally expanding instead of normally hyperbolic: ###### Theorem 0.1. Let $r\geq 1$ and let $(L,\mathcal{L})$ be a compact lamination $C^{r}$-immersed by $i$ into a manifold $M$. Let $f$ be a $C^{r}$-endomorphism of $M$ which preserves and $r$-normally expands $\mathcal{L}$. Then the immersed lamination is $C^{r}$-persistent. If moreover $i$ is an embedding and $f$ is forward plaque-expansive at $\mathcal{(}L,\mathcal{L})$ then the embedded lamination is $C^{r}$-persistent. A second result is a generalization of the HPS theorem: ###### Theorem 0.2. Let $r\geq 1$ and let $(L,\mathcal{L})$ be a compact lamination $C^{r}$-immersed by $i$ into a manifold $M$. Let $f$ be a $C^{r}$-endomorphism of $M$ which preserves and is $r$-normally hyperbolic to $\mathcal{L}$ with a bijective pullback $f^{}$. Then the immersed lamination is $C^{r}$-persistent. If moreover $i$ is an embedding and $f^{}$ is plaque-expansive at $\mathcal{(}L,\mathcal{L})$ then the embedded lamination is $C^{r}$-persistent. We recall that a _pullback_ of an endomorphism $f$ preserving a lamination $(L,\mathcal{L})$ immersed by $i$ is an endomorphism $f^{}$ of $(L,\mathcal{L})$ such that: $i\circ f^{}=f\circ i.$ In the complex analytic world, we prove the following theorem: ###### Theorem 0.3. Let $(L,\mathcal{L})$ be a complex lamination immersed by i into a complex manifold $M$. Let $f$ be a holomorphic endomorphism of $M$ which preserves and $0$-normally expands $\mathcal{L}$. Then for any complex $k$-ball $B$, and any complex analytic family $(f_{t})_{t\in B}$ of endomorphisms of $M$, such that $f_{0}$ is $f$, there exist an open neighborhood $B_{0}$ of $0\in B$ and a complex analytic family of laminations $(L,\mathcal{L}_{t})_{t\in B_{0}}$ immersed by $(i_{t})_{t}$ such that: * - $f_{0}=f$, $i_{0}=i$ and $\mathcal{L}_{0}=\mathcal{L}$, * - $f_{t}$ preserves the lamination $(L,\mathcal{L}_{t})$ holomorphically immersed by $i_{t}$ into $M$, for any $t\in B_{0}$. If moreover $i$ is an embedding and $f$ is forward plaque-expansive, then $i_{t}$ is an embedding for every $t\in B_{0}$. A _complex analytic family of endomorphisms_ $(f_{t})_{t\in B}$ of $M$ means that $(t,x)\in B\times M\mapsto(t,f_{t}(x))$ is complex analytic. By complex analytic family of laminations $(L,\mathcal{L}_{t})_{t\in B_{0}}$ immersed by $(i_{t})_{t}$, we mean that there exists a complex lamination structure $\mathcal{D}$ on $D:=B\times L$ such that: * - $\overline{w}\;:\;(x,t)\in(D,\mathcal{D})\mapsto t\in B_{0}$ is a complex analytic submersion and $\overline{w}^{-1}(t)=(L,\mathcal{L}_{t})$, * - $(x,t)\in(D,\mathcal{D})\mapsto(i_{t}(x),t)\in M\times B$ is a complex analytic immersion. ###### Remark 0.4. In Example 5.5, we show that the complex structure $\mathcal{D}$ on $L\times B_{0}$ is not necessarily the trivial product structure $\mathcal{L}\times B_{0}$. Nevertheless, we give a sufficient condition to have such a product structure (see Proposition 5.3). ###### Remark 0.5. The above theorem remains true if $L$ is not necessarily compact but a pullback $f^{}$ of $f$ sends $cl(L)$ into $L$. The mirror result of the generalization of HPS’s theorem is the following: ###### Theorem 0.6. Let $(L,\mathcal{L})$ be a compact complex lamination immersed by $i$ into a complex manifold $M$. Let $f$ be a holomorphic endomorphism of $M$ which preserves, has bijective pull back $f^{}$ and is $1$-normally hyperbolic to $(L,\mathcal{L})$. Then for any complex $k$-ball $B$ and complex analytic family $(f_{t})_{t\in B}$ of endomorphisms of $M$ such that $f_{0}$ is $f$, there exist an open neighborhood $B_{0}$ of $0$ and a complex analytic family of laminations $(L,\mathcal{L}_{t})_{t\in B_{0}}$ immersed by $(i_{t})_{t}$ such that: * - $f_{0}=f$, $i_{0}=i$ and $\mathcal{L}_{0}=\mathcal{L}$, * - $f_{t}$ preserves the lamination $(L,\mathcal{L}_{t})$ holomorphically immersed by $i_{t}$ into $M$, for any $t\in B_{0}$. If moreover $i$ is an embedding and $f^{}$ is plaque-expansive, then $i_{t}$ is an embedding for every $t\in B_{0}$. ###### Remark 0.7. Under the hypothesis of the above theorem, we have the same conclusion if $f$ is $0$-normally _contracting_ instead of $1$-normally hyperbolic. We will provide some extensions of all the above results in the _non-compact cases_ (see Theorems 3.1 and 3.4). We notice that in all the above theorems, the hypotheses are open. Let us finish this introduction by giving two examples. ###### Example 0.8. Let $R_{1}$ and $R_{2}$ be two rational functions of the Riemannian sphere $\mathbb{S}^{2}$. Let $K$ be a compact subset of $\mathbb{S}^{2}$ expanded by $R_{1}$: $R_{1}$ sends $K$ into itself and there exists $\lambda>1$ such that for a Riemannian metric on $\mathbb{S}^{2}$, for every $x\in K$ $\|d_{x}R_{1}\|\geq\lambda$. We suppose that for every $z\in\mathbb{S}^{2}$, $\|d_{z}R_{2}\|<\lambda$. Let $f:\;(z,z^{\prime})\in\mathbb{S}^{2}\times\mathbb{S}^{2}\rightarrow(R_{1}(z),R_{2}(z^{\prime}))$. Let $\mathcal{L}$ be the lamination on $K\times\mathbb{S}^{2}$ whose leaves are of the form $\mathcal{L}_{k}:=\\{k\\}\times\mathbb{S}^{2}$, for $k\in K$. We notice that this compact lamination is preserved by $f$ and $1$-normally expanded by $f$. By Theorem 0.1, this lamination is $C^{1}$-persistent. In other words, for any $C^{1}$-perturbation $f^{\prime}$ of $f$, there exists a family of manifolds $(\mathcal{L}^{\prime}_{k})_{k\in K}$ such that for any $k\in K$: - $\mathcal{L}^{\prime}_{k}$ is $C^{1}$-close to $\mathcal{L}_{k}$: there exists a $C^{1}$-embedding of $\mathcal{L}_{k}$ onto $\mathcal{L}^{\prime}_{k}$ which is $C^{1}$-close to the canonical inclusion, * - the endomorphism $f^{\prime}$ sends $\mathcal{L}^{\prime}_{k}$ into $\mathcal{L}^{\prime}_{R_{1}(k)}$, * - for $k^{\prime}$ close to $k$ the sphere $\mathcal{L}^{\prime}_{k^{\prime}}$ is $C^{1}$-close to $\mathcal{L}_{k}$. Moreover, since $f$ is plaque-expansive at $\mathcal{L}$, the submanifolds $(\mathcal{L}^{\prime}_{k})_{k}$ are disjoint from each other. This implies that there exists a $C^{1}$-embedding, close to the canonical inclusion, of $\mathcal{L}$ onto the laminations $\mathcal{L}^{\prime}$ formed by the leaves $(\mathcal{L}^{\prime}_{k})_{k}$. ###### Example 0.9. Let $M:=\mathbb{S}^{1}\times\mathbb{S}^{1}\times\mathbb{R}^{2}$ be the product of the 2-torus with the real plane. Let $\alpha\in\mathbb{S}^{1}$. $\mathrm{Let}\;f:\;M:=\mathbb{S}^{1}\times\mathbb{S}^{1}\times\mathbb{R}^{2}\rightarrow M$ $(\theta,\phi,x,y)\mapsto(2\theta,\phi+\alpha,0,10y).$ We notice that $f$ is $4$-normally hyperbolic at the torus $\mathbb{T}^{2}:=\mathbb{S}^{1}\times\mathbb{S}^{1}\times\\{0\\}$. However it is not injective restricted to this torus, and so we cannot apply Theorem 0.2. That is why we consider the Smale’s solenoid $\tilde{\mathbb{S}}$ which projects onto $\mathbb{S}^{1}$ by sending the points of a same stable manifold to a same point of $\mathbb{S}^{1}$. Let $\pi$ be this projection. We endow $\tilde{\mathbb{T}}^{2}:=\tilde{\mathbb{S}}\times\mathbb{S}^{1}$ with the 2-dimensional lamination structure whose leaves are the product of the unstable manifolds with the circle $\mathbb{S}^{1}$. We notice that the map $i:\;(x,\phi)\in\tilde{\mathbb{S}}\times\mathbb{S}^{1}\mapsto(\pi(x),\phi,0,0)\in M$ is an immersion of this lamination onto the torus $\mathbb{T}^{2}$. Also the dynamics $f_{\|\mathbb{T}^{2}}$ lifts to the product $\tilde{f}$ of the usual dynamics of the solenoid with the rotation of angle $\alpha$. In other words the following diagram commutes: $\begin{array}[]{rcccl}&&f&&\\\ &\mathbb{T}^{2}\subset M&\rightarrow&\mathbb{T}^{2}\subset M&\\\ i&\uparrow&&\uparrow&i\\\ &\tilde{\mathbb{T}}^{2}&\rightarrow&\tilde{\mathbb{T}}^{2}&\\\ &&\tilde{f}&&\\\ \end{array}$ Theorem 0.2 implies that the immersed lamination $\tilde{\mathbb{T}}^{2}$ is $C^{4}$-persistent. In particular, for any $C^{4}$-perturbation $f^{\prime}$ of $f$, there exists an immersion $i^{\prime}$ $C^{4}$-close to $i$ such that $f^{\prime}$ sends each leaf of $\tilde{\mathbb{T}}^{2}$ immersed by $i^{\prime}$ to the immersion by $i^{\prime}$ of the image by $\tilde{f}$ of this leaf. Actually one can show that the torus $\mathbb{T}^{2}$ is not persistent. I thank J.-C. Yoccoz who suggested that I study the persistence of complex laminations. I am also grateful to C. Bonatti wonder me about the persistence of normally hyperbolic laminations by endomorphisms, by looking at the preorbits spaces. I thank also D. Varolin, J. Milnor, M. Lyubich, J. Kahn and S. Bonnot for many discussions and valuable suggestions. ## 1 Geometry of lamination ### 1.1 Definitions Throughout this chapter, $r$ refers to a fixed positive integer or to ${\mathcal{H}}$. The vector space $\mathbb{K}$ refers to the real line $\mathbb{R}$ when $r$ is an integer or to $\mathbb{C}$ when $r$ is ${\mathcal{H}}$. For our purpose it is convenient to denote by $C^{\mathcal{H}}$ the class of holomorphic maps. Let us consider a locally compact and second-countable metric space $L$ covered by open sets $(U_{i})_{i}$, called _distinguished open sets_ , endowed with homeomorphisms $h_{i}$ from $U_{i}$ onto $V_{i}\times T_{i}$, where $V_{i}$ is an open set of $\mathbb{K}^{d}$ and $T_{i}$ is a metric space. We say that the _charts_ $(U_{i},h_{i})_{i}$ define a $C^{r}$-_atlas_ of a lamination structure on $L$ of dimension $d$ if the _coordinate change_ $h_{ij}=h_{j}\circ h_{i}^{-1}$ can be written in the form $h_{ij}(x,t)=(\phi_{ij}(x,t),\psi_{ij}(x,t)),$ where $\phi_{ij}$ takes its values in $\mathbb{K}^{d}$, $\psi_{ij}(\cdot,t)$ is locally constant for any $t$ and: * - if $r$ is an integer then the partial derivatives $(\partial_{x}^{s}\phi_{ij})_{s=1}^{r}$ exist and are continuous on the domain of $\phi_{ij}$, * - if $r$ is ${\mathcal{H}}$ then $(\phi_{ij}(\cdot,t))_{t}$ is a continuous family of complex analytic maps. A $(C^{r})$-_lamination_ is a metric space $L$ endowed with a maximal $C^{r}$-atlas $\mathcal{L}$. A _plaque_ is a subset of $L$ which can be written in the form $h_{i}^{-1}(V_{i}^{0}\times\\{t\\})$, for a chart $h_{i}$, a point $t\in T_{i}$, and a connected component $V_{i}^{0}$ of $V_{i}$. A plaque that contains a point $x\in L$ will be denoted by $\mathcal{L}_{x}$; the union of the plaques containing $x$ and of diameter less than $\epsilon>0$ will be denoted by $\mathcal{L}_{x}^{\epsilon}$. As the diameter is given by the metric of $L$, the set $\mathcal{L}_{x}^{\epsilon}$ is – in general – not homeomorphic to a manifold. The _leaves_ of $\mathcal{L}$ are the smallest subsets of $L$ which contain any plaque that intersects them. If $V$ is an open subset of $L$, the set of the charts $(U,\phi)\in\mathcal{L}$ such that $U$ is included in $V$, forms a lamination structure on $V$, which is denoted by $\mathcal{L}_{\|V}$. A subset $P$ of $L$ is _saturated_ if it is a union of leaves. An _admissible_ subset $A$ of $L$ is a locally compact and saturated subset of the restriction of the lamination to some open subset. ###### Examples 1.1. * - A manifold of dimension $d$ is a lamination of the same dimension. * - A $C^{r}$-foliation on a connected manifold induces a $C^{r}$-lamination structure. * - A locally compact and second-countable metric space defines a lamination of dimension zero. * - The Smale solenoid attractor is endowed with a structure of laminations whose leaves are the stable manifolds. * - Let $M$ be the cylinder $\mathbb{S}^{1}\times\mathbb{R}$. Let $\pi$ be the canonical projection of $\mathbb{R}$ onto $\mathbb{S}^{1}\cong\mathbb{R}/\mathbb{Z}$. $\mathrm{Let}\;L:=\Big{\\{}({\theta},y)\in M:\;y=\arctan(\overline{\theta}),\;\pi(\overline{\theta})=\theta\Big{\\}}\cup\mathbb{S}^{1}\times\\{-\pi/2,\pi/2\\}.$ The compact space $L$ is canonically endowed with a $1$-dimensional lamination structure which consists of the leaves $\mathbb{S}^{1}\times\\{-\pi/2\\}$, $\mathbb{S}^{1}\times\\{\pi/2\\}$, and a last one which spirals down to these two circles. * - The stable foliation of an Anosov $C^{r}$-diffeomorphism defines a $C^{r}$-lamination structure whose leaves are the stable manifolds. ###### Property 1.2. If $(L,\mathcal{L})$ and $(L^{\prime},\mathcal{L}^{\prime})$ are two laminations, then $L\times L^{\prime}$ is endowed with the lamination structure whose leaves are the product of the leaves of $(L,\mathcal{L})$ with the leaves of $(L^{\prime},\mathcal{L}^{\prime})$. We denote this structure by $\mathcal{L}\times\mathcal{L}^{\prime}$. ### 1.2 Morphisms of laminations A $C^{r}$-morphism (of laminations) from $(L,\mathcal{L})$ to $(L^{\prime},\mathcal{L}^{\prime})$ is a continuous map $f$ from $L$ to $L^{\prime}$ such that, seen via charts $h$ and $h^{\prime}$, it can be written in the form: $h^{\prime}\circ f\circ h^{-1}(x,t)=(\phi(x,t),\psi(x,t)),$ where $\phi$ takes its values in $\mathbb{K}^{d^{\prime}}$, $\psi(\cdot,t)$ is locally constant and: * - if $r$ an integer, then $\partial_{x}^{s}\phi$ exists and is continuous on the domain of $\phi$, for all $s\in\\{1,\dots,r\\}$, * - if $r$ is ${\mathcal{H}}$, then $(\phi(\cdot,t))_{t}$ is a continuous family of complex analytic maps. If moreover the linear map $\partial_{x}\phi(x,t)$ is always one-to-one (resp. onto), we will say that $f$ is an _immersion (of laminations)_ (resp. _submersion_). An _isomorphism (of laminations)_ is a bijective morphism of laminations whose inverse is also a morphism of laminations. An _embedding (of laminations)_ is an immersion which is a homeomorphism onto its image. An _endomorphism of $(L,\mathcal{L})$_ is a morphism from $(L,\mathcal{L})$ into itself. We denote by: * - $Mor^{r}(\mathcal{L},\mathcal{L}^{\prime})$ the set of the $C^{r}$-morphisms from $\mathcal{L}$ into $\mathcal{L}^{\prime}$, * - $Im^{r}(\mathcal{L},\mathcal{L}^{\prime})$ the set of the $C^{r}$-immersions from $\mathcal{L}$ into $\mathcal{L}^{\prime}$, * - $Emb^{r}(\mathcal{L},\mathcal{L}^{\prime})$ the set of the $C^{r}$-embeddings from $\mathcal{L}$ into $\mathcal{L}^{\prime}$, * - $End^{r}(\mathcal{L})$ the set of the $C^{r}$-endomorphisms of $\mathcal{L}$. We denote by $T\mathcal{L}$ the vector bundle over $L$, whose fiber at $x\in L$, denoted by $T_{x}\mathcal{L}$, is the tangent space at $x$ to its leaf. The differential $Tf$ of a morphism $f$ from $\mathcal{L}$ into $\mathcal{L}^{\prime}$ is the bundle morphism from $T\mathcal{L}$ into $T\mathcal{L}^{\prime}$ over $f$, induced by the differential of $f$ along the leaves of $\mathcal{L}$. ###### Remark 1.3. If $M$ is a manifold, we notice that $End^{r}(M)$ denotes the set of $C^{r}$-maps from $M$ into itself, possibly non-bijective and possibly with singularities. ###### Example 1.4. Let $f$ be a $C^{r}$-diffeomorphism of a manifold $M$. Let $K$ be a hyperbolic compact subset of $M$. Then, the union $W^{s}(K)$ of the stable manifolds of points in $K$ is a $C^{r}$-lamination $(L,\mathcal{L})$ immersed injectively. Moreover, if every stable manifold does not accumulate on $K$, then $(L,\mathcal{L})$ is a $C^{r}$-embedded lamination. ###### Proof. See [Ber] Example 1.1.5.∎ ### 1.3 Riemannian metric on a lamination A _Riemannian metric_ $g$ on a $C^{r}$-lamination $(L,\mathcal{L})$ is an inner product $g_{x}$ on each fiber $T_{x}\mathcal{L}$ of $T\mathcal{L}$, which depends continuously on the base point $x$. It follows from the local compactness and the second-countability of $L$ that any lamination $(L,\mathcal{L})$ can be endowed with some Riemannian metric. A Riemannian metric induces – in a standard way – a metric on each leaf. For two points $x$ and $y$ in a same leaf, the distance between $x$ and $y$ is defined by: $d_{g}(x,y)=\inf_{\\{\gamma\in Mor([0,1],\mathcal{L});\gamma(0)=x,\gamma(1)=y\\}}\int_{0}^{1}\sqrt{g(\partial_{t}\gamma(t),\partial_{t}\gamma(t))},dt.$ ### 1.4 Equivalent Classes of morphisms We will say that two morphisms $f$ and $f^{\prime}$ in $Mor^{r}(\mathcal{L},\mathcal{L}^{\prime})$ (resp. $Im^{r}(\mathcal{L},\mathcal{L}^{\prime})$ and $End^{r}(\mathcal{L}))$ are _equivalent_ if they send each leaf of $\mathcal{L}$ into the same leaf of $\mathcal{L}^{\prime}$. The equivalence class of $f$ will be denoted by $Mor^{r}_{f}(\mathcal{L},\mathcal{L}^{\prime})$ (resp. $Im_{f}^{r}(\mathcal{L},\mathcal{L}^{\prime})$ and $End_{f}^{r}(\mathcal{L})$). Given a Riemannian metric $g$ on $(L^{\prime},\mathcal{L}^{\prime})$, we endow the equivalence class with the $C^{r}$-compact-open topology. Let us describe elementary open sets which generate this topology. Let $K$ be a compact subset of $L$ such that $K$ and $f(K)$ are included in distinguished open subsets endowed with charts $(h,U)$ and $(h^{\prime},U^{\prime})$. We define $(\phi,\psi)$ by $h^{\prime}\circ f\circ h^{-1}=(\phi,\psi)$ on $h(K)$. Let $\epsilon>0$. The following subset is an elementary open set of the topology: $\Omega:=\Big{\\{}f^{\prime}\in Mor^{r}_{f}(\mathcal{L},\mathcal{L}^{\prime})\;:\;f^{\prime}(K)\subset U^{\prime}\;\mathrm{and\;there\;is}\;\phi^{\prime}\;\mathrm{s.t.}\;\quad$ $h^{\prime}\circ f^{\prime}\circ h^{-1}=(\phi^{\prime},\psi)\;\mathrm{and}\;\max_{h(K)}\big{(}\sum_{s=1}^{r}\\|\partial_{x}^{s}\phi-\partial_{x}^{s}\phi^{\prime}\\|\big{)}<\epsilon\Big{\\}}.$ with the convention that the sum $\sum_{s=1}^{\mathcal{H}}$ is zero. Thus for $r=\mathcal{H}$, this topology is the $C^{0}$-compact-open topology induced by the metrics of the leaves. For any manifold $M$, each of the spaces $Im^{r}(\mathcal{L},M)$, $Emb^{r}(\mathcal{L},M)$ and $End^{r}(M)$ contains a unique equivalence class. We endow these spaces with the topology of their unique equivalence class. In particular the topology on $C^{r}(M,M)=End^{r}(M)$ is the (classical) $C^{r}$-compact-open topology. ### 1.5 Tubular neighborhood of an immersed lamination Let $r\in[\\![1,\infty[\\![\cup\\{\mathcal{H}\\}$. Let $(L,\mathcal{L})$ be a lamination $C^{r}$-immersed by $i$ into a $C^{r}$-manifold $M$. Via $i$, we can identify the bundle $T\mathcal{L}$ over $L$ to a subbundle of $i^{}TM\rightarrow L$ whose fiber at $x\in L$ is $T_{i(x)}M$. The quotient bundle $\pi\;:\;F:=i^{}TM/T\mathcal{L}\rightarrow L$ is called the normal bundle of the immersed lamination $(L,\mathcal{L})$. Let $d$ be the dimension of $\mathcal{L}$ and let $n$ be the dimension of $M$. Thus the dimension of the fibers is $n-d$. A _$C^{r}$ -tubular neighborhood_ of the immersed lamination $(L,\mathcal{L})$ is the data of a $C^{r}$-lamination structure $\mathcal{F}$ on a neighborhood $F^{\prime}$ of the $0$-section, with a $C^{r}$-immersion $I$ from $(F^{\prime},\mathcal{F})$ into $M$, such that: * - The preimage by $\pi_{\|F^{\prime}}$ of the plaques of $\mathcal{L}$ are plaques of $\mathcal{F}$. Hence the dimension of $\mathcal{F}$ is $n$. * - the restriction $\pi_{\|F^{\prime}}$ is a $C^{r}$-submersion, * - $I\circ 0_{F}=i$. We denote such a tubular neighborhood by $(F,\mathcal{F},I,\pi)$. ###### Proposition 1.5. Every $C^{r}$-immersed lamination has a tubular neighborhood, when $r$ is a positive integer. ###### Remark 1.6. The above proposition is not always true in the holomorphic case. For example, the equation $y^{2}=4X^{3}+aX+b$ defines a torus of $\mathbb{C}\mathbb{P}^{2}$, whose modulus varies with $(a,b)\in\mathbb{C}^{2}$. Such tori depend (differentially) smoothly on $(a,b)$ but are not all biholomorphic. Thus, such tori cannot have a holomorphic tubular neighborhood. Nevertheless, all Stein submanifolds (see [Siu77]) among others are endowed with an analytic tubular neighborhood. ###### Proof of Proposition 1.5. We showed in [Ber] appendix A.2.1 that there exists a $C^{r}$-lifting $N$ of $i$ in the Grassmannian of $(n-d)$-planes of $TM$ such that, for every $x\in L$: $N(x)\oplus Ti(T_{x}\mathcal{L})=T_{i(x)}M.$ As the above sum is direct, we shall canonically identify the fiber $F_{x}$ of $F$ at $x$, with the subspace $N(x)$. Let $g$ be the smooth Riemannian metric on $M$, and exp be the exponential map associated to $g$. $\mathrm{Let}\;I\;:\;F\rightarrow M$ $(x,u)\mapsto\text{exp}_{x}(u^{\prime}),$ where $u^{\prime}$ is a representative of $u$ in $N(x)$. Such a map $I$ is well defined on a neighborhood of the zero section of $F$. The existence of a suitable laminar structure $\mathcal{F}$ follows from Lemma 9.2. ∎ ## 2 Persistence of laminations ### 2.1 Preserved laminations A lamination $(L,\mathcal{L})$ embedded by $i$ into a manifold $M$ is _preserved_ by an endomorphism $f$ of $M$ if each embedded leaf of $\mathcal{L}$ is sent by $f$ into an embedded leaf of $\mathcal{L}$. This is equivalent to suppose the existence of an endomorphism $f^{}$ of $(L,\mathcal{L})$ such that the following diagram commutes: (1) $\begin{array}[]{rcccl}&&f&&\\\ &M&\rightarrow&M&\\\ i&\uparrow&&\uparrow&i\\\ &L&\rightarrow&L&\\\ &&f^{}&&\end{array}$ The endomorphism $f^{}$ is called the _pullback of $f$ via $i$_. When the lamination is only immersed by $i$, these two definitions are not equivalent. A lamination $(L,\mathcal{L})$ immersed by $i$ into a manifold $M$ is _preserved_ by an endomorphism $f$ of $M$ if there exists a _pull back of $f$ in $(L,\mathcal{L})$ via $i$_. That is an endomorphism $f^{}$ of $(L,\mathcal{L})$ such that the diagram (1) commutes. We notice that the pull back $f^{}$ has at least the minimum of the regularities of $f$ and $i$, as soon as $f^{}$ exists and is continuous. The leaves of a lamination $(L,\mathcal{L})$ immersed by $i$ into a manifold $M$ are _preserved_ by an endomorphism $f$ of $M$ if the immersion of each leaf of $\mathcal{L}$ is sent by $f$ into the immersion of a leaf of $\mathcal{L}$. Clearly, if $f$ preserves an immersed lamination, then it preserves its leaves. We give in [Ber] two examples of diffeomorphisms preserving the leaves of an immersed lamination but not the lamination. ### 2.2 Persistence of laminations Let us fix $r\in[\\![1,\infty[\\![\cup\\{\mathcal{H}\\}$. Let $(L,\mathcal{L})$ be a lamination $C^{r}$-embedded by $i$ into a manifold $M$. Let $f$ be a $C^{r}$-endomorphism of $M$ which preserves $\mathcal{L}$. The _embedded lamination_ $(L,\mathcal{L})$ is _$C^{r}$ -persistent_ if for any endomorphism $f^{\prime}$ $C^{r}$-close to $f$, there exists an embedding $i^{\prime}$ $C^{r}$-close to $i$ such that $f^{\prime}$ preserves the lamination $(L,\mathcal{L})$ embedded by $i^{\prime}$ and such that each point of $i^{\prime}(L)$ is sent by $f^{\prime}$ into the image by $i^{\prime}$ of a small plaque containing $f(x)$. This implies that the pullback $f^{\prime}$ of $f^{\prime}$ is equivalent and $C^{r}$-close to the pullback $f^{}$ of $f$. Let $(L,\mathcal{L})$ be a lamination immersed by $i$ into a manifold $M$. Let $f$ be a $C^{r}$-endomorphism of $M$ which preserves $\mathcal{L}$. Let $f^{}$ be a pull back of $f$ in $(L,\mathcal{L})$. The _immersed lamination_ $(L,\mathcal{L})$ is $C^{r}$-_persistent_ if for any endomorphism $f^{\prime}$ $C^{r}$-close to $f$, there exists an immersion $i^{\prime}$ $C^{r}$-close to $i$, such that $f^{\prime}$ preserves the lamination $(L,\mathcal{L})$ immersed by $i^{\prime}$, and has a pullback $f^{\prime}$ in $(L,\mathcal{L})$ equivalent and $C^{r}$-close to $f^{}$. In other words, for every $f^{\prime}\in End^{r}(M)$ close to $f$ there exists $i^{\prime}\in Im^{r}(\mathcal{L},M)$ and $f^{\prime}\in End_{f^{}}^{r}(\mathcal{L})$ close to respectively $i$ and $f^{}$ such that the following diagram commutes: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ i^{\prime}&\uparrow&&\uparrow&i^{\prime}\\\ &L&\rightarrow&L&\\\ &&f^{\prime}&&\end{array}$ In the above definitions, the topologies of the spaces $End^{r}(M)$, $Im^{r}(\mathcal{L},M)$, $Emb^{r}(\mathcal{L},M)$ and $End_{f^{}}^{r}(\mathcal{L})$ are described in section 1.4. ### 2.3 Normal hyperbolicity Let $(L,\mathcal{L})$ be a lamination imbedded by $i$ into be a Riemannian manifold $(M,g)$. Let $f$ be an endomorphism of $M$ which preserves the immersion $i$ of $(L,\mathcal{L})$. Let $f^{}$ be a pullback of $f$. We identify, via the immersion $i$, the bundle $T\mathcal{L}\rightarrow L$ to a subbundle of $\pi\;:\;i^{}TM\rightarrow L$. We remind that the fiber of $i^{}TM$ at $x\in L$ is $T_{i(x)}M$. ###### Definition 2.1. For every $r\geq 0$, we say that $f$ _is $r$-normally hyperbolic to the lamination $(L,\mathcal{L})$ (immersed by $i$ over $f^{}$)_, if the following conditions hold: * • there exists a splitting $i^{}TM=E^{s}\oplus T\mathcal{L}\oplus E^{u}$ such that $E^{s}$ is $i^{}Tf$-stable: $Tf$ sends $E^{s}_{x}$ into $E^{s}_{f^{}(x)}$, for every $x\in L$, • there exist a norm on $i^{}TM$ and $\lambda<1$ such that, for any unit vectors $v_{s}\in E^{s}_{x}$, $v_{c}\in E^{c}_{x}$ and $v_{u}\in E^{u}_{x}$, we have: $\\|Tf(v_{s})\\|<\lambda\cdot\min(1,\\|Tf(v_{c})\\|^{r}),$ $\\|p_{u}\circ Tf(v_{u})\\|>\lambda^{-1}\cdot\max(1,\\|Tf(v_{c})\\|^{r}),$ where $p_{u}$ is the projection of $i^{}TM_{f^{}(x)}=T_{i\circ f^{}(x)}M$ onto $E^{u}$, parallelly to $E^{s}\oplus T\mathcal{L}$. When $E^{u}$ has dimension zero, we say that $f$ $r$-_normally contracts_ $(L,\mathcal{L})$. When $E^{s}$ has dimension zero, we say that $f$ $r$-_normally expands_ $(L,\mathcal{L})$. ###### Remark 2.2. Usually, one defines the normal hyperbolicity without changing the metric, but by looking some iterate of the dynamics. The definition stated above is actually more general. It is even strictly more general when the lamination is non-compact. Both of them are strictly more general that the one of HPS, even if their proof works as well for the usual definition. ###### Remark 2.3. When $f^{}$ is bijective, the subbundle $E^{u}$ of $i^{}TM$ can be chosen $i^{}Tf$-invariant: $Tf(E^{u}_{x})=E^{u}_{f^{}(x)}$, for every $x\in L$. ###### Remark 2.4. For the non-compact case, sometime we need to generalize the concept of normal hyperbolicity as follow. Let $f$ be an endomorphism of a manifold $M$ and let $(L,\mathcal{L})$ be a lamination immersed by $i$ into $M$. We say that $f$ is _$r$ -normally hyperbolic_ to $(L,\mathcal{L})$ over a morphism $f^{}$ from the restriction of $(L,\mathcal{L})$ to an open subset $D\subset L$ to $\mathcal{L}$ if: • the following diagram commutes $\begin{array}[]{rcccl}&&f&&\\\ &M&\rightarrow&M&\\\ i&\uparrow&&\uparrow&i\\\ &L^{\prime}&\rightarrow&L&\\\ &&f^{}&&\end{array}$, • the properties of Definition 2.1 are satisfied. ###### Property 2.5. Let $(L,\mathcal{L})$ be a lamination immersed by $i$ into a manifold $M$ and $1$-normally hyperbolic. Then, for all $x$, $y\in L$ with the same images by $i$, the spaces $E^{s}_{x}$ and $E^{s}_{x}\oplus T_{x}\mathcal{L}$ have the same images by $Ti$ as respectively $E^{s}_{y}$ and $E^{s}_{y}\oplus T_{y}\mathcal{L}$. Thus we can denote by respectively $E^{s}_{i(x)}$ and $(T\mathcal{L}\oplus E^{s})_{i(x)}$ the subspaces of $T_{i(x)}M$ corresponding to $E^{s}_{x}$ and $T_{x}\mathcal{L}\oplus E^{s}_{x}$, via the canonical identification of $i^{}TM_{x}$ with $T_{i(x)}M$. Moreover, for every compact subset $K$ of $L$, the section of the Grassmannian $z\in i(K)\rightarrow(E^{s}\oplus T\mathcal{L})_{z}$ is continuous. ###### Proof. See [Ber] Property 2.1.7.∎ ###### Remark 2.6. There exist compact laminations $(L,\mathcal{L})$ that are immersed by some $i$, normally contracted (and so normally hyperbolic), with bijective pullback, but for which there are two different points $x$ and $y$ in $L$ whose images by $i$ are the same whereas the images of $T_{x}\mathcal{L}$ and $T_{y}\mathcal{L}$ by $Ti$ are different. For instance, the following map of the cylinder $\mathbb{R}\times\mathbb{S}^{1}$: $\phi\;:\;(x,\theta)\mapsto\Big{(}\frac{x}{2}+\frac{1}{2}\cdot\sin(\theta),2\cdot\theta\Big{)}$ preserves such an immersed solenoid around $\\{0\\}\times\mathbb{S}^{1}$; we represent it in figure 1 by identifying the cylinder to the punctured plane. Figure 1: An attractor of the punctured plane. ###### Example 2.7 (Suspension of a holomorphic diffeomorphism). Let $H\;:\;(z,z^{\prime})\in\mathbb{C}^{2}\rightarrow(z^{2}+c+bz^{\prime},z)\in\mathbb{C}^{2}$ be a complex Hénon map which lets invariant a compact subset $K$ of $\mathbb{C}^{2}$ ($H(K)=K$) and is hyperbolic at $K$. Such a compact subset can be the Julia set of $K$ when $\|c\|>2(\|b\|+1)^{2}$ ($H$ is a horseshoe). Let $M$ be the quotient of $\mathbb{C}\setminus\\{0\\}\times\mathbb{C}^{2}$ by the proper, discontinuous and fixed-point-free group action: $\mathcal{G}:=\Big{\\{}g_{n}\;:\;g_{n}(x,z,z^{\prime})\mapsto(2^{n}\cdot x,H^{n}(z,z^{\prime})),n\in\mathbb{Z}\Big{\\}}.$ The complex $1$-lamination $(\tilde{L},\tilde{\mathcal{L}}):=\mathbb{C}\setminus\\{0\\}\times K$ is $\mathcal{G}$-equivariant and so its image by the projection $\mathbb{C}\setminus\\{0\\}\times\mathbb{C}^{2}\rightarrow M$ is a compact (complex) lamination $(L,\mathcal{L})$. The endomorphism $\tilde{h}:=(x,z,z^{\prime})\mapsto(a\cdot x,H(z,z^{\prime}))$ preserves and is normally hyperbolic to $(\tilde{L},\tilde{\mathcal{L}})$, for any $a\in\mathbb{C}\setminus\\{0\\}$. Moreover, $\tilde{h}$ is $\mathcal{G}$-equivariant. Thus $\tilde{h}$ defines a diffeomorphism $h$ of $M$ which is $r$-normally hyperbolic to $(L,\mathcal{L})$, for any $r\geq 0$. By Theorem 0.2, this (immersed) lamination is $C^{r}$-persistent for any $r\geq 1$. All of the above laminations are persistent as embedded laminations since the dynamics are plaque-expansive. Let us recall the definition of plaque- expansiveness. ### 2.4 Plaque-expansiveness ###### Definition 2.8 (Pseudo-orbit which respect $\mathcal{L}$). Let $(L,\mathcal{L})$ be a lamination and $f$ be an endomorphism of $(L,\mathcal{L})$. Let $\epsilon$ be a positive continuous function on $L$. An _$\epsilon$ -pseudo-orbit (resp. backward-pseudo-orbit, resp. forward-pseudo- orbit) which respects $\mathcal{L}$_ is a sequence $(x_{n})_{n}\in L^{\mathbb{Z}}$ (resp. $(x_{n})_{n}\in L^{\mathbb{Z}^{-}}$, resp. $(x_{n})_{n}\in L^{\mathbb{N}}$) such that, for any $n$, the point $f(x_{n})$ belongs to a plaque of $\mathcal{L}$ containing $x_{n+1}$ whose diameter is less than $\epsilon(x_{n+1})$. ###### Definition 2.9 (Plaque-expansiveness). Let $\epsilon>0$. The endomorphism $f$ is _$\epsilon$ -plaque-expansive (resp. backward-plaque-expansive, resp. forward-plaque-expansive) at $(L,\mathcal{L})$_ if for any positive function $\eta$ less than $\epsilon$, for all $\eta$-pseudo-orbits $(x_{n})_{n}$ and $(y_{n})_{n}$ which respect $\mathcal{L}$, such that for any $n$ the distance between $x_{n}$ and $y_{n}$ is less than $\eta(x_{n})$, then $x_{0}$ and $y_{0}$ belong to a same small plaque of $\mathcal{L}$. ###### Remark 2.10. We remark that every map $f^{\prime}\in End_{f}(\mathcal{L})$ close enough to $f^{}$ is plaque-expansive (resp. backward-plaque-expansive, resp. forward- plaque-expansive). ###### Remark 2.11. Forward or backward plaque-expansiveness implies plaque-expansiveness. We do not know if normal hyperbolicity implies plaque-expansiveness, even when $L$ is compact. But in many cases this is true. Let us recall the results on plaque-expansiveness in the diffeomorphism case. Let $f$ be a diffeomorphism which preserves and is normally hyperbolic to a compact lamination $(L,\mathcal{L})$. If one of the following situations occur, then $f$ is plaque-expansive at $\mathcal{L}$. * • [HPS77] the leaves of the lamination are the fibers of a bundle, * • [HPS77] the lamination is locally a saturated subset of $C^{1}$-foliation, * • [RHRHU07] if for any $\epsilon>0$, there exists $\delta>0$ such that for any $x\in L$, $f^{n}(\mathcal{L}_{x}^{\delta})$ is included in $\mathcal{L}_{f^{n}(x)}^{\epsilon}$, for any $n\in\mathbb{Z}$ (or $n\geq 0$ in the normally expanded case). In [Ber], we establish a generalization of the last result in the endomorphism context: we suppose the lamination normally expanded. Then if there exist $\epsilon>0$ and $\delta>0$ such that: * • for every $x$ the subset $\mathcal{L}_{x}^{\epsilon}$ is precompact in its leaf, * • for any $x\in L$, the map $f^{n}$ sends $\mathcal{L}_{x}^{\delta}$ into $\mathcal{L}_{f^{n}(x)}^{\epsilon}$, for any $n\geq 0$. Then the endomorphism $f$ is forward plaque-expansive. ## 3 Main theorem on differentiable persistence ### 3.1 Statement The following is our main theorem on $C^{r}$-persistence of non compact laminations; it implies Theorems 0.1 and 0.2. ###### Theorem 3.1. Under the hypotheses of Theorem 0.1 (resp. 0.2) except that the lamination $(L,\mathcal{L})$ is not necessarily compact, for every precompact open subset $L^{\prime}$ of $L$, for every $f^{\prime}$ $C^{r}$-close to $f$ there exist an immersion $i(f^{\prime})$ of $(L,\mathcal{L})$ into $M$ satisfying: 1. 1. there exists a $C^{r}$-morphism $f^{\prime}$ from $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$ into $(L,\mathcal{L})$ such that the following diagram commutes: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ i(f^{\prime})&\uparrow&&\uparrow&i(f^{\prime})\\\ &L^{\prime}&\rightarrow&L&\\\ &&f^{\prime}&&\end{array}$ 2. 2. the morphism $f^{\prime}$ is equivalent and $C^{r}$-close to the restriction $f^{}_{\|L^{\prime}}$ of a pullback $f^{}$ of $f$; the immersion $i^{\prime}$ is $C^{r}$-close and is equal to $i$ on the complement of a compact subset of $L$ independent of $f^{\prime}$. 3. 3. There exists $\epsilon>0$ such that for any $f^{\prime}$-orbit $(y_{n})_{n\geq 0}\in M^{\mathbb{N}}$ (resp. $(y_{n})_{n}\in M^{\mathbb{Z}}$) which is $\epsilon$-close to the image by $i$ of an $\epsilon$-pseudo-orbit $(x_{n})_{n}\in L^{\prime\mathbb{N}}$ (resp. $(x_{n})_{n}\in L^{\prime\mathbb{Z}}$) of $f^{}$, respecting the plaques, then there exists $z_{0}$ uniformly close to $x_{0}$ in its leaf sent to $y_{0}$ by $i(f^{\prime})$. ###### Remark 3.2. Conclusion (3) implies the uniqueness of $(i(f^{\prime}),f^{\prime})$ in Theorems 0.1 and 0.2, up to reparametrization. ###### Remark 3.3. This theorem implies the persistence of immersed compact laminations stated in Theorems 0.1 and 0.2 by taking $L^{\prime}=L$. The persistence of compact embedded laminations, given in these same theorems, is consequence of Corollary 3.9 (see below). In the normally contracting case, we have the following similar theorem: ###### Theorem 3.4. Let $r\geq 1$. Let $(L,\mathcal{L})$ be a lamination $C^{r}$-immersed by $i$ into a manifold $M$. Let $D$ be an open subset of $L$. Let $f$ be a $C^{r}$-endomorphism of $M$ which $r$-normally contracts the lamination $(L,\mathcal{L})$ over an injective, open immersion $f^{}$ from $(D,\mathcal{L}_{\|D})$ into $(L,\mathcal{L})$. Then for any open precompact subset $L^{\prime}$ of $D$ and for every $f^{\prime}$ $C^{r}$-close to $f$, there exist an immersion $i(f^{\prime})$ of $(L,\mathcal{L})$ into $M$ and a $C^{r}$-morphism $f^{\prime}$ of $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$ into $(L,\mathcal{L})$ satisfying conclusions 1 and 2 of Theorem 3.1. Conclusion $(3)$ is replaced by $(3^{\prime})$ there exists a real number $\epsilon>0$ such that for any $f^{\prime}$-preorbit $(y_{n})_{n\leq 0}\in M^{\mathbb{Z}^{-}}$ which is $\epsilon$-close to the image by $i$ of an $\epsilon$-pseudo-preorbit $(x_{n})_{n\leq 0}\in M^{\mathbb{Z}^{-}}$ of $f^{}$ respecting the plaques, there exists then $z_{0}$ close to $x_{0}$ in its leaf sent to $y_{0}$ by $i(f^{\prime})$. ###### Remark 3.5. The proofs of these Theorems do not use the fact that $f$ and its perturbation are defined on the complement of a neighborhood of $i(cl(L^{\prime}))$. Thus, these theorems remain true if $f$ is a $C^{r}$-map from a neighborhood $U$ of $i(cl(L^{\prime}))$ into $M$, for any $C^{r}$-perturbation $f^{\prime}\in C^{r}(U,M)$ (except that the domain of $f^{\prime}$ in the diagram has to be replaced by $U$). Also in the normally expanded case, we do not need $f^{}$ to be defined outside of any neighborhood of $cl(L^{\prime})$. ### 3.2 Non-compact Examples ###### Example 3.6. Let $r\geq 1$ and let $f_{1}$ be a $C^{r}$ diffeomorphism of a manifold $N_{1}$. Let $\Lambda$ be a hyperbolic compact subset. Then, by Example 1.4, $W^{s}(\Lambda)$ is an invectively $C^{r}$-immersed lamination $(L_{1},\mathcal{L}_{1})$ whose leaves are stable manifolds. Let $E_{u}$ be the unstable direction of $\Lambda$ and let $m:=\min_{u\in E_{u}\setminus\\{0\\}}\frac{\\|Tf(u)\\|}{\\|u\\|}.$ We may suppose $m>1$. Let $M_{2}$ be a compact Riemannian manifold and let $f_{2}$ be a $C^{r}$-endomorphism of $M_{2}$ whose differential has norm less than $\sqrt[r]{m}$ (hence $f_{2}$ has possibly many singularities and is not necessarily bijective). Thus, the product dynamics $f:=(f_{1},f_{2})$ on $M:=M_{1}\times M_{2}$ $r$-normally expands the $C^{r}$-immersed lamination $(L,\mathcal{L}):=(L_{1}\times M_{2},\mathcal{L}_{1}\times M_{2})$ over a pullback $f^{}$. Let $L^{\prime}$ be a precompact, open subset of $L$, whose closure is sent into itself by $f^{}$ (there exists arbitrarily big such subsets). By Theorem 3.1, the $C^{r}$-immersed lamination $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$ is $C^{r}$-persistent, since $f^{\prime}(L^{\prime})$ is included in $L^{\prime}$ for $f^{\prime}$ $C^{r}$-close to $f$. Actually, this lamination is forward plaque-expansive and we will see that this implies its persistence as an embedded lamination. ###### Example 3.7. Let $f\;:(z,z^{\prime})\in\mathbb{C}^{2}\mapsto(z(z+e^{i\theta}),4z^{\prime})\in\mathbb{C}^{2}$, with $\theta\in\mathbb{R}$. The endomorphism $1$-normally expands the immersed submanifold $\mathbb{C}\hookrightarrow\mathbb{C}\times\\{0\\}$ (for some special Riemannian metric). Let $L^{\prime}$ be a neighborhood of the filled Julia set of the polynomial function $z\mapsto z(z+1)$. By Theorem 3.1, for every $f^{\prime}$ $C^{1}$-close to $f$, there exists a $C^{r}$-immersion $i(f^{\prime})$ of $\mathbb{C}\times\\{0\\}$ into $\mathbb{C}^{2}$ such that any point $x\in L^{\prime}$ has its image by $i(f^{\prime})$ sent by $f^{\prime}$ to the image by $i(f^{\prime})$ of some point $y\in\mathbb{C}\times\\{0\\}$. Let $K_{f^{\prime}}$ be the set consisting of the points of $\mathbb{C}^{2}$ with bounded $f^{\prime}$-orbits. One easily shows that $K_{f^{\prime}}$ is included in any fixed neighborhood $W$ of $L^{\prime}\times\\{0\\}$, for $f^{\prime}$ close enough to $f$. Thus, by conclusion (3) of Theorem 3.1, $K_{f^{\prime}}$ is included in $i^{\prime}(L^{\prime})$. It is well known that for some perturbation $f^{\prime}$ of $f$, the subset $K_{f^{\prime}}$ is a Cantor set. ###### Example 3.8. In Example 2.7 we recalled that for a large open subset $L$ of parameters $(b,c)\in\mathbb{C}^{2}$, the Hénon map $H_{c,b}(z,z^{\prime})\mapsto(z^{2}+c+b\cdot z^{\prime},z)$ has an uniformly hyperbolic horseshoe. This means that $H_{c,b}$ preserves a hyperbolic compact subset $K_{c,b}$ and its restriction to $K_{c,b}$ is conjugated to a shift map on the set $\Sigma_{2}:=\\{0,1\\}^{\mathbb{Z}}$. Let $\mathbb{A}$ be an annulus (with finite modulus) holomorphically immersed into $L$. Let $M$ be equal to the product of $\mathbb{A}$ with $\mathbb{C}^{2}$. Let $f\;:\;(t,z,z^{\prime})\in M\mapsto\big{(}t,H_{t}(z,z^{\prime})\big{)}\in M$. By the hyperbolic continuation theorem (or Theorem 0.6), the subset $\cup_{t\in\mathbb{A}}K_{t}$ is canonically endowed with a structure of complex one-dimensional lamination $(L,\mathcal{L})$, whose structure is given by the hyperbolic continuation theorem. The diffeomorphism $f$ preserves this lamination and is $r$-normally hyperbolic to it, for any $r\geq 1$. We endow $(L,\mathcal{L})$ with the holomorphic tubular neighborhood $(F,\mathcal{F},I,\pi)$ induced by the bundle $\mathbb{A}\times\mathbb{C}^{2}\rightarrow\mathbb{A}$. Let $\mathbb{A}^{\prime}$ be an open precompact subset of $\mathbb{A}$ preserved by any complex automorphism of $\mathbb{A}$. Thus $\mathbb{A}^{\prime}$ is also an annulus. Let $L^{\prime}$ be the open precompact set equal to $L\cap(\mathbb{A}^{\prime}\times\mathbb{C}^{2})$. By Theorem 3.1, for any diffeomorphism $f^{\prime}$ $C^{r}$-close to $f$, there exists an immersion $i(f^{\prime})$ $C^{r}$-close to the canonical inclusion, respecting the fibers of $\mathbb{A}\times\mathbb{C}^{2}\rightarrow\mathbb{A}$, and such that any point $x\in L^{\prime}$ has its image by $i(f^{\prime})$ sent by $f^{\prime}$ into the image by $i(f^{\prime})$ of a small plaque of $f(x)$. If $f^{\prime}$ is holomorphic then it is of the form $f^{\prime}(t,z,z^{\prime})=(g(t),H^{\prime}_{t}(z,z^{\prime}))$, with $g$ a complex automorphism of $\mathbb{A}$. Thus $g(\mathbb{A}^{\prime})$ is equal to $\mathbb{A}^{\prime}$ and $f^{\prime}(\mathbb{A}^{\prime})$ is also equal to $\mathbb{A}$. Consequently, $f^{\prime}$ preserves the lamination $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$ $C^{r}$-immersed by $i(f^{\prime})$. ### 3.3 Plaque-expansiveness implies injectivity ###### Corollary 3.9. Under the hypotheses of Theorem 3.1 in the normally hyperbolic case (resp. normally expanded case, resp. under the hypotheses of Theorem 3.4), if $i$ is an embedding and $f^{}$ is plaque-expansive (resp. forward plaque-expansive, resp. backward plaque-expansive), then the restriction of $i(f^{\prime})$ to $\Lambda^{}:=\cap_{n\in\mathbb{Z}}f^{\prime^{n}}(cl(L^{\prime}))$ (resp. $\cap_{n\geq 0}f^{\prime^{n}}(cl(L^{\prime}))$, resp. $\cap_{n\leq 0}f^{\prime^{n}}(cl(L^{\prime}))$) is a homeomorphism, for every $f^{\prime}$ $C^{r}$-close to $f$. ###### Proof. As the three proofs are very similar, we shall only show the one of Theorem 3.1 in the normally hyperbolic case. As $\Lambda^{}$ is compact, we can take the function $\epsilon$ (of the plaque-expansiveness definition) equal to the constant $\epsilon_{0}:=\min_{\Lambda^{}}\epsilon$. For $V_{f}$ sufficiently small, for any $f^{\prime}\in V_{f}$, the endomorphism $f^{\prime}$ sends every point $x\in L^{\prime}$ into the plaque $\mathcal{L}^{\epsilon_{0}}_{f^{}(x)}$ containing $f^{}(x)$ and with diameter less than $\epsilon_{0}$. On the other hand, the following map is continuous: $\phi\;:\;V_{f}\rightarrow\mathbb{R}$ $f^{\prime}\mapsto\min\big{\\{}d\big{(}i^{\prime}(x),i^{\prime}(y)\big{)},\;(x,y)\in\Lambda^{^{2}}\;d(x,y)>\epsilon_{0}\big{\\}},\quad\mathrm{with}\;i^{\prime}:=i(f^{\prime}).$ As $\phi$ is positive at $f$, by restricting $V_{f}$, we may suppose that $\phi$ is positive on $V_{f}$. Consequently, for every $f^{\prime}\in V_{f}$, if $x,y\in\Lambda^{}$ are sent by $i^{\prime}$ to a same point, then they are $\epsilon_{0}$-distant. By commutativity of the diagram of Theorem 3.1, for all $n\geq 0$, $f^{\prime^{n}}(x)$ and $f^{\prime^{n}}(y)$ are sent by $i^{\prime}$ to a same point and hence, are $\epsilon_{0}$-distant. As we firstly notice, $(f^{\prime^{n}}(x))_{n}$ and $(f^{\prime^{n}}(y))_{n}$ are two $\epsilon_{0}$-pseudo-orbits. Thus, by plaque-expansiveness, $x$ and $y$ belong to a same small plaque. As $i^{\prime}$ is an immersion which depends continuously on $f^{\prime}$, we may suppose that $i^{\prime}$ is injective on such small plaques. Thus, $x$ and $y$ are equal and $i^{\prime}$ is injective. By compactness of $\Lambda^{}$ and continuity of $i^{\prime}$, the map $i^{\prime}_{\|\Lambda^{}}$ is a homeomorphism onto its image. Thus, $i^{\prime}$ is an embedding.∎ ## 4 Space of preorbits The following proposition provides a way to construct bijective pullbacks in order to use Theorems 0.2 and 3.1 on persistence of normally hyperbolic laminations and Theorem 3.4 on persistence of normally contracted laminations. ###### Proposition 4.1. Let $r\in[\\![1,\infty[\\![\cup\\{{\mathcal{H}}\\}$ and let $(L,\mathcal{L})$ be a lamination of class $C^{r}$. Let $D$ be an open subset of $L$ and let $f^{}$ be a proper, open $C^{r}$-immersion from $\mathcal{L}_{\|D}$ into $\mathcal{L}$. Let $\tilde{L}$ be the preorbits space of $f$: $\tilde{L}:=\\{(x_{i})_{i\geq 0}\in L^{\mathbb{N}}\;:\;f(x_{i+1})=x_{i},\;\forall i\\}.$ Then there exists a canonical $C^{r}$-lamination structure $\tilde{\mathcal{L}}$ on $\tilde{L}$ such that the canonical projection $\pi\;:\;(x_{i})_{i}\in\tilde{L}\mapsto x_{0}\in L$ is an injective, open $C^{r}$-immersion and the map $(x_{i})_{i}\in\pi^{-1}(D)\mapsto(f(x_{i}))_{i}$ is a $C^{r}$-immersion onto $\tilde{L}$. ###### Proof. Let $(U,\phi)\in\mathcal{L}$ be a chart of the form $\phi\;:\;U\rightarrow W\times T$, with $U$ a precompact, open subset of $L$, $T$ a locally compact metric space and $W$ a simply connected, open subset of $\mathbb{K}^{n}$. We want to find a metric space $\tilde{T}$ and a homeomorphism $\tilde{\phi}\;:\;\pi^{-1}(U)\rightarrow W\times\tilde{T}$ such that the following diagram commutes: $\begin{array}[]{rccl}&\pi^{-1}(U)&\tilde{\phi}_{1}&\\\ \phi&\downarrow&\searrow&\\\ &U&\rightarrow&W\\\ &&\phi_{1}&\end{array}$ where $\phi_{1}$ and $\tilde{\phi}_{1}$ are the first coordinates of respectively $\phi$ and $\tilde{\phi}$. Let $(x_{n})_{n}\in\tilde{L}^{\mathbb{N}}$ be such that $x_{0}$ belongs to $U$. For each $n\leq 0$, let $U_{n}$ be the union of plaques of $x_{n}$ included in $f^{-n}(U)$. We notice that $U_{n}$ is the connected component of $x_{n}$ in the intersection of $f^{-n}(U)$ with the leaf of $x_{n}$. Let us show that the restriction of $f^{n}$ to $U_{n}$ is a homeomorphism onto $U_{0}$. First of all, $f^{-n}(U)$ is an open subset of $L$ and so $U_{n}$ is a connected open subset of the leaf of $x_{n}$. Let us show that $f^{n}_{\|U_{n}}$ is injective. Let $y,y^{\prime}\in U_{n}$ be sent by $\phi_{1}\circ f^{n}$ to a same point $z\in U_{0}$. By connectedness of $U_{n}$, there exists a path in $U_{n}$ from $y$ to $y^{\prime}$. As $U_{0}\cong W$ is simply connected, the image by $f^{n}$ of this path is homotopic to the trivial path $\\{z\\}$. As $f^{n}$ is an immersion, we can pull back this homotopy to $U_{N}$. This shows that $y$ and $y^{\prime}$ are equal. As the differential of $f^{n}$ along the leaves at each point of $U_{n}$ is an isomorphism, it follows from the global inversion theorem that the restriction of $f^{n}$ to $U_{n}$ is a diffeomorphism onto its image which is open. To show that its image is equal to $U_{0}$, by connectedness of $U_{0}$, it only remains to prove that $f^{n}(U_{n})$ is a closed subset of $U_{0}$. Let $z\in cl(f^{n}(U_{n}))\cap U_{0}$. Then there exists a sequence $(y_{k})_{k}\in U_{n}^{\mathbb{N}}$ such that the sequence $(f^{n}(y_{k}))_{k}$ converges to $z$. Let us admit that $U_{n}$ is precompact in the leaf of $x_{n}\in\mathcal{L}_{\|D}$. Then we may suppose that $(y_{k})_{k}$ converges to some $y$ in the closure of $U_{n}$ in the leaf of $x_{n}\in\mathcal{L}_{\|D}$. By continuity of $f$, the image by $f^{n}$ of $y$ is $z$. By openness of $f$ and $U_{0}$, there exists a small plaque $\mathcal{L}_{y}$ containing $y$ in the leaf of $x_{n}$, which is sent into $U_{0}$. Since $(y_{n})_{n}$ converges to $y$ in the leaf of $x_{n}$, the union of $U_{n}$ with $\mathcal{L}_{y}$ is connected. As this union is sent into $U_{0}$, this union is contained in $U_{n}$. Hence $y$ belongs to $U_{0}$ and $z$ belongs to $U_{0}$. Therefore, $f^{n}(U_{n})$ is closed in $U_{0}$, and so $f_{\|U_{n}}$ is a homeomorphism onto $U_{0}$. Let us show that $U_{n}$ is precompact in the leaf of $x_{n}$. We endow $\mathcal{L}$ and $\mathcal{L}_{\|D}$ with respectively two (possibly different) complete metrics. Then, for these Riemannian distances, the bounded and closed subsets of leaves of $\mathcal{L}$ and $\mathcal{L}_{\|D}$ are compact. Thus, to show that $U_{n}$ is precompact in the leaf of $x_{n}$ in $\mathcal{L}_{\|D}$, it is sufficient to show that $U_{n}$ is bounded for the complete distance of $\mathcal{L}_{\|D}$. We note that: $\text{diam}\>_{\mathcal{L}_{\|D}}U_{k+1}\leq\sup_{U_{k+1}}\\|Tf^{-1}\\|\cdot\text{diam}\>_{\mathcal{L}}U_{k},$ where $\text{diam}\>_{\mathcal{L}_{\|D}}$ and $\text{diam}\>_{\mathcal{L}}$ refer to the diameters with respect to the complete metrics of respectively $\mathcal{L}_{\|D}$ and $\mathcal{L}$. We remind that the finiteness of $\text{diam}\>_{\mathcal{L}_{\|D}}U_{k}$ is equivalent to its precompactness in a leaf of $\mathcal{L}_{\|D}$, and so its precompactness in a leaf of $\mathcal{L}$. Moreover, for $k\geq 1$, by precompactness of $U_{0}$ and properness of $f$, the subset $U_{k}$ is included in $f^{-k}(U_{0})$ which is precompact in $D$. Thus $\sup_{U_{k}}\\|Tf^{-1}\\|$ is finite. Using these two last remarks, an induction proves that, for $k\geq 1$, $U_{k}$ is precompact in a leaf of $\mathcal{L}_{\|D}$. Let $T_{n}$ be the set of the connected components of each intersection of $f^{-n}(U)$ with some leaf of $\mathcal{L}$. For instance, $T_{0}$ is the set $T$. As the elements of $T_{n}$ are homeomorphic to $W$ and hence are locally compact, we can endow $T_{n}$ with the following distance: $d_{T_{n}}\;:\;(t,t^{\prime})\in T_{n}\mapsto\sup_{x\in t}d(x,t^{\prime})+\sup_{x^{\prime}\in t^{\prime}}d(x^{\prime},t),$ where the metric $d$ refers to the one of the metric space $L$. We notice that $f$ induces a continuous map from $T_{n+1}$ to $T_{n}$, since $f$ is uniformly continuous on $U_{n+1}$. Thus, we shall consider the projective limit: $\tilde{T}:=\lim_{\leftarrow}T_{n}.$ The topology of $\tilde{T}$ is given, for instance, by the following metric: $d_{\tilde{T}}((t_{n})_{n\geq 0},(t^{\prime}_{n})_{n\geq 0})=\sum_{n\geq 0}\frac{\min(d_{T_{n}}(t_{n},t^{\prime}_{n}),1)}{2^{n}}.$ For such a metric $\tilde{T}$ is locally compact. Therefore, the following map satisfies the requested properties: $\tilde{\phi}\;:\;\pi^{-1}(U)\rightarrow W\times\tilde{T}$ $\underline{x}=(x_{n})_{n\geq 0}\mapsto(\tilde{\phi}_{1}(x_{0}),(t_{n}(x))_{n}),$ where $t_{n}(x)$ is the point of $T_{n}$ containing $x_{n}$. The above map is well defined and is a homeomorphism since the projective limit $\tilde{L}$ has its topology given by the following metric: $\tilde{d}:\;\big{(}(x_{n})_{n},(y_{n})_{n}\big{)}\mapsto\sum_{n\geq 0}\frac{\min\big{(}d(x_{n},y_{n}),1\big{)}}{2^{n}}$ For such metric space $(\tilde{L},\tilde{d})$ is secondly countable and locally compact. Also the family $(\pi^{-1}(U),\tilde{\phi})_{U}$ constructed as above forms an atlas $\tilde{\mathcal{L}}$ of laminations on $(\tilde{L},\tilde{d})$. ∎ ###### Example 4.2. Let $P$ be a polynomial function of $\mathbb{C}$. Let $C_{0}$ be the closure of the forward orbits of the critical values. Let $L_{0}$ be the nonempty open subset $\mathbb{C}\setminus C_{0}$ and $D_{0}$ be the preimage of $L_{0}$. The open set $L_{0}$ is canonically endowed with the complex one-dimensional lamination structure consisting of a single leaf. We remark that the restriction of $P$ to $D_{0}$ satisfies the hypotheses of the above proposition. Therefore, the set: $L:=\\{(z_{i})_{i\leq 0}\in L^{\mathbb{N}}\;:\>P(z_{i-1})=z_{i}\\}$ is endowed with a structure of lamination $\mathcal{L}$. $\mathrm{Let}\quad\pi\;:\;(z_{i})_{i}\in L\mapsto z_{0}\in L_{0}\quad\mathrm{and}\quad f^{}\;:\;(z_{i})_{i\leq 0}\in D\mapsto(P(z_{i}))_{i\leq 0}\in L,$ with $D$ be the preimage by $\pi$ of $D_{0}$. We notice that the maps $\pi$ and $f^{}$ are immersions for the structure induced by $\mathcal{L}$. The map $H\;:\;(z,z^{\prime})\mapsto(P(z)+z^{\prime},0)$ preserves and $r$-normally contracts the immersion $z\in L\mapsto(\pi(z),0)$ of $\mathcal{L}$ over $f^{}$, for any $r\geq 1$. The strong stable direction is the tangent space at $z^{\prime}=0$ of the foliation whose leaves are $\\{(z,z^{\prime}):\;z^{\prime}=-P(z)\\}$. Let $L^{\prime}_{0}$ be a precompact open subset of $D_{0}$ and let $L^{\prime}$ be the preimage of $L^{\prime}_{0}$ by $\pi$. By Theorem 3.4, for any $C^{r}$-perturbation $H^{\prime}$ of $H$, there exists an immersion $i(H^{\prime})$ of $(L,\mathcal{L})$ into $\mathbb{C}^{2}$, $C^{r}$-close to $i$, such that any point $z\in L^{\prime}$ has its image by $i(H^{\prime})$ sent by $H^{\prime}$ to the image by $i(H^{\prime})$ of a point $y\in L$. Moreover $y$ is close to $f^{}(x)$ in its leaf of $\mathcal{L}$. We notice that for $P(z)=z^{2}+c$, possible perturbations are the complex Hénon maps $H^{\prime}\;:\;(z,z^{\prime})\mapsto(z^{2}+z^{\prime},bz)$, for small $b\in\mathbb{C}$. Also conclusion $(3)$ of Theorem 3.4 implies that there exists an open neighborhood $W$ of $L^{\prime}_{0}\times\\{0\\}$ in $\mathbb{C}^{2}$ such that any point of $W$ is in the image of $i(H^{\prime})$ if it has a $H^{\prime}$-preorbit in $W$. This phenomenon was already known by Hubbard and Oberste-Vorth ([HOV94]) when $P$ is a hyperbolic map. ## 5 Persistence of complex laminations Before giving examples of our Theorems 0.3 and 0.6 on persistence of complex laminations by deformation, let us explain and illustrate how to use Theorems 3.1 and 3.4 on persistence of non-compact laminations in the complex analytic context. The problem is to transform a $C^{1}$ immersion given by these theorems to a holomorphic immersion. Obviously the $J$-invariance of the tangent space to its immersed leaves is necessarily, where $J$ is the automorphism of the tangent bundle of the manifold provided by its complex structure ($J^{2}=-1$). Conversely, the $1$-normal hyperbolicity of a lamination implies the $J$-invariance of the tangent space to its immersed leaves as stated in the following propositions: ###### Proposition 5.1. Under the hypotheses of Theorems 3.1 or 3.4 with $r=1$, if $i$ and $f$ are moreover holomorphic, then for any holomorphic perturbation $f^{\prime}$ of $f$, for any $\mathcal{L}$-admissible111See definition in Subsection 1.1 subset222See Corollary 3.9 for a more precise definition of $\Lambda^{}$ $A$ of $\Lambda^{}:=\cap_{n}f^{n}(L^{\prime})$, the tangent space to the immersed lamination $(A,\mathcal{L}_{\|A})$ by $i(f^{\prime})$ is $J$-invariant. ###### Proof. We only show the normal contracting case (Theorem 3.4), the proofs of the normal expanding or hyperbolic cases are very similar. We remind that a cone of a vector space $E$ is an open set invariant by nonzero and real scalar multiplications. As the immersion by $i$ of $(L,\mathcal{L})$ is $1$-normally contracted by $f$, for $f^{\prime}$-close enough to $f$, there exists a family of cones $(C_{x})_{x\in\Lambda^{}}$ such that $C_{x}$ is formed by vectors of $T_{i(f^{\prime})(x)}M$ satisfying the following conditions for every $x\in\Lambda^{}$: 1. 1. $\chi_{x}$ contains $Ti(f^{\prime})(T_{x}\mathcal{L})$, 2. 2. $Tf^{\prime}$ sends the closure of the cone $\chi_{f^{\prime-1}(x)}$ into $\chi_{x}\cup\\{0\\}$, 3. 3. $Ti(f^{\prime})(T_{x}\mathcal{L})$ is a maximal vector subspace included in $\chi_{x}$, As $i$ is holomorphic and close to $i(f^{\prime})$, we may also suppose that: 4. 4. $\chi(x)$ contains a $J$-invariant $\dim(\mathcal{L})$-plane. By using the Hilbert metric, one can show that assertions $2$ and $3$ imply that for any $x\in\Lambda^{}$, the push-forward of $\chi$ by $f$ converges to $Ti(f^{\prime})(T_{x}\mathcal{L})$: (2) $\cap_{n\geq 0}Tf^{n}(\chi_{f^{-n}(x)})=Ti(f^{\prime})(T_{x}\mathcal{L}).$ One the other hand, as $f^{\prime}$ is holomorphic, its tangent map $Tf$ commutes with $J$ ($Tf\circ J=J\circ Tf$). Thus assertions $2$ and $4$ imply that $\cap_{n\geq 0}Tf^{n}(\chi_{f^{-n}(x)})$ contains a $J$-invariant $\dim(\mathcal{L})$-plane. By equation (2), such a plane is necessarily $Ti(f^{\prime})(T_{x}\mathcal{L})$.∎ All the relevance of the above property follows from this basic proposition: ###### Proposition 5.2. Let $(L,\mathcal{L})$ be a lamination $C^{1}$-immersed into a complex manifold $(M,J)$. If for every $x\in L$, $Ti(T_{x}\mathcal{L})$ is $J$-invariant, then there exists a unique complex structure $\mathcal{L}_{\mathcal{H}}$ on $L$, which is compatible with the $C^{1}$-structure $\mathcal{L}$ and such that $i$ is a holomorphic immersion. ###### Proof. Uniqueness Let $\mathcal{L}_{H}^{1}$ and $\mathcal{L}_{H}^{2}$ be two complex structures compatible with $\mathcal{L}$ such that $i$ is holomorphic for both structures. Let $x\in L$. Let $\phi_{1}:\;U_{1}\rightarrow V_{1}\times T_{1}$ and $\phi_{2}:\;U_{2}\rightarrow V_{2}\times T_{2}$ be two charts of neighborhoods of $x$ in the structures $\mathcal{L}_{H}^{1}$ and $\mathcal{L}_{H}^{2}$ respectively. Let us denote by $J_{1}$ and $J_{2}$ the complex structures of $V_{1}\subset\mathbb{C}^{d}$ and $V_{2}\subset\mathbb{C}^{d}$ respectively. We have: $T_{x}i\circ T\phi_{1}\circ J_{1}=J\circ T_{x}i\circ T\phi_{1}\quad\mathrm{and}\quad T_{x}i\circ T\phi_{2}\circ J_{2}=J\circ T_{x}i\circ T\phi_{2}.$ Consequently: $(T_{x}\phi_{2})^{-1}\circ T_{x}\phi_{1}\circ J_{1}=J_{2}.$ In other worlds, the coordinate change $\phi_{2}^{-1}\circ\phi_{1}$ is holomorphic and so by maximality of $\mathcal{L}_{H}^{1}$ and $\mathcal{L}_{H}^{2}$, these two structures are equal. Existence Let us construct the holomorphic structure of $L$. Let $x\in L$. Via a chart of $M$, we may identify a neighborhood of $i(x)\in M$ to an open subset $V$ of $\mathbb{C}^{n}$, such that $i(x)$ is $0$ and $Ti(T_{x}\mathcal{L})$ is $(\mathbb{C}^{d}\times\\{0\\})\cap V$. Let $\phi:\;U\rightarrow W\times T$ be a chart of a neighborhood of $x\in L$. We may suppose $U$ sufficiently small such that $i(U)$ is included in $V$ and such that for every $t\in T$, the set $i\circ\phi^{-1}(W\times\\{t\\})$ is the graph of a function from an open subset $W_{t}$ of $\mathbb{C}^{d}$ into $\mathbb{C}^{n-d}$. By restricting $T$, we may suppose the existence of an open neighborhood $W^{\prime}$ of $0$ in the intersection $\cap_{t\in T}W_{t}$. Let $U^{\prime}$ be the open subset of $L$ formed by the point $x^{\prime}\in U$ such that the projection of $i(x^{\prime})$ by $p:\;\mathbb{C}^{n}\rightarrow\mathbb{C}^{d}\times\\{0\\}$ is in $W^{\prime}$. We notice that $U^{\prime}$ contains $x$ and that $\phi^{\prime}:\;U^{\prime}\rightarrow W^{\prime}\times T$ $x^{\prime}\mapsto(p\circ i(x),\phi_{2}(x^{\prime}))$ is a chart of $\mathcal{L}$, with $\phi_{2}$ the second coordinate of $\phi$. Therefore one can easily show that the family of maps $\phi^{\prime}$ constructed in such a way is a atlas of complex lamination. ∎ Therefore the complex structure of a $C^{1}$-persistent complex lamination may change, but for our purpose the complex structure remains the same if the lamination is endowed with a holomorphic tubular neighborhood: ###### Proposition 5.3. Let $(L,\mathcal{L})$ be a complex lamination immersed into a complex manifold $M$. Suppose that $(L,\mathcal{L})$ is endowed with a holomorphic tubular neighborhood $(F,\mathcal{F},I,\pi)$. Let $i^{\prime}$ be a $C^{1}$-immersion of $(L,\mathcal{L})$ equal to the composition of $I$ with a $C^{1}$-section of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$. If for every $x\in L$, $Ti^{\prime}(T_{x}\mathcal{L})$ is $J$-invariant, then $i^{\prime}$ is holomorphic. ###### Proof. Since $i^{\prime}$ is differentiable, it is sufficient to show that for any $x\in L$, for any $u\in T_{x}\mathcal{L}$: $T_{x}i^{\prime}(Ju)=J\circ T_{x}i^{\prime}(u).$ Since $i^{\prime}$ is equal to the composition of a $C^{1}$-section with a holomorphic tubular neighborhood $I$, there exists a plaque $\mathcal{L}_{x}$ of $x\in L$ and an open neighborhood $U$ of $i^{\prime}(x)$ such that $U$ is biholomorphic to $\mathcal{L}_{x}\times V$, where $V$ is an open supset of $\mathbb{C}^{p}$ and in such a trivialization: $i^{\prime}_{\|\mathcal{L}_{x}}:\;y\in\mathcal{L}_{x}\mapsto(y,\sigma(y)).$ Since $i^{\prime}(U)$ has its tangent space $J$-invariant, for any $u\in T_{x}U$ there exists $v\in T_{x}U$ such that: $J\circ(u,T_{x}\sigma(u))=(v,T_{x}\sigma(v)).$ This implies that $T_{x}\sigma$ commutes with $J$ and so that $T_{x}i^{\prime}$ commutes with $J$. ∎ Each complex lamination of Examples 2.7, 3.7, 3.8 or 4.2 is endowed with a holomorphic tubular neighborhood. Also each of them is preserved by a holomorphic dynamics and satisfies the hypotheses of Theorems 3.1 or 3.4. Thus we can apply to each them the three above propositions. Let us proceed in detail for Example 3.8 about fibered hyperbolic horseshoes. In this example, we saw that the restriction of the lamination $(L,\mathcal{L})$ to some subset $L^{\prime}:=L\cap(\mathbb{A}^{\prime}\times\mathbb{C}^{2})$ is $C^{1}$-persistent for any holomorphic perturbation $f^{\prime}$ of the dynamics. In other words, there exists a $C^{1}$-embedding $i(f^{\prime})$ of $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$ that $f^{\prime}$ lets invariant ($f^{\prime}\big{(}i(f^{\prime})(L^{\prime})\big{)}=i(f^{\prime})(L^{\prime})\big{)}$. Thus $\Lambda^{}$ is equal to $L^{\prime}$, and the embedding $i(f^{\prime})$ is holomorphic for a possibly non canonical complex structure on the lamination $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$, by Propositions 5.1 and 5.2. But, since $i(f^{\prime})$ is the composition of $I$ with a $C^{1}$-section of a holomorphic tubular neighborhood $(F,\mathcal{F},I,\pi)$ of $(L,\mathcal{L})$ (induced by the bundle $\mathbb{A}\times\mathbb{C}^{2}\rightarrow\mathbb{A}$), this complex structure is equal to the initial one. ###### Question 5.4. Find a non trivial holomorphic perturbation of the suspended Hénon map of Example 2.7. Let us now give an example of complex submanifold, normally expanded by a biholomorphic map, which is differentially persistent, but for which the complex structure is not persistent. ###### Example 5.5 (Hopf surface). Let $\tilde{M}:=\mathbb{C}\times\big{(}\mathbb{C}^{2}\setminus\\{(0,0)\\}\big{)}$ and $\alpha\in\mathbb{C}$ with modulus in $]0,1[$. Let $g$ be the following biholomorphic map of $\tilde{M}$: $g\;:\;(t,z_{1},z_{2})\mapsto(t,\alpha\cdot z_{1}+t\cdot z_{2},\alpha\cdot z_{2}).$ We remark that the action on $\tilde{M}$ of the group $\mathcal{G}:=\\{g^{n},\;n\in\mathbb{Z}\\}$ is proper, discontinuous and fixed-point-free on $\tilde{M}$. Thus the quotient $M:=\tilde{M}/\mathcal{G}$ is a complex manifold. For $s\in\mathbb{C}$, the following holomorphic endomorphism of $\tilde{M}$: $\tilde{f}_{s}\;:\;(t,z_{1},z_{2})\mapsto(2(t-s),z_{1},z_{2})$ is $\mathcal{G}$-equivariant. Thus there exists $f_{s}\in End^{\mathcal{H}}(M)$ such that the following diagram commutes: $\begin{array}[]{rcccl}&\tilde{M}&\stackrel{{\scriptstyle\tilde{f}_{s}}}{{\rightarrow}}&\tilde{M}&\\\ \pi&\downarrow&&\downarrow&\pi\\\ &M&\stackrel{{\scriptstyle f_{s}}}{{\rightarrow}}&M&\end{array},$ with $\pi\;:\;\tilde{M}\rightarrow M$ the canonical projection. We notice that the (Hopf) surface $M_{s}:=\pi\big{(}\\{s\\}\times(\mathbb{C}^{2}\setminus\\{(0,0)\\})\big{)}$ is normally expanded by $f_{s}$ since $\tilde{f}_{s}$ normally expands $\\{s\\}\times(\mathbb{C}^{2}\setminus\\{(0,0)\\})$. Actually, for every $r\geq 1$ and every small $s$, $M_{s}$ is the unique surface $C^{r}$-close to $M_{0}$ that $f_{s}$ preserves. Nevertheless, for $s\not=0$, the Hopf surface $M_{s}$ is not biholomorphic to $M_{0}$ (see [MK06], ex. 3, p.23). Thus, $M_{0}$ is a compact complex submanifold of $M$, which is $1$-normally expanded by $f_{0}$ and hence $C^{1}$-persistent but not holomorphically persistent. Nevertheless, by Theorem 0.3 the torus $M_{0}$ is persistent by deformation. Let us now give an example of a normally expanded submanifold by an endomorphism, which is persistent by deformation, but not differentially neither biholomorphically persistent. ###### Example 5.6. In the above Example, the Hopf surface $M_{0}$ preserved and $0$-normally expanded by the endomorphism of induced by: $\tilde{f}_{s}\;:\;(t,z_{1},z_{2})\mapsto(4(t-s),10z_{1},10z_{2})$ the submanifold $M_{0}$ is persistent by deformation, but not differentially nor biholomorphically persitent. Also Theorems 0.3 and 0.6 enable us to construct new laminations as in the following example. ###### Example 5.7. Let $P$ be a polynomial function of $\mathbb{C}$ which expands some repulsive compact subset $K$ of $\mathbb{C}$. For instance $P$ can be Collet-Eckmann or a hyperbolic quadratic polynomial function. Let $f\;:\;(z,(z_{i})_{i})\in\mathbb{C}\times\mathbb{C}^{n}\mapsto(P(z),0)$, for $n\in\mathbb{N}$. The $n$-dimensional embedded lamination $(L,\mathcal{L})=K\times\mathbb{C}$ is preserved and $1$-normally expanded. As $K$ does not contain more than one point, $(L,\mathcal{L})$ is endowed with a tubular neighborhood. Thus, by Theorem 3.1 and Propositions 5.1, 5.2 and 5.3, the $n$-dimensional lamination $(L^{\prime},\mathcal{L}^{\prime})=K\times B(0,R)$ is holomorphically persistent, for $R>0$ and with $B(0,R)$ the ball of $\mathbb{C}^{n}$ centered at $0$ of radius $R$. This means that for every holomorphic map $f^{\prime}$ close to $f$, there exists a holomorphic immersion $i(f^{\prime})$ of $(L^{\prime},\mathcal{L}^{\prime})$ into $\mathbb{C}^{n+1}$, close to $i_{\|L^{\prime}}$ and such that for every $k\in K$, $f^{\prime}$ sends $i(f^{\prime})(\\{k\\}\times\mathbb{C}^{n})$ into $i(f^{\prime})(\\{P(k)\\}\times B(0,R))$. On the other hand, by Theorem 0.3, the lamination $(L^{\prime},\mathcal{L}^{\prime})$ is persistent by deformation. This means that for a complex analytic family of holomorphic endomorphisms $(f_{t})_{t\in B}$ of $\mathbb{C}^{n+1}$ such that $f_{t_{0}}=f$ for some $t_{0}\in B$, there exists a neighborhood $B_{0}$ of $t_{0}$ such that $B_{0}$ is included in $V_{f}$ and $i\;:\;(t,x)\in B_{0}\times L^{\prime}\mapsto i(f_{t})(x)\in\mathbb{C}^{n+1}$ is a holomorphic immersion for the a complex laminar structure $\mathcal{D}$ on $B_{0}\times L^{\prime}$. By uniqueness of the structure $\mathcal{D}$ is $B_{0}\times\mathcal{L}^{\prime}$. With $n=0$ and with $B$ the open set of parameters $c\in\mathbb{C}$ such that $P_{c}(z)=z^{2}+c$ is hyperbolic, the above discussion implies the well known result that $\cup_{c\in B}\\{c\\}\times J_{c}$ is endowed with a complex one- dimensional structure of lamination, where the Julia set $J_{c}$ of $P_{c}$ is the transverse space. ## 6 Proof of differentiable persistences of normally contracted laminations In this section we prove Theorem 3.4. ### 6.1 The setting We fix $r\geq 1$. Let us denote by $K$ the closure of $L^{\prime}$. Let $(F,\mathcal{F},I,\pi)$ be a $C^{r}$-tubular neighborhood of the immersed lamination $(L,\mathcal{L})$ into $M$. We identify $L$ to the zero section of $F$. We endow $i^{}TM$ with a norm satisfying the normal contraction inequality. We endow $F$ with the norm associating to a vector $v\in F$ the one of the projection of $T_{0}I(v)$ into $E^{s}$ parallelly to $T\mathcal{L}$ in $i^{}TM$. Let $d$ be the dimension of the leaves of $\mathcal{L}$. We will always suppose $\eta>0$ small enough such that for every $x$ in a neighborhood of $K\cup f^{}(K)$: * - the ball $B_{F_{x}}(0,\eta)$ centered at $0$ and with radius $\eta$ is diffeomorphically sent by $I$ to a submanifold $F_{x}^{\eta}$, * - the union $\mathcal{L}_{x}^{\eta}$ (resp. $\mathcal{F}_{x}^{\eta}$) of plaques of $\mathcal{L}$ (resp. $\mathcal{F}$) containing $x$ and with diameter less than $\eta$ is a plaque. ### 6.2 The algorithm The idea of the proof is to use the following lemma which parametrizes the image by the dynamics of perturbations of the immersed lamination $\mathcal{L}$. ###### Lemma 6.1. For any small $\eta>0$, there exist an open neighborhood $V$ of $f^{}(K)$, a neighborhood $V_{i}^{0}$ of $i\in Mor^{r}(\mathcal{L},M)$, a neighborhood $V_{f}$ of $f\in End^{r}(M)$ and a continuous map $S^{0}\;:\;V_{f}\times V_{i}^{0}\rightarrow Im^{r}(\mathcal{L}_{\|V},M)$ satisfying: 1. 1. the morphism $S^{0}(f,i)$ is equal to $i_{\|V}$, 2. 2. for all $x\in V$, $i^{\prime}\in V_{i}^{0}$ and $f^{\prime}\in V_{f}$, the image of $i^{\prime}(\mathcal{L}_{f^{^{-1}}(x)}^{\eta})$ by $f^{\prime}$ intersects transversally $F_{x}^{\eta}$ at a unique point $S^{0}(f^{\prime},i^{\prime})(x)$. ###### Proof. Let us remind that $f^{}$ is an injective immersion from $\mathcal{L}_{\|D}$ into $\mathcal{L}$. Thus, we may suppose $\eta>0$ small enough such that $f^{\prime}$ sends diffeomorphically $i^{\prime}(\mathcal{L}_{f^{-1}(x)}^{\eta})$ to a submanifold transverse at a unique point $v$ to the submanifold $F_{x}^{\eta}$, for all $x$ close to $f^{}(K)$, $i^{\prime}$ $C^{r}$-close to $i$ and $f^{\prime}$ $C^{r}$-close to $f$. Figure 2: Definition of $S^{0}$. Let $S^{0}(f^{\prime},i^{\prime})(x)$ be the point $v$. Such a map $S^{0}$ satisfies conclusions 1 and 2 of Lemma 6.1. Let us show that $S^{0}$ takes continuously its values in the set of morphisms from the lamination $\mathcal{L}$ restricted to $V$ into $M$. A small open neighborhood $U$ of $x$ has the closure of its backward image by $f^{}$ included in a distinguish subset of $\mathcal{L}$. Let $\mathcal{L}_{y}$ be some of the plaques provided by a chart of this distinguish subset, containing ${y\in f^{^{-1}}(U)}$. We notice that the submanifold $f^{\prime}\circ i^{\prime}(\mathcal{L}_{y})$ depends $C^{r}$-continuously on ${y\in f^{^{-1}}(U)}$, $i^{\prime}$ $C^{r}$-close to $i$, and $f^{\prime}$ $C^{r}$-close to $f$. For $\eta>0$ small enough, the manifolds $(F_{x^{\prime\prime}}^{\eta})_{x^{\prime\prime}\in\mathcal{L}_{x^{\prime}}^{\eta}}$ are the leaves of a $C^{r}$-foliation, which depends $C^{r}$-continuously on $x^{\prime}\in V$: the foliation associated to a point $x^{\prime}_{1}$ close to $x^{\prime}$ is sent to the one of $x^{\prime}$ by a $C^{r}$-diffeomorphism of $M$ close to the identity. For $\eta>0$ sufficiently small and then $f^{\prime}$ and $i^{\prime}$ $C^{r}$-close enough to $f$ and $i$, each submanifold $F_{x^{\prime\prime}}^{\eta}$ intersects transversally at a unique point the submanifold $f^{\prime}\big{(}i^{\prime}(\mathcal{L}_{f^{-1}(x^{\prime})})\big{)}$, where $x^{\prime}$ belongs to $U$ and $x^{\prime\prime}$ to $\mathcal{L}_{x^{\prime}}^{\eta}$. As we know $S^{0}(f^{\prime},i^{\prime})(x^{\prime\prime})$ is this intersection point. In other words, $S^{0}(f^{\prime},i^{\prime})_{\|\mathcal{L}_{x^{\prime}}^{\eta}}$ is the composition of $i$ with the holonomy along the $C^{r}$-foliation $(F_{x^{\prime\prime}}^{\eta})_{x^{\prime\prime}\in\mathcal{L}_{x^{\prime}}^{\eta}}$, from $i(\mathcal{L}_{x^{\prime}}^{\eta})$ to the transverse section $f^{\prime}\circ i^{\prime}\big{(}\mathcal{L}_{f^{-1}(x^{\prime})}\big{)}$, Thus, the map $S^{0}(f^{\prime},i^{\prime})$ is of class $C^{r}$ along the $\mathcal{L}$-plaques contained in $U$. As these manifolds vary $C^{r}$-continuously with $x^{\prime}\in U$, the map $S^{0}(f^{\prime},i^{\prime})_{\|U}$ is a $\mathcal{L}_{\|U}$-morphism into $M$. Also by transversality, the map $S^{0}_{\|U}$ takes its values in the subset of immersions. As these foliations and manifolds also depend $C^{r}$-continuously on $f^{\prime}$ and $i^{\prime}$, there are neighborhood $V_{f}$ of $f\in End^{r}(M)$ and $V_{i}^{0}$ of $i_{\|L^{\prime}}\in Im^{r}(\mathcal{L}_{\|L^{\prime}},M)$ such that the map $S^{0}_{\|U}\;:\;(f^{\prime},i^{\prime})\in V_{f}\times V_{i}^{0}\mapsto S^{0}(f^{\prime},i^{\prime})_{\|U}$ is well defined and continuous into $Im^{r}(\mathcal{L}_{\|U},M)$. As $cl(L^{\prime})$ is compact, it has a finite open cover of by such open subsets $U$ on which the restriction of $S^{0}$ satisfies the above regularity property. By taking $V_{i}^{0}$ and $V_{f}$ small enough to be convenient for all the subsets of this finite cover, we get the continuity of the following map: $S^{0}\;:\;(f^{\prime},i^{\prime})\in V_{f}\times V_{i}^{0}\mapsto S^{0}(f^{\prime},i^{\prime})\in Im^{r}(\mathcal{L}_{\|V},M).$ where $V$ is the union of the open cover of $f^{}(K)$. ∎ By conclusion 2 of Lemma 6.1, for any $(f^{\prime},i^{\prime})\in V_{f}\times V_{i}^{0}$, we shall identify $S^{0}(f^{\prime},i^{\prime})$ with the $C^{r}$ section of the bundle: $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$ restricted to $V$. For a function $\rho\in Mor^{r}(\mathcal{L},[0,1])$, with compact support included in $V$ and equal to $1$ on a neighborhood $V^{\prime}$ of $L^{\prime}$, we define by using the above identification: $S\;:\;V_{f}\times V_{i}^{0}\rightarrow Im^{r}(\mathcal{L},M)$ $(f^{\prime},i^{\prime})\mapsto S_{f^{\prime}}(i^{\prime}):=\left\\{\begin{array}[]{cl}\rho(x)\cdot S^{0}(f^{\prime},i^{\prime})(x)&\mathrm{if}\;x\in V\\\ i(x)&\mathrm{else}\end{array}\right.$ For $f^{\prime}\in V_{f}$, we want to show that $S_{f^{\prime}}$ has a unique fixed point $i(f^{\prime})$. The space $\Gamma$ of $C^{r}$-sections of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$ with support in $V$ is a Banach space. Unfortunately $S$ is not a contraction for such a space. Nevertheless it is a contraction for the $C^{0}$-uniform norm $d_{C^{0}}$ of $\Gamma$ induced by the norm of $F$. ### 6.3 The $C^{0}$-contraction By definition of $S$, the $C^{0}$-contraction of $S^{0}$ implies the one of $S$. Let $x\in f^{^{-1}}(V)\subset D$. We shall identify both foliations $(F_{x^{\prime}}^{\eta})_{x^{\prime}\in\mathcal{L}_{x}^{\eta}}$ and $(F_{y^{\prime}}^{\eta})_{y^{\prime}\in\mathcal{L}_{f^{}(x)}^{\eta}}$ to $\mathbb{R}^{n-d}\times\mathbb{R}^{d}$. Let $p_{1}\;:\;\mathbb{R}^{n-d}\times\mathbb{R}^{d}\rightarrow\mathbb{R}^{n-d}$ and $p_{2}\;:\;\mathbb{R}^{n-d}\times\mathbb{R}^{d}\rightarrow\mathbb{R}^{d}$ be the canonical projections. For $f^{\prime}\in V_{f}$, $i^{\prime}\in V_{i}^{0}$ and $x\in f^{-1}(V)$, let $y:=p_{2}\circ f^{\prime}\circ i^{\prime}(x)$. The point $z=S_{f^{\prime}}^{0}(i^{\prime})(y)$ is a solution of the equation: $f^{\prime}\circ i^{\prime}(x)=z.$ Thus, the point $z=S_{f^{\prime}}^{0}(i^{\prime})(y)$ is also a solution of: $p_{1}\circ f^{\prime}\circ i^{\prime}(x)=p_{1}(z).$ As $S^{0}(f^{\prime},i^{\prime})(y)$ belongs to $F_{y}^{\eta}$, this point is identified to $(v,y)\in\mathbb{R}^{n-d}\times\mathbb{R}^{d}$. As $\mathcal{L}_{x}^{\eta}$ is precompact in $L$, the restriction $p_{1}\circ i^{\prime}_{\|\mathcal{L}_{x}^{\eta}}$ is identified to a vector of the Banach space of $C^{1}$-bounded maps $C^{1}_{b}(\mathcal{L}^{\eta}_{x},\mathbb{R}^{n-d})$. Let us apply the implicit function theorem with the following $C^{1}$-map from a neighborhood of $(0,0)$: $\Psi_{x,f^{\prime}}\;:\;C^{1}_{b}(\mathcal{L}_{x}^{\eta},\mathbb{R}^{n-d})\times\mathbb{R}^{n-d}\rightarrow\mathbb{R}^{n-d}$ $(j,v)\mapsto p_{1}\circ f^{\prime}\circ j(x)-v.$ We may assume that $(\Psi_{x,f^{\prime}})_{x\in f^{-1}(V),f^{\prime}\in V_{f}}$ is locally a continuous family of $C^{1}$-maps, since the charts of $\mathcal{L}$ provide continuous identifications of the $\eta$-plaques centered at points of $f^{-1}(V)$. Also, $\Psi_{x,f}(0,0)$ is zero for any $x\in f^{-1}(V)$ and $\partial_{v}\Psi_{x,f^{\prime}}=-id$ is invertible. Thus, by precompactness of $f^{-1}(V)$, the implicit function theorem implies that $i^{\prime}\in\Gamma\mapsto S^{0}_{f^{\prime}}(i^{\prime})(y)$ is of class $C^{1}$, for any $y\in V$ and $f^{\prime}$ close to $f$. Moreover, $\Psi_{x,f^{\prime}}$ depends $C^{1}$-continuously on $x$ and $f^{\prime}$. The differential $\partial_{i}(S^{0}(f,i)(y))(j)=-\partial_{v}\Psi(x,f)^{-1}\circ T\Psi(x,f)(j)=p_{1}\circ Tf\circ j(x)$ is contracting by normal contraction. Thus, $\partial_{i}(S^{0}(f^{\prime},i^{\prime})(y))$ is also $\lambda$-contracting for $f^{\prime}$ close to $f$ and $i^{\prime}\in\Gamma$ close to $i$. Thus, by the mean values theorem, $S^{0}(f^{\prime},\cdot)$ and so $S_{f^{\prime}}$ are $\lambda$-contracting on a fixed neighborhood of $i$, for every $f^{\prime}\in V_{f}$. Also $S_{f}$ sends the intersection of $V_{i}^{0}$ with the $\epsilon$-$C^{0}$-ball centered at $i$ into a $\lambda\cdot\epsilon$-ball, for any small $\epsilon>0$. By continuity of $f^{\prime}\mapsto S_{f^{\prime}}$, by restricting $V_{f}$, for every $f^{\prime}\in V_{f}$, $S_{f^{\prime}}$ sends $i$ into $B_{C^{0}}(i,(1-\lambda)\epsilon)$ and so $cl(B_{C^{0}}(i,\epsilon))\cap V_{i}^{0}$ into $B_{C^{0}}(i,\epsilon)\cap\Gamma$. ### 6.4 The $1$-jet space Contrarily to what happened for the $C^{0}$-topology, the map $S_{f}$ is not necessarily contracting in the $C^{1}$-topology, even when $L$ is compact (and so $S=S^{0}$). However, we are going to show that the forward action of $Tf$ on the Grassmannian of $d$-planes of $T\mathcal{F}$ is contracting, at the neighborhood of the distribution induced by $T\mathcal{L}$. Let us describe how $Tf$ acts on $(F,\mathcal{F})$. By compactness of $cl(V)$, there exists $\epsilon>0$ such that the restriction of $I$ to any $\epsilon$-plaque $\mathcal{F}_{x}^{\epsilon}$ of $\mathcal{F}$, centered at $x\in cl(V)$, is a diffeomorphism onto its image which is open in $M$. We denote by $I^{-1}_{x}$ its inverse. Thus, by restricting $V_{f}$, on a neighborhood $U^{}$ of the zero section of $F_{\|f^{-1}(V)}$, we can define, for every $f^{\prime}\in V_{f}$: $\hat{f}^{\prime}\;:\;z\in U^{}\mapsto I^{-1}_{f^{}\circ\pi(z)}\circ f^{\prime}\circ I(z),$ which is a $C^{r}$-morphism from $\mathcal{F}_{\|U^{}}$ into $\mathcal{F}$. Let us define a normed vector bundle structure on a neighborhood of $T\mathcal{L}$ in the Grassmannian of $d$-planes of $T\mathcal{F}$. Let $\chi_{0}$ be a continuous, horizontal distribution of $d$-planes of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$. By horizontal we mean that for every $y\in F$, the $d$-plane $\chi_{0}(y)$ of $T_{y}\mathcal{F}$ is sent onto $T_{\|\pi(y)}\mathcal{L}$ by $T_{y}\pi$. Let $\Psi\;:\;(y,t)\in F\times\mathbb{R}\mapsto t\cdot y\in F$. The map $\Psi$ is a $C^{r}$-morphism of $(F,\mathcal{F})\times\mathbb{R}$ onto $(F,\mathcal{F})$. Let $\chi$ be the continuous, horizontal distribution defined by: $\chi(y)=\left\\{\begin{array}[]{cl}\partial_{y}\Psi_{(y,\\|y\\|)}\Big{(}\chi\big{(}\frac{y}{\\|y\\|}\big{)}\Big{)}&\mathrm{if}\;y\not=0,\\\ T_{y}\mathcal{L}&\mathrm{if}\;y=0.\end{array}\right.$ For all $x\in L$ and $y\in F_{x}$, the vector space $T_{x}\mathcal{L}^{}\otimes F_{x}$ of linear maps from $T_{x}\mathcal{L}$ into $F_{x}$ is isomorphic to the space $\chi(y)^{}\otimes F_{x}$, by the isomorphism which associates to $l\in T_{x}\mathcal{L}^{}\otimes F_{x}$ the map $l\circ T\pi_{\|\chi(y)}$. As $F_{x}$ is canonically isomorphic to the subspace of $T_{y}\mathcal{F}$ tangent to the fiber $F_{x}$, via the graph map we shall identify $\chi(y)^{}\otimes F_{x}$ (and so $T_{x}\mathcal{L}^{}\otimes F_{x}$) to a neighborhood of $\chi(y)$ in the Grassmannian of $d$-planes of $T_{y}\mathcal{F}$. We denote by $P^{1}_{y}$ such a vector space that we endow with the norm subordinate to those of $T_{x}\mathcal{L}$ and $F_{x}$. We denote by $P^{1}$ the vector bundle over $F$ whose fiber at $y\in F$ is $P^{1}_{y}$. As $f^{}$ is an immersion and $f$ preserves the immersion of $(L,\mathcal{L})$, for $f^{\prime}$ close enough to $f$ and for all small $z\in(f^{}\circ\pi)^{-1}(V)$, for any $d$-plane $l\in P^{1}_{z}$, the image by $T\hat{f}^{\prime}$ is a small $d$-plane $l^{\prime}\in P^{1}_{\hat{f}^{\prime}(z)}$. Let us show the following lemma: ###### Lemma 6.2. For all small $\epsilon>0$ and then $V_{f}$ small enough, for all $f^{\prime}\in V_{f}$, $z\in(f^{}\circ\pi)^{-1}(V)$ and $l\in P^{1}_{z}$ both with norm not greater than $\epsilon>0$, the norm of $\phi_{f^{\prime}z}(l):=l^{\prime}\in P^{1}_{\hat{f}(z)}$ is less than $\epsilon$. Moreover, the map $\phi_{\hat{f}^{\prime}z}$ is $\lambda$-contracting. ###### Proof. Let $\pi_{v}$ be the projection of $T\mathcal{F}$ onto $F$ parallelly to $\chi$. Let $\pi_{h}$ be the projection of $T\mathcal{F}$ onto $\chi$ parallelly to $F$. Let $Tf^{\prime}_{h}:=\pi_{h}\circ T\hat{f}^{\prime}$ and $Tf^{\prime}_{v}:=\pi_{v}\circ T\hat{f}^{\prime}$. For any vector $e\in\chi(z)$, the point $(e,l(e))$ is sent by $T_{z}\hat{f}^{\prime}$ onto $\big{(}Tf^{\prime}_{h}(e,l(e)),Tf^{\prime}_{v}(e,l(e))\big{)}$. Let $e^{\prime}:=Tf^{\prime}_{h}(e,l(e))$. By definition of $l^{\prime}$, the point $(e,l(e))$ is sent by $T_{x}\hat{f}^{\prime}$ to $(e^{\prime},l^{\prime}(e^{\prime}))$. As $f^{}$ is an immersion, by restricting $V_{f}$ and $\epsilon$, the map $e\mapsto Tf_{h}^{\prime}(e,l(e))$ is invertible. Thus $e=(Tf^{\prime}_{h}(\cdot,l(\cdot)))^{-1}(e^{\prime})$. Therefore, the map $l^{\prime}$ is $e^{\prime}\mapsto Tf^{\prime}_{v}(\cdot,l(\cdot))\circ(Tf^{\prime}_{h}(\cdot,l(\cdot))^{-1}(e^{\prime})$. Hence, the expression of $l^{\prime}=\phi_{f^{\prime}z}(l)$ depends algebraically on $l$, and the coefficients of this algebraic expression depends continuously on $f^{\prime}$ or $z$, with respect to some trivializations. When $f^{\prime}$ is equal to $f$ and $z$ belongs moreover to the zero section, the map $\phi_{fz}\;:\;l^{\prime}\mapsto T_{z}f_{v}\circ l\circ\big{(}T_{z}f_{h}(\cdot,l(\cdot))\big{)}^{-1}$ is $\lambda$-contracting for $l$ small, by normal contraction, since $T_{z}f_{h}(\cdot,l(\cdot))$ is close to $T_{z}f^{}$. Thus, for $\epsilon$ small enough, for all $z\in(f^{}\circ\pi)^{-1}(V)$ and $l\in P^{1}_{z}$ both with norm less than $\epsilon$, the tangent map of $\phi_{fz}$ has a norm less than $\lambda$. Therefore, for $V_{f}$ small enough, the tangent map $T\phi_{f^{\prime}z}$ has a norm less than $\lambda$ and hence $\phi_{f^{\prime}z}$ is a $\lambda$-contraction on $cl\big{(}{B}_{P_{z}^{1}}(0,\epsilon)\big{)}$. As, for $z$ in the zero section, the map $\phi_{fz}$ vanishes at $0$, for $\epsilon$ and $V_{f}$ small enough, the norm $\phi_{f^{\prime}z}(0)$ is less than $(1-\lambda)\cdot\epsilon$. Consequently, by $\lambda$-contraction, the closed $\epsilon$-ball centered at the $0$-section is sent by $\phi_{f^{\prime}}$ into the $\epsilon$-ball centered at $0$, for all $z\in(f^{}\circ\pi)^{-1}(V)$ with norm less than $\epsilon$ and $f^{\prime}\in V_{f}$. ∎ ###### Remark 6.3. If $r\geq 1$, we notice that the action of $\hat{f}$ on $P^{1}$ is contracting and $(r-1)$-times more than $f^{}$. For any $C^{1}$-section $i^{\prime}$ of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$ and for every $x\in L$, the tangent space to the image of $i^{\prime}$ at $i^{\prime}(x)$ is an element $\nabla i^{\prime}(x)$ of $P^{1}_{i^{\prime}(x)}\approx T_{x}\mathcal{L}^{}\oplus F_{x}$. By conclusion 2 of Lemma 6.1, for $f^{\prime}\in V_{f}$, for $j\in V_{i}^{0}$, for $x_{0}\in(\pi\circ\hat{f}^{\prime}\circ j)^{-1}(V)$, the $d$-plane $\nabla j(x_{0})$ is sent by $T_{x}\hat{f^{\prime}}$ to $\nabla S^{0}(f^{\prime},j)(y_{0})$, with $x:=j(x_{0})$ and $y_{0}:=\pi\circ\hat{f}^{\prime}(x)$. Thus $\nabla S^{0}(f^{\prime},j)(y_{0})$ is equal to $\phi_{f^{\prime}x}(\nabla j(y_{0}))$. The interest of such a distribution $\chi$ is that any morphism $p\in Mor^{1}(\mathcal{L},\mathbb{R})$ and $i^{\prime}\in\Gamma$ satisfy: (3) $\nabla(p\cdot i^{\prime})(x)=Tp(x)\cdot i^{\prime}(x)+p(x)\cdot\nabla i^{\prime}(x),\quad\forall x\in L$ ###### Proof of Equation (3). Let $\pi_{v}$ be the projection of $T\mathcal{F}$ onto $F$ parallelly to $\chi$. The linear morphism $\pi_{v}$ commutes with the partial derivative $\partial_{1}\Psi$ with respect to the first variable since $\pi_{v}\circ\partial_{1}\Psi$ and $\partial_{1}\Psi\circ\pi_{v}$ have the same kernel $\chi$ and are equal on the complement $F$ in $T\mathcal{F}$. Thus we have: $\nabla(pi^{\prime})=\pi_{v}\circ T(\Psi(i^{\prime},p))=\pi_{v}\big{(}\partial_{1}\Psi_{(i^{\prime},p)}\circ Ti^{\prime}+\partial_{2}\Psi_{(i^{\prime},p)}\circ Tp\big{)}=\partial_{1}\Psi_{(i^{\prime},p)}\circ\pi_{v}\circ Ti^{\prime}+\pi_{v}\circ\partial_{2}\Psi_{(i^{\prime},p)}\circ Tp.$ Also we have: $\partial_{1}\Psi_{(y,t)}(u)=u\cdot t$ and $\partial_{2}\Psi_{(y,t)}(t^{\prime})=t^{\prime}y$, for every $y\in F_{\pi(y)}$. Consequently: $\nabla(pi^{\prime})=p\cdot\pi_{v}\circ Ti^{\prime}+T\rho\cdot i^{\prime}=p\cdot\nabla i^{\prime}+Tp\cdot i^{\prime}.$ ∎ Thus by Equation (3), we have for any $x_{0}\in(\pi\circ\hat{f}^{\prime}\circ j)^{-1}(V)$: $\nabla S_{f^{\prime}}(j)(y_{0})=\rho(y_{0})\cdot\phi_{f^{\prime}x}(\nabla j(x_{0}))+T\rho(y_{0})\cdot\hat{f}^{\prime}(x),\quad\mathrm{with}\;x:=j(x_{0})\;\mathrm{and}\;y_{0}:=\pi\circ\hat{f}^{\prime}(x).$ Since the support of $\rho$ is in $V$, for every $y_{0}\in L$: $\nabla S_{f}(j)(y_{0})=\phi^{\prime}_{fx}(\nabla j(y_{0})),$ with $\phi^{\prime}_{fx}\;:\;l\in P^{1}_{x}\mapsto\left\\{\begin{array}[]{cl}0&\mathrm{if}\;\pi(y_{0})\notin V\\\ \big{(}\rho(y_{0})\cdot\phi_{f^{\prime}x}(l)+T_{y_{0}}\rho\cdot x\big{)}\in P^{1}_{\hat{f}^{\prime}(x)}&\mathrm{if}\;\pi(y_{0})\in V.\end{array}\right.$. We notice that $\phi^{\prime}_{f^{\prime}x}$ is at least as much contracting as $\phi_{f^{\prime}x}$ and sends the $\epsilon$-ball of $P^{1}_{x}$ centered at 0 into the $\epsilon$-ball of $P^{1}_{\hat{f}^{\prime}(x)}$ centered at $0$, for $x$ small and $f^{\prime}$ close to $f$. Also $(\phi^{\prime}_{f^{\prime}x})_{x}$ is a continuous family of maps. ### 6.5 Proof of Theorem 3.4 when $r=1$ Let us begin by proving the existence of a closed neighborhood $V_{i}$ of $i\in\Gamma$ sent by $S_{f^{\prime}}$ into itself, for any $f^{\prime}\in End^{r}(M)$ close to $f$. For $\epsilon,\epsilon^{\prime}>0$ small enough, the following set is included in $V_{i}^{0}$: $V_{i}:=\Big{\\{}i^{\prime}\in\Gamma;\quad d_{C^{0}}(i,i^{\prime})\leq\epsilon^{\prime}\;\mathrm{and}\;\;\\|\nabla i^{\prime}(x)\\|_{P^{1}}\leq\epsilon,\;\forall x\in L\Big{\\}}$ where $\\|\cdot\\|$ refers to the $P^{1}$-norm. Therefore, by the two last steps, for $\epsilon$ then $\epsilon^{\prime}$ and finally $V_{f}$ small enough, $S_{f^{\prime}}$ sends $V_{i}$ into itself, for every $f^{\prime}\in V_{f}$. We showed that $S$ is contracting for the $C^{0}$-topology. Hence, for every $f^{\prime}$ in $v_{f}$, the sequence $(S_{f^{\prime}}^{n}(i))_{n\geq 0}$ converges to a Lipschitz section $i(f^{\prime})$. If this section is continuously differentiable, then the morphism $f^{\prime}:=\;x\in L^{\prime}\mapsto\pi\circ\hat{f}^{\prime}\circ i^{\prime}(x)$ is also of class $C^{1}$. Moreover Conclusions 1 and 2 of Theorem 3.4 are satisfied. Let us show that $i(f^{\prime})$ is of class $C^{1}$. We remind that for every $x\in V$ and every $i^{\prime}\in V_{i}$, there exists $y\in L$ such that: $\rho(x)\cdot\hat{f}^{\prime}\circ i^{\prime}(y)=S_{f}(i^{\prime})(x).$ Consequently, for every $x\in V$, there exists $y$ such that: $\rho(x)\circ\hat{f}\circ i(f^{\prime})(y)=i(f^{\prime})(x).$ Thus there are two possibilities for each point $x\in V$: 1. 1. there exists an infinite sequence $(x_{n})_{n\leq 0}\in V^{\mathbb{N}}$ such that: (4) $x_{0}=x\;\mathrm{and}\;\rho(x_{n+1})\cdot\hat{f}^{\prime}\circ i^{\prime}(x_{n})=S_{f}(i^{\prime})(x_{n+1})$ 2. 2. there is not such an infinite sequence. In the second case, let $(x_{n})_{n=0}^{-N}\in V^{N}$ be a maximal chain satisfying (4). Then we notice that $i(f^{\prime})(x)=S_{f^{\prime}}^{N+1}(i)(x).$ Since $\rho$ has compact support in $V$, the above equality holds on a neighborhood of $x$. This implies that $i(f^{\prime})$ is of class $C^{1}$ on a neighborhood of $x$. The first case is more delicate. Let $(x_{n})_{n\leq 0}$ be as in the definition. By Lemma 6.2, the intersection $\bigcap_{n\geq 0}\phi^{\prime n}_{f^{\prime}}(B_{P^{1}_{i(f^{\prime})(x_{n})}}(0,\epsilon))$ is decreasing and consists of a single point $l(x)$. Also for $x^{\prime}$ close to $x$, the length of the preorbit of $x^{\prime}$ in $V$ is large. Let $u$ be a tangent vector at $i(f^{\prime})(x)$ to $i(f^{\prime})(\mathcal{L}_{x}^{\eta})$. This means that there exists a sequence $(y_{n})_{n}\in\big{(}i(f^{\prime})(\mathcal{L}_{x}^{\eta})\setminus\\{y\\}\big{)}^{\mathbb{N}}$ such that: $y_{n}\stackrel{{\scriptstyle n\rightarrow\infty}}{{\longrightarrow}}y\quad\mathrm{and}\quad y_{n}=y+ud(y_{n},y)+o(d(y_{n},y)),\quad\mathrm{with}\;y:=i(f^{\prime})(x).$ Also, for every $k\leq 0$ and then $n$ sufficiently large, there exists $(y_{n}^{j})_{j=0}^{k}\in V^{-k}$ such that $y_{n}^{0}=y$ and: $\rho\circ\pi(y_{n}^{j+1})\circ\hat{f}^{\prime}(y_{n}^{j})=y_{n}^{j+1},\quad\forall j<0$ Moreover the sequence $(y_{n}^{j})_{n}$ converges to $y^{j}:=i(f^{\prime})(x_{j})$ and there exists $u_{j}$ such that: $y_{n}^{j}=y_{j}+u_{j}d(y_{n}^{j},y_{j})+o(d(y_{n}^{j},y_{j})).$ Consequently, $u_{j}$ is tangent to $i(f^{\prime})(\mathcal{L}_{x_{j}}^{\eta})$. Also $u_{j}$ belongs to a $d$-plane of $B_{P^{1}_{y_{j}}}(0,\epsilon)$. Therefore $u_{0}$ is a $\epsilon\cdot\lambda^{-j}$-close to $l$. When $j$ approaches infinity, we get that the tangent space of $i(f^{\prime})(x)$ is $l$. Thus we showed that at any point $x\in L$, the tangent space $\nabla i(f^{\prime})(x)$ of $i(f^{\prime})(\mathcal{L}_{x}^{\eta})$ at $x$ is a $d-$plane of $P^{1}_{i(f^{\prime})(x)}(0,\epsilon)$. To show that $i(f^{\prime})$ is of class $C^{1}$, it only remains to show that such a distribution of $d$-planes is continuous at every point $x$ with infinite preorbit $(x_{n})_{n\leq 0}$. Let $\delta>0$. Let $N\geq 0$ such that $\lambda^{N}\epsilon\leq\delta/3$. Let $x^{\prime}$ be sufficiently close to $x$ such that it has a preorbit $(x_{n}^{\prime})_{n=0}^{-N}\in V^{\mathbb{N}}$ well defined. By continuity of $\phi_{f^{\prime}}^{\prime}$, for $x^{\prime}$ sufficiently close to $x$, the chain $(x_{n}^{\prime})_{n=0}^{-N}$ is sufficiently close to $(x_{n})_{n=0}^{-N}$ such that $l_{1}:=\phi^{\prime}_{f^{\prime}y_{-1}}\circ\cdots\circ\phi^{\prime}_{f^{\prime}y_{-N}}(0)$ and $l^{\prime}_{1}:=\phi^{\prime}_{f^{\prime}y^{\prime}_{-1}}\circ\cdots\circ\phi^{\prime}_{f^{\prime}y^{\prime}_{-N}}(0)$ are $\delta/3$-close, with $(y^{\prime}_{n})_{n}:=i(f^{\prime})(x^{\prime}_{n})$. Also by contraction of $\phi^{\prime}_{f^{\prime}}$, $d(l_{1},\nabla i(f^{\prime})(x))\leq\frac{\delta}{3}\quad\mathrm{and}\quad d(l^{\prime}_{1},\nabla i(f^{\prime})(x^{\prime}))\leq\frac{\delta}{3}.$ Therefore the distance between $\nabla i(f^{\prime})(x)$ and $\nabla i(f^{\prime})(x^{\prime})$ is less than $\delta$. This finishes the proof of the regularity of $i(f^{\prime})$. Let us prove conclusion $(3^{\prime})$ of Theorem 3.4. Let $(y_{j})_{j\leq 0}\in M^{\mathbb{Z}^{-}}$ be a $f^{\prime}$-preorbit which is $\epsilon$-close to the image by $i$ of a $\epsilon$-pseudo-orbit $(x_{j})_{j\leq 0}\in L^{\prime\mathbb{Z}^{-}}$ of $f^{}$ respecting the plaques. Let $y_{j}^{\prime}:=(I_{x_{j}})^{-1}(y_{j})$ and $z_{j}:=\pi(y^{\prime}_{j})$. We want to show that $i(f^{\prime})(z_{j})$ is equal to $y_{j}$, for every $j\leq 0$. Let $n<0$ be such that $i(f^{\prime})(z_{n})$ and $y_{n}$ are at distance in $F_{z_{n}}$ at least: (5) $\frac{\lambda+1}{2}\sup_{j<0}d_{F_{z_{j}}}\big{(}i(f^{\prime})(z_{j}),y_{j}\big{)}.$ For $\epsilon>0$ small, the sequence $(z_{j})_{j<0}$ is close to $L^{\prime}$. Also we can construct a section $i^{\prime}\in V_{i}$ equal to $y_{j}$ at $z_{j}$, for each $j<0$, and such that the $C^{0}$-distance between $i^{\prime}$ and $i(f^{\prime})$ is $\sup_{j<0}d_{F_{z_{j}}}\big{(}i(f^{\prime})(z_{j}),y_{j}\big{)}$. We remind that on $V^{\prime}\subset V$ the maps $S$ and $S^{0}$ are equal, since $\rho$ is there equal to 1. By contraction of $S^{0}$, the $C^{0}$-distance between $i(f^{\prime})_{\|V^{\prime}}$ and $S^{0}(f^{\prime},i^{\prime})_{\|V^{\prime}}$ is less than $\lambda\cdot d(i(f^{\prime})(z_{n}),y_{n})$. By conclusion $(2)$ of Lemma 6.1, the immersion $S^{0}_{f^{\prime}}(i^{\prime})$ sends $(z_{j})_{j\leq 0}$ to $(y_{j})_{j\leq 0}$. By Equation (5): $\sup_{j<0}d_{F_{z_{j}}}\big{(}i(f^{\prime})(z_{j}),y_{j}\big{)}\geq d_{C^{0}}\big{(}i(f^{\prime}),i^{\prime}\big{)}\geq\frac{1}{\lambda}d_{C^{0}}\big{(}i(f^{\prime}),S_{f^{\prime}}(i^{\prime})\big{)}\geq\frac{\lambda+1}{2\lambda}\sup_{j<0}d_{F_{z_{j}}}\big{(}i(f^{\prime})(z_{j}),y_{j}\big{)}.$ Therefore for every $j<0$, the point $z_{j}$ is sent by $i(f^{\prime})$ to $z_{j}$. By commutativity of the diagram, the point $z_{0}$ is sent by $i(f^{\prime})$ to $z_{0}$. ### 6.6 Proof of Theorem 3.4 when $r>1$ Instead of regarding the immersion $i$ of $\mathcal{L}$ into $M$, we can consider the following immersion into the Grassmannian bundle $G_{r}(d,TM)$ of $d$-planes of $TM$: $i_{1}\;:\;x\in L\mapsto\big{(}i(x),Ti(T_{x}\mathcal{L})\big{)}.$ Since $f^{}$ is an immersion, $Tf$ acts canonically on the Grassmannian $G$ of $d$-planes of $TM$, on a neighborhood of $\\{Ti(T_{x}\mathcal{L}),\;x\in L\\}$. As $f$ is $1$-normally contracting, $Tf$ sends into itself a nice such a neighborhood. We define by induction the $l^{th}$-Grassmannian $G^{l}$: $G^{1}$ is $G$ and $G^{l+1}$ is the Grassmannian of $d$-planes of $TG^{l}$. By the same way, we define the immersion $i_{l}$ of $(L,\mathcal{L})$ into $G^{l}$. We notice that $Tf$ acts canonically on $G^{l}$. We denote by $f_{l}$ its action. Let us describe the constructions $(G^{j})_{j}$, $(i_{j})_{j}$ and $(f_{j})_{j}$. $\mathrm{Let}\;G^{0}:=M\;\mathrm{and}\;G^{j+1}:=\\{l\in C^{1}(\mathbb{R}^{d},G^{j}):\;T_{0}l\;\mathrm{is\;injective}\\}/\sim,$ with $l\sim l^{\prime}$ if there exists a diffeomorphism $\phi$ of $\mathbb{R}^{d}$ fixing the origin such that $l^{\prime}(x)=l\circ\phi(x)+o(x)$. We notice that we have a bundle $\pi_{j+1,j}:\;G^{j+1}\rightarrow G^{j}$ and so $\pi_{j}:=\pi_{j,j-1}\circ\dots\pi_{1,0}:\;G^{j}\rightarrow M$. For each $i^{\prime}\in V_{i}$, we define the following immersion of $(L,\mathcal{L})$ into $G^{j}$: $i_{0}^{\prime}:=i^{\prime}$ and $i^{\prime}_{j+1}:=x\in(L,\mathcal{L})\mapsto[i^{\prime}_{j\|\mathcal{L}_{x}^{\eta}}]$, by identifying $\mathcal{L}_{x}^{\eta}$ to $\mathbb{R}^{d}$ and $x$ to $0$. We notice that $i^{\prime}_{j}$ is well defined for $0<j\leq r$ since $\pi_{j-1}\circ i^{\prime}_{j-1}=i^{\prime}$ is an immersion and so $i^{\prime}$ is an immersion. Similarly the action $f^{\prime}_{j}$ of $f^{\prime}\in V_{f}$ on $G^{j}$ is defined inductively: $f^{\prime}_{0}$ is $f^{\prime}$ and $f^{\prime}_{j+1}([l])$ is equal to $[f^{\prime}_{j}\circ l]$. This action is well defined on a neighborhood of the image of $i_{j}$ since $\pi_{j}\circ f^{\prime}_{j}\circ i_{j}=f\circ i$ and so $f^{\prime}_{j}\circ i_{j}$ is well an immersion. By Remark (6.3), $f_{1}$ $r-1$-contracts the immersion $i_{1}$ of $(L,\mathcal{L})$ over $f^{}$. Thus, by induction, the action of $f^{\prime}_{l}$ on the $l^{th}$-Grassmannian $G^{l}$ of $TM$ is contracting at a neighborhood of $i_{l}(L)$, for every $l\in\\{1,\dots,r\\}$. As before we shall use the formalism of connexions since the lamination $(L,\mathcal{L})$ is not compact, and so we are going to study the consequences of the multiplication by $\rho$ of after the action of $T^{j}f^{\prime}$ on $G^{j}$. In a similar way to which it is done in section 6.4, for $i^{\prime}\in\Gamma$ and $l\in\\{1,\dots,r\\}$, let $\nabla^{l}i^{\prime}$ be the map $p_{v}\circ T^{l}i^{\prime}$, where $\pi_{v}$ is the projection of $T\mathcal{F}$ onto $F$ parallelly to $\chi$, we notice that $T^{l}i^{\prime}$ is the $l^{th}$-differential of $i^{\prime}$. Thus $\nabla^{l}i^{\prime}$ is a vector of the space $L_{sym}^{l}(T\mathcal{L},F)$ of $l$-linear symmetric morphisms from the bundle $(T\mathcal{L})^{l}$ to $F$ over $L$. Let us show the identification of $i^{\prime}_{j}$ with $\nabla^{j}i^{\prime}(x)$, by induction. For $j\leq 1$ this is already done. We remind that $i^{\prime}_{j+1}(x)=[i_{j\|\mathcal{L}_{x}^{\eta}}]$, for $x\in L$. By induction, we have $i^{\prime}_{j+1}(x)=[p_{v}\circ T^{j}i^{\prime}_{\|\mathcal{L}_{x}^{\eta}}]$ and so $i^{\prime}_{j+1}(x)=Tp_{v}\circ T_{x}T^{j}i^{\prime}$. As $p_{v}$ is the tangent map of $T_{i^{\prime}(x)}\mathcal{F}$, we have: $i^{\prime}_{j+1}(x)=p_{v}\circ T^{j+1}_{x}i^{\prime}.$ We notice that this identification is coherent with the topology of $G^{j}$ and $Im^{j}(\mathcal{L},M)$. By the above discussion, we can define an endomorphism $\phi^{l}_{f^{\prime}}$ from the $\epsilon$-neighborhood of the zero section of $L_{sym}^{l}(T\mathcal{L},F)$ into itself, contracting each fiber and such that: $\forall x\in V,\quad\nabla^{l}S^{0}(f^{\prime},i^{\prime})(x):=\phi^{l}_{f^{\prime}y}(\nabla^{l}i^{\prime}(y)),\quad\mathrm{with}\;\hat{f}^{\prime}(i^{\prime}(y))=x.$ An induction proves similarly as for Equation (3), the following Leibniz formula: $\nabla^{l}S_{f^{\prime}}(i^{\prime}):=\sum_{k=0}^{l}C_{l}^{k}(T^{l-k}\rho)\nabla^{k}S^{0}(f^{\prime},i^{\prime}).$ Thus $(\nabla^{l}S_{f^{\prime}}(i^{\prime}))_{l=1}^{r}$ is equal to a continuous, contracting function of $(\nabla^{k}i^{\prime})_{k=1}^{l}$ and we can proceed as in section 6.5 to conclude the proof. ## 7 Proof of differentiable persistences of normally expanded laminations In this section we prove Theorem 3.1 in the normally expanded case. ### 7.1 The setting We fix $r\geq 1$. Let us denote by $K$ the closure of $L^{\prime}$. Let $(F,\mathcal{F},I,\pi)$ be a $C^{r}$-tubular neighborhood of the immersed lamination $(L,\mathcal{L})$ in $M$. We identify $L$ to the zero section of $F$. We endow $i^{}TM$ with a norm satisfying the normal expansion inequality. We endow $F$ with the norm associating to a vector $v\in F$ the one of the projection of $T_{0}I(v)$ into $E^{u}$ parallelly to $T\mathcal{L}$ in $i^{}TM$. Let $d$ be the dimension of the leaves of $\mathcal{L}$. We will always suppose $\eta>0$ small enough such that for every $x$ in a neighborhood of $K\cup f^{}(K)$: - the ball $B_{F_{x}}(0,\eta)$ centered at $0$ and with radius $\eta$ is sent by $I$ to a submanifold $F_{x}^{\eta}$, * - the union $\mathcal{L}_{x}^{\eta}$ (resp. $\mathcal{F}_{x}^{\eta}$) of plaques of $\mathcal{L}$ (resp. $\mathcal{F}$) containing $x$ and with diameter less than $\eta$, is a precompact plaque. ### 7.2 The algorithm The idea of the proof is to use the following lemma which constructs the preimages of perturbations of the immersed lamination $\mathcal{L}$. ###### Lemma 7.1. For any small $\eta>0$, there exists a neighborhood $V$ of $K$ in $L$, a neighborhood $V_{i}^{0}$ of $i\in Mor^{r}(\mathcal{L},M)$, a neighborhood $V_{f}$ of $f\in End^{r}(M)$ and a continuous map $S^{0}\;:\;V_{f}\times V_{i}^{0}\rightarrow Im^{r}(\mathcal{L}_{\|V},M)$ satisfying: 1. 1. the morphism $S^{0}(f,i)$ is equal to $i_{\|V}$, 2. 2. for all $x\in L^{\prime}$, $i^{\prime}\in V_{i}^{0}$ and $f^{\prime}\in V_{f}$, the preimage of $i^{\prime}(\mathcal{L}_{f^{}(x)}^{\eta})$ by $f^{\prime}$ intersects $F_{x}^{\eta}$ at a unique point $S^{0}(f^{\prime},i^{\prime})(x)$, for all $f^{\prime}\in V_{f}$ and $i^{\prime}\in V_{i}^{0}$. ###### Proof. By normal expansion, we may suppose $\eta>0$ small enough and then $(i^{\prime},f)$ close to $(i,f)$ such that, by normal expansion, the restriction of $f^{\prime}$ to $F_{x}^{\eta}$ is a diffeomorphism onto its image, and this image intersects transversally at a unique point the image of the plaque $\mathcal{L}_{f^{}(x)}^{\eta}$ by $i^{\prime}$, for all $x$ in a neighborhood $V$ of $K$, for all $i^{\prime}$ $C^{r}$-close to $i$ and for all $f^{\prime}$ $C^{r}$-close to $f$. Writing this intersection point in the form $f^{\prime}(v)=i^{\prime}(x^{\prime})$, we define $S^{0}(f^{\prime},i^{\prime})(x):=v$. Figure 3: Definition of $S$. Such a map $S^{0}$ satisfies conclusions 1 and 2 of Lemma 7.1. Let us show that $S^{0}$ takes continuously its values in the set of morphisms from the lamination $\mathcal{L}_{\|V}$ into $M$. Let $x\in K$ and let $(U_{y},\phi_{y})\in\mathcal{L}$ be a chart of a neighborhood of $y:=f^{}(x)$. We may suppose that $\phi_{y}$ can be written in the form: $\phi_{y}\;:\;U_{y}\rightarrow\mathbb{R}^{d}\times T_{y},$ where $T_{y}$ is a locally compact metric space. Let $(u_{y},t_{y})$ be defined by $\phi_{y}(y)=:(u_{y},t_{y})$. We remark that $\mathbb{R}^{d}$ is $C^{r}$-immersed by $\psi:=u\in\mathbb{R}^{d}\mapsto i\circ\phi^{-1}(u,t_{y})$. By normal expansion, the restriction of $f$ to a small neighborhood of $i(x)$ is transverse to the above immersed manifold, at $z:=f\circ i(x)$. In other words, $Tf(T_{i(x)}M)+T\Psi(T_{u_{y}}\mathbb{R}^{d})=T_{z}M.$ Thus, by transversality, there exist nice open neighborhoods $V_{u_{y}}$ of $u_{y}\in\mathbb{R}^{d}$ and $V_{i(x)}$ of $i(x)\in M$ such that the preimage by $f_{\|V_{i(x)}}$ of $\psi(V_{u_{y}})$ is a $C^{r}$-submanifold. Moreover, such a submanifold depends continuously on $f$ and $\psi$, with respect to the $C^{r}$-topologies. More precisely, there exist neighborhoods $V_{u_{y}}$ of $u_{y}$, $V_{f}$ of $f\in End^{r}(M)$, $V_{\psi}$ of $\psi\in C^{r}(\mathbb{R}^{d},M)$, and $V_{i(x)}$ of $i(x)$ such that for all $f^{\prime}\in V_{f}$ and $\psi^{\prime}\in V_{\psi}$, $f^{\prime}_{\|V_{i(x)}}$ is transverse to $\Psi^{\prime}_{\|V_{u_{y}}}$ and also the preimage by $f^{\prime}_{\|V_{i(x)}}$ of $\psi^{\prime}(V_{u_{y}})$ is a manifold which depends continuously on $f^{\prime}$ and $\psi^{\prime}$, in the compact-open $C^{r}$-topologies. There exist neighborhoods $V_{t_{y}}$ of $t_{y}$ in $T_{y}$ and $V_{i}^{0}$ of $i\in Im^{r}(\mathcal{L},M)$, such that $\psi_{i^{\prime},t}\;:\;u\in\mathbb{R}^{d}\mapsto i^{\prime}\circ\phi(u,t)$ belongs to $V_{\psi}$, for all $t\in V_{t_{y}}$ and $i^{\prime}\in V_{i}^{0}$. Thus, the preimage by every $f^{\prime}\in V_{f}$, restricted to $V_{i(x)}$, of the plaque $\mathcal{L}_{t}:=\phi^{-1}_{y}(V_{u_{y}}\times\\{t\\})$, immersed by $i^{\prime}\in V_{i}^{0}$, depends $C^{r}$-continuously on $f^{\prime}$, $i^{\prime}$ and $t$. For $\eta>0$ small enough, the manifolds $(F_{x^{\prime\prime}}^{\eta})_{x^{\prime\prime}\in\mathcal{L}_{{t}}^{\eta}}$ form the leaves of a $C^{r}$-foliation on a open subset of $M$, which depends continuously on $t\in V_{t_{y}}$: the foliation associated to $t^{\prime}$ close to $t$ is sent to the one of $t$ by a $C^{r}$-diffeomorphism of $M$ close to the identity. We may suppose $U_{x}$ and $\eta$ small enough in order that the closure of $\cup_{x^{\prime}\in U_{x}}\mathcal{L}_{x^{\prime}}^{\eta}$ can be sent by $f^{}$ into $\phi_{y}^{-1}(V_{u_{y}}\times V_{t_{y}})$. For all $\eta>0$ and then $V_{i}^{0}$ and $V_{f}$ small enough, each submanifold $F_{x^{\prime\prime}}^{\eta}$ intersects transversally at a unique point the submanifold $f^{\prime-1}_{\|V_{i(x)}}\big{(}i^{\prime}(\mathcal{L}_{t^{\prime}})\big{)}$, where $t^{\prime}$ is the second coordinate of $\phi_{y}\circ f^{}(x^{\prime\prime})$ and $x^{\prime\prime}$ belongs to $\cup_{x^{\prime}\in U_{x}}\mathcal{L}_{x^{\prime}}^{\eta}$. As we know $S^{0}(f^{\prime},i^{\prime})(x^{\prime\prime})$ is this intersection point. In other words, $S^{0}(f^{\prime},i^{\prime})_{\|\mathcal{L}_{x^{\prime}}^{\eta}}$ is the composition of $i$ with the holonomy along the $C^{r}$-foliation $(F_{x^{\prime\prime}}^{\eta})_{x^{\prime\prime}\in\mathcal{L}_{x^{\prime}}^{\eta}}$, from $i(\mathcal{L}_{x^{\prime}}^{\eta})$ to the transverse section $f^{\prime-1}_{\|V_{i(x)}}(i^{\prime}(\mathcal{L}_{t^{\prime}}))$, where $t^{\prime}$ is the second coordinate of $\phi_{y}\circ f^{}(x^{\prime})$, for every $x^{\prime}\in U_{x}$. Thus, the map $S^{0}(f^{\prime},i^{\prime})$ is of class $C^{r}$ along the $\mathcal{L}$-plaques contained in $U_{x}$. As these manifolds vary $C^{r}$-continuously with $x^{\prime}\in U_{x}$, the map $S^{0}(f^{\prime},i^{\prime})_{\|U_{x}}$ is a $\mathcal{L}_{\|U_{x}}$-morphism into $M$. As these foliations and manifolds depend $C^{r}$-continuously on $x^{\prime}\in U_{x}$, $i^{\prime}\in V_{i}^{0}$ and $f^{\prime}\in V_{f}$, the map $(f^{\prime},i^{\prime})\in V_{f}\times V_{i}^{0}\mapsto S^{0}(f^{\prime},i^{\prime})_{\|U_{x}}$ is continuous into $Mor^{r}(\mathcal{L}_{\|U_{x}},M)$. As, $K$ is compact, we get a finite open cover of $K$ by such open subsets $U_{x}$ on which the restriction of $S^{0}$ satisfies the above regularity property. By taking $V_{i}^{0}$ and $V_{f}$ small enough to be convenient for all the subsets of this finite cover, we get the continuity of the following continuous map: $S^{0}\;:\;(f^{\prime},i^{\prime})\in V_{f}\times V_{i}^{0}\mapsto S^{0}(f^{\prime},i^{\prime})\in Mor^{r}(\mathcal{L}_{\|V},M).$ where $V$ is the union of the finite open cover of $K$. As $S^{0}$ takes its values in the set of the composition of the immersion $I$ with the section of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$, $S^{0}$ takes its values in the set of immersions from $\mathcal{L}_{\|V}$ into $M$. ∎ By conclusion 2 of Lemma 6.1, for any $f^{\prime}\in V_{f}$ and $i^{\prime}\in V_{i}^{0}$, we shall identify $S^{0}(f^{\prime},i^{\prime})$ with the $C^{r}$ section of the bundle: $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$ restricted to $V$. For a function $\rho\in Mor(\mathcal{L},[0,1])$, with compact support included in $V$ and equal to $1$ on a neighborhood $V^{\prime}$ of $L^{\prime}$, we define by using the above identification: $S\;:\;V_{f}\times V_{i}^{0}\rightarrow Im^{r}(\mathcal{L},M)$ $(f^{\prime},i^{\prime})\mapsto S_{f^{\prime}}(i^{\prime}):=\left\\{\begin{array}[]{cl}\rho(x)\cdot S^{0}(f^{\prime},i^{\prime})(x)&\mathrm{if}\;x\in V\\\ i(x)&\mathrm{else}\end{array}\right.$ For $f^{\prime}\in V_{f}$, we want to show that $S_{f^{\prime}}$ has a unique fixed point $i(f^{\prime})$. The space $\Gamma$ of $C^{r}$ sections of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$ whose restriction to the complement of $V$ is equal to $0$ (that is $i$ in the identification) is a Banach space. Unfortunately $S$ is not a contraction for such a space. But it is a contraction when $\Gamma$ is endowed with the $C^{0}$-uniform norm $d_{C^{0}}$ induced by the norm of $F$. ### 7.3 The $C^{0}$-contraction By definition of $S$, the $C^{0}$-contraction of $S$ follows from the $C^{0}$-contraction of $S^{0}$. For $x\in V$, we shall identify the foliations $(F_{x^{\prime}}^{\eta})_{x^{\prime}\in\mathcal{L}_{x}^{\eta}}$ and $(F_{y}^{\eta})_{y\in\mathcal{L}_{f^{}(x)}^{\eta}}$ to $\mathbb{R}^{n-d}\times\mathbb{R}^{d}$. For $f^{\prime}\in V_{f}$ and $i^{\prime}\in V_{i}^{0}$, the point $y=S^{0}(f^{\prime},i^{\prime})(x)$ is a solution of the equation: $f^{\prime}(y)=i^{\prime}\circ p_{2}\circ f^{\prime}(y),$ where $p_{2}\;:\;\mathbb{R}^{n-d}\times\mathbb{R}^{d}\rightarrow\mathbb{R}^{d}$ is the canonical projection. Thus, the point $y=S(f^{\prime},i^{\prime})(x)$ is also a solution of: $p_{1}\circ f^{\prime}(y)=p_{1}\circ i^{\prime}\circ p_{2}\circ f^{\prime}(y),$ with $p_{1}\;:\;\mathbb{R}^{n-d}\times\mathbb{R}^{d}\rightarrow\mathbb{R}^{n-d}$ is the canonical projection. As $S^{0}(f^{\prime},i^{\prime})(x)$ belongs to $F_{x}^{\eta}$, in these identifications, it is equal to some $(v,x)\in\mathbb{R}^{n-d}\times\mathbb{R}^{d}$. As $\mathcal{L}_{f^{}(x)}^{\eta}$ is precompact, in these identifications, $p_{1}\circ i^{\prime}$ belongs to the Banach space of $C^{1}$-bounded maps $C^{1}_{b}(\mathcal{L}^{\eta}_{f^{}(x)},\mathbb{R}^{n-d})$. Let us apply the implicit function theorem with the following $C^{1}$-map from a neighborhood of $0$: $\Psi_{x,f^{\prime}}\;:\;C^{1}_{b}(\mathcal{L}_{f^{}(x)}^{\eta},\mathbb{R}^{n-d})\times\mathbb{R}^{n-d}\rightarrow\mathbb{R}^{n-d}$ $(j,v)\mapsto p_{1}\circ f^{\prime}(y)-p_{1}\circ j\circ p_{2}\circ f^{\prime}(y),$ where $y=(x,v)$. We may assume that $(\Psi_{x,f^{\prime}})_{x\in K,f^{\prime}\in V_{f}}$ is locally a continuous family of $C^{1}$-maps, since the charts of $\mathcal{L}$ and the trivializations of $F$ locally provide identifications of $(F_{x^{\prime}}^{\eta})_{x^{\prime}\in\mathcal{L}_{x}^{\eta}}$ and $(F_{y}^{\eta})_{y\in\mathcal{L}_{f^{}(x)}^{\eta}}$ which vary continuously with $x\in V$. Also, $\Psi_{x,f}(0,0)$ is zero for any $x\in V$, and $T\Psi_{x,f^{\prime}}(0,0)(\partial v)=p_{1}\circ Tf^{\prime}(\partial v)$ is invertible. Thus, by precompactness of $V$, the implicit function theorem implies that $i^{\prime}\mapsto S^{0}(f^{\prime},i^{\prime})(x)$ is of class $C^{1}$, for any $x\in V$ and $f^{\prime}$ close to $f$. Moreover, its differential depends continuously on $x$, $i^{\prime}$ and $f^{\prime}$. The differential: $TS^{0}({f},i)(x)(j)=\partial_{v}\Psi_{(x,f)}(0,0)^{-1}\circ T\Psi_{(x,f)}(0,0)(j)=(p_{2}\circ Tf_{\|TI(F_{x})})^{-1}\circ j\big{(}f^{}(x)\big{)}$ is contracting by normal expansion. By continuity, $TS_{f^{\prime}}(j)(x)(\partial i)$ is also $\lambda$-contracting for $f^{\prime}$ close to $f$ and $i^{\prime}$ close to $i$. Thus, by the mean values theorem, $S^{0}_{f}$ and so $S_{f^{\prime}}$ are $\lambda$-contracting on the intersection of $V_{i}^{0}$ with the $\epsilon$-ball centered at $i$, and send this ball into a $\lambda\cdot\epsilon$-ball, for any $\epsilon>0$. Since $S_{f}(i)=i$, by continuity of $f^{\prime}\mapsto S_{f^{\prime}}$ and by restricting $V_{f}$, for every $f^{\prime}\in V_{f}$, $S_{f}$ sends $cl(B_{C^{0}}(i,\epsilon))$ into $B_{C^{0}}(i,\epsilon)$. ### 7.4 The $1$-jet space Contrarily to what happened for the $C^{0}$-topology, the map $S_{f}$ is not necessarily contracting in the $C^{1}$-topology, even when $L$ is compact (and so $S=S^{0}$). However, we are going to show that the backward action of $Tf$ on the Grassmannian of $d$-planes of $T\mathcal{F}$ is contracting, at the neighborhood of the distribution induced by $T\mathcal{L}$. Actually, it is the $C^{r}$-endomorphism of a neighborhood of the zero section $U^{}$ of $(F,\mathcal{F})$, $\hat{f}^{\prime}\;:\;z\in U^{}\mapsto I^{-1}_{f^{}\circ\pi(z)}\circ f^{\prime}\circ I(z)$ defined as in section 6.4 which acts on the Grassmannian of $T\mathcal{F}$ by backward image. Let us define a norm and a vector structure on a neighborhood of $T\mathcal{L}$ in the Grassmannian of $d$-planes of $T\mathcal{F}$. Let $\chi_{0}$ be a continuous, horizontal distribution of $d$-planes of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$. By horizontal we mean that for every $y\in F$, the $d$-plane $\chi_{0}(y)$ is sent onto $T_{\|\pi(y)}\mathcal{L}$ by $T_{y}\pi$. Let $\Psi\;:\;(y,t)\in F\times\mathbb{R}\mapsto t\cdot y\in F$. The map $\Psi$ is a $C^{r}$-morphism of $(F,\mathcal{F})\times\mathbb{R}$ onto $(F,\mathcal{F})$. Let $\chi$ be the horizontal distribution defined by: $\chi(y)=\left\\{\begin{array}[]{cl}T\Psi_{(y,\\|y\\|)}\Big{(}\chi_{0}\big{(}\frac{y}{\\|y\\|}\big{)}\Big{)}&\mathrm{if}\;y\not=0\\\ T_{y}\mathcal{L}&\mathrm{if}\;y=0\end{array}\right.$ For all $x\in L$ and $y\in F_{x}$, the vector space $T_{x}\mathcal{L}^{}\otimes F_{x}$ of linear maps from $T_{x}\mathcal{L}$ to $F_{x}$ is isomorphic to the space $\chi(y)^{}\otimes F_{x}$, by the isomorphism which associates to $l\in T_{x}\mathcal{L}^{}\otimes F_{x}$ the map $l\circ T\pi_{\|\chi(y)}$. As $F_{x}$ is canonically isomorphic to the subspace of $T_{y}\mathcal{F}$ tangent to $F_{x}$, via the graph map we shall identify $\chi(y)^{}\otimes F_{x}$ (and so $T_{x}\mathcal{L}^{}\otimes F_{x}$) to a neighborhood of $\chi(y)$ in the Grassmannian of $d$-planes of $T_{y}\mathcal{F}$. We denote by $P^{1}_{y}$ such a vector space that we endow with the norm subordinate to those of $T_{x}\mathcal{L}$ and $F_{x}$. We denote by $P^{1}$ the vector bundle over $F$ whose fiber at $y\in F$ is $P^{1}_{y}$. By normal expansion, for all $f^{\prime}$ $C^{1}$-close to $f$ and for all small $x\in\pi^{-1}(V)$ for any small $d$-plane $l\in P^{1}_{\hat{f}^{\prime}(x)}$, the preimage by $T_{x}\hat{f}^{\prime}$ of the $d$-plane $l$ is a small $d$-plane $l^{\prime}\in P^{1}_{x}$. Let us show the following lemma: ###### Lemma 7.2. For all small $\epsilon>0$, and then $V_{f}$ small enough, for all $f^{\prime}\in V_{f}$, $x\in\pi^{-1}(V)$ and $l\in P^{1}_{\hat{f}^{\prime}(x)}$ both with norm at most $\epsilon>0$, the norm of $\phi_{f^{\prime}x}(l):=l^{\prime}\in P^{1}_{x}$ is then less than $\epsilon$. Moreover, the map $\phi_{f^{\prime}x}$ is $\lambda$-contracting. ###### Proof. Let $\pi_{v}$ be the projection of $T\mathcal{F}$ onto $F$ parallelly to $\chi$. Let $\pi_{h}$ be the projection of $T\mathcal{F}$ onto $\chi$ parallelly to $F$. For any vector $e^{\prime}\in\chi(x)$, the point $(e^{\prime},l^{\prime}(e^{\prime}))$ is sent by $T_{x}\hat{f}^{\prime}$ to $\big{(}Tf^{\prime}_{h}(e^{\prime},l^{\prime}(e^{\prime})),Tf^{\prime}_{v}(e^{\prime},l^{\prime}(e^{\prime}))\big{)},\quad\mathrm{with}\;Tf^{\prime}_{h}:=\pi_{h}\circ T\hat{f}^{\prime}\;\mathrm{and}\;Tf^{\prime}_{v}:=\pi_{v}\circ T\hat{f}^{\prime}.$ Let $e:=Tf^{\prime}_{h}(e^{\prime},l^{\prime}(e^{\prime}))$. By definition of $l^{\prime}$, the point $(e^{\prime},l^{\prime}(e^{\prime}))$ is sent by $T_{x}f^{\prime}$ to $(e,l(e))$. Therefore, we have $l(e)=Tf^{\prime}_{v}(e^{\prime},l^{\prime}(e^{\prime}))$ and $l(e)=l\circ Tf^{\prime}_{h}(e^{\prime},l^{\prime}(e^{\prime}))$. $\Rightarrow Tf^{\prime}_{v}(e^{\prime},l^{\prime}(e^{\prime}))=l\circ Tf^{\prime}_{h}(e^{\prime},l^{\prime}(e^{\prime}))$ $\Rightarrow(Tf^{\prime}_{v}-l\circ Tf^{\prime}_{h})(l^{\prime}(e^{\prime}))=(l\circ Tf^{\prime}_{h}-Tf^{\prime}_{v})(e^{\prime})$ By normal expansion, for $\epsilon$ and $V_{f}$ small enough, the map $(Tf^{\prime}_{v}-l\circ Tf^{\prime}_{h})_{\|F}$ is bijective. Consequently: $l^{\prime}(e^{\prime})=(Tf^{\prime}_{v}-l\circ Tf^{\prime}_{h})_{\|F}^{-1}(l\circ Tf^{\prime}_{h}-Tf^{\prime}_{v})(e^{\prime}).$ Hence, the expression of $l^{\prime}=\phi_{f^{\prime}x}(l)$ depends algebraically on $l$, and the coefficients of this algebraic expression depends continuously on $f^{\prime}$ or $x$, with respect to some trivializations. When $f^{\prime}$ is equal to $f$ and $x$ belongs to the zero section of $F_{\|V}$, the map $\phi_{fx}\;:\;l^{\prime}\mapsto(Tf_{v}-l\circ Tf_{h})_{\|F}^{-1}(l\circ Tf_{h})$ is $\lambda$-contracting for $l$ small, by normal expansion. Thus, for $\epsilon$ small enough, for all $x\in B_{F\|V}(0,\epsilon)$ and $l\in B_{P^{1}_{\hat{f}(x)}}(0,\epsilon)$, the tangent map of $\phi_{fx}$ has a norm less than $\lambda$. Therefore, by taking $\epsilon$ slightly smaller, for $V_{f}$ small enough, the tangent map $T\phi_{f^{\prime}x}$ has a norm less than $\lambda$ and hence $\phi_{f^{\prime}x}$ is a $\lambda$-contraction on $cl\big{(}{B}_{P_{\hat{f}(x)}}(0,\epsilon)\big{)}$. As, for $x\in V$, the map $\phi_{fx}$ vanishes at $0$, for $V_{f}$ small enough, the norm $\phi_{f^{\prime}x}(0)$ is less than $(1-\lambda)\cdot\epsilon.$ Consequently, by $\lambda$-contraction, the closed $\epsilon$-ball centered at the $0$-section is sent by $\phi_{f^{\prime}x}$ into the $\epsilon$-ball centered at $0$, for all $x\in B_{F\|V}(0,\epsilon)$ and $f^{\prime}\in V_{f}$. ∎ For any $C^{1}$-section $i^{\prime}$ of $(F,\mathcal{F})\rightarrow(L,\mathcal{L})$, the tangent space to the image of $i^{\prime}$ at $i^{\prime}(x)$, for every $x\in L$, is an element $\nabla i^{\prime}(x)$ of $P^{1}_{i^{\prime}(x)}$ and so of $T_{x}\mathcal{L}^{}\oplus F_{x}$. By conclusion 2 of Lemma 6.1, for $x_{0}\in V$, for $f^{\prime}\in V_{f}$, for $j\in V_{i}^{0}\cap\Gamma$, the $d$-planes $\nabla j(x_{0})$ is sent by $T_{x}\hat{f^{\prime}}$ to $\nabla S^{0}(f^{\prime},j)(y_{0})$, with $x:=j(x_{0})$ and $y_{0}:=\pi\circ\hat{f}^{\prime}(x)$. Thus $\nabla S^{0}(f^{\prime},j)(y_{0})$ is equal to $\phi_{f^{\prime}x}(\nabla j(y_{0}))$. We remind that the interest of such a distribution $\chi$ is that for any morphism $p\in Mor^{1}(\mathcal{L},\mathbb{R}^{+})$, we have: (6) $\nabla(p\cdot i^{\prime})(x)=Tp(x)\cdot i^{\prime}(x)+p(x)\cdot\nabla i^{\prime}(x).$ Such equality is proved in section 6.4, Equation (3). Thus we have for any $x_{0}\in V$: $\nabla S_{f^{\prime}}(j)(x_{0})=\rho(x_{0})\cdot\phi_{f^{\prime}x}(\nabla j(y_{0}))+T\rho(x_{0})\cdot z,\quad\mathrm{with}\;z:=S_{f^{\prime}}(j)(x_{0})\;\mathrm{and}\;y_{0}:=\pi\circ\hat{f}^{\prime}(x)$ In other notations: $\nabla S_{f}(j)(x_{0})=\phi^{\prime}_{f^{\prime}z}(\nabla j(y_{0})),\quad\mathrm{with}\;\phi^{\prime}_{f^{\prime}z}\;:\;l\in P^{1}_{\hat{f}^{\prime}(z)}\mapsto\rho\circ\pi(z)\cdot\phi_{f^{\prime}z}(l)+T_{\pi(z)}\rho\cdot z\in P^{1}_{z},$ for any $z\in B_{F\|V}(0,\epsilon)$, for $z$ small enough. We notice that $\phi^{\prime}_{f^{\prime}z}$ is as much contracting as $\phi_{f^{\prime}z}$ and sends the $\epsilon$-ball of $P^{1}_{\hat{f}(z)}$ centered at 0 into the $\epsilon$-ball of $P^{1}_{z}$ centered at $0$. ### 7.5 Proof of Theorem 3.1 when $r=1$ and in the normally expanded case Let us begin by proving the existence of closed neighborhood $V_{i}$ of $i\in\Gamma$ sent by $S_{f^{\prime}}$ into itself, for any $f^{\prime}\in End^{r}(M)$ close to $f$. For $\epsilon>0$, then $\epsilon^{\prime}>0$ and finally $V_{f}$ small enough, the following set is included in $V_{i}^{0}$: $V_{i}:=\Big{\\{}i^{\prime}\in\Gamma;\quad d_{0}(i,i^{\prime})\leq\epsilon^{\prime},\;\mathrm{and}\;\;\\|\nabla i^{\prime}(x)\\|\leq\epsilon,\;\forall x\in V\Big{\\}}.$ Therefore, by the two last steps, $S_{f^{\prime}}$ sends $V_{i}$ into itself, for every $f^{\prime}\in V_{f}$. We already showed that $S$ is $C^{0}$-contracting for the $C^{0}$-topology. ###### Lemma 7.3. For every $f^{\prime}\in V_{f}$, for every $\delta>0$, there exists $N>0$ such that: for all $i^{\prime},i^{\prime\prime}\in S^{N}(V_{i})$, for every $x\in L$: $d(\nabla i^{\prime}(x),\nabla i^{\prime\prime}(x))<\delta.$ Therefore, the sequence $(S_{f^{\prime}}^{k}(i))_{k}$ converges in $Im^{1}(\mathcal{L},M)$ to some $C^{1}$-immersion $i^{\prime}\in V_{i}$. Such an immersion is unique by the $C^{0}$-contraction of $S_{f^{\prime}}$ and so, is a fixed point of $S_{f^{\prime}}$. We define the $C^{1}$-morphism $f^{\prime}:=x\mapsto\pi\circ\hat{f}^{\prime}\circ i^{\prime}(x)$ from $\mathcal{L}_{\|L^{\prime}}$ to $\mathcal{L}$. Such a morphism is equivalent to $f^{}_{\|L^{\prime}}$, and satisfies with $i(f^{\prime})$ conclusions $(1)$ and $(2)$ of Theorem 3.1. ###### Proof of Lemma 7.3. For any $\delta>0$, there exists $N^{\prime}\geq 0$ such that $\lambda^{N^{\prime}}\cdot\epsilon$ is less than $\delta$. For $k\geq 0$, let $U_{k}$ be the closure of $\cup_{i^{\prime}\in V_{i}}S^{k}_{f^{\prime}}(i^{\prime})$. Thus $U_{N^{\prime}}$ is identified to a compact subset of $F$. Let $\sigma^{\epsilon,k}$ be the space of continuous sections of the bundle $P^{1}_{\|U_{k}}$ over $U_{k}$ whose fiber at $x\in U_{k}$ consists of planes of $P^{1}_{x}$ with norm less than $\epsilon$. The following map is well defined and contracting: $f^{\prime\\#}:\sigma^{\epsilon,0}\longrightarrow\sigma^{\epsilon,1}$ $\sigma\longmapsto\left[x\in U_{1}\mapsto\left\\{\begin{array}[]{cl}\phi_{f^{\prime}x}\Big{(}\sigma\circ\hat{f}\big{(}x/\rho\circ\pi(x)\big{)}\Big{)}&\mathrm{if}\;\rho\circ\pi(x)\not=0\\\ 0&\mathrm{else}\end{array}\right.\right]$ Thus there exists a continuous section $\sigma_{f^{\prime}}\in\sigma^{\epsilon,N^{\prime}}$ in $f^{\prime\\#^{N^{\prime}}}(\sigma^{\epsilon,0})$. Moreover, the diameter of $\sigma^{\epsilon,N^{\prime}}$ is less than $\delta$ for the uniform norm induced by the one of $P^{1}$. By continuity of $\sigma_{f^{\prime}}$ and compactness of $U_{N^{\prime}}$, there exist $e>0$ such that for every $x\in V$, $(y,y^{\prime})\in U_{N^{\prime}}\cap F_{x}$, if $\\|y-y^{\prime}\\|<e$ then $\\|\sigma_{f^{\prime}}(y)-\sigma_{f^{\prime}}(y^{\prime})\\|$ is less than $\delta$. Let $N\geq N^{\prime}$ such that $\lambda^{N}\cdot\epsilon$ is less than $e$. Thus the $C^{0}$-diameter of $S^{N}_{f^{\prime}}(V_{i})$ is less than $e$. Consequently, for $(i^{\prime},i^{\prime\prime})\in S^{N}_{f^{\prime}}(V_{i})$, the norm $\\|\nabla i^{\prime}(x)-\nabla i^{\prime\prime}(x)\\|$ is less than: $d\Big{(}\nabla i^{\prime\prime},\sigma_{f^{\prime}}\big{(}i^{\prime\prime}(x)\big{)}\Big{)}+d\Big{(}\sigma_{f^{\prime}}(i^{\prime\prime}(x)),\sigma_{f^{\prime}}(i^{\prime}(x))\Big{)}+d\Big{(}\sigma_{f^{\prime}}\big{(}i^{\prime}(x)),\nabla i^{\prime}(x)\Big{)}$ $\Rightarrow d(Ti^{\prime\prime}(T_{x}\mathcal{L}),Ti^{\prime}(T_{x}\mathcal{L}))\leq 3\delta$ The last inequality concludes the proof of the lemma. ∎ Let us prove conclusion $(3)$ of Theorem 3.1. Let $(y_{n})_{n\geq 0}\in M^{\mathbb{N}}$ be a $f^{\prime}$-orbit which is $\epsilon$-close to the image by $i$ of a $\epsilon$-pseudo-orbit $(x_{n})_{n\geq 0}\in L^{\prime\mathbb{N}}$ of $f^{}$ respecting the plaques. Let $y_{n}^{\prime}:=(I_{x_{n}}^{\eta})^{-1}(y_{n})$ and $z_{n}:=\pi(y^{\prime}_{n})$. We want to show that $i(f^{\prime})(z_{n})$ is equal to $y_{n}$, for every $n\geq 0$. Let $n>0$ be such that $i(f^{\prime})(z_{n})$ and $y_{n}$ are at distance in $F_{z_{n}}$ at least: (7) $\frac{\lambda+1}{\lambda}\sup_{j>0}d_{F_{z_{j}}}\big{(}i(f^{\prime})(z_{j}),y_{j}\big{)}.$ For $\epsilon>0$ small, the sequence $(z_{j})_{j\geq 0}$ is close to $L^{\prime}$. Also we can construct a section $i^{\prime}\in V_{i}$ equal to $y_{j}$ at $z_{j}$ for $j>0$, and such that the $C^{0}$-distance between $i^{\prime}$ and $i(f^{\prime})$ is the one between $\sup_{j>0}d_{F_{z_{j}}}\big{(}i(f^{\prime})(z_{j}),y_{j}\big{)}$. We remind that on $V^{\prime}\subset V$ the maps $S$ and $S^{0}$ are equal, since $\rho$ is there equal to 1. By contraction of $S^{0}$, the $C^{0}$-distance between $i(f^{\prime})_{\|V^{\prime}}$ and $S^{0}(f^{\prime},i^{\prime})_{\|V^{\prime}}$ is less than $\lambda\cdot d(i(f^{\prime})(z_{n}),y_{n})$. By conclusion $(2)$ of Lemma 7.1, $S^{0}_{f^{\prime}}(i^{\prime})$ sends $(z_{j})_{j\geq 0}$ to $(y_{j})_{j>0}$. By Equation (7), for every $j>0$, the point $z_{j}$ is sent by $i(f^{\prime})$ to $z_{j}$. By normal expansion, the point $z_{0}$ is sent by $i(f^{\prime})$ to $z_{0}$. ### 7.6 General case: $r>1$ Unfortunately, as we deal with maps which have possibly singularities along the leaves, the action of $Tf$ on the Grassmannian is not as well defined as in Theorem 3.4. I do not know any other way to prove this result than the following calculus. As we deal with the $C^{r}$-topology, we shall generalize the Grassmannian concept as follow: For $x\in U^{}$, let $G_{x}^{r}$ be the set of the $C^{r}$-$d$-submanifolds of $M$ which contain $I(x)$, quotiented by the following $r$-tangent relation: Two such submanifolds $N$ and $N^{\prime}$ are equivalent if there exists a chart $(U,\phi)$ of a neighborhood of $I(x)\in M$, which sends $N\cap U$ onto $\mathbb{R}^{d}\times\\{0\\}$ and sends $N^{\prime}\cap U$ onto the graph of a map from $\mathbb{R}^{d}$ into $\mathbb{R}^{n-d}$, whose $r$-first derivatives vanish at $\phi(x)$. We notice that for $r=1$, this space is the Grassmannian of $d$-planes of $T_{x}\mathcal{F}\approx T_{I(x)}M$. As we are interested in the submanifolds “close” to the embedding of small $\mathcal{L}$-plaques by $i$, we restrict our study to the $d$-submanifolds containing $I(x)$ and transverse to $F_{\pi(x)}^{\epsilon}$. The preimage of a submanifold of this equivalent class by the map $\text{exp}\circ T_{x}I$ is a graph of a $C^{r}$-map $\overline{l}$ from $\chi(x)$ to $F_{\pi(x)}$, with exp the exponential associated to a Riemannian metric of $M$. Moreover, its $r$-tangent equivalence class can be identified to the Taylor polynomial of $\overline{l}$ at $0$: $\overline{l}(u)=T_{0}\overline{l}(u)+\frac{1}{2}T^{2}_{0}\overline{l}(u^{2})+\cdots+\frac{1}{r!}T^{r}_{0}\overline{l}(u^{r})+o(\\|u\\|^{r}),$ where $u$ belongs to $\chi(x)$, the $k^{th}$-derivative $T_{0}^{k}\overline{l}$ belongs to the space $L_{sym}^{k}(\chi(x),F_{\pi(x)})$ of $k$-linear symmetric maps from $\chi(x)^{k}$ into $F_{\pi(x)}$. We notice that we abused of notation by writing $u^{k}$ instead of $(u,\dots,u)$. The map $l(u):=\sum_{k=1}^{n}\frac{1}{k}T^{k}_{0}\overline{l}(u)$ is an element of the vector space: $P_{x}^{r}:=\oplus_{k=1}^{r}L_{sym}^{k}(\chi(x),F_{\pi(x)}).$ Conversely, any vector $l\in P_{x}^{r}$ that we write in the form: $l\>:\;u\in\chi(x)\mapsto\sum_{k=1}^{r}l_{k}(u^{k})$ is the class of the following $C^{r}$-$d$-submanifold of $M$: $\text{exp}\circ TI\Big{(}\\{(u+l(u));\;u\in\chi(x)\;\mathrm{and}\;\\|u\\|<r_{i}(x)\big{\\}}\Big{)},$ where $r_{i}$ depends continuously on the injectivity radius of exp. The linear map $l_{1}$ from $\chi(x)$ to $F_{\pi(x)}$ is called the _linear part of $l$_. We notice that $l_{1}$ belongs to $P^{1}_{x}$. We denote by $P^{r}$ the vector bundle over $U^{}$, whose fiber at $x$ is $P^{r}_{x}$. By normal expansion, for $U^{}$ small enough and $f^{\prime}$ close enough to $f$, for any point $x\in U^{}$ sent by $\hat{f}^{\prime}$ to some $y\in U^{}$, any $l\in P^{r}_{y}$ whose linear part is small enough, the preimage by $f^{\prime}$ of a representative of $l$ is a representative of a vector $\phi_{f^{\prime}x}(l)\in P^{r}_{x}$, which depends only on $l$. Let us show the following lemma: ###### Lemma 7.4. For every $\epsilon>0$ small enough and then $V_{f}$ small enough, for all $f^{\prime}\in V_{f}$, $x\in\pi^{-1}(V)\cap U^{}$ and $l\in P^{r}_{\hat{f}^{\prime}(x)}$ with linear part of norm at most $\epsilon>0$, the norm of the linear part of $\phi_{f^{\prime}}(l)$ is then less than $\epsilon$. Moreover, the map $\phi_{f^{\prime}}$ is of the form $(l_{m})_{m}\in P^{r}_{\hat{f}(x)}\mapsto\Big{[}C_{m}^{f^{\prime}}(l_{m})+\Psi_{m}\big{(}(l_{k})_{1\leq k<m}\big{)}\Big{]}_{m}\in P^{r}_{x}$ where $\Psi_{m}\big{(}(l_{k})_{1\leq k\leq m}\big{)}$ does not depend on $(l_{i})_{i>m}$. Also $C_{m}^{f^{\prime}}$ is a $\lambda$-contracting endomorphism of $L_{sym}^{m}(\chi(x),F_{\pi(x)})$ endowed with the norm subordinate to $\chi(x)$ and $F_{\pi(x)}$, and $\chi(x)$ is endowed with the norm induced by $T_{x}\mathcal{F}$. For $i^{\prime}\in\Gamma$ and $l\in\\{1,\dots,r\\}$, let $\nabla^{l}i^{\prime}$ be the section $\pi_{v}\circ T^{l}i^{\prime}$, where $\pi_{v}$ is the projection of $T\mathcal{F}$ onto $F$ parallelly to $\chi$ and $T^{l}i^{\prime}$ is the $l$-tangent maps of $i^{\prime}$. Thus $\nabla^{l}i^{\prime}(x)$ is a vector of $L_{sym}^{l}(T_{x}\mathcal{L},F_{x})$, for every $x\in L$. As in section 6.4, we have for any $i^{\prime}\in\Gamma$: $\nabla^{l}(\rho\cdot i^{\prime})(x)=\sum_{k=0}^{l}C_{l}^{k}T^{l-k}\rho\cdot\nabla^{k}i^{\prime}(x).$ Thus, $\nabla^{l}(S_{f^{\prime}}(i^{\prime}))(x)=\rho(x)\cdot\nabla^{l}S^{0}(f^{\prime},i^{\prime})(x)+\Psi_{m}\big{(}(\nabla^{k}S^{0}(f^{\prime},i^{\prime})(x))_{1\leq k<l}\big{)}$, where $\Psi^{\prime}_{m}$ is a continuous function which does not depend on $\big{(}\nabla^{k}S^{0}(f^{\prime},i^{\prime})(x)\big{)}_{k\geq l}$. On the other hand, by conclusion $2$ of Lemma 7.1 and by Lemma 7.4, for any $x\in V$ and $i^{\prime}\in V_{i}^{0}$, $(\nabla^{k}S^{0}(f^{\prime},i^{\prime})(x_{0}))_{k=1}^{r}$ is equal to the image by $\phi_{f^{\prime}x}$ of $(\nabla^{k}i^{\prime}(y_{0}))_{k=1}^{r}$, with $x_{0}\in V$, $x=S^{0}(f^{\prime},i^{\prime})(x_{0})$ and $y_{0}:=\pi\circ\hat{f}^{\prime}\circ S^{0}(f^{\prime},i^{\prime})(x_{0})$. Thus $(\nabla^{k}S_{f^{\prime}}(i^{\prime})(x_{0}))_{k=1}^{r}$ is equal to the image of $\big{(}(\nabla^{k}i^{\prime})(y_{0})\big{)}_{k=1}^{r}$ by a continuous function $\phi^{\prime}_{f^{\prime}x}$ of the form: $\phi^{\prime}_{f^{\prime}x}:\;(l_{m})_{m}\in P^{r}_{\hat{f}(x)}\mapsto\Big{[}\rho(x)\cdot C_{m,x}^{f^{\prime}}(l_{m})+\Psi^{\prime}_{m}((l_{k})_{k=1}^{m})\Big{]}_{m}$ Such a function is contracting for a good norm on $P^{r}$ independent on the base point. To conclude the proof, we proceed as in the previous section. ###### Proof of Lemma 7.4. We showed in the case $r=1$ that $\phi_{f^{\prime}x}$ preserves the subset of vectors of $P^{1}$ with norm at most $\epsilon$, and hence preserves the subset of $P^{r}$ formed by maps of linear part at most $\epsilon$. Let us show the $\lambda$-contraction of $\phi_{f^{\prime}x}$. Let $l^{\prime}:=\phi_{f^{\prime}}(l)$. Let $J_{z}^{r}f^{\prime}$ be the $r$-jet map of $f^{\prime}$ at $z:=I(x)$ (see [Mic80]). We remind that the $r$-jet $J_{z}^{r}f^{\prime}$ of $f^{\prime}$ at $z$ is a vector of $\bigoplus_{j=1}^{r}L_{sym}^{j}(T_{z}M,T_{f^{\prime}(z)}M)$ such that, if we denote by $f_{j}^{\prime}$ its component in $L_{sym}^{j}(T_{z}M,T_{f^{\prime}(z)}M)$, we have: $\text{exp}^{-1}_{f^{\prime}(z)}\circ f^{\prime}\circ\text{exp}_{z}(u)=\sum_{j=1}^{r}f^{\prime}_{j}(u^{j})+o(\\|u\\|^{r}),\;\mathrm{for}\;u\in T_{z}M$ Let $J_{x}^{r}\hat{f}^{\prime}:=(T_{\hat{f}^{\prime}(x)}I)^{-1}\circ J_{z}^{r}f^{\prime}\circ T_{x}I$ and $J_{x}^{r}\hat{f}^{\prime}=:\sum_{j=1}^{r}\hat{f}^{\prime}_{j}\in\bigoplus_{j=1}^{r}L_{sym}^{j}(T_{x}\mathcal{F},T_{\hat{f}^{\prime}(x)}\mathcal{F})$. By definition of $l^{\prime}:=\phi_{f^{\prime}x}(l)$, for any $u^{\prime}\in\chi(x)$, there exists $u\in\chi(\hat{f}^{\prime}(x))$ such that: (8) $J_{x}^{r}\hat{f}^{\prime}(u^{\prime}+l^{\prime}(u^{\prime}))=u+l(u)+o(\\|u\\|^{r}).$ We recall that $\pi_{v}$ and $\pi_{h}$ denote the projection of $T\mathcal{F}$ onto respectively $F$ and $\chi$. By (8), we have: $u:=\pi_{h}\circ J^{r}_{x}\hat{f}^{\prime}(u^{\prime}+l^{\prime}(u^{\prime}))+o(\\|u^{\prime}\\|^{r})\;\mathrm{and}\;l(u)=\pi_{v}\circ J^{r}_{x}\hat{f}^{\prime}(u^{\prime}+l^{\prime}(u^{\prime}))+o(\\|u^{\prime}\\|^{r}).$ Thus, we have: (9) $l\circ\pi_{h}\circ J^{r}_{x}\hat{f}^{\prime}(u^{\prime}+l^{\prime}(u^{\prime}))=\pi_{v}\circ J^{r}_{x}\hat{f}^{\prime}(u^{\prime}+l^{\prime}(u^{\prime}))+o(\\|u^{\prime}\\|^{r}).$ We have: (10) $J^{r}_{x}\hat{f}^{\prime}(u^{\prime}+l^{\prime}(u^{\prime}))=\sum_{I\in R}\hat{f}^{\prime}_{\|I\|}\Big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\Big{]}+o(\\|u^{\prime}\\|^{r}),$ where $R$ is the set $\cup_{k=1}^{r}\\{0,\dots,r\\}^{k}$; $l^{\prime}_{0}(u^{\prime 0})$ is equal to $u^{\prime}$; and for $I\in R$, $\|I\|$ denotes the length of $I$. Let $f^{\prime}_{kv}$ and $f^{\prime}_{kh}$ be respectively the linear maps $\pi_{v}\circ\hat{f}^{\prime}_{k}$ and $\pi_{h}\circ\hat{f}^{\prime}_{k}$ respectively, for every $k\in\\{0,\dots,r\\}$. It follows from Equations (9) and (10) that: (11) $l\Big{(}\sum_{I\in R}f^{\prime}_{\|I\|h}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}\Big{)}=\sum_{I\in R}f^{\prime}_{\|I\|v}\big{[}\prod_{k\in I}l_{k}^{\prime}(u^{\prime k})\big{]}+o(\\|u^{\prime}\\|^{k})$ On the one hand, we have: (12) $\sum_{I\in R}f^{\prime}_{\|I\|v}\big{[}\prod_{k\in I}l_{k}^{\prime}(u^{\prime k})\big{]}=\sum_{m=1}^{r}\left[\sum_{I\in R,\;\Sigma I=m}f^{\prime}_{\|I\|v}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}\right]+o(\\|u^{\prime}\\|^{r})$ with, for every $I\in R$, the integer $\Sigma I$ equal to $\sum_{j\in I}j$ plus the number of times that $0$ appears in $I$. On the other, as $l\Big{(}\sum_{I\in R}f^{\prime}_{\|I\|h}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}\Big{)}=\sum_{a=1}^{r}l_{a}\Big{(}\sum_{I\in R}f^{\prime}_{\|I\|h}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}\Big{)}^{a}$ and as, $\Big{(}\sum_{I\in R}f^{\prime}_{\|I\|h}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}\Big{)}^{a}=\sum_{(I_{\alpha})_{\alpha}\in R^{a}}\prod_{\alpha=1}^{a}f^{\prime}_{\|I_{\alpha}\|h}\big{[}\prod_{k\in I_{\alpha}}l_{k}^{\prime}(u^{\prime k})\big{]}$ the polynomial map $l\Big{(}\sum_{I\in R}f^{\prime}_{\|I\|h}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}\Big{)}$ is equal to: (13) $\sum_{m=1}^{r}\quad\sum_{\\{(I_{\alpha})_{\alpha\in A}\in R^{},\;\sum_{\alpha}\Sigma I_{\alpha}=m\\}}l_{\|A\|}\Big{[}\prod_{\alpha\in A}f^{\prime}_{\|I_{\alpha}\|h}\big{[}\prod_{k\in I_{\alpha}}l_{k}^{\prime}(u^{\prime k})\big{]}\Big{]}$ with $R^{}:=\cup_{a=1}^{r}R^{a}$. By identification, it follows from equations (11), (12) and (13) and that, for every $m\in\\{1,\dots,r\\}$: $\underbrace{\sum_{\\{(I_{\alpha})_{\alpha\in A}\in R^{},\;\sum_{\alpha}\Sigma I_{\alpha}=m\\}}l_{\|A\|}\Big{[}\prod_{\alpha\in A}f^{\prime}_{\|I_{\alpha}\|h}\big{[}\prod_{k\in I_{\alpha}}l_{k}^{\prime}(u^{\prime k})\big{]}\Big{]}}_{l^{\prime}_{m}\;\mathrm{only\;occurs\;for}\;(I_{\alpha})_{\alpha}=((m));\;l_{m}\;\mathrm{only\;for}\;(I_{\alpha})_{\alpha}\in\\{\\{0\\},\\{1\\}\\}^{m}}=\underbrace{\sum_{I\in R,\;\Sigma I=m}f^{\prime}_{\|I\|v}\big{[}\prod_{k\in I}l^{\prime}_{k}(u^{\prime k})\big{]}}_{\mathrm{Here}\;l^{\prime}_{m}\;\mathrm{only\;occurs\;for}\;I=(m).}$ Thus, there exists an algebraic function $\phi$, such that $f^{\prime}_{1v}\circ l^{\prime}_{m}(u^{\prime m})$ is equal to: $l_{1}\circ f_{1h}^{\prime}\circ l^{\prime}_{m}(u^{\prime m})+\sum_{(i_{\alpha})_{\alpha=1}^{m}\in\\{0,1\\}^{m}}l_{m}\Big{(}\prod_{\alpha=1}^{m}f_{1h}^{\prime}\circ l_{i\alpha}^{\prime}(u^{\prime i_{\alpha}})\Big{)}+\phi\big{(}(l_{i})_{i<m},(l_{i}^{\prime})_{i<m},(f^{\prime}_{i})_{i=1}^{r}\big{)}$ Since the linear part $l_{1}$ of $l$ is small, we have: $l_{m}^{\prime}=(f^{\prime}_{1v}-l_{1}\circ f^{\prime}_{1h})_{\|F_{\pi(x)}}^{-1}\Big{[}\sum_{I\in\\{0,1\\}^{m}}l_{m}\circ\prod_{k\in I}f^{\prime}_{1h}\circ l_{k}+\phi\big{(}(l_{i})_{i<m},(l_{i}^{\prime})_{i<m},(f_{i})_{i=1}^{r}\big{)}\Big{]}.$ For $x\in V$ and $f^{\prime}=f$, we have $f^{\prime}_{1h\|F_{x}}=0$. Thus, $\sum_{I\in\\{0,1\\}^{m}}l_{m}\prod_{k\in I}f^{\prime}_{1h}\circ l_{k}=l_{m}\circ\big{(}f^{\prime}_{1h}\big{)}^{m}.$ It follows from the $r$-normal expansion that the map $C\;:\;l_{m}\mapsto(f^{\prime}_{1v}-l_{1}\circ f^{\prime}_{1h})_{\|F_{x}}^{-1}\circ l_{m}\circ\big{(}f^{\prime}_{1h}\big{)}^{m}$ is $\lambda$-contracting, when $l_{1}$ is small. Also, the map $l_{s}^{\prime}$ is an algebraic function of only $(l_{k})_{k\leq s}$ and $(f^{\prime}_{k})_{j\leq r}$, for $s\leq m$. Thus, $l_{m}^{\prime}=C_{m}(l_{m})+\phi\big{(}(l_{i})_{i<m},(f_{i}^{\prime})_{i}\big{)}$ This implies that for a norm on $P^{r}$, the map $\phi_{fx}$ is contracting, for $U^{}$ small enough and then $f^{\prime}$ $C^{r}$-close to $f$. ∎ ## 8 Proof of persistences of complex laminations by deformation Let us prove the persistence of $0$-normally expanded complex laminations by deformation. The persistence of $0$-normally contracted laminations is proved similarly. Let $(F,\mathcal{F},\pi,I)$ be a differentiable tubular neighborhood of the none necessarily compact lamination $(L,\mathcal{L})$ immersed by $i$. Let $L^{\prime}$ be a precompact open subset of $L$ whose closure is sent into $L^{\prime}$ by the pullback $f^{}$. Let $B_{0}$ be a precompact, open subset of $B$, that we will restrict. Let $\hat{F}$ be the product $B_{0}\times F_{\|L^{\prime}}$ and $\hat{\pi}\;:\;(x,t)\in B_{0}\times F_{\|L^{\prime}}\rightarrow(x,t)\in B_{0}\times L^{\prime}$. Let $Z$ be the set of maps of the form: $i_{\sigma}\;:\;(t,x)\in B_{0}\times L^{\prime}\mapsto(t,I\circ\sigma(t,x))\in B_{0}\times M,$ where $\sigma$ is a $C^{0}$-bounded, differentiable section of $\hat{\pi}\;:\;\hat{F}\rightarrow B_{0}\times L^{\prime}$, such that $i_{\sigma}$ sends any small plaque of the product lamination $B_{0}\times\mathcal{L}_{\|L^{\prime}}$ onto a complex submanifold of $B_{0}\times M$. We notice that $Z$ is not empty since it contains $i_{0}\;:\;(t,x)\mapsto(t,i(x))$. ###### Lemma 8.1. The subspace $Z$ equipped with the $C^{0}$-norm: $\\|i_{\sigma}\\|=\sup_{(t,x)\in B_{0}\times L^{\prime}}\\|\sigma(t,x)\\|$ is a complete space ###### Proof. This follows from the Cauchy integral theorem.∎ Let $\hat{S}\;:\;i_{\sigma}\in Z\longmapsto\Big{[}(t,x)\mapsto\big{(}t,S^{0}(f_{t},i_{\sigma}(t,\sigma(t,\cdot))\big{)}\Big{]}$, with $S^{0}$ defined by Lemma 7.1 with $r=1$. By restricting $B_{0}$, such a map is well defined since this lemma only uses the $0$-normal expansion. By conclusion 2 of Lemma 7.1, the image $V^{\prime}$ by $\hat{S}(i_{\sigma})$ of a small plaque of $B_{0}\times\mathcal{L}_{\|L^{\prime}}$ is sent by $\hat{f}\;:\;(t,x)\in B_{0}\times M\mapsto(t,f_{t}(x))$ into the image $V$ by $i_{\sigma}$ of a small plaque of $B_{0}\times\mathcal{L}_{\|L^{\prime}}$. By $0$-normal expansion, the map $\hat{f}$ is transverse to $V$ at $V^{\prime}$: $\forall y\in V^{\prime},\;T\hat{f}\big{(}T_{y}(B_{0}\times M)\big{)}+T_{\hat{f}(y)}V=T_{\hat{f}(y)}(B_{0}\times M).$ Thus, $V^{\prime}$ is an open subset of $\hat{f}^{-1}(V)$, since $V$ and $V^{\prime}$ have the same dimension. As $\hat{f}$ and $V$ are complex analytic, $V^{\prime}$ is also complex analytic. Thus, $\hat{S}(i_{\sigma})$ belongs to $Z$. Finally, by restricting $B_{0}$, the map $\hat{S}$ is $C^{0}$-contracting, since subsection 7.3 needs only the $0$-normal expansion. Consequently, $\hat{S}$ has a unique fixed point $\underline{i}\;:\;(t,x)\in B_{0}\times L^{\prime}\mapsto(t,i_{t}(x))$ in $Z$. By Lemma 5.2, $B_{0}\times M$ induces via $\underline{i}$ a unique complex laminar structure $\mathcal{D}$ on $D$, compatible with the $C^{1}$-laminar structure and the complex structure $B_{0}\times\mathcal{L}_{\|L^{\prime}}$. By compatibility of the structure, the projection $B_{0}\times L^{\prime}\mapsto B_{0}$ induces a holomorphic submersion $\overline{w}\;:\;(D,\mathcal{D})\rightarrow B_{0}$. We notice that $S(f,i)=i$ and that $\underline{i}_{\|\\{t_{0}\\}\times L^{\prime}}$ is $i$: by uniqueness stated in Lemma 5.2, the structure $\overline{w}^{-1}(t_{0})=(L^{\prime},\mathcal{L}_{t_{0}})$ is equal to $(L^{\prime},\mathcal{L}_{\|L^{\prime}})$. We notice that $f_{t}$ preserves the immersed lamination $(L,\mathcal{L}_{t})=\overline{w}^{-1}(t)$ by $i_{t}=\underline{i}_{\|\overline{w}^{-1}(t)}$ since $\underline{i}$ is a fixed point of $S$ and hence since $S^{0}(f_{t},i_{t})$ equals $i_{t}$. ## 9 Proof of persistences of normally hyperbolic laminations ### 9.1 Existence of laminar structures on the stable and unstable sets of normally hyperbolic laminations ###### Proposition 9.1. Under the hypotheses of Theorem 3.1 in the normally hyperbolic case, there exist two laminations $(L^{s},\mathcal{L}^{s})$ and $(L^{u},\mathcal{L}^{u})$ $C^{r}$-immersed by respectively $i^{s}$ and $i^{u}$ into $M$, such that: - $f$ $r$-normally expands $(L^{s},\mathcal{L}^{s})$ over a morphism $f_{s}^{}$ from a the restriction of $\mathcal{L}^{s}$ to an open, precompact subset $L^{\prime s}$ into $\mathcal{L}^{s}$. - $f$ $r$-normally contracts $(L^{u},\mathcal{L}^{u})$ over an injective, open immersion $f_{u}^{}$ from the restriction of $\mathcal{L}^{u}$ to a precompact open subset $D^{u}$ to $\mathcal{L}^{u}$, - there exist two canonical $(C^{r})$-inclusions of $(L,\mathcal{L})$ into $(L^{s},\mathcal{L}^{s})$ and $(L^{u},\mathcal{L}^{u})$ such that the following diagram is well defined and commutes: $\begin{array}[]{rclll}L^{\prime s}&\stackrel{{\scriptstyle f_{s}^{}}}{{\rightarrow}}&L^{s}&&\\\ \uparrow&&\uparrow&\stackrel{{\scriptstyle i^{s}}}{{\searrow}}&\\\ L^{\prime}&\stackrel{{\scriptstyle f^{}}}{{\rightarrow}}&L&\stackrel{{\scriptstyle i}}{{\rightarrow}}&M\\\ \downarrow&&\downarrow&\stackrel{{\scriptstyle i^{u}}}{{\nearrow}}&\\\ D^{u}&\stackrel{{\scriptstyle f_{u}^{}}}{{\rightarrow}}&L^{u}&&\\\ \end{array}$ - for every $\epsilon>0$ small enough, all points $x\in L^{\prime}$ satisfy333We remind that $\mathcal{L}^{\epsilon}_{x}$ is the union of $\mathcal{L}$-plaques with diameter less than $\epsilon$ which contain $x$. $i(\mathcal{L}_{x}^{\epsilon})=i^{s}(\mathcal{L}_{x}^{s\epsilon})\;{\overline{\pitchfork}}\;i^{u}(\mathcal{L}_{x}^{u\epsilon}).$ * - let $x\in L^{\prime}$ with $f^{}$-forward orbit included in $L^{\prime}$. Then a local strong stable manifold of $i(x)$ is included in the immersion by $i^{s}$ of a $\mathcal{L}^{s}$-plaque of $x$, - let $x\in L^{\prime}$ with $f^{}$-backward orbit included in $L^{\prime}$. Then a local strong unstable manifold of $i(x)$ is included in the immersion by $i^{u}$ of a $\mathcal{L}^{u}$-plaque of $x$, ###### Proof of Proposition 9.1. We remind that $r$ is positive. Definition of $(L^{u},\mathcal{L}^{u})$ Since $f^{}$ is bijective, we can suppose $Tf$-invariance of the strong unstable direction $E^{u}$. we notice that the section of the Grassmannian $x\in L\mapsto E^{u}_{x}$ is continuous but not necessarily differentiable. However by Appendix A.2.1 of [Ber], there exists a $C^{r}$-lifting $E^{\prime u}$ of $i$ into the Grassmannian of $TM$ of class $C^{r}$ which is $C^{0}$-close to the section $E^{u}$. By Lemma 9.2 (see below) the section $E^{\prime u}$ defines a structure of $C^{r}$-lamination $\mathcal{L}^{u}$ on the vector bundle $\pi^{u}\;:\;L^{u}\rightarrow L$ whose fiber at $x\in L$ is $E^{u}_{x}$ and such that: * - the leaves of $\mathcal{L}^{u}$ are the preimages by $\pi$ of the leaves of $(L,\mathcal{L})$, * - $(L,\mathcal{L})$ is canonically $C^{r}$-embedded into $(L^{u},\mathcal{L}^{u})$ as the $0$-section of this bundle, * - for a small, smooth, positive function $\delta$ on $(L^{u},\mathcal{L}^{u})$, the map: $j_{0}\;:\;(x,u)\in L^{u}\mapsto\text{exp}_{x}\left(\frac{\delta(x)\cdot u}{\sqrt{1+u^{2}}}\right)$ is a $C^{r}$-immersion of $\mathcal{L}^{u}$ into $M$, with exp the exponential map associated to a Riemmanian metric of $M$.. Let $(F,\mathcal{F},I,\pi)$ be a tubular neighborhood of the immersed lamination $(L^{u},\mathcal{L}^{u})$ by $j_{0}$. Let $V$ be a small neighborhood of $f^{}(cl(L^{\prime}))$ in $L^{u}$. Let $\eta>0$ be small enough such that for every $x\in V$, the family of submanifolds $(F_{y}^{\eta})_{y\in\mathcal{L}^{u\eta}_{x}}$ are the leaves of a $C^{r}$-foliation, with $F_{y}^{\eta}$ the image by $I$ of the ball of radius $\eta$ and centered at $0$ in $F_{y}$. Let $\Gamma$ be the subset of the $C^{r}$-immersions $j$ of $(L^{u},\mathcal{L}^{u})$ such that: - the restriction of $j$ to $L$ is equal to $i$, * - the restriction of $j$ to the complement of $V$ is equal to $j_{0}$, * - the image by $j$ of $y\in L^{u}$ belongs to $F_{y}^{\eta}$, * - the immersion $j$ is $C^{r}$-close to $j_{0}$ in the $C^{r}$-topology: $\Gamma$ is a small neighborhood of $j_{0}$. By normal hyperbolicity, for $E^{u^{\prime}}$ close enough to $E^{u}$ and also $\Gamma$, and $\delta$ small enough, for every $x\in V$, for every $j\in\Gamma$, $f$ sends $j\circ\pi^{u-1}(\mathcal{L}_{f^{^{-1}}\circ\pi^{u}(x)}^{\eta})$ to a submanifold which intersects transversally at a unique point $S^{0}(j)(x)$ the submanifold $F_{x}^{\eta}$. We notice that $S^{0}(j)$ is plaquewise a $C^{r}$-immersion since it is the composition of the immersion $j_{0}$ of a $\mathcal{L}^{u}-$plaque of $x$ with the holonomy along a $C^{r}$-foliation between two transverse $C^{r}$-sections. As these foliations and transverse sections depend continuously on $x\in V$, the map $S^{0}(j)$ is a $C^{r}$-immersion of $(V,\mathcal{L}^{u}_{\|V})$ into $M$. Let us fix a $C^{r}$-morphism $\rho$ from $(L^{u},\mathcal{L}^{u})$ into $[0,1]$ with support in $V$ and equal to $1$ on a neighborhood $V^{\prime}$ of $f^{}(cl(L^{\prime}))$. Let $S(j)$ be equal to $j_{0}$ on $V^{c}$ and to $\rho\cdot S^{0}(j)$ on $V$, where the scalar product use the vector space structure of the tubular neighborhood $(F,\mathcal{F},I,\pi)$. Such a map $S(j)$ is a $C^{r}$-immersion of $(L^{u},\mathcal{L}^{u})$ into $M$. The existence of a fixed point $i^{u}$ of $S$ is proved similarly as in the proof of the persistence of normally contracted laminations, except that: * - we do not need to deal with the $C^{0}$-topology, since $i$ is equal to the restriction to $L$ of any immersion of the image of $S$, * - we need possibly to take $E^{\prime u}$ closer to $E^{u}$ and $V$ smaller, in order to use the contraction of some map $\phi_{f}$ defined on the neighborhood of $E^{u}\oplus T\mathcal{L}$ in the Grassmannian of $T\mathcal{F}$. Let $D^{u}$ be an open neighborhood of $L^{\prime}$ in $L^{u}$, small enough in order that every point $x\in D^{u}$ have its image by $i^{u}$ sent by $f$ into the image by $i^{u}$ of a small plaque containing $f^{}_{u}\circ\pi(x)$ and included in $V^{\prime}$. Let $f^{}_{u}(x)$ be the point of this plaque such that $i^{u}\circ f^{}_{u}(x)=f\circ i^{u}(x)$. Since the restriction of $Tf$ to $Ti(T\mathcal{L})\oplus E^{u}$ is open and injective, for $E^{\prime u}$ close enough to $E^{u}$, and also $V$ small enough, $f^{}_{u}$ is an open and injective immersion of $(D^{u},\mathcal{L}^{u}_{\|D^{u}})$ into $(L^{u},\mathcal{L}^{u})$. From the construction by graph transform of $i^{u}$, one easily shows that any points $x\in\cap_{n\geq 0}f^{-n}(L^{\prime})$ is sent by $i$ to a point with local strong unstable manifold included in the image by $i^{u}$ of a plaque of $\mathcal{L}^{u}$. The existence of $i^{s}$ and $(L^{s},\mathcal{L}^{s})$ is proved similarly by using this time the proof of Theorem 3.1 in the normally expanded case. Also since $\Gamma$ is a small neighborhood of $j_{0}$, the tangent space $Ti^{u}(\mathcal{L}^{u}_{\|L})$ is close to $Ti(T\mathcal{L})\oplus E^{u}$. Similarly, the tangent space $Ti^{s}(T\mathcal{L}^{s}_{\|L})$ is close to $Ti(T\mathcal{L})\oplus E^{s}$. Therefore $i^{u}$ and $i^{s}$ are transverse at $L^{\prime}$ and so the last conclusion of the proposition is proved. ### 9.2 Differentiable persistence of normally hyperbolic laminations ###### Proof of Theorem 3.1 when the lamination is normally hyperbolic. We apply Theorem 3.1 in the normally expanded case to $(i^{s},f^{}_{s},f,L^{s})$ and Theorem 3.4 to $(i^{u},f^{}_{u},f,D^{u})$ with some precompact, open neighborhoods $L^{\prime s}\subset L^{s}$ and $L^{\prime u}\subset D^{u}$ of $L^{\prime}$. These theorems give a neighborhood $V_{f}$ of $f\in End^{r}(M)$ formed by endomorphisms $f^{\prime}$ for which there exist $i^{s}(f^{\prime})\in Im^{r}(\mathcal{L}^{s},M)$, $i^{u}(f^{\prime})\in Im^{r}(\mathcal{L}^{u},M)$, $f^{\prime}_{s}\in Mor_{f^{}_{s}}(\mathcal{L}^{s}_{\|L^{\prime s}},\mathcal{L}^{s})$ and $f^{\prime}_{u}\in Im^{r}_{f^{}_{u}}(\mathcal{L}^{u}_{\|D^{u}},\mathcal{L}^{u})$ $C^{r}$-close to respectively $i^{s}$, $i^{u}$, $f^{}_{s}$, $f^{}_{u}$ such that the following diagrams commute: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ i^{s}(f^{\prime})&\uparrow&&\uparrow&i^{s}(f^{\prime})\\\ &L^{\prime s}&\rightarrow&L^{s}&\\\ &&f^{\prime}_{s}&&\end{array}\qquad\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ i^{u}(f^{\prime})&\uparrow&&\uparrow&i^{u}(f^{\prime})\\\ &L^{\prime u}&\rightarrow&L^{u}&\\\ &&f^{\prime}_{u}&&\end{array}.$ Moreover $i^{u}(f^{\prime})$ and $i^{s}(f^{\prime})$ are equal to respectively $i^{u}$ and $i^{s}$ on the complement of compact subsets $V^{s}$ and $V^{u}$ independent of $f^{\prime}\in V_{f}$. For $\epsilon>0$ sufficiently small and then $V_{f}$ small enough, for every $x\in L\cap(V^{s}\cup V^{u})$ and $f^{\prime}\in V_{f}$, the intersection of $i^{s}(f^{\prime})(\mathcal{L}_{x}^{s\epsilon})$ with $i^{u}(f^{\prime})(\mathcal{L}_{x}^{u\epsilon})$ is transverse. $\mathrm{Let}\quad\mathcal{L}_{x}^{f^{\prime}\epsilon}=i^{s}(f^{\prime})(\mathcal{L}_{x}^{s\epsilon})\;{\overline{\pitchfork}}\;i^{u}(f^{\prime})(\mathcal{L}_{x}^{u\epsilon}).$ We want that $i(f^{\prime})$ sends $x\in L\cap(V^{s}\cup V^{u})$ into this intersection. In order to obtain the smoothness of $i(f^{\prime})$, we use a tubular neighborhood $(F,\mathcal{F},I,\pi)$ of the immersed lamination $(L,\mathcal{L})$. Thus, for $\epsilon$ small and then $V_{f}$ small enough, for every $y\in\mathcal{L}_{x}^{\epsilon}$, the submanifold $F_{y}^{\epsilon}:=I\big{(}B_{F_{x}}(0,\epsilon)\big{)}$ intersects transversally at a unique point $i(f^{\prime})(x)$ the submanifold $\mathcal{L}_{x}^{f^{\prime}\epsilon}$. In other words, $i(f^{\prime})_{\|\mathcal{L}_{x}^{\epsilon/2}}$ is the composition of $i_{\|\mathcal{L}_{x}^{\epsilon/2}}$ with the holonomy along the $C^{r}$-foliation formed by the leaves $(F_{y}^{\epsilon})_{y\in\mathcal{L}_{x}^{\epsilon}}$, from $i(\mathcal{L}_{x}^{\epsilon})$ to the transverse section $\mathcal{L}_{x}^{f^{\prime}\epsilon}$. Thus, the map $i(f^{\prime})$ is of class $C^{r}$ along the $\mathcal{L}$-plaques. As these manifolds depend $C^{r}$-continuously on $x\in L$, the map $i(f^{\prime})_{\|L\cap(V^{s}\cup V^{u})}$ is a $\mathcal{L}$-morphism into $M$. For $x\in L\setminus(V^{s}\cap V^{u})$ we put $i(f^{\prime})(x)=i(x)$. We notice that $i(f^{\prime})$ is $C^{r}$-close to $i$ when $f^{\prime}$ is $C^{r}$-close to $f$, since $i^{s}(f^{\prime})$ and $i^{u}(f^{\prime})$ are $C^{r}$-close to respectively $i^{s}$ and $i^{u}$. As $i(f^{\prime})$ and $i$ restricted to $L\setminus(V^{s}\cup V^{u})$ are equal, by precompactness of $L\cap(V^{s}\cup V^{u})$, for $V_{f}$ sufficiently small, the morphism $i(f^{\prime})$ is an immersion, for every $f^{\prime}\in V_{f}$. Let us construct $f^{\prime}$ such that the diagram of Theorem 3.1 commutes. For all $\delta>0$ and then $V_{f}$ small enough, for every $f^{\prime}\in V_{f}$, $i(f^{\prime})$ sends every $x\in L^{\prime}$ into the intersection of $i^{s}(f^{\prime})(\mathcal{L}_{x}^{s\delta})$ with $i^{u}(f^{\prime})(\mathcal{L}_{x}^{u\delta})$. Thus, for $V_{f}$ sufficiently small, by Theorems 3.1 and 3.4, every $f^{\prime}\in V_{f}$ sends the point $i(f^{\prime})(x)$ into the intersection $\mathcal{L}^{f^{\prime}\epsilon}_{f^{}(x)}$ of $i^{s}(f^{\prime})(\mathcal{L}_{f^{}(x)}^{s\epsilon})$ with $i^{u}(f^{\prime})(\mathcal{L}_{f^{}(x)}^{u\epsilon})$. Let $f^{\prime}(x)$ be the point of $\mathcal{L}_{f^{}(x)}^{\epsilon}$ such that $f^{\prime}\circ i(f^{\prime})(x)$ belongs to $F_{f^{\prime}(x)}^{\epsilon}$. We notice that the diagram of Theorem 3.1 commutes well. To see the regularity of $f^{\prime}\mapsto f^{\prime}$, we proceed as before: $f^{\prime}$ is locally the composition of $i(f^{\prime})$ with $f^{\prime}$ with the holonomy along a $C^{r}$-foliation between two transverse sections, all of them depending continuously on the point. Thus, $f^{\prime}$ is well a $C^{r}$-endomorphism of $\mathcal{L}$, equivalent to $f^{}_{\|L^{\prime}}$. As $i(f^{\prime})$ is close to $i$, the morphism $f^{\prime}$ is close to $f^{}_{\|L^{\prime}}$. ∎ ### 9.3 Proof of persistence of normally hyperbolic complex laminations by deformations (Theorem 0.6) Let $(L^{u},\mathcal{L}^{u})$ be the lamination $C^{1}$-immersed by $i^{u}$ provided by Proposition 9.1, with $r=1$ and $L=L^{\prime}$. By Lemma 5.2, the complex manifold $M$ induces on $\mathcal{L}^{\prime u}:=\mathcal{L}^{u}_{L^{\prime u}}$ a complex analytic structure such that $i^{u}$ is complex analytic, as soon as $Ti^{u}(T\mathcal{L}^{\prime u})$ is $J$-invariant, with $J$ the automorphism of $TM$ provided by its complex structure. To show the $J$-invariance of $Ti(T\mathcal{L}^{\prime u})$, we endow $i^{u}TM\rightarrow L^{u}$ with a cone field $\chi$ over the zero section of $i^{u}(TM)$ such that: * - $\chi$ contains $Ti^{u}(T_{x}\mathcal{L}^{\prime u})$ canonically embedded into $i^{u}TM_{x}$, for all $x\in L^{\prime u}$, - $Ti^{u}(T_{x}\mathcal{L}^{\prime u})$ is a maximal vector subspace of $i^{u}TM_{x}$ included in $\chi$, for all $x\in L^{\prime u}$, - the angle of $\chi$ is small. Let $x\in L^{\prime u}$. We want to prove that $Ti^{u}(T_{x}\mathcal{L}^{\prime u})$ is $J$-invariant. For this end, since $L$ is compact, we may suppose that $L^{\prime u}$ is preserved by $f^{-1}$ and so we notice that $Ti^{u}(T_{x}\mathcal{L}^{u})$ is equal to the intersection: $\bigcap_{n\geq 0}Tf^{n}(\chi_{f^{-n}(x)}).$ Such an intersection is equal to $\bigcap_{n\geq 0}Tf^{n}\big{(}Ti^{u}(T_{f^{-n}(x)}\mathcal{L}^{\prime u})\big{)}.$ As $Ti^{u}(T_{y}\mathcal{L}^{u})$ is $J$ ,invariant for $y\in L$, for $n\geq 0$ sufficiently large $J\circ Ti^{u}(T_{f^{-n}(x)}\mathcal{L}^{\prime u})$ is contained in $\chi$. Thus $\bigcap_{n\geq 0}Tf^{n}\big{(}J\circ Ti^{u}(T_{f^{-n}(x)}\mathcal{L}^{\prime u})\big{)}$ is included in $Ti^{u}(T_{x}\mathcal{L}^{\prime u})$. As $Tf$ commutes with $J$, we have: $Ti^{u}(T_{x}\mathcal{L}^{\prime u})\supset\cap_{n\geq 0}Tf^{n}\big{(}J\circ Ti^{u}(T_{f^{-n}(x)}\mathcal{L}^{u})\big{)}=J\circ Ti^{u}(T_{x}\mathcal{L}^{\prime u}).$ Thus, $Ti^{u}(T_{x}\mathcal{L}^{u})$ is $J$-invariant. To endow $(L_{s},\mathcal{L}_{s})$ with a complex analytic structure such that $i^{s}$ is holomorphic, we proceed similarly (actually it is even simpler). We finally remark that the inclusions: $i^{s}\;:\;(L,\mathcal{L})\hookrightarrow(L_{s},\mathcal{L}_{s})\quad\mathrm{and}\quad i^{u}\;:\;(L,\mathcal{L})\hookrightarrow(L^{u},\mathcal{L}^{u})$ are complex analytic by uniqueness of the complex structure in Lemma 5.2. ∎ ###### Lemma 9.2. Let $r\geq 1$. Let $(L,\mathcal{L})$ be a $d$-dimensional lamination $C^{r}$-immersed by $i$ into a Riemannian manifold $(M,g)$. Let $N$ be a $C^{r}$-lifting of $i$ into the Grassmannian of $k$-planes of $TM$. If for any $x\in L$, $N(x)$ is in direct sum with $Ti(T_{x}\mathcal{L})$ then there exists a structure of lamination $\mathcal{F}$ on the vector bundle $\pi\;:\;F\rightarrow L$ whose fiber at $x\in L$ is $N(x)$, such that: * - $\pi$ is a $C^{r}$-submersion, * - $I\;:\;(x,u)\in F\mapsto\text{exp}_{x}(u)$ is a $C^{r}$-immersion on a neighborhood of the $0$-section of $F$, * - $(F,\mathcal{F})$ is of dimension $d+k$. ###### Proof. For $x\in L$, let $\phi\in\mathcal{L}$ be a chart of a neighborhood $U$ of $x$. We may suppose that $\phi$ can be written in the form: $\phi\;:\;U\rightarrow\mathbb{R}^{d}\times T,$ where $d$ is the dimension of the leaves of $\mathcal{L}$ and $T$ is a locally compact metric space. We may also suppose $U$ small enough to be sent by $i$ into a distinguish open subset of $M$ that we can identify to $\mathbb{R}^{n}$. Therefore, the restriction of the tangent space of $M$ to this distinguish open is identified to $\mathbb{R}^{n}\times\mathbb{R}^{n}$. Finally, in this identification, we suppose $F_{x}$ identified to $\\{0\\}\times\mathbb{R}^{k}\times\\{0\\}$. Let $p$ be the canonical projection of $\mathbb{R}^{n}\times\mathbb{R}^{n}$ onto $\\{0\\}\times(\mathbb{R}^{k}\times\\{0\\})$. For $U$ small enough, $p$ is a linear bijection from $F_{y}$ onto $\mathbb{R}^{k}$, for every $y\in U$. We can now define the chart: $\phi_{U,x}\;:\;\pi^{-1}(U)\rightarrow\mathbb{R}^{d}\times\mathbb{R}^{k}\times T$ $(y,v)\mapsto(\phi_{1}(y),p(v),\phi_{2}(y))$ where $\phi_{1}$ and $\phi_{2}$ are the first and the second coordinates of $\phi$. As we already saw, we can identify the restriction of the Grassmannian of $TM$ at a neighborhood of $F_{x}$ to the product of $\mathbb{R}^{n}$ with the space $L(\mathbb{R}^{k},\mathbb{R}^{n-k})$ of linear maps from $\mathbb{R}^{k}$ to $\mathbb{R}^{n-k}$. In such identifications, for $U$ sufficiently small, the restriction of $F$ to $U$ is a $C^{r}$-morphism from $U$ into $\mathbb{R}^{n}\times L(\mathbb{R}^{k},\mathbb{R}^{n-k})$, which is of the form $F(y)=(i(y),l_{y})$. For another such a chart $\phi_{U^{\prime},x^{\prime}}=(\phi_{1}^{\prime},p^{\prime},\phi_{2}^{\prime})$, the changing coordinate: $\phi_{U^{\prime},x^{\prime}}\circ\phi_{U,x}^{-1}\;:\;\mathbb{R}^{d}\times\mathbb{R}^{k}\times T\rightarrow\mathbb{R}^{d}\times\mathbb{R}^{k}\times T^{\prime}$ $(u,w,t)\mapsto(\phi_{1}^{\prime}(y),p^{\prime}\circ l_{y}(w),\phi_{2}^{\prime}(y))$ with $y:=\phi^{-1}(u,t)$, satisfies the requested properties in order that $(\phi_{U,x})_{x\in L}$ defines a $C^{r}$-lamination structure $\mathcal{F}$ on $F$. ∎ ## References * [Ber] Pierre Berger. Persistence of stratification of normally expanded laminations. arXiv:math.DS. * [HOV94] John H. Hubbard and Ralph W. Oberste-Vorth. Hénon mappings in the complex domain. I. The global topology of dynamical space. Inst. Hautes Études Sci. Publ. Math., (79):5–46, 1994. * [HPS77] M. W. Hirsch, C. C. Pugh, and M. Shub. Invariant manifolds. Springer-Verlag, Berlin, 1977. Lecture Notes in Mathematics, Vol. 583. * [Mañ78] Ricardo Mañé. Persistent manifolds are normally hyperbolic. Trans. Amer. Math. Soc., 246:261–283, 1978. * [Mic80] Peter W. Michor. Manifolds of differentiable mappings, volume 3 of Shiva Mathematics Series. Shiva Publishing Ltd., Nantwich, 1980. * [MK06] James Morrow and Kunihiko Kodaira. Complex manifolds. AMS Chelsea Publishing, Providence, RI, 2006. Reprint of the 1971 edition with errata. * [Prz76] Feliks Przytycki. Anosov endomorphisms. Studia Math., 58(3):249–285, 1976. * [RHRHU07] Federico Rodriguez Hertz, Maria Alejandra Rodriguez Hertz, and Raul Ures. A survey of partially hyperbolic dynamics. In Partially hyperbolic dynamics, laminations, and Teichmüller flow, volume 51 of Fields Inst. Commun., pages 35–87. Amer. Math. Soc., Providence, RI, 2007. * [Rob76] Clark Robinson. Structural stability of $C^{1}$ diffeomorphisms. J. Differential Equations, 22(1):28–73, 1976. * [Shu69] Michael Shub. Endomorphisms of compact differentiable manifolds. Amer. J. Math., 91:175–199, 1969. * [Siu77] Yum Tong Siu. Every Stein subvariety admits a Stein neighborhood. Invent. Math., 38(1), 1976/77.	arxiv-papers	2008-07-30T22:19:29	2024-09-04T02:48:57.055507	{ "license": "Public Domain", "authors": "Pierre Berger", "submitter": "Pierre Berger", "url": "https://arxiv.org/abs/0807.4952" }
0807.5000	# Signals for strange quark contributions to the neutrino (antineutrino) scattering in quasi-elastic region Myung-Ki Cheoun1), K. S. Kim2) 111Corresponding author : kyungsik@hau.ac.kr 1)Department of Physics, Soongsil University, Seoul, 156-743, Korea 2)School of Liberal Arts and Science, Korea Aerospace University, Koyang 412-791, Korea ###### Abstract Strange quark contributions to the neutrino (antineutrino) scattering are investigated on the elastic neutrino-nucleon scattering and the neutrino- nucleus scattering for 12C target in the quasi-elastic region on the incident energy of 500 MeV, within the framework of a relativistic single particle model. For the neutrino-nucleus scattering, the effects of final state interaction for the knocked-out nucleon are included by a relativistic optical potential. In the cross sections we found some cancellations of the strange quark contributions between the knocked-out protons and neutrons. Consequently, the asymmetries between the incident neutrino and antineutrino which is the ratio of neutral current to charged current, and the difference between the asymmetries are shown to be able to yield more feasible quantities for the strangeness effects. In order to explicitly display importance of the cancellations, results of the exclusive reaction ${}^{16}O(\nu,\nu^{{}^{\prime}}p)$ are additionally presented for detecting the strangeness effects. ###### pacs: 25.30. Pt; 13.15.+g; 24.10.Jv Since the exploration of a spin structure of proton at the EMC measurement emc89 , the deep-inelastic scattering of leptons from nucleons has played important roles for studying the distribution of quarks and gluons. In the quark parton model for the nucleon, the cross section of the scattering with polarized leptons on polarized nucleons is usually given by four structure functions, $F_{1}(x),F_{2}(x),g_{1}(x)$, and $g_{2}(x)$ $F_{2}(x)=x{\mathop{\Sigma}_{q}}e_{q}^{2}q(x)=2xF_{1}(x)~{},~{}g_{1}(x)={1\over 2}{\mathop{\Sigma}_{q}}e_{q}^{2}\Delta q(x)$ (1) with $x=Q^{2}/2p\cdot q$ and $Q^{2}=-{(p^{{}^{\prime}}-p)}^{2}$ for the momentum $p$ of the initial nucleon and the momentum transfer $q$ from leptons. $e_{q},q(x)$, and $\Delta q(x)$ denote a charge, a parton distribution and a polarized parton distribution function for quarks flavor $q$, respectively Albe02 ; garvey . In experimental aspect, the structure functions are usually measured as their integral forms because of their $Q^{2}$ dependence $\Gamma_{1}(Q^{2})=\int_{0}^{1}g_{1}(x,Q^{2})dx~{}.$ (2) Available experimental data for the $\Gamma_{1}(Q^{2})$ are located in $0.115\sim 0.126$ depending on the region, $3\leq Q^{2}\leq 10.7$ Albe02 . If $\Delta q$ is defined as a difference of the total numbers of quarks and antiquarks in the nucleon with a helicity equal and opposite to the spin of the nucleon $\Delta q=\int_{0}^{1}{\mathop{\Sigma}_{r=\pm 1}}r[q^{(r)}(x)+{\bar{q}}^{(r)}(x)]dx~{},$ (3) $\Delta q$ stands for the contribution of $q$-quarks and ${\bar{q}}$-antiquarks to the spin of the nucleon. For example, in the infinite momentum frame (quark-parton model), $\Gamma_{1}(Q^{2})$ is evaluated in terms of the $\Delta q$ with some constraints among them $\Gamma_{1}^{p}={1\over 2}({4\over 9}\Delta u+{1\over 9}\Delta d+{1\over 9}\Delta s)~{},~{}\Delta u+\Delta d-2\Delta s=3F-D~{},~{}\Delta u-\Delta d=g_{A}$ (4) with $D$ and $F$ constants in SU(3) symmetry. On the other hand, the axial form factors $G_{A}^{q}(Q^{2})$ and the axial charges $g_{A}^{q}$ are given by the matrix elements and the diagonal matrix elements of the axial current $<p^{{}^{\prime}}\|{\bar{q}}\gamma_{\mu}\gamma_{5}q\|p>={\bar{u}}(p^{{}^{\prime}})\gamma_{\mu}\gamma_{5}G_{A}^{q}(Q^{2})u(p)~{},~{}<p\|{\bar{q}}\gamma_{\mu}\gamma_{5}q\|p>=2Ms_{\mu}g_{A}^{q}~{},$ (5) where $M$ and $s_{\mu}$ are respectively the mass and spin vector of the nucleon. In particular, for our primary interest, the axial coupling constant $g_{A}^{s}$ (or axial charge) induced by strangeness quark turned out to be equal to the $\Delta s$, which is related to the nucleon spin by the $s$ quark contribution. Though its value extracted from both experimental and/or theoretical results has not been fixed yet, several groups reported the value as $0.0$ Young06 , $-0.08\pm 0.06,-0.12\pm 0.06$ Pate05 and $-0.18\pm 0.05$ emc89 . Therefore, its value is allowed to vary in $0<g_{A}^{s}<-0.19$ for further discussions of its effects in the elastic neutrino (antineutrino) $(\nu{(\bar{\nu}}))$ scattering. In this paper, although the strangeness effects in vectorial parts can be studied by parity not only violating (or polarized electron scattering) but $\nu$ scattering, we focus on the strangeness $g_{A}^{s}$ effect in the axial part by studying the $\nu{(\bar{\nu}})$ scattering on the quasi-elastic (QE) region, where inelastic processes like pion production and delta resonance are excluded. Before going further, we briefly summarize the present status for signals of the strangeness effects in the $\nu{(\bar{\nu}})$ scattering. Since the vectorial parts are rarely contributed due to the given Weinberg angle in the reaction, the elastic cross section on the proton $\sigma(\nu p\rightarrow\nu p)$ mediated by a neutral current (NC) reaction is sensitive mainly on the $g_{A}^{s}$ value. But the measurement of the cross section is not experimentally easy, so that one usually resorts to the ratio of proton to neutron cross sections $R_{p/n}={{\sigma(\nu p\rightarrow\nu p)}/{\sigma(\nu n\rightarrow\nu n)}}$ garvey . The measurement of this ratio has also some difficulties in the neutron detection. Since the charged current (CC) cross section is relatively insensitive to the strangeness, the ratio $R_{NC/CC}={{\sigma(\nu p\rightarrow\nu p)}/{\sigma(\nu n\rightarrow\mu^{-}p)}}$ is suggested as a plausible signal for the nucleon strangeness although the signal goes down by a factor 2 Vent06 . In this paper, more precise analysis for those quantities are carried out in searching other feasible quantities for the strangeness effects in the $\nu{(\bar{\nu}})$-scattering. The elastic neutrino-nucleon $(\nu{(\bar{\nu}})-N)$ cross section by the NC reaction is expressed in terms of the Sachs (vector and axial) form factors Albe02 ; cheoun08 $\displaystyle{({{d\sigma}\over{dQ^{2}}})}_{\nu({\bar{\nu}})}^{NC}$ $\displaystyle=$ $\displaystyle{{G_{F}^{2}}\over{2\pi}}[{1\over 2}y^{2}{(G_{M}^{V})}^{2}+(1-y-{{M}\over{2E_{\nu}}}y){{{(G_{E}^{V})}^{2}+{E_{\nu}\over 2M}y{(G_{M}^{V})}^{2}}\over{1+{E_{\nu}\over 2M}y}}$ (6) $\displaystyle+({1\over 2}y^{2}+1-y+{{M}\over{2E_{\nu}}}y){(G_{A})}^{2}+h2y(1-{1\over 2}y)G_{M}^{V}G_{A}]~{}.$ Here, $E_{\nu}$ is the incident $\nu({\bar{\nu}})$ energy in the laboratory frame, and $y={{p\cdot q}/{p\cdot k}}={{Q^{2}}/{2p\cdot k}}$ with initial four momenta $k$ of $\nu({\bar{\nu}})$, target nucleon $p$, and four momentum transfer $q$ to the nucleon, respectively. $G_{F}\simeq 1.16639\times 10^{-11}$ MeV-2 is the Fermi constant. Vectorial Sachs form factors are written as electric and magnetic form factors of the nucleon $\displaystyle G_{M,E}^{V,~{}p(n)}(Q^{2})$ $\displaystyle=$ $\displaystyle({\frac{1}{2}}-2\sin^{2}\theta_{W})G_{M,E}^{p(n)}(Q^{2})-{\frac{1}{2}}G_{M,E}^{n(p)}(Q^{2})-{\frac{1}{2}}G_{M,E}^{s}(Q^{2})~{}~{}\mbox{for~{} NC}~{}$ $\displaystyle=$ $\displaystyle(G_{M,E}^{p}(Q^{2})-G_{M,E}^{n}(Q^{2}))~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}\mbox{for~{} CC}~{},$ $\displaystyle G_{M}^{s}(Q^{2})$ $\displaystyle=$ $\displaystyle{{Q^{2}F_{1}^{s}+\mu_{s}}\over{(1+\tau){(1+Q^{2}/M_{V}^{2})}^{2}}},~{}G_{E}^{s}(Q^{2})={{Q^{2}F_{1}^{s}-\mu_{s}\tau}\over{(1+\tau){(1+Q^{2}/M_{V}^{2})}^{2}}}~{},$ (7) where $\mu_{s}=G_{M}^{s}(0)$ is a strange magnetic moment. The axial form factor is usually presumed as a dipole form $\displaystyle G^{NC}_{A}(Q^{2})$ $\displaystyle=$ $\displaystyle{\frac{1}{2}}(\mp g_{A}+g_{A}^{s})/(1+Q^{2}/M_{A}^{2})^{2}~{}~{}\mbox{for~{} NC}~{}$ $\displaystyle G_{A}^{CC}(Q^{2})$ $\displaystyle=$ $\displaystyle- g_{A}/{(1+Q^{2}/M_{A}^{2})}^{2}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}\mbox{for~{} CC}~{},$ (8) where $g_{A}=1.262$ and $M_{A}$ are the axial coupling constant and the axial cut off mass, respectively. $-(+)$ coming from the isospin dependence denotes the knocked-out proton (neutron), respectively. The $g_{A}^{s}$ appeared explicitly in Eq.(8) represents the strange quark contents in the nucleon. Of course, the strangeness effects could contribute to other form factors, but their effects turned out to contribute only a few % to those physical quantities cheoun08 . In order to calculate the $\nu({\bar{\nu}})$-nucleus $(\nu-A)$ scattering, we choose the nucleus fixed frame where the target nucleus is seated at the origin of the coordinate system. The four-momenta of the incident and outgoing neutrinos (antineutrinos), the target nucleus, the residual nucleus, and the knocked-out nucleon are labelled $p_{i}^{\mu}$ and $p_{f}^{\mu}$, $p_{A}^{\mu}$, $p_{A-1}^{\mu}$, and $p^{\mu}$, respectively. In the laboratory frame, the inclusive cross section, which does not detect the outgoing $\nu$ (${\bar{\nu}}$), is given by the contraction between lepton and hadron tensor kimplb ; umino ; udias ; giusti1 $\displaystyle{\frac{d\sigma}{dT_{p}}}$ $\displaystyle=$ $\displaystyle 4\pi^{2}{\frac{M_{N}M_{A-1}}{(2\pi)^{3}M_{A}}}\int\sin\theta_{l}d\theta_{l}\int\sin\theta_{p}d\theta_{p}pf^{-1}_{rec}\sigma^{Z,W^{\pm}}_{M}[v_{L}R_{L}+v_{T}R_{T}+hv^{\prime}_{T}R^{\prime}_{T}]~{},$ (9) where $\theta_{l}$ denotes the scattering angle of the lepton. $\sigma^{Z,W^{\pm}}_{M}$ for NC and CC is defined by $\sigma^{Z}_{M}={\left({\frac{G_{F}\cos(\theta_{l}/2)E_{f}M_{Z}^{2}}{{\sqrt{2}}\pi(Q^{2}+M^{2}_{Z})}}\right)}^{2}~{}~{},~{}\sigma^{W^{\pm}}_{M}=\sqrt{1-{\frac{M^{2}_{l}}{E_{f}}}}\left({\frac{G_{F}\cos(\theta_{C})E_{f}M_{W}^{2}}{2\pi(Q^{2}+M^{2}_{W})}}\right)^{2}~{}~{},$ (10) where $M_{Z}$ and $M_{W}$ are the rest mass of $Z$-boson and $W$-boson, respectively. $\theta_{C}$ denotes the Cabibbo angle given by $\cos^{2}{\theta_{C}}\simeq 0.9749$. The recoil factor $f_{rec}$ is given as $f_{rec}={\frac{E_{A-1}}{M_{A}}}\left\|1+{\frac{E_{p}}{E_{A-1}}}\left[1-{\frac{{\bf q}\cdot{\bf p}}{p^{2}}}\right]\right\|.$ (11) Nuclear response functions to the weak current are denoted as $R_{L},R_{T}$ and $R_{T}^{{}^{\prime}}$, whose detailed forms are referred from Ref. kimplb . Detailed results of the cross sections on the proton and the neutron target for $g_{A}^{s}=-0.19$ and $0.0$ are shown in the upper parts of Fig. 1. Cross sections by the incident $\nu({\bar{\nu}})$ on the proton are usually enhanced in the whole $Q^{2}$ region by the $g_{A}^{s}$, 15% maximally, while they are reduced on the neutron. Other quantities related to these cross sections, such as asymmetries and ratios between $\nu$ and ${\bar{\nu}}$ in the NC reaction, also show similar effects cheoun08 . Therefore the strangeness effect is not large enough to be discernible in the cross sections, asymmetries, and ratios if we recollect the persisting experimental error bars in the BNL data garvey . The lowest part of Fig. 1 exhibits the strange quark contributions to the cross section on 12C$(\nu({\bar{\nu}}),\nu^{{}^{\prime}}({\bar{\nu}^{{}^{\prime}}}))$ reaction. For directly comparing with the nucleon case, we present them in terms of outgoing nucleon kinetic energy $T_{N}$ because $Q^{2}=2MT_{P}$ would hold on the free nucleon inside nuclei. Thick and thin curves are the results for $g^{s}_{A}=-0.19$ and $0.0$, respectively. Note that a relativistic optical potential for the final nucleon is taken into account for the FSI clark . Detailed discussions about the FSI are done at Ref. kim07 . Peak positions for 12C nuclei, which does not show up in the nucleon case, just come from the binding energy of nucleons inside nuclei. To analyze cross sections (solid curves) for the $\nu-^{12}$C scattering, we present separately each contribution via final neutrons (dotted curves) and final protons (dashed curves) in the figure. The effect of $g^{s}_{A}$ for the proton is increased by about 30%, but for the neutron it is decreased by 30%, maximally, independently of the incident energy. These individual $g_{A}^{s}$ effects on each nucleon resemble exactly those of each nucleon in the upper parts of Fig. 1, that is, the $g_{A}^{s}$ effect enhances the cross section by protons and decreases that by neutrons. However, total net $g_{A}^{s}$ effects severely depend on the competition between the protons and neutrons processes because of the summation of all nucleons. In the case of 12C, the enhancement by the proton is nearly compensated by the neutron process, so that the net effect increases the cross section only below 3% for the given $E_{\nu}$. For the ${\bar{\nu}}-^{12}$C process, the $g_{A}^{s}$ effect enhances the cross section by about 20% on the protons, but reduces them by 20% the neutrons. The reduction by neutrons is nearly balanced by the enhancement due to protons. Consequently, the net effect of the strange quark reduces also the cross section only below 2% for the given $E_{\bar{\nu}}$, similarly to the $\nu$ case. From these results, in the $\nu-A$ cross sections, the $g^{s}_{A}$ effect contributes more positively to the proton while it does negatively to the neutron, and is cancelled out finally. In the case of 12C, the resultant effects are only within a few % by the cancellation between the final protons and neutrons. Therefore, the $g_{A}^{s}$ effect in nuclei is too small to be deduced from the A$(\nu,\nu^{{}^{\prime}})$ cross section. On the other hand, the cross section for the CC scattering is given with the following replacement into the NC cross sections of Eqs. (6) and (9) ${({{d\sigma}\over{dQ^{2}}})}_{\nu({\bar{\nu}})}^{CC}={({{d\sigma}\over{dQ^{2}}})}_{\nu({\bar{\nu}})}^{NC}(G_{E}^{V}\rightarrow G_{E}^{CC},G_{M}^{V}\rightarrow G_{M}^{CC},G_{A}\rightarrow G_{A}^{CC})~{},$ (12) with $G_{E}^{CC}=G_{E}^{p}(Q^{2})-G_{E}^{n}(Q^{2})~{},~{}G_{M}^{CC}=G_{M}^{p}(Q^{2})-G_{M}^{n}(Q^{2})$. Since the relevant form factors are expressed only by the electro-magnetic form factors the CC scattering is not nearly influenced by the strangeness in the form factors. Therefore, the ratios of the NC and CC reactions given by $R_{NC/CC}={\frac{\sigma(\nu,\nu^{\prime}p)}{\sigma(\nu,\mu^{-}p)}}={\frac{\sigma_{NC}^{\nu p}}{\sigma_{CC}^{\nu}}}~{},{\bar{R}_{NC/CC}}={\frac{\sigma({\bar{\nu}},{\bar{\nu}}^{\prime}n)}{\sigma({\bar{\nu}},\mu^{+}n)}}={\frac{\sigma_{NC}^{{\bar{\nu}}n}}{\sigma_{CC}^{{\bar{\nu}}}}}~{},$ (13) have been suggested for probing the strangeness on the nucleon or nuclei. Moreover any possible nuclear structure effects in these reactions are expected to be cancelled out. Since these ratios are focused on the knocked- out nucleon, we introduce another definition of the ratios by focusing on the nucleon inside the target nucleus as follows $R^{\prime}_{NC/CC}={\frac{\sigma(\nu,\nu^{\prime}n)}{\sigma(\nu,\mu^{-}p)}}={\frac{\sigma_{NC}^{\nu n}}{\sigma_{CC}^{\nu}}}~{},~{}{\bar{R}}^{\prime}_{NC/CC}={\frac{\sigma({\bar{\nu}},{\bar{\nu}}^{\prime}p)}{\sigma({\bar{\nu}},\mu^{+}n)}}={\frac{\sigma_{NC}^{{\bar{\nu}}p}}{\sigma_{CC}^{{\bar{\nu}}}}}.$ (14) Since the charge exchange is not included, these $R^{{}^{\prime}}$ and ${\bar{R}}^{{}^{\prime}}$ quantities have the same meaning as $R_{NC/CC}={\sigma(\nu n\rightarrow\nu n)}/{\sigma(\nu n\rightarrow\mu^{-}p)},{\bar{R}}_{NC/CC}={\sigma({\bar{\nu}}p\rightarrow{\bar{\nu}}p)}/{\sigma({\bar{\nu}}p\rightarrow{\mu^{+}}n)}$ on nucleon level, namely, $R^{{}^{\prime}}$ and ${\bar{R}}^{{}^{\prime}}$ mean the ratios on the nucleon inside nuclei bombarded by incident $\nu({\bar{\nu}})$. Results and relevant discussions for the ratios on the 12C target are given at Fig. 2. The strangeness effects are not so large than those of the cross sections at Fig. 1. Divergence of high $T_{P}$ is associated with the effect of outgoing lepton mass in the CC reaction Albe02 . In order to find more feasible $g_{A}^{s}$ signals, the difference of the cross sections between incident $\nu$ and ${\bar{\nu}}$ is suggested as a candidate for the effectsAlbe02 ; kimplb . Since $h=-1$ $(h=+1)$ in Eqs.(6) and (9) corresponds to the helicity of the incident $\nu$ (${\bar{\nu}}$), a difference of the cross sections is summarized as a simple form. For instance, for the $\nu-N$ scattering, the difference of the cross section is given as ${({{d\sigma}\over{dQ^{2}}})}_{\nu}^{NC}-{({{d\sigma}\over{dQ^{2}}})}_{{\bar{\nu}}}^{NC}=-{{G_{F}^{2}}\over{2\pi}}~{}4y(1-{1\over 2}y)G_{M}^{V}G_{A}~{}.$ (15) Similar conjectures can be done for the $\nu-A$ scattering from Eq. (9). Since the difference of the cross sections between $\nu$ and ${\bar{\nu}}$ can be expressed only by a few form factors, more efficient observable are possible if we consider asymmetries between $\nu$ and ${\bar{\nu}}$ relative to the CC reactions. Moreover it enables us to distinguish each strangeness contribution in the two strangeness effects, i.e. vectorial and axial strangeness parts at small $Q^{2}$ region $\displaystyle A_{NC/CC}^{p}$ $\displaystyle=$ $\displaystyle{{\sigma_{NC}^{\nu p}-\sigma_{NC}^{{\bar{\nu}}p}}\over{(\sigma_{CC}^{\nu n}-\sigma_{CC}^{{\bar{\nu}}p})}}=0.12-0.12{{g_{A}^{s}}\over{g_{A}}}-0.13{{G_{M}^{s}}\over{G_{M}^{3}}}~{},$ (16) $\displaystyle A_{NC/CC}^{n}$ $\displaystyle=$ $\displaystyle{{\sigma_{NC}^{\nu n}-\sigma_{NC}^{{\bar{\nu}}n}}\over{(\sigma_{CC}^{\nu n}-\sigma_{CC}^{{\bar{\nu}}p})}}=0.16+0.16{{g_{A}^{s}}\over{g_{A}}}+0.13{{G_{M}^{s}}\over{G_{M}^{3}}}~{},$ where the third terms with $G_{M}^{s}(Q^{2}=0)=\mu_{s}$, $G_{M}^{3}={(G_{M}^{p}-G_{M}^{n}})/2$, $\mu_{p}=2.79$, and $\mu_{n}=-1.91$ come from the vectorial part. Since $\|G_{M}^{s}/G_{M}^{3}\|$ and $\|g_{A}^{s}/g_{A}\|$ are approximately 0.2, the strangeness effects from the vector and axial parts are comparable, in principle. It could be questioned if Eq.(16) still holds even in the case of a nucleus. But, we assume that it holds because the outgoing nucleon can be described as quasi-freely bombarded by the incident $\nu({\bar{\nu}})$. Here two interesting remarks are possible. One is that $A_{NC/CC}^{p(n)}$ could be a constant if there would be no strangeness effects. The second point is related to the $\mu_{s}$ sign. If $g_{A}^{s}$ and $G_{M}^{s}$ have different $(\pm)$ sign, the values of $A_{NC/CC}^{p}$ and $A_{NC/CC}^{n}$ become constants 0.12 and 0.16, respectively, because the last two terms in Eq. (16) are nearly cancelled out. Any deviations from these constants would imply that both signs are same. For instance, we presented the case of $g_{A}^{s}=-0.19$ and $\mu_{s}=-0.4$ showing such a tendency at the upper panels in Fig. 3. Both nucleon and nuclei cases show nearly the same trend. It means that $g_{A}^{s}$ effect is nearly evadable from any observable suggested until now. Divergence of high $T_{P}$ in the nuclear case of Fig. 3 also comes from the CC reaction in the denominator. As another way to study the strangeness, we introduce the difference and the sum of asymmetries $\displaystyle DA_{NC/CC}$ $\displaystyle=$ $\displaystyle A_{NC/CC}^{p}-A_{NC/CC}^{n}\simeq-0.04-0.28({{g_{A}^{s}}\over{g_{A}}}+{{G_{M}^{s}}\over{G_{M}^{3}}})~{},$ (17) $\displaystyle SA_{NC/CC}$ $\displaystyle=$ $\displaystyle A_{NC/CC}^{p}+A_{NC/CC}^{n}=0.28+0.04{{g_{A}^{s}}\over{g_{A}}}~{}.$ The sum of asymmetry in Eq. (17) is given only in terms of the axial part. But the second term is very small by the factor 0.04, so that the SA is nearly independent of the strangeness, but the DA depends strongly on the axial strangeness. Figure 4 represents the calculation for the difference and the summation of the asymmetries. On the summation, the $g_{A}^{s}$ effect is negligible, so that the summation of the asymmetry behaves a constant about 0.28 at low and middle kinetic energies, independently of free or bound nucleons cheoun08 . On the difference, the $g_{A}^{s}$ effects clearly appear to be larger than any other observable. If the values of $g_{A}^{s}$ and $\mu_{s}$ have different signs, the DA would be constant, but it depends on $Q^{2}$ if they have same signs as in Fig. 4. Finally, we consider the exclusive $\nu-A$ scattering i.e. $A(\nu,\nu^{{}^{\prime}}N)$ because we expect that the strangeness effects via each nucleon process can be survived in the exclusive reaction without mutual cancellations. Our terminology for the exclusive reaction is based on the concept in the electron scattering kimepja . In the laboratory frame, the exclusive differential cross section is given by the contraction between the lepton tensor and the hadron tensor udias ; kimepja $\displaystyle{\frac{d^{5}\sigma}{dE_{f}d\Omega_{f}d\Omega_{p}}}$ $\displaystyle=$ $\displaystyle K[v_{L}R_{L}+v_{T}R_{T}+v_{TT}R_{TT}\cos 2\phi_{p}+v_{LT}R_{LT}\cos\phi_{p}$ (18) $\displaystyle+$ $\displaystyle h(v^{\prime}_{T}R^{\prime}_{T}+v^{\prime}_{LT}R^{\prime}_{LT}\sin\phi_{p})],$ where $\phi_{p}$ denotes the azimuthal angle of the knocked-out proton as measured with respect to the incident $\nu$ scattering plane ($\bf x-\bf z$ plane). $R_{L}$, $R_{T}$, $R_{TT}$, $R_{LT}$, $R^{\prime}_{T}$, and $R^{\prime}_{LT}$ are called the longitudinal, transverse, transverse- transverse, longitudinal-transverse, polarized transverse, and polarized longitudinal-transverse response functions, respectively. The kinematics factor $K$ is given by $K={\frac{M_{N}M_{A-1}p}{(2\pi)^{3}M_{A}}}f^{-1}_{rec}\sigma^{Z}_{M}$. Our results in Fig. 5, where incident $E_{\nu}=2.4$ GeV is adopted as a JLAB type’s electron energy, show clearly the strangeness effects, as expected. However, in the experimental side, this reaction may have actual difficulties in detecting incident and outgoing neutrinos. However, if one integrates the scattering angle and averages over the neutrino energies, the five folded cross section in Eq. (18) becomes a two folded form, that is, the cross section is going to be the angular distribution of the knocked-out nucleon by detecting only solid angles of the knocked-out nucleon. In conclusion, the $g_{A}^{s}$ effect in the elastic $\nu({\bar{\nu}})-N$ scattering for the NC reaction enhances the cross section for the proton, but reduces it for the neutron. But for the $\nu-A$ scattering in the QE region, the contributions of both protons and neutrons compensate each other in the cross section because of the summation over all nucleons, so that the net effects on the cross sections are nearly indiscernible due to the possible cancellations. However the asymmetries between incident $\nu$ and ${\bar{\nu}}$ divided by the CC reaction could be a prominent signal for the strangeness effect. In specific, differences of the asymmetries could be a feasible quantity for the signal. Finally the exclusive reaction like 12C$(\nu({\bar{\nu}}),\nu^{\prime}({\bar{\nu}^{{}^{\prime}}})N)$ could be more efficient tests for the effect rather than the inclusive reaction 12C$(\nu({\bar{\nu}}),\nu^{\prime}({\bar{\nu}^{{}^{\prime}}}))$, because there are no competitions of the $g_{A}^{s}$ effects for the knocked-out protons and neutrons in the exclusive reaction. Our results for the exclusive reaction justify clearly this conjecture. ## Acknowledgements This work was supported by the Korea Science and Engineering Foundation(KOSEF) grant funded by the Korea government(MOST) (R01-2007-000-20569) and one of author, Cheoun, was supported by the Soongsil University Research Fund. ## References * (1) J. Ashman et al., EMC Collaboration, Nucl. Phys. B 328, 1 (1989). * (2) W. M. Alberico, S. M. Bilenky, and C. Maieron, Phys. Rep. 358, 227, (2002). * (3) G. T. Garvey, S. Krewald, E. Kolbe, and K. Langanke, Phys. Rev. C 48, 1919 (1993); Phys. Lett. B 289, 249 (1992). * (4) R. D. Young, J. Roche, R. D. Carlini, and A. W Thomas, Phys. Rev. Lett., 97, 102002, (2006). * (5) Stephen F. Pate, Eur. Phys. J. A 24, s2, 67, (2005). * (6) B. I. S. van der Ventel, and J. Piekarewicz, Phys. Rev. C 69, 035501, (2004). * (7) Myung-Ki Cheoun and K. S. Kim, J. Phys. G 35, 065107, (2008). * (8) K. S. Kim, Myung Ki Cheoun, and B. G. Yu, to be appeared, Phys. Rev C, arXiv:0707.2767 (2008). * (9) Y. Umino, J. M. Udias, Phys. Rev. C 52, 3399 (1995); Y. Umino, J. M. Udias, and P. J. Mulders, Phys. Rev. Lett. 74, 4993 (1995). * (10) Andrea Meucci, Carlotta Giusti, and Franco Davide Pacati, Nucl. Phys. A739, 277 (2004); Nucl. Phys. A744, 307 (2004); Nucl. Phys. A773, 250(2006). * (11) M. C. Martinez, P. Lava, N. Jahowicz, J. Ryckebusch, K, Vantournhout, and J. M. Udias, Phys. Rev. C 73, 024607 (2006). * (12) E. D. Cooper, S. Hama, B. C. Clark, and R. L. Mercer, Phys. Rev. C 47, 297 (1993). * (13) K. S. Kim, B. G. Yu, M. K. Cheoun, T. K. Choi, and M. T. Cheon, J. Phys. G 34, 2643 (2007). * (14) J. M. Udias, P. Sarriguren, E. Moya de Guerra, E. Garrido, and J. A. Caballero, Phys. Rev. C 53, R1488 (1996). * (15) K. S. Kim, Myung Ki Cheoun, Yeungun Chung, and Hyung Joo Nam, Eur. Phys. J. A 11, 147 (2001). Figure 1: (Color online) Upper 4 figures are differential cross sections by the NC scattering on proton (left) and neutron (right), Eq. (6), in a unit ($G_{F}^{2}/9\simeq 6.05\times 10^{-39}~{}[{{cm}^{2}/{(GeV/c)}^{2}}]$), as a function of $Q^{2}$ for $E_{\nu({\bar{\nu}})}=500$ MeV. They are calculated for $g_{A}^{s}=-0.19$ and 0.0 cases, respectively. The lowest panel is for 12C($\nu(\bar{\nu}),\nu^{\prime}(\bar{\nu})$) cross section, Eq.(9), as a function of the knocked out nucleon kinetic energy $T_{p}$ at $E_{\nu({\bar{\nu}})}=500$ MeV, where left (right) panel is for the $\nu$ ($\bar{\nu}$) scattering, respectively. Solid curves are the results for the cross sections, dashed and dotted lines are the contributions of the proton and the neutron, respectively. Thick and thin lines are calculations with $g^{s}_{A}=-0.19$ and $g^{s}_{A}=0.0$. Figure 2: (Color online) Ratios of the NC to the CC cross sections of the $\nu-A$ scattering for 12C as a function of the knocked-out nucleon kinetic energy. For the NC reaction, solid (red) curves represent the results with $g_{A}^{s}=0.0$, dashed (black) lines are with $g_{A}^{s}=-0.19$. Figure 3: (Color online) Asymmetries $A_{NC/CC}^{p}$ and $A_{NC/CC}^{n}$, Eq.(16), between the NC and CC reactions for an incident $E_{\nu({\bar{\nu}})}$ = 500 MeV. They are calculated for $g_{A}^{s}=-0.19$ (dashed curve) and 0.0 cases (solid curve), respectively. Upper panel is for the nucleon, while lower panel is for ${}^{12}C$. Figure 4: (Color online) The differences and summations of the asymmetries between the NC and CC cross sections of ${\nu-N}$ (upper part) and ${\nu-A}$ scattering for 12C target (lower parts) as a function of the knocked-out nucleon kinetic energy. Solid (red) curves represent the results with $g_{A}^{s}=0.0$, dashed (black) lines are with $g_{A}^{s}=-0.19$. Figure 5: (Color online) Neutral current 16O($\nu,\nu^{\prime}p$) cross sections as a function of the missing momentum at $E_{\nu}=2.4$ GeV. Solid (red) curves are the results for $g_{A}^{s}=0.0$ and dashed (black) lines are for $g_{A}^{s}=-0.19$, respectively. Thick and thin lines are the results for the incident $\nu$ and ${\bar{\nu}}$. States of the outgoing nucleon, $p_{1/2}$ and $p_{3/2}$, are indicated, respectively.	arxiv-papers	2008-07-31T07:26:22	2024-09-04T02:48:57.068087	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Myung-Ki Cheoun and K. S. Kim", "submitter": "Kyungsik Kim", "url": "https://arxiv.org/abs/0807.5000" }
0807.5089	11institutetext: Astrophysikalisches Institut Potsdam, An der Sternwarte 16, D-14482 Potsdam, Germany 11email: adaglio@aip.de # An unbiased measurement of the UV background and its evolution via the proximity effect in quasar spectra ††thanks: Based on data collected at the European Southern Observatory and obtained from the ESO Science Archive. Aldo Dall’Aglio Lutz Wisotzki Gábor Worseck (Received … / Accepted …) We investigated a set of high-resolution ($R\sim 45\ 000$), high signal-to- noise ($S/N\sim 70$) quasar spectra to search for the signature of the so- called proximity effect in the H i Ly$\alpha$ forest. The sample consists of 40 bright quasars with redshifts in the range $2.1<z<4.7$. Using the flux transmission statistic, we determined the redshift evolution of the H i effective optical depth in the Lyman forest between $2\la z\la 4.5$, finding good agreement with previous measurements based on smaller samples. We also see the previously reported dip in $\tau_{\mathrm{eff}}(z)$ around redshift $z\sim 3.3$, but as the significance of that feature is only 2.6$\sigma$, we consider this detection tentative. Comparing the flux transmission near each quasar with what was expected from the overall trend of $\tau_{\mathrm{eff}}(z)$, we clearly detect the proximity effect not only in the combined quasar sample, but also towards each individual line of sight at high significance, albeit with varying strength. We quantify this strength using a simple prescription based on a fiducial value for the intensity of the metagalactic UV background (UVB) radiation field at 1 Ryd, multiplied by a free parameter that varies from QSO to QSO. The observed proximity effect strength distribution (PESD) is asymmetric, with an extended tail towards values corresponding to a weak effect. We demonstrate that this is not simply an effect of gravitational clustering around quasars, as the same asymmetry is already present in the PESD predicted for purely Poissonian variance in the absorption lines. We present the results of running the same analysis on simulated quasar spectra generated by a simple Monte-Carlo code. Comparing the simulated PESD with observations, we argue that the standard method of determining the UVB intensity $J_{\nu_{0}}$ by averaging over several lines of sight is heavily biased towards high values of $J_{\nu_{0}}$ because of the PESD asymmetry. Using instead the _mode_ of the PESD provides an estimate of $J_{\nu_{0}}$ that is unbiased with respect to his effect. For our sample we get a modal value for the UVB intensity of $\log J_{\nu_{0}}=-21.51\pm 0.15$ (in units of $\mathrm{erg}\,\mathrm{cm}^{-2}\,\mathrm{s}^{-1}\,\mathrm{Hz}^{-1}\,\mathrm{sr}^{-1}$) for a median quasar redshift of 2.73. With $J_{\nu_{0}}$ fixed we then corrected $\tau_{\mathrm{eff}}$ near each quasar for local ionisation and estimated the amount of excess H i absorption attributed to gravitational clustering. On scales of $\sim 3$ Mpc, only a small minority of quasars show substantial overdensities of up to a factor of a few in $\tau_{\mathrm{eff}}$; these are exactly the objects with the weakest proximity effect signatures. After removing those quasars residing in overdense regions, we redetermined the UVB intensity using a hybrid approach of sample averaging and statistical correction for the PESD asymmetry bias, arriving at $\log J_{\nu_{0}}=-21.46^{+0.14}_{-0.21}$. This is the most accurate measurement of $J_{\nu_{0}}$ to date. We present a new diagnostic based on the shape and width of the PESD that strongly supports our conclusion that there is no systematic overdensity bias for the proximity effect. This additional diagnostic breaks the otherwise unavoidable degeneracy of the proximity effect between UVB and overdensity. We then applied our hybrid approach to estimate the redshift evolution of the UVB intensity and found tentative evidence of a mild decrease in $\log J_{\nu_{0}}$ with increasing redshift, by a factor of $\sim 0.4$ from $z=2$ to $z=4$. Our results are in excellent agreement with earlier predictions for the evolving UVB intensity, and they also agree well with other methods of estimating the UVB intensity. In particular, our measured UVB evolution is much slower than the change in quasar space densities between $z=4$ and $z=2$, supporting the notion of a substantial contribution of star-forming galaxies to the UVB at high redshift. ###### Key Words.: diffuse radiation – intergalactic medium – quasars: absorptionlines Figure 1: A typical quasar spectrum from our sample (Q 0002$-$422 $z=2.768$) together with the power law continuum fit (dashed red line) and the local continuum fit (blue line). ## 1 Introduction High-resolution spectra of high-redshift, quasi-stellar objects (QSOs) show a plethora of narrow absorption lines from various species along their lines of sight through the intergalactic medium (IGM). Most absorption lines blueward of an Ly$\alpha$ emission line stem from the Ly$\alpha$ transition of intervening neutral hydrogen, giving rise to the Ly$\alpha$ forest (Sargent et al. 1980; Weymann et al. 1981; Rauch 1998). High-resolution spectra of the Ly$\alpha$ forest enabled a statistical analysis of the absorbers, as well as their physical properties (e.g. Kim et al. 2001; Schaye et al. 2003; Kim et al. 2004). The vast majority of H i Ly$\alpha$ absorbers are optically thin to ionising photons and are thus kept highly photoionised by the metagalactic UV background (UVB) generated by the overall population of quasars and star- forming galaxies (Haardt & Madau 1996; Fardal et al. 1998; Haardt & Madau 2001). The UVB intensity at a given frequency is determined by the evolving population of UV sources, modified by cosmological expansion and absorption (e.g. Davé et al. 1999). It is important to determine the amplitude of the UVB and its spectral shape as a function of redshift in order to constrain the relative contributions of quasars and star-forming galaxies to the UVB, as well as to provide a key ingredient to numerical simulations of structure formation. There are different ways to constrain the H i UVB photoionisation rate in the IGM. Assuming its spectral energy distribution it is possible to infer the UVB intensity at the Lyman limit $J_{\nu_{0}}$. First, the UVB can be predicted by integrating the contributions of the observed source population. Given the luminosity functions of the sources, a characteristic intrinsic spectral energy distribution and the observed absorber distribution functions in the IGM, the amplitude and spectral shape of the UVB can be calculated numerically as the radiation is filtered by the IGM (Bechtold et al. 1987). In particular, comparing these models to observations provides constraints on the relative contributions of quasars (Haardt & Madau 1996; Fardal et al. 1998) and star- forming galaxies (Haardt & Madau 2001). Furthermore, the H i photoionisation rate at $2<z<4$ has been constrained by matching the Ly$\alpha$ forest absorption and the IGM temperature evolution from numerical simulations of structure formation to observations (Rauch et al. 1997; Theuns et al. 1998; McDonald et al. 2000; Meiksin & White 2004; Tytler et al. 2004; Bolton et al. 2005; Kirkman et al. 2005; Jena et al. 2005). A more direct approach to estimate the UVB photoionisation rate is based on the so-called proximity effect. In the vicinity of a luminous quasar the UV radiation field is expected to be enhanced, resulting in a statistical deficit of Ly$\alpha$ absorbers near the quasar compared to far away from it along the line of sight (Weymann et al. 1981; Carswell et al. 1982; Murdoch et al. 1986). Knowing the UV luminosity of the quasar, the deficit of lines (or generally Ly$\alpha$ opacity) yields an estimate of the UVB intensity at the H i Lyman limit (Carswell et al. 1987; Bajtlik et al. 1988). Traditionally, the proximity effect has been considered as a statistical effect and large samples of up to $\sim 100$ quasars have been compiled to measure the mean UVB at $2\la z\la 4$ (e.g. Bajtlik et al. 1988; Lu et al. 1991; Giallongo et al. 1996; Cooke et al. 1997; Scott et al. 2000; Liske & Williger 2001). Nevertheless, the proximity effect can also be detected towards individual quasars (e.g. Williger et al. 1994; Cristiani et al. 1995; Lu et al. 1996; Savaglio et al. 1997; Dall’Aglio et al. 2008). Proximity effect analyses at different redshifts can constrain the redshift evolution of the UVB, showing that it broadly peaks at $z\sim 3$. The decrease in the UVB at $z<2$ (Kulkarni & Fall 1993; Scott et al. 2002) and $z\ga 4$ (Williger et al. 1994; Lu et al. 1996; Savaglio et al. 1997) is likely due to the declining space density of quasars (Haardt & Madau 1996; Fardal et al. 1998). Whereas the UVB evolution at low redshifts is sufficiently fast to be detectable within a single sample (Scott et al. 2002), there is presently little direct evidence of evolution within the redshift range $2\la z\la 4$. It is currently an open issue whether the measured UVB at $z\la 3$ is consistent with the integrated emission of quasars, or whether a significant share of ionising photons has to be provided by star-forming galaxies (e.g. Madau et al. 1999; Sokasian et al. 2003; Schirber & Bullock 2003; Hunt et al. 2004). One difficulty lies in the fact that the UVB estimates via the proximity effect are subject to several systematic uncertainties, all of which can lead to an overestimate of the inferred UV background: 1. 1. Quasar variability: The size of the proximity effect zone scales as the average quasar luminosity over the photoionisation equilibrium timescale in the IGM ($\sim 10^{4}$ yr). Quasar variability on shorter timescales will tend to overestimate the UVB due to Malmquist bias in the selected quasar samples (Schirber et al. 2004). It is not known how important this effect is in practice. 2. 2. Underestimated quasar redshifts: Quasar emission redshifts determined from broad high-ionisation lines are likely underestimated (e.g. Gaskell 1982; Richards et al. 2002), causing the UVB to be overestimated by a factor $\sim 3$ (Espey 1993). However, this effect can be largely avoided by using low- ionisation lines to determine systemic redshifts. 3. 3. Overdense environment: In the standard ionisation model of the proximity effect (Bajtlik et al. 1988), the Ly$\alpha$ forest line density is extrapolated into the quasar vicinity, assuming that the matter distribution is not altered by the presence of the QSO. If quasars typically reside in high intrinsic overdensities instead, the UVB will be overestimated by a factor of up to $\sim 3$ (Loeb & Eisenstein 1995). Conversely, the disagreement between the UVB estimates from the proximity effect and those obtained by matching simulations to the mean Ly$\alpha$ forest absorption has recently been used as an argument that quasars are surrounded by substantial overdensities (Rollinde et al. 2005; Guimarães et al. 2007; Faucher-Giguère et al. 2008b). In the present paper we pursue a new approach to employ the proximity effect as a tool to investigate the UVB, based on measuring its signature towards individual QSO sight lines. We demonstrated recently (Dall’Aglio et al. 2008, hereafter Paper I) that indeed essentially all quasars show this signature, even at relatively low spectral resolution. Here we follow up on this finding, now based exclusively on high resolution, high S/N quasar spectra. The plan of the paper is as follows. We begin with a brief description of our spectroscopic data (Sect. 2). In Sect. 3 we explain the Monte Carlo simulations used to assess uncertainties and to interpret the observed proximity effect strength distribution. We then determine the redshift evolution of the effective optical depth in the Ly$\alpha$ forest in Sect. 4, followed by a comparison between the two most commonly adopted methods of revealing the proximity effect in Sect. 5. We very briefly report on the analysis of the combined QSO sample in Sect. 6. In Sect. 7 we investigate the proximity effect in individual sightlines, showing that the traditional sample-combining method is biased. We then estimate the excess Ly$\alpha$ absorption near quasars (Sect. 8), and use this to constrain the redshift evolution of the UVB (Sect. 9). We present our conclusions in Sect. 10. Throughout this paper we assume a flat universe with Hubble constant $H_{0}=70$ $\mathrm{km}\,\mathrm{s}^{-1}\,\mathrm{Mpc}^{-1}$ and density parameters $\left(\Omega_{\mathrm{m}},\Omega_{\Lambda}\right)=\left(0.3,0.7\right)$. All distances are expressed in physical units. ## 2 Data ### 2.1 Sample description and data reduction The sample consists of 40 quasar sight lines observed with the UV-Visual Echelle Spectrograph (UVES, Dekker et al. 2000) at VLT/UT2 on Cerro Paranal, Chile. All 40 spectra were taken from the ESO archive and are publicly available. We selected the QSO sample based on the following criteria: (i) a minimum of 10 exposures and (ii) a complete or nearly complete coverage of the Ly$\alpha$ forest region especially close to the quasar. Table 1 lists the complete sample of quasars with adopted redshifts (Sect. 2.4) and quasar luminosities at the Lyman limit (Sect. 2.3). The data were reduced within the ECHELLE/UVES environment (version 2.2) of the software package MIDAS and following the procedures described in Kim et al. (2004). After the bias and inter-order background subtraction from each science frame, an optimal extraction of the spectrum was performed, order by order, assuming a Gaussian distribution along the spatial direction. The sky was estimated and subtracted together with the Gaussian fit, maximising the quality of the extracted spectrum (e.g. Kim et al. 2001). Cosmic rays were identified and removed using a median filter. The wavelength calibration was performed with the available ThAr lamp frames. Relative flux calibration was performed using the master-response functions publicly available on the ESO website and accounting for different airmasses. This allowed us to place our spectra onto a relative flux scale reasonably well, while the absolute scale had to be tied to external photometry (Sect. 2.2). In each final coadded spectrum the individual extractions were weighed corresponding to their S/N and re-sampled on 0.05 Å bins. The resolving power is $\sim 45\,000$ in the regions of interest, corresponding to a velocity resolution of $\sim 6$ $\mathrm{km}\,\mathrm{s}^{-1}$ (FWHM). The wavelengths in the final spectra are vacuum heliocentric, and the fluxes are corrected for galactic reddening. An example of a final spectrum as well as our continuum definitions (Sect. 2.3) is shown in Fig. 1. Table 1: List of analysed QSOs, ordered by redshift. QSO \| Mag \| Filter \| S/N \| $f_{\nu_{0}}^{{\ddagger}}$ \| $\log L_{\nu_{0}}^{\star}$ \| ${z_{\mathrm{em}}}^{\dagger}$ ---\|---\|---\|---\|---\|---\|--- HE~1341$-$1020 \| 17.1 \| $B$ \| $70$ \| 139 \| 31.17 \| 2.137 Q~0122$-$380 \| 16.7 \| $B$ \| $60$ \| 291 \| 31.51 \| 2.192 PKS~1448$-$232 \| 17.0 \| $V$ \| $70$ \| 282 \| 31.50 \| 2.222 PKS~0237$-$233 \| 16.6 \| $V$ \| $100$ \| 750 \| 31.93 \| 2.224 HE~0001$-$2340 \| 16.7 \| $V$ \| $60$ \| 403 \| 31.68 \| 2.278 HE~1122$-$1648$\ {}^{\rm{a}}$ \| 16.5 \| $V$ \| $120$ \| 866 \| 32.05 \| 2.407 Q~0109$-$3518 \| 16.4 \| $B$ \| $70$ \| 404 \| 31.72 \| 2.406 HE~2217$-$2818 \| 16.0 \| $V$ \| $80$ \| 796 \| 32.02 \| 2.414 Q~0329$-$385 \| 17.0 \| $V$ \| $50$ \| 226 \| 31.48 \| 2.437$\ {}^{\rm{b}}$ HE~1158$-$1843 \| 16.9 \| $V$ \| $60$ \| 268 \| 31.56 \| 2.459 Q~2206$-$1958 \| 17.3 \| $V$ \| $70$ \| 255 \| 31.57 \| 2.567 Q~1232$+$0815 \| 18.4 \| $r$ \| $50$ \| 055 \| 30.91 \| 2.575 HE~1347$-$2457 \| 16.8 \| $B$ \| $60$ \| 772 \| 32.06 \| 2.615 HS~1140$+$2711$\ {}^{\rm{a}}$ \| 16.7 \| $r$ \| $80$ \| 581 \| 31.94 \| 2.628 Q~0453$-$423 \| 17.1 \| $V$ \| $70$ \| 168 \| 31.41 \| 2.664 PKS~0329$-$255 \| 17.1 \| $V$ \| $50$ \| 243 \| 31.58 \| 2.706 Q~1151$+$0651 \| 18.1 \| $r$ \| $50$ \| 100 \| 31.21 \| 2.758 Q~0002$-$422 \| 17.2 \| $V$ \| $60$ \| 346 \| 31.76 \| 2.769 Q~0913$+$0715 \| 17.8 \| $r$ \| $60$ \| 181 \| 31.48 \| 2.788 HE~0151$-$4326 \| 17.2 \| $B$ \| $70$ \| 499 \| 31.92 \| 2.787 Q~1409$+$095 \| 18.6 \| $r$ \| $40$ \| 067 \| 31.06 \| 2.843 HE~2347$-$4342 \| 16.7 \| $V$ \| $100$ \| 630 \| 32.05 \| 2.886 Q~1223$+$1753 \| 18.1 \| $r$ \| $40$ \| 794 \| 31.16 \| 2.955 Q~0216$+$080$\ {}^{\rm{a}}$ \| 18.1 \| $V$ \| $40$ \| 084 \| 31.20 \| 2.996 HE~2243$-$6031 \| 16.4 \| $V$ \| $100$ \| 776 \| 32.17 \| 3.012 CTQ~0247 \| 17.4 \| $V$ \| $70$ \| 239 \| 31.66 \| 3.025 HE~0940$-$1050 \| 16.4 \| $V$ \| $50$ \| 691 \| 32.13 \| 3.089 Q~0420$-$388 \| 16.9 \| $V$ \| $100$ \| 426 \| 31.93 \| 3.120 CTQ~0460 \| 17.5 \| $V$ \| $60$ \| 281 \| 31.76 \| 3.141 Q~2139$-$4434 \| 17.7 \| $V$ \| $40$ \| 117 \| 31.39 \| 3.207 Q~0347$-$3819 \| 17.7 \| $V$ \| $60$ \| 213 \| 31.66 \| 3.229 PKS~2126$-$158 \| 17.0 \| $V$ \| $60$ \| 380 \| 31.92 \| 3.285$\ {}^{\rm{c}}$ Q~1209$+$0919 \| 18.6 \| $r$ \| $40$ \| 120 \| 31.42 \| 3.291$\ {}^{\rm{c}}$ Q~0055$-$2659 \| 17.5 \| $V$ \| $60$ \| 245 \| 31.81 \| 3.665$\ {}^{\rm{c}}$ Q~1249$-$0159$\ {}^{\rm{a}}$ \| 18.6 \| $g$ \| $60$ \| 071 \| 31.27 \| 3.668 Q~1621$-$0042 \| 17.3 \| $r$ \| $70$ \| 307 \| 31.91 \| 3.709 Q~1317$-$0507 \| 17.7 \| $R$ \| $50$ \| 084 \| 31.35 \| 3.719 PKS~2000$-$330$\ {}^{\rm{a}}$ \| 17.0 \| $V$ \| $60$ \| 178 \| 31.69 \| 3.786 BR~1202$-$0725 \| 17.8 \| $R$ \| $100$ \| 460 \| 32.24 \| 4.697 Q~1451$-$1512 \| 17.3 \| $I$ \| $60$ \| 332 \| 32.12 \| 4.766 * ${\ddagger}$: Lyman limit flux in units of $\mu\mathrm{Jy}$ with uncertainties around 7%. * $\star$: Lyman limit luminosities in units of $\mathrm{erg}\,\mathrm{s}^{-1}\,\mathrm{Hz}^{-1}$ with uncertainties of the order of $\sigma_{\log L_{\nu_{0}}}\simeq^{+0.05}_{-0.1}$ dex. * $\dagger$: Systemic redshift estimated from Si ii$+$O i at $\lambda_{\mathrm{rest}}=1305.77$Å (Morton 2003). Redshift uncertainty is $\sigma_{z}=0.003$. * a: Gaps in the spectral range of the Ly$\alpha$ forest. * b: Redshift measured from C iv emission line and corrected according to Richards et al. (2002). * c: Redshift taken from Paper I. ### 2.2 Quasar magnitudes Table 1 provides our adopted apparent magnitudes and Lyman limit fluxes. The photometric data were collected from various sources, in particular SDSS, Véron-Cetty & Véron (2006), Rollinde et al. (2005), Cooke et al. (1997), Ellison et al. (2005), and Paper I. Where more than one measurement was available, the data were consistent to within $\sigma_{m}<0.1$ mag. Note that while knowledge of the Lyman limit luminosity is indispensable for the quantitative interpretation of the proximity effect, even perfectly simultaneous high-S/N photometry would not account for the effects of quasar variability averaged over the photoionisation time scale. By randomly changing the flux scale of some objects by up to 20% we confirmed that our conclusions concerning the proximity effect are not affected by photometric uncertainties. ### 2.3 Continuum definition In order to search for the proximity effect, we needed to convert the quasar spectra into continuum-rectified transmission spectra. We also needed to know the quasar flux at the Lyman limit. To achieve this we determined two types of continua: (i) a global power law ($f_{\nu}\propto\nu^{\alpha}$), excluding emission and absorption regions, used to estimate the quasar flux at the Lyman limit and (ii) a more local estimate that also includes the broad emission lines as a quasi-continuum, used to compute the transmission spectrum. We constrained the Lyman limit flux via a power law continuum fit since in most cases this wavelength is either not covered by the observations or is located in the far blue range of the spectrum where the S/N significantly drops. We developed an algorithm to automatically fit the local continuum, building on the work by Young et al. (1979) and Carswell et al. (1982). A cubic spline was interpolated on adaptive intervals along the spectrum. The fixpoints for the spline interpolation were chosen starting from a regular sampling of the spectrum with a binning that becomes finer whenever the slope of the computed continuum exceeds a given threshold. This was done in order to better reproduce the wings of emission lines. To assess the continuum uncertainties we proceeded in the same way as described in Paper I. With the help of Monte Carlo simulations we computed the average ratio between the recovered and the input continua at different redshifts. The results are presented in Fig. 2. The top panel shows the mean systematic error of the fitted continuum with respect to the input continuum for five different redshifts. The bottom panel presents the intrinsic uncertainties of the continuum. Due to the high resolution and the high S/N of the spectra, the statistical error associated with continuum placement is limited to a few percent at most in the Ly$\alpha$ line wing. The automatically fitted continuum was corrected for the systematic bias and used to compute the transmission $T=F_{\rm{qso}}/F_{\rm{cont}}$. Regarding the Lyman limit fluxes, the uncertainties are of the order of $7\%$, dominated by the somewhat subjective choice of the continuum wavelength ranges redward of the Ly$\alpha$ emission used to fit the power laws. Figure 2: Top panel: Average ratio between the fitted and input continuum for the five sets of 500 simulated QSOs at $z$ = 2.0, 2.5, 3.0, 3.5, 4.0 with $S/N=70$, denoted with numbers 1, 2, 3, 4 and 5 respectively. Bottom panel: Standard deviation profiles relative to the above systematic bias. ### 2.4 Systemic quasar redshift The spectral range covered by the UVES observations can maximally extend from $\sim$3 000 to $\sim$10 000 Å. However, not all QSOs have such wide wavelength coverage. Most of them extend far enough to the red of Ly$\alpha$ to allow the detection of several emission lines (Si ii+O i, C ii, Si iv+O iv], C iv) from which a redshift can be measured. Several authors (e.g. Gaskell 1982; Tytler & Fan 1992; Richards et al. 2002) pointed out the existence of a systematic shift between high and low ionisation lines attributed to the relative motion of gas near the AGN. In order to measure a redshift as close as possible to the systemic one, we used low-ionisation lines if these were available. In almost all cases we adopted Si ii+O i measurements for the final emission redshift, as listed in Table 1. For four objects (PKS 2126$-$158, Q 0329$-$385, Q 1209$+$0919 and Q 0055$-$2659) this was not possible due to either a non-detection of these lines or lack of wavelength coverage. For PKS 2126$-$158, Q 1209$+$0919 and Q 0055$-$2659 we adopted the redshifts based on Si ii+O i from the low-resolution spectra in Paper I, while for Q 0329$-$385 we could only take the C iv based redshift and corrected it for the mean systematic shift determined by Richards et al. (2002). Figure 3: Observed effective optical depth as a function of redshift after removal of pixels contaminated by Ly$\alpha$ absorption lines with damping wings, Ly limit systems and the proximity effect. Measurements from each individual quasar are shown in the upper panel, while those from the combined sample of QSOs are shown in the lower panel. The solid curve shows the power law least-squares fit to the data while the dashed curve indicates the evolution of the Ly$\alpha$ optical depth fitted by Schaye et al. (2003) and the dotted line that by Kim et al. (2007). ## 3 Monte Carlo simulations of artificial spectra A deeper understanding of statistical and systematic effects contributing to the proximity effect signature is made possible by comparing observations with simulated spectra. Typically, numerical 3D simulations are successful in reproducing the properties of the Ly$\alpha$ forest at any redshift. However, such computations are very time consuming and their major shortcoming is the limited redshift coverage of the simulated sight lines, especially at high resolution. We developed a straightforward Monte Carlo code to generate synthetic spectra and employed the simulations for three main purposes. First, to estimate the influence of the quasar continuum placement in the spectra (Sect. 2.3); then to compute the line-of-sight variations in the evolution of the optical depth in the Ly$\alpha$ forest (Sect. 4); and finally to study the influence of the fact that only a finite number of absorbers contributes to the measured optical depth (Sect. 5.3). The procedure used to generate synthetic spectra is based on the assumption that the Ly$\alpha$ forest is well represented by three observed distributions: 1. 1. The line number density distribution, approximated by a power law of the form $\mathrm{d}n/\mathrm{d}z\propto(1+z)^{\gamma}$ where $\gamma$ will be measured in Sect. 4. 2. 2. The column density distribution, given by $f(N_{\mbox{\scriptsize{H {i}}}})\propto N_{\mbox{\scriptsize{H {i}}}}^{-\beta}$ where the slope is $\beta\simeq 1.5$ (Kim et al. 2001). 3. 3. The Doppler parameter distribution, given by $\mathrm{d}n/\mathrm{d}b\propto b^{-5}\mathrm{exp}\left[{-{b_{\sigma}^{4}}/{b^{4}}}\right]$ where $b_{\sigma}\simeq 24\;\mathrm{km/s}$ (Kim et al. 2001). The simulated absorbers have column densities within the range $10^{12}<N_{\mbox{\scriptsize{H {i}}}}<10^{18}$ cm-2 and Doppler parameters between $10<b<100$ km/s. The slope of the column density distribution was fixed to $\beta=1.5$. For our purposes these approximations are sufficient to yield the accuracy we need in the following analysis. Every simulated sight line was populated with absorption features drawn from the above distributions until the computed effective optical depth was consistent with its evolution presented in Sect. 4. Once the simulated transmission was computed, an artificial quasar spectral energy distribution, including emission lines of varying strengths and widths, was generated via the principal components method described by Suzuki (2006) and already employed in Paper I. Gaussian noise was added to the final quasar spectrum, in order to match the S/N level of our observed objects. Table 2: Effective optical depth in the Ly$\alpha$ forest of our combined QSO sample. $<z>$ \| $\Delta z$ \| $\tau_{\mathrm{eff}}$ \| $\sigma_{\tau_{\mathrm{eff}}}$ ---\|---\|---\|--- 1.874 \| 0.15 \| 0.166 \| 0.021 2.024 \| 0.15 \| 0.180 \| 0.026 2.174 \| 0.15 \| 0.185 \| 0.028 2.324 \| 0.15 \| 0.217 \| 0.031 2.474 \| 0.15 \| 0.270 \| 0.035 2.624 \| 0.15 \| 0.327 \| 0.040 2.774 \| 0.15 \| 0.349 \| 0.042 2.924 \| 0.15 \| 0.378 \| 0.046 3.074 \| 0.15 \| 0.420 \| 0.051 3.224 \| 0.15 \| 0.388 \| 0.054 3.374 \| 0.15 \| 0.439 \| 0.061 3.524 \| 0.15 \| 0.569 \| 0.068 3.749 \| 0.30 \| 0.726 \| 0.054 3.974 \| 0.15 \| 0.807 \| 0.084 4.124 \| 0.15 \| 0.861 \| 0.093 4.274 \| 0.15 \| 0.967 \| 0.095 4.483 \| 0.27 \| 1.093 \| 0.099 ## 4 Evolution of the Ly$\alpha$ effective optical depth The H i effective optical depth is related to the mean transmission as $\tau_{\rm eff}\equiv-\ln\left<T\right>\equiv-\ln\left<e^{-\tau_{H\textsc{i}}}\right>$, where $\left<\>\right>$ indicates the average over predefined redshift intervals. Note that the effective optical depth is defined as the negative logarithm of the average pixel by pixel transmission, and not as the average value of the pixel by pixel optical depth in a given redshift interval. Its redshift evolution (Zuo 1993) is usually well represented by a power law $\tau_{\rm{eff}}=\tau_{0}(1+z)^{\gamma+1}.$ (1) We explored two methods of determining the evolution of $\tau_{\rm eff}$ with redshift. In the first approach each Ly$\alpha$ forest spectrum was divided into two redshift intervals, each covering the same redshift path length to sample the increasing opacity along single sight lines. The upper panel of Fig. 3 shows the results obtained using all those pixels in the Ly$\alpha$ forest range decontaminated from damped Ly$\alpha$ or Lyman limit systems (DLAs and LLSs respectively) and the proximity effect region ($\Delta v<5000$km s-1 were excluded). The second approach consists of measuring the mean $\tau_{\rm eff}$ in the forest of all quasars intersecting a particular redshift slice ($\Delta z=0.15$). The results are shown in Table 2 and in the lower panel of Fig. 3, yielding best-fit values of $(\log\tau_{0},\gamma)=(-2.21\pm 0.09,2.04\pm 0.17)$. For both approaches the uncertainties in the effective optical depth, $\sigma_{\tau_{\mathrm{eff}}}$, quantify the amount of variance between the individual lines of sight for the considered redshift range. In Figure 3 we present the results from the two approaches in comparison to the evolution of the Ly$\alpha$ optical depth as measured by Schaye et al. (2003), using a similar technique on 19 QSO spectra taken with UVES and HIRES and the fit by Kim et al. (2007) combining 18 UVES spectra. Our results agree well with both Schaye et al. (2003) and Kim et al. (2007) fits. There is less agreement at high redshift, probably due to the limited number of QSOs at $z>3.6$, but our results are formally consistent within the errors. Although both panels use the same data, these were combined in different ways and thus return slightly different results. It is clear from the top panel that the first method suffers from a lack of data points in the range $3.6<z<4$, leading us to slightly underestimate the effective optical depth at high redshift. The second approach significantly reduces the scatter caused by cosmic variance as already pointed out by Kim et al. (2002). It is also in good agreement with literature estimates. Therefore we adopt the best-fit parameters from the second method to describe the redshift dependence of the normalised effective optical depth in our quasar spectra. In the bottom panel of Fig. 3 we identify a marginally significant departure of $\tau_{\rm eff}$ from the power law evolution in the redshift range $3.2\lesssim z\lesssim 3.4$. This feature, first discovered by Bernardi et al. (2003) using SDSS quasar spectra (but see McDonald et al. 2005), has been been interpreted to be a result of one or more of the following effects: (i) a change in the IGM temperature; (ii) a change in the free electron number density; or (iii) an enhancement in the UV background possibly connected to He ii reionisation. Recently, Faucher-Giguère et al. (2008c) found the same dip in $\tau_{\rm eff}(z)$ using a sample of ESI and HIRES spectra. In the light of this debate and using a completely independent data sample, we confirm the detection of this feature. We note, however, that its significance is only 2.6$\sigma$ (two pixels with each 1.8$\sigma$), so the detection can be called no more than tentative. ## 5 Methods of quantifying the proximity effect The direct signature of an enhanced ionisation field near bright quasars is a reduction of the neutral hydrogen fraction on a physical scale of several Mpc. Two principal techniques to detect it have been developed in the last two decades, which we briefly review in the following subsections. ### 5.1 Flux transmission statistics The method used in this work to reveal the proximity effect along single lines of sight is based on the comparison between the observed effective optical depth and the expected one in the Ly$\alpha$ forest (Liske & Williger 2001). Approaching the quasar systemic redshift, the photoionisation rate of the source starts to dominate over the UVB. This effect leads to an enhancement in H+ relative to H0, reducing the opacity of absorbers close to the QSOs. This influence modifies the effective optical depth, which becomes $\tau_{\mathrm{eff}}=\tau_{0}(1+z)^{\gamma+1}(1+\omega)^{1-\beta}$ (2) (Liske & Williger 2001) where $\beta$ is the slope in the column density distribution and $\omega$ is the ratio between the quasar and background photoionisation rates. Following the assumptions outlined in Bajtlik et al. (1988) to compute $\omega$, we get $\omega(z)=\frac{f_{\nu}(\lambda_{0}(1+z))}{4\pi J_{\nu_{0}}}\frac{1}{(1+z)}\left(\frac{d_{L}(z_{\mathrm{q}},0)}{d_{L}(z_{\mathrm{q}},z)}\right)^{2}$ (3) with $z$ as the redshift along the LOS such that $z<z_{\mathrm{q}}$, $d_{L}(z_{\mathrm{q}},0)$ the luminosity distance of the QSO to the observer, and $d_{L}(z_{\mathrm{q}},z)$ as seen at any foreground redshift in the LOS. Equation 3 implies that the spectral shape of the QSO and the background are approximately the same around 1 Ryd, which might not be exactely true. Assuming a quasar spectral shape $f_{\nu}\propto\nu^{\alpha}$, with $\alpha_{\mathrm{q}}\simeq-0.5$ (Vanden Berk et al. 2001) and taking a background spectral index of $\alpha_{\mathrm{b}}\simeq-1.4$ (Agafonova et al. 2005), the resulting scale of $\omega$ is shifted by $\sim 0.1$ dex. However, there may be an indication of a break in the spectral slope of quasars around the Lyman limit (Telfer et al. 2002). Such a break would make the quasar spectrum more similar again to that of the background, reducing the effect on $\omega$. We adopt the following notation: The ratio of the observed optical depth to the one expected in the Ly$\alpha$ forest, or the _normalised effective optical depth_ $\xi$, is given by $\xi=\frac{\tau_{\mathrm{eff}}}{\tau_{0}(1+z)^{\gamma+1}}=(1+\omega)^{1-\beta}.$ (4) with the parameters $(\log\tau_{0},\gamma)$ measured in Sect. 4. The proximity effect is apparent as a decrease in the normalised optical depth below unity as $\omega\rightarrow\infty$, i.e. towards the quasar. We make use of this technique, first to detect the proximity effect on our combined sample of quasars (Sect. 6), then towards single lines of sight (Sect. 7) and finally to assess the dependence of the measured UV background on redshift and overdensities (Sect. 9). Figure 4: Statistical errors associated with individual lines of sight using the line statistic (LS) and the flux statistic (FS) method. The different column density thresholds represent the completeness limits from Scott et al. (2000) and Hu et al. (1995) (solid squares and triangles, respectively). The filled circles show the level of uncertainty if the line list were complete up to the minimum column density at our resolution and S/N. ### 5.2 Line counting Counting absorption lines stronger than a given threshold is historically the first technique used to measure the deficit of absorption near high redshift quasars. Typically represented by a Voigt profile, each absorption line has to be fitted and associated with a set of parameters $(N_{{H\textsc{i}}},z,b)$ describing its column density, redshift, and Doppler parameter, respectively. At medium resolution only the equivalent width can be measured. Several authors (e.g., Giallongo et al. 1996; Scott et al. 2000) used this approach to detect the proximity effect and to measure the UV background on combined samples of QSOs with the formalism introduced by Bajtlik et al. (1988). It is clear that at low spectral resolution, line counting is vastly inferior to the flux transmisson statistic because of line blending. On the other hand, at the high resolution of UVES and similar instruments, line counting could still be a reasonable approach. We therefore conducted a performance comparison of the two techniques, following the prescription in Sect. 3. We simulated the Ly$\alpha$ forest of a quasar at redshift $z=3$, obtaining not only the spectrum but also keeping the _complete_ line list. We repeated this step to create 500 independent simulated spectra. The principal limitation of the absorption line fitting approach is line blending, leading to incompleteness of weak lines at low column densities. Even at high resolution where lines with $\log N_{{H\textsc{i}}}\simeq 12.5$ are detectable, the completeness is $<25\%$ for column densities $\log N_{{H\textsc{i}}}<13.2$ (Hu et al. 1995). Another problem in fitting absorption lines is that there is no unique solution, and much is left to the experience and ability of the scientist/software code in finding a _good_ fit. For our simulation study we employed the code AUTOVP111Developed by R. Davé: http://ursa.as.arizona.edu/$\sim$rad to fully automatically fit an array of absorption lines to each spectrum. The proximity effect signature is then a decrease in the number density of lines $N$ relative to that expected for the redshift. Thus we can define an alternative quantity $\xi^{\prime}=\frac{N}{N_{0}}=(1+\omega)^{1-\beta},$ (5) which again is expected to decrease from unity to zero as $\omega$ increases. We measured $\xi$ (using the flux transmission statistic) and $\xi^{\prime}$ (from the automated line counts, adopting different column density thresholds) for all 500 simulated spectra, and computed mean and dispersion values for several bins of $\log\omega$. Figure 4 presents the results of this study, showing the standard deviation of $\xi$ or $\xi^{\prime}$, i.e. the statistical error of the measurement for a single line of sight, as a function of $\omega$ (approaching the quasar). For a realistic completeness limit in column density $\log N_{H\textsc{i}}>13$, the flux transmission approach is always more sensitive than line counting. In the unrealistic ideal case where we used the true input line lists, the two methods are equivalent at low $\omega$, but even then the line counting approach fails in probing $\xi^{\prime}$ very close to the emission redshift of the quasar. This can easily be understood as the quasar radiation reduces the column densities by a factor $1+\omega$; thus, close to the quasar, where $\omega\ga 100$, original column densities of $\log N_{{H\textsc{i}}}\ga 15$ are needed to produce a detectable absorption line. Such absorbers are very rare, which leads to increased uncertainties. We conclude that line counting is less sensitive to the expected proximity effect signature than the flux transmission method, even at high spectral resolution, and all our results will be based exclusively on the flux transmission technique. Figure 5: Distribution of the normalised optical depth $\xi$ under the influence of increasing $\omega$ values. The histograms represent the $\xi$ distribution of 500 simulated spectra with $z_{\mathrm{QSO}}=3$. ### 5.3 Poissonian variance between lines of sight At the spectral resolution in our sample, the major source of uncertainty is due to the variation in the number of absorption features along individual lines of sight. The probability of finding or not finding an absorption line at a given redshift is to first approximation a Poissonian process which can be easily modelled by our Monte Carlo simulations. While a full accounting of the expected effects of cosmic variance due to the large-scale structure of the IGM would require cosmological 3D simulations, it is actually useful to start with the Poissonian approximation, as it allows us to appreciate to what extent the observed level of variance is consistent with pure Poisson noise. Equation 3 shows that $\omega$ depends on the redshift and luminosity of the quasar, implying that each QSO will have its own particular $\log\omega$ scale. Consequently, in a fixed $\log\omega$ interval the influence of Poisson noise will vary from line of sight to line of sight, since the corresponding redshift range in the Ly$\alpha$ forest is determined by the quasar emission redshift and its Lyman limit luminosity. Therefore, the distribution of the normalised optical depth $\xi$ as a function of $\log\omega$ will depend on (i) Poisson noise, (ii) quasar redshift, and (iii) quasar luminosity. We studied how the distribution of $\xi$ is affected by different UV radiation fields (i.e. in different $\log\omega$ bins) at a given redshift. We computed the standard deviation $\sigma$ given by $\sigma^{2}=\int(\xi-\mu)^{2}\ P(\xi)\,\mathrm{d}\xi$ (6) where $P(\xi)$ describes the probability density function of $\xi$ estimated using our mock catalogue of sight lines as displayed in Fig. 5, and $\mu$ is its mean value. For illustration purposes, we assumed a quasar at redshift $z=3$ and luminosity $\log L_{\nu_{0}}=31.5$ with a binning of $\Delta\log\omega=1$. Each panel displays how Poisson noise affects a different portion of the Ly$\alpha$ forest according to the quasar properties, i.e. for different values of $\omega$. The final error bars in the proximity effect analysis are a composition of the following three effects, ordered by descending relative importance: (i) the standard deviation computed above, (ii) the continuum uncertainties (Sect. 2.3) and (iii) the intrinsic noise present in the spectrum. The Poisson noise in the placement of lines dominates the total error. Continuum uncertainties and the intrinsic noise are almost negligible. Figure 6: Normalised effective optical depth versus $\omega$ for the combined sample of 40 quasars, binned in steps of $\Delta\log\omega=1$. The curved lines are the best fit of the simple photoionisation model to the data, corresponding to a UV background of $\log(J_{\nu_{0}})=-21.10$ (solid line) in units of erg cm-2 s-1 Hz-1 sr-1, while the dotted curve denotes to the fiducial UVB ($J^{\star}$) for the computation of $\omega$. The horizontal dashed line illustrates the case of no proximity effect. Figure 7: The proximity effect signatures in individual lines of sight, for a subsample of 9 QSOs. Each panel shows the normalised effective optical depth $\xi$ versus $\omega$, binned in steps of $\Delta\log\omega=1$, with the best- fit model of the combined analysis superimposed as the dotted red line (see Sect. 6). The solid lines delineate the best fit to each individual QSO as described in the text. This subsample was chosen for presentation purposes to show the variable strength of the proximity effect, going from _strong_ (top- left panel) to _weak_ (bottom-right panel). ## 6 The proximity effect in the combined sample of quasars We first present briefly the evidence of the proximity effect in the traditional way, as a signal in the combined full sample of QSOs. The result is displayed in Fig. 6. As expected, the normalised effective optical depth $\xi$ starts at unity for $\omega\ll 1$ and then goes down to zero for $\omega\gg 1$. In the definition of $\omega$ (Eq. 3) we had to assume a fiducial value of the UVB intensity, $J^{\star}=10^{-21}\,\mathrm{erg}\,\mathrm{cm}^{-2}\,\mathrm{s}^{-1}\,\mathrm{Hz}^{-1}\,\mathrm{sr}^{-1}$. This uniquely converts the redshift scale into an $\omega$ scale. We defined a regularly spaced grid along the $\log\omega$ axis and in each bin we determined the average transmission and effective optical depth values, and using Eq. 4 the corresponding values of $\xi$. We then modelled the decrease in $\xi$ with $\log(\omega)$ according to the formula $F(\omega)=\left(1+\frac{\omega}{a}\right)^{1-\beta}$ (7) where $a$ is the single free parameter which expresses the _observed_ turnover of $\xi$ and thereby provides the best-fit value of $J_{\nu_{0}}$, since obviously $J_{\nu_{0}}=a\times J^{\star}$. The slope of the column density distribution was fixed to $\beta=1.5$ (see Sect. 3 for details). For the median redshift of our sample of $z=2.73$ we thus obtain $J_{\nu_{0}}=(7.9\pm 3.1)\times 10^{-22}\,\mathrm{erg}\,\mathrm{cm}^{-2}\,\mathrm{s}^{-1}\,\mathrm{Hz}^{-1}\,\mathrm{sr}^{-1}$, or in logarithmic units $\log J_{\nu_{0}}=-21.10^{+0.14}_{-0.22}$. This number is very close to the value of $\log J_{\nu_{0}}=-21.03$ reported by us in Paper I and is also consistent with several other literature estimates. A detailed comparison with the literature is hampered by the fact that the vast majority of previous UVB measurements were derived adopting the almost obsolete Einstein-de Sitter model. Typically, values for $J_{\nu_{0}}$ are then about a factor of 1.4 (0.15 dex) higher than in a $\Lambda-$Universe. Cooke et al. (1997) measured $\log J_{\nu_{0}}=-21.00_{-0.15}^{+0.18}$ on 11 high resolution quasar spectra using line statistics. Scott et al. (2000) obtained $\log J_{\nu_{0}}=-21.15_{-0.43}^{+0.17}$, applying the same method on more than a hundred spectra at $\sim 1$ Å resolution. Liske & Williger (2001) used the flux statistic on 10 QSO spectra with $\sim 2$ Å resolution and a S/N of $\sim 40$, obtaining $\log J_{\nu_{0}}=-21.45_{-0.20}^{+0.30}$. While the formal errors of this measurement are small, there are reasons to believe that the proximity effect exploited this way delivers systematically too high values of $J_{\nu_{0}}$. Overdense environments around quasars have been suggested as the prime reason; we shall demonstrate in the next section that in fact the averaging process inherent in the sample combination is responsible for a major bias in the UVB determination. Figure 8: Observed distribution of the proximity effect strength (given by $\log a$) for our sample of objects. The distribution is characterised by a prominent peak and a skewed profile which extends over several orders of magnitude in strength. Table 3: The proximity effect strength $\log a$ along single lines of sight. QSO \| $z_{\rm{em}}$ \| $\log L_{\nu_{0}}^{\dagger}$ \| $\log a$ ---\|---\|---\|--- Q 1232$+$0815 \| 2.575 \| 30.91 \| $-1.89\pm 0.37$ Q 0329$-$385 \| 2.438 \| 31.48 \| $-1.86\pm 0.29$ PKS 0329$-$255 \| 2.706 \| 31.58 \| $-1.86\pm 0.25$ HE 0001$-$2340 \| 2.278 \| 31.68 \| $-1.46\pm 0.36$ HE 1347$-$2457 \| 2.615 \| 32.06 \| $-1.43\pm 0.25$ Q 0216$+$080 \| 2.996 \| 31.20 \| $-1.26\pm 0.36$ HE 2217$-$2818 \| 2.414 \| 32.02 \| $-1.21\pm 0.31$ Q 0109$-$3518 \| 2.406 \| 31.72 \| $-1.10\pm 0.38$ Q 1249$-$0159 \| 3.668 \| 31.27 \| $-1.08\pm 0.22$ PKS 2126$-$158 \| 3.285 \| 31.92 \| $-0.99\pm 0.22$ Q 1209$+$0919 \| 3.291 \| 31.42 \| $-0.85\pm 0.29$ Q 0913$+$0715 \| 2.788 \| 31.48 \| $-0.84\pm 0.34$ Q 0002$-$422 \| 2.769 \| 31.76 \| $-0.78\pm 0.32$ PKS 1448$-$232 \| 2.222 \| 31.50 \| $-0.75\pm 0.47$ Q 0055$-$2659 \| 3.665 \| 31.81 \| $-0.67\pm 0.20$ HS 1140$+$2711 \| 2.628 \| 31.94 \| $-0.63\pm 0.35$ HE 2243$-$6031 \| 3.012 \| 32.17 \| $-0.61\pm 0.24$ CTQ 0460 \| 3.141 \| 31.76 \| $-0.61\pm 0.27$ HE 0940$-$1050 \| 3.089 \| 32.13 \| $-0.49\pm 0.26$ Q 1621$-$0042 \| 3.709 \| 31.91 \| $-0.49\pm 0.20$ Q 1151$+$0651 \| 2.758 \| 31.21 \| $-0.45\pm 0.46$ PKS 2000$-$330 \| 3.786 \| 31.69 \| $-0.45\pm 0.23$ Q 0122$-$380 \| 2.192 \| 31.51 \| $-0.42\pm 0.44$ Q 1409$+$095 \| 2.843 \| 31.06 \| $-0.33\pm 0.54$ Q 0347$-$3819 \| 3.229 \| 31.66 \| $-0.23\pm 0.30$ Q 0420$-$388 \| 3.120 \| 31.93 \| $-0.18\pm 0.32$ Q 1451$-$1512 \| 4.766 \| 32.12 \| $-0.01\pm 0.14$ Q 1317$-$0507 \| 3.719 \| 31.35 \| $+0.13\pm 0.31$ Q 0453$-$423 \| 2.664 \| 31.41 \| $+0.18\pm 0.45$ HE 1122$-$1648 \| 2.407 \| 32.05 \| $+0.30\pm 0.33$ HE 1158$-$1843 \| 2.459 \| 31.56 \| $+0.35\pm 0.38$ BR 1202$-$0725 \| 4.697 \| 32.24 \| $+0.51\pm 0.14$ Q 1223$+$1753 \| 2.955 \| 31.16 \| $+0.58\pm 0.36$ CTQ 0247 \| 3.025 \| 31.66 \| $+0.76\pm 0.25$ HE 2347$-$4342 \| 2.886 \| 32.05 \| $+0.87\pm 0.17$ Q 2206$-$1958 \| 2.567 \| 31.57 \| $+1.15\pm 0.21$ HE 1341$-$1020 \| 2.137 \| 31.17 \| $+1.51\pm 0.14$ PKS 0237$-$233 \| 2.224 \| 31.93 \| $+2.07\pm 0.12$ Q 2139$-$4434 \| 3.207 \| 31.39 \| $+2.32\pm 0.26$ HE 0151$-$4326 \| 2.787 \| 31.92 \| $+2.57\pm 0.15$ * $\dagger$: Lyman limit luminosities and uncertainties $\left(\sigma_{\log L_{\nu_{0}}}\simeq^{+0.05}_{-0.1}\right)$ in units of $\mathrm{erg}\,\mathrm{s}^{-1}\,\mathrm{Hz}^{-1}$ ## 7 The proximity effect in individual lines of sight Figure 9: Proximity effect strength distributions in our synthetic spectra for different quasar redshifts (left-hand panel) and luminosities (right-hand panel). The distributions are computed from 500 artificial sight lines generated by our Monte Carlo code as described in Sect. 3 and are affected only by Poissonian shot noise. ### 7.1 The distribution of the proximity effect strengths For our investigation of the proximity effect towards single lines of sight, the approach was essentially identical to that of the combined analysis: We computed the normalised effective optical depths for individual sight lines within a given range $\Delta\log\omega$ and checked whether $\xi$ systematically decreases for high values of $\omega$. For presentation purposes, the results of a subsample of 9 objects are displayed in Fig. 7, one panel per quasar (see Appendix A for the complete sample). The error bars are dominated by Poissonian shot noise, estimated from the simulations as described in Sect. 5.3. In each panel, $\xi(\omega)$ is shown for the fiducial reference value $J_{\nu_{0}}=J^{\star}$. Figures 7 and 18 illustrate that in all cases, $\xi$ decreases substantially as $\omega$ increases. Thus we conclude that the proximity effect is detected in all of our quasar spectra. This is in excellent agreement with our finding in Paper I, where in only 1 out of 17 low-resolution quasar spectra we failed to detect the proximity effect (and for that object, HE 2347$-$4342, we now detect a weak effect in the high-resolution UVES spectrum; see Appendix A). We then applied the above fitting procedure (Eq.7) to each spectrum separately. The results are shown as the solid curves in Fig. 7, while Table 3 summarises the fit results. The value of the fitting parameter $a$, which describes the horizontal offset of the solid curve relative to the dotted one, will be regarded in the following as a measure of the strength of the proximity effect signal. It shows significant scatter between the different quasars. Note that the zeropoint of $\log a$ is arbitrary and depends only on our assumed fiducial UVB value $J^{\star}$. Figure 8 presents the measured proximity effect strength distribution (PESD), i.e. the distribution of $\log a$ for our sample. Three major characteristics emerge: (i) the distribution covers almost five orders of magnitude in $a$, ranging from a very weak proximity effect ($a\gg 1$) to a very strong one ($a\ll 1$), (ii) it shows a well-defined peak, and (iii) it is significantly skewed. In the following we will investigate the effects behind these properties. ### 7.2 The effects of variance between sight lines The most obvious contributor to a spread in proximity effect strengths is the variance in the absorber distribution between different lines of sight. To investigate this effect we employed our Monte Carlo simulated spectra, for which by design this variance can be accurately described as a pure Poissonian process. The proximity effect was introduced in the synthetic spectra as a reduction in the optical depth by a factor $1+\omega$, with $\omega$ given by Eq. 3. We first fixed the luminosity of the quasar to a value of $\log(L_{\nu_{0}}/\mathrm{erg\ s^{-1}\ Hz^{-1}})=31.5$ and considered three different redshifts: $z$ = 2.0, 3.0, and 4.0, in order to see how the distribution of $a$ changes with $z$. Then we took a constant redshift of $z=3.0$ and considered three quasar luminosities: $\log(L_{\nu_{0}}/\mathrm{erg\ s^{-1}\ Hz^{-1}})$ = 31, 31.5, and 32. The UV background was fixed to $J_{\nu_{0}}=J^{\star}$ in both cases. Our results are presented in Fig. 9; recall that high $a$ values imply a weak proximity effect. Two main characteristics become apparent in Fig. 9: The distributions are significantly skewed, and the skewness decreases towards higher redshifts and higher luminosities. This asymmetry is a direct result of the non-linear dependence of $\omega$ on redshift as expressed in eq. 3. Towards the quasar redshift equal $\Delta\log\omega$ bins correspond to progressively smaller $\Delta z$ intervals. Very close to the QSO, the (Poissonian) distribution of the number of absorbers per $\Delta\log\omega$ bin deviates significantly from a Gaussian. Similarly the distribution of effective optical depth becomes skewed at $z\rightarrow z_{\mathrm{em}}$, resulting in high $\tau_{\mathrm{eff}}$ values being more likely than expected for a normal distribution. The asymmetry of the PESD directly reflects this statistical effect. The dependence with redshift can be explained by the fact that on the one hand, Poisson noise has a weaker impact at high redshifts and on the other hand, an increasing luminosity is capable of a substantial reduction in the optical depth further away from the quasar emission, overionising even strong systems in the quasar vicinity. Because of the asymmetric nature of the PESD at any given redshift and luminosity, one tends to overestimate the best fit values of Eq. 7 when combining multiple sight lines together in an averaging process. By comparing how the distribution widths evolve with redshift and luminosity, we conclude that luminosity has a weaker effect than redshift in changing the shape of the PESD. A remarkable feature of Fig. 9 is the fact that the mode of the distribution always stays at the input value corresponding to $\log a=0$. We exploit this property in the next section. Figure 10: Comparison between observed (thick) and simulated (thin) distribution of the proximity effect strength (given by $\log(a)$) for our sample of quasars. We mark the position of the mode of the observed distribution and the relative error associated to it. The error bars are computed from the dispersion of the mode in the 500 realisations of the simulated sample. We also display the value of the UVB computed in the combined analysis of the proximity effect. ### 7.3 An unbiased measurement of the UV background Unfortunately, we do not have 500 observed quasar spectra at one particular redshift or luminosity, but rather a quasar sample distributed in redshift and luminosity. We created another set of simulated spectra to predict the ‘PESD as observed’, assuming that only Poissonian variance contributes to the spread of $\log a$, in the following way: We constructed 40 simulated QSO spectra with the condition that their redshifts and luminosities were exactly the same as the observed ones, and we constructed the PESD for this ‘sample’. We repeated the process 500 times and averaged the PESD. The result is shown in Fig. 10 as a thin histogram, superimposed on the observed data of our quasar sample. The predicted distribution of the proximity effect strength parameter $\log a$ looks amazingly similar to the observed one. In particular, the degree of asymmetry is very nearly the same. The observed PESD seems to show a slightly narrower core, and it certainly has some outliers on both sides, at $\log a\simeq-2$ and $\log a>2$. We interpret this excellent agreement as an indication that the statistical fluctuations of the distribution of absorption features between individual sight lines is responsible for most of the spread and in fact for most of the asymmetry in the PESD. What does this imply for the UV background? We again find that the modal value of the ‘as observed’ distribution recovers the input UVB intensity, whereas the combined analysis heavily overestimates it. From the 500 realisations we derive that the mode of the PESD of 40 quasars is consistent with the value of input model to within $\pm 0.15$ dex. Similarly we can quantify the bias introduced by the combined analysis. For this purpose, we computed the best- fit $\log a$ as in Sect. 6 and averaged it over the 500 realisations. Typically, the averaged value of $J_{\nu_{0}}$ is overestimated by $\Delta\log a=0.3$ dex, or a factor of 2. Figure 11: Redshift evolution of the UVB modelled by Haardt & Madau (1996) in comparison to previous measurements of the UVB via the proximity effect (Bajtlik et al. 1988; Lu et al. 1991; Giallongo et al. 1993; Williger et al. 1994; Bechtold 1994; Fernández-Soto et al. 1995; Srianand & Khare 1996; Giallongo et al. 1996; Lu et al. 1996; Savaglio et al. 1997; Cooke et al. 1997; Scott et al. 2000; Liske & Williger 2001), all for an Einstein-de Sitter universe. Figure 12: Measured intensity of the UVB for a $\Lambda$ universe. Our best estimate is the modal value of $J_{\nu_{0}}$ (filled red circle); the biased ‘combined value’ (open square) as well as the similarly obtained value from Paper I (open circle) are shown for comparison. Also shown are the most recent evolving UVB model by Haardt & Madau (2001) (continuous green line) and the determination by Bolton et al. (2005). The horizontal bars indicate the lower and upper quartile of the quasar redshifts. From Fig. 10 we estimated the position of the mode of the observed PESD to be located at $\log a=-0.51$ and converted it into a measurement of the UVB, yielding a value of $\log J_{\nu_{0}}=-21.51\pm 0.15$. The uncertainty was derived from the dispersion of the mode in the simulated PESDs. This value of the UVB is 0.4 dex lower than the best-fit of the combined proximity effect analysis given in Sect. 6. For completeness we also considered what happens when one computes the average of the best-fit $\log a$ values for individual quasars (as listed in Table 3), instead of fitting the average $\xi$-$\omega$ profile. This results in a UVB value of $\log J_{\nu_{0}}=-21.24\pm 0.17$, lower than the outcome of the combined proximity effect analysis as well, but still much higher than the modal value (and also systematically offset in the simulations). We compiled several literature results in measuring the UVB by the proximity effect together with our new our results in Figs. 11 and 12. As said above, the cosmological model employed does have an effect on the UVB result. Figure 11 collects literature measurements that were performed for an Einstein-de Sitter model. Most of them agree poorly with the UVB model by Haardt & Madau (1996) (which was computed for the same cosmological model), in the sense that the measurements are systematically higher. Figure 12 puts our new determination of $J_{\nu_{0}}$ into context. It is remarkable that our modal value is almost exactly on the predicted curve based on the newer UVB model by Haardt & Madau (2001). Their computation was based on assuming a combined contribution of QSOs and star-forming galaxies to the UV background, now performed for a $\Lambda$ universe. It is also fully consistent with the determination of the UVB photoionisation rate by Bolton et al. (2005) obtained by matching the observed Ly forest transmission to cosmological simulations. It is equally evident that our ‘combined value’ is too high, as is our similarly obtained best-fit value in Paper I. We summarise this section by stating that the distribution of proximity effect strengths as determined in individual quasar spectra is largely consistent with a purely Poissonian random process. We thus understand why the PESD is asymmetric, and why previous measurements based on averaging sample properties were bound to overestimate the UVB background. So far we have identified the modal value of the PESD as a largely unbiased estimator of the UVB. In the next section we will improve on this estimator by proposing a hybrid approach that also takes physical overdensities into account. ## 8 The role of overdensities around quasars ### 8.1 Quantifying overdensities We have demonstrated above that much of the disagreement between proximity effect and other measurements of the UVB is due to the asymmetric shape of the PESD, even for the idealised case of purely Poissonian variance between sight lines. However, this does not imply that other effects can be neglected altogether. The most suspicious additional bias in this context is the degree of typical H0 overdensity within a few Mpc around quasars. Obviously, if a quasar environment is overdense and thus more likely to show high column density absorption than the general Lyman forest, the proximity effect will appear to be weakened (Loeb & Eisenstein 1995). This issue has been recently addressed by Rollinde et al. (2005) and Guimarães et al. (2007) using both observations and numerical simulations. Their basic approach consisted of a comparison of the cumulative optical depth probability distributions at difference distances to their quasars. Close to the quasars, that distribution deviates from the one for the general Lyman forest. Assuming that the UVB photoionisation rate is well constrained from the cosmic mean H0 opacity and accounting for the expected proximity effect signature based on the known quasar luminosities, they concluded from the discrepancy between the measured and predicted distributions that quasars reside in physical overdensities of a factor of a few (averaged over $\sim 2.5h^{-1}$ proper Mpc). The amount of inferred overdensity depends, as one would expect, on the adopted value of the general UVB photoionisation rate. Here we approach the same problem from a different angle. We also assume that the mean UVB photoionisation rate, or the mean intensity $J_{\nu_{0}}$, can be measured. We adopt the modal value of the PESD estimated in Sect. 7.3, which we demonstrated above to be in excellent agreement with other methods of estimating the UVB. We then again consider each individual quasar in our sample in turn, aiming at measuring the _distribution_ of (over)density properties. For the present paper we are not interested in estimating physical overdensities, but only in identifying the quasars with strongly overdense environments. We therefore adopted a pragmatic definition of ‘density’ that simply involves the absorber optical depth within a certain filter length as a proxy. We computed this ‘overdensity degree’ $\Xi$ in the following way: Figure 13: Average overdensity degree $\Xi$ as function of proper distance for the complete sample of quasars (solid line) and in a subsample with $\Xi$ consistent with no sign of overdensities within 3 Mpc from the quasar (dotted line). The two horizontal lines mark the $\pm 1\sigma$ scatter expected from our simulation in the case of no overdense environment. 1. 1. We reconstructed the optical depth profile $\tau^{}(z)$, _removing the radiative influence of the quasar_ by multiplying the optical depth $\tau(z)$ in the Ly$\alpha$ forest by $1+\omega(z)$. For the UVB intensity we used the modal value of the PESD, $\log J_{\nu_{0}}=-21.51$. Despite the high S/N of our spectra, this procedure required particular attention for transmission points falling below zero and above unity due to noise and continuum uncertainties. The negative points belong to saturated lines which will become even more saturated after the inverse correction for the proximity effect; such points were left uncorrected. Pixels with transmission values $T>1$ were also disregarded in the correction in order to prevent numerical problems or an artificially increased S/N level. Both cases apply only to a small number of pixels, and their in- or exclusion does not affect the results. Figure 14: Dependence of the overdensity degree on quasar luminosity (panel [a]), redshift (panel [b]), and on the proximity effect strength (panel [c]). The uncertainties relative to each measurements are represented by the solid cross in the respective panel. The solid line in the right-hand panels shows the least-square linear fit to the data. In each panel the dashed horizontal lines give an estimate of the amount of Poissonian shot noise in the simulations of the forest for $R_{0}=3$ Mpc from the emission redshift, as $\pm 1\sigma$ envelopes around the expected $\tau_{\mathrm{eff}}$ in the case of no overdensity ($\Xi=1$). 2. 2. From the corrected $\tau^{}(z)$ we computed the effective optical depth in bins along the line of sight and divided it by the expected values at the appropriate values of $z$ (given by eq. 1). This procedure had to be carried out for each object separately. We define this ratio of effective optical depths as the _overdensity degree_ $\Xi$. We performed this calculation for different bin sizes, corresponding to increments of 2, 3, 5 and 8 proper Mpc. Qualitatively, the results are the same, with a bin size of 3 Mpc showing the clearest trends. We only quote results for that bin size in the following. 3. 3. We then merged all lines of sight, resulting in 40 estimates of $\Xi$ per distance bin. In each bin we estimated the mean and the standard deviation of $\Xi$. We used our Monte-Carlo simulated spectra to reproduce our QSO sample and thereby estimated the amount of variance in $\Xi$ attributed to pure Poisson noise. ### 8.2 Overdensity distribution The dependence of the average $\Xi$ as a function of distance is presented in Fig. 13 (solid line). At large distances ($>5$ Mpc), $\Xi$ fluctuates around unity as predicted by perfect Poissonian noise. Only in the innermost bin for $\la 3$ Mpc does the mean value of $\Xi$ increase towards an overdensity degree of a factor of 2 (in units of optical depth). So far this can be seen as supporting the notion of quasars _typically_ residing in overdense regions (e.g. D’Odorico et al. 2008). However, in the same panel we also show the average $\Xi$ if the 10 objects (25 % of the sample) with the highest individual overdensities are removed (see below for details of this deselection). For all bins beyond $\sim 5$ Mpc, the run of $\Xi$ with distance is statistically indistinguishable for the two samples. Only for the nearest two bins the two profiles deviate, in the sense that the dotted profile (based on 75 % of the sample) is fully consistent with no overdensities at all. Figure 14 shows the individually measured ‘overdensity degrees’ within the innermost 3 Mpc bin as a function of quasar luminosity at the Lyman limit, quasar redshift, and proximity effect strength $\log a$ (cf. Sect. 7.1), respectively. This diagram reveals that the corrected effective optical depths for individual quasars range from close to zero to up to 6 times larger than in the mean Ly$\alpha$ forest. The horizontal dashed lines again indicate the amount of scatter ($\pm 1\sigma$) expected for pure Poissonian variance within the same physical scale around each of the QSOs. Panel [a] shows that there is no systematic tendency of $\Xi$ to vary with the quasar Lyman limit luminosity. The spread appears to be somewhat larger at high luminosities, which may reflect the expected trend of luminous quasars to reside in more massive environments. But there are also several high-$L$ objects with $\Xi$ values around unity. Regarding the redshift dependence depicted in Panel [b] of Fig. 14, no significant trend can be seen here either. There may be a deficit of high overdensities for $z>3$, but this could equally be an effect of small number statistics. However, panel [c] of Fig. 14 reveals some dependence between $\Xi$ and proximity effect strength $\log(a)$. The four QSOs with $\Xi>3$ are among the six objects with the weakest proximity effect (highest values of $\log a$). A quantitative test using the standard linear correlation coefficient $r$ shows that $\log\Xi$ is indeed correlated with $\log a$, $r=0.6$ (with a probability of $<1$ % of no correlation), but that this trend is strongly driven by the few points with $\Xi>3$. Without these points, the correlation coefficient reduces to $r=0.4$, and there is a 3 % probability of no intrinsic correlation. ### 8.3 Effects of overdensities on the proximity effect strength distribution We now investigate how the presence of some fraction of the QSOs residing in dense environments changes the properties of the PESD. We generated a new sample of synthetic spectra at $z=$ 2, 3 and 4, respectively, at a fixed luminosity of $\log(L_{\nu_{0}}/\mathrm{erg\ s^{-1}\ Hz^{-1}})=31.5$, with the difference that we now required that our sight lines show excess absorption near the emission redshift. We did not assume any particular radial distribution for the overdensity profile, but simply continued to populate the spectrum within 3 Mpc with lines until the effective optical depth reached a given threshold. We constructed two samples, for overdensities of $\Xi=2$ and 4, respectively, to approximately mimic our observational results (500 spectra for every $\Xi$). These simulations show that if all quasars had the same enhanced expectation value $\left<\Xi\right>$, the _shape_ of the PESD would not be greatly affected for all studied redshifts. The main effect would be a global shift of the PESD towards higher $\log a$ values, with the shift increasing as $\left<\Xi\right>$ becomes larger. This is shown in the top panel of Fig. 15, where we plotted the PESD for three expectation values of $\left<\Xi\right>=1$, 2, and 4, respectively. Already for an average enhancement of $\left<\Xi\right>=2$, the peak of the PESD is shifted by $\sim 0.6$ dex. It follows that if quasar environments are systematically overdense by some uniform factor, we would be unable to tell this from our data alone: Our empirical determination of $\Xi$ depends on assuming the correct global $J_{\nu_{0}}$. If the mode of the PESD is biased by some factor, we would _underestimate_ the $\Xi$ values of the individual quasars by the same factor, effectively renormalising the _observed_ $\Xi$ scale to an expectation value of unity. Of course, this is nothing but yet another manifestation of the well-known degeneracy between the UVB and overdensities around quasars. However, such a simple picture of all quasars having the same density enhancement factor would be highly unrealistic. If quasars reside in overdensities, the enhancement will have a distribution stretching over at least a factor of few. Because the shift of the PESD depends on $\left<\Xi\right>$, this will lead to a broadening of the PESD by an amount depending on the physical spread of overdensities. Notice that this spread will come on top of the inevitable scatter due to Poissonian variance. As an example toy model, we show in the bottom panel of Fig. 15 the modified PESD for an assumed physical rms scatter in $\left<\Xi\right>$ of 0.25 dex. The predicted distribution is considerably different; it is much broader and more box-shaped, and certainly much less in agreement with the observed PESD than the one for $\left<\Xi\right>=1$. While the assumption of a scatter of 0.25 dex is entirely ad-hoc, the example shows that the PESD reacts quite sensitively, and a very small physical dispersion around the expectation value of $\Xi$ is required to make observed and predicted PESD consistent. This is naturally obtained only if the expectation value for most quasars is close to $\left<\Xi\right>\approx 1$. These considerations strongly support the notion that there is no strong _systematic_ overdensity bias for the proximity effect. While the excellent agreement between our modal estimate and other ways of determining the UVB intensity (Fig. 12) is already highly suggestive, we cannot of course use this agreement both to justify our method _and_ regard it as an indepenent measurement of the UVB. The comparison of expected and observed PESD shapes adds independent information from the distribution properties and breaks the UVB-overdensity degeneracy. Clearly some fraction of objects sits in overdense regions. This, however, only produces a tail to the PESD by adding some high values of $\log a$. A comparison of Fig. 14c with Fig. 10 shows that the objects with strong overdensities are exactly those lying outside the prediction for the simple Poissonian model. We note in passing that there is no reason to interpret the few quasars showing an extremely strong proximity effect as sitting in ‘underdensities’. The three objects with the lowest $\log a$ values have a mean $\Xi$ very close to unity (see Fig. 14c), and only one object is located outside of the $\pm 2\sigma$ envelope of $\Xi$ expected for pure Poissonian fluctuations around the mean density. Figure 15: Top panel: Expected PESD computed from Monte-Carlo simulations for different expectation values of the overdensity degree ($\Xi=1$, 2, 4, respectively). The main effect is a substantial shift towards an on average weaker proximity effect. Bottom panel: Comparison of the observed PESD and the expectation for an assumed Gaussian distribution of overdensities (see text for details). ### 8.4 A hybrid method of estimating the UV background In the following we adopt the notion that only a relatively small fraction of at most $\sim 25$ % of our quasars is affected by significant overdensities. If this is true, then for a completely unbiased estimation of the UVB we need to: (i) remove the quasars with biased environments; (ii) correct for the asymmetry bias in the proximity effect strength distribution. In order to conduct the first step we had to identify outliers in the $\Xi$ distribution. However, recall that $\Xi$ is based on the assumption that we already know the UVB background. As an initial guess we adopted the modal value from the PESD, from which we then could compute $\Xi$ for all quasars. There is no obvious cutoff, and many quasars have $\Xi$ values that are consistent with weak overdensities. We decided to make a conservative cut in the sense that we would rather avoid too many even mildly overdense objects. We set the threshold at $+1\sigma$ (predicted scatter for Poissonian variance), implying that quasars with $\Xi>\Xi_{0}=1.8$ would be excluded from the analysis. This was the case for 10 objects, or 25 % of the sample. We then constructed the PESD based on the remaining 30 objects. The modal value remains unchanged, but the uncertainty obviously increases (0.2 rather than 0.15 dex). The fact that the mode of the PESD is robust against the inclusion or exclusion of 25 % of the sample is certainly encouraging. It is also clear that it would not be possible to use a much smaller fraction of objects, as the uncertainties of estimating the mode would rapidly go up. Next, we determined the systematic offset of the ‘combined sample averaging’ estimator (used in Sect. 6) due to the asymmetry of the PESD. We proceeded as follows: We generated mock quasar samples from the Monte Carlo simulated spectra, assuming no systematic overdensity bias, with the same redshift and luminosity distributions as the observed sample. We removed those objects showing apparent overdensities (occurring because of statistical noise) and determined the best-fit UVB value using the standard sample combination method as described in Sect. 6. Repeating this step 500 times, we averaged over the differences to the input UVB, obtaining a mean offset of $\Delta\log a=0.27$ dex. Finally, we conducted the same procedure for the observed sample with $\Xi<1.8$, performing a least-squares fit of Eq. 7 to the data and correcting the resulting value by the above offset. The inferred value is $\log J_{\nu_{0}}=-21.46^{+0.14}_{-0.21}$, which we consider as the best and most robust estimate of the global UV background intensity for our sample. In order to find out how much the method depends on the adopted criterion to eliminiate outliers, we repeated the described procedure once more, but now setting the threshold at $+2\sigma$, corresponding to $\Xi_{0}=3.1$. This implies that only the four objects with the highest $\Xi$ values were excluded from our sample. We redetermined the mean offset, which remained unchanged. The least-squares fit of Eq. 7 to the data, corrected for the relative offset, yields again a value of $\log J_{\nu_{0}}=-21.49^{+0.14}_{-0.21}$, demonstrating that our hybrid method is highly robust and quite insensitive to the outlier criterion. Determining the UVB intensity using this hybrid method has one important advantage over the modal PESD estimator: It depends less critically on the sample size. The mode is a robust statistical quantity only if the histogram from which it is estimated is reasonably well defined. In the next section we will split the sample into subsets to investigate the redshift evolution of the UVB. It will be seen that the PESD of such subsamples can become rather broad, inhibiting any confident mode estimation. The hybrid technique, on the other hand, involves straightforward averaging and therefore also works for smaller samples. Table 4 summarises all our different methods of estimating the UVB intensity and the resulting values. Table 4: Summary of our UV background measurements and employed methods, for the sample median redshift $z=2.73^{\dagger}$. Method \| $\log J_{\nu_{0}}$ ---\|--- fit to combined sample \| $-21.10^{+0.14}_{-0.22}$ mode of PESD \| $\mathbf{-21.51\pm 0.15}$ mean of individual $\log a$ \| $-21.24\pm 0.17$ fit to combined subsample ($\Xi<1.8$) \| $-21.19^{+0.14}_{-0.21}$ fit to combined subsample ($\mathbf{\Xi<1.8}$,corrected) \| $\mathbf{-21.46^{+0.14}_{-0.21}}$ * $\dagger$: The two boldfaced rows represent the unbiased estimates. ## 9 The redshift evolution of the UV background The high S/N of our data and the size and redshift range of the sample prompted us to explore whether we could split the sample into subsets of different redshifts and still get meaningful constraints on the UVB. As a first attempt we divided the sample into two subsets with $z<3$ and $z>3$. For each subset we constructed the PESD separately, shown in Fig. 16. The distributions look substantially different from each other. The PESD of the lower redshift subset is very broad, and no clear peak can be recognised. On the other hand, the high redshift PESD is significantly narrower and displays a well-defined maximum. The overall behaviour is qualitatively consistent with the predictions of Fig. 9 based on our Monte Carlo simulations, where we found that the widths of the PESD should decrease as redshift increases (Sect. 7.2). The modal value obtained from the high redshift sample is $\log J_{\nu_{0}}\simeq-21.6$, slightly lower than the overall best estimate. In the lower redshift subset, even a rough guess of the distribution mode is impossible. Figure 16: Observed PESD for our sample of quasars divided into two redshift intervals. The shapes of the two PESDs differ substantially from each other as predicted in Sect. 7.2. Table 5: UV background intensity as function of redshift for the set of quasars that show overdensity degree $\Xi\leq 1.8$ as plotted in Fig. 17. $<z>$ \| $\log J_{\nu_{0}}$ \| $\sigma_{\log J_{\nu_{0}}}$ ---\|---\|--- 2.10 \| -21.15 \| 0.20 2.30 \| -21.89 \| 0.20 2.50 \| -21.20 \| 0.19 2.70 \| -21.62 \| 0.19 2.90 \| -21.46 \| 0.20 3.10 \| -21.50 \| 0.20 3.30 \| -21.91 \| 0.20 3.55 \| -21.61 \| 0.20 3.80 \| -21.71 \| 0.21 Figure 17: Evolution of the mean intensity of the UV background with redshift from those quasars with overdensity degree $\Xi<1.8$ (see text for details). We estimated the UVB using the combined analysis of the proximity effect in $\Delta z=0.2$ intervals over the range $2\lesssim z\lesssim 4$. _Top panel_ : $\log J_{\nu_{0}}(z)$ in comparison to a constant evolutionary model (dashed) and a linear fit to the data (solid). _Bottom panel_ : $\log J_{\nu_{0}}(z)$ in comparison to the model by Haardt & Madau (2001) and the more recent determination of the UVB evolution from Bolton et al. (2005). The errors in the $z$ direction represent the range used to measure the average UVB. A more powerful way of estimating the UVB intensity is the hybrid method described in the previous subsection. We now employed this method to investigate the redshift evolution of the UVB. After merging all quasar spectra in the same way as for measuring the evolution of the Ly$\alpha$ effective optical depth (Sect. 4) we sampled the Ly$\alpha$ forest in redshift bins of $\Delta z=0.2$. We used only the forest in the redshift range $2\la z\la 4$ in order not to be affected by small number statistics. For every bin we fitted Eq. 7 to the data and obtained a preliminary estimate of $J_{\nu_{0}}$. These values are still systematically high, both as a result of the PESD asymmetry and because of the influence of overdense lines of sight. In particular, the redshift ranges at $z\sim 2.2$ and at $z\sim 2.9$ are heavily affected by the latter effect, as at these redshifts we find the majority of objects with a strongly overdense environment (cf. Fig. 14). In such narrow redshift bins, the values of the UVB are extremely sensitive to locally enhanced absorption, causing deviations of up to 1.3 dex. We now followed the same recipe as in Sect. 8.4 above: We removed from the sample those objects with $\Xi>1.8$ within the innermost 3 Mpc. We then applied the combined average estimator for each redshift interval and applied the PESD asymmetry bias determined from the simulations. We obtained in total eight values of $J_{\nu_{0}}$, for eight different redshift bins as listed in Table 5. These result are presented as black points in Fig. 17, where the horizontal bars indicate the redshift interval and the vertical bars are the statistical errors. The top panel of Fig. 17 shows that our measurements do suggest a certain evolution of the UVB intensity with redshift. Fitting a simple least-squares linear relation to the data points we get $\log J_{\nu_{0}}(z)=(-0.20\pm 0.14)\ z+(-21.0\pm 0.4)\>.$ (8) This fit (depicted as the solid line in the top panel of Fig. 17) yields residuals fully consistent with random errors. However, the slope is not very steep, and in view of the size of the error bars it is appropriate to ask whether evolution is actually _demanded_ by the data. For comparison we considered also the null hypothesis of no evolution, which is represented by the horizontal dashed line in Fig. 17). The quality of that model is somewhat poorer, but a KS test gives still 10 % acceptance probability, which is at best a marginal rejection. We can therefore only tentatively claim to detect the UVB evolution within this redshift interval. In a different take, we can confidently exclude a very steep redshift evolution, such as would be predicted from a UV background made predominantly of quasar light. How do these results compare with other works aimed at estimating or predicting the UVB evolution at similar redshifts? We first consider the prediction by Haardt & Madau (2001). They constrained $J_{\nu_{0}}(z)$ by integrating the contributions of the observed quasar and young star-forming galaxy populations, accounting also for intergalactic absorption and re- emission. Their predicted redshift evolution of the UVB intensity is depicted as the solid green line in the lower panel of Fig. 17). The agreement with our data points could hardly be better; in particular, the shallow slope is almost exactly reproduced. We roughly estimate a slope of $\mathrm{d}J/\mathrm{d}z=-0.16$ for the Haardt & Madau model at $z\ga 2.2$, which is extremely close to our best-fit value of $-0.2$. Nevertheless it should be emphasised that the galaxy contribution to the UVB crucially depends on the poorly constrained escape fraction of Lyman continuum photons from galaxies (e.g. Shapley et al. 2006; Gnedin et al. 2007). Recently, several studies calibrated the outcome of hydrodynamical IGM simulations through the measured Ly$\alpha$ forest opacity to estimate the photoionisation history of intergalactic hydrogen (e.g., Bolton et al. 2005; Faucher-Giguère et al. 2008a). These studies typically gave an approximatly constant UVB intensity within the range $2<z<4$. As an example we show the results of Bolton et al. (2005) by the crosses in the lower panel of Fig. 17). As stated above, a no-evolution scenario as found by these studies is also consistent with the errors, and our results are also consistent with those of Bolton et al. (2005) and Faucher-Giguère et al. (2008a). At any rate, our measurements confirm the prevailing assumption that quasars alone are unable to keep the IGM at a highly ionised state at redshifts larger than three. The integrated UV emissivity of AGN has a maximum very close to $z=2$ and drops by about an order of magnitude towards $z=4$ (Wolf et al. 2003; Hopkins et al. 2007), whereas we find the UVB intensity to fall by only $\sim 0.2$ dex (2$\sigma$ limit is 0.48 dex). This strengthens the notion for a non-negligible contribution by star-forming galaxies to the UV background radiation field; their fractional contribution increases when going to higher redshifts. We can make a simple back-of-an-envelope estimate of the mixing ratio between quasars and stars by assuming that the absolute UV emissivity of star-forming galaxies is roughly constant with redshift between $z=2$ and $z=4$. Since the integrated quasar emissivity at $z=4$ falls to $\sim 10$ % of its $z=2$ value (Hopkins et al. 2007), and using our best-fit measurement of the overall UV background evolution (adopting a factor 0.4 between $z=2$ and $z=4$), we obtain that the H0 photoionisation rate at $z=2$ is dominated by quasars by a factor of $\sim 2$ over star forming galaxies. At $z=4$, galaxies would then dominate by a factor of 5 over quasars. Any change towards making the total UVB evolution more constant with redshift would increase the fraction of starlight in the UVB. ## 10 Conclusions We have analysed the largest sample of high-resolution quasar spectra presented in a single study to date. We demonstrated that the line-of-sight proximity effect is a universal phenomenon that can be found in the spectrum of essentially every individual high-redshift quasar. While in Paper I we arrived at the same conclusion using low-resolution spectra, we now can significantly reinforce this claim. In particular, continuum placing uncertainties, which were a bit of an issue in Paper I, play absolutely no role now, given the high spectral resolution and high S/N data exploited in this paper. Our study rehabilitates the proximity effect as a tool to investigate the intensity at 1 Ryd (more accurately: the H0 photoionisation rate) of the cosmic ultraviolet background. Previous investigations using the proximity effect nearly always produced UVB estimates that were suspiciously high and indeed at variance with other means to estimate or predict the UVB. It has been suggested in the literature that quasars typically reside in denser than average large-scale environments, and that therefore excess absorption biases the proximity effect to appear too weak, resulting in an overestimated UVB. We argue that this is not the main reason for the discrepancies between the proximity effect and other methods. The actually measured strength of the proximity effect in a single quasar line of sight depends on the UVB, but also on the number of absorption lines present in the ‘proximity effect zone’. That number is generally small, and to first approximation the presence or absence of a line at a given redshift can be described as a Poissonian random process. We have demonstrated that the distribution of the resulting ‘proximity effect strength’ parameter (which directly relates to the UVB) is significantly skewed, even without invoking any physical overdensities. Any direct averaging over a sample of sightlines will inevitably bias the resulting UVB. Only by looking at individual lines of sight separately can this effect be detected and removed. We proposed and used two different methods of estimating the UVB intensity in the presence of this bias. In the first method, the UVB is taken directly from the mode of the proximity effect strength distribution (PESD). This method is simple and robust against a small number of outliers; on the other hand it is limited to large samples and depends on the condition that the mode is not affected by the outliers. Our second method involves Monte Carlo simulations to calibrate the asymmetry bias of the PESD and then uses straightforward sample averaging, possibly combined with a prior rejection of outliers, and the result is then corrected for the bias. This ‘hybrid’ method is quite stable and works also for smaller samples; it is, however, more complicated to use and depends on a successful removal of outliers. Our best estimates of the UVB intensity using these two methods give very similar results. The modal estimate is $\log J_{\nu_{0}}=-21.51\pm 0.15$, and the hybrid estimator gives $\log J_{\nu_{0}}=-21.46^{+0.14}_{-0.21}$, which is statistically indistinguishable. This value is in excellent agreement with other methods to estimate or predict the UVB. If the UV background is known, the data used for the proximity effect can also be employed to reconstruct the absorption pattern without the radiative influence of the quasar. Doing this for our sample of 40 quasars we found that only 10 % of them showed very significant excess absorption (by more than a factor of $\sim 3$) while most are distributed in a way more or less consistent with Poissonian fluctuations around the cosmic mean absorption expected at these redshifts. Given the widespread notion of quasars residing in massive galaxies, which in turn should sit in massive haloes with significantly clustered environments, this sounds surprising. We have no intention of questioning that picture, and we stress that our measurements constrain exclusively the neutral hydrogen distribution averaged on scales of $\sim 3$ proper Mpc towards the quasar. Nevertheless, our results may bear some relevance to the question of what are the typical halo masses for quasars. Most of the QSOs in our sample are optically selected and therefore presumably radio-quiet, which could mean that their haloes are not exceptionally massive. In a recent paper, Mandelbaum et al. (2008) estimated halo masses of AGN in the Sloan survey from clustering and galaxy-galaxy lensing, finding that optically selected AGN follow the same relation between stellar and halo mass as normal galaxies, whereas radio-loud AGN have much higher halo masses. (Incidentally, the outstanding object in our sample in terms of overdensity is the radio quasar PKS 0237$-$233; but the next-ranked objects are all optically selected.) Our argument against significant overdensities of quasars and in favour of the applicability of the proximity effect to measure the UV background gains weight by the excellent agreement between our estimates of the UVB intensity and those obtained by completely different methods. This notion is supported independently by considering the _shape_ of the PESD. The observed distribution width clearly favours average densities close to unity; if significant overdensities were involved, a much broader distribution would be expected than the one observed. This additional constraint breaks the degeneracy between UV background and overdensities that has haunted the field for so long. Finally, we also attempted to constrain the redshift evolution of the UVB intensity. Although the results are still quite uncertain, we could for the first time from a single quasar sample derive useful limits on the amount of evolution. We can rule out at high significance that quasar light dominates the UVB at redshifts $z\ga 4$. Our best fit suggests a mild decrease in the UVB intensity towards higher redshifts, and the derived slope is in astonishingly good agreement with the predictions by Haardt & Madau (2001). Nevertheless, an approximately constant UVB as found by Bolton et al. (2005) and others is also formally consistent with our data. We have shown that the proximity effect holds the potential to derive important cosmological quantities. While our sample is large in comparison to those of previous studies using high-resolution spectra, it still suffers in several aspects from small number statistics. As the public archives are growing, substantial progress can be expected from applying the proximity effect analysis to further spectra, in particular at higher redshift. ###### Acknowledgements. We would like to thank ESO for making the ESO data archive publicly available. We also thank Tae$-$Sun Kim for assisting us in the data reduction process. A.D. and G.W. were partly supported by a HWP grant from the International Helmholtz Institute for Supercomputational Physics. We finally acknowledge support by the Deutsche Forschungsgemeinschaft under Wi 1369/21-1. ## References * Agafonova et al. (2005) Agafonova, I. I., Centurión, M., Levshakov, S. A., & Molaro, P. 2005, A&A, 441, 9 * Bajtlik et al. (1988) Bajtlik, S., Duncan, R. C., & Ostriker, J. P. 1988, ApJ, 327, 570 * Bechtold (1994) Bechtold, J. 1994, ApJS, 91, 1 * Bechtold et al. (1987) Bechtold, J., Weymann, R. J., Lin, Z., & Malkan, M. A. 1987, ApJ, 315, 180 * Bernardi et al. (2003) Bernardi, M., Sheth, R. K., SubbaRao, M., et al. 2003, AJ, 125, 32 * Bolton et al. (2005) Bolton, J. S., Haehnelt, M. G., Viel, M., & Springel, V. 2005, MNRAS, 357, 1178 * Carswell et al. (1987) Carswell, R. F., Webb, J. K., Baldwin, J. A., & Atwood, B. 1987, ApJ, 319, 709 * Carswell et al. (1982) Carswell, R. F., Whelan, J. A. J., Smith, M. G., Boksenberg, A., & Tytler, D. 1982, MNRAS, 198, 91 * Cooke et al. (1997) Cooke, A. J., Espey, B., & Carswell, R. F. 1997, MNRAS, 284, 552 * Cristiani et al. (1995) Cristiani, S., D’Odorico, S., Fontana, A., Giallongo, E., & Savaglio, S. 1995, MNRAS, 273, 1016 * Dall’Aglio et al. (2008) Dall’Aglio, A., Wisotzki, L., & Worseck, G. 2008, A&A, 480, 359 * Davé et al. (1999) Davé, R., Hernquist, L., Katz, N., & Weinberg, D. H. 1999, ApJ, 511, 521 * Dekker et al. (2000) Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B., & Kotzlowski, H. 2000, in Presented at the Society of Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol. 4008, Proc. SPIE Vol. 4008, p. 534-545, Optical and IR Telescope Instrumentation and Detectors, Masanori Iye; Alan F. Moorwood; Eds., ed. M. Iye & A. F. Moorwood, 534–545 * D’Odorico et al. (2008) D’Odorico, V., Bruscoli, M., Saitta, F., et al. 2008, ArXiv e-prints, 0806.3075 * Ellison et al. (2005) Ellison, S. L., Hall, P. B., & Lira, P. 2005, AJ, 130, 1345 * Espey (1993) Espey, B. R. 1993, ApJ, 411, L59 * Fardal et al. (1998) Fardal, M. A., Giroux, M. L., & Shull, J. M. 1998, AJ, 115, 2206 * Faucher-Giguère et al. (2008a) Faucher-Giguère, C.-A., Lidz, A., Hernquist, L., & Zaldarriaga, M. 2008a, ApJ, 682, L9 * Faucher-Giguère et al. (2008b) Faucher-Giguère, C.-A., Lidz, A., Zaldarriaga, M., & Hernquist, L. 2008b, ApJ, 673, 39 * Faucher-Giguère et al. (2008c) Faucher-Giguère, C.-A., Prochaska, J. X., Lidz, A., Hernquist, L., & Zaldarriaga, M. 2008c, ApJ, 681, 831 * Fernández-Soto et al. (1995) Fernández-Soto, A., Barcons, X., Carballo, R., & Webb, J. K. 1995, MNRAS, 277, 235 * Gaskell (1982) Gaskell, C. M. 1982, ApJ, 263, 79 * Giallongo et al. (1996) Giallongo, E., Christiani, S., D’Odorico, S., Fontana, A., & Savaglio, S. 1996, ApJ, 466, 46 * Giallongo et al. (1993) Giallongo, E., Christiani, S., Fontana, A., & Trèvese, D. 1993, ApJ, 416, 137 * Gnedin et al. (2007) Gnedin, N. Y., Kravtsov, A. V., & Chen, H.-W. 2007, ArXiv e-prints, 707 * Guimarães et al. (2007) Guimarães, R., Petitjean, P., Rollinde, E., et al. 2007, MNRAS, 377, 657 * Haardt & Madau (1996) Haardt, F. & Madau, P. 1996, ApJ, 461, 20 * Haardt & Madau (2001) Haardt, F. & Madau, P. 2001, in Clusters of Galaxies and the High Redshift Universe Observed in X-rays, ed. D. M. Neumann & J. T. V. Tran * Hopkins et al. (2007) Hopkins, P. F., Richards, G. T., & Hernquist, L. 2007, ApJ, 654, 731 * Hu et al. (1995) Hu, E. M., Kim, T.-S., Cowie, L. L., Songaila, A., & Rauch, M. 1995, AJ, 110, 1526 * Hunt et al. (2004) Hunt, M. P., Steidel, C. C., Adelberger, K. L., & Shapley, A. E. 2004, ApJ, 605, 625 * Jena et al. (2005) Jena, T., Norman, M. L., Tytler, D., et al. 2005, MNRAS, 361, 70 * Kim et al. (2007) Kim, T.-S., Bolton, J. S., Viel, M., Haehnelt, M. G., & Carswell, R. F. 2007, MNRAS, 382, 1657 * Kim et al. (2002) Kim, T.-S., Carswell, R. F., Cristiani, S., D’Odorico, S., & Giallongo, E. 2002, MNRAS, 335, 555 * Kim et al. (2001) Kim, T.-S., Cristiani, S., & D’Odorico, S. 2001, A&A, 373, 757 * Kim et al. (2004) Kim, T.-S., Viel, M., Haehnelt, M. G., Carswell, R. F., & Cristiani, S. 2004, MNRAS, 347, 355 * Kirkman et al. (2005) Kirkman, D., Tytler, D., Suzuki, N., et al. 2005, MNRAS, 360, 1373 * Kulkarni & Fall (1993) Kulkarni, V. P. & Fall, S. M. 1993, ApJ, 413, L63 * Liske & Williger (2001) Liske, J. & Williger, G. M. 2001, MNRAS, 328, 653 * Loeb & Eisenstein (1995) Loeb, A. & Eisenstein, D. J. 1995, ApJ, 448, 17 * Lu et al. (1996) Lu, L., Sargent, W. L. W., Womble, D. S., & Takada-Hidai, M. 1996, ApJ, 472, 509 * Lu et al. (1991) Lu, L., Wolfe, A. M., & Turnshek, D. A. 1991, ApJ, 367, 19 * Madau et al. (1999) Madau, P., Haardt, F., & Rees, M. J. 1999, ApJ, 514, 648 * Mandelbaum et al. (2008) Mandelbaum, R., Li, C., Kauffmann, G., & White, S. D. M. 2008, ArXiv e-prints:astro-ph/0806.4089 * McDonald et al. (2000) McDonald, P., Miralda-Escudé, J., Rauch, M., et al. 2000, ApJ, 543, 1 * McDonald et al. (2005) McDonald, P., Seljak, U., Cen, R., et al. 2005, ApJ, 635, 761 * Meiksin & White (2004) Meiksin, A. & White, M. 2004, MNRAS, 350, 1107 * Morton (2003) Morton, D. C. 2003, ApJS, 149, 205 * Murdoch et al. (1986) Murdoch, H. S., Hunstead, R. W., Pettini, M., & Blades, J. C. 1986, ApJ, 309, 19 * Rauch (1998) Rauch, M. 1998, ARA&A, 36, 267 * Rauch et al. (1997) Rauch, M., Miralda-Escude, J., Sargent, W. L. W., et al. 1997, ApJ, 489, 7 * Richards et al. (2002) Richards, G. T., Vanden Berk, D. E., Reichard, T. A., et al. 2002, AJ, 124, 1 * Rollinde et al. (2005) Rollinde, E., Srianand, R., Theuns, T., Petitjean, P., & Chand, H. 2005, MNRAS, 361, 1015 * Sargent et al. (1980) Sargent, W. L. W., Young, P. J., Boksenberg, A., & Tytler, D. 1980, ApJS, 42, 41 * Savaglio et al. (1997) Savaglio, S., Christiani, S., D’Odorico, S., et al. 1997, A&A, 318, 347 * Schaye et al. (2003) Schaye, J., Aguirre, A., Kim, T.-S., et al. 2003, ApJ, 596, 768 * Schirber & Bullock (2003) Schirber, M. & Bullock, J. S. 2003, ApJ, 584, 110 * Schirber et al. (2004) Schirber, M., Miralda-Escudé, J., & McDonald, P. 2004, ApJ, 610, 105 * Scott et al. (2000) Scott, J., Bechtold, J., Dobrzycki, A., & Kulkarni, V. P. 2000, ApJ, 130, 67 * Scott et al. (2002) Scott, J., Bechtold, J., Morita, M., Dobrzycki, A., & Kulkarni, V. P. 2002, ApJ, 571, 665 * Shapley et al. (2006) Shapley, A. E., Steidel, C. C., Pettini, M., Adelberger, K. L., & Erb, D. K. 2006, ApJ, 651, 688 * Sokasian et al. (2003) Sokasian, A., Abel, T., & Hernquist, L. 2003, MNRAS, 340, 473 * Srianand & Khare (1996) Srianand, R. & Khare, P. 1996, MNRAS, 280, 767 * Suzuki (2006) Suzuki, N. 2006, ApJS, 163, 110 * Telfer et al. (2002) Telfer, R. C., Zheng, W., Kriss, G. A., & Davidsen, A. F. 2002, ApJ, 565, 773 * Theuns et al. (1998) Theuns, T., Leonard, A., Efstathiou, G., Pearce, F. R., & Thomas, P. A. 1998, MNRAS, 301, 478 * Tytler & Fan (1992) Tytler, D. & Fan, X.-M. 1992, ApJS, 79, 1 * Tytler et al. (2004) Tytler, D., Kirkman, D., O’Meara, J. M., et al. 2004, ApJ, 617, 1 * Vanden Berk et al. (2001) Vanden Berk, D. E., Richards, G. T., Bauer, A., et al. 2001, AJ, 122, 549 * Véron-Cetty & Véron (2006) Véron-Cetty, M.-P. & Véron, P. 2006, A&A, 455, 773 * Weymann et al. (1981) Weymann, R. J., Carswell, R. F., & Smith, M. G. 1981, ARA&A, 19, 41 * Williger et al. (1994) Williger, G. M., Baldwin, J. A., Carswell, R. F., et al. 1994, ApJ, 428, 574 * Wolf et al. (2003) Wolf, C., Wisotzki, L., Borch, A., et al. 2003, A&A, 408, 499 * Young et al. (1979) Young, P. J., Sargent, W. L. W., Boksenberg, A., Carswell, R. F., & Whelan, J. A. J. 1979, ApJ, 229, 891 * Zuo (1993) Zuo, L. 1993, A&A, 278, 343 ## Appendix A Proximity effect on individual lines of sight Figure 18: The proximity effect signatures in individual lines of sight. Each panel shows the normalised optical depth $\xi$ versus $\omega$, binned in steps of $\Delta\log\omega=1$, with the best-fit model of the combined analysis superimposed as dotted red lines. The solid lines delineate the best fit to each individual QSO as described in the text. Figure 18: Continued	arxiv-papers	2008-07-31T15:37:18	2024-09-04T02:48:57.074781	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Aldo Dall'Aglio, Lutz Wisotzki, Gabor Worseck", "submitter": "Aldo Dall'Aglio", "url": "https://arxiv.org/abs/0807.5089" }
0808.0049	# A note on the invariant subspace problem relative to a type ${\rm II}_{1}$ factor Junsheng Fang Don Hadwin ###### Abstract Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor with a faithful normal tracial state $\tau$ and let $\mathcal{M}^{\omega}$ be the ultrapower algebra of $\mathcal{M}$. In this paper, we prove that for every operator $T\in\mathcal{M}^{\omega}$, there is a family of projections $\\{P_{t}\\}_{0\leq t\leq 1}$ in $\mathcal{M}^{\omega}$ such that $TP_{t}=P_{t}TP_{t}$, $P_{s}\leq P_{t}$ if $s\leq t$, and $\tau_{\omega}(P_{t})=t$. Let $\mathfrak{M}=\\{Z\in\mathcal{M}:\,\text{there is a family of projections}\,\\{P_{t}\\}_{0\leq t\leq 1}\newline \text{in}\,\mathcal{M}\,\text{such that}\,ZP_{t}=P_{t}ZP_{t},P_{s}\leq P_{t}\,\text{if}\,s\leq t,\,\text{and}\,\tau(P_{t})=t\\}$. As an application we show that for every operator $T\in\mathcal{M}$ and $\epsilon>0$, there is an operator $S\in\mathfrak{M}$ such that $\\|S\\|\leq\\|T\\|$ and $\\|S-T\\|_{2}<\epsilon$. We also show that $\prod_{n}^{\omega}M_{n}(\mathbb{C})$ is not $\ast$-isomorphic to the ultrapower algebra of the hyperfinite type ${\rm II}_{1}$ factor. Keywords: Invariant subspaces, type ${\rm II}_{1}$ factors, ultrapower algebras. MSC: 46L10, 47C15 ## 1 Introduction Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor acting on a Hilbert space $\mathcal{H}$. The invariant subspace problem relative to a factor von Neumann algebra $\mathcal{M}$ asks for every operator $T\in\mathcal{M}$, does there exists a projection $P\in\mathcal{M}$, $0<P<I$, such that $TP=PTP$. The hyperinvariant subspace problem relative to $\mathcal{M}$ asks for every operator $T\in\mathcal{M}\setminus\mathbb{C}I$, does there exists a projection $P$, $0<P<I$, such that $SP=PSP$ for every operator $S$ in $\mathcal{B}(\mathcal{H})$ with $ST=TS$. It is easy to see that if a projection $P$ is hyperinvariant for $T$, then $P$ is in the von Neumann algebra generated by $T$ and therefore in $\mathcal{M}$. A huge advance on the (hyper)invariant subspace problem relative to a factor of type ${\rm II}_{1}$ has been made during past ten years (see for example [2, 3, 6, 13]). In 1983, Brown [1] introduced a spectral distribution measure for non-normal elements in a finite von Neumann algebra with respect to a fixed normal faithful tracial state, which is called the Brown measure of the operator. Recently, Haagerup and Schultz [6] proved a remarkable result which states that if the support of Brown measure of an operator in a type ${\rm II}_{1}$ factor contains more than two points, then the operator has a non-trivial hyperinvariant subspace affiliated with the type ${\rm II}_{1}$ factor. However, the invariant subspace problem relative to a type ${\rm II}_{1}$ factor still remains open for operators with single point Brown measure support (for this case, we refer to Dykema and Haagerup’s paper [2]). Suppose that each $\mathcal{M}_{n}$ is a finite von Neumann algebra with a faithful normal tracial state $\tau_{n}$. Let $\prod_{n\in\mathcal{N}}\mathcal{M}_{n}$ be the $l^{\infty}$-product of the $\mathcal{M}_{n}$’s. Then $\prod_{n}\mathcal{M}_{n}$ is a von Neumann algebra (with pointwise multiplication). Let $\omega$ be a free ultrafilter on $\mathcal{N}$ ($\omega$ may be viewed as an element in $\beta\mathcal{N}\setminus\mathcal{N}$, where $\beta\mathcal{N}$ is the Stone- Céch compactification of $\mathcal{N}$). If $\\{X_{n}\\}$ and $\\{Y_{n}\\}$ are two elements in $\prod_{n}\mathcal{M}_{n}$, then we define $\\{X_{n}\\}\sim\\{Y_{n}\\}$ when $\lim_{n\to\omega}\\|X_{n}-Y_{n}\\|_{2}=0$. Recall that for an operator $T_{n}\in\mathcal{M}_{n}$, $\\|T_{n}\\|_{2}=\tau_{n}(T_{n}^{}T_{n})^{1/2}$. Then the ultraproduct, denoted by $\prod^{\omega}\mathcal{M}_{n}$, of $\mathcal{M}_{n}$ (with respect to the free ultrafilter $\omega$) is the quotient von Neumann algebra of $\prod_{n}\mathcal{M}_{n}$ modulo the equivalence relation $\sim$ and the limit of $\tau_{n}$ at $\omega$ gives rise to a tracial state on $\prod^{\omega}\mathcal{M}_{n}$. We shall use $\tau_{\omega}$ to denote the tracial state on $\prod^{\omega}\mathcal{M}_{n}$. When $\mathcal{M}_{n}=\mathcal{M}$ for all $n$, then $\prod^{\omega}\mathcal{M}_{n}$ is called the ultrapower of $\mathcal{M}$, denoted by $\mathcal{M}^{\omega}$. The initial algebra $\mathcal{M}$ is embedded into $\mathcal{M}^{\omega}$ as constant sequences given by elements in $\mathcal{M}$. Ultrapowers for finite von Neumann algebras were first introduced and studied by McDuff [8]. Sakai [12] showed that an ultrapower of a finite von Neumann algebra with respect to a faithful normal trace is again a finite von Neumann algebra, and the ultrapower algebra $\mathcal{M}^{\omega}$ of a type ${\rm II}_{1}$ factor is also a type ${\rm II}_{1}$ factor. Ultrapowers of type ${\rm II}_{1}$ factors play an important role in the study of type ${\rm II}_{1}$ factors. This paper is organized as follows. In section 2 of this paper, we prove that every operator in an ultrapower algebra of a type ${\rm II}_{1}$ factor $\mathcal{M}$ has a nontrivial invariant space affiliated with the ultrapower algebra. Precisely, we prove that for every operator $T\in\mathcal{M}^{\omega}$, there is a family of projections $\\{P_{t}\\}_{0\leq t\leq 1}$ in $\mathcal{M}^{\omega}$ such that $TP_{t}=P_{t}TP_{t}$, $P_{s}\leq P_{t}$ if $s\leq t$, and $\tau_{\omega}(P_{t})=t$. This result is more or less trivial if $\mathcal{M}$ has property $\Gamma$. Recall that $\mathcal{M}$ is said to have property $\Gamma$ if for any finite elements $T_{1},\cdots,T_{n}$ in $\mathcal{M}$ and $\epsilon>0$, there is a unitary operator $U$ in $\mathcal{M}$ such that $\tau(U)=0$ and $\\|T_{i}U-UT_{i}\\|_{2}<\epsilon$ for $1\leq i\leq n$. If $\mathcal{M}$ is a separable (with separable predual) type ${\rm II}_{1}$ factor, then $\mathcal{M}$ has property $\Gamma$ if and only if $\mathcal{M}^{\prime}\cap\mathcal{M}^{\omega}$ is non-trivial. Dixmier [Di] proved that if $\mathcal{M}^{\prime}\cap\mathcal{M}^{\omega}$ is non-trivial, then it is non-atomic. This implies that if $\mathcal{M}$ has property $\Gamma$, then for every operator $T\in\mathcal{M}^{\omega}$, there is a family of projections $\\{P_{t}\\}_{0\leq t\leq 1}$ in $\mathcal{M}^{\omega}$ such that $TP_{t}=P_{t}T$, $P_{s}\leq P_{t}$ if $s\leq t$, and $\tau_{\omega}(P_{t})=t$. To prove the result for non-$\Gamma$ factors, we need combine techniques developed by Haagerup and Schultz [6] and a result of Popa [10]. As an application, in section 3 we show that for every operator $T$ in the unit ball of $\mathcal{M}$ and $\epsilon>0$, there is an operator $S\in\mathfrak{M}$ such that $\\|S\\|\leq 1$ and $\\|S-T\\|_{2}<\epsilon$, where $\mathfrak{M}=\\{Z\in\mathcal{M}:\,\text{there is a family of projections}\,\\{P_{t}\\}_{0\leq t\leq 1}\,\text{in}\,\mathcal{M}\,\text{such that}\,ZP_{t}=P_{t}ZP_{t},P_{s}\leq P_{t}\,\text{if}\,s\leq t,\,\text{and}\,\tau(P_{t})=t\\}$. In particular, this implies that $\mathfrak{M}$ is dense in $\mathcal{M}$ in the strong operator topology. In section 4, we give a very simple proof of $\prod_{n}^{\omega}M_{n}(\mathbb{C})$ is not $\ast$-isomorphic to the ultrapower algebra of the hyperfinite type ${\rm II}_{1}$ factor (this result might be known to specialists, however we can not find it in the existed literature). This result relies on a result of Herrero and Szarek [5] (also see [17]). Thanks to the existence of a faithful normal tracial state on a type ${\rm II}_{1}$ factor, in section 5 we show that if two operators $S$ and $T$ are quasi-similar in a type ${\rm II}_{1}$ factor $\mathcal{M}$, then $LatS\cap\mathcal{M}$ is not trivial if and only if $LatT\cap\mathcal{M}$ is not trivial. As a corollary, we show that for two operator $S,T$ in $\mathcal{M}$, $Lat(ST)\cap\mathcal{M}$ is not trivial if and only if $Lat(TS)\cap\mathcal{M}$ is not trivial. On the other hand, if the same result also holds for arbitrary two operators in $\mathcal{B}(\mathcal{H})$, then the answer to the classical invariant subspace problem is affirmative (see Remark 5.7). _Acknowledgment:_ The authors thank David Sherman for his comments on Lemma 4.2 and Theorem 4.3 and for poiting out to us von Neumann’s paper [17]. ## 2 Invariant subspaces for operators in the ultrapower algebras The main result of this section is the following result. ###### Theorem 2.1. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor and let $\mathcal{M}^{\omega}$ be the ultrapower algebra of $\mathcal{M}$. For every operator $T\in\mathcal{M}^{\omega}$, there is a family of projections $\\{P_{t}\\}_{0\leq t\leq 1}$ in $\mathcal{M}^{\omega}$ such that $TP_{t}=P_{t}TP_{t}$, $P_{s}\leq P_{t}$ if $s\leq t$, and $\tau_{\omega}(P_{t})=t$. ###### Corollary 2.2. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor with a faithful normal tracial state $\tau$. For every operator $T\in\mathcal{M}$ and $0\leq t\leq 1$, there is a sequence of projections $P_{n}\in\mathcal{M}$ such that $\lim_{n\rightarrow\infty}\\|TP_{n}-P_{n}TP_{n}\\|_{2}=0$ and $\tau(P_{n})=t$. To prove Theorem 2.1, we need the following lemmas. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor and let $T\in\mathcal{M}$. We regard $\mathcal{M}$ as a subfactor of $\mathcal{M}_{1}=\mathcal{M}\ast\mathcal{L}(\mathbb{F}_{4})$. The faithful normal tracial state on $\mathcal{M}_{1}$ will also be denoted by $\tau$. We choose a circular system $\\{x,y\\}$ (in the sense of [16]) that generates $\mathcal{L}(\mathbb{F}_{4})$ and which therefore is free from $\mathcal{M}$. By Theorem 5.2 of [7], the unbounded operator $z=xy^{-1}$ is in $L^{p}(\mathcal{M}_{1},\tau)$ for $0<p<1$. Let $T_{n}=T+\frac{1}{n}z$. Then $T_{n}\in L^{p}(\mathcal{M}_{1},\tau)$ for $0<p<1$. We will need the following lemma, which follows from Proposition 4.5, Corollary 4.6, Theorem 5.1 and Theorem 6.9 of [6]. ###### Lemma 2.3. With the above assumption, we have 1. 1. $\lim_{n\rightarrow\infty}\\|T-T_{n}\\|_{p}^{p}=0$; 2. 2. for every $n$, there is a projection $P_{n}\in\mathcal{M}_{1}$ such that $T_{n}P_{n}=P_{n}T_{n}P_{n}$ and $\tau(P_{n})=\frac{1}{2}$. The next lemma follows from the main theorem of [10]. ###### Lemma 2.4. Let $\mathcal{M}$ be a separable type ${\rm II}_{1}$ factor. Then there is a unitary operator $u\in\mathcal{M}^{\omega}$ such that $\\{\mathcal{M},u\mathcal{M}u^{}\\}^{\prime\prime}\cong\mathcal{M}(u\mathcal{M}u^{}).$ ###### Lemma 2.5. Let $\mathcal{M}$ be a separable type ${\rm II}_{1}$ factor and let $T\in\mathcal{M}$. Then for every $\epsilon>0$, there is a projection $P\in\mathcal{M}$, $\tau(P)=1/2$, such that $\\|TP-PTP\\|_{2}<\epsilon$. ###### Proof. Note that $\mathcal{M}$ is a von Neumann subalgebra of $\mathcal{M}^{\omega}$ if we identify $T\in\mathcal{M}$ with the constant sequence $(T)\in\mathcal{M}^{\omega}$. To prove the lemma, it is sufficient to show that there is a projection $P\in\mathcal{M}^{\omega}$, $\tau(P)=1/2$, such that $\\|TP-PTP\\|_{2}<\epsilon$. By Lemma 2.4, there is a unitary operator $u\in\mathcal{M}^{\omega}$ such that $\\{\mathcal{M},u\mathcal{M}u^{}\\}^{\prime\prime}\cong\mathcal{M}(u\mathcal{M}u^{})$. So it is sufficient to show that there is a projection $P\in\\{\mathcal{M},u\mathcal{M}u^{}\\}^{\prime\prime}$, $\tau(P)=1/2$, such that $\\|TP-PTP\\|_{2}<\epsilon$. Note that $T\in\mathcal{M}$ and therefore $T$ is free with $u\mathcal{M}u^{}$ in $\\{\mathcal{M},u\mathcal{M}u^{}\\}^{\prime\prime}$. Repeat the above arguments twice if necessary, we may assume that $\mathcal{M}\supseteq\mathcal{L}(\mathbb{F}_{4})$ and $T$ is free with $\mathcal{L}(\mathbb{F}_{4})$. We choose a circular system $\\{x,y\\}$ in $\mathcal{L}(\mathbb{F}_{4})$. Let $z=xy^{-1}$ and $T_{n}=T+\frac{1}{n}z$. By Lemma 4.2, for every $n\geq 1$, there is a projection $P_{n}\in\mathcal{M}$ with $\tau(P_{n})=1/2$ and $T_{n}P_{n}=P_{n}T_{n}P_{n}$. By Lemma 4.2, $\lim_{n\rightarrow\infty}\\|T_{n}-T\\|_{p}^{p}=0$ for $0<p<1$. Note that $\displaystyle\\|P_{n}TP_{n}-TP_{n}\\|_{2}^{2}$ $\displaystyle=$ $\displaystyle\tau(\|P_{n}TP_{n}-TP_{n}\|^{2})$ $\displaystyle=$ $\displaystyle\tau(\|P_{n}TP_{n}-TP_{n}\|^{p/2}\|P_{n}TP_{n}-TP_{n}\|^{2-p/2})$ $\displaystyle\leq$ $\displaystyle\tau(\|P_{n}TP_{n}-TP_{n}\|^{p})^{1/2}\tau(\|P_{n}TP_{n}-TP_{n}\|^{4-p})^{1/2}$ $\displaystyle=$ $\displaystyle\\|P_{n}TP_{n}-TP_{n}\\|_{p}^{p/2}\\|P_{n}TP_{n}-TP_{n}\\|_{4-p}^{(4-p)/2}$ $\displaystyle\leq$ $\displaystyle\left(\\|P_{n}TP_{n}-TP_{n}\\|_{p}^{p}\right)^{1/2}\\|2T\\|_{4-p}^{(4-p)/2},$ and $\displaystyle\\|P_{n}TP_{n}-TP_{n}\\|_{p}^{p}$ $\displaystyle\leq$ $\displaystyle\\|P_{n}(T-T_{n})P_{n}-(T-T_{n})P_{n}\\|_{p}^{p}$ $\displaystyle\leq$ $\displaystyle\\|P_{n}(T-T_{n})P_{n}\\|_{p}^{p}+(T-T_{n})P_{n}\\|_{p}^{p}$ $\displaystyle\leq$ $\displaystyle 2\\|T-T_{n}\\|_{p}^{p}\rightarrow 0.$ Therefore, $\lim_{n\rightarrow\infty}\\|P_{n}TP_{n}-TP_{n}\\|_{2}^{2}=0$. ∎ ###### Lemma 2.6. Let $\mathcal{M}$ be a separable type ${\rm II}_{1}$ factor, $T\in\mathcal{M}$ and $\epsilon>0$. For every positive integer $n$, there are projections $\\{P_{j}\\}_{j=0}^{2^{n}}$ in $\mathcal{M}$ such that $0=P_{0}<P_{1}<P_{2}<\cdots<P_{2^{n}-1}<P_{2^{n}}=I$, $\tau(P_{j})=j/2^{n}$, and $\\|TP_{j}-P_{j}TP_{j}\\|_{2}\leq\epsilon$ for all $0\leq j\leq 2^{n}$. ###### Proof. If $n=1$, then the lemma follows from Lemma 2.5. Suppose $n=2$. By Lemma 2.5, there are projections $P,Q$ in $\mathcal{M}$ such that $\tau(P)=\tau(Q)=1/2$, $P+Q=1$ and $\\|TP-PTP\\|_{2}<\epsilon/2$. Let $a=PTP$, $b=PTQ$, $c=QTP$, and $d=QTQ$. We can write $T=\left(\begin{array}[]{cc}a&b\\\ c&d\end{array}\right)$ with respect to the decomposition $I=P+Q$. Then $\\|c\\|_{2}<\epsilon/2$. Note that both $P\mathcal{M}P$ and $Q\mathcal{M}Q$ are type ${\rm II}_{1}$ factors. We apply Lemma 2.5 to $a\in P\mathcal{M}P$ and $b\in Q\mathcal{M}Q$, respectively. There are projections $P_{1}\leq P$, $Q_{1}\leq Q$ such that $\tau(P_{1})=\tau(Q_{1})=1/4$, $\\|aP_{1}-P_{1}aP_{1}\\|_{2}<\epsilon/2$ and $\\|bQ_{1}-Q_{1}bQ_{1}\\|_{2}<\epsilon/2$. Let $P_{0}=0$, $P_{2}=P$, $P_{3}=P+Q_{1}$, and $P_{4}=I$. Then $0=P_{0}<P_{1}<P_{2}<P_{3}<P_{4}=I$ and $\tau(P_{j})=j/4$ for $0\leq j\leq 4$. Simple computations show that $\\|TP_{j}-P_{j}TP_{j}\\|_{2}\leq\epsilon$ for all $0\leq j\leq 4$. The general case can be proved by using the induction on $n$ with similar arguments as the above. ∎ Combining Lemma 2.6 and the noncommutative H$\ddot{\text{o}}$lder’s inequality, we have the following: ###### Corollary 2.7. Let $\mathcal{M}$ be a separable type ${\rm II}_{1}$ factor and let $T\in\mathcal{M}$. Then for every $\epsilon>0$ and every $t$ with $0\leq t\leq 1$, there is a projection $P\in\mathcal{M}$, $\tau(P)=t$, such that $\\|TP- PTP\\|_{2}<\epsilon$. The following lemma extends Lemma 2.5 to arbitrary type ${\rm II}_{1}$ factors. ###### Lemma 2.8. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor and let $T\in\mathcal{M}$. Then for every $\epsilon>0$, there is a projection $P\in\mathcal{M}$, $\tau(P)=1/2$, such that $\\|TP-PTP\\|_{2}<\epsilon$. ###### Proof. Let $\mathcal{N}$ be the von Neumann subalgebra generated by $T$. Then $\mathcal{N}$ is separable. If $\mathcal{N}^{\prime}\cap\mathcal{M}$ is a diffuse von Neumann algebra, then for every $t$, $0\leq t\leq 1$, there is a projection $P\in\mathcal{N}^{\prime}\cap\mathcal{M}$ such that $PT=TP$ and $\tau(P)=t$. Hence Lemma 2.8 follows. If $\mathcal{N}^{\prime}\cap\mathcal{M}$ is not a diffuse von Neumann algebra, let $P_{0},P_{1},P_{2},\cdots$ be a sequence of projections in $\mathcal{N}^{\prime}\cap\mathcal{M}$ such that $P_{0}+P_{1}+P_{2}+\cdots=I$, $P_{0}(\mathcal{N}^{\prime}\cap\mathcal{M})P_{0}$ is diffuse, and $P_{1},P_{2},\cdots$ are non-zero minimal projections in $(1-P_{0})(\mathcal{N}^{\prime}\cap\mathcal{M})(1-P_{0})$. Note that $(\mathcal{N}P_{n})^{\prime}\cap(P_{n}\mathcal{M}P_{n})=P_{n}(\mathcal{N}^{\prime}\cap\mathcal{M})P_{n}=\mathbb{C}P_{n}$ for $n\geq 1$. This implies that $\mathcal{N}P_{n}$ is a separable type ${\rm II}_{1}$ factor for $n\geq 1$. There is an $n\geq 0$ such that $\sum_{k=1}^{n}\tau(P_{k})\leq t\leq\sum_{k=1}^{n+1}\tau(P_{k})$. Applying Corollary 2.7 to $\mathcal{N}P_{n+1}$, $t^{\prime}=t-\sum_{k=1}^{n}\tau(P_{k})$, and $TP_{n+1}$, there is a projection $Q_{n+1}\in\mathcal{N}P_{n+1}$ such that $\tau(Q_{n+1})=t^{\prime}$ and $\\|TP_{n+1}Q_{n+1}-Q_{n+1}TP_{n+1}Q_{n+1}\\|_{2}<\epsilon.$ Let $P=P_{0}+P_{1}+\cdots+P_{n}+Q_{n+1}$. Then $P\in\mathcal{M}$, $\tau(P)=t$, and $\\|TP-PTP\\|_{2}<\epsilon.$ ∎ As a consequence of Lemma 2.8, Lemma 2.6 is also true for arbitrary type ${\rm II}_{1}$ factors. ###### Proof of Theorem 2.1. Let $T=(T_{n})\in\mathcal{M}^{\omega}$. By Lemma 2.6, for each $n$, there are projections $\\{P_{n,j}\\}_{0\leq j\leq 2^{n}}$ in $\mathcal{M}$ such that $0=P_{n,0}<P_{n,1}<P_{n,2}<\cdots<P_{n,2^{n}-1}<P_{n,2^{n}}=I$, $\tau(P_{n,j})=j/2^{n}$, and $\\|T_{n}P_{n,j}-P_{n,j}T_{n}P_{n,j}\\|_{2}\leq 1/n$ for all $0\leq j\leq 2^{n}$. For every $t$, $0\leq t\leq 1$, choose $P_{n,j}$ such that $\tau(P_{n,j})\leq t<\tau(P_{n,j+1})$. Let $P_{t}=(P_{n,j})\in\mathcal{M}^{\omega}$. Then $P_{s}\leq P_{t}$ if $s\leq t$, $\tau_{\omega}(P_{t})=t$, and $TP_{t}=P_{t}TP_{t}$. ∎ ## 3 Operators with non-trivial invariant subspaces relative to a type ${\rm II}_{1}$ factor Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor with a faithful normal tracial state $\tau$, and let $\mathfrak{M}=\\{S^{\prime}\in\mathcal{M}:\,\text{there is a family of projections}\,\\{P_{t}\\}_{0\leq t\leq 1}\,\text{in}\,\mathcal{M}\,\text{such that}\,ZP_{t}=P_{t}ZP_{t},P_{s}\leq P_{t}\,\text{if}\,s\leq t,\,\text{and}\,\tau(P_{t})=t\\}$. Let $(\mathcal{M})_{1}$ be the set of operators $T$ in $\mathcal{M}$ such that $\\|T\\|\leq 1$. As an application of Theorem 2.1, we prove the following result. ###### Theorem 3.1. For every operator $T\in(\mathcal{M})_{1}$ and every $\epsilon>0$, there is an operator $S\in\mathfrak{M}\cap(\mathcal{M})_{1}$ such that $\\|T-S\\|_{2}<\epsilon$. In particular, the set $\mathfrak{M}$ is dense in $\mathcal{M}$ in the strong operator topology. To prove Theorem 3.1, we need the following lemmas. The following lemma is well known. ###### Lemma 3.2. Suppose $\\{T_{n}\\}_{n}\subseteq(\mathcal{M})_{1}$ is a Cauchy sequence with respect to $\\|\cdot\\|_{2}$. Then there is an operator $T\in(\mathcal{M})_{1}$ such that $\lim_{n\rightarrow\infty}\\|T_{n}-T\\|_{2}=0.$ For an operator $T\in\mathcal{M}$, let $N(T)$ be the projection onto the kernel space of $T$. ###### Lemma 3.3. Let $\epsilon,\delta>0$ and $T\in\mathcal{M}$. If $\\|T\\|_{2}<\delta$, then there is a projection $P\in\mathcal{M}$ such that $P\geq N(T)$, $\\|TP\\|\leq\epsilon$, and $\tau(I-P)<\delta^{2}/\epsilon^{2}$. ###### Proof. By applying the polar decomposition theorem, we may assume that $T$ is a positive operator. Let $\nu$ be the Borel measure on $[0,\infty)$ induced by the composition of $\tau$ with the spectral projections of $T$. Then $\\|T\\|_{2}^{2}=\int_{0}^{\infty}t^{2}d\nu(t)<\delta^{2}.$ Let $P=\chi_{[0,\epsilon]}(T)$. Then $P\geq N(T)$, $\\|TP\\|\leq\epsilon$ and $\epsilon^{2}\tau(I-P)\leq\int_{\epsilon}^{\infty}t^{2}d\nu(t)\leq\\|T\\|_{2}^{2}<\delta^{2}.$ Hence, $\tau(I-P)<\delta^{2}/\epsilon^{2}$. ∎ ###### Lemma 3.4. For every operator $T\in(\mathcal{M})_{1}$ and every $\epsilon>0$, there is an operator $S\in(\mathcal{M})_{1}$ such that 1. 1. $\\|T-S\\|_{2}<\epsilon$ and 2. 2. there is a projection $P\in\mathcal{M}$ such that $\tau(P)=1/2$ and $SP=PSP$. ###### Proof. Choose $\delta,\epsilon_{1}>0$ such that $\epsilon_{1}+\epsilon_{1}/\delta+\delta<\epsilon.$ By Corollary 2.2, there is a projection $P_{1}$ in $\mathcal{M}$ such that $\\|TP_{1}-P_{1}TP_{1}\\|_{2}<\delta.$ (3.1) Let $P_{2}=I-P_{1}$ and $T_{ij}=P_{i}TP_{j}$ for $i,j=1,2$. Then we can write $T=\left(\begin{array}[]{cc}T_{11}&T_{12}\\\ T_{21}&T_{22}\end{array}\right)$ with respect to the decomposition $I=P_{1}+P_{2}$. Since $\\|T\\|\leq 1$, $\\|T_{ij}\\|\leq 1$ for all $i,j=1,2$. Note that ( 3.1) implies $\\|T_{21}\\|_{2}<\delta$ and also note that $N(T_{2,1})\geq P_{2}$. By Lemma 3.3, there is a projection $Q\in\mathcal{M}$, $Q\geq P_{2}$, $\\|T_{21}Q\\|\leq\epsilon_{1}$ and $\tau(I-Q)<\epsilon_{1}^{2}/\delta^{2}$. Write $Q=P_{1}^{\prime}+P_{2}$. Then $P_{1}^{\prime}\leq P_{1}$ and $\tau(P_{1}-P_{1}^{\prime})<\epsilon_{1}^{2}/\delta^{2}$. Let $R=T_{11}P_{1}^{\prime}+T_{12}+T_{22}$, i.e., we can write $R=\left(\begin{array}[]{cc}T_{11}P_{1}^{\prime}&T_{12}\\\ 0&T_{22}\end{array}\right)$ with respect to the decomposition $I=P_{1}+P_{2}$. Then $R-TQ=T_{21}Q$. Therefore, $\\|R\\|=\\|TQ+T_{21}Q\\|\leq 1+\epsilon_{1}.$ (3.2) On the other hand, $R-T=T_{11}(P_{1}-P_{1}^{\prime})+T_{21}$. This implies that $\\|R-T\\|_{2}\leq\\|T_{11}(P_{1}-P_{1}^{\prime})\\|_{2}+\\|T_{21}\\|_{2}\leq\epsilon_{1}/\delta+\delta.$ (3.3) Let $S=(1+\epsilon_{1})^{-1}R$. Then ( 3.2) implies that $\\|S\\|\leq 1$ and (3.3) implies that $\\|S-T\\|_{2}\leq\\|S-R\\|_{2}+\\|R-T\\|_{2}\leq\epsilon_{1}\\|S\\|_{2}+\epsilon_{1}/\delta+\delta\leq\epsilon_{1}+\epsilon_{1}/\delta+\delta<\epsilon.$ Note that $SP_{1}=P_{1}SP_{1}$ and $\tau(P_{1})=1/2$. Let $P=P_{1}$. We prove the lemma. ∎ ###### Proof of Theorem 3.1. We use the induction to construct operators $T_{n}$ and $\\{P_{n,j}\\}_{j=1}^{2^{n}}$ for each $n\geq 0$ satisfying the following conditions: 1. 1. for each $n$, $\\{P_{n,j}\\}_{j=1}^{2^{n}}$ is a family of projections in $\mathcal{M}$ such that $\sum_{j=1}^{2^{n}}P_{n,j}=I$ and $\tau(P_{n,j})=1/2^{n}$ for $1\leq j\leq 2^{n}$; 2. 2. $P_{n,j}=P_{n+1,2j-1}+P_{n+1,2j}$ for $1\leq j\leq 2^{n}$; 3. 3. $\\|T_{n}\\|\leq 1$, $T_{0}=T$, and $\\|T_{n}-T_{n+1}\\|_{2}<\epsilon/2^{n+1}$; 4. 4. for each $k$, $1\leq k\leq 2^{n}$, $\sum_{j=1}^{k}P_{n,j}$ is an invariant subspace of $T_{n}$. For $n=0$, let $T_{0}=T$ and $P_{0,1}=I$. For $n=1$, by Lemma 3.4, there is an operator $S\in\mathcal{M}$, $\\|S\\|\leq 1$, $\\|S-T\\|_{2}<\epsilon/2$ and there is a projection $P\in\mathcal{M}$, $\tau(P)=1/2$ and $SP=PSP$. Let $T_{1}=S$, $P_{1,1}=P$ and $P_{1,2}=I-P$. Now for $n=2$, we construct $T_{2}$ and $\\{P_{2,j}\\}_{j=1}^{4}$ satisfying the above conditions 1,2,3 and 4. Since $P_{1,1}$ is an invariant subspace of $T_{1}$, we can write $T_{1}=\left(\begin{array}[]{cc}A&T_{12}\\\ 0&B\end{array}\right)$ with respect to the decomposition $I=P_{1,1}+P_{1,2}$. Let $\epsilon_{1},\delta>0$ such that $\epsilon_{1}+3\epsilon_{1}/\delta+2\delta<\epsilon/4.$ Applying Corollary 2.2 to $A\in P_{1,1}\mathcal{M}P_{1,1}$ and $B\in P_{1,2}\mathcal{M}P_{1,2}$, there are projections $Q_{1},Q_{2},Q_{3},Q_{4}$ such that $\tau(Q_{j})=1/4$ for $1\leq j\leq 4$, $Q_{1}+Q_{2}=P_{1,1}$, $Q_{3}+Q_{4}=P_{1,2}$, $\\|AQ_{1}-Q_{1}AQ_{1}\\|_{2}<\delta$ and $\\|BQ_{3}-Q_{3}BQ_{3}\\|_{2}<\delta$. Now we can write $T_{1}=\left(\begin{array}[]{cc}\left(\begin{array}[]{cc}A_{11}&A_{12}\\\ A_{21}&A_{22}\end{array}\right)&T_{12}\\\ 0&\left(\begin{array}[]{cc}B_{11}&B_{12}\\\ B_{21}&B_{22}\end{array}\right)\end{array}\right)$ with respect to the decomposition $I=Q_{1}+Q_{2}+Q_{3}+Q_{4}$. Note that $\\|AQ_{1}-Q_{1}AQ_{1}\\|_{2}<\delta$ implies $\\|A_{21}\\|_{2}<\delta$ and $\\|BQ_{3}-Q_{3}BQ_{3}\\|_{2}<\delta$ implies $\\|B_{21}\\|_{2}<\delta$. By Lemma 3.3 and similar arguments as the proof of Lemma 3.4, there are projections $Q_{1}^{\prime}\leq Q_{1}$, $Q_{3}^{\prime}\leq Q_{3}$ such that $\\|A_{21}Q_{1}^{\prime}\\|<\epsilon_{1}$, $\\|B_{21}Q_{3}^{\prime}\\|<\epsilon_{1}$, $\tau(Q_{1}-Q_{1}^{\prime})\leq\epsilon_{1}^{2}/\delta^{2}$ and $\tau(Q_{3}-Q_{3}^{\prime})\leq\epsilon_{1}^{2}/\delta^{2}$. Let $R=\left(\begin{array}[]{cc}\left(\begin{array}[]{cc}A_{11}Q_{1}^{\prime}&A_{12}\\\ 0&A_{22}\end{array}\right)&T_{12}(Q_{3}^{\prime}+Q_{4})\\\ 0&\left(\begin{array}[]{cc}B_{11}Q_{3}^{\prime}&B_{12}\\\ 0&B_{22}\end{array}\right)\end{array}\right)$ with respect to the decomposition $I=Q_{1}+Q_{2}+Q_{3}+Q_{4}$. Then $\\|R-T_{1}(Q_{1}^{\prime}+Q_{2}+Q_{3}^{\prime}+Q_{4})\\|=\\|A_{21}Q_{1}^{\prime}+B_{21}Q_{3}^{\prime}\\|<\epsilon_{1}$ and $\\|R-T_{1}\\|_{2}=\\|A_{11}(Q_{1}-Q_{1}^{\prime})+A_{21}+T_{12}(Q_{3}-Q_{3}^{\prime})+B_{11}(Q_{3}-Q_{3}^{\prime})+B_{21}\\|_{2}\leq 3\epsilon_{1}/\delta+2\delta.$ Therefore, $\\|R\\|\leq\\|T_{1}(Q_{1}^{\prime}+Q_{2}+Q_{3}^{\prime}+Q_{4})\\|+\\|R-T_{1}(Q_{1}^{\prime}+Q_{2}+Q_{3}^{\prime}+Q_{4})\\|<1+\epsilon_{1}.$ Let $T_{2}=(1+\epsilon_{1})^{-1}R$. Then $\\|T_{2}\\|\leq 1$ and $\\|T_{2}-T_{1}\\|_{2}\leq\\|T_{2}-R\\|_{2}+\\|R-T_{1}\\|_{2}<\epsilon_{1}\\|T_{2}\\|+3\epsilon_{1}/\delta+2\delta<\epsilon_{1}+3\epsilon_{1}/\delta+2\delta<\epsilon/4.$ Let $P_{2,j}=Q_{j}$ for $1\leq j\leq 4$. Then $T_{2}$ and $\\{P_{2,j}\\}_{j=1}^{4}$ satisfy the conditions 1,2,3 and 4. The general case can be proved similarly by using the induction. Suppose $T_{n}$ and $\\{P_{n,j}\\}_{j=1}^{2^{n}}$ satisfy the above conditions 1,2,3 and 4. By 3 and Lemma 3.2, there is an operator $S\in(\mathcal{M})_{1}$ such that $\lim_{n\rightarrow\infty}\\|S-T_{n}\\|_{2}=0$ and $\\|S-T\\|_{2}<\epsilon$. By 2 and 4, for each $n$ and $k$, $1\leq k\leq 2^{n}$, $\sum_{j=1}^{k}P_{n,j}$ is an invariant subspace of $T_{N}$ for $N\geq n$ and therefore an invariant subspace of $S$. By 1, $\tau(\sum_{j=1}^{k}P_{n,j})=k/2^{n}$. Note that $\\{k/2^{n}:n\geq 0,1\leq k\leq 2^{n}\\}$ is dense in $[0,1]$. For every $t$, $0\leq t\leq 1$, let $\displaystyle{P_{t}=\bigvee_{k/2^{n}\leq t}\left(\sum_{j=1}^{k}P_{n,j}\right)}.$ By 1, $P_{s}\leq P_{t}$ if $s\leq t$, $\tau(P_{t})=t$ and $SP_{t}=P_{t}SP_{t}$. ∎ ## 4 $\prod^{\omega}M_{n}(\mathbb{C})$ is not $\ast$-isomorphic to $\mathcal{R}^{\omega}$ Throughout this section $\mathcal{M}$ is a separable type ${\rm II}_{1}$ factor. Recall that a separable type ${\rm II}_{1}$ factor $\mathcal{M}$ has property $\Gamma$ if for every $n$, $T_{1},\cdots,T_{n}\in\mathcal{M}$, and every $\epsilon>0$, there is a projection $P\in\mathcal{M}$ such that $\tau(P)=1/2$ and $\\|T_{i}P-PT_{i}\\|_{2}<\epsilon$ (cf. [4]). ###### Lemma 4.1. Suppose $\mathcal{M}$ has property $\Gamma$. Then for every operator $T\in\mathcal{M}^{\omega}$ and $t$, $0\leq t\leq 1$, there is a projection $P\in\mathcal{M}^{\omega}$ such that $PT=TP$ and $\tau_{\omega}(P)=1/2$. ###### Proof. Write $T=(T_{n})$. Since $\mathcal{M}$ has property $\Gamma$, there exists a projection $P_{n}\in\mathcal{M}$ such that $\\|P_{n}T_{n}-T_{n}P_{n}\\|_{2}<1/n$ and $\tau(P_{n})=1/2$. Let $P=(P_{n})\in\mathcal{M}^{\omega}$. Then $PT=TP$ and $\tau_{\omega}(P_{n})=1/2$. ∎ Let $(M_{n}(\mathbb{C}))_{1}$ be the set of matrices $T\in M_{n}(\mathbb{C})$ such that $\\|T\\|\leq 1$, and let $\nu((M_{n}(\mathbb{C}))_{1},\omega)$ be the covering number of $(M_{n}(\mathbb{C}))_{1}$ with respect to the normalized trace norm $\\|\cdot\\|_{2}$. There are universal constants $c_{1},c_{2}$ [14, 15] such that $\left(\frac{c_{1}}{\omega}\right)^{2n^{2}}\leq\nu((M_{n}(\mathbb{C}))_{1},\omega)\leq\left(\frac{c_{2}}{\omega}\right)^{2n^{2}}.$ (4.1) The next lemma follows from Theorem 9 of Herrero and Szarek [5] (also see [17]). For the sake of completeness, we include a direct proof. ###### Lemma 4.2. There exists a universal constant $\alpha>0$ with the following property: for each $n\geq 2$, there exists a matrix $T_{n}\in M_{n}(\mathbb{C})$, $\\|T_{n}\\|=1$, such that $\\|PT_{n}-T_{n}P\\|_{2}\geq\alpha$ for every projection $P\in M_{n}(\mathbb{C})$ with ${\rm rank}P=\left[\frac{n}{2}\right]$, where $\left[\frac{n}{2}\right]$ is the maximal integer less or equal to $\frac{n}{2}$. ###### Proof. Suppose the lemma is false. Then for every $\epsilon>0$, there is an $n\geq 2$, for every matrix $T\in M_{n}(\mathbb{C})$, $\\|T\\|\leq 1$, there is a projection $P\in M_{n}(\mathbb{C})$ such that ${\rm rank}P=\left[\frac{n}{2}\right]$ and $\\|PT-TP\\|_{2}<\epsilon$. Without of loss of generality we may assume that $n=2k$. Let $(M_{n}(\mathbb{C}))_{1}$ be the set of $n\times n$ complex matrices $T$ such that $\\|T\\|\leq 1$. For $T\in M_{n}(\mathbb{C})$, let $\\|T\\|_{2}$ be the trace norm with respect to the normalized trace $\tau_{n}=\frac{Tr}{n}$ on $M_{n}(\mathbb{C})$. By (4.1), $\left(\frac{c_{1}}{2\epsilon}\right)^{2n^{2}}\leq\nu((M_{n}(\mathbb{C}))_{1},2\epsilon)\leq\left(\frac{c_{2}}{2\epsilon}\right)^{2n^{2}}$ (4.2) and $\left(\frac{c_{1}}{\epsilon}\right)^{2k^{2}}\leq\nu((M_{k}(\mathbb{C}))_{1},\epsilon)\leq\left(\frac{c_{2}}{\epsilon}\right)^{2k^{2}}.$ Let $\\{T_{t}\\}_{t\in\mathbb{T}}$ be an $\epsilon$-net of $(M_{k}(\mathbb{C}))_{1}$ such that $\\#\mathbb{T}\leq\left(\frac{c_{2}}{\epsilon}\right)$. Now for every $T\in(M_{n}(\mathbb{C}))_{1}$, $\\|TP-PT\\|_{2}<\epsilon$ for some projection $P\in M_{n}(\mathbb{C})$ with rank $k$. Write $T=\left(\begin{array}[]{cc}T_{11}&T_{12}\\\ T_{21}&T_{22}\end{array}\right)$ with respect to the decomposition $I=P+(I-P)$. Since $\\|T\\|\leq 1$, $\\|T_{11}\\|,\\|T_{22}\\|\leq 1$. Choose $t_{1},t_{2}\in\mathbb{T}$ such that $\\|T_{11}-T_{t_{1}}\\|_{2}<\epsilon$ and $\\|T_{22}-T_{t_{2}}\\|_{2}<\epsilon$ with respect to the normalized trace norm on $M_{k}(\mathbb{C})$. Since $\\|TP- PT\\|_{2}<\epsilon$, $\\|T-\left(\begin{array}[]{cc}T_{t_{1}}&0\\\ 0&T_{t_{2}}\end{array}\right)\\|_{2}<2\epsilon.$ This implies that, $\nu((M_{n}(\mathbb{C}))_{1},2\epsilon)\leq\left(\frac{c_{2}}{\epsilon}\right)^{2k^{2}}\cdot\left(\frac{c_{2}}{\epsilon}\right)^{2k^{2}}=\left(\frac{c_{2}}{\epsilon}\right)^{4k^{2}}.$ (4.3) Note that $n=2k$. By (4.2), $\left(\frac{c_{1}}{2\epsilon}\right)^{2n^{2}}\leq\left(\frac{c_{2}}{\epsilon}\right)^{n^{2}}.$ By taking $\ln$ on both sides, we have $\frac{2(\ln c_{1}-\ln 2-\ln\epsilon)}{-\ln\epsilon}\leq\frac{\ln c_{2}-\ln\epsilon}{-\ln\epsilon}.$ Let $\epsilon\rightarrow 0+$. This implies $2\leq 1$. This is a contradiction. ∎ ###### Theorem 4.3. The von Neumann algebra $\prod^{\omega}M_{n}(\mathbb{C})$ is not $\ast$-isomorphic to $\mathcal{R}^{\omega}$, the ultrapower algebra of the hyperfinite ${\rm II}_{1}$ factor. ###### Proof. Choose $T_{n}\in M_{n}(\mathbb{C})$ as in Lemma 4.2. Let $T=(T_{n})\in\prod^{\omega}M_{n}(\mathbb{C})$. Claim if $P$ is a projection in $\prod^{\omega}M_{n}(\mathbb{C})$ such that $TP=PT$, then $\tau_{\omega}(P)\neq 1/2$. Otherwise, suppose $P=(P_{n})\in\prod^{\omega}M_{n}(\mathbb{C})$ is a projection such that $TP=PT$ and $\tau_{\omega}(P)=1/2$. We may assume that $P_{n}$ is a projection in $M_{n}(\mathbb{C})$ with ${\rm rank}P=\left[\frac{n}{2}\right]$. By Lemma 4.2, $\\|T_{n}P_{n}-P_{n}T_{n}\\|_{2}\geq\alpha>0$. Hence $\\|PT- TP\\|_{2}\geq\alpha>0$. This is a contradiction. On the other hand, for every operator $T\in\mathcal{R}^{\omega}$, there is a projection $Q\in\mathcal{R}^{\omega}$ such that $TQ=QT$ and $\tau_{\omega}(Q)=1/2$ by Lemma 4.1. So $\prod^{\omega}M_{n}(\mathbb{C})$ is not $\ast$-isomorphic to $\mathcal{R}^{\omega}$. ∎ ###### Remark 4.4. _By Theorem 9 of[5], there is an operator $T$ in $\prod^{\omega}M_{n}(\mathbb{C})$ such that if $TP=PT$ for some projection $P$ in $\prod^{\omega}M_{n}(\mathbb{C})$, then $P=0$ or $P=I$._ Question: Can $\mathcal{R}^{\omega}$ be embedded into $\prod^{\omega}M_{n}(\mathbb{C})$? If $\mathcal{M}$ is a separable type ${\rm II}_{1}$ factor and $\mathcal{M}^{\omega}\cong\mathcal{R}^{\omega}$, is $\mathcal{M}\cong\mathcal{R}$? ## 5 The lattice of invariant subspaces of an operator affiliated with a type ${\rm II}_{1}$ factor Let $\mathcal{M}$ be a factor (not necessarily type ${\rm II}_{1}$) acting on a Hilbert space $\mathcal{H}$ and $T\in\mathcal{M}$. We denote by $Lat_{\mathcal{M}}T$ the set of projections $P\in\mathcal{M}$ such that $TP=PTP$. So $P\in Lat_{\mathcal{M}}$ if and only if $P\mathcal{H}$ is an invariant subspace of $T$. Recall that a hyperinvariant subspace of $T$ is a (closed) subspace invariant under every operator in $\\{T\\}^{\prime}$. It is easy to see that the projection onto a hyperinvariant subspace of $T$ is in the von Neumann algebra generated by $T$. Suppose $S,T$ are two operators in $\mathcal{M}$. Recall that $S$ and $T$ are _quasi-similar_ in $\mathcal{M}$ if there are operators $X,Y\in\mathcal{M}$ which are one-to-one and have dense range such that $SX=XT$ and $YS=TY$. The following theorem is given in [11](Theorem 6.19). ###### Theorem 5.1. If $S$ and $T$ are quasi-similar in $\mathcal{B}(\mathcal{H})$ and $S$ has a nontrivial hyperinvariant subspace, then $T$ has a nontrivial hyperinvariant subspace. It is still not known that if we replace the hyperinvariant subspace by the invariant subspace in the above theorem, the theorem still holds or not. However, in this section we will show that if we replace $\mathcal{B}(\mathcal{H})$ by a type ${\rm II}_{1}$ factor and replace the hyperinvariant subspace by the invariant subspace, then the above theorem still holds. We denote by $N(T)$ the kernel space of $T$ and $R(T)$ the closure of range space of $T$. ###### Lemma 5.2. Let $\mathcal{M}$ be a finite von Neumann algebra with a faithful normal trace $\tau$, and let $T\in\mathcal{M}$. Then $\tau(R(T))+\tau(N(T))=1$. In particular, $N(T)=0$ if and only if $R(T)=I$. ###### Proof. By the polar decomposition theorem, there is a unitary operator $U$ and a positive operator $\|T\|$ in $\mathcal{M}$ such that $T=U\|T\|$. So $T^{}=\|T\|U^{}$. Now, we have $T^{}T=\|T\|^{2}=U^{}TT^{}U$. Thus, $\tau(R(T))=\tau(R(TT^{}))=\tau(R(T^{}T))=\tau(R(T^{}))=1-\tau(N(T))$. ∎ ###### Corollary 5.3. Let $\mathcal{M}$ be a finite von Neumann algebra with a faithful normal trace $\tau$. Let $T\in\mathcal{M}$ be an operator such that $N(T)=0$, and let $E\in\mathcal{M}$ be a projection. Then $\tau(R(TE))=\tau(E)$. In particular, if $0<E<I$, then $0<R(TE)<I$. ###### Proof. Since $N(T)=0$, $N(TE)=I-E$. By lemma 5.2, $\tau(R(TE))=1-\tau(N(TE))=1-\tau(I-E)=\tau(E)$. ∎ ###### Proposition 5.4. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor with a faithful normal trace $\tau$ and $S,T\in\mathcal{M}$. If there is an operator $X\in\mathcal{M}$ such that $N(X)=0$ and $XS=TX$, then $LatS$ is isomorphic to a sublattice of $LatT$ and $LatT$ is isomorphic to a sublattice of $LatS$. In particular, $S$ has a nontrivial invariant subspace if and only if $T$ has a nontrivial invariant subspace. ###### Proof. For $E\in Lat_{\mathcal{M}}S$, let $F=R(XE)$. The assumption $XS=TX$ implies that $F\in Lat_{\mathcal{M}}T$. Define $\phi(E)=F$. By corollary 5.3, $\tau(F)=\tau(E)$. We want to show that $\phi$ is a lattice isomorphism from $Lat_{\mathcal{M}}S$ onto a sublattice of $Lat_{\mathcal{M}}T$. Let $E_{1},E_{2}\in LatS$. Then $\phi(E_{1}\vee E_{2})=R(X(E_{1}\vee E_{2}))=R(XE_{1})\vee R(XE_{2})=\phi(E_{1})\vee\phi(E_{2})$ and $\phi(E_{1}\wedge E_{2})=R(X(E_{1}\wedge E_{2}))\leq R(X(E_{1}))\wedge R(X(E_{2}))=\phi(E_{1})\wedge\phi(E_{2})$. By corollary 5.3, $\tau(\phi(E_{1})\wedge\phi(E_{2}))=\tau(\phi(E_{1})\vee\phi(E_{2}))-\tau(\phi(E_{1}))-\tau(\phi(E_{2}))$ $=\tau(E_{1}\vee E_{2})-\tau(E_{1})-\tau(E_{2})=\tau(E_{1}\wedge E_{2})=\tau(\phi(E_{1}\wedge E_{2})).$ So $\phi(E_{1}\wedge E_{2})=\phi(E_{1})\wedge\phi(E_{2})$. Thus $\phi$ is a lattice homomorphism. Let $E_{1},E_{2}\in LatS$ and $E_{1}\neq E_{2}$. We may assume that $E=E_{1}\vee E_{2}>E_{1}$. So $\tau(E)>\tau(E_{1})$. If $\phi(E_{1})=\phi(E_{2})=F\in LatT$. Then $F=\phi(E_{1}\vee E_{2})$. By corollary 5.3, $\tau(F)=\tau(E_{1})=\tau(E_{1}\vee E_{2})=\tau(E)$. This is a contradiction. So $\phi$ is a lattice isomorphism from $Lat_{\mathcal{M}}S$ onto a sublattice of $Lat_{\mathcal{M}}T$. Similarly, by $X^{}T^{}=S^{}X^{}$, there is a lattice isomorphism from $Lat_{\mathcal{M}}T^{}$ onto a sublattice of $Lat_{\mathcal{M}}S^{}$. Since $Lat_{\mathcal{M}}T$ is isomorphism to $Lat_{\mathcal{M}}T^{}$ and $Lat_{\mathcal{M}}S$ is isomorphic to $Lat_{\mathcal{M}}S^{}$. So there is a lattice isomorphic from $Lat_{\mathcal{M}}T$ onto a sublattice of $Lat_{\mathcal{M}}S$. ∎ ###### Proposition 5.5. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor and $S,T\in\mathcal{M}$. If $S$ and $T$ are quasi-similar, then the lattice of hyperinvariant subspaces of $S$ and the lattice of hyperinvariant subspaces of $T$ are isomorphic. ###### Proof. Let $X,Y$ in $\mathcal{M}$ be one to one operators with dense ranges such that $XS=TX$ and $SY=YT$. Let $E$ be a hyperinvariant subspace of $S$. Let $F$ the closure of the linear span of $R(AXE)$, where $AT=TA$. Then clearly $F$ is a hyperinvariant subspace of $T$. Note that $\tau(F)\geq\tau(XE)=\tau(E)$ by corollary 5.3. Since $YAXS=YATX=YTAX=SYAX$ and $E$ is a hyperinvariant subspace of $S$, $R(YAXE)\leq E$ and therefore, $R(YF)\leq E$. By corollary 5.3, $\tau(E)\geq\tau(F)$. So $\tau(F)=\tau(E)$, $F=R(XE)$, and $E=R(YF)$. Now $E\rightarrow F=R(XE)$ is a lattice isomorphism (the inverse is $F\rightarrow E=R(YF)$) from the lattice of hyperinvariant subspaces of $S$ onto the lattice of hyperinvariant subspaces of $T$. ∎ ###### Corollary 5.6. Let $\mathcal{M}$ be a type ${\rm II}_{1}$ factor and $S,T\in\mathcal{M}$. Then $Lat_{\mathcal{M}}ST$ is not trivial iff $Lat_{\mathcal{M}}TS$ is not trivial. Furthermore, if $N(S)=N(T)=0$, then $Lat_{\mathcal{M}}ST$ is isomorphic to $Lat_{\mathcal{M}}TS$ and the lattice of hyperinvariant subspaces of $ST$ is isomorphic to the lattice of hyperinvariant subspaces of $TS$ as lattices. ###### Proof. Suppose $Lat_{\mathcal{M}}ST$ is not trivial. If $TS=0$, then $Lat_{\mathcal{M}}TS$ is not trivial. We assume that $TS\neq 0$. If $N(S)\neq 0$ or $R(T)\neq I$, then $N(S)$ or $R(T)$ is a non trivial invariant subspace of $TS$. if $N(S)=0$ and $R(T)=I$, then by lemma 5.2, $R(S)=I$ and $N(T)=0$. Thus $ST,TS$ are quasisimilar. By Proposition 5.4, $Lat_{\mathcal{M}}TS$ is not trivial. If $N(S)=N(T)=0$, then $R(S)=R(T)=I$ by lemma 5.2. For $E\in Lat_{\mathcal{M}}ST$, let $F=R(TE)$ and $E_{1}=R(SF)$. Then $E_{1}=R(SF)=R(STE)\leq E$ since $E\in Lat_{\mathcal{M}}ST$. By corollary 5.3, $\tau(E)=\tau(F)=\tau(E_{1})$. This implies that $E=E_{1}$. Note that $R(TSF)=R(TSTE)\leq R(TE)=F$, $F\in Lat_{\mathcal{M}}TS$. Define $\phi(E)=R(TE)$ and $\psi(F)=R(SF)$ for $E\in Lat_{\mathcal{M}}ST$ and $F\in Lat_{\mathcal{M}}TS$, respectively. Then $\psi=\phi^{-1}$. So $\phi$ is a lattice isomorphism from $Lat_{\mathcal{M}}ST$ onto $Lat_{\mathcal{M}}TS$. The lattice of hyperinvariant subspaces of $ST$ is isomorphic to the lattice of hyperinvariant subspaces of $TS$ as lattices is a corollary of Proposition 5.5. ∎ ###### Remark 5.7. _Let $T\in\mathcal{B}(H)$ and $V\in\mathcal{B}(H)$ such that $VV^{}=I$ but $V^{}V\neq I$. Then $R(V^{})$ is a nontrivial invariant subspace of $V^{}TV$. Note that $T=TVV^{}$. If the first part of Corollary 5.6 is true for $\mathcal{M}=\mathcal{B}(\mathcal{H})$, then the answer to the invariant subspace question (relative to $\mathcal{B}(\mathcal{H})$ is affirmative._ ## References [1] L.G. Brown, Lidskii’s theorem in the type ${\rm II}$ case, Geometric methods in operator algebras, H. Araki and E. Effros (Eds.) _Pitman Res. notes in Math. Ser_ 123, Longman Sci. Tech. (1986), 1-35. * [2] K. Dykema and U. Haagerup, Invariant subspaces of Voiculescu’s circular operator, _Geom. Funct. Anal._ 11 (2001), 693-741. * [3] K. Dykema and U. Haagerup, Invariant subspaces of the quasinilpotent DT-operator, _J. Funct. Anal._ 209 no.2, (2004), 332-366. * [4] J. Dixmier, Quelques propriétés des suites centrales dans les facteurs de type ${\rm II}_{1}$, (French) _Invent. Math._ 7 (1969) 215–225. * [5] D. A. Herrero and S. J. Szarek How well can an $n\times n$ matrix be approximated by reducible ones? _Duke Math. J._ 53 (1986), no. 1, 233–248. * [6] U. Haagerup and H. Schultz, Invariant Subspaces for Operators in a General ${\rm II}_{1}$-factor, preprint available at http://www.arxiv.org/pdf/math.OA/0611256. * [7] U. Haagerup and H. Schultz, Brown measures of unbounded operators affiliated with a finite von Neumann algebra, _Math Scand._ , 100 (2007), no. 2, 209–263. * [8] D. McDuff, Central sequences and the hyperfinite factor, _Proc. London Math. Soc._ 21 (1970), 443–461. * [9] F. Murray and J. von Neumann, On rings of operators, IV,_Ann. of Math._ 44 (1943), 716–808. * [10] S. Popa, Free independent sequences in type II1 fractors and related problems, _Asterisque_ , 232 (1995), 187-202. * [11] H. Radjavi and P. Rosenthal, “Invariant Subspaces”, Springer-Verlag, New York, 1973. * [12] S. Sakai, “The Theory of W* Algebras”, Lecture notes, Yale University, 1962. * [13] P. Sniady and R. Speicher, Continuous family of invariant subspaces for $R-$diagonal operators, _Invent. Math._ 146 (2001), 329-363. * [14] S. J. Szarek, Nets of Grassmann manifold and orthogonal group, _Proceedings of research workshop on Banach space theory_ (Iowa City, Iowa, 1981), 169 C185, Univ. of Iowa, Iowa City, Iowa, 1982. * [15] S. J. Szarek, The finite-dimensional basis problem with an appendix on nets of Grassmann manifolds, _Acta Math._ 151 (1983), no. 3-4, 153–179. * [16] D.V. Voiculescu, K. Dykema and A. Nica, “Free Random Variables”, CRM Monograph Series, vol. 1, AMS, Providence, R.I., 1992\. * [17] J. von Neumann, Approximative properties of matrices of high finite order, _Portugaliae Math._ 3, (1942). 1–62. E-mail address: [Junsheng Fang] jfang@math.tamu.edu Address: Department of Mathematics, Texas A&M University, College Station, TX,77843. E-mail address: [Don Hadwin] don@math.unh.edu Address: Department of Mathematics, University of New Hampshire, Durham, NH, 03824.	arxiv-papers	2008-08-01T01:15:26	2024-09-04T02:48:57.087034	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Junsheng Fang, Don Hadwin", "submitter": "Junsheng Fang", "url": "https://arxiv.org/abs/0808.0049" }
0808.0091	# Are there the VP couplings in the $\psi(3770)$ non-charmed decays hidden behind the current measurements? D. Zhang, G. Rong, J.C. Chen Institute of High Energy Physics, Beijing 100049, China zhangdh@mail.ihep.ac.cn ###### Abstract A global analysis of the full amplitudes for $e^{+}e^{-}\rightarrow$ VP (Vector and Pseudoscalar) channels at $\sqrt{s}$ =3.773 GeV and 3.670 GeV, which were measured by the CLEO-c Collaboration, shows that those measurements are essentially nontrivial for searching for the $\psi(3770)$ non-$D\overline{D}$ decays. Unlike the nearly negative verdict on the $\psi(3770)$ strong decays to the VP channels in the original analysis of the CLEO-c data, there exist some unusual solutions that predict the remarkable strength of $SU(3)$ symmetry VP decay of $\psi(3770)$ resonance, which give some clue to understand the mechanism of $\psi(3770)$ non-$D\overline{D}$ decays and to reexplain the well-known $\rho-\pi$ puzzle in the J/$\psi$ and $\psi(3686)$ decays. ###### pacs: 13.20.Gd, 13.66.Bc, 14.40.Gx ## I Motivation There is a long-standing puzzle in understanding the exist measurements for $\psi(3770)$ and $D\overline{D}$ production cross sections at the peak of $\psi(3770)$ production in $e^{+}e^{-}$ annihilation nondd ; R-Z-C . Potential Models predict that $\psi(3770)$ decays into $D\overline{D}$ with branching fraction of $\sim 100\%$. Recently careful investigation shows that the branching fraction of $\psi(3770)$ non-$D\overline{D}$ decay would be up to more than $10\%$ PhysRevLett_97_121801 ; PhysLettB_641_145 . It is very interesting to know what are the exclusive $\rm{non-}D\overline{D}$ final states of $\psi(3770)$ decays. Except the electromagnetic transitions and hadronic transitions of $\psi(3770)$ to lower charmonium states, are there indeed other significant exclusive non-charmed decay modes from $\psi(3770)$ decays? In the charmonium decays, there is another long-standing puzzle in understanding the $\rho\pi$ decays of J/$\psi$ and $\psi(3686)$. The partial widths of $\rho\pi$ channel and other VP channels in the $\psi(3686)$ decays are unexpectedly lower than those in J/$\psi$ decays. This is so called “$\rho\pi$” puzzle. Are the J/$\psi$ decay rates enhanced by some unknown mechanism or the $\psi(3686)$ decay rates are suppressed abnormally? To investigate the possible source of this puzzle, it is also important to measure the $\psi(3770)$ VP decay amplitude. Recently, the BES Collaboration BES_KK observed a large production cross section for $e^{+}e^{-}\rightarrow K^{}(892)^{0}\overline{K}^{0}$+c.c. $\sigma(e^{+}e^{-}\rightarrow{K^{0}\overline{K}^{0}}+c.c.)=(15.0\pm 4.6\pm 3.3)~{}~{}{\rm{pb}},$ at center-of-mass energy of $\sqrt{s}$=3.773 GeV and found that the $K^{\pm}(892)K^{\mp}$ production is suppressed. Taking into account the possible interference between the strong decay amplitude and the continuum production amplitude at $\sqrt{s}$=3.773 GeV, the BES Collaboration set an upper limit on the strong decay partial width for $\psi(3770)\rightarrow K^{}(892)\overline{K}$+c.c. to be $\Gamma(\psi(3770)\rightarrow K^{}(892)\overline{K}~{}+~{}\rm{c.c.})<29.0~{}~{}{\rm keV}$ at $90\%$ confidence level. The CLEO-c Collaboration made more careful studies of twelve exclusive VP decay channels for $\psi(3770)\rightarrow\rho\pi$, $K^{}(892)\overline{K}+c.c.$, $\omega\pi^{0}$, $\rho\eta$, $\rho\eta^{\prime}$, $\omega\eta$, $\omega\eta^{\prime}$, $\phi\eta$, $\phi\eta^{\prime}$ and $\phi\pi^{0}$ reported in Ref. CLEO_c_VP . The CLEO-c Collaboration measured the cross sections for all of the channels at the energies $\sqrt{s}$=3.773 GeV and $\sqrt{s}$=3.670 GeV. The CLEO-c results show that the measured cross sections at $\sqrt{s}$=3.773 GeV are almost equal to or even less than the ones measured at $\sqrt{s}$=3.670 GeV, which mean that the net cross sections for the $\psi(3770)$ decays are consistent with zero except only for the channel $\psi(3770)\rightarrow\phi\eta$. The negative results about the $\psi(3770)$ strong VP decays led people ignoring the important strong decay component existing in $\psi(3770)$ and only focusing their attention on the form factors of those channels as well as the isospin violation in electromagnetic interaction CLEO_c_VP . In this paper, we develop a model to account for both the amplitudes of electromagnetic (E-M) production and $\psi(3770)$ strong decay in the process of $e^{+}e^{-}\rightarrow$ VP. By analyzing the cross sections for the exclusive VP channels, which were measured by the CLEO-c Collaboration, we extract out the branching fractions for $\psi(3770)$ decay to these VP final states. ## II The Model and the Formulae In the $\psi(3686)$ decay sector, because of the smallness of the strong VP decay coupling, the E-M decay component as well as the continuum (E-M) component of the VP channel would be no longer the small amounts comparing with those of strong decay. People have to deal with the two components properly Wang_2003 . At the resonance peak, the production amplitude consists of two parts, one is the decay amplitude of charmonium resonance and the other is continuum E-M production amplitude. In the resonance decay part, there are two components as well. They are the E-M decay amplitudes and the strong decay amplitude. Totally, there are three components involved in the $e^{+}e^{-}$ annihilation process at the resonance peak, which are the strong decay component, the E-M decay component and the continuum production component. Unlike the VP decays of J/$\psi$ and $\psi(3686)$, the E-M decay amplitudes of $\psi(3770)$ can be neglected due to the little tiny dileptonic decay branching fraction. There are only the strong decay amplitude and the continuum production amplitude in the $e^{+}e^{-}$ collision production at $\sqrt{s}$=3.773 GeV. Typically, according to the conventional point of view, the partial widths of the $\psi(3770)$ VP decay channels could be up to keV order of magnitude, like their cousins in J/$\psi$ decays. However, due to the large width of $\psi(3770)$ the decay amplitudes of those channels can not get large amplification as the ones at the narrow resonance states. Associated with the measurements of the form factors of channel $\omega\pi^{0}$ at the energies of $\psi(3686)$ and $\psi(3770)$ resonance vicinities BES_ompi0 ; CLEO_3686_VP ; CLEO_c_VP the decay process with only a few keV partial width of the rather wide resonance is really hard to be measured if one does not consider the interference between the amplitudes of the strong decay and the continuum production. As the measurements by CLEO-c CLEO_c_VP , both the evident yield excess of channel $\phi\eta$ and the rather large yield deficit of the $\rho\pi$ channel at the resonance peak hint that there must be rather complex interference between the two kinds of amplitudes acting globally on the VP channels. Some destructive interference just shows up at $\rho\pi$ channel in the “deficit” way. And more complex interferences cause the $\phi\eta$ yields enhanced at resonance peak. In fact, in such complicate interference case the decay contributions may easily be covered up by the continuum contribution and the interference contributions. If one completely neglects the buried decay contribution, the single E-M amplitude assumption would not describe the measured cross sections well. In practice, it is dangerous to measure the branching fractions for the $\psi(3770)$ non-$D\overline{D}$ decays by simply considering the net yields for the channels observed at the peak of $\psi(3770)$ over that at the nearby off resonance region. In this analysis we introduce the strong decay amplitudes in the analysis formalism to see how the strong decay affects the VP production at the $\psi(3770)$ resonance peak. We describe the global decay of $\psi(3770)$ and the continuum production process still based on the flavor $SU(3)$ invariant model, which was developed thirty years ago haber_su3 , In this model the strange quark mass correction in both of the strong coupling and the E-M coupling, the wave function nonet symmetry breaking and the double Okubo-Zweig-Iizuka (DOZI) suppression effects are all taken into account. As for the continuum production at the two energy points $\sqrt{s}$=3.773 GeV and 3.670 GeV, except the coherent strong decay amplitudes from the J/$\psi$ and $\psi(3686)$ tails, which can safely be neglected from the calculations, there is only the continuum E-M amplitudes itself. In addition, we guess that the incoherent component contributions which are mainly from the initial state radiative (ISR) return to J/$\psi$ and $\psi(3686)$ resonances have efficiently been rejected in the work reported in Ref. CLEO_c_VP and can be neglected in our analysis too. Following the convention given in Ref. Seiden_1988 , we define that ${\bf g}$ represents the VP strong decay amplitude in the flavor $SU(3)$ symmetry limit; ${{\bf{g}}_{s}}$ represents the strong decay amplitude from $s$ quark, $s_{g}=1-({\bf{g}}_{s}/\|{\bf{g}}\|)/2,$ (1) characterizes the $SU(3)$ mass violation, which is as the same as the parameter “$s$” given in the Tab. VIII of Ref. Seiden_1988 ; $\theta_{P}$ represents the $\eta-\eta^{\prime}$ mixing angle; the product $r\cdot{\bf{g}}$ represents the amplitude correction of the $SU(3)$ nonet symmetry violation with the factor $(1-s_{\rm P})$ for a strange pseudoscalar production and with the factor $(1-s_{\rm V})$ for a strange vector production. If $s_{\rm V}=s_{\rm P}$=0, (exactly $s_{\rm V}+s_{\rm P}$=0), $r\cdot{\bf{g}}$ measures the pure DOZI amplitude correction. Unlike the case in Ref. Seiden_1988 , because the E-M amplitude is no longer small comparing with the strong amplitude, we have to consider both the isoscalar and isovector components of the E-M amplitude. We define the E-M amplitude in the form of $SU(3)$ octet matrix representation as ${\bf E}={\bf e}_{1}\cdot\rm{I}_{3}+{\bf e}_{0}\cdot\rm Y$ in which ${\bf e}_{0}$ and ${\bf e}_{1}$ are the isoscalar and isovector components, respectively, and $\rm{I}_{3}$ and Y are, respectively, the isospin third component and the hypercharge matrices in flavor $SU(3)$ octet space. We define $\theta_{0}$ as the phase of ${\bf e}_{0}$ relative to ${\bf{g}}$, $\delta_{1}$ as the phase shift difference of ${\bf e}_{1}$ to ${\bf e}_{0}$ and a factor $(1/2-s_{e})$ as the correction for strange quark coupling to E-M isoscalar part ${\bf{e}}_{0}$. We assume that the couplings ${\bf e}$’s and their phases do not change in the all VP channels, and their moduli at the two different energy points $\sqrt{s}$=3.773 GeV and $\sqrt{s}$=3.671 GeV only change with a $1/s^{3}$ dependence. If ${\bf e}_{1}={\bf e}_{0}$, we return to common definition as Refs. haber_su3 ; Seiden_1988 did. Table 1: The Amplitudes for VP production in $e^{+}e^{-}$ annihilation at $\sqrt{s}=3.773$ GeV. The coupling ${\bf{g}}_{K}$ defined in channels $K^{}\bar{K}+c.c.$ can be considered as a free parameter if one of ${\bf{g}}$ and ${\bf{g}}_{s}$ is fixed. Channel($ch$) \| $M^{ch}_{res,3770}$ \| $M^{ch}_{ctm,3770}$ ---\|---\|--- $\rho^{0}\pi^{0},\rho^{\pm}\pi^{\mp}$ \| ${\bf{g}}$ \| ${\bf{e}}_{0}$ $\omega\eta$ \| ${\bf{g}}X_{\eta}+\sqrt{2}r{\bf{g}}[\sqrt{2}X_{\eta}+(1-s_{\rm P})Y_{\eta}]$ \| ${\bf{e}}_{0}X_{\eta}$ $\phi\eta$ \| ${\bf{g}}_{s}Y_{\eta}+r{\bf{g}}(1-s_{\rm V})[\sqrt{2}X_{\eta}+(1-s_{\rm P})Y_{\eta}]$ \| $-2{\bf{e}}_{0}(1-s_{e})Y_{\eta}$ $\omega\eta^{\prime}$ \| ${\bf{g}}X_{\eta^{\prime}}+\sqrt{2}r{\bf{g}}[\sqrt{2}X_{\eta^{\prime}}+(1-s_{\rm P})Y_{\eta^{\prime}}]$ \| ${\bf{e}}_{0}X_{\eta^{\prime}}$ $\phi\eta^{\prime}$ \| ${\bf{g}}_{s}Y_{\eta^{\prime}}+r{\bf{g}}(1-s_{\rm V})[\sqrt{2}X_{\eta^{\prime}}+(1-s_{\rm{P}})Y_{\eta^{\prime}}]$ \| $-2{\bf{e}}_{0}(1-s_{e})Y_{\eta^{\prime}}$ $\omega\pi^{0}$ \| 0 \| $3{\bf{e}}_{1}$ $\phi\pi^{0}$ \| 0 \| 0 $\rho^{0}\eta$ \| 0 \| $3{\bf{e}}_{1}X_{\eta}$ $\rho^{0}\eta^{\prime}$ \| 0 \| $3{\bf{e}}_{1}Y_{\eta}$ $K^{0}\bar{K}^{0}+c.c.$ \| ${\bf{g}}_{K}=({\bf{g}}+{\bf{g}}_{s})/2$ \| $-{\bf{e}}_{0}(1/2-s_{e})-3/2{\bf e}_{1}$ $K^{\pm}K^{\mp}$ \| ${\bf{g}}_{K}=({\bf{g}}+{\bf{g}}_{s})/2$ \| $-{\bf{e}}_{0}(1/2-s_{e})+3/2{\bf e}_{1}$ For the channels “$ch$”, ($ch$=$\rho^{0}\pi^{0},K^{0}\overline{K}^{0}+c.c.$, etc) at energy $\sqrt{s}$, $M^{ch}_{res,\sqrt{s}}$ denotes the resonance decay amplitude and $M^{ch}_{ctm,\sqrt{s}}$ denotes the continuum production amplitude. The total production amplitudes are then written as $M^{ch}_{\sqrt{s}}=M^{ch}_{res,\sqrt{s}}+M^{ch}_{ctm,\sqrt{s}},$ (2) in which $M^{ch}_{res,3670}=0,$ and $M^{ch}_{ctm,3670}=M^{ch}_{ctm,3770}\cdot f_{d}$ at the energy of $\sqrt{s}$=3.670 GeV, where $f_{d}=3.773^{3}/3.670^{3}$ is the scaling factor for the $1/s^{3}$ energy dependence of the cross section. The amplitudes $M^{ch}_{ctm,3770}$ and $M^{ch}_{res,3770}$ defined at $\sqrt{s}$=3.773 GeV for all of the channels are listed in Tab. 1. In the table, $X_{\eta}={\rm{cos}}(54.736^{o}+\theta_{P})$, $Y_{\eta}={\rm{sin}}(54.736^{o}+\theta_{P})$, $X_{\eta^{\prime}}=-{\rm{sin}}(54.736^{o}+\theta_{P})$ and $Y_{\eta^{\prime}}={\rm{cos}}(54.736^{o}+\theta_{P})$, which are the same as those given in Ref. Seiden_1988 . Those amplitudes completely control the correlations among the VP channel productions. If any significant decay amplitude $M^{ch}_{res,3773}$ given in Tab. 1 has been measured to be non- zero, which means that the measured couplings ${\bf{g}}$, ${\bf{g}}_{s}$ etc. are non-zero, this indicates that $\psi(3770)$ has a significant branching fraction for decay to the non-charmed channel “$ch$”. Table 2: The numbers of the events observed by CLEO-c, the subscripts $sw$ and $sb$ indicate the signal window and the side band window, respectively; the upper script 3.67 and 3.77 indicate the c.m. energies. Channel($ch$) \| $N^{3.67}_{sw}$ \| $N^{3.67}_{sb}$ \| $N^{3.77}_{sw}$ \| $N^{3.77}_{sb}$ \| $\epsilon(\%)$ ---\|---\|---\|---\|---\|--- $\rho\pi$ \| 43 \| 5.4 \| 314 \| 44.8 \| 26.3 $~{}\rho^{0}\pi^{0}$ \| 21 \| 3.4 \| 130 \| 33.0 \| 32.5 $~{}\rho^{\pm}\pi^{\mp}$ \| 22 \| 2.0 \| 184 \| 11.8 \| 23.1 $\omega\pi^{0}$ \| 54 \| 6.2 \| 696 \| 39.2 \| 19.0 $\phi\pi^{0}$ \| 1 \| 1.6 \| 2 \| 40 \| 16.5 $\rho^{0}\eta$ \| 36 \| 3.1 \| 508 \| 31.0 \| 19.6 $\omega\eta$ \| 4 \| 0.0 \| 15 \| 6.0 \| 9.9 $\phi\eta$ \| 5 \| 1.0 \| 132 \| 15.9 \| 11.0 $\rho^{0}\eta^{\prime}$ \| 1 \| 0.0 \| 27 \| 0.9 \| 2.9 $\omega\eta^{\prime}$ \| 0 \| 0.0 \| 2 \| 0.0 \| 1.5 $\phi\eta^{\prime}$ \| 0 \| 0.0 \| 9 \| 2.0 \| 1.2 $K^{0}\bar{K}^{0}+c.c.$ \| 38 \| 0.4 \| 501 \| 18.1 \| 8.8 $K^{\pm}K^{\mp}$ \| 4 \| 1.0 \| 36 \| 32.4 \| 16.0 We start the analysis from the observed numbers of the events for the VP channels, their errors and their corresponding detection efficiencies at the energies of $\sqrt{s}$=3.773 GeV and 3.670 GeV, which were published by CLEO-c Collaboration CLEO_c_VP . The numbers and errors of the events are obtained in both the signal windows and the side bands. Tab. 2 shows those numbers and detection efficiencies which are given in Ref. CLEO_c_VP . Taking the numbers of the events from Tab. 2, we obtain the numbers, $N^{obs,ch}_{\sqrt{s}}=(N^{\sqrt{s}}_{sw}-N^{\sqrt{s}}_{sb})_{ch},$ (3) of the observed events at $\sqrt{s}$ for the channel “$ch$”. Usually the determinations of the detection efficiencies and the ISR corrections are all energy dependent and relate to the production line shapes for those channels. We guess that the determinations of the detection efficiencies given in Ref. CLEO_c_VP were done under the assumption of that the continuum cross section line shape is in the $1/s^{3}$ energy dependence and the energy cut is at $\sqrt{s^{\prime}}\geq J/\psi$ mass, where $\sqrt{s^{\prime}}$ is the center of mass energy of the ISR return system. It should be stressed that, for the ISR and FSR (Final State Radiative) corrections in the continuum processes, our calculation gives $\eta_{ctm}$=1.19 at $\sqrt{s}$=3773 GeV and $\eta_{ctm}$=1.11 at $\sqrt{s}$=3670 GeV, while Ref. CLEO_c_VP gives $\eta_{ctm}$=1/1.20=0.833 at both of the two energy points. As for the resonance decay, the ISR correction is quite different from the one for the continuum process. In our calculation, the ISR correction factor for the resonance is $\eta_{res}$=0.824 including the FSR correction at $\sqrt{s}$ =3.773 GeV for all channels. In our ISR correction calculations, the $v^{3}$ phase space dependences have been taken into account, where $v=\sqrt{[1-(m_{V}+m_{P})^{2}/s][(1-(m_{V}-m_{P})^{2}/s]}$ is the velocity of the vector daughter in the CM decay system. However, in the calculation of $\eta_{ctm}$’s, a mean production threshold of the channels has been set to serve as the common threshold for all of those channels. For this reason, the channel dependences of the corrections $\eta_{ctm}$’s are ignored in this analysis. For the calculation of the contribution of the interference terms among the amplitudes describing different processes, we have to know their own detection efficiencies and ISR corrections. In the analysis we simply take the geometric average of the related coefficients as the effective ones. Using the amplitudes of VP channels in Eq.(2) and Tab. 1, the efficiencies for those channels, the ISR and FSR correction as given above, the $v^{3}$ phase space dependences and the luminosities $L_{\sqrt{s}}$ accumulated at the two collision energy points of $\sqrt{s}$ =3.773 GeV and 3.670 GeV, we can calculate the expected numbers $N^{exp,ch}_{3670}$ and $N^{exp,ch}_{3773}$ of the event yields for those channels under the simplification assumption concerning the interference terms mentioned above. For channel “$ch$” $N^{exp,ch}_{3670}=L_{3670}\cdot v^{3}_{3670}\cdot\epsilon^{ch}_{ctm}\cdot\eta_{ctm,3670}^{ch}\cdot\|M^{ch}_{ctm,3670}\|^{2}$ and $\displaystyle N^{exp,ch}_{3773}=L_{3773}\cdot v^{3}_{3773}\times$ $\displaystyle\left\|\sqrt{\epsilon^{ch}_{res}\cdot\eta_{res}^{ch}}\cdot M^{ch}_{res,3773}+\sqrt{\epsilon^{ch}_{ctm}\cdot\eta_{ctm}^{ch}}\cdot M^{ch}_{ctm,3773}\right\|^{2}.$ Comparing the numbers $N^{obs,ch}_{\sqrt{s}}$ defined in Eq.(3) with the expected one $N^{exp,ch}_{\sqrt{s}}$, we get the equation set $N^{exp,ch}_{\sqrt{s}}=N^{obs,ch}_{\sqrt{s}}$ (4) in which $ch$ = $\rho\pi$, $K^{}(892)\overline{K}$+c.c., $\omega\pi^{0}$, $\rho\eta$, $\rho\eta^{\prime}$, $\omega\eta$,$\omega\eta^{\prime}$,$\phi\eta$, $\phi\eta^{\prime}$ and $\phi\pi^{0}$, $\sqrt{s}$=3.670 and 3.773 GeV. Because of the zero observation and zero expectation for the $\phi\pi^{0}$ channel, listed in Tab. 2 and Tab. 1, we can get rid of this channel in our analysis. So we only focus our attention on the rest eleven channels. In the maximum likelihood fit, leaving the parameters ${\bf g}$, ${\bf g}_{s}$, ${\bf e_{0}}$, ${\bf e_{1}}$, $r$, $s_{e}$, $s_{\rm V}$, $s_{\rm P}$, $\theta_{P}$ $\delta_{1}$ and ${\rm{cos}}\theta_{0}$ free, we can solve the Eq.(4) by maximizing the probability function $Prob=\prod_{ch,\sqrt{s}}^{2N_{ch}}P_{ch}(N^{obs,ch}_{\sqrt{s}},N^{exp,ch}_{\sqrt{s}})$ (5) where $P_{ch}(N^{obs,ch}_{\sqrt{s}},N^{exp,ch}_{\sqrt{s}})$ is the probability of finding $N^{obs,ch}_{\sqrt{s}}$ events with the assumed mean number $N^{exp,ch}_{\sqrt{s}}$ at the energy $\sqrt{s}$, and get the solution of Eq.(4) with most probable values of the parameters which control the VP channel production at $\sqrt{s}$=3.773 and 3.670 GeV. From the solutions of ${\bf{g}}$, $r$, ${\bf{e}}_{0}$ and/or ${\bf{e}}_{1}$ etc, one can get the Born cross sections for the VP channel production at $\sqrt{s}$=3.773 or 3.670 GeV. For example, the Born cross section for decay channel $\psi(3770)\rightarrow$ “$ch$” at $\sqrt{s}$ =3.773 GeV can be written as $\sigma^{ch}_{res}=v^{3}_{3773}\cdot\|M^{ch}_{res,3773}\|^{2}.$ (6) The total Born cross section for $e^{+}e^{-}$ annihilation to the channel “$ch$” at $\sqrt{s}$=3.773 GeV is given by $\sigma^{ch}_{T}=v^{3}_{3773}\cdot\|M^{ch}_{res,3773}+M^{ch}_{ctm,3773}\|^{2}.$ (7) While the Born cross section for the channel “$ch$” in continuum production at $\sqrt{s}$ is then given by $\sigma^{ch}_{ctm,\sqrt{s}}=v^{3}_{\sqrt{s}}\cdot\|M^{ch}_{ctm,\sqrt{s}}\|^{2}.$ (8) ## III The results According to the different coupling configurations of the amplitudes for the VP channels, shown in Eq.(2) and Tab. 1, the twelve VP channels can be divided into three sub-sets. The first one consists of the channels without E-M isovector components, such as $\rho\pi$, $\omega\eta$, $\phi\eta$, $\omega\eta^{\prime}$ and $\phi\eta^{\prime}$. The second one consists of the E-M isovector component only, which is the pure E-M channels $\omega\pi^{0}$, $\rho\eta$, and $\rho\eta^{\prime}$. The third subset includes only the channels $K^{}\overline{K}+c.c.$, for which the amplitudes involve all of the two E-M coupling parts and the strong coupling component as well as their interferences in the VP production. Since there is no common coupling parameter in the first two sub-sets despite of the pseudoscalar mixing angle $\theta_{P}$, one can simply try to solve Eq.(4) separately in the two sub- sets at first. ### III.1 The channels without isovector E-M amplitude We start the analysis from the first set in which the channels are without the isovector E-M amplitude contribution. From the numbers of events, $N^{obs,ch}_{\sqrt{s}}$, observed at the two energy points of $\sqrt{s}$=3.671 GeV and 3.773 GeV for the six channels $\rho^{0}\pi^{0}$, $\rho^{\pm}\pi^{\mp}$, $\omega\eta$, $\phi\eta$, $\omega\eta^{\prime}$ and $\phi\eta^{\prime}$, we solve the Eq.(4). Leaving all of the related parameters, such as $\|{\bf g}\|$, ${\bf g}_{s}$, $\|{\bf e}_{0}\|$, $cos\theta_{0}$, $s_{e}$, $\theta_{P}$, $r$, $s_{V}$ and $s_{P}$ free, the fitting yields $\|{\bf{g}}\|=2.752\pm{0.291}$, ${\bf{g}}_{s}=1.279\pm{0.706}$, $\rm{cos}\theta_{0}=-0.935\pm{0.141}$, $\|{\bf{e}}_{0}\|=1.495\pm{0.126}$, $s_{e}=0.264\pm{0.247}$, $r=-0.236\pm{0.070}$, $\theta_{\rm P}=(-17.83\pm{12.12})^{o}$, $s_{V}=0.092\pm{0.941}$ and $s_{P}=-0.091\pm{0.587}$. Indeed, as hinted by the measurements from the CLEO-c CLEO_c_VP and guessed above, the solution of $\rm{cos}\theta_{0}=-0.935\pm{0.141}$ shows that ${\bf{g}}$ and ${\bf{e}}_{0}$ are almost opposite in the VP production at $\psi(3770)$ resonance peak. Associated with earlier measurements in the J/$\psi$ production and decays, in which the E-M decay amplitudes and the strong decay amplitude are more likely with the phase difference of $\phi_{e,res}-\phi_{g}\sim+90^{o}$ kopke ; achasov , and the phase difference between the continuum amplitude and the one of resonance E-M decay at the resonance peak is $\phi_{e,ctm}-\phi_{e,res}=+90^{o}$ too BES_95 ; Wang_03 , the measurement of the phase difference $\theta_{0}=\phi_{e,ctm}-\phi_{g}=(160^{+20}_{-17})^{o}$ between ${\bf{e}}_{0}$ and ${\bf{g}}$ here is reasonable. The $90^{o}$ phase difference between the E-M decay amplitudes and the strong decay amplitude shown in J/$\psi$ decays kopke ; achasov is mainly due to the exist of an original $90^{o}$ phase difference from the short distance force range in the on-shell three gluon annihilation of $1^{--}$ quarkonium states. This argument should be kept in the $\psi(3770)$ decays. Out of the hard quark-gluon interaction level the strong phase shift of long distance final state interaction should essentially be small zheng . The resolutions of $s_{V}=0.092\pm{0.941}$ and $s_{P}=-0.091\pm{0.587}$, which are consistent with zero, seem to mean the small nonet symmetry breaking between the singlet and octet wave functions. Table 3: The fitted parameters. \| \| Solution 1 \| Solution 2 \| Solution 3 \| Solution 4 ---\|---\|---\|---\|---\|--- \| \| the channels \| the channels \| all channels \| all channels \| \| without isovector \| with isovector \| with \| without \| \| E-M component \| E-M component \| counter terms \| counter terms $\|{\bf g}\|$ \| \| $2.64^{+0.41}_{-0.54}$ \| 2.64(fixed) \| $2.67^{+0.38}_{-0.32}$ \| $2.43^{+0.31}_{-0.29}$ ${\bf g}_{s}$ \| \| $1.23^{+0.67}_{-0.74}$ \| $-1.79^{+0.42}_{-0.36}$ \| $1.18^{+0.71}_{-0.78}$ \| $-1.23^{+0.52}_{-0.49}$ $s_{g}$ \| \| $0.27^{+0.14}_{-0.13}$ \| $0.84^{+0.17}_{-0.20}$ \| $0.28\pm{0.16}$ \| $0.75\pm{0.42}$ ${\bf g}_{s}^{\rm{add}}$ \| \| \| \| $-1.51^{+0.43}_{-0.40}$ \| 0.0(fixed) $\rm{cos}\theta_{0}$ \| \| $-0.91^{+0.12}_{-0.13}$ \| -0.91(fixed) \| $-0.91\pm{0.07}$ \| $-0.86^{+0.06}_{-0.04}$ $\|{\bf e}_{0}\|$ \| \| $1.49^{+0.12}_{-0.13}$ \| 1.49(fixed) \| $1.49^{+0.12}_{-0.13}$ \| $1.44^{+0.13}_{-0.14}$ $s_{e}$ \| \| $0.25^{+0.19}_{-0.22}$ \| $-0.35\pm{0.11}$ \| $0.20^{+0.20}_{-0.23}$ \| $-0.29^{+0.13}_{-0.14}$ $a^{\mu}$ \| \| \| \| $-0.55^{+2.38}_{-0.22}$ \| 0.0(fixed) $\|{\bf e}_{1}\|$ \| \| \| $1.223^{+0.016}_{-0.018}$ \| $1.226^{+0.017}_{-0.019}$ \| $1.22\pm{0.02}$ $\delta_{1}$ \| \| \| $(7.39^{+5.90}_{-7.69})^{o}$ \| $(6.70^{+6.59}_{-8.19})^{o}$ \| $(7.62^{+8.88}_{-6.02})^{o}$ $\theta_{P}$ \| \| $(-18.7^{+5.8}_{-4.7})^{o}$ \| $(-23.2\pm{2.2})^{o}$ \| $(-22.5^{+1.9}_{-2.0})^{o}$ \| $(-24.1^{+2.1}_{-2.3})^{o}$ $r$ \| \| $-0.24\pm{0.06}$ \| \| $-0.26\pm{0.06}$ \| $-0.32^{+0.09}_{-0.08}$ $s_{V}$ \| \| 0.0(fixed) \| \| 0.0(fixed) \| $-0.42^{+0.36}_{-0.59}$ $s_{P}$ \| \| 0.0(fixed) \| \| 0.0(fixed) \| $0.45\pm{0.70}$ $\chi^{2}/n_{dof}$ \| \| 4.63/5=0.93 \| 3.98/5=0.80 \| 9.26/11=0.84 \| 13.2/11=1.20 So we can assume that the nonet symmetry maintains for the wave functions, which means that the nonet symmetry violation is only in the DOZI coupling. If we fix $s_{V}$= $s_{P}$ =0 in the fit, we get the parameters $\|{\bf g}\|$, ${\bf g}_{s}$, $\|{\bf e}_{0}\|$, $cos\theta_{0}$, $s_{e}$, $\theta_{P}$ and $r$, which are listed in the second column (“Solution 1”) of Tab. 3, where $\theta_{\rm P}=(-18.73^{+5.8}_{-4.7})^{o}$ indicates $X_{\eta}=Y_{\eta^{\prime}}=0.809^{+0.045}_{-0.064}~{}\rm{and}~{}Y_{\eta}=-X_{\eta^{\prime}}=-0.588^{+0.068}_{-0.079}.$ The fit gives $\chi^{2}/n_{dof}=4.63/5=0.93$, which is also listed in Tab. 3. With the ${\bf{g}}$ and ${\bf{g}}_{s}$, we obtain the $SU(3)$ mass correction $s_{g}=0.268^{+0.140}_{-0.127}.$ Except for the three coupling strengths, $\|{\bf e}_{0}\|$, $\|{\bf g}\|$ and $\|{\bf g}_{s}\|$, which have their own dynamics in the higher energy position, the other relative correction parameters, $s_{g}$, $s_{e}$, $r$ and the mixing angle $\theta_{\rm P}$ obtained from the fit are all reasonable comparing with those obtained from J/$\psi$ decays measured by Mark-III and DM2 MARK3_DM2 . However, from Eqs.(2) and (6), we find that the strong decay coupling $\|{\bf g}\|$ gives quite large cross section, branching fraction and the partial width for $\psi(3770)\rightarrow\rho\pi$ decay, which are $\left.\begin{aligned} &\sigma^{\rho\pi}_{res}=(18.2^{+6.11}_{-6.70})~{}~{}{\rm{pb}}\\\ &B^{\rho\pi}=(1.83^{+0.62}_{-0.67})\times 10^{-3}\\\ &\Gamma^{\rho\pi}=49.7^{+16.9}_{-18.3}~{}\rm{keV}.\end{aligned}\right\\}$ (9) The partial width is almost in two order of magnitude higher than that of the conventional typical partial width of the J/$\psi$ VP decays. The latter one is at the order of 1 keV. If we assume that the fraction of the width of $\psi(3770)\rightarrow\rho\pi$ to the width of $\psi(3770)\rightarrow light~{}hadrons$ is roughly as the same as the one in the J/$\psi$ decays, the huge partial width of $\psi(3770)\rightarrow\rho\pi$ would predict that about $10\%$ of $\psi(3770)$ decays to non-$D\overline{D}$ final states. This predicted branching fraction for $\psi(3770)\rightarrow{non-}D\overline{D}$ is almost as the same as the one measured by BES-II Collaboration PhysRevLett_97_121801 ; PhysLettB_641_145 . The large cross section $\sigma^{\rho\pi}_{res}$ as shown in Eq.(9) is a factor of more than 3 of the total $\rho\pi$ production cross section, $\sigma^{\rho\pi}_{T}=5.35^{+1.45}_{-1.58}~{}~{}{\rm{pb}},$ (10) obtained from Eq.(7). This cross section is consistent with the measurement reported in Ref. CLEO_c_VP , (see Tab. 4). In the Tab. 4 we list all of the production cross sections and the branching fractions for $\psi(3770)\rightarrow$ VP predicted in this work, and also listed the production cross sections of the VP channels given in Ref. CLEO_c_VP as the comparison. As for the $\rho\pi$ production cross section at $\sqrt{s}=3.670$ GeV in this measurement, we get $\sigma^{\rho\pi}_{T,3670}=6.34^{+1.05}_{-1.03}~{}~{}{\rm{pb}},$ (11) which is indeed higher than the one given in Eq.(10) at resonance peak. Because of the different determinations of the ISR corrections in Ref. CLEO_c_VP and in our work, the decrease of the Born cross section of $\rho\pi$ channel at $\psi(3770)$ resonance peak is not so large as that obtained by CLEO-c Collaboration CLEO_c_VP (see also Tab. 4). Owing to the large cancellation between the two amplitudes ${\bf g}$ and ${\bf e}_{0}$, the large cross section $\sigma^{\rho\pi}_{res}$ for $\psi(3770)\rightarrow\rho\pi$ disappeared without the global amplitude analysis. ### III.2 The pure E-M channels $\omega\pi^{0}$, $\rho^{0}\eta$ and $\rho^{0}\eta^{\prime}$ plus the channel $K^{}\overline{K}$+c.c We consider together the last two sub-sets of the channels in which the amplitudes of E-M production contain isovector component. The isoscalar E-M component and the strong decay coupling only serve in the channel $K^{}\overline{K}$+c.c., and the strong coupling here is with a combined form ${\bf{g}}_{K}=({\bf g}+{\bf g}_{s})/2$. Inserting the ten numbers of the events observed at the two energy points of 3.670 GeV and 3.773 GeV for the rest five channels into Eq.(4), leaving ${\bf e}_{1}$, $\delta_{1}$, $s_{e}$, $\theta_{\rm P}$ and ${\bf g}_{K}$ free (or instead of ${\bf g}_{K}$, leaving ${\bf g}_{s}$ free but fixing ${\bf g}$ at some reasonable value such as $\|{\bf g}\|$=2.635 obtained from last solution independently), and fixing $\|{\bf e}_{0}\|$=1.494 and cos$\theta_{0}$=-0.907 obtained also from last solution in assumption of that there is no more correction added to the couplings ${\bf g}$ and $\|{\bf e}_{0}\|$ measured in last solution in the channel $K^{}\overline{K}$, we fit the numbers of events observed in the five channels and obtain the solution of the free parameters. The results are listed in the third column (“Solution 2”) of Tab. 3. The fit gives $\chi^{2}/n_{dof}=3.98/5=0.80$. From above two solutions, we see that the measured pseudoscalar mixing angles $\theta_{\rm P}$ in the two independent measurements are consistent with each other. The isovector E-M component, ${\bf{e}}_{1}$, is really split from the isoscalar one, ${\bf{e}}_{0}$, with almost 2$\sigma$ deviation in magnitude and with a non-zero phase shift difference. However, we note that the $s$ quark strong decay coupling ${\bf{g}}_{s}$ and its E-M coupling correction $s_{e}$ in the solution 2 are both with the negative values, while they are supposed to be positive in the conventional $SU$(3) invariant model with simple static mass corrections. The minus $s_{e}$ is formally very likely to be some anomalous “magnetic moments” term added to the $s$ quark E-M coupling. As for the minus ${\bf g}_{s}$, which is obviously irrelative to the three pure E-M channels, it seems that the $s$ quark strong coupling undergo almost $180^{o}$ phase shift from the conventional $SU(3)$ strong interaction wave function. The odd behavior of the minus ${\bf g}_{s}$ as well as the E-M “anomalous magnetic moments” term for the $s$ quark indicate that there might be some other dynamic sources or more complicated interaction correction contributing to the production of $K^{}\overline{K}$+c.c. in $e^{+}e^{-}$ annihilation. Owing to the cancellation of ${\bf{g}}$ and the opposite ${\bf{g}}_{s}$ (small ${\bf g}_{K}$), the decay cross section, branching fraction and partial width $\left.\begin{aligned} &\sigma^{K^{}\overline{K}~{}}_{res}=(0.558^{+0.702}_{-0.379})~{}~{}{\rm{pb}}\\\ &B^{K^{}\overline{K}~{}}=(0.56^{+0.70}_{-0.38})\times 10^{-4}\\\ &\Gamma^{K^{}\overline{K}~{}}=(1.52^{+1.92}_{-1.03})~{}~{}{\rm{keV}},\end{aligned}\right\\}$ (12) for $\psi(3770)\rightarrow K^{}\overline{K}$+c.c in this solution are quite small comparing with that of $\psi(3770)\rightarrow\rho\pi$ measured in last solution given in Eq.(9). These results are also listed in Tab. 4. The remarkable increase of the E-M coupling $e_{0}$ by negative correction $s_{e}$ and the large interferences between the two E-M amplitudes and the strong amplitudes result a serious asymmetry between the total production cross sections of channels $K^{0}\overline{K}^{0}$+c.c. and $K^{\pm}K^{\mp}$. For example, at $\sqrt{s}=3.773$ GeV the cross sections $\left.\begin{aligned} &\sigma^{K^{0}\overline{K}^{0}~{}+c.c.}_{T}=19.30^{+3.55}_{-1.85}~{}~{}{\rm{pb}}\\\ &\sigma^{K^{\pm}K^{\mp}~{}}_{T}<0.50~{}~{}({\rm{pb,~{}~{}at~{}90\%~{}confidence~{}level}}),\end{aligned}\right\\}$ (13) are consistent with the observed values reported in Refs. CLEO_c_VP and CLEO_3686_VP . Using the parameters $\|{\bf e}_{1}\|$ and $\theta_{\rm P}$, and from Eq.(8) we get the production cross sections of the three pure E-M channels, $\left.\begin{aligned} &\sigma^{\omega\pi^{0}}_{ctm}=(11.74^{+0.31}_{-0.34})~{}~{}{\rm{pb}}\\\ &\sigma^{\rho\eta}_{ctm}=(8.00{\pm{0.22}})~{}~{}{\rm{pb}}\\\ &\sigma^{\rho\eta^{\prime}}_{ctm}=(2.57{\pm{0.28}})~{}~{}{\rm{pb}}.\end{aligned}\right\\}$ (14) These are also listed in Tab. 4. The E-M production cross sections and the total production cross sections obtained in this subsection are all systematically lower than those measured by CLEO-c Collaboration CLEO_c_VP by about $30\%$, (see Tab. 4). Those differences are also due to the different determinations of ISR corrections in the two works as mentioned above. ### III.3 The global fit including all of the measured VP channels We can introduce two additional effective counter terms, ${\bf{g}}^{\rm{add}}_{s}$ and ${\bf{a}}^{\mu}$, to compensate the odd behavior appeared in the channel $K^{}(892)\overline{K}$+c.c. for both the strong coupling and E-M coupling of $s$ quark. We can simply assume that ${\bf{g}}^{\rm{add}}_{s}$ and ${\bf g}$, ${\bf{a}}^{\mu}$ and ${\bf e}_{0}$ are collinear, respectively. In this case, we define the amplitudes of channels $K^{}(892)\overline{K}$+c.c. as $\displaystyle M^{K^{+}K^{-},K^{-}K^{+}}_{res,3770}=({\bf{g}}+{\bf{g}}_{s})/2+{\bf{g}}^{\rm{add}}_{s},$ (15) $\displaystyle M^{K^{+}K^{-},K^{-}K^{+}}_{ctm,3770}=-{\bf{e}}_{0}(1/2-s_{e})+3/2{\bf e}_{1}+{\bf{a}}^{\mu},$ $\displaystyle M^{K^{0}\overline{K}^{0}+c.c.}_{res,3770}=({\bf{g}}+{\bf{g}}_{s})/2+{\bf g}^{\rm{add}}_{s},$ $\displaystyle M^{K^{0}\overline{K}^{0}+c.c.}_{ctm,3770}=-{\bf{e}}_{0}(1/2-s_{e})-3/2{\bf e}_{1}+{\bf{a}}^{\mu},$ as given in Tab. 1. With those amplitudes we globally solve the Eq.(4) with all of the observed channels. Fixing $s_{V}$= $s_{P}$ =0 and leaving all of the other parameters including ${\bf{g}}^{\rm{add}}_{s}$ and ${\bf{a}}^{\mu}$ free, the fit gives the most probable values of parameters $\|{\bf{g}}\|$, ${\bf{g}}_{s}$, ${\bf{g}}^{\rm{add}}_{s}$, $\rm{cos}\theta_{0}$, $\|{\bf{e}}_{0}\|$, $s_{e}$, ${\bf{a}}^{\mu}$ $\|{\bf{e}}_{1}\|$, $\delta_{1}$, $r$, $\theta_{\rm P}$. Those results are listed in the fourth column (“Solution 3”) of Tab. 3 with $\chi^{2}/n_{dof}$=9.26/11=0.84. From the parameters of solution 3 and Eqs.(2),(6),(7), (8) and (15), we can get the production cross sections of the twelve VP channels, including the zero measurement of channel $\phi\pi^{0}$. For example, for the channels $\rho\pi$ and $K^{}(892)\overline{K}$+c.c., the decay cross sections from $\psi(3770)$ resonance, and their decay branching fractions and partial widths can be calculated as $\left.\begin{aligned} &\sigma^{\rho\pi}_{res}=(18.68^{+4.12}_{-4.19})~{}~{}{\rm{pb}}\\\ &B^{\rho\pi}=(1.87^{+0.41}_{-0.42})\times 10^{-3}\\\ &\Gamma^{\rho\pi}=51.1^{+11.2}_{-11.5}~{}\rm{keV}\\\ &\sigma^{K^{}\overline{K}~{}}_{res}=(0.56^{+1.28}_{-0.22})~{}~{}{\rm{pb}}\\\ &B^{K^{}\overline{K}~{}}=(0.56^{+1.28}_{-0.22})\times 10^{-4}\\\ &\Gamma^{K^{}\overline{K}~{}}=(1.52^{+3.50}_{-0.60})~{}~{}{\rm{keV}}.\end{aligned}\right\\}$ (16) And the production cross sections of the three pure E-M channels can be calculated as $\left.\begin{aligned} &\sigma^{\omega\pi^{0}}_{ctm}=(11.79^{+0.34}_{-0.36})~{}~{}{\rm{pb}}\\\ &\sigma^{\rho\eta}_{ctm}=(7.92\pm{0.41})~{}~{}{\rm{pb}}\\\ &\sigma^{\rho\eta^{\prime}}_{ctm}=(2.69\pm{{0.29}})~{}~{}{\rm{pb}}.\end{aligned}\right\\}$ (17) The measured values in Eqs.(16),(17) are all consistent with those measured in the last two subsections. As for the measurements of other channels, we have $\left.\begin{aligned} &B^{\omega\eta}=(2.12^{+0.75}_{-0.48})\times 10^{-4}\\\ &B^{\omega\eta^{\prime}}<0.25\times 10^{-4},~{}({\rm~{}{at~{}90\%~{}c.l.}})\\\ &B^{\phi\eta}=(0.87^{+0.75}_{-0.58})\times 10^{-4}\\\ &B^{\phi\eta^{\prime}}<0.60\times 10^{-4},~{}({\rm~{}{at~{}90\%~{}c.l.}}),\end{aligned}\right\\}$ (18) obtained in the assumption of that there exist the counter terms ${\bf{g}}^{\rm{add}}_{s}$ and ${\bf{a}}^{\mu}$ and $s_{V}$=$s_{P}$=0. If we fix the strong couplings ${\bf g}$ and ${\bf g}_{s}$ to be the values which force the partial widths of $\rho\pi$ and $K^{}\overline{K}$ etc. to be at the order of 1 keV which is at the same order of the J/$\psi$ VP decay coupling dynamics, the fitted $\chi^{2}$ is 47.27 for 15 degree of freedom which corresponds to 5.3 standard deviation worse than that of solution 3. If there essentially were no the two counter terms specially for channel $K^{}\overline{K}$+c.c (${\bf{g}}^{\rm{add}}_{s}={\bf{a}}^{\mu}=0$), i.e. there existed the negative coupling ${\bf{g}}_{s}$ and negative correction $s_{e}$ universally allowed on all of other relative channels with the value $s$ quark, the fit gives a somehow poor solution with large $\chi^{2}$=19.2 for 13 degree of freedom, which gives 2.7 standard deviation away from the counter term assumption (solution 3). However, in this case, the effects of the opposite ${\bf{g}}_{s}$ and negative $s_{e}$ in the related channels of ${\phi\eta}$ and ${\phi\eta}^{\prime}$ might be compensated by somehow larger nonet symmetry breaking with non-zero $s_{V}$ and $s_{P}$. Ignoring the terms ${\bf{g}}^{\rm{add}}_{s}$ and ${\bf{a}}^{\mu}$ used in the last solution and leaving the nonet symmetry breaking parameters $s_{V}$ and $s_{P}$ free, we solve the Eq.(4) again to obtain a new solution. The results are listed in the fifth column (“Solution 4”) of Tab. 3. The fit gives $\chi^{2}/{\rm{n_{d.o.f.}}}$=13.2/11=1.2, which is a little bit higher than those in other solutions. However, the solution still has the statistical significance more than 5.6 standard deviation to the one obtained in the assumption that only the E-M components (i.e. the ${\bf e}_{0}$, ${\bf e}_{1}$ and $\delta_{1}$, and the $\eta-\eta^{\prime}$ mixing angle) act on the VP channel production. As for the solution 4, we obtain the unusual solution again, which is like solution 2 with the negative ${\bf{g}}_{s}$ and negative E-M correction $s_{e}$ for the $s$ quark couplings. The parameters of $r$, $s_{\rm V}$ and $s_{\rm P}$ which associate with the nonet symmetry breaking measurements are now with unexpected larger values. Nevertheless, as guessed above, those unusual numbers may suggest more complicate dynamics for the $s$ quark production either in the channel $K^{}\overline{K}$ or in all of the relative channels, involving value $s$ quark production. This solution predicts the decay cross sections, branching fractions and decay width of channels $\rho\pi$ and $K^{}(892)\overline{K}$ and the production cross sections of three pure E-M production channels $\omega\pi^{0}$, $\rho\eta$ and $\rho\eta^{\prime}$, $\left.\begin{aligned} &\sigma^{\rho\pi}_{res}=(15.50_{-3.52}^{+4.21})~{}~{}{\rm{pb}}\\\ &B^{\rho\pi}=(1.55^{+0.42}_{-0.35})\times 10^{-3}\\\ &\Gamma^{\rho\pi}=42.4^{+11.5}_{-9.6}~{}\rm{keV}\\\ &\sigma^{K^{}\overline{K}~{}}_{res}=(1.14^{+1.12}_{-0.57})~{}~{}{\rm{pb}}\\\ &B^{K^{}\overline{K}~{}}=(1.14^{+1.12}_{-0.57})\times 10^{-4}\\\ &\Gamma^{K^{}\overline{K}~{}}=(3.11^{+3.06}_{-1.56})~{}~{}{\rm{keV}}\\\ &\sigma^{\omega\pi^{0}}_{ctm}=(11.67^{+0.35}_{-0.38})~{}~{}{\rm{pb}}\\\ &\sigma^{\rho\eta}_{ctm}=(8.10^{+0.60}_{-0.61})~{}~{}{\rm{pb}}\\\ &\sigma^{\rho\eta^{\prime}}_{ctm}=(2.43^{0.39}_{-0.37})~{}~{}{\rm{pb}},\end{aligned}\right\\}$ (19) which are consistent with those obtained in the last solution with the counter terms. However, for the channels $\omega\eta$, $\omega\eta^{\prime}$, $\phi\eta$ and $\phi\eta^{\prime}$, we obtain the branching fractions of $\left.\begin{aligned} &B^{\omega\eta}=(0.92^{+0.75}_{-0.92})\times 10^{-4}\\\ &B^{\omega\eta^{\prime}}<1.66\times 10^{-4},~{}({\rm~{}{at~{}90\%~{}c.l.}})\\\ &B^{\phi\eta}<1.57\times 10^{-4},~{}({\rm~{}{at~{}90\%~{}c.l.}})\\\ &B^{\phi\eta^{\prime}}=3.46^{+8.93}_{-1.75}\times 10^{-4},\end{aligned}\right\\}$ (20) which are quite different comparing with those given in Eq.(18). As for the branching fractions for $\psi(3770)\rightarrow\phi\eta$, the measured values obtained by this analysis are also somehow inconsistent with that measured by CLEO-c Collaboration CLEO_c_VP , (see Tab. 4). However both the results show the large $\psi(3770)$ VP decay couplings. The summed total cross sections over all VP channels of $\sigma^{ch}_{T}$ at both energy points of $\sqrt{s}$=3.670 GeV and 3.773 GeV are listed in Tab. 4. The summed values of solution 3 and solution 4 as well as the results of CLEO-c measurement show in almost equals at the two energy points. The real resonance decays have been “hidden”. This is mainly owing to the destructive nature of interference between the couplings ${\bf g}$ and ${\bf e}_{0}$. Table 4: Summary of the results. The errors for our measurements are statistical only. The branching fraction of channel $\phi\eta$ measured by CLEO-c Collaboration is directly from work CLEO_c_VP and the upper limits of other channels measured by CLEO-c Collaboration are from the “Method II” of paper CLEO_c_VP \| \| Solution 1 \| Solution 2 \| Solution 3 \| Solution 4 \| CLEO-c ---\|---\|---\|---\|---\|---\|--- $\rho\pi$ \| $\sigma_{3670}$ (pb) \| $6.86^{+1.14}_{-1.12}$ \| \| $6.89^{+1.61}_{-1.60}$ \| $6.43^{+1.53}_{-1.35}$ \| $8.0^{+1.7}_{-1.4}\pm 0.9$ \| $\sigma_{T}$ (pb) \| $5.35^{+1.45}_{-1.58}$ \| \| $5.42^{+3.38}_{-2.00}$ \| $5.08^{+3.57}_{-1.26}$ \| $4.4\pm 0.3\pm 0.5$ \| $\sigma_{R}$ (pb) \| $18.24^{+6.11}_{-6.70}$ \| \| $18.68^{+4.56}_{-3.36}$ \| $15.50^{+5.04}_{-2.96}$ \| $<0.04$ \| Br $10^{-3}$ \| $1.83^{+0.61}_{-0.67}$ \| \| $1.87^{+0.46}_{-0.34}$ \| $1.55^{+0.50}_{-0.30}$ \| $<0.004$ $K^{0}\bar{K}^{0}+c.c.$ \| $\sigma_{T}$ (pb) \| \| $19.30^{+3.55}_{-1.85}$ \| $19.34^{+11.29}_{-6.78}$ \| $19.44\pm 3.61$ \| $23.3\pm 1.1\pm 3.1$ \| $\sigma_{R}$ (pb) \| \| $0.28^{+0.37}_{-0.19}$ \| $0.28^{+0.64}_{-0.11}$ \| $0.57^{+0.56}_{-0.29}$ \| $<20.8$ \| Br $10^{-4}$ \| \| $0.28^{+0.37}_{-0.19}$ \| $0.28^{+0.64}_{-0.11}$ \| $0.57^{+0.56}_{-0.29}$ \| $<20.8$ $K^{\pm}K^{\mp}$ \| $\sigma_{T}$ (pb) \| \| $<0.502$ \| $<2.59$ \| $<0.77$ \| $<0.6$ \| $\sigma_{R}$ (pb) \| \| $0.28^{+0.37}_{-0.19}$ \| $0.28^{+0.64}_{-0.11}$ \| $0.57^{+0.56}_{-0.29}$ \| $<0.1$ \| Br $10^{-4}$ \| \| $0.28^{+0.37}_{-0.19}$ \| $0.28^{+0.64}_{-0.11}$ \| $0.57^{+0.56}_{-0.29}$ \| $<0.1$ $\omega\pi^{0}$ \| $\sigma_{T}$ (pb) \| \| $11.74^{+0.31}_{-0.34}$ \| $11.79^{+0.33}_{-0.36}$ \| $11.67^{+0.36}_{-0.38}$ \| $14.6\pm 0.6\pm 1.5$ \| $\sigma_{R}$ (pb) \| \| $0.0$ \| $0.0$ \| $0.0$ \| $<0.06$ \| Br $10^{-4}$ \| \| $0.0$ \| $0.0$ \| $0.0$ \| $<0.06$ $\rho^{0}\eta$ \| $\sigma_{T}$ (pb) \| \| $8.00\pm 0.22$ \| $7.92\pm 0.41$ \| $8.10\pm 0.44$ \| $10.3\pm 0.5\pm 1.0$ \| $\sigma_{R}$ (pb) \| \| $0.0$ \| $0.0$ \| $0.0$ \| $<1.3$ \| Br $10^{-4}$ \| \| $0.0$ \| $0.0$ \| $0.0$ \| $<1.3$ $\rho^{0}\eta^{\prime}$ \| $\sigma_{T}$ (pb) \| \| $2.57\pm 0.28$ \| $2.69\pm 0.29$ \| $2.43\pm 0.32$ \| $3.8^{+0.9}_{-0.8}\pm 0.6$ \| $\sigma_{R}$ (pb) \| \| $0.0$ \| $0.0$ \| $0.0$ \| $<0.4$ \| Br $10^{-4}$ \| \| $0.0$ \| $0.0$ \| $0.0$ \| $<0.4$ $\omega\eta$ \| $\sigma_{T}$ (pb) \| $0.45^{+1.07}_{-0.23}$ \| \| $0.38^{+0.62}_{-0.20}$ \| $0.32^{+0.42}_{-0.10}$ \| $0.4\pm 0.2\pm 0.1$ \| $\sigma_{R}$ (pb) \| $2.19^{+1.13}_{-0.74}$ \| \| $2.12\pm 0.21$ \| $0.92^{+0.92}_{-0.77}$ \| $<0.1$ \| Br $10^{-4}$ \| 2.19 \| \| $2.12\pm 0.21$ \| $0.92^{+0.92}_{-0.77}$ \| $<0.1$ $\omega\eta^{\prime}$ \| $\sigma_{T}$ (pb) \| $0.44^{+0.46}_{-0.26}$ \| \| $0.59^{+0.59}_{-0.39}$ \| $0.44^{+1.84}_{-0.41}$ \| $0.6^{+0.8}_{-0.3}\pm 0.6$ \| $\sigma_{R}$ (pb) \| $<0.33$ \| \| $<0.27$ \| $<0.47$ \| $<1.9$ \| Br $10^{-4}$ \| $<0.33$ \| \| $<0.27$ \| $<0.47$ \| $<1.9$ $\phi\eta$ \| $\sigma_{T}$ (pb) \| $3.99^{+2.75}_{-1.27}$ \| \| $3.91^{+2.75}_{-1.27}$ \| $3.70^{+2.78}_{-1.02}$ \| $4.5\pm 0.5\pm 0.5$ \| $\sigma_{R}$ (pb) \| $0.84^{+0.79}_{-0.58}$ \| \| $0.87^{+0.85}_{-0.52}$ \| $<1.45$ \| $2.4\pm{0.6}$ \| Br $10^{-4}$ \| $0.84^{+0.79}_{-0.58}$ \| \| $0.87^{+0.85}_{-0.52}$ \| $<1.45$ \| $3.1\pm{0.7}$ $\phi\eta^{\prime}$ \| $\sigma_{T}$ (pb) \| $1.95^{+4.49}_{-2.83}$ \| \| $2.29^{+4.20}_{-1.76}$ \| $1.69^{+1.37}_{-0.51}$ \| $2.5^{+1.5}_{-1.1}\pm 0.4$ \| $\sigma_{R}$ (pb) \| $<0.52$ \| \| $<0.66$ \| $3.46^{+4.90}_{-1.76}$ \| $<3.8$ \| Br $10^{-4}$ \| $<0.52$ \| \| $<0.66$ \| $3.46^{+4.90}_{-1.76}$ \| $<3.8$ All “$ch$” summed \| $\sigma^{all}_{3670}$ (pb) \| \| \| $58.3\pm 6.2$ \| $62.3\pm 6.9$ \| $64.2^{+15.8}_{-6.8}$ \| $\sigma^{all}_{T,3770}$ (pb) \| \| \| $54.4^{+13.0}_{-7.4}$ \| $52.9^{+6.3}_{-4.1}$ \| $64.6^{+4.5}_{-4.3}$ ## IV Discussion and Summary From above global analyses of the production cross sections for twelve channels of $e^{+}e^{-}\rightarrow$ VP measured by CLEO-c Collaboration at the $\psi(3770)$ resonance peak of $\sqrt{s}$ =3.773 GeV and at the energy of $\sqrt{s}$=3.670 GeV in the continuum region, we obtained four different solutions of the parameters to control the VP channel production. From Tab. 4 we see that the measured branching fractions or the production cross sections except for the channels involving the nonet symmetry breaking are all consistent with each other in the four solutions. It is remarkable that the large $SU(3)$ symmetry strong decay strength $\|{\bf g}\|$ leads to the huge branching fraction of a level of $10^{-3}$ for the typical channel $\rho\pi$, which corresponds to the decay width of two order of magnitude higher than that in J/$\psi$ decays. The large strong decay coupling hidden behind the VP channel production in the $e^{+}e^{-}$ annihilation at the $\psi(3770)$ resonance peak might help people to understand the sources of the $\psi(3770)$ non-$D\overline{D}$ decays and give people some useful information to reexplain the long-standing $\rho\pi$ puzzle in the $1^{--}$ charmonium state VP decays. Furthermore, the large strong decay coupling with the opposite $s$-quark strong coupling and the minus E-M correction $s_{e}$ or equivalently, the $s$-quark “anomalous magnetic moments” ${\bf{a}}^{\mu}$, required by the $K^{}\overline{K}$ production as presented in those solutions with different treatments and different assumptions in this work are all unusual comparing with the conventional hard gluon annihilation picture plus single $\psi(3770)$ resonance assumption. If those measurements and analyses are all correct, one has to re-understand the strong interaction dynamics which leads to the large Okubo-Zweig-Iizuka rule violation in the vector meson $\psi(3770)$ decays and the strange behavior of the $s$ quark couplings to the light hadron production in the energy region around $\psi(3770)$ resonance. It seems that people has to seriously consider the role of the long distance strong interaction corrections including the $D(D_{s})$ meson exchange scheme to describe those anomalous phenomena and the large Okubo-Zweig-Iizuka breaking in this energy region, like many authors did Lipkin ; Q.Zhao . The destructive interferences mechanism might manifest the possibility of the significant existences of buried $\psi(3770)$ non-$D\overline{D}$ decays. However, how do people understand the large net inclusive non- $D\overline{D}$ hadron branching fraction of $\psi(3770)$ decays measured by BES Collaboration PhysRevLett_97_121801 ; PhysLettB_641_145 recently. Phenomenologically, for example, if the behavior of the reversed $s$ quark strong coupling appeared in VP channel is still maintained in those cases, it would lead to constructive interference with the parallel continuum E-M amplitudes and cause abundant strangeness meson production at the $\psi(3770)$ peak, resulting in the large net cross section excess. The one of the exceptions is the channel $K^{}\overline{K}$+c.c., in which the E-M production is due to the magnetic moments coupling. The minus E-M coupling correction for $s$ quark or the counter term ${\bf{a}}^{\mu}$ as a special “anomalous magnetic moments” enhances the E-M production at the continuum region and leads to the observation of the equal cross sections of channel $K^{}\overline{K}$+c.c. at the two energy points $\sqrt{s}$=3.670 and 3.773 GeV in work CLEO_c_VP . This argument can be cleared up by coming more precisely experimental measurements. Of course, another probable outlet would be that there are more complicated structures or contents in the $\psi(3770)$ resonance scope which are evident in a measurement of cross sections for $e^{+}e^{-}\rightarrow$ hadrons by the BES Collaboration BES_two ; BES_two_DD . This measurement of the cross section indicates that there are somehow complicate “diresonance” structure instead of the conventional single $\psi(3770)$ resonance assumption BES_two ; voloshin . The exist of the extra substances might respond to the unusual behavior of $\psi(3770)$ VP decays and the measured large non-$D\overline{D}$ branching fraction of $\psi(3770)$ decays PhysRevLett_97_121801 ; PhysLettB_641_145 . ## V Acknowledgments This work is supported in part by the National Natural Science Foundation of China under contract No. 10935007. ## References * (1) G. Rong, (BES Collaboration), Presented at 39th Rencontres de Moriond on Electroweak Interactions and Unified Theories, La Thuile, Aosta Valley, Italy, 21-28 Mar. 2004, p401. * (2) G. Rong, et al., arXiv:hep-ex/0506051. * (3) M. Ablikim et al. (BES Collaboration), Phys. Rev. Lett. 97, 121801 (2006); Phys. Rev. D 76, 122002 (2007). * (4) M. Ablikim et al. (BES Collaboration), Phys. Lett. B 641, 145 (2006); Phys. Lett. B 659, 74 (2008). * (5) G. Rong (BES Collaboration), Proceedings of the 32nd International Conference on High Energy Physics, Beijing, China, 16-22 Aug. (2004), p1200. * (6) N. E. Adams et al., (CLEO Collaboration), Phys. Rev. D 73, 012002 (2006). * (7) P. Wang, C.Z. Yuan and X.H. Mo, Phy. Lett. B 574, 41 (2003). * (8) M. Ablikim et al. (BES Collaboration), Phys. Rev. D 70, 112007 (2004). * (9) N. E. Adams et al., (CLEO Collaboration), Phys. Rev. Lett. 94, 012005 (2005). * (10) H.E. Haber and J. Perrier, Phys. Rev. D 32, 2961 (1985). * (11) A. Seiden, et al., Phys. Rev. D 38, 824 (1988). * (12) L. K$\ddot{\rm{o}}$pke and N. Wermes, Phys. Rept. 174, 67 (1989). * (13) N.N. Achasov hep-ph/0110057. * (14) J.Z. Bai et al. (BES Collaboration) Phys. Lett. B 355, 374 (1995). * (15) P. Wang, Hadron 03, Achaffenburg, Germany, 2003, AIP Conference Proceedings 717, 571 (2004). * (16) H.Q. Zheng, Phys. Lett. B 356, 107 (1995). * (17) R. M. Bartrusaitis, et al., (Mark III Collaboration), Phs. Rev. D 32, 2883 (1985); D. Coffman, et al., (Mark III Collaboration), Phys. Rev. D 38, 2695 (1988); J. Jousset, et al., (DM2 Collaboration), Phys. Rev. D 41, 1389 (1990). * (18) H.J. Lipkin, Nucl. Phys. B 179, 278 (1987). * (19) N.N. Achasov and A.A. Kozhevnikov, Phys. Rev. D 49, 275(1994); Y.J. Zhang, G. Li, Q. Zhao, Phys. Rev. Lett. 102, 172001 (2009); X. Liu, B.Zhang and X.Q. Li, arXiv:0902.0480. * (20) M. Ablikim et al., (BES Collaboration), arXiv:hep-ex/0808.0494; Phys. Rev. Lett. 101, 102004(2008). * (21) M. Ablikim et al., (BES Collaboration), Phys. Lett. B 668, 263(2008). * (22) S.Dubynskiy and M.B. Voloshin Phys. Rev. D 78, 116014(2008).	arxiv-papers	2008-08-01T10:58:26	2024-09-04T02:48:57.093923	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "D. Zhang, G. Rong, J.C. Chen", "submitter": "Jiangchuan Chen", "url": "https://arxiv.org/abs/0808.0091" }
0808.0123	Existence and asymptotics of solutions of the Debye-Nernst-Planck system in $\mathbb{R}^{2}$ Agnieszka Herczak 111Państwowa Wyższa Szkola Zawodowa w Nysie, ul. Grodzka 19, 48-300 Nysa, Poland E-mail: herczak@math.uni.wroc.pl, Michal Olech 222Instytut Matematyczny, Uniwersytet Wroclawski, pl. Grunwaldzki 2/4, 50-384 Wroclaw, Poland; E-mail: olech@math.uni.wroc.pl, The preparation of this article has been partially supported by a KBN/MNiI grant 1 P03A 008 30 ###### Abstract In this paper we investigate a system describing electrically charged particles in the whole space $\mathbb{R}^{2}$. Our main goal is to describe large time behavior of solutions which start their evolution from initial data of small size. This is achieved using radially symmetric self-similar solutions. Key words: parabolic-elliptic system, existence of solutions, asymptotic behavior of solutions, long time behavior of solutions. AMS subject classification: Primary 35B40, 35K15, 35Q99; Secondary 78A35, 92E99. Acknowledgment: Authors are grateful to Professors Piotr Biler and Grzegorz Karch for a very fruitful discussions and their invaluable help. ## 1 Introduction We analyze the Debye-Nernst-Planck system of the form $\begin{split}u_{t}&=\Delta u+\nabla\cdot\big{(}u\nabla\phi_{u}\big{)},\\\ \phi_{u}&=-E_{2}\ast u\text{\quad in \quad}\mathbb{R}^{2},\end{split}$ (1a) where $u(x,t):\mathbb{R}^{2}\times\mathbb{R}^{+}\rightarrow\mathbb{R}$ and $E_{2}(z)=\tfrac{1}{2\pi}\log\|z\|$ is the fundamental solution of the Laplace equation in $\mathbb{R}^{2}$. We supplement this system of equations with the initial condition $u(x,0)=u_{0}(x)\text{\quad in\quad}\mathbb{R}^{2}.$ (1b) One may think about this system as of a single parabolic equation with a nonlocal nonlinearity. The history of studies of this model is quite long. The system was introduced by Nernst and Planck in the nineteenth century, and then modified by Debye in twenties of the twentieth century (see [10]). It describes the evolution of the density $u(x,t)$ of electrically charged particles interacting through the potential $\phi_{u}$. We encounter systems of this type, also for several unknown densities, in the electrochemistry, theory of semiconductors and physics of plasma [1, 9, 16, 20]. Many different methods have been used for the analysis of this system since that time (see [2, 3, 4, 5] and references therein). However some problems still remain unsolved. Similar problems where the first equation of (1a) is replaced by $\phi_{u}=E_{2}\ast u$, appear in the chemotaxis theory and they are used for describing the evolution of particles which interact via gravitational forces [6, 14, 15]. Some reasonings from that theory can be adopted without essential changes but questions related to the long time asymptotics have completely different answers. For example, Banach type theorems giving the local existence can be applied in both cases (see the next section). On the other hand, we know that gravitational problem does not have solutions for large initial masses (see [24] and the references therein), which is not the case of electric forces. Whenever it would not cause any confusion we omit time or spatial variable. Moreover $(f\ast g)\,(x)$ denotes the usual convolution, i.e. $(f\ast g)\,(x)=\int_{\mathbb{R}^{2}}f(x-y)g(y)\,\text{d}y$, and $\mathcal{S}(t)$ is the heat semigroup given by $\big{(}\mathcal{S}(t)u\big{)}(x)=\int_{\mathbb{R}^{2}}G(x-y,t)u(y)\,\text{d}y,$ (2) where $G(x,t)=\tfrac{1}{4\pi t}\,{\,\text{e}}^{-\nicefrac{{\|x\|^{2}}}{{4t}}}$ is the Gauss kernel. Here and subsequently $\\|u\\|_{p}$ denotes the usual $L^{p}(\mathbb{R}^{2})$ norm of the function $u(x,t)$, with respect to the spatial variable. $C$ is a generic constant which can vary from line to line. The Schwartz class of rapidly decreasing functions is denoted by $\mathscr{S}(\mathbb{R}^{2})$ and its dual by $\mathscr{S}^{\prime}(\mathbb{R}^{2})$. $\delta_{0}(x)$ denotes Dirac delta function at the point $0$. In each dimension if $u(x,t)$ and $\phi(x,t)$ solve the system (1a) then $\lambda^{2}u(\lambda x,\lambda^{2}t)$ and $\phi(\lambda x,\lambda^{2}t)$ also do. As we will see the total charge of particles, namely the integral $\int_{\mathbb{R}^{2}}u(x,t)\,\text{d}x$, is preserved in time. However, solutions invariant under the above scaling preserve the charge only in $\mathbb{R}^{2}$. So that we expect essentially nonlinear asymptotics of generic solutions. Moreover, self-similar solutions which describe long time behavior decay faster than the Gauss kernel when $\|x\|\rightarrow\infty$. The two dimensional case differs much from the higher dimensional case, where the long time behavior is described simply by the heat kernel [4]. However, in [4] authors use a completely different approach involving energy and entropy dissipation methods. First we prove (following methods used in [3, 17]) the existence of solutions for the system (1a). Then we study self-similar solutions in radially symmetric and general setting. The last section is devoted to the large time behavior of solutions. ## 2 Existence of solutions for singular initial data We are mainly interested in mild solutions framework. We look for solutions of the integral equation $u(t)=\mathcal{S}(t)u_{0}+\int_{0}^{t}\mathcal{S}(t-\tau)\,\nabla\cdot\big{(}u\nabla\phi\big{)}(\tau)\,\text{d}\tau,$ (3) where $\phi(x,t)$ and $u_{0}(x)$ are as in (1). Two major problems connected with this approach appear. One is: in what sense the solutions of the integral equation (3) satisfy the differential problem (1a). It is obvious that the classical solutions of (1a) satisfy the equality (3). Unfortunately, the reverse implication is not always true (see also a discussion of this problem in [17]). The second problem is: in what sense the initial condition is fulfilled. The answer for this question will come with the definition of the space in which we solve the integral problem (3). Let us notice here that the term $\nabla\cdot\big{(}u\nabla\phi_{u})$ can be written as $\mathcal{B}(u,u)=\nabla\cdot\big{(}u\nabla\phi_{u}\big{)}$ where $\mathcal{B}(u,v)$ is a bilinear form defined by $\mathcal{B}(u,v)=\nabla\cdot\big{(}u\nabla\phi_{v}\big{)},\quad\text{with}\quad\phi_{v}=-E_{2}\ast v,$ for $u,v$ from some Banach space $\mathcal{X}$, which in our case is a space $L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})$ (see a discussion before a proof of Lemma 2.1 below). We introduce here the notation of the rescaled version of a given function $f(x)$ $f_{\lambda}(x)=f(\lambda x),$ where $\lambda$ is some real and positive parameter. In the case of the function depending on space and time, this kind of scaling (and, therefore, the notation) is used only with respect to the spatial variable. By a simple calculation we show that the bilinear form $\mathcal{B}(u,v)$ has scaling order equal to 0 (cf. [17]) which means that for every $\lambda>0$ and every $u,v\in\mathcal{X}$ $\mathcal{B}\big{(}u_{\lambda},v_{\lambda}\big{)}=\big{(}\mathcal{B}(u,v)\big{)}_{\lambda}.$ We use frequently very well known estimates of the heat semigroup $\begin{split}&\big{\\|}\mathcal{S}(t)f\big{\\|}_{p}\leqslant Ct^{\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{q}}}\\|f\\|_{q}\hskip 5.0pt,\\\ &\big{\\|}\nabla\mathcal{S}(t)f\big{\\|}_{p}\leqslant Ct^{\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{q}}-\nicefrac{{1}}{{2}}}\\|f\\|_{q}\hskip 5.0pt,\end{split}$ (4) where $1\leqslant q\leqslant p\leqslant\infty$, $C=C(p,q)$ is some positive constant, $t>0$ and $f(x)\in L^{p}(\mathbb{R}^{2})$. Let us also recall the weak Young inequality $\\|g\ast h\\|_{\gamma}\leqslant C({\beta},{\gamma},g)\\|h\\|_{\alpha}\hskip 3.0pt.$ (5) for all $h\in L^{\alpha}(\mathbb{R}^{2})$, $\|g(x)\|\leqslant\|x\|^{-\nicefrac{{2}}{{\beta}}}$ and $\nicefrac{{1}}{{\alpha}}+\nicefrac{{1}}{{\beta}}=1+\nicefrac{{1}}{{\gamma}}$. For the proof see, e.g., [19, Remark (2) in Theorem 4.3]. The inclusions $\mathscr{S}(\mathbb{R}^{2})\subset L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})\subset\mathscr{S}^{\prime}(\mathbb{R}^{2})$ are continuous and the norm $\\|\cdot\\|_{\nicefrac{{4}}{{3}}}$ is translation invariant. Moreover the estimation $\big{\\|}\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant C\,t^{-\nicefrac{{3}}{{4}}}\\|u\\|_{\nicefrac{{4}}{{3}}}\\|v\\|_{\nicefrac{{4}}{{3}}}$ (6) holds, for every $u,v\in L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})$. Indeed, using the second inequality in (4) with $p=\nicefrac{{4}}{{3}}$ and $q=1$ and the Hölder inequality, we get $\big{\\|}\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant C\,t^{-\nicefrac{{3}}{{4}}}\\|u\\|_{\nicefrac{{4}}{{3}}}\\|\nabla\phi_{v}\\|_{4}.$ We use then inequality (5) with $\gamma=4$, $\beta=2$ and $\alpha=\nicefrac{{4}}{{3}}$ to obtain $\\|u\\|_{\nicefrac{{4}}{{3}}}\\|\nabla\phi_{v}\\|_{4}=\\|u\\|_{\nicefrac{{4}}{{3}}}\big{\\|}\|x\|^{-1}\ast v\big{\\|}_{4}\leqslant\\|u\\|_{\nicefrac{{4}}{{3}}}\\|v\\|_{\nicefrac{{4}}{{3}}},$ which concludes the proof of (6). It means that the space $L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})$ is adequate to the problem (1) in the sense of [17, Definition 3.1]. We construct a homogeneous Besov space modeled on the Banach space $\mathcal{X}=L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})$, as $B\mathcal{X}=\Big{\\{}u\in\mathscr{S}^{\prime}(\mathbb{R}^{2}):\ \sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(t)u\big{\\|}_{\nicefrac{{4}}{{3}}}<\infty\Big{\\}}\hskip 5.0pt,$ and we introduce a norm on this space by $\\|u\\|_{B\mathcal{X}}=\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(t)u\big{\\|}_{\nicefrac{{4}}{{3}}}\hskip 5.0pt.$ The space $B\mathcal{X}$ is in fact the classical homogeneous Besov space $\dot{B}^{\nicefrac{{1}}{{2}}}_{\nicefrac{{4}}{{3}},\infty}$. The norm $\\|\cdot\\|_{B\mathcal{X}}$ is equivalent to the original norm in $\dot{B}^{\nicefrac{{1}}{{2}}}_{\nicefrac{{4}}{{3}},\infty}$ introduced by the dyadic decomposition. For a further discussion on this topic we refer the reader to [17, Example 4.2] and the references therein. As it was in [17, Lemma 5.2] let us prove here the a priori estimate for the bilinear form $\mathcal{B}(u,v)$. ###### Lemma 2.1 There exists a constant $C>0$ independent of $u,\,v$ and $t$ such that $\big{\\|}\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{B\mathcal{X}}\leqslant C\,t^{-\nicefrac{{1}}{{2}}}\\|u\\|_{\nicefrac{{4}}{{3}}}\,\\|v\\|_{\nicefrac{{4}}{{3}}}\hskip 5.0pt,$ (7) for all $u,v\in\mathcal{X}$. Proof We begin with the fact, that there exists some function $C(t)$, independent of $u$ and $v$ such that $\big{\\|}\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{B\mathcal{X}}=\sup_{s>0}s^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(s)\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant C(t)\\|u\\|_{\nicefrac{{4}}{{3}}}\,\\|v\\|_{\nicefrac{{4}}{{3}}}$ for every $t>0$. The proof consists of two steps. First let us assume that $0<s<1$. Then also $0<s^{\nicefrac{{1}}{{4}}}<1$ and because $\\|\mathcal{S}(t)f\\|_{\nicefrac{{4}}{{3}}}\leqslant\\|f\\|_{\nicefrac{{4}}{{3}}}$, the inequality (6) yields $s^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(s)\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant\big{\\|}\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant Ct^{-\nicefrac{{3}}{{4}}}\\|u\\|_{\nicefrac{{4}}{{3}}}\,\\|v\\|_{\nicefrac{{4}}{{3}}}\hskip 5.0pt.$ In the case $s>1$ we proceed in a similar way but first we use the semigroup property $\mathcal{S}(s)\mathcal{S}(t)=\mathcal{S}(s+t)=\mathcal{S}(t)\mathcal{S}(s)$ to obtain $\displaystyle s^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(s)\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}=s^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(t)\mathcal{S}(s)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}$ $\displaystyle\leqslant s^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(s)\mathcal{B}(u,v)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant Cs^{-\nicefrac{{1}}{{2}}}\\|u\\|_{\nicefrac{{4}}{{3}}}\,\\|v\\|_{\nicefrac{{4}}{{3}}}\leqslant C\\|u\\|_{\nicefrac{{4}}{{3}}}\,\\|v\\|_{\nicefrac{{4}}{{3}}}\hskip 5.0pt.$ Now, we show that $C(t)$ is some positive constant. First let us check the scaling property $\mathcal{S}(t)\mathcal{B}(u_{\lambda},v_{\lambda})=\big{(}\mathcal{S}(\lambda^{2}t)\mathcal{B}(u,v)\big{)}_{\lambda}\hskip 5.0pt,$ (8) for all $u,\ v\in L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})$ and for all real $\lambda>0$. For the proof let $\xi\in\mathscr{S}^{\prime}(\mathbb{R}^{2})$. Then $\begin{split}\big{(}\mathcal{S}(t)\xi_{\lambda}\big{)}(x)&=\frac{1}{2\pi t}\int_{\mathbb{R}^{2}}{\,\text{e}}^{-\frac{\|x-z\|^{2}}{4t}}\xi(\lambda z)\,\text{d}z=\frac{1}{2\pi\lambda^{2}t}\int_{\mathbb{R}^{2}}{\,\text{e}}^{-\frac{\lambda^{2}\|x-y/\lambda\|^{2}}{4\lambda^{2}t}}\xi(y)\,\text{d}y\\\ &=\frac{1}{2\pi\lambda^{2}t}\int_{\mathbb{R}^{2}}{\,\text{e}}^{-\frac{\|\lambda x-y\|^{2}}{4\lambda^{2}t}}\xi(y)\,\text{d}y=\big{(}\mathcal{S}(\lambda^{2}t)\xi(x)\big{)}_{\lambda}\hskip 5.0pt.\end{split}$ (9) As we mentioned before the scaling order of the form $\mathcal{B}(u,v)$ equals zero which yields $\displaystyle\mathcal{S}(t)\mathcal{B}(u_{\lambda},v_{\lambda})=\mathcal{S}(t)\big{(}\mathcal{B}(u,v)\big{)}_{\lambda}=\Big{(}\mathcal{S}(\lambda^{2}t)\mathcal{B}(u,v)\Big{)}_{\lambda}.$ Then we use scaling properties $\\|f_{\lambda}\\|_{\nicefrac{{4}}{{3}}}=\lambda^{-\nicefrac{{3}}{{2}}}\\|f\\|_{\nicefrac{{4}}{{3}}}\quad\text{and}\quad\\|f\\|_{B\mathcal{X}}=\lambda^{2}\\|f_{\lambda}\\|_{B\mathcal{X}}\hskip 3.0pt,$ to show that $\displaystyle\big{\\|}\mathcal{S}(\lambda^{2}t)\mathcal{B}(u,v)\big{\\|}_{B\mathcal{X}}=\lambda^{2}\big{\\|}\big{(}\mathcal{S}(\lambda^{2}t)\mathcal{B}(u,v)\big{)}_{\lambda}\big{\\|}_{B\mathcal{X}}=\lambda^{2}\big{\\|}\mathcal{S}(t)\mathcal{B}(u_{\lambda},v_{\lambda})\big{\\|}_{B\mathcal{X}}$ $\displaystyle\leqslant\lambda^{2}C(t)\\|u_{\lambda}\\|_{\nicefrac{{4}}{{3}}}\,\\|v_{\lambda}\\|_{\nicefrac{{4}}{{3}}}=C(t)\lambda^{-1}\\|u\\|_{\nicefrac{{4}}{{3}}}\\|v\\|_{\nicefrac{{4}}{{3}}}\hskip 5.0pt.$ Now we fix $t=t_{0}>0$ and we put $\lambda=(\nicefrac{{t}}{{t_{0}}})^{\nicefrac{{1}}{{2}}}$. Then we obtain the estimate $\big{\\|}\mathcal{S}(t)\mathcal{B}(u,v)\big{\\|}_{B\mathcal{X}}\leqslant C(t_{0})\,t_{0}^{\nicefrac{{1}}{{2}}}t^{-\nicefrac{{1}}{{2}}}\\|u\\|_{\nicefrac{{4}}{{3}}}\\|v\\|_{\nicefrac{{4}}{{3}}},$ which concludes the proof. $\blacksquare$ The space where we are looking for the solutions of the integral equation (3) is defined as $\begin{split}\mathscr{X}=\mathscr{F}\big{(}[0,\infty):&B\mathcal{X}\big{)}\ \cap\\\ &\cap\,\Big{\\{}f(t):\ (0,\infty)\rightarrow L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2}):\ \sup_{t>0}t^{\nicefrac{{3}}{{4}}}\\|f(t)\\|_{\mathcal{X}}<\infty\Big{\\}},\end{split}$ (10) and we equip this space with the norm $\\|f\\|_{\mathscr{X}}=\max\Big{\\{}\sup_{t>0}\\|f(t)\\|_{B\mathcal{X}},\ \sup_{t>0}t^{\nicefrac{{3}}{{4}}}\\|f(t)\\|_{\nicefrac{{4}}{{3}}}\Big{\\}}.$ The space $\mathscr{X}$ with the above norm is a Banach space. We understand $\mathscr{F}\big{(}[0,\infty):B\mathcal{X}\big{)}$ as the space of $B\mathcal{X}$-valued measurable functions which belong to $L^{\infty}\big{(}[0,\infty):B\mathcal{X}\big{)}$ and attain the initial condition of the problem (1) as $t\rightarrow 0^{+}$ in the sense of tempered distributions, namely $\int_{\mathbb{R}^{2}}\phi(x)u(x,t)\,\text{d}x\rightarrow\int_{\mathbb{R}^{2}}\phi(x)u_{0}(x)\,\text{d}x,$ for all $\phi\in\mathscr{S}(\mathbb{R}^{2})$ as $t$ goes to $0^{+}$. Now let us formulate and prove the main result of this section. In the proof we combine the reasonings presented in [3] and [17, 18]. ###### Theorem 2.2 There exists an $\varepsilon>0$ such that for all $u_{0}\in B\mathcal{X}$ satisfying $\\|u_{0}\\|_{B\mathcal{X}}<\varepsilon$, there is a global in time solution $u(x,t)$ of the problem (1) in the space (10). This solution is unique among all the functions $u\in B\mathcal{X}$ satisfying the condition $\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\\|u(t)\\|_{\mathcal{X}}<2\varepsilon$. Proof We show that the operator $\mathscr{N}\big{(}u\big{)}(t)=\mathcal{S}(t)u_{0}+\int_{0}^{t}\mathcal{S}(t-\tau)\,\mathcal{B}\big{(}u,u\big{)}(\tau)\,\text{d}\tau$ is a contraction in the set $B_{\varepsilon}=\Big{\\{}u\in\mathscr{X}:\ \\|u(t)\\|_{\mathscr{X}}\leqslant 2\varepsilon\Big{\\}}\hskip 5.0pt$ for some small $\varepsilon>0$, to be specified later. In the following estimates we use frequently the elementary inequalities such as the Hölder inequality, the heat semigroup estimates (4) and the estimates of the bilinear form $\mathcal{B}(u,v)$ from (6) and (7) $\displaystyle t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathscr{N}\big{(}u\big{)}(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant$ $\displaystyle\,\big{\\|}u_{0}\big{\\|}_{B\mathcal{X}}+C\,t^{\nicefrac{{1}}{{4}}}\int_{0}^{t}(t-\tau)^{-\nicefrac{{3}}{{4}}}\tau^{-\nicefrac{{1}}{{2}}}\bigg{(}\tau^{\nicefrac{{1}}{{4}}}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\bigg{)}^{2}\,\text{d}\tau$ $\displaystyle\leqslant$ $\displaystyle\,\big{\\|}u_{0}\big{\\|}_{B\mathcal{X}}+C\,\bigg{(}\sup_{\tau>0}\tau^{\nicefrac{{1}}{{4}}}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\bigg{)}^{2}\int_{0}^{1}(1-w)^{-\nicefrac{{3}}{{4}}}w^{-\nicefrac{{1}}{{2}}}\,\text{d}w$ $\displaystyle\leqslant$ $\displaystyle\,\big{\\|}u_{0}\big{\\|}_{B\mathcal{X}}+C\varepsilon^{2},$ and similarly $t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathscr{N}\big{(}u\big{)}(t)-\mathscr{N}\big{(}v\big{)}(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\\\ \leqslant\ C\,t^{\nicefrac{{1}}{{4}}}\int_{0}^{t}(t-\tau)^{-\nicefrac{{3}}{{4}}}\big{\\|}(u-v)(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\Big{(}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}+\big{\\|}v(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\Big{)}\,\text{d}\tau\\\ \leqslant\ C\,\varepsilon\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}(u-v)(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\hskip 5.0pt.$ In the same way we deal with the estimate $\displaystyle\big{\\|}\mathscr{N}\big{(}u\big{)}(t)\big{\\|}_{B\mathcal{X}}\leqslant$ $\displaystyle\,\big{\\|}u_{0}\big{\\|}_{B\mathcal{X}}+C\int_{0}^{t}(t-\tau)^{-\nicefrac{{1}}{{2}}}\tau^{-\nicefrac{{1}}{{2}}}\bigg{(}\tau^{\nicefrac{{1}}{{4}}}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\bigg{)}^{2}\,\text{d}\tau$ $\displaystyle\leqslant$ $\displaystyle\,\big{\\|}u_{0}\big{\\|}_{B\mathcal{X}}+C\bigg{(}\sup_{\tau>0}\tau^{\nicefrac{{1}}{{4}}}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\bigg{)}^{2}\int_{0}^{1}(1-w)^{-\nicefrac{{1}}{{2}}}w^{-\nicefrac{{1}}{{2}}}\,\text{d}w$ $\displaystyle\leqslant$ $\displaystyle\,\big{\\|}u_{0}\big{\\|}_{B\mathcal{X}}+C\varepsilon^{2}$ and $\big{\\|}\mathscr{N}\big{(}u\big{)}(t)-\mathscr{N}\big{(}v\big{)}(t)\big{\\|}_{B\mathcal{X}}\leqslant\,C\,\varepsilon\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}(u-v)(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\hskip 5.0pt.$ Combining the above inequalities and choosing $\varepsilon>0$ such that $C\varepsilon^{2}<\varepsilon$ and $C\varepsilon<1$, we obtain $\displaystyle\big{\\|}\mathscr{N}(u)\big{\\|}_{\mathscr{X}}\leqslant 2\varepsilon\hskip 5.0pt,$ $\displaystyle\big{\\|}\mathscr{N}(u)-\mathscr{N}(v)\big{\\|}_{\mathscr{X}}\leqslant\big{\\|}(u-v)(t)\big{\\|}_{\mathscr{X}}\hskip 5.0pt,$ for every $u,\ v\in L^{\nicefrac{{4}}{{3}}}(\mathbb{R}^{2})$. It follows that $\mathscr{N}:\ B_{\varepsilon}\rightarrow B_{\varepsilon}$ is a contraction in $B_{\varepsilon}$. Thanks to the Banach fixed point theorem we obtain the existence of a mild solution of the problem (1). This solution is unique in the set $B_{\varepsilon}$ and is global in time. $\blacksquare$ Since we are particularly interested in self-similar solutions, it is desired to know what does the condition $\\|u_{0}\\|_{B\mathcal{X}}<\varepsilon$ mean for the initial condition being a distribution. Namely, let $u_{0}(x)=M\delta_{x_{0}}(x)$. Then $\begin{split}\\|M\mathcal{S}(t)\delta_{x_{0}}\\|_{\nicefrac{{4}}{{3}}}^{\nicefrac{{4}}{{3}}}=&\ \int_{\mathbb{R}^{2}}\Bigg{(}\frac{M}{4\pi t}\int_{\mathbb{R}^{2}}{\,\text{e}}^{-\frac{\|x-y\|^{2}}{4t}}\,\delta_{x_{0}}(y)\,\text{d}y\Bigg{)}^{\nicefrac{{4}}{{3}}}\,\text{d}x\\\ =&\ \int_{\mathbb{R}^{2}}\bigg{(}\frac{M}{4\pi t}\;{\,\text{e}}^{-\frac{\|x-x_{0}\|^{2}}{4t}}\bigg{)}^{\nicefrac{{4}}{{3}}}\\!\\!\\!\,\text{d}x=\frac{M^{\nicefrac{{4}}{{3}}}}{(4\pi t)^{\nicefrac{{1}}{{3}}}}\int_{\mathbb{R}^{2}}\frac{1}{4\pi t}\;{\,\text{e}}^{-\frac{4\|x-x_{0}\|^{2}}{3\cdot 4t}}\,\text{d}x\\\ =&\ \frac{4}{3}\cdot\frac{M^{\nicefrac{{4}}{{3}}}}{(4\pi t)^{\nicefrac{{1}}{{3}}}}\cdot\frac{1}{t^{\nicefrac{{1}}{{4}}}}\int_{\mathbb{R}^{2}}\frac{1}{4\pi t}\;{\,\text{e}}^{-\nicefrac{{\|z\|^{2}}}{{4t}}}\,\text{d}z.\end{split}$ Then $\limsup_{t\rightarrow 0}t^{\nicefrac{{1}}{{4}}}\\|M\mathcal{S}(t)\delta_{x_{0}}\\|_{\nicefrac{{4}}{{3}}}\sim M^{\nicefrac{{4}}{{3}}},$ which means that atomic measures are admissible as initial data (1b) but only with small masses of single atoms. Moreover let us notice, that if $u_{0}(x)$ is a function from $L^{1}(\mathbb{R}^{2})$ – it does not contain singular part – then $\limsup_{t\rightarrow 0}t^{\nicefrac{{1}}{{4}}}\\|u_{0}\\|_{\nicefrac{{4}}{{3}}}=0.$ Above shows that the part $\limsup_{t\rightarrow 0}t^{\nicefrac{{1}}{{4}}}\\|\cdot\\|_{\nicefrac{{4}}{{3}}}$ of the norm $\\|\cdot\\|_{\mathscr{X}}$ measures only the singular part. ## 3 Local and global existence in $L^{p}(\mathbb{R}^{2})$ space Using the same framework of mild solutions we prove the local in time existence of solutions of the problem (3) in the space $C([0,T];L^{1}(\mathbb{R}^{2}))\cap C([0,T];L^{p}(\mathbb{R}^{2}))$ for some $p>1$ and an arbitrary initial condition $u_{0}\in L^{1}(\mathbb{R}^{2})\cap L^{p}(\mathbb{R}^{2})$. Moreover, we show that these solutions can be extended to global in time solutions. This situation differs from the case where particles interact through gravitational forces and through the chemotactic attraction. The solutions in that case blow up in $L^{p}(\mathbb{R}^{2})$ spaces for sufficiently large initial data (see [14, 15]). In the following we use the same inequalities as before, i.e. the heat semigroup estimates (4) and the weak Young inequality (5). Let us define the space $\mathcal{X}_{T}=C\big{(}[0,T];L^{1}(\mathbb{R}^{2})\big{)}\cap C\big{(}[0,T];L^{p}(\mathbb{R}^{2})\big{)},$ which is Banach space with the norm $\\|u\\|_{\mathcal{X}_{T}}=\sup_{0\leqslant t\leqslant T}\\|u\\|_{1}+\sup_{0\leqslant t\leqslant T}\\|u\\|_{p}\hskip 5.0pt.$ ###### Proposition 3.1 Let $\frac{4}{3}\leqslant p<4$. For any $u_{0}\in L^{1}(\mathbb{R}^{2})\cap L^{p}(\mathbb{R}^{2})$ there exists some $T>0$ and a unique solution of the integral equation $u(t)=\mathcal{S}(t)u_{0}+\int_{0}^{t}\mathcal{S}(t-\tau)\nabla\cdot\big{(}u\nabla\phi\big{)}(\tau)\,\text{d}\tau,$ such that $u(x,t)\in\mathcal{X}_{T}$. Proof Let $\mathscr{N}\big{(}u\big{)}(t)=\mathcal{S}(t)u_{0}+\int_{0}^{t}\nabla\mathcal{S}(t-\tau)\cdot\big{(}u\nabla\phi\big{)}(\tau)\,\text{d}\tau.$ (11) Then (4) implies that $\mathcal{I}_{p}(t)=\\|\mathscr{N}\big{(}u\big{)}(t)-\mathcal{S}(t)u_{0}\\|_{p}\leqslant C\int_{0}^{t}(t-\tau)^{\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{q}}-\nicefrac{{1}}{{2}}}\big{\\|}\big{(}u\nabla\phi\big{)}(\tau)\big{\\|}_{q}\,\text{d}\tau,$ for every $1\leqslant p,\,q\leqslant\infty$. We use the Hölder inequality to obtain $\mathcal{I}_{p}(t)\leqslant C\int_{0}^{t}(t-\tau)^{\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{q}}-\nicefrac{{1}}{{2}}}\big{\\|}u(\tau)\big{\\|}_{p}\big{\\|}\big{(}\nabla\phi\big{)}(\tau)\big{\\|}_{\frac{pq}{p-q}}\,\text{d}\tau.$ We can exploit (5) with $\alpha=\frac{2pq}{pq-2p-2q},\beta=2$ and $\gamma=\frac{pq}{p-q}$, which yields $\mathcal{I}_{p}(t)\leqslant C\int_{0}^{t}(t-\tau)^{\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{q}}-\nicefrac{{1}}{{2}}}\big{\\|}u(\tau)\big{\\|}_{p}\big{\\|}u(\tau)\big{\\|}_{\frac{2pq}{pq-2p-2q}}\,\text{d}\tau.$ We would like to have $\frac{2pq}{pq-2p-2q}=p$ which is possible for $q=\frac{2p}{4-p}$ and all $\nicefrac{{4}}{{3}}\leqslant p<4$. Consequently, we have $\sup_{0\leqslant t\leqslant T}\\|\mathscr{N}\big{(}u\big{)}(t)-\mathcal{S}(t)u_{0}\\|_{p}\\\ \leqslant C\bigg{(}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)\big{\\|}_{p}\bigg{)}^{2}\int_{0}^{t}(t-\tau)^{-\nicefrac{{1}}{{p}}}\,\text{d}\tau\\\ \leqslant CT^{1-\nicefrac{{1}}{{p}}}\bigg{(}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)\big{\\|}_{p}\bigg{)}^{2}\int_{0}^{1}(1-s)^{-\nicefrac{{1}}{{p}}}\,\text{d}s,$ where the last integral is finite since $-\nicefrac{{3}}{{4}}\leqslant-\nicefrac{{1}}{{p}}\leqslant-\nicefrac{{1}}{{4}}$. Similarly, we obtain $\sup_{0\leqslant t\leqslant T}\\|\mathscr{N}\big{(}u\big{)}(t)-\mathcal{S}(t)u_{0}\\|_{1}\\\ \leqslant C\bigg{(}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)\big{\\|}_{p}\bigg{)}^{2}\int_{0}^{t}(t-\tau)^{1-\nicefrac{{2}}{{p}}}\,\text{d}\tau\\\ \leqslant CT^{2-\nicefrac{{2}}{{p}}}\bigg{(}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)\big{\\|}_{p}\bigg{)}^{2}\int_{0}^{1}(1-s)^{1-\nicefrac{{2}}{{p}}}\,\text{d}s,$ with the same constraint for $p$, namely $\nicefrac{{4}}{{3}}\leqslant p<4$, which yields the inequality $-\nicefrac{{1}}{{2}}\leqslant 1-\nicefrac{{2}}{{p}}\leqslant\nicefrac{{1}}{{2}}$. An analogous calculation gives $\sup_{0\leqslant t\leqslant T}\\|\mathscr{N}\big{(}u\big{)}(t)-\mathscr{N}\big{(}v\big{)}(t)\\|_{p}\\\ \leqslant CT^{1-\nicefrac{{1}}{{p}}}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)-v(t)\big{\\|}_{p}\bigg{(}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)\big{\\|}_{p}+\sup_{0\leqslant t\leqslant T}\big{\\|}v(t)\big{\\|}_{p}\bigg{)}$ and $\sup_{0\leqslant t\leqslant T}\\|\mathscr{N}\big{(}u\big{)}(t)-\mathscr{N}\big{(}v\big{)}(t)\\|_{1}\\\ \leqslant CT^{2-\nicefrac{{2}}{{p}}}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)-v(t)\big{\\|}_{p}\bigg{(}\sup_{0\leqslant t\leqslant T}\big{\\|}u(t)\big{\\|}_{p}+\sup_{0\leqslant t\leqslant T}\big{\\|}v(t)\big{\\|}_{p}\bigg{)},$ for all $u,v\in\mathcal{X}_{T}$ and $\nicefrac{{4}}{{3}}\leqslant p<4$. Without loss of generality we can assume that $0<T<1$. Then $\big{\\|}\mathscr{N}\big{(}u\big{)}(t)\big{\\|}_{\mathcal{X}_{T}}\leqslant\\|u_{0}\\|_{\mathcal{X}_{T}}+CT^{1-\nicefrac{{1}}{{p}}}\big{\\|}u(t)\big{\\|}_{\mathcal{X}_{T}}$ and $\rule{20.0pt}{0.0pt}\big{\\|}\mathscr{N}\big{(}u\big{)}(t)-\mathscr{N}\big{(}u\big{)}(t)\big{\\|}_{\mathcal{X}_{T}}\\\ \leqslant CT^{1-\nicefrac{{1}}{{p}}}\big{\\|}u(t)-v(t)\big{\\|}_{\mathcal{X}_{T}}\Big{(}\big{\\|}u(t)\big{\\|}_{\mathcal{X}_{T}}+\big{\\|}v(t)\big{\\|}_{\mathcal{X}_{T}}\Big{)}.\rule{20.0pt}{0.0pt}$ For arbitrary $R>0$ we can choose $T>0$ so small that the operator $\mathscr{N}$ is a contraction in the ball $B_{R}(0)\subset\mathcal{X}_{T}$. The Banach fixed point theorem gives the existence of a local in time solution in the space $\mathcal{X}_{T}$. From the construction this solution is unique in the ball $B_{R}(0)$. But the uniqueness is a local property and the above solution is also unique in the space $\mathcal{X}_{\hat{T}}$ for $\hat{T}>0$ possibly smaller than $T$. Indeed, let us suppose that there is another solution in the space $\mathcal{X}_{T}$ contained in some ball $B_{\hat{R}}(0)$ with $R<\hat{R}$. Taking $\hat{T}<T$ we construct the unique solution on the time interval $[0,\hat{T}]$ but in the ball $B_{\hat{R}}(0)$, which contradicts the uniqueness of the previous solution. $\blacksquare$ ###### Proposition 3.2 Let $\nicefrac{{4}}{{3}}\leqslant p<4$ and $u_{0}(x)\in L^{1}(\mathbb{R}^{2})\cap L^{p}(\mathbb{R}^{2})$. If additionally $u_{0}(x)$ is nonnegative, then for all $0<t<T$, $u(x,t)\geqslant 0$ and $\int_{\mathbb{R}^{2}}u(x,t)\,\text{d}x=\int_{\mathbb{R}^{2}}u_{0}(x)\,\text{d}x.$ Proof The conservation of positivity for the solutions of the problem (1) is a consequence of the analogous property for the system on bounded domains. Indeed, there has been proved in [2, Proposition 1 and Proposition 2] that solutions with initial conditions $0\leqslant u_{0}(x)\in L^{1}(\Omega)\cap L^{p}(\Omega)$ are nonnegative for $t>0$. The proof involved Stampacchia truncation method. The same approach permits to prove nonnegativity of the solution $u(x,t)$ either by approximating $u(x,t)$ on the sets $\big{\\{}\|x\|<N\big{\\}}$ where $N\in\mathbb{N}$, by solutions $u(x,t)=u_{k}(x,t)$ of the problem posed on the sets $\big{\\{}\|x\|<N+k\big{\\}}$ where $k\in\mathbb{N}$ or by a direct rewriting of the Stampacchia scheme (see also [12, Lemma 1] and [11, Theorem I, Lemma 4.1]). The $L^{1}(\mathbb{R}^{2})$ norm of the solution is properly controlled, as it was justified in the proof of Proposition 3.1, namely $\begin{split}\sup_{0\leqslant t<T}\\|\mathcal{N}(u)(t)\\|_{1}\leqslant\\|u_{0}\\|_{1}+CT^{1-\nicefrac{{1}}{{p}}}\big{\\|}u(t)\big{\\|}_{\mathcal{X}_{T}}.\end{split}$ Integrating the Duhamel formula (11) over $\mathbb{R}^{2}$ we get $\int_{\mathbb{R}^{2}}u(x,t)\,\text{d}x=\int_{\mathbb{R}^{2}}\mathcal{S}(t)u_{0}(x)\,\text{d}x+\int_{\mathbb{R}^{2}}\int_{0}^{t}\nabla\mathcal{S}(t-\tau)\cdot\big{(}u\nabla\phi\big{)}(\tau)\,\text{d}\tau\,\text{d}x.$ But the last integral on the right hand side is equal to zero, since we can apply the Fubini theorem, and the first one is equal to $\int_{\mathbb{R}^{2}}u_{0}(x)\,\text{d}x$. $\blacksquare$ We obtain the global in time existence of the solutions using the following a priori estimate. ###### Lemma 3.3 Let $u(x,t)$ be a solution of the problem (1) with nonnegative initial condition $u_{0}(x)\in L^{1}(\mathbb{R}^{2})$. Let $M=\int_{\mathbb{R}^{2}}u_{0}(x)\,\text{d}x$. For every $p\in(1,\infty)$ such that $2^{k-1}<p<2^{k}$ for some $k\in\mathbb{N}$, the inequality $\\|u(\cdot,t)\\|_{p}\leqslant C(k)\,M\,t^{-(1-\nicefrac{{1}}{{p}})}$ holds with a constant $C(k)$ given explicitly. Moreover $\\|u\\|_{\infty}\leqslant C\,M\,t^{-1}$ holds for some constant $C>0$. We base the proof of the above lemma on the following interpolation inequality ###### Lemma 3.4 Let $1\leqslant q,r\leqslant\infty,\ j,m\in\mathbb{N},\ 0\leqslant j<m$ and $a\in[\nicefrac{{j}}{{m}},1]$. Moreover $\frac{1}{p}=\frac{j}{n}+a\Big{(}\frac{1}{r}-\frac{m}{n}\Big{)}+(1-a)\frac{1}{q},$ where $n$ is the space dimension. Then $\sum_{\|\alpha\|=j}\\|D^{\alpha}u\\|_{p}\leqslant C\bigg{(}\sum_{\|\alpha\|=m}\\|D^{\alpha}u\\|_{r}\bigg{)}^{a}\,\\|u\\|_{q}^{1-a}.$ (12) This inequality was introduced by Gagliardo-Nirenberg-Sobolev and sometimes is referred to as the Nash inequality [21]. Proof of Lemma 3.3 Step 1. Let $p=2$. Multiplying the first equation in (1a) by $u$ and integrating by parts, we get $\begin{split}0=&\int_{\mathbb{R}^{2}}\big{(}uu_{t}-u\Delta u-u\nabla\cdot(u\nabla\phi)\big{)}\,\text{d}x\\\ &=\frac{1}{2}\frac{\text{d}}{\text{d}t}\\|u\\|^{2}_{2}+\\|\nabla u\\|^{2}_{2}+\frac{1}{2}\int_{\mathbb{R}^{2}}\nabla u^{2}\cdot\nabla\phi\,\text{d}x\\\ &=\frac{1}{2}\frac{\text{d}}{\text{d}t}\\|u\\|^{2}_{2}+\\|\nabla u\\|^{2}_{2}+\frac{1}{2}\int_{\mathbb{R}^{2}}u^{3}\,\text{d}x.\end{split}$ (13) From Proposition 3.2 the solution $u(x,t)\geqslant 0$ in $\mathbb{R}^{2}$ for all $t>0$. Then we obtain the inequality $\frac{\text{d}}{\text{d}t}\\|u\\|^{2}_{2}+2\\|\nabla u\\|^{2}_{2}\leqslant 0.$ (14) Then taking $a=\nicefrac{{1}}{{2}},\ q=m=1,\ n=r=p=2,\ j=0$ in (12) we get $\\|u\\|_{2}\leqslant C\\|\nabla u\\|^{\nicefrac{{1}}{{2}}}_{2}\\|u\\|_{1}^{\nicefrac{{1}}{{2}}}$ (15) which is equivalent to $\frac{1}{CM_{0}^{2}}\\|u\\|_{2}^{4}\leqslant\\|\nabla u\\|^{2}_{2}$ (16) where $M_{0}=\\|u(0)\\|_{1}$. Using (14) and (16) we get $\frac{\text{d}}{\text{d}t}\\|u\\|_{2}^{2}+\frac{2}{CM_{0}^{2}}\\|u\\|_{2}^{4}\leqslant 0.$ Integrating the above inequality yields $\\|u(t)\\|_{2}^{2}\leqslant\frac{1}{\frac{1}{\\|u(0)\\|_{2}^{2}}+\frac{2}{CM_{0}^{2}}t}\leqslant\frac{CM_{0}^{2}}{2}\>t^{-1},$ which ends this part of the proof. Step 2. $p=2^{k}$. We prove it by recurrence. Let as assume that $\\|u\\|_{2^{k-1}}^{2^{k-1}}\leqslant a_{k-1}M_{0}^{2^{k-1}}t^{-(2^{k-1}-1)},$ (17) where $a_{k}=\begin{cases}\nicefrac{{C}}{{2}}&\text{\quad for\quad}k=1,\\\ C2^{k-2}\big{(}a_{k-1}\big{)}^{2}&\text{\quad for\quad}k\geqslant 2.\end{cases}$ (18) In the inequality (15) we set $u^{2^{k-1}}$ instead of $u$ which yields $\big{(}\\|u\\|_{2^{k}}^{2^{k}}\big{)}^{2}\leqslant C\\|\nabla u^{2^{k-1}}\\|_{2}^{2}\big{(}\\|u\\|^{2^{k-1}}_{2^{k-1}}\big{)}^{2}.$ (19) The same calculation as in (13) with the multiplication by $u^{2^{k}-1}$ gives $\frac{\text{d}}{\text{d}t}\\|u\\|_{2^{k}}^{2^{k}}+\frac{4(2^{k}-1)}{2^{k}}\\|\nabla u^{2^{k-1}}\\|_{2}^{2}\leqslant 0.$ (20) Combining together inequalities (17), (19) and (20) we get $\displaystyle\\|\nabla u^{2^{k-1}}\\|_{2}^{2}\geqslant\frac{\big{(}\\|u\\|_{2^{k}}^{2^{k}}\big{)}^{2}}{C\big{(}\\|u\\|_{2^{k-1}}^{2^{k-1}}\big{)}^{2}},$ $\displaystyle\frac{\text{d}}{\text{d}t}\\|u\\|_{2^{k}}^{2^{k}}+\frac{2^{k}-1}{C2^{k-2}}\frac{\big{(}\\|u\\|_{2^{k}}^{2^{k}}\big{)}^{2}}{\big{(}\\|u\\|_{2^{k-1}}^{2^{k-1}}\big{)}^{2}}\leqslant 0,$ $\displaystyle\frac{\frac{\text{d}}{\text{d}t}\\|u\\|_{2^{k}}^{2^{k}}}{\big{(}\\|u\\|_{2^{k}}^{2^{k}}\big{)}^{2}}\leqslant-\frac{2^{k}-1}{CM_{0}^{2^{k}}2^{k-2}(a_{k-1})^{2}}\>t^{2^{k}-2}\hskip 3.0pt.$ Let us denote $f(t)=\\|u(t)\\|_{2^{k}}^{2^{k}}$ and integrate the above inequality over the interval $(0,t)$. We obtain $\displaystyle-\frac{1}{f(t)}+\frac{1}{f(0)}\leqslant-\frac{1}{CM_{0}^{2^{k}}2^{k-2}\big{(}a_{k-1}\big{)}^{2}}\>t^{2^{k}-1},$ $\displaystyle\frac{1}{f(t)}\geqslant\frac{1}{CM_{0}^{2^{k}}2^{k-2}\big{(}a_{k-1}\big{)}^{2}}t^{2^{k}-1}+\frac{1}{f(0)}\geqslant\frac{1}{CM_{0}^{2^{k}}2^{k-2}\big{(}a_{k-1}\big{)}^{2}}\>t^{2^{k}-1},$ $\displaystyle f(t)\leqslant C2^{k-2}\big{(}a_{k-1}M_{0}^{2^{k}}\big{)}^{2}t^{-(2^{k}-1)}$ which gives $\\|u\\|_{2^{k}}^{2^{k}}\leqslant a_{k}M_{0}^{2^{k}}t^{-(2^{k}-1)}.$ (21) Step 3. General case. For each $p\in(1,\infty)$ there exists some $k\in\mathbb{N}$ such that $2^{k-1}<p<2^{k}$. Then it is sufficient to use inequality (21) and very well known interpolation inequality $\\|u\\|_{p}\leqslant\\|u\\|_{2^{k}}^{1-a}\\|u\\|_{1}^{a}$ with $a=\frac{1-\nicefrac{{1}}{{p}}}{1-\nicefrac{{1}}{{2^{k}}}}$. Then $\\|u\\|_{p}\leqslant M_{0}a_{k}^{(\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{2^{k}}})/(2^{k}-1)}t^{-(1-\nicefrac{{1}}{{p}})}.$ At the end let us notice that $a_{k}^{(\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{2^{k}}})/(2^{k}-1)}\leqslant a_{k}^{\nicefrac{{1}}{{p}}-\nicefrac{{1}}{{2^{k}}}}\leqslant a_{k}^{\nicefrac{{1}}{{2^{k}}}}$ Step 4. $p=\infty$. In the last step we use the property $\\|u\\|_{\infty}=\limsup_{p\,\rightarrow\,\infty}\\|u\\|_{p}.$ Then it is enough to show that $\lim_{k\rightarrow\infty}a_{k}^{\nicefrac{{1}}{{2^{k}}}}<\infty$. But $a_{k}=C^{v_{k}}2^{w_{k}}$, where $w_{k}=\begin{cases}-1&\text{\quad for\quad}k=1,\\\ 2w_{k-1}+(k-2)&\text{\quad for\quad}k\geqslant 2.\end{cases}$ and $v_{k}=2^{k}-1$. It is not difficult to see that for all $k\in\mathbb{N}$, $w_{k}<0$. Then $\lim_{k\rightarrow\infty}a_{k}^{\nicefrac{{1}}{{2^{k}}}}=\lim_{k\rightarrow\infty}\frac{C^{1-\nicefrac{{1}}{{2^{k}}}}}{2^{\nicefrac{{\|w_{k}\|}}{{2^{k}}}}}\leqslant C<\infty$ which ends the proof . $\blacksquare$ As a simple corollary from Proposition 3.1 and Lemma 3.3 we obtain the main result of this section, namely ###### Theorem 3.5 Let $\nicefrac{{4}}{{3}}\leqslant p<4$ and $u_{0}(x)\in L^{1}(\mathbb{R}^{2})\cap L^{p}(\mathbb{R}^{2})$. Then, there exists the unique solution of the integral problem (3) in the space $C\big{(}[0,\infty):\,L^{1}(\mathbb{R}^{2})\big{)}\cap C\big{(}[0,\infty):\,L^{p}(\mathbb{R}^{2})\big{)}$. ## 4 Radially symmetric self-similar solutions As we mentioned in the introduction the system (1a) is invariant under the scaling $\lambda^{2}u(\lambda x,\lambda^{2}t),\quad\phi(\lambda x,\lambda^{2}t),$ (22) where $0<\lambda\in\mathbb{R}$. In this section we assume that $u(x,t),\ \phi(x,t)$ are self-similar, i.e. they are invariant under the scaling (22). The most important from our point of view are the solutions beginning their evolution from the initial conditions of the form $M\delta_{0}(x)$. Such an initial condition is obviously radial. Thus we assume from now on the radial symmetry of the solutions $u(x,t),\ \phi(x,t)$. Following the integrated density method (see [3, 7, 22]) we introduce the new variable $Q(r,t)=\int_{B_{r}(0)}u(x,t)\,\text{d}x=2\pi\int_{0}^{r}s\,u(s,t)\,\text{d}s,$ where $B_{r}(0)$ is the closed ball of radius $r$ centered at the origin $(0,0)$. Integrating the system (1a) over such balls and performing simple computations, we rewrite the problem (1a)–(1b) in the form $\displaystyle Q_{t}=Q_{rr}-\frac{1}{r}Q_{r}-\frac{1}{2\pi r}QQ_{r},$ $\displaystyle Q(0,t)=0,\quad\text{for}\quad t\geqslant 0$ $\displaystyle\lim_{r\rightarrow\infty}Q(r,t)=\int_{\mathbb{R}^{2}}u_{0}(x)\,\text{d}x=M,\quad\text{for}\quad t\geqslant 0.$ Using self-similarity we can again change the variables: $Q(r,t)=2\pi\xi(y)$ for $y=\nicefrac{{r^{2}}}{{t}}$. This way we transform our problem to the following one $\displaystyle\xi^{\prime\prime}(y)+\frac{1}{4}\xi^{\prime}(y)-\frac{1}{2y}\xi(y)\xi^{\prime}(y)=0,$ (23a) $\displaystyle\xi(0)=0,\quad\xi^{\prime}(0)=a.$ (23b) Formally, the boundary conditions implied by (1b) are $\xi(0)=0\quad\text{and}\quad\lim_{y\rightarrow\infty}\xi(y)=M.$ But it is more convenient to investigate equation (23a) with the boundary conditions (23b) for some positive real $a$. As we will see, both problems are equivalent. The reader can find similar results considering the problem (23) (existence of solutions, regularity, etc.) in [7] but methods used there are different. The main result we prove in this paper is the following ###### Theorem 4.1 For all $M>0$, there exists a unique solution of (23a), satisfying the initial conditions $\xi(0)=0$ and $\lim_{y\rightarrow\infty}\xi(y)=M$. First we justify some a priori properties of the solutions of (23). Then we are able to construct a suitable space where the Banach fixed point theorem can be applied. ###### Lemma 4.2 Let $\xi(y)$ be a solution of (23) with $y\in[0,Y_{0})$. Then * i) $\xi^{\prime}(0)>\nicefrac{{1}}{{2}}$ implies $\xi^{\prime}(y)>\nicefrac{{1}}{{2}}$, for all $y\in[0,Y_{0})$; * ii) if $0<\xi^{\prime}(0)=a<\nicefrac{{1}}{{2}}$ then $0<\xi^{\prime}(y)<a$ and $-\nicefrac{{1}}{{8}}<\xi^{\prime\prime}(y)<0$ for all $y\in[0,Y_{0}]$. Moreover, in such a case $0<\xi(y)<a\,y$ on the whole existence interval; * iii) if $\xi(y)$ is global and $0<\xi^{\prime}(0)<\nicefrac{{1}}{{2}}$, then $\lim_{y\rightarrow\infty}\xi(y)$ exists; Proof (i) In order to compute $\xi^{\prime\prime}(0)$, we let $y$ tend to $0^{+}$ in the equation (23a). Then $\xi^{\prime\prime}(0)=\xi^{\prime}\big{(}0\big{)}\bigg{(}\frac{1}{2}\,\xi^{\prime}(0)-\frac{1}{4}\bigg{)}=\frac{a}{2}\bigg{(}a-\frac{1}{2}\bigg{)}$ since $\xi^{\prime}(0)=\lim_{y\rightarrow 0^{+}}\frac{\xi(y)-\xi(0)}{y}=\lim_{y\rightarrow 0^{+}}\frac{\xi(y)}{y}.$ In the case $a>\nicefrac{{1}}{{2}}$ we have $\xi^{\prime\prime}(y)>0$ for some right neighbourhood of $y=0$. Now let us define $y_{0}=\sup\Big{\\{}y\in[0,Y_{0}):\ \xi^{\prime}(s)>\tfrac{1}{2}\text{\quad for all\quad}s\in[0,y)\Big{\\}},$ (24) and suppose that we have $\xi^{\prime}(y_{0})=\nicefrac{{1}}{{2}}$ at a point $y_{0}<Y_{0}$. It follows immediately that $\xi^{\prime}(y)$ is decreasing on some subinterval of $[0,y_{0})$. Since for $y\in[0,y_{0})$ we have $\xi(y)>\tfrac{1}{2}y$, then by (23a), the second derivative $\xi^{\prime\prime}(y)$ is positive on the interval $[0,y_{0})$. It follows that $\xi^{\prime}(y)$ cannot decrease on this interval, which contradicts our assumption. Thus, we have $\xi^{\prime}(y)>\nicefrac{{1}}{{2}}$ for all $y\in[0,Y_{0})$. (ii) It is easy to check, that if $0<a<\nicefrac{{1}}{{2}}$ then $\xi^{\prime}(y)<\nicefrac{{1}}{{2}}$ on $[0,Y_{0})$. It is enough to note that $\xi^{\prime\prime}(y)<0$, which implies that the first derivative $\xi^{\prime}(y)$ is strictly decreasing, so $\xi^{\prime}(y)<\xi^{\prime}(0)=a$. It remains to prove that $\xi^{\prime}(y)$ is positive. Integrating $\xi^{\prime}(y)<\nicefrac{{1}}{{2}}$, we get that $\xi(y)<\nicefrac{{y}}{{2}}$ for $y\in[0,Y_{0})$. Assume on the contrary, that there exists a point $\tilde{y}\in[0,Y_{0})$, such that $\xi^{\prime}(\tilde{y})=0$. Then $\xi^{\prime\prime}(\tilde{y})=0$. We have to consider two possibilities: 1. $\xi^{\prime}(y)<0$; 2. $\xi^{\prime}(y)>0$; both, for $y$ in a small, right neighbourhood of $\tilde{y}$, i.e. $[\tilde{y},\tilde{y}+\varepsilon)=:\mathcal{R}_{\varepsilon}^{+}(\tilde{y})$, with sufficiently small $\varepsilon>0$. * ad 1. We know that $\xi^{\prime}(y)$ is decreasing in $\mathcal{R}_{\varepsilon}^{+}(\tilde{y})$. On the other hand, since $\xi^{\prime\prime}(y)=\xi^{\prime}(y)\left(\frac{\xi(y)}{2y}-\frac{1}{4}\right)>0,$ and $\xi(y)<\nicefrac{{y}}{{2}}$ we deduce that $\xi^{\prime\prime}(y)>0$ for $y>\tilde{y}$. Then the function $\xi^{\prime}(y)$ cannot decrease for any $y>\tilde{y}$; * ad 2. Here $\xi^{\prime}(y)$ is increasing in $\mathcal{R}^{+}_{\varepsilon}(\tilde{y})$. But in the same manner as in the previous case, we see that $\xi^{\prime\prime}(y)<0$. A contradiction ends this part of the proof since integrating the above inequality and using the initial conditions we obtain the remaining part of the conclusion. (iii) We have shown that $\xi^{\prime}(y)>0$ and $\xi^{\prime\prime}(y)<0$. We have to consider two possibilities: either the limit of $\xi^{\prime}(y)$ at $\infty$ is equal to $0$ or it is a positive constant less than $\nicefrac{{1}}{{2}}$. We will exclude the second case. First of all we rewrite (23a) as $\frac{\xi^{\prime\prime}(y)}{\xi^{\prime}(y)}=\left(\frac{\xi(y)}{2y}-\frac{1}{4}\right).$ Using (ii) we show that the right hand side of the above equality is strictly negative, namely $\frac{\xi^{\prime\prime}(y)}{\xi^{\prime}(y)}\leqslant-\delta=-\frac{1}{2}\bigg{(}\frac{1}{2}-a\bigg{)}<0.$ (25) We can write (25) in the form $\left(\ln\xi^{\prime}(y)\right)^{\prime}\leqslant-\delta,$ and after the integration from $0$ to some $y>y_{0}$, we obtain $\begin{split}\ln\xi^{\prime}(y)&<\ln\xi^{\prime}(0)-\delta y,\\\ \xi^{\prime}(y)&<{\,\text{e}}^{\ln a}{\,\text{e}}^{-\delta y},\\\ 0<\xi^{\prime}(y)&<a{\,\text{e}}^{-\delta y}.\end{split}$ Let $y$ go to $\infty$. Then $\lim_{y\rightarrow\infty}\xi^{\prime}(y)\leqslant a\lim_{y\rightarrow\infty}{\,\text{e}}^{-\delta y}\leqslant 0.$ Integrating again the inequality $\xi^{\prime}(y)<a{\,\text{e}}^{-\delta y}$ on the interval $[0,y]$, we obtain $\begin{split}\xi(y)-\xi(0)&<a\int_{0}^{y}{\,\text{e}}^{-\delta s}\,\text{d}s,\\\ \xi(y)&<\frac{a}{\delta}\big{(}1-{\,\text{e}}^{-\delta y}\big{)}=\frac{4a}{1-2a}\big{(}1-{\,\text{e}}^{\nicefrac{{1}}{{2}}(a-\nicefrac{{1}}{{2}})y}\big{)},\end{split}$ (26) which concludes the argument. $\blacksquare$ To prove the existence of solutions of the problem (23) we transform it into an integral one and apply a Banach fixed point theorem. Multiplying equation (23a) by ${\,\text{e}}^{\nicefrac{{y}}{{4}}}$, integrating and using the initial conditions (1b), we obtain $\displaystyle\Big{(}{\,\text{e}}^{\nicefrac{{y}}{{4}}}\xi^{\prime}(y)\Big{)}^{\prime}=$ $\displaystyle\ \frac{{\,\text{e}}^{\nicefrac{{y}}{{4}}}}{2y}\xi(y)\xi^{\prime}(y),$ $\displaystyle\xi^{\prime}(y)=$ $\displaystyle\ a{\,\text{e}}^{-\nicefrac{{y}}{{4}}}+\frac{1}{2}{\,\text{e}}^{-\nicefrac{{y}}{{4}}}\int_{0}^{y}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\xi(s)\xi^{\prime}(s)\,\text{d}s,$ $\displaystyle\xi(y)=$ $\displaystyle\ 4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{y}}{{4}}}\Big{)}+\frac{1}{2}\int_{0}^{y}\bigg{(}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\xi(s)\xi^{\prime}(s)\,\text{d}s\bigg{)}\,\text{d}t.$ We look for fixed points of the operator $\mathcal{H}(\xi)(y)=4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{y}}{{4}}}\Big{)}+\frac{1}{2}\int_{0}^{y}\bigg{(}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\xi(s)\xi^{\prime}(s)\,\text{d}s\bigg{)}\,\text{d}t,$ in the space ${\mathcal{T}}=C^{1}[0,Y_{0}]\cap\Big{\\{}\xi:\ \xi(0)=0,\ \xi^{\prime}(0)=a,\ 0<\xi^{\prime}(y)<\tfrac{1}{2},\text{\ for\ }y\in[0,Y_{0}]\Big{\\}},$ endowed with the usual $C^{1}$ norm $\\|\xi\\|_{{\mathcal{T}}}=\sup_{0\leqslant y\leqslant Y_{0}}\xi(y)+\sup_{0\leqslant y\leqslant Y_{0}}\xi^{\prime}(y)=\\|\xi\\|_{C}+\\|\xi\\|_{C^{1}}\hskip 10.0pt.$ ###### Lemma 4.3 There exists $Y_{0}>0$ such that for a given $0<a<\nicefrac{{1}}{{2}}$ the problem (23) has a solution in the space ${\mathcal{T}}$. Moreover this solution is unique. Proof We widely use here the properties of $\xi(y)$, which follow from the choice of the space ${\mathcal{T}}$ and Lemma 4.2. Let us begin with $\begin{split}\big{\\|}\mathcal{H}(\xi)\big{\\|}_{C}\leqslant&\ 4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}+\frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\|\xi(s)\xi^{\prime}(s)\|\,\text{d}s\,\text{d}t\\\ \leqslant&\ 4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}+\frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\|\xi(s)\|\,\\|\xi\\|_{C^{1}}\,\text{d}s\,\text{d}t\\\ \leqslant&\ 4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}+\frac{1}{2}\\|\xi\\|_{C^{1}}^{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}{\,\text{e}}^{\nicefrac{{s}}{{4}}}\,\text{d}s\,\text{d}t\\\ \leqslant&\ 4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}+2\\|\xi\\|_{C^{1}}^{2}\sup_{0\leqslant y\leqslant Y_{0}}\big{(}y+4{\,\text{e}}^{-\nicefrac{{y}}{{4}}}-4\big{)}\\\ \leqslant&\ 4a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}+2Y_{0}\\|\xi\\|_{C^{1}}^{2}\hskip 10.0pt.\end{split}$ Similarly, we have $\begin{split}\\|\mathcal{H}(\xi)\\|_{C^{1}}=&\ \sup_{0\leqslant y\leqslant Y_{0}}\bigg{\|}a{\,\text{e}}^{-\nicefrac{{y}}{{4}}}+\frac{1}{2}\frac{\text{d}}{\text{d}y}\bigg{(}\int_{0}^{y}\Big{(}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\xi(s)\xi^{\prime}(s)\,\text{d}s\Big{)}\,\text{d}t\bigg{)}\Bigg{\|}\\\ \leqslant&\ a+\frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}{\,\text{e}}^{-\nicefrac{{y}}{{4}}}\int_{0}^{y}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\|\xi(s)\|\,\|\xi^{\prime}(s)\|\,\text{d}s\\\ \leqslant&\ a+\frac{1}{2}\\|\xi\\|_{C^{1}}\sup_{0\leqslant y\leqslant Y_{0}}{\,\text{e}}^{-\nicefrac{{y}}{{4}}}\int_{0}^{y}as\,\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\text{d}s=\\\ \leqslant&\ a+2a\\|\xi\\|_{C^{1}}\sup_{0\leqslant y\leqslant Y_{0}}\Big{(}1-{\,\text{e}}^{-\nicefrac{{y}}{{4}}}\Big{)}=a+2a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}\\|\xi\\|_{C^{1}}\hskip 10.0pt,\end{split}$ which gives $\\|\mathcal{H}(\xi)\\|_{{\mathcal{T}}}\leqslant a+2a\Big{(}1-{\,\text{e}}^{-\nicefrac{{Y_{0}}}{{4}}}\Big{)}\big{(}2+\\|\xi\\|_{C^{1}}\big{)}+2Y_{0}\\|\xi\\|_{C^{1}}.$ To show the contraction property for the operator $\mathcal{H}$ we estimate $\begin{split}\\|\mathcal{H}(\xi)-\mathcal{H}(\eta)\\|_{C}\leqslant&\ \frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\|\xi(s)\xi^{\prime}(s)-\eta(s)\eta^{\prime}(s)\|\,\text{d}s\,\text{d}t\\\ \leqslant&\ \frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\|\xi(s)\|\cdot\|\xi^{\prime}(s)-\eta^{\prime}(s)\|\,\text{d}s\,\text{d}t\\\ &+\frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{{\,\text{e}}^{\nicefrac{{s}}{{4}}}}{s}\,\|\xi(s)-\eta(s)\|\cdot\|\eta^{\prime}(s)\|\,\text{d}s\,\text{d}t\\\ \leqslant&\ \frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}a{\,\text{e}}^{\nicefrac{{s}}{{4}}}\sup_{0\leqslant s\leqslant Y_{0}}\|\xi^{\prime}(s)-\eta^{\prime}(s)\|\,\text{d}s\,\text{d}t\\\ &+\frac{1}{2}\sup_{0\leqslant y\leqslant Y_{0}}\int_{0}^{y}{\,\text{e}}^{-\nicefrac{{t}}{{4}}}\int_{0}^{t}\frac{1}{2}\,{\,\text{e}}^{\nicefrac{{s}}{{4}}}\sup_{0\leqslant s\leqslant Y_{0}}\|\xi^{\prime}(s)-\eta^{\prime}(s)\|\,\text{d}s\,\text{d}t\\\ \leqslant&\ \\|\xi-\eta\\|_{C^{1}}\big{(}2a+1\big{)}\sup_{0\leqslant y\leqslant Y_{0}}\Big{(}y+4{\,\text{e}}^{-\nicefrac{{y}}{{4}}}-4\Big{)}\\\ \leqslant&\ Y_{0}\big{(}2a+1\big{)}\\|\xi-\eta\\|_{C^{1}}\hskip 5.0pt.\end{split}$ We estimate the term $\\|\mathcal{H}(\xi)-\mathcal{H}(\eta)\\|_{\mathcal{C}^{1}}$ in the same way, to get $\\|\mathcal{H}(\xi)-\mathcal{H}(\eta)\\|_{{\mathcal{T}}}\leqslant C(Y_{0})\\|\xi^{\prime}-\eta^{\prime}\\|_{\mathcal{T}}.$ Choosing a sufficiently small $Y_{0}$ we justify, as in Theorem 2.2, the existence of the solution to the integral problem. To complete the proof we have to show that $\xi(y)$ constructed above solves the differential equation (23a). For all $y\in(0,Y_{0})$ our solution is a smooth function since the singularity in the coefficient $\nicefrac{{1}}{{y}}$ disapears. Then we can consider the same equation but on some interval not containing zero. The uniqueness of the solutions ends the argument. $\blacksquare$ ###### Lemma 4.4 The solution obtained in Lemma 4.3 exists for all real $y\in[0,\infty)$. Proof Let us assume that the solution $\overline{\xi}(y)$ is not a global one. Then, there exists a finite $Y_{\text{max}}=\sup\\{y:\ \overline{\xi}(y)\text{\ exists and is in the space\ }\mathcal{T}\\}\quad.$ (27) Using monotonicity and boundedness of the solution on the existence interval given by Lemma 4.2, we can let $y$ tend to $Y_{\text{max}}$. Let us denote $\overline{\xi}(y_{\text{max}})$ by $\alpha$ and $\overline{\xi}^{\prime}(y_{\text{max}})$ by $\beta$. Then $\displaystyle\xi^{\prime\prime}(y)+\frac{1}{4}\xi^{\prime}(y)-\frac{1}{2y}\xi(y)\xi^{\prime}(y)=0,$ (28) $\displaystyle\xi(Y_{\text{max}})=\alpha,\quad\xi^{\prime}(Y_{\text{max}})=\beta.$ (29) is an ordinary differential Cauchy problem which is not a singular one on the interval $[Y_{\text{max}},Y_{0}^{\prime})$ for some $Y_{0}^{\prime}>Y_{\text{max}}$. Then there exists the solution $\tilde{\xi}(y)$ of this problem, on the interval $[Y_{\text{max}},Y_{0}^{\prime})$ (see, e.g. [13]). Since the function $\eta(y)=\begin{cases}\overline{\xi}(y)&\text{\quad for \quad}y\in[0,Y_{0}),\\\ \tilde{\xi}(y)&\text{\quad for \quad}y\in[Y_{0},Y_{0}^{\prime})\end{cases}$ belongs to $\mathcal{T}$ on the interval $[0,Y_{0}^{\prime})$, we get a contradiction with the maximality of $Y_{\text{max}}$. $\blacksquare$ Now we can prove the main result of this paper. Proof of Theorem 4.1 Lemmas 4.3 and 4.4 assure the existence of the unique solution of the problem (23a)–(23b) for $y\in[0,\infty)$ and $a\in[0,\nicefrac{{1}}{{2}})$. Moreover, from the inequality (26) we obtain the estimate from above for total mass of the solution, i.e. $\xi(+\infty)$. As in [23, Theorem 21], we use the comparison principle for a second order ordinary differential equation, to show the estimates from below of our solution. In order to do this we have to regularize our problem and show that all the estimates are independent of the introduced regularization parameter. Indeed, let $\xi_{\varepsilon}$ solve the equation $F(\xi_{\varepsilon}):=\xi_{\varepsilon}^{\prime\prime}(y)+\frac{1}{4}\xi_{\varepsilon}^{\prime}(y)-\frac{1}{2y+\varepsilon}\;\xi_{\varepsilon}(y)\;\xi_{\varepsilon}^{\prime}(y)=0,$ (30) with the initial conditions $\xi_{\varepsilon}(0)=0,\ \xi_{\varepsilon}^{\prime}(0)=\theta$ for some positive $\varepsilon$ and $\theta$. Then for $h(y)=A\big{(}1-{\,\text{e}}^{-\mu y}\big{)}$, with $A=\tfrac{4a}{1-2a}$ $\begin{split}F\big{(}h(y)\big{)}=&-\mu^{2}A{\,\text{e}}^{-\mu y}-A\mu{\,\text{e}}^{-\mu y}\Big{(}\frac{1}{2y+\varepsilon}A\big{(}1-{\,\text{e}}^{-\mu y}\big{)}-\frac{1}{4}\Big{)}\\\ =&-\frac{\mu A}{4(2y+\varepsilon)}{\,\text{e}}^{-\mu y}\Big{(}(2y+\varepsilon)(4\mu-1)+4A\big{(}1-{\,\text{e}}^{-\mu y}\big{)}\Big{)},\end{split}$ which is not positive for all $y\in\mathbb{R}^{+}$, uniformly with respect to $\varepsilon$. Existence of the solutions to the problem (30) is a standard fact (see, e.g., [13]). It suffices to show that family $\\{\xi_{\varepsilon}\\}_{0<\varepsilon<1}$ is equicontinuous on every finite interval $[0,Y_{0}]$. Then, passing to the limit $\varepsilon\rightarrow 0$ we conclude the proof. $\blacksquare$ Now we see explicitly the equivalence of the both formulations of our problem: with the initial conditions $\xi(0)=0,\ \xi^{\prime}(0)=a$ and the boundary conditions $\xi(0)=0,\ \xi(\infty)=M$. Taking $\xi^{\prime}(0)=a$ we solve our system with total mass $\tfrac{4a}{1-2a}$ and if we want to know solutions with mass M we have to take $\xi^{\prime}(0)=\tfrac{M}{2M+4}$ in (23b). ## 5 Asymptotics Before proving the main result of this section we recall one more technical fact (for the proof see [17, Lemma 6.1]. ###### Lemma 5.1 Let $h(x)$ be a nonnegative function in $L^{1}(0,1)$ such that $\int_{0}^{1}h(x)\,{\rm d}x<1$. Moreover, we assume that $\alpha(t),\ \beta(t):\ (0,\infty)\rightarrow\mathbb{R}^{+}$ are bounded and such that $\alpha(t)\leqslant\beta(t)+\int_{0}^{1}h(x)\alpha(xt)\,{\rm d}x,$ (31) for all $t\in(0,\infty)$. Then $\lim_{t\rightarrow\infty}\beta(t)=0$ implies $\lim_{t\rightarrow\infty}\alpha(t)=0$. In the case of the problem (1) we are able to prove analogue of [18, Theorem 3], namely the following theorem. ###### Theorem 5.2 Under the assumptions of Theorem 2.2, let $u,\ v$ be two solutions of the problem (1) with the initial conditions $u_{0},\ v_{0}\in B\mathcal{X}$, respectively, obtained in Theorem 2.2. Then, for sufficiently small $\varepsilon>0$ the relations $\lim_{t\rightarrow\infty}t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(t)(u_{0}-v_{0})\big{\\|}_{\nicefrac{{4}}{{3}}}=0$ (32) and $\lim_{t\rightarrow\infty}t^{\nicefrac{{1}}{{4}}}\big{\\|}u(\cdot,t)-v(\cdot,t)\big{\\|}_{\nicefrac{{4}}{{3}}}=0$ (33) are equivalent. Proof The integral representation of the solution (3) yields $\displaystyle u(t)-v(t)$ $\displaystyle=\mathcal{S}(t)(u_{0}-v_{0})+\int_{0}^{t}\mathcal{S}(t-\tau)\big{(}\mathcal{B}(u,u)-\mathcal{B}(v,v)\big{)}(\tau)\,\text{d}\tau$ $\displaystyle=\mathcal{S}(t)(u_{0}-v_{0})+\int_{0}^{t}\mathcal{S}(t-\tau)\big{(}\mathcal{B}(u,u-v)-\mathcal{B}(v,u-v)\big{)}(\tau)\,\text{d}\tau.$ Then $\big{\\|}u(t)-v(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant\big{\\|}\mathcal{S}(t)(u_{0}-v_{0})\big{\\|}_{\nicefrac{{4}}{{3}}}\\\ +C\int_{0}^{t}(t-\tau)^{-\nicefrac{{3}}{{4}}}\Big{(}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}+\big{\\|}v(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\Big{)}\big{\\|}(u-v)(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\,\text{d}\tau.$ (34) According to assumptions in Theorem 2.2 $\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}u(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant 2\varepsilon\text{\quad and \quad}\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}v(t)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant 2\,\varepsilon.$ Multiplying the inequality (34) by $t^{\nicefrac{{1}}{{4}}}$ and denoting $\alpha(t)=t^{\nicefrac{{1}}{{4}}}\big{\\|}u(t)-v(t)\big{\\|}_{\nicefrac{{4}}{{3}}},\quad\beta(t)=t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(t)(u_{0}-v_{0})\big{\\|}_{\nicefrac{{4}}{{3}}},$ we get $\displaystyle\alpha(t)\leqslant$ $\displaystyle\ \beta(t)+C\int_{0}^{t}(t-\tau)^{-\nicefrac{{3}}{{4}}}t^{\nicefrac{{1}}{{4}}}\tau^{-\nicefrac{{1}}{{4}}}\times$ $\displaystyle\times\Big{(}\tau^{\nicefrac{{1}}{{4}}}\big{\\|}u(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}+\tau^{\nicefrac{{1}}{{4}}}\big{\\|}v(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\Big{)}\big{\\|}(u-v)(\tau)\big{\\|}_{\nicefrac{{4}}{{3}}}\,\text{d}\tau$ $\displaystyle=$ $\displaystyle\ \beta(t)+4\,\varepsilon\,C\int_{0}^{1}(1-w)^{-\nicefrac{{3}}{{4}}}w^{-\nicefrac{{1}}{{2}}}\alpha(wt)\,\text{d}w.$ Since $h(w)=4\,\varepsilon\,C(1-w)^{-\nicefrac{{3}}{{4}}}w^{-\nicefrac{{1}}{{2}}}\in L^{1}(0,1)$ and, if necessary, at the expense of decreasing of $\varepsilon>0$, the quantity $4\,\varepsilon\,C\int_{0}^{1}(1-w)^{-\nicefrac{{3}}{{4}}}w^{-\nicefrac{{1}}{{2}}}\,\text{d}w$ is strictly less then $1$, we can apply Lemma 5.1 to obtain that the relation (32) implies (33). To obtain the reverse implication we proceed in a similar way. First let us estimate $\begin{split}t^{\nicefrac{{1}}{{4}}}&\big{\\|}u(t)-v(t)-\mathcal{S}(t)\big{(}u_{0}-v_{0}\big{)}\big{\\|}_{\nicefrac{{4}}{{3}}}\\\ &\leqslant C\int_{0}^{1}(1-w)^{-\nicefrac{{3}}{{4}}}w^{-\nicefrac{{1}}{{2}}}(wt)^{\nicefrac{{1}}{{4}}}\big{\\|}u(wt)-v(wt)\big{\\|}_{\nicefrac{{4}}{{3}}}\,\text{d}w.\end{split}$ (35) Then we notice that $\displaystyle\sup_{w\in[0,1]}(wt)^{\nicefrac{{1}}{{4}}}\big{\\|}u(wt)-v(wt)\big{\\|}_{\nicefrac{{4}}{{3}}}\leqslant\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}u(t)-v(t)\big{\\|}_{\nicefrac{{4}}{{3}}}$ $\displaystyle\leqslant\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}u(t)\big{\\|}_{\nicefrac{{4}}{{3}}}+\sup_{t>0}t^{\nicefrac{{1}}{{4}}}\big{\\|}v(t)\big{\\|}_{\nicefrac{{4}}{{3}}}<\infty,$ and use the Lebesgue dominated convergence theorem. We obtain that the right hand side of the inequality (35) converges to zero as $t$ tends to $\infty$. Then we deduce that the relation (33) implies (32). $\blacksquare$ We derive as a corollary of Theorem 5.2 a remark on local asymptotic stability of self–similar solutions or, in other words, the self–similar asymptotics of solutions to (3) with small initial data. ###### Corollary 5.3 Suppose that $u_{0}$ satisfies the assumptions of Theorem 2.2, $\int_{\mathcal{R}^{2}}u_{0}(x)\,\text{d}x=M$ and $v_{0}(x)=U_{M}(x,1)$, where $U_{M}$ is the unique self–similar solution with $\int_{\mathcal{R}^{2}}U_{M}(x,1)\,\text{d}x=M$. Then $\lim_{t\rightarrow\infty}t^{\nicefrac{{1}}{{4}}}\big{\\|}u(\cdot,t)-U_{M}(\cdot,t)\big{\\|}_{\nicefrac{{4}}{{3}}}=0$ holds. Proof Indeed, if $v_{0}(x)$ is a regular initial condition with $\int_{\mathcal{R}^{2}}v_{0}(x)\,\text{d}x=M$, then $\lim_{t\rightarrow\infty}t^{\nicefrac{{1}}{{4}}}\big{\\|}\mathcal{S}(t)(u_{0}-v_{0})\big{\\|}_{\nicefrac{{4}}{{3}}}=0,$ cf. [17], because the $L^{p}$–asymptotics of the heat semigroup is determined by the Gauss kernel multiplied by $M$, namely $\lim_{t\rightarrow\infty}t^{1-\nicefrac{{1}}{{p}}}\big{\\|}\mathcal{S}(t)v_{0}-M\mathcal{S}(t)\delta_{0}\big{\\|}_{p}=0,\quad 1\leqslant p\leqslant\infty.$ This is the reason why we took the self–similar profile at the time $t=1$, not the singular data at $t=0$. $\blacksquare$ We expect that the local result in the above corollary can be extended to the global asymptotic stability of self–similar solutions on all $L^{p}$–spaces, that is $\lim_{t\rightarrow\infty}t^{1-\nicefrac{{1}}{{p}}}\big{\\|}u(\cdot,t)-U_{M}(\cdot,t)\big{\\|}_{p}=0$ (36) (which is in fact implied by (36) for $p=1$), for each $u_{0}\in B\mathcal{X}$ and $U_{M}$ with $M=\int_{\mathcal{R}^{2}}u_{0}(x)\,\text{d}x$. The proof of this result seems to be beyond the scope of methods in that paper. Indeed, imitating the variational approach as was in [8] for the problem with the gravitational interactions (in particular: energy and entropy functionals as well as entropy dissipation relation), one may first prove a priori estimates on $u$ and then a bound on $\int_{\mathcal{R}^{2}}u(x,t)\log\frac{u(x,t)}{U_{M}(x,t)}\,\text{d}x$, leading to (36). ## References * [1] V. Babskii, Z. Zhukov, and V. Yudovich. Mathematical Theory of Electrophoresis – Applications to Methods of Fractionation of Biopolymers. Naukova Dumka, Kiev, 1983. * [2] P. Biler. Existence and asymptotics of solutions for a parabolic-elliptic system with nonlinear no-flux boundary condition. Nonlinear Analysis, 19(12):1121–1136, 1992. * [3] P. Biler. The Cauchy problem and self-similar solutions for a nonlinear parabolic equation. Studia Math., 114(2):181–205, 1995. * [4] P. Biler and J. Dolbeault. Long time behavior of solutions of Nernst-Planck and Debye-Hueckel drift-diffusion systems. Ann. Henri Poincaré, 1(3):461–472, 2000. * [5] P. Biler, W. Hebisch, and T. Nadzieja. The Debye system: existence and large time behavior of solutions. Nonlinear Analysis, 23(9):1189–1209, 1994. * [6] P. Biler and T. Nadzieja. Existence and nonexistence of solutions for a model of gravitational interaction of particles. I. Colloq. Math., 66(2):319–334, 1994. * [7] P. Biler, T. Nadzieja, and A. Raczyński. Nonlinear singular parabolic equations. In Reaction diffusion systems (Trieste, 1995), volume 194 of Lecture Notes in Pure and Appl. Math., pages 21–36. Dekker, New York, 1998\. * [8] A. Blanchet, J. Dolbeault, and B. Perthame. Two-dimensional Keller-Segel model: optimal critical mass and qualitative properties of the solutions. Electron. J. Diff. Eq., 44, 2006. * [9] Y. S. Choi and R. Lui. Multi-dimensional electrochemistry model. Arch. Rational Mech. Anal., 130(4):315–342, 1995. * [10] P. Debye and E. Hueckel. Zur Theorie der Electrolyte II. Phys. Zeit., 24:305–325, 1923. * [11] H. Gajewski. On existence, uniqueness and asymptotic behavior of solutions of the basic equations for carrier transport in semiconductors. Z. Angew. Math. Mech., 65(2):101–108, 1985. * [12] H. Gajewski and K. Groeger. On the basic equations for carrier transport in semiconductors. J. Math. Anal. Appl., 113(1):12–35, 1986. * [13] P. Hartman. Ordinary Differential Equations. S. M. Hartman, Baltimore, 1973. Corrected reprint. * [14] D. Horstmann. From 1970 until present: the Keller-Segel model in chemotaxis and its consequences. I. Jahresber. Deutsch. Math.-Verein., 105(3):103–165, 2001. * [15] D. Horstmann. From 1970 until present: the Keller-Segel model in chemotaxis and its consequences. II. Jahresber. Deutsch. Math.-Verein., 106(2):51–69, 2001. * [16] A. Jüngel. Quasi-hydrodynamic Semiconductor Equations. Progress in Nonlinear Differential Equations and their Applications, 41\. Birkhäuser Verlag, Basel, 2001. * [17] G. Karch. Scaling in nonlinear parabolic equations. J. Math. Anal. Appl., 234(2):534–558, 1999. * [18] G. Karch. Scaling in nonlinear parabolic equations: applications to Debye system. In Disordered and complex systems (London, 2000), volume 553 of AIP Conf. Proc., pages 243–248. Amer. Inst. Phys., Melville, NY, 2001. * [19] E. H. Lieb and M. Loss. Analysis. Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, second edition, 2001. * [20] P. A. Markowich, C. A. Ringhofer, and C. Schmeiser. Semiconductor Equations. Springer-Verlag, Vienna, 1990. * [21] J. Nash. Continuity of solutions of parabolic and elliptic equations. Amer. J. Math., 80:931–954, 1958. * [22] M. Olech. Nonuniqueness of steady states in annular domains for Streater equations. Appl. Math. (Warsaw), 31(3):303–312, 2004. * [23] M. H. Protter and H. F. Weinberger. Maximum Principles in Differential Equations. Springer-Verlag, New York, 1984. Corrected reprint of the 1967 original. * [24] T. Suzuki. Free Energy and Self-interacting Particles. Progress in Nonlinear Differential Equations and Their Applications, 62\. Birkhäuser, Boston, MA, 2004.	arxiv-papers	2008-08-01T14:13:40	2024-09-04T02:48:57.100275	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "A. Herczak and M. Olech", "submitter": "Michal Olech Mr", "url": "https://arxiv.org/abs/0808.0123" }
0808.0127	# New CP phase of $B_{s}-\bar{B}_{s}$ mixing on ${\rm T}$ violation in $B_{d(s)}\to K^{}(\phi)\ell^{+}\ell^{-}$ Chuan-Hung Chen1,2111Email: physchen@mail.ncku.edu.tw, Chao-Qiang Geng3222Email: geng@phys.nthu.edu.tw and Lin Li1333Email:lilin@phys.sinica.edu.tw 1Department of Physics, National Cheng- Kung University, Tainan 701, Taiwan 2National Center for Theoretical Sciences, Hsinchu 300, Taiwan 3Department of Physics, National Tsing-Hua University, Hsinchu 300, Taiwan ###### Abstract A large CP-violating phase uncovered recently by CDF and D$\O$ collaborations in the time-dependent CP asymmetry (CPA) of the $B_{s}\to J/\Psi\phi$ decay clearly indicates that a non-Cabibbo-Kobayashi-Maskawa (CKM) phase has to be brought into the $b\to s$ transition. We find that the model with $SU(2)_{L}$ singlet exotic quarks can not only provide the new phase induced from the Z-mediated flavor changing neutral current (FCNC) at tree level, but also strongly relate the $B_{s}-\bar{B}_{s}$ mixing, $B_{q}\to V_{q}\ell^{+}\ell^{-}$ ($V_{d[s]}=K^{}[\phi])$ and $B_{s}\to\mu^{+}\mu^{-}$ together. In particular, we show that the new CP phase can be unambiguously exposed by the large statistical significances of $T$ violating observables in $B_{q}\to V_{q}\ell^{+}\ell^{-}$, while the branching ratio of $B_{s}\to\mu^{+}\mu^{-}$ can be enhanced to be $O(10^{-8})$. CP violation (CPV) has been one of the most mysterious phenomena in high energy physics since it was discovered in the K system CP . At B factories, BABAR and BELLE have observed both the mixing-induced time-dependent CP asymmetry (CPA) in the $B_{d}$ oscillation through the golden mode of $B\to J/\Psi K_{S}$ and the direct CPAs in exclusive $B\to\pi\pi$ and $B\to\pi K$ decays HFAG , where the former is dictated by the $b\to d$ transition while the latter $b\to s$. Although three generations of the standard model (SM) can provide a unique CP violating phase in the Cabibbo-Kobayashi-Maskawa (CKM) matrix CKM to interpret the observed CPAs, it does not provide a solution to understand the matter-antimatter asymmetry in the Universe and it does not stop people to search for a new CPV source. Nevertheless, it is very difficult to unfold a new CPV phase in the direct CPAs of the nonleptonic exclusive decays due to the inevitable large uncertainty of nonperturbative QCD effects. Hence, the best environment to look for the new phase is that the QCD effects are less involved while the SM contributions are highly suppressed. Now, the dawn to see the new effects could be in the $B_{s}$ system. By the $B_{s}$ production in Tevatron Run II, besides CDF and $D{\O}$ observations on the $B_{s}$ oscillation of $\Delta m_{s}=17.77\pm 0.10\pm 0.07$ ps-1 CDF and $\Delta m_{s}=18.56\pm 0.87$ ps-1 D0 , respectively, and the large direct CPA of $0.39\pm 0.17$ for $B_{s}\to K^{-}\pi^{+}$ BsCP , an unexpected large CPV phase has been detected in the mixing-induced CPA for $B_{s}\to J/\Psi\phi$. To explain the new phase, we write the transition matrix element for $\bar{B}_{s}\to B_{s}$ as $\displaystyle M^{s}_{12}$ $\displaystyle=$ $\displaystyle A^{\rm SM}_{12}e^{-2i\beta_{s}}+A^{\rm NP}_{12}e^{2i(\theta^{\rm NP}_{s}-\beta_{s})}$ (1) where $\beta_{s}\equiv arg(-V_{ts}V^{}_{tb}/V_{cs}V^{}_{cb})$ is the CPV phase in the SM and $\theta^{NP}_{s}$ is the new CPV phase in some extension of the SM. Here, the convention has been chosen to be the same as that in Ref. UTfit . Due to $\Delta\Gamma_{s}\ll\Delta m_{s}$ in the B-system PDG06 , the time-dependent CPA could be simplified to be $\displaystyle-S_{J/\Psi\phi}$ $\displaystyle\simeq$ $\displaystyle{\rm Im}\left(\sqrt{\frac{M^{s^{}}_{12}}{M^{s}_{12}}}\right)=\sin(2\beta_{s}-\phi^{\rm NP}_{s})\,,$ $\displaystyle\phi^{NP}_{s}$ $\displaystyle=$ $\displaystyle\arctan\left(\frac{r\sin 2\theta^{\rm NP}_{s}}{1-r\cos 2\theta^{\rm NP}_{s}}\right)$ (2) with $r=A^{NP}_{12}/A^{\rm SM}_{12}$. By adopting Wolfenstein parametrization Wolfenstein of the CKM matrix up to $O(\lambda^{4})$, in which $V_{tb}=1-A^{2}\lambda^{4}/2$, $V_{ts}=-A\lambda^{2}+1/2A\lambda^{4}\left(1-2(\rho+i\eta)\right)$, $V_{cb}=A\lambda^{2}$ and $V_{cs}=1-\lambda^{2}/2-1/8\lambda^{4}(1+4A^{2})$ Buras , one can easily find $\displaystyle\beta_{s}\approx\lambda^{2}\eta\approx 0.019\,,$ (3) where $\lambda=0.2272$ and $\eta=0.359$ PDG06 have been used. Clearly, the mixing-induced CPA of $B_{s}$ in the SM is only few percent. Astonishingly, the nonvanished CP phase measured by CDF CDFCP and D$\O$ D0CP is $\displaystyle\phi_{s}=2\beta_{s}-\phi^{NP}_{s}=\left\\{\begin{array}[]{c}[0.24,\,1.36]\cup[1.78,\,2.82]\ (\textrm{CDF})\\\ 0.57^{+0.30+0.02}_{-0.24-0.07}\hskip 48.36958pt(\textrm{D$\O$})\end{array}\right.$ (6) at the $68\%$ confidence level (CL), while the allowed range at the $90\%$ CL by D$\O$ is given to be $\phi_{s}\in[-0.06,\,1.20]$. Recently, the UTfit collaboration has combined all available information in the $B_{s}$ system and concluded that the CPV phase of the $B_{s}$ mixing amplitude deviates more than $3\sigma$ from the SM value UTfit . To explain the large CPV phase in the $b\to s$ transition, several new physics models have been proposed models . In this paper, we explore the effects of the large phase in the decays of $B_{q}\to V_{q}\ell^{+}\ell^{-}$ ($V_{d(s)}=K^{}(\phi)$), corresponding to $b\to s\ell^{+}\ell^{-}$ at the quark level. In particular, we will show that the large CPV phase can be directly probed by measuring T-odd observables in the decays. To comprehend the beauty of using T violation to probe the CPV phase, we briefly summarize the characters of CP-odd and T-odd observables. In a decay process, the direct CPA or CP-odd observable is defined by ${\cal A}_{CP}\equiv(\bar{\Gamma}-\Gamma)/(\bar{\Gamma}+\Gamma)$, where $\Gamma$ ($\bar{\Gamma}$) is the partial decay rate of the (CP-conjugate) process. As a result, ${\cal A}_{CP}\propto\sin\theta_{w}\sin\theta_{st}$ with $\theta_{w(st)}$ being the weak (strong) phase. Clearly, to have a nonvanished CPA, both phases are needed. The efficiency on the CPA is mainly dictated by the uncertain calculations of the strong phase. Another way to probe the CPV phase is through the spin-momentum triple correlation, such as $\vec{s}\cdot(\vec{p}_{i}\times\vec{p}_{j})$ Garisto-CGL ; geng-BRD ; CG_NPB636 for a three-body decay, where $\vec{s}$ is the spin carried by one of outgoing particles and $\vec{p}_{i}$ and $\vec{p}_{j}$ denote any two independent momenta. The triple correlation is a T-odd observable since it changes sign under the time reversal (${\rm T}$) transformation of $t\rightarrow-t$. We note that the T transformation defined here is different from the real time-reversal transformation which also contains the interchange of initial and final states. By the CPT invariant theorem, T violation (TV) implies CPV. Therefore, the study of the T-odd observable can help us to understand the origin of CPV. Intriguingly, the T-odd triple correlation is proportional to $\sin(\theta_{w}+\theta_{st})$, which indicates that the strong phase is not necessary to achieve a nonzero T-odd observable. It has been shown in Ref. CG_NPB636 that T-violating effects in the exclusive $b\to s\ell^{+}\ell^{-}$ processes are sensitive to new physics with small QCD uncertainties, which could provide a good place to directly observe the new phase revealed by CDF and D$\O$. The transition amplitudes for $B_{q}\to V_{q}\ell^{+}\ell^{-}$ ($\ell=e,\,\mu$) are given by CG_NPB636 ; CG_PRD66 $\displaystyle{\cal M}_{V_{q}}^{(\lambda)}$ $\displaystyle=$ $\displaystyle-\frac{G_{F}\alpha\lambda_{t}}{2\sqrt{2}\pi}\left\\{{\cal M}_{1\mu}^{(\lambda)}L^{\mu}+{\cal M}_{2\mu}^{(\lambda)}L^{5\mu}\right\\}\,,$ $\displaystyle{\cal M}_{a\mu}^{(\lambda)}$ $\displaystyle=$ $\displaystyle if_{1}\varepsilon_{\mu\nu\alpha\beta}\epsilon^{\nu}(\lambda)P^{\alpha}q^{\beta}+f_{2}\epsilon_{\mu}^{}(\lambda)+f_{3}\epsilon^{}\cdot qP_{\mu}\,,$ where $\lambda_{t}=V_{tb}V_{ts}^{}$, $L_{\mu}=\bar{\ell}\gamma_{\mu}\ell$, $L^{5}_{\mu}=\bar{\ell}\gamma_{\mu}\gamma_{5}\ell$, $P=p_{B}+p_{V}$, $q=p_{B}-p_{V}$, $a=1\,(2)$ while $f_{i}=h_{i}$ $(g_{i})$ and $\displaystyle h_{1}$ $\displaystyle=$ $\displaystyle{C^{\rm eff}_{9}V\over m_{B}+m_{V}}+\frac{2m_{b}}{q^{2}}C_{7}T_{1}\,,$ $\displaystyle h_{2}$ $\displaystyle=$ $\displaystyle-\frac{1}{2}(m_{B}+m_{V})C^{\rm eff}_{9}A_{1}-\frac{1}{2}\frac{2m_{b}}{q^{2}}P\cdot qC_{7}T_{2}\,,$ $\displaystyle h_{3}$ $\displaystyle=$ $\displaystyle{C^{\rm eff}_{9}A_{2}\over m_{B}+m_{V}}+\frac{2m_{b}}{q^{2}}C_{7}\Big{(}T_{2}(q^{2})+\frac{q^{2}}{P\cdot q}T_{3}\Big{)}\,,$ $\displaystyle g_{i}$ $\displaystyle=$ $\displaystyle h_{i}\|_{C^{\rm eff}_{9}\to C_{10},C_{7}=0}\,,\ \ (i=1,2,3)\,.$ (7) Here, $m_{B}(m_{V})\equiv m_{B_{q}}(m_{V_{q}})$, $C^{\rm eff}_{9}$, $C_{7}$ and $C_{10}$ are the Wilson coefficients BBL and the definitions of the form factors in Eq. (7) can be found in Ref. CG_NPB636 . Furthermore, to obtain T-odd terms, the polarizations of $V_{q}$ should be kept in the averaged squared-amplitude. To achieve the requirement, we have to consider the decay chain $B_{q}\to V_{q}(\to P_{1}P_{2})\ell^{+}\ell^{-}$ in which $P_{1}P_{2}$ is $K\pi(KK)$ as $V_{q}=K^{}(\phi)$. Consequently, the differential decay rate associated with these terms is given by $\displaystyle\frac{d\Gamma}{d\cos\theta_{K}d\cos\theta_{\ell}d\phi dq^{2}}=\frac{3\alpha^{2}G_{F}^{2}\left\|\lambda_{t}\right\|^{2}\left\|\vec{p}\right\|}{2^{14}\pi^{6}m_{B}^{2}}$ $\displaystyle\times\\{4\cos^{2}\theta_{K}\sin^{2}\theta_{\ell}\sum_{i=1,2}\|{\cal M}_{i}^{0}\|^{2}+\sin^{2}\theta_{K}(1+\cos^{2}\theta_{\ell})$ $\displaystyle\sum_{i=1,2}\left(\|{\cal M}_{i}^{+}\|^{2}+\|{\cal M}_{i}^{-}\|^{2}\right)-\sin 2\theta_{K}\sin 2\theta_{\ell}\sin\phi$ $\displaystyle\sum_{i=1,2}{Im}\left({\cal M}_{i}^{+}-{\cal M}_{i}^{-}\right){\cal M}_{i}^{0}-2\sin^{2}\theta_{K}\sin^{2}\theta_{\ell}\sin 2\phi$ $\displaystyle\sum_{i=1,2}{Im}\left({\cal M}_{i}^{+}{\cal M}_{i}^{-}\right)+2\sin 2\theta_{K}\sin\theta_{\ell}\sin\phi(Im{\cal M}_{1}^{0}$ $\displaystyle({\cal M}_{2}^{+}+{\cal M}_{2}^{-})-Im({\cal M}_{1}^{+}+{\cal M}_{1}^{-}){\cal M}_{2}^{0})+\cdots]\\}\,,$ (8) where $\theta_{\ell}(\theta_{K})$ is the polar angle of the lepton (K-meson) in the $q^{2}$ ($V_{q}$) rest frame, $\|\vec{p}\|=[((m_{B}^{2}+m_{V}^{2}-q^{2})/(2m_{B}))^{2}-m_{V}^{2}]^{1/2}$ and ${\cal M}_{i}^{0}$ and ${\cal M}_{i}^{\pm}$ denote the longitudinal and transverse polarizations of $V_{q}$ with their explicit expressions being $\displaystyle{\cal M}_{a}^{0}$ $\displaystyle=$ $\displaystyle\sqrt{q^{2}}\left(\frac{E_{V}}{m_{V}}f_{2}+2\sqrt{q^{2}}\frac{\left\|\vec{p}_{V}\right\|^{2}}{m_{V}}f_{3}\right),$ $\displaystyle{\cal M}_{a}^{\pm}$ $\displaystyle=$ $\displaystyle\sqrt{q^{2}}\left(\pm 2\left\|\vec{p}_{V}\right\|\sqrt{q^{2}}f_{1}+f_{2}\right)\,,$ (9) respectively. To shorten the expressions, we have only presented the relevant pieces in Eq. (8), where the three imaginary terms denote the T-violating effects. The whole expression for differential decay rate could refer to Refs. CG_NPB636 ; BHP . From Eqs. (7), (8) and (9), it is easy to show that the large contributions to the T-violating effects arise from $Im{\cal M}_{1}^{0}({\cal M}_{2}^{+}+{\cal M}_{2}^{-})-Im({\cal M}_{1}^{+}+{\cal M}_{1}^{-}){\cal M}_{2}^{0}$. To explore the effects, we examine the T-odd observable, defined by $\left\langle{\cal O}_{T}\right\rangle=\int{\cal O}_{T}d\Gamma$ where ${\cal O}_{T}$ is a T-odd five-momentum correlation, given by ${\cal O}_{T}=\frac{\vec{p}_{B}\cdot\vec{p}_{K}}{\left\|\vec{p}_{B}\right\|\left\|\vec{p}_{K}\right\|}\frac{\vec{p}_{B}\cdot(\vec{p}_{K}\times\vec{p}_{\ell^{+}})}{\left\|\vec{p}_{B}\right\|\left\|\vec{p}_{K}\right\|\omega_{\ell^{+}}}\,$ (10) with $\omega_{\ell^{+}}=q\cdot p_{\ell^{+}}/\sqrt{q^{2}}$. In the $V_{q}$ rest frame, we note that ${\cal O}_{T}=\cos\theta_{K}\sin\theta_{K}\sin\theta_{\ell}\sin\phi$. To signal the nonvanished CPV phase, we employ the statistical significance of the observable, defined by $\varepsilon_{T}(q^{2})={\frac{\int{\cal O}_{T}d\Gamma}{\sqrt{(\int d\Gamma)(\int{\cal O}^{2}_{T}d\Gamma)}}}\,.$ (11) Integrating all relevant angles in Eq. (11), we obtain $\displaystyle\varepsilon_{T}(q^{2})$ $\displaystyle\simeq$ $\displaystyle\frac{0.76}{\sqrt{{\cal D}_{1}{\cal D}_{2}}}[Im{\cal M}_{1}^{0}({\cal M}_{2}^{+}+{\cal M}_{2}^{-})-$ $\displaystyle Im({\cal M}_{1}^{+}+{\cal M}_{1}^{-}){\cal M}_{2}^{0}]\,,$ $\displaystyle{\cal D}_{a}$ $\displaystyle=$ $\displaystyle\sum_{i=1,2}\left[\left\|{\cal M}_{i}^{0}\right\|^{2}+\frac{1}{a}\left(\left\|{\cal M}_{i}^{+}\right\|^{2}+\left\|{\cal M}_{i}^{-}\right\|^{2}\right)\right].\,\ $ (12) To observe the effect at $n\sigma$ level, the required number of B mesons is $N_{B}=n^{2}/(Br\cdot\varepsilon^{2}_{T})$. We now use the clue of the current data to illustrate our model-independent analysis. Although the principle of the minimal flavor violation (MFV) MFV could be as a dogma to rule the new source of CPV MFVCP , to focus on the criterion of the minimal extension of the SM, we employ the vector-like-quark model (VQM), in which the vector-like quarks (VQs) are $SU(2)_{L}$ singlet exotic quarks, as the ones naturally realized in $E_{6}$ models E_6 . Explicitly, we include $SU(2)_{L}$ singlet VQs in the SM, where the right- handed component is the same as the right-handed down-quark. Since the singlet quarks do not couple to W-bosons directly, one of the fascinating characters of the model is that the corresponding CKM matrix is not a unitary matrix. By introducing flavor mixing matrices to diagonalize the $4\times 4$ down-type quark mass matrix, we will display that the model interestingly leads to Z-mediated flavor changing neutral currents (FCNCs) at tree level ZFCNC , which clearly have significant impacts on the $B_{s}-\bar{B}_{s}$ mixing, $B_{q}\to V_{q}\ell^{+}\ell^{-}$ and $B_{s}\to\mu^{+}\mu^{-}$ processes. We note that the non-unitary CKM matrix could result in new contributions to the processes from box and penguin diagrams. However, to simply illustrate the new phase, only the Z-mediated effects at tree level are considered here, while those from the box and penguin diagrams could be referred to Ref. Aguilar- Saavedra . In the mass eigenstates, the coupling of the $Z$-boson to fermions is written by $\displaystyle{\cal L}_{Z}$ $\displaystyle=$ $\displaystyle-\frac{gc^{f}_{L}}{2\cos\theta_{W}}\bar{F}\gamma^{\mu}\left(V^{L}_{F}X_{F}V^{L^{\dagger}}_{F}\right)P_{L}FZ_{\mu}\,,$ $\displaystyle X_{Q}$ $\displaystyle=$ $\displaystyle\left[\begin{array}[]{ccc}\begin{array}[]{c}\openone_{3\times 3}\\\ \end{array}&\|&{\bf 0}_{3\times 1}\\\ -\ -\ -&-&\ -\\\ {\bf 0}_{1\times 3}&\|&\xi_{Q}\\\ \end{array}\right]\,,\ \ \ X_{\ell}=\openone_{3\times 3}\,,$ (17) where $g$ is the coupling constant of $SU(2)_{L}$, $\theta_{W}$ is the Weinberg’s angle, $P_{R(L)}=(1\pm\gamma_{5})/2$, $F^{T}=(q_{1},q_{2},q_{3},q_{4})$ and $(e,\mu,\tau)$ represent quarks (Qs) and leptons $(\ell s)$, $c^{f}_{L}$ is defined as $c^{f}_{L}=c^{f}_{V}+c^{f}_{A}$ with $\displaystyle c^{f}_{V}$ $\displaystyle=$ $\displaystyle T^{3}_{f}-2\sin^{2}\theta_{W}Q_{f}\,,\ \ \ c^{f}_{A}=T^{3}_{f}$ (18) in which $T^{3}_{f}$ and $Q_{f}$ are the third component of the weak isospin and the electric charge of the particle, respectively, and $\xi_{f}=-2\sin^{2}\theta_{W}Q_{f}/c^{f}_{L}$. Here, as the quantum number of the new right-handed VQ is the same as the right-handed down-quark, the tree FNCNs only occur in the left-handed quarks. Accordingly, the interaction for $b$-$s$-$Z$ is given by $\displaystyle{\cal L}_{b\to s}=\frac{gc^{D}_{L}\lambda_{23}}{2\cos\theta_{W}}\bar{s}\gamma^{\mu}P_{L}bZ_{\mu}+H.c.$ (19) with $\lambda_{23}=(1-\xi_{D})(V^{L}_{D})_{24}(V^{L}_{D})^{}_{34}\equiv\|\lambda_{23}\|\exp[i(\theta^{\rm NP}_{s}-\beta_{s})]$. By Eqs. (1) and (19), the transition matrix element for the $\Delta B=2$ process is obtained as $\displaystyle A^{\rm NP}_{12}$ $\displaystyle=$ $\displaystyle\frac{G_{F}\left(\lambda_{23}\right)^{2}}{3\sqrt{2}}m_{B_{s}}f^{2}_{B_{s}}\hat{B_{s}}\,.$ (20) As a result, the $B_{s}-\bar{B}_{s}$ mixing in the VQM is $\displaystyle\Delta m_{s}=\Delta m^{\rm SM}_{s}\left(1+2r\cos\theta^{\rm NP}_{s}+\|r\|^{2}\right)^{1/2}\,.$ (21) From the above equation, it is clear that a large new CPV phase can have a significant influence on the $B_{s}-\bar{B}_{s}$ mixing. By combining the SM and Z-mediated FCNC, the branching ratio (BR) of $B_{s}\to\mu^{+}\mu^{-}$ is found to be $\displaystyle{\cal B}_{\ell^{+}\ell^{-}}=\tau_{B_{s}}\frac{G^{2}_{F}}{16\pi}m_{B_{s}}f^{2}_{B_{s}}m^{2}_{\ell}\left(1-\frac{4m^{2}_{\ell}}{m^{2}_{B_{s}}}\right)^{1/2}\left(\|\Re\|^{2}+\|\Im\|^{2}\right),$ $\displaystyle\Re$ $\displaystyle=$ $\displaystyle-\frac{\|\lambda_{t}\|\alpha}{\pi\sin^{2}\theta_{W}}Y(x_{t})+\|\lambda_{23}\|c^{D}_{L}\cos\theta^{\rm NP}_{s},$ $\displaystyle\Im$ $\displaystyle=$ $\displaystyle\|\lambda_{23}\|c^{D}_{L}\sin\theta^{\rm NP}_{s}.$ (22) We can also obtain the effects of the Z-mediated FCNCs on $B_{q}\to V_{q}\ell^{+}\ell^{-}$ decays by utilizing the replacement: $\displaystyle C^{\rm eff}_{9}[V,A_{1(2)}]$ $\displaystyle\to$ $\displaystyle\left(C^{\rm eff}_{9}+\frac{4\lambda_{23}}{\alpha\lambda_{t}}c^{D}_{L}c^{\ell}_{V}\right)[V,A_{1(2)}],$ $\displaystyle C_{10}[V,A_{1(2)}]$ $\displaystyle\to$ $\displaystyle\left(C_{10}-\frac{4\lambda_{23}}{\alpha\lambda_{t}}c^{D}_{L}c^{\ell}_{A}\right)[V,A_{1(2)}].$ (23) We note that in the following numerical analysis we will concentrate on $B_{d}\to K^{}\ell^{+}\ell^{-}$ as they have already been observed. However, the same discussions can be easily applied to $B_{s}\to\phi\ell^{+}\ell^{-}$. In the Z-mediated $b\to s$ transition, the magnitude $\|\lambda_{23}\|$ and the phase $\theta^{\rm NP}_{s}$ can be determined by the observed $B_{s}$ mixing and BR of $B_{d}\to K^{0}\mu^{+}\mu^{-}$. We adopt the average value of $\Delta m_{s}=18.17\pm 0.86$ ps-1 CDF ; D0 and ${\cal B}(B_{d}\to K^{0}\mu^{+}\mu^{-})=(1.22^{+0.38}_{-0.32})\times 10^{-6}$ PDG06 as the inputs, while the SM results are taken to be $\Delta m^{\rm SM}_{s}=19.3\pm 6.7$ ps-1 LN , ${\cal B}(B_{d}\to K^{0}\mu^{+}\mu^{-})_{\rm SM}=1.3\times 10^{-6}$ CGA21 and ${\cal B}(B_{s}\to\mu^{+}\mu^{-})_{\rm SM}\approx 0.33\times 10^{-8}$ with $V_{ts}=-0.04$ and $f_{B_{s}}=0.23$ GeV. In order to reveal the strong correlations among $\Delta m_{s}$, ${\cal B}(B_{d}\to K^{0}\ell^{+}\ell^{-})$ and ${\cal B}(B_{s}\to\mu^{+}\mu^{-})$ influenced by the same parameters, we present ${\cal B}(B_{s}\to\mu^{+}\mu^{-})$ versus $\Delta m_{s}$ [${\cal B}(B_{d}\to K^{0}\ell^{+}\ell^{-})$] in Fig. 1(a)[(b)]. From the figures, we see clearly that the Z-mediated effects could enhance the BR of $B_{s}\to\mu^{+}\mu^{-}$ to be $O(10^{-8})$, which is close to the current upper limit of $4.7\times 10^{-8}$ CDFBsmumu . Figure 1: (a)[(b)] ${\cal B}(B_{s}\to\mu^{+}\mu^{-})$ versus $\Delta m_{s}$ [${\cal B}(B_{d}\to K^{0}\ell^{+}\ell^{-})$]. With the constrained values of $\lambda_{23}$ and $\theta^{\rm NP}_{s}$, in Fig. 2(a) we show the allowed $\phi_{s}$ given by Eq. (6). It is wroth mentioning that ${\cal B}(B_{s}\to K^{0}\ell^{+}\ell^{-})$ excludes $\phi_{s}$ larger than 0.2 rad. Furthermore, we present the contributions of the new CP violating source to the T-odd observable of Eq. (12) in Fig. 2(b). Intriguingly, the new phase can lead to a large statistical significance of the T-odd observable in $B_{q}\to V_{q}\ell^{+}\ell^{-}$. In summary, motivated by the large CP phase found by CDF and D$\O$ in the $B_{s}-\bar{B}_{s}$ mixing, we have investigated the $SU(2)_{L}$ singlet VQM. This model can not only provide the large phase through the Z-mediated FCNCs at tree level, but also strongly relate $\Delta m_{s}$, $B_{q}\to V_{q}\ell^{+}\ell^{-}$ and $B_{s}\to\mu^{+}\mu^{-}$ processes. In particular, we have shown that the new CP phase can be unambiguously exposed by the large statistical significances of the T-violating observables in $B_{q}\to V_{q}\ell^{+}\ell^{-}$ ($V_{q}=K^{},\,\phi)$. In addition, we have found that ${\cal B}(B_{s}\to\mu^{+}\mu^{-})$ can be enhanced as large as $O(10^{-8})$. Finally, we remark that the T-violating effects in $B_{q}\to V_{q}\ell^{+}\ell^{-}$ as well as the result on ${\cal B}(B_{s}\to\mu^{+}\mu^{-})$ are accessible at future super-B factories, such as the SuperKEKB SuperKEKB and LHCb LHCb . For example, 4400 events/year for $B\to K^{}\ell^{+}\ell^{-}$ decays will be produced at the LHCb, corresponding to the accuracy of the T-odd observable being around percent level. Figure 2: (a) Correlation of $\phi_{s}=2\beta_{s}-\phi^{\rm NP}_{s}$ and $\Delta m_{s}$. (b) Statistical significance $\varepsilon_{T}$ of $B_{d}\to K^{0}\mu^{+}\mu^{-}$ as a function of $q^{2}$. ###### Acknowledgements. This work is supported in part by the National Science Council of R.O.C. under Grant #s: NSC 97-2112-M-006 -001-MY3 and NSC-95-2112-M-007-059-MY3. ## References * (1) J. H. Christenson et al., Rev. Lett. 13, 138 (1964). * (2) E. Barberio et al., arXiv:0704.3575 [hep-ex]; online update at http://www.slac.stanford.edu/xorg/hfag. * (3) N. Cabibbo, Phys. Rev. Lett. 10, 531 (1963); M. Kobayashi and T. Maskawa, Prog. Theor. Phys. 49, 652 (1973). * (4) A. Abulencia et al. (CDF Collaboration), Phys. Rev. Lett. 97, 242003 (2006). * (5) D$\O$ note 5474-conf at http://www-d0.fnal.gov/Run2Physics/WWW/results/prelim/B/B51. * (6) M. Morello (CDF Collaboration), Phys. Conf. Ser. 110, 052040 (2008). * (7) M. Bona et al., arXive:0803.0659 [hep-ph]. * (8) Particle Data Group, W.M. Yao et al., J. Phys. G: Nucl. Part. Phys. 33, 1 (2006). * (9) L. Wolfenstein, Phys. Rev. Lett. 51, 1945 (1983). * (10) A. J. Buras, arXiv:hep-ph/0505175. * (11) T. Aaltonen et al. ( CDF Collaboration), Phys. Rev. Lett. 100, 161802 (2008). * (12) V. M. Abazov et al. (D$\O$ Collaboration), arXiv:0802.2255 [hep-ex]. * (13) B. Dutta and Y. Mimura, arXiv:0805.2988 [hep-ph]; F. J. Botella, G. C. Branco and M. Nebot, arXiv:0805.3995 [hep-ph], M. Blanke et al., arXiv:0805.4393 [hep-ph]; A. Soni et al., arXiv:0807.1971 [hep-ph]. * (14) G. Belanger and C. Q. Geng, Phys. Rev. D44, 2789 (1991). C. H. Chen, C. Q. Geng and C. C. Lih, Phys. Rev. D56, 6856 (1997). * (15) C. H. Chen and C. Q. Geng, Phys. Rev. D64, 074001 (2001). * (16) C. H. Chen and C. Q. Geng, Nucl. Phys. B636, 338 (2002); Phys. Rev. D66, 094018 (2002). * (17) C. H. Chen and C. Q. Geng, Phys. Rev. D66, 014007 (2002). * (18) G. Buchalla, A.J. Buras and M.E. Lautenbacher, Rev. Mod. Phys. 68, 1125 (1996). * (19) C. Bobeth, G. Hiller and G. Piranishvili, arXiv:0805.2525 [hep-ph]; U. Egede et al., arXiv:0807.2589 [hep-ph]. * (20) R. S. Chivukula and H. Georgi, Phys. Lett. B 188, 99 (1987); L. J. Hall and L. Randall, Phys. Rev. Lett. 65, 2939 (1990); M. Ciuchini et al., Nucl. Phys. B534, 3 (1998); G. D’Ambrosio et al., ibid. B645, 155 (2002). * (21) G. Isidori et al., JHEP 08, 064 (2006); G. Colangelo, E. Nikolidakis and C. Smith, arXiv:0807.0801 [hep-ph]. * (22) J.L. Hewett and T.G. Rizzo, Phys. Rept. 183, 193 (1989). * (23) T. G. Rizzo, Phys. Rev. D33, 3329 (1986); G. C. Branco and L. Lavoura, Nucl. Phys. B278, 738 (1986); P. Langacker and D. London, Phys. Rev. D38, 886 (1988); Y. Nir and D. Silverman, Phys. Rev. D42, 1477 (1990). * (24) J. A. Aguilar-Saavedra, Phys. Rev. D67, 035003 (2003); J. A. Aguilar-Saavedra et al., Nucl. Phys. B706, 204 (2005). * (25) A. Lenz and U. Nierste, JHEP 06, 072 (2007); A. Lenz, arXiv:0802.0977 [hep-ph]. * (26) C. H. Chen and C. Q. Geng, Mod. Phys. Lett. A21, 1137 (2006). * (27) T. Aaltonen et al. (CDF Collabpration), Phys. Rev. Lett. 100, 101802 (2008); T. Kuhr, arXiv:0804.2743 [hep-ex]. * (28) A. G. Akeroyd et al., arXiv:hep-ex/0406071. * (29) M. Calvi, arXiv:hep-ex/0506046.	arxiv-papers	2008-08-01T14:34:35	2024-09-04T02:48:57.106620	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Chuan-Hung Chen, Chao-Qiang Geng, Lin Li", "submitter": "Chuan Hung Chen", "url": "https://arxiv.org/abs/0808.0127" }
0808.0146	# $H^{1}$ and $BMO$ for certain nondoubling metric measure spaces Andrea Carbonaro, Giancarlo Mauceri and Stefano Meda A. Carbonaro, G. Mauceri: Dipartimento di Matematica Università di Genova via Dodecaneso 35, 16146 Genova Italia E-mail addresses: carbonaro@dima.unige.it mauceri@dima.unige.it S. Meda: Dipartimento di Matematica e Applicazioni Università di Milano-Bicocca via R. Cozzi 53 20125 Milano Italy E-mail address: stefano.meda@unimib.it ###### Abstract. Suppose that $(M,\rho,\mu)$ is a metric measure space, which possesses two “geometric” properties, called “isoperimetric” property and approximate midpoint property, and that the measure $\mu$ is locally doubling. The isoperimetric property implies that the volume of balls grows at least exponentially with the radius. Hence the measure $\mu$ is not globally doubling. In this paper we define an atomic Hardy space $H^{1}(\mu)$, where atoms are supported only on “small balls”, and a corresponding space $BMO(\mu)$ of functions of “bounded mean oscillation”, where the control is only on the oscillation over small balls. We prove that $BMO(\mu)$ is the dual of $H^{1}(\mu)$ and that an inequality of John–Nirenberg type on small balls holds for functions in $BMO(\mu)$. Furthermore, we show that the $L^{p}(\mu)$ spaces are intermediate spaces between $H^{1}(\mu)$ and $BMO(\mu)$, and we develop a theory of singular integral operators acting on function spaces on $M$. Finally, we show that our theory is strong enough to give $H^{1}(\mu)$-$L^{1}(\mu)$ and $L^{\infty}(\mu)$-$BMO(\mu)$ estimates for various interesting operators on Riemannian manifolds and symmetric spaces which are unbounded on $L^{1}(\mu)$ and on $L^{\infty}(\mu)$. ###### Key words and phrases: Singular integrals, $BMO$, atomic Hardy space, Riesz transform, metric measure spaces, multipliers. Work partially supported by the Progetto Cofinanziato “Analisi Armonica”. ## 1\. Introduction Suppose that $(M,\rho,\mu)$ is a metric measure space. Assume temporarily that $\mu$ is a doubling measure; then $(M,\rho,\mu)$ is a space of homogeneous type in the sense of Coifman and Weiss. Harmonic analysis on spaces of homogeneous type has been the object of many investigations. In particular, the atomic Hardy space $H^{1}(\mu)$ and the space $BMO(\mu)$ of functions of bounded mean oscillation have been defined and studied in this setting. We briefly recall their definitions. An atom $a$ is a function in $L^{1}(\mu)$ supported in a ball $B$ which satisfies appropriate “size” and cancellation condition. Then $H^{1}(\mu)$ is the space of all functions in $L^{1}(\mu)$ that admit a decomposition of the form $\sum_{j}\lambda_{j}\,a_{j}$, where the $a_{j}$’s are atoms and the sequence of complex numbers $\\{\lambda_{j}\\}$ is summable. A locally integrable function $f$ is in $BMO(\mu)$ if $\sup_{B}\mbox{\small$\displaystyle\frac{1}{\mu(B)}$}\int_{B}\left\|{f-f_{B}}\right\|\,\text{\rm d}\mu<\infty,$ where the supremum is taken over _all_ balls $B$, and $f_{B}$ denotes the average of $f$ over $B$. These spaces enjoy many of the properties of their Euclidean counterparts. In particular, the topological dual of $H^{1}({\mu})$ is isomorphic to $BMO({\mu})$, an inequality of John–Nirenberg type holds for functions in $BMO(\mu)$, the spaces $L^{p}(\mu)$ are intermediate spaces between $H^{1}(\mu)$ and $BMO(\mu)$ for the real and the complex interpolation methods. Furthermore, some important operators, which are bounded on $L^{p}(\mu)$ for all $p$ in $(1,\infty)$, but otherwise unbounded on $L^{1}(\mu)$ and on $L^{\infty}(\mu)$, turn out to be bounded from $H^{1}({\mu})$ to $L^{1}(\mu)$ and from $L^{\infty}(\mu)$ to $BMO({\mu})$. We remark that the doubling property is key in establishing these results. There is a huge literature on this subject: we refer the reader to [CW, St2] and the references therein for further information. There are interesting cases where $\mu$ is not doubling; then $\mu$ may or may not be locally doubling. An important case in which $\mu$ is not even locally doubling is that of nondoubling measures of polynomial growth treated, for instance, in [NTV, To, V], where new spaces $H^{1}$ and $BMO$ are defined, and a rich theory is developed (see also [MMNO] for more general measures on $\mathbb{R}^{n}$). We also mention recent works of X.T. Duong and L. Yan [DY1, DY2], who define an Hardy space $H^{1}$ and a space $BMO$ of bounded mean oscillation associated to a given operator satisfying suitable estimates. This is done in metric measure spaces with the doubling property, but it is a remarkable fact that the theory works also for “bad domains” in the ambient space, to which the restriction of the measure $\mu$ may be nondoubling. In this paper we consider the case where $\mu(M)=\infty$, and $\mu$ is a nondoubling locally doubling measure. By this we roughly mean that for every $R$ in $\mathbb{R}^{+}$ balls of radius at most $R$ satisfy a doubling condition, with doubling constant that may depend on $R$ (see (2.1) in Section 2 for the precise definition). Important examples of this situation are complete Riemannian manifolds with Ricci curvature bounded from below, a class which includes all Riemannian symmetric spaces of the noncompact type and Damek–Ricci spaces. In recent years, analysis on complete Riemannian manifolds satisfying the local doubling condition has been the object of many investigations. For instance, see [Sa] and the references therein for the equivalence between a scale-invariant parabolic Harnack inequality and a scaled Poincaré inequality, and [CD, ACDH, Ru] for recents results on the boundedness of Riesz transforms on such manifolds. Our approach to the case of locally doubling measures is inspired by a result of A.D. Ionescu [I] on rank one symmetric spaces of the noncompact type and by a recent paper of the second and third named authors concerning the analysis of the Ornstein–Uhlenbeck operator [MM]. For each “scale” $b$ in $\mathbb{R}^{+}$, we define spaces $H_{b}^{1}(\mu)$ and $BMO_{b}(\mu)$ much as in the case of spaces of homogeneous type, the only difference being that we require that the balls involved have at most radius $b$. So, for instance, an $H_{b}^{1}(\mu)$ atom is an atom supported in a ball of radius at most $b$. We remark that in the case where $M$ is a symmetric space of the noncompact type and real rank one, the space $BMO_{1}(\mu)$ agrees with the space defined by Ionescu. Ionescu also proved that if $p$ is in $(1,2)$, then $L^{p}(\mu)$ is an interpolation space between $L^{2}(\mu)$ and $BMO_{1}(\mu)$ for the complex method of interpolation. In the case where $M$ is a complete noncompact Riemannian manifold with locally doubling Riemannian measure and satisfying certain additional assumptions E. Russ [Ru] defined an Hardy space that agrees with the space $H_{1}^{1}(\mu)$ defined above, but he did not investigate its structural properties. We prove that under a mild “geometric” assumption, which we call property (AMP) (see Section 2), $H_{b}^{1}(\mu)$ and $BMO_{b}(\mu)$, in fact, do not depend on the parameter $b$ provided that $b$ is large enough (see Section 4). Furthermore, we show that $BMO_{b}(\mu)$ is isomorphic to the topological dual of $H_{b}^{1}(\mu)$ (see Section 6), and that functions in $BMO_{b}(\mu)$ satisfy an inequality of John–Nirenberg type (see Section 5). As far as interpolation is concerned, there is no reason to believe that in this generality $L^{p}(\mu)$ spaces with $p$ in $(1,\infty)$ are interpolation spaces between $H_{b}^{1}(\mu)$ and $BMO_{b}(\mu)$. However, this is true under a simple geometric assumption on $M$, called property (I). Roughly speaking, $M$ possesses property (I) if a fixed ratio of the measure of any bounded open set is concentrated near its boundary. If $M$ possesses property (I), then a basic relative distributional inequality for the local sharp function and the local Hardy–Littlewood maximal function holds. We prove this in Section 7, by adapting to our setting some ideas of Ionescu [I]. We remark that our approach, which makes use of the dyadic cubes of M. Christ and G. David [Ch, Da], simplifies considerably the original proof in [I]. As a consequence of the relative distributional inequality we prove an interpolation result for analytic families of operators, analogous to that proved by C. Fefferman and E.M. Stein [FS] in the classical setting. An interesting application of the aforementioned interpolation result is to singular integral operators (Theorem 8.2). We prove that if $\mathcal{T}$ is a bounded self adjoint operator on $L^{2}(\mu)$ and its kernel $k$ is a locally integrable function off the diagonal in $M\times M$ and satisfies a _local Hörmander type condition_ (i.e. if $\nu_{k}<\infty$ and $\upsilon_{k}<\infty$, where $\nu_{k}$ and $\upsilon_{k}$ are defined in the statement of Theorem 8.2), then $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $(1,\infty)$, from $H^{1}({\mu})$ to $L^{1}(\mu)$ and from $L^{\infty}(\mu)$ to $BMO({\mu})$. Applications of this results to multipliers for the spherical Fourier transform on Riemannian symmetric spaces of the noncompact type, to spectral operators of the Laplace–Beltrami operator and to local Riesz transforms on certain noncompact Riemannian manifolds are given in Section 10. Our results complement earlier results of J.Ph. Anker [A], M. Taylor [Ta] and Russ [Ru]. Of course, there are many other interesting operators on measured metric spaces with properties (AMP), (LDP) and (I), which are unbounded on $L^{1}(\mu)$ and on $L^{\infty}(\mu)$, but satisfy $H^{1}(\mu)$-$L^{1}(\mu)$ and $L^{\infty}(\mu)$-$BMO(\mu)$ estimates. Some of these will be considered in a forthcoming paper. It is interesting to speculate about the range of applicability of the theory we develop. In particular, a natural problem is to find conditions (possibly easy to verify) under which a complete Riemannian manifold possesses all the three properties, local doubling, (I), and (AMP), needed to prove the results of Sections 2-8. This problem is considered in Section 9. Suppose that $M$ is a complete Riemannian manifold with Riemannian distance $\rho$ and Riemannian density $\mu$. A known fact, which is a straightforward consequence of the Bishop–Gromov comparison Theorem, is that if $M$ has Ricci curvature bounded from below, then $(M,\rho,\mu)$ is locally doubling. Furthermore, since $\rho$ is a length distance, $(M,\rho,\mu)$ has property (AMP). We shall prove that $M$ possesses property (I) if and only if the Cheeger isoperimetric constant $h(M)$ (see (9.1) for the definition) is strictly positive. As a consequence we shall prove that if $M$ has Ricci curvature bounded from below, then $M$ possesses property (I) if and only if the bottom $b(M)$ of the spectrum of $M$ is strictly positive. Similar results on graphs with bounded geometry will appear elsewhere. Finally, it would be interesting to consider the case where $\mu(M)<\infty$. To keep the length of this paper reasonable, we shall postpone the detailed study of the case where $\mu(M)<\infty$ to a forthcoming paper [CMM]. ## 2\. Geometric assumptions Suppose that $(M,\rho,\mu)$ is a metric measure space, and denote by $\mathcal{B}$ the family of all balls on $M$. We assume that $\mu(M)>0$ and that every ball has finite measure. For each $B$ in $\mathcal{B}$ we denote by $c_{B}$ and $r_{B}$ the centre and the radius of $B$ respectively. Furthermore, we denote by $\kappa\,B$ the ball with centre $c_{B}$ and radius $\kappa\,r_{B}$. For each $b$ in $\mathbb{R}^{+}$, we denote by $\mathcal{B}_{b}$ the family of all balls $B$ in $\mathcal{B}$ such that $r_{B}\leq b$. For any subset $A$ of $M$ and each $\kappa$ in $\mathbb{R}^{+}$ we denote by $A_{\kappa}$ and $A^{\kappa}$ the sets $\bigl{\\{}x\in A:\rho(x,A^{c})\leq\kappa\bigr{\\}}\qquad\hbox{and}\qquad\bigl{\\{}x\in A:\rho(x,A^{c})>\kappa\bigr{\\}}$ respectively. In Sections 2-8 we assume that $M$ is _unbounded_ and possesses the following properties: 1. (i) _local doubling property_ (LDP): for every $b$ in $\mathbb{R}^{+}$ there exists a constant $D_{b}$ such that (2.1) $\mu\bigl{(}2B\bigr{)}\leq D_{b}\,\mu\bigl{(}B\bigr{)}\qquad\forall B\in\mathcal{B}_{b}.$ This property is often called _local doubling condition_ in the literature, and we adhere to this terminology. Note that if (2.1) holds and $M$ is bounded, then $\mu$ is doubling. 2. (ii) _isoperimetric property_ (I): there exist $\kappa_{0}$ and $C$ in $\mathbb{R}^{+}$ such that for every bounded open set $A$ (2.2) $\mu\bigl{(}A_{\kappa}\bigr{)}\geq C\,\kappa\,\mu(A)\qquad\forall\kappa\in(0,\kappa_{0}].$ Suppose that $M$ has property (I). For each $t$ in $(0,\kappa_{0}]$ we denote by $C_{t}$ the supremum over all constants $C$ for which (2.2) holds for all $\kappa$ in $(0,t]$. Then we define $I_{M}$ by $I_{M}=\sup\bigl{\\{}C_{t}:t\in(0,\kappa_{0}]\bigr{\\}}.$ Note that the function $t\mapsto C_{t}$ is decreasing on $(0,\kappa_{0}]$, so that (2.3) $I_{M}=\lim_{t\to 0^{+}}C_{t};$ 3. (iii) _property_ (AMP) (approximate midpoint property): there exist $R_{0}$ in $[0,\infty)$ and $\beta$ in $(1/2,1)$ such that for every pair of points $x$ and $y$ in $M$ with $\rho(x,y)>R_{0}$ there exists a point $z$ in $M$ such that $\rho(x,z)<\beta\,\rho(z,y)$ and $\rho(x,y)<\beta\,\rho(x,y)$. This is clearly equivalent to the requirement that there exists a ball $B$ containing $x$ and $y$ such that $r_{B}<\beta\,\rho(x,y)$. ###### Remark 2.1. Observe that the isoperimetric property (I) implies that for every open set $A$ of _finite measure_ $\mu\bigl{(}A_{\kappa}\bigr{)}\geq C\,\kappa\,\mu(A)\qquad\forall\kappa\in(0,\kappa_{0}],$ where $\kappa_{0}$ and $C$ are as in (2.2). Indeed, suppose that $A$ is an open set of finite measure. Fix a reference point $o$ in $M$ and denote by $B(o,j)$ the ball with centre $o$ and radius $j$, and by $A(j)$ the set $A\cap B(o,j)$. For each $\kappa$ in $(0,\kappa_{0}]$ denote by $A_{j,\kappa}$ the set $\bigl{\\{}x\in A(j):\rho\bigl{(}x,B(o,j)^{c}\bigr{)}\leq\kappa,\rho(x,A^{c})>\kappa\bigr{\\}}.$ First we prove that (2.4) $\lim_{j\to\infty}\mu\bigl{(}A_{j,\kappa}\bigr{)}=0.$ Since $\mu(A)<\infty$, for each $\epsilon>0$ there exists $J$ such that $\mu\bigl{(}A\cap B(o,J)^{c}\bigr{)}<\epsilon.$ Now, if $j>J+\kappa$ and $x$ is in $A_{j,\kappa}$, then $x$ belongs also to $A\cap B(o,J)^{c}$, whence $\mu(A_{j,\kappa})<\epsilon$ for all $j\geq J$, as required. Observe that $A_{j,\kappa}$ is contained in $A(j)_{\kappa}$ and that $A_{j,\kappa}=A(j)_{\kappa}\setminus\bigl{(}B(o,j)\cap A_{\kappa}\bigr{)}$. Therefore (2.5) $\mu\bigl{(}B(o,j)\cap A_{\kappa}\bigr{)}=\mu\bigl{(}A(j)_{\kappa}\bigr{)}-\mu\bigl{(}A_{j,\kappa}\bigr{)}.$ Since $\mu(A_{\kappa})=\lim_{j\to\infty}\mu\bigl{(}B(o,j)\cap A_{\kappa}\bigr{)}$, $\mu(A_{\kappa})=\lim_{j\to\infty}\mu\bigl{(}A(j)_{\kappa}\bigr{)}$ by (2.5) and (2.4). Since $A(j)$ is a bounded open set, we may conclude that $\displaystyle\mu(A_{\kappa})$ $\displaystyle\geq\lim_{j\to\infty}C\,\kappa\,\mu\bigl{(}A(j)\bigr{)}$ $\displaystyle=C\,\kappa\,\mu\bigl{(}A\bigr{)},$ as required. ###### Remark 2.2. The local doubling property is needed for all the results in this paper, but many results in Sections 2-8 depend only on some but not all the properties (i)-(iii). In particular, Theorem 3.2 and Proposition 3.4 require only the local doubling property, Propositions 3.1 and 3.5, Lemma 7.2 and Theorem 7.3, which are key in proving the interpolation result Theorem 7.4, require property (I), but not property (AMP), all the results in Sections 4, 5 and 6 require property (AMP) but not property (I). In particular, property (AMP) is key to prove the scale invariance of the spaces $H^{1}(\mu)$ and $BMO(\mu)$ defined below (Proposition 4.3). Finally, all the properties (i)-(iii) above are needed for the interpolation results in Section 7 and for the results in Section 8. ###### Remark 2.3. The local doubling property implies that for each $\tau\geq 2$ and for each $b$ in $\mathbb{R}^{+}$ there exists a constant $C$ such that (2.6) $\mu\bigl{(}B^{\prime}\bigr{)}\leq C\,\mu(B)$ for each pair of balls $B$ and $B^{\prime}$, with $B\subset B^{\prime}$, $B$ in $\mathcal{B}_{b}$, and $r_{B^{\prime}}\leq\tau\,r_{B}$. We shall denote by $D_{\tau,b}$ the smallest constant for which (2.6) holds. In particular, if (2.6) holds (with the same constant) for all balls $B$ in $\mathcal{B}$, then $\mu$ is doubling and we shall denote by $D_{\tau,\infty}$ the smallest constant for which (2.6) holds. ###### Remark 2.4. There are various “structural constants” which enter the proofs of our results. We have made the choice to keep track of these constants, which often appear explicitly in the statements. For the reader’s convenience we give here a list of all the relevant constants used is Sections 2-8: $\begin{array}[]{ll}D_{\tau,b}&\qquad\hbox{$\tau\geq 2$, $b\in\mathbb{R}^{+}\cup\\{\infty\\}$ (see Remark~{}\ref{r: geom I})}\\\ I_{M}&\qquad\hbox{the isoperimetric constant (see \hbox{\rm(ii)}\ above)}\\\ \hbox{$R_{0}$ and $\beta$}&\qquad\hbox{appear in the (AMP) property (see \hbox{\rm(iii)}\ above)}\\\ \hbox{$\delta$, $C_{1}$ and $a_{0}$}&\qquad\hbox{appear in the construction of dyadic cubes (see Thm~{}\ref{t: dyadic cubes})}.\end{array}$ ## 3\. Preliminary results Roughly speaking, if $M$ has property (I), then a fixed ratio of the measure of any bounded open set is concentrated near its boundary. The following proposition contains a quantitative version of this statement. ###### Proposition 3.1. The following hold: 1. (i) the volume growth of $M$ is at least exponential; 2. (ii) for every bounded open set $A$ $\mu(A_{t})\geq\bigl{(}1-\mathrm{e}^{-I_{M}t}\bigr{)}\,\mu(A)\qquad\forall t\in\mathbb{R}^{+}.$ ###### Proof. First we prove (i). Denote by $o$ a reference point in $M$. For every $r>0$ denote by $V_{r}$ the measure of the ball with centre $o$ and radius $r$. It is straightforward to check that $B(o,r)_{\kappa}\subset B(o,r)\setminus B(o,r-\kappa)$, so that for all sufficiently large $r$ $\displaystyle V_{r}-V_{r-\kappa}$ $\displaystyle\geq\mu\bigl{(}B(o,r)_{\kappa}\bigr{)}$ $\displaystyle\geq C\,\kappa\,V_{r}\qquad\forall\kappa\in(0,\kappa_{0}]$ by property (I) ($C$ and $\kappa_{0}$ are as in (2.2)). Hence $\displaystyle V_{r}$ $\displaystyle\geq C\,\kappa\,V_{r}+V_{r-\kappa}$ $\displaystyle\geq\eta\,V_{r-\kappa}\qquad\forall\kappa\in(0,\kappa_{0}],$ where $\eta=C\kappa+1$. Denote by $n$ the positive integer for which $r-n\kappa$ is in $[\kappa,2\kappa)$. Then $\displaystyle V_{r}$ $\displaystyle\geq\eta^{n}\,V_{r-n\kappa}$ $\displaystyle\geq\eta^{r/\kappa-2}\,V_{\kappa},$ as required. Now we prove (ii). Suppose that $A$ is a bounded open subset of $M$, and that $t$ is in $(0,\kappa_{0}]$. Note that $A^{s}$ is a bounded open subset of $M$. It is straightforward to check that $(A^{s})_{t}\subset A^{s}\setminus A^{s+t}$. Therefore $\displaystyle\mu(A^{s+t})-\mu(A^{s})$ $\displaystyle\leq-\mu\bigl{(}(A^{s})_{t}\bigr{)}$ $\displaystyle\leq- C_{t}\,t\,\mu(A^{s})$ by property (I) (see (2.2) above). Since $s\mapsto\mu(A^{s})$ is monotonic, it is differentiable almost everywhere. The inequality above and (2.3) imply that for almost every $s$ in $\mathbb{R}^{+}$ $\frac{\textrm{d}}{\textrm{d}s}\mu(A^{s})\leq-I_{M}\,\mu(A^{s}).$ Notice also that $\lim_{s\to 0^{+}}\mu(A^{s})=\mu(A)$. Therefore $\mu(A^{s})\leq\mathrm{e}^{-I_{M}\,s}\,\mu(A)\qquad\forall s\in\mathbb{R}^{+},$ and finally $\mu(A_{s})\geq\bigl{(}1-\mathrm{e}^{-I_{M}\,s}\bigr{)}\,\mu(A)\qquad\forall s\in\mathbb{R}^{+},$ as required. ∎ We shall make use of the analogues in our setting of the so-called dyadic cubes $Q_{\alpha}^{k}$ introduced by G. David and M. Christ [Da, Ch] on spaces of homogeneous type. It may help to think of $Q_{\alpha}^{k}$ as being essentially a cube of diameter $\delta^{k}$ with centre $z_{\alpha}^{k}$. ###### Theorem 3.2. There exist a collection of open subsets $\\{Q_{\alpha}^{k}:k\in\mathbb{Z},\alpha\in I_{k}\\}$ and constants $\delta$ in $(0,1)$, $a_{0}$, $C_{1}$ in $\mathbb{R}^{+}$ such that 1. (i) $\bigcup_{\alpha}Q_{\alpha}^{k}$ is a set of full measure in $M$ for each $k$ in $\mathbb{Z}$; 2. (ii) if $\ell\geq k$, then either $Q_{\beta}^{\ell}\subset Q_{\alpha}^{k}$ or $Q_{\beta}^{\ell}\cap Q_{\alpha}^{k}=\emptyset$; 3. (iii) for each $(k,\alpha)$ and each $\ell<k$ there is a unique $\beta$ such that $Q_{\alpha}^{k}\subset Q_{\beta}^{\ell}$; 4. (iv) $\mathrm{diam}(Q_{\alpha}^{k})\leq C_{1}\,\delta^{k}$; 5. (v) $Q_{\alpha}^{k}$ contains the ball $B(z_{\alpha}^{k},a_{0}\,\delta^{k})$. ###### Proof. The proof of (ii)–(v) is as in [Ch]. In fact, the proof depends only on the metric structure of the space and not on the properties of the measure $\mu$ and is even easier in our case, because $\rho$ is a genuine distance, rather than a quasi-distance. The proof of (i) is again as in [Ch]; observe that only a local doubling property is used in the proof. ∎ Note that (iv) and (v) imply that for every integer $k$ and each $\alpha$ in $I_{k}$ $B(z_{\alpha}^{k},a_{0}\,\delta^{k})\subset Q_{\alpha}^{k}\subset B(z_{\alpha}^{k},C_{1}\,\delta^{k}).$ ###### Remark 3.3. When we use dyadic cubes, we implicitly assume that for each $k$ in $\mathbb{Z}$ the set $M\setminus\bigcup_{\alpha\in I_{k}}Q_{\alpha}^{k}$ has been permanently deleted from the space. We shall denote by $\mathcal{Q}^{k}$ the class of all dyadic cubes of “resolution” $k$, i.e., the family of cubes $\\{Q_{\alpha}^{k}:\alpha\in I_{k}\\}$, and by $\mathcal{Q}$ the set of all dyadic cubes. We shall need the following additional properties of dyadic cubes. ###### Proposition 3.4. Suppose that $b$ is in $\mathbb{R}^{+}$ and that $\nu$ is in $\mathbb{Z}$, and let $C_{1}$ and $\delta$ are as in Theorem 3.2. The following hold: 1. (i) suppose that $Q$ is in $\mathcal{Q}^{k}$ for some $k\geq\nu$, and that $B$ is a ball such that $c_{B}\in Q$. If $r_{B}\geq C_{1}\,\delta^{k}$, then (3.1) $\mu(B\cap Q)=\mu(Q);$ if $r_{B}<C_{1}\,\delta^{k}$, then (3.2) $\mu(B\cap Q)\geq D_{C_{1}/(a_{0}\delta),\delta^{\nu}}^{-1}\,\mu(B);$ 2. (ii) suppose that $\tau$ is in $[2,\infty)$. For each $Q$ in $\mathcal{Q}$ the space $\bigl{(}Q,\rho_{\|Q},\mu_{\|Q}\bigr{)}$ is of homogeneous type. Denote by $D_{\tau,\infty}^{Q}$ its doubling constant (see Remark 2.3 for the definition). Then $\sup\,\Bigl{\\{}D_{\tau,\infty}^{Q}:Q\in\bigcup_{k=\nu}^{\infty}\mathcal{Q}^{k}\Bigr{\\}}\leq D_{\tau,C_{1}\delta^{\nu}}\,D_{C_{1}/(a_{0}\delta),\delta^{\nu}};$ 3. (iii) for each ball $B$ in $\mathcal{B}_{b}$, let $k$ be the integer such that $\delta^{k}\leq r_{B}<\delta^{k-1}$, and and let $\widetilde{B}$ denote the ball with centre $c_{B}$ and radius $\bigl{(}1+C_{1}\bigr{)}\,r_{B}$. Then $\widetilde{B}$ contains all dyadic cubes in $\mathcal{Q}^{k}$ that intersect $B$ and $\mu(\widetilde{B})\leq D_{1+C_{1},b}\,\mu(B);$ 4. (iv) suppose that $B$ is in $\mathcal{B}_{b}$, and that $k$ is an integer such that $\delta^{k}\leq r_{B}<\delta^{k-1}$. Then there are at most $D_{(1+C_{1})/(a_{0}\delta),b}$ dyadic cubes in $\mathcal{Q}^{k}$ that intersect $B$. ###### Proof. First we prove (i). Our proof is a version of the proof given by Christ [Ch, p. 613] that keeps track of the various structural constants involved. First we prove (3.1). By Theorem 3.2 (iv) the diameter of $Q$ is at most $C_{1}\,\delta^{k}$, so that $Q\subset B$, whence $B\cap Q=Q$, and the required formula is obvious. To prove (3.2), denote by $j$ the unique integer such that $\delta^{j}<\frac{r_{B}}{C_{1}}\leq\delta^{j-1}$ and by $Q_{\beta}^{j}$ the unique dyadic cube of resolution $j$ that contains $c_{B}$. Then $j\geq k$, because $C_{1}\,\delta^{j}<r_{B}\leq C_{1}\delta^{k}$. The cubes $Q_{\beta}^{j}$ and $Q$ have nonempty intersection, because, they both contain $c_{B}$. Thus $Q_{\beta}^{j}\subset Q$. By Theorem 3.2 (iv) the diameter of $Q_{\beta}^{j}$ is at most $C_{1}\,\delta^{j}$, which is $<r_{B}$ by the definition of $j$, so that $Q_{\beta}^{j}\subset B$. Therefore $Q_{\beta}^{j}\subset B\cap Q$, and $\displaystyle\mu(B\cap Q)$ $\displaystyle\geq\mu\bigl{(}Q_{\beta}^{j}\bigr{)}$ $\displaystyle\geq\mu\bigl{(}B(z_{\beta}^{j},a_{0}\,\delta^{j})\bigr{)}.$ Observe that $\mbox{\small$\displaystyle\frac{r_{B}}{a_{0}\,\delta^{j}}$}\leq\mbox{\small$\displaystyle\frac{C_{1}}{a_{0}\,\delta}$}$ and that $B(z_{\beta}^{j},a_{0}\,\delta^{j})\subset B$. Hence $\mu\bigl{(}B(z_{\beta}^{j},a_{0}\,\delta^{j})\bigr{)}\geq D_{C_{1}/(a_{0}\delta),\delta^{\nu}}^{-1}\,\mu(B),$ as required to conclude the proof of (3.2), and of (i). Next we prove (ii). Suppose that $Q$ is a dyadic cube in $\mathcal{Q}^{k}$, with $k\geq\nu$. Suppose that $B$ and $B^{\prime}$ are balls in $\mathcal{B}$ with $B\subset B^{\prime}$ such that $c_{B}$ and $c_{B^{\prime}}$ belong to $Q$ and $r_{B^{\prime}}\leq\tau\,r_{B}$. We treat the cases where $C_{1}\,\delta^{k}\leq r_{B}$ and $r_{B}<C_{1}\,\delta^{k}$ separately. If $C_{1}\,\delta^{k}\leq r_{B}$, then $\mu(B^{\prime}\cap Q)=\mu(Q)=\mu(B\cap Q)$. If $r_{B}<C_{1}\,\delta^{k}$, then $\displaystyle\mu(B^{\prime}\cap Q)$ $\displaystyle\leq\mu(B^{\prime})$ $\displaystyle\leq D_{\tau,C_{1}\delta^{k}}\,\mu(B)$ $\displaystyle\leq D_{\tau,C_{1}\delta^{k}}\,D_{C_{1}/(a_{0}\delta),\delta^{\nu}}\,\mu(B\cap Q),$ by the local doubling property of $M$ and (3.2). Therefore $Q$ is a homogeneous space with doubling constant at most $D_{\tau,C_{1}\delta^{k}}\,D_{C_{1}/(a_{0}\delta),\delta^{\nu}}$. Since $k\geq\nu$, these doubling constants are dominated by $D_{\tau,C_{1}\delta^{\nu}}\,D_{C_{1}/(a_{0}\delta),\delta^{\nu}}$, as required. Now we prove (iii). Denote by $Q$ a cube in $\mathcal{Q}^{k}$ that intersects $B$. By the triangle inequality and Theorem 3.2 (iv), $Q$ is contained in $\widetilde{B}$. The required estimate of the measure of $\widetilde{B}$ follows from the local doubling condition (see Remark 2.3). Finally we prove (iv). Denote by $Q_{1},\ldots,Q_{N}$ the cubes in $\mathcal{Q}^{k}$ that intersect $B$. By (iii) each of these cubes is contained in $\widetilde{B}$. By Theorem 3.2 (v) each cube $Q_{j}$ contains a ball, $B_{j}^{\prime}$ say, of radius $a_{0}\,\delta^{k}$, and these balls are pairwise disjoint because they are contained in disjoint dyadic cubes. By the local doubling condition $\mu(\widetilde{B})\leq D_{(1+C_{1})/(a_{0}\delta),b}\,\mu(B_{j}^{\prime})$ for all $j$. Therefore $\displaystyle N\,\mu(\widetilde{B})$ $\displaystyle\leq D_{(1+C_{1})/(a_{0}\delta),b}\,\sum_{j=1}^{N}\mu(B_{j}^{\prime})$ $\displaystyle=D_{(1+C_{1})/(a_{0}\delta),b}\,\mu\Bigl{(}\bigcup_{j=1}^{N}B_{j}^{\prime}\Bigr{)}$ $\displaystyle\leq D_{(1+C_{1})/(a_{0}\delta),b}\,\mu(\widetilde{B}),$ from which the desired estimate follows. ∎ Our next result is a covering property enjoyed by spaces with property (I). It is key in proving relative distributional inequalities for the sharp maximal operator (see Lemma 7.2 below). ###### Proposition 3.5. Suppose that $\nu$ is an integer. For every $\kappa$ in $\mathbb{R}^{+}$, every open subset $A$ of $M$ of finite measure and every collection $\mathcal{C}$ of dyadic cubes of resolution at least $\nu$ such that $\bigcup_{Q\in\mathcal{C}}Q=A$, there exist mutually disjoint cubes $Q_{1},\ldots,Q_{k}$ in $\mathcal{C}$ such that 1. (i) $\sum_{j=1}^{k}\mu(Q_{j})\geq\bigl{(}1-\mathrm{e}^{-I_{M}\,\kappa}\bigr{)}\,\mu(A)/2$; 2. (ii) $\rho(Q_{j},A^{c})\leq\kappa$ for every $j$ in $\\{1,\ldots,k\\}$. ###### Proof. Denote by $\widetilde{\mathcal{C}}$ the subcollection of all cubes in $\mathcal{C}$ that intersect $A_{\kappa}$. Clearly the cubes in $\widetilde{\mathcal{C}}$ cover $A_{\kappa}$ and satisfy (ii). Next we prove (i). Since two dyadic cubes are either disjoint or contained one in the other, we may consider the sequence $\\{Q_{j}\\}$ of cubes in $\widetilde{\mathcal{C}}$ which are not contained in any other cube of $\widetilde{\mathcal{C}}$. The existence of these “maximal” cubes is guaranteed by the assumption that the resolution of the cubes in $\mathcal{C}$ is bounded from below. The cubes $\\{Q_{j}\\}$ are mutually disjoint and cover $A_{\kappa}$. Therefore (3.3) $\displaystyle\mu(A)$ $\displaystyle\geq\sum_{j}\mu(Q_{j})$ $\displaystyle\geq\mu\bigl{(}A_{\kappa}\bigr{)}$ $\displaystyle\geq\bigl{(}1-\mathrm{e}^{-I_{M}\,\kappa}\bigr{)}\,\mu(A),$ where the last inequality holds because $M$ possesses property (I). To conclude the proof of (i) take $k$ so large that $\sum_{j=1}^{k}\mu(Q_{j})\geq(1/2)\sum_{j=1}^{\infty}\mu(Q_{j})$. Then $\sum_{j=1}^{k}\mu(Q_{j})\geq\bigl{(}1-\mathrm{e}^{-I_{M}\,\kappa}\bigr{)}\,\mu(A)/2,$ as required. ∎ ## 4\. The Hardy space $H^{1}$ ###### Definition 4.1. Suppose that $r$ is in $(1,\infty]$. A $(1,r)$-_atom_ $a$ is a function in $L^{1}(\mu)$ supported in a ball $B$ in $\mathcal{B}$ with the following properties: 1. (i) ${\\|{a}\\|_{\infty}}\leq\mu(B)^{-1}$ if $r$ is equal to $\infty$ and $\Bigl{(}\frac{1}{\mu(B)}\int_{B}\left\|{a}\right\|^{r}\,\text{\rm d}{\mu}\Bigr{)}^{1/r}\leq\mu(B)^{-1}$ if $r$ is in $(1,\infty)$; 2. (ii) $\displaystyle\int_{B}a\,\text{\rm d}\mu=0$. ###### Definition 4.2. Suppose that $b$ is in $\mathbb{R}^{+}$. The _Hardy space_ $H_{b}^{1,r}({\mu})$ is the space of all functions $g$ in $L^{1}(\mu)$ that admit a decomposition of the form (4.1) $g=\sum_{k=1}^{\infty}\lambda_{k}\,a_{k},$ where $a_{k}$ is a $(1,r)$-atom _supported in a ball $B$ of $\mathcal{B}_{b}$_, and $\sum_{k=1}^{\infty}\left\|{\lambda_{k}}\right\|<\infty$. The norm ${\\|{g}\\|_{H_{b}^{1,r}({\mu})}}$ of $g$ is the infimum of $\sum_{k=1}^{\infty}\left\|{\lambda_{k}}\right\|$ over all decompositions (4.1) of $g$. Clearly a function in $H_{c}^{1,r}(\mu)$ is in $H_{b}^{1,r}(\mu)$ for all $c<b$. We shall prove in Proposition 4.3 below that, in fact, the reverse inclusion holds whenever $c$ is large enough. Hence for $b$ large the space $H_{b}^{1,r}(\mu)$ does not depend on the parameter $b$, and for each pair of sufficiently large parameters $b$ and $c$ the norms ${\\|{\cdot}\\|_{H_{b}^{1,r}(\mu)}}$ and ${\\|{\cdot}\\|_{H_{c}^{1,r}(\mu)}}$ are equivalent. There are cases where $H_{c}^{1,r}(\mu)$ and $H_{b}^{1,r}(\mu)$ are isomorphic spaces for each pair of parameters $c$ and $b$. This happens, for instance, if $M$ is the upper half plane, $\rho$ the Poincaré metric and $\mu$ the associated Riemannian measure. However, if $M$ is a homogeneous tree of degree $q\geq 1$, $\rho$ denotes the natural distance and $\mu$ the counting measure, it is straightforward to check that $H_{1}^{1}(\mu)$ consists of the null function only, whereas $H_{2}^{1}(\mu)$ is a much richer space. Recall that $R_{0}$ and $\beta$ are the constants which appear in the definition of the (AMP) property. ###### Proposition 4.3. Suppose that $r$ is in $(1,\infty]$, $b$ and $c$ are in $\mathbb{R}^{+}$ and satisfy $R_{0}/(1-\beta)<c<b$. The following hold: 1. (i) there exist a constant $C$ and a nonnegative integer $N$, depending only on $M$, $b$ and $c$, such that for each ball $B$ in $\mathcal{B}_{b}$ and each $(1,r)$-atom $a$ supported in $B$ there exist at most $N$ atoms $a_{1},\ldots,a_{N}$ with supports contained in balls $B_{1},\ldots,B_{N}$ in $\mathcal{B}_{c}$ and $N$ constants $\lambda_{1},\ldots,\lambda_{N}$ such that $\left\|{\lambda_{j}}\right\|\leq C$, $a=\sum_{j=1}^{N}\lambda_{j}\,a_{j}\qquad\hbox{and}\qquad{\\|{a}\\|_{H_{c}^{1,r}(\mu)}}\leq C\,N;$ 2. (ii) a function $f$ is in $H_{c}^{1,r}(\mu)$ if and only if $f$ is in $H_{b}^{1,r}(\mu)$. Furthermore, there exists a constant $C$ such that ${\\|{f}\\|_{H_{b}^{1,r}(\mu)}}\leq{\\|{f}\\|_{H_{c}^{1,r}(\mu)}}\leq C\,{\\|{f}\\|_{H_{b}^{1,r}(\mu)}}\qquad\forall f\in H_{c}^{1,r}(\mu).$ ###### Proof. Choose $\beta^{\prime}$ in $(0,1-\beta)$ such that $R_{0}/\beta^{\prime}<c$. First we prove (i). Suppose that $B$ is a ball in $\mathcal{B}_{b}$ and that $r_{B}>c$, for otherwise there is nothing to prove. Denote by $\\{z_{1},\ldots,z_{N_{1}}\\}$ a maximal set of points in $B$ such that $\rho(z_{j},z_{k})\geq\beta^{\prime}r_{B}$ for all $j\neq k$ and each point of $B$ is at distance at most $\beta^{\prime}r_{B}$ from the set $\\{z_{1},\ldots,z_{N_{1}}\\}$. Denote by $B_{j}$ the ball with centre $z_{j}$ and radius $\beta^{\prime}r_{B}$, and by $B_{0}$ the ball with centre $c_{B}$ and radius $\beta^{\prime}r_{B}$. Note that (4.2) $\mu(B)\leq D_{1/\beta^{\prime},b}\,\mu(B_{0}),$ where $D_{1/\beta^{\prime},b}$ is as in Remark 2.3. We consider the partition of unity $\\{\psi_{1},\ldots,\psi_{N_{1}}\\}$ of ${\mathbf{1}}_{\bigcup_{j}B_{j}}$ subordinated to the covering $\\{B_{1},\ldots,B_{N_{1}}\\}$ defined by $\psi_{j}={\mathbf{1}}_{B_{j}}/\sum^{N_{1}}_{k=1}{\mathbf{1}}_{B_{k}}.$ For each $j$ in ${\left\\{1,\dots,{N_{1}}\right\\}}$ we define $A_{j}=\frac{1}{\mu(B_{0})}\int_{M}a\,\psi_{j}\,\text{\rm d}\mu\qquad\hbox{and}\qquad\phi_{j}=a\,\psi_{j}-A_{j}\,{\mathbf{1}}_{B_{0}}.$ It is straighforward to check that $a=\sum_{j=1}^{N_{1}}\phi_{j}$ and each function $\phi_{j}$ has integral $0$ and is supported in $B_{j}\cup B_{0}$. Define $J^{\prime}$ and $J^{\prime\prime}$ by $J^{\prime}=\\{j:z_{j}\notin B_{0}\\}\qquad\hbox{and}\qquad J^{\prime\prime}=\\{j:z_{j}\in B_{0}\\}.$ If $j$ is in $J^{\prime}$, then $z_{j}$ is not in $B_{0}$, so that $\rho(c_{B},z_{j})>\beta^{\prime}r_{B}>\beta^{\prime}c>R_{0}$. Therefore we may use property (AMP) and conclude that there exists a ball $B_{j}^{\prime}$ containing $c_{B}$ and $z_{j}$ with radius $<\beta\,\rho(c_{B},z_{j})$. Denote by $\widetilde{B}_{j}$ the ball centred at $c_{B_{j}^{\prime}}$ with radius $r_{B_{j}^{\prime}}+\beta^{\prime}r_{B}$. By using the triangle inequality we see that $\widetilde{B}_{j}$ contains $B_{j}\cup B_{0}$. Observe that $r_{\tilde{B}_{j}}\leq(\beta+\beta^{\prime})\,r_{B}$, which is strictly less than $r_{B}$, because we assumed that $\beta^{\prime}<1-\beta$. Next we check that if $j$ is in $J^{\prime}$, then $\phi_{j}$ is a multiple of a $(1,r)$-atom: we give details in the case where $r=2$; the cases where $r\in(1,\infty)\setminus\\{2\\}$ may be treated similarly, and the variations needed to treat the case where $r=\infty$ are straightforward and are omitted. By the triangle inequality $\displaystyle\Bigl{(}\mbox{\small$\displaystyle\frac{1}{\mu(\widetilde{B}_{j})}$}\int_{\widetilde{B}_{j}}\left\|{\phi_{j}}\right\|^{2}\,\text{\rm d}\mu\Bigr{)}^{1/2}$ $\displaystyle\leq\Bigl{(}\mbox{\small$\displaystyle\frac{1}{\mu(\widetilde{B}_{j})}$}\int_{\widetilde{B}_{j}}\left\|{a\,\psi_{j}}\right\|^{2}\,\text{\rm d}\mu\Bigr{)}^{1/2}+\left\|{A_{j}}\right\|\Bigl{(}\mbox{\small$\displaystyle\frac{1}{\mu(\widetilde{B}_{j})}$}\int_{\widetilde{B}_{j}}{\mathbf{1}}_{B_{0}}\,\text{\rm d}\mu\Bigr{)}^{1/2}$ $\displaystyle\leq\sqrt{\mbox{\small$\displaystyle\frac{\mu(B)}{\mu(\widetilde{B}_{j})}$}}\,\Bigl{(}\mbox{\small$\displaystyle\frac{1}{\mu(B)}$}\int_{B}\left\|{a}\right\|^{2}\,\text{\rm d}\mu\Bigr{)}^{1/2}\ +\sqrt{\mbox{\small$\displaystyle\frac{\mu(B_{0})}{\mu(\widetilde{B}_{j})}$}}\,\mbox{\small$\displaystyle\frac{1}{\mu(B_{0})}$}\int_{M}\left\|{a\,\psi_{j}}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq\Bigl{(}\sqrt{\mbox{\small$\displaystyle\frac{\mu(B)}{\mu(B_{0})}$}}+\mbox{\small$\displaystyle\frac{\mu(B)}{\mu(B_{0})}$}\Bigr{)}\,\mbox{\small$\displaystyle\frac{1}{\mu(B)}$}$ $\displaystyle\leq\mbox{\small$\displaystyle\frac{2\,D_{1/\beta^{\prime},b}}{\mu(B)}$}.$ Observe that $\mbox{\small$\displaystyle\frac{1}{\mu(B)}$}\leq\mbox{\small$\displaystyle\frac{1}{\mu(B_{0})}$}\leq\mbox{\small$\displaystyle\frac{D_{\beta/\beta^{\prime}+1,b}}{\mu(\widetilde{B}_{j})}$},$ because $B_{0}$ is contained both in $B$ and $\widetilde{B}_{j}$ and the ratio between the radii of $\widetilde{B}_{j}$ and $B_{0}$ is at most $\beta/\beta^{\prime}+1$. Therefore we may conclude that $\Bigl{(}\mbox{\small$\displaystyle\frac{1}{\mu(\widetilde{B}_{j})}$}\int_{\widetilde{B}_{j}}\left\|{\phi_{j}}\right\|^{2}\,\text{\rm d}\mu\Bigr{)}^{1/2}\leq\frac{2\,D_{1/\beta^{\prime},b}\,D_{\beta/\beta^{\prime}+1,b}}{\mu(\widetilde{B}_{j})},$ i.e., $\phi_{j}/(2\,D_{1/\beta^{\prime},b}\,D_{\beta/\beta^{\prime}+1,b})$ is an atom supported in the ball $\widetilde{B}_{j}$ of radius at most $(\beta+\beta^{\prime})\,r_{B}$. Now suppose that $j$ is in $J^{\prime\prime}$. Then $B_{j}\cup B_{0}$ is contained in $2B_{0}$. Notice that $\beta^{\prime}<1-\beta<1/2$, so that $r_{2B_{0}}=2\beta^{\prime}r_{B}<(\beta+\beta^{\prime})\,r_{B}$. By arguing much as above, we see that $\phi_{j}/(2\,D_{1/\beta^{\prime},b}\,D_{\beta/\beta^{\prime}+1,b})$ is an atom supported in the ball $2B_{0}$ of radius $<(\beta+\beta^{\prime})\,r_{B}$. We have written $a$ as the sum of $N_{1}$ functions $\phi_{j}$, each of which is a multiple of an atom with constant $2\,D_{1/\beta^{\prime},b}\,D_{\beta/\beta^{\prime}+1,b}$. Thus, we have proved that ${\\|{a}\\|_{H_{b(\beta+\beta^{\prime})}^{1,r}}}\leq 2\,D_{1/\beta^{\prime},b}\,D_{\beta/\beta^{\prime}+1,b}\,{N_{1}}$. Now, if $j$ is in $J^{\prime}$, and $r_{\tilde{B}_{j}}<c$, then $\widetilde{B}_{j}$ is in $\mathcal{B}_{c}$. Similarly, if $j$ is in $J^{\prime\prime}$, and $r_{2B_{0}}<c$, then $2B_{0}$ is in $\mathcal{B}_{c}$. If $2B_{0}$ and all the balls $\widetilde{B}_{j}$ are in $\mathcal{B}_{c}$, then the proof is complete. Otherwise either $2B_{0}$ or some of the $\widetilde{B}_{j}$’s is not in $\mathcal{B}_{c}$, and we must iterate the construction above. It is clear that after a finite number of steps, depending on the ratio $b/c$, we end up with the required decomposition. Next we prove (ii). Obviously ${\\|{f}\\|_{H_{b}^{1,r}}}\leq{\\|{f}\\|_{H_{c}^{1,r}}}$, so we only have to show that ${\\|{f}\\|_{H_{c}^{1,r}}}\leq C\,{\\|{f}\\|_{H_{b}^{1,r}}}$ for some constant $C$ depending only on $b$ and $c$ and $M$. But this follows directly from (i). ∎ ###### Definition 4.4. Suppose that $r$ is in $(1,\infty)$. Then for every $b$ and $c$ in $\mathbb{R}^{+}$ such that $R_{0}/(1-\beta)<c<b$ the spaces $H_{b}^{1,r}(\mu)$ and $H_{c}^{1,r}(\mu)$ are isomorphic (in fact, they contain the same functions) by Proposition 4.3 (ii), and will simply be denoted by $H^{1,r}(\mu)$. Later (see Section 6) we shall prove that $H^{1,r}(\mu)$ does not depend on the parameter $r$ in $(1,\infty)$, and we shall denote $H^{1,r}(\mu)$ simply by $H^{1}(\mu)$. ## 5\. The space $BMO$ Suppose that $q$ is in $[1,\infty)$. For each locally integrable function $f$ we define $N_{b}^{q}(f)$ by $N_{b}^{q}(f)=\sup_{B\in\mathcal{B}_{b}}\Bigl{(}\frac{1}{\mu(B)}\int_{B}\left\|{f-f_{B}}\right\|^{q}\,\text{\rm d}\mu\Bigr{)}^{1/q},$ where $f_{B}$ denotes the average of $f$ over $B$. We denote by $BMO_{b}^{q}(\mu)$ the space of all equivalence classes of locally integrable functions $f$ modulo constants, such that $N_{b}^{q}(f)$ is finite, endowed with the norm ${\\|{f}\\|_{BMO_{b}^{q}(\mu)}}=N_{b}^{q}(f).$ Notice that only “small” balls enter in the definition of $BMO_{b}^{q}(\mu)$. It is a nontrivial fact, proved in Proposition 5.1 below, that $BMO_{b}^{q}(\mu)$ is independent of the parameter $b$, provided $b$ is large enough, and that the norms $N_{b}^{q}$ are all equivalent. ###### Proposition 5.1. Suppose that $q$ is in $[1,\infty)$, and $b$ and $c$ are positive constant such that $R_{0}/(1-\beta)<c<b$. Then $BMO_{b}^{q}(\mu)$ and $BMO_{c}^{q}(\mu)$ coincide and the norms $N_{b}^{q}$ and $N_{c}^{q}$ are equivalent. ###### Proof. Obviously, if $0<c<b$ and $f$ is in $BMO_{b}^{q}(\mu)$, then $f$ is in $BMO_{c}^{q}(\mu)$ and $N_{c}^{q}(f)\leq N_{b}^{q}(f)$. Thus, we only have to show that $N_{b}^{q}(f)\leq C\,N_{c}^{q}(f)$ for some constant $C$ depending only on $b$ and $c$ and $M$. We give the proof in the case where $q=1$; the proof in the other cases is similar. Suppose that $B$ is a ball in $\mathcal{B}_{b}$. Observe that $\displaystyle\frac{1}{\mu(B)}\int_{B}\left\|{f-f_{B}}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq\frac{2}{\mu(B)}\,\inf_{c\in\mathbb{C}}\int_{B}\left\|{f-c}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq\frac{2}{\mu(B)}\,{\\|{f}\\|_{L^{1}(B)/\mathbb{C}}},$ where ${L^{1}(B)/\mathbb{C}}$ is the quotient of the space $L^{1}(B)$ modulo the constants. Since the dual of ${L^{1}(B)/\mathbb{C}}$ is $L^{\infty}_{0}(B)$ (the space of all functions in $L^{\infty}(B)$ with vanishing integral, endowed with the $L^{\infty}(B)$ norm), ${\\|{f}\\|_{L^{1}(B)/\mathbb{C}}}=\sup\Bigl{\\{}\Bigl{\|}{\int_{B}f\,\phi\,\,\text{\rm d}\mu}\Bigr{\|}:\phi\in L^{\infty}_{0}(B),{\\|{\phi}\\|_{\infty}}\leq 1\Bigr{\\}}.$ Suppose that $\phi$ is a function in $L^{\infty}_{0}(B)$ with ${\\|{\phi}\\|_{\infty}}\leq 1$. Then $\phi/\mu(B)$ is a $(1,\infty)$-atom and, by Proposition 4.3 (i), there exist $(1,\infty)$-atoms $a_{1},\ldots,a_{N}$ supported in balls $B_{j}$ in $\mathcal{B}_{c}$ whose union contains $B$ such that $\phi/\mu(B)=\sum_{j=1}^{N}\lambda_{j}\,a_{j}$, with $\left\|{\lambda_{j}}\right\|\leq C$ and ${\\|{a_{j}}\\|_{\infty}}\leq 1/\mu(B_{j})$, where $C$ and $N$ are constants which depend only on $b$, $c$ and $M$. Thus $\displaystyle\frac{1}{\mu(B)}\Bigl{\|}{\int_{B}f\,\phi\,\text{\rm d}\mu}\Bigr{\|}$ $\displaystyle\leq C\sum_{j=0}^{N}\,\int_{B_{j}}\bigl{\|}{f-f_{B_{j}}}\bigr{\|}\,\left\|{a_{j}}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq C\,\sum_{j=0}^{N}\,\frac{1}{\mu(B_{j})}\int_{B_{j}}\left\|{f-f_{B_{j}}}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq C\,N\,N_{c}^{1}(f).$ Hence $N_{b}^{1}(f)\leq 2\,C\,N\,N_{c}^{1}(f)$, as required. ∎ ###### Remark 5.2. For the rest of this paper, we fix a constant $b_{0}>R_{0}/(1-\beta)$. For each $q$ in $[1,\infty)$ we denote by $BMO^{q}(\mu)$ the space $BMO_{b_{0}}^{q}(\mu)$ endowed with any of the equivalent norms $N_{b}^{q}$, where $b>R_{0}/(1-\beta)$. Next, we want to show that $BMO^{q}(\mu)$ is independent of $q$ (see the remark at the end of this section). The strategy is the same as in the classical case: it hinges on a John–Nirenberg type inequality for functions in $BMO^{1}(\mu)$. The original inequality was proved in [JN], where classical functions of bounded mean oscillation appeared for the first time. We need the following generalization of the John–Nirenberg inequality to doubling spaces which is stated in [CW] and proved in [Buc, Thm 2.2] (see also [MMNO, MP]). ###### Proposition 5.3. Suppose that $(X,d,\mu)$ is a doubling metric measure space, with doubling constant $D$. There exist constants $J_{D}$ and $\eta_{D}$, which depend only on $D$, such that $\mu\bigl{(}\bigl{\\{}x\in B:\bigl{\|}{f(x)-f_{B}}\bigr{\|}>s\bigr{\\}}\bigr{)}\leq J_{D}\,\mathrm{e}^{-\eta_{D}s/{\\|{f}\\|_{BMO(X)}}}\,\mu(B)\qquad\forall s\in\mathbb{R}^{+}\quad\forall B.$ By Proposition 3.4 (ii) for each dyadic cube $Q$ the measured metric space $(Q,\rho_{\|Q},\mu_{\|Q})$ is a space of homogeneous type. We denote by $BMO(Q)$ the classical $BMO$ space on $Q$. ###### Theorem 5.4. Denote by $\nu$ the unique integer such that $\delta^{\nu}\leq b_{0}<\delta^{\nu-1}$, and by $N$ the norm $N_{2\max(C_{1},b_{0})}^{1}$ on $BMO(\mu)$. The following hold: 1. (i) for each dyadic cube $Q$ in $\bigcup_{k=\nu}^{\infty}\mathcal{Q}^{k}$ and for each $f$ in $BMO^{1}(\mu)$ the restriction of $f$ to $Q$ is in $BMO(Q)$ and ${\\|{f}\\|_{BMO(Q)}}\leq 2\,D_{C_{1}/(a_{0}\delta),b_{0}\max(1,a_{0})}\,N(f);$ 2. (ii) there exist positive constants $J$ and $\eta$ such that for every function $f$ in $BMO^{1}(\mu)$ and for every ball $B$ in $\mathcal{B}_{b_{0}}$ $\mu\bigl{(}\\{x\in B:\left\|{f(x)-f_{B}}\right\|>s\\}\bigr{)}\leq J\,\mathrm{e}^{-\eta\,s/N(f)}\,\mu(B).$ ###### Proof. First we prove (i). Suppose that $Q$ is in $\mathcal{Q}^{k}$. Recall that a ball in $Q$ is the intersection of $Q$ with a ball $B$ in $\mathcal{B}$ whose centre belongs to $Q$. We have to estimate the oscillation $\mathrm{osc}_{f}(B\cap Q)$ of $f$ over $B\cap Q$ defined by (5.1) $\mathrm{osc}_{f}(B\cap Q)=\frac{1}{\mu(B\cap Q)}\int_{B\cap Q}\bigl{\|}{f-f_{B\cap Q}}\bigr{\|}\,\text{\rm d}\mu.$ We shall prove that $\mathrm{osc}_{f}(B\cap Q)\leq 2\,D_{C_{1}/(a_{0}\delta),b_{0}\max(1,a_{0})}\,N(f)\qquad\forall f\in BMO^{1}(\mu),$ from which (i) follows. We consider the cases where $r_{B}<C_{1}\,\delta^{k}$ and $r_{B}\geq C_{1}\,\delta^{k}$ separately. In the case where $r_{B}<C_{1}\,\delta^{k}$ we compare (5.1) with the oscillation of $f$ over $B$. By the triangle inequality $\displaystyle\mathrm{osc}_{f}(B\cap Q)$ $\displaystyle\leq\frac{1}{\mu(B\cap Q)}\int_{B\cap Q}\bigl{\|}{f-f_{B}}\bigr{\|}\,\text{\rm d}\mu+\bigl{\|}{f_{B}-f_{B\cap Q}}\bigr{\|}$ $\displaystyle\leq\frac{2}{\mu(B\cap Q)}\int_{B\cap Q}\bigl{\|}{f-f_{B}}\bigr{\|}\,\text{\rm d}\mu.$ By Proposition 3.4 (i) we know that $\mu(B\cap Q)\geq D_{C_{1}/(a_{0}\delta),\delta^{\nu}}^{-1}\,\mu(B)$; hence the right hand side in the displayed formula above may be estimated from above by $\frac{2\,D_{C_{1}/(a_{0}\delta),\delta^{\nu}}}{\mu(B)}\int_{B}\bigl{\|}{f-f_{B}}\bigr{\|}\,\text{\rm d}\mu,$ which, in turn, may be majorised by $D_{C_{1}/(a_{0}\delta),b_{0}}\,N(f)$. Now suppose that $r_{B}\geq C_{1}\,\delta^{k}$. Since $\mathrm{diam}(Q)<C_{1}\,\delta^{k}$ by Theorem 3.2 (iv), $Q\cap B=Q$. For the sake of definitess, suppose that $Q$ is the dyadic cube $Q_{\beta}^{k}$. Then $Q_{\beta}^{k}$ contains the ball $B(z_{\beta}^{k},a_{0}\,\delta^{k})$. Denote by $\widetilde{B}$ the ball centred at $z_{\beta}^{k}$ and radius $C_{1}\,\delta^{k}$. Now, $\displaystyle\mathrm{osc}_{f}(B\cap Q)$ $\displaystyle\leq\frac{1}{\mu(B\cap Q)}\int_{B\cap Q}\bigl{\|}{f-f_{\widetilde{B}}}\bigr{\|}\,\text{\rm d}\mu+\bigl{\|}{f_{\widetilde{B}}-f_{B\cap Q}}\bigr{\|}$ $\displaystyle\leq\frac{2}{\mu(B\cap Q)}\int_{B\cap Q}\bigl{\|}{f-f_{\widetilde{B}}}\bigr{\|}\,\text{\rm d}\mu$ $\displaystyle\leq\frac{2}{\mu\bigl{(}B(z_{\beta}^{k},a_{0}\,\delta^{k})\bigr{)}}\int_{\widetilde{B}}\bigl{\|}{f-f_{\widetilde{B}}}\bigr{\|}\,\text{\rm d}\mu$ $\displaystyle\leq\frac{2\,D_{C_{1}/a_{0},a_{0}\delta^{\nu}}}{\mu\bigl{(}\widetilde{B}\bigr{)}}\int_{\widetilde{B}}\bigl{\|}{f-f_{\widetilde{B}}}\bigr{\|}\,\text{\rm d}\mu,$ which is majorised by $2\,D_{C_{1}/a_{0},a_{0}b_{0}}\,N(f)$. The proof of (i) is complete. Now we prove (ii). Suppose that $B$ is $\mathcal{B}_{b_{0}}$. Denote by $k$ the unique integer such that $\delta^{k}\leq r_{B}<\delta^{k-1}$ and by $Q_{1},\ldots,Q_{N}$ the dyadic cubes of resolution $k$ that intersect $B$. By Proposition 3.4 (iv) we have the estimate $N\leq D_{(1+C_{1})/(a_{0}\delta),b_{0}}$. Then (5.2) $\mu\bigl{(}\\{x\in B:\left\|{f(x)-f_{B}}\right\|>s\\}\bigr{)}\leq\sum_{j=1}^{N}\mu\bigl{(}\\{x\in Q_{j}:\left\|{f(x)-f_{B}}\right\|>s\\}\bigr{)}.$ We estimate each of the summands on the right hand side from above by (5.3) $\mu\bigl{(}\\{x\in Q_{j}:\left\|{f(x)-f_{Q_{j}}}\right\|>s/2\\}\bigr{)}+\mu\bigl{(}\\{x\in Q_{j}:\left\|{f_{B}-f_{Q_{j}}}\right\|>s/2\\}\bigr{)}.$ By Proposition 5.3 and (i) the first summand in this formula is majorised by $J_{Q_{j}}\,\mathrm{e}^{-\eta_{Q_{j}}s/{\\|{f}\\|_{BMO(Q_{j})}}}\,\mu(Q_{j})\leq J_{Q_{j}}\,\mathrm{e}^{-\eta_{Q_{j}}s/(2D_{C_{1}/(a_{0}\delta),b_{0}\max(1,a_{0})}N(f))}\,\mu(Q_{j}).$ Here we use the fact that since $\mathrm{diam}(Q_{j})$ is finite, then $Q_{j}$ is a ball in the doubling space $(Q_{j},\rho_{\|Q_{j}},\mu_{\|Q_{j}})$. By Proposition 3.4 (ii) all the spaces $(Q_{j},\rho_{\|Q},\mu_{\|Q})$ are spaces of homogeneous type with doubling constant dominated by $D_{\tau,C_{1}b_{0}}\,D_{C_{1}/(a_{0}\delta),b_{0}}$, which we simply denote by $D^{\prime}$. Also, denote by $\eta^{\prime}$ the constant $\eta_{D^{\prime}}/(2D_{C_{1}/(a_{0}\delta),b_{0}\max(1,a_{0})})$. By Proposition 3.4 (iii) the ball $\widetilde{B}$ with centre $c_{B}$ and radius $(1+C_{1})\,r_{B}$ contains $Q_{1},\ldots,Q_{N}$ and $\mu(\widetilde{B})\leq D_{C_{1}+1,{b_{0}}}\,\mu(B)$. Thus, by summing over $j$, we see that (5.4) $\displaystyle\sum_{j=1}^{N}\mu\bigl{(}\\{x\in Q_{j}:\left\|{f(x)-f_{Q_{j}}}\right\|>s/2\\}\bigr{)}$ $\displaystyle\leq J_{D^{\prime}}\,\mathrm{e}^{-\eta^{\prime}s/N(f)}\,\sum_{j=1}^{N}\mu(Q_{j})$ $\displaystyle\leq J_{D^{\prime}}\,\mathrm{e}^{-\eta^{\prime}s/N(f)}\,\mu(\widetilde{B})$ $\displaystyle\leq J_{D^{\prime}}\,\mathrm{e}^{-\eta^{\prime}s/N(f)}D_{C_{1}+1,b_{0}}\,\mu(B).$ Now we estimate the second summand in (5.3). We claim that $\left\|{f_{B}-f_{Q_{j}}}\right\|\leq 2\,D_{(1+C_{1})/(a_{0}\delta),b_{0}}\,N(f).$ Indeed, $\displaystyle\left\|{f_{B}-f_{Q_{j}}}\right\|$ $\displaystyle\leq\left\|{f_{B}-f_{\widetilde{B}}}\right\|+\left\|{f_{\widetilde{B}}-f_{Q_{j}}}\right\|$ $\displaystyle\leq\frac{1}{\mu(B)}\int_{B}\left\|{f-f_{\widetilde{B}}}\right\|\,\text{\rm d}\mu+\frac{1}{\mu(Q_{j})}\int_{Q_{j}}\left\|{f-f_{\widetilde{B}}}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq\frac{1}{\mu(B)}\int_{B}\left\|{f-f_{\widetilde{B}}}\right\|\,\text{\rm d}\mu+\frac{1}{\mu\bigl{(}B(z_{j},a_{0}\delta^{k})\bigr{)}}\int_{\widetilde{B}}\left\|{f-f_{\widetilde{B}}}\right\|\,\text{\rm d}\mu$ $\displaystyle\leq\frac{2\,D_{(1+C_{1})/(a_{0}\delta),b_{0}}}{\mu(\widetilde{B})}\int_{\widetilde{B}}\left\|{f-f_{\widetilde{B}}}\right\|\,\text{\rm d}\mu,$ which is dominated by $2\,D_{(1+C_{1})/(a_{0}\delta),b_{0}}\,N(f)$, as claimed. Thus $\mu\bigl{(}\\{x\in Q_{j}:\left\|{f_{B}-f_{Q_{j}}}\right\|>s/2\\}\bigr{)}\leq\mu\bigl{(}\\{x\in Q_{j}:2D_{(1+C_{1})/(a_{0}\delta),b_{0}}\,N(f)>s/2\\}\bigr{)}.$ Now, the right hand side is equal to $\mu(Q_{j})$ when $s$ is in $\bigl{(}0,4D_{(1+C_{1})/(a_{0}\delta),b_{0}}\,N(f)\bigr{)}$, and to $0$ when $s$ is in $\bigl{[}4D_{(1+C_{1})/(a_{0}\delta),b_{0}}\,N(f),\infty\bigr{)}$, so that $\mu\bigl{(}\\{x\in Q_{j}:\left\|{f_{B}-f_{Q_{j}}}\right\|>s/2\\}\bigr{)}\leq\mathrm{e}^{4D_{(1+C_{1})/(a_{0}\delta),b_{0}}}\,\mathrm{e}^{-s/N(f)}\,\mu(Q_{j})\qquad\forall s\in\mathbb{R}^{+}.$ Therefore (5.5) $\displaystyle\sum_{j=1}^{N}\mu\bigl{(}\\{x\in Q_{j}:\left\|{f_{B}-f_{Q_{j}}}\right\|>s/2\\}\bigr{)}$ $\displaystyle\leq\mathrm{e}^{4D_{(1+C_{1})/(a_{0}\delta),b_{0}}}\,\mathrm{e}^{-s/N(f)}\,\mu(\widetilde{B})$ $\displaystyle\leq\mathrm{e}^{4D_{(1+C_{1})/(a_{0}\delta),b_{0}}}\,D_{C_{1}+1,b_{0}}\,\mathrm{e}^{-s/N(f)}\,\mu(B).$ Now, (5.4) and (5.5) imply that $\displaystyle\mu\bigl{(}\\{x\in B:\left\|{f(x)-f_{B}}\right\|>s\\}\bigr{)}$ $\displaystyle\leq\bigl{(}J_{D^{\prime}}\,\mathrm{e}^{-\eta^{\prime}s/N(f)}+\mathrm{e}^{4D_{(1+C_{1})/(a_{0}\delta),b_{0}}}\,\mathrm{e}^{-s/N(f)}\bigr{)}\,D_{C_{1}+1,b_{0}}\,\mu(B)$ $\displaystyle\leq J\,\mathrm{e}^{-\eta\,s/N(f)}\,\mu(B),$ where $J=\bigl{(}J_{D^{\prime}}+\mathrm{e}^{4D_{(1+C_{1})/(a_{0}\delta),b_{0}}}\bigr{)}\,D_{C_{1}+1,b_{0}}$ and $\eta=\min(1,\eta^{\prime})$, as required. ∎ A standard consequence of the John–Nirenberg type inequality is the following. ###### Corollary 5.5. Denote by $\nu$ the unique integer such that $\delta^{\nu}\leq b_{0}<\delta^{\nu-1}$, and by $N$ the norm $N_{2\max(C_{1},b_{0})}^{1}$ on $BMO(\mu)$. The following hold: 1. (i) for every $c<\eta$ $\int_{B}\mathrm{e}^{c\,\left\|{f-f_{B}}\right\|/N(f)}\,\text{\rm d}\mu\leq\Bigl{(}1+\frac{Jc}{\eta-c}\Bigr{)}\,\mu(B)\qquad\forall f\in BMO(\mu)\quad\forall B\in\mathcal{B}_{b_{0}},$ where $\eta$ and $J$ are as in Theorem 5.4 (ii); 2. (ii) for each $q$ in $(1,\infty)$ there exists a constant $C$ such that $\Bigl{(}\frac{1}{\mu(B)}\int_{B}\left\|{f-f_{B}}\right\|^{q}\,\text{\rm d}\mu\Bigr{)}^{1/q}\leq C\,N(f)\qquad\forall f\in BMO(\mu)\quad\forall B\in\mathcal{B}_{b_{0}}.$ ###### Proof. First we prove (i). Observe that the left hand side of (i) is equal to $\mu(B)+\int_{1}^{\infty}\mu\left(\\{x\in B:\left\|{f-f_{B}}\right\|>N(f)\,(\log\beta)/c\\}\right)\,\text{\rm d}\beta.$ Changing variables and using the John–Nirenberg type inequality proved in Theorem 5.4 we see that the last integral may be estimated by $\mu(B)\Bigl{[}1+J\,\int_{0}^{\infty}\mathrm{e}^{(c-\eta)v/c}\,\text{\rm d}v\Bigr{]}.$ The above integral is finite if and only if $c<\eta$ and it is equal to $c/(\eta-c)$: the required inequality follows. Now we prove (ii). By elementary calculus, for each $q$ in $(1,\infty)$ there exists a constant $C_{q}$ such that $e^{s}\geq C_{q}\,s^{q}$ for every $s$ in $\mathbb{R}^{+}$. Therefore (i) implies that $C_{q}\,\Bigl{(}\frac{c}{N(f)}\Bigr{)}^{q}\int_{B}\,\left\|{f-f_{B}}\right\|^{q}\,\text{\rm d}\mu\leq\Bigl{(}1+\frac{Jc}{\eta-c}\Bigr{)}\,\mu(B),$ which is equivalent to the required estimate. The proof of the corollary is complete. ∎ ###### Remark 5.6. By Corollary 5.5 (ii), if $f$ is in $BMO^{1}(\mu)$, then $f$ is in $BMO^{q}(\mu)$ for all $q$ in $(1,\infty)$. Conversely, if $f$ is in $BMO^{q}(\mu)$ for some $q$ in $(1,\infty)$, then trivially it is in $BMO^{1}(\mu)$, hence in $BMO^{r}(\mu)$ for all $r$ in $(1,\infty)$ by Corollary 5.5 (ii). Furthermore, the norms $N_{b_{0}}^{1}$ and $N_{b_{0}}^{q}$ are equivalent. In view of this observation, all spaces $BMO^{q}(\mu)$, $q$ in $[1,\infty)$, coincide. We shall denote $BMO^{1}(\mu)$ simply by $BMO(\mu)$. We endow $BMO(\mu)$ with any of the equivalent norms $N_{b}^{q}$, where $q$ is in $[1,\infty)$ and $b>R_{0}/(1-\beta)$. This remark will be important in the proof of the duality between the Hardy space $H^{1}(\mu)$ and $BMO(\mu)$ (see Section 6 below). ## 6\. Duality We shall prove that the topological dual of $H^{1,r}({\mu})$ may be identified with $BMO^{r^{\prime}}({\mu})$, where $r^{\prime}$ denotes the index conjugate to $r$. Suppose that $1<r<s<\infty$. Then $\bigl{(}H^{1,r}({\mu})\bigr{)}^{}=\bigl{(}H^{1,s}({\mu})\bigr{)}^{}$, because we have proved that $BMO^{r^{\prime}}({\mu})=BMO^{s^{\prime}}({\mu})$ (see Remark 5.6). Observe that the identity is a continuous injection of $H^{1,s}(\mu)$ into $H^{1,r}(\mu)$, and that $H^{1,s}(\mu)$ is a dense subspace of $H^{1,r}(\mu)$. Then we may conclude that $H^{1,s}(\mu)=H^{1,r}(\mu)$. We need some more notation and some preliminary observation. For each ball $B$ in $\mathcal{B}_{b_{0}}$ let $L^{2}_{0}(B)$ denote the Hilbert space of all functions $f$ in $L^{2}(\mu)$ such that the support of $f$ is contained in $B$ and $\int_{B}f\,\text{\rm d}{\mu}=0$. We remark that a function $f$ in $L^{2}_{0}(B)$ is a multiple of a $(1,2)$-atom, and that (6.1) ${\\|{f}\\|_{H^{1,2}({\mu})}}\leq\mu(B)^{1/2}\,{\\|{f}\\|_{L^{2}(B)}}.$ Suppose that $\ell$ is a bounded linear functional on $H^{1,2}({\mu})$. Then, for each $B$ in $\mathcal{B}_{b_{0}}$ the restriction of $\ell$ to $L^{2}_{0}(B)$ is a bounded linear functional on $L^{2}_{0}(B)$. Therefore, by the Riesz representation theorem there exists a unique function $\ell^{B}$ in $L^{2}_{0}(B)$ which represents the restriction of $\ell$ to $L^{2}_{0}(B)$. Note that for every constant $\eta$ the function $\ell^{B}+\eta$ represents the same functional, though it is not in $L^{2}_{0}(B)$ unless $\eta$ is equal to $0$. Denote by $\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,2}(\mu)}$ the norm of $\ell$. Observe that (6.2) $\displaystyle{\\|{\ell^{B}}\\|_{L^{2}_{0}(B,{\mu})}}$ $\displaystyle=\sup_{{\\|{f}\\|_{L^{2}_{0}(B)=1}}}\Bigl{\|}{\int_{B}\ell^{B}\,f\,\text{\rm d}{\mu}}\Bigr{\|}$ $\displaystyle\leq\sup_{{\\|{f}\\|_{L^{2}_{0}(B)=1}}}\,\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,2}({\mu})}\,{\\|{f}\\|_{H^{1,2}({\mu})}}$ $\displaystyle\leq\mu(B)^{1/2}\,\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,2}({\mu})},$ the last inequality being a consequence of (6.1). For every $f$ in $BMO^{r^{\prime}}({\mu})$ and every finite linear combination $g$ of $(1,r)$-atoms the integral $\int_{\mathbb{R}^{d}}f\,g\,\text{\rm d}{\mu}$ is convergent. Denote by $H_{\mathrm{fin}}^{1,r}({\mu})$ the subspace of $H^{1,r}({\mu})$ consisting of all finite linear combinations of $(1,r)$-atoms. Then $g\mapsto\int_{\mathbb{R}^{d}}f\,g\,\text{\rm d}{\mu}$ defines a linear functional on $H_{\mathrm{fin}}^{1,r}({\mu})$. We observe that $H_{\mathrm{fin}}^{1,r}({\mu})$ is dense in $H^{1,r}({\mu})$. ###### Theorem 6.1. Suppose that $r$ is in $(1,\infty)$. The following hold 1. (i) for every $f$ in $BMO^{r^{\prime}}({\mu})$ the functional $\ell$, initially defined on $H_{\mathrm{fin}}^{1,r}({\mu})$ by the rule $\ell(g)=\int_{\mathbb{R}^{d}}f\,g\,\text{\rm d}{\mu},$ extends to a bounded functional on $H^{1,r}({\mu})$. Furthermore, $\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,r}({\mu})}\leq{\\|{f}\\|_{BMO^{r^{\prime}}({\mu})}};$ 2. (ii) there exists a constant $C$ such that for every continuous linear functional $\ell$ on $H^{1,r}({\mu})$ there exists a function $f^{\ell}$ in $BMO^{r^{\prime}}({\mu})$ such that ${\\|{f^{\ell}}\\|_{BMO^{r^{\prime}}({\mu})}}\leq C\,\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,r}({\mu})}$ and $\ell(g)=\int_{\mathbb{R}^{d}}f^{\ell}\,g\,\text{\rm d}{\mu}\qquad\forall g\in H_{\mathrm{fin}}^{1,r}({\mu}).$ ###### Proof. The proof of (i) follows the line of the proof of [CW] which is based on the classical result of C. Fefferman [F, FS]. We omit the details. Now we prove (ii) in the case where $r$ is equal to $2$. The proof for $r$ in $(1,\infty)\setminus\\{2\\}$ is similar and is omitted. Recall that for each $b>R_{0}/(1-\beta)$ the space $H^{1,2}(\mu)$ is isomorphic to $H_{b}^{1,2}(\mu)$ with norm ${\\|{\cdot}\\|_{H_{b}^{1,2}(\mu)}}$. Thus, we may interpret $\ell$ as a continuous linear functional on $H_{b}^{1,2}(\mu)$ for each $b>R_{0}/(1-\beta)$. Fix a reference point $o$ in $M$. For each $b$ there exists a function $f_{b}^{\ell}$ in $L^{2}_{0}(B(o,b))$ that represents $\ell$ in $B(o,b)$. Since both $f_{1}^{\ell}$ and the restriction of $f_{b}^{\ell}$ represent $\ell$ on $B(o,1)$, there exists a constant $\eta_{b}$ such that $f_{1}^{\ell}-f_{b}^{\ell}=\eta_{b}$ on $B(o,1)$. By integrating both sides of this equality on $B(o,1)$ we see that $\eta_{b}=-\frac{1}{\mu\bigl{(}B(o,1)\bigr{)}}\int_{B(o,1)}f_{b}^{\ell}\,\text{\rm d}\mu.$ Define $f^{\ell}(x)=f_{b}^{\ell}(x)+\eta_{b}\qquad\forall x\in B(o,b)\quad\forall b\in[1,\infty).$ It is straightforward to check that this is a good definition. We claim that the function $f^{\ell}$ is in $BMO({\mu})$ and there exists a constant $C$ such that ${\\|{f^{\ell}}\\|_{BMO({\mu})}}\leq C\,\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,2}({\mu})^{}}\qquad\forall\ell\in H^{1,2}({\mu})^{}.$ Indeed, choose a ball $B$ in $\mathcal{B}_{b_{0}}$. Then there exists a constant $\eta^{B}$ such that (6.3) $f^{\ell}\big{\|}_{B}=\ell^{B}+\eta^{B},$ where $\ell^{B}$ is in $L^{2}_{0}(B)$ and represents the restriction of $\ell$ to $L^{2}_{0}(B)$. By integrating both sides on $B$, we see that $\eta^{B}=\bigl{(}f^{\ell})_{B}$. Then, by (6.3), $\displaystyle\Bigl{(}\frac{1}{\mu(B)}\int_{B}\bigl{\|}{f^{\ell}-\bigl{(}f^{\ell}\bigr{)}_{B}}\bigr{\|}^{2}\,\text{\rm d}{\mu}\Bigr{)}^{1/2}$ $\displaystyle=\Bigl{(}\frac{1}{\mu(B)}\int_{B}\bigl{\|}{\ell^{B}}\bigr{\|}^{2}\,\text{\rm d}{\mu}\Bigr{)}^{1/2}$ $\displaystyle\leq\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,2}({\mu})},$ so that $N_{b_{0}}^{2}(f^{\ell})\leq\|\\!\|\\!\|{\ell}\|\\!\|\\!\|_{H^{1,2}({\mu})}$, as required. ∎ ###### Remark 6.2. Note that the proof of Theorem 6.1 does not apply, strictly speaking, to the case where $r$ is equal to $\infty$. However, a straightforward, though tedious, adaptation to the case where $\mu$ is only locally doubling of a classical result [CW], show that $H^{1,\infty}(\mu)$ and $H^{1,2}(\mu)$ agree, with equivalence of norms. Consequently, the dual space of $H^{1,\infty}(\mu)$ is $BMO(\mu)$. ## 7\. Estimates for the sharp function and interpolation The main step in the proof of Fefferman–Stein’s interpolation result for analytic families of operators is a certain relative distributional inequality (also referred to as “good $\lambda$ inequality” in the literature) [FS, Thm 5, p. 153], [St2], which is a modified version of the original relative distributional inequality of D.L. Burkholder and R.F. Gundy [BG, Bur] for martingales. Extensions of Fefferman–Stein’s distributional inequality to spaces of homogeneous type are available in the literature (see, e.g., Macías’ thesis [Ma]). It may be worth observing that the doubling property plays a key rôle in their proof. An extension of this theory to rank one symmetric spaces of the noncompact type is due to Ionescu [I]. In this section we adapt Ionescu’s ideas and arguments to our setting. For each integer $k$, and each locally integrable function $f$, the _noncentred dyadic local Hardy–Littlewood maximal function_ $\mathcal{M}_{k}f$ is defined by (7.1) $\mathcal{M}_{k}f(x)=\sup_{Q}\frac{1}{\mu(Q)}\int_{Q}\left\|{f}\right\|\,\text{\rm d}\mu\qquad\forall x\in M,$ where the supremum is taken over all dyadic cubes of resolution $\geq k$ that contain $x$. For each $p$ in $M$ we denote by $\mathcal{B}_{b}(p)$ the subcollection of all balls in $\mathcal{B}_{b}$ which contain $p$. For each $b$ in $\mathbb{R}^{+}$ we define a _local sharp function_ $f^{\sharp,b}$ of a locally integrable function $f$ thus: $f^{\sharp,b}(p)=\sup_{B\in\mathcal{B}_{b}(p)}\frac{1}{\mu(B)}\int_{B}\left\|{f-f_{B}}\right\|\,\text{\rm d}\mu\qquad\forall p\in M.$ Observe that $f$ is in $BMO(\mu)$ if and only if ${\\|{f^{\sharp,b}}\\|_{\infty}}$ is finite for some (hence for all) $b$ in $(R_{0}/(1-\beta),\infty)$. We shall need the following result, whose proof, _mutatis mutandis_ , is the same as that of its Euclidean analogue. ###### Theorem 7.1. Suppose that $k$ is an integer. Then the noncentred dyadic local Hardy–Littlewood maximal operator $\mathcal{M}_{k}$ is bounded on $L^{p}(\mu)$ for every $p$ in $(1,\infty]$ and of weak type $1$. ###### Lemma 7.2. Define constants $C_{0}$, $b^{\prime}$, $\sigma$ and $D$ by $C_{0}=\max(C_{1}/\delta,\delta),\quad b^{\prime}=\max(b_{0},2C_{1}+C_{0}),\quad\sigma=\bigl{(}1-\mathrm{e}^{-I_{M}\,\delta^{3}}\bigr{)}/2\quad\hbox{and}\quad D=D_{b^{\prime}/a_{0},a_{0}},$ where $a_{0}$, $C_{1}$ and $\delta$ are as in Theorem 3.2, and $D_{b^{\prime}/a_{0},a_{0}}$ is defined in Remark 2.3. For every $\eta^{\prime}$ in $(0,1)$, for all positive $\varepsilon<(1-\eta^{\prime})/(2D)$, and for every $f$ in $L^{1}(\mu)$ $\mu\bigl{(}\\{\mathcal{M}_{2}f>\alpha,\,f^{\sharp,b^{\prime}}\leq\varepsilon\,\alpha\\}\bigr{)}\leq\eta\,\mu\bigl{(}\\{\mathcal{M}_{2}f>\eta^{\prime}\,\alpha\\}\bigr{)}\qquad\forall\alpha\in\mathbb{R}^{+},$ where $\eta=1-\sigma+\mbox{\small$\displaystyle\frac{2\varepsilon\,D}{\sigma\,(1-\eta^{\prime})}$}.$ ###### Proof. For each $\beta>0$ we denote by $A(\beta)$ and $S(\beta)$ the sets $\\{\mathcal{M}_{2}f>\beta\\}$ and $\\{f^{\sharp,b^{\prime}}>\beta\\}$ respectively. The inequality to prove may then be rewritten as follows: $\mu\bigl{(}A({\alpha})\cap S({\varepsilon\alpha})^{c}\bigr{)}\leq\eta\,\mu\bigl{(}A({\eta^{\prime}\alpha})\bigr{)}\qquad\forall\alpha\in\mathbb{R}^{+}.$ To each $x$ in $A({\eta^{\prime}\alpha})$ we associate the maximal dyadic cube $Q_{x}$ containing $x$ of resolution at least $2$ such that $\left\|{f}\right\|_{Q_{x}}>\eta^{\prime}\alpha$. Here $\left\|{f}\right\|_{Q_{x}}$ denotes the average of $\left\|{f}\right\|$ on the cube $Q_{x}$. We denote by $\mathcal{C}_{\eta^{\prime}\alpha}$ the collection of cubes $\\{Q_{x}\\}_{x\in A({\eta^{\prime}\alpha})}$. Clearly $A(\eta^{\prime}\alpha)=\bigcup_{x\in A(\eta^{\prime}\alpha)}Q_{x}$, and $\mu\bigl{(}A(\eta^{\prime}\alpha)\bigr{)}<\infty$, because $\mathcal{M}_{2}$ is of weak type 1. By Proposition 3.5 (with $\kappa=\delta^{3}$) there exist mutually disjoint cubes $Q_{1},\ldots,Q_{k}$ in $\mathcal{C}_{\eta^{\prime}\alpha}$ such that $\rho\bigl{(}Q_{j},A(\eta^{\prime}\alpha)^{c}\bigr{)}\leq\delta^{3}$ and (7.2) $\sum_{j=1}^{k}\mu(Q_{j})\geq\sigma\,\mu\bigl{(}A(\eta^{\prime}\alpha)\bigr{)}.$ We claim that if $0<\varepsilon<(1-\eta^{\prime})/(2D)$ then (7.3) $\mu\bigl{(}Q_{j}\cap A(\alpha)\cap S(\varepsilon\alpha)^{c}\bigr{)}\leq\mbox{\small$\displaystyle\frac{2\varepsilon\,D}{\sigma\,(1-\eta^{\prime})}$}\,\mu\bigl{(}Q_{j}\bigr{)}\qquad\forall j\in\\{1,\ldots,k\\}.$ We postpone for a moment the proof of the claim and show how (7.3) implies the required conclusion. Observe that $A(\alpha)\subset A(\eta^{\prime}\alpha)$ and that $\displaystyle\mu\bigl{(}A(\alpha)\cap S(\varepsilon\alpha)^{c}\bigr{)}$ $\displaystyle=\mu\Bigl{(}\bigl{(}A(\eta^{\prime}\alpha)\setminus\bigcup_{j=1}^{k}Q_{j}\bigr{)}\cap A(\alpha)\cap S(\varepsilon\alpha)^{c}\Bigr{)}+\mu\Bigl{(}\bigl{(}\bigcup_{j=1}^{k}Q_{j}\bigr{)}\cap A(\alpha)\cap S(\varepsilon\alpha)^{c}\Bigr{)}$ $\displaystyle\leq(1-\sigma)\,\mu\bigl{(}A(\eta^{\prime}\alpha)\bigr{)}+\mbox{\small$\displaystyle\frac{2\varepsilon\,D}{\sigma\,(1-\eta^{\prime})}$}\,\sum_{j=1}^{k}\mu(Q_{j})$ $\displaystyle\leq\eta\,\mu\bigl{(}A(\eta^{\prime}\alpha)\bigr{)}.$ The penultimate inequality is a consequence of (7.2) and of (7.3), and the last inequality follows from the fact that the $Q_{j}$’s are mutually disjoint cubes contained in $A(\eta^{\prime}\alpha)$. Thus, to conclude the proof of the lemma it remains to prove the claim (7.3). For the rest of the proof we shall denote any of the cubes $Q_{1},\ldots,Q_{k}$ simply by $Q$. Denote by $\nu_{0}$ the resolution of $Q$. We claim that there exists a dyadic cube $\widetilde{Q}$ of the same resolution as $Q$ such that $\left\|{f}\right\|_{\widetilde{Q}}\leq\eta^{\prime}\alpha$ and $\rho(Q,\widetilde{Q})\leq C_{0}\,\delta^{\nu_{0}}$. We treat the cases where $\nu_{0}$ is equal to $2$ or $\geq 3$ separately. Suppose that $\nu_{0}=2$. Since $\rho\big{(}Q,A(\eta^{\prime}\alpha)^{c})\big{)}\leq\delta^{3}$, there exists a point $y$ in $A(\eta^{\prime}\alpha)^{c}$ such that $\rho(Q,y)\leq\delta^{3}$. Denote by $\widetilde{Q}$ the dyadic cube with resolution $2$ which contains $y$. Then $\rho(Q,\widetilde{Q})\leq\delta^{3}\leq C_{0}\delta^{2}$ and $\left\|{f}\right\|_{\widetilde{Q}}\leq\eta^{\prime}\alpha$, because $\widetilde{Q}\cap A(\eta^{\prime}\alpha)^{c}\not=\emptyset$. Now suppose that $\nu_{o}\geq 3$. Then the father $Q^{\sharp}$ of $Q$ contains a point $y$ in $A(\eta^{\prime}\alpha)^{c}$, for otherwise $Q^{\sharp}$ would be contained in $A(\eta^{\prime}\alpha)$, thereby contradicting the maximality of $Q$. Denote by $\widetilde{Q}$ the dyadic cube of resolution $\nu_{0}$ which contains $y$. Then $y$ is in $\widetilde{Q}\cap Q^{\sharp}$ and therefore $\widetilde{Q}\subset Q^{\sharp}$. Thus $\rho(Q,\widetilde{Q})\leq\,\mathrm{diam}(Q^{\sharp})\leq C_{1}\,\delta^{\nu_{0}-1}\leq C_{0}\,\delta^{\nu_{0}}$ and $\left\|{f}\right\|_{\widetilde{Q}}\leq\eta^{\prime}\alpha$, because $\widetilde{Q}\cap A(\eta^{\prime}\alpha)^{c}\not=\emptyset$. This completes the proof of the claim. To each point $y$ in $Q\cap A(\alpha)$ we associate a maximal dyadic cube $Q_{y}^{\prime}$ of resolution at least $2$ containing $y$ such that $\left\|{f}\right\|_{Q_{y}^{\prime}}>\alpha$. Denote $\mathcal{C}^{\prime}$ the collection of all these cubes. By Proposition 3.5 we may select mutually disjoint cubes $Q_{1}^{\prime},\ldots,Q_{k^{\prime}}^{\prime}$ in $\mathcal{C}^{\prime}$ such that (7.4) $\sum_{j=1}^{k^{\prime}}\mu(Q_{j}^{\prime})\geq\sigma\,\mu\bigl{(}Q\cap A(\alpha)\bigr{)}.$ Note also that $Q_{1}^{\prime},\ldots,Q_{k^{\prime}}^{\prime}$ are contained in $Q$. Denote by $B^{}$ a ball with centre at a point of $Q$ and radius $b^{\prime}\delta^{\nu_{0}}$ (recall that $b^{\prime}\geq 2C_{1}+C_{0})$. Then $B^{}$ contains both $Q$, whence the cubes $Q^{\prime}_{1},\ldots,Q^{\prime}_{k^{\prime}}$, and $\widetilde{Q}$. Hence (7.5) $\mu(B^{})\leq D\,\mu(Q)\qquad\hbox{and}\qquad\mu(B^{})\leq D\,\mu(\widetilde{Q})$ by the local doubling property. If $Q\cap A(\alpha)\cap S(\varepsilon\alpha)^{c}$ is nonempty, then (7.6) $\int_{B^{}}\left\|{f-f_{B^{}}}\right\|\,\text{\rm d}\mu\leq\varepsilon\,\alpha\,\mu(B^{}).$ Since $\widetilde{Q}\subset B^{}$ and $\left\|{f}\right\|_{\widetilde{Q}}\leq\eta^{\prime}\,\alpha$, $\mu(\widetilde{Q})\,\bigl{(}\left\|{f_{B^{}}}\right\|-\eta^{\prime}\,\alpha\bigr{)}\leq\int_{B^{}}\left\|{f-f_{B^{}}}\right\|\,\text{\rm d}\mu$ by the triangle inequality. Now (7.6) implies that (7.7) $\mu(\widetilde{Q})\,\bigl{(}\left\|{f_{B^{}}}\right\|-\eta^{\prime}\,\alpha\bigr{)}\leq\varepsilon\,\alpha\,\mu(B^{}).$ By a similar argument (7.8) $\bigl{(}\alpha-\left\|{f_{B^{}}}\right\|\bigr{)}\sum_{j=1}^{k^{\prime}}\mu(Q_{j}^{\prime})\,\leq\varepsilon\,\alpha\,\mu(B^{}).$ From (7.7) we see that $\left\|{f_{B^{}}}\right\|\leq\alpha\,\bigl{(}\eta^{\prime}+\varepsilon\,\mbox{\small$\displaystyle\frac{\mu(B^{})}{\mu(\widetilde{Q})}$}\bigr{)}$. By inserting this inequality in (7.8), we obtain that $\Bigl{(}1-\eta^{\prime}-\varepsilon\,\frac{\mu(B^{})}{\mu(\widetilde{Q})}\Bigr{)}\sum_{j=1}^{k^{\prime}}\mu(Q_{j}^{\prime})\leq\varepsilon\,\mu(B^{}),$ whence $\sigma\,\Bigl{(}1-\eta^{\prime}-\varepsilon\,\frac{\mu(B^{})}{\mu(\widetilde{Q})}\Bigr{)}\,\mu(Q\cap A(\alpha)\cap S(\varepsilon\alpha)^{c})\leq\varepsilon\,\mu(B^{}),$ by (7.4). Now, since $\varepsilon<(1-\eta^{\prime})/(2D)$, we may use (7.5) and conclude that $\,\mu(Q\cap A(\alpha)\cap S(\varepsilon\alpha)^{c})\leq\frac{2\varepsilon\,D}{\sigma\,(1-\eta^{\prime})}\,\mu(Q),$ as required. ∎ ###### Theorem 7.3. For each $p$ is in $(1,\infty)$ there exists a positive constant $C$ such that ${\\|{f^{\sharp,b^{\prime}}}\\|_{L^{p}(\mu)}}\geq C\,{\\|{f}\\|_{L^{p}(\mu)}}\qquad\forall f\in L^{p}(\mu),$ where $b^{\prime}=\max(b_{0},2C_{1}+C_{0})$ is as in the statement of Lemma 7.2. ###### Proof. Observe that it suffices to show that (7.9) ${\\|{f^{\sharp,b^{\prime}}}\\|_{L^{p}(\mu)}}\geq C\,{\\|{\mathcal{M}_{2}f}\\|_{L^{p}(\mu)}},$ because $\mathcal{M}_{2}f\geq\left\|{f}\right\|$ by the differentiation theorem of the integral, which is a standard consequence of Proposition 7.1. Let $\eta$ and $\eta^{\prime}$ be as in the statement of Lemma 7.2. By Lemma 7.2 $\displaystyle{\\|{\mathcal{M}_{2}f}\\|_{L^{p}(\mu)}}^{p}$ $\displaystyle=p\int_{0}^{\infty}\alpha^{p-1}\,\mu\bigl{(}A(\alpha)\bigr{)}\,\text{\rm d}\alpha$ $\displaystyle=p\int_{0}^{\infty}\alpha^{p-1}\,\mu\bigl{(}A(\alpha)\cap S(\varepsilon\alpha)^{c}\bigr{)}\,\text{\rm d}\alpha+p\int_{0}^{\infty}\alpha^{p-1}\,\mu\bigl{(}A(\alpha)\cap S(\varepsilon\alpha)\bigr{)}\,\text{\rm d}\alpha$ $\displaystyle\leq p\,\eta\int_{0}^{\infty}\alpha^{p-1}\,\mu\bigl{(}A(\eta^{\prime}\alpha)\bigr{)}\,\text{\rm d}\alpha+p\int_{0}^{\infty}\alpha^{p-1}\,\mu\bigl{(}S(\varepsilon\alpha)\bigr{)}\,\text{\rm d}\alpha$ $\displaystyle=p\,\eta\,\eta^{\prime-p}\int_{0}^{\infty}\beta^{p-1}\,\mu\bigl{(}A(\beta)\bigr{)}\,\text{\rm d}\beta+p\,\varepsilon^{-p}\int_{0}^{\infty}\beta^{p-1}\,\mu\bigl{(}S(\beta)\bigr{)}\,\text{\rm d}\beta$ $\displaystyle\leq\eta\,\eta^{\prime-p}\,{\\|{\mathcal{M}_{2}f}\\|_{L^{p}(\mu)}}^{p}+\varepsilon^{-p}\,{\\|{f^{\sharp,b^{\prime}}}\\|_{L^{p}(\mu)}}^{p}.$ Now, for a given $p$, we choose $\eta^{\prime}$ such that $\eta^{\prime p}=1-\sigma/4$, and then we choose $\varepsilon$ small enough so that $\eta\leq 1-\sigma/2$. Therefore $\eta\,\eta^{\prime-p}<1$ and (7.9) follows. ∎ As a consequence of the relative distributional inequality proved in Theorem 7.3 we establish an interpolation result for analytic families of operators. We need the following notation. In the following, when $X$ and $Y$ are Banach spaces, and $\theta$ is in $(0,1)$, we denote by $(X,Y)_{[\theta]}$ the complex interpolation space between $X$ and $Y$ with parameter $\theta$. ###### Theorem 7.4. Suppose that $\theta$ is in $(0,1)$. The following hold: 1. (i) if $p_{\theta}$ is $2/(1-\theta)$, then $\bigl{(}L^{2}(\mu),BMO(\mu)\bigr{)}_{[\theta]}=L^{p_{\theta}}(\mu)$; 2. (ii) if $p_{\theta}$ is $2/(2-\theta)$, then $\bigl{(}H^{1}(\mu),L^{2}(\mu)\bigr{)}_{[\theta]}=L^{p_{\theta}}(\mu)$. ###### Proof. The proof of (i) is an adaptation of the proof of [FS, Cor. 1, p.156]. We omit the details. Now (ii) follows from (i) and the duality theorem [BL, Corollary 4.5.2]. ∎ ###### Theorem 7.5. Denote by $S$ the strip $\\{z\in\mathbb{C}:\operatorname{\mathrm{Re}}z\in(0,1)\\}$. Suppose that $\\{\mathcal{T}_{z}\\}_{z\in\bar{S}}$ is a family of uniformly bounded operators on $L^{2}(\mu)$ such that $z\mapsto\int_{\mathbb{R}^{d}}\mathcal{T}_{z}f\,g\,\text{\rm d}\mu$ is holomorphic in $S$ and continuous in $\bar{S}$ for all functions $f$ and $g$ in $L^{2}(\mu)$. Further, assume that there exists a constant $A$ such that $\|\\!\|\\!\|{\mathcal{T}_{is}}\|\\!\|\\!\|_{L^{2}(\mu)}\leq A\qquad\hbox{and}\qquad\|\\!\|\\!\|{\mathcal{T}_{1+is}}\|\\!\|\\!\|_{L^{\infty}(\mu);BMO({\mu})}\leq A.$ Then for every $\theta$ in $(0,1)$ the operator $\mathcal{T}_{\theta}$ is bounded on $L^{p_{\theta}}({\mu})$, where $p_{\theta}=2/(1-\theta)$, and $\|\\!\|\\!\|{\mathcal{T}_{\theta}}\|\\!\|\\!\|_{L^{p_{\theta}}({\mu})}\leq A_{\theta},$ where $A_{\theta}$ depends only on $A$ and on $\theta$. ###### Proof. This follows directly from (i) of the previous theorem and [CJ, Thm 1]. Alternatively, we may follow the line of the proof of the classical case (see, for instance, [St2, Thm 4, p.175], or [FS]). ∎ ## 8\. Singular integrals In this section we develop a theory of singular integral operators acting on $L^{p}(\mu)$ spaces. Preliminarly, we observe the following. Recently, M. Bownik [Bow], following up earlier work of Y. Meyer, produced in the classical Euclidean setting an example of an operator $\mathcal{T}_{B}$ defined on $(1,\infty)$-atoms with $\sup\\{{\\|{\mathcal{T}_{B}a}\\|_{L^{1}(\lambda)}}:\hbox{$a$ is a $(1,\infty)$-atom}\\}<\infty,$ that does not extend to a bounded operator from $H^{1}(\lambda)$ to $L^{1}(\lambda)$: here $\lambda$ denotes the Lebesgue measure on $\mathbb{R}^{n}$. The problem of giving sufficient conditions for an operator uniformly bounded on atoms to extend to a bounded operator from $H^{1}(\lambda)$ to $L^{1}(\lambda)$ has been considered independently in [MSV, YZ]. The paper [YZ] and most of [MSV] focuse on the Euclidean case. However, in the last part of [MSV] more general settings are considered. In particular, suppose that $(M,\rho,\mu)$ is a $\sigma$-finite metric measure space with properties (LDP), (I) and (AMP). Then the following holds. ###### Proposition 8.1. Suppose that $q$ is in $(1,\infty)$, and that $\mathcal{T}$ is a linear operator defined on finite linear combinations of $(1,q)$-atoms, satisfying $\sup\\{{\\|{\mathcal{T}a}\\|_{L^{1}(\mu)}}:\hbox{\emph{$a$ is a $(1,q)$-atom}}\\}<\infty.$ The following hold: 1. (i) $\mathcal{T}$ extends to a bounded operator $\widetilde{\mathcal{T}}$ from $H^{1}(\mu)$ to $L^{1}(\mu)$ and the transpose operator $\mathcal{T}^{}$ extends to a bounded operator $(\mathcal{T}^{})\widetilde{\phantom{a}}$ from $L^{\infty}(\mu)$ to $BMO(\mu)$; 2. (ii) if $\mathcal{T}$ is bounded on $L^{2}(\mu)$, then $\mathcal{T}$ and $\widetilde{\mathcal{T}}$ are consistent operators on $H^{1}(\mu)\cap L^{2}(\mu)$. ###### Proof. The result [MSV, Thm 4.1 and Prop. 4.2] is stated for spaces of homogeneous type. However, the proof extends _verbatim_ to our setting. ∎ Now we assume that $\mathcal{T}$ is bounded on $L^{2}(\mu)$ and that there exists a locally integrable function $k$ off the diagonal in $M\times M$ such that for every function $f$ with support of finite measure $\mathcal{T}f(x)=\int_{M}k(x,y)\,f(y)\,\text{\rm d}\mu(y)\qquad\forall x\notin\mathrm{supp}\,f.$ We refer to $k$ as to the kernel of $\mathcal{T}$. A straightforward computation shows that the kernel $k^{}$ of the (Hilbert space) adjoint $\mathcal{T}^{}$ of $\mathcal{T}$ is related to the kernel $k$ of $\mathcal{T}$ by the formula (8.1) $k^{}(y,x)=\overline{k(x,y)}.$ In particular, if $\mathcal{T}$ is self adjoint on $L^{2}(\mu)$, then (8.2) $k(y,x)=\overline{k(x,y)}.$ The next theorem is a version in our case of a classical result which holds on spaces of homogeneous type. _Mutatis mutandis_ , its proof is similar to the proof in the classical case. However, we include a sketch of the proof for the reader’s convenience. See also [MM] for a detailed proof of the analogous result in the Gaussian case. ###### Theorem 8.2. Suppose that $b$ is in $\mathbb{R}^{+}$ and $b>R_{0}/(1-\beta)$, where $R_{0}$ and $\beta$ appear in the definition of property (AMP). Suppose that $\mathcal{T}$ is a bounded operator on $L^{2}(\mu)$ and that its kernel $k$ is locally integrable off the diagonal of $M\times M$. Define $\upsilon_{k}$ and $\nu_{k}$ by $\upsilon_{k}=\sup_{B\in\mathcal{B}_{b}}\sup_{x,x^{\prime}\in B}\int_{(2B)^{c}}\left\|{k(x,y)-k(x^{\prime},y)}\right\|\,\text{\rm d}\mu(y),$ and $\nu_{k}=\sup_{B\in\mathcal{B}_{b}}\sup_{y,y^{\prime}\in B}\int_{(2B)^{c}}\left\|{k(x,y)-k(x,y^{\prime})}\right\|\,\text{\rm d}{\mu}(x).$ The following hold: 1. (i) if $\nu_{k}$ is finite, then $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $(1,2]$ and from $H^{1}({\mu})$ to $L^{1}(\mu)$. Furthermore, there exists a constant $C$ such that $\|\\!\|\\!\|{\mathcal{T}}\|\\!\|\\!\|_{H^{1}({\mu});L^{1}(\mu)}\leq C\,\bigl{(}\nu_{k}+\|\\!\|\\!\|{\mathcal{T}}\|\\!\|\\!\|_{L^{2}(\mu)}\bigr{)};$ 2. (ii) if $\upsilon_{k}$ is finite, then $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $[2,\infty)$ and from $L^{\infty}(\mu)$ to $BMO({\mu})$. Furthermore, there exists a constant $C$ such that $\|\\!\|\\!\|{\mathcal{T}}\|\\!\|\\!\|_{L^{\infty}(\mu);BMO({\mu})}\leq C\,\bigl{(}\upsilon_{k}+\|\\!\|\\!\|{\mathcal{T}}\|\\!\|\\!\|_{L^{2}(\mu)}\bigr{)};$ 3. (iii) if $\mathcal{T}$ is self adjoint on $L^{2}(\mu)$ and $\nu_{k}$ is finite, then $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $(1,\infty)$, from $H^{1}({\mu})$ to $L^{1}(\mu)$ and from $L^{\infty}(\mu)$ to $BMO({\mu})$. ###### Proof. First we prove (i). In view of Proposition 8.1 it suffices to show that $\mathcal{T}$ maps $(1,2)$-atoms in $H_{1}^{1}(\mu)$ uniformly into $L^{1}(\mu)$. This is done exactly as in the classical case, except that we need to consider only atoms supported in balls of $\mathcal{B}_{b}$. Then $\mathcal{T}$ maps $H^{1}(\mu)$ into $L^{1}(\mu)$, and, by interpolation, on $L^{p}(\mu)$ for all $p$ in $(1,2)$. Next we prove (ii). Since the kernel $k^{}(x,y)$ of the (Hilbert space) adjoint $\mathcal{T}^{}$ of $\mathcal{T}$ is $\overline{{k(y,x)}}$, by (i) $\upsilon_{k^{}}=\nu_{k}$ is finite and the operator $\mathcal{T}^{}$ is bounded from $H^{1}(\mu)$ to $L^{1}(\mu)$. Hence $\mathcal{T}$ is bounded from $L^{\infty}({\mu})$ to $BMO({\mu})$. Moreover $\|\\!\|\\!\|{\mathcal{T}^{}}\|\\!\|\\!\|_{L^{\infty}(\mu);BMO({\mu})}\leq C\,\bigl{(}\nu_{k}+\|\\!\|\\!\|{\mathcal{T}}\|\\!\|\\!\|_{L^{2}(\mu)}\bigr{)}.$ By interpolation $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $(2,\infty)$, Finally, we prove (iii). By (ii), $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $(1,2)$ and from $H^{1}({\mu})$ to $L^{1}(\mu)$. By (8.2) also $\upsilon_{k}$ is finite. Hence, by (i), $\mathcal{T}$ extends to a bounded operator on $L^{p}(\mu)$ for all $p$ in $[2,\infty)$ and from $L^{\infty}(\mu)$ to $BMO({\mu})$, thereby concluding the proof of (iii) and of the theorem. ∎ ###### Remark 8.3. It is worth observing that in the case where $M$ is a Riemannian manifold and the kernel $k$ is “regular”, then the condition $\upsilon_{k}<\infty$ of Theorem 8.2 (i) may be replaced by the condition $\upsilon_{k}^{\prime}<\infty$, where (8.3) $\upsilon^{\prime}_{k}=\sup_{B\in\mathcal{B}_{b}}\,r_{B}\sup_{x\in B}\int_{(2B)^{c}}\left\|{\operatorname{\rm{grad}}_{x}k(x,y)}\right\|\,\text{\rm d}\mu(y).$ Similarly, the condition $\nu_{k}<\infty$ of Theorem 8.2 (ii) may be replaced by the condition $\nu_{k}^{\prime}<\infty$, where (8.4) $\nu^{\prime}_{m}=\sup_{B\in\mathcal{B}_{b}}\,r_{B}\sup_{y\in B}\int_{(2B)^{c}}\left\|{\operatorname{\rm{grad}}_{y}k(x,y)}\right\|\,\text{\rm d}{\mu}(x).$ Indeed, by the mean value theorem we see that the condition $\sup_{B\in\mathcal{B}_{1}}\sup_{x,x^{\prime}\in B}\left\|{x-x^{\prime}}\right\|\,\int_{(2B)^{c}}\left\|{\operatorname{\rm{grad}}_{x}k(x,y)}\right\|\,\text{\rm d}\mu(y)<\infty$ implies the condition $\upsilon_{k}<\infty$ of the theorem. Since $\left\|{x-x^{\prime}}\right\|<2\,r_{B}$, (8.3) follows. We note also that formula (8.2) imply that if $\mathcal{T}$ is self adjoint, then $\upsilon^{\prime}_{k}<\infty$ holds if and only if $\nu^{\prime}_{k}<\infty$ does. ## 9\. Cheeger’s isoperimetric constant and property (I) In this section we show that the theory developed in the previous sections may be applied to an interesting class of complete noncompact Riemannian manifolds. First we recall that if $(M,\rho)$ is a complete Riemannian manifold with Ricci curvature bounded from below, then $M$ is a locally doubling metric space with respect to the Riemannian measure and the geodesic distance. The proof of this fact is a direct consequence of M. Gromov’s variant [Gr] of R.L. Bishop’s comparison theorem (see, for instance, [BC]). It is natural to investigate the dependence of property (I) on other geometric or analytic properties of $M$. Denote by $b(M)$ the bottom of the spectrum of $M$, defined by $b(M)=\inf_{f\neq 0}\,\frac{\int_{M}\left\|{\operatorname{\rm{grad}}f}\right\|^{2}\,\text{\rm d}V}{\int_{M}f^{2}\,\text{\rm d}V},$ where $V$ denotes the Riemannian measure of $M$ and $f$ runs over all sufficiently smooth functions with compact support. We denote by $h(M)$ the Cheeger isoperimetric constant of $M$ defined by (9.1) $h(M)=\inf\frac{\sigma(\partial A)}{V(A)},$ where the infimum runs over all bounded open subsets $A$ with smooth boundary $\partial A$ and where $\sigma$ denotes the $(d-1)$-dimensional measure. Cheeger proved that (9.2) $b(M)\geq h(M)^{2}/4.$ In this section we shall relate $b(M)$ and $h(M)$ to the isoperimetric constant $I_{M}$ defined in Section 2. First we need to show that in Cheeger’s isoperimetric inequality $\sigma(\partial A)\geq h(M)\,V(A)$, we may replace the bounded open set $A$ with smooth boundary by any set $E$ of finite measure and the $(d-1)$-dimensional measure $\sigma$ of the boundary by the perimeter of $E$. The definition of perimeter of a set in a Riemannian manifold mimics closely the definition in the Euclidean setting [EG, MPPP]. If $U$ is an open subset of $M$ we shall denote by $\Lambda^{k}_{c}(U)$ the space of smooth $k$-forms with compact support contained in $U$. The divergence is the formal adjoint of the the exterior derivative d, i.e. the operator $\operatorname{\rm{div}}$ mapping $k+1$-forms to $k$-forms defined by (9.3) $\int_{M}\langle\operatorname{\rm{div}}\omega,\eta\rangle_{x}\,\text{\rm d}V(x)=-\int_{M}\langle\omega,\,\text{\rm d}\eta\rangle_{x}\,\text{\rm d}V(x)$ for all smooth $k+1$-forms $\omega$ and all smooth $k$-forms $\eta$ with compact support. Given a real valued function $f$ in $L^{1}(M)$, the variation of $f$ in $U$ is $\operatorname{\rm{Var}}(f,U)=\sup{\left\\{\int_{M}f\,(\operatorname{\rm{div}}\omega)\,\text{\rm d}V:\omega\in\Lambda^{1}_{c}(U),{\\|{\omega}\\|_{\infty}}\leq 1\right\\}}.$ We say that $f$ has bounded variation in $U$ if $\operatorname{\rm{Var}}(f,U)<\infty$. We shall denote by $BV(U)$ the space of function of bounded variation in $U$. As in the Euclidean case, if $f$ is in $BV(M)$, then the map $U\mapsto\operatorname{\rm{Var}}(f,U)$ extends to a finite Borel measure on $M$. A measurable set $E\subset M$ has finite perimeter if its indicator function ${\bf 1}_{E}$ is in $BV(M)$. If $U$ is a Borel set the perimeter of $E$ in $U$ is $P(E,U)=\operatorname{\rm{Var}}({\bf 1}_{E},U).$ Since the manifold $M$ is complete, by [He, Theorem 2.7] the Sobolev space $H^{1,1}(M)$ is the completion of the space $C^{\infty}_{c}(M)$ of smooth functions with compact support on $M$ with respect to the norm ${\\|{f}\\|_{H^{1,1}}}={\\|{f}\\|_{1}}+\int_{M}\left\|{\operatorname{\rm{grad}}f}\right\|\,\text{\rm d}V.$ It is an easy matter to show that $H^{1,1}(M)\subset BV(M)$ and $\operatorname{\rm{Var}}(f,M)=\int_{M}\left\|{\operatorname{\rm{grad}}f}\right\|\,\text{\rm d}V\qquad\forall f\in H^{1,1}(M).$ Note also that the space $\operatorname{\rm{Lip}}_{c}(M)$ of Lipschitz functions with compact support on $M$ is contained in $H^{1,1}(M)$, by [He, Lemma 2.5]. In the Euclidean setting it is well known that $BV$ functions may be approximated in variation by smooth functions in $L^{1}$ (see, for instance [EG, Theor. 3.9]). The same result holds in the Riemannian setting [MPPP, Prop. 1.4]. ###### Lemma 9.1. For every $f$ in $BV(M)$ there exists a sequence $(f_{n})$ in $C^{\infty}_{c}(M)$ which converges to $f$ in $L^{1}(M)$ and such that $\operatorname{\rm{Var}}(f,M)=\lim_{n\to\infty}\operatorname{\rm{Var}}(f_{n},M).$ Now we can prove that Cheeger’s isoperimetric inequality for smooth compact hypersurfaces implies an analogous inequality for the perimeters of sets of finite measure. ###### Proposition 9.2. Suppose that $M$ is a complete Riemannian manifold. If $h(M)>0$ then for every measurable set $E$ $P(E,M)\geq h(M)\,V(E).$ ###### Proof. It is well known that Cheeger’s isoperimetric inequality for smooth submanifolds is equivalent to the Sobolev inequality (9.4) $\operatorname{\rm{Var}}(f,M)=\int_{M}\left\|{\operatorname{\rm{grad}}f}\right\|\,\text{\rm d}V\geq h(M)\int_{M}\left\|{f}\right\|\,\text{\rm d}V$ for all real valued functions in $C^{1}_{c}(M)$ [Cha]. Suppose that $E$ is a measurable set of finite perimeter. By Lemma 9.1 there exists a sequence $(f_{n})$ of functions in $C^{1}_{c}(M)$ such that $f_{n}\to{\bf 1}_{E}$ in $L^{1}(M)$ and $\operatorname{\rm{Var}}(f_{n},M)\to\operatorname{\rm{Var}}({\bf 1}_{E},M)=P(E,M)$. Hence the desired conclusion follows from (9.4). ∎ The following lemma is the coarea formula for functions in $H^{1,1}(M)$. The proof uses the density of $C^{1}_{c}(M)$ in $H^{1,1}(M)$ and mimics closely the argument in the Euclidean setting [EG]. ###### Lemma 9.3. Suppose that $M$ is a complete Riemannian manifold. Assume that $f\in H^{1,1}(M)$. Then for every open subset $U$ of $M$ the sets $A(t):={\left\\{x\in U:f(x)>t\right\\}}$ have finite perimeter for a.e. $t$ in $\mathbb{R}$ and $\int_{U}\left\|{\operatorname{\rm{grad}}f}\right\|\,\text{\rm d}V=\int_{\mathbb{R}}P\big{(}A(t),U\big{)}\,\text{\rm d}t.$ ###### Remark 9.4. We observe en passant that using Lemma 9.1, one can prove the following coarea formula for functions in $BV(M)$ $\operatorname{\rm{Var}}(f,M)=\int_{\mathbb{R}}P\big{(}A(t),M\big{)}\,\text{\rm d}t.$ Now we are ready to prove the equivalence of property (I) and Cheeger’s inequality. We recall that the constant $I_{M}$ is defined in Section 2. ###### Theorem 9.5. Suppose that $M$ is a complete Riemannian manifold. The following hold: 1. (i) $M$ possesses property (I) if and only if $h(M)>0$. Furthermore $I_{M}=h(M)$; 2. (ii) if the Ricci curvature of $M$ is bounded from below, then $M$ possesses property (I) if and only if $b(M)>0$. ###### Proof. First we prove (i). Assume that $M$ possesses property (I) and denote by $A$ a bounded open subset of $M$ with smooth boundary. By Proposition 3.1 $\frac{V(A_{\kappa})}{\kappa}\geq\,\frac{1-\mathrm{e}^{-I_{M}t}}{t}\,V(A)\qquad\forall t\in\mathbb{R}^{+}.$ Hence $\liminf_{\kappa\to 0}\frac{V(A_{\kappa})}{\kappa}\geq I_{M}\,V(A).$ Since the limit in the left hand side is the lower inner Minkowski content of $\partial A$, which coincides with $\sigma(\partial A)$ because $\partial A$ is smooth, we have proved that Cheeger’s isoperimetric constant $h(M)$ is at least $I_{M}$. To prove the converse, assume that $h(M)>0$ and let $A$ be a bounded open set in $M$. Since the manifold $M$ is complete, the function $f$ defined by $f(x)=\rho(x,A^{c})$ (here $A^{c}$ denotes the complementary set of $A$ in $M$) is Lipschitz and $\left\|{\operatorname{\rm{grad}}f}\right\|=1$ almost everywhere. Recall that for each $t$ in $\mathbb{R}$ we denote by $A^{t}$ the set ${\left\\{x\in A:f(x)>t\right\\}}$. Notice that $P\big{(}A^{s},A^{t}\big{)}=\begin{cases}P\big{(}A^{s},M\big{)}&{\text{ if}}\ s>t;\\\ 0&{\text{ if}}\ s<t.\\\ \end{cases}$ Thus, by the coarea formula $V\big{(}A^{t}\big{)}=\int_{A^{t}}\left\|{\operatorname{\rm{grad}}f}\right\|\,\text{\rm d}V=\int_{t}^{\infty}P\big{(}A^{s},M\big{)}\,\text{\rm d}s.$ Hence, by Proposition 9.2 $\frac{\,\text{\rm d}}{\,\text{\rm d}t}V\big{(}A^{t}\big{)}=-P\big{(}A^{t},M\big{)}\leq-h(M)\,V\big{(}A^{t}\big{)}\qquad\text{for a.e.}\ t\in\mathbb{R}.$ This differential inequality implies that $V\big{(}A^{t}\big{)}\leq e^{-h(M)\,t}V(A)$ for all $t>0$, i.e. $\displaystyle V(A_{t})$ $\displaystyle=V(A)-V\big{(}A^{t}\big{)}$ $\displaystyle\geq(1-e^{-h(M)\,t})\,V(A)$ $\displaystyle\geq h(M)\,t\,V(A)\qquad\forall t\in(0,1).$ Thus $M$ possesses property (I), and $h(M)\leq I_{M}$, as required to conclude the proof of (i). To prove (ii) we recall that if $M$ has Ricci curvature bounded below by $-K$, for some $K\geq 0$ then (9.5) $b(M)\leq C\big{(}\sqrt{K}\ h(M)+h(M)^{2}\big{)},$ where $C$ is a constant which depends only on the dimension of $M$ [Bu, Le]. This inequality, together with Cheeger’s inequality (9.2), shows that the constants $h(M)$ and $b(M)$ are equivalent. The required conclusion follows directly from (i). ∎ ###### Remark 9.6. We remark that property (I) is invariant under quasi-isometries. Indeed, the fact that $h(M)$ is positive is invariant under quasi-isometries [Cha, Remark V.2.2]. ###### Remark 9.7. Suppose that $M$ is a complete Riemannian manifold. If $b(M)=0$, then $M$ has not property (I) by Theorem 9.5 (i). Observe also that if $M$ has Ricci curvature bounded from below, and a spectral gap, then $M$ has property (I). In particular noncompact Riemannian symmetric spaces and Damek-Ricci spaces have property (I). ## 10\. Applications In this section we illustrate some applications of the theory developed in the previous sections. Other applications will appear in [CMM]. The first application we consider is to spectral multipliers on certain Riemannian manifolds. Suppose that $M$ is a complete Riemannian manifold with positive injectivity radius $\mathrm{inj}(M)$, Ricci curvature bounded from below, and positive Cheeger isoperimetric constant $h(M)$. By Cheeger’s inequality the bottom $b(M)$ of the $L^{2}(M)$ spectrum of the Laplace–Beltrami operator $\mathcal{L}$ on $M$ is positive. We denote by $\mu$ the Riemannian measure on $M$. As shown in Section 9, under these assumptions $M$ possesses property (I). Furthermore, there exist constants $\alpha$, $\alpha^{\prime}$, $\beta$, $\beta^{\prime}$, $C_{1}$ and $C_{2}$ such that $C_{1}\,(1+r^{2})^{\alpha^{\prime}/2}\,\mathrm{e}^{\beta^{\prime}\,r}\leq\mu\bigl{(}B(p,r)\bigr{)}\leq C_{2}\,(1+r^{2})^{\alpha/2}\,\mathrm{e}^{\beta\,r}\qquad\forall r\in\mathbb{R}^{+}.$ We say that $M$ has ($N$-)bounded geometry provided that the derivatives of the Riemann tensor up to the order $N$ are uniformly bounded. Clearly if $M$ has $N$-bounded geometry, then the Ricci curvature of $M$ is bounded from below. For each $\sigma$ in $\mathbb{R}^{+}$ we denote by $\mathbf{S}_{\sigma}$ the strip $\\{\zeta\in\mathbb{C}:\left\|{\operatorname{\mathrm{Im}}\zeta}\right\|<\sigma\\}$. ###### Definition 10.1. Suppose that $\kappa$ is a positive integer and that $\sigma$ is in $\mathbb{R}^{+}$. The space $H^{\infty}(\mathbf{S}_{\sigma};\kappa)$ is the vector space of all functions $f$ in $H^{\infty}(\mathbf{S}_{\sigma})$ for which there exists a positive constant $C$ such that for each $\varepsilon$ in $\\{-1,1\\}$ (10.1) $\left\|{D^{k}f(\zeta)}\right\|\leq{C}\,(1+\left\|{\zeta}\right\|)^{-k}\qquad\forall k\in\\{0,1,\ldots,\kappa\\}\quad\forall\zeta\in\mathbf{S}_{\sigma}.$ If (10.1) holds, we say that $f$ satisfies a _Mihlin–Hörmander condition of order $\kappa$_ on the strip $\mathbf{S}_{\sigma}$. We endow $H^{\infty}(\mathbf{S}_{\sigma};\kappa)$ with the norm ${\\|{f}\\|_{\sigma;\kappa}}=\max_{k\in\\{0,1,\ldots,\kappa\\}}\sup_{\zeta\in\mathbf{S}_{\sigma}}\,(1+\left\|{\zeta}\right\|)^{k}\,\left\|{D^{k}f(\zeta)}\right\|.$ The following result complements a celebrated result of Taylor [Ta, Thm 1]. ###### Theorem 10.2. Suppose that $M$ is an $n$ dimensional complete Riemannian manifold with $N$-bounded geometry, where $N$ is an integer $\geq n/2+1$. Assume that the injectivity radius $\mathrm{inj}(M)$ and the bottom $b(M)$ of the $L^{2}(M)$ spectrum of $M$ are positive. Suppose that $f$ is in $H^{\infty}(\mathbf{S}_{\sigma};\kappa)$, where $\sigma\geq\beta/2$ and $\kappa>\max(\alpha/2+1,n/2+1)$. Then $f(\mathcal{L}_{b}^{1/2})$ extends to a bounded operator from $H^{1}(\mu)$, from $L^{\infty}(M)$ to $BMO(\mu)$ and on $L^{p}(M)$ for all $p$ in $(1,\infty)$. ###### Proof. Denote by $\mathcal{L}_{b}$ the operator $\mathcal{L}-b\,\mathcal{I}$, formally defined on $L^{2}(\mu)$. The strategy of the proof of [Ta, Thm 1] is to decompose the operator $f(\mathcal{L}_{b}^{1/2})$ as the sum of two operators, $f_{0}(\mathcal{L}_{b}^{1/2})$ and $f_{\infty}(\mathcal{L}_{b}^{1/2})$, where $f_{\infty}(\mathcal{L}_{b}^{1/2})$ is bounded on $L^{1}(\mu)$ and $L^{\infty}(\mu)$ and $f_{0}(\mathcal{L}_{b}^{1/2})$ is of weak type $1$. To prove the latter, Taylor [Ta] and Cheeger, Gromov and Taylor [CGT] prove that the integral kernel of $f_{0}(\mathcal{L}_{b}^{1/2})$, which is compactly supported, satisfies a Hörmander type integral condition. Taking this for granted, by Theorem 8.2, the operator $f_{0}(\mathcal{L}_{b}^{1/2})$ is bounded from $H^{1}(\mu)$ to $L^{1}(\mu)$, and from $L^{\infty}(\mu)$ to $BMO(\mu)$. Therefore the same is true for $f(\mathcal{L}_{b}^{1/2})$. The boundedness of $f(\mathcal{L}_{b}^{1/2})$ on $L^{p}(\mu)$ the follows by interpolation. ∎ Note that this result applies to Riemannian symmetric spaces of the noncompact type, and to Damek–Ricci spaces. In the case where $M$ is a symmetric space of the noncompact type and real rank $>1$, J.Ph. Anker [A] extended Taylor’s result [Ta, Thm 1] to certain multiplier operators for the spherical Fourier transform. Suppose that $G$ is a noncompact semisimple Lie group with finite centre, and denote by $K$ a maximal compact subgroup thereof, and by $X$ the associated Riemannian symmetric space of the noncompact type $G/K$ (the $G$ invariant metric on $X$ is induced by the Killing form of $G$). Denote by $\mu$ a $G$-invariant measure on $X$. We complement Anker’s result by showing that (some) of the multiplier operators he considers satisfy natural $H^{1}(\mu)$-$L^{1}(\mu)$ and $L^{\infty}(\mu)$-$BMO(\mu)$ estimates. Suppose that $G=KAN$ is an Iwasawa decomposition of $G$, where $A$ is abelian and $N$ is nilpotent. Denote by $\mathfrak{g}$ and $\mathfrak{a}$ the Lie algebras of $G$ and $A$ respectively, and by $\rho$ the half sum of the positive roots of $(\mathfrak{g},\mathfrak{a})$ with multiplicities. Denote by $\mathfrak{a}^{}$ the dual of $\mathfrak{a}$. An important rôle in what follows is played by a certain tube $\mathbf{T}$ in the complexified dual $\mathfrak{a}_{\mathbb{C}}^{}$ of $\mathfrak{a}$. To define $\mathbf{T}$, denote by $\mathbf{W}$ the convex hull of the vectors $\\{w\cdot\rho:w\in W\\}$ in $\mathfrak{a}^{}$, where $W$ denotes the Weyl group of $G$. Then define $\mathbf{T}=\mathfrak{a}^{}+i\mathbf{W}$. ###### Definition 10.3. Suppose that $\kappa$ is a positive integer. The space $H^{\infty}(\mathbf{T};\kappa)$ is the vector space of all Weyl invariant bounded holomorphic functions $m$ on $\mathbf{T}$ for which there exists a positive constant $C$ such that for every multiindex $\beta$ with $\left\|{\beta}\right\|\leq\kappa$ (10.2) $\left\|{D^{\beta}m(\zeta)}\right\|\leq{C}\,(1+\left\|{\zeta}\right\|)^{-\left\|{\beta}\right\|}\qquad\forall\zeta\in\mathbf{T}.$ If (10.2) holds, we say that $f$ satisfies a _Mihlin–Hörmander condition of order $\kappa$_ on the tube $\mathbf{T}$. We endow $H^{\infty}(\mathbf{T};\kappa)$ with the norm ${\\|{f}\\|_{\kappa}}=\max_{\left\|{\beta}\right\|\leq\kappa}\sup_{\zeta\in\mathbf{T}}\,(1+\left\|{\zeta}\right\|)^{\left\|{\beta}\right\|}\,\left\|{D^{\beta}f(\zeta)}\right\|.$ Suppose that $k$ is a $K$-bi-invariant distribution on $G$, and denote by $m$ its spherical Fourier transform: $m$ may be thought of as a distribution on $\mathfrak{a}^{}$. If $m$ is bounded on $\mathfrak{a}^{}$, then the convolution operator $\mathcal{T}_{k}$, defined by $\mathcal{T}_{k}f=fk\qquad\forall f\in C_{c}^{\infty}(G)$ extends to a bounded operator on $L^{2}(\mu)$. ###### Theorem 10.4. Suppose that $\kappa>(\dim X)/2+1$. Suppose that $k$ is a $K$-bi-invariant distribution such that its spherical transform $m$ is in $H^{\infty}(\mathbf{T};\kappa)$. Then the operator $\mathcal{T}_{k}$ extends to a bounded operator from $H^{1}(\mu)$ to $L^{1}(\mu)$ and from $L^{\infty}(\mu)$ to $BMO(\mu)$. ###### Proof. Denote by $\psi$ a $K$-bi-invariant smooth function on $G$ with compact support which is identically $1$ in a neighbourhood of the identity, and define the distributions $k_{0}$ and $k_{\infty}$ by $k_{0}=\psi\,k\qquad\hbox{and}\qquad k_{\infty}=(1-\psi)\,k.$ Anker [A, Thm 1] shows that $k_{\infty}$ is, in fact, a function in $L^{1}(G)$. Therefore, the operator $\mathcal{T}_{k_{\infty}}$, defined by $\mathcal{T}_{k_{\infty}}f=f\mapsto fk_{\infty}$ extends to a bounded operator on $L^{1}(\mu)$ and on $L^{\infty}(\mu)$, and _a fortiori_ to a bounded operator from $H^{1}(\mu)$ to $L^{1}(\mu)$ and from $L^{\infty}(\mu)$ to $BMO(\mu)$. Furthermore, Anker proves that $k_{0}$ is locally integrable off the origin, and satisfies the following Hörmander type integral inequality $\sup_{B\in\mathcal{B}_{1}}\sup_{y\in B}\int_{(2B)^{c}}\left\|{k_{0}(y^{-1}x)-k_{0}(x)}\right\|\,\text{\rm d}\mu(x)<\infty.$ Define the operator $\mathcal{T}_{k_{0}}$ by $\mathcal{T}_{k_{0}}f=fk_{0}$. Denote by $\Delta$ the diagonal in $X\times X$, and define the locally integrable function $t$ on $X\times X\setminus\Delta$ by $t(x,y)=k_{0}(y^{-1}x).$ It is straightforward to check that $t$ is the kernel (see the definition at the beginning of Section 8) of the operator $\mathcal{T}_{k_{0}}$, and that $t$ satisfies conditions $\nu_{t}<\infty$ and $\upsilon_{t}<\infty$. By Theorem 8.2 (iii) the operator $\mathcal{T}_{k_{0}}$ extends to a bounded operator from $H^{1}(\mu)$ to $L^{1}(\mu)$ and from $L^{\infty}(\mu)$ to $BMO(\mu)$. Since $\mathcal{T}_{k}=\mathcal{T}_{k_{0}}+\mathcal{T}_{k_{\infty}}$, the required boundedness properties of $\mathcal{T}_{k}$ follow directly from those of $\mathcal{T}_{k_{0}}$ and $\mathcal{T}_{k_{\infty}}$. ∎ Our last application is to the boundedness of Riesz transforms. This is a very fashionable and interesting subject: see [CD, ACDH] for recent results on manifolds, and the references therein for less recent results. Suppose that $M$ is a complete Riemannian manifolds satisfying the following assumptions: the Riemannian measure $\mu$ is locally doubling, and the following scaled local Poincaré inequality holds: for every positive $b$ there exists a constant $C$ such that for every $B$ in $\mathcal{B}_{b}$ and for every $f$ in $C^{\infty}\bigl{(}2B\bigr{)}$ $\int_{B}\left\|{f-f_{B}}\right\|^{2}\,\text{\rm d}\mu\leq C\,r^{2}\int_{2B}\left\|{\nabla f}\right\|^{2}\,\text{\rm d}\mu.$ Suppose also that the volume growth of $M$ is at most exponential. Note that these assumptions hold if $M$ is a Riemannian manifold with Ricci curvature bounded from below. We may define the “localised” Riesz transforms $\nabla(\mathcal{L}+\varepsilon)^{-1/2}$, where $\varepsilon$ is in $\mathbb{R}^{+}$. Russ [Ru] proved that the localised Riesz transforms map local atoms uniformly into $L^{1}(M)$. However, in general, there is no indication that this result interpolates with the trivial $L^{2}(M)$ estimate to produce $L^{p}(M)$ boundedness for $p$ in $(1,2)$. Our theory complements Russ’ results. Indeed, Proposition 8.1 implies that $\nabla(\mathcal{L}+\varepsilon)^{-1/2}$ extends to a bounded operator from $H^{1}(M)$ into $L^{1}(M)$. In the case where $M$ possesses property (I) these result interpolate with the trivial $L^{2}(M)$ estimate and give that $\nabla(\mathcal{L}+\varepsilon)^{-1/2}$ extends to a bounded operator on $L^{p}(M)$ for all $p$ in $(1,2)$, a fact already known, but whose proof is far from being trivial. ## References * [A] J.-Ph. Anker, $L_{p}$ Fourier multipliers on Riemannian symmetric spaces of the noncompact type, _Ann. of Math._ 132 (1990), 597–628. * [ACDH] P. Auscher, T. Coulhon, X.T. Duong and S. Hofmann, Riesz transforms on manifolds and heat kernel regularity, _Ann. Sci. École Norm. Sup._ 37 (2004), 911–957. * [BL] J. Bergh and J. Löfström, Interpolation Spaces. An Introduction. Grundlehren der mathematischen Wissenschaften, Bd. 223, Springer–Verlag, Berlin Heidelberg New York, 1976 * [BC] R.L. Bishop and R. Crittenden, “Geometry of manifolds”, Academic Press, New York, 1964. * [Bow] Bownik, Boundedness of operators on Hardy spaces via atomic decomposition, _Proc. Amer. Math. Soc._ , 133 (2005), 3535–3542. * [Buc] S. M. Buckley, Inequalities of John-Nirenberg type in doubling spaces, _J. Analyse Math._ , 79 (1999), 215–240. * [Bur] D.L. Burkholder, Distribution function inequalities for martingales, _Ann. Prob._ 1 (1973), 19–42. * [BG] D.L. Burkholder and R.F. Gundy, Extrapolation and interpolation of quasilinear operators on martingales, _Acta Math._ 124 (1970), 249–304. * [Bu] P. Buser, A note on the isoperimetric constant, _Ann. Sci. École Norm. Sup._ 15 (1982), 213–230. * [Ca] A.P. Calderón, Intermediate spaces and interpolation, the complex method, _Studia Math._ 24 (1964), 113–165. * [CMM] A. Carbonaro, G. Mauceri and S. Meda, $H^{1}$, $BMO$ and the complementary isoperimetric property, preprint, 2007. * [Cha] I. Chavel, _Isoperimetric inequalities. Differential geometric and analytic perspectives._ , vol. 145 of Cambridge Tracts in Mathematics. Cambridge University Press, Cambridge, 2001. * [CGT] J. Cheeger, M. Gromov and M. Taylor, Finite propagation speed, kernel estimates for functions of the Laplace operator, and the geometry of complete Riemannian manifolds, _J. Diff. Geom._ 17 (1982), 15–53. * [Ch] M.Christ, A $T(b)$ theorem with remarks on analytic capacity and the Cauchy integral, _Coll. Math._ 61 (1990), 601–628. * [CW] R.R. Coifman and G. Weiss, Extensions of Hardy spaces and their use in analysis, _Bull. Amer. Math. Soc._ 83 (1977), 569–645. * [CD] T. Coulhon and X.T. Duong, Riesz transforms for $1<p\leq 2$, _Trans. Amer. Math. Soc._ 351 (1999), 1151–1169. * [CJ] M. Cwickel and S. Janson, Interpolation of analytic families of operators, _Studia Math._ 79 (1984), 61–71. * [Da] G. David, Morceaux de graphes lipschitziens et intégrales singuliéres sur une surface, _Rev. Mat. Iberoamericana_ 4 (1989), 73–114. * [DY1] X.T. Duong and L.X. Yan, New function spaces of $BMO$ type, the John-Nirenberg inequality, interpolation, and applications, _Comm. Pure Appl. Math._ , (2005), –. * [DY2] X.T. Duong and L.X. Yan, Duality of Hardy and $BMO$ spaces associated inequality, interpolation, and applications th operators with heat kernel bounds, _J. Amer. Math. Soc._ , 18 (2005), 943–973. * [EG] L. C. Evans, R. F. Gariepy, _Measure Theory and Fine Properties of Functions_ , Studies in Advanced Mathematics, CRC Press, Boca Raton, FL, 1992. * [F] C. Fefferman, Characterizations of bounded mean oscillation, _Bull. Amer. Math. Soc._ 77 (1971), 587–588. * [FS] C. Fefferman and E.M. Stein, $H^{p}$ spaces of several variables, _Acta Math._ 87 (1972), 137–193. * [Gr] M. Gromov, Curvature, diameter and Betti numbers, _Comment. Math Helv._ 56 (1981), 179–195. * [Gun] R.F. Gundy, Sur les transformations de Riesz pour le semigroupe d’Ornstein–Uhlenbeck, _C. R. Acad. Sci. Paris Sci. Ser. I Math._ 303 (1986), 967–970. * [He] E. Hebey, _Sobolev Spaces on Riemannian Manifolds_ , Lecture Notes in Mathematics, 1635, Springer-Verlag, Berlin, 1996. * [H1] L. Hörmander, Estimates for translation invariant operators in $L^{p}$ spaces, _Acta Math._ 104 (1960), 93–140. * [I] A.D. Ionescu, Fourier integral operators on noncompact symmetric spaces of real rank one, _J. Funct. Anal._ 174 (2000), no. 2, 274–300. * [JN] F. John and L. Nirenberg, On functions of bounded mean oscillation, _Comm. Pure Appl. Math._ 14 (1961), 415–426. * [Le] M. Ledoux, A simple analytic proof of an inequality of P. Buser, _Proc. Amer. Math. Soc._ 121 (1994), 951–959. * [MP] P. MacManus and C. Pérez, Trudinger inequalities without derivatives, _Trans. Amer. Math. Soc._ 354 (2002), 1997–2002. * [Ma] R. Macías, _Interpolation theorems on generalized Hardy spaces_ , Ph. D. thesis, Washington University, 1974. * [MMNO] J. Mateu, P. Mattila, A. Nicolau and J. Orobitg, BMO for nondoubling measures, _Duke Math. J._ 102 (2000), 533–565. * [MM] G. Mauceri and S. Meda, $H^{1}$ and $BMO$ for the Ornstein–Uhlenbeck operator, to appear in _J. Funct. Anal._. * [MSV] S. Meda, P. Sjögren and M. Vallarino, On the $H^{1}$-$L^{1}$ boundedness of operators, to appear in _Proc. Amer. Math. Soc._. * [MPPP] M. Miranda Jr, D. Pallara, F. Paronetto, M. Preunkert, Heat Semigroup and Functions of Bounded Variation on Riemannian Manifolds, Preprint (2005), available at the address http://cvgmt.sns.it/cgi/get.cgi/papers/mirpalpar05a/ * [NTV] Nazarov, Treil and Volberg, The $Tb$-theorem on non-homogeneous spaces, _Acta Math._ 190 (2003), no. 2, 151–239. * [Ru] E. Russ, $H^{1}$–$L^{1}$ boundedness of Riesz transforms on Riemannian manifolds and on graphs, _Pot. Anal._ 14 (2001), 301–330. * [Sa] L. Saloff-Coste, _Aspects of Sobolev-type inequalities_ , London Math. Soc. Lecture Note Series No. 289, Cambridge University Press, 2002. * [St1] E.M. Stein, _Topics in Harmonic Analysis Related to the Littlewood–Paley Theory_ , Annals of Math. Studies, No. 63, Princeton N. J., 1970. * [St2] E.M. Stein, _Harmonic Analysis. Real variable methods, orthogonality and oscillatory integrals_ , Princeton Math. Series No. 43, Princeton N. J., 1993. * [Ta] M. E. Taylor, $L^{p}$-estimates on functions of the Laplace operator, _Duke Math. J._ 58 (1989), 773–793. * [To] X. Tolsa, BMO, $H^{1}$, and Calderón-Zygmund operators for non doubling measures, _Math. Ann._ 319 (2001), no. 1, 89–149. * [T] A. Torchinsky, _Real variable methods in Harmonic Analysis_ , Pure and Applied Mathematics 123, Academic Press, 1986. * [V] J. Verdera, On the $T(1)$-theorem for the Cauchy integral, _Ark. Mat._ 38 (2000), 183–199. * [YZ] D. Yang and Y. Zhou, A boundedness criterion via atoms for linear operators in Hardy spaces, to appear in _Constr. Approx._.	arxiv-papers	2008-08-01T16:42:50	2024-09-04T02:48:57.112689	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Andrea Carbonaro, Giancarlo Mauceri, Stefano Meda", "submitter": "Giancarlo Mauceri", "url": "https://arxiv.org/abs/0808.0146" }
0808.0169	Viscous Cosmology and Thermodynamics of Apparent Horizon M. Akbar **Email address: ak64bar@yahoo.com †††Email address: makbar@camp.edu.pk Centre for Advanced Mathematics and Physics National University of Sciences and Technology Peshawar Road, Rawalpindi, Pakistan Abstract It is shown that the differential form of Friedmann equations of FRW universe can be recast as a similar form of the first law ,$T_{A}dS_{A}=dE+WdV$, of thermodynamics at the apparent horizon of FRW universe filled with the viscous fluid. It is also shown that the generalized second law of thermodynamics holds at the apparent horizon of FRW universe and preserves dominant energy condition. PACS numbers: 04.70.Dy, 97.60.Lf ### Introduction: In the cosmological setting, one can associate Hawking temperature and entropy with the apparent horizon analogous to the hawking temperature and entropy associated with the black hole horizon [1, 2, 3]. In the case of de Sitter space, the event horizon matches with the apparent horizon of FRW universe with $k=0$ however for more general cosmological models, the event horizon may not exist but the apparent horizon associated with the Hawking temperature and entropy always exists. The thermodynamic properties associated with the apparent horizon of FRW universe filled with the perfect fluid has been studied by many authors (see for examples [1, 2, 4, 6, 5, 7, 8]). The extension of this connection between thermodynamics and gravity has also been carried out in the braneworld cosmology [9, 10, 11, 12]. For a general static spherically symmetric and stationary axisymmetric spacetimes, It was shown that the Einstein field equations can be rewritten [18, 19, 20] as a first law of thermodynamics. More recently, by considering a masslike function, it has been shown [13] that the equilibrium thermodynamics should exist in the extended theories of gravity and the Friedman equations in various theories of gravity can be rewritten as a first law $TdS=dE$ at the apparent horizon of FRW universe. More recently, Cai et al [5] has shown that by employing Clausius relation, $\delta Q=TdS$, to the apparent horizon of a FRW universe, they are able to derive the modified Friedmann equation by using quantum corrected entropy-area relation. The cosmological models with perfect cosmic fluid has been studied widely in literature however viscous cosmic fluid came much later in the study of the universe [17]. It has been revealed by Barrow [14](also see [15]) that the thermodynamic entropy associated with bulk viscosity violates the dominant energy condition and can decrease very well. This fact was resumed [16] in the framework of the solutions given in reference [15]. By taking into account the entropy of the sources only, it has been deduced that the entropy of the bulk viscous sources cannot decrease since it would violate the second law of thermodynamics for large cosmic time values. However, in order to state the generalized second law, one has to include both the entropy connected with the sources and the entropy associated with background geometry. The most of the study of the second law in the cosmological context has been carried out at the event horizon (see for examples [15, 16]) however the validity of second law within the apparent horizon of FRW universe is of great important to investigate. In this letter we will also investigate this issue. The cosmological models with bulk viscosity have been studied in the literature from various point of view (see for examples [15, 21, 22]). The purpose of this letter is twofold. The first is to show that the Friedmann equations of FRW universe pertaining cosmic bulk viscosity can be rewritten as a similar form of the first law of thermodynamics. The other is to discuss the generalized second law within the apparent horizon of FRW universe. Let us start with a FRW universe of metric $ds^{2}=-dt^{2}+a^{2}(t)(\frac{dr^{2}}{1-kr^{2}}+r^{2}d\Omega^{2}),$ (1) where $d\Omega^{2}=d\theta^{2}+sin^{2}\theta d\phi^{2}$ stands for the line element of 2-dimensional unit sphere and the spatial curvature constant $k=+1$, 0 and $-1$ represents a closed, flat and open universe, respectively. The above metric (1) can be rewritten in spherical form $ds^{2}=h_{ab}dx^{a}dx^{b}+\tilde{r}^{2}d\Omega^{2},$ (2) where $\tilde{r}=a(t)r$, $x^{0}=t$, $x^{1}=r$ and $h_{ab}=diag(-1,\frac{a^{2}}{1-kr^{2}})$. One can work out the dynamical apparent horizon from the relation $h^{ab}\partial_{a}\tilde{r}\partial_{b}\tilde{r}=0$ which turns out $\frac{1}{\tilde{r}^{2}_{A}}=H^{2}+k/a^{2}$, where $H\equiv\frac{\dot{a}}{a}$ is the Hubble parameter and the dots denote derivatives with respect to the cosmic time. let the universe be filled with a viscous fluid and $u^{\mu}=(u^{0},u^{i})$ is the four-velocity of the fluid. In comoving coordinates $u^{0}=1$ and $u^{i}=0$. Define $h_{\mu\nu}=g_{\mu\nu}+u_{\mu}u_{\nu}$, $w_{\mu\nu}=h^{\alpha}_{\mu}h^{\beta}_{\nu}u_{[\alpha;\beta]}$ and $\theta_{\mu\nu}=h^{\alpha}_{\mu}h^{\beta}_{\nu}u_{(\alpha;\beta)}$ as a projection tensor, rotation tensor and expansion tensor, respectively. The scalar expansion is $\theta\equiv\theta^{\mu}_{\mu}=u^{\mu}_{;\mu}$ and $\sigma_{\mu\nu}=\theta_{\mu\nu}-\frac{1}{3}h_{\mu\nu}\theta$ is the shear tensor. Taking into consideration of metric (1), $u^{\mu}_{;\nu}=h^{\mu}_{\nu}\frac{\dot{a}}{a}$ which implies that the rotation and shear tensors vanish, i.e. $w_{\mu\nu}=\sigma_{\mu\nu}=0$. While the scalar expansion is $\theta=3\frac{\dot{a}}{a}=3H$. Hence the energy- momentum tensor of viscous fluid in the background of metric (1) finally becomes [21] $T_{\mu\nu}=(\rho+P-\zeta\theta)u_{\mu}u_{\nu}+(P-\zeta\theta)g_{\mu\nu},$ (3) where $\rho$, $P$ and $\zeta$ are the energy density, thermodynamic pressure and bulk viscosity respectively. The components of energy-momentum tensor are $T_{00}=\rho$, $T_{0i}=0$ and $T_{ij}=(P-\zeta\theta)g_{ij}$. Therefore, the total effect of the bulk viscosity is to reduce the pressure $P$ of the perfect fluid by an amount $\zeta\theta$, so that the effective pressure of the viscous fluid turns out to be $\tilde{P}=P-\zeta\theta$. The energy conservation $T^{\mu\nu}_{;\nu}=0$, yields $\dot{\rho}+3H(\rho+\tilde{P})=0$. Now we turn to discuss the thermodynamic interpretation of Friedman equations at the apparent horizon of FWR universe. For this purpose, we first define the surface gravity $\kappa=\frac{1}{2\sqrt{-h}}\partial_{a}(\sqrt{-h}h^{ab}\partial_{b}\tilde{r})$ at the apparent of FRW universe. By utilizing above relation, one can easily find that the surface gravity at the apparent horizon of FRW universe yields $\kappa=\frac{-1}{\tilde{r}_{A}}(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}}).$ (4) When $\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}}\leq 1$ implies the surface gravity $\kappa\leq 0$ which leads to the temperature $T=\kappa/2\pi\leq 0$. However, in reference [1], an approximation, $\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}}\ll 1$, has been used while determining horizon temperature. They found $T=\frac{1}{2\pi\tilde{r}_{A}}$ by approximating surface gravity $\|\kappa\|=\frac{1}{\tilde{r}_{A}}$. Now we turn to define energy of the universe within the apparent horizon. We take total matter energy $E=V\rho$ inside a sphere of radius $\tilde{r}$ which is also the Miser-Sharp energy [25], $E=\frac{\tilde{r}}{2}(1-h^{ab}\partial_{a}\tilde{r}\partial_{b}\tilde{r})$, within the apparent horizon. Since at the apparent horizon $(h^{ab}\partial_{a}\tilde{r}\partial_{b}\tilde{r}=0)$, so the Mizner-Sharp energy is in fact total matter energy inside the sphere of radius $\tilde{r}_{A}$ and is given by $E=V\rho.$ (5) The entropy $S=A/4$ is the horizon entropy, where $A=4\pi\tilde{r}_{A}^{2}$ is the horizon area. Note that we use the units $\hbar=c=G=\kappa_{B}=1$. From the Einstein field equations $G_{\mu\nu}=8\pi T_{\mu\nu}$, the Friedman equations for the viscous fluid of stress-energy tensor (3) can be written as $H^{2}+\frac{k}{a^{2}}=\frac{8\pi}{3}\rho,$ (6) $\dot{H}-\frac{k}{a^{2}}=-4\pi(\rho+\tilde{P}).$ (7) In terms of the apparent horizon, the Friedman equation (6) can be rewritten as $\frac{1}{\tilde{r}^{2}_{A}}=\frac{8\pi}{3}\rho.$ (8) Let the apparent horizon surface acts as the boundary of the thermal system of the FRW universe. In general, the apparent horizon is not constant but changes with time. Let $d\tilde{r}_{A}$ be an infinitesimal change in radius of the apparent horizon in time interval dt. This change in the apparent horizon will cause a small change $dV$ in the volume $V$ of the universe. This constitutes two spherical system of space-time with radii $\tilde{r}_{A}$ and $\tilde{r}_{A}+d\tilde{r}_{A}$ with a common source of fluid with non-zero pressure and energy density near horizon. Each space-time constituting a thermal system and satisfying Einstein’s equations, differs infinitesimally in the extensive variables volume, energy and entropy by $dV$, $dE$ and $dS$, respectively, while having the same values of intensive variables temperature $T$ and pressure $\bar{P}$. Thus for these two thermal states of the space- time, there must exist a certain relation connecting these thermodynamic quantities. To establish connection among the thermal quantities and Friedman equations, we differentiate equation (8) which implies $\frac{1}{\tilde{r}_{A}^{3}}d\tilde{r}_{A}=4\pi(\rho+\tilde{P})Hdt$ (9) Multiplying both sides of this equation with a factor $\tilde{r}_{A}^{3}(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}})$, one can rewrite this equation in a form $\frac{\kappa}{2\pi}\frac{d(4\pi\tilde{r}_{A}^{2})}{4}=-4\pi\tilde{r}_{A}^{2}(\rho+P)H(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}})dt$ (10) One can recognize that the quantities $\frac{\kappa}{2\pi}$ and $\frac{4\pi\tilde{r}_{A}^{2}}{4}$ are the temperature $T$ and entropy $S$ respectively. Therefore the above equation can be rewritten as $TdS=-4\pi\tilde{r}_{A}^{3}(\rho+\tilde{P})H(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}})dt$ (11) Now we turn to the total matter energy (5) and taking differential of it, one gets $dE=4\pi\tilde{r}_{A}^{2}\rho d\tilde{r}_{A}-4\pi\tilde{r}_{A}^{3}(\rho+\tilde{P})Hdt.$ (12) Using equations (11) and (12) one yields $dE=TdS+WdV,$ (13) where $W=\frac{1}{2}(\rho-\tilde{P})$ is the work density. The above equation is the unified first law [24] of relativistic thermodynamics. Note that the above thermal identity is obtained from the Friedman equation with viscous fluid together with the characteristics of the apparent horizon, while the author of reference [24] studied the thermodynamics of trapping horizon of dynamical black hole. It has been found that the entropy associated with apparent horizon is proportional to the horizon area which is originally initiated from the black hole horizon entropy that satisfies the so-called area formula [26]. In the case of perfect fluid, $\zeta=0$, one can compare the above thermal identity with the standard form of the first law, $dE=TdS- PdV$, of thermodynamics. In fact, the negative pressure term $-P$ in the first law is taken the place of the work density $W$. Notice that for pure de Sitter spacetime, $\rho=-P$, then one acquires the standard form $dE=TdS-PdV$. Note that we are considering the universe as a thermal object with apparent horizon as its boundary. So it useful to define other thermodynamic quantities like Enthalpy $H$ and Gibbs free energy $G$. The Enthalpy is defined by the equation $H=U+PV$, where $U=E$ is the total internal energy which is taken to be the total matter energy of the universe bounded by the apparent horizon. So the Enthalpy of the universe bounded by the apparent horizon can be written in terms of apparent radius $H=\frac{4\pi}{3}(\rho+P)\tilde{r}_{A}^{3}$. So the heat capacity of the universe enveloped by the apparent horizon at constant pressure of the perfect is defined as $C_{P}=(\frac{\partial H}{\partial T})_{P}=(\frac{\partial H}{\partial\tilde{r}_{A}}\frac{\partial\tilde{r}_{A}}{\partial T})_{P},$ (14) which yields $C_{P}=4\pi\tilde{r}_{A}^{2}(\rho+P)(\frac{\partial\tilde{r}_{A}}{\partial T})_{P}$. In general the temperature, $T=\kappa/2\pi\equiv\frac{-1}{2\pi\tilde{r}_{A}}(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}})$. However, if one considers the approximation, $\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}}\ll 1$ (also see [1, 9]), the temperature $T=\mid\kappa\mid/2\pi\equiv 1/2\pi\tilde{r}_{A}$ which implies that the heat capacity of the universe is negative provided the dominant energy condition holds. On the other hand, if one consider general temperature $T=\frac{-1}{2\pi\tilde{r}_{A}}(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}})$ associated with the apparent horizon which implies the heat capacity of the universe is positive definite provided $\frac{\dot{\tilde{r}}_{A}}{H\tilde{r}_{A}}\leq 1$. It is also interesting to note that the total matter energy $E=V\rho$, entropy $S=A/4$ and temperature $T=1/2\pi\tilde{r}_{A}$ satisfy the relation $E=TS$ . ### Generalized Second Law: Recently a lot of attention has been granted to the generalized second law of thermodynamics in the accelerating universe handled by the dark energy [27]. Using a particular model for dark energy, the generalized second law has been studied in reference [28] with the boundaries of universe enveloped by the apparent as well event horizon. It is important to investigate the generalized second law as defined in the region of the universe bounded by the apparent horizon in more general context. Let us now consider a region of FRW universe bounded by the apparent horizon filled by a perfect fluid of energy density $\rho$ and pressure $P$. We assume that the region bounded by the apparent horizon acts as a thermal system with boundary defined by the apparent horizon. Since the apparent horizon is not constant but varies with time. As the apparent radius changes , the volume enveloped by the apparent horizon will also change, however the thermal system bounded by the apparent horizon remains in equilibrium when it moves from one state to another so that the temperature of the system must be uniform and the same as the temperature of its surroundings. This requires that the temperature of total energy inside the apparent horizon should be in equilibrium with the temperature associated with the apparent horizon because we are not considering the flow of energy through the horizon. In order to state that a given cosmological solution satisfies/violates the second law of thermodynamics one should define first what is the generalization of the second law to the case of FRW universe. It should include both the entropy due to sources and the entropy associated with the background geometry. So the generalized second law can be expressed as $\dot{S}_{A}+\dot{S}_{m}\geq 0,$ (15) where $S_{A}=A/4G$ is the Bekenstein-Hawking entropy associated with the apparent horizon of FRW universe and $S_{m}$ is the entropy due to the matter sources inside the apparent horizon. From the Friedmann equations (6) and (7), it can easily be shown that $\dot{\tilde{r}}_{A}=4\pi\tilde{r}_{A}^{3}H(\rho+P)\geq 0$ provided the universe driven by the source of perfect fluid preserves dominant energy condition. Since we are assuming the region of a FRW universe enveloped by the apparent horizon as thermal system so that the change of energy $dE$ of the universe from one state to another must be connected through the Friedmann equations together with the characteristics of the apparent horizon. It has been shown [9, 3, 10] that the Friedmann equations at apparent horizon of FRW universe filled with perfect fluid satisfied the thermal identity $dE=T_{A}dS_{A}+WdV$ instead of satisfying the standard first law $T_{A}dS_{A}=dE+PdV$, where $T_{A}$ is the temperature associated with the apparent horizon. Since we are assuming the local equilibrium so that the change of matter energy $dE_{m}$ satisfies the first law of thermodynamics, $dE_{m}=T_{A}dS_{m}-PdV$, where $T_{A}=\|\kappa\|/2\pi$ is the temperature associated with the apparent horizon and is given by $T=\frac{1}{2\pi\tilde{r}_{A}}(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}}),$ (16) where $\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}}\leq 1$ ensures that the temperature is positive. From the first law, $dE_{m}=T_{A}dS_{m}-PdV$, one can obtain $T_{A}\dot{S}_{m}=4\pi\tilde{r}_{A}^{2}\dot{\tilde{r}}_{A}(\rho+P)-4\pi\tilde{r}_{A}^{3}H(\rho+P).$ (17) Now we turn to find out the entropy change $\dot{S}_{H}$ associated with the apparent horizon of FRW universe which yields $T_{A}\dot{S}_{A}=4\pi\tilde{r}_{A}^{3}H(\rho+P)-2\pi\tilde{r}_{A}^{2}(\rho+P)\dot{\tilde{r}}_{A}.$ (18) Adding equations (17) and (18), one gets the expression for the generalized second law $T_{A}(\dot{S}_{m}+\dot{S}_{H})=2\pi\tilde{r}_{A}^{2}\dot{\tilde{r}}_{A}(\rho+P).$ (19) It is obvious from the above equation that $\dot{S}_{m}+\dot{S}_{H}\geq 0$ implies that the generalized second law holds provided $\rho+P\geq 0$. ### Conclusion: It is shown that the differential form of Friedmann equations of FRW universe filled with a viscous fluid can be rewritten as a similar form of the first law ,$TdS=dE+WdV$, of thermodynamics at the apparent horizon of FRW universe. It can easily be seen that the similar identity also holds for the perfect fluid when bulk viscosity of the fluid is zero. It is also shown that the heat capacity of the universe filled with perfect fluid is negative if one considers the approximate temperature $T=\|\kappa\|/2\pi=1/2\pi\tilde{r}_{A}$. However if one utilizes the general expression $T=\frac{-1}{2\pi\tilde{r}_{A}}(1-\frac{\dot{\tilde{r}}_{A}}{2H\tilde{r}_{A}})$, the heat capacity of the universe is positive definite provided $\frac{\dot{\tilde{r}}_{A}}{H\tilde{r}_{A}}\leq 1$ which implies that the universe within the apparent horizon of FRW universe is thermodynamically stable. We also verify that the generalized second law of thermodynamics at the apparent horizon of FRW universe holds provided that the matter source satisfies the dominant energy condition. It is interesting to investigate the generalized second law for the extended theories of gravity in a region bounded by the apparent horizon of FRW universe. The work in this respect is under progress. ## Acknowledgments I would like to thank Rong-Gen Cai for his useful comments. The work is supported by a grant from National university of Sciences and Technology. ## References [1] R. G. Cai and S. P. Kim, JHEP 0502, 050 (2005)[arXiv:hep-th/0501055] * [2] G. W. Gibbons and S. W. Hawking, Phys.Rev. D15, 2738 (1977). * [3] M. Akbar and R. G. Cai, Phys. Rev. D75, 081003(2007) * [4] C. Eling, R. Guedens and T. Jacobson, Phys. Rev. Lett. 96,121301(2006). * [5] R. G. Cai, L. M. Cao, and Y. P. Hu, hep-th/0807.1232 * [6] M. Akbar and R. G. Cai, Phys. Lett. B648, 243(2007) * [7] R. G. Cai and L. M. Cao, Phys. Rev. D75, 064008(2007) * [8] M. Akbar and R. G. Cai, Phys. Lett. B635, 7 (2006). [arXiv:hep-th/0602156] * [9] R. G. Cai and L. M. Cao, hep-th/0612144 * [10] A. Sheykhi, B. Wang and R. G. Cai, Nucl. Phys. B779, 1(2007) [hep-th/0701198] * [11] A. Sheykhi, B. Wang and R. G. Cai, hep-th/0701261 * [12] R. G. Cai, hep-th/0712.2142 * [13] Y. Gong and A. Wang, hep-th/0704.0793 * [14] J. D. Barrow, Nucl. Phys. B310, 743(1988) * [15] M. Giovannini, Phys. Rev. D 59, 121301(1999) * [16] W. Zimdahl and D. Pavon, Phys. Rev. D 61, 108301(2000) * [17] C. W. Misner, Astrophys. J. 151, 431(1968) * [18] D. kothawala, S. Sarkar, and T. Padmanabhan, arXiv:gr-qc/0701002. * [19] T. Padmanabhan, Class. Quant. Grav. 19, 5387 (2002)[arXiv: gr-qc/0204019]; Phys. Rept. 406, 49 (2005)[arXiv: gr-qc/0311036]. * [20] A. Paranjape, S. Sarkar and T. Padmanabhan, Phys. Rev. D74, 104015(2006)[arXiv:hep-th/0607240]. * [21] Iver Brevik, gr-qc/0404095 * [22] W. Zimdahl, Phys. Rev. D 53, 5483(1996) * [23] M. Akbar and R. G. Cai, Phys. Rev. D75, 084003(2007); M. Akbar and R. G. Cai, Phys. Lett. B635, 7(2006)[hep-th/0602156]; M. Akbar and R. G. Cai, Phys. Lett. B648, 243(2007)[gr-qc/0612089]; A. V. Frolov and L. Kofman, JCAP 0305, 009 (2003); R. Bousso, Phys. Rev. D71, 064024(2005)[arXiv: hep-th/0412197]; * [24] S. A. Hayward, S. Mukohyana, and M. C. Ashworth, Phys. Lett. A256, 347(1999); S. A. Hayward, Class. Quant. Grav. 15, 3147(1998) * [25] C. W. Misner and D. H. Sharp, Phys. Rev. 136, B571(1964) * [26] R. M. Wald, Phys. Rev. D48, 3427(1993) * [27] G. Izquierdo and D. Pavon, Phys. Lett. B633, 420(2006); G. Izquierdo and D. Pavon, Phys. Lett. B639, 1(2006); M. R. Setare, JCAP 0701, 023(2007); H. Mohseni Sadjadi, Phys. Lett. B645, 108(2007); Y. Gong, B. Wang and A. Wang, JCAP 0701, 024(2007); J. Zhou, B. Wang, Y. Gong and E. Abdalla, gr-qc/0705.1264 * [28] B. Wang, Y. Gong and E. Abdalla, Phys. Rev. D74, 083520(2006)	arxiv-papers	2008-08-01T18:35:17	2024-09-04T02:48:57.121599	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Akbar", "submitter": "Muhammad Akbar", "url": "https://arxiv.org/abs/0808.0169" }
0808.0213	# Damped wave equations with dynamic boundary conditions Delio Mugnolo Institut für Angewandte Analysis, Universität Ulm, Helmholtzstraße 18, D-89081 Ulm, Germany delio.mugnolo@uni-ulm.de ###### Abstract. We discuss several classes of linear second order initial-boundary value problems, where damping terms appear in the main wave equation as well as in the dynamic boundary condition. We investigate their well-posedness and describe some qualitative properties of their solutions, including boundedness, stability, or almost periodicity. In particular, we are able to characterize the analyticity of certain $C_{0}$-semigroups associated to such problems. Applications to several problems on domains and networks are shown, mostly borrowed from [10, 39]. ###### Key words and phrases: Second order damped initial-boundary value problems; Operator matrices; Dynamical or Wentzell boundary conditions; Semigroups of operators ###### 1991 Mathematics Subject Classification: 47D06,35L20 ## 1\. Introduction In recent years, wave and beam equations with dynamic boundary conditions have been studied by many authors, see e.g. [27, 21, 3, 18, 19, 39, 40, 6, 10, 22, 31] and references therein. Wave equations with such boundary conditions are motivated by physical models incorporating the effect of frictions, as shown in [20]. Oscillating models involving dynamic boundary conditions for networks or more general polygonal domains have been considered, among others, by Ali Mehmeti, cf. [1, Chapt. 4] and references therein, as well as by Lagnese–Leugering–Schmidt in [26, § 2.7]. In fact, our setting can be adapted to problems on networks, interface problems and domains alike, cf. examples below. Most recent papers deal with abstract methods based on the theories of operator matrices and $C_{0}$-semigroups. These theories may have advantages over more usual methods based on Hilbert space methods and energy estimates. They allow more flexibility in treating non-dissipative systems. While most of the above mentioned papers only treat undamped wave equations, aim of this paper is to apply known methods in order to investigate a class of _damped_ problems. More precisely, we consider second order problems where the damping effect can be observed in both the waveguide and its boundary. While the first-order counterpart of this setting, i.e., _diffusion problems_ with dynamic boundary conditions, has been often discussed both on domains and on ramified structures due to its relations to stochastic analysis (in particular to the theory of Feller semigroups), comparatively less attention has been devoted to _wave equations_ with dynamic boundary conditions. Though, the connections between wave equations with further kinds of oscillatory boundary conditions (which are well-known in the mathematical physics of acoustic waves and mixed water/ice systems, cf. [25, 7, 8]) and dynamic ones has been thoroughly shown in [19]. Wave equations with different kinds of time-dependent boundary conditions have been recently considered by Nicaise and coauthors, cf. [37] and references therein. A numeric approach to this class of problems based on a Trotter–Kato-type result has been proposed in [28]. See also the conference procee! dings [2]. We stress that our theory is formulated in the abstract context of Banach spaces, whereas waves equations are usually discussed in a Hilbert space framework. This is due to the fact that, by a celebrated result of Littman, cf. [29], _undamped_ wave equations are well-posed in an $L^{p}$-setting if and only if $p=2$ or the space dimension is 1. However, more recent results indicate that this limitation does not apply to the _damped_ case, see e.g. [14, 15]. ## 2\. Mathematical framework ###### Assumption 2.1. We impose the following throughout this paper. 1. (1) $X$, $Y$, and $\partial X$ are Banach spaces such that $Y\hookrightarrow X$. 2. (2) ${A}:D({A})\subset X\to X$ and $C:D({C})\subset X\to X$ are linear operators. 3. (3) $L:D(A)\cap D(C)\to{\partial X}$ is a linear and surjective operator. 4. (4) $B_{1}:D(A)\to\partial X$ and $B_{2}:D(C)\to\partial X$ are linear operators. 5. (5) $B_{3}:D(B_{3})\subset{\partial X}\to{\partial X}$ and $B_{4}:D(B_{4})\subset{\partial X}\to{\partial X}$ are linear and closed operators. Functions on the main waveguide will be throughout this paper vectors in a Banach space $X$. We introduce a complete abstract second order problem $\ddot{u}(t)=Au(t)+C\dot{u}(t),\qquad t{\geq 0}$ and equip such a problem with second order dynamic boundary conditions represented by an equation (2.1) $\ddot{w}(t)=B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}w(t)+B_{4}\dot{w}(t),\qquad t{\geq 0},$ on another Banach space $\partial X$. Here the relation between the variables $u$ and $w$ is expressed by $w(t)=Lu(t)\qquad\hbox{and/or by}\qquad\dot{w}(t)=L\dot{u}(t),\qquad t\geq 0,$ where $L$ is some operator from $X$ to $\partial X$. We want to investigate (analytic) well-posedness and asymptoptic behavior of such a system. To this aim we re-write it in an abstract form and are eventually led to the complete second order abstract Cauchy problem (c${\mathcal{ACP}}^{2}$) $\left\\{\begin{array}[]{rcl}\ddot{\mathfrak{u}}(t)&=&\mathcal{A}{\mathfrak{u}}(t)+\mathcal{C}\dot{\mathfrak{u}}(t),\qquad t\geq 0,\\\ {\mathfrak{u}}(0)&=&{\mathfrak{f}}\in{\mathcal{X}},\qquad\dot{\mathfrak{u}}(0)={\mathfrak{g}}\in{\mathcal{X}},\end{array}\right.$ on the product space ${\mathcal{X}}:=X\times\partial X.$ Here (2.2) $\mathcal{A}:=\begin{pmatrix}A&0\\\ B_{1}&B_{3}\end{pmatrix}\qquad\hbox{and}\qquad\mathcal{C}:=\begin{pmatrix}C&0\\\ B_{2}&B_{4}\end{pmatrix}$ are operator matrices on $\mathcal{X}$, and their domains will depend on _how unbounded the damping term $C$ is with respect to the elastic term $A$_, as we see next. Moreover, the new variable $\mathfrak{u}$ is to be understood as ${\mathfrak{u}}(t):=\begin{pmatrix}u(t)\\\ Lu(t)\end{pmatrix},\qquad t\geq 0.$ If we reduce the second order evolution problem in $({\rm c}\mathcal{ACP}^{2})$ to a first order abstract Cauchy problem, our goal becomes to discuss the well-posedness of (${\mathbb{ACP}}$) $\left\\{\begin{array}[]{rcl}\dot{\mathbb{u}}(t)&=&{\mathbb{A}}{\mathbb{u}}(t),\qquad t\geq 0,\\\ {\mathbb{u}}(0)&=&{\mathbb{f}},\end{array}\right.$ in some suitable phase space, where $\mathbb{A}$ is the reduction matrix defined by (2.3) ${\mathbb{A}}:=\begin{pmatrix}0&I_{D({\mathcal{C}})}\\\ {\mathcal{A}}&{\mathcal{C}}\end{pmatrix}$ with suitable domain, and accordingly ${\mathbb{u}}(t)\equiv\begin{pmatrix}{\mathfrak{u}}(t)\\\ \dot{\mathfrak{u}}(t)\end{pmatrix},\quad t\geq 0,\qquad{\mathbb{f}}\equiv\begin{pmatrix}{\mathfrak{f}}\\\ {\mathfrak{g}}\end{pmatrix}.$ A setting of this kind permits to treat damped wave equations on domains or, more generally, on networks or ramified structures. ###### Example 2.2. Consider the initial value problems associated with a one dimensional damped plate-like equation on a bounded interval (2.4) $\left\\{\begin{array}[]{rcll}\ddot{u}(t,x)&=&-u^{\prime\prime\prime\prime}(t,x)+\dot{u}^{\prime\prime}(t,x),&t\geq 0,\;x\in(0,1),\\\ u^{\prime\prime}(t,j)&=&(-1)^{j}u^{\prime}(t,j)-u(t,j),&t\geq 0,\;j=0,1,\\\ \ddot{u}(t,j)&=&(-1)^{j+1}u^{\prime\prime\prime}(t,j)+(-1)^{j}u^{\prime}(t,j)\\\ &&\qquad+(-1)^{j}\dot{u}^{\prime}(t,j)-u(t,j)-\dot{u}(t,j),&t\geq 0,\;j=0,1,\end{array}\right.$ or with a strongly damped wave equation on an _open book_ consisting of $N$ copies of a domain $\Omega\subset{\mathbb{R}}^{n}$ (2.5) $\left\\{\begin{array}[]{rcll}\ddot{u}_{j}(t,x)&=&\Delta(\alpha u_{j}+\dot{u}_{j})(t,x),&t\geq 0,\;x\in\Omega,\;j=1,\ldots,N,\\\ u_{j}(t,z)&=&u_{\ell}(t,z)=:u(t,z),&t\geq 0,\;z\in\partial\Omega,\;j,\ell=1,\ldots,N,\\\ \ddot{u}_{j}(t,z)&=&-\frac{\partial}{\partial\nu}(\beta u_{j}+\dot{u}_{j})(t,z)+\gamma u(t,z)+\delta\dot{u}(t,z),&t\geq 0,\;z\in\partial\Omega,\;j=1,\ldots,N,\end{array}\right.$ for $\alpha,\beta,\gamma,\delta\in\mathbb{C}$. Here the second equation prescribes continuity along the _binding_ of the book. The initial value problems associated to both systems can be reduced to $(c{\mathcal{ACP}}^{2})$ on the Hilbert spaces $L^{2}(0,1)\times{\mathbb{C}}^{2}$ and $L^{2}(\Omega)\times L^{2}(\partial\Omega)$, respectively. Certain assumptions relating boundedness of elastic and damping term are satisfied, and by known results on damped wave equations we deduce analytic well-posedness.∎ In most usual examples (like in (2.4) and (2.5)), known energy estimates (cf. [18, 5]) permit to apply known results on damped second order problems – c.f. [36, 12, 13, 38, 34], [16, § XVIII.5.1], or [17, § VI.3], for methods based on spectral theory, functional calculus, operator matrix theory, of forms. All these methods require certain boundedness assumptions relating $A,C$ and $B_{3},B_{4}$. Aim of this paper is to give sufficient conditions on $A,C,L,B_{1},B_{2},B_{3},B_{4}$ that ensure that the reduction matrix $\mathbb{A}$ generates a $C_{0}$-semigroup, independently of dissipativeness. The theorems we are going to present in the remainder of this paper are of the following form: _If the damped wave equations with homogeneous boundary conditions associated with (suitable restrictions of) the matrix operators $\mathcal{A}$ and $\mathcal{C}$ are well-posed, then so is $(c\mathcal{ACP}^{2})$._ We will be specially concerned with investigatig further qualitative properties (boundedness, compactness, almost periodicity…) enjoyed by such a semigroup. In the following we will pay special attention to second order problems that are governed by analytic semigroups, a feature often discussed in applications, cf. [11, § 1]. Furthermore, a parabolic character directly implies regularity results, and can thus be exploited in order to discuss semilinear problems, e.g. by the techniques presented in [30]. Generation of analytic semigroups in the context of damped wave equations with dynamic boundary conditions has also been investigated in [39, 40] by different methods. Let us explain the plan of this paper. We have introduced in the Assumption 2.1.(1) a Banach space $Y$. Such a space $Y$ is in common applications somehow related to the domain of the operator $C$ and to the phase space of the second order problem – it was fact $Y=H^{2}(0,1)\cap H^{1}_{0}(0,1)$ and $Y=H^{1}_{0}(\Omega)$ in (2.4) and (2.5), respectively. (In fact, in concrete cases it will be a Sobolev space of the same order of the so-called _Kisyński space_ of the wave equation, i.e., the first factor of the phase space.) Depending on $Y$ and on the operator $L$, we need as in [31] to distinguish three different cases: $L$ can be – unbounded from $Y$ to $\partial X$, – unbounded from $X$ to $\partial X$ but bounded from $Y$ to $\partial X$, or – bounded from $X$ to $\partial X$. In this paper _we only consider the first two cases_ , in Sections 3 and 4, respectively. These occur, e.g., when we consider a wave equation on an $L^{p}$-space and $L$ is the normal derivative (see Example 3.5) or the trace operator (see Example 4.8), respectively. Our results should be compared with those of [10, 39, 40]. Instead, the case of $L\in{\mathcal{L}}(X,\partial X)$ is typical for spaces $X$ where the point evaluation is a bounded operator. The strongly damped case, i.e., the case of an operator $C$ that is “more unbounded” than $A$, is technically slightly different and will be treated in Sections 5. Finally, in Section 6 we prove a technical lemma on the exponential stability of semigroups generated by operator matrices. This seems to be new and of independent interest. ## 3\. The damped case $L\not\in{\mathcal{L}}(Y,\partial X)$ Of concern in this section are second order abstract problems with dynamic boundary conditions of the form (AIBPV${}_{a}^{2}$) $\left\\{\begin{array}[]{rcll}\ddot{u}(t)&=&Au(t)+C\dot{u}(t),&t{\geq 0},\\\ \ddot{w}(t)&=&B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}w(t)+B_{4}\dot{w}(t),&t{\geq 0},\\\ w(t)&=&Lu(t),&t{\geq 0},\\\ u(0)&=&f\in X,\qquad\;\;\dot{u}(0)=g\in X,&\\\ x(0)&=&h\in\partial X,\qquad\dot{x}(0)=j\in\partial X.&\end{array}\right.$ ###### Assumption 3.1. We complement the Assumptions 2.1 by the following. 1. (1) $\begin{pmatrix}A\\\ L\end{pmatrix}:D(A)\subset X\to X\times\partial X$ is closed. 2. (2) $A_{0}:=A_{\|{\rm ker}(L)}$ has nonempty resolvent set. 3. (3) $C$ is closed, $D(A)\subset D(C)$, and $[D(C)]$ is isomorphic to $Y$. 4. (4) $\partial Y$ is a Banach space, $[D(B_{4})]$ is isomorphic to $\partial Y$ and $\partial Y\hookrightarrow\partial X$. We denote by $[D(A)_{L}]$ the Banach space obtained by endowing $D(A)$ with the graph norm of the closed operator $A\choose L$. Observe that by the Closed Graph Theorem the embeddings $[D(A_{0})]\hookrightarrow[D(A)_{L}]\hookrightarrow Y$ hold. Furthermore, for $\lambda\in\rho(A_{0})$ we consider the _Dirichlet operators_ $D^{A,L}_{\lambda}$ associated with $A,L$, cf. [9, Lemma 2.3]. Such operators are right inverses of $L$ for all $\lambda\in\rho(A_{0})$ and are such that $AD_{\lambda}^{A,L}=\lambda D_{\lambda}^{A,L}$. Moreover, they are linear and bounded from $\partial X$ to $[D(A)_{L}]$, cf. [31, Lemma 3.2]. The following can be verified by a direct matrix computation. ###### Lemma 3.2. Consider the operator (3.1) ${\mathbb{A}}:=\begin{pmatrix}0&0&I_{Y}&0\\\ 0&0&0&I_{\partial Y}\\\ A&0&C&0\\\ B_{1}&B_{3}&B_{2}&B_{4}\end{pmatrix}.$ with domain (3.2) $D({\mathbb{A}}):=\left\\{\begin{pmatrix}u\\\ x\\\ v\\\ y\end{pmatrix}\in D(A)\times D(B_{3})\times D(C)\times D(B_{4}):Lu=x\right\\}$ on the Banach space (3.3) ${\mathbb{X}}:=Y\times\partial Y\times X\times\partial X.$ Then the well-posedness of the first order abstract Cauchy problem $(\mathbb{ACP})$ on $\mathbb{X}$ is equivalent to the well-posedness of $({\rm AIBPV}_{a}^{2})$ on $X$ and $\partial X$. We can identify a general function ${\mathbb{u}}:{\mathbb{R}}_{+}\to{\mathbb{X}}$ by $\mathbb{u}(t)\equiv\begin{pmatrix}u(t)\\\ x(t)\\\ v(t)\\\ y(t)\end{pmatrix},\qquad t\geq 0.$ Hence, if ${\mathbb{u}}$ is a classical solution to $(\mathbb{ACP})$, so that ${\mathbb{u}}\in C^{1}({\mathbb{R}}_{+},{\mathbb{X}})\cap C({\mathbb{R}}_{+},[D({\mathbb{A}})])$, then moreover $\dot{u}(t)=v(t)$, $t\geq 0$, and $v\in C^{1}({\mathbb{R}}_{+},X)$, and we conclude that $u\in C^{2}({\mathbb{R}}_{+},X)\cap C^{1}({\mathbb{R}}_{+},Y)$. ###### Lemma 3.3. Let $\lambda\in\rho(A_{0})$. The operator matrix $({\mathbb{A}},D({\mathbb{A}}))$ on $\mathbb{X}$ is similar to (3.4) ${\mathbb{G}}:=\begin{pmatrix}0&I_{Y}&\vrule height=14.0pt,depth=5.0pt&0&-D^{A,L}_{\lambda}\\\ \vskip-3.0pt\cr A_{0}&C&\vrule height=14.0pt,depth=5.0pt&\lambda D^{A,L}_{\lambda}&0\\\ \hrule\cr 0&0&\vrule height=14.0pt,depth=5.0pt&0&I_{\partial Y}\\\ \vskip-3.0pt\cr B_{1}&B_{2}&\vrule height=14.0pt,depth=5.0pt&B_{3}+B_{1}D^{A,L}_{\lambda}&B_{4}\\\ \end{pmatrix}$ with domain $D({\mathbb{G}}):=D(A_{0})\times D(C)\times D(B_{3})\times Y)$ on the Banach space ${\mathbb{Y}}:=Y\times X\times\partial Y\times\partial X.$ ###### Proof. Let $\lambda\in\rho(A_{0})$. First of all, we consider an isomorphism of $\mathbb{X}$ onto $\mathbb{Y}$ given by ${\mathbb{U}}_{\lambda}:=\begin{pmatrix}I_{Y}&-D^{A,L}_{\lambda}&0&0\\\ 0&0&I_{X}&0\\\ 0&I_{\partial Y}&0&0\\\ 0&0&0&I_{\partial X}\\\ \end{pmatrix},\quad\hbox{with inverse}\quad{\mathbb{U}}_{\lambda}^{-1}:=\begin{pmatrix}I_{Y}&&D_{\lambda}^{A,L}&0\\\ 0&0&I_{\partial Y}&0\\\ 0&I_{X}&0&0\\\ 0&0&0&I_{\partial X}\\\ \end{pmatrix}.$ We want to compute ${\mathbb{G}}={\mathbb{U}}_{\lambda}{\mathbb{A}}{\mathbb{U}}^{-1}_{\lambda}$. Observe that $\displaystyle D({\mathbb{G}})$ $\displaystyle=$ $\displaystyle\\{{\mathbb{x}}\in{\mathbb{X}}:{\mathbb{U}}^{-1}_{\lambda}{\mathbb{x}}\in D({\mathbb{A}})\\}$ $\displaystyle=$ $\displaystyle\left\\{\begin{pmatrix}u\\\ v\\\ x\\\ y\end{pmatrix}\in{\mathbb{X}}:\begin{pmatrix}u+D^{A,L}_{\lambda}x\\\ x\\\ v\\\ y\end{pmatrix}\\!\\!\\!\begin{array}[]{ll}&\in D(A)\times D(B_{3})\times D(C)\times D(B_{4})\\\ &\qquad\hbox{and }L(u+D^{A,L}_{\lambda}x)=x\end{array}\right\\}$ $\displaystyle=$ $\displaystyle\left\\{\begin{pmatrix}u\\\ v\\\ x\\\ y\end{pmatrix}\in{\mathbb{X}}:\begin{pmatrix}u+D^{A,L}_{\lambda}x\\\ x\\\ v\\\ y\end{pmatrix}\\!\\!\\!\begin{array}[]{ll}&\in D(A)\times D(B_{3})\times D(C)\times D(B_{4})\\\ &\qquad\hbox{and }Lu=0\end{array}\right\\}$ $\displaystyle=$ $\displaystyle D(A_{0})\times D(C)\times D(B_{3})\times D(B_{4}).$ Moreover, $\displaystyle{\mathbb{U}}_{\lambda}{\mathbb{A}}{\mathbb{U}}^{-1}_{\lambda}\begin{pmatrix}u\\\ v\\\ x\\\ y\end{pmatrix}$ $\displaystyle=$ $\displaystyle{\mathbb{U}}_{\lambda}{\mathbb{A}}\begin{pmatrix}u+D^{A,L}_{\lambda}x\\\ x\\\ v\\\ y\end{pmatrix}$ $\displaystyle=$ $\displaystyle{\mathbb{U}}_{\lambda}\begin{pmatrix}v\\\ y\\\ A_{0}u+\lambda D^{A,L}_{\lambda}x+Cv\\\ B_{1}(u+D^{A,L}_{\lambda}x)+B_{3}x+B_{2}v+B_{4}y\end{pmatrix}$ $\displaystyle=$ $\displaystyle\begin{pmatrix}v-D^{A,L}_{\lambda}y\\\ A_{0}u+\lambda D^{A,L}_{\lambda}x+Cv\\\ y\\\ B_{1}(u+D^{A,L}_{\lambda}x)+B_{3}x+B_{2}v+B_{4}y\end{pmatrix}.$ This finally shows the claimed representation of the operator matrix ${\mathbb{G}}$. ∎ We are now in the position to prove the main result of this section. ###### Theorem 3.4. Under the Assumptions 2.1 and 3.1, let $B_{1}\in{\mathcal{L}}([D(A)_{L}],\partial X)$. Then the following assertions hold. 1. (1) Assume that $B_{1}\in{\mathcal{L}}([D(A_{0})],\partial Y)$, or else that $B_{1}\in{\mathcal{L}}(Y,\partial X)$, and moreover that $B_{2}\in{\mathcal{L}}(Y,\partial Y)$, or else that $B_{2}\in{\mathcal{L}}(X,\partial X)$. Then the operator matrix $\mathbb{A}$ generates a $C_{0}$-semigroup on $\mathbb{X}$ if and only if (3.7) $\begin{pmatrix}0&I_{Y}\\\ A_{0}&C\end{pmatrix}\quad\hbox{with domain}\quad D({A}_{0})\times Y$ and (3.8) $\begin{pmatrix}0&I_{\partial Y}\\\ B_{3}&B_{4}\end{pmatrix}\quad\hbox{with domain}\quad D(B_{3})\times D(B_{4})$ generate $C_{0}$-semigroups on $Y\times X$ and $\partial Y\times\partial X$, respectively. 2. (2) Let $B_{2}\in{\mathcal{L}}(Y,\partial X)$. Then the reduction matrices introduced in (3.7)–(3.8) both generate analytic semigroups if and only if $\mathbb{A}$ generates an analytic semigroup. 3. (3) Let $B_{1}=B_{2}=0$. If the semigroup generated by either of the matrices defined in (3.7)–(3.8) is bounded and the other one is uniformly exponentially stable, then the semigroup generated by ${\mathbb{A}}$ is bounded. 4. (4) Let $B_{1}=B_{2}=0$. Assume the semigroups generated by matrices in (3.7)–(3.8) to be bounded. Let further the semigroup generated by the matrix in (3.7) be analytic. If the matrices in (3.7)–(3.8) have no common purely imaginary spectral values, then the semigroup generated by ${\mathbb{A}}$ is bounded. 5. (5) Let $B_{1}\in{\mathcal{L}}(Y,\partial X)$ and $B_{2}\in{\mathcal{L}}(X,\partial X)$. If both semigroups generated by matrices in (3.7)–(3.8) are uniformly exponentially stable, then there exists $\epsilon>0$ such that also the semigroup generated by ${\mathbb{A}}$ is uniformly exponentially stable, provided that $\\|B_{1}\\|_{{\mathcal{L}}(Y,\partial X)}+\\|B_{2}\\|_{{\mathcal{L}}(X,\partial X)}<\epsilon$. 6. (6) The operator matrix $\mathbb{A}$ on $\mathbb{X}$ has compact resolvent if and only if all the embeddings $[D({A}_{0})]\hookrightarrow Y\hookrightarrow X$ and $[D(B_{3})]\hookrightarrow\partial Y\hookrightarrow\partial X$ are compact. ###### Proof. Let $\lambda\in\rho(A_{0})$. By Lemma 3.3 instead of $\mathbb{A}$ on $\mathbb{X}$ it suffices to investigate the similar operator matrix $\mathbb{G}$ on $\mathbb{Y}$. We consider $\mathbb{G}$ as a $2\times 2$ operator matrix with diagonal domain. More precisely, ${\mathbb{G}}={\mathbb{G}}_{0}+{\mathbb{G}}_{1}:=\begin{pmatrix}{\bf A}&0\\\ {\bf C}&{{\bf D}}\end{pmatrix}+\begin{pmatrix}0&{\bf B}\\\ 0&0\end{pmatrix},\qquad D(\mathbb{G})=D({\bf A})\times D({{\bf D}}),$ where the $2\times 2$ block-entries ${\bf A},{\bf B},{\bf C},{\bf D}$ are defined as in (3.4). Also observe that by assumption the operator $B_{1}D^{A,L}_{\lambda}$ is bounded from $\partial Y$ to $\partial X$ for all $\lambda\in\rho(A_{0})$, so that we can discuss the generator property of the reduction matrix introduced in (3.8) instead of $\bf D$. By the Assumptions 3.1 the Dirichlet operator $D^{A,L}_{\lambda}$ is bounded from $\partial X$ to $Y$, hence the block entry $\bf B$ is bounded from $\partial Y\times\partial X$ to $Y\times X$. Thus, by the bounded perturbation theorem we only have to care about the lower triangular operator matrix ${\mathbb{G}}_{0}$. (1) The off-diagonal block-entry $\bf C$ is bounded from $[D({\bf A})]$ to $[D({{\bf D}})]$ or from $Y\times X$ to $\partial Y\times\partial X$. It follows by a perturbation result by Desch–Schappacher or by the bounded perturbation theorem, respectively, that ${\mathbb{G}}_{0}$ generates a $C_{0}$-semigroup on $\mathbb{Y}$ if and only if both diagonal block-entries ${\bf A},{\bf D}$ of ${\mathbb{G}}_{0}$ generate $C_{0}$-semigroups on $Y\times X$ and on $\partial Y\times\partial X$, respectively. (2) The diagonal block-entries of ${\mathbb{G}}_{0}$ both generate analytic semigroups. Moreover, the off-diagonal entry $\bf B$ is bounded from $[D({\bf A})]$ to $\partial Y\times\partial X$. It follows by [35, Cor. 3.3] that ${\mathbb{G}}_{0}$ generates an analytic semigroup on $\mathbb{Y}$. (3)–(4)–(5) These assertions follow directly from Proposition 6.1 below. (6) Since $D({\mathbb{G}})=D({\bf A})\times D({\bf D})$, $\mathbb{G}$ has compact resolvent if and only its diagonal block-entries have compact resolvent. ∎ ###### Example 3.5. We discuss the initial value problem associated with $\left\\{\begin{array}[]{rcll}\ddot{u}(t,x)&=&-\Delta^{2}u(t,x)+\Delta\dot{u}(t,x),&t{\geq 0},\;x\in\Omega\\\ \ddot{w}(t,z)&=&q_{1}(z)\Delta u(t,z)+q_{2}(z)\dot{u}(t,z)\\\ &&\qquad-\Delta_{\partial\Omega}^{2}w(t,z)-w(t,z)+\Delta_{\partial\Omega}\dot{w}(t,z)-\dot{w}(t,z),&t{\geq 0},\;z\in\partial\Omega,\\\ w(t,z)&=&\frac{\partial\Delta u}{\partial\nu}(t,z)-p(z)\Delta u(t,z),&t{\geq 0},\;z\in\partial\Omega,\\\ \frac{\partial{u}}{\partial\nu}(t,z)&=&p(z){u}(t,z),&t{\geq 0},\;z\in\partial\Omega,\\\ \frac{\partial\dot{u}}{\partial\nu}(t,z)&=&p(z)\dot{u}(t,z),&t{\geq 0},\;z\in\partial\Omega,\end{array}\right.$ similar to that discussed in [40, Exa. 4.3]. Here $\Omega\subset{\mathbb{R}}^{n}$ is a bounded open domain with smooth boundary $\partial\Omega$ and $p,q_{1},q_{2}\in L^{\infty}(\partial\Omega)$, $p<0$. Observe that – whenever re-written as $({\rm c}\mathcal{ACP}^{2})$ – the operator matrix $\mathcal{C}$ is in general neither self-adjoint, nor strictly negative definite, and $\mathcal{A}\not=-{\mathcal{C}}^{2}$, thus it is not possible to directly apply the results presented in [13], [16, § XVIII.5.1], or [38, § 6.4]. In order to apply the results presented above, we consider $Y:=\left\\{u\in H^{2}(\Omega):\frac{\partial u}{\partial\nu}=pu_{\|\partial\Omega}\right\\},\quad X:=L^{2}(\Omega),\quad\partial Y:=H^{2}(\partial\Omega),\quad\partial X:=L^{2}(\partial\Omega),$ and further $C:=\Delta,\qquad D(C):=Y,$ i.e., $C$ is the Laplacian with Robin boundary conditions, and $A:=-\Delta^{2},\qquad D(A):=\left\\{u\in H^{\frac{7}{2}}(\Omega):\frac{\partial u}{\partial\nu}=pu_{\|\partial\Omega}\right\\}\subset D(C).$ Let $Lu(z):=\frac{\partial\Delta u}{\partial\nu}(z)-p(z)\Delta u(z)\qquad\hbox{for}\;u\in H^{\frac{7}{2}}(\Omega),\;z\in\partial\Omega.$ Such an operator is well defined in the sense of traces. Then by usual boundary regularity results one sees that $-A_{0}=-A_{\|\ker(L)}$ is the square of $C$. The operator $C$ is self-adjoint and strictly negative definite, and we obtain by [38, Thm. 6.4.3 and Thm. 6.4.4] that the operator matrix defined in (3.7) generates an analytic, uniformly exponentially stable semigroup on $Y\times X$. Let now $(B_{1}u)(z):=q_{1}(z)\Delta u(z)\quad(B_{2}u)(z):=q_{2}(z)u(z),\qquad u\in H^{2}(\Omega),\;z\in\partial\Omega.$ It is clear that $B_{1},B_{2}$ are bounded from $Y$ to $\partial X$ and from $X$ to $\partial X$, respectively. Consider moreover the operators $B_{3}$ and $B_{4}$ defined by $\begin{array}[]{lll}&B_{3}:=-\Delta^{2}_{\partial\Omega}-I,&D(B_{3}):=H^{4}(\partial\Omega),\\\ &B_{4}:=\Delta_{\partial\Omega}-I,&D(B_{4}):=H^{2}(\partial\Omega),\end{array}$ where $\Delta_{\partial\Omega}$ denotes the Laplace–Beltrami operator, which is self-adjoint and negative definite. By [13, Thm. 1.1] the operator matrix defined in (3.8) generates a uniformly exponentially stable analytic semigroup on $\partial Y\times\partial X$. If $\\|q_{1}\\|_{\infty}+\\|q_{2}\\|_{\infty}\to 0$, then by Theorem 3.4.(5) the solution $u$ converges to 0 in the energy norm. By Theorem 3.4.(4), the semigroup governing the problem is also compact. ∎ ## 4\. The damped case $L\in{\mathcal{L}}(Y,\partial X)$ The case of $L\in{\mathcal{L}}(Y,\partial X)$ introduces some technical difficulties. In particular, we will show that our damped wave equations is well-posed on a phase space that is _not_ a product space (as $\mathbb{X}:=Y\times\partial Y\times X\times\partial X$ in Section 3 indeed was). We thus slightly modify our setting. ###### Assumption 4.1. We complement the Assumptions 2.1 by the following. 1. (1) $V$ is a Banach space such that $V\hookrightarrow Y$. 2. (2) $L$ can be extended to an operator that is bounded from $Y$ to $\partial X$, which we denote again by $L$, and such that $\ker(L)=V$. 3. (3) $\begin{pmatrix}A\\\ L\end{pmatrix}:D(A)\subset X\to X\times\partial X$ is closed. 4. (4) ${A}_{0}:=A\|_{D(A)\cap\ker(L)}$ has nonempty resolvent set. 5. (5) $C$ is bounded from $[D(A)_{L}]$ to $X$. 6. (6) $B_{1},B_{2}$ are bounded from $[D(A)_{L}]$ to $\partial X$. 7. (7) $B_{3}$ is bounded on $\partial X$. ###### Remark 4.2. If $u$ is a classical solution to $({\rm AIBVP}^{2}_{a})$, then $u\in C^{1}({\mathbb{R}}_{+},Y)$. Since by Assumption 4.1.(2) the operator $L$ is bounded from $Y$ to $\partial X$, we obtain that $L$ and the derivation with respect to time commute. In other words, if $w(t)=Lu(t)$ holds for all $t\geq 0$, then also $\dot{w}(t)=L\dot{u}(t)$ holds for all $t\geq 0$. Thus, we are led to consider in this section a modified version of $({\rm AIBVP}^{2}_{a})$, namely (AIBPV${}_{b}^{2}$) $\left\\{\begin{array}[]{rcll}\ddot{u}(t)&=&Au(t)+C\dot{u}(t),&t{\geq 0},\\\ \ddot{w}(t)&=&B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}w(t)+B_{4}\dot{w}(t),&t{\geq 0},\\\ w(t)&=&Lu(t),\qquad\dot{w}(t)=L\dot{u}(t),&t{\geq 0},\\\ u(0)&=&f\in X,\qquad\;\;\dot{u}(0)=g\in X,&\\\ x(0)&=&h\in\partial X,\qquad\dot{x}(0)=j\in\partial X.&\end{array}\right.$ As before, we can perform a first order reduction of such a problem, re- writing it as $({\mathbb{ACP}})$. We investigate $(\mathbb{ACP})$ on the non-diagonal Banach space $\underline{\mathbb{X}}$ defined by $\underline{\mathbb{X}}:=\left\\{\begin{pmatrix}u\\\ x\end{pmatrix}\in Y\times\partial X:Lu=x\right\\}\times X\times\partial X\subset{\mathbb{X}},$ instead of $\mathbb{X}$ as in Section 3. A general function ${\mathbb{u}}:\mathbb{R}_{+}\to\underline{\mathbb{X}}$ is of the form $\mathbb{u}(t)\equiv\begin{pmatrix}u(t)\\\ Lu(t)\\\ v(t)\\\ y(t)\end{pmatrix},\qquad t\geq 0.$ Observe that if $\mathbb{u}(\cdot)$ is a classical solution to the problem $(\mathbb{ACP})$ in $\underline{\mathbb{X}}$, then by definition we obtain $\frac{du}{dt}(\cdot)=v(\cdot)\qquad\hbox{and}\qquad\frac{d(Lu)}{dt}(\cdot)=y(\cdot)$ Again because $L$ and $\frac{d}{dt}$ commute for $u\in C^{1}({\mathbb{R}}_{+},Y)$, we thus conclude that $Lv(\cdot)=y(\cdot).$ ###### Proposition 4.3. Consider the operator matrix $\underline{\mathbb{A}}$ on $\underline{\mathbb{X}}$ defined by (4.1) $\underline{\mathbb{A}}=\begin{pmatrix}0&0&I_{Y}&0\\\ 0&0&0&I_{\partial X}\\\ A&0&C&0\\\ B_{1}&B_{3}&B_{2}&B_{4}\end{pmatrix}.$ with domain (4.2) $D(\underline{\mathbb{A}})=\left\\{\begin{pmatrix}u\\\ x\\\ v\\\ y\end{pmatrix}\in D(A)\times\partial X\times D(C)\times D(B_{4}):Lu=x,\;Lv=y\right\\}.$ Then $\mathbb{u}\in C^{1}({\mathbb{R}}_{+},\underline{\mathbb{X}})$ is a classical solution to the initial-value problem associated with $\dot{\mathbb{u}}(t)={\mathbb{A}}\mathbb{u}(t),\qquad t{\geq 0},$ if and only if it is a classical solution to the initial-value problem associated with $\dot{\mathbb{u}}(t)=\underline{\mathbb{A}}\mathbb{u}(t),\qquad t{\geq 0}.$ ###### Corollary 4.4. The well-posedness (in the classical sense) of the first order abstract Cauchy problem ($\underline{\mathbb{ACP}}$) $\left\\{\begin{array}[]{rcl}\dot{\mathbb{u}}(t)&=&\underline{\mathbb{A}}{\mathbb{u}}(t),\qquad t\geq 0,\\\ {\mathbb{u}}(0)&=&{\mathbb{f}},\end{array}\right.$ on $\mathbb{X}$ is equivalent to the well-posedness (in the classical sense) of the $({\rm AIBPV}_{b}^{2})$ on $X$ and $\partial X$. The following can be shown essentially like in the proof of [31, Lemma 5.2]. We denote by $D_{\lambda}^{A,L}$ the Dirichlet operators associated with $A,L$ introduced in Section 3. ###### Lemma 4.5. The product Banach space $\underline{\mathbb{V}}:=V\times X\times\partial X\times\partial X$ is isomorphic to $\underline{\mathbb{X}}$ via $\underline{\mathbb{U}}_{\lambda}:=\begin{pmatrix}I_{Y}&-D_{\lambda}^{A,L}&0&0\\\ 0&0&I_{X}&-D_{\lambda}^{A,L}\\\ 0&I_{\partial X}&0&0\\\ 0&0&0&I_{\partial X}\\\ \end{pmatrix},\qquad\lambda\in\rho(A_{0}).$ The inverse of $\underline{\mathbb{U}}_{\lambda}$ is the operator matrix $\underline{\mathbb{U}}^{-1}_{\lambda}:=\begin{pmatrix}I_{V}&&D_{\lambda}^{A,L}&0\\\ 0&0&I_{\partial X}&0\\\ 0&I_{X}&0&D_{\lambda}^{A,L}\\\ 0&0&0&I_{\partial X}\\\ \end{pmatrix},\qquad\lambda\in\rho(A_{0}).$ In the remainder of this section we hence take $\lambda\in\rho(A_{0})$ and investigate properties of the similar operator matrix $\underline{\mathbb{U}}_{\lambda}\underline{\mathbb{A}}\underline{\mathbb{U}}^{-1}_{\lambda}$ on the product space $\mathbb{V}$. A tedious but direct matrix computation, similar to that performed in the proof of Lemma 3.3, yields the following. ###### Lemma 4.6. Let $\lambda\in\rho(A_{0})$. Then the operator matrix $\underline{\mathbb{A}}$ on $\underline{\mathbb{X}}$ defined in (4.1)–(4.2) is similar to (4.3) $\underline{\mathbb{G}}:=\begin{pmatrix}0&I_{V}&\vrule height=14.0pt,depth=5.0pt&()\\\ \vskip-3.0pt\cr A_{0}-D^{A,L}_{\lambda}B_{1}&C_{0}-D^{A,L}_{\lambda}B_{2}&\vrule height=14.0pt,depth=5.0pt&&\\\ \hrule\cr 0&0&\vrule height=14.0pt,depth=5.0pt&0&I_{\partial X}\\\ \vskip-3.0pt\cr B_{1}&B_{2}&\vrule height=14.0pt,depth=5.0pt&B_{3}+B_{1}D_{\lambda}^{A,L}&B_{4}+B_{2}D_{\lambda}^{A,L}\\\ \end{pmatrix}$ with domain $D(\underline{\mathbb{G}}):=D(A_{0})\times D(C_{0})\times\partial X\times D(B_{4})$ on the Banach space $\underline{\mathbb{V}}$. Here the upper-right block entry $()$ is given by $()=\begin{pmatrix}0&0\\\ -D_{\lambda}^{A,L}(B_{1}D_{\lambda}^{A,L}+B_{3})&-D_{\lambda}^{A,L}(B_{2}D_{\lambda}^{A,L}+B_{4})+(C-\lambda)D_{\lambda}^{A,L}\end{pmatrix}.$ The similarity transformation is performed by means of the operator matrix $\underline{\mathbb{U}}_{\lambda}$ introduced in Lemma 4.5. Observe that if $B_{4}\in{\mathcal{L}}(\partial X)$, then the lower-right entry (4.4) $\begin{pmatrix}0&I_{\partial X}\\\ B_{3}+B_{1}D_{\lambda}^{A,L}&B_{4}+B_{2}D_{\lambda}^{A,L}\end{pmatrix}$ in (4.3) is by assumption a bounded operator on $\partial X\times\partial X$. The following parallels parallels Theorem 3.4. ###### Theorem 4.7. Under the Assumptions 2.1 and 4.1 the following assertions hold. 1. (1) Let $B_{4}\in{\mathcal{L}}(\partial X)$. Then the operator matrix $\underline{\mathbb{A}}$ generates a $C_{0}$-semigroup (resp., an analytic semigroup) on $\underline{\mathbb{X}}$ if and only if (4.5) $\begin{pmatrix}0&I_{V}\\\ {A}_{0}-D^{A,L}_{\lambda}B_{1}&C_{0}-D^{A,L}_{\lambda}B_{2}\end{pmatrix}\quad\hbox{with domain}\quad D({A}_{0})\times V$ generates a $C_{0}$-semigroup (resp., an analytic semigroup) on $V\times X$ for some $\lambda\in\rho(A_{0})$. 2. (2) Let $B_{1}=B_{2}=0$ and $D_{\lambda}^{A,L}B_{4}\in{\mathcal{L}}(\partial X)$ for some $\lambda\in\rho(A_{0})$. If the semigroup generated by either of the matrices defined in (4.4)–(4.5) is bounded and the other one is uniformly exponentially stable, then the semigroup generated by $\underline{\mathbb{A}}$ is bounded. 3. (3) Let $B_{1}=B_{2}=0$. Assume the semigroups generated by matrices in (4.4)–(4.5) to be bounded. Let further the semigroup generated by the matrix in (4.4) be analytic. If the matrices in (4.4)–(4.5) have no common purely imaginary spectral values, then the semigroup generated by ${\mathbb{A}}$ is bounded. 4. (4) Let $B_{1}\in{\mathcal{L}}(V,\partial X)$ and $B_{2}\in{\mathcal{L}}(X,\partial X)$. Assume both semigroups generated by matrices in (4.4)–(4.5) to be uniformly exponentially stable. Then there is $\epsilon>0$ such that the semigroup generated by $\underline{\mathbb{A}}$ is uniformly exponentially stable whenever $\\|B_{1}\\|+\\|B_{2}\\|<\epsilon$. 5. (5) The operator matrix $\underline{\mathbb{A}}$ on $\underline{\mathbb{X}}$ has compact resolvent if and only if both the embeddings $[D({A}_{0})]\hookrightarrow V\hookrightarrow X$ are compact and ${\rm dim}\;\partial X<\infty$. ###### Proof. By Lemma 4.6 $\underline{\mathbb{A}}$ is a generator on $\underline{\mathbb{X}}$ if and only if $\underline{\mathbb{G}}$ is a generator on $\underline{\mathbb{Y}}$. Hence the operator matrix with diagonal domain $\underline{\mathbb{G}}$ can be studied by means of the results in [35, § 3]. Observe that the upper-right block entry $()$ of $\underline{\mathbb{G}}$ is a bounded operator from $\partial X\times\partial X$ to $V\times X$, by assumption. (1) Since $C\in{\mathcal{L}}([D(A)_{L}],X)$, the upper-right block-entry of (4.3) is a bounded operator from $\partial X\times\partial X$ to $V\times X$. Also the lower-left block-entry is bounded from $D(A_{0})\times V$ to $\partial X\times\partial X$, and the claim follows by [35, Cor. 3.2]. (3)–(4)–(5) The claims follow by Proposition 6.1 below. (6) The operator matrix $\underline{\mathbb{G}}$ has compact resolvent if and only if its domain is compactly embedded in $\underline{\mathbb{X}}$, i.e., if the embedding $[D(\underline{\mathbb{G}})]=[D(A_{0})]\times[D(C_{0})]\times[\partial X]\times[D(B_{4})]\hookrightarrow V\times X\times\partial X\times\partial X$ is compact. ∎ ###### Example 4.8. We discuss the initial value problem associated with $\left\\{\begin{array}[]{rcll}\ddot{u}_{j}(t,x)&=&u^{\prime\prime}_{j}(t,x),&t\in{\mathbb{R}},\;x\in(0,1),\;j=1,\ldots,E,\\\ u_{j}(t,\mathsf{v}_{i})&=&u_{\ell}(t,\mathsf{v}_{i})=:d^{u}_{i}(t),&t\in\mathbb{R},\;j,\ell=1,\ldots,E,\;i=1\ldots,V,\\\ \ddot{u}(t,i)&=&\sum_{j=1}p_{ih}\phi_{hj}\frac{\partial u_{j}}{\partial\nu}(t,h)+\sum_{h=1}^{V}m_{ih}d^{u}_{h}(t)+\sum_{h=1}^{V}n_{ih}d^{\dot{u}}_{h}(t),&t\in{\mathbb{R}},\;i=1,\ldots,V,\\\ \end{array}\right.$ on a network $G$ with $E$ edges and $V$ vertices. Here $M=(m_{ih})$, $N=(n_{ih})$, and $P=(p_{ih})$ are $V\times V$ matrices. (We refer to [24, 33] for the graph-theoretical notation as well as for references to this kind of problems.) Let $Y:=(W^{1,p}(0,1))^{E}\cap C(G),\quad X:=(L^{p}(0,1))^{E},\quad\hbox{and}\quad\partial X:={\mathbb{C}}^{V},$ for any $1\leq p<\infty$. Furthermore, we set $Au:=u^{\prime\prime},\qquad\hbox{for all}\;u\in D(A):=(W^{2,p}(0,1))^{E}\cap C(G),\qquad C:=0.$ Thus, the damping effect only appears in the boundary conditions. Moreover we consider an operator $(B_{1}u)_{i}:=\sum_{j=1}\phi_{ij}\frac{\partial u_{j}}{\partial\nu}(t,i)\qquad\hbox{for all}\;u\in D(B_{1}):=D(A)$ of Kirchhoff-type, and $B_{2}:=0,\qquad B_{3}:=M,\qquad B_{4}:=N.$ Let $Lu:=d^{u},\qquad D(L):=Y,$ so that $V=(W^{1,p}_{0}(0,1))^{E}\qquad\hbox{and}\qquad D(A_{0})=(W^{2,p}(0,1)\cap W^{1,p}_{0}(0,1))^{E},$ i.e., $A_{0}$ can be seen as a diagonal operator matrix consisting of second derivatives on $E$ unconnected intervals, each equipped with Dirichlet boundary conditions. Following the proof of [31, Prop. 7.1] one can show that $A_{0}-D^{A,L}_{\lambda}B_{1}$ generates a cosine operator function on $X$ for some $\lambda\in\rho(A_{0})$, hence the operator matrix in (4.5) generates a $C_{0}$-group. Thus, $\underline{\mathbb{A}}$ generates a $C_{0}$-group on $\underline{\mathbb{X}}$. This system is a generalisation of that considered in [10, § 2] for $E=1$, $V=2$, and $B_{1}=B_{3}=0$. Let $P=0$ and $M,N$ be negative definite. Then the matrix defined in (4.4) has negative spectrum, hence it generates a semigroup that is uniformly exponentially stable. Since moreover the group generated by the matrix in (4.5) is bounded, we conclude by Theorem 4.7.(2) that the solution to the problem is bounded and asymptotically almost periodic (for positive time). In particular, the problem admits a unique classical (backward as well as forward) solution.∎ We may sometimes interpret our dynamic boundary conditions as Wentzell-type ones. ###### Proposition 4.9. Let $\underline{\mathbb{A}}$ generate an analytic semigroup on $\mathbb{X}$. Then the solution $u$ to $({\rm AIBVP}^{2}_{b})$ satisfies the abstract Wentzell-type boundary conditions (4.6) $L\left(Au(t)+C\dot{u}(t)\right)=B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}Lu(t)+B_{4}L\dot{u}(t),\qquad t>0.$ If further $C$ maps $D(A)$ into $Y$, then $u$ satisfies in fact $LAu(t)+LC\dot{u}(t)=B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}Lu(t)+B_{4}L\dot{u}(t),\qquad t>0.$ ###### Proof. By assumption $(\mathbb{ACP})$ is governed by an analytic semigroup, thus for all initial data $\mathbb{f}\in\underline{\mathbb{X}}$ the orbit ${\mathbb{u}}(\cdot):=e^{\cdot\mathbb{A}}{\mathbb{f}}$ is of class $C^{\infty}\big{(}(0,\infty);[D({\mathbb{A}})]\big{)}$. In particular, taking the first coordinate $u$ of $\mathbb{u}$ and recalling that $u$ is by definition the solution to $({\rm AIBVP}^{2}_{b})$, we deduce that (4.7) $Au(t)+C\dot{u}(t)=\ddot{u}(t)\in D(A)\qquad\hbox{for all }t>0.$ By Assumption 2.1.(3) we can apply the operator $L$ to $Au(t)+C\dot{u}(t)$, $t>0$. By Remark 4.2 $L$ commutes with the derivation with respect to time, so that (4.8) $\ddot{w}(t)=L\ddot{u}(t)=L\big{(}Au(t)+C\dot{u}(t)\big{)}\qquad\hbox{for all }t>0.$ Plugging (4.8) into (2.1) we finally obtain (4.6). Let now $C$ map $D(A)$ into $Y$. Since also $\dot{u}(t)\in D(A)$, $t>0$, we obtain that $C\dot{u}(t)\in Y$, $t>0$, and we conclude that $Au(t)=\ddot{u}(t)-C\dot{u}(t)\in Y,\qquad t>0.$ Summing up, we can apply $L$ to each addend on the LHS of (4.7). ∎ ###### Example 4.10. We revisit the first system considered in Example 2.2. The associated initial value problem system is governed by an analytic semigroup and Proposition 4.9.(4) applies. The solution $u$ satisfies $\frac{\partial^{k}u}{\partial t^{k}}(t,\cdot)\in D(A)=H^{4}(0,1)$ for all $t>0$ and $k\in\mathbb{N}$, and in particular $\dot{u}(t,\cdot),\ddot{u}(t,\cdot)\in H^{4}(0,1)$, $t>0$. If follows that $\dot{u}^{\prime\prime}(t,\cdot)\in H^{2}(0,1)$ and $u^{\prime\prime\prime\prime}(t,\cdot)=\dot{u}^{\prime\prime}(t,\cdot)-\ddot{u}(t,\cdot)\in H^{2}(0,1)$ for all $t>0$. Thus, for $t>0$ we can evaluate $u^{\prime\prime\prime\prime}(t,\cdot)$ and $\dot{u}^{\prime\prime}(t,\cdot)$ at the endpoints of the interval $[0,1]$. We conclude that the solution to the initial value problem associated with (2.4) also satisfies $\begin{array}[]{rl}&u^{\prime\prime\prime\prime}(t,j)-\dot{u}^{\prime\prime}(t,j)+(-1)^{j+1}u^{\prime\prime\prime}(t,j)+(-1)^{j}u^{\prime}(t,j)\\\ &\qquad\qquad+(-1)^{j}\dot{u}^{\prime}(t,j)-u(t,j)-\dot{u}(t,j)=0,\qquad j=0,1,\;t>0,\end{array}$ a Wentzell-type boundary condition. ∎ ## 5\. The strongly damped case In this section we discuss the problem in a strongly damped setting, i.e., we assume that $C$ is “more unbounded” than $A$, and modify our assumptions accordingly. We treat both case $L\not\in{\mathcal{L}}(Y,\partial X)$ and $L\in{\mathcal{L}}(Y,\partial X)$ More precisely, we consider a complete second order abstract initial-boundary value problems with dynamic boundary conditions of the form (AIBPV${}_{c}^{2}$) $\left\\{\begin{array}[]{rcll}\ddot{u}(t)&=&Au(t)+C\dot{u}(t),&t{\geq 0},\\\ \ddot{w}(t)&=&B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}w(t)+B_{4}\dot{w}(t),&t{\geq 0},\\\ \dot{w}(t)&=&L\dot{u}(t),&t{\geq 0},\\\ u(0)&=&f\in X,\qquad\;\;\dot{u}(0)=g\in X,&\\\ x(0)&=&h\in\partial X,\qquad\dot{x}(0)=j\in\partial X.&\end{array}\right.$ Observe that the coupling relation expressed by the third equation is not the same of $({\rm AIBPV}_{a}^{2})$ or $({\rm AIBPV}_{b}^{2})$. ###### Assumption 5.1. We complement the Assumptions 2.1 by the following. 1. (1) $\begin{pmatrix}C\\\ L\end{pmatrix}:D(C)\subset X\to X\times\partial X$ is closed. 2. (2) $C_{0}:=C_{\|{\rm ker}(L)}$ has nonempty resolvent set. 3. (3) $A$ is closed, $D(C)\subset D(A)$, and $[D(A)]$ is isomorphic to $Y$. 4. (4) $\partial Y$ is a Banach space such that $[D(B_{4})]\hookrightarrow\partial Y\hookrightarrow\partial X$. As in Section 3, we denote by $[D(C)_{L}]$ the Banach space obtained by endowing $D(C)$ with the graph norm of the $C\choose L$, and for $\lambda\in\rho(C_{0})$ we consider the Dirichlet operators $D^{C,L}_{\lambda}$ associated with $C,L$, which are bounded from $\partial X$ to $[D(C)_{L}]$. By the Closed Graph Theorem we further have that $[D(C)_{L}]\hookrightarrow Y$. ###### Lemma 5.2. Define the linear space (5.1) $D_{d}({\mathbb{A}}):=\left\\{\begin{pmatrix}u\\\ x\\\ v\\\ y\end{pmatrix}\in D(A)\times D(B_{3})\times D(C)\times D(B_{4}):Lv=y\right\\}.$ Consider the operator $\mathbb{A}$ with domain $D_{d}({\mathbb{A}})$ on the Banach space $\mathbb{X}$, where $\mathbb{A}$ and $\mathbb{X}$ are defined as in (3.1) and (3.3). Then the well-posedness of the first order abstract Cauchy problem $(\mathbb{ACP})$ on $\mathbb{X}$ is equivalent to the well-posedness of $({\rm AIBPV}_{c}^{2})$ on $X$ and $\partial X$. With a proof similar to that of Lemma 3.3, one can see that the following holds. ###### Lemma 5.3. Let $\lambda\in\rho(C_{0})$. Then the operator matrix $({\mathbb{A}},D_{d}({\mathbb{A}})$ on $\mathbb{X}$ is similar to (5.2) ${\mathbb{H}}:=\begin{pmatrix}0&I_{Y}&\vrule height=14.0pt,depth=5.0pt&0&-D^{C,L}_{\lambda}\\\ \vskip-3.0pt\cr A-D^{C,L}_{\lambda}B_{1}&C_{0}-D^{C,L}_{\lambda}B_{2}&\vrule height=14.0pt,depth=5.0pt&-D^{C,L}_{\lambda}B_{3}&D^{C,L}_{\lambda}(\lambda- B_{3}D^{C,L}_{\lambda}-B_{4})\\\ \hrule\cr 0&0&\vrule height=14.0pt,depth=5.0pt&0&I_{\partial Y}\\\ \vskip-3.0pt\cr B_{1}&B_{2}&\vrule height=14.0pt,depth=5.0pt&B_{3}&B_{4}+B_{2}D^{C,L}_{\lambda}\\\ \end{pmatrix}$ with domain $D({\mathbb{H}}):=D(A)\times D(C_{0})\times D(B_{3})\times D(B_{4})$ on the Banach space ${\mathbb{Y}}:=Y\times X\times\partial Y\times\partial X.$ The similarity transformation is performed by means of the operator matrix ${\mathbb{V}}_{\lambda}:=\begin{pmatrix}I_{Y}&0&0&0\\\ 0&0&I_{X}&-D^{C,L}_{\lambda}\\\ 0&I_{\partial Y}&0&0\\\ 0&0&0&I_{\partial X}\\\ \end{pmatrix},$ which is an isomorphism from $\mathbb{X}$ onto $\mathbb{Y}$, for any $\lambda\in\rho(C_{0})$. ###### Theorem 5.4. Under the Assumptions 2.1 and 5.1 the following assertions hold. 1. (1) Assume that $B_{1}\in{\mathcal{L}}(Y,\partial X)$, and moreover that $B_{2}\in{\mathcal{L}}([D(C_{0})],[D(B_{4})])\cap{\mathcal{L}}([D(C)_{L}],\partial X)$ or else $B_{2}\in{\mathcal{L}}(X,\partial X)$. If $D^{C,L}_{\lambda}B_{3}\in{\mathcal{L}}(\partial Y,X)$ and $D^{C,L}_{\lambda}B_{4}\in{\mathcal{L}}(\partial X,X)$ for some $\lambda\in\rho(C_{0})$, then $\mathbb{A}$ generates a $C_{0}$-semigroup on $\mathbb{X}$ if and only if both $C_{0}-D^{C,L}_{\lambda}B_{2}$ and (5.3) $\begin{pmatrix}0&I_{\partial Y}\\\ B_{3}&B_{4}\end{pmatrix}$ generate $C_{0}$-semigroups on $X$ and $\partial Y\times\partial X$, respectively. 2. (2) Let $B_{1}\in{\mathcal{L}}(Y,\partial X)$ and $B_{2}\in{\mathcal{L}}(X,\partial X)$. If for some $\lambda\in\rho(C_{0})$ both $C_{0}-D^{C,L}_{\lambda}B_{2}$ and the reduction matrix defined in (5.3) generate analytic semigroups on $X$ and $\partial Y\times\partial X$, respectively, then $\mathbb{A}$ generates an analytic semigroup on $\mathbb{X}$. 3. (3) Let $B_{1}\in{\mathcal{L}}(Y,\partial X)$ and $B_{2}\in{\mathcal{L}}([D(C)_{L}],\partial X)$. Assume that for some $\lambda\in\rho(C_{0})$ $D^{C,L}_{\lambda}B_{3}\in{\mathcal{L}}(\partial Y,X)$ and $D^{C,L}_{\lambda}B_{4}\in{\mathcal{L}}(\partial X,X)$. If $C_{0}-D^{C,L}_{\lambda}B_{2}$ and the reduction matrix defined in (5.3) generate analytic semigroups on $X$ and $\partial Y\times\partial X$, respectively, then $\mathbb{A}$ generates an analytic semigroup on $\mathbb{X}$. 4. (4) Assume that $B_{1},B_{2}\in{\mathcal{L}}(Y,\partial X)$ and $B_{3}\in{\mathcal{L}}(\partial Y,\partial X)$. Then $\mathbb{A}$ generates a cosine operator function on $\mathbb{X}$ if and only both $C_{0}$ and $B_{4}$ generate cosine operator functions with associated phase spaces $Y\times X$ and $\partial Y\times\partial X$, respectively. ###### Proof. Let $\lambda\in\rho(C_{0})$. By Lemma 5.3 instead of $\mathbb{A}$ on $\mathbb{X}$ it suffices to investigate the similar operator matrix $\mathbb{H}$ on $\mathbb{Y}$. We consider $\mathbb{H}$ as a $2\times 2$ operator matrix with diagonal domain, i.e., ${\mathbb{H}}={\mathbb{H}}_{0}+{\mathbb{H}}_{1}:=\begin{pmatrix}{\bf A}&{\bf B}\\\ {\bf C}&{{\bf D}}\end{pmatrix},\qquad D(\mathbb{H})=D({\bf A})\times D({{\bf D}}),$ where the $2\times 2$ block-entries ${\bf A},{\bf B},{\bf C},{\bf D}$ are defined as in (5.2). (1) Under our assumptions we have $B_{2}D^{C,L}_{\lambda}\in{\mathcal{L}}(\partial X)$, so that both $\mathbf{A}$ and $\mathbf{D}$ generate $C_{0}$-semigroups on $Y\times X$ and $\partial Y\times\partial X$, respectively. Now the off-diagonal block-entries of $\mathbb{H}$ define an additive perturbation which is bounded either on $[D({\mathbb{H}})]$ or on $\mathbb{X}$, and by [35, Cor. 3.2] the claim follows. (2)–(3) By assumption, both $\mathbf{A}$ and $\mathbf{D}$ generate analytic semigroups on $Y\times X$ and $\partial Y\times\partial X$, respectively. Then in (2) $\mathbf{C}$ is bounded from $Y\times X$ to $\partial Y\times\partial X$ while $\mathbf{B}$ is bounded from $[D({\mathbf{D}})]$ to $Y\times X$, and in (3) $\mathbf{B}$ is bounded from $\partial Y\times\partial X$ to $Y\times X$ while $\mathbf{C}$ is bounded from $[D({\mathbf{A}})]$ to $\partial Y\times\partial X$, and by [35, Cor. 3.3] the claim follows. (4) By assumption, [32, Prop. 6.1] applies and $\bf A$ and $\bf D$ generate cosine operator functions with associated phase spaces $(Y\times Y)\times(Y\times X)$ and $(\partial Y\times\partial Y)\times(\partial Y\times\partial X)$, respectively. Moreover, observe that ${\bf B}\in{\mathcal{L}}([D({\bf D})],Y\times Y)$ and ${\bf C}\in{\mathcal{L}}(Y\times Y,\partial Y\times\partial X)$. Thus, by [32, Prop. 3.2] $\mathbb{H}$ generates a cosine operator function on $\mathbb{X}$. ∎ Recall that any generator of a cosine operator function also generates an analytic semigroup of angle $\frac{\pi}{2}$, cf. [4, Thm. 3.14.17]. Let us now modify our framework in order to deal with a setting where the boundary operator $L$ is bounded from $Y$ to $\partial X$. ###### Assumption 5.5. We complement the Assumptions 2.1 by the following. 1. (1) $V$ is a Banach space such that $V\hookrightarrow Y$. 2. (2) $L$ can be extended to an operator that is bounded from $Y$ to $\partial X$, which we denote again by $L$, and such that $\ker(L)=V$. 3. (3) $\begin{pmatrix}C\\\ L\end{pmatrix}:D(C)\subset X\to X\times\partial X$ is closed. 4. (4) ${C}_{0}:=C\|_{D(C)\cap\ker(L)}$ has nonempty resolvent set. 5. (5) $A$ is bounded from $[D(C)_{L}]$ to $X$. 6. (6) $B_{1},B_{2}$ are bounded from $[D(C)_{L}]$ to $\partial X$. 7. (7) $B_{3}$ is bounded on $\partial X$. Under the Assumptions 5.5 Remark 4.2, Proposition 4.3, and Corollary 4.4 still hold. Thus, we discuss the generator property of the same operator matrix $(\underline{\mathbb{A}},D(\underline{\mathbb{A}}))$ on $\underline{\mathbb{X}}$ introduced in Proposition 4.3. Moreover, also Lemma 4.5 remains valid up to replacing $\underline{\mathbb{U}}_{\lambda}$ therein by $\underline{\mathbb{V}}_{\lambda}:=\begin{pmatrix}I_{Y}&-D_{\lambda}^{C,L}&0&0\\\ 0&0&I_{X}&-D_{\lambda}^{C,L}\\\ 0&I_{\partial X}&0&0\\\ 0&0&0&I_{\partial X}\\\ \end{pmatrix},\qquad\lambda\in\rho(A_{0}).$ ###### Lemma 5.6. The operator matrix $\underline{\mathbb{A}}$ on $\underline{\mathbb{X}}$ defined in (4.1)–(4.2) is similar to $\underline{\mathbb{H}}:=\begin{pmatrix}0&I_{V}&\vrule height=14.0pt,depth=5.0pt&()\\\ \vskip-3.0pt\cr A_{0}-D^{C,L}_{\lambda}B_{1}&C_{0}-D^{C,L}_{\lambda}B_{2}&\vrule height=14.0pt,depth=5.0pt&&\\\ \hrule\cr 0&0&\vrule height=14.0pt,depth=5.0pt&0&I_{\partial X}\\\ \vskip-3.0pt\cr B_{1}&B_{2}&\vrule height=14.0pt,depth=5.0pt&B_{3}+B_{1}D_{\lambda}^{C,L}&B_{4}+B_{2}D_{\lambda}^{C,L}\\\ \end{pmatrix}$ with domain $D(\underline{\mathbb{G}}):=D(A_{0})\times D(C_{0})\times D(B_{3})\times D(B_{4})$ on the Banach space $\underline{\mathbb{V}}$. Here the upper-right block entry $()$ is given by $()=\begin{pmatrix}0&0\\\ -D_{\lambda}^{C,L}(B_{1}D_{\lambda}^{C,L}+B_{3})&-D_{\lambda}^{C,L}(B_{2}D_{\lambda}^{C,L}+B_{4})+(\lambda-A)D_{\lambda}^{C,L}\end{pmatrix}.$ ###### Theorem 5.7. Under the Assumptions 2.1 and 5.5 the following assertions hold. 1. (1) Let $B_{4}\in{\mathcal{L}}(\partial X)$. Then the operator matrix $\underline{\mathbb{A}}$ generates a $C_{0}$-semigroup (resp., an analytic semigroup) on $\underline{\mathbb{X}}$ if and only if (5.4) $\begin{pmatrix}0&I_{V}\\\ {A}_{0}-D^{C,L}_{\lambda}B_{1}&C_{0}-D^{C,L}_{\lambda}B_{2}\end{pmatrix}\quad\hbox{with domain}\quad V\times D({C}_{0})$ generates a $C_{0}$-semigroup (resp., an analytic semigroup) on $V\times X$ for some $\lambda\in\rho(C_{0})$. 2. (2) Let $A\in{\mathcal{L}}(V,X)$, $B_{1},B_{2}\in{\mathcal{L}}(V,\partial X)$ and $B_{4}\in{\mathcal{L}}(\partial X)$. Then $\underline{\mathbb{H}}$ generates a cosine operator function on $\underline{\mathbb{X}}$ if and only if $C_{0}$ generates a cosine operator function with associated phase space $V\times X$. ###### Proof. The assertion in (1) can be proved in a way similar to Theorem 4.1. To show that (2) holds, observe that the lower-right block-entry of $\underline{\mathbb{H}}$ is a bounded operator on $\partial X\times\partial X$ (hence the generator of a cosine operator function on $\partial X\times\partial X$), and that by [32, Prop. 6.1] the upper-left block-entry of $\mathbb{H}$ generates a cosine operator function with associated phase space $(V\times V)\times(V\times X)$. Then, by assumption the lower-left block-entry of $\underline{\mathbb{H}}$ is bounded from $V\times V$ to $\partial X\times\partial X$ and the upper-right one is bounded from $\partial X\times\partial X$ to $X\times X$. Thus, by [32, Prop. 3.2] also $\mathbb{H}$ generates a cosine operator function on $\mathbb{X}$. ∎ ###### Remark 5.8. Stability criteria like those stated in Theorems 3.4 and 4.7 could be easily formulated also in the contexts of Theorems 5.4 and 5.7. However, little is currently known about the asymptotical behavior of strongly damped systems, thus such criteria could be hardly checked in concrete cases. The following can be proved similarly to Proposition 4.9. ###### Proposition 5.9. Let Theorem 5.7 apply. Then the solution $u$ to $({\rm AIBVP}^{2}_{b})$ satisfies abstract Wentzell-type boundary conditions $L\left(Au(t)+C\dot{u}(t)\right)=B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}Lu(t)+B_{4}L\dot{u}(t),\qquad t>0.$ If further $A$ maps $D(C)$ into $Y$, then $u$ satisfies in fact $LAu(t)+LC\dot{u}(t)=B_{1}u(t)+B_{2}\dot{u}(t)+B_{3}Lu(t)+B_{4}L\dot{u}(t),\qquad t>0.$ ###### Example 5.10. We consider the initial value problem associated with the system $\left\\{\begin{array}[]{rcll}\ddot{u}(t,x)&=&\alpha u^{\prime\prime}(t,x)+\dot{u}^{\prime\prime}(t,x),&t\geq 0,\;x\in(0,1),\\\ u(t,0)&=&\dot{u}(t,0)=0,&t\geq 0,\\\ \ddot{u}(t,1)&=&-\beta_{1}u^{\prime}(t,1)-\beta_{2}\dot{u}^{\prime}(t,1)+\beta_{3}u(t,1)+\beta_{4}\dot{u}(t,1),&t\geq 0,\\\ \end{array}\right.$ where $\alpha,\beta_{1},\beta_{2},\beta_{3},\beta_{4}\in\mathbb{C}$. A similar problem has been discussed in [21, § 4] and [10, § 4], in a Hilbert space setting only. Reformulate it as $({\mathbb{ACP}})$ by setting $X:=L^{p}(0,1),\qquad Y:=\left\\{u\in W^{2,p}(0,1):u(0)=0\right\\},\qquad{\partial X}:={\mathbb{C}},$ for $1\leq p<\infty$. We define the linear operators $A:=\alpha\frac{d^{2}}{dx^{2}},\qquad C:=\frac{d^{2}}{dx^{2}},\qquad D({A}):=D(C):=Y,$ $Lu:=u(1)\qquad\qquad\hbox{for all }u\in D(L):=Y,$ $B_{i}u:=-\beta_{i}u^{\prime}(1),\qquad\hbox{for all }u\in D(B_{i}):=Y,\;i=1,2,$ $B_{i}:=\beta_{i},\qquad i=3,4.$ Thus, $\ker(L)=W^{2,p}(0,1)\cap W^{1,p}_{0}(0,1)$ and $C_{0}:=C_{\arrowvert\ker(L)}$ is the second derivative with Dirichlet boundary conditions, the generator of a cosine operator function on $L^{p}(0,1)$. One sees moreover that all the Assumptions 2.1 and 5.5 are satisfied. Since $B_{1},B_{2}\in{\mathcal{L}}(W^{2,p}(0,1),{\mathbb{C}})$, the above initial-boundary value problem is governed by an analytic semigroup. This semigroup yields a solution that satisfies Wentzell-type boundary conditions for $t>0$. ∎ ## 6\. A technical result Consider the operator matrix ${\mathcal{H}}:=\begin{pmatrix}H&J\\\ K&L\end{pmatrix},\qquad D({\mathcal{H}}):=D(H)\times D(L).$ on a product Banach space $E\times F$, where $H:E\to E$ and $L:F\to F$ are closed operators. Our aim is to discuss the stability of the semigroup generated by $\mathcal{H}$, in the spirit of [35]. In fact, if $J\in{\mathcal{L}}(F,E)$ and $K\in{\mathcal{L}}(E,F)$, then $\mathcal{H}$ generates a $C_{0}$-semigroup on $E\times F$ if and only if $H$ and $L$ generate $C_{0}$-semigroups on $E$ and $F$, respectively, and in this case $(e^{t\mathcal{H}})_{t\geq 0}$ is given by the Dyson–Phillips series $\sum_{k=0}^{\infty}S_{k}(t),\qquad t\geq 0,$ where $S_{0}(t):=\begin{pmatrix}e^{tH}&0\\\ 0&e^{tL}\end{pmatrix},\qquad t\geq 0,$ and $S_{k}(t):=\int_{0}^{t}S_{0}(t-s)\begin{pmatrix}0&J\\\ K&0\end{pmatrix}S_{k-1}(s)ds,\qquad t\geq 0,\;k=1,2,\ldots.$ If we denote by $S_{k}^{(ij)}(t)$ the $(i,j)$-entry of the operator matrix $S_{k}(t)$, $t\geq 0$, $1\leq i,j\leq 2$, then a direct matrix computation shows that $\displaystyle S_{k}(t)\begin{pmatrix}x\\\ y\end{pmatrix}$ $\displaystyle=$ $\displaystyle\int_{0}^{t}\begin{pmatrix}e^{(t-s)H}&0\\\ 0&e^{(t-s)L}\end{pmatrix}\begin{pmatrix}JS_{k-1}^{(21)}(s)x+JS_{k-1}^{(22)}(s)y\\\ KS_{k-1}^{(11)}(s)x+KS_{k-1}^{(12)}(s)y\end{pmatrix}ds$ $\displaystyle=$ $\displaystyle\int_{0}^{t}\begin{pmatrix}e^{(t-s)H}JS_{k-1}^{(21)}(s)x+e^{(t-s)H}JS_{k-1}^{(22)}(s)y\\\ e^{(t-s)L}KS_{k-1}^{(11)}(s)x+e^{(t-s)L}KS_{k-1}^{(12)}(s)y\end{pmatrix}ds,$ i.e., $(S_{k}(t))_{t\geq 0}$ can be expressed in terms of vector-valued convolution, (6.1) $S_{k}(t)=\begin{pmatrix}e^{\cdot H}JS_{k-1}^{(21)}(\cdot)&e^{\cdot H}JS_{k-1}^{(22)}(\cdot)\\\ e^{\cdot L}KS_{k-1}^{(11)}(\cdot)&e^{\cdot L}KS_{k-1}^{(12)}(\cdot)\end{pmatrix}(t),\qquad t\geq 0.$ By known results on vector-valued convolution we can now obtain the following. ###### Proposition 6.1. Let $M_{1},M_{2}\geq 1$ and $\epsilon_{1},\epsilon_{2}\leq 0$ be constants such that (6.2) $\\|e^{tH}\\|\leq M_{1}e^{\epsilon_{1}t}\qquad\hbox{and}\qquad\\|e^{tL}\\|\leq M_{2}e^{\epsilon_{2}t},\qquad t\geq 0.$ Then the following assertions hold. 1. (1) Let $J=0$. If $\epsilon_{1}<0$ or $\epsilon_{2}<0$, then the semigroup $(e^{t\mathcal{H}})_{t\geq 0}$ is bounded. 2. (2) Let $J=0$. If $(e^{tL})_{t\geq 0}$ is analytic and $\sigma(H)\cap\sigma(L)\cap i{\mathbb{R}}=\emptyset$, then $(e^{t\mathcal{H}})_{t\geq 0}$ is bounded. 3. (3) Let both $\epsilon_{1}<0$ and $\epsilon_{2}<0$. Assume that $M:=\frac{M_{1}M_{2}\\|J\\|\\|K\\|}{\epsilon_{1}\epsilon_{2}}<1.$ Then $(e^{t\mathcal{H}})_{t\geq 0}$ is uniformly exponentially stable. ###### Proof. To begin with, one can prove by complete induction over $n$ that $S_{2n}^{(12)}=S_{2n}^{(21)}=S_{2n+1}^{(11)}=S_{2n+1}^{(22)}=0,\qquad n\in\mathbb{N}.$ (1) If $\epsilon_{1}<0$ and $K=0$, then one can check that (6.3) $S^{(12)}_{1}=e^{\cdot H}f:=e^{\cdot H}Je^{\cdot L}$ and moreover $S^{21}_{1}=0$ as well as $S_{k}=0$ for all $k=2,3,\ldots$. We first consider the case $\epsilon_{1}<0$. Then, by the Datko–Pazy theorem $e^{\cdot H}x\in L^{1}({\mathbb{R}}_{+},E)$ for all $x\in E$. Since $e^{\cdot L}y\in L^{\infty}({\mathbb{R}}_{+},F)$ for all $y\in F$, by the vector-valued Young inequality (see [4, Prop. 1.3.5]) we see that $S^{(12)}_{1}y\in L^{\infty}({\mathbb{R}}_{+},E)$ for all $y\in F$. The case $\epsilon_{2}<0$ can be treated similarly. Thus, we have shown that $e^{\cdot\mathcal{H}}=S_{0}(\cdot)+S_{1}(\cdot)\in L^{\infty}({\mathbb{R}}_{+},{\mathcal{L}}(E\times F))$. (2) The Laplace transform $\hat{f}(\lambda)$ of $f$ defined in (6.3) is given by $KR(\lambda,L)$ for all $\lambda$ with ${\rm Re}\lambda>0$. This yields that the half-line spectrum ${\rm sp}(f)$ of $f$, defined as in [4, § 4.4], is given by $\\{\eta\in{\mathbb{R}}:i\eta\in\sigma(H)\\}$. Thus, [4, Thm. 5.6.5] yields that $S^{(12)}_{1}y\in L^{\infty}({\mathbb{R}}_{+},E)$ for all $y\in F$, and again $e^{\cdot\mathcal{H}}\in L^{\infty}({\mathbb{R}}_{+},{\mathcal{L}}(E\times F))$. (3) We prove by complete induction over $n$ that the estimates (6.4) $\\|S_{2n}^{(11)}(\cdot)x\\|_{L^{1}({\mathbb{R}}_{+},E)}\leq M^{n}\frac{M_{1}}{\|\epsilon_{1}\|}\\|x\\|,\qquad x\in E,$ (6.5) $\\|S_{2n}^{(22)}(\cdot)y\\|_{L^{1}({\mathbb{R}}_{+},F)}\leq M^{n}\frac{M_{2}}{\|\epsilon_{2}\|}\\|y\\|,\qquad y\in F,$ (6.6) $\\|S_{2n+1}^{(12)}(\cdot)y\\|_{L^{1}({\mathbb{R}}_{+},E)}\leq M^{n}\frac{M_{1}M_{2}\\|J\\|}{\epsilon_{1}\epsilon_{2}}\\|y\\|,\qquad y\in F,$ (6.7) $\\|S_{2n+1}^{(21)}(\cdot)x\\|_{L^{1}({\mathbb{R}}_{+},F)}\leq M^{n}\frac{M_{1}M_{2}\\|K\\|}{\epsilon_{1}\epsilon_{2}}\\|x\\|,\qquad x\in E,$ hold for all $k\in\mathbb{N}$. For $n=0$, one sees that (6.4)–(6.5) are satisfied, since by (6.2) $\\|e^{\cdot H}x\\|_{L^{1}}\leq-\frac{M_{1}}{\epsilon_{1}}\\|x\\|$ and $\\|e^{\cdot L}y\\|_{L^{1}}\leq-\frac{M_{2}}{\epsilon_{2}}\\|x\\|$. Moreover, $S_{1}^{(12)}(t)y=(e^{\cdot H}Je^{\cdot L})(t)y$ by (6.1). By the Young inequality we also obtain $\\|S_{1}^{(12)}(\cdot)y\\|_{L^{1}}=\\|e^{\cdot H}Je^{\cdot L}y\\|_{L^{1}}\leq\\|e^{\cdot H}\\|_{L^{1}}\\|J\\|\\|e^{\cdot L}y\\|_{L^{1}}\leq\frac{M_{1}M_{2}\\|J\\|}{\epsilon_{1}\epsilon_{2}}\\|y\\|.$ Likewise one can show that $\\|S_{1}^{(21)}(\cdot)x\\|_{L^{1}}=\\|e^{\cdot L}Ke^{\cdot H}x\\|_{L^{1}}\leq\frac{M_{1}M_{2}\\|K\\|}{\epsilon_{1}\epsilon_{2}}\\|x\\|$, thus the above inequalities hold for $n=0$. Assume now that they hold for $n$. Then for one applies the Young inequality and obtains $\displaystyle\\|S_{2n+2}^{(11)}(\cdot)x\\|_{L^{1}}$ $\displaystyle=$ $\displaystyle\\|e^{\cdot H}JS_{2n+1}^{(21)}(\cdot)x\\|_{L^{1}}\leq\\|e^{\cdot H}\\|_{L^{1}}\\|J\\|\\|S_{2n+1}^{(21)}(\cdot)x\\|_{L^{1}}$ $\displaystyle\leq$ $\displaystyle M^{n}\frac{M_{1}M_{2}\\|J\\|\\|K\\|}{\epsilon_{1}\epsilon_{2}}\frac{M_{1}}{\|\epsilon_{1}\|}\\|x\\|=M^{n+1}\frac{M_{1}}{\|\epsilon_{1}\|}\\|x\\|.$ The remaining three estimates can be proven likewise, leading to $\\|S_{2n}(\cdot)\begin{pmatrix}x\\\ y\end{pmatrix}\\|_{L^{1}}+\\|S_{2n+1}(\cdot)\begin{pmatrix}x\\\ y\end{pmatrix}\\|_{L^{1}}\leq M^{n}M_{0}\left\\|\begin{pmatrix}x\\\ y\end{pmatrix}\right\\|,\quad n\in{\mathbb{N}},\;\begin{pmatrix}x\\\ y\end{pmatrix}\in E\times F.$ where $M_{0}:=\left(\frac{M_{1}}{\|\epsilon_{1}\|}+\frac{M_{2}}{\|\epsilon_{2}\|}+\frac{M_{1}M_{2}\\|J\\|}{\epsilon_{1}\epsilon_{2}}+\frac{M_{1}M_{2}\\|K\\|}{\epsilon_{1}\epsilon_{2}}\right)$. We now prove the proposition’s claim. By assumption $M<1$, thus by the dominated convergence theorem $\displaystyle\int_{0}^{\infty}\\|e^{t\mathcal{H}}\begin{pmatrix}x\\\ y\end{pmatrix}\\|dt$ $\displaystyle\leq$ $\displaystyle\int_{0}^{\infty}\left(\sum_{k=0}^{\infty}\\|S_{k}(t)\begin{pmatrix}x\\\ y\end{pmatrix}\\|\right)dt\leq\sum_{k=0}^{\infty}\\|S_{k}(\cdot)\begin{pmatrix}x\\\ y\end{pmatrix}\\|_{L^{1}({\mathbb{R}}_{+},E\times F)}$ $\displaystyle\leq$ $\displaystyle M_{0}\sum_{k=0}^{\infty}M^{k}\left\\|\begin{pmatrix}x\\\ y\end{pmatrix}\right\\|=\frac{M_{0}}{1-M}\left\\|\begin{pmatrix}x\\\ y\end{pmatrix}\right\\|,\qquad\begin{pmatrix}x\\\ y\end{pmatrix}\in E\times F.$ By the theorem of Datko–Pazy, this concludes the proof. ∎ ## References [1] F. Ali Mehmeti, _Nonlinear Waves in Networks_ , Math. Research 80, Akademie Verlag, Berlin, 1994. * [2] F. Ali Mehmeti, J. von Below, and S. Nicaise (eds.), _PDE’s on Multistructures_ (Proceedings Luminy/France 1999), Lect. Notes Pure Appl. Math. 219, Marcel Dekker 2001. * [3] K.T. Andrews, K.L. Kuttler, and M. Shillor, _Second order evolution equations with dynamic boundary conditions_ , J. Math. Anal. Appl., 197 (1996), 781–795. * [4] W. Arendt, C.J.K. Batty, M. Hieber, and F. Neubrander, “Vector-valued Laplace Transforms and Cauchy Problems,” Monographs in Mathematics 96, Birkhäuser, Basel, 2001. * [5] W. Arendt, G. Metafune, D. Pallara, and S. Romanelli, _The Laplacian with Wentzell–Robin boundary conditions on spaces of continuous functions_ , Semigroup Forum, 67 (2003), 247–261. * [6] A. Bátkai and K.-J. Engel, _Abstract wave equations with generalized Wentzell boundary conditions_. J. Diff. Equations, 207 (2004), 1–20. * [7] J.T. Beale, _Spectral properties of an acoustic boundary condition_ , Indiana Univ. Math. J. 25 (1976), 895–917. * [8] B.P. Belinsky, _Wave propagation in the ice-covered ocean wave guide and operator polynomials_ , in “Proceedings of the Second ISAAC Congress” (Fukuoka 1999) (eds. H.G.W. Begehr, R.P. Gilbert, and J. Kajiwara), Kluwer Academic Publishers, Dordrecht, 2000, 1319–1333. * [9] V. Casarino, K.-J. Engel, R. Nagel, and G. Nickel, _A semigroup approach to boundary feedback systems_ , Integral Equations Oper. Theory, 47 (2003), 289–306. * [10] V. Casarino, K.-J. Engel, G. Nickel, and S. Piazzera, _Decoupling techniques for wave equations with dynamic boundary conditions_ , Disc. Cont. Dyn. Syst. A, 12 (2005), 761–772. * [11] G. Chen and D.L. Russell, _A mathematical model for linear elastic systems with structural damping_ , Q. Appl. Math., 39 (1982), 433–454. * [12] S.P. Chen and R. Triggiani, _Proof of two conjectures by G. Chen and D. L. Russell on structural damping for elastic systems_ , in “Approximation and optimization” (Proceedings Havana 1987) (eds. J.A. Gómez Fernández, F. Guerra, M. A. Jiménez Pozo, and G. López Lagomasino), Lect. Notes Math. 1354, Springer-Verlag, Berlin, 1988, 234–256. * [13] S.P. Chen and R. Triggiani, _Proof of extensions of two conjectures on structural damping for elastic systems_ , Pacific J. Math,. 136 (1989), 15–55. * [14] R. Chill and S. Srivastava, _$L^{p}$ -maximal regularity for second order Cauchy problems_, Math. Z., 251 (2005), 751–781. * [15] R. Chill and S. Srivastava, _$L^{p}$ maximal regularity for second order Cauchy problems is independent of $p$_, Boll. Unione Mat. Ital. 9 (2008), 147–157. * [16] R. Dautray and J.-L. Lions, _Analyse mathématique et calcul numérique pour les sciences et les techniques, vol. 8_ , Masson, Paris, 1988. * [17] K.-J. Engel and R. Nagel, “One-Parameter Semigroups for Linear Evolution Equations,” Graduate Texts in Mathematics 194, Springer-Verlag, New York, 2000. * [18] A. Favini, G.R. Goldstein, J.A. Goldstein, and S. Romanelli, _The one dimensional wave equation with Wentzell boundary conditions_ , in “Differential Equations and Control Theory” (Proceedings Athens 2000) (eds. S. Aizicovici and N.H. Pavel), Lecture Notes in Pure and Applied Mathematics 225, Marcel Dekker, New York, 2001, 139–145. * [19] C. Gal, G.R. Goldstein, and J.A. Goldstein, _Oscillatory boundary conditions for acoustic wave equations_ , J. Evol. Equations 3 (2003), 623–636. * [20] G.R. Goldstein, _Derivation and physical interpretation of general boundary conditions_ , Adv. Differ. Equ.. 11 (2006), 457-480. * [21] M. Grobbelaar-Van Dalsen, _On fractional powers of a closed pair of operators and a damped wave equation with dynamic boundary conditions_ , Appl. Anal., 53 (1994), 41–54. * [22] V. Keyantuo and M. Warma, _The wave equation with Wentzell–Robin boundary conditions on $L^{p}$-spaces_, J. Diff. Equations, 229 (2006), 680-697. * [23] M. Kramar, D. Mugnolo, and R. Nagel, _Semigroups for initial-boundary value problems_ , in “Evolution Equations 2000: Applications to Physics, Industry, Life Sciences and Economics” (Proceedings Levico Terme 2000) (eds. M. Iannelli and G. Lumer), Progress in Nonlinear Differential Equations, Birkhäuser, Basel, 2003, 277–297. * [24] M. Kramar Fijavž and D. Mugnolo, and E. Sikolya, _Variational and semigroup methods for waves and diffusion in networks_ , Appl. Math. Optimization 55, 219–240. * [25] V.N. Krasil’nikov, _On the solution of some boundary-contact problems of linear hydrodynamics_ , J. Appl. Math. Mech. 25 (1961), 1134–1141. * [26] J.E. Lagnese, G. Leugering, and E.J.P.G. Schmidt, _Modeling, Analysis, and Control of Dynamic Elastic Multi-Link Structures_ , Syst. Contr.: Found. Appl., Birkhäuser 1994. * [27] P. Lancaster, A. Shkalikov, and Q. Ye, _Strongly definitizable linear pencils in Hilbert space_ , Integral Equations Oper. Theory, 17 (1993), 338–360. * [28] J. Liang, R. Nagel, and T.-J. Xiao, _Approximation theorems for the propagators of higher order abstract Cauchy problems_ , Trans. Am. Math. Soc. 360 2008, 1723-1739. * [29] W. Littman, _The wave operator and $L_{p}$ norms_, J. Math. Mech. 12 (1963), 55–68. * [30] A. Lunardi, _Analytic Semigroups and Optimal Regularity in Parabolic Problems_ , Progress in Nonlinear Differential Equations and their Applications 16, Birkhäuser, Basel, 1995. * [31] D. Mugnolo, _Operator matrices as generators of cosine operator functions_ , Integral Equations Oper. Theory, 54 (2006), 441-464. * [32] D. Mugnolo, _Matrix methods for wave equations_. Math. Z., 253 (2006), 667-680. * [33] D. Mugnolo, _Gaussian estimates for a heat equation on a network_ , Netw. Heterog. Media 2 (2007), 55-79. * [34] D. Mugnolo, _A variational approach to strongly damped wave equations_. in “Functional Analysis and Evolution Equations – The Günter Lumer Volume” (eds. H. Amann et al.), Birkhäuser, Basel, 2008, 503–514. * [35] R. Nagel, _Towards a “matrix theory” for unbounded operator matrices_ , Math. Z., 201 (1989), 57–68. * [36] F. Neubrander, _Well-posedness of higher order abstract Cauchy problems_ , Trans. Am. Math. Soc., 295 (1986), 257-290. * [37] S. Nicaise and J. Valein, _Stabilization of the wave equation on 1-D networks with a delay term in the nodal feedbacks_ , Netw. Heterog. Media 2 (2007), 425-479. * [38] T.-J. Xiao and J. Liang, _The Cauchy Problem for Higher-Order Abstract Differential Equations_ , Lecture Notes in Mathematics 1701, Springer-Verlag, Berlin 1998. * [39] T.-J. Xiao and J. Liang, _Complete second order evolution equations in Banach spaces with dynamic boundary conditions_ , J. Differential Equations, 200 (2004), 105–136. * [40] T.-J. Xiao and J. Liang, _Second order parabolic equations in Banach spaces with dynamic boundary conditions_ , Trans. Am. Math. Soc., 356 (2004), 4787–4809.	arxiv-papers	2008-08-01T23:11:50	2024-09-04T02:48:57.126632	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Delio Mugnolo", "submitter": "Delio Mugnolo", "url": "https://arxiv.org/abs/0808.0213" }
0808.0242	# Geometric Phase and Quantum Phase Transition : Two-Band Model H. T. Cui cuiht@aynu.edu.cn Department of Physics, Anyang Normal University, Anyang 455000, China Jie Yi School of Physical Science and Technology, Heilongjiang University, Harbin 150018, China ###### Abstract The connection between the geometric phase and quantum phase transition has been discussed extensively in the two-band model. By introducing the twist operator, the geometric phase can be defined by calculating its ground-state expectation value. In contrast to the previous numerical examinations, our discussion presents an exact calculation for the determination of the geometric phase. Through two representative examples, our calculation shows the intimate connection between the geometric phase and phase transition: different behaviors of the geometric phase can be identified in this paper, which are directly related to the energy gap above the ground state. ###### pacs: 75.10.Pq, 03.65.Vf, 05.30.Pr, 42.50.Vk ## I introduction Recently, quantum phase transition sachdev has received great attention due to its intimate correlation to the fundamental principles of quantum mechanics, especially to the concept of quantum entanglement preskill ; osterloh ; wu ; vidal ( see Ref.afov07 for a comprehensive review ). In general the quantum phase transition happens when the degeneracy of the ground states occurs, which cannot be characterized completely by the pattern of symmetry broken ( order parameters of some kind ). Instead the universal quantum order or topological order is needed for the description of properties of the ground state in many-body systems wen . Recently the connection between the geometric phase of the ground state and the quantum criticality has been displayed in spin-chain systems by displaying the singularity of the geometric phase closed to the critical points carollo ; zhu . Moreover in Ref zhu the author showed that the scaling behavior of the geometric phase of the ground state near the critical points can also display the universal class of the phase transitions. Many works have been devoted to this interesting issue cui ; rhp07 ; Zanardi ; pv ( or see Ref.zhu08 for a review ). Although great progresses has been made in the understanding of quantum phase transition from the fundamental principles of quantum mechanics, an important question is not yet resolved: whether there exists a universal way to characterize the different phases and their boundaries. More specifically, could the geometric phase of the ground state in many-body systems serve this purpose? This conjecture is natural since the quantum phase transition generally emerges from the degeneracy of the ground state in many-body systems. Geometric phase as a measurement of the curvature of the Hilbert space, could mark the fantastic changes of the ground state when degeneracy happens. However the critical point is how to obtain the geometric phase of the ground state. The earlier method is to impose a local rotation about some special orientations, as has been done in carollo ; zhu ; cui . In our own viewpoints, the connection built by this method is fragile; the geometric phase is trivial when the system is symmetrical about this rotation. Recently the differential information-geometry analysis of quantum fidelity in many-body systems displayed the intimate correlation between quantum phase transition and the singularity of fidelity between the states across the transition point Zanardi . In these papers the quantum geometric tensor, which is intimately related to the degeneracy of the ground state, was introduced for the determination of the fidelity. As shown in these papers, the imaginary part of this tensor actually described the curvature two-form whose holonomy is the Berry phase, and the degeneracy of the ground state would induce the singularity of fidelity Zanardi . However, in our opinion, it seems not transparent to directly define the geometric phase in this coupling-parameter space since a cyclic evolution may be difficult to construct. Moreover, the explicit expression for the ground state is necessary for the construction of this tensor, which in general is difficult. Furthermore it is also unknown in the case where the degeneracy of the ground state is broken. In Ref. rhp07 , the Bargmann phase, a generalization of the Berry phase, has also been constructed for detecting the phase transition in many-body systems. However the connection between degeneracy of the ground state and the Bargmann phase was unclear in this case since there was a lack of simple interpretational underpinnings for the Bargmann phase in terms of physical adiabatic processes rhp07 . With respect to the points stated above, it is urgent to find a popular way for construction of geometric phase in many-body systems. For this purpose, a nonlocal operator -the twist operator- is introduced to obtain the geometric phase of the ground state in this paper. Our calculation shows that the geometric phase, decided by the ground-state expectation value of the twist operator, can serve as the quantity to distinguish different phases and boundaries between them, as shown below. The general form for the twist operator can be written as in the lattice systems, $\eta=\exp(\frac{2\pi i}{N}\sum_{x}xn_{x}),$ (1) in which $N$ is the number of lattice sites, $x$ denotes the coordinates of lattice sites and $n_{x}$ is generally related to the physical quantities located at site $x$, such as the total spin or charge, or particle number at site $x$ and so on. The twist operator was first introduced by Lieb, Schultz and Mattis for the proof of gapless excitation in one-dimensional spin-1/2 chainslsm . Then Resta pointed out that its ground state expectation was direct to the Berry-phase theory of polarization in strongly correlated electron systems resta . Moreover Aligia found that the ground-state expectation value of the twist operator Eq. (1) allows one to discriminate conducting from nonconducting phase in the extended quantum systems aligia . The vanishing of the ground-state expectation value, i.e. $\eta=0$, has been shown the ability to detect the boundaries of different valence-bond-solid phases in spin chains naka . However these studies were implemented in some special examples and a general discussion was absent. Moreover since the previous calculations were numerical or approximate, the details of $\eta$ adjacent to the phase transition points are unclear. Hence it is of great interest to find the exact expression for $\eta$, even for special cases. Our paper serves this purpose, and the exact results can be found for a special case. It is of great interest to note that the twist operator $\eta$ actually creates a wave-like excitation since it rotates all the particles with a relative angle between the neighboring lattices $2\pi/N$lsm . Under the large $N$ limit the ground state has an adiabatic variation, and its ground-state expectation value is exactly a geometric factor, of which an argument is the geometric phase. Applying $\eta$ to the unique ground state, one obtains a low-lying excited state. The important quantity is the overlap between the ground state and this excited state, i.e. $z=\langle g\|\eta\|g\rangle,$ (2) in which $\|g\rangle$ denotes the many-body ground state. In general $z$ is a complex number and its argument is just the geometric phase, determined by $\gamma_{g}=\text{Arg}[z]=i\int_{0}^{2\pi}d\phi\langle g(\phi)\|\partial_{\phi}\|g(\phi)\rangle,$ (3) in which $\|g(\phi)\rangle=\exp(\frac{i\phi}{N}\sum_{x}x\hat{n}_{x})\|g\rangle$. Since $\gamma_{g}$ in fact came from the continued deformation of the boundary condition of systems aligia and then slightly related to the symmetry of the Hamiltonian, this construction of the geometric phase is more popular than the previous method. An important character is that $\gamma_{g}$ is related to the correlation functions for the ground state, and numerical evaluation could be implemented efficiently resta ; aligia ; naka . It is an immediate speculation that $z$ and $\gamma_{g}$ may be singular near the critical points, where the degeneracy of the ground state happens and the macroscopic properties of the system have fantastic changes. However, our findings are more subtle; the exact calculations show an unexpected ability for $\gamma_{g}$ or $z$ to distinguish the different phases in many-body systems; one case is that $z$ tends to be zero and then $\gamma_{g}$ is ill- defined when one approaches the phase transition points, in which the degeneracy of the ground state happens. The other is that $\gamma_{g}$ has different values for different phases and displays the singularity at the transition points, where no degeneracy happens. The physical reason, as shown in the following discussions, is directly related to the energy gap above the ground state. The paper is organized as follows. In Sec.II, the exact expression of $z$ and the geometric phase $\gamma_{g}$ are presented in the two-band model for a special case, in which the ground state is the filled Fermi sea. In Sec. III, two representative examples are provided for the demonstration of this connection. One is the $D$-dimensional free-fermion model, in which there are quantum phase transitions originated from the ground-state-energy degeneracy. The other example is the Su-Schrieffer-Heeger ( SSH ) model, in which the energy gap is non-vanishing at the quantum phase transition points and a topological order, defined by the geometric phase $\gamma_{g}$, provides a clear description of the phase diagram for this model. ## II two-band model Consider the one-dimensional (1D) translational invariant Hamiltonian with two bands separated by a finite gap ryu ; ryu2 , $H=\sum_{x,x^{\prime}}\mathbf{c}_{x}^{\dagger}\mathscr{H}_{x,x^{\prime}}\mathbf{c}_{x^{\prime}},$ (4) in which $\mathbf{c}_{x}^{(\dagger)}=(c_{+},c_{-})_{x}^{\textrm{T}(\dagger)}$ defines a pair of fermion annihilation (creation) operators for each site $x,x^{\prime}=1,2,\dots,N$ and the form of $\mathbf{c}_{x}$ is decided completely by the Hamiltonian. $\mathscr{H}_{x,x^{\prime}}$ is a $2\times 2$ matrix and its elements can be determined by the hermiticity of Eq. (4), $\mathscr{H}_{x,x^{\prime}}=\left(\begin{array}[]{cc}A_{+}&B\\\ C&A_{-}\end{array}\right)_{x,x^{\prime}}$ (5) in which $A_{\pm,xx^{\prime}}=A_{\pm,x^{\prime}x}^{}$ and $B_{x,x^{\prime}}=C^{}_{x^{\prime},x}$. Although our discussion is restricted to 1D system in this section, it should point out that this situation can be easily generalized to higher dimension systems. Without the loss of generality, it is conventional for the translational invariant system to impose the periodic boundary condition. In spite of the simplicity, the Hamiltonian Eq. (4) has a wide range of applications, such as the Bogoliubov- de Gennes Hamiltonian in superconductivity, graphite systems ryu . Applying the Fourier transformation and considering the periodic boundary condition, $\mathbf{c}_{x}=1/\sqrt{N}\sum_{k}e^{ikx}\mathbf{c}_{k}$ in which $k=2\pi n/N$ with $n=1,2,\dots,N$. Then the Hamiltonian in the momentum space can be written generally as $H=\sum_{k}\mathbf{c}_{k}^{\dagger}\mathscr{H}(k)\mathbf{c}_{k}$. If one introduces the four-vector $R_{\mu}(k)(\mu=0,x,y,z)$, then the Hamiltonian can be rewritten as, $H=\sum_{\mu}\sum_{k}\mathbf{c}_{k}^{\dagger}R_{\mu}(k)\sigma_{\mu}\mathbf{c}_{k},$ (6) in which $\sigma_{0}$ is a $2\times 2$ unit matrix and $\sigma_{i}(i=x,y,z)$ is the Pauli operators. Obviously Eq. (6) can be diagonalized by finding the eigenvectors $\nu_{\pm}$ of $\sum_{i}R_{i}(k)\sigma_{i}=\mathbf{R}(k)\cdot\mathbf{\sigma}$, in which the vector $\mathbf{R}(k)$ is similar to the Bloch vector for the density operator in the $2\times 2$ Hilbert space footnote , and furthermore should also satisfy the relation $(R_{x}(k),R_{y}(k))\neq(0,0)$ since in this case one has a trivial geometric phase berry , $\nu_{\pm}=\frac{1}{\sqrt{2R(k)(R(k)\mp R_{z}(k))}}\left(\begin{array}[]{c}R_{x}(k)-iR_{y}(k)\\\ \pm R(k)-R_{z}(k)\end{array}\right)$ (7) in which $R(k)=\|\mathbf{R}(k)\|$ and the corresponding eigenvalues are $E_{\pm}=R_{0}(k)\pm R_{z}(k)$. Obviously there is a finite gap between the two bands since $E_{+}>E_{-}$. The ground state is defined as the filled Fermi sea $\|g\rangle=\prod_{k}\beta_{-,k}^{\dagger}\|0\rangle_{k}$, in which $\beta_{-,k}^{\dagger}=\mathbf{c}_{k}^{\dagger}\nu_{-}$ and $\|0\rangle_{k}$ is the vacuum state of $c_{\pm k}$. Now it is time to determine the geometric phase Eq. (3), given by the ground- state expectation value of the twist operator Eq. (2). In this model the twist operator can be expressed explicitly as $\eta=\exp(\frac{2\pi i}{N}\sum_{x}x\mathbf{c}_{x}^{\dagger}\mathbf{c}_{x}),$ (8) in which $n_{x}=\mathbf{c}_{x}^{\dagger}\mathbf{c}_{x}=c_{+}^{\dagger}c_{+}+c_{-}^{\dagger}c_{-}$. Seemingly $\mathbf{c}_{x}^{\dagger}\mathbf{c}_{x}$ could define the particle number at the site $x$, however the physical meanings for it may be different for different systems, dependent on one’s interests; for spin systems it may denote the total spin at site $x$, and for electron systems it may also denote the total charge number at site $x$. The geometric phases in the both situations have been defined respectively as the spin Berry phase and the charge Berry phase, which have extensive applications in determining the phase diagram in strongly correlated electron systems yama . It is a crucial step to determine $z$. First with the periodic boundary condition, one can rewrite $\eta$ in the moment space $\eta=\exp(-\frac{2\pi}{N}\sum_{k}\mathbf{c}_{k}^{\dagger}\partial_{k}\mathbf{c}_{k}).$ (9) Now, introduce the new fermion operators $\beta_{\pm,k}=\nu_{\pm}^{\dagger}\mathbf{c}_{k},$ (10) in which $\nu_{\pm}$ is defined in Eq. (7) and both of $\beta_{\pm,k}$ satisfy the anti-commutative relation. Then the ground state is defined as the filled Fermi sea $\|g\rangle=\prod_{k}\beta^{\dagger}_{-,k}\|0\rangle_{k}$. Substitute Eq. (10) into Eq. (9) $\eta=\prod_{k}\exp[-\frac{2\pi}{N}(\mathbf{\beta}_{k}^{\dagger}\mathcal{M}\mathbf{\beta}_{k}+\beta_{-,k}^{\dagger}\partial_{k}\beta_{-,k}+\beta_{+,k}^{\dagger}\partial_{k}\beta_{+,k})],$ (11) in which $\mathbf{\beta}_{k}=(\beta_{-,k},\beta_{+,k})^{T}$ and $\mathcal{M}=\left(\begin{array}[]{cc}K^{\prime}&K\\\ -K^{}&K^{\prime}\end{array}\right),$ (12) where $\displaystyle\cos\theta_{k}$ $\displaystyle=$ $\displaystyle\frac{R_{z}(k)}{R(k)},\gamma_{k}=\arctan\frac{R_{y}(k)}{R_{x}(k)}$ $\displaystyle K^{\prime}$ $\displaystyle=$ $\displaystyle-i\sin^{2}\frac{\theta_{k}}{2}\partial_{k}\gamma_{k}$ $\displaystyle K$ $\displaystyle=$ $\displaystyle\frac{e^{i\gamma_{k}}}{2}(\partial_{k}\theta_{k}+i\sin\theta_{k}\partial_{k}\gamma_{k})$ (13) The last two terms in Eq. (11) precludes the further exact calculations. An important case is that if $R_{y}(k)=-R_{y}(-k)$, one then can properly choose $\nu_{\pm}$ so that the new Fermi operator $\beta_{\pm,k}$ can be converted into each other by exchanging $k\leftrightarrow-k$. Then the terms in Eq. (11), $\beta_{-,k}^{\dagger}\partial_{k}\beta_{-,k},\beta_{+,k}^{\dagger}\partial_{k}\beta_{+,k}$ can cancel each other. An exact result in this special case can be obtained for $z$ $z=\langle g\|\eta\|g\rangle=\prod_{k}[1-\frac{\|K\|^{2}}{C_{+}^{2}}(e^{\frac{2\pi}{N}\lambda_{+}}-1)-\frac{\|K\|^{2}}{C_{-}^{2}}(e^{\frac{2\pi}{N}\lambda_{-}}-1)]$ (14) in which $\displaystyle\lambda_{\pm}$ $\displaystyle=$ $\displaystyle\pm i\sqrt{\|K^{\prime}\|^{2}+\|K\|^{2}}$ $\displaystyle C_{\pm}^{2}$ $\displaystyle=$ $\displaystyle-\|K\|^{2}+(\lambda_{\pm}-K^{\prime})^{2}$ (15) and from this formula, the geometric phase $\gamma_{g}$ can also be obtained exactly. The exact determining of $z$ provides the ability to detect the distinguished behaviors of $\gamma_{g}$ or $z$ near the phase transition points. Although the exact results can be obtained only for this special case, some different connections between the geometric phase $\gamma_{g}$ or $z$ and quantum phase transitions are disclosed, as shown in the following calculations. ## III exemplifications In this section two representative models are presented to display the distinguished characteristics of $\gamma_{g}$ or $z$ for the determination of the phase diagram in many-body systems. One is the $D$-dimensional free- fermion model, in which the quantum phase transition is originated from the degeneracy of the ground-state energy li06 . The other is the Su-Schrieffer- Heeger (SSH) model, in which the quantum phase transition happens with the non-vanishing energy gap above the ground state ryu . It is obvious that these examples include two important cases of quantum phase transition; one is the degeneracy of the ground-state energy, whereas not for the other. Moreover the both models are exactly solvable. Figure 1: ( Color online ) The ground-state expectation value of twist operator $z$ for the $D$-dimensional free-fermion model vs the parameter $\lambda$. We have chosen $\gamma=1$ for this plot. The plotting in the inset of (c) shows the details of $\lambda>3$. ### III.1 $D$-dimensional free-fermion model The Hamiltonian is read as $H=\sum_{\langle{\mathbf{i,j}}\rangle}[c_{\mathbf{i}}^{\dagger}c_{\mathbf{j}}-\gamma(c_{\mathbf{i}}^{\dagger}c_{\mathbf{j}}^{\dagger}+h.c.)]-2\lambda\sum_{\mathbf{i}}c_{\mathbf{i}}^{\dagger}c_{\mathbf{i}},$ (16) in which $\langle\mathbf{i,j}\rangle$ denotes the nearest-neighbor lattice sites and $c_{\mathbf{i}}$ is the fermion operator. This Hamiltonian, first introduced in Ref.li06 , depicts the hopping and pairing between nearest- neighbor lattice sites, in which $\lambda$ is the chemical potential and $\gamma$ is the pairing potential. Eq.(16) could be considered as a $D$-dimensional generalization of one-dimensional spin-$1/2$ $XY$ model. However for the $D>1$ case, this model shows some novel phase characteristics li06 . Eq. (16) can be resolved exactly by transforming into moment space with periodic boundary condition. Wiht the help of the Bogoliubov transformation li06 , one has $H=\sum_{\mathbf{k}}2\Lambda_{\mathbf{k}}\eta_{\mathbf{k}}^{\dagger}\eta_{\mathbf{k}}+const.$ (17) in which $\Lambda_{\mathbf{k}}=\sqrt{t_{\mathbf{k}}^{2}+\Delta_{\mathbf{k}}^{2}}$, $t_{\mathbf{k}}=\sum_{\alpha=1}^{D}\cos k_{\alpha}-\lambda$ and $\Delta_{\mathbf{k}}=\gamma\sum_{\alpha=1}^{D}\sin k_{\alpha}$. The phase diagram can be determined based on the gapless excitation $\Lambda_{k}=0$ li06 . For $D=1$, which corresponds to the one dimensional spin-$1/2$ $XY$ model, the energy gap above the ground state is non-vanishing except at $\lambda_{c}=1$ for $\gamma\neq 0$, where a second-order quantum phase transition occurs. For $\gamma=0$ the energy of the ground state is degenerate in the region $\|\lambda\|\leq 1$ and the transition occurs at $\lambda=\pm 1$. When $D=2$, the phases diagram should be identified with respect to two different situations; for $\gamma=0$, the degeneracy of the ground state occurs when $\lambda\in[0,2]$, whereas the gap above the ground state is non- vanishing for $\lambda>2$. However for $\gamma\neq 0$ three different phases can be identified as $\lambda=0$, $\lambda\in(0,2]$ and $\lambda>2$. The first two phases correspond to case that the energy gap for the ground state vanishes, whereas not for $\lambda>2$. One should note that $\gamma=0$ means a well-defined Fermi surface with $k_{x}=k_{y}\pm\pi$, whose symmetry is lowered by the presence of $\gamma$ terms. For $D=3$ two phases can be identified as $\lambda\in[0,3]$ with the vanishing energy gap above the ground state and $\lambda>3$ with a non-vanishing energy gap above ground state. In a word the critical points can be identified as $\lambda_{c}=D(D=1,2,3)$ for any anisotropy of $\gamma$, and $\lambda=0$ for $D=2$ with $\gamma>0$. Defining the fermion-pair operator $\mathbf{c}_{\mathbf{i}}=(c,c^{\dagger})^{T}_{\mathbf{i}}$ and transforming the system into moment space, one then obtain $\displaystyle H=\sum_{\mathbf{k}}\mathbf{c}^{\dagger}_{\mathbf{k}}\left(\begin{array}[]{cc}\lambda-\sum_{\alpha}\cos k_{\alpha}&-i\gamma\sum_{\alpha}\sin k_{\alpha}\\\ i\gamma\sum_{\alpha}\sin k_{\alpha}&-\lambda+\sum_{\alpha}\cos k_{\alpha}\end{array}\right)\mathbf{c}_{\mathbf{k}}$ (20) It is worth noting $\mathbf{R(k)}=(0,\gamma\sum_{\alpha}\sin k_{\alpha},\lambda-\sum_{\alpha}\cos k_{\alpha})$ and $R\mathbf{(k)}=\Lambda_{\mathbf{k}}$. $R_{y}(\mathbf{k})$ is obviously satisfied the requirement $R_{y}(\mathbf{k})=-R_{y}(-\mathbf{k})$. Substituted into Eq. (14), one obtain $\displaystyle\gamma_{g}=\text{Arg}z_{\mu}=\text{Arg}\prod_{\alpha=1}^{D}\prod_{k_{\alpha},k_{\mu}}\cos\frac{2\pi}{N}\|K_{\mu}\|,$ (21) in which $\mu=1\dots D$ and $\displaystyle K_{\mu}=-\frac{i\gamma(\lambda-\sum_{\alpha}\cos k_{\alpha})[\cos k_{\mu}(\lambda-\sum_{\alpha}\cos k_{\alpha})-\sin k_{\mu}\sum_{\alpha}\sin k_{\alpha}]}{2[\gamma^{2}(\sum_{\alpha}\sin k_{\alpha})^{2}+(\lambda-\sum_{\alpha}\cos k_{\alpha})^{2}]^{3/2}}.$ (22) The schematic drawing of $z$ for $D=1,2,3$ have been presented in Figs. 1, in which we have chosen $\gamma=1$ for specification. Some different characters can be found in the figures note . Figure 2: ( Color online )(a) $z$ for $\lambda=1$ ( plotted by a logarithm ) vs the particle number $N$ for $D=1$; (b) the asymptotic behavior of $z$ closed to transition point $\lambda=1$ for $D=1$; (c) the geometric phase $\gamma_{g}$ for $D=1$. We have chosen $\gamma=1$ for all plots. Figure 3: $z$ for $D=2$ vs particle number $N$ at phase transition point $\lambda=2$. $D=1$. It is the well-known one-dimensional $XY$ model for this case, in which there is a quantum phase transition at $\lambda=1$ because of the degeneracy of the ground-state energy. $z$ are plotted with $\lambda$ respectively in Fig. 1 (a). It is obvious that $z$ has an dropping and then an abrupt increment when one approaches the phase transition point $\lambda_{c}=1$. Moreover our calculation shows that $z$ tends to be zero with $\lambda\rightarrow 1$ as shown in Fig. 2 (b), and at exact $\lambda_{c}=1$ $z$ tends to be 1 with the increase of the lattice site number as shown in Fig. 2 (a) note2 . $\gamma_{g}$ have also been plotted in Fig. 2 (c), in which a rapid oscillation happens closed to $\lambda_{c}=1$. These phenomena mean that $\gamma_{g}$ is ill-defined closed to $\lambda=1$ and has an abrupt change at the phase transition point. Since the energy gap vanished only at $\lambda_{c}=1$, the singularity of $\gamma_{g}$ and $z$ would be directly related to the degeneracy of the ground state energy. It also hints that one could mark the transition point by detecting the point where $z=0$. $D=2$. With the increment of dimensionality, the situation becomes more complex. We have plotted $z$ in Fig. 1 (b). It is obvious that two different regions can be identified as $\lambda\in[0,2]$ in which $z$ is disordered, and $\lambda>2$ in which $z$ is an increasing function of $\lambda$. With respect that the disappearance of the energy gap above the ground state happens when $\lambda\in[0,2]$, $z$ presents a clear identification of the phase diagram. It is a reasonable speculation from Fig. 1 (b) that $z$ may tend to be zero under the thermodynamic limit when $\lambda\in[0,2]$. Then $\gamma_{g}$ is under thermodynamic limit $\displaystyle\gamma_{g}=\begin{cases}\text{undetermined},&\lambda\leq 2\\\ 0,&\lambda>2.\end{cases}$ (23) Our calculation also shows that $z$ tends to be zero with the increment of $N$ at the exact transition point $\lambda_{c}=2$, as shown in Fig. 3. Similar to the case of $D=1$, it may be desirable to find the point $z=0$ as a way of detecting the phase transition. $D=3$. This case is very similar to that of $D=2$, except the phase transition happens at $\lambda_{c}=3$. $z$ has been shown in Fig. 1 (c). However in this case $z$ seems unlikely to detect the phase transition since the data in the figure has a smoothing changes at the phase transition point for large $N$, as shown in the inset of Fig. 1 (c). Only for $N=11^{3}$, there is an abrupt changs of $z$ near to the phase transition point. One should note that $z$ tends to be zero when $\lambda\in[0,3]$, in which the energy gap above the ground state disappears. From the discussions above, one can note the great impact of the degeneracy of the ground-state energy on the geometric phase $\gamma_{g}$ or $z$; The degeneracy of the ground-state energy leads to $z=0$ or the ill-defined $\gamma_{g}$. However, this conclusion would not be made until the next example is studied in which the energy gap above the ground state does not disappear. It is interesting to give a further discussion of the geometric phase in this nontrivial case. Unfortunately Eq. (22) seems unsuccessful in characterizing the transitions for $\gamma=0$ (the tight-binding model) since in this case $K_{\mu}$ is completely undetermined when $\lambda-\sum_{\alpha}\cos k_{\alpha}=0$. With respect to Eqs. (16) and (8), it means that $[H,\eta]=0$ since $[\sum_{\langle{\mathbf{i,j}}\rangle}c_{\mathbf{i}}^{\dagger}c_{\mathbf{j}},\sum_{\mathbf{i}}c_{\mathbf{i}}^{\dagger}c_{\mathbf{i}}]=0$ in this special case. Hence our discussion excludes this special situation since one has trivial results. One should note that the phase of $\lambda=0$ with $\gamma>0$ in $D=2$ also cannot be identified by $z$ since the transition comes from the deformation of the Fermi surface instead of the degeneracy of the ground state. ### III.2 Su-Schrieffer-Heeger (SSH) model Figure 4: ( Color online ) $\gamma_{g}$ and $\|z\|$ for the SSH model vs. the parameter $\phi$. Another example is the Su-Schrieffer-Heeger (SSH) model, which is also exactly solvable. The 1D tight-binding Hamiltonian for the SSH model for a chain of polyacetylene is given by heeger $H=\sum_{l=1}^{N}t(-1+(-1)^{l}\phi_{l})(c_{l}^{\dagger}c_{l+1}+h.c.),$ (24) in which $\phi_{l}$ represents the dimerization at the $l$th site and an alternating sign of the hopping elements reflects dimerization between the carbon atoms in the molecule. Without the loss of generality, it is convenient to neglect the kinetic energy in the system and take $\phi_{l}=\phi$, $t=1$ ryu2 . There is a critical point at $\phi=0$ ,which divides the ground states into two different types. One should point out that SSH has a gaped excitation for any $\phi\in[-1,1]$ and the quantum phase transition comes from the excitation of the boundary statesryu2 . One can find from the following calculation that $\gamma_{g}$ can discriminate the two phases and the boundary between them. Defining $c_{x}=(c_{l},c_{l+1})^{T}$, the Hamiltonian becomes $\displaystyle H$ $\displaystyle=$ $\displaystyle\sum_{x=1}^{N}c_{x}^{\dagger}\left(\begin{array}[]{cc}0&-(1+\phi)\\\ -(1+\phi)&0\end{array}\right)c_{x}$ (27) $\displaystyle+$ $\displaystyle\left[c_{x}^{\dagger}\left(\begin{array}[]{cc}0&0\\\ -1+\phi&0\end{array}\right)c_{x+1}+h.c.\right]$ (30) Imposing the periodic boundary condition and Fourier transformation, we then have $\mathscr{H}(k)=\left(\begin{array}[]{cc}0&-(1+\phi)-(1-\phi)e^{-ik}\\\ -(1+\phi)-(1-\phi)e^{ik}&0\end{array}\right),$ (31) It is obvious that $\mathbf{R}(k)=(-(1+\phi)-(1-\phi)\cos k,-(1-\phi)\sin k,0)$ and $R_{y}(k)=-R_{y}(-k)$ is satisfied. Then $z=\prod_{k}(\cos\frac{2\sqrt{2}\pi}{N}\|K^{\prime}\|-\frac{i\text{Im}[K^{\prime}]}{\sqrt{2}\|K^{\prime}\|}\sin\frac{2\sqrt{2}\pi}{N}\|K^{\prime}\|),$ (32) in which $K^{\prime}=-\frac{i}{2}\frac{(1-\phi)^{2}+(1-\phi^{2})\cos k}{(1-\phi)^{2}\sin^{2}k+[1+\phi+(1-\phi)\cos k]^{2}}$. We plot $\gamma_{g}$ against $\phi$ with different site numbers in Fig. 4. It is obvious that $\gamma_{g}$ is $-\pi$ for $\phi\in[-1,0)$ and zero for $\phi\in(0,1]$ from Fig. 4. Moreover $\gamma_{g}$ tends to be $\pi$ at exact phase transition point $\phi=0$. Furthermore our calculation shows that $\|z\|$ is not zero, which means that one cannot detect the phase transition by finding the point $z=0$. Since the energy gap for the ground state is nonvanishing in this model, the geometric phase $\gamma_{g}$ can be well- defined for any $\phi$. With respect to the discussion for the $D$-dimensional free-fermion system, it is evident that $\gamma_{g}$ or $z$ are directly related to the degeneracy of the ground state. ## IV discussions and conclusions Given the two examples, some comments should be presented in this section. In this paper the twist operator Eq. (1) has been introduced and its ground-state expectation $z$ has been calculated to define the geometric phase Eq. (3) for the two-band model. Although the absent of general results, the exact expression of $z$ can be obtained in a special case, which provides the ability to detect the details of the geometric phase adjacent to the phase transition points. With respect to the discussions for two representative examples-D-dimensional free-fermions model and the Su-Schrieffer-Heeger model, some distinguished properties of $\gamma_{g}$ or $z$ have been found in our calculations. First when the degeneracy of the ground state happens, $z$ tends to be zero and the geometric phase $\gamma_{g}$ is ill-defined in this case, as shown in Figs. 1, while $\gamma_{g}$ can be well-defined when the energy of the ground state is nondegenerate, as shown in Fig. 4. This phenomenon clearly displays the intimation connection between geometric phase, defined by the ground-state expectation value of the twist operator, and the degeneracy of the ground state in many-body systems. Consequently one can find the nodal structure of the geometric phase (the situation that the geometric phase is ill-defined because of $z=0$) to detect the phase transition originated from the degeneracy of the ground state. The nodal structure of geometric phase is introduced by Fillip and Sjöqvist for the description of experimental measure of the geometric phase based on the interference, which characterizes the condition for the disappearance of the fringes and then the geometric phase is ill-defined. Second geometric phase can also present the phase diagram even if there is a energy gap above the ground state, as shown in Fig. 4, in which two different phases defined by $\gamma_{g}=-\pi$ and $0$ can be identified and the phase transition point is marked by the discontinued variation of geometric phase. In a word the geometric phase $\gamma_{g}$ displays the ability to mark the phase diagram in this discussion, whether the phase is determined by the degeneracy of the ground state or not. Hence the geometric phase can provide a more popular depiction for the phase transition. However there still exist some problems. First is that the geometric phase seems to fail to characterize the tight-bond model ($\gamma=0$ in Eq. (16)). It is the reason that $[H,\eta]=0$, and the twist operator has a trivial effect on the ground state. Second the geometric phase fails to detect some phase transitions not originated from the degeneracy of the ground state, such as the transition from the deformation of the Fermi surface. Thirdly $\gamma_{g}$ or $z$ seems unable to detect the broken of symmetry which happens in the 1D spin-1/2 $XY$ model, as shown in Fig. 1 (a). Although there exists some defects, the geometric phase defined by the twist operator provides one another way of detecting the phase diagram for many-body systems. Moreover the flexibility of choosing $n_{x}$ in the definition of the twist operator implies that one could properly choose different physical quantities $n_{x}$ for the description of different properties of the system. _Note added_. Recently we become aware of a paper which also focuses on the connection between the geometric phase and the quantum phase transition by numerical evaluation hkh . In this paper, three different phases in gapped spin chains can be defined by the geometric phase $\gamma=0,\pi,$ and undefined respectively, which is similar to our conclusions. _Acknowledgement_ The author(H.T.C.) acknowledges the support of NSF of China, Grant No. 10747195. ## References * (1) Subir Sachdev, Quantum Phase Transition(Cambridge University Press, Cambridge, 1999). * (2) J. Preskill, J. Mod. Opt. 47, 127 (2000). * (3) A. Osterloh, L. Amico, G. Falci, R. Fazio, Nature, 416, 608(2002); T. J. Osborne and M. A. Nielsen, Phys. Rev. A 66, 032110 (2002). * (4) L. A. Wu, M. S. Sarandy, D. A. Lidar, Phys. Rev. Lett. 93, 250404 (2004). * (5) J. I. Latorre, E. Rico and G. Vidal, Quantum Inf. Comput. 4, 48 (2004). * (6) L. Amico, R. Fazio, A. Osterloh, V. Vedral, Rev. Mod. Phys. 80, 517 (2008) . * (7) X. G. Wen, Phys. Rev. B 40, R7387 (1989); X.G. Wen and Q. Niu, Phys. Rev. B 41, 9377 (1990); X. G. Wen, Quantum field Theory of Many-body Systems (Oxford University Press, 2005). * (8) Angelo C. M. Carollo and J. K. Pachos, Phys. Rev. Lett. 95, 157203 (2005); J. K. Pachos and Angelo C. M. Carollo, Phil. Trans. R. Soc. A 364, 3463 (2006). * (9) S. L. Zhu, Phys. Rev. Lett. 96, 077206 (2006). * (10) H. T. Cui, K, Li, X. X. Yi, Phys. Lett. A 360, 243 (2006). * (11) M. E. Reuter, M. J. Hartmann, M.B. Plenio, Proc. Roy. Soc. Lond. A 463, 1271 (2007). * (12) P. Zanardi, P. Giorda, M. Cozzini, Phys. Rev. Lett. 99, 100603 (2007); L. C. Venuti and P. Zanardi, ibid. 99, 095701 (2007). * (13) N. Paunković and V. R. Vieira, Phys. Rev. E 77, 011129 (2008). * (14) S.-L. Zhu, Int. J. Mod. Phys. B 22, 561(2008). * (15) E. Lieb, T. Schultz, D. Mattis, Annals of Physics, 16, 407 (1961). * (16) R. Resta, Phys. Rev. Lett. 80, 1800 (1998); R. Resta and S. Sorella, ibid. 82, 370 (1999); R. Resta, J. Phys: Condens. Matter 12, R107 (2000). * (17) A. A. Aligia and G. Ortiz, Phys. Rev. Lett. 82, 2560 (1999); A. A. Aligia, Europhys. Lett. 45, 411 (1999). * (18) M. Nakamura and J. Voit, Phys. Rev. B 65, 153110 (2002); M. Nakamura and S. Todo, Phys. Rev. Lett. , (2002) * (19) S. Ryu and Y. Hatsugai, Phys. Rev. Lett. 89, 077002 (2002). * (20) S. Ryu and Y. Hatsugai, Phys. Rev. B 73 , 245115 (2006). * (21) It is convenient for a two-levle system to write the desity matrix as $\rho=1/2(1+\mathbf{R}\cdot\mathbf{\sigma})$, in which the vector $\mathbf{R}$ is real three-dimensional vector and $\|\mathbf{R}\|\leq 1$. This vectoer is well known as the Bloch vector. * (22) M.V. Berry, Proc. R. Soc. Lond. A 392, 45(1984). * (23) M. Yamanaka, M. Oshikawa, I. Affleck, Phys. Rev. Lett. 79, 1110 (1997); A. A. Aligia and E. R. Gaglinao, Physica C 304, 29 (1998); Euro. Phys. Journ. B 5, 371 (1998). * (24) W. Li, L. Ding, R. Yu, T. Roscide and S. Haas, Phys. Rev. B 74, 073103 (2006). * (25) Since Eq. (16) is spatially isotropic, it has no difference for $z_{\mu}(\mu=1\dots D)$ with $D>1$ and these plottings are chosen for $z_{1}$. * (26) The indetermination of $K_{\mu}$ when $\lambda=1,k_{\alpha}=0$ can be resolvable by L’Hospital rule. Our calculation shows that $K_{\mu}$ is 0 for this case. For $D=2,3$, the similar calculations are applied. * (27) A. Heeger, S. Kivelson, J. R. Schrieffer, W. P. Su, Rev. Mod. Phys. 60, 781 (1988). * (28) S. Filipp and E. Sjöqvist, Phys. Rev. Lett. 90, 050403 (2003); Phys. Rev. A 68, 042112 (2003). * (29) T. Hirano, H. Katsura, Y. Hatsugai, Phys. Rev. B 77, 094431 (2008)	arxiv-papers	2008-08-02T09:22:10	2024-09-04T02:48:57.133186	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "H.T.Cui and Jie Yi", "submitter": "Hai-tao Cui Dr.", "url": "https://arxiv.org/abs/0808.0242" }
0808.0250	Contraction in $L^{1}$ and large time behavior for a system arising in chemical reactions and molecular motors Michel Chipot 111Angewandte Mathematik, Universitaet Zurich, CH-8057 Zurich E-mail: m.m.chipot@math.unizh.ch, Danielle Hilhorst 222Laboratoire de Mathématiques, CNRS and Université de Paris-Sud, 91405 Orsay Cédex, France; E-mail: Danielle.Hilhorst@math.u-psud.fr, David Kinderlehrer 333Center for Nonlinear Analysis and Department of Mathematical Sciences, Carnegie Mellon University, Pittsburgh, PA 15213; E-mail: davidk@andrew.cmu.edu, Michał Olech 444Instytut Matematyczny Uniwersytetu Wrocławskiego, pl. Grunwaldzki 2/4, 50-384 Wrocław, Polska; Laboratoire de Mathématiques, CNRS Université de Paris-Sud, 91405 Orsay Cédex, France; E-mail: olech@math.uni.wroc.pl. The preparation of this article has been partially supported by a KBN/MNiI grant 1 P03A 008 30 and a Marie Curie Transfer of Knowledge Fellowship of the European Community’s Sixth Framework Programme under the contracts MTKD- CT-2004-013389, DMS 0305794, DMS 0405343, and DMS 0806703. We would also like to mention the support of the Swiss National Science Foundation under the contract # 20-113287/1. ###### Abstract We prove a contraction in $L^{1}$ property for the solutions of a nonlinear reaction–diffusion system whose special cases include intercellular transport as well as reversible chemical reactions. Assuming the existence of stationary solutions we show that the solutions stabilize as $t$ tends to infinity. Moreover, in the special case of linear reaction terms, we prove the existence and the uniqueness (up to a multiplicative constant) of the stationary solution. Key words: weakly coupled system, molecular motor, transport, parabolic systems, contraction property. AMS subject classification: 34D23, 35K45, 35K50, 35K55, 35K57, 92C37, 92C45. ## 1 Introduction We start with two specific reaction-diffusion systems. The first one describes a reversible reaction and the other one a molecular motor. We first consider the reversible chemical reaction (see also Bothe [4], Bothe and Hilhorst [5], Desvillettes and Fellner [10] and Érdi and Tóth [11]). It involves a reaction- diffusion system of the form $\begin{split}u_{t}=&\,d_{1}\Delta u-\alpha k\big{(}r_{A}(u)-r_{B}(v)\big{)}\quad\text{in}\quad\Omega\times(0,T),\ \ \Omega\subset\mathbb{R}^{d},\\\ v_{t}=&\,d_{2}\Delta v+\beta k\big{(}r_{A}(u)-r_{B}(v)\big{)}\quad\text{in}\quad\Omega\times\ (0,T),\ \ \Omega\subset\mathbb{R}^{d},\end{split}$ (1.1) together with the homogeneous Neumann boundary conditions, where $d_{1}$, $d_{2}$, $\alpha$, $\beta$, $k$ and $T$ are positive constants and where $\Omega$ is a bounded subset of $\mathbb{R}^{d}$ with smooth boundary. Such systems describe, with a suitable choice of the functions $r_{A}$ and $r_{B}$, chemical reactions for two mobile species. For example, functions $r_{A}(u)=u^{k},\ r_{B}(v)=v^{m}$ correspond to a reversible reaction $kA\rightleftharpoons mB$. Reactions of the type $q_{1}A_{1}+\ldots q_{k}A_{k}\rightleftharpoons q_{1}B_{1}+\ldots q_{m}B_{m}$ can also be described by similar systems with more complicated reactions terms. Another model problem is a system in $d=1$ space dimension and $n$ unknown variables $u_{1},\dots,u_{n}$, $n>1$, for intercellular transport, namely $\begin{split}\frac{\partial u_{i}}{\partial t}&=\frac{\partial}{\partial x}\left(\sigma\frac{\partial u_{i}}{\partial x}+u_{i}\psi_{i}^{\prime}\right)\\\ &\qquad\qquad+\sum_{j=1}^{n}a_{ij}u_{j}\quad\text{in}\quad Q_{T}=[0,1]\times(0,T)\\\ &\sigma\frac{\partial u_{i}}{\partial x}+u_{i}\psi_{i}^{\prime}=0\quad\text{on}\quad\partial Q_{T}=\\{0,1\\}\times(0,T),\end{split}$ where $\begin{split}&a_{ii}\leq 0,\ a_{ij}\geq 0\mbox{~{}~{}for~{}~{}all~{}~{}}\ \ i\in\\{1,\ldots,n\\},i\neq j,\\\ &\sum_{i=1}^{n}a_{ij}=0\mbox{~{}~{}for~{}~{}all~{}~{}}i,j\in\\{1,\ldots,n\\}.\end{split}$ (1.2) It models transport via motor proteins in the eukaryotic cell where chemical energy is transduced into directed motion. A derivation of the system from a mass transport viewpoint is given in [7]. For an analysis of the steady state solutions and for further references we refer to [6], [12], [13], and [20]. In this paper we study the corresponding system in higher space dimension, namely $\begin{split}\frac{\partial u_{i}}{\partial t}=\text{div}\big{(}\sigma_{i}\nabla u_{i}&+u_{i}\nabla\psi_{i}\big{)}\\\ &+\alpha_{i}\bigg{(}\sum_{j=1}^{n}\lambda_{ij}r_{j}\big{(}u_{j}(x,t),x\big{)}\bigg{)}\quad\mbox{~{}~{}in~{}~{}}Q_{T},\end{split}$ (1.3a) where $i\in\\{1,\ldots,n\\}$, and $u_{i}(x,t):Q_{T}\to\mathbb{R}^{+}$, with $Q_{T}=\Omega\times(0,T)$, $\Omega$ an open bounded subset of $\mathbb{R}^{d}$ with smooth boundary, and $T$ some positive constant. We supplement this system with the Robin (no-flux) boundary conditions $\sigma_{i}\frac{\partial u_{i}}{\partial\nu}+u_{i}\frac{\partial\psi_{i}}{\partial\nu}=0,\quad i\in\\{1,\ldots,n\\},\quad\text{on}\quad\partial\Omega\times(0,T),$ (1.3b) where $\nu$ is the outward normal vector to $\partial\Omega$, and the initial conditions $u_{1}(x,0)=u_{0,1}(x),\ \ldots\ ,u_{n}(x,0)=u_{0,n}(x),\quad x\in\Omega.$ (1.3c) We assume that the following hypotheses hold 1. 1. The constants $\sigma_{i}$ and $\alpha_{i}\in\mathbb{R}$, where $i\in\\{1,\ldots,n\\}$, are strictly positive; 2. 2. For $i,j\in\\{1,\ldots,n\\},\ \lambda_{ii}\leq 0,\ \lambda_{ij}\geq 0$ if $i\neq j$, $\sum_{k=1}^{n}\lambda_{kj}=0$; 3. 3. for all $i\in\\{1,\ldots,n\\}$, the smooth functions $r_{i}$ are nondecreasing with respect to the first variable; $r_{i}(0,x)=0$ and we assume that the functions $\psi_{i}$ are smooth as well; 4. 4. $u_{i}(.,0)=u_{0i}\in C(\overline{\Omega}),\ u_{0i}\geqslant 0$. In the linear case of the molecular motors, it amounts to choosing $r_{i}(s,x)=s,\ \lambda_{ij}=a_{ij}\ \text{and}\ \alpha_{i}=1\ \text{for all}\ i,j\in\\{1,\ldots,n\\}.$ (1.4) We denote by Problem (P) the system (1.3a) together with the boundary and initial conditions (1.3b), (1.3c), and admit without proof that Problem (P) possesses a unique smooth and bounded solution on each time interval $(0,T]$. An essential idea for proving the existence of a solution would be to apply the Comparison principle Theorem 2.2 below to deduce that any solution of Problem (P) has to be nonnegative and bounded from above by a stationary solution. Finally, we note that because of the boundary conditions (1.3b) the quantity $\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\int_{\Omega}u_{i}(x,t)\,\text{d}x$ (1.5) is conserved in time. The organization of this paper is as follows. In Section 2 we prove a comparison principle for Problem (P). The main idea, which permits to show that Problem (P) is cooperative, is a change of functions which transforms the Robin boundary conditions into the homogeneous Neumann boundary conditions. In Section 3 we establish a contraction in $L^{1}$ property for the corresponding semigroup solution. Let us point out the similarity with an old result due to Crandall and Tartar [8] where they proved in a scalar case that in the presence of a conservation of the integral property such as (1.5), a comparison principle such as Theorem 2.2 is equivalent to a contraction in $L^{1}$ property such as the inequality (3.4) below. As far as we know such an abstract result is not known in the case of systems. Section 4 deals with the large time behavior of the solutions. Supposing the existence of a stationary solution, we construct a continuum of stationary solutions and prove that the solutions stabilize as $t$ tends to infinity. Let us mention a result by Perthame [19] who proved the stabilization in the case of the two component one-dimensional molecular motor problem. Finally in Section 5, show the existence and uniqueness (up to a multiplicative constant) of the stationary solution of the molecular motor problem. Acknowledgment: The authors acknowledge the preliminary master thesis work of Aude Brisset about the corresponding two component system. They are grateful to the professors Piotr Biler, Stuart Hastings, Annick Lesne and Hiroshi Matano for very fruitful discussions. ## 2 Comparison principle First, we remark that the system of equations (1.3a) is cooperative. However, since nothing is known about the sign of the coefficients $\displaystyle{\frac{\partial\psi_{i}}{\partial\nu}}$ in the Robin boundary conditions (1.3b), we cannot decide whether the Problem (P) is cooperative. This leads us to perform a change of variables which transforms the Robin boundary conditions into the homogeneous Neumann boundary conditions. ### 2.1 The change of unknown functions Performing the change of variables $w_{i}(x,t)=u_{i}(x,t){\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}},\quad i\in\\{1,\ldots,n\\},$ (2.1) we deduce from (1.3) that ${\vec{w}}:=(w_{1},\ldots,w_{n})$ satisfies the parabolic problem $\begin{split}&\frac{\partial w_{i}}{\partial t}=\sigma_{i}{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\text{div}\big{(}{\,\text{e}}^{\nicefrac{{-\psi_{i}(x)}}{{\sigma_{i}}}}\nabla w_{i}\big{)}\\\ &+\alpha_{i}{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\bigg{(}\sum_{j=1}^{n}\lambda_{ij}r_{j}\big{(}w_{j}(x,t){\,\text{e}}^{\nicefrac{{-\psi_{j}(x)}}{{\sigma_{j}}}},x\big{)}\bigg{)}\quad\mbox{~{}~{}in~{}~{}}Q_{T},\end{split}$ (2.2) together with the homogeneous Neumann boundary conditions $\frac{\partial w_{i}}{\partial\nu}=0,\quad i\in\\{1,\ldots,n\\},\quad\text{on}\quad\partial\Omega,$ (2.3) and the initial conditions $w_{i}(x,0)=u_{0,i}(x){\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}},\quad i\in\\{1,\ldots,n\\},\quad x\in\Omega.$ (2.4) In the following, we denote by Problem $P_{N}$ — the problem (2.2), (2.3), (2.4). To begin with we define the operators $\begin{split}&{\cal L}_{i}(w_{i})=\frac{\partial w_{i}}{\partial t}-\sigma_{i}{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\text{div}\big{(}{\,\text{e}}^{\nicefrac{{-\psi_{i}(x)}}{{\sigma_{i}}}}\nabla w_{i}\big{)}\\\ &-\alpha_{i}{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\bigg{(}\sum_{j=1}^{n}\lambda_{ij}r_{j}\big{(}w_{j}(x,t){\,\text{e}}^{\nicefrac{{-\psi_{j}(x)}}{{\sigma_{j}}}},x\big{)}\bigg{)}\quad\mbox{~{}~{}in~{}~{}}Q_{T}.\end{split}$ (2.5) We say that $\left(\underline{w}_{1},\ldots,\underline{w}_{n}\right)$ is a subsolution of Problem $P_{N}$ if $\begin{split}{\cal L}_{i}(\underline{w}_{i})\leqslant&\;0\quad\text{in}\quad Q_{T},\\\ \frac{\partial\underline{w}_{i}}{\partial\nu}\leqslant&\;0\quad\text{on}\quad\partial\Omega\times(0,T),\\\ \underline{w}_{i}(x,0)\leqslant&\;w_{i}(x,0),\quad x\in\Omega\end{split}$ (2.6) for all $i\in\\{1,\ldots,n\\}$. We define similarly a supersolution $\left(\overline{u}_{1},\ldots,\overline{u}_{n}\right)$ of Problem $P_{N}$ by the inequalities $\begin{split}{\cal L}_{i}(\overline{w}_{i})\geqslant&\;0\quad\text{in}\quad Q_{T},\\\ \frac{\partial\overline{w}_{i}}{\partial\nu}\geqslant&\;0\quad\text{on}\quad\partial\Omega\times(0,T),\\\ \overline{w}_{i}(x,0)\geqslant&\;w_{i}(x,0),\quad x\in\Omega.\end{split}$ (2.7) The following comparison theorem holds ([2], [21]). ###### Theorem 2.1. Let $(\underline{w}_{1},\ldots,\underline{w}_{n})$ and $(\overline{w}_{1},\ldots,\overline{w}_{n})$, be a sub - and a super - solution, respectively, for the operators ${\cal L}_{j}$ defined by (2.5) with $j\in\\{1,\ldots,n\\}$, which means that (2.6) and (2.7) hold for $i\in\\{1,\ldots,n\\}$. Then $\underline{w}_{i}\leqslant\overline{w}_{i}$ in $Q_{T}$. Moreover, for all $i\in\\{1,\ldots,n\\}$ such that $\underline{w}_{i}\leqslant\overline{w}_{i}$ and $\underline{w}_{i}\not\equiv\overline{w}_{i}$ on $\\{t=0\\}\times\Omega$ then $\underline{w}_{i}<\overline{w}_{i}$ in $Q_{T}$. $\blacksquare$ This comparison theorem immediately translates into a comparison theorem for solutions of the original Problem (P). For all $i\in\\{1,\ldots,n\\}$, we define the operators $\begin{split}L_{i}(u_{i})=(u_{i})_{t}-\text{div}\big{(}&\sigma_{i}\nabla u_{i}+u_{i}\nabla\psi_{i}\big{)}\\\ &-\alpha_{i}\bigg{(}\sum_{j=1}^{n}\lambda_{ij}\,r_{j}\,(u_{j},x)\bigg{)}\quad\text{in}\quad Q_{T}.\end{split}$ (2.8) The following result holds. ###### Theorem 2.2. Let $(\underline{u}_{1},\ldots,\underline{u}_{n})$ and $(\overline{u}_{1},\ldots,\overline{u}_{n})$, be a sub - and a super - solution, respectively, for the operators $L_{j}$, defined by (2.8) with $j\in\\{1,\ldots,n\\}$. Then $\underline{u}_{i}\leqslant\overline{u}_{i}$ in $Q_{T}$. Moreover, for all $i\in\\{1,\ldots,n\\}$ such that $\underline{u}_{i}\leqslant\overline{u}_{i}$ and $\underline{u}_{i}\not\equiv\overline{u}_{i}$ on $\\{t=0\\}\times\Omega$ then $\underline{u}_{i}<\overline{u}_{i}$ in $Q_{T}$. $\blacksquare$ Next we state two immediate corollaries of Theorem 2.2. ###### Corollary 2.3. (uniqueness) If $(u_{1}^{1},\ldots,u_{n}^{1})$ and $(u_{1}^{2},\ldots,u_{n}^{2})$ are solutions of Problem (P) with the same initial condition $(u_{0,1},\ldots,u_{0,n})\in\big{(}C(\overline{\Omega}))^{n}$, then for all $i\in\\{1,\ldots,n\\},\ u_{i}^{1}=u_{i}^{2}$ . $\blacksquare$ ###### Corollary 2.4. (positivity) If $(u_{1},\ldots,u_{n})$ is the solution of Problem (P) with the nonnegative initial condition $(u_{0,1},\ldots,u_{0,n})\in\big{(}C(\overline{\Omega})\big{)}^{n}$, then for all $i\in\\{1,\ldots,n\\},$ $u_{i}\geqslant 0$. Moreover, for all $i\in\\{1,\ldots,n\\}$, such that $u_{0,i}\geqslant 0$ and $u_{0,i}\not\equiv 0$, $u_{i}>0$ in $\Omega$. $\blacksquare$ ## 3 Contraction property The purpose of this section is to show a contraction in $\big{(}L^{1}(\Omega)\big{)}^{n}$ property for the solutions of Problem (P) with the initial conditions belonging to $\big{(}L^{\infty}(\Omega)\big{)}^{n}$. The main steps of the proof rely upon arguments due to [3] and [18]. We first introduce some notation. We suppose that the functions $(u_{1}^{1},\ldots,u_{n}^{1})$ and $(u_{1}^{2},\ldots,u_{n}^{2})$ are the solutions of Problem (P) with the initial conditions $(u_{0,1}^{1},\ldots,u_{0,n}^{1})$ and $(u_{0,1}^{2},\ldots,u_{0,n}^{2})$, respectively. Define $(U_{1},\ldots,U_{n}):=(u_{1}^{1}-u_{1}^{2},\ldots,u_{n}^{1}-u_{n}^{2}).$ (3.1) Then $\begin{split}&(U_{i})_{t}=\text{div}\big{(}\sigma_{i}\nabla U_{i}+U_{i}\nabla\psi_{i}\big{)}\\\ &+\ \alpha_{i}\sum_{j=1}^{n}\lambda_{ij}\big{(}r_{j}(u_{j}^{1}(x,t),x)-r_{j}(u_{j}^{2}(x,t),x)\big{)}\quad\text{in}\quad Q_{T},\\\ &\sigma_{i}\frac{\partial U_{i}}{\partial\nu}+U_{i}\frac{\partial\psi_{i}}{\partial\nu}=0\quad\text{on}\quad\partial\Omega\times(0,T),\\\ &U_{i}(x,0)=U_{0,i}(x)\quad\text{for}\quad x\in\Omega,\end{split}$ (3.2) together with $U_{0,i}=u_{0,i}^{1}-u_{0,i}^{2},$ (3.3) for each $i\in\\{1,\ldots,n\\}$. Next we prove the following contraction in $L^{1}$ property. ###### Theorem 3.1. For all $t>0$, $\frac{1}{\alpha_{1}}\\|U_{1}(\cdot,t)\\|_{L^{1}(\Omega)}+\ldots+\frac{1}{\alpha_{n}}\\|U_{n}(\cdot,t)\\|_{L^{1}(\Omega)}\\\ \leqslant\frac{1}{\alpha_{1}}\\|U_{0,1}(\cdot)\\|_{L^{1}(\Omega)}+\ldots+\frac{1}{\alpha_{n}}\\|U_{0,n}(\cdot)\\|_{L^{1}(\Omega)},$ (3.4) where $U_{i}$ and $U_{0,i},\ i\in\\{1,\ldots,n\\}$, are defined by (3.1) and (3.3), respectively. Proof Dividing each partial differential equation of (3.2) by $\alpha_{i}$ and summing them up, we obtain $\begin{split}\frac{\text{d}}{\text{d}t}\bigg{(}\sum_{i=1}^{n}\frac{1}{\alpha_{i}}&U_{i}\bigg{)}=\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\,\text{div}\left(\sigma_{i}\nabla U_{i}+U_{i}\nabla\psi_{i}\right)\\\ &+\sum_{i=1}^{n}\sum_{j=1}^{n}\lambda_{ij}\Big{(}r_{j}(u_{j}^{1}(x,t),x)-r_{j}(u_{j}^{2}(x,t),x)\Big{)}\\\ =&\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\,\text{div}\left(\sigma_{i}\nabla U_{i}+U_{i}\nabla\psi_{i}\right)\\\ &+\sum_{j=1}^{n}\bigg{\\{}\Big{(}r_{j}(u_{j}^{1}(x,t),x)-r_{j}(u_{j}^{2}(x,t),x)\Big{)}\sum_{i=1}^{n}\lambda_{ij}\bigg{\\}}\\\ =&\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\,\text{div}\left(\sigma_{i}\nabla U_{i}+U_{i}\nabla\psi_{i}\right),\end{split}$ where we have used Hypothesis 2. This, together with the boundary conditions (1.3b), implies the conservation in time of the quantity $\frac{\text{d}}{\text{d}t}\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\int_{\Omega}U_{i}(x,t)\,\text{d}x=0.$ (3.5) Let us look closer at the nonlinear term in (3.2). We can write, for fixed index $i$ $\sum_{j=1}^{n}\lambda_{ij}\big{(}r_{j}(u_{j}^{1}(x,t),x)-r_{j}(u_{j}^{2}(x,t),x)\big{)}\\\ =\sum_{j=1}^{n}\lambda_{ij}U_{j}\int_{0}^{1}\frac{\partial}{\partial u}r_{j}(\theta u_{j}^{1}+(1-\theta)u_{j}^{2},x)\text{d}\theta=\sum_{j=1}^{n}A_{ij}U_{j}.$ Freezing the functions $u_{i}^{k}$ for $i\in\\{1,\ldots,n\\},\ k\in\\{1,2\\}$, we deduce that the functions $U_{1},\ldots,U_{n}$ satisfy a system of the form $(U_{i})_{t}=\text{div}\Big{(}\sigma_{i}\nabla U_{i}+U_{i}\nabla\psi_{i}\Big{)}+\sum_{j=1}^{n}A_{ij}U_{j}\quad\text{in}\quad Q_{T},$ (3.6) with the boundary and initial conditions $\begin{split}&\sigma_{i}\frac{\partial U_{i}}{\partial\nu}+U_{i}\frac{\partial\psi_{i}}{\partial\nu}=0\quad\text{on}\quad\partial\Omega\times(0,T),\\\ &U_{i}(x,0)=U_{0,i}(x),\quad x\in\Omega.\end{split}$ (3.7) for $i\in\\{1,\ldots,n\\}$, where $A_{ij}$ are functions of space and time. In order to make the notation more concise, we write $\begin{split}\vec{U}_{0}=&\,\big{(}U_{0,1},\ldots,U_{0,n}\big{)},\\\ \vec{U}=&\,\big{(}U_{1},\ldots,U_{n}\big{)},\\\ \vec{U}_{0}^{\pm}=&\,\big{(}U_{0,1}^{\pm},\ldots,U_{0,n}^{\pm}\big{)},\\\ \vec{U}^{\pm}=&\,\big{(}U_{1}^{\pm},\ldots,U_{n}^{\pm}\big{)},\\\ \end{split}$ where $s^{+}=\max\\{s,0\\},\ s^{-}=\max\\{-s,0\\}$. By (3.6), (3.7) and Corollary 2.3 we can write $\vec{U}$ in the form $(\vec{U})(x,t)=\mathcal{S}(t)\,\vec{U}_{0}(x)=\big{(}{\mathcal{S}}_{1}(t)\vec{U}_{0},\ldots,{\mathcal{S}}_{n}(t)\vec{U}_{0}\big{)}(x)$ with some operator ${\mathcal{S}}(t)$. We set $\big{(}W_{1},\ldots,W_{n}\big{)}=-\big{(}U_{1}{\,\text{e}}^{\nicefrac{{\psi_{1}(x)}}{{\sigma_{1}}}},\ldots,U_{n}{\,\text{e}}^{\nicefrac{{\psi_{n}(x)}}{{\sigma_{n}}}}\big{)},$ and $\widetilde{A}_{ij}=A_{ij}\,{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\,{\,\text{e}}^{-\nicefrac{{\psi_{j}(x)}}{{\sigma_{j}}}}$. Then, the system of equations (3.6) can be expressed in the form $\big{(}W_{i}\big{)}_{t}={\sigma_{i}}{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\text{div}\Big{(}{\,\text{e}}^{-\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}}\nabla W_{i}\Big{)}+\sum_{j=1}^{n}\widetilde{A}_{ij}W_{j}\quad\text{in}\quad Q_{T},$ (3.8) with the boundary and initial conditions $\displaystyle\frac{\partial W_{i}}{\partial\nu}$ $\displaystyle=0\quad\text{on}\quad\partial\Omega\times(0,T),$ (3.9) $\displaystyle W_{i}(x,0)$ $\displaystyle=-U_{0,i}{\,\text{e}}^{\nicefrac{{\psi_{i}(x)}}{{\sigma_{i}}}},\quad x\in\Omega,$ (3.10) for $i\in\\{1,\ldots,n\\}$. Next we show that the solutions $W_{i}$ of the problem (3.8) – (3.10) with nonpositive initial conditions are nonpositive in $\overline{\Omega}$ for all $t\in(0,T)$. To that purpose we consider the auxiliary problem $\displaystyle\big{(}W_{i}\big{)}_{t}-\vartheta_{i}(x)\text{div}\Big{(}\zeta_{i}(x)\nabla W_{i}\Big{)}-\sum_{j=1}^{n}\gamma_{ij}W_{j}\leqslant 0\quad\text{in}\quad Q_{T},$ (3.11) $\displaystyle\frac{\partial W_{i}}{\partial\nu}\leqslant 0\quad\text{on}\quad\partial\Omega\times(0,T),$ (3.12) $\displaystyle W_{i}(x,0)=W_{0,i}(x)\leqslant 0\quad x\in\Omega,$ (3.13) for $i\in\\{1,\ldots,n\\}$. We assume that $\vartheta_{i}(x)$ and $\zeta_{i}(x)$ are nonnegative in $\overline{\Omega}$ and that the coefficients $\gamma_{ij}$ satisfy the same assumptions as the coefficients $\lambda_{ij}$ in Problem (P). The following result holds. ###### Lemma 3.2. Let $(W_{1},\ldots,W_{n})$ be a smooth and bounded solution of the problem (3.11) – (3.13) with nonpositive initial conditions $W_{0,i}$ on a time interval $[0,T]$. Then $W_{i}(x,t)\leqslant 0$ in $\overline{\Omega}\times(0,T]$. Moreover, for each $i\in\\{1,\ldots,n\\}$ such that $W_{0,i}\leqslant 0$ and $W_{0,i}\not\equiv 0,\ W_{i}<0$ in $\overline{\Omega}\times(0,T]$. Proof The result of Lemma 3.2 follows from the fact that the system (3.11), (3.12), (3.13), with the inequalities $\\{\leqslant\\}$ replaced by the equalities $\\{=\\}$, is a cooperative system. However, for the sake of completeness, we present a proof below. We first remark that, in view of [21, Remark (i), p. 191], one can always satisfy the condition $\sum_{j=1}^{n}\gamma_{ij}\leqslant 0\text{\ for all\ }i\in\\{1,\ldots,n\\},$ (3.14) for the matrix of coefficients $\big{(}\gamma_{ij}\big{)}_{i,j=1}^{n}$ by performing the change of variables $\overline{W}_{i}=W_{i}{\,\text{e}}^{-ct}$ for all $i\in\\{1,\ldots,n\\}$ and $c>0$ large enough. Thanks to the regularity of each $W_{i}$, we can apply Theorem 15, p. 191 from [21] to conclude that $W_{i}-M\leqslant 0$ in $\overline{\Omega}\times[0,T]$ for some $M\>0$ and all $i\in\\{1,\ldots,n\\}$. In fact, we can deduce that $W_{i}-M<0$ in $\overline{\Omega}\times(0,T)$. Indeed, if for some $k\in\\{1,\ldots,n\\}$, $W_{k}=M$ in an interior point $(\tilde{x},\tilde{t})\in\Omega\times(0,T)$, then Theorem 15, p. 191 in [21] implies that $W_{k}\equiv M$ for all $0\leqslant t<\tilde{t}$, which is impossible since $W_{k}(x,0)\leqslant 0$. If the maximum $M$ of $W_{k}$ is attained at a boundary point $P\in\partial\Omega\times(0,T)$ then either there exists an open ball $K\subset\Omega\times(0,T)$ such that $P\in\partial K$ and $W_{k}-M<0$ in $K$, and the last part of Theorem 15, p. 191 in [21] contradicts the boundary inequality (3.12), or for all open balls $K\subset\Omega\times(0,T)$ such that $P\in\partial K$ there exists a point $(\tilde{x},\tilde{t})\in K$ such that $W_{i}(\tilde{x},\tilde{t})=M$, and we proceed as in the case before. Hence, there exists $\widetilde{M}>0$, such that $W_{i}\leqslant\widetilde{M}<M$ in $\overline{\Omega}\times[0,T]$ for all $i\in\\{1,\ldots,n\\}$. Then we can repeat the reasoning for all $M>0$ until $M=0$. Indeed, if this would not be the case, we find the least real number $\overline{M}>0$, with $W_{i}\leqslant\overline{M}\leqslant\widetilde{M}$ in $\overline{\Omega}\times[0,T]$, which leads again to the existence of a real number $0\leqslant\widehat{M}<\overline{M}$ with the same property. This contradicts the fact that $\overline{M}$ was defined as the least such real number. $\blacksquare$ Since the functions $u_{i}^{1},\ u_{i}^{2}$ are bounded on $\overline{\Omega}\times[0,T]$, it follows that the functions $W_{i}$ are bounded on $\overline{\Omega}\times[0,T]$ for all $i\in\\{1,\ldots,n\\}$. Then we are in a position to apply Lemma 3.2 with $\vartheta_{i}(x)={\,\text{e}}^{\nicefrac{{\psi_{i}}}{{\sigma_{i}}}}$, $\zeta_{i}(x)=\sigma_{i}{\,\text{e}}^{-\nicefrac{{\psi_{i}}}{{\sigma_{i}}}}$ and $\gamma_{ij}=\widetilde{A}_{ij}$ for $i,j\in\\{1,\ldots,n\\}$. We deduce that the solutions $W_{i}$ of the problem (3.8) – (3.10) with nonpositive initial conditions are nonpositive in $\overline{\Omega}$ for all $t\in(0,T)$. Next we remark that the above reasoning can be applied either with $\vec{U}_{0}$ replaced by $U^{+}_{0}$ or with $\vec{U}_{0}$ replaced by $U^{-}_{0}$. This permits to show that ${\mathcal{S}}_{i}(t)\vec{U}_{0}^{+},{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\geqslant 0$ and that ${\mathcal{S}}_{i}(t)\vec{U}_{0}^{\pm}>0\quad\text{if}\quad\vec{U}_{0}^{\pm}\not\equiv 0.$ (3.15) We easily compute $\begin{split}\sum_{i=1}^{n}\frac{1}{\alpha_{i}}&\big{\\|}U_{i}(\cdot,t)\big{\\|}_{L^{1}(\Omega)}-\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\big{\\|}U_{0,i}(\cdot)\big{\\|}_{L^{1}(\Omega)}\\\ =&\ \sum_{i=1}^{n}\frac{1}{\alpha_{i}}\big{\\|}{\mathcal{S}}_{i}(t)\vec{U}_{0}^{+}-{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\big{\\|}_{L^{1}(\Omega)}-\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\big{\\|}U_{0,i}(\cdot)\big{\\|}_{L^{1}(\Omega)}\\\ =&\ \sum_{i=1}^{n}\int_{\Omega}\frac{1}{\alpha_{i}}\Big{\\{}\max\big{\\{}{\mathcal{S}}_{i}(t)\vec{U}_{0}^{+},{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\big{\\}}\end{split}$ (3.16) $\begin{split}&-\frac{1}{\alpha_{i}}\min\big{\\{}{\mathcal{S}}_{i}(t)\vec{U}_{0}^{+},{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\big{\\}}\Big{\\}}\,\text{d}x-\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\int_{\Omega}\big{\\{}U_{i,0}^{+}+U_{i,0}^{-}\big{\\}}\,\text{d}x\\\ =&\ \sum_{i=1}^{n}\int_{\Omega}\frac{1}{\alpha_{i}}\big{(}{\mathcal{S}}_{i}(t)\vec{U}_{0}^{+}+{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\big{)}\,\text{d}x-\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\int_{\Omega}\big{\\{}U_{i,0}^{+}+U_{i,0}^{-}\big{\\}}\,\text{d}x\\\ &-2\sum_{i=1}^{n}\int_{\Omega}\frac{1}{\alpha_{i}}\min\big{\\{}{\mathcal{S}}_{i}(t)\vec{U}_{0}^{+},{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\big{\\}}\,\text{d}x\\\ =&\ -2\sum_{i=1}^{n}\int_{\Omega}\frac{1}{\alpha_{i}}\min\big{\\{}{\mathcal{S}}_{i}(t)\vec{U}_{0}^{+},{\mathcal{S}}_{i}(t)\vec{U}_{0}^{-}\big{\\}}\,\text{d}x\leqslant 0,\end{split}$ (3.17) which completes the proof of (3.4). $\blacksquare$ ###### Corollary 3.3. Let $(u_{0,1}^{1},\ldots,u_{0,n}^{1}),\ (u_{0,1}^{2},\ldots,u_{0,n}^{2})\in\big{(}C(\overline{\Omega})\big{)}^{n}$ be as in Theorem 3.1. Moreover, let us assume that for at least one index $k\in\\{1,\ldots,n\\}$ the difference $u^{1}_{0,k}-u^{2}_{0,k}$ changes the sign. Then, the inequality (3.4) is strict for all $t>0$, so that solution satisfies a strict contraction property. ## 4 Large time behavior of solutions In this section we assume the existence and uniqueness of a positive solution $\vec{v}=(v_{1},\ldots,v_{n})\in\big{(}C(\overline{\Omega})\cap C^{2}(\Omega)\big{)}^{n}$ of the elliptic problem $\displaystyle\text{div}\big{(}\sigma_{i}\nabla v_{i}+v_{i}\nabla\psi_{i}\big{)}+\alpha_{i}\bigg{(}\sum_{j=1}^{n}\lambda_{ij}r_{j}\big{(}v_{j}(x),x\big{)}\bigg{)}$ $\displaystyle=0\quad\text{in}\quad\Omega,$ (4.1) $\displaystyle\sigma_{i}\frac{\partial v_{i}}{\partial\nu}+v_{i}\frac{\partial\psi_{i}}{\partial\nu}$ $\displaystyle=0\quad\text{on}\quad\partial\Omega,$ (4.2) $\displaystyle\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\int_{\Omega}v_{i}(x)\,\text{d}x$ $\displaystyle=1,$ (4.3) for $i\in\\{1,\ldots,n\\}$. ###### Definition 4.1. We say that a vector function $\vec{v}=(v_{1},\ldots,v_{n})\in\big{(}C(\overline{\Omega})\big{)}^{n}$ is nonnegative (resp. positive) if $v_{i}(x)\geqslant 0$ (resp. $v_{i}(x)>0$) for all $x\in\overline{\Omega}$ and all $i\in\\{1,\ldots,n\\}$. Next we introduce the semigroup notation for the unique solution of Problem (P), namely $\vec{u}(t)=\mathcal{T}(t)\,\vec{u}_{0}=\Big{(}\mathcal{T}_{1}(t)\,\vec{u}_{0},\ldots,\mathcal{T}_{n}(t)\,\vec{u}_{0}\Big{)},$ with the initial data $\vec{u}_{0}\in\big{(}C(\overline{\Omega})\big{)}^{n}$. The method of the proof is based upon an idea of Osher and Ralston [18]. It mainly exploits the contraction properties for the nonlinear semigroup $\mathcal{T}(t)\,$ given by Theorem 3.1 and Corollary 3.3. A similar reasoning was developed in other contexts by Bertsch and Hilhorst [3], Hilhorst and Hulshof [14] and Hilhorst and Peletier [15]. We suppose there exists a set $\mathscr{H}\subset\big{(}C(\overline{\Omega})\cap C^{2}(\Omega)\big{)}^{n}$ of positive stationary solutions with the following property which we denote by $\mathscr{S}$: For each $\vec{f}=(f_{1},\ldots,f_{n})\in\big{(}C(\overline{\Omega})\cap C^{2}(\Omega)\big{)}^{n}$ either $\vec{f}\in\mathscr{H}$ or there exists $(\xi_{1},\ldots,\xi_{n})\in\mathscr{H}$, such that $f_{i}-\xi_{i}$ changes the sign for at least one index $i\in\\{1,\ldots,n\\}$. One can prove that a set $\mathscr{H}$ satisfying Property $\mathscr{S}$ exists in at least two cases: * i) In the case of the system (1.1) where the Robin boundary conditions reduce to the homogeneous Neumann boundary conditions, the set $\mathscr{H}$ is given by $\rule{20.0pt}{0.0pt}\mathscr{H}=\Big{\\{}(a,b):\ a>0,\ b=r_{B}^{-1}(r_{A}(a))\\\ \text{and}\ \frac{a}{\alpha}+\frac{b}{\beta}=\int_{\Omega}\Big{(}\frac{u}{\alpha}+\frac{v}{\beta}\Big{)}\,\text{d}x\Big{\\}}.\rule{20.0pt}{0.0pt}$ For more details we refer to [5]. * ii) In the case of the molecular motor with a linear $n$-component system the set $\mathscr{H}$ is given by $\mathscr{H}=\big{\\{}c\vec{v}:\ c\in\mathbb{R}^{+}\big{\\}},$ where $\vec{v}$ is a unique solution of the elliptic problem (4.1) – (4.3). ###### Proposition 4.2. The continuum $\mathscr{H}$ is such that for each $\vec{f}=(f_{1},\ldots,f_{n})\in\big{(}C(\overline{\Omega})\cap C^{2}(\Omega)\big{)}^{n}$ either $\vec{f}\in\mathscr{H}$, or there exists $(\xi_{1},\ldots,\xi_{n})\in\mathscr{H}$ such that $f_{i}-\xi_{i}$ changes the sign for at least one index $i\in\\{1,\ldots,n\\}$. Proof * i) In the case of system (1.1) the proof is rather obvious since the continuum $\mathscr{H}$ is composed of constant pairs. * ii) In the case of the molecular motor, let us assume that $\vec{f}\not\in\mathscr{H}$. Then there does not exist any positive constant $c$ such that $c\,\vec{v}=\vec{f}$. In particular, there exists an index $i\in\\{1,\ldots,n\\}$ such that $v_{i}$ is not proportional to $f_{i}$, or in other words $cv_{i}\neq f_{i}$ for all $c>0$. Without loss of generality we can assume that the first coordinate has this property. Let $x_{0}\in\Omega$ be arbitrary. Since $v_{1}$ is strictly positive in $\overline{\Omega}$, we can define $c_{0}=\frac{f_{1}(x_{0})}{v_{1}(x_{0})},$ so that $\big{(}f_{1}-c_{0}\,v_{1}\big{)}(x_{0})=0.$ Let $\mathcal{Z}=\big{\\{}x\in\overline{\Omega}:\ \big{(}f_{1}-c_{0}\,v_{1}\big{)}(x)=0\big{\\}}$. From the continuity of $f_{1}$ and $v_{1}$, $\mathcal{Z}$ is closed as a subset of $\Omega$. If there exist $x_{1},\,x_{2}\in\mathcal{Z}^{c}$, such that $\big{(}f_{1}-c_{0}\,v_{1}\big{)}(x_{1})$ and $\big{(}f_{1}-c_{0}\,v_{1}\big{)}(x_{2})$ are of different signs, then the proof is complete. Now suppose that $\big{(}f_{1}-c_{0}\,v_{1}\big{)}(x)$ is positive for all $x\in\mathcal{Z}^{c}$. In particular $\big{(}f_{1}-c_{0}\,v_{1}\big{)}(\tilde{x})=d>0$ for some fixed $\tilde{x}\in\mathcal{Z}^{c}$. Then choosing $\displaystyle\varepsilon=\frac{d}{2v_{1}(\tilde{x})}$ we see that $\big{(}f_{1}-(c_{0}+\varepsilon)v_{1}\big{)}(\tilde{x})=\frac{d}{2}>0.$ However $\big{(}f_{1}-(c_{0}+\varepsilon)v_{1}\big{)}(x_{0})<0.$ We proceed similarly when $\big{(}f_{1}-c_{0}\,v_{1}\big{)}(x)$ is negative for all $x\in\mathcal{Z}^{c}$. $\blacksquare$ In the sequel we suppose that the initial data $\vec{u}_{0}=(u_{0,1},\ldots,u_{0,n})$ from $\big{(}C(\overline{\Omega})\big{)}^{n}$ also satisfy the following property: $\text{{There exists\ }}\vec{h}\in\mathscr{H}\text{{\ such that\ }}0\leqslant\vec{u}_{0}\leqslant\vec{h}\quad\text{in}\quad\overline{\Omega},$ (4.4) and remark that this property is satisfied in both the cases (i) and (ii). ###### Proposition 4.3. Let $\vec{u}_{0}=(u_{0,1},\ldots,u_{0,n})\in\big{(}C(\overline{\Omega})\big{)}^{n}$ satisfy the property (4.4). Then the solution $(u_{1},\ldots,u_{n})$ of Problem (P) is such that $0\leqslant\vec{u}(t)\leqslant\vec{h}$ for all $t>0$. Proof We remark that $\vec{0}$ is a subsolution of Problem (P) and that $\vec{h}$ is a supersolution, and apply Theorem 2.2. $\blacksquare$ Next we prove the main result of this section. To that purpose we first define the norm $\boldsymbol{\big{\\|}}\cdot\boldsymbol{\big{\\|}_{1}}$ by $\boldsymbol{\big{\\|}}\vec{f}\ \boldsymbol{\big{\\|}_{1}}=\sum_{i=1}^{n}\frac{1}{\alpha_{i}}\big{\\|}f_{i}\big{\\|}_{L^{1}(\Omega)}\hskip 5.0pt.$ Note that this norm is equivalent to the usual product norm in the space $\big{(}L^{1}(\Omega)\big{)}^{n}$. ###### Theorem 4.4. For all nonnegative $\vec{u}_{0}=(u_{0,1},\ldots,u_{0,n})\in\big{(}C(\overline{\Omega})\big{)}^{n}$ there exists $\vec{f}=(f_{1},\ldots,f_{n})\in\mathscr{H}$, such that $\lim_{t\rightarrow\infty}\boldsymbol{\big{\\|}}\ \mathcal{T}(t)\,\vec{u}-\vec{f}\ \boldsymbol{\big{\\|}_{1}}=0.$ Proof The proof consists of several steps. To begin with we define the $\omega$-limit set $\omega(\vec{u}_{0})=\Big{\\{}\vec{g}\in\big{(}L^{1}(\Omega)\big{)}^{n}:\ \text{there exists a sequence \ }t_{k}\rightarrow\infty\\\ \text{\ as\ }k\rightarrow\infty,\text{\ such that\ }\lim_{k\rightarrow\infty}\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{g}\ \boldsymbol{\big{\\|}_{1}}=0\Big{\\}},$ (4.5) The organization of the proof is as follows. First we show that $\omega(\vec{u}_{0})$ is not empty. In the second step we define the Lyapunov functional $\mathcal{V}(\vec{\xi})=\boldsymbol{\big{\\|}}\vec{\xi}-\vec{w}\boldsymbol{\big{\\|}_{1}},$ where $\vec{w}$ is a stationary solution and check that it is constant on $\omega(\vec{u}_{0})$. We then deduce that $\omega(\vec{u}_{0})\subset\mathscr{H}$, and finally prove that $\omega(\vec{u}_{0})$ consists of exactly one function. Step 1. $\omega(\vec{u}_{0})$ is not empty. Let $\varepsilon>0$ be arbitrary. Suppose that $\Omega^{\prime}\subset\subset\Omega$ satisfy $\big{\|}\Omega\setminus\Omega^{\prime}\big{\|}\leqslant\frac{\varepsilon}{2K}.$ and set $K=\sum_{i=1}^{n}\frac{2}{\alpha_{i}}\\|h_{i}\\|_{C(\overline{\Omega})},$ (4.6) where $\vec{h}$ has been introduced in (4.4). We have already proved in Proposition 4.3 that $\mathcal{T}(t)\,\vec{u}_{0}$ is bounded in $\big{(}L^{\infty}(\Omega)\big{)}^{n}$. Therefore there exist a vector function $\vec{g}\in\big{(L^{\infty}}(\Omega))^{n}$ and a sequence $\\{\vec{u}(t_{k})\\}$ such that $\vec{u}(t_{k})\rightharpoonup\vec{g}\mbox{~{}~{}weakly in ~{}~{}}(L^{2}(\Omega))^{n},$ (4.7) as $t_{k}\rightarrow\infty$. Next we deduce from [16, Chap. III, Theorem 10.1] that there exists a positive constant $C$ such that $\big{\|}u_{i}(x_{1},t)-u_{i}(x_{2},t)\big{\|}\leqslant C\|x_{1}-x_{2}\|^{\alpha}$ for all $x_{1},\ x_{2}\in\Omega^{\prime}$ and all $t>0$. Therefore, it follows from the Ascoli-Arzelà Theorem (see, e.g., [1, Theorem 1.33]) that $\vec{u}(t_{k})\rightarrow\vec{g}$ as $t_{k}\rightarrow\infty$, uniformly in $\overline{\Omega}^{\prime}$. We choose $t_{0}$ large enough such that for all $t_{k}\geqslant t_{0}$ $\boldsymbol{\big{\\|}}\vec{u}(\cdot,t_{k})-\vec{g}(\cdot)\boldsymbol{\big{\\|}}_{\boldsymbol{1},\Omega^{\prime}}\leqslant\frac{\varepsilon}{2},$ (4.8) where $\boldsymbol{\big{\\|}}\cdot\boldsymbol{\big{\\|}}_{\boldsymbol{1},\Omega^{\prime}}$ corresponds to the $L^{1}$ norm in $\Omega^{\prime}$. We deduce that, in view of (4.6) and (4.7) that $\boldsymbol{\big{\\|}}\vec{u}(\cdot,t_{k})-\vec{g}(\cdot)\boldsymbol{\big{\\|}}_{\boldsymbol{1},\Omega\setminus\Omega^{\prime}}\leqslant K\big{\|}\Omega\setminus\Omega^{\prime}\big{\|}\leqslant\frac{\varepsilon}{2},$ which together with (4.8) yields $\boldsymbol{\big{\\|}}\vec{u}(\cdot,t_{k})-\vec{g}(\cdot)\boldsymbol{\big{\\|}_{1}}\leqslant\varepsilon.$ Step 2. $\omega(\vec{u}_{0})\subset\mathscr{H}$. Indeed, let $\vec{g}\in\omega(\vec{u}_{0})$ and suppose $\vec{g}\notin\mathscr{H}$. According to Proposition 4.2 we can find a steady state solution $\vec{w}\in\mathscr{H}$, such that at least one component of $\vec{w}-\vec{g}$ changes the sign. Without loss of generality we can assume that it happens for the first component, namely that $f_{1}-w_{1}$ changes the sign. We remark that, by the contraction property in Theorem 3.1, the functional $\mathcal{V}(\vec{\xi})=\boldsymbol{\big{\\|}}\vec{\xi}-\vec{w}\boldsymbol{\big{\\|}_{1}}$ is a Lyapunov functional for Problem (P), where $\vec{\xi}\in\big{(}L^{1}(\Omega)\big{)}^{n}$. Next we describe some of its properties. Property (a) The functional $\mathcal{V}$ is constant on $\omega(\vec{u}_{0})$. Since $\mathcal{T}(t)\,\vec{w}=\vec{w}$ and $\mathcal{T}(t)\,$ has the contraction property (3.4), the functional $\mathcal{V}$ is nonincreasing in time along the trajectory $\mathcal{T}(t)\,\vec{u}_{0}$, which yields $\begin{split}\mathcal{V}\big{(}\mathcal{T}(t)\,\vec{u}_{0}\big{)}&=\boldsymbol{\big{\\|}}\mathcal{T}(t)\,\vec{u}_{0}-\vec{w}\boldsymbol{\big{\\|}_{1}}\\\ &=\boldsymbol{\big{\\|}}\mathcal{T}(t)\,\vec{u}_{0}-\mathcal{T}(t)\,\vec{w}\boldsymbol{\big{\\|}_{1}}\leqslant\boldsymbol{\big{\\|}}\vec{u}_{0}-\vec{w}\boldsymbol{\big{\\|}_{1}}<\infty\hskip 5.0pt.\end{split}$ Thus there exists a finite limit $\mathcal{V}^{\ast}$ of $\mathcal{V}\big{(}\mathcal{T}(t)\,\vec{u}_{0}\big{)}$ as $t\rightarrow\infty$. Let $\vec{h}_{1},\ \vec{h}_{2}\in\omega(\vec{u}_{0})$. We can find a sequence $t_{k}\rightarrow\infty$ as $k\rightarrow\infty$, such that $\boldsymbol{\big{\\|}}\mathcal{T}(t_{2k})\,\vec{u}_{0}-\vec{h}_{1}\boldsymbol{\big{\\|}_{1}}\rightarrow 0\quad\text{and}\quad\boldsymbol{\big{\\|}}\mathcal{T}(t_{2k+1})\,\vec{u}_{0}-\vec{h}_{2}\boldsymbol{\big{\\|}_{1}}\rightarrow 0,$ as $k$ tends to $\infty$. It follows that $\mathcal{V}\big{(}\vec{h}_{1}\big{)}=\mathcal{V}\big{(}\vec{h}_{2}\big{)}=\mathcal{V}^{\ast}$. Property (b) The $\omega$-limit set $\omega(\vec{u}_{0})$ is invariant with respect to the semigroup $\mathcal{T}(t)\,$, namely if $\vec{h}\in\omega(\vec{u}_{0})$, then for all $t>0$ also $\mathcal{T}(t)\,\vec{h}\in\omega(\vec{u}_{0})$. Let the sequence $t_{k}\rightarrow\infty$ as $k\rightarrow\infty$ be such that $\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{h}\boldsymbol{\big{\\|}_{1}}\rightarrow 0$. From the contraction property (3.4) $\displaystyle\boldsymbol{\big{\\|}}\mathcal{T}(t_{k}+t)\,\vec{u}_{0}-\mathcal{T}(t)\,\vec{h}\boldsymbol{\big{\\|}_{1}}=\boldsymbol{\big{\\|}}\mathcal{T}(t)\,\mathcal{T}(t_{k})\,\vec{u}_{0}-\mathcal{T}(t)\,\vec{h}\boldsymbol{\big{\\|}_{1}}$ $\displaystyle\leqslant\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{h}\boldsymbol{\big{\\|}_{1}}.$ Since the last term above tends to $0$ as $k$ tends to $\infty$ this shows that $\mathcal{T}(t)\,\vec{h}\in\omega(\vec{u}_{0})$. Now, remember that $\vec{g}\in\omega(\vec{u}_{0})$ is such that $\vec{g}\notin\mathscr{H}$ and $\vec{w}\in\mathscr{H}$ is such that the first component of $\vec{w}-\vec{g}$ changes the sign in $\Omega$. Then, Corollary 3.3 yields $\rule{20.0pt}{0.0pt}\mathcal{V}(\mathcal{T}(t)\,\vec{g})=\boldsymbol{\big{\\|}}\mathcal{T}(t)\,\vec{g}-\vec{w}\boldsymbol{\big{\\|}_{1}}\\\ =\boldsymbol{\big{\\|}}\mathcal{T}(t)\,\vec{g}-\mathcal{T}(t)\,\vec{w}\boldsymbol{\big{\\|}_{1}}<\boldsymbol{\\|}\vec{g}-\vec{w}\boldsymbol{\\|_{1}}=\mathcal{V}(\vec{g}),\rule{20.0pt}{0.0pt}$ for all $t>0$, which contradicts Property (a). Therefore $\vec{g}\in\mathscr{H}$. Step 3. The set $\omega(\vec{u}_{0})$ contains only one element. Suppose that $\vec{g}_{1},\vec{g}_{2}\in\omega(\vec{u}_{0})$. Then we can find two sequences $t_{k},\,s_{k}$ tending to $\infty$ as $k\rightarrow\infty$, such that $s_{k}\leqslant t_{k}$ and $\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{g}_{1}\boldsymbol{\big{\\|}_{1}},\,\boldsymbol{\big{\\|}}\mathcal{T}(s_{k})\,\vec{u}_{0}-\vec{g}_{2}\boldsymbol{\big{\\|}_{1}}\rightarrow 0$ as $t_{k}\rightarrow\infty$. Since $\omega(\vec{u}_{0})\subset\mathscr{H}$, it follows that $\displaystyle\boldsymbol{\\|}\vec{g}_{1}-\vec{g}_{2}\boldsymbol{\\|_{1}}\leqslant\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{g}_{1}\boldsymbol{\big{\\|}_{1}}+\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{g}_{2}\boldsymbol{\big{\\|}_{1}}$ $\displaystyle=\,\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{g}_{1}\boldsymbol{\big{\\|}_{1}}+\boldsymbol{\big{\\|}}\mathcal{T}(t_{k}-s_{k})\,\mathcal{T}(s_{k})\,\vec{u}_{0}-\mathcal{T}(t_{k}-s_{k})\,\vec{g}_{2}\boldsymbol{\big{\\|}_{1}}$ $\displaystyle\leqslant\boldsymbol{\big{\\|}}\mathcal{T}(t_{k})\,\vec{u}_{0}-\vec{g}_{1}\boldsymbol{\big{\\|}_{1}}+\boldsymbol{\big{\\|}}\mathcal{T}(s_{k})\,\vec{u}_{0}-\vec{g}_{2}\boldsymbol{\big{\\|}_{1}}\hskip 3.0pt,$ which tends to $0$ as $k\rightarrow\infty$. $\blacksquare$ ## 5 Stationary solutions for the linear molecular motor problem In this section we show the existence and the uniqueness (up to a multiplicative constant) of the classical stationary solution of the problem for the molecular motor. We suppose that $\Omega$ is an open bounded subset of $\mathbb{R}^{d}$ with smooth boundary $\partial\Omega$. We consider the linear system $\text{div}\big{(}\sigma_{i}\nabla v_{i}(x)+v_{i}(x)\nabla\psi_{i}(x)\big{)}+\sum_{j=1}^{n}a_{ij}v_{j}(x)=0\quad\text{in}\quad\Omega,$ (5.1) where $i\in\\{1,\ldots,n\\}$, $n>1$. The system (5.1) is supplemented with the Robin boundary conditions $\sigma_{i}\frac{\partial v_{i}}{\partial\nu}+v_{i}\frac{\partial\psi_{i}}{\partial\nu}=0\quad\text{on}\quad\partial\Omega,$ (5.2) where $i\in\\{1,\ldots,n\\}$. Thus, the problem can be written as ${\cal A}\vec{v}=0,$ with a linear operator $\cal A$ in a suitable Banach space $\cal X$ of functions on $\Omega$, to be made precise later. Moreover, we impose the integral constraint $\sum_{i=1}^{n}\int_{\Omega}v_{i}(x)\,\text{d}x=1.$ (5.3) The adjoint problem ${\cal A}^{\ast}\vec{\varphi}=0$ to (5.1), in a dual space ${\cal X}^{\ast}$, is now $\sigma_{i}\Delta\varphi_{i}-\nabla\psi_{i}\cdot\nabla\varphi_{i}+\sum_{j=1}^{n}a_{ji}\varphi_{j}=0,\quad\text{in}\quad\Omega,$ (5.4) with the Neumann boundary conditions for each $i=1,\,\dots,\,n$ $\frac{\partial\varphi_{i}}{\partial\nu}=0\quad\text{on}\quad\partial\Omega.$ (5.5) Since $\sum_{j=1}^{n}a_{ji}=0$, the problem (5.4) has the obvious solution $\vec{\varphi}=(\varphi_{1},\dots,\varphi_{n})=(1,\dots,1).$ (5.6) We are going to apply the Krein-Rutman theorem on the first eigenvalues and eigenvectors of positive operators, and this will permit us to conclude that the problem (5.1)–(5.2) has a one-dimensional space of solutions. Therefore, under the additional constraint (5.3), the original problem (5.1)–(5.2) has a unique solution. Perthame and Souganidis sketched this argument for $n>1$ and $d=1$ in [20]. ###### Theorem 5.1. Under the assumption $\sum_{j=1}^{n}a_{ji}=0$, there exists a unique smooth solution $\vec{v}$ of the system (5.1)–(5.3). Before proving Theorem 5.1 we recall some basic definitions as well as the Krein-Rutman theorem from [9, Ch. VIII, p. 188–191]. ###### Definition 5.2 (Reproducing cone). We say that a closed set $K$ in $\cal X$ is a cone, if it possesses the following properties: * i) $0\in K$, * ii) $u,\ v\in K\Longrightarrow\alpha u+\beta v\in K$, for all $\alpha,\ \beta\geqslant 0$, * iii) $v\in K$ and $-v\in K\Longrightarrow v=0$. A cone $K\subset\cal X$ is said to be reproducing if ${\cal X}=K-K\equiv\big{\\{}k_{1}-k_{2}:\ k_{1},\ k_{2}\in K\big{\\}}$. ###### Definition 5.3 (Dual cone). If $K$ is a cone in $\cal X$, then the set $K^{\ast}\subset{\cal X}^{\ast}$ is said to be a dual cone if $\langle f^{\ast},v\rangle\geqslant 0,$ for every $v\in K$. ###### Definition 5.4 (Strict positivity). Let $\cal B$ be a linear operator on $\cal X$. Then $\cal B$ is said to be strongly positive if ${\cal B}v\in K^{o}$ for all $v\in K$ such that $v\neq 0$. ###### Theorem 5.5. Let $K$ be a reproducing cone in a Banach space $\cal X$, with nonempty interior $K^{o}\neq\emptyset$, and let $\cal B$ be a strongly positive compact operator on $K$ in a sense of Definition 5.4. Then the spectral radius of $\cal B$, $r(\cal B)$, is a simple eigenvalue of $\cal B$ and $\cal B^{\ast}$, and their associated eigenvectors belong to $K^{o}$ and $(K^{\ast})^{o}$. More precisely, there exists a unique associated eigenvector in $K^{o}$ (resp. $(K^{\ast})^{o}$) of norm $1$. Furthermore, all other eigenvalues are strictly less in absolute value than $r(\cal B)$. Proof We will apply Theorem 5.5 to the space ${\cal X}=\big{(}C(\overline{\Omega})\big{)}^{n}\subset\big{(}L^{1}(\Omega)\big{)}^{n}$ endowed with the usual supremum norm, and the operators $\displaystyle{\cal B}=(\lambda I-{\cal A})^{-1}:\ {\cal X}\to{\cal X},$ $\displaystyle{\cal B}^{\ast}=(\lambda I-{\cal A}^{\ast})^{-1}:\ {\cal X}^{\ast}\to{\cal X}^{\ast},$ where $\lambda>0$ is a strictly positive real number to be fixed later. Let $K=\big{\\{}\vec{u}\in{\cal X}:\ u_{i}(x)\geqslant 0\ {\rm for\ each\ }x\in\overline{\Omega},\ i=1,\dots,n\big{\\}}.$ We remark that $K$ is a reproducing cone, with nonempty interior $K^{o}=\big{\\{}\vec{u}\in{\cal X}:\ \inf_{x\in\overline{\Omega}}u_{i}(x)>0,\ i=1,\dots,n\big{\\}}.$ From the standard theory [17, Theorem 2.1 and Theorem 3.1, Ch. 7] for elliptic partial differential linear systems, the boundary value problem $\sigma_{i}\Delta\varphi_{i}-\nabla\psi_{i}\cdot\nabla\varphi_{i}+\sum_{i=1}^{n}a_{ji}\varphi_{j}-\lambda\varphi_{i}=f_{i}\ \ {\rm in\ \ }\Omega,$ (5.7) with the homogeneous Neumann conditions (5.5) on $\partial\Omega$, for $\lambda=\widetilde{\lambda}>0$ sufficiently large, has a solution $\vec{\varphi}=(\varphi_{1},\dots,\varphi_{n})\in{\cal X}$ for each $\vec{f}=(f_{1},\dots,f_{n})\in{\cal X}$. Moreover, if $f_{i}(x)\geqslant 0$ for each $i=1,\dots,n$, and $x\in\overline{\Omega}$, then $\varphi_{i}(x)\geqslant 0$ (in fact, $\varphi_{i}(x)>0$ in $\Omega$), which is a consequence of the maximum principle (cf. also Example 3 on p. 196–197 in [9]). Thus, the operator ${\cal B}^{\ast}=\big{(}\widetilde{\lambda}I-{\cal A}^{\ast}\big{)}^{-1}$ is a strongly positive and compact operator, and by Theorem 5.5, the largest eigenvalue $\mu$ of $\cal B$ and ${\cal B}^{\ast}$ is simple. Since $\displaystyle-\sigma_{i}\Delta\varphi_{i}+\nabla\psi_{i}\cdot\nabla\varphi_{i}-\sum_{j=1}^{n}a_{ji}\varphi_{j}+\widetilde{\lambda}\varphi_{i}=\widetilde{\lambda}\varphi_{i}\quad\text{in}\quad\Omega$ $\displaystyle\frac{\partial\varphi_{i}}{\partial\nu}=0\quad\text{on}\quad\partial\Omega,$ for all $i\in\\{1,\ldots,n\\}$, with $\vec{\varphi}=(\varphi_{1},\ldots,\varphi_{n})=(1,\ldots,1)$, and since $(1,\ldots,1)\in(K^{\ast})^{o}$, it follows that $\displaystyle\frac{1}{\widetilde{\lambda}}=r\Big{(}\big{(}\widetilde{\lambda}I-{\cal A}^{\ast}\big{)}^{-1}\Big{)}$ is a simple eigenvalue of the operator $\big{(}\widetilde{\lambda}I-{\cal A}^{\ast}\big{)}^{-1}$. Applying again Theorem 5.5, we deduce that $\displaystyle\frac{1}{\widetilde{\lambda}}$ is the largest eigenvalue of the operator $\big{(}\widetilde{\lambda}I-{\cal A}\big{)}^{-1}$ and that it is simple, and that there exists $\vec{v}\in K^{o}\subset\cal X$ such that $\Big{(}\widetilde{\lambda}I-{\cal A}\Big{)}^{-1}\vec{v}=\frac{1}{\widetilde{\lambda}}\,\vec{v},$ which is equivalent to ${\cal A}\vec{v}=0.$ This proves the existence of the solution of the problem (5.1)–(5.3). $\blacksquare$ ## References * [1] R. A. Adams, Sobolev Spaces, Pure and Applied Mathematics, vol. 65, Academic Press, New York-London, 1975\. * [2] N. D. Alikakos, P. Hess, H. Matano, Discrete order preserving semigroups and stability for periodic parabolic differential equations, J. Differential Equations, 1989, 2, 82, 322–341. * [3] M. Bertsch and D. Hilhorst, A density dependent diffusion equation in population dynamics: stabilization to equilibrium, SIAM J. Math. Anal., 1986, 4, 17, 863–883. * [4] D. Bothe, Instantaneous limits of reversible chemical reactions in presence of macroscopic convection, J. Differential Equations, 2003, 1, 193, 27–48. * [5] D. Bothe and D. Hilhorst, A reaction-diffusion system with fast reversible reaction, J. Math. Anal. Appl., 2003, 1, 268, 125–135. * [6] M. Chipot, S. Hastings, and D. Kinderlehrer, Transport in a molecular motor system, M2AN Math. Model. Numer. Anal., 2004, 38, 6, 1011- 1034. * [7] M. Chipot, D. Kinderlehrer, and M. Kowalczyk, 2004 A variational principle for molecular motors, Meccanica, 38, 505–518 * [8] M. G. Crandall and L. Tartar, Some relations between nonexpansive and order preserving mappings, Proc. Amer. Math. Soc., 1980, 3, 78, 385–390. * [9] R. Dautray, J.-L. Lions, Mathematical Analysis and Numerical Methods for Science and Technology, Volume 3: Spectral Theory and Applications, Springer-Verlag, Berlin, 1990. * [10] L. Desvillettes and K. Fellner, Exponential decay toward equilibrium via entropy methods for reaction-diffusion equations, J. Math. Anal. Appl., 2006, 1, 319, 157–176. * [11] P. Érdi and J. Tóth, Mathematical Models of Chemical Reactions, Nonlinear Science: Theory and Applications, Princeton University Press, Princeton, NJ, 1989\. * [12] Hastings, S., Kinderlehrer, D. and McLeod, J.B. Diffusion mediated transport in multiple state systems, An elliptic-parabolic problem in combustion theory: convergence to travelling waves, SIAM J. Math. Anal., 2007, 39.4, 1208-1230. * [13] Hastings, S., Kinderlehrer, D. and McLeod, J.B. Diffusion mediated transport with a look at motor proteins, Recent advances in nonlinear analysis, (Chipot, M., Lin, C-S,., and Tsai, D-H., eds) (2008) World Scientific, 95 - 112. * [14] D. Hilhorst and J. Hulshof, An elliptic-parabolic problem in combustion theory: convergence to travelling waves, Nonlinear Analysis TMA, 1991, 6, 17, 519–546. * [15] D. Hilhorst and M. A. Peletier, Convergence to travelling waves in a reaction-diffusion system arising in contaminant transport, J. Differential Equations, 2000, 1, 163, 89–112. * [16] O. A. Ladyženskaja and V. A. Solonnikov and N. N. Ural’ceva, Linear and Quasilinear Equations of Parabolic Type, Translations of Mathematical Monographs, Vol. 23, American Mathematical Society, Providence, R.I., 1967\. * [17] O. A. Ladyženskaja and N. N. Ural’ceva, Linear and Quasilinear Elliptic Equations, Academic Press, New York, 1968\. * [18] S. Osher and J. Ralston, $L^{1}$ stability of travelling waves with applications to convective porous media flow, Comm. Pure Appl. Math., 1982, 6, 35, 737–749. * [19] B. Perthame, The general relative entropy principle—applications in Perron-Frobenius and Floquet theories and a parabolic system for biomotors, Rend. Accad. Naz. Sci. XL Mem. Mat. Appl. (5), 2005, 29, 307–325. * [20] B. Perthame, P. E. Souganidis, Asymmetric potentials and motor effect: a large deviation approach, 1–16, Arch. Rat. Mech. Anal., to appear, 2008. * [21] M. H. Protter and H. F. Weinberger, Maximum Principles in Differential Equations, Springer-Verlag, New York, 1984\.	arxiv-papers	2008-08-02T11:19:47	2024-09-04T02:48:57.138503	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Chipot, D. Hilhorst, D. Kinderlehrer and M. Olech", "submitter": "Michal Olech Mr", "url": "https://arxiv.org/abs/0808.0250" }
0808.0251	Numerical approximation of a reaction-diffusion system with fast reversible reaction Robert Eymard 111Université Paris-Est, 77454 Marne-la-Vallée Cédex 2, France; E-mail: Robert.Eymard@univ-mlv.fr, Danielle Hilhorst 222Laboratoire de Mathématiques, CNRS and Université de Paris-Sud 11, 91405 Orsay Cédex, France; E-mail: Danielle.Hilhorst@math.u-psud.fr, Michal Olech 333Instytut Matematyczny Uniwersytetu Wroclawskiego, pl. Grunwaldzki 2/4, 50-384 Wroclaw, Polska; Laboratoire de Mathématiques, CNRS Université de Paris-Sud, 91405 Orsay Cédex, France; E-mail: olech@math.uni.wroc.pl. The preparation of this article has been partially supported by a Marie Curie Transfer of Knowledge Fellowship of the European Community’s Sixth Framework Programme under contract number MTKD-CT-2004-013389 ###### Abstract We consider the finite volume approximation of a reaction-diffusion system with fast reversible reaction. We deduce from a priori estimates that the approximate solution converges to the weak solution of the reaction – diffusion problem and satisfies estimates which do not depend on the chemical kinetics factor. It follows that the solution converges to the solution of a nonlinear diffusion problem, as the size of the volume elements and the time steps converge to zero while the kinetic rate tends to infinity. Key words: instantaneous reaction limit, mass-action kinetics, finite volume methods, convergence of approximate solutions, discrete a priori estimates, Kolmogorov’s theorem. AMS subject classification: 35K45, 35K50, 35K55, 65M12, 65N12, 65N22, 80A30, 92E20. ## 1 Introduction In this paper, we consider chemical systems with fast reactions where mean reaction times vary from approximately $10^{-14}$ second to 1 minute. In particular, reactions that involve bond making or breaking are not likely to occur in less than $10^{-13}$ second. Moreover, chemical systems almost always involve some elementary reaction steps that are reversible and fast. The study of reactions with rates that are outside of the time frame of ordinary laboratory operations requires specialized instrumentation, techniques and ways of proceeding (see for example Espenson [4, Chapter 11]). This work tries to give an efficient, quick and cheap way for numerical investigations of such reactions. In this article, we consider a reversible chemical reaction between mobile species $\mathcal{A}$ and $\mathcal{B}$, that takes place inside a bounded region $\Omega\subset\mathbb{R}^{d}$ where $d=1,2$ or $3$. If the region is isolated and diffusion is modelled by Fick’s law, this leads to the reaction- diffusion system of partial differential equations $\begin{split}u_{t}&=a\Delta u-\alpha k\big{(}r_{A}(u)-r_{B}(v)\big{)}\text{\quad in \quad}\Omega\times(0,T),\\\ v_{t}&=b\Delta v+\beta k\big{(}r_{A}(u)-r_{B}(v)\big{)}\text{\quad in \quad}\Omega\times(0,T),\end{split}$ (1) where $T>0$ and $\Omega$ is a bounded set of $\mathbb{R}^{d}$. An example of explicit expressions and values for $\alpha,\ \beta,\ k,\ r_{A},\ r_{B},\ a,\ b$ is given in Section 6. We supplement the system (1) by the homogeneous Neumann boundary conditions $\nabla u\cdot\boldsymbol{n}=\nabla v\cdot\boldsymbol{n}=0\text{\quad on \quad}\partial\Omega\times(0,T),$ (2) and the initial conditions of the form $u(x,0)=u_{0}(x),\ v(x,0)=v_{0}(x)\text{\quad in \quad}\Omega.$ (3) In the sequel we call the system (1) together with the boundary conditions (2) and the initial conditions (3), Problem $\mathcal{P}^{k}$. For a reversible reaction $m\mathcal{A}\rightleftharpoons n\mathcal{B}$ one has $\alpha=-m,\ \beta=n$ and the rate functions are of the form $r_{A}(u)=u^{m}$ and $r_{B}(v)=v^{n}$. Further discussion about this motivation and some concrete examples can be found in Érdi and Tóth [9] and Espenson [4]. In practice, especially for ionic or radical reactions, changes due to reaction are often very fast compared to diffusive effects. This corresponds to a large rate constant $k$. Bothe and Hilhorst [1] study the limit to an instantaneous reaction. They exploit a natural Lyapunov functional and use compactness arguments to prove that $u^{k}\rightarrow u\quad\text{and}\quad v^{k}\rightarrow v\quad\text{in}\quad L^{2}\big{(}\Omega\times(0,T)\big{)},$ as $k$ tends to infinity, where $(u^{k},v^{k})$ is the solution of Problem $\mathcal{P}^{k}$ and the limit $(u,v)$ is determined by $r_{A}(u)=r_{B}(v)\quad\text{and}\quad\frac{u}{\alpha}+\frac{v}{\beta}=w,$ (4) where $w$ is the unique weak solution of the nonlinear diffusion problem $\begin{split}w_{t}&=\Delta\phi(w)\quad\text{in}\quad\Omega\times(0,T)\\\ \frac{\partial\phi(w)}{\partial\boldsymbol{n}}&=0\quad\text{on}\quad\partial\Omega\times(0,T)\\\ w(x,0)&=w_{0}(x):=\frac{1}{\alpha}u_{0}(x)+\frac{1}{\beta}v_{0}(x)\quad\text{in}\quad\Omega\hskip 3.0pt,\end{split}$ (5) with $\begin{split}\phi:=\bigg{(}\frac{a}{\alpha}\text{id}+\frac{b}{\beta}\eta\bigg{)}&\circ\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\quad\text{on}\quad\mathbb{R}^{+},\\\ \eta&=r_{B}^{-1}\circ r_{A}.\end{split}$ (6) The identities in (4) can be explained as follows: the first one states that the system is in chemical equilibrium, while the second one defines $w$ as the quantity that is conserved under the chemical reaction. Given a function $w$, the system (4) can be uniquely solved for $(u,v)$ if $r_{A},\,r_{B}$ are strictly increasing with for instance $r_{A}(\mathbb{R}^{+})\subset r_{B}(\mathbb{R}^{+})$ so that $\eta=r_{B}^{-1}\circ r_{A}$ is well defined and strictly increasing. Under these assumptions $u$ is the unique solution of $\frac{1}{\alpha}u+\frac{1}{\beta}\eta(u)=w,$ which gives the explicit representation of $u$ and $v$ $u=\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\hskip-4.0pt\big{(}w\big{)},\quad v=\eta\circ\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\hskip-4.0pt\big{(}w\big{)}.$ (7) We assume the following hypotheses, which we denote by $\mathcal{H}$: 1. 1. Let $\Omega$ be an open, connected and bounded subset of $\mathbb{R}^{d}$, where $d=1,2$ or $3$, with a smooth boundary $\partial\Omega$, 2. 2. $u_{0}(x),\ v_{0}(x)\in L^{\infty}(\Omega)$ and there exist constants $U,\ V>0$ such that $0\leqslant u_{0}(x)\leqslant U$ and $0\leqslant v_{0}(x)\leqslant V$ in $\Omega$, 3. 3. $\alpha$, $\beta$, $a$, $b$ and $k$ are strictly positive real values (sometimes we use the notation $k\alpha=\hat{\alpha}$ and $k\beta=\hat{\beta}$), 4. 4. Let $r_{A}(x),\ r_{B}(x)\in C^{1}(\mathbb{R})$ be strictly increasing functions, such that $r_{A}(0)=r_{B}(0)=0$, and assume furthermore that $r_{A}(\mathbb{R}^{+})\subset r_{B}(\mathbb{R}^{+})$. We recall from Bothe and Hilhorst [1, Section 2] that Problem $\mathcal{P}^{k}$ has a unique classical solution $(u^{k},v^{k})$ on every finite time interval $[0,T]$, for all nonnegative bounded initial data. By classical solution, we mean a function pair $(u^{k},\,v^{k})$ such that $u^{k},\,v^{k}\in C^{2,1}\big{(}\Omega\times(0,T]\big{)}\cap C^{1,0}\big{(}\overline{\Omega}\times(0,T]\big{)}$ with $u^{k},v^{k}\in C\big{(}[0,T];\,L^{2}(\Omega)\big{)}$ (see also Ladyženskaja, Solonnikov and Ural’ceva [11]). Next we present a notion of a weak solution of Problem $\mathcal{P}^{k}$, which will be used in the sections 4 and 5. ###### Definition 1.1. We say that $(u^{k},v^{k})$ is a weak solution to Problem $\mathcal{P}^{k}$ if and only if 1. 1. $u^{k},v^{k}\in L^{2}\big{(}0,T;H^{1}(\Omega)\big{)}$ and $u^{k}_{t},v^{k}_{t}\in L^{2}\big{(}0,T;(H^{1}(\Omega))^{\prime}\big{)}$; 2. 2. Let $\Psi$ be the set of test functions, defined as $\Psi=\Big{\\{}\psi\in C^{2,1}(\overline{\Omega}\times[0,T]):\nabla\psi\cdot\boldsymbol{n}=0\text{\quad on \quad}\partial\Omega\times[0,T]\text{\quad and \quad}\psi(T)=0\Big{\\}}.$ For a.e. $t\in(0,T)$ and all $\psi\in\Psi$ $\int_{\Omega}u_{0}(x)\psi(x,0)\,\text{d}x+\int_{\Omega}u^{k}(x,t)\psi_{t}(x,t)\,\text{d}x+a\int_{\Omega}u^{k}(x,t)\Delta\psi(x,t)\,\text{d}x\\\ -\hat{\alpha}\int_{\Omega}\psi(x,t)\Big{(}r_{A}\big{(}u^{k}(x,t)\big{)}-r_{B}\big{(}v^{k}(x,t)\big{)}\Big{)}\,\text{d}x=0$ (8) and $\int_{\Omega}v_{0}(x)\psi(x,0)\,\text{d}x+\int_{\Omega}v^{k}(x,t)\psi_{t}(x,t)\,\text{d}x+b\int_{\Omega}u^{k}(x,t)\Delta\psi(x,t)\,\text{d}x\\\ +\hat{\beta}\int_{\Omega}\psi(x,t)\Big{(}r_{A}\big{(}u^{k}(x,t)\big{)}-r_{B}\big{(}v^{k}(x,t)\big{)}\Big{)}\,\text{d}x=0.$ (9) We remark that every essentially bounded weak solution of Problem $\mathcal{P}^{k}$, in the sense of Definition 1.1, is also a classical solution. This paper is organized as follows. In section 2 we define a finite volume discretisation and an approximate solution $(u_{\cal D}^{k},v_{\cal D}^{k})$ for Problem $\mathcal{P}^{k}$. In section 3 we prove a discrete comparison principle which yields discrete $L^{\infty}$ estimates, and we show the existence and uniqueness of the approximate solution. Section 4 contains technical lemmas used further in the convergence proofs. The convergence of the approximate solution to the classical solution of Problem $\mathcal{P}^{k}$ in the case of fixed $k$ is proved in section 5. In section 6 we use a suitable Lyapunov function and we obtain a discrete $L^{2}\big{(}0,T;H^{1}(\Omega)\big{)}$ estimate, which does not depend on $k$. We then apply Kolmogorov’s theorem and deduce the convergence of the approximate solutions to the classical solution of Problem $\mathcal{P}^{k}$. Afterwards we show that the approximate solution $(u_{\cal D}^{k},v_{\cal D}^{k})$ converges to $(u,v)$ defined in (7) as $k$ tends to $\infty$ and the size of the discretisation parameters tends to zero. In Section 6 we present numerical results obtained with our finite volume scheme, for the reversible dimerisation of $o$-phenylenedioxydimethylsilane ($2,2$-dimethyl-$1,2,3$-benzodioxasilole) which is a reaction of the type $2\mathcal{A}\rightleftharpoons\mathcal{B}$ (see Meyer, Klein and Weiss [12]). On the one hand we compute the approximate solution $(u_{\cal D}^{k},v_{\cal D}^{k})$ of the solution $(u^{k},v^{k})$ of Problem $\mathcal{P}^{k}$ and on the other hand the numerical approximation $w_{\cal D}$ of the solution $w$ of the problem (5) – (6), and we check that $\rule{20.0pt}{0.0pt}u_{\cal D}^{k}\approx\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\hskip-4.0pt\big{(}w_{\cal D}\big{)}\quad\text{and}\quad v_{\cal D}^{k}\approx\eta\circ\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\hskip-4.0pt\big{(}w_{\cal D}\big{)}\rule{20.0pt}{0.0pt}$ for $k$ large enough and $\hbox{\rm size}\,({\cal D})$ small enough. ###### Remark 1.2. In what follows we denote by $C$, $C_{k}$ and $C_{\psi}$ positive generic constants which may vary from line to line. ## 2 The finite volume scheme The finite volume method has first been developed by engineers in order to study complex, coupled physical problems where the conservation of quantities such as masses, energy or impulsion must be carefully respected by the approximate solution. Another advantage of this method is that a large variety of meshes can be used in the computations. The finite volume methods are particularly well suited for numerical investigations of conservations laws. They are one of the most popular methods among the engineers performing computations for industrial purposes: the modelling of flows in porous media, problems related to oil recovery, questions related to hydrology, such as the numerical approximation of a stationary incompressible Navier – Stokes equations. For a comprehensive discussion about the finite volume method, we refer to Eymard, Gallouët and Herbin [6] and the references therein. Following [6], we define a finite volume discretization of $Q_{T}$. ###### Definition 2.1 (Admissible mesh of $\Omega$). An admissible mesh ${\cal M}$ of $\Omega$ is given by a set of open, bounded subsets of $\Omega$ (control volumes) and a family of points (one per control volume), satisfying the following properties 1. 1. The closure of the union of all the control volumes is $\overline{\Omega}$. We denote by $m_{K}$ the measure of each volume element $K$ and $\hbox{\rm size}\,({\cal M})=\max_{K\in{\cal M}}m_{k}.$ 2. 2. $K\cap L=\emptyset$ for any $(K,L)\in{\cal M}^{2}$, such that $K\neq L$. If $\overline{K}\cap\overline{L}\neq\emptyset$, then it is a subset of a hyperplane in $\mathbb{R}^{d}$. Let us denote by ${\cal E}\subset{\cal T}^{2}$ the set of pairs $(K,L)$, such that the $d-1$ Lebesgue measure of $\overline{K}\cap\overline{L}$ is strictly positive. For $(K,L)\in{\cal E}$ we write $K\|L$ for the set $\overline{K}\cap\overline{L}$ and $m_{K\|L}$ for the $d-1$ Lebesgue measure of $K\|L$. 3. 3. For any $K\in{\cal M}$ we also define ${\cal N}_{K}=\\{L\in{\cal T},\ (K,L)\in{\cal E}\\}$ and assume that $\partial K=\overline{K}\backslash K=\big{(}\overline{K}\cap\partial\Omega\big{)}\cup\Big{(}\bigcup_{L\in{\cal N}_{K}}K\|L\Big{)}$. 4. 4. There exists a family of points $(x_{K})_{K\in{\cal M}}$, such that $x_{K}\in K$ and if $L\in{\cal N}_{K}$ then the straight line $(x_{K},x_{L})$ is orthogonal to $K\|L$. We set $d_{K\|L}=d(x_{K},x_{L})\quad\text{and}\quad T_{K\|L}=\frac{m_{K\|L}}{d_{K\|L}},$ where the last quantity is sometimes called the transmissibility across the edge $K\|L$. Since Problem Problem $\mathcal{P}^{k}$is a time evolution problem, we also need to discretize the time interval $(0,T).$ ###### Definition 2.2 (Time discretization). A time discretization of the interval $(0,T)$ is given by an integer value $N$ and by a strictly increasing sequence of real values $(t^{(n)})_{n\in\\{0,\ldots,N+1\\}}$ with $t^{(0)}=0$ and $t^{(N+1)}=T$. The time steps are defined by $t_{\delta}^{(n)}=t^{(n+1)}-t^{(n)}\quad\text{for}\quad n\in\\{0,\ldots,N\\}.$ We may then define a discretization of the whole domain $Q_{T}$ in the following way. ###### Definition 2.3 (Discretization of $Q_{T}$). A finite volume discretization ${\cal D}$ of $Q_{T}$ is defined as ${\cal D}=\Big{(}{\cal M},{\cal E},(x_{K})_{K\in{\cal T}},(t^{(n)})_{n\in\\{0,\ldots,N+1\\}}\Big{)},$ where ${\cal M}$, ${\cal E}$ and $(x_{K})_{K\in{\cal T}}$ are given in Definition 2.1 and the sequence $(t^{(n)})_{n\in\\{0,\ldots,N+1\\}}$ is a time discretization of $(0,T)$ in the sense of Definition 2.2. One then sets $\hbox{\rm size}\,({\cal D})=\max\Big{\\{}\hbox{\rm size}\,({\cal M}),t_{\delta}^{(n)}:\ {n\in\\{0,\ldots,N\\}}\Big{\\}}.$ We present below the finite volume scheme which we use and define approximate solutions. We assume that the hypotheses $\mathcal{H}$ hold and suppose that ${\cal D}$ be an admissible discretization of $Q_{T}$ in the sense of Definition 2.3. We prescribe the approximate initial conditions $u_{K}^{(0)}=\frac{1}{m_{K}}\int_{K}u_{0}(x)\,\text{d}x\quad\text{and}\quad v_{K}^{(0)}=\frac{1}{m_{K}}\int_{K}v_{0}(x)\,\text{d}x,$ (10) where $K\in{\cal M}$, and associate to Problem $\mathcal{P}^{k}$ the finite volume scheme $\begin{split}m_{K}\big{(}u_{K}^{(n+1)}&-u_{K}^{(n)}\big{)}-t_{\delta}^{(n)}a\sum_{L\in{\cal N}_{K}}T_{K\|L}(u_{L}^{(n+1)}-u_{K}^{(n+1)})+t_{\delta}^{(n)}m_{K}\hat{\alpha}\Big{(}r_{A}\big{(}u_{K}^{(n+1)}\big{)}-r_{B}\big{(}v_{K}^{(n+1)}\big{)}\Big{)}=0,\\\ m_{K}\big{(}v_{K}^{(n+1)}&-v_{K}^{(n)}\big{)}-t_{\delta}^{(n)}b\sum_{L\in{\cal N}_{K}}T_{K\|L}(v_{L}^{(n+1)}-v_{K}^{(n+1)})-t_{\delta}^{(n)}m_{K}\hat{\beta}\Big{(}r_{A}\big{(}u_{K}^{(n+1)}\big{)}-r_{B}\big{(}v_{K}^{(n+1)}\big{)}\Big{)}=0.\end{split}$ (11) Note that (11) is a nonlinear system of equations in the unknowns $(u_{K}^{(n+1)},v_{K}^{(n+1)})_{K\in{\cal M},\ n\in\\{0,\ldots,N\\}}.$ For $x\in\Omega$ and $t\in(0,T)$ let $K\in\cal M$ be such that $x\in K$ and $n\in\\{0,\ldots,N\\}$ be such that $t^{(0)}=0$, $t^{(N+1)}=T$ and $t\in\big{(}t^{(n)},t^{(n+1)}\big{]}$. We can then define the approximate solutions $u_{\cal D}(x,t)=u_{K}^{(n+1)}\mbox{\quad and \quad}v_{\cal D}(x,t)=v_{K}^{(n+1)}\hskip 5.0pt.$ (12) In the next section, we will prove the existence and uniqueness of the solution of the discrete problem (11), together with the initial values (10). ## 3 The approximate solution In this section we prove the existence and uniqueness of the solution of the system (11). Let us start with a discrete version of the comparison principle. ###### Proposition 3.1 (Discrete comparison principle). We suppose that the hypotheses $\mathcal{H}$ are satisfied. Let $\cal D$ be a discretization as in Definition 2.3. Let $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$ and $(\tilde{u}_{K}^{(0)},\tilde{v}_{K}^{(0)})_{K\in{\cal M}}$ be given sequences of real values such that $u_{K}^{(0)}\leqslant\tilde{u}_{K}^{(0)}\quad\text{and}\quad v_{K}^{(0)}\leqslant\tilde{v}_{K}^{(0)},$ for all $K\in{\cal M}$. If the sequences $(u_{K}^{(n+1)},v_{K}^{(n+1)})_{K\in{\cal M},\ n\in\\{0,\ldots,N\\}}$ and $(\tilde{u}_{K}^{(n+1)},\tilde{v}_{K}^{(n+1)})_{K\in{\cal M},\ n\in\\{0,\ldots,N\\}}$ satisfy the equations (11) with the initial values $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$ and $(\tilde{u}_{K}^{(0)},\tilde{v}_{K}^{(0)})_{K\in{\cal M}}$, respectively, then for $K\in{\cal M}$ and $n\in\\{0,\ldots,N\\}$ $u_{K}^{(n+1)}\leqslant\tilde{u}_{K}^{(n+1)}\quad\text{and}\quad v_{K}^{(n+1)}\leqslant\tilde{v}_{K}^{(n+1)}.$ (13) Proof We set $\hat{u}_{K}^{(n)}=u_{K}^{(n)}-\tilde{u}_{K}^{(n)}$ and $\hat{v}_{K}^{(n)}=v_{K}^{(n)}-\tilde{v}_{K}^{(n)}$ for all $K\in{\cal M}$ and $n\in\\{0,\ldots,N+1\\}$ and define $\begin{split}\hat{A}_{K}^{(n+1)}=&\Big{(}r_{A}\big{(}u_{K}^{(n+1)}\big{)}-r_{A}\big{(}\tilde{u}_{K}^{(n+1)}\big{)}\Big{)}/\hat{u}_{K}^{(n+1)},\\\ \hat{B}_{K}^{(n+1)}=&\Big{(}r_{B}\big{(}v_{K}^{(n+1)}\big{)}-r_{B}\big{(}\tilde{v}_{K}^{(n+1)}\big{)}\Big{)}/\hat{v}_{K}^{(n+1)},\end{split}$ whenever $\hat{u}_{K}^{(n+1)}\neq 0$ (else $\hat{A}_{K}^{(n+1)}=0$) or $\hat{v}_{K}^{(n+1)}\neq 0$ (else $\hat{B}_{K}^{(n+1)}=0$). Since the functions $r_{A}$ and $r_{B}$ are monotone increasing, it follows that $\hat{A}_{K}^{(n+1)}$ and $\hat{B}_{K}^{(n+1)}$ are nonnegative. We then have, by subtracting the discrete equation (11) for $u_{K}^{(n+1)}$ and for ${\tilde{u}_{K}}^{(n+1)}$, $\begin{split}m_{K}\Bigg{(}1+t_{\delta}^{(n)}&\bigg{(}\hat{\alpha}\hat{A}_{K}^{(n+1)}+\frac{a}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\Bigg{)}\hat{u}_{K}^{(n+1)}\\\ =&\ m_{K}\hat{u}_{K}^{(n)}+t_{\delta}^{(n)}a\sum_{L\in{\cal N}_{K}}T_{K\|L}\hat{u}_{L}^{(n+1)}\\\ &\ +t_{\delta}^{(n)}m_{K}\hat{\alpha}\Big{(}r_{B}\big{(}v_{K}^{(n+1)}\big{)}-r_{B}\big{(}\tilde{v}_{K}^{(n+1)}\big{)}\Big{)},\end{split}$ (14) for $K\in{\cal M}$ and $n\in\\{0,\ldots,N\\}$. Setting $s^{+}=\max(s,0)$ and using that $s\leqslant s^{+},\ (s+t)^{+}\leqslant s^{+}+t^{+}$ we obtain $\begin{split}m_{K}\Bigg{(}1+t_{\delta}^{(n)}\bigg{(}&\hat{\alpha}\hat{A}_{K}^{(n+1)}+\tfrac{a}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\Bigg{)}\hat{u}_{K}^{(n+1)}\\\ \leqslant&\ m_{K}\big{(}\hat{u}_{K}^{(n)}\big{)}^{+}+t_{\delta}^{(n)}a\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{u}_{L}^{(n+1)}\big{)}^{+}+t_{\delta}^{(n)}m_{K}\hat{\alpha}\Big{(}r_{B}\big{(}v_{K}^{(n+1)}\big{)}-r_{B}\big{(}\tilde{v}_{K}^{(n+1)}\big{)}\Big{)}^{+},\end{split}$ (15) where $K\in{\cal M}$ and $n\in\\{0,\ldots,N\\}$. Next we multiply the inequality (15) by indicator of the set where $\hat{u}_{K}^{(n+1)}$ is nonnegative. Since the right-hand-side of (15) is nonnegative as well, we obtain, acting similarly for both components, $\begin{split}m_{K}\Bigg{(}1+t_{\delta}^{(n)}\bigg{(}&\hat{\alpha}\hat{A}_{K}^{(n+1)}+\frac{a}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\Bigg{)}\big{(}\hat{u}_{K}^{(n+1)}\big{)}^{+}\\\ \leqslant&\ m_{K}\big{(}\hat{u}_{K}^{(n)}\big{)}^{+}+t_{\delta}^{(n)}a\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{u}_{L}^{(n+1)}\big{)}^{+}+t_{\delta}^{(n)}m_{K}\hat{\alpha}\Big{(}r_{B}\big{(}v_{K}^{(n+1)}\big{)}-r_{B}\big{(}\tilde{v}_{K}^{(n+1)}\big{)}\Big{)}^{+},\\\ m_{K}\Bigg{(}1+t_{\delta}^{(n)}\bigg{(}&\hat{\beta}\hat{B}_{K}^{(n+1)}+\frac{b}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\Bigg{)}\big{(}\hat{v}_{K}^{(n+1)}\big{)}^{+}\\\ \leqslant&\ m_{K}\big{(}\hat{v}_{K}^{(n)}\big{)}^{+}+t_{\delta}^{(n)}b\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{v}_{L}^{(n+1)}\big{)}^{+}+t_{\delta}^{(n)}m_{K}\hat{\beta}\Big{(}r_{A}\big{(}u_{K}^{(n+1)}\big{)}-r_{A}\big{(}\tilde{u}_{K}^{(n+1)}\big{)}\Big{)}^{+}.\end{split}$ (16) Since $\begin{split}\hat{A}_{K}^{(n+1)}(\hat{u}_{K}^{(n+1)})^{+}=&\big{(}r_{A}(u_{K}^{(n+1)})-r_{A}(\tilde{u}_{K}^{(n+1)})\big{)}^{+}\\\ \hat{B}_{K}^{(n+1)}(\hat{v}_{K}^{(n+1)})^{+}=&\big{(}r_{B}(v_{K}^{(n+1)})-r_{B}(\tilde{v}_{K}^{(n+1)})\big{)}^{+}\end{split}$ we add the first equation of (16) divided by $\hat{\alpha}$ and the second equation of (16) divided by $\hat{\beta}$, which yields $m_{K}\bigg{(}\frac{1}{\hat{\alpha}}+t_{\delta}^{(n)}\frac{a}{m_{K}\hat{\alpha}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\big{(}\hat{u}_{K}^{(n+1)}\big{)}^{+}+\ m_{K}\bigg{(}\frac{1}{\hat{\beta}}+t_{\delta}^{(n)}\frac{b}{m_{K}\hat{\beta}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\big{(}\hat{v}_{K}^{(n+1)}\big{)}^{+}\\\ \leqslant m_{K}\frac{1}{\hat{\alpha}}\big{(}\hat{u}_{K}^{(n)}\big{)}^{+}+t_{\delta}^{(n)}\frac{a}{\hat{\alpha}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{u}_{L}^{(n+1)}\big{)}^{+}+\ m_{K}\frac{1}{\hat{\beta}}\big{(}\hat{v}_{K}^{(n)}\big{)}^{+}+t_{\delta}^{(n)}\frac{b}{\hat{\beta}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{v}_{L}^{(n+1)}\big{)}^{+},$ (17) for $K\in{\cal M}$ and $n\in\\{0,\ldots,N\\}$. Let us note that $\rule{40.0pt}{0.0pt}\sum_{K\in{\cal M}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{u}_{K}^{(n+1)}\big{)}^{+}=\sum_{L\in{\cal M}}\sum_{K\in{\cal N}_{L}}T_{L\|K}\big{(}\hat{u}_{L}^{(n+1)}\big{)}^{+}\\\ =\sum_{L\in{\cal M}}\sum_{K\in{\cal N}_{L}}T_{K\|L}\big{(}\hat{u}_{L}^{(n+1)}\big{)}^{+}=\sum_{K\in{\cal M}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}\hat{u}_{L}^{(n+1)}\big{)}^{+}.\rule{40.0pt}{0.0pt}$ Summing the inequalities (17) over $K\in{\cal M}$, we get $\sum_{K\in{\cal M}}\bigg{[}m_{K}\Big{(}\frac{1}{\hat{\alpha}}\big{(}\hat{u}_{K}^{(n+1)}\big{)}^{+}+\frac{1}{\hat{\beta}}\big{(}\hat{v}_{K}^{(n+1)}\big{)}^{+}\Big{)}\bigg{]}\leqslant\sum_{K\in{\cal M}}\bigg{[}m_{K}\Big{(}\frac{1}{\hat{\alpha}}\big{(}\hat{u}_{K}^{(n)}\big{)}^{+}+\frac{1}{\hat{\beta}}\big{(}\hat{v}_{K}^{(n)}\big{)}^{+}\Big{)}\bigg{]},$ which therefore leads, by induction, to $\sum_{K\in{\cal M}}\bigg{[}m_{K}\Big{(}\frac{1}{\hat{\alpha}}\big{(}\hat{u}_{K}^{(n+1)}\big{)}^{+}+\frac{1}{\hat{\beta}}\big{(}\hat{v}_{K}^{(n+1)}\big{)}^{+}\Big{)}\bigg{]}=0,$ where $n\in\\{0,\ldots,N\\}$. It implies that $(\hat{u}_{K}^{(n+1)})^{+}=(\hat{v}_{K}^{(n+1)})^{+}=0$, which completes the proof. $\blacksquare$ ###### Corollary 3.2 (Discrete contraction in $L^{1}$ property). With the notation from Proposition 3.1, we have that $\sum_{K\in\cal M}m_{K}\bigg{(}\frac{\big{\|}u_{k}^{(n+1)}-\tilde{u}_{k}^{(n+1)}\big{\|}}{\hat{\alpha}}+\frac{\big{\|}v_{k}^{(n+1)}-\tilde{v}_{k}^{(n+1)}\big{\|}}{\hat{\beta}}\bigg{)}\leqslant\ \sum_{K\in\cal M}m_{K}\bigg{(}\frac{\big{\|}u_{k}^{(n)}-\tilde{u}_{k}^{(n)}\big{\|}}{\hat{\alpha}}+\frac{\big{\|}v_{k}^{(n)}-\tilde{v}_{k}^{(n)}\big{\|}}{\hat{\beta}}\bigg{)}$ for $n\in\\{0,\ldots,N\\}$. In other words, the discrete counterpart of the $L^{1}(\Omega)$-contraction property for solutions of (1) (see e.g. [1]) is preserved by the numerical scheme (11). Proof The proof directly follows from the proof of Proposition 3.1. Let us consider the term $\hat{u}_{K}$. We multiply the equation (14) by $\text{sgn\,}\big{(}\hat{u}_{K}^{(n+1)}\big{)}$. Then, the inequality $x\leqslant\|x\|$ yields $\begin{split}m_{K}\Bigg{(}1+t_{\delta}^{(n)}\bigg{(}&\hat{\alpha}\hat{A}_{K}^{(n+1)}+a\sum_{L\in{\cal N}_{K}}T_{K\|L}\bigg{)}\Bigg{)}\big{\|}\hat{u}_{K}^{(n+1)}\big{\|}\\\ \leqslant&\ m_{K}\big{\|}\hat{u}_{K}^{(n)}\big{\|}+t_{\delta}^{(n)}\tfrac{a}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{\|}\hat{u}_{L}^{(n+1)}\big{\|}+t_{\delta}^{(n)}m_{K}\hat{\alpha}\Big{\|}r_{B}(v_{K}^{(n+1)})-r_{B}(\tilde{v}_{K}^{(n+1)})\Big{\|}.\end{split}$ We proceed in the same way for $\hat{v}_{K}^{(n+1)}$ and remark that $\begin{split}\hat{A}_{K}^{(n+1)}\big{\|}\hat{u}_{K}^{(n+1)}\big{\|}=&\Big{\|}r_{A}(u_{K}^{(n+1)})-r_{A}(\tilde{u}_{K}^{(n+1)})\Big{\|},\\\ \hat{B}_{K}^{(n+1)}\big{\|}\hat{v}_{K}^{(n+1)}\big{\|}=&\Big{\|}r_{B}(v_{K}^{(n+1)})-r_{B}(\tilde{v}_{K}^{(n+1)})\Big{\|},\end{split}$ which enable us to obtain the counterpart of the inequalities in (17) which we sum over $K\in\cal M$, as in the proof of Proposition 3.1. This yields the result. $\blacksquare$ We now are in a position to prove a discrete $L^{\infty}$ estimate for the approximate solution. ###### Theorem 3.3. Let ${\cal D}=\Big{(}{\cal M},{\cal P},{\cal E},(t_{\delta}^{(n)})_{n\in\\{0,\ldots,N+1\\}}\Big{)}$ be an admissible discretization of $Q_{T}$ in the sense of Definition 2.3. We suppose that the hypotheses $\mathcal{H}$ are satisfied. Let $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$ be given by (10) and $(u_{K}^{(n+1)},v_{K}^{(n+1)})$ satisfy (11) for $K\in{\cal M}$ and $n\in\\{0,\ldots,N\\}$. Then $0\leqslant u_{K}^{(n+1)}\leqslant U+\frac{\alpha}{\beta}V\quad\text{and}\quad 0\leqslant v_{K}^{(n+1)}\leqslant V+\frac{\beta}{\alpha}U,\ $ (18) for all $K\in{\cal M}$ and $\ n\in\\{0,\ldots,N\\}$, where $U$ and $V$ are the positive constants from the hypothesis $\mathcal{H}\,2$. Proof From Proposition 3.1 we immediately obtain that $u_{K}^{(n+1)}$ and $v_{K}^{(n+1)}$ are nonnegative for $K\in{\cal M}$ and $\ n\in\\{0,\ldots,N\\}$. In order to find a discrete upper solution, we consider approximate solutions of the corresponding system of ordinary differential equations. More precisely, we consider sequences $(\bar{u}^{(n)})_{n\in\\{0,\ldots,N+1\\}}$, $(\bar{v}^{(n)})_{n\in\\{0,\ldots,N+1\\}}$ (we postpone for a moment the proof that they exist) such that $\bar{u}^{(0)}=U,\quad\bar{v}^{(0)}=V$ and $\begin{split}\bar{u}^{(n+1)}-\bar{u}^{(n)}&=\alpha k\,t_{\delta}^{(n)}\Big{(}r_{B}(\bar{v}^{(n+1)}\big{)}-r_{A}\big{(}\bar{u}^{(n+1)})\Big{)},\\\ \bar{v}^{(n+1)}-\bar{v}^{(n)}&=\beta k\,t_{\delta}^{(n)}\Big{(}r_{A}\big{(}\bar{u}^{(n+1)})-r_{B}(\bar{v}^{(n+1)}\big{)}\Big{)},\end{split}$ (19) for $n\in\\{0,\ldots,N\\}$. We note that the sequences $(\bar{u}^{(n+1)})_{n\in\\{0,\ldots,N\\}}$, $(\bar{v}^{(n+1)})_{n\in\\{0,\ldots,N\\}}$ satisfy (11) with the initial data $U,V$. Therefore they satisfy the comparison principle from Proposition 3.1 which yields $0\leqslant\bar{u}^{(n+1)}\quad\text{and}\quad 0\leqslant\bar{v}^{(n+1)}\quad\text{for all}\quad n\in\\{0,\ldots,N\\}.$ (20) Adding up the first equation of (19) divided by $\alpha$ and the second one divided by $\beta$, we obtain $\frac{\bar{u}^{(n+1)}}{\alpha}+\frac{\bar{v}^{(n+1)}}{\beta}=\frac{\bar{u}^{(n)}}{\alpha}+\frac{\bar{v}^{(n)}}{\beta}=\ldots=\frac{U}{\alpha}+\frac{V}{\beta}\hskip 5.0pt.$ We deduce from the previous equation and from (20) that $0\leqslant\bar{u}^{(n+1)}\leqslant U+\frac{\alpha}{\beta}V\text{\quad and\quad}0\leqslant\bar{v}^{(n+1)}\leqslant V+\frac{\beta}{\alpha}U,$ (21) for $n\in\\{0,\ldots,N\\}$. $\blacksquare$ In order to prove the existence of the sequences $(\bar{u}^{(n)})_{n\in\\{0,\ldots,N+1\\}}$ and $(\bar{v}^{(n)})_{n\in\\{0,\ldots,N+1\\}}$ we use the topological degree theory in finite dimensional spaces. The reader can find basic definitions as well as further informations about this powerful theory in Deimling [3]. An example of the application of this tool to the analysis of finite volume schemes can be found in Eymard, Gallouët, Ghilani and Herbin [5]. With $\mathcal{F},\mathcal{G}:\mathbb{R}^{2}\rightarrow\mathbb{R}^{2}$ defined as $\begin{split}\mathcal{F}\big{(}\bar{u}^{(n)},\bar{v}^{(n)}\big{)}=&\ \big{(}\bar{u}^{(n)},\bar{v}^{(n)}\big{)},\\\ \mathcal{G}\big{(}\bar{u}^{(n)},\bar{v}^{(n)}\big{)}=&\ \bigg{(}\hat{\alpha}t_{\delta}^{(n-1)}\Big{(}r_{A}\big{(}\bar{u}^{(n)})-r_{B}(\bar{v}^{(n)}\big{)}\Big{)},\\\ &-\hat{\beta}t_{\delta}^{(n-1)}\Big{(}r_{A}\big{(}\bar{u}^{(n)})-r_{B}(\bar{v}^{(n)}\big{)}\Big{)}\bigg{)},\end{split}$ we rewrite the system (19) in the form $\mathcal{F}\big{(}\bar{u}^{(n+1)},\bar{v}^{(n+1)}\big{)}+\mathcal{G}\big{(}\bar{u}^{(n+1)},\bar{v}^{(n+1)}\big{)}=y:=\big{(}\bar{u}^{(n)},\bar{v}^{(n)}\big{)}.$ Moreover we see that setting $\mathcal{O}=B(0,r)\subset\mathbb{R}^{2}$ a ball centered at $(0,0)$ with a radius $r>\sqrt{\big{(}U+\tfrac{\alpha}{\beta}V\big{)}^{2}+\big{(}V+\tfrac{\beta}{\alpha}U\big{)}^{2}}$ we fulfill all the assumptions of Theorem [3, Theorem 3.1, page 16]. For the continuous function $\mathcal{H}:[0,1]\times B(0,r)\rightarrow\mathbb{R}^{2}$ given by $\mathcal{H}(\lambda,\bar{u}^{(n+1)},\bar{v}^{(n+1)})=\mathcal{F}(\bar{u}^{(n+1)},\bar{v}^{(n+1)})+\lambda\mathcal{G}(\bar{u}^{(n+1)},\bar{v}^{(n+1)}),$ we see that $d\big{(}\mathcal{F}+\lambda\mathcal{G},B,(\bar{u}^{(n)},\bar{v}^{(n)})\big{)}=d\big{(}\mathcal{H}(\lambda),B,(\bar{u}^{(n)},\bar{v}^{(n)})\big{)}$ for all $\lambda\in[0,1]$ and $(\overline{u}^{(n)}),\,(\overline{v}^{(n)})$ such that $0\leqslant\bar{u}^{(n)}\leqslant U-\tfrac{\alpha}{\beta}V$ and $0\leqslant\bar{v}^{(n)}\leqslant V-\tfrac{\beta}{\alpha}U$ for $n\in\\{0,\ldots,N\\}$. On the other hand we deduce from [3, Theorem 3.1, page 16] (d1) that $d\big{(}\mathcal{H}(0),B,(\bar{u}^{(n)},\bar{v}^{(n)})\big{)}=1.$ (22) In view of [3, Theorem 3.1, page 16] (d3) and (d4), (22) implies that the equality $\mathcal{F}(\bar{u}^{(n+1)},\bar{v}^{(n+1)})+\mathcal{G}(\bar{u}^{(n+1)},\bar{v}^{(n+1)})=\big{(}\bar{u}^{(n)},\bar{v}^{(n)}\big{)}$ has a solution or, in other words, that there exists a solution of (19). The uniqueness of this solution immediately follows from Proposition 3.1. We can prove in the same way the existence and uniqueness of the solution of the system (11). Indeed, we rewrite (11) in the form $\widetilde{\cal F}\Big{(}(u^{(n+1)}_{K})_{K\in\cal M},(v^{(n+1)}_{K})_{K\in\cal M}\Big{)}+\widetilde{\cal G}\Big{(}(u^{(n+1)}_{K})_{K\in\cal M},(v^{(n+1)}_{K})_{K\in\cal M}\Big{)}=\ \Big{(}(u^{(n)}_{K})_{K\in\cal M},(v^{(n)}_{K})_{K\in\cal M}\Big{)},$ (23) where $\widetilde{\cal F},\widetilde{\cal G}:\mathbb{R}^{2\Theta}\rightarrow\mathbb{R}^{2\Theta}$, with $\Theta$ the number of control volumes for the discretization $\cal D$, are continuous functions given by $\begin{split}&\widetilde{\cal F}\Big{(}(u^{(n)}_{K})_{K\in\cal M},(v^{(n)}_{K})_{K\in\cal M}\Big{)}=\Big{(}(u^{(n)}_{K})_{K\in\cal M},(v^{(n)}_{K})_{K\in\cal M}\Big{)},\\\ &\widetilde{\cal G}\Big{(}(u^{(n)}_{K})_{K\in\cal M},(v^{(n)}_{K})_{K\in\cal M}\Big{)}=\big{(}\mathcal{W}_{1},\mathcal{W}_{2}\big{)},\end{split}$ where $\rule{15.0pt}{0.0pt}\mathcal{W}_{1}=-\frac{t_{\delta}^{(n)}a}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}(u_{L}^{(n)}-u_{K}^{(n)})+t_{\delta}^{(n-1)}\hat{\alpha}\big{(}r_{A}(u_{K}^{(n)})-r_{B}(v_{K}^{(n)})\big{)},\rule{15.0pt}{0.0pt}$ and where $\rule{15.0pt}{0.0pt}\mathcal{W}_{2}=-\frac{t_{\delta}^{(n)}b}{m_{K}}\sum_{L\in{\cal N}_{K}}T_{K\|L}(v_{L}^{(n)}-v_{K}^{(n)})-t_{\delta}^{(n-1)}\hat{\beta}\big{(}r_{A}(u_{K}^{(n)})-r_{B}(v_{K}^{(n)})\big{)}.\rule{15.0pt}{0.0pt}$ We set $\widetilde{\mathcal{O}}=B(0,R)\subset\mathbb{R}^{2\Theta}$ a ball centered at zero with a radius $R>\sqrt{\Theta\big{(}U+\tfrac{\alpha}{\beta}V\big{)}^{2}+\Theta\big{(}V+\tfrac{\beta}{\alpha}U\big{)}^{2}}.$ Since $\Theta>1$, we deduce from the discrete $L^{\infty}(Q_{T})$ estimate of Theorem 3.3 that the equation (23) does not have any solutions on $\partial\widetilde{\mathcal{O}}$. Applying again [3, Theorem 3.1, page 16] with $\widetilde{\cal H}(\lambda)=\widetilde{\cal F}+\lambda\widetilde{\cal G}$ and $\lambda\in[0,1]$ completes the proof of the following result. ###### Theorem 3.4. We suppose that the hypotheses $\mathcal{H}$ are satisfied. Let $\cal D$ be a discretization as in Definition 2.3. Let $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$ be given by (10). Then there exists one and only one sequence $(u_{K}^{(n+1)},v_{K}^{(n+1)})_{K\in{\cal M},\ n\in\\{0,\ldots,N\\}},$ which satisfies (11), with the initial condition $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$. $\blacksquare$ ## 4 Convergence proof with $k$ fixed We begin with the discrete version of $L^{2}(Q_{T})$ estimates of the gradient of the approximate solutions. ###### Proposition 4.1. We suppose that the hypotheses $\mathcal{H}$ are satisfied. Let $\cal D$ be a discretization as in Definition 2.3. Let (10) and (11) give the sequences $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$ and $(u_{K}^{(n+1)},v_{K}^{(n+1)})_{K\in{\cal M},\ n\in\\{0,\ldots,N\\}}$, respectively. Then, there exists a constant $C_{k}>0$, which does not depend on $\cal D$, but which depend on all the data of the continuous Problem $\mathcal{P}^{k}$ (namely, the constants $\alpha,\ \beta,\ U,\ V$ including $k$ and the functions $r_{A},\ r_{B}$), such that $\rule{20.0pt}{0.0pt}\frac{1}{2}\sum_{K\in{\cal M}}m_{K}\Big{(}u_{K}^{(N+1)}\Big{)}^{2}-\frac{1}{2}\sum_{K\in{\cal M}}m_{K}\Big{(}u_{K}^{(0)}\Big{)}^{2}+a\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{(K,L)\in{\cal E}}T_{K\|L}\Big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\Big{)}^{2}\leqslant C_{k}\rule{20.0pt}{0.0pt}$ (24) and $\rule{20.0pt}{0.0pt}\frac{1}{2}\sum_{K\in{\cal M}}m_{K}\Big{(}v_{K}^{(N+1)}\Big{)}^{2}-\frac{1}{2}\sum_{K\in{\cal M}}m_{K}\Big{(}v_{K}^{(0)}\Big{)}^{2}+b\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{(K,L)\in{\cal E}}T_{K\|L}\Big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\Big{)}^{2}\leqslant C_{k}\rule{20.0pt}{0.0pt}$ (25) Proof For the sake of simplicity we only present the proof for the $u$-component. We multiply the first equation in the finite volume scheme (11) by $u_{K}^{(n+1)}$ and sum the result over all $K\in\cal M$ and over all $n\in\\{0,\ldots,N\\}$ to obtain $\mathcal{S}_{1}+\mathcal{S}_{2}+\mathcal{S}_{3}=0,$ (26) where $\displaystyle\mathcal{S}_{1}=\sum_{n=0}^{N}\sum_{K\in{\cal M}}m_{K}\Big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\Big{)}u_{K}^{(n+1)},$ $\displaystyle\mathcal{S}_{2}=-a\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{K\in{\cal M}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\Big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\Big{)}u_{K}^{(n+1)},$ $\displaystyle\mathcal{S}_{3}=\hat{\alpha}\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{K\in{\cal M}}\Big{(}r_{A}\big{(}u_{K}^{(n+1)}\big{)}-r_{B}\big{(}v_{K}^{(n+1)}\big{)}\Big{)}u_{K}^{(n+1)}.$ Since $\Big{(}u_{K}^{(n+1)}\Big{)}^{2}-u_{K}^{(n)}u_{K}^{(n+1)}=\frac{1}{2}\Big{(}u_{K}^{(n+1)}\Big{)}^{2}-\frac{1}{2}\Big{(}u_{K}^{(n)}\Big{)}^{2}+\frac{1}{2}\Big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\Big{)}^{2},$ we deduce that $\mathcal{S}_{1}=\frac{1}{2}\sum_{n=0}^{N}\sum_{K\in{\cal M}}m_{K}\bigg{(}\Big{(}u_{K}^{(n+1)}\Big{)}^{2}-\Big{(}u_{K}^{(n)}\Big{)}^{2}\bigg{)}+\frac{1}{2}\sum_{n=0}^{N}\sum_{K\in{\cal M}}m_{K}\Big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\Big{)}^{2}\\\ \geqslant\frac{1}{2}\sum_{n=0}^{N}\sum_{K\in{\cal M}}m_{K}\bigg{(}\Big{(}u_{K}^{(n+1)}\Big{)}^{2}-\Big{(}u_{K}^{(n)}\Big{)}^{2}\bigg{)},$ (27) All the terms in the sum on $n$ on the right hand side of (27) simplify except for the first and the last ones. We have that $\mathcal{S}_{1}\geqslant\frac{1}{2}\sum_{K\in{\cal M}}m_{K}\Big{(}u_{K}^{(N+1)}\Big{)}^{2}-\frac{1}{2}\sum_{K\in{\cal M}}m_{K}\Big{(}u_{K}^{(0)}\Big{)}^{2}.$ (28) We can perform a discrete integration by parts to obtain $\mathcal{S}_{2}=a\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{(K,L)\in{\cal E}}T_{K\|L}\Big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\Big{)}^{2}.$ (29) Finally we use the hypothesis $\mathcal{H}\,4$ and the inequalities in (18) to estimate the last term, namely $-\mathcal{S}_{3}\leqslant\hat{\alpha}\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{K\in{\cal M}}m_{K}\Big{(}r_{A}\big{(}U+\frac{\alpha}{\beta}V\big{)}+r_{B}\big{(}V+\frac{\beta}{\alpha}U\big{)}\Big{)}\big{(}U+\frac{\alpha}{\beta}V\big{)}\\\ \leqslant\alpha kC\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{K\in{\cal M}}m_{K}=\alpha kC\|\Omega\|T,$ (30) with some positive constant $C$. Identities (26) and (29) together with the inequalities (28) and (30) immediately give (24). Since the argument in the case of the $v$-component is similar, we omit the proof. $\blacksquare$ ### 4.1 Space and time translates of approximate solutions We now turn to the space translates estimates. We use here methods which have been presented for example by Eymard, Gutnic and Hilhorst [8] and by Eymard, Gallouët, Hilhorst and Slimane [7]. The results of the current and the next subsection together with the technical Proposition 7.2 will imply the relative compactness of the sequence of approximate solutions. ###### Proposition 4.2 (Space translates estimates). We assume that 1. 1. ${\cal D}=\Big{(}{\cal M},{\cal E},(x_{K})_{K\in{\cal T}},(t^{(n)})_{n\in\\{0,\ldots,N+1\\}}\Big{)}$ is an admissible discretization of $Q_{T}$ in the sense of Definition 2.3, 2. 2. the hypotheses $\mathcal{H}$ and assumptions (36) are satisfied, 3. 3. the functions $(u_{\cal D}$ and $v_{\cal D})$ are derived from the scheme (10) – (11) and given by the formulas (12). Then there exists a positive constant $C_{k}$, which does not depend on $\cal D$, but depends on all the data of the continuous Problem $\mathcal{P}^{k}$, including $k$, such that $\rule{15.0pt}{0.0pt}\int_{0}^{T}\int_{\Omega_{\xi}}\big{(}u_{\cal D}(x+\xi,t)-u_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{k}\|\xi\|\big{(}2\,\hbox{\rm size}\,({\cal D}\big{)}+\|\xi\|),\rule{15.0pt}{0.0pt}$ (31) and $\rule{15.0pt}{0.0pt}\int_{0}^{T}\int_{\Omega_{\xi}}\big{(}v_{\cal D}(x+\xi,t)-v_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{k}\|\xi\|\big{(}2\,\hbox{\rm size}\,({\cal D})\big{)}+\|\xi\|),\rule{15.0pt}{0.0pt}$ (32) for all $\xi\in\mathbb{R}^{d}$ and for $\Omega_{\xi}$ defined as in Proposition 7.2. Proof Inequalities (31) and (32) follow from the estimates (24) and (25), respectively. We refer to [6, Lemma 3.3] for a details. $\blacksquare$ ###### Proposition 4.3 (Time translates estimates). Let the assumptions of Proposition 4.2 are satisfied. Then, there exists some constant $C_{k}>0$, which does not depend on $\cal D$, but which depend on all the data including $k$, such that $\rule{20.0pt}{0.0pt}\int_{\Omega\times(0,T-\tau)}\big{(}u_{\cal D}(x,t+\tau)-u_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{k}\big{(}\hbox{\rm size}\,({\cal D})+\tau\big{)}\rule{20.0pt}{0.0pt}$ (33) and $\rule{20.0pt}{0.0pt}\int_{\Omega\times(0,T-\tau)}\big{(}v_{\cal D}(x,t+\tau)-v_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{k}\big{(}\hbox{\rm size}\,({\cal D})+\tau\big{)},\rule{20.0pt}{0.0pt}$ (34) for all $\tau\in(0,T)$. Proof In order to apply Lemma 7.1 (see appendix), we follow the same steps as in [8, Lemma 5.5]. The only difference appear in the nonlinear part of the equations. However, these can be easily estimated using the regularity properties of functions $r_{A}(\cdot)$ and $r_{B}(\cdot)$, as well as $L^{\infty}$ estimates (18) in Theorem 3.3. $\blacksquare$ ### 4.2 Convergence proof In this section, we state convergence results with $k$ fixed. This differs from next section where we will introduce additional hypotheses about the nonlinear reaction terms and obtain convergence results which permit us to pass to the limit as $k\rightarrow\infty$. ###### Theorem 4.4. We suppose that the hypotheses $\mathcal{H}$ are satisfied. Let $(u_{{\cal D}},v_{{\cal D}})$ be the approximate solution defined by (10), (11) and (12). There exist a pair of functions $(u^{k},v^{k})$ and a sequence $(u_{{\cal D}_{m}},v_{{\cal D}_{m}})_{m\in\mathbb{N}}$ of $(u_{{\cal D}},v_{{\cal D}})$ such that $(u_{{\cal D}_{m}},v_{{\cal D}_{m}})_{m\in\mathbb{N}}\mbox{~{}~{}converges to~{}~{}}(u^{k},v^{k})\in\big{(}L^{2}\big{(}0,T;H^{1}(\Omega)\big{)}\big{)}^{2}$ strongly in $L^{2}\big{(}Q_{T}\big{)}$ as $\hbox{\rm size}\,({\cal D}_{m})$ tends to zero. The function pair $(u^{k},v^{k})$ is a weak solution of Problem $\mathcal{P}^{k}$ in the sense of Definition 1.1. Since Problem $\mathcal{P}^{k}$ is a uniformly parabolic system, $(u^{k},v^{k})$ must coincide with the unique classical solution of Problem $\mathcal{P}^{k}$ . This immediately yields the following result. ###### Corollary 4.5. The pair $(u_{{\cal D}},v_{{\cal D}})$ converges to the unique classical solution $(u^{k},v^{k})$ of Problem $\mathcal{P}^{k}$ as $\hbox{\rm size}\,({\cal D})$ tends to zero. $\blacksquare$ Proof of Theorem 4.4 In view of the estimates (31), (33) and Proposition 7.2 which is a consequence of the Fréchet–Kolmogorov Theorem [2, Theorem IV.25, page 72], we deduce the relative compactness of the set $\big{(}u_{{\cal D}}\big{)}$ so that there exists a sequence of $\big{(}u_{{\cal D}_{m}}\big{)}_{m=1}^{\infty}$ and a function ${\cal U}_{k}$, such that $u_{{\cal D}_{m}}\rightarrow{\cal U}_{k}$ strongly in $L^{2}(Q_{T})$ and weakly in $L^{2}\big{(}0,T;H^{1}(\Omega)\big{)}$, as $m\rightarrow\infty$. The same conclusion holds for the $v$-component. Indeed, the inequalities (32) and (34) permit to apply the compactness result in Proposition 7.2 for the sequence $\big{(}v_{{\cal D}_{m}}\big{)}_{m=1}^{\infty}$. There exists a function ${\cal V}_{k}$ such that $v_{{\cal D}_{m}}\rightarrow{\cal V}_{k}$ strongly in $L^{2}(Q_{T})$ and weakly in $L^{2}\big{(}0,T;H^{1}(\Omega)\big{)}$ as $m\rightarrow\infty$. Next we show that $(\mathcal{U}_{k}$, $\mathcal{V}_{k})$ is a weak solution of Problem $\mathcal{P}^{k}$, in the sense of Definition 1.1. Since the proof for the $v$-component is similar, we only present here the detailed proof in case of the $u$-component. Let $\psi\in\Psi$, where $\Psi$ is the class of test functions from Definition 1.1. We multiply the first equation of (11) by $\psi(x_{K},t^{(n)})$, where $\psi\in\Psi$. Then we sum over all $K\in\cal M$ and $n\in\\{0\ldots N-1\\}$ to obtain ${\cal T}_{1m}^{u}-{\cal T}_{2m}^{u}+{\cal T}_{3m}^{u}=0,$ where $\displaystyle{\cal T}_{1m}^{u}$ $\displaystyle=\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}m_{K}\big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\big{)}\psi(x_{K},t^{(n)}),$ $\displaystyle{\cal T}_{2m}^{u}$ $\displaystyle=a\sum_{n=0}^{N-1}t_{\delta}^{(n)}\sum_{K\in{\cal M}}\sum_{L\in{\cal N}_{K}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}\psi(x_{K},t^{(n)}),$ $\displaystyle{\cal T}_{3m}^{u}$ $\displaystyle=\sum_{n=0}^{N-1}t_{\delta}^{(n)}\sum_{K\in{\cal M}}m_{K}\hat{\alpha}\Big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\Big{)}\psi(x_{K},t^{(n)}).$ The complete proof, that $\lim_{m\rightarrow\infty}{\cal T}_{1m}^{u}=-\int_{\Omega}u_{0}(x)\psi(x,0)\,\text{d}x-\int_{0}^{T}\int_{\Omega}{\cal U}_{k}(x,t)\psi_{t}(x,t)\,\text{d}x\,\text{d}t$ and $\lim_{m\rightarrow\infty}{\cal T}_{2m}^{u}=-a\int_{0}^{T}\int_{\Omega}{\cal U}_{k}(x,t)\Delta\psi(x,t)\,\text{d}x\,\text{d}t,$ can by found in [8, Lemma 5.5]. Let us focus on the proof that $\lim_{m\rightarrow\infty}{\cal T}_{3m}^{u}=\alpha k\int_{0}^{T}\int_{\Omega}\big{(}r_{A}({\cal U}_{k})-r_{B}({\cal V}_{k})\big{)}\psi(x,t)\,\text{d}x\,\text{d}t.$ We write $\displaystyle\hat{\alpha}\sum_{n=0}^{N-1}t_{\delta}^{(n)}$ $\displaystyle\sum_{K\in{\cal M}}m_{K}\Big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\Big{)}\psi(x_{K},t^{(n)})$ $\displaystyle-\hat{\alpha}\int_{0}^{T}\int_{\Omega}\psi(x,t)\big{(}r_{A}({\cal U}_{k})-r_{B}({\cal V}_{k})\big{)}\,\text{d}x\,\text{d}t$ $\displaystyle=$ $\displaystyle\,\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\Big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\Big{)}\psi(x_{K},t^{(n)})\,\text{d}x\,\text{d}t$ $\displaystyle-\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\psi(x,t)\big{(}r_{A}({\cal U}_{k})-r_{B}({\cal V}_{k})\big{)}\,\text{d}x\,\text{d}t$ $\displaystyle-\hat{\alpha}\sum_{K\in{\cal M}}\int_{t^{(N)}}^{t^{(N+1)}}\int_{K}\psi(x,t)\big{(}r_{A}({\cal U}_{k})-r_{B}({\cal V}_{k})\big{)}\,\text{d}x\,\text{d}t.$ Thanks to the regularity of the function $\psi$, the last sum above converges to zero . Moreover, $\begin{split}\hat{\alpha}\sum_{n=0}^{N-1}&\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\Big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\Big{)}\psi(x_{K},t^{(n)})\,\text{d}x\,\text{d}t\\\ &-\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\psi(x,t)\big{(}r_{A}(\mathcal{U}_{k})-r_{B}(\mathcal{V}_{k})\big{)}\,\text{d}x\,\text{d}t\\\ =&\,\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\big{(}\psi(x_{K},t^{(n)})-\psi(x,t)\big{)}\times\\\ &\rule{120.0pt}{0.0pt}\times(r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\,\text{d}x\,\text{d}t\\\ &+\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\psi(x,t)\big{(}r_{A}(u_{K}^{(n+1)})-r_{A}({\cal U}_{k})\big{)}\,\text{d}x\,\text{d}t\\\ &-\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\psi(x,t)\big{(}r_{B}(v_{K}^{(n+1)})-r_{B}({\cal V}_{k})\big{)}\,\text{d}x\,\text{d}t.\end{split}$ (35) Next we show that the three terms above tend to zero as $m\rightarrow\infty$. First we take their absolute value and apply the triangle inequality. The Cauchy-Schwarz inequality applied to the first sum of the right hand side of (35) yields $\displaystyle\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\big{\|}\psi(x_{K},t^{(n)})-\psi(x,t)\big{\|}\big{\|}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)}\big{\|}\,\text{d}x\,\text{d}t$ $\displaystyle\leqslant\bigg{(}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\big{(}\psi(x_{K},t^{(n)})-\psi(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\bigg{)}^{1/2}\times$ $\displaystyle\times\hat{\alpha}\bigg{(}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)}\big{)}^{2}\,\text{d}x\,\text{d}t\bigg{)}^{1/2}.$ The first term of above product converges to zero, as $m\rightarrow\infty$, since $\psi(x,t)$ is smooth enough. The second term is bounded. Indeed, it is sufficient to remark that $r_{A}(u_{K}^{(n+1)})$, $r_{B}(v_{K}^{(n+1)})$ are bounded for all $K\in\mathcal{M}$ and $n\in\\{0,\ldots,N\\}$. The last two terms in (35) are similar and we show how the proof goes with the first one. Indeed, $\displaystyle\hat{\alpha}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\|\psi(x,t)\|\big{\|}r_{A}(u_{K}^{(n+1)})-r_{A}({\cal U}_{k})\big{\|}\,\text{d}x\,\text{d}t$ $\displaystyle\leqslant\,\hat{\alpha}\\|\psi\\|_{L^{\infty}(Q_{T})}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\big{\|}r_{A}(u_{K}^{(n+1)})-r_{A}({\cal U}_{k})\big{\|}\,\text{d}x\,\text{d}t$ $\displaystyle\leqslant\,\hat{\alpha}\\|\psi\\|_{L^{\infty}(Q_{T})}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\\|r_{A}^{\prime}\\|_{L^{\infty}(Q_{T})}\big{\|}u(x_{K},t^{(n)})-{\mathcal{U}_{k}}(x,t)\big{\|}\,\text{d}x\,\text{d}t$ $\displaystyle\leqslant\,\hat{\alpha}\\|\psi\\|_{L^{\infty}(Q_{T})}\bigg{(}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\\|r_{A}^{\prime}\\|_{L^{\infty}(Q_{T})}^{2}\,\text{d}x\,\text{d}t\bigg{)}^{1/2}\ \times$ $\displaystyle\times\bigg{(}\sum_{n=0}^{N-1}\sum_{K\in{\cal M}}\int_{t^{(n)}}^{t^{(n+1)}}\int_{K}\big{\|}u(x_{K},t^{(n)})-{\mathcal{U}_{k}}(x,t)\big{\|}^{2}\,\text{d}x\,\text{d}t\bigg{)}^{1/2}.$ The $L^{\infty}$ norm of the function $r_{A}(x)$ is taken over the finite interval $[0,U+\frac{\alpha}{\beta}V]$. Because $r_{A}(x)$ is of class $C^{1}(\mathcal{R})$ the first term of the above product is bounded. The second one converges to zero since $u_{{\cal D}_{m}}\rightarrow{\cal U}_{k}$ as $m\rightarrow\infty$ in $L^{2}(Q_{T})$. $\blacksquare$ ## 5 The case that $k$ tends to infinity In order to prove the convergence of the finite volume scheme when $size~{}(\cal D)$ tends to zero and $k$ tends to infinity, we impose some additional conditions on the nonlinear terms $r_{A}(x)$ and $r_{B}(x)$. At first we prove a counterpart of Proposition 4.1. ###### Proposition 5.1. Let us assume hypotheses $\mathcal{H}$. Moreover, we assume that the functions $r_{A}(x),\ r_{B}(x)$ satisfy $\begin{split}&r_{\kappa}\in C^{1}(\mathbb{R}),\ r_{\kappa}^{\prime}(\cdot)>0\text{\quad on \quad}(0,+\infty),\\\ &r_{\kappa}(0)=0,\text{\quad and \quad}\limsup_{s\rightarrow 0^{+}}\frac{sr_{\kappa}^{\prime}(s)}{r_{\kappa}(s)}<\infty,\end{split}$ (36) where $\kappa\in\\{A,\ B\\}$. Let ${\cal D}=\Big{(}{\cal M},{\cal E},(x_{K})_{K\in{\cal T}},(t^{(n)})_{n\in\\{0,\ldots,N+1\\}}\Big{)}$ be an admissible discretization of $Q_{T}$ in the sense of Definition 2.3, and the sequences $(u_{K}^{(0)},v_{K}^{(0)})_{K\in{\cal M}}$ and $\big{(}u_{K}^{(n+1)},v_{K}^{(n+1)}\big{)}_{K\in{\cal M},\ n\in\\{0,\ldots,N\\}}$ are given by (10) and (11), respectively. Then, there exists some positive constant $C$ which is independent of the discretization $\cal D$ and of the reaction rate $k$, such that $\sum_{n=0}^{N}t_{\delta}^{(n)}\bigg{(}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}+\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\big{)}^{2}\bigg{)}\leqslant C$ (37) and $k\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{K\in{\cal M}}m_{K}\big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\big{)}^{2}\leqslant C.$ (38) ###### Remark 5.2. Observe that the condition (36) holds, for example, in the case that the rate functions $r_{A}(s)$ and $r_{B}(s)$ behave like $s^{\gamma}$, for some positive $\gamma$, whenever $s\rightarrow 0^{+}$. Proof of Proposition 5.1 Let $(a,b)\in(\mathbb{R}^{+})^{2}$ be such that $r_{A}(a)=r_{B}(b)$. We define two functions $\begin{split}V_{A}(s)=\frac{1}{\alpha}\Bigg{(}s\ln\frac{r_{A}(s)}{r_{A}(a)}+\int_{s}^{a}\frac{\sigma r_{A}^{\prime}(\sigma)}{r_{A}(\sigma)}\,\text{d}\sigma\Bigg{)},\\\ V_{B}(s)=\frac{1}{\beta}\Bigg{(}s\ln\frac{r_{B}(s)}{r_{B}(b)}+\int_{s}^{b}\frac{\sigma r_{B}^{\prime}(\sigma)}{r_{B}(\sigma)}\,\text{d}\sigma\Bigg{)},\end{split}$ which are continuous on $\mathbb{R}^{+}$ because of hypotheses $\mathcal{H}$ and the assumptions (36). We can extend these functions to also be continuous at $s=0$. To do so for the function $V_{A}(s)$ we check that the hypotheses (36) give the integrability of $\displaystyle\frac{sr_{A}^{\prime}(s)}{r_{A}(s)}$ on the interval $[0,a]$ and we pass to the limit $\lim_{s\rightarrow 0^{+}}V_{A}(s)\stackrel{{\scriptstyle H}}{{=}}\frac{1}{\alpha}\int_{0}^{a}\frac{\sigma r_{A}^{\prime}(\sigma)}{r_{A}(\sigma)}\,\text{d}\sigma-\frac{1}{\alpha}\lim_{s\rightarrow 0^{+}}s\cdot\frac{s\,r_{A}^{\prime}(s)}{r_{A}(s)}<\infty,$ where we have applied de l’Hospital theorem as formulated in [13, Theorem 2, p. 174]. For a given $\varepsilon\in(0,1)$ and $n\in\\{0,\ldots,N\\}$ we consider $A^{(n+1)}(\varepsilon)=\sum_{K\in{\cal M}}m_{K}\Big{(}V_{A}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}-V_{A}\big{(}u_{K}^{(n)}+\varepsilon\big{)}\Big{)}.$ Since $\displaystyle V_{A}^{\prime\prime}(s)=\frac{r_{A}^{\prime}(s)}{\alpha\,r_{A}(s)}>0$ for all $s\in\mathbb{R}^{+}$ we deduce that $\displaystyle V_{A}\big{(}u_{K}^{(n)}+\varepsilon\big{)}-V_{A}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}=\big{(}u_{K}^{(n)}-u_{K}^{(n+1)}\big{)}V_{A}^{\prime}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}+\frac{1}{2}\big{(}u_{K}^{(n)}-u_{K}^{(n+1)}\big{)}^{2}V_{A}^{\prime\prime}\big{(}s\big{)}$ $\displaystyle\geqslant\big{(}u_{K}^{(n)}-u_{K}^{(n+1)}\big{)}V_{A}^{\prime}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}\hskip 5.0pt.$ As a consequence $A^{(n+1)}(\varepsilon)\leqslant\sum_{K\in{\cal M}}m_{K}\big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\big{)}V_{A}^{\prime}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}.$ We substitute $m_{K}\big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\big{)}$ from the scheme (11), which yields $A^{(n+1)}(\varepsilon)\leqslant A_{1}^{(n+1)}(\varepsilon)+A_{2}^{(n+1)}(\varepsilon),$ with $A_{1}^{(n+1)}(\varepsilon)=-t_{\delta}^{(n)}\frac{a}{4}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}\Big{\\{}V_{A}^{\prime}\big{(}u_{L}^{(n+1)}+\varepsilon\big{)}-V_{A}^{\prime}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}\Big{\\}}$ and $A_{2}^{(n+1)}(\varepsilon)=-t_{\delta}^{(n)}\sum_{K\in{\cal M}}m_{K}\hat{\alpha}\,V_{A}^{\prime}\big{(}u_{K}^{(n+1)}+\varepsilon\big{)}\Big{\\{}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\Big{\\}}.$ Since there exists some constant $C>0$ such that $V_{A}^{\prime\prime}(s)\geqslant C$ for all $s\in[0,U+\frac{\alpha}{\beta}V]$ we can use the $L^{\infty}$ bound (18) and the mean value theorem to obtain $A_{1}^{(n+1)}(\varepsilon)\leqslant- t_{\delta}^{(n)}C\frac{a}{4}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}.$ Following the same steps for the function $B^{(n+1)}(\varepsilon)=\sum_{K\in{\cal M}}m_{K}\Big{(}V_{B}\big{(}v_{K}^{(n+1)}+\varepsilon\big{)}-V_{B}\big{(}v_{K}^{(n)}+\varepsilon\big{)}\Big{)},$ we arrive at $A^{(n+1)}(\varepsilon)+B^{(n+1)}(\varepsilon)\leqslant-\ C^{(n+1)}-D^{(n+1)}(\varepsilon),$ (39) with $C^{(n+1)}=t_{\delta}^{(n)}C\frac{a}{4}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}+t_{\delta}^{(n)}C\frac{b}{4}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\big{)}^{2},$ and $D^{(n+1)}(\varepsilon)=t_{\delta}^{(n)}k\sum_{K\in{\cal M}}m_{K}\big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\big{)}\big{(}\ln r_{A}(u_{K}^{(n+1)}+\varepsilon)-\ln r_{B}(v_{K}^{(n+1)}+\varepsilon)\big{)},$ where we used that $\displaystyle\alpha V_{A}^{\prime}(s)=\ln\frac{r_{A}(s)}{r_{A}(a)}$ and that $\displaystyle\beta V_{B}^{\prime}(s)=\ln\frac{r_{B}(s)}{r_{B}(b)}$. Let $D_{1}^{(n+1)}(\varepsilon)=t_{\delta}^{(n)}k\sum_{K\in{\cal M}}m_{K}\big{(}r_{A}(u_{K}^{(n+1)}+\varepsilon)-r_{B}(v_{K}^{(n+1)}+\varepsilon)\big{)}\big{(}\ln r_{A}(u_{K}^{(n+1)}+\varepsilon)-\ln r_{B}(v_{K}^{(n+1)}+\varepsilon)\big{)}.$ Since the inequality $(c-d)(\ln c-\ln d)\geqslant\frac{(c-d)^{2}}{c+d}$ holds for all $\displaystyle(c,d)\in\Big{(}0,U+\frac{\alpha}{\beta}V+1\Big{)}\times\Big{(}0,V+\frac{\beta}{\alpha}U+1\Big{)}$, then $D_{1}^{(n+1)}(\varepsilon)\geqslant t_{\delta}^{(n)}\frac{k}{C_{b}}\\!\sum_{K\in{\cal M}}\\!m_{K}\big{(}r_{A}(u_{K}^{(n+1)}+\varepsilon)-r_{B}(v_{K}^{(n+1)}+\varepsilon)\big{)}^{2},$ where $C_{b}$ is an upper bound for $r_{A}(c)+r_{B}(d)$ with $\displaystyle(c,d)\in\big{(}0,U+\frac{\alpha}{\beta}V+1\big{)}\times\big{(}0,V+\frac{\beta}{\alpha}U+1\big{)}$. Such bounds exist in view of Theorem 3.3 and regularity of the functions $r_{A}(x)$ and $r_{B}(x)$. Let us define $E^{(n+1)}(\varepsilon):=t_{\delta}^{(n)}\frac{k}{C_{b}}\\!\sum_{K\in{\cal M}}\\!m_{K}\big{(}r_{A}(u_{K}^{(n+1)}+\varepsilon)-r_{B}(v_{K}^{(n+1)}+\varepsilon)\big{)}^{2}.$ The assumptions (36) and Lemma 7.3 in the appendix, imply that for all $K\in\cal M$ and $\varepsilon>0$ small enough, there exist constants $C_{1},\ C_{2}$ such that $\big{\|}r_{A}(u_{K}^{(n+1)}+\varepsilon)-r_{A}(u_{K}^{(n+1)})\big{\|}\leqslant C_{1}\varepsilon,\quad\big{\|}r_{B}(v_{K}^{(n+1)}+\varepsilon)-r_{B}(v_{K}^{(n+1)})\big{\|}\leqslant C_{1}\varepsilon,$ and $\big{\|}\ln r_{A}(u_{K}^{(n+1)}+\varepsilon)\big{\|}\leqslant C_{2}\big{(}\|\ln\varepsilon\|+1\big{)},\quad\big{\|}\ln r_{B}(v_{K}^{(n+1)}+\varepsilon)\big{\|}\leqslant C_{2}\big{(}\|\ln\varepsilon\|+1\big{)}.$ Then $-D^{(n+1)}\leqslant C\varepsilon\|\Omega\|k\big{(}\|\ln\varepsilon\|+1\big{)}-D^{(n+1)}(\varepsilon),$ for some positive constant $C$. As a consequence $\displaystyle A^{(n+1)}(\varepsilon)+B^{(n+1)}(\varepsilon)$ $\displaystyle\leqslant-\,C^{(n+1)}-D^{(n+1)}(\varepsilon)$ $\displaystyle\leqslant-\,C^{(n+1)}-E^{(n+1)}(\varepsilon)+C\varepsilon\|\Omega\|k\big{(}\|\ln\varepsilon\|+1\big{)}.$ Now it is possible to pass to the limit in (39). We obtain $A^{(n+1)}(0)+B^{(n+1)}(0)\leqslant-C^{(n+1)}-E^{(n+1)}(0).$ which is $\displaystyle\sum_{K\in{\cal M}}m_{K}\Big{(}V_{A}\big{(}u_{K}^{(n+1)}\big{)}-V_{A}\big{(}u_{K}^{(n)}\big{)}\Big{)}+\sum_{K\in{\cal M}}m_{K}\Big{(}V_{B}\big{(}v_{K}^{(n+1)}\big{)}-V_{B}\big{(}v_{K}^{(n)}\big{)}\Big{)}$ $\displaystyle+t_{\delta}^{(n)}C\frac{a}{4}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}+t_{\delta}^{(n)}C\frac{b}{4}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\big{)}^{2}$ $\displaystyle+t_{\delta}^{(n)}\frac{k}{C_{b}}\\!\sum_{K\in{\cal M}}\\!m_{K}\big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\big{)}^{2}\leqslant 0.$ Now we sum the above inequalities over $n\in\\{0,\ldots,N\\}$ to obtain $\displaystyle\sum_{K\in{\cal M}}m_{K}\Big{(}V_{A}\big{(}u_{K}^{(N+1)}\big{)}+V_{B}\big{(}u_{K}^{(N+1)}\big{)}\Big{)}++C\frac{a}{4}\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}$ $\displaystyle+C\frac{b}{4}\sum_{n=0}^{N}t_{\delta}^{(n)}\sum_{(K,L)\in{\cal E}}T_{K\|L}\big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\big{)}^{2}+\frac{k}{C_{b}}\sum_{n=0}^{N}t_{\delta}^{(n)}\\!\sum_{K\in{\cal M}}\\!m_{K}\big{(}r_{A}(u_{K}^{(n+1)})-r_{B}(v_{K}^{(n+1)})\big{)}^{2}$ $\displaystyle\leqslant\sum_{K\in{\cal M}}m_{K}\Big{(}V_{A}\big{(}u_{K}^{(0)}\big{)}+V_{B}\big{(}u_{K}^{(0)}\big{)}\Big{)}.$ Since $V_{A}(s)$ and $V_{B}(s)$ are nonnegative and continuous on $[0,+\infty)$, and since $u_{K}^{(n+1)},\ v_{K}^{(n+1)}$ are nonnegative and bounded for all $K\in\mathcal{M}$ and $n\in\\{0,\ldots,N\\}$, this concludes the proof. $\blacksquare$ ### 5.1 Space and time translates of the approximate solutions Since we have already presented the general methods in section 4 we only give here some essential ideas, leaving out the details of the proofs. We begin with a counterpart of Proposition 4.2. ###### Proposition 5.3 (Space translates estimates). Let us assume that 1. 1. $\Big{(}{\cal M},{\cal E},(x_{K})_{K\in{\cal T}},(t^{(n)})_{n\in\\{0,\ldots,N+1\\}}\Big{)}$ is an admissible discretization of $Q_{T}$ in the sense of Definition 2.3, 2. 2. hypotheses $\mathcal{H}$ and assumptions (36) are satisfied, 3. 3. functions $(u_{\cal D}$ and $v_{\cal D})$ are derived from the scheme (10) – (11) and given by the formulas (12). Then there exists a positive constant $C$ which is independent of $\cal D$ and $k$, such that $\rule{15.0pt}{0.0pt}\int_{0}^{T}\int_{\Omega_{\xi}}\big{(}u_{\cal D}(x+\xi,t)-u_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C\|\xi\|\big{(}2\,\hbox{\rm size}\,({\cal D}\big{)}+\|\xi\|),\rule{15.0pt}{0.0pt}$ (40) and $\rule{15.0pt}{0.0pt}\int_{0}^{T}\int_{\Omega_{\xi}}\big{(}v_{\cal D}(x+\xi,t)-v_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C\|\xi\|\big{(}2\,\hbox{\rm size}\,({\cal D}\big{)}+\|\xi\|),\rule{15.0pt}{0.0pt}$ (41) for all $\xi\in\mathbb{R}^{d}$ and for $\Omega_{\xi}$ defined as in Proposition 7.2. Proof As it was in the proof of Proposition 4.2, we refer to [6][Lemma 3.3] for a complete proof. The only difference is to apply the result of Proposition 5.1 instead of that of Proposition 4.1. $\blacksquare$ Let us now prove an analogue of Proposition 4.3 ###### Lemma 5.4 (Time translates estimate). Let the assumptions 1. 2. and 3. of Lemma 5.3 be satisfied. Set $\displaystyle w_{\cal D}=\frac{1}{\alpha}u_{\cal D}+\frac{1}{\beta}v_{\cal D}$. Then there exists a positive constant $C$ which is independent of $\cal D$ and $k$, such that $\int_{\Omega\times(0,T-\tau)}\big{(}w_{\cal D}(x,t+\tau)-w_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C\big{(}\hbox{\rm size}\,({\cal D})+\tau\big{)},$ for all $\tau\in(0,T)$. Proof The proof is similar to that of Proposition 4.3. We present below the essential steps of the argument. We define $\mathcal{A}(t):=\int_{\Omega}\big{(}w_{\cal D}(x,t+\tau)-w_{\cal D}(x,t)\big{)}^{2}\,\text{d}x,$ which can be easily transformed into $\displaystyle\mathcal{A}(t)$ $\displaystyle=\sum_{K\in\mathcal{M}}m_{K}\big{(}w_{K}^{(n(t+\tau)+1)}-w_{K}^{(n(t)+1)}\big{)}^{2}$ $\displaystyle=\sum_{k\in\mathcal{M}}\bigg{(}\big{(}w_{K}^{(n(t+\tau)+1)}-w_{K}^{(n(t)+1)}\big{)}\sum_{n=n(t)+1}^{n(t+\tau)}m_{K}\big{(}w_{K}^{(n+1)}-w_{K}^{(n)}\big{)}\bigg{)}.$ Since $w^{(n+1)}_{K}-w^{(n)}_{K}=\frac{1}{\alpha}\Big{(}u_{K}^{(n+1)}-u_{K}^{(n)}\Big{)}+\frac{1}{\beta}\Big{(}v_{K}^{(n+1)}-v_{K}^{(n)}\Big{)},$ we can apply discrete integration by parts in the scheme (11) to obtain $\displaystyle\mathcal{A}(t)=\frac{a}{\alpha}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}\big{(}w_{K}^{(n(t+\tau)+1)}-w_{L}^{(n(t+\tau)+1)}\big{)}$ $\displaystyle+\frac{a}{\alpha}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}\big{(}w_{L}^{(n(t)+1)}-w_{K}^{(n(t)+1)}\big{)}$ $\displaystyle\frac{b}{\beta}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\big{)}\big{(}w_{K}^{(n(t+\tau)+1)}-w_{L}^{(n(t+\tau)+1)}\big{)}$ $\displaystyle+\frac{b}{\beta}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}v_{L}^{(n+1)}-v_{K}^{(n+1)}\big{)}\big{(}w_{L}^{(n(t)+1)}-w_{K}^{(n(t)+1)}\big{)}.$ Next we estimate the second term in the sum above, to obtain $\displaystyle\frac{a}{\alpha}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}\sqrt{T_{K\|L}}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}\cdot\sqrt{T_{K\|L}}\big{(}w_{L}^{(n(t)+1)}-w_{K}^{(n(t)+1)}\big{)}$ $\displaystyle\leqslant\frac{a}{2\alpha}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}$ $\displaystyle+\frac{a}{2\alpha}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}w_{L}^{(n(t)+1)}-w_{K}^{(n(t)+1)}\big{)}^{2}$ $\displaystyle\leqslant\frac{a}{2\alpha}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}u_{L}^{(n+1)}-u_{K}^{(n+1)}\big{)}^{2}$ $\displaystyle+\frac{a}{\alpha^{3}}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}u_{L}^{(n(t)+1)}-u_{K}^{(n(t)+1)}\big{)}^{2}$ $\displaystyle+\frac{a}{\alpha\beta^{2}}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\sum_{(K,L)\in\cal E}T_{K\|L}\big{(}v_{L}^{(n(t)+1)}-v_{K}^{(n(t)+1)}\big{)}^{2},$ where the first inequality follows from the relation $\displaystyle\pm s_{1}s_{2}\leqslant\frac{1}{2}(s_{1}^{2}+s_{2}^{2})$ and the second one follows from the simple inequality $(s_{1}+s_{2})^{2}\leqslant 2(s_{1}^{2}+s_{2}^{2})$. To conclude the proof we integrate above inequalities over $\mathbb{R}$ with respect to the time variable $t$. Next we apply Proposition 7.1 (for details see [8, Lemma 5.5]). $\blacksquare$ Proposition 5.3 together with Proposition 5.4 immediately give the following corollary. ###### Corollary 5.5. Let the assumptions 1. 2. and 3. of Proposition 5.3 be satisfied. We set $\displaystyle w_{\cal D}=\frac{1}{\alpha}u_{\cal D}+\frac{1}{\beta}v_{\cal D}$. Then, there exists a constant $C>0$, which is independent of the discretization parameters ${\cal D}$ and of $k$, such that $\int_{\Omega_{\xi}\times(0,T)}\big{(}w_{\cal D}(x+\xi,t)-w_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C\|\xi\|\big{(}2\hbox{\rm size}\,({\cal D}\big{)}+\|\xi\|),$ (42) for all $\xi\in\mathbb{R}^{d}$ and $\Omega_{\xi}=\\{x\in\mathbb{R}^{d},[x,x+\xi]\subset\Omega\\}$. Moreover $\int_{\Omega\times(0,T-\tau)}\big{(}w_{\cal D}(x,t+\tau)-w_{\cal D}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C\big{(}\hbox{\rm size}\,({\cal D})+\tau\big{)},$ (43) where $\tau\in(0,T)$. $\blacksquare$ ### 5.2 The limit as $size(\cal D)$ tends to zero and $k$ tends to infinity We state below the main convergence results of this paper, first only letting the size of the volume elements and the time steps tend to zero, and then also letting the kinetic rate tend to infinity. ###### Proposition 5.6. We suppose that the hypotheses $\mathcal{H}$ are satisfied. Let $k>0$ be arbitrary and let $(u^{k},\,v^{k})\in\Big{(}L^{2}\big{(}0,T;\,H^{1}(\Omega)\big{)}\Big{)}^{2}$ be the unique classical solution of Problem $\mathcal{P}^{k}$. Then the sequence $(u^{k}_{{\cal D}},\,v^{k}_{{\cal D}})$ of approximate solutions of Problem $\mathcal{P}^{k}$ given by (10), (11) and (12) converges strongly in $\Big{(}L^{2}\big{(}Q_{T}\big{)}\Big{)}^{2}$ to $(u^{k},\,v^{k})$ as $\hbox{\rm size}\,({\cal D})\rightarrow 0$. In particular, the sequence $\displaystyle w^{k}_{{\cal D}}=\frac{1}{\alpha}u^{k}_{{\cal D}}+\frac{1}{\beta}v^{k}_{{\cal D}}$ converges strongly in $L^{2}\big{(}Q_{T}\big{)}$ to the function $\displaystyle w^{k}=\frac{1}{\alpha}u^{k}+\frac{1}{\beta}v^{k}$ as $\hbox{\rm size}\,({\cal D})\rightarrow 0$ and there exist positive constants $C_{1},\,C_{2}$ which do not depend on $k$, such that $\displaystyle k\int_{\Omega\times(0,T)}\big{(}r_{A}(u^{k})-r_{B}(v^{k})\big{)}^{2}\,\text{d}x\leqslant C_{1},$ (44) $\displaystyle\int_{\Omega_{\xi}\times(0,T)}\big{(}w^{k}(x+\xi,t)-w^{k}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{2}\|\xi\|^{2},$ (45) $\displaystyle\int_{\Omega\times(0,T-\tau)}\big{(}w^{k}(x,t+\tau)-w^{k}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{2}\tau,$ (46) where $\tau\in(0,T),\ \xi\in\mathbb{R}^{d}$ and $\Omega_{\xi}=\\{x\in\mathbb{R}^{d},[x,x+\xi]\subset\Omega\\}$. Proof To prove the result we use Corollary 5.5. The method of proof is similar to that of Theorem 4.4. $\blacksquare$ It is now possible to pass to the limit as $k\rightarrow+\infty$. ###### Theorem 5.7. Let $\big{(}u_{\mathcal{D}}^{k},\,v_{\mathcal{D}}^{k}\big{)}$ be the sequence of approximate solutions of Problem $\mathcal{P}^{k}$, defined by (10), (11) and (12). Then $u^{k}_{\mathcal{D}}\longrightarrow\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\\!\\!\\!(w)$ and $v^{k}_{\mathcal{D}}\longrightarrow\eta\circ\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\\!\\!\\!(w)$ as $\hbox{\rm size}\,(\mathcal{D})\rightarrow 0$ and $k$ tends to $\infty$, where $\eta=r_{B}^{-1}\circ r_{A}$ and where $w$ is the unique weak solution of the problem (5) – (6). Proof Let $\displaystyle w^{k}_{{\cal D}}=\frac{1}{\alpha}u^{k}_{{\cal D}}+\frac{1}{\beta}v^{k}_{{\cal D}}$. The estimates from Corollary 5.5, which are uniform with respect to $k$, permit to apply Proposition 7.2. As a consequence we deduce the relative compactness in $L^{2}(Q_{T})$ of the sequence $\\{w^{k}_{{\cal D}}\\}$. Then there exist a function $w\in L^{2}(Q_{T})$ and a subsequence $\\{w^{k_{i}}_{{\cal D}_{m}}\\}$ such that $w^{k_{i}}_{{\cal D}_{m}}$ converges to $w$ strongly in $L^{2}(Q_{T})$ as $k_{i}$ tends to infinity and $size({\cal D}_{m})$ tends to zero. Theorem 3.3 implies that $w$ is nonnegative and bounded in $Q_{T}$. The inequality (38), namely ${k_{i}}\big{\\|}r_{A}(u^{k_{i}}_{{\cal D}_{m}})-r_{B}(v^{k_{i}}_{{\cal D}_{m}})\big{\\|}^{2}_{L^{2}(Q_{T})}\leqslant C$ where the positive constant $C$ is independent of $k_{i}$ and $size({\cal D}_{m})$, implies that $r_{A}(u^{k_{i}}_{{\cal D}_{m}})-r_{B}(v^{k_{i}}_{{\cal D}_{m}})\rightarrow 0\quad\text{in}\quad L^{2}(Q_{T}),$ and consequently almost everywhere, as $k_{i}$ tends to infinity. Then $v^{k_{i}}_{{\cal D}_{m}}=\eta(u^{k_{i}}_{{\cal D}_{m}})+e^{k_{i}}_{{\cal D}_{m}},$ where $\eta(s)=r^{-1}_{B}\big{(}r_{A}(s)\big{)}$ and $e^{k_{i}}_{{\cal D}_{m}}$ tends to zero almost everywhere as $size({\cal D}_{m})$ tends to zero and $k_{i}$ tends to infinity. In view of the hypotheses $\mathcal{H}\,4$ the function $\eta(s)$ is well defined on $[0,\infty)$. Moreover, $H(u^{k_{i}}_{{\cal D}_{m}})=w^{k_{i}}_{{\cal D}_{m}}-\frac{1}{\beta}e^{k_{i}}_{{\cal D}_{m}}\rightarrow w\quad\text{a.e.\ in}\quad Q_{T},$ where $\displaystyle H(s)=\frac{1}{\alpha}s+\frac{1}{\beta}\eta(s)$. Hypotheses $\mathcal{H}\,4$ ensures that the function $H(s)$ has the continuous inverse function. Then Lebesgue’s dominated convergence theorem implies that there exists a function pair $(\tilde{u},\,\tilde{v})\in\big{(}L^{2}(Q_{T})\big{)}^{2}$ $u^{k_{i}}_{{\cal D}_{m}}\rightarrow\tilde{u}\quad\text{and}\quad v^{k_{i}}_{{\cal D}_{m}}\rightarrow\tilde{v}\quad\text{in}\quad L^{2}(Q_{T})$ (47) as $size({\cal D}_{m})$ tends to zero and $k_{i}$ tends to infinity. Next we identify the limit pair $(\tilde{u},\,\tilde{v})$. Let $\overline{w}$ be the solution of the problem (5) – (6) and the functions $\overline{u}$ and $\overline{v}$ are defined as in (7), namely $\overline{u}=\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\hskip-4.0pt\big{(}\overline{w}\big{)},\quad\overline{v}=\eta\circ\bigg{(}\frac{1}{\alpha}\text{id}+\frac{1}{\beta}\eta\bigg{)}^{-1}\hskip-4.0pt\big{(}\overline{w}\big{)}.$ Let $\varepsilon>0$ be arbitrary. We have that $\rule{20.0pt}{0.0pt}\big{\\|}\overline{u}-\tilde{u}\big{\\|}_{L^{2}(Q_{T})}\leqslant\big{\\|}\overline{u}-u^{k_{i}}\big{\\|}_{L^{2}(Q_{T})}+\big{\\|}u^{k_{i}}-u_{\mathcal{D}_{m}}^{k_{i}}\big{\\|}_{L^{2}(Q_{T})}+\big{\\|}u_{\mathcal{D}_{m}}^{k_{i}}-\tilde{u}\big{\\|}_{L^{2}(Q_{T})}\rule{5.0pt}{0.0pt}.$ From (47) we deduce that there exists $k_{0}$ and $\delta_{0}$ such that, for all $k_{i}\geq k_{0}$ and all $size({\cal D}_{m})\leq\delta_{0}$, the last term of the inequality above is less then $\varepsilon/3$. From Theorem 1 in [1] there exists some ${{\bar{k}}_{0}}$, such that for all ${k_{i}}>{\bar{k}}_{0}$, $\big{\\|}\overline{u}-u^{k_{i}}\big{\\|}_{L^{2}(Q_{T})}\leqslant\varepsilon/3$. Then, fixing $k_{i}=\max(k_{0},{\bar{k}}_{0})$, we can take $size({\cal D}_{m})\leq\delta_{0}$ small enough so that by Proposition 5.6 $\big{\\|}u^{k_{i}}-u_{\mathcal{D}_{m}}^{k_{i}}\big{\\|}_{L^{2}(Q_{T})}\leqslant\varepsilon/3$. Since the argument for the $v$-component is similar, this completes the proof. $\blacksquare$ ## 6 Numerical example In this section we give an example of an application of the finite volume scheme (11) in one space dimension. For the numerical experiments we choose the reaction of the reversible dimerisation of $o$-phenylenedioxydimethylsilane ($2,2$-dimethyl-$1,2,3$-benzodioxasilole) which has been studied by 1H NMR spectroscopy. The kinetics of this reaction can be described quantitatively by a bimolecular lO-ring formation reaction and a monomolecular backreaction (for further details we refer to Meyer, Klein and Weiss [12]). Since the reaction is of the type $2\mathcal{A}\rightleftharpoons\mathcal{B}$, the reaction terms take the form $r_{A}(u)=k_{1}u^{2}\quad\text{and}\quad r_{B}(v)=k_{2}v.$ Moreover $\alpha=2$ and $\beta=1$. For this particular process benzene was chosen as a solvent. Then it was possible to estimate rate constants for both reactions at the temperature $T=298K$, $k_{1}\approx 1,072\cdot 10^{-4}\text{L}^{2}\text{mol}^{-2}\quad\text{and}\quad k_{2}\approx 2,363\cdot 10^{-6}\text{L}^{2}\text{mol}^{-2}$ and diffusion coefficients $a\approx 1,579\cdot 10^{-9}\text{m}^{2}\text{s}^{-1}\quad\text{and}\quad b\approx 1,042\cdot 10^{-9}\text{m}^{2}\text{s}^{-1}.$ In the experiment we set $k=1$ for the chemical kinetics factor. We remark that it is equivalent to the situation when coefficients $a,\ b,\ k_{1}$ and $k_{2}$ are of order $1$ and $k$ is of order $10^{4}$. In fact, we can multiply the system (1) by $10^{9}$ and change the time scale as $t\mapsto 10^{9}t$. However the above reasoning is formally correct and shows in an explicit way the order of the kinetics factor $k$; in our example we decided to keep constants in the form given by the spectroscopic analysis. Figure 1 shows the initial conditions $u_{0}(x)$ and $v_{0}(x)$, defined as $u_{0}(x)=\begin{cases}0&\text{ for }x\in[0,0.03]\\\ \frac{1}{2}\sin\Big{(}\frac{50\pi}{7}(x-0.03)\Big{)}&\text{ for }x\in[0.03,0.1],\end{cases}$ (48) and $v_{0}(x)=\begin{cases}\frac{1}{4}\cos\Big{(}\frac{50\pi}{7}x\Big{)}&\text{ for }x\in[0,0.07],\\\ 0&\text{ for }x\in[0.07,0.1]\end{cases}$ (49) On Figures 3 and 5 we see the time evolution of the solution $\big{(}u_{{\cal D}}^{k},v_{{\cal D}}^{k}\big{)}$ until the times $T_{\text{max}}^{1}=10^{5}\ [s]$ and $T_{\text{max}}^{2}=10^{11}\ [s]$, respectively. Then we follow the evolution of the solution $w_{{\cal D}}$ of the nonlinear diffusion problem (5) – (6) for initial condition deduced from that used in the reaction – diffusion Problem $\mathcal{P}^{k}$. We have used a uniform mesh with $h=0,002$ and initial time step $t_{\delta}=10^{-8}$ to obtain the approximate solution $\big{(}u_{{\cal D}}^{k},v_{{\cal D}}^{k}\big{)}$ and $t_{\delta}=10^{-6}$ to obtain the approximate solution $w_{{\cal D}}$to the nonlinear diffusion problem. We can use the approximate solution $w_{{\cal D}}$ and the formulas (7) to define functions $u_{{\cal D}}^{w}$ and $v_{{\cal D}}^{w}$. Indeed, let $u_{{\cal D}}^{w}=h\big{(}w_{{\cal D}}\big{)}\quad\text{and}\quad v_{{\cal D}}^{w}=g\big{(}h\big{(}w_{{\cal D}}\big{)}\big{)},$ where $h(y)=\frac{1}{2}\bigg{(}\sqrt{\Big{(}\frac{\alpha k_{1}}{\beta k_{2}}\Big{)}^{2}+y\,\frac{4k_{2}}{\beta k_{1}}}-\frac{\alpha k_{2}}{\beta k_{1}}\bigg{)}$ (50) and $g(h)=h\,\frac{a}{\alpha}+h^{2}\,\frac{bk_{1}}{\beta k_{2}}\hskip 3.0pt.$ (51) Proceeding in the similar way as in the proof of Theorem 5.7, we write for the $u$-component that $\big{\\|}u_{{\cal D}}^{k}-h(w_{{\cal D}})\big{\\|}_{L^{2}(Q_{T})}\leqslant\big{\\|}u_{{\cal D}}^{k}-u^{k}\big{\\|}_{L^{2}(Q_{T})}\\\ +\big{\\|}u^{k}-h(w)\big{\\|}_{L^{2}(Q_{T})}+\big{\\|}h(w)-h(w_{{\cal D}})\big{\\|}_{L^{2}(Q_{T})}.$ (52) We simultaneously pass to the limit as $\hbox{\rm size}\,({\cal D})\rightarrow 0$ and $k\rightarrow\infty$. From Theorem 5.7 we immediately deduce that the first term on the right hand side of (52) tends to zero. The same conclusion holds for the two other terms. Indeed, [1, Theorem 1, Sec. 3] implies that $\big{\\|}u^{k}-h(w)\big{\\|}_{L^{2}(Q_{T})}$ tends to zero as $k\rightarrow\infty$. Moreover, [7, Theorem 5.1] yields that $\big{\\|}w-w_{{\cal D}}\big{\\|}_{L^{2}(Q_{T})}\rightarrow 0$ as $\hbox{\rm size}\,({\cal D})\rightarrow 0$ and since the function $h$ is well defined and continuous, we conclude that for every small $\varepsilon>0$ there exist ${\cal D}$ small enough and $k$ large enough so that $\big{\\|}u_{{\cal D}}^{k}-h(w_{{\cal D}})\big{\\|}_{L^{2}(Q_{T})}\leqslant\varepsilon.$ We proceed in the same way to show that for every small $\varepsilon>0$ there exist ${\cal D}$ small enough and $k$ large enough so that $\big{\\|}v_{{\cal D}}^{k}-g(h(w_{{\cal D}}))\big{\\|}_{L^{2}(Q_{T})}\leqslant\varepsilon.$ The results from our numerical experiment agree with above analysis, since $\displaystyle\max_{x\in\Omega}\Big{\|}u_{{\cal D}}^{k}(x,T_{\text{max}}^{1})-h(w_{{\cal D}})(x,T_{\text{max}}^{1})\Big{\|}_{\infty}\simeq 4,74\cdot 10^{-3},$ $\displaystyle\max_{x\in\Omega}\Big{\|}v_{{\cal D}}^{k}(x,T_{\text{max}}^{1})-g(h(w_{{\cal D}})(x,T_{\text{max}}^{1}))\Big{\|}_{\infty}\simeq 4,032\cdot 10^{-3},$ whereas $\displaystyle\max_{x\in\Omega}\Big{\|}u_{{\cal D}}^{k}(x,T_{\text{max}}^{2})-h(w_{{\cal D}})(x,T_{\text{max}}^{2})\Big{\|}_{\infty}\simeq 3,121\cdot 10^{-14},$ $\displaystyle\max_{x\in\Omega}\Big{\|}v_{{\cal D}}^{k}(x,T_{\text{max}}^{2})-g(h(w_{{\cal D}})(x,T_{\text{max}}^{2}))\Big{\|}_{\infty}\simeq 1,84\cdot 10^{-13}.$ Figure 1: Initial conditions $u_{0}$ and $v_{0}$, defined in (48) and (49). Figure 2: Initial condition $w_{0}=\frac{1}{\alpha}u_{0}+\frac{1}{\beta}v_{0}$, where $u_{0},\ v_{0}$ are defined in (48) and (49). Figure 3: Evolution of $u_{\text{ap}}$ and $v_{\text{ap}}$ on the time interval $(0,\ldots,T_{\text{max}}^{1})$. Figure 4: Evolution of $w_{\text{ap}}$ on the time interval $(0,\ldots,T_{\text{max}}^{1})$. Figure 5: Evolution of $u_{\text{ap}}$ and $v_{\text{ap}}$ on the time interval $(0,\ldots,T_{\text{max}}^{2})$. Figure 6: Evolution of $w_{\text{ap}}$ on the time interval $(0,\ldots,T_{\text{max}}^{2})$. In order to show the accuracy of our method in the fast reaction limit, let us increase the kinetics parameter $k$ in Problem $\mathcal{P}^{k}$ keeping all other data as previously. Let $\mathcal{J}_{u}^{k}=\max_{x\in\Omega}\Big{\|}u_{{\cal D}}^{k}(x,T_{\text{max}}^{2})-h(w_{{\cal D}})(x,T_{\text{max}}^{2})\Big{\|}_{\infty}$ and $\mathcal{J}_{v}^{k}=\max_{x\in\Omega}\Big{\|}v_{{\cal D}}^{k}(x,T_{\text{max}}^{2})-g(h(w_{{\cal D}})(x,T_{\text{max}}^{2}))\Big{\|}_{\infty}.$ Table 1 below shows the numerical results. $k$ \| $\mathcal{J}_{u}^{k}$ \| $\mathcal{J}_{v}^{k}$ ---\|---\|--- $10^{-7}$ \| $2,3498$$\cdot$ \| $10^{-2}$ \| $4,5272$. \| $10^{-2}$ $10^{-6}$ \| $3,976$$\cdot$ \| $10^{-3}$ \| $1,988$. \| $10^{-3}$ $10^{-5}$ \| $4,699$$\cdot$ \| $10^{-8}$ \| $2,3498$. \| $10^{-8}$ $10^{-4}$ \| $9,3725$$\cdot$ \| $10^{-10}$ \| $4,6862$. \| $10^{-10}$ $10^{-3}$ \| $2,498$$\cdot$ \| $10^{-16}$ \| $9,992$. \| $10^{-16}$ $10^{-2}$ \| $1,7892$$\cdot$ \| $10^{-12}$ \| $8,95$. \| $10^{-13}$ $10^{-1}$ \| $4,163$$\cdot$ \| $10^{-17}$ \| $9,992$. \| $10^{-16}$ $10^{0}$ \| $3,121$$\cdot$ \| $10^{-17}$ \| $1,84$. \| $10^{-16}$ Table 1: The accuracy of our method in the fast reaction limit, when the kinetics parameter $k$ in Problem $\mathcal{P}^{k}$ increases and all other data are unchanged. ## 7 Appendix The proof of the following result can be found in [8, Lemma 5.3 and Lemma 5.4]. ###### Lemma 7.1. Let $(t^{(n)})_{n\in{\mathbb{Z}}}$ be a strictly increasing sequence of real numbers such that $\lim_{n\rightarrow-\infty}t^{(n)}=-\infty$ and $\lim_{n\rightarrow\infty}t^{(n)}=\infty$. Moreover, let $t_{\delta}^{(n)}:=t^{(n+1)}-t^{(n)}$ be uniformly bounded. For all $t\in\mathbb{R}$ we denote by $n(t)$ an integer $n$, such that $t\in[t^{(n)},t^{(n+1)})$. Let $(a^{(n)})_{n\in{\mathbb{Z}}}$ be a family of nonnegative real values such that $a^{(n)}\neq 0$ for finitely many $n\in\mathbb{Z}$. Then, for all $\tau\in(0,+\infty)$ and $\zeta\in\mathbb{R}$ $\displaystyle\int_{\mathbb{R}}\sum_{n=n(t)+1}^{n(t+\tau)}\big{(}t_{\delta}^{(n)}a^{(n+1)}\big{)}\,\text{d}t=\tau\sum_{n\in{\mathbb{Z}}}\big{(}t_{\delta}^{(n)}a^{(n+1)}\big{)},$ (53) $\displaystyle\int_{\mathbb{R}}\bigg{(}\sum_{n=n(t)+1}^{n(t+\tau)}t_{\delta}^{(n)}\bigg{)}a^{(n(t+\zeta)+1)}\,\text{d}t\leqslant\big{(}\tau+\max_{n\in{\mathbb{Z}}}t_{\delta}^{(n)}\big{)}\sum_{n\in{\mathbb{Z}}}\big{(}t_{\delta}^{(n)}a^{(n+1)}\big{)}.$ (54) The following proposition is a direct corollary from Fréchet–Kolmogorov Theorem (see Brezis [2, Theorem IV.25, page 72]). ###### Proposition 7.2. Let $\mathcal{O}$ be a bounded and open subset of $\mathbb{R}^{d+1}$, $d=1,2$ or $3$. Let $(w_{n})_{n\in\mathbb{N}}$ be a sequence of functions $w_{n}(x,t):\ \mathbb{R}^{d}\times\mathbb{R}\rightarrow\mathbb{R}$, such that 1. 1. for all $n\in\mathbb{N}$, $w_{n}\in L^{\infty}(\mathcal{O})$ and there exists a constant $C_{b}>0$ which does not depend on $n$, such that $\\|w_{n}\\|_{L^{\infty}(\mathcal{O})}\leqslant C_{b}$, 2. 2. there exist positive constants $C_{1},\,C_{2}$ and a sequence of nonnegative real values $(\mu_{n})_{n\in\mathbb{N}}$, such that $\lim_{n\rightarrow\infty}\mu_{n}=0$ and $\int_{\mathcal{O}_{\xi,\tau}}\big{(}w_{n}(x+\xi,t+\tau)-w_{n}(x,t)\big{)}^{2}\,\text{d}x\,\text{d}t\leqslant C_{1}\|\xi\|+C_{2}\tau+\mu_{n},$ for $\xi\in\mathbb{R}^{d},\ \tau\in\mathbb{R},\ n\in\\{0,\ldots,N\\}$ and $\rule{20.0pt}{0.0pt}\mathcal{O}_{\xi,\tau}=\big{\\{}(x,t)\in\mathbb{R}^{d+1}:\text{\ the interval\ }\big{[}(x,t),(x+\xi,t+\tau)\big{]}\text{\ lies in\ }{\mathcal{O}}\big{\\}}.\rule{20.0pt}{0.0pt}$ Then there exists a subsequence of $(w_{n})_{n\in\mathbb{N}}$, denoted again by $(w_{n})_{n\in\mathbb{N}}$ and a function $w\in L^{\infty}(\mathcal{O})$ such that $w_{n}\rightarrow w$ in $L^{2}(\mathcal{O})$, as $n\rightarrow\infty$. ###### Lemma 7.3. Let $A>0$ and a function $r(s)$ satisfying $\begin{split}&r\in C^{1}(\mathbb{R}),\ r^{\prime}(\cdot)>0\text{\quad on \quad}(0,+\infty),\\\ &r(0)=0,\text{\quad and \quad}\limsup_{s\rightarrow 0^{+}}\frac{sr^{\prime}(s)}{r(s)}<\infty\end{split}$ (55) be given. Then there exists $\varepsilon_{0}>0$ only depending on $r(s)$ and $C>0$, depending only on $r(s)$ and $A$, such that for all $\varepsilon\in(0,\varepsilon_{0})$ and for all $u\in[0,A]$ the inequality $\big{\|}\ln r(u+\varepsilon)\big{\|}\leqslant C\big{(}\|\ln\varepsilon\|+1\big{)}$ holds. Proof Let $\alpha=\displaystyle\limsup_{s\rightarrow 0^{+}}\frac{sr^{\prime}(s)}{r(s)}+1$. There exists a constant $\varepsilon_{0}>0$, such that for all $s\in(0,\varepsilon_{0})$ $\frac{r^{\prime}(s)}{r(s)}\leqslant\frac{\alpha}{s}.$ (56) Let $\varepsilon\in(0,\varepsilon_{0})$ and $u\in[0,A]$. Then, either $r(u+\varepsilon)\geqslant 1$, which implies $\ln r(u+\varepsilon)\leqslant\ln r(A+\varepsilon_{0})$ or $r(u+\varepsilon)\leqslant 1$. In that case $\big{\|}\ln r(u+\varepsilon)\big{\|}\leqslant\big{\|}\ln r(\varepsilon)\big{\|},$ holds. Integrating inequality (56) over the interval $[\varepsilon,\varepsilon_{0}]$, we obtain $\big{\|}\ln r(\varepsilon)\big{\|}\leqslant\big{\|}\ln r(\varepsilon_{0})\big{\|}+\alpha\big{(}\|\ln\varepsilon_{0}\|+\|\ln\varepsilon\|\big{)}.$ Setting $C=\max\big{\\{}\alpha,\,\alpha\|\ln\varepsilon_{0}\|+\|\ln r(\varepsilon_{0})\|,\,\ln\big{(}r(A+\varepsilon_{0})\big{)}\big{\\}}$ we conclude the proof. $\blacksquare$ ## References * [1] D. Bothe and D. Hilhorst, A reaction-diffusion system with fast reversible reaction, J. Math. Anal. Appl., 2003, 1, 268, 125–135. * [2] H. Brezis, Analyse fonctionnelle, Collection Math ematiques Appliquées pour la Maîtrise, Masson, Paris, 1983\. * [3] K. Deimling, Nonlinear functional analysis, Springer-Verlag, Berlin, 1985\. * [4] J.H. Espenson, Chemical Kinetics and Reaction Mechanisms, Mc Graw-Hill, 1995\. * [5] R. Eymard, T. Gallouët, M. Ghilani, and R. Herbin, Error estimate for the finite volume approximate of the solution to a nonlinear convective equation, Theory Appl. Transp. Porous Media, 1998, 11, 13–24. * [6] R. Eymard, T. Gallou et, and R. Herbin, Finite Volume Methods, Handbook of Numerical Analysis, VII, North-Holland, Amsterdam, 2000\. * [7] R. Eymard, T. Gallou et, D. Hilhorst, and Y. Naït Slimane Finite volumes and nonlinear diffusion equations, RAIRO Mod el. Math. Anal. Num er., 1998, 32, 6, 747 761. * [8] R. Eymard, M. Gutnic, and D. Hilhorst, The finite volume method for Richards equation, Comput. Geosci., 2000, 3, 3-4, 259 294. * [9] P. Érdi and J. Tóth, Mathematical models of chemical reactions, Nonlinear Science: Theory and Applications, Princeton University Press, Princeton, NJ, 1989\. * [10] G. B. Folland, Real analysis Pure and Applied Mathematics John Wiley & Sons Inc., New York Modern techniques and their applications, 1984\. * [11] O. A. Ladyženskaja and V. A. Solonnikov and N. N. Ural’ceva, Linear and Quasilinear Equations of Parabolic Type, Translations of Mathematical Monographs, Vol. 23, American Mathematical Society, Providence, R.I., 1967\. * [12] H. Meyer, J. Klein, and A. Weiss, Kinetiche untersuchung reversiblen dimerisierung von o-phenylenedioxydimethylsilan, J. Organometallic Chem., 1979, 117, 323 328. * [13] S. M. Nikolsky, A course of mathematical analysis, Vol. 1, Mir Publishers, Moscow, Translated from the second Russian edition by V. M. Volosov, 1977	arxiv-papers	2008-08-02T11:27:14	2024-09-04T02:48:57.145051	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "R. Eymard, D. Hilhorst and M. Olech", "submitter": "Michal Olech Mr", "url": "https://arxiv.org/abs/0808.0251" }
0808.0275	# Trivial extensions defined by Prüfer conditions C. Bakkari Department of Mathematics, Faculty of Sciences and Technology, P. O. Box 2202, University S. M. Ben Abdellah, Fez 30000, Morocco cbakkari@hotmail.com , S. Kabbaj Department of Mathematics, King Fahd University of Petroleum & Minerals, P. O. Box 5046, Dhahran 31261, KSA kabbaj@kfupm.edu.sa and N. Mahdou Department of Mathematics, Faculty of Sciences and Technology, P. O. Box 2202, University S. M. Ben Abdellah, Fez 30000, Morocco mahdou@hotmail.com ###### Abstract. This paper deals with well-known extensions of the Prüfer domain concept to arbitrary commutative rings. We investigate the transfer of these notions in trivial ring extensions (also called idealizations) of commutative rings by modules and then generate original families of rings with zerodivisors subject to various Prüfer conditions. The new examples give further evidence for the validity of Bazzoni-Glaz conjecture on the weak dimension of Gaussian rings. Moreover, trivial ring extensions allow us to widen the scope of validity of Kaplansky-Tsang conjecture on the content ideal of Gaussian polynomials. ###### Key words and phrases: Prüfer domain, semihereditary ring, arithmetical ring, Gaussian ring, Prüfer ring, weak global dimension, Gaussian polynomial, content ideal, trivial ring extension ###### 2000 Mathematics Subject Classification: 13F05, 13B05, 13A15, 13D05, 13B25 The second author was supported by KFUPM under Research Project # MS/RING/368. The third author would like to thank KFUPM for its hospitality ## 1\. Introduction All rings considered in this paper are commutative with identity elements and all modules are unital. In 1932, Prüfer introduced and studied integral domains in which every non-zero finitely generated ideal is invertible [24]. In 1936, Krull [20] named these rings after H. Prüfer and stated equivalent conditions that make a domain Prüfer. Since then, “Prüfer domains have assumed a central role in the development of multiplicative ideal theory through numeral equivalent forms. These touched on many areas of commutative algebra, e.g., valuation theory, arithmetic relations on the set of ideals, $$-operations, and polynomial rings; in addition to several homological characterizations” (Gilmer [9]). The extension of this concept to rings with zerodivisors gave rise to five classes of Prüfer-like rings featuring some homological apsects (Bazzoni-Glaz [2] and Glaz [13]); namely: (1) Semihereditary ring, i.e., every finitely generated ideal is projective (Cartan-Eilenberg [5]); (2) Ring with weak dimension at most one (Glaz [10, 12]); (3) Arithmetical ring, i.e., every finitely generated ideal is locally principal (Fuchs [6]); (4) Gaussian ring, i.e., $c(fg)=c(f)c(g)$ for any polynomials $f,g$ with coefficients in the ring, where $c(f)$ denotes the content of $f$ (Tsang [26]); (5) Prüfer ring, i.e., every finitely generated regular ideal is invertible or, equivalently, projective (Butts-Smith [4] and Griffin [15]). While, in the domain context, all these forms coincide with the definition of a Prüfer domain, Glaz [13] provided examples which show that all these notions are distinct in the context of arbitrary rings. The following diagram of implications summarizes the relations between them [2, 3, 12, 13, 21, 22, 26]: Semihereditary $\Rightarrow$ weak dimension $\leq 1$ $\Rightarrow$ Arithmetical $\Rightarrow$ Gaussian $\Rightarrow$ Prüfer It is notable that original examples -for each one of the above classes- are rare in the literature. This paper investigates the transfer of the above- mentioned Prüfer conditions to trivial ring extensions. Our results generate new examples which enrich the current literature with new families of Prüfer- like rings with zerodivisors. Particularly, we obtain further evidence for the validity of Bazzoni-Glaz conjecture sustaining that “the weak dimension of a Gaussian ring is 0, 1, or $\infty$” [3]. Moreover, trivial ring extensions offer the possibility to widen the scope of validity of the content conjecture of Kaplansky and Tsang which was extended to investigate rings where “every Gaussian polynomial has locally principal content ideal” [1, 2, 14, 16, 21, 22, 26]. Notice that both conjectures share the common context of rings with zerodivisors. This very fact lies behind our motivation for studying the Gaussian condition and related concepts in trivial ring extensions. Let $A$ be a ring and $E$ an $A$-module. The trivial ring extension of $A$ by $E$ (also called the idealization of $E$ over $A$) is the ring $R:=A\propto~{}E$ whose underlying group is $A\times E$ with multiplication given by $(a,e)(a^{\prime},e^{\prime})=(aa^{\prime},ae^{\prime}+a^{\prime}e)$. For the reader’s convenience, recall that if $I$ is an ideal of $A$ and $E^{\prime}$ is a submodule of $E$ such that $IE\subseteq E^{\prime}$, then $J:=I\propto E^{\prime}$ is an ideal of $R$; ideals of $R$ need not be of this form [19, Example 2.5]. However, prime (resp., maximal) ideals of $R$ have the form $p\propto E$, where $p$ is a prime (resp., maximal) ideal of $A$ [17, Theorem 25.1(3)]. Suitable background on commutative trivial ring extensions is [10, 17]. Section 2 deals with trivial ring extensions of the form $R:=A\propto~{}B$, where $A\subseteq B$ is an extension of integral domains. Precisely, we examine the possible transfer of Prüfer ring conditions to $R$. The main result asserts that “$R$ is Gaussian (resp., Arithmetical) if and only if $A$ is Prüfer with $K\subseteq B$ (resp., $K=B$).” This generates new examples of non-arithmetical Gaussian rings as well as arithmetical rings with weak dimension strictly greater than one. Recall that classical examples of non- semihereditary arithmetical rings stem from Jensen’s 1966 result [18] as non- reduced principal rings, e.g., $\mathbb{Z}/n^{2}\mathbb{Z}$ for any integer $n\geq 2$. In this line, we provide a new family of examples of non-finite conductor arithmetical rings, hence quite far from being principal. We also establish a result on the weak dimension of these constructions which happens to corroborate Bazzoni-Glaz conjecture (cited above). In their recent paper devoted to Gaussian properties, Bazzoni and Glaz have proved that a Prüfer ring satisfies any of the other four Prüfer conditions if and only if its total ring of quotients satisfies that same condition [3, Theorems 3.3 & 3.6 & 3.7 & 3.12]. This fact narrows the scope of study to the class of total rings of quotients. Section 3 investigates Prüfer conditions in a special class of total rings of quotients; namely, those arising as trivial ring extensions of local rings by vector spaces over the residue fields. The main result establishes that “if $(A,M)$ is a non-trivial local ring and $E$ a nonzero $\frac{A}{M}$-vector space, then $R:=A\propto~{}E$ is a non- arithmetical total ring of quotients. Moreover, $R$ is a Gaussian ring if and only if so is $A$.” This enables us to build new examples of non-arithmetical Gaussian total rings of quotients or non-Gaussian total rings of quotients (which are necessarily Prüfer). Further the weak dimension of these constructions turns out to be infinite which subjects, here too, the new examples of Gaussian rings to Bazzoni-Glaz conjecture. A problem initially associated with Kaplansky and his student Tsang [1, 2, 14, 22, 26] and also termed as Tsang-Glaz-Vasconcelos conjecture in [16] sustained that “every nonzero Gaussian polynomial over a domain has an invertible (or, equivalently, locally principal) content ideal.” It is well-known that a polynomial over any ring is Gaussian if its content ideal is locally principal. The converse is precisely the object of Kaplansky-Tsang-Glaz- Vasconcelos conjecture extended to those rings where “every Gaussian polynomial has locally principal content ideal. The objective of Section 4 is to validate this conjecture in a large family of rings distinct from the three classes of arithmetical rings, of locally domains, and of locally approximately Gorenstein rings, where the conjecture holds so far. This was made possible by the main result which states that a trivial ring extension of a domain by its quotient field satisfies the condition that “every Gaussian polynomial has locally principal content ideal.” We end up with a conjecture that equates the latter condition with the local irreducibility of the zero ideal. This would offer an optimal solution to the Kaplansky-Tsang-Glaz- Vasconcelos conjecture that recovers all previous results. The section closes with a discussion -backed with examples- which attempts to rationalize this statement. ## 2\. Extensions of domains This section explores trivial ring extensions of the form $R:=A\propto~{}B$, where $A\subseteq B$ is an extension of integral domains. Notice in this context that $(a,b)\in R$ is regular if and only if $a\not=0$. The main result (Theorem 2.1) examines the transfer of Prüfer conditions to $R$ and hence generates new examples of non-arithmetical Gaussian rings and of arithmetical rings with weak dimension $\gvertneqq 1$. In 1969, Osofsky proved that the weak dimension of an arithmetical ring is either $\leq 1$ or infinite [23]. In 2005, Glaz proved Osofsky’s result in the class of coherent Gaussian rings [12, Theorem 3.3]. Recently, Bazzoni and Glaz conjectured that “the weak dimension of a Gaussian ring is 0, 1, or $\infty$” [3]. Theorem 2.1 validates this conjecture for the class of all Gaussian rings emanating from these constructions. Moreover, Example 2.7 widens its scope of validity beyond coherent Gaussian rings. ###### Theorem 2.1. Let $A\subseteq B$ be an extension of domains and $K:=\operatorname{qf}(A)$. Let $R:=A\propto~{}B$ be the trivial ring extension of $A$ by $B$. Then: (1) $R$ is Gaussian if and only if $R$ is Prüfer if and only if $A$ is Prüfer with $K\subseteq B$. (2) $R$ is arithmetical if and only if $A$ is Prüfer with $K=B$. (3) $\operatorname{w.dim}.dim(R)=\infty$. The proof of the theorem involves the following lemmas of independent interest. ###### Lemma 2.2. Let $A$ be a ring, $E$ an nonzero $A$-module, and $R:=A\propto~{}E$. If $R$ is Gaussian (resp., arithmetical), then so is $A$. ###### Proof. Straightforward since the arithmetical and Gaussian properties are stable under factor rings (here $A\cong\frac{R}{0\propto E}$). ∎ Notice that Lemma 2.2 does not hold for the Prüfer property as shown by Example 2.8. ###### Lemma 2.3. Let $K$ be a field, $E$ a nonzero $K$-vector space, and $R:=K\propto~{}E$. Then $\operatorname{w.dim}.dim(R)=\infty$. ###### Proof. Let $\\{f_{i}\\}_{i\in I}$ be a basis of the $K$-vector space $E$ and let $J:=0\propto E$. Consider the $R$-map $R^{(I)}\buildrel u\over{\rightarrow}J$ defined by $u((a_{i},e_{i})_{i\in I})=(0,\sum_{i\in I}a_{i}f_{i})$. Clearly, $\operatorname{Ker}(u)=0\propto E^{(I)}$. Here we are identifying $R^{(I)}$ with $A^{(I)}\propto E^{(I)}$ as $R$-modules. We have the exact sequence of $R$-modules: $0\rightarrow 0\propto E^{(I)}\rightarrow R^{(I)}\buildrel u\over{\rightarrow}J\rightarrow 0.$ We claim that $J$ is not flat. Otherwise, by [25, Theorem 3.55], we obtain $0\propto E^{(I)}=J^{(I)}=JR^{(I)}=(0\propto E^{(I)})\cap JR^{(I)}=(0\propto E^{(I)})J=0,$ a contradiction. Therefore the above exact sequence yields $\operatorname{fd}(J)=\operatorname{fd}(J^{(L)})\leq\operatorname{fd}(J)-1.$ This forces the flat dimension of $J$ and hence the weak dimension of $R$ to be infinite. ∎ ###### Proof of Theorem 2.1. (1) We need only prove the following implications: $R$ Prüfer $\Rightarrow$ $A$ Prüfer with $K\subseteq B$ $\Rightarrow$ $R$ Gaussian Assume $R$ is a Prüfer ring. We wish to show first that $K\subseteq B$ in the case when $A$ is local. Let $x\not=0\in A$ and let $I:=((x,0),(x,1))R$, a finitely generated regular ideal of $R$. Then $I$ is invertible and hence principal (since $R$ is local too). Write $I=(a,b)R$ for some $a\in A$ and $b\in B$. Clearly, $a=ux$ for some invertible element $u$ in $A$, hence $I=(ux,b)R=(x,u^{-1}b)R$. Further $(x,0)\in I$ yields $u^{-1}b=b^{\prime}x$ for some $b^{\prime}\in B$. It follows that $I=(x,b^{\prime}x)R=(x,0)(1,b^{\prime})R=(x,0)R$ since $(1,b^{\prime})$ is invertible. But $(x,1)\in I$ yields $1=xb^{\prime\prime}$ for some $b^{\prime\prime}\in B$. Therefore $K\subseteq B$. Next suppose $A$ is not necessarily local and let $q\in\operatorname{Spec}(B)$ and $p:=q\cap A$. Clearly, $S:=(A\setminus p)\times 0$ is a multiplicatively closed subset of $R$ with the feature that $\frac{r}{1}$ is regular in $S^{-1}R$ if and only if $r$ is regular in $R$. So finitely generated regular ideals of $S^{-1}R$ originate from finitely generated regular ideals of $R$. Hence $A_{p}\propto B_{p}=S^{-1}R$ is a Prüfer ring. Whence $K=\operatorname{qf}(A_{p})\subseteq B_{p}\subseteq B_{q}$. It follows that $K\subseteq B=\bigcap B_{q}$, where $q$ ranges over $\operatorname{Spec}(B)$, as desired. Now, one can easily check that $K\subseteq B$ implies $K\propto B=Q(R)$, the total ring of quotients of $R$. Moreover, let $f=\sum(k_{i},b_{i})x^{i}$ and $g=\sum(k^{\prime}_{j},b^{\prime}_{j})x^{j}$ be two polynomials in $Q(R)[x]$. If there is $i$ or $j$ such that $k_{i}\not=0$ or $k^{\prime}_{j}\not=0$, then $(k_{i},b_{i})$ or $(k^{\prime}_{j},b^{\prime}_{j})$ is invertible, hence $c(f)=Q(R)$ or $c(g)=Q(R)$, whence $c(fg)=c(f)c(g)$ (this is Gauss lemma which asserts that a polynomial with unit content is Gaussian). If $k_{i}=k^{\prime}_{j}=0$ for all $i$ and $j$, then $c(fg)=0=c(f)c(g)$. Consequently, $Q(R)$ is a Gaussian ring and so is $R$ by [3, Theorem 3.3]. By Lemma 2.2, $A$ is a Prüfer domain, completing the proof of the first implication. Assume $A$ is a Prüfer domain with $K\subseteq B$. Let $I$ be a nonzero finitely generated ideal of $R$ minimally generated by $(a_{1},b_{1}),\ldots,(a_{n},b_{n})$. For any $a\not=0\in A$ and any $b,b^{\prime}\in B$, we have $(0,b^{\prime})=(a,b)(0,a^{-1}b^{\prime})$. So minimality forces either $a_{i}=0$ for each $i$ or $a_{i}\not=0$ for each $i$. In the first case, $I^{2}=0$ and hence $I$ is not a regular ideal. Next assume $a_{i}\not=0$ for each $i$. It follows that $I=(\sum Aa_{i})\propto B$ since $(a_{i},b)=(a_{i},b_{i})(1,a_{i}^{-1}(b-b_{i}))$ for each $i$ and any $b\in B$. Since $A$ is a Prüfer domain, $J:=\sum Aa_{i}$ is invertible and $aJ^{-1}$ is an ideal of $A$ for some $a\not=0\in A$. $\begin{array}[]{lcl}\hbox{So }(a,0)^{-1}(aJ^{-1}\propto B)I&=&(a,0)^{-1}(aJ^{-1}\propto B)(J\propto B)\\\ &=&(a,0)^{-1}(aJ^{-1}J\propto B)\\\ &=&(a,0)^{-1}(aA\propto B)\\\ &=&R.\end{array}$ Consequently, $R$ is a Prüfer ring and hence Gaussian by [3, Theorem 3.3], completing the proof of (1). (2) Assume $R$ is an arithmetical ring. By (1), $A$ is Prüfer with $K\subseteq B$. So $K\propto B=Q(R)$ is arithmetical since it is a localization of $R$. Let $b\not=0\in B$. Then $I:=((0,1),(0,b))Q(R)$ is principal and hence $I:=(0,b^{\prime})Q(R)$ for some $b^{\prime}\not=0\in B$. Further $(0,b)\in I$ yields $b=kb^{\prime}$ for some some $k\not=0\in K$, and then $I:=(0,k^{-1}b)Q(R)$. Moreover $(0,1)\in I$ yields $1=k^{\prime}k^{-1}b$ for some $k^{\prime}\not=0\in K$. It follows that $b\in K$ and thus $K=B$. Conversely, assume $A$ is Prüfer with $K=B$. By (1), $R=A\propto K$ is Gaussian. Moreover $Q(R)=K\propto K$ is a principal ring (a fortiori arithmetical) since it has a unique nonzero proper ideal $M:=0\propto K=T(0,1)$. By [3, Theorem 3.5], $R$ is arithmetical, completing the proof of (2). (3) Let $S:=A\setminus\\{0\\}$. So $T:=S\times 0$ is a multiplicatively closed subset of $R$. By Lemma 2.3, $\operatorname{w.dim}.dim(T^{-1}R)=\operatorname{w.dim}.dim(K\propto KB)=\infty$. So $\operatorname{w.dim}.dim(R)=\infty$. This completes the proof of the theorem. ∎ The following corollary is an immediate consequence of Theorem 2.1. ###### Corollary 2.4. Let $D$ be a domain, $K:=\operatorname{qf}(D)$, and $R:=D\propto~{}K$. Then the following statements are equivalent: (1) D is a Prüfer domain; (2) $R$ is an arithmetical ring; (3) $R$ is a Gaussian ring; (4) $R$ is a Prüfer ring. ∎ Recall Jensen’s 1966 result: “for a ring $R$, $\operatorname{w.dim}.dim(R)\leq 1$ if and only if $R$ is arithmetical and reduced” [18]. Classical examples of arithmetical rings with weak dimension $\gneqq 1$ stem from Jensen’s result as non-reduced principal rings, e.g., $\mathbb{Z}/n^{2}\mathbb{Z}$ for any integer $n\geq 2$. In this vein, Theorem 2.1 generates a new family of examples quite far from being principal as shown below. For this purpose, recall that a ring $R$ is finite conductor if $aR\cap bR$ and $(0:c)$ are finitely generated for any $a,b,c\in R$ [11]. The class of finite conductor rings properly contains the class of coherent (a fortiori, Noetherian and hence principal) rings [11, 19]. ###### Example 2.5. Let $D$ be any Prüfer domain which is not a field and $K:=\operatorname{qf}(D)$. Then $R:=D\propto~{}K$ is an arithmetical ring with $\operatorname{w.dim}.dim(R)=\infty$. Moreover, $R$ is not a finite conductor ring by [19, Thoerem 2.8] and hence not coherent. Also, Theorem 2.1 enriches the literature with new examples of non- arithmetical Gaussian rings, as shown below. ###### Example 2.6. Let $K\subsetneqq L$ be a field extension. Then $R:=K\propto~{}L$ is a Gaussian ring which is not arithmetical. The next example shows that Theorem 2.1 widens the scope of validity of Bazzoni-Glaz conjecture beyond the class of coherent Gaussian rings. ###### Example 2.7. Let $\mathbb{Z}$ and $\mathbb{R}$ denote the ring of integers and field of real numbers, respectively. Then $R:=\mathbb{Z}_{(2)}\propto~{}\mathbb{R}$ satisfies the following statements: (1) $R$ is a Gaussian ring, (2) $R$ is not an arithmetical ring, (3) $R$ is not a coherent ring, (4) $\operatorname{w.dim}.dim(R)=\infty$. ###### Proof. Assertions (1), (2), and (4) hold by direct application of Theorem 2.1. It remains to prove (3). Indeed, consider the following exact sequence over $R$ $0\rightarrow 0\propto~{}\mathbb{R}\rightarrow R\buildrel u\over{\rightarrow}R(0,1)=0\propto~{}\mathbb{Z}_{(2)}\rightarrow 0$ where $u$ is defined by $u(a,b)=(a,b)(0,1)=(0,a)$. Now $0\propto~{}\mathbb{R}$ is not finitely generated as an $R$-module (otherwise $\mathbb{R}$ would be finitely generated as a $\mathbb{Z}_{(2)}$-module). Hence $0\propto~{}\mathbb{Z}_{(2)}$ is a finitely generated ideal of $R$ that is not finitely presented. Whence $R$ is not coherent, as desired. ∎ The next example illustrates the failure of Theorem 2.1, in general, beyond the context of domain extensions. ###### Example 2.8. Let $(A,M)$ be a non-valuation local domain, $E$ a nonzero $A$-module with $ME=0$, and $B:=A\propto~{}E$. Then $R:=A\propto~{}B$ is a Prüfer ring which is not Gaussian. ###### Proof. Indeed, one can easily check that $R$ is a total ring of quotients and hence a Prüfer ring. By Lemma 2.2, $R$ is not Gaussian. ∎ ## 3\. A class of total rings of quotients In a recent paper devoted to Gaussian properties, Bazzoni and Glaz have proved that a Prüfer ring satisfies any of the other four Prüfer conditions (mentioned above) if and only if its total ring of quotients satisfies that same condition [3, Theorems 3.3 & 3.6 & 3.7 & 3.12]. This fact narrows the scope of study to the class of total rings of quotients. This section investigates Prüfer conditions in a particular class of total rings of quotients; namely, those arising as trivial ring extensions of local rings by vector spaces over the residue fields. The main result (Theorem 3.1) enriches the literature with original examples of non-arithmetical Gaussian total rings of quotients as well as non-Gaussian total rings of quotients (which are necessarily Prüfer). The theorem validates Bazzoni-Glaz conjecture for the class of Gaussian rings emanating from these constructions. ###### Theorem 3.1. Let $(A,M)$ be a local ring and $E$ a nonzero $\frac{A}{M}$-vector space. Let $R:=A\propto~{}E$ be the trivial ring extension of $A$ by $E$. Then: (1) $R$ is a total ring of quotients and hence a Prüfer ring. (2) $R$ is Gaussian if and only if $A$ is Gaussian. (3) $R$ is arithmetical if and only if $A:=K$ is a field and $\dim_{K}E=1$. (4) $\operatorname{w.dim}.dim(R)\gvertneqq 1$. If $M$ admits a minimal generating set, then $\operatorname{w.dim}.dim(R)=\infty$. ###### Proof. (1) Straightforward. (2) By Lemma 2.2, only the sufficiency has to be proved. Assume $A$ is a Gaussian ring and let $F=\sum(a_{i},e_{i})x^{i}$ be a polynomial in $R[x]$. If $a_{i}\notin M$ for some $i$, then $(a_{i},e_{i})$ is invertible in $R$, hence $F$ is Gaussian. Now assume $a_{i}\in M$ for each $i$ and let $G=\sum(a^{\prime}_{j},e^{\prime}_{j})x^{j}\in R[x]$. We may suppose, without loss of generality, that $a^{\prime}_{j}\in M$ for each $j$. Let $f=\sum a_{i}x^{i}$ and $g=\sum a^{\prime}_{j}x^{j}$ in $A[x]$. One can easily check that $ME=0$ yields the following $\begin{array}[]{lcl}c(FG)&=&c(fg)\propto c(fg)E\\\ &=&c(fg)\propto 0\\\ &=&c(f)c(g)\propto 0\\\ &=&c(F)c(G).\end{array}$ Therefore $F$ is Gaussian, as desired. (3) Sufficiency is clear since $K\propto~{}K$ is a principal ring. Next assume $R$ is an arithmetical ring. We claim that $A$ is a field. Deny and let $a\not=0\in M$ and $e\not=0\in E$. Therefore the ideal $I:=R(a,0)+R(0,e)$ is principal in $R$ (since $R$ is local). So $I=R(a^{\prime},e^{\prime})$ for some $(a^{\prime},e^{\prime})\in R$. Clearly, $(a,0)\in I$ forces $a^{\prime}$ to be nonzero and belong to $M$. Further, $(0,e)\in I$ yields $ba^{\prime}=0$ and $e=be^{\prime}$ for some $b\in A$. Necessarily, $b\in M$ since $a^{\prime}\not=0$. It follows that $e=be^{\prime}=0$, the desired contradiction. Now, let $e,e^{\prime}$ be two nonzero vectors in $E$. Then $I=R(0,e)+R(0,e^{\prime})$ is a principal ideal of $R$. Similar arguments used in the proof of Theorem 2.1(2) yields $e=ke^{\prime}$ for some $k\in K$. So that $\dim_{K}E=1$. (4) Let $J:=0\propto E$ and let $\\{(f_{i})\\}_{i\in I}$ be a basis of the $(A/M)$-vector space $E$. Consider the exact sequence of $R$-modules: (1) $0\rightarrow Ker(u)\rightarrow R^{(I)}\buildrel u\over{\rightarrow}J\rightarrow 0$ where $u((a_{i},e_{i})_{i\in I})=(0,\sum_{i\in I}a_{i}f_{i})$. Hence, $Ker(u)=(M\propto E)^{(I)}$. Here, too, we identify $R^{(I)}$ with $A^{(I)}\propto E^{(I)}$ as $R$-modules. We claim that $J$ is not flat. Otherwise, by [25, Theorem 3.55], we obtain $J^{(I)}=(M\propto E)^{(I)}\cap JR^{(I)}=J(M\propto E)^{(I)}=0$, absurd. By [25, Theorem 2.4], $\operatorname{w.dim}.dim(R)\gvertneqq 1$. Next assume that $M$ admits a minimal generating set. Then one can easily check that $M\propto E$ admits a minimal generating set too. Let $(b_{i},g_{i})_{i\in L}$ denote a minimal generating set of $M\propto E$ and consider the exact sequence of $R$-modules: $0\rightarrow Ker(v)\rightarrow R^{(L)}\buildrel v\over{\rightarrow}M\propto E\rightarrow 0$ where $v((a_{i},e_{i})_{i\in L})=\sum_{i\in L}(a_{i},e_{i})(b_{i},g_{i})$. The minimality assumption yields (see the proof of [25, Lemma 4.43]) $Ker(v)\subseteq(M\propto E)^{(L)}.$ It follows that $\operatorname{Ker}(v)=V\propto E^{(L)}=(V\propto 0)\oplus J^{(L)}$, where $V:=\\{(a_{i})_{i\in L}\in M^{(L)}\ \|\sum_{i\in L}a_{i}b_{i}=0\\}.$ We obtain (2) $\operatorname{fd}\big{(}(V\propto 0)\oplus J^{(L)}\big{)}\leq\operatorname{fd}(M\propto E).$ On the other hand, from the exact sequence in (1) we get (3) $\operatorname{fd}(M\propto E)=\operatorname{fd}(M\propto E)^{(I)}\leq\operatorname{fd}(J)-1.$ A combination of (2) and (3) yields $\operatorname{fd}(J)\leq\operatorname{fd}(J)-1$. Consequently, the flat dimension of $J$ (and a fortiori the weak dimension of $R$) has to be infinite, completing the proof of the theorem. ∎ Theorem 3.1 generates new and original examples of rings with zerodivisors subject to Prüfer conditions as shown below. ###### Example 3.2. Let $(V,M)$ be a non-trivial valuation domain. Then $R:=V\propto~{}\frac{V}{M}$ is a non-arithmetical Gaussian total ring of quotients. ###### Example 3.3. Let $K$ be a field and $E$ a $K$-vector space with $\dim_{K}E\geq 2$. Then $R:=K\propto~{}E$ is a non-arithmetical Gaussian total ring of quotients. ###### Example 3.4. Let $(A,M)$ be a non-valuation local domain. Then $R:=A\propto~{}\frac{A}{M}$ is a non-Gaussian total ring of quotients. Recently, Bazzoni and Glaz proved that a Gaussian ring, with a maximal ideal $M$ such that the nilradical of $R_{M}$ is non-null and nilpotent, has infinite weak dimension [3, Theorem 6.4]. The next example widens the scope of validity of Bazzoni-Glaz conjecture as well as illustrates the setting of this result beyond coherent Gaussian rings. ###### Example 3.5. Let $\mathbb{R}$ denote the field of real numbers and $x$ an indeterminate over $\mathbb{R}$. Then $R:=\mathbb{R}\propto~{}\mathbb{R}[x]$ satisfies the following statements: (1) $R$ is a Gaussian ring, (2) $R$ is not an arithmetical ring, (3) $R$ is not a coherent ring, (4) $R$ is local with nonzero nilpotent maximal ideal, (5) $\operatorname{w.dim}.dim(R)=\infty$. ###### Proof. Assertions (1) and (2) hold by direct application of Theorem 3.1. Assertion (3) is handled by [19, Theorem 2.6(2)]. Clearly, (4) holds since the maximal ideal of $R$ is $M:=0\propto~{}\mathbb{R}[x]$ (by [17, Theorem 25.1(3)]) with $M^{2}=0$. Finally, (5) is satisfied by Theorem 3.1(4), [3, Proposition 6.3], or [3, Theorem 6.4]. ∎ ## 4\. Kaplansky-Tsang-Glaz-Vasconcelos conjecture Let $R$ be a ring and $Q(R)$ its total ring of quotients. An ideal $I$ of $R$ is said to be invertible if $II^{-1}=R$, where $I^{-1}:=\\{x\in Q(R)\ \|\ xI\subseteq R\\}$. A nonzero ideal is invertible if and only if it is regular, finitely generated, and locally principal. Particularly, for finitely generated ideals of domains, invertibility coincides with the locally principal condition. A polynomial $f$ over $R$ is said to be Gaussian if $c(fg)=c(f)c(g)$ holds for any polynomial $g$ over $R$. A problem initially associated with Kaplansky and his student Tsang [1, 2, 14, 22, 26] and also termed as Tsang-Glaz-Vasconcelos conjecture in [16] sustained that “every nonzero Gaussian polynomial over a domain has an invertible (or, equivalently, locally principal) content ideal.” It is well-known that a polynomial over any ring is Gaussian if its content ideal is locally principal. The converse is precisely the object of Kaplansky-Tsang-Glaz- Vasconcelos conjecture extended to those rings where “every Gaussian polynomial has locally principal content ideal.” Notice for convenience that the conjecture has a local character since the Gaussian condition is a local property (i.e., a polynomial is Gaussian over a ring $R$ if and only if its image is Gaussian over $R_{M}$ for each maximal ideal $M$ of $R$). It is this very fact that has enabled a natural extension of the conjecture from domains to rings (recall, for instance, that a Von Neumann regular ring is locally a field). Significant progress has been made on this conjecture. Glaz and Vasconcelos proved it for normal Noetherian domains [14]. Then Heinzer and Huneke established its veracity over locally approximately Gorenstein rings (see definition below) and over locally Noetherian domains [16, Theorem 1.5 & Corollary 3.4]. Recently, Loper and Roitman settled the conjecture for (locally) domains [21, Theorem 4], and then Lucas extended their result to arbitrary rings by restricting to polynomials with regular content [22, Theorem 6]. Obviously, the conjecture is true in arithmetical rings. Moreover, trivial ring extensions offer the possibility to widen the scope of its validity to a large family of rings distinct from the above contexts. This gives birth to a new class of rings containing strictly the three classes of arithmetical rings, of locally domains, and of locally approximately Gorenstein rings (see Figure 1). We term the new concept as follows: ###### Definition 4.1. A ring $R$ is pseudo-arithmetical if every Gaussian polynomial over $R$ has locally principal content ideal. We first prove a transfer result (Theorem 4.2) on trivial ring extensions. Then Conjecture 4.5 will equate the pseudo-arithmetical notion with the local irreducibility of the zero ideal. If true, this conjecture would offer an optimal solution to the Kaplansky-Tsang-Glaz-Vasconcelos conjecture that recovers all previous results. ###### Theorem 4.2. (1) Let $R:=A\propto~{}K$ be the trivial ring extension of a domain $A$ by its quotient field $K$. Then $R$ is a pseudo-arithmetical ring. (2) Let $A\subseteq B$ be an extension of rings and $R:=A\propto~{}B$. If $R$ is a pseudo-arithmetical ring, then so is $A$. ###### Proof. (1) Let $F:=\sum(a_{i},k_{i})x^{i}$ be a nonzero Gaussian polynomial in $R[x]$. Assume $a_{i}\not=0$ for some $i$. Then $(a_{i},k_{i})$ is regular in $R$ and so is $c(F)$ in $R$. Hence the Gaussian property forces $F$ to be regular in $R[x]$. Then $c(F)$ is locally principal by [22, Theorem 6]. Next assume $a_{i}=0$ for each $i$. Let $a$ be a nonzero element of $A$ such that $ak_{i}\in A$ for each $i$ and set $F^{\prime}:=(a,0)F=\sum(0,ak_{i})x^{i}$ in $R[x]$. We claim that $f^{\prime}:=\sum ak_{i}x^{i}$ is a (nonzero) Gaussian polynomial of $A[x]$. Indeed, consider $g=\sum a^{\prime}_{i}x^{i}\in A[x]$ and set $G:=\sum(a^{\prime}_{i},0)x^{i}$ in $R[x]$. Then $0\propto c(f^{\prime}g)=c(F^{\prime}G)=c(F^{\prime})c(G)$. Moreover, $c(F^{\prime})=\sum R(0,ak_{i})=0\propto c(f^{\prime})$ and $c(G)=c(g)\propto K$ (see proof of Lemma 2.2). It follows that $0\propto c(f^{\prime}g)=0\propto c(f^{\prime})c(g)$ and hence $c(f^{\prime}g)=c(f^{\prime})c(g)$. Whence $c(f^{\prime})$ is locally principal since $A$ is a domain [21]. Let $P:=p\propto K\in\operatorname{Max}(R)$ for some maximal ideal $p$ of $A$ and set $S:=(A\setminus p)\times 0\subseteq R\setminus P$. Since $c(f^{\prime})A_{p}=a^{\prime}A_{p}$ for some $a^{\prime}\in A$, we get $\begin{array}[]{lcl}(a,0)c(F)R_{P}&=&c(F^{\prime})R_{P}\\\ &=&\big{(}0\propto c(f^{\prime})\big{)}R_{P}\\\ &=&\big{(}S^{-1}(0\propto c(f^{\prime}))\big{)}R_{P}\\\ &=&\big{(}0\propto c(f^{\prime})A_{p}\big{)}R_{P}\\\ &=&\big{(}0\propto a^{\prime}A_{p}\big{)}R_{P}\\\ &=&(0,a^{\prime})R_{P}\\\ &=&(a,0)(0,\frac{a^{\prime}}{a})R_{P}.\end{array}$ Consequently, $c(F)R_{P}=(0,\frac{a^{\prime}}{a})R_{P}$ since $(a,0)$ is regular in $R$. Thus $c(F)$ is locally principal and therefore $R$ is a pseudo-arithmetical ring. (2) Let $f=\sum a_{i}x^{i}$ be a Gaussian polynomial over $A$ and set $F:=\sum(0,a_{i})x^{i}$. Let $G=\sum(a^{\prime}_{i},b_{i})x^{i}\in R[x]$ and set $g:=\sum a^{\prime}_{i}x^{i}$ in $A[x]$. Since $f$ is Gaussian, we have $c(F)c(G)=(0\propto c(f))c(G)=0\propto c(f)c(g)=0\propto c(fg)$. On the other hand, one can see that $c(FG)=0\propto c(fg)$. Therefore, $c(FG)=c(F)c(G)$, hence $F$ is a Gaussian polynomial over $R$. So $c(F)=0\propto I$ is a locally principal ideal of $R$ where $I:=c(f)$. Now ape the proof of the arithmetical statement in Lemma 2.2, to get that $I$ is locally principal, as desired. ∎ Obviously, a ring is arithmetical if and only if it is Gaussian and pseudo- arithmetical. In this context, note that Examples 2.6 & 3.3 illustrate the failure of Theorem 4.2(1) for trivial ring extensions $R:=A\propto~{}E$ with $E\not=\operatorname{qf}(A)$. ###### Example 4.3. Let $(A,M)$ be a local ring which is not a field and $E$ a nonzero vector space over $\frac{A}{M}$. Then $R:=A\propto~{}E$ is a Prüfer ring which is not pseudo-arithmetical. Indeed, Theorem 3.1 ensures that $R$ is a non- arithmetical total ring of quotients (hence Prüfer). We claim that the polynomial $f:=(a,0)+(0,e)x$, where $a\not=0\in M$ and $e\not=0\in E$, is Gaussian but $c(f)$ is not principal in $R$. To see this, let $g\in R[x]$. If $g\notin(M\times~{}E)[x]$, then Gauss lemma ensures that $c(fg)=c(f)c(g)$ since $R$ is local with ideal maximal $M\propto~{}E$. Assume $g\in(M\times~{}E)[x]$. Then $ME=0$ yields $c(f)c(g)=(a,0)c(g)=c((a,0)g)=c(fg).$ Now ape the proof of Theorem 3.1(3), to obtain that $c(f)$ is not principal, and therefore $R$ is not pseudo-arithmetical. ###### Remark 4.4. (1) Now pick any non-Prüfer domain $A$ with $K:=\operatorname{qf}(A)$ and consider the trivial extension $R:=A\propto~{}K$. Then by Corollary 2.4, $R$ is not a Prüfer ring (a fortiori, $R$ is not arithmetical). Moreover, there are plenty of non-regular Gaussian polynomials over $R$, e.g., $f:=\sum(0,k_{i})x^{i}$. However, Theorem 4.2 ensures that every Gaussian polynomial over $R$ has locally principal content ideal (i.e., $R$ is pseudo- arithmetical). (2) Next we examine the Noetherian case. From [16], a local ring $(R,M)$ is said to be approximately Gorenstein if $R$ is Noetherian and for every integer $n>0$ there is an ideal $I\subseteq M^{n}$ such that $R/I$ is Gorenstein (e.g., any local Noetherian ring $(R,M)$ with the $M$-adic completion $\hat{R}$ reduced). Heinzer and Huneke proved that every locally approximately Gorenstein ring is pseudo-arithmetical [16, Theorem 1.5]. This result combined with [16, Remark 1.6] asserts that Noetherianity has no direct effect on the pseudo-arithmetical notion even in low dimension, in the sense that non- Gorenstein Artinian local rings are not pseudo-arithmetical. Finally, notice that the above example $R:=A\propto~{}K$ is not Noetherian since it is not coherent by [19, Theorem 2.8]. (3) From [12], a ring $R$ is called a PF ring if all principal ideals of $R$ are flat, or, equivalently, if $R$ is locally a domain [10, Theorem 4.2.2(3)]. A ring $R$ is called a PP ring or a weak Baer ring if all principal ideals of $R$ are projective. In the class of Gaussian rings, the PP and PF properties coincide, respectively, with the notions of semihereditary ring and ring with weak dimension at most 1. Clearly, note that the above example $R:=A\propto~{}K$ is not locally a domain. In view of Example 4.3 and Remark 4.4, Figure 1 summarizes the relations between all these classes of rings where the implications are irreversible in general. Figure 1. Pseudo-arithmetical rings in perspective From the above discussion, it turns out that the pseudo-arithmetical notion must have a characterization that accommodates the three disparate classes of arithmetical rings, of locally domains, and of locally approximately Gorenstein rings (see Figure 1). This new characterization will offer a “happy end” to the Kaplansky-Tsang-Glaz-Vasconcelos conjecture. In this vein, we conjecture the following: ###### Conjecture 4.5. A ring $R$ is pseudo-arithmetical if and only if the zero ideal is locally irreducible. ###### Remark 4.6. (1) Fuchs, Heinzer and Olberding have recently studied irreducibility in commutative rings [7, 8] and noticed that “it is readily seen that a ring $R$ is an arithmetical ring if and only if for each proper ideal $I$ of $R$, $I_{M}$ is an irreducible ideal of $R_{M}$ for every maximal ideal $M$ of $R$ containing $I$ [7].” (2) Assume that Conjecture 4.5 is true. If $R$ is locally a domain or locally approximately Gorenstein, then a polynomial over $R$ is Gaussian if and only if its content is locally principal [21, Theorem 4] & [16, Theorem 1.5]. In particular, a nonzero polynomial over an integral domain is Gaussian if and only if its content is invertible. Indeed the locally domain statement follows from the obvious fact that the zero ideal in a domain is irreducible. Next assume $R$ is locally approximately Gorenstein. Recall that a Gaussian polynomial $f:=\sum a_{i}x^{i}$ over a ring $R$ forces its image $\overline{f}:=\sum\overline{a_{i}}x^{i}$ to be Gaussian over $R/I$, for every ideal $I$ of $R$. Using this fact and the fact that the Gaussian condition is a local property, in combination with the definition of a locally approximately Gorenstein ring, Heinzer and Huneke showed that the proof reduces to the case where $R$ is a zero-dimensional local Gorenstein ring (see the beginning of the proof of [16, Theorem 1.5]). But in this setting the zero ideal is irreducible, as desired. ## References [1] D. D. Anderson and B. J. Kang, Content formulas for polynomials and power series and complete integral closure, J. Algebra 181 (1996) 82–94. * [2] S. Bazzoni and S. Glaz, Prüfer rings, Multiplicative Ideal Theory in Commutative Algebra, pp. 263–277, Springer, 2006. * [3] S. Bazzoni and S. Glaz, Gaussian properties of total rings of quotients, J. Algebra 310 (2007) 180–193. * [4] H. S. Butts and W. Smith, Prüfer rings, Math. Z. 95 (1967) 196–211. * [5] H. Cartan and S. Eilenberg, Homological Algebra, Princeton University Press, 1956. * [6] L. Fuchs, Uber die Ideale arithmetischer Ringe, Comment. Math. Helv. 23 (1949) 334–341. * [7] L. Fuchs, W. Heinzer and B. Olberding, Commutative ideal theory without finiteness conditions: Primal ideals, Trans. Amer. Math. Soc. 357 (2005) 2771-2798. * [8] L. Fuchs, W. Heinzer and B. Olberding, Commutative ideal theory without finiteness conditions: Completely irreducible ideals, Trans. Amer. Math. Soc. 358 (2006) 3113-3131. * [9] R. Gilmer, Multiplicative Ideal Theory, Marcel Dekker, New York, 1972. * [10] S. Glaz, Commutative Coherent Rings, Lecture Notes in Mathematics, 1371, Springer-Verlag, Berlin, 1989. * [11] S. Glaz, Finite conductor rings, Proc. Amer. Math. Soc. 129 (2000) 2833–2843. * [12] S. Glaz, The weak dimension of Gaussian rings, Proc. Amer. Math. Soc. 133 (9) (2005) 2507–2513. * [13] S. Glaz, Prüfer conditions in rings with zero-divisors, CRC Press Series of Lectures in Pure Appl. Math. 241 (2005) 272–282. * [14] S. Glaz and W. Vasconcelos, The content of Gaussian polynomials, J. Algebra 202 (1998) 1–9. * [15] M. Griffin, Prüfer rings with zero-divisors, J. Reine Angew Math. 239/240 (1969) 55–67. * [16] W. Heinzer and C. Huneke, Gaussian polynomials and content ideals, Proc. Amer. Math. Soc. 125 (1997) 739–745. * [17] J. A. Huckaba, Commutative Rings with Zero-Divisors, Marcel Dekker, New York, 1988. * [18] C. U. Jensen, Arithmetical rings, Acta Math. Hungr. 17 (1966) 115–123. * [19] S. Kabbaj and N. Mahdou, Trivial extensions defined by coherent-like conditions, Comm. Algebra 32 (10) (2004) 3937–3953. * [20] W. Krull, Beitrage zur arithmetik kommutativer integritatsbereiche, Math. Z. 41 (1936) 545–577. * [21] K. A. Loper and M. Roitman, The content of a Gaussian polynomial is invertible, Proc. Amer. Math. Soc. 133 (5) (2004) 1267–1271. * [22] T. G. Lucas, Gaussian polynomials and invertibility, Proc. Amer. Math. Soc. 133 (7) (2005) 1881–1886. * [23] B. Osofsky, Global dimension of commutative rings with linearly ordered ideals, J. London Math. Soc. 44 (1969) 183–185. * [24] H. Prüfer, Untersuchungen uber teilbarkeitseigenschaften in korpern, J. Reine Angew. Math. 168 (1932) 1–36. * [25] J. J. Rotman, An Introduction to Homological Algebra, Academic Press, New York, 1979. * [26] H. Tsang, Gauss’s Lemma, Ph.D. thesis, University of Chicago, Chicago, 1965.	arxiv-papers	2008-08-02T18:20:26	2024-09-04T02:48:57.152623	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "C. Bakkari, S. Kabbaj, N. Mahdou", "submitter": "Salah-Eddine Kabbaj", "url": "https://arxiv.org/abs/0808.0275" }
0808.0361	# Coulomb excitation at intermediate energies H. Esbensen Physics Division, Argonne National Laboratory, Argonne, Illinois 60439, USA ###### Abstract Straight line trajectories are commonly used in semi-classical calculations of the first-order Coulomb excitation cross section at intermediate energies, and simple corrections are often made for the distortion of the trajectories that is caused by the Coulomb field. These approximations are tested by comparing to numerical calculations that use exact Coulomb trajectories. In this paper a model is devised for including relativistic effects in the calculations. It converges at high energies towards the relativistic straight-line trajectory approximation and approaches the non-relativistic Coulomb trajectory calculation at low energies. The model is tested against a number of measurements and analyses that have been performed at beam energies between 30 and 70 MeV/nucleon, primarily of quadrupole excitations. Remarkably good agreement is achieved with the previous analyses, and good agreement is also achieved in the few cases, where the B(E$\lambda$) value is known from other methods. The magnitudes of the relativistic and Coulomb distortion effects are discussed. ###### pacs: PACS number(s): 23.20.-g; 25.70.De; 25.70.-z ## I Introduction The first-order excitation of a nucleus, induced by the Coulomb field from another nucleus, has been discussed in detail in Ref. AW-book . The semi- classical perturbation theory developed there was based on a classical, non- relativistic Coulomb orbit for the relative motion of the projectile and target nuclei. A relativistic description was later developed but it was restricted to straight-line trajectories AW-NPA . This leaves a gap in the theoretical description at intermediate energies (say, at 20-200 MeV/nucleon), where both relativistic effects and Coulomb distortions of the trajectory can be important. This is very unfortunate because many experiments have been performed in recent years at intermediate-energy, radioactive beam facilities. These include Coulomb dissociation experiments, for example, of 8B, 11Be and 11Li, and Coulomb excitation measurements in inverse kinematics, mostly of the lowest $2^{+}$ state in light and medium heavy nuclei glass . The purpose of this paper is to devise a model that interpolates smoothly between the non-relativistic description of the Coulomb excitation which is based on Coulomb trajectories AW-book and the relativistic description which is based on straight-line trajectories AW-NPA . To develop an exact theory that contains the two extreme descriptions as limits is in general a difficult task. It has been studied by Bertulani et al. aleixo ; bert68 , who considered the effects of retardation explicitly for a Coulomb trajectory. The method proposed here is much simpler but it is sufficiently accurate for analyzing the data that have been obtained at intermediate energies. The point is that the experimental uncertainties are often large, typically of the order of 10%, and there are also theoretical uncertainties, which can distort the analysis of data, for example, the influence of nuclear and higher-order processes. The accuracy of the proposed model of the intermediate-energy Coulomb excitation is tested by analyzing the results obtained with the commonly used ‘Coulomb corrected’ straight-line trajectory method, which originally was proposed by Winther and Alder AW-NPA . Another test is to use B(E2) values that are known from other measurements (for example, of the life-time) and compare the calculated cross sections to measurements that have been performed at intermediate energies. There has recently been a debate in the literature bertpl ; scheit about the validity of the analyses of Coulomb excitation experiments that have been performed in the past. It turned out to be caused by a misunderstanding of the experimental conditions, as pointed out in Ref. scheit . However, independent of the controversy, it was claimed that corrections to the low-energy and high-energy theories of Coulomb excitation could be as large as 20% or 30% at intermediate energies bert68 ; bertpl . In this paper it will be shown how this large uncertainty can be brought under control and reduced to only a few percent. The semi-classical theory of the non-relativistic Coulomb excitation is summarized in sections II to III. The expressions that are obtained from straight-line and Coulomb trajectories, respectively, are compared in section IV. The analytic expression for the relativistic Coulomb excitation in the straight line trajectory approximation is quoted in Section V, and it is used to devise a model which includes the combined effect of relativity and Coulomb trajectories. The model is tested in section VI against measurements and other calculations, and section VII contains the conclusions. ## II Non-relativistic Coulomb excitation The semi-classical, non-relativistic description of Coulomb excitation AW-book is summarized in the following. Thus we consider a target nucleus with atomic number $Z_{2}$ which is being excited by the Coulomb field from a projectile nucleus with atomic number $Z_{1}$. It is assumed that the projectile and target do not overlap during the collision. We can therefore use the so-called far-field approximation which assumes that the intrinsic coordinate $r$ of the target nucleus is smaller that the distance $R(t)$ between projectile and target. The first-order amplitude for the electric excitation of the target nucleus, from an initial state $\|i>$ to a final state $\|f>$, is given by the multipole expansion AW-book , $a_{fi}=\frac{Z_{1}e^{2}}{i\hbar}\sum_{\lambda\mu}\sqrt{\frac{4\pi}{2\lambda+1}}S_{\lambda\mu}(\omega)<f\|M_{\lambda\mu}^{}\|i>,$ (1) where $\hbar\omega=\Delta E_{fi}$ is the excitation energy, $M_{\lambda\mu}$ is the multipole operator $r^{\lambda}Y_{\lambda\mu}({\hat{r}})$, and $S_{\lambda\mu}(\omega)=\sqrt{\frac{4\pi}{2\lambda+1}}\int_{-\infty}^{\infty}dt\ e^{i\omega t}\ \frac{Y_{\lambda\mu}({\hat{R}}(t))}{R(t)^{\lambda+1}}=\int_{-\infty}^{\infty}dt\ e^{i\omega t}\ \frac{d_{\mu 0}^{\lambda}({\beta})e^{i\mu\phi}}{(R(t))^{\lambda+1}}$ (2) is the so-called orbital integral. It depends on the orbit ${\vec{R}}(t)$ of the projectile with respect to the target nucleus. The unit vector ${\hat{R}}={\vec{R}}/R$ is expressed in terms of the spherical coordinates $(\beta,\phi)$ in the last expression. It is noted that this definition of the orbital integral, Eq. (2), differs by the factor $\sqrt{4\pi/(2\lambda+1)}$ from the convention used in Ref. AW-book . The calculation of the orbital integrals is discussed in the following sections for two choices of the coordinate system. The first choice, system A, is the most convenient for numerical calculations that are based on a Coulomb orbit AW-book . The second choice, referred to as system H, is more convenient at high energies, where a straight line trajectory becomes an accurate approximation and relativistic effects can be treated exactly AW-NPA . The transformation between the two representations will be discussed in order to be able to compare the two extreme methods and devise a model that interpolates smoothly between them. ### II.1 Cross sections The Coulomb excitation cross section is calculated as the product of the first-order excitation probability $P_{fi}$ and the elastic Rutherford cross section $(d\sigma/d\Omega)_{R}$, $\Bigl{(}\frac{d\sigma}{d\Omega}\Bigr{)}_{fi}=P_{fi}\ \Bigl{(}\frac{d\sigma}{d\Omega}\Bigr{)}_{\rm R}.$ (3) The Rutherford cross section can be obtained from the Rutherford scattering formula, $\tan(\theta/2)=a/b$, where $b$ is the impact parameter and $a=\frac{Z_{1}Z_{2}e^{2}}{M_{0}v^{2}},$ (4) is half the of distance of closest approach in a head-on collision, and $M_{0}$ is the reduced mass, $M_{0}=M_{1}M_{2}/(M_{1}+M_{2})$. The Rutherford cross section is $\Bigl{(}\frac{d\sigma}{d\Omega}\Bigr{)}_{\rm R}=\frac{2\pi bdb}{d\Omega}=\pi a^{2}\ \frac{d}{d\Omega}\Bigl{(}\frac{1}{\sin^{2}(\theta/2)}\Bigr{)}.$ (5) The average excitation probability obtained from Eq. (1) is the average over the initial magnetic sub-states $M_{i}$ and the sum over the final m-sub- states $M_{f}$, $P_{fi}=\frac{4\pi Z_{1}^{2}e^{4}}{(2\lambda+1)\hbar^{2}}\ \sum_{\mu}\|S_{\lambda\mu}\|^{2}\ \frac{1}{2I_{i}+1}\sum_{M_{i}M_{f}}\|\langle I_{f}M_{f}\|M_{\lambda\mu}^{}\|I_{i}M_{i}\rangle\|^{2}.$ The last sum divided by $(2I_{i}+1)$ is equal to $(2\lambda+1)^{-1}$ times the multipole strength (or reduced transition probability) for the excitation, i. e., $P_{fi}=4\pi\ \Bigl{(}\frac{Z_{1}e^{2}}{(2\lambda+1)\hbar}\Bigr{)}^{2}\ B(E\lambda)\ \sum_{\mu}\|S_{\lambda\mu}\|^{2}.$ (6) ## III Orbital integrals in coordinate system A In the coordinate system denoted by A (c.f. Ref. AW-book ), the z-axis is perpendicular to the scattering plane so the orbit has the form ${\vec{R}}(t)=(x(t),y(t),0)$. The angle $\beta$ is fixed at $\pi/2$, and the orbital integral (2) can be written as $S_{\lambda\mu}^{A}=d_{\mu 0}^{\lambda}({\pi\over 2})\int_{-\infty}^{\infty}dt\ e^{i\omega t}\ {(x(t)+iy(t))^{\mu}\over(R(t))^{\lambda+\mu+1}}.$ (7) The factor $d_{\mu 0}^{\lambda}(\pi/2)$ ensures that the orbital integral is non-zero only when $\lambda+\mu$ is even. In the coordinate system A one chooses the x-axis as the symmetry axis of the Coulomb orbit so that $x(-t)=x(t)$ and $y(-t)=-y(t)$. From this it follows that $S_{\lambda\mu}^{A}=\Bigl{(}S_{\lambda\mu}^{A}\Bigr{)}^{}$ is a real quantity. To calculate $S_{\lambda-\mu}^{A}$ when $\mu>0$ one can use the expression $S_{\lambda-\mu}^{A}=d_{\mu 0}^{\lambda}({\pi\over 2})\int_{-\infty}^{\infty}dt\ e^{i\omega t}\ {(x(t)-iy(t))^{\mu}\over(R(t))^{\lambda+\mu+1}}.$ (8) Below we discuss the calculation for a Coulomb trajectory and compare it to the result of the straight line approximation. ### III.1 Coulomb trajectories To evaluate the orbital integrals numerically for a Coulomb orbit one makes use of a dimensionless time-variable $w$ (see Ref. AW-book ) so that $R(t)=a[\epsilon\cosh(w)+1],\ \ \ t={a\over v}[\epsilon\sinh(w)+w],$ (9) where $a$ is defined in Eq. (4) and $\epsilon$ is the eccentricity of the orbit, which can be expressed in terms of the impact parameter $b$ or the scattering angle $\theta$ in the center of mass system as follows, $\epsilon=\sqrt{1+(b/a)^{2}}=\frac{1}{\sin(\theta/2)}.$ (10) Inserting the Cartesian coordinates of the orbit (see AW-book ): $x=a[\cosh(w)+\epsilon],\ \ y=a\sqrt{\epsilon^{2}-1}\sinh(w),\ \ z=0,$ (11) into Eq. (7) one obtains, $S_{\lambda\mu}^{A}={1\over va^{\lambda}}d_{\mu 0}^{\lambda}\bigl{(}{\pi\over 2}\bigr{)}\ I_{\lambda\mu},$ (12) where $I_{\lambda\mu}=\int_{-\infty}^{\infty}dw\ \exp[i\xi_{a}(\epsilon\sinh(w)+w)]{[\cosh(w)+\epsilon+i\sqrt{\epsilon^{2}-1}\sinh(w)]^{\mu}\over(\epsilon\cosh(w)+1)^{\lambda+\mu}},$ (13) and $\xi_{a}=\omega a/v$. Properties and even tabulations of $I_{\lambda\mu}$ are given in Ref. AW56 . It will be calculated using a simple numerical integration with respect to $w$ over the finite interval [-5:5], and using just a few thousand steps. The accuracy of the numerical integration can be tested in the case of a straight-line trajectory against the analytic expressions, which are discussed in appendix A. ## IV Coordinate system H At high energies it is more convenient to use the coordinates system H where the z-axis is along the beam direction and the x-axis is along the impact parameter so that the y-axis is perpendicular to the scattering plane. The coordinates of the trajectory are therefore ${\vec{R}}(t)$ = $(x(t),0,z(t))$, which implies that $Y_{\lambda\mu}({\hat{R}})$ is real. Since $Y_{\lambda\mu}^{}=(-1)^{\mu}Y_{\lambda-\mu}$, it follows directly from the definition (2) that the orbital integral in coordinate system H must have the property, $S_{\lambda-\mu}^{H}(\omega)=(-1)^{\mu}S_{\lambda\mu}^{H}(\omega).$ (14) The coordinate system H is convenient at high energies because the analytic expressions for the orbital integrals that exist for straight-line trajectories are relatively simple in this representation even when the relativistic effects are included AW-NPA . This feature will be exploited in section V to devise an expression that contains the effects of relativity and the Coulomb distortion of the trajectory. To do that, we will need to transform the orbital integrals in coordinate system A to the new coordinate system H. To go from the H to the A coordinate system consists of a rotation of $\pi/2$ around the z-axis, followed by a rotation of $-\pi/2$ around the new y-axis, and finally a rotation of $-\pi/2$ around the final z-axis. The required transformation is therefore $S_{\lambda\mu}^{H}=\sum_{\mu^{\prime}}i^{\mu-\mu^{\prime}}\ d_{\mu\mu^{\prime}}^{\lambda}\bigl{(}-{\pi\over 2}\bigr{)}S_{\lambda\mu^{\prime}}^{A}.$ (15) Inserting Eq. (12) into this transformation we can write that $S_{\lambda\mu}^{H}=\frac{i^{\lambda+\mu}}{va^{\lambda}}\ I_{\lambda\mu}^{H},\ \ \ {\rm where}\ \ I_{\lambda\mu}^{H}=\sum_{\mu^{\prime}}i^{-(\lambda+\mu^{\prime})}\ d_{\mu^{\prime}\mu}^{\lambda}\bigl{(}{\pi\over 2}\bigr{)}\ d_{\mu^{\prime}0}^{\lambda}\bigl{(}{\pi\over 2}\bigr{)}\ I_{\lambda\mu^{\prime}}.$ (16) Values of $d_{\mu^{\prime}\mu}^{\lambda}(\pi/2)$ can be obtained from appendix D of Ref. AW-book . The explicit expressions one obtains for dipole and quadrupole excitations are $I_{10}^{H}=\frac{I_{1-1}-I_{11}}{2},\ \ \ I_{11}^{H}=\frac{I_{1-1}+I_{11}}{2\sqrt{2}},$ (17) $I_{20}^{H}=\frac{3}{8}\Bigl{[}I_{22}+I_{2-2}-\frac{2}{3}I_{20}\Bigr{]},\ \ I_{21}^{H}=\frac{1}{2}\sqrt{\frac{3}{8}}\Bigl{[}I_{2-2}-I_{22}\Bigr{]},\ \ I_{22}^{H}=\frac{1}{4}\sqrt{\frac{3}{8}}\Bigl{[}I_{22}+I_{2-2}+2I_{20}\Bigr{]}.$ (18) ### IV.1 Straight line trajectories In the straight-line trajectory approximation, the projectile moves with constant velocity $v$ along the z-axis at an impact parameter $b$ with respect to the target, ${\vec{R}}(t)=(b,0,vt)$. The non-relativistic orbital integrals have the analytic form AW-NPA , ${\tilde{S}}_{\lambda\mu}^{H}={2\over v}\ {i^{\lambda+\mu}\over\sqrt{(\lambda+\mu)!(\lambda-\mu)!}}\ \Bigl{(}{\omega\over v}\Bigr{)}^{\lambda}\ K_{\mu}(\xi_{b}),$ (19) where $\xi_{b}=\omega b/v$ is the adiabaticity parameter associated with the impact parameter $b$. The tilde on ${\tilde{S}}_{\lambda\mu}^{H}$ is a reminder of the straight-line trajectory approximation, and the superscript H refers to the coordinate system used here. It should be emphasized that the expression Eq. (19) can be derived by inserting the straight-line approximation, Eq. (A3), into the transformation, Eq. (16). ### IV.2 Coulomb trajectories In order to test the accuracy of Eq. (19) it is useful to write the general Coulomb trajectory result, Eq. (16), in a similar form, $S_{\lambda\mu}^{H}={2\over v}{i^{\lambda+\mu}\over\sqrt{(\lambda+\mu)!(\lambda-\mu)!}}\ \Bigl{(}{\omega\over v}\Bigr{)}^{\lambda}K_{\lambda\mu}^{eff}(b/a,\xi_{a}),$ (20) where $K_{\lambda\mu}^{eff}(b/a,\xi_{a})={1\over 2}\Bigl{(}{1\over\xi_{a}}\Bigr{)}^{\lambda}\ \sqrt{(\lambda+\mu)!(\lambda-\mu)!}\ I_{\lambda\mu}^{H},$ (21) and $I_{\lambda\mu}^{H}$ are the orbital integrals defined in coordinate system H according to Eq. (16). They are given explicitly by Eqs. (17) and (18) for dipole and quadrupole excitations. One can check that $K_{\lambda\mu}^{eff}$ gives the correct modified Bessel function $K_{\mu}(\xi)$ when one inserts the straight-line trajectory results in coordinate system A, Eqs. (A4-A6) of appendix A, into the definition (16) of $I_{\lambda\mu}^{H}$. In the limit: $\xi_{a}<<1$ and $b/a>>1$, i. e., at high energies and large impact parameters, one should recover the result for a straight line, i. e. $K_{\lambda\mu}^{eff}(b/a,\xi_{a})$ $\rightarrow$ $K_{\mu}(\xi_{b})$. This convergence will be illustrated in the next section. To evaluate the excitation probability (6), we need the expression $\sum_{\mu}\|S_{\lambda\mu}^{H}\|^{2}=\frac{4}{v^{2}\ b^{2\lambda}}\ F_{\lambda}(b/a,\xi_{a}),$ where $F_{\lambda}(b/a,\xi_{a})=\sum_{\mu}\frac{\xi_{b}^{2\lambda}}{(\lambda+\mu)!(\lambda-\mu)!}\ \Bigr{(}K_{\lambda\mu}^{eff}(b/a,\xi_{a})\Bigr{)}^{2}.$ (22) In the straight-line trajectory approximation, where $K_{\lambda\mu}^{eff}(b/a,\xi_{a})$ $\rightarrow$ $K_{\mu}(\xi_{b})$, one obtains $F_{1}(\xi_{b})=\xi_{b}^{2}\Bigl{(}K_{0}^{2}(\xi_{b})+K_{1}^{2}(\xi_{b})\Bigr{)},\ \ \ {\rm and}\ \ \ F_{2}(\xi_{b})=\frac{\xi_{b}^{4}}{12}\ \Bigl{(}3K_{0}^{2}(\xi_{b})+4K_{1}^{2}(\xi_{b})+K_{2}^{2}(\xi_{b})\Bigr{)}.$ (23) for dipole and quadrupole excitations, respectively. For $\xi_{b}\rightarrow$ 0 these functions approach the constant values $F_{1}\rightarrow$1 and $F_{2}\rightarrow$1/3. ### IV.3 Comparison of results Here the convergence of the Coulomb trajectories calculations towards the straight-line trajectory calculation is illustrated by comparing the functions $F_{\lambda}(b/a,\xi_{a})$ defined in Eq. (22) to the analytic expressions (23) for dipole and quadrupole excitations. The solid curves in Fig. 1 show the results of the straight-line trajectory approximation, Eq. (23), for dipole ($\lambda=1$) and quadrupole ($\lambda=2$) excitations, respectively, as functions of the adiabaticity parameter $\xi_{b}=\omega b/v$. The results of the Coulomb trajectory calculations defined in Eq. (22) are shown by dashed curves for three values of $b/a$. It is seen that the Coulomb trajectory results approach the straight-line result in a smooth manner for increasing values of $b/a$. The solid circles in Fig. 1 are the results one obtains by multiplying the straight-line trajectory expressions, Eq. (23), with the factor $\exp(-\pi\xi_{a})=\exp(-\pi\xi_{b}a/b)$ for $b/a$=5. This correction factor was suggested by Winther and Alder AW-NPA and it is seen to reproduce the dipole results ($\lambda$=1) for the Coulomb trajectory with $b/a$=5 fairly well. However, it does not work so well for quadrupole excitations ($\lambda$=2). The problem is that the Coulomb trajectory results depend on $b/a$ for $\xi_{b}\rightarrow 0$, whereas the simple correction factor $\exp(-\pi\xi_{a})=\exp(-\pi\xi_{b}a/b)$ is 1 in this limit. The non-relativistic cross sections one obtains for the dipole excitation of the 11Be $1/2^{+}$ ground state to the $1/2^{-}$ excited state and for the quadrupole excitation of 42S to the $2^{+}$ state are illustrated in Fig. 2. The cross sections were calculated for a gold target with a minimum impact parameter of 14 fm and they are shown as functions of the beam energy. Although these cross sections are referred to as non-relativistic, it should be emphasized that the velocity that has been used here was actually obtained from the relativistic expression, Eq. (B1), in terms of the beam energy. The excitation energies and B(E$\lambda$) values that have been used for 11Be and 42S are shown in Table I, which will be discussed later. The top dashed curves in Fig. 2 show the straight-line trajectory calculations, and the thin dashed curves are the same results multiplied by the factor $\exp(-\pi\xi_{a})$. The solid curves are based on Coulomb trajectories. They approach the straight-line trajectory approximation fairly quickly for the dipole excitation but somewhat slower for the quadrupole excitation. It is also seen that the approximation of multiplying the straight line calculation with the factor $\exp(-\pi\xi_{a})$ works quite well for the dipole excitation when compared with the Coulomb trajectory calculation, whereas this approximation is poorer for quadrupole excitations. Other approximations have been applied to correct the straight line trajectory approximation for the distortion that is caused by a Coulomb trajectory and some of them will be discussed in section VI.D. ## V Relativistic expression The relativistic expressions for the orbital integrals in the straight-line trajectory approximation are AW-NPA , ${\tilde{S}}_{\lambda\mu}^{H}(rel)={2\over\gamma v}{i^{\lambda+\mu}{\bar{G}}_{\lambda\mu}\over\sqrt{(\lambda+\mu)!(\lambda-\mu)!}}\ \Bigl{(}{\omega\over v}\Bigr{)}^{\lambda}\ K_{\mu}(\xi_{b}={\omega b\over\gamma v}).$ (24) where ${\bar{G}}_{\lambda\mu}$ can be extracted from Ref. AW-NPA . The notation used here is such that ${\bar{G}}_{\lambda\mu}=1$ in the non- relativistic limit, whereas Ref. AW-NPA uses a different normalization. For electric dipole and quadrupole excitations one finds that ${\bar{G}}_{10}={\bar{G}}_{20}={\bar{G}}_{2\pm 2}={1\over\gamma},\ \ {\bar{G}}_{1\pm 1}=1,\ \ {\bar{G}}_{2\pm 1}={1\over 2}(1+{1\over\gamma^{2}}).$ (25) It is seen that the relativistic effects on the Coulomb excitation enter into Eq. (24) through the factor $\gamma^{-1}{\widetilde{G}}_{\lambda\mu}$, and in the adiabatic distance $\gamma v/\omega$ of the adiabaticity parameter $\xi_{b}=\omega b/(\gamma v)$, which is the argument of the modified Bessel functions. To complete the discussion of relativistic effects one should also specify the kinematics of Coulomb scattering at relativistic energies. This is done in appendix B. One of the reasons large relativistic effects have been observed is actually due to the difference between relativistic and non-relativistic kinematics, the main one being the determination of the velocity from the beam energy, Eq. (B1). A relativistic effect in Coulomb scattering is the $\gamma$-factor which appears in the definition (B3) of half the distance of closest approach. This effect is commonly agreed upon AW-NPA ; bert68 . There is also a relativistic correction to the reduced mass, Eq. (B2), which is less significant at intermediate energies, and it is usually ignored AW-NPA ; bert68 . However, for completeness, it is better to keep it in the case of really high energies. Finally, there are also relativistic effects in the transformation (B4) from the center-of-mass to laboratory scattering angles. ### V.1 Interpolating model To accurately calculate the Coulomb excitation at intermediate beam energies it is important to include relativistic effects and the effect of the Coulomb distortion on the relative trajectory of projectile and target. It may be difficult to derive in a general expression for the Coulomb excitation amplitude because of the acceleration in a Coulomb orbit. However, one can devise a formula which gives the correct expression in the non-relativistic regime for a Coulomb orbit, and which reproduces the relativistic expressions for a straight line trajectory in the high-energy regime. We shall see that the following expression, $S_{\lambda\mu}^{H}(rel)={2\over\gamma v}{i^{\lambda+\mu}\ {\bar{G}}_{\lambda\mu}\over\sqrt{(\lambda+\mu)!(\lambda-\mu)!}}\ \Bigl{(}{\omega\over v}\Bigr{)}^{\lambda}\ K_{\lambda\mu}^{eff}(b/a,\xi_{a}={\omega a\over\gamma v}),$ (26) serves the purpose of interpolating between the two energy regimes. It is first noted that Eq. (26) is identical to Eq. (20) if we insert $\gamma=1$. The low-energy regime is therefore correctly described. In the high-energy regime, we can assume that $b/a>>1$ and $\xi_{a}<<1$, which implies that the straight-line trajectory limit will be reached, $K_{\lambda\mu}^{eff}(b/a,\xi_{a})\rightarrow K_{\mu}(\xi_{b}),\ \ \ {\rm where}\ \ \xi_{b}=\frac{b}{a}\xi_{a},$ according to the discussions in section IV.B and IV.C. Since the value of $\xi_{a}$ in Eq. (26) is chosen as $\xi_{a}$ = $\omega a/(\gamma v)$ we obtain $\xi_{b}=\frac{b}{a}\times\frac{\omega a}{\gamma v}=\frac{\omega b}{\gamma v},$ which is the correct adiabaticity parameter for a straight-line trajectory in the relativistic regime, according to Eq. (24), so the high-energy regime will also be described correctly. ### V.2 Total cross sections A great advantage of the straight-line trajectory approximation is that one can obtain analytic expressions for the total cross section, evaluated for all impact parameters larger that a certain minimum impact parameter $b_{0}$. Thus one obtains AW-NPA $\sigma_{\lambda}=4\pi\Bigl{(}\frac{Z_{1}e^{2}}{(2\lambda+1)\hbar v}\Bigr{)}^{2}\ B(E\lambda)\sum_{\mu}\frac{4{\bar{G}}_{\lambda\mu}^{2}}{(\lambda+\mu)!(\lambda-\mu)!}\ \Bigl{(}{\omega\over v}\Bigr{)}^{2(\lambda-1)}\ g_{\mu}(\frac{\omega b_{0}}{\gamma v}),$ (27) where $g_{\mu}(\xi)=2\pi\int_{\xi}^{\infty}\xi d\xi\ K_{\mu}^{2}(\xi)=\pi\xi^{2}\Bigl{[}K_{\mu+1}(\xi)K_{\mu-1}(\xi)-K_{\mu}^{2}(\xi)\Bigr{]},$ (28) according to Eq. (3.4) of Ref. AW-NPA . For dipole and quadrupole excitations this yields $\sigma_{\lambda=1}=16\pi\Bigl{(}\frac{Z_{1}e^{2}}{3\hbar v}\Bigr{)}^{2}\ B(E1)\ \Bigl{[}\frac{1}{\gamma^{2}}\ g_{0}(\xi)+g_{1}(\xi)\Bigr{]},$ (29) $\sigma_{\lambda=2}=\frac{4\pi}{3}\Bigl{(}\frac{Z_{1}e^{2}}{5\hbar v}\Bigr{)}^{2}\ B(E2)\ (\frac{\omega}{\gamma v})^{2}\ \Bigl{[}3g_{0}(\xi)+g_{1}(\xi)\gamma^{2}(1+\gamma^{-2})^{2}+g_{2}(\xi)\Bigr{]}.$ (30) At intermediate and high energies, where one can assume that $\xi_{b}<<1$, one obtains the following simple asymptotic expression for the quadrupole excitation cross section, $\sigma_{\lambda=2}=\frac{1}{3}\ \Bigl{(}\frac{4\pi Z_{1}e^{2}}{5\hbar vb_{0}}\Bigr{)}^{2}\ B(E2).$ (31) This expression gives a surprisingly good estimate of the cross section at high energies and it provides a simple way of testing more elaborate numerical calculations. The expression (31) shows that the high-energy cross section is insensitive to the excitation energy. One can also conclude that relativistic effects are not dramatic for E2 transitions because all of the $\gamma$ factors that appear in Eq. (30) disappear when one takes the limit $\xi_{b}\rightarrow$0. We shall see in the next section that the large relativistic effects, which have been pointed out in the literature, are primarily caused by plotting the cross sections as a function of the beam energy $T$ per nucleon. The cross section (31), which is proportional to $v^{-2}$, will therefore be very sensitive to whether one uses non-relativistic ($v=\sqrt{2T/m}$) or relativistic kinematics (Eq. (B1)) to determine the velocity. For completeness it is noted that some relativistic effects do survive in the dipole cross section, Eq. (29), when one takes the limit $\xi_{b}\rightarrow$0 in the dipole cross section, Eq. (29), $\sigma_{\lambda=1}=\Bigl{(}\frac{4\pi Z_{1}e^{2}}{3\hbar v}\Bigr{)}^{2}\ B(E1)\ \Bigl{[}2\ln(\frac{1.123\gamma v}{\omega b_{0}})-(\frac{v}{c})^{2}\Bigr{]}.$ (32) This expression shows the well-know logarithmic dependence on $\gamma$. ## VI Applications Two examples of calculated cross sections are shown in Fig. 3, namely, for the dipole excitation of 11Be to the low-lying $1/2^{-}$ state, and the quadrupole excitation of 42S to the lowest $2^{+}$ state, respectively. The cross sections are shown as functions of the beam energy. In both cases the minimum impact parameter was set to $b_{0}$ = 14 fm and the target was Au. The excitation energy and multipole strength of the two transitions are given in Table I, which will be discussed below. The upper thick dashed curves in Fig. 3 show the relativistic Coulomb excitation cross section for straight-line trajectories. The solid curves are based on Coulomb trajectories and make use of the expression, Eq. (26), for the interpolating model. The lower, thin dashed curves in Fig. 3 are the cross sections one obtains by inserting $\xi_{a}=\omega a/v$ into the expression, Eq. (26), for the interpolating model. This is seen to be a poor approximation. Inserting instead $\xi_{a}=\omega a/(\gamma v)$ (the thick solid curves) one obtains a smooth transition from the low-energy to the high- energy theory. ### VI.1 Comparison to experiments A number of Coulomb excitation experiments have been performed at intermediate energies and some of them are quoted in Table I. Those considered here are of interest because sufficient experimental information was provided to repeat the analysis. The measured cross sections, $\sigma_{exp}$, can be compared to the straight-line trajectory approximation, $\sigma_{Strl}$, and to the cross section, $\sigma_{Coul}$, obtained from the interpolating model, Eq. (26). It is seen that the latter model performs very well in comparison to the measured cross sections. The deviations are insignificant in comparison to the experimental uncertainty. It should be emphasized that most of the B(E2) values quoted in Table I were extracted from the data using the relativistic straight-line trajectory approximation with some corrections included for the Coulomb distortion of the trajectory. The good agreement between $\sigma_{Coul}$ and the measured cross sections therefore shows that the previous analyses were very reasonable. The results of the straight-line trajectory approximation, $\sigma_{Strl}$, are quoted in the last column. They are typically 5 to 10% higher than the measured values. The ratio of measured and calculated cross sections is illustrated in Fig. 4. The solid circles show the ratio $\sigma_{exp}/\sigma_{Coul}$ with respect to the interpolating relativistic Coulomb excitation cross section. The average ratio is about 2% less than 1. The triangles in Fig. 4 show the experimental ratio $\sigma_{exp}/\sigma_{Strl}$ to the relativistic straight-line trajectory calculation. Here the deviation from 1 is much larger. The deviation from the solid points reflects the significance of the Coulomb distortion of the trajectory. It amounts to about 3 - 9%. The largest deviations from the solid points occur in the low energy experiments, runs no. 4-8 (see Table I.) There are two examples in Table I where the B(E2) values are known from other sources, namely, 24Mg and 26Mg. The calculated cross sections, $\sigma_{Coul}$, are also in these cases consistent with the measured cross sections. This is very fortunate because the uncertainties are small (about 5% on the measured cross sections and 3% on the B(E2) values). These two measurements therefore provide an independent test of the interpolating model, Eq. (26). The last example on a comparison with data is the Coulomb excitation of 46Ar ar46 , which was measured for a range of maximum acceptance angles. The measurements are compared to two calculations in Fig. 5, namely, the straight- line and the ‘interpolated’ Coulomb trajectory calculation, Eq. (26). Both calculations are in good agreement with the data because the experimental uncertainties are much larger than the difference between the two calculations. ### VI.2 Comparison to other methods The relativistic description for straight-line trajectories developed by Winther and Alder AW-NPA was based on the Liénard-Wiechert potential. An different method was used by Aleixo and Bertulani aleixo who calculated the retardation effects explicitly for Coulomb trajectories. The latter approach was used by Bertulani et al. bert68 ; bertpl to investigate the effects of relativity and Coulomb distortions at intermediate energies in much the same way it is done here. It is therefore of interest to compare Eq. (26) to the predictions of the more elaborate theory. An example of a comparison of calculated cross sections is shown in Fig. 6 for the Coulomb excitation of 46Ar to the $2^{+}$ on a Pb target. The input is the same as used in producing table 2 of Ref. bertpl , and it was assumed bertcom that the distance of closest approach for a Coulomb trajectory, $r_{0}(b)=a+\sqrt{a^{2}+b^{2}}$ according to Eqs. (9) and (10), has the minimum value $r_{min}$ = $R_{1}+R_{2}=1.2(A_{1}^{1/3}+A_{2}^{1/3})$. The solid line is the cross section obtained from the interpolating model, Eq. (26). The upper dashed curve is the result for straight line trajectories with minimum impact parameter $b_{min}=\sqrt{r_{min}(r_{min}-2a})$. The lower dashed curve is the non-relativistic cross section for Coulomb trajectories. It is seen that the solid curve interpolates smoothly between the non- relativistic Coulomb trajectory calculation at low energy and the relativistic straight-line trajectory calculation at high energy. The solid points in Fig. 6A are based on the cross sections published in table 2 of Ref. bertpl . They have here been multiplied by the factor 1.44 because a factor of $e^{2}$ was unfortunately omitted bertcom in the calculations of Ref. bertpl . The results are presented in Fig. 6B in terms of the ratio to the interpolating relativistic Coulomb trajectory calculation, Eq. (26). The average value of the solid points is close to one (actually 1.005 $\pm$ 0.002, to be precise) which shows that the two theories are in excellent agreement. This may not be so surprising because both theories approach the relativistic straight-line trajectory approximation at high energy and the non-relativistic Coulomb trajectory calculation at low energy. The lower dashed curve in Fig. 6B shows that relativistic effects are enormous at 500 MeV/nucleon. It is interesting that the large relativistic effects have a very simple explanation. Thus, according to the asymptotic expression, Eq. (31), the quadrupole excitation cross section is proportional to $v^{-2}$ at high energies. In the non-relativistic calculation this implies $\sigma_{2}^{\rm NR}\propto m/(2T)$. In the relativistic calculation one obtains instead $\sigma_{2}^{\rm rel}\propto$ $(m+T)^{2}/[T(2m+T)]$, according to Eq. (B1) of appendix B. The ratio of the two cross sections is therefore $\frac{\sigma_{2}^{\rm NR}}{\sigma_{2}^{\rm rel}}\approx\frac{m(2m+T)}{2(m+T)^{2}}.$ (33) (There is also a difference in the minimum impact parameter in the relativistic and non-relativistic calculations but the effect is small.) The expression (33) is illustrated in Fig. 6B by the triangles which explain quite accurately the behavior of the non-relativistic calculation at high energy. ### VI.3 Significance of relativistic effects Another way to illustrate the effects of relativity is to recalculate the cross sections shown in Fig. 3 assuming that $\gamma=1$ everywhere in the underlying equations, Eqs. (24-26). The velocity $v$, which appears explicitly in these equations, will be determined from the relativistic expression, Eq. (B1), in order to avoid the trivial relativistic effect described by Eq. (33). The results of such calculations show that the cross sections for the excitation of the $2^{+}$ state in 42S, which were shown in Fig. 3, change by less than 0.5%, both for the straight-line trajectory calculation (24) and also for the interpolated Coulomb trajectory model, Eq. (26). Similar calculations performed for the dipole excitation of 11Be (which were shown in Fig. 3) change the cross section by less than 1% at energies below 200 MeV/nucleon. The change is about 10% at 1 GeV/u but that is not so surprising because the asymptotic dipole cross section, Eq. (32), does contain a $\gamma$ factor, whereas the asymptotic quadrupole cross section, Eq. (31), does not. It is concluded that the relativistic effects in the Coulomb excitation of nuclei are large at intermediate and high energies but most of the effect is trivial and can easily be avoided by using the correct relativistic expression to determine the velocity from the beam energy. In the analysis of measurements it is also important to use relativistic kinematics when converting scattering angles into impact parameters. The non-trivial relativistic effects on the Coulomb quadrupole excitation, on the other hand, are surprisingly small. ### VI.4 Significance of Coulomb distortion Let us finally take a look at how the straight-line trajectory approximation can be corrected for Coulomb distortions. Two examples are shown in Fig. 7, namely, the Coulomb excitation to the $2^{+}$ state of 16N with excitation energy $E_{x}$ = 0.12 MeV (A), and also to the $2^{+}$ state of 54Ni with excitation energy $E_{x}$ = 1.4 MeV (B). The B(E2) values were taken from table 2 of Ref. bertpl . The results are shown in terms of the ratio to the interpolating, relativistic Coulomb excitation cross section, Eq. (26). The calculations were again performed with the minimum distance of closest approach $r_{min}=1.2(A_{1}^{1/3}+A_{2}^{1/3})$. The comparison with Ref. bertpl is illustrated in Fig. 7 by the solid points. The cross sections from table 2 of Ref. bertpl were again multiplied with the factor 1.44. It is seen that the solid points are very close to 1 for the excitation of 54Ni. There are some discrepancies for 16N, where the average ratio is 1.028 $\pm$ 0.012. It is therefore concluded that the two theories of relativistic Coulomb excitation agree within a few percent. The thick solid curves in Fig. 7 show the ratio for the relativistic straight- line trajectory approximation, and the dashed curves show the results of various ways to correct this approximation for Coulomb distortions. The top dashed curves show the straight-line approximation multiplied by the factor $\exp(-\pi\xi_{a})$. This factor has a large effect for the heavier nucleus 54Ni with the relatively large excitation energy but it has essentially no effect for the lighter nucleus 16N with the small excitation energy. The lowest dashed curves in Fig. 7 show the straight-line trajectory result one obtains by replacing the minimum impact parameter $b_{min}$ by the effective value $b_{eff}$ = $b_{min}+\pi a/2$. This approximation was justified in Ref. AW-NPA for large impact parameters, where the excitation probability falls off exponentially as $\exp(-2\omega b/(\gamma v))$. Thus by multiplying the excitation probability for a straight line trajectory, $P_{Strl}$, with the factor $\exp(-\pi\xi_{a})$ one obtains approximately, $P_{Strl}\ \exp(-\pi\xi_{a})\propto\exp(-\frac{2\omega b}{\gamma v})\ \exp(-\pi\frac{\omega a}{\gamma v})=\exp[-\frac{2\omega}{\gamma v}(b+\frac{\pi a}{2})].$ This argument does not always apply to the Coulomb excitation of low-lying states at intermediate energies because the minimum impact parameter is usually much smaller than the adiabatic distance $\gamma v/\omega$. Using the effective minimum impact parameter has a very large effect on the calculated cross section. It produces a ratio in Fig. 7 that is almost as far below 1 as the ratio for the pure straight line trajectory calculation is above 1. The approximation is therefore not very useful for quadrupole excitations. It works apparently better for dipole excitations, as discussed in connection with Figs. 2 and 3, but that will not be discussed here. The second lowest curves in Fig. 7A and 7B are the results one obtains by choosing the effective minimum impact parameter, $b_{eff}=r_{min}$, which is the minimum distance of closest approach that is used in the relativistic Coulomb trajectory calculation. This approximation was recommended by Goldberg goldberg , and it is evidently the best choice out of the four examples of approximations shown in Fig. 7. In the analysis of an actual experiment the cross section ratio discussed above would usually be closer to one because one would always choose a small acceptance angle (i. e., a large minimum impact parameter) in order to avoid the influence of nuclear and higher-order processes. ## VII Conclusions A model has been devised for including relativistic effects in calculations of the Coulomb excitation cross section at intermediate energies. The model interpolates smoothly between the theory of non-relativistic Coulomb excitation at low energy and the relativistic, straight-line trajectory approximation high energy. The results that were obtained with this model compare very well with the calculations performed by Bertulani et al., who included the relativistic retardation effects explicitly for Coulomb trajectories. It was demonstrated that the large relativistic effects, which have been pointed out in the literature, are mainly caused by comparing calculations that are based on a relativistic and a non-relativistic velocity, respectively. The non-trivial relativistic effects, which are beyond the simple relativistic kinematics of two-body scattering, are surprisingly small for quadrupole excitations. The Coulomb distortion, which is responsible for the deviation between the straight line trajectory approximation and calculations that are based on a Coulomb trajectory, can also have a very large effect. However, the effects of the Coulomb distortion are usually suppressed by the experimental conditions and simple corrections can be made to improve the accuracy of the straight line trajectory approximation. The proposed interpolating model reproduces fairly well the analyses that have been performed previously, primarily of quadrupole excitation experiments, at beam energies in the range of 30 to 70 MeV/nucleon. The average deviation is only a few percent. The model also reproduces the measured cross sections in the few cases where the quadrupole excitation strength is known accurately from other sources. The good agreement with the previous analyses is partly due to the experimental conditions, which suppress the effects of the Coulomb distortion, and partly to the fact that some corrections for the distortion were made in the analyses. However, if high precision Coulomb excitation experiments were pursued, it would be necessary to treat the Coulomb distortion more accurately in the analysis. It is believed that the interpolating Coulomb excitation model proposed here would provide a sufficiently accurate description. Acknowledgments. This work was supported by the U.S. Department of Energy, Office of Nuclear Physics, under Contract No. DE-AC02-06CH11357. ## VIII Appendix A: Straight line trajectory In coordinate system A, a straight line trajectory has the coordinates $x(t)=b$, where $b$ is the (constant) impact parameter, and $y(t)=vt$. Using the dimensionless integration variable $s=vt/b$, the orbital integral, Eq. (7), is therefore ${\tilde{S}}_{\lambda\mu}^{A}=d_{\mu 0}^{\lambda}({\pi\over 2})\ {1\over vb^{\lambda}}\int_{-\infty}^{\infty}ds\ e^{i\xi_{b}s}\ {(1+is)^{\mu}\over(1+s^{2})^{(\lambda+\mu+1)/2}}$ $=d_{\mu 0}^{\lambda}({\pi\over 2})\ {1\over vb^{\lambda}}\Bigl{(}1+{d\over d\xi_{b}}\Bigr{)}^{\mu}\int_{-\infty}^{\infty}ds\ {\cos(\xi_{b}s)\over(1+s^{2})^{(\lambda+\mu+1)/2}},$ $None$ where $\xi_{b}=\omega b/v$ is the adiabaticity parameter. The tilde on ${\tilde{S}}_{\lambda\mu}^{A}$ is a reminder that we are using the straight- line approximation. An analytic expression for the integral is given in the book by Gragshteyn and Ryzhik gradshteyn , Eqs. 8.432 no. 5, and one obtains ${\tilde{S}}_{\lambda\mu}^{A}=d_{\mu 0}^{\lambda}({\pi\over 2}){2\over vb^{\lambda}}{1\over(2n-1)!!}\Bigl{(}1+{d\over d\xi_{b}}\Bigr{)}^{\mu}\xi_{b}^{n}K_{n}(\xi_{b}),$ $None$ which is expressed in terms of modified Bessel functions of order $n=(\lambda+\mu)/2$. Here $\lambda+\mu$ is even as mentioned earlier so $n$ is an integer. We can express the result in a form similar to Eq. (12) with ${\tilde{I}}_{\lambda,\pm\mu}=\frac{2}{(2n-1)!!}\ \bigl{(}\frac{a}{b}\bigr{)}^{\lambda}\ (1\pm\frac{d}{d\xi_{b}})^{\mu}\ \xi_{b}^{n}K_{n}(\xi_{b}).$ $None$ To evaluate this expression one can make use of the relations: $\frac{d}{dx}(x^{n}K_{n}(x))=-x^{n}K_{n-1}(x)$. For dipole and quadrupole excitations one obtains ${\tilde{I}}_{1,\pm 1}=2\xi_{a}\ \Bigl{[}K_{1}(\xi_{b})\mp K_{0}(\xi_{b})\Bigr{]},$ $None$ ${\tilde{I}}_{20}=\xi_{a}^{2}\Bigl{[}K_{2}(\xi_{b})-K_{0}(\xi_{b})\Bigr{]}=2\ (\frac{a}{b})^{2}\ \xi_{b}\ K_{1}(\xi_{b}),$ $None$ ${\tilde{I}}_{2\pm 2}=\frac{1}{3}\xi_{a}^{2}\Bigl{[}K_{2}(\xi_{b})\mp 4K_{1}(\xi_{b})+3K_{0}(\xi_{b})\Bigr{]},$ $None$ where the relation $K_{2}(x)=K_{0}(x)+2/xK_{1}(x)$ has been used in Eq. (A5). To improve the straight-line approximation, one can multiply the results, Eqs. (A4-A6), by the factor $\exp(-\pi\xi_{a}/2)$, according to Winther and Alder AW-NPA . In fact, the exact analytic expression, which has been obtained for a Coulomb trajectory in the case of dipole excitations, supports this suggestion; see Eq. (12) in appendix H of Ref. AW-book . We will later on investigate how good this improvement and other approximations are for quadrupole excitations. ## IX Appendix B: Relativistic kinematics Here we specify the expressions that are used to calculate the relativistic Coulomb scattering. Most of them are taken from Jackson’s book jackson . First of all, the kinetic energy T of the projectile is commonly given in units of MeV/nucleon so the $\gamma$ factor and the beam velocity $v$ can be obtained from $\gamma m=m+T,\ \ \ \beta=v/c=\frac{\sqrt{T(2m+T)}}{m+T},$ $None$ where $m$ = 931.5 MeV is the nucleon mass (using the notation $c$=1.) The masses of projectile and target are denoted by $M_{1}$ and $M_{2}$, and the total energy in the center of mass system is (Jackson (12.31)) $E^{\prime}=\sqrt{M_{1}^{2}+M_{2}^{2}+2\gamma M_{1}M_{2}}.$ where $\gamma M_{1}$ is the laboratory energy of the projectile. Moreover, the energy and momentum of the projectile in the center of mass system are (Jackson, Eqs. (12.31-34)) $E_{1}^{\prime}=\frac{M_{1}^{2}+E_{1}M_{2}}{E^{\prime}}=\frac{M_{1}(M_{1}+\gamma M_{2})}{E^{\prime}},$ $p_{1}^{\prime}=\frac{M_{2}}{E^{\prime}}p=\frac{M_{1}M_{2}}{E^{\prime}}\ \gamma v.$ where $p=\gamma M_{1}v$ is the momentum of the projectile in the laboratory frame. Note that the non-relativistic reduced mass $M_{1}M_{2}/(M_{1}+M_{2})$ has been replaced in the last expression by the relativistic reduced mass, $M_{0}=\frac{M_{1}M_{2}}{E^{\prime}}=\frac{M_{1}M_{2}}{\sqrt{M_{1}^{2}+M_{2}^{2}+2\gamma M_{1}M_{2}}}.$ $None$ Rutherford’s scattering formula in the center-of-mass system is derived from the transverse momentum transfer in elastic Coulomb scattering, estimated in the straight line approximation by $\Delta p_{\perp}=p_{1}^{\prime}\ \sin(\theta)=\frac{2Z_{1}Z_{2}e^{2}}{vb},$ This is a reasonable approximation in high-energy forward-angle scattering but to match the non-relativistic expression, $\tan(\theta/2)=a/b$, one should consider the Coulomb distortion of the trajectory. This would give a factor of $\cos^{2}(\theta/2)$ on the right-hand side, so we obtain $\tan(\theta/2)=\frac{Z_{1}Z_{2}e^{2}}{p_{1}^{\prime}vb}=\frac{Z_{1}Z_{2}e^{2}}{\gamma M_{0}v^{2}b}.$ Thus we recover the usual scattering formula, $\tan(\theta/2)=a/b$, but the definition of $a$, Eq. (4), must be replaced by $a=\frac{Z_{1}Z_{2}e^{2}}{\gamma M_{0}\beta^{2}}.$ $None$ There are two corrections compared to Eq. (4). One is the factor $1/\gamma$, which is commonly considered. The other is the relativistic reduced mass $M_{0}$, which is often replaced by the non-relativistic value. The scattering angle in the laboratory frame is determined by (see Jackson (12.50)) $\tan(\phi)=\frac{\sin(\theta)}{\gamma_{c.m.}(\cos(\theta)+\alpha)},$ $None$ where $\gamma_{c.m.}=(\gamma M_{1}+M_{2})/E^{\prime}$ (Jackson (12.35)), and $\alpha=\frac{M_{1}}{M_{2}}\ \frac{M_{1}+\gamma M_{2}}{\gamma M_{1}+M_{2}},$ $None$ according to Jackson (12.54) for elastic scattering. It is seen that the transformation from the c.m. to the laboratory system reduces to the usual expression for $\gamma\rightarrow 1$. ## References * (1) K. Alder and A. Winther, Electromagnetic Excitation (North-Holland, NY, 1975). * (2) A. Winther and K. Alder, Nucl. Phys. A319, 518 (1979). * (3) T. Glasmacher, Annual Rev. Nucl. Part. Sci. 48, 1 (1998). * (4) A. N. F. Aleixo and C. A. Bertulani, Nucl. Phys. A505, 448 (1989). * (5) C. A. Bertulani, A. E. Stuchbery, T. J. Mertzimekis, and A. D. Davies, Phys. Rev. C 68, 044609 (2003). * (6) C. A. Bertulani, G. Cardella, M. D. Napoli, G. Raciti, and E. Rapisarda, Phys. Lett. B 650, 233 (2007). * (7) H. Scheit, A. Gade, Th. Glasmacher, T. Motobayashi, Phys. Lett. B 659, 515 (2007). * (8) K. Alder, A. Bohr, T. Huus, B. Mottelson, and A. Winther, Rev. Mod. Phys. 28, 432 (1956). * (9) P. D. Cottle et al., Phys. Rev. C 64, 057304 (2001). * (10) Evaluated Nuclear Structure Data Files (ENSDF), National Nuclear Data Center, Brookhaven National Laboratory, http//:www.nndc.bnl.gov/ * (11) J. A. Church et al., Phys. Rev. C 72, 054320 (2005). * (12) H. Scheit et al., Phys. Rev. Lett 77, 3967 (1996). * (13) A. Gade et al., Phys. Rev. C 68, 014302 (2003). * (14) M. Fauerbach et al., Phys. Rev. C 56, R1-4 (1997). * (15) C. A. Bertulani (private communications). * (16) A. Goldberg, Nucl. Phys. A240, 636 (1984). * (17) I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products (Academic Press, NY, 1965). * (18) J. D. Jackson, Classical Electrodynamics (Wiley, NY, 1967). Table 1: Cross sections for the quadrupole excitation ($\lambda=2$) of different nuclei on a Au or Bi target, and the dipole excitation ($\lambda=1$) of 11Be (last row). The experimental conditions are from the quoted references. The $T$ is the beam energy per nucleon at mid-target and $\phi_{max}^{Lab}$ ($\theta_{max}^{c.m}$) is the maximum laboratory (center-of-mass) acceptance angle. The adopted B(E2) values NDS for 24Mg and 26Mg are also shown. The the last two columns show the calculated cross sections for the relativistic Coulomb ($\sigma_{Coul}$) and straight-line ($\sigma_{Strl}$) trajectories. run \| \| Ex \| B(E$\lambda$) \| T \| $\phi_{max}^{Lab}$ \| $\sigma_{exp}$ \| $\sigma_{Coul}$ \| $\sigma_{Strl}$ ---\|---\|---\|---\|---\|---\|---\|---\|--- no. \| Nucleus \| (MeV) \| (e2fm2λ) \| (MeV/nucleon) \| (deg) \| (mb) \| (mb) \| (mb) 0 \| 24Mg+Au simg \| 1.3687 \| 467(28) \| 36 \| $\theta_{max}^{c.m.}\leq$ 4.48 \| 78.7(48) \| 81.7 \| 88.0 \| adopted NDS \| \| 436(10) \| \| \| \| 76.3(18) \| 82.2 0 \| 26Si+Au simg \| 1.7959 \| 336(33) \| 41.8 \| $\theta_{max}^{c.m.}\leq$ 4.48 \| 55.8(55) \| 56.3 \| 61.0 1 \| 26Mg+Bi mg \| 1.8087 \| 315 \| 67 \| 2.38 \| 44(2) \| 45.9 \| 47.8 \| adopted NDS \| \| 307(9) \| \| \| \| 44.7(13) \| 46.6 2 \| 32Mg+Au mg \| 0.885 \| 447 \| 71.2 \| 2.26 \| 91(10) \| 92.7 \| 95.9 3 \| 34Mg+Bi mg \| 0.659 \| 541 \| 67 \| 2.38 \| 126(22) \| 130.1 \| 134.9 4 \| 38S+Au sar \| 1.292 \| 235 \| 34.6 \| 4.10 \| 59(7) \| 60.1 \| 64.7 5 \| 40S+Au sar \| 0.891 \| 334 \| 35.3 \| 4.10 \| 94(9) \| 96.9 \| 103.4 6 \| 42S+Au sar \| 0.890 \| 397 \| 36.6 \| 4.10 \| 128(19) \| 131.1 \| 139.9 7 \| 44Ar+Au sar \| 1.144 \| 345 \| 30.9 \| 4.10 \| 81(9) \| 83.0 \| 89.6 8 \| 46Ar+Au sar \| 1.554 \| 196 \| 32.8 \| 4.10 \| 53(10) \| 53.6 \| 58.3 9 \| 46Ar+Au ar46 \| 1.554 \| 212 \| 73.2 \| 2.90 \| 68(8 ) \| 68.6 \| 71.9 \| 11Be+Au be11 \| 0.32 \| 0.079 \| 57.6 \| 3.80 \| 244(31) \| 246.1 \| 247.2 Figure 1: The adiabaticity functions $F_{\lambda}(b/a,\xi)$ are illustrated for $\lambda$ = 1 and 2. The top solid curves are the results for a straight line trajectory, Eq. (23). The dashed curves are obtained from Coulomb trajectories, Eq. (22), using the indicated values of $b/a$. The solid points were obtained by multiplying the straight line trajectory with $exp(-\pi\xi_{a})$ = $exp(-\pi\frac{a}{b}\xi_{b})$ for $a/b$=5. Figure 2: Energy dependence of the non-relativistic cross sections for the dipole excitation of 11Be and the quadrupole excitation of 42S discussed in the text. The target is gold, and the minimum impact parameter was set to 14 fm in both cases. The upper thick dashed curves show the straight-line trajectory approximation, whereas the thin dashed curves have been corrected with the factor $\exp(-\pi\xi_{a})$. The solid curves are the results for Coulomb trajectories. Figure 3: Energy dependence of the relativistic cross sections for the dipole excitation of 11Be and the quadrupole excitation of 42S discussed in the text. The target is Au, and the minimum impact parameter was set to 14 fm in both cases. The thick dashed curves show the straight-line trajectory approximation, Eqs. (29,30). The solid curves show the interpolating, relativistic Coulomb excitation cross section, Eq. (26). The thin dashed curves are also based on Eq. (26) but use $\xi_{a}=\omega a/v$ in the second argument of $K_{\lambda\mu}^{eff}(b/a,\xi_{a})$. Figure 4: Ratio of measured and calculated Coulomb excitation cross sections for the different runs shown in Table I. Shown are the results for the interpolated relativistic Coulomb excitation (solid points) and the straight-line trajectory calculations (triangles). Figure 5: Cross section for the Coulomb excitation of the $2^{+}$ state in 46Ar on a Au target at 73.2 MeV/nucleon as a function of the maximum laboratory acceptance angle $\phi_{max}$. The relativistic straight-line and Coulomb trajectory calculations are compared to the data ar46 . Figure 6: (A): cross sections for the $2^{+}$ excitation of 46Ar on a Pb target. The minimum distance of closest approach was set to 11.41 fm. The relativistic straight-line (upper dashed curve) and Coulomb trajectory calculations (solid curve) are compared to the non-relativistic Coulomb trajectory calculation (lower dashed curve), and to the results of Ref. bertpl (solid points) which have been multiplied by 1.44. (B): ratios of the cross sections in (A) to the relativistic Coulomb trajectory calculation. The triangles show the estimate, Eq. (33). Figure 7: Ratios of different cross sections to the relativistic Coulomb trajectory calculation. Results are shown for the excitation of the $2^{+}$ state in 16N and 54Ni, respectively, on a Pb target. The thick solid curves are the ratios for the straight line trajectory approximation. The dashed curves show various ways of correcting the straight-line approximation, such as multiplying with the factor $\exp(-\pi\xi_{a})$, or using the effective minimum impact parameters discussed in the text. The solid points are the ratios for the cross sections given in table 2 of Ref. bertpl multiplied by 1.44.	arxiv-papers	2008-08-04T15:37:56	2024-09-04T02:48:57.158933	{ "license": "Public Domain", "authors": "H. Esbensen", "submitter": "Henning Esbensen", "url": "https://arxiv.org/abs/0808.0361" }
0808.0394	# Magnetoplasmon spectrum of a weakly modulated two-dimensional electron gas system M. Tahir1,∗, K. Sabeeh2,† and A. MacKinnon1 1Department of Physics, Blackett 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 The magnetoplasmon spectrum of a magnetically modulated two-dimensional electron gas (MM2DEG) is investigated. We derive the inter and intra Landau band magnetoplasmon spectrum within the self consistent field approach. The derivation is performed at zero temperature as well as at finite temperature. Results are presented for the inter and intra Landau band magnetoplasmon spectrum as a function of the inverse magnetic field. As a result of magnetic modulation, magnetic Weiss oscillations are found to occur in the magnetoplasmon spectrum. Furthermore, our finite temperature theory facilitates analysis of effects of temperature on the magnetoplasmon spectrum. The results are compared with those obtained for an electrically modulated 2DEG system. In addition, we derive and discuss the effects of simultaneous electric and magnetic modulations on the magnetoplasmon spectrum of 2DEG when the modulations are in phase as well as when they are out of phase. Magnetic 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 in the former case while we find complete suppression of Weiss oscillations for particular relative strength of the modulations in the latter case. ###### pacs: 72.80.Rj, 71.45.Gm, 73.21.-b Date text]date ## I Magnetoplasmon spectrum of magnetically modulated two-dimensional electron gas (MM2DEG) In the past two decades, remarkable progress has been made in epitaxial crystal growth techniques which have made possible the fabrication of novel semiconductor heterostructures. These modern microstructuring techniques can be used to laterally confine quasi-two-dimensional electron gas (2DEG) in e.g.,GaAs/AlGaAs heterostructure on a submicrometer scale. Furthermore, magnetic modulation of these systems can be realized by depositing an array of ferromagnetic strips on top of the heterostructure or by superconducting layers beneath the substrate. These magnetically modulated 2DEG systems realized in semiconductor heterostructures have attracted a lot of attention in the past and continue to do so today due to their novel properties1 ; 2 ; 3 ; 4 ; 5 ; 6 ; 7 ; 8 ; 9 ; 10 ; 11 ; 12 ; 13 ; 14 ; 15 ; 16 ; 17 ; 18 ; 19 ; 20 ; 21 ; 22 ; 23 ; 24 ; 25 ; 26 ; 27 ; 28 . Primarily, it is due to the introduction of another length scale in the system, the period of modulation, as a result new phenomena arise due to commensurability of the period of modulation and the other characteristic length scale of the system, cyclotron diameter at the Fermi level. Resistivity measurements on these systems are found to exhibit commensurability oscillations (magnetic Weiss oscillations) since the magnetic modulation broadens the Landau levels into minibands whose width oscillates as a function of the magnetic field. These oscillations are periodic as a function of the inverse magnetic field with a larger period than that of the Shubnikov-de Hass (SdH) oscillations. The period of Weiss oscillations depends on both the modulation period and the square root of the number density of the MM2DEG, in contrast to the linear dependence on the number density of the SdH oscillations. Moreover, the amplitude of Weiss oscillations is weakly affected by temperature as compared to SdH oscillations. We investigate the effects of magnetic modulation on the collective excitations (magnetoplasmons) of a 2DEG. Plasmons are a very general phenomenon and have been studied extensively in a wide variety of systems including ionized gases, simple metals and semiconductor 2DEG systems. In a 2DEG, these collective excitations are induced by the electron-electron interactions. These collective excitations, plasmons, are among the most important electronic properties of a system. In the presence of an external magnetic field, these collective excitations are known as magnetoplasmons. Magnetic oscillations of the plasmon frequency occur in a magnetic field. Single particle magneto-oscillatory phenomena such as the Shubnikov-de Haas and de Haas-van Alphen effects have provided very important probes of the electronic structure of solids. Their collective analog yields important insights into collective phenomena. For this reason we study the effects of magnetic modulation on the collective excitations (magnetoplasmons) in a 2DEG. We present the magnetoplasmon spectrum of a MM2DEG in the presence of a perpendicular magnetic field using the self consistent field approach. Both the inter and intra Landau band magnetoplasmons are determined. Inter Landau band magnetoplasmons arise due to electronic transitions between different Landau bands whereas intra Landau band magnetoplasmons are a result of electronic transitions in a single Landau band. We evaluate the dynamic, nonlocal density-density response function to obtain these results. The collective excitations in a MM2DEG system have been investigated in the past12 ; 16 ; 20 ; 21 ; 24 . All of these studies primarily investigate the inter Landau band magnetoplasmon mode. Reference16 is an experimental study where the inter Landau band magnetoplasmons are investigated by infrared optical measurements whereas 12 ; 20 ; 21 ; 24 are theoretical studies. In 12 only the primary inter Landau band magnetoplasmon was considered. Far-infrared absorption of MM2DEG was theoretically investigated in 20 to explore plasma oscillations in the system. Similarly plasma oscillations were also theoretically investigated in 21 ; 24 taking into account primarily the inter Landau band magnetoplasmons. Since the existence of an intra-Landau band magnetoplasmon is a result of the finite Landau bandwidth caused by modulation, the study of collective excitations and in particular modulation induced effects in this system requires taking into account the intra Landau band magnetoplasmons. To overcome this shortcoming in previous work mentioned above, we determine both the intra and inter-Landau band magnetoplasmon spectrum in this work. In addition, our finite temperature theory facilitates the analysis of effects of temperature on the Weiss & SdH oscillations in the magnetoplasmon spectrum of MM2DEG. Furthermore, we carry out a detailed comparison of phase and amplitude of magnetic Weiss oscillations and the electric Weiss oscillations in a 2DEG. We also present the effects of simultaneous electric and magnetic modulations on the magnetoplasmon spectrum of a 2DEG. To the best of our knowledge, the complete study of both the inter and intra Landau band magnetoplasmon spectrum in this system and comparison of these with the results for the electrically modulated two-dimensional electron gas system (EM2DEG) 29 ; 30 ; 31 ; 32 has not been carried out so far. To this end, we have undertaken the present study. In section II, we present the formulation of the problem. Section III contains the magnetoplasmon spectrum of a MM2DEG and comparison with an EM2DEG at zero temperature ($T=0$) whereas in section IV we discuss the temperature dependence of magnetoplasmon spectrum of a MM2DEG and its comparison with an EM2DEG including the asymptotic description. In the following two sections, magnetoplasmon spectrum in the presence of simultaneous electric and magnetic modulations is presented. We discuss the effects of in-phase modulations (electric and magnetic) in section V while out of phase is discussed in section VI. Concluding remarks are made in section VII. ## II Formulation The system that we are considering is a 2DEG in the presence of a perpendicular magnetic field that is modulated weakly and periodically along one direction. We take the magnetic field $B$ to be perpendicular to the $x-y$ plane in which electrons with unmodulated areal density $n_{D}$, effective mass $m^{\ast}$ and charge $-e$ are confined. We employ the Landau gauge and write the vector potential as $A=(0,Bx+(B_{0}/K)\sin Kx,0)$, where $K$ is $2\pi/a,$ $a$ is the period of modulation and $B_{0}$ is the magnetic modulation strength such that $B_{0}\ll B.$ The Hamiltonian in the Landau gauge is5 ; 8 ; 26 $H_{0}=\frac{1}{2m^{\ast}}[-\hslash^{2}\frac{\partial^{2}}{\partial x^{2}}+(-i\hslash\frac{\partial}{\partial y}+eBx+(eB_{0}/K)\sin Kx)^{2}].$ (1) Since the Hamiltonian does not depend on the $y$ coordinate, the unperturbed wavefunctions are plane waves in the the $y$-direction. This allows us to write for the wavefunctions, $\phi_{nk_{y}}(\bar{x})=\frac{1}{\sqrt{L_{y}}}e^{ik_{y}y}u_{n}(x),$ (2) with $L_{y}$ being a normalization length in $y$-direction and $\bar{x}$ a 2D position vector in the $x$-$y$ plane. The hamiltonian in equation (1) can be expressed as $\displaystyle H_{0}$ $\displaystyle=-\frac{\hslash^{2}}{2m^{\ast}}\frac{\partial^{2}}{\partial x^{2}}+\frac{1}{2}m^{\ast}\omega_{c}^{2}(x-x_{0})^{2}$ $\displaystyle+(\omega_{0}/K)(p_{y}+eBx)\sin Kx+(m^{\ast}\omega_{0}^{2}/4K^{2})[1-\cos 2Kx],$ (3) where $\omega_{c}=\frac{eB}{m^{\ast}}$ is the cyclotron frequency, $\omega_{0}=eB_{0}/m^{\ast}$ is the modulation frequency, $x_{0}=-l^{2}k_{y}=-\frac{\hslash k_{y}}{m^{\ast}\omega_{c}}$ is the coordinate of cyclotron orbit center, $l=\sqrt{\frac{\hslash}{m^{\ast}\omega_{c}}}$ is the magnetic length, and $m^{\ast}$ is the effective mass. We can write the unmodulated eigenstates in the form $\phi_{nk_{y}}(\bar{x})=\frac{1}{\sqrt{L_{y}}}e^{ik_{y}y}u_{n}(x;x_{0}),$ with $u_{n}(x;x_{0})=(\sqrt{\pi}2^{n}n!l)^{\frac{-1}{2}}\exp(-\frac{1}{2l^{2}}(x-x_{0})^{2})H_{n}(\frac{x-x_{0}}{l}),$ where $u_{n}(x;x_{0})$ is a normalized harmonic oscillator wavefunction centered at $x_{0}$ and $H_{n}(x)$ are Hermite polynomials with $n$ the Landau level quantum number33 . 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 $\varepsilon(n,x_{0})=(n+1/2)\hslash\omega_{c}+G_{n}\cos(\frac{2\pi}{a}x_{0})$ (4) where $G_{n}=\hslash\omega_{0}\exp(-u/2)[L_{n}(u)/2+L_{n-1}^{1}(u)],u=\frac{K^{2}l^{2}}{2}=(\frac{2\pi}{a})^{2}\frac{\hslash}{2m^{\ast}\omega_{c}}$ and $L_{n}(u),$ $L_{n-1}^{l}(u)$ are Laguerre and associated Laguerre polynomials. This result has been obtained previously5 ; 8 . The above equation shows that the formerly sharp Landau levels are now broadened into minibands by the modulation potential. Furthermore, the Landau bandwidth ($\sim\mid G_{n}\mid$) oscillates as a function of $n$, since $L_{n}(u)$ is an oscillatory function of its index33 . Hereafter, we employ the Ehrenreich-Cohen Self-Consistent Field (SCF) approach to determine the density-density response function34 . Following the SCF approach, $\Pi_{0}(\bar{q},\omega)$ is the density-density response function of the non-interacting electron system, given by $\Pi_{0}(\bar{q},\omega)=\frac{1}{A}\underset{n,n^{\prime}}{\sum}\underset{k_{y}}{\sum}C_{nn^{\prime}}(\frac{\hslash\bar{q}^{2}}{2m^{\ast}\omega_{c}})\frac{f(\varepsilon(n^{\prime},k_{y}-q_{y}))-f(\varepsilon(n,k_{y}))}{\varepsilon(n^{\prime},k_{y}-q_{y})-\varepsilon(n,k_{y})+\hslash\omega+i\hslash\eta},$ (5) where $C_{nn^{\prime}}(x)=\frac{n_{2}!}{n_{1}!}e^{-x}x^{n_{1}-n_{2}}[L_{n_{2}}^{n_{1}-n_{2}}(x)]^{2}$ with $n_{1}$= max($n$, $n^{\prime}$), $n_{2}$= min($n$, $n^{\prime}$), $f(\varepsilon)$ is the Fermi-Dirac distribution function, $\bar{q}$ is the 2D wave number, $A$ is the area of the system, $x=\frac{\hslash\bar{q}^{2}}{2m^{\ast}\omega_{c}}$ and $L_{n}^{l}(x)$ are associated Laguerre polynomials. The density-density response function of the interacting system can be expressed as $\Pi(\bar{q},\omega)=\frac{\Pi_{0}(\bar{q},\omega)}{1-v_{c}(\bar{q})\Pi_{0}(\bar{q},\omega)}$ (6) with $v_{c}(\bar{q})=$ $\frac{2\pi e^{2}}{k\bar{q}}$ is the 2-D Coulomb potential, $k$ is the background dielectric constant. Using the transformation $k_{y}\rightarrow-k_{y},$ realizing that $\varepsilon(n,k_{y})$ is an even function of $k_{y},$ interchanging $n\leftrightarrow n^{\prime}$ we can write the non-interacting density-density response function given by equation (5) as $\displaystyle\Pi_{0}(\bar{q},\omega)$ $\displaystyle=\frac{m^{\ast}\omega_{c}}{\pi\hslash a}\underset{n,n^{\prime}}{\sum}C_{nn^{\prime}}(\frac{\hslash\bar{q}^{2}}{2m^{\ast}\omega_{c}})\overset{a}{\underset{0}{\int}}dx_{0}[f(\varepsilon(n,x_{0}+x_{0}^{\prime})-f(\varepsilon(n^{\prime},x_{0}))]$ $\displaystyle\times[\varepsilon(n,x_{0}+x_{0}^{\prime})-\varepsilon(n^{\prime},x_{0})+\hslash\omega+i\hslash\eta]^{-1}.$ (7) In writing the above equation we converted the $k_{y}$-sum into an integral over $x_{0}$ and $x_{0}^{\prime}=-\frac{\hslash q_{y}}{m^{\ast}\omega_{c}}$. ## III Magnetoplasmons modes of MM2DEG and comparison with EM2DEG at zero temperature The plasma modes are obtained from the roots of the longitudinal dispersion relation from Eqs. (6, 7) $1-v_{c}(\bar{q})\operatorname{Re}\Pi_{0}(\bar{q},\omega)=0$ (8) along with the condition $Im\Pi_{0}(\bar{q},\omega)=0$ to ensure long-lived excitations. The roots of Eq. (8) give the plasma modes of a MM2DEG as $1=\frac{2\pi e^{2}}{\kappa\bar{q}}\frac{2}{\pi l^{2}}\underset{n,n^{\prime}}{\sum}C_{nn^{\prime}}\left(x\right)\\{I_{1}(n,n^{\prime},x_{0}^{\prime};\omega)+I_{1}(n,n^{\prime},x_{0}^{\prime};-\omega)\\},$ (9) with $I_{1}(n,n^{\prime},x_{0}^{\prime};\omega)=P\overset{a}{\underset{0}{\int}}dx_{0}\frac{f(\varepsilon(n,x_{0}))}{\varepsilon(n^{\prime},x_{0}+x_{0}^{\prime})-\varepsilon(n,x_{0}+x_{0}^{\prime})+\hslash\omega}.$ (10) and P is the principal value. From here on we will only show the dependence of $I_{1}$ on $\omega$ and suppress the rest such that $I_{1}(n,n^{\prime},x_{0}^{\prime};\omega)\rightarrow I_{1}(\omega).$ SdH and Weiss oscillations are found to occur in the magnetoconductivity of both electric and magnetically modulated 2DEG. These transport measurements can be explained without taking into account electron-electron interactions. In order to investigate collective excitations of the system such as magnetoplasmons it is essential to consider electron-electron interactions. Magnetoplasmons arise due to the coherent motion of electrons as a result of electron-electron interactions. Two types of magnetoplasmons can be identified: Those arising from electronic transitions involving different Landau bands (inter Landau band plasmons) and those within a single Landau band (intra Landau band plasmons). Inter-Landau band plasmons involve the local 1D magnetoplasma mode and the Bernstein-like plasma resonances35 ; 36 , all of which involve excitation energies greater than the Landau-band separation ($\sim\hslash\omega_{c}$). On the other hand, intra-Landau band magnetoplasmons resonate at energies comparable to the bandwidths, and the existence of this new class of modes is due to finite width of the Landau levels. In a MM2DEG considered here, the Landau bandwidth ($\sim\mid G_{n}\mid$) oscillates as a function of the band index $n,$ since $L_{n}(u),$ $L_{n}^{l}(u)$ are oscillatory functions of the index $n.$ Such oscillating bandwidths affect the plasmon spectrum of the intra-Landau band type, when Landau-band separation is larger than the bandwidth as is the case considered here, resulting in magnetic Weiss oscillations similar to the electric Weiss oscillations found in the electrically modulated system. These oscillations are accompanied by SdH type of oscillatory behavior29 ; 37 . Both these oscillations are periodic as a function of inverse magnetic field ($1/B$) but occur with different periods and amplitudes. As we show below, Weiss oscillations in the magnetoplasmon spectrum of a MM2DEG differ in phase and amplitude with those of an EM2DEG29 . We now examine the inter-Landau-band transitions. These transitions occur between different Landau bands so we consider $n$ $\neq n^{\prime}$in Eq.(10) which yields $I_{1}(\omega)=\frac{f(\varepsilon(n))}{(\hslash\omega-\Delta)},$ where $\Delta=\left(\varepsilon(n)-\varepsilon(n^{\prime})\right)$ with $\varepsilon(n)=(n+\frac{1}{2})\hslash\omega_{c}$, which permits us to write the following term in Eq (9) as $(I_{1}(\omega)+I_{1}(-\omega))=2\frac{\Delta f(\varepsilon(n))}{(\hslash\omega)^{2}-(\Delta)^{2}}.$ (11) Next, we consider the coefficient $C_{nn^{\prime}}(x)$ in Eq.(9) and expand it to lowest order in its argument (low wave-number expansion). In this case, we are only considering the $n^{\prime}=n\pm 1$ terms. The inter-Landau band plasmon modes under consideration arise from neighboring Landau bands. Hence for $n^{\prime}=n+1$ and $x\ll$ 1, using the following associated Laguerre polynomial expansion38 $L_{n}^{{}^{l}}(x)={\displaystyle\sum\limits_{m=0}^{n}}(-1)^{m}\frac{(n+l)!}{(l+m)!(n-m)!}\frac{x^{m}}{m!}$ for $l>0$ and retaining the first term in the expansion for $x\ll$ 1, $C_{nn^{\prime}}(x)$ reduces to $C_{n,n+1}(x)\rightarrow(n+1)x,$ (12) and for $n^{\prime}=n-1$ and $x\ll 1,$ it reduces to $C_{n,n-1}(x)\rightarrow nx.$ (13) Substitution of equations (11) and (12, 13) into equation (9) and replacing $x=\frac{\hslash\bar{q}^{2}}{2m^{\ast}\omega_{c}}$ yields $1=\frac{2\pi e^{2}}{km^{\ast}}\bar{q}\frac{1}{\omega^{2}-\omega_{c}^{2}}(\frac{m\omega_{c}}{\pi\hslash}\underset{n}{\sum}f(\varepsilon(n)).$ The term in parenthesis is easily recognized as the unmodulated particle density $n_{D}=\frac{m\omega_{c}}{\pi\hslash}\underset{n}{\sum}f(\varepsilon(n)),$ where the summation is over all occupied Landau bands. Defining the plasma frequency through $\omega_{p,2D}^{2}=\frac{2\pi n_{D}e^{2}}{km}\bar{q},$ we obtain the inter-Landau-band plasmon dispersion relation $1=\frac{\omega_{p,2D}^{2}}{\omega^{2}-\omega_{c}^{2}}$ or $\omega^{2}=\omega_{c}^{2}+\omega_{p,2D}^{2},$ (14) here $\omega$ is the well-known local 2D principal plasma frequency. Since we are interested in the modulation induced effects on the magnetoplasmons in this system, our focus will be on the intra-Landau band magnetoplasmons as the dispersion relation for inter-Landau band magnetoplasmons have been discussed and results displayed in 12 ; 29 . For the intra-Landau-band excitations spectrum, we need to consider transitions within a Landau miniband, i.e. $n=n^{\prime}$, $\varepsilon(n^{\prime})-\varepsilon(n)=0$ and $C_{nn^{\prime}}(x)\rightarrow 1$. An analytical expression of the intra-Landau band plasmon energy $\hslash\overset{\sim}{\omega}$ can be obtained $\hslash^{2}\overset{\sim}{\omega}^{2}=\frac{16e^{2}}{k\bar{q}\pi}\frac{m^{\ast}\omega_{c}}{\hslash a}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)\times A_{n},$ (15) where $A_{n}=\underset{n}{\sum}G_{n}\times\int_{0}^{a/2}dx_{0}f(\varepsilon(n,x_{0}))\cos(Kx_{0}).$ At zero temperature ($T=0$), $\hslash^{2}\overset{\sim}{\omega}^{2}=\frac{8e^{2}}{k\bar{q}}\frac{m^{\ast}\omega_{c}}{\pi\hslash}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)\times\underset{n}{\sum\mid}G_{n}\mid\sqrt{1-\Delta_{n}^{2}}\theta(1-\Delta_{n}),$ (16) with $\Delta_{n}=\mid\frac{\varepsilon_{F}-\varepsilon(n)}{G_{n}}\mid,\theta(x)$ the Heaviside unit step function. If we replace the magnetic modulation term ($G_{n}$) by the electric one, Eq. (16) has the same structure as Eq. (8) of 29 that pertains to electric modulation. The above expression for $\hslash\overset{\sim}{\omega}$ has been obtained under the condition $\hslash\omega>>\mid\varepsilon(n,x_{0}+x_{0}^{\prime})-\varepsilon(n,x_{0})\mid$ as $x_{0}^{\prime}\rightarrow 0$ which leads to a relation between the energy and the Landau level broadening $\hslash\omega>>\mid 2G_{n}\sin(\frac{\pi}{a}x_{0}^{\prime})\sin[(\frac{2\pi}{a})(x_{0}+\frac{x_{0}^{\prime}}{2})]\mid$. This ensures that $Im\Pi_{0}(\bar{q},\omega)=0$ and the intra-Landau-band magnetoplasmons are undamped. For a given $G_{n}$, this can be achieved with a small but nonzero $q_{y}$ (recall that $x_{0}^{\prime}=-\frac{\hslash q_{y}}{m^{\ast}\omega_{c}}$). In general, the inter- and intra-Landau-band modes are coupled for arbitrary magnetic field strengths. The general dispersion relation is : $1=\frac{\omega_{p,2D}^{2}}{\omega^{2}-\omega_{c}^{2}}+\frac{\overset{\sim}{\omega}^{2}}{\omega^{2}}.$ This equation yields two modes which are given by $\displaystyle\omega_{\pm}^{2}$ $\displaystyle=\frac{1}{2}(\omega_{c}^{2}+\omega_{p,2D}^{2}+\overset{\sim}{\omega}^{2})\pm\frac{1}{2}\\{(\omega_{c}^{2}+\omega_{p,2D}^{2}+\overset{\sim}{\omega}^{2}+2\omega_{c}\overset{\sim}{\omega})$ $\displaystyle\times(\omega_{c}^{2}+\omega_{p,2D}^{2}+\overset{\sim}{\omega}^{2}-2\omega_{c}\overset{\sim}{\omega})\\}^{1/2}$ which reduces to $\omega_{+}^{2}=\omega_{c}^{2}+\omega_{p,2D}^{2},$ and $\omega_{-}^{2}=\overset{\sim}{\omega}^{2}$ with corrections of order $\overset{\sim}{\omega}^{2}/\omega_{c}^{2}$ and $\overset{\sim}{\omega}^{2}/\omega_{p,2D}^{2}$. So long as $\mid G_{n}\mid<\hslash\omega_{c},$ mixing of the inter-and intra-band modes is small. Only the intra-Landau-band mode ($\hslash\overset{\sim}{\omega})$ will be excited in the frequency regime $\hslash\omega_{c}>\hslash\omega\sim$ $\mid G_{n}\mid.$ The intra-Landau-band plasma energy given by equation (11) is shown graphically in Fig.(1) as a function of $1/B$. The parameters used are 5 ; 7 ; 8 ; 26 ; 29 ; 30 ; 31 : $m^{\ast}=0.07m_{e}$, $k=12$, $n_{D}=3.16\times 10^{15}$ m-2, $a=380$ nm. We also take $q_{x}=0$ and $q_{y}=0.01k_{F},$ with $k_{F}=(2\pi n_{D})^{1/2}$ being the Fermi wave number of the unmodulated 2DEG in the absence of magnetic field. In numerical evaluation we have taken the sum over thirty Landau levels which ensures convergence of numerical results. Numerical evaluation of the dispersion relation was performed in Mathematica. In Fig(1), modulation induced oscillations in the intra-Landau band mode are apparent, superimposed on SdH-type oscillations. These oscillations are periodic as a function of inverse magnetic field ($1/B$). To gain further insight into the results presented in Fig.(1), we consider equation (16). In the regime $\hslash\omega_{c}>\left\|G_{n}\right\|$, the unit step function vanishes for all but the highest occupied Landau band, corresponding ,say, to the band index $N$. The sum over $n$ is trivial and plasma energy is given as $\hslash\overset{\sim}{\omega}=\mid G_{N}\mid^{1/2}(1-\Delta_{N}^{2})^{1/4}\theta(1-\Delta_{N})$. The analytic structure primarily responsible for the SdH type of oscillations is the function $\theta(1-\Delta_{N})$, which jumps periodically from zero (when the Fermi level is above the highest occupied Landau band) to unity (when the Fermi level is contained with in the highest occupied Landau band). On the other hand, the periodic modulation of the amplitude of the SdH type oscillations is due to the oscillatory nature of the factor $\mid G_{N}\mid^{1/2}$, which has been shown to exhibit commensurability oscillations 6 ; 19 ; 21 ; 31 . In the same figure, we also show the intra- Landau band plasma energy for the electrically modulated 2DEG29 . The magnetoplasmons spectrum of the MM2DEG at a specific Landau level is minimum when the corresponding spectrum for the EM2DEG is maximum, this confirms that magnetic oscillations in the two system are out of phase. To have comparable results for the two systems, at zero temperature, the strength of magnetic modulation potential, $\hslash\omega_{0},$ has to be $\sim 8.5$ times smaller than that of the electric modulation potential, $V_{0}$. This can be understood if we realize that this constraint arises due to the step function $\theta(1-\Delta_{N})$ appearing in the dispersion relation at zero temperature. Therefore, comparable results for the two systems when they are subjected to modulation of equal strength requires that we carry out a finite temperature calculation to avoid the constraint imposed by the step function at zero temperature. ## IV Temperature dependent magnetoplasmons mode of MM2DEG and comparison with EM2DEG For the finite temperature calculation of the intra-landau band plasma energy, we invoke the condition of weak modulation and perform the following expansion in equation (15) $f(\varepsilon(n,x_{0}))\simeq f(\varepsilon(n))+G_{n}f^{\prime}(\varepsilon(n))\cos(Kx_{0}),$ (17) where $f^{\prime}(x)=\frac{d}{dx}f(x)$ is the derivative of the Fermi-Dirac distribution function. With the substitution of this expansion in equation (15), the integral over $x_{0}$ contains two terms. The integral over $x_{0}$ of the first term vanishes and the integral of the second term yields intra- Landau band dispersion relation $\hslash^{2}\overset{\sim}{\omega}^{2}=\frac{4e^{2}m^{\ast}\omega_{c}}{k\bar{q}\pi\hslash}\sin^{2}[\frac{\pi}{a}(x_{0}^{\prime})]\times B_{n},$ (18) where $B_{n}=\underset{n}{\sum}G_{n}^{2}\times[-f^{\prime}(\varepsilon(n))]\ .$ To facilitate comparison of the above dispersion relation with the results obtained for an electrically modulated 2DEG we will determine the asymptotic expressions of intra-Landau band magnetoplasmon spectrum, where analytic results in terms of elementary functions can be obtained. Moreover, these asymptotic expressions will allow us to identify terms responsible for SdH and Weiss oscillations and how they are affected by temperature. The asymptotic expression can be obtained by using the following asymptotic expression for the Laguerre polynomials5 ; 7 ; 8 ; 26 $\exp^{-u/2}L_{n}(u)\rightarrow\frac{1}{\sqrt{\pi\sqrt{nu}}}\cos(2\sqrt{nu}-\frac{\pi}{4}).$ (19) Note that the above asymptotic expression for $L_{n}(u)$ is valid for $n\gg 1$, at low magnetic fields when many Landau Levels are filled. We now take the continuum limit: $n-->\frac{\varepsilon(n)}{\hslash\omega_{c}},\overset{\infty}{\underset{n=0}{{\displaystyle\sum}}}-->{\displaystyle\int\limits_{0}^{\infty}}\frac{d\varepsilon}{\hslash\omega_{c}}.$ (20) In the asymptotic limit, $B_{n}$ that appears in equation (18) can be written as $B_{n}=\frac{\hslash^{2}\omega_{0}^{2}}{\pi\hslash\omega_{c}}\sqrt{\frac{\hslash\omega_{c}}{u}}\left(\frac{aK_{F}}{2\pi}\right)^{2}{\displaystyle\int\limits_{0}^{\infty}}\frac{d\varepsilon}{\sqrt{\varepsilon}}\frac{\beta g(\varepsilon)}{[g(\varepsilon)+1)]^{2}}\sin^{2}\left(2\sqrt{nu}-\frac{\pi}{4}\right)$ (21) where $g(\varepsilon)=\exp[\beta(\varepsilon-\varepsilon_{F})],\beta=\frac{1}{K_{B}T}.$ Now assuming that temperature is low such that $\beta^{-1}\ll\varepsilon_{F}$ and replacing $\varepsilon=\varepsilon_{F}+s\beta^{-1}$, we can express the above integral as $B_{n}=\frac{\hslash^{2}\omega_{0}^{2}}{\pi\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\left(\frac{aK_{F}}{2\pi}\right)^{2}{\displaystyle\int\limits_{0}^{\infty}}ds\frac{e^{s}}{[e^{s}+1)]^{2}}\sin^{2}\left(2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}+\sqrt{\frac{u}{\hslash\omega_{c}\varepsilon_{F}}}s\beta^{-1}\right)$ with the result $B_{n}=\frac{\hslash^{2}\omega_{0}^{2}}{2\pi\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\left(\frac{aK_{F}}{2\pi}\right)^{2}\left[1-A\left(\frac{T}{T_{a}}\right)+2A\left(\frac{T}{T_{a}}\right)\sin^{2}\left[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right]\right]$ (22) where $T_{a}$ is the characteristic damping temperature of Weiss oscillationgiven by $k_{B}T_{a}=\frac{\hslash\omega_{c}aK_{F}}{4\pi^{2}},$ $\frac{T}{T_{a}}=\frac{4\pi^{2}k_{B}T}{\hslash\omega_{c}aK_{F}}$ and $A(x)=\frac{x}{\sinh(x)}-^{(x-->\infty)}->=2xe^{-x}.$ From equation(18), the asymptotic expression for intra-Landau band plasmon spectrum is obtained $\displaystyle\hslash^{2}\overset{\sim}{\omega}^{2}$ $\displaystyle=\frac{4\hslash^{2}\omega_{0}^{2}e^{2}m^{\ast}\omega_{c}}{k\bar{q}\hslash 2\pi^{2}\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\left(\frac{aK_{F}}{2\pi}\right)^{2}\sin^{2}\left[\frac{\pi}{a}(x_{0}^{\prime})\right]$ $\displaystyle\times\left[1-A\left(\frac{T}{T_{a}}\right)+2A\left(\frac{T}{T_{a}}\right)\sin^{2}\left(2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right)\right]$ (23) The above expression is not able to account for the SdH type of oscillations in the magnetoplasmon spectrum. These oscillations can be accounted for by expressing the density of states (in the absence of disorder)5 as $D(\varepsilon)=\frac{m^{\ast}}{\pi\hslash}\left(1-2\cos\left[\frac{2\pi\varepsilon}{\hslash\omega_{c}}\right]\right)$ (24) and inserting the continuum approximation as $\overset{\infty}{\underset{n=0}{{\displaystyle\sum}}}-->2\pi l^{2}{\displaystyle\int\limits_{0}^{\infty}}D(\varepsilon)d\varepsilon,$ this yields the asymptotic expression for the intra-Landau band magnetoplasmon dispersion relation for MM2DEG $\displaystyle\hslash^{2}\overset{\sim}{\omega}^{2}$ $\displaystyle=\frac{4\hslash^{2}\omega_{0}^{2}e^{2}m^{\ast}\omega_{c}}{k\bar{q}\hslash 2\pi^{2}\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\left(\frac{aK_{F}}{2\pi}\right)^{2}\sin^{2}\left[\frac{\pi}{a}(x_{0}^{\prime})\right]\times\\{[1-A\left(\frac{T}{T_{a}}\right)$ $\displaystyle+2A\left(\frac{T}{T_{a}}\right)\sin^{2}\left(2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right)]-4A\left(\frac{T}{T_{s}}\right)\cos\left[\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}\right]\sin^{2}\left[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right]\\}$ (25) where $\frac{T}{T_{s}}=\frac{2\pi^{2}k_{B}T}{\hslash\omega_{c}},$ $T_{s}$ defines the characteristic damping temperature of the SdH oscillations in the magnetoplasmon spectrum of MM2DEG. Following the same approach as discussed above for MM2DEG, we can obtain the intra-Landau band magnetoplasmon spectrum for EM2DEG $\hslash^{2}\overset{\sim}{\omega}^{2}=\frac{4e^{2}m^{\ast}\omega_{c}}{\hslash k\bar{q}\pi}\sin^{2}[\frac{\pi}{a}(x_{0}^{\prime})]\times B_{n},$ (26) where $B_{n}=\sum F_{n}^{2}\times[-f^{\prime}(\varepsilon(n))]$, and $F_{n}=V_{0}e^{-u/2}L_{n}(u)$ is the modulation width of the EM2DEG with $V_{0}$ the amplitude of electric modulation. The corresponding asymptotic result for EM2DEG is $\displaystyle\hslash^{2}\overset{\sim}{\omega}^{2}$ $\displaystyle=\frac{4V_{0}^{2}e^{2}m^{\ast}\omega_{c}}{k\bar{q}2\pi^{2}\hslash\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)\times\\{[1-A\left(\frac{T}{T_{a}}\right)$ $\displaystyle+2A\left(\frac{T}{T_{a}}\right)\cos^{2}\left(2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right)]-4A\left(\frac{T}{T_{s}}\right)\cos[\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}]\cos^{2}[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}]\\}$ (27) The intra-landau band plasmon dispersion relations obtained for the MM2DEG and the EM2DEG systems given by equations (25, 27) allow us to identify the terms responsible for Weiss and SdH oscillations. Moreover, the characteristic damping temperatures appearing in these expressions carry the effects of temperature on these oscillations. Comparing equations (25,27), the following differences can be highlighted: 1) Amplitude of the oscillations (Weiss and SdH) are larger by the factor $\frac{aK_{F}}{2\pi}$ in MM2DEG compared to those of EM2DEG. 2) The factor $\sin^{2}\left[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right]$ that appears in equation (25) for MM2DEG and the corresponding factor $\cos^{2}\left[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}\right]$ in equation (27) for EM2DEG results in a $\pi/2$ phase difference in the oscillations in the magnetoplasmon spectrum of the two systems Since equations (25, 27) are the key results of this work, we show the intra- Landau band magnetoplasmon energy for both the magnetically and electrically modulated systems as a function of inverse magnetic field in Fig.(2). The results presented are for equal strength of the two modulations which is taken to be $V_{0}=\hslash\omega_{0}=1$ meV. The temperature is $0.4$ K. Rest of the parameters are the same as the case for the zero temperature results presented in Fig.(1). We observe the modulation induced effects in the intra-Landau band mode, Weiss oscillations modulating the SdH oscillations in the magnetoplasmon spectrum. From the figure, we see that the amplitude of Weiss oscillations in the magnetically modulated system is greater by a factor of $\sim 8.5$ compared to the electrically modulated system. This can be seen from equations (25, 27) where the difference in the amplitudes is the factor $\frac{aK_{F}}{2\pi}$ and it is $\sim 8.5$ for the parameters that we have used. Therefore, in the magnetically modulated system the amplitude is larger by this factor compared to the electrically modulated system. We also observe that Weiss oscillations in the MM2DEG are out of phase by $\pi/2$ compared to those in EM2DEG. To see the effects of temperature on Weiss and SdH oscillations in a MM2DEG, we plot the intra-Landau band plasmon energy for a MM2DEG as a function of inverse magnetic field at two different temperatures in Fig.(3). The modulation strength is $1$ meV. The results are shown at the following two temperatures: $0.3$ K and $3$ K. The SdH oscillations are completely damped at $3$ K whereas Weiss oscillations persist at this temperature. Eqs.(25, 27) also allow us to determine the temperature scales for damping of Weiss and SdH oscillations in the magnetoplasmon spectrum. For a MM2DEG, from Eq.(25), $\frac{T_{a}}{T_{s}}=\frac{aK_{F}}{2}\gg 1$; e.g., $n_{D}=3.16\times 10^{15}$m-2 and $a=380$ nm, we have $\frac{T_{a}}{T_{s}}=27$ for the experimentally relevant parameters considered here. Hence, SdH oscillations are completely damped at a much lower temperature compared to Weiss oscillations. These results are consistent with and complement those obtained from electron transport studies of a magnetically modulated 2DEG 5 ; 6 ; 7 ; 8 ; 19 ; 20 ; 21 ; 26 ; 27 . ## V Magnetoplasmon spectrum with Periodic Electric and Magnetic Modulation: In-phase In this section, we calculate the magnetoplasmon spectrum when electric and magnetic modulations are in-phase. We take the magnetic modulation to have the same phase as given in the previous section with the in-phase electric modulation. The energy eigenvalues are5 ; 26 $\varepsilon(n,x_{0})=(n+1/2)\hbar\omega_{c}+(G_{n}+F_{n})\cos(Kx_{o})$ (28) and the bandwidth can be written as $\Delta(in- phase)=\frac{2\hslash\omega_{0}ak_{F}\times\sqrt{1+\delta^{2}}}{2\pi\sqrt{\pi\sqrt{nu}}}\times\sin\left(2\sqrt{nu}-\frac{\pi}{4}+\Phi\right)$ (29) where the ratio between the two modulation strengths $\delta=\frac{2\pi V_{0}}{\hslash\omega_{0}ak_{F}}=\tan(\Phi).$ 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 $n=n_{F}=\frac{\varepsilon_{F}}{\hbar\omega_{c}}-\frac{1}{2}$ is the highest Fermi integer. We see that the flat band condition in this case depends on the relative strength of the two modulations. Following the same approach as discussed in the previous section for the MM2DEG, we can obtain the intra-Landau band magnetoplasmon spectrum in the presence of in phase modulations as $\hslash^{2}\overset{\sim}{\omega}^{2}=\frac{4e^{2}m^{\ast}\omega_{c}}{k\bar{q}\pi\hslash}\sin^{2}[\frac{\pi}{a}(x_{0}^{\prime})]\times I_{n},$ (30) where $I_{n}=\underset{n}{\sum}(G_{n}+F_{n})^{2}\times[-f^{\prime}(\varepsilon(n))]\ .$ In Fig.(4) we show the in-phase magnetoplasmon spectrum (the magnetic and the electric modulations are in-phase) $\hslash\overset{\sim}{\omega}$ given by Eq.(30) as a function of the inverse magnetic field for temperature $T=0.3$ K, electron density $n_{e}=3\times 10^{11}$ cm${}^{-2},$ the period of modulation $a=380$ nm. The strength of the electric modulation $V_{0}=0.2$ meV whereas $B_{0}=0.02$ T which corresponds to $\hslash\omega_{0}=0.03$ meV. In the same figure we have also shown the magnetoplasmon spectrum 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 magnetoplason spectrum for electric and magnetic modulations as can be seen in the figure. To better understand the effects of in-phase modulations on the magnetoplasmon spectrum we consider the asymptotic expression of the magnetoplasmon spectrum given by Eq.(30). The asymptotic expression is given by $\displaystyle\hslash^{2}\overset{\sim}{\omega}^{2}$ $\displaystyle=\frac{2V_{0}^{2}e^{2}m^{\ast}\omega_{c}\delta^{-2}}{k\bar{q}2\pi^{2}\hslash\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)\times(1+\delta^{2})\\{[1-A\left(\frac{T}{T_{a}}\right)$ $\displaystyle+(2A\left(\frac{T}{T_{a}}\right)-4A\left(\frac{T}{T_{s}}\right)\cos[\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}])\sin^{2}[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}+\Phi]\\}$ (31) From the asymptotic expression given by Eq.(31), we observe that in the presence of in-phase electric and magnetic modulatio the magnetoplasmon energy acquires a dependence on the phase factor $\Phi$ and $\delta$ which depend on the relative modulation strengths. The shift in the Weiss oscillations when in-phase electric and magnetic modulations are present can be seen in Fig.(5). How the Weiss oscillations are affected as $\Phi$ as well as the magnetic field is varied can be seen in Fig.(5). The results shown are for a fixed magnetic modulation of strength $\hslash\omega_{0}=0.03$ meV and the electric modulation is varied. The change in $V_{0}$ results in a corresponding change in both $\delta$ and $\Phi$. From Fig.(5), we observe that the position of the extrema in the magnetoplasmon spectrum as a function of the inverse magnetic field depend on the relative strength of the modulations. The effects of electric and magnetic modulations that are out-of-phase on the magnetoplasmon spectrum can be better appreciated if we consider the asymptotic expression in this case. This is taken up in the next section. ## VI Magnetoplasmons with Periodic Electric and Magnetic Modulation: Out-of- phase In this section, we calculate the magnetoplasmon spectrum 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 are5 ; 26 $\varepsilon(n,x_{0})=(n+1/2)\hbar\omega_{c}+\sin(Kx_{0})G_{n}+F_{n}\cos(Kx_{0}),$ (32) and the bandwidth is $\Delta(\text{out of phase})=\frac{2\hslash\omega_{0}ak_{F}}{2\pi\sqrt{\pi\sqrt{nu}}}\times\sqrt{\delta^{2}+(1-\delta^{2})\sin\left(2\sqrt{nu}-\frac{\pi}{4}\right)}\,.$ (33) 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. Following the same approach discussed in the previous section for MM2DEG, we can obtain the intra-Landau band magnetoplasmon spectrum in the presence of out of phase modulations as $\hslash^{2}\overset{\sim}{\omega}^{2}=\frac{4e^{2}m^{\ast}\omega_{c}}{k\bar{q}\pi\hslash}\sin^{2}[\frac{\pi}{a}(x_{0}^{\prime})]\times O_{n},$ (34) where $O_{n}=\underset{n}{\sum}(G_{n}^{2}+F_{n}^{2})\times[-f^{\prime}(\varepsilon(n))]\ .$ The asymptotic expression 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 $\displaystyle\hslash^{2}\overset{\sim}{\omega}^{2}$ $\displaystyle=\frac{2V_{0}^{2}e^{2}m^{\ast}\omega_{c}\delta^{-2}}{k\bar{q}2\pi^{2}\hslash\sqrt{u\hslash\omega_{c}\varepsilon_{F}}}\sin^{2}\left(\frac{\pi}{a}(x_{0}^{\prime})\right)\times\\{2\delta^{2}+(1-\delta^{2})[1-A\left(\frac{T}{T_{a}}\right)$ $\displaystyle+2A\left(\frac{T}{T_{a}}\right)\sin^{2}[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}]]-4A\left(\frac{T}{T_{s}}\right)\cos[\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}])[\delta^{2}-(\delta^{2}-1)\sin^{2}[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}]]\\}$ (35) From the expression of the out-of-phase bandwidth given by Eq.(33) we find that Weiss oscillations in the bandwidth are absent for relative modulation strength $\delta=\pm 1$, the same is reflected in the magnetoplasmon spectrum as the term responsible for Weiss oscillations ($\sin^{2}[2\sqrt{\frac{u\varepsilon_{F}}{\hslash\omega_{c}}}-\frac{\pi}{4}]$) vanishes for $\delta=\pm 1$ as can be seen from the above equation. Therefore the magnetoplasmon spectrum does not exhibit Weiss oscillations when the relative modulation strength $\delta=\pm 1.$The spectrum as a function of magnetic field when the electric and magnetic modulations are out-of-phase is shown in Fig.(6). The results shown are for a fixed magnetic modulation of strength $\hslash\omega_{0}=0.03$ meV and the electric modulation $V_{0}$ is allowed to vary between positive and negative values. The other parameters are the same as in Figs.(4, 5). As $V_{0}$ is varied there is a corresponding change in $\delta.$ We find that the positions of the extrema of the spectrum 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 magnetoplasmon spectrum when the two modulations are out of phase. It is also observed in Fig.(6) 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 magnetoplasmon spectrum. ## VII Conclusions In conclusion, we have determined the inter and intra-Landau band magnetoplasmon spectrum for a magnetically modulated two-dimensional electron gas in the presence of a perpendicular magnetic field. Our results show that magnetic Weiss oscillations occur in intra-Landau band magnetoplasmon spectrum. Their origin lies in the interplay of the two physical length scales of the system i.e. the modulation period, and cyclotron diameter at the Fermi level. When the strength of magnetic modulation potential is equal to the electric one, the magnetic Weiss oscillations in the magnetoplasmon spectrum are out of phase and occur with a larger amplitude compared to the electric Weiss oscillations in a 2DEG. These results also exhibit that the Weiss oscillations depend on the temperature much less than that of the SdH oscillations. We have determined the effects of both the electric and magnetic modulations on the magnetoplasmon spectrum of 2DEG. These oscillations are affected by the relative phase of the two modulations and positions of the extrema of the oscillations depend 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. 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). Electronic address: m.tahir06@imperial.ac.uk. †Electronic address: ksabeeh@qau.edu.pk, kashifsabeeh@hotmail.com ## References (1) Izawa S, Katsamoto S, Endo A and Iye Y, J. Phys. Soc. Japan 64, 706-710 (1995). * (2) Ye P D, Weiss D, Gerhardts R R, Seeger M, von Klitzing K, Eberl K and Nickel H, Phys. Rev. Lett. 74, 3013-3016 (1995). * (3) Carmona H A, Geim A K, Nogaret A, Main P C, Foster T J, Henini M, Beaumont S P and Blamire M G, Phys. Rev. Lett. 74, 3009-3012 (1995). * (4) Yagi R and Iye Y, J. Phys. Soc. Japan 62, 1279-1285 (1993). * (5) Peeters F M and Vasilopoulos P, Phys. Rev. B 47, 1466-1473 (1993). * (6) Wu Xiaoguang and Ulloa Sergio E, Solid State Commun. 82 945-949 (1992). * (7) Xue Deng Ping and Xiao Gang, Phys. Rev. B 45, 5986-5990 (1992). * (8) P. Vasilopoulos and F. M. Peeters, Superlattices Microstruct. 7, 393-395 (1990).; P. Vasilopoulos and F. M. Peeters, Physica Scripta. Vol. T39, 177-181 (1991). * (9) Yoshioka D and Iye Y, J. Phys. Soc. Japan 56, 448-451 (1987). * (10) J. M. Luttinger, Phys. Rev. 84, 814-817 (1951). * (11) Q. P. Li and S. Das Sarma, Phys. Rev. B 44, 6277-6238 (1991). * (12) S. M. Stewart and C. Zhang, J. Phys.: Condens. Matter 8, 6019-6033 (1996). * (13) P. Vasilopoulos, X. F. Wang and F. M. Peeters, Phys. Rev. B 70, 155312-1-8 (2004).; I. S. Ibrahim and F. M. Peeters, Phys. Rev. B 52, 17321-17334 (1995). * (14) Andrey Krakovsky, Phys. Rev. B 53, 8469-8472 (1996).; M. C. Chang and Q. Niu, Phys. Rev. B 52, 10843-10850 (1995). * (15) A. Y. Rom, Phys. Rev. B 55, 11025-11028 (1997). * (16) S. Cina, D. M. Whittaker, D. D. Arnone, T. Burke, H. P. Hughes, M. Leadbeater, M. Pepper, and D. A. Ritchie, Phys. Rev. Lett. 83, 4425-4428 (1999). * (17) V. Fessatidis, H. L. Cui, and P. Vasilopoulos, Superlattices and Microstruct. 23, 47-54 (1998). * (18) Mayumi Kato, Akira Endo, Makoto Sakairi, Shingo Katsumoto, and Yasuhiro Iye, J. Phys. Soc. Jpn. 68, 1492-1495 (1999). * (19) Tong-Zhong Li, Shi-Wei Gu, Xue-Hua Wang and Jian-Ping Peng, J. Phys.: Condens. Matter 8, 313-327 (1996). * (20) A. Manolescu and V. Gudmundsson, Superlattices Microstruct. 23, 1169-1180 (1998). * (21) X. Wu and S. E. Ulloa, Phys. Rev. B 47, 7182-7186 (1993). * (22) Iye Y, Endo A, Izawa S, Kato M and Katsumoto S, Physica B 227 122-126 (1996). * (23) J. E. Muller, Phys. Rev. Lett. 68, 385-388 (1992). * (24) X. G. Wu and S. E. Ulloa, Phys. Rev. B 47, 10 028-10031 (1993). * (25) Jirong Shi, F. M. Peeters, K. W. Edmonds, and B. L. Gallagher, Phys. Rev. B 66, 035328-1-9 (2002). * (26) F M Peeters, P Vasilopoulos, Jirong Shi, Journal of Physics Condensed Matter 14, 8803-8816 (2002). * (27) K. W. Edmonds, B. L. Gallagher, P. C. Main, N. Overend, R. Wirtz, A. Nogaret, M. Henini, C. H. Marrows, B. J. Hickey, and S. Thoms, Phys. Rev. B 64, 041303-1-4 (2001). * (28) A. Matulis and F. M. Peeters, Phys. Rev. B 62, 91-94 (2000). * (29) H. L. Cui, V. Fessatidis, and N. J. M. Horing, Phys. Rev. Lett. 63, 2598-2601 (1989). * (30) W-M Que, G. Kirczenow, Phys. Rev. Lett. 62, 1687 (1989). * (31) Weiss D, Klitzing K V, Ploog K and Weiman G, Europhys. Lett. 8 179-184 (1989).; R. W. Winkler and J. P. Kotthaus, K. Ploog, Phys. Rev. Lett 62, 1177-1180 (1989).; Gerhardts R R, Weiss D and von Klitzing K, Phys. Rev. Lett. 62, 1173-1176 (1989). * (32) O. G. Balev, N. Studart, and P. Vasilopoulos, Phys. Rev. B 62, 15834 (2000),; M. S. Kushwaha, Surf. Sci. Rep. 41, 1-416 (2001), and references therein. * (33) Handbook of Mathematical Functions, edited by M. Abramowitz and I. A. Stegun (Dover, New York, 1972). * (34) H. Ehrenreich and M. H. Cohen, Phys. Rev. 115, 786-790 (1959). * (35) R. Krahne, M. Hochgräfe, Ch. Heyn, and D. Heitmann, Phys. Rev. B 61, R16319-R16322 (2000). * (36) Wei-ming Que and George Kirczenow, Phys. Rev. B 36, 6596-6601 (1987). * (37) T. Demel, D. Heitmann, P. Grambow, and K. Ploog, Phys. Rev. B 38, 12 732-12735 (1988). * (38) I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products (Academic Press, New York, 1980).	arxiv-papers	2008-08-04T16:09:29	2024-09-04T02:48:57.165249	{ "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/0808.0394" }
0808.0734	# From Equilibrium to Transport Properties of Strongly Correlated Fermi Liquids Thomas Schäfer Department of Physics, North Carolina State University, Raleigh, NC 27695 ###### Abstract We summarize recent results regarding the equilibrium and non-equilibrium behavior of cold dilute atomic gases in the limit in which the two body scattering length $a$ goes to infinity. In this limit the system is described by a Galilean invariant (non-relativistic) conformal field theory. We discuss the low energy effective lagrangian appropriate to the limit $a\to\infty$, and compute low energy coefficients using an $\epsilon$-expansion. We also show how to combine the effective lagrangian with kinetic theory in order to compute the shear viscosity, and compare the kinetic theory predictions to experimental results extracted from the damping of collective modes in trapped Fermi gases. ###### keywords: cold atomic gases, conformal symmetry, shear viscosity ## 1 Introduction Over the last ten years there has been remarkable progress in the study of “designer fluids”, dilute, non-relativistic Bose and Fermi gases in which the scattering length between the Bosons or Fermions can be continuously adjusted. In the following we are particularly interested in Fermi gases, since these systems are stable for both positive and negative values of the scattering length, including the strongly correlated limit in which the scattering length is taken to infinity. The scattering length is controlled through a Feshbach resonance. Alkali atoms such as 6Li and 40K have a single valence electron. When a dilute gas of atoms is cooled to very low temperatures, we can view the atoms as pointlike particles interacting via interatomic potentials which depend on the hyperfine quantum numbers. A Feshbach resonance arises if a molecular bound state in a “closed” hyperfine channel crosses near the threshold of an energetically lower “open” channel. Because the magnetic moments of the open and closed states are in general different, Feshbach resonances can be tuned using an applied magnetic field. At resonance the two-body scattering length in the open channel diverges, and the cross section $\sigma$ is limited only by unitarity, $\sigma(k)=4\pi/k^{2}$ for low momenta $k$. In the unitarity limit, details about the microscopic interaction are irrelevant, and the system displays universal properties. Near a Feshbach resonance the scattering length behaves as $a=a_{0}\left(1+\frac{\Delta B}{B-B_{0}}\right)\,$ (1) where $a_{0}$ is the non-resonant value of the scattering length (typically on the order of the effective range of the interatomic potential), $B$ is the magnetic field, $B_{0}$ the position of the resonance, and $\Delta B$ the width. A small negative scattering length corresponds to a weak attractive interaction between the atoms. This case is known as the BCS (Bardeen-Cooper- Schrieffer) limit. On the other side of the resonance the scattering length is positive. In the BEC (Bose-Einstein condensation) limit the interaction is strongly attractive and the fermions form deeply bound molecules. For this reason the unitarity limit $a\to\infty$ is also known at the BCS/BEC crossover. The unitarity limit is of interest to QCD practitioners for a for a number of reasons: * • The unitarity limit provides an approximate description of dilute neutron matter. The neutron-neutron scattering length is $a_{nn}=-18$ fm, and the effective range is $r_{nn}=2.8$ fm. This means that there is a range of densities, relevant to the outer layers of neutron stars, for which the interparticle spacing is large compared to the effective range, but small compared to the scattering length. * • The Fermi gas at unitarity is a high $T_{c}$ superconductor. There is an attractive interaction in the spin singlet channel which leads to s-wave superconductivity below some critical temperature $T_{c}$. In the unitarity limit the only energy scale in the problem is the Fermi energy $E_{F}$, and we must have $k_{B}T_{c}=\alpha E_{F}$ with some numerical constant $\alpha$. Quantum Monte Carlo calculations (and experimental results) indicate that $\alpha\simeq 0.15$ [1, 2], much larger than in ordinary (or even high $T_{c}$) electronic superconductors, but comparable to what might be achieved in color superconducting quark matter [3]. * • The limit $a\to\infty$ corresponds to a non-relativistic conformal field theory [4]. In the unitarity limit there is no scale in the problem (other than the thermodynamic variables temperature and density). Indeed, one can show that the theory is not only scale invariant, but invariant under the full conformal group. This raises the question whether there are any physical consequences of conformal symmetry that go beyond results that follow from scale invariance. It also raises the possibility that a holographic description, similar to the $AdS/CFT$ correspondence, can be obtained [5, 6]. * • Non-relativistic fermions at unitarity behave as a very good fluid and show interesting transport properties, including a very small shear viscosity. Kinetic theory suggests that the shear viscosity is inversely proportional to the scattering cross section, and reaches a minimum at unitarity. This expectation is confirmed by experiments that demonstrate large elliptic flow and a very small damping rate for collective oscillations [7, 8]. ## 2 Equilibrium Properties We begin by analyzing equilibrium properties of the dilute Fermi gas at unitarity. If the temperature is large, $k_{B}T>E_{F}$, then the scattering cross section is regularized by the thermal wave length, and the effective interaction is weak. Here the Fermi energy is defined by $E_{F}=(3\pi^{2}n)^{2/3}/(2m)$, where $n$ is the density, and $m$ is the mass of the atoms. In the high temperature regime the equation of state is well described by the Virial expansion, and the system has single particle excitations with the quantum numbers of the fundamental fermions. In the regime $k_{B}T\sim E_{F}$ the interactions are strong. As noted above, superfluidity occurs at $k_{B}T_{c}\simeq 0.15E_{F}$. Below the critical temperature the excitations are Goldstone bosons. In following section we will discuss the effective theory of the Goldstone bosons, and relate the parameters in the effective lagrangian to static properties of the system. ### 2.1 Low Energy Effective Theory and Density Functional The Goldstone boson field can be defined as the phase of the difermion condensate $\langle\psi\psi\rangle=e^{2i\varphi}\|\langle\psi\psi\rangle\|$. The effective Lagrangian at next-to-leading order (NLO) in derivatives of $\varphi$ and the external potential is [9] ${\cal L}=c_{0}m^{3/2}X^{5/2}+c_{1}m^{1/2}\frac{(\vec{\nabla}X)^{2}}{\sqrt{X}}+\frac{c_{2}}{\sqrt{m}}\left[\left(\nabla^{2}\varphi\right)^{2}-9m\nabla^{2}V\right]\sqrt{X}\,,$ (2) where we have defined $X=\mu-V-\dot{\varphi}-\frac{(\vec{\nabla}\varphi)^{2}}{2m}\,.$ (3) Here, $\mu$ is the chemical potential and $V(\vec{x},t)$ is an external potential. The functional form of the effective lagrangian is fixed by the symmetries of the problem, Galilean invariance, $U(1)$ symmetry, and conformal symmetry. The NLO effective lagrangian is characterized by three dimensionless parameters, $c_{0},c_{1},c_{2}$. These parameters can be related to physical properties of the system. The first parameter, $c_{0}$, can be related to the equation of state. We have $c_{0}=\frac{2^{5/2}}{15\pi^{2}\xi^{3/2}}\,,$ (4) where $\xi$ determines the chemical potential in units of the Fermi energy, $\mu=\xi E_{F}$. The two NLO parameters $c_{1},c_{2}$ are related to the momentum dependence of correlation functions. The phonon dispersion relation, for example, is given by $q_{0}=v_{s}q\left[1-\pi^{2}\sqrt{2\xi}\left(c_{1}+\frac{3}{2}c_{2}\right)\frac{q^{2}}{k_{F}^{2}}+O(q^{4}\log(q^{2}))\right]$ (5) where $v_{s}=\sqrt{\xi/3}v_{F}$ and $v_{F}=k_{F}/m$. The static susceptibility $\chi(q)=-i\int dt\,d^{3}x\;e^{-i\vec{q}\cdot\vec{x}}\,\langle\psi^{\dagger}\psi(0)\psi^{\dagger}\psi(x)\rangle$ (6) involves a different linear combination of $c_{1}$ and $c_{2}$, $\chi(q)=-\frac{mk_{F}}{\pi^{2}\xi}\left[1+2\pi^{2}\sqrt{2\xi}\left(c_{1}-\frac{9}{2}c_{2}\right)\frac{q^{2}}{k_{F}^{2}}+O(q^{4}\log(q^{2}))\right].$ (7) Higher derivative terms in the effective lagrangian can also be used to compute the energy of inhomogeneous matter. At NLO in an expansion in derivatives of the density we find the following energy density functional [10] $\displaystyle{\cal E}(x)$ $\displaystyle=$ $\displaystyle n(x)V(x)+\frac{3\cdot 2^{2/3}}{5^{5/3}mc_{0}^{2/3}}n(x)^{5/3}-\frac{4}{45}\frac{2c_{1}+9c_{2}}{mc_{0}}\frac{\left(\nabla n(x)\right)^{2}}{n(x)}$ $\displaystyle\mbox{}-\frac{12}{5}\frac{c_{2}}{mc_{0}}\nabla^{2}n(x)\,.$ The first two terms correspond to the local density approximation (LDA) and the terms proportional to $c_{1}$ and $c_{2}$ are the leading correction to the LDA involving derivatives of the density. Figure 1: Leading order contributions to the effective potential in the epsilon expansion. Solid lines denote fermions propagators, dashed lines denote boson propagators, and the cross is an insertion of the chemical potential. ### 2.2 Epsilon Expansion At unitarity the determination of $c_{1}$ and $c_{2}$ is a non-perturbative problem, and we will perform the calculation using an expansion around $d=4-\epsilon$ spatial dimensions [11, 12]. Our starting point is the lagrangian ${\cal L}=\Psi^{\dagger}\left[i\partial_{0}+\sigma_{3}\frac{\vec{\nabla}^{2}}{2m}\right]\Psi+\mu\Psi^{\dagger}\sigma_{3}\Psi+\left(\Psi^{\dagger}\sigma_{+}\Psi\phi+h.c.\right)-\frac{1}{C_{0}}\phi^{\dagger}\phi\ ,$ (9) where $\Psi=(\psi_{\uparrow},\psi_{\downarrow}^{\dagger})^{T}$ is a two- component Nambu-Gorkov field, $\sigma_{i}$ are Pauli matrices acting in the Nambu-Gorkov space, $\sigma_{\pm}=(\sigma_{1}\pm i\sigma_{2})/2$, $\phi$ is a complex boson field, and $C_{0}$ is a coupling constant. In dimensional regularization the fermion-fermion scattering length becomes infinite for $1/C_{0}\to 0$. The epsilon expansion is based on the observation that the fermion-fermion scattering amplitude near $d=4$ dimensions is saturated by the propagator of a boson with mass $2m$. The coupling of the boson to pairs of fermions is given by $g=\frac{\sqrt{8\pi^{2}\epsilon}}{m}\left(\frac{m\phi_{0}}{2\pi}\right)^{\epsilon/4}\,.$ (10) In the superfluid phase $\phi$ acquires an expectation value $\phi_{0}=\langle\phi\rangle$. We write the boson field as $\phi=\phi_{0}+g\varphi$. The lagrangian is split into a free part ${\cal L}_{0}=\Psi^{\dagger}\left[i\partial_{0}+\sigma_{3}\frac{\vec{\nabla}^{2}}{2m}+\phi_{0}(\sigma_{+}+\sigma_{-})\right]\Psi+\varphi^{\dagger}\left(i\partial_{0}+\frac{\vec{\nabla}^{2}}{4m}\right)\varphi\,,$ (11) and an interacting part ${\cal L}_{I}+{\cal L}_{ct}$, where $\displaystyle{\cal L}_{I}$ $\displaystyle=$ $\displaystyle g\left(\Psi^{\dagger}\sigma_{+}\Psi\varphi+h.c\right)+\mu\Psi^{\dagger}\sigma_{3}\Psi+2\mu\varphi^{\dagger}\varphi\,,$ (12) $\displaystyle{\cal L}_{ct}$ $\displaystyle=$ $\displaystyle-\varphi^{\dagger}\left(i\partial_{0}+\frac{\vec{\nabla}^{2}}{4m}\right)\varphi-2\mu\varphi^{\dagger}\varphi\,.$ (13) Note that the leading self energy corrections to the boson propagator generated by the interaction term ${\cal L}_{I}$ cancel against the counterterms in ${\cal L}_{ct}$. The chemical potential term for the fermions is included in ${\cal L}_{I}$ rather than in ${\cal L}_{0}$. This is motivated by the fact that near $d=4$ the system reduces to a non-interacting Bose gas and $\mu\to 0$. We will count $\mu$ as a quantity of $O(\epsilon)$. The Feynman rules are quite simple. The fermion and boson propagators are $\displaystyle G(p_{0},p)$ $\displaystyle=$ $\displaystyle\frac{i}{p_{0}^{2}-E_{p}^{2}}\left[\begin{array}[]{cc}p_{0}+\epsilon_{p}&-\phi_{0}\\\ -\phi_{0}&p_{0}-\epsilon_{p}\end{array}\right]\ ,$ (16) $\displaystyle D(p_{0},p)$ $\displaystyle=$ $\displaystyle\frac{i}{p_{0}-\epsilon_{p}/2}\ ,$ (17) where $E_{p}^{2}=\epsilon_{p}^{2}+\phi_{0}^{2}$ and $\epsilon_{p}=p^{2}/(2m)$. The fermion-boson vertices are $ig\sigma^{\pm}$. Insertions of the chemical potential are $i\mu\sigma_{3}$. Both $g^{2}$ and $\mu$ are corrections of order $\epsilon$. Figure 2: Scalar self energy at LO in the epsilon expansion. In order to determine $c_{0},c_{1},c_{2}$ we have to compute three physical observables. We have studied $\xi=\mu/E_{F}$, and the curvature terms in the phonon dispersion relation and the static susceptibility. The universal parameter $\xi$ was originally calculated by Nishida and Son. They computed the effective potential to NLO in the epsilon expansion, see Fig. 1. The derivative of the effective potential with respect to $\mu$ determines the density $n$, and the relation between $n$ and $\mu$ fixes $\xi$. The result is $\xi=\frac{\epsilon^{3/2}}{2}\left[1+\frac{1}{8}\epsilon\log(\epsilon)-\frac{1}{4}\left(12C-5+5\log(2)\right)\epsilon+O(\epsilon^{2})\right]\,,$ (18) with $C=0.144$. The phonon dispersion relation can be extracted from the scalar propagator. We introduce a two-component scalar field $\Phi=(\varphi,\varphi^{})$. The scalar propagator satisfies a Dyson- Schwinger equation [13] $\begin{array}[]{c}\includegraphics[width=199.16928pt]{prop_phonon_eps.eps}\end{array}$ (19) At LO in the epsilon expansion the self energy is determined by the diagrams shown in Fig. 2. NLO contributions were calculated in [10]. The phonon dispersion relation is $p_{0}=\sqrt{\mu\epsilon_{p}}\left(1+\frac{\epsilon}{8}\right)\left\\{1+\frac{\epsilon_{p}}{8\mu}\left(1-\frac{\epsilon}{4}\right)+\ldots\right\\}$ (20) We note that the dispersion relation curves up (unlike 4He, but similar to weakly interacting Bose gases). This implies that there is $\varphi\to\varphi+\varphi$ decay. Finally, we can determine the static susceptibility. Computing the diagrams in Fig. 3 we get [10, 14] $\displaystyle\chi(q)$ $\displaystyle=$ $\displaystyle-\frac{2Z}{\epsilon\mu}\left\\{1-\frac{1}{8}\left(\frac{q^{2}}{m\mu}\right)\left(1-\frac{\epsilon}{4}\right)+O(q^{4})\right\\}\left(\frac{m\phi_{0}}{2\pi}\right)^{d/2}\,,$ $\displaystyle\mbox{}Z\,=\,1-\frac{1}{2}\left(\gamma-\log(2)\right)\epsilon\,.$ The coefficient $c_{0}$ follows from the result for $\xi$ ($\xi=0.475$ at NLO in the $\epsilon$-expansion) using equ. (4). Matching equ. (20,2.2) against equ. (5,7) gives $c_{2}=0$ and $c_{1}/c_{0}=3/8-\epsilon/4$. The corresponding energy density functional was studied in [10]. Compared to a free Fermi gas the local density term is reduced by a factor $\sim 2$ (the interaction is attractive), while the gradient correction proportional to $(\nabla n)^{2}/n$ is enhanced by a factor $\sim 2$. + + Figure 3: Leading order contributions to the static susceptibility. The wiggly line denotes an external current. The double line is the scalar propagator defined in equ. (19). ## 3 Transport Properties In the following we will discuss transport properties of the Fermi gas at unitarity. The interest in non-equilibrium properties arises from the observation that transport coefficients are much more sensitive to the strength of the interaction than thermodynamic quantities. A renewed interest in transport properties was also sparked the AdS/CFT correspondence and the experimental limits on the shear viscosity of the quark gluon plasma obtained at RHIC. In the following we shall focus on the shear viscosity of the Fermi gas at unitarity. Close to equilibrium the (coarse grained) energy momentum tensor can be written as $\displaystyle T_{ij}$ $\displaystyle=$ $\displaystyle(P+\epsilon)v_{i}v_{j}-P\delta_{ij}+\delta T_{ij}\ ,$ (22) $\displaystyle\delta T_{ij}$ $\displaystyle=$ $\displaystyle-\eta(\nabla_{i}v_{j}+\nabla_{i}v_{j}-\frac{2}{3}\delta_{ij}\vec{\nabla}\cdot\vec{v})+\cdots,$ where $\epsilon$ and $P$ are the energy density and pressure, and $v_{i}$ is the local flow velocity. The first term is the ideal gas contribution, and $\delta T_{ij}$ is the leading order (in gradients of $v_{i}$) dissipative correction. The traceless part of $\delta T_{ij}$ is proportional to the shear viscosity $\eta$. ### 3.1 Kinetic Theory a) b) Figure 4: Leading order processes that contribute to the shear viscosity at low temperature (Fig. a) and high temperature (Fig. b). Dashed lines are phonon propagators and solid lines are fermion propagators. We first consider the case that the fluid is composed of weakly interacting quasi-particles. In the unitarity limited Fermi gas this is the case at $T\ll T_{c}$ (phonons) and $T\gg T_{c}$ (atoms). In these limits we can compute the shear viscosity using kinetic theory. In the following we will concentrate on the low temperature case discussed in [15]. In kinetic theory the stress- energy tensor is given by $T_{ij}=v_{s}^{2}\int\frac{d^{3}p}{(2\pi)^{3}}\frac{p_{i}p_{j}}{E_{p}}f_{p}\,,$ (23) where $f_{p}$ is the distribution function of the phonons, $v_{s}$ is the speed of sound, $p_{i}$ is the momentum and $E_{p}$ the quasi-particle energy. Close to equilibrium $f_{p}=f_{p}^{(0)}+\delta f_{p}$, where $f_{p}^{(0)}$ is the Bose-Einstein distribution and $\delta f_{p}$ is a small departure from equilibrium. We write $\delta f_{p}=-\chi(p)f_{p}^{(0)}(1+f_{p}^{(0)})/T$. In the case of shear viscosity we can further decompose $\chi(p)=g(p)(p_{i}p_{j}-\frac{1}{3}\delta_{ij}p^{2})(\nabla_{i}v_{j}+\nabla_{j}v_{i}-\frac{2}{3}\delta_{ij}\vec{\nabla}\cdot\vec{v})\,.$ (24) Inserting equ. (24) into equ. (23) we get $\eta=\frac{4v^{2}}{15T}\int\frac{d^{3}p}{(2\pi)^{3}}\frac{p^{4}}{2E_{p}}f_{p}^{(0)}(1+f_{p}^{(0)})g(p)\,.$ (25) The non-equilibrium distribution $g(p)$ is determined by the Boltzmann equation $\frac{df_{p}}{dt}=\frac{\partial f_{p}}{\partial t}+\vec{v}\cdot\vec{\nabla}f_{p}=C[f_{p}],$ (26) relating the rate of change of the distribution function $f_{p}$ to the collision operator $C[f_{p}]$. The $2\leftrightarrow 2$ collision integral is given by $\displaystyle C_{2\leftrightarrow 2}[f_{p}]$ $\displaystyle=$ $\displaystyle\frac{1}{2E_{p}}\int\frac{d^{3}k}{(2\pi)^{3}2E_{k}}\frac{d^{3}k^{\prime}}{(2\pi)^{3}2E_{k^{\prime}}}\frac{d^{3}p^{\prime}}{(2\pi)^{3}2E_{p^{\prime}}}$ $\displaystyle\times(2\pi)^{4}\delta^{(4)}(p+k-p^{\prime}-k^{\prime})\|{\cal M}\|^{2}D_{2\leftrightarrow 2},$ where $D_{2\leftrightarrow 2}$ contains the distribution functions and $\|{\cal M}\|$ is the $2\leftrightarrow 2$ scattering amplitude shown in Fig. 4. The three and four-phonon vertices are fixed by the effective lagrangian (2). Linearizing $D_{2\leftrightarrow 2}$ in $\delta f_{p}$ one finds $D_{2\leftrightarrow 2}=\frac{1}{T}f_{k^{\prime}}^{(0)}f_{p^{\prime}}^{(0)}(1+f_{k}^{(0)})(1+f_{p}^{(0)})\,\left(\chi(p)+\chi(k)-\chi(p^{\prime})-\chi(k^{\prime})\right)\,.$ (28) There are a variety of methods for solving the linearized Boltzmann equation. A standard technique is based on expanding $g(p)$ in a complete set of functions. A nice feature of this method is that the truncated expansion gives a variational estimate $\eta\geq\frac{4v^{4}}{25T^{2}}\frac{(b_{0}A_{00})^{2}}{\sum_{s,t}b_{s}b_{t}M_{st}}\,$ (29) where $b_{s}$ is a set of expansion coefficients, $A_{00}$ is a normalization integral, and $M_{st}$ are matrix elements of the linearized collision operator. For the best trial function we find [15] $\eta/s=7.7\times 10^{-6}\xi^{5}\frac{T_{F}^{8}}{T^{8}}\,,$ (30) where $\xi$ is the universal parameter introduced in Sect. 2.1 and we have normalized the result to the entropy density $s$ of a weakly interacting phonon gas. A similar estimate can be obtained in the high temperature limit. In this case the relevant degrees of freedom are atoms, and the dominant scattering process is shown in Fig. 4b. The result is [16, 17] $\eta/s=\frac{45\pi^{3/2}}{64\sqrt{2}}\left(\frac{T}{T_{F}}\right)^{3/2}\left[\log\left(\frac{3\sqrt{\pi}}{4}\frac{T^{3/2}}{T_{F}^{3/2}}\right)+\frac{5}{2}\right]^{-1}\,.$ (31) The high and low temperature limits of $\eta/s$ are shown in Fig. 5, together with the proposed lower bound $\eta/s=1/(4\pi)$ [18] and experimental data which we will discuss in the next section. ### 3.2 Hydrodynamics Figure 5: Viscosity to entropy density ratio of a cold atomic gas in the unitarity limit. This plot is based on the damping data published in [8] and the thermodynamic data in [19, 20]. The dashed line shows the conjectured viscosity bound $\eta/s=1/(4\pi)$, and the solid lines show the high and low temperature limits. Hydrodynamics describes the evolution of long-wavelength, slow-frequency modes. The hydrodynamic description remains valid even if there is no underlying kinetic theory. The hydrodynamic equations follow from conservation of mass (particle number), energy and momentum. In a non-relativistic system the equations of continuity and of momentum conservation are given by $\displaystyle\frac{\partial n}{\partial t}+\vec{\nabla}\cdot\left(n\vec{v}\right)$ $\displaystyle=$ $\displaystyle 0,$ (32) $\displaystyle mn\frac{\partial\vec{v}}{\partial t}+mn\left(\vec{v}\cdot\vec{\nabla}\right)\vec{v}$ $\displaystyle=$ $\displaystyle-\vec{\nabla}P-n\vec{\nabla}V,$ (33) where $n$ is the number density, $m$ is the mass of the atoms, $\vec{v}$ is the fluid velocity, $P$ is the pressure and $V$ is the external potential. In an ideal fluid the equation of energy conservation can be rewritten as conservation of entropy, $\frac{\partial ns}{\partial t}+\vec{\nabla}\cdot\left(ns\vec{v}\right)=0\,.$ (34) A non-zero shear viscosity leads to dissipation, converting kinetic energy to heat and increasing the entropy. The shear viscosity of the dilute Fermi gas in the unitarity limit can be measured by studying the damping of collective modes in trapped systems [21]. The frequency of these modes agrees well the prediction of ideal hydrodynamics. The dissipated energy is given by $\dot{E}=-\frac{1}{2}\int d^{3}x\,\eta(x)\,\left(\partial_{i}v_{j}+\partial_{j}v_{i}-\frac{2}{3}\delta_{ij}\partial_{k}v_{k}\right)^{2}\,.$ (35) The damping rate is given by the ratio of the energy dissipated to the total energy of the collective mode. The kinetic energy is $E_{kin}=\frac{m}{2}\,\int d^{3}x\,n(x)\vec{v}^{\,2}\,.$ (36) If the damping rate is small both $\dot{E}$ and $E_{kin}$ can be computed using the solution of ideal hydrodynamics. We recently performed an analysis [22] which is based on measurements of the damping rate of the lowest radial breathing mode performed by the Duke group [8]. We showed that can relate the dimensionless ratio $\Gamma/\omega$, where $\Gamma$ is the damping rate and $\omega$ is the trap frequency, to the shear viscosity to entropy density ratio $\frac{\eta}{s}=\frac{3}{4}\xi^{1/2}(3N)^{1/3}\left(\frac{\bar{\omega}\Gamma}{\omega_{\perp}^{2}}\right)\left(\frac{E}{E_{T=0}}\right)\left(\frac{N}{S}\right).$ (37) Here $N$ is the total number of particles in the trap ($2\cdot 10^{5}$ in [8]), $\xi$ is the universal parameter defined in Sec. 2.1, $E/E_{T=0}$ is the ratio of the total energy to the energy at $T=0$ (which can be extracted using a Virial theorem from the measured cloud size), and $S/N$ is the entropy per particle (which is measured using adiabatic sweeps to the BCS limit [20]). The results are compared to theoretical prediction in the high and low temperature limit in Fig. 5. The data show a minimum near $T/T_{F}\simeq 0.2$. At the minimum $\eta/s\sim 1/2$. This should probably be considered as an upper bound, since dissipative mechanism other than shear viscosity may be present. In the high $T$ limit there is fairly good agreement with kinetic theory. The temperature dependence implied by the low $T$ prediction is not seen in the data. This is maybe not very surprising, since the mean free path in the low $T$ regime quickly exceeds the system size. ## 4 Outlook There are many promising directions for further study. Clearly, it is desirable to obtain additional experimental constraints on the shear viscosity, and to improve the theoretical analysis of the existing data sets. It would also be interesting to confirm that the bulk viscosity vanishes in the normal phase, and to measure the thermal conductivity. We would also like to improve the theoretical tools for computing transport properties in the interesting regime near $T_{c}$. There are some recent ideas for applying holography and the $AdS/CFT$ correspondence to Galilean invariant conformal field theories [5, 6], but there are also many purely field theoretic methods ($\epsilon$ expansions, large $N$ methods) that have yet to be pursued. Acknowledgments: Much of this work was carried out in collaboration with G. Rupak. The work is supported in part by the US Department of Energy grant DE- FG02-03ER41260. ## References [1] E. Burovski, N. Prokof’ev, B. Svistunov, M. Troyer, Phys. Rev. Lett. 96, 160402 (2006) [cond-mat/0602224]. * [2] A. Bulgac, J. E. Drut, P. Magierski, arXiv:0803.3238 [cond-mat.stat-mech]. * [3] M. G. Alford, A. Schmitt, K. Rajagopal and T. Schäfer, Rev. Mod. Phys., in press, arXiv:0709.4635 [hep-ph]. * [4] T. Mehen, I. W. Stewart and M. B. Wise, Phys. Lett. B 474, 145 (2000) [arXiv:hep-th/9910025]. * [5] D. T. Son, arXiv:0804.3972 [hep-th]. * [6] K. Balasubramanian and J. McGreevy, arXiv:0804.4053 [hep-th]. * [7] K. M. O’Hara, S. L. Hemmer, M. E. Gehm, S. R. Granade, J. E. Thomas, Science 298, 2179 (2002) [cond-mat/0212463]. * [8] J. Kinast, A. Turlapov, J. E. Thomas, Phys. Rev. Lett. 94, 170404 (2005) [cond-mat/0502507]. * [9] D. T. Son and M. Wingate, Annals Phys. 321, 197 (2006) [arXiv:cond-mat/0509786]. * [10] G. Rupak and T. Schäfer, arXiv:0804.2678 [nucl-th]. * [11] Z. Nussinov and S. Nussinov, Phys. Rev. A 74, 053622 (2006), cond-mat/0410597. * [12] Y. Nishida and D. T. Son, Phys. Rev. Lett. 97, 050403 (2006) [arXiv:cond-mat/0604500]. * [13] Y. Nishida, Phys. Rev. A 75, 063618 (2007) [arXiv:cond-mat/0608321]. * [14] A. Kryjevski, arXiv:0804.2919 [nucl-th]. * [15] G. Rupak and T. Schäfer, Phys. Rev. A 76, 053607 (2007) [arXiv:0707.1520 [cond-mat.other]]. * [16] P. Massignan, G. M. Bruun, and H. Smith, Phys. Rev. A 71, 033607 (2005). * [17] G. M. Bruun and H.‘Smith, Phys. Rev. A 72, 043605 (2005). * [18] P. Kovtun, D. T. Son and A. O. Starinets, Phys. Rev. Lett. 94, 111601 (2005) [hep-th/0405231]. * [19] J. Kinast, A. Turlapov, J. E. Thomas, Q. Chen, J. Stajic, and K. Levin, Science 307, 1296 (2005) [cond-mat/0502087]. * [20] L. Luo, B. Clancy, J. Joseph, J. Kinast, J. E. Thomas, preprint, cond-mat/0611566. * [21] J. Kinast, S. L. Hemmer, M. E. Gehm, A. Turlapov, and J. E. Thomas, Phys. Rev. Lett. 92, 150402 (2004). * [22] T. Schäfer, Phys. Rev. A 76, 063618 (2007) [arXiv:cond-mat/0701251].	arxiv-papers	2008-08-05T22:21:28	2024-09-04T02:48:57.175088	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Thomas Schaefer (North Carolina State University)", "submitter": "Thomas Schaefer", "url": "https://arxiv.org/abs/0808.0734" }
0808.0764		arxiv-papers	2008-08-06T17:25:02	2024-09-04T02:48:57.179345	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Scott Morrison, Emily Peters, Noah Snyder", "submitter": "Scott Morrison", "url": "https://arxiv.org/abs/0808.0764" }
0808.0800	# Sub-barrier fusion excitation for the system 7Li+28Si Mandira Sinha,1,2 H. Majumdar,1,∗ P. Basu,1 Subinit Roy,1 R. Bhattacharya,2 M. Biswas,1 M. K. Pradhan,1 S. Kailas3 E.mail: harashit.majumdar@saha.ac.in 1 Saha Institute of Nuclear Physics, 1/AF, Bidhan Nagar, Kolkata-700064, India 2 Gurudas College, Narikeldanga, Kolkata-700054, India 3 Nuclear Physics Division, Bhabha Atomic Research Centre, Mumbai-400085, India harashit.majumdar@saha.ac.in ###### Abstract The sub-barrier fusion excitation functions are measured for the first time for the system 7Li +28Si by the characteristic $\gamma$-ray method in the energy range $E_{\textit{lab}}$= 7-11.5 MeV. The results show an enhancement, below the barrier, by about a factor of two when compared with the one- dimensional barrier penetration (1D BPM) model. Introduction of coupling with the rotational 2+ state (1.779MeV) of the target improves the fit somewhat, but still an enhancement of about 25-40$\%$ remains. ###### pacs: 25.60.Pj;25.70.$-$z;25.70.Gh ††preprint: APS/123-QED Exploring the structure and reaction mechanism with loosely bound stable projectiles or with radioactive ion beams (e.g. halo/skin nuclei), at sub- and near-barrier energies, appears to be an interesting and challenging problem at the present time. Recent theoretical studies Dass96 ; Huss93 ; Taki93 ; Hagi00 ; NZag ; Huss06 have yielded new insights into the fusion reaction dynamics leading to enhancement/suppression of fusion, weakening of usual threshold anomaly found with tightly bound projectiles and appearance of new type of break-up threshold anomaly around the barrier energy. The interplay among fusion, loose structure, breakup to the continuum, and transfer channels are considered to be responsible for the above phenomena. Precise measurements exist for the fusion of loosely bound stable projectiles with heavy targets like 165Ho, 206,208Pb, 209Bi Mdas99 ; YW ; Sign97 ; VT ; YuE . Most of these experiments found appreciable enhancement of complete fusion (CF) compared to the one-dimensional barrier penetration model (1D BPM) calculation at the sub-barrier energy region. However role of the breakup and other nonelastic channels are not explicitly and unambiguously discerned. In the medium mass range there have been some recent experimental investigations SMores ; Padro02 ; Anjos ; cbeck ; mromo ; Gomes where only the total fusion could be measured owing to limitations of the techniques used. Most of these studies were done at energies above the Coulomb barrier and observed that TF is not affected by breakup. However Beck $et$ $al$. cbeck reported some small enhancement of 6Li induced fusion with 59Co at energies very close to the barrier. The experimental attempts for the sub- or near-barrier fusion studies in the light mass region (A$\sim$20-50) are rare. Most of the recent investigations Fig06 ; Anjos1 ; Padro02 ; GMar05 are pursued at well above the barrier energies. The fusion cross sections were found not to be hindered by breakup and agree well with 1D BPM predictions. Reference Padro02 showed that fusion excitations induced by stable weakly bound projectiles like 6,7Li and 9Be, at energies above the barrier, are almost similar to those produced by strongly bound nuclei 11B and 16O. But this observation is in contradiction to the findings of Figuiera $et$ $al$. Fig06 , where it has been shown that a hindrance to fusion cross section is systematically larger for reactions induced by weakly bound projectiles (e.g., 9Be) than for those with strongly bound nuclei (e.g., 11B and 12C). Of all the existing studies, only one experiment GMar05 measured the total fusion cross section by the time of flight technique, at energies very close to but above the nominal barrier. Here also total fusion was seen not to be affected by the break-up process. However, the deduced reaction cross section, even nearest to barrier energy, was found to be larger than the fusion cross section. Most of the above works showed neither enhancement nor suppression of excitation function in the near-barrier energies. Moreover, none of these experiments explored the fusion behaviour below the Coulomb barrier. In this perspective we present, for the first time, an experimental measurement of fusion cross section for the system 7Li +28Si, at sub-barrier energies, and extend it to just above the barrier. This study is complementary to our earlier work mand for the same system, measuring excitation function at energies well above the barrier, where some sort of suppression was observed beyond twice the barrier energy. To measure the total fusion (TF) cross-sections for 7Li +28Si, an experiment was done at 3 MV Pelletron accelerator of Institute of Physics (Bhubaneswar) with 7Li (2+, 3+) beam (8-30 pnA) at energies, $E_{\textit{lab}}$= 7, 8, 8.5, 10, and 11.5 MeV. A self-supported thin target of 28Si (175$\mu$g/cm2) was used. A specially designed small thin walled target chamber made of stainless steel (in the 0o beam line) was used to measure the fusion cross section using the characteristic $\gamma$-ray method. The $\gamma$-rays emitted from the evaporation residues were detected using a HPGe detector placed at 125o with respect to the beam direction. A long insulated metallic cylinder with proper electron suppressor was used as Faraday cup and standard current integrator was employed to measure the incident beam current. Efficiency runs were taken both at the beginning and at the end of the main experiment with a number of standard sources (152Eu, 133Ba, 207Bi) spanning the energy range 81-1770 keV. Some of the important residues in the fusion of 7Li +28Si detected are 30Si, 32S, 33S and 30P. Their characteristic $\gamma$-ray cross sections are shown in Fig. 1. The solid lines show CASCADE six predictions. The agreement between experiment and the compound nuclear evaporation estimation is apparent for data at energies above the Coulomb barrier, but at Coulomb barrier and below, the $\gamma$-ray augmentation is noticeable. These cross sections (${\sigma}_{\gamma}$) were extracted after analyzing the $\gamma$-ray spectra and using relevant efficiencies, beam and target specifications as described in Ref. mand . To subtract/correct for $\gamma$-rays arising out of beam impingement on the slit, beam line or the Faraday cup, one additional spectra was taken with a beam using a Ta-frame having a hole in place of the target position. The main contributions to fusion come from channels like $p$$n$+33S, $d$$n$+32S, $\alpha$$p$+30Si for E $\leq$ 9 MeV and $p$$n$+33S, $\alpha$$p$+30Si, $\alpha$$n$+30P, $d$$n$+32S for E$\geq$9 MeV. Some of the prominent identified $\gamma$-rays are 1.263 MeV(30Si), 2.230 + 2.235 MeV (32S+30Si), 1.967 MeV (33S), 0.677 MeV(30P). The contributions of the observed channels are almost about 85-80$\%$ of the total fusion cross section for 7Li+28Si system, in the energy region 7-11.5 MeV, respectively. The total fusion cross section was extracted as the ratio of the total experimentally measured $\gamma$-ray cross-sections and the corresponding branching factor $F_{\gamma}$. As there were overlapping $\gamma$-rays and weak transitions $F_{\gamma}$ was estimated theoretically, following a procedure used in Refs. Day ; Mukh as the ratio of the total theoretical $\gamma$-ray cross sections and the corresponding theoretical fusion cross section, both obtained from a statistical model calculation using the code CASCADE .The uncertainty in the measurement of the fusion cross-section was estimated to be about 16$\%$ for all energies, except for the lowest energy, where it was nearly 20$\%$, owing to a very poor yield. Figure 1: ${\sigma}_{\gamma}$ vs $E_{c.m.}$ for 7Li+28Si . Theoretical CASCADE predictions are shown by solid lines. Figure 2: Experimental fusion excitation function and theoretical predictions. The estimated projectile energy loss in the half thickness target is about 134 keV at 7 MeV and 125 to 102 keV in the high energy regime 8-11.5 MeV. The intrinsic energy resolution and uncertainty in beam energy calibration yields an error of about 30 keV. These factors were taken into account and fusion cross sections were plotted as a function of effective projectile energy in Fig. 2. The effective projectile energy was also used in the other figures. The overall resulting uncertainty in projectile energy is also shown. The one- dimensional barrier penetration model (1D BPM) estimates were found out using the coupled channel code CCFULL Ccfull in the no coupling mode, and are shown in Fig. 2 for comparison. The input optical model parameters ($V_{0}$ = 130 MeV, $r_{0}$ = 0.97 fm and $a_{0}$ = 0.63 fm) were extracted as described in ref. mand . It is seen that below the nominal barrier ($V_{b}$ = 6.79 MeV) the theoretical prediction underestimates the experimental excitation function and the difference is more near the barrier, pointing to the effective role of coupling in this region. Sub-barrier enhancement with respect to 1D BPM is apparent and it is more prominent just below the barrier. The experimental results are also compared with Wong’s phenomenological prediction wong using the input parameters like, barrier and barrier radius from the prescription of Vaz $et$ $al$. vaz and curvature from Wong parametrization. These values are respectively 6.74 MeV ($V_{b}$), 8.18 fm ($R_{b}$) and 3.24 ($\hbar$$w$). As expected, the Wong formulation overestimates the fusion excitation well below the barrier owing to assumption of parabolic nature of the potential, whereas the shape of real nucleus-nucleus potential may be asymmetric and broad at lower energies. Our experimental observations are somewhat similar to the recent findings of Penionzhkevick $et$ $al$. YuE for 6He + 208Pb having large enhancement and of Beck $et$ $al$. cbeck for 6Li + 59Co yielding small enhancement. We have explored the effects of rotational coupling employing the exact coupled channels calculation with CCFULL where the rotational state $2^{+}$ (1.779 MeV) of 28Si (with g.s. deformation $\beta_{2}$= -0.407) was coupled to the g.s. The results are also shown in Fig. 2. Though it yields a reasonable fit to the experimental data there is still a 25-40$\%$ under prediction in the sub-barrier energy range $E_{c.m}$= 5.6-6.4 MeV. Effect of projectile deformation is seen to be small and is not shown. It is possible that other types of coupling, e.g., transfer and/or breakup are responsible for the remaining discrepancy. A recent observation Jing at sub-barrier energies showed that the product of the fusion cross section ($\sigma$) and the c.m. energy ($E$) for 60Ni+89Y falls much faster than the usually accepted exponential falloff. They analyzed this steep falloff in terms of the logarithmic derivative ($L$) of the product $\sigma$$E$ defind by $L(E)$= $dln$($\sigma$$E$)/$dE$=(1/$\sigma$$E$)[d($\sigma$$E$)/$dE$]. Their results showed a continuous increase with decreasing bombarding energy in contradiction to theoretical prediction with Wong’s prescription wong . This discrepancy was attributed by Hagino $et$ $al$. hag to a deviation of the parabolic shape of potential assumed by Wong wong , from the asymmetric shape of the Coulomb barrier and was explained by using a large diffuseness of the ion-ion potential. To investigate the nature of the fall of $\sigma$$E$ for our system we have plotted in Fig. 3 the experimental values of $L$ obtained from consecutive fusion data points together with the Wong prediction. Here also we find increasing $L$ with decreasing $E$ below the barrier while the theoretical prediction saturates to a constant value below the barrier. However the increase is not that steep as is observed by Jiang $et$ $al$. Jing . Figure 3: Experimental slope function L(E), extracted from measured fusion cross-section alongwith the theoretical prediction from Wong model (solid line). To summarize, we have experimentally found the excitation function for 7Li + 28Si at near- and mostly sub-barrier energies, for the first time, employing the usual characteristic $\gamma$-ray method. Below the barrier our results show some sort of enhancement when compared with the 1D BPM prediction. Introduction of coupling to target rotational motion improves the fit with experiment to some extent. Recently Shrivastava $et$ $al$. Asriv have advocated in their work on 7Li +65Co that neutron transfer is more probable than all other possible direct reactions and hence an $n$-transfer followed by fusion may be a possibility. Sub-barrier enhancement owing to $n$-transfer (with positive $Q$-value) has been shown by Zagrebaev NZag for the fusion of 6He with 206Pb. However Pakou $et$ $al$. Pako pointed out in their work on the direct and compound contribution in the reaction 7Li+28Si that $d$-transfer is the dominant mechanism at near-barrier energies. These imply that the picture is not yet clear. So it is necessary to do a detailed theoretical analysis (utilizing a more realistic coupled reaction channels model) introducing these possible couplings for a better and complete understanding of the phenomenon. We would like to acknowledge the Department of Atomic Energy (DAE) and Board of Research in Nuclear Science (BRNS) for the financial support under the project (grant no. 2002/37/38/BRNS). We also would like to thank B. Mallick of Institute of Physics (Bhubaneswar), A. K. Mitra and S. Chatterjee of Saha Institute of Nuclear Physics for their technical support during the experiment and J. Panja of the same Institute for the preparation of Si Target. ## References * (1) C. H. Dasso $et$ $al$., Nucl. Phys. A597, 473 (1996); C. H. Dasso and A. Vitturi., Phys. Rev. C50, R12 (1994). * (2) M. S. Hussein $et$ $al$., Phys. Rev. C47, 2398 (1993); M. S. Hussein $et$ $al$., Nucl. Phys. A588, 85c(1995). * (3) N. Takigawa, M. Kuratani and H. Sagawa., Phys. Rev. C47, R2470 (1993). * (4) K. Hagino $et$ $al$., Phys. Rev. C61, 037602 (2000). * (5) V. I. Zagrebaev, Phys. Rev. C67, 061601(R) (2003). * (6) M. S. Hussein $et$ $al$., Phys.Rev. C73, 044610 (2006). * (7) M. Dasgupta $et$ $al$., Phys. Rev. Lett. 82, 1395 (1999). * (8) Y. W. Wu $et$ $al$., Phys. Rev. C68, 044605 (2003). * (9) C. Signorini $et$ $al$., Eur. Phys. J A5, 7 (1999). * (10) V. Tripathi $et$ $al$., Phys. Rev. Lett. 88, 172701 (2002); Phys. Rev. C72, 017601 (2005). * (11) Yu. E. Penionzhkevich $et$ $al$., Phys. Rev. Lett. 96, 162701 (2006). * (12) S. B. Moraes $et$ $al$., Phys. Rev. C61, 064608 (2000). * (13) I. Padron $et$ $al$., Phys. Rev. C66, 044608(2002). * (14) R. M. Anjos $et$ $al$., Phys. Lett. B534, 45 (2002). * (15) C. Beck $et$ $al$., Phys. Rev. C67, 054602 (2003). * (16) M. Romoli $et$ $al$., Nucl. Phys. A746, 522c (2004). * (17) P. R. S. Gomes $et$ $al$., Phys. Lett. B601, 20 (2004); Phys. Rev. C71, 034608 (2005). * (18) M. C. S. Figueira $et$ $al$., Nucl. Phys. A561, 453 (1993). * (19) R. M. Anjos $et$ $al$., Phys. Rev. C42, 354 (1990). * (20) G. V. Marti $et$ $al$., Phys. Rev. C71, 027602 (2005). * (21) Mandira Sinha $et$ $al$., Phys. Rev. C76, 027603 (2007). * (22) E. Pülhofer., Nucl. Phys. A280, 267 (1977). * (23) R. A. Dayras $et$ $al$., Nucl. Phys. A265,153 (1976). * (24) A. Mukherjee $et$ $al$., Nucl. Phys. A645,13 (1999). * (25) K. Hagino $et$ $al$., Comput. Phys. Commun. 123, 143 (1999). * (26) C. Y. Wong., Phys. Rev. Lett. 31, 766 (1973). * (27) L. C. Vaz, J. M. Alexander and G. R. Satchler., Physics reports. 69, 373 (1981). * (28) C. L. Jiang $et$ $al$., Phys. Rev. Lett. 89, 052701 (2002). * (29) K. Hagino, N. Rowley and M. Dasgupta., Phys. Rev. C67, 054603 (2003). * (30) A. Shrivastava $et$ $al$., Phys. Lett. B633, 463 (2006). * (31) A. Pakou $et$ $al$., Phys. Rev. C71, 064602 (2005).	arxiv-papers	2008-08-06T09:38:24	2024-09-04T02:48:57.182695	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Mandira Sinha, H. Majumdar, P. Basu, Subinit Roy, R. Bhattacharya, M.\n Biswas, M. K. Pradhan, S. Kailas", "submitter": "Mandira Sinha", "url": "https://arxiv.org/abs/0808.0800" }
0808.0831	# The Mixmaster Universe in a Generalized Uncertainty Principle framework Marco Valerio Battisti battisti@icra.it ICRA - International Center for Relativistic Astrophysics Dipartimento di Fisica (G9), Università di Roma “Sapienza” P.le A. Moro 5, 00185 Rome, Italy Giovanni Montani montani@icra.it ICRA - International Center for Relativistic Astrophysics Dipartimento di Fisica (G9), Università di Roma “Sapienza” P.le A. Moro 5, 00185 Rome, Italy ENEA C.R. Frascati (Dipartimento F.P.N.), Via E. Fermi 45, 00044 Frascati, Rome, Italy ICRANET C.C. Pescara, P.le della Repubblica 10, 65100 Pescara, Italy ###### Abstract The Bianchi IX cosmological model is analyzed in a generalized uncertainty principle framework. The Arnowitt-Deser-Misner reduction of the dynamics is performed and a time-coordinate, namely the volume of the Universe, naturally arises. Such a variable is treated in the ordinary way while the anisotropies (the physical degrees of freedom) are described by a deformed Heisenberg algebra. The analysis of the model (passing through Bianchi I and II) is performed at classical level by studying the modifications induced on the symplectic geometry by the deformed algebra. We show that, the Universe can not isotropize because of the deformed Kasner dynamics, the triangular allowed domain is asymptotically stationary with respect to the particle (Universe) and its bounces against the walls are not interrupted by the deformed effects. Furthermore, no reflection law can be in general obtained since the Bianchi II model is no longer analytically integrable. This way, the deformed Mixmaster Universe can be still considered a chaotic system. ###### pacs: 04.60.Bc; 98.80.Qc; 11.10.Nx ## I Introduction The existence of a fundamental scale, by which the continuum space-time picture that we have used from our experience at large scales probably breaks down, may be taken as a general feature of any quantum theory of gravity (for a review see gar ). An intuitive approach to introduce such a cut-off is based on deforming the canonical uncertainty relations leading to the so-called generalized uncertainty principle111Over the paper we adopt units such that $\hbar=c=16\pi G=1$. (GUP) $\Delta q\Delta p\geq\frac{1}{2}\left(1+\beta(\Delta p)^{2}+\beta\langle{\bf p}\rangle^{2}\right),$ (1) where $\beta>0$ is a deformation parameter such that for $\beta=0$ the ordinary relation is recovered. The uncertainty principle (1) has appeared in perturbative string theory String , considerations on the proprieties of black holes Mag and de Sitter space Sny . From the string theory point of view, a minimal observable length is a consequence of the fact that strings can not probe distances below the string scale. The relation (1) implies a finite minimal uncertainty in the position $\Delta q_{0}=\sqrt{\beta}$ and therefore this approach entails a minimal scale in the quantum framework. However, the cut-off predicted by the GUP is, by its nature, different from the minimal length predicted by other approaches, for example the minimal eigenvalue of the geometric operators in loop quantum gravity lqg . Recently, such an approach received notable interest and a wide work has been made on this field in a large variety of directions (see for example GUP and the references therein). The generalized uncertainty principle (1) can be immediately reproduced modifying the canonical Heisenberg algebra by the following one Kem1 ; Kem2 $[{\bf q},{\bf p}]=i(1+\beta{\bf{p}}^{2}).$ (2) Although such a deformed commutation relation, differently from the GUP itself, has not been so far derived directly from string theory, it represents a possible way in which certain features of a more fundamental theory may manifest themselves in some mechanical models. In this work we analyze the Bianchi IX cosmological model (the Mixmaster Universe) in the GUP framework. This study improves a research line of ours which is centered in the investigation of cosmological models with a minimal scale BM07c ; BM07a ; BM07b . The Friedmann-Robertson-Walker Universe filled with a scalar field and the Taub model have been analyzed in previous works. In the first case BM07a , the big-bang singularity appears to be probabilistically removed but no evidences for a big-bounce (as predicted by the loop approach APS ) arise. In the second case BM07b also, the Universe is singularity-free and furthermore the GUP wave packets provide the right behavior in the establishment of a quasi-isotropic configuration for the model. The Bianchi IX model, together with Bianchi VIII, is the most general homogeneous model and its physical relevance relies on the fact that it represents a general solution of the Einstein equations toward the singularity BKL . In fact, via the Belinski-Khalatnikov-Lifshitz (BKL) scenario, when the cosmological singularity is approached in the context of a generic inhomogeneous framework, the spatial points (causal horizons) dynamically decouple and each of them evolves independently as a Bianchi IX model BKL . The approach to the singularity of the Mixmaster model is described by a particle in two dimensions (the two physical degree of freedom of the Universe, i.e. the anisotropies) moving in a potential having exponential walls bounding a triangle Mis . Such a particle is reflected by the walls and the dynamics appears to be chaotic chaos . Such a model has been then used to describe the (classical) physics near the cosmological singularity. The application of the GUP framework in quantum cosmology is well-motivated. By the minisuperspace reduction, a genuine quantum field theory (quantum general relativity) reduces to a quantum mechanical system (homogeneous quantum cosmology). As well-known, the homogeneous models, in the vacuum case, are characterized by only three degrees of freedom and therefore they are nothing but three-dimensional mechanical systems. In this respect, the GUP approach to quantum cosmology appears to be physically grounded since it can be reproduced modifying the canonical Heisenberg algebra. The Bianchi IX model will be studied in the context of the Arnowitt-Deser- Misner (ADM) reduction of the dynamics (for a review see rev ). Such a representation allows us to regard one variable, mainly the Universe volume, as a time for the dynamics. This model will be described by the motion of a two-dimensional particle in a triangular allowed domain. These variables, describing the physical degrees of freedom of the system, will be treated in the GUP formalism, while the time-variable in a canonical way. To perform the analysis two necessary steps, i.e. the study of the Bianchi I and II cosmological models, are necessary. The main results we obtain are in order. (i) The Bianchi I dynamics is still Kasner-like but is deeply modified since the GUP effects act in an opposite way with respect to a massless scalar field. Moreover, the deformed particle (Universe) moves faster than the ordinary case and when the Universe shrinks toward the singularity, the distances can contract along one direction while growing along the other two, i.e. two negative Kasner indices are allowed. (ii) The Bianchi II model is no longer analytically integrable and therefore no BKL map can be obtained. In other words, a relation which describes the details of the bounce of the particle against the potential walls can not be analytically found. (iii) The potential walls of Bianchi IX become stationary with respect to the particle when its momentum is of the same order of the cut-off. The triangular domain is “dynamically closer” than the standard one and no way for the particle to escape from the bounces arises. We conclude that the deformed evolution of the Mixmaster Universe is still chaotic. The paper is organized as follows. In Section II the Bianchi cosmological models are reviewed and the deformed picture applied to them. Section III and IV are devoted to the analysis of the Bianchi I and II models in the GUP scheme, respectively. In Section V the deformed Bianchi IX model is investigated. Concluding remarks follow. ## II Deformed Bianchi models In this Section we discuss how the equations of motion for the Bianchi models are modified by a minimal cut-off on the anisotropies. We analyze the deformations induced on the (reduced) phase space by a generalization of (2) in which both the two degrees of freedom of the Universe have a non-zero minimal uncertainty. The Bianchi Universes are spatially homogeneous cosmological models such that the symmetry group acts simply transitively222Let $G$ a Lie group, $G$ is said to act simply transitively on the spatial manifold $\Sigma$ if, for all $p,q\in\Sigma$, there is a unique element $g\in G$ such that $g(p)=q$. on each spatial manifold RS . The dynamics of these models is summarized in the scalar constraint which, in the Misner scheme Mis , reads $H=-p_{\alpha}^{2}+p_{+}^{2}+p_{-}^{2}+e^{4\alpha}V(\gamma_{\pm})=0,$ (3) where the lapse function $N=N(t)$ has been fixed by the time gauge $\dot{\alpha}=1$ as $N=-e^{3\alpha}/2p_{\alpha}$. The variable $\alpha=\alpha(t)$ describes the isotropic expansion of the Universe while its shape changes (the anisotropies) are determinated via $\gamma_{\pm}=\gamma_{\pm}(t)$. Therefore, homogeneity reduces the phase space of general relativity to six dimensions. In this framework the cosmological singularity appears for $\alpha\rightarrow-\infty$ and the differences between the Bianchi models are summarized in the potential term $V(\gamma_{\pm})$ which is related to the three-dimensional scalar of curvature. To describe the time evolution of the models a choice of time has to be performed. As well-known Ish in general relativity it is possible to trace the dynamics in a relational way (with respect to an other field) or with respect to an internal time which is constructed from phase space variables. The ADM reduction of the dynamics relies on the idea to solve the scalar constraint with respect to a suitably chosen momentum. This way, we obtain an effective Hamiltonian which depends only on the physical degrees of freedom of the system. Since the volume $\mathcal{V}$ of the Universe is $\mathcal{V}\propto e^{3\alpha}$, the variable $\alpha$ can be regarded as a good clock for the evolution and therefore the ADM picture arises as soon as the constraint (3) is solved with respect to $p_{\alpha}$. Explicitly, we obtain $-p_{\alpha}=\mathcal{H}=\left(p_{+}^{2}+p_{-}^{2}+e^{4\alpha}V(\gamma_{\pm})\right)^{1/2},$ (4) where $\mathcal{H}$ is a time-dependent Hamiltonian from which is possible to extract, for a given symplectic structure, all the dynamical informations about the homogeneous cosmological models. Let us now analyze the modifications induced on the phase space by the GUP approach. In particular, we consider the $N$-dimensional generalization of the relation (2) as Kem2 $[{\bf q}_{i},{\bf p}_{j}]=i\delta_{ij}(1+\beta{\bf p}^{2})+i\beta^{\prime}{\bf p}_{i}{\bf p}_{j},\qquad{\bf p}^{2}={\bf p}_{i}{\bf p}^{i},$ (5) $\beta^{\prime}>0$ being a new parameter. Furthermore, assuming that the translation group is not deformed, i.e. $[{\bf p}_{i},{\bf p}_{j}]=0$, the commutation relations among the coordinates are almost uniquely determined by the Jacobi identity. The deformed classical dynamics is thus summarized in the modified symplectic geometry arising from the classical limit of the quantum- mechanical commutators, as soon as the parameters $\beta$ and $\beta^{\prime}$ are regarded as independent constants with respect to $\hbar$. Therefore, the phase space algebra we consider is the one in which the fundamental Poisson brackets are Ben $\displaystyle\\{q_{i},p_{j}\\}$ $\displaystyle=$ $\displaystyle\delta_{ij}(1+\beta p^{2})+\beta^{\prime}p_{i}p_{j},$ (6) $\displaystyle\\{p_{i},p_{j}\\}$ $\displaystyle=$ $\displaystyle 0,$ $\displaystyle\\{q_{i},q_{j}\\}$ $\displaystyle=$ $\displaystyle\frac{(2\beta-\beta^{\prime})+(2\beta+\beta^{\prime})\beta p^{2}}{1+\beta p^{2}}(p_{i}q_{j}-p_{j}q_{i}).$ From a string theory point of view, keeping the parameters $\beta$ and $\beta^{\prime}$ fixed as $\hbar\rightarrow 0$ corresponds to keeping the string momentum scale fixed while the string length scale shrinks to zero String . In order to obtain the deformed Poisson bracket, some natural requirements have to be considered. It must posses the same proprieties as the quantum mechanical commutator, i.e. it has to be anti-symmetric, bilinear and satisfy the Leibniz rules as well as the Jacobi identity. This way, the Poisson bracket for any phase space function reads $\\{F,G\\}=\left(\frac{\partial F}{\partial q_{i}}\frac{\partial G}{\partial p_{j}}-\frac{\partial F}{\partial p_{i}}\frac{\partial G}{\partial q_{j}}\right)\\{q_{i},p_{j}\\}+\frac{\partial F}{\partial q_{i}}\frac{\partial G}{\partial q_{j}}\\{q_{i},q_{j}\\}.$ (7) It is worth noting, that for $\beta^{\prime}=2\beta$ the coordinates $q_{i}$ become commutative up to higher order corrections, i.e. $\\{q_{i},q_{j}\\}=0+\mathcal{O}(\beta^{2})$ and the isotropic minimal uncertainty in position reads $\Delta q_{0}=2\sqrt{\beta}$. This can be considered a preferred choice of parameters and from now on we analyze this case. However, although we neglect terms like $\mathcal{O}(\beta^{2})$, the case in which $\beta p^{2}\gg 1$ is allowed since in such a framework no restrictions on the $p$-domain arise, i.e. $p\in\mathbb{R}$. The deformed classical dynamics of the Bianchi models can be obtained from the symplectic algebra (6) for $\beta^{\prime}=2\beta$. The time evolution of the anisotropies and momenta, with respect to the ADM Hamiltonian (4), is thus given by ($i,j=\pm$) $\displaystyle\dot{\gamma}_{i}$ $\displaystyle=$ $\displaystyle\\{\gamma_{i},\mathcal{H}\\}=\frac{1}{\mathcal{H}}\left[(1+\beta p^{2})\delta_{ij}+2\beta p_{i}p_{j}\right]p_{j},$ (8) $\displaystyle\dot{p}_{i}$ $\displaystyle=$ $\displaystyle\\{p_{i},\mathcal{H}\\}=-\frac{e^{4\alpha}}{2\mathcal{H}}\left[(1+\beta p^{2})\delta_{ij}+2\beta p_{i}p_{j}\right]\frac{\partial V}{\partial\gamma_{j}},$ where the dot denotes differentiation with respect to the time variable $\alpha$ and $p^{2}=p_{+}^{2}+p_{-}^{2}$. These are the deformed equations of motion for the homogeneous Universes and the ordinary ones are recovered in the $\beta=0$ case. In what follow such a dynamics for the Bianchi I, II and IX models will be investigated in detail. ## III Deformed Bianchi I model The Bianchi I model is the simplest homogeneous model and describes a Universe with flat space sections rev ; RS . Its line element is invariant under the group of three-dimensional translations and therefore the spatial Cauchy surfaces can be identified with $\mathbb{R}^{3}$. This Universe contains as a special case the flat FRW model which is obtained as soon as the isotropy condition is taken into account. In the above scheme, this Universe corresponds to the case $V(\gamma_{\pm})=0$ and thus, from the Hamiltonian (4), it is described by a two-dimensional free particle (more precisely a massless scalar relativistic particle). The deformed equations of motion (8) are immediately solved by $\dot{\gamma}_{\pm}=C_{\pm}(\beta),\qquad\dot{p}_{\pm}=0,$ (9) $C(\beta)$ being a function of $\beta$. Therefore, the solution is Kasner- like. The velocity of the particle (Universe), however, is modified by the deformed geometry and, from the first equation of (8), it reads $\dot{\gamma}^{2}=\dot{\gamma}_{+}^{2}+\dot{\gamma}_{-}^{2}=\frac{p^{2}}{\mathcal{H}^{2}}\left(1+6\mu+9\mu^{2}\right)=1+6\mu+9\mu^{2},$ (10) where $\mu=\beta p^{2}$. In the last step we have used the fact the for the Binchi I model the ADM Hamiltonian (4) is given by $\mathcal{H}^{2}=p^{2}=$ const. As expected, for $\beta\rightarrow 0$ (or better when $\mu\ll 1$), the standard Kasner velocity $\dot{\gamma}^{2}=1$ is recovered. The effects of an anisotropies cut-off imply that the particle moves faster than the ordinary case. Let us now analyze how the Kasner behavior is modified by the deformed framework. As well-known rev ; RS , the Kasner solution is such that the spatial metric reads $dl^{2}=t^{2s_{1}}dx_{1}^{2}+t^{2s_{2}}dx_{2}^{2}+t^{2s_{3}}dx_{3}^{2},$ (11) where $s_{1},s_{2},s_{3}$ are the so-called Kasner indices satisfying the relations $s_{1}+s_{2}+s_{3}=1$ and $s_{1}^{2}+s_{2}^{2}+s_{3}^{2}=1$. Only one of them is independent and except for the case $(0,0,1)$ and $(-1/3,2/3,2/3)$, such indices are never equal, but one negative and two positive. It is worth noting that the first Kasner-relation arises from the arbitrariness in choosing the tetrads, and thus is still valid in the deformed approach, while the second one is directly related to the anisotropy velocity $\dot{\gamma}$ by the equations rev $\dot{\gamma}_{+}=\frac{1}{2}(1-3s_{3}),\qquad\dot{\gamma}_{-}=\frac{\sqrt{3}}{2}(s_{1}-s_{2}).$ (12) From formula (10), the second Kasner-relation is then deformed as $s_{1}^{2}+s_{2}^{2}+s_{3}^{2}=1+4\mu+6\mu^{2},$ (13) and, as usual, for $\beta=0$ the standard one is recovered. Two remarks are in order. (i) From this equation, it is easy to verify that the GUP acts in an opposite way as a massless scalar field (or alternatively stiff-fluid with pressure equal to density) in the standard model. In the ordinary case, a massless scalar field allows only a finite number of oscillations in Bianchi IX before the evolution is changed into a state in which all directions shrink monotonically to zero as the curvature singularity is reached. In this case the chaotic behavior of the Mixmaster is tamed scafield . On the other hand, in the GUP framework, all the terms on the right hand side of (13) are positive and it means that the Universe cannot isotropize, i.e. it can not reach the stage such that the Kasner indices are equal. (ii) For every non- zero $\mu$, the modifications induced on the standard Kasner behavior are significant since two indices can be negative at the same time. In other words, as the volume of the Universe contracts toward the classical singularity, distances can shrink along one direction and grow along the other two. In the ordinary case the contraction is along two directions. Therefore, even if a “quasi-standard” regime ($\mu\ll 1$) is addressed, the Kasner dynamics is deeply modified by such an approach. ## IV Deformed Bianchi II Model Let us now investigate the dynamics of Bianchi II in the framework of the deformed phase space discussed above. This model is a fundamental step toward the Bianchi IX one. It represents a bridge from the flat homogeneous model (Bianchi I) and the Mixmaster Universe (Bianchi IX). Its dynamics is the one of a two-dimensional particle bouncing against a single wall. More precisely, it corresponds to the Mixmaster dynamics when only one of the three equivalent potential walls is taken into account rev ; RS . The main features of Bianchi IX, as the BKL map, are obtained considering such a simplified model since it is, in the ordinary framework, an integrable system differently from Bianchi IX itself Misq . In the Hamiltonian framework, Bianchi II is the homogeneous model for which the potential term is given by $V(\gamma_{\pm})=e^{-8\gamma_{+}}$ and such an expression can be obtained from the one of Bianchi IX in a given asymptotic region. The ADM Hamiltonian (4) in this case reads $\mathcal{H}=\left(p_{+}^{2}+p_{-}^{2}+e^{4(\alpha-2\gamma_{+})}\right)^{1/2}.$ (14) Our aim is to describe the bounce of the particle (Universe) against the potential wall in the GUP scheme. A fundamental difference with respect to the ordinary case is that $\mathcal{H}$ is no longer a constant of motion near the classical singularity ($\alpha\rightarrow-\infty$). In the undeformed scheme the anisotropy velocities are simply given by $\dot{\gamma}_{\pm}=p_{\pm}/\mathcal{H}$ and, from (14), one immediately obtains $\dot{\mathcal{H}}=0$ for $\alpha\rightarrow-\infty$. In our scheme such a feature is modified by the deformation terms and, in particular, by the velocity relation (10) which replaces the ordinary one $\dot{\gamma}^{2}=p^{2}/\mathcal{H}^{2}$. The equation $\dot{\mathcal{H}}=\partial_{\alpha}\mathcal{H}$ in the deformed framework gives $\frac{\partial}{\partial\alpha}(\ln\mathcal{H}^{2})=4\left(1-\frac{p^{2}(\dot{\gamma})}{\mathcal{H}^{2}}\right)\neq 4\left(1-\dot{\gamma}^{2}\right),$ (15) and is no longer equal to zero. Of course, with $p(\dot{\gamma})$ we have indicated the solution of the velocity equation (10). From the above relation (15) it is possible to compute the velocity of the potential wall. The condition that the potential term $V(\gamma_{\pm})=e^{-8\gamma_{+}}$ be important near the cosmological singularity is easily seen to be $e^{4(\alpha-2\gamma_{+})}\simeq\mathcal{H}^{2}$. The potential (wall) velocity $\dot{\gamma}_{w}$ then reads $\dot{\gamma}_{+}\simeq\dot{\gamma}_{w}=\frac{1}{2}-\frac{1}{8}\frac{\partial}{\partial\alpha}(\ln\mathcal{H}^{2})=\frac{p^{2}(\dot{\gamma})}{2\mathcal{H}^{2}}.$ (16) As in the ordinary case, $\gamma_{w}$ defines the equipotentials in the anisotropy plane, where the potential term is relevant. In the standard picture, since $\dot{\gamma}^{2}=1$, the wall velocity is equal to one half of the particle one, i.e. $\dot{\gamma}_{w}=1/2$. The undeformed dynamics toward the classical singularity ($\alpha\rightarrow-\infty$) is as follows. The anisotropy particle $\gamma(\alpha)$ moves with velocity $\dot{\gamma}=1$ except when it approaches the equipotentials $e^{4(\alpha-2\gamma_{+})}\simeq\mathcal{H}^{2}$. This wall moves outward with a velocity $\dot{\gamma}_{w}=1/2$ and therefore, in a finite time interval, the particle will bounce against it. The dynamics of the particle before and after this collision is the one described by the Bianchi I model. In the deformed case, both the particle and the potential wall move faster than the ordinary one. The main point is to establish if there exists a range in which the wall moves faster than the anisotropy particle. In fact, in such a case, the point-Universe no longer bounces against the wall and the Kasner behavior remains unaltered. From the above relations (10) and (16), it is possible to derive the explicit form of the wall velocity, which reads (near the singularity) $\dot{\gamma}_{w}=\frac{1}{36\mu}\left(-4+2^{1/3}2g^{-1/3}+2^{2/3}g^{1/3}\right),$ (17) where $g=g(\mu)$ is defined as $g=2+81\mu\dot{\gamma}^{2}+9\sqrt{\mu\dot{\gamma}^{2}(4+81\mu\dot{\gamma}^{2})}$. We stress that, near the cosmological singularity ($\alpha\rightarrow-\infty$) we have $\mathcal{H}^{2}\simeq p^{2}$ and therefore the particle velocity $\dot{\gamma}$ is the same as in the Bianchi I case. Moreover, it is not difficult to see that for $\beta\rightarrow 0$ the ordinary velocity of the potential wall is recovered and that the bounce always occurs also in the deformed scheme. Let us now discuss the details of the bounce. In the standard case the particle (Universe) moves twice as fast as the receding potential wall, independently of its momentum (namely its energy). In the deformed framework the particle velocity, as well as the potential velocity, depends on the anisotropy momenta and on the deformation parameter $\beta$. In this case also the particle moves faster than the wall, since the relation $\dot{\gamma}_{w}<\dot{\gamma}$ is always verified. Thus, a bounce takes place also in the deformed picture. Furthermore, the wall appears asymptotically stationary when the particle has a growing energy, i.e. when the region $\mu\gg 1$ is investigated (we recall that $p\in\mathbb{R}$). In this limit, the relation $\dot{\gamma}_{w}/\dot{\gamma}\sim 1/(6\mu)$ holds (see Fig. 1) and then, for $\mu\gg 1$, no limit angle for the collision appears. More precisely, let us indicate with $\theta_{i}$ and $\theta_{f}$ the angles of incidence and of reflection for the bounce, respectively. The velocity $\dot{\gamma}$ is parametrized as follows Misq : in the initial state we have $(\dot{\gamma}_{+})_{i}=-\dot{\gamma}\cos\theta_{i}$, $(\dot{\gamma}_{-})_{i}=\dot{\gamma}\sin\theta_{i}$ and in the final one $(\dot{\gamma}_{+})_{f}=\dot{\gamma}\cos\theta_{f}$, $(\dot{\gamma}_{-})_{f}=\dot{\gamma}\sin\theta_{f}$. Thus, the maximum angle in order the bounce against the wall to occur is given by $\|\theta_{i}\|<\|\theta_{\max}\|=\cos^{-1}(\dot{\gamma}_{w}/\dot{\gamma})$ and hence, in the asymptotic limit $\mu\gg 1$, we have $\|\theta_{\max}\|=\pi/2$. In the ordinary case ($\dot{\gamma}_{w}/\dot{\gamma}=1/2$), the maximum incidence angle is given by $\|\theta_{\max}\|=\pi/3$ Misq . Figure 1: The potential wall velocity $\dot{\gamma}_{w}$ with respect to the particle one $\dot{\gamma}$ in function of $\mu=\beta p^{2}$. In the $\mu\rightarrow 0$ limit, the ordinary behavior $\dot{\gamma}_{w}/\dot{\gamma}=1/2$ is recovered. The next step would be to obtain the reflection law (the BKL map) which connects the initial ($\theta_{i}$) and final ($\theta_{f}$) between the particle-velocity and the wall. In order to integrate the model (14), we have to recover two first integrals of motion. In the ordinary case, one of them is immediately found in $p_{-}$. On the other hand, in the deformed framework, the equations of motion (8) become coupled by $\beta$-terms and in this case $\dot{p}_{-}=\frac{8\beta}{\mathcal{H}}p_{+}p_{-}e^{4(\alpha-2\gamma_{+})},$ (18) which no longer vanishes unless the $\beta=0$ case is considered. In the undeformed scheme, the other constant of motion can be recovered by a linear combination of $p_{+}$ and $\mathcal{H}$, in particular by $\Omega=\mathcal{H}-p_{+}/2$. From such constants of motion, is not difficult to obtain the reflection law as $2(\sin\theta_{f}-\sin\theta_{i})=\sin(\theta_{i}+\theta_{f})$ Misq . Contrarily, in the deformed picture, the remaining equations of motion read $\displaystyle\dot{p}_{+}$ $\displaystyle=$ $\displaystyle\frac{4}{\mathcal{H}}e^{4(\alpha-2\gamma_{+})}(1+3\beta p_{+}^{2}+\beta p_{-}^{2}),$ (19) $\displaystyle\dot{\mathcal{H}}$ $\displaystyle=$ $\displaystyle\frac{2}{\mathcal{H}}e^{4(\alpha-2\gamma_{+})},$ and $\Omega$ is a constant of motion if and only if the $\beta=0$ case is taken into account. This way, differently from the standard case, the Bianchi II model in such a framework appears to be a non (analytically) integrable system. As a matter of fact no first integrals of motion can be recovered, i.e. an equation which gives $\theta_{f}$ in terms of $\theta_{i}$ can not be in general obtained. To gain insight onto the physical features of the model, we can consider the special cases for which a reflection law can be fixed. We analyze the different situations in which $p_{+}\gg p_{-}$ ($p_{-}\gg p_{+}$) and $p_{+}=p_{-}$. In the first case333The complementary case ($p_{-}\gg p_{+}$) is qualitatively the same. ($p_{+}\gg p_{-}$ corresponds to $\|(\dot{\gamma}_{+})_{i}\|\gg\|(\dot{\gamma}_{-})_{i}\|$), two constants of motion can be obtained and read $\displaystyle\Omega$ $\displaystyle=$ $\displaystyle\mathcal{H}-\frac{1}{2\sqrt{3\beta}}\tan^{-1}(\sqrt{3\beta}p_{+}),$ (20) $\displaystyle K$ $\displaystyle=$ $\displaystyle\frac{1}{3}\ln(1+3\beta p_{+}^{2})-\ln(p_{-}).$ and as $\mu\rightarrow 0$ the ordinary framework is recovered. By the use of the equations of motion $\dot{\gamma}_{\pm}$ given by (8), it is possible to obtain the required reflection law between $\theta_{i}$ and $\theta_{f}$. An interesting feature appears as soon as the ultra-deformed case is considered, i.e. when $\mu\gg 1$. In such a range, the two constants of motion (20) become $\Omega\simeq\mathcal{H}$ and $K\simeq\ln{(3\beta p_{+}^{2})}/3$ and thus the reflection law is given by $\|\theta_{i}\|=\|\theta_{f}\|$ as in the usual framework. In the second peculiar case ($p_{+}=p_{-}$), the two first integrals of motion are $\displaystyle\Omega$ $\displaystyle=$ $\displaystyle\mathcal{H}-\frac{1}{4\sqrt{\beta}}\tan^{-1}(2\sqrt{\beta}p_{+}),$ (21) $\displaystyle K$ $\displaystyle=$ $\displaystyle\frac{p_{+}}{2}+\frac{1}{4\sqrt{\beta}}\tan^{-1}(2\sqrt{\beta}p_{+}).$ Also in this case, considering equations (8) the map $\theta_{f}=\theta_{f}(\theta_{i})$ can be obtained and in the ultra-deformed regime ($\mu\gg 1$) it reads $\|\theta_{i}\|=\|\theta_{f}\|$ as in the ordinary scheme. Let us now summarize the effects of the deformed framework on the dynamics of the Bianchi II model. The main difference with respect to the ordinary picture is that such a model is no longer integrable and therefore no reflection map can be in general inferred. It can be obtained only in few peculiar cases. The other important feature of our model is that the potential wall becomes stationary, with respect to the $\gamma$-particle, in the asymptotic regime $\mu\gg 1$ (see Fig. 1). Therefore, when the momenta of the particle (Universe) reaches the cut-off value $\beta$, its bounce against the wall is improved in the sense that no longer maximum limit angle appears. ## V Deformed Mixmaster Universe In this section we describe the deformed Mixmaster Universe. We analyze the deformed phase space of the Bianchi IX cosmological model in agreement with the previous discussion. The Bianchi IX geometry is invariant under the three- dimensional rotation group and therefore the space-time manifold can be topologically written as $\mathcal{M}=\mathbb{R}\otimes SO(3)$ rev ; RS . Thus, this Universe is the generalization of the closed FRW model when the isotropy hypothesis is relaxed. In the Hamiltonian formulation, it appears as soon as the potential term $V(\gamma_{\pm})=e^{4(\gamma_{+}+\sqrt{3}\gamma_{-})}+e^{4(\gamma_{+}-\sqrt{3}\gamma_{-})}+e^{-8\gamma_{+}}$ in (4) is taken into account. Such a potential delimits a triangular domain in the $\gamma$-plane where the dynamics is restricted Mis ; Misq . As well known, the evolution of the Mixmaster Universe is that of a two-dimensional particle bouncing infinite times against three walls which rise steeply toward the singularity. In particular, between two succeeding bounces the system is described by the Kasner evolution and the permutations of the expanding- contracting directions is given by the BKL map BKL . Such a dynamics is also chaotic chaos . From the analysis of the deformed Bianchi I and II models we know several features of the deformed Mixmaster Universe. Inside the closed domain the $\gamma$-particle moves freely and therefore its velocity is given by the formula (10). The Bianchi II model, appearing as soon as only one of the three equivalent walls is taken into account, is recovered when the asymptotic region $\gamma_{+}\rightarrow-\infty$, $\|\gamma_{-}\|<-\sqrt{3}\gamma_{+}$ of the Bianchi IX model is considered. The velocity of the potential walls is then the same as previously computed, see equation (17). In Bianchi II, because of the presence of a single potential wall, the particle performs only one bounce and then it runs freely toward the singularity. Differently, in the Bianchi IX case the particle will collide infinite times against the three walls. Two conclusions on the deformed Mixmaster Universe can be inferred. * • When the ultra-deformed regime is reached ($\mu\gg 1$), i.e. when the $\gamma$-particle (Universe) has the momentum bigger than the cut-off one, the triangular closed domain appears to be stationary with respect to the particle itself. This way, the bounces of such a particle are increased by the presence of deformation terms, i.e. by the non-zero minimal uncertainty in the anisotropies. * • No BKL map (reflection law) can be in general obtained. It arises analyzing the single bounce against a given wall of the equilateral-triangular domain and the Bianchi II model is no longer an integrable system in the deformed picture. In other words, the chaotic behavior of the Bianchi IX model is not tamed by GUP effects, i.e. the deformed Mixmaster Universe is still a chaotic system. We have to stress a point. The chaoticity of Bianchi IX arises from the analysis of the stochastic proprieties of such a model and in particular from the BKL map. As we have seen, no reflection law for the Bianchi II model can be obtained in the deformed framework. Therefore, since the BKL map is constructed from the reflection law of Bianchi II, no deformed map arises at all and no quantitative predictions can be made for the model. We can however use qualitative arguments to realize that the chaoticity of Bianchi IX is not tamed by the GUP effects. In the ordinary framework, Bianchi II is an integrable model which is a part of a chaotic system (Bianchi IX). On the other hand, in the deformed framework Bianchi II, which is still a part of Bianchi IX, is no longer an integrable model. The deformations induced by a minimal uncertainty make the model much more complicated and surely they are not able to cast a chaotic system in a non-chaotic one. This is the meaning when we claim that the deformed Mixmaster Universe is still a chaotic system. It is worth noting that effects due to two negative Kasner indices arise in the modified dynamics. This new issue would require additional investigation, but it seems no way related to a possible removal of the chaoticity of this model. It is interesting to point out the differences between our model and the loop Mixmaster dynamics Mixl . Loop quantum cosmology lqc is based on the discrete structure of space predicted by loop quantum gravity. When such a framework is applied to the Bianchi IX model the classical reflections of the $\gamma$-particle stop after a finite amount of time and, when the Planck scale is reached, the height of the potential walls rapidly decreases until they completely disappear. This way, the Mixmaster chaos is suppressed by (loop) quantum effects Mixl . In the loop framework, although the analysis is performed through the ADM reduction of the dynamics as we did, all the three scale factors are quantized using the loop techniques. On the other hand, in our approach the time variable (related to the volume of the Universe) is treated in the standard way and only the two physical degrees of freedom of the Universe (the anisotropies) are considered as deformed. This makes clear the differences between these two cut-off approaches. In fact, we expect that if we implement the deformed framework to the whole phase space, modifications on the Universe volume, i.e. on the height of the potential walls, can appear. However, on the basis of bat , to reproduce the loop phenomenology a Snyder- deformed Heisenberg algebra should be taken into account. ## VI Concluding remarks In this paper we have shown the effects of a modified Heisenberg algebra, which reproduces a GUP as arises from studies on string theory String , on the Bianchi I, II and IX cosmological models (for other analysis of low-energy- string-effective cosmological models see lowstring ). The dynamics of these Universes is analyzed in the ADM formalism by which the variable $\alpha$ (namely the volume of the Universe) is regarded as the time-coordinate for the dynamics. Such a time variable is described in the standard way. On the other hand, the two physical degrees of freedom of the Universe (the shape changes $\gamma_{\pm}$) are treated according to the GUP prescription, i.e. by using a deformed Heisenberg algebra. A fundamental scale is then introduced in these models by the appearance of a non-zero minimal uncertainty in the anisotropies. The analysis of the dynamics is performed at classical level taking into account the modifications induced on the phase space by the deformed algebra. In particular, the deformed dynamics of the particle as well as of the potential walls is investigated in detail. Three main conclusions can be inferred. * • The velocity of the $\gamma$-particle (Universe) inside the allowed domain of the Mixmaster model grows with respect to the undeformed case. The deformation effects, acting as opposite to a stiff-matter, imply that the Universe cannot isotropize. Furthermore, although the dynamics is still Kasner-like, two negative Kasner indices are now allowed. During each Kasner era, the volume of the Universe can contract in one direction while expands in the other two. * • The velocity $\dot{\gamma}_{w}$ of the potential walls, bounding the triangular domain of Bianchi IX, is increased by the deformation terms. However, it no rises so much to avoid the bounces of the $\gamma$-particle against the walls, i.e. the particle bounces are not stopped by the GUP effects. As matter of fact, when the ultra-deformed regime is reached (when $\mu\gg 1$) the dynamics is that of a particle which bounces against stationary walls (no maximum incidence angle appears). * • No BKL map (reflection law $\theta_{f}=\theta_{f}(\theta_{i})$) can be in general analytically computed. In fact, such a map arises from the analysis of the Bianchi II model which is no longer analytically integrable in the deformed scheme. A non-vanishing minimal uncertainty in the anisotropies complicates so much the Mixmaster dynamics in such a way that each of its wall-side is no longer an integrable system. We can then conclude that the chaoticity of the Bianchi IX model is not tamed by the GUP effects on the Universe anisotropies. ## References * (1) L.J.Garay, Int.J.Mod.Phys.A 10 (1995) 145, gr-qc/9403008. * (2) D.J.Gross and P.F.Mendle, Nucl.Phys.B 303 (1988) 407; E.Witten, Phys.Today 49 (1996) 24; D.J.Gross, Phys.Rev.Lett. 60 (1988) 1229; D.Amati, M.Ciafaloni and G.Veneziano, Phys.Lett.B 216 (1989) 41; G.Amelino-Camelia et al., Mod.Phys.Lett.A 12 (1997) 2029. * (3) M.Maggiore, Phys.Lett.B 304 (1993) 65; Phys.Rev.D 49 (1994) 5182, hep-th/9305163. * (4) H.S.Snyder, Phys.Rev. 71 (1947) 38. * (5) C.Rovelli, Quantum Gravity (CUP, Cambridge, 2004); T.Thiemann, Modern Canonical Quantum General Relativity (CUP, Cambridge, 2007). * (6) I.Dadic, L.Jonke and S.Meljanac, Phys.Rev.D 67 (2003) 087701, hep-th/0210264; A.Kempf and L.Lorenz, Phys.Rev.D 74 (2006) 103517; A.Ashoorioon, A.Kempf and R.B.Mann, Phys.Rev.D 71 (2005) 023503; L.N.Chang et al., Phys.Rev.D 65 (2002) 125027; F.Brau and F.Buisseret, Phys.Rev.D 74 (2006) 036002; J.Y.Bang and M.S.Berger, Phys.Rev.D 74 (2006) 125012; M.M.Stetsko, Phys.Rev.A 74 (2006), 062105; B.Vakili, Phys.Rev.D 77 (2008) 044023, arXiv:0801.2438; B.Vakili and H.R.Sepangi, Phys.Lett.B 651 (2007) 79, arXiv:0706.0273; N.Khosravi and H.R.Sepangi, Phys.Lett.A 372 (2008) 3356, arXiv:0802.0767. * (7) A.Kempf, G.Mangano and R.B.Mann, Phys.Rev.D 52 (1995) 1108, hep-th/9412167. * (8) A.Kempf, J.Phys.A 30 (1997) 2093, hep-th/9604045; A.Kempf and G.Mangano, Phys.Rev.D 55 (1997) 7909, hep-th/9612084. * (9) M.V.Battisti and G.Montani, AIP Conf.Proc. 966 (2008) 219, arXiv:0709.4610; Int.J.Mod.Phys.A 23 (2008) 1257, arXiv:0802.0688. * (10) M.V.Battisti and G.Montani, Phys.Lett.B 656 (2007) 96, gr-qc/0703025. * (11) M.V.Battisti and G.Montani, Phys.Rev.D 77 (2008) 023518, arXiv:0707.2726. * (12) A.Ashtekar, T.Pawlowski and P.Singh, Phys.Rev.Lett. 96 (2006) 141301; Phys.Rev.D 73 (2006) 124038, gr-qc/0604013. * (13) V.A.Belinski, I.M.Khalatnikov and E.M.Lifshitz, Adv.Phys. 19 (1970) 525; Adv.Phys. 31 (1982) 639. * (14) C.Misner, Phys.Rev.Lett. 22 (1969) 1071. * (15) G.P.Imponente and G.Montani, Phys.Rev.D 63 (2001) 103501; T.Damour, M.Henneaux and H.Nicolai, Class.Quant.Grav. 20 (2003) R145. * (16) G.Montani, M.V.Battisti, R.Benini and G.Imponente, Int.J.Mod.Phys.A 23 (2008) 2353, arXiv:0712.3008. * (17) M.P.Ryan and L.C.Shapley, Homogeneous Relativistic Cosmologies (PUP, Princeton, 1975). * (18) C.J.Isham, Canonical quantum gravity and the problem of time (1992), gr-qc/9201011; K.V.Kuchar, in Quantum Gravity II, a second Oxford symposium, (1981), eds C. J. Isham et al., Clarendom Press., Oxford. * (19) S.Benczik et al., Phys.Rev.D 66 (2002) 026003, hep-th/0204049. * (20) V.A.Belinski and I.M.Khalatnikov, Sov.Phys.JETP 36 (1973) 591; B.Berger, Phys.Rev.D 61 (2000) 023508. * (21) C.Misner, Phys.Rev. 186 (1969) 1319. * (22) M.Bojowald and G.Date, Phys.Rev.Lett. 92 (2004) 071302, gr-qc/0311003; M.Bojowald, G.Date and G.M.Hossain, Class.Quant.Grav. 21 (2004) 3541, gr-qc/0404039. * (23) M.Bojowald, Living Rev.Rel. 8 (2005) 11, gr-qc/0601085; A.Ashtekar, Nuov.Cim.B 122 (2007) 135, gr-qc/0702030. * (24) M.V.Battisti, Phys.Rev.D 79 (2009) 083506. * (25) J.D.Barrow and M.P.Dabrowski, Phys.Rev.D 57 (1998) 7204; T.Damour and M.Henneaux, Phys.Rev.Lett. 85 (2000) 920; Phys.Lett.B 488 (2000) 108.	arxiv-papers	2008-08-06T13:26:00	2024-09-04T02:48:57.186979	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Marco Valerio Battisti, Giovanni Montani", "submitter": "Marco Valerio Battisti", "url": "https://arxiv.org/abs/0808.0831" }
0808.1035	# Global One–Dimensionality conjecture within Quantum General Relativity Łukasz Andrzej Glinka111Previously at Bogoliubov Laboratory of Theoretical Physics of Joint Institute for Nuclear Research in Dubna, Russian Federation E-mail: laglinka@gmail.com _International Institute for Applicable_ _Mathematics & Information Sciences,_ _Hyderabad (India) & Udine (Italy),_ _B.M. Birla Science Centre,_ _Adarsh Nagar, 500 063 Hyderabad, India_ ###### Abstract The simple quantum gravity model, based on a new conjecture within the canonically quantized $3+1$ general relativity, is presented. The conjecture states that matter fields are functionals of an embedding volume form only, and reduces the quantum geometrodynamics. By dimensional reduction the resulting theory is presented in the form of the Dirac equation, and application of the Fock quantization with the diagonalization procedure yields construction of the appropriate quantum field theory. The 1D wave function is derived, the corresponding 3-dimensional manifolds are discussed, and physical scales are associated with quantum correlations. Keywords general relativity; 3+1 splitting ; quantum gravity models ; low dimensional quantum field theories ; quantization methods ; global one- dimensionality PACS 04.60.-m ; 05.30.Jp ; 05.70.Ce; 11.10.Kk ; 98.80.Qc ## 1 Introduction Quantum gravity is one of the fundamental problems of modern theoretical physics. In spite of the significant efforts and various approaches, we are still very far of understanding the role of quantized gravitational fields in physical phenomena at high energies (for different approaches to quantum gravity see _e.g._ Ref. [1]). In this paper we propose a simple model of quantum gravity which can be useful for clarifying its some important aspects. The celebrated field-theoretic formalism yields plausible phenomenology for numerous experimental data of all areas of physics. In this paper this point of view is used for construction of a simple quantum gravity model. The $3+1$ splitting of a general relativistic metric tensor and the canonical quantization of the appropriate action functional are used in the well- grounded way. In straightforward and strict analogy with the generic cosmological model [2], the new conjecture within the Wheeler–DeWitt quantum geometrodynamics is proposed. The model is based on the ansatz composed by the steps 1. 1. global one–dimensionality conjecture, _i.e._ one-dimensional matter fields, 2. 2. reduced quantum geometrodynamics, yielding one-dimensional theory, 3. 3. dimensional reduction, resulting in the Dirac equation formulation, and expressing the supposition that the quantum geometrodynamics in itself is a one-dimensional field theory. The dimensional reduction leads to the appropriate Dirac equation and the Euclidean Clifford algebra. Its the Fock quantization with the diagonalization procedure, consisting of the Bogoliubov transformation and the Heisenberg equations of motion, yields correctly defined quantum field theory. The resulting model describes quantum gravity in terms of a quantum field theory formulated in the Fock static operator reper associated with initial data. The 1D wave function is derived, the corresponding 3-dimensional manifolds are discussed, and quantum correlations are associated with physical scales. Mathematically, we employ the one- dimensional functional integrals, so that the proposing quantum gravity model is methodologically corresponding to the trend initiated by Hartle and Hawking in the paper [3]. An organization of the paper is as follows. In the preliminary section 2 historically first quantized 3+1 general relativity is presented. Section 3 is devoted to the ansatz presentation. Next, the sections 4 and 5 discuss field quantization and some implications of general formulation, respectively. Finally, in the section 6 the entire paper’s results are summarized in condensed way. ## 2 Quantum 3+1 General Relativity In general relativity (See _e.g._ [4]) a pseudo-Riemannian manifold $(M,g)$ with a metric tensor $g_{\mu\nu}$, the Christoffel symbols $\Gamma^{\rho}_{\mu\nu}$, the Riemann curvature $R^{\lambda}_{\mu\alpha\nu}$, the Einstein curvature $G_{\mu\nu}=R_{\mu\nu}-\frac{1}{2}g_{\mu\nu}{{}^{(4)}}\\!R$, where $R_{\mu\nu}=R^{\lambda}_{\mu\lambda\nu}$, ${{}^{(4)}}\\!R=R^{\kappa}_{\kappa}$, satisfying the Einstein field equations222We use the units $c=\hbar=8\pi G/3=1$ in this text. $G_{\mu\nu}+\Lambda g_{\mu\nu}=3T_{\mu\nu},$ (1) where $\Lambda$ is cosmological constant and $T_{\mu\nu}$ is stress-energy tensor, models a spacetime333In (1) the coefficient of $T_{\mu\nu}$ usually equals $8\pi G/c^{4}$ that is exactly 3 in the our units.. For a compact $M$ with a boundary $(\partial M,h)$ and curvature $K_{ij}$, the usual variational principle is corrected [5] and (1) arise by the action $S[g]=\int_{M}d^{4}x\sqrt{-g}\left\\{-\dfrac{1}{6}R+\dfrac{\Lambda}{3}\right\\}+S_{\psi}[g]-\dfrac{1}{3}\int_{\partial M}d^{3}x\sqrt{h}K,$ (2) where $K=h^{ij}K_{ij}$, $S_{\psi}[g]$ is Matter fields action, $T_{\mu\nu}=-\frac{2}{\sqrt{-g}}\frac{\delta S_{\psi}[g]}{\delta g^{\mu\nu}}$. For $\Lambda=0$, a global timelike Killing field on $M$ $\mathcal{K}$ exists, the foliation $t=const$ is spacelike, $\partial M$ is the Nash embedding [6], and $3+1$ splitting [7] holds $\displaystyle g_{\mu\nu}=\left[\begin{array}[]{cc}-N^{2}+N_{i}N^{i}&N_{j}\\\ N_{i}&h_{ij}\end{array}\right]\quad,\quad h_{ik}h^{kj}=\delta_{i}^{j}\quad,\quad N^{j}=h^{ij}N_{i}$ (5) For $\Lambda>0$, $\mathcal{K}$ does not exist, spacelike $\partial M$ only foliates an exterior to the horizons on geodesic lines, (5) is a gauge. In both cases (2) takes the form $\displaystyle S[g]=\int dt\int_{\partial M}d^{3}x\left\\{\pi\dot{N}+\pi^{i}\dot{N_{i}}+\pi_{\psi}\dot{\psi}+\pi^{ij}\dot{h}_{ij}-NH- N_{i}H^{i}\right\\},$ (6) where dot means $t$-differentiation, $H$ and $H^{i}$ are defined as $\displaystyle H=\sqrt{h}\left\\{K^{2}-K_{ij}K^{ij}+{{}^{(3)}R}-2\Lambda-6\varrho\right\\},\qquad H^{i}=-2\pi^{ij}_{\leavevmode\nobreak\ ;j}\leavevmode\nobreak\ ,$ (7) with ${{}^{(3)}R}=h^{ij}R_{ij}$, $\varrho=n^{\mu}n^{\nu}T_{\mu\nu}$, $n^{\mu}=[1/N,-N^{i}/N]$, and particularly $\pi^{ij}=-\sqrt{h}\left(K^{ij}-h^{ij}K\right).$ (8) The curvature $K_{ij}$ satisfies the Gauss–Codazzi equations [8] $2NK_{ij}=N_{i\|j}+N_{j\|i}-\dot{h}_{ij}.$ (9) where stroke means intrinsic covariant differentiation. Time-preservation [9] of the primary constraints [10] leads to the secondary ones – (scalar) Hamiltonian constraint yielding dynamics, and (vector) diffeomorphism one merely reflecting spatial diffeoinvariance $\pi\approx 0\quad,\quad\pi^{i}\approx 0\quad\longrightarrow\quad H\approx 0\quad,\quad H^{i}\approx 0.$ (10) DeWitt [10] showed that $H^{i}$ generate the diffeomorphisms $\widetilde{x}^{i}=x^{i}+\xi^{i}$ $\displaystyle i\left[h_{ij},\int_{\partial M}H_{a}\xi^{a}d^{3}x\right]$ $\displaystyle=$ $\displaystyle- h_{ij,k}\xi^{k}-h_{kj}\xi^{k}_{\leavevmode\nobreak\ ,i}-h_{ik}\xi^{k}_{\leavevmode\nobreak\ ,j}\leavevmode\nobreak\ \leavevmode\nobreak\ ,$ (11) $\displaystyle i\left[\pi_{ij},\int_{\partial M}H_{a}\xi^{a}d^{3}x\right]$ $\displaystyle=$ $\displaystyle-\left(\pi_{ij}\xi^{k}\right)_{,k}+\pi_{kj}\xi^{i}_{\leavevmode\nobreak\ ,k}+\pi_{ik}\xi^{j}_{\leavevmode\nobreak\ ,k}\leavevmode\nobreak\ \leavevmode\nobreak\ ,$ (12) and consequently the first-class constraints algebra can be derived $\displaystyle i\left[H_{i}(x),H_{j}(y)\right]$ $\displaystyle=$ $\displaystyle\int_{\partial M}H_{a}c^{a}_{ij}d^{3}z,$ (13) $\displaystyle i\left[H(x),H_{i}(y)\right]$ $\displaystyle=$ $\displaystyle H\delta^{(3)}_{,i}(x,y),$ (14) $\displaystyle i\left[\int_{\partial M}H\xi_{1}d^{3}x,\int_{\partial M}H\xi_{2}d^{3}x\right]$ $\displaystyle=$ $\displaystyle\int_{\partial M}H^{a}\left(\xi_{1,a}\xi_{2}-\xi_{1}\xi_{2,a}\right)d^{3}x.$ (15) Here $H_{i}=h_{ij}H^{j}$, and $c^{a}_{ij}=\delta^{a}_{i}\delta^{b}_{j}\delta^{(3)}_{,b}(x,z)\delta^{(3)}(y,z)-(x\rightarrow y)$ are structure constants of diffeomorphism group. The scalar constraint reduced by (8) with using of the canonical primary quantization [9, 11] $\displaystyle i\left[\pi^{ij}(x),h_{kl}(y)\right]=\frac{1}{2}\left(\delta_{k}^{i}\delta_{l}^{j}+\delta_{l}^{i}\delta_{k}^{j}\right)\delta^{(3)}(x,y),$ (16) $\displaystyle i\left[\pi^{i}(x),N_{j}(y)\right]=\delta^{i}_{j}\delta^{(3)}(x,y)\quad,\quad i\left[\pi(x),N(y)\right]=\delta^{(3)}(x,y),$ (17) yields the Wheeler–DeWitt equation [12, 10] $\left\\{G_{ijkl}\dfrac{\delta^{2}}{\delta h_{ij}\delta h_{kl}}+h^{1/2}\left({{}^{(3)}R}-2\Lambda-6\varrho\right)\right\\}\Psi[h_{ij},\phi]=0,$ (18) where $G_{ijkl}$ is the Wheeler metric on superspace $S(\partial M)$ [12, 13] $G_{ijkl}=\dfrac{1}{2\sqrt{h}}\left(h_{ik}h_{jl}+h_{il}h_{jk}-h_{ij}h_{kl}\right),$ (19) and first class constraints are conditions on $\Psi[h_{ij},\phi]$ $\pi\Psi[h_{ij},\phi]=0\quad,\quad\pi^{i}\Psi[h_{ij},\phi]=0\quad,\quad H^{i}\Psi[h_{ij},\phi]=0.$ (20) In fact quantum general relativity, given by the Wheeler–DeWitt equation, historically is one of the first attempts of quantum gravity theory construction. Actually, however, quantum geometrodynamics has became the motivation for development of the quantum gravity idea and building novel formulations. We are going to build the our model basing on the Wheeler–DeWitt theory (18). Strictly speaking, however, the proposing toy model will possess a reductionist character. ## 3 The Ansatz For construction of a simple quantum gravity model let us apply the following three step ansatz. Global one–dimensionality conjecture. Suppose that Matter fields are one- variable functionals $\phi=\phi[h]$ where $h$ is a volume form of $\partial M$ $h\equiv\det h_{ij}=\dfrac{1}{3}\epsilon^{ijk}\epsilon^{lmn}h_{il}h_{jm}h_{kn},$ (21) and $\epsilon$ is the Levi-Civita symbol. Also we assume, as an element of the model, that gravitational field is described only variable h. In result a wave function $\Psi[h_{ij},\phi]$ becomes $\Psi[h_{ij},\phi]\rightarrow\Psi[h],$ (22) so that, the proposed quantum gravity model is $\left\\{-G_{ijkl}\dfrac{\delta^{2}}{\delta h_{ij}\delta h_{kl}}-h^{1/2}\left({{}^{(3)}R}-2\Lambda-6\varrho[h]\right)\right\\}\Psi[h]=0.$ (23) In analogy to the generic cosmology [2], (22) describes isotropic spacetimes444Assumption (22) means that we consider a strata of full superspace, _i.e._ the DeWitt minisuperspace model where the wave function depends only one variable $h$.. Reduced quantum geometrodynamics. Using $3+1$ splitting (5) within the Jacobi formula [4] $\delta g=gg^{\mu\nu}\delta g_{\mu\nu},$ (24) establishes the Jacobian matrix for for the reduction of variables $h_{ij}$ to $h$ $\displaystyle N^{2}\delta h=N^{2}hh^{ij}\delta h_{ij}\longrightarrow\mathcal{J}\left(h_{ij},h\right)=\dfrac{\delta(h)}{\delta(h_{ij})}=\dfrac{\delta h}{\delta h_{ij}}\equiv hh^{ij}.$ (25) Because of the approximation (22) the variational derivative $\delta/\delta h_{ij}$ acts on functional depending only on $h$. It allows us to express the derivative with respect $h_{ij}$ through the derivative with respect $h$. Therefore $\dfrac{\delta\Psi[h]}{\delta h_{ij}}=hh^{ij}\dfrac{\delta\Psi[h]}{\delta h}.$ (26) Consequently, application of (26) within the differential operator in (23) gives $\displaystyle G_{ijkl}\dfrac{\delta^{2}}{\delta h_{ij}\delta h_{kl}}=G_{ijkl}h^{ij}h^{kl}h^{2}\dfrac{\delta^{2}}{\delta h^{2}}.$ (27) So that the reduction is given by the double contraction $\displaystyle G_{ijkl}h^{ij}h^{kl}=\dfrac{1}{2\sqrt{h}}\left(h_{ik}h_{jl}+h_{il}h_{jk}-h_{ij}h_{kl}\right)h^{ij}h^{kl}=-\dfrac{3}{2}h^{-1/2},$ (28) where we have used the relations for 3-dimensional embedding $h^{ab}h_{bc}=h^{a}_{c}$, $h^{a}_{a}=\mathrm{Tr}h_{ab}=3$. Jointing (27) and (28) one obtains finally the relation555Because the relation (25) arises due to $3+1$ approximation, so (28) is an approximation within the ansatz. $\displaystyle G_{ijkl}\dfrac{\delta^{2}}{\delta h_{ij}\delta h_{kl}}=-\dfrac{3}{2}h^{3/2}\dfrac{\delta^{2}}{\delta h^{2}},$ (29) which leads to the reduced theory $\left\\{\dfrac{3}{2}h^{3/2}\dfrac{\delta^{2}}{\delta h^{2}}+h^{1/2}\left({{}^{(3)}R}-2\Lambda-6\varrho[h]\right)\right\\}\Psi[h]=0.$ (30) Dimensional reduction. The model (30) can be rewritten as $\left(\dfrac{\delta^{2}}{\delta{h^{2}}}-m^{2}\right)\Psi=0,$ (31) where $m^{2}$ is a squared (variable) mass of $\Psi$ $m^{2}=\dfrac{2}{3h}\left({}^{(3)}R-2\Lambda-6\varrho\right)=\dfrac{2}{3h}(K_{ij}K^{ij}-K^{2}),$ (32) and scalar constraint was used. Eq. (31) arises by stationarity of the action666Here $S[\Psi]$ is a field-theoretic action functional in $\Psi$ so that any dependence on $h$ of the mass $m$ does not play a role for equations of motion $\delta S/\delta\Psi=0$. $S[\Psi]=\int\delta hL\left(\Psi,\dfrac{\delta\Psi}{\delta h}\right)\quad,\quad L=\dfrac{1}{2}\left(\Pi_{\Psi}^{2}+m^{2}\Psi^{2}\right)\quad,\quad\Pi_{\Psi}=\dfrac{\delta\Psi}{\delta h}.$ (33) By using $\Pi_{\Psi}$ one rewrites the equation (31) as $\dfrac{\delta\Pi_{\Psi}}{\delta h}-m^{2}\Psi=0,$ (34) which together with $\Pi_{\Psi}$ in (33) yields the appropriate Dirac equation $\left(i\gamma\dfrac{\delta}{\delta h}-M\right)\Phi=0\quad,\quad\Phi=\left[\begin{array}[]{c}\Pi_{\Psi}\\\ \Psi\end{array}\right]\quad,\quad M=\left[\begin{array}[]{cc}-1&0\\\ 0&m^{2}\end{array}\right].$ (35) The $\gamma$ matrices algebra consists only one element - the Pauli matrix $\sigma_{y}$ $\gamma=\left[\begin{array}[]{cc}0&-i\\\ i&0\end{array}\right]\quad,\quad\gamma^{2}=I\quad,\quad\left\\{\gamma,\gamma\right\\}=2\delta_{E}\quad,\quad\delta_{E}=\left[\begin{array}[]{cc}1&0\\\ 0&1\end{array}\right].$ (36) The matrix algebra (36) forms the Euclidean Clifford algebra [14] $\mathcal{C}\ell_{1,1}(\mathbb{R})$ which has a $2D$ complex representation. Restricting to $Pin_{1,1}(\mathbb{R})$ yield a complex representation of 2D Pin group (2D spin representation); restricting to $Spin_{1,1}(\mathbb{R})$ splits it onto a sum of two half spin $1D$ representations (1D Weyl representation). The algebra decomposes into a direct sum of central simple algebras isomorphic to matrix algebra over $\mathbb{R}$ $\displaystyle\mathcal{C}\ell_{1,1}(\mathbb{R})=\mathcal{C}\ell^{+}_{1,1}(\mathbb{R})\oplus\mathcal{C}\ell^{-}_{1,1}(\mathbb{R})\quad,\quad\mathcal{C}\ell_{1,1}(\mathbb{R})\cong\mathbb{R}(2)\quad,\quad\mathcal{C}\ell^{\pm}_{1,1}(\mathbb{R})\cong\mathbb{R},$ (37) and moreover has a decomposition into a tensor product $\displaystyle\mathcal{C}\ell_{1,1}(\mathbb{R})=\mathcal{C}\ell_{2,0}(\mathbb{R})\otimes\mathcal{C}\ell_{0,0}(\mathbb{R})\quad,\quad\mathcal{C}\ell_{0,0}(\mathbb{R})\cong\mathbb{R}.$ (38) ## 4 Quantization The Dirac equation (35) can be canonically quantized (See _e.g._ [15]) $\displaystyle i\left[\mathbf{\Pi}_{\Psi}[h^{\prime}],\mathbf{\Psi}[h]\right]=\delta(h^{\prime}-h)\leavevmode\nobreak\ ,\leavevmode\nobreak\ i\left[\mathbf{\Pi}_{\Psi}[h^{\prime}],\mathbf{\Pi}_{\Psi}[h]\right]=0\leavevmode\nobreak\ ,\leavevmode\nobreak\ i\left[\mathbf{\Psi}[h^{\prime}],\mathbf{\Psi}[h]\right]=0.$ (39) Using of the Fock space allows to derive the solution in the form $\mathbf{\Phi}=\mathbb{Q}\mathfrak{B}\quad,\quad\mathbb{Q}=\dfrac{1}{\sqrt{2}}\left[\begin{array}[]{cc}\sqrt{1/\|m\|}&\sqrt{1/\|m\|}\\\ -i\sqrt{\|m\|}&i\sqrt{\|m\|}\end{array}\right],$ (40) where $\mathfrak{B}=\mathfrak{B}[h]$ is a dynamical reper $\mathfrak{B}=\left\\{\left[\begin{array}[]{c}\mathsf{G}[h]\\\ \mathsf{G}^{\dagger}[h]\end{array}\right]:\left[\mathsf{G}[h^{\prime}],\mathsf{G}^{\dagger}[h]\right]=\delta\left(h^{\prime}-h\right),\left[\mathsf{G}[h^{\prime}],\mathsf{G}[h]\right]=0\right\\},$ (41) and yields non Heisenberg-like dynamics of (41) $\dfrac{\delta\mathfrak{B}}{\delta h}=\mathbb{X}\mathfrak{B}\quad,\quad\mathbb{X}=\left[\begin{array}[]{cc}-im&\dfrac{1}{2m}\dfrac{\delta m}{\delta h}\\\ \dfrac{1}{2m}\dfrac{\delta m}{\delta h}&im\end{array}\right].$ (42) Supposing that there is an other reper $\mathfrak{F}$ determined by the Bogoliubov transformation and the Heisenberg equations of motion $\displaystyle\mathfrak{F}$ $\displaystyle=$ $\displaystyle\left[\begin{array}[]{cc}u&v\\\ v^{\ast}&u^{\ast}\end{array}\right]\mathfrak{B},\qquad\|u\|^{2}-\|v\|^{2}=1,$ (45) $\displaystyle\dfrac{\delta\mathfrak{F}}{\delta h}$ $\displaystyle=$ $\displaystyle\left[\begin{array}[]{cc}-i\Omega&0\\\ 0&i\Omega\end{array}\right]\mathfrak{F},$ (48) where $u$, $v$, $\Omega$ are functionals of $h$, one obtains $\dfrac{\delta\mathbf{b}}{\delta h}=\mathbb{X}\mathbf{b}\quad,\quad\mathbf{b}=\left[\begin{array}[]{c}u\\\ v\end{array}\right],$ (49) and $\Omega\equiv 0$, so that $\mathfrak{F}$ is the Fock static reper with respect to initial data ($I$) $\mathfrak{F}=\left\\{\left[\begin{array}[]{c}\mathsf{G}_{I}\\\ \mathsf{G}^{\dagger}_{I}\end{array}\right]:\left[\mathsf{G}_{I},\mathsf{G}^{\dagger}_{I}\right]=1,\left[\mathsf{G}_{I},\mathsf{G}_{I}\right]=0\right\\},$ (50) and vacuum state $\left\|\mathrm{VAC}\right\rangle$ is correctly defined $\mathsf{G}_{I}\left\|\mathrm{VAC}\right\rangle=0\quad,\quad 0=\left\langle\mathrm{VAC}\right\|\mathsf{G}_{I}^{\dagger}.$ (51) Integrability of Eqs. (49) is crucial. The transformation (45) suggests employing the superfluid parametrization which yield777In (52) the functional measure $\delta h$ for the case of a fixed space configuration transits into the Riemann–Lebesgue measure $dh$. However, $h$ in general is a smooth function of space parameters, $\delta h$ is a total variation and has a sense of the Stieltjes measure. $u=\dfrac{1+\lambda}{2\sqrt{\lambda}}\exp\left\\{im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\}\quad,\quad v=\dfrac{1-\lambda}{2\sqrt{\lambda}}\exp\left\\{-im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\},$ (52) where $\lambda\equiv\lambda[h]$, $\lambda^{\prime}=\lambda[h^{\prime}]$ is a length scale _i.e._ inverted mass scale $\mu=m/m_{I}=1/\lambda$. Consequently the integrability problem is solved by $\mathbf{\Phi}=\mathbb{Q}\mathbb{G}\mathfrak{F},$ (53) where $\mathbb{G}$ is the monodromy matrix $\mathbb{G}=\left[\begin{array}[]{cc}\dfrac{\lambda+1}{2\sqrt{\lambda}}\exp\left\\{-im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\}\vspace{10pt}&\dfrac{\lambda-1}{2\sqrt{\lambda}}\exp\left\\{im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\}\\\ \dfrac{\lambda-1}{2\sqrt{\lambda}}\exp\left\\{-im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\}&\dfrac{\lambda+1}{2\sqrt{\lambda}}\exp\left\\{im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\}\end{array}\right].$ (54) One sees now that the presented model expresses quantum gravity as a quantum field theory, where the quantum field associated with a space configuration is given by the relation (53). In this manner one can write out some straightforward conclusions following form the simple model. ## 5 Some implications of general formulation The proposed field-theoretic model was solved. However, still we do not know what it the role of the 1D wave function given by the equation (23). The same problem is to define any geometric quantities related to this model. The quantum field theory (53) has also unclear significance. Let us present now some conclusions arising from the previous section’s model, which will clarify these questions in some detail. Global 1D wave function. The Dirac equation (35) can be rewritten in the form of Schrödinger-like evolution equation $\dfrac{\delta\Phi}{\delta h}=H\Phi\quad,\quad H=-\left[\begin{array}[]{cc}0&\dfrac{m_{I}^{2}}{\lambda^{2}}\\\ 1&0\end{array}\right],$ (55) yielding unitary evolution operator $U=U(h,h_{I})=\exp\int_{h_{I}}^{h}H\delta h$ given by $U=\left[\begin{array}[]{cc}\cosh f[h,h_{I}]&\left(-m_{I}^{2}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime 2}}\right)\dfrac{\sinh f[h,h_{I}]}{f[h,h_{I}]}\\\ (h_{I}-h)\dfrac{\sinh f[h,h_{I}]}{f[h,h_{I}]}&\cosh f[h,h_{I}]\end{array}\right],$ (56) where $f[h,h_{I}]=\|m_{I}\|\sqrt{{(h-h_{I})\int_{h_{I}}^{h}\frac{\delta h^{\prime}}{\lambda^{\prime 2}}}}$, so that Eq. (55) is solved by $\Phi[h]=U(h,h_{I})\Phi[h_{I}].$ (57) Straightforward elementary algebraic manipulations allow to determine the global one-dimensional wave function as $\Psi=\Psi^{I}\cosh f[h,h_{I}]-\Pi_{\Psi}^{I}(h-h_{I})\mathrm{sgn}(h-h_{I})\dfrac{\sinh f[h,h_{I}]}{f[h,h_{I}]},$ (58) and similarly the canonical conjugate momentum is $\Pi_{\Psi}=\Pi_{\Psi}^{I}\cosh f[h,h_{I}]-\Psi^{I}m_{I}^{2}\mathrm{sgn}(h-h_{I})\left(\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime 2}}\right)\dfrac{\sinh f[h,h_{I}]}{f[h,h_{I}]},$ (59) where $\Psi^{I}=\Psi[h_{I}]$ and $\Pi_{\Psi}^{I}=\Pi_{\Psi}[h_{I}]=\left.\dfrac{\delta\Psi}{\delta h}\right\|_{h=hI}$ are initial data. In this manner the probability density in the classical reduced model is $\displaystyle\Omega[h]$ $\displaystyle=$ $\displaystyle(\Psi^{I})^{2}\cosh^{2}f[h,h_{I}]+(\Pi_{\Psi}^{I})^{2}(h-h_{I})^{2}\left[\dfrac{\sinh{f[h,h_{I}]}}{f[h,h_{I}]}\right]^{2}-$ (60) $\displaystyle-$ $\displaystyle 2\Psi^{I}\Pi_{\Psi}^{I}(h-h_{I})\mathrm{sgn}(h-h_{I})\dfrac{\sinh{2f[h,h_{I}]}}{2f[h,h_{I}]},$ and $\Psi^{I}$ and $\Pi_{\Psi}^{I}$ are determined by the normalization condition $\int_{h_{I}}^{\infty}\Omega[h^{\prime}]\delta h^{\prime}=1\quad\longrightarrow\quad C(\Pi_{\Psi}^{I})^{2}-2B\Psi^{I}\Pi_{\Psi}^{I}+A(\Psi^{I})^{2}-1=0,$ (61) where the constants $A$, $B$, $C$ are given by the integrals $\displaystyle A$ $\displaystyle=$ $\displaystyle\int_{h_{I}}^{\infty}\cosh^{2}f[h^{\prime},h_{I}]\delta h^{\prime},$ (62) $\displaystyle B$ $\displaystyle=$ $\displaystyle\int_{h_{I}}^{\infty}(h^{\prime}-h_{I})\mathrm{sgn}(h^{\prime}-h_{I})\dfrac{\sinh{2f[h^{\prime},h_{I}]}}{2f[h^{\prime},h_{I}]}\delta h^{\prime},$ (63) $\displaystyle C$ $\displaystyle=$ $\displaystyle\int_{h_{I}}^{\infty}(h^{\prime}-h_{I})^{2}\left[\dfrac{\sinh{f[h^{\prime},h_{I}]}}{f[h^{\prime},h_{I}]}\right]^{2}\delta h^{\prime},$ (64) The equation (61) can be solved straightforwardly. In result one obtains $\displaystyle\Pi_{\Psi}^{I}=\dfrac{B}{C}\Psi^{I}\pm\sqrt{{\left[\left(\dfrac{B}{C}\right)^{2}-\dfrac{A}{C}\right](\Psi^{I})^{2}+\dfrac{1}{C}}}.$ (65) Using $\Pi_{\Psi}^{I}=\dfrac{\delta\Psi^{I}}{\delta h_{I}}$ in (65) yields differential equation for $\Psi^{I}$, with the solution $\Psi^{I}=f_{\pm}^{(-1)}\left(\pm\dfrac{h_{I}}{C}+C_{1}\right),$ (66) where $C_{1}$ is integration constant, and $f_{\pm}(h_{I})$ are the functions $\displaystyle f_{\pm}(h_{I})=\dfrac{B}{AC}\Bigg{\\{}\mathrm{artanh}\dfrac{Bh_{I}}{\sqrt{{\left(B^{2}-AC\right)h_{I}^{2}+C}}}\pm\ln\sqrt{{Ah_{I}^{2}-1}}-$ $\displaystyle-\dfrac{\sqrt{{B^{2}-AC}}}{B}\ln\left[\left(B^{2}-AC\right)h_{I}+\sqrt{{B^{2}-AC}}\sqrt{{\left(B^{2}-AC\right)h_{I}^{2}+C}}\right]\Bigg{\\}}.$ (67) The 3-dimensional manifolds. The model (31) can be rewritten as $\left(\dfrac{\delta^{2}}{\delta{h^{2}}}-\dfrac{2}{3h}{{}^{(3)}R}\right)\Psi[h]=-\dfrac{4}{h}\left(\varrho[h]+\dfrac{\Lambda}{3}\right)\Psi[h].$ (68) and considered as the equation for the 3-dimensional scalar curvature ${{}^{(3)}\\!R}$ $^{(3)}\\!R=-6\left(\varrho[h]+\dfrac{\Lambda}{3}\right)+\varphi(\Psi)h\quad,\quad\varphi(\Psi)=\dfrac{3}{2}\dfrac{1}{\Psi}\dfrac{\delta^{2}\Psi}{\delta{h^{2}}}.$ (69) In the vacuum case, _i.e._ for the conditions $\left(\varrho[h]\equiv 0\cap\Lambda\equiv 0\right)$ or $\varrho[h]=-\dfrac{\Lambda}{3}$, one obtains from (69) that $^{(3)}\\!R=\varphi_{n}h,$ (70) where $\varphi_{n}$ in an eigenvalue determined by the equation $\dfrac{\delta^{2}\Psi}{\delta{h^{2}}}-\dfrac{2}{3}\varphi_{n}\Psi=0.$ (71) Supposing analytical form of $\Psi$ one establishes $\varphi_{n}$ $\Psi=\sum_{n=0}^{\infty}a_{n}(h-h_{I})^{n}\longrightarrow\varphi_{n}=\dfrac{3}{2}\left.\left(\dfrac{\delta^{n}}{\delta h^{n}}\left(\dfrac{m_{I}^{2}}{\lambda^{2}[h]}\Psi[h]\right)\left/\dfrac{\delta^{n}\Psi[h]}{\delta h^{n}}\right.\right)\right\|_{h=h_{I}}.$ (72) Let us assume that there are generalized functional Fourier transforms $\widetilde{\Psi}[s]=\int\delta he^{-2\pi ish}\Psi[h]\quad,\quad\widetilde{\dfrac{1}{\lambda^{2}}}[s]=\int\delta he^{-2\pi ish}\dfrac{1}{\lambda^{2}[h]},$ (73) as well as the generalized Leibniz product formula for functional derivatives $\dfrac{\delta^{n}}{\delta h^{n}}\left(\dfrac{1}{\lambda^{2}[h]}\Psi[h]\right)=\sum_{r=0}^{n}\binom{n}{r}\left(\dfrac{\delta^{r}}{\delta h^{r}}\dfrac{1}{\lambda^{2}[h]}\right)\left(\dfrac{\delta^{n-r}}{\delta h^{n-r}}\Psi[h]\right).$ (74) Using (73), (74) and $\sum_{r=0}^{n}\binom{n}{r}x^{r}=(1+x)^{n}$, within (72) yields $\varphi_{n}=\dfrac{3}{2}m_{I}^{2}\iint\delta s^{\prime}\delta se^{2i\pi(s^{\prime}+s)h_{I}}\left(1+\dfrac{s^{\prime}}{s}\right)^{n}\widetilde{\dfrac{1}{\lambda^{2}}}[s^{\prime}]\widetilde{\Psi}[s],$ (75) so that applying the inverted Fourier transforms $\Psi[h]=\int\delta se^{2\pi ish}\widetilde{\Psi}[s]\quad,\quad\dfrac{1}{\lambda^{2}[h]}=\int\delta se^{2\pi ish}\widetilde{\dfrac{1}{\lambda^{2}}}[s],$ (76) within the relation (75) one receives finally $\displaystyle\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!\varphi_{n}$ $\displaystyle=$ $\displaystyle\dfrac{3}{2}\iint\delta h\delta h\mathcal{G}(h-h_{I})\dfrac{m_{I}^{2}}{\lambda^{2}[h]}\Psi[h]=\dfrac{3}{2}\iint\delta h\delta h\mathcal{G}(h-h_{I})\dfrac{\delta^{2}\Psi[h]}{\delta h^{2}}=$ (77) $\displaystyle=$ $\displaystyle\dfrac{3}{2}\iint\delta h\delta h\mathcal{G}(h-h_{I})\dfrac{\delta\Pi_{\Psi}[h]}{\delta h}=-\dfrac{3}{2}\iint\delta h\delta h\dfrac{\delta}{\delta h}\mathcal{G}(h-h_{I})\Pi_{\Psi}[h],$ where we have used equations of motion, partial integration method. In (77) the kernel $\mathcal{G}(h-h_{I})$ and its derivative can be derived straightforwardly as $\displaystyle\mathcal{G}(h-h_{I})$ $\displaystyle=$ $\displaystyle\iint\delta s^{\prime}\delta se^{-2i\pi(s^{\prime}+s)(h-h_{I})}\left(1+\dfrac{s^{\prime}}{s}\right)^{n},$ (78) $\displaystyle\dfrac{\delta}{\delta h}\mathcal{G}(h-h_{I})$ $\displaystyle=$ $\displaystyle-\iint\delta s^{\prime}\delta s\dfrac{e^{-2i\pi(s^{\prime}+s)(h-h_{I})}}{2i\pi(s^{\prime}+s)}\left(1+\dfrac{s^{\prime}}{s}\right)^{n}.$ (79) Estimation of the functional integrals (78) or (79), and using of (58) or (59), leads to (77), which is a crucial for the relation (70). Quantum correlations. With using of the matrices (54) and (40), and the relation (53) one derives the quantum field $\displaystyle\mathbf{\Psi}[h]=\frac{\lambda[h]}{2\sqrt{2m_{I}}}\left(\exp\left\\{-im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda[h^{\prime}]}\right\\}\mathsf{G}_{I}+\exp\left\\{im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda[h^{\prime}]}\right\\}\mathsf{G}_{I}^{\dagger}\right).$ (80) Taking into account the $n$-field one-point quantum states determined as $\displaystyle\|h,n\rangle\equiv\mathbf{\Psi}^{n}\left\|\mathrm{VAC}\right\rangle=\left(\frac{\lambda}{2\sqrt{2m_{I}}}\exp\left\\{im_{I}\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\}\right)^{n}\mathsf{G}^{\dagger n}_{I}\left\|\mathrm{VAC}\right\rangle,$ (81) yields two-point correlators $\mathrm{Cor}_{n^{\prime}n}(h^{\prime},h)\equiv\langle n^{\prime},h^{\prime}\|h,n\rangle$ or explicitly $\displaystyle\mathrm{Cor}_{n^{\prime}n}(h^{\prime},h)$ $\displaystyle=$ $\displaystyle\dfrac{\lambda^{\prime n^{\prime}}\lambda^{n}}{\left(\sqrt{8m_{I}}\right)^{n^{\prime}+n}}\exp\left\\{im_{I}\left(n^{\prime}\int_{h^{\prime}}^{h_{I}}+n\int_{h_{I}}^{h}\right)\dfrac{\delta h^{\prime\prime}}{\lambda^{\prime\prime}}\right\\}\times$ (82) $\displaystyle\times$ $\displaystyle\left\langle\mathrm{VAC}\right\|\mathsf{G}_{I}^{n^{\prime}}\mathsf{G}^{\dagger n}_{I}\left\|\mathrm{VAC}\right\rangle,$ where $\lambda^{\prime}\equiv\lambda[h^{\prime}]$ and so on. Basically one obtains $\displaystyle\mathrm{Cor}_{00}(h,h)=\mathrm{Cor}_{00}(h^{\prime},h)=\mathrm{Cor}_{00}(h_{I},h_{I})=\left\langle\mathrm{VAC}\|\mathrm{VAC}\right\rangle,$ (83) $\displaystyle\mathrm{Cor}_{11}(h_{I},h_{I})=\dfrac{1}{8m_{I}}\quad,\quad\dfrac{\mathrm{Cor}_{n^{\prime}n}(h_{I},h_{I})}{\left[\mathrm{Cor}_{11}(h_{I},h_{I})\right]^{(n^{\prime}+n)/2}}=\left\langle\mathrm{VAC}\right\|\mathsf{G}_{I}^{n^{\prime}}\mathsf{G}^{\dagger n}_{I}\left\|\mathrm{VAC}\right\rangle,$ (84) so by elementary algebraic manipulations one receives $\displaystyle\mathrm{Cor}_{11}(h^{\prime},h)=\dfrac{\sqrt{{\mathrm{Cor}_{11}(h^{\prime},h^{\prime})\mathrm{Cor}_{11}(h,h)}}}{\mathrm{Cor}_{11}(h_{I},h_{I})}\exp\left\\{im_{I}\int_{h^{\prime}}^{h}\dfrac{\delta h^{\prime\prime}}{\lambda^{\prime\prime}}\right\\},$ (85) $\displaystyle\dfrac{\mathrm{Cor}_{nn}(h^{\prime},h)}{\mathrm{Cor}_{00}(h_{I},h_{I})}=\left[\dfrac{\mathrm{Cor}_{11}(h^{\prime},h)}{\mathrm{Cor}_{00}(h_{I},h_{I})}\right]^{n}\quad,\quad\dfrac{\mathrm{Cor}_{11}(h,h)}{\mathrm{Cor}_{00}(h_{I},h_{I})}=\lambda^{2}\mathrm{Cor}_{11}(h_{I},h_{I}).$ (86) Straightforwardly from (86) one relate a size scale with quantum correlations $\displaystyle\lambda=\sqrt{{\dfrac{\mathrm{Cor}_{11}(h,h)}{\mathrm{Cor}_{11}(h_{I},h_{I})\mathrm{Cor}_{00}(h_{I},h_{I})}}},$ (87) that consequently leads to the formulas $\dfrac{\mathrm{Cor}_{n^{\prime}n}(h,h)}{\mathrm{Cor}_{n^{\prime}n}(h_{I},h_{I})}=\lambda^{n^{\prime}+n}\exp\left\\{-im_{I}(n^{\prime}-n)\int_{h_{I}}^{h}\dfrac{\delta h^{\prime}}{\lambda^{\prime}}\right\\},$ (88) $\dfrac{\mathrm{Cor}_{11}(h^{\prime},h)}{\mathrm{Cor}_{00}(h_{I},h_{I})\mathrm{Cor}_{11}(h_{I},h_{I})}=\lambda^{\prime}\lambda\exp\left\\{im_{I}\int_{h^{\prime}}^{h}\dfrac{\delta h^{\prime\prime}}{\lambda^{\prime\prime}}\right\\},$ (89) $\dfrac{\mathrm{Cor}_{nn}(h^{\prime},h)}{\mathrm{Cor}_{00}(h_{I},h_{I})}=\lambda^{\prime n}\lambda^{n}[\mathrm{Cor}_{11}(h_{I},h_{I})]^{n}\exp\left\\{im_{I}n\int_{h^{\prime}}^{h}\dfrac{\delta h^{\prime\prime}}{\lambda^{\prime\prime}}\right\\}.$ (90) A whole information on the system is contained in $\lambda$, $\mu$, and $m_{I}$. Quantum correlations are determined by these quantities only. By measurement of quantum correlations one deduces $\lambda$, $\mu$, $m_{I}$. The presented conclusions have a formal character, however, they show a general feature of the proposed simple model of quantum gravity. We have solved the model by the 1D wave function (58). We have discussed the 3-dimensional manifolds (70) corresponding to this model, and we have found the relation between quantum correlations and physical scales (87). ## 6 Summary This paper discussed some consequences of the global one–dimensionality conjecture within the Wheeler–DeWitt theory. We have applied a field theory for formulation of the simple model of quantum gravity. The model was constructed by the following steps 1. 1. We have started from general relativity of compact manifold with boundaries; its action was written in $3+1$ splitting and the scalar constraint was canonically quantized. Resulting theory was the Wheeler–DeWitt model of quantum gravity; 2. 2. Next stage we have proposed the ansatz based on the global one-dimensionality conjecture; 3. 3. With using of the ansatz the quantum geometrodynamics was reduced to one- dimensional global evolution, with the dimension being an embedding volume form; 4. 4. Employing canonical formalism, we have used a field-theoretic action corresponding to the model, and by dimensional reduction the appropriate Dirac equation was received; 5. 5. Finally the Fock quantization was applied. Static reper of creators and annihilators was found by using of the diagonalization procedure employing the Bogoliubov transformation and the Heisenberg equations of motion. Consequently, the proposed model is defining quantum gravity as a quantum field theory. The quantum field was derived in a straightforward way (53). In result, we have obtained correctly defined integrability problem, which allowed to study its formal consequences. Particularly, we have discussed 1. 1. The integrability problem and global one-dimensional wave function. It was shown that by integration of the model in the Schrödinger equation form there is a possibility to obtain an exact solutions of the model. 2. 2. 3-dimensional manifolds corresponding to the model. It was shown that the model is defining a 3-dimensional manifolds ${}^{(3)}\\!R=\phi_{n}h$, where for given wave function the parameter $\phi_{n}$ can be derived by the assumption of the appropriate Fourier transforms and its inverted transforms. 3. 3. Relation between quantum correlations and physical scales. We have connected one-point quantum correlations with size and mass scales. The presented conclusions are partial, but they show possible physical and geometric implications following from the simple quantum gravity model. From a mathematical point of view we have applied a strategy of one-dimensional integration, with the Riemann–Lebesgue measure for fixed space configuration or the Stieltjes measure in general case. Both physical and mathematical sides of the model were emphasized in this paper. ## Acknowledgements Special thanks are directed to Prof. I. L. Buchbinder for very constructive discussion and his helpful comments to primary notes of the author. The author benefitted also many valuable discussions from Profs. A. B. Arbuzov, I. Ya. Aref’eva, B. M. Barbashov, K. A. Bronnikov, and V. N. Pervushin. ## References [1] I. L. Buchbinder, S. D. Odintsov, and I. L. Shapiro, Effective Action in Quantum Gravity. Institute of Physics Publishing (1992); D. J. Gross, T. Piran, and S. Weinberg (eds.), Two Dimensional Quantum Gravity and Random Surfaces. World Scientific (1992); G. W. Gibbons and S. W. Hawking (eds.), Euclidean Quantum Gravity. World Scientific (1993); G. Esposito, Quantum Gravity, Quantum Cosmology and Lorentzian Geometries. Springer (1994); J. Ehlers and H. Friedrich (eds.), Canonical Gravity: From Classical to Quantum. Springer (1994); E. Prugovečki, Principles of Quantum General Relativity. World Scientific (1995); R. Gambini and J. Pullin, Loops, Knots, Gauge Theories and Quantum Gravity. Cambridge University Press (1996); G. Esposito, A. Yu. Kamenshchik, and G. Pollifrone, Euclidean Quantum Gravity on Manifolds with Boundary. Springer (1997); P. Fré, V. Gorini, G. Magli, and U. Moschella, Classical and Quantum Black Holes. Institute of Physics Publishing (1999); I. G. Avramidi, Heat Kernel and Quantum Gravity. Springer (2000); B. N. Kursunoglu, S. L. Mintz, and A. Perlmutter (eds.), Quantum Gravity, Generalized Theory of Gravitation and Superstring Theory-Based Unification. Kluwer (2002); S. Carlip, Quantum Gravity in 2+1 Dimensions. Cambridge University Press (2003); D. Giulini, C. Kiefer and C. Lämmerzahl (eds.), Quantum Gravity. From Theory To Experimental Search. Springer (2003); C. Rovelli, Quantum Gravity. Cambridge University Press (2004); G. Amelino-Camelia and J. Kowalski-Glikman (eds.), Planck Scale Effects in Astrophysics and Cosmology. Springer (2005); A. Gomberoff and D. Marolf (eds.), Lectures on Quantum Gravity. Springer (2005); D. Rickles, S. French, and J. Saatsi (eds.), The Structural Foundations of Quantum Gravity. Clarendon Press (2006); D. Gross, M. Henneaux, and A. Sevrin (eds.), The Quantum Structure of Space and Time. World Scientific (2007); C. Kiefer, Quantum Gravity. 2nd ed., Oxford University Press (2007); T. Thiemann, Modern Canonical Quantum General Relativity. Cambridge University Press (2007); D. Oriti, Approaches to Quantum Gravity. Toward a New Understanding of Space, Time, and Matter. Cambridge University Press (2009). * [2] L. A. Glinka, AIP Conf. Proc. 1018, 94, (2008). arXiv:0801.4157 [gr-qc]; SIGMA 3, 087, (2007). arXiv:0707.3341 [gr-qc]; arXiv:gr-qc/0612079 * [3] J. B. Hartle and S. W. Hawking, Phys. Rev. D 28, 2960 (1983). * [4] C. W. Misner, K. S. Thorne, and J. A. Wheeler, Gravitation. Freeman (1973); R. M. Wald, General Relativity. University of Chicago (1984); S. Carroll, Spacetime and Geometry. An introduction to General Relativity. Addison–Wesley (2004). * [5] J. W. York, Phys. Rev. Lett. 28, 1082 (1972); G. W. Gibbons and S. W. Hawking, Phys. Rev. D 15, 2752 (1977). * [6] J. F. Nash, Ann. Math. 56, 405 (1952); _ibid._ 63, 20 (1956); S. Masahiro, Nash Manifolds. Springer (1987); M. Günther, Ann. Global Anal. Geom. 7, 69 (1989); Math. Nachr. 144, 165 (1989). * [7] R. Arnowitt, S. Deser and C. W. Misner, in Gravitation: An Introduction to Current Research, ed. by L. Witten, p. 227, Wiley (1962); B. DeWitt, The Global Approach to Quantum Field Theory, Vol. 1 & 2, Clarendon Press (2003). * [8] A. Hanson, T. Regge, and C. Teitelboim, Constrained Hamiltonian Systems. Accademia Nazionale dei Lincei (1976). * [9] P. A. M. Dirac, Lectures on Quantum Mechanics. Belfer Graduate School of Science, Yeshiva University (1964). * [10] B. S. DeWitt, Phys. Rev. 160, 1113 (1967). * [11] L. D. Faddeev, Usp. Fiz. Nauk 136, 435 (1982). * [12] J. A. Wheeler, Geometrodynamics. Academic Press (1962); Einsteins Vision. Springer (1968). * [13] A. E. Fischer, Gen. Rel. Grav. 15, 1191 (1983); J. Math. Phys. 27, 718 (1986). * [14] V. V. Fernández, A. M. Moya, and W. A. Rodrigues Jr, Adv. Appl. Clifford Alg. 11, 1 (2001). * [15] N. N. Bogoliubov and D. V. Shirkov, Introduction to the theory of quantized fields. 3rd ed., John Wiley and Sons (1980).	arxiv-papers	2008-08-07T18:48:08	2024-09-04T02:48:57.194505	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "L. A. Glinka", "submitter": "L. A. Glinka", "url": "https://arxiv.org/abs/0808.1035" }
0808.1210	# Improving Wang-Landau sampling with adaptive windows A. G. Cunha-Netto111e-mail address: agcnetto@fisica.ufmg.br1,2, A. A. Caparica1, Shan-Ho Tsai3, Ronald Dickman2 and D. P. Landau3 1Instituto de Física, Universidade Federal de Goiás, C.P. 131, 74001-970, Goiânia, Brazil 2Departamento de Física, Instituto de Ciências Exatas, Universidade Federal de Minas Gerais, C.P.702, 30123-970 Belo Horizonte, Minas Gerais, Brazil 3Center for Simulational Physics, University of Georgia, Athens, Georgia, 30602 USA ###### Abstract Wang-Landau sampling (WLS) of large systems requires dividing the energy range into “windows” and joining the results of simulations in each window. The resulting density of states (and associated thermodynamic functions) are shown to suffer from boundary effects in simulations of lattice polymers and the five-state Potts model. Here, we implement WLS using adaptive windows. Instead of defining fixed energy windows (or windows in the energy-magnetization plane for the Potts model), the boundary positions depend on the set of energy values on which the histogram is flat at a given stage of the simulation. Shifting the windows each time the modification factor $f$ is reduced, we eliminate border effects that arise in simulations using fixed windows. Adaptive windows extend significantly the range of system sizes that may be studied reliably using WLS. Monte Carlo simulation, Wang-Landau sampling, lattice polymer, Potts model ###### pacs: 05.10.Ln, 64.60.Cn, 64.60.De In recent years Wang-Landau sampling (WLS) landau ; landaupre ; wanglandau_other ; wang_landau_ajp has become an important algorithm for Monte Carlo (MC) simulations, and is being applied to a vast array of models in statistical physics and beyondwang_landau_ajp . WLS uses a random walk in energy ($E$) space to estimate the density of states, $g(E)$; to sample well the full range of energies, the walk is adjusted to spend approximately equal time intervals at each energy value. The estimate for $g(E)$ is refined successively, in a series of random walks. WLS has been used, for example, to simulate polymers 3d ; jain ; rampf_binder_paul ; parsons_williams and proteins rathore ; wuest_landau and to calculate the joint density of states (JDOS) of two or more variableszhouetal , e.g., $g(E,M)$ of magnetic systems ($M$ is the magnetization). For models with a complex energy landscape, WLS permits simulation of larger systems than can be studied using conventional MC approaches. Because sampling the full range of energies of a large system is not viable using a single walk, one divides the energy range into slightly overlapping subintervals (“windows”) and samples each separately. The density of states $g(E)$ for the full energy range is then obtained via a matching procedure that consists of multiplying $g(E)$ in each window by a factor to force continuity of this function at the borders. (This yields $g(E)$ to within an overall multiplicative factor, permitting evaluation of canonical averages. If the absolute density of states is needed, the normalization must use some reference value for which the density of states is known exactly, for example, the ground state degeneracy.) The question naturally arises whether sampling restricted to windows yields reliable estimates for $g(E)$, near the boundary energy values. Wang and Landau suggested that boundary effects would be negligible if adjacent windows were defined with a suitably large overlap landaupre . On the other hand, Schulz et al.schulz found it advantageous to introduce an additional rule, namely, whenever a configuration is rejected because its energy is greater than the maximum value, $E_{>}$, of a given window, one should update $g(E)$ for the current energy value. These prescriptions are effective for Ising models but do not eliminate boundary effects in all instances, e.g., in simulations of polymers preprint and systems where the JDOS is required shan . Here we present an implementation of WLS that eliminates border effects. For the sake of clarity, we consider two examples with severe border problems: a lattice polymer with attractive interactions between nonbonded nearest- neighbor monomers baumgartner ; 3d ; douglas_ishinabe , simulated via reptationlandau_binder , and the calculation of $g(E,M)$ in the five-state Potts model fywu on the square lattice. (We note that while the reptation method is not suitable for sampling the most compact configurations, this limitation does not affect the conclusions presented here.) Figure 1: Specific heat of a polymer of $N=100$ monomers on the square lattice, using two windows and border energies $E_{b}$ = -68 and -60. The inset shows the neighborhood of the maximum. Average values and uncertainties are calculated using ten independent runs. To begin, we document the border effects that arise in fixed-window simulations of polymers, focusing on the specific heat $c(T)$ (calculated as usual from the variance of the energy). The polymer chain is modeled as a self-avoiding walk on the square lattice. The energy is $E=-N_{nb}$, with $N_{nb}$ the number of nearest-neighbor contacts between nonconsecutive monomers. Our simulations sample the entire energy range, from the ground state $E_{min}$ to $E_{max}=0$. We encounter strong border effects, even if, following the suggestion of Ref. schulz , we use three overlapping levels at the border(s). Varying the border energy $E_{b}$, we find that the temperature $T_{m}$ at which $c(T)$ takes its maximum varies in an analogous manner (a smaller $E_{b}$ corresponds to a smaller $T_{m}$), signaling a systematic error (see Fig. 1). Evidently WLS using fixed windows yields distorted results for $g(E)$ in this system. (We stress that such a distortion does not arise in WLS of the 2d Ising model landaupre , for which simulations using fixed windows have been performed on lattices of up to $256\times 256$ sites.) We shall now describe a procedure that eliminates the boundary effects described above. Recall first that in WLS, the simulation begins with an arbitrary configuration, and with $g(E)=1$ for all values of energy $E$. A random walk in energy space is realized by generating trial configurations and accepting them with probability $p(E_{1}\rightarrow E_{2})=\min(g(E_{1})/g(E_{2}),1)$, where $E_{1}$ is the current energy value and $E_{2}$ the energy of the trial configuration ${\cal C}_{2}$. If ${\cal C}_{2}$ is accepted, we update $g(E_{2})\rightarrow f\times g(E_{2})$ and $H(E_{2})\rightarrow H(E_{2})+1$; otherwise, $g(E_{1})\rightarrow f\times g(E_{1})$ and $H(E_{1})\rightarrow H(E_{1})+1$, where the histogram $H(E)$ records the number of visits to energy $E$, and $f$ is the modification factor, initially set to $e=2.71828\ldots$. When $H(E)$ is sufficiently flat, it is reset to zero and a new random walk is initiated, with a smaller $f$, e.g., $f\rightarrow\sqrt{f}$, used to update $g(E)$. The simulation ends when $f$ is sufficiently close to $1$, e.g., $f\approx 1+10^{-6}$. The histogram is said to be flat if, for all energies in the window of interest, $H(E)>\kappa\overline{H}$, where the overline denotes an average over energies. (Typically, $\kappa=0.8$, as used here.) In WLS with fixed windows, this stopping criterion is applied in each window. In our method, by contrast, we determine the range over which the histogram has attained the desired degree of uniformity at various stages of the simulation. Initially, we allow the WLS random walk to visit all possible energies $E_{min}\leq E\leq E_{max}$, accumulating the histogram in the usual manner. After a certain number $N_{1}$ of Monte Carlo steps (in practice we use $N_{1}=10^{4}$), we check if the histogram, restricted to the interval $E_{w}\leq E\leq E_{max}$, is flat. (Here $E_{w}=E_{max}-W$, with $W$ defining the minimum acceptable window size.) If it is not flat, we perform an additional $N_{1}$ Monte Carlo steps, and check again, repeating until the histogram is flat on the minimal window. Once this condition is satisfied, we check whether the existing histogram is in fact flat on a larger window. (This is done by including the energy level just below $E_{w}$ in $\overline{H}$ and checking if the flatness criterion, $H(E)>\kappa\overline{H}$, holds in the enlarged window. In this manner, adding levels one by one, we identify the largest window over which flatness is satisfied.) Let $E^{}\leq E_{w}$ be the smallest energy such that the histogram, restricted to the interval $[E^{},E_{max}]$, is flat, and let $E_{1}=E^{}+\Delta E$, where $\Delta E$ determines the overlap between adjacent windows. In the next stage of the simulation, we restrict the random walk to energies $E_{min}\leq E\leq E_{1}+\Delta E$. After $N_{1}$ steps we check if the histogram is flat on the interval $[E_{1}+\Delta E-W,E_{1}+\Delta E]$, and proceed as above, until all possible energies have been included in a window with a flat histogram (see Fig. 2). To avoid problems that might arise with very small windows, the final window (with lower limit $E_{min}$) always has a width $\geq W$. In the studies shown below we use $W=(E_{max}-E_{min})/10$ and $\Delta E=3$ for lattice polymers ($\Delta E=1$ or $\Delta M=1$ for the Potts model). Once all allowed energies have been sampled, we connect the functions $g(E)$ associated with the various windows by imposing continuity at $E_{1}$, $E_{2}$,…,$E_{n}$. Then the entire procedure is repeated using the next value of the modification factor $f$, that is, $f\to\sqrt{f}$; the simulation ends once $f-1<10^{-6}$. Figure 2: Lower panel: schematic of adaptive windows; the value of $E_{max}$ depends on the model, while $\Delta E$ is chosen to ensure sufficient overlap. $E^{}$ and $E_{1}$, $E_{2}$, etc., are determined during the simulation not fixed beforehand. Upper panel: example of a histogram and associated window in the lattice polymer simulation. Note that at each iteration (using a different value of $f$), the borders $E_{j}$ are chosen differently: repetition of a border value in successive iterations is prohibited. This point is crucial to the functioning of our method; if the borders were fixed, errors incurred at a given iteration (similar to those seen in Fig. 1) would accumulate, rather than being corrected in subsequent iterations. (Here it is well to remember that WLS functions precisely because errors in $g(E)$ incurred at a given value of $f$ tend to be corrected at subsequent stages.) Figure 3: Density of states $g(E)$ of two-dimensional polymers, $N=200$. Solid line: adaptive-window sampling; dashed line: Metropolis sampling, using the ratios $g(E)/g(E_{ref})$. Circles denote the limits of the energy ranges associated with each temperature in Metropolis sampling. It is important to stress that energy windows are not merely a device for accelerating the simulation, but are in fact necessary in studies of large systems; without them, WLS simply does not converge. Thus, the largest lattice polymer we are able to study without windows is $N=70$; using the adaptive- window scheme, studies including the entire range of energies are possible on chains of at least 300 monomers. To test the validity of our method, we compare $g(E)$ (for $N=200$), given by our scheme with that obtained using high-resolution Metropolis Monte Carlolandau_binder ; excellent agreement is found (see Fig. 3). (Metropolis simulations are performed at temperatures $T=1.0,1.5,3.0$ and $10.0$, using $10^{8}$ MC steps per study, using the RAN2 random number generator numericalrecipes . In Metropolis MC, the expected number $m(E)$ of visits to energy $E$ satisfies $m(E)/m(E_{ref})=e^{-(E-E_{ref})/k_{B}T}g(E)/g(E_{ref})$. We use this relation to determine the ratios $g(E)/g(E_{ref})$, by equating $m(E)$ to the actual number of visits to energy $E$ during the simulation. At each temperature, a certain range of energies are well sampled; for each temperature studied, we take the reference energy $E_{ref}$ as the most visited value. The resulting values for $g(E)$ are normalized by equating $g(E_{ref})$ to the corresponding value obtained via WLS.) We turn now to the $Q$-state Potts modelfywu , with Hamiltonian ${\cal H}=-J\sum_{\langle ij\rangle}\delta_{\sigma_{i},\sigma_{j}}-h\sum_{i}\delta_{\sigma_{i},1}$ (1) where $\sigma_{i}=1,2,...,Q$; $\langle ij\rangle$ denotes pairs of nearest- neighbor spins, $J>0$ is a ferromagnetic coupling, and $h$ is an external field that couples to one of the states. (Our units are such that $J/k_{B}=1$.) We use WLS with a 2d random walk to determine the JDOS $g(E,M)$. [Here $E=-\sum_{\langle ij\rangle}\delta_{\sigma_{i},\sigma_{j}}$ and $M=\sum_{i}\delta_{\sigma_{i},1}$. Using $g(E,M)$, thermodynamic properties may be obtained for arbitrary $(T,h)$.] We study the five-state model on square lattices of 32 $\times$ 32 sites with periodic boundaries; we use the R1279 shift register random number generator landau_binder . We tried to parallelize the 2d WLS by performing independent random walks on overlapping $(E,M)$ regions. We divided the parameter space into strips with: (i) restricted values of $E$ and unrestricted $M$; (ii) restricted $M$ and unrestricted $E$; and (iii) rectangles with both $E$ and $M$ restricted. The independent random walks were carried out in parallel and the densities of states from adjacent regions were normalized at a single common $(E,M)$ point, or by minimizing the least-square distance between an overlapping $(E,M)$ region. We found, however, that the resulting density of states exhibits some small discontinuities which affect the probability distributions of energy and magnetization, as illustrated by the dashed line in Fig.4 for $P(M,T,h)$ for the case (ii) above. This curve was obtained with one simulation (hence no error bars are displayed); the temperature is taken such that $P(M,T,h)$ has two peaks of approximately equal height. When we average $P(M,T,h)$ from multiple independent simulations the jumps are somewhat smoothed out; nevertheless, the thermodynamic quantities obtained with these implementations of parallel WLS suffer from a small systematic error, as illustrated in Fig.5 for the specific heat. Figure 4: Magnetization probability distribution $P(M,T,h)$ for $T=0.86177$ and $h=0.005$ computed from $g(E,M)$ obtained via WLS with fixed windows using case (ii) above (dashed line), adaptive windows (solid line), and from a hybrid Monte Carlo (MC) method (dot-dashed line). In the inset only a few typical error bars are shown for the hybrid MC result (error bars for the WLS with adaptive windows are slightly smaller). Figure 5: Specific heat versus temperature for $h=0.005$, obtained with different sampling methods. Error bars are smaller than the symbol sizes. The discontinuities in the probability distributions, and the systematic error in thermodynamic quantities arise because the slope of the JDOS is discontinuous at the borders between neighboring fixed windows. These problems can be partly circumvented by using a larger overlap region, so that the JDOS used in the calculation of thermodynamic properties does not come from the immediate vicinity of any border. The disadvantage of such an approach is that the overlap between adjacent windows has to be quite large. There is, moreover, no guarantee that these regions will be sampled properly. Using adaptive windows, by contrast, the slopes of the JDOS at the borders of neighboring windows are equal to within numerical uncertainty. (In WLS with adaptive windows, we normalize the JDOS by applying a least-square-difference criterion along the line of overlap between neighboring sampling windows.) The magnetization probability distribution computed from $g(E,M)$ (obtained from an average over ten independent runs using WLS with adaptive windows), is also shown in Figure 4 (solid line). This result is in good agreement with the distribution obtained using a hybrid Monte Carlo method landau_binder with $10^{8}$ MC steps. Here one MC step comprises $4L^{2}$ single spin-change trials with the Metropolis algorithm and $6$ Wolff cluster updates. (In the main graph of Fig.4 the result from adaptive-window WLS is almost indistinguishable from that using hybrid MC.) The specific heat obtained with WLS using adaptive windows is also in excellent agreement with the result of the hybrid Monte Carlo method, as shown in Fig.5. Although we illustrate the effectiveness of WLS with adaptive windows for $L=32$, a relatively small lattice size, the reader should recall that we are performing two-dimensional random walks; therefore, the parameter space is much larger than for a one- dimensional random walk using the same system size. The resulting JDOS allows us to obtain thermodynamic quantities for the entire $(T,h)$ space. We note in passing that because the 5-state Potts model has a very weak first-order phase transition, the hybrid Monte Carlo method, employed here to test our new method, can be used to sample equilibrium states; however, as $Q$ increases, it becomes forbiddingly difficult for this method to equilibrate the system using currently available computational resources. In contrast, the WLS method works well even in the presence of strong first-order phase transitions and, when combined with the adaptive windows method presented here, can be used to simulate quite large systems. In summary, we show how WLS, arguably the most promising method presently available for simulating complex spin and fluid models, may be applied reliably to large systems without border effects. For the models considered here, such effects are strong and effectively prohibit the study of large systems using fixed windows. In our method, errors that may arise near the border of a given window are corrected in subsequent stages, in which the border positions are shifted. We expect this improvement of WLS to find broad application in studies of polymers, spin systems with complex phase diagrams, and complex fluids. We are grateful to CAPES, CNPq, FUNAPE-UFG, Fapemig (Brazil), and NSF grant number DMR-0341874 (USA) for financial support. ## References * (1) F. Wang and D. P. Landau, Phys. Rev. Lett. 86, 2050 (2001). * (2) F. Wang and D. P. Landau, Phys. Rev. E 64, 056101 (2001). * (3) D. P. Landau and F. Wang, Comput. Phys. Commun. 147, 674 (2002); F. Wang and D. P. Landau, Braz. J. Phys. 34, 354 (2004). * (4) For a partial list of applications of WLS, see e.g., D. P. Landau, S.-H. Tsai, and M. Exler, Amer. J. Phys. 72, 1294 (2004). * (5) A. G. Cunha-Netto, C. J. Silva, A. A. Caparica and R. Dickman, Braz. J. Phys. 36, 619 (2006). * (6) T.S. Jain and J.J. de Pablo, J. Chem. Phys. 116, 7238 (2002); J. Chem. Phys. 118, 4226 (2003). * (7) F. Rampf, K. Binder, and W. Paul, J. Polym. Sci. Part B: Polym. Phys. 44, 2542 (2006); W. Paul, T. Strauch, F. Rampf, and K. Binder, Phys. Rev. E 75, 060801(R), (2007). * (8) D. F. Parsons and D. R. M. Williams, Phys. Rev. E 74, 041804 (2006); J. Chem. Phys. 124, 221103 (2006). * (9) N. Rathore and J.J. de Pablo, J. Chem. Phys. 116, 7225 (2002); J. Chem. Phys. 118, 4285 (2003). * (10) T. Wüst and D. P. Landau, Comput. Phys. Commun. 179, 124 (2008). * (11) C. Zhou, T. C. Schulthess, S. Torbrügge, and D. P. Landau, Phys. Rev. Lett. 96, 120201 (2006); C. J. Silva, A. A. Caparica, and J. A. Plascak, Phys. Rev. E 73, 036702 (2006). * (12) B.J. Schulz, K. Binder, M. Müller, and D.P. Landau, Phys. Rev. E 67, 067102 (2003). * (13) A. G. Cunha-Netto, A. A. Caparica and R. Dickman, unpublished. * (14) S.-H. Tsai, F. Wang, and D. P. Landau, Braz. J. Phys. 38, 6 (2008). * (15) A. Baumgärtner, J. Phys. (Paris) 43, 1407 (1982). * (16) J. F. Douglas and T. Ishinabe, Phys. Rev. E 51, 1791 (1995). * (17) See, e.g., D. P. Landau and K. Binder, A Guide to Monte Carlo Simulation in Statistical Physics, 2nd Edition (Cambridge University Press, Cambridge, 2005). * (18) F. Y. Wu, Rev. Mod. Phys. 54, 235 (1982). * (19) W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes, The Art of Scientific Computing, 2nd Edition (Cambridge University Press, Cambridge, 1992).	arxiv-papers	2008-08-08T13:09:26	2024-09-04T02:48:57.201368	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "A. G. Cunha-Netto, A. A. Caparica, Shan-Ho Tsai, Ronald Dickman and D.\n P. Landau", "submitter": "Ronald Dickman", "url": "https://arxiv.org/abs/0808.1210" }
0808.1221	# Neutral Current induced $\pi^{0}$ production and neutrino magnetic moment M. Sajjad Athar, S. Chauhan and S. K. Singh Department of Physics, Aligarh Muslim University, Aligarh-202 002, India ###### Abstract We have studied the total cross section, $Q^{2}$, momentum and angular distributions for pions in the $\nu$($\bar{\nu}$) induced $\pi^{0}$ production from nucleons. The calculations have been done for the weak production induced by the neutral current in the standard model and the electromagnetic production induced by neutrino magnetic moment. It has been found that with the present experimental limits on the muon neutrino magnetic moment $\mu_{\nu_{\mu}}$, the electromagnetic contribution to the cross section for the $\pi^{0}$ production is small. The neutrino induced neutral current production of $\pi^{0}$, while giving an alternative method to study the magnetic moment of neutrino $\mu_{\nu_{\mu}}$, does not provide any improvement over the present experimental limit on $\mu_{\nu_{\mu}}$ from the observation of this process in future experiments at T2K and NO$\nu$A. ###### pacs: 13.15.+g, 14.60.Pq, 14.60.St ## I Introduction The neutral current $\pi^{0}$ production in neutrino interactions plays an important role in the background studies of $\nu_{\mu}\rightarrow\nu_{e}$ oscillations in the appearance mode as well as in discriminating between $\nu_{\mu}\rightarrow\nu_{\tau}$ and $\nu_{\mu}\rightarrow\nu_{s}$ modes Aguiler1 ; Kajita ; Mauger ; Raaf . This process can also help to distinguish between production of $\nu_{\tau}$ and $\bar{\nu}_{\tau}$ in some oscillation scenarios at neutrino energies much below the $\tau$ production threshold but above the pion threshold Nieves . The recent results on neutral current induced pion production in neutrino oscillation experiments at K2K aip ; nakayama and MiniBooNE Aguiler ; nguyen ; link have generated great interest in studying these processes. In this context, the proposed experiments by T2K Yury ; nakadaira and NO$\nu$A norman collaborations plan to study this process with better statistics. The neutral pions can also be produced by electromagnetic interactions if $\nu(\bar{\nu})$ have diagonal and/or transition magnetic moments. This process would in principle contribute additional events to the neutral current reaction and would modify the energy and angular distributions of the neutral pions, which may be observed in future experiments. It is thus possible, in principle, to get information about the magnetic moment of neutrinos(antineutrinos) from studying neutral current induced $\pi^{0}$ production from nucleons and nuclei. While the minimal extensions of Standard model predict very tiny diagonal magnetic moments Fujikawa , there are models of electroweak interactions which predict enhanced transition magnetic moment Mohapatra . The present limits for the magnetic moment of neutrinos come from neutrino - electron scattering for $\nu_{\mu}$ and from $e^{+}e^{-}\rightarrow\nu\bar{\nu}\gamma$ for $\nu_{\tau}$. These limits for $\nu_{\mu}$ and $\nu_{\tau}$ magnetic moments are $\mu_{\nu_{\mu}}<6.8\times 10^{-10}\mu_{B}$ and $\mu_{\nu_{\tau}}<3.9\times 10^{-7}\mu_{B}$ review . The data from neutrino oscillation experiments from Sudbury Neutrino Observatory and Super-Kamiokande have also been analysed to obtain improved limits on the neutrino magnetic moments for $\nu_{\mu}$ and $\nu_{\tau}$ grifols ; grimus ; joshipura ; beacon ; pulido . A recent analysis of the Borexino experiment claims to improve these limits on the magnetic moments of $\nu_{\mu}$ and $\nu_{\tau}$ by 3 orders of magnitudemontanino . We would like to consider the possibility of obtaining new bounds on the neutrino magnetic moment using high statistics data on neutral current induced $\pi^{0}$ production from nucleons and nuclei in future experiments. Such a possibility was earlier discussed by Kang et al. kang using first result on $\pi^{0}$ production from Superkamiokande experiment on atmospheric neutrinosFukuda . We study in this paper, the $\pi^{0}$ production induced by weak neutral current and magnetic moment interaction of neutrinos and antineutrinos in the energy region of few GeV, relevant for K2K, MiniBooNE, T2K and NO$\nu$A experiments. In Sec.II, we give the formalism and present our results for the total cross section $\sigma$, $Q^{2}$ distribution ($\frac{d\sigma}{dQ^{2}}$), momentum distribution ($\frac{d\sigma}{dp_{\pi}}$) and angular distribution($\frac{d\sigma}{dcos\theta_{\pi q}}$) for $\pi^{0}$ in Sec.III, where we also discuss the possibility of obtaining improved limits on $\nu(\bar{\nu})$ magnetic moments. ## II Formalism In the energy region of one GeV relevant for atmospheric neutrinos and present accelerator neutrino experiments the dominant process of pion production is through the excitation of $\Delta$ resonance and its subsequent decay to pions, i.e. $\nu N\rightarrow\nu\Delta\rightarrow\nu N\pi^{0}.$ (1) Figure 1: Feynman diagram for the process $\nu+p(p)\rightarrow\nu(k^{\prime})+p(p^{\prime})+\pi^{0}(k_{\pi})$. The differential scattering cross section for the reaction $\nu(k)+p(p)\rightarrow\nu(k^{\prime})+p(p^{\prime})+\pi^{0}(k_{\pi})$ shown in Fig.1. is given by $d\sigma=\frac{(2\pi)^{4}\delta^{4}(p_{i}-p_{f})}{4{p\cdot k}}\prod_{j=1}^{3}\frac{d^{3}p_{j}}{(2\pi)^{3}2E_{j}}\|\mathcal{M}_{fi}\|^{2},$ (2) where $p_{i}(=k+p)$ and $p_{f}(=\sum_{j=1}^{3}p_{j})$ are the four momenta of the initial and final states, respectively. The transition matrix element $\mathcal{M}_{fi}$ is written using ${\it L}^{W}=\frac{G_{F}}{\sqrt{2}}l_{\alpha}j^{\alpha}$ (3) for the weak $ZN\Delta$ interaction and ${\it L}_{\pi N\Delta}=\frac{f_{\pi N\Delta}}{m_{\pi}}{\bar{\Psi}}_{\mu}{\vec{T}}^{\dagger}(\partial^{\mu}{\vec{\phi}})\Psi+h.c.$ (4) for the strong $\pi N\Delta$ interaction. $\Psi_{\mu}$ is a Rarita-Schwinger field for spin-$\frac{3}{2}$ particle, ${\vec{T}}^{\dagger}$ is the isospin transition operator, $\vec{\phi}$ is the pion field. The matrix element of the leptonic current $l_{\alpha}$ and the hadronic current $j^{\alpha}$ are defined as $<k\|l_{\alpha}\|k^{\prime}>=\bar{u}({\bf k}^{\prime})\gamma_{\alpha}(1-\gamma^{5})u({\bf k}),$ (5) and $<\Delta(P)\|j^{\alpha}\|p>=\bar{\Psi}_{\beta}({\bf P}){\bf\mathcal{O}}^{\beta\alpha}u({\bf p})$ (6) $u({\bf p})$ is the Dirac spinor for the proton. $\mathcal{O}^{\beta\alpha}=(1-2sin^{2}\theta_{W}){\mathcal{O}}_{V}^{\beta\alpha}+{\mathcal{O}}_{A}^{\beta\alpha}$ for the neutral current process with ${\mathcal{O}}_{V}^{\beta\alpha}$ and ${\mathcal{O}}_{A}^{\beta\alpha}$ given by $\displaystyle{\mathcal{O}}_{V}^{\beta\alpha}$ $\displaystyle=\left(\frac{C_{3}^{V}(q^{2})}{M}(g^{\alpha\beta}\not q-q^{\beta}\gamma^{\alpha})+\frac{C_{4}^{V}(q^{2})}{M^{2}}(g^{\alpha\beta}q\cdot P-q^{\beta}P^{\alpha})+\frac{C_{5}^{V}(q^{2})}{M^{2}}(g^{\alpha\beta}q\cdot p-q^{\beta}p^{\alpha})\right)\gamma_{5}$ (7) and $\displaystyle{\mathcal{O}}_{A}^{\beta\alpha}$ $\displaystyle=$ $\displaystyle\left(\frac{C_{4}^{A}(q^{2})}{M^{2}}(g^{\alpha\beta}{\not q}-q^{\beta}\gamma^{\alpha})+C_{5}^{A}(q^{2})g^{\alpha\beta}+\frac{C_{6}^{A}(q^{2})}{M^{2}}q^{\beta}q^{\alpha}\right)$ (8) where $C_{i}^{V}(q^{2})$ and $C_{i}^{A}(q^{2})$ are the vector and axial vector transition form factors, and $\theta_{W}$ is the Weinberg angle ($sin^{2}\theta_{W}$= 0.23122). q(=$k-k^{\prime}$) is the four momentum transfer. M is the mass of the nucleon. Using Eqs. (3)-(6) the matrix element for the process $\nu+p\rightarrow\nu+p+\pi^{0}$ in the $\Delta$ dominance model is written as $\mathcal{M}_{fi}=\sqrt{\frac{2}{3}}\frac{G_{F}}{\sqrt{2}}\frac{f_{\pi N\Delta}}{m_{\pi}}\bar{u}({\bf p}^{\prime})k^{\sigma}_{\pi}{\mathcal{P}}_{\sigma\lambda}\mathcal{O}^{\lambda\alpha}l_{\alpha}u({\bf p})$ (9) where $\sqrt{\frac{2}{3}}$ has appeared because of the isospin factor coming at the vertex $\Delta^{+}\rightarrow p\pi^{0}$. $G_{F}(=1.16637\times 10^{-5}GeV^{-2})$ is the Fermi coupling constant. ${\mathcal{P}}^{\sigma\lambda}$ is the $\Delta$ propagator in momentum space given by ${\mathcal{P}}^{\sigma\lambda}=\frac{{\it P}^{\sigma\lambda}}{P^{2}-M_{\Delta}^{2}+iM_{\Delta}\Gamma},$ (10) where ${\it P}^{\sigma\lambda}$ is the spin-3/2 projection operator given by $\displaystyle{\it P}^{\sigma\lambda}=\sum_{spins}\psi^{\sigma}\bar{\psi}^{\lambda}=(\not P+M_{\Delta})\left(g^{\sigma\lambda}-\frac{2}{3}\frac{P^{\sigma}P^{\lambda}}{M_{\Delta}^{2}}+\frac{1}{3}\frac{P^{\sigma}\gamma^{\lambda}-P^{\sigma}\gamma^{\lambda}}{M_{\Delta}}-\frac{1}{3}\gamma^{\sigma}\gamma^{\lambda}\right),$ (11) and the delta decay width $\Gamma$ is taken to be an energy dependent P-wave decay width taken as $\Gamma(W)=\frac{1}{6\pi}\left(\frac{f_{\pi N\Delta}}{m_{\pi}}\right)^{2}\frac{M}{W}\|q_{cm}\|^{3}.$ (12) $\|q_{cm}\|$ is the pion momentum in the rest frame of the resonance and is given by $\|q_{cm}\|=\frac{\sqrt{(W^{2}-m_{\pi}^{2}-M^{2})^{2}-4m_{\pi}^{2}M^{2}}}{2W},$ with W as the center of mass energy. If the reaction shown in Eq.(1) is induced by neutrino magnetic moment then the matrix element given by Eq.(9) would modify to $\mathcal{M}_{fi}=\sqrt{\frac{2}{3}}\frac{f_{\pi N\Delta}}{m_{\pi}}\bar{u}({\bf p}^{\prime})k^{\sigma}_{\pi}{\mathcal{P}}_{\sigma\lambda}\mathcal{O}^{\lambda\alpha}_{V}l_{\alpha}^{em}u({\bf p}),$ (13) where $l_{\alpha}^{em}=\mu_{\nu}^{eff}\bar{u}({\bf k}^{\prime}){\frac{\sigma^{\alpha\beta}}{q^{2}}}q_{\beta}u({\bf k}),$ (14) with $\mu_{\nu}^{eff}$ as the effective magnetic moment of the neutrino, which is given in terms of the magnetic moments of the mass eigen states and oscillation probabilities that depend upon the specific oscillation models used for analysing the neutrino oscillation experiments grifols ; grimus ; joshipura ; beacon ; pulido . Figure 2: Total scattering cross section for the reaction $\nu+p\rightarrow\nu+p+\pi^{0}$ induced by weak neutral current and the magnetic moment induced processes. Figure 3: $Q^{2}$ distribution for the weak neutral current and the magnetic moment induced processes at $E_{\nu}=1GeV$. Figure 4: Pion momentum distribution for the weak neutral current and the magnetic moment induced processes at $E_{\nu}=1GeV$. Figure 5: Pion angular distribution for the weak neutral current and the magnetic moment induced processes at $E_{\nu}=1GeV$. We have taken the following form of the $N-\Delta$ transition form factors lalakulich $\displaystyle C_{i}^{V}(Q^{2})$ $\displaystyle=$ $\displaystyle C_{i}^{V}(0)~{}(1+\frac{Q^{2}}{M_{V}^{2}})^{-2}~{}{D_{i}}$ $\displaystyle D_{i}$ $\displaystyle=$ $\displaystyle(1+\frac{Q^{2}}{4M_{V}^{2}})^{-1};~{}~{}i=3,4$ $\displaystyle D_{i}$ $\displaystyle=$ $\displaystyle(1+\frac{Q^{2}}{0.776M_{V}^{2}})^{-1};~{}~{}i=5$ (15) $C_{3}^{V}(0)=2.13,C_{4}^{V}(0)=-1.51,$ $C_{5}^{V}(0)=0.48$ where $M_{V}(=0.84GeV)$ is the vector dipole mass. The axial vector form factors are parametrised as $\displaystyle C_{i}^{A}(Q^{2})$ $\displaystyle=$ $\displaystyle C_{i}^{A}(0)~{}(1+\frac{Q^{2}}{M_{A}^{2}})^{-2}~{}{D_{i}}$ $\displaystyle D_{i}$ $\displaystyle=$ $\displaystyle(1+\frac{Q^{2}}{3M_{A}^{2}})^{-1};$ (16) $C_{4}^{A}(Q^{2})=-\frac{C_{5}^{A}(Q^{2})}{4},C_{5}^{A}(0)=1.2,$ $C_{6}^{A}(Q^{2})=C_{5}^{A}(Q^{2})\frac{M^{2}}{m_{\pi}^{2}+Q^{2}}$ where $M_{A}(=1.05GeV)$ is axial vector dipole mass. The differential scattering cross section $\frac{d^{5}\sigma}{dQ^{2}d\Omega_{\pi}dp_{l}}$ is calculated using Eq.(2), and is written as $\displaystyle\frac{d^{5}\sigma}{dQ^{2}d\Omega_{\pi}dp_{l}}=\frac{1}{(4\pi)^{5}}\frac{\pi}{E_{\nu}E_{l}}\frac{\|\vec{k}^{\prime}\|\|\vec{k}_{\pi}\|}{ME_{\nu}}\frac{1}{E_{p}^{\prime}+E_{\pi}\left(1-\frac{\|\vec{q}\|}{\|\vec{k}_{\pi}\|}cos(\theta_{\pi q})\right)}\bar{\sum}\sum\|\mathcal{M}_{fi}\|^{2}$ (17) where $\|\vec{k}_{\pi}\|$ is the pion momentum. Similarly we get an expression for the pion distribution using Eq. (2). ## III Results and Discussions The numerical results for the total cross section $\sigma$, the differential cross sections $\frac{d\sigma}{dQ^{2}}$, $\frac{d\sigma}{dcos\theta_{\pi q}}$ and $\frac{d\sigma}{dp_{\pi}}$ for the neutral current production of $\pi^{0}$ induced by neutrinos(antineutrinos) are presented in Figs.2-5 along with the contributions of the electromagnetic production induced by the neutrino(antineutrino) magnetic moment. For the neutral current production, the numerical values of the vector and axial vector form factors given in Eqs.(II) & (II) have been used while for the electromagnetic production the numerical values of the vector form factors given in Eq.(II) along with the neutrino magnetic moment $\mu_{\nu}^{eff}=6.8\times 10^{-10}\mu_{B}$ is used. A momentum dependent strong form factor with $f_{\pi N\Delta}(m_{\pi}^{2})=2.12$ Penner has been used in numerical calculations. We show in Fig 2., the total cross section $\sigma$, for the neutral current induced $\nu(\bar{\nu})$ production of $\pi^{0}$. The present results are in agreement with the results of Leitner et al.ruso and also with the results of Hernandez et al.nieves if their values for the parameter $C_{5}^{A}(0)$ and $M_{A}$ are used but are in disagreement with results of Kang et al.kang who find a smaller value for $\sigma$. We also show in this figure, the total cross section $\sigma$ for electromagnetic production of $\pi^{0}$ induced by neutrino magnetic moment which is in agreement with the results of Kang et al.kang if neutrino magnetic moment $\mu_{\nu}^{eff}=6\times 10^{-9}\mu_{B}$ as used by them is taken. We see in Fig. 2. that with the present limits on the magnetic moment of $\nu_{\mu}$, the electromagnetic production of $\pi^{0}$ is $10^{-3}$ times smaller than the neutral current induced $\pi^{0}$ production. It is, therefore, not feasible to improve the present limit on neutrino magnetic moment from $\pi^{0}$ production cross section measurements as earlier expected from the work of Kang et al. kang . In Figs. 3-5, we also show the differential cross sections $\frac{d\sigma}{dQ^{2}}$, $\frac{d\sigma}{dcos\theta_{\pi q}}$ and $\frac{d\sigma}{dp_{\pi}}$ for the neutral current induced $\pi^{0}$ production by $\nu$ and $\bar{\nu}$ as well as the $\pi^{0}$ production induced by magnetic moment of $\nu$($\bar{\nu}$). The present experiments at MiniBooNE Aguiler see neutral current induced $\pi^{0}$ events of the order of $2.8\times 10^{4}$ which can be further increased by an order of magnitude at T2K and NO$\nu$A. These pions are produced on nuclear targets like ${}^{12}C$. In the case of nuclear targets, there are incoherent as well as coherent production of $\pi^{0}$ which have different angular and momentum distributions. An analysis of these experiments in order to study the neutrino magnetic moment would require an understanding of nuclear effects in the incoherent and coherent production of $\pi^{0}$ induced by the neutral currents as well as by the neutrino magnetic moment on nuclear targets in the energy region of 1 GeV. We would like to conclude that it is possible in principle to study the neutrino magnetic moment from the observations of neutral current induced $\pi^{0}$ production from nuclear targets in the near detector in future neutrino oscillation experiments by T2K & NO$\nu$A collaborations. However, with the present limits on $\mu_{\nu_{\mu}}$ the magnetic moment induced $\pi^{0}$ production cross sections are quite smaller then the weak neutral current induced cross sections. It is, therefore, not a feasible method to constrain the neutrino magnetic moment beyond the present experimental limits. ## IV Acknowledgments One of the authors (S.C.) is thankful to the Jawaharlal Nehru Memorial Fund for the Doctoral Fellowship. ## References * (1) A. A. Aguilar-Arevalo et al., Phys, Rev. Lett. 98 231801 (2007). * (2) T. Kajita, Nucl. Phys. B Proc. Suppl. 77 123 (1999). * (3) C. M. Mauger, Ph.D. Thesis, SUNY, Stony Brook [UMI-30-88072, (2002)(unpublished)]. * (4) J. L. Raaf, Ph.D. Thesis, Cincinnati U. [FERMILAB-THESIS-2005-20, (2005)(unpublished)]. * (5) E. Hernandez, J. Nieves, and M. Valverde, Phys. Lett. B 647, 452 (2007). * (6) C. Mariani, A. I. P. Conf. Proc. 967, 174 (2007). * (7) S. Nakayama et al., Phys. Lett. B 619, 255 (2005). * (8) A. A. Aguilar-Arevalo et al., Phys. Lett. B 664, 41 (2008). * (9) V. T. Nguyen, A. I. P. Conf. Proc. 981, 250 (2008). * (10) J. M. Link, A. I. P. Conf. Proc. 967, 151 (2007). * (11) T. Nakadaira, A. I. P. Conf. Proc. 981, 222 (2008). * (12) Y. Kudenko, physics.ins-det/0805.0411, (2008). * (13) A. Norman, A. I. P. Conf. Proc. 981, 225 (2008). * (14) K. Fujikawa, and R. Shrock, Phys. Rev. Lett. 45, 963 (1980). * (15) R. N. Mohapatra and P. B. Pal, “Massive Neutrinos in Physics and Astrophysics”(World Scientific, 1998); J. W. F. Valle, “Gauge Theories and the Physics of Neutrino mass”, Prog. Part. Nucl. Phys. 26, 91 (2001). * (16) W.-M. Yao et al., J. of Phys. G 33, 1 (2006). * (17) J. A. Grifols, E. Masso, and S. Mohanty, Phys. Lett. B 587, 184 (2004). * (18) W. Grimus, M. Maltoni, T. Schwetz, M. A. Tortola, and J. W. F. Valle, Nucl. Phys. B 648, 376 (2003). * (19) A. S. Joshipura and S. Mohanty, Phys. Rev. D 66, 012003 (2002). * (20) J. F. Beacom and P. Vogel, Phys. Rev. Lett. 83, 5222 (1999). * (21) J. Pulido and A. M. Mourao, Phys. Rev. D 57, 7108 (1998). * (22) D. Montanino, M. Picariello, and J. Pulido, Phys. Rev. D 77, 093011(2008). * (23) S. K. Kang, J. E. Kim, and J. S. Lee, Phys. Rev. D 60, 033008 (1999). * (24) Y. Fukuda et al., Phys, Rev. Lett. 81, 1562 (1998). * (25) O. Lalakulich, E. A. Paschos, and G. Piranishvili, Phys. Rev. D 74, 014009 (2006). * (26) G. Penner and U. Mosel, Phys. Rev. C 66, 055212 (2002); M. Post, Ph.D. Thesis, Universitat Giessen, 2004. * (27) T. Leitner, L. Alvarez-Ruso, and U. Mosel, Phys. Rev. C 74, 065502 (2006). * (28) E. Hernandez, J. Nieves, and M. Valverde, Phys. Rev. D 76, 033005 (2007).	arxiv-papers	2008-08-08T14:31:34	2024-09-04T02:48:57.205319	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Sajjad Athar, S. Chauhan and S. K. Singh", "submitter": "Shri Singh krishna", "url": "https://arxiv.org/abs/0808.1221" }
0808.1331	# Semistar dimension of polynomial rings and Prüfer-like domains Parviz Sahandi Department of Mathematics, University of Tabriz, Tabriz, Iran, and School of Mathematics, Institute for Research in Fundamental Sciences (IPM), P.O. Box: 19395-5746, Tehran Iran sahandi@ipm.ir ###### Abstract. Let $D$ be an integral domain and $\star$ a semistar operation stable and of finite type on it. In this paper we define the semistar dimension (inequality) formula and discover their relations with $\star$-universally catenarian domains and $\star$-stably strong S-domains. As an application we give new characterizations of $\star$-quasi-Prüfer domains and UM$t$ domains in terms of dimension inequality formula (and the notions of universally catenarian domain, stably strong S-domain, strong S-domain, and Jaffard domains). We also extend Arnold’s formula to the setting of semistar operations. ###### Key words and phrases: Semistar operation, Krull dimension, strong S-domain, Jaffard domain, quasi- Prüfer domain, UM$t$ domain ###### 2000 Mathematics Subject Classification: Primary 13C15, 13G05, 13A15 P. Sahandi was supported in part by a grant from IPM (No. 88130034) ## 1\. Introduction All rings considered in this paper are (commutative integral) domains (with 1); throughout, $D$ denotes a domain with quotient field $K$. In [22], Okabe and Matsuda introduced the concept of a semistar operation. Let $D$ be an integral domain and $\star$ a semistar operation on $D$. In [24] we defined and studied the $\widetilde{\star}$-Jaffard domains and proved that every $\widetilde{\star}$-Noetherian and P$\star$MD of finite $\widetilde{\star}$-dimension is a $\widetilde{\star}$-Jaffard domain. In [25] we defined and studied two subclasses of $\widetilde{\star}$-Jaffard domains, namely the $\widetilde{\star}$-stably strong S-domains and $\widetilde{\star}$-universally catenarian domains and showed how these notions permit studies of $\widetilde{\star}$-quasi-Prüfer domains in the spirit of earlier works on quasi-Prüfer domains. The next natural step is to seek a semistar analogue of dimension (inequality) formula [15]. In Section 2 of this paper we define the $\widetilde{\star}$-dimension (inequality) formula and show that each $\widetilde{\star}$-universally catenarian domain satisfies the $\widetilde{\star}$-dimension formula and each $\widetilde{\star}$-stably strong S-domain satisfies the $\widetilde{\star}$-dimension inequality formula. In Section 3 we give new characterizations of $\star$-quasi-Prüfer domains and UM$t$ domains in terms of the classical notions of dimension inequality formula, universally catenarian domain, stably strong S-domain, strong S-domain, and Jaffard domains. In the last section we extend Arnold’s formula to the setting of semistar operations (see Theorem 4.6). To facilitate the reading of the introduction and of the paper, we first review some basic facts on semistar operations. Denote by $\overline{\mathcal{F}}(D)$ the set of all nonzero $D$-submodules of $K$, and by $\mathcal{F}(D)$ the set of all nonzero _fractional ideals_ of $D$; i.e., $E\in\mathcal{F}(D)$ if $E\in\overline{\mathcal{F}}(D)$ and there exists a nonzero element $r\in D$ with $rE\subseteq D$. Let $f(D)$ be the set of all nonzero finitely generated fractional ideals of $D$. Obviously, $f(D)\subseteq\mathcal{F}(D)\subseteq\overline{\mathcal{F}}(D)$. As in [22], a semistar operation on $D$ is a map $\star:\overline{\mathcal{F}}(D)\rightarrow\overline{\mathcal{F}}(D)$, $E\mapsto E^{\star}$, such that, for all $x\in K$, $x\neq 0$, and for all $E,F\in\overline{\mathcal{F}}(D)$, the following three properties hold: $\star_{1}$: $(xE)^{\star}=xE^{\star}$; $\star_{2}$: $E\subseteq F$ implies that $E^{\star}\subseteq F^{\star}$; $\star_{3}$: $E\subseteq E^{\star}$ and $E^{\star\star}:=(E^{\star})^{\star}=E^{\star}$. Let $\star$ be a semistar operation on the domain $D$. For every $E\in\overline{\mathcal{F}}(D)$, put $E^{\star_{f}}:=\bigcup F^{\star}$, where the union is taken over all finitely generated $F\in f(D)$ with $F\subseteq E$. It is easy to see that $\star_{f}$ is a semistar operation on $D$, and ${\star_{f}}$ is called _the semistar operation of finite type associated to_ $\star$. Note that $(\star_{f})_{f}=\star_{f}$. A semistar operation $\star$ is said to be of _finite type_ if $\star=\star_{f}$; in particular ${\star_{f}}$ is of finite type. We say that a nonzero ideal $I$ of $D$ is a _quasi- $\star$-ideal_ of $D$, if $I^{\star}\cap D=I$; a _quasi- $\star$-prime_ (ideal of $D$), if $I$ is a prime quasi-$\star$-ideal of $D$; and a _quasi- $\star$-maximal_ (ideal of $D$), if $I$ is maximal in the set of all proper quasi-$\star$-ideals of $D$. Each quasi-$\star$-maximal ideal is a prime ideal. It was shown in [14, Lemma 4.20] that if $D^{\star}\neq K$, then each proper quasi-$\star_{f}$-ideal of $D$ is contained in a quasi-$\star_{f}$-maximal ideal of $D$. We denote by $\operatorname{QMax}^{\star}(D)$ (resp., $\operatorname{QSpec}^{\star}(D)$) the set of all quasi-$\star$-maximal ideals (resp., quasi-$\star$-prime ideals) of $D$. If $\Delta$ is a set of prime ideals of a domain $D$, then there is an associated semistar operation on $D$, denoted by $\star_{\Delta}$, defined as follows: $E^{\star_{\Delta}}:=\cap\\{ED_{P}\|P\in\Delta\\}\text{, for each }E\in\overline{\mathcal{F}}(D).$ If $\Delta=\emptyset$, let $E^{\star_{\Delta}}:=K$ for each $E\in\overline{\mathcal{F}}(D)$. When $\Delta:=\operatorname{QMax}^{\star_{f}}(D)$, we set $\widetilde{\star}:=\star_{\Delta}$. It has become standard to say that a semistar operation $\star$ is stable if $(E\cap F)^{\star}=E^{\star}\cap F^{\star}$ for all $E$, $F\in\overline{\mathcal{F}}(D)$. All spectral semistar operations are stable [14, Lemma 4.1(3)]. In particular, for any semistar operation $\star$, we have that $\widetilde{\star}$ is a stable semistar operation of finite type [14, Corollary 3.9]. The most widely studied (semi)star operations on $D$ have been the identity $d_{D}$, and $v_{D}$, $t_{D}:=(v_{D})_{f}$, and $w_{D}:=\widetilde{v_{D}}$ operations, where $E^{v_{D}}:=(E^{-1})^{-1}$, with $E^{-1}:=(D:E):=\\{x\in K\|xE\subseteq D\\}$. For each quasi-$\star$-prime $P$ of $D$, the $\star$-height of $P$ (for short, $\star$-$\operatorname{ht}(P)$) is defined to be the supremum of the lengths of the chains of quasi-$\star$-prime ideals of $D$, between prime ideal $(0)$ (included) and $P$. Obviously, if $\star=d_{D}$ is the identity (semi)star operation on $D$, then $\star$-$\operatorname{ht}(P)=\operatorname{ht}(P)$, for each prime ideal $P$ of $D$. If the set of quasi-$\star$-prime of $D$ is not empty, the $\star$-dimension of $D$ is defined as follows: $\star\text{-}\operatorname{dim}(D):=\sup\\{\star\text{-}\operatorname{ht}(P)\|P\text{ is a quasi-}\star\text{-prime of }D\\}.$ If the set of quasi-$\star$-primes of $D$ is empty, then pose $\star\text{-}\operatorname{dim}(D):=0$. Thus, if $\star=d_{D}$, then $\star\text{-}\operatorname{dim}(D)=\operatorname{dim}(D)$, the usual (Krull) dimension of $D$. It is known (see [12, Lemma 2.11]) that $\widetilde{\star}\text{-}\operatorname{dim}(D)=\sup\\{\operatorname{ht}(P)\mid P\text{ is a quasi-}\widetilde{\star}\text{-prime ideal of }D\\}.$ Let $\star$ be a semistar operation on a domain $D$. Recall from [12, Section 3] that $D$ is said to be a _$\star$ -Noetherian domain_, if $D$ satisfies the ascending chain condition on quasi-$\star$-ideals. Also recall from [16] that, $D$ is called a _Prüfer $\star$-multiplication domain_ (for short, a P$\star$MD) if each finitely generated ideal of $D$ is _$\star_{f}$ -invertible_; i.e., if $(II^{-1})^{\star_{f}}=D^{\star}$ for all $I\in f(D)$. When $\star=v$, we recover the classical notion of P$v$MD; when $\star=d_{D}$, the identity (semi)star operation, we recover the notion of Prüfer domain. Finally recall from [7] that $D$ is said to be a _$\star$ -quasi-Prüfer domain_, in case, if $Q$ is a prime ideal in $D[X]$, and $Q\subseteq P[X]$, for some $P\in\operatorname{QSpec}^{\star}(D)$, then $Q=(Q\cap D)[X]$. This notion is the semistar analogue of the classical notion of the quasi-Prüfer domains. By [7, Corollary 2.4], $D$ is a $\star_{f}$-quasi-Prüfer domain if and only if $D$ is a $\widetilde{\star}$-quasi-Prüfer domain. ## 2\. The $\star$-dimension (inequality) formula We begin with the following definition. Recall that if $D\subseteq T$ are domains, then $\operatorname{tr.deg.}_{D}(T)$ is defined as the _transcendence degree_ of the quotient field of $T$ over the quotient field of $D$. If $P$ is a prime ideal of $D$, then $\mathbb{K}(P)$ is denoted the residue field of $D$ in $P$, i.e., $D_{P}/PD_{P}$, which is canonically isomorphic to the field of quotients of the integral domain $D/P$. ###### Definition 2.1. Let $D\subseteq T$ be an extension of domain and $\star$ and $\star^{\prime}$ are semistar operation on $D$ and $T$ respectively. We say that $D\subseteq T$ satisfies the _$(\star,\star^{\prime})$ -dimension formula (resp. $(\star,\star^{\prime})$-dimension inequality formula)_ if for all $Q\in\operatorname{QSpec}^{\star^{\prime}}(T)$ such that $(Q\cap D)^{\star}\subsetneq D^{\star}$, $\star^{\prime}\text{-}\operatorname{ht}(Q)+\operatorname{tr.deg.}_{\mathbb{K}(Q\cap D)}(\mathbb{K}(Q))=\star\text{-}\operatorname{ht}(Q\cap D)+\operatorname{tr.deg.}_{D}(T).$ $(\text{resp. }\star^{\prime}\text{-}\operatorname{ht}(Q)+\operatorname{tr.deg.}_{\mathbb{K}(Q\cap D)}(\mathbb{K}(Q))\leq\star\text{-}\operatorname{ht}(Q\cap D)+\operatorname{tr.deg.}_{D}(T).)$ The domain $D$ is said to satisfy the _$\star$ -dimension formula (resp. $\star$-dimension inequality formula)_ if for all finitely generated domain $T$ over $D$, $D\subseteq T$ satisfies the $(\star,d_{T})$-dimension formula (resp. $(\star,d_{T})$-dimension inequality formula). If $\star=d_{D}$ and $\star^{\prime}=d_{T}$, then these definitions coincides with the classical ones (see [20], [AC], and [15]). ###### Proposition 2.2. Let $D$ be a domain and $\star$ a semistar operation on $D$. Then the following conditions are equivalent: * (1) $D$ satisfy the $\widetilde{\star}$-dimension formula (resp. $\widetilde{\star}$-dimension inequality formula); * (2) $D_{P}$ satisfy the dimension formula for each $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$ (resp. dimension inequality formula); * (3) $D_{M}$ satisfy the dimension formula for each $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$ (resp. dimension inequality formula). ###### Proof. We only prove the case of dimension formula and the other case is the same. $(1)\Rightarrow(2)$ Let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. Let $T$ be a finitely generated domain over $D_{P}$. So that there exist finitely many elements $\theta_{1},\cdots,\theta_{r}\in T$ such that $T=D_{P}[\theta_{1},\cdots,\theta_{r}]$. Set $T^{\prime}=D[\theta_{1},\cdots,\theta_{r}]$. Then $T=T^{\prime}_{D\backslash P}$ and $T^{\prime}$ is a finitely generated domain over $D$. Let $Q$ be a prime ideal of $T$ and set $qD_{P}:=Q\cap D_{P}$, where $q(\subseteq P)$ be a prime ideal of $D$. Thus there exists a prime ideal $Q^{\prime}$ of $T^{\prime}$ such that $Q^{\prime}\cap(D\backslash P)=\emptyset$ and $Q=Q^{\prime}T^{\prime}_{D\backslash P}$. Thus $Q^{\prime}\cap D=q$. Since $q\subseteq P$, we have $q$ is a quasi-$\widetilde{\star}$-prime ideal of $D$. Since $\widetilde{\star}\text{-}\operatorname{ht}(q)=\operatorname{ht}(q)$, then by the hypothesis we have: $\operatorname{ht}(Q^{\prime})+\operatorname{tr.deg.}_{\mathbb{K}(q)}(\mathbb{K}(Q^{\prime}))=\operatorname{ht}(q)+\operatorname{tr.deg.}_{D}(T^{\prime}).$ Since $\operatorname{ht}(Q^{\prime})=\operatorname{ht}(Q)$ we see that $\operatorname{ht}(Q)+\operatorname{tr.deg.}_{\mathbb{K}(q)}(\mathbb{K}(Q))=\operatorname{ht}(q)+\operatorname{tr.deg.}_{D_{P}}(T).$ $(2)\Rightarrow(3)$ is trivial. $(3)\Leftarrow(1)$ Suppose that $T$ is a finitely generated domain over $D$. Let $Q\in\operatorname{Spec}(T)$ and set $P:=Q\cap D$ such that $P^{\widetilde{\star}}\subsetneq D^{\widetilde{\star}}$. Thus $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\cup\\{0\\}$. Let $M$ be a quasi-$\widetilde{\star}$-maximal ideal of $D$ containing $P$. Note that $T_{D\backslash M}$ is a finitely generated domain over $D_{M}$ and that $Q\cap(D\backslash M)\neq\emptyset$. Thus $QT_{D\backslash M}\in\operatorname{Spec}(T_{D\backslash M})$ and $PD_{M}=QT_{D\backslash M}\cap D_{M}$. Therefore by the (3), we have $\operatorname{ht}(QT_{D\backslash M})+\operatorname{tr.deg.}_{\mathbb{K}(PD_{M})}(\mathbb{K}(QT_{D\backslash M}))=\operatorname{ht}(PD_{M})+\operatorname{tr.deg.}_{D_{M}}(T_{D\backslash M}).$ Now since $\widetilde{\star}\text{-}\operatorname{ht}(Q)=\operatorname{ht}(Q)=\operatorname{ht}(QT_{D\backslash M})$, $\operatorname{ht}(P)=\operatorname{ht}(PD_{M})$, $\operatorname{tr.deg.}_{\mathbb{K}(P)}(\mathbb{K}(Q))=\operatorname{tr.deg.}_{\mathbb{K}(PD_{M})}(\mathbb{K}(QT_{D\backslash M}))$ and $\operatorname{tr.deg.}_{D}(T)=\operatorname{tr.deg.}_{D_{M}}(T_{D\backslash M})$ the proof is complete. ∎ Let $D$ be an integral domain with quotient field $K$, let $X$, $Y$ be two indeterminates over $D$ and let $\star$ be a semistar operation on D. Set $D_{1}:=D[X]$, $K_{1}:=K(X)$ and take the following subset of $\operatorname{Spec}(D_{1})$: $\Theta_{1}^{\star}:=\\{Q_{1}\in\operatorname{Spec}(D_{1})\|\text{ }Q_{1}\cap D=(0)\text{ or }(Q_{1}\cap D)^{\star_{f}}\subsetneq D^{\star}\\}.$ Set $\mathfrak{S}_{1}^{\star}:=\mathcal{S}(\Theta_{1}^{\star}):=D_{1}[Y]\backslash(\bigcup\\{Q_{1}[Y]\|Q_{1}\in\Theta_{1}^{\star}\\})$ and: $E^{\circlearrowleft_{\mathfrak{S}_{1}^{\star}}}:=E[Y]_{\mathfrak{S}_{1}^{\star}}\cap K_{1},\text{ for all }E\in\overline{\mathcal{F}}(D_{1}).$ It is proved in [24, Theorem 2.1] that the mapping $\star[X]:=\circlearrowleft_{\mathfrak{S}_{1}^{\star}}:\overline{\mathcal{F}}(D_{1})\to\overline{\mathcal{F}}(D_{1})$, $E\mapsto E^{\star[X]}$ is a stable semistar operation of finite type on $D[X]$, i.e., $\widetilde{\star[X]}=\star[X]$. It is also proved that $\widetilde{\star}[X]=\star_{f}[X]=\star[X]$, $d_{D}[X]=d_{D[X]}$ and $\operatorname{QSpec}^{\star[X]}(D[X])=\Theta_{1}^{\star}\backslash\\{0\\}$. If $X_{1},\cdots,X_{r}$ are indeterminates over $D$, for $r\geq 2$, we let $\star[X_{1},\cdots,X_{r}]:=(\star[X_{1},\cdots,X_{r-1}])[X_{r}],$ where $\star[X_{1},\cdots,X_{r-1}]$ is a stable semistar operation of finite type on $D[X_{1},\cdots,X_{r-1}]$. For an integer $r$, put $\star[r]$ to denote $\star[X_{1},\cdots,X_{r}]$ and $D[r]$ to denote $D[X_{1},\cdots,X_{r}]$. Following [25], the domain $D$ is called _$\star$ -catenary_, if for each pair $P\subset Q$ of quasi-$\star$-prime ideals of $D$, any two saturated chain of quasi-$\star$-prime ideals between $P$ and $Q$ have the same finite length. If for each $n\geq 1$, the polynomial ring $D[n]$ is $\star[n]$-catenary, then $D$ is said to be _$\star$ -universally catenarian_. Every P$\star$MD which is $\widetilde{\star}$-LFD (that is $\operatorname{ht}(P)<\infty$ for all $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$), is $\widetilde{\star}$-universally catenarian by [25, Theorem 3.4]. ###### Corollary 2.3. Let $D$ be an $\widetilde{\star}$-universally catenarian domain. Then $D$ satisfies the $\widetilde{\star}$-dimension formula. ###### Proof. Let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. Hence $D_{P}$ is a universally catenarian domain by [25, Lemma 3.3]. Thus by [5, Corollary 4.8], $D_{P}$ satisfies the dimension formula. Now Proposition 2.2 completes the proof. ∎ The domain $D$ is called a _$\star$ -strong S-domain_, if each pair of adjacent quasi-$\star$-prime ideals $P_{1}\subset P_{2}$ of $D$, extend to a pair of adjacent quasi-$\star[X]$-prime ideals $P_{1}[X]\subset P_{2}[X]$, of $D[X]$. If for each $n\geq 1$, the polynomial ring $D[n]$ is a $\star[n]$-strong S-domain, then $D$ is said to be an _$\star$ -stably strong S-domain_. Every $\widetilde{\star}$-Noetherian, $\widetilde{\star}$-quasi- Prüfer or $\widetilde{\star}$-universally catenarian domain is $\widetilde{\star}$-stably strong S-domain by [25, Corollaries 2.6 and 3.6]. ###### Corollary 2.4. Let $D$ be an $\widetilde{\star}$-stably strong S-domain. Then $D$ satisfies the $\widetilde{\star}$-dimension inequality formula. ###### Proof. Use [25, Proposition 2.5] and [20, Theorem 1.6] and the same argument as proof of Corollary 2.3. ∎ A valuation overring $V$ of $D$ is called a _$\star$ -valuation overring of $D$_ provided $F^{\star}\subseteq FV$, for each $F\in f(D)$. Following [24], the _$\star$ -valuative dimension_ of $D$ is defined as: $\star\text{-}\operatorname{dim}_{v}(D):=\sup\\{\operatorname{dim}(V)\|V\text{ is }\star\text{-valuation overring of }D\\}.$ Although Example 4.4 of [24] shows that $\star$-$\operatorname{dim}(D)$ is not always less that or equal to $\star$-$\operatorname{dim}_{v}(D)$, but it is observed in [24] that $\widetilde{\star}$-$\operatorname{dim}(D)\leq\widetilde{\star}$-$\operatorname{dim}_{v}(D)$. We say that $D$ is a _$\star$ -Jaffard domain_, if $\star\text{-}\operatorname{dim}(D)=\star\text{-}\operatorname{dim}_{v}(D)<\infty$. When $\star=d$ the identity operation then $d$-Jaffard domain coincides with the classical Jaffard domain (cf. [1]). It is proved in [24], that $D$ is a $\widetilde{\star}$-Jaffard domain if and only if $\star[X_{1},\cdots,X_{n}]\text{-}\operatorname{dim}(D[X_{1},\cdots,X_{n}])=\widetilde{\star}\text{-}\operatorname{dim}(D)+n,$ for each positive integer $n$. ###### Lemma 2.5. For each domain $D$, $\widetilde{\star}\text{-}\operatorname{dim}_{v}(D)=\sup\\{\operatorname{dim}_{v}(D_{P})\|P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\\}$. ###### Proof. We can assume that $\widetilde{\star}\text{-}\operatorname{dim}_{v}(D)$ is a finite number. Suppose that $n=\widetilde{\star}\text{-}\operatorname{dim}_{v}(D)$. Then there exists a $\widetilde{\star}$-valuation overring $V$, with maximal ideal $N$, of $D$ such that $\operatorname{dim}(V)=n$. Set $P:=N\cap D$. So that $V$ is a valuation overring of $D_{P}$. Hence $n=\operatorname{dim}(V)\leq\operatorname{dim}_{v}(D_{P})\leq\widetilde{\star}\text{-}\operatorname{dim}_{v}(D)=n$, where the second inequality is true since each valuation overring of $D_{P}$ is a $\widetilde{\star}$-valuation overring of $D$ ([17, Theorem 3.9]). ∎ In [1, Page 174] it is proved that a finite-dimensional domain satisfying the dimension inequality formula is a Jaffard domain. In the following result we give the semistar analogue of the mentioned result. ###### Theorem 2.6. Let $D$ be a domain of finite $\widetilde{\star}$-dimension. If $D$ satisfies the $\widetilde{\star}$-dimension inequality formula, then $D$ is a $\widetilde{\star}$-Jaffard domain. ###### Proof. Let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. Then $D_{P}$ is a finite dimensional domain and satisfies the dimension inequality formula by Proposition 2.2. Consequently $D_{P}$ is a Jaffard domain by [1]. Thus using Lemma 2.5, we have $\displaystyle\widetilde{\star}\text{-}\operatorname{dim}(D)=$ $\displaystyle\sup\\{\operatorname{dim}(D_{P})\|P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\\}$ $\displaystyle=$ $\displaystyle\sup\\{\operatorname{dim}_{v}(D_{P})\|P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\\}$ $\displaystyle=$ $\displaystyle\widetilde{\star}\text{-}\operatorname{dim}_{v}(D).$ Thus $D$ is a $\widetilde{\star}$-Jaffard domain. ∎ Therefore we have the following implications for finite $\widetilde{\star}$-dimensional domains: $\begin{array}[]{ccc}\textsf{$\widetilde{\star}$-Noetherian}\text{ or }\textsf{$\widetilde{\star}$-quasi-Pr\"{u}fer}&&\textsf{P$\star$MD}\\\ \Downarrow&&\Downarrow\\\ \textsf{$\widetilde{\star}$-stably strong S-domain}&\Leftarrow&\textsf{$\widetilde{\star}$-universally catenary}\\\ \Downarrow&&\Downarrow\\\ \textsf{$\widetilde{\star}$-dimension inequality formula}&\Leftarrow&\textsf{$\widetilde{\star}$-dimension formula}\\\ \Downarrow&&\\\ \textsf{$\widetilde{\star}$-Jaffard}&&\end{array}$ Let $D$ be a domain with quotient field $K$, let $X$ be an indeterminate over $D$, let $\star$ be a semistar operation on D, and let $P$ be a quasi-$\star$-prime ideal of $D$ (or $P=0$). Set $\mathcal{S}_{P}^{\star}:=(D/P)[X]\backslash\bigcup\\{(Q/P)[X]\mid Q\in\operatorname{QSpec}^{\star_{f}}(D)\text{ and }P\subseteq Q\\}.$ Clearly, $\mathcal{S}_{P}^{\star}$ is a multiplicatively closed subset of $(D/P)[X]$. For all $E\in\overline{\mathcal{F}}(D/P)$, set $E^{\circlearrowleft_{\mathcal{S}_{P}^{\star}}}:=E(D/P)[X]_{\mathcal{S}_{P}^{\star}}\cap(D_{P}/PD_{P}).$ It is proved in [10, Theorem 3.2] that the mapping $\star/P:=\circlearrowleft_{\mathcal{S}_{P}^{\star}}:\overline{\mathcal{F}}(D/P)\to\overline{\mathcal{F}}(D/P)$, $E\mapsto E^{\circlearrowleft_{\mathcal{S}_{P}^{\star}}}$, is a stable semistar operation of finite type on $D/P$; i.e., $\widetilde{\star/P}=\star/P$, $\operatorname{QMax}^{\star/P}(D/P)=\\{Q/P\in\operatorname{Spec}(D/P)\mid Q\in\operatorname{QMax}^{\star_{f}}(D)\text{ and }P\subseteq Q\\}$, $\widetilde{\star}/P=\star_{f}/P=\star/P$ and $d_{D}/P=d_{D/P}$. ###### Lemma 2.7. A domain $D$ is $\widetilde{\star}$-universally catenarian if and only if $D/P$ is $(\star/P)$-universally catenarian for each $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. ###### Proof. $(\Rightarrow)$ Let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. By [10, Theorem 3.2 (a)], $\star/P=\widetilde{\star/P}$. Hence, by [25, Proposition 3.2 and Lemma 3.3], $D/P$ is $(\star/P)$-universally catenarian if and only if $(D/P)_{\mathcal{M}}$ is a universally catenarian domain for each $\mathcal{M}\in\operatorname{QMax}^{\star/P}(D/P)$, that is (by [10, Theorem 3.2 (b)]), if and only if $D_{M}/PD_{M}$ is a universally catenarian domain whenever $P$ is a subset of $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$. But by [25, Proposition 3.2 and Lemma 3.3], $D_{M}$ is a universally catenarian domain for all $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$. This, in turn, is immediate since any factor domain of a universally catenarian domain must be a universally catenarian domain. $(\Leftarrow)$ It is enough to consider $P=0$, since we have $\star/0=\widetilde{\star}$. ∎ In [21, Corollary 14.D] it is proved that a Noetherian domain $D$ is an universally catenarian domain if and only if $D$ is catenary and $D/P$ satisfies the dimension formula for each $P\in\operatorname{Spec}(D)$. In the following result we give the semistar analogue of this result. ###### Theorem 2.8. Let $D$ be a $\widetilde{\star}$-Noetherian domain. Then $D$ is an $\widetilde{\star}$-universally catenarian domain if and only if $D$ is $\widetilde{\star}$-catenary and $D/P$ satisfies the $(\star/P)$-dimension formula for each $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. ###### Proof. $(\Rightarrow)$ Let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. Then $D/P$ is $(\star/P)$-universally catenarian by Lemma 2.7. Hence $D/P$ satisfies the $(\star/P)$-dimension formula by Corollary 2.3. $(\Leftarrow)$ Let $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$. It is enough to show that $D_{M}$ is a universally catenarian domain. To this end let $PD_{M}$ be a prime ideal of $D_{M}$. Thus $P$ is a quasi-$\widetilde{\star}$-prime ideal of $D$. Since $D/P$ satisfies the $(\star/P)$-dimension formula, then $(D/P)_{M/P}=D_{M}/PD_{M}$ satisfies the dimension formula by Theorem 2.2. On the other hand $D_{M}$ is a Noetherian domain by [12, Proposition 3.8] and catenary by [25, Proposition 3.2]. Consequently $D_{M}$ is a universally catenarian domain by [21, Corollary 14.D]. ∎ Recall that the celebrated theorem of Ratliff [23, Theorem 2.6] says that a Noetherian ring $R$ is universally catenarian if and only if $R[X]$ is catenarian. On the other hand it is proved in [6, Theorem 1] that the Noetherian assumption in Ratliff’s theorem can be replaced with going-down condition by proving that: for a going-down domain $D$, we have $D$ is universally catenarian if and only if $D[X]$ is catenarian if and only if $D$ is an LFD strong S-domain. As a semistar analogue in [25, Theorem 3.7] we proved that: suppose that $D$ is $\widetilde{\star}$-Noetherian. Then $D[X]$ is $\star[X]$-catenary if and only if $D$ is $\widetilde{\star}$-universally catenarian. In the last theorem of this section we treat the second case. Let $D\subseteq T$ be an extension of domains. Let $\star$ and $\star^{\prime}$ be semistar operations on $D$ and $T$, respectively. Following [9], we say that $D\subseteq T$ satisfies $(\star,\star^{\prime})$-$\operatorname{GD}$ if, whenever $P_{0}\subset P$ are quasi-$\star$-prime ideals of $D$ and $Q$ is a quasi-$\star^{\prime}$-prime ideal of $T$ such that $Q\cap D=P$, there exists a quasi-$\star^{\prime}$-prime ideal $Q_{0}$ of $T$ such that $Q_{0}\subseteq Q$ and $Q_{0}\cap D=P_{0}$. The integral domain $D$ is said to be a _$\star$ -going-down domain_ (for short, a $\star$-$\operatorname{GD}$ domain) if, for every overring $T$ of $D$ the extension $D\subseteq T$ satisfies $(\star,d_{T})$-$\operatorname{GD}$. These concepts are the semistar versions of the “classical” concepts of going-down property and the going-down domains (cf. [8]). It is known by [9, Propositions 3.5 and 3.2(e)] that every P$\star$MD and every integral domain $D$ with $\star$-$\operatorname{dim}(D)=1$ is a $\star$-$\operatorname{GD}$ domain. ###### Theorem 2.9. Let $D$ be a $\widetilde{\star}$-$\operatorname{GD}$ domain. The following statements are equivalent: * (1) $D$ is a $\widetilde{\star}$-LFD $\widetilde{\star}$-strong S-domain. * (2) $D$ is $\widetilde{\star}$-universally catenarian. * (3) $D[X]$ is $\star[X]$-catenarian. ###### Proof. $(1)\Rightarrow(2)$ holds by [25, Theorem 4.1] and $(2)\Rightarrow(3)$ is trivial. For $(3)\Rightarrow(1)$ let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. Then $D_{P}$ is a going- down domain by [10, Proposition 2.5] and $D_{P}[X]$ is catenarian by [25, Lemma 3.3]. Thus $D_{P}$ is a LFD strong S-domain by [6, Theorem 1]. Hence $D$ is a $\widetilde{\star}$-LFD $\widetilde{\star}$-strong S-domain by [25, Proposition 2.4]. ∎ ## 3\. Characterizations of $\star$-quasi-Prüfer domains In this section we give some characterization of $\widetilde{\star}$-quasi- Prüfer domains. We need to recall the definition of $(\star,\star^{\prime})$-linked overrings. Let $D$ be a domain and $T$ an overring of $D$. Let $\star$ and $\star^{\prime}$ be semistar operations on $D$ and $T$, respectively. One says that $T$ is _$(\star,\star^{\prime})$ -linked to_ $D$ (or that $T$ is a $(\star,\star^{\prime})$-linked overring of $D$) if $F^{\star}=D^{\star}\Rightarrow(FT)^{\star^{\prime}}=T^{\star^{\prime}}$, when $F$ is a nonzero finitely generated ideal of $D$ (cf. [11]). In particular we are interested in the case $\star^{\prime}=d_{T}$. We first recall the following characterization of $\widetilde{\star}$-quasi-Prüfer domains. ###### Theorem 3.1. ([25, Theorem 4.3]) Let $D$ be an integral domain. Suppose that $\widetilde{\star}$-$\operatorname{dim}(D)$ is finite. Consider the following statements: * (1’) Each $(\star,\star^{\prime})$-linked overring $T$ of $D$ is an $\widetilde{\star^{\prime}}$-universally catenarian domain. * (1) Each $(\star,\star^{\prime})$-linked overring $T$ of $D$ is an $\widetilde{\star^{\prime}}$-stably strong S-domain. * (2) Each $(\star,\star^{\prime})$-linked overring $T$ of $D$ is an $\widetilde{\star^{\prime}}$-strong S-domain. * (3) Each $(\star,\star^{\prime})$-linked overring $T$ of $D$ is an $\widetilde{\star^{\prime}}$-Jaffard domain. * (4) Each $(\star,\star^{\prime})$-linked overring $T$ of $D$ is an $\widetilde{\star^{\prime}}$-quasi-Prüfer domain. * (5) $D$ is an $\widetilde{\star}$-quasi-Prüfer domain. Then $(1^{\prime})\Rightarrow(1)\Leftrightarrow(2)\Leftrightarrow(3)\Leftrightarrow(4)\Leftrightarrow(5)$. ###### Proof. The implication $(1^{\prime})\Rightarrow(1)$ holds by [25, Corollary 3.6] and $(1)\Rightarrow(2)$ is trivial. For $(2)\Rightarrow(5)$ see proof of [25, Theorem 4.3] part $(3)\Rightarrow(6)$. The implication $(5)\Rightarrow(1)$ holds by [25, Corollary 2.6]. For $(4)\Leftrightarrow(5)\Leftrightarrow(6)$ see [24, Theorem 4.14]. ∎ Now we have the following theorem; a result reminiscent of the well-known result of Ayache, Cahen and Echi [4] (see also [15, Theorem 6.7.8]) for quasi- Prüfer domains. ###### Theorem 3.2. Let $D$ be an integral domain. Suppose that $\widetilde{\star}$-$\operatorname{dim}(D)$ is finite. Then the following statements are equivalent: * (1) Each $(\star,d_{T})$-linked overring $T$ of $D$ is a stably strong S-domain. * (2) Each $(\star,d_{T})$-linked overring $T$ of $D$ is a strong S-domain. * (3) Each $(\star,d_{T})$-linked overring $T$ of $D$ is a Jaffard domain. * (4) Each $(\star,d_{T})$-linked overring $T$ of $D$ is a quasi-Prüfer domain. * (5) $D$ is an $\widetilde{\star}$-quasi-Prüfer domain. ###### Proof. We only prove the equivalence of $(1)\Leftrightarrow(5)$ and the proofs of $(2)\Leftrightarrow(5)$, $(3)\Leftrightarrow(5)$, and $(4)\Leftrightarrow(5)$, are the same. The implication $(5)\Rightarrow(1)$ holds by Theorem 3.1. For $(1)\Rightarrow(5)$ let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. It is enough for us to show that $D_{P}$ is a quasi-Prüfer domain by [7, Theorem 2.16]. To this end let $T$ be an overring of $D_{P}$. Then $T_{D\backslash P}=T$ and therefore $T$ is $(\star,d_{T})$-linked overring of $D$ by [11, Example 3.4 (1)]. Thus by the hypothesis we have $T$ is a stably strong S-domain. Therefore $D_{P}$ is a quasi-Prüfer domain by [15, Theorem 6.7.8]. ∎ ###### Theorem 3.3. Let $D$ be an integral domain. Suppose that $\widetilde{\star}$-$\operatorname{dim}(D)$ is finite. Then the following statements are equivalent: * (1) $D$ is an $\widetilde{\star}$-quasi-Prüfer domain. * (2) For each $(\star,\star^{\prime})$-linked overring $T$ of $D$, every extension of domains $T\subseteq S$, satisfies the $(\widetilde{\star^{\prime}},\widetilde{\star^{\prime\prime}})$-dimension inequality formula, where $\star^{\prime}$ and $\star^{\prime\prime}$ are semistar operations on $T$ and $S$ respectively. ###### Proof. $(1)\Rightarrow(2)$ If $D$ is an $\widetilde{\star}$-quasi-Prüfer domain and $T$ is $(\star,\star^{\prime})$-linked to $D$, then $T$ is a $\widetilde{\star^{\prime}}$-Jaffard domain by [24, Theorem 4.14]. Let $Q\in\operatorname{QSpec}^{\widetilde{\star^{\prime\prime}}}(S)$ such that $(Q\cap T)^{\widetilde{\star^{\prime}}}\subsetneq T^{\widetilde{\star^{\prime}}}$ and set $q:=Q\cap T$. Then we have $q\in\operatorname{QSpec}^{\widetilde{\star^{\prime}}}(T)\cup\\{0\\}$. Set $P:=q\cap D$. Thus we have $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\cup\\{0\\}$. Therefore $D_{P}$ and hence $T_{q}$, are quasi-Prüfer domains by [7, Theorem 1.1]. In particular $T_{q}$ is a Jaffard domain. So that we have $\displaystyle\operatorname{dim}(S_{Q})+\operatorname{tr.deg.}_{\mathbb{K}(q)}(\mathbb{K}(Q))\leq$ $\displaystyle\operatorname{dim}_{v}(S_{Q})+\operatorname{tr.deg.}_{\mathbb{K}(q)}(\mathbb{K}(Q))$ $\displaystyle\leq$ $\displaystyle\operatorname{dim}_{v}(T_{q})+\operatorname{tr.deg.}_{T}(S),$ where the first inequality holds since $\operatorname{dim}(S_{Q})\leq\operatorname{dim}_{v}(S_{Q})$ and the second one is by [15, Lemma 6.7.3]. The conclusion follows easily from the fact that $\operatorname{dim}(T_{q})=\operatorname{dim}_{v}(T_{q})$. $(2)\Rightarrow(1)$ Let $T$ be an overring of $D$ and $\star^{\prime}$ be a semistar operation on $T$ such that $T$ is $(\star,\star^{\prime})$-linked to $D$. Let $(V,N)$ be any $\widetilde{\star^{\prime}}$-valuation overring of $T$. Then $V$ is $(\star^{\prime},d_{V})$-linked to $T$ by [12, Lemma 2.7]. Set $Q:=N\cap T$. Then by assumption we have $\operatorname{dim}(V)\leq\operatorname{dim}(T_{Q})-\operatorname{tr.deg.}_{\mathbb{K}(Q)}(\mathbb{K}(N)).$ In particular $\operatorname{dim}(V)\leq\operatorname{dim}(T_{Q})\leq\widetilde{\star^{\prime}}\text{-}\operatorname{dim}(T)$, and hence $\widetilde{\star^{\prime}}\text{-}\operatorname{dim}_{v}(T)=\widetilde{\star^{\prime}}\text{-}\operatorname{dim}(T)$, that is $T$ is a $\widetilde{\star^{\prime}}$-Jaffard domain. Thus $D$ is an $\widetilde{\star}$-quasi-Prüfer domain by [24, Theorem 4.14]. ∎ ###### Corollary 3.4. Let $D$ be an integral domain. Suppose that $\widetilde{\star}$-$\operatorname{dim}(D)$ is finite. Then the following statements are equivalent: * (1) $D$ is an $\widetilde{\star}$-quasi-Prüfer domain. * (2) For each $(\star,d_{T})$-linked overring $T$ of $D$, every extension of domains $T\subseteq S$, satisfies the dimension inequality formula. ###### Proof. $(1)\Rightarrow(2)$ holds by Theorem 3.3. For $(2)\Rightarrow(1)$ let $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. It is enough for us to show that $D_{P}$ is a quasi-Prüfer domain by [7, Theorem 2.16]. To this end let $T$ be an overring of $D_{P}$. Then $T_{D\backslash P}=T$ and therefore $T$ is $(\star,d_{T})$-linked overring of $D$ by [11, Example 3.4 (1)]. If $T\subseteq S$ is any extension of domains, then $T\subseteq S$ satisfies the dimension inequality formula by the hypothesis. Therefore $D_{P}$ is a quasi- Prüfer domain by [15, Theorem 6.7.4]. ∎ Recall that an integral domain $D$ is called a _UM $t$-domain_ if every upper to zero in $D[X]$ is a maximal $t$-ideal and has been studied by several authors (See [19], [13] and [7]). It is observed in [7, Corollary 2.4 (b)] that $D$ is a $w$-quasi-Prüfer domain if and only if $D$ is a UM$t$-domain. The following corollary is a new characterization of UM$t$ domains. ###### Corollary 3.5. Let $D$ be an integral domain. Suppose that $w$-$\operatorname{dim}(D)$ is finite. Then the following statements are equivalent: * (1) Each $(t_{D},d_{T})$-linked overring $T$ of $D$ is a stably strong S-domain. * (2) Each $(t_{D},d_{T})$-linked overring $T$ of $D$ is a strong S-domain. * (3) Each $(t_{D},d_{T})$-linked overring $T$ of $D$ is a Jaffard domain. * (4) Each $(t_{D},d_{T})$-linked overring $T$ of $D$ is a quasi-Prüfer domain. * (5) For each $(t_{D},d_{T})$-linked overring $T$ of $D$, every extension of domains $T\subseteq S$, satisfies the dimension inequality formula. * (6) $D$ is a UM$t$ domain. ## 4\. Arnold’s formula In the last section we extends some results of J. Arnold of the dimension of polynomial rings to the setting of the semistar operations. First we wish to give the following lemma which is a new property of semistar valuative dimension. ###### Lemma 4.1. (see [24, Theorem 4.2]) Let $D$ be an integral domain and $n$ be an integer. Then the following statements are equivalent: * (1) Each $(\star,d_{T})$-linked overring $T$ of $D$ has dimension at most $n$. * (2) Each $\widetilde{\star}$-valuation overring of $D$ has dimension at most $n$. ###### Proof. The implication $(1)\Rightarrow(2)$ is trivial. For $(2)\Rightarrow(1)$ let $T$ be a $(\star,d_{T})$-linked overring of $D$ and $V$ be a valuation overring of $T$. Then it is easy to see that $V$ is $(\star,d_{V})$-linked overring of $D$. Thus by [12, Lemma 2.7], $V$ is an $\widetilde{\star}$-valuation overring of $D$. Hence $\operatorname{dim}(V)\leq n$. Consequently $\operatorname{dim}(T)\leq\operatorname{dim}_{v}(T)\leq n$ as desired. ∎ When $\star=d_{D}$, the equivalence of (1) and (3) of the following theorem is due to J. Arnold [2, Theorem 6]. ###### Theorem 4.2. Let $D$ be an integral domain, and $n$ be an integer. Then the following statements are equivalent: * (1) $\widetilde{\star}$-$\operatorname{dim}_{v}(D)=n$. * (2) $\star[n]\text{-}\operatorname{dim}(D[n])=2n$. * (3) $\star[r]\text{-}\operatorname{dim}(D[r])=r+n$ for all $r\geq n-1$. * (4) Each $(\star,d_{T})$-linked overring $T$ of $D$ has dimension at most $n$, and $n$ is minimal. ###### Proof. The equivalence $(1)\Leftrightarrow(2)$ follows from [24, Theorem 4.5], and $(3)\Rightarrow(2)$ is trivial. For $(1)\Rightarrow(3)$ suppose that $\widetilde{\star}$-$\operatorname{dim}_{v}(D)=n$. Then For all $r\geq n$ we have $\star[r]\text{-}\operatorname{dim}(D[r])=\star[r]\text{-}\operatorname{dim}_{v}(D[r])=r+\widetilde{\star}\text{-}\operatorname{dim}_{v}(D)=r+n$, by [24, Corollary 4.7 and Theorem 4.8]. Now assume that $r=n-1$. Since $\widetilde{\star}$-$\operatorname{dim}_{v}(D)=n$, there exists a quasi-$\widetilde{\star}$-prime ideal $M$ of $D$ such that $n=\operatorname{dim}_{v}(D_{M})$, by Lemma 2.5. So that by [2, Theorem 6] we have $\operatorname{dim}(D_{M}[r])=r+n$. Let $\mathcal{P}\in\operatorname{QSpec}^{\star[r]}(D[r])$ be such that $\star[r]\text{-}\operatorname{dim}(D[r])=\operatorname{ht}(\mathcal{P})$. Set $P:=\mathcal{P}\cap D$. Then by [24, Remark 2.3] we have $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\cup\\{(0)\\}$. Thus $\displaystyle r+n\leq$ $\displaystyle\star[r]\text{-}\operatorname{dim}(D[r])=\operatorname{ht}(\mathcal{P})$ $\displaystyle=$ $\displaystyle\operatorname{dim}(D[r]_{\mathcal{P}})=\operatorname{dim}(D_{P}[r]_{\mathcal{P}D_{P}[r]})$ $\displaystyle\leq$ $\displaystyle\operatorname{dim}(D_{P}[r])\leq\operatorname{dim}(D_{M}[r])=r+n,$ where the first inequality holds by [24, Theorem 3.1]. Hence $\star[r]\text{-}\operatorname{dim}(D[r])=r+n$ for all $r\geq n-1$. The equivalence $(1)\Leftrightarrow(4)$ follows from Lemma 4.1. ∎ As an immediate consequence we have: ###### Corollary 4.3. $\widetilde{\star}\text{-}\operatorname{dim}_{v}(D)=\sup\\{\operatorname{dim}(T)\|T\text{ is }(\star,d_{T})\text{-linked overring of }D\\}$. One of the famous formulas in the dimension theory of commutative rings is the Arnold’s formula [2, Theorem 5] which states as $\operatorname{dim}(D[n])=n+\sup\\{\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])\|\\{\theta_{i}\\}_{1}^{n}\subseteq K\\}.$ Now we prove the semistar analogue of Arnold’s formula. ###### Lemma 4.4. Let $D$ be an integral domain and $n$ be an integer. Then $\star[n]\text{-}\operatorname{dim}(D[n])=\sup\\{\operatorname{dim}(D_{M}[n])\|M\in\operatorname{QMax}^{\widetilde{\star}}(D)\\}.$ ###### Proof. If $P$ is a quasi-$\widetilde{\star}$-prime ideal of $D$, and if $QD_{P}[n]$ is a non-zero prime ideal of $D_{P}[n](=D[n]_{D\backslash P})$, then $Q\cap D\subseteq P$ and hence $Q\in\operatorname{QSpec}^{\star[n]}(D[n])$ by [24, Remark 2.3]. So that the inequality $\geq$ is true. Now let $Q\in\operatorname{QMax}^{\star[n]}(D[n])$ be such that $\star[n]\text{-}\operatorname{dim}(D[n])=\operatorname{ht}(Q)$, and set $P:=Q\cap D$. Then by [24, Remark 2.3] we have $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)\cup\\{0\\}$. So that $\displaystyle\star[n]\text{-}\operatorname{dim}(D[n])=$ $\displaystyle\operatorname{ht}(Q)=\operatorname{dim}(D[n]_{Q})$ $\displaystyle=$ $\displaystyle\operatorname{dim}(D_{P}[n]_{QD_{P}[n]})\leq\operatorname{dim}(D_{P}[n])$ $\displaystyle\leq$ $\displaystyle\star[n]\text{-}\operatorname{dim}(D[n]).$ Therefore the proof is complete. ∎ ###### Corollary 4.5. Let $D$ be an integral domain and $n$ be an integer. Then there exist a quasi-$\widetilde{\star}$-maximal ideal $M$ of $D$ and a quasi-$\star[n]$-maximal ideal $Q$ of $D[n]$ such that $M=Q\cap D$ and $\star[n]\text{-}\operatorname{dim}(D[n])=\operatorname{ht}(Q)=n+\operatorname{ht}(M[n]).$ ###### Proof. By Lemma 4.4 there exists a quasi-$\widetilde{\star}$-maximal ideal $M$ of $D$ such that $\star[n]\text{-}\operatorname{dim}(D[n])=\operatorname{dim}(D_{M}[n])$. Thus there exists a prime ideal $Q$ of $D[n]$ such that $Q\cap(D\backslash M)=\emptyset$, $\operatorname{dim}(D_{M}[n])=\operatorname{ht}(QD_{M}[n])$ and that $QD_{M}[n]$ is a maximal ideal of $D_{M}[n]$. Since $Q\cap D\subseteq M$ we have $Q$ is a quasi-$\star[n]$-prime of $D[n]$ by [24, Remark 2.3], and since $\star[n]\text{-}\operatorname{dim}(D[n])=\operatorname{ht}(Q)$, we have $Q$ is a quasi-$\star[n]$-maximal ideal of $D[n]$. Set $PD_{M}:=QD_{M}[n]\cap D_{M}$ for some $P\in\operatorname{QSpec}^{\widetilde{\star}}(D)$. Then by [3, Corollary 2.9] we have $\operatorname{ht}(QD_{M}[n])=n+\operatorname{ht}(PD_{M}[n])$ and that $PD_{M}$ is a maximal ideal of $D_{M}$. Thus we have $P=M$ and $\star[n]\text{-}\operatorname{dim}(D[n])=\operatorname{ht}(QD_{M}[n])=n+\operatorname{ht}(MD_{M}[n])=n+\operatorname{ht}(M[n]),$ which ends the proof. ∎ We are ready to prove the semistar analogue of Arnold’s formula. ###### Theorem 4.6. Let $D$ be an integral domain and $n$ be a positive integer. Then $\star[n]\text{-}\operatorname{dim}(D[n])=n+\sup\\{\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])\|\\{\theta_{i}\\}_{1}^{n}\subseteq K\\}.$ where $\iota$ is the inclusion map of $D$ in $D[\theta_{1},\cdots,\theta_{n}]$. ###### Proof. Let $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$ and $\\{\theta_{i}\\}_{1}^{n}\subseteq K$. Let $Q$ be a maximal ideal of $D_{M}[\theta_{1},\cdots,\theta_{n}]$ such that $\operatorname{dim}(D_{M}[\theta_{1},\cdots,\theta_{n}])=\operatorname{ht}(Q)$. Let $Q_{0}$ be a prime ideal of $D[\theta_{1},\cdots,\theta_{n}]$ such that $Q_{0}\cap(D\backslash M)=\emptyset$ and $Q=Q_{0}D_{M}[\theta_{1},\cdots,\theta_{n}]$. Thus $Q_{0}$ is a quasi-$\widetilde{\star}_{\iota}$-prime ideal of $D[\theta_{1},\cdots,\theta_{n}]$ since $Q_{0}\cap D\subseteq M$ ([24, Remark 2.3]). Hence we obtain that $\operatorname{dim}(D_{M}[\theta_{1},\cdots,\theta_{n}])=\operatorname{ht}(Q)=\operatorname{ht}(Q_{0})\leq\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])$. Using Lemma 4.4 and Arnold’s formula [2, Theorem 5], we have $\star[n]\text{-}\operatorname{dim}(D[n])=n+\sup\\{\operatorname{dim}(D_{M}[\theta_{1},\cdots,\theta_{n}])\\},$ where the supremum is taken over $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$ and $\\{\theta_{i}\\}_{1}^{n}\subseteq K$. So that $\star[n]\text{-}\operatorname{dim}(D[n])\leq n+\sup\\{\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])\|\\{\theta_{i}\\}_{1}^{n}\subseteq K\\}$. Now choose $M\in\operatorname{QMax}^{\widetilde{\star}}(D)$ and $\\{\theta_{i}\\}_{1}^{n}\subseteq K$ such that $\star[n]\text{-}\operatorname{dim}(D[n])=n+\operatorname{dim}(D_{M}[\theta_{1},\cdots,\theta_{n}])$. Let $Q^{\prime}$ be a quasi-$\widetilde{\star}_{\iota}$-prime ideal of $D[\theta_{1},\cdots,\theta_{n}]$ such that $\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])=\operatorname{ht}(Q^{\prime})$ and set $P^{\prime}:=Q^{\prime}\cap D$. Thus $\displaystyle\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])=$ $\displaystyle\operatorname{ht}(Q^{\prime})=\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}]_{Q^{\prime}})$ $\displaystyle=$ $\displaystyle\operatorname{dim}(D_{P^{\prime}}[\theta_{1},\cdots,\theta_{n}]_{Q^{\prime}D_{P^{\prime}}[\theta_{1},\cdots,\theta_{n}]})$ $\displaystyle\leq$ $\displaystyle\operatorname{dim}(D_{P^{\prime}}[\theta_{1},\cdots,\theta_{n}])\leq\operatorname{dim}(D_{M}[\theta_{1},\cdots,\theta_{n}]).$ Hence by the first part of the proof $\operatorname{dim}(D_{M}[\theta_{1},\cdots,\theta_{n}])=\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])$. Thus we have $\star[n]\text{-}\operatorname{dim}(D[n])=n+\widetilde{\star}_{\iota}\text{-}\operatorname{dim}(D[\theta_{1},\cdots,\theta_{n}])$ to complete the proof. ∎ ## References * [1] D. F. Anderson, A. Bouvier, D. Dobbs, M. Fontana and S. Kabbaj, On Jaffard domain, Expo. Math., 6, (1988), 145–175. * [2] J. T. Arnold, On the dimension theory of overrings of an integral domain, Trans. Amer. Math. Soc., 138, (1969), 313–326. * [3] J. T. Arnold and R. Gilmer, The dimension sequence of a commutative ring, Amer. J. Math., 96, (1974), 385–408. * [4] A. Ayache and P. Cahen and O. Echi, Anneaux quasi-Prüfériens et P-anneaux. Boll. Un. Mat. Ital 10-B, (1996), 1–24. * [5] A. Bouvier, D. E. Dobbs and M. Fontana, Universally catenarian integral domains, Advances in Math., 72 (1988), 211–238. * [6] A. Bouvier, D. E. Dobbs and M. Fontana, Two sufficient conditions for universal catenarity, Comm. Alg., 15 (1987), 861–872. * [7] G.W. Chang and M. Fontana, Uppers to zero in polynomial rings and Prüfer-like domains, Comm. Algebra 37 (2009), 164–192. * [8] D. E. Dobbs and I. J. Papick, On going-down for simple overrings III, Proc. Amer. Math. Soc. 54 (1976), 35–38. * [9] D. E. Dobbs and P. Sahandi, Going-down and Semistar operations, J. Algebra Appl. 8 (2009), 83–104. * [10] D. E. Dobbs and P. Sahandi, On semistar Nagata rings, Prüfer-like domains and semistar going-down domains, Houston J. Math. to appear. * [11] S. El Baghdadi and M. Fontana, Semistar linkedness and flatness, Prüfer semistar multiplication domains, Comm. Algebra 32 (2004), 1101–1126. * [12] S. El Baghdadi, M. Fontana and G. Picozza, Semistar Dedekind domains, J. Pure Appl. Algebra 193 (2004), 27–60. * [13] M. Fontana, S. Gabelli and E. Houston, UMT-domains and domains with Prüfer integral closure, Comm. Algebra 26, (1998), 1017–1039. * [14] M. Fontana and J. A. Huckaba, Localizing systems and semistar operations, in: S. Chapman and S. Glaz (Eds.), Non Noetherian Commutative Ring Theory, Kluwer, Dordrecht, 2000, 169–197. * [15] M. Fontana, J. Huckaba, and I. Papick, Prüfer domains, New York, Marcel Dekker, 1997. * [16] M. Fontana, P. Jara and E. Santos, Prüfer $\star$-multiplication domains and semistar operations, J. Algebra Appl. 2 (2003), 21–50. * [17] M. Fontana and K. A. Loper, Nagata rings, Kronecker function rings and related semistar operations, Comm. Algebra 31 (2003), 4775–4801. * [18] R. Gilmer, Multiplicative ideal theory, New York, Dekker, 1972. * [19] E. Houston and M. Zafrullah, On $t$-invertibility, II, Comm. Algebra 17 (1989), 1955–1969. * [20] S. Kabbaj, La formule de la dimension pour les S-domaines forts universels, Boll. Un. Mat. Ital., D (6) 5, No. 1, (1986), 145–161. * [21] H. Matsumura, Commutative algebra, Benjamin, New York, 1970. * [22] A. Okabe and R. Matsuda, Semistar-operations on integral domains, Math. J. Toyama Univ. 17 (1994), 1–21. * [23] L. J. Ratliff, Jr., On quasi-unmixed local domains, the altitude formula, and the chain condition for prime ideals, II, Amer. J. Math. 92, (1970), 99–144. * [24] P. Sahandi, Semistar-Krull and valuative dimension of integral domains, Ricerche Mat. 58, no. 2, (2009), 219–242. * [25] P. Sahandi, Universally catenarian integral domains, strong S-domains and semistar operations, Comm. Algebra 38, No. 2, (2010), 673–683.	arxiv-papers	2008-08-09T05:59:49	2024-09-04T02:48:57.210906	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Parviz Sahandi", "submitter": "Parviz Sahandi Dr.", "url": "https://arxiv.org/abs/0808.1331" }
0808.1408	There are infinitely many prime numbers in all arithmetic progressions with first term and difference coprime.111Originally published in Abhandlungen der Königlich Preussischen Akademie der Wissenschaften von 1837, 45–81. Translated by Ralf Stephan, eMail: mailto:ralf@ark.in-berlin.de. The scanned images are available at http://bibliothek.bbaw.de/bibliothek- digital/digitalequellen/schriften/ anzeige?band=07-abh/1837&seite:int=00000286 (By Mr. Lejeune-Dirichlet) [Read to the Academy of Sciences the 27th of July, 1837] MSC-Class: 01A55; 11-03; 11N13; 11B25 Observant investigation of the series of primes will perceive several properties with a generality that can be lifted to any degree of probability by using continued induction, while the discovery of their proof with all required strictness presents the greatest difficulties. One of the most curious results of this kind appears when we divide the members of the series through an arbitrary number. If we ignore the primes dividing the divisor, which are among the first members of the series, all other primes will leave a residue that is coprime to the divisor. The main result after continued division is that every residue appears to return infinitely often, and in particular, that the ratio of the values that indicate how often two arbitrary residues have come up until a specific position is reached will have unity as limit if we continue division indefinitely. After abstracting away the constancy of appearance of single residues and limiting the result to the never ending of the appearance of each residue, we can state the latter as the theorem: “that each unlimited arithmetic progression, with the first member and the difference being coprime, will contain infinitely many primes.” No proof existed for this simple theorem until now, however desirable such a proof would have been for the numerous applications that can be made of the theorem. The only mathematician who has tried a justification of this theorem, as far as I know, is Legendre[4], who should not only have been interested in investigating it because the difficulty of the subject would have appealed to him, but especially also because he used the mentioned property of arithmetic progressions as lemma with some earlier works. Legendre bases a possible proof on the task to find the longest run of members of an arithmetic progression that are divisible by given primes, but he solves it only by induction. If one tries to prove the solution of that task, which was thus found by him and which is highly strange because of its simplicity, then one encounters great difficulties that I did not succeed to overcome. Only after I entirely left the line taken by Legendre I arrived at a completely strict proof of the theorem on arithmetic progressions. The proof so found, which I have the honor to present to the Academy in this paper, is not fully arithmetical but based partly on the study of continous variables. With the novelty of the applied principles it appeared useful to me to start with the treatment of the special case where the difference of the progression is an odd prime, before proving the theorem in its entire generality. §. 1. Let $p$ an odd prime and $c$ a primitive root of it such that the residues of the powers: $c^{0},c^{1},c^{2},\ldots,c^{p-2},$ when divided by $p$, and ordered, are identical to the numbers: $1,2,3,\ldots,p-1$. Let $n$ a number not divisible by $p$, then, after Gauss, we will call the exponent $\gamma<p-1$ which satisfies the congruence $c^{\gamma}\equiv n$ (mod. $p$) the index of $n$ and, if necessary, denote it as $\gamma_{n}$. The choice of the primitive root shall be arbitrary as long as it is fixed. Regarding the above defined indices the easily provable theorem holds that the index of a product equals the sum of indices of its factors minus the included multiple of $p-1$. Further we notice that always $\gamma_{1}=0$, $\gamma_{p-1}=\frac{1}{2}(p-1)$, as well as $\gamma_{n}$ is even or odd, according to $n$ being quadratic residue of $p$ or not, or applying Legendre’s symbol, according to $\bigl{(}\frac{n}{p}\bigr{)}=+1$ or $\bigl{(}\frac{n}{p}\bigr{)}=-1$. Now let $q$ a prime different from $p$ (not excluding $2$) and $s$ a positive variable greater than unity. Further we denote as $\omega$ an arbitrary root of the equation: (1) $\omega^{p-1}-1=0,$ and we construct the geometrical series: (2) $\frac{1}{1-\omega^{\gamma}\frac{1}{q^{s}}}=1+\omega^{\gamma}\frac{1}{q^{s}}+\omega^{2\gamma}\frac{1}{q^{2s}}+\omega^{3\gamma}\frac{1}{q^{3s}}+\cdots,$ where $\gamma$ means the index of $q$. If we substitute for $q$ every prime different from $p$ and multiply the so formed equations in each other we get a series on the right hand with a structure that is easily perceived. Namely, let $n$ any integer not divisible by $p$, and let $n=q^{\prime m^{\prime}}q^{\prime\prime m^{\prime\prime}}\ldots$, where $q^{\prime}$, $q^{\prime\prime},\ldots$ denote different primes, then the general term will be of the form: $\omega^{m^{\prime}\gamma_{q^{\prime}}+m^{\prime\prime}\gamma_{q^{\prime\prime}}+\cdots}\frac{1}{n^{s}}.$ But now it holds that: $m^{\prime}\gamma_{q^{\prime}}+m^{\prime\prime}\gamma_{q^{\prime\prime}}+\cdots\equiv\gamma_{n}\mbox{ (mod.\ $p-1$)},$ and because of (1): $\omega^{m^{\prime}\gamma_{q^{\prime}}+m^{\prime\prime}\gamma_{q^{\prime\prime}}+\cdots}=\omega^{\gamma_{n}}.$ Therefore we have the equation: (3) $\prod\frac{1}{1-\omega^{\gamma}\frac{1}{q^{s}}}=\sum\omega^{\gamma}\frac{1}{n^{s}}=L,$ where the multiplication sign applies to the whole series of primes with the only exception of $p$, while the sum is over all integers from $1$ to $\infty$ not divisible by $p$. The letter $\gamma$ means on the left $\gamma_{q}$, but on the right $\gamma_{n}$. The equation just found represents $p-1$ different equations that result if we put for $\omega$ its $p-1$ values. It is known that these $p-1$ different values can be written using powers of the same $\Omega$ when it is chosen correctly, to wit: $\Omega^{0},\Omega^{1},\Omega^{2},\ldots,\Omega^{p-2}.$ According to this notation, we will write the different values $L$ of the series or product as: (4) $L_{0},L_{1},L_{2},\ldots,L_{p-2}$ where it is obvious that $L_{0}$ and $L_{\frac{1}{2}(p-1)}$ have a meaning independent of the choice of $\Omega$ and that they relate to $\omega=1$ and $\omega=-1$, respectively. Before we go on it is necessary to state the reason for the condition made above, that $s>1$ should hold. We can convince ourselves of the necessity of this limitation if we respect the essential difference which exists between two kinds of infinite series. If we regard each value instead of each term or, it being imaginary, its module, then two cases can happen. Either it is possible to give a finite value which is greater than the sum of any of however many of these values or moduli, or this condition cannot be satisfied by any finite number. In the first case, the series always converges and has a completely defined sum regardless how the series terms are ordered, be it that they continue to two and more dimensions or that they comprise a double or multiple series. In the second case the series can converge too but convergence is essentially dependent on the kind of order of terms. Does convergence hold for a specific order then it can stop when this order is changed, or, if this does not happen, then the sum of the series might become completely different. So, for example, of the two series made from the same terms: $\displaystyle 1-\frac{1}{\sqrt{2}}+\frac{1}{\sqrt{3}}-\frac{1}{\sqrt{4}}+\frac{1}{\sqrt{5}}-\frac{1}{\sqrt{6}}+\cdots,$ $\displaystyle 1+\frac{1}{\sqrt{3}}-\frac{1}{\sqrt{2}}+\frac{1}{\sqrt{5}}+\frac{1}{\sqrt{7}}-\frac{1}{\sqrt{4}}+\cdots,$ only the first converges while of the following: $\displaystyle 1-\frac{1}{2}+\frac{1}{3}-\frac{1}{4}+\frac{1}{5}-\frac{1}{6}+\cdots,$ $\displaystyle 1+\frac{1}{3}-\frac{1}{2}+\frac{1}{5}+\frac{1}{7}-\frac{1}{4}+\cdots,$ both converge, but with different sums. Our infinite series $L$, as can be easily seen, belongs only then to the first of the classes we just differentiated if we assume $s>1$, such that if we put, under this condition, $L=\lambda+\mu\sqrt{-1}$, then $\lambda$ and $\mu$ will get completely defined finite values. Denoting now with $f_{m}+g_{m}\sqrt{-1}$ the product of the first $m$ factors of the form $\frac{1}{1-\omega^{\gamma}\frac{1}{q^{s}}}$, where the order of the factors may be arbitrary, we can always make $m$ of a size such that among these first $m$ factors will be all those satisfying $q<h$ with $h$ an arbitrary integer. As soon as $m$ reaches this size, each of both differences $f_{m}-\lambda$, $g_{m}-\mu$, will obviously, ignoring the sign, always stay smaller than $\frac{1}{h^{s}}+\frac{1}{(h+1)^{s}}+\cdots$, however large $m$ may be imagined to grow further. Under the assumption $s>1$ however, the value $\frac{1}{h^{s}}+\frac{1}{(h+1)^{s}}+\cdots$ may shrink arbitrarily small with a correspondingly huge $h$. Therefore, it is proved that the infinite product in (3) has a value equal to the series $L$, independent of the order of factors. With $s=1$ or $s<1$ however, this proof can be no longer applied and, in fact, the infinite product has then in general no longer a definite value, regardless of factor ordering. If we could prove the existence of a limit for the multiplication continued to infinity, given a specific ordering of the factors, then the equation (3), understood correctly, would still hold but would have no use for the statement of the value. We would then, $q^{\prime},q^{\prime\prime},q^{\prime\prime},\ldots$ being the values of $q$ according to the assumed ordering, have to view the series $L$ as a multiple series that has to be ordered such that first those members would be taken where $n$ only contains the prime factor $q^{\prime}$, then those of the rest where $n$ has no other factors than $q^{\prime},q^{\prime\prime}$ and so on. From the necessity to order the members this way the summation of the series would become as difficult as the investigation of the product itself already is. §. 2. If we put $s=1+\varrho$ the equation (3) still holds, however small the positive value $\varrho$ is assumed. We want to study now how the series $L$ in (3) changes if $\varrho$ is let become infinitely small. The behaviour of the series with respect to this is quite different, according to $\omega$ being equal to positive unity or having any other value. To begin with the first case, or the investigation of $L_{0}$, we look at the sum: $S=\frac{1}{k^{1+\varrho}}+\frac{1}{(k+1)^{1+\varrho}}+\frac{1}{(k+2)^{1+\varrho}}+\cdots,$ where $k$ denotes a positive constant. If we substitute in the well-known formula: $\int_{0}^{1}x^{k-1}\log^{\varrho}\left(\frac{1}{x}\right)\,dx=\frac{\Gamma(1+\varrho)}{k^{1+\varrho}}$ for $k$ sequentially $k,\,k+1,\,k+2,\ldots$ and add, we get: $S=\frac{1}{\Gamma(1+\varrho)}\int_{0}^{1}\log^{\varrho}\left(\frac{1}{x}\right)\frac{x^{k-1}}{1-x}\,dx.$ If we add $\frac{1}{\varrho}$ and subtract at the same time: $\frac{1}{\varrho}=\frac{\Gamma(\varrho)}{\Gamma(1+\varrho)}=\frac{1}{\Gamma(1+\varrho)}\int_{0}^{1}\log^{\varrho-1}\left(\frac{1}{x}\right)\,dx,$ the equation transforms into: $S=\frac{1}{\varrho}+\frac{1}{\Gamma(1+\varrho)}\int_{0}^{1}\left(\frac{x^{k-1}}{1-x}-\frac{1}{\log\left(\frac{1}{x}\right)}\right)\log^{\varrho}\left(\frac{1}{x}\right)\,dx,$ where the second member, for $\varrho$ infinitely small, approaches the finite limit: $\int_{0}^{1}\left(\frac{x^{k-1}}{1-x}-\frac{1}{\log\left(\frac{1}{x}\right)}\right)\,dx.$ Regarding instead of the series $S$ the more general one which has two positive constants $a$, $b$: $\frac{1}{b^{1+\varrho}}+\frac{1}{(b+a)^{1+\varrho}}+\frac{1}{(b+2a)^{1+\varrho}}+\cdots,$ we need only transform it into: $\frac{1}{a^{1+\varrho}}\Biggl{(}\frac{1}{\bigl{(}\frac{b}{a}\bigr{)}^{1+\varrho}}+\frac{1}{\bigl{(}\frac{b}{a}+1\bigr{)}^{1+\varrho}}+\frac{1}{\bigl{(}\frac{b}{a}+2\bigr{)}^{1+\varrho}}+\cdots\Biggr{)}$ and compare with $S$ to see immediately that it equals to an expression of the following form: $\frac{1}{a}\cdot\frac{1}{\varrho}+\varphi(\varrho),$ where $\varphi(\varrho)$ approaches a finite limit with $\varrho$ becoming infinitely small. The studied series $L_{0}$ consists of $p-1$ partial series like: $\frac{1}{m^{1+\varrho}}+\frac{1}{(p+m)^{1+\varrho}}+\frac{1}{(2p+m)^{1+\varrho}}+\cdots,$ where we have to assume successively $m=1,\,2,\ldots,\,p-1$. We have thus: (5) $L_{0}=\frac{p-1}{p}\cdot\frac{1}{\varrho}+\varphi(\varrho),$ where again $\varphi(\varrho)$ is a function of $\varrho$ that, whenever $\varrho$ gets infinitely small, has a finite value which could be easily expressed through a definite integral, given what we found so far. This is not necessary for our task, however. The equation (5) shows that, for infinitely small $\varrho$, $L_{0}$ will become $\infty$ such that $L_{0}-\frac{p-1}{p}\cdot\frac{1}{\varrho}$ remains finite. §. 3. After we have found how our series, with $\omega=1$ assumed, changes with $s$ approaching unity from above it remains for us to extend the same study to the other roots $\omega$ of the equation $\omega^{p-1}-1=0$. Although the sum of the series $L$ is independent of the ordering of its members, as long as $s>1$, it will still be of advantage if we imagine the members ordered such that the values of $n$ will continously increase. On this condition, $\sum\omega^{\gamma}\frac{1}{n^{s}}$ will be a function of $s$ that remains continous and finite for all positive values of $s$. Thus the limit that is approached by the value of the series if $s=1+\varrho$ and $\varrho$ is let become infinitely small, and which is independent of the ordering of the members is expressed by: $\sum\omega^{\gamma}\frac{1}{n}$ which wouldn’t necessarily be with a different ordering as $\sum\omega^{\gamma}\frac{1}{n}$ would then differ from $\sum\omega^{\gamma}\frac{1}{n^{1+\varrho}}$ by a finite amount or might not even have a value. To prove the statement just made we denote as $h$ an arbitrary positive integer and express the sum of the first $h(p-1)$ members of the series: $\sum\omega^{\gamma}\frac{1}{n^{s}}$ with the help of the formula already used above, which holds for any positive $s$: $\int_{0}^{1}x^{n-1}\log^{s-1}\left(\frac{1}{x}\right)\,dx=\frac{\Gamma(s)}{n^{s}}$ by a definite integral. We therefore get for the sum: $\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{\frac{1}{x}\,f(x)}{1-x^{p}}\log^{s-1}\left(\frac{1}{x}\right)\,dx-\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{\frac{1}{x}\,f(x)}{1-x^{p}}x^{hp}\log^{s-1}\left(\frac{1}{x}\right)\,dx$ where we have used the abbreviation: $f(x)=\omega^{\gamma_{1}}x+\omega^{\gamma_{2}}x^{2}+\cdots+\omega^{\gamma_{p-1}}x^{p-1}.$ If we assume now $\omega$ not $=1$, then the polynomial $\frac{1}{x}\,f(x)$ is divisible by $1-x$ because we have: $f(1)=\omega^{\gamma_{1}}+\omega^{\gamma_{2}}+\cdots+\omega^{\gamma_{p-1}}=1+\omega+\cdots+\omega^{p-2}=0.$ If we eliminate thus the factor $1-x$ from numerator and denominator of the fraction under the integral sign the fraction becomes: $\frac{t+u\sqrt{-1}}{1+x+x^{2}+\cdots+x^{p-1}},$ where $t$ and $u$ are polynomials with real coefficients. If we write $T$ and $U$ for the largest possible values of $t$ and $u$ between $x=0$ and $x=1$ then obviously the real and imaginary parts of the second integral: $\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{\frac{1}{x}\,f(x)}{1-x^{p}}x^{hp}\log^{s-1}\left(\frac{1}{x}\right)\,dx$ are smaller than $\displaystyle\frac{T}{\Gamma(s)}\int_{0}^{1}x^{hp}\log^{s-1}\left(\frac{1}{x}\right)\,dx$ $\displaystyle=$ $\displaystyle\frac{T}{(hp+1)^{s}},$ $\displaystyle\frac{U}{\Gamma(s)}\int_{0}^{1}x^{hp}\log^{s-1}\left(\frac{1}{x}\right)\,dx$ $\displaystyle=$ $\displaystyle\frac{U}{(hp+1)^{s}},$ respectively, and so the integral disappears for $h=\infty$. The series: $\sum\omega^{\gamma}\frac{1}{n^{s}},$ with the assumed ordering of its members, converges therefore, and for its sum we have the expression: $\sum\omega^{\gamma}\frac{1}{n^{s}}=\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{\frac{1}{x}\,f(x)}{1-x^{p}}\log^{s-1}\left(\frac{1}{x}\right)\,dx.$ This function of $s$ not only itself remains continous and finite as long as $s>0$, but the same property applies also to its derivative with respect to $s$. To convince oneself of this it is enough to remember that $\Gamma(s)$, $\frac{d\Gamma(s)}{ds}$, is continous and finite too, and that $\Gamma(s)$ never disappears as long as $s$ remains positive. Thus, if we put: $\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{\frac{1}{x}\,f(x)}{1-x^{p}}\log^{s-1}\left(\frac{1}{x}\right)\,dx=\psi(x)+\chi(s)\sqrt{-1},$ where $\psi(s)$ und $\chi(s)$ are real functions, we have for positive $\varrho$, after a well known theorem: (6) $\psi(1+\varrho)=\psi(1)+\varrho\psi^{\prime}(1+\delta\varrho),\qquad\chi(1+\varrho)=\chi(1)+\varrho\chi^{\prime}(1+\varepsilon\varrho),$ where we abbreviated: $\psi^{\prime}(s)=\frac{d\psi(s)}{ds},\qquad\chi^{\prime}(s)=\frac{d\chi(s)}{ds}$ and denoted as $\delta$ and $\varepsilon$ positive fractions independent of $\varrho$. Incidentally, it is easily understandable that with $\omega=-1$ we get: $\chi(s)=0$, and that if we go from an imaginary root $\omega$ to its conjugate $\frac{1}{\omega}$ then $\psi(s)$ will have the same value while $\chi(s)$ will become its negative. §. 4. We have to prove now that the finite limit approached by $\sum\omega^{\gamma}\frac{1}{n^{1+\varrho}}$, with the positive $\varrho$ becoming infinitely small, and given that $\omega$ does not mean the root $1$, will be nonzero. This limit is, after the last section $\sum\omega^{\gamma}\frac{1}{n}$ and expressed by the integral $\sum\omega^{\gamma}\frac{1}{n}=-\int_{0}^{1}\frac{\frac{1}{x}\,f(x)}{x^{p}-1}\,dx$ which can be easily written using logarithms and circular functions. Let us take an arbitrary linear factor of the denominator $x^{p}-1$: $x-e^{\frac{2m\pi}{p}\sqrt{-1}},$ where $m$ is of the series $0,\,1,\,2,\ldots,\,p-1$. If we decompose: $\frac{\frac{1}{x}\,f(x)}{x^{p}-1}$ into partial fractions then, after known formulae, the numerator of the fraction: $\frac{A_{m}}{x-e^{\frac{2m\pi}{p}\sqrt{-1}}}$ is given by the expression: $\frac{\frac{1}{x}\,f(x)}{px^{p-1}}$ where $x=e^{\frac{2m\pi}{p}\sqrt{-1}}$. So we have: $A_{m}=\frac{1}{p}f\left(e^{\frac{2m\pi}{p}\sqrt{-1}}\right).$ If we substitute this value and note that $A_{0}=0$ we get: $\sum\omega^{\gamma}\frac{1}{n}=-\frac{1}{p}\sum f\left(e^{\frac{2m\pi}{p}\sqrt{-1}}\right)\int_{0}^{1}\frac{dx}{x-e^{\frac{2m\pi}{p}\sqrt{-1}}},$ where the sum on the right goes from $m=1$ to $m=p-1$. The function: $f\left(e^{\frac{2m\pi}{p}\sqrt{-1}}\right)$ is well known from cyclotomy and can be easily related to: $f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right).$ It namely holds that $f\left(e^{\frac{2m\pi}{p}\sqrt{-1}}\right)=\sum\omega^{\gamma_{g}}e^{gm\frac{2\pi}{p}\sqrt{-1}},$ where the sum is from $g=1$ to $g=p-1$. If we substitute for $gm$ the respective residue $h$ modulo $p$ then $1,\,2,\ldots,\,p-1$ become the different values of $h$, and we have, because of $gm\equiv h$ (mod. $p$): $\gamma_{g}\equiv\gamma_{h}-\gamma_{m}\mbox{ (mod.\ $p-1$)}.$ Thus if we write $\gamma_{h}-\gamma_{m}$ for $\gamma_{g}$, which is allowed because of the equation $\omega^{p-1}-1=0$, then we get: $f\left(e^{\frac{2m\pi}{p}\sqrt{-1}}\right)=\omega^{-\gamma_{m}}\sum\omega^{\gamma_{h}}e^{h\frac{2\pi}{p}\sqrt{-1}}=\omega^{-\gamma_{m}}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right).$ The equation above becomes therefore: $\sum\omega^{\gamma}\frac{1}{n}=-\frac{1}{p}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)\sum\omega^{-\gamma_{m}}\int_{0}^{1}\frac{dx}{x-e^{\frac{2m\pi}{p}\sqrt{-1}}}.$ Now, for any positive fraction $\alpha$: $\int_{0}^{1}\frac{dx}{x-e^{2\alpha\pi\sqrt{-1}}}=\log(2\sin\alpha\pi)+\frac{\pi}{2}(1-2\alpha)\sqrt{-1},$ therefore: $\sum\omega^{\gamma}\frac{1}{n}=-\frac{1}{p}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)\sum\omega^{-\gamma_{m}}\left(\log\left(2\sin\frac{m\pi}{p}\right)+\frac{\pi}{2}\left(1-\frac{2m}{p}\right)\sqrt{-1}\right).$ Although this expression is very simple for $\sum\omega^{\gamma}\frac{1}{n}$, in general we cannot conclude that $\sum\omega^{\gamma}\frac{1}{n}$ has a nonzero value. What is missing are fitting principles for the statement of conditions under which transcendent compounds containing undefined integers can disappear. But our desired proof succeeds for the specific case where $\omega=-1$. For imaginary values of $\omega$ we will give another method in the following section that, however, cannot be applied to the mentioned specific case. On the condition $\omega=-1$, and with $\gamma_{m}$ even or odd according to: $\bigl{(}\frac{m}{p}\bigr{)}=+1$ or $-1$, and thus $(-1)^{-\gamma_{m}}=\bigl{(}\frac{m}{p}\bigr{)}$, as well as $(-1)^{\gamma_{n}}=\bigl{(}\frac{n}{p}\bigr{)}$, we get as limit of $L_{\frac{1}{2}(p-1)}$ for $\varrho$ becoming infinitely small: $\sum\left(\frac{n}{p}\right)\frac{1}{n}\\\ =-\frac{1}{p}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)\sum\left(\frac{m}{p}\right)\left(\log\left(2\sin\frac{m\pi}{p}\right)+\frac{\pi}{2}\left(1-\frac{2m}{p}\right)\sqrt{-1}\right),$ or more simple, since $\sum\bigl{(}\frac{m}{p}\bigr{)}=0$ if we sum from $m=1$ to $m=p-1$: $\sum\left(\frac{n}{p}\right)\frac{1}{n}=-\frac{1}{p}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)\sum\left(\frac{m}{p}\right)\left(\log\left(2\sin\frac{m\pi}{p}\right)-\frac{\pi}{p}m\sqrt{-1}\right).$ We have to distinguish two cases, depending on the prime $p$ having the form $4\mu+3$ or $4\mu+1$. In the first case it holds for two values, like $m$ und $p-m$, adding to $p$ that: $\left(\frac{m}{p}\right)=-\left(\frac{p-m}{p}\right)\mbox{ and }\sin\frac{m\pi}{p}=\sin\frac{(p-m)\pi}{p}.$ Therefore the real part of the sum disappears and we get, denoting with $a$ those values of $m$ for which $\bigl{(}\frac{m}{p}\bigr{)}=1$ and with $b$ those for which $\bigl{(}\frac{m}{p}\bigr{)}=-1$, or in other words, $a$ and $b$ denoting the quadratic residues and nonresidues of $p$ that are smaller than $p$: $\sum\left(\frac{n}{p}\right)\frac{1}{n}=\frac{\pi}{p^{2}}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)\bigl{(}\sum a-\sum b\bigr{)}\sqrt{-1}.$ With $p=4\mu+1$ the imaginary part of the sum disappears because then $\bigl{(}\frac{m}{p}\bigr{)}=\bigl{(}\frac{p-m}{p}\bigr{)}$, and we get: $\sum\left(\frac{n}{p}\right)\frac{1}{n}=\frac{1}{p}f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)\log\frac{\prod\sin\frac{b\pi}{p}}{\prod\sin\frac{a\pi}{p}},$ where the multiplication extends over all $a$ or $b$. Notice know that, under the assumption $\omega=-1$, using well-known formulae[1] we get for $f\left(e^{\frac{2\pi}{p}\sqrt{-1}}\right)$ in the first case $\sqrt{p}\sqrt{-1}$, and in the last $\sqrt{p}$, so respectively: $\sum\left(\frac{n}{p}\right)\frac{1}{n}=\frac{\pi}{p\sqrt{p}}(\sum b-\sum a),\qquad\sum\left(\frac{n}{p}\right)\frac{1}{n}=\frac{1}{\sqrt{p}}\log\frac{\prod\sin\frac{b\pi}{p}}{\prod\sin\frac{a\pi}{p}}.$ In the case of $p=4\mu+3$ we see immediately that $\sum\bigl{(}\frac{n}{p}\bigr{)}\frac{1}{n}$ is nonzero since $\sum a+\sum b=\frac{1}{2}p(p-1)$ is odd and thus $\sum a=\sum b$ cannot hold. To prove the same for $p=4\mu+1$ we use the equations which are known from cyclotomy[3] $2\prod\left(x-e^{\frac{2a\pi}{p}\sqrt{-1}}\right)=Y-Z\sqrt{p},\qquad 2\prod\left(x-e^{\frac{2b\pi}{p}\sqrt{-1}}\right)=Y+Z\sqrt{p},$ where $Y$ and $Z$ are polynomials with integer coefficients. If we substitute in these equations and those following from them: $4\frac{x^{p}-1}{x-1}=Y^{2}-pZ^{2}$ $x=1$ and denote then with $g$ and $h$ those integer values taken by $Y$ and $Z$, we get after several easy reductions: $2^{\frac{p+1}{2}}\prod\sin\frac{a\pi}{p}=g-h\sqrt{p},\quad 2^{\frac{p+1}{2}}\prod\sin\frac{b\pi}{p}=g+h\sqrt{p},\quad g^{2}-ph^{2}=4p.$ From the last equation follows that $g$ is divisible by $p$. If we therefore put $g=pk$ and divide the first two [equations] through each other we get: $\frac{\prod\sin\frac{b\pi}{p}}{\prod\sin\frac{a\pi}{p}}=\frac{k\sqrt{p}+h}{k\sqrt{p}-h},\qquad h^{2}-pk^{2}=-4.$ From the second of these equations, $h$ cannot be zero, thus both sides of the first [equation] are different from unity, from which immediately follows, respecting the expression we obtained above, that: $\sum\bigl{(}\frac{n}{p}\bigr{)}\frac{1}{n}$ cannot have the value zero, q. e. d. We can add that the sum $\sum\bigl{(}\frac{n}{p}\bigr{)}\frac{1}{n}$, because as limit of a product of only positive factors, namely as limit of: $\prod\frac{1}{1-\left(\frac{q}{p}\right)\frac{1}{q^{1+\varrho}}},$ for $\varrho$ becoming infinitely small, it never can become negative, so it will be necessarily positive. From this statement two important theorems follow directly that are probably not provable otherwise, of which the one related to the case $p=4\mu+3$ is that for a prime of this form $\sum b>\sum a$ always holds. However, we don’t want to stay with these results of our method since we will have occasion to get back to the subject with another investigation. §. 5. To prove for $L_{m}$, if $m$ is neither $0$ nor $\frac{1}{2}(p-1)$, that its limit, which corresponds to $\varrho$ being infinitely small, is different from zero we take the logarithm of: $\prod\frac{1}{1-\omega^{\gamma}\frac{1}{q^{1+\varrho}}}$ and develop the logarithm of each factor using the formula: $-\log(1-x)=x+\frac{1}{2}x^{2}+\frac{1}{3}x^{3}+\cdots.$ We so find: $\sum\omega^{\gamma}\frac{1}{q^{1+\varrho}}+{\textstyle\frac{1}{2}}\sum\omega^{2\gamma}\frac{1}{(q^{2})^{1+\varrho}}+{\textstyle\frac{1}{3}}\sum\omega^{3\gamma}\frac{1}{(q^{3})^{1+\varrho}}+\cdots=\log L,$ where the summation is with repect to $q$ and $\gamma$ means the index of $q$. If we substitute for $\omega$ its values: $1,\Omega,\Omega^{2},\ldots,\Omega^{p-2},$ add and remember that the sum: $1+\Omega^{h\gamma}+\Omega^{2h\gamma}+\cdots+\Omega^{(p-2)h\gamma}$ always disappears except when $h\gamma$ is divisible by $p-1$ but has in this case the value $p-1$, and that the condition $h\gamma\equiv 0$ (mod. $p-1$) is identical with $q^{h}\equiv 1$ (mod. $p$), then we get: $(p-1)\left(\sum\frac{1}{q^{1+\varrho}}+{\textstyle\frac{1}{2}}\sum\frac{1}{q^{2+2\varrho}}+{\textstyle\frac{1}{3}}\sum\frac{1}{q^{3+3\varrho}}+\cdots\right)=\log(L_{0}L_{1}\ldots L_{p-2}),$ where the first, second,… summation relates to those values of $q$, the first, second,… powers of which are contained in the form $\mu p+1$, respectively. Because the left side is real it follows that the product under the log sign is positive, which is also obvious otherwise, and that we have to take for the logarithm its arithmetical value that has no ambiguity. The series on the left hand remains always positive, and we will show now that the right side to the contrary would be $-\infty$ with vanishing $\varrho$, if we would try to assume the limit for $L_{m}$ to disappear. The right side can be written as: $\log L_{0}+\log L_{\frac{1}{2}(p-1)}+\log L_{1}L_{p-2}+\log L_{2}L_{p-3}+\cdots,$ where $\log L_{0}$ after (5) is equal to the expression: $\log\left(\frac{p-1}{p}\cdot\frac{1}{\varrho}+\varphi(\varrho)\right)$ or: $\log\left(\frac{1}{\varrho}\right)+\log\left(\frac{p-1}{p}+\varrho\varphi(\varrho)\right),$ the second term of which approaches the finite limit: $\log\bigl{(}\frac{p-1}{p}\bigr{)};$ likewise, $\log L_{\frac{1}{2}(p-1)}$ remains finite, since the limit of $L_{\frac{1}{2}(p-1)}$ is nonzero with §. 4. One of the other logarithms $\log L_{m}L_{p-1-m}$, is after §. 3: $\log\bigl{(}\psi^{2}(1+\varrho)+\chi^{2}(1+\varrho)\bigr{)},$ which expression, if $L_{m}$ and thus $L_{p-1-m}$ too would have zero as limit such that at the same time $\psi(1)=0$, $\chi(1)=0$, would transform into: $\log\bigl{(}\varrho^{2}(\psi^{\prime 2}(1+\delta\varrho)+\chi^{\prime 2}(1+\varepsilon\varrho)\bigr{)}$ or: $-2\log\left(\frac{1}{\varrho}\right)+\log\bigl{(}\psi^{\prime 2}(1+\delta\varrho)+\chi^{\prime 2}(1+\varepsilon\varrho)\bigr{)}.$ Combining the term $-2\log\bigl{(}\frac{1}{\varrho}\bigr{)}$ with the first term of $\log L_{0}$, there results: $-\log\bigl{(}\frac{1}{\varrho}\bigr{)}$, which value would become $-\infty$ with infinitely small $\varrho$, and it is clear that this infinitely large negative value cannot be cancelled by e.g.: $\log\bigl{(}\psi^{\prime 2}(1+\delta\varrho)+\chi^{\prime 2}(1+\varepsilon\varrho)\bigr{)}$ because this expression either remains finite or itself becomes $-\infty$, namely when simultaneously $\psi^{\prime}(1)=0$, $\chi^{\prime}(1)=0$. Just as evident is that, would we try to view some other pairs of related $L$ than $L_{m}$ and $L_{p-1-m}$ as mutually cancelling, the contradiction would be even intensified. Therefore it is proved that the limit of $L_{m}$ for $m>0$, corresponding to infinitely small $\varrho$, is finite and different from zero. Also, $L_{0}$ in the same case, becomes $\infty$ from which immediately follows that the series: (7) $\sum\omega^{\gamma}\frac{1}{q^{1+\varrho}}+{\textstyle\frac{1}{2}}\sum\omega^{2\gamma}\frac{1}{q^{2+2\varrho}}+{\textstyle\frac{1}{3}}\sum\omega^{3\gamma}\frac{1}{q^{3+3\varrho}}+\cdots=\log L$ always approaches a finite limit if $\omega$ not $=1$, but becomes infinitely large for $\omega=1$ if we let $\varrho$ become infinitely small. Would we want to have the limit itself, which is not necessary for our task however, its calculation (for $\omega$ not $-1$) using the expression $\log\bigl{(}\psi(1)+\chi(1)\sqrt{-1}\bigr{)}$ would be afflicted with an ambiguity that could be lifted easily with specialisation, i.e., as soon as $p$ and $\omega$ is given numerically. If we equal the series (7) with $u+v\sqrt{-1}$ and therefore: $u+v\sqrt{-1}=\log L=\log\bigl{(}\psi(1+\varrho)+\chi(1+\varrho)\sqrt{-1}\bigr{)},$ we have: $u={\textstyle\frac{1}{2}}\log\bigl{(}\psi^{2}(1+\varrho)+\chi^{2}(1+\varrho)\bigr{)},$ $\cos v=\frac{\psi(1+\varrho)}{\sqrt{\psi^{2}(1+\varrho)+\chi^{2}(1+\varrho)}},\qquad\sin v=\frac{\chi(1+\varrho)}{\sqrt{\psi^{2}(1+\varrho)+\chi^{2}(1+\varrho)}},$ and thus the limit of $u$ is no longer ambigous: ${\tfrac{1}{2}}\log\bigl{(}\psi^{2}(1)+\chi^{2}(1)\bigr{)}.$ To get the same with $v$ we note that the series, however small $\varrho$ may be, is continously variable with respect to this value, which can be easily proved. Therefore also $v$ is a continous function of $\varrho$. Because $\psi(1)=0$, $\chi(1)=0$ cannot hold at the same time, it will be possible to derive from the expressions of $\psi(1+\varrho)$ and $\chi(1+\varrho)$ given above as definite integrals always a positive finite value $R$ of such character that at least one of the functions $\psi(1+\varrho)$, $\chi(1+\varrho)$ retains its sign for each $\varrho$ smaller than $R$. Therefore $\cos v$ or $\sin v$, as soon as $\varrho$ decreases below $R$, will no longer change signs and thus the continously variable arc $v$ will no longer be able to increase or decreases by $\pi$. If we thus determine the finite value of $v$ corresponding to $\varrho=R$, let’s call it $V$, which we can easily find by numerical computation from the series (7) because [the series] for each finite value of $\varrho$ belongs to one of the classes differentiated in §. 1 and thus has a completely defined sum, the limit $v_{0}$ of $v$ is given by the equations $\cos v_{0}=\frac{\psi(1)}{\sqrt{\psi^{2}(1)+\chi^{2}(1)}},\qquad\sin v_{0}=\frac{\chi(1)}{\sqrt{\psi^{2}(1)+\chi^{2}(1)}},$ under the condition that the difference $V-v_{0}$, ignoring signs, has to be smaller than $\pi$. §. 6. We are now able to prove that each arithmetic progression with difference $p$ whose first member is not divisible by $p$ contains infinitely many primes; or, in other words, that there are infinitely many primes of the form $\mu p+m$, where $\mu$ is an arbitrary number and $m$ one of the numbers $1,\,2,\,3,\ldots,\,p-1$. If we multiply the equations contained in (7) that correspond consecutively to the roots: $1,\Omega,\Omega^{2},\ldots,\Omega^{p-2}$ with: $1,\Omega^{-\gamma_{m}},\Omega^{-2\gamma_{m}},\ldots,\Omega^{-(p-2)\gamma_{m}}$ and add we get on the left side: $\displaystyle\sum(1+\Omega^{\gamma-\gamma_{m}}+\Omega^{2(\gamma-\gamma_{m})}+\cdots+\Omega^{(p-2)(\gamma-\gamma_{m})})\frac{1}{q^{1+\varrho}}$ $\displaystyle+$ $\displaystyle{\textstyle\frac{1}{2}}\sum(1+\Omega^{2\gamma-\gamma_{m}}+\Omega^{2(2\gamma-\gamma_{m})}+\cdots+\Omega^{(p-2)(2\gamma-\gamma_{m})})\frac{1}{q^{2+2\varrho}}$ $\displaystyle+$ $\displaystyle{\textstyle\frac{1}{3}}\sum(1+\Omega^{3\gamma-\gamma_{m}}+\Omega^{2(3\gamma-\gamma_{m})}+\cdots+\Omega^{(p-2)(3\gamma-\gamma_{m})})\frac{1}{q^{3+3\varrho}}$ $\displaystyle+$ $\displaystyle\cdots,$ where the summation is over $q$ and $\gamma$ denotes the index of $q$. But now it holds that: $1+\Omega^{h\gamma-\gamma_{m}}+\Omega^{2(h\gamma-\gamma_{m})}+\cdots+\Omega^{(p-2)(h\gamma-\gamma_{m})}=0,$ except when $h\gamma-\gamma_{m}\equiv 0$ (mod. $p-1$), in which case the sum equals $p-1$. This congruence however is identical with $q^{h}\equiv m$ (mod. $p$). We therefore have the equation: $\sum\frac{1}{q^{1+\varrho}}+{\textstyle\frac{1}{2}}\sum\frac{1}{q^{2+2\varrho}}+{\textstyle\frac{1}{3}}\sum\frac{1}{q^{3+3\varrho}}+\cdots\\\ =\tfrac{1}{p-1}(\log L_{0}+\Omega^{-\gamma_{m}}\log L_{1}+\Omega^{-2\gamma_{m}}\log L_{2}+\cdots+\Omega^{-(p-2)\gamma_{m}}\log L_{p-2}),$ where the first summation is over all primes $q$ of form $\mu p+m$, the second over all primes $q$ with squares of that form, the third over all primes $q$ with cubes of that form etc. If we assume now $\varrho$ becoming infinitely small, the right side will become infinitely large through the term $\log L_{0}$. Thus also the left hand has to become infinity. But on this side the sum of all terms, except the first, remains finite because, as is well-known, ${\textstyle\frac{1}{2}}\sum\frac{1}{q^{2}}+{\textstyle\frac{1}{3}}\sum\frac{1}{q^{3}}+\cdots$ is even finite if we substitute for $q$ not certain primes, as we did, but all integers larger than $1$. Thus the series $\sum\frac{1}{q^{1+\varrho}}$ has to grow beyond any positive limit; it has to have infinitely many terms, that means, there are infinitely many primes $q$ of form $\mu p+m$, q. e. d. §. 7. To extend the previously given proof to arithmetic progressions with the difference being any composite integer, several theorems from the theory of residues will be necessary which we want to collect now to be able to refer to them more easily in the following. Justification of these results can be looked up in _Disq. arith. sect. III._ where the subject is treated in depth. I. The existence of primitive roots is not limited to odd primes $p$ but also applies to any power $p^{\pi}$ of it. With $c$ a primitive root for the modulus $p^{\pi}$, the residues of the powers: $c^{0},c^{1},c^{2},\ldots,c^{(p-1)p^{\pi-1}-1}$ with respect to [that modulus] are all mutually different and identical to the series of the numbers smaller than $p^{\pi}$ and coprime to $p^{\pi}$. If we have now an arbitrary number $n$ not divisible by $p$, then the exponent $\gamma_{n}<(p-1)p^{\pi-1}$ which satisfies the congruence $c^{\gamma_{n}}\equiv n\hbox{ (mod.~{}$p^{\pi}$)}$ is completely determined and we shall call it the index of $n$. Regarding such indices the easy to prove theorems hold that the index of a product equals the sum of factor indices minus the largest contained multiple of $(p-1)p^{\pi-1}$, and that $\gamma_{n}$ is even or odd corresponding to $\bigl{(}\frac{n}{p}\bigr{)}=+1$ or $-1$. II. The prime number $2$ shows an essentially different behaviour in the theory of primitive roots than the odd primes, and we note about this prime the following, if we ignore the first power of $2$ which is not important here. 1) For the modulus $2^{2}$ we have the primitive root $-1$. If we denote the index of an arbitrary odd number $n$ with $\alpha_{n}$ such that then: $(-1)^{\alpha_{n}}\equiv n\hbox{ (mod.~{}$4$)},$ we have $\alpha_{n}=0$ oder $\alpha_{n}=1$, according to $n$ having the form $4\mu+1$ or $4\mu+3$, and we get the index of a product if we subtract its largest contained even number from the sum of factor indices. 2) For a modulus of the form $2^{\lambda}$, with $\lambda\geq 3$, there does not exist any primitive root anymore, i.e., there is no number such that the period of its power residues after the divisor $2^{\lambda}$ contains every odd number smaller than $2^{\lambda}$. It is possible only to express one half of these numbers as such residues. If we choose any number of form $8\mu+5$ or especially $5$ as basis then the residues of the powers $5^{0},5^{1},5^{2},\ldots,5^{2^{\lambda-2}-1}$ modulo $2^{\lambda}$ are mutually different and coincide with the numbers of form $4\mu+1$ and being smaller than $2^{\lambda}$. If we therefore have a number $n$ of form $4\mu+1$ then the congruence: $5^{\beta_{n}}\equiv n\hbox{ (mod.~{}$2^{\lambda}$)}$ can always be satisfied by exactly one exponent or index $\beta_{n}$ that should be smaller than $2^{\lambda-2}$. If $n$ is of form $4\mu+3$ then this congruence is impossible. Because however on this condition $-n$ is of form $4\mu+1$ then in general we will denote as the index of an odd number $n$ the completely defined exponent $\beta_{n}$ that is smaller than $2^{\lambda-2}$ and satisfies the congruence: $5^{\beta_{n}}\equiv\pm n\hbox{ (mod.~{}$2^{\lambda}$)}$ where the upper or lower sign is applied corresponding to $n$ being of form $4\mu+1$ or $4\mu+3$. Because of this double sign the residue of $n$ modulo $2^{\lambda}$ is thus no longer completely determined by the index $\beta_{n}$, since the same index matches two residues that complement to $2^{\lambda}$. For the so defined indices obviously the theorems hold that the index of a product equals the sum of factor indices minus the largest multiple of $2^{\lambda-2}$ contained therein, as well as that $\beta_{n}$ is even or odd according to $n$ being of form $8\mu\pm 1$ or $8\mu\pm 5$. To remove the aforementioned ambiguity it is sufficient to study not only the index $\beta_{n}$ related to the modulus $2^{\lambda}$ and base $5$ but also the index $\alpha_{n}$ related to the modulus $4$ and base $-1$. In that case, according to $\alpha_{n}=0$ or $\alpha_{n}=1$, we will have to apply the upper or lower sign in: $5^{\beta_{n}}\equiv\pm n\hbox{ (mod.~{}$2^{\lambda}$)}.$ We could join, if we want, both indices into one formula and write: $(-1)^{\alpha^{n}}5^{\beta_{n}}\equiv n\hbox{ (mod.~{}$2^{\lambda}$)},$ by which congruence the residue of $n$ modulo $2^{\lambda}$ will be completely determined. III. Let now: $k=2^{\lambda}p^{\pi}p^{\prime\pi^{\prime}}\ldots,$ where, as in II. 2, $\lambda\geq 3$ and $p,p^{\prime},\ldots$ denote mutually different odd primes. If we take any number $n$ not divisible by the primes $2,p,p^{\prime},\ldots$ and the indices: $\alpha_{n},\beta_{n},\gamma_{n},\gamma^{\prime}_{n},\ldots,$ corresponding to the moduli: $4,2^{\lambda},p^{\pi},p^{\prime\pi^{\prime}},\ldots$ and their primitive roots: $-1,5,c,c^{\prime},\ldots$ then we have the congruences: $(-1)^{\alpha_{n}}\equiv n\hbox{ (mod.~{}$4$)},\qquad 5^{\beta_{n}}\equiv\pm n\hbox{ (mod.~{}$2^{\lambda}$)},$ $c^{\gamma_{n}}\equiv n\hbox{ (mod.~{}$p^{\pi}$)},\qquad c^{\prime\gamma^{\prime}_{n}}\equiv n\hbox{ (mod.~{}$p^{\pi}$)},\ldots,$ by which the residue of $n$ modulo $k$ is completely determined, which follows at once from well-known theorems if we remember that the double sign in the second of these congruences is fixed by the first. We will call the indices $\alpha_{n},\beta_{n},\gamma_{n},\gamma^{\prime}_{n},\ldots$ or $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ the system of indices for the number. Because the Indices: $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ or: $2,\,2^{\lambda-2},\,(p-1)p^{\pi-1},\,(p^{\prime}-1)p^{\prime\pi^{\prime}-1},\ldots,$ respectively, can have different values it holds that: (8) $2\cdot 2^{\lambda-2}(p-1)p^{\pi-1}(p^{\prime}-1)p^{\prime\pi^{\prime}-1}\cdots=k\left(1-\frac{1}{2}\right)\left(1-\frac{1}{p}\right)\left(1-\frac{1}{p^{\prime}}\right)\cdots\\\ =K$ is the number of possible systems of this kind, which agrees with the well- known theorem that by $K$ is expressed the number of those numbers smaller than $k$ and coprime to $k$. §. 8. When trying to prove the theorem on arithmetic progressions in its full generality we note that we can, without loss of it, assume the difference $k$ of the progressions as divisible by $8$ and thus of the form included with §. 7, III. Is the theorem proven on this condition it will obviously hold as well with the difference odd or only divisible by $2$ or $4$. Let $\theta,\varphi,\omega,\omega^{\prime},\ldots$ any roots of the equations: $\theta^{2}-1=0,\quad\varphi^{2^{\lambda-2}}-1=0,$ (9) $\omega^{(p-1)p^{\pi-1}}-1=0,\quad\omega^{\prime(p^{\prime}-1)p^{\pi^{\prime}-1}}-1=0,\ldots,$ and let $q$ an arbitrary prime not equal $2,p,p^{\prime},\ldots$. If we write the equation: $\frac{1}{1-\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{q^{s}}}\\\ =1+\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{q^{s}}+\theta^{2\alpha}\varphi^{2\beta}\omega^{2\gamma}\omega^{\prime 2\gamma^{\prime}}\cdots\frac{1}{q^{2s}}+\cdots,$ where $s>1$, and the system of indices $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ is with respect to $q$, and if we multiply all equations of this form, which we get if we substitute for $q$ every prime different from $2,p,p^{\prime},\ldots$, with each other then we get, remembering the abovementioned properties of indices and equations (9): (10) $\prod\frac{1}{1-\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{q^{s}}}=\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{n^{s}}=L,$ where the product is over the primes except $2,p,p^{\prime}\ldots$, and the sum is over all positive integers not divisible by the primes $2,p,p^{\prime}\ldots$. The system of indices $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ is on the left side with respect to $q$, on the right side to $n$. The general equation (10), in which the different roots $\theta,\varphi,\omega,\omega^{\prime},\ldots$ can be mutually combined arbitrarily, apparently contains a number $K$ of special equations. To denote the series $L$ corresponding to each of the combinations in a comfortable way we can think of the roots of each of these equations (9) expressed as powers of one of them. Let $\Theta=-1,\Phi,\Omega,\Omega^{\prime},\ldots$ roots suitable for that purpose, then: $\theta=\Theta^{\mathfrak{a}},\qquad\varphi=\Phi^{\mathfrak{b}},\qquad\omega=\Omega^{\mathfrak{c}},\qquad\omega^{\prime}=\Omega^{\prime\mathfrak{c}^{\prime}},\ldots,$ where: $\mathfrak{a}<2,\qquad\mathfrak{b}<2^{\lambda-2},\qquad\mathfrak{c}<(p-1)p^{\pi-1},\qquad\mathfrak{c}^{\prime}<(p^{\prime}-1)p^{\prime\pi^{\prime}-1},\ldots$ and, using this notation, denote the series $L$ with: (11) $L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots}.$ The necessity for the condition $s>1$ in the equation (10) is the same as already developed in §. 1. §. 9. The series denoted with $L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots}$, of which the number equals $K$, can be divided into the following three classes according to the different combinations $\theta,\varphi,\omega,\omega^{\prime},\ldots$ of their roots. The first class contains only one series, namely $L_{0,0,0,0,\ldots}$, that is, the one where: $\theta=1,\quad\varphi=1,\quad\omega=1,\quad\omega^{\prime}=1,\ldots$ holds. The second class shall cover all other series with only real solutions to the equations (9) such that therefore to express those series we have to combine the signs in: $\theta=\pm 1,\quad\varphi=\pm 1,\quad\omega=\pm 1,\quad\omega^{\prime}=\pm 1,\ldots$ in every possible way excepting only the combination corresponding to the first class. The third class finally includes all series $L$ where at least one of the roots $\varphi,\omega,\omega^{\prime},\ldots$ is imaginary, and it is evident that the series of this class come in pairs since the two root combinations: $\theta,\varphi,\omega,\omega^{\prime},\ldots;\quad\frac{1}{\theta},\frac{1}{\varphi},\frac{1}{\omega},\frac{1}{\omega^{\prime}},\ldots$ are mutually different given the just mentioned condition. We will now have to study the behaviour of those series on substitution of $s=1+\varrho$ and letting become $\varrho$ infinitely small. Let us look first at the series constituting the first class, and clearly, we can see it as sum of $K$ partial series of form: $\frac{1}{m^{1+\varrho}}+\frac{1}{(k+m)^{1+\varrho}}+\frac{1}{(2k+m)^{1+\varrho}}+\cdots,$ where $m<k$ and coprime to $k$. Thus the series of this class equals after §. 2: (12) $\frac{K}{k}\cdot\frac{1}{\varrho}+\varphi(\varrho),$ where $\varphi(\varrho)$ remains finite for infinitely small $\varrho$. Regarding the series of second or third class we find, if we order their terms such that the values of $n$ are increasing and with $s>0$, for them the equation: (13) $\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{n^{s}}=\frac{1}{\Gamma(s)}\int_{0}^{1}\frac{\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\ldots x^{n-1}}{1-x^{k}}\log^{s-1}\left(\frac{1}{x}\right)\,dx,$ where the sum on the right hand is over all positive integers $n$ smaller than and coprime to $k$, and $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ stands for the system of indices for $n$. We easily prove that the right side has a finite value, because for it we only need to mention that the polynomial $\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\ldots x^{n-1}$ involves the factor $1-x$, which illustrates immediately that if we put $x=1$ by which the polynomial is transformed to the product: $(1+\theta)(1+\varphi+\cdots+\varphi^{2^{\lambda-2}-1})(1+\omega+\cdots+\omega^{(p-1)p^{\pi-1}-1})\times\\\ \times(1+\omega^{\prime}+\cdots+\omega^{\prime(p^{\prime}-1)p^{\prime\pi^{\prime}-1}-1})\cdots,$ at least one of its factors will disappear, which would be impossible with the root combination corresponding to the first class: $\theta=1,\quad\varphi=1,\quad\omega=1,\quad\omega^{\prime}=1,\ldots.$ We convince ourselves as easily that the right side of the equation (13), as well as its differential quotient with respect to $s$, are continous functions of $s$. It follows at once that, for $\varrho$ getting infinitely small, each series of second or third class approaches a finite limit expressed by: (14) $\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{n}=\int_{0}^{1}\frac{\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\ldots x^{n-1}}{1-x^{k}}\,dx.$ It remains only to prove that this limit is always nonzero. §. 10. Although the limit for an $L$ of the second or third class is easily expressed using logarithms and circular functions, like in §. 4, such an expression would have no use for the desired proof, also not if $L$ belongs to the second class, even though this case elsewise is mostly analogous to that studied in the last half of §. 4. Let us just now assume the mentioned property was proved for each $L$ of the second class. We will now show how the same requirement can be satisfied for an $L$ of the third class. To this end we take logarithms of both sides of the equation (10) and develop; we so get $\sum\theta^{\alpha}\varphi^{\beta}\omega^{\gamma}\omega^{\prime\gamma^{\prime}}\cdots\frac{1}{q^{1+\varrho}}+{\textstyle\frac{1}{2}}\sum\theta^{2\alpha}\varphi^{2\beta}\omega^{2\gamma}\omega^{\prime 2\gamma^{\prime}}\cdots\frac{1}{q^{2+2\varrho}}+\cdots=\log L,$ where the indices $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ are with respect to $q$, as well as the sum. If we express the roots $\theta,\varphi,\omega,\omega^{\prime},\ldots$ in the way mentioned in §. 8 and put: $\theta=\Theta^{\mathfrak{a}},\qquad\varphi=\Phi^{\mathfrak{b}},\qquad\omega=\Omega^{\mathfrak{c}},\qquad\omega^{\prime}=\Omega^{\prime\mathfrak{c}^{\prime}},\ldots,$ then the general term on the left side becomes: $\frac{1}{h}\sum\Theta^{h\alpha\mathfrak{a}}\Phi^{h\beta\mathfrak{b}}\Omega^{h\gamma\mathfrak{c}}\Omega^{\prime h\gamma^{\prime}\mathfrak{c}^{\prime}}\cdots\frac{1}{q^{h+h\varrho}},$ while after (11) we have to write for the right side: $\log L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots}$ Let now $m$ an integer smaller than and coprime to $k$. If we multiply both sides with: $\Theta^{-\alpha_{m}\mathfrak{a}}\Phi^{-\beta_{m}\mathfrak{b}}\Omega^{-\gamma_{m}\mathfrak{c}}\Omega^{\prime-\gamma_{m}^{\prime}\mathfrak{c}^{\prime}}\cdots$ and abbreviate the left side to the general term we get: $\displaystyle\cdots+\frac{1}{h}\sum\Theta^{(h\alpha-\alpha_{m})\mathfrak{a}}\Phi^{(h\beta-\beta_{m})\mathfrak{b}}\Omega^{(h\gamma-\gamma_{m})\mathfrak{c}}\Omega^{\prime(h\gamma^{\prime}-\gamma_{m}^{\prime})\mathfrak{c}^{\prime}}\cdots\frac{1}{q^{h+h\varrho}}+\cdots$ $\displaystyle=$ $\displaystyle\Theta^{-\alpha_{m}\mathfrak{a}}\Phi^{-\beta_{m}\mathfrak{b}}\Omega^{-\gamma_{m}\mathfrak{c}}\Omega^{\prime-\gamma_{m}^{\prime}\mathfrak{c}^{\prime}}\cdots\log L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots}$ Summing now, to include all root combinations, from: $\mathfrak{a}=0,\qquad\mathfrak{b}=0,\qquad\mathfrak{c}=0,\qquad\mathfrak{c}^{\prime}=0,\ldots$ to: $\mathfrak{a}=1,\quad\mathfrak{b}=2^{\lambda-2}-1,\quad\mathfrak{c}=(p-1)p^{\pi-1}-1,\quad\mathfrak{c}^{\prime}=(p^{\prime}-1)p^{\prime\pi^{\prime}-1}-1,\ldots,$ the general term on the left hand becomes: $\frac{1}{h}\sum W\frac{1}{q^{h+h\varrho}},$ where the sum is over all primes $q$ and $W$ means the product of the sums taken over $\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime},\ldots$ or respectively over: $\sum\Theta^{(h\alpha-\alpha_{m})\mathfrak{a}},\qquad\sum\Phi^{(h\beta-\beta_{m})\mathfrak{b}},\qquad\sum\Omega^{(h\gamma-\gamma_{m})\mathfrak{c}},\qquad\sum\Omega^{\prime(h\gamma^{\prime}-\gamma_{m}^{\prime})\mathfrak{c}^{\prime}},\ldots.$ We can now see from §. 7 that the first of these sums is $2$ or $0$, corresponding to if the congruence $h\alpha-\alpha_{m}\equiv 0$ (mod. $2$) or, equally, the congruence $q^{h}\equiv m$ (mod. $4$) holds or not; that the second is $2^{\lambda-2}$ or $0$ corresponding to if the congruence $h\beta-\beta_{m}\equiv 0$ (mod. $2^{\lambda-2}$) or, equally, the congruence $q^{h}=\pm m$ (mod. $2^{\lambda}$) holds or not; that the third is $(p-1)p^{\pi-1}$ or $0$, corresponding to if the congruence $h\gamma-\gamma_{m}\equiv 0$ (mod. $(p-1)p^{\pi-1}$) or, equally, the congruence $q^{h}\equiv m$ (mod. $p^{\pi}$) holds or not, and so on; that therefore $W$ always disappears except when the congruence $q^{h}\equiv m$ holds modulo each of the modules $2^{\lambda},p^{\pi},p^{\prime\pi^{\prime}},\ldots$, that is, when $q^{k}\equiv m$ (mod. $k$) holds, in which case $W=K$. Our equation thus becomes: (15) $\displaystyle\sum\frac{1}{q^{1+\varrho}}+{\textstyle\frac{1}{2}}\sum\frac{1}{q^{2+2\varrho}}+{\textstyle\frac{1}{3}}\sum\frac{1}{q^{3+3\varrho}}+\cdots$ $\displaystyle=$ $\displaystyle\frac{1}{K}\sum\Theta^{-\alpha_{m}\mathfrak{a}}\Phi^{-\beta_{m}\mathfrak{b}}\Omega^{-\gamma_{m}\mathfrak{c}}\Omega^{\prime-\gamma_{m}^{\prime}\mathfrak{c}^{\prime}}\cdots\log L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots},$ where the summation on the left is over all primes $q$ the first, second, third powers of which are contained in the form $\mu k+m$, while the summation on the right is over $\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime},\ldots$ and extends between the given limits. For $m=1$, we get $\alpha_{m}=0$, $\beta_{m}=0$, $\gamma_{m}=0$, $\gamma_{m}^{\prime}=0$,…, and the right side reduces to: $\frac{1}{K}\sum\log L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots}.$ The term of this sum corresponding to the $L$ of the first class, $L_{0,0,0,0,\ldots}$, will, because of the expression (12), contain $\log\bigl{(}\frac{1}{\varrho}\bigr{)}$. Those terms corresponding to different $L$ of the second class will, on condition of the desired proof above, remain finite for infinitely small $\varrho$. Would the limit for an arbitrary $L$ of the third class be zero the study of the continuity, as in §. 5, of the expression (13) for the logarithm of this $L$, together with its $L$, would result in the term: $-2\log\left(\frac{1}{\varrho}\right)$ which, combined with $\log\bigl{(}\frac{1}{\varrho}\bigr{)}$ in $\log L_{0,0,0,0,\ldots}$ would result in $-\log\bigl{(}\frac{1}{\varrho}\bigr{)}$ which would become $-\infty$ for infinitely small $\varrho$, while the left side would consist of only positive terms. Therefore no $K$ of the third class can have the limit zero, and it follows (excepting the missing proof for the series of second class) that: $\log L_{\mathfrak{a},\mathfrak{b},\mathfrak{c},\mathfrak{c}^{\prime}\cdots}$ always approaches a finite limit for infinitely small $\varrho$, except when simultaneously $\mathfrak{a}=0$, $\mathfrak{b}=0$, $\mathfrak{c}=0$, $\mathfrak{c}^{\prime}=0$,…in which case the logarithm gets a value that is infinitely large. Applying this result to the general equation (15) we see at once that its right side becomes infinite for infinitely small $\varrho$, namely by the term $\frac{1}{K}\log L_{0,0,0,0,\ldots}$ which grows beyond all limits, while all other remain finite. Therefore also the left side must exceed any finite limit, from which follows, as in §. 6, that the series $\sum\frac{1}{q^{1+\varrho}}$ has infinitely many terms or, in other words, that the number of primes $q$ of form $k\mu+m$, with $\mu$ an arbitrary integer and $m$ a given number coprime to $k$, is infinite q. e. d. §. 11. Regarding now the demonstration necessary for the completion of the just developed proof, it reduces, according to the expression given in (14) for the limit of an $L$ of second or third class, to showing that for any of the root combinations of form: $\pm 1,\,\pm 1,\,\pm 1,\,\pm 1,\,\ldots,$ with the only exception of: $+1,\,+1,\,+1,\,+1,\,\ldots,$ the sum: (16) $\sum(\pm 1)^{\alpha}(\pm 1)^{\beta}(\pm 1)^{\gamma}(\pm 1)^{\gamma^{\prime}}\cdots\frac{1}{n},$ has a nonzero value, where $\alpha,\beta,\gamma,\gamma^{\prime},\ldots$ means the system of indices for $n$, and where for $n$ are substituted all positive integers not divisible by any of the primes $2,p,p^{\prime},p^{\prime\prime},\ldots$, ordered by size. In the originally presented paper I proved this property using indirect and quite complicated considerations. Later however I convinced myself that the same object can be reached otherwise far shorter. The principles from where we started can be applied to several other problems, between which and the subject here treated one first would guess to be no connection. For example, we could solve the very interesting task to determine the number of different quadratic forms which correspond to an arbitrary positive or negative determinant, and we find that this number (which however is not the end result of this investigation) can be expressed as product of two numbers, the first of which is a very simple function of the determinant that has a finite value for every determinant, while the other factor is expressed by a series that coincides with the above (16). From this then follows immediately that the sum (16) never can be zero since otherwise the number of quadratic forms for the respective determinant would reduce to zero, while this number actually always is $\geq 1$. For this reason I will omit my earlier proof for the said property of the series (16) here, and refer on the subject to the mentioned investigation on the number of quadratic forms222Preliminary notice on this can be found in [2] from which emerges the necessary theorem as corollary.. ## References * [1] Dirichlet, Peter Gustav Lejeune: Über eine neue Anwendung bestimmter Integrale auf die Summation endlicher oder unendlicher Reihen. Abh. Königl. Pr. Wiss. Berlin (1835) 391 ff.; see also Comm. Soc. Gott. Vol. 1. * [2] Dirichlet, Peter Gustav Lejeune: Sur l’usage des séries infinies dans la théorie des nombres, Journ. f. Math. 18 (1838) 259–74. * [3] Gauss, Carl Friedrich: Disquisitiones arithmeticae. Lipsiae, Fleischer, 1801. art. 357. * [4] Legendre, Adrien Marie: Théorie des Nombres. Paris, Didot, $3^{\text{i\\`{e}me}}$ ed. ? 1830, $4^{\text{i\\`{e}me}}$ Partie. §. IX.	arxiv-papers	2008-08-10T14:57:30	2024-09-04T02:48:57.217043	{ "license": "Public Domain", "authors": "Peter Gustav Lejeune Dirichlet", "submitter": "Ralf Stephan", "url": "https://arxiv.org/abs/0808.1408" }
0808.1437	# Neutrino nucleus cross sections M. Sajjad Athar, S. Chauhan, S. K. Singh Department of Physics, Aligarh Muslim University, Aligarh-202 002, India. M. J. Vicente Vacas Departamento de Física Teórica and IFIC, Centro Mixto Universidad de Valencia-CSIC, 46100 Burjassot (Valencia), Spain ###### Abstract We present the results of our calculation which has been performed to study the nuclear effects in the quasielastic, inelastic and deep inelastic scattering of neutrinos(antineutrinos) from nuclear targets. These calculations are done in the local density approximation. We take into account the effect of Pauli blocking, Fermi motion, Coulomb effect, renormalization of weak transition strengths in the nuclear medium in the case of the quasielastic reaction. The inelastic reaction leading to production of pions is calculated in a $\Delta$\- dominance model taking into account the renormalization of $\Delta$ properties in the nuclear medium and the final state interaction effects of the outgoing pions with the residual nucleus. We discuss the nuclear effects in the $F_{3}^{A}(x)$ structure function in the deep inelastic neutrino(antineutrino) reaction using a relativistic framework to describe the nucleon spectral function in the nucleus. ###### pacs: 12.15.-y,13.15.+g,13.60.Rj,23.40.Bw,25.30.Pt It is now well established that neutrinos oscillate. The next target of the experimentalist is to determine precisely the various parameters of neutrino mass matrix given by Pontecarvo-Maki-Nakagawa-Sakata (PMNS), absolute masses of different flavors of neutrinos, pattern of the known neutrino mass differences, CP violation in neutrino sector, etc. To address some of these problems several experiments like CNGS, MINOS and SciBooNE are taking data. The eperimental analyses of neutrino oscillation data are going on, for example, at K2K and MiniBooNE, and several experiments are planned to be done in future like T2K and NO$\nu$A. Besides the accelerator experiments, experiments with neutrinos from $\nu$\- factories, $\beta$-beams, etc., are also planned, as well as some experiments with natural $\nu$-sources like solar neutrinos, atmospheric neutrinos, or (anti)neutrinos from nuclear reactors are also planned. These experiments use various nuclear targets like ${}^{12}C$ by MiniBooNE, SciBooNE, MINER$\nu$A and NO$\nu$A, ${}^{16}O$ by SuperKamiokande, T2K, UNO, Hyper-K, K2K and MEMPHYS, ${}^{40}Ar$ by ICARUS and NO$\nu$A, ${}^{56}Fe$ by MINOS, INO, MINER$\nu$A, and ${}^{208}Pb$ by CNGS, MINER$\nu$A and OPERA collaborations. Most of these experiments are being done in the neutrino energy region of $E_{\nu}<2GeV$, for example at MiniBooNE, the average energy($<E_{\nu}>$) is 750 MeV while at K2K it is 1.3 GeV. At the energies of a few GeV the contribution to the cross section comes from the quasielastic, inelastic and deep inelastic processes and the analysis of the data of neutrino experiments is based on Monte Carlo generator of events like NUANCE, NEUGEN, NEUT, etc. In these Monte Carlo generators, neutrino cross sections are used which are based on the model of Llewellyn Smith Smith and Smith and Moniz Moniz for the quasielastic reactions, Rein and Sehgal Rein for the inelastic reactions and GRV98Grv98 along with the modifications suggested by Bodek et al. Bodek for the deep inelastic reactions. The importance of a better knowledge of neutrino-nucleus cross section to be used in the Monte Carlo generators has been realized and discussed in a series of neutrino conferences like NuInt, NuFact, NOW, etc. There are, now, various theoretical calculations for the quasielastic process which make use of nuclear models like shell model with pairing correlations, random phase approximation, relativistic mean field approximation, etc. In the case of inelastic neutrino-nuclear reactions pion production processes have been studied. These are generally studied in a $\Delta$-dominance model in which pions are dominantly produced through the excitation of $\Delta$ and its subsequent decay leading to pions. Some of the calculations have also been done by taking background as well as higher resonance terms. In the case of deep inelastic scattering of neutrinos(antineutrinos) from nuclear targets, there are very few calculations where the dynamical origin of the nuclear medium effects have been studied. In some phenomenological analyses, nuclear medium effects have been described in terms of few parameters which are determined from fitting the experimental data of charged leptons and neutrino(antineutrino) deep inelastic scattering from various nuclear targets. Here we discuss the nuclear effects in the quasielastic reaction, inelastic one $\pi$ production in the $\Delta$ dominance model and the nuclear effects on the $F_{3}^{A}(x,Q^{2})$ structure function in the deep inelastic reaction. These calculations are done in the local density approximation. For the quasielastic process this model takes into account the effect of Pauli blocking, Fermi motion, Coulomb effect, renormalization of weak transition strengths in the nuclear medium. The inelastic reaction leading to production of pions is calculated in a $\Delta$\- dominance model taking into account the renormalization of $\Delta$ properties in the nuclear medium and the final state interaction effects on the outgoing pions. We discuss the nuclear effects in the $F_{3}^{A}(x)$ structure function in the deep inelastic neutrino(antineutrino) reaction using a relativistic framework to describe the nucleon spectral function in the nucleus. The details of these calculations may be found in the Refs. Singh1 ; Athar2 for the quasielastic process, Refs. Singh2 ; Athar3 for the inelastic process and Ref. Athar5 for the deep inelastic process. Similar calculations for nuclear effects in the quasielastic and inelastic processes have also been recently done by many other groups Nieves -Amaro . In the following we describe, in brief, the formalism for calculating the nuclear effects in quasielastic, inelastic and deep inelastic processes in Sec.1, and present the numerical results in Sec.2 with concluding remarks given in Sec.3. ## I NEUTRINO NUCLEUS REACTIONS ### I.1 QUASIELASTIC REACTION The basic reaction for the quasielastic process is a neutrino interacting with a neutron inside the nucleus which is given by $\nu_{\mu}(k)+n(p)\rightarrow\mu^{-}(k^{\prime})+p(p^{\prime})$ (1) The cross section for quasi-elastic charged lepton production is calculated in the local density approximation by taking into account the Fermi motion and the Pauli blocking effects through the imaginary part of the Lindhard function for the particle hole excitations in the nuclear medium. The renormalization of the weak transition strengths are calculated in the random phase approximation(RPA) through the interaction of the p-h excitations as they propagate in the nuclear medium using a nucleon-nucleon potential described by pion and rho exchanges. The effect of the Coulomb distortion of muon in the field of final nucleus is taken into account using a local version of the modified effective momentum approximation. The total cross section $\sigma(E_{\nu})$ for the charged current neutrino induced reaction on a nucleon inside the nucleus in a local Fermi gas model is written as Athar2 : $\displaystyle\sigma(E_{\nu})$ $\displaystyle=$ $\displaystyle-\frac{2{G_{F}}^{2}\cos^{2}{\theta_{c}}}{\pi}\int^{r_{max}}_{r_{min}}r^{2}dr\int^{p_{\mu}^{max}}_{p_{\mu}^{min}}{p_{\mu}}^{2}dp_{\mu}\int_{-1}^{1}d(cos\theta)\frac{1}{E_{\nu_{\mu}}E_{\mu}}L_{\mu\nu}J^{\mu\nu}Im{U_{N}(q_{0},{\bf q})}.$ (2) where $L_{\mu\nu}=\sum L_{\mu}{L_{\nu}}^{\dagger}$ and ${J^{\mu\nu}}={\bar{\sum}}\sum J^{\mu}{J^{\nu}}^{\dagger}$. The leptonic current $L_{\mu}$ and the hadronic current $J^{\mu}$ are given by $L_{\mu}=\bar{u}(k^{\prime})\gamma_{\mu}(1-\gamma_{5})u(k)$ (3) $J^{\mu}=\bar{u}(p^{\prime})[F_{1}(q^{2})\gamma^{\mu}+F_{2}(q^{2})i{\sigma^{\mu\nu}}{\frac{q_{\nu}}{2M}}+F_{A}(q^{2})\gamma^{\mu}\gamma_{5}+F_{P}(q^{2})q^{\mu}\gamma_{5}]u(p).$ (4) where $q(=k-k^{\prime})$ is the four momentum transfer, M is the mass of the nucleon, $G_{F}(=1.16637\times 10^{-5}GeV^{-2})$ is the Fermi coupling constant and $\theta$ is the lepton angle. $U_{N}$ is the Lindhard function for the particle hole excitation Singh1 . The form factors $F_{1}$, $F_{2}$, $F_{A}$ and $F_{P}$ are isovector electroweak form factors and for our numerical calculations we have used the parameterisation of Bradford et al. bradford with axial dipole mass ${M}_{A}$=1.05GeV and vector dipole mass ${M}_{V}$=0.84GeV. Inside the nucleus, the Q-value of the reaction and Coulomb distortion of outgoing lepton are taken into account by modifying the imaginary part of the Lindhard function $Im{U_{N}(q_{0},{\bf q})}$ by $Im{U_{N}(q_{0}-V_{c}(r)-Q,{\bf q})}$. Furthermore, the renormalization of weak transition strength in the nuclear medium in a random phase approximation(RPA) is taken into account by considering the propagation of particle hole(ph) as well as delta-hole($\Delta h$) excitations. These considerations lead to modified terms involving the bilinear terms in the weak coupling constant in the hadronic tensor $J^{\mu\nu}_{RPA}$ for which expressions are given in Ref. Athar2 . ### I.2 INELASTIC RESONANCE PRODUCTION OF PIONS The basic reaction for the inelastic one pion production in nuclei, for a neutrino interacting with a nucleon inside a nuclear target is given by $\nu_{\mu}(k)+N(p)\rightarrow\mu^{-}(k^{\prime})+N^{\prime}(p^{\prime})+\pi^{+}(k_{\pi})~{}~{}~{}N,N^{\prime}=p/n$ (5) The cross sections for pion production is calculated using the $\Delta$ dominance model. In this model, the weak hadronic currents interacting with the nucleons in the nuclear medium excite a $\Delta$ resonance which decays into pions and nucleons. The pions interact with the nucleus inside the nuclear medium before coming out. The final state interaction of pions leading to elastic, charge exchange scattering and the absorption of pions lead to reduction of pion yield. The nuclear medium effects on $\Delta$ properties lead to modification in its mass and width which have been discussed earlier by Oset et al. Oset to explain the pion and electron induced pion production processes from nuclei. In the local density approximation the expression for the total cross section for the charged current one pion production is given by $\displaystyle\sigma$ $\displaystyle=$ $\displaystyle\frac{1}{(4\pi)^{5}}\int_{r_{min}}^{r_{max}}(\rho_{p}(r)+\frac{1}{9}\rho_{n}(r))d\vec{r}\int_{Q^{2}_{min}}^{Q^{2}_{max}}dQ^{2}\int_{0}^{\infty}dk^{\prime}\int_{-1}^{+1}d(cos\theta_{\pi})\int_{0}^{2\pi}d\phi_{\pi}\times$ (6) $\displaystyle\frac{\pi\|\vec{k}^{\prime}\|\|\vec{k}_{\pi}\|}{ME_{\nu}^{2}E_{l}}\frac{1}{E_{p}^{\prime}+E_{\pi}\left(1-\frac{\|\vec{q}\|}{\|\vec{k}_{\pi}\|}cos(\theta_{\pi})\right)}\bar{\sum}\sum\|\mathcal{M}_{fi}\|^{2}$ where the proton density $\rho_{p}(r)=\frac{Z}{A}\rho(r)$ and the neutron density $\rho_{n}(r)=\frac{A-Z}{A}\rho(r)$ with $\rho(r)$ as the nuclear density taken as 3-parameter Fermi Density taken from Ref.Vries . The transition matrix element $\mathcal{M}_{fi}$ is given by $\mathcal{M}_{fi}=\sqrt{3}\frac{G_{F}a}{\sqrt{2}}\frac{f_{\pi N\Delta}}{m_{\pi}}\bar{\Psi}({\bf P})k^{\sigma}_{\pi}{\mathcal{P}}_{\sigma\lambda}\mathcal{O}^{\lambda\alpha}L_{\alpha}u({\bf p})$ (7) where $L^{\alpha}$ is the leptonic current defined by Eq.(3), $a=cos\theta_{c}$ and $\mathcal{O}^{\beta\alpha}={\mathcal{O}}_{V}^{\beta\alpha}+{\mathcal{O}}_{A}^{\beta\alpha}$ for the charged current induced $\pi^{\pm}$ production process while for the neutral current induced $\pi^{0}$ production process a=1 and $\mathcal{O}^{\beta\alpha}=(1-2sin^{2}\theta_{W}){\mathcal{O}}_{V}^{\beta\alpha}+{\mathcal{O}}_{A}^{\beta\alpha}$. ${\mathcal{O}}_{V}^{\beta\alpha}$ and ${\mathcal{O}}_{A}^{\beta\alpha}$ are the vector and axial vector N-$\Delta$ transition operators given by ${\mathcal{O}}_{V}^{\beta\alpha}=\left(\frac{C_{3}^{V}(q^{2})}{M}(g^{\alpha\beta}\not q-q^{\beta}\gamma^{\alpha})+\frac{C_{4}^{V}(q^{2})}{M^{2}}(g^{\alpha\beta}q\cdot P-q^{\beta}P^{\alpha})+\frac{C_{5}^{V}(q^{2})}{M^{2}}(g^{\alpha\beta}q\cdot p-q^{\beta}p^{\alpha})\right)\gamma_{5}$ (8) and ${\mathcal{O}}_{A}^{\beta\alpha}=\frac{C_{4}^{A}(q^{2})}{M^{2}}(g^{\alpha\beta}{\not q}-q^{\beta}\gamma^{\alpha})+C_{5}^{A}(q^{2})g^{\alpha\beta}+\frac{C_{6}^{A}(q^{2})}{M^{2}}q^{\beta}q^{\alpha}$ (9) where $C_{i}^{V}(q^{2})$ and $C_{i}^{A}(q^{2})$ are the vector and axial vector transition form factors and for our numerical calculations these have been taken from the work of Lalakulich et al. Lalakulich . $\theta_{W}$ is the weak mixing angle. ${\mathcal{P}}^{\sigma\lambda}$ is the $\Delta$ propagator in momentum space given by ${\mathcal{P}}^{\sigma\lambda}=\frac{{\it P}^{\sigma\lambda}}{P^{2}-M_{\Delta}^{2}+iM_{\Delta}\Gamma}$ (10) where ${\it P}^{\sigma\lambda}$ is the spin-3/2 projection operator given by ${\it P}^{\sigma\lambda}=\sum_{spins}\psi^{\sigma}\bar{\psi}^{\lambda}=(\not P+M_{\Delta})\left(g^{\sigma\lambda}-\frac{2}{3}\frac{P^{\sigma}P^{\lambda}}{M_{\Delta}^{2}}+\frac{1}{3}\frac{P^{\sigma}\gamma^{\lambda}-P^{\sigma}\gamma^{\lambda}}{M_{\Delta}}-\frac{1}{3}\gamma^{\sigma}\gamma^{\lambda}\right)$ (11) and the delta decay width $\Gamma$ is taken to be an energy dependent P-wave decay width taken as Oset : $\Gamma(W)=\frac{1}{6\pi}\left(\frac{f_{\pi N\Delta}}{m_{\pi}}\right)^{2}\frac{M}{W}\|{\bf q}_{cm}\|^{3}$ (12) $\|q_{cm}\|$ is the pion momentum in the rest frame of the resonance and W is the center of mass energy. Inside the nuclear medium the mass and width of delta are modified which in the present calculation are taken into account by using a modified mass $M_{\Delta}\rightarrow M_{\Delta}+{Re}\Sigma_{\Delta}$ and modified width $\Gamma_{\Delta}\rightarrow\tilde{\Gamma}_{\Delta}-2{Im}\Sigma_{\Delta}$ from the model developed by Oset et al. Oset , where $\tilde{\Gamma}_{\Delta}$ is reduced width of $\Delta$ due to Pauli blocking of nucleons in the $\Delta\rightarrow{N}\pi$ decay and $\Sigma_{\Delta}$ is the self energy of $\Delta$ calculated in nuclear many body theory using local density approximation. The expressions of ${Re}\Sigma_{\Delta}$ and ${Im}\Sigma_{\Delta}$ are taken from the Ref. Oset . The pions produced in this process are scattered and absorbed in the nuclear medium. This is treated in a Monte Carlo simulation which has been taken from the Ref. Vicente . ### I.3 DEEP INELASTIC REACTION The basic process for a neutrino interacting with a nucleon inside the nucleus is $\nu_{\mu}(k)+N(p)\rightarrow\mu^{-}(k^{\prime})+X(p^{\prime}).$ (13) where X is the jet of partons. The differential scattering cross section for the deep inelastic scattering of (anti)neutrinos from unpolarized nucleons in the limit of lepton mass $m_{l}\rightarrow 0$, is described in terms of three structure functions, $F^{\nu}_{1}$($x$,$Q^{2}$), $F^{\nu}_{2}$($x$,$Q^{2}$) and $F^{\nu}_{3}$($x$,$Q^{2}$), where $x=\frac{Q^{2}}{2M\nu}=-\frac{q^{2}}{2M\nu}$ is the Bjorken variable, $\nu$ and $q$ being the energy and momentum transfer of leptons. In the asymptotic region of Bjorken scaling i.e. $Q^{2}\rightarrow\infty$, $\nu\rightarrow\infty$, $x$ finite, all the structure functions depend only on the Bjorken variable $x$. In this scaling limit, $F^{\nu}_{1}(x)$ and $F^{\nu}_{2}(x)$ are related by the Callan-Gross relation Callan leading to only two independent structure functions $F^{\nu}_{2}$($x$) and $F^{\nu}_{3}$($x$) which are determined from the experimental data on deep inelastic scattering of (anti)neutrinos in the asymptotic region. We have studied nuclear medium effects on the nucleon structure function $F^{A}_{3}$($x$,$Q^{2}$) in iron using spectral function to describe the momentum distribution of nucleons in the nucleus. The spectral function has been calculated using the Lehmann’s representation for the relativistic nucleon propagator and nuclear many body theory is used to calculate it for an interacting Fermi sea in nuclear matter. A local density approximation is then applied to translate these results to finite nuclei Marco . Here we consider the modifications of nucleonic contributions to $F^{A}_{3}$($x$,$Q^{2}$) arising due to binding energy, off mass shell and Fermi motion of the nucleon in the nuclear medium which dominate in the region of $x\geq$0.3. In this model, in the region of 0.3$>x>$0.1, corresponding to the anti-shadowing region, the nuclear medium modification effects on $F^{A}_{3}$($x$,$Q^{2}$) are expected to be small due to vanishing of the pion contribution and we have not considered the shadowing region of 0.0$<x<$0.1. Therefore, our results should be able to describe the dominant contribution of nuclear medium effects to $F^{A}_{3}$($x$,$Q^{2}$) in the range of 0.1$<x<$1. The average structure function $F_{3}^{N}(x)$ on isoscalar nucleon target defined as $\displaystyle F_{3}^{N}(x)=\frac{1}{2}\left(F^{\nu N}_{3}+F^{\bar{\nu}N}_{3}\right)$ is given by $F_{3}^{N}(x)=[u_{v}(x)+d_{v}(x)+s(x)-{\bar{s}}(x)+c(x)-{\bar{c}}(x)],$ where $u_{v}(x)$ and $d_{v}(x)$ are the valence quark parton distributions. For an isoscalar target and a symmetric sea, $F_{3}^{N}(x)$ structure function is given in terms of valence quarks $u_{v}$ and $d_{v}$ which satisfy the Gross- Llewellyn Smith sum rule GLS : $\int_{0}^{1}F^{N}_{3}(x)dx=3.$ (14) In the local density approximation the reaction given by Eq.(13) takes place at a point ${\bf r}$, lying inside the nucleus in a volume element $d^{3}r$ with local density $\rho_{p}({\bf r})$ and $\rho_{n}({\bf r})$ corresponding to the proton and neutron. The expression for $F^{A}_{3}(x,Q^{2})$ in the nuclear medium is given by Athar5 : $\displaystyle F^{A}_{3}(x,Q^{2})$ $\displaystyle=$ $\displaystyle 4\int d^{3}r\;\int\frac{d^{3}p}{(2\pi)^{3}}\int^{\mu}_{-\infty}\;dp^{0}S_{h}(p^{0},{\bf p},\rho(r))F(p,Q^{2})F^{N}_{3}(x_{N},Q^{2}),$ (15) where $x_{N}$ is the Bjorken variable expressed in terms of the nucleon variables, $(p^{0},{\bf p})$, in the nucleus i.e. $x_{N}=\frac{Q^{2}}{2p.q}$, $F(p,Q^{2})=\frac{M}{E({\bf p})}\left(\frac{p_{0}\gamma- p_{z}}{(p_{0}-p_{z}\gamma)\gamma}\right);~{}\gamma=\left(1+\frac{4M^{2}x^{2}}{Q^{2}}\right)^{1/2}~{}~{}~{}and$ $S_{h}(\omega,{\bf p})$ is the hole spectral function, the expression for which is taken from Ref. Fernandez . ## II RESULTS AND DISCUSSION ### II.1 QUASIELASTIC REACTION In Fig.1, we present the ratio R of the charged current quasielastic lepton production cross section to the cross section on free nucleon defined as $R=\frac{1}{N}\frac{\sigma(^{12}C)}{\sigma(free)}$ as a function of neutrino(Fig.1a) and antineutrino(Fig.1b) energies when $\sigma(^{12}C)$ is calculated using Eq.(2). We find that with the incorporation of nuclear medium effects without the RPA correlations the reduction in the cross section is around $45\%$ at $E_{\nu}$=0.2GeV, $16\%$ at $E_{\nu}$=0.4GeV, $10\%$ at $E_{\nu}\approx$ 1.0GeV and $7\%$ at $E_{\nu}$=2GeV from the cross sections calculated for the free case. However, when we also incorporate the RPA effects, the total reduction in the cross section is around $40\%$ at $E_{\nu}$=0.4GeV, $20\%$ at $E_{\nu}$=1.0GeV, $18\%$ at $E_{\nu}$=2GeV from the cross sections calculated for the free case. In the case of antineutrinos these reductions are larger as shown in Fig.1(b). We have compared our results with the results obtained in the Fermi gas model which has been used in the NUANCE Monte Carlo generator nuance by the MiniBooNE collaboration boone1 . We find that the present results in the local Fermi gas model are similar to the results used in the NUANCE generator, but when RPA effects are included the cross sections are reduced. Figure 1: $R=\frac{1}{N}\frac{\sigma(^{12}C)}{\sigma(free)}$ vs Neutrino (Antineutrino) Energy. Figure 2: $\frac{d\sigma}{dp_{\mu}}$ vs $p_{\mu}$ for the $\nu_{\mu}$(${\bar{\nu}}_{\mu}$) induced reactions on ${}^{12}C$ target at $E_{\nu}=1GeV$. In Fig.2, we have shown the results for the lepton momentum distribution $\frac{d\sigma}{dp_{\mu}}$ for the $\nu_{\mu}$ and $\bar{\nu}_{\mu}$ induced charged current quasielastic processes. We find that when nuclear medium effects are taken into account there is a reduction as well as shift in the peak region towards the lower value of lepton momentum. This reduction in $\frac{d\sigma}{dp_{\mu}}$ when calculated in the local Fermi gas model without the RPA correlation effects as compared to the cross section calculated without the nuclear medium effects is around $10\%$ in the peak region of lepton momentum, which further reduces by around $30\%$ when RPA effects are also taken into account. In the case of antineutrino the reduction in $\frac{d\sigma}{dp_{\mu}}$ in the local Fermi gas model is around $30\%$ which further reduces by $30\%$ when RPA effects are also taken into account. ### II.2 INELASTIC REACTION In Fig.3, we present the results for $Q^{2}$-distribution $\frac{d\sigma}{dQ^{2}}$ and momentum distribution $\frac{d\sigma}{dp_{\pi}}$ for the charged current $\nu_{\mu}$(${\bar{\nu}}_{\mu}$) induced one $\pi^{+}$($\pi^{-}$) production cross section. These results are presented for the differential scattering cross section calculated with and without the nuclear medium effects and with nuclear medium effects including the pion absorption effects. For the $Q^{2}$\- distribution shown in Fig.3a, we find that the reduction in the cross section as compared to the cross section calculated without the nuclear medium effects is around $35\%$ in the peak region. When pion absorpion effects are also taken into account there is a further reduction of around $15\%$. The results for the antineutrino induced one $\pi^{-}$ production cross section are qualititatively similar in nature but quantitatively we find that the peak shifts towards a slightly lower value of $Q^{2}$. In Fig.3b, the results for the pion momentum distribution have been shown. We find that in the peak region the reduction in the cross section is around $40\%$ when nuclear medium effects are taken into account, which further reduces by about $15\%$ when pion absorption effects are also taken into account. Figure 3: $\frac{d\sigma}{dQ^{2}}$ and $\frac{d\sigma}{dp_{\pi}}$ for the $\nu_{\mu}$(${\bar{\nu}}_{\mu}$) induced charged current one $\pi^{+}(\pi^{-})$ process on ${}^{12}C$ target at $E_{\nu}=1GeV$. Figure 4: $\frac{d\sigma}{dp_{\pi}}$ and $\frac{d\sigma}{dcos_{\theta q}}$ for the neutal current neutrino(antineutrino) induced $\pi^{0}$ production on ${}^{12}C$ target at $E_{\nu}=1GeV$. Figure 5: $\sigma$ for $\nu_{\mu}$ (${\bar{\nu}}_{\mu}$) induced charged current one $\pi^{+}(\pi^{-})$ production on ${}^{12}C$ target. Figure 6: (a)Results for the ratio $R=\frac{F_{3}^{A}(x)}{AF_{3}^{N}(x)}$ at $Q^{2}=5GeV^{2}$ by different authors. Solid line: Result of the present calculation using MRST2004 NNLO parton distribution function; short dashed line: NuTeV Collaboration Tzanov , dashed-double dotted line is the result of Kulagin Kulagin , double dashed- dotted line: Hirai et al. Hirai and the results of Kulagin and Petti Petti1 are shown by dashed-dotted line.(b)$\Delta GLS=\frac{1}{3}(3-\int_{0}^{1}F_{3}^{A}(x,Q^{2})dx)$ vs $Q^{2}$. In Fig.4, we present the results for the neutral current $\nu$($\bar{\nu}$) induced one $\pi^{0}$ production cross section. These results are presented for the pion momentum $\frac{d\sigma}{dp_{\pi}}$ and angular distributions $\frac{d\sigma}{dcos\theta_{\pi}}$, with and without the nuclear medium effects and with nuclear medium and pion absorption effects. For the pion momentum distribution shown in Fig.4a, we find that in the peak region the reduction in the cross section is around $40\%$ when nuclear medium effects are taken into account, which further reduces by about $15\%$ when pion absorption effects are also taken into account. The results with antineutrinos are similar in nature, except that in the case of $\bar{\nu}$, the angular distribution are more forward peaked than in the case of $\nu$. In Fig.5, we present the results for the total scattering cross section $\sigma$ for charged current $\nu_{\mu}$(${\bar{\nu}}_{\mu}$) induced one $\pi^{+}$($\pi^{-}$) production cross section. These results have been presented for the cross sections calculated without(with) the nuclear medium effects and also when pion absorption effect is included along with the nuclear medium effects. We find that with the inclusion of nuclear medium effects the reduction in the cross section from the cross section calculated without the nuclear medium effects for neutrino energies between 1-2 GeV is 30-35$\%$ which further reduces by 15$\%$ when pion absorption effects are also taken into account. The results with antineutrinos are similar in nature. ### II.3 DEEP INELASTIC REACTION In Fig.6(a), we compare our results for R($x$,$Q^{2}$) at $Q^{2}=5GeV^{2}$, where $R=\frac{F^{A}_{3}(x,Q^{2})}{AF^{N}_{3}(x,Q^{2})}$, with the results of Tzanov et al. Tzanov , Kulagin and Petti Petti , Kulagin Kulagin and Hirai et al. Hirai . While the work of Kulagin Kulagin and Kulagin and Petti Petti ; Petti1 use a nuclear model to calculate the nuclear effects which shows a $Q^{2}$ dependence, the work of Tzanov et al. Tzanov and Hirai et al. Hirai are phenomenological analyses, which assume the nuclear effects to be independent of $Q^{2}$. We find a suppression in $F^{A}_{3}(x,Q^{2})$ for $x<$0.7 and an enhancement thereafter, which are respectively smaller than the results of Kulagin Kulagin , but are larger than the recent results of Kulagin and Petti Petti1 . It should be noted that these latter results Petti1 give suppression in the region of 0.4$<x<$0.8 and enhancement for $x>$0.8, which are smaller than the present results and the results obtained earlier in Ref. Kulagin . When compared with the results of Tzanov et al. Tzanov and Hirai et al. Hirai , we find a smaller suppression in the region 0.5$<x<$0.7. In the region 0.7$<x<$0.8, we find an enhancement while they obtain a suppression. In Fig.6(b), we show the $Q^{2}$ dependence of the nuclear effects of the GLS integral, where we plot $\Delta$GLS=$\frac{1}{3}(3-\int_{0}^{1}F^{A}_{3}(x,Q^{2})dx)$ as a function of $Q^{2}$. The experimental results from CCFR collaborations Kim , CHARM collaborations Bergsma and IHEP-JINR collaborations Barabash are also shown. The $Q^{2}$ behaviour of $\Delta$GLS has been found to be in reasonable agreement with the present available experimental results. In this figure, we have also shown the theoretical results obtained by Qiu and Vitev Qiu . Our results are in agreement with the results of Qiu and Vitev Qiu for $Q^{2}>5GeV^{2}$ where theoretically the suppression is found to be larger than the experimental results. For $Q^{2}<5GeV^{2}$, we find a larger suppression compared to the central value of the experimental result and both theoretical values are within the experimental errors. ## III CONCLUSIONS We will like to conclude that : (i) In the case of charged current quasielastic lepton production, the role of nuclear medium effects like Pauli blocking, Fermi motion is to reduce the cross section. When nuclear correlations in the nuclear medium are taken into account in a Random Phase Aprroximation (RPA) there is further reduction in the cross section. The total reduction in the cross section is about $20\%$ in case of $\nu$ and slightly larger for $\bar{\nu}$ around $E_{\nu}$ = 1GeV. (ii) In the case of charged current one pion production, the nuclear medium and pion absorption effects lead to a reduction in the cross section about 45$\%$ at $E_{\nu}$=1 GeV and give appreciable distortion in the energy and angular distribution of pions and leptons. (iii) In the case of deep inelastic scattering, the nuclear effects decrease the value of the structure function $F^{A}_{3}(x,Q^{2})$ in the iron nucleus for $x\leq x_{min}$=0.7 and increase it at higher $x>x_{min}$. In general nuclear medium effects decrease the value of GLS integral for all $Q^{2}$. ACKNOWLEDGMENTS The work presented here is mainly supported by Government of India through the grant DST-SP/S2K-07/2000 and the academic exchange program between Aligarh Muslim University, Aligarh and University of Valencia, Spain. S. C. would like to thank the Jawaharlal Nehru Memorial Fund for the Doctoral Fellowship. ## References * (1) C. H. Llewellyn Smith, Phys. Rep. 3, 261 (1972). * (2) R. A. Smith and E. J. Moniz, Nucl. Phys. B 43, 605 (1972). * (3) D. Rein and L. M. Sehgal, Ann. Phys. 133, 79 (1981). * (4) M. Gluck, E. Reya and A. Vogt, Eur. Phys. J C5, 461 (1998). * (5) A. Bodek, I. Park and U. Yang, Nucl. Phys. Proc. Suppl. 139, 113 (2005). * (6) S. K. Singh and E. Oset, Phys. Rev. C 48, 1246 (1993); Nucl. Phys. A 542, 587 (1992). * (7) M. Sajjad Athar, S. Ahmad and S. K. Singh, Nucl. Phys. A 764, 551 (2006). * (8) S. K. Singh, M. J. Vicente Vacas and E. Oset, Phys. Lett. B 416, 23 (1998). * (9) M. Sajjad Athar, S. Ahmad and S. K. Singh, Phys. Rev. D 75, 093003 (2007); Phys. Rev. D 74, 073008 (2006). * (10) M. Sajjad Athar, S. K. Singh and M. J. Vicente Vacas, arXiv:nucl-th/0711.4443. * (11) J. Nieves, J. E. Amaro and M. Valverde, Phys. Rev. C 70, 055503 (2004); J. Nieves, M. Valverde and M. J. Vicente Vacas, Phys. Rev. C 73, 025504 (2006). * (12) T. Leitner, L. Alvarez-Ruso and U. Mosel, Phys. Rev. C 73, 065502 (2006). * (13) O. Benhar and D. Meloni, Nucl. Phys. A 789, 379 (2007). * (14) O. Benhar, N. Farina, H. Nakamura, M. Sakuda and R. Seki, Phys. Rev. D 72, 053005 (2005). * (15) C. Praet, O. Lalakulich, N. Jachowicz and J. Ryckebusch, arXiv:0804.2750. * (16) A. Meucci, C. Giusti and F. D. Pacati, Nucl. Phys. A 739, 277 (2004). * (17) J. E. Amaro et al., Phys. Rev. C 71, 015501 (2005). * (18) R. Bradford et al., Nucl. Phys. Proc.Suppl. 159, 127 (2006). * (19) E. Oset and L. L. Salcedo, Nucl. Phys. A 468, 631 (1987); C. Garcia Recio et al., Nucl. Phys. A 526, 685 (1991). * (20) O. Lalakulich, E. A. Paschos and G. Piranishvili, Phys. Rev. D 74, 014009 (2006). * (21) C. W. de Jager, H. de Vries and C. de Vries, At. Data Nucl. Data Tables 14, 479 (1974). * (22) M. J. Vicente Vacas, M. Kh. Khankasaev and S. G. Mashnik, arXiv:nucl-th/9412023. * (23) C. G. Callan and D. J. Gross, Phys. Rev. Lett. 22, 156 (1969). * (24) E. Marco, E. Oset and P. Fernandez de Cordoba, Nucl. Phys. A 611, 484 (1996). * (25) D.J. Gross and C. H. Llewellyn Smith, Nucl. Phys. B 14, 337 (1969). * (26) P. Fernandez de Cordoba and E. Oset, Phys. Rev. C 46, 1697 (1992). * (27) D. Casper, Nucl. Phys. Proc. Suppl. 112, 161 (2002). * (28) M. O. Wascko, Nucl. Phys. Proc. Suppl. 159, 79 (2006). * (29) M. Tzanov et al., Phys. Rev. D 74, 012008 (2006). * (30) S. A. Kulagin and R. Petti, Nucl. Phys. A765, 126 (2006). * (31) S. A. Kulagin, Nucl. Phys.A 640, 435 (1998). * (32) M. Hirai, S. Kumano and T. H. Nagai, Phys. Rev. C 70, 044905 (2004). * (33) S. A. Kulagin and R. Petti, Phys. Rev. D 76, 094033 (2007). * (34) J. H. Kim et al., Phys. Rev. Lett. 81, 3595 (1998). * (35) F. Bergsma et al., Phys. Lett. B 123, 269 (1988). * (36) L. S. Barabash et al., arXiv:hep-ex/9611012. * (37) J. W. Qiu and I. Vitev, Phys. Lett. B 587, 52 (2004).	arxiv-papers	2008-08-11T03:02:54	2024-09-04T02:48:57.223784	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Sajjad Athar, S. Chauhan, S. K. Singh and M. J. Vicente Vacas", "submitter": "Shri Singh krishna", "url": "https://arxiv.org/abs/0808.1437" }
0808.1555	11institutetext: Universitá di Torino, via P. Giuria 1, 10125 Torino, Italy 11email: baldi@oato.inaf.it 22institutetext: INAF - Osservatorio Astronomico di Torino, Strada Osservatorio 20, I-10025 Pino Torinese, Italy 22email: capetti@oato.inaf.it # Recent star formation in nearby 3CR radio-galaxies from UV HST observations ††thanks: Based on observations obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS 5-26555. Ranieri D. Baldi 11 Alessandro Capetti 22 We analyzed HST images of 31 nearby (z $\mathchar 13358\relax$ 0.1) 3CR radio- galaxies. We compared their UV and optical images to detect evidence of recent star formation. Six objects were excluded because they are highly nucleated or had very low UV count rates. After subtracting the emission from their nuclei and/or jets, 12 of the remaining 25 objects, presenting an UV/optical colors NUV - r $<$ 5.4, are potential star-forming candidates. Considering the contamination from other AGN-related processes (UV emission lines, nebular continuum, and scattered nuclear light), there are 6 remaining star-forming “blue” galaxies. We then divide the radio galaxies, on the basis of the radio morphology, radio power, and diagnostic optical line ratios, into low and high excitation galaxies, LEG and HEG. While there is no correlation between the FR type (or radio power) and color, the FR type is clearly related to the spectroscopic type. In fact, all HEG (with one possible exception) show morphological evidence of recent star formation in UV compact knots, extended over 5-20 kpc. Conversely, there is only 1 “blue” LEG out of 19, including in this class also FR I galaxies. The picture that emerges, considering color, UV, optical, and dust morphology, is that only in HEG recent star formation is associated with these relatively powerful AGN, which are most likely triggered by a recent, major, wet merger. Conversely, in LEG galaxies the fraction of actively star-forming objects is not enhanced with respect to quiescent galaxies. The AGN activity in these sources can be probably self-sustained by their hot interstellar medium. ###### Key Words.: Galaxies: active – Galaxies: elliptical and lenticular, cD – Galaxies: photometry – Galaxies: evolution – Galaxies: interactions – Galaxies: starburst – Ultraviolet: galaxies ††offprints: R.D. Baldi ## 1 Introduction Both observational and theoretical studies have supported the idea of a co- evolution between supermassive black holes (SMBH) and their host galaxies. The observational evidence is derived from several pieces of evidence: the dynamical signature of the widespread presence of SMBH in galaxies (e.g Kormendy & Richstone 1995; Richstone et al. 1998; Kormendy & Gebhardt 2001), the relation between SMBH mass and the spheroid mass (Magorrian et al. 1998; McLure & Dunlop 2002; Marconi & Hunt 2003; Häring & Rix 2004), stellar velocity dispersion (Ferrarese & Merritt 2000; Gebhardt et al. 2000; Tremaine et al. 2002) and concentration index (Graham et al. 2001; Graham & Driver 2007). These relations were interpreted by e.g. Hopkins et al. (2007b) as various projections of a fundamental plane, which included the SMBH mass, analogous to a similar relation for elliptical galaxies (Dressler et al. 1987; Djorgovski & Davis 1987); this appears to be a sign of a common evolutionary process involving SMBH and galaxies. The growth of a SMBH can occur via either gas accretion or coalescence with another SMBH during a merger event. Similarly, the growth of a galaxy can be associated with the capture of the stellar populations of another galaxy, due to either a merger or to star formation, which might be self-sustained or triggered by a merger. In this framework, nuclear activity (i.e. the manifestation of gas accretion onto a SMBH) and star formation are expected to be related, and for both processes mergers are likely to play a crucial role. Indeed, the hierarchical galaxy evolution scenarios support the idea that gas flows associated with galaxy mergers trigger both starburst and AGN activity (Kauffmann & Haehnelt 2000; Di Matteo et al. 2005). A connection between mergers, galactic starburst, and AGN activity has been well-established and modeled (e.g. Barnes & Hernquist 1991, 1996; Mihos & Hernquist 1994, 1996; Springel et al. 2005; Kapferer et al. 2005). Hopkins et al. (2006) presented a simulation in which starburst, AGN activity, and SMBH growth were connected by an evolutionary sequence, due to mergers between gas-rich galaxies. Observationally, studies of ultraluminous infrared galaxies (ULIRGs), distant submillimeter galaxies (SMGs), and quasars demonstrate that they are the remnants of major mergers, in which massive starbursts occur in combination with a dust enshrouded AGN (Barnes & Hernquist 1996; Schweizer 1998; Jogee 2006; Sanders et al. 1988a, b; Sanders & Mirabel 1996; Dasyra et al. 2007). Studies of QSO host galaxies also revealed the widespread presence of a young stellar population (e.g. Brotherton et al. 1999; Kauffmann et al. 2003; Sánchez et al. 2004; Zakamska et al. 2006). In Seyfert galaxies, evidence of nuclear, dusty, compact starbursts was found (Heckman et al. 1997; González Delgado et al. 1998). Apparently, AGN activity and star formation are connected also from a quantitative point of view, since the most luminous quasars have the youngest host stellar populations (Jahnke et al. 2004; Vanden Berk et al. 2006) and the most significant post-merger tidal features and disturbances (e.g. Canalizo & Stockton 2001; Hutchings et al. 2003; Letawe et al. 2007). Most of studies focused on the general AGN population when radio-quiet AGN are not studied in isolation. However, there have been plentiful studies of analysis of radio-loud AGN, which are the focus of this paper. Observations of radio galaxies in the local Universe (Heckman et al. 1986; Smith & Heckman 1989; Colina & de Juan 1995) detected morphological features such as double nuclei, arcs, tails, bridges, shells ripples, and tidal plumes. In a substantial subset of radio galaxies, this suggests that the AGN activity is triggered by the accretion of gas during major mergers and/or tidal interactions. Furthermore, gas kinematics studies (Tadhunter et al. 1989 and Baum et al. 1992) supported the interpretation that mergers can represent the triggering process of SMBH/galaxies co-evolution. Studies of the spectral energy distribution (SEDs) of radio galaxies discovered that young stellar populations, indicative of a recent starburst, provide a significant contribution to the optical/UV continua, up to 25-40 %, of powerful galaxies at low and intermediate redshift (Lilly & Longair 1984; Smith & Heckman 1989; Tadhunter et al. 1996; Melnick et al. 1997; Aretxaga et al. 2001; O’Dea et al. 2001; Tadhunter et al. 2002; Wills et al. 2002; Allen et al. 2002; Wills et al. 2004; Raimann et al. 2005; Tadhunter et al. 2005; Holt et al. 2007). Apparently, AGN activity initiates at a later stage of the merger event than star formation (Emonts et al. 2006), which agrees with predications of numerical simulations (Hopkins et al. 2007a), although the precise length of the delay is not well constrained. However, there are still several open questions, such as: (i) what is the fraction of nearby radio galaxies showing evidence for young stars?; (ii) is there a relationship between AGN properties and galaxy star-formation characteristics?; (iii) what is the role of mergers in triggering both AGN activity and star formation in radio-galaxies? We consider these issues by adopting an alternative approach based on the analysis of UV images with the aim of detecting a young stellar population. More specifically, we analyzed HST/STIS observations of a sample of nearby radio galaxies from the 3CR catalogue. The flux level in the UV band ($\sim$ 2500 Å), relative to the optical emission, is sensitive to very low levels of star formation, as demonstrated by GALEX data (the GALEX/NUV band is similar to the filter used to acquire the STIS images used in this study), by detecting fraction of as low as 1-3 % of young stars formed within the last billion years (Silk 1977 and Somerville et al. 2000). The HST UV images were already used for this purpose by O’Dea et al. (2001), who studied the radio- galaxy 3C 236. The sample considered is sufficiently large (31 objects) for statistical conclusions and is representative of the overall population of nearby radio galaxies. Furthermore, Schawinski et al. (2006) recently analyzed GALEX and SDSS images of a significant number of non-active early-type galaxies. This study provides a well defined control sample that we can use as a benchmark for the star formation properties of quiescent galaxies to be compared with the measurements we will derive for radio-galaxies. We note that Schawinski et al. (2006) focused only on quiescent galaxies to avoid contamination by UV light from an active nucleus. For the sample that we consider here, the light produced by AGN activity (UV nuclei, and jets) can however be masked by taking advantage of the high resolution of the HST images. The outline of this paper is the following. In Sect. 2, we describe the properties of the sample and HST observations, on which this study is based, as well as the reduction method used. In Sect. 3, we present our results by classifying the galaxies in terms of their NUV-r color profiles and UV morphology. In Sect. 4, we discuss the link between AGN and star formation properties, while in Sect. 5 we summarize our findings. The cosmological parameters used in this paper are H0 = 75 $\rm{\,km\,s}^{-1}$ Mpc-1 and q0 = 0.5. ## 2 Observations and data reduction We selected all 3CR radio-galaxies for which STIS near-UV observations were available in the public archive. The sample consists of 31 low-z radio- galaxies with z $<$ 0.1, apart form 3C 346 at z = 0.16. For 25 objects, the UV data were published by Allen et al. (2002). To this sample we added a subset of 6 objects observed by various HST projects. Our final sample represents $\sim$60 % of the entire 3CR sample with z $<$ 0.1 (49 galaxies) that contains an almost equal fraction (and coverage) of FR I and FR II galaxies. Since the objects discussed in Allen et al. (2002) sample were selected at random (since the data were obtained as part of a HST snapshot proposal) and only 6 objects were added our sample is representative of the entire low-z 3CR sample and has no evident selection biases. We analyzed UV and optical HST images of our sample. All UV observations employed the STIS near-ultraviolet (NUV) Cs2Te Multianode Microchannel Array (MAMA) detector, which has a field of view of 25${}^{\prime\prime}\times$25′′ and a pixel size of $\sim 0\aas@@fstack{\prime\prime}$024\. The filters used during observations were the following: most objects were observed with the F25SRF2 filter (centered on 2320 Å with a FWHM of 1010 Å); for three additional galaxies, the F25CN182 filter was used (with a pivot wavelength of 1983 Å and a FWHM of 630 Å), while the remaining 5 observations were obtained with the F25QTZ filter (centered on 2364.8 Å with a FWHM of 995.1 Å), which provides more effective rejection of the OI$\lambda$1302 Å geocoronal emission than F25SRF2. All of these filter band cutoffs remove geocoronal Ly$\alpha$ emission. The observations log is given in Table 1. The optical observations were acquired using the WFPC2 camera. Most objects were located in the PC field with a field of view of $\sim 36\aas@@fstack{\prime\prime}4\times 36\aas@@fstack{\prime\prime}4$ and pixel size of $\sim 0\aas@@fstack{\prime\prime}$04553\. Some objects were located in the WF fields, which individually have a field of view of 1.3${}^{\prime}\times$1.3′ and pixel size of $\sim 0\aas@@fstack{\prime\prime}1$. For all galaxies, the used filter was F702W with a pivot wavelength of 6919 Å and a bandwidth of 1385 Å 111Only one object (3C 192) lacked the WFPC2 images and then its SDSS optical image has been employed. The WFPC2 exposure times were 280 s or 560 s. The optical observations were published by de Koff et al. (1996) and Martel et al. (1999). The UV and optical images were processed by the standard HST pipeline, developed at the STScI, in the package STSDAS in IRAF222IRAF is distributed by the National Optical Astronomical Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. which corrects for flat- fielding and, for optical images, also bias subtraction. For optical data, the CCD images were combined and the cosmic-ray events were cleaned in a single step using the IRAF/STSDAS task crrej, when multiple exposures of the same target were available; otherwise cosmic rays were removed using the task cosmicrays of the IRAF package. The images were flux calibrated by using the keyword PHOTFLAM in the calibrated science header file, and divided by the value of the keyword EXPTIME (exposure time). The expected calibration error was given by the error in the value of PHOTFLAM: for optical band images, this was 2 % and, for the UV band, this was 5 %. ### 2.1 UV / optical colors To estimate the UV / optical colors, we measured the total UV and optical flux for each object in the sample. The optical images were registered onto the UV images; the registration was performed after re-binning the UV images to a pixel size of 0$\aas@@fstack{\prime\prime}$45 as to enable detection of regions of low brightness UV emission. We registered the images by aligning, when present in the field of view, optical and UV nuclei and other objects (companion galaxies and/or stars). Otherwise, in the case of diffuse sources (e.g. 3C 40), we cross-correlated the UV with the R images to estimate the relative shift. Different masks, defined individually for each galaxy, were applied to both images, which identified areas containing nearby galaxies or stars. More importantly, we flagged regions contaminated by emission from extended jets. The presence of a optical and/or UV jet is noted in the Table 1. Similarly, we masked the nuclear regions (with a typical size of 0$\aas@@fstack{\prime\prime}$12-0$\aas@@fstack{\prime\prime}$25) in all objects where the studies of Chiaberge et al. (1999) and Chiaberge et al. (2002) identified the presence of unresolved nuclei in the optical and/or UV images (as marked in Table 1). The background level was estimated from a circular annulus of 12$\aas@@fstack{\prime\prime}$5 in radius and 3$\aas@@fstack{\prime\prime}$6 in width for both the optical and UV images. This was the largest region in the UV images that had not been contaminated by enhanced dark current at the edges of the STIS detector. For objects located close to the edges of the frame, we used instead circular sectors for the background estimate. The UV background rate measurements were compared with those obtained from the same images by Allen et al. (2002): our values were in good agreement with their estimates, although they are lower on average by 20% (and as much as 40 % for one object). This is most likely because the region that we used to measure the background was located at larger radii from the center of the galaxy. We measured UV and optical fluxes within a set of concentric apertures, centered on each object, with different radii 2′′, 4′′, 6′′, 8′′, and 10′′. From the total counts within each region, we subtracted the estimated background and applied the appropriate photometric conversion to obtain a flux in units of erg s-1 cm-2 Å-1. These fluxes were then transformed onto the AB magnitude system and used to estimate the NUV-R color. With this approach, we measured the integrated NUV-R color integrated over different apertures out to the distance at which the error in the color was smaller than 0.2 mag333The error in the NUV-R color increases with increasing aperture due to the uncertainty in the UV background.. At this stage, we discarded 6 galaxies of the sample. For two objects (namely 3C 353 and 3C 452), the integrated UV count rate was insufficient to measure accurately (with an error smaller than 0.5 mag) the galaxy’s color. Four targets (namely 3C 227, 3C 371, 3C 382, and 3C 390.3) were highly nucleated. The color of these galaxies were strongly dependent on the nuclear subtraction and so it was difficult to separate any genuine diffuse emission from the halo of the bright nuclear point source that might dominate even at significant distances from the galaxy center. After these procedures, we obtained the NUV-R color profiles for 25 3CR radio- galaxies, which are shown in Fig. 1 (grouped into the categories described in Sect. 3). The NUV-R colors obtained were corrected for Galactic absorption by an amount equal to 2.48$\times$E(B-V), derived from the extinction in the form given by Cardelli et al. (1989). In most cases (and we highlight the few exceptions below), the color profiles were flat with color variations smaller than $\pm 0.2$ mag. Therefore, the color for each galaxy was almost independent of the aperture used. Table 1: Observation log Name \| Filter Exp. time \| Date \| Opt CCC \| UV CCC \| Opt Jet \| UV Jet ---\|---\|---\|---\|---\|---\|--- 3C 015 \| F25QTZ \| 7230 \| 01-01-27 \| YES \| YES \| NO \| YES 3C 029 \| F25SRF2 \| 1440 \| 01-01-27 \| YES \| YES \| NO \| NO 3C 035 \| F25SRF2 \| 1440 \| 99-10-21 \| NO \| NO \| NO \| NO 3C 040 \| F25SRF2 \| 1440 \| 00-06-03 \| NO \| NO \| NO \| NO 3C 066B \| F25SRF2 \| 1440 \| 99-07-13 \| YES \| YES \| NO \| YES 3C 078 \| F25QTZ \| 2000 \| 00-03-15 \| YES \| YES \| YES \| YES 3C 192 \| F25SRF2 \| 1440 \| 00-03-23 \| NO \| NO \| NO \| NO 3C 198 \| F25SRF2 \| 1440 \| 00-04-23 \| YES \| YES \| NO \| NO 3C 227 \| F25SRF2 \| 1440 \| 00-01-25 \| YES \| YES \| NO \| NO 3C 236 \| F25SRF2 \| 1440 \| 99-10-05 \| NO \| NO \| NO \| NO 3C 264 \| F25QTZ \| 3600 \| 00-02-12 \| YES \| YES \| YES \| YES 3C 270 \| F25SRF2 \| 1440 \| 00-03-05 \| YES \| NO \| NO \| NO 3C 274 \| F25QTZ \| 640 \| 03-06-08 \| YES \| YES \| YES \| YES 3C 285 \| F25SRF2 \| 1440 \| 00-04-16 \| YES \| NO \| NO \| NO 3C 293 \| F25SRF2 \| 1440 \| 00-06-14 \| NO \| NO \| NO \| NO 3C 296 \| F25SRF2 \| 1440 \| 00-04-15 \| YES \| NO \| NO \| NO 3C 305 \| F25SRF2 \| 1440 \| 00-04-27 \| NO \| NO \| NO \| NO 3C 310 \| F25SRF2 \| 1440 \| 00-06-10 \| YES \| YES \| NO \| NO 3C 317 \| F25SRF2 \| 1440 \| 99-07-26 \| YES \| YES \| NO \| NO 3C 321 \| F25SRF2 \| 1440 \| 00-06-05 \| NO \| NO \| NO \| NO 3C 326 \| F25SRF2 \| 1440 \| 00-03-12 \| NO \| NO \| NO \| NO 3C 338 \| F25SRF2 \| 1440 \| 00-06-04 \| YES \| YES \| NO \| NO 3C 346 \| F25QTZ \| 3600 \| 00-08-22 \| YES \| YES \| YES \| YES 3C 353 \| F25SRF2 \| 1440 \| 00-06-22 \| NO \| NO \| NO \| NO 3C 371 \| F25QTZ \| 7989 \| 00-09-21 \| YES \| YES \| NO \| NO 3C 382 \| F25CN182 \| 1440 \| 00-02-23 \| YES \| YES \| NO \| NO 3C 388 \| F25SRF2 \| 1440 \| 00-06-02 \| YES \| YES \| NO \| YES 3C 390.3 \| F25CN182 \| 1440 \| 99-08-10 \| YES \| YES \| NO \| NO 3C 449 \| F25SRF2 \| 1440 \| 00-04-16 \| YES \| YES \| NO \| NO 3C 452 \| F25CN182 \| 1440 \| 00-01-30 \| NO \| NO \| NO \| NO 3C 465 \| F25SRF2 \| 1440 \| 00-05-25 \| YES \| YES \| NO \| NO Column description. Col. (1): 3CR source name. Col. (2): STIS filter used. Col. (3): exposure time [s]. Col. (4): observation dates. Col. (5)-(8): the presence of optical and UV central compact core (CCC) and jets in each object as taken by Chiaberge et al. (1999) and Chiaberge et al. (2002). ### 2.2 Cross-calibration of NUV-R colors The previous studies of galaxy UV/optical color analysis were performed using GALEX and SDSS observations, by measuring the NUV-r color, while we used NUV and optical HST observations. Before we can compare our HST-based NUV-R color, which for clarity we refer to as (NUV-R)HST, with results present in the literature, we must complete a cross-calibration between GALEX, SDSS, and HST data. The GALEX near-ultraviolet filter is centered on $\lambda_{eff}$ = 2271 Åwith $\Delta\lambda$ = 732 Å. The images correspond to a circular field of view of radius $\sim$38$\arcmin$ and a spatial resolution of $\sim 5\arcsec$. The Sloan Digital Sky Survey (SDSS) images (Fourth Data Release) have a field of view of 13.51$\arcmin$$\times$8.98$\arcmin$ and a pixel size of 0$\aas@@fstack{\prime\prime}$396\. The r-filter band is centered at 6231 Å with a FWHM of 1373 Å. The filter transmission curves are shown in Fig. 2, with a typical spectrum of an old (11 Gyr) stellar population. We note that the STIS passbands are wider than the GALEX/NUV passband, while there is a small shift between the centers of the SDSS and HST/F702W optical filters. Figure 1: The NUV-R color (HST based) profiles of the sample: (a),(b),(c) red quiescent galaxies; (d), and (e) blue UV clumpy galaxies, separated into NUV-r color increasing and decreasing with radius, (f) blue UV-disky galaxies. For these last objects we present also the NUV-R color at 1′′ and at the aperture corresponding to their dusty disks in order to appreciate better the color profiles bluing to smaller radii. Figure 2: The transmission curves of HST (STIS/F25SRF2 filter and WFPC2/F702W filter are marked with a solid line), GALEX-SDSS (NUV filter and r filter are marked with a dashed line) overplotted on a galaxy spectrum model taken from Bruzual & Charlot (2003) with 11 Gyr and Z = 0.008Z⊙. Ideally, cross-calibrations should be performed by directly comparing images of the same objects taken with different telescopes in the band of interest. However, the object with the highest value of total counts in the GALEX NUV image in our sample is 3C 040, observed as part of the MIS survey. However, even for this object the flux ratio between HST and GALEX (0.7 $\pm$ 0.2) has an associated error that it is too large for a meaningful cross-correlation. We therefore decided to compare the GALEX and HST data using stellar population synthesis models; this was similar to the method described by Proffitt (2006). We used 18 models from Bruzual & Charlot (2003) in a grid of different age (1,3,5,7,9, and 11 Gyr) and metallicity (0.008, 0.02 [= Z⊙], and 0.05). We convolved these spectra with the different transmission curves in NUV and optical bands to compare the resulting fluxes and estimate the color correction to be applied to the HST colors. As shown in Fig. 3, left panel, the conversion of the NUV magnitude from STIS to GALEX has a significant dependence on the stellar population age and metallicity. For a redder stellar population, the larger width of the STIS filter towards longer wavelength, increases the measured STIS flux with respect to the GALEX data. The correction that has to be applied to the STIS data has an almost linear dependence on (NUV-R)HST. Conversely, the correction to convert the optical band images from the WFPC2 to SDSS bandpasses is far smaller and has only a weak trend with stellar population, since in this band the spectrum is flat. In this case, we were able to check directly the results of the stellar population analysis by comparing the fluxes for the same aperture in the HST/WFPC2 (opportunely convolved) and SDSS r-band images. The results obtained confirmed the need for only a small magnitude correction (a few hundredths of a magnitude). In the middle panel of Figure 3, we show the correspondence between the NUV-R HST colors and the NUV-r color in the GALEX/SDSS system. It can be reproduced with a linear relation in the form: (NUV-r) - (NUV-R)HST = 0.16 $\times$ [(NUV-R)HST \- 4] - 0.07 In the right panel of Figure 3, we present the residuals from this linear relation of $\sim$ 0.2 mag, which we adopt as a conservative error in the color conversion. We applied the same method used for the STIS/F25SRF2 filter to the STIS/F25QTZ filter, which differs only slightly with short wavelength cutoff at longer wavelength. The color conversion in this case is (NUV-r) - (NUV-R)HST = 0.15 $\times$ [(NUV-R)HST \- 4] - 0.04 We also tested the cross-calibration between GALEX-SDSS and HST data by using a two-stellar-populations galaxy model and varying the relative contribution of a 1 Gyr and 11 Gyr stellar population models at three different metallicities. We obtained a quasi-linear cross-calibration relation, which links the points of the two individual populations in Fig. 3. The maximum residual is 0.2 mag, similar to our typical single-stellar-population model cross-calibration error. In Table 2, we indicate the aperture size used for each object and the corresponding color. This (NUV-R)HST color was transformed into the NUV- r color using the prescription described above. Figure 3: Conversion from the HST NUV-R color to GALEX/SDSS NUV-r color, estimated from models of stellar population synthesis from Bruzual & Charlot (2003) as marked in the figure. The left panel shows the linear relation between the NUV magnitudes and the observed HST color. The middle panel shows the connection between HST and GALEX-SDSS color with overplotted the linear relation we used for the color correction. The right panel shows the residuals from this linear relation. Table 2: Properties of the 3CR sample. 3C \| z \| log $P_{178}$ \| FR - E.L. \| MR \| E(B-V) \| $\sigma$ \| (NUV-R)HST \| aperture (′′ kpc) \| UV morphology ---\|---\|---\|---\|---\|---\|---\|---\|---\|--- 15 \| 0.073 \| 26.2 \| I/II - LEG \| -22.41 \| 0.022 \| \| 5.87 $\pm$ 0.06∗ \| 10 \| 14.1 \| disky 29 \| 0.0447 \| 25.8 \| I \| -22.73 \| 0.036 \| 231 \| 6.2 $\pm$ 0.1 \| 10 \| 7.0 \| 35 \| 0.0670 \| 26.0 \| II - LEG \| -22.34 \| 0.141 \| \| 5.6 $\pm$ 0.2 \| 6 \| 2.6 \| 40 \| 0.0177 \| 25.2 \| II - LEG \| -22.62 \| 0.041 \| 242 \| 5.58 $\pm$ 0.02 \| 10 \| 3.5 \| 66B \| 0.0215 \| 25.4 \| I \| -23.42 \| 0.080 \| \| 5.76 $\pm$ 0.03 \| 8 \| 3.3 \| 78 \| 0.029 \| 25.5 \| I \| -23.23 \| 0.173 \| 263 \| 4.96 $\pm$ 0.03∗ \| 10 \| 5.6 \| disky 192 \| 0.060 \| 26.2 \| II - HEG \| -21.60 \| 0.054 \| \| 4.93 $\pm$ 0.05 \| 6 \| 6.9 \| knotty 198 \| 0.0815 \| 26.1 \| II - HEG \| -20.82 \| 0.026 \| 174 \| 3.09 $\pm$ 0.09 \| 8 \| 12.6 \| knotty 227 \| 0.0861 \| 26.7 \| II - WQ \| -21.42 \| 0.026 \| \| nucleated \| \| \| 236 \| 0.0989 \| 26.5 \| II - HEG \| -23.10 \| 0.011 \| \| 5.04 $\pm$ 0.08 \| 8 \| 11.7 \| knotty 264 \| 0.0217 \| 25.4 \| I \| -22.51 \| 0.023 \| 271 \| 5.87 $\pm$ 0.02∗ \| 10 \| 4.2 \| disky 270 \| 0.0073 \| 24.8 \| I \| -22.11 \| 0.018 \| 309 \| 5.21 $\pm$ 0.01 \| 8 \| 1.2 \| 274 \| 0.0037 \| 25.6 \| I \| -22.39 \| 0.022 \| 333 \| 5.14 $\pm$ 0.01 \| 10 \| 0.85 \| 285 \| 0.0794 \| 26.1 \| II - HEG \| -22.26 \| 0.017 \| 181 \| 3.79 $\pm$ 0.03 \| 8 \| 12.3 \| knotty 293 \| 0.0452 \| 25.8 \| I/II - LEG \| -22.10 \| 0.017 \| 201 \| 4.30 $\pm$ 0.03 \| 10 \| 8.7 \| knotty 296 \| 0.0237 \| 25.2 \| I \| -23.38 \| 0.025 \| \| 5.59 $\pm$ 0.01 \| 10 \| 4.8 \| 305 \| 0.0414 \| 25.8 \| I/II - HEG \| -22.88 \| 0.029 \| 193 \| 4.62 $\pm$ 0.02 \| 10 \| 8.1 \| knotty 310 \| 0.0540 \| 26.5 \| I \| -22.30 \| 0.042 \| 209 \| 5.4 $\pm$ 0.1 \| 10 \| 6.3 \| 317 \| 0.0350 \| 26.1 \| I \| -23.18 \| 0.037 \| 216 \| 5.21 $\pm$ 0.04 \| 10 \| 6.7 \| 321 \| 0.0960 \| 26.4 \| II - HEG \| -22.58 \| 0.044 \| \| 3.32 $\pm$ 0.03 \| 6 \| 11.2 \| knotty 326 \| 0.0885 \| 26.5 \| II - LEG \| -21.71 \| 0.053 \| \| 6.0 $\pm$ 0.2 \| 4 \| 3.5 \| 338 \| 0.0298 \| 26.0 \| I \| -23.50 \| 0.012 \| 310 \| 5.41 $\pm$ 0.02 \| 8 \| 4.7 \| 346 \| 0.1610 \| 26.8 \| II - LEG \| -22.11 \| 0.067 \| \| 5.4 $\pm$ 0.2 \| 4 \| 12.6 \| 353 \| 0.0304 \| 26.7 \| II - LEG \| -21.36 \| 0.439 \| \| 5.0 $\pm$ 0.5 \| 2 \| 1.2 \| 371 \| 0.051 \| 26.5 \| II - BLLAC \| -22.66 \| 0.036 \| \| nucleated \| \| \| 382 \| 0.0578 \| 26.2 \| II - WQ \| -23.50 \| 0.070 \| \| nucleated \| \| \| 388 \| 0.0908 \| 26.6 \| II - LEG \| -23.07 \| 0.080 \| 408 \| 5.8 $\pm$ 0.2 \| 8 \| 10.7 \| 390.3 \| 0.0561 \| 26.5 \| II - WQ \| -22.21 \| 0.071 \| \| nucleated \| \| \| 449 \| 0.0171 \| 24.9 \| I \| -21.96 \| 0.167 \| 253 \| 4.57 $\pm$ 0.03∗ \| 10 \| 3.3 \| disky 452 \| 0.0811 \| 26.9 \| II - HEG \| -22.21 \| 0.137 \| \| 6.0 $\pm$ 1.6 \| 2 \| 3.1 \| 465 \| 0.0293 \| 25.8 \| I \| -23.64 \| 0.069 \| 356 \| 5.37 $\pm$ 0.03 \| 10 \| 5.9 \| Column description. Col. (1): 3CR source name. Col. (2): redshift (from NED). Col. (3): Log radio power (W Hz-1) at 178 MHz taken by Allen et al. (2002) and Kellermann et al. (1969). Col (4): radio morphological and optical spectral classification, as taken from Jackson & Rawlings (1997) and from the website http://www.science.uottawa.ca/∼cwillott/3crr/3crr.html with the only variation from the literature is the classification of 3C 236 in HEG derived from SDSS data. Col. (5): magnitude in R band as taken from Donzelli et al. (2007) and from NED. Col. (6): galactic extinction (from NED). Col. (7): stellar velocity dispersion taken from Hyperleda, the website http://www.mpa- garching.mpg.de/SDSS/DR4/raw_data.html and Smith et al. (1990). Cols. (8): NUV-r integrated color on the greater available aperture with the corresponding scale radius in (9) arcsec and (10) kpc. The objects labelled with * are the UV-disky galaxies, whose (NUV-R)HST colors, corrected excluding the UV disks contaminated by the emission lines (see Sect. 3.1), are: 6.1 $\pm$ 0.1 for 3C 015 , 5.24 $\pm$ 0.08 for 3C 078, 6.06 $\pm$ 0.06 for 3C 264, and 5.2 $\pm$ 0.1 for 3C 449. These NUV-r colors are those that we finally use in the plots. Col. (11): UV morphology. ## 3 RESULTS As described by Schawinski et al. (2006), the NUV-r color can be used as a means of identifying recent star formation. They adopted a threshold at NUV-r $<$ 5.4 to separate “red” quiescent galaxies from “blue” galaxies with active star formation. This value is suggested by both theoretical population synthesis models, although the availability of these models is limited, and empirical data (Lee et al. 2005). Despite the possible presence of an UV upturn, observed often in early-type galaxies (Code & Welch 1979 and Burstein et al. 1988), passively evolving galaxies do not appear to have NUV-r colors bluer than 5.4. Given our 0.2 mag calibration uncertainty, we consider only galaxies bluer than NUV-r $<$ 5.2 to be bona-fide “blue” galaxies. Figure 4: a) The UV images of 4 illustrative examples of red diffuse galaxies. Note the UV nucleus of 3C 317, masked in our analysis, superposed to the diffuse emission. b) The UV images of the blue UV-clumpy galaxies show present star formation on large scale ($\sim$5-20 kpc) arranged in discrete knots. We also show the orientation of their radio-axis. c) Images of the UV-disky galaxies which show, apart from their UV jets and nuclei, UV emission in a small scale circumnuclear disks ($\sim$0.5-1 kpc). On the basis of this criteria for the integrated colors and UV morphology, the galaxies of the sample can be divided in 3 main categories: 1. 1. quiescent red galaxies: 14 object have red and flat integrated color profiles (Fig. 1 a,b,c). They lack any sign of significant recent star formation, since, apart from emission from the active nucleus (and from extended UV/optical jets), we detect only diffuse UV light tracing the optical emission (Fig. 4, a panel). We note that this description applies to the 7 galaxies with 5.2 $<$ NUV-r $<$ 5.6, whose classification is uncertain due to our calibration accuracy of 0.2 mag. 2. 2. blue UV-clumpy galaxies: 7 objects have NUV-r $<$ 5.2 and show a clumpy UV morphology (and they are labeled as “knotty” in Table 2), which are usually associated with dust lane structures (see Fig. 4, b panel). Their color profiles show different trends, since in some cases NUV-r increases toward larger radii (Fig. 1, d panel), but the opposite behavior is also observed (e panel). Their UV emission is extended over 5-20 kpc. 3. 3. UV-disky galaxies: 2 objects (namely 3C 078 and 3C 449) have a blue NUV-r color, but the UV emission is located co-spatially with the circum-nuclear dusty disk (Fig. 4, c panel). Another 2 galaxies (namely 3C 015 and 3C 264), although globally red, show similar UV emission, which is associated with their disks. In all cases, the NUV-r color increases toward larger radii and, by extending the color profile to apertures smaller than 2′′ in 3C 015 and 3C 264, they also reach a blue color. All of these bright UV disks have scales of 0.5-1 kpc and these galaxies are labeled as disky in Table 2. ### 3.1 Origin of UV excess As explained above, in quiescent early-type galaxies a color bluer than NUV-r $<$ 5.4 can be considered as evidence for the presence of a young stellar populations. However, in active galaxies, it is fundamental to assess the origin of this UV emission. Other emission processes related to the presence of an AGN can contribute to this excess, in particular UV nuclear scattered light, nebular continuum, and emission lines. Emission line images represent a useful tool to tackle the issue of AGN UV contamination. The sites where emission lines are produced mark the geometrical intersection between the interstellar medium and the nuclear ionizing light, i.e. the regions where both UV lines and UV scattered light can be found. This analysis is crucial in particular for the “blue” objects. We focused on the UV-disky galaxies with UV emission, that is cospatial with a circum-nuclear dusty disk. We then considered H$\alpha$ images from the HST archive, which are available for 3C 078, 3C 264, and 3C 449. In all 3 galaxies, there was a clear morphological correspondence between the UV and H$\alpha$ emitting regions (for example see Fig. 5), which suggests a possible contribution from UV emission lines. We then quantified the fraction of contamination of emission lines to the total UV light by opportunely scaling the H$\alpha$ flux. We adopted the line ratios measured by Dopita et al. (1997) in the emission line disk of M87, a well studied analogue to our disked sources, for which HST UV spectroscopy is available. We found that UV emission lines contribute for between 50 and 80 % of the UV light within the area covered by the disk, with the strongest contribution originatin in C IV$\lambda 1550$. The contamination would be even stronger adopting the UV/optical line ratios measured by Ferland & Osterbrock (1986) in the NLR region of nearby AGN, where the C IV line is more prominent. These results (that we extend for analogy to the fourth galaxy in this group, 3C 015, already classified as a red object) suggest that the disk-like structures seen in UV light are due to line emission and not to a young stellar population. Excluding the disk regions, both “blue” UV-disky objects have a redder integrated NUV-r color of 5.3 and 5.4, for 3C 449 and 3C 078 respectively. Figure 5: The images (from the left to the right panel) of 3C 264 in H$\alpha$, in optical band (HST-F702W) and in NUV band (HST-F25QTZ). We note the spatial correspondence between the H$\alpha$ and UV emission region with the dusty disk. Figure 6: The images (from the upper to lower panel) of 3C 285, 3C 321, and 3C 305 in (from the left to the right panel) H$\alpha$ and [O III] (available only for 3C 285), in optical band and in NUV band. The size of all images are 10$\arcsec$ $\times$ 10$\arcsec$, corresponding to physical scales of 15.4, 18.6, and 8.1 kpc for 3C 285, 3C 321, and 3C 305 respectively. We also show the orientation of their radio-axis. We then focused on galaxies that present a clumpy morphology in UV emission. In analogy with the analysis performed on UV-disky galaxies, we selected emission-line images (in H$\alpha$ and [O III]) from the HST archive, which were available for 3C 285, 3C 305, and 3C 321. We then considered each object individually: 1) in 3C 285, the H$\alpha$ emission was cospatial with the dust feature bisecting its nuclear region, while the [O III] emission is confined to a one-sided elongated structure well aligned with its radio axis (see Fig. 6 top panels). There appears to be no significant contribution by AGN related processes to the large scale UV emission in this galaxy, since this is far more extended and not cospatial with the line-emitting regions; 2) in 3C 321 (reminding that the companion galaxy on the NW side was masked in our analysis), the H$\alpha$-line image presents a well defined conical morphology centered on the SE radio jet. The same structure is seen in UV light and this provides strong support for AGN contamination (either scattered light or line emission) in this region. However the UV light in this object originates for about 2/3 of the total flux from the chain of knots associated with the nuclear dust lane that most likely represents an extended region of star formation, since it has no counterpart in the H$\alpha$ image and is unrelated to the radio-structure; 3) in 3C 305, the emission line takes the form a S-shaped structure, aligned with the radio axis, reminiscent of what is often found in Seyfert galaxies (e.g. Mrk 3, Capetti et al. 1999). The UV emission has a similar structure and this indicates possible AGN contamination444As discussed below not from nebular processes, but eventually from scattered light.. However, the location of the UV emission knots, cospatial with the nuclear dust structure, might support the possiblity that they are star- forming regions. Based on our data alone, we cannot reach a robust conclusion. However, stellar population synthesis models applied to long slit spectra of 3C 305 imply that a young, spatially extended, stellar population is present(Tadhunter et al. 2005) and we therefore consider this object to be an actively star-forming galaxy. For the remaining 4 objects, whose HST emission-line images are unavailable, we note that, in the case of scattered nuclear UV light, as expected by the AGN unification model, the UV light should be aligned with the radio-axis as observed for 3C 285 and 3C 321. Although such UV/radio alignment might possibly be present in 3C 192, no association between radio and UV light in the other 3 UV-knotty galaxies is present. A stronger link is instead present with the dust structures, as already noted by Allen et al. (2002) from the inspection of the optical HST images. For example, in both 3C 305 and 3C 321 the brighter UV knots closely follow the edges of the dust lanes bisecting their host galaxies. A similar association is seen also in 3C 285 and 3C 293. We now consider the possibility of an emission line origin for the UV excess. All seven “blue” UV-clumpy galaxies were observed with the F25SFR2 filter, including Ly$\alpha$ at the edge of the transmission curve. Given their redshift, this corresponds to a low transmission efficiency (0.02-0.07 relative to the transmission peak). With a filter width of about 1000 Å, significant contamination is expected only for extreme values of the Ly$\alpha$ equivalent width of approximately $\sim$15000-50000 Å. More quantitatively, the Ly$\alpha$ fluxes can be measured from the H$\beta$ fluxes (see Table 3) taken from our own optical spectroscopy (Buttiglione et al. 2008), by adopting a scaling of FLyα/FHβ = 55 (Ferland & Osterbrock 1986). Line contamination is estimated to be typically 1 - 4 % and has a negligible effect on the integrated colors. Furthermore, since the line emission tends to be strongly nuclear concentrated (Mulchaey et al. 1996), most is contained within the nuclear region and is therefore excluded from our UV flux measurements. Concerning the other bright UV lines (i.e. C IV$\lambda 1549$, C III]$\lambda 1909$, and Mg II$\lambda 2798$), they are a factor of 5 to 30 times fainter than Ly$\alpha$; despite the higher throughput of the F25SFR2 filter at their respective wavelengths, their contribution to the UV flux is therefore also negligible. Nebular continuum emission also contributes to the UV emission. We estimated that the UV contamination from nebular continuum at $\sim$ 2300 Å (the center of the passband of the STIS images) is 0.004 $\AA^{-1}$ H$\beta$ in the low density limit and for a temperature of 104 K, adopting the coefficients given by Osterbrock (1989). The contamination from nebular continuum is estimated on a source-by-source basis (from the values reported in Table 3). In a similar way to the UV emission-line contamination, the nebular continuum affects the global NUV-r color by less than approximately 0.1 mag (see the percentages of nebular contamination in Table 3). Table 3: Contribution of the nebular continuum in the blue UV knotty galaxies Name \| UV cont. \| H$\beta$ \| neb. cont. (%) \| neb. cont (%) ---\|---\|---\|---\|--- \| \| \| photometric \| spectroscopic 3C 192 \| 10 \| 32 \| 11 \| 9.7 (1) 3C 198 \| 22 \| 34 \| 6 \| $<$1 (2) 3C 236 \| 10 \| 13 \| 5 \| $<$1 (2) 3C 285 \| 24 \| 6 \| 1 \| 1.8 (2) 3C 293 \| 46 \| 6 \| 1 \| $<$10 (3) 3C 305 \| 68 \| 26 \| 2 \| $<$10 (3) 3C 321 \| 33 \| 4 \| 1 \| 0.3-25.2 (2) Column description: (1): source name; (2): STIS UV flux in $10^{-17}\>{\rm erg}\,{\rm s}^{-1}\,{\rm cm}^{-2}\,{\rm\AA}^{-1}$ units; (3) H$\beta$ fluxes in $10^{-16}\>{\rm erg}\,{\rm s}^{-1}\,{\rm cm}^{-2}$ units; (4) percentage of UV contamination from nebular continuum with our photometric analysis and (5) with spectroscopic analysis from literature (6): 1, Wills et al. (2002); 2, Holt et al. (2007); 3, Tadhunter et al. 2005 (the range depends on the size of the aperture used). Until now, we have not taken into account the effects of internal reddening, which is clearly an important issue considering the large $\lambda$ coverage of the NUV-R color and the presence of dust in several targets. We discuss the reddening effects for our sample, considering separately the three different classes of galaxies: 1. 1. knotty blue galaxies: they all indeed have extended dust structures. However, the reddening-corrected NUV-R color would be even bluer than the uncorrected values. The presence of dust makes us overestimate the contamination from UV line emission and nebular continuum, strengthening the case for a young stellar population in these objects. 2. 2. Red quiescent galaxies: they typically have no significant dust lanes that can absorb UV and optical light. This is confirmed by the flatness of their NUV-R color profiles at different radii. Therefore, the internal reddening effect in these objects is likely to be negligible. 3. 3. Blue/red disk galaxies: the correction for internal reddening associated with their dusty disks would indeed make these objects bluer, and, furthermore, the absorption from this disk leads to an overestimate of the contamination from UV line emission, which we call their UV excess. We estimated the typical absorption of these disks, taking from the literature the V-R and R-I color excesses that they cause (Martel et al. 2000). The corresponding color excess is E(UV-R) = 0.15 - 0.52. Therefore the line contamination is not significantly altered by absorption within these disks (since the line flux is reduced only by 15-40%) and it is sufficient to cause their blue colors, even after reddening correction. We conclude that internal reddening causes to appear the knotty galaxies even bluer and is negligible for the other objects. Therefore it does not affect our main results. All of these arguments imply that only a young stellar population can explain the UV excess in these early-type galaxies, apart from the possible exceptions of 3C 192. ## 4 Discussion We focus on the global integrated NUV-r color of the objects in the 3C sample. For the UV-disky galaxies, the NUV-r global color is measured from the region of the galaxies outside their UV disks, contaminated by emission lines. The NUV-r color distribution of our sample (see Fig. 7) is peaked at NUV-r $\sim$ 5.4 but with a substantial blue tail. Comparing our results with those found by Kauffmann et al. (2006) for a volume-limited sample of AGN hosted by massive bulge-dominated galaxies, we note that the color distribution of their AGN, limited to those for the same range of stellar velocity dispersion $\sigma$ ($\sigma$ $>$ 180 km/s), is similar to that estimated from our sample. Figure 7: The NUV-r color distribution of the 3CR host galaxies. This distribution is peaked at NUV-r $\sim$ 5.4 but with a substantial blue tail. We consider in more detail the connection between AGN and recent star formation. First of all, the NUV-r color is not simply related to the radio power (see Fig. 8). While the bluest galaxies are all at relatively high values of radio power at 178 MHz (P178MHz), we also find many of the reddest galaxies at the high end of radio luminosities. Figure 8: NUV-r color versus radio power (178 MHz) taken by Allen et al. (2002) and Kellermann et al. (1969) for the sample. The solid line, drawn at NUV-r = 5.4, marks the separation between blue and red galaxies. The two dashed lines represent the level of accuracy of the color calibration. The circles represent the FR I objects, the triangles the intermediate FR I/II galaxies, and the diamonds FR II galaxies. The empty points correspond to low excitation galaxies (LEG), while the filled ones to high excitation galaxies (HEG). In Fig. 9 (upper panel), we separate the objects depending on their FR morphological class (Fanaroff & Riley 1974). All bona-fide blue galaxies belong to the FR II class or intermediate FR I/FR II objects, while no blue galaxy of FR I class is found. However, in the FR II classes, we find almost equal numbers of “blue” and “red” galaxies. No simple trend between NUV-r color and radio morphology is therefore present. We now investigate whether there is a connection between the extent of the radio emission and color. At least in the case of FR II, the radio-source size is related to its age. A small radio source size could relate to a recent onset of activity, triggered by a more recent merger, with a higher possibility of detecting the induced event of star formation with respect to large radio sources. On the basis of these considerations, we measured the radio source sizes of the FR II galaxies from images in the literature. For the blue galaxies, two objects, 3C293 and 3C305, were associated with a small- scale, 4 and 5 kpc, respectively, radio sources. All of the other 5 blue objects were instead several hundreds of kiloparsec in size. They included also 3C 236, which O’Dea et al. (2001) suggested was just re-ignited radio- source, based on the presence of a small-scale radio emission superimposed on a large-scale emission. For the red galaxies, 3C 353 and 3C 388 have small- scale radio emission (12 and 8 kpc, respectively), while the remaining sources have far larger dimensions. There appears to be no simple relation between the color and the radio emission size for FR II galaxies. Instead, if we separate the sample on the basis of their excitation level using the diagnostic optical line ratios (i.e. into low and high excitation galaxies, LEG and HEG, Laing et al. 1994 and Jackson & Rawlings 1997), the situation is far clearer (see Fig. 9, lower panel). All six high-excitation galaxies are blue and, apart from possibly 3C 192, they all show evidence of recent star formation; the sub-sample of HEG is in fact essentially coincident with the blue UV-clumpy sub-sample. The fraction of star-forming HEG is far in excess of what found by Schawinski et al. (2006) in their sample of quiescent galaxies, which is 20-25 % for objects in the same range of velocity dispersion or galaxy luminosity. Figure 9: The NUV-r color distribution of 3CR galaxies. In the upper panel we point out the FR-morphological separation: the FR II distribution is black, the FR I one is empty and the FR I/II intermediate one is shaded. In the lower panel we point out the excitation level classification: the HEG distribution is black, the FR II-LEG one is shaded and the distribution of the FR I objects is empty (for simplicity we sorted the FR I/II intermediate-LEG objects as FR II-LEG). Conversely, there is only 1 blue LEG (namely 3C 293) out of 19, including all FR I in this class. The substantial tail of blue objects found by Schawinski et al. (their Fig. 1) does not appear to be present in this class of objects, and this strongly disfavours the possibility of enhanced star formation in LEG hosts, although the possible effects of the different environment should also be considered (see the discussion below). Therefore, it appears that a connection between recent star formation and nuclear activity in 3CR hosts is only present in high excitation galaxies. We now consider in more detail the properties and differences between the HEG and LEG classes. Both the UV emission and dust structures in HEG suggest that these galaxies underwent a recent major merger and the highly chaotic and unsettled morphologies imply an external origin for the dust (Tremblay et al. 2007). The amount of dust derived by de Koff et al. (2000), using optical absorption maps, and Müller et al. (2004), from ISO and IRAS observations, in HEG galaxies is also quantitatively larger than in LEG (see Fig. 10). This indicates that a significant amount of gas is provided by the cannibalized object and appears to correspond to a “wet” merger. Conversely, with one exception, we found no evidence of recent star formation in the LEG galaxies. Dust is observed in many of these galaxies, although only in the form of circum-nuclear disks. While the formation mechanisms of these structures are still largely unknown, Temi et al. (2007) proposed that the accumulation of cold dust in the central regions of massive ellipticals is due to mass loss from red giant stars in the galaxy core and is therefore of internal origin. Their optical and UV morphologies are also relaxed and lack clear signs of a recent merger. Colina & de Juan (1995) found evidence of subtle isophotal disturbances in FR I galaxies, which could represent “dry” mergers at a late stage. Figure 10: The NUV-r color vs. dust mass (de Koff et al. 2000 and Müller et al. 2004) of the 3CR sample. Note that the HEG galaxies have dust content greater than the other objects. Symbols as in Fig. 8. Figure 11: Left panel: absolute magnitude MR vs NUV-r color. Right panel: stellar velocity dispersion vs NUV-r color. Symbols as in Fig. 8. In terms of nuclear activity, HEG are associated to brighter AGN than LEG. Within our sample, HEG have a median [O III] luminosity555 Narrow emission- line luminosities, of in particular the [O III] line, provide robust orientation-independent measures of the intrinsic AGN luminosity (e.g. Mulchaey et al. 1994). higher by a factor of 30 with respect to LEG (Chiaberge et al. 2002). This is reminiscent of the luminosity segregation between high and low excitation AGN found by Kewley et al. (2006), since they find that the median [O III] luminosity for LINERs and Seyfert galaxies (the radio-quiet analogous to LEG and HEG) differ by a factor of 16, Seyfert galaxies being brighter. We note that Kewley et al. (2006) also found that LINERs are older, more massive, less dusty, and have higher velocity dispersions than Seyfert galaxies, which is analogous to differences found between LEG and HEG. In low luminosity radio galaxies or LEGs from a spectroscopic point of view, the results by Allen et al. (2006); Balmaverde et al. (2008) indicate that the accretion rate from the hot interstellar medium is proportional to the kinetic output of their jets. This is an indication that this mode of accretion (self- sustained by the interstellar medium) provides a sufficient energy input to power their jets. The higher AGN luminosity of HEG suggests that for these objects there is the need of gas of external origin to satisfy their energy requirements. Another element of difference between HEG and LEG is their environment. It is well known that FR I are located in regions of higher galaxy density, often at the center of galaxy cluster, than FR II (Zirbel 1997). Furthermore, among the FR II radio-galaxies, LEG are often found in clusters of galaxies (Chiaberge et al. 2000), while HEG usually inhabit groups. The different galactic environments can lead, at least from a statistical point of view, to the different star formation manifestations between HEG and LEG. In fact, in groups of galaxies, the low relative velocity between objects and the higher fraction of gas-rich galaxies increase the probability of major “wet” merger events, whereas in cluster of galaxies, mostly “dry” mergers occur. This favors the conditions to trigger, via a “wet” merger, a powerful AGN and significant star formation in the group galaxies and explains the different environments between HEG and LEG. The picture that emerges, considering i) NUV-r color, ii) UV and optical morphology, iii) dust content, iv) nuclear luminosity, and v) environment is that a recent major “wet” merger is needed to trigger the relatively powerful AGN and copious star formation associated with the HEG. Conversely, in LEG galaxies the fraction of actively star-forming objects is not enhanced with respect to that observed in quiescent galaxies. The accretion of the interstellar medium also provides a sufficient energy input to power LEGs, without the need for an external gas supply. These results add to the list of differences between high and low excitation radio-galaxies, which include the presence, only in HEG, of a prominent Broad Line Region, an obscuring torus, and a radiatively efficient accretion disk. All of these features are linked to their availability of sufficiently high amounts of gas to form these structures. It is also possible that HEG and LEG are linked by an evolutionary sequence. A FR I (or LEG) galaxy might become a HEG when affected by a merger; at a later stage, when the fresh gas is exhausted in the galaxy by star formation and accretion onto the SMBH, the galaxy may revert to a quiescent LEG state. We now compare our results with previous studies of the recent star formation in radio-galaxies based the optical spectroscopy. Unfortunately, it is difficult to perform a robust comparison of the fraction of actively star- forming galaxies and its relation to the AGN properties. This is, in part, due to the different selection criteria of the samples, the subjective definition of the various sub-groups of radio galaxies, and the lack of classification into the HEG and LEG classes for many objects studied in the literature. Similarly, an object-by-object comparison of the level of AGN and nebular continuum is not conclusive. There is a strong band mismatch between our UV data and the B band, where the contamination estimates are usually made. Furthermore, the ground-based spectroscopy provides information that can be matched reasonably only for our central aperture, and not for all galaxies. In many cases we also excluded the central regions, where AGN contamination is probably stronger, which are instead included in the ground-based data. Nonetheless, the levels of nebular continuum contamination from spectroscopic analysis are comparable to our photometric estimates and is tipically less 10% (see Table 3). Finally, we note that, quite reassuringly, the two independent approaches converge to the same result for all blue 7 objects in common. We conclude the discussion with a further comparison with previous studies of the NUV-r colors derived from the analysis of GALEX and SDSS images of early- type galaxies. Schawinski et al. (2006) found that the fraction of RSF galaxies decreases with increasing stellar velocity dispersion. They explained this trend to be an AGN-feedback effect powered by supermassive black holes that halts star formation in the most massive galaxies. A similar effect is observed for our 3CR sample (see Fig. 11, right panel), which appear to show a connection between the stellar velocity dispersion and NUV-r color since the bluest objects have preferentially low stellar velocity dispersion ($\sigma$ $\mathchar 13358\relax$ 250 km/s), while the red galaxies cover the entire range of $\sigma$. However, a closer inspection shows that 4 of the galaxies with the highest values of $\sigma$ are cD galaxies located at the center of clusters, while one (namely 3C 270) is the dominant member of a rich group. Apparently, the zone of avoidance (there are no blue galaxies at large $\sigma$) is mostly driven by an environmental effect rather than a different level of AGN feedback, and a blue NUV-r color is a manifestation of a recent merger. A quantitative comparison, however, is not straightforward due to the relatively small size of our sample and the cross-calibration uncertainties, and would require the selection of quiescent galaxies only in a rich environment. The study of Kauffmann et al. (2006) was instead based on a volume-limited sample of massive bulge-dominated galaxies, which also included AGN. Their GALEX-SDSS color profiles demonstrated that the UV excess light is almost always associated with an extended disk for galaxies with young bulges and strongly accreting black holes. They suggested that the presence of an extended gas structure is a necessary condition for AGN activity. Unfortunately, the small field-of-view of the STIS images used in our study does not allow us to investigate directly whether low surface brightness UV disks are associated with our radio-galaxies. Furthermore, the Kauffmann et al. (2006) sample cannot be used as a reliable reference for our 3CR sources, since it is dominated by galaxies of lower $\sigma$ and contains mostly radio- quiet AGN. ## 5 SUMMARY AND CONCLUSIONS We have analyzed images of 3CR radio galaxies for which both optical and UV HST images are available. The sample is composed of 31 radio-galaxies, all but one with z $<$ 0.1, representing $\sim$ 60 % of all 3CR sources below this redshift limit. We have excluded six objects, four are very highly nucleated and two with data of too low signal-to-noise ratio, and have 25 remaining objects. To perform a rigorous comparison with previous studies based on GALEX (NUV band) and SDSS (r band) observations, we derived a cross-calibration between GALEX, SDSS, and HST colors. We conservatively estimated that the error associated with this procedure was $\mathchar 13358\relax$ 0.2 mag. On the basis of the integrated colors and UV morphology, the galaxies of the sample can be divided into 3 main categories: (1) the quiescent red galaxies (14 objects) have “red” colors, with only diffuse UV emission, tracing the optical light; (2) the “blue” UV-clumpy galaxies (7 objects) show a clumpy UV morphology (extended over 5-20 kpc); (3) the UV-disky galaxies (4 objects) have UV emission that is cospatial with their circum-nuclear (on a scale of 0.5-1 kpc) dusty disks. To recognize the presence of recent star formation, it is necessary to account for other sources of UV emission. We estimated the level of UV contamination from emission lines and nebular continuum. In UV-disky galaxies, we found that UV emission lines contributed $\sim$ 50 - 80 % of the total UV flux within the region coincident with the dusty disks, causing the observed blue NUV-r color. For the 7 UV-clumpy galaxies, the emission line or nebular continuum contamination were both negligible. There was a spatial association between UV light and radio-axis only in one galaxy, which is a possible sign of contamination from scattered nuclear light. Summarizing, we confirm the presence of a young stellar population in at least 6 of the 7 UV-clumpy galaxies, in full agreement (on an object-by-object comparison) with results obtained for ground-based optical spectroscopy. The NUV-r color is not simply related to the radio power since, while the bluest galaxies have all relatively large values of P178MHz, at the high end of radio luminosities we found many of the reddest galaxies. Concerning the FR type, we found no blue galaxy among the FR I sources, but among the FR II (or transiction FRI/II) galaxies there was an almost equal number of “blue” and “red” galaxies. The clearer association between galaxy color and AGN type is related to the optical spectroscopic classification into HEG and LEG: all six HEG are blue, while there is only 1 blue LEG out of 19 including all FR I into this class. The fraction of star forming HEG is far in excess of that found by Schawinski et al. (2006) for their sample of quiescent galaxies, while in LEG there appear to be even less star-forming galaxies than in quiescent galaxies. These results can be summarized as follows: 1. 1. While all HEG have a blue UV-r color, the opposite is true for LEG, with only one exception; 2. 2. UV, optical, and dust morphology of HEG are highly chaotic and correspond to unsettled morphologies. In LEG, the UV emission is typically diffuse and traces the optical light (leaving aside the UV-disky galaxies contaminated by line emission), while dust (when present) is mostly arranged in small circum- nuclear disk structures; 3. 3. HEG have a higher dust content; 4. 4. HEG have higher nuclear luminosity. All of these findings can be explained if for HEG we are seeing the effects of a recent, major, “wet” merger. The fresh input of gas and dust causes both the higher star formation rate and stronger nuclear activity. The most favorable situation for the occurrence of a “wet” merger is in groups of galaxies. This can explain the different environment between HEG, usually located in groups, and LEG, more often found in clusters of galaxies. Conversely, in LEG we did not find evidence for recent star formation (with only one exception). A quantitative comparison with the star-formation fraction of quiescent galaxies is not straightforward due to the relatively small size of our sample and the cross-calibration uncertainties. However, we note that the substantial tail of blue objects found in inactive early-type galaxies is apparently not present for our objects, and this strongly disfavours the possibility of enhanced star formation in LEG hosts. It appears that there is no clear connection between the recent star formation and the presence of low excitation AGN. This is in agreement with the idea that in LEG the hot interstellar medium is able to substain the AGN activity via quasi-spherical accretion. This process does not require an external cold gas supply that might become detectable in a star formation event. The results presented here demonstrate the potential of these studies for investigating the triggering mechanism of nuclear activity and star formation in radio galaxies, a method that can be adopted also for other classes of AGN. However, this approach can not provide a quantitative estimate of the star formation history, an essential ingredient in the study of the coupling between the growth of galaxies and SMBH. This crucial information can only be derived by complementing the existing UV images with further data, e.g. optical and UV spectra, that can be used to constraints the age of the recently formed stellar population. ###### Acknowledgements. We would like to thank the anonymous referee, Clive Tadhunter, and Marco Chiaberge for their very useful comments and suggestions. The authors acknowledge partial financial support by PRIN - INAF 2006 grant. ## References * Allen et al. (2002) Allen, M. G., Sparks, W. B., Koekemoer, A., et al. 2002, ApJS, 139, 411 * Allen et al. (2006) Allen, S. W., Dunn, R. J. H., Fabian, A. C., Taylor, G. B., & Reynolds, C. S. 2006, MNRAS, 372, 21 * Aretxaga et al. (2001) Aretxaga, I., Terlevich, E., Terlevich, R. J., Cotter, G., & Díaz, Á. I. 2001, MNRAS, 325, 636 * Balmaverde et al. (2008) Balmaverde, B., Baldi, R. D., & Capetti, A. 2008, A&A, 486, 119 * Barnes & Hernquist (1996) Barnes, J. E. & Hernquist, L. 1996, ApJ, 471, 115 * Barnes & Hernquist (1991) Barnes, J. E. & Hernquist, L. E. 1991, ApJ, 370, L65 * Baum et al. (1992) Baum, S. A., Heckman, T. M., & van Breugel, W. 1992, ApJ, 389, 208 * Brotherton et al. (1999) Brotherton, M. S., van Breugel, W., Stanford, S. A., et al. 1999, ApJ, 520, L87 * Bruzual & Charlot (2003) Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000 * Burstein et al. (1988) Burstein, D., Bertola, F., Buson, L. M., Faber, S. M., & Lauer, T. R. 1988, ApJ, 328, 440 * Buttiglione et al. (2008) Buttiglione, S., Capetti, A., Baldi, R., et al. 2008, A&A, submitted * Canalizo & Stockton (2001) Canalizo, G. & Stockton, A. 2001, ApJ, 555, 719 * Capetti et al. (1999) Capetti, A., Axon, D. J., Macchetto, F. D., Marconi, A., & Winge, C. 1999, ApJ, 516, 187 * Cardelli et al. (1989) Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 * Chiaberge et al. (1999) Chiaberge, M., Capetti, A., & Celotti, A. 1999, A&A, 349, 77 * Chiaberge et al. (2000) Chiaberge, M., Capetti, A., & Celotti, A. 2000, A&A, 355, 873 * Chiaberge et al. (2002) Chiaberge, M., Capetti, A., & Celotti, A. 2002, A&A, 394, 791 * Code & Welch (1979) Code, A. D. & Welch, G. A. 1979, ApJ, 228, 95 * Colina & de Juan (1995) Colina, L. & de Juan, L. 1995, ApJ, 448, 548 * Dasyra et al. (2007) Dasyra, K. M., Tacconi, L. J., Davies, R. I., et al. 2007, ApJ, 657, 102 * de Koff et al. (1996) de Koff, S., Baum, S. A., Sparks, W. B., et al. 1996, ApJS, 107, 621 * de Koff et al. (2000) de Koff, S., Best, P., Baum, S. A., et al. 2000, ApJS, 129, 33 * Di Matteo et al. (2005) Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604 * Djorgovski & Davis (1987) Djorgovski, S. & Davis, M. 1987, ApJ, 313, 59 * Donzelli et al. (2007) Donzelli, C. J., Chiaberge, M., Macchetto, F. D., et al. 2007, ApJ, 667, 780 * Dopita et al. (1997) Dopita, M. A., Koratkar, A. P., Allen, M. G., et al. 1997, ApJ, 490, 202 * Dressler et al. (1987) Dressler, A., Lynden-Bell, D., Burstein, D., et al. 1987, ApJ, 313, 42 * Emonts et al. (2006) Emonts, B. H. C., Morganti, R., Tadhunter, C. N., et al. 2006, A&A, 454, 125 * Fanaroff & Riley (1974) Fanaroff, B. L. & Riley, J. M. 1974, MNRAS, 167, 31P * Ferland & Osterbrock (1986) Ferland, G. J. & Osterbrock, D. E. 1986, ApJ, 300, 658 * Ferrarese & Merritt (2000) Ferrarese, L. & Merritt, D. 2000, ApJ, 539, L9 * Gebhardt et al. (2000) Gebhardt, K., Bender, R., Bower, G., et al. 2000, ApJ, 539, L13 * González Delgado et al. (1998) González Delgado, R. M., Heckman, T., Leitherer, C., et al. 1998, ApJ, 505, 174 * Graham & Driver (2007) Graham, A. W. & Driver, S. P. 2007, ApJ, 655, 77 * Graham et al. (2001) Graham, A. W., Erwin, P., Caon, N., & Trujillo, I. 2001, ApJ, 563, L11 * Häring & Rix (2004) Häring, N. & Rix, H.-W. 2004, ApJ, 604, L89 * Heckman et al. (1997) Heckman, T. M., Gonzalez-Delgado, R., Leitherer, C., et al. 1997, ApJ, 482, 114 * Heckman et al. (1986) Heckman, T. M., Smith, E. P., Baum, S. A., et al. 1986, ApJ, 311, 526 * Holt et al. (2007) Holt, J., Tadhunter, C. N., González Delgado, R. M., et al. 2007, MNRAS, 381, 611 * Hopkins et al. (2006) Hopkins, P. F., Hernquist, L., Cox, T. J., et al. 2006, ApJS, 163, 1 * Hopkins et al. (2007a) Hopkins, P. F., Hernquist, L., Cox, T. J., & Keres, D. 2007a, ArXiv e-prints, 706 * Hopkins et al. (2007b) Hopkins, P. F., Hernquist, L., Cox, T. J., Robertson, B., & Krause, E. 2007b, ArXiv Astrophysics e-prints * Hutchings et al. (2003) Hutchings, J. B., Maddox, N., Cutri, R. M., & Nelson, B. O. 2003, AJ, 126, 63 * Jackson & Rawlings (1997) Jackson, N. & Rawlings, S. 1997, MNRAS, 286, 241 * Jahnke et al. (2004) Jahnke, K., Kuhlbrodt, B., & Wisotzki, L. 2004, MNRAS, 352, 399 * Jogee (2006) Jogee, S. 2006, in Lecture Notes in Physics, Berlin Springer Verlag, Vol. 693, Physics of Active Galactic Nuclei at all Scales, ed. D. Alloin, 143–+ * Kauffmann & Haehnelt (2000) Kauffmann, G. & Haehnelt, M. 2000, MNRAS, 311, 576 * Kauffmann et al. (2006) Kauffmann, G., Heckman, T. M., Budavari, T., et al. 2006, ArXiv Astrophysics e-prints * Kauffmann et al. (2003) Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MNRAS, 346, 1055 * Kellermann et al. (1969) Kellermann, K. I., Pauliny-Toth, I. I. K., & Williams, P. J. S. 1969, ApJ, 157, 1 * Kewley et al. (2006) Kewley, L. J., Groves, B., Kauffmann, G., & Heckman, T. 2006, MNRAS, 372, 961 * Kormendy & Gebhardt (2001) Kormendy, J. & Gebhardt, K. 2001, in American Institute of Physics Conference Series, Vol. 586, 20th Texas Symposium on relativistic astrophysics, ed. J. C. Wheeler & H. Martel, 363–+ * Kormendy & Richstone (1995) Kormendy, J. & Richstone, D. 1995, ARA&A, 33, 581 * Laing et al. (1994) Laing, R. A., Jenkins, C. R., Wall, J. V., & Unger, S. W. 1994, in The First Stromlo Symposium: The Physics of Active Galaxies. ASP Conference Series, Vol. 54, 1994, G.V. Bicknell, M.A. Dopita, and P.J. Quinn, Eds., p.201, 201–+ * Lee et al. (2005) Lee, Y.-W., Ree, C. H., Rich, R. M., et al. 2005, ApJ, 619, L103 * Letawe et al. (2007) Letawe, G., Magain, P., Courbin, F., et al. 2007, MNRAS, 378, 83 * Lilly & Longair (1984) Lilly, S. J. & Longair, M. S. 1984, MNRAS, 211, 833 * Magorrian et al. (1998) Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285 * Marconi & Hunt (2003) Marconi, A. & Hunt, L. K. 2003, ApJ, 589, L21 * Martel et al. (1999) Martel, A. R., Baum, S. A., Sparks, W. B., et al. 1999, ApJS, 122, 81 * Martel et al. (2000) Martel, A. R., Turner, N. J., Sparks, W. B., & Baum, S. A. 2000, ApJS, 130, 267 * McLure & Dunlop (2002) McLure, R. J. & Dunlop, J. S. 2002, MNRAS, 331, 795 * Melnick et al. (1997) Melnick, J., Gopal-Krishna, & Terlevich, R. 1997, A&A, 318, 337 * Mihos & Hernquist (1994) Mihos, J. C. & Hernquist, L. 1994, ApJ, 431, L9 * Mihos & Hernquist (1996) Mihos, J. C. & Hernquist, L. 1996, ApJ, 464, 641 * Mulchaey et al. (1994) Mulchaey, J. S., Koratkar, A., Ward, M. J., et al. 1994, ApJ, 436, 586 * Mulchaey et al. (1996) Mulchaey, J. S., Wilson, A. S., & Tsvetanov, Z. 1996, ApJS, 102, 309 * Müller et al. (2004) Müller, S. A. H., Haas, M., Siebenmorgen, R., et al. 2004, A&A, 426, L29 * O’Dea et al. (2001) O’Dea, C. P., Koekemoer, A. M., Baum, S. A., et al. 2001, AJ, 121, 1915 * Osterbrock (1989) Osterbrock, D. E. 1989, Astrophysics of Gaseous Nebulae and Active Galactic Nuclei (Mill Valley: Univ. Sci. Books) * Proffitt (2006) Proffitt, C. R. 2006, in The 2005 HST Calibration Workshop: Hubble After the Transition to Two-Gyro Mode, ed. A. M. Koekemoer, P. Goudfrooij, & L. L. Dressel, 234–+ * Raimann et al. (2005) Raimann, D., Storchi-Bergmann, T., Quintana, H., Hunstead, R., & Wisotzki, L. 2005, MNRAS, 364, 1239 * Richstone et al. (1998) Richstone, D., Ajhar, E. A., Bender, R., et al. 1998, Nature, 395, A14+ * Sánchez et al. (2004) Sánchez, S. F., Jahnke, K., Wisotzki, L., et al. 2004, ApJ, 614, 586 * Sanders & Mirabel (1996) Sanders, D. B. & Mirabel, I. F. 1996, ARA&A, 34, 749 * Sanders et al. (1988a) Sanders, D. B., Soifer, B. T., Elias, J. H., et al. 1988a, ApJ, 325, 74 * Sanders et al. (1988b) Sanders, D. B., Soifer, B. T., Elias, J. H., Neugebauer, G., & Matthews, K. 1988b, ApJ, 328, L35 * Schawinski et al. (2006) Schawinski, K., Khochfar, S., Kaviraj, S., et al. 2006, Nature, 442, 888 * Schweizer (1998) Schweizer, F. 1998, in Saas-Fee Advanced Course 26: Galaxies: Interactions and Induced Star Formation, ed. R. C. Kennicutt, Jr., F. Schweizer, J. E. Barnes, D. Friedli, L. Martinet, & D. Pfenniger, 105–+ * Silk (1977) Silk, J. 1977, ApJ, 211, 638 * Smith & Heckman (1989) Smith, E. P. & Heckman, T. M. 1989, ApJ, 341, 658 * Smith et al. (1990) Smith, E. P., Heckman, T. M., & Illingworth, G. D. 1990, ApJ, 356, 399 * Somerville et al. (2000) Somerville, R. S., Lemson, G., Kolatt, T. S., & Dekel, A. 2000, MNRAS, 316, 479 * Springel et al. (2005) Springel, V., Di Matteo, T., & Hernquist, L. 2005, MNRAS, 361, 776 * Tadhunter et al. (2002) Tadhunter, C., Dickson, R., Morganti, R., et al. 2002, MNRAS, 330, 977 * Tadhunter et al. (2005) Tadhunter, C., Robinson, T. G., González Delgado, R. M., Wills, K., & Morganti, R. 2005, MNRAS, 356, 480 * Tadhunter et al. (1996) Tadhunter, C. N., Dickson, R. C., & Shaw, M. A. 1996, MNRAS, 281, 591 * Tadhunter et al. (1989) Tadhunter, C. N., Fosbury, R. A. E., & Quinn, P. J. 1989, MNRAS, 240, 225 * Temi et al. (2007) Temi, P., Brighenti, F., & Mathews, W. G. 2007, ApJ, 660, 1215 * Tremaine et al. (2002) Tremaine, S., Gebhardt, K., Bender, R., et al. 2002, ApJ, 574, 740 * Tremblay et al. (2007) Tremblay, G. R., Chiaberge, M., Donzelli, C. J., et al. 2007, ArXiv e-prints, 705 * Vanden Berk et al. (2006) Vanden Berk, D. E., Shen, J., Yip, C.-W., et al. 2006, AJ, 131, 84 * Wills et al. (2004) Wills, K. A., Morganti, R., Tadhunter, C. N., Robinson, T. G., & Villar-Martin, M. 2004, MNRAS, 347, 771 * Wills et al. (2002) Wills, K. A., Tadhunter, C. N., Robinson, T. G., & Morganti, R. 2002, MNRAS, 333, 211 * Zakamska et al. (2006) Zakamska, N. L., Strauss, M. A., Krolik, J. H., et al. 2006, AJ, 132, 1496 * Zirbel (1997) Zirbel, E. L. 1997, ApJ, 476, 489	arxiv-papers	2008-08-11T18:50:37	2024-09-04T02:48:57.231647	{ "license": "Public Domain", "authors": "Ranieri D. Baldi (1) and Alessandro Capetti (2) ((1) Universita' degli\n Studi di Torino, Italy (2) INAF - Osservatorio Astronomico di Torino, Italy)", "submitter": "Ranieri Diego Baldi", "url": "https://arxiv.org/abs/0808.1555" }
0808.1618	# Quantum computation with graphene nanoribbon Guo-Ping Guo gpguo@ustc.edu.cn Zhi-Rong Lin Xiao-Peng Li Tao Tu tutao@ustc.edu.cn Guang-Can Guo Key Laboratory of Quantum Information, University of Science and Technology of China, Chinese Academy of Sciences, Hefei 230026, People’s Republic of China ###### Abstract We propose a scalable scheme to implement quantum computation in graphene nanoribbon. It is shown that electron or hole can be naturally localized in each zigzag region for a graphene nanoribbon with a sequence of Z-shaped structure without exploiting any confined gate. An one-dimensional graphene quantum dots chain is formed in such graphene nanoribbon, where electron or hole spin can be encoded as qubits. The coupling interaction between neighboring graphene quantum dots is found to be always-on Heisenberg type. Applying the bang-bang control strategy and decoherence free subspaces encoding method, universal quantum computation is argued to be realizable with the present techniques. ###### pacs: 03.67.Lx, 03.67.Pp, 42.50.Dv, 03.67.Bg Electron spin is one of the leading candidates for the realization of a practical solid qubit Loss1998 . The coherent manipulation of electron spins in GaAs quantum dots has been efficiently realized Petta ; Koppens2006 . However due to the interaction with the environment, the decoherence time is often in nanoseconds scale in GaAs quantum dots Hanson ; Petta ; Koppens2005 . Even by applying the complex technique to prepare nuclear state, the dephasing time for spin qubits is just about 1$\mu s$ Reilly . The decoherence is one of the most challenges in the way to quantum computer in GaAs quantum dots. Due to the weak spin-orbit coupling and hyperfine interactions in carbon, graphene is argued to be an excellent candidate for quantum computation Loss2007 . However, due to the special band structure of graphene Neto , the low-energy quasiparticles in graphene behave as Dirac fermions, and the Klein tunneling and Chiral effect lead to the fact that it is non-trivial to form good quantum dot (localized electron states) in graphene. It has been shown that the massless Dirac fermions in graphene can ben confined by using suitable transverse states in graphene nanoribbons (GNR) Loss2007 ; Silvestrov , by combining single and bilayer regions of graphene Nilsson ; Peeters or by using inhomogeneous magnetic fields Martino . Recently, there was an experiment report that GNR with well defined zigzag or armchair edge structures can be chemically produced HJDai . It has also been discovered that localized states exist in the zigzag region in Z-shaped GNR Wang . Here we present a scalable quantum computation scheme based on Z-shaped GNR quantum dot system without exploiting any confined gates. The localized particle can be chosen to be electron or hole by adjusting the back gate even in the room temperature. The qubit is encoded on the electron (hole) spin states localized in the zigzag region of the GNR with a sequence of Z-shaped structure. The interaction between qubits is determined by the GNR geometrical structure and found to be Heisenberg form. By exploiting bang-bang (BB) control strategy and decoherence free subspaces (DFSs) encoding method, universal quantum gates are shown to be realizable in this system with the present techniques. Based on the $\pi$ orbital tight-binding model, the local density of state (LDOS) and the band structure of the zigzag region in a GNR with a sequence of Z-shaped structure can be obtained by the direct diagonalization of the single particle Hamiltonian $H_{0}=\sum_{ij}{\tau_{ij}}\|i\rangle\langle j\|$, where the hopping matrix element $\tau_{ij}=-\tau$ if the orbits $\|i\rangle$ and $\|j\rangle$ are nearest neighboring on the honeycomb lattice, otherwise $\tau_{ij}=0$ Brey ; Nakada . From the calculated band structure, we can see that there are several localized states with electron-hole symmetry around the zero energy point as shown in Fig. 1a. Thus we can choose to get one localized electron or hole in the zigzag region by adjusting the Fermi level through the back gate. The electron ground state energy and the energy gap between the ground state and the first excitation state are very sensitive to the size of the zigzag region, as shown in Fig 2. It has been known that the width of the armchair GNR ($N$ unit cells) decides whether the system is metallic or semiconducting Brey ; Nakada . If $N=3m-1$ ($m$ being an integer), the armchair GNR is metallic, otherwise it is semiconducting. In addition, for the present Z-shaped structure the boundaries along the ribbon of armchair region is unsymmetrical when $N$ is even. Actually, in our calculation we find there is no confined state in the zigzag region of Z-shaped GNR when $N=3m-1$ or $N=2m$ as shown in Fig. 2. On the other hand, when $N$ is $7$ and the length of the zigzag region $L$ (unit cells) (see Fig. 1c) is $3,4,5,6,7$ and $N=9$, $L=3,4$ both the ground level and energy gap are above $0.1$eV. Thus we can confine electron (hole) to form quantum dot even in the room temperature. Figure 1: (Color online) Schematic of the proposed architecture of GNR for quantum computation. (a) The Z-shaped GNR quantum dot can localize one electron (left figure) or hole (right figure) in the zigzag region by adjusting the Fermi level through the back gate. (b) Local density of states of GNR with two Z-shaped structure in series. (c) A GNR with two Z-shaped structure in series, each zigzag region confines one electron. The physical qubit is encoded into the spin of the confined electron. (d) Special encoding method to eliminate the interaction between logical qubits. Physical qubits $1$ and $2$ form logical qubit $L_{1}$; physical qubits $3$ and $4$ form logical qubit $L_{2}$; physical qubits $5$ and $6$ form logical qubit $L_{3}$. The $G^{z}$ , $G^{x}$, $G^{y}$ are the BB operation sets of $L_{1}$, $L_{2}$ and $L_{3}$ respectively. Figure 2: (Color online) The ground energy level ($E_{0}$) and energy gap ($E_{1}-E_{0}$) between the ground state and the first excitation state of the Z-shaped GNR quantum dot device with different width of nanoribbon ($N$ unit cells) and different length of quantum dot region ($L$ unit cells). Fig. 1b shows the spatial distribution of local density of ground state for a GNR with two Z-shaped structure in series. Each zigzag region (quantum dot) confines one electron and the quantum dots are coupled by the exchange coupling $J_{1}$. We can obtain $J_{1}$ by calculating the exchange integral $J_{1}=\int\varphi_{1}^{\ast}(\vec{r_{1}})\varphi_{2}^{\ast}(\vec{r_{2}})\frac{e^{2}}{\|\vec{r_{1}}-\vec{r_{2}}\|}\varphi_{1}(\vec{r_{2}})\varphi_{2}(\vec{r_{1}})d\vec{r_{1}}d\vec{r_{2}}$, where $\varphi_{1}(\vec{r})$ and $\varphi_{2}(\vec{r})$ are the wavefunction of neighboring graphene quantum dots. We can also calculate the next nearest neighboring exchange coupling $J_{2}$ by the same method. Obviously, the exchange coupling $J_{1}$, $J_{2}$ are determined by the geometrical structure of the nanoribbon. For each $N$ and $L$, $J_{1}$ and $J_{1}/J_{2}$ depend on the number of unit cells ($D$) between two neighboring qubits. By numerical calculations, $J_{1}$, $J_{1}/J_{2}$ are obtained with different $N$, $L$, $D$, as shown in Fig. 3. For $N=7$, $L=6$, $D=18$, $J_{1}=8\mu$eV, $J_{1}/J_{2}=10^{5}$, we can safely neglect this non-nearest neighboring qubits coupling. For clarity, in the following discussion we focus on the atomic structure with $N=7$, $L=6$, $D=18$. Figure 3: (Color online) (a) The coupling energy $J_{1}$ of two nearest neighboring qubits dependence on the number of unit cells between two qubits is presented for different size of quantum dot region. (b) The ratio of $J_{1}$ to the next nearest neighboring exchange energy $J_{2}$ depend on the distance $D$ of two neighboring qubits for different size of quantum dot region. To carry out quantum computation, we have to form the logical qubit and realize universal quantum gate. It has been shown that single qubit rotations combined with two-qubit operations can be used to create basic quantum gates DiVincenzo . The spin of the localized electron or hole can be used as the physical qubit and the GNR with a sequence of Z-shaped structure forms an one- dimensional qubit chain as shown in Fig. 1d. The neighboring qubits in this chain have an always-on Heisenberg interaction $H=J_{1}\vec{S_{1}}\cdot\vec{S_{2}}$. Here $\vec{S_{1}}$ and $\vec{S_{2}}$ are the spin operator of the neighboring localized electron (hole). It has been known that BB control strategy and DFSs encoding method do not require directly controlling the interaction between qubits Benjamin ; Zhang . The quantum information in qubits can be protected from decoherence induced by the environment and undesired disturbance induced by the inherent qubit-qubit interaction with these strategies. For a sequence of Z-shaped structure GNR with $N=7$, $L=6$, $D=18$, the Hamiltonian of the system can be expressed as $H_{I}=\sum_{i,j}J_{i,j}(\sigma_{i}^{x}\otimes\sigma_{j}^{x}+\sigma_{i}^{y}\otimes\sigma_{j}^{y}+\sigma_{i}^{z}\otimes\sigma_{j}^{z}),$ (1) where $\sigma_{i,j}^{x,y,z}$ are the spin Pauli operators of the localized electron (hole) in the quantum dots, $i$ and $j$ represent two neighboring dots. Here we have neglected the interaction between non-neighboring dots, which has been shown to be $5$ orders smaller than the neighboring interaction. To avoid the spin qubits to entangle with the environment, we can apply a BB operation $U_{z}=exp(-i\sigma_{z}\pi/2)$ to each quantum dot region. Such rotation operations can be realized if a pulsed magnetic field could be applied exclusively Loss1998 . To counteract phase decoherence, we can use DFSs encoding Duan . For a simply DFSs encoding, two physical qubits can encode a logical qubit: $\|0\rangle_{L}=\|\uparrow_{1}\downarrow_{2}\rangle,\|1\rangle_{L}=\|\downarrow_{1}\uparrow_{2}\rangle.$ (2) As shown in Fig. 1c, we use localized electron in the two neighboring zigzag regions to form a logical qubit. In order to protect quantum information in the logical qubits, we must decouple the always-on Heisenberg interaction between two physical qubits within a logical qubits and interaction between two neighboring logical qubits. A nonsynchronous BB pulse operations and a special encoding method can be exploited to eliminate these interactions Zhang . Here we propose an architecture in which the one-dimensional GNR chain form a periodic structure $L_{1}L_{2}L_{3}L_{1}L_{2}L_{3}$ $\cdots$ with three logical qubits as a unit, as shown in Fig. 1d. $L_{1}$ represents a logical qubit encoded as Eq.(2). $L_{2}$ is a logical qubit encoded as $\|0\rangle_{L2}=\frac{1}{2}(\|\uparrow\rangle_{3}+\|\downarrow\rangle_{3})(\|\uparrow\rangle_{4}-\|\downarrow\rangle_{4}),$ (3) $\|1\rangle_{L2}=\frac{1}{2}(\|\uparrow\rangle_{3}-\|\downarrow\rangle_{3})(\|\uparrow\rangle_{4}+\|\downarrow\rangle_{4}).$ (4) And $L_{3}$ is a logical qubit encoded as $\|0\rangle_{L3}=\frac{1}{2}(\|\uparrow\rangle_{5}+i\|\downarrow\rangle_{5})(\|\uparrow\rangle_{6}-i\|\downarrow\rangle_{6}),$ (5) $\|1\rangle_{L3}=\frac{1}{2}(\|\uparrow\rangle_{5}-i\|\downarrow\rangle_{5})(\|\uparrow\rangle_{6}+i\|\downarrow\rangle_{6}).$ (6) With this periodic architecture, we have to apply nonsynchronous BB pluse operations respectively to $L_{1}$, $L_{2}$, $L_{3}$ from the operation set $G^{z}=\\{I,U_{z},R_{z}\\}$, $G^{x}=\\{I,U_{x},R_{x}\\}$, $G^{y}=\\{I,U_{y},R_{y}\\}$, where $U_{z}=-\sigma_{1}^{z}\otimes\sigma_{2}^{z}$, $R_{z}=-iI_{1}^{z}\otimes\sigma_{2}^{z}$, $U_{x}=-\sigma_{1}^{x}\otimes\sigma_{2}^{x}$, $R_{x}=-iI_{1}\otimes\sigma_{2}^{x}$, $U_{y}=-\sigma_{1}^{y}\otimes\sigma_{2}^{y}$, and $R_{y}=-iI_{1}\otimes\sigma_{2}^{y}$. Then we obtain a quantum computation system with entirely decoupled logical qubits. Now we show how to carry out universal quantum gates of the logical qubits defined above. Logical operations $\bar{X}$ and $\bar{Z}$ can generate all SU(2) transformations of logical qubit. For logical qubit $L_{1}$, $\bar{X}=\frac{1}{2}(\sigma_{1}^{x}\otimes\sigma_{2}^{x}+\sigma_{1}^{y}\otimes\sigma_{2}^{y})$, $\bar{Z}=\frac{1}{2}(\sigma_{1}^{z}-\sigma_{2}^{z})$. $\bar{X}$ can be easily achieved by recoupling qubits $1$ and $2$ by adjusting the BB pulses of both qubits to be synchronous Zhang . The operation time can be obtained by $J\Delta t=\hbar\pi/4$ , for $N=7$, $L=6$ , $D=18$, $\Delta t=0.2$ns. $\bar{Z}$ can be achieved by directly varying the Zeeman splitting on the two physical qubits individually analogous to single-qubit operations in the Loss- DiVincenzo quantum computer Loss1998 . The operation time of this $\bar{Z}$ gate can be about $1$ns when $20$mT magnetic field could be pulsed exclusively onto each quantum dot region. The fidelity of the $\bar{X}$ gate can be effected by the fluctuation or inhomogeneity in the exchange coupling $J_{1}$ between different dots. The charge noise in the back gate can cause the fluctuation of $J_{1}$. The main sources of the $J_{1}$ inhomogeneity between different dots are the disorder, irregular edges and defect of the GNR. Short- range disorder scarcely affects the LDOS of the ground state. The irregular edge effect and long-range disorder change the LDOS but do not destroy the confined states Wang . To get high fidelity for $\bar{X}$ operation, we should avoid the long-range disorder and irregular edge. Actually, if we know the coupling $J_{1}$ between different dots exactly, inhomogeneity of $J_{1}$ can not effect the fidelity of $\bar{X}$ gate when corresponding inhomogeneous operation times are used. In addition, we find the effect of the $J_{1}$ inhomogeneity or fluctuation to the fidelity of the $\bar{X}$ gate is small as shown in Fig. 4. Because the nuclear field would change the evolution of the spin states, the fidelity of the $\bar{Z}$ gate is dominated by the nuclear field Koppens2006 . The fidelity of the $\bar{Z}$ gate can be very high due to small nuclear field in graphene system. Similarly, high fidelity operation $\bar{X}$ and $\bar{Z}$ can be also realized for logical qubits $L_{2}$ and $L_{3}$. Figure 4: The fidelity of $\bar{X}$ gate against the fluctuation or inhomogeneity of the exchange coupling energy $J_{1}$. We can construct CNOT gate between two neighboring logical qubits, for example $L_{1}$ and $L_{2}$, by W gate $W=\|0\rangle\langle 0\|\otimes I+\|1\rangle\langle 1\|\otimes e^{i2\theta\bar{Z}}=e^{i\theta\bar{Z}\otimes\bar{Z}}$ and Hadamard operation Bremner . By performing Hadamard transformation $H=\frac{1}{\sqrt{2}}\left[\begin{array}[]{cc}1&1\\\ 1&-1\end{array}\right],$ (7) to the two physical qubits of the second logical qubit $L_{2}$ and changing the BB control pulse to be the same with $L_{1}$, we can recouple the two neighboring logical qubits and implement W gate of logical qubits of $L_{1}$ and $L_{2}$. For the present graphene quantum dots chain with $N=7$, $L=6$, $D=18$, the total operation time of a CNOT gate can be implemented in about $1$ns with an oscillating magnetic field of $100$mT to achieve the Hadamard operation. Similar to the above discussion for $\bar{X}$ and $\bar{Z}$ operation, we can find that the fluctuation or inhomogeneity of $J_{1}$ and the nuclear field have trivial effect to the fidelity of the CNOT gate in the present protocol. The major decoherence sources of spin qubits in solid state system have been identified as the spin-orbit interaction and hyperfine interaction. The weak spin-orbit coupling have been predicted in carbon material due to the low atomic weight Min . Since the primary component of natural carbon is the zero spin isotope ${}^{12}C$, the very long coherence time given by hyperfine coupling has been theoretically argued Loss2007 . Assuming the abundance of ${}^{13}C$ is about $1\%$ as in the nature carbon material, the decoherence time has been predicted to be more than $10\mu$s in the graphene quantum dot Loss2007 ; Coish . This decoherence time is $4$ orders longer than the gates operation time of the present protocol. In addition, the decoherence time can be much longer if the percentage of ${}^{13}C$ is decreased by isotopic purification. In this paper we have presented a scalable scheme of quantum computation based on GNR with a sequence of Z-shaped structure. No confined gates is needed to localize the particle, which can be chosen to be electron or hole by adjusting back gate. The qubit is encoded in electron or hole spin states, which is naturally localized in the zigzag region of GNR even in room temperature. The neighboring qubits are found to have an always-on Heisenberg interaction and the dynamical decoupling techniques with DFSs is exploited to achieve universal quantum computation in this system. Due to recent achievement in production of graphene nanoribbon, this proposal may be implementable within the present techniques. We thank Prof. Q. W. Shi and Dr. Z. F. Wang for helpful discussions. This work was funded by National Fundamental Research Program, the Innovation funds from Chinese Academy of Sciences, NCET-04-0587, and National Natural Science Foundation of China (Grant No. 60121503, 10574126, 10604052). ## References * (1) D. Loss and D. P. DiVincenzo, Phys. Rev. A 57, 120 (1998). * (2) J. R. Petta, A. C. Johnson, J. M. Taylor, E. A. Laird, A. Yacoby, M. D. Lukin, C. M. Marcus, M. P. Hanson, and A. C. Gossard, Science 309, 2180 (2005). * (3) F. H. L. Koppens, C. Buizert, K. J. Tielrooij, I. T. Vink, K. C. Nowack, T. Meunier, and L. P. Kouwenhoven, Nature 442, 766 (2006). * (4) R. Hanson, L. P. Kouwenhoven, J. R. Petta, S. Tarucha, and L. M. K. Vandersypen, Rev. Mod. Phys. 79, 1217 (2007). * (5) F. H. L. Koppens, J. A. Folk, J. M. Elzerman, R. Hanson, L. H. Willems van Beveren, I. T. Vink, H. P. Tranitz, W. Wegscheider, L. P. Kouwenhoven, L. M. K. Vandersypen, Science 309, 1346 (2005). * (6) D. J. Reilly, J. M. Taylor, J. R. Petta, C. M. Marcus, M. P. Hanson, and A. C. Gossard, Science 321, 817 (2008). * (7) B. Trauzettel, D. V. Bulaev, D. Loss, and G. Burkard, Nature Physics 3, 192 (2007). * (8) A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov and A. K. Geim, arXiv:0709.1163. * (9) P. G. Silvestrov and K. B. Efetov, Phys. Rev. Lett. 98, 016802 (2007). * (10) J. Nilsson, A. H. Castro Neto, F. Guinea, and N. M. Peres, Phys. Rev. B 76, 165416 (2007). * (11) A. Matulis and F. M. Peeters, Phys. Rev. B 77, 115423 (2008). * (12) A. D. Martino, L. Dell’Anna, and R. Egger, Phys. Rev. Lett. 98, 066802 (2007). * (13) X. L. Li, X. R. Wang, S. Lee, and H. J. Dai, Science 319, 1229 (2008). * (14) Z. F. Wang, H. X. Zheng, Q. W. Shi, J. Chen, Q. X. Li, and J. G. Hou, Appl. Phys. Lett. 91, 053109 (2007). * (15) L. Brey and H. A. Fertig, Phys. Rev. B 73, 235411 (2006). * (16) K. Nakada, M. Fujita, G. Dresselhaus, and M. S. Dresselhaus, Phys. Rev. B. 54, 17954 (1996). * (17) D. P. DiVincenzo, Phys. Rev. A 51, 1015 (1995). * (18) S. C. Benjamin, and S. Bose, Phys. Rev. Lett. 90, 247901 (2003). * (19) Y. Zhang, Z. W. Zhou, B. Yu, and G. C. Guo, Phys. Rev. A 69, 042315 (2004). * (20) L. M. Duan, G. C. Guo, Phys. Rev. Lett. 79, 1953 (1997). * (21) M. J. Bremner, C. M. Dawson, J. L. Dodd, A. Gilchrist, A. W. Harrow, D. Mortimer, M. A. Nielsen, and T. J. Osborne, Phys. Rev. Lett. 89, 247902 (2002). * (22) H. Min, J. E. Hill, N. A. Sinitsyn, B. R. Sahu, L. Kleinman, and A. H. MacDonald, Phys. Rev. B 74, 165310 (2006). * (23) W. A. Coish, and D. Loss, Phys. Rev. B 70, 195340 (2004).	arxiv-papers	2008-08-12T08:48:58	2024-09-04T02:48:57.240712	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Guo-Ping Guo, Zhi-Rong Lin, Xiao-Peng Li, Tao Tu, Guang-Can Guo", "submitter": "Tao Tu", "url": "https://arxiv.org/abs/0808.1618" }
0808.1635	# Resonant Transmission of Electromagnetic Fields Through Subwavelength Zero-$\epsilon$ Slits Klaus Halterman Simin Feng Physics Division, Naval Air Warfare Center, China Lake, California 93555 ###### Abstract We theoretically investigate the transmission of electromagnetic radiation through a metal plate with a zero-$\epsilon$ metamaterial slit, where the permittivity tends towards zero over a given bandwidth. Our analytic results demonstrate that the transmission coefficient can be substantial for a broad range of slit geometries, including subwavelength widths that are many wavelengths long. This novel resonant effect has features quite unlike the Fabry-Pérot-like resonances that have been observed in conductors with deep channels. We further reveal that ultranarrow zero-$\epsilon$ channels can have significantly greater transmission compared to slits with no wave impedance difference across them. With the current state of the art in nanofabrication technologies, and the recent observation of resonant optical transmission through small metal holes ebb , there has been a renewed interest in electromagnetic wave transmission and diffraction in metallic nanostructures. Although diffraction effects involving optical slits and gratings has a long history, scientific curiosity coupled with the potential for device applications has spurred considerable research activity involving the manipulation and confinement of electromagnetic waves in nanoscale resonant structures. With the concurrent advent of metamaterials, or composite structures with tailored electromagnetic properties, the interplay of classical transmission systems with these new composite media has become an important issue. By engineering metallic nano- apertures, gratings, or channels to incorporate metamaterials, it is anticipated that various enhanced or exotic transmission characteristics will ensue. Lately, there has been interest in zero-$\epsilon$ metamaterials, which are structures that exhibit an effective permittivity of zero (or nearly so) in the passband, akin to the property of some precious metals palik near their plasma frequency. It was shown that waveguide devices containing zero-$\epsilon$ inclusions could potentially be more efficient via the reduction of unwanted reflections if one of the physical dimensions was made smaller eng . This can possibly negate the ill effects of impedance mismatch. A multilayer structure having nonlinear electric and magnetic responses, and a refractive index close to zero, was shown to effectively shield feng electromagnetic fields. Experimental work has verified liu ; eng3 electromagnetic tunneling through zero-$\epsilon$ metamaterials at microwave frequencies. It was shown eng2 that the effective magnetic permeability, $\mu$, can also be resonantly tuned to vanish, creating a matched metamaterial with an effective zero index of refraction. These media have been linked to applications in miniaturized resonators, highly directive antennas enoch , delay lines with zero-phase difference I/O, and transformers that convert small-curvature wave fronts into output beams with planar-like wave fronts zio . A fundamental system in which to investigate zero-$\epsilon$ diffraction and transmission effects is a subwavelength channel through a metal film. If the slit is filled with air, it has been established that for light polarized perpendicular to the slit in an optically thick perfect metal of length, $l$, Fabry-Pérot-like waveguide modes arise when $l$ is approximately a half integer number of wavelengths: $l\approx n\lambda/2$. These harmonic modes, which follow from geometrical arguments, result in transmission peaks when the waves coherently superimpose over the given path length. As the geometrical parameters vary, the resonant wavelength can shift takakura ; mybad in metals with slit perforations, as observed with microwaves suck0 . For ultrashort incident pulses, the Fabry-Pérot-like modes can be resonantly activated mechler through the Fourier components of the wave packet inside the slit, leading in some cases to enhanced transmission. The resonant enhancement of a nanometer scale electromagnetic pulse was also shown to be spatially and temporally localized mechler2 in the near field. For perfect metals with holes rather than slits, transmission is strictly limited to incident wavelengths that are less than twice the diameter of the openings. In this Letter we reveal some unexpected and exotic transmission behavior of light through a subwavelength zero-$\epsilon$ slit. Our theoretical framework demonstrates that significant transmission can occur in these structures for a considerable range of ultranarrow widths that are sufficiently deep. The inverse relationship between the slit length and width that achieves maximal transmission is shown to be highly nontrivial. We show that zero-$\epsilon$ metamaterials with inherently large intrinsic impedances, can have greater transmission than matched zero index slits, where $\mu$ is also vanishingly small. We also take advantage of the subwavelength geometry (that is, the slit width is smaller than about half the incident wavelength), to invoke the single mode approximation, which has been shown to yield valuable physical insight into resonance phenomena for perfect mata and real gordy metals. Our analytic results thus permit one to efficiently map out the relevant sector of parameter space deemed appropriate for any future experimental endeavors. We consider a planar perfect metallic structure that is translationally invariant in the $x-z$ plane and has channel length $l$ normal to the plane with width $p$ (along $x$). The incident beam is TM polarized, so that the magnetic field ${\bm{H}}$ is directed along $z$. The wavevector, ${\bm{k}}$, forms an angle $\theta$ with the normal to the plane. For this configuration, the $z$ component of the magnetic field, $H_{z}$, must satisfy the scalar Helmholtz equation, $\displaystyle\frac{\partial^{2}H_{z}}{\partial x^{2}}+\frac{\partial^{2}H_{z}}{\partial y^{2}}+\epsilon\mu k_{0}^{2}H_{z}=0,$ (1) where $k_{0}=\omega/c$. Our focus is primarily metamaterial slits very near the characteristic plasma frequency and, as in recent works, represent the frequency dispersive electrical response by an effective Drude model, $\epsilon=1-\omega_{p}^{2}/[\omega(\omega+i\Gamma)]$, where $\Gamma$ is correlated to the mean free path in the filling material. Other than when discussing absorption loss, we generally take $\Gamma$ equal to zero in order to isolate the electromagnetic effects inherent to vanishing constitutive relations. We solve (1) in a given region, and then match each solution at the corresponding interfaces using the appropriate boundary conditions. To construct the solution in the “continuum” regions of free space surrounding the metal sheet, we Fourier transform Eq. (1) along $x$ and use separation of variables. This renders an expression written in terms of plane wave expansions, $\displaystyle H_{z,1}$ $\displaystyle=e^{i(k_{0x}x-k_{0y}y)}+\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\infty}d\alpha R(\alpha)e^{i\alpha x}e^{i\beta y}$ (2) $\displaystyle H_{z,3}$ $\displaystyle=\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\infty}d\alpha T(\alpha)e^{i\alpha x}e^{-i\beta y},$ (3) where $k_{0x}=k_{0}\sin\theta$, $k_{0y}=k_{0}\cos\theta$, and $\beta=\sqrt{k_{0}^{2}-\alpha^{2}}$. Here the subscripts 1 and 3 denote the entrance and exit regions respectively. The unknown coefficients $R(\alpha)$ and $T(\alpha)$ are determined below. The general eigenmode expansion for the field within the slit region is takakura , $\displaystyle H_{z,2}=\sum_{m=0}^{\infty}\cos\bigl{[}k_{m}(x-p/2)\bigr{]}\bigl{[}a_{m}e^{i\tau_{m}y}+b_{m}e^{-i\tau_{m}y}\bigr{]},$ (4) where $k_{m}\equiv m\pi/p$, and $\tau_{m}\equiv\sqrt{\epsilon\mu k_{0}^{2}-k_{m}^{2}}$. The summation is over a single index $m$ reflecting the interconnection among the wavenumbers consistent with the wave equation. The electric field is obtained directly via application of Maxwell’s equation: ${\bm{E}}=i/(\epsilon k_{0})\nabla\times{\bm{H}}$. It is generally valid to retain only the lowest order mode in the expansion (4) for the ultranarrow channels considered in this paper. Moreover, as $\epsilon$ goes to zero, the summation results in higher order evanescent fields that have a negligible contribution to the time-averaged energy flux. Thus, for the electric field we have, $\displaystyle E_{x,2}=\eta(-ae^{i\tau y}+be^{-i\tau y})\zeta(x),$ (5) where $\tau\equiv\tau_{0}$, $\eta=\sqrt{\mu/\epsilon}$ is the intrinsic impedance, and the function $\zeta(x)$ is unity in the slit and vanishes elsewhere. Matching the electric field at the entrance and exit interfaces, yields the $R$ and $T$ coefficients: $\displaystyle R(\alpha)$ $\displaystyle=\frac{\sqrt{2\pi}k_{0y}\delta(k_{0x}-\alpha)}{\beta}+\frac{w}{\beta}(a-b)\Phi(\alpha p),$ (6a) $\displaystyle T(\alpha)$ $\displaystyle=\frac{w}{\beta}(-ae^{-i\tau l}+be^{i\tau l})e^{-i\beta l}\Phi(\alpha p),$ (6b) where we define $\Phi(x)\equiv\sqrt{2/\pi}\sin(x/2)/x$. A convenient technique mata that utilizes the boundary conditions to determine the remaining unknown $a$ and $b$ coefficients is the use of the following relationships at the appropriate openings: $\langle H_{z,1}(x,0),\zeta(x)\rangle=\langle H_{z,2}(x,0),\zeta(x)\rangle$, and $\langle H_{z,2}(x,-l),\zeta(x)\rangle=\langle H_{z,3}(x,-l),\zeta(x)\rangle$. After some tedious algebra, we have for $b$, $\displaystyle b=\frac{2\sqrt{2\pi}\Phi(k_{0x}p)(1+wI_{0})}{(1+wI_{0})^{2}-e^{2i\xi}(1-wI_{0})^{2}},$ (7) where $w\equiv k_{0}p\eta$, $\xi\equiv\tau l$, and $a=be^{2i\xi}(wI_{0}-1)/(wI_{0}+1)$. The complex valued integral, $I_{0}$ is a function solely of $q$: $\displaystyle I_{0}$ $\displaystyle=\frac{1}{\pi q^{2}}\int_{0}^{\infty}du\frac{\sin^{2}(uq)}{u^{2}\sqrt{1-u^{2}}},$ (8) where $q\equiv k_{0}p/2$ is the dimensionless measure characterizing the slit width. The series expansion for the imaginary component of $I_{0}$ has the following leading terms: $\displaystyle{\rm Im}[I_{0}]\approx-\frac{1}{6\pi}\Bigl{[}(q^{2}-6)(\gamma+\ln q)+9-\frac{19}{12}q^{2}\Bigr{]},$ (9) where $\gamma$ is Euler’s constant. Similarly for the real part of $I_{0}$, we have (to fourth order), ${\rm Re}[I_{0}]\approx\frac{1}{2}-\frac{1}{12}q^{2}+\frac{1}{120}q^{4}$. It will be seen shortly that $I_{0}$ is a very important quantity in the determination of the transmission. Figure 1: Real and imaginary components of $I_{0}$ as a function of the dimensionless width $q\equiv\pi p/\lambda$. The approximate solutions and exact integral [Eq. (8)] have good agreement over the relevant range of widths. As $q$ vanishes, the imaginary component has a slow divergence [see Eq. (9)]. The accuracy of these expansions is demonstrated graphically in Fig. 1, where the truncated expansions are plotted alongside the numerically integrated $I_{0}$ [Eq. (8)]. Clearly, the approximations for $I_{0}$ are satisfactory for the small $q$ of interest here, deviating only slightly for the larger $q$, as would be expected for a power series centered about the origin. Next, we insert the calculated $a$ and $b$ coefficients into Eq. (5), to determine the electric field at the bottom of the slit, $\displaystyle E^{\rm bot}_{x}=\frac{2i\sin(q\sin\theta)\eta}{q\sin\theta[\sin\xi+4I_{0}q\eta_{2}(i\cos\xi+I_{0}q\eta_{2}\sin\xi)]},$ (10) which is valid for arbitrary $\mu$ and $\epsilon$. The magnetic field is easily obtained via the relationship, $H^{\rm bot}_{z}=2qI_{0}E^{\rm bot}_{x}$. The transmission coefficient, $\tau$, is defined as the time- averaged Poynting vector at the bottom of the slit, $\langle S_{y}\rangle_{y=-l}$, integrated over the exit opening, and divided by the Poynting vector of the incident plane wave: $\displaystyle\tau=\frac{16q^{2}{\rm Re}[I_{0}]\|\eta\|^{2}{\cal F}(\theta)}{\|\sin\xi+4I_{0}q\eta(i\cos\xi+I_{0}q\eta\sin\xi)\|^{2}},$ (11) which is dimensionless after incorporating a normalization factor, $k_{0}$. The angular dependence is encapsulated entirely by the expression, ${\cal F}(\theta)={\rm sinc}^{2}(q\sin\theta)/\cos\theta$, which depends weakly on $q$ for subwavelength slits, and is close to unity for source waves near normal incidence. For small $q$, we can further write (for $\sin\xi\neq 0$): $\tau\approx 8(\mu/\epsilon)q^{2}\csc^{2}(k_{0}l\sqrt{\epsilon\mu})$, otherwise if we consider channels with some integer multiple of the Fabry- Pérot length, $l=\lambda/(2\sqrt{\epsilon\mu})$, Eq. (11) approximately reduces to, $\tau\approx\frac{1}{2}\|I_{0}\|^{-2}$, which clearly only depends on the ratio of the slit width to the wavelength [see Eq. (8)]. Having derived the general expression for transmission of electromagnetic fields through a subwavelength channel containing, to this point, conventional material parameters, we now examine the effects of taking the limit of vanishing $\epsilon$. It is relatively straightforward to show that Eq. (10) for the electric field at the slit exit now reduces to the succinct form, $\displaystyle E^{\rm bot}_{x}=\sqrt{\frac{\pi}{2}}\frac{\Phi(2q\sin\theta)}{qI_{0}(1-iI_{0}k_{0}lq\mu)},$ (12) giving a transmission of, $\displaystyle\tau=\frac{{\rm Re}[I_{0}]}{\|I_{0}\|^{2}\bigl{\|}1-iI_{0}k_{0}lq\mu\bigr{\|}^{2}},$ (13) where $H_{z}=2qI_{0}E^{\rm bot}_{x}$, and is spatially constant in the slit. It can be deduced from Faraday’s Law, ${\bm{H}}=-i/(\mu k_{0})\nabla\times{\bm{E}}$, that $E_{x}$ must therefore be a linear function of the coordinate $y$ within the slit: $E_{x}=(y/l)\delta E_{x}+E_{x}^{\rm top}$, where $\delta E_{x}\equiv E_{x}^{\rm top}-E_{x}^{\rm bot}$. The electric field at the top of the slit, $E_{x}^{\rm top}$, is simply, $E_{x}^{\rm top}=(1-2iqI_{0}k_{0}l\mu)E_{x}^{\rm bot}$. For $q\ll 1$, we can write Eq. (13) strictly in terms of elementary functions, $\displaystyle\tau\approx\frac{2\pi[\pi+k_{0}l\mu q(3-2\gamma-2\ln q)]}{4\ln q(2\gamma-3+\ln q)+\pi^{2}+(2\gamma-3)^{2}},$ (14) demonstrating that for a given ratio of slit width to wavelength, $\tau$ is simply a linear function of the channel length $l$ and permeability $\mu$. Figure 2: Transmission, $\tau$, through a zero-$\epsilon$ slit as a function of the dimensionless geometrical parameters, $l/\lambda$ and $p/\lambda$. To illustrate how the transmission depends upon the geometry of the slit, we present in Fig. 2, a three-dimensional representation of $\tau$ [Eq. (13)] as a function of the slit dimensions, $l$ and $p$, scaled by the wavelength. In this figure, the source field is normally incident with wavelength, $\lambda=5\mu{\rm m}$, which is very close to the resonant plasma wavelength, so that the frequency dispersive $\epsilon$ nearly vanishes. It is evident from the plot that the transmission does not possess Fabry-Pérot-like resonant oscillations as a function of $l$, as would be expected from a slit with conventional material. The figure shows that energy flow is generally restricted unless the geometrical parameters coincide with the brightly peaked regions that can decay quite rapidly. Figure 3: Transmission characteristics for a zero-$\epsilon$ metamaterial slit, demonstrating the $l$ and $p$ that give the optimal transmission at a given wavelength. Brighter regions indicate higher transmission. Relatively high transmission can persevere in deep slits ($l\gg\lambda$) and small openings ($p\ll\lambda$). The wave impedance, $Z$, defined as $Z=E_{x}/H_{z}$, is calculated from the field expressions above to yield the following impedance relation between the top and bottom of the openings, $Z^{\rm top}=Z^{\rm bot}-ik_{0}l\mu$, with $Z^{\rm bot}=1/(2qI_{0})$. A remarkable property of a narrow zero-$\epsilon$ slit is that despite the considerable wave impedance for $p\ll\lambda$, energy can still be transmitted if the slit length $l$ is increased. This behavior is further illustrated in Fig. 3, where the transmission is shown to be robust for an extended range of channel lengths and widths. The maximum transmission, $\tau_{\rm max}$, seen contained within the continuous bright curve, is determined by examining the extrema of Eq. (13). By taking the appropriate derivative of the denominator, we find that $\tau_{\rm max}$ is determined by the transcendental equation, $\mu qk_{0}l=\gamma(q)$, where $\gamma(q)=-{\rm Im}[I_{0}]/\|I_{0}\|^{2}$ is a gradually increasing monotonic function of $q$. Thus, as a function of $q$, and for each $l$, the intercept of $\gamma(q)$ with lines of slope $\mu k_{0}l$ yields the permissible electrical slit widths that achieve peak light transmission. This technique is shown graphically in the inset of Fig. 4, where the opposite relationship between slit width and length are clearly seen. Figure 4: Transmission through deep metamaterial slits as a function of $p/\lambda$. A considerable resonance peak emerges for a small subwavelength width of zero-$\epsilon$ material. Also shown is a matched zero index medium where $\mu$ and $\epsilon$ are both nearly zero. The inset shows the intersections of the lines with the $\gamma(q)$ curve, corresponding to electrical widths that yield peak transmission. Differing ratios of $l/\lambda$ (see legend) yield lines with different slopes. These results are consistent with the main plot, where the channel length corresponds to $l=50\lambda$. It is also of interest to determine if the transmission can be enhanced for light incident upon a a matched zero index material, whereby both $\epsilon$ and $\mu$ are vanishingly small. In the main plot of Fig. 4, the transmission is therefore plotted as a function of slit width for a zero-$\epsilon$ slit (with $\mu=1$ as usual), and matched zero index slit. Surprisingly, for zero-$\epsilon$ media, the shown subwavelength widths support transmission resonances that are absent in matched zero index slits. Although the $l$ dependence washes out for matched zero index channels, $\tau$ still depends strongly on the width dimension, and Eq. (13) approximately reduces to the expression for a slit filled with a conventional dielectric and length satisfying the Fabry-Pérot geometric resonance condition. We can thus conclude that for zero-$\epsilon$ media and narrow enough openings, energy flow can be much greater than for slits loaded with metamaterials having no wave impedance mismatch between the ends of the slit. We also considered the effects of finite $\Gamma$, and concluded that the presence of absorption loss simply reduces the overall magnitude of $\tau$ by a factor that becomes greater with increasing slit depth, but having no effect on the geometrical parameters leading to the particular resonance location. This bodes well for future development of miniaturized devices used for energy transport and narrow band frequency selective surfaces. In conclusion, we have shown that significant transmission arises for a range of subwavelength widths in sufficiently deep slits containing zero-$\epsilon$ inclusions. We also demonstrated that these resonant channels can transmit greater energy compared to the corresponding matched zero index slits. Although there is currently limited metamaterial fabrication in the IR and optical frequencies, some recent progress has been made with negative-index metamaterials involving metal layers separated by a dielectric ref11 , alternating layers of InGaAs and AlInAs semiconductors ref12 , and photonic crystals in the near-infrared range ref13 . ###### Acknowledgements. This work is supported by NAVAIR s ILIR program sponsored by ONR and by a grant of HPC resources from ARSC at the University of Alaska, Fairbanks, as part of the DOD HPCMP. ## References * (1) T.W. Ebbesen, et al., Nature (London) 391, 667 (1998). * (2) E. D. Palik, Handbook of Optical Constants of Solids (Academic Press, Washington, DC, 1985). * (3) M. Silveirinha and N. Engheta, Phys. Rev. Lett. 97, 157403 (2006); Phys. Rev. B76 245109 (2007). * (4) S. Feng and K. Halterman, Phys. Rev. Lett. 100, 063901 (2008). * (5) R. Liu, Q. Cheng, T. Hand, and J.J. Mock, Phys. Rev. Lett. 100, 023093 (2008). * (6) B. Edwards, et al., Phys. Rev. Lett. 100, 033903 (2008). * (7) M. Silveirinha and N. Engheta, Phys. Rev. B75, 075119 (2007). * (8) S. Enoch, et al., Phys. Rev. Lett. 89, 213902 (2002). * (9) R.W. Ziolkowski, Phys. Rev. E70, 046608 (2004). * (10) Y. Takakura, Phys. Rev. Lett. 86, 5601 (2001). * (11) J. Bravo-Abad, L. Martin-Moreno, and F.J. Garcia-Vidal, Phys. Rev. E69, 026601 (2004). * (12) F. Yang and J.R. Sambles, Phys. Rev. Lett. 89, 063901 (2002). * (13) M. Mechler, O. Samek, and S.V. Kukhlevsky, Phys. Rev. Lett. 98, 163901 (2007). * (14) S.V. Kukhlevsky, et al., Phys. Rev. B70, 195428 (2004). * (15) O. Mata-Mendez and J. Avendano, J. Opt. Soc. Am. A. 24, 1687 (2007). * (16) R. Gordon, Phys. Rev. B73, 153405 (2006). * (17) S. Zhang et al., Phys. Rev. Lett. 95, 137404 (2005). * (18) A. J. Hoffman et al., Nat. Mater. 6, 946 (2007). * (19) R. Chatterjee et al., Phys. Rev. Lett. 100, 187401 (2008).	arxiv-papers	2008-08-12T15:20:07	2024-09-04T02:48:57.244721	{ "license": "Public Domain", "authors": "Klaus Halterman and Simin Feng", "submitter": "Klaus Halterman", "url": "https://arxiv.org/abs/0808.1635" }
0808.1734	# Observation of a continuous phase transition in a shape-memory alloy J. C. Lashley Los Alamos National Laboratory, Los Alamos, NM 87545, USA S. M. Shapiro Brookhaven National Laboratory, Upton, NY 11973, USA B. L. Winn Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA C. P. Opeil Boston College, Department of Physics, 140 Commonwealth Avenue, Chestnut Hill, MA 02467, USA M. E. Manley Lawrence Livermore National Laboratory, Livermore, CA 94550, USA A. Alatas Advanced Photon Source, Argonne National Laboratory, Argonne, IL 60439, USA W. Ratcliff National Institute of Standards and Technology, Gaithersburg, MD 20899, USA T. Park Los Alamos National Laboratory, Los Alamos, NM 87545, USA Department of Physics, Sungkyunkwan University, Suwon 440-746, Korea R. A. Fisher Los Alamos National Laboratory, Los Alamos, NM 87545, USA B. Mihaila Los Alamos National Laboratory, Los Alamos, NM 87545, USA P. Riseborough Department of Physics, Temple University, Philadelphia, PA 19122, USA E. K. H. Salje Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EQ, UK J. L. Smith Los Alamos National Laboratory, Los Alamos, NM 87545, USA ###### Abstract Elastic neutron-scattering, inelastic x-ray scattering, specific-heat, and pressure-dependent electrical transport measurements have been made on single crystals of AuZn and Au0.52Zn0.48 above and below their martensitic transition temperatures ($T_{M}$ = 64 K and 45 K, respectively). In each composition, elastic neutron scattering detects new commensurate Bragg peaks (modulation) appearing at $Q=(1.33,0.67,0)$ at temperatures corresponding to each sample’s $T_{M}$. Although the new Bragg peaks appear in a discontinuous manner in the Au0.52Zn0.48 sample, they appear in a continuous manner in AuZn. Surprising us, the temperature dependence of the AuZn Bragg peak intensity and the specific-heat jump near the transition temperature are in favorable accord with a mean-field approximation. A Landau-theory-based fit to the pressure dependence of the transition temperature suggests the presence of a critical endpoint in the AuZn phase diagram located at $T_{M}^{}$ = 2.7 K and $p^{}$ = 3.1 GPa, with a quantum saturation temperature $\theta_{s}$ = 48.3$\pm$3.7 K. ###### pacs: 81.30.Kf, 71.20.Be, A class of materials exhibiting martensitic (diffusionless) phase transformations yields properties used in a range of technological applications including implants to increase flow in restricted blood vessels Duerig , actuators for the treatment of high myopia Yu , voltage generators Suorsa , and orthodontic arch-wires Jafari . These properties often depend on the history of the material and may allow it to recover its previous shape after deformation, known as the shape-memory effect (SME). It has long been recognized that these transformations are all thermodynamically first order (discontinuous) Albers ; Boyd ; Gash . Special cases arise when the order parameter is coupled to an external field in a complicated way, leading to a weakly first-order transition Rubini , which is believed to result from a complicated coupling between strain and order parameter fluctuations. Because shape-memory alloys and other multiferroic materials owe their functionality to complicated cross-field responses between two (or more) pairs of conjugate thermodynamic variables, a description of the free-energy landscape and the ability to predict functionality become one and the same. Phenomenological descriptions of the martensitic transformation pathway, based on reciprocal space Bowles and real space David geometries, have established rules to determine relative twin and crystallographic orientations between the austenite (high-temperature) and martensite (low-temperature) modifications. Knowledge of the symmetry breaking allows for a definition of the energetic driving force in terms of the difference in free energies between phases. In the Ginzburg-Landau (GL) approach, the free energy, $G$, is expressed as a sum of symmetry invariants. In its simplest form, $G$ is approximated by a polynomial expansion in even powers of an order parameter, $Q$, as $G(Q)=aQ^{2}+bQ^{4}+cQ^{6}+\frac{g}{2}\|{\nabla Q}\|^{2}$. For martensitic transformations, $Q$ is taken to be strain or strain coupled to shuffle (displacements involving quasi-static phonons with fractional commensurate wave vectors with uniform shears Burgers ). The coefficients $a,b,c$, and $g$ are material parameters to be determined experimentally or by first principle simulations. For the majority of shear-induced transformations, the GL expansion captures the essential physics Toledano ; Salje , such as constitutive response Avadh1 ; Iwata and the occurrence of anti-phase boundaries Cao . When applied to shape-memory alloys, the presence of intervening (premartensitic) phases Sethna presents difficulties for both experiment and theory as to a unique characterization of the order parameter. Further issues have arisen in materials exhibiting quantum mechanical effects in the band structure and strong electron-phonon coupling Avadh2 . The thermodynamic behavior is best explored when no further transitions occur between TM and 0 K. This effect was recently reported for Fe-doped NiTi, where all thermodynamic parameters could be correlated self-consistently within mean field theory Toni . In order to integrate these cases into GL theory, it is necessary to measure and access each thermodynamic property and infer cross-field couplings of the order parameter. The ability to distinguish subtle differences between first-order and weakly first-order depends on the ability to resolve the behavior of pertinent physical properties. In this Letter we examine the AuZn system for thermodynamic properties that contribute to the first-order or weakly first-order nature of the free-energy landscape. We measure elastic neutron-scattering, inelastic x-ray scattering, specific-heat, and the pressure dependence of the transition by electrical transport. While we observed classical first-order behavior at 45 K for Au0.52Zn0.48, we observed a mean field continuous transformation for AuZn near 64 K in the temperature dependence of the Bragg peak intensity, and a Landau step with a $\lambda$ anomaly in the excess specific heat. Figure 1: (a) The excess specific heat divided by temperature versus temperature in the vicinity of the martensitic transitions for Au0.52Zn0.48 and AuZn. (b) Temperature dependence of the satellite-peak intensity along $Q=(1.33,0.67,0)$. The satellite peak intensity is proportional to the square of the order parameter. Figure 2: Temperature dependence of the satellite peak intensity on heating along $Q=(1.33,0.67,0)$ in AuZn, near the transition temperature. A fit to mean-field theory for $T\leq T_{c}$. The points above 64 K are background. Single crystals were prepared by fusion of the elements in a Bridgman furnace and were oriented by back-reflection Laue diffraction. The neutron experiments were performed on the BT9 triple-axis spectrometer at the NIST research reactor. Measurements of the phonon dispersion were made on XOR of sector 3 at the Advanced Photon Source, Argonne National Laboratory Sinn1 ; Sinn2 . Specific-heat measurements were measured using a thermal-relaxation calorimeter by Quantum Design Lashley . The pressure dependence of the transition temperature was made by a four-terminal ac-transport method in a mechanical pressure cell designed to reach pressures of 3 GPa. Figure 3: Temperature dependence of the phonon dispersion in AuZn along the TA2 branch. At 300 K one notices considerable phonon softening at $q=(0.33,0.33,0)$ (vertical line). The dashed lines between points are provided as guides to the eye. Figure 4: Inelastic x-ray scattering data showing the temperature dependence near the soft mode, $q$ = (0.35, 0.35, 0). The 300 and 200 K data are offset by 0.004 and 0.0005, respectively on the y-axis for clarity. The energy positions are fit to a double Lorentzian (solid curve) at 300 K to determine the phonon energy. The intrinsic resolution for inelastic x-ray scattering is 2 meV. Figure 5: (a) The electrical-resisitvity data taken on cooling and warming is shown in the vicinity of $T_{M}$. (b) Plot of $ln~{}p$ versus $1/T$ to test the applicability Clausius-Clapeyron equation. Previous investigations have shown that AuZn exhibits a martensitic transition as seen with low-temperature electron microscopy Ted and by recoverable transformation strain (shape-memory effect) Darling . Figure 1 shows the temperature dependence of (a) the excess specific heat and (b) the satellite peak intensity for Au0.52Zn0.48 and AuZn. Although we have measured the specific heat previously Ross , it is plotted along with the elastic neutron- scattering data, since it is revealing to compare both sets of measurements that were made on the same samples. We find a latent heat anomaly at 45 K for Au0.52Zn0.48. A $\lambda$-anomaly was observed for AuZn, but no latent heat was observed. The specific heat near the transition can be influenced by intrinsic defects, including triple-point and anti-structure defects, both known to be present in the B2 structure Austin ; Gupta . It was therefore necessary to use a microscopic probe to determine whether or not the transition could be a broadened first-order transition. Elastic neutron- scattering measurements indicate that in each composition new commensurate Bragg peaks (modulation) appear at $Q=(1.33,0.67,0)$ at temperatures corresponding to each sample’s $T_{M}$ Figure 1(b). Mirroring the specific- heat data, the temperature dependence of the satellite peak intensity shows a rapid jump at 45 K in Au0.52Zn0.48, while a continuous variation occurs for AuZn on warming starting at 64 K. The satellite peak intensity is proportional to the square of the order parameter, since it leads to the low-temperature rhombohedral phase. The specific-heat data also indicate thermal hysteresis (not shown) of 6 K at 45 K for Au0.52Zn0.48. The satellite peak intensity for AuZn depicted in Fig. 1(b) was measured by neutrons on heating and cooling. Unlike the specific-heat data Fig. 1(a) we observe a small discontinuity and 1.5 K difference in transition temperature on cooling. Figure 2 shows the satellite peak intensity versus temperature in the vicinity of $T_{M}$ for AuZn. Near the transition temperature, the satellite peak evolves continuously with increasing temperature, showing a temperature dependence given by $I=A\left\|{T-T_{M}}\right\|^{2\beta};$ where the satellite peak intensity is $I$, the critical temperature is $T_{M}$, and the critical exponent is $\beta$. We obtain TM=64.08 K $\pm$0.11 K and 2$\beta$= 1.02$\pm$0.08. The value of the critical exponent obtained from this fit ($\beta$=0.51) is close to the mean-field value of $\beta$=0.50 Binney . One could argue that mean-field exponents were obtained by coincidence given that the critical regime, as defined by the Ginzburg-criterion, is restricted to a range of temperatures too small to be resolved by our experiments, or that the correlation length does not diverge, since it is limited by other extrinsic disorder (i.e., martensite twins) or intrinsic fluctuations. The latter possibility seems more plausible to us, because the inverse correlation length, or width of the Bragg peaks (not shown), increases as the temperature is lowered below $T_{M}$. As the temperature is lowered, the unit cell is modulated in the [110] shear direction. A commensurate shuffle of every third unit cell results in a hexagonal primitive unit cell formed from nine primitive cubic cells of the parent phase. This structure can also be described in terms of its conventional rhombohedral unit cell. In Fig. 3 we show the TA2 phonon- dispersion curves along [110], measured by inelastic x-ray scattering at temperatures of 300 K and 200 K. The inflection in the phonon frequency near $\xi=1/3$ indicates the low shear instability along the TA2[110] branch. This inflection is comparable to an earlier investigation Makita of off- stoichiometric AuZn samples. At 300 K, the phonon energy positions have been fit over the energy range -4 meV $\leq E\leq$ 4 meV to a double Lorentzian, as shown in Figure 4. The phonons continuously soften with decreasing temperature to the point where separation becomes difficult below 200 K. Figure 5 (a) shows the change in resistivity between cooling and warming. The data were collected with warming and cooling rates of 0.2 K min-1. In the region of $T_{M}$ there is no hysteresis, to within experimental error. Because of the lack of hysteresis, we elected to measure the pressure dependence of the transition. For a first-order transition, the Clausius- Clapeyron relation, $\ln p/p_{0}=-\Delta H/(RT)+A$, predicts a linear relationship for $\ln p/p_{0}$ versus $1/T$ with the slope $\left(-\Delta H/R\right)$. Here, $p$ denotes the pressure, $H$ the enthalpy, $R$ the universal gas constant, and $p_{0}$ is a reference pressure. Figure 5(b) shows the plot of $\ln p$ versus $1/T$. It appears that there is no observable linear region. Assuming that the entropy does not vary significantly with pressure, the conditions for the applicability of the Clausius-Clapeyron relation are not met. This result provides further evidence that the transition is not first order. Figure 6: Pressure dependence of the transition temperature up to pressures of nearly 3 GPa, together with a fit to Salje’s model Qsat which accounts for quantum saturation effects close to zero absolute temperature. The difference between the linear (classical) and quantum-dominant behavior is illustrated in the inset. We find that quantum effects become important for temperatures lower than $\theta_{s}$ = 48.3$\pm$3.7 K, and a critical endpoint is predicted at $T_{M}^{}$ = 2.7 K and $p^{}$ = 3.1 GPa. We use a modified GL free energy expression to describe the quantum saturation in the order parameter expected at sufficiently low temperature Qsat . This saturation effect reflects the departure from classical behavior as the absolute temperature approaches zero and is a direct consequence of the third law of thermodynamics. Following Salje _et al._ Qsat , we fit the pressure dependence of the transition as $\theta_{s}/T_{M}(p)=\coth^{-1}\bigl{[}\coth(\theta_{s}/T_{M}(0))-p/p_{0}\bigr{]}$. Here $\theta_{s}$ is a phenomenological temperature below which quantum mechanical effects are dominant. We find $p_{0}$ = 5.6$\pm$0.7 GPa and $\theta_{s}$=48.3$\pm$3.7 K. The resulting phase diagram predicts a critical endpoint at $T_{M}^{}$ = 2.7 K and $p^{}$ = 3.1 GPa. Figure 6 depicts the experimental pressure dependence of the transition temperature for pressures up to nearly 3 GPa, together with the fit to Salje’s classical regime. We find a classical slope of -11.7$\pm$0.2 K GPa-1. The difference between the classical and quantum regimes is important for temperatures below $\theta_{s}$. In conclusion, we report results of a detailed experimental study of the martensitic transition in AuZn. While the temperature dependence of the satellite peak intensity (proportional to the square of the order parameter) show marked first-order behavior at 45 K for Au0.52Zn0.48, we observed a continuous feature well-described by a mean-field exponent ($\beta$=0.51) for AuZn at 64 K. This result is contrary to the established definition of a martensitic transition. Providing supporting evidence to the continuous nature of the phase transition in AuZn are the lack of thermal hysteresis from electrical-resistivity measurements, the $\lambda$-anomaly in the specific heat, and the disagreement of pressure data with the Clausius-Clapeyron relation. Using Salje’s model to describe the quantum saturation effects close to zero absolute temperature, we predict the presence of a critical endpoint in the phase diagram located at $T_{M}^{}$ = 2.7 K and $p^{}$ = 3.1 GPa, with a phenomenological temperature $\theta_{s}$=48.3$\pm$3.7 K for the onset of quantum effects. ###### Acknowledgements. This work was performed under the auspices of the United States Department of Energy and Department of Commerce and supported in part by the Trustees of Boston College. Work at Brookhaven is supported by the Office of Science, United States Department of Energy under Contract No. DE-ACO2-98CH10886. The use of the Advanced Photon Source was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. ## References * (1) P. W. Duerig, MRS Bulletin 27, 101 (2002). * (2) I. Yu et al., Mater. Sci. Eng. A 481-482, 651 (2008). * (3) I. Suorsa et al., J. Appl. Phys. 95, 8054 (2004). * (4) J. Jafari, S. M. Zebarjad, and S. A. Sajjadi, Mater. Sci. Eng. A. 473, 42 (2008). * (5) R. C. Albers et al., Comp. and Appl. Math. 23, 345 (2004). * (6) J. G. Boyd and D. C. Lagoudas, Int. J. Plasticity 12, 805 (1996). * (7) P. W. Gash, J. Appl. Phys. 54, 6900 (1983). * (8) S. Rubini and P. Ballone, Phys. Rev. B. 48, 99, (1993). * (9) J. S. Bowles and J. K. Mackenzie, Act. Met. 2, 129 (1954). * (10) M. S. Wechsler, D. S. Lieberman, and T. A. Read, J. of Metals 5, 1503 (1953). * (11) W.G. Burgers, Physica Utrecht 1, 561 (1934). * (12) P. Toledano and P. Toledano, The Landau Theory of Phase Transformations (World Scientific, Singapore, 1987). * (13) E. K. H. Salje, Phase Transformations in Ferrelastic and Co-elastic Solids (Cambridge University Press, Cambridge, UK, 1990). * (14) R. Ahluwalia et al., Acta. Mat. 52, 209 (2004). * (15) M. Iwata and Y. Ishibashi, J. Phys. Soc. Jap. 72, 2843 (2003). * (16) W. Cao, A. Saxena, and D. M. Hatch, Phys. Rev. B. 64, 024106 (2001). * (17) S. Kartha et al., Phys. Rev. Lett. 67, 3630 (1991). * (18) K. H. Ahn et al., Phys. Rev. B. 71, 212102 (2005). * (19) E. K. H. Salje et al., J. Phys. Cond. Matt., 20, 275216 (2008). * (20) H. Sinn et al., Nucl. Inst. and Meth. in Phys. Res. A 467, 1545 (2001). * (21) H. Sinn, J. Phys. Cond. Matter 13, 7525 (2001). * (22) J. C. Lashley et al., Cryogenics 43, 369 (2003). * (23) H. Pops and T. B. Massalski, Trans. Met. Soc. AIME 223, 728 (1965). * (24) T. Darling et al., Phil. Mag. B. 82, 825 (2002). * (25) R. McDonald et al., J. Phys. Cond. Mat. 17, L69 (2005). * (26) Y. A. Chang and J. P. Neumann, Prog. Sol. St. Chem. 14, 221 (1985). * (27) D. Gupta and D. S. Lieberman, Phys. Rev. B. 4, 1070 (1971). * (28) J. J. Binney, N, J, Dowrick, A. J. Fisher, and M. E. J. Newman, Theory of Critical Phenomena (Oxford Science Publications, Oxford University Press Inc., New York, 1993). * (29) T. Makita et al., Phys. B. 213, 430 (1995). * (30) S. A. Hayward and E. K. H. Salje, J. Phys.: Condens. Matter 10, 1421 (1998); J. M. P$\acute{e}$rez-Mato and E. K. H. Salje, ibid. 12, L29 (2000).	arxiv-papers	2008-08-12T20:31:55	2024-09-04T02:48:57.249798	{ "license": "Public Domain", "authors": "J. C. Lashley, S. M. Shapiro, B. L. Winn, C. P. Opeil, M. E. Manley,\n A. Alatas, W. Ratcliff, T. Park, R. A. Fisher, B. Mihaila, P. Riseborough, E.\n K. H. Salje, and J. L. Smith", "submitter": "Bogdan Mihaila", "url": "https://arxiv.org/abs/0808.1734" }
0808.1777	# Spin-charge Separation in Nodal Antiferromagnetic Insulator Su-Peng Kou Department of Physics, Beijing Normal University, Beijing, 100875 P. R. China ###### Abstract In this paper, by using two dimensional (2D) Hubbard models with $\pi$-flux phase and that on a hexagonal lattice as examples, we explore spin-charge- separated solitons in nodal antiferromagnetic (AF) insulator - an AF order with massive Dirac fermionic excitations (see detail in the paper). We calculate fermion zero modes and induced quantum numbers on solitons (half skyrmions) in the continuum limit, which are similar to that in the quasi one- dimensional conductor polyacetylene (CH)x and that in topological band insulator. In particular, we find some novel phenomena : thanks to an induced staggered spin moment, a mobile half skyrmion becomes a fermionic particle; when a hole or an electron is added, the half skyrmion turns into a bosonic particle with charge degree of freedom only. Our results imply that nontrivial induced quantum number on solitons may be a universal feature of spin-charge separation in different systems. PACS numbers : 71.10.Fd, 75.10.Jm, 75.40.Gb The Fermi liquid based view of the electronic properties has been very successful as a basis for understanding the physics of conventional solids. The quasi-particles of Fermi liquid carry both spin and charge quantum numbers. However, in some cases, spin-charge separation occurs, providing a new framework for thinking about the given systems. It indicates that the systems have two independent elementary excitations, neutral spinon and spinless holon, respectively, as opposed to single quasi-particle excitation in conventional solids. The first example is electronic systems in one spatial dimensionle . The idea of solitons with induced quantum numbers starts with a beautiful result obtained in the context of relativistic quantum field theories by Jackiw and Rebbijackiw . Based on this idea, spin-charge separated solitons had a lasting impact on condensed matter physics. In the long molecule-chain of trans- Polyacetylene, spin-charge separation can occur in term of soliton statespolrev . Due to induced fermion quantum numbers, the soliton may be neutral particles with spin 1/2, or spinless with charge $\pm e$. In two dimensional electronic systems, spin-charge separation has been a basic concept in understanding doped Mott-Hubbard insulator related to high-$T_{c}$ cuprates pwa ; KRS . It is supposed that the particles can be liberated at low energies, with spin-charge separation being an upshot in the “resonating valence bond” (RVB) spin liquid state which is proposed by Anderson as a new state of matterpwa . Recently topological band insulator (TBI) has attracted considerable attention because of their relevance to the quantum spin Hall effectkm ; sh . It is pointed out that there may exist spin-charge separated solitons in the presence of $\pi$ flux with induced quantum numbersqi ; ran ; ran1 . In this paper we focus on a special class of antiferromagnetic (AF) ordered state - nodal AF insulator, and we will show how spin-charge separation occurs. Nodal AF insulator is an AF order (long range or short range) with massive Dirac fermionic excitations. When there is no AF order, fermionic excitations become nodal quasi-particles. There are two examples of nodal AF insulator in condensed matter physics - one is an AF order on a honeycomb lattice, the other is $\pi$-flux phase together with a nonzero Neel order parameter. Based on these examples, our results confirm that induced quantum number on solitons is an important feature of the spin-charge separation in different systems. _Formulation_ $\mathcal{-}$ To develop a systematical formulation, we start with the extended Hubbard models, $H=-\sum\limits_{\langle i,j\rangle,\sigma}t_{ij}\hat{c}_{i,\sigma}^{\dagger}\hat{c}_{j,\sigma}+U\sum_{j}\hat{n}_{j\uparrow}\hat{n}_{j\downarrow}-\mu\sum\limits_{i,\sigma}\hat{c}_{i,\sigma}^{\dagger}\hat{c}_{i,\sigma}+h.c..$ (1) Here $\hat{c}_{i,\sigma}^{\dagger}$ and $\hat{c}_{j,\sigma}$ are electronic creation and annihilation operators. $U$ is the on-site Coulomb repulsion. $\sigma$ are the spin-indices for electrons. $\mu$ is the chemical potential. $\langle i,j\rangle$ denotes two sites on a nearest-neighbor link. $\hat{n}_{j\uparrow}$ and $\hat{n}_{j\downarrow}$ are the number operators of electrons with up-spin and down spin. On a honeycomb lattice, the nearest neighbor hopping is a constant, $t_{i,j}=t;$ on a square lattice with $\pi$-flux phase, it can be chosen as $t_{i,i+\hat{x}}=\chi,$ $t_{i,i+\hat{y}}=i\chi$ st ; st0 ; pi ; wen . The partition function of the extended Hubbard models is written as $\mathcal{Z}=\int\mathcal{D}\overline{c}\mathcal{D}ce^{-\int_{0}^{\beta}d\tau L},$ where $\displaystyle L$ $\displaystyle=$ $\displaystyle\sum_{j,\sigma}\bar{c}_{j,\sigma}\left(\partial_{\tau}-\mu\right)c_{j,\sigma}+\sum_{\langle i,j\rangle,\sigma}t_{ij}\bar{c}_{i,\sigma}c_{j,\sigma}$ $\displaystyle-U\sum_{j}n_{j\uparrow}n_{j\downarrow}.$ $\bar{c}_{i,\sigma}$ and $c_{j,\sigma}$ are Grassmann variables describing the electronic fields. Firstly let us derive long wave-length effective Lagrangian of the hopping term in the extended Hubbard models. Although $\pi$-flux phase does not break translational symmetry, we may still divide the square lattice into two sublattices, $A$ and $B$. After transforming the hopping term into momentum space, we obtain $E_{f}=2\chi\sqrt{\cos^{2}k_{x}+\cos^{2}k_{y}}$. So there exist two nodal fermi-points at $\mathbf{k}_{1}=(\frac{\pi}{2},\frac{\pi}{2}),$ $\mathbf{k}_{2}=(\frac{\pi}{2},-\frac{\pi}{2})$ and the spectrum of fermions becomes linear in the vicinity of the two nodal points. On a honeycomb lattice, after dividing the lattice into two sublattices, $A$ and $B$, the dispersion is obtained in Ref.her ; her1 ; pei . There also exist two nodal points, $\mathbf{k}_{1}=\frac{2\pi}{\sqrt{3}}(1,\frac{1}{\sqrt{3}})$ and $\mathbf{k}_{2}=\frac{2\pi}{\sqrt{3}}(-1,-\frac{1}{\sqrt{3}})$ and the spectrum of fermions becomes linear near $\mathbf{k}_{1,2}$. In the continuum limit, the Dirac-like effective Lagrangian describes the low energy fermionic modes for both cases $\mathcal{L}_{f}=i\bar{\psi}_{1}\gamma_{\mu}\partial_{\mu}\psi_{1}+i\bar{\psi}_{2}\gamma_{\mu}\partial_{\mu}\psi_{2}$ (3) where $\bar{\psi}_{1}=\psi_{1}^{\dagger}\gamma_{0}=(\begin{array}[]{llll}\bar{\psi}_{\uparrow 1A},&\bar{\psi}_{\uparrow 1B},&\bar{\psi}_{\downarrow 1A},&\bar{\psi}_{\downarrow 1B}\end{array})$ and $\bar{\psi}_{2}=\psi_{2}^{\dagger}\gamma_{0}=(\begin{array}[]{llll}\bar{\psi}_{\uparrow 2B},&\bar{\psi}_{\uparrow 2A},&\bar{\psi}_{\downarrow 2B},&\bar{\psi}_{\downarrow 2A}\end{array})$her ; her1 ; pei . $\gamma_{\mu}$ is defined as $\gamma_{0}=\sigma_{0}\otimes\tau_{z},$ $\gamma_{1}=\sigma_{0}\otimes\tau_{y},$ $\gamma_{2}=\sigma_{0}\otimes\tau_{x},$ $\sigma_{0}=\left(\begin{array}[]{ll}1&0\\\ 0&1\end{array}\right)$. $\tau^{x},$ $\tau^{y},$ $\tau^{z}$ are Pauli matrices. We have set the Fermi velocity to be unit $v_{F}=1$. In the strongly coupling limit, $U>>t$, there always exists an AF order in the extended Hubbard models. Introducing Stratonovich-Hubbard fields for the spin degrees of freedom wen , we obtain the partition function as $Z=\int\mathcal{D}\overline{c}\mathcal{D}c\mathcal{D}\mathbf{B}e^{-\int_{0}^{\beta}d\tau L}$, where the Lagrangian is given by $\displaystyle L$ $\displaystyle=$ $\displaystyle\sum_{j,\sigma}\bar{c}_{j,\sigma}\left(\partial_{\tau}-\mu\right)c_{j,\sigma}+\sum_{\langle i,j\rangle,\sigma}t_{ij}\bar{c}_{i,\sigma}c_{j,\sigma}$ $\displaystyle-\frac{3}{2U}\sum_{j}\mathbf{B}_{j}^{2}+U\sum_{j}(-1)^{j}\mathbf{B}_{j}\cdot\bar{c}_{j}\mathbf{\sigma}c_{j}$ with Pauli matrices $\mathbf{\sigma=}(\sigma^{x},\sigma^{y},\sigma^{z})\mathbf{.}$ Here $\mathbf{B}_{j}$ is a vector denoting spin configurations, $\mathbf{B}_{j}=\left\|B_{j}\right\|\mathbf{n}_{j}$ where $\left\|B_{j}\right\|=\phi_{0}$ represents the value of localized spin moments and $\mathbf{n}_{j}$ is a unit vector describing the Néel order parameter. In AF ordered state, the mass gap of the electrons is given as $m=\phi_{0}U$. Then starting from Eq.(Spin-charge Separation in Nodal Antiferromagnetic Insulator), we get the same long wave-length effective model of nodal AF insulator at ; ki ; senthil $\mathcal{L}_{\mathrm{eff}}=\sum\limits_{\alpha}i\bar{\psi}_{\alpha}\gamma_{\mu}\partial_{\mu}\psi_{\alpha}+m(\bar{\psi}_{1}\mathbf{n\cdot\sigma}\psi_{1}-\bar{\psi}_{2}\mathbf{n\cdot\sigma}\psi_{2}).$ (5) $\alpha=1,$ $2$ labels the two Fermi points _Zero modes on half skyrmions_ \- In this section we will study the properties of topological solitons. Instead of considering topological solitons with integer topological charge (skyrmions), we focus on solitons with a half topological charge, $\int\frac{1}{2\pi}\mathbf{n}\cdot\partial_{x}\mathbf{n}\times\partial_{y}\mathbf{n}$ $d^{2}\mathbf{r}=\pm\frac{1}{2}$ with $\mathbf{r=(}x,$ $y\mathbf{).}$ Such soliton is called a half skyrmion (meron or anti-meron). A meron with a narrow core size (the lattice size $a$) is characterized by $\mathbf{n=r}/\mid\mathbf{r}\mid,$ $\mathbf{r}^{2}=x^{2}+y^{2}$ pol ; meron ; half ; ba ; ng ; ots ; weng1 ; kou . To stabilize such a solitons, one may add a small easy-plane anisotropy of the Néel order. Around a meron configuration, the fermionic operators are expanded as $\displaystyle\hat{\psi}_{\mathbf{\alpha}}(\mathbf{r},t)$ $\displaystyle=$ $\displaystyle\sum_{\mathbf{k}\neq 0}\hat{b}_{\mathbf{\alpha k}}e^{-iE_{\mathbf{k}}t}\psi_{{\mathbf{\alpha}}\mathbf{k}}(\mathbf{r})$ $\displaystyle+\sum_{\mathbf{k}\neq 0}\hat{d}_{\mathbf{\alpha k}}^{\dagger}e^{iE_{\mathbf{k}}t}\psi_{\mathbf{\alpha k}}^{\dagger}(\mathbf{r})+\hat{a}_{\mathbf{\alpha}}^{0}\psi_{\mathbf{\alpha}}^{0}(\mathbf{r}),\text{ }$ where $\hat{b}_{\alpha\mathbf{k}}$ and $\hat{d}_{\alpha\mathbf{k}}^{\dagger}$ are operators of $\mathbf{k}\neq 0$ modes that are irrelevant to the soliton states discussed below. $\psi_{\mathbf{\alpha k}}^{\dagger}(\mathbf{r})=(\begin{array}[]{llll}\psi_{\uparrow\mathbf{\alpha Ak}}^{0},&\psi_{\uparrow\mathbf{\alpha}B\mathbf{k}}^{0},&\psi_{\downarrow\mathbf{\alpha Ak}}^{0},&\psi_{\downarrow\mathbf{\alpha}B\mathbf{k}}^{0}\end{array})$ are the functions of zero modes. $\hat{a}_{\alpha}^{0}$ are annihilation operators of zero modes. To obtain the zero modes, we write down two Dirac equations from Eq.(5) $i\partial_{x}\gamma_{1}\psi_{1}^{0}+i\partial_{y}\gamma_{2}\psi_{1}^{0}+m\mathbf{n\cdot\sigma}\psi_{1}^{0}=0$ (7) and $i\partial_{x}\gamma_{1}\psi_{2}^{0}+i\partial_{y}\gamma_{2}\psi_{2}^{0}-m\mathbf{n\cdot\sigma}\psi_{2}^{0}=0.$ (8) Firstly we solve the Dirac equation for $\psi_{1}^{0}$. With the ansatz $\psi_{1}^{0}=\left(\begin{array}[]{l}\xi_{1}(\tilde{x})e^{-i\theta}\\\ \xi_{2}(\tilde{x})\\\ \xi_{3}(\tilde{x})\\\ \xi_{4}(\tilde{x})e^{i\theta}\end{array}\right),$ we have $\displaystyle\partial_{\tilde{x}}\xi_{2}$ $\displaystyle=$ $\displaystyle\xi_{3},\text{ }\partial_{\tilde{x}}\xi_{3}=\xi_{2},$ (9) $\displaystyle\partial_{\tilde{x}}\xi_{1}$ $\displaystyle=$ $\displaystyle-\frac{\xi_{1}}{\tilde{x}}+\xi_{4},\text{ }\partial_{\tilde{x}}\xi_{4}=-\frac{\xi_{4}}{\tilde{x}}+\xi_{1}$ where $\mathbf{r=}\mid\mathbf{r}\mid(\cos\theta,\sin\theta)$ and $\tilde{x}=\frac{\mid\mathbf{r}\mid}{m}$. The solution has been obtained in Ref.zeromode as $\xi_{1}(\tilde{x})=\xi_{4}(\tilde{x})=0,\text{ }\xi_{2}(\tilde{x})=\xi_{3}(\tilde{x})=\tilde{x}^{\frac{1}{2}}K_{\frac{1}{2}}(\tilde{x})$ where $K_{\frac{1}{2}}(\tilde{x})$ is the modified Bessel function. So the solution of $\psi_{1}^{0}$ becomes $\left(\begin{array}[]{l}0\\\ \tilde{x}^{\frac{1}{2}}K_{\frac{1}{2}}(\tilde{x})\\\ \tilde{x}^{\frac{1}{2}}K_{\frac{1}{2}}(\tilde{x})\\\ 0\end{array}\right).$ To solve $\psi_{2}^{0},$ we transform the equation $i\partial_{i}\gamma_{i}\psi_{2}^{0}-m\mathbf{n\cdot\sigma}\psi_{2}^{0}=0$ into $Ui\partial_{i}\gamma_{i}\tilde{\psi}_{2}^{0}U^{-1}+mU\mathbf{n\cdot\sigma}\tilde{\psi}_{2}^{0}U^{-1}=0,$ (10) where $U=e^{i\pi\gamma_{0}/2},$ $U\gamma_{i}U^{-1}=-\gamma_{i}$ and $U^{-1}\psi_{2}^{0}U=\tilde{\psi}_{2}^{0}.$ Then the solution of $\psi_{2}^{0}\ $is obtained as $\left(\begin{array}[]{l}0\\\ -\tilde{x}^{\frac{1}{2}}K_{\frac{1}{2}}(\tilde{x})\\\ \tilde{x}^{\frac{1}{2}}K_{\frac{1}{2}}(\tilde{x})\\\ 0\end{array}\right).$ It is noticable that from above solutions of zero modes, the components $\psi_{\uparrow 1A}^{0},$ $\psi_{\downarrow 1B}^{0},$ $\psi_{\downarrow 2A}^{0}$ and $\psi_{\uparrow 2B}^{0}$ are all zero. _Topological mechanism of spin-charge separation -_ For the solutions of zero modes, there are four zero-energy soliton states $\mid\mathrm{sol}\rangle$ around a half skyrmion which are denoted by $\mid 1_{+}\rangle\otimes\mid 2_{+}\rangle,$ $\mid 1_{-}\rangle\otimes\mid 2_{-}\rangle,$ $\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle$ and $\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle.$ $\mid 1_{-}\rangle$ and $\mid 2_{-}\rangle$ are empty states of the zero modes $\psi_{1}^{0}(\mathbf{r})$ and $\psi_{2}^{0}(\mathbf{r});$ $\mid 1_{+}\rangle$ and $\mid 2_{+}\rangle$ are occupied states of them. Thus we have the relationship between $\hat{a}_{\mathbf{\alpha}}^{0}$ and $\mid\mathrm{sol}\rangle$ as $\hat{a}_{1}^{0}\mid 1_{+}\rangle=\mid 1_{-}\rangle,\text{ }\hat{a}_{1}^{0}\mid 1_{-}\rangle=0,\text{ }\hat{a}_{2}^{0}\mid 2_{+}\rangle=\mid 2_{-}\rangle,\text{ }\hat{a}_{2}^{0}\mid 2_{-}\rangle=0,$ (11) Firstly we define total induced fermion number operators of the soliton states, $\hat{N}_{F}=\sum\limits_{\alpha}\hat{N}_{\alpha,F}$ with $\displaystyle\hat{N}_{\alpha,F}$ $\displaystyle\equiv$ $\displaystyle\int:\hat{\psi}_{\alpha}^{\dagger}\hat{\psi}_{\alpha}:d^{2}\mathbf{r}$ $\displaystyle=$ $\displaystyle(\hat{a}_{\alpha}^{0})^{\dagger}\hat{a}_{\alpha}^{0}+\sum\limits_{\mathbf{k}\neq 0}(\hat{b}_{\alpha\mathbf{k}}^{\dagger}\hat{b}_{\alpha\mathbf{k}}-\hat{d}_{\alpha\mathbf{k}}^{\dagger}\hat{d}_{\alpha\mathbf{k}})-\frac{1}{2}.$ $:\hat{\psi}_{\alpha}^{\dagger}\hat{\psi}_{\alpha}:$ means normal product of $\hat{\psi}_{\alpha}^{\dagger}\hat{\psi}_{\alpha}$. From the relation between $\hat{a}_{\mathbf{\alpha}}^{0}$ and $\mid\mathrm{sol}\rangle$ in Eq.(11), we find that $\mid 1_{\pm}\rangle$ or $\mid 2_{\pm}\rangle$ have eigenvalues of $\pm\frac{1}{2}$ of the total induced fermion number operator $\hat{N}_{F}$, $\displaystyle\hat{N}_{1,F}\|1_{\pm}\rangle$ $\displaystyle=$ $\displaystyle\pm{\frac{1}{2}}\|1_{\pm}\rangle,\text{ }\hat{N}_{1,F}\|2_{\pm}\rangle=0,\text{ }$ (13) $\displaystyle\hat{N}_{2,F}\|2_{\pm}\rangle$ $\displaystyle=$ $\displaystyle\pm{\frac{1}{2}}\|2_{\pm}\rangle,\text{ }\hat{N}_{2,F}\|1_{\pm}\rangle=0.$ Another important induced quantum number operator is staggered spin operator, $\hat{S}_{(\pi,\pi)}^{z}=\frac{1}{2}\sum\limits_{i\in A}\hat{c}_{i}^{\dagger}\sigma_{z}\hat{c}_{i}-\frac{1}{2}\sum\limits_{i\in B}\hat{c}_{i}^{\dagger}\sigma_{z}\hat{c}_{i}=\frac{1}{2}\int:[(\hat{\psi}_{\uparrow 1A}^{\dagger}\hat{\psi}_{\uparrow 1A}+\hat{\psi}_{\downarrow 1B}^{\dagger}\hat{\psi}_{\downarrow 1B}-\hat{\psi}_{\downarrow 1A}^{\dagger}\hat{\psi}_{\downarrow 1A}$ $-\hat{\psi}_{\uparrow 1B}^{\dagger}\hat{\psi}_{\uparrow 1B})+(\hat{\psi}_{\uparrow 2A}^{\dagger}\hat{\psi}_{\uparrow 2A}+\hat{\psi}_{\downarrow 2B}^{\dagger}\hat{\psi}_{\downarrow 2B}-\hat{\psi}_{\downarrow 2A}^{\dagger}\hat{\psi}_{\downarrow 2A}-\hat{\psi}_{\uparrow 2B}^{\dagger}\hat{\psi}_{\uparrow 2B})]:d^{2}\mathbf{r}.$ For the four degenerate zero modes, it can be simplified into $\hat{S}_{(\pi,\pi)}^{z}\mid\mathrm{sol}\rangle=\frac{1}{2}(\hat{N}_{2,F}-\hat{N}_{1,F})\mid\mathrm{sol}\rangle.$ Let us show the detailed calculations. From the zero solutions of $\psi_{\uparrow 1A}^{0},$ $\psi_{\downarrow 1B}^{0},$ $\psi_{\downarrow 2A}^{0}$ and $\psi_{\uparrow 2B}^{0}$, we obtain four equations $\displaystyle(\int$ $\displaystyle:$ $\displaystyle\hat{\psi}_{\uparrow 1A}^{\dagger}\hat{\psi}_{\uparrow 1A}:d^{2}\mathbf{r})\mid\mathrm{sol}\rangle\equiv 0,$ (14) $\displaystyle(\int$ $\displaystyle:$ $\displaystyle\hat{\psi}_{\downarrow 1B}^{\dagger}\hat{\psi}_{\downarrow 1B}:d^{2}\mathbf{r})\mid\mathrm{sol}\rangle\equiv 0,$ $\displaystyle(\int$ $\displaystyle:$ $\displaystyle\hat{\psi}_{\downarrow 2A}^{\dagger}\hat{\psi}_{\downarrow 2A}:d^{2}\mathbf{r})\mid\mathrm{sol}\rangle\equiv 0,$ $\displaystyle(\int$ $\displaystyle:$ $\displaystyle\hat{\psi}_{\uparrow 2B}^{\dagger}\hat{\psi}_{\uparrow 2B}:d^{2}\mathbf{r})\mid\mathrm{sol}\rangle\equiv 0.$ Using above four equations, we obtain $\hat{S}_{(\pi,\pi)}^{z}\mid\mathrm{sol}\rangle=\frac{1}{2}\int d^{2}\mathbf{r}:(-\hat{\psi}_{\uparrow 1A}^{\dagger}\hat{\psi}_{\uparrow 1A}-\hat{\psi}_{\downarrow 1B}^{\dagger}\hat{\psi}_{\downarrow 1B}-\hat{\psi}_{\downarrow 1A}^{\dagger}\hat{\psi}_{\downarrow 1A}-\hat{\psi}_{\uparrow 1B}^{\dagger}\hat{\psi}_{\uparrow 1B}$ $+\hat{\psi}_{\uparrow 2A}^{\dagger}\hat{\psi}_{\uparrow 2A}+\hat{\psi}_{\downarrow 2B}^{\dagger}\hat{\psi}_{\downarrow 2B}+\hat{\psi}_{\downarrow 2A}^{\dagger}\hat{\psi}_{\downarrow 2A}+\hat{\psi}_{\uparrow 2B}^{\dagger}\hat{\psi}_{\uparrow 2B}):\mid\mathrm{sol}\rangle=-\frac{1}{2}(\hat{N}_{1,F}-\hat{N}_{2,F})\mid\mathrm{sol}\rangle.$ Then we calculate two induced quantum numbers defined above. Without doping, the soliton states of a half skyrmion are denoted by $\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle$ and $\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle$. One can easily check that the total induced fermion number on the solitons is zero from the cancelation effect between two nodals $\hat{N}_{F}\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle=\hat{N}_{F}\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle=0.$ It is consistent to the earlier results that forbid a Hopf term for the low energy theory of two dimensional Heisenberg modelwenhopf . On the other hand, there exists an induced staggered spin moment on the soliton states $\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle$ and $\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle,$ $\displaystyle\hat{S}_{(\pi,\pi)}^{z}$ $\displaystyle\mid$ $\displaystyle 1_{-}\rangle\otimes\mid 2_{+}\rangle=\frac{1}{2}\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle,$ (15) $\displaystyle\hat{S}_{(\pi,\pi)}^{z}$ $\displaystyle\mid$ $\displaystyle 1_{+}\rangle\otimes\mid 2_{-}\rangle=-\frac{1}{2}\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle.$ The induced staggered spin moment may be straightforwardly obtained by combining the definition of $\hat{S}_{(\pi,\pi)}^{z}$ and Eq.(13) together. When half skyrmions become mobile, their quantum statistics becomes important. Let us examine the statistics of a half skyrmion with an induced staggered spin moment. In CP(1) representation of $\mathbf{n,}$ a ”bosonic spinon” is introduced by $\mathbf{n}=\mathbf{\bar{z}\sigma z}$ with $\mathbf{z}=\left(\begin{array}[]{l}z_{\uparrow}\\\ z_{\downarrow}\end{array}\right)$ and $\mathbf{\bar{z}z=1}$. Since each ”bosonic spinon” $\mathbf{z}$ carries $\pm\frac{1}{2}$ staggered spin moment, an induced staggered spin moment corresponds to a trapped ”bosonic spinon” $\mathbf{z}$. On the other hand, a half skyrmion can be regarded as a $\pi-$flux of the ”bosonic spinon”, $\int\frac{1}{2\pi}\mathbf{n}\cdot\partial_{x}\mathbf{n}\times\partial_{y}\mathbf{n}$ $d^{2}\mathbf{r}=\frac{1}{2\pi}\int\epsilon_{\mu\nu}\partial_{\mu}a_{\nu}d^{2}\mathbf{r}=\pm\frac{1}{2}$ with $a_{\mu}\equiv\frac{i}{2}(\mathbf{\bar{z}}\partial_{\mu}\mathbf{z}-\partial_{\mu}\mathbf{\bar{z}z}).$ To be more explicit, moving a ”bosonic spinon” $z$ around a half skyrmion generates a Berry phase $\phi$ to $\mathbf{z}\rightarrow\mathbf{z}^{\prime}=\left(\begin{array}[]{l}z_{\uparrow}e^{i\phi}\\\ z_{\downarrow}e^{i\phi}\end{array}\right)$ where $\phi=\int\epsilon_{\mu\nu}\partial_{\mu}a_{\nu}d^{2}\mathbf{r}=\pm\pi$. As a result, a ”bosonic spinon” $\mathbf{z}$ and a half skyrmion (meron or antimeron ) share mutual semion statistics. Binding the trapped ”bosonic spinon”, a mobile half skyrmion becomes a fermionic particle. We may use the operator $\hat{f}_{\sigma}$ to describe such neutral fermionic particle with half spin. The relation between the zero energy states and the fermionic states is given as $\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle=\hat{f}_{\downarrow}^{\dagger}\mid 0\rangle_{f}\text{ and }\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle=\hat{f}_{\uparrow}^{\dagger}\mid 0\rangle_{f}$ (The state $\|0\rangle_{f}$ is defined through $\hat{f}_{\uparrow}\|0\rangle_{f}=\hat{f}_{\downarrow}\|0\rangle_{f}=0$). We call such neutral object (fermion with $\pm\frac{1}{2}$ spin degree freedom) a (fermionic) ”spinon”. Now we go away from half filling. It is known that when a hole (electron) is doped, it is equivalence to removing (adding) an electron. Without considering the existence of half skyrmions, the hole (electron) will be doped into the lower (upper) Hubbard band. The existence of zero modes on half skyrmions leads to the appearance of bound levels in the middle of the Mott-Hubbard gap meron . The hole (electron) will be doped onto the bound states on the half skyrmion and then one of the zero modes is occupied. When one hole is doped, the soliton state is denoted by $\|1_{-}\rangle\otimes\|2_{-}\rangle$. One can easily check the result by calculating its induce quantum numbers. On the one hand, there is no induced staggered spin moment, $\hat{S}_{(\pi,\pi)}^{z}\mid 1_{-}\rangle\otimes\mid 2_{-}\rangle=0.$ On the other hand, the total fermion number is not zero, $\hat{N}_{F}\|1_{-}\rangle\otimes\|2_{-}\rangle=-\|1_{-}\rangle\otimes\|2_{-}\rangle.$ These results mean that such soliton state is a spinless ”holon” with positive charge degrees of freedom. After binding a fermionic hole, the soliton state (holon) does not have an induced staggered spin moment. Thus the holon obeys bosonic statistics and becomes a charged bosonic particles. When one electron is doped, the soliton state is denoted by $\|1_{+}\rangle\otimes\|2_{+}\rangle$. The induced quantum numbers of it are $\hat{N}_{F}\|1_{+}\rangle\otimes\|2_{+}\rangle=+\|1_{+}\rangle\otimes\|2_{+}\rangle$ and $\hat{S}_{(\pi,\pi)}^{z}\mid 1_{+}\rangle\otimes\mid 2_{+}\rangle=0.$ Such soliton state is also a bosonic particle with a negative charge but without spin degrees of freedom. We call such a soliton state an ”electon” to mark difference with the word ”electron”. \| $\mid 1_{+}\rangle\otimes\mid 2_{+}\rangle$ \| $\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle$ \| $\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle$ \| $\mid 1_{-}\rangle\otimes\mid 2_{-}\rangle$ ---\|---\|---\|---\|--- $N_{F}$ \| $1$ \| 0 \| 0 \| -1 $S_{(\pi,\pi)}^{z}$ \| $0$ \| $1/2$ \| $-1/2$ \| 0 Table 1: Quantum numbers of the degenerate soliton states. Finally we get a __ topological mechanism of spin-charge separation in nodal AF insulators. There exist two types of topological objects - one is the fermionic spinon, the other is the bosonic holon ( or the bosonic electon ). In 1D system, real spin-charge separation may occur. As far as the low energy physics is concerned, the spin and charge dynamics are completely decoupled from each other. In 2D, real spin-charge separation in a nodal AF insulators can not occur in long range AF order. In the future we will study the deconfinement condition of spin-charge separated solitons and explore the properties of deconfined phases with real spin-charge separation. _Summary_ \- By using 2D $\pi$-flux phase Hubbard model and the Hubbard model on a honeycomb lattice as examples, we explore spin-charge separation in nodal AF insulator. The crus crux of the matter in this paper is the discovery of induced staggered spin moment $S_{(\pi,\pi)}^{z}$ on half skyrmions in nodal AF insulators. Based on such nontrivial induced quantum number, we classify four degenerate soliton states with zero energy - two of them ($\mid 1_{-}\rangle\otimes\mid 2_{+}\rangle$ and $\mid 1_{+}\rangle\otimes\mid 2_{-}\rangle$) represent the up-spin and down-spin states for a fermionic ”spinon”, another state ($\|1_{-}\rangle\otimes\|2_{-}\rangle$) represents a ”holon” and the last one ($\|1_{+}\rangle\otimes\|2_{+}\rangle$) denotes an ”electon”. This research is supported by NFSC Grant no. 10574014. ## References * (1) V. J. Emery and A. Luther, Phys. Rev. Lett. 33, 589 (1974). * (2) R. Jackiw and C. Rebbi, Phys. Rev. D 13, 3398 (1976). * (3) W. P. Su, J. R. Schrieffer and A. J. Heeger, Phys. Rev. B 22, 2099 (1980); A.J. Heeger, S. Kivelson, J. R. Schrieffer and W.-P. Su, Rev. Mod. Phys. 60, 781 (1988); D. J. Van Harlingen, Rev. Mod. Phys. 67, 515 (1995). * (4) P. W. Anderson, Science 235, 1196 (1987). * (5) S.A. Kivelson, D.S. Rokhsar, and J.R. Sethna, Phys. Rev. B 35, 8865(1987). * (6) C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005). * (7) S. Murakami, N. Nagaosa and S.C. Zhang, Science 301, 1348 (2003). * (8) X. L. Qi and S. C. Zhang, cond-mat/08010252. * (9) Y. Ran, A. Vishwanath and D. H. Lee, cond-mat/08010627. * (10) Y. Ran, D. H. Lee and A. Vishwanath, cond-mat/08062321. * (11) G. Kotliar, Phys. Rev. B 37, 3664 (1988). * (12) T. C. Hsu, Phys. Rev. B 41, 11379 (1990). * (13) X. G. Wen, _Quantum Field Theory of Many-Body Systems_(Oxford University Press, 2004). * (14) I. Affleck and J. B. Marston, Phys. Rev. B 37, 3774 (1988); J. B. Marston and I. Affleck, Phys. Rev. B 39, 11538 (1989). * (15) C-Y. Hou, C. Chamon, and C. Mudry, Phys. Rev. Lett. 98, 186809 (2007). * (16) Igor F. Herbut, Phys. Rev. Lett. 99, 206404 (2007). * (17) R. Jackiw, S.-Y. Pi, cond-mat/0701760. * (18) A. Tanaka and Xiao Hu, Phys. Rev. Lett. 95, 036402 (2005). * (19) Ki-Seok Kim, Phys. Rev. B 72, 214401 (2005). * (20) T. Senthil and M. P. A. Fisher, Phys. Rev. B 74, 064405 (2006). * (21) A. A. Belavin and A. M. Polyakov, JETP Lett. 22, 245 (1975). * (22) J.A. Verges, _et al_ , Phys. Rev. B 43, 6099, (1991); S. John, M. Berciu and A. Golubentsev, Europhys. Lett. 41, 31 (1998); M. Berciu and S. John, Phys. Rev. B 57, 9521 (1998); M. Berciu and S. John, Phys. Rev. B 61, 16454 (2000). * (23) T. Morinari, Phys. Rev. B 72, 104502 (2005). * (24) G. Baskaran, Phys. Rev. B 68, 212409 (2003). * (25) T.-K. Ng, Phys. Rev. Lett. 82, 3504 (1999). * (26) Y. Otsuka and Y. Hatsugai, Phys. Rev. B 65, 073101 (2002). * (27) Z. Y. Weng, D. N. Sheng, and C. S. Ting, Phys. Rev. Lett. 80, 5401 (1998); Zheng-Yu Weng, Int. J. Mod. Phys. B21, 773 (2007). * (28) S. P. Kou and Z. Y. Weng, Phy. Rev. Lett. 90, 157003 (2003); S. P. Kou, X. L. Qi, Z. Y. Weng, Phys. Rev. B 71, 235102 (2005). * (29) M. Carena, _et al_ , Phys. Rev. D 42, 2120 (1990). * (30) X. G. Wen and A. Zee, Phys. Rev. Lett, 61, 1025 (1988).	arxiv-papers	2008-08-13T04:41:07	2024-09-04T02:48:57.256305	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Su-Peng Kou", "submitter": "Supeng Kou", "url": "https://arxiv.org/abs/0808.1777" }
0808.1782	# Architectural design for a topological cluster state quantum computer Simon J. Devitt devitt@nii.ac.jp National Institute for Informatics, 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo 101-8430, Japan. Austin G. Fowler Institute for Quantum Computing, University of Waterloo, Waterloo, Canada. Ashley M. Stephens Centre for Quantum Computer Technology, School of Physics, University of Melbourne, Victoria 3010, Australia Andrew D. Greentree Centre for Quantum Computer Technology, School of Physics, University of Melbourne, Victoria 3010, Australia Lloyd C.L. Hollenberg Centre for Quantum Computer Technology, School of Physics, University of Melbourne, Victoria 3010, Australia William J. Munro Hewlett-Packard Laboratories, Filton Road, Stoke Gifford, Bristol BS34 8QZ, United Kingdom National Institute for Informatics, 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo 101-8430, Japan. Kae Nemoto National Institute for Informatics, 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo 101-8430, Japan. ###### Abstract The development of a large scale quantum computer is a highly sought after goal of fundamental research and consequently a highly non-trivial problem. Scalability in quantum information processing is not just a problem of qubit manufacturing and control but it crucially depends on the ability to adapt advanced techniques in quantum information theory, such as error correction, to the experimental restrictions of assembling qubit arrays into the millions. In this paper we introduce a feasible architectural design for large scale quantum computation in optical systems. We combine the recent developments in topological cluster state computation with the photonic module, a simple chip based device which can be used as a fundamental building block for a large scale computer. The integration of the topological cluster model with this comparatively simple operational element addresses many significant issues in scalable computing and leads to a promising modular architecture with complete integration of active error correction exhibiting high fault-tolerant thresholds. ## I Introduction The scientific effort to construct a large scale device capable of Quantum Information Processing (QIP) has advanced significantly over the past decade NPT99 ; CNHM03 ; YPANT03 ; CLW04 ; HHB05 ; GHW05 ; G06 ; H06 ; G07 ; O'Brien2 . However, the design of a viable processing architecture capable of housing and manipulating the millions of physical qubits necessary for scalable QIP is hampered by an architectural design gap that exists between advanced techniques in error correction and fault-tolerant computing and the comparatively small scale devices under consideration by experimentalists. While some progress in scalable system design has been made Kielpinski1 ; Taylor1 ; Hollenberg1 ; Fowler3 ; Knill1 ; Kok1 , the development of a truly large scale quantum architecture, able to implement programmable QIP with extensive error correction, hinges on the ability to adapt well established computational and error correction models to the experimental operating conditions and fabrication restrictions of physical systems, well beyond 1000 physical qubits. The recent introduction by Raussendorf, Harrington and Goyal of Topological Cluster state Quantum Computation (TCQC) Raussendorf4 ; Raussendorf5 ; Fowler2 marks an important milestone in the advance of theoretical techniques for quantum information processing (QIP). This model of QIP differs significantly from both the circuit model of computation and the traditional cluster state model NC00 ; Raussendorf1 in that the entanglement resource defines a topological “backdrop” where information qubits can be defined in very non- local ways. By disconnecting the physical qubits (used to construct the lattice) from the information qubits (non-local correlations within the lattice), information can be topologically protected, which is inherently robust against local perturbations caused by errors on the physical system. This exploitation of topological protection utilizing qubits, rather than exotic anyonic systems, which are particularly difficult to experimentally create and manipulate Nayak1 , leads to a computation model that exhibits high fault-tolerant thresholds where problematic error channels such as qubit loss are naturally corrected. In terms of architecturally implementing certain topological coding models, there are two general techniques which can be used, depending on the physical system under consideration. Surface codes BK01 ; DKLP02 ; Fowler1 , appropriate for matter based systems and the 3D TCQC model Raussendorf4 ; Raussendorf5 , which is far more useful in optics based architectures due to the high mobility and comparatively inexpensive cost of photonic qubits. This paper introduces a computational architecture for the 3D topological model, utilizing photonic qubits. By making use of the TCQC model, error correction protocols are automatically incorporated as a property of the computational scheme and photon loss (arguably the dominant error channel) becomes correctable without the incorporation of additional coding schemes or protocols Dawson2 ; Ralph1 ; Varnava1 ; Kok1 . As error correction within the TCQC model is predicated on the presence of a large cluster lattice before logical qubits are defined, standard probabilistic techniques for preparing entangled photons are generally not appropriate without employing extensive photon routing and/or storage. Therefore, a deterministic photon/photon entangling gate is desired to drastically simplify any large scale implementation of this model. Such operations could for instance include the C-Phase and parity gates Beenakker4 ; Nemoto4 ; Spiller6 , and may be via direct photon-photon interactions, cavity QED techniques Domokos1 ; Duan1 or an indirect bus-mediated quantum nondemolition-like interaction Spiller6 ; Devitt1 . As an architectural building block, we are going to focus on the later with the recently introduced idea of the photonic module and chip Devitt1 ; Stephens1 . This small scale, chip based quantum device is an illustrative example of a technology with all the essential components to act as an architectural building block for the TCQC model. This device has the flexibility to entangle an arbitrary number of photons with no dynamic change in its operation and the manner in which photons flow in and out of the unit make it an ideal component to realize a modular structure for large scale TCQC. In recent years, experimental work in the both cavity/photon coupling E1 ; F1 ; F2 and chip based single photon technologies P1 ; C1 ; M1 has advanced significantly. While the high fidelity construction of the photonic chip is still a daunting task, the continued effort in these areas allows for optimism that such a device is an experimental possibility. The following will detail the lattice preparation network, utilizing the photonic chip illustrated in Fig. 1 as the basic building block of the computer. We detail the physical layout of the network, the optical switching sequence required and several techniques that are used to optimize the preparation of the cluster. We complete the discussion with a resource analysis, examining the number of fabricated devices required to perform a large scale quantum algorithm. As this paper is focused on the possible construction of a large scale architecture, we utilize the photonic chip as a building block for this architectural design and we direct readers to Refs. Devitt1 ; Stephens1 ; Su1 ; Yang1 which examine the micro-analytics of the photonic chip in more depth. Additionally, we will also assume (for the sake of conceptual simplicity) that appropriate, high fidelity, on demand single photon sources and detectors are also feasible on the same developmental time frame as a photonic chip. However, as the photonic module is essentially a non-demolition photon detector, sources and detectors can be constructed directly with the chip Nemoto1 allowing us to relax these assumptions, if required. Figure 1: Schematic design for the photonic chip. The chip is a 3-in 3-out integrated device containing one photonic module Devitt1 , classical single photon routing and two optical waveplates allowing for the optional application of single photon Hadamard gates to specific photons. Not shown are Hadamard plates potentially required on the input or output port of the chip. ## II The flowing computer design The general structure of the optical TCQC is illustrated in Fig. 2. The high mobility of single photons and the nature of the TCQC model allows for an essentially “flowing” model of computation. Initially, unentangled photons enter the preparation network from the rear, flow through a static network consisting of layers of photonic chips arranged in 4 stages and exit the preparation network where they can be immediately consumed for computation. Figure 2: General layout of the optical architecture. Layers of photonic chips are arranged in 4 stages to prepare the entanglement links between qubits. Photons enter the network from appropriate single photon sources at the rear and exit the front linked up in the required cluster state. Single photon detectors can then be placed immediately after the preparation network to consume the cluster, performing computation. This eliminates the need for photon routing and storage as each individual photon essentially follows a linear trajectory from the source, through the photonic chip preparation network and into detectors. The box in the left hand figure indicates the small section of stages 1 and 2 of the preparation network illustrated in Fig. 5. Figure 3: Layout of the optical architecture. This is a magnified section of Fig. 2 where only the last (of four) stages of photonic chips is illustrated. Figure 4: Structure of each unit cell of the cluster. The cluster is not a Body Centered Cubic (BCC) lattice. The green links, representing entanglement bonds specified by the stabilizer generators of Eq. 1, ensure that each photon is connected to four neighbors, not six. In the flowing network, photons along the $z$-axis (for a given $(x,y)$ co-ordinate) are carried by common optical waveguides. As the lattice is not fully connected, there are two groups of waveguides, one set runs at a fixed repetition rate along the $z$-axis (the red photons shown above) and the other operates at half that rate (black photons). We denote these waveguides as full and half rate lines respectively. Fig 3. shows a smaller region of computer where only one (of four) stages of the preparation network is illustrated for clarity while Fig 4. illustrates a single unit cell of the cluster state which repeats and extends in all three dimensions. In the TCQC model, computation proceeds by measuring, sequentially, cross sectional layers ($x-y$ plane) of the lattice which acts to simulate time through the computation. Qubit information is defined within the lattice via the creation of holes (known as defects). As the cluster is consumed along the $z$-axis, defects are deformed via measurement such that they can be braided around each other, enacting CNOT gates between qubits Raussendorf5 ; Fowler2 . Each unit cell of the cluster does not have a Body Centered Cubic (BCC) structure, each photon is only connected to four neighbors, not six. Therefore, in a flowing network design, where each photon pulse along the $z$-axis is carried by common optical waveguides, there are two sets of photon pulse repetition rates. One set (waveguides carrying red photons in Fig 4) runs at a fixed repetition rate which we will denote as full rate lines. The other set (waveguides carrying black photons in Fig 4) runs at half that rate, denoted half rate lines. If the lattice is examined along the $z$-axis, these full and half rate lines form a checkerboard pattern. The global cluster is uniquely defined utilizing the stabilizer formalism Gottesman1 . The lattice is a unique state defined as the simultaneous $+1$ eigenstate of the operators, $K^{x,y,z}=X_{x,y,z}\bigotimes_{b\in n(x,y,z)}Z_{b},$ (1) where $X$ and $Z$ are the $2\times 2$ Pauli operators, $(x,y,z)$ are the co- ordinates of each of the $N$ photons in the 3D lattice, $n(x,y,z)$ is the set of 4 qubits linked to the node $(x,y,z)$ and the $N-5$ identity operators are implied [Fig. 4]. For photons that do not have four nearest neighbors (i.e. at edges of the lattice), the associated stabilizer retains this form but excludes the operator(s) associated with the missing neighbor(s). The optical network to prepare such a state requires the deterministic projection of a group of unentangled photons into the entangled state defined via the stabilizers. Fig 1 illustrates the basic structure of the photonic chip, introduced in Refs. Devitt1 ; Stephens1 which we utilize to perform these deterministic projections. The photonic module, which lies at the heart of each chip, is designed to project an arbitrary $N$ photon state into a $\pm 1$ eigenstate of the operator $X^{\otimes N}$ Devitt1 , where $N$ is the number of photons sent through the module between initialization and measurement of the atomic system. In order to perform a parity check of the operator $ZZXZZ$, single photon Hadamard operations are applied to every photon before and after passage through a photonic chip. For every group of five photons sent through between initialization and measurement, the central photon in the stabilizer operator will be routed through a second set of Hadamard gates before and after passing through the module. Note, as each chip is linked in series, the Hadamard rotations on each input/output are not required for all stages of the preparation network as they cancel. ### Photon stream initialization In total there are four stages required to prepare the 3D lattice. The four stages are partitioned into two groups which have identical layout and switching patterns. These two groups stabilize the lattice with respect to the stabilizer operators along the $y-z$ planes of the cluster and the $x-z$ planes respectively. If an arbitrary input state is utilized to prepare the cluster, two additional stages are required to stabilize the cluster with respect to each operator associated with the $x-y$ plane of each unit cell. However we can eliminate the need to perform these parity checks by carefully choosing the initial product state that is fed into the preparation network. Each stabilizer operator, $ZZXZZ$, in the $x-y$ plane has the $X$ operator centered on the photons in half rate lines (i.e. the black photons in Fig. 3). Each photon in these lines is prepared in the $+1$ eigenstate of $X$, $\|+\rangle=(\|H\rangle+\|V\rangle)/\sqrt{2}$, while photons in full rate lines are initialized in the state $+1$ eigenstate of $Z$, $\|H\rangle$, (we assume a polarization basis for computation). This initialization ensures that the stabilizer set for the photon stream (before entering the preparation network) is described by the stabilizers, $K^{i}=X_{i},\ i\in{h},\quad K^{j}=Z_{j},\ j\in{f},$ (2) where $h$ and $f$ are the sets containing photons in the half and full rate optical waveguides respectively. As any product of stabilizers is also a stabilizer of the system, all of the stabilizers in the $x-y$ planes of the cluster are automatically satisfied. It can be easily checked that the only element in the group generated by Eq. 2 that commutes with the stabilizer projections associated with the $y-z$ and $x-z$ planes of the cluster is the $ZZXZZ$ term associated with each stabilizer along the $x-y$ plane. ## III Preparation network and photonic chip operation Illustrated in Fig 5 is a 2D cross section of the preparation network (the section of the global preparation network indicated in Fig. 2). We illustrate only two stages of the network since the preparation network in the $x-z$ plane is identical to the $y-z$ plane. Each stage requires a staggered arrangement of photonic chips and in Fig 5, after a specific stage, we have detailed the stabilizer that has been measured where the central photon corresponds to the $X$ term in the measured operator. The temporal staggering of the photonic state, for each optical line, is also detailed where the spacing interval, $T$, is bounded below by the minimum interaction time required for the operation of the photonic module and bounded above by the coherence time of the atomic systems in each chip Devitt1 ; Stephens1 . Figure 5: Chip orientation for stages one and two of the preparation network, utilized to measure the stabilizer operators along the $y-z$ plane of the lattice. Photons flow from left to right through each chip. The temporal staggering of each photon is required such that only one photon passes through each photonic module at any given time step. The fundamental temporal staggering is the atom/photon interaction time $T$, however each row has to be temporally offset to ensure proper temporal ordering for future stages. Each stage measures a specific stabilizer, illustrated with a cross after the specific stage. The operator that is measured is $ZZXZZ$, where the $X$ operator associated with the photon at the center of the cross (insert). Extending this network to more and more cells requires extending each column vertically. The network for stages three and four, required to stabilize the system with respect to the $x-z$ plane, is identical where the switching sequence is offset by $2T$. ### III.1 Temporal Asynchrony Before detailing the switching sequence for these stages we have to address a slight complication that arises when creating the 3D lattice. In general, each photon is involved in five separate parity checks. By utilizing a specific input state we can remove the need to measure the stabilizers associated with the $x-y$ planes of each unit cell. This implies that we are measuring four operators for each photon present in half rate lines (we no longer perform a parity check associated with the $x-y$ plane stabilizers for these photons) and three operators for each photon in full rate lines (each of these photons are involved in two parity checks associated with $x-y$ plane stabilizers). As each photon suffers a temporal delay of $T$ by passing through a photonic module, there is a temporal discrepancy with the photons in the half rate and full rate lines which, if not compensated, leads to two photons being temporally synchronous in later parity checks. The flexibility of the photonic chips allows us to solve the delay problem in a convenient way. If a delay were not required, a given module would have a window of $3T$ where it is idle between successive parity checks. Normally this window would be utilized for measurement and re-initialization of the atomic system. For a given stage of the preparation network, every fourth photon in the full rate lines is not involved in any parity check (see Fig. 5) and must therefore be delayed. Hence we partition this $3T$ window into three steps. Immediately after the last photon from the previous parity check exits the chip, the atomic system is measured (now utilizing a temporal window of $T$ rather than $3T$). As the delayed photon is the next one to enter the chip, we simply do not initialise the module in the required atomic superposition state. As detailed in Ref. Stephens1 , the atom/photon interaction required for the module requires initializing a 3-level atomic system into a superstition of its two ground states, $(\|1\rangle+\|2\rangle)/\sqrt{2}$, with a resonant RF field. The presence of a photon in the cavity mode (which is detuned with the $\|1\rangle\leftrightarrow\|3\rangle$ transition) induces a phase shift on the state $\|1\rangle$ which oscillates the system between the states $(\pm\|1\rangle+\|2\rangle)/\sqrt{2}$. If we instead keep the atomic system in the $\|1\rangle$ (or $\|2\rangle$) state, the presence of a photon will have no effect on the atomic system. Delaying the required photon by $T$ therefore requires operating the module as usual, but for this first step we do not initialize the atomic system. This stage again takes time $T$ and we refer to it as the “holding” stage. In the next temporal window, $T$, the atomic system is re-initialized in the appropriate superposition state and the next 5 photon parity check proceeds as normal. This delay trick will maintain the temporal staggering of the photons without the need for additional technology and ensures that every photonic chip is active for all time steps (no idle time for any module). ### III.2 Switching sequence Illustrated in Fig. 5 and Tab. 1 is the network and switching sequence for the stabilizers associated with the $y-z$ planes of the unit cell. The stabilizers associated with the $x-z$ planes are measured using precisely the same network and switching sequence (when viewing the lattice along the $y$-axis). The difference in the switching sequence is a global offset of $2T$ to account for the delay from stages one and two. Chip \| 1 \| 2 \| 3 \| 4 \| 5 \| 6 \| 7 \| 8 \| 9 \| 10 \| 11 \| 12 \| 13 ---\|---\|---\|---\|---\|---\|---\|---\|---\|---\|---\|---\|---\|--- 1 \| U \| C \| B \| L \| M \| H \| I \| R \| U \| C \| B \| L \| M 1* \| M \| H \| I \| R \| U \| C \| B \| L \| M \| H \| I \| R \| U 2 \| B \| L \| M \| $H^{U}$ \| I \| $R^{B}$ \| U \| C \| B \| L \| M \| $H^{U}$ \| I 2* \| I \| $R^{B}$ \| U \| C \| B \| L \| M \| $H^{U}$ \| I \| $R^{B}$ \| U \| C \| B 3 \| B \| L \| M \| H \| I \| R \| U \| C \| B \| L \| M \| S \| I 3* \| I \| R \| U \| C \| B \| L \| M \| H \| I \| L \| U \| C \| B 4 \| M \| HU \| I \| RB \| U \| C \| B \| L \| M \| HU \| I \| RB \| U 4* \| U \| C \| B \| L \| M \| HU \| I \| RB \| U \| C \| B \| L \| M Table 1: Switching sequence for stages one and two of the photonic chip network. We have not illustrated stages three and four as the spatial arrangement of chips and the switching sequences are identical when examining a cross section the cluster along the $y$-axis (the switching sequence for stages three and four will have a $2T$ offset from the one shown above). The temporal location of each of the photons at $t=\\{T_{1},T_{2},T_{3},T_{4}\\}$ is indicated in Fig. 5. The labels $R,L,C,U,B$ refer to the geometric arrangement of the photons in the lattice stabilizer (i.e. right, left, center, upper and bottom) while the switching settings are $R=C=L=\text{central chip port}$, $U=\text{upper chip port}$, $B=\text{lower chip port}$, $M=\text{Measure the atomic system}$ $I=\text{Initialize the atomic system in the }(\|1\rangle+\|2\rangle)/\sqrt{2}\text{ state}$ and $H$ indicating to “hold” the atomic system in the $\|1\rangle$ state in order to delay a photon not involved in any stabilizer check by $T$. In chip layers two and four, temporally synchronous photons enter a particular chip, however only one of these photons is routed into the module. These stages are indicated by $\\{.\\}^{U,B}$, denoting the port (either U or B) that is routed through the bypass in that time step. Additionally, for a given column of chips, there are two switching sequences, (one denoted by ) which simply alternates down any given column. By examining Fig. 5 there is the potential for photon collisions for chip layers two and four. However, whenever two photons enter a chip simultaneously, one interacts with the module while the other is required to bypass the unit. Hence the notation used in Tab. 1, $\\{.\\}^{U,B}$, refers to switching the central port $\\{.\\}$ to the module while switching the photon in the (U)pper or (B)ottom port to a bypass line. ## IV Resource requirements. While we have only illustrated the network for a $5\times 5$ continuously generated lattice, the patterning of photonic chips and the switching sequence for each unit extends in two dimensions allowing for the continuous generation of an arbitrary large 3D lattice in a modular way [Fig. 2]. The total number of photonic chips required for the preparation of a large cross section of the lattice is easily calculated. In general, for an $N\times N$ cross section of cells, $4N^{2}+4N$ photonic chips are required to continually generate the lattice. We are able to choose the optimal clock cycle of the computer, $T$, as the fundamental atom/photon interaction time within each photonic module. As shown in Tab. 1, the switching sequence for the preparation network allows for a temporal window of $T$ for the measurement of the atomic system, within each device. As estimated in Devitt1 ; Su1 , depending on the system used, this rate can be approximately 10ns to 1$\mu$s. If we choose $T$ to be the optimal operating rate of the photonic module, then there is the potential that atomic measurement in the preparation network is too slow. This can be overcome by the availability of more photonic chips. If more chips are available then we construct multiple copies of the preparation network with optical switches placed between each preparation stage. While the atomic systems are being measured from one round of parity checks the next set of incoming photons are switched to a different group of chips. For a given ratio of the atom/photon interaction time to atomic measurement time, $T_{\text{atom}}/T_{\text{module}}\geq 1$, photons are routed to multiple copies of the preparation network. Therefore the number of additional photonic modules/chips will increase by a factor of $\Gamma=\lceil T_{\text{atom}}/T_{\text{module}}\rceil$ to compensate for slow measurement. Resource estimates are consequently related to the total 2D cross section of the lattice and $\Gamma$. Given that the number of photonic chips required in the preparation network, for a continuously generated $N\times N$ cross section of unit cells, is $4N^{2}+4N$, the number of fabricated photonic chips required when atomic measurement is slow will be approximately $\Gamma(4N^{2}+4N)$. To put these resource costs in context we can make a quick estimate of the resources required to build a quantum computer capable of solving interesting problems. Let us choose a logical error rate per time step of the lattice of $10^{-16}$ to be our target error rate, where a logical time step is the creation and measurement of a single layer of the cluster. This gives an approximate error rate, per logical non-Clifford $R_{z}(\pi/8)$ rotation of approximately $O(10^{-11})$ [Appendix B]. This error rate would be sufficiently low to enable the factoring of integers several thousand binary digits long using Shor’s algorithm Shor1 ; MI05 ; Z98 . Let us assume that increasing the separation between defects and the circumference of defects by two cells reduces the logical error rate per time step by a factor of 100 [Appendix A]. Given that the current threshold error rate of the 3D topological cluster state scheme is $6.7\times 10^{-3}$ Raussendorf3 , albeit in the absence of loss, this assumption is equivalent to assuming qubits are affected by un-correlated $Z$ errors with a probability between $10^{-4}$ and $10^{-5}$ per time step. Correlated errors, which can be produced by the photonic chip when preparing the cluster, effect qubits on two separate lattice structures defined by the cluster (the primal and dual lattices) which are corrected independently and can be treated as such Raussendorf6 . Complete failure of a photonic chip is heralded as the eigenvalue conditions for cells prepared by a faulty chip are not satisfied. Therefore the measurement of a small region the $x-y$ plane of the cluster (along the $z$-axis) will identify errors in cluster cells at a much higher rate than the rest of the computer, indicating chip-failure. Figure 6: Large scale lattice structure required for long term computation via topological cluster states. Here we illustrate a basic concept of the lattice structure that is required for lengthy operations of the topological optical computer (when examining the lattice along the $z$-axis). Based on the estimates in the main text, each logical qubit (defined via a pair of defects in the lattice) requires a $40\times 20$ array of cells. The layout for each logical qubit is shown on the left, with the larger lattice, comprising thousands (if not millions) of logical qubits shown on the right. Each logical qubit requires just over 3000 fabricated photonic chips. Assuming single qubit error rates in the $10^{-4}$ to $10^{-5}$ range, this lattice would be sufficient to permit approximately $10^{11}$ logical non-Clifford $R_{z}(\pi/8)$ operations. Given the above assumptions, the desired logical error rate per time step could be achieved with defects measuring $4\times 4$ cells in cross-section and separated by 16 cells. Fig. 6 shows a section of a semi-infinite lattice of sufficiently well error corrected logical qubits. Each logical qubit occupies a cross-sectional region of $40\times 20$ cells. To prepare such a lattice (for an arbitrary number of time steps and setting $\Gamma=1$) we require the fabrication of 3320 integrated photonic chips. This preliminary estimate illustrates that the resource requirements of this architecture are promising. The high fidelity construction of approximately $3\times 10^{3}$ photonic modules, per logical qubit, to prepare a lattice that has sufficient topological protection to perform on the order of $10^{11}$ non-Clifford logical operations is arguably a less significant challenge than the high fidelity manufacturing of other proposed quantum computer architectures. Other proposed systems not only require comparable (if not more) physical qubits to achieve the same error protection, but depending on the system they will also require interconnected quantum bus systems for qubit transport, very non-trivial classical control structures and most likely the fabrication of the entire computer at once, with little flexibility to expand the size of the computer as more resources become available. In Appendix B it is estimated that each $R_{z}(\pi/8)$ rotation requires approximately 250 logical qubits (required for ancilla state distillation in order to enact non-Clifford gates). It should be noted that the logical qubit requirements for non-Clifford gates are identical to all computational models employing state injection, state distillation and teleportation as a method for fault-tolerant universal computation. ## V Conclusions We have presented a detailed architecture for topological cluster state computation in optical systems. While this architectural design is not appropriate for all systems, it does contain does contain several key elements, easing the conceptual design of a large scale computer. For example: 1. 1. The utilization of a computational model fundamentally constructed from error correction, rather than implementing error correction codes on top of an otherwise independent computational model. 2. 2. The modular construction of the architecture, where architectural expansion is achieved via the addition of comparatively simple elements in a regular and known manner. 3. 3. Utilization of a computational model exhibiting high fault-tolerant thresholds and correcting otherwise pathological error channels (such as qubit loss). 4. 4. Utilization of a measurement based computational model. As the preparation network is designed simply to prepare the quantum resource, programming such a device is a problem of software, not hardware. By constructing this architecture with the photonic module and photonic chip as the primary operational elements, this design addresses many of the significant hurdles limiting the practical scalability of optical quantum computation. These include inherent engineering problems associated with probabilistic photon/photon interactions and the apparent intractability of designing a large scale, programmable system which can be scaled to millions of qubits. The experimental feasibility of constructing this type of computer is promising. High fidelity coupling of single photons to colour centers in cavities is a significant area of research in the quantum computing community and the nano engineering required to produce a high fidelity photonic module is arguably much simpler than a full scale atom/cavity based quantum computer which would require the fabrication of all data qubits and interconnected transport buses simultaneously. An additional benefit is the continuous nature of the lattice preparation. The required number of photonic modules and photonic chips only depends on the 2D cross sectional size of the lattice. This is important. Unlike other computational systems, this model does not require us to penalize individual photonic qubits within our physical resource analysis. Working under the assumption of appropriate on-demand sources, resource costs become a function of the total number of photonic chips (and single photon sources), rather that the total number of actual qubits required for a large 3D cluster. This highlights the advantages of having photons as disposable computational resources and we hope recasts the question of quantum resources into how many active quantum components are required to construct a large scale computer. ## VI Acknowledgments The authors thank C.-H. Su, Z. W. E. Evans and C. D. Hill for helpful discussions. The authors acknowledge the support of MEXT, JST, the EU project QAP, the Australian Research Council (Centre of Excellence Scheme and Fellowships DP0880466, DP0770715), and the Australian Government, the US National Security Agency, Army Research Office under Contract No. W911NF-05-1-0284. ## Appendix A Error Scaling of the Topological Lattice This appendix briefly reviews how to use the three-dimensional cluster state that is prepared by the network. In particular we outline techniques for error correction and universal computation and explain how logical failure rates are suppressed and how resource requirements are calculated. To extract the error syndrome, one measures each face qubit in every unit cell (see Fig. 4) in the $X$ basis. In the absence of $Z$ errors on these qubits, the parity of these six measurements will be even. A single $Z$ error on one face qubit, or an error during measurement, will flip the parity that is recorded from even to odd. When a chain of more than one error occurs, only cells at the end points of the chain will record odd parity. Note that no information about the path of any error chains is obtained during error correction and that error chains can terminate on boundaries of the cluster. Identifying the minimal set of $Z$ errors that is consistent with the syndrome proceeds by representing each odd parity unit cell as a node in a completely connected, weighted graph. The weight of the edge between any two nodes is the minimum number of errors that could connect the two cells which they represent. For each node in the graph a partner node is added which represents a boundary where an error chain could have terminated. The weight of the edge connecting each node to its partner node is the minimum number of errors that could connect the cell to the boundary. Partner nodes also form a fully connected graph with the weight between any two nodes equal to zero. The entire graph is then solved for the minimum weight pairing of the nodes. Each pair in the solution represents a $Z$ error or chain of $Z$ errors which can be corrected for with classical post-processing. We note that only $Z$ error correction is required to correct $Z$, $X$, measurement, initialization errors and photon loss. The measurements required for syndrome extraction are made in the $X$ basis and so are not affected by $X$ errors. Errors during preparation of the cluster state are equivalent to $Z$ errors on the prepared state. Photon loss could be treated as a random measurement result, equivalent to a measurement error or alternatively by calculating the parity of a closed surface of face qubits enclosing the lost photon or photons. This latter method is preferable given recent results establishing a high tolerance to heralded loss events Stace1 . A single logical qubit is associated with a pair of defects in the lattice. A defect is a connected volume of the lattice in which all qubits have been measured in the $Z$ basis. Two types of defects can be made - primal and dual. Primal defects are regions where one or more unit cells (see Fig. 4) have been measured. Dual defects are defined identically in the dual space of the lattice, where a center of a dual unit cell corresponds to a vertex of a primal unit cell. Valid operations on a single logical qubit include joining the pair of defects via a chain of physical $Z$ operations and encircling a single defect via a chain of physical $Z$ operations. Technically, these logical operations have additional $X$ operators associated with them to ensure anticommutation Fowler2 , however these do not affect $X$ basis measurements and thus can be ignored in the current discussion. Note that logical $X$ and $Z$ operations can be realized by simply tracing their effect through the circuit until logical measurement and then appropriately adjusting the measurement outcome, as opposed to performing gated operations directly on the physical qubits within the three-dimensional lattice. A controlled-not interaction between a dual and primal qubit can be effected by braiding one of the dual defects around one of the primal defects. Braiding is performed by changing the shape of the region of measurements defining the location of a defect. Logical CNOT, combined with initialization, readout, state injection, and state distillation allow universal quantum computation as detailed in Ref Fowler2 . Protection against logical errors is achieved by increasing the separation between defects and the circumference of defects. A logical error occurs when the error correction routine causes the application of a set of corrections that make or complete a chain of physical operations forming an erroneous logical operation. Just considering error chains joining defects, it is clear that the number of physical errors required to make such a chain increases linearly with the defect separation. The probability of forming such a chain therefore decreases exponentially with separation. As the distance of a code can be defined as the weight of its minimal weight logical operator, increasing the separation between defects by two cells increases the code distance by two. Increasing the circumference of defects has an analogous effect on the probability of logical errors in the conjugate basis. The efficacy of error correction depends on the amount by which the physical error rate, $p$, is below some threshold error rate. As the distance of a code is increased from $d$ to $d+2$, the encoded error rate is transformed from $n(d)p^{(d+1)/2}$ to $n(d+2)p^{(d+1)/2+1}$, where higher order terms are neglected and $n(d)$ and $n(d+2)$ are constants related to the code and the circuits required to extract the syndrome. Increasing the distance of the code will lower the encoded error rate only if $p<n(d)/n(d+2)$. If $p$ is a factor of $x$ below this threshold ($p=n(d)/xn(d+2)$), then the encoded error rate is reduced by a factor $n(d)p^{(d+1)/2}/n(d+2)p^{(d+1)/2+1}=n(d)/pn(d+2)=x$. In general, provided that the quantity $n(d)/n(d+2)$ is constant for all $d$, increasing the distance of a code by two will result in a reduction in the encoded error rate by a factor equal to the difference between $p$ and the threshold error rate. In our analysis of topological error correction we assume that $n(d)/n(d+2)$ is constant for all $d$. The threshold for the topological error correction code described above is $6.7\times 10^{-3}$, where qubit loss is neglected and where all other errors are assumed to be equally likely to occur Raussendorf5 . If we assume that the physical error rate is two orders of magnitude below this threshold, equal to $6.7\times 10^{-5}$, then increasing the separation of defects and the circumference of defects by two cells reduces the encoded error by a factor of 100. If our target encoded error rate, per time step, is $10^{-16}$ we require a minimum code distance of 17, which corresponds to a defect separation of 16 unit cells, or alternatively that we can correct error chains up to 8 errors long. Similarly the perimeter of the cells should also be 16, and hence we require defects that are $4\times 4$ cells in cross-section. This directly leads to the resource estimates per logical qubit given in the body of the paper. ## Appendix B Error rate on non-Clifford logical operations In the main body of the text, two logical error rates are presented. The first is the effective logical error rate per single layer of the cluster along the direction of simulated time [Appendix A] and the second is the effective error rate per non-Clifford rotation $R_{z}(\pi/8)$. We present these two values separately as the optimization of the measurement sequence for applying non- Clifford gates is still incomplete. This calculation instead provides a rough order of magnitude estimate of the effective non-Clifford error rate given the topological protection afforded by the lattice specified in the body of the paper. Within the topological model, only a subset of universal gates can be applied directly to the lattice, these include initialization and measurement in the $X$ and $Z$ basis, single qubit $X$ and $Z$ operations (although these are realized by tracing their effect through the circuit until logical measurement and then appropriately adjusting the results) and braided CNOT gates. Completing the universal set is achieved through state injection, magic state distillation and teleportation protocols Raussendorf5 ; BK05 . As non-Clifford rotations require the implementation of state distillation protocols, the effective logical gate rate is then dependent on the volume of cluster used to inject and distill a sufficiently high fidelity ancilla state for use in the teleportation protocol. In order to estimate the failure rate of a non-Clifford $R_{z}(\pi/8)$ operation, we examine the volume required to distill high fidelity ancilla states from low fidelity injected sources and to perform the required teleportation circuits to implement the $R_{z}(\pi/8)$ gate. Tab. 2 from Ref. Raussendorf5 specifies the volume of the 3D cluster that is required to implement five specific gate operations within the TCQC model, namely the CNOT, state distillation circuits for the singular qubit states, $\|Y\rangle=\frac{1}{\sqrt{2}}\left(\|0\rangle+i\|1\rangle\right),\quad\|A\rangle=\frac{1}{\sqrt{2}}\left(\|0\rangle+e^{i\frac{\pi}{4}}\|1\rangle\right)$ (3) and teleportation circuits to implement the single qubit gates, $P=\begin{pmatrix}1&0\\\ 0&i\end{pmatrix},\quad R_{z}(\pi/8)\equiv T=\begin{pmatrix}1&0\\\ 0&e^{i\frac{\pi}{4}}\end{pmatrix}$ (4) Gate \| Volume ---\|--- CNOT \| $V_{2}=12$ (16) Telegate, $T$ \| $V_{1,z}=2$ Telegate, $P$ \| $V_{1,x}=4$ $\|A\rangle$-distillation \| $V_{A}=336$ (168) $\|Y\rangle$-distillation \| $V_{Y}=120$ (60) Table 2: Total volume utilized in the 3D cluster to perform logical gates. In brackets are revised volume estimates Austin that are currently unpublished. Our calculations will assume the original volume estimates from Ref. Raussendorf5 , overestimating the total volume consumed for the gate $R_{z}(\pi/8)$. The volume estimates are given in terms of a scaled logical cell, where defects are now defined such that they have a sufficiently large circumference and separation to suppress the probability of logical failure, Fig. 7 (from Ref. Raussendorf5 ) illustrates. The cluster volumes quoted in Tab. 2 give the total number of logical cells required to implement specific gates. Fig. 7 shows defects of the same dimensions to Fig. 6, but translated along the $x-y$ plane of the lattice. A re-scaled logical cell is now a volume of $\lambda^{3}=20^{3}$ with a cylindrical defect of volume $d^{2}\lambda=16\times 20$ passing through its center. Figure 7: (From Ref. Raussendorf5 ) Cross section of a re-scaled logical cell providing topological protection. To achieve sufficient fault-tolerant protection, each elementary cell of the cluster [Fig. 4] is rescaled to a larger logical cell, where a central defect has sufficient circumference and separation from other defects to be highly protected against error. Shown here is the 2D cross section of a rescaled logical cell in terms of the lattice parameters utilized in the main text [Fig. 6]. To implement the non-Clifford gate, $T$, injected low fidelity states [Eq. 3] need to be purified using magic state distillation BK05 before teleportation circuits can be used to enact the gate. Shown in Fig. 8 is the quantum circuit required to perform a single qubit $T$ gate on an arbitrary state $\|\psi\rangle$ given an appropriate ancilla $\|A\rangle$. The measurement result of the ancilla qubit after the CNOT determines if the gate $T$ or $T^{\dagger}$ is applied. If the gate $T^{\dagger}$ is applied, then the further application of a single qubit $P$ gate transforms the rotation from $T^{\dagger}$ to $PT^{\dagger}=T$. As single qubit $P$ gates also require distilled ancilla and teleportation protocols and the application of the teleported gate, $T$, occurs with a probability of 0.5, every two $T$ gates within the quantum circuit will, on average, require the application of one $P$ gate. Hence not only do we need to distill one $\|A\rangle$ state, but we also need to distill “half” a $\|Y\rangle$ state. Figure 8: Standard Telegate circuit to implement single qubit $T$ rotation. Given the availability of the ancilla state $\|A\rangle$, this circuit allows for the teleported implementation of the $T$ gate on an unknown state, $\|\psi\rangle$. The measurement result, $M_{z}$, determines if the rotation $T$ or $T^{\dagger}\equiv R_{z}(-\pi/8)$ is applied. As this measurement result is random, with a $T$ gate applied 50% of the time, if $T^{\dagger}$ is applied a second teleportation circuit is run to implement the gate $P$ ($PT^{\dagger}=T$). Assuming that the residual error on all injected qubits within the lattice is equal to our assumed operational error rate of the computer $p=6.7\times 10^{-5}$, two concatenated levels of state distillation are required. To leading order in $p$, the recursion relations and success probabilities ($P_{A,Y}$) for $\|A\rangle$ and $\|Y\rangle$ state distillation, in the limit of negligible topological error, are given by, $\displaystyle p_{l+1}^{A}=35(p_{l}^{A})^{3},\quad p_{l+1}^{Y}=7(p_{l}^{Y})^{3},$ (5) $\displaystyle P_{l+1}^{A}=1-15p_{l}^{A},\quad P_{l+1}^{Y}=1-7p_{l}^{Y},$ for concatenation level $l$ Raussendorf6 . For $p_{0}^{A}=6.7\times 10^{-5}$ and $p_{0}^{Y}=6.7\times 10^{-5}$, the residual error after two rounds of distillation is $p_{2}^{A}\approx O(10^{-32})$ and $p_{2}^{Y}\approx O(10^{-35})$ and the probability of obtaining an incorrect syndrome result for each level is $[P_{1}^{A}\approx O(10^{-3})$, $P_{2}^{A}\approx O(10^{-10})]$ and [$P_{1}^{Y}\approx O(10^{-4})$, $P_{2}^{Y}\approx O(10^{-11})$]. In total, $15/(1-P_{1}^{A})+1/(1-P_{2}^{A})\approx 16$ distillation circuits are required for two levels of $\|A\rangle$ state distillation and $7/(1-P_{1}^{Y})+1/(1-P_{2}^{Y})\approx 8$ are required for two levels of $\|Y\rangle$ state distillation. Run in parallel, this requires a total volume of $16V_{A}+(1/2)\times 8V_{Y}=5856$ logical cells, where the factor of $1/2$ accounts for the probabilistic implementation of the $P$ gate. The teleportation circuit is then applied requiring a logical volume of $V_{1,z}+(1/2)V_{1,x}=4$ cells. Given that the scaling factor for the computer detailed in Fig. 6 is $\lambda=20$, the total number of elementary cells, in the direction of simulated time, is $\Omega=\lambda(5865+4)\approx 1.1\times 10^{5}$ and the failure rate of the logical $T$ gate is approximately, $1-(1-10^{-16})^{\Omega}\approx O(10^{-11})$ (6) Hence, this basic estimate illustrates that the logical non-Clifford failure rate for the topological computer is reduced by approximately five orders of magnitude from the logical failure rate per temporal layer. Note that this value for the logical gate assumes no specific optimization of non-Clifford group gates and represents a conservative estimate on the expected logical gate fidelity. In addition to the error rate of the logical gate, a single application of the $T$ gate consumes, on average, $15^{2}+7^{2}/2\approx 250$ logical qubits in the lattice. However, the qubit resources at the logical level are equivalent for all computational models employing state distillation and teleportation of $P$ and $T$ gates to achieve universality. ## References (1) Y. Nakamura, Yu. A. Pashkin and J.S. Tsai. Nature (London) 398, 786 (1999). * (2) I. Chiorescu, Y. Nakamura, C.J.P.M. Harmans and J.E. Mooij, Science 299, 1869 (2003). * (3) T. Yamamoto, Yu. A. Pashkin, O. Astafiev, Y. Nakamura and J.S. Tsai Nature (London) 425, 941 (2003). * (4) J. Chiaverini, D. Leibfried, T. Schaetz, M.D. Barrett, R.B. Blakestad, J. Britton, W.M. Itano, J.D. Jost, E. Knill, C. Langer, R. Ozeri and D.J. Wineland. Nature (London) 432, 602 (2004). * (5) H. Häffner, W. Hänsel, C.F. Roos, J. Benhelm, D. Chek-al-kar, M. Chwalla, T. Körber, U.D. Rapol, M. Riebe, P.O. Schmidt, C. Becher, O. Gühne, W. Dür and R. Blatt. Nature (London) 438, 643 (2005). * (6) J. Gorman, D.G. Hasko and D.A. Williams. Phys. Rev. A. 95, 090502 (2005). * (7) T. Gaebel, M. Domhan, I. Popa, C. Wittmann, P. Neumann, F. Jelezko, J.R. Rabeau, N. Stavrias, A.D. Greentree, S. Prawer, J. Meijer, J. Twamley, P.R. Hemmer and J. Wrachtrup. Nature (Physics) 2, 408 (2006). * (8) R. Hanson, F.M. Mendoza, R.J. Epstein, and D.D. Awschalom. Phys. Rev. Lett. 97, 087601 (2006). * (9) M.V. Gurudev Dutt, L. Childress, L. Jiang, E. Togan, J. Maze, F. Jelezko, A.S. Zibrov, P.R. Hemmer, and M.D. Lukin. Science 316, 1312 (2007). * (10) J.L. O Brien, G.J. Pryde, A.G. White, T.C. Ralph, and D. Branning. Nature (London) 426, 264 (2003). * (11) D. Kielpinski, C. Monroe, and D. J. Wineland. Nature 417, 709 (2002). * (12) J. M. Taylor, H.-A. Engel, W. Dür, A. Yacoby, C. M. Marcus, P. Zoller, and M. D. Lukin. Nature Phys. 1, 177 (2005). * (13) L. C. L. Hollenberg, A. D. Greentree, A. G. Fowler, and C. J. Wellard. Phys. Rev. B 74, 045311 (2006). * (14) A. G. Fowler, W. F. Thompson, Z. Yan, A. M. Stephens, B. L. T. Plourde, and F. K. Wilhelm. Phys. Rev. B 76, 174507 (2007). * (15) E. Knill, R. Laflamme, and G. J. Milburn. Nature 409, 46 (2001). * (16) P. Kok, W. J. Munro, K. Nemoto, T. C. Ralph, J. P. Dowling, and G. J. Milburn Rev. Mod. Phys. 79, 135 (2007). * (17) R. Raussendorf and J. Harrington, Phys. Rev. Lett. 98, 190504 (2007). * (18) R. Raussendorf, J. Harrington and K. Goyal, New J. Phys. 9, 199 (2007). * (19) A.G. Fowler and K. Goyal, arXiv:0805.3202 (2008). * (20) R. Raussendorf, H.J. Briegel, Phys. Rev. Lett. 86, 5188 (2001). * (21) M.A. Nielsen and I.L. Chuang, Quantum Information and Computation. Cambridge University Press (2000). * (22) C. Nayak, S.H. Simon, A. Stern, M. Freedman and S. Das Sarma, arXiv:0707.1889 (2007). * (23) S.B. Bravyi and A.Y. Kitaev, Quant. Computers and Computing 2, 43 (2001). * (24) E. Dennis, A.Y. Kitaev, A. Landahl and J. Preskull J. Math. Phys. 43, 4452 (2002). * (25) A.G. Fowler, A.M. Stephens and P. Groszkowski, arXiv:0803.0272 (2008). * (26) C.M. Dawson, H.L. Haselgrove and M.A. Nielsen, Phys. Rev. A, 73, 052306 (2006). * (27) T.C. Ralph, A.J.F. Hayes and A. Gilchrist, Phys. Rev. Lett. 95, 100501 (2005). * (28) M. Varnava, D.E. Browne and T. Rudolph, Phys. Rev. Lett., 100, 160502 (2008). * (29) C.W.J. Beenakker, D.P. DiVincenzo, C. Emary, and M. Kindermann, Phys. Rev. Lett. 93, 020501 (2004); * (30) K. Nemoto and W.J. Munro, Phys. Rev. Lett. 93, 250502 (2004). * (31) T.P. Spiller, K. Nemoto, S.L. Braunstein, W.J. Munro, P. van Loock, G.J. Milburn, New J. Phys. 8, 30 (2006). * (32) P. Domokos, J.M. Raimond, M. Brune and S. Haroche. Phys. Rev. A. 52, 3554 (1995). * (33) L.-M. Duan and H.J. Kimble. Phys. Rev. Lett. 92, 127902 (2004). * (34) S.J. Devitt, A.D. Greentree, R. Ionicioiu, J.L. O’Brien, W.J. Munro and L.C.L. Hollenberg Phys. Rev. A. 76, 052312 (2007). * (35) A.M. Stephens, Z.W.E. Evans, S.J. Devitt, A.G. Fowler, A.D. Greentree, W.J. Munro, J.L. O’Brien, K. Nemoto and L.C.L. Hollenberg, Phys. Rev. A., 78, 032318 (2008). * (36) C-H. Su, A.D. Greentree, W.J. Munro, K. Nemoto, L.C.L. Hollenberg Phys. Rev. A., 78, 062336 (2008). * (37) W.L. Yang, H. Wei, F. Zhou and M. Feng J. Phys. B: At. Mol. Opt. Phys. 42, 055503 (2009) * (38) D. Englund, A. Faron, I. Fushman, N. Stoltz, P. Petroff, and J. Vuckovic Nature (London), 450, 857 (2007). * (39) I. Fushman, D. Englund, A. Faron, N. Stoltz, P. Petroff, and J. Vuckovic Science, 320, 769 (2008). * (40) A. Faron, I. Fushman, D. Englund, N. Stoltz, P. Petroff, and J. Vuckovic Optics Express, 16, 12154 (2008). * (41) A. Politi, M.J. Cryan, J.G. Rarity, S. Yu and J.L. O’Brien Science 320, 646 (2008). * (42) A.S. Clark, J. Fulconis, J.G. Rarity, W.J. Wadsworth and J.L. O’Brien arXiv:0802.1676, (2008). * (43) G.D. Marshall, A. Politi, J.C.F. Matthews, P. Dekker, M. Ams, M.J. Withford and J.L. O’Brien arXiv:0902.4357, (2009). * (44) D. Gottesman, Ph.D Thesis (Caltech), quant-ph/9705052, (1997). * (45) P.W. Shor, SIAM J. of Computing, 26, 1484 (1997). * (46) R. Van Meter and K.M. Itoh Phys. Rev. A. 71, 052320 (2005). * (47) C. Zalka quant-ph/9806084 (1998). * (48) R. Raussendorf, J. Harrington and K. Goyal, Ann. Phys. 321, 2242 (2006). * (49) K. Nemoto and W.J. Munro, Phys. Lett. A. 344, 104 (2005). * (50) T.M. Stace, S.D. Barrett and A.C. Doherty, arXiv:0904.3556 (2009) * (51) S. Bravyi and A. Kitaev Phys. Rev. A. 71, 022316 (2005). * (52) A.G. Fowler et. al. in Preparation (2009). * (53) R. Raussendorf, J. Harrington and K. Goyal Ann. Phys. 321, 2242 (2006)	arxiv-papers	2008-08-13T05:50:05	2024-09-04T02:48:57.261107	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Simon J. Devitt, Austin G. Fowler, Ashley M. Stephens, Andrew D.\n Greentree, Lloyd C.L. Hollenberg, William J. Munro and Kae Nemoto", "submitter": "Simon Devitt Dr", "url": "https://arxiv.org/abs/0808.1782" }
0808.1806	# Ballistic propagation of thermal excitations near a vortex in superfluid 3He-B C. F. Barenghi1, Y. A. Sergeev2 and N. Suramlishvili1,3 1School of Mathematics and Statistics, Newcastle University, Newcastle upon Tyne, NE1 7RU 2School of Mechanical and Systems Engineering, Newcastle University, Newcastle upon Tyne, NE1 7RU 3Andronikashvili Institute of Physics, Tbilisi, 0177, Georgia ###### Abstract Andreev scattering of thermal excitations is a powerful tool for studying quantized vortices and turbulence in superfluid 3He-B at very low temperatures. We write Hamilton’s equations for a quasiparticle in the presence of a vortex line, determine its trajectory, and find under wich conditions it is Andreev reflected. To make contact with experiments, we generalize our results to the Onsager vortex gas, and find values of the intervortex spacing in agreement with less rigorous estimates. ###### pacs: 67.40.Vs Quantum fluids: vortices and turbulence, 67.30.em Excitations in He3 67.30.hb Hydrodynamics in He3 67.30.he Vortices in He3 ## I Motivation Superfluid turbulence consists of a disordered tangle of quantized vortex filaments which move under the velocity field of each otherDonnelly-book ; BDV-book . If the temperature, $T$ is sufficiently smaller than the critical temperature, $T_{c}$, then the normal fluid can be neglected and the vortices do not experience any friction effectBDV . The simplicity of the vortex structures (discrete vortex lines) and the absence of dissipation mechanisms, such as friction and viscosity, make superfluid turbulence a remarkable fluid system, particularly when compared to turbulence in ordinary fluids. Current experimental, theoretical and numerical investigations attempt to determine the similarities and the dissimilarities between superfluid turbulence and ordinary turbulence. Questions which are currently addressed concern (i) the existence of a Kolmogorov energy cascade at length scales larger than the typical intervortex spacingVinen-Niemela ; Hulton , (ii) the existence of a Kelvin wave cascade at length scales smaller than the Kolmogorov lengthKivotides-cascade ; Vinen-cascade ; Kozik-cascade ; Nazarenko-cascade followed by (iii) acoustic emission at even shorter length scalesVinen-sound ; Parker , (iv) the possible existence of a bottleneckNazarenko-bottleneck ; Svistunov-bottleneck between the Kolmogorov cascade and the Kelvin wave cascade, (v) the nature of the fluctuations of the observed vortex line densityRoche ; Roche-Barenghi ; Tabeling ; Bradley3 and (vi) their decaytowed-grid ; decay , (vii) whether there are two forms of turbulenceVinen-Kolmogorov , a structured one, which consists of many length scales (Kolmogorov turbulence), and an unstructured, more random one (Vinen turbulence), (viii) the effects of rotation on turbulenceFinne ; Mongiovi ; Tsubota ; Eltsov . Most of these questions refer to the important limit $T/T_{c}\ll 1$, where fundamental distinctions between a perfect Euler fluid and a superfluid becomes apparentBarenghi . Superfluid turbulence experiments are currently performed in both 4He Golov ; Roche-Barenghi ; McClintock ; Skrbek ; Roche and in 3He-B Bradley3 ; Fisher ; Finne ; Yano . In the last few years it has been recognized that, to make progress in answering the above questions, it is necessary to develop better measurement techniques which are suitable for turbulence in quantum fluids. In 4He, the application of the classical PIV methodVanSciver-piv ; Bewley-piv ; Kivotides-piv was a breakthrough. In 3He-B, a non–classical, powerful measurement technique which is suitable in the limit $T/T_{c}\ll 1$ is the Andreev scatteringFisher , developed at the University of Lancaster. This article is concerned with the Andreev scattering. The plan of the paper is the following. In Section II we shall describe the basic ideas behind the Andreev scattering and review what is a quantized vortex line. In Section III we shall write down the governing equations of motion. In Section IV we shall determine the ballistic trajectories of excitations in the vicinity of the velocity field of a vortex line, and, in Section V, we shall study the transport of heat by ballistic quasiparticles through a tangle of vortices. Section VI will apply our result to the current experiments. Finally, in Section VII, we shall draw the conclusions. ## II Andreev scattering and quantized vortices The study of the motion of quasiparticle excitations in a superfluid was pioneered by AndreevAndreev . Consider an excitation which moves in the direction of increasing excitation gap. The excitation propagates at constant energy, and gradually reaches the minimum of the rising excitation spectrum, where its group velocity becomes zero. Thereafter it retraces its path but as an excitation on the other side of the minimum. An incoming quasiparticle is thus reflected as a quasihole and an incoming quasihole is reflected as a quasiparticle. The effect is a consequence of the fact that the minimum of the energy spectrum of the excitation lies at nonzero momentum. The case of $p$-wave triplet pairing appropriate to superfluid 3He has been discussed by various authors studying the interaction of excitations with the boundariesKR , motion of quasiparticles through the $A$-$B$ phase boundary in 3HeYip , ballistic motion of quasi–particle in slow varying textural field of 3He-ALeggett , scattering of ballistic quasiparticles in 3He-B by a moving solid surface Guenault ; Enrico , and calculation of the friction force on quantized vorticesKopnin-Kravtsov ; Stone . Ref. Stone and Volovik are concerned with Andreev reflection within the vortex core and therefore apply to the bound states. Our concern is the propagation of thermal excitations outside vortex cores. Collisions between the quasiparticles can cause some spreading of the incoming beam. However, the spreading can be made arbitrary small by lowering the density of the excitations, that is to say, by lowering the temperature. At low enough temperatures the mean free path exceeds the dimensions of the experimental cell and we can consider undamped excitations moving along straight paths until they hit a boundary or any potential barrier, particularly a barrier formed by a vortex. Andreev reflection of excitations thus gives the opportunity to probe flows in superfluid 3He at ultra–low temperatures. The most fruitful and promising application of Andreev scattering is thus superfluid turbulence in 3He-B in the low temperature limit, that is to say for $T/T_{c}\leq 0.4~{}\textrm{K}$ Bradley3 ; Fisher ; Bradley1 ; Bradley2 ; QFS-Fisher . Superfluid 3He-B is described by a macroscopic wave function, called the order parameter, with a well defined phase $\phi$. The superfluid velocity ${\bf v}_{s}$ is proportional to the gradient of the phase, ${\bf v}_{s}=\frac{\hbar}{2m}\nabla\phi,$ (2.1) where $m$ is the mass of one 3He atom. Consequently, in contrast to classical fluids, superfluid motion is irrotational and vorticity exist only in the form of quantized vortices. Quantized vortices are line defects around which the phase $\phi$ changes by $2\pi$. The superfluid order parameter is distorted within the relatively narrow core of the vortex, and the superfluid flows around the core with speed which is inversally proportional to the distance from the vortex core. Since both the real and the imaginary parts of the order parameter are zero on the axis of a vortex, vortex lines can be considered as topological defects. Vortices cannot terminate in the middle of the flow, so they are either closed loops or extend to the walls. Superfluid turbulence consists of a tangle of quantized vortices. The complex flow field within the tangle acts as a potential barrier for quasiparticles, causing the Andreev reflection of a fraction of a beam of thermal excitations incident upon the tangle. The use of Andreev scattering as a visualization technique of ultra–low temperature turbulence requires to find out exactly what happens to a single quasiparticle which moves in the velocity field of a vortex, which is what we set out to do. ## III Equations of motion of thermal excitations Our first aim is to formulate, in the $(x,\,y)$-plane, the equations of motion of a single excitation moving in the velocity field of a single straight vortex which we assume to be fixed and aligned along the $z$-axis. We are thus concerned with a two-dimensional problem only. The quantities (here and below the numerical values of the quantities are taken at the $0$ bar pressureGreywall ) which are necessary to describe the motion of the excitation are the Fermi velocity, ${\mbox{$v_{F}$}}\approx 5.48\times 10^{3}~{}\rm cm/s$, the Fermi momentum, ${\mbox{$p_{F}$}}=m^{}{\mbox{$v_{F}$}}\approx 8.28\times 10^{-20}~{}\rm g~{}cm/s$, and the Fermi energy, ${\mbox{$\epsilon_{F}$}}={\mbox{$p_{F}$}}^{2}/(2m^{})\approx 2.27\times 10^{-16}~{}\rm erg$. The quantity ${\mbox{$\epsilon_{p}$}}=\frac{p^{2}}{2m^{}}-{\mbox{$\epsilon_{F}$}}$ (3.1) is the ”kinetic” energy of the excitation measured with respect to the Fermi energy, $\epsilon_{F}$, where $m^{}\approx 3.01\times m=1.51\times 10^{-23}~{}\rm g$ is the effective mass of the excitation, and ${\bf{p}}$ the momentum, $p=\|{\bf{p}}\|$. Let $\Delta_{0}$ be the magnitude of the superfluid energy gap. Near the vortex axis, at radial distances $r$ smaller than the zero–temperature coherence length $\xi_{0}=\hbar v_{F}/\pi\Delta_{0}\approx 0.8\times 10^{-5}~{}\rm cm$, the energy gap falls to zero and can be approximated by $\Delta(r)\approx\Delta_{0}\tanh(r/\xi_{0})$ Bardeen ; TsunetoBook . Since we are mainly concerned with what happens for $r\gg\xi_{0}$, we neglect the spatial dependence of the energy gap and assume the constant value $\Delta_{0}=1.76k_{B}T_{c}\approx 2.43\times 10^{-19}~{}\rm erg$. Using polar coordinates $(r,\,\phi)$ in the $(x,\,y)$ plane , the velocity field of a superfluid vortex set along the $z-$axis is ${\bf{v}}_{s}=\frac{\kappa}{2\pi r}{\mbox{\boldmath$\hat{e}_{\phi}$}},$ (3.2) where $\kappa=\frac{h}{2m}=\frac{\pi\hbar}{m}=0.662\times 10^{-3}~{}\textrm{cm}^{2}\textrm{/s}$ (3.3) is the quantum of circulation, and $\hat{e}_{\phi}$ is the unit vector in the azimuthal direction on the $(x,\,y)$-plane. In the presence of the vortex, the energy of the excitation becomes $E=\sqrt{{\mbox{$\epsilon_{p}$}}^{2}+\Delta_{0}^{2}}+{\bf{p}}\cdot{\bf{v}}_{s}.$ (3.4) In writing Eq. (3.4), the spatial variation of the order parameter is not taken into account for the sake of simplicity. We also assume that the interaction term ${\bf{p}}\cdot{\bf{v}}_{s}$ varies on a spatial scale which is larger than $\xi_{0}$, and that the excitation can be considered a compact object of momentum ${\bf{p}}={\bf{p}}(t)$, position ${\bf{r}}={\bf{r}}(t)$, and energy $E=E({\bf{p}},\,{\bf{r}})$. This gives us the opportunity to use the method developed in Ref.Leggett , and consider Eq. (3.4) as an effective Hamiltonian, for which the equations of motion are $\frac{d{\bf{r}}}{dt}=\frac{\partial E({\bf{p}},{\bf{r}})}{\partial{\bf{p}}}=\frac{{\mbox{$\epsilon_{p}$}}}{\sqrt{{\mbox{$\epsilon_{p}$}}^{2}+\Delta_{0}^{2}}}\frac{{\bf{p}}}{m^{}}+{\bf{v}}_{s},$ (3.5) $\frac{d{\bf{p}}}{dt}=-\frac{\partial E({\bf{p}},{\bf{r}})}{\partial{\bf{r}}}=-\frac{\partial}{\partial{\bf{r}}}({\bf{p}}\cdot{\bf{v}}_{s}).$ (3.6) Eqs. (3.5) and (3.6) have one immediate integral of motion, the energy: $E({\bf{p}},\,{\bf{r}})=E=\textrm{constant}.$ (3.7) Eq. (3.5) represents the group velocity of the excitation in the velocity field of the vortex. Excitations such that ${\mbox{$\epsilon_{p}$}}>0$ are called quasiparticles, and excitations such that ${\mbox{$\epsilon_{p}$}}<0$ are called quasiholes. The right-hand-side of Eq. (3.6) is thus the force acting on the excitation. ## IV Propagation of excitation in the velocity field of a vortex We want to determine the trajectory of an excitation which moves in the two- dimensional velocity field of the vortex. It is convenient to rewrite the Hamiltonian, Eq. (3.4), and Hamilton’s equations (3.5) and (3.6) in polar cordinates $(r,\,\phi)$. We notice that the system consisting of a single excitation and a single vortex has a second integral of motion: the component of the angular momentum in the $z$-direction, perpendicular to the plane of motion, $(x,\,y)$. Consequently, we can introduce two pairs of canonically conjugated variables, $(p_{r};\,r)$ and $(J=p_{\phi}r;\,\phi)$, where $p_{r}$ and $p_{\phi}$ are the radial and azimuthal components of ${\bf{p}}$, and $J$ is the angular momentum. Since $J$ is constant, it is convenient to write it in the form $J={\mbox{$p_{F}$}}\rho_{0}$, thereby defining the constant $\rho_{0}$ for a particular trajectory. Under special initial conditions, as we shall see, $\rho_{0}$ can be interpreted as the impact parameter. Eqs. (3.4), (3.1) and (3.5) become $E=\sqrt{{\mbox{$\epsilon_{p}$}}^{2}+\Delta_{0}^{2}}+{\mbox{$p_{F}$}}\rho_{0}\frac{\kappa}{2\pi r^{2}}\,,$ (4.1) ${\mbox{$\epsilon_{p}$}}=\frac{p_{r}^{2}}{2m^{}}+\frac{({\mbox{$p_{F}$}}\rho_{0})^{2}}{2m^{}r^{2}}-{\mbox{$\epsilon_{F}$}},$ (4.2) ${\dot{r}}=\frac{dr}{dt}=\frac{{\mbox{$\epsilon_{p}$}}}{\sqrt{{\mbox{$\epsilon_{p}$}}^{2}+\Delta_{0}^{2}}}\frac{p_{r}}{m^{}}\,,$ (4.3) ${\dot{\phi}}=\frac{d\phi}{dt}=\frac{{\mbox{$\epsilon_{p}$}}}{\sqrt{{\mbox{$\epsilon_{p}$}}^{2}+\Delta_{0}^{2}}}\frac{{\mbox{$p_{F}$}}\rho_{0}}{m^{}r^{2}}+\frac{\kappa}{2\pi r^{2}}\,.$ (4.4) By setting $dE/dt=0$ and using Eq. (4.3) we find $\dot{{\mbox{$\epsilon_{p}$}}}=\frac{d{\mbox{$\epsilon_{p}$}}}{dt}={\mbox{$p_{F}$}}\rho_{0}\frac{p_{r}}{m^{}}\frac{\kappa}{\pi r^{3}}\,,$ (4.5) and from Eq. (4.2) we have $\|p_{r}\|={\mbox{$p_{F}$}}\left(1+\frac{{\mbox{$\epsilon_{p}$}}}{{\mbox{$\epsilon_{F}$}}}-\frac{\rho_{0}^{2}}{r^{2}}\right)^{1/2}\,.$ (4.6) Eqs. (4.1)-(4.6) form a closed set which allows us to determine the trajectory of the excitation. It is apparent from Eq. (4.3) that a quasiparticle incident upon the vortex has ${\mbox{$\epsilon_{p}$}}>0$ and $p_{r}<0$, whereas a quasiparticle moving away from the vortex has ${\mbox{$\epsilon_{p}$}}>0$ and $p_{r}>0$. Vice–versa, a quasihole incident upon the vortex is characterized by ${\mbox{$\epsilon_{p}$}}<0$ and $p_{r}>0$, whereas a quasihole moving away from the vortex has ${\mbox{$\epsilon_{p}$}}<0$ and $p_{r}<0$. Later we shall consider a quasiparticle which leaves a point of the wall of the cylindrical experimental cell; this quasiparticle is initially characterized by $r=R$ (where $R$ is the radius of the cell), $p={\mbox{$p_{F}$}}$ and $p_{r}<0$. The axis of the vortex will still be at the centre of the coordinate system. In such case the quasiparticle with initial momentum directed along the $x$-axis will feel the effective pairing potential $\Delta_{eff}\approx\Delta_{0}-{\mbox{$p_{F}$}}y\kappa/(2\pi r)$ (Fig. 1). It is obvious from Eq. (4.3) that unless $\rho_{0}$ is exactly zero ($J=0$), the radial velocity of the excitation will eventually vanish. This may happen either because $p_{r}=0$ (classical turning point) or because ${\mbox{$\epsilon_{p}$}}=0$ (Andreev turning point). It can be seen from Eqs. (4.1) and (4.6) that the classical turning point is reached first when $E>\Delta_{0}+{\mbox{$p_{F}$}}\frac{\kappa}{2\pi\rho_{0}}\approx\Delta_{0}\biggl{(}1+\frac{3\pi\xi_{0}}{2\rho_{0}}\biggr{)}$ (4.7) (here and in the equations below the numerical factor 3 is introduced by the ratio between the effective mass of quasiparticle and the bare mass of a 3He atom: $m^{}/m\approx 3$) in which case a quasiparticle with this energy follows a trajectory which is of the ”normal” type: the quasiparticle retains its ”particle” nature and moves past the vortex, across the experimental cell to the wall on the opposite side. On the contrary, a quasiparticle with energy $E$ such that $\Delta_{0}<E<\Delta_{0}+{\mbox{$p_{F}$}}\frac{\kappa}{2\pi\rho_{0}}\approx\Delta_{0}\biggl{(}1+\frac{3\pi\xi_{0}}{2\rho_{0}}\biggr{)}$ (4.8) reaches the Andreev turning point first, undergoes Andreev reflection, and returns to a point near its starting point after changing its nature and becoming a quasihole. Of these two cases, our concern is the case of Andreev reflection. We first determine the locus of Andreev turning points, defined by the minimum radial distance from the vortex core: $r_{min}=\left(\frac{\kappa}{2\pi}\frac{{\mbox{$p_{F}$}}\rho_{0}}{(E-\Delta_{0})}\right)^{1/2}=\left(\frac{3\pi\xi_{0}\rho_{0}}{2}\frac{\Delta_{0}}{(E-\Delta_{0})}\right)^{1/2}.$ (4.9) Consider a quasiparticle which has reached $r=r_{min}$. At this point the radial velocity $\dot{r}$ vanishes, but the excitation does not stop. It has still a nonzero azimuthal velocity, $r\dot{\phi}$. Thereafter the excitation propagates as a quasihole (characterized by a negative value of $\epsilon_{p}$). In order to calculate the trajectory of a reflected quasiparticle it is convenient to simplify the governing equations of motion using the fact that at the ultra–low temperatures which interest us, $T\ll T_{c}$, most quasiparticles have energies ${\mbox{$\epsilon_{p}$}}\ll\Delta_{0}$. We can then make the following approximation: $\sqrt{{\mbox{$\epsilon_{p}$}}^{2}+\Delta_{0}^{2}}\approx\Delta_{0}+\frac{{\mbox{$\epsilon_{p}$}}^{2}}{2\Delta_{0}}=\Delta_{0}+\frac{(p^{2}-p_{F}^{2})^{2}}{8m^{2}\Delta_{0}}\approx\Delta_{0}+\frac{(p-p_{F})^{2}}{2\Delta_{0}/v_{F}^{2}}\,.$ (4.10) This spectrum is similar to Landau’s spectrum of excitations in superfluid He II near the roton minimum ($p=p_{0}$), $E\approx\Delta_{0}+(p-p_{0})^{2}/(2m_{r})$ (where $m_{r}$ is the effective roton mass), which was used to calculate the mutual friction force Samuels ; note that in Eq. (4.10) the role of the roton mass is played by the ratio $\Delta_{0}/v_{F}^{2}$. Using Eqs. (4.10), (4.3) and (4.4), and the smallness of the ratios ${\mbox{$\epsilon_{p}$}}/{\mbox{$\epsilon_{F}$}}$ and $\Delta_{0}/{\mbox{$\epsilon_{F}$}}$, we obtain $dt=\frac{m^{}\Delta_{0}}{{\mbox{$\epsilon_{p}$}}p_{r}}dr=-\frac{m^{}}{{\mbox{$p_{F}$}}}\frac{r_{min}}{(3\pi\xi_{0}\rho_{0})^{1/2}}\frac{r^{2}dr}{(r^{2}-r_{min}^{2})^{1/2}(r^{2}-\rho_{0}^{2})^{1/2}}\,,$ (4.11) $d\phi=-\left(\frac{\rho_{0}}{r^{2}(1-\rho_{0}^{2}/r^{2})^{1/2}}\pm\frac{3br_{min}}{2(3\pi\xi_{0}\rho_{0})^{1/2}(r^{2}-r_{min}^{2})^{1/2}(r^{2}-\rho_{0}^{2})^{1/2}}\right)dr,$ (4.12) where $b=\hbar/{\mbox{$p_{F}$}}$, the sign plus is used for quasiparticles and the sign minus for quasiholes. From Eqs. (4.11) and (4.12) we obtain the Andreev return time $\tau$ of the excitation (the time it takes to travel from the radial distance $R$ to the Andreev reflection point and back) and the Andreev reflection angle $\Delta\phi$: $\tau=2\frac{r_{min}}{{\mbox{$v_{F}$}}}\frac{1}{(3\pi\xi_{0}\rho_{0})^{1/2}}\int_{r_{min}}^{R}\frac{r^{2}dr}{(r^{2}-r_{min}^{2})^{1/2}(r^{2}-\rho_{0}^{2})^{1/2}}\,,$ (4.13) $\Delta\phi=3\frac{br_{min}}{(3\pi\xi_{0}\rho_{0})^{1/2}}\int_{r_{min}}^{R}\frac{dr}{(r^{2}-r_{min}^{2})^{1/2}(r^{2}-\rho_{0}^{2})^{1/2}}\,.$ (4.14) The evaluation of these elliptic integrals is shown in the Appendix. We obtain $\tau=2\frac{r_{min}}{{\mbox{$v_{F}$}}}\frac{1}{(3\pi\xi_{0}\rho_{0})^{1/2}}\left(\frac{(R^{2}-r_{min}^{2})^{1/2}(R^{2}-\rho_{0}^{2})^{1/2}}{R}+\frac{\pi}{4}\left(\frac{\rho_{0}}{r_{min}}\right)^{1/2}r_{min}\right)\,,$ (4.15) which becomes, assuming $R\gg r$ and $R\gg\rho_{0}$, $\tau\approx 2\frac{Rr_{min}}{{\mbox{$v_{F}$}}(3\pi\xi_{0}\rho_{0})^{1/2}}=\frac{R}{{\mbox{$v_{F}$}}}\left(\frac{2\Delta_{0}}{E-\Delta_{0}}\right)^{1/2}\approx\frac{R}{{\mbox{$v_{F}$}}}\frac{2\Delta_{0}}{{\mbox{$\epsilon_{p}$}}}.$ (4.16) We conclude that the Andreev return time is longer if the excitation’s energy is lower. Similarly, assuming $\rho_{0}/R\ll 1$ and $r_{min}/R\ll 1$, the Andreev reflection angle is $\Delta\phi\approx\frac{\pi b}{(3\pi\xi_{0}\rho_{0})^{1/2}}\,.$ (4.17) To apply these results we assume that the initial momentum of the quasiparticle is directed along the $x$-axis, and that the angular momentum $J=-py_{0}={\mbox{$p_{F}$}}\rho_{0}$. From Eq. (3.1) it follows that the momentum $p={\mbox{$p_{F}$}}(1+2m^{}{\mbox{$\epsilon_{p}$}}/{\mbox{$p_{F}$}})^{1/2}$ and, in the ultra-low temperature limit, $(p-{\mbox{$p_{F}$}})/{\mbox{$p_{F}$}}\leq 10^{-4}$. For $y_{0}$ we have $y_{0}=\rho_{0}(1+2m^{}{\mbox{$\epsilon_{p}$}}/{\mbox{$p_{F}$}})^{-1/2}\approx\rho_{0}$. In this case $\rho_{0}$ becomes the impact parameter (Fig. 2), and Eq. (4.17) shows that quasiparticles with smaller impact parameter (hence smaller angular momentum) are Andreev reflected by smaller angles. As it is seen from Eq. (4.9), the Andreev radius depends strongly on the initial energy of the excitation: $r_{min}=(3\pi\xi_{0}\rho_{0})^{1/2}\frac{\Delta_{0}}{{\mbox{$\epsilon_{p}$}}}\,.$ (4.18) The same arguments apply to the critical value $\rho_{0c}$ defined as a maximum value of $\rho_{0}$ which causes the Andreev reflection of quasiparticles with the given initial energy $\epsilon_{p}$. To calculate $\rho_{0c}$, we assume that at the starting point of the trajectory the quasiparticle has coordinates $(R,\,\phi_{0})$, where $\phi_{0}=\arcsin(y_{0}/R)\approx-\arcsin(\rho_{0c}/R)$; the coordinates of the Andreev reflection point in this case should be $(r_{min},\,-\pi/2)$. Thus the difference between the reflection angle and the starting angle is $\Delta\phi=-\frac{\pi}{2}+\arcsin\left(\frac{\rho_{0c}}{R}\right).$ (4.19) This difference can also be calculated from Eq. (4.12) where the second term (of the order of $\hbar/{\mbox{$p_{F}$}}$) in the integrand can be neglected. We obtain $\Delta{\phi}=-\arcsin\left(\frac{\rho_{0c}}{r_{min}}\right)+\arcsin\left(\frac{\rho_{0c}}{R}\right)\,.$ (4.20) By comparing Eqs. (4.19) and (4.20) we find $\rho_{0c}\approx 3\pi\xi_{0}\left(\frac{\Delta_{0}}{{\mbox{$\epsilon_{p}$}}}\right)^{2}\,.$ (4.21) In the typical low temperatures experiments $k_{B}T/\Delta_{0}\approx 0.1$, and, for quasiparticles with initial energy ${\mbox{$\epsilon_{p}$}}\approx k_{B}T$, we find $r_{min}\approx 10(3\pi\xi_{0}\rho_{0})^{1/2}$ and $\rho_{0c}\sim 10^{3}\xi_{0}$, while the same quantities for the quasiparticles with ${\mbox{$\epsilon_{p}$}}\approx(\Delta_{0}k_{B}T)^{1/2}$ are $r_{min}\sim 3(3\pi\xi_{0}\rho_{0})^{1/2}$ and $\rho_{0c}\sim 10^{2}\xi_{0}$. ## V Heat transport through the velocity field of a vortex In the experimental studies of superfluid turbulence in 3He-B at the ultra-low temperatures the vortex tangle is studied by detecting the fraction of quasiparticles which are Andreev reflected by the vortices and measuring the heat which is transported by the quasiparticles. Using the results of previous Sections, it is straightforward to calculate the fraction of energy (or heat) transmitted across the velocity field of a vortex. Once we know this fraction, we shall generalize it to a system of many vortices. In Section IV it was explained that the quasiparticles characterized by the particular impact parameter $\rho_{0}$ are Andreev reflected by a vortex if their energies satisfy the condition $\Delta_{0}\leq E\leq\Delta_{0}(1+3\pi\xi_{0}/2\rho_{0})$. If this condition is not satisfied, the quasiparticles pass freely across the vortex velocity field. If the system which consists of the vortex and quasiparticles is in thermal equilibrium, there is no preferred direction around the vortex. Incident and transmitted fluxes at one side of the vortex are canceled by the fluxes in the opposite direction, and no net flow of energy exists when the temperature everywhere around the vortex has the same value. A net flux of quasiparticles and of energy results only if there is some small temperature difference, $\delta T\ll T$ between the two sides. If this is the case, the heat carried by the incident quasiparticles is described by the expression: $\delta Q_{inc}=\int_{\Delta_{0}}^{\infty}N(E)v_{g}(E)E\frac{\partial f(E)}{\partial T}\,\delta T\,dE,$ (5.1) where $N(E)=N_{F}\frac{E}{(E^{2}-\Delta^{2})^{1/2}}\,.$ (5.2) Here $N_{F}=\frac{mp_{F}}{\pi^{2}\hbar^{3}}$ (5.3) is the density of states at the Fermi energy with corresponding Fermi momentum $p_{F}$. The group velocity of a Bogolubov quasiparticle $v_{g}$ is given by the expression: $v_{g}=\frac{{\mbox{$\epsilon_{p}$}}}{E}{\mbox{$v_{F}$}}=\frac{(E^{2}-\Delta^{2})^{1/2}}{E}{\mbox{$v_{F}$}}\,,$ (5.4) and $f(E)$ is the Fermi distribution function, which, at ultra–low temperatures, is transformed into the Boltzman distribution, and describes the mean occupation number of a state with energy $E$: $f(E)=e^{-\frac{E}{k_{B}T}}.$ (5.5) The thermal flux of quasiparticles incident on the vortex velocity field per unit length per unit time is obtained with the help of Eqs. (5.2), (5.4) and (5.5); one has $\delta Q_{inc}=N_{F}v_{F}\frac{\delta T}{k_{B}T^{2}}\int_{\Delta}^{\infty}E^{2}e^{-\frac{E}{k_{B}T}}\,dE\approx N_{F}v_{F}\Delta_{0}^{2}\frac{\delta T}{T}e^{-\frac{\Delta_{0}}{k_{B}T}}.$ (5.6) If there is a plane current of quasiparticles with transverse cross section $R_{0}$, then the total heat current incident on the vortex per unit time will be: $Q_{inc}=2R_{0}\delta Q_{inc}=2R_{0}N_{F}v_{F}\Delta_{0}^{2}\frac{\delta T}{T}e^{-\frac{\Delta_{0}}{k_{B}T}}.$ (5.7) We assume that, in the $(x,\,y)$-plane orthogonal to the straight vortex line, the polarity of the vortex located at $(0,\,0)$ is positive and consider quasiparticles incoming in the positive $x-$direction. As discussed earlier, in this case the upper half-plane will be absolutely transparent for quasiparticles so that the heat transferred by quasiparticles through this half-plane will meet no resistance. The lower half-plane of vortex flow field will reflect a fraction of quasiparticles and induce some thermal resistance. A quasiparticle with the impact parameter $\rho_{0}$ is transmitted through the vortex velocity field if it carries the energy $E>\Delta_{0}(1+\frac{3}{2}\pi\xi_{0}/\rho_{0})$, in which case the heat which is transferred per unit time by such a quasiparticle can be calculated as $\delta Q(\rho)=\int_{\Delta_{0}(1+\frac{3\pi\xi_{0}}{2\rho_{0}})}^{\infty}N(E)v_{g}(E)E\frac{\partial f(E)}{\partial T}\,\delta T\,dE\simeq Q_{inc}\frac{1}{2R_{0}}\biggl{(}1+\frac{3\pi\xi_{0}}{\rho_{0}}\biggr{)}e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{2\rho_{0}}}.$ (5.8) Notice that estimating the ratio $\xi_{0}/\rho_{0}$ we kept only the linear term. The total amount of energy transferred through the vortex by quasiparticles originated within the interval $-R_{0}\leq y\leq R_{0}$ is: $Q_{tr}=\frac{Q_{inc}}{2}\biggl{[}1+\frac{1}{R_{0}}\int_{0}^{R_{0}}\left(1+\frac{3\pi\xi_{0}}{\rho_{0}}\right)e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{2\rho_{0}}}d\rho_{0}\biggr{]}\,.$ (5.9) The integral in Eq.(5.9) can be estimated as $\approx R_{0}e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{2R_{0}}}.$ (5.10) Thus the fraction of heat which is transferred through the velocity field of the vortex is $\delta f_{tr}=\frac{1}{2}\biggl{(}1+e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{2R_{0}}}\biggr{)}.$ (5.11) In experiments at ultra–low temperatures we have $\Delta_{0}/k_{B}T\sim 10$, so that the cross-section of the thermal flux is $R_{0}\sim 10\xi_{0}$ and approximately $52\%$ of the total heat is transferred through the vortex. If the heat current has the cross-section $\sim 10^{2}\xi_{0}$, the fraction of the transferred heat is approximately $0.82$. Therefore the reflection of the heat flux takes place only in the close vicinity of the vortex core. ## VI Andreev reflection in a vortex gas. To apply our result to experiments, we consider for simplicity a system of random parallel-antiparallel vortices (i.e. a system of vortex points in the $(x,\,y)$-plane; such a system is known as the Onsager vortex gas). This vortex system is penetrated by a quasiparticle current created by a temperature difference $\delta T$. It is convenient to introduce the effective radius $R_{0}$ of each vortex as the half of the mean intervortex distance, i.e. $R_{0}=\ell/2$. We divide the vortex configuration in parallel layers of width $\ell$ each perpendicular to the quasiparticle current. Clearly, the transmittability of each layer is equal to the transmittability of a vortex within a region of radius $\ell/2$. Thus the fraction of heat transmitted by each layer is $\delta f_{tr}=\frac{1}{2}\biggl{(}1+e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{\ell}}\biggr{)}\,.$ (6.1) If we assume now that vortices are well separated and that their velocity fields do not overlap significantly, we obtain the conditions $\xi_{0}\ll\ell,\quad\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{\ell}\ll 1.$ (6.2) Eq. (6.1) becomes $\delta f_{tr}\approx\frac{1}{2}\biggl{(}1+1-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{\ell}\biggr{)}=1-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{2\ell}\approx e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}}{2\ell}}.$ (6.3) Driven by the temperature difference, the heat flux $Q_{0}$ reduces, after penetrating the first layer, to $Q_{1}=Q_{0}\,\delta f_{tr}$; after penetrating the second layer, it becomes $Q_{2}=Q_{1}\,\delta f_{tr}$. Hence, after penetrating the last $n$th layer, we obtain $Q_{n}=Q_{n-1}\,\delta f_{tr}$. Thus we have $Q_{n}=Q_{n-1}\,\delta f_{tr}=...=Q_{0}\,\delta f_{tr}^{n}.$ (6.4) We conclude that the fraction of heat which is transferred through the system of vortices is $f_{tr}=(\delta f_{tr})^{n}.$ (6.5) If the total vorticity is confined within a region of size $S$, the number of layers, $n$ can be estimated as a $n\approx S/\ell$. From Eq. (6.5) we obtain: $f_{tr}=e^{-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}S}{2\ell^{2}}}.$ (6.6) Finally we obtain the intervortex distance: $\ell=\biggl{(}-\frac{\Delta_{0}}{k_{B}T}\frac{3\pi\xi_{0}S}{2\ln f_{tr}}\biggr{)}^{\frac{1}{2}}.$ (6.7) The quantities $S$ (the size of the vortex system) and $f_{tr}$ (the fraction of reflected quasiparticles) in Eq. (6.7) can be observed experimentally. From the available description of one experimentFisher , the maximum transmitted fraction of quasiparticle current is $f_{tr}\approx 0.75$ and the spatial extension of the vorticity is $S\sim 2\cdot 10^{-1}\,\textrm{cm}$. Since the zero temperature coherence length is $\xi_{0}\approx 0.8\times 10^{-5}\,\textrm{cm}$, we conclude that in the case where $\Delta_{0}/k_{B}T\sim 10$ the average intervortex distance is $\ell\sim 1.62\cdot 10^{-2}~{}\textrm{cm}$, which is in good agreement with existing estimatesBradley1 . ## VII Conclusions In conclusion, starting from Hamilton’s equations, we have calculated the trajectories of quasiparticles which move in the velocity field of a quantized vortex in 3He-B and determined the Andreev reflection point. Generalizing the result to a disordered system of many vortices, we have determined the precise location of turning point and showed how to recover the typical intervortex spacing in the turbulent 3He-B. Our result is in good agreement with less rigorous estimates. Future work will investigate Andreev reflection of quasiparticles by a system of moving vortices. We shall also study how the Andreev reflection technique can be used to visualize vortex structures (e.g. coherent bundles of vortices) and determine turbulent fluctuations and turbulence statistics. ## VIII Acknowledgments This work was supported by EPSRC grant GR/T08876/01. NS also was supported by Georgian National Science Foundation grant GNSF/ST06/4-018. We are also grateful to Professor S.N. Fisher for stimulating discussions and for reading the manuscript. ## Appendix A The Andreev return time, $\tau$, and the Andreev reflection angle, $\Delta\phi$, are defined by formulae (4.13) and (4.14), where $0<\rho_{0}<r_{min}<r<R$. To evaluate these formulae we use the following integrals: $I_{1}=\int_{r_{min}}^{R}\frac{r^{2}\,dr}{(r^{2}-r_{min}^{2})^{1/2}(r^{2}-\rho_{0}^{2})^{1/2}}=\frac{\sqrt{R^{2}-r_{min}^{2}}}{\sqrt{R^{2}-\rho_{0}^{2}}}+r_{min}G,$ (A.1) where $G=K\left(\frac{\rho_{0}}{r_{min}}\right)-F\left(\arcsin{\frac{r_{min}}{R}},\,\frac{\rho_{0}}{r_{min}}\right)-E\left(\frac{\pi}{2},\,\frac{\rho_{0}}{r_{min}}\right)+E\left(\arcsin{\frac{r_{min}}{R}},\,\frac{\rho_{0}}{r_{min}}\right),$ (A.2) and $I_{2}=\int_{r_{min}}^{R}\frac{dr}{(r^{2}-r_{min}^{2})^{1/2}(r^{2}-\rho_{0}^{2})^{1/2}}=\frac{1}{r_{min}}\left[K\left(\frac{\rho_{0}}{r_{min}}\right)-F\left(\arcsin{\frac{r_{min}}{R}},\,\frac{\rho_{0}}{r_{min}}\right)\right]\,,$ (A.3) where $K$, $F$ and $E$ are elliptic integrals, defined as $F(k,\,\theta)=\int_{0}^{\theta}\frac{d\phi}{\sqrt{1-k^{2}\sin^{2}{(\phi)}}}\,,$ (A.4) $K(k)=F\left(\frac{\pi}{2},\,k\right),$ (A.5) $E(k,\,\theta)=\int_{0}^{\theta}\sqrt{1-k^{2}\sin^{2}{(\theta)}}\,d\phi,$ (A.6) with $\theta=\arcsin{(r_{min}/R)}$ and $k=\rho_{0}/R$. For $k^{2}<1$ the elliptic integrals (A.4), (A.5) and (A.6) are represented by the series $F(k,\,\theta)=\frac{2\theta}{\pi}K(k)-\sin{\theta}\cos{\theta}\frac{k^{2}}{4}+...\,,$ (A.7) $K(k)=\frac{\pi}{2}+\frac{\pi^{2}}{8}k^{2}+...\,,$ (A.8) $E(k,\,\theta)=\frac{2\theta}{\pi}E(k)+\sin{\theta}\cos{\theta}\frac{k^{2}}{4}+...\,,$ (A.9) $E\left(k,\,\frac{\pi}{2}\right)=E(k)=\frac{\pi}{2}-\frac{\pi}{8}k^{2}+...\,,$ (A.10) using which we obtain $\tau\approx\frac{2r_{min}}{{\mbox{$v_{F}$}}(3\pi\xi_{0}\rho_{0})^{1/2}}\left[\frac{\sqrt{R^{2}-r_{min}^{2}}\sqrt{R^{2}-\rho_{0}^{2}}}{R}+\frac{\pi}{4}\left(\frac{\rho_{0}}{r_{min}}\right)^{2}r_{min}\right]\,.$ (A.11) Assuming $R\gg r_{min}$, $R\gg\rho_{0}$ and $\rho_{0}<r_{min}$ we have $\tau\approx\frac{2R}{{\mbox{$v_{F}$}}}\frac{r_{min}}{(3\pi\xi_{0}\rho_{0})^{1/2}}\,.$ (A.12) Similarly, $\Delta\phi\approx\pi\frac{b}{(3\pi\xi_{0}\rho_{0})^{1/2}}\ll 1.$ (A.13) ## References * (1) R. J. Donnelly, Quantised Vortices In Helium II (Cambridge University Press, Cambridge, England, 1991). * (2) C. F. Barenghi, R. J. Donnelly, and W. F. Vinen, Quantized Vortex Dynamics And Superfluid Turbulence (Springer, Berlin. 2001). * (3) C. F. Barenghi. R. J. Donnelly, and W. F. Vinen J. Low Temp. Phys. 52, 189 (1982). * (4) W. F. Vinen and J. J. Niemela, J. Low Temp. Phys. 128, 167 (2002), and Erratum, 129, 213 (2002). * (5) S. Hulton, C. F. Barenghi, and D. C. Samuels, Phys. Rev. Lett. 89, 275301 (2002). * (6) D. Kivotides, J. C. Vassilicos, D. C. Samuels, and C. F. Barenghi, Phys. Rev. Lett. 86, 3080 (2001). * (7) W. F. Vinen, M. Tsubota, and A. Mitani, Phys. Rev. Lett. 91, 135301 (2003). * (8) E. Kozik and B. Svistunov, Phys. Rev. Lett. 92, 035301 (2004). * (9) V. S. L’vov, S. V. Nazarenko, and G. E. Volovik, JETP Lett. 80, 479 (2004). * (10) W. F. Vinen, Phys. Rev. B 64, 134520 (2001). * (11) C. F. Barenghi, N. G. Parker, N. P. Proukakis, and C. S. Adams, J. Low Temp. Phys. 138, 629 (2005). * (12) V. S. L’vov, S. V. Nazarenko, and O. Rudenko, Phys. Rev. B 76, 024520 (2007). * (13) E. Kozik and B. Svistunov, arXiv:0710.4572 (2007). * (14) P. E. Roche, P. Diribarne, T. Didelot, O. Francais, L. Rousseau, and W. H. Willaime, Europhys. Lett. 77, 66002 (2007). * (15) P. E. Roche and C. F. Barenghi, to appear in Europhys. Lett. * (16) J. Maurer and P. Tabeling, Europhys. Lett. 43, 29 (1998). * (17) D. I. Bradley, D. O. Clubb, S. N. Fisher, A. M. Guénault, R. P. Haley, C. J. Matthews, G. R. Pickett, V. Tsepelin, and K. Zaki, Phys. Rev. Lett. 96, 035301 (2006). * (18) S. R. Stalp, L. Skrbek, and R. J. Donnelly, Phys. Rev. Lett. 82, 4831 (1999). * (19) C. F. Barenghi, A. V. Gordeev, and L. Skrbek, Phys. Rev. E 74, 026309 (2006). * (20) G. E. Volovik, J. Low Temp. Phys. 136, 309 (2004). * (21) A. P. Finne, T. Araki, R. Blaauwgeers, V. B. Eltsov, N. B. Kopnin, M. Krusius, L. Skrbek, M. Tsubota, and G. E. Volovik, Nature 424, 1022 (2003). * (22) D. Jou and M. S. Mongiovi, Phys. Rev. B 69, 094513 (2004). * (23) M. Tsubota, T. Araki, and C. F. Barenghi, Phys. Rev. Lett. 90, 205301 (2003). * (24) V. B. Eltsov, A. P. Finne, R. Hanninen, J. Kopu, M. Krusius, M. Tsubota, and E. V. Thuneberg, Phys. Rev. Lett. 96, 215302 (2001). * (25) C. F. Barenghi, to be published in Physica D (2008). * (26) P. M. Walmsley, A. I. Golov, A. A. Levchenko, and B. White, J. Low Temp. Phys. 148, 317 (2007). * (27) D. Charalambous, L. Skrbek, P. C. Hendry, P. V. E. McClintock, and W. F. Vinen, Phys. Rev. E 74, 036307 (2006). * (28) T. V. Chagovets, A. V. Gordeev, and L. Skrbek, Phys. Rev. E 76, 027301 (2007). * (29) S. N. Fisher, A. J. Hall, A. M. Guénault and G. R. Pickett, Phys. Rev. Lett. 86, 244 (2001). * (30) H. Yano, A. Handa, M. Nakagawa, K. Obara, O. Ishikawa, and T. Hata, J. Low Temp. Phys. 138, 561 (2005). * (31) T. Zhang and S. W. Van Sciver, Nature Phys. 1, 36 (2005). * (32) G. P. Bewley, D. P. Lathrop, and K. R. Sreenivasan, Nature 441, 588 (2006). * (33) Y. A. Sergeev, C. F. Barenghi, and D. Kivotides,, Phys. Rev. B 74, 184506 (2006), and Erratum, Phys. Rev. B 75, 019904 (2006). * (34) A. F. Andreev, Zh. Eksp. Teor. Fiz. 46, 1823 (1964). * (35) G. Kieselmann and D. Rainer, J. Phys. B (Condensed Matter) 52, 267 (1983). * (36) S. Yip, Phys. Rev. B 32, 2915 (1985). * (37) N. A. Greaves and A. J. Leggett, J. Phys. C (Solid State Physics) 16, 4383 (1983). * (38) A. M. Guénault and G. R. Pickett, Dynamic and Thermal Behaviour of Quasi-particles in Superfluid 3He-B, in: Helium Three, 659, edited by W. P. Halperin and L. P. Pitaevskii (Elsevier Science Publishers B.V., 1990). * (39) M. P. Enrico, S. N. Fisher and R. J. Watts-Tobin, J. Low Temp. Phys. 98, 81, (1995). * (40) N. B. Kopnin and V. E. Kravtsov, Soviet Phys. JETP 44, 861 (1976). * (41) M. Stone, Phys. Rev. B 54, 13222 (1995). * (42) G. E. Volovik, The universe in a helium droplet (Clarendon Press, Oxford, 2003). * (43) D. I. Bradley, S. N. Fisher, A. M. Guénault, M. R. Lowe, G. R. Pickett, A. Rahm, and R. C. V. Whitehead, Phys. Rev. Lett. 93, 235302 (2004). * (44) D. I. Bradley, D. O. Clubb, S. N. Fisher, A. M. Guénault, R. P. Haley, C. J. Matthews, G. R. Pickett, V. Tsepelin, and K. Zaki, Phys. Rev. Lett. 95, 035302 (2005). * (45) D. I. Bradley, S. N. Fisher, A. M. Guénault, R. P. Haley, C. J. Matthews, G. R. Pickett, J. Roberts, S. O. Sullivan and V. Tsepelin, J. Low Temp. Phys. 148, 235 (2007). * (46) D. S. Greywall, Phys. Rev. 33, 7520 (1986). * (47) J. Bardeen, R. Kummel, A. E. Jacobs, and L. Tewordt, Phys. Rev. 187, 556 (1969). * (48) T. Tsuneto, Superconductivity and Superfluidity (Cambridge University Press, Cambridge, England, 1998). * (49) D. C. Samuels and R. J. Donnelly, Phys. Rev. Lett. 65, 187 (1990). Figure 1: (Color online) The dimensionless effective potential $D=\Delta_{eff}(r/\xi_{0})/\Delta_{0}$ seen by quasiparticles with momentum parallel to the $x$-axis and moving from $x=-\infty$. The dimensionless coordinates $x$ and $y$ are in units of $\xi_{0}$. Figure 2: (Color online) Schematic trajectory of the quasiparticle which starts at position A, is Andreev–reflected by the vortex (at the origin) at position B (where it becomes a quasihole), then traces its way back with a small Andreev angle $\Delta\phi$ (not to scale).	arxiv-papers	2008-08-13T10:51:28	2024-09-04T02:48:57.267949	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "C.F. Barenghi, Y.A. Sergeev and N. Suramlishvili", "submitter": "Carlo F. Barenghi", "url": "https://arxiv.org/abs/0808.1806" }
0808.1929	[noPostScript] # Existence of minimal models for varieties of log general type II Christopher D. Hacon Department of Mathematics University of Utah 155 South 1400 East JWB 233 Salt Lake City, UT 84112, USA hacon@math.utah.edu and James McKernan Department of Mathematics University of California at Santa Barbara Santa Barbara, CA 93106, USA mckernan@math.ucsb.edu Department of Mathematics MIT 77 Massachusetts Avenue Cambridge, MA 02139, USA mckernan@math.mit.edu ###### Abstract. Assuming finite generation in dimension $n-1$, we prove that pl-flips exist in dimension $n$. The first author was partially supported by NSF research grant no: 0456363 and an AMS Centennial fellowship and the second author was partially supported by NSA grant no: H98230-06-1-0059 and NSF grant no: 0701101. We would like to thank F. Ambro, C. Birkar, P. Cascini, J. A. Chen, A. Corti, O. Fujino, S. Keel, and J. Kollár for valuable suggestions. ###### Contents 1. 1 Introduction 2. 2 Notation and conventions 3. 3 Preliminary results 4. 4 Multiplier ideal sheaves 5. 5 Asymptotic multiplier ideal sheaves 6. 6 Lifting sections 7. 7 Rationality of the restricted algebra 8. 8 Proof of (1.2) ## 1\. Introduction This is the second of two papers whose purpose is to establish: ###### Theorem 1.1. The canonical ring $R(X,K_{X})=\bigoplus_{m\in\mathbb{N}}H^{0}(X,\mathcal{O}_{X}(mK_{X})),$ is finitely generated for every smooth projective variety $X$. Note that Siu has announced a proof of finite generation for varieties of general type, using analytic methods, see [11]. Our proof relies on the ideas and techniques of the minimal model program and roughly speaking in this paper we will show that finite generation in dimension $n-1$ implies the existence of flips in dimension $n$. More precisely, assuming the following: ###### Theorem F. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety. Let $(X,\Delta=A+B)$ be a $\mathbb{Q}$-factorial kawamata log terminal pair of dimension $n$, where $A\geq 0$ is an ample $\mathbb{Q}$-divisor and $B\geq 0$. If $K_{X}+\Delta$ is pseudo-effective, then 1. (1) The pair $(X,\Delta)$ has a log terminal model $\mu\colon X\dasharrow Y$. In particular if $K_{X}+\Delta$ is $\mathbb{Q}$-Cartier then the log canonical ring $R(X,K_{X}+\Delta)=\bigoplus_{m\in\mathbb{N}}H^{0}(X,\mathcal{O}_{X}(\llcorner m(K_{X}+\Delta)\lrcorner)),$ is finitely generated. 2. (2) Let $V\subset\operatorname{WDiv}_{\mathbb{R}}(X)$ be the vector space spanned by the components of $\Delta$. Then there is a constant $\delta>0$ such that if $G$ is a prime divisor contained in the stable base locus of $K_{X}+\Delta$ and $\Xi\in\mathcal{L}_{A}(V)$ such that $\\|\Xi-\Delta\\|<\delta$, then $G$ is contained in the stable base locus of $K_{X}+\Xi$. 3. (3) Let $W\subset V$ be the smallest affine subspace of $\operatorname{WDiv}_{\mathbb{R}}(X)$ containing $\Delta$, which is defined over the rationals. Then there is a constant $\eta>0$ and a positive integer $r>0$ such that if $\Xi\in W$ is any divisor and $k$ is any positive integer such that $\\|\Xi-\Delta\\|<\eta$ and $k(K_{X}+\Xi)/r$ is Cartier, then every component of $\operatorname{Fix}(k(K_{X}+\Xi))$ is a component of the stable base locus of $K_{X}+\Delta$. we prove the existence of pl-flips: ###### Theorem A. Pl-flips exist in dimension $n$. that is, we prove: ###### Theorem 1.2. Theorem Fn-1 implies Theorem An. With the results of [2], (1.2) completes the proof of (1.1). The main ideas used in this paper have their origins in the work of Shokurov on the existence of flips [10] together with the use of the extension theorem of [4] which in turn was inspired by the work of Kawamata, Siu and Tsuji (cf. [6], [12] and [15]). For further history about the details of this problem see [3, §2.1]. In this paper, however we do not make use of the concept of “asymptotic saturation” introduced by Shokurov, and in fact we prove a more general result which does not require the relative weak log Fano condition (see also [1]). Further treatments of the results of this paper may be found in [1] and [5] (which follows Shokurov’s approach more explicitly). We now turn to a more detailed description of the results and techniques used in this paper. Recall the following: ###### Definition 1.3. Let $(X,\Delta)$ be a purely log terminal pair and $f\colon X\longrightarrow Z$ be a projective morphism of normal varieties. Then $f$ is a pl-flipping contraction if $\Delta$ is a $\mathbb{Q}$-divisor and 1. (1) $f$ is small, of relative Picard number one, 2. (2) $-(K_{X}+\Delta)$ is $f$-ample, 3. (3) $X$ is $\mathbb{Q}$-factorial, 4. (4) $S=\llcorner\Delta\lrcorner$ is irreducible and $-S$ is $f$-ample. The flip of a pl-flipping contraction $f\colon X\longrightarrow Z$ is a small projective morphism $g\colon Y\longrightarrow Z$ of relative Picard number one, such that $K_{Y}+\Gamma$ is $g$-ample, where $\Gamma$ is the strict transform of $\Delta$. The flip $g$ is unique, if it exists at all, and it is given by $Y=\operatorname{Proj}_{Z}\mathfrak{R}\qquad\text{where}\qquad\mathfrak{R}=\bigoplus_{m\in\mathbb{N}\,:\,k\|m}f_{}\mathcal{O}_{X}(m(K_{X}+\Delta)),$ and $k$ is any positive integer such that $k(K_{X}+\Delta)$ is integral. Therefore, in order to prove the existence of pl-flips, it suffices to show that $\mathfrak{R}$ is a finitely generated $\mathcal{O}_{Z}$-algebra. Since this problem is local over $Z$, we may assume that $Z=\operatorname{Spec}A$ is affine and it suffices to prove that $R(X,k(K_{X}+\Delta))=\bigoplus_{m\in\mathbb{N}\,:\,k\|m}H^{0}(X,\mathcal{O}_{X}(m(K_{X}+\Delta))),$ is a finitely generated $A$-algebra. It is then natural to consider the restricted algebra $R_{S}(X,k(K_{X}+\Delta))=\operatorname{Im}\left(R(X,k(K_{X}+\Delta))\longrightarrow R(S,k(K_{S}+\Omega))\right),$ whose graded pieces correspond to the images of the restriction homomorphisms $H^{0}(X,\mathcal{O}_{X}(m(K_{X}+\Delta)))\longrightarrow H^{0}(S,\mathcal{O}_{S}(m(K_{S}+\Omega))),$ where $m=kl$ is divisible by $k$ and $\Omega$ is defined by the adjunction formula $(K_{X}+\Delta)\|_{S}=K_{S}+\Omega,$ and $k(K_{X}+\Delta)$ is Cartier. Shokurov has shown, cf. (3.2), that the algebra $R(X,k(K_{X}+\Delta))$ is finitely generated if and only if the restricted algebra is finitely generated. Now, if the natural inclusion $R_{S}(X,k(K_{X}+\Delta))\subset R(S,k(K_{S}+\Omega)),$ were an isomorphism, then (1.2) would follow from (1) of Theorem Fn-1. In fact the pair $(S,\Omega)$ is kawamata log terminal, $\operatorname{dim}S=\operatorname{dim}X-1=n-1$ and since $f\|_{S}$ is birational, $\Omega$ is automatically big so that, by a standard argument, (1) of Theorem Fn-1 applies and $R(S,k(K_{S}+\Omega))$ is finitely generated. (3.2) also implies $\mathfrak{R}$ is finitely generated. Unluckily this is too much to hope for. However it does suggest that one should concentrate on the problem of lifting sections and the main focus of this paper is to prove the extension result (6.3). In fact (1.2) is a straightforward consequence of (6.3). To fix ideas, let us start with an example where we cannot lift sections. Let $X$ be the blow up of $\mathbb{P}^{2}$ at a point $o$, with exceptional divisor $E$. Let $S$ be the strict transform of a line through $o$, let $L_{1}$, $L_{2}$ and $L_{3}$ be the strict transforms of general lines in $\mathbb{P}^{2}$, let $p=E\cap S$ and let $p_{i}=L_{i}\cap S$. Then the pair $(X,\Delta=S+(2/3)(E+L_{1}+L_{2}+L_{3})),$ is purely log terminal but the homomorphism $H^{0}(\mathcal{O}_{X}(3l(K_{X}+\Delta)))\longrightarrow H^{0}(\mathcal{O}_{S}(3l(K_{S}+\Omega)))\simeq H^{0}(\mathcal{O}_{\mathbb{P}^{1}}(2l)),$ is never surjective, where $\Omega=(\Delta-S)\|_{S}=2/3(p+p_{1}+p_{2}+p_{3})$ and $l$ is a positive integer. The problem is that the stable base locus of $K_{X}+\Delta$ contains $E$ and yet $\|3(K_{S}+\Omega)\|$ is base point free. Notice, however, that $\|3l(K_{X}+\Delta)\|_{S}=\|3l(K_{S}+\Theta)\|+3l(\Omega-\Theta),$ where $\Theta=(2/3)(p_{1}+p_{2}+p_{3})$ is obtained from $\Omega$ by throwing away $p$. In other words, $\Theta$ is obtained from $\Omega$ by removing some part of each components contained in the stable base locus of $K_{X}+\Delta$. Returning to the general setting, one may then hope that the restricted algebra $R_{S}(S,l(K_{X}+\Delta))$ is given by an algebra of the form $R(S,l(K_{S}+\Theta))$ for some kawamata log terminal pair $(S,\Theta)$ where $0\leq\Theta\leq\Omega$ is a $\mathbb{Q}$-divisor obtained from $\Omega$ by subtracting components of $\Omega$ contained in the stable base locus of $K_{X}+\Delta$. We will now explain how this may be achieved. The tricky thing is to determine exactly how much of the stable base locus to throw away. It is not hard to reduce to the following situation: $\pi\colon X\longrightarrow Z$ is a projective morphism to a normal affine variety $Z$, where $(X,\Delta=S+A+B)$ is a purely log terminal pair of dimension $n$, $S=\llcorner\Delta\lrcorner$ is irreducible, $X$ and $S$ are smooth, $A\geq 0$ is an ample $\mathbb{Q}$-divisor, $B\geq 0$, $(S,\Omega=(\Delta-S)\|_{S})$ is canonical and the stable base locus of $K_{X}+\Delta$ does not contain $S$. Let $\Theta_{m}=\Omega-\Omega\wedge F_{m}\qquad\text{where}\qquad F_{m}=\operatorname{Fix}(\|m(K_{X}+\Delta)\|_{S})/m,$ and $m(K_{X}+\Delta)$ is Cartier. Then $m(\Omega-\Theta_{m})$ is the biggest divisor contained in $\operatorname{Fix}(\|m(K_{X}+\Delta)\|_{S})$ such that $0\leq\Theta_{m}\leq\Omega$. It follows that $\|m(K_{S}+\Theta_{m})\|+m(\Omega-\Theta_{m})\supset\|m(K_{X}+\Delta)\|_{S}.$ A simple consequence of the main lifting result (6.3) of this paper implies that this tautological inclusion (1) is actually an equality, $\|m(K_{S}+\Theta_{m})\|+m(\Omega-\Theta_{m})=\|m(K_{X}+\Delta)\|_{S}.$ A technical, but significant, improvement on the proof of the existence of flips which appears in [5] is that the statement of (1) and of (6.3) involves only linear systems and divisors on $X$, even though the proof of (6.3) involves passing to a higher model. The key point is that since $(S,\Omega)$ is canonical, it suffices to keep track only of the fixed divisor on $S$ and not of the whole base locus. To prove (1) we use the method of multiplier ideal sheaves. In fact the main point is to establish an inclusion of multiplier ideal sheaves, (5.3). A proof of (5.3) appeared originally in [4]. We chose to include a proof of this result for the convenience of the reader and we decided to use notation closer to the well established notation used in [9]. Note however that the multiplier ideal sheaves we use, see (4.2), must take into account the divisor $\Delta$ (for example consider the case worked out above) and the fact that $(S,\Omega)$ is canonical. In fact (1) follows from the MMP. Indeed, if one runs $f\colon X\dasharrow Y$ the $(K_{X}+\Delta)$-MMP, almost by definition this will not change the linear systems $\|m(K_{X}+\Delta)\|$. Since $K_{Y}+\Gamma=K_{X}+f_{}\Delta$ is nef, one can lift sections on $Y$ from the strict transform $T$ of $S$, by an easy application of Kawamata-Viehweg vanishing. In general, however, the linear systems $\|m(K_{T}+g_{}\Theta)\|$ are bigger than the linear systems $\|m(K_{S}+\Theta)\|$, since the induced birational map $g\colon S\dasharrow T$ might extract some divisors. However any such divisor must have log discrepancy at most one, so this cannot happen, almost by definition, if $K_{S}+\Theta$ is canonical. In order to establish that $R_{S}(X,k(K_{X}+\Delta))$ is finitely generated, cf. (7.1), and thereby to finish the proof of (1.2), it is necessary and sufficient to show that $\Theta=\lim(\Theta_{m!}/m!)$ is rational (the seemingly strange use of factorials is so that we can use limits rather than limsups). At this point we play off two facts. The first is that since we are assuming that Theorem F holds on $S$, if $m>0$ is sufficiently divisible and $\Phi$ is an appropriately chosen $\mathbb{Q}$-divisor sufficiently close to $\Theta$, then the base locus of $\|m(K_{S}+\Phi)\|$ and the stable base locus of $K_{S}+\Theta$ are essentially the same (basically because $K_{S}+\Theta$ and $K_{S}+\Phi$ share a log terminal model $\mu\colon S\dasharrow S^{\prime}$ and these two sets of divisors are precisely the divisors contracted by $\mu$). The second is that using (4.1), (6.3) is slightly stronger than (1); one is allowed to overshoot $\Theta_{m}$ by an amount $\epsilon/m$, where $\epsilon>0$ is fixed. (It seems worth pointing out that (4.1) seems to us a little mysterious. In particular, unlike (6.3), we were unable to show that this result follows from the MMP.) More precisely, since the base locus of $\|m(K_{S}+\Theta_{m})\|$ contains no components of $\Theta_{m}$, by (2) of Theorem F it follows that the stable base locus of $K_{S}+\Theta$ contains no components of $\Theta$. If $\Theta$ is not rational, then by Diophantine approximation there is a $\mathbb{Q}$-divisor $0\leq\Phi\leq\Omega$ very close to $\Theta$ and an integer $k>0$ such that $k\Phi$ is integral and $\operatorname{mult}_{G}\Phi>\operatorname{mult}_{G}\Theta$, for some prime divisor $G$. By (6.3), it actually follows that $\|k(K_{S}+\Phi)\|+k(\Omega-\Phi)=\|k(K_{X}+\Delta)\|_{S}.$ The condition $\operatorname{mult}_{G}\Phi>\operatorname{mult}_{G}\Theta$ ensures that $G$ is a component of $\operatorname{Fix}(k(K_{S}+\Phi))$, and hence of the stable base locus of $K_{S}+\Phi$. But then $G$ is a component of $\Theta$ and of the stable base locus of $\Theta$. This is the required contradiction. ## 2\. Notation and conventions We work over the field of complex numbers $\mathbb{C}$. Let $X$ be a normal variety. A $\left.\begin{array}[]{r}\text{{(integral) divisor}}\\\ \text{{$\mathbb{Q}$-divisor}}\\\ \text{{$\mathbb{R}$-divisor}}\\\ \end{array}\right\\}\quad\text{is a}\quad\left\\{\begin{array}[]{l}\text{$\mathbb{Z}$-linear}\\\ \text{$\mathbb{Q}$-linear}\\\ \text{$\mathbb{R}$-linear,}\\\ \end{array}\right.$ combination of prime divisors. Given an integral Weil divisor $D$, we let $R(X,D)=\bigoplus_{m\in\mathbb{N}}H^{0}(X,\mathcal{O}_{X}(mD)).$ Set $\displaystyle\operatorname{WDiv}_{\mathbb{Q}}(X)$ $\displaystyle=\operatorname{WDiv}(X)\underset{\mathbb{Z}}{\otimes}\mathbb{Q}$ $\displaystyle\operatorname{WDiv}_{\mathbb{R}}(X)$ $\displaystyle=\operatorname{WDiv}(X)\underset{\mathbb{Z}}{\otimes}\mathbb{R},$ where $\operatorname{WDiv}(X)$ is the group of Weil divisors on $X$. The definitions below for $\mathbb{R}$-divisors reduce to the usual definitions for $\mathbb{Q}$-divisors and integral divisors, see [2]. Note that the group of $\mathbb{R}$-divisors forms a vector space, with a canonical basis given by the prime divisors. If $C=\sum c_{i}B_{i}$ and $D=\sum d_{i}B_{i}$, where $B_{i}$ are distinct prime divisors, then we write $D\geq 0$ if $d_{i}\geq 0$ and we will denote by $\displaystyle\\|C\\|$ $\displaystyle=\max_{i}c_{i}$ $\displaystyle C\wedge D$ $\displaystyle=\sum_{i}\min\\{c_{i},d_{i}\\}B_{i}$ $\displaystyle\llcorner C\lrcorner$ $\displaystyle=\sum_{i}\llcorner c_{i}\lrcorner B_{i}$ $\displaystyle\\{C\\}$ $\displaystyle=C-\llcorner C\lrcorner.$ Two $\mathbb{R}$-divisors $C$ and $D$ are $\left.\begin{array}[]{rl}\text{linearly equivalent,}&C\sim D\\\ \text{$\mathbb{Q}$-linearly equivalent,}&C\sim_{\mathbb{Q}}D\\\ \text{$\mathbb{R}$-linearly equivalent,}&C\sim_{\mathbb{R}}D\\\ \end{array}\right\\}\quad\text{if $C-D$ is a}\quad\left\\{\begin{array}[]{l}\text{$\mathbb{Z}$-linear}\\\ \text{$\mathbb{Q}$-linear}\\\ \text{$\mathbb{R}$-linear,}\\\ \end{array}\right.$ combination of principal divisors. Note that if $C\sim_{\mathbb{Q}}D$ then $mC\sim mD$ for some positive integer $m$, but this fails in general for $\mathbb{R}$-linear equivalence. Note also that if two $\mathbb{Q}$-divisors are $\mathbb{R}$-linearly equivalent then they are in fact $\mathbb{Q}$-linearly equivalent, but that two integral divisors might be $\mathbb{Q}$-linearly equivalent without being linearly equivalent. Let $\displaystyle\|D\|$ $\displaystyle=\\{\,C\in\operatorname{WDiv}(X)\,\|\,C\geq 0\,,\,C\sim D\,\\}$ $\displaystyle\|D\|_{\mathbb{Q}}$ $\displaystyle=\\{\,C\in\operatorname{WDiv}_{\mathbb{Q}}(X)\,\|\,C\geq 0\,,\,C\sim_{\mathbb{Q}}D\,\\}$ $\displaystyle\|D\|_{\mathbb{R}}$ $\displaystyle=\\{\,C\in\operatorname{WDiv}_{\mathbb{R}}(X)\,\|\,C\geq 0\,,\,C\sim_{\mathbb{R}}D\,\\}.$ If $T$ is a subvariety of $X$, not contained in the base locus of $\|D\|$, then $\|D\|_{T}$ denotes the image of the linear system $\|D\|$ under restriction to $T$. If $D$ is an integral divisor, $\operatorname{Fix}(D)$ denotes the fixed divisor of $D$ so that $\|D\|=\|D-\operatorname{Fix}(D)\|+\operatorname{Fix}(D)$ where the base locus of $\|D-\operatorname{Fix}(D)\|$ contains no divisors. More generally $\operatorname{Fix}(V)$ denotes the fixed divisor of the linear system $V$. The stable base locus of $D$, denoted by $\mathbf{B}(D)$, is the intersection of the support of the elements of $\|D\|_{\mathbb{R}}$ (if $\|D\|_{\mathbb{R}}$ is empty then by convention the stable base locus is the whole of $X$). The stable fixed divisor is the divisorial support of the stable base locus. The augmented stable base locus of $D$, denoted by $\mathbf{B}_{+}(D)$, is given by the stable base locus of $D-\epsilon A$ for some ample divisor $A$ and any rational number $0<\epsilon\ll 1$. The diminished stable base locus is defined by $\mathbf{B}_{-}(D)=\bigcup_{\epsilon>0}\mathbf{B}(D+\epsilon A).$ In particular we have $\mathbf{B}_{-}(D)\subset\mathbf{B}(D)\subset\mathbf{B}_{+}(D).$ An $\mathbb{R}$-Cartier divisor $D$ is an $\mathbb{R}$-linear combination of Cartier divisors. An $\mathbb{R}$-Cartier divisor $D$ is nef if $D\cdot\Sigma\geq 0$ for any curve $\Sigma\subset X$. An $\mathbb{R}$-Cartier divisor $D$ is ample if it is $\mathbb{R}$-linearly equivalent to a positive linear combination of ample divisors (in the usual sense). An $\mathbb{R}$-Cartier divisor $D$ is big if $D\sim_{\mathbb{R}}A+B$, where $A$ is ample and $B\geq 0$. A $\mathbb{Q}$-Cartier divisor $D$ is a general ample $\mathbb{Q}$-divisor if there is an integer $m>0$ such that $mD$ is very ample and $mD\in\|mD\|$ is very general. A log pair $(X,\Delta)$ is a normal variety $X$ and $\mathbb{R}$-Weil divisor $\Delta\geq 0$ such that $K_{X}+\Delta$ is $\mathbb{R}$-Cartier. We say that a log pair $(X,\Delta)$ is log smooth, if $X$ is smooth and the support of $\Delta$ is a divisor with global normal crossings. A projective birational morphism $g\colon Y\longrightarrow X$ is a log resolution of the pair $(X,\Delta)$ if $X$ is smooth and the inverse image of $\Delta$ union the exceptional locus is a divisor with global normal crossings. Note that in the definition of log resolution we place no requirement that the indeterminacy locus of $g$ is contained in the locus where the pair $(X,\Delta)$ is not log smooth. If $V$ is a linear system on $X$, a log resolution of $V$ and $(X,\Delta)$ is a log resolution of the pair $(X,\Delta)$ such that if $\|M\|+F$ is the decomposition of $g^{}V$ into its mobile and fixed parts, then $\|M\|$ is base point free and $F$ union the exceptional locus union the strict transform of $\Delta$ is a divisor with simple normal crossings support. If $g$ is a log resolution, then we may write $K_{Y}+\Gamma=g^{}(K_{X}+\Delta)+E,$ where $\Gamma\geq 0$ and $E\geq 0$ have no common components, $g_{}\Gamma=\Delta$ and $E$ is $g$-exceptional. Note that this decomposition is unique. The log discrepancy of a divisor $F$ over $X$ $a(X,\Delta,F)=1+\operatorname{mult}_{F}(E-\Gamma).$ Note that with this definition, a component $F$ of $\Delta$ with coefficient $b$ has log discrepancy $1-b$. The log discrepancy does not depend on the choice of model $Y$, so that the log discrepancy is also a function defined on valuations. A log canonical place is any valuation of log discrepancy at most zero and the centre of a log canonical place is called a log canonical centre. Note that every divisor on $X$ is by definition a canonical centre, so the only interesting canonical centres are of codimension at least two. The pair $(X,\Delta)$ is kawamata log terminal if there are no log canonical centres. We say that the pair $(X,\Delta)$ is purely log terminal (respectively canonical or terminal) if the log discrepancy of any exceptional divisor is greater than zero (respectively at least one or greater than one). We say that the pair is divisorially log terminal if there is a log resolution $g\colon Y\longrightarrow X$ such that all exceptional divisors $E\subset Y$ have log discrepancy greater than zero. ## 3\. Preliminary results In this section we recall several results about finitely generated algebras and in particular we will give a proof of Shokurov’s result that the pl-flip exists if and only if the restricted algebra is finitely generated. ###### Definition 3.1. Let $X$ be a normal variety, $S$ be a prime divisor and $B$ an integral Weil divisor which is $\mathbb{Q}$-Cartier and whose support does not contain $S$. The restricted algebra $R_{S}(X,B)$ is the image of the homomorphism $R(X,B)\longrightarrow R(S,B\|_{S})$. We remark that as $B$ is $\mathbb{Q}$-Cartier then $B\|_{S}$ is a well defined $\mathbb{Q}$-Cartier divisor on $S$. ###### Theorem 3.2. Let $f\colon X\longrightarrow Z$ be a pl-flipping contraction with respect to $(X,\Delta)$. Pick an integer $k$ such that $k(K_{X}+\Delta)$ is Cartier. Then 1. (1) The flip of $f$ exists if and only if the flip of $f$ exists locally over $Z$. 2. (2) If $Z=\operatorname{Spec}A$ is affine then the flip $f^{+}\colon X^{+}\longrightarrow Z$ exists if and only if the restricted algebra $R_{S}(X,k(K_{X}+\Delta))$ is a finitely generated $A$-algebra. We start with the following well known result: ###### Lemma 3.3. Let $R$ be a graded algebra which is an integral domain and let $d$ be a positive integer. Then $R$ is finitely generated if and only if $R_{(d)}$ is finitely generated. ###### Proof. Suppose that $R$ is finitely generated. It is easy to write down an action of the cyclic group $\mathbb{Z}_{d}$ on $R$ so that the invariant ring is $R_{(d)}$. Thus $R_{(d)}$ is finitely generated by the Theorem of E. Noether which states that the ring of invariants of a finitely generated ring under the action of a finite group is finitely generated. Suppose now that $R_{(d)}$ is finitely generated. Let $f\in R_{i}$. Then $f$ is a root of the monic polynomial $x^{d}-f^{d}\in R_{(d)}[x]$. It follows that $R$ is integral over $R_{(d)}$ and the result follows by another Theorem of E. Noether on finiteness of integral closures. ∎ ###### Lemma 3.4. Let $S$ be a normal prime divisor on $X$ and let $B$ an integral Weil divisor which is $\mathbb{Q}$-Cartier and whose support does not contain $S$. * • If $R(X,B)$ is finitely generated then $R_{S}(X,B)$ is finitely generated. * • If $S\sim B$ and $R_{S}(X,B)$ is finitely generated then $R(X,B)$ is finitely generated. ###### Proof. Since there is a surjective homomorphism $\phi\colon R(X,B)\longrightarrow R_{S}(X,B)$, it is clear that if $R(X,B)$ is finitely generated then $R_{S}(X,B)$ is finitely generated. Suppose now that $R_{S}(X,B)$ is finitely generated and $S\sim B$. Then there is a rational function $g_{1}$ such that $(g_{1})=S-B$. If we consider the elements of $R(X,B)_{m}$ as rational functions, then a rational function $g$ belongs to $R(X,B)_{m}$ if and only if $(g)+mB\geq 0$. But if $g$ is in the kernel of $\phi$, then there is a divisor $S^{\prime}\geq 0$ such that $(g)+mB=S+S^{\prime}$. It follows that $(g/g_{1})+(m-1)B=S^{\prime}$ so that $g/g_{1}=h\in R(X,B)_{m-1}$. But then the kernel of $\phi$ is the principal ideal generated by $g_{1}$. ∎ ###### Proof of (3.2). It is well known that the flip $f^{+}\colon X^{+}\longrightarrow Z$ exists if and only if the sheaf of graded $\mathcal{O}_{Z}$-algebras $\bigoplus_{m\in\mathbb{N}\,:\,k\|m}f_{}\mathcal{O}_{X}(m(K_{X}+\Delta)),$ is finitely generated, cf. [7, 6.4]. Since this can be checked locally, this gives (1). If $Z=\operatorname{Spec}A$ is affine it suffices to check that $R(X,k(K_{X}+\Delta))$ is a finitely generated $A$-algebra. Since the relative Picard number is one, there are real numbers $a$ and $b$ such that $a(K_{X}+\Delta)$ and $bS$ are numerically equivalent over $Z$. As both $-(K_{X}+\Delta)$ and $-S$ are ample $\mathbb{Q}$-divisors we may assume that $a$ and $b$ are both positive integers. Moreover, as $a(K_{X}+\Delta)-bS$ is numerically trivial over $Z$, it is semiample over $Z$ by the base point free theorem. In particular, we may replace numerical equivalence by linear equivalence, $a(K_{X}+\Delta)\sim_{Z}bS.$ But then there is a rational function $g$ and a divisor $D$ on $Z$ such that $a(K_{X}+\Delta)=bS+f^{}D+(g).$ As any line bundle on a quasi-projective variety is locally trivial, possibly passing to an open subset of $Z$, and using (1), we may assume that $D\sim 0$, so that $a(K_{X}+\Delta)\sim bS.$ By (3.3) it follows that $R(X,k(K_{X}+\Delta))$ is finitely generated if and only if $R(X,S)$ is finitely generated. Since $Z$ is affine and $f$ is small, $S$ is mobile so that $S\sim S^{\prime}$ where $S^{\prime}\geq 0$ is a divisor whose support does not contain $S$. By (3.4), $R(X,S)$ is finitely generated if and only if $R_{S}(X,S^{\prime})$ is finitely generated. Since $a(K_{X}+\Delta)\|_{S}\sim bS^{\prime}\|_{S}$ the result follows by (3.3). ∎ ## 4\. Multiplier ideal sheaves The main result of this section is: ###### Theorem 4.1. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$, where $(X,\Delta=S+A+B)$ is a log pair, $S=\llcorner\Delta\lrcorner$ is irreducible, $(X,S)$ is log smooth, and both $A\geq 0$ and $B\geq 0$ are $\mathbb{Q}$-divisors. Let $k$ be any positive integer and $0\leq\Phi\leq\Omega=(\Delta-S)\|_{S}$ be any divisor such that both $k(K_{S}+\Phi)$ and $k(K_{X}+\Delta)$ are Cartier. Let $C=A/k$. If there is an integer $l>1$ and an integral divisor $P\geq 0$ such that $lA$ is Cartier, $C-\frac{(k-1)}{m}P$ is ample, $(X,\Delta+\frac{k-1}{m}P)$ is purely log terminal and $l\|k(K_{S}+\Phi)\|+m(\Omega-\Phi)+(mC+P)\|_{S}\subset\|m(K_{X}+\Delta+C)+P\|_{S},$ where $m=kl$, then $\|k(K_{S}+\Phi)\|+k(\Omega-\Phi)\subset\|k(K_{X}+\Delta)\|_{S}.$ To prove (4.1), we need a variant of multiplier ideal sheaves: ###### Definition-Lemma 4.2. Let $(X,\Delta)$ be a log smooth pair where $\Delta$ is a reduced divisor and let $V$ be a linear system whose base locus contains no log canonical centres of $(X,\Delta)$. Let $\mu\colon Y\longrightarrow X$ be a log resolution of $V$ and $(X,\Delta)$ and let $F$ be the fixed divisor of the linear system $\mu^{}V$. Let $K_{Y}+\Gamma=\mu^{}(K_{X}+\Delta)+E$ where $\Gamma=\sum P_{i}$ is the sum of the divisors on $Y$ of log discrepancy zero. Then for any real number $c\geq 0$, define the multiplier ideal sheaf $\mathcal{J}_{\Delta,c\cdot V}:=\mu_{}\mathcal{O}_{Y}(E-\llcorner cF\lrcorner).$ If $\Delta=0$ we will write $\mathcal{J}_{c\cdot V}$ and if $D=cG$, where $G>0$ is a Cartier divisor, we define $\mathcal{J}_{\Delta,D}:=\mathcal{J}_{\Delta,c\cdot V},$ where $V=\\{G\\}$. ###### Proof. We have to show that the definition of the multiplier ideal sheaf is independent of the choice of log resolution. Let $\mu\colon Y\longrightarrow X$ and $\mu^{\prime}\colon Y^{\prime}\longrightarrow X$ be two log resolutions of $(X,\Delta)$ and $V$. We may assume that $\mu^{\prime}$ factors through $\mu$ via a morphism $\nu\colon Y^{\prime}\longrightarrow Y$. Then $F^{\prime}=\nu^{}F$ as $\mu^{}V-F$ is free, and $\displaystyle E^{\prime}-cF^{\prime}$ $\displaystyle=K_{Y^{\prime}}+\Gamma^{\prime}-\mu^{\prime}(K_{X}+\Delta)-cF^{\prime}$ $\displaystyle=K_{Y^{\prime}}+\Gamma^{\prime}-\nu^{}(K_{Y}+\Gamma-E+cF)$ $\displaystyle=\nu^{}(E-\llcorner cF\lrcorner)+K_{Y^{\prime}}+\Gamma^{\prime}-\nu^{}(K_{Y}+\Gamma+\\{cF\\})$ $\displaystyle=\nu^{}(E-\llcorner cF\lrcorner)+G.$ Since $(Y,\Gamma+E+F)$ is log smooth, it follows that $(Y,\Gamma+\\{cF\\})$ is log canonical and has the same log canonical places as $(Y,\Gamma)$ and hence as $(X,\Delta)$. Thus $\ulcorner{G}\urcorner\geq 0$ and since $\nu_{}(K_{Y^{\prime}}+\Gamma^{\prime})=K_{Y}+\Gamma$, $\ulcorner{G}\urcorner$ is $\nu$-exceptional. Then $\displaystyle\mu^{\prime}_{}\mathcal{O}_{Y^{\prime}}(E^{\prime}-\llcorner cF^{\prime}\lrcorner)$ $\displaystyle=\mu_{}(\nu_{}\mathcal{O}_{Y^{\prime}}(E^{\prime}-\llcorner cF^{\prime}\lrcorner))$ $\displaystyle=\mu_{}(\nu_{}\mathcal{O}_{Y^{\prime}}(\nu^{}(E-\llcorner cF\lrcorner)+\ulcorner{G}\urcorner))$ $\displaystyle=\mu_{}\mathcal{O}_{Y}(E-\llcorner cF\lrcorner).\qed$ We need to develop a little of the theory of multiplier ideal sheaves. ###### Lemma 4.3. Let $(X,\Delta)$ be a log smooth pair where $\Delta$ is reduced, let $V$ be a linear system whose base locus contains no log canonical centres of $(X,\Delta)$ and let $G\geq 0$ and $D\geq 0$ be $\mathbb{Q}$-Cartier divisors whose supports contain no log canonical centres of $(X,\Delta)$. Then 1. (1) $\mathcal{J}_{\Delta,D}=\mathcal{O}_{X}$ if and only if $(X,\Delta+D)$ is divisorially log terminal and $\llcorner D\lrcorner=0$. 2. (2) If $0\leq\Delta^{\prime}\leq\Delta$ then $\mathcal{J}_{\Delta,c\cdot V}\subset\mathcal{J}_{\Delta^{\prime},c\cdot V}$. In particular, $\mathcal{J}_{\Delta,c\cdot V}\subset\mathcal{J}_{c\cdot V}\subset\mathcal{O}_{X}$. 3. (3) If $\Sigma\geq 0$ is a Cartier divisor, $D-\Sigma\leq G$ and $\mathcal{J}_{\Delta,G}=\mathcal{O}_{X}$ then $\mathcal{I}_{\Sigma}\subset\mathcal{J}_{\Delta,D}$. ###### Proof. (1) follows easily from the definitions. (2) follows from the fact that $a(P,X,\Delta^{\prime})\geq a(P,X,\Delta)$ for all divisors $P$ on $Y$. To see (3), notice that as $\Sigma$ is Cartier and $\mathcal{J}_{\Delta,G}=\mathcal{O}_{X}$, we have $\mathcal{J}_{\Delta,G}(-\Sigma)=\mathcal{O}_{X}(-\Sigma)=\mathcal{I}_{\Sigma}.$ But since $D\leq G+\Sigma$, we also have $\mathcal{J}_{\Delta,G}(-\Sigma)=\mathcal{J}_{\Delta,G+\Sigma}\subset\mathcal{J}_{\Delta,D}.\qed$ We have the following extension of (9.5.1) of [8] or (2.4.2) of [14]: ###### Lemma 4.4. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$. Let $(X,\Delta)$ be a log smooth pair where $\Delta$ is reduced, let $S$ be a component of $\Delta$, let $D\geq 0$ be a $\mathbb{Q}$-Cartier divisor whose support does not contain any log canonical centres of $(X,\Delta)$ and let $\Theta=(\Delta-S)\|_{S}$. Let $N$ be a Cartier divisor. 1. (1) There is a short exact sequence $0\longrightarrow\mathcal{J}_{\Delta-S,D+S}\longrightarrow\mathcal{J}_{\Delta,D}\longrightarrow\mathcal{J}_{\Theta,D\|_{S}}\longrightarrow 0.$ 2. (2) (Nadel Vanishing) If $N-D$ is ample then $H^{i}(X,\mathcal{J}_{\Delta,D}(K_{X}+\Delta+N))=0,$ for $i>0$. 3. (3) If $N-D$ is ample then $H^{0}(X,\mathcal{J}_{\Delta,D}(K_{X}+\Delta+N))\longrightarrow H^{0}(S,\mathcal{J}_{\Theta,D\|_{S}}(K_{X}+\Delta+N)),$ is surjective. ###### Proof. By the resolution lemma of [13], we may find a log resolution $\mu\colon Y\longrightarrow X$ of $(X,\Delta+D)$ which is an isomorphism over the generic point of each log canonical centre of $(X,\Delta)$. If $T$ is the strict transform of $S$ then we have a short exact sequence $0\longrightarrow\mathcal{O}_{Y}(E-\llcorner\mu^{}D\lrcorner-T)\longrightarrow\mathcal{O}_{Y}(E-\llcorner\mu^{}D\lrcorner)\longrightarrow\mathcal{O}_{T}(E-\llcorner\mu^{}D\lrcorner)\longrightarrow 0.$ Now $\mu_{}\mathcal{O}_{Y}(E-\llcorner\mu^{}D\lrcorner)=\mathcal{J}_{\Delta,D}$. If $\Gamma$ is the sum of the divisors of log discrepancy zero then $E-\mu^{}D=(K_{Y}+\Gamma)-\mu^{}(K_{X}+\Delta+D).$ But then $E-\mu^{}D-T=(K_{Y}+\Gamma-T)-\mu^{}(K_{X}+\Delta-S+(D+S)),$ so that $\mu_{}\mathcal{O}_{Y}(E-\llcorner\mu^{}D\lrcorner-T)=\mathcal{J}_{\Delta-S,D+S},$ and $(E-\mu^{}D)\|_{T}=K_{T}+(\Gamma-T)\|_{T}-\mu^{}(K_{S}+\Theta+D\|_{S}),$ so that $\mu_{}\mathcal{O}_{T}(E-\llcorner\mu^{}D\lrcorner)=\mathcal{J}_{\Theta,D\|_{S}}.$ Since $(Y,\Gamma+\mu^{}D)$ is log smooth and $\Gamma$ and $\mu^{}D$ have no common components, $(Y,\Gamma+\\{\mu^{}D\\})$ is divisorially log terminal. Therefore we may pick an exceptional divisor $F\geq 0$ such that $K_{Y}+\Gamma+\\{\mu^{}D\\}+F$ is divisorially log terminal and $-F$ is $\mu$-ample. As $E-\llcorner\mu^{}D\lrcorner-T-(K_{Y}+\Gamma-T+\\{\mu^{}D\\}+F)=-\mu^{}(K_{X}+\Delta+D)-F,$ is $\mu$-ample, Kawamata-Viehweg vanishing implies that $R^{1}\mu_{}\mathcal{O}_{Y}(E-\llcorner\mu^{}D\lrcorner-T)=0,$ and this gives (1). Similarly, Kawamata-Viehweg vanishing implies that $R^{i}\mu_{}\mathcal{O}_{Y}(\mu^{}(K_{X}+\Delta+N)+E-\llcorner\mu^{}D\lrcorner)=0,$ for $i>0$. As $N-D$ is ample then, possibly replacing $F$ by a small multiple, we may assume that $\mu^{}(N-D)-F$ is ample. As $\mu^{}(K_{X}+\Delta+N)+E-\llcorner\mu^{}D\lrcorner-(K_{Y}+\Gamma+\\{\mu^{}D\\}+F)=\mu^{}(N-D)-F,$ is ample, Kawamata-Viehweg vanishing implies that $H^{i}(Y,\mathcal{O}_{Y}(\mu^{}(K_{X}+\Delta+N)+E-\llcorner\mu^{}D\lrcorner))=0,$ for $i>0$. Since the Leray-Serre spectral sequence degenerates, this gives (2), and (3) follows from (2). ∎ ###### Proof of (4.1). Since $(X,\Delta+\frac{k-1}{m}P)$ is purely log terminal, $(S,\Omega+\frac{k-1}{m}P\|_{S})$ is kawamata log terminal and $S$ is contained in the support neither of $A$ nor of $P$. If $\Sigma\in\|k(K_{S}+\Phi)\|$ then we may pick a divisor $G\in\|m(K_{X}+\Delta+C)+P\|\quad\text{such that}\quad G\|_{S}=l\Sigma+m(\Omega-\Phi+C\|_{S})+P\|_{S}.$ Let $\Lambda=\frac{k-1}{m}G+B\qquad\text{and}\qquad N=k(K_{X}+\Delta)-K_{X}-S.$ As the support of the $\mathbb{Q}$-divisor $\Lambda\geq 0$ does not contain $S$ and by assumption $N-\Lambda\sim_{\mathbb{Q}}C-\frac{k-1}{m}P,$ is ample, (4.4) implies that sections of $H^{0}(S,\mathcal{J}_{\Lambda\|_{S}}(k(K_{S}+\Omega)))$ extend to sections of $H^{0}(X,\mathcal{O}_{X}(k(K_{X}+\Delta)))$. Now $\displaystyle\phantom{\leq}\Lambda\|_{S}-(\Sigma+k(\Omega-\Phi))$ $\displaystyle=\frac{k-1}{m}(l\Sigma+m(\Omega-\Phi+C\|_{S})+P\|_{S})+B\|_{S}-(\Sigma+k(\Omega-\Phi))$ $\displaystyle\leq\Omega+\frac{k-1}{m}P\|_{S}.$ As $(S,\Omega+\frac{k-1}{m}P\|_{S})$ is kawamata log terminal, $\mathcal{J}_{\Omega+\frac{k-1}{m}P\|_{S}}=\mathcal{O}_{S}$ and we are done by (3) of (4.3). ∎ ## 5\. Asymptotic multiplier ideal sheaves ###### Definition 5.1. Let $X$ be a normal variety and let $D$ be a divisor. An additive sequence of linear systems associated to $D$ is a sequence $V_{\bullet}$, such that $V_{m}\subset\mathbb{P}(H^{0}(X,\mathcal{O}_{X}(mD)))$ and $V_{i}+V_{j}\subset V_{i+j}.$ ###### Definition-Lemma 5.2. Suppose that $(X,\Delta)$ is log smooth, where $\Delta$ is reduced and let $V_{\bullet}$ be an additive sequence of linear systems associated to a divisor $D$. Assume that there is an integer $k>0$ such that no log canonical centre of $(X,\Delta)$ is contained in the base locus of $V_{k}$. If $c$ is a positive real number and $p$ and $q$ are positive integers divisible by $k$ then $\mathcal{J}_{\Delta,\frac{c}{p}\cdot V_{p}}\subset\mathcal{J}_{\Delta,\frac{c}{q}\cdot V_{q}}\qquad\text{$\forall q$ divisible by $p$.}$ In particular the asymptotic multiplier ideal sheaf of $V_{\bullet}$ $\mathcal{J}_{\Delta,c\cdot V_{\bullet}}=\bigcup_{p>0}\mathcal{J}_{\Delta,\frac{c}{p}\cdot V_{p}},$ is given by $\mathcal{J}_{\Delta,c\cdot V_{\bullet}}=\mathcal{J}_{\Delta,\frac{c}{p}\cdot V_{p}}$, for $p$ sufficiently large and divisible. If we take $V_{m}=\|mD\|$ the complete linear system, then define $\mathcal{J}_{\Delta,c\cdot\\|D\\|}=\mathcal{J}_{\Delta,c\cdot V_{\bullet}},$ and if $S$ is a component of $\Delta$ and we take $W_{m}=\|mD\|_{S}$, then define $\mathcal{J}_{\Theta,c\cdot\\|D\\|_{S}}=\mathcal{J}_{\Theta,c\cdot W_{\bullet}},$ where $\Theta=(\Delta-S)\|_{S}$. ###### Proof. If $p$ divides $q$ then pick a common log resolution $\mu\colon Y\longrightarrow X$ of $V_{p}$, $V_{q}$ and $(X,\Delta)$ and note that $\frac{1}{q}F_{q}\leq\frac{1}{p}F_{p},$ where $F_{p}$ is the fixed locus of $\mu^{}V_{p}$ and $F_{q}$ is the fixed locus of $\mu^{}V_{q}$. Therefore $\mathcal{J}_{\Delta,\frac{c}{p}\cdot V_{p}}\subset\mathcal{J}_{\Delta,\frac{c}{q}\cdot V_{q}}$. The equality $\mathcal{J}_{\Delta,c\cdot V_{\bullet}}=\mathcal{J}_{\Delta,\frac{c}{p}\cdot V_{p}},$ now follows as $X$ is Noetherian. ∎ We are now ready to state the main result of this section: ###### Theorem 5.3. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$. Suppose that $(X,\Delta=S+B)$ is log smooth and purely log terminal of dimension $n$, where $S=\llcorner\Delta\lrcorner$ is irreducible and let $k$ be a positive integer such that $D=k(K_{X}+\Delta)$ is integral. Let $A$ be any ample $\mathbb{Q}$-divisor on $X$. Let $q$ and $r$ be any positive integers such that $Q=qA$ is very ample, $rA$ is Cartier and $(j-1)K_{X}+\Xi+rA$ is ample for every Cartier divisor $0\leq\Xi\leq j\ulcorner{\Delta}\urcorner$ and every integer $1\leq j\leq k+1$. If the stable base locus of $D$ does not contain any log canonical centre of $(X,\ulcorner{\Delta}\urcorner)$, then $\mathcal{J}_{\\|mD\|_{S}\\|}\subset\mathcal{J}_{\Theta,\\|mD+P\\|_{S}}\qquad\text{for all}\qquad m\in\mathbb{N},$ where $\Theta=\ulcorner{B}\urcorner\|_{S}$, $p=qn+r$ and $P=pA$. Moreover, we have $\pi_{}\mathcal{J}_{\\|mD\|_{S}\\|}(mD+P)\subset\operatorname{Im}\left(\pi_{}\mathcal{O}_{X}(mD+P)\to\pi_{}\mathcal{O}_{S}(mD+P)\right)$ for all $m\in\mathbb{N}$. We will need some results about the sheaves $\mathcal{J}_{\Delta,c\cdot V_{\bullet}}$, most of which are easy generalisations of the corresponding facts for the usual asymptotic multiplier ideal sheaves. ###### Lemma 5.4. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$ and let $D$ be a $\mathbb{Q}$-Cartier divisor. Suppose that $(X,\Delta)$ is log smooth, $\Delta$ is reduced and the stable base locus of $D$ contains no log canonical centre of $(X,\Delta)$. Then 1. (1) for any real numbers $0<c_{1}\leq c_{2}$ there is a natural inclusion $\mathcal{J}_{\Delta,c_{2}\cdot\\|D\\|}\subset\mathcal{J}_{\Delta,c_{1}\cdot\\|D\\|},$ and 2. (2) if $D$ is Cartier and $S$ is a component of $\Delta$, then the image of the map $\pi_{}\mathcal{O}_{X}(D)\longrightarrow\pi_{}\mathcal{O}_{S}(D),$ is contained in $\pi_{}\mathcal{J}_{\Theta,\\|D\\|_{S}}(D)$ where $\Theta=(\Delta-S)\|_{S}$. ###### Proof. (1) is immediate from the definitions. Suppose that $D$ is Cartier. Pick an integer $p$ such that $\mathcal{J}_{\Theta,\\|D\\|_{S}}=\mathcal{J}_{\Theta,\frac{1}{p}\cdot\|pD\|_{S}},$ and a log resolution $\mu\colon Y\longrightarrow X$ of $\|D\|$, $\|pD\|$ and $(X,\Delta)$. Let $T$ be the strict transform of $S$, let $F_{1}$ be the fixed locus of $\mu^{}\|D\|$ and let $F_{p}$ be the fixed locus of $\mu^{}\|pD\|$. We have $(\pi\circ\mu)_{}\mathcal{O}_{Y}(\mu^{}D-F_{1})=\pi_{}\mathcal{O}_{X}(D)=(\pi\circ\mu)_{}\mathcal{O}_{Y}(E+\mu^{}D).$ The first equality follows by definition of $F_{1}$ and the second follows as $E\geq 0$ is exceptional. As there are inequalities $\mu^{}D-F_{1}\leq\mu^{}D-\llcorner F_{p}/p\lrcorner\leq E+\mu^{}D-\llcorner F_{p}/p\lrcorner\leq E+\mu^{}D,$ the image of $\pi_{}\mathcal{O}_{X}(D)$ is equal to the image of $(\pi\circ\mu)_{}\mathcal{O}_{Y}(E+\mu^{}D-\llcorner F_{p}/p\lrcorner).$ Thus the image of $\pi_{}\mathcal{O}_{X}(D)$ is contained in $(\pi\circ\mu)_{}\mathcal{O}_{T}(E+\mu^{}D-\llcorner F_{p}/p\lrcorner)=\pi_{}\mathcal{J}_{\Theta,\\|D\\|_{S}}(D).\qed$ ###### Lemma 5.5. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$ and let $D$ be a Cartier divisor. Suppose that $(X,\Delta)$ is log smooth and $\Delta$ is reduced. Let $S$ be a component of $\Delta$ and $\Theta=(\Delta-S)\|_{S}$. If $\mathbf{B}_{+}(D)$ contains no log canonical centres of $(X,\Delta)$ then the image of the map $\pi_{}\mathcal{O}_{X}(K_{X}+\Delta+D)\longrightarrow\pi_{}\mathcal{O}_{S}(K_{S}+\Theta+D),$ contains $\pi_{}\mathcal{J}_{\Theta,\\|D\\|_{S}}(K_{S}+\Theta+D).$ ###### Proof. Pick an integer $p>1$ such that $\mathcal{J}_{\Theta,\\|D\\|_{S}}=\mathcal{J}_{\Theta,\frac{1}{p}\cdot\|pD\|_{S}},$ and there is a divisor $A+B\in\|pD\|$ where $A\geq 0$ is a general very ample divisor and $B\geq 0$ contains no log canonical centres of $(X,\Delta)$. By the resolution lemma of [13], we may find a log resolution $\mu\colon Y\longrightarrow X$ of $\|pD\|$ and of $(X,\Delta)$ which is an isomorphism over every log canonical centre of $(X,\Delta)$. Let $F_{p}$ be the fixed divisor of $\mu^{}\|pD\|$, $M_{p}=p\mu^{}D-F_{p}$ and let $\Gamma$ and $T$ be the strict transforms of $\Delta$ and $S$. We have a short exact sequence $0\longrightarrow\mathcal{O}_{Y}(G-T)\longrightarrow\mathcal{O}_{Y}(G)\longrightarrow\mathcal{O}_{T}(G)\longrightarrow 0,$ where $G=K_{Y}+\Gamma+\mu^{}D-\llcorner F_{p}/p\lrcorner$. As $\mu^{}A$ is base point free and $\mu^{}(A+B)\in\mu^{}\|pD\|$, the divisor $C:=\mu^{}B-F_{p}$ is effective. Note that $M_{p}-C\sim\mu^{}A$. As no component of $C$ is a component of $\Gamma$, we may pick $0<\delta\leq 1/p$ and an exceptional $\mathbb{Q}$-divisor $F\geq 0$ such that $(Y,\Gamma-T+\\{F_{p}/p\\}+\delta(C+F))$ is divisorially log terminal and $\mu^{}A-F$ is ample. As $\|M_{p}\|$ is free, $M_{p}/p$ is nef and so $\displaystyle G-T-(K_{Y}+\Gamma-T+\\{\frac{1}{p}F_{p}\\}+\delta(C+F))$ $\displaystyle=\frac{1}{p}M_{p}-\delta(C+F)$ $\displaystyle\sim_{\mathbb{Q}}(\frac{1}{p}-\delta)M_{p}+\delta(\mu^{}A-F),$ is ample. In particular Kawamata-Viehweg vanishing implies that $R^{1}\phi_{}\mathcal{O}_{Y}(G-T)=0$ where $\phi=\pi\circ\mu$. Therefore the homomorphism $\pi_{}\mathcal{O}_{X}(K_{X}+\Delta+D)\supset\phi_{}\mathcal{O}_{Y}(G)\longrightarrow\phi_{}\mathcal{O}_{T}(G)=\pi_{}\mathcal{J}_{\Theta,\\|D\\|_{S}}(K_{S}+\Theta+D),$ is surjective. ∎ ###### Theorem 5.6. Let $\pi\colon X\longrightarrow Z$ be a projective morphism, where $Z$ is affine and $X$ is a smooth variety of dimension $n$. If $D$ is a Cartier divisor whose stable base locus is a proper subset of $X$, $A$ is an ample Cartier divisor and $H$ is a very ample divisor then $\mathcal{J}_{\\|D\\|}(D+K_{X}+A+nH)$ is globally generated. ###### Proof. Pick an integer $p>0$ such that if $pB\in\|pD\|$ is a general element, then $\mathcal{J}_{\\|D\\|}=\mathcal{J}_{\frac{1}{p}\cdot\|pD\|}=\mathcal{J}_{B}.$ Then by (2) of (4.4), $H^{i}(X,\mathcal{J}_{\\|D\\|}(D+K_{X}+A+mH))=0$ for all $i>0$ and $m\geq 0$ and we may apply (5.7). ∎ ###### Lemma 5.7. Let $\pi\colon X\longrightarrow Z$ be a projective morphism where $X$ is smooth of dimension $n$, $Z$ is affine and let $H$ be a very ample divisor. If $\mathcal{F}$ is any coherent sheaf such that $H^{i}(X,\mathcal{F}(mH))=0$, for $i>0$ and for all $m\geq-n$ then $\mathcal{F}$ is globally generated. ###### Proof. Pick $x\in X$. Let $\mathcal{T}\subset\mathcal{F}$ be the torsion subsheaf supported at $x$, and let $\mathcal{G}=\mathcal{F}/\mathcal{T}$. Then $H^{i}(X,\mathcal{G}(mH))=0$ for $i>0$ and for all $m\geq-n$ and $\mathcal{F}$ is globally generated if and only if $\mathcal{G}$ is globally generated. Replacing $\mathcal{F}$ by $\mathcal{G}$ we may therefore assume that $\mathcal{T}=0$. Pick a general element $Y\in\|H\|$ containing $x$. As $\mathcal{T}=0$ there is an exact sequence $0\longrightarrow\mathcal{F}(-Y)\longrightarrow\mathcal{F}\longrightarrow\mathcal{G}\longrightarrow 0,$ where $\mathcal{G}=\mathcal{F}\otimes\mathcal{O}_{Y}$. As $H^{i}(Y,\mathcal{G}(mH))=0$, for $i>0$ and for all $m\geq-(n-1)$, $\mathcal{G}$ is globally generated by induction on the dimension. As $H^{1}(X,\mathcal{F}(-Y))=0$ it follows that $\mathcal{F}$ is globally generated. ∎ ###### Proof of (5.3). We follow the argument of [4] which in turn is based on the ideas of [6], [12] and [15]. We proceed by induction on $m$. The statement is clear for $m=0$, and so it suffices to show that $\mathcal{J}_{\\|(m+1)D\|_{S}\\|}\subset\mathcal{J}_{\Theta,\\|(m+1)D+P\\|_{S}},$ assuming that $\mathcal{J}_{\\|tD\|_{S}\\|}\subset\mathcal{J}_{\Theta,\\|tD+P\\|_{S}}\qquad\text{for all}\qquad t\leq m.$ If $\Delta=\sum\delta_{i}\Delta_{i}$, where each $\Delta_{i}$ is a prime divisor, then for any $1\leq s\leq k$, put $\Delta^{s}=\sum_{i\,:\,\delta_{i}>(k-s)/k}\Delta_{i}.$ We have * • each $\Delta^{s}$ is integral, * • $\displaystyle{S=\Delta^{1}\leq\Delta^{2}\leq\cdots\leq\Delta^{k}=\ulcorner{\Delta}\urcorner}$, and * • $\displaystyle{\Delta=\frac{1}{k}\sum_{s=1}^{k}\Delta^{s}}$, and these properties uniquely determine the divisors $\Delta^{s}$. We let $\Delta^{k+1}=\ulcorner{\Delta}\urcorner$. We recursively define integral divisors $D_{\leq s}$ by the rule $D_{\leq s}=\begin{cases}0&\text{if $s=0$}\\\ K_{X}+\Delta^{s}+D_{\leq s-1}&1\leq s\leq k.\end{cases}$ Note that $D_{\leq k}=D$. By (1) of (5.4) there is an inclusion $\mathcal{J}_{\\|(m+1)D\|_{S}\\|}\subset\mathcal{J}_{\\|mD\|_{S}\\|},$ and so it suffices to prove that there are inclusions $\mathcal{J}_{\\|mD\|_{S}\\|}\subset\mathcal{J}_{\Theta^{s+1},\\|mD+D_{\leq s}+P\\|_{S}},$ for $0\leq s\leq k$, where $\Theta^{i}=(\Delta^{i}-S)\|_{S}$ for $1\leq i\leq k+1$. Thus $\Theta^{k}=\Theta^{k+1}=\Theta$ and $\Theta^{1}=0$. We proceed by induction on $s$. Now $\mathcal{J}_{\\|mD\|_{S}\\|}\subset\mathcal{J}_{\Theta,\\|mD+P\\|_{S}}\subset\mathcal{J}_{\Theta^{1},\\|mD+P\\|_{S}}.$ The first inclusion holds by assumption and since $\Theta^{1}\leq\Theta$, (2) of (4.3) implies the second inclusion. Thus (5) holds when $s=0$. Now suppose that (5) holds for $s\leq t-1$. Note that $\displaystyle mD+D_{\leq t}+P$ $\displaystyle=K_{X}+\Delta^{t}+(D_{\leq t-1}+P)+mD$ ($\dagger$) $\displaystyle=mD+K_{X}+(\Delta^{t}+D_{\leq t-1}+rA)+nQ,$ where, by assumption, both $D_{\leq t-1}+P$ and $\Delta^{t}+D_{\leq t-1}+rA$ are ample for any $1\leq t\leq k+1$. In particular $\mathbf{B}_{+}(mD+D_{\leq t-1}+P)$ contains no log canonical centres of $(X,\ulcorner{\Delta}\urcorner)$. Then $\displaystyle\pi_{}\mathcal{J}_{\\|mD\|_{S}\\|}(mD+D_{\leq t}+P)$ $\displaystyle\subset\pi_{}\mathcal{J}_{\Theta^{t},\\|mD+D_{\leq t-1}+P\\|_{S}}(mD+D_{\leq t}+P)$ $\displaystyle\subset\operatorname{Im}\left(\pi_{}\mathcal{O}_{X}(mD+D_{\leq t}+P)\longrightarrow\pi_{}\mathcal{O}_{S}(mD+D_{\leq t}+P)\right)$ $\displaystyle\subset\pi_{}\mathcal{J}_{\Theta^{t+1},\\|mD+D_{\leq t}+P\\|_{S}}(mD+D_{\leq t}+P).$ The first inclusion holds as we are assuming (5) for $s=t-1$, the second inclusion holds by (5) and (5.5) and the last inclusion follows from (2) of (5.4). But (5) and (5.6) imply that $\mathcal{J}_{\\|mD\|_{S}\\|}(mD+D_{\leq t}+P),$ is generated by global sections and so $\mathcal{J}_{\\|mD\|_{S}\\|}\subset\mathcal{J}_{\Theta^{t+1},\\|mD+D_{\leq t}+P\\|_{S}}.$ The inclusion $\pi_{}\mathcal{J}_{\\|mD\|_{S}\\|}(mD+P)\subset\operatorname{Im}\left(\pi_{}\mathcal{O}_{X}(mD+P)\to\pi_{}\mathcal{O}_{S}(mD+P)\right),$ is part of the inclusions proved above when $s=k$. ∎ ## 6\. Lifting sections ###### Lemma 6.1. Let $D\geq 0$ be a Cartier divisor on a normal variety $X$, and let $Z\subset X$ be an irreducible subvariety. Then $\liminf\frac{\operatorname{mult}_{Z}(\|mD\|)}{m}=\lim\frac{\operatorname{mult}_{Z}(\|m!D\|)}{m!}.$ ###### Proof. Note that if $a$ divides $b$ then $\frac{\operatorname{mult}_{Z}(\|aD\|)}{a}\geq\frac{\operatorname{mult}_{Z}(\|bD\|)}{b},$ whence the result. ∎ ###### Lemma 6.2. Let $D\subset X$ be a divisor on a smooth variety and $Z$ a closed subvariety. If $\lim\operatorname{mult}_{Z}(\|m!D\|)/m!=0$ then $Z$ is not contained in $\mathbf{B}_{-}(D)$. ###### Proof. Let $A$ be any ample divisor. Pick $l>0$ such that $lA-K_{X}$ is ample. If $m>l$ is sufficiently divisible then $\mathcal{J}_{\\|mD\\|}(m(D+A))$ is globally generated by (5.6). But if $p>0$ is sufficiently large and divisible and $D_{mp}\in\|mpD\|$ is general, then $\operatorname{mult}_{Z}D_{mp}=\operatorname{mult}_{Z}\|mpD\|<p$ and $\mathcal{J}_{\\|mD\\|}=\mathcal{J}_{(1/p)D_{mp}}.$ But since $\operatorname{mult}_{Z}D_{mp}/p<1$ it follows that $(X,D_{mp}/p)$ is kawamata log terminal, in a neighbourhood of the generic point of $Z$. Thus $Z$ is not contained in the co-support of $\mathcal{J}_{\\|mD\\|}$ and so $Z$ is not contained in the base locus of $m(D+A)$. ∎ ###### Theorem 6.3. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$, where $(X,\Delta=S+A+B)$ is a purely log terminal pair, $S=\llcorner\Delta\lrcorner$ is irreducible, $(X,S)$ is log smooth, $A\geq 0$ is a general ample $\mathbb{Q}$-divisor, $B\geq 0$ is a $\mathbb{Q}$-divisor and $(S,\Omega+A\|_{S})$ is canonical, where $\Omega=(\Delta-S)\|_{S}$. Assume that the stable base locus of $K_{X}+\Delta$ does not contain $S$. Let $F=\lim F_{l!}$, where, for any positive and sufficiently divisible integer $m$, we let $F_{m}=\operatorname{Fix}(\|m(K_{X}+\Delta)\|_{S})/m.$ If $\epsilon>0$ is any rational number such that $\epsilon(K_{X}+\Delta)+A$ is ample and if $\Phi$ is any $\mathbb{Q}$-divisor on $S$ and $k>0$ is any integer such that 1. (1) both $k\Delta$ and $k\Phi$ are Cartier, and 2. (2) $\Omega\wedge\lambda F\leq\Phi\leq\Omega$, where $\lambda=1-\epsilon/k$, then $\|k(K_{S}+\Omega-\Phi)\|+k\Phi\subset\|k(K_{X}+\Delta)\|_{S}.$ ###### Proof. By assumption $A=H/m$, where $H$ is very ample and a very general element of $\|H\|$ and $m\geq 2$ is an integer. If $C=A/k$, then $A+(k-1)C=\frac{2k-1}{km}H,$ and so $(X,\Delta+(k-1)C=S+\frac{2k-1}{km}H+B)$ is purely log terminal, as $\frac{2k-1}{km}<1.$ On the other hand, $(S,\Omega+C\|_{S}),$ is canonical as we are even assuming that $(S,\Omega+A\|_{S})$ is canonical. Pick $\eta>\epsilon/k$ rational so that $\eta(K_{X}+\Delta)+C$ is ample and let $\mu=1-\eta<\lambda=1-\epsilon/k$. If $l>0$ is any sufficiently divisible integer so that $O=l(\eta(K_{X}+\Delta)+C)$ is very ample, then $\displaystyle G_{l}$ $\displaystyle=\operatorname{Fix}(\|l(K_{X}+\Delta+C)\|_{S})/l$ $\displaystyle=\operatorname{Fix}(\|l\mu(K_{X}+\Delta)+O\|_{S})/l$ $\displaystyle\leq\operatorname{Fix}(\|l\mu(K_{X}+\Delta)\|_{S})/l$ $\displaystyle=\mu F_{\mu l}.$ Thus $\lim G_{l!}\leq\mu\lim F_{l!}=\mu F.$ On the other hand (6.2) implies that there is a positive integer $l$ such that every prime divisor on $S$ which does not belong to the support of $F$ does not belong to the base locus of $\|l(K_{X}+\Delta+C)\|$. Thus we may pick a positive integer $l$ such that * • $k$ divides $l$, * • $lC$ is Cartier, and * • $G_{l}\leq\lambda F$. Let $f\colon Y\longrightarrow X$ be a log resolution of the linear system $\|l(K_{X}+\Delta+C)\|$ and of $(X,\Delta+C)$. We may write $K_{Y}+\Gamma=f^{}(K_{X}+\Delta+C)+E,$ where $\Gamma\geq 0$ and $E\geq 0$ have no common components, $f_{}\Gamma=\Delta+C$ and $f_{}E=0$. Then $H_{l}=\operatorname{Fix}(l(K_{Y}+\Gamma))/l=\operatorname{Fix}(lf^{}(K_{X}+\Delta+C))/l+E.$ If $\Xi=\Gamma-\Gamma\wedge H_{l}$ then $l(K_{Y}+\Xi)$ is Cartier and $\operatorname{Fix}(l(K_{Y}+\Xi))$ and $\Xi$ share no common components. Since the mobile part of $\|l(K_{Y}+\Xi)\|$ is free and the support of $\operatorname{Fix}(l(K_{Y}+\Xi))+\Xi$ has normal crossings it follows that the stable base locus of $K_{Y}+\Xi$ contains no log canonical centres of $(Y,\ulcorner{\Xi}\urcorner)$ (which are nothing but the strata of $\ulcorner{\Xi}\urcorner$). Let $H\geq 0$ be any ample divisor on $Y$. Pick positive integers $m$ and $q$ such that $l$ divides $m$ and $Q=qH$ is very ample. Let $T$ be the strict transform of $S$, let $\Gamma_{T}=(\Gamma-T)\|_{T}$ and let $\Xi_{T}=(\Xi-T)\|_{T}$. If $\tau\in H^{0}(T,\mathcal{O}_{T}(m(K_{T}+\Xi_{T})))=H^{0}(T,\mathcal{J}_{\\|m(K_{T}+\Xi_{T})\\|}(m(K_{T}+\Xi_{T}))),$ and $\sigma\in H^{0}(T,\mathcal{O}_{T}(Q))$ then $\sigma\cdot\tau\in H^{0}(T,\mathcal{J}_{\\|m(K_{T}+\Xi_{T})\\|}(m(K_{T}+\Xi_{T})+Q)).$ On the other hand, if $q$ is sufficiently large and divisible then by (5.3) $H^{0}(T,\mathcal{J}_{\\|m(K_{T}+\Xi_{T})\\|}(m(K_{T}+\Xi_{T})+Q))$ is contained in the image of $H^{0}(Y,\mathcal{O}_{Y}(m(K_{Y}+\Xi)+Q))\longrightarrow H^{0}(T,\mathcal{O}_{T}(m(K_{T}+\Xi_{T})+Q)).$ Hence there is a fixed $q$ such that whenever $l$ divides $m$, we have $\|m(K_{T}+\Xi_{T})\|+m(\Gamma_{T}-\Xi_{T})+\|Q\|_{T}\|\subset\|m(K_{Y}+\Gamma)+Q\|_{T}.$ If $g=f\|_{T}\colon T\longrightarrow S$ then $g_{}\Gamma_{T}=\Omega+C\|_{S}$ and since $g_{}\Xi_{T}\leq\Omega+C\|_{S}$ and $(S,\Omega+C\|_{S})$ is canonical, we have $\|m(K_{S}+g_{}\Xi_{T})\|=g_{}\|m(K_{T}+\Xi_{T})\|$. Therefore, applying $g_{}$, we obtain $\|m(K_{S}+g_{}\Xi_{T})\|+m(\Omega+C\|_{S}-g_{}\Xi_{T})+P\|_{S}\subset\|m(K_{X}+\Delta+C)+P\|_{S},$ where $P=f_{}Q$. Since for every prime divisor $L$ on $S$ we have $\operatorname{mult}_{L}G_{l}=\operatorname{mult}_{L^{\prime}}\operatorname{Fix}(\|l(K_{Y}+\Gamma)\|_{T})/l=\operatorname{mult}_{L^{\prime}}H_{l}\|_{T},$ where $L^{\prime}$ is the strict transform of $L$ on $T$, it follows that $g_{}\Xi_{T}-C\|_{S}=\Omega-\Omega\wedge G_{l}\geq\Omega-\Omega\wedge\lambda F\geq\Omega-\Phi\geq 0.$ Therefore $\|m(K_{S}+\Omega-\Phi)\|+m\Phi+(mC+P)\|_{S}\subset\|m(K_{X}+\Delta+C)+P\|_{S},$ for any $m$ divisible by $l$. In particular if we pick $m$ so that $C-\frac{k-1}{m}P$ is ample and $(X,\Delta+\frac{k-1}{m}P)$ is purely log terminal then the result follows by (4.1). ∎ ## 7\. Rationality of the restricted algebra In this section we will prove: ###### Theorem 7.1. Assume Theorem Fn-1. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$, where $(X,\Delta=S+A+B)$ is a purely log terminal pair of dimension $n$, $S=\llcorner\Delta\lrcorner$ is irreducible, $(X,S)$ is log smooth, $A\geq 0$ is a general ample $\mathbb{Q}$-divisor, $B\geq 0$ is a $\mathbb{Q}$-divisor and $(S,\Omega+A\|_{S})$ is canonical, where $\Omega=(\Delta-S)\|_{S}$. Assume that the stable base locus of $K_{X}+\Delta$ does not contain $S$. Let $F=\lim F_{l!}$ where, for any positive and sufficiently divisible integer $m$, we let $F_{m}=\operatorname{Fix}(\|m(K_{X}+\Delta)\|_{S})/m.$ Then $\Theta=\Omega-\Omega\wedge F$ is rational. In particular if both $k\Delta$ and $k\Theta$ are Cartier then $\|k(K_{S}+\Theta)\|+k(\Omega-\Theta)=\|k(K_{X}+\Delta)\|_{S},$ and $R_{S}(X,k(K_{X}+\Delta))\simeq R(S,k(K_{S}+\Theta)).$ ###### Proof. Suppose that $\Theta$ is not rational. Let $V\subset\operatorname{WDiv}_{\mathbb{R}}(S)$ be the vector space spanned by the components of $\Theta$. Then there is a constant $\delta>0$ such that if $\Phi\in V$ and $\\|\Phi-\Theta\\|<\delta$ then $\Phi\geq 0$ has the same support as $\Theta$ and moreover, by (2) of Theorem Fn-1, if $G$ is a prime divisor contained in the stable base locus of $K_{S}+\Theta$ then it is also contained in the stable base locus of $K_{S}+\Phi$. If $l(K_{X}+\Delta)$ is Cartier and $\Theta_{l}=\Omega-\Omega\wedge F_{l}$ then $\|l(K_{X}+\Delta)\|_{S}\subset\|l(K_{S}+\Theta_{l})\|+l(\Omega\wedge F_{l}).$ Hence $\operatorname{Fix}(l(K_{S}+\Theta_{l}))$ does not contain any components of $\Theta_{l}$. In particular the stable base locus of $K_{S}+\Theta_{l}$ does not contain any components of $\Theta_{l}$. But we may pick $l>0$ so that $\Theta_{l}\in V$ and $\\|\Theta_{l}-\Theta\|\|<\delta$. It follows that no component of $\Theta$ is in the stable base locus of $K_{S}+\Theta$. Let $W\subset V$ be the smallest rational affine space which contains $\Theta$. (3) of Theorem Fn-1 implies that there is a positive integer $r>0$ and a positive constant $\eta>0$ such that if $\Phi\in W$, $k\Phi/r$ is Cartier and $\\|\Phi-\Theta\\|<\eta$ then every component of $\operatorname{Fix}(k(K_{S}+\Phi))$ is in fact a component of the stable base locus of $K_{S}+\Theta$. Pick a rational number $\epsilon>0$ such that $\epsilon(K_{X}+\Delta)+A$ is ample. By Diophantine approximation, we may find a positive integer $k$, a divisor $\Phi$ on $S$ and a prime divisor $G$ (necessarily a component of $\Theta$ whose coefficient is irrational) such that 1. (1) $0\leq\Phi\in W$, 2. (2) both $k\Phi/r$ and $k\Delta/r$ are Cartier, 3. (3) $\\|\Phi-\Theta\\|<\min(\delta,\eta,f\epsilon/k)$ where $f$ is the smallest non- zero coefficient of $F\neq 0$, and 4. (4) $\operatorname{mult}_{G}\Phi>\operatorname{mult}_{G}\Theta$. ###### Claim 7.2. $\Omega\wedge\lambda F\leq\Omega-\Phi\leq\Omega$, where $\lambda=1-\epsilon/k$. ###### Proof of (7.2). Let $P$ be a prime divisor on $S$ and let $\omega$, $f$, $\phi$ and $\theta$ be the multiplicities of $\Omega$, $F$, $\Phi$ and $\Theta$ along $P$. We just need to check that $\min(\omega,\lambda f)\leq\omega-\phi.$ There are two cases. If $\omega\leq f$, then $\theta=0$ so that $\phi=0$ and (7) holds. If $\omega\geq f$, then $\theta=\omega-f$ and since $\\|\Phi-\Theta\\|<f\epsilon/k$, $\min(\omega,\lambda f)=\left(1-\frac{\epsilon}{k}\right)f\leq f-(\phi-\theta)=\omega-\phi.\qed$ (7.2), (2) and (6.3) imply that $\|k(K_{S}+\Phi)\|+k(\Omega-\Phi)\subset\|k(K_{X}+\Delta)\|_{S}.$ (4) implies that $G$ is a component of $\operatorname{Fix}(k(K_{S}+\Phi))$. (2) and $\\|\Phi-\Theta\\|<\eta$ imply that $G$ is a component of the stable base locus of $K_{S}+\Theta$, a contradiction. Thus $\Theta$ is rational. Hence $\Omega\wedge F$ is rational, and we are done by (6.3). ∎ ## 8\. Proof of (1.2) ###### Theorem 8.1. Assume Theorem Fn-1. Let $\pi\colon X\longrightarrow Z$ be a projective morphism to a normal affine variety $Z$. Suppose that $(X,\Delta=S+A+B)$ is a purely log terminal pair of dimension $n$, $S=\llcorner\Delta\lrcorner$ is irreducible and not contained in the stable base locus of $K_{X}+\Delta$, $A\geq 0$ is a general ample $\mathbb{Q}$-divisor and $B\geq 0$ is a $\mathbb{Q}$-divisor. Then there is a birational morphism $g\colon T\longrightarrow S$, a positive integer $l$ and a kawamata log terminal pair $(T,\Theta)$ such that $K_{T}+\Theta$ is $\mathbb{Q}$-Cartier and $R_{S}(X,l(K_{X}+\Delta))\cong R(T,l(K_{T}+\Theta)).$ ###### Proof. If $f\colon Y\longrightarrow X$ is a log resolution of $(X,\Delta)$ then we may write $K_{Y}+\Gamma^{\prime}=f^{}(K_{X}+\Delta)+E,$ where $\Gamma^{\prime}\geq 0$ and $E\geq 0$ have no common components, $f_{}\Gamma^{\prime}=\Delta$ and $f_{}E=0$. If $T$ is the strict transform of $S$ then we may choose $f$ so that $(T,\Psi^{\prime}=(\Gamma^{\prime}-T)\|_{T})$ is terminal. Note that $T$ is not contained in the stable base locus of $K_{Y}+\Gamma^{\prime}$ as $S$ is not contained in the stable base locus of $K_{X}+\Delta$. Pick a $\mathbb{Q}$-divisor $F$ such that $f^{}A-F$ is ample and $(Y,\Gamma^{\prime}+F)$ is purely log terminal. Pick $m>1$ so that $m(f^{}A-F)$ is very ample and pick $mC\in\|m(f^{}A-F)\|$ very general. Then $(Y,\Gamma=\Gamma^{\prime}-f^{}A+F+C\sim_{\mathbb{Q}}\Gamma^{\prime}),$ is purely log terminal and if $m$ is sufficiently large $(T,\Psi+C\|_{T})$ is terminal, where $\Psi=(\Gamma-T)\|_{T}$. On the other hand $\displaystyle R(X,k(K_{X}+\Delta))$ $\displaystyle\cong R(Y,k(K_{Y}+\Gamma))\qquad\text{and}$ $\displaystyle R_{S}(X,k(K_{X}+\Delta))$ $\displaystyle\cong R_{T}(Y,k(K_{Y}+\Gamma)),$ for any $k$ sufficiently divisible. Now apply (7.1) to $(Y,\Gamma)$. ∎ ###### Proof of (1.2). By (3.2) we may assume that $Z$ is affine and by (3.4), it suffices to prove that the restricted algebra is finitely generated. As $Z$ is affine, $S$ is mobile and as $f$ is birational, the divisor $\Delta-S$ is big. But then $\Delta-S\sim_{\mathbb{Q}}A+B,$ where $A$ is a general ample $\mathbb{Q}$-divisor and $B\geq 0$. As $S$ is mobile, we may assume that the support of $B$ does not contain $S$. Now $K_{X}+\Delta^{\prime}=K_{X}+S+(1-\epsilon)(\Delta-S)+\epsilon A+\epsilon B\sim_{\mathbb{Q}}K_{X}+\Delta,$ is purely log terminal, where $\epsilon$ is any sufficiently small positive rational number. By (3.3), we may replace $\Delta$ by $\Delta^{\prime}$. We may therefore assume that $\Delta=S+A+B$, where $A$ is a general ample $\mathbb{Q}$-divisor and $B\geq 0$. Since we are assuming Theorem Fn-1, (8.1) implies that the restricted algebra is finitely generated.∎ ## References * [1] F. Ambro, _Restrictions of log canonical algebras of general type_ , J. Math. Sci. Univ. Tokyo 13 (2006), no. 3, 409–437. * [2] C. Birkar, P. Cascini, C. Hacon, and J. McKernan, _Existence of minimal models for varieties of log general type_ , arXiv:math.AG/0610203. * [3] A. Corti, _3-fold flips after Shokurov_ , Flips for 3-folds and 4-folds (Alessio Corti, ed.), Oxford University Press, 2005, pp. 13–40. * [4] C. Hacon and J. McKernan, _Boundedness of pluricanonical maps of varieties of general type_ , Invent. Math. 166 (2006), no. 1, 1–25. * [5] by same author, _Extension theorems and the existence of flips_ , Flips for 3-folds and 4-folds (Alessio Corti, ed.), Oxford University Press, 2007, pp. 79–100. * [6] Y. Kawamata, _On the extension problem of pluricanonical forms_ , Algebraic geometry: Hirzebruch 70 (Warsaw, 1998), Contemp. Math., vol. 241, Amer. Math. Soc., Providence, RI, 1999, pp. 193–207. * [7] J. Kollár and S. Mori, _Birational geometry of algebraic varieties_ , Cambridge tracts in mathematics, vol. 134, Cambridge University Press, 1998. * [8] R. Lazarsfeld, _Positivity in algebraic geometry. I_ , Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 48, Springer-Verlag, Berlin, 2004, Classical setting: line bundles and linear series. * [9] by same author, _Positivity in algebraic geometry. II_ , Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 49, Springer-Verlag, Berlin, 2004, Positivity for vector bundles, and multiplier ideals. * [10] V. V. Shokurov, _Prelimiting flips_ , Proc. Steklov Inst. of Math. 240 (2003), 82–219. * [11] Y-T. Siu, _A General Non-Vanishing Theorem and an Analytic Proof of the Finite Generation of the Canonical Ring_ , arXiv:math.AG/0610740. * [12] by same author, _Invariance of plurigenera_ , Invent. Math. 134 (1998), no. 3, 661–673. * [13] E. Szabó, _Divisorial log terminal singularities_ , J. Math. Sci. Univ. Tokyo 1 (1994), no. 3, 631–639. * [14] S. Takayama, _Pluricanonical systems on algebraic varieties of general type_ , Invent. Math. 165 (2006), no. 3, 551–587. * [15] H. Tsuji, _Pluricanonical systems of projective varieties of general type_ , arXiv:math.AG/9909021.	arxiv-papers	2008-08-14T02:00:41	2024-09-04T02:48:57.275283	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Christopher D. Hacon, James McKernan", "submitter": "James McKernan", "url": "https://arxiv.org/abs/0808.1929" }
0808.1941	# Perfect coupling of light to surface plasmons with ultra-narrow linewidths M. Sukharev1∗, P. R. Sievert2, T. Seideman3, and J. B. Ketterson2,4 1Department of Applied Sciences and Mathematics, Arizona State University at the Polytechnic Campus, Mesa AZ, 85212, USA 2Department of Physics and Astronomy, Northwestern University, Evanston IL, 60208, USA 3Department of Chemistry, Northwestern University, Evanston IL, 60208, USA 4Department of Electrical and Computer Engineering, Northwestern University, Evanston IL, 60208, USA ∗corresponding author: maxim.sukharev@asu.edu ###### Abstract We examine the coupling of electromagnetic waves incident normal to a thin silver film that forms an oscillatory grating embedded between two otherwise uniform, semi-infinite half spaces. Two grating structures are considered, in one of which the mid point of the Ag film remains fixed whereas the thickness varies sinusoidally, while in the other the mid point oscillates sinusoidally whereas the film thicknesses remains fixed. On reducing the light wavelength from the long wavelength limit, we encounter signatures in the transmission, $T$, and reflection, $R$, coefficients associated with: i) the short-range surface plasmon mode, ii) the long-range surface plasmon mode, and iii) electromagnetic diffraction tangent to the grating. The first two features can be regarded as generalized (plasmon) Wood’s anomalies whereas the third is the first-order conventional (electromagnetic) Wood’s anomaly. The energy density at the film surface is enhanced for wavelengths corresponding to these three anomalies, particularly for the long range plasmon mode in thin films. When exciting the silver film with a pair of waves incident from opposite directions, we find that by adjusting the grating oscillation amplitude and fixing the relative phase of the incoming waves to be even or odd, $T+R$ can be made to vanish for one or the other of the plasmon modes; this corresponds to perfect coupling (impedance matching in the language of electrical engineering) between the incoming light and these modes. ## I Introduction There is currently much interest in the physical properties and possible applications of plasmons excited in various metallic structures AtwaterJAppPhys2005 ; OzbayScience2006 ; AbajoRMP2007 ; BarnesAdvMat2007 ; CatrysseNatPhys2007 ; Lezec98 ; Kawata08 , particularly those made from silver or gold, where relatively narrow resonances are observed relative to some other metals ZhangJPCC2008 . Much of this interest is associated with the high electric fields generated by various resonant responses. Common among the many applications of these fields is the enhancement of nonlinear optical responses for optical devices FendlerAdvMat2004 or spectroscopic applications VanDuyneExpRevMolD2004 . The simplest plasmonic structure is a single metallic sphere. In the limit where the sphere diameter is small compared to the wavelength of the incident light, the polarization is well described by the leading term in the Mie expansion, corresponding to a single peak at $\omega=\omega_{bp}/\sqrt{3}$, where $\omega_{bp}$ is the bulk plasmon frequency KreibigBook . As the sphere size increases higher order terms in the Mie expansion become important Mie . Smaller sphere radii yield larger field enhancements at the particle surface. For ellipsoids Nitzan82 , when the exciting field lies along the major axis of a prolate ellipsoid of revolution, the high curvature for the surface normal to that direction further enhances the field; this is the so-called lightning rod effect. One is then encouraged to examine structures with sharp corners, such as pyramids SchatzPyramids , or assemblies of these containing adjacent sharp corners leading to higher local field enhancements. Although the local fields associated with small finite structures and their arrays can be relatively high, they are in practice limited by various effects including boundary scattering and radiation losses. With regard to the latter, one is encouraged to examine structures where radiation is suppressed. The specific case of surface plasmons has long been studied RaetherBook . When propagating at a half space separating a bulk metal from an adjacent dielectric, radiation is kinematically forbidden; damping only involves scattering from inhomogeneities (e.g. a rough surface) and dissipation in the metal itself. In fact the very existence of a surface plasmon requires that its frequency, $\omega_{p}$, and wavevector, $k_{p}$, not match light propagating at any angle in the surrounding dielectric. For this reason it is common to couple to surface plasmons via an evanescent light wave and carefully control the coupling strength. In the commonly used Otto geometry OttoOriginal1968 this is accomplished by forming a thin dielectric layer between a prism and the free surface of the metal adjacent to it. By adjustment of the thickness of the coupling layer, the reflected beam in the prism can be made to vanish at some critical angle QueizerATR1973 , a condition we will refer to as _critical coupling_. If, in the Otto geometry, the metal layer (generally a vapor deposited thin film) has a finite thickness, plasmons involving both sides can be excited. If the dielectric constant of the coupling layer and that of the far side differ, modes with different wavelengths for a given frequency are present, which tend to localize on opposing sides of the metal film. When the dielectric constants on both sides are identical, however, the modes become degenerate; if in addition the film is thin, the modes are strongly coupled and split, with one being symmetric and the other antisymmetric with respect to the mid point of the film RaetherBook . The structure of these modes was first examined by Economou Economou69 . Later work by Sarid SaridPaper showed that the damping of the symmetric modes is greatly reduced as the film thins leading to long range propagation; the antisymmetric mode turns out to have a much shorter range. This paper focuses on coupling to the long- and short-range modes using a sinusoidal grating for the case of perpendicular incidence involving a special case of Wood’s anomaly, which we discuss at greater length below. Grating coupling has been used to experimentally couple to plasmon modes at a free surface, in connection with a study of the associated one-dimensional plasmon bandstructure Seymore83 . More recently the plasmon bandstructure for thin square-wave modulated films has been examined Kawata08 . Structures similar to those modeled here have recently been studied experimentally and will be reported elsewhere Ketterson09 . On the practical side, when coupled with an appropriate dye, the structures may be used to make future vertically emitting plasmonic band-gap lasers Kawata04 . The paper is arranged as follows. In Section II we consider electromagnetic modes in the Otto geometry in symmetric dielectric environments modeled with a dielectric constant, $\varepsilon_{Ag}\left(\omega\right)$ corresponding to silver embedded in a water environment. We also present calculations of intensity enhancements and line widths for critical coupling. Section III describes the methodology applied in our numerical simulations and illustrates results for in-phase and out-of-phase oscillatory gratings using the same silver and dielectric environment parameters used for the Otto geometry simulations. Specifically, we calculate surface averaged field enhancements, demonstrating large enhancements at the Wood’s anomalies, in particular for the long-range mode. Our conclusions are summarized in Section IV. ## II Long- and short-range plasmons in thin, flat metal films Although the magnetic field amplitude is largest within the metal for the symmetric mode, the corresponding electric field has a node at the mid point and its over-all amplitude is smaller than for the antisymmetric mode; hence it has a smaller dissipation, leading to a longer range (and correspondingly a longer lifetime). One of our goals in the present study is to achieve the largest possible electric field enhancements in a spatially resonant, film- based structure and therefore low dissipation is a desirable feature. We will refer to this symmetric mode as the _long-range plasmon_. In addition the energy density lies largely outside the film, with the result that the mode velocity approaches that of light in the surrounding dielectric as the film is thinned RaetherBook ; the mode can be thought of as being “stiff” in this sense. On the other hand the electric field for the antisymmetric mode is, over-all, larger within the metal. This leads to greater damping as the film is thinned; simultaneously the velocity falls (being “loaded” by the metal) and the mode is “soft” in this sense. We therefore refer to this mode as the _short-range plasmon_. In modeling the overall behavior of the incoming beam in an Otto geometry experiment, one can regard the system as a stack of dielectric layers and apply the Fresnel conditions at all interfaces, while simultaneously accounting for the propagation between interfaces. In the present study, we consider a silver film and approximate the dielectric function by the Drude model, $\varepsilon_{1}(\omega)=\varepsilon_{\infty}-\frac{\omega_{bp}^{2}}{(\omega^{2}+i\omega\Gamma)}$ (1) where the parameters are $\omega_{bp}=1.7901\times 10^{16}$ rad/sec, $\Gamma=3.0841\times 10^{14}$ rad/sec, and $\varepsilon_{\infty}=8.926$. This model is a good representation of silver films for wavelengths between $300$ nm and $700$ nm. We define $k_{0}=2\pi/\lambda_{0}=\omega/c$, the wave vector of the exciting light of wavelength $\lambda_{0}$ in vacuum, where $c$ is the velocity of light. For very thin films the effects of scattering from an additional boundary would enter; no attempt to incorporate such effects is made in what follows. Note the effects of scattering from a single boundary are implicitly contained in the measured reflectivity data from which the dielectric constant is obtained. Using dispersion relations for the plasmon modes in the Sarid geometry and our model parameters one can show that the high (long-range) and low (short-range) frequency modes lie, respectively, above and below that for a surface plasmon propagating at a half space, where the latter is given by $\frac{\omega}{\omega_{bp}}=\frac{1}{\sqrt{\varepsilon_{\infty}+\varepsilon_{2}}},$ (2) here we will take the dielectric constant of the host medium, $\varepsilon_{2}$, is $1.76$, corresponding to water. Eq. (2) results in $0.306$ for our model parameters. Both families of curves approach this limit as the film thicknesses increases. This indicates the validity of our solution, although it is not strictly within the range of applicability of the Drude model. We also note that as the real in-plane plasma momentum decreases, all curves merge with the ”light-line” for the dielectric medium, which is given by $(\omega/\omega_{bp})=(1/\sqrt{\varepsilon_{2}})(k_{\rho}^{\prime}/k_{bp})$; here $k_{bp}$ is the in-plane wave vector of the bulk plasmon and $k_{\rho}^{\prime}$ is the real part of the in-plane complex wave vector of the plasmon. As it was found in SaridPaper , the high frequency mode has a very long range for thin films. As noted above, this is due to destructive interference of the overlapping evanescent fields in the metal, which minimizes the energy dissipation for this mode and enhances the lifetime of the plasmons at the surface. The dissipation decreases quadratically with the film thickness and linearly with the imaginary part of $\varepsilon_{1}$. For the purpose of our paper it is desirable to couple into the long-range plasmon mode with the minimum radiation loss, as this should maximize the field strength on the metal film. For a featureless, flat, metal film and fixed excitation frequency, the only method available is ”tuning”, i.e. varying the thickness, of the dielectric (coupling) layer between the coupling prism and the metal film. The total reflection coefficient is then calculated from the generalized Fresnel coefficients, accounting for the multiple reflections from the layers involved. The incident light wavelength is swept while the nominal angle of incidence is varied to correspond to a fixed in- plane momentum $k_{\rho}^{\prime}$ of $(2\pi/400)$ nm-1 (this fixed value of $k_{\rho}^{\prime}$ is arbitrary but corresponds to the periodicity of the grating couplers considered in Section III below). Excitation of the plasmon modes is indicated by minima in the total reflection. The minimum corresponding to either mode can be ”tuned” to zero by adjusting the thickness of the dielectric layer. This is the critical coupling layer thickness and corresponds to perfect coupling or impedance matching. The energy reflection at critical coupling for the long-lived mode versus the wavelength for a number of silver film thicknesses is given in Fig. 1A. One can determine a linewidth at the half-power line at various film thicknesses and calculate the $Q$-factor, defined as $Q=\lambda/\Delta\lambda$. The results are shown in Fig. 1B. Note that the $Q$ significantly increases as the silver layer thins and increases by $4$ orders of magnitude at a thickness of $5$ nm, implying a corresponding build up of the stored energy; the electric field, which scales as the square root of the energy density, is similarly enhanced. ## III Optics of metal gratings at impedance matching conditions An alternative way of coupling into surface plasmons is to arrange for the metal film to act as a diffraction grating, a ”line grating” in the present study. The choice of the grating structure introduces a wide range of physical scenarios. A relatively simple choice involves a sinusoidal variation of the position of the metal/dielectric interfaces and here we will study two limiting cases the two limiting cases shown in Figs. 2A, B. In Fig. 2A the mid point of the metal film remains fixed but the thickness varies sinusoidally between values $2D_{1}$ and $2D_{2}$ KorovinOptLett08 ; in Fig. 2B we have a case where the mid point oscillates sinusoidally with amplitude $2D_{2}$ but the film thicknesses remains fixed at a value $2D_{1}$. The two sinusoids are, respectively, out-of-phase and in-phase for these two cases, as shown in Fig. 2A. Intuitively we expect the modes for the constant thickness case to closely resemble the Sarid modes on a flat film while simultaneously allowing grating coupling to an external light wave. For an incoming direction and wavelength for which the grating diffracts the light tangent to the grating, such that the diffracted wave is “in step” with the grating itself, one expects a stronger coupling; this enhanced interaction manifests itself as anomalies in other diffracted orders including the zeroth- order transmitted, $T$, and reflected, $R$, light intensity. Note that when energy is absorbed by the film, the condition $T+R=1$, which one would have for a stack of lossless dielectric layers, does not apply. The phenomena, first discovered by R. W. Wood, is referred to as _Wood’s anomaly_ ; the theoretical explanation was first given by Rayleigh and for this reason it is also called the Rayleigh-Wood anomaly. For a rigorous discussion the reader is referred to reference Oliner65 . The Wood’s phenomena is shown schematically in Fig. 3 for the case of a wave entering perpendicular to the grating. In this case we have equal and opposite diffracted waves (for a total of two, unlike the case for an arbitrary incident angle) that form a standing wave, and one may anticipate an even stronger interaction. As a generalization of the Rayleigh-Wood phenomena, we will include the two cases where the wavelength of the long- and short-range plasmons (rather than the electromagnetic wave) matches the grating spacing, which we can interpret as _plasmon Wood’s anomalies_. We note in passing that when the incoming wave is perpendicular to the film one couples to the plasmons at the second Brillouin zone point of the associated one-dimensional plasmonic band structure within the film that is associated with the presence of the periodic grating. The optical properties of the periodic metal structures depicted in the Fig. 2 were simulated using FDTD TafloveBook in two dimensions. We have restricted ourselves to the case of normal incidence; furthermore we limit the calculations to the transverse magnetic mode, TM, which couples to plasmons propagating in the same direction as the static grating waves. For two dimensions this mode is sometimes designated as the transverse electric mode TafloveBook , TEz, and has two electric and one magnetic component. The form given in Eq. (1), with identical parameters, forms the basis for the dielectric response of the metal to external EM excitation. The FDTD unit cell used in the simulations is shown schematically in the Fig. 2B. The upper and lower parts of the grid are terminated by horizontal absorbing boundaries, for which we use perfectly matched layers (PML) Berenger with a depth of $16$ spatial steps and exponential differences in order to avoid diffusion instabilities. In all simulations we use a uniform Yee grid with spatial steps $\delta x=\delta y=0.98$ nm and a time step $\delta t=\delta x/\left(1.5c\right)$. The metal film is assumed to be embedded in a dielectric with refractive index of $1.33$, corresponding to $\epsilon_{2}=1.76$. We note that the latter leads to a Wood’s anomaly at $532$ nm for the $400$ nm periodic gratings as discussed in the previous section. The gratings are excited by $x$-polarized plane waves that are generated along horizontal lines near the PML regions. All simulations have been performed in parallel using $128$ processors on the DataStar cluster at the San Diego Supercomputer Center and Saguaro cluster at the ASU High Performance Computing Center. An average execution time of our codes is approximately $20$ minutes or less. Since the FDTD algorithm is a time domain integrator, electromagnetic (EM) field components at some pre-defined incident wavelength can be obtained either by applying a long excitation pulse, with duration much longer than the characteristic resonant lifetime, or by exciting the system with an ultra- short incident pulse and subsequently Fourier transforming the resulting EM fields. The latter method requires a single FDTD run within which one obtains the entire spectrum. Here we employ simulations of phasor functions using a discrete Fourier transform ”on-the-fly” procedure with a short pulse excitation scheme TafloveBook ; SukharevPRB07 . In order to calculate the transmission, $T$, and reflection, $R$, coefficients, we numerically integrate the Poynting vector normal to the horizontal lines shown in Fig. 2B as dashed lines. Transmission, $T$, is calculated using the total fields, which consists of the EM waves scattered by the metal film and the incident field, whereas reflection, $R$, is simulated using the scattered fields only, so as to exclude standing waves generated by interference between the incident field and the EM waves scattered back by the film. Numerical convergence is verified by ascertaining that $T+R=1$ for a lossless dielectric for all wavelengths of interest. For the double-ended excitation scheme, i.e. for the case of excitation of the grating from both sides of the film, we perform simulations of the reflection for both sides of the structure using scattered fields only. Finally, in order to assure numerical convergence especially in the vicinity of sharp resonances we integrate Maxwell equations for as long as $2$ ps. Unlike the Otto geometry discussed above, we do not in general expect to be able to achieve critical coupling for our gratings, in the sense that all of the incoming energy is absorbed by the film; this would require two quantities to simultaneously vanish, namely the transmission and reflection coefficients, whereas we have only a single parameter, the amplitude of the grating oscillation, to vary (here we assume the thickness, or average thickness for the out-of-phase gratings, is held constant). For this reason we performed simulations where the beams simultaneously enter normal to the grating but from opposite sides. On symmetry grounds we anticipated that if these waves were in phase or out of phase we might selectively excite either the short- range or long-range plasmon modes, but not both; this was confirmed by the simulations, although the couplings for in-phase and out-of-phase sinusoids were reversed as will be discussed below. For such a case we then have only a single quantity, $T+R$, which has the same magnitude on both sides and simulations show that we may cause it to vanish (corresponding to critical coupling) at some critical value of our single grating amplitude parameter. We note that if the oscillation amplitudes on the upper and lower surfaces were independently varied (while maintaining constant average thickness), critical coupling should be achievable under single ended excitation, since we then have two independent parameters with which to simultaneously tune $T$ and $R$ through zero. Fig. 4A shows simulations for an out-of-phase grating (see Fig. 2A) and calculate $T+R$ as a function of the free space incident wavelength, $\lambda_{0}$, for three excitation schemes: i) single-ended excitation (solid line); ii) symmetric double-ended excitation (dash-dotted line) (where two in- phase plane waves emanating from symmetrically placed horizontal lines near the upper and lower PML regions excite the metal); and iii) anti-symmetric double-ended excitation (dashed line) (which is similar to the symmetric scheme except that the phase difference between the two incident waves is $\pi$). The single-ended excitation scheme clearly generates all three modes: an electromagnetic Wood’s anomaly located at $\lambda=532$ nm and the long- range and short-range surface plasmon modes localized at $545$ nm and $624$ nm, respectively. We note that the resonant wavelength for the electromagnetic Wood’s anomaly accurately corresponds to the grating period of $400$ nm in the dielectric media with the refractive index of $1.33$. It is apparent in Fig. 4A that the phase delay between the two incident pulses in the double-ended excitation scheme gives us a control over the symmetry of the excited surface plasmon mode. The observed behavior can be summarized as follows: for the case of out-of-phase sinusoidal gratings (which are symmetric with respect to the film center), in-phase incoming waves excite the symmetric mode whereas out-of-phase waves excite the antisymmetric mode. On the other hand, Fig. 4B shows calculations for an in-phase sinusoidal grating where the converse is true: in-phase incoming waves excite the antisymmetric mode whereas out-of-phase waves excite the symmetric mode. Provided that our gratings do not significantly depart from flat films, we expect the anti-symmetric and symmetric resonant wavelengths to be approximately given by the long- and short-range modes, which are governed by the film thickness as given in Fig. 1. For the in-phase gratings the thickness is given by $2D_{1}$ and we can then use the parameter $2D_{2}$ to control the coupling. For the out-of-phase sinusoids the resonant wavelength would, in the first approximation, be governed by the average thickness, $\left(D_{1}+D_{2}\right)$; the difference, $\left(D_{1}-D_{2}\right)$, can then be adjusted to control the coupling. Fig. 4A shows an example of an out-of-phase sinusoidal grating described by the parameters $D_{1}=5$ nm and $D_{2}=31$ nm. Here we achieve a near perfect coupling for both the short- and long-range modes with the same parameter set. Note that with the single ended excitation the maximum coupling is approximately $50\%$. Fig. 4B shows an example of an in-phase grating where the parameters $D_{1}=10$ nm and $D_{2}=14.4$ nm minimize $T+R$ for the long- range surface plasmon mode. As discussed in previous sections, at the impedance matching conditions incident EM radiation is efficiently transformed into propagating surface plasmons leading to strong field enhancements on the surface of the gratings. Fig. 5A presents the intensity enhancement factors averaged over the metal surface for the out-of-phase (solid line) and in-phase (dashed line) sinusoidal grating under the double-ended excitation conditions (here we consider only long-range modes). We note that high EM fields, significantly enhanced at the impedance matching conditions, are not localized near specific spatial regions, as in the case of tip-enhanced optical probes TERS , as expected fort a standing sinusoidal wave. Following the results of Section II we simulated the enhancement for several grating thicknesses as shown in Fig. 5B. The results confirm our general conclusions that with thinner films the intensity enhancements are much higher as follows from the high quality factors $Q$ (see Fig. 1B for details). ## IV Discussion and Conclusions The above calculations demonstrate that the largest enhancements occur for the long-range mode for an in-phase oscillation of the grating amplitude. In evaluating nonlinear optics applications requiring rapid response, one must keep in mind that resonance is always accompanied by dispersion and the associated distortion of waveforms; the number of field cycles required to build up a response is proportional to the $Q$ of the resonance and this may be particularly important when, e.g., trying to exploit plasmon waveguiding. Hence we envision the large enhancements obtained here would primarily be exploited for sensor applications. We recall that the large intensity enhancements for our grating structures given in section III were calculated on a per-unit-area basis. This property is particularly important when considering sensor applications involving fluorescence-based spectroscopy or Surface Enhanced Raman Spectroscopy (SERS) where we assume we have no control over where the target molecule will actually attach. It is sometimes assumed that the enhancement of SERS scales as the square of the intensity; although this appears physically unreasonable, we note that the surface averaged square of the intensity is also greatly enhanced. Although various plasmon resonant nanostructures can have “hot spots” with rather large intensity enhancements, exploiting these requires that the target molecules “find” these hot spots, which in turn requires effective mixing or long exposure times. If receptors are involved one would want them attached only on the hot spots; if they are widely distributed they will compete with the hot spots for target molecules thereby depleting the latter. Many other one-dimensional grating structures come to mind, including grooves and slots; both have multiple parameters (thickness, width, vertical position) associated with them, which can be tuned to achieve critical coupling under single ended excitation. The sharp edges associated with these structures can lead to further, but highly local, enhancements. Two-dimensional (e.g. hole) arrays are natural extensions. Our group is currently examining several of these structures. For SERS, and especially for fluorescence, the response of the grating at the emission (shifted) wavelength is also important, an issue under study. In connection with possible SERS applications we point out that it should be possible to engineer one- and two-dimensional structures where light is i) gathered and ii) concentrated in a “two step” process involving a periodic array of bumps, ridges or holes. Firstly, the periodic grating structure so formed would be engineered to efficiently collect the light in an impedance matched manner, as we have described. Second, the edges of the structures so patterned would optimize to concentrate these fields. We now discuss several technical issues regarding the implementation of the structures we have modeled. We used water for the dielectric in our simulations; i.e. we are implicitly assuming water-based sensor applications. In order to support the film, we require an insulating coating on a substrate whose dielectric constant can be adjusted to equal that of water. Such a material is Teflon AF© manufactured by Dupont. Although forming a sinusoidal grating is challenging, one can pattern and etch a grooved structure in, e.g., a glass or Si substrate by a variety of techniques, where the groove edges naturally develop a slope under simple etching protocols. Subsequently spinning on of Teflon AF would, via surface tension, further smooth the contour, thereby better approximating a sine wave. Depositing a uniform Ag film on a low surface tension material like Teflon AF may also be problematic. We note that silver reacts with water over time and so one might use Au (for which the enhancements will be lower) or spin on a very thin protective layer of Teflon AF over or within which any receptors would be incorporated Nenninger . Finally we must address the issue of matching the grating resonance frequency for the long-range mode to that of an available laser, say an argon ($517$ nm) or a YAG ($532$ nm) laser. Forming the structure with the required $d$-spacing is not an option, given the narrowness of the resonances. One can, however, take advantage of the degree of freedom mentioned in the Introduction, namely tipping the incoming laser beam out of the scattering plane, in particular in a direction parallel to the grating lines (corresponding to $\theta=\pi/2$). By making the $d$-spacing slightly smaller than that which would match the resonance one can tune through the resonance by decreasing $\theta$ slightly from the plane normal. At resonance we would retain a standing wave along the grating plane and directed normal to its lines but the plasmon wavevector would then have a small component parallel to the lines. ###### Acknowledgements. This research was supported in part by the NCLT program of the National Science Foundation at the Materials Research Institute of Northwestern University (grant number ESI-0426328) and the AFOSR DARPA program (grant number FA9550-08-1-0221). The numerical work was enabled by the resources of the National Science Foundation San Diego Supercomputer Center under grant number MCA06N016. ## References * (1) S. A. Maier and H. A. Atwater, J. App. Phys. 98, 011101 (2005). * (2) E. Ozbay, Science 311, 189 (2006). * (3) F. J. Garcia de Abajo, Rev. Mod. Phys. 79, 1267 (2007). * (4) H. F. Ghaemi, Tineke Thio, D. E. Grupp, T. W. Ebbesen, and H. J. Lezec, Phys. Rev. B 58, 6779 (1998). * (5) T. Okamoto, J. Simonen and S. Kawata, Phys. Rev. B 77, 115425 (2008). * (6) W. A. Murray and W. L. Barnes, Adv. Mat. 19, 3771 (2007). * (7) P. B. Catrysse, Nature Phys. 3, 839 (2007). * (8) A. M. Schwartzberg and J. Z. Zhang, J. Phys. Chem. C 112, 10323 (2008). * (9) E. Hutter and J. H. Fendler, Adv. Mat. 16, 1685 (2004). * (10) A. J Haes and R. P. Van Duyne, Expert Rev. Mol. Diagn. 4, 527 (2004). * (11) U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters (Springer, New York, 1995). * (12) G. Mie, Ann. Phys. 25, 377 (1908). * (13) J. I. Gersten and A. Nitzan, in Surface Enhanced Raman Scattering (ed. R. K. Chang and T. Furtak, Plenum, New York, 1982), p 89\. * (14) R. Jin, Y. W. Cao, C. A. Mirkin, K. L. Kelly, G. C. Schatz, J. G. Zheng, Science 294, 1901 (2001). * (15) H. Raether, Surface Plasmons on Smooth and Rough Surfaces and on Gratings (Springer, Berlin, 1988). * (16) A. Otto, Z. Phys. 216, 398 (1968). * (17) B. Fischer, N. Marschall, and H. J. Queisser, Surf. Sci. 34, 50 (1973). * (18) E. N. Economu, Phys. Rev. 182, 539 (1969). * (19) D. Sarid, Phys. Rev. Lett. 47, 1927 (1981). * (20) Y. J. Chen, E. S. Koteles and R. J. Seymore, Solid State Comm. 46, 95 (1983). * (21) W-q Mu, D. B. Buchholz, M. Sukharev, R. P. H. Chang, and J. B. Ketterson (submitted). * (22) T. Okamoto, F. H’Dhili and S. Kawata, App. Phys. Lett. 85, 3968 (2004). * (23) N. L. Dmitruk and A. V. Korovin, Opt. Lett. 33, 893 (2008). * (24) A. Hessel and A. A. Oliner, App. Opt. 4, 1275 (1965). * (25) A. Taflove and S. C. Hagness, Computational Electrodynamics: The Finite-Difference Time-Domain Method, 3rd ed. (Artech House, Boston, 2005). * (26) J.-P. Berenger, J. Comput. Phys. 114, 185 (1994). * (27) M. Sukharev, J. Sung, K. G. Spears, and T. Seideman, Phys. Rev. B 76, 184302 (2007). * (28) J. Steidtner and B. Pettinger, Phys. Rev. Lett. 100, 236101 (2008). * (29) G. G. Nenninger, P. Tobiska, J. Homola, S. S. Yee, Sensors and Actuators B 74, 145 (2001). Figure 1: (A) Energy reflection coefficient vs wavelength at critical coupling for film thicknesses of $5$ nm (solid curve), $10$ nm (dashed curve), and $15$ nm (dash-dotted curve). (B) Quality factor, $Q=\lambda/\Delta\lambda$, for the surface plasmon resonance vs silver film thickness derived from the linewidths shown in Fig. 1A. Figure 2: (A) Schematic setup of out-of-phase and in-phase sinusoidal gratings. (B) Unit cell of FDTD simulations. The vertical dashed lines depict the periodic boundaries. Two horizontal dashed lines represent the detection contours. The metal film shown in the center of the unit cell is excited either by a laser pulse generated along a single horizontal line placed a few spatial steps beneath the upper PML region (we refer to this excitation scheme as a single-ended excitation), or by two incident plane waves generated symmetrically on both sides of the film with a fixed phase delay between the incident pulses (referred to as double-ended excitation scheme). Figure 3: The electromagnetic Wood’s phenomenon in which an incoming wave (entering perpendicular) is diffracted tangent to the grating such that the wave fronts match the grating spacing resulting in anomalies in the reflected and transmitted waves. Figure 4: Sum of the transmission, $T$, and reflection, $R$, coefficients as a function of the incident wavelength. The solid curve illustrates the single- ended excitation scheme, whereas the dashed and dash-dotted curves show the double-ended excitation with phase shifts $\pi$ and $0$. Panel A corresponds to the out-phase sinusoidal grating with $D_{1}=5$ nm and $D_{2}=31$ nm. Panel B shows data for the in-phase sinusoidal grating with parameters $D_{1}=10$ nm and $D_{2}=14.4$ nm. The structural parameters for both gratings have been optimized so as to minimize $T+R$ for the double ended excitation scheme and the long-range mode. (Panel A - dashed curve, panel B - dash-dotted curve). Figure 5: (A) Surface averaged intensity enhancement as a function of the incident wavelength for out-of-phase (solid curve) and in-phase (dashed curve) gratings at the impedance matching conditions for the long-range plasmon mode. (B) Surface averaged intensity enhancement as a function of the film thickness at the impedance matching conditions for long-range plasmon modes for the out- of-phase (squares) and in-phase (circles) sinusoidal grating.	arxiv-papers	2008-08-14T04:58:28	2024-09-04T02:48:57.281840	{ "license": "Public Domain", "authors": "M. Sukharev, P. R. Sievert, T. Seideman, J. B. Ketterson", "submitter": "Maxim Sukharev", "url": "https://arxiv.org/abs/0808.1941" }
0808.1952	# Correlation of spin and velocity in granular gases W. T. Kranz Max Planck Institute for Dynamics and Self Organization, Bunsenstr. 10, 37073 Göttingen, Germany N. V. Brilliantov Department of Mathematics, University of Leicester, University Road, Leicester LE1 7RH UK T. Pöschel Universität Bayreuth, Physikalisches Institut, 95440 Bayreuth, Germany A. Zippelius Max Planck Institute for Dynamics and Self Organization, Bunsenstr. 10, 37073 Göttingen, Germany Institute of Theoretical Physics, University of Göttingen, Friedrich-Hund-Platz 1, 37077 Göttingen ###### Abstract In a granular gas of rough particles the spin of a grain is correlated with its linear velocity. We develop an analytical theory to account for these correlations and compare its predictions to numerical simulations, using Direct Simulation Monte Carlo as well as Molecular Dynamics. The system is shown to relax from an arbitrary initial state to a quasi-stationary state, which is characterized by time-independent, finite correlations of spin and linear velocity. The latter are analysed systematically for a wide range of system parameters, including the coefficients of tangential and normal restitution as well as the moment of inertia of the particles. For most parameter values the axis of rotation and the direction of linear momentum are perpendicular like in a sliced tennis ball, while parallel orientation, like in a rifled bullet, occurs only for a small range of parameters. The limit of smooth spheres is singular: any arbitrarily small roughness unavoidably causes significant translation-rotation correlations, whereas for perfectly smooth spheres the rotational degrees of freedom are completely decoupled from the dynamic evolution of the gas. ###### pacs: 45.70.-n, 45.70.Qj, 47.20.-k ## I Introduction Materials which are composed of macroscopic objects, i.e. granular media, attract increasing scientific interest due to their importance in nature and technology, e.g. Goldhirsch (2003); Levy and Kalman (2001). The latter may be exemplified by transport and storage of sand, cereals, granular chemicals, etc. the former–by avalanches, land slides, dust devils, etc. Spectacular celestial objects, like planetary rings or interstellar dust clouds, can serve as another example of natural granular systems Greenberg and Brahic (1984). The granular matter exists there in a gaseous state and exhibits many properties of a common molecular gas, e.g. Pöschel and Luding (2001); Pöschel and Brilliantov (2003); Goldhirsch (2003); Brilliantov and Pöschel (2004). The main (and very important) difference of a granular gas from a molecular gas is the dissipative nature of particle interactions, which are describe by macroscopic mechanics of solids rather than by a microscopic interaction potential. The consequences of the dissipative interactions are quite substantial: A spatially homogeneous state is unstable McNamara and Young (1992); Goldhirsch and Zanetti (1993); Brito and Ernst (1998), velocities are not distributed according to a Maxwell-Boltzmann distribution Goldshtein and Shapiro (1995); van Noije and Ernst (1998); Esipov and Pöschel (1997); Brey et al. (1999a); Deltour and Barrat (1997); Huthmann et al. (2000); Brilliantov and Pöschel (2000a); Goldhirsch et al. (2003); Pöschel et al. (2006) and the diffusion or self-diffusion is anomalous Brey et al. (1999b, 2000); Brilliantov and Pöschel (2000b); Santos and Dufty (2001); Garzo and Montanero (2004). These properties of a granular gas have been observed for the case of smooth particles, when grain collisions do not affect their rotational motion. This is, certainly, an oversimplified model, since real grains have a rough surface and exchange rotational and translational energy in collisions. Real granular particles experience frictional forces when colliding. Hence, a more adequate model takes into account the rotational motion of particles and the exchange of rotational and translational energy in collisions Goldhirsch (2003); Brilliantov and Pöschel (2004); Huthmann and Zippelius (1997); Goldshtein and Shapiro (1995); Aspelmeier et al. (2001); Goldhirsch et al. (2005a); Elperin and Golshtein (1997); Jenkins and Richman (1985); Lun and Savage (1987); Jaeger et al. (1990); Luding (1995); Jenkins and Louge (1997); Bardenhagen et al. (2000); Cafiero et al. (2002); Mitarai et al. (2002). Dissipative frictional gases exhibit additional unusual features which are not present in molecular gases. For instance, equipartition between rotational and translational motion does not hold Huthmann and Zippelius (1997) and the hydrodynamic description requires an additional field and a dynamic equation to account for its evolution Goldhirsch et al. (2005b, a). Moreover, the rotational and translational motion of particles are correlated as mentioned in a very implicit way in Appendix E of Goldhirsch et al. (2005a) and worked out in Brilliantov et al. (2007). In the present study we analyze the latter effect in detail. In Sec II we introduce a model of frictional particles and the observables of interest. Subsequently in Sec. III an approximate analytical theory is developed and in Sec. IV we briefly explain the simulation techniques. The main results are presented in Sec. V, where we compare predictions of the analytical theory with data from simulations. The emphasis lies on the correlations in the quasi-steady state, but we also briefly discuss the relaxation to the steady state. The technical details of the calculations are given in the Appendix. ## II Model and Observables We consider a granular gas consisting of $N$ inelastic hard spheres of radius $a$, mass $m$, and moment of inertia $I=qma^{2}$. Here the dimensionless variable q is determined by the mass distribution within the disc. The state of the system is fully described by the particles’ positions $\\{\bm{r}_{i}\\}$, velocities $\\{\bm{v}_{i}\\}$, and angular velocities $\\{\bm{\omega}_{i}\\}$ for $i=1,\ldots,N$. The particles move freely in between instantaneous collisions, whereupon their linear and angular velocities change according to the collision rule: the relative velocity at the point of contact of colliding particles is $\bm{g}\equiv\bm{v}_{1}-\bm{v}_{2}+a\hat{\bm{n}}\times(\bm{\omega}_{1}+\bm{\omega}_{2})\,,$ (1) with $\hat{\bm{n}}\equiv\hat{\bm{n}}_{12}\equiv\left(\bm{r}_{1}-\bm{r}_{2}\right)/\left\|\bm{r}_{1}-\bm{r}_{2}\right\|$. The post-collisional (primed) velocity is related to the pre-collisional one by $\displaystyle\bm{g}^{\prime}\cdot\hat{\bm{n}}$ $\displaystyle=-\varepsilon_{n}\,\bm{g}\cdot\hat{\bm{n}}$ (2) $\displaystyle\bm{g}^{\prime}\times\hat{\bm{n}}$ $\displaystyle=\varepsilon_{t}\,\bm{g}\times\hat{\bm{n}}\,.$ The coefficient of normal restitution is denoted by $\varepsilon_{n}$ with $0\leq\varepsilon_{n}\leq 1$. The value $\varepsilon_{n}=0$ implies no relative motion in the normal direction after the collision, whereas for $\varepsilon_{n}=1$ no dissipation of the normal component of the relative motion occurs. The coefficient of tangential restitution has two elastic limits, namely $\varepsilon_{t}=1$ corresponding to smooth spheres and $\varepsilon_{t}=-1$ corresponding to perfectly rough (reflecting) collisions without loss of energy for the tangential motion. For all other values energy is lost in the tangential component. In general, both coefficients of restitution, $\varepsilon_{n}$ and $\varepsilon_{t}$, depend on the impact velocity Brilliantov et al. (1996); Schwager and Pöschel (1998); Ramírez et al. (1999); Becker et al. (2008). Together with the conservation of linear and angular momentum the collision rule, Eq. (2), determines the post-collisional velocities in terms of the pre- collisional ones: $\displaystyle\bm{v}^{\prime}_{1}=\bm{v}_{1}-\bm{\delta},\quad\bm{\omega}^{\prime}_{1}=\bm{\omega}_{1}+\frac{1}{qa}(\hat{\bm{n}}\times\bm{\delta})$ (3) $\displaystyle\bm{v}^{\prime}_{2}=\bm{v}_{2}+\bm{\delta},\quad\bm{\omega}^{\prime}_{2}=\bm{\omega}_{2}+\frac{1}{qa}(\hat{\bm{n}}\times\bm{\delta})\,,$ where $m\bm{\delta}$ denotes the exchange of linear momentum with $\bm{\delta}\equiv\eta_{t}\bm{g}+(\eta_{n}-\eta_{t})(\hat{\bm{n}}\cdot\bm{g})\hat{\bm{n}}\,,$ (4) $\eta_{n}\equiv\frac{1+\varepsilon_{n}}{2}\,,\qquad\qquad\eta_{t}\equiv\frac{q}{2}\,\frac{1-\varepsilon_{t}}{1+q}\,.$ (5) In the present study we address only non-driven systems. Moreover, we focus on the homogeneous cooling state (HCS) of a gas, which is characterized by two time-dependent granular temperatures, one for the translational and one for the rotational motion, $T=\frac{m}{3N}\sum_{i=1}^{N}\bm{v}_{i}^{2}\quad\text{and}\quad R=\frac{I}{3N}\sum_{i=1}^{N}\bm{\omega}_{i}^{2}.$ (6) One generally observes that after a transient period the system reaches a quasi-stationary state where $r\equiv R(t)/T(t)=\text{const.}$, that is, both temperatures decay with the same rate. In general, $r\neq 1$ so that equipartition is violated. The value of $r$ depends on the collision parameters as well as on the moment of inertia Huthmann and Zippelius (1997); Luding et al. (1998). In this paper we focus on the correlation between the axis of rotation of a granular particle and the direction of its linear velocity, which may be quantified by the angle $\theta_{i}$ between the linear and rotational velocity, $\cos{\theta}_{i}=\frac{\bm{v}_{i}\cdot\bm{\omega}_{i}}{\|\bm{v}_{i}\|\|\bm{\omega}_{i}\|}.$ (7) All information on the angle is contained in the distribution $f(\cos{\theta})=\frac{1}{N}\sum_{i=1}^{N}\delta(\cos{\theta}-\cos{\theta}_{i}),$ (8) In a molecular gas all values of $\cos{\theta}$ occur with equal probability due to equipartition. In contrast for a granular gas we know that equipartition is violated and we expect to observe deviations from the equi- distribution. Because of symmetry, the average of $\cos{\theta_{i}}$ over all particles vanishes. Thus, a measure of correlations is the second moment, $\left\langle\cos^{2}\theta\right\rangle=\frac{1}{N}\sum_{i}\frac{(\bm{v}_{i}\cdot\bm{\omega}_{i})^{2}}{\bm{v}_{i}^{2}\bm{\omega}_{i}^{2}}\,.$ (9) If the angular and linear velocities are not correlated in their direction, $\left<\cos^{2}\theta\right>=1/3$. Hence, any deviation of $\left<\cos^{2}\theta\right>$ from $1/3$ indicates correlations. Moreover, if $\left<\cos^{2}\theta\right><1/3$ the angular and linear velocities are preferably perpendicular, like in a sliced tennis ball, while for $\left<\cos^{2}\theta\right>>1/3$ they are preferably aligned like in a rifled bullet. ## III Analytical theory The evolution of any observable $F(t)=F\left(\left\\{\bm{r}_{i}(t),\bm{v}_{i}(t),\bm{\omega}_{i}(t)\right\\}\right)$ (10) may be obtained by means of the pseudo-Liouville operator $\mathcal{L}_{+}$ via $\partial_{t}F(t)=i\mathcal{L}_{+}F(t)\quad{\mbox{for}}\quad t>0.$ (11) For hard spheres the pseudo-Liouville operator decomposes into two parts, $\mathcal{L}_{+}=\mathcal{L}_{0}+\mathcal{L}^{\prime}_{+}$, where $\mathcal{L}_{0}=\mathcal{L}_{0}^{\text{tr}}+\mathcal{L}_{0}^{\text{rot}}$ describes the free streaming of translational and rotational motion of particles. Here $\mathcal{L}_{0}^{\text{tr}}=\sum_{i}\bm{v}_{i}\cdot\nabla_{i}$ and a similar expression for $\mathcal{L}_{0}^{\text{rot}}$. The latter is not needed here, because we never specify the orientation of our particles , which are perfect spheres. The interaction part of the pseudo-Liouville operator reads, $\mathcal{L}^{\prime}_{+}=\sum_{i<j}\mathcal{T}_{ij}$, where the binary collision operator $\mathcal{T}_{ij}$ reads Huthmann and Zippelius (1997); van Noije and Ernst (1998) $i\mathcal{T}_{ij}=-\hat{\bm{n}}_{ij}\cdot\bm{v}_{ij}\Theta\left(-\hat{\bm{n}}_{ij}\cdot\bm{v}_{ij}\right)\delta\left(r_{ij}-2a\right)\left(\hat{b}_{ij}-1\right)\,.$ (12) The operator $\hat{b}_{ij}$ replaces unprimed by primed values according to the collision rule, Eq. (3). For example, $\hat{b}_{12}\bm{v}_{1}=\bm{v}_{1}^{\prime}\,,\qquad\hat{b}_{12}\bm{v}_{2}=\bm{v}_{2}^{\prime}\,,\qquad\hat{b}_{12}\bm{v}_{k}=\bm{v}_{k}\,,\quad k\neq 1,\,2$ (13) with $\bm{v}_{1}^{\prime}$ and $\bm{v}_{2}^{\prime}$ given by Eq. (3) and with similar relations for the rotational velocities. The ensemble average of a dynamic variable is defined by $\left\langle F\right\rangle_{t}=\int d\Gamma\rho(0)F(t)=\int d\Gamma\rho(t)F(0)$ (14) with $d\Gamma=\prod_{i}(d^{3}r_{i}d^{3}v_{i}d^{3}\omega_{i})$. Here $F(t)=\exp{(-i\mathcal{L}_{+}t)}F(0)$ and $\rho(t)=\exp{(-i\mathcal{L}_{+}^{\dagger}t)}\rho(0)$ denotes the $N$-particle distribution, whose evolution is governed by the adjoint $\mathcal{L}_{+}^{\dagger}$ of the evolution operator $\mathcal{L}_{+}$. Differentiating Eq. (14) one obtains $\displaystyle\frac{d}{dt}\left\langle F\right\rangle_{t}$ $\displaystyle=$ $\displaystyle\int d\Gamma\rho(0)\frac{d}{dt}F(t)=\int d\Gamma\rho(0)i\mathcal{L}_{+}F(t)$ (15) $\displaystyle=$ $\displaystyle\int d\Gamma\rho(0)\exp{(i\mathcal{L}_{+}t)}i\mathcal{L}_{+}F(0)$ $\displaystyle=$ $\displaystyle\int d\Gamma\rho(t)i\mathcal{L}_{+}F(0)=\left\langle i\mathcal{L}_{+}F\right\rangle_{t}\,.$ It is impossible to compute the time-dependent $N$-particle distribution exactly, so that we have to resort to approximations. A standard procedure in the analytical treatment of granular gases is to assume homogeneity and molecular chaos, e.g. Brilliantov and Pöschel (2004) (see also Pöschel et al. (2003)). Under these assumptions the $N$-particle velocity distribution function takes the form $\rho(t)=g_{N}(\bm{r}_{1},\ldots,\bm{r}_{N})\prod_{i}\rho_{1}(\bm{v}_{i},\bm{\omega}_{i},t)\,,$ (16) where the $N$-particle correlation function of a hard sphere system, $g_{N}(\bm{r}_{1},\ldots,\bm{r}_{N})$, is not affected by the particle roughness. For the HCS it may be approximated by the corresponding function of an equilibrium hard-sphere system (e.g. Brilliantov and Pöschel (2004)). For an isotropic system $\rho_{1}(\bm{v},\bm{\omega})$ depends in general on $v=\|\bm{v}\|,\omega=\|\bm{\omega}\|$ and the angle $\theta$ ($\cos\theta=\bm{v}\cdot\bm{\omega}/\left(\left\|\bm{v}\right\|\,\left\|\bm{\omega}\right\|\right)$. Here we are particularly interested in the dependence on $\cos\theta$ and expand $\rho_{1}$ in Legendre polynomials $P_{n}(\cos\theta)$ $\displaystyle\rho_{1}(\bm{v},\bm{\omega},t)\propto$ $\displaystyle\exp\left(-\frac{m\bm{v}^{2}}{2T(t)}\right)\exp\left(-\frac{I\bm{\omega}^{2}}{2R(t)}\right)$ (17) $\displaystyle\times$ $\displaystyle\sum_{n=0}^{\infty}b_{n}(t)\bm{v}^{n}\bm{\omega}^{n}P_{n}(\cos\theta)\,,$ where the $b_{n}(t)$ are time dependent expansion coefficients and the distribution function has to be normalized according to $\int d\Gamma\rho_{1}=1$. We use a simple Gaussian even though the distributions are non-Gaussian for strong dissipation and high densities. Deviations have been handled by an expansion in Sonine polynomials Noskowicz et al. (2007). Here we concentrate on the dependence on $\cos\theta$ and leave a more general ansatz with both, angular correlations and non-Gaussian distributions, to future work. To keep the calculations tractable, we limit the calculation to the lowest non-trivial order $\displaystyle\rho_{1}(\bm{v},\bm{\omega},t)\propto$ $\displaystyle\exp\left(-\frac{m\bm{v}^{2}}{2T(t)}\right)\exp\left(-\frac{I\bm{\omega}^{2}}{2R(t)}\right)$ (18) $\displaystyle\times$ $\displaystyle\left[1+b(t)\bm{v}^{2}\bm{\omega}^{2}P_{2}(\cos\theta)\right]\,,$ where $b(t)\equiv b_{2}(t)$ and $P_{2}(\cos\theta)=\nicefrac{{3}}{{2}}(\cos^{2}\theta-\nicefrac{{1}}{{3}})$. The terms for odd $n$ vanish by symmetry. The lowest order coefficient $b(t)$ is simply related to the quantity of interest $\left\langle\cos^{2}\theta\right\rangle_{t}$. Using $P_{0}(\cos\theta)=1$ and expressing $\left\langle\cos^{2}\theta\right\rangle_{t}$ in terms of Legendre polynomials we can write $\displaystyle\left\langle\cos^{2}\theta\right\rangle_{t}=\frac{1}{3}\int_{v}\int_{\omega}$ $\displaystyle[P_{0}(\cos\theta)+2P_{2}(\cos\theta)]$ (19) $\displaystyle\times$ $\displaystyle[P_{0}(\cos\theta)+b(t)v^{2}\omega^{2}P_{2}(\cos\theta)]\,$ where for brevity we introduce the shorthand notation $\int_{v}=\left(\frac{m}{2\pi T}\right)^{3/2}\int d^{3}v\,\exp\left(-\frac{mv^{2}}{2T}\right)$ (20) and similarly for $\int_{\omega}.$ The angular integration in the Eq. (19) may be performed using the orthogonality relation for Legendre polynomials, yielding $\left\langle\cos^{2}\theta\right\rangle_{t}=\frac{1}{3}+b(t)\frac{6T(t)R(t)}{5qm^{2}a^{2}}\,.$ (21) Hence, the correlations of interest manifest themselves through the coefficient $b(t)$–the larger the coefficient, the more pronounced are deviations from the value $\left\langle\cos^{2}\theta\right\rangle=1/3$ of the uncorrelated case. To summarize our analytical approach so far: The time dependent $N$-particle distribution has been parametrised by three time-dependent functions $T(t)$, $R(t)$ and $b(t)$, which have to be calculated self-consistently. This is achieved by applying the general equation (15) for the evolution of an observable to $T(t)$, $R(t)$ and $b(t)$ and using our ansatz for $\rho(t)$, see Eqs. (16,18). Even with all these simplifying assumptions, the analytical calculations are rather cumbersome and all the details of the calculation have been relegated to the Appendix. The results are three first order differential equations for $T(t),R(t)$ and $b(t)$. These simplify, if we measure times in units of the Enskog collision frequency $\omega_{E}=16(\pi T/m)^{1/2}na^{2}g_{2}(2a)$. In other words we rescale time according to $d\tau=\omega_{E}dt$ and obtain: $\displaystyle\frac{dT}{d\tau}=-AT(\tau)+B$ $\displaystyle\left[1-\frac{b(\tau)}{2}\frac{T(\tau)R(\tau)}{qm^{2}a^{2}}\right]R(\tau)$ (22) $\displaystyle\frac{dR}{d\tau}=BT(\tau)-C$ $\displaystyle\left[1-\frac{b(\tau)}{2}\frac{T(\tau)R(\tau)}{qm^{2}a^{2}}\right]R(\tau)$ where $\displaystyle A$ $\displaystyle\equiv\eta_{n}(1-\eta_{n})+\eta_{t}(1-\eta_{t}),$ (23) $\displaystyle B$ $\displaystyle\equiv\frac{\eta_{t}^{2}}{q}\,,\qquad C\equiv\frac{\eta_{t}}{q}\left(1-\frac{\eta_{t}}{q}\right).$ and $20\frac{db}{d\tau}=-b(\tau)\left[A^{(1)}+B^{(1)}\frac{R(\tau)}{T(\tau)}+\frac{40}{T(\tau)}\frac{dT}{d\tau}(\tau)\right.\\\ \left.+\frac{40}{R(\tau)}\frac{dR}{d\tau}(\tau)\right]\\\ -\frac{qm^{2}a^{2}}{T(\tau)R(\tau)}\left[A^{(0)}+B^{(0)}\frac{R(\tau)}{T(\tau)}+C^{(0)}\frac{T(\tau)}{R(\tau)}\right].$ (24) The constants are given by: $A^{(0)}\equiv\frac{16}{3}\frac{\eta_{t}^{3}}{q}\left(\frac{2\eta_{t}}{q}-1\right)-\frac{2}{3}\frac{\eta_{t}^{2}}{q}\left(\frac{8\eta_{t}}{q}-3\right)\\\ +\frac{1}{3}\frac{\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)+\frac{8}{3}\frac{\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)\eta_{n}(\eta_{n}-1)$ (25a) $\displaystyle B^{(0)}$ $\displaystyle\equiv\frac{1}{3}\frac{\eta_{t}^{2}}{q}\left[\frac{16\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)+5\right]$ (25b) $\displaystyle C^{(0)}$ $\displaystyle\equiv\frac{2}{3}\frac{\eta_{t}^{2}}{q}\left[8\eta_{t}(\eta_{t}-1)+4\eta_{n}(\eta_{n}-1)+3\right]$ (25c) $A^{(1)}\equiv-\frac{8\eta_{t}^{3}}{q}\left(\frac{2\eta_{t}}{q}-1\right)+\frac{1}{3}\frac{\eta_{t}^{2}}{q}\left(\frac{24\eta_{t}}{q}-37\right)\\\ -\frac{5}{6}\frac{\eta_{t}}{q}\left(\frac{9\eta_{t}}{q}-29\right)-\frac{4\eta_{t}\eta_{n}^{2}}{q}\left(\frac{\eta_{t}}{q}-1\right)\\\ +\frac{4}{3}\frac{\eta_{t}\eta_{n}}{q}\left(\frac{3\eta_{t}}{q}-14\right)-12\eta_{t}\eta_{n}\\\ +22(\eta_{t}+\eta_{n})-6(\eta_{t}^{2}+\eta_{n}^{2})$ (25d) $B^{(1)}\equiv-\frac{2}{3}\frac{\eta_{t}^{2}}{q}\left[\frac{8\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)+1\right]\,.$ (25e) Eqs. (22) and (24) constitute a set of self-consistent equations for the observables $T(t),R(t)$, and $b(t)$. Fig. 1 illustrates the dependence of the above coefficients on the coefficient of tangential restitution. Figure 1: (Color online) The coefficients in Eq. (25) as a function of $\varepsilon_{t}$ for $\varepsilon_{n}=0.9$ and $q=2/5$. Note that $B^{(1)}$ is slightly negative for $\varepsilon_{t}\gtrsim 0.7$. Except for the coefficient $A^{(0)}$ all coefficients vanish in the limit $\varepsilon_{t}\to 1$. ## IV Simulations We performed both Direct Simulation Monte Carlo (DSMC) Bird (1994) and event- driven Molecular Dynamics (MD) Pöschel and Schwager (2005) calculations to check the predictions of the analytical theory. DSMC determines the stationary distribution of the scaled velocities by numerically solving the kinetic Boltzmann equation which is based on the assumption of molecular chaos. Consequently, for its application it is assumed that the gas is uniform, thus, spatial correlations of the particles are neglected. If this precondition is given, DSMC yields very precise statistical results because of the large number of particles which can be simulated (here we use $N=2\times 10^{7}$ particles111To be precise, although the mathematical operations in DSMC looks like a particle simulation, the particles in the simulation do not correspond to real particles. They are better considered as quanta of probability Brey and Cubero (2001).). Molecular Dynamics calculates the trajectories of the particles using the collision rule, Eq. (2), therefore, MD allows to trace the evolution of the correlation. On the other hand, MD is restricted to much smaller systems as compared to DSMC. Although MD is free from the mentioned assumptions, DSMC is significantly more efficient for a homogeneous granular gas. Moreover, in the limit of low density both methods provide, in principle, identical results for the stationary state Pöschel and Schwager (2005). In practice, we use MD for $N=8000$ particles to study the transient process of the system’s relaxation to its steady-state and up to $N=10^{5}$ for steady state correlations. The volume fraction is $\frac{N}{V}\frac{4\pi a^{3}}{3}=0.0146$ or even smaller, such that the gas is always in the HCS. ## V Results Starting from a random distribution of velocities and angular velocities with mean $\left<\bm{v}\right>=\left<\omega\right>=0$, after some transient period the system relaxes to a steady state where the correlation of the spin and the translational velocity as well as the ratio of translational and rotational temperatures adopt stationary values. We quantify these correlations by means of the second moment $\left<\cos^{2}\theta\right>$, see Eq. (9) and analyze this quantity as a function of three parameters, $\varepsilon_{n}$, $\varepsilon_{t}$, and $q$ in Sec. V.1. The relaxation to the steady state is discussed in Sec. V.2 and in Sec. V.3 we consider correlations beyond the second moment and investigate the distribution of $\cos\theta$. ### V.1 Steady-state correlations To study the steady-state properties it is convenient to introduce an auxiliary variable $x(\tau)\equiv b(\tau)\frac{T(\tau)R(\tau)}{qm^{2}a^{2}}=\frac{5}{6}\left(\left<\cos^{2}\theta\right>_{\tau}-\frac{1}{3}\right)\,.$ (26) Using $x(\tau)$ and $r(\tau)=R(\tau)/T(\tau)$ we recast the set of three equations (22,24) for $b$, $R$ and $T$ into a set of two equations for $x$ and $r$. The result reads $\displaystyle\frac{dr}{d\tau}$ $\displaystyle=$ $\displaystyle B-C\left[1-\frac{x(\tau)}{2}\right]r(\tau)+Ar(\tau)-B\left[1-\frac{x(\tau)}{2}\right]r^{2}(\tau)$ (27) $\displaystyle 20\frac{dx}{d\tau}$ $\displaystyle=$ $\displaystyle-x(\tau)\left\\{A^{(1)}+B^{(1)}r(\tau)-20A-20C+20B[r(\tau)+r^{-1}(\tau)]+20x(\tau)[C-B\,r(\tau)]/2\right\\}$ $\displaystyle-$ $\displaystyle A^{(0)}-B^{(0)}r(\tau)-C^{(0)}r^{-1}(\tau)\,.$ Setting the left hand side of Eqs. (27) and (V.1) to zero one arrives at a set of coupled nonlinear equations for the stationary values $r_{\infty}\equiv r(\tau\to\infty)$ and $x_{\infty}\equiv x(\tau\to\infty)$. Instead of solving these equations directly, we resort to an iteration scheme: At the outset we calculate a first approximation of the temperature ratio $r^{(0)}_{\infty}$ neglecting correlations, that is, for $x=0$. Hence we assume that for moderate inelasticity and roughness the temperature ratio is not noticeably affected by the rotational-translational coupling. The result reads $r^{(0)}_{\infty}=\frac{A-C}{2B}+\sqrt{1+\frac{(A-C)^{2}}{4B^{2}}}\,.$ (29) Using this value for the stationary temperature ratio we then proceed to calculate an approximate value of $x_{\infty}$ $x^{(0)}_{\infty}=-\frac{A^{(0)}+B^{(0)}r^{(0)}_{\infty}+C^{(0)}/r^{(0)}_{\infty}}{A^{(1)}+B^{(1)}r^{(0)}_{\infty}-40C+40B/r^{(0)}_{\infty}}$ (30) where we use the fact, that $B\left[r^{(0)}_{\infty}+1/r^{(0)}_{\infty}\right]=A-C+2B/r^{(0)}_{\infty}$ (31) and neglect the terms quadratic in $x_{\infty}$ since they are presumably small. In principle, one could further iterate to get better approximations, but we find that the results are reasonably good already at this stage. For the more intuitive variable, $\cos^{2}\theta$, Eq. (30) implies $\left\langle\cos^{2}\theta\right\rangle_{\infty}\approx\frac{1}{3}-\frac{6}{5}\frac{A^{(0)}+B^{(0)}r^{(0)}_{\infty}+C^{(0)}/r^{(0)}_{\infty}}{A^{(1)}+B^{(1)}r^{(0)}_{\infty}+40B/r^{(0)}_{\infty}-40C}\,.$ (32) Fig. 2 shows the steady-state value of the correlation factor $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ as a function of $\varepsilon_{t}$ for different values of $\varepsilon_{n}$ in comparison with DSMC results. Figure 2: (Color online) Steady-state value of $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ as a function of the coefficient of tangential restitution, $\varepsilon_{t}$, for different $\varepsilon_{n}$. The predictions of the analytical theory, Eq. (32), are depicted by lines and points indicate the simulation data by DSMC. The line of vanishing correlations, $\left\langle\cos^{2}\theta\right\rangle=1/3$ is shown, as well as the isolated point $\varepsilon_{t}=1$, which refers to the system of perfectly smooth hard spheres. Note the existence of non-vanishing correlations even in the limit of smooth spheres, $\varepsilon_{t}\to 1$ (see Eq. 33). Obviously, theory as well as simulations show that both types of correlations may occur, $\left\langle\cos^{2}\theta\right\rangle<1/3$, as for a sliced tennis ball or $\left\langle\cos^{2}\theta\right\rangle>1/3$ as for a rifled bullet The dependence of the correlations on $\varepsilon_{t}$ is nonmonotonic with the strongest correlations for $\varepsilon_{t}\sim 0$ and $\varepsilon_{t}\to 1$. Even though the dependence on $\varepsilon_{n}$ is also not strictly monotonic, the dominant tendency is an increase of correlations with decreasing $\varepsilon_{n}$, i.e. increasing inelasticity. The agreement between theory and computer experiment is excellent for small inelasticity. Moreover, even for significant dissipation the theory is able to reproduce qualitatively the simulation results. Decreasing the moment of inertia, $q$, turns the magnitude of the correlations more sensitive to changes in the coefficients of tangential restitution, as one can see from Fig. 3. Interestingly, varying the moment of inertia can even alter the type of the correlations: For instance, for $q=1/5$ there exists a region for $\varepsilon_{t}>0$, where the rotation axis is preferably directed along the linear velocity, $\left\langle\cos^{2}\theta\right\rangle_{\infty}>1/3$, while for $q=2/3$ there is no such region. Figure 3: Steady-state value of $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ for $\varepsilon_{n}=0.9$ as a function of $\varepsilon_{t}$ and for different moments of inertia of a grain (see also Fig. 8). With the decreasing moment of inertia the correlations become more sensitive to variations of the coefficient of tangential restitution Fig. 4 (upper panel) illustrates the analytical result, Eq. (32) for the whole range of parameters $\varepsilon_{t}$ and $\varepsilon_{n}$. Figure 4: Stationary value $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ (color coded) as a function of normal ($\varepsilon_{n}$) and tangential ($\varepsilon_{t}$) coefficients of restitution. The stationary value of the temperature ratio $r$ is superimposed through the dashed contour lines. The solid lines indicate vanishing correlations ($\left\langle\cos\theta\right\rangle_{\infty}=1/3$). The moment of inertia $q=2/5$ (upper panel) corresponds to homogeneous spheres. The middle and bottom panel show the same data for $q=1/5$ and $q=2/3$, respectively Note that for the majority of values of the coefficients, $\left\langle\cos^{2}\theta\right\rangle_{\infty}<\nicefrac{{1}}{{3}}$, that is, in most cases the axes tend to be perpendicular to each other. Only in two small regions of the parameter space the axes are preferably parallel. The correlations vanish only for combinations of $\varepsilon_{n}$ and $\varepsilon_{t}$ indicated by full lines. Dashed lines show curves of constant $r$. Strong correlations appear for large deviations from equipartition. This is shown more clearly in the middle and bottom panels of Fig. 4 which demonstrate the rather strong influence of the moment of inertia $I$ on the correlation factor $\left\langle\cos^{2}\theta\right\rangle_{\infty}$. To check the assumption that strong correlations occur for strong deviations from equipartition, we plot in Fig. 5 the correlation factor $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ as a function of $r_{\infty}$ and $\varepsilon_{t}$. Technically this may be done, using $\varepsilon_{n}=\varepsilon_{n}(r_{\infty})$–the inverse function of $r_{\infty}=r_{\infty}(\varepsilon_{n})$, given by Eq. (29), for each fixed $\varepsilon_{t}$. Note that pronounced correlations are present mainly for strong dissipation and large temperature ratios. Also note the small range of admissible temperature ratios for very rough spheres. Figure 5: Stationary value of $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ (color coded) as a function of the temperature ratio $r_{\infty}$ and the coefficient of tangential restitution $\varepsilon_{t}$. As previously, the solid lines indicate vanishing correlations and the dashed lines follow constant values of the coefficient of normal restitution $\varepsilon_{n}$. Note the logarithmic scale for the $r$-axis. The ragged border is an artifact of the limited numerical resolution Analyzing Eq. (32) in the limit of vanishing roughness, $K^{(0)}\equiv\lim_{\varepsilon_{t}\to 1}\left\langle\cos^{2}\theta\right\rangle_{\infty}-\frac{1}{3}=-\frac{3}{8}\,\frac{1-\varepsilon_{n}}{7-\varepsilon_{n}}$ (33) we see that even the smallest roughness induces finite correlations, for any given (fixed) value of the coefficient of normal restitution, $\varepsilon_{n}\neq 1$. For $\varepsilon_{t}=1$, that is, for perfectly smooth spheres, the initial rotational velocity of the particles is preserved. Therefore, the initial rotational energy is preserved as well and $r$ does not reach a steady state. On the other hand $\left\langle\cos^{2}\theta\right\rangle$ relaxes to the stationary value $\nicefrac{{1}}{{3}}$ once the correlations in the initial values of the translational velocities are lost due to collisions. Hence a straightforward expansion around $\varepsilon_{t}=1$ is problematic, or at least should be done with much care, as long as there is a finite inelasticity $\varepsilon_{n}\neq 1$. [See also the discussion of relaxation times in the following paragraph.] ### V.2 Relaxation to the steady-state So far we have discussed the quasi-stationary state, which is characterized by constant $r$ and $\left\langle\cos^{2}\theta\right\rangle$. It is also of interest to understand, how this stationary state is reached—starting from arbitrary initial conditions. Of particular interest is the limit of almost smooth spheres $\eta_{t}\propto\varepsilon_{t}-1\ll 1$ [see the definition, Eq. (5)]. While the decay of the rotational temperature $R$ and the translational temperature $T$ takes place extremely slowly, that is, with a rate $\sim\eta_{t}\ll 1$ [see Eq. (22) with $r=R/T\simeq A/B$ in this limit], the relaxation of the temperature ratio, $r=R/T$ as well as of the correlation factor $x\sim(\left\langle\cos^{2}\theta\right\rangle-1/3)$ occurs on the collision time scale. Indeed, in this limit one can write using Eqs. (27), (V.1) and the definitions of the coefficients (23) (25), $dr/d\tau\simeq-\eta_{n}(1-\eta_{n})\left(r/r_{\infty}^{(0)}\right)\left(r-r_{\infty}^{(0)}\right)\,,$ with $r_{\infty}^{(0)}\simeq A/B\sim 1/\eta_{t}^{2}\gg 1$ from Eq. (29). This implies that $r$ relaxes to its stationary value exponentially fast with a rate $\eta_{n}(1-\eta_{n})={\cal O}(1)$ (that is, on the collision time scale), while both temperatures $T$ and $R$ continue to decay with the same small rate. To analyse the relaxation of $x(\tau)$ to its steady state value, we use Eq. (V.1) and approximate $r(\tau)$ by its steady state value $r_{\infty}$: $\frac{dx}{d\tau}=-a_{0}-a_{1}x-a_{2}x^{2}$ (34) where $\displaystyle a_{0}$ $\displaystyle=$ $\displaystyle\frac{1}{20}\left[A^{(0)}+B^{(0)}r_{\infty}+C^{(0)}r^{-1}_{\infty}\right]$ (35) $\displaystyle a_{1}$ $\displaystyle=$ $\displaystyle\left[\frac{1}{20}A^{(1)}+\frac{1}{20}B^{(1)}r_{\infty}-A-C+B(r_{\infty}+r^{-1}_{\infty})\right]$ $\displaystyle a_{2}$ $\displaystyle=$ $\displaystyle\frac{1}{2}(C-Br_{\infty})\,.$ The above equation with the initial condition $x(0)=0$ is solved by $x(\tau)-x_{\infty}=-\frac{x_{\infty}}{1-\tanh\phi}\left[1-\tanh\left(\frac{\tau}{\tau_{\text{rel}}}+\phi\right)\right]\,,$ (36) with the relaxation time $\tau_{\text{rel}}=\frac{1}{2}\sqrt{a_{1}^{2}-4a_{0}a_{2}}$ (37) and $\tanh\phi=a_{1}/\sqrt{a_{1}^{2}-4a_{0}a_{2}}$. Evaluating the coefficients for typical values of $\varepsilon_{t}$ and $\varepsilon_{n}$, we find that the relaxation of the correlation factor $\left\langle\cos^{2}\theta\right\rangle_{t}$ to its steady-state also occurs within a few collisions per particle. This is illustrated in Fig. 6, where we plot the relaxation time $\tau_{\rm rel}$ given by Eq. (37). We wish to stress here again, that the relaxation on the collisional time scale to the steady state values applies only to the temperature ratio and the mean square cosine of the angle between linear and angular velocity. For nearly smooth particles, $\varepsilon_{t}\to 1$, the relaxation of the rotational and translational temperatures is, nevertheless, a very slow process, which proceeds with a small rate, tending to zero as $\varepsilon_{t}\to 1$. Figure 6: Relaxation time $\tau_{\text{rel}}$ (in the collision units) of $\left\langle\cos^{2}\theta\right\rangle_{\tau}$ when it approaches the steady-state value $\left\langle\cos^{2}\theta\right\rangle_{\infty}$. Note the narrow range of possible values for $\tau_{\rm rel}$ To demonstrate the existence of several time regimes we discuss in the following an instructive example. We initialize the particles with $\bm{\omega}=0$ corresponding to $r=0$. The collision parameters are $\varepsilon_{n}=\varepsilon_{t}=0.8$ so that the asymptotic value of the ratio of temperatures is $r_{\infty}>1$. We expect $r$ to monotonically increase as a function of time—and this is indeed observed as shown in Fig. 7. Figure 7: (Color online) Relaxation of $\left\langle\cos^{2}\theta\right\rangle_{t}$ and of the ratio of temperatures $r(t)=R(t)/T(t)$ to the steady state. Dots: molecular dynamics data for $8000$ particles, lines: analytical theory. To show that vanishing correlations $\left\langle\cos^{2}\theta\right\rangle_{t}$ coincide with equipartition, we have chosen the vertical axes, such that the point $r=1$ on the right axis (blue) and the point $\left\langle\cos^{2}\theta\right\rangle_{t}=1/3$ on the left axis (red) have the same vertical height as indicated by a horizontal line. Now, we can check our hypothesis that correlations are small for values of $r$ close to equipartition. If the hypothesis is correct, we should observe non- monotonic behavior of $\left\langle\cos^{2}\theta\right\rangle_{t}$. For short times the correlations should be large and of tennis ball type, because grazing collisions are the most effective for spinless particles to gain angular momentum. At intermediate times, when $r\sim 1$, the correlations should be very small or vanishing. In the asymptotic state with $r_{\infty}>1$, one should again observe finite correlations. These three time regimes are clearly born out in the time dependent correlations, shown in Fig. 7: (a) In the short time regime ($0<t<10^{3}$) correlations are strong and $0<r<1$. (b) At intermediate times ($10^{3}<t<10^{5}$) equipartition holds approximately $r\approx 1$ and correlations are small or vanishing. (c) The steady state ($t>10^{6}$) is characterized by $r\gg 1$ and finite $\left\langle\cos^{2}\theta\right\rangle_{\infty}<1/3$. The agreement between analytical theory and molecular dynamics is good also for the time-dependent quantities. Figure 8: Impact of the grains’ moments of inertia on the evolution and steady-state of $\left\langle\cos^{2}\theta\right\rangle_{t}$. The system parameters are $\varepsilon_{n}=0.9$, $\varepsilon_{t}=0.9$, $r(0)=0.001$ and vanishing initial correlations. The values of $q$ ($I=qma^{2}$) represent spheres with the mass concentrated towards the center ($q=1/5$), the homogeneous spheres ($q=2/5$) and spheres with the mass concentrated mainly in the outer shell ($q=2/3$). Fig. 8 demonstrates that the moment of inertia of the particles does not change the evolution of $\left\langle\cos^{2}\theta\right\rangle_{t}$ qualitatively. For the particular choice of the coefficients of restitution the correlations are more pronounced for larger $q=I/ma^{2}$ and fade with decreasing $q$. This however is not a general rule; depending on the coefficients $\varepsilon_{n}$ and $\varepsilon_{t}$, this tendency may reverse. Figure 9: The ratio of the temperatures calculated with $b(t)$ according to Eq. (24) to those with $b(t)\equiv 0$. The coefficients of restitution are $\varepsilon_{n}=\varepsilon_{t}=0.8$, and the initial ratio of rotational to translational temperatures was set to the steady-state value $r(0)=r_{\infty}$. The inset shows $r(t)/r(0)-1$ as a function of time. Note that the deviation of $r(t)$ from $r_{\infty}$ is always very small. The correlations between translational and rotational motion also have a noticeable, albeit small impact on the basic characteristics of granular gases—the translational and rotational temperatures. In Fig. 9 we present the time dependence of $R(t)/R^{(0)}(t)$—the ratio of the rotational temperature $R(t)$ with correlations to the corresponding value $R^{(0)}(t)$ without correlations. The respective ratio $T(t)/T^{(0)}(t)$ for the translational temperature is also plotted. Here we choose the case of large $r\simeq 24$ ($\varepsilon_{n}=\varepsilon_{t}=0.8$), which correspond to $\left\langle\cos^{2}\theta\right\rangle_{\infty}<1/3$, that is, for preferably perpendicular rotational and translational velocity. Fig. 9 demonstrates that the effect of the correlations on the granular temperatures $R(t)$ and $T(t)$ is indeed small. The corresponding quantity $r(t)=R(t)/T(t)$ is also not sensitive to these correlations. Moreover $r(t)$ does not deviate noticeably from its steady-state value throughout the system’s evolution, that is, $\|r(\tau)/r_{\infty}-1\|\ll 1$, as shown in the inset of Fig. 9. ### V.3 Beyond the second moment A complete one-particle picture includes the distribution ${\cal W}(\cos{\theta},v,\omega)=\frac{1}{N}\sum_{i=1}^{N}\delta(\cos{\theta}-\cos{\theta}_{i})\delta(v-v_{i})\delta(\omega-\omega_{i}).$ (38) Since correlations are developed in collisions, one intuitively expects that particles with larger velocities, that suffer stronger collisions, would show more pronounced orientational correlations; we study these effects by binning the particles velocities. So far we discussed the correlation factor $\left\langle\cos^{2}\theta\right\rangle_{t}$, which is a second moment of the distribution function ${\cal W}(\cos{\theta},v,\omega)$. Let us now analyze the distribution function itself. Due to the limited statistics of our numerical data we discriminate only between two classes of particles: the class of fast particles comprising those particles whose linear velocity belongs to the set of the 1/3 largest values and whose angular velocity belongs to the set of the 1/3 largest values. The class of slow particles is defined correspondingly as the set of particles whose linear velocity belongs to the set of the 1/3 smallest values and the angular velocity belongs to the set of the 1/3 smallest values. In Fig. 10 we show the distributions $f(\|\cos\theta\|)$ for the two classes in comparison with the distribution for all particles using both methods, MD and DSMC. In both cases we skipped the first 20 collisions per particle such that the ratio of temperatures, $r$, has reached its stationary value. For the MD simulation we used a system of $N=10^{5}$ particles at low density (filling factor $<1\%$. Then we averaged over 200 snapshots in distance of 1 collision per particle. In case of DSMC we used a system of $N=10^{7}$ particles and made the statistics based on a single snapshot. Both results agree very well. Figure 10: The angular distribution $f(\|\cos\theta\|)$ for the system of rough spheres with $\varepsilon_{t}=0.9$, $\varepsilon_{n}=0.9$ and $q=2/5$ in the stationary state. Note that while there is no preferable angle between $\bm{v}$ and $\bm{\omega}$ for slow particles, correlations are clearly visible for fast particles favouring perpendicular linear and angular velocities. The angular distribution is almost flat for slow particles and cannot be distinguished from the distribution of all particles (within statistical accuracy). On the other hand the fast particles exhibit a nonuniform distribution with a maximum around $\cos\theta=0$. Physically this means that the angle $\theta$ between $\bm{v}$ and $\bm{\omega}$ for slow particles is uniformly distributed within the interval $(0,\,\,\pi)$, while for fast particles it lies preferentially around $\theta=\pi/2$. In other words, for the particular choice of $\varepsilon_{t}=0.9$ and $\varepsilon_{n}=0.9$ the fast particles tend to behave like sliced tennis balls, with $\bm{\omega}$ perpendicular to $\bm{v}$. ## VI Conclusions and Outlook We have analysed in detail the correlations between rotational and translational motion in a granular gas of frictional particles. Under the assumption of molecular chaos and homogeneity we have developed an analytical theory which accounts for the correlations $\left\langle\cos^{2}\theta\right\rangle_{t}$ in addition to the rotational $R(t)$ and translational $T(t)$ temperature. We have also performed large scale DSMC simulations as well as event driven simulations to study the evolution of a gas of rough spheres and in particular the above correlations. We observe that the gas of rough particles always relaxes to a steady-state with constant correlation $\left\langle\cos^{2}\theta\right\rangle_{\infty}$ and constant ratio $r_{\infty}=R(t)/T(t)$. While the relaxation of $\left\langle\cos^{2}\theta\right\rangle$ and $r$ to their steady-state values happens on the collisional time scale, the evolution of the rotational and translational temperature in the near-smooth limit $\varepsilon_{t}\to 1$ is a slow process with a vanishingly small rate $\sim\eta_{t}\sim(1-\varepsilon_{t})\ll 1$. Physically, this may be explained as follows. In the near-smooth limit the coupling of the rotational modes to the translational ones becomes very weak. The energy of the rotational motion of the particles is almost conserved in collisions and the exchange of energy between the translational and rotational degrees of freedom becomes very slow. Consequently the rotational temperature as well as the translational temperature have a slowly decaying component, governed by this weak exchange of energies. However both temperatures decay with the same slow timescale so that their ratio, $r$, is stationary - after it has reached its steady state on the fast time scale of a few collisions. Simultaneously, $\left\langle\cos^{2}\theta\right\rangle_{t}$ relaxes to its steady state with a similar rate of the order of a few collisions. We conclude that the relaxation of the temperature ratio, $r$, and the angular correlations is rapid, – independent of the strength of the coupling $(1-\varepsilon_{t})$ as long as it is finite. Furthermore the correlations persist up to a vanishingly small roughness and are absent only for perfectly smooth particles, $\varepsilon_{t}=1$, which makes expansions around the smooth limit questionable. Our main results concern the correlation between the directions of rotational and translational velocity in the stationary state: The correlations depend sensitively on the values of the coefficients of restitution and the moment of inertia; for most of the system parameters $\left\langle\cos^{2}\theta\right\rangle<1/3$, implying that linear and angular velocities are preferably orthogonal, like in a sliced tennis ball. Only for a small part of the parameter space $\left\langle\cos^{2}\theta\right\rangle>1/3$, which means that $\bm{v}$ and $\bm{\omega}$ are preferably parallel like in a rifled bullet; the manifold of vanishing correlations (in $\varepsilon_{n},\varepsilon_{t}$ space) has seemingly zero measure. The correlations are more pronounced for strong deviations from equipartition. Our approach can be extended in several directions. In the simulations it is straightforward to use more advanced models for the coefficients of restitution as functions of the impact velocity, e.g. Brilliantov et al. (1996); Schwager and Pöschel (1998); Ramírez et al. (1999); Becker et al. (2008). It would also be of interest to study the full one-particle distribution. Our results already indicate that more energetic particles have stronger correlations, but a systematic study has yet to be done. Furthermore, one expects to observe correlations not only in very dilute gases, but also in rapidly moving denser systems. Our approximate analytical theory is based on the assumption of homogeneity and the density only enters into the Enskog collision frequency, which sets the time scale. Hence our results for the stationary state are independent of the density. This cannot hold true in a rapidly moving dense system, yet we expect to observe correlations as well. These could be analysed in a molecular dynamics simulation either for a driven Gayen and Alam (2008) or undriven system. Finally, the observed correlations may have important consequences for the stability theory of dilute granular flows: they possibly alter the domain of stability of granular system with respect to shear fluctuations—the main instability of granular flows of smooth particles. #### Acknowledgement We thank Isaac Goldhirsch for interesting discussions; TK and AZ thank Timo Aspelmeier for help with the MD simulations; TP acknowledges support by a grant from G.I.F., the German-Israeli Foundation for Scientific Research and Development. * ## Appendix A Analytical Calculations ### A.1 Correlation factor We present the details of the analytical calculations, leading to the three self-consistent equations (22) and (24) for $T(t),R(t)$ and $b(t)$. First, we note that he computation of $b(t)$ or $\left\langle\cos^{2}\theta\right\rangle_{t}$ is severely hampered by the denominator in Eq. (9). Fortunately one can carry out the calculations with the auxiliary observable $\left\langle\Delta\right\rangle_{t}\equiv\frac{2}{3N}\sum_{i=1}^{N}\bm{v}_{i}^{2}\bm{\omega}_{i}^{2}P_{2}(\cos\theta_{i})\,.$ (39) Its relation to our set of observables can be established by essentially the same steps as leading from Eq. (19) to Eq. (21): $30b(t)\frac{T(t)R(t)}{qm^{2}a^{2}}=\left\langle\Delta\right\rangle_{t}\frac{qm^{2}a^{2}}{T(t)R(t)}\,.$ (40) In the case of vanishing correlations we have $\left\langle\Delta\right\rangle_{t}=0$. Positive (negative) values correspond to a preference of a parallel (perpendicular) orientation. Owing to the assumptions of spatial homogeneity and molecular chaos it suffices to consider the phase space of only a single pair of particles (without loss of generality these shall be labeled $1$ and $2$). Integrating out the spatial degrees of freedom and using the definition of the pair correlation function $N(N-1)\int d\bm{r}_{3}\ldots d\bm{r}_{N}g_{N}(\bm{r}_{1},\ldots\bm{r}_{N})=n^{2}g_{2}(r_{12})\,,$ (41) with $n$ being the number density of the gas (e.g. Brilliantov and Pöschel (2004)) we obtain $\displaystyle\left\langle i\mathcal{L}_{+}v\Delta\right\rangle_{t}=$ $\displaystyle\nu\,N\,\int\limits_{v_{1}}\int\limits_{v_{2}}\int\limits_{\omega_{1}}\int\limits_{\omega_{2}}\left(\hat{\bm{n}}\cdot\bm{v}_{12}\right)\Theta\left(-\hat{\bm{n}}\cdot\bm{v}_{12}\right)$ (42) $\displaystyle\times\left[1+b(t)\bm{v}_{1}^{2}\bm{\omega}_{1}^{2}P_{2}\left(\cos\theta_{1}\right)\right]$ $\displaystyle\times\left[1+b(t)\bm{v}_{2}^{2}\bm{\omega}_{2}^{2}P_{2}\left(\cos\theta_{2}\right)\right]$ $\displaystyle\times\left(\hat{b}_{12}-1\right)\Delta\,,$ where $\hat{\bm{n}}$ is an arbitrary but fixed unit vector, $\nu=-8\pi na^{2}g_{2}(2a)$ and we used the shorthand notations $\begin{split}\int_{v_{i}}&\equiv\left(\frac{m}{2\pi T}\right)^{3/2}\int d^{3}v_{i}\exp\left(-\frac{m\bm{v}_{i}^{2}}{2T}\right)\\\ \int_{\omega_{i}}&\equiv\left(\frac{I}{2\pi R}\right)^{3/2}\int d^{3}\omega_{i}\exp\left(-\frac{I\bm{\omega}_{i}^{2}}{2R}\right)\,.\end{split}$ (43) In the following we will drop the $b^{2}(t)$-term stemming from the product of the two one particle distribution functions $\rho_{1}$ since it was assumed to be small and we only want to go to first order in $b(t)$. The calculation of $\left(\hat{b}_{12}-1\right)\Delta$ is obviously rather involved and, thus, it needs to be broken up to stay tractable. It is convenient to introduce relative integration variables $\begin{split}\bm{v}\equiv\bm{v}_{12}/\sqrt{2}~{}~{}~{}~{}~{}&\bm{V}\equiv(\bm{v}_{1}+\bm{v}_{2})/\sqrt{2}\\\ \bm{\omega}\equiv\bm{\omega}_{12}/\sqrt{2}~{}~{}~{}~{}~{}&\bm{\Omega}\equiv(\bm{\omega}_{1}+\bm{\omega}_{2})/\sqrt{2}\,.\end{split}$ (44) The term $\left\langle i\mathcal{L}_{+}\Delta\right\rangle_{t}$ can be broken up along two different principles. First, one can make the dependence on $b(t)$ explicit, that is, $\left\langle i\mathcal{L}_{+}\Delta\right\rangle_{t}=\left\langle\left(\hat{b}_{12}-1\right)\Delta\right\rangle^{(0)}\\\ +b(t)\left\langle\left(\hat{b}_{12}-1\right)\Delta\right\rangle^{(1)}+\mathcal{O}\left(b^{2}\right)\,,$ (45) where for any function $F$ $\left\langle F\right\rangle^{(0)}=\nu\int\limits_{v_{1}}\int\limits_{v_{2}}\int\limits_{\omega_{1}}\int\limits_{\omega_{2}}\left(\hat{\bm{n}}\cdot\bm{v}_{12}\right)\Theta\left(-\hat{\bm{n}}\cdot\bm{v}_{12}\right)\,F$ (46) and $\left\langle F\right\rangle^{(1)}=\nu\int\limits_{v_{1}}\int\limits_{v_{2}}\int\limits_{\omega_{1}}\int\limits_{\omega_{2}}\left(\hat{\bm{n}}\cdot\bm{v}_{12}\right)\Theta\left(-\hat{\bm{n}}\cdot\bm{v}_{12}\right)\\\\[2.84544pt] \times\left[\bm{v}_{1}^{2}\bm{\omega}_{1}^{2}P_{2}\left(\cos\theta_{1}\right)+\bm{v}_{2}^{2}\bm{\omega}_{2}^{2}P_{2}\left(\cos\theta_{2}\right)\right]\,F\,.$ (47) In order to be able to exploit some further symmetries it is advisable to split up the last average again, $\left\langle F\right\rangle^{(1)}=\left\langle F\right\rangle^{\text{even}}+\left\langle F\right\rangle^{\text{odd}}$ (48) where $\left\langle F\right\rangle^{\text{even}}=\frac{3\sqrt{2}}{4}\nu\int\limits_{v}\int\limits_{V}\int\limits_{\omega}\int\limits_{\Omega}\left(\hat{\bm{n}}\cdot\bm{v}\right)\Theta\left(-\hat{\bm{n}}\cdot\bm{v}\right)\\\ \times\left[\left(\bm{V}\cdot\bm{\Omega}\right)^{2}+\left(\bm{V}\cdot\bm{\omega}\right)^{2}\right.\\\ +\left(\bm{v}\cdot\bm{\Omega}\right)^{2}+\left(\bm{v}\cdot\bm{\omega}\right)^{2}\\\ -\left.\frac{1}{3}\left(\bm{V}^{2}+\bm{v}^{2}\right)\left(\bm{\Omega}^{2}+\bm{\omega}^{2}\right)\right]\,F$ (49) involves only even powers of $\bm{V},\bm{\omega},\bm{\Omega}$ and $\left\langle F\right\rangle^{\text{odd}}=\frac{3\sqrt{2}}{2}\nu\int\limits_{v}\int\limits_{V}\int\limits_{\omega}\int\limits_{\Omega}\left(\hat{\bm{n}}\cdot\bm{v}\right)\Theta\left(-\hat{\bm{n}}\cdot\bm{v}\right)\\\ \times\left[\left(\bm{V}\cdot\bm{\Omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)+\left(\bm{V}\cdot\bm{\omega}\right)\left(\bm{v}\cdot\bm{\Omega}\right)\right.\\\ -\left.\frac{2}{3}\left(\bm{V}\cdot\bm{v}\right)\left(\bm{\Omega}\cdot\bm{\omega}\right)\right]F$ (50) in contrast involves only the odd powers of these quantities. Independently we can write $\Delta=\Delta_{A}-\Delta_{B}/3$ (51) where $\Delta_{A}\equiv\sum_{i}\left(\bm{v}_{i}\cdot\bm{\omega}_{i}\right)^{2}~{}~{}~{}~{}\text{and}~{}~{}~{}~{}\Delta_{B}\equiv\sum_{i}\bm{v}_{i}^{2}\bm{\omega}_{i}^{2}\,.$ (52) First we address the $\Delta_{A}$-part. Applying the collision rule to $\Delta_{A}$ yields $\left(\hat{b}_{12}-1\right)\Delta_{A}=\left(\bm{\delta}\cdot\bm{\omega}\right)^{2}+\left(\bm{\delta}\cdot\bm{\Omega}\right)^{2}+\frac{1}{q^{2}a^{2}}\left[\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{v}\right]^{2}+\frac{1}{q^{2}a^{2}}\left[\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{V}\right]^{2}-\sqrt{2}\left(\bm{\delta}\cdot\bm{\Omega}\right)\left(\bm{v}\cdot\bm{\Omega}\right)\\\ -\sqrt{2}\left(\bm{\delta}\cdot\bm{\Omega}\right)\left(\bm{V}\cdot\bm{\omega}\right)-\sqrt{2}\left(\bm{\delta}\cdot\bm{\omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)-\sqrt{2}\left(\bm{\delta}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\Omega}\right)+\frac{\sqrt{2}}{qa}\left(\bm{v}\cdot\bm{\Omega}\right)\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{v}\\\ +\frac{\sqrt{2}}{qa}\left(\bm{V}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{v}+\frac{\sqrt{2}}{qa}\left(\bm{v}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{V}+\frac{\sqrt{2}}{qa}\left(\bm{V}\cdot\bm{\Omega}\right)\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{V}\\\ -\frac{2}{qa}\left(\bm{\delta}\cdot\bm{\Omega}\right)\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{v}-\frac{2}{qa}\left(\bm{\delta}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\times\bm{\delta}\right)\cdot\bm{V}$ (53) and invoking the definition of $\bm{\delta}$, Eq. (4), we obtain $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{A}\right\rangle^{(0)}=2\left(2\eta_{t}^{2}-2\eta_{t}+\frac{\eta_{t}^{2}}{q^{2}}-\frac{\eta_{t}}{q}+\frac{2\eta_{t}^{2}}{q}\right)\left\langle\left(\bm{v}\cdot\bm{\omega}\right)^{2}\right\rangle^{(0)}+2\frac{\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)\left\langle\left(\bm{\omega}\cdot\bm{V}\right)^{2}\right\rangle^{(0)}\\\ +2\left[2\left(\eta_{n}-\eta_{t}\right)^{2}+\frac{\eta_{t}^{2}}{q^{2}}-\frac{2\eta_{t}}{q}\left(\eta_{n}-\eta_{t}\right)\right]\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\cdot\bm{\omega}\right)^{2}\right\rangle^{(0)}\\\ +2\frac{\eta_{t}^{2}}{q^{2}}\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)^{2}\left(\hat{\bm{n}}\cdot\bm{\omega}\right)^{2}\right\rangle^{(0)}+2\eta_{t}^{2}a^{2}\left\langle\left[\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{\omega}\right]^{2}\right\rangle^{(0)}+2\frac{\eta_{t}^{2}}{q^{2}a^{2}}\left\langle\left[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{V}\right]^{2}\right\rangle^{(0)}\\\ +4\left[\left(2\eta_{t}-1\right)\left(\eta_{n}-\eta_{t}\right)-\frac{\eta_{t}^{2}}{q^{2}}+\frac{1}{2}\frac{\eta_{t}}{q}-\frac{\eta_{t}^{2}}{q}+\frac{\eta_{t}}{q}\left(\eta_{n}-\eta_{t}\right)\right]\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)\right\rangle^{(0)}\\\ -\frac{2\eta_{t}}{q}\left(\frac{2\eta_{t}}{q}-1\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\omega}\right)\right\rangle^{(0)}\,.$ (54) The terms that vanish by symmetry are already left out at this point. The contributions to $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{A}\right\rangle^{\text{even}}$ have exactly the same form. For $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{A}\right\rangle^{\text{odd}}$ one finds the following contributions $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{A}\right\rangle^{\text{odd}}=\left(\frac{4\eta_{t}^{2}}{q}-\frac{4\eta_{t}}{q}-4\eta_{t}\right)\left\langle\left(\bm{v}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}-\frac{4\eta_{t}}{q}\left(\eta_{n}-\eta_{t}\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}\\\ +\frac{2\eta_{t}}{q}\left(1-2\eta_{t}\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)\right\rangle^{\text{odd}}-\frac{4\eta_{t}^{2}}{q}\left\langle\left[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{V}\right]\left[\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{\omega}\right]\right\rangle^{\text{odd}}\\\ +\left[4\left(\frac{\eta_{t}}{q}-1\right)\left(\eta_{n}-\eta_{t}\right)+\frac{2\eta_{t}}{q}\right]\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}\,.$ (55) Correspondingly, the $\Delta_{B}$-part may be written as $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{B}\right\rangle^{(0)}=\frac{2\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)\left(2\eta_{t}-1\right)^{2}\left\langle\bm{v}^{2}\left(\bm{n}\times\bm{\Omega}\right)^{2}\right\rangle^{(0)}+\frac{2\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)\left\langle\bm{V}^{2}\left(\bm{n}\times\bm{\Omega}\right)^{2}\right\rangle^{(0)}\\\ +\frac{2\eta_{t}^{2}}{q^{2}a^{2}}\left(2\eta_{t}-1\right)^{2}\left\langle\bm{v}^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{(0)}+\frac{2\eta_{t}^{2}}{q^{2}a^{2}}\left\langle\bm{V}^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{(0)}+4\eta_{t}\left(\eta_{t}-1\right)\left\langle\bm{v}^{2}\bm{\omega}^{2}\right\rangle^{(0)}\\\ +4\left(\eta_{n}^{2}-\eta_{n}-\eta_{t}^{2}+\eta_{t}\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\bm{\omega}^{2}\right\rangle^{(0)}+\frac{8\eta_{t}}{q}\left(\frac{\eta_{t}}{q}-1\right)\left(\eta_{n}^{2}-\eta_{n}-\eta_{t}^{2}+\eta_{t}\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{(0)}\\\ +\frac{8\eta_{t}^{2}}{q}\left(\frac{2\eta_{t}}{q}-1\right)\left(2\eta_{t}-1\right)\left\langle\left[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{\Omega}\right]^{2}\right\rangle^{(0)}+\frac{8\eta_{t}^{2}}{q^{2}a^{2}}\left(\eta_{n}^{2}-\eta_{n}-\eta_{t}^{2}+\eta_{t}\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{(0)}\\\ +2\eta_{t}^{2}a^{2}\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\bm{\omega}^{2}\right\rangle^{(0)}+2\eta_{t}^{2}a^{2}\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\bm{\Omega}^{2}\right\rangle^{(0)}+\frac{8\eta_{t}^{3}}{q}a^{2}\left(\frac{\eta_{t}}{q}-1\right)\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{4}\right\rangle^{(0)}\\\ +\frac{8\eta_{t}^{4}}{q^{2}}\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{(0)}\,.$ (56) The contributions to $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{B}\right\rangle^{\text{even}}$ again are formally equivalent to the above expression. This leaves us with $\left\langle\left(\hat{b}_{12}-1\right)\Delta_{B}\right\rangle^{\text{odd}}=-\frac{4\eta_{t}}{q}\left(2\eta_{t}-1\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\left(\bm{v}\cdot\bm{V}\right)\right\rangle^{\text{odd}}+\frac{8\eta_{t}^{2}}{q}\left\langle\left[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{\omega}\right]\left[\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{V}\right]\right\rangle^{\text{odd}}\\\ -\frac{8\eta_{t}}{q}\left(\eta_{n}-\eta_{t}\right)\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}\,.$ (57) We have now reduced the problem to the tedious but straightforward calculation of a considerable number of averages. This task is best suited for a computer algebra system and thus we only tabulate the results. To simplify the notation we introduce the abbreviations $\tilde{\nu}\equiv\nu\sqrt{T/m\pi}$, $\tilde{T}\equiv T/m$, and $\tilde{R}\equiv R/I$ $\left\langle\left(\bm{v}\cdot\bm{\omega}\right)^{2}\right\rangle^{\left(0\right)}=-4\tilde{\nu}\tilde{T}\tilde{R}$ (58a) $\left\langle\left(\bm{V}\cdot\bm{\omega}\right)^{2}\right\rangle^{\left(0\right)}=-3\tilde{\nu}\tilde{T}\tilde{R}$ (58b) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\cdot\bm{\omega}\right)^{2}\right\rangle^{\left(0\right)}=-2\tilde{\nu}\tilde{T}\tilde{R}$ (58c) $\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)^{2}\left(\hat{\bm{n}}\cdot\bm{\omega}\right)^{2}\right\rangle^{\left(0\right)}=-\tilde{\nu}\tilde{T}\tilde{R}$ (58d) $\left\langle[\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{\omega}]^{2}\right\rangle^{\left(0\right)}=-2\tilde{\nu}\tilde{R}^{2}$ (58e) $\left\langle[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{V}]^{2}\right\rangle^{\left(0\right)}=-2\tilde{\nu}\tilde{T}^{2}$ (58f) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)\right\rangle^{\left(0\right)}=-2\tilde{\nu}\tilde{T}\tilde{R}$ (58g) $\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\omega}\right)\right\rangle^{\left(0\right)}=-\tilde{\nu}\tilde{T}\tilde{R}$ (58h) $\left\langle\bm{v}^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\left(0\right)}=-8\tilde{\nu}\tilde{T}\tilde{R}$ (58i) $\left\langle\bm{V}^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\left(0\right)}=-6\tilde{\nu}\tilde{T}\tilde{R}$ (58j) $\left\langle\bm{v}^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{\left(0\right)}=-12\tilde{\nu}\tilde{T}^{2}$ (58k) $\left\langle\bm{V}^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{\left(0\right)}=-6\tilde{\nu}\tilde{T}^{2}$ (58l) $\left\langle\bm{v}^{2}\bm{\omega}^{2}\right\rangle^{\left(0\right)}=-12\tilde{\nu}\tilde{T}\tilde{R}$ (58m) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\bm{\omega}^{2}\right\rangle^{\left(0\right)}=-6\tilde{\nu}\tilde{T}\tilde{R}$ (58n) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\left(0\right)}=-4\tilde{\nu}\tilde{T}\tilde{R}$ (58o) $\left\langle[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{\Omega}]^{2}\right\rangle^{\left(0\right)}=-2\tilde{\nu}\tilde{T}\tilde{R}$ (58p) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{\left(0\right)}=-4\tilde{\nu}\tilde{T}^{2}$ (58q) $\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\bm{\omega}^{2}\right\rangle^{\left(0\right)}=-6\tilde{\nu}\tilde{R}^{2}$ (58r) $\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\bm{\Omega}^{2}\right\rangle^{\left(0\right)}=-10\tilde{\nu}\tilde{R}^{2}$ (58s) $\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{4}\right\rangle^{\left(0\right)}=-8\tilde{\nu}\tilde{R}^{2}$ (58t) $\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\right\rangle^{\left(0\right)}=-4\tilde{\nu}\tilde{T}\tilde{R}$ (58u) $\left\langle\left(\bm{v}\cdot\bm{\omega}\right)^{2}\right\rangle^{\text{even}}=-24\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59a) $\left\langle\left(\bm{V}\cdot\bm{\omega}\right)^{2}\right\rangle^{\text{even}}=-15\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59b) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\cdot\bm{\omega}\right)^{2}\right\rangle^{\text{even}}=-6\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59c) $\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)^{2}\left(\hat{\bm{n}}\cdot\bm{\omega}\right)^{2}\right\rangle^{\text{even}}=-3\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59d) $\left\langle[\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{\omega}]^{2}\right\rangle^{\text{even}}=2\tilde{\nu}\tilde{T}\tilde{R}^{3}$ (59e) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)\right\rangle^{\text{even}}=-12\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59f) $\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\omega}\right)\right\rangle^{\text{even}}=-6\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59g) $\left\langle\bm{v}^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=6\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59h) $\left\langle\bm{V}^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=3\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59i) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=6\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59j) $\left\langle[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{\Omega}]^{2}\right\rangle^{\text{even}}=6\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (59k) $\left\langle\bm{\omega}^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=3\tilde{\nu}\tilde{T}\tilde{R}^{3}$ (59l) $\left\langle\bm{\Omega}^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=7\tilde{\nu}\tilde{T}\tilde{R}^{3}$ (59m) $\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{4}\right\rangle^{\text{even}}=8\tilde{\nu}\tilde{T}\tilde{R}^{3}$ (59n) $\left\langle\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=0$ (59o) $\left\langle\left(\bm{v}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}=-20\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (60a) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\bm{V}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}=-10\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (60b) $\left\langle\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\left(\bm{v}\cdot\bm{\omega}\right)\right\rangle^{\text{odd}}=-7\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (60c) $\left\langle\left(\hat{\bm{n}}\cdot\bm{v}\right)\left(\hat{\bm{n}}\cdot\bm{V}\right)\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\right\rangle^{\text{odd}}=-4\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (60d) $\left\langle[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{V}][\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{\omega}]\right\rangle^{\text{odd}}=0$ (60e) $\left\langle\left(\hat{\bm{n}}\cdot\bm{\omega}\right)\left(\hat{\bm{n}}\cdot\bm{\Omega}\right)\left(\bm{v}\cdot\bm{V}\right)\right\rangle^{\text{odd}}=-2\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}$ (60f) $\left\langle[\left(\hat{\bm{n}}\times\bm{v}\right)\cdot\bm{\omega}][\left(\hat{\bm{n}}\times\bm{\Omega}\right)\cdot\bm{V}]\right\rangle^{\text{odd}}=-5\tilde{\nu}\tilde{T}^{2}\tilde{R}^{2}\,.$ (60g) ### A.2 The correction terms for the temperatures To calculate $dT/dt=\left\langle i\mathcal{L}_{+}T\right\rangle_{t}$ one essentially proceeds along the same lines of reasoning as detailed above. First of all, it is again advantageous to write the corrections to the Gaussian distribution function explicitly, that is, $\left\langle i\mathcal{L}_{+}T\right\rangle_{t}=\left\langle\left(\hat{b}_{12}-1\right)T\right\rangle^{(0)}+b(t)\left\langle\left(\hat{b}_{12}-1\right)T\right\rangle^{(1)}$ (61) where $\frac{3}{4m}\left(\hat{b}_{12}-1\right)T=\eta_{t}\left(\eta_{t}-1\right)\left(\hat{\bm{n}}\times\bm{v}\right)^{2}\\\ +\eta_{n}\left(\eta_{n}-1\right)\left(\hat{\bm{n}}\cdot\bm{v}\right)^{2}\\\ +\eta_{t}^{2}a^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\\\ +\eta_{t}\left(2\eta_{t}-1\right)a\bm{v}\cdot\left(\hat{\bm{n}}\times\bm{\Omega}\right)$ (62) and $\displaystyle\frac{3}{4m}\left(\hat{b}_{12}-1\right)R$ $\displaystyle=\frac{\eta_{t}^{2}}{q}\left(\hat{\bm{n}}\times\bm{v}\right)^{2}$ (63) $\displaystyle+\eta_{t}\left(\frac{\eta_{t}}{q}+1\right)a^{2}\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}$ $\displaystyle+\eta_{t}\left(\frac{2\eta_{t}}{q}+1\right)a\bm{v}\cdot\left(\hat{\bm{n}}\times\bm{\Omega}\right)\,.$ The term $\left\langle\left(\hat{b}_{12}-1\right)T\right\rangle^{(0)}$ is already known Aspelmeier et al. (2001) and the only other contribution is $\left\langle\left(\hat{\bm{n}}\times\bm{\Omega}\right)^{2}\right\rangle^{\text{even}}=2\tilde{\nu}\tilde{T}\tilde{R}^{2}$. ## References * Goldhirsch (2003) I. Goldhirsch, Ann. Rev. Fluid Mech. 35, 267 (2003). * Levy and Kalman (2001) A. Levy and H. Kalman, _Handbook of Conveying and Handling of Particulate Solids_ (Elsevier, Amsterdam, 2001). * Greenberg and Brahic (1984) R. Greenberg and A. Brahic, eds., _Planetary Rings_ (Arizona Univ. Press., Tucson, 1984). * Pöschel and Luding (2001) T. Pöschel and S. Luding, eds., _Granular Gases_ , vol. 564 of _Lecture Notes in Physics_ (Springer, Berlin, 2001). * Pöschel and Brilliantov (2003) T. Pöschel and N. V. Brilliantov, eds., _Granular Gas Dynamics_ , vol. 624 of _Lecture Notes in Physics_ (Springer, Berlin, 2003). * Brilliantov and Pöschel (2004) N. V. Brilliantov and T. Pöschel, _Kinetic Theory of Granular Gases_ (Oxford University Press, Oxford, 2004). * McNamara and Young (1992) S. McNamara and W. R. Young, Phys. Fluids A 4, 496 (1992). * Goldhirsch and Zanetti (1993) I. Goldhirsch and G. Zanetti, Phys. Rev. Lett. 70, 1619 (1993). * Brito and Ernst (1998) R. Brito and M. H. Ernst, Europhys. Lett. 43, 497 (1998). * Goldshtein and Shapiro (1995) A. Goldshtein and M. Shapiro, J. Fluid Mech. 282, 75 (1995). * van Noije and Ernst (1998) T. P. C. van Noije and M. H. Ernst, Granular Matter 1, 57 (1998). * Esipov and Pöschel (1997) S. E. Esipov and T. Pöschel, J. Stat. Phys. 86, 1385 (1997). * Brey et al. (1999a) J. J. Brey, D. Cubero, and M. J. Ruiz-Montero, Phys. Rev. E 59, 1256 (1999a). * Deltour and Barrat (1997) P. Deltour and J.-L. Barrat, J. Physique I 7, 137 (1997). * Huthmann et al. (2000) M. Huthmann, J. Orza, and R. Brito, Granular Matter 2, 189 (2000). * Brilliantov and Pöschel (2000a) N. V. Brilliantov and T. Pöschel, Phys. Rev. E 61, 2809 (2000a). * Goldhirsch et al. (2003) I. Goldhirsch, H. S. Noskowicz, and O. Bar-Lev, in Pöschel and Brilliantov (2003), pp. 37–63. * Pöschel et al. (2006) T. Pöschel, N. V. Brilliantov, and A. Formella, Phys. Rev. E 74, 041302 (2006). * Brey et al. (1999b) J. J. Brey, M. J. Ruiz-Montero, and R. Garcia-Rojo, Phys. Rev. E 60, 7174 (1999b). * Brey et al. (2000) J. J. Brey, M. J. Ruiz-Montero, D. Cubero, and R. Garcia-Rojo, Physics of Fluids 12, 876 (2000). * Brilliantov and Pöschel (2000b) N. V. Brilliantov and T. Pöschel, Phys. Rev. E 61, 1716 (2000b). * Santos and Dufty (2001) A. Santos and J. W. Dufty, Phys. Rev. Lett. 86, 4823 (2001). * Garzo and Montanero (2004) V. Garzo and J. M. Montanero, Phys. Rev. E 69, 021301 (2004). * Huthmann and Zippelius (1997) M. Huthmann and A. Zippelius, Phys. Rev. E 56, R6275 (1997). * Aspelmeier et al. (2001) T. Aspelmeier, M. Huthmann, and A. Zippelius, in Pöschel and Luding (2001), p. 31. * Goldhirsch et al. (2005a) I. Goldhirsch, S. H. Noskowicz, and O. Bar-Lev, J. Phys. Chem. 109, 21449 (2005a). * Elperin and Golshtein (1997) T. Elperin and E. Golshtein, Physica A 247, 67 (1997). * Jenkins and Richman (1985) J. T. Jenkins and M. W. Richman, Physics of Fluids 28, 3485 (1985). * Lun and Savage (1987) C. K. K. Lun and S. B. Savage, J. Appl. Mech. Trans. ASME 54, 47 (1987). * Jaeger et al. (1990) H. M. Jaeger, C. Liu, S. R. Nagel, and T. A. Witten, Europhys. Lett. 11, 619 (1990). * Luding (1995) S. Luding, Phys. Rev. E 52, 3416 (1995). * Jenkins and Louge (1997) J. T. Jenkins and M. Louge, Physics of Fluids 9 (10), 2835 (1997). * Bardenhagen et al. (2000) S. G. Bardenhagen, J. U. Brackbill, and D. Sulsky, Phys. Rev. E 62, 3882 (2000). * Cafiero et al. (2002) R. Cafiero, S. Luding, and H. J. Herrmann, Europhys. Lett. 60, 854 (2002). * Mitarai et al. (2002) N. Mitarai, H. Hayakawa, and H. Nakanishi, Phys. Rev. Lett. 88, 174301 (2002). * Goldhirsch et al. (2005b) I. Goldhirsch, S. H. Noskowicz, and O. Bar-Lev, Phys. Rev. Lett. 95, 068002 (2005b). * Brilliantov et al. (2007) N. V. Brilliantov, T. Pöschel, W. T. Kranz, and A. Zippelius, Phys. Rev. Lett. 98, 128001 (2007). * Brilliantov et al. (1996) N. V. Brilliantov, F. Spahn, J.-M. Hertzsch, and T. Pöschel, Phys. Rev. E 53, 5382 (1996). * Schwager and Pöschel (1998) T. Schwager and T. Pöschel, Phys. Rev. E 57, 650 (1998). * Ramírez et al. (1999) R. Ramírez, N. V. Brilliantov, T. Schwager, and T. Pöschel, Phys. Rev. E 60, 4465 (1999). * Becker et al. (2008) V. Becker, T. Schwager, and T. Pöschel, Phys. Rev. E 77, 011304 (2008). * Luding et al. (1998) S. Luding, M. Huthmann, S. McNamara, and A. Zippelius, Phys. Rev. E 58, 3416 (1998). * Pöschel et al. (2003) T. Pöschel, N. V. Brilliantov, and T. Schwager, Int. J. Mod. Phys. C 13, 1263 (2003). * Noskowicz et al. (2007) S. H. Noskowicz, O. Bar-Lev, D. Serero, and I. Goldhirsch, Europhys. Lett. 79, 60001 (2007). * Bird (1994) G. A. Bird, _Molecular Gas Dynamics and the Direct Simulation of Gas Flows_ (Oxford University Press, 1994). * Pöschel and Schwager (2005) T. Pöschel and T. Schwager, _Computational Granular Dynamics_ (Springer, New York, 2005). * Gayen and Alam (2008) B. Gayen and M. Alam, Phys. Rev. Lett. 1, 068002 (2008). * Brey and Cubero (2001) J. J. Brey and D. Cubero, in _Granular Gases_ , edited by T. Pöschel and S. Luding (Springer, Berlin, 2001), vol. 564 of _Lecture Notes in Physics_ , p. 59.	arxiv-papers	2008-08-14T07:19:23	2024-09-04T02:48:57.288725	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "W. T. Kranz, N. V. Brilliantov, T. Poeschel, A. Zippelius", "submitter": "Wolf Till Kranz", "url": "https://arxiv.org/abs/0808.1952" }
0808.1969	# Determination of the single-ion anisotropy energy in a $S$ = 5/2 kagome antiferromagnet using x-ray absorption spectroscopy. M. A. de Vries m.a.devries@physics.org Laboratory for Quantum Magnetism, École Polytechnique Fédérale de Lausanne, Station 3, CH-1015 Lausanne, Switzerland CSEC and School of Chemistry, The University of Edinburgh, Edinburgh EH9 3JZ, UK T. K. Johal Daresbury Laboratory, Warrington WA4 4AD, UK A. Mirone ESRF, 6 rue Jules Horowitz, F-38000 Grenoble, France J. S. Claydon Department of Engineering Materials, University of Sheffield, Sheffield S1 3JD, UK G. J. Nilsen Laboratory for Quantum Magnetism, École Polytechnique Fédérale de Lausanne, Station 3, CH-1015 Lausanne, Switzerland CSEC and School of Chemistry, The University of Edinburgh, Edinburgh EH9 3JZ, UK H. M. Rønnow Laboratory for Quantum Magnetism, École Polytechnique Fédérale de Lausanne, Station 3, CH-1015 Lausanne, Switzerland G. van der Laan Daresbury Laboratory, Warrington WA4 4AD, UK Diamond Light Source, Chilton, Didcot OX11 0DE, UK A. Harrison CSEC and School of Chemistry, The University of Edinburgh, Edinburgh EH9 3JZ, UK Institut Laue-Langevin, 6 rue Jules Horowitz, F-38000 Grenoble, France ###### Abstract We report x-ray absorption and x-ray linear dichroism measurements at the Fe $L_{2,3}$ edges of the geometrically frustrated systems of potassium and hydronium iron jarosite. Comparison with simulated spectra, involving ligand- field multiplet calculations modelling the $3d$-$2p$ hybridization between the iron ion and the oxygen ligands, has yielded accurate estimates for the ligand metal-ion hybridization and the resulting single-ion crystal field anisotropy energy. Using this method we provide an experimentally verified scenario for the appearance of a single-ion anisotropy in this nominally high-spin $3d^{5}$ orbital singlet ${}^{6}S$ system, which accounts for features of the spin-wave dispersion in the long-range ordered ground state of potassium iron jarosite. ###### pacs: 75.25.+z, 71.70.Ch, 71.70.Ej, 75.10.Dg ## I Introduction The jarosite group of minerals has received much attention for being close to ideal realizations of the kagome antiferromagnet. The kagome topology, which is named after a Japanese basket weaving pattern, frustrates the antiferromagnetic Néel ordering Diep:04 . Due to the resulting “underconstraint” neither in the quantum case of $S$ = 1/2 nor in the classical limit is the kagome antiferromagnet expected to show a symmetry- breaking transition, even at $T=0$ K Reimers:91 ; Chalker:92 ; Reimers:93 . Physical realizations of this system are highly valued because they allow for the investigation of the Mott insulating phase in the absence of Néel-like magnetic order. The jarosite group of general stoichiometry $AM_{3}$(SO4)2(OH)6, where $A$ can be—amongst others—K, Na, Rb, Ag, NH3, or H3O, provides physical models for the kagome antiferromagnet with spin $S$ = 3/2 and 5/2, where the $M$ site is occupied by trivalent Cr KerenJar:96 ; Inami:01 ; Morimoto:03 and Fe Takano:68 ; Townsend:86 ; Wills:96 ; WillsThesis ; KerenJar:96 ; WillsKjar:00 ; WillsKjar:01 ; Grohol:05 , respectively. $M=$ V is a $S$ = 1 system with ferromagnetic near-neighbor interactions Grohol:02 ; Grohol:03van . Fig. 1 illustrates the jarosite structure of well separated kagome layers consisting of $M$O6 octahedra. Potassium iron jarosite is representative for most of the iron analogues. In this system a transition to a non-collinear long-range ordered state is observed at 64 K, despite that the magnitude of the Curie-Weiss temperature, $\Theta_{\rm{CW}}$, is as high as $-$800 K Grohol:03 ; WillsKjar:00 . The ground state has shown to be a so-called $q=0$ state of positive chirality, in which the spins lie within the kagome plane and all point either in or out of each shared kagome triangle Townsend:86 ; Inami:00 ; WillsKjar:01 ; Inami:03 ; Grohol:05 . The large figure $\Theta_{\text{CW}}/T_{\rm{N}}=12.5$ is an indication of relatively strong geometric frustration Ramirez:94 , but clearly the observed long-range ordered ground state points to additional terms in the Hamiltonian of this kagome antiferromagnet, beyond near-neighbor exchange. It has been shown that the $q=0$ ground state can arise due to magnetic anisotropies such as easy-plane single-ion crystal field (CF) anisotropy $D_{z}\hat{S}_{z}^{2}-E_{xy}(\hat{S}_{x}^{2}-\hat{S}_{y}^{2}),$ Nishiyama:03 ; Yildirim:06 where $D_{z}$ is the zero-field splitting parameter, or a Dzyaloshinsky-Moriya interaction (DMI) Elhajal:02 ; Yildirim:06 $D_{ij}\hat{S}_{i}\times\hat{S}_{j}$. The DMI term has shown to be symmetry allowed on the kagome lattice and has been argued to prevail over the CF anisotropy in the spherical ($L=0$) 3$d^{5}$ Fe3+ ion Elhajal:02 . Fits to the spin-wave dispersion curves measured with neutron spectroscopy on single crystal Matan:06 and powder samples Coomer:06 of potassium jarosite have found to be slightly better in a model with DMI compared to the model with CF anisotropy. The evidence so far has, however, been inconclusive. Given the similarity in crystal structure of potassium (iron) jarosite and hydronium (iron) jarosite the above effects could expected to be equally important in the latter. However, hydronium jarosite has a spin-glass ground state with a freezing temperature of $T_{g}=17(2)$ K Bisson:unpub07 . The Weiss temperature for hydronium jarosite is less accurately known ($-$700 to $-$1400 K) because the inverse susceptibility does not enter a linear regime sufficiently far below the decomposition temperature of the compound Wills:96 . It has been suggested that the spin-glass state is the result of proton transfer from the hydronium groups to the Fe-(O${}_{\text{eq}}$H)-Fe super- exchange mediating hydroxy groups Grohol:03 . One indication that this could happen might be that the OH groups at the crystallographically analogous location in the related kagome compound zinc paratacamite can form a muonium state by chemically binding a positive muon Mendels:07 . The aim of the work described here is to investigate the origins of the magnetic anisotropy (of either single-ion or DMI character) in the iron jarosites, and to understand the essential differences between the salts such as potassium jarosite with a long range ordered ground state and hydronium jarosite. Figure 1: (color online) The jarosite structure viewed perpendicular to the $c$ axis (left picture) and along the $c$ axis (right picture), revealing the kagome network of magnetic ions. The $M$O6 octahedra are drawn in transparent brown with the $M^{3+}$ ions at the center in dark brown. The SO4 tetrahedra are transparent yellow. The large spheres at the origin of the unit cells in blue are the $A$ site ions, and the hydrogen of the OH groups are drawn as small white spheres. Using Mössbauer spectroscopy Afanasev:74 ; Bonville:06 the spin state of the Fe3+ in the iron jarosites was found to be $S$ = 5/2, thus nominally high-spin $3d^{5}$. The free-ion 3$d^{5}$ state is ${}^{6}S$, i.e. $L=0$. Hence, there is no orbital angular momentum due to the spin-orbit coupling in the presence of a crystal field as long as the system is high-spin 3$d^{5}$. This holds at least up to third order perturbation theory of the trigonal crystal-field distortions, spin-spin and spin-orbit couplings111It has been shown Watanabe:57 that in 4th, 5th and 6th order perturbation theory of the crystal field and spin-spin interactions a small orbital angular momentum and single- ion anisotropy arises.. Even in a $D_{4h}$ crystal field the 3$d^{5}$ $S$ = 5/2 spin should therefore be practically isotropic apart from the possible effects of dipole-dipole interactions. Both DMI and single-ion anisotropy arise only when the spin-orbit operator $L\cdot S$ can induce a finite orbital angular momentum Moriya:60 ; Watanabe:57 . That $\langle L\rangle>0$ in the iron jarosites is clear from Curie-Weiss fits to the high temperature magnetic susceptibility which indicate $\mu_{\text{eff}}$ between 6.2 and 6.7 $\mu_{\text{B}}$ in potassium iron jarosite Grohol:05 and 6.6(2) $\mu_{\text{B}}$ in the hydronium analogue Wills:96 ; WillsH3Ojar:00 (compared to a spin only value of the effective moment of 5.92 $\mu_{\text{B}}$). Clearly, this opens up the possibility for both DMI and single ion anisotropies. However, the origin of the orbital angular momentum itself has so far received little attention. A very strong crystal field, such as in iron phthalocyanine Thole:88Spinmix can give rise to a low-spin ground state in which the spin-orbit coupling will in general reinstate part of the orbital angular momentum of the free-ion configuration which is, in zero order, quenched by the crystal field. It could be that in weaker crystal fields, such as those present in iron oxides, some orbital angular momentum appears due to a mixing in of the low-spin state into the ground state. The crystal field alone does, however, not provide matrix elements between high-spin and low-spin states. The most likely explanation is therefore that an Fe2+ configuration mixes in due to charge transfer Zaanen:85 from the surrounding ligands. For hematite, e.g., the effective valence of the Fe ion has been estimated at 2.43 due to charge transfer from the oxygen ligands Coey:71 . These effects have earlier been shown to explain zero field splitting in Mn2+ and Fe3+ substituted crystals Wan-Lun:89 ; Wan-Lun:94 and in the Mn2+ compounds MnF3 Fransisco:88 , MnPS3 and MnPSe3 Jeevanandam:99 , despite the lower covalence of the fluoride and phosphorus-containing ligands compared to oxygen. Transition metal $2p\to 3d$ x-ray spectroscopy studies have played a key role in the understanding of the effect of charge-transfer and crystal-field in determining the electronic and magnetic properties in transition metals and their compounds Zaanen:85 ; vanderLaan:86 ; FdG:94Rev ; FdG:05Rev . We have measured the x-ray absorption and x-ray linear dichroism (XLD) at the Fe $L_{2,3}$ ($2p\to 3d$) edges on single-crystalline and powder samples of potassium and hydronium iron jarosite. By comparison with atomic multiplet calculations, taking into account ligand-field effects and multiple ionic configurations, we obtain accurate values for the charge transfer and single- ion anisotropy in the $S$ = 5/2 kagome antiferromagnet. The results are compared to magnetic susceptibility and neutron spectroscopy measurements on in particular potassium iron jarosite. ## II Experimental Single crystals of potassium iron jarosite, KFe3(OH)6(SO4)2, were grown using a hydrothermal reduction-oxidation method as described in Ref. Grohol:03, ; 4.88 g (28.0 mmol) K2SO4 and 2.2 mL (40 mmol) H2SO4 were dissolved in 50 mL distilled water, and transferred into a 125 mL PTFE liner of a stainless steel bomb. 0.56 g (10 mmol) iron wire with a diameter of 2 mm was added to the solution. The bomb was placed in an oven at 202∘C for 4 days, then cooled down to room temperature at a rate of 0.3∘C/min. The precipitate was washed, filtered and dried, yielding 0.37 g, which is 22% based on Fe. In order to obtain sufficiently large crystals it was important that the surface area of the solution exposed to air inside the vessel was minimized. This was achieved using a glass container which fitted inside the PTFE liner, with only a small hole at the top. Using this method single crystals of up to 0.7 mm in diameter were obtained, which was sufficiently large for the application in polarized x-ray spectroscopy. A number of crystals were characterized using single- crystal x-ray diffraction. These crystals were identified as potassium iron jarosite and face-indexed. The original solvothermal synthesis method Dutrizac:76 ; WillsThesis for the jarosites was used for the preparation of H3OFe3(OH)6(SO4)2. In this preparation 6.6 g (22 mmol) of Fe2(SO4)${}_{3}\cdot$5H2O was dissolved in 50 mL water. This solution was transferred to a 125 mL PTFE liner of a stainless steel bomb. The solution was heated in the bomb to 140∘C for 12 hours. The hydronium iron jarosite which precipitated during the reaction was washed, filtered and dried, yielding $\sim$0.27 g of product. Polarized soft x-ray absorption measurements were carried out using the liquid helium cryostat, housed in the high-field superconducting magnet on ID08 at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France. Total- electron-yield spectra were measured by recording the drain current from the sample as a function of photon energy in the region of the Fe $L_{2,3}$ absorption edges. Potassium iron jarosite is strongly insulating, so the resonance enhanced total-electron-yield signal at the iron edge is much smaller than the signal from the surrounding metal of the sample holder. To avoid this source of background intensity a potassium iron jarosite single crystal of 0.7 mm diameter was attached using silver glue to the end of a graphite tip, which in turn was glued to the end of a standard ESRF sample holder. The sample was aligned in the x-ray beam with the crystallographic $c$ axis pointing in the horizontal plane and perpendicular to the incident beam. This allowed us to measure absorption spectra with the incident x-ray polarization both perpendicular and parallel to the $c$ axis. Spectra were recorded with alternating horizontal and vertical polarization, and for each polarization averages were taken over at least eight scans. In this way the polarization dependent spectra were taken at a number of temperatures between 20 and 290 K, in zero external field. Isotropic absorption spectra at the Fe $L_{2,3}$ edges were obtained on beamline 5U.1 of the Synchrotron Radiation Source (SRS) at Daresbury Laboratory, UK, using potassium and hydronium iron jarosite powder dispersed over double-sided UHV-compatible carbon tape. ## III Results Figure 2 shows the spectra obtained from single-crystal and powder samples of potassium iron jarosite at 290 K. The upper curves in Fig. 2 display the x-ray absorption spectra from a single crystal measured on beamline ID08 at the ESRF, with the incident x-ray polarization parallel to the $c$ axis, $I_{c}$ (black circles), and perpendicular to the $c$ axis, $I_{\text{ab}}$ (red crosses). The curves in the middle of the figure give the corresponding isotropic absorption spectrum $I_{\text{iso}}=\frac{2}{3}I_{\text{ab}}+\frac{1}{3}I_{c}$ (black line) and the isotropic spectrum directly measured from a powder sample on station 5U.1 at the SRS (broken red line). The curve at the bottom gives the x-ray linear dichroism $I_{\text{xld}}$ defined as $I_{\text{ab}}-I_{\text{c}}$ (thin black line). The $2p\to 3d$ transition consists of two edges, the first one at $\sim$$709$ eV ($L_{3}$ edge) corresponds to the creation of a $2p_{3/2}$ core hole. The transition giving rise to a $2p_{1/2}$ core hole ($L_{2}$ edge) is at $\sim$$12$ eV higher photon energy due to the $2p$ spin-orbit coupling. The inset of Fig. 2 shows the $I_{\text{c}}$ and $I_{\text{ab}}$ spectra in close up at the $L_{3}$ edge, which is the most significant region of the spectrum. Figure 2: (color online) x-ray absorption spectra of potassium iron jarosite at 290 K measured with x-ray polarization parallel and perpendicular to the $c$ axis, $I_{\text{c}}$ (black circles) and $I_{\text{ab}}$ (red crosses), respectively. The lower trace shows the corresponding values of the isotropic absorption, $I_{\text{iso}}=\frac{2}{3}I_{\text{ab}}+\frac{1}{3}I_{\text{c}}$ (black line) and the x-ray linear dichroism, $I_{\text{xld}}=I_{\text{ab}}-I_{\text{c}}$ (thin black line). Also shown $I_{\text{iso}}$ from a powder sample measured on 5U.1 at SRS Daresbury with a lower energy resolution (broken, red line). The inset shows a close up of $I_{\text{c}}$ and $I_{\text{ab}}$ at the $L_{3}$ edge. As seen in Fig. 3, the structure at the $L_{3}$ edge becomes more prominent in the isotropic spectrum as the temperature is lowered. At each stabilized temperature, the reproducibility of up to eight spectra demonstrated the absence of any temporal variation in the spectral line shape or intensity, therefore charging effects and drift in sample alignment can be eliminated as the cause of the observed temperature dependence. These isotropic spectra obtained by weighted averaging of $I_{\text{ab}}$ and $I_{\text{c}}$ are in good agreement with the spectra from powder samples measured for the SRS at all temperatures. That this change in line shape occurs in the powder averaged spectrum might be an indication of a slight change in the electronic structure as the temperature is lowered. Only very small changes were observed in the XLD spectral line shape on cooling the sample, as shown in Fig. 4. This change in the spectral line shape is accompanied by a slight reduction in the integrated XLD spectrum Scho:98 $\int I_{\text{xld}}(\omega)d\omega/\int I_{\text{iso}}(\omega)d\omega$ (1) corresponding to a small change of the total quadrupole moment of 0.05(2). This might be interpreted as that the Fe $3d$ shell gains between 3 and 7% net $z^{2}-r^{2}$ character at base temperature. Figure 3: (color online) Temperature dependence of the isotropic x-ray absorption measured at the Fe $L_{2,3}$ edges from a synthesized potassium iron jarosite single crystal (black lines), compared to calculated spectra (as discussed in Secs. IV and V) with simulated temperatures as listed in the legend. The room temperature spectra of hydronium iron jarosite were identical to the spectra obtained on potassium iron jarosite within the experimental resolution as measured at the SRS (the red, broken line in the main panel of Fig. 2 for comparison). Figure 4: (color online) Temperature dependence of the x-ray linear dichroism measured at the Fe $L_{2,3}$ edges from a synthesized potassium iron jarosite single crystal (black lines), compared to calculated linear dichroism corresponding to the isotropic simulated spectra of Fig. 3, scaled in intensity to fit the experimental spectra by factors ranging from $1/2$ at 290 K down to $1/3$ for the 60 and 40 K data. Possible reasons for this mismatch are discussed in the text. ## IV Ligand-field multiplet calculations Due to the strong electrostatic interactions between the $2p$ core hole and the $3d$ levels in the final state, the $2p$ $\to$ $3d$ absorption spectrum is not simply proportional to the density of unoccupied $3d$ levels vanderlaan:91 as a function of energy. For ferromagnets and magnetically soft materials, where all the magnetic moments can be aligned by applying an external field, the x-ray magnetic circular dichroism (XMCD) and x-ray magnetic linear dichroism (XMLD) sum rules can be applied to obtain the expectation values of the orbital and spin magnetic moments Thole:92 and the magneto-crystalline anisotropy vanderlaan:99 , respectively. In the case of the jarosites it is not possible to align the magnetic moments parallel with an external field due to the large energy scale of the antiferromagnetic interactions ($\Theta_{\rm{CW}}\approx-800$ K). The only way to obtain accurate information about the relevant expectation values is by explicitly calculating the transition probabilities $f(E_{\hbar\omega},\mathbf{q})=\frac{\|\bra{\psi_{g}}\mathbf{\hat{r}_{q}}\ket{\psi_{e}}\|^{2}}{E_{\hbar\omega}-\Delta E-i\Gamma/2},$ (2) between the ground state $\psi_{g}$ and excited state $\psi_{e}$, where $\mathbf{\hat{r}_{q}}$ is the electric-dipole operator for x-ray absorption with polarization $\mathbf{q}$, and $\Gamma$ is the lifetime broadening, which is treated as a fitting parameter. The experimental x-ray spectra were used to obtain good approximations for $\psi_{g}$ and $\psi_{e}$ with the aid of the Hilbert++ code Mirone:00 ; Mirone:code . The Hilbert space for $\ket{\psi_{g}}$ is spanned by determinants of the lowest atomic $LSJ$ configurations of $\ket{3d^{n}}$ and $\ket{3d^{n+1}\underline{L}}$, where $\underline{L}$ denotes a ligand hole. Likewise, the Hilbert space for $\ket{\psi_{e}}$ is spanned by determinants of the atomic configurations of $\ket{2p^{5}3d^{n+1}}$ and $\ket{2p^{5}3d^{n+2}\underline{L}}$. The Slater integrals for the $3d$-$3d$ and $2p$-$3d$ interactions and the $2p$ and $3d$ spin-orbit parameters for the ground state and final state configurations calculated using Cowan’s atomic multiplet program Cowan:81 are tabulated in Ref. vanderlaan:92, . Approximate solutions for the eigenfunctions of the iron ion in the compound are then obtained by diagonalization of the Hamiltonian including charge-transfer and crystal-field terms arising from the surrounding oxygen ligands Mirone:code $\displaystyle\mathcal{H}$ $\displaystyle=$ $\displaystyle\mathcal{H}_{\text{atom}}$ (3) $\displaystyle+$ $\displaystyle\sum_{b}[t_{\sigma,b}d^{+}_{3\tilde{z}^{2}-r^{2}}p^{-}_{\tilde{z}}+t_{\pi,b}(d^{+}_{\tilde{x}\tilde{z}}p^{-}_{\tilde{x}}+d^{+}_{\tilde{y}\tilde{z}}p^{-}_{\tilde{y}})+\text{cc}]$ $\displaystyle+$ $\displaystyle\sum_{b}[V_{\sigma,b}d^{+}_{3\tilde{z}^{2}-r^{2}}d^{-}_{3\tilde{z}^{2}-r^{2}}+V_{\pi,b}(d^{+}_{\tilde{x}\tilde{z}}d^{-}_{\tilde{x}\tilde{z}}+d^{+}_{\tilde{y}\tilde{z}}d^{-}_{\tilde{y}\tilde{z}})]$ $\displaystyle+$ $\displaystyle\epsilon_{p}\sum_{b}(p^{+}_{\tilde{x}}p^{-}_{\tilde{x}}+p^{+}_{\tilde{y}}p^{-}_{\tilde{y}}+p^{+}_{\tilde{z}}p^{-}_{\tilde{z}}).$ In this Hamiltonian, the first term represents the atomic Hamiltonian. The second term gives the hybridization between metal $3d$ and oxygen $2p$ orbitals, where $t_{\sigma,b}$ and $t_{\pi,b}$ are the Slater-Koster hopping parameters. For each bond, $t_{\sigma,\pi}$ are rescaled with the bond length using $t_{\sigma,\pi,b}=t_{\sigma,\pi}(R_{\text{b}}/R_{\text{ref}})^{\alpha}$ where $R_{b}/R_{\text{ref}}$ is the normalized length of Fe-O bond $b$ and the rescaling exponent $\alpha$ is treated as a free parameter. The third term gives the contribution of the electrostatic crystal field which was not used in our case. The last term gives the energy of the $2p$ electrons in the oxygen valence band, where $\epsilon_{p}$ is a simulation parameter related to the charge-transfer gap $\Delta_{pd}$ via $\Delta_{pd}=U_{3d^{5}}-\epsilon_{p}$. The total $dd$ Coulomb energy for the addition of a sixth electron in the Fe $3d$ shell $U_{3d^{5}}=\sum_{n=1}^{5}U_{n}$ was found to be 44.3 eV by considering the case $t_{\sigma,\pi}=0$, where no charge transfer occurs unless the Fe2+ $d^{6}\underline{L}$ configuration has a lower energy than the Fe3+ $d^{5}$ configuration. The sum is made over all bonds $b$, where the local coordinate frame $\tilde{x},\tilde{y},\tilde{z}$ is oriented with the $\tilde{z}$ axis along the bond direction. Numerous comparisons between experiment and theory Thole:PRB85 ; FdG:05Rev have shown that this approach works well, provided that the Slater integrals for the Coulomb interactions and spin-orbit couplings are reduced to 70-80% of the atomic Hartree-Fock values Thole:88Spinmix . Here, all $pd$ and $dd$ Slater integrals were reduced to 80% of the atomic values calculated using Cowan’s program. The lifetime broadening $\Gamma$ was set to $0.24$ eV for the $L_{3}$ edge and $0.3$ eV for the $L_{2}$ edge. These are values as commonly used in the calculation of x-ray absorption spectra, with $L_{2}$ always slightly larger than $L_{3}$. This spectral broadening is due to the coupling with the (infinite dimensional) electromagnetic field. Figure 5: (color online) Experimentally obtained isotropic spectrum and x-ray linear dichroism (black lines) at 290 K and simulated XAS and XLD (red lines) Boltzmann averaged at 110 K. The simulated XLD is roughly twice as large as the experimentally obtained XLD which is ascribed to mis-alignments and sample imperfections. The crystal field at the iron atom in potassium iron jarosite is approximately of $D_{4h}$ symmetry, with an elongation of the bond length $R_{\text{ap}}$ along the fourfold symmetry axis ($c^{\prime}$) of $\sim$4% with respect to the bond length $R_{\text{eq}}$ in the equatorial plane Grohol:03 . The local symmetry axes of the FeO6 octahedra are canted with respect to the crystallographic $c$ axis by an angle of 19.2∘, resulting in buckled FeO4 planes forming the kagome layers. Our simulation shows that the spectra are mainly determined by the strongly anisotropic $\sigma$-type hybridization between the iron $3d$ and oxygen $2p$ orbitals. The electrostatic part of the crystal field could be left out. An excellent agreement between experimental and simulated spectra for $I_{iso}$ was obtained using only four simulation parameters, namely the charge-transfer gap $\Delta_{pd}$, $t_{\sigma}$, $t_{\pi}$ and the rescaling constant $\alpha$ as defined in $t_{\pi,\sigma}^{\text{ap}}=t_{\pi,\sigma}^{\text{eq}}(R_{\text{ap}}/R_{\text{eq}})^{\alpha}$ (with $R_{\text{ap}}/R_{\text{eq}}=1.035$ in the present case). When the simulated isotropic spectra is in good agreement with the experimental spectrum, at the same time optimal agreement is obtained for the linear dichroism spectrum. At high temperature there is good agreement between the experimental and simulated linear dichroism spectrum, apart from an intensity scaling factor of 0.5 as shown in Figs. 4 and 5. With the fitting parameters for the simulation there is no way to reduce the overall intensity of the linear dichroism without radically modifying the isotropic spectrum. For up to 6% the intensity difference might be attributable to the local canting of the FeO6 octahedra with respect to the $ab$-plane. In addition to this there is an inevitable loss of intensity in the XLD due to misalignments and sample imperfection. Reduced energy resolution can also lead to a reduction of the measured XLD. We conclude that the difference in overall intensity of the experimental and calculated $I_{\mathrm{xld}}$ is not due to fundamental shortcomings of the simulation. For reasons we do not quite understand, the agreement between theory and experiment as evident in the temperature dependence of the isotropic spectra is not reflected in the XLD spectra at lower temperatures. We expect that these spectra, which are averages over increasingly fewer states, are also increasingly sensitive to shortcomings in our simulations. Since the XLD is an orientational energy dependence it will be particularly sensitive for any shortcomings in our theoretical model. It should be emphasised that, despite the differences between the experimental and simulated spectra at low temperatures, the overall agreement between experiment and simulation as obtained here is remarkably good. The parameters as found in our simulation are given in Table 1 (left-hand side). The large error bar on the value for the charge-transfer gap of 3.5(2.0) eV reflects the adjustment of the total $dd$ Coulomb energy on the iron site $U_{dd}^{\text{tot}}=U_{3d^{5}}(\langle n\rangle-5)$ to any increase in $\epsilon_{p}$ by transfer of additional electron weight into the Fe ion. This is reflected in the total error in $\langle n\rangle$ of $\sim$0.1 electron. The values found here for $\Delta_{pd}$ and $t_{\sigma}$ are similar to those found previously for other Fe3+ compounds FdG:03JACS ; FdG:05Rev . Fe2O3, which is in many respects probably the compound most closely related to iron jarosite, was previously found to have a band gap of predominantly charge- transfer character of 2 to 2.5 eV Ciccacci:91 . With an unusually small value for $t_{\pi}$ of 0.3(2) eV the lowest energy peak, at 707.8 eV is reproduced in the simulated spectrum. The main peak at 709 eV splits into two peaks with the introduction of a trigonal distortion of the FeO6 octahedron. A common value for the rescaling exponent $\alpha$ is $-$3 and the small value found here ($-$7.5) is an indication that the charge anisotropy is larger than the crystal field anisotropy as defined by the oxygen positions around the iron. The lowest energy low-spin configuration, corresponding to the ${}^{4}F_{11/2}$ free-ion configuration, lies 1.2 eV higher in energy and is not found to mix into the ground state. Table 1: Results from simulations of the Fe $L_{2,3}$ spectra of potassium iron jarosite, with ligand-field multiplet calculations using the Hilbert++ codeMirone:code . The left-hand side lists all the parameters varied in the simulation. The right-hand side gives the resulting expectation values for the calculated ground-state properties of the Fe ion. The errors indicate the range over which the parameters could be varied whilst maintaining a good agreement with the data. The right-hand side of the lowest panel gives the effective moment in the case of full orbital polarization, while the calculation indicates only a very weak orbital polarization of the added electron weight due to charge transfer. We use the convention that upper case $J,L,S$ represent operators and lower case $j,l,s$ represent the quantum numbers given by $\langle L^{2}\rangle=l(l+1)$ etc. Simulation pars. \| \| Expectation values \| ---\|---\|---\|--- $\Delta_{pd}$ (eV) \| 3.5(2.0) \| $D_{z}$ (meV) \| 0.5(1) $t_{\pi}$ (eV) \| 0.3(2) \| $\sqrt{\langle S^{2}\rangle}=\langle S\rangle$ \| 2.77(6) $t_{\sigma}$ (eV) \| 3.0(2) \| $\leadsto s=2.32$ \| $\alpha$ \| $-$7.5 \| $\sqrt{\langle L^{2}\rangle}(\neq\langle L\rangle)$ \| 1.5(2) \| \| $\leadsto l=1.11$ \| \| \| but $\langle L\rangle$ \| 0.008(3) \| \| $\langle n\rangle$ \| 5.39(7) From $t_{\sigma,\pi}$ and $\alpha$: \| (in eV) \| Using $\langle L\rangle$ we obtain \| $t_{\sigma}^{\text{eq}}$ \| 3.0(2) \| $j=s(+l\approx 0)$ \| 2.3 $t_{\sigma}^{\text{ap}}$ \| 2.3(2) \| $g_{J}$ \| $\sim 2$ $t_{\pi}^{\text{eq}}$ \| $<0.5$ \| $\mu_{\text{eff}}$ \| $5.5(1)\mu_{\text{B}}$ $t_{\pi}^{\text{ap}}$ \| $<0.4$ \| and using $\sqrt{\langle L^{2}\rangle}$ \| \| \| $j=s+l$ \| 3.4 \| \| $g_{J}$ \| 1.67(5) \| \| $\mu_{\text{eff}}$ \| $6.51(7)\mu_{\text{B}}$ The single-ion anisotropy $D_{z}$ is obtained as the difference between the energy of the $S_{z}$ = $\pm 1/2$ and $\pm 5/2$ levels and was found to be of the easy-plane type. This easy-plane anisotropy was also confirmed by rotating the spin quantisation axis to various directions with respect to the crystal field. In this way we have also attempted to obtain estimates for $E_{xy}$. The $D_{4h}$ symmetry of the ligand field was lowered to $C_{2}$, to model the actual positions of the oxygen anions around the cation as obtained from crystallographical data Grohol:03 ; WillsKjar:00 . However, this had no effect on the agreement of the calculated spectra with experiment, nor on the values as listed in Table 1 and hence we consider an estimate for $E_{xy}$ from our data presently not attainable. It should in this respect be noted that the value for $D_{z}$ was found to change with the splitting of the main peak at 709 meV, introduced by the trigonal distortion of the ligand field. Its value is therefore tightly constrained by the experimentally obtained spectra and the 3d spin-orbit coupling strength. As is detailed in Table 1 our calculation indicates that for the isolated FeO6 cluster most of the orbital angular momentum of the Fe 3$d^{6}$ configuration is quenched by the ligand field, with $\langle L\rangle=0.008(3)$ while the maximum orbital angular momentum which could in this case be realised as $\sqrt{\langle L^{2}\rangle}=1.5(2)$. Due to the lower resolution at which the hydronium jarosite powder spectrum has been measured we cannot provide detailed information about this system, apart from a confirmation that the general crystal field is similar, with comparable charge-transfer and ($\sigma$-type) hybridization. This is a good indication that the magnetic exchange in hydronium jarosite is likely to be comparable with that of the other jarosites, i.e. with a Weiss temperature around $-$800 K. There is also a slight evidence that the $D_{4h}$ ligand- field anisotropy and hence the zero-field splitting, is slightly smaller in hydronium jarosite. ## V Discussion The expectation values of the ground-state obtained in our calculation are given in Table 1 (right-hand side). The corresponding free-ion configuration is $d^{5}$ ${}^{6}S_{5/2}$ with an additional electron weight of $\sim$0.4 electron that is mainly in the $d(x^{2}-y^{2})$ orbital. The spin-orbit coupling is much smaller than the $t_{\sigma}$ hopping bandwidth and hence the fraction of the electron transferred to the ligands to the Fe cation is only very weakly orbitally polarized, with $\langle L\rangle=0.008(3)$. This value is much smaller than the orbital angular momentum as inferred from high- temperature magnetic susceptibility data, which indicate that $\mu_{\text{eff}}\approx 6.5$ $\mu_{\text{B}}$ Grohol:05 . As shown in the lower panel of Table 1 (right-hand side) the experimental value is remarkably close to the case where the added electron weight due to charge transfer is fully orbitally polarized. It is not expected that in this $3d$ transition metal compound $l$ and $j(=s+l)$ are good quantum numbers, but there is sufficient evidence that $\langle L\rangle$ is significantly larger than zero 222Iron jarosite is in this sense not a unique case. For $3d$ transition metal compounds with more than half-filled shells (such as e.g. Herbertsmithite deVries:07 ) the effective moments are in general significantly larger than the spin-only moment.. It might be that comparison of experimental results with the Mott-Hubbard-like calculations as described here, but with larger clusters containing two or three iron centres, can provide an explanation for the discrepancy between the calculated and experimentally obtained orbital angular momentum. The energy of $D_{z}$ = 0.5(1) meV for the easy-plane single-ion anisotropy is in good agreement with the values found by Matan _et al._ Matan:06 (0.428(5) meV) and Coomer _et al._ Coomer:06 (0.47(2) meV) in their spin-wave analysis of inelastic neutron data from single crystals and powders, respectively. This implies that an easy-plane anisotropy, along with further neighbor interactions, can explain the magnetic ground state in potassium iron jarosite. Further evidence might be found in the temperature dependence of the experimental and calculated isotropic spectra. The spectra in Fig. 5 were taken at room temperature and were compared with the calculated spectra for the Boltzmann averaged configurations of the FeO6 cluster. As is visible in Fig. 3 the relative intensity of the two maxima in the split peak at 709 eV changes in a similar way as was observed in the experimentally obtained spectra. At the same time an (unexplained) increase in intensity of the peak at 707.8 eV peak relative to the intensity of the main peak is observed in the experimental spectra. We have not been able to account for the latter (which was also observed in powder samples) either by changing the temperature of the calculated spectra or with simulations with new values for $\delta_{pd}$, $t_{\sigma,\pi}$ and $\alpha$. The change in the main peak at 709 eV where we obtain good agreement, is a direct result of the spin-orbit coupling which causes a change in the shape of the electronic $3d$ orbital as the spins align in the easy plane. The change in shape of the main peak in the experimental spectra was found to occur between 200 K and 40 K. This is in agreement with previous experiments which have shown that in potassium jarosite the alignment of spin with the kagome planes sets in above 120 K, i.e. well above the transition temperature of 65 K Grohol:05 . The temperature dependence in the experimentally obtained spectra was simulated with the temperature (given in units of eV) used for the Boltzmann averaging over excited states as the only extra parameter, as shown in Fig. 3. All other simulation parameters were fixed to the values given in Table 1. In this way we found the expectation values for $\langle\|S_{z}\|\rangle$ at each corresponding experimental temperature, arising from the thermal population of the $m_{S}$ levels in the $S=5/2$ multiplet. The temperature dependence of $\langle\|S_{z}\|\rangle$ (Fig. 6) shows a gradual alignment of the spins into the kagome planes, at temperatures well above the transition temperature to a long-range ordered state at 64 K. This is in rough agreement with the degree of co-planarity of the spins as measured using neutron diffraction on large single crystals Grohol:05 . Figure 6: The expectation values $\langle\|S_{z}\|\rangle$ of the simulated spectra, against the temperature of the corresponding experimental spectrum. For a single $S=5/2$ spin the minimum for $\langle\|S_{z}\|\rangle$ is 0.5 and the maximum is 1.5 (equal population of all $m_{S}$. The gradual increase of $\langle\|S_{z}\|\rangle$ with temperature is in agreement with earlier neutron measurements which show a gradual alignment of the spins in the kagome plane at temperatures well above the transition temperature of 64 K Grohol:05 . The simulated linear dichroism spectra at the lowest temperatures, corresponding to states with the spins aligned in the equatorial plane (the states $\ket{s,m_{z^{\prime}}}=\ket{2.5,\pm 0.5}$) do not agree well the experimentally obtained spectra as shown in Fig. 4, although some features in the temperature dependence of the XLD are reproduced in the simulated spectra. Without success we have attempted to improve on this situation by calculating the spectra for the Fe spin aligned along the $x^{\prime}$ axis in the equatorial plane (i.e. $\ket{s,m_{x^{\prime}}}=\ket{2.5,\pm 2.5}$) which resembles more closely the actual ground state in potassium jarosite. The remaining disagreement could be in part due to experimental error due to sample imperfections and in part due to intrinsic limitations of the simulation. It might be naive to expect that the ligand-field can be inferred from crystallographic data alone. A first indication for this is the large value of $\alpha$ obtained here, which amplifies the effect of the crystallographic trigonal distortion of the FeO6 octahedra. Even when the symmetry of the ligand field is further lowered from $D_{4h}$ and accurately modelled on the locations of the oxygen anions around the central iron 333Though all Fe-O${}_{\text{eq}}$ bond lengths are the same, the equatorial plane is rectangular rather than square, with the short side pointing into the kagome triangles. Furthermore, the Fe-O${}_{\text{ap}}$ axis is canted by $4^{\circ}$ away from the local $c^{\prime}$ axis. neither the calculated spectra nor the simulation parameters change significantly. Therefore, it is possible that the actual charge anisotropy follows more closely the $xy$ plane rather than the equatorial oxygens. The easy-plane magnetic anisotropy of $D_{z}=0.5(1)$ meV found here (see Table 1), is then not directly related to the trigonal distortion of the crystal field but more to the 2D character of the electronic hopping and magnetic exchange, the virtual absence of $\pi$ symmetry hybridization and the Fe $3d$ spin-orbit coupling strength. In that case the easy plane could coincide more closely with the kagome plane. The strong 2D character can in part be attributed to the sulphate groups separating the kagome layers. The formal valence of the sulphur cation is 6+ with the [Ne] electronic configuration, but this large electrostatic charge is likely to attract electron weight from the neighboring oxygen anions, including the oxygens apical in the FeO6 octahedra. The hopping parameters $t_{\sigma,\pi}$ from these oxygens to the iron is then further reduced with respect to the in- plane hopping. This translates into the large value for $\alpha$ found here. The large difference between $t_{\pi}$ and $t_{\sigma}$ obtained here might be explained by the triangular arrangement of the neighboring FeO6 octahedra. This dramatically lowers the symmetry of the local crystal field around the oxygen ligands and reduces the $\pi$ symmetry overlap of the oxygen $2p$ orbitals with the iron $3d$ orbitals. Due to the low symmetry of the oxygen sites the symmetry of the crystal field can be expected to be even lower than indicated by the crystallographic positions of the anions. It is in this sense fortunate that the situation can be modelled so well by the anisotropic $\pi$ and $\sigma$ hopping whose ratio is equal for all six anions around the central iron. It could well be that the last differences between the calculated and experimental spectra, such as the low temperature x-ray linear dichroism, some of the difference in the overall intensity of the XLD and the missing bump just above 707 eV, could be brought into agreement by allowing a different balance between $t_{\pi}$ and $t_{\sigma}$ for the apical oxygen anions. In neutron spin-wave studies Matan:06 ; Yildirim:06 ; Coomer:06 the crystal field anisotropy scenario has been compared in close detail with that of DMI. In the present context such a discussion would require calculations on clusters containing two to three iron centres. Such a calculation would certainly be of great interest, but has not been attempted here, since the present approach already accounts so well for the experimental observations on potassium iron jarosite. The present result closes the last missing causal link, from electronic structure to collective magnetic ground state. ## VI Conclusion The isotropic and x-ray linear dichroism spectra taken at the Fe $L_{2,3}$ edges in potassium iron jarosite indicate that an orbital angular momentum and zero-field splitting arise due to an anisotropic, mainly $\sigma$ type, charge-transfer from the oxygen ligands. This closely resembles the scenario described in Ref. Wan-Lun:94, to explain the zero-field splitting as observed in Mn2+ salts. The zero-field splitting parameter of the crystal field anisotropy of 0.5 meV as found here is in excellent agreement with the value found from fits of a crystal field anisotropy model to the spin-wave spectrum in the ground state of potassium iron jarosite Matan:06 ; Yildirim:06 ; Coomer:06 . However, the most important result obtained here is that we have verified in detail the mechanism by which this magnetic anisotropy arises, closing the remaining missing link between the electronic structure in potassium iron jarosite and its magnetic ground state. We have even measured a change in the electronic structure as the magnetic ground state becomes co-planar, with the change of the isotropic line shape over the relevant temperature range. Of course, an additional in-plane anisotropy $E_{xy}$ and further- neighbor interactions are needed to fully explain the magnetic ground state. The present study does not provide more hints to the origin of the glassy state in hydronium jarosite. We found that the energy scale of the magnetic interaction in hydronium jarosite should be comparable to that in potassium jarosite. It might be that temperature dependent spectra taken at a higher resolution could provide new insights in the case of hydronium jarosite. A final interesting observation is that the maximum orbital angular momentum polarization possible for the amount of charge-transfer observed here, but not the actually calculated orbital polarization, is in rough agreement with the experimentally observed effective moment. ## VII Acknowledgement Simon Parsons and Clivia Hejny (Edinburgh University) are acknowledged for their help with the characterization and alignment of single crystals of potassium jarosite. We would like to thank Nicholas Harrison (Imperial College, London and Daresbury Laboratory) and Barry Searle (Daresbury Laboratory) for fruitful discussions about LSDFT calculations on potassium jarosite. Peter Bencok (ESRF and Diamond) and Nicola Farley (Daresbury Laboratory) are acknowledged for their help during x-ray spectroscopy measurements. We further thank Claudine Lacroix (CNRS, Grenoble), Andrew Wills (UCL), Frank de Groot (Utrecht University) and Paul Attfield (Edinburgh University) for fruitful discussions. MdV gratefully acknowledges financial support from the Centre for Materials Physics and Chemistry (Science and Technology Facility Council, UK) and the Highly Frustrated Magnetism network of the European Science Foundation. ## References * (1) Frustrated spin systems, Ed. H. T. Diep (World Scientific, Singapore, 2004). * (2) J. N. Reimers, A. J. Berlinsky, and A.-C. Shi, Phys. Rev. B 43, 865 (1991). * (3) J. T. Chalker, P. C. W. Holdsworth, and E. F. Shender, Phys. Rev. Lett. 68, 855 (1992). * (4) J. N. Reimers and A. J. Berlinsky, Phys. Rev. B 48, 9539 (1993). * (5) A. Keren, K. Kojima, L. P. Le, G. M. Luke, W. D. Wu, Y. J. Uemura, M. Takano, H. Dabkowska, and M. J. P. Gingras, Phys. Rev. B 53, 6451 (1996). * (6) T. Inami, T. Morimoto, M. Nishiyama, S. Maegawa, Y. Oka, and H. Okumura, Phys. Rev. B 64, 054421 (2001). * (7) T. Morimoto, M. Nishiyama, S. Maegawa, and Y. Oka, J. Phys. Soc. Jpn 72, 2085 (2003). * (8) M. Takano, T. Shinjo, M. Kiyama, and T. Takada, J. Phys. Soc. Jpn 25, 902 (1968). * (9) M. G. Townsend, G. Longworth, and E. Roudaut, Phys. Rev. B 33, 4919 (1986). * (10) A. S. Wills and A. Harrison, J Chem. Soc. Faraday Frans. 92, 2161 (1996). * (11) A. S. Wills, A. Harrison, C. Ritter, and R. I. Smith, Phys. Rev. B 61, 6156 (2000). * (12) A. S. Wills, Phys. Rev. B 63, 064430 (2001). * (13) D. Grohol, K. Matan, J. H. Cho, S. H. Lee, J. W. Lynn, D. G. Nocera, and Y. S. Lee, Nature Materials 4, 323 (2005). * (14) D. Grohol and D. Nocera, J. Am. Chem. Soc. 124, 2640 (2002). * (15) D. Grohol, Q. Huang, B. H. Toby, J. W. Lynn, Y. S. Lee, and D. G. Nocera, Phys. Rev. B 68, 094404 (2003). * (16) D. Grohol, D. G. Nocera, and D. Papoutsakis, Phys. Rev. B 67, 064401 (2003). * (17) T. Inami, M. Nishiyama, S. Maegawa, and Y. Oka, Phys. Rev. B 61, 12181 (2000). * (18) M. Nishiyama, S. Maegawa, T. Inami, and Y. Oka, Phys. Rev. B 67, 224435 (2003). * (19) A. P. Ramirez, S.-W. Cheong, and M. L. Kaplan, Phys. Rev. Lett. 72, 3108 (1994). * (20) N. Nishiyama and S. Maegawa, Physica B 329-333, 1065 (2003). * (21) T. Yildirim and A. B. Harris, Phys. Rev. B 73, 214446 (2006). * (22) M. Elhajal, B. Canals, and C. Lacroix, Phys. Rev. B 66, 014422 (2002). * (23) K. Matan, D. Grohol, D. G. Nocera, T. Yildirim, A. B. Harris, S. H. Lee, S. E. Nagler, and Y. S. Lee, Phys. Rev. Lett. 96, 247201 (2006). * (24) F. C. Coomer, A. Harrison, G. S. Oakley, J. Kulda, J. R. Stewart, J. A. Stride, B. Fåk, J. W. Taylor, and D. Visser, J. Phys.: Condens. Mat. 18, 8847 (2006). * (25) W. G. Bisson and A.S. Wills, to be published (2008). * (26) P. Mendels, F. Bert, M. A. de Vries, A. Olariu, A. Harrison, F. Duc, J. C. Trombe, J. S. Lord, A. Amato, and C. Baines, Phys. Rev. Lett. 98, 077204 (2007). * (27) A. M. Afanasev, V. D. Gorobchenko, D. S. Kulgawczuk, and I. I. Lukashevich, Phys. Status Solidi (A) 26, 697 (1974). * (28) P. Bonville et al., Hyperfine Interact. 168, 1085 (2006). * (29) T. Moriya, Phys. Rev. 120, 91 (1960). * (30) H. Watanabe, Progr. Theor. Phys. 18, 405 (1957). * (31) A. S. Wills, V. Dupuis, E. Vincent, J. Hammann, and R. Calemczuk, Phys. Rev. B 62, R9264 (2000). * (32) B. T. Thole, G. van der Laan, and P. H. Butler, Chem. Phys. Lett. 149, 295 (1988). * (33) J. Zaanen, G. A. Sawatzky, and J. W. Allen, Phys. Rev. Lett. 55, 418 (1985). * (34) J. M. D. Coey and G. A. Sawatzky, J. Phys. C: Solid St. Phys. 4, 2386 (1971). * (35) YuWan-Lun, Phys. Rev. B 39, 622 (1989). * (36) YuWan-Lun and TanTao, Phys. Rev. B 49, 3243 (1994). * (37) E. Francisco and L. Pueyo, Phys. Rev. B 37, 5278 (1988). * (38) P. Jeevanandam and S. Vasudevan, J. Phys.: Cond. Mat. 11, 3563 (1999). * (39) G. van der Laan, J. Zaanen, G. A. Sawatzky, R. Karnatak,, and J. M. Esteva, Phys. Rev. B 33, 4253 (1986). * (40) F. de Groot, J. El. Spec. Rel. Phen. 67, 529 (1994). * (41) F. de Groot, Coordination Chemistry Reviews 249, 31 (2005). * (42) J. E. Dutrizac and S. Kaiman, Canadian Mineralogist 14, 151 (1976). * (43) A. S. Wills, Ph.D. thesis, The University of Edinburgh, 1996. * (44) P. F. Schofield, G. van der Laan, C. M. B. Henderson, and G. Cressey, Mineralogical Magazine 62, 65 (1998). * (45) G. van der Laan and B. T. Thole, Phys. Rev. B 43, 13401 (1991). * (46) B. T. Thole, P. Carra, F. Sette, and G. van der Laan, Phys. Rev. Lett. 68, 1943 (1992). * (47) G. van der Laan, Phys. Rev. Lett. 82, 640 (1999). * (48) A. Mirone, M. Sacchi, and S. Gota, Phys. Rev. B 61, 13540 (2000). * (49) A. Mirone, Hilbert++ manual, arXiv:0706.4170v1, 2007. * (50) R. D. Cowan, The Theory of Atomic Structure and Spectra (University of California Press, Berkeley, 1981). * (51) G. van der Laan and I. W. Kirkman, J. Phys.: Condens. Mat. 4, 4189 (1992). * (52) B. T. Thole, G. van der Laan, J. C. Fuggle, G. A. Sawatzky, R. C. Karnatak and J.-M. Esteva, Phys. Rev. B 32, 5107 (1985). * (53) E. C. Wasinger,F. M. F. de Groot, B. Hedman, K. O. Hodgson, and E. I. Solomon, J. Am. Chem. Soc. 125, 12894 (2003). * (54) F. Ciccacci, L. Braicovich, E. Puppin, and E. Vescovo, Phys. Rev. B 44, 10444 (1991). * (55) M. A. de Vries, K. V.Kamenev, W. A. Kockelmann, J. Sanchez-Benitez, and A. Harrison, Phys. Rev. Lett. 100, 157205 (2008).	arxiv-papers	2008-08-14T10:01:42	2024-09-04T02:48:57.296338	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. A. de Vries, T. K. Johal, A. Mirone, J. S. Claydon, G. J. Nilsen,\n H. M. Ronnow, G. van der Laan and A. Harrison", "submitter": "Mark Vries de", "url": "https://arxiv.org/abs/0808.1969" }
0808.1995	# Local complementation rule for continuous-variable four-mode unweighted graph states Jing Zhang† State Key Laboratory of Quantum Optics and Quantum Optics Devices, Institute of Opto-Electronics, Shanxi University, Taiyuan 030006, P.R.China ###### Abstract The local complementation rule is applied for continuous-variable (CV) graph states in the paper, which is an elementary graph transformation rule and successive application of which generates the orbit of any graph states. The corresponding local Gaussian transformations of local complementation for four-mode unweighted graph states were found, which do not mirror the form of the local Clifford unitary of qubit exactly. This work is an important step to characterize the local Gaussian equivalence classes of CV graph states. Entanglement lies at the heart of quantum mechanics and plays a crucial role in quantum information processing. Recently, special types of multipartite entangled states, the so-called the graph states one ; two , have moved into the center of interest. A graph quantum state is described by a mathematical graph, i.e. a set of vertices connected by edges. A vertex represents a physical system, e. g. a qubit (2-dimensional Hilbert space), qudit (d-dimensional Hilbert space), or CV (continuous Hilbert space). An edge between two vertices represents the physical interaction between the corresponding systems. An interesting feature is that many entanglement properties of graph states are closely related to their underlying graphs. They not only provide an efficient model to study multiparticle entanglement one , but also find applications in quantum error correction three ; four , multi-party quantum communication five and most prominently, serve as the initial resource in one-way quantum computation six . Considerable efforts have been stepped toward generating and characterizing cluster state with linear optics experimentally seven ; eight ; nine ; ten . The principle feasibility of one-way quantum computing model has been experimentally demonstrated through photon cluster state successfully seven ; ten . Most of the concepts of quantum information and computation have been initially developed for discrete quantum variables, in particular two-level or spin-$\frac{1}{2}$ quantum variables (qubits). In parallel, quantum variables with a continuous spectrum, such as the position and momentum of a particle or amplitude and phase quadrature of an electromagnetic field, in informational or computational processes have attracted a lot of interest and appears to yield very promising perspectives concerning both experimental realizations and general theoretical insights eleven ; twelve , due to relative simplicity and high efficiency in the generation, manipulation, and detection of CV state. Although up to six-qubit single-photon cluster states have been created via postselection using nonlinear and linear optics, the deterministic, unconditional realization of optical cluster states would be based on continuous variables. CV cluster and graph states have been proposed thirteen , which can be generated by squeezed state and linear optics thirteen1 ; thirteen2 , and demonstrated experimentally for four-mode cluster state forteen . The one-way CV quantum computation was also proposed with CV cluster state fifteen . Moreover, the protocol of CV anyonic statistics implemented with CV graph states is proposed fifteen1 . One of the interesting issues on entanglement is how to define the equivalence of two entangled states. The transformations of qubit graph states under local Clifford operations were studied by Hein one and Van den Nest sixteen . They translate the action of local Clifford operations on qubit graph states into transformations on their associated graphs, that is, to derive transformations rules called the local complement rule, stated in purely graph theoretical terms, which completely characterize the evolution of graph states under local Clifford operations. The corresponding local Clifford unitary is a single and simple form. The successive application of this rule suffices to generate the complete orbit of any qubit graph state under local Clifford operations. In this paper, the local complement rule for CV four-mode unweighted graph state is applied and the corresponding local Clifford transformations (also called local Gaussian transformation for CV) for four-mode graph state were found. The local Gaussian equivalence classes of CV four-mode unweighted graph states can be obtained by this way. It was shown that the corresponding local Gaussian unitary can not exactly mirror that for qubit, which is not a single form compared with qubit. This result shows the complexity of CV quantum systems and stimulate the research on the local Gaussian equivalence of CV graph states. Although only focusing on the CV four-mode unweighted graph states, this work makes an important step in the direction of addressing the general question ”What are the graph transformation rules that describe local unitary equivalence of any CV graph states?”. The CV operations are reviewed firstly that follow the standard prescription given in Ref.seventeen . The Pauli $X$ and $Z$ operators of qubit are generalized to the Weyl-Heisenberg group, which is the group of phase-space displacements. For CVs, this is a Lie group with generators $\hat{x}=(\hat{a}+\hat{a}^{\dagger})/\sqrt{2}$ (quadrature-amplitude or position) and $\hat{p}=-i(\hat{a}-\hat{a}^{\dagger})/\sqrt{2}$ (quadrature- phase or momentum) of the electromagnetic field as the CV system. These operators satisfy the canonical commutation relation $[\hat{x},\hat{p}]=i$ (with $\hbar=1$). In analogy to the qubit Pauli operators, the single mode Pauli operators are defined as $X(s)=exp[-is\hat{p}]$ and $Z(t)=exp[it\hat{x}]$ with $s,t\in\mathbb{R}$. The Pauli operator $X(s)$ is a position-translation operator, which acts on the computational basis of position eigenstates $\\{\|q\rangle;q\in\mathbb{R}\\}$ as $X(s)\|q\rangle=\|q+s\rangle$, whereas $Z$ is a momentum-translation operator, which acts on the momentum eigenstates as $Z(t)\|p\rangle=\|p+t\rangle$. These operators are non-commutative and obey the identity $X(s)Z(t)=e^{-ist}Z(t)X(s)$. The Pauli operators for one mode can be used to construct a set of Pauli operators $\\{X_{i}(s_{i}),Z_{i}(t_{i});i=1,...,n\\}$ for n-mode systems. This set generates the Pauli group $\mathcal{C}_{1}$. The clifford group $\mathcal{C}_{2}$ is the normalizer of the Pauli group, whose transformations acting by conjugating, preserve the Pauli group $\mathcal{C}_{1}$; i.e., a gate U is in the Clifford group if $\emph{UR}\emph{U}^{-1}\in\mathcal{C}_{1}$ for every $\emph{R}\in\mathcal{C}_{1}$. The clifford group $\mathcal{C}_{2}$ for CV is shown seventeen to be the (semidirect) product of the Pauli group and linear symplectic group of all one-mode and two-mode squeezing transformations. Transformation between the position and momentum basis is given by the Fourier transform operator $F=exp[i(\pi/4)(\hat{x}^{2}+\hat{p}^{2})]$, with $F\|q\rangle_{x}=\|q\rangle_{p}$. The action $FRF^{-1}$ of the Fourier transform on the Pauli operators is $\displaystyle F:X(s)$ $\displaystyle\rightarrow$ $\displaystyle Z(s),$ $\displaystyle Z(t)$ $\displaystyle\rightarrow$ $\displaystyle X(-t).$ (1) This is the generalization of the Hadamard gate for qubits. The phase gate $P(\eta)=exp[i(\eta/2)\hat{x}^{2}]$ with $\eta\in\mathbb{R}$ is a squeezing operation for CV and the action on the Pauli operators is $\displaystyle P(\eta):X(s)$ $\displaystyle\rightarrow$ $\displaystyle e^{-is^{2}\eta/2}Z(s\eta)X(s),$ $\displaystyle Z(t)$ $\displaystyle\rightarrow$ $\displaystyle Z(t),$ (2) in analogy to the phase gate of qubit eighteen . The controlled operation C-Z is generalized to controlled-$Z(C_{Z})$. This gate $C_{Z}=exp[i\hat{x}_{1}\bigotimes\hat{x}_{2}]$ provides the basic interaction for two mode 1 and 2, and describes the quantum nondemolition (QND) interaction. This set $\\{X(s),F,P(\eta),C-Z;s,\eta\in\mathbb{R}\\}$ generates the Clifford group. Transformations in the Clifford group do not form a universal set of gates for CV quantum computation. However, Clifford group transformation (Gaussian transformations) together with any higher-order nonlinear transformation (non-Gaussian transformation) acting on a single-mode form a universal set of gates seventeen . The local Gaussian group only was concerned here, which can be obtained by repeated application of Fourier and phase gates. In the following, another type of the phase gate will be used $P_{X}(\eta)=FP(\eta)F^{-1}=exp[i(\eta/2)\hat{p}^{2}]$ and the action on the Pauli operators is $\displaystyle P_{X}(\eta):X(s)$ $\displaystyle\rightarrow$ $\displaystyle X(s),$ $\displaystyle Z(t)$ $\displaystyle\rightarrow$ $\displaystyle e^{-it^{2}\eta/2}X(-t\eta)Z(t),$ (3) where $P_{X}(\eta)^{\dagger}=P_{X}(\eta)^{-1}=P_{X}(-\eta)$. A graph quantum state is described by a mathematical graph $G=(V,E)$, i.e. a finite set of $n$ vertices $V$ connected by a set of edges $E$ ninteen . An $\left\\{a,c\right\\}$-path is a order list of vertices $a=a_{1},a_{2},\ldots,a_{n-1},a_{n}=c$, such that for all $i$, $a_{i}$ and $a_{i+1}$ are adjacent. A connected graph is a graph that has an $\left\\{a,c\right\\}$-path for any two $a,c\in V$. Otherwise it is referred to as disconnected. The neighborhood $N_{a}\subset V$ is defined as the set of vertices $b$ for which $\left\\{a,b\right\\}\in E$. When a vertex a is deleted in a graph G, together with all edges incident with a, one obtains a new graph, denoted by $G-a$. For a subset of vertices $U\subset V$ of a graph $G=(V,E)$ let us denote with $G-U$ the graph that is obtained from $G$ by deleting the set $U$ of vertices and all edges which are incident with an element of $U$. Similarly, an subgraph $G[C]$ of a graph $G=(V,E)$, where $C\subset V$, is obtained by deleting all vertices and the incident edges that are not contained in $C$. The preparation procedure of CV graph states thirteen can exactly mirror that for qubit graph states only using the Clifford operations: first, prepare each mode (or graph vertex) in a phase- squeezed state, approximating a zero-phase eigenstate (analog of Pauli-X eigenstates), then, apply a QND interaction (C-Z gate) to each pair of modes $(j,k)$ linked by an edge in the graph. All C-Z gates commute. Thus, the resulting CV graph state becomes, in the limit of infinite squeezing, $g_{a}=(\hat{p}_{a}-\sum_{b\in N_{a}}\hat{x}_{b})\rightarrow 0$, where the modes $a\in V$ correspond to the vertices of the graph of $n$ modes, while the modes $b\in N_{a}$ are the nearest neighbors of mode $a$. This relation is as a simultaneous zero-eigenstate of the position-momentum linear combination operators. The stabilizers $G_{a}(\xi)=exp[-i\xi g_{a}]=X_{a}(\xi)\prod_{b\in N_{a}}Z_{b}(\xi)$ with $\xi\in\mathbb{R}$ for CV graph states are analogous to $n$ independent stabilizers $G_{a}=X_{a}\prod_{b\in N_{a}}Z_{b}$ for qubit graph states. Note that the CV graph states that is discussed here are unweighted since the QND interactions all have the same strength. For the CV weighted graph states generated by the different QND interaction strength, the stabilizers become $G_{a}(\xi)=X_{a}(\xi)\prod_{b\in N_{a}}Z_{b}(\Omega_{ab}\xi)$, where $\Omega_{ab}$ is the interaction strength between mode a and b. The CV weighted graph states are more complex, which is not considered in this paper. Figure 1: The connected four-vertex graphs for an successive application of the local complementation. The rule is successively applied to the vertex, which is circle in the figure. The action of the local complement rule, can be described as: letting $G=(V,E)$ be a graph and $a\in V$ be a vertex, the local complement of $G$ for $a$, denoted by $\lambda_{a}(G)$, is obtained by complementing the subgraph of $G$ generated by the neighborhood $N_{a}$ of $a$ and leaving the rest of the graph unchanged. The successive application of this rule suffices to generate the complete orbit of any graph. Here, the corresponding local Gaussian unitary for CV four-mode graph state were examined. The corresponding four- mode graph state $\|\lambda_{a}(G)\rangle$ by local complement of a graph $G$ at some vertex $a\in V$, is given by a local Gaussian unitary operation $\displaystyle\|\lambda_{a}(G)\rangle=U_{\lambda_{a}}\|G\rangle,$ (4) where $U_{\lambda_{a}}$ is local Gaussian operation. A form of the local Gaussian unitary comprising two types of phase gate is defined $\displaystyle U_{LG_{a}}=P_{Xa}(1)\prod_{b\in N_{a}}P_{b}(-1),$ (5) which mirrors the form of qubit local Clifford operation for local complementation. Fig.1 depicts connected four-mode graphs by such a successive application of the local complement rule. The four independent stabilizers of the first graph state $\|G^{(1)}\rangle$ are given by $\displaystyle G_{1}^{(1)}(\xi)$ $\displaystyle=$ $\displaystyle X_{1}(\xi)Z_{2}(\xi),$ $\displaystyle G_{2}^{(1)}(\xi)$ $\displaystyle=$ $\displaystyle X_{2}(\xi)Z_{1}(\xi)Z_{3}(\xi),$ $\displaystyle G_{3}^{(1)}(\xi)$ $\displaystyle=$ $\displaystyle X_{3}(\xi)Z_{2}(\xi)Z_{4}(\xi),$ $\displaystyle G_{4}^{(1)}(\xi)$ $\displaystyle=$ $\displaystyle X_{4}(\xi)Z_{3}(\xi).$ (6) with $G_{i}^{(1)}(\xi)\|G^{(1)}\rangle=\|G^{(1)}\rangle$ in the limit of infinite squeezing, where $i=1,...,4$. Applying the local Gaussian unitary $U_{LG_{3}}$ to the vertex 3, I can compute the four independent stabilizers of the resulting graph state $\|G^{(2)}\rangle$ by Eqs. (2,3,4,5), for example calculating $G_{2}^{(2)}(\xi)$, $\displaystyle\|G^{(2)}\rangle$ $\displaystyle=$ $\displaystyle\|\lambda_{3}(G^{(1)})\rangle$ (7) $\displaystyle=$ $\displaystyle U_{LG_{3}}G_{2}^{(1)}(\xi)\|G^{(1)}\rangle$ $\displaystyle=$ $\displaystyle U_{LG_{3}}G_{2}^{(1)}(\xi)U_{LG_{3}}^{-1}U_{LG_{3}}\|G^{(1)}\rangle$ $\displaystyle=$ $\displaystyle[e^{i\xi^{2}/2}Z_{2}(-\xi)X_{2}(\xi)]Z_{1}(\xi)$ $\displaystyle[e^{-i\xi^{2}/2}X_{3}(-\xi)Z_{3}(\xi)]U_{LG_{3}}\|G^{(1)}\rangle$ $\displaystyle=$ $\displaystyle X_{2}(\xi)Z_{1}(\xi)Z_{3}(\xi)U_{LG_{3}}$ $\displaystyle[Z_{2}(-\xi)X_{3}(-\xi)]\|G^{(1)}\rangle$ $\displaystyle=$ $\displaystyle X_{2}(\xi)Z_{1}(\xi)Z_{3}(\xi)U_{LG_{3}}$ $\displaystyle[Z_{2}(-\xi)X_{3}(-\xi)]G_{3}^{(1)}(\xi)\|G^{(1)}\rangle$ $\displaystyle=$ $\displaystyle X_{2}(\xi)Z_{1}(\xi)Z_{3}(\xi)Z_{4}(\xi)\|\lambda_{3}(G^{(1)})\rangle$ $\displaystyle=$ $\displaystyle G_{2}^{(2)}(\xi)\|G^{(2)}\rangle$ to obtain $\displaystyle G_{1}^{(2)}(\xi)$ $\displaystyle=$ $\displaystyle X_{1}(\xi)Z_{2}(\xi),$ $\displaystyle G_{2}^{(2)}(\xi)$ $\displaystyle=$ $\displaystyle X_{2}(\xi)Z_{1}(\xi)Z_{3}(\xi)Z_{4}(\xi),$ $\displaystyle G_{3}^{(2)}(\xi)$ $\displaystyle=$ $\displaystyle X_{3}(\xi)Z_{2}(\xi)Z_{4}(\xi),$ $\displaystyle G_{4}^{(2)}(\xi)$ $\displaystyle=$ $\displaystyle X_{4}(\xi)Z_{2}(\xi)Z_{3}(\xi),$ (8) which exactly correspond to the stabilizers of No.2 graph state in Fig.1. The complete orbit of the first graph can be obtained by applying the local complement rule repeatedly to the vertices and the corresponding local Gaussian unitary is shown in the following forms: $No.1\stackrel{{\scriptstyle U_{LG_{3}}}}{{\longrightarrow}}No.2\stackrel{{\scriptstyle U_{LG_{3}}^{2}F^{2}_{1}U_{LG_{2}}^{\dagger}}}{{\longrightarrow}}No.3\stackrel{{\scriptstyle U_{LG_{3}}^{\dagger}}}{{\longrightarrow}}No.4\stackrel{{\scriptstyle U_{LG_{1}}}}{{\longrightarrow}}No.5\stackrel{{\scriptstyle U_{LG_{2}}^{2}F^{2}_{1}U_{LG_{3}}^{\dagger}}}{{\longrightarrow}}No.6\stackrel{{\scriptstyle U_{LG_{1}}^{\dagger}}}{{\longrightarrow}}No.7\stackrel{{\scriptstyle U_{LG_{3}}}}{{\longrightarrow}}No.8\stackrel{{\scriptstyle U_{LG_{4}}^{\dagger}}}{{\longrightarrow}}No.9\stackrel{{\scriptstyle U_{LG_{1}}}}{{\longrightarrow}}No.10\stackrel{{\scriptstyle U_{LG_{2}}^{\dagger}}}{{\longrightarrow}}No.11$. Here the complete orbit means the local complement rule is applied on the graph until exhaust all possibilities. Figure 2: The set of four-vertex graphs is equivalent to Fig.1 under local Gaussian transformation and graph isomorphisms. The graph No.7, which is repeated and placed in the dash-line box behind the No.10, is used for generating the graph No.11 directly. Notice the difference in the Gaussian operations of $2\rightarrow 3$, and $5\rightarrow 6$. In the qubit case, these would have been of identical form. This shows the added richness of CV graph states over their qubit counterparts. Note that Hein et al. one classify the equivalence of the graph states by considering the local complementation and additional graph isomorphisms, which corresponds to the permutations of the vertices. Fig.2 shows another set of graphs, which are not equivalent to any graph in the equivalence class represented in Fig.1 only considering the local complementation. However, they belong to the same equivalence class when considering both, local Gaussian unitary and graph isomorphisms. The corresponding local Gaussian unitary in Fig.2 is shown in the following forms: $No.1\stackrel{{\scriptstyle U_{LG_{1}}}}{{\longrightarrow}}No.2\stackrel{{\scriptstyle U_{LG_{3}}^{2}F^{2}_{1}U_{LG_{2}}^{\dagger}}}{{\longrightarrow}}No.3\stackrel{{\scriptstyle U_{LG_{1}}^{\dagger}}}{{\longrightarrow}}No.4\stackrel{{\scriptstyle U_{LG_{2}}}}{{\longrightarrow}}No.5\stackrel{{\scriptstyle U_{LG_{2}}^{2}F^{2}_{3}U_{LG_{1}}^{\dagger}}}{{\longrightarrow}}No.6\stackrel{{\scriptstyle U_{LG_{4}}^{2}F^{2}_{2}U_{LG_{3}}^{\dagger}}}{{\longrightarrow}}No.7\stackrel{{\scriptstyle U_{LG_{4}}^{\dagger}}}{{\longrightarrow}}No.8\stackrel{{\scriptstyle U_{LG_{3}}}}{{\longrightarrow}}No.9\stackrel{{\scriptstyle U_{LG_{1}}^{\dagger}}}{{\longrightarrow}}No.10;$ $No.7\stackrel{{\scriptstyle U_{LG_{2}}^{\dagger}}}{{\longrightarrow}}No.11$. Figure 3: The set of four-vertex graphs is not equivalent to Fig.1 and 2 under local Gaussian transformation and graph isomorphisms. The graph No.1, which is placed in the dash-line box, is used repeatedly by the local complementation. The set of graphs in Fig.3, usually called GHZ (Greenberger-Horne-Zeilinger) entangled states, is not equivalent with Fig.1 and 2 under local Gaussian transformation and graph isomorphisms. The local Gaussian unitary is applied to four-mode graph states in Fig.3, which is written above the arrows of the following diagram: $No.1\stackrel{{\scriptstyle U_{LG_{1}}^{\dagger}}}{{\longrightarrow}}No.2\stackrel{{\scriptstyle U_{LG_{1}}}}{{\longrightarrow}}No.1\stackrel{{\scriptstyle U_{LG_{2}}^{\dagger}}}{{\longrightarrow}}No.3\stackrel{{\scriptstyle U_{LG_{2}}}}{{\longrightarrow}}No.1\stackrel{{\scriptstyle U_{LG_{3}}^{\dagger}}}{{\longrightarrow}}No.4\stackrel{{\scriptstyle U_{LG_{3}}}}{{\longrightarrow}}No.1\stackrel{{\scriptstyle U_{LG_{4}}^{\dagger}}}{{\longrightarrow}}No.5$. Fig. 4 lists the graphs with up to four vertices that are not equivalent under local Gaussian transformation and graph isomorphisms. Figure 4: The connected graphs with up to four vertices are not equivalent under local Gaussian transformation and graph isomorphisms. In summary, the local complement rule was extended for CV graph states and the corresponding local Gaussian transformations of four-mode unweighted graph states were given. Thus the local Gaussian equivalence classes of CV four-mode unweighted graph states can be obtained. It was shown that the corresponding local Clifford unitary can not exactly mirror that for qubit and demonstrate the complexity of CV quantum systems. It is worth remarking that, whether the local complementation for any CV graph states can be implemented completely by the local Gaussian transformations and the general form of the corresponding local Gaussian unitary can be found, still need be further investigated. This work not only contribute to a deeper and more complete understanding of CV multipartite entanglement, but also stimulate the research on CV graph states theoretically and experimentally. †Corresponding author’s email address: jzhang74@sxu.edu.cn, jzhang74@yahoo.com ## I ACKNOWLEDGMENTS 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), NSFC (Grant No. 60678029), Program for the Top Young and Middle-aged Innovative Talents of Higher Learning Institutions of Shanxi and NSF of Shanxi Province (Grant No. 2006011003). ## II Reference ## References * (1) M. Hein, et al., Phys. Rev. A 69, 062311 (2003). * (2) H. J. Briegel and R. Raussendorf, Phys. Rev. Lett. 86, 910 (2001). * (3) D. Schlingemann, R. F. Werner, Phys. Rev. A 65, 012308 (2002). * (4) S. Yu, Q. Chen, and C. H. Oh, quant-ph:0709.1780. * (5) R. Cleve, et al., Phys. Rev. Lett. 83, 648 (1999). * (6) R. Raussendorf and H. J. Briegel, Phys. Rev. Lett. 86, 5188 (2001); R. Raussendorf, et al., Phys. Rev. A 68, 022312 (2003). * (7) P. Walther et al., Nature (London) 434, 169 (2005); * (8) N. Kiesel et al., Phys. Rev. Lett. 95, 210502 (2005); * (9) C. Y. Lu et al., Nature Physics 3, 91-95 (2007). * (10) R. Prevedel et al., Nature (London) 445, 65 (2007). * (11) S. L. Braunstein and A. K. Pati, _Quantum Information with Continuous Variables_ (Kluwer Academic, Dordrecht, 2003). * (12) S. L. Braunstein, P. van Loock, Rev. Mod. Phys. 77, 513 (2005). * (13) J. Zhang, S. L. Braunstein, Phys. Rev. A 73, 032318 (2006). * (14) P. van Loock, C. Weedbrook and M. Gu, Phys. Rev. A 76, 032321 (2007). * (15) N. C. Menicucci et al.,Phys. Rev. A 76, 010302 (2007). * (16) X. Su, et al., Phys. Rev. Lett. 98, 070502 (2007); M. Yukawa, et al., arXiv: 0804.0289 * (17) N. C. Menicucci et al., Phys. Rev. Lett. 97, 110501 (2006); P. van Loock, J. Opt. Soc. Am. B 24, 340 (2007). * (18) J. Zhang, C. Xie, and K. Peng, quant-ph/0711.0820. * (19) M. Van den Nest, J. Dehaene, and B. De Moor, Phys. Rev. A 69, 022316 (2004). * (20) S. D. Bartlett, et al., Phys. Rev. Lett. 88, 097904 (2002). * (21) D. Gottesman, et al., Phys. Rev. A 64, 012310 (2001). * (22) M. Hein, et al., quant-ph/0602096.	arxiv-papers	2008-08-14T14:17:12	2024-09-04T02:48:57.301606	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Jing Zhang", "submitter": "Jing Zhang", "url": "https://arxiv.org/abs/0808.1995" }
0808.2057	# Weighing matrices and optical quantum computing Steven T. Flammia1 and Simone Severini2 1Perimeter Institute for Theoretical Physics, 31 Caroline St. N, Waterloo, Ontario, N2L 2Y5 Canada 2Institute for Quantum Computing and Department of Combinatorics & Optimization, University of Waterloo, 200 University Ave. W, Waterloo, Ontario, N2L 3G1 Canada sflammia@perimeterinstitute.ca simoseve@gmail.com , ###### Abstract Quantum computation in the one-way model requires the preparation of certain resource states known as cluster states. We describe how the construction of continuous-variable cluster states for optical quantum computing relate to the existence of certain families of matrices. The relevant matrices are known as weighing matrices, with a few additional constraints. We prove some results regarding the structure of these matrices, and their associated graphs. ###### pacs: 02.10.Ox, 03.67.Lx, 42.50.Ex ## 1 Introduction In the standard model of quantum computing [1], the computation proceeds by coherently performing unitary dynamics on a simple initial state of a quantum system before being read out by a sequence of local measurements. By contrast, the one-way model of quantum computing [2] eliminates the need for coherent quantum evolution, substituting instead a highly entangled, but easily prepared initial state as the resource for the computation. The typical resource state used is the cluster state [3], whose preparation consists of, for example, coupling a system of qubits with a nearest-neighbor Ising interaction. The computation proceeds by a sequence of local measurements on the cluster state and the basis chosen at each step determines the computation. In this way, a cluster state is like a quantum breadboard on which the quantum circuit is inscribed via measurements. Although cluster states were originally defined using qubits, they generalize to $d$-level quantum systems [4] and continuous variables [5, 6]. Recently, Ref. [7] proposed a method for efficiently generating continuous-variable cluster states (CVCS). The method relies on pumping an optical cavity containing a special nonlinear medium at certain carefully chosen frequencies. The resulting CVCS is encoded in the quadratures of the photons that emerge from the cavity. The method of CVCS generation proposed in [7] required, on grounds of experimental and theoretical tractability, the existence of certain families of matrices. The required matrices are weighing matrices, defined as matrices $W$ whose entries are in $\\{0,\pm 1\\}$ and satisfy $WW^{T}=kI$, where $k$ is called the _weight_. The weighing matrices were obliged to have _Hankel_ form, meaning that the skew-diagonals of the matrix were all constant, and any skew- diagonal with support on the main diagonal must be identically 0. A matrix with vanishing main diagonal is said to be _hollow_. As we show later, when viewing the matrices as adjacency matrices for an associated graph, this is implied when the graph is connected and bipartite. In this article, we investigate the mathematical structure behind the construction in [7] and begin to classify the associated matrices. The discussion is organized as follows. In section 2 we give a brief glimpse of the physics which motivates the restrictions given in the previous paragraph. We begin the mathematical discussion in section 3, where we discuss anticirculant graphs, and section 4, where we discuss anticirculant weighing matrices, a special case of the Hankel weighing matrices. In section 5, we complete the rather trivial classification of the case of weight 2, while section 6 provides several examples about the case of weight 4. We discuss a number of open problems in section 7, and conclude in section 8. ## 2 Quantum computing in the frequency comb In this section we outline the physics that naturally leads us to consider Hankel hollow weighing matrices. For a more detailed discussion of the physics, see [7, 8, 9, 10]. The resonant frequencies of an optical cavity are defined classically as the modes which constructively interfere inside the cavity. Because they are evenly spaced, they form a so-called a frequency comb. By placing a nonlinear medium inside the cavity, pump photons with frequency $\omega_{p}$ can downconvert into an entangled pair of photons, so long as the new photons’ frequencies satisfy an energy conservation constraint, $\omega_{p}=\omega_{m}+\omega_{n}\,,$ (1) where $\omega_{m}$ and $\omega_{n}$ are the $m^{\rm th}$ and $n^{\rm th}$ resonant frequencies of the cavity. The range over which the downconversion can occur is limited by the phasematching bandwidth of the cavity. More generally, multiple frequencies can be simultaneously phasematched inside the cavity, and upconversion is allowed as well as downconversion. The total Hamiltonian is then idealized by the following $\mathcal{H}=i\hbar\kappa\sum_{p\in P}\ \sum_{m+n=p}\\!M_{mn}(\hat{a}^{\dagger}_{m}\hat{a}^{\dagger}_{n}-\hat{a}^{\phantom{{\dagger}}}_{m}\hat{a}^{\phantom{{\dagger}}}_{n}),$ (2) where $P$ is the (discrete) spectrum of the polychromatic pump, $\kappa$ is a global coupling strength, $\hat{a}^{\dagger}_{n}$ is the creation operator for the $n^{\rm th}$ mode, and $M_{mn}$ are matrix elements of a symmetric matrix. In units of the fundamental frequency of the cavity, the restriction of the inner sum to terms with $m+n=p$ enforces photon energy conservation, and the magnitude of $M_{mn}$ determines the strength of the coupling in units of $\hbar\kappa$, while the sign determines whether the photons are upconverted or downconverted. If all of the coupling strengths are equal, then $M$ is a Hankel matrix whose elements are either $0$ or $\pm c$, where $c$ is a constant. Experimental simplicity further demands that there is no single-mode squeezing and hence we require that the elements along the main diagonal of $M$ are all zero, i.e. $M$ must be a hollow matrix. The Hamiltonian (2) can be used to create continuous-variable cluster states, defined as Gaussian states satisfying the relation $\mathbf{p}-A\mathbf{q}\rightarrow\mathbf{0}\;,$ (3) where $\mathbf{p}$ and $\mathbf{q}$ are vectors of amplitude and phase quadratures for the $k^{\rm th}$ cavity mode, $q_{k}=\hat{a}^{\dagger}_{k}+\hat{a}^{\phantom{{\dagger}}}_{k}$ and $p_{k}=i(\hat{a}_{k}^{\dagger}-\hat{a}^{\phantom{{\dagger}}}_{k})$, $A$ is the symmetric weighted adjacency matrix of the graph of the CVCS, and the arrow denotes the limit of large squeezing. Reference [8] first proved a nonunique relationship between the adjacency matrix $A$, which describes the cluster state, and the adjacency matrix $M$, which describes the coupling between the modes inside the cavity. The general relationship is somewhat complicated, so references [9, 7, 10] introduced the simplifying ansatz that $AA^{T}=A^{2}=1,$ (4) from which it follows [9, 10] that (up to a trivial relabeling) $M=A.$ (5) In short, assuming $A$ is an orthogonal matrix implies the graph describing the couplings between photons in the cavity ($M$) is _identical_ to the graph describing the cluster state ($A$), which is a vast simplification over the general relationship. The derivation requires that $M$ be the adjacency matrix of a bipartite graph, but as we will see, one can derive this from the Hankel and hollow constraints. We now see how the twin demands of experimental and theoretical simplicity for creating CVCS in an optical cavity naturally lead us to consider Hankel hollow orthogonal matrices, all of whose non-zero elements are either $\pm c$, for some constant $c$. We can of course consider renormalized matrices instead by dividing out $c$, so that all entries are $\pm 1$; after finding these renormalized matrices, we can then reintroduce the constant afterwards so that (4) holds. Thus, we see that finding and classifying Hankel hollow weighing matrices is our primary interest. For completeness, we mention one further elaboration that is possible for this scheme. If the photons’ polarization, transverse, or spatial degrees of freedom are taken into account, then the Hamiltonian (2) can be modified by adding additional indices for these modes. If these additional modes are all frequency-degenerate, then an additional sum over these degenerate modes appears in (2), which allows for the symmetric intercoupling between the degenerate degrees of freedom. If such an interaction could be simultaneously phasematched by the nonlinear medium inside the cavity, then one is naturally lead to consider _block_ -Hankel matrices, where the size of the blocks is equal to the number of degenerate degrees of freedom. We leave the consideration of block-Hankel hollow weighing matrices to future work. ## 3 Anticirculant graphs In this section we introduce the definition of anticirculant graphs and study some of their first properties. For the sake of self-containedness, we will define here all the graph-theoretic concepts we require. For further background on theory of graphs the interested reader is referred to the book by Diestel [11]. Our reference about permutations and finite groups is Cameron [12]. We will work with simple graphs. A (simple) _graph_ $G=(V,E)$ is an ordered pair of sets defined as follows: $V(G)$ is a non-empty set, whose elements are called _vertices_ ; $E(G)$ is a non-empty set of unordered pairs of vertices, whose elements are called _edges_. An edge of the form $\\{i,i\\}$ is called a _loop_. Two vertices $i$ and $j$ are said to be _adjacent_ if $\\{i,j\\}\in E(G)$. Often, this is simply denoted by writing $ij$. A graph is _loopless_ if it is has no loops. The _adjacency matrix_ of a graph $G$ is denoted by $A(G)$ and defined by $[A(G)]_{i,j}:=\left\\{\begin{tabular}[]{ll}$1,$&if $ij\in E(G);$\\\ $0,$&if $ij\notin E(G).$\end{tabular}\right.$ An adjacency matrix of a loopless graph is said to be _hollow_. The _degree_ of a vertex $i$ in a graph $G$ is the number of edges incident with the vertex, that is, $\\#\\{j:ij\in E(G)\\}$. A graph is _regular_ if each of its vertices has the same degree. We say $d$_-regular_ to specify that each vertex has degree $d$. A square matrix is _Hankel_ if it has constant skew-diagonals. These are diagonals that traverse the matrix from North-East to South-West. (For this reason, these are also called _antidiagonals_). Notice that an $n\times n$ matrix has $2n-1$ skew-diagonals. An $n\times n$ Hankel matrix is said to be _anticirculant_(or, equivalently, _skew-circulant_ or _backcirculant_) if the $i$-th and $(i+n)$-th skew diagonals are equal. A _permutation_ of _length_ $n$ is a bijection $\pi:[n]\longrightarrow[n]$, where $[n]=\\{1,2,...,n\\}$. A _permutation matrix_ of _dimension_ $n$ is a $n\times n$ matrix $P$ with the following two properties: $[P]_{i,j}\in\\{0,1\\}$ for every $i,j$; $P$ has a unique $1$ in each row and in each column. A permutation $\pi$ is said to _induce_ a permutation matrix $P$ if $[P]_{i,\pi(i)}=1$ for every $i$. Two graphs $G$ and $H$ are _isomorphic_ if there is a permutation matrix $P$ such that $A(G)=PA(H)P^{-1}$. We write $G\cong H$ to denote that graphs $G$ and $H$ are isomorphic. A graph $G$ is _bipartite_ if there is a permutation matrix $P$ such that $PA(G)P^{-1}=\left(\begin{array}[]{cc}0&M\\\ M^{T}&0\end{array}\right),$ (6) for some matrix $M$. If $M=J_{n}$, where $J_{n}$ is the all-ones matrix of size $n$, then $G\cong K_{n,n}$, the _complete bipartite graph_. A graph is said to be _anticirculant_ if it is isomorphic to a graph whose adjacency matrix is anticirculant. Let us denote by $\mathcal{L}_{n,d}$ and by $\mathcal{L}_{n}$ the set of all $d$-regular anticirculant graphs on $n$ vertices and the set of anticirculant graphs on $n$ vertices, respectively. Notice that $\\#\mathcal{L}_{n,d}={n\choose d}$. Therefore, $\\#\mathcal{L}_{n}=\sum\nolimits_{d=1}^{n-1}\\#\mathcal{L}_{n,d}=\sum\nolimits_{d=1}^{n-1}{n\choose d}=2^{n}-2.$ Given $G\in\mathcal{L}_{n}$, the ordered set $S(G)=(s:[A(G)]_{1,s}=1)$ is said to be the _symbol_ of $G$, in analogy with the terminology used for circulant graphs. Indeed, $S(G)$ specifies $G$ completely. In cycle notation, each element $s\in S(G)$ is associated to a permutation of the form $\pi_{s}=(1,s)(2,s-1)\cdots(s+1,n)(s+2,n-1)\cdots,$ (7) where the subtraction is modulo $n$. In line notation, $\pi_{s}=(s)(s-1)(s-2)\cdots(n)(n-1)\cdots(s+1)$. On the basis of the definitions, we can collect the facts below: ###### Proposition 1 Let $G\in\mathcal{L}_{n}$ be a graph. The following statements hold true: 1. 1. The graph $G$ is $d$-regular with $d=\\#S(G)$. 2. 2. If $G$ is loopless then $n$ is even. 3. 3. If $G$ is loopless then each $s\in S(G)$ is even. 4. 4. The graph $G$ is bipartite. Proof. In order: 1. 1. Let $\pi_{s}$ be the permutation associated to the element $s\in S(G)$. Then $\pi_{s}$ induces a symmetric permutation matrix $P_{s}$. Since each $P_{s}$ is defined by $s$, we have $[P_{s}]_{i,j}=1$ if and only if $[P_{t}]_{i,j}=0$ for all $t\not=s$. Consequently, $A(G)=\sum_{s\in S(G)}P_{s}$, and $G$ is a regular graph with degree $d=\\#S$. 2. 2. If the size of $P_{s}$ is odd, then $[P_{s}]_{i,i}=1$ for some $i$. This is because $\pi_{s}$ is necessarily an involution of odd length, and for this reason it has at least one fixed point. (Recall that an _involution_ is a permutation that is its own inverse.) 3. 3. Suppose on the contrary that $s=2k+1$, for some $k\geq 1$. Then, $[A(G)]_{1,s}=[A(G)]_{1,2k+1}=[A(G)]_{k+1,k+1}=1$, contradicting the hypothesis of $G$ being loopless. 4. 4. By the form of each $\pi_{s}$ (see Eq. 7), it is sufficient to observe that vertices labeled by odd numbers are adjacent only to vertices labeled by even numbers and _viz_. At this stage, we are ready to prove that the members of $\mathcal{L}_{n}$ are Cayley graphs. The _Cayley graph_ $G=G(\Gamma,T)$ of a group $\Gamma$ _w.r.t._ the set $T\subseteq\Gamma$ is the graph in which $V(G)=\\{\Gamma\\}$ and $\\{g,h\\}\in E(G)$, if there is $s\in T$ such that $gs=h$. The dihedral group $D_{2k}$ of order $2k$ is the group of symmetries of a regular $k$-gon. The dihedral group $D_{2k}$ is nonabelian and presented as $D_{2k}=\langle s,t:t^{k}=e,s^{2}=e,sts=t^{-1}\rangle,$ where $t$ is a rotation and $s$ is a reflection. The elements of $D_{2k}$ are $k$ rotations $t^{0},t,t^{2},...,t^{k-1}$ and $k$ reflections $s,st,st^{2},...,st^{k-1}$. A graph $G$ is said to be _connected_ if there is no permutation matrix $P$ such that $PA(G)P^{-1}=\bigoplus_{i}M_{i}$, for some matrices $M_{i}$. If a Cayley graph of a group $\Gamma$ is not connected then it is the disjoint union of isomorphic Cayley graphs of a subgroup of $\Gamma$, with each isomorphic component corresponding to a coset of the subgroup. ###### Proposition 2 Let $G$ be a loopless anticirculant graph. Then $G$ is a Cayley graph of the dihedral group with respect to a set of reflections. Proof. The adjacency matrix of a Cayley graph $G=G(\Gamma,T)$ can be written as $A(G)=\sum_{t\in T}\rho_{\rm{reg}}(t)$, where $\rho_{\rm{reg}}$ is the regular permutation representation of $\Gamma$. When the order of $\Gamma$ is $n$, this is an homomorphism of the form $\rho_{\rm{reg}}:\Gamma\longrightarrow\Sigma_{n}$, where $\Sigma_{n}$ is the set of all $n\times n$ permutation matrices. Each $\rho_{\rm{reg}}(t)$ describes the action of the group element $t$ on the set $\\{1,2,...,n\\}$. Let $G\in\mathcal{L}_{n}$ be a loopless graph. By Proposition 1, we need to take $n=2k$. Let us label the lines (rows and columns) of $A(G)$ with the group elements of $D_{n}$, of order $n=2k$, in the following order: $\begin{tabular}[]{lllllllll}$t^{0}=e$&$s$&$t$&$st$&$t^{2}$&$st^{2}$&$\cdots$&$t^{k-1}$&$st^{k-1}$\end{tabular},$ where $e$ is the identity element of the group. Consider $\rho_{\rm{reg}}(st^{l})$, where $1\leq l\leq k$. By applying the generating relations of $D_{n}$, it is straighforward to verify that $\rho_{\rm{reg}}(s)=P_{2}$, and $\rho_{\rm{reg}}(st^{l})=P_{2+2l}$, when $1\leq l\leq k-1$. The elements of $S(G)$ are then associated to reflections in $D_{2k}$: $\begin{tabular}[]{lllll}$2\longrightarrow s$&$4\longrightarrow st$&$6\longrightarrow st^{2}$&$\cdots$&$2k\longrightarrow st^{k-1}$\end{tabular}.$ This concludes the proof of the proposition. An example is useful to clarify this result. Let $\mathbb{Z}_{n}$ be the additive group of integers modulo $n$. The Cayley graph $X(\mathbb{Z}_{n},\\{1,n-1\\})$ is also called the $n$_-cycle_. This is the unique connected $2$-regular graph up to isomorphism. Let $G\in\mathcal{L}_{6,2}$ be the graph with adjacency matrix $A(G)=\rho_{\rm{reg}}(s)+\rho_{\rm{reg}}(st)=\left(\begin{array}[]{cccccc}0&1&0&1&0&0\\\ 1&0&1&0&0&0\\\ 0&1&0&0&0&1\\\ 1&0&0&0&1&0\\\ 0&0&0&1&0&1\\\ 0&0&1&0&1&0\end{array}\right).$ where $s=(1,2)(3,6)(4,5)$ and $t=(1,4)(2,3)(5,6)$. Then $s^{2}=st^{2}=e$ and $\langle s,st\rangle=D_{6}$. Additionally, it is immediate to see that $G\cong X(\mathbb{Z}_{6},\\{1,5\\})$. ## 4 Anticirculant weighing matrices In this section we will highlight the interplay between weighing matrices and anticirculant graphs supporting orthogonal matrices. A Weighing matrix $W$ of _order_ $n$ and _weight_ $k$, denoted by $W(n,k)$, is a square $n\times n$ matrix with entries $[W]_{i,j}\in\\{-1,0,1\\}$, satisfying $WW^{T}=kI$, where $I$ is the identity matrix. When $k=n$, then $W(n,k)$ is said to be a _Hadamard matrix._ It is generally recognized that weighing matrices were first discussed by Frank Yates in 1935 [13], while Hadamard matrices were introduced by James Sylvester and Jacques Hadamard during the second half of the nineteenth century (see [14]). Geramita and Seberry [15], and Koukouvinos and Seberry [16] are general surveys on this topic, as well as applications. We are specifically interested in circulant weighing matrices. A _circulant weighing matrix_ of order $n$ and weight $k$ is a $W(n,k)$ which is also a circulant matrix. In analogy with the previous section, we define the ordered set $T(W(n,k))=\\{s:[A(CW(n,k))]_{i,s}=1\\}$ to be the symbol of $CW(n,k)$. Arasu and Seberry overview the subject in [17]. We shall give particular importance to graphs associated to weighing matrices. The _graph of a_(real symmetric) _matrix_ $M$, denoted by $G(M)$, is the graph defined by $[A(G(M))]_{i,j}:=\left\\{\begin{tabular}[]{ll}$1,$&if $[M]_{i,j}\neq 0;$\\\ $0,$&if $[M]_{i,j}=0.$\end{tabular}\right.$ The graph $G(CW(n,k))$ is a Cayley graph of the cyclic group $\mathbb{Z}_{n}$ with respect to the set $T(W(n,k))$. For instance, let us look at the matrix $CW(6,4)=\left(\begin{array}[]{rrrrrr}0&1&1&0&1&-1\\\ -1&0&1&1&0&1\\\ 1&-1&0&1&1&0\\\ 0&1&-1&0&1&1\\\ 1&0&1&-1&0&1\\\ 1&1&0&1&-1&0\end{array}\right)$ (8) The graph $G(CW(6,4))$ is the Cayley graph of $\mathbb{Z}_{6}$ with respect to the set $\\{1,2,4,5\\}$. This is illustrated in Figure 1. Figure 1: The Cayley graph $X(\mathbb{Z}_{6},\\{1,2,4,5\\})$. The following statement is one of the main tools in our discussion. For the purposes of this paper, it is sufficient to focus on matrices of even order. ###### Proposition 3 For each circulant weighing matrix $M=CW(n,k)$ of order $n=2l$, with $l\geq 1$, there is a weighing matrix $U$ such that $G(U)\in\mathcal{L}_{n,k}$. The graph $G(U)$ is loopless if each element of $T(W(n,k))$ is odd. Proof. Consider the permutation $\pi=\left(1,n\right),(2,n-1),...,(l,l+1)$ (9) on a set $[n]$. Let $P_{\pi}$ be the permutation matrix induced by $\pi$. Labeling the rows and the columns of $P_{\pi}$ with the elements of $[n]$ in the lexicographic order, we can write $P_{\pi}=\left(\begin{array}[]{cccc}0&\cdots&0&1\\\ \vdots&0&1&0\\\ 0&&\ddots&\vdots\\\ 1&0&\cdots&0\end{array}\right).$ Now let $U=MP_{\pi}$. Since $P_{\pi}$ and $M$ are orthogonal matrices, $U$ is also orthogonal. By the action of the permutation $\pi$ and the fact that $M$ is circulant, the graph $G(U)\in\mathcal{L}_{n,k}$ and $S(G(U))=\\{s=n-[A(CW(n,k))]_{i,t}+1\\}$, where $t\in T(W(n,k))$. Notice that $G(U)$ can have self-loops if we do not impose further restrictions. On the other hand $G(U)$ is loopless if each $s\in S(G(U))$ is even (Proposition 1). Such a property is satisfied only if each element of $T(W(n,k))$ is odd. It is important to remark that $P_{\pi}$ acts on columns and hence it does not generally preserve isomorphism of graphs. Proposition 3 prompts us to some definitions. An _anticirculant weighing_ _matrix_ of order $n$ and weight $k$, denoted by $AW(n,k)$, is a $W(n,k)$ which is also anticirculant. A matrix $AW(n,k)$ is a special kind of Hankel weighing matrix. A _Hankel weighing_ _matrix_ of order $n$ and weight $k$, denoted by $M=HW(n,k)$, is a $W(n,k)$ such that $G(M)$ is anticirculant. While it is obvious that there exists an $AW(n,k)$ if and only if there exists a $CW(n,k)$, we need to take into account the following counterexample: ###### Proposition 4 Not every $HW(n,k)$ is an $AW(n,k)$. Proof. Let $P_{\pi}$ be the permutation matrix as in the proof of Proposition 3. The matrix $U=\left(\begin{array}[]{cc}P_{\pi}&P_{\pi}\\\ P_{\pi}&-P_{\pi}\end{array}\right)$ (10) is a $HW(n,2)$, for every $n$, since $UU^{T}=U^{2}=\left(\begin{array}[]{cc}2P_{\pi}^{2}&0\\\ 0&2P_{\pi}^{2}\end{array}\right)=\left(\begin{array}[]{cc}2I&0\\\ 0&2I\end{array}\right),$ because $P_{\pi}=P_{\pi}^{T}$. Proposition 4 seems to indicate that classification theorems for circulant weighing matrices help only partially when attempting to classify Hankel weighing matrices. In the next two sections, our classification begins by considering graphs of small degree. This corresponds physically to the number of pump beams required to build a given CVCS. If a graph is $r$-regular, it requires $2r-1$ different pump beams to build the associated CVCS, since each of the $2r-1$ bands of the adjacency matrix requires a different pump frequency. Thus, we are primarily interested in graphs that have an interesting topology, but also have small values of $r$, since greater values would have greater experimental complexity. Cases of small $r$ are also more tractable theoretically, which is another reason to focus on them. ## 5 2-regular graphs The unique – up to isomorphism – connected $2$-regular graph on $n$ vertices is the $n$_-cycle_ , $C_{n}$. A graph $G$ is said to be the _disjoint union_ of graphs $H_{1},...,H_{l}$ if there is a permutation matrix $P$ such that $PA(G)P^{-1}=\bigoplus_{i=1}^{l}A(H_{i})$. In such a case, we write $G\cong\biguplus\nolimits_{i=1}^{l}H_{i}$. Equivalently, let $G$ and $H$ be graph such that $V(G)\cap V(H)=\emptyset$. The set of vertices of $G\uplus H$ is $V(G)\cup V(H)$ and the set of edges, $E(G)\cup E(G)$. All useful information about graphs of matrices $HW(n,2)$ is stated as follows: ###### Proposition 5 Let $D_{n}$ be the dihedral group of order $n=2k$. Let $T=\\{st^{i},st^{j}\\}\subset D_{n}$, where $1\leq i,j\leq k$, be a set of reflections. There is a unitary matrix $U$ such that $X(D_{n},T)\cong G(U)$ if and only if $(i-j)\bmod k=(j-i)\bmod k$. Moreover, if this is the case then $k$ is even and therefore $n$ is a multiple of $4$. It follows that the graph $G\in\mathcal{L}_{n,2}$ and $G$ is the disjoint union of $k/2$ copies of $C_{4}$. Proof. It is known that given a group $\Gamma$ and a set $T=\\{g,h\\}\subset\Gamma$, there is a unitary matrix $U$ such that $A(X(\Gamma,T))=A(G(U))$ if and only if $gh^{-1}=g^{-1}h$ (see, _e.g._ , [18, 19]). From this, $st^{i}(st^{j})^{-1}=(st^{i})^{-1}st^{j}$. Since $st^{l}$ is a reflection, for every $1\leq l\leq k$, we can write $(st^{l})^{-1}=st^{l}$. Thus $st^{i}st^{j}=st^{j}st^{i}$, that is, $st^{i}$ and $st^{j}$ commute. Since $st^{l}s=t^{k-l}$, for every $1\leq l\leq k$, it follows that $t^{(k-i+j)\bmod k}=t^{(k-j+i)\bmod k}$. This proves that $(i-j)\bmod k=(j-i)\bmod k$, for $st^{i}$ and $st^{j}$. Now, without loss of generality, assume that $j>i$. Note that $(i-j)\bmod k=k-j+i$. So, we need that $j-i=k-j+i,$ which implies $j-i=k/2$. It follows that $U$ has the form described in Eq. (10), up to isomorphism. Figure 2 illustrates the graph $X(D_{8},\\{st,st^{3}\\})$ associated to the matrix $HW(8,2)$. Since in general only a disjoint union of squares can be achieved with 2-regular graphs, and these graphs have only limited interest from a quantum computing perspective, we need to consider larger graphs. A proposal for creating such graphs was given in Ref. [9]. Figure 2: The graph $X(D_{8},\\{st,st^{3}\\})\cong C_{4}\uplus C_{4}$. ## 6 4-regular graphs In this section we consider several constructions of 4-regular graphs of anticirculant weighing matrices. Unfortunately, we are unable to prove that this set of examples is exhaustive, so a complete characterization of the degree 4 case remains open. The CVCS graphs in this section can all be implemented with an experimental setup requiring no more than 7 pump beams. ### 6.1 Graphs from $AW(4,4)$ The matrix $M_{4,4}=AW(4,4)=\left(\begin{array}[]{rrrr}1&1&1&-1\\\ 1&1&-1&1\\\ 1&-1&1&1\\\ -1&1&1&1\end{array}\right)$ is a Hadamard matrix of order $4$. Since the entries of a Hadamard matrix are all nonzero, $G(M_{4,4})\cong K_{4}^{+}$, where $K_{n}^{+}$ denotes the complete graph on $n$ vertices with a self-loop at each vertex. We can construct other anticirculant weighing matrices of higher order from $M_{4,4}$. Given that we are interested in anticirculant matrices whose graph is without self-loops, the first row of these matrices must have nonzero entries only at even positions. Let $\mathcal{M}_{4,4}$ be a family of matrices obtained from $M_{4,4}$ be adding new lines (_i.e._ , rows and columns) between the rows and the columns of $M_{4,4}$. If we add the same number of lines between each column (resp. row), we obtain matrices of order $4k$, for $k\geq 1$. For small orders, the first rows of these matrices are $\begin{tabular}[]{l}$0,1,0,1,0,1,0,-1$\\\ $0,0,1,0,0,1,0,0,1,0,0,-1$\\\ $0,0,0,1,0,0,0,1,0,0,0,1,0,0,0,-1$\\\ $0,0,0,0,1,0,0,0,0,1,0,0,0,0,1,0,0,0,0,-1$\end{tabular}.$ Clearly, for each matrix $M$ constructed in this way, we have $G(M)\in L_{n}$. When $n=4k$, with $k$ odd, the graph $G(M)$ has self-loops. For this reason, we consider only the case $n=4k$, with $k$ even. In other words, we consider the case $n=8l$, for any $l\geq 1$. Let $S(M)=\\{s_{1},s_{2},s_{3},s_{4}\\}$. By the above construction $s_{i+1}-s_{i}=d+1$ for $i=1,2,3$, and $n-s_{4}+2s_{1}=s_{2}$. This last condition provides that the number of columns between $s_{4}$ and $s_{1}$ is the same as one between $s_{2}$ and $s_{1}$ when looking at the matrix as wrapped on a torus, by gluing together the first and the last column. In this way, we can define a unitary matrix $U$ such that $G(U)\cong G(M)$. Let us denote by $U_{1},...,U_{n}$ the rows of $U$. The nonzero entries in $U_{1}$ are exactly $U_{1,s_{1}},U_{1,s_{2}},U_{1,s_{3}}$ and $U_{1,s_{4}}$. In particular, $U_{1,s_{1}}=U_{1,s_{2}}=U_{1,s_{3}}=1$ and $U_{1,s_{4}}=-1$. Suppose $s_{i+1}-s_{i}-1=d$ for all $i=1,2,3$ and $n-s_{4}+2s_{1}=s_{2}$. Thus the inner product between the rows $U_{1}$ and $U_{s_{2}-s_{1}}$ is $\displaystyle\langle U_{1},U_{s_{2}-s_{1}}\rangle$ $\displaystyle=$ $\displaystyle U_{1,s_{1}}U_{l,s_{1}}+U_{1,s_{2}}U_{l,s_{2}}+U_{1,s_{3}}U_{l,s_{3}}+U_{1,s_{1}}U_{l,s_{4}}$ $\displaystyle=$ $\displaystyle U_{1,s_{1}}U_{1,s_{2}}+U_{1,s_{2}}U_{1,s_{3}}+U_{1,s_{3}}U_{1,s_{4}}+U_{1,s_{4}}U_{1,s_{1}}$ $\displaystyle=$ $\displaystyle 1+1-1-1$ $\displaystyle=$ $\displaystyle 0.$ The _distance_ $d$ guarantees that each row (resp. column) of $U$ has exactly $4$ nonzero entries contributing to the zero inner product with other $3$ rows (resp. columns). Orthogonality is guaranteed since no nonzero entries contribute to the inner product with all remaining rows (resp. columns). The numbers $s_{1},...,s_{4}$ must be even and $d=(n-4)/4$ must be odd, since $n=8l=\left(4d+4\right)$. The matrices of order $8l$ constructed in with this method will be denoted by $M_{4,4,l}$. Essentially, orthogonality arises since we _interlace_ $M_{4,4}$ with itself $d+1$ times. This can be seen directly in the matrix $\displaystyle M_{4,4,1}=AW(8,4)=$ $\displaystyle\left(\begin{array}[]{rrrrrrrr}0&1&0&1&0&1&0&-1\\\ \framebox{$1$}&0&\framebox{$1$}&0&\framebox{$1$}&0&\framebox{$-1$}&0\\\ 0&1&0&1&0&-1&0&1\\\ \framebox{$1$}&0&\framebox{$1$}&0&\framebox{$-1$}&0&\framebox{$1$}&0\\\ 0&1&0&-1&0&1&0&1\\\ \framebox{$1$}&0&\framebox{$-1$}&0&\framebox{$1$}&0&\framebox{$1$}&0\\\ 0&-1&0&1&0&1&0&1\\\ \framebox{$-1$}&0&\framebox{$1$}&0&\framebox{$1$}&0&\framebox{$1$}&0\end{array}\right),$ (19) which is constructed by interlacing two copies of $M_{4,4}$. The boxed numbers are the nonzero entries in one of the two copies. At this stage, we can ask information about the graphs of matrices in $\mathcal{M}_{4,4}$. It is sufficient to observe that $M_{4,4,l}=M_{4,4}\otimes F_{l+1}$, where the matrix $F_{n}=\left(\begin{array}[]{cccc}0&\cdots&0&1\\\ \vdots&&1&0\\\ 0&&&\vdots\\\ 1&0&\cdots&0\end{array}\right),$ is $n\times n$. Note that this is just the matrix $P_{\pi}$ from Proposition 3, but we have changed notation to make the dependence on the size of the matrix explicit. The matrix $F_{n}$ is the adjacency matrix of the disjoint union of $n/2$ graphs $K_{2}$, if $n$ is even and the disjoin union of $(n-1)/2$ graphs $K_{2}$ and a single vertex with a self-loop, if $n$ is odd. The _tensor product_ $G\otimes H$ of graphs $G$ and $H$ is the graph such that $A(G\otimes H)=A(G)\otimes A(H)$. The set of vertices of $G\otimes H$ is the Cartesian product $V(G)\times V(H)$ and two vertices $\\{u,u^{\prime}\\}$ and $\\{v,v^{\prime}\\}$ are adjacent in $G\otimes H$ if and only if $\\{u,v\\}\in E(G)$ and $\\{u^{\prime},v^{\prime}\\}\in E(H)$. The graph $G\otimes H$ is connected if and only if both factors are connected and at least one factor is nonbipartite. Here, $K_{2}$ and $K_{4}$ are both connected and $K_{2}$ is bipartite. When $l>1$, one of the factors is not connected. Therefore $G(M_{4,4,l})$ is connected only if $l=1$. In fact $G(M_{4,4,l})\cong\biguplus\nolimits_{\\#l}K_{4,4}$, where $K_{n,n}$ denotes the complete bipartite graph on $2n$ vertices. In Figure 3 is illustrated $G(M_{4,4,1})$; in Figure 4, $G(M_{4,4,4})$. Figure 3: The graph $G(M_{4,4,1})\cong K_{4,4}$ Figure 4: The graph $G(M_{4,4,4})\cong K_{4,4}\uplus K_{4,4}$. The graph $G(M_{4,4,l})$ is made of $l$ connected components. Can we find two permutation matrices $P$ and $Q$ such that $G(PM_{4,4,l}Q)\ncong G(M_{4,4,l})$ and it is connected? In particular, for the moment, we consider only permutation matrices that preserves the distance $d$ between the elements of $S(M)$. From $M_{4,4,l}$, we can construct exactly $l-1$ further matrices, by permutation the lines under this constraint. For example, when $n=24=8\cdot 3$, the possible first rows are $\begin{tabular}[]{l}$0,0,0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,-1$\\\ $0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,-1,0,0$\\\ $0,1,0,0,0,0,0,1,0,0,0,0,0,1,0,0,0,0,0,-1,0,0,0,0$\end{tabular}.$ The position of the entry $-1$ does not matter, because we are interested in the graph $G(M_{4,4,l})$. Given $s_{1},...,s_{4}$, in matrices of the form $M_{4,4,l}$, when $l>1$, $\gcd(\\{s_{i}-s_{j}:\mbox{for all }i,j\mbox{ such that }j>i\\}\cup 8l)>2.$ This implies that $G(PM_{4,4,l}Q)\cong\biguplus\nolimits_{\\#l}K_{4,4}$, for any two permutations $P$ and $Q$ preserving the distance $d$. ### 6.2 Graphs from $AW(6,4)$ It is natural to analyse the construction in the previous section, but replacing the matrix $AW(4,4)$ with some other anticirculant weighing matrix. Recalling that there is no $AW(5,4)$, the smallest available matrix of order $n>4$ is $AW(6,4)$: $M_{6,4}=AW(6,4)=\left(\begin{array}[]{rrrrrr}-1&1&0&1&1&0\\\ 1&0&1&1&0&-1\\\ 0&1&1&0&-1&1\\\ 1&1&0&-1&1&0\\\ 1&0&-1&1&0&1\\\ 0&-1&1&0&1&1\end{array}\right).$ This is obtained from the $CW(6,4)$ in (8). The graph $G(M_{6,4})$, illustrated in Figure 5, has $4$ self-loops. Figure 5: The graph $G(M_{6,4})$. The graph $G(M_{6,4,l})$, if loopless, is on $12l$ vertices, with $l\geq 1$. The graph $G(M_{6,4,1})\cong G(M_{6,4})\otimes K_{2}$ is connected since $K_{2}$ and $G(M_{6,4})$ are connected and $G(M_{6,4})$ is nonbipartite. (The chromatic number of $G(M_{6,4})$ is $3$.) A drawing of $G(M_{6,4,1})$ is in Figure 6. Figure 6: The graph $G(M_{6,4,1})\cong G(M_{6,4})\otimes K_{2}$. In general, $M_{6,4,l}=M_{6,4}\otimes F_{l+1}$. If we avoid self-loops, $G(M_{6,4,l})=\biguplus\nolimits_{l}G(M_{6,4,1}).$ By permuting the lines of $M_{6,4}$ (_shifting to the right_ the columns), we have the matrix $M_{6,4}^{1}=AW(4,4)_{1}=\left(\begin{array}[]{cccccc}0&-1&1&0&1&1\\\ -1&1&0&1&1&0\\\ 1&0&1&1&0&-1\\\ 0&1&1&0&-1&1\\\ 1&1&0&-1&1&0\\\ 1&0&-1&1&0&1\end{array}\right).$ Apply our construction to $M_{6,4}^{1}$, we obtain graphs on $6k$ vertices, with $k$ odd. However, since $G(M_{6,4,1})\cong G(M_{6,4,1}^{1})$, these graphs will be a disjoint union of copies of $G(M_{6,4,1})$. So the permutation is of no use if our intention is to get different graphs from $M_{6,4}$. On the other side, the structure of $M_{6,4}^{1}$ suggests a different construction, which will allow us to obtain matrices of the form $AW(4m,4)$, whose graph is loopless and connected for every $m\geq 3$. Notice that $M_{6,4}=\left(\begin{array}[]{cc}A&B\\\ B&A\end{array}\right),$ where $A=\left(\begin{array}[]{rrr}0&-1&1\\\ -1&1&0\\\ 1&0&1\end{array}\right)$ \| \| $B=\left(\begin{array}[]{rrr}0&1&1\\\ 1&1&0\\\ 1&0&-1\end{array}\right).$ ---\|---\|--- Then $2AB=0$ and $A^{2}+B^{2}=4I$ and indeed $M_{6,4}M_{6,4}=\left(M_{6,4}\right)^{2}=\left(\begin{array}[]{cc}A^{2}+B^{2}&2AB\\\ 2AB&A^{2}+B^{2}\end{array}\right)=4I.$ We can add an even number of extra lines to $M_{6,4}$ and still preserve its block structure. For example, the matrices $A^{\prime}=\left(\begin{array}[]{rrrrrr}0&-1&0&1&0&0\\\ -1&0&1&0&0&0\\\ 0&1&0&0&0&1\\\ 1&0&0&0&1&0\\\ 0&0&0&1&0&1\\\ 0&0&1&0&1&0\end{array}\right)\ ,\ B^{\prime}=\left(\begin{array}[]{cccccc}0&1&0&1&0&0\\\ 1&0&1&0&0&0\\\ 0&1&0&0&0&-1\\\ 1&0&0&0&-1&0\\\ 0&0&0&-1&0&1\\\ 0&0&-1&0&1&0\end{array}\right)$ satisfy the above conditions. Hence the matrix $M_{12,4}=\left(\begin{array}[]{cc}A^{\prime}&B^{\prime}\\\ B^{\prime}&A^{\prime}\end{array}\right)$ is an $AW(12,4)$. In general, let $\mathcal{M}_{6,4}$ be the family of matrices obtained with this construction. If $M\in\mathcal{M}_{6,4}$ then $s_{1}=2$, $s_{2}=4$, $s_{3}=n/2+2$ and $s_{4}=n/2+4$, where $S(M)=\\{s_{1},s_{2},s_{3},s_{4}\\}$. These are $AW(4k,4)$ when $s_{3}-s_{2}=\left(n-6\right)/2$ is an odd number. The construction gives also matrices of order $2l\geq 6$, for any $l$. Yet the associated graphs will have loops when $l$ is odd. This is the reason for taking multiples of $4$. If $M\in\mathcal{M}_{6,4}$ then $G(M)$ is connected, in virtue of the fact that $s_{2}-s_{1}=4-2=2$ and then $\gcd(\\{s_{i}-s_{j}:$ for all $i,j$ such that $j>i\\}\cup 4k\\}=2$. The graphs of matrices in the set $\mathcal{M}_{6,4}$, with order $4m$, are the direct product of $C_{2m}$ with the graph $K_{2}^{+}$. (Recall that $K_{n}^{+}$ is the complete graph on $n$ vertices with self-loops; these are the graphs with adjacency matrix $J_{n}$.) The _direct product_ $G\times H$ of graphs $G$ and $H$ has set of vertices $V(G\times H)=V(G)\times V(H)$ and two vertices $\\{u,u^{\prime}\\}$ and $\\{v,v^{\prime}\\}$ are adjacent in $G\times H$ when $\\{u,v\\}\in E(G)$ and $\\{u^{\prime},v^{\prime}\\}\in E(H)$. By the definition, $A(G\times H)=A(G)\otimes A(H)$. For this reason $G\times H$ is sometimes called the _Kronecker product of graphs_ (see Imrich and Klavžar [20], Ch. 5). If $G\in\mathcal{M}_{6,4}$ then $G\cong K_{2}^{+}\times C_{2m}$ and $M(G)=J_{2}\otimes A(C_{2m})$. ### 6.3 Graphs from $AW(7,4)$ Let us consider the matrix $M_{7,4}=AW(7,4)=\left(\begin{array}[]{rrrrrrr}0&1&0&1&1&-1&0\\\ 1&0&1&1&-1&0&0\\\ 0&1&1&-1&0&0&1\\\ 1&1&-1&0&0&1&0\\\ 1&-1&0&0&1&0&1\\\ -1&0&0&1&0&1&1\\\ 0&0&1&0&1&1&-1\end{array}\right).$ The graph of $M_{7,4}$ is in Figure 7. Figure 7: The graph $G(M_{7,4})$. Figure 8: The graph $G(M_{7,4,1})$. Interlacing $M_{7,4}$ with itself, by the same method used for $M_{4,4}$ (see section 6.1, Eq. (19)), we obtain the matrix $M_{7,4,1}=AW(14,4)=AW(7,4)\otimes F_{2}$ (20) The graph of $M_{7,4,1}$ is in Figure 8. This is on $14$ vertices, $4$-regular and connected. At a first analysis, $G(M_{7,4,1})$ does not seem to be a direct product of graphs. It is nontrivial to extend $G(M_{7,4,1})$ to an infinite family of connected regular graphs, as we did by taking $M_{6,4}$. Examples show that graphs obtained by $M_{7,4}$ are not connected, but $G(M_{7,4,1})$. ## 7 Open problems * • Study the structure of the graph $G(M_{7,4,1})$. Determine if it belongs to a family of connected graphs arising from $AW(7,4)$, having an infinite number of members, as it is for the graphs arising from $AW(6,4)$. * • Since there is no general classification of weighing matrices, characterizing graphs of Hankel weighing matrices seems to be out of reach. This could be however done for graphs of degree $4$ and $9$, on the basis of present knowledge on $CW(n,4)$ and $CW(n,9)$ (see [17]). * • We have seen that there are examples of Hankel weighing matrices which are not anticirculant. A way to study the relation between these matrices and anticirculant ones, would be to prove that the graphs of Hankel weighing matrices are always isomorphic to graphs of $AW(n,d)$, except for some special cases which will include the examples in proposition 4. Figure 9: Hamiltonian cycles in graphs of matrices $AW(12,4)$ and $AW(14,4)$. * • A _Hamiltonian cycle_ in a graph is an ordered set of sequentially adjacent vertices, in which every vertex of the graph appears exactly once. Graphs with Hamiltonian cycles are said to be _Hamiltonian_. Alspach and Zhang [21] conjectured that a connected Cayley graph $G(D_{n},T)$ is Hamiltonian whenever $T$ is a set of reflections. According to this conjecture, any connected anticirculant graph is Hamiltonian whenever it is loopless. Additionally, Gutin _et al._ [22] conjectured that graphs of unitary matrices are Hamiltonian if connected. According to these conjectures the graphs of Hankel weighing matrices are Hamiltonian if connected. This is the case for the graphs considered here. In Figure 9 are drawn the graphs $K_{2}^{+}\times C_{8}$ and $G(M_{7,4,1})$. The highlighted edges represent Hamiltonian cycles. Perhaps the physical picture that motivates our study could provide some insight toward proving these conjectures. * • We have seen that graphs of the form $K_{2}^{+}\times C_{2m}$ are graphs of unitary matrices. The quantum dynamics induced by these matrices is trivial because of symmetry. However symmetry can be broken with the action of a diagonal matrix without altering the zero pattern of the unitary. This suggests the possibility of constructing unitaries with a nontrivial dynamics whose graph is a undirected Cayley graph of the dihedral group. * • Sections 5 and 6 make extensive use of the following trick. Begin with a circulant weighing matrix, then reverse the columns. If the resulting graph has loops, one can create a loopless graph by the “interlacing” trick. In fact the interlaced graph will again be connected if the original graph was connected but not bipartite. Can one classify the Hankel and anticirculant weighing matrices which cannot be obtained by this trick? ## 8 Conclusion We have demonstrated connections between an approach to quantum computation with optical modes and the existence of certain weighing matrices with Hankel structure. Because the degree of the graph associated to a matrix corresponds to the experimental resources required to implement the graph as a CVCS, it is important to characterize which graphs with small degree have adjacency matrices with Hankel structure. We proved some general theorems about such matrices, classified completely the case of degree 2, and provided some examples with degree 4. We also raised some interesting open problems that might help shed some light on which graphs can be implemented in a single OPO using the scheme in Ref. [7, 10]. ## Acknowledgments STF was supported by the Perimeter Institute for Theoretical Physics. Research at Perimeter is supported by the Government of Canada through Industry Canada and by the Province of Ontario through the Ministry of Research & Innovation. SS was supported by the Institute for Quantum Computing. Research at the Institute for Quantum Computing is supported by DTOARO, ORDCF, CFI, CIFAR, and MITACS. ## References ## References * [1] M. A. Nielsen and I. L. Chuang. Quantum Computation and Quantum Information. Cambridge University Press, Cambridge, 2000. * [2] R. Raussendorf and H. J. Briegel. A one-way quantum computer. Phys. Rev. Lett., 86(22):5188–5191, 2001. * [3] H. J. Briegel and R. Raussendorf. Persistent entanglement in arrays of interacting particles. Phys. Rev. Lett., 86(5):910–913, Jan 2001. * [4] D. L. Zhou, B. Zeng, Z. Xu, and C. P. Sun. Quantum computation based on d-level cluster states. Phys. Rev. A, 68:062303, 2003. * [5] J. Zhang and S. L. Braunstein. Continuous-variable gaussian analog of cluster states. Phys. Rev. A, 73(3):32318, 2006. * [6] N. C. Menicucci, P. van Loock, M. Gu, C. Weedbrook, T. C. Ralph, and M. A. Nielsen. Universal quantum computation with continuous-variable cluster states. Phys. Rev. Lett., 97(11):110501, 2006. * [7] N. C. Menicucci, S. T. Flammia, and O. Pfister. One-way quantum computing in the optical frequency comb. Phys. Rev. Lett., 101:130501, 2008. * [8] N. C. Menicucci, S. T. Flammia, H. Zaidi, and O. Pfister. Ultracompact generation of continuous-variable cluster states. Phys. Rev. A, 76(1):010302, Jul 2007. * [9] H. Zaidi, N. C. Menicucci, S. T. Flammia, R. Bloomer, M. Pysher, and O. Pfister. Entangling the optical frequency comb: simultaneous generation of multiple 2x2 and 2x3 continuous-variable cluster states in a single optical parametric oscillator. Laser Phys., 18(5):659, May 2008. * [10] S. T. Flammia, N. C. Menicucci, and O. Pfister. The optical frequency comb as a one-way quantum computer. arXiv: 0811.2799 [quant-ph], 2008. * [11] R. Diestel. Graph theory. Graduate Texts in Mathematics. Springer-Verlag, New York, 2nd edition, 2000. * [12] P. J. Cameron. Permutation groups. Number 45 in London Mathematical Society Student Texts. Cambridge University Press, Cambridge, 1999. * [13] F. Yates. Complex experiments. J. Roy. Statist. Soc. (Suppl)., 2:181–247, 1935. * [14] K. J. Horadam. Hadamard matrices and their applications. Princeton University Press, Princeton, NJ, 2007. * [15] A. V. Geramita and J. Seberry. Orthogonal designs, Quadratic forms and Hadamard matrices. Number 45 in Lecture Notes in Pure and Applied Mathematics. Marcel Dekker, Inc., New York, 1979. * [16] C. Koukouvinos and J. Seberry. Weighing matrices and their applications. J. Statist. Plann. Inference, 62(1):91–101, 1997. * [17] K. T. Arasu and J. Seberry. On circulant weighing matrices. Australas. J. Combin., 17:21–37, 1998. * [18] S. Severini. On the digraph of a unitary matrix. SIAM J. Matrix Anal. Appl. (SIMAX), 25(1):295–300, 2003. * [19] S. Severini. Graphs of unitary matrices. Ars Combin., 89, Oct 2008. * [20] W. Imrich and S. Klavžar. Product graphs. Structure and recognition. Wiley-Interscience Series in Discrete Mathematics and Optimization. Wiley-Interscience, New York, 2000. * [21] B. Alspach and C.-Q. Zhang. Hamilton cycles in cubic cayley graphs on dihedral groups. Ars Combin., 28:101–108, 1989. * [22] G. Gutin, A. Rafiey, S. Severini, and A. Yeo. Hamilton cycles in digraphs of unitary matrices. Discrete Appl. Math., 173(1-3):67–78, 2006.	arxiv-papers	2008-08-14T20:04:15	2024-09-04T02:48:57.306830	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Steven T. Flammia, Simone Severini", "submitter": "Steve Flammia", "url": "https://arxiv.org/abs/0808.2057" }
0808.2103	# Quasielastic and inelastic neutrino reactions in ${}^{12}C$ at K2K energies M. Sajjad Athar, S. Chauhan and S. K. Singh Department of Physics, Aligarh Muslim University, Aligarh-202 002, India ###### Abstract In this paper, we present the results of a study made for the effect of nuclear medium in the charged current induced quasielastic lepton production(CCQE) and the incoherent and coherent one pion production (CC1$\pi^{+}$)processes from ${}^{12}C$ in the $\nu_{\mu}$ energy region of 0.4-3GeV. The theoretical results are compared with the recent experimental results for the ratio of charged current $\nu_{\mu}$ induced one pion production cross section to the quasielastic lepton production cross section reported by K2K collaboration. We also present the results for the angular and momentum distributions of leptons and pions produced in these processes. ###### pacs: 12.15.-y,13.15.+g,13.60.Rj,23.40.Bw,25.30.Pt ## I INTRODUCTION Recently K2K collaboration has reported experimental results on the inclusive single pion production induced by charged current by neutrinos in the energy region of 0.4-3 GeVRodriguez . Similar results were earlier reported by MiniBooNE collaboration in the energy region of 1GeV Miniboone . The study of the energy dependence of the pion production cross sections in neutrino reactions for nuclei in this energy is important in modeling the neutrino nucleus cross sections for various Monte Carlo neutrino event generators used in analyzing the present neutrino oscillation experiment at MiniBooNE, K2K and future experiments to be done by T2KT2K and NO$\nu$A Nova collaborations. The present experimental result by K2K collaboration reports the energy dependence of the ratio of inclusive cross sections of the charge current induced pion production and quasielastic reaction in neutrino reaction on polystyrene($C_{8}H_{8}$) target. Both of these reactions are influenced by the nuclear medium effects in the energy region of this experiment. The Monte Carlo simulation of single pion production in this reaction uses the Rein and Sehgal modelRein which does not include any nuclear medium effect in the production process. The NEUT generator used in the analysis of K2K experimentRodriguez , however, includes the nuclear effects arising due to final state interactions of pions with the nucleus like pion absorption and pion scattering Neut . The quasielastic reaction, on the other hand, uses Smith-Moniz model Moniz which does not include the effect of nuclear medium arising due to nucleon-nucleon correlations but includes only the effect of Pauli principle and Fermi motion in a Fermi gas model. There are many calculations for the nuclear medium effects in the quasielastic reaction llewelyn -Leitner , and quite a few calculations in the case of single pion production which take into account explicitly the nuclear medium and final state interaction effects in various energy region a2 -RusoCoh . In this paper we present a study of nuclear medium effects in these reactions in the energy region of 0.4-3 GeV averaged over the neutrino flux of the K2K experiment. The single pion production in this region is dominated by the resonance production in which a $\Delta$ resonance is excited and decays subsequently to a pion and a nucleon. When this process takes place inside the nucleus then there are two possibilities, the target nucleus remains in the ground state leading to coherent production of pions or is excited and/or broken up leading to incoherent production of pions. In the present paper, we have considered both the production processes in the $\Delta$ resonance model in the local density approximation to calculate single pion production from nuclei. The effect of nuclear medium on the production of $\Delta$ is treated by including the modification of $\Delta$ properties in the medium. Once pions are produced, they undergo final state interactions with the final nuclei. For the incoherent pion production final state interaction has been treated using a Monte Carlo code for pion nucleus interaction Vicente , while for the coherent case, the distortion of the pion plane wave is calculated by using the Eikonal approximation Carrasco . In the case of quasielastic reaction, the effects of Pauli principle and Fermi motion are included through the Lindhard function calculated in local density approximation Singh1 . The details of these calculations are given in earlier publications Athar2 ,a2 ,Athar4 . In this paper we report the results for the nuclear medium effects on total cross section $\sigma(E)$, $Q^{2}$ distribution $\frac{d\sigma}{dQ^{2}}$ and momentum distribution $\frac{d\sigma}{dp_{l}}$ for the muon, in the case of quasielastic and single pion production induced by charged currents. We compare our results of the ratio $R(E)=\frac{\sigma^{CC1\pi^{+}}(E)}{\sigma^{CCQE}(E)}$, with the recent observations of K2K collaboration. The results for the pion momentum and angular distributions have also been shown for the incoherent pion production process. In Secs. II & III, we briefly describe the various formula used in the present calculations for quasielastic and inelastic reactions and present the results and discussion in Sec.IV. ## II QUASIELASTIC REACTION The basic reaction for the quasielastic process is a neutrino interacting with a neutron inside the nucleus i.e. $\nu_{\mu}(k)+n(p)\rightarrow\mu^{-}(k^{\prime})+p(p^{\prime})$ (1) The cross section for quasi-elastic charged lepton production is calculated in the local density approximation by taking into account the Fermi motion and the Pauli blocking effects through the imaginary part of the Lindhard function for the particle hole excitations in the nuclear medium. The renormalization of the weak transition strengths are calculated in the Random Phase Approximation(RPA) through the interaction of the p-h excitations as they propagate in the nuclear medium using a nucleon-nucleon potential described by pion and rho exchanges Oset1 . The effect of the Coulomb distortion of muon in the field of final nucleus is also taken into account using a local version of the modified effective momentum approximation(MEMA) Athar2 . Figure 1: Kinematical variables for $\nu_{\mu}(k)+N(p)\rightarrow\mu^{-}(k^{\prime})+N^{\prime}(p^{\prime})+\pi^{+}(k_{\pi})$ process. The total cross section $\sigma(E_{\nu})$ for the charged current neutrino induced reaction on a nucleon inside the nucleus in a local Fermi gas model is written as Athar2 : $\displaystyle\sigma(E_{\nu})$ $\displaystyle=$ $\displaystyle-\frac{2{G_{F}}^{2}\cos^{2}{\theta_{c}}}{\pi}\int_{r_{min}}^{r_{max}}r^{2}dr\int^{{k^{\prime}}_{max}}_{{k^{\prime}}_{min}}{k^{\prime}}^{2}dk^{\prime}\int_{-1}^{1}dcos\theta\frac{1}{E_{\nu_{\mu}}E_{\mu}}L_{\mu\nu}J^{\mu\nu}_{RPA}Im{U_{N}(q_{0},{\bf q})},$ (2) where $L_{\mu\nu}=\sum L_{\mu}{L_{\nu}}^{\dagger}$ and ${J^{\mu\nu}_{RPA}}={\bar{\sum}}\sum J^{\mu}{J^{\nu}}^{\dagger}$, calculated with RPA correlations in nuclei. The leptonic current $L_{\mu}$ and the hadronic current $J^{\mu}$ are given by $L_{\mu}=\bar{u}(k^{\prime})\gamma_{\mu}(1-\gamma_{5})u(k)$ (3) $\displaystyle J^{\mu}$ $\displaystyle=$ $\displaystyle\bar{u}(p^{\prime})[F_{1}(q^{2})\gamma^{\mu}+F_{2}(q^{2})i{\sigma^{\mu\nu}}{\frac{q_{\nu}}{2M}}+F_{A}(q^{2})\gamma^{\mu}\gamma_{5}+F_{P}(q^{2})q^{\mu}\gamma_{5}]u(p).$ (4) Figure 2: Total scattering cross section for $\nu_{\mu}$ induced charged current quasielastic process in ${}^{12}C$. Figure 3: Total scattering cross section for the $\nu_{\mu}$ induced incoherent charged current one pion production process in ${}^{12}C$. Figure 4: Total scattering cross section for $\nu_{\mu}$ induced coherent charged current one pion production process in ${}^{12}C$. Figure 5: Ratio of the cross sections for $\nu_{\mu}$ induced charged current one pion production process and charged current quasielastic process in polystyrene. Experimental points have been taken from Ref.Rodriguez for K2K results in Scibar and from Refs.Radecky , Barish for ANL results. Figure 6: $Q^{2}$ distribution for $\nu_{\mu}$ induced charged current quasielastic process in ${}^{12}C$ averaged over the K2K spectrum. where $q(=k-k^{\prime})$ is the four momentum transfer, M is the mass of the nucleon, $G_{F}(=1.16637\times 10^{-5}GeV^{-2})$ is the Fermi coupling constant and $\theta$ is the lepton angle. $U_{N}$ is the Lindhard function for the particle hole excitation Singh1 . The form factors $F_{1}(Q^{2})$, $F_{2}(Q^{2})$, $F_{A}(Q^{2})$ and $F_{P}(Q^{2})$ are isovector electroweak form factors. For our numerical calculations we have used the parameterization of Bradford et al. bradford for the vector form factors $F_{1}(Q^{2})$ and $F_{2}(Q^{2})$ with dipole mass ${M}_{V}$=0.84GeV and a dipole form for $F_{A}(Q^{2})$ with dipole mass $M_{A}$=1.1GeV. Inside the nucleus, the Q-value of the reaction and Coulomb distortion of outgoing lepton are taken into account by modifying the Lindhard function ${U_{N}(q_{0},{\bf q})}$ by ${U_{N}(q_{0}-V_{c}(r)-Q,{\bf q})}$, where $V_{c}(r)$ is the Coulomb potential of the final nucleus. Furthermore, the renormalization of weak transition strength in the nuclear medium in a random phase approximation(RPA) is taken into account by considering the propagation of particle hole(ph) as well as delta-hole($\Delta h$) excitations Oset2 . These considerations lead to modified hadronic tensor components $(J^{\mu\nu}_{RPA})$ involving the bilinear terms, for which expressions are given in Refs. Nieves ,Athar2 . Figure 7: Muon momentum distribution for $\nu_{\mu}$ induced charged current quasielastic process in ${}^{12}C$ averaged over the K2K spectrum. Figure 8: $Q^{2}$ distribution for $\nu_{\mu}$ induced incoherent charged current one pion production process in ${}^{12}C$ averaged over the K2K spectrum. ## III INELASTIC RESONANCE PRODUCTION OF PIONS ### III.1 INCOHERENT PION PRODUCTION The basic reaction for the inelastic one pion production in nuclei, for a neutrino interacting with a nucleon N inside a nuclear target is given by $\nu_{\mu}(k)+N(p)\rightarrow\mu^{-}(k^{\prime})+N^{\prime}(p^{\prime})+\pi^{+}(k_{\pi}).$ (5) In the case of incoherent pion production in $\Delta$ dominance model, the weak hadronic currents interacting with the nucleons in the nuclear medium excite a $\Delta$ resonance which decays into pions and nucleons. The pions interact with the nucleus inside the nuclear medium before coming out. The final state interaction of pions leading to elastic, charge exchange scattering and the absorption of pions lead to reduction of pion yield. The nuclear medium effects on $\Delta$ properties lead to modification in its mass and width which have been discussed earlier by Oset et al. Oset and applied to explain the pion and electron induced pion production processes from nuclei. In the local density approximation the expression for the total cross section for the charged current one pion production (kinematics is shown in Fig.1) is written as $\displaystyle\sigma$ $\displaystyle=$ $\displaystyle\frac{1}{(4\pi)^{5}}\int_{r_{min}}^{r_{max}}(\rho_{p}(r)+\frac{1}{9}\rho_{n}(r))d\vec{r}\int_{Q^{2}_{min}}^{Q^{2}_{max}}dQ^{2}\int^{{k^{\prime}}_{max}}_{{k^{\prime}}_{min}}dk^{\prime}\int_{-1}^{+1}dcos\theta_{\pi}\int_{0}^{2\pi}d\phi_{\pi}~{}~{}\frac{\pi\|\vec{k}^{\prime}\|\|\vec{k}_{\pi}\|}{ME_{\nu}^{2}E_{l}}$ (6) $\displaystyle\times\frac{1}{E_{p}^{\prime}+E_{\pi}\left(1-\frac{\|\vec{q}\|}{\|\vec{k}_{\pi}\|}cos\theta_{\pi}\right)}\bar{\sum}\sum\|\mathcal{M}_{fi}\|^{2}$ where the proton density $\rho_{p}(r)=\frac{Z}{A}\rho(r)$ and the neutron density $\rho_{n}(r)=\frac{A-Z}{A}\rho(r)$ with $\rho(r)$ as the nuclear density taken as 3-parameter Fermi Density Vries . Figure 9: Muon momentum distribution for $\nu_{\mu}$ induced incoherent charged current one pion production process in ${}^{12}C$ averaged over the K2K spectrum. Figure 10: Pion angular distribution for $\nu_{\mu}$ induced incoherent charged current one pion production process in ${}^{12}C$ averaged over the K2K spectrum. Figure 11: Pion Momentum distribution for $\nu_{\mu}$ induced incoherent charged current one pion production process in ${}^{12}C$ averaged over the K2K spectrum. The transition matrix element $\mathcal{M}_{fi}$ is given by $\mathcal{M}_{fi}=\sqrt{3}\frac{G_{F}}{\sqrt{2}}\frac{f_{\pi N\Delta}}{m_{\pi}}\bar{u}({\bf p}^{\prime})k^{\sigma}_{\pi}{\mathcal{P}}_{\sigma\lambda}\mathcal{O}^{\lambda\alpha}L_{\alpha}u({\bf p})$ (7) where $L^{\alpha}$ is the leptonic current defined by Eq.(3), and $\mathcal{O}^{\beta\alpha}$ is the $N-\Delta$ transition operator taken from Lalakulich et al. Lalakulich1 . $\theta_{W}$ is the weak mixing angle. ${\mathcal{P}}^{\sigma\lambda}$ is the $\Delta$ propagator in momentum space and is given as : ${\mathcal{P}}^{\sigma\lambda}=\frac{{\it P}^{\sigma\lambda}}{P^{2}-M_{\Delta}^{2}+iM_{\Delta}\Gamma}$ (8) where ${\it P}^{\sigma\lambda}$ is the spin-3/2 projection operator and the delta decay width $\Gamma$ is taken to be an energy dependent P-wave decay width Oset : $\Gamma(W)=\frac{1}{6\pi}\left(\frac{f_{\pi N\Delta}}{m_{\pi}}\right)^{2}\frac{M}{W}\|{\bf q}_{cm}\|^{3}$ (9) $\|q_{cm}\|$ is the pion momentum in the rest frame of the resonance and W is the center of mass energy. Inside the nuclear medium, the mass and width of delta are modified which in the present calculation are taken by considering the following effects. In nuclear medium $\Delta$s decay mainly through the $\Delta\rightarrow N\pi$ channel. The final nucleons have to be above the Fermi momentum $p_{F}$ of the nucleon in the nucleus which leads to a modification in the decay width of delta. The modified delta decay width $\tilde{\Gamma}$ has been taken from the works of Oset et al. Oset which is given by $\tilde{\Gamma}=\Gamma\times F(p_{F},E_{\Delta},k_{\Delta})$ (10) where $F(p_{F},E_{\Delta},k_{\Delta})=\frac{k_{\Delta}\|{{\bf q}_{cm}}\|+E_{\Delta}{E^{\prime}_{p}}_{cm}-E_{F}{W}}{2k_{\Delta}\|{\bf q^{\prime}}_{cm}\|}$ (11) $E_{F}=\sqrt{M^{2}+p_{F}^{2}}$, $k_{\Delta}$ is the $\Delta$ momentum and $E_{\Delta}=\sqrt{W+k_{\Delta}^{2}}$. Furthermore, in the nuclear medium there are additional decay channels open due to two and three body absorption processes like $\Delta N\rightarrow NN$ and $\Delta NN\rightarrow NNN$ through which $\Delta$ disappears in nuclear medium without producing a pion, while a two body $\Delta$ absorption process like $\Delta N\rightarrow\pi NN$ gives rise to some more pions. These nuclear medium effects on the $\Delta$ propagation are included by modifying the mass and the decay width of $\Delta$ in nuclear medium as $\Gamma\rightarrow\tilde{\Gamma}-2Im\Sigma_{\Delta}~{}~{}\text{and}~{}~{}M_{\Delta}\rightarrow\tilde{M}_{\Delta}=M_{\Delta}+Re\Sigma_{\Delta}.$ (12) The expressions for the real and imaginary part of the $\Delta$ self energy are taken from Oset et al. Oset : $\displaystyle Re{\Sigma}_{\Delta}$ $\displaystyle=$ $\displaystyle 40\frac{\rho}{\rho_{0}}MeV~{}~{}and$ $\displaystyle-Im{{\Sigma}_{\Delta}}$ $\displaystyle=$ $\displaystyle C_{Q}\left(\frac{\rho}{{\rho}_{0}}\right)^{\alpha}+C_{A2}\left(\frac{\rho}{{\rho}_{0}}\right)^{\beta}+C_{A3}\left(\frac{\rho}{{\rho}_{0}}\right)^{\gamma}~{}~{}~{}~{}$ (13) In the above equation $C_{Q}$ accounts for the $\Delta N\rightarrow\pi NN$ process, $C_{A2}$ for the two-body absorption process $\Delta N\rightarrow NN$ and $C_{A3}$ for the three-body absorption process $\Delta NN\rightarrow NNN$. The coefficients $C_{Q}$, $C_{A2}$, $C_{A3}$ and $\alpha$, $\beta$ and $\gamma$ are taken from Ref. Oset . The pions produced in this process are scattered and absorbed in the nuclear medium. This is treated in a Monte Carlo simulation which has been taken from Ref. Vicente . ### III.2 COHERENT PION PRODUCTION $\nu_{\mu}$ induced coherent one pion production on ${}^{12}C$ target is given by $\nu_{\mu}+_{6}^{12}C\rightarrow\mu^{-}+_{6}^{12}C+\pi^{+}$ for which the basic reaction is given by Eq.(5). The total cross section (kinematics is shown in Fig.1) is given by $\displaystyle\sigma$ $\displaystyle=$ $\displaystyle\frac{1}{(2\pi)^{5}}\int_{Q^{2}_{min}}^{Q^{2}_{max}}dQ^{2}\int^{{k^{\prime}}_{max}}_{{k^{\prime}}_{min}}dk^{\prime}\int_{-1}^{+1}dcos\theta_{\pi}\int_{0}^{2\pi}d\phi_{\pi}\frac{\pi\|\vec{k}^{\prime}\|\|\vec{k}_{\pi}\|}{8E_{\nu}^{2}E_{l}}~{}~{}\frac{M}{E_{p}^{\prime}+E_{\pi}\left(1-\frac{\|\vec{q}\|}{\|\vec{k}_{\pi}\|}cos(\theta_{\pi})\right)}\bar{\sum}\|\mathcal{M}_{fi}\|^{2}$ (14) where the matrix element for the reaction is given by $\mathcal{M}_{fi}=\frac{G_{F}}{\sqrt{2}}cos\theta_{c}L^{\mu}J_{\mu}{\cal F}(\vec{q}-\vec{k}_{\pi})$ (15) $L_{\mu}$ is the leptonic current given by Eq.(3) and $J_{\mu}$ is the hadronic current given by $J_{\mu}=\sqrt{3}\frac{f_{\pi N\Delta}}{m_{\pi}}F(P^{2})\sum_{r,s}{\bar{u}_{s}}(p^{\prime})k_{\pi\sigma}\mathcal{P}^{\sigma\lambda}\mathcal{O}_{\lambda\mu}u_{r}({\bf p})$ (16) where F($P^{2}$) (P=p+q) is $\Delta N\pi$ form factor given byPenner : $F(P)=\frac{\Lambda^{4}}{\Lambda^{4}+(P^{2}-M_{\Delta}^{2})^{2}}$ (17) with $\Lambda=1GeV$. ${\cal F}(\vec{q}-\vec{k}_{\pi})$ is the nuclear form factor, given by ${\cal F}(\vec{q}-\vec{k}_{\pi})=\int d^{3}{\vec{r}}\left[{\rho_{p}({\vec{r}})}+\frac{1}{3}{\rho_{n}({\vec{r}})}\right]e^{-i({\vec{q}}-{\vec{k}}_{\pi}).{\vec{r}}}$ (18) When pion absorption effects are taken into account using the Eikonal approximation then the nuclear form factor ${\cal F}({\vec{q}}-{\vec{k}_{\pi}})$ is modified to $\tilde{\cal F}({\vec{q}}-{\vec{k}_{\pi}})$, which is taken to be Athar4 : $\displaystyle\tilde{\cal F}({\vec{q}}-{\vec{k}_{\pi}})$ $\displaystyle=$ $\displaystyle 2\pi\int_{0}^{\infty}b~{}db\int_{-\infty}^{\infty}dz~{}\rho({\vec{b}},z)~{}J_{0}(k_{\pi}^{t}b)~{}e^{i(\|{\vec{q}}\|-k_{\pi}^{l})z}e^{-if({\vec{b}},z)}$ (19) where $f({\vec{b}},z)=\int_{z}^{\infty}\frac{1}{2\|{\vec{k}_{\pi}}\|}{\Pi(\rho({\vec{b}},z^{\prime}))}dz^{\prime}$ (20) $k_{\pi}^{l}$ and $k_{\pi}^{t}$ are the longitudinal and transverse component of the pion momentum and the pion self-energy $\Pi$ is given by $\Pi(\rho({\vec{b}},z^{\prime}))=\frac{4}{9}\left(\frac{f_{\pi N\Delta}}{m_{\pi}}\right)^{2}\frac{M^{2}}{W^{2}}\|{\vec{p}_{\pi}}\|^{2}~{}\rho({\vec{b}},z^{\prime})~{}\frac{1}{W-{\tilde{M}}_{\Delta}+\frac{i{\tilde{\Gamma}}}{2}}$ (21) with W as the center of mass energy in the $\Delta$ rest frame. ## IV RESULTS AND DISCUSSIONS ### IV.1 TOTAL CROSS SECTIONS For the quasielastic reaction, the numerical results are obtained from Eq.(2) using vector form factors $F_{1}(Q^{2})$ and $F_{2}(Q^{2})$ given by Bradford et al. bradford with vector dipole mass ${M}_{V}$=0.84GeV and a dipole form for $F_{A}(Q^{2})$ with axial dipole mass $M_{A}$=1.1GeV, and are shown in Fig.2. Our results for the total scattering cross section $\sigma$ without the RPA effects (dotted line) in the case of charged current quasielastic scattering is in fair agreement with the calculation done in Fermi Gas Model Moniz ,llewelyn ,Gaisser in the energy region of 0.4-3 GeV. When RPA correlations are included in our model, we see that they give rise to a reduction in the total cross section $\sigma(E)$ (solid line) which is 12-15$\%$ in the energy region of 0.4-3 GeV. In the case of charged current induced incoherent and coherent pion productions the results for the total cross sections are obtained from Eq.(6) and Eq.(14) respectively using the N-$\Delta$ transition form factors given by Lalakulich et al.Lalakulich1 with $M_{A}$=1.1 GeV. The numerical results are shown in Fig.3 and Fig.4 for the incoherent and coherent pion productions. We see that the effect of nuclear medium in the pion production process as well as the final state interaction of pions both reduce the cross section. In the case of incoherent production of pions, the reduction due to nuclear medium effects in the production process is larger than due to final state interaction while in the case of coherent pion production, the reduction due to final state interaction is quite large as compared to the reduction due to the nuclear medium effects. However, the contribution of coherent process to the total pion production is small(5-6$\%$) in the energy region of $0.4GeV<E<3GeV$, therefore, the reduction in the total cross section is mainly given by the reduction in the cross section of the incoherent process. In Fig.5, we show the results for the ratio of total one pion production and the quasielastic production cross sections for $\nu_{\mu}$ induced reaction in polystyrene($C_{8}H_{8}$), i.e. $R(E)=\frac{\sigma^{CC1\pi^{+}}(E)}{\sigma^{CCQE}(E)}$ (solid line) as a function of neutrino energy. A comparison with the experimental results of K2K Rodriguez and ANL Radecky , Barish for R(E) are also shown in the figure. In this figure, we also show the ratio R(E)(dashed line) when no nuclear medium effect is taken into account either in the quasielastic process or in the inelastic process. When the cross sections are calculated without any nuclear medium modifications except the Fermi motion and Pauli principle for the quasielastic process and without any modification of $\Delta$ properties in the nuclear medium in case of inelastic process, the results for R(E) are shown by dashed-double dotted line. This should be compared with the Monte Carlo predictions used in K2K analysisRodriguez . We see from Figs. 2-5 that the nuclear medium modifications play an important role in quasielastic and inelastic reactions and tend to reduce the cross section ratio R(E). We have also studied the effect of the uncertainties in the ratio R(E) due to the use of various vector form factors given by Budd et al. BBBA2 , Bosted BBBA3 and dipole form factorsGalster in the quasielastic case and the N-$\Delta$ transition form factors given by Schreiner & von HippelSchreiner and Paschos et al.pas in the inelastic case. The effect of varying the axial dipole mass $(M_{A})$ in the parameterization of axial form factor in quasielastic as well as in the inelastic processes in the region of $1.05<M_{A}<1.21$ GeV on the ratio R(E) has also been studied. These effects lead to a small change of about 4-6$\%$ in the ratio R(E). ### IV.2 ANGULAR AND MOMENTUM DISTRIBUTIONS OF LEPTONS AND PIONS In Figs. 6 & 7, we show the effect of nuclear medium on $Q^{2}$ distribution and the momentum distribution of the leptons averaged over the neutrino flux at K2K in the case of neutrino induced charged current quasielastic scattering on ${}^{12}C$ target. A reduction in $Q^{2}$ distribution in the peak region is found to be about 30$\%$ when RPA effects are taken into account while this reduction is about 12$\%$ in the momentum distribution of leptons in the peak region. In the case of incoherent one pion production process, the $Q^{2}$ distribution as well as the momentum distribution of leptons are shown in Figs.8 & 9\. In these reactions charge pions are also produced and we show in Figs.10 & 11, the angular distribution as well as momentum distribution of pions and the effect of nuclear medium on these distributions. We find that the reduction in $Q^{2}$ distribution in the peak region is around 30$\%$ when nuclear medium effects are taken into account which further reduces by about 14$\%$ when pion absorption effects are incorporated. For the pion momentum distribution the reduction due to nuclear medium effects is around 40$\%$ which further reduces by about 15$\%$ when pion absorption effects are taken into account. We have not shown these results for the coherent process as the contribution to the total pion production events have been found to be quite small. However, these results are interesting in their own right and will be reported elsewhere. To conclude, we have studied in this paper the effect of nuclear medium on the inclusive quasielastic and one pion charged current production in nuclei and compared our results with the recent experimental results from the K2K experiment Rodriguez . The nuclear medium effects are important in this energy region and reduce the ratio by 30$\%$ around the energy region of 1GeV which becomes 20-25$\%$ in the neutrino energy region of 2-3GeV. The results for nuclear medium effects on the $Q^{2}$ and the momentum distribution of lepton in the quasielastic as well as inelastic reaction and the results for the nuclear medium effects on the angular and momentum distribution of pions have also been presented. ## V ACKNOWLEDGMENTS One of the authors(M. S. A.) is thankful to T. Kajita and Y. Hayato (I. C. R. R., University of Tokyo) for many useful discussions and the warm hospitality provided during his stay at ICRR where part of this work was done. S. C. is thankful to the Jawaharlal Nehru Memorial Fund for the Doctoral Fellowship. ## References * (1) A. Rodriguez et al. (K2K Collaboration), Phys. Rev. D 78, 032003 (2008). * (2) M. O. Wascko (MiniBooNE Collaboration), Nucl. Phys. Proc. Suppl. 159, 50 (2006). * (3) Y. Obayashi (T2K Collaboration), arXiv:hep-ex/0807.4012v2(2008), Y. Itow et al., arXiv:hep-ex/0106019(2001). * (4) D. S. Ayres et al. (NO$\nu$A Collaboration), arXiv:hep-ex/0503053(2005). * (5) D. Rein and L. M. Sehgal, Annals. Phys. 79, 133 (1981); Nucl. Phys. B 223, 29 (1983). * (6) Y. Hayato, Nucl. Phys. Proc. Suppl. B 112, 171 (2002). * (7) R. A. Smith and E. J. Moniz, Nucl. Phys. B 43, 605 (1972). * (8) C. H. Llewellyn Smith, Phys. Rep. 3, 261 (1972). * (9) T. K. Gaisser and J. S. O’Connell, Phys. Rev. D 34, 822 (1986). * (10) T. Kuramoto, M. Fukugita, Y. Kohyama and K. Kubodera, Nucl. Phys. A 512, 711 (1990). * (11) S. K. Singh and E. Oset, Nucl. Phys. A 542, 587 (1992); Phys. Rev. C 48, 1246 (1993). * (12) H. Kim, J. Piekarewicz and C. J. Horowitz, Phys. Rev. C 51, 2739 (1995). * (13) N. Auerbach, N. Van Giai and O. K. Vorov, Phys. Rev. C 56, 2368 (1997). * (14) A. C. Hayes and I. S. Towner, Phys. Rev. C 61, 044603 (2000). * (15) J. Marteau, J. Delorme and M. Ericson, Nucl. Instrum. Meth. A 451, 76 (2000). * (16) O. Benhar, N. Farina, H. Nakamura, M. Sakuda and R. Seki, Phys. Rev. D 72, 053005 (2005). * (17) J. Nieves, J. E. Amaro and M. Valverde, Phys. Rev. C 70, 055503 (2004); J. Nieves, M. Valverde and M. J. Vicente Vacas, Phys. Rev. C 73, 025504 (2006). * (18) A. Meucci, C. Giusti and F. D. Pacati, Nucl. Phys. A 739, 277 (2004). * (19) M. Sajjad Athar, S. Ahmad and S. K. Singh, Nucl. Phys. A 764, 551 (2006). * (20) C. Giusti, A. Meucci and F.D. Pacati, arXiv:nucl-th/0607037(2006). * (21) M. C. Martinez, P. Lava, N. Jachowicz, J. Ryckebusch, K. Vantournhout, J.M. Udias, Phys. Rev. C 73, 024607 (2006). * (22) J. A. Caballero, J. E. Amaro, M. B. Barbaro, T. W. Donnelly and J. M. Udias, Phys. Lett. B 653, 366 (2007). * (23) A. V. Butkevich and S. A. Kulagin, arXiv:nucl-th/0705.1051(2007). * (24) M. Martini, G. Co’, M. Anguiano and A. M. Lallena, Phys. Rev. C 75, 034604 (2007). * (25) J. E. Amaro, M. B. Barbaro, J. A. Caballero, T. W. Donnelly and J. M. Udias, Phys. Rev. C 75, 034613 (2007). * (26) M. Sajjad Athar, S. Ahmad and S.K.Singh, Eur. Phys. J. A 24, 459 (2005); S. Ahmad, M. Sajjad Athar and S. K. Singh, Phys. Rev. D 74, 073008 (2006). * (27) T. Leitner, L. Alvarez-Ruso and U. Mosel, Phys. Rev. C 73, 065502 (2006). * (28) O. Benhar and D. Meloni, Nucl. Phys. A 789, 379 (2007). * (29) H. Kim, S. Schramm and C. J. Horowitz, Phys. Rev. C 53, 3131 (1996). * (30) S. K. Singh, M. J. Vicente Vacas and E. Oset, Phys. Lett. B 416, 23 (1998). * (31) J. Marteau, Eur. Phys. J. A 5, 183 (1999). * (32) E. A. Paschos, J. Y. Yu and M. Sakuda, Phys. Rev. D 69, 014013 (2004). * (33) W. Cassing, M. Kant, K. Langanke and P. Vogel, Phys. Lett. B 639, 32 (2006). * (34) E. A. Paschos, I. Schienbein and J.-Y. Yu, arXiv:hep-ph/0704.1991 (2007). * (35) C. Praet, O. Lalakulich, N. Jachowicz and J. Ryckebusch, arXiv:nucl-th/0804.2750 (2008) . * (36) A. Kartavtsev, E. A. Paschos and G. J. Gounaris, Phys. Rev. D 74, 054007 (2006). * (37) S. K. Singh, M. Sajjad Athar and S. Ahmad, Phys. Rev. Lett., 96, 241807 (2006). * (38) L. Alvarez-Ruso, L. S. Geng, S. Hirenzaki and M. J. Vicente Vacas, Phys. Rev. C 75, 055501 (2007). * (39) M. J. Vicente Vacas, M. Kh. Khankhasayev and S. G. Mashnik, arXiv:nucl-th/9412023(1994). * (40) R. C. Carrasco, J. Nieves and E. Oset, Nucl. Phys. A 565, 797 (1993). * (41) E. Oset. H. Toki and W. Weise, Phys. Rep. 83, 281 (1982). * (42) R. Bradford, A. Bodek, H. Budd and J. Arrington, Nucl. Phys. Proc. Suppl. 159, 127 (2006). * (43) E. Oset, P. Fernandez de Cordoba, L. L. Salcedo and R. Brockmann, Phys. Rep. 188, 79 (1990); E. Oset, D. Strottman, H. Toki and J. Navarro, Phys. Rev. C 48, 2395 (1993). * (44) E. Oset and L. L. Salcedo, Nucl. Phys. A 468, 631 (1987); C. Garcia Recio, E. Oset, L. L. Salcedo, D. Strottman, M. J. Lopez, Nucl. Phys. A 526, 685 (1991). * (45) C. W. de Jager, H. de Vries and C. de Vries, At. Data Nucl. Data Tables 14, 479 (1974). * (46) O. Lalakulich, E. A. Paschos and G. Piranishvili, Phys. Rev. D 74, 014009 (2006). * (47) G. Penner and U. Mosel, Phys. Rev. C 66, 055211 (2002); M. Post, Ph.D. Thesis, Universitat Giessen, 2004(unpublished). * (48) G. M. Radecky et al., Phys. Rev. D 25, 1161 (1982). * (49) S. J. Barish, A. Engler, R. W. Kraemer, K. Miller and B. J. Stacey, Phys. Rev. D 16, 3103 (1977). * (50) H. Budd, A. Bodek and J. Arrington, arXiv:hep-ex/0308005. * (51) P. E. Bosted, Phys. Rev. C 51, 409 (1995). * (52) S. Galster, H. Klein, J. Moritz, K. H. Schmidt, D. Wegener and J. Bleckwenn, Nucl. Phys. B 32, 221 (1971). * (53) P. A. Schreiner and F. von Hippel, Nucl. Phys. B 58, 333 (1973).	arxiv-papers	2008-08-15T07:02:46	2024-09-04T02:48:57.312654	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Sajjad Athar, S. Chauhan and S. K. Singh", "submitter": "Shri Singh krishna", "url": "https://arxiv.org/abs/0808.2103" }
0808.2113	# Exclusive semileptonic decays of $\Lambda_{b}\to\Lambda l^{+}l^{-}$ in supersymmetric theories M. Jamil Aslam1,2 Yu-Ming Wang1 Cai-Dian Lü 1 1Institute of High Energy Physics, P.O. Box 918(4), Beijing 100049, China 2National Center for Physics, Quaid-i-Azam University, Islamabad, Pakistan ###### Abstract The weak decays of $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=e,\,\,\mu$) are investigated in Minimal Supersymmetric Standard Model (MSSM) and also in Supersymmetric (SUSY) SO(10) Grand Unified Models. In MSSM the special attention is paid to the Neutral Higgs Bosons (NHBs) as they make quite a large contribution in exclusive $B\to X_{s}l^{+}l^{-}$ decays at large $\tan{\beta}$ regions of parameter space of SUSY models, since part of SUSY contributions is proportional to $\tan^{3}{\beta}$. The analysis of decay rate, forward-backward asymmetries, lepton polarization asymmetries and the polarization asymmetries of $\Lambda$ baryon in $\Lambda_{b}\to\Lambda l^{+}l^{-}$ show that the values of these physical observables are greatly modified by the effects of NHBs. In SUSY SO(10) GUT model, the new physics contribution comes from the operators which are induced by the NHBs penguins and also from the operators having chirality opposite to that of the corresponding SM operators. SUSY SO(10) effects show up only in the decay $\Lambda_{b}\to\Lambda+\tau^{+}\tau^{-}$ where the longitudinal and transverse lepton polarization asymmetries are deviate significantly from the SM value while the effects in the decay rate, forward-backward asymmetries and polarization asymmetries of final state $\Lambda$ baryon are very mild. The transverse lepton polarization asymmetry in $\Lambda_{b}\to\Lambda+\tau^{+}\tau^{-}$ is almost zero in SM and in MSSM model. However, it can reach to $-0.1$ in SUSY SO(10) GUT model and could be seen at the future colliders; hence this asymmetry observable will provide us useful information to probe new physics and discriminate between different models. ###### pacs: 13.30.Ce, 14.20.Mr, 11.30.Pb ## I Introduction From last decade, rare decays induced by flavor changing neutral currents (FCNCs) $b\to s\,l^{+}l^{-}$ have become the main focus of the studies due to the CLEO measurement of the radiative decay $b\to s\gamma$ CLEO1 . In the standard model (SM) these decays are forbidden at tree level and can only be induced by Glashow-Iliopoulos-Maiani mechanism Glashow:1970gm via loop diagrams. Hence, such decays will provide helpful information about the parameters of Cabbibo-Kobayashi-Maskawa (CKM) matrix CKM 1 ; CKM 2 elements as well as various hadronic form factors. In the literature there have been intensive studies on the exclusive decays $B\to P(V,A)\,l^{+}l^{-}$ b to s in theory 16 ; b to s in theory 17 ; b to s in theory 18 ; b to s in theory 19 ; b to s in theory 20 ; b to s in theory 21 ; jamil1 both in the SM and beyond, where the notions $P,V$ and $A$ denote the pseudoscalar, vector and axial vector mesons respectively. It is generally believed that supersymmetry (SUSY) is not only one of the strongest competitor of the SM but is also the most promising candidate of new physics. The reason is that it offers a unique scheme to embed the SM in a more fundamental theory where many theoretical problems such as gauge hierarchy, origin of mass, and Yukawa couplings can be resolved. One direct way to search for SUSY is to discover SUSY particles at high energy colliders, but unfortunately, so far no SUSY particles have been found. Another way is to search for its effects through indirect methods. The measurement of invariant mass spectrum, forward-backward asymmetry and polarization asymmetries are the suitable tools to probe new physics effects. For most of the SUSY models, the SUSY contributions to an observable appear at loop level due to the $\ R$-parity conservation. Therefore, it has been realized for a long time that rare processes can be used as a good probe for the searches of SUSY, since in these processes the contributions of SUSY and SM arises at the same order in perturbation theory Yan . Motivated from the fact that in two Higgs doublet model and in other SUSY models, Neutral Higgs Bosons (NHBs) could contribute largely to the inclusive processes $B\to X_{s}l^{+}l^{-}$, as part of supersymmetric contributions is proportional to the $\tan^{3}{\beta}$ Huang . Subsequently, the physical observables, like branching ratio and forward-backward asymmetry, in the large $\mathrm{\tan{\beta}}$ region of parameter space in SUSY models can be quite different from that in the SM. In addition, similar effects in exclusive $B\to K(K^{})\,\ l^{+}l^{-}$ decay modes are also investigated Yan , where the analysis of decay rates, forward-backward asymmetries and polarization asymmetries of final state lepton indicates the significant role of NHBs. It is believed that physics beyond the SM is essential to explain the problem of neutrino oscillation. To this purpose, a number of SUSY SO(10) models have been proposed in the literature Babu ; Mahanthappa ; Gut ; Senjanovic . One such model is the SUSY SO(10) Grand Unified Models (GUT), in which there is a complex flavor non-diagonal down-type squark mass matrix element of 2nd and 3rd generations of order one at the GUT scale Gut . This can induce large flavor off-diagonal coupling such as the coupling of gluino to the quark and squark which belong to different generations. These couplings are in general complex and may contribute to the process of flavor changing neutral currents (FCNCs). The above analysis of physical observables in $B\to K(K^{})\,\ l^{+}l^{-}$ decay is extended in SUSY SO(10) GUT model in Ref. Li . It is believed that the effects of the counterparts of usual chromo-magnetic and electromagnetic dipole moment operators as well as semileptonic operators with opposite chirality are suppressed by $m_{s}/m_{b}$ in the SM, but in SUSY SO(10) GUTs their effect can be significant, since $\delta^{dRR}_{23}$ can be as large as 0.5 Li ; Gut . Apart from this, $\delta^{dRR}_{23}$ can induce new operators as the counterparts of usual scalar operators in SUSY models due to NHB penguins with gluino-down type squark propagator in the loop. It has been shown Li that the forward-backward asymmetries as well as the longitudinal and transverse decay widths of $B\to K(K^{})\,\ l^{+}l^{-}$ decay, are sensitive to these NHBs effect in SUSY SO(10) GUT model which can be detected in the future $B$ factories. Compared to the $B$ meson decays, the investigations of FCNC $b\rightarrow s$ transition for bottom baryon decays $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ are much behind because more degrees of freedom are involved in the bound state of baryon system at the quark level. From the experimental point of view, the only drawback of bottom baryon decays is that the production rate of $\Lambda_{b}$ baryon in $b$ quark hadronization is about four times less than that of the $B$ meson. Theoretically, the major interest in baryonic decays can be attributed to the fact that they can offer a unique ground to extract the helicity structure of the effective Hamiltonian for $b\rightarrow s$ transition in the SM and beyond, which is lost in the hadronization of mesonic case. The key issue in the study of exclusive baryonic decays is to properly evaluate the hadronic matrix elements for $\Lambda_{b}\to\Lambda$, namely the transition form factors which are obviously governed by non-perturbative QCD dynamics. Currently, there has been some studies in the literature on $\Lambda_{b}\rightarrow\Lambda$ transition form factors in different models including pole model (PM) Mannel , covariant oscillator quark model (COQM)Mohanta , MIT bag model (BM)Cheng and non-relativistic quark model Cheng 2 , QCD sum rule approach (QCDSR) Huang1 , perturbative QCD (pQCD) approach HLLW and also in the light-cone sum rules approach (LCSR) YuMing . Using these form factors, the physical observables like decay rates, forward- backward asymmetries and polarization asymmetries of $\Lambda$ baryon as well as of the final state leptons in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ were studied in great details in the literature c.q. geng 1 ; c.q. geng 2 ; c.q. geng 3 ; c.q. geng 4 ; Aliev 1 ; Aliev 2 ; Aliev 3 ; Aliev 4 . It is pointed out that these observables are very sensitive to the new physics, for instance, the polarization asymmetries of $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decays heavily depend on the right handed current, which is much suppressed in the SM c.q. geng 4 . In this paper, we will investigate the exclusive decay $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ( $l$ = $\mu$, $\tau$) both in the Minimal Supersymmetric Standard Model (MSSM) as well as in the SUSY SO(10) GUT model Gut . We evaluate the branching ratios, forward-backward asymmetries, lepton polarization asymmetries and polarization asymmetries of $\Lambda$ baryon with special emphasis on the effects of NHBs in MSSM. It is pointed out that different source of the vector current could manifest themselves in different regions of phase space. For low value of momentum transfer, the photonic penguin dominates, while the $Z$ penguin and $W$ box become important towards high value of momentum transfer Yan . In order to search the region of momentum transfer with large contributions from NHBs, the above decay in certain large $\mathrm{{tan{\ \beta}}}$ region of parameter space has been analyzed in SuperGravity (SUGRA) and M-theory inspired models CSHuang . We extend this analysis to the SUSY SO(10) GUT model Yan , where there are some primed counterparts of the usual SM operators. For instance, the counterparts of usual operators in $\ B\rightarrow\ X_{s}\ \gamma$ decay are suppressed by $m_{s}/m_{b}$ and consequently negligible in the SM because they have opposite chiralities. These operators are also suppressed in Minimal Flavor Violating (MFV) models Bobeth ; Wu , however, in SUSY SO(10) GUT model their effects can be significant. The reason is that the flavor non-diagonal squark mass matrix elements are the free parameters and some of them have significant effects in rare decays of $B$ mesons Lunghi . In our numerical analysis for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decays, we shall use the results of the form factors calculated by LCSR approach in Ref. YuMing , and the values of the relevant Wilson coefficient for MSSM and SUSY SO(10) GUT models are borrowed from Ref. Yan ; Li . The effects of SUSY contributions to the decay rate and zero position of forward-backward asymmetry are also explored in this work. Our results show that not only the decay rates are sensitive to the NHBs contribution but the zero point of the forward-backward asymmetry also shifts remarkably. It is known that the hadronic uncertainties associated with the form factors and other input parameters have negligible effects on the lepton polarization asymmetries and polarization asymmetries of $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decays. We have also studied these asymmetries in the SUSY models mentioned above and found that the effects of NHBs are quite significant in some regions of parameter space of SUSY. The paper is organized as follows. In Sec. II, we present the effective Hamiltonian for the dilepton decay $\Lambda_{b}\to\Lambda l^{+}l^{-}$. Section III contains the definitions and numbers of the form factors for the said decay using the LCSR approach. In Sec. IV we present the basic formulas of physical observables like decays rate, forward-backward asymmetries (FBAs) and polarization asymmetries of lepton and that of the $\Lambda$ baryon in $\Lambda_{b}\to\Lambda l^{+}l^{-}$. Section V is devoted to the numerical analysis of these observables and the brief summary and concluding remarks are given in Sec. VI. ## II Effective Hamiltonian After integrating out the heavy degrees of freedom in the full theory, the general effective Hamiltonian for $\ b\rightarrow\ sl^{+}l^{-}$ in SUSY SO(10) GUT model, can be written as Li $\displaystyle H_{eff}$ $\displaystyle=$ $\displaystyle-\frac{G_{F}}{2\sqrt{2}}V_{tb}V_{ts}^{\ast}\bigg{[}{\sum\limits_{i=1}^{2}}C_{i}({\mu})O_{i}({\mu})+{\sum\limits_{i=3}^{10}(}C_{i}({\mu})O_{i}({\mu})+C_{i}^{\prime}({\mu})O_{i}^{\prime}({\mu}))$ (1) $\displaystyle+\sum\limits_{i=1}^{8}{(}C_{Q_{i}}({\mu})Q_{i}({\mu})+C_{Q_{i}}^{\prime}({\mu})Q_{i}^{\prime}({\mu}))\bigg{]},$ where $O_{i}({\mu})$ $(i=1,\ldots,10)$ are the four-quark operators and $C_{i}({\mu})$ are the corresponding Wilson coefficients at the energy scale ${\mu}$ Goto . Using renormalization group equations to resume the QCD corrections, Wilson coefficients are evaluated at the energy scale ${\mu=m}_{b}$. The theoretical uncertainties associated with the renormalization scale can be substantially reduced when the next-to-leading- logarithm corrections are included Buchalla . The new operators $Q_{i}({\mu})$ $(i=1,\ldots,8)$ come from the NHBs exchange diagrams, whose manifest forms and corresponding Wilson coefficients can be found in Ewerth ; Feng . The primed operators are the counterparts of the unprimed operators, which can be obtained by flipping the chiralities in the corresponding unprimed operators. It needs to point out that these primed operators will appear only in SUSY SO(10) GUT model and are absent in SM and MSSM Yan . The explicit expressions of the operators responsible for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ transition are given by $\displaystyle O_{7}$ $\displaystyle=$ $\displaystyle\frac{e^{2}}{16\pi^{2}}m_{b}\left(\bar{s}\sigma_{\mu\nu}P_{R}b\right)F^{\mu\nu},\,\qquad O_{7}^{\prime}=\frac{e^{2}}{16\pi^{2}}m_{b}\left(\bar{s}\sigma_{\mu\nu}P_{L}b\right)F^{\mu\nu}$ $\displaystyle O_{9}$ $\displaystyle=$ $\displaystyle\frac{e^{2}}{16\pi^{2}}(\bar{s}\gamma_{\mu}P_{L}b)(\bar{l}\gamma^{\mu}l),\,\qquad\ \ \ O_{9}^{\prime}=\frac{e^{2}}{16\pi^{2}}(\bar{s}\gamma_{\mu}P_{R}b)(\bar{l}\gamma^{\mu}l)$ $\displaystyle O_{10}$ $\displaystyle=$ $\displaystyle\frac{e^{2}}{16\pi^{2}}(\bar{s}\gamma_{\mu}P_{L}b)(\bar{l}\gamma^{\mu}\gamma_{5}l),\,\ \ \ \ \ \ O_{10}^{\prime}=\frac{e^{2}}{16\pi^{2}}(\bar{s}\gamma_{\mu}P_{R}b)(\bar{l}\gamma^{\mu}\gamma_{5}l)$ $\displaystyle Q_{1}$ $\displaystyle=$ $\displaystyle\frac{e^{2}}{16\pi^{2}}(\bar{s}P_{R}b)(\bar{l}l),\qquad\qquad\ \ \ \ \ Q_{1}^{\prime}=\frac{e^{2}}{16\pi^{2}}(\bar{s}P_{L}b)(\bar{l}l)$ $\displaystyle Q_{2}$ $\displaystyle=$ $\displaystyle\frac{e^{2}}{16\pi^{2}}(\bar{s}P_{R}b)(\bar{l}\gamma_{5}l),\qquad\ \ \ \ \ \ \ \ Q_{2}^{\prime}=\frac{e^{2}}{16\pi^{2}}(\bar{s}P_{L}b)(\bar{l}\gamma_{5}l)$ (2) with $P_{L,R}=\left(1\pm\gamma_{5}\right)/2$. In terms of the above Hamiltonian, the free quark decay amplitude for $b\rightarrow s$ $l^{+}l^{-}$ can be derived as Huang : $\displaystyle\mathcal{M}(b$ $\displaystyle\rightarrow$ $\displaystyle sl^{+}l^{-})=-\frac{G_{F}\alpha}{\sqrt{2}\pi}V_{tb}V_{ts}^{\ast}\bigg{\\{}C_{9}^{eff}(\bar{s}\gamma_{\mu}P_{L}b)(\bar{l}\gamma^{\mu}l)+C_{10}(\bar{s}\gamma_{\mu}P_{L}b)(\bar{l}\gamma^{\mu}\gamma_{5}l)$ $\displaystyle-2m_{b}C_{7}^{eff}(\bar{s}i\sigma_{\mu\nu}\frac{q^{\nu}}{s}P_{R}b)(\bar{l}\gamma^{\mu}l)+C_{Q_{1}}(\bar{s}P_{R}b)(\bar{l}l)+C_{Q_{2}}(\bar{s}P_{R}b)(\bar{l}\gamma_{5}l)+(C_{i}(m_{b})\longleftrightarrow C_{i}^{\prime}(m_{b}))\bigg{\\}}$ where $s=q^{2}$ and $q=p_{\Lambda_{b}}-p_{\Lambda}$ is the momentum transfer. Due to the absence of $Z$ boson in the effective theory, the operator $O_{10}$ can not be induced by the insertion of four-quark operators. Therefore, the Wilson coefficient $C_{10}$ does not renormalize under QCD corrections and hence it is independent on the energy scale. Moreover, the above quark level decay amplitude can receive additional contributions from the matrix element of four-quark operators, $\sum_{i=1}^{6}\langle l^{+}l^{-}s\|O_{i}\|b\rangle$, which are usually absorbed into the effective Wilson coefficient $C_{9}^{eff}(\mu)$. To be more specific, we can decompose $C_{9}^{eff}(\mu)$ into the following three parts b to s in theory 3 ; b to s in theory 4 ; b to s in theory 5 ; b to s in theory 6 ; b to s in theory 7 ; b to s in theory 8 ; b to s in theory 9 $C_{9}^{eff}(\mu)=C_{9}(\mu)+Y_{SD}(z,s^{\prime})+Y_{LD}(z,s^{\prime}),$ where the parameters $z$ and $s^{\prime}$ are defined as $z=m_{c}/m_{b},\,\,\,s^{\prime}=q^{2}/m_{b}^{2}$. $Y_{SD}(z,s^{\prime})$ describes the short-distance contributions from four-quark operators far away from the $c\bar{c}$ resonance regions, which can be calculated reliably in the perturbative theory. The long-distance contributions $Y_{LD}(z,s^{\prime})$ from four-quark operators near the $c\bar{c}$ resonance cannot be calculated from first principles of QCD and are usually parameterized in the form of a phenomenological Breit-Wigner formula making use of the vacuum saturation approximation and quark-hadron duality. The manifest expressions for $Y_{SD}(z,s^{\prime})$ and $Y_{LD}(z,s^{\prime})$ can be written as YuMing ; c.q. geng 1 ; c.q. geng 2 ; c.q. geng 3 ; c.q. geng 4 $\displaystyle Y_{SD}(z,s^{\prime})$ $\displaystyle=$ $\displaystyle h(z,s^{\prime})(3C_{1}(\mu)+C_{2}(\mu)+3C_{3}(\mu)+C_{4}(\mu)+3C_{5}(\mu)+C_{6}(\mu))$ (4) $\displaystyle-\frac{1}{2}h(1,s^{\prime})(4C_{3}(\mu)+4C_{4}(\mu)+3C_{5}(\mu)+C_{6}(\mu))$ $\displaystyle-\frac{1}{2}h(0,s^{\prime})(C_{3}(\mu)+3C_{4}(\mu))+{\frac{2}{9}}(3C_{3}(\mu)+C_{4}(\mu)+3C_{5}(\mu)+C_{6}(\mu)),$ $\displaystyle Y_{LD}(z,s^{\prime})$ $\displaystyle=$ $\displaystyle\frac{3}{\alpha_{em}^{2}}(3C_{1}(\mu)+C_{2}(\mu)+3C_{3}(\mu)+C_{4}(\mu)+3C_{5}(\mu)+C_{6}(\mu))$ (5) $\displaystyle\sum_{j=\psi,\psi^{\prime}}\omega_{j}(q^{2})k_{j}\frac{\pi\Gamma(j\rightarrow l^{+}l^{-})M_{j}}{q^{2}-M_{j}^{2}+iM_{j}\Gamma_{j}^{tot}},$ with $\displaystyle h(z,s^{\prime})$ $\displaystyle=$ $\displaystyle-{\frac{8}{9}}\mathrm{ln}z+{\frac{8}{27}}+{\frac{4}{9}}x-{\frac{2}{9}}(2+x)\|1-x\|^{1/2}\left\\{\begin{array}[]{l}\ln\left\|\frac{\sqrt{1-x}+1}{\sqrt{1-x}-1}\right\|-i\pi\quad\mathrm{for}{{\ }x\equiv 4z^{2}/s^{\prime}<1}\\\ 2\arctan\frac{1}{\sqrt{x-1}}\qquad\mathrm{for}{{\ }x\equiv 4z^{2}/s^{\prime}>1}\end{array}\right.,$ (8) $\displaystyle h(0,s^{\prime})$ $\displaystyle=$ $\displaystyle{\frac{8}{27}}-{\frac{8}{9}}\mathrm{ln}{\frac{m_{b}}{\mu}}-{\frac{4}{9}}\mathrm{ln}s^{\prime}+{\frac{4}{9}}i\pi\,\,.$ (9) The non-factorizable effects b to s 1 ; b to s 2 ; b to s 3 ; NF charm loop from the charm loop can bring about further corrections to the radiative $b\rightarrow s\gamma$ transition, which can be absorbed into the effective Wilson coefficient $C_{7}^{eff}$. Specifically, the Wilson coefficient $C^{eff}_{7}$ is given by c.q. geng 4 $C_{7}^{eff}(\mu)=C_{7}(\mu)+C_{b\rightarrow s\gamma}(\mu),$ with $\displaystyle C_{b\rightarrow s\gamma}(\mu)$ $\displaystyle=$ $\displaystyle i\alpha_{s}\bigg{[}{\frac{2}{9}}\eta^{14/23}(G_{1}(x_{t})-0.1687)-0.03C_{2}(\mu)\bigg{]},$ (10) $\displaystyle G_{1}(x)$ $\displaystyle=$ $\displaystyle{\frac{x(x^{2}-5x-2)}{8(x-1)^{3}}}+{\frac{3x^{2}\mathrm{ln}^{2}x}{4(x-1)^{4}}},$ (11) where $\eta=\alpha_{s}(m_{W})/\alpha_{s}(\mu)$, $x_{t}=m_{t}^{2}/m_{W}^{2}$, $C_{b\rightarrow s\gamma}$ is the absorptive part for the $b\rightarrow sc\bar{c}\rightarrow s\gamma$ rescattering and we have dropped out the tiny contributions proportional to CKM sector $V_{ub}V_{us}^{\ast}$. In addition, $C_{7}^{\prime eff}(\mu)$ and $C_{9}^{\prime eff}(\mu)$ can be obtained by replacing the unprimed Wilson coefficients with the corresponding prime ones in the above formula. ## III Matrix elements and form factors in Light Cone Sum Rules With the free quark decay amplitude available, we can proceed to calculate the decay amplitudes for $\Lambda_{b}\to\Lambda\gamma$ and $\Lambda_{b}\to\Lambda l^{+}l^{-}$ at hadron level, which can be obtained by sandwiching the free quark amplitudes between the initial and final baryon states. Consequently, the following four hadronic matrix elements $\displaystyle\langle\Lambda(P)\|\bar{s}\gamma_{\mu}b\|\Lambda_{b}(P+q)\rangle$ , $\displaystyle\,\,\,\langle\Lambda(P)\|\bar{s}\gamma_{\mu}\gamma_{5}b\|\Lambda_{b}(P+q)\rangle,$ $\displaystyle\langle\Lambda(P)\|\bar{s}\sigma_{\mu\nu}b\|\Lambda_{b}(P+q)\rangle$ , $\displaystyle\,\,\,\langle\Lambda(P)\|\bar{s}\sigma_{\mu\nu}\gamma_{5}b\|\Lambda_{b}(P+q)\rangle,$ (12) need to be computed. Generally, the above matrix elements can be parameterized in terms of the form factors as c.q. geng 4 ; Aliev 1 ; Aliev 2 ; Aliev 3 ; Aliev 4 : $\displaystyle\langle\Lambda(P)\|\bar{s}\gamma_{\mu}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)(g_{1}\gamma_{\mu}+g_{2}i\sigma_{\mu\nu}q^{\nu}+g_{3}q_{\mu})\Lambda_{b}(P+q),\,\,$ (13) $\displaystyle\langle\Lambda(P)\|\bar{s}\gamma_{\mu}\gamma_{5}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)(G_{1}\gamma_{\mu}+G_{2}i\sigma_{\mu\nu}q^{\nu}+G_{3}q_{\mu})\gamma_{5}\Lambda_{b}(P+q),\,\,$ (14) $\displaystyle\langle\Lambda(P)\|\bar{s}\sigma_{\mu\nu}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)[h_{1}\sigma_{\mu\nu}-ih_{2}(\gamma_{\mu}q_{\nu}-\gamma_{\nu}q_{\mu})$ (15) $\displaystyle- ih_{3}(\gamma_{\mu}P_{\nu}-\gamma_{\nu}P_{\mu})-ih_{4}(P_{\mu}q_{\nu}-P_{\nu}q_{\mu})]\Lambda_{b}(P+q),\,\,$ $\displaystyle\langle\Lambda(P)\|\bar{s}\sigma_{\mu\nu}\gamma_{5}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)[H_{1}\sigma_{\mu\nu}-iH_{2}(\gamma_{\mu}q_{\nu}-\gamma_{\nu}q_{\mu})$ (16) $\displaystyle- iH_{3}(\gamma_{\mu}P_{\nu}-\gamma_{\nu}P_{\mu})-iH_{4}(P_{\mu}q_{\nu}-P_{\nu}q_{\mu})]\gamma_{5}\Lambda_{b}(P+q),$ where all the form factors $g_{i}$, $G_{i}$, $h_{i}$ and $H_{i}$ are functions of the square of momentum transfer $q^{2}$. Contracting Eqs. (15-16) with the four momentum $q^{\mu}$ on both side and making use of the equations of motion $\displaystyle q^{\mu}(\bar{\psi}_{1}\gamma_{\mu}\psi_{2})$ $\displaystyle=$ $\displaystyle(m_{1}-m_{2})\bar{\psi}_{1}\psi_{2}$ (17) $\displaystyle q^{\mu}(\bar{\psi}_{1}\gamma_{\mu}\gamma_{5}\psi_{2})$ $\displaystyle=$ $\displaystyle-(m_{1}+m_{2})\bar{\psi}_{1}\gamma_{5}\psi_{2}$ (18) we have $\displaystyle\langle\Lambda(P)\|\bar{s}i\sigma_{\mu\nu}q^{\nu}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)(f_{1}\gamma_{\mu}+f_{2}i\sigma_{\mu\nu}q^{\nu}+f_{3}q_{\mu})\Lambda_{b}(P+q),\,\,$ (19) $\displaystyle\langle\Lambda(P)\|\bar{s}i\sigma_{\mu\nu}\gamma_{5}q^{\nu}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)(F_{1}\gamma_{\mu}+F_{2}i\sigma_{\mu\nu}q^{\nu}+F_{3}q_{\mu})\gamma_{5}\Lambda_{b}(P+q),$ (20) with $\displaystyle f_{1}$ $\displaystyle=$ $\displaystyle{\frac{2h_{2}-h_{3}+h_{4}(m_{\Lambda_{b}}+m_{\Lambda})}{2}}q^{2},$ (21) $\displaystyle f_{2}$ $\displaystyle=$ $\displaystyle{\frac{2h_{1}+h_{3}(m_{\Lambda}-m_{\Lambda_{b}})+h_{4}q^{2}}{2}},$ (22) $\displaystyle f_{3}$ $\displaystyle=$ $\displaystyle{\frac{m_{\Lambda}-m_{\Lambda_{b}}}{q^{2}}}f_{1},$ (23) $\displaystyle F_{1}$ $\displaystyle=$ $\displaystyle{\frac{2H_{2}-H_{3}+H_{4}(m_{\Lambda_{b}}-m_{\Lambda})}{2}}q^{2},$ (24) $\displaystyle F_{2}$ $\displaystyle=$ $\displaystyle{\frac{2H_{1}+H_{3}(m_{\Lambda}+m_{\Lambda_{b}})+H_{4}q^{2}}{2}},$ (25) $\displaystyle F_{3}$ $\displaystyle=$ $\displaystyle{\frac{m_{\Lambda}+m_{\Lambda_{b}}}{q^{2}}}F_{1}.$ (26) Due to the conservation of vector current, the form factors $f_{3}$ and $g_{3}$ do not contribute to the decay amplitude of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$. To incorporate the NHBs effect one need to calculate the matrix elements involving the scalar $\bar{s}b$ and the pseudoscalar $\bar{s}\gamma_{5}b$ currents, which can be parameterized as $\displaystyle\langle\Lambda(P)\|\bar{s}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle{\frac{1}{m_{b}+m_{s}}}\overline{\Lambda}(P)[g_{1}(m_{\Lambda_{b}}-m_{\Lambda})+g_{3}q^{2}]\Lambda_{b}(P+q),\,\,$ (27) $\displaystyle\langle\Lambda(P)\|\bar{s}\gamma_{5}b\|\Lambda_{b}(P+q)\rangle$ $\displaystyle=$ $\displaystyle{\frac{1}{m_{b}-m_{s}}}\overline{\Lambda}(P)[G_{1}(m_{\Lambda_{b}}+m_{\Lambda})-G_{3}q^{2}]\gamma_{5}\Lambda_{b}(P+q).$ (28) The various form factors $f_{i}$ and $g_{i}$ appearing in the above equations are not independent in the heavy quark limit and one can express them in terms of two independent form factors $\xi_{1}$ and $\xi_{2}$ in HQET defined by YuMing $\langle\Lambda(P)\|\bar{b}\Gamma s\|\Lambda_{b}(P+q)\rangle=\overline{\Lambda}(P)[\xi_{1}(q^{2})+{\not v}\xi_{2}(q^{2})]\Gamma\Lambda_{b}(P+q),$ (29) with $\Gamma$ being an arbitrary Lorentz structure and $v_{\mu}$ being the four-velocity of $\Lambda_{b}$ baryon. Comparing Eqs. (13-14), (19\- 20) and the Eq. (29), one can arrive at c.q. geng 4 ; Aliev 1 ; Aliev 2 ; Aliev 3 ; Aliev 4 $\displaystyle f_{1}$ $\displaystyle=$ $\displaystyle F_{1}={\frac{q^{2}}{m_{\Lambda_{b}}}}\xi_{2},$ (30) $\displaystyle f_{2}$ $\displaystyle=$ $\displaystyle F_{2}=g_{1}=G_{1}=\xi_{1}+{\frac{m_{\Lambda}}{m_{\Lambda_{b}}}}\xi_{2},$ (31) $\displaystyle f_{3}$ $\displaystyle=$ $\displaystyle{\frac{m_{\Lambda}-m_{\Lambda_{b}}}{m_{\Lambda_{b}}}}\xi_{2},$ (32) $\displaystyle F_{3}$ $\displaystyle=$ $\displaystyle{\frac{m_{\Lambda}+m_{\Lambda_{b}}}{m_{\Lambda_{b}}}}\xi_{2},$ (33) $\displaystyle g_{2}$ $\displaystyle=$ $\displaystyle G_{2}=g_{3}=G_{3}={\frac{\xi_{2}}{m_{\Lambda_{b}}}}.$ (34) Due to our poor understanding towards non-perturbative QCD dynamics, one has to rely on some approaches to calculate the form factors answering for $\Lambda_{b}\to\Lambda$ transition. It is suggested that the soft non- perturbative contribution to the transition form factor can be calculated quantitatively in the framework of LCSR approach LCSR 1 ; LCSR 2 ; LCSR 3 ; LCSR 4 ; LCSR 5 , which is a fully relativistic approach and well rooted in quantum field theory, in a systematic and almost model-independent way. As a marriage of standard QCDSR technique SVZ 1 ; SVZ 2 ; SVZ 3 and theory of hard exclusive process hard exclusive process 1 ; hard exclusive process 2 ; hard exclusive process 3 ; hard exclusive process 4 ; hard exclusive process 5 ; hard exclusive process 6 ; hard exclusive process 7 ; hard exclusive process 8 , LCSR cure the problem of QCDSR applying to the large momentum transfer by performing the operator product expansion (OPE) in terms of twist of the relevant operators rather than their dimension braun talk . Therefore, the principal discrepancy between QCDSR and LCSR consists in that non-perturbative vacuum condensates representing the long-distance quark and gluon interactions in the short-distance expansion are substituted by the light cone distribution amplitudes (LCDAs) describing the distribution of longitudinal momentum carried by the valence quarks of hadronic bound system in the expansion of transverse-distance between partons in the infinite momentum frame. Considering the distribution amplitude up to twist-6, the form factors for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ have been calculated in YuMing to the accuracy of leading conformal spin, where the pole model was also employed to extend the results to the whole kinematical region. Specifically, the dependence of form factors on transfer momentum are parameterized as $\xi_{i}(q^{2})={\frac{\xi_{i}(0)}{1-a_{1}q^{2}/m_{\Lambda_{b}}^{2}+a_{2}q^{4}/m_{\Lambda_{b}}^{4}}},$ (35) where $\xi_{i}$ denotes the form factors $f_{2}$ and $g_{2}$. The numbers of parameters $\xi_{i}(0),\,\,a_{1},\,\,a_{2}$ have been collected in Table 1. Table 1: Numerical results for the form factors $f_{2}(0)$, $g_{2}(0)$ and parameters $a_{1}$ and $a_{2}$ involved in the double-pole fit of eq. (35) for both twist-3 and twist-6 sum rules with $M_{B}^{2}\in[3.0,6.0]~{}\mbox{GeV}^{2}$, $s_{0}=39\pm 1~{}\mbox{GeV}^{2}$. parameter \| twist-3 \| up to twist-6 ---\|---\|--- $f_{2}(0)$ \| $0.14_{-0.01}^{+0.02}$ \| $0.15_{-0.02}^{+0.02}$ $a_{1}$ \| $2.91_{-0.07}^{+0.10}$ \| $2.94_{-0.06}^{+0.11}$ $a_{2}$ \| $2.26_{-0.08}^{+0.13}$ \| $2.31_{-0.10}^{+0.14}$ $g_{2}(0)(10^{-2}\mathrm{{GeV^{-1}})}$ \| $-0.47_{-0.06}^{+0.06}$ \| $1.3_{-0.4}^{+0.2}$ $a_{1}$ \| $3.40_{-0.05}^{+0.06}$ \| $2.91_{-0.09}^{+0.12}$ $a_{2}$ \| $2.98_{-0.08}^{+0.09}$ \| $2.24_{-0.13}^{+0.17}$ To the leading order and leading power, the other form factors can be related to these two as $\displaystyle F_{1}(q^{2})$ $\displaystyle=$ $\displaystyle f_{1}(q^{2})=q^{2}g_{2}(q^{2})=q^{2}G_{2}(q^{2}),$ $\displaystyle F_{2}(q^{2})$ $\displaystyle=$ $\displaystyle f_{2}(q^{2})=g_{1}(q^{2})=G_{1}(q^{2}),$ (36) where the form factors $F_{3}(q^{2})$ and $G_{3}(q^{2})$ are dropped out here due to their tiny contributions. ## IV Formula for Observables In this section, we proceed to perform the calculations of some interesting observables in phenomenology including decay rates, forward-backward asymmetry, polarization asymmetries of final state lepton and of $\Lambda$ baryon. From Eq. (LABEL:quark-amplitude), it is straightforward to obtain the decay amplitude for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ as $\mathcal{M}_{\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}}=-\frac{G_{F}\alpha}{2\sqrt{2}\pi}V_{tb}V_{ts}^{\ast}\left[T_{\mu}^{1}(\bar{l}\gamma^{\mu}l)+T_{\mu}^{2}(\bar{l}\gamma^{\mu}\gamma_{5}l)+T^{3}(\bar{l}l)\right],$ (37) where the auxiliary functions $T_{\mu}^{1}$, $T_{\mu}^{2}$ and $T^{3}$ are given by $\displaystyle T_{\mu}^{1}$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)\bigg{[}\left\\{\gamma_{\mu}\left(g_{1}-G_{1}\gamma_{5}\right)+i\sigma_{\mu\nu}q^{\nu}\left(g_{2}-G_{2}\gamma_{5}\right)\right\\}C_{9}^{eff}$ (38) $\displaystyle+\left\\{\gamma_{\mu}\left(g_{1}+G_{1}\gamma_{5}\right)+i\sigma_{\mu\nu}q^{\nu}\left(g_{2}+G_{2}\gamma_{5}\right)\right\\}C_{9}^{\prime eff}$ $\displaystyle-2m_{b}/s\left\\{\gamma_{\mu}\left(f_{1}+F_{1}\gamma_{5}\right)+i\sigma_{\mu\nu}q^{\nu}\left(f_{2}+F_{2}\gamma_{5}\right)\right\\}C_{7}^{eff}$ $\displaystyle-2m_{b}/s\left\\{\gamma_{\mu}\left(f_{1}-F_{1}\gamma_{5}\right)+i\sigma_{\mu\nu}q^{\nu}\left(f_{2}-F_{2}\gamma_{5}\right)\right\\}C_{7}^{\prime eff}\bigg{]}\Lambda_{b}\left(P+q\right),$ $\displaystyle T_{\mu}^{2}$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)\bigg{[}\left\\{\gamma_{\mu}\left(g_{1}-G_{1}\gamma_{5}\right)+i\sigma_{\mu\nu}q^{\nu}\left(g_{2}-G_{2}\gamma_{5}\right)+\left(g_{3}-G_{3}\gamma_{5}\right)q_{\mu}\right\\}C_{10}$ $\displaystyle+\left\\{\gamma_{\mu}\left(g_{1}+G_{1}\gamma_{5}\right)+i\sigma_{\mu\nu}q^{\nu}\left(g_{2}+G_{2}\gamma_{5}\right)+\left(g_{3}+G_{3}\gamma_{5}\right)q_{\mu}\right\\}C_{10}^{\prime}$ $\displaystyle-\frac{q_{\mu}}{2m_{l}\left(m_{b}+m_{s}\right)}\left\\{\left(g_{1}\left(m_{\Lambda_{b}}-m_{\Lambda}\right)+g_{3}q^{2}+G_{1}\left(m_{\Lambda_{b}}+m_{\Lambda}\right)-G_{3}q^{2}\right)\right\\}C_{Q_{2}}$ $\displaystyle-\frac{q_{\mu}}{2m_{l}\left(m_{b}+m_{s}\right)}\left\\{\left(g_{1}\left(m_{\Lambda_{b}}-m_{\Lambda}\right)+g_{3}q^{2}-G_{1}\left(m_{\Lambda_{b}}+m_{\Lambda}\right)+G_{3}q^{2}\right)\right\\}C_{Q_{2}}^{\prime}\bigg{]}\Lambda_{b}\left(P+q\right),$ and $\displaystyle T^{3}$ $\displaystyle=$ $\displaystyle\overline{\Lambda}(P)\bigg{[}\left\\{\left(g_{1}\left(m_{\Lambda_{b}}-m_{\Lambda}\right)+g_{3}q^{2}+G_{1}\left(m_{\Lambda_{b}}+m_{\Lambda}\right)-G_{3}q^{2}\right)\right\\}C_{Q_{1}}$ (40) $\displaystyle+\left\\{\left(g_{1}\left(m_{\Lambda_{b}}-m_{\Lambda}\right)+g_{3}q^{2}-G_{1}\left(m_{\Lambda_{b}}+m_{\Lambda}\right)+G_{3}q^{2}\right)\right\\}C_{Q_{1}}^{\prime}\bigg{]}\Lambda_{b}\left(P+q\right).$ It needs to point out that the terms proportional to $q_{\mu}$ in $T_{\mu}^{1}$ do not contribute to the decay amplitude with the help of the equation of motion for lepton fields. Besides, one can also find that the above results can indeed reproduce that obtained in the SM with $C_{i}^{\prime}=0$ and $T^{3}=0$. ### IV.1 The differential decay rates of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ The differential decay width of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ in the rest frame of $\Lambda_{b}$ baryon can be written as PDG , ${\frac{d\Gamma({\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}})}{ds}}={\frac{1}{(2\pi)^{3}}}{\frac{1}{32m_{\Lambda_{b}}^{3}}}\int_{u_{min}}^{u_{max}}\|{\widetilde{M}}_{\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}}\|^{2}du,$ (41) where $u=(p_{\Lambda}+p_{l^{-}})^{2}$ and $s=(p_{l^{+}}+p_{l^{-}})^{2}$; $p_{\Lambda}$, $p_{l^{+}}$ and $p_{l^{-}}$ are the four-momenta vectors of $\Lambda$, $l^{+}$ and $l^{-}$ respectively. ${\widetilde{M}}_{\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}}$ denotes the decay amplitude after performing the integration over the angle between the $l^{-}$ and $\Lambda$ baryon. The upper and lower limits of $u$ are given by $\displaystyle u_{max}$ $\displaystyle=$ $\displaystyle(E_{\Lambda}^{\ast}+E_{l}^{\ast})^{2}-(\sqrt{E_{\Lambda}^{\ast 2}-m_{\Lambda}^{2}}-\sqrt{E_{l}^{\ast 2}-m_{l}^{2}})^{2},$ $\displaystyle u_{min}$ $\displaystyle=$ $\displaystyle(E_{\Lambda}^{\ast}+E_{l}^{\ast})^{2}-(\sqrt{E_{\Lambda}^{\ast 2}-m_{\Lambda}^{2}}+\sqrt{E_{l}^{\ast 2}-m_{l}^{2}})^{2},$ (42) where $E_{\Lambda}^{\ast}$ and $E_{l}^{\ast}$ are the energies of $\Lambda$ and $l^{-}$ in the rest frame of lepton pair $E_{\Lambda}^{\ast}={\frac{m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2}-s}{2\sqrt{s}}},\hskip 28.45274ptE_{l}^{\ast}={\frac{\sqrt{s}}{2}}.$ (43) Putting everything together, we can achieve the decay rates and invariant mass distributions of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ with and without long distance contributions as $\displaystyle{\frac{d\Gamma}{ds}}$ $\displaystyle=$ $\displaystyle\frac{\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}\left(u_{\max}-u_{\min}\right)}{128m_{\Lambda_{b}}^{3}\pi^{5}}\times$ $\displaystyle\bigg{\\{}(f_{2}m_{l}(f_{2}-g_{2}((m_{\Lambda}+m_{\Lambda_{b}}))(C_{Q_{1}}C_{7}^{\ast eff}+C_{Q_{1}}^{\ast}C_{7}^{eff})+\frac{m_{l}}{2m_{b}}(f_{2}(m_{\Lambda}+m_{\Lambda_{b}})-g_{2}s)(C_{9}^{\ast eff}C_{Q_{1}}+C_{Q_{1}}^{\ast}C_{9}^{eff})$ $\displaystyle+\frac{s}{2m_{\Lambda_{b}}}\bigg{(}(f_{2}-g_{2}m_{\Lambda})^{2}-g_{2}^{2}m_{\Lambda_{b}}^{2})(C_{9}^{\ast eff}C_{10}+C_{10}^{\ast}C_{9}^{eff})$ $\displaystyle+m_{b}(f_{2}^{2}-g_{2}^{2}s)(C_{10}^{\ast}C_{7}^{eff}+C_{7}^{\ast eff}C_{10})\bigg{)}(2m_{l}^{2}+m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2}-s-u_{\max}-u_{\min})$ $\displaystyle+\frac{m_{l}}{2m_{b}}f_{2}^{2}(m_{\Lambda}-m_{\Lambda_{b}})((m_{\Lambda}+m_{\Lambda_{b}})^{2}-s)(C_{10}^{\ast}C_{Q_{2}}+C_{Q_{2}}^{\ast}C_{10})$ $\displaystyle-\frac{m_{b}}{s}(2m_{l}^{2}+s)((m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)f_{2}^{2}-4sm_{\Lambda}f_{2}g_{2}+g_{2}^{2}s(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s))(C_{9}^{\ast eff}C_{7}^{eff}+C_{7}^{\ast eff}C_{9}^{eff})$ $\displaystyle+\frac{sm_{\Lambda_{b}}}{2m_{b}^{2}}f_{2}^{2}((m_{\Lambda}+m_{\Lambda_{b}})^{2}-s)(C_{Q_{2}}C_{Q_{2}}^{\ast})+\frac{2m_{b}^{2}}{3m_{\Lambda_{b}}s}\bigg{(}(6m_{l}^{2}(sm_{l}^{2}+(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2})^{2}-s(s+u_{\max}+u_{\min}))$ $\displaystyle+s(3m_{\Lambda}^{4}+3m_{\Lambda_{b}}^{4}-3(m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2})(s+u_{\max}+u_{\min})+2u_{\max}^{2}$ $\displaystyle+3s(u_{\max}+u_{\min})+2u_{\max}u_{\min}))f_{2}^{2}+6f_{2}g_{2}m_{\Lambda}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}-s)s(2m_{l}^{2}+s)$ $\displaystyle+g_{2}^{2}s^{2}(-6m_{l}^{4}+6(u_{\max}+u_{\min})m_{l}^{2}-3s^{2}-2(u_{\max}^{2}-u_{\min}^{2})$ $\displaystyle-3s(u_{\max}+u_{\min})+3m_{\Lambda_{b}}^{2}(s+u_{\max}+u_{\min})-2u_{\max}u_{\min}$ $\displaystyle+3m_{\Lambda}^{2}(-2m_{\Lambda_{b}}^{2}+s+u_{\max}+u_{\min}))\bigg{)}\left\|C_{7}^{eff}\right\|^{2}$ $\displaystyle+\frac{1}{4m_{\Lambda_{b}}}\bigg{(}(-6m_{l}^{4}+(6m_{l}^{2}-3s)(u_{\max}+u_{\min})-3s^{2}-2(u_{\max}^{2}+u_{\min}^{2}+u_{\max}u_{\min})+3m_{\Lambda_{b}}^{2}(s+u_{\max}+u_{\min})$ $\displaystyle+3m_{\Lambda}^{2}(-2m_{\Lambda_{b}}^{2}+s+u_{\max}+u_{\min}))f_{2}^{2}-6f_{2}g_{2}m_{\Lambda}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}-s)s(2m_{l}^{2}+s)$ $\displaystyle+g_{2}^{2}(6sm_{l}^{4}+6m_{l}^{2}((m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2})^{2}-s(s+u_{\max}+u_{\min}))+s(3m_{\Lambda}^{4}+3m_{\Lambda_{b}}^{4}$ $\displaystyle-3(m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2})(s+u_{\max}+u_{\min})+2(u_{\max}^{2}+u_{\min}^{2}+u_{\max}u_{\min})+3t(u_{\max}+u_{\min}))))\bigg{)}\left\|C_{9}^{eff}\right\|^{2}$ $\displaystyle\frac{1}{4m_{\Lambda_{b}}}\bigg{(}(6m_{l}^{4}+6(2m_{\Lambda}^{2}+2m_{\Lambda_{b}}^{2}-2s-u_{\max}-u_{\min})m_{l}^{2}+2(u_{\max}^{2}+u_{\min}^{2}+u_{\max}u_{\min})$ $\displaystyle+3(s-m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2})(s+u_{\max}+u_{\min}))f_{2}^{2}+6g_{2}f_{2}(4m_{l}^{2}-s)(m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2}+s)$ $\displaystyle+g_{2}^{2}(-6sm_{l}^{4}+6(m_{\Lambda}^{4}-(2m_{\Lambda_{b}}^{2}+s)m_{\Lambda}^{2}+m_{\Lambda_{b}}^{4}-m_{\Lambda_{b}}^{2}s+s(u_{\max}+u_{\min})m_{l}^{2}$ $\displaystyle+s(-3m_{\Lambda}^{4}+3(m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2})(s+u_{\max}+u_{\min})-3m_{\Lambda_{b}}^{4}$ $\displaystyle-2(u_{\max}^{2}+u_{\min}^{2}+u_{\max}u_{\min})-3t(u_{\max}+u_{\min})))\bigg{)}\left\|C_{10}\right\|^{2}\bigg{\\}},$ where $u_{max}$ and $u_{min}$ are defined in Eq. (42). In Eq. (LABEL:drate) we have given the result in MSSM with NHBs and ignored the contribution from the primed operators which appear SUSY SO(10) GUT model as the results are very tiny. ### IV.2 FBAs of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ Now we are in a position to explore the FBAs of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$, which is an essential observable sensitive to the new physics effects. To calculate the forward-backward asymmetry, we consider the following double differential decay rate formula for the process $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ${\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}={\frac{1}{(2\pi)^{3}}}{\frac{1}{64m_{\Lambda_{b}}^{3}}}\lambda^{1/2}(m_{\Lambda_{b}}^{2},m_{\Lambda}^{2},s)\sqrt{1-{\frac{4m_{l}^{2}}{s}}}\|{\widetilde{M}}_{\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}}\|^{2}\,\,,$ (45) where $\theta$ is the angle between the momentum of $\Lambda_{b}$ baryon and $l^{-}$ in the dilepton rest frame; $\lambda(a,b,c)=a^{2}+b^{2}+c^{2}-2ab-2ac-2bc$. Following Refs. c.q. geng 4 ; b to s in theory 9 , the differential and normalized FBAs for the semi- leptonic decay $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ are defined as ${\frac{dA_{FB}(q^{2})}{ds}}=\int_{0}^{1}d\cos\theta{\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}-\int_{-1}^{0}d\cos\theta{\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}$ (46) and $A_{FB}(q^{2})={\frac{\int_{0}^{1}d\cos\theta{\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}-\int_{-1}^{0}d\cos\theta{\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}}{\int_{0}^{1}d\cos\theta{\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}+\int_{-1}^{0}d\cos\theta{\frac{d^{2}\Gamma(s,\cos\theta)}{dsd\cos\theta}}}}.$ (47) Following the same procedure as we did for the differential decay rate, one can easily get the expression for the forward-backward asymmetry. ### IV.3 Lepton Polarization asymmetries of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ In the rest frame of the lepton $l^{-}$, the unit vectors along longitudinal, normal and transversal component of the $l^{-}$ can be defined as Aliev UED : $\displaystyle s_{L}^{-\mu}$ $\displaystyle=$ $\displaystyle(0,\vec{e}_{L})=\left(0,\frac{\vec{p}_{-}}{\left\|\vec{p}_{-}\right\|}\right),$ $\displaystyle s_{N}^{-\mu}$ $\displaystyle=$ $\displaystyle(0,\vec{e}_{N})=\left(0,\frac{\vec{p}_{\Lambda}\times\vec{p}_{-}}{\left\|\vec{p}_{\Lambda}\times\vec{p}_{-}\right\|}\right),$ (48) $\displaystyle s_{T}^{-\mu}$ $\displaystyle=$ $\displaystyle(0,\vec{e}_{T})=\left(0,\vec{e}_{N}\times\vec{e}_{L}\right),$ where $\vec{p}_{-}$ and $\vec{p}_{\Lambda}$ are the three-momenta of the lepton $l^{-}$ and $\Lambda$ baryon respectively in the center mass (CM) frame of $l^{+}l^{-}$ system. Lorentz transformation is used to boost the longitudinal component of the lepton polarization to the CM frame of the lepton pair as $\left(s_{L}^{-\mu}\right)_{CM}=\left(\frac{\|\vec{p}_{-}\|}{m_{l}},\frac{E_{l}\vec{p}_{-}}{m_{l}\left\|\vec{p}_{-}\right\|}\right)$ (49) where $E_{l}$ and $m_{l}$ are the energy and mass of the lepton in the CM frame. The normal and transverse components remain unchanged under the Lorentz boost. The longitudinal ($P_{L}$), normal ($P_{N}$) and the transverse ($P_{T}$) polarizations of lepton can be defined as: $P_{i}^{(\mp)}(s)=\frac{\frac{d\Gamma}{ds}(\vec{\xi}^{\mp}=\vec{e}^{\mp})-\frac{d\Gamma}{ds}(\vec{\xi}^{\mp}=-\vec{e}^{\mp})}{\frac{d\Gamma}{ds}(\vec{\xi}^{\mp}=\vec{e}^{\mp})+\frac{d\Gamma}{ds}(\vec{\xi}^{\mp}=-\vec{e}^{\mp})}$ (50) where $i=L,\;N,\;T$ and $\vec{\xi}^{\mp}$ is the spin direction along the leptons $l^{\mp}$. The differential decay rate for polarized lepton $l^{\mp}$ in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decay along any spin direction $\vec{\xi}^{\mp}$ is related to the unpolarized decay rate (41) with the following relation $\frac{d\Gamma(\vec{\xi}^{\mp})}{ds}=\frac{1}{2}\left(\frac{d\Gamma}{ds}\right)[1+(P_{L}^{\mp}\vec{e}_{L}^{\mp}+P_{N}^{\mp}\vec{e}_{N}^{\mp}+P_{T}^{\mp}\vec{e}_{T}^{\mp})\cdot\vec{\xi}^{\mp}].$ (51) We can achieve the expressions of longitudinal, normal and transverse polarizations for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decays as collected below, where only the results in MSSM models with NHB are given for the conciseness of paper. Thus the longitudinal lepton polarization can be written as $\displaystyle P_{L}(s)$ $\displaystyle=$ $\displaystyle(1/{\frac{d\Gamma}{ds}})\frac{\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}u_{\max}}{768(m_{b}^{2}-m_{s}^{2})m_{\Lambda_{b}}^{3}\pi^{5}}\sqrt{1-\frac{4m_{l}^{2}}{s}}\times$ $\displaystyle\bigg{\\{}-6m_{l}(m_{b}-m_{s})m_{\Lambda_{b}}f_{2}^{2}(m_{\Lambda}-m_{\Lambda_{b}})(s-(m_{\Lambda}+m_{\Lambda_{b}})^{2})(C_{Q_{1}}C_{10}^{\ast}+C_{Q_{1}}^{\ast}C_{10})$ $\displaystyle-6m_{\Lambda_{b}}^{2}f_{2}s(s-(m_{\Lambda}+m_{\Lambda_{b}})^{2})(C_{Q_{1}}^{\ast}C_{Q_{2}}+C_{Q_{1}}C_{Q_{2}}^{\ast})$ $\displaystyle+(m_{b}^{2}-m_{s}^{2})((3m_{\Lambda}^{4}+3m_{\Lambda_{b}}^{4}-3s^{2})(f_{2}^{2}+g_{2}^{2}s)-6m_{\Lambda_{b}}^{2}m_{\Lambda}^{2}(f_{2}^{2}+g_{2}^{2}s)$ $\displaystyle-\sqrt{\frac{\lambda(m_{\Lambda_{b}}^{2},m_{\Lambda}^{2},s)}{1-4m_{l}^{2}/s}}(f_{2}^{2}-g_{2}^{2}s))(C_{9}^{\ast eff}C_{10}+C_{9}^{eff}C_{10}^{\ast})$ $\displaystyle-12(m_{b}^{2}-m_{s}^{2})(m_{b}m_{\Lambda_{b}}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)-m_{s}m_{\Lambda}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}-s)$ $\displaystyle(f_{2}^{2}+g_{2}^{2}s)+f_{2}g_{2}(m_{s}((m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2})^{2}-s)-4m_{b}m_{\Lambda_{b}}m_{\Lambda}s))(C_{10}^{\ast}C_{7}^{eff}+C_{7}^{\ast eff}C_{10})\bigg{\\}}.$ Similarly, the normal lepton polarization is $\displaystyle P_{N}(s)$ $\displaystyle=$ $\displaystyle(1/{\frac{d\Gamma}{ds}})\frac{\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}u_{\max}}{1024(m_{b}+m_{s})m_{\Lambda_{b}}^{3}\pi^{4}}\sqrt{\frac{\lambda(m_{\Lambda_{b}}^{2},m_{\Lambda}^{2},s)}{s}}\times$ (53) $\displaystyle\bigg{\\{}(s-4m_{l}^{2})(g_{2}s-f_{2}(m_{\Lambda}+m_{\Lambda_{b}}))m_{\Lambda_{b}}(C_{Q_{1}}C_{10}^{\ast}+C_{Q_{1}}^{\ast}C_{10})$ $\displaystyle-2sm_{\Lambda_{b}}\frac{(m_{b}+m_{s})^{2}}{m_{b}-m_{s}}(g_{2}(m_{\Lambda}+m_{\Lambda_{b}})-f_{2})f_{2}(C_{Q_{2}}^{\ast}C_{7}^{eff}+C_{7}^{\ast eff}C_{Q_{2}})$ $\displaystyle+s\frac{m_{b}+m_{s}}{m_{b}-m_{s}}m_{\Lambda_{b}}(f_{2}(m_{\Lambda}+m_{\Lambda_{b}})-g_{2}s)f_{2}(C_{Q_{2}}^{\ast}C_{9}^{eff}+C_{9}^{\ast eff}C_{Q_{2}})$ $\displaystyle-2f_{2}m_{l}(m_{b}+m_{s})(f_{2}(m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2})+g_{2}m_{\Lambda}s)(C_{9}^{eff}C_{10}^{\ast}+C_{9}^{\ast eff}C_{10})$ $\displaystyle+8m_{l}(m_{b}+m_{s})(m_{s}m_{\Lambda}(f_{2}^{2}+g_{2}^{2}s)$ $\displaystyle+m_{b}m_{\Lambda_{b}}(f_{2}^{2}-g_{2}^{2}s)-f_{2}g_{2}m_{s}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s))(C_{9}^{eff}C_{7}^{\ast eff}+C_{9}^{\ast eff}C_{7}^{eff})$ $\displaystyle+4m_{l}(m_{b}+m_{s})f_{2}(g_{2}m_{s}m_{\Lambda_{b}}^{2}-f_{2}m_{b}m_{\Lambda_{b}}+m_{\Lambda}m_{s}(f_{2}-g_{2}m_{\Lambda}))(C_{7}^{\ast eff}C_{10}+C_{7}^{eff}C_{10}^{\ast})$ $\displaystyle-\frac{16(m_{b}+m_{s})^{2}(m_{b}-m_{s})m_{l}}{s}(f_{2}(m_{\Lambda_{b}}-m_{\Lambda})+g_{2}s)(g_{2}s-f_{2}(m_{\Lambda_{b}}-m_{\Lambda}))\left\|C_{7}^{eff}\right\|^{2}$ $\displaystyle+4m_{l}(m_{b}+m_{s})s((f_{2}-g_{2}m_{\Lambda})^{2}-g_{2}^{2}m_{\Lambda_{b}}^{2})\left\|C_{9}^{eff}\right\|^{2}\bigg{\\}},$ and the transverse one is given by $\displaystyle P_{T}(s)$ $\displaystyle=$ $\displaystyle(1/{\frac{d\Gamma}{ds}})\frac{i\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}u_{\max}}{4096m_{\Lambda_{b}}^{3}\pi^{4}\sqrt{s}}\sqrt{1-\frac{4m_{l}^{2}}{s}}\sqrt{\lambda(m_{\Lambda_{b}}^{2},m_{\Lambda}^{2},s)}(m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2}+s)\times$ $\displaystyle\bigg{\\{}2m_{\Lambda_{b}}f_{2}(g_{2}(m_{\Lambda}+m_{\Lambda_{b}})-f_{2})(C_{Q_{1}}C_{7}^{\ast eff}-C_{Q_{1}}^{\ast}C_{7}^{eff})$ $\displaystyle-\frac{m_{\Lambda_{b}}}{m_{b}+m_{s}}f_{2}(g_{2}s-f_{2}(m_{\Lambda}+m_{\Lambda_{b}}))(C_{Q_{1}}C_{9}^{\ast eff}-C_{9}^{eff}C_{Q_{1}})$ $\displaystyle-\frac{m_{\Lambda_{b}}}{m_{b}-m_{s}}f_{2}(g_{2}s-f_{2}(m_{\Lambda}+m_{\Lambda_{b}}))(C_{Q_{2}}C_{10}^{\ast}-C_{10}C_{Q_{2}}^{\ast})$ $\displaystyle-2m_{l}((f_{2}-g_{2}m_{\Lambda})^{2}-g_{2}^{2}m_{\Lambda_{b}}^{2})(C_{9}^{\ast eff}C_{10}-C_{9}^{eff}C_{10}^{\ast})$ $\displaystyle+\frac{4m_{l}}{s}((m_{s}m_{\Lambda}+m_{b}m_{\Lambda_{b}})f_{2}^{2}-f_{2}g_{2}m_{s}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)$ $\displaystyle+g_{2}^{2}(m_{s}m_{\Lambda}-m_{b}m_{\Lambda_{b}})s)(C_{10}^{\ast}C_{7}^{eff}-C_{7}^{\ast eff}C_{10})\bigg{\\}}.$ The ${\frac{d\Gamma}{ds}}$ appearing in the above equation is the one given in Eq. (LABEL:drate). ### IV.4 $\Lambda$ polarization in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ To study the $\Lambda$ spin polarization, one needs to express the $\Lambda$ four spin vector in terms of a unit vector $\hat{\xi}$ along the $\Lambda$ spin in its rest frame as c.q. geng 2 $s_{0}=\frac{\vec{p}_{\Lambda}\cdot\vec{\xi}}{m_{\Lambda}},\text{ }\vec{s}=\vec{\xi}+\frac{s_{0}}{E_{\Lambda}+m_{\Lambda}}\vec{p}_{\Lambda},$ (55) where the unit vectors along the longitudinal, normal and transverse components of the $\Lambda$ polarization are chosen to be $\displaystyle\hat{e}_{L}$ $\displaystyle=$ $\displaystyle\frac{\vec{p}_{\Lambda}}{\left\|\vec{p}_{\Lambda}\right\|},$ $\displaystyle\hat{e}_{N}$ $\displaystyle=$ $\displaystyle\frac{\vec{p}_{\Lambda}\times(\vec{p}_{-}\times\vec{p}_{\Lambda})}{\left\|\vec{p}_{\Lambda}\times(\vec{p}_{-}\times\vec{p}_{\Lambda})\right\|},$ $\displaystyle\hat{e}_{T}$ $\displaystyle=$ $\displaystyle\frac{\vec{p}_{-}\times\vec{p}_{\Lambda}}{\left\|\vec{p}_{-}\times\vec{p}_{\Lambda}\right\|}.$ Similar to the lepton polarization, the polarization asymmetries for $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ can be defined as $P_{i}^{(\mp)}(s)=\frac{\frac{d\Gamma}{ds}(\vec{\xi}=\hat{e})-\frac{d\Gamma}{ds}(\vec{\xi}=-\hat{e})}{\frac{d\Gamma}{ds}(\vec{\xi}=\hat{e})+\frac{d\Gamma}{ds}(\vec{\xi}=-\hat{e})}$ (56) where $i=L,\;N,\;T$ and $\vec{\xi}$ is the spin direction along the $\Lambda$ baryon. The differential decay rate for polarized $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decay along any spin direction $\vec{\xi}$ is related to the unpolarized decay rate (41) through the following relation $\frac{d\Gamma(\vec{\xi})}{ds}=\frac{1}{2}\left(\frac{d\Gamma}{ds}\right)[1+(P_{L}\vec{e}_{L}+P_{N}\vec{e}_{N}+P_{T}\vec{e}_{T})\cdot\vec{\xi}].$ (57) Following the same procedure as we did for the lepton polarizations, we can derive the formulae for the longitudinal, normal and transverse polarizations of $\Lambda$ baryon in the MSSM as $\displaystyle P_{L}(s)$ $\displaystyle=$ $\displaystyle(1/{\frac{d\Gamma}{ds}})\frac{\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}u_{\max}}{64m_{\Lambda}m_{\Lambda_{b}}^{3}\pi^{5}s^{3/2}}\times$ $\displaystyle\bigg{\\{}\frac{m_{\Lambda}m_{\Lambda_{b}}m_{l}}{2m_{b}}(m_{\Lambda}+m_{\Lambda_{b}})s^{3/2}\sqrt{\lambda_{p}}f_{2}^{2}(C_{Q_{2}}C_{10}^{\ast}+C_{10}^{\ast}C_{Q_{2}})$ $\displaystyle+m_{\Lambda}m_{\Lambda_{b}}m_{l}(2m_{l}^{2}+s)(g_{2}^{2}s-f_{2}^{2})\sqrt{\lambda_{p}}\sqrt{s}(C_{7}^{\ast eff}C_{9}^{eff}+C_{7}^{eff}C_{9}^{\ast eff})$ $\displaystyle+\frac{m_{b}^{2}}{3}\bigg{[}\frac{m_{\Lambda}}{\sqrt{s}}(12m_{l}^{2}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2})\sqrt{\lambda_{p}}+s\sqrt{1-\frac{4m_{l}^{2}}{s}}u_{\max}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)-3\sqrt{\lambda_{p}}s(s-m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2}))f_{2}^{2}$ $\displaystyle- g_{2}f_{2}(24m_{l}^{2}\sqrt{s}\sqrt{\lambda_{p}}m_{\Lambda}^{2}-\sqrt{s}u_{\max}^{2}\sqrt{\lambda_{p}}+\sqrt{s}u_{\max}\sqrt{1-\frac{4m_{l}^{2}}{s}}\lambda_{p})$ $\displaystyle+g_{2}^{2}m_{\Lambda}s^{3/2}(12\sqrt{\lambda_{p}}m_{l}^{2}+u_{\max}\sqrt{1-\frac{4m_{l}^{2}}{s}}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)+3\sqrt{\lambda_{p}}(s-m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2}))\bigg{]}\left\|C_{7}^{eff}\right\|^{2}$ $\displaystyle+\frac{s}{12}\bigg{[}-m_{\Lambda}(-12m_{l}^{2}\sqrt{s}\sqrt{\lambda_{p}}+\sqrt{s}\sqrt{1-\frac{4m_{l}^{2}}{s}}u_{\max}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)+3u_{\max}\sqrt{s}(s-m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2}))f_{2}^{2}$ $\displaystyle+g_{2}f_{2}u_{\max}\sqrt{s}(\sqrt{1-\frac{4m_{l}^{2}}{s}}\lambda_{p}-u_{\max}\sqrt{\lambda_{p}})-g_{2}^{2}m_{\Lambda}\sqrt{s}(12m_{l}^{2}\sqrt{\lambda_{p}}(s-m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2})$ $\displaystyle+u_{\max}s\sqrt{1-\frac{4m_{l}^{2}}{s}}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)-3s\sqrt{\lambda_{p}}(s-m_{\Lambda}^{2}+m_{\Lambda_{b}}^{2}))\bigg{]}\left\|C_{10}\right\|^{2}$ $\displaystyle+\frac{s}{12}\bigg{[}-m_{\Lambda}(12m_{l}^{2}\sqrt{s}\sqrt{\lambda_{p}}+\sqrt{s}\sqrt{1-\frac{4m_{l}^{2}}{s}}u_{\max}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)-3u_{\max}\sqrt{s}(m_{\Lambda_{b}}^{2}-s+m_{\Lambda}^{2}))f_{2}^{2}$ $\displaystyle+g_{2}f_{2}u_{\max}\sqrt{s}(\sqrt{1-\frac{4m_{l}^{2}}{s}}\lambda_{p}-u_{\max}\sqrt{\lambda_{p}})-g_{2}^{2}m_{\Lambda}\sqrt{s}(-12m_{l}^{2}\sqrt{\lambda_{p}}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2})$ $\displaystyle+u_{\max}s\sqrt{1-\frac{4m_{l}^{2}}{s}}(m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2}-s)-3s\sqrt{\lambda_{p}}(m_{\Lambda}^{2}-s-m_{\Lambda_{b}}^{2}))\bigg{]}\left\|C_{9}^{eff}\right\|^{2}\bigg{\\}},$ $\displaystyle P_{N}(s)$ $\displaystyle=$ $\displaystyle(1/{\frac{d\Gamma}{ds}})\frac{\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}u_{\max}}{512m_{b}m_{\Lambda_{b}}^{3}\pi^{4}\sqrt{s}}\sqrt{1-\frac{4m_{l}^{2}}{s}}\times$ $\displaystyle\bigg{\\{}2m_{b}m_{l}m_{\Lambda_{b}}(s-(m_{\Lambda}+m_{\Lambda_{b}})^{2})(f_{2}(m_{\Lambda_{b}}-m_{\Lambda})+g_{2}s)(C_{Q_{1}}C_{7}^{\ast eff}+C_{Q_{1}}^{\ast}C_{7}^{eff})$ $\displaystyle+m_{l}m_{\Lambda_{b}}s(s-(m_{\Lambda}+m_{\Lambda_{b}})^{2})(f_{2}+g_{2}(m_{\Lambda_{b}}-m_{\Lambda}))(C_{Q_{1}}C_{9}^{\ast eff}+C_{Q_{1}}^{\ast}C_{9}^{eff})$ $\displaystyle- m_{b}s(m_{\Lambda_{b}}^{2}-m_{\Lambda}^{2}+s)(m_{\Lambda}f_{2}^{2}-g_{2}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)f_{2}+g_{2}m_{\Lambda}t)(C_{10}C_{9}^{\ast eff}+C_{10}^{\ast}C_{9}^{eff})$ $\displaystyle+4m_{b}m_{\Lambda_{b}}s(-m_{\Lambda}f_{2}^{2}+g_{2}(m_{\Lambda}^{2}-m_{\Lambda_{b}}^{2}+s)f_{2}-g_{2}^{2}m_{\Lambda}t)(C_{10}C_{7}^{\ast eff}+C_{10}^{\ast}C_{7}^{eff})\bigg{\\}},$ $\displaystyle P_{T}(s)$ $\displaystyle=$ $\displaystyle(1/{\frac{d\Gamma}{ds}})\frac{i\alpha^{2}G_{F}^{2}\left\|V_{tb}V_{ts}^{\ast}\right\|^{2}u_{\max}\lambda_{p}^{1/2}}{512m_{b}m_{\Lambda_{b}}^{2}\pi^{4}\sqrt{s}}m_{l.}\sqrt{1-\frac{4m_{l}^{2}}{s}}\times\bigg{\\{}2m_{b}(f_{2}(m_{\Lambda}+m_{\Lambda_{b}})-g_{2}s)(C_{Q_{1}}C_{7}^{\ast eff}-C_{Q_{1}}^{\ast}C_{7}^{eff})$ (58) $\displaystyle+s(g_{2}(m_{\Lambda}+m_{\Lambda_{b}})-f_{2})(C_{Q_{1}}C_{9}^{\ast eff}-C_{Q_{1}}^{\ast}C_{9}^{eff})\bigg{\\}},$ where $\lambda_{p}=\lambda(m_{\Lambda_{b}}^{2},m_{\Lambda}^{2},s)$ and the mass of strange quark is neglected to make the expressions more compact. ## V Numerical Analysis In this section, we would like to present the numerical analysis of decay rates, FBAs and polarization asymmetries of the lepton and $\Lambda$ baryon. The numerical values of Wilson coefficients and other input parameters used in our analysis are borrowed from Ref. Yan ; Li ; YuMing and collected in Tables II, III and IV. Table 2: Values of input parameters used in our numerical analysis $G_{F}=1.166\times 10^{-2}$ GeV-2 \| $\left\|V_{ts}\right\|=41.61_{-0.80}^{+0.10}\times 10^{-3}$ ---\|--- $\left\|{V_{tb}}\right\|=0.9991$ \| $m_{b}=\left(4.68\pm 0.03\right)$ GeV $m_{c}\left(m_{c}\right)=1.275_{-0.015}^{+0.015}$ GeV \| $m_{s}\left(1\text{ GeV}\right)=\left(142\pm 28\right)$ MeV $m_{\Lambda_{b}}=5.62$ GeV \| $m_{\Lambda}=1.12$ GeV $f_{\Lambda_{b}}=3.9_{-0.2}^{+0.4}\times 10^{-3}$ GeV2 \| $f_{\Lambda}=6.0_{-0.4}^{+0.4}\times 10^{-3}$ GeV2 Table 3: Wilson Coefficients in SM and different SUSY models but without Neutral Higgs boson contributions. The primed Wilson coefficients corresponds to the operators which are opposite in helicities from those of the SM operators and these comes only in SUSY SO(10) GUT model. Wilson Coefficients \| $C_{7}^{eff}$ \| $C_{7}^{\prime eff}$ \| $C_{9}$ \| $C_{9}^{\prime}$ \| $C_{10}$ \| $C_{10}^{\prime}$ ---\|---\|---\|---\|---\|---\|--- SM \| $-0.313$ \| $0$ \| $4.334$ \| $0$ \| $-4.669$ \| $0$ SUSYI \| $+0.3756$ \| $0$ \| $4.7674$ \| $0$ \| $-3.7354$ \| $0$ SUSYII \| $+0.3756$ \| $0$ \| $4.7674$ \| $0$ \| $-3.7354$ \| $0$ SUSYIII \| $-0.3756$ \| $0$ \| $4.7674$ \| $0$ \| $-3.7354$ \| $0$ SUSY SO(10) $\left(A_{0}=-1000\right)$ \| $-0.219+0i$ \| $0.039-0.038i$ \| $4.275+0i$ \| $0.011+0.0721i$ \| $-4.732-0i$ \| $-0.075-0.670i$ Table 4: Wilson coefficient corresponding to NHBs contributions. SUSYI corresponds to the regions where SUSY can destructively contribute and can change the sign of $C_{7}$, but contribution of NHBs are neglected, SUSYII refers to the region where $\tan\beta$ is large and the masses of the superpartners are relatively small. SUSY III corresponds to the regions where $\tan\beta$ is large and the masses of superpartners are relatively large. The primed Wilson coefficients are the contribution of NHBs in SUSY SO(10) GUT model. As the neutral Higgs bosons are proportional to the lepton mass, and the values shown in the table are for $\mu$ and $\tau$ case. The values in the bracket are for the $\tau$. Wilson Coefficients \| $C_{Q_{1}}$ \| $C_{Q_{1}}^{\prime}$ \| $C_{Q_{2}}$ \| $C_{Q_{2}}^{\prime}$ ---\|---\|---\|---\|--- SM \| $0$ \| $0$ \| $0$ \| $0$ SUSYI \| $0$ \| $0$ \| 0 \| $0$ SUSYII \| $6.5\left(16.5\right)$ \| $0$ \| $-6.5\left(-16.5\right)$ \| $0$ SUSYIII \| $1.2\left(4.5\right)$ \| $0$ \| $-1.2\left(-4.5\right)$ \| $0$ SUSY SO(10) $\left(A_{0}=-1000\right)$ \| $\begin{array}[]{c}0.106+0i\\\ \left(1.775+0.002i\right)\end{array}$ \| $\begin{array}[]{c}-0.247+0.242i\\\ \left(-4.148+4.074i\right)\end{array}$ \| $\begin{array}[]{c}-0.107+0i\\\ \left(-1.797-0.002i\right)\end{array}$ \| $\begin{array}[]{c}-0.250+0.246i\\\ \left(-4.202+4.128i\right)\end{array}$ In the subsequent analysis, we will focus on the parameter space of large $\tan\beta$, where the NHBs effects are significant owing to the fact that the Wilson coefficients corresponding to NHBs are proportional to $(m_{b}m_{l}/m_{h})\tan^{3}\beta$ $(h=h^{0}$, $A^{0})$. Here, one $\tan\beta$ comes from the chargino-up-type squark loop and $\tan^{2}\beta$ comes from the exchange of the NHBs. At large value of $\tan\beta$ the $C_{Q_{i}}^{(^{\prime})}$ compete with $C_{i}^{(^{\prime})}$ and can overwhelm $C_{i}^{(^{\prime})}$ in some region as can be seen from the Tables III and IV Huang . Apart from the large $\tan\beta$ limit, the other two conditions responsible for the large contributions from NHBs are: (i) the mass values of the lighter chargino and lighter stop should not be too large; (ii) the mass splitting of charginos and stops should be large, which also indicate large mixing between stop sector and chargino sector Yan . Once these conditions are satisfied, the process $B\rightarrow X_{s}\gamma$ will not only impose constraints on $C_{7}$ but it also puts very stringent constraint on the possible new physics. It is well known that the SUSY contribution is sensitive to the sign of the Higgs mass term $\mu$ and SUSY contributes destructively when the sign of this term becomes minus. It is pointed out in literature Yan that there exist considerable regions of SUSY parameter space in which NHBs can largely contribute to the process $b\rightarrow sl^{+}l^{-}$ due to change of the sign of $C_{7}$ from positive to negative, while the constraint on $b\rightarrow s\gamma$ is respected. Also, when the masses of SUSY particles are relatively large, say about $450$ GeV, there exist significant regions in the parameter space of SUSY models in which NHBs could contribute largely. However, in these cases $C_{7}$ does not change its sign, because contributions of charged Higgs and charginos cancel each other. Hopefully, we can distinguish between these two regions of SUSY by observing $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ with ($l=\mu,\tau$). \| \| ---\|---\|--- (a)(b)(c)(d) \| \| Figure 1: The differential width for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$ without long-distance contributions (a, c) and with long-distance contributions (b, d). The solid, dashed, dashed-dot, dashed-double dot and dashed-triple dot line represents, SM, SUSY I, SUSY II, SUSY III and SUSY SO(10) GUT model. The numerical results for the decay rates, FBAs and polarization asymmetries of the lepton and $\Lambda$ baryon are presented in Figs. 1-8. Fig. 1 describes the differential decay rate of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$, from which one can see that the supersymmetric effects are quite significant for SUSY I and SUSY II model in the high momentum transfer regions for the final states being muon pair, whereas these effects are extremely small for SUSY III and SUSY SO(10) GUT models in this case. The reason for the increase of differential decay width in SUSY I model is the relative change in the sign of $C_{7}^{eff}$; while the large change in SUSY II model is due to the contribution of the NHBs. As for the SUSY III and SUSY SO (10) models, the value of the Wilson coefficients corresponding to NHBs is small and hence one expects small deviations from SM. For the tauon case, the values of Wilson coefficients corresponding to NHBs in SUSY III are larger than that for the muon case and therefore their effects are quite significant as shown in Fig. 1(c, d). The numerical values of the branching fractions for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) with and without long-distance contribution in SM and different SUSY models are given in Table V. Table 5: Branching ratio for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}\left(l=\mu,\tau\right)$ in units of $10^{-6}$ in SM and different SUSY models. Branching Ratio \| $\begin{array}[]{c}\Lambda_{b}\rightarrow\Lambda\mu^{+}\mu^{-}\\\ \text{without LD}\end{array}$ \| $\begin{array}[]{c}\Lambda_{b}\rightarrow\Lambda\mu^{+}\mu^{-}\\\ \text{with LD}\end{array}$ \| $\begin{array}[]{c}\Lambda_{b}\rightarrow\Lambda\tau^{+}\tau^{-}\\\ \text{without LD}\end{array}$ \| $\begin{array}[]{c}\Lambda_{b}\rightarrow\Lambda\tau^{+}\tau^{-}\\\ \text{with LD}\end{array}$ ---\|---\|---\|---\|--- SM \| $5.9$ \| $39$ \| $2.1$ \| $4$ SUSYI \| $7.9$ \| $47$ \| $3.5$ \| $5.7$ SUSYII \| $25$ \| $65$ \| $31$ \| $33$ SUSYIII \| $5.6$ \| $45$ \| $3.2$ \| $5.6$ SUSY SO(10) $\left(A_{0}=-1000\right)$ \| $5.92$ \| $23$ \| $2$ \| $2.6$ \| \| ---\|---\|--- (a)(b)(c)(d) \| \| Figure 2: Forward-backward asymmetry for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$ without long-distance contributions (a, c) and with long-distance contributions (b, d). The line conventions are same as given in the legend of Fig. 1. \| \| ---\|---\|--- (a)(b)(c)(d) \| \| Figure 3: Longitudinal lepton polarization asymmetries for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$ without long-distance contributions (a, c) and with long-distance contributions (b, d). The line conventions are same as given in the legend of Fig. 1 In Fig. 2, the FBAs for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ are presented. Fig. 2(a,b) describe the FBAs for $\Lambda_{b}\rightarrow\Lambda\mu^{+}\mu^{-}$ with and without long-distance contributions, from which one can easily distinguish different SUSY models. It is known that in the SM the zero position of FBAs is due to the opposite sign of $C_{7}^{eff}$ and $C_{9}^{eff}$. In SUSY I and SUSY II models, the sign of $C_{7}^{eff}$ and $C_{9}^{eff}$ are the same and hence the zero point of the FBAs disappears. Whereas, in SUSY III model due to the opposite sign of $C_{7}^{eff}$ and $C_{9}^{eff}$ forward-backward asymmetry passes from the zero but this zero position shifts to the right from that of the SM value due to the contribution from the NHBs. Similar behavior is expected in SUSY SO(10) GUT model but in this case the shifting is very mild as the contribution from the NHBs is very small. For $\Lambda_{b}\rightarrow\Lambda\tau^{+}\tau^{-}$ the FBAs with and without long-distance contributions are represented in Fig. 2(c,d). Again, one can easily distinguish between the contributions from different SUSY models. Here, the most interesting point is that the FBAs pass through zero point in SUSY SO(10) GUT model. This is due to the same sign of the $C_{Q_{1}}^{\prime}$ and $C_{Q_{2}}^{\prime}$ which suppress the large contribution coming from the $C_{7}^{eff}$ and $C_{9}^{eff}$ in this model. Though SUSY effects are more distinguishable in FBAs in this case, however, it is too difficult to measure it experimentally due to its small value. (a)(b) --- Figure 4: Normal lepton polarization asymmetries for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$. Figs. (3-5) describe the lepton polarization asymmetries for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$. Before, we try to explain the behavior of different polarization asymmetries with the help the formulas (LABEL:expression-LP, 53, LABEL:expression-TP) given above. Eq. (LABEL:expression-LP) shows the dependence of the longitudinal lepton polarization on different Wilson coefficients, from which one can expect that the value of lepton polarization asymmetries in the SUSY I model should be greatly modified from that of the SM due to the change of sign for the term proportional to $C_{7}^{eff}C_{10}$. Due to this change in sign, the large positive contribution comes and the magnitude of the longitudinal polarization asymmetry decreases from that of the SM value. However, this value is expected to increase in SUSY II model because of the NHBs contribution which lies in the first and second term of Eq. (LABEL:expression-LP). In SUSY III model, this asymmetry lies close to that of the SM value due to the same sign of $C_{7}^{eff}C_{10}$ and small contribution from the NHBs. As we have considered all the primed Wilson coefficients to be zero therefore the effect of SUSY SO(10) on longitudinal lepton polarization asymmetry will be explained by plotting it with the square of the momentum transfer. Now, Fig. 3(a,b) shows the dependence of longitudinal polarization asymmetry for the $\Lambda_{b}\to\Lambda\mu^{+}\mu^{-}$ on the square of momentum transfer. The value in SUSY I model is significantly different from that of the SM, however, this value is close to that in the SM for SUSY II and SUSY III models. Furthermore, the absolute value of longitudinal polarization asymmetry in the SUSY SO(10) is small compared to the SM model due to the complex part of the Wilson coefficients and also due to small contributions of the NHBs in this model. --- (a)(b) Figure 5: Transverse lepton polarization asymmetries for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$. Fig. 3(c,d) are for the longitudinal lepton polarization asymmetries of $\Lambda_{b}\rightarrow\Lambda\tau^{+}\tau^{-}$ with and without long-distance contributions, where the different SUSY models are easily distinguishable. Contrary to the muon case, the values of this asymmetry in SUSY II and SUSY III models are even larger in magnitude than that obtained in the SM owing to the large contributions form NHBs which is attributed to the first and second term of Eq. (LABEL:expression-LP). Though the large contributions come from the first term which is proportional to $m_{l}$ in Eq. (LABEL:expression-LP), but this is overshadowed by the much larger term proportional to $m_{\Lambda_{b}}(C_{Q_{1}}^{\ast}C_{Q_{2}}+C_{Q_{2}}^{\ast}C_{Q_{1}})$. The dependence of lepton normal polarization asymmetries for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ on the momentum transfer square are presented in Fig. 4. In terms of Eq. (53), one can observe that this asymmetry is sensitive to the contribution of NHBs in SUSY II and SUSY III models, while it is insensitive to the contributions from SUSY I and SUSY SO(10) model. It can be seen that $P_{N}$ changes its sign in the case of large contributions from NHBs as indicated in Fig. 4 and this is also clear from the first three terms of Eq. (53). As expected, the contribution of NHBs from the $\tau^{+}\tau^{-}$ channel is much more significant than that from the $\mu^{+}\mu^{-}$ channel. Now, the normal polarization is proportional to the $\lambda$ which approaches to zero at large momentum transfer region and hence the normal polarization is suppressed by $\lambda$ in this region. (a)(b) \| \| ---\|---\|--- Figure 6: Longitudinal $\Lambda$ polarization asymmetries for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$. Fig. 5 shows the dependence of transverse polarization asymmetries for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ on the square of momentum transfer. From Eq. (LABEL:expression-TP) we can see that it is proportional to the imaginary part of the Wilson coefficient which are negligibly small in SM as well as in SUSY I, SUSY II and SUSY III models. However, complex flavor non- diagonal down-type squark mass matrix elements of 2nd and 3rd generations are of order one at GUT scale in SUSY SO(10) model, which induce complex couplings and Wilson coefficients. As a result, non zero transverse polarization asymmetries for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ exist in this model. Even though in this case, the asymmetry effects are quite small in $\tau^{+}\tau^{-}$ channel, but the value of transverse polarization asymmetry can reach to $-0.1$ when the momentum transfer is around $15\mathrm{GeV^{2}}$. Experimentally, to measure $\left\langle P_{T}\right\rangle$ of a particular decay branching ratio $\mathcal{B}$ at the $n\sigma$ level, the required number of events are $N=n^{2}/(\mathcal{B}\left\langle P_{T}\right\rangle^{2})$ and if $\left\langle P_{T}\right\rangle\sim 0.1$, then the required number of events are almost $10^{8}$ for $\Lambda_{b}$ decays. Since at LHC and BTeV machines, the expected number of $b\bar{b}$ production events is around $10^{12}$ per year, so the measurement of transverse polarization asymmetries in the $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ decays could discriminate the SUSY SO(10) model from the SM and other SUSY models. Fig. 6 shows the dependence of longitudinal polarization of $\Lambda$ baryon on the square of momentum transfer. One can see that the effects of NHBs are quite distinguishable in SUSY II and SUSY III models both for the $\mu^{+}\mu^{-}$ and $\tau^{+}\tau^{-}$ channels; but the values for SUSY I and SUSY II are almost close to that of the SM. As observed from Eq. (IV.4), the effects of NHBs are proportional to the mass of leptons, therefore the large deviations from the SM are expected for the tauon as presented in Fig. 6. In the SUSY II model, the value of the longitudinal polarization even changes its sign for $\tau^{+}\tau^{-}$ channel. (a)(b) \| \| ---\|---\|--- Figure 7: Normal $\Lambda$ polarization asymmetries for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$. With the help of Eq. (IV.4), one can see that the normal polarization asymmetry of $\Lambda$ baryon is sensitive to the $C_{Q_{1}}$ and the sign of the $C_{7}^{eff}$. It is shown that the sign of $C_{7}^{eff}$ is negative in SUSY I and II models. In particular, this asymmetry in the SUSY II model differs from that in the SM remarkably due to the large value of $C_{Q_{1}}$. Moreover, the contributions from SUSY III and SUSY SO(10) models are also quite distinguishable from the SM as shown in Fig. 7. Therefore, the measurements of normal polarization asymmetries for $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ in future experiments will help to distinguish different scenarios beyond the SM . (a)(b) \| \| ---\|---\|--- Figure 8: Transverse $\Lambda$ polarization asymmetries for the $\Lambda_{b}\to\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays as functions of $q^{2}$. Similar to the transverse polarization asymmetry of lepton, the transverse polarization asymmetry of $\Lambda$ baryon is also proportional to the imaginary part of $C_{Q_{1}}C_{7}^{eff}$ and $C_{Q_{1}}C_{9}^{eff}$ (c.f. Eq. (58)). It is known that these imaginary parts are quite small in SM, SUSY I, SUSY II and SUSY III model and hence the values of the transverse polarization asymmetries of $\Lambda$ baryon are almost zero in these models. However, the imaginary part of Wilson coefficient in SUSY SO(10) model is large and hence its effects are quite different from the other models discussed above as collected in Fig. 8. For the muon case, the transverse polarization asymmetry can reach the number $-0.1$ in SUSY SO(10) model; whereas the value is too small to measure experimentally for the tauon case. ## VI Conclusion We have carried out the study of invariant mass spectrum, FBAs, polarization asymmetries of lepton and $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays in SUSY theories including SUSY SO(10) GUT model. Particularly, we analyze the effects of NHBs to this process and our main outcomes can be summarized as follows: • The differential decay rates deviate sizably from that of the SM especially in the large momentum transfer region. These effects are significant in SUSY II model where the value of the Wilson coefficients corresponding to the NHBs is large. However, the SUSY SO(10) effects in differential decay rate of $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ are negligibly small . * • The SUSY effects show up for the FBA of the $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays and the deviations from the SM are very large especially in SUSY I and SUSY II model where the FBAs did not pass from zero. The reason is the same sign of $C_{7}^{eff}$ and $C_{9}^{eff}$ in these two models, but in SUSY III scenario it passes from the zero-point different from that of the SM. The effects of SUSY SO(10) are quite distinguishable when the final state leptons are the tauon pair but these are too small to be measured experimentally. * • The longitudinal, normal and transverse polarizations of leptons are calculated in different SUSY models. It is found that the SUSY effects are very promising which could be measured at future experiments and shed light on the new physics beyond the SM. * • Following the same line, the longitudinal, normal and transverse polarizations of $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decays are calculated at length. The different SUSY effects are clearly distinguishable from each other and also from that of the SM. The transverse polarization asymmetries are proportional to the imaginary part of the Wilson coefficients. Hence it is almost zero in SM as well as in MSSM model, however, the Wilson coefficients in SUSY SO(10) GUT model have large imaginary part and hence the value of the transverse polarization is expected to be non-zero. The maximum value of transverse polarization of $\Lambda_{b}\rightarrow\Lambda\mu^{+}\mu^{-}$ decay reaches to $-0.1$ for the square of momentum transfer around $10\mathrm{GeV^{2}}$ and hence can be measurable at future experiments like LHC and BTeV machines where a large number of $\ b\bar{b}$ pairs are expected to be produced. In short, the experimental investigation of observables, like decay rates, FBAs, lepton polarization asymmetries and the polarization asymmetries of $\Lambda$ baryon in $\Lambda_{b}\rightarrow\Lambda l^{+}l^{-}$ ($l=\mu,\tau$) decay will be used to search for the SUSY effects, in particular the NHBs effect, encoded in the MSSM as well as SUSY SO(10) models. ## Acknowledgements This work is partly supported by National Science Foundation of China under Grant No.10735080 and 10625525. The authors would like to thank Cheng Li, Yue- Long Shen and Qi-Shu Yan for helpful discussions. ## References * (1) CLEO Collaboration, M. S. Alam et al., Phys. Rev. Lett. 74 (1995) 2885. * (2) S. L. Glashow, J. Iliopoulos and L. Maiani, Phys. Rev. D 2 (1970) 1285. * (3) N. Cabibbo, Phys. Rev. Lett. 10 (1963) 531. * (4) M. Kobayashi and T. Maskawa, Prog. Theor. Phys. 49 (1973) 652. * (5) T. M. Aliev, A. Ozpineci and M. Savci, Phys. Rev. D 56 (1997) 4260 [arXiv:hep-ph/9612480]. * (6) P. Ball and V. M. Braun, Phys. Rev. D 58 (1998) 094016 [arXiv:hep-ph/9805422]. * (7) T. M. Aliev, C. S. Kim and Y. G. Kim, Phys. Rev. D 62 (2000) 014026 [arXiv:hep-ph/9910501]. * (8) T. M. Aliev, D. A. Demir and M. Savci, Phys. Rev. D 62 (2000) 074016 [arXiv:hep-ph/9912525]. * (9) W. Jaus and D. Wyler, Phys. Rev. D 41 (1990) 3405. * (10) A. Ali, P. Ball, L. T. Handoko and G. Hiller, Phys. Rev. D 61 (2000) 074024 [arXiv:hep-ph/9910221]. * (11) M. A. Paracha, I. Ahmed and M. J. Aslam, Eur. Phys. J. C 52 (2007) 967 [arXiv. hep-ph/0707.0733]; I. Ahmed, M. A. Paracha and M. J. Aslam, Eur. Phys. J. C 54 (2008) 591 [arXiv. hep-ph/0802.0740]; A. Saddique, M. J. Aslam and C. D. Lü, Appear in EPJC, [arXiv. hep-ph/0803.0192] * (12) Q. S. Yan, C. S. Huang, L. Wei and S. H. Zhu, Phys. Rev. D 62 (2000) 094023 [arXiv:hep-ph/0004262] * (13) C. S. Huang and Q. S. Yan, Phys. Lett. B 442 (1998) 209; C. S. Huang, W. Liao and Q. S. Yan, Phys. Rev. D 59 (1999) 011701 * (14) K. S. Babu and S. M. Bar, Phys. Lett. B 381 (1996) 202 [arXiv:hep-ph/9511446]; C. H. Albright, K. S. Babu and S. M. Bar, Phys. Rev. Lett. 81 1998 1167 [arXiv:hep-ph/9802314]; J. Sato and T. Yanagida, Phys. Lett. B 430 (1998) 127 [arXiv:hep-ph/9710516]; N. Irges, S. Lavignac and P. Ramond, Phys. Rev. D 58 (1998) 035003 [arXiv:hep-ph/9802334] * (15) M. C. Chen and K. T. Mahanthappa, Phys. Rev. D 62 (2000) 113007 [arXiv:hep-ph/0005292] * (16) D. Cheng, A. Masiro and H. Murayama, Phys. Rev. D 62(2003) 075013 [arXiv: hep-ph/0205111] * (17) B. Bajc, G. Senjanovic and F. Vissani, Phys. Rev. Lett. 90 (2003) 051802 [arXiv: hep-ph/0210207]; H. S. Goh, R. N. Mohapatra and S. -P. Ng, Phys. Lett. B 570 (2003) 215 [arXiv: hep-ph/0303055] * (18) W. J. Li, Y. B. Dai and C. S. Huang, Eur. Phys. J. C 40 (2005) 565 [arXiv: hep-ph/0410317] * (19) T. Mannel and S. Recksiegel, J. Phys. G 24 (1998) 979 [arXiv:hep-ph/9701399]. * (20) R. Mohanta, A. K. Giri, M. P. Khanna, M. Ishida and S. Ishida, Prog. Theor. Phys. 102 (1999) 645 [arXiv:hep-ph/9908291]. * (21) H. Y. Cheng, C. Y. Cheung, G. L. Lin, Y. C. Lin, T. M. Yan and H. L. Yu, Phys. Rev. D 51 (1995) 1199 [arXiv:hep-ph/9407303]. * (22) H. Y. Cheng and B. Tseng, Phys. Rev. D 53, 1457 (1996) [Erratum-ibid. D 55, 1697 (1997)] [arXiv:hep-ph/9502391]. * (23) C. S. Huang and H. G. Yan, Phys. Rev. D 59 (1999) 114022 [Erratum-ibid. D 61 (2000) 039901] [arXiv:hep-ph/9811303]. * (24) X. G. He, T. Li, X. Q. Li and Y. M. Wang, Phys. Rev. D 74 (2006) 034026 [arXiv:hep-ph/0606025]. * (25) Y. M. Wang, Y. Li and C. D. Lü, arXiv:hep-ph/0804.0648. * (26) C. H. Chen and C. Q. Geng, Phys. Rev. D 63 (2001) 054005 [arXiv:hep-ph/0012003]. * (27) C. H. Chen and C. Q. Geng, Phys. Rev. D 63 (2001) 114024 [arXiv:hep-ph/0101171]. * (28) C. H. Chen and C. Q. Geng, Phys. Lett. B 516 (2001) 327 [arXiv:hep-ph/0101201]. * (29) C. H. Chen and C. Q. Geng, Phys. Rev. D 64 (2001) 074001 [arXiv:hep-ph/0106193]. * (30) T. M. Aliev, A. Ozpineci and M. Savci, Nucl. Phys. B 649 (2003) 168 [arXiv:hep-ph/0202120]. * (31) T. M. Aliev, A. Ozpineci and M. Savci, Phys. Rev. D 67 (2003) 035007 [arXiv:hep-ph/0211447]. * (32) T. M. Aliev, V. Bashiry and M. Savci, Nucl. Phys. B 709 (2005) 115 [arXiv:hep-ph/0407217]. * (33) T. M. Aliev and M. Savci, JHEP 0605 (2006) 001 [arXiv:hep-ph/0507324]. * (34) C. S. Huang et al, Commun. Theor. Phys. 32 (1999) 499\. * (35) C. Bobeth, T. Ewerth, F. Kruger and J. Urban, Phys. Rev. D 66 (2002) 0740231 * (36) C. S. Huang and X. H. Wu, Nucl. Phys. B 657 (2003) 304 [arXiv:hep-ph/0212220]. * (37) E. Lunghi, A. Masiero, I. Scimemi and L. Silvestrini, Nucl. Phys. B 568 (2000) 120. * (38) T. Goto et al., Phys. Rev. D 55 (1997) 4273; T. Goto, Y. Okada and Y. Shimizu, Phys. Rev. D 58 (1998) 094006; S. Bertolini, F. Borzynatu, A. Masiero and G. Ridolfi, Nucl. Phys. B 353 (1991) 591. * (39) G. Buchalla, A. J. Buras and M. E. Lauthenbache, Rev. Mod. Phys. 68 (1996) 1125; A. Buras et al., Nucl. Phys. B 400 (1993) 75; S. Bertonlini et al., Nucl. Phys. B 353 (1991) 591. * (40) C. S. Huang, Nucl. Phys. Proc. Suppl. 93 (2001) 73; C. Bobeth, T. Ewerth, F. Kruger, J. Urban, Phys. Rev. D 64 (2001) 074014; Y. B. Dai, C. S. Huang, H. W. Huang, Phys. Lett. B 390 (1997) 257; Erratum-ibid. B 513 (2001) 429. * (41) C. S. Huang, X. H. Wu, Nucl. Phys. B 657 (2003) 304 [arXiv: hep-ph/0212220]; J. F. Cheng, C. S. Huang, X. H. Wu, arXiv. 0404055 [hep-ph] * (42) C.S. Kim, T. Morozumi, A.I. Sanda, Phys. Lett. B 218 (1989) 343. * (43) X. G. He, T. D. Nguyen and R. R. Volkas, Phys. Rev. D 38 (1988) 814. * (44) B. Grinstein, M.J. Savage, M.B. Wise, Nucl. Phys. B 319 (1989) 271. * (45) N. G. Deshpande, J. Trampetic and K. Panose, Phys. Rev. D 39 (1989) 1461. * (46) P. J. O’Donnell and H. K. K. Tung, Phys. Rev. D 43 (1991) 2067. * (47) N. Paver and Riazuddin, Phys. Rev. D 45 (1992) 978. * (48) A. Ali, T. Mannel and T. Morozumi, Phys. Lett. B273 (1991) 505. * (49) D. Melikhov, N. Nikitin and S. Simula, Phys. Lett. B 430 (1998) 332 [arXiv:hep-ph/9803343]. * (50) J. M. Soares, Nucl. Phys. B 367 (1991) 575. * (51) G. M. Asatrian and A. Ioannisian, Phys. Rev. D 54 (1996) 5642 [arXiv:hep-ph/9603318]. * (52) M. R. Ahmady, Phys. Rev. D 53 (1996) 2843 [arXiv:hep-ph/9508213]. * (53) F. Kruger and L. M. Sehgal, Phys. Lett. B 380 (1996) 199 [arXiv:hep-ph/9603237]. * (54) Z. Ligeti, I. W. Stewart and M. B. Wise, Phys. Lett. B 420 (1998) 359 [arXiv:hep-ph/9711248]. * (55) N. G. Deshpande, X. G. He and J. Trampetic, Phys. Lett. B 367 (1996) 362. * (56) J. M. Soares, Phys. Rev. D 53 (1996) 241 [arXiv:hep-ph/9503285]. * (57) I. I. Balitsky, V. M. Braun and A. V. Kolesnichenko, Nucl. Phys. B 312 (1989) 509. * (58) I. I. Balitsky, V. M. Braun and A. V. Kolesnichenko, Sov. J. Nucl. Phys. 44 (1986) 1028 [Yad. Fiz. 44 (1986) 1582]. * (59) V. M. Braun and I. E. Filyanov, Z. Phys. C 44 (1989) 157 [Sov. J. Nucl. Phys. 50 (1989 YAFIA,50,818-830.1989) 511.1989 YAFIA,50,818]. * (60) V. L. Chernyak and I. R. Zhitnitsky, Nucl. Phys. B 345 (1990) 137. * (61) P. Ball, V. M. Braun and H. G. Dosch, Phys. Rev. D 44 (1991) 3567. * (62) M. A. Shifman, A. I. Vainshtein and V. I. Zakharov, Nucl. Phys. B 147 (1979) 385. * (63) M. A. Shifman, A. I. Vainshtein and V. I. Zakharov, Nucl. Phys. B 147 (1979) 519. * (64) M. A. Shifman, A. I. Vainshtein and V. I. Zakharov, Nucl. Phys. B 147 (1979) 448. * (65) G. P. Lepage and S. J. Brodsky, Phys. Lett. B 87 (1979) 359. * (66) G. P. Lepage and S. J. Brodsky, Phys. Rev. D 22 (1980) 2157. * (67) S. J. Brodsky and G. P. Lepage, Phys. Rev. D 24 (1981) 1808. * (68) A. V. Efremov and A. V. Radyushkin, Phys. Lett. B 94 (1980) 245. * (69) A. V. Efremov and A. V. Radyushkin, Theor. Math. Phys. 42 (1980) 97 [Teor. Mat. Fiz. 42 (1980) 147]. * (70) V. L. Chernyak and A. R. Zhitnitsky, JETP Lett. 25 (1977) 510 [Pisma Zh. Eksp. Teor. Fiz. 25 (1977) 544]. * (71) V. L. Chernyak and A. R. Zhitnitsky, Sov. J. Nucl. Phys. 31 (1980) 544 [Yad. Fiz. 31 (1980) 1053]. * (72) V. L. Chernyak and A. R. Zhitnitsky, Phys. Rept. 112 (1984) 173. * (73) V. M. Braun, Plenary talk given at the IVth International Workshop on Progress in Heavy Quark Physics Rostock, Germany, 20–22 September 1997 arXiv:hep-ph/9801222. * (74) W.M. Yao et al., J. Phys. G 33, 1 (2006). * (75) T. M. Aliev and M. Savci, Eur. Phys. J. C 50 (2007) 91 [arXiv. hep-ph/0606225].	arxiv-papers	2008-08-15T13:06:29	2024-09-04T02:48:57.317710	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Jamil Aslam, Yu-Ming Wang and Cai-Dian Lu", "submitter": "Yuming Wang", "url": "https://arxiv.org/abs/0808.2113" }
0808.2144	# Flow Results and Hints of Incomplete Thermalization ###### Abstract We classified $v_{2}$ measurements according to their sensitivities w.r.t. to two planes, namely, reaction plane and participant plane. Likewise, in $v_{2}/\epsilon$ scaling, we showed that one needs to choose a $\epsilon$ that is sensitive to the same plane as that $v_{2}$ is sensitive to. We presented our $v_{2}/\epsilon$ as a function of centrality and transverse momentum. We studied the ratio of $v_{4}/v_{2}^{2}$. We discussed the applicable range for hydrodynamics, as well as implications to an incomplete thermalization. ## 1 Introduction In ultra-relativistic heavy ion collisions, spectators pass through each other quickly, and the system begins its evolution with what is left behind - the overlap region of two nuclei. The pressure gradient convert the spacial anisotropy, quantified by eccentricity $\epsilon$, of the overlap region into anisotropy in momentum space, quantified by azimuthal anisotropy $v_{2}$. $v_{2}$ is defined as the second Fourier coefficient in the description of particles distribution w.r.t. the reaction plane [1], and it is largely determined by the collective motion in the in-plane direction. $v_{2}$ at Relativistic Heavy Ion Collider (RHIC) is reported to be large and, for the first time in heavy ion collisions, can be described by ideal hydrodynamics [2]. Theoretical calculation shows that in order to explain the large $v_{2}$ observed at RHIC, one has to assume that the shear viscosity is extremely small [3]. That is one of the important reasons for which scientists think that a perfect liquid has been formed in relativistic heavy ion collisions [4]. Since the announcement of the discovery of a perfect liquid, our understanding of the matter created at RHIC has continued to advance. Elliptic flow analyses have been extended to great details by all four experiments at RHIC. At the same time, due to different techniques used, it becomes an increasingly amount of work to understand/compare results across different experiments and, sometimes, even different analyses within the same experiment. Therefore it becomes important to understand what is the relation between various $v_{2}$ measurements. For the physics side, in order to quantify how perfect the liquid is, it is necessary to re-examine the hydrodynamics limit. In this paper, we try to study the hydrodynamic behavior under a more general context, namely, the transport approach which recovers hydrodynamics when mean free path is extremely small if compared to the system size [5] ## 2 Choosing the right $v_{2}$ and $\epsilon$ pairs The ratio of $v_{2}/\epsilon$ reflects how well the initial anisotropy is converted into momentum anisotropy [6]. This conversion process is directly affected by dynamics of the system, e.g. Equation Of State, thermalization etc. It is important to measure this quantity as accurate as possible. However, there exist many $v_{2}$ measurements, for example, $v_{2}$ measured by event plane method ($v_{2}\mathrm{\\{EP\\}})$, by cumulants (${v_{2}\\{2\\}}$,${v_{2}\\{4\\}}$), by Lee-Yang zero method ($v_{2}\mathrm{\\{LYZ\\}}$), and by using event plane reconstructed with Shower Maximum Detectors at Zero Degree Calorimeters ($v_{2}\mathrm{\\{ZDCSMD\\}}$), etc. Likewise there exist many $\epsilon$ calculations. In this section, we will try to make connections between various $v_{2}$ and $\epsilon$ methods, and make the justification for the right combination of them. Define $\mbox{\boldmath$\varepsilon$}=\\{\varepsilon_{x},\varepsilon_{y}\\}=\left\\{\left<{\frac{\sigma_{y}^{2}-\sigma_{x}^{2}}{\sigma_{x}^{2}+\sigma_{y}^{2}}}\right>_{part},\left<{\frac{2\sigma_{xy}}{\sigma_{x}^{2}+\sigma_{y}^{2}}}\right>_{part}\right\\},$ (1) where $\sigma_{x}^{2}=\left<{x^{2}}\right>-\left<{x}\right>^{2}$, $\sigma_{y}^{2}=\left<{y^{2}}\right>-\left<{y}\right>^{2}$, and $\sigma_{xy}^{2}=\left<{xy}\right>-\left<{y}\right>\left<{x}\right>$, and the average is taken over the coordinates of the participants in a given event. With this definition, $\varepsilon_{x}$ is the eccentricity of reaction plane (defined by the impact parameter), and $\varepsilon_{x}$ is also called $\varepsilon_{RP}$. $\varepsilon_{y}$ has a distribution centered at zero with finite width. The participant eccentricity measures the asymmetry in the participant plane (defined by the principle axis of the ellipsoid), and is given by $\varepsilon_{part}=\sqrt{\varepsilon_{x}^{2}+\varepsilon_{y}^{2}}\equiv\varepsilon_{PP}$ When both $\varepsilon_{x}$ and $\varepsilon_{y}$ has a Gaussian distribution, to the first order this is true and it is supported by Glauber Monte Carlo simulations, then the probability density function for $\varepsilon_{PP}$ is $\frac{dn}{d\varepsilon_{part}}=\frac{\varepsilon_{part}}{\sigma_{\varepsilon}^{2}}I_{0}\left(\frac{\varepsilon_{part}\left<{\varepsilon_{RP}}\right>}{\sigma_{\varepsilon}^{2}}\right)\exp\left(-\frac{\varepsilon_{part}^{2}+\left<{\varepsilon_{RP}}\right>^{2}}{2\sigma_{\varepsilon}^{2}}\right)\equiv\mathrm{BG}(\varepsilon_{part};\left<{\varepsilon_{RP}}\right>,\sigma_{\varepsilon}),$ (2) With this p.d.f., one can show that $\varepsilon_{part}\\{4\\}=\left<{\varepsilon_{RP}}\right>$ [7]. Similarly, under the assumption that $v_{2}$ is proportional to the initial system eccentricity (this is not true over a broad centrality range, but for a fine centrality bin, it still ensures a good Gaussian for $v_{2}$ and $s_{2}(\equiv\left<{\mathrm{sin2}(\phi-\Psi_{RP})}\right>)$, which is what is required for the derivation of BG formula shown in Eq.( 2) ), one can find that ${v_{2}\\{4\\}}=\left<{v_{2}\\{RP\\}}\right>$ [7, 8]. That means that, ${v_{2}\\{4\\}}$ and $\varepsilon_{part}\\{4\\}$ are measurements sensitive to the reaction plane, not the participant plane. Indeed, that explains the reason that STAR’s $v_{2}\mathrm{\\{ZDCSMD\\}}$ agrees with ${v_{2}\\{4\\}}$, as shown in the left panel of Fig.1. $v_{2}\mathrm{\\{ZDCSMD\\}}$ is measured with the first order event plane reconstructed by spectator neutrons thus is more sensitive to $v_{2}$ in the reaction plane (not participant plane). Fig. 1.: Left: $v_{2}$ measurements as a function of centrality. This plot is from [9]. Right: $v_{2}$ of Au+Au collisions at $\sqrt{s_{NN}}=130\mathrm{GeV}$. This plot is made based on datapoints from [10] Because the p.d.f. of the magnitude of the flow vector, the $q-$distribution [10], shares an almost identical formula as Eq. (2), the $v_{2}$ obtained from fitting the $q-$distribution should be equivalent to ${v_{2}\\{4\\}}$, as confirmed by experimental data in the right panel of Fig.1. To summarize this section, we find that ${v_{2}\\{4\\}}$, $v_{2}\mathrm{\\{ZDCSMD\\}}$ and $v_{2}\mathrm{\\{q-dist\\}}$ are sensitive to anisotropy in the reaction plane, they should be scaled with the standard eccentricity or 4-particle cumulant eccentricity. Other $v_{2}$ measurements that are based on two particle correlations, that includes $v_{2}\\{2\\}$, $v_{2}\mathrm{\\{EP\\}}$ and $v_{2}\mathrm{\\{scalarProduct\\}}$ [10], etc., should be scaled with participant eccentricity or the 2-particle cumulant eccentricity. ## 3 Hints of Incomplete Thermalization The large data set from run IV Au+Au collisions at $\sqrt{s_{NN}}=200\mathrm{GeV}$ allows us to extend ${v_{2}\\{4\\}}$ measurement to large $p_{t}$ and in fine centrality bins (see Fig. 2). For the reason stated in the previous section, we scale ${v_{2}\\{4\\}}$ by initial standard eccentricities. The left plot of Fig. 2 shows ${v_{2}\\{4\\}}$ scaled by the eccentricity from Color Glass Condensate (CGC) [11], and the right plot shows ${v_{2}\\{4\\}}$ scaled by the Monte Carlo Glauber eccentricity for wounded nucleons. As expected, for the CGC case, the magnitude of $v_{2}/\epsilon$ is lower if compared to the ratio in which a Glauber eccentricity is used. For both cases, we see the ratio rises from peripheral events to central events, indicating that stronger flow has been developed in central collisions. We also notice that the $p_{t}$ where $v_{2}$ reaches its maximum increases from peripheral collisions to central collisions, which is consistent with the expectation that the applicable range for hydrodynamics extends to large pt in central collisions. Note that $v_{2}/\epsilon$ shows sign of saturation in central collisions for the CGC case but not much for the Glauber case. This is explained by [11] as the following: in very peripheral collisions, due to little asymmetry in the saturation scales, the CGC eccentricity approaches the same value as that in the Glauber model, but in central collisions, CGC predicts a larger eccentricity than the Glauber model when there is a large asymmetry in the local saturation scales of the collisions partners, along a path in impact-parameter direction away from the origin. Note that $v_{2}\\{2\\}$ is not suitable for this study because it is more susceptible to nonflow at large $p_{t}$. Fig. 2.: $v_{2}$ scaled by initial CGC eccentricity (left) and Glauber eccentricity (right) as a function of $p_{t}$. This plot is from [12]. To understand how well hydrodynamics describes STAR’s $v_{2}$, we investigated the behavior of $v_{2}$ under the contex of the transport model, which will be reduced to hydrodynamics when the mean free path is much smaller than the system size [5]. In such approach, the dependence of $v_{2}/\epsilon$ on particle’s density in the transverse plane ($1/SdN/dy$) can be described by: $\frac{v_{2}}{\epsilon}=\left[\frac{v_{2}}{\epsilon}\right]_{hydro}\frac{1}{1+K/K_{0}}=\left[\frac{v_{2}}{\epsilon}\right]_{hydro}\frac{1}{1+\left(\sigma\frac{c_{s}}{c}\frac{1}{4S}\frac{dN}{dy}\right)^{-1}\frac{1}{K_{0}}}$ (3) Where $K$ is Knudsen number defined by the mean free path divided by the system size(sometimes it is more convenient to use $K^{-1}$ which means number of collisions a particle encounters before it escapes), and $K_{0}$ is a constant can be determined through transport calculations. In our study, we take $K_{0}=0.7$ following the suggestion of Ollitrault [11]. A factor of 4 in front of $S$ is to take into account the different definition of $S$ between STAR ($S=\pi\sqrt{\left<{x^{2}}\right>\left<{y^{2}}\right>}$) and [11]. (($S=4\pi\sqrt{\left<{x^{2}}\right>\left<{y^{2}}\right>}$). $\left[\frac{v_{2}}{\epsilon}\right]_{hydro}$ and $\sigma$ are free parameters that have to be determined from fitting the data. In this approach hydrodynamic limit of $v_{2}/\epsilon$ can be never reached, but can be only asymptotic approached. Fig. 3.: $v_{2}\\{4\\}$ scaled by initial eccentricities, as a function of centrality. Fitted hydrodynamic limits for $v_{2}/\epsilon$ are indicated by horizontal lines. CGC eccentricity and overlap area S are from [11]. Fig. 3 shows ${v_{2}\\{4\\}}$ scaled by CGC initial eccentricity and Glauber initial eccentricity for wounded nucleons, as a function of particle density in the transverse plane, with fits to Eq. 3. The fitted hydrodynamic limits for the ratio of $v_{2}/\epsilon$ are $0.23$ and $0.36$, for CGC case and Glauber case, respectively. For the same reason mentioned above, we see that the curve shows a hint of saturation for the CGC case, but not for the Glauber case, due to the relatively larger $\epsilon$ for CGC case in central collisions. It is interesting to see that, for central Au+Au collisions, the ratio of $v_{2}/\epsilon$ is about $20-30\%$ away from hydrodynamic limits. It means that, there is still significant room for flow to grow before the system saturates at hydrodynamic limits. From the simple observation that both $v_{2}$ and $v_{4}$ are proportional to $K^{-1}$ for small $K^{-1}$, one expects that $v_{4}/v_{2}^{2}$ decreases with $K^{-1}$, reaching a minimum when the hydrodynamical regime is reached. For this reason the ratio of $v_{4}/v_{2}^{2}$ has been argued as a probe to test the degree of thermalization. Fig. 4 shows STAR’s measurement of $v_{4}/v_{2}^{2}$ as a function of transverse momentum. The major systematic uncertainty in this measurement comes from $v_{2}$ [12]. In this analysis, the induced systematic error from $v_{2}$ uncertainty is estimated by studying the difference between ${v_{2}\\{4\\}}$ and $v_{2}$ measured with event plane constructed by tracks from STAR’s Forward Time Projection Chamber (FTPC). This advanced study reduces the previously-reported systematic error at QM06 conference by $40\%$ relatively. The dashed lines are ratio come out of calculations by solving Boltzmann equations with Monte Carlo simulation, with different Knudsen number $K$. When the Knudsen number is small, it recovers the hydrodynamic limit as indicated the solid line. The plot shows that the system exhibits, again, significant deviation from ideal hydrodynamic limit ($K<<1$), and the data is consistent with a incomplete thermalized system with $K>0.5$. Fig. 4.: $v_{4}\mathrm{\\{EP_{2}\\}}/v_{2}^{2}\\{4\\}$ as a function of $p_{t}$. This plot is from [12] with an advanced estimation on systematics represented by half box-brackets. ## 4 Conclusion To summarize, we found that ${v_{2}\\{4\\}}$, $v_{2}\mathrm{\\{ZDCSMD\\}}$ and $v_{2}\mathrm{\\{q-dist\\}}$ are all sensitive to azimuthal correlation w.r.t the reaction plane, not the participant plane,thus they should be scaled by eccentricities that are sensitive to reaction plane too. That includes standard eccentricity and 4-particle cumulant eccentricity. For $v_{2}$ methods that are based on two particle correlations, they are sensitive to the azimuthal correlation w.r.t the participant plane, and they need to be scaled by the corresponding eccentricities that are sensitive to the participant plane. That includes participant eccentricity or 2-particle cumulant eccentricity. We found that from peripheral to central Au+Au collisions flow increases, and the applicable range for hydrodynamics extends to larger $p_{t}$. However, $v_{2}/\epsilon$ and $v_{4}/v_{2}^{2}$ shows significant deviation from ideal hydrodynamic limit, when that limit is extracted from fitting the data itself with a Boltzmann equation motivated formula. Our study shows that although in general hydrodynamic does a good job in terms of describing $v_{2}$ at RHIC, there are features that are not consistent with a complete thermalization and they cannot be easily dismissed. ## References * [1] A.M. Poskanzer and S.A. Voloshin, Phys. Rev. C58 (1998) 1671. * [2] K.H. Ackermann _et al._ , [STAR Collaboration] Phys. Rev. Lett. 86 (2001) 402. * [3] D. Teaney Phys. Rev. C68 (2003) 034913. * [4] J. Adams _et al._ , [STAR Collaboration] Nucl. Phys. A757 (2005) 102; “New Discoveries at RHIC”, Proceedings of RIKEN BNL Research Center Workshop, 62 (2004). * [5] C. Gombeaud and J.-Y. Ollitrault, e-print nucl-th/0702075; * [6] J.-Y. Ollitrault, Phys. Rev. D46 (1992) 229; P.F. Kolb, J. Sollfrank, and U. Heinz, Phys. Rev. C62 (2000) 054909. * [7] S. Voloshin, A. Poskanzer, A. Tang, and G. Wang, Phys. Lett. B659 (2008) 537. * [8] R. Bhalerao and J.-Y. Ollitrault, Phys. Lett. B614 (2006) 260; * [9] G. Wang for the STAR collaboration,Nucl. Phys. A774, (2006) 515. * [10] C. Adler _et al._ , [STAR Collaboration], Phys. Rev. C66, (2002) 034904. * [11] H.-J. Drescher, A. Dumitru, C. Gombeaud, and J.-Y. Ollitrault, Phys. Rev. C76 (2007) 024905. * [12] Y. Bai, Ph. D. Thesis, NIKHEF and Utrecht University, the Netherlands (2007)	arxiv-papers	2008-08-15T13:43:09	2024-09-04T02:48:57.325235	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Aihong Tang (for the STAR Collaboration)", "submitter": "Aihong Tang", "url": "https://arxiv.org/abs/0808.2144" }
0808.2149	# Entangled state for constructing generalized phase space representation and its statistical behavior 1Li-yun Hu and 1,2Hong-yi Fan 1Department of Physics, Shanghai Jiao Tong University, Shanghai, 200030, China 2Department of Material Science and Engineering, University of Science and Technology of China, Hefei, Anhui 230026, China ###### Abstract Based on the conception of quantum entanglement of Einstein-Podolsky-Rosen we construct generalized phase space representation associated with the entangled state $\left\|\Gamma\right\rangle_{e}$, which is endowed with definite physical meaning. The set of states make up a complete and non-orthogonal representation. The Weyl ordered form of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ is derived which clearly exhibit the statistical behavior of marginal distribution of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|.$ The minimum uncertainty relation obeyed by $\left\|\Gamma\right\rangle_{e}$ is also demonstrated. ## 1 Introduction Phase space formalism of quantum mechanics has many applications in quantum statistics, quantum optics and quantum information theory. It began with Wigner’s celebrated paper in 1932 [1]. Among many kinds of pseudo-probability distribution functions, Wigner function $W(q,p)$ of a quantum state (pure or mixed states) is the most popularly used, since in phase space it exhibits two marginal distribution as the following way [2], $\mathrm{P}(p)=\int_{-\infty}^{\infty}W(q,p)dq,\;\mathrm{P}\left(q\right)=\int_{-\infty}^{\infty}W(q,p)dp,$ (1) where $\mathrm{P}\left(q\right)\;\left[\mathrm{P}(p)\right]\;$is proportional to the probability for finding the particle at $q$ [at $p$ in momentum space]. Besides, the Wigner operator also serves as an integral kernel of the Weyl rule [3, 4] which is a quantization scheme connecting classical functions of ($q,p$) with their quantum correspondence operators of ($Q,P$). The single- mode Wigner operator in the coordinate representation is $\Delta(q,p)=\frac{1}{2\pi}\int_{-\infty}^{\infty}\left\|q-\frac{v}{2}\right\rangle\left\langle q+\frac{v}{2}\right\|e^{-ipv}dv,$ (2) where $\left\|q\right\rangle$ is the coordinate eigenvector, $Q\left\|q\right\rangle=q\left\|q\right\rangle,$ $\left\|q\right\rangle=\pi^{-1/4}\exp\left[-\frac{q^{2}}{2}+\sqrt{2}qa^{\dagger}-\frac{1}{2}a^{\dagger 2}\right]\left\|0\right\rangle.$ (3) Using the normally ordered form of vacuum projector $\left\|0\right\rangle\left\langle 0\right\|=\colon\exp[-a^{\dagger}a]\colon,$ where $\colon\colon$ denotes normal ordering, and the technique of integration within an ordered product (IWOP) of operators [5, 6], the integration in Eq. (2) can be performed, leading to the explicit operator $\Delta(q,p)=\frac{1}{\pi}\colon e^{-(q-Q)^{2}-(p-P)^{2}}\colon=\frac{1}{\pi}\colon e^{-2(a^{\dagger}-\alpha^{\ast})(a-\alpha)}\colon\equiv\Delta(\bar{\alpha},\bar{\alpha}^{\ast}),\;\;$ (4) where $Q$ and $P$ are related to the Bose creation and annihilation operators $\left(a^{\dagger},a\right)$ by $Q=(a+a^{\dagger})/\sqrt{2}$ and $P=(a-a^{\dagger})/(i\sqrt{2}),$ respectively, $[a,a^{\dagger}]=1,$ $\bar{\alpha}=\left(q+ip\right)/\sqrt{2}.$ Obviously, $\int_{-\infty}^{\infty}\Delta(q,p)dp=\frac{1}{\sqrt{\pi}}\colon e^{-(q-Q)^{2}}\colon=\left\|q\right\rangle\left\langle q\right\|,$ $\int_{-\infty}^{\infty}\Delta(q,p)dq=\frac{1}{\sqrt{\pi}}\colon e^{-(p-P)^{2}}\colon=\left\|p\right\rangle\left\langle p\right\|,$ where $\left\|p\right\rangle$ is the momentum eigenvector. This formalism helps us to understand the Wigner function’s role better[2][7]-[10]. In two-mode case when two systems are prepared in an entangled state, measuring one of the two canonically conjugate variables on one system, the value for a physical variable in the another system may be inferred with certainty, this is the quantum entanglement. Because entanglement is now widely used in quantum information and quantum computation, it has been paid much attention by physicists [11]. The original idea of quantum entanglement began with Einstein-Podolsky-Rosen’s observation in EPR’s treatment [12], who noticed that two particles’ relative coordinate $Q_{1}-Q_{2}$ and total momentum $P_{1}+P_{2}$ can be simultaneously measured, (also their conjugate variables $\left[Q_{1}+Q_{2},P_{1}-P_{2}\right]=0),$ therefore, the corresponding Wigner function should be such that its two marginal distributions are respectively proportional to the probability for finding the two particles which possess certain total momentum value $\left[\text{relative momentum value}\right]$ and simultaneously relative position value $\left[\text{center-of-mass position value}\right]$ (see also Eqs. (9) and (12)). The investigation of Wigner functions for entangled states is not only just for the convenience of some calculations, but also for revealing the intrinsic entanglement property inherent to some physical systems. In Ref. [13] by virtue of the well-behaved properties of the entangled state representation $\left\langle\eta\right\|$, $\left\|\eta\right\rangle=\exp\left\\{-\frac{1}{2}\left\|\eta\right\|^{2}+\eta a_{1}^{\dagger}-\eta^{\ast}a_{2}^{\dagger}+a_{1}^{\dagger}a_{2}^{\dagger}\right\\}\left\|00\right\rangle,\;\eta=\eta_{1}+i\eta_{2},$ (5) we have successfully established the so-called entangled Wigner operator for correlated two-body systems [13], $\Delta_{w}\left(\rho,\varsigma\right)=\int\frac{d^{2}\eta}{\pi^{3}}\left\|\rho-\eta\right\rangle\left\langle\rho+\eta\right\|\exp(\eta\varsigma^{\ast}-\eta^{\ast}\varsigma),$ (6) $\left\|\eta\right\rangle$ is the common eigenvector of $Q_{1}-Q_{2}$ and $P_{1}+P_{2}$ [14], which obeys the eigenvector equations $(Q_{1}-Q_{2})\left\|\eta\right\rangle=\sqrt{2}\eta_{1}\left\|\eta\right\rangle,\;\;(P_{1}+P_{2})\left\|\eta\right\rangle=\sqrt{2}\eta_{2}\left\|\eta\right\rangle.$ (7) Using the IWOP technique we have shown in [13] that $\Delta_{w}\left(\rho,\varsigma\right)$ is just the product of two independent single-mode Wigner operators $\Delta_{w}\left(\rho,\varsigma\right)=\Delta(\bar{\alpha},\bar{\alpha}^{\ast})\Delta(\bar{\beta},\bar{\beta}^{\ast})$ provided we take $\varsigma=\bar{\alpha}+\bar{\beta}^{\ast},\;\rho=\bar{\alpha}-\bar{\beta}^{\ast},\text{\ \ }\bar{\alpha}=\frac{q_{1}+ip_{1}}{\sqrt{2}},\text{ }\bar{\beta}=\frac{q_{2}+ip_{2}}{\sqrt{2}}.$ (8) Performing the integration of $\Delta_{w}\left(\rho,\varsigma\right)$ over $d^{2}\varsigma$ leads to the projection operator of the entangled state $\left\|\eta\right\rangle$ $\int d^{2}\varsigma\Delta_{w}(\rho,\varsigma)=\frac{1}{\pi}\left\|\eta\right\rangle\left\langle\eta\right\|\|_{\eta=\rho},\text{ }$ (9) and the marginal distribution in ($\eta_{1},\eta_{2}$) phase space is $\left\langle\psi\right\|\int d^{2}\varsigma\Delta_{w}(\rho,\varsigma)\left\|\psi\right\rangle=\frac{1}{\pi}\|\psi(\eta)\|^{2}\|_{\eta=\rho}.$(in reference to (7)). Similarly, we can introduce the common eigenvector of$\;Q_{1}+Q_{2}$ and $P_{1}-P_{2}$ [14] (the conjugate state of $\left\|\eta\right\rangle)$ $\left\|\xi\right\rangle=\exp\left\\{-\frac{1}{2}\left\|\xi\right\|^{2}+\xi a_{1}^{\dagger}+\xi^{\ast}a_{2}^{\dagger}-a_{1}^{\dagger}a_{2}^{\dagger}\right\\}\left\|00\right\rangle,\;\;\xi=\xi_{1}+i\xi_{2},$ (10) which obeys another pair of eigenvector equations $\left(Q_{1}+Q_{2}\right)\left\|\xi\right\rangle=\sqrt{2}\xi_{1}\left\|\xi\right\rangle,\;\;\left(P_{1}-P_{2}\right)\left\|\xi\right\rangle=\sqrt{2}\xi_{2}\left\|\xi\right\rangle.$ (11) Performing the integration of $\Delta_{w}\left(\rho,\varsigma\right)$ over $d^{2}\rho$ yields $\int d^{2}\rho\Delta_{w}(\rho,\varsigma)=\frac{1}{\pi}\left\|\xi\right\rangle\left\langle\xi\right\|\|_{\xi=\varsigma},\;\left\langle\psi\right\|\int d^{2}\rho\Delta_{w}(\rho,\varsigma)\left\|\psi\right\rangle=\frac{1}{\pi}\|\psi(\xi)\|^{2}\|_{\xi=\varsigma}.$ (12) The introduction of the entangled Wigner operator also brings much convenience for calculating the Wigner function of some entangled states. Working in the $\left\|\eta\right\rangle or\left\|\xi\right\rangle$ representation one can interrelate some physical systems. For example, the Einstein-Podolsky-Rosen arrangement relies on free propagation of quantum- coupled particles (described by $\left\|\eta\right\rangle$ or $\left\|\xi\right\rangle)$, on the other hand, the two-mode squeezed state dealing with oscillators which are bound systems, these two seemingly very different physical systems can be interrelated by constructing the following ket-bra integration in terms of $\left\|\eta\right\rangle,$ $\int\frac{d^{2}\eta}{\pi\mu}\|\frac{\eta}{\mu}\rangle\langle\eta\|=\exp\left[\left(a_{1}^{\dagger}a_{2}^{\dagger}-a_{1}a_{2}\right)\ln\mu\right],$ (13) the right hand-side of (13) is just the two-mode squeezing operator [7, 15]. In the following we shall employ the $\left\|\eta\right\rangle$ state to formulating generalized phase space representation and then study its statistical behaviors, i.e. enlightened by Eqs. (7) and (11) we construct generalized phase space representation characteristic of the properties under the action of $Q_{1}-Q_{2}$ and $P_{1}+P_{2}$, associated with a two-mode state vector $\left\|\Gamma\right\rangle_{e}$, ${}_{e}\left\langle\Gamma\right\|\frac{Q_{1}-Q_{2}}{\sqrt{2}}$ $\displaystyle=$ $\displaystyle\left(\alpha\sigma_{1}+i\beta\frac{\partial}{\partial\tau_{2}}\right)\text{ }_{e}\left\langle\Gamma\right\|,$ ${}_{e}\left\langle\Gamma\right\|\frac{P_{1}-P_{2}}{\sqrt{2}}$ $\displaystyle=$ $\displaystyle\left(\gamma\tau_{2}+i\delta\frac{\partial}{\partial\sigma_{1}}\right)\text{ }_{e}\left\langle\Gamma\right\|,$ (14) where the subscript “$e$” implies the entanglement, $\alpha,\beta,\gamma$ and $\delta$ are all real parameters, satisfying $\beta\gamma-\alpha\delta=1,$ (15) and $\left[Q_{1}-Q_{2},P_{1}-P_{2}\right]=2i.$ Simultaneously, under the action of the center-of-mass operator and the relative momentum operator, the state ${}_{e}\left\langle\Gamma\right\|$ behaves ${}_{e}\left\langle\Gamma\right\|\frac{Q_{1}+Q_{2}}{\sqrt{2}}$ $\displaystyle=$ $\displaystyle\left(\gamma\tau_{1}-i\delta\frac{\partial}{\partial\sigma_{2}}\right)\text{ }_{e}\left\langle\Gamma\right\|,$ ${}_{e}\left\langle\Gamma\right\|\frac{P_{1}+P_{2}}{\sqrt{2}}$ $\displaystyle=$ $\displaystyle\left(\alpha\sigma_{2}-i\beta\frac{\partial}{\partial\tau_{1}}\right)\text{ }_{e}\left\langle\Gamma\right\|.$ (16) The present paper is arranged as follows. In Sec. 2 using the newly developed bipartite entangled state representation $\|\eta\rangle$ of continuum variables we shall derive the concrete form of entangled state $\left\|\Gamma\right\rangle_{e}$ in two-mode Fock space and then analyze its properties, in so doing the phase space theory can be developed to the entangled case. In Sec. 3, the completeness relation and non-orthonormal property of $\left\|\Gamma\right\rangle_{e}$ are proved. In Sec. 4. the Weyl ordered form of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ is derived, which yields the classical correspondence of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$. In Sec. 5 we examine marginal distributions of the operator $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ by using the properties of the entangled state $\|\eta\rangle$ and its conjugate state $\left\|\xi\right\rangle$. The uncertainty relation of coordinate and momentum quadratures in $\left\|\Gamma\right\rangle_{e}$ and the Wigner function of $\left\|\Gamma\right\rangle_{e}$ are calculated in sections 6 and 7, respectively. ## 2 The state $\left\|\Gamma\right\rangle_{e}$ in two-mode Fock space We find that the explicit form of the state $\left\|\Gamma\right\rangle_{e}$ in two-mode Fock space is (see the Appendix), $\left\|\Gamma\right\rangle_{e}\equiv 2\sqrt{-\alpha\beta\gamma\delta}\exp\left[\frac{\alpha\left\|\sigma\right\|^{2}}{2\delta}-\frac{\gamma\left\|\tau\right\|^{2}}{2\beta}+\left(\alpha\sigma+\gamma\tau\right)a_{1}^{\dagger}+\left(\gamma\tau^{\ast}-\alpha\sigma^{\ast}\right)a_{2}^{\dagger}-\left(\beta\gamma+\alpha\delta\right)a_{1}^{\dagger}a_{2}^{\dagger}\right]\left\|00\right\rangle,$ (17) where $\sigma=\sigma_{1}+i\sigma_{2},$ $\tau=\tau_{1}+i\tau_{2}$; real numbers ($\alpha,\beta,\gamma$ and $\delta$) satisfy the relation Eq.(15); $(a_{i},a_{i}^{{\dagger}}),$ $i=1,2,$ are the two-mode Bose annihilation and creation operators obeying $\left[a_{i},a_{j}^{\dagger}\right]=\delta_{ij}$. To satisfy the square integrable condition for wave function in phase space $\left\|\Gamma\right\rangle_{e},$ $\frac{\alpha}{\delta}<0$ and $\frac{\gamma}{\beta}>0$ are demanded. In order to certify that Eq.(17) really obeys Eqs. (14) and (16) we operate $a_{i}$ on $\left\|\Gamma\right\rangle_{e},$ $\displaystyle a_{1}\left\|\Gamma\right\rangle_{e}$ $\displaystyle=$ $\displaystyle\left[\left(\alpha\sigma+\gamma\tau\right)-\left(\beta\gamma+\alpha\delta\right)a_{2}^{\dagger}\right]\left\|\Gamma\right\rangle_{e},$ $\displaystyle a_{2}\left\|\Gamma\right\rangle_{e}$ $\displaystyle=$ $\displaystyle\left[\left(\gamma\tau^{\ast}-\alpha\sigma^{\ast}\right)-\left(\beta\gamma+\alpha\delta\right)a_{1}^{\dagger}\right]\left\|\Gamma\right\rangle_{e}.$ (18) Then noting the relation between $Q_{i},P_{j}$ and $a_{i},a_{j}^{{\dagger}}$, $Q_{i}=(a_{i}+a_{i}^{{\dagger}})/\sqrt{2},\ P_{i}=(a_{i}-a_{i}^{{\dagger}})/(\sqrt{2}\mathtt{i}),$ (19) and Eq.(15) as well as $\displaystyle\frac{\partial}{\partial\sigma}\left.{}_{e}\left\langle\Gamma\right\|\right.$ $\displaystyle=$ $\displaystyle\left.{}_{e}\left\langle\Gamma\right\|\left(\frac{\alpha\sigma^{\ast}}{2\delta}-\alpha a_{2}\right)\right.,\text{ }\frac{\partial}{\partial\sigma^{\ast}}\left.{}_{e}\left\langle\Gamma\right\|\right.=\left.{}_{e}\left\langle\Gamma\right\|\left(\frac{\alpha\sigma}{2\delta}+\alpha a_{1}\right)\right.,$ $\displaystyle\frac{\partial}{\partial\tau}\left.{}_{e}\left\langle\Gamma\right\|\right.$ $\displaystyle=$ $\displaystyle\left.{}_{e}\left\langle\Gamma\right\|\left(-\frac{\gamma\tau^{\ast}}{2\beta}+\gamma a_{2}\right)\right.,\text{ }\frac{\partial}{\partial\tau^{\ast}}\left.{}_{e}\left\langle\Gamma\right\|\right.=\left.{}_{e}\left\langle\Gamma\right\|\left(-\frac{\gamma\tau}{2\beta}+\gamma a_{1}\right)\right.,$ (20) we see, for example, ${}_{e}\left\langle\Gamma\right\|\frac{Q_{1}+Q_{2}}{\sqrt{2}}$ $\displaystyle=$ $\displaystyle\left.{}_{e}\left\langle\Gamma\right\|\right.\left[-\delta\left(\alpha a_{1}+\alpha a_{2}\right)-i\alpha\sigma_{2}+\gamma\tau_{1}\right]$ (21) $\displaystyle=$ $\displaystyle\left[\gamma\tau_{1}+\delta\left(\frac{\partial}{\partial\sigma}-\frac{\partial}{\partial\sigma^{\ast}}\right)\right]\left.{}_{e}\left\langle\Gamma\right\|\right.$ $\displaystyle=$ $\displaystyle\left(\gamma\tau_{1}-i\delta\frac{\partial}{\partial\sigma_{2}}\right)\left.{}_{e}\left\langle\Gamma\right\|\right.,$ which is the first equation in Eq.(16). In a similar way, ${}_{e}\left\langle\Gamma\right\|$ satisfying the other equations in Eqs.(14) and (16) can be checked. Using Eqs.(14), (16) and noticing the quantum commutator $\left[\frac{Q_{1}\pm Q_{2}}{\sqrt{2}},\frac{P_{1}\pm P_{2}}{\sqrt{2}}\right]=i,$ (22) in the phase space representation, we have ${}_{e}\left\langle\Gamma\right\|\left[\frac{Q_{1}\pm Q_{2}}{\sqrt{2}},\frac{P_{1}\pm P_{2}}{\sqrt{2}}\right]=i\left(\beta\gamma-\alpha\delta\right)_{e}\left\langle\Gamma\right\|,$ (23) which results in the condition shown in Eq.(15). ## 3 The properties of $\left\|\Gamma\right\rangle_{e}$ ### 3.1 The completeness relation of $\left\|\Gamma\right\rangle_{e}$ Next we prove the completeness relation of Eq.(17). Using the normally ordered vacuum projector $\left\|00\right\rangle\left\langle 00\right\|=\colon\exp\left(-a_{1}^{\dagger}a_{1}-a_{2}^{\dagger}a_{2}\right)\colon,$ (24) where $\colon\colon$ denotes the normal product, which means all the bosonic creation operators are standing on the left of annihilation operators in a monomial of $a^{\dagger}$ and $a$ [16]. It should be emphasized that a normally ordered product of operators can be integrated with respect to $c$-numbers provided the integration is convergent. Then we can use Eq.(17) and the IWOP technique to perform the following integration $\displaystyle\frac{1}{\beta^{2}\delta^{2}}\int\frac{d^{2}\sigma d^{2}\tau}{4\pi^{2}}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ (25) $\displaystyle=$ $\displaystyle-\frac{\alpha\gamma}{\beta\delta}\int\frac{d^{2}\sigma d^{2}\tau}{\pi^{2}}\colon\exp\left[\frac{\alpha\left\|\sigma\right\|^{2}}{\delta}+\sigma\alpha\left(a_{1}^{\dagger}-a_{2}\right)+\sigma^{\ast}\alpha\left(a_{1}-a_{2}^{\dagger}\right)-a_{1}^{\dagger}a_{1}\right.$ $\displaystyle\left.-\frac{\gamma\left\|\tau\right\|^{2}}{\beta}+\tau\gamma\left(a_{1}^{\dagger}+a_{2}\right)+\tau^{\ast}\gamma\left(a_{2}^{\dagger}+a_{1}\right)-\left(\beta\gamma+\alpha\delta\right)\left(a_{1}^{\dagger}a_{2}^{\dagger}+a_{1}a_{2}\right)-a_{2}^{\dagger}a_{2}\right]\colon$ $\displaystyle=$ $\displaystyle-\frac{\alpha\gamma}{\beta\delta}\int\frac{d^{2}\sigma d^{2}\tau}{\pi^{2}}\colon\exp\left\\{\frac{\alpha}{\delta}\left[\sigma+\delta\left(a_{1}-a_{2}^{\dagger}\right)\right]\left[\sigma^{\ast}+\delta\left(a_{1}^{\dagger}-a_{2}\right)\right]\right.$ $\displaystyle\left.-\frac{\gamma}{\beta}\left[\tau-\beta\left(a_{2}^{\dagger}+a_{1}\right)\right]\left[\tau^{\ast}-\beta\left(a_{1}^{\dagger}+a_{2}\right)\right]\right\\}\colon$ $\displaystyle=$ $\displaystyle\colon\exp\left[-\left(a_{1}^{\dagger}a_{1}+a_{2}^{\dagger}a_{2}\right)\left(\alpha\delta-\beta\gamma+1\right)\right]\colon=1,$ where we have used the integral formula [17] $\int\frac{d^{2}\beta}{\pi}\exp\left[\varsigma\left\|\beta\right\|^{2}+\xi\beta+\eta\beta^{\ast}\right]=-\frac{1}{\varsigma}\exp\left[-\frac{\xi\eta}{\varsigma}\right],\text{ Re}\varsigma<0.$ (26) Thus $\left\|\Gamma\right\rangle_{e}$ is capable of making up a new quantum mechanical representation. ### 3.2 The non-orthonormal property of $\left\|\Gamma\right\rangle_{e}$ Noticing the overlap relation ${}_{e}\left\langle\Gamma\right\|\left.z_{1},z_{2}\right\rangle$ $\displaystyle=$ $\displaystyle 2\sqrt{-\alpha\beta\gamma\delta}\exp\left[-\frac{\left\|z_{1}\right\|^{2}}{2}+\frac{\alpha\left\|\sigma\right\|^{2}}{2\delta}-\frac{\gamma\left\|\tau\right\|^{2}}{2\beta}+\left(\alpha\sigma^{\ast}+\gamma\tau^{\ast}\right)z_{1}\right]$ (27) $\displaystyle\times\exp\left[-\frac{\left\|z_{2}\right\|^{2}}{2}+\left(\gamma\tau-\alpha\sigma\right)z_{2}-\left(\beta\gamma+\alpha\delta\right)z_{1}z_{2}\right],$ where $\left\|z\right\rangle=\exp\left(-\left\|z\right\|^{2}/2+za^{\dagger}\right)\left\|0\right\rangle$ is the coherent state [18, 19] and using the over-completeness relation of coherent states $\int\frac{d^{2}z_{1}d^{2}z_{2}}{\pi^{2}}\left\|z_{1},z_{2}\right\rangle\left\langle z_{1},z_{2}\right\|=1,$ we can derive the inner-product ${}_{e}\left\langle\Gamma\right\|\left.\Gamma^{\prime}\right\rangle_{e}$, ($\left\|\Gamma^{\prime}\right\rangle_{e}$ has the same $\beta,\gamma,\alpha$ and $\delta$ with $\left\|\Gamma\right\rangle_{e}$), ${}_{e}\left\langle\Gamma\right\|\left.\Gamma^{\prime}\right\rangle_{e}$ $\displaystyle=$ $\displaystyle\int\frac{d^{2}z_{1}d^{2}z_{2}}{\pi^{2}}\left.{}_{e}\left\langle\Gamma\right.\left\|z_{1},z_{2}\right\rangle\right.\left\langle z_{1},z_{2}\right\|\left.\Gamma^{\prime}\right\rangle_{e}$ (28) $\displaystyle=$ $\displaystyle-4\alpha\beta\gamma\delta\int\frac{d^{2}z_{1}d^{2}z_{2}}{\pi^{2}}\exp\left[-\left\|z_{1}\right\|^{2}+\left(\alpha\sigma^{\ast}+\gamma\tau^{\ast}\right)z_{1}+\left(\alpha\sigma^{\prime}+\gamma\tau^{\prime}\right)z_{1}^{\ast}\right]$ $\displaystyle-\left\|z_{2}\right\|^{2}+\left(\gamma\tau-\alpha\sigma\right)z_{2}+\left(\gamma\tau^{\prime\ast}-\alpha\sigma^{\prime\ast}\right)z_{2}^{\ast}-\left(\beta\gamma+\alpha\delta\right)z_{1}z_{2}$ $\displaystyle\left.+\frac{\alpha}{2\delta}\left(\left\|\sigma\right\|^{2}+\left\|\sigma^{\prime}\right\|^{2}\right)-\frac{\gamma}{2\beta}\left(\left\|\tau\right\|^{2}+\left\|\tau^{\prime}\right\|^{2}\right)-\left(\beta\gamma+\alpha\delta\right)z_{1}^{\ast}z_{2}^{\ast}\right].$ With the aid of the integral formula Eq.(26), we perform the integral over $d^{2}z_{1}d^{2}z_{2}$ in Eq.(28) and finally obtain ${}_{e}\left\langle\Gamma\right\|\left.\Gamma^{\prime}\right\rangle_{e}$ $\displaystyle=$ $\displaystyle\exp\left[\frac{\alpha}{4\beta\gamma\delta}\left\|\sigma-\sigma^{\prime}\right\|^{2}-\frac{1}{4\beta\delta}\left(\tau^{\prime}\sigma^{\ast}-\sigma\tau^{\prime\ast}+\sigma^{\prime}\tau^{\ast}-\tau\sigma^{\prime\ast}\right)\right.$ (29) $\displaystyle\left.+\frac{\gamma}{4\alpha\beta\delta}\left\|\tau-\tau^{\prime}\right\|^{2}-\frac{\left(\beta\gamma+\alpha\delta\right)}{4\beta\delta}\left(\tau^{\prime}\sigma^{\prime\ast}-\sigma^{\prime}\tau^{\prime\ast}+\sigma\tau^{\ast}-\tau\sigma^{\ast}\right)\right].$ From Eq.(29) one can see that ${}_{e}\left\langle\Gamma\right\|\left.\Gamma^{\prime}\right\rangle_{e}$ is non-orthogonal, only when $\sigma=\sigma^{\prime}$ and $\tau=\tau^{\prime}$, ${}_{e}\left\langle\Gamma\right\|\left.\Gamma\right\rangle_{e}=1.$ ## 4 The Weyl ordered form of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ For a density operator $\rho$ of bipartite system, we can convert it into its Weyl ordered form [3, 4, 20] by using the formula $\rho=4\int\frac{d^{2}z_{1}d^{2}z_{2}}{\pi^{2}}\genfrac{}{}{0.0pt}{}{:}{:}\left\langle- z_{1},-z_{2}\right\|\rho\left\|z_{1},z_{2}\right\rangle\exp\left[2\sum_{i=1}^{2}\left(a_{i}^{\dagger}a_{i}+a_{i}z_{i}^{\ast}-z_{i}a_{i}^{\dagger}\right)\right]\genfrac{}{}{0.0pt}{}{:}{:},$ (30) where the symbol $\genfrac{}{}{0.0pt}{}{:}{:}\genfrac{}{}{0.0pt}{}{:}{:}$ denotes the Weyl ordering, $\left\|z_{i}\right\rangle$ is the coherent state, $\left\langle- z_{i}\right\|\left.z_{i}\right\rangle=\exp\\{-2\left\|z_{i}\right\|^{2}\\}$. Note that the order of Bose operators $a_{i}$ and $a_{i}^{\dagger}$ within a Weyl ordered product can be permuted. That is to say, even though $\left[a,a^{\dagger}\right]=1$, we can have $\genfrac{}{}{0.0pt}{}{:}{:}aa^{\dagger}\genfrac{}{}{0.0pt}{}{:}{:}=\genfrac{}{}{0.0pt}{}{:}{:}a^{\dagger}a\genfrac{}{}{0.0pt}{}{:}{:}.$ Substituting Eq.(17) into Eq.(30) and performing the integration by virtue of the technique of integration within a Weyl ordered product (IWWOP) of operators [21], we finally obtain $\displaystyle\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ $\displaystyle=$ $\displaystyle-16\alpha\beta\gamma\delta\int\frac{d^{2}z_{1}d^{2}z_{2}}{\pi^{2}}\genfrac{}{}{0.0pt}{}{:}{:}\exp\left[-\left\|z_{1}\right\|^{2}+\left(\sigma^{\ast}\alpha+\tau^{\ast}\gamma-2a_{1}^{\dagger}\right)z_{1}+\left(2a_{1}-\sigma\alpha-\tau\gamma\right)z_{1}^{\ast}\right.$ (31) $\displaystyle-\left\|z_{2}\right\|^{2}+\left(\tau\gamma-\sigma\alpha-2a_{2}^{\dagger}\right)z_{2}+\left(2a_{2}-\tau^{\ast}\gamma+\sigma^{\ast}\alpha\right)z_{2}^{\ast}$ $\displaystyle\left.-\left(\beta\gamma+\alpha\delta\right)\left(z_{1}^{\ast}z_{2}^{\ast}+z_{1}z_{2}\right)+\frac{\alpha\left\|\sigma\right\|^{2}}{\delta}-\frac{\gamma\left\|\tau\right\|^{2}}{\beta}+2a_{1}^{\dagger}a_{1}+2a_{2}^{\dagger}a_{2}\right]\genfrac{}{}{0.0pt}{}{:}{:}$ $\displaystyle=$ $\displaystyle 4\genfrac{}{}{0.0pt}{}{:}{:}\exp\left\\{\frac{\alpha\delta}{\beta\gamma}\left(\frac{\sigma}{\delta}+\left(a_{1}-a_{2}^{\dagger}\right)\right)\left(\frac{\sigma^{\ast}}{\delta}+\left(a_{1}^{\dagger}-a_{2}\right)\right)\right.$ $\displaystyle\left.+\frac{\gamma\beta}{\alpha\delta}\left(\frac{\tau}{\beta}-\left(a_{1}+a_{2}^{\dagger}\right)\right)\left(\frac{\tau^{\ast}}{\beta}-\left(a_{2}+a_{1}^{\dagger}\right)\right)\right\\}\genfrac{}{}{0.0pt}{}{:}{:},$ or $\displaystyle\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ $\displaystyle=$ $\displaystyle 4\genfrac{}{}{0.0pt}{}{:}{:}\exp\left\\{\frac{\alpha\delta}{\beta\gamma}\left[\left(\frac{\sigma_{1}}{\delta}+\frac{Q_{1}-Q_{2}}{\sqrt{2}}\right)^{2}+\left(\frac{\sigma_{2}}{\delta}+\frac{P_{1}+P_{2}}{\sqrt{2}}\right)^{2}\right]\right.$ (32) $\displaystyle\left.+\frac{\beta\gamma}{\alpha\delta}\left[\left(\frac{\tau_{1}}{\beta}-\frac{Q_{1}+Q_{2}}{\sqrt{2}}\right)^{2}+\left(\frac{\tau_{2}}{\beta}-\frac{P_{1}-P_{2}}{\sqrt{2}}\right)^{2}\right]\right\\}\genfrac{}{}{0.0pt}{}{:}{:},$ which is the Weyl ordered form of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|.$ Noting the difference between Eq.(31) and Eq.(25), they are in different operator ordering. The merit of Weyl ordering lies in the Weyl ordered operators’ invariance under similar transformations, which was proved in Ref.[22]. In addition, it is very convenient for us to obtain the marginal distributions of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ (see the next section). In Ref. [23] we have derived the Weyl ordering form of two-mode Wigner operator $\Delta_{w}\left(\rho;\varsigma\right)$ $\Delta_{w}\left(\rho;\varsigma\right)=\genfrac{}{}{0.0pt}{}{:}{:}\delta\left(a_{1}-a_{2}^{\dagger}-\rho\right)\delta\left(a_{1}^{\dagger}-a_{2}-\rho^{\ast}\right)\delta\left(a_{1}+a_{2}^{\dagger}-\varsigma\right)\delta\left(a_{1}^{\dagger}+a_{2}-\varsigma^{\ast}\right)\genfrac{}{}{0.0pt}{}{:}{:}.$ (33) Eq.(33) indicates that the Weyl quantization scheme, for bipartite entangled operator, is to take the following correspondence, $\rho\rightarrow\left(a_{1}-a_{2}^{\dagger}\right),\text{ }\varsigma\rightarrow\left(a_{1}+a_{2}^{\dagger}\right),$ (34) then the form of Eq.(31) indicates that the classical Weyl function corresponding to $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ is $4\exp\left[\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}+\frac{\gamma\beta}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right]\equiv h\left(\rho;\varsigma\right).$ (35) Thus the Weyl quantization rule in this case is embodied as $\displaystyle\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ $\displaystyle=$ $\displaystyle 4\int d^{2}\rho d^{2}\varsigma\genfrac{}{}{0.0pt}{}{:}{:}\delta\left(a_{1}-a_{2}^{\dagger}-\rho\right)\delta\left(a_{1}^{\dagger}-a_{2}-\rho^{\ast}\right)\delta\left(a_{1}+a_{2}^{\dagger}-\varsigma\right)$ (36) $\displaystyle\times\delta\left(a_{1}^{\dagger}+a_{2}-\varsigma^{\ast}\right)\genfrac{}{}{0.0pt}{}{:}{:}\exp\left[\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}+\frac{\gamma\beta}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right]$ $\displaystyle=$ $\displaystyle 4\int d^{2}\rho d^{2}\varsigma\Delta_{w}\left(\rho;\varsigma\right)\exp\left[\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}+\frac{\gamma\beta}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right].$ Using the IWOP technique and Eq.(24), in [13] we have shown that the normally ordered form of $\Delta_{w}\left(\rho;\varsigma\right)$ is $\Delta_{w}\left(\rho;\varsigma\right)=\frac{1}{\pi^{2}}\colon\exp\left[-\left(a_{1}-a_{2}^{\dagger}-\rho\right)\left(a_{1}^{\dagger}-a_{2}-\rho^{\ast}\right)-\left(a_{1}+a_{2}^{\dagger}-\varsigma\right)\left(a_{1}^{\dagger}+a_{2}-\varsigma^{\ast}\right)\right]\colon.$ (37) Substituting Eq.(37) into Eq.(36) yields $\displaystyle\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ $\displaystyle=$ $\displaystyle 4\int\frac{d^{2}\rho d^{2}\varsigma}{\pi^{2}}\colon\exp\left\\{-\left(a_{1}-a_{2}^{\dagger}-\rho\right)\left(a_{1}^{\dagger}-a_{2}-\rho^{\ast}\right)+\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}\right.$ (38) $\displaystyle\left.-\left(a_{1}+a_{2}^{\dagger}-\varsigma\right)\left(a_{1}^{\dagger}+a_{2}-\varsigma^{\ast}\right)+\frac{\gamma\beta}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right\\}\colon$ $\displaystyle=$ $\displaystyle-4\alpha\beta\gamma\delta\colon\exp\left\\{\frac{\alpha}{\delta}\left[\sigma+\delta\left(a_{1}-a_{2}^{\dagger}\right)\right]\left[\sigma^{\ast}+\delta\left(a_{1}^{\dagger}-a_{2}\right)\right]\right.$ $\displaystyle\left.-\frac{\gamma}{\beta}\left[\tau-\beta\left(a_{2}^{\dagger}+a_{1}\right)\right]\left[\tau^{\ast}-\beta\left(a_{1}^{\dagger}+a_{2}\right)\right]\right\\}\colon,$ which confirms Eq.(25). In particular, when $\beta=-\delta=1,$ and $\alpha=\frac{\kappa}{1+\kappa},$ $\gamma=\frac{1}{1+\kappa},$ Eq.(36) becomes $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|\rightarrow 4\int d^{2}\rho d^{2}\varsigma\Delta_{w}\left(\rho,\varsigma\right)\exp\left[\allowbreak-\kappa\left\|\rho-\sigma\right\|^{2}-\frac{1}{\kappa}\left\|\varsigma-\tau\right\|^{2}\right],$ (39) which is the generalization of single-mode Husimi operator [24, 25]. ## 5 Marginal distributions of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ As mentioned above, based on the Weyl ordered form Eq.(32) it is very convenient for us to obtain the marginal distributions of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ , $\int_{-\infty}^{\infty}\frac{d^{2}\sigma}{\pi}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|=-\frac{4\beta\gamma\delta}{\alpha}\genfrac{}{}{0.0pt}{}{:}{:}\exp\left\\{\frac{\beta\gamma}{\alpha\delta}\left[\left(\frac{\tau_{1}}{\beta}-\frac{Q_{1}+Q_{2}}{\sqrt{2}}\right)^{2}+\left(\frac{\tau_{2}}{\beta}-\frac{P_{1}-P_{2}}{\sqrt{2}}\right)^{2}\right]\right\\}\genfrac{}{}{0.0pt}{}{:}{:}.$ (40) Noting $\left[Q_{1}+Q_{2},P_{1}-P_{2}\right]=0,$ there is no operator ordering problem involved in Eq.(40), so the symbol $\genfrac{}{}{0.0pt}{}{:}{:}\genfrac{}{}{0.0pt}{}{:}{:}$ in Eq.(40) can be neglected, $\int_{-\infty}^{\infty}\frac{d^{2}\sigma}{\pi}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|=-\frac{4\beta\gamma\delta}{\alpha}\exp\left\\{\frac{\beta\gamma}{\alpha\delta}\left[\left(\frac{\tau_{1}}{\beta}-\frac{Q_{1}+Q_{2}}{\sqrt{2}}\right)^{2}+\left(\frac{\tau_{2}}{\beta}-\frac{P_{1}-P_{2}}{\sqrt{2}}\right)^{2}\right]\right\\}.$ (41) The completeness relation of $\|\xi\rangle$ expressed in Eq.(10) is $\displaystyle\int\frac{d^{2}\xi}{\pi}\left\|\xi\right\rangle\left\langle\xi\right\|$ $\displaystyle=$ $\displaystyle 1,\text{ }d^{2}\xi=d\xi_{1}d\xi_{2},$ (42) $\displaystyle\left\langle\xi^{\prime}\right.\left\|\xi\right\rangle$ $\displaystyle=$ $\displaystyle\pi\delta\left(\xi^{\prime}-\xi\right)\delta\left(\xi^{\prime\ast}-\xi^{\ast}\right),$ (43) we see that the marginal distribution of function $\|_{e}\left\langle\Gamma\right.\left\|\Psi\right\rangle\|^{2}$ in “$\xi$-direction” is given by $\displaystyle\left\langle\Psi\right\|\int_{-\infty}^{\infty}\frac{d^{2}\sigma}{\pi}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right.\left\|\Psi\right\rangle$ $\displaystyle=$ $\displaystyle\left\langle\Psi\right\|\int\frac{d^{2}\xi}{\pi}\left\|\xi\right\rangle\left\langle\xi\right\|\int_{-\infty}^{\infty}\frac{d^{2}\sigma}{\pi}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right.\int\frac{d^{2}\xi^{\prime}}{\pi}\left\|\xi^{\prime}\right\rangle\left\langle\xi^{\prime}\right.\left\|\Psi\right\rangle$ (44) $\displaystyle=$ $\displaystyle-\frac{4\beta\gamma\delta}{\alpha}\int_{-\infty}^{\infty}\frac{d^{2}\xi}{\pi}\left\|\Psi\left(\xi\right)\right\|^{2}\exp\left[\frac{\beta\gamma}{\alpha\delta}\left\|\frac{\tau}{\beta}-\xi\right\|^{2}\right],$ which is a Gaussian-broadened version of quantal distribution $\left\|\Psi\left(\xi\right)\right\|^{2}$ (measuring two particles’ relative momentum and center-of-mass coordinate). Similarly, we can obtain another marginal distribution by performing the integral $d^{2}\tau$ over $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|,$ $\int_{-\infty}^{\infty}\frac{d^{2}\tau}{\pi}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|=-\frac{4\alpha\beta\delta}{\gamma}\exp\left\\{\frac{\alpha\delta}{\beta\gamma}\left[\left(\frac{\sigma_{1}}{\delta}+\frac{Q_{1}-Q_{2}}{\sqrt{2}}\right)^{2}+\left(\frac{\sigma_{2}}{\delta}+\frac{P_{1}+P_{2}}{\sqrt{2}}\right)^{2}\right]\right\\}.$ (45) By using the completeness relation of $\|\eta\rangle$, $\displaystyle\int\frac{d^{2}\eta}{\pi}\|\eta\rangle\langle\eta\|$ $\displaystyle=$ $\displaystyle 1,\ \ d^{2}\eta=d\eta_{1}d\eta_{2},$ (46) $\displaystyle\left\langle\eta^{\prime}\right.\left\|\eta\right\rangle$ $\displaystyle=$ $\displaystyle\pi\delta\left(\eta^{\prime}-\eta\xi\right)\delta\left(\eta^{\prime\ast}-\eta^{\ast}\right),$ (47) we see that the other marginal distribution of $\|_{e}\left\langle\Gamma\right.\left\|\Psi\right\rangle\|^{2}$ in “$\eta$-direction” is $\left\langle\Psi\right\|\int_{-\infty}^{\infty}\frac{d^{2}\tau}{\pi}\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right.\left\|\Psi\right\rangle=-\frac{4\alpha\beta\delta}{\gamma}\int_{-\infty}^{\infty}\frac{d^{2}\eta}{\pi}\left\|\Psi\left(\eta\right)\right\|^{2}\exp\left\\{\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\eta\right\|^{2}\right\\},$ (48) a Gaussian-broadened version of quantal distribution $\left\|\Psi\left(\eta\right)\right\|^{2}$ (measuring two particles’ relative coordinate and total momentum), Eqs.(44) and (48) describe the relationship between wave functions in the ${}_{e}\left\langle\Gamma\right\|$ representation and those in EPR entangled state $\left\|\xi\right\rangle$ ($\left\|\eta\right\rangle$) representation, respectively. Note that $\|\eta\rangle$ and $\left\|\xi\right\rangle$ are related to each other by $\left\langle\xi\right.\left\|\eta\right\rangle=\frac{1}{2}\exp\left(\frac{\xi^{\ast}\eta-\xi\eta^{\ast}}{2}\right).$ (49) ## 6 Minimum uncertainty relation for $\left\|\Gamma\right\rangle_{e}$ From the marginal distributions of $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ we have seen that its phase space representation involves both the center-of mass (relative) coordinate and the relative (total) momentum. In order to see clearly how the state $\left\|\Gamma\right\rangle_{e}$ obeys uncertainty relation, we introduce two pairs of quadrature phase amplitudes for two-mode field: $Q_{\pm}\equiv\frac{Q_{1}\pm Q_{2}}{\sqrt{2}},\text{ }P_{\pm}\equiv\frac{P_{1}\pm P_{2}}{\sqrt{2}},\text{ }\left[Q_{\pm},P_{\pm}\right]=i.$ (50) In similar to deriving Eq.(29), using Eqs. (5), (10) and (17), we calculate the overlap between $\left\langle\eta\right\|$ and $\left\|\Gamma\right\rangle_{e},$ $\left\langle\eta\right.\left\|\Gamma\right\rangle_{e}=\sqrt{-\frac{\alpha\delta}{\beta\gamma}}\exp\left\\{\frac{\alpha\delta}{2\beta\gamma}\left\|\frac{\sigma}{\delta}+\eta\right\|^{2}+\frac{1}{2\beta}\left[\tau\left(\eta^{\ast}-\alpha\sigma^{\ast}\right)-\tau^{\ast}\left(\eta-\alpha\sigma\right)\right]\right\\},$ (51) and the overlap between $\left\langle\xi\right\|$ and $\left\|\Gamma\right\rangle_{e},$ $\left\langle\xi\right.\left\|\Gamma\right\rangle_{e}=\sqrt{-\frac{\beta\gamma}{\alpha\delta}}\exp\left\\{\allowbreak\frac{\beta\gamma}{2\alpha\delta}\left\|\frac{\tau}{\beta}-\xi\right\|^{2}-\frac{1}{2\delta}\left[\sigma\left(\xi^{\ast}-\gamma\tau^{\ast}\right)-\sigma^{\ast}\left(\xi-\gamma\tau\right)\right]\right\\}.$ (52) Then employing the completeness relation of $\left\|\eta\right\rangle$ and Eq.(51), we evaluate $\displaystyle\left\langle Q_{-}\right\rangle$ $\displaystyle=$ $\displaystyle\int\frac{d^{2}\eta}{\pi}\eta_{1}\left\|\left\langle\eta\right.\left\|\Gamma\right\rangle_{e}\right\|^{2}=-\frac{\sigma_{1}}{\delta},$ $\displaystyle\left\langle Q_{-}^{2}\right\rangle$ $\displaystyle=$ $\displaystyle\int\frac{d^{2}\eta}{\pi}\eta_{1}^{2}\left\|\left\langle\eta\right.\left\|\Gamma\right\rangle_{e}\right\|^{2}=\frac{\sigma_{1}^{2}}{\delta^{2}}-\frac{\beta\gamma}{2\alpha\delta},$ (53) and $\left\langle P_{-}\right\rangle=\frac{\tau_{2}}{\beta},\text{ }\left\langle P_{-}^{2}\right\rangle=\frac{\tau_{2}^{2}}{\beta^{2}}-\frac{\alpha\delta}{2\beta\gamma}.$ (54) It then follows $\displaystyle\left\langle\Delta Q_{-}^{2}\right\rangle$ $\displaystyle=$ $\displaystyle\left\langle Q_{-}^{2}\right\rangle-\left\langle Q_{-}\right\rangle^{2}=-\frac{\beta\gamma}{2\alpha\delta},$ $\displaystyle\left\langle\Delta P_{-}^{2}\right\rangle$ $\displaystyle=$ $\displaystyle\left\langle P_{-}^{2}\right\rangle-\left\langle P_{-}\right\rangle^{2}=-\frac{\alpha\delta}{2\beta\gamma},$ (55) and $\sqrt{\left\langle\Delta Q_{-}^{2}\right\rangle\left\langle\Delta P_{-}^{2}\right\rangle}=\frac{1}{2}.$ (56) In a similar way, using (52) we can derive $\displaystyle\left\langle Q_{+}\right\rangle$ $\displaystyle=$ $\displaystyle\frac{\sigma_{2}}{\delta},\text{ }\left\langle Q_{+}^{2}\right\rangle=\frac{\sigma_{2}^{2}}{\delta^{2}}-\frac{\beta\gamma}{2\alpha\delta},$ $\displaystyle\left\langle P_{+}\right\rangle$ $\displaystyle=$ $\displaystyle\frac{\tau_{1}}{\beta},\text{ }\left\langle P_{+}^{2}\right\rangle=\frac{\tau_{1}^{2}}{\beta^{2}}-\frac{\alpha\delta}{2\beta\gamma},$ (57) which also leads to $\sqrt{\left\langle\Delta Q_{+}^{2}\right\rangle\left\langle\Delta P_{+}^{2}\right\rangle}=\frac{1}{2}.$ (58) Eqs. (55)-(58) show that $\left\|\Gamma\right\rangle_{e}$ is a minimum uncertainty state for the two pairs of quadrature operators. ## 7 The Wigner function of $\left\|\Gamma\right\rangle_{e}$ For a bipartite system, the two-mode Wigner operator in entangled state $\left\|\eta\right\rangle$ representation is expressed in Eq.(6), so the Wigner function $W\left(\rho,\varsigma\right)$ of $\left\|\Gamma\right\rangle_{e}$ is given by $W\left(\rho,\varsigma\right)=Tr\left[\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|\Delta_{w}\left(\rho,\varsigma\right)\right]=\int\frac{d^{2}\eta}{\pi^{3}}\left.{}_{e}\left\langle\Gamma\right.\left\|\rho-\eta\right\rangle\right.\left\langle\rho+\eta\right\|\left.\Gamma\right\rangle_{e}e^{\eta\varsigma^{\ast}-\varsigma\eta^{\ast}}.$ (59) Substituting Eq.(51) into Eq.(59) and using the formula in Eq.(26), we obtain $\displaystyle W\left(\rho,\varsigma\right)$ $\displaystyle=$ $\displaystyle\int\frac{d^{2}\eta}{\pi^{3}}\left.{}_{e}\left\langle\Gamma\right.\left\|\rho-\eta\right\rangle\right.\left\langle\rho+\eta\right\|\left.\Gamma\right\rangle_{e}e^{\eta\varsigma^{\ast}-\varsigma\eta^{\ast}}$ (60) $\displaystyle=$ $\displaystyle-\frac{\alpha\delta}{\beta\gamma}\int\frac{d^{2}\eta}{\pi^{3}}\exp\left\\{\frac{\alpha\delta}{\beta\gamma}\left\|\eta\right\|^{2}+\left(\varsigma^{\ast}-\frac{\tau^{\ast}}{\beta}\right)\eta+\left(\frac{\tau}{\beta}-\varsigma\right)\eta^{\ast}\right.$ $\displaystyle\left.+\frac{\alpha}{2\beta\gamma\delta}\left(\allowbreak 2\left\|\sigma\right\|^{2}+2\delta^{2}\allowbreak\left\|\rho\right\|^{2}+2\delta\left(\sigma\rho^{\ast}+\rho\sigma^{\ast}\right)\allowbreak\right)\right\\}$ $\displaystyle=$ $\displaystyle\frac{1}{\pi^{2}}\exp\left[\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}+\frac{\gamma\beta}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right].$ From Eq.(60) one can see that the Wigner function of $\left\|\Gamma\right\rangle_{e}$ possesses the well-behaved feature in the sense that its marginal distribution in “$\sigma$-direction” is a general Gaussian form $\exp\left\\{\allowbreak\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}\right\\}$, while its marginal distribution in “$\tau$-direction” is $\exp\left\\{\frac{\beta\gamma}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right\\}.$ When $\beta\gamma+\alpha\delta=0\ $and $\beta=-\delta=1,$ Eq.(60) reduces to $W\left(\rho,\varsigma\right)=\frac{1}{\pi^{2}}\exp\left(-\left\|\sigma+\rho\right\|^{2}-\left\|\tau-\varsigma\right\|^{2}\right),$ (61) which is just the Wigner function of two-mode canonical coherent state. Before ending this work, we mention that in 1990 Torres-Vega and Frederick introduced the state $\left\|\Gamma\right\rangle$ which satisfies [30] $\displaystyle\left\langle\Gamma\right\|Q_{1}$ $\displaystyle=$ $\displaystyle\left(\alpha q+i\beta\frac{\partial}{\partial p}\right)\left\langle\Gamma\right\|,$ (62) $\displaystyle\left\langle\Gamma\right\|P_{1}$ $\displaystyle=$ $\displaystyle\left(\gamma p+i\delta\frac{\partial}{\partial q}\right)\left\langle\Gamma\right\|.$ (63) these two equations well satisfy the correspondence between the classical and quantum Liouville equations in single-mode case and have many applications in chemical physics and quantum chemistry [30]-[34]. Recently $\left\|\Gamma\right\rangle$ has been identified as a coherent squeezed state [35] $\left\|\Gamma\right\rangle\equiv\left(2\sqrt{-\alpha\beta\gamma\delta}\right)^{1/2}\exp\left[\frac{\alpha q^{2}}{2\delta}-\frac{\gamma p^{2}}{2\beta}+\sqrt{2}\left(\alpha q+i\gamma p\right)a_{1}^{\dagger}+\frac{\beta\gamma+\alpha\delta}{2}a_{1}^{\dagger 2}\right]\left\|0\right\rangle_{1}.$ (64) The $\left\|\Gamma\right\rangle_{e}$ in this paper is the non-trivial generalization of $\left\|\Gamma\right\rangle$. In summary, based on the conception of quantum entanglement of Einstein- Podolsky-Rosen, we have introduced the entangled state $\left\|\Gamma\right\rangle_{e}$ for constructing generalized phase space representation, which possesses well-behaved properties. The set of $\left\|\Gamma\right\rangle_{e}$ make up a complete and non-orthogonal representation, so it may have new applications, for examples: 1) It can be chosen as a good representation for solving dynamic problems for some Hamiltonians which include explicitly the function of quadrature operators $Q_{\pm},$ and/or $P_{\pm};$ 2) $\left\|\Gamma\right\rangle_{ee}\left\langle\Gamma\right\|$ may be considered as a generalized Wigner operator, since from Eq. (36) we see that it is expressed as smoothing out the usual Wigner operator by averaging over a “course graining” function $\exp\left[\frac{\alpha\delta}{\beta\gamma}\left\|\frac{\sigma}{\delta}+\rho\right\|^{2}+\frac{\gamma\beta}{\alpha\delta}\left\|\frac{\tau}{\beta}-\varsigma\right\|^{2}\right],$ and the corresponding generalized Wigner function is positive definite. 3) The representation $\left\|\Gamma\right\rangle_{e}$ may be used to analyze entanglement degree for some entangled states. 4) The $\left\|\Gamma\right\rangle_{e}$ state can be taken as a quantum channel for quantum teleportation, such channel may make the teleportation fidelity flexible, since it involves adjustable parameters $\alpha,\beta,\gamma,\delta.$ We hope these applications could be studied in the near future. ACKNOWLEDGEMENTS The work is supported by the National Natural Science Foundation of China under grant no. 10775097. APPENDIX Derivation of Eq.(17) Here we show how to derive Eq.(17). In the $\|\eta\rangle$ representation we have $\displaystyle\frac{Q_{1}+Q_{2}}{\sqrt{2}}\|\eta\rangle$ $\displaystyle=-i\frac{\partial}{\partial\eta_{2}}\|\eta\rangle=\left(\frac{\partial}{\partial\eta}-\frac{\partial}{\partial\eta^{\ast}}\right)\|\eta\rangle,$ (A1) $\displaystyle\frac{P_{1}-P_{2}}{\sqrt{2}}\|\eta\rangle$ $\displaystyle=i\frac{\partial}{\partial\eta_{1}}\|\eta\rangle=i\left(\frac{\partial}{\partial\eta}+\frac{\partial}{\partial\eta^{\ast}}\right)\|\eta\rangle,$ (A2) so according to the requirement in Eqs. (14)-(16), we see $\displaystyle\left[\frac{\gamma}{2}\left(\tau+\tau^{\ast}\right)+\delta\left(\frac{\partial}{\partial\sigma}-\frac{\partial}{\partial\sigma^{\ast}}\right)\right]\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=\left(\frac{\partial}{\partial\eta}-\frac{\partial}{\partial\eta^{\ast}}\right)\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.,$ (A3) $\displaystyle\left[\frac{\alpha}{2i}\left(\sigma-\sigma^{\ast}\right)-i\beta\left(\frac{\partial}{\partial\tau}+\frac{\partial}{\partial\tau^{\ast}}\right)\right]\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=\frac{\eta-\eta^{\ast}}{2i}\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.,$ (A4) and $\displaystyle\left[\frac{\alpha}{2}\left(\sigma+\sigma^{\ast}\right)-\beta\left(\frac{\partial}{\partial\tau}-\frac{\partial}{\partial\tau^{\ast}}\right)\right]\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=\frac{\eta+\eta^{\ast}}{2}\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.,$ (A5) $\displaystyle\left[\frac{\gamma}{2i}\left(\tau-\tau^{\ast}\right)+i\delta\left(\frac{\partial}{\partial\sigma}+\frac{\partial}{\partial\sigma^{\ast}}\right)\right]\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=i\left(\frac{\partial}{\partial\eta}+\frac{\partial}{\partial\eta^{\ast}}\right)\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.,$ (A6) Combining Eqs.(A3)-(A6) yields $\displaystyle\left(\alpha\sigma+2\beta\frac{\partial}{\partial\tau^{\ast}}\right)\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=\eta\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.,$ (A7) $\displaystyle\left(\alpha\sigma^{\ast}-2\beta\frac{\partial}{\partial\tau}\right)\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=\eta^{\ast}\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.,$ (A8) $\displaystyle\left(\gamma\tau^{\ast}+2\delta\frac{\partial}{\partial\sigma}\right)\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=2\frac{\partial}{\partial\eta}\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ (A9) $\displaystyle\left(\gamma\tau-2\delta\frac{\partial}{\partial\sigma^{\ast}}\right)\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right.$ $\displaystyle=-2\frac{\partial}{\partial\eta^{\ast}}\left.{}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\right..$ (A10) The solution to Eqs.(A7)-(A10) is ${}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle=C\exp\left\\{\frac{\alpha\delta}{2\beta\gamma}\left\|\frac{\sigma}{\delta}+\eta\right\|^{2}+\frac{1}{2\beta}\left[\tau^{\ast}\left(\eta-\alpha\sigma\right)-\tau\left(\eta^{\ast}-\alpha\sigma^{\ast}\right)\right]\right\\},$ (A11) where $C$ is the normalization constant determined by ${}_{e}\left\langle\Gamma\right\|\left.\Gamma\right\rangle_{e}=1.$ Using the completeness relation of EPR entangled state (46) and the integral formula in Eq.(26), we obtain ${}_{e}\left\langle\Gamma\right\|$ $\displaystyle=\int\frac{d^{2}\eta}{\pi}_{e}\left\langle\Gamma\right\|\left.\eta\right\rangle\left\langle\eta\right\|$ $\displaystyle=C\left\langle 00\right\|\int\frac{d^{2}\eta}{\pi}\exp\left[-\frac{1}{2}\|\eta\|^{2}+\eta^{\ast}a_{1}-\eta a_{2}+a_{1}a_{2}\right]$ $\displaystyle\times\exp\left\\{\frac{\alpha}{2\beta\gamma\delta}\left\|\sigma+\delta\eta\right\|^{2}+\frac{1}{2\beta}\left[\tau^{\ast}\left(\eta-\alpha\sigma\right)-\tau\left(\eta^{\ast}-\alpha\sigma^{\ast}\right)\right]\right\\}$ $\displaystyle=\left\langle 00\right\|C\exp\left[\frac{\alpha\left\|\sigma\right\|^{2}}{2\delta}-\frac{\gamma\left\|\tau\right\|^{2}}{2\beta}+\left(\alpha\sigma^{\ast}+\gamma\tau^{\ast}\right)a_{1}+\left(\gamma\tau-\alpha\sigma\right)a_{2}-\left(\beta\gamma+\alpha\delta\right)a_{1}a_{2}\right],$ (A12) which is just Eq.(17) when $C$ is taken as $C=2\sqrt{-\alpha\beta\gamma\delta}.$ ## References * [1] Wigner E P 1932 Phys. Rev. 40 749 * [2] Feynman R P 1972 StatisticalMechanics (New York: Benjamin) * [3] Weyl H 1927 Z. Phys. 46 1 * [4] Weyl H 1953 The Classical Groups (Princeton University Press) * [5] Fan H Y, Lu H L and Fan Y 2006 Ann. Phys. 321 480; Fan H Y 2003 J. Opt. B: Quantum Semiclass. Opt. 5 R147 * [6] Wünsche A 1999 J. Opt. B: Quantum Semiclass. Opt. 1 R11 * [7] Walls D F and Milburn G 1994 Quantum Optics (Berlin:Springer) * [8] Smithey D T, Beck M and Raymer M G 1993 Phys. Rev. Lett. 70 1244 * [9] Wu H J and Fan H Y 1997 Mod. Phys. Lett. B 11 544; Fan H Y 2001 J. Opt. B: Quantum Semiclass. Opt. 3 388; Fan H Y and Sun Z H 2000 Phys. Lett. A 272 219 * [10] Vogel K and Risken H 1989 Phys. Rev. A 40 2847; Tatarsky T 1983 Sov. Phys.–Usp. 26 311 * [11] Bouwmeester D et al 2000 The Physics of Quantum Information (Springer, Berlin); Nielsen M A and Chuang I L 2000 Quantum Computation and Quantum Information (Cambridge University Press). * [12] Einstein A, Podolsky B and Rosen N 1935 Phys. Rev. 47 777 * [13] Fan H Y 2002 Phys. Rev. A. 65 064102; Wu H J and Fan H Y 1997 Mod. Phys. Lett. B 11 549 * [14] Fan H Y and Klauder J R 1994 Phys. Rev. A. 47 704; Fan H Y and Chen B Z 1996 Phys. Rev. A. 53 2948; Fan H Y and Ye X 1995 Phys. Rev A. 51 3343 * [15] Bužek V 1990 J. Mod. Opt. 37 303; Loudon R and Knight P L 1987 J. Mod. Opt. 34 709; Fan H Y and Fan Y, 1996 Phys. Rev. A. 54 958; Dodonov V V 2002 J. Opt. B: Quant. Semiclass. Opt. 4 R1 * [16] Louisell W H 1973 Quantum Statistical Properties of Radiation (New York: Wiley) * [17] Puri R R 2000 Mathematical Methods of Quantum Optics (Springer-Verlag, Berlin), Appendix A. * [18] Glauber R J 1963 Phys. Rev. 130 2529; 1963 Phys. Rev. 131 2766 * [19] Klauder J R and Skargerstam B S 1985 Coherent States (World Scientific, Singapore) * [20] Fan H Y 1992 J. Phys. A, 25 3443 * [21] Fan H Y 2007 Ann. Phys. doi: 10.1016/j.aop.2007.08.009 * [22] Fan H Y 2008 Ann. Phys. 323 500 * [23] Fan H Y, Fan Y 2002 Int. J. Mod. Phys. A, 17 701; Fan H Y 2000 Mod. Phys. Lett. A 15 2297 * [24] Husimi K 1940 Proc. Phys. Math. Soc. Jpn. 22 264 * [25] Fan H Y and Yang Y L 2006. Phys. Lett. A 353 439 * [26] Wolfgang P. Schleich 2001 Quantum Optics in Phase Space (Wiley-VCH, Birlin); Dodonov V V, Man’ko V I 2003 Theory of Nonclassical States of Light (Taylor & Francis, New York) * [27] Agawal G S, Wolf E 1972 Phys. Rev. D 2 2161; Agawal G S, Wolf E 1972 Phys. Rev. D 2 2187; Agawal G S, Wolf E 1972 Phys. Rev. D 2 2206 * [28] Bužek V, Knight P L 1995 Prog. Opt. 34 1; Bužek V, Keitel C H, Knight P L 1995 Phys. Rev. A 51 2575 * [29] O’Connell R F, Wigner E P 1981 Phys. Lett. A 83 145; Hillery M, O’Connell R F, Scully M O and Wigner E P 1984 Phys. Rep. 106 121; Vogel K, Risken H 1989 Phys. Rev. A 40 2847 * [30] Go. Torres-Vega and Frederick J H 1993 J. Chem. Phys. 98 3103 * [31] Go. Torres-Vega and Frederick J H 1990 J. Chem. Phys. 93, 8862 * [32] Bopp F 1961 in Werner Heisenberg und die Physik unserer Zeit (Vieweg, Braunschweig) * [33] Jaffé C and Brumer P 1985 J. Chem. Phys. 82 2330; Jaffé C, Kanfer S and Brumer P 1985 Phys. Rev. Lett. 54 8 * [34] Davis M J 1988 J. Phys. Chem. 92 3124; Gray S K 1987 J. Chem. Phys. 87 2051; Klaus B. Møller, Thomas G. Jørgensen and Go. Torres-Vega 1997 J. Phys. Chem. 106 7228 * [35] Hu L Y, Fan H Y and Lu H L 2008 J. Chem. Phys. 128 054101	arxiv-papers	2008-08-15T14:11:09	2024-09-04T02:48:57.329907	{ "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/0808.2149" }
0808.2329	# Local Coarse-grained Approximation to Path Integral Monte Carlo Integration for Fermion Systems D. Y. Sun State Key Laboratory of Precision Spectroscopy and Department of Physics, East China Normal University, Shanghai 200062, China ###### Abstract An approximate treatment of exchange in finite-temperature path integral Monte Carlo simulations for fermions has been proposed. In this method, some of the fine details of density matrix due to permutations have been smoothed over or averaged out by using the coarse-grained approximation. The practical usefulness of the method is tested for interacting fermions in a three dimensional harmonic well. The results show that, the present method not only reduces the sign fluctuation of the density matrix, but also avoid the fermion system collapsing into boson system at low temperatures. The method is substantiated to be exact when applied to free particles. ###### pacs: 02.70.Ss, 31.15.xk, 02.70.-c ## I Introduction The path integral Monte Carlo method (PIMC) provides a nonperturbative, basis- set-independent, and fully correlated calculation for quantum many-body systems at both zero and finite temperature.Ceperley95 ; Foulkes01 ; sign However for many-fermion systems, PIMC suffers from uncontrollable errors arising from the notorious _sign problem_ ,sign ; Loh90 which limits the accuracy or stability of the method. The origin of the _sign problem_ comes from the fact that the density matrix can be positive or negative by even or odd permutations. At low temperatures, contributions from positive and negative parts of the density matrix almost perfectly cancel each other so that there is no hope of extracting any useful information. A few methods have been proposed to deal with the _sign problem_. For problems in the continuous space, there are fixed-node approximation,Anderson76 ; Ceperley92 released node methods,Ceperley80 exact cancelation methods with Green’s function sampling,Chen95 multilevel blocking algorithm,Mak98 ; Egger00 a hybrid path integral and basis set method,Chiles84 various pseudopotential approximations,Schnitker87 ; Landman87 ; Hall88 ; Kuki87 ; Coker87 ; Bartholomew85 ; Sprik85 ; Oh98 ; Miura01 the general method for replacing integration over pure states by integration over idempotent density matrices,Newman91 and method by introducing several images of the system,Lyubartsev global stationary phase approach.Moreira There are also a number of methods for lattice models.Zhang97 ; Zhang99 ; Helenius00 So far many efforts have been devoted to tackle this problem, however, it remains being the key bottleneck in using PIMC for many-fermion systems. In this paper, we use a methodology to reduce the rapid oscillation of integrand in the evaluation of high dimensional integrals. The idea is that, for the region in which the function is rapidly oscillating, the coarse- grained approximations are used to kill fluctuations. Applying this technique to PIMC, we found that, the sign fluctuations of density matrix can be reduced, thus the Metropolis MC integration algorithm converges efficiently. More importantly, after carrying out this approximation, the exchange determinant becomes a nonlocal form in imaginary time, thus the collapse behavior can be avoided (see below). The basic strategy can also be used in the evaluation of other integrals, where the integrands exhibit the rapidly oscillating characters. This paper is organized as follows. The methodology is described in Sec. II. The numerical tests are presented in Sec. III. Discussion and conclusion are given in Sec. IV. ## II Methodology To illustrate the coarse-grained approximation used in present paper, we first consider the following integral, $I=\int_{a}^{c}\int_{A}^{C}f(x,y)g(x,y)h(x,y)dxdy.$ (1) We assume that, except for _f(x,y)_ is a rapidly oscillating function for variable _y_ , the rest parts of integrands are well behavior (or slow varied) function of _x_ and _y_. Due to the rapid oscillation of _f(x,y)_ , it could cause the difficulty on evaluating the integral (Eq. 1) in MC simulation. To overcome the difficulty, our strategy is to make coarse-grained approximations to _f(x,y)_. To do this, we rewrite the above integral as, $I=\int_{a}^{c}\int_{A}^{C}F(x)g(x,y)h(x,y)dxdy,$ (2) with $F(x)=\frac{\int_{A}^{C}f(x,y^{\prime})g(x,y^{\prime})h(x,y^{\prime})dy^{\prime}}{\int_{A}^{C}g(x,y^{\prime})h(x,y^{\prime})dy^{\prime}}.$ (3) One can see that, _F(x)_ is a kind of coarse-grained functions, and the rapid oscillation of _f(x,y)_ is smoothed over. _F(x)_ also can be viewed as an average of _f(x,y)_ weighted by _g(x,y)__h(x,y)_. If _F(x)_ can be evaluated either exactly or approximately, the rapid fluctuation due to _f(x,y)_ could be reduced effectively. For real problems, _F(x)_ usually is hard to be evaluated exactly. However, it is possible to determine _F(x)_ under some reasonable approximations, as we have done in present paper. Considering a three-dimensional system consisting of _N_ spinless, indistinguishable quantum fermions, the standard PIMC is based on the following expansion of partition function: $Z=\frac{C}{N!}\lim_{M\rightarrow\infty}\int\prod_{i=1}^{N}\prod_{\nu=1}^{M}d\vec{r}_{i}^{(\nu)}detAexp(-\beta H)$ (4) with $H=\sum_{i=1}^{N}\frac{1}{2}m\omega_{M}^{2}L_{i}^{2}+\sum_{\nu=1}^{M}\frac{1}{M}V(\\{\vec{r}_{i}^{(\nu)}\\})$ where $\omega_{M}=\frac{\sqrt{M}}{\beta\hbar}$, $C=(\frac{mM}{2\pi\beta\hbar^{2}})^{\frac{3}{2}NM}$, $\vec{r}_{i}^{(M+1)}=\vec{r}_{i}^{(1)}$, _V_ is the potential energy, and the square _length_($L_{i}^{2}$) is defined as, $L_{i}^{2}=\sum_{\nu=1}^{M}(\vec{r}_{i}^{(\nu+1)}-\vec{r}_{i}^{(\nu)})^{2}$. The subscript _i_ refers to the particle number while the superscript $\nu$ refers to different slit of imaginary time. $\beta$, $m$ and $M$ are reciprocal temperature($1/k_{B}T$), mass of particles and total number of beads respectively. $\\{\vec{r}_{i}^{(\nu)}\\}$ refers to $(\vec{r}_{1}^{(\nu)},\vec{r}_{2}^{(\nu)},......,\vec{r}_{N-1}^{(\nu)},\vec{r}_{N}^{(\nu)})$.To make the expression compact, we have introduced a $N\times N$ matrix _A_ whose element reads $A_{ij}=exp(-\frac{1}{2}\beta m\omega^{2}_{M}((\vec{r}_{i}^{(1)}-\vec{r}_{j}^{(M)})^{2}-(\vec{r}_{i}^{(1)}-\vec{r}_{i}^{(M)})^{2})).$ (5) _detA_ is the determinant of the matrix _A_ , which accounts the contribution of permutations to the partition function. It is _detA_ , which can be positive or negative, that causes the so-called _sign problem_. Previous studies based on pseudopotential methods have shown that, direct use of Eq.4-like formula usually results in a fermion system _collapsing_ into a bosonic state at low temperature.Hall88 The physical reason comes from the fact that the matrix _A_ approaching to unit at low temperature. To prevent this undesirable behavior, people usually recast matrix _A_ in a nonlocal form as suggested by Hall,Hall88 or directly use a nonlocal pseudopotential as suggested by Miura and Okazaki.Miura01 Although these schemes do give a good solution, the computational cost also increases. From Eq. 5, one can see that, if $\vec{r}_{i}^{(M)}$ is close to $\vec{r}_{j}^{(M)}$, $A_{ij}$ will be a rapidly oscillating function of $\vec{r}_{i}^{(1)}$. Although, for N$>$2, it is difficult to prove the direct relation between the rapid oscillation of $A_{ij}$ and the sign problem, it is quite clear that the rapid oscillation of $A_{ij}$ directly results in the sign fluctuation for N=2. To smooth the rapid oscillation, the coarse-grained approximation is made for $A_{ij}$ by integration over $\vec{r}_{i}^{(1)}$ for all possible configurations with fixed $\vec{r}_{i}^{(M)}$, $\vec{r}_{j}^{(M)}$ and $L_{i}$. Now Eq. 4 is replaced by $Z=\frac{C}{N!}\lim_{M\rightarrow\infty}\int\prod_{i=1}^{N}\prod_{\nu=1}^{M}d\vec{r}_{i}^{(\nu)}det\Xi exp(-\beta H)$ (6) where the new $N\times N$ matrix $\Xi$ is the coarse-grained approximation of matrix _A_. According to the idea presented in Eq. 1, 2 and 3, the elements of $\Xi$ are $\Xi_{\alpha\gamma}=\frac{\int^{\prime}\prod_{\nu=1}^{M-1}d\vec{r}_{\alpha}^{(\nu)}A_{\alpha\gamma}exp(-\beta H)}{\int^{\prime}\prod_{\nu=1}^{M-1}d\vec{r}_{\alpha}^{(\nu)}exp(-\beta H)}$ (7) Since $A_{\alpha\gamma}$ is only relevant to $\vec{r}_{\alpha}^{(1)}$, $\vec{r}_{\alpha}^{(M)}$ and $\vec{r}_{\gamma}^{(M)}$, Eq. 7 can be rewritten as, $\Xi_{\alpha\gamma}=\frac{\int^{\prime}\prod_{\nu=1}^{M-1}d\vec{r}_{\alpha}^{(\nu)}A_{\alpha\gamma}exp(-\frac{1}{2}\beta m\omega_{M}^{2}L_{\alpha}^{2}-\beta U)}{\int^{\prime}\prod_{\nu=1}^{M-1}d\vec{r}_{\alpha}^{(\nu)}exp(-\frac{1}{2}\beta m\omega_{M}^{2}L_{\alpha}^{2}-\beta U)}$ (8) with $U=\sum_{\nu=1}^{M}\frac{1}{M}V(\\{\vec{r}_{i}^{(\nu)}\\}))$ $\int^{\prime}$ in Eq. 7 and 8 refers the integral under the constraint of fixed $\vec{r}_{\alpha}^{(M)}$, $\vec{r}_{\gamma}^{(M)}$ and $L_{\alpha}$. Since the kinetic energy relevant part($\frac{1}{2}\omega_{M}^{2}L_{\alpha}^{2}$) is a constant for fixed $L_{\alpha}$. The Eq. 8 can be further simplified as, $\Xi_{\alpha\gamma}=\frac{\int^{\prime}\prod_{\nu=1}^{M-1}d\vec{r}_{\alpha}^{(\nu)}A_{\alpha\gamma}exp(-\beta U)}{\int^{\prime}\prod_{\nu=1}^{M-1}d\vec{r}_{\alpha}^{(\nu)}exp(-\beta U)}$ (9) To calculate the Eq. 9 under the constraint of fixed $L_{\alpha}$, we can rewrite it as, $\Xi_{\alpha\gamma}=\frac{\int d\vec{r}_{\alpha}^{(1)}A_{\alpha\gamma}\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})\bar{U}}{\int d\vec{r}_{i}^{(1)}\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})\bar{U}}$ (10) where $\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})=\int^{\prime}\prod_{\nu=2}^{M-1}d\vec{r}_{\alpha}^{(\nu)}$, which is the number of configurations for fixed $L_{\alpha}$, $\vec{r}_{\alpha}^{(1)}$ and $\vec{r}_{\alpha}^{(M)}$. $\bar{U}$ reads, $\bar{U}=\int\prod_{\nu=2}^{M-1}d\vec{r}_{\alpha}^{(\nu)}exp(-\beta U)/\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})$ (11) The above integral is also under the constraint of fixed $L_{\alpha}$. There is almost no hope to evaluate Eq. 10 exactly. In this paper, we calculate it with approximations, $\Xi_{\alpha\gamma}\approx\frac{1}{\Upsilon(L_{\alpha},\vec{r}_{\alpha}^{(M)})}\int d\vec{r}_{\alpha}^{(1)}A_{\alpha\gamma}\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)}),$ (12) $\Upsilon(L_{\alpha},\vec{r}_{\alpha}^{(M)})$ is the total number of configurations for fixed $L_{\alpha}$ and $\vec{r}_{\alpha}^{(M)}$, _i.e_ , $\Upsilon(L_{\alpha},\vec{r}_{\alpha}^{(M)})=\int d\vec{r}_{\alpha}^{(1)}\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)}).$ By replacing Eq. 10 with Eq. 12, we have assumed that $\bar{U}$ is weakly dependent of $\vec{r}_{\alpha}^{(1)}$ for fixed $\vec{r}_{\alpha}^{(M)}$, $\vec{r}_{\gamma}^{(M)}$ and $L_{\alpha}$. This approximation works well if _M_ is not too small. The reason lies on the fact that, only the configurations, in which $\vec{r}_{\alpha}^{(1)}$ is close to $\vec{r}_{\alpha}^{(M)}$, make the $\Delta(L_{i},\vec{r}_{i}^{(1)},\vec{r}_{i}^{(M)})$ be significant(see below Eq. 13 and 14). Our numerical test (see below) also demonstrates this point. $\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})$ can be written as an integral over three Cartesian directions, $\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})=\int\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})\bar{\Delta}(L_{\alpha y},y_{\alpha}^{(1)},y_{\alpha}^{(M)})\bar{\Delta}(L_{\alpha z},z_{\alpha}^{(1)},z_{\alpha}^{(M)})dL_{\alpha x}dL_{\alpha y}dL_{\alpha z},$ (13) where the integral is evaluated under the constraint: $L_{\alpha}^{2}=L_{\alpha x}^{2}+L_{\alpha y}^{2}+L_{\alpha z}^{2}$. $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$ is the number of configurations for fixed $L_{\alpha x}$ ($L_{\alpha x}^{2}=\sum_{\nu=1}^{M}(x_{\alpha}^{(\nu+1)}-x_{\alpha}^{(\nu)})^{2}$) ,$x_{\alpha}^{(1)}$ and $x_{\alpha}^{(M)}$. $\bar{\Delta}(L_{\alpha y},y_{\alpha}^{(1)},y_{\alpha}^{(M)})$ and $\bar{\Delta}(L_{\alpha z},z_{\alpha}^{(1)},z_{\alpha}^{(M)})$ are the counterparts of $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$ along _y_ and _z_ direction respectively. To calculate $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$, we define a (M-1)-dimensional vector $\vec{R}$, of which Cartesian components are $(x_{\alpha}^{(1)}-x_{\alpha}^{(2)},....,x_{\alpha}^{(M-1)}-x_{\alpha}^{(M)})$. First, for a given $L_{\alpha x}$, it requires $\|\vec{R}\|$=$\sqrt{L_{\alpha x}^{2}-(x_{\alpha}^{(1)}-x_{\alpha}^{(M)})^{2}}$, all the configurations satisfying this condition lie on a surface of (M-1)-dimensional super-sphere with radius equal to $\sqrt{L_{\alpha x}^{2}-(x_{\alpha}^{(1)}-x_{\alpha}^{(M)})^{2}}$; Second, since the Cartesian components of $\vec{R}$ are not independent, _i.e_ , the project of $\vec{R}$ on the (M-1)-dimensional unit vector is $\frac{(x_{\alpha}^{(1)}-x_{\alpha}^{(M)})}{\sqrt{M-1}}$, this condition defines a (M-1)-dimensional super-plane; Thus, all the (M-1)-dimensional points, which attribute to $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$, lie on a (M-2)-dimensional super- spherical surface intersected by the (M-1)-dimensional super-sphere and the (M-1)-dimensional super-plane. According to analytic geometry in high dimensional space, $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$ is the proportional area of the (M-2)-dimensional super-spherical surface with radius equal to $(R^{2}-\frac{(x_{\alpha}^{(1)}-x_{\alpha}^{(M)})^{2}}{(M-1)})^{\frac{1}{2}}$. We end up with: $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})dRdx_{i}^{(1)}\propto C_{\bar{\Delta}}(R^{2}-\frac{(x_{\alpha}^{(1)}-x_{\alpha}^{(M)})^{2}}{(M-1)})^{\frac{M-3}{2}}dRdx_{\alpha}^{(1)}$ with $C_{\bar{\Delta}}=(M-2)\pi^{\frac{M-2}{2}}/((M-2)/2)!$.StaMec By changing integration variable from ($R$, $x_{\alpha}^{1}$) to ($L_{\alpha x}$, $x_{\alpha}^{1}$), we have, $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})dL_{\alpha x}dx_{\alpha}^{(1)}\propto C_{\bar{\Delta}}L_{\alpha x}(L_{\alpha x}^{2}-(x_{\alpha}^{(1)}-x_{\alpha}^{(M)})^{2}\frac{M}{M-1})^{\frac{M-4}{2}}dL_{\alpha x}dx_{\alpha}^{(1)},$ (14) Similarly, we can obtain $\bar{\Delta}(L_{\alpha y},y_{\alpha}^{(1)},y_{\alpha}^{(M)})$ and $\bar{\Delta}(L_{\alpha z},z_{\alpha}^{(1)},z_{\alpha}^{(M)})$, which have the same formula as $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$. Substituting $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$ in Eq. 14 for $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$ in Eq. 13, as well as replacing the counterparts of $\bar{\Delta}(L_{\alpha y},y_{\alpha}^{(1)},y_{\alpha}^{(M)})$ and $\bar{\Delta}(L_{\alpha z},z_{\alpha}^{(1)},z_{\alpha}^{(M)})$ in Eq. 13, $\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})$ can be obtained. As a result, $\Xi_{\alpha\gamma}$ ought to be calculated numerically. One can see that, $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$ quickly decays as a function of $\|x_{\alpha}^{(1)}-x_{\alpha}^{(M)}\|$. The behaviors of $\bar{\Delta}(L_{\alpha y},y_{\alpha}^{(1)},y_{\alpha}^{(M)})$ and $\bar{\Delta}(L_{\alpha z},z_{\alpha}^{(1)},z_{\alpha}^{(M)})$ are the same as that of $\bar{\Delta}(L_{\alpha x},x_{\alpha}^{(1)},x_{\alpha}^{(M)})$. Accordingly, $\Delta(L_{\alpha},\vec{r}_{\alpha}^{(1)},\vec{r}_{\alpha}^{(M)})$ also quickly decays as a function of $\|\vec{r}_{\alpha}^{(1)}-\vec{r}_{\alpha}^{(M)}\|$. By employing the change of variables as $\vec{r}_{\alpha}^{(1)}=\vec{r}_{\alpha}^{(M)}+\vec{\delta}_{\alpha}$, and making coordinate transformation in spherical coordinates, we can see that $\Xi_{\alpha\gamma}$ is a function of $L_{\alpha}$ and $r_{\alpha\gamma}^{(M)}$ ($r_{\alpha\gamma}^{(M)}=\|\vec{r}_{\alpha}^{(M)}-\vec{r}_{\alpha}^{(M)}\|$). After carrying out above coarse-grained approximation, we find that, the off- diagonal element $\Xi_{\alpha\gamma}$ is a function of both $r_{\alpha\gamma}^{(M)}$ and $L_{\alpha}$, which has an explicit nonlocal form. In contrast, the off-diagonal element $A_{\alpha\gamma}$ is a local function in imaginary time. More importantly, the off-diagonal element $A_{\alpha\gamma}$ could be much larger or much smaller than 1.0, while the off-diagonal element $\Xi_{\alpha\gamma}$ is less than 1.0 when $L_{\alpha}$ is not too long (see Fig. 1). Thus, after replacing $A_{\alpha\gamma}$ with $\Xi_{\alpha\gamma}$, at least for two-particle system, a lot of sign fluctuations are well canceled. However, since the length of path could be much longer at low temperature, the off-diagonal element $\Xi_{\alpha\gamma}$ can be larger than 1.0 (see Fig. 1), which will result in the presence of negative $det\Xi$. To further reduce the negative sign, we have made the second stage of coarse-grained approximation for longer paths. Using the similar idea as above, we can replace $\Xi$ by a $N\times N$ matrix $\Lambda$, $\displaystyle\Lambda_{\alpha\gamma}$ $\displaystyle=\left\\{\begin{array}[]{ll}\Xi_{\alpha\gamma},forL_{\alpha}\leq L_{\alpha}^{\ast}\\\ \eta(r_{\alpha\gamma}^{(M)}),forL_{\alpha}>L_{\alpha}^{\ast}\\\ \end{array}\right.$ (17) $L_{\alpha}^{\ast}$ is a function of $r_{\alpha\gamma}^{(M)}$, which is determined by, $\eta(r_{\alpha\gamma}^{(M)})=\frac{\int_{L_{\alpha}^{\ast}}^{\infty}\prod_{\nu=1}^{M-1}D\vec{r}_{\alpha}^{(\nu)}\Xi_{\alpha\gamma}exp(-\beta H)}{\int_{L_{\alpha}^{\ast}}^{\infty}\prod_{\nu=1}^{M-1}D\vec{r}_{\alpha}^{(\nu)}exp(-\beta H)}$ (18) The above integral is made over all the configurations with $L_{\alpha}>L_{\alpha}^{\ast}$. For an arbitrary interact potential, the above integral is almost no hope to be evaluated exactly. However it can be calculated with approximations, $\eta(r_{\alpha\gamma}^{(M)})\approx\frac{\int_{L_{\alpha}^{\ast}}^{\infty}dL_{\alpha}\Xi_{\alpha\gamma}exp(-\frac{1}{2}\beta m\omega_{M}^{2}L_{\alpha}^{2})\Upsilon(L_{\alpha},\vec{r}_{\alpha}^{(M)})}{\int_{L_{\alpha}^{\ast}}^{\infty}dL_{\alpha}exp(-\frac{1}{2}\beta m\omega_{M}^{2}L_{\alpha}^{2})\Upsilon(L_{\alpha},\vec{r}_{\alpha}^{(M)})}.$ (19) In the evaluation of the above coarse-grained approximation, we have assumed that total potential energy is constant when $L_{\alpha}$ is longer than certain value, _i.e_ , $L_{\alpha}^{\ast}$. In current work, we have taken $\eta(r_{\alpha\gamma}^{(M)})=1$ through the whole paper. Within the current approximations, $\Lambda$ has a few advantages over the original matrix _A_. First, our calculations have shown that the off-diagonal element of $\Lambda$ is not larger than 1.0 anywhere (see Fig. 1), in constrast the off-diagonal element of _A_ could be much larger than 1.0. Thus at least for two-particle system, $\Lambda$ is always non-negative, the _sign problem_ completely vanishes. Second, different from _A_ , $\Lambda$ is nonlocal, which depends on the whole path. The nonlocal behavior of $\Lambda$ can effectively avoid the _collapse_ of fermion system into boson system at low temperature. At low temperature, the length of path becomes longer and longer, so the off-diagonal element of $\Lambda$ has more chance being 1.0. This situation makes $\Lambda$ have little chance being unit matrix. Our calculation also demonstrates this point. It should be pointed out that the current formula is exact for free particles (see APPENDIX). Now we end up with the final formula for real calculations, $Z\cong C\int\prod_{i=1}^{N}\prod_{\nu=1}^{M}D\vec{r}_{i}^{(\nu)}det\Lambda exp[-\beta\sum_{i=1}^{N}\sum_{\nu=1}^{M}\frac{1}{2}m\omega_{M}^{2}(\vec{r}_{i}^{(\nu+1)}-\vec{r}_{i}^{(\nu)})^{2}+\sum_{\nu=1}^{M}\frac{1}{M}V(\\{\vec{r}_{i}^{(\nu)}\\})]$ (20) In real calculation, the element of $\Lambda$ is first numerically integrated. At the same time, the derivative of $\Lambda$ respective to temperature is also numerically calculated to account the contribution to thermal energy. Eq. 18 can not be directly used in standard MC, since for the fermionic systems, $det\Lambda$ is not always positive. However Eq. 18 can be integrated using modified MC technique, which is widely used previously.DeR81 ; Takahashi84 To achieve this result, we first defined pseudo-Hamiltonian, $H_{p}$, which is, $H_{p}=\sum_{i=1}^{N}\sum_{\nu=1}^{M}\frac{1}{2}m\omega_{M}^{2}(\vec{r}_{i}^{(\nu+1)}-\vec{r}_{i}^{(\nu)})^{2}+\sum_{\nu=1}^{M}\frac{1}{M}V(\\{\vec{r}_{i}^{(\nu)}\\})+ln\|det\Lambda\|.$ The thermodynamic average of a physical quantity Q is $\langle Q\rangle=\frac{\int\prod_{i=1}^{N}\prod_{\nu=1}^{M}D\vec{r}_{i}^{(\nu)}Q(\\{r_{i}^{(\nu)}\\})sgn(det\Lambda)exp(-\beta H_{p}\\})}{\int\prod_{i=1}^{N}\prod_{\nu=1}^{M}D\vec{r}_{i}^{(\nu)}sgn(det\Lambda)exp(-\beta H_{p}\\})\\})},$ (21) where $sgn(det\Lambda)$ stands for the sign of $det\Lambda$ at a configuration. It needs to be pointed out that, we have used the similar technique as most pseudopotential methods,Schnitker87 ; Landman87 ; Hall88 ; Kuki87 ; Coker87 ; Bartholomew85 ; Sprik85 ; Oh98 ; Miura01 but we do not recast matrix $\Lambda$ or extend $\Lambda$ into each imaginary time. ## III The Numerical Tests To illustrate the usefulness of the current method, we have considered _N_ interacting spinless fermions confined in a three-dimensional harmonic well, which Hamiltonian reads, $H=\sum_{j=1}^{N}(\frac{\vec{p}_{j}^{2}}{2m}+\frac{m\omega^{2}}{2}\vec{r}_{j}^{2})+\sum_{i<j}^{N}V(r_{ij}),$ (22) where _m_ , $\vec{r}_{j}$, $\vec{p}_{j}$ and $V(r_{ij})$ are mass, positions, momenta of the particles, and the inter-particle interaction potential, respectively. For computational simplicity, the units by which $m=\hbar=k_{B}=1$ are used through the rest of this paper. In current calculations, we consider three cases, _i.e_ , Case 1: $V(r_{ij})=0$, N=6 and $\omega^{2}=1$, no interaction between particles, reflecting a standard harmonic system; Case 2: N=6, $V(r_{ij})=-\frac{m\Omega^{2}}{2}r_{ij}^{2}$, where the interaction is also harmonic one with $\Omega^{2}=\frac{1.0}{4.0}$ and $\omega^{2}=4$; Case 3: $V(r_{ij})=\frac{q^{2}}{\|\vec{r}_{i}-\vec{r}_{j}\|}$, $\omega^{2}=0.320224986$, $q=1$ and N=2, interaction between particles is the Coulomb potential, the parameters correspond to hydrogen-like ion ($H^{+}$) of Kestner-Sinanoḡlu model.Kestner62 The exact results of all three cases can be found elsewhere,Kestner62 ; Brosens98 which is easy to check the validity of the current method. These models are widely used as a benchmark for checking the usefulness of various methods for _sign problem_ , see for examples Ref.Lyubartsev ; Newman91 ; Hall88 ; Miura01 Our Metropolis MC scheme is preformed based on Eq. 19. At each step, $H_{p}$ are calculated to determine the rejection and acceptance. $det\Lambda$ is calculated by a certain algorithm with the computational cost scaled by $N^{3}$.nr This kind of numerical technique enables us to perform the fermionic simulations with reasonable computational time. There are two basic types of moves in current simulations: (1) Displacement move, where all the coordinates for a single particle are displaced uniformly; (2) Standard bisection moves.Ceperley95 ; Chakravarty The MC procedure used in this work is wildly used by others. One MC step is defined as one application of each procedure. Ten million MC steps of calculation were carried out for each temperature. For a few cases, 100 million MC steps are made to check the ergodic problem. The results agree with the short runs within the error bars. To further check the ergodic problem, the simulations are carried out by a few random generated starting configurations. All simulations converge to the same results. The energy is calculated based on the thermodynamic estimator.Ceperley95 The energies are well converged at M/$\beta$=20, 22 and 5 for case 1, 2 and 3 respectively. The calculated thermal energy is in good agreement with the exact one for all three cases studied. In case 3, the exact energy 2.647 of Ref. Kestner62 has been almost accurately reproduced, which is 2.652$\pm$0.003 in current simulations. Fig. 2 shows the thermal energy per particle as a function of temperature for Case 1 and 2, the corresponding exact results are also shown in Fig. 2 with lines. As can be seen from the figure, the overall temperature dependence is well reproduced by current calculations. The calculated thermal energies agree very well with the exact value at low temperature. The slight deviation at high temperature is due to the fact that the first stage of approximation will result in error when the number of beads is too small, which is the case for high temperature. We have calculated the pair correlation function (PCF) between beads, which is defined as, $g(r)=\langle\frac{2}{MN(N-1)}\sum_{\nu}^{M}\sum_{i}^{N-1}\sum_{j>i}^{N}\delta(r-\|\vec{r}_{i}^{(\nu)}-\vec{r}_{j}^{(\nu)}\|)\rangle.$ It is known that,Miura01 comparing with boson and Boltzmann systems, the fermionic PCF has a hole around the origin, which reflects the Pauli exclusion principle. In Fig. 3, we present PCF for case 1 and 2 at temperature of 0.2. From this figure, we can see that, the pair correlation function clearly represents the effect due to the Pauli exclusion principle. The similar behaviors are observed for other temperature and systems. The average sign reflects the signal-to-noise ratio, which directly affects calculation precision and computation time needed. The average sign is defined as $Sign=(N_{+}-N_{-})/(N_{+}+N_{-})$, where $N_{+}$ and $N_{-}$ are the total positive and negative configuration respectively. The lower panel of Fig. 4 shows the average sign of current simulations via temperature. It can be seen that, $Sign$ decreases with the decrease of the temperature. However, for the studied systems, even at lowest temperature (T=0.1), the average sign is quite high (around 0.1). We also calculated the average sign via the number of particles at T=0.5 for both case 1 and 2, which is shown in the upper panel of Fig. 4. Similarly, $Sign$ also decreases with the increase of the number of particles. Although we have not completely solved the _sign problem_ , our approach does much improve the sign decay rate with both temperature and number of particles. Direct using of Eq. 4, $Sign$ is about 0.01 at temperature of 0.8 for case 1. And for temperature lower than 0.8, the large sign fluctuation makes MC simulation difficult to obtain any useful information. According to the data shown in Fig. 4, the maximum number of particles, which can be handled in current method, should be in order of ten. Considering both spin-up and -down, the maximum number of particles can be around twenty, which could be particularly useful for atom and molecular systems. ## IV Discussion and Conclusion In this paper, we have introduced an approach to reduce the fermion sign fluctuation in finite temperature PIMC simulations. By this method, configurations, which probably cause the sign fluctuation, are pre-calculated within two stages of coarse-grained approximations, while the rest are treated exactly. After two stages of coarse-grained approximations, at least for two- particle system, the _sign problem_ is solved completely. Since the exchange matrix $A$ is replaced by a non-local one ($\Lambda$), the collapse of fermion system into a boson one at low temperature has been effectively avoided. The pilot calculation was performed on three model systems: six independent particles in a three-dimensional harmonic well, six interacting particles in a three-dimensional harmonic well, and hydrogen-like ion ($H^{+}$) of Kestner- Sinanoḡlu model. The calculation shows that the current approach not only dramatically drops the sign fluctuation, but also gives an excellent description to real systems. Our method could be particularly useful for atom and molecular systems. Although our approach suffers from the _sign problem_ for large number of particles, we believe that it provide an alternative thought on the sign problem. We also believe that a similar approach can also be helpful in other path integral methods. The current formula can be easily extended to systems consisting of both spin-up and -down fermions.(see for example, Oh98 ; Takahashi ) Our approximation breaks down for systems including particles more than twenty (including both spin up and down particles). It would be possible to generalize our method for problems of larger numbers of fermions. Although we have used $\eta(r_{ij}^{(M)})=1$ through out this paper, other values are also possible. For example, if $\eta(r_{ij}^{(M)})=exp(-\alpha\frac{\beta}{2M}m\omega_{M}^{2}(r_{ij}^{(M)})^{2}$) is chosen, the current method can be more flexible. For $\alpha=0$, it is the case used in current work. For $\alpha=1$, $det\Lambda$ becomes the exact density matrix of free particles(see APPENDIX), thus the sign problem can be avoided completely. In fact, with $\alpha$ increasing from 0 to 1, the approximation becomes more and more crude, but the negative parts become less and less. To further improve the current method, a better form or value for $\eta(r_{ij}^{(M)})$ could be found. It is actually the issue on which we are working now. ###### Acknowledgements. I am very grateful to Prof. X. G. Gong and Prof. T. Xiang for valuable discussions and encouragements. And thank Prof. Feng Zhou for interesting discussions. I also would like to thank Guanwen Zhang for reading the manuscript prior to publication and for helpful suggestions. This work is supported by the National Natural Science Foundation of China, Shanghai Project for the Basic Research. The computation is performed in the Supercomputer Center of Shanghai. APPENDIX In this appendix, we will prove that the current formula is exact for free particles. Since the second stage of approximation is just a straightforward integration for free particles, we only prove the formula of first stage is correct for free particles. All Cartesian coordinates are equivalent for free particles, for simplicity we only prove it in one Cartesian direction, say, _x_. The partition function for free particles in one dimension has the form, $Z=\frac{1}{N!}\int\prod_{i=1}^{N}Dx_{i}^{(M)}\rho(\\{x_{i}^{(M)}\\},\\{x_{i}^{(M)}\\})$ (23) where $\rho(x_{i}^{(M)},x_{i}^{(M)})$ is the density matrix, of which element with current formula reads, $\rho_{ij}(\\{x_{i}^{M}\\},\\{x_{i}^{M}\\})=C_{1D}\int\prod_{i=1}^{N}\prod_{\nu=1}^{M-1}Dx_{i}^{(\nu)}\Xi_{ij}^{(1D)}exp(-\beta\sum_{i=1}^{N}\frac{1}{2}m\omega_{M}^{2}L_{ix}^{2})$ (24) where $C_{1D}=(\frac{mM}{2\pi\beta\hbar^{2}})^{\frac{M}{2}}$, and $\Xi_{ij}^{(1D)}$ is the one-dimensional counterpart of $\Xi_{ij}$, which is $\Xi_{ij}^{(1D)}=\frac{\int dx_{i}^{(1)}\bar{\Delta}(L_{ix},x_{i}^{(1)},x_{i}^{(M)})e^{-\frac{1}{2}\beta m\omega^{2}_{M}((x_{i}^{(1)}-x_{j}^{(M)})^{2}-(x_{i}^{(1)}-x_{i}^{(M)})^{2})}}{\Upsilon_{1D}(L_{xi},x_{i}^{(M)})}$ (25) $\Upsilon_{1D}(L_{ix},x_{i}^{(M)})=\int dx_{i}^{(1)}\bar{\Delta}(L_{ix},x_{i}^{(1)},x_{i}^{(M)})$ Eq. 22 is only relevant to $\\{L_{ix}\\}$ and $\\{x_{i}^{(1)}\\}$, the integration over $x_{i}^{(\nu)}$ ($\nu$=1,…M-1) can be replaced by (M-1)-dimensional spherical polar coordinates, _i.e_ , integration over $L_{ix}$ multiplying $\Upsilon_{1D}(L_{ix},x_{i}^{M})$, Eq. 22 becomes, $\rho_{ij}(\\{x_{i}^{M}\\},\\{x_{i}^{M}\\})\propto C_{1D}\int dL_{xi}dx_{i}^{(1)}\bar{\Delta}(L_{ix},x_{i}^{(1)},x_{i}^{(M)})$ (26) $\times exp(-\frac{1}{2}\beta m\omega_{M}^{2}(L_{ix}^{2}+(x_{i}^{(1)}-x_{j}^{(M)})^{2}-(x_{i}^{(1)}-x_{i}^{(M)})^{2}))$ Remembering for fixed $x_{i}^{(1)}$ and $x_{i}^{(M)}$, the minimum value of $L_{ix}^{2}$ equals to $(x_{i}^{(1)}-x_{i}^{(M)})^{2}\frac{M}{M-1}$. We first do the integration over variable $\\{L_{ix}\\}$ by change of variables as $L_{ix}^{{}^{\prime}2}=L_{ix}^{2}-(x_{i}^{(1)}-x_{i}^{(M)})^{2}\frac{M}{M-1}$, $\rho_{ij}(\\{x_{i}^{M}\\},\\{x_{i}^{M}\\})$ becomes, $\rho_{ij}(\\{x_{i}^{M}\\},\\{x_{i}^{M}\\})\propto C_{1D}C_{\bar{\Delta}}\frac{(\frac{M-4}{2})!}{(\frac{\beta m\omega_{M}^{2}}{2})^{\frac{M-2}{2}}}\int dx_{i}^{(1)}$ (27) $\times exp(-\frac{1}{2}\beta m\omega^{2}_{M}((x_{i}^{(1)}-x_{j}^{(M)})^{2}+\frac{1}{M-1}(x_{i}^{(1)}-x_{i}^{(M)})^{2}))$ $=C_{1D}C_{\bar{\Delta}}\frac{(\frac{M-4}{2})!}{(\frac{\beta m\omega_{M}^{2}}{2})^{\frac{M-2}{2}}}\sqrt{\frac{2(M-1)\pi}{M\beta m\omega_{M}^{2}}}exp(-\frac{1}{2M}\beta m\omega^{2}_{M}(x_{i}^{(M)}-x_{j}^{(M)})^{2})$ $=C_{M}(\frac{m}{2\pi\beta\hbar^{2}})^{\frac{1}{2}}exp(-\frac{1}{2M}\beta m\omega^{2}_{M}(x_{i}^{(M)}-x_{j}^{(M)})^{2}),$ where $C_{M}$ is an irrelevant constant. The above express is the exact formula for free particles. It needs to noted that, since we only can get the relative value for $\bar{\Delta}$, we could not obtain the absolute value of $C_{M}$. However the absolute value of $C_{M}$ is irrelevant to our calculation, which is a function of _M_ only. Our numerical test also shows that the calculated element of density matrix based on the current formula is in excellent agreement with the exact data. In fact, the current formula must be exact for free particles, since all the approximations become exact without potential part. ## References * (1) D. M. Ceperley, Rev. Mod. Phys. 67, 279 (1995). * (2) W. M. C. Foulkes et al, Rev. Mod. Phys. 73, 33 (2001). * (3) See, e.g., Quantum Monte Carlo Methods in Condensed Matter Physics, edited by M. Suzuki (World Scientific, Singapore, 1993), and references therein. * (4) E. Y. Loh, Jr., J. E. Gubernatis, R. T. Scalettar, S. R. White, D. J. Scalapino, and R. L. Sugar, Phys. Rev. B 41, 9301 (1990). * (5) J. B. Anderson, J. Chem. Phys. 63, 1499(1975); 65, 4121 (1976) * (6) D. M. Ceperley, Phys. Rev. Lett. 69, 331(1992) * (7) D. M. Ceperley and B. J. Alder, Phys. Rev. Lett. 45, 566 (1980); J. Chem. Phys. 81, 5833 (1984). * (8) B. Chen and J. B. Anderson, J. Chem. Phys. 102, 4491 (1995). * (9) C. H. Mak, R. Egger and H. Weber-Gottschick, Phys. Rev. Lett. 81 4533(1998). * (10) R. Egger, L. Mühlbacher and C. H. Mak, Phys. Rev. E 61 5961 (2000). * (11) R. A. Chiles, G. A. Jongeward, M. A. Bolton, and P. G. Wolynes, J. Chem. Phys. 81,2039 (1984). * (12) J. Schnitker and P. J. Rossky, J. Chem. Phys. 86, 347l (1987). * (13) U. Landman, R. N. Barnett, C. L. Cleveland, D. Scharf, and J. Jortner, International J. Quantum Chem., Quantum Chem. Symp. 21, 573 (1987). * (14) R. W. Hall, J. Chem. Phys. 89, 4212 (1988); J. Phys. Chem. 93,5628 (1989). * (15) A. Kuki and P. G. Wolynes, Science 236, 1647 (1987). * (16) D. F. Coker, B. J. Berne, and D. Thirumalai, J. Chem. Phys. 86, 5689 (1987). * (17) J. Bartholomew, R. Hall, and B. J. Berne, Phys. Rev. B 32, 548 (1985). * (18) M. Sprik, M. L. Klein, and D. Chandler, Phys. Rev. B 32, 545 ( 1985); Phys. Rev. B 31, 4234 (1985); J. Chem. Phys. 83,3042 ( 1985). * (19) Ki-dong Oh and P. A. Deymier, Phys. Rev. Lett. 81, 3104 (1998); Phys. Rev. B 58, 7577 (1998). * (20) S. Miura and S. Okazaki, J. Chem. Phys. 112, 10116 (2000); 115, 5353 (2001). * (21) W. H. Newman and A. Kuki, J. Chem. Phys. 96, 1409(1991). * (22) A. P. Lyubartsev, J. Phys. A: Math. Gen. 38, 6659 (2005); J. Phys. A: Math. Theor. 40, 7151 (2007). * (23) A. G. Moreira, S. A. Baeurle and G. H. Fredrickson, Phys. Rev. Lett. 91, 150201 (2003). * (24) S. Zhang, J. Carlson and J. E. Gubernatis, Phys. Rev. B 55, 7464(1997) * (25) S. Zhang, Phys. Rev. Lett. 83, 2777 (1999). * (26) P. Henelius and A. W. Sandvik Phys. Rev. B 62 1102(2000). * (27) R. P. Pathria, _Statistical Mechanics_ , (Elsevier(Singapore) Pte Ltd. 2003), pp. 504. * (28) See, for example, W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in Fortran, 2nd ed. (Cambridge U.P., New York, 1992). * (29) H. De Raedt and A. Lagendijk, Phys. Rev. Lett. 46, 77 (1981). * (30) M. Takahashi and M. Imada, J. Phys. Soc. Jpn. 53, 963 (1984). * (31) N. R. Kestner and O. Sinanoḡlu, Phys. Rev. 128, 2687 (1962). * (32) F. Brosens, J. T. Devreese and L. F. Lemmens, Phys. Rev. E 57, 3871 (1998). * (33) C. Chakravarty, M. C. Gordillo and D. M. Ceperley, J. Chem. Phys. 109, 2124 (1998). * (34) M. Takahashi and M. Imada, J. Phys. Soc. Jpn. 53, 963 (1984). Figure 1: (color online) The off-diagonal element of $\Lambda$ and $\Xi$ ($\Lambda_{ij}$ and $\Xi_{ij}$) as a function of particle separation ($r_{ij}$) for a few selected lengths of path ($L$) at T=0.5. When the length of path is short, $\Lambda_{ij}$ and $\Xi_{ij}$ are the same. As length being longer, $\Xi_{ij}$ can be larger than one for short particle separation. In contrast, $\Lambda_{ij}$ is not larger than one for any case. Figure 2: (color online) The thermal energy of case 1 (Circle) and 2 (Squares) as a function of temperature calculated by PIMC. The solid and dash lines are the exact energies of case 1 and 2 respectively. The agreement is quite well. Figure 3: The pair correlation functions between beads at T=0.2 for both case 1 and 2. The hole around original is the reflection of Pauli exclusion principle. Figure 4: The average sign of case 1 (circles) and 2 (squares) via temperature (lower panel) and number of particles (upper panel) as calculated by our PIMC simulation. _Sign_ decreases with the decrease of temperature, and the increase of number of particles.	arxiv-papers	2008-08-18T02:50:14	2024-09-04T02:48:57.338099	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "D. Y. Sun", "submitter": "Deyan Sun", "url": "https://arxiv.org/abs/0808.2329" }
0808.2344	# The $I^{G}J^{PC}=0^{+}1^{-+}$ Tetraquark State Hua-Xing Chen1,2,3 hxchen@rcnp.osaka-u.ac.jp Atsushi Hosaka2 hosaka@rcnp.osaka-u.ac.jp Shi-Lin Zhu1 zhusl@phy.pku.edu.cn 1Department of Physics, Peking University, Beijing 100871, China 2Research Center for Nuclear Physics, Osaka University, Ibaraki 567–0047, Japan 3State Key Laboratory of Nuclear Physics and Technology, Peking University, Beijing 100871, China ###### Abstract We study the tetraquark state with $I^{G}J^{PC}=0^{+}1^{-+}$ in the QCD sum rule. We exhaust all possible flavor structures by using a diquark-antidiquark construction and find that the flavor structure $(\mathbf{\bar{3}}\otimes\mathbf{\bar{6}})\oplus(\mathbf{6}\otimes\mathbf{3})$ is preferred. There are altogether four independent currents which have the quark contents $qs\bar{q}\bar{s}$. By using both the Shifman-Vainshtein- Zakharov (SVZ) sum rule and the finite energy sum rule, these currents lead to mass estimates around $1.8-2.1$ GeV, where the uncertainty is due to the mixing of two single currents. Its possible decay modes are $S$-wave $b_{1}(1235)\eta$ and $b_{1}(1235)\eta^{\prime}$, and $P$-wave $KK$, $\eta\eta$, $\eta\eta^{\prime}$ and $\eta^{\prime}\eta^{\prime}$, etc. The decay width is around 150 MeV through a rough estimation. exotic mesons, tetraquark, QCD sum rule ###### pacs: 12.39.Mk, 11.40.-q, 12.38.Lg Manifestly exotic hadron states which are not reached by three quarks for baryons and a quark-antiquark pair for mesons provide one of the most important subjects in hadron physics. The confirmation of their existence (or nonexistence) and the study of their structure are of great importance for the understanding of strong interaction dynamics at low energy exotic . Quantum numbers can tell whether a hadron is exotic or not. For instance baryons with strangeness $S=+1$ and mesons with $J^{PC}=1^{-+}$ are such states. For the baryon sector, the pentaquark $\Theta^{+}$ has been studied intensively since 2003 Nakano:2003qx . But the existence is still controversial. For the meson sector, the $\pi_{1}$ mesons of $I^{G}J^{PC}=1^{-}1^{-+}$ are listed as manifestly exotic states in the PDG for some time Yao:2006px ; experiments ; Lu:2004yn , and a lot of theoretical considerations have been made theory ; isoscalar . So far, many of them are for the isovector $I=1$ states. In principle, an isoscalar state is also possible, though not observed experimentally isoscalar . We have performed the QCD sum rule analyses of the light scalar mesons ($\sigma$, $\kappa$, $f_{0}$ and $a_{0}$), $Y(2175)$ and $\pi_{1}$ mesons Chen ; Chen:2008qw . All our results are consistent with the experimental observations. Encouraged by this, we would like to extend the QCD sum rule analysis using tetraquark currents for these $I^{G}J^{PC}=0^{+}1^{-+}$ states. Figure 1: Weight diagrams for $\mathbf{6_{f}}\otimes\mathbf{\bar{6}_{f}}(\mathbf{S})$ (top panel), $\mathbf{\bar{3}_{f}}\otimes\mathbf{3_{f}}(\mathbf{A})$ (middle panel), and $\mathbf{\bar{3}_{f}}\otimes\mathbf{\bar{6}_{f}}(\mathbf{M})$ (bottom panel). The weight diagram for $\mathbf{6_{f}}\otimes\mathbf{3_{f}}(\mathbf{M})$ is the charge-conjugation transformation of the bottom one. The QCD sum rule requires a computation of a two-point correlation function in the form of operator product expansion (OPE), which is then fitted by a phenomenological function to extract physical hadron properties sumrule . To calculate the OPE, we need employ an interpolating field (current) which couples to the physical state we consider. For tetraquarks, there are several independent currents and it is important to establish how one or some of them should be chosen. We have systematically performed the classification of currents by using the diquark-antidiquark ($(qq)(\bar{q}\bar{q})$) construction Chen ; Chen:2008qw . The currents constructed from the quark- antiquark pairs ($(\bar{q}q)(\bar{q}q)$) can be written as a combination of these ($(qq)(\bar{q}\bar{q})$) currents. We note here that the mixing can happen between hybrid states, tetraquark states, and even six-quark states, while the currents can also couple to all these states. However, it always makes sense to clarify what a single channel problem tells us before entering more sophisticated coupled channel problems. Therefore, here we concentrate exclusively on the tetraquark properties with some details. The tetraquark currents with the quantum numbers $J^{PC}=1^{-+}$ have been constructed in our previous paper Chen:2008qw . Now we need construct the isoscalar ones. The flavor structures are shown in Fig. 1 in terms of $SU(3)$ weight diagrams. The ideal mixing scheme is used since it is expected to work well for hadrons except for the pseudoscalar mesons. In order to have a definite charge-conjugation parity, the diquark and antidiquark inside can have the same flavor symmetry, which is either symmetric $\mathbf{6_{f}}\otimes\mathbf{\bar{6}_{f}}$ ($\mathbf{S}$) or antisymmetric $\mathbf{\bar{3}_{f}}\otimes\mathbf{3_{f}}$ ($\mathbf{A}$). Another option is the combination of $\mathbf{\bar{3}_{f}}\otimes\mathbf{\bar{6}_{f}}$ and $\mathbf{6_{f}}\otimes\mathbf{3_{f}}$ ($\mathbf{M}$), which can also have a definite charge-conjugation parity. From Fig. 1, we find that there are althgether six isospin singlets: $\displaystyle qq\bar{q}\bar{q}(\mathbf{S})\,,qs\bar{q}\bar{s}(\mathbf{S})\,,ss\bar{s}\bar{s}(\mathbf{S})\sim\mathbf{6_{f}}\otimes\mathbf{\bar{6}_{f}}~{}~{}~{}(\mathbf{S})\,,$ $\displaystyle qq\bar{q}\bar{q}(\mathbf{A})\,,qs\bar{q}\bar{s}(\mathbf{A})\sim\mathbf{\bar{3}_{f}}\otimes\mathbf{3_{f}}~{}~{}~{}(\mathbf{A})\,,$ (1) $\displaystyle qs\bar{q}\bar{s}(\mathbf{M})\sim(\mathbf{\bar{3}_{f}}\otimes\mathbf{\bar{6}_{f}})\oplus(\mathbf{6_{f}}\otimes\mathbf{3_{f}})~{}(\mathbf{M})\,,$ where $q$ represents an $up$ or $down$ quark, and $s$ represents a $strange$ quark. For each state, there are several independent currents. We list them in the following. 1. 1. For the three isospin singlets of $\mathbf{6}_{f}\otimes\mathbf{\bar{6}}_{f}$ ($\mathbf{S}$): $\displaystyle\begin{array}[]{ll}\eta^{S}_{1\mu}&\sim u_{a}^{T}C\gamma_{5}d_{b}(\bar{u}_{a}\gamma_{\mu}\gamma_{5}C\bar{d}_{b}^{T}+\bar{u}_{b}\gamma_{\mu}\gamma_{5}C\bar{d}_{a}^{T})\\\ &+u_{a}^{T}C\gamma_{\mu}\gamma_{5}d_{b}(\bar{u}_{a}\gamma_{5}C\bar{d}_{b}^{T}+\bar{u}_{b}\gamma_{5}C\bar{d}_{a}^{T})\,,\\\ \eta^{S}_{2\mu}&\sim u_{a}^{T}C\gamma^{\nu}d_{b}(\bar{u}_{a}\sigma_{\mu\nu}C\bar{d}_{b}^{T}-\bar{u}_{b}\sigma_{\mu\nu}C\bar{d}_{a}^{T})\\\ &+u_{a}^{T}C\sigma_{\mu\nu}d_{b}(\bar{u}_{a}\gamma^{\nu}C\bar{d}_{b}^{T}-\bar{u}_{b}\gamma^{\nu}C\bar{d}_{a}^{T})\,,\end{array}$ (6) $\displaystyle\begin{array}[]{ll}\eta^{S}_{3\mu}&\sim u_{a}^{T}C\gamma_{5}s_{b}(\bar{u}_{a}\gamma_{\mu}\gamma_{5}C\bar{s}_{b}^{T}+\bar{u}_{b}\gamma_{\mu}\gamma_{5}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\gamma_{\mu}\gamma_{5}s_{b}(\bar{u}_{a}\gamma_{5}C\bar{s}_{b}^{T}+\bar{u}_{b}\gamma_{5}C\bar{s}_{a}^{T})\,,\\\ \eta^{S}_{4\mu}&\sim u_{a}^{T}C\gamma^{\nu}s_{b}(\bar{u}_{a}\sigma_{\mu\nu}C\bar{s}_{b}^{T}-\bar{u}_{b}\sigma_{\mu\nu}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\sigma_{\mu\nu}s_{b}(\bar{u}_{a}\gamma^{\nu}C\bar{s}_{b}^{T}-\bar{u}_{b}\gamma^{\nu}C\bar{s}_{a}^{T})\,.\end{array}$ (11) $\displaystyle\begin{array}[]{ll}\eta^{S}_{5\mu}&\sim s_{a}^{T}C\gamma_{5}s_{b}(\bar{s}_{a}\gamma_{\mu}\gamma_{5}C\bar{s}_{b}^{T}+\bar{s}_{b}\gamma_{\mu}\gamma_{5}C\bar{s}_{a}^{T})\\\ &+s_{a}^{T}C\gamma_{\mu}\gamma_{5}s_{b}(\bar{s}_{a}\gamma_{5}C\bar{s}_{b}^{T}+\bar{s}_{b}\gamma_{5}C\bar{s}_{a}^{T})\,,\\\ \eta^{S}_{6\mu}&\sim s_{a}^{T}C\gamma^{\nu}s_{b}(\bar{s}_{a}\sigma_{\mu\nu}C\bar{s}_{b}^{T}-\bar{s}_{b}\sigma_{\mu\nu}C\bar{s}_{a}^{T})\\\ &+s_{a}^{T}C\sigma_{\mu\nu}s_{b}(\bar{s}_{a}\gamma^{\nu}C\bar{s}_{b}^{T}-\bar{s}_{b}\gamma^{\nu}C\bar{s}_{a}^{T})\,.\end{array}$ (16) where $\eta^{S}_{1\mu}$ and $\eta^{S}_{2\mu}$ are the two independent currents containing only light flavors; $\eta^{S}_{3\mu}$ and $\eta^{S}_{4\mu}$ are the two independent ones containing one $s\bar{s}$ pair; $\eta^{S}_{5\mu}$ and $\eta^{S}_{6\mu}$ are the two independent ones containing two $s\bar{s}$ pairs. 2. 2. For the two isospin singlets of $\mathbf{\bar{3}}_{f}\otimes\mathbf{3}_{f}$ ($\mathbf{A}$): $\displaystyle\begin{array}[]{ll}\eta^{A}_{1\mu}&\sim u_{a}^{T}C\gamma_{5}d_{b}(\bar{u}_{a}\gamma_{\mu}\gamma_{5}C\bar{d}_{b}^{T}-\bar{u}_{b}\gamma_{\mu}\gamma_{5}C\bar{d}_{a}^{T})\\\ &+u_{a}^{T}C\gamma_{\mu}\gamma_{5}d_{b}(\bar{u}_{a}\gamma_{5}C\bar{d}_{b}^{T}-\bar{u}_{b}\gamma_{5}C\bar{d}_{a}^{T})\,,\\\ \eta^{A}_{2\mu}&\sim u_{a}^{T}C\gamma^{\nu}d_{b}(\bar{u}_{a}\sigma_{\mu\nu}C\bar{d}_{b}^{T}+\bar{u}_{b}\sigma_{\mu\nu}C\bar{d}_{a}^{T})\\\ &+u_{a}^{T}C\sigma_{\mu\nu}d_{b}(\bar{u}_{a}\gamma^{\nu}C\bar{d}_{b}^{T}+\bar{u}_{b}\gamma^{\nu}C\bar{d}_{a}^{T})\,,\end{array}$ (21) $\displaystyle\begin{array}[]{ll}\eta^{A}_{3\mu}&\sim u_{a}^{T}C\gamma_{5}s_{b}(\bar{u}_{a}\gamma_{\mu}\gamma_{5}C\bar{s}_{b}^{T}-\bar{u}_{b}\gamma_{\mu}\gamma_{5}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\gamma_{\mu}\gamma_{5}s_{b}(\bar{u}_{a}\gamma_{5}C\bar{s}_{b}^{T}-\bar{u}_{b}\gamma_{5}C\bar{s}_{a}^{T})\,,\\\ \eta^{A}_{4\mu}&\sim u_{a}^{T}C\gamma^{\nu}s_{b}(\bar{u}_{a}\sigma_{\mu\nu}C\bar{s}_{b}^{T}+\bar{u}_{b}\sigma_{\mu\nu}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\sigma_{\mu\nu}s_{b}(\bar{u}_{a}\gamma^{\nu}C\bar{s}_{b}^{T}+\bar{u}_{b}\gamma^{\nu}C\bar{s}_{a}^{T})\,,\end{array}$ (26) where $\eta^{A}_{1\mu}$ and $\eta^{A}_{2\mu}$ are the two independent currents containing only light flavors; $\eta^{A}_{3\mu}$ and $\eta^{A}_{4\mu}$ are the two independent ones containing one $s\bar{s}$ pair. 3. 3. For the isospin singlet of $(\mathbf{\bar{3}}_{f}\otimes\mathbf{\bar{6}}_{f})\oplus(\mathbf{6}_{f}\otimes\mathbf{3}_{f})$ ($\mathbf{M}$), $\displaystyle\begin{array}[]{ll}\eta^{M}_{1\mu}&\sim u_{a}^{T}C\gamma_{\mu}s_{b}(\bar{u}_{a}C\bar{s}_{b}^{T}+\bar{u}_{b}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}Cs_{b}(\bar{u}_{a}\gamma_{\mu}C\bar{s}_{b}^{T}+\bar{u}_{b}\gamma_{\mu}C\bar{s}_{a}^{T})\,,\\\ \eta^{M}_{2\mu}&\sim u_{a}^{T}C\sigma_{\mu\nu}\gamma_{5}s_{b}(\bar{u}_{a}\gamma^{\nu}\gamma_{5}C\bar{s}_{b}^{T}+\bar{u}_{b}\gamma^{\nu}\gamma_{5}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\gamma^{\nu}\gamma_{5}s_{b}(\bar{u}_{a}\sigma_{\mu\nu}\gamma_{5}C\bar{s}_{b}^{T}+\bar{u}_{b}\sigma_{\mu\nu}\gamma_{5}C\bar{s}_{a}^{T})\,,\\\ \eta^{M}_{3\mu}&\sim u_{a}^{T}Cs_{b}(\bar{u}_{a}\gamma_{\mu}C\bar{s}_{b}^{T}-\bar{u}_{b}\gamma_{\mu}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\gamma_{\mu}s_{b}(\bar{u}_{a}C\bar{s}_{b}^{T}-\bar{u}_{b}C\bar{s}_{a}^{T})\,,\\\ \eta^{M}_{4\mu}&\sim u_{a}^{T}C\gamma^{\nu}\gamma_{5}s_{b}(\bar{u}_{a}\sigma_{\mu\nu}\gamma_{5}C\bar{s}_{b}^{T}-\bar{u}_{b}\sigma_{\mu\nu}\gamma_{5}C\bar{s}_{a}^{T})\\\ &+u_{a}^{T}C\sigma_{\mu\nu}\gamma_{5}s_{b}(\bar{u}_{a}\gamma^{\nu}\gamma_{5}C\bar{s}_{b}^{T}-\bar{u}_{b}\gamma^{\nu}\gamma_{5}C\bar{s}_{a}^{T})\,,\end{array}$ (35) where $\eta^{M}_{i\mu}$ are the four independent ones containing one $s\bar{s}$ pair. The above structure has some implications on their decay patterns. The expressions of Eqs. (6)-(35) are not exactly correct, since they do not have a definite isospin. For instance, the current $\eta^{A}_{3\mu}$ should contain $(us\bar{u}\bar{s}+ds\bar{d}\bar{s})$ in order to have $I=0$. However, in the following QCD sum rule analysis, we find that there is no difference between these two cases in the limit that the masses and condensates of the $up$ and $down$ quarks are the same. Actually we also ignore a small quark mass effect ($m_{u}\sim m_{d}\lesssim 10$ MeV). By using these tetraquark currents, we have performed the OPE calculation up to dimension 12. Values for various condensates and $m_{s}$ follow the references Yao:2006px ; values . There are altogether 14 currents. It turns out that some of them lead to the same results of OPEs as the previous ones in Ref. Chen:2008qw : $\eta^{S}_{1,2,3,4\mu}\sim\eta^{S}_{1,2,3,4\mu}$ Chen:2008qw , $\eta^{A}_{3,4\mu}\sim\eta^{A}_{1,2\mu}$ Chen:2008qw , and $\eta^{M}_{1,2,3,4\mu}\sim\eta^{M}_{5,6,7,8\mu}$ Chen:2008qw . Therefore, we just need calculate the OPEs of $\eta^{S}_{5,6\mu}$ and $\eta^{A}_{1,2\mu}$. The full OPE expressions are too lengthy and are omitted here. In our previous paper Chen:2008qw we have found that the OPEs of the currents $\eta^{S}_{i\mu}$’s and $\eta^{A}_{i\mu}$’s lead to unphysical results where the spectral densities $\rho(s)$ become negative in the region of $2$ GeV2 $\lesssim s\lesssim 4$ GeV2. We find this to be the case also for the isoscalar currents. Therefore, our QCD sum rule analysis does not support a tetraquark state which has a flavor structure either $\mathbf{6_{f}}\otimes\mathbf{\bar{6}_{f}}$ or $\mathbf{\bar{3}_{f}}\otimes\mathbf{3_{f}}$ and a mass less than 2 GeV. We shall discuss only the currents of the mixed flavor symmetry. We find there is only one set of four independent currents as given in Eqs. (35), unlike the isovector case which have two sets. The spectral densities calculated by the mixed currents are positive for a wide range of $s$, and the convergence of OPE is very good in the region of $2$ GeV${}^{2}<M_{B}^{2}<$ 5GeV2 as in our previous study Chen:2008qw . In general, the pole contribution should be large enough in the SVZ sum rule. However, the pole contributions of multiquark states are rather small due to the large continuum contribution. Therefore a careful choice of the threshold parameter is important in order to subtract the continuum contribution. When using the SVZ sum rule, the mass is obtained as functions of Borel mass $M_{B}$ and threshold value $s_{0}$. As an example, we show the mass calculated from currents $\eta^{M}_{2\mu}$ in Fig. 2. The Borel mass dependence is weak, as shown in the upper figure; the $s_{0}$ dependence has a minimum where the stability is the best, as shown in the bottom figure. The minimum is around 2.0 GeV, which we choose to be our prediction. The other three independent currents $\eta^{M}_{1\mu}$, $\eta^{M}_{3\mu}$ and $\eta^{M}_{4\mu}$ lead to similar results, which are around 2.1 GeV, 1.9 GeV and 2.0 GeV respectively. Figure 2: The mass of the state $qs\bar{q}\bar{s}$ calculated by using the current $\eta^{M}_{2\mu}$, as functions of $M_{B}^{2}$ (left) and $s_{0}$ (right) in units of GeV. When using the finite energy sum rule, the mass is obtained as a function of the threshold value $s_{0}$, which is shown in Fig. 3. There is also a mass minimum around 2.1 GeV, 1.9 GeV, 1.9 GeV and 2.0 GeV for currents $\eta^{M}_{1\mu}$, $\eta^{M}_{2\mu}$, $\eta^{M}_{3\mu}$ and $\eta^{M}_{4\mu}$ respectively. In a short summary, we have performed a QCD sum rule analysis for $qs\bar{q}\bar{s}$. The mass obtained is around 2.0 GeV. We label this state $\sigma_{1}(2000)$. Figure 3: The mass calculated using the finite energy sum rule. The labels besides the lines indicate the suffix i of the current $\eta^{M}_{i\mu}$ ($i=1,\cdots,4$). We can also study the mixing of these four currents. The currents $\eta^{M}_{1\mu}$ and $\eta^{M}_{3\mu}$ have the largest mass difference, so we study their mixing as an example: $\displaystyle\eta^{M}_{mix}={\rm cos}\theta\eta^{M}_{1\mu}+{\rm sin}\theta\eta^{M}_{3\mu}\,,$ (36) where $\theta$ is the mixed angle. We calculate its OPE, and find that the resulting spectral density is just: $\displaystyle\rho^{M}_{mix}={\rm cos}^{2}\theta\rho^{M}_{1\mu}+{\rm sin}^{2}\theta\rho^{M}_{3\mu}\,,$ (37) The obtained mass is shown in Fig. 4 as functions of $\theta$. When we take $s_{0}=3.5$ GeV2 (solid line), the mass maximum is 2.05 GeV, and the minimum is 1.85 GeV. Therefore, we arrive at the similar result which produces the mass around 2 GeV. We can also consider the mixing of other currents, which would not change the results significantly due to the similarity of single currents. The mass estimates are around $1.8-2.1$ GeV, where the uncertainty is due to the mixing of two single currents. Figure 4: The mass calculated using the finite energy sum rule, and for the mixed current $\eta^{M}_{mix}$. The curves are obtained by setting $s_{0}=3.5$ GeV2 (solid line), 4 GeV2 (short-dashed line) and 5 GeV2 (long-dashed line). Now let us discuss its decay properties as expected from a naive fall-apart process. As shown in Eqs. (35) the currents contain one $s\bar{s}$ pair. Therefore, we expect that the final states should also contain one $s\bar{s}$ pair. In order to spell out the possible spin of decaying particles and their orbital angular momentum, we need perform a Fierz rearrangement to change $(qq)(\bar{q}\bar{q})$ currents to $(\bar{q}q)(\bar{q}q)$ ones. For illustration, we use one of the four independent $(\bar{q}q)(\bar{q}q)$ currents Chen:2008qw : $\displaystyle\xi^{M}_{2\mu}$ $\displaystyle=$ $\displaystyle(\bar{s}_{a}\gamma^{\mu}\gamma_{5}s_{a})(\bar{u}_{b}\gamma_{5}u_{b})-(\bar{s}_{a}\gamma_{5}s_{a})(\bar{u}_{b}\gamma^{\mu}\gamma_{5}u_{b})$ (38) $\displaystyle+\cdots\,.$ All terms of this current have the structure $(\bar{q}_{a}\gamma^{\mu}\gamma_{5}q_{a})(\bar{q}_{b}\gamma_{5}q_{b})$. Therefore, the expected decay patterns are: (1) $1^{+}$ and $0^{-}$ particles with relative angular momentum $L=0$, and (2) $0^{-}$ and $0^{-}$ particles with $L=1$. For the $S$-wave decay, we expect the following two-body decay patterns $\displaystyle\sigma_{1}(I^{G}J^{PC}=0^{+}1^{-+})$ $\displaystyle\rightarrow$ $\displaystyle a_{1}(1260)\eta,a_{1}\eta^{\prime},\cdots\,,$ (39) $\displaystyle b_{1}(1235)\eta,b_{1}\eta^{\prime}\cdots\,.$ If we consider, however, the $G$ parity conservation, the fist line is forbidden and the second line is the only one allowed. These modes can be observed in the final states $\omega\pi\eta$ and $\omega\pi\eta^{\prime}$. For the $P$-wave decay, we expect (with the $G$ parity conservation): $\displaystyle\sigma_{1}(I^{G}J^{PC}=0^{+}1^{-+})$ $\displaystyle\rightarrow$ $\displaystyle KK,\eta\eta,\eta\eta^{\prime},\eta^{\prime}\eta^{\prime}\cdots\,.$ (40) We can also estimate the (partial) decay width through the comparison with the observed $\pi_{1}(2015)$ Lu:2004yn , which has $\Gamma_{\rm tot}\sim 230$ MeV. Assuming that the decay of $\pi_{1}(2015)$ solely goes through $S$-wave $b_{1}\pi$ and that of $\sigma_{1}(2000)$ through $b_{1}\eta$, we expect $\Gamma_{\sigma_{1}\rightarrow b_{1}\eta}\sim 160$ MeV, as they are proportional to the $S$-wave phase space. For the $P$-wave decay there is an information $\pi_{1}(2015)\rightarrow\eta^{\prime}\pi$, which corresponds to $\sigma_{1}(2000)\rightarrow\eta^{\prime}\eta$ (Because both $\pi_{1}(1600)$ and $\pi_{1}(2015)$ have been observed in the final states $\pi\eta^{\prime}$ other than $\pi\eta$, we choose $\eta\eta^{\prime}$ to be the final states of $\sigma_{1}(2000)$ other than $KK$ and $\eta\eta$). Assuming once again that this is the unique decay mode, we expect that the decay width is approximately 130 MeV. If the decay occurs $50\%$ through $b_{1}\pi$ ($b_{1}\eta$) and $50\%$ through $\eta^{\prime}\pi$ ($\eta^{\prime}\eta$), we expect that $\Gamma_{\sigma_{1}}\sim 150$ MeV. In summary, we have performed the QCD sum rule analysis of the exotic tetraquark states with $I^{G}J^{PC}=0^{+}1^{-+}$. We test all possible flavor structures in the diquark-antidiquark $(qq)(\bar{q}\bar{q})$ construction, $\mathbf{6}\otimes\mathbf{\bar{6}}$, $\mathbf{\bar{3}}\otimes\mathbf{3}$ and $(\mathbf{\bar{3}}\otimes\mathbf{\bar{6}})\oplus(\mathbf{6}\otimes\mathbf{3})$. We find that only the mixed currents of the flavor structure $(\mathbf{\bar{3}}\otimes\mathbf{\bar{6}})\oplus(\mathbf{6}\otimes\mathbf{3})$ allow a positive and convergent OPE, and there is only one choice with the quark content $qs\bar{q}\bar{s}$, which have four independent currents. We have then performed both the SVZ sum rule and the finite energy sum rule. The mass estimates are around $1.8-2.1$ GeV, where the uncertainty is due to the mixing of two single currents. The possible decay modes are $S$-wave $b_{1}(1235)\eta$ and $b_{1}(1235)\eta^{\prime}$, and $P$-wave $KK$, $\eta\eta$, $\eta\eta^{\prime}$ and $\eta^{\prime}\eta^{\prime}$, etc. The decay width is around 150 MeV through a rough estimation. Here we want to note that we do not know how to determine the mixing angle, which is an interesting problem. ## Acknowledgments H.X.C. is grateful for Monkasho support for his stay at the Research Center for Nuclear Physics where this work was done. This project was supported by the National Natural Science Foundation of China under Grants No. 10625521, 10721063, the Ministry of Education of China, and the Grant for Scientific Research ((C) No. 19540297) from the Ministry of Education, Culture, Science and Technology, Japan. ## References * (1) R. L. Jaffe, Phys. Rev. D 15, 267 (1977); R. L. Jaffe, Phys. Rev. D 15, 281 (1977); E. Klempt and A. Zaitsev, Phys. Rept. 454, 1 (2007); S. L. Zhu, Int. J. Mod. Phys. E 17, 283 (2008). * (2) T. Nakano et al. [LEPS Collaboration], Phys. Rev. Lett. 91, 012002 (2003). * (3) W. M. Yao et al. [Particle Data Group], J. Phys. G 33, 1 (2006). * (4) D. R. Thompson et al. [E852 Collaboration], Phys. Rev. Lett. 79, 1630 (1997); A. Abele et al. [Crystal Barrel Collaboration], Phys. Lett. B 446, 349 (1999); G. S. Adams et al. [E862 Collaboration], Phys. Lett. B 657, 27 (2007); M. Nozar et al. [CLAS Collaboration], arXiv:0805.4438 [hep-ex]. * (5) M. Lu et al. [E852 Collaboration], Phys. Rev. Lett. 94, 032002 (2005). * (6) C. McNeile et al., Nucl. Phys. Proc. Suppl. 73, 264 (1999); P. Lacock and K. Schilling [TXL collaboration], Nucl. Phys. Proc. Suppl. 73, 261 (1999); K. G. Chetyrkin and S. Narison, Phys. Lett. B 485, 145 (2000); H. Y. Jin, J. G. Korner and T. G. Steele, Phys. Rev. D 67, 014025 (2003); K. C. Yang, Phys. Rev. D 76, 094001 (2007). * (7) C. Bernard et al., Phys. Rev. D 68, 074505 (2003); F. Iddir and L. Semlala, arXiv:0710.5352 [hep-ph]; P. R. Page, E. S. Swanson and A. P. Szczepaniak, Phys. Rev. D 59, 034016 (1999). * (8) H. X. Chen, A. Hosaka and S. L. Zhu, Phys. Lett. B 650, 369 (2007); H. X. Chen, X. Liu, A. Hosaka and S. L. Zhu, Phys. Rev. D 78, 034012 (2008). * (9) H. X. Chen, A. Hosaka and S. L. Zhu, Phys. Rev. D 78, 054017 (2008). * (10) M. A. Shifman, A. I. Vainshtein and V. I. Zakharov, Nucl. Phys. B 147, 385 (1979); L. J. Reinders, H. Rubinstein and S. Yazaki, Phys. Rept. 127, 1 (1985). * (11) K. C. Yang, W. Y. P. Hwang, E. M. Henley and L. S. Kisslinger, Phys. Rev. D 47, 3001 (1993); S. Narison, Camb. Monogr. Part. Phys. Nucl. Phys. Cosmol. 17, 1 (2002); V. Gimenez, V. Lubicz, F. Mescia, V. Porretti and J. Reyes, Eur. Phys. J. C 41, 535 (2005); M. Jamin, Phys. Lett. B 538, 71 (2002); B. L. Ioffe and K. N. Zyablyuk, Eur. Phys. J. C 27, 229 (2003); A. A. Ovchinnikov and A. A. Pivovarov, Sov. J. Nucl. Phys. 48, 721 (1988) [Yad. Fiz. 48, 1135 (1988)]; W. Y. P. Hwang and K. C. Yang, Phys. Rev. D 49, 460 (1994).	arxiv-papers	2008-08-18T07:03:31	2024-09-04T02:48:57.343823	{ "license": "Public Domain", "authors": "Hua-Xing Chen, Atsushi Hosaka, and Shi-Lin Zhu", "submitter": "Hua-Xing Chen", "url": "https://arxiv.org/abs/0808.2344" }
0808.2421	# Dynamics of driven vortex-antivortex matter in superconducting films with a magnetic dipole array Cléssio L. S. Lima Departamento de Física, Universidade Federal de Pernambuco, 50670-901 Recife-PE, Brazil Instituto de Física, Universidade Federal do Rio de Janeiro, C.P. 68528, 21945-970, Rio de Janeiro-RJ, Brazil Clécio C. de Souza Silva Departamento de Física, Universidade Federal de Pernambuco, 50670-901 Recife-PE, Brazil ###### Abstract We investigate theoretically vortex-antivortex (v-av) matter moving in thin superconducting films with a regular array of in-plane magnetic dipoles. Our model considers v-av pair creation induced by the local current density generated by the magnetic texture and the transport current and simulates the dynamics of vortices and antivortices by numerical integration of the Langevin equation of motion. Calculations of the transport properties at zero applied field show a strong dependence of the v-av dynamics on the current intensity and direction. The dynamics of the v-av matter is characterized by a series of creation and annihilation processes, which reflect on the time dependence of the electrical field, and by guided motion, resulting in a zero-field transverse resistance. ###### pacs: 74.25.Qt,74.78.Na A small magnetic dot placed in the vicinity of a thin superconducting film has the ability of creating vortex-antivortex (v-av) pairs in the superconductor and keep them from annihilating each other.Erdin02 ; Milosevic02 Thanks to this, phases of vortex matter in which vortices and antivortices coexist can be stabilized in bilayers made of thin superconducting films and arrays of nanomagnets, even in the absence of macroscopic magnetic fields. This topic has attracted a great deal of attention lately. Several calculations using the Ginzbug-Landau approach have been carried out for arrays of polarized magnetic dots on top of a superconducting film, predicting a great variety of equilibrium phases of vortices and anti-vortices.Priour03 ; Milosevic0405 Most early experimental work on superconductor-nanomagnet hybrids were aimed at the flux pinning properties of the magnetic dots.Schuller ; VanBael ; Villegas08 Recently, experiments on superconducting films with polarized magnetic dot arrays corroborate the existence of vortex-antivortex phases in these structures and show that such phases induce novel phenomena such as the field-induced superconductivity and the zero-magnetic-field ratchet effect.Lange03 ; Souza07 A problem that has received much less attention is the existence in these systems of dynamical states of vortex matter in which vortices and antivortices coexist. These states may result when a transport current is applied to the film, and the v-av matter moves. In this paper we study this problem theoretically. We expect that during the motion v-av pairs annihilate and that new pairs are created. The fundamental question is how these processes, together with vortex-vortex and vortex-nanomagnet interactions, determine the properties of the moving vortex matter. Here we address this point by using numerical simulations of a simple model for vortices interacting with a periodic array of permanent point dipoles. The details of our model is as follows. The bilayer comprises a thin superconducting film of thickness $d$, penetration depth $\lambda\gg d$, and coherence length $\xi<d$. A square array of magnetic dipoles, with magnetic moment ${\bf m}$ and unit cell of dimensions $a_{p}\times a_{p}$ (Fig. 1), Figure 1: (Color online) (a) Schematic view of the in-plane dipole array. (b) Definition of the transport current J and driving force F orientations. is placed above the film, at a distance $z_{0}>d/2$ from it. Experimentally, the space between the superconducting film and the dipole layer is usually filled with an insulating buffer layer to prevent proximity effects. Here we assume that the dipoles are polarized in plane, parallel to the $x$ direction (${\bf m}=m\hat{\bf x}$). The total energy of the vortex matter is given by $E=\sum_{i,j}\,q_{i}q_{j}\,\Big{[}\frac{1}{2}U_{\rm vv}({\bf r}_{ij})+E_{c}\delta_{i,j}\Big{]}+\sum_{j}q_{j}U_{\rm vm}({\bf r}_{j}),$ (1) where $q$ is the vorticity ($q=1$ for vortices, $q=-1$ for antivortices). Here we assume that no external magnetic field is applied. In other words, the vortex-matter “neutrality” condition, $\sum_{j}q_{j}=0$, must be satisfied. $U_{vm}({\bf r})$ is the usual interaction energy between a vortex located at ${\bf r}$ and the dipole array calculated in the London limit.Carneiro05 The interaction energy $U_{\rm vv}({\bf r}_{ij})$ between an arbitrary pair of vortices at positions ${\bf r}_{i}$ and ${\bf r}_{j}$ is calculated following Clem’s variational approach for the order parameter (${\bf r}_{ij}={\bf r}_{i}-{\bf r}_{i}$).Clem74 In the limit $\lambda\gg d$ considered here, this gives $U_{\rm vv}({\bf r})=\epsilon_{0}d\ln(\|{\bf r}_{ij}\|^{2}+2\xi_{c}^{2})$, where $\epsilon_{0}=\phi_{0}^{2}/4\pi\mu_{0}\lambda^{2}$ and $\xi_{c}$ is a variational parameter (in general, $\xi_{c}\sim\xi(T)$). The vortex core energy can be calculated straightforwardly by using Clem’s trial order parameter to compute the gain in condensation energy due to an isolated vortex, which results in $E_{c}=0.375\epsilon_{0}$. $-E_{c}$ acts as a chemical potential of the vortex system and thus plays an important role in the nucleation of vortex-antivortex pairs.MinhagenRMP Our numerical algorithm is based on the assumption that the time evolution of the vortex matter can be broken into two distinct parts: motion of the vortices and antivortices, and creation of v-av pairs. The first is governed by the Langevin equations $\eta\frac{d{\bf r}_{j}(t)}{dt}=-{\bf{\nabla}}_{j}E+{\bf F}_{j}+{\bf{\Gamma}}_{j}\;,$ (2) where $\eta$ is the Bardeen-Stephen friction coefficient, $j$ runs over all vortices and anti-vortices present at time $t$, ${\bf F}_{j}=(q_{j}\phi_{0}/c){\bf J}\times\hat{\bf z}$ is the driving force, and ${\mbox{\boldmath$\Gamma$}_{j}}$ is the random force appropriate to temperature $T$. ${\bf J}$ is the applied current density which makes and angle $\alpha$ with the positive $y$-axis, as shown in Fig. 1(b). These equations are solved numerically by a finite difference method. Eventually, a vortex and an antivortex collide and annihilate each other. This is taken into account by assuming that such a collision occurs every time the vortex and the antivortex are separated by a distance $\xi_{c}$. The procedure for the second part, v-av pair creation, is to freeze the vortex matter configuration after every Langevin dynamics time step and create v-av pairs using a Monte Carlo procedure. This assumes that the actual nucleation process takes place in a time interval so short that the motion of the vortex matter during it can be neglected. Trial v-av pairs are then placed at every point of an auxiliary grid, with unit cell of dimensions $a_{c}\times a_{c}$ ($a_{c}\simeq\xi_{c}$), commensurate with the dipole array. The vortex and antivortex are placed at a distance $\xi_{c}$ from each other at a given axis, with the pair center of mass fixed at the grid point. At high $T$, this axis is chosen randomly, whereas at low $T$ it is determined by the direction of the total current density calculated at the grid point (at that instant of time) as to favor pair unbinding. Then, a Metropolis code is used to either accept or reject the pair. The energy entering this algorithm includes the energy due to the interaction with the transport current, that is $E_{T}=E-\sum_{j}{\bf F_{j}}\cdot{\bf r}_{j}$. Such a procedure accounts for both deterministic (induced by the local current distribution) and thermal pair nucleation. Unpub The simulations are carried out on a $L\times L$ section of the bilayer, with $L=8a_{p}$, and assuming periodic boundary conditions in both $x$ and $y$ directions. The time step used in the numerical integration is $dt=2\cdot 10^{-4}t_{0}$ ($t_{0}=\phi_{0}\epsilon_{0}/(\eta{a_{p}})$) and we chose $z_{0}=0.2a_{p}$. Equilibrium configurations of the vortex matter for small $T$ and ${\bf J}=0$, were obtained minimizing $E$ (Eq. 1) by means of a simulated annealing scheme. Fig. 2 Figure 2: (Color Online) (Top) Vortex(black)-antivortex(red) configurations for (a) $m=1.0\phi_{0}z_{0}$ ($n_{pair}=1$), (b) $m=1.65\phi_{0}z_{0}$ ($n_{pair}=2$) and (c) $m=2.25\phi_{0}z_{0}$ ($n_{pair}=3$). (Bottom) Number of v-av pairs per dipole, $n_{pair}$, versus magnetic moment of the dipoles, $m$, for different values of $\xi_{c}$. shows the equilibrium states of the superconducting film obtained by this method. The bottom panel presents the number of vortex-antivortex pairs per dipole ($n_{pair}$) as a function of the magnetic moment of the dipoles for several values of $\xi_{c}$. These curves are characterized by plateaus at integer $n_{pair}$. Examples of stable vortex-antivortex configurations for $\xi_{c}=7\times 10^{-2}a_{p}$ are depicted in the upper panels. Notice that by increasing $\xi_{c}$ (for instance, by increasing the temperature) the magnetic moment necessary for the creation of a new v-av pair per dipole decreases. This is consistent with the fact that the force necessary to separate a v-av pair decreases with $\xi_{c}$. Therefore, for lower $\xi_{c}$ values, a strong magnetic moment is necessary to hold the pair from annihilating each other. A similar temperature dependence of the critical magnetic moment has been also observed in numerical calculation for magnetic dot arrays in the off-plane geometryMilosevic0405 . Now we discuss the response of the vortex matter to an applied current flow. For simplicity, we shall consider hereafter only the cases where each dipole stabilize one v-av pair in equilibrium, by assuming $\xi_{c}=7\times 10^{-2}a_{p}$ and $m=1.0\phi_{0}z_{0}$. Fig. 3 (a) Figure 3: (Color online) Time averages of the $x$ and $y$ components of the electric field (top) and of the v-av pair density (Bottom) as functions of the current density applied at different orientations $\alpha$. Inset: angular dependence of the critical current, $J_{c}(\alpha)$. presents the $x$ and $y$ components of the mean electric field induced by vortex motion, expressed by the time average of ${\bf E}=\phi_{0}\sum{q_{i}\dot{{\bf r}_{i}}}\times\hat{z}$, as a function of the current density J (in units of the depairing current $J_{d}$) for different orientations of J. Averages were taken over $15\cdot 10^{6}$ time steps after a relaxation waiting time of $5\cdot 10^{6}$ time steps. The $E_{x}(J)$ and $E_{y}(J)$ curves evidence a strong dependence of the v-av dynamics on the current orientation $\alpha$. Such anisotropy becomes clearer when one plots the critical current $J_{c}$ as a function of $\alpha$. Vortices are depinned more easily when J is at an angle $\alpha=180^{\circ}$. At this direction, ${\bf J}=J\hat{y}$ induces a Lorentz force on the vortices (antivortices) parallel to $+\hat{x}$ ($-\hat{x}$) in such a way as to separate the v-av pairs. Such separation is favored by the supercurrent generated right below each dipole, which is also parallel to $-\hat{y}$ and is responsible for the stabilization of the v-av pairs. At $\alpha=90^{\circ}$, on the other hand, the applied current has to overcome alone the v-av mutual attraction, with no help from the dipole-induced supercurrent. For $\alpha<90^{\circ}$, the component $J_{y}$ inverts sign and tends to bind the v-av pair. At $\alpha=75^{\circ}$, $J_{x}$ is still strong enough to unbind the pairs. At lower $\alpha$, however, all v-av pairs annihilate each other and the current necessary to create new pairs and stablish a dynamical state of moving v-av matter is close to $J_{d}$. At such current range, our model is no longer valid and therefore $J_{c}$ could not be estimated for $\alpha<75^{\circ}$. Fig. 3 (b) presents the mean v-av pair density $\langle n_{pair}\rangle$ as a function of J. Interestingly, for all angles studied the onset of v-av matter motion is accompanied by an increase in $\langle n_{pair}\rangle$ with respect to the static (equilibrium) value $\langle n_{pair}\rangle=1$. Here we analyze in more detail the dynamics of vortices and antivortices in the film and how it reflects on the macroscopic electrical response and the mean pair density. We shall restrict ourselves to four directions, $\alpha=180^{\circ}$, $\alpha=150^{\circ}$, $\alpha=120^{\circ}$ and $\alpha=90^{\circ}$. Fig. 4 Figure 4: Time series of the electric field $E_{x}$, $E_{y}$ and v-av pair density $n_{pair}$ for $\alpha=90^{\circ}$ (a), $120^{\circ}$ (b), $150^{\circ}$ (c) and $180^{\circ}$ (d). present the time evolution of the number of v-av pairs per dipole and the $x$ and $y$ components of the mean electric field for fixed ${\bf J}$ values. For $J>J_{c}(\alpha)$, we observed steady states of the moving v-av matter characterized by an oscillatory time dependence of $E_{x}$ and $E_{y}$, even though the applied current is constant. For $\alpha=90^{\circ}$, $n_{pair}(t)$ rapidly reaches 1.5 and keeps essentially constant at this value. The corresponding steady state is illustrated in Fig. 5(a), Figure 5: (Color online) (a) Time-averaged vortex density for $J=0.5J_{d}$ and different current orientations $\alpha$. Light (dark) shades correspond to vortices (antivortices). The circles indicate creation ($\star$) and annihilation (${\dagger}$) of v-av pairs. (b) Electric field direction $\theta$ as a function of $\alpha$ for several driving current intensities (symbols). The $\theta=\alpha$ line is shown for comparison. The cartoons illustrate the dominating v-av motion for $\theta\sim 180^{\circ}$ and $\theta\sim 90^{\circ}$. where we present contour plots of the time-averaged vortex-antivortex density, defined as $\rho({\bf r})=\langle\sum_{i}q_{i}\delta({\bf r-r}_{i}(t))\rangle_{t}$. The average is taken at a fixed ${\bf J}$ over $N_{S}$ time steps. The trajectories of vortices (antivortices) are given by the light (dark) color shades. As it is clear, vortices and antivortices follow distinct tracks, perpendicular to the applied current, and collide very rarely. For $\alpha>90^{\circ}$, the dynamical steady states are much reacher. $n_{pair}(t)$ is also oscillatory, indicating a cyclic series of creation and annihilation of v-av pairs. The general picture is as follows: when a current $J>J_{c}(\alpha)$ is applied, the vortices and antivortices stabilized by the dipoles in the equilibrium ($J=0$) state, which we shall refer to as first- generation vortices, depin and move apart from each other, leaving space for the creation of new (second-generation) v-av pairs right below the dipoles. At this point, $n_{pair}(t)$ reaches twice its original value ($n_{pair}(t)=2$). These new pairs, give an additional push to the preexistent ones, inducing a sudden increase in $\|{\bf E}\|$. All first-generation vortices and antivortices run towards annihilation, which will take place at the midpoint between first- neighbor dipoles, for $135^{\circ}\leq\alpha\leq 180^{\circ}$, or at the face center of the dipole array unit cells, for $105^{\circ}\leq\alpha<135^{\circ}$, as revealed by the $\rho({\bf r})$ plots presented in Fig. 5(a). Just before annihilation, however, new v-av pairs spawn at the above mentioned positions and subsequently annihilate the first- generation vortices and antivortices. We believe these extra pairs arise from the fact that our model assumes a rigid vortex core. At distances of order $\xi$, the v-av annihilation in superconductors is characterized by strong vortex core deformation. In our model, the extra v-av pairs appear naturally as a way of compensating the vortex core rigidity and accelerate the annihilation process. As a side effect, a spurious, high-frequency noise arises in $E_{x}(t)$ and $E_{y}(t)$. Notwithstanding, we have checked that such noise has only a neglectable influence on the macroscopic quantities shown in Fig. 3. Finally, after the annihilation process, the second- generation v-av pairs become first-generation ones and the whole process repeats. Notice that the lifetime, $t_{\rm L}$, for a vortex of the first or second generation is always larger than the oscillation period $P$ (see Fig. 4). The angular dependence of the v-av dynamics described above suggests that motion occurs preferentially along the $x$ directions for $\alpha$ close to 180∘. Such guided motion of the v-av matter should reflect directly on the electric field E generated by vortex motion, giving rise to a deflection of E with respect to J and thus inducing transverse resistance. To demonstrate this statement, we plot the direction of ${\bf E}$, $\theta$, as a function of the direction of ${\bf J}$, $\alpha$, for several values of $J$ (left panel of Fig. 5). Indeed, $\theta(\alpha)$ diverges considerably from the $\theta=\alpha$ line. For $\alpha$ close to $180^{\circ}$, $\theta$ tends to lock at $\theta\simeq 180^{\circ}$ in the whole $J$ range, whereas for $\alpha=75^{\circ}$, E locks at $\theta=90^{\circ}$. This transverse resistance resulting from the guided motion of v-av matter is essentially different from that observed in superconducting films with non-magnetic pinning arraysguidance and other Hall-like effects in the sense that here no macroscopic magnetic field is required. In summary we have studied the dynamics of vortices and antivortices in superconducting films interacting with an array of in-plane magnetic dipoles by means of a model which explicitly takes into account creation and annihilation of v-av pairs. The current-driven dynamics of the vortex- antivortex matter at zero applied field is characterized by a series of creation and annihilation processes producing an oscillatory behavior of the time dependent electric field. Other predictions of our model include the anisotropy in the critical current and the transverse electric field induced by vortex-antivortex guidance. These results can be checked out experimentally by means of conventional transport measurements at zero applied magnetic field or by direct high-resolution magneto-optical imaging. We anticipate that the zero-field guidance effect should also occur for other dipole polarizations, such as the off-plane geometry, as long as the magnetic texture captures v-av pairs. ###### Acknowledgements. The authors are grateful to Gilson Carneiro for fruitful discussions. Research supported in part by the Brazilian agencies CNPq and CAPES. ## References * (1) S. Erdin et al., Phys. Rev. B66, 014414 (2002). * (2) M.V. Milosevic, S. V. Yampolskii, and F.M. Peeters, Phys. Rev. B66, 174519 (2002). * (3) D. J. Priour, Jr. and H. A. Fertig, Phys. Rev. Lett. 93, 057003 (2004). * (4) M.V. Milosevic and F. M. Peeters, Phys. Rev. Lett. 93, 267006 (2004); 94, 227001 (2005). * (5) J. I. Martin et al., Phys. Rev. Lett. 79, 1929 (1997); A. Hoffmann, P. Prieto, and I. K. Schuller, Phys. Rev. B61, 6958 (2000). * (6) M. J. Van Bael et al., Phys. Rev. B59, 14674 (1999); M. J. Van Bael et al., Phys. Rev. Lett. 86, 155 (2001); * (7) J. E. Villegas et al., Phys. Rev. B77, 134510 (2008). * (8) M. Lange et al., Phys. Rev. Lett. 90, 197006 (2003). * (9) C. C. de Souza Silva et al., Phys. Rev. Lett. 98, 117005 (2007). * (10) G. Carneiro, Phys. Rev. B72, 144514 (2005). * (11) J.R. Clem, in: K.D. Timmerhaus, W.J. O’Sullivan, and E.F. Hammel (Eds.), Low Temperature Physics, vol. 3, Plenum, New York, 1974, p. 102. * (12) P. Minhagen, Rev. Mod. Phys. 59, 1001 (1987). * (13) C. L. S. Lima, C. C. de Souza Silva, and G. Carneiro (private communication). * (14) A. V. Silhanek et al., Phys. Rev. B68, 214504 (2003); J. E. Villegas et al., Phys. Rev. B68, 224504 (2003)	arxiv-papers	2008-08-18T16:39:49	2024-09-04T02:48:57.348556	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Cl\\'essio L. S. Lima (1 and 2) and Cl\\'ecio C. de Souza Silva (1).\n ((1) Departamento de F\\'isica, Universidade Federal de Pernambuco, (2)\n Instituto de F\\'isica, Universidade Federal do Rio de Janeiro)", "submitter": "Clessio Lima Leao", "url": "https://arxiv.org/abs/0808.2421" }
0808.2506	Physics and Astrophysics of Planetary Systems # Telescopes versus Microscopes: the puzzle of iron-60 Jonathan Williams Institute for Astronomy, University of Hawaii, Honolulu, USA; jpw@ifa.hawaii.edu ###### Abstract The discovery that the short-lived radionucleide 60Fe was present in the oldest meteorites suggests that the formation of the Earth closely followed the death of a massive star. I discuss three astrophysical origins: winds from an AGB star, injection of supernova ejecta into circumstellar disks, and induced star formation on the boundaries of HII regions. I show that the first two fail to match the solar system 60Fe abundance in the vast majority of star forming systems. The cores and pillars on the edges of HII regions are spectacular but rare sites of star formation and larger clumps with masses $10^{3-4}$ $M_{\odot}$ at tens of parsec from a supernova are a more likely birth environment for our Sun. I also examine $\gamma$-ray observations of 60Fe decay and show that the Galactic background could account for the low end of the range of meteoritic measurements if the massive star formation rate was at least a factor of 2 higher 4.6 Gyr ago. ## 1 Introduction The study of planet formation can be approached from two sides: the large scale encompassing the molecular core that collapses to a star and surrounding planetary disk, and the small scale in which the planets and the remnants of their formation, meteorites, are examined in detail to learn about the conditions of the early solar system (ESS). Astronomers have identified each of the main stages by which the interstellar medium (ISM) becomes molecular, fragments, and individual cores collapse to protostars. We have imaged the disks of planet forming material around young stars and witnessed the debris from planetesimal collisions. Although many details remain to be worked out, the basic properties of a proto-planetary disk, such as mass, size, composition, and lifetime, are well characterized. We have also identified over 200 extrasolar planetary systems around nearby stars. The surprising diversity of these systems, however, raises the vexing issue of how typical is our solar system. Cosmochemists have identified the oldest rocks in the solar system and determined their age, 4.567 Gyr, to astonishing precision. Careful study of their mineralogy shows the detailed conditions, including, for example, thermal history, radiation field, and transport processes, in the ESS. As with cosmologists studying the Universe, however, there is one and only one system at hand which raises the vexing issue of how typical is our solar system. The areas where these two different lines of inquiry, telescopic and microscopic, overlap provide interesting comparisons. Short lived radionucleides (SLR), with half-lives less than 3 Myr (see below), are of particular interest for understanding planet formation and the astrophysical environment of the ESS. In this chapter, I focus on the puzzle presented by the presence of 60Fe in the most primitive meteorites. As I show, it may be hard to reconcile the astronomical and cosmochemical pictures but it is important to try since it may show the limitations in our knowledge or understanding and it may also help show whether our solar system is indeed typical, or not. I begin by reviewing the astronomical and cosmochemical background to this subject. These sections are brief as they are covered in other chapters in this book. I discuss three different scenarios which have been proposed for the delivery of SLR into the ESS and assess the probability for incorporation of 60Fe. I conclude that the only viable mechanism that might apply to a large number of planetary systems is the rapid collapse of large cluster forming clumps neighboring a massive star forming region. Finally I return to a basic assumption that the 60Fe abundance is greater than the Galactic background and show a discrepancy between the predicted level and recent $\gamma$-ray observations. ## 2 Astronomical observations of star and planet formation Stars form in molecular clouds, strung out along the spiral arms of the Galaxy. Our current understanding of the processes by which these large and massive clouds condense to stellar scales was recently summarized by McKee & Ostriker ([2007]). A rotationally supported disk inevitably accompanies the protostar due to conservation of angular momentum and the magnification of any initial spin. These disks initially funnel material onto the growing star but subsequently become the sites of planet formation. The observational properties and the theory behind the coagulation of sub-micron sized ISM dust grains to planetesimals are reviewed in the chapters by Hogerheijde and Youdin. For the purposes of the discussion here it is important to note that several states, or phases, of gas coexist in the ISM (Cox [2005]). The lowest density state is a hot ionized phase produced by supernovae and this fills most of the volume. Most of the mass is in atomic clouds at intermediate temperatures. The highest density state is cold and molecular. Most stars form in giant molecular clouds with typical sizes111I use $\sim$ to indicate an order of magnitude variation around a value and $\approx$ for a factor of 3., $L\approx 50$ pc and masses, $M\sim 10^{5}$ $M_{\odot}$. The clouds possess considerable substructure, generically characterized as clumps with sizes $L\approx 5$ pc and masses, $M\sim 10^{3}$ $M_{\odot}$ (Williams, Blitz, & McKee [2000]). Individual stars form in dense cores within clumps, with typical radii $R\approx 0.05$ pc and mass $M\approx 1$ $M_{\odot}$ (di Francesco et al. [2007]). Protostellar disks have radii $R\approx 100$ AU and masses ranging from $M\approx 10^{-3}-10^{-1}$ $M_{\odot}$ (Andrews & Williams [2007]). Although many observational and theoretical studies are directed toward the formation of isolated, individual stars, as this is the simplest system to understand, most stars form in large clusters (Lada & Lada [2003]). The stellar mass distribution is heavily skewed toward the low end in terms of numbers but the few massive stars dominate the luminosity and their radiation, winds, and eventual supernovae may significantly affect the properties of neighboring primordial planetary systems. The timescales are also critical for our comparison with cosmochemistry. It is surprisingly difficult to identify young or old clouds (or even tell the difference – see Williams & Maddalena [1996]) and consequently the lifetimes of giant molecular cloud are highly uncertain. Current estimates, based on theories of cloud formation and destruction, range from $\approx 1-20$ Myr (Hartmann, Ballesteros-Paredes, & Bergin [2001]; Matzner [2002]). The evolutionary state of cores within star forming clouds are easier to characterize and several lines of evidence suggest that as soon as they become sufficiently dense, $\mbox{$n_{\rm H_{2}}$}\sim 10^{5}$ ${\rm cm^{-3}}$, stars will rapidly form within a few free-fall timescales, $\sim 10^{5}$ yr (Jorgensen et al. [2007]). The presence of a circumstellar disk around a star is most readily detected as a long wavelength excess above the stellar photosphere. Large surveys of clusters with different ages show that the fraction of stars with disks decreases from an initial value close to unity to zero by by 6 Myr, and that the median disk lifetime is about 3 Myr (Haisch, Lada, & Lada [2001]; Hernandez et al. [2008]). This is the characteristic timescale for the formation of planetesimals although the last steps in the growth of fully fledged planets may take considerably longer (see chapters by Kalas and Beichmann). ## 3 Short lived radionucleides and the importance of 60Fe Astronomers are limited to studying the light that molecular cores, young stars, and dusty disks either emit naturally or absorb and scatter from a background or nearby source. Having rocks in the laboratory, however, allows for far more targeted and detailed investigations. Cosmochemists can break down individual grains within meteorites to the atomic level and measure their structure, mineralogy, and isotopic composition to learn about the precise conditions in which they formed. An element can have unstable isotopes that are chemically identical but decay over cosmic timescales. A dust grain with a particular mineralogy can therefore be incorporated into a planetesimal but subsequently change its composition. These isotopic anomalies are then frozen into the material and, unless the planetesimal undergoes further processing, are immutable. SLR, defined here as having half-lives less than the 3 Myr characteristic lifetime of disks, therefore provide a natural chronometer for planet formation. I refer the reader to the chapters by Aleon and Gounelle for more detail and the broader aspects of such work. 60Fe decays to 60Ni with a half life of 1.5 Myr and its presence in the ESS was inferred by the correlation222 If 60Fe had not been present, the nickel ratio would have been independent of iron. of the isotope ratio, 60Ni/61Ni, with the main isotope of iron, 56Fe/61Ni (Tachibana & Huss [2003]). The exact abundance of 60Fe in the ESS is not yet firmly established, with published measurements ranging from [60Fe]/[56Fe]$=3-10\times 10^{-7}$ (Tachibana et al. [2006]; Moynier et al. [2005]) but it appears to be greater than the expected background level. The Galactic background is set by a balance between production in massive stellar winds and supernovae and destruction by radioactive decay. The equilibrium abundance of any radionucleide, normalized by the ratio of production rates, is then a simple fuction of its decay time (Schramm & Wasserburg [1970]). Jacobsen ([2005]) plots the normalized abundances of many radionucleides with half-lives from 0.1 Myr to almost 10 Gyr against the expected background. Solar system levels are generally lower than the ISM average but most radionucleides with half-lives greater than 3 Myr fit a model with a slow exchange, over a $\approx 60$ Myr timescale, between the hot ionized medium and cold, molecular clouds. The background level decreases rapidly with decay time and many SLR lie above this model, however. The origin of radionucleides with abundances above the background can be broadly categorized in two ways: production by energetic particles from the active protosun and rapid transport from an external source into the ESS (Wadhwa et al. [2007]). Both mechanisms have considerable flexibility and can be adjusted to match the abundances of most SLRs. Note that these are not mutually exclusive and both may have played a role. For instance, 10Be can only be produced by spallation reactions, and is strong evidence for disk processing by protosolar radiation. On the other hand, the neutron rich iron isotope, 60Fe, can only be formed in the cores of massive stars and must have then been delivered into the ESS on a timescale comparable to its half life. The possibility of an external influence on the Sun’s formation predates the discovery of 60Fe (see Cameron & Truran [1977] and references therein) but the success of local irradiation models (e.g. Lee et al. [1998]; see also the chapter by Gounelle in this volume) allowed most astronomers to downplay the issue. The importance of 60Fe is that it is most clear-cut example of an SLR that could not have been produced by the protosun and therefore constitutes indisputable evidence that the products of nucleosynthesis from the core of massive star polluted the protosolar nebula. The puzzle of 60Fe is that astrophysical contexts in which this occurs do not appear to be common. We are faced with either a significant gap in our understanding of star and planet formation or the realization that our solar system formed in a very unusual situation. ## 4 Possible origins of 60Fe in the early solar system ### 4.1 Pollution of the protosolar nebula by an AGB star Asymptotic Giant Branch (AGB) stars are evolved low and intermediate mass stars that have used up all the hydrogen in the central core and are now burning hydrogen and helium in shells around it. Mixing in these shells dredges up newly synthesized material from the core which may be blown out in a powerful stellar wind. In principle, SLRs with abundances comparable to ESS levels can be injected into the ISM through AGB winds (Cameron [1993]). Figure 1 shows a wind-blown bubble from an evolved massive star in a star-forming molecular cloud; a simultaneous demonstration that this situation can occur but that its effect is limited to a very small region of the cloud. In regard to the 60Fe problem discussed here, its production in an AGB star requires faster reactions than that required for 26Al. Wasserburg et al. ([2006]) show that only stars with masses $\lower 2.15277pt\hbox{$\;\buildrel>\over{\sim}\;$}5$ $M_{\odot}$ produce enough 60Fe to match the ESS abundance. Stars with masses $>10$ $M_{\odot}$ end their lives as supernovae and the possibility that they are the origin of the 60Fe is discussed below. The stellar mass range of interest for the AGB hypothesis is therefore $5-10$ $M_{\odot}$. By their nature, AGB stars have ages a little greater than the main sequence lifetime appropriate to their mass, $\sim 25-110$ Myr (Schaller et al. [1992]). The molecular cloud in which they were born will have dispersed and any connection to ongoing star and planet formation will be serendipitous. Based on a census of the known mass-losing AGB stars and molecular clouds within 1 kpc of the sun, Kastner & Myers ([1994]) estimated the probability of a chance encounter to be $\sim 1$% per Myr. However, the ratio of AGB wind to molecular mass is very small, $3\times 10^{-4}$, and they concluded – in the context of 26Al – that the probability that the ESS was polluted in this way was $\leq 3\times 10^{-6}$. Figure 1: The relative scale of AGB ejecta compared to molecular clouds. The two panels are from the Digitized Sky Survey and show optical extinction and nebulosity associated with the IC1318 star forming region in the left panel. The right panel shows a close-up of the wind-blown bubble from the AGB star, HD 192163. Note the relative scales that graphically demonstrates how small a part of a molecular cloud can be polluted by a massive star wind. The Kastner & Myers work is an often cited argument against an AGB stellar origin for 26Al in the ESS. The same reasoning applies to 60Fe with an even lower likelihood due to the higher stellar masses necessary. Their result was based on the catalogs of AGB stars and molecular clouds in the solar neighborhood (1 kpc), however, and it is worth making a related but more general case for the entire Galaxy. Jura & Kleinmann ([1990]) find that the average surface density of AGB stars with high mass loss rates is $\sim 10$ kpc-2, independent of Galactic radius. This implies a total population of about $3\times 10^{3}$ in the Galaxy. Each lasts for about 1 Myr, a timescale comparable to the 60Fe half-life, and ejects a total of about 3 $M_{\odot}$ into the ISM. Based on the predicted wind SLR abundances, these winds can pollute about 100 times more mass to ESS levels or $\sim 10^{6}$ $M_{\odot}$ in total. The total molecular mass in the Galaxy is $1\times 10^{9}$ $M_{\odot}$ (Williams & McKee [1997]) so the proportion of star forming material that can be enriched is no more than 0.1%. This assumes that the star formation efficiency is independent of the presence of an AGB star which is entirely consistent with observations. Note that this is a generous upper limit since it does not include the low likelihood that an AGB star lie within or near a molecular cloud as in Kastner & Myers, nor does it include SLR decay during the (potentially $\gg 1$ Myr) passage from AGB wind to planetesimal. Finally the Jura & Kleinmann AGB number count likely includes many stars with masses too low to produce appreciable 60Fe. Even without these additional factors, however, this simple calculation provides a robust and independent demonstration that the pollution of the proto-solar system by the winds from an AGB star is a very unlikely scenario for the origin of 60Fe in the ESS. ### 4.2 Injection of supernova ejecta into the protoplanetary disk Circumstellar disks can survive a supernova explosion as close as 0.2 pc and potentially capture significant levels of SLR from the ejecta (Chevalier [2000]). Subsequent numerical simulations by Ouellette, Desch, & Hester ([2007]) show that solid grains can be efficiently mixed into the disk material. The mechanism works but how often does it occur? A fundamental constraint is the similarity in the timescales for massive star and disk evolution (Figure 2). Even the most massive stars take 3 Myr to burn their hydrogen and evolve off the main sequence, at which point half of the disks around neighboring stars have disappeared. Further, only stars with masses greater than 30 $M_{\odot}$ explode within 6 Myr, the maximum disk lifetime. Stars this massive are extremely rare and only found in the largest clusters. Figure 2: Disk lifetimes versus main-sequence stellar timescales. The solid diagonal line decreasing from 100% to 0% represents the decrease in the disk fraction in clusters of varying age and is taken from Haisch et al. ([2001]). The curved dashed line plots the percentage of stars with main sequence lifetimes from 3 to 10 Myr. The hashed areas show that there are no supernovae before 3 Myr and no disks remain after 6 Myr. This leaves only a small range in time, $3-6$ Myr, when there are both disks and a supernova. Moreover, only stars with masses greater than 30 Msun satisfy this constraint and they are extremely rare, numbering only about one in $10^{4}$ stars. An even stronger constraint comes matching the abundances of the supernova ejecta with the meteoritic record. The amount of captured material is equal to the product of the disk area and the surface density of the supernova ejecta. The abundance of a particular SLR is therefore proportional to the supernova yield times the square of the ratio of the disk radius to the distance from the source. Ouellette et al. ([2007]) and Looney, Tobin, & Fields ([2006]) show this criterion requires the disk be within a “radioactivity distance” of 0.3 pc of a supernova with progenitor mass $10-40$ Msun. This is only slightly greater than the survival distance. The disk injection scenario therefore only works over a very narrow range of distances and this makes it extremely unlikely. In Williams & Gaidos ([2007]), we explicitly calculate the disk injection likelihood, taking into account the massive star lifetimes, disk evolution, and the constraints on distance from the source. We considered a cluster with $N_{}$ stars sampled from the initial mass function (IMF), $dN_{}/dM_{}\propto M_{}^{-2.5}$ (Scalo 1986), and calculate the most likely supernova progenitor mass. This defines a supernova timescale and a corresponding disk fraction (see Figure 2). To take the disk-supernova distance into account, we model the cluster as expanding linearly with time and parameterize the expansion via the surface density at 3 Myr, an observationally defined quantity. We also extended the enrichment range out to 0.4 pc to allow for more massive supernovae than considered by Ouellette et al. and Looney et al. The results are shown in the left panel of Figure 3. The probability that a disk is injected peaks at $0.5-1.5$%, depending on stellar density, at $N_{}\simeq 10^{4}$. Small clusters are unlikely to harbor a supernova that explodes before all the disks have disappeared and most disks are too far from the supernova to be enriched in large clusters. In fact, the enrichment fraction is the dominant factor in the low overall probability; clusters with $10^{4}$ stars have about about a 50% chance of containing a high mass star that will become a supernova within 4 Myr implying a disk fraction of 1/3 at the time of the blast, but the cluster will be several parsecs in radius and only a few percent of disks will be close enough to the supernova to be enriched. Figure 3: Probability of enrichment of a protostellar disk by a supernova versus cluster size. In each panel three curves are plotted corresponding to different expansion rates, parameterized by the stellar number density, $\Sigma_{3Myr}$. The shading shows the range of probabilities over cluster formation timescales from 0 to 3 Myr. The left panel shows the disk enrichment likelihood for a single supernova event. Small clusters are unlikely to have a supernova before all disks have dissipated and most disks in large clusters are too far away from the supernova to be enriched to ESS levels. The right panel shows the disk enrichment probability in the case where multiple supernovae can enrich multiple disks. In this case, the probability rises strongly in large clusters where many supernovae occur (see inset) and many disks may be impacted. It is also possible to relax an implicit assumption in the model that all the stars in the cluster form at once. This does not greatly change the disk injection probability. The greyscale associated with each surface density in Figure 3 shows the range of probabilities for a cluster formation timescale varying from $0-3$ Myr. Except for the largest clusters, $N_{}\lower 2.15277pt\hbox{$\;\buildrel>\over{\sim}\;$}10^{5}$ stars, the probability decreases with formation time since the rarer, massive stars will likely form after many low mass stars and their disks will have an evolutionary head start. This is for an unbiased sampling of the IMF. The probability can be increased by putting in a bias toward forming high mass stars first (equivalent to shifting the stellar curve leftward in Figure 2, but there is no observational evidence for this. Young high mass stars are always surrounded by many low mass stars and no O star has been observed in isolation at less than 1 Myr. Large clusters can host many supernovae and these can enrich multiple disks. In this case, and generously assuming that the low and high mass stars are spatially mixed333 In practice, high mass stars tend to lie together at the cluster center and their combined impact on the other stars is reduced., the right panel in Figure 3 shows that the disk injection probability rises substantially to $\approx 10$% in the largest known clusters, $N_{}=5\times 10^{5}$. However, most stars form in smaller clusters (actually an equal number per logarithmic bin) and the overall disk injection probability, for any given star in the Galaxy, is very small, $\lower 2.15277pt\hbox{$\;\buildrel<\over{\sim}\;$}1$%. Nevertheless, this shows that if the proto-solar system disk was impacted by a supernova blast, it most likely happened in an enormous cluster, similar in size or greater than W49 or $\eta$ Carinae today. This simple model can be expanded in a number of ways but it is hard to escape the fact that very few disks are enriched with SLR by direct incorporation of supernova ejecta. Gounelle & Meibom ([2008]) include an additional factor to allow for external photoevaporation of the disk by massive stars in the above formalism. This reduces the disk injection probability further, particularly in the most massive clusters where it negates the effect of multiple supernovae. Certainly massive stars can rapidly erode the outer parts of disks, as observed in the Trapezium Cluster in Orion (O’dell & Wen [1993]) but it is not yet clear that this prevents planet formation in the inner parts. Indeed Throop & Bally ([2005]) postulate that the preferential removal of gas might enhance dust sedimentation and speed up the growth of planetesimals. Williams, Andrews, & Wilner ([2005]) show that enough mass remains in several Trapezium Cluster disks to form solar system scale architectures but more observations are required to show the statistics and significance of photoevaporation on disk mass (Mann & Williams, in prep). On the other hand, mid-infrared observations by Balog et al. ([2007]) and Hernandez et al. ([2008]) show marginally significant evidence for a decrease in the inner disk fraction of about a factor of 2 within 0.5 pc of massive stars. Our model shows that the disk injection probability increases for higher cluster surface densities. We bracketed a range, $\Sigma_{}\simeq 30-300$ stars pc-2, that is observed in infrared surveys (Adams et al. [2006], Carpenter et al. [2000]). This is also consistent with Lada & Lada ([2003]) who show that the number of detectable clusters declines with age and estimate that only about 10% survive as recognizable entities beyond 10 Myr (i.e. with surface densities significantly above the field star background). Our Sun could not have formed in a long-lived cluster, however, because the solar system would have been disrupted by stellar encounters (Adams & Laughlin [2001]). Finally, we have treated supernova ejecta as isotropic but observations show a great deal of inhomogeneity (e.g. Hwang et al. [2004]). In principle, this allows disk enrichment at greater distances but the overall disk injection probability decreases because more disks nearby are not enriched. Specifically, let the filling factor be $f$. Then the ejecta are spread over an area $4\pi d^{2}f$ at distance $d$ from the supernova. Compared to the homogeneous case, a disk can be enriched at a greater distance, $\propto f^{-1/2}$, but the disks are distributed isotopically and the number that are impacted decreases linearly with $f$. The overall number of injected disks is the product of these and therefore proportional to $f^{1/2}<1$. Consider the extreme case where all the ejecta are shot out in a single beam; at most one disk may be enriched very far from the star but all the other disks in the cluster are not enriched at all. There does not appear to be any way to make this mechanism apply in a general way and I therefore conclude that the direct injection of supernova ejecta into the proto-solar system disk is an unlikely scenario for the origin of 60Fe in the ESS. ### 4.3 Induced star formation The fundamental puzzle of 60Fe is that it associates the birth of our Sun with the death of a massive star. Yet stellar evolutionary timescales are generally much longer than formation timescales. Even the most massive stars, with the shortest lives, require 3 Myr before becoming a supernova and ejecting SLR into their surroundings. This is much greater than the $<1$ Myr timescale for a dense core to form and collapse to a protostar. As shown above, the timescale problem is one of the factors in the low likelihood of AGB wind contamination of the proto-solar system or direct injection of supernova ejecta into the proto-planetary disk. This problem would be mitigated, however, if there were a causal relation between the death of a massive star and the birth of lower mass stars. This led to the idea of induced star formation as a solution to the SLR problem. It has long been known that large stellar associations in molecular clouds often consist of spatially and kinematically distinct subgroups ordered in age (Blauuw 1964). The canonical example is the four Orion groups, OB1a–d, spaced sequentially from north to south and with ages from $\approx 10$ to $\approx 1$ Myr respectively. Elmegreen & Lada ([1977]) proposed that such sequences could be explained by the induced collapse of molecular gas on the boundary of an expanding HII region. There is a large body of literature on the observations and theory of star formation on the boundaries of HII regions (e.g., see the recent conference proceedings by Elmegreen & Palous [2007]) and clearly some stars are induced. Hester & Desch ([2005]) have advocated that our Sun formed in this manner and incorporated SLR from the massive stellar winds and subsequent supernovae. As with the other SLR transport mechanisms, however, the essential question is not whether this might occur but how likely is it? The answer depends on the size scale involved. Core scales, $L<0.1$ pc The expansion and photoionization of an HII region sweeps away the inter-clump material in a molecular cloud and reveals the denser regions. Indeed part of the reason why images of star formation on the boundaries of HII regions are so spectacular is because fingers of dusty, molecular gas are dramatically silhouetted against a bright background (e.g., Hester et al. [1996]; Smith, Stassun, & Bally [2005]). However, these structures amount to only a small fraction of the total mass of molecular gas in the cloud, are very short- lived, and they do not appear to have a significantly enhanced star formation efficiency. For instance, the famous “pillars of creation” in M16 have a total mass $M_{\rm gas}=200$ $M_{\odot}$ (White et al. [1999]) and their high velocity gradients indicate dynamical timescales $\sim 10^{5}$ yr (Pound [1998]). Only 11 of the 73 globules within the pillars contain a young stellar object (McCaughrean & Andersen [2002]) which indicates a star forming efficiency, $M_{}/M_{\rm gas}\approx 3$%, similar to molecular clouds in general and a total number of induced stars that is dwarfed by the $\approx 10^{4}$ stars in the host HII region (Hillenbrand et al. [1993]). Even if there were 100 such star-forming pillars at the cloud interface over the lifetime of the HII region, the total number of stars that were induced to form would amount to only about 10% of the stars in the central cluster. Obviously this is an extremely rough estimate but new high quality infrared and optical archival datasets should allow a more precise accounting of the amount of stars that are induced versus those that form spontaneously, and those in the associated HII regions. As with the other SLR transport mechanisms, the geometric dilution of SLR as they travel from source to planetesimal is the strongest constraint. Looney et al. ([2006]) show that the radioactivity distance that matches the 60Fe abundance with the ESS is about 100 times the core radius. For a typical core with radius $\approx 0.05$ pc this corresponds to $\approx 5$ pc. Note also that this assumes 100% transport efficiency from the supernova to the collapsing core which is likely to be overestimated by an order of magnitude (Vanhala & Boss [2002]). A 10% efficiency would imply a supernova-core distance of $\approx 2$ pc. The M16 pillars are about 2 pc from the ionizing star so appear to be an ideal analog for the solar birthplace (Hester & Desch [2005]). However, HII regions expand rapidly to tens of parsecs in size and any molecular cores or circumstellar disks must survive the direct impact of stellar wind and radiation for at least 3 Myr until the supernova occurs. Numerical simulations by Mellema et al. ([2006]) show that most molecular globules at the center of an HII region are photoablated within 0.4 Myr. Freyer, Hensler, & Yorke ([2003]) include the effect of stellar winds and show that they sweep away molecular filaments inside the HII region to beyond 10 pc by 3 Myr. The impact on a stellar wind on a core and whether it might trigger star formation was studied by Foster & Boss ([1996]). They found that the core could be induced to collapse if it remained isothermal (i.e., radiates away the impact energy) during the passage of the shock and that this occurs only for speeds less than 100 ${\rm km~{}s^{-1}}$. Stellar winds are much faster than this and their impact will shred a core. A simple visual comparison of their simulations with images of molecular globules inside HII regions look closer to the fast shock case suggesting that the dominant effect is core destruction and not star formation (Figure 4), Figure 4: Comparison of numerical simulations of induced core collapse with observations. The left two panels show results from Foster & Boss ([1996]) on the density distribution of dense cores impacted by slow and fast stellar ejecta. Speeds greater than 100 ${\rm km~{}s^{-1}}$ shred cores rather than induce them to collapse. The right panel shows an image of the Keyhole nebula in the $\eta$ Carina nebula. This is a dense core that appears to be in the process of being destroyed by the HII region rather than compressed to form new stars. Observations of the molecular gas toward HII regions with Myr ages often show a ring of enhanced column density on the boundary with the cloud but very little if any molecular material in the center (e.g., Lang et al. [2000]). Individual stars in clusters have an age spread of $\approx 1$ Myr but even if some stars were induced during the early expansion of the HII region their surrounding core has gone and any future supernova will, at most, impact a circumstellar disk. That is the scenario discussed above and shown to affect less than 1% of all protoplanetary systems. Since the evolution of the disk fraction includes observations of OB associations, any early induced star formation of this nature is implicitly taken into account. Clump and cloud scales, $L>0.1$ pc Larger targets capture more ejecta and can be further from the supernova. On the scales that characterize clumps and clouds, the radioactivity distance ranges from $50-10^{4}$ pc and better matches the sizes of evolved HII regions. Figure 5: The radioactivity distance at which an object of size $L$ will receive 60Fe to produce an abundance relative to 56Fe that matches the ESS level. The dashed line shows the Looney et al. ([2006]) calculation with no decay and the solid curved line incorporates decay during the free-fall collapse. The issue here is that the large regions must collapse to stellar scales and there will be significant decay of 60Fe as this happens. Larger regions have lower average densities and longer free-fall timescales. Using the size- linewidth relation in Heyer & Brunt ([2004]), and deriving a density from the virial mass, implies $t_{\rm ff}{\rm(Myr)}=3.7L{\rm(pc)}^{0.41}$. The initial amount of 60Fe must be higher to compensate for the decay and this reduces the clump-supernova distance by a factor $e^{t_{\rm ff}/2t_{1/2}}$ where $t_{1/2}=1.5$ Myr (Figure 5)444Note that the free-fall timescale is much greater than the travel time for the ejecta to reach the source and the latter is therefore neglected.. The longer collapse times at large size scales offsets the greater amount of initial 60Fe and the radioactivity distance has a maximum of 45 pc. Objects further away than this will be either too small to capture enough 60Fe or too large to collapse before significant decay. However clumps with sizes $L=2-11$ pc and masses $M=10^{3}-4\times 10^{4}$ $M_{\odot}$ that are within 40 pc of the massive stars and collapse as (or because) a supernova explodes could, in principle, produce numerous planetary systems enriched with ESS levels of 60Fe. Of all the possibilities considered so far, these objects represent the most promising candidates to match the cosmochemical constraints on the birth environment of the solar system. The case of the Rosette The Rosette molecular cloud, with mass $2\times 10^{5}$ $M_{\odot}$, sits adjacent to the well known Rosette nebula, a luminous 2–3 Myr old HII region powered by the NGC2244 cluster containing 7 O stars and about $10^{4}$ stars in total (Townsley et al. [2003]). The cloud contains several embedded clusters (Roman-Zuniga et al. [2008]) indicative of ongoing star formation. Mid-infrared observations by Poulton et al. ([2008]) reveals circumstellar disks in the nebula in addition to protostars in the cloud (Figure 6). Figure 6: Young stars in the Rosette molecular cloud. The greyscale plots CO emission from the cold gas in the cloud (Heyer, Williams, & Brunt [2006]) and the dots show the location of stars with mid-infrared excesses (Poulton et al. [2008]). These dusty sources stars form two distinct groups; a cluster of disks within the HII region remnants of its formation $2-3$ Myr ago, and a distributed population of protostellar disks and envelopes in the cloud, concentrated toward the molecular peaks. The dashed lines show the 5 pc radius around the cluster at which dense cores could be enriched (if they survived) and the maximum 45 pc radioactivity distance at which clumps with size 5 pc could be enriched. There are 293 identified disks in the NGC2244 cluster and their detection statistics are in proportion to that expected for its $2-3$ Myr age (Balog et al. [2007]). 458 sources lie in the cloud, mostly concentrated toward molecular peaks, and possess spectral energy distributions indicative of protostellar disks or envelopes and ages $\approx 1$ Myr. The HII region has evacuated a noticeable spherical cavity in the molecular gas with a radius of about 10 pc. However, the clear spatial segregation between the cluster and cloud sources suggest that the fraction of induced star formation at early times is small. If stars formed as the HII region expanded to its current size, we would expect dusty sources between the cluster and cloud. There are no identifiable cores within 5 pc of the cluster center. However, most of the cloud lies within the maximum radioactivity distance of 45 pc from the cluster and is structured into clumps with sizes and masses that match the conditions required for enrichment in Figure 5 (Williams, Blitz, & Stark [1995]). ## 5 The Galactic background revisited An essential point in the discovery of live 60Fe in the ESS is that its abundance is above the expected Galactic background level. Recent $\gamma$-ray observations suggest they may not be so different, however. The measured flux of photons at the decay energy of an SLR can be directly converted to an abundance. Diehl et al. ([2006]) show that 26Al is produced on a Galactic scale with an average abundance, $[^{26}{\rm Al}]/[^{27}{\rm Al}]=8.4\times 10^{-6}$, about one sixth the ESS value (Lee, Papanastassiou, & Wasserburg [1977]). 60Fe is less abundant and the line flux about ten times weaker but its decay has been detected and the inferred abundance, $[^{60}{\rm Fe}]/[^{26}{\rm Al}]=0.23$ (Harris et al. [2005]). This implies $[^{60}{\rm Fe}]/[^{56}{\rm Fe}]=[^{60}{\rm Fe}/[^{26}{\rm Al}]\times[^{26}{\rm Al}]/[^{27}{\rm Al}]\times[^{27}{\rm Al}]/[^{56}{\rm Fe}]=0.23\times 8.4\times 10^{-6}\times 0.092=1.8\times 10^{-7}$ which is between one half to one sixth of the ESS values (Tachibana et al. [2006]; Moynier et al. [2005]). Normalized by the production ratio, the observed abundances for both 26Al and 60Fe are both $1.5\times 10^{-3}$. These are, respectively, a factor of 6 and 3 higher than the predicted ISM average (Jacobsen [2005]). The reason for this discrepancy is not clear. Nevertheless, it seems hard to deny the $\gamma$-ray evidence and the implication that the observed 60Fe background level may be within a factor of 2 of the low end of the range of meteoritic measurements, [60Fe]/[56Fe]$=3\times 10^{-7}$. Further, the Galactic star formation rate – and therefore the background – may have been about two times higher at the time the Sun was born (Rocha-Pinto et al. [2000]). The match between the numbers is tantalizing and, if they hold, the 60Fe puzzle goes away. Even the high end of the 60Fe abundance measurements, [60Fe]/[56Fe]$=10^{-6}$, is only a factor of 3 greater than the enhanced, early Galactic background. Given the known inhomogeneity in the 26Al emission (Knodlseder et al. [1999]) might a potential explanation be that we formed in a “radioactive overdensity”? The likelihood of being in an overdensity of 3 is no more than 1/3 if the new stars are uniformly distributed. In fact, young stars congregate in spiral arms close to the supernova that produce the background. However, the structure of the multiple phases of the ISM is complex and not well characterized (Cox [2005]). The definitive answer will ultimately have to wait for higher resolution, higher sensitivity $\gamma$-ray observations but it may also be possible to address the issue via extragalactic observations of the ionized, atomic, and molecular gas components (e.g., Scoville et al. [2001]; Calzetti et al. [2005]; Schuster et al. [2007]). A crucial missing step is the journey from $\sim 10^{7}$ K hot ionized gas in supernovae remnants to $\approx 30$ K circumstellar disk to planetesimal. There is some evidence for rapid cloud formation from atomic gas on Myr timescales (Hartmann et al. [2001]; Elmegreen [2000]) but there is no viable mechanism to convert the hot ionized medium to atomic gas on comparable timescales. Note that each delay of 1 Myr raises the required overdensity by a factor of 4 and 2 for 26Al and 60Fe respectively. Finally, solving the 60Fe conundrum in this way, results in the opposite problem of enormous excesses in the abundances of the longer lived radionucleides. This reflects the difficulty, as many studies have noted, of matching all the abundances of all the SLR from a single source (e.g., Harper [1996]). ## 6 Discussion and Conclusions I have critically evaluated several proposed mechanisms for the delivery of 60Fe to the ESS and found that most are highly improbable. AGB stars have moved away from their birthsite and any cross-pollination of a molecular core or circumstellar disk would be serendipitous. Further, their output is too small to affect a significant fraction of the ISM. No more than 0.1%, and probably far less, of stars could receive 60Fe in this way. Supernova may inject SLR directly into disks but the range of distances over which disks both survive the blast and are enriched to ESS levels is very small. Further, few disks remain by the time of the supernova for all but the most massive, shortest-lived progenitors. In this case, I estimate that no more than about 1% of all protoplanetary systems could receive 60Fe in this way. I was unable to assign probabilities to induced star formation scenarios since the statistics are not well known. The population of young stars on the boundary of HII regions appears to be only a small fraction of that within the central cluster and the number that form deeper in the surrounding molecular cloud. HII regions expand rapidly and are tens of parsecs in size before any supernova explosion. Matching the required high levels of 60Fe places individual star forming cores well within the boundary of mature HII regions and they would be shredded by stellar winds. Even if a core were to collapse and form a star, the surrounding envelope would be photoablated or swept away by the time of a supernova which would impact, at most, a circumstellar disk. This scenario is effectively the same as the direct disk injection hypthesis since that used a prescription for disk evolution that included observations of clusters with massive stars. Larger, cluster forming clumps in the molecular cloud beyond the HII region can capture enough supernova material and collapse fast enough to deliver ESS levels of 60Fe to planetary systems. The large scale “collect-and-collapse” scenario for molecular clumps neighboring a massive star forming region may be the most promising solution to the 60Fe puzzle. Molecular clouds are found around many supernova remnants and, although they may be are strongly disrupted, star formation can occur in the surviving gas (Huang & Thaddeus [1986]; Reach & Rho [1999]; Reynoso & Mangum [2001]). Perhaps the Orion subgroups are an example where this happened in the past and the Rosette cloud an example where this will happen in the future. However, only through detailed surveys of the molecular and protostellar content in young cloud- supernova interactions that quantify the amount and efficiency of star formation can we assess the likelihood that the Sun was born in such an environment. The discrepancy between the $\gamma$-ray measurements of the 26Al and 60Fe background with model calculations needs to be understood. Also, more sensitive cosmochemical measurements will reduce the uncertainty on the initial abundance of 60Fe and constrain the timing of its injection (Bizzarro et al. [2007], Dauphas et al. [2008]). If the high end of of the current range is confirmed, the background can be ruled out as the source although it may be a significant component that reduces the amount required from a discrete source. Finally, if there really is no common scenario that can deliver large amounts of 60Fe from a massive stellar core to a protoplanetary system, we are faced with the prospect that our solar system may be a one-in-a-hundred rarity. Astronomers have a long history of rejecting the idea that we are special but what are the implications in this case? It seems reasonable to assume that however 60Fe was delivered to the ESS, many other SLR were incorporated in the same manner. But if this was an unusual occurrence, then most planetary systems have lower levels not just of 60Fe but of other SLR, particularly 26Al. The latter is the dominant heating source for planetesimals at early times (Hevey & Sanders [2006]) and a reduced abundance would imply a different thermal history and, potentially, result in a higher water content of terrestrial planets (Desch & Leshin [2004]; Gaidos, Raymond, & Williams [2008]). ###### Acknowledgements. I thank Thierry Montmerle and the organizers for inviting me to a wonderful meeting in a spectacular location. Inspiration for this work came from stimulating conversations with Matthieu Gounelle and Ed Young at this meeting, Jeff Hester, Steve Desch and John Bally at a 2007 workshop, and Eric Gaidos, Sasha Krot, and Gary Huss in Hawaii. This work is supported by the NASA Astrobiology Institute under Cooperative Agreement No. NNA04CC08A. ## References [2001] Adams, F. C., & Laughlin, G. 2001, Icarus, 150, 151 * [2006] Adams, F. C., Proskow, E. M., Fatuzzo, M., & Myers, P. C. 2006, ApJ, 641, 504 * [2007] Andrews, S. M., & Williams, J. P. 2007, ApJ, 659, 705 * [2007] Balog, Z., Muzerolle, J., Rieke, G. H., Su, K. Y. L., Young, E. T., & Megeath, S. T. 2007, ApJ, 660, 1532 * [1990] Bertoldi, F., & McKee, C. F. 1990, ApJ, 354, 529 * [2007] Bizzarro, M., Ulfbeck, D., Trinquier, A., Thrane, K., Connelly, J. N., & Meyer, B. S. 2007, Science, 316, 1178 * [2005] Calzetti, D., et al. 2005, ApJ, 633, 871 * [1977] Cameron, A. G. W., & Truran, J. W. 1977, Icarus, 30, 447 * [1993] Cameron, A. G. W. 1993, Protostars and Planets III, 47 * [2000] Carpenter, J. M. 2000, AJ, 120, 3139 * [2000] Chevalier, R. A. 2000, ApJ, 538, L151 * [2005] Cox, D. P. 2005, ARA&A, 43, 337 * [2008] Dauphas, N., et al. 2008, ApJ, in press (astro-ph/0805.2607) * [2004] Desch, S. J., & Leshin, L. A. 2004, Lunar and Planetary Institute Conference Abstracts, 35, 1987 * [2007] di Francesco, J., Evans, N. J., II, Caselli, P., Myers, P. C., Shirley, Y., Aikawa, Y., & Tafalla, M. 2007, Protostars and Planets V, 17 * [2006] Diehl, R., et al. 2006, Nature, 439, 45 * [1977] Elmegreen, B. G., & Lada, C. J. 1977, ApJ, 214, 725 * [2000] Elmegreen, B. G. 2000, ApJ, 530, 277 * [2007] Elmegreen, B. G., & Palous, J. 2007, IAU Symposium, 237 * [1996] Foster, P. N., & Boss, A. P. 1996, ApJ, 468, 784 * [2003] Freyer, T., Hensler, G., & Yorke, H. W. 2003, ApJ, 594, 888 * [2008] Gaidos, E., Raymond, S. N., & Williams, J. P. 2008, ApJ, submitted * [2008] Gounelle, M., & Meibom, A. 2008, ApJ, 680, 781 * [2001] Haisch, K. E., Jr., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, L153 * [1996] Harper, C. L., Jr. 1996, ApJ, 466, 1026 * [2005] Harris, M. J., et al. 2005, A&A, 433, L49 * [2001] Hartmann, L., Ballesteros-Paredes, J., & Bergin, E. A. 2001, ApJ, 562, 852 * [2008] Hernandez, J., Hartmann, L., Calvet, N., Jeffries, R. D., Gutermuth, R., Muzerolle, J., & Stauffer, J. 2008, ApJ, in press (astro-ph/0806.2639) * [1996] Hester, J. J., et al. 1996, AJ, 111, 2349 * [2005] Hester, J. J., & Desch, S. J. 2005, Chondrites and the Protoplanetary Disk, 341, 107 * [2006] Hevey, P. J., & Sanders, I. S. 2006, Meteoritics and Planetary Science, 41, 95 * [2004] Heyer, M. H., & Brunt, C. M. 2004, ApJ, 615, L45 * [2006] Heyer, M. H., Williams, J. P., & Brunt, C. M. 2006, ApJ, 643, 956 * [1993] Hillenbrand, L. A., Massey, P., Strom, S. E., & Merrill, K. M. 1993, AJ, 106, 1906 * [1986] Huang, Y.-L., & Thaddeus, P. 1986, ApJ, 309, 804 * [2004] Hwang, U., et al. 2004, ApJ, 615, L117 * [2005] Jacobsen, S. B. 2005, Chondrites and the Protoplanetary Disk, 341, 548 * [2007] Jørgensen, J. K., Johnstone, D., Kirk, H., & Myers, P. C. 2007, ApJ, 656, 293 * [1990] Jura, M., & Kleinmann, S. G. 1990, ApJ, 364, 663 * [1994] Kastner, J. H., & Myers, P. C. 1994, ApJ, 421, 605 * [1999] Knödlseder, J., et al. 1999, A&A, 344, 68 * [2003] Lada, C. J., & Lada, E. A. 2003, ARA&A, 41, 57 * [2000] Lang, W. J., Masheder, M. R. W., Dame, T. M., & Thaddeus, P. 2000, A&A, 357, 1001 * [1998] Lee, T., Shu, F. H., Shang, H., Glassgold, A. E., & Rehm, K. E. 1998, ApJ, 506, 898 * [1977] Lee, T., Papanastassiou, D. A., & Wasserburg, G. J. 1977, ApJ, 211, L107 * [2006] Looney, L. W., Tobin, J. J., & Fields, B. D. 2006, ApJ, 652, 1755 * [2007] Mac Low, M.-M., Toraskar, J., Oishi, J. S., & Abel, T. 2007, ApJ, 668, 980 * [1998] Madau, P., Pozzetti, L., & Dickinson, M. 1998, ApJ, 498, 106 * [2002] Matzner, C. D. 2002, ApJ, 566, 302 * [2002] McCaughrean, M. J., & Andersen, M. 2002, A&A, 389, 513 * [2007] McKee, C. F., & Ostriker, E. C. 2007, ARA&A, 45, 565 * [2006] Mellema, G., Arthur, S. J., Henney, W. J., Iliev, I. T., & Shapiro, P. R. 2006, ApJ, 647, 397 * [2005] Moynier, F., Blichert-Toft, J., Telouk, P., & Albarède, F. A. 2005, 36th Annual Lunar and Planetary Science Conference, 36, 1593 * [1993] O’dell, C. R., Wen, Z., & Hu, X. 1993, ApJ, 410, 696 * [2007] Ouellette, N., Desch, S. J., & Hester, J. J. 2007, ApJ, 662, 1268 * [1998] Pound, M. W. 1998, ApJ, 493, L113 * [2008] Poulton, C. J., Robitaille, T. P., Greaves, J. S., Bonnell, I. A., Williams, J. P., & Heyer, M. H. 2008, MNRAS, 384, 1249 * [1999] Reach, W. T., & Rho, J. 1999, ApJ, 511, 836 * [2001] Reynoso, E. M., & Mangum, J. G. 2001, AJ, 121, 347 * [2000] Rocha-Pinto, H. J., Scalo, J., Maciel, W. J., & Flynn, C. 2000, A&A, 358, 869 * [2008] Román-Zúñiga, C. G., Elston, R., Ferreira, B., & Lada, E. A. 2008, ApJ, 672, 861 * [1986] Scalo, J. M. 1986, Fundamentals of Cosmic Physics, 11, 1 * [1992] Schaller, G., Schaerer, D., Meynet, G., & Maeder, A. 1992, A&A Suppl., 96, 269 * [1970] Schramm, D. N., & Wasserburg, G. J. 1970, ApJ, 162, 57 * [2007] Schuster, K. F., Kramer, C., Hitschfeld, M., Garcia-Burillo, S., & Mookerjea, B. 2007, A&A, 461, 143 * [2001] Scoville, N. Z., Polletta, M., Ewald, S., Stolovy, S. R., Thompson, R., & Rieke, M. 2001, AJ, 122, 3017 * [2005] Smith, N., Stassun, K. G., & Bally, J. 2005, AJ, 129, 888 * [2003] Tachibana, S., & Huss, G. R. 2003, ApJ, 588, L41 * [2006] Tachibana, S., Huss, G. R., Kita, N. T., Shimoda, G., & Morishita, Y. 2006, ApJ, 639, L87 * [2003] Townsley, L. K., Feigelson, E. D., Montmerle, T., Broos, P. S., Chu, Y.-H., & Garmire, G. P. 2003, ApJ, 593, 874 * [2005] Throop, H. B., & Bally, J. 2005, ApJ, 623, L149 * [2002] Vanhala, H. A. T., & Boss, A. P. 2002, ApJ, 575, 1144 * [2007] Wadhwa, M., Amelin, Y., Davis, A. M., Lugmair, G. W., Meyer, B., Gounelle, M., & Desch, S. J. 2007, Protostars and Planets V, 835 * [2006] Wasserburg, G. J., Busso, M., Gallino, R., & Nollett, K. M. 2006, Nuclear Physics A, 777, 5 * [1999] White, G. J., et al. 1999, A&A, 342, 233 * [1995] Williams, J. P., Blitz, L., & Stark, A. A. 1995, ApJ, 451, 252 * [1996] Williams, J. P., & Maddalena, R. J. 1996, ApJ, 464, 247 * [1997] Williams, J. P., & McKee, C. F. 1997, ApJ, 476, 166 * [2000] Williams, J. P., Blitz, L., & McKee, C. F. 2000, Protostars and Planets IV, 97 * [2005] Williams, J. P., Andrews, S. M., & Wilner, D. J. 2005, ApJ, 634, 495 * [2007] Williams, J. P., & Gaidos, E. 2007, ApJ, 663, L33	arxiv-papers	2008-08-19T00:16:20	2024-09-04T02:48:57.353993	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Jonathan P. Williams", "submitter": "Jonathan Williams", "url": "https://arxiv.org/abs/0808.2506" }
0808.2508	# Infrared behaviour of the fermion propagator in unquenched QED3 with finite threshold effects Yuichi Hoshino Kushiro National College of Technology,Otanoshike Nishi 2-32-1,Kushiro City,HOkkaido 084,Japan ###### Abstract To remove the linear infrared divergences in quenched approximation we include the massive fermion loop to the photon spectral function.Spectral function of fermion has no one particle singularity if we fix the anomalous dimension to be unity.In the case of $N$ flavour,$N$ dependence of order parameter is mild which may be due to screening effects. confinement,low-dimensional field theory ###### pacs: PACS number ††preprint: identifier LABEL:FirstPage101 LABEL:LastPage#1102 ###### Contents 1. I Introduction 2. II Spectral representation of the propagator 1. II.1 Fermion 2. II.2 Photon 3. III Analysis in position space 4. IV Analysises in momentum space 1. IV.1 structure in Euclid space 2. IV.2 structure in Minkowski space 5. V Renormalization constant and order parameter 6. VI Summary 7. VII Aknowledgement 8. VIII References ## I Introduction To study the infrared behavior of the propagator in QED2+1,we evaluated the spectral function which is known as the Bloch-Nordsieck approximation in four dimension.In quenched case linear and logarithmic infrared divergences appear in the order $e^{2}$ spectral function $F$.Exponentiation of $F$ yields full propagator with all order of infrared divergences.In four dimension anomalous dimension modifies the short distance singularity at least for weak coupling,which leads cut structure near the mass shell.For fixed infrared cut- off we expand the function $F$ and find a mass shift,its log correction and anomalous dimension of wave function.To avoid the infrared divergences we choose the gauge $d=-1$ and include effects of massive fermion loops to the photon spectral function.The resutls have no infrared divergence but the effects of finite threshold seems to make oscillation of the propagator.If we set the anomalous dimension to be unity short distance singularity disappears,chiral symmetry breaks dynamically, and the spectral function has $-\delta^{\prime}(p^{2}/m^{2}-1)$ like singularity which shows the absence of one particle state and vanishment of $Z_{2}^{-1}$. ## II Spectral representation of the propagator ### II.1 Fermion In this section we show how to evalute the fermion propagator non pertubatively by the spectral represntation which preserves unitarity and analyticity [1,2,3]. Assuming parity conservation we adopt 4-component spinor. The spectral function of the fermion in (2+1) dimension is defined $\displaystyle\left\langle 0\|T(\psi(x)\overline{\psi}(y)\|0\right\rangle$ $\displaystyle=i\int\frac{d^{3}p}{(2\pi)^{3}}e^{-ip\cdot(x-y)}\int_{0}^{\infty}ds\frac{\gamma\cdot p\rho_{1}(s)+\rho_{2}(s)}{p^{2}-s+i\epsilon},$ (1) $\displaystyle\rho(p)$ $\displaystyle=-\frac{1}{\pi}\operatorname{Im}S_{F}(p)=\gamma\cdot p\rho_{1}(p)+\rho_{2}(p)$ $\displaystyle=(2\pi)^{2}\sum_{n}\delta^{(3)}(p-p_{n})\left\langle 0\|\psi(x)\|n\right\rangle\left\langle n\|\overline{\psi}(0)\|0\right\rangle.$ (2) In the quenched approximation the state $\|n>$ stands for a fermion and arbitrary numbers of photons, $\|n>=\|r;k_{1},...,k_{n}>,r^{2}=m^{2}.$ (3) In deriving the matrix element $\left\langle 0\|\psi(x)\|n\right\rangle$ we must take into occount the soft photon emission vertex which is written in the textbook for the scattering of charged particle by external electromagnetic fied or collision of charged particles.Based on low-energy theorem the most singular contribution for the matrix element $T_{n}=\left\langle\Omega\|\psi\|r;k_{1},....,k_{n}\right\rangle$ is known as the soft photons attached to external line.Ward-Takahashi-identity $\displaystyle T_{n}$ $\displaystyle=\epsilon_{\mu}^{n}T_{n}^{\mu},$ $\displaystyle k_{\mu}^{n}T_{n}^{\mu}(r;k_{1},..kn)$ $\displaystyle=eT_{n-1}(r;k_{1,}..k_{n-1}),k_{\mu}^{n2}\neq 0$ (4) is proved to be satisfied with the use of LSZ reduction formula[2].We have an approximate solution of (4) $T_{n}\|_{k_{n}^{2}=0}=e\frac{\gamma\cdot\epsilon}{\gamma\cdot(r+k_{n})-m}T_{n-1}.$ (5) From this relation the n-photon matrix element is replaced by the products of $T_{1}$ $T_{n}T_{n}^{+}\gamma_{0}\rightarrow{\displaystyle\prod\limits_{j=1}^{n}}T_{1}(k_{j})T_{1}^{+}(k_{j})\gamma_{0}.$ (6) By one-photon matrix element[4] $T_{1}=-ie\frac{\gamma\cdot(r+k)+m}{(r+k)^{2}-m^{2}+i\epsilon}\gamma_{\mu}\epsilon_{\lambda}^{\mu}(k)U_{S}(r),$ (7) and the function $F=\int\frac{d^{3}k}{(2\pi)^{2}}\exp(ik\cdot x)\theta(k_{0})\delta(k^{2})\sum\limits_{\lambda,S}T_{1}\overline{T_{1}},$ (8) where polarization sum we have $\Pi^{\mu\nu}(k)=\sum_{\lambda}\epsilon_{\lambda}^{\mu}(k)\epsilon_{\lambda}^{\nu}(k)=-(g^{\mu\nu}-\frac{k^{\mu}k^{\nu}}{k^{2}})-d\frac{k^{\mu}k^{\nu}}{k^{2}}.$ (9) The infinite sum $\sum_{n=0}T_{n}T_{n}^{+}\gamma_{0}/n!$ leads an $\exp(F)$.In this way we obtain a dressed fermion propagator with soft photon $S_{F}(x)=-(\frac{i\gamma\cdot\partial}{m}+1)\int\frac{md^{2}r}{(2\pi)^{2}\sqrt{r^{2}+m^{2}}}\exp(ir\cdot x)\exp(F),$ (10) where $F$ is known as model independent $F=-\frac{e^{2}}{2}(\frac{\gamma\cdot r+m}{m})\int\frac{d^{3}k}{(2\pi)^{2}}\exp(ik\cdot x)\theta(k^{0})\delta(k^{2})[\frac{m^{2}}{(r\cdot k)^{2}}+\frac{1}{(r\cdot k)}+\frac{d-1}{k^{2}}],$ (11) here we used covaiant$\ d$ gauge photon propagator and $\delta(k^{2})$ is read as the imaginary part of the free photon propagator at on shell.In our approximation two kinds of spectral function satisfy $\rho_{2}=m\rho_{1}.$ ### II.2 Photon For unquenched case we use the dressed photon with massive fermion loop with $N$ flavours.Spectral functions for dressed photon are given by vacuum polarization [5,6] $\displaystyle\Pi_{\mu\nu}(k)$ $\displaystyle\equiv ie^{2}\int\overline{d^{3}}pTr(\gamma_{\mu}\frac{1}{\gamma\cdot p-m}\gamma_{\nu}\frac{1}{\gamma\cdot(p-k)-m})$ $\displaystyle=-e^{2}\frac{T_{\mu\nu}}{8\pi}[(\sqrt{k^{2}}+\frac{4m^{2}}{\sqrt{k^{2}}})\ln(\frac{2m+\sqrt{k^{2}}}{2m-\sqrt{k^{2}}})-4m],$ (12) $\displaystyle T_{\mu\nu}$ $\displaystyle=(g_{\mu\nu}-\frac{k_{\mu}k_{\nu}}{k^{2}}),\overline{d^{3}}p=\frac{d^{3}p}{(2\pi)^{3}},$ $D_{\mu\nu}^{-1}(k)\equiv-(T_{\mu\nu}k^{2}+dk_{\mu}k_{\nu})+\Pi_{\mu\nu}(k).$ (13) Polarization function $\Pi(k)$ is $\displaystyle\Pi(k)$ $\displaystyle=-\frac{e^{2}}{8\pi}[(\sqrt{-k^{2}}+\frac{4m^{2}}{\sqrt{-k^{2}}})\ln(\frac{2m+\sqrt{-k^{2}}}{2m-\sqrt{-k^{2}}})-4m],$ $\displaystyle=-\frac{e^{2}}{8}i\sqrt{-k^{2}}(-k^{2}>0,m=0),$ (14) $\displaystyle=\frac{e^{2}}{6\pi m}k^{2}+O(k^{4})(-k^{2}/m\ll 1).$ (15) Fermion mass is assumed to be generated dynamically.In quenched case it is shown that $m$ is proportional to $e^{2}$[2].For massless case or high-energy limit we have for number of $N$ fermion flavour $\rho_{\gamma}^{D}(k)=\frac{1}{\pi}\operatorname{Im}D_{F}(k)=\frac{c\sqrt{k^{2}}}{k^{2}(k^{2}+c^{2})},c=\frac{e^{2}N}{8},$ (16) $D_{\mu\nu}(k)=-T_{\mu\nu}\int_{0}^{\infty}\frac{\rho_{\gamma}(\mu^{2})d\mu^{2}}{k^{2}-\mu^{2}+i\epsilon}-d\frac{k_{\mu}k_{\nu}}{(k^{2}-i\epsilon)^{2}}.$ (17) If we include finite threshold effects of massive fermion pair we have $\rho_{\gamma}^{F}(k)=\frac{1}{\pi}\Im D_{F}(k)=\delta(k^{2})+\frac{1}{\pi}\Im\frac{1}{(-k^{2}-\Pi(-k^{2}))}\theta(-k^{2}-4m^{2}).$ (18) ## III Analysis in position space To evaluate the function $F$ it is helpful to use the exponential cut- off(infrared cut-off)[2,3].Using the bare photon propagator with bare mass $\mu$ ;$D_{F}^{(0)}(x)_{+}=\exp(-\mu\|x\|)/8\pi i\|x\|.$The function $F$ is written in the following form $F=ie^{2}m^{2}\int_{0}^{\infty}\alpha d\alpha D_{F}(x+\alpha r)-e^{2}\int_{0}^{\infty}d\alpha D_{F}(x+\alpha r)-i(d-1)e^{2}\frac{\partial}{\partial\mu^{2}}D_{F}(x,\mu^{2}).$ (19) The above formulea are derived by the parameter tric $\displaystyle\lim_{\epsilon\rightarrow 0}\int_{0}^{\infty}d\alpha\exp(-\alpha(\epsilon-ik\cdot r))$ $\displaystyle=\frac{i}{k\cdot r},$ $\displaystyle\lim_{\epsilon\rightarrow 0}\int_{0}^{\infty}\alpha d\alpha\exp(-\alpha(\epsilon-ik\cdot r))$ $\displaystyle=-\frac{1}{(k\cdot r)^{2}}.$ (20) In this case we have $\displaystyle F$ $\displaystyle=\frac{e^{2}}{8\pi}[\frac{\exp(-\mu\|x\|)-\mu\|x\|E_{1}(\mu\|x\|)}{\mu}-\frac{E_{1}(\mu\|x\|)}{m}$ $\displaystyle+\frac{(d-1)\exp(-\mu\left\|x\right\|))}{2\mu}],\|x\|=\sqrt{x^{2}}$ (21) where $\mu$ is a bare photon mass.Short distance behaiviour of $F$ has the following form $F\sim\frac{e^{2}(1+d)}{16\pi\mu}+\frac{e^{2}}{8\pi m}(\gamma+(1+m\|x\|)\ln(\mu\left\|x\right\|))-\frac{(d+1-2\gamma)e^{2}\|x\|}{16\pi},(\mu\|x\|\ll 1).$ (22) Long distance behaviour is given by the asymptotic expansion of $E_{1}(\mu\|x\|)$ $E_{1}(z)\sim\frac{\exp(-z)}{z}\\{1-\frac{1}{z}+\frac{1\cdot 2}{z^{2}}-\frac{1\cdot 2\cdot 3}{z^{3}}+...\\},(\|\arg z\|<\frac{3}{2}\pi),$ (23) $F\sim-\frac{e^{2}}{8\pi}[\frac{\exp(-\mu\|x\|)}{\mu^{2}\|x\|}+\frac{\exp(-\mu\|x\|)}{m\mu\|x\|}+\frac{(d-1)\exp(-\mu\|x\|)}{2\mu}],(\mu\|x\|\gg 1).$ (24) where $\gamma$ is an Euler constant.In (20) linear term in $\left\|x\right\|$ is understood as the finite mass shift from the form of the propagator in position space (9,25) and $\left\|x\right\|\ln(\mu\left\|x\right\|)$ term is position dependent mass $\displaystyle m$ $\displaystyle=m_{0}+\frac{e^{2}}{16\pi}(1+d-2\gamma)+\frac{e^{2}}{8\pi m}\mu,$ (25) $\displaystyle m(x)$ $\displaystyle=m-\frac{e^{2}}{8\pi}\ln(\mu\left\|x\right\|),$ (26) which has mass changing effects at short distance and it will be discussed in section 4.These mass terms has different gauge dependence from that obtained by self-energy in the Dyson-Schwinger equation.Here we notice that the propagator in poition space can be written $\displaystyle S_{F}(x)$ $\displaystyle=-(i\gamma\cdot\partial+m)\frac{1}{4\pi\|x\|}\left(\begin{array}[c]{c}\exp(-m\|x\|)A(\mu\|x\|)^{D+C\|x\|}(\mu\|x\|\ll 1)\\\ \exp(-m\|x\|)(\mu\|x\|\gg 1)\end{array}\right),$ (29) $\displaystyle=i\gamma\cdot\partial S_{V}(x)+S_{S}(x),$ (30) where $A=\exp(\frac{e^{2}(1+d)}{16\pi\mu}+\frac{e^{2}\gamma}{8\pi m}),C=\frac{e^{2}}{8\pi},D=\frac{e^{2}}{8\pi m}.$ (31) For the finitenes of the value $S_{F}(0)$ we imply $D=1.$In this case the physical mass equals to $m=e^{2}/8\pi$ and fermion may be confined for finite $\mu$.Here we apply the spectral function of photon to evaluate the unquenched fermion proagator.We simply integrate the function $F(x,\mu)$ for quenched case which is given in (19),where $\mu$ is a photon mass.Spectral function of photon with massless fermion loop is given in (13) and we have $\rho_{\gamma}^{D}(\mu)=\frac{c}{\pi\mu(\mu^{2}+c^{2})},Z_{3}^{-1}=\int_{0}^{\infty}2\rho_{\gamma}^{D}(\mu)\mu d\mu=1.$ (32) In this case the spectral function of the fermion is given $\widetilde{\rho}(x)=\int_{0}^{\infty}2\rho_{\gamma}^{D}(\mu)\exp(F(x,\mu))d\mu.$ (33) In this way the short distance fermion propagator with $N$ flavours is modified in the gauge $d=-1$ $S_{F}(x)=-(i\gamma\cdot\partial+m)\frac{\exp(-m\left\|x\right\|)}{4\pi\left\|x\right\|}\widetilde{\rho}(x).$ (34) For the case of $N$ fermion flavour we assume the physical mass $m$ as $m=c/N\pi.$In the whole region of $\|x\|$ we evaluate $\widetilde{\rho}(x)$ numerically including finite threshold effect for the photon spectral function with massive fermion loop $\widetilde{\rho}(x)=\int_{4m^{2}}^{\infty}\exp(F(s,x))\rho_{\gamma}^{F}(s)ds.$ (35) The renormalization constant $Z_{3}^{-1}$ is defined as the residue of pole $Z_{3}^{-1}=1+\int_{4m^{2}}^{\infty}\rho_{\gamma}^{F}(s)ds.$ (36) We set the mass $m=c/N\pi$ and see the $N$ dependence of $Z_{3}^{-1}$ for weak coupling $c=1/8$ in Fig.1.In Fig.2 we see the screenig effect leads infrared finite spectral function with dynamical fermion mass $m=c/N\pi$ from $N=1$ to $4$ in the gauge $d=-1,c=1/8.$ In comparison with quenched case with finite cut-off $\mu$ these are reduced from unity by screening effects at large distance. . Fig.1 $Z_{3}^{-1}-1$ for $N=1/2..10,c=1/8.$ Fig.2.$\widetilde{\rho}(x)$ for $c=1/8,N=1(bottom),2,3,4(top)\|.$ ## IV Analysises in momentum space ### IV.1 structure in Euclid space Now we turn to the fermion propagator in momentum space.The momentum space propagator at short distance is given by Fourier transform $S_{F}(p)=-\int d^{3}x\exp(-ip\cdot x)(i\gamma\cdot\partial+m)\frac{\exp(-m\left\|x\right\|}{4\pi\|x\|}\exp(F(x)),$ (37) where $\exp(F(x))=A\exp(D\gamma)(\mu\left\|x\right\|)^{D+C\left\|x\right\|},D=\frac{c}{N\pi m},C=\frac{c}{N}.$ (38) First we examine the perturbative effects by expanding $F$ in powers of $e^{2}$ for quenched case at short distace $\displaystyle\exp(F)$ $\displaystyle=1+F(x)=1+A-B\|x\|+C\|x\|\ln(\mu\|x\|)+D\ln(\mu\|x\|),$ $\displaystyle A$ $\displaystyle=\exp(\frac{e^{2}(1+d)}{16\pi\mu}+\frac{e^{2}\gamma}{8\pi m}),B=\frac{(d+1-2\gamma)}{16\pi},C=\frac{e^{2}}{8\pi},D=\frac{e^{2}}{8\pi m}.$ (39) Terms proportional to $B,C$ represents dynamical mass generation correctly.On the other hand it is known that $\int d^{3}x\exp(-ip\cdot x)\frac{\exp(-m\left\|x\right\|)}{4\pi\left\|x\right\|}(\mu\left\|x\right\|)^{D}=\frac{2m\mu}{(p^{2}+m^{2})^{2}}\text{ for }D=1.$ for Euclidean momentum $p^{2}\geq 0.$However above formule are restricted for small $\|x\|,$we may include the finite range effect and test the high ernergy behaviour for small $\mu$ $\displaystyle\int_{0}^{1/\mu}\frac{\exp(-m\|x\|)}{\|x\|}(\mu\|x\|)\frac{\sin(p\|x\|)}{p\|x\|}x^{2}dx$ $\displaystyle\rightarrow-\frac{\mu\exp(-m/\mu)\cos(p/\mu)}{p^{2}}+\frac{\mu\exp(-m/\mu)\sin(p/\mu)(m-\mu)}{p^{3}}+O(1/p^{4}).$ (40) Above formula shows the oscillation and the propagator does not dump as $1/p^{4}$.For large $N$ this oscillation effects becomes small as $\mu\rightarrow 0.$Numerical solutions of the scalar part of the propagator $S_{S}(p)$ with various $N$ are shown in Fig.3-1, Fig.3-2, for $c=1/8,1,N=1,2,3$ respectively.The boundary condition at $p=0$ is not specfied in our approximation.For weak coupling $c=1/8$ and $c=1$ zero momentum mass for $N=1$ is the largest among them, which has been seen in the Dyson- Schwinger analysis[8]. Figure.3-1 $S_{S}(p)$ for $c=1/8,$ $N=1(top),2(middle),3(bottom),p=\exp(\pi/2\sinh(k/5)).$ Figure.3-2 $S_{S}(p)$ for $c=1,$ $N=1(bottom),2(middle),3(top),p=\exp(\pi/2\sinh((m/5))$ ### IV.2 structure in Minkowski space Now we derive the propagator in Minkowski momentum region.To do this it is helpful in real time to derive the spectral function[1,9].In Minkowski space we change the variable from $r=\sqrt{x^{2}}$ to $iT.$We may define $\rho(s^{\prime 2})=\frac{1}{2\pi}\int_{-\infty}^{\infty}\exp(-i(s^{\prime 2}/m^{2}-1)T)\operatorname{Im}(\exp(\widetilde{F}(iT))dT,$ (41) which is normalized to $\delta(s^{2}/m^{2}-1)$ for free case where $\exp(\widetilde{F}(iT))=\int_{4m^{2}}^{\infty}d\sigma\rho^{F}(\sigma)\exp(F(iT,\sqrt{\sigma})).$ (42) It is understood that the Fourier transformation is performed in $s^{2}$ by the following form $\frac{1}{2\pi}\int_{-\infty}^{\infty}dT\exp(-i(m-s)T)+\exp(-i(m+s)T)=2\|s\|\delta(s^{2}-m^{2}).$ (43) Taking the imaginary part of the function $\exp(F(\mu,iT))$ at short distance in the gauge $d=-1$ for quenched case for $D=1$ $\operatorname{Im}\exp(F(\mu,iT))\approx\exp(-\frac{\pi}{2}CT)\mu T\cos(CT\ln(\mu T)),$ (44) provided $(i\mu T)^{iCT+D}=\exp(\frac{\pi}{2}(-CT+Di))(\mu T)^{D}(\cos(CT\ln(\mu T))+i\sin(CT\ln(\mu T))).$ (45) In this way we see the oscillation of the propagator in Minkowski space by the effects of position dependent mass as $m=iCT\ln(iCT).$In Fig.4 we see the profile of real and imaginary part of the function $\exp(F(iT))$ for $c=1,N=1$. Fig.4 $\Re(\exp(\widetilde{F}(iT))$, $\Im(\exp(\widetilde{F}(iT))..$ Fig.5 $\rho(s),N=1(bottom),2,3(top).$ In Fig.5 the spectral function $\rho(s)$ in unit of $e^{2}$ looks symmetric about the point $s=1$.The function $\exp(F(iT))$ is approximated as $i\mu T$ for small $T$.This fact suggests the singularity of $\rho(s)$ such as the $-\delta^{\prime}(s-1)$. ## V Renormalization constant and order parameter In this section we consider the renormalization constant in our model.It is easy to evaluate the renormalization constant and bare mass which are defined by assuming multiplicative renormalization $\frac{Z_{2}^{-1}}{\gamma\cdot p-m_{0}}=S_{F}(p).$ (46) $\displaystyle Z_{2}^{-1}$ $\displaystyle=\lim_{p\rightarrow\infty}\frac{1}{4}tr(\gamma\cdot pS_{F}(p))$ $\displaystyle=\lim_{p\rightarrow\infty}\int\frac{p^{2}\rho(s)ds}{p^{2}-s+i\epsilon}=\int\rho(s)ds,$ (47) where $Z_{2\text{ }}$is defined for one particle state in the intial and final state in a weak sense $\psi(x)_{t\rightarrow+\infty,-\infty}\rightarrow\sqrt{Z_{2}}\psi(x)_{out,in}$ $Z_{2}(2\pi)^{-2}\theta(p_{0})\delta(p^{2}-m^{2})\equiv\|\left\langle p\|\psi(0)\|0\right\rangle\|_{p^{2}=m^{2}}^{2}.$ (48) However pole part is absent in our approximation and $Z_{2\text{ }}$vanishes for $D>0$ case.For $D=1$ we have $\displaystyle Z_{2}^{-1}$ $\displaystyle=\int_{0}^{\infty}ds\int_{-\infty}^{\infty}\frac{dt}{2\pi}\exp(-i(s-1)t)\Im(\exp(\widetilde{F}(it))$ $\displaystyle=\int_{-\infty}^{\infty}dt\delta_{+}(t)\Im(\exp(\widetilde{F}(it))=0,$ (49) provided $\Im(\exp(\widetilde{F}(it))=0$ in the limit $t_{+}\rightarrow 0$ by eq(38).Order parameter for each flavour $\left\langle\overline{\psi}\psi\right\rangle$ is given $\left\langle\overline{\psi}\psi\right\rangle=-trS_{F}(x).$ (50) The value of $\left\langle\overline{\psi}\psi\right\rangle$ is $3.1(1.5.,1.0)\times 10^{-3}e^{4}$ for $N=1(2,3)$ in $1/N$ which may be compared with $1.2(0.13,0.0002)\times 10^{-3}e^{4}$ for $N=1(2,3)$ in the CP vertex with massless loop correction [8].In our approximation $N$ dependence is mild.This may be understood as the large screening effect at small $N.$Recently we solved the Bethe-Salpeter equation in the ladder approximation for axial-scalar which corresponds the psedoscalar in four dimension.For massless boundstate solution we find the $D=e^{2}/4\pi m=1$ from the normalization condition at short distance[9]. ## VI Summary We evaluated the fermion propagator in three dimensional QED with dressed photon by the dispersion method.In the evaluation of lowest order matrix element for fermion spectral function we obtain finite mass shift,wave function renormalization and gauge invariant position dependent mass with bare photon mass $\mu$ as an infrared cut-off..To remove linear infrared divergences we include finite threshold effects of massive fermion loop to the photon spectral function.We set the coupling constant mass ratio $c/N\pi m$ to be unity which is consistent with perurabative analysis of the mass at high energy which is proportional to$1/p^{4}$.In this case the order parameter $\left\langle\overline{\psi}\psi\right\rangle$ becomes finite.In our approximation vacumm expectation value and infrared mass are not sensitive to flavour number $N$ which may due to the screening effects of $Z_{3}^{-1\text{ }}$for small $N$.In Minkowski space.the spectral fuction is not positive definite and has a form such as $-\delta^{\prime}(p^{2}/m^{2}-1)$.In the strong coupling there seems to be no particle content in the model.In our analysis chiral symmetry breaks dyamically by anomalous dimension.Our results suggests the importance of anomalous dimension to understand confinement. ## VII Aknowledgement The author would like to than Professor Rober Delbourgo for critiques of linear divegrgence. ## VIII References [1]R.Jackiw,L.Soloviev,Phys.Rev.137.3(1968)1485. [2]Y.Hoshino,JHEP0409:048(2004),whereIhave a wrong sign of $\|x\|\ln(\mu\|x\|)$ term. [3]Y.Hoshino,Nucl.Phys.A790(2007)613c-618c. [4]K.Nishijima,Fields and Particles,W.A.BENJAMIN,INC(1969). [5]A.B.Waites,R.Delbourgo,Int.J.Mod.Phys.A7(1992)6857. [6]C.Itzykson,J.B.Zuber,Quantum field theory,McGRAW-HILL. [7]J.Schwinger,Particle Sources and Fields,vol.I,Pereseus Books Publishing, L,L,C(1970). [8]C.S.Fischer,R.Alkofer,T.Dahm,P.Maris,Phys.Rev.D70,073007(2004):[arXiv:hep- th/0407014]. [9]Y.Hoshino:[arXiv:0706.1063]	arxiv-papers	2008-08-19T00:13:44	2024-09-04T02:48:57.360128	{ "license": "Public Domain", "authors": "Yuichi Hoshino", "submitter": "Yuichi Hoshino", "url": "https://arxiv.org/abs/0808.2508" }
0808.2734	# Elasticity of Spider Dragline Silks Viewed as Nematics: Yielding Induced by Isotropic-Nematic Phase Transition Lin-ying Cui cly05@mails.tsinghua.edu.cn Department of physics, Tsinghua University, Beijing, 100084, China Fei Liu Center for Advanced Study, Tsinghua University, Beijing, 100084, China Zhong-can Ou-Yang Center for Advanced Study, Tsinghua University, Beijing, 100084, China Institute of Theoretical Physics, The Chinese Academy of Sciences, P.O.Box 2735 Beijing 100190, China ###### Abstract Spider dragline silk shows well-known outstanding mechanical properties. However, its sigmoidal shape of the measured stress-strain curves (i.e. the yield) can not be described by classical polymer theories and recent hierarchical chain model. To solve the long lasting problem, we generalized the Maier-Saupe theory of nematics to construct an elastic model for the polypeptide chain network of the dragline silk. The comprehensive agreement between theory and experiments on the stress-strain curve strongly indicates the dragline silks to belong to liquid crystal elastomers. Especially, the remarkable yielding elasticity of the silk is understood for the first time as the force-induced isotropic-nematic phase transition of the chain network. Our theory also predicts a drop of the stress in supercontracted dragline silk, an early found effect of humidity on the mechanical property in many silks. ###### pacs: 81.05.Lg, 81.40.Jj, 82.35.Pq, 61.30.Dk Spider dragline silks (SDSs), the main structural web silk regarded as the “spider’s lifeline”, exhibit fascinating mechanical properties, such as a tactful combination of high tensile strength and high extensibility gosline , thus showing a remarkably sigmoidal shape of the measured stress-strain curves Vollrath2 . Several experimental studies have been carried out to determine the supra-molecular structure organization of the SDS Anja ; Hayashi ; Oroudjev ; Simmons ; Krasnov and tried to produce mimic silks with similar properties Li . It is now widely accepted that SDSs are semicrystalline polymers with $\beta$-sheet nanocrystals embedded in amorphous region, which is a polypeptide chain network Termonia ; Vollrath1 ; Lefevre ; see Fig. 1(a). However, the deformation mechanism, which is essential for understanding the SDS’s extraodinary mechanical properties and mimicking the silk, is still in intense debate Anja ; Porter ; Vollrath1 ; Vehoff ; Papa . On the theoretical side, to understand the exceptional mechanical properties of SDS is of longstanding interest, and many models have been proposed Termonia ; Porter ; Vollrath1 ; Krasnov and some insights attained Termonia ; Porter ; Vollrath1 ; Krasnov . For example, the model by Termonia Termonia treated the SDS as a hydrogen-bonded amorphous region embedded with stiff crystals as cross-links. In the interfacial region, an extremely high modulus is required to get the dragline’s overall behavior on deformation. While, in the model of Porter and Vallrath Porter ; Vollrath1 , parameters linking to chemical compositions and morphological order were used to interpret thermo- mechanical properties. But some parameters such as ordered/disordered fractions are difficult to be obtained from experiments. A recent model Krasnov connecting deformations on macroscopic and molecular length scales still did not consider the change of the orientation of nanocomposites during deformation. Especially, as pointed out by Vehoff at al. recently Vehoff , basic polymer theories such as the freely jointed chain, the freely rotating chain and the worm-like chain, as well as a hierarchical chain model of spider capture silk haijun can not reproduce the sigmoidal shape or even the steep initial regime of the spider dragline silk [ Fig. 2(a)] Vehoff . In one word, a unified description for SDS as a model biomaterial still seems to be lacking. Quite a few works Vollrath2 ; Knight ; Anja ; Simmons ; Li ; Kerkam ; willcox have pointed out that spider silk is liquid crystalline material and liquid crystal (LC) phase plays a vital role in both its spinning process and mechanical properties. In the spinning process, the liquid crystalline ‘spinning dope’ helps spider to control the folding and crystallization of the main protein constituents at benign condition (close to ambient temperatures and pressures using water as solvent) Vollrath2 ; Kerkam ; willcox . The liquid crystalline phase also plays an important role in the solid silk’s properties Knight . For instance, several works have found out that the orientation of nanocomposites can affect SDS’s mechanical properties significantly Anja ; Papa ; Simmons . A recent experiment also suggested the existence of conformational transition and the liquid crystalline state of regenerated silk fibroin in water Li . Therefore, to present an analytically tractable LC model of the SDS that can catch the main physical factors is a current challenge to theorists. In this work, we generalized the Maier-Saupe theory Maier of nematic LC to construct an elastic model for the polypeptide chain network of the SDS. We show that on deformation the SDS undergoes significant changes with orientation of the chain network increased and the dimension of the silk along force direction elongated. The comprehensive agreement between theory and experiments on the stress-strain curve strongly indicates the SDSs to belong to LC elastomers, described as a new class of matter recently warner . Especially, the remarkable yielding elasticity of the SDS is understood for the first time as the force-induced isotropic-nematic phase transition of the chain network and the self-consistently obtained yield point agrees with experimental data well. The present theory also predicts a drop of the stress in supercontracted SDS, an early found effect of humidity on the mechanical properties in many silks Bell ; Vehoff ; Vollrath1 . We take the polypeptide chain network in the amorphous region of the SDS as a molecular LC field with each chain section corresponding to a mesogenic molecule; see Fig. 1. Because the SDS’s high extensibility results primarily from the disordered region Krasnov ; Oroudjev ; Hayashi , and many experiments showed that the deformation of the crystals is at least a factor of 10 smaller than that of the bulk Anja and that the orientation of the $\beta$-sheets is almost unchanged (usually very high) under stress Philip ; Papa , we can neglect the deformations and rotations of the $\beta$-sheet crystals in current work. Following the LC continuum theory in the absence of forces, the potential of a mesogenic molecule takes the Maier-Saupe interaction form Maier $\displaystyle V(\cos\theta)=-aS(\frac{3}{2}\cos^{2}\theta-\frac{1}{2}),$ (1) where $\theta$ is the angle between the long axis of the molecule and the silk axis (the z-axis), which is also the direction of $\hat{\bf n}$ [ Fig. 1 (b)], $a$ is the strength of the mean field, and $S$ is the orientation order parameter of the LC, defined as the average of second Legendre polynomial gennes $\displaystyle S=\left\langle\frac{3}{2}\cos^{2}\theta-\frac{1}{2}\right\rangle.$ (2) We notice that the Maier-Saupe potential has been used by Pincus and de Gennes in investigating LC phase transition in a polypeptide system Pincus . When a uniform force field $\bf f$ along z-axis is applied, the potential of a molecule is written as $\displaystyle U(\cos\theta)=V-fl\cos\theta,$ (3) where $l$ denotes the length of the mesogenic molecule. From the definition of the order parameter $S$, we get a self-consistency equation $\displaystyle\begin{array}[]{lll}S&=&\int_{-1}^{1}(\frac{3}{2}\cos^{2}\theta-\frac{1}{2})\exp(\frac{3aS}{2{\rm k_{B}T}}\cos^{2}\theta+\alpha\cos\theta)d\cos\theta\\\ &&\left/\int_{-1}^{1}\exp(\frac{3aS}{2{\rm k_{B}T}}\cos^{2}\theta+\alpha\cos\theta)d\cos\theta\right.,\end{array}$ (6) with $\alpha=fl/{\rm k_{B}}T$. The solution of the above equation may not be unique, in order to obtain physically sound solution we still need the requirement of minimization of the free energy given by $\displaystyle F_{MS}=-{\rm k_{B}}T\ln Z+\frac{1}{2}aS^{2},$ (7) where $Z$ is the partition function $Z=\int_{-1}^{1}e^{-U(\cos\theta)/{\rm k_{B}}T}d\cos\theta$, and the second term at the right-hand side corrects for the double counting arising from the mean field method warner . We calculate the orientation function $S$ numerically at temperature $T^{}=T/T_{\rm ni}$ and force $f$ and show the results in Fig. 2(b). Here $T_{\rm ni}=a/(4.541\rm k_{B})$ is the isotropic-nematic transition temperature in the absence of forces Maier . We see that, at temperatures below $T_{\rm ni}$ the molecules have spontaneous nematic order, and the force does not induce further order significantly. While for the molecules initially in paranematic states, the applied force field will induce a first-order phase transition, which means $S$ jumps discontinuously to a higher value at a certain critical force $f_{\rm C}(T^{})$. At even higher temperatures, nematic field is weaker and the effect of the force is less dramatic. Interestingly, the $\alpha-S$ curves at different temperatures are qualitatively similar to the stress-orientation curves given by a much more complex nematic elastomer theory warner2 (Fig. 5 in Ref. [25]). To compare with the mechanical experiments of the SDS, we give the expressions for stress and strain in our theoretical framework. Apparently, the stress $\sigma$ of a bulk is $\sigma\equiv F/A=Nf$: $F$ is the force on the surface of the bulk, $A$ is the area of the surface, and $N$ is the number of molecules per area. The strain $\varepsilon$ of the bulk is defined as $\varepsilon=[L(f)-L_{0}]/L_{0}$, where $L(f)$ is the length of the bulk along z-axis when the force field $\bf f$ is applied and we can take it as $L(f)=\langle l\|\cos\theta\|\rangle$, and $L_{0}=L(f=0)=l/2$. Then the strain $\varepsilon$ is $\varepsilon=2\langle\|\cos\theta\|\rangle-1$. We show $\varepsilon$ versus $\sigma$ at different temperatures in Fig. 2(c). We see that at temperatures below $T_{\rm ni}$, the strain grows smoothly with the stress. While for temperatures just above $T_{\rm ni}$, the strain grows with the stress in almost a linear way under small forces, and then a jump in the strain occurs at the critical force $f_{\rm C}(T^{})$, after which the strain increases smoothly with the stress again. At even higher temperatures, the jump is replaced by a smooth increase in strain, but there is a plateau in a certain range of force. In our model, the reduced temperature $T^{}$ is an essential parameter, and we need to choose a proper value for it in order to predict the stress-strain curve of the SDS. Experiments showed that the solution from which the SDS was drawn was in liquid crystalline state at ambient temperature and pressure Vollrath2 ; Kerkam ; willcox ; Li , while the orientation in the amorphous region of solid silk was very low Papa . Thus, we assign the isotropic-nematic transition temperature $T_{\rm ni}$ of the cross-linked chain network slightly lower than the room temperature $T_{r}$. Namely, the SDS is in paranematic state at ambient temperature and $T^{}$ is just above 1. The $\sigma-\varepsilon$ curves in the paranematic states in our model indeed exhibit main features of the stress-strain relation of the SDS: there is a linear increase in stress with strain at small values, and then at a certain strain and afterwards, the material becomes softer with lower Young’s modulus Knight ; Du . We then reveal the beginning of the isotropic-nematic phase transition as the yield point. We see that the curves with $T^{}=1.01$ and 1.02 agree well with the measured curve in the beginning linear region and the yield point. But because the actual deformation process of the silk is more complicated and additional factors may be involved, such as the viscoelasticity, defects and poly-domain effect warner ,the curve with $T^{}=1.1$ agrees better with the overall stress-strain measurement topologically. We calculate the yield point by choosing curves with $T^{}=1.01$ and 1.02. We get the yield strain $\varepsilon_{y}\approx 0.04$, the yield stress $\sigma_{y}=\alpha N{\rm k_{B}}T/l\approx 8.4{\rm MPa}$, and the Young’s modulus at the linear region $E\equiv\sigma_{y}/\varepsilon_{y}\approx 210{\rm MPa}$, given $\alpha\sim 0.2$, $N/l\sim 10{\rm nm}^{-3}$, and ${\rm k_{B}}T_{r}\sim 4.1{\rm pNnm}$. These results agree with experimental data Du ; Vollrath1 [Fig. 2(d)] satisfactorily. We notice that our results agree much better with the mechanical properties of the silks with low spinning speed. That is because the spinning speed can induce a low orientation in the amorphous region which makes the silk more stiff. Since this additional order in the amorphous region is not taken into account in the current work, the silks in our model are generally a bit softer than the silks with high spinning speed. Another thing needs pointing out is that we predict there is a phase transition at the yield point which is supported by a few experiments. For instance, in the polarized FTIR spectroscopy experiment by Papadopoulos et al., the orientation of some components in the amorphous region increased by 0.3 when the strain reached 24%. Besides, we would like to discuss about the isotropic-nematic transition temperature $T_{\rm ni}$ of the cross-linked chain network. If we choose $T^{}=1.1$ in our calculation, the transition temperature $T_{\rm ni}\approx 270{\rm K}$, which is reasonable. In addition to describing the stress-strain relation of the SDS, our simple theory can also qualitatively account for the drop of the stress in the wet SDS, i.e. the supercontracted SDS. We take $L_{0}$ and $R_{0}$ as the initial length and radius of the silk, and $L$ and $R$ as those under stress. Under the assumption of volume conservation we have $\pi R_{0}^{2}L_{0}=\pi R^{2}L$, so $R/R_{0}=\sqrt{1/(1+\varepsilon)}$. The free energy of the bulk can be written as $\displaystyle\begin{array}[]{lll}F&=&U-f_{ext}(L-L_{0})+2\pi RL\gamma\\\ &=&U-f_{ext}(L-L_{0})+2\pi R_{0}L_{0}\gamma\sqrt{1+\varepsilon},\end{array}$ (10) where $U$ is the internal energy of the bulk, $f_{ext}$ is the external force on the bulk and $\gamma$ is the surface energy coefficient. Minimizing $F$ with respect to $\varepsilon$, we get $\displaystyle\sigma=\frac{f}{\pi R_{0}^{2}}=\frac{1}{\pi R_{0}^{2}L_{0}}\frac{\partial U}{\partial\varepsilon}+\frac{\gamma}{R_{0}\sqrt{1+\varepsilon}}.$ (11) When the silk is immersed in water, the surface energy coefficient $\gamma$ increases, so with the same stress $\sigma$ we get a bigger strain. Thus our theory can predict the softening of supercontracted silk, an effect observed in many experiments Bell ; Vehoff ; Vollrath1 ; work2 ; work3 . In conclusion, we investigate the mechanical properties of the SDS from a point of view of the LC continuum theory. We found out that the deformation process is a force-induced isotropic-nematic phase transition process. Remarkably, such a simple model with Maier-Saupe theory is able to reproduce the stress-strain curve of the SDS, get the yield point, and qualitatively interpret the drop of the stress in the supercontracted silk. This work is supported by the National Innovation Research Project for Undergraduates. ## References (1) J. M. Gosline, M. W. Denny, and M. E. Demont, Nature 309, 551 (1984). * (2) F. Vollrath and D. P. Knight, Nature 410, 541 (2001). * (3) A. Glis ovic, T. Vehoff, R. J. Davies, and T. Salditt, Macromolecules 41, 390 (2008). * (4) I. Krasnov, et al., Phys. Rev. Lett. 100, 048104 (2008). * (5) C. Y. Hayashi and R. V. Lewis, J. Mol. Biol. 275, 773 (1998). * (6) E. Oroudjev, et al., Proc. Natl. Acad. Sci. USA 99, 6460 (2002). * (7) A. H. Simmons, C. A. Michal, and L. W. Jelinski, Science 271, 84 (1996). * (8) X. G. Li, L. Y. Wu, M. R. Huang, H. L. Shao, and X. C. Hu , Biopolymers 89, 497 (2007). * (9) Y. Termonia, Macromolecules 27, 7378 (1994). * (10) T. Lefevre, M. E. Rousseau, and M. Pezolet, Biophys. J. 92, 2885 (2007). * (11) F. Vollrath and D. Porter, Soft Matter 2, 377 (2006). * (12) D. Porter, F. Vollrath, and Z. Shao, Eur. Phys. J. E 16, 199 (2005). * (13) T. Vehoff, A. Gli sovi c, H. Schollmeyer, A. Zippelius, and T. Salditt, Biophys. J. 93, 4425 (2007). * (14) P. Papadopoulos, J. Solter, and F. Kremer, Eur. Phys. J. E 24, 193 (2007). * (15) HJ Zhou and Y. Zhang, Phys. Rev. Lett. 94, 028104 (2005). * (16) K. Kerkam, C. Viney, D. Kaplan, and S. Lombardi, Nature 349, 596 (1991). * (17) P. J. Willcox, S. P. Gido, W. Muller, and D. L. Kaplan, Macromolecules 29, 5106 (1996). * (18) D. P. Knight and F. Vollrath, Phil. Trans. R. Soc. Lond. B 357, 155 (2002). * (19) W. Maier and A. Saupe, Z. Naturforsch. A., 14, 882 (1959). * (20) M. Warner, E. M. Terentjev, Liquid Crystal Elastomers, (Clarendon Press, Oxford, 2003). * (21) F. I. Bell, I. J. McEwen, and C. Viney, Nature 416, 37 (2002). * (22) P. T. Eles and C. A. Michal, Biomacromolecules 5, 661 (2004). * (23) P. G. de Gennes, J. Prost, The Physics of Liquid Crystals(2nd edt), (Clarendon Press, Oxford, 1993). * (24) P. Pincus and P. G. de Gennes, J. Polymer Sciences: Polymer Symposium 65, 85 (1978). * (25) M. Warner and X. J. Wang, Macromolecules 24, 4932 (1991). * (26) N. Du, et al., Biophys. J. 91, 4528 (2006). * (27) R. W. Work, J. exp. Biol. 118, 379 (1985). * (28) R. W. Work, Textile Res. J. 47, 650 (1977). * (29) R. W. Work, J. Arachnol. 9, 299 (1981). Figure 1: (a). A schematic diagram of the structure of the dragline silk. The bold lines represent the $\beta$-sheet crystals, and the thin lines represent the polypeptide chains in the amorphous region. The z-axis is along the silk axis. (b). The coordinate system of the nematics. $\hat{\bf n}$ is the director of the nematics, $\hat{\bf u}$ is the director of the mesogenic molecule, and $\theta$ is the angle between the long axis of the molecule and the silk axis $z$. Figure 2: (Color online.) (a) Comparison of a typical measured dragline silk’s stress-strain curve (black solid line) with theoretical curves evaluated by the freely jointed chain (red dash dot line) and the hierarchical chain model (olive dash line) [After T. Vehoff et al.]. (b) The orientation order parameter $S$ as a function of $\alpha$($=fl/{\rm k_{B}}T$). (c) The stress-strain curves at different temperatures $T^{*}(=T/T_{\rm ni})$. (d) The stress-strain curve of the SDS spinned with the speed of $1{\rm mms}^{-1}$ [After N. Du et al.].	arxiv-papers	2008-08-20T11:42:58	2024-09-04T02:48:57.367324	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Lin-ying Cui, Fei Liu, Zhong-can Ou-Yang", "submitter": "Lin-ying Cui", "url": "https://arxiv.org/abs/0808.2734" }
0808.2926	# Relation between Fresnel transform of input light field and Radon transform of Wigner function of the field Hong-yi Fan and Li-yun Hu∗ Department of Physics, Shanghai Jiao Tong University, Shanghai, 200030, China Corresponding author: hlyun@sjtu.edu.cn; hlyun@sjtu.org ###### Abstract We prove a new theorem about the relationship between optical field Wigner function’s Radon transform and optical Fresnel transform of the field, i.e., when an input field $\psi\left(x^{\prime}\right)$ propagates through an optical $\left[D\left(-B\right)\left(-C\right)A\right]$ system, the energy density of the output field is equal to the Radon transform of the Wigner function of the input field, where the Radon transform parameters are $D,B.$ We prove this theorem in both spatial-domain and frequency-domain. In optical communication theory every signal or image can be uniquely and indirectly described by a Wigner distribution function (WDF) [1, 2, 3]. The WDF (or named Wigner transform) of an optical signal field $\psi\left(x^{\prime}\right)$ is defined as $W_{\psi}(\nu^{\prime},x^{\prime})=\int_{-\infty}^{+\infty}\frac{du}{2\pi}e^{i\nu^{\prime}u}\psi^{\ast}\left(x^{\prime}+\frac{u}{2}\right)\psi\left(x^{\prime}-\frac{u}{2}\right),$ (1) $W_{\psi}(\nu^{\prime},x^{\prime})$ involves both spatial distribution information and space-frequency distribution information of the signal. $\nu$ is named space frequency. $W_{\psi}(\nu^{\prime},x^{\prime})$ is said to be bilinear in the signal because the signal enters twice in its definition. The WDF undergoes certain variations if something happens to the signal. For examples, passage through a lens corresponds to a vertical shearing of the WDF, propagation in free space means a horizontal shearing of the WDF [4]. However, the WDF preserves space and space frequency marginal properties of any signal, $\displaystyle\int_{-\infty}^{+\infty}d\nu^{\prime}W_{\psi}(\nu^{\prime},x^{\prime})$ $\displaystyle=$ $\displaystyle\|\psi\left(x^{\prime}\right)\|^{2},$ (2) $\displaystyle\int_{-\infty}^{+\infty}dx^{\prime}W_{\psi}(\nu^{\prime},x^{\prime})$ $\displaystyle=$ $\displaystyle\|\tilde{\psi}\left(\nu^{\prime}\right)\|^{2},\text{ }$ (3) where $\tilde{\psi}\left(\nu\right)=\int_{-\infty}^{+\infty}\frac{dx}{\sqrt{2\pi}}\psi\left(x\right)e^{ix\nu}.$ If one wants to reconstruct the Wigner function by using various probability distribution, obviously the position density $\|\psi\left(x^{\prime}\right)\|^{2}$ and the space-frequency density $\|\tilde{\psi}\left(\nu^{\prime}\right)\|^{2}$ were not enough, so the Radon transform [5, 6] of the Wigner function is introduced [7], $R\left(x\right)\equiv\iint\limits_{-\infty}^{\infty}dx^{\prime}d\nu^{\prime}\delta\left(x-Dx^{\prime}+B\nu^{\prime}\right)W_{\psi}(\nu^{\prime},x^{\prime}),$ (4) $R\left(x\right)$ is also a probability distribution along an infinitely thin phase space strip denoted by the real parameters $D,B$. The inverse relation of (4) is the foundation of optical tomographic imaging techniques (the techniques derive two-dimensional data from a three-dimensional object to obtain a slice image of the internal structure and thus have the ability to peer inside the object noninvasively.) On the other hand, an optical system can be analyzed by using either diffraction theory or ray optics. It is worth setting up a connection between ray transfer matrix and diffraction theory, i.e., the diffraction integration written in terms of ray transfer matrix—named Collins’ diffraction integral formula [8, 9, 10] which describes the propagation of a general beam $\psi\left(x^{\prime}\right)$ through an $\left(ABCD\right)$ optical paraxial system, $\phi\left(x\right)=\int_{-\infty}^{\infty}\mathcal{K}\left(x,x^{\prime}\right)\psi\left(x^{\prime}\right)dx^{\prime},$ (5) where $\phi\left(x\right)$ is the output field, $AD-BC=1,$ and $\mathcal{K}\left(x,x^{\prime}\right)$ is the integral kernel $\mathcal{K}\left(x,x^{\prime}\right)=\frac{1}{\sqrt{2\pi iB}}\exp\left[\frac{i}{2B}\left(Ax^{\prime 2}-2x^{\prime}x+Dx^{2}\right)\right].$ (6) If the energy in the initial beam is normalized, $\int_{-\infty}^{+\infty}\left\|\psi\left(x^{\prime}\right)\right\|^{2}dx^{\prime}=1,$ then the output beam is normalized too, $\int_{-\infty}^{+\infty}\left\|\phi\left(x\right)\right\|^{2}dx=1.$ Clearly, if the $\left[ABCD\right]$ system is changed to $\left[D\left(-B\right)\left(-C\right)A\right]$ system, then Eq. (5) should read $\phi\left(x\right)=\int_{-\infty}^{\infty}\mathcal{\vec{K}}\left(x,x^{\prime}\right)\psi\left(x^{\prime}\right)dx^{\prime},$ (7) where the integral kernel $\mathcal{\vec{K}}\left(x,x^{\prime}\right)$ is $\mathcal{\vec{K}}\left(x,x^{\prime}\right)=\frac{1}{\sqrt{-2\pi iB}}\exp\left[\frac{-i}{2B}\left(Dx^{\prime 2}-2x^{\prime}x+Ax^{2}\right)\right].$ (8) In this Letter, we shall reveal the following theorem: When an input field $\psi\left(x^{\prime}\right)$ propagates through an optical $\left[D\left(-B\right)\left(-C\right)A\right]$ system, the energy density of the output field $\phi\left(x\right)$ is equal to the Radon transform of the Wigner function of the input field, where the Radon transform parameters are $D,B$. The proof is demonstrated as follows. With the use of Dirac $\delta$-function, we can re-express the Wigner function of input field $\psi\left(x^{\prime}\right)$ in Eq. (1) as $\displaystyle W_{\psi}(\nu^{\prime},x^{\prime})$ $\displaystyle=$ $\displaystyle\iint\limits_{-\infty}^{\infty}dx^{\prime\prime}dx^{\prime\prime\prime}\int_{-\infty}^{+\infty}\frac{du}{2\pi}e^{i\nu^{\prime}u}\psi^{\ast}\left(x^{\prime\prime}\right)\psi\left(x^{\prime\prime\prime}\right)\delta\left(x^{\prime\prime}-x^{\prime}-\frac{u}{2}\right)\delta\left(x^{\prime}-\frac{u}{2}-x^{\prime\prime\prime}\right)$ (9) $\displaystyle=$ $\displaystyle\frac{1}{\pi}\iint\limits_{-\infty}^{\infty}dx^{\prime\prime}dx^{\prime\prime\prime}\psi^{\ast}\left(x^{\prime\prime}\right)\psi\left(x^{\prime\prime\prime}\right)e^{i2\nu^{\prime}\left(x^{\prime\prime}-x^{\prime}\right)}\delta\left(2x^{\prime}-x^{\prime\prime}-x^{\prime\prime\prime}\right).$ Substituting (9) into (4) we rewrite the Radon transform of $W_{\psi}(\nu^{\prime},x^{\prime})$ as $\displaystyle R\left(x\right)$ $\displaystyle=$ $\displaystyle\frac{1}{\pi}\iint\limits_{-\infty}^{\infty}dx^{\prime}d\nu^{\prime}\delta\left(x-Dx^{\prime}+B\nu^{\prime}\right)\iint\limits_{-\infty}^{\infty}dx^{\prime\prime}dx^{\prime\prime\prime}\psi^{\ast}\left(x^{\prime\prime}\right)\psi\left(x^{\prime\prime\prime}\right)e^{i2\nu^{\prime}\left(x^{\prime\prime}-x^{\prime}\right)}\delta\left(2x^{\prime}-x^{\prime\prime}-x^{\prime\prime\prime}\right)$ (10) $\displaystyle=$ $\displaystyle\frac{1}{2\pi B}\iint\limits_{-\infty}^{\infty}dx^{\prime\prime}dx^{\prime\prime\prime}\psi^{\ast}\left(x^{\prime\prime}\right)\psi\left(x^{\prime\prime\prime}\right)\exp\left\\{\frac{i}{2B}[D\left(x^{\prime\prime 2}-x^{\prime\prime\prime 2}\right)-2x\left(x^{\prime\prime}-x^{\prime\prime\prime}\right)\right\\}.$ On the other hand, when the beam $\psi\left(x^{\prime}\right)$ propagates through the $\left[D\left(-B\right)\left(-C\right)A\right]$ optical system, according to the Fresnel integration (7)-(8), we have $\displaystyle\left\|\phi\left(x\right)\right\|^{2}$ $\displaystyle=$ $\displaystyle\int_{-\infty}^{\infty}dx^{\prime}\mathcal{\vec{K}}\left(x,x^{\prime}\right)\psi\left(x^{\prime}\right)\int_{-\infty}^{\infty}dx^{\prime\prime}\psi^{\ast}\left(x^{\prime\prime}\right)\mathcal{\vec{K}}^{\ast}\left(x,x^{\prime\prime}\right)$ (11) $\displaystyle=$ $\displaystyle\frac{1}{2\pi B}\int_{-\infty}^{\infty}dx^{\prime}\exp\left[\frac{-i}{2B}\left(Dx^{\prime 2}-2x^{\prime}x+Ax^{2}\right)\right]\psi\left(x^{\prime}\right)$ $\displaystyle\times\int_{-\infty}^{\infty}dx^{\prime\prime}\exp\left[\frac{i}{2B}\left(Dx^{\prime\prime 2}-2x^{\prime\prime}x+Ax^{2}\right)\right]\psi^{\ast}\left(x^{\prime\prime}\right)$ $\displaystyle=$ $\displaystyle\frac{1}{2\pi B}\iint\limits_{-\infty}^{\infty}dx^{\prime}dx^{\prime\prime}\psi\left(x^{\prime}\right)\psi^{\ast}\left(x^{\prime\prime}\right)\exp\left\\{\frac{i}{2B}\left[D\left(x^{\prime\prime 2}-x^{\prime 2}\right)+2x\left(x^{\prime}-x^{\prime\prime}\right)\right]\right\\},$ which is the same as $R\left(x\right)$ in (10). So combining (4), (10)-(11), (7) and (8) we reach the conclusion $\left\|\frac{1}{\sqrt{-2\pi iB}}\int_{-\infty}^{\infty}\exp\left[\frac{-i}{2B}\left(Dx^{\prime 2}-2x^{\prime}x+Ax^{2}\right)\right]\psi\left(x^{\prime}\right)dx^{\prime}\right\|^{2}=\iint\limits_{-\infty}^{\infty}dx^{\prime}d\nu^{\prime}\delta\left(x-Dx^{\prime}+B\nu^{\prime}\right)W_{\psi}(\nu^{\prime},x^{\prime}),$ (12) where $DA-BC=1$. The physical meaning of Eq. (12) is: when an input field propagates through an optical $\left[D\left(-B\right)\left(-C\right)A\right]$ system, the energy density of the output field is equal to the Radon transform of the Wigner function of the input field. So far as our knowledge is concerned, this conclusion seems new. Eq. (8) is the relationship between the input amplitude and output one in spatial-domain. Now we turn the above discussion to the case of space- frequency domain. For a $\left[D\left(-B\right)\left(-C\right)A\right]$ optical system, the Collins’ diffraction integral formula in space-frequency (angle spectrum [11]) domain is [12] $\widetilde{\phi}\left(v\right)=\int_{-\infty}^{\infty}\widetilde{\mathcal{K}}\left({\nu},\nu^{\prime}\right)\widetilde{\psi}\left({\nu}^{\prime}\right)d{\nu}^{\prime},$ (13) where the kernel is $\widetilde{\mathcal{K}}\left({\nu},{\nu}^{\prime}\right)=\frac{1}{\sqrt{2\pi iC}}\exp\left[\frac{i}{2C}\left(D{\nu}^{2}-2{\nu}^{\prime}{\nu}+A{\nu}^{\prime 2}\right)\right].$ (14) On the other hand, the Wigner function expressed in terms of the space- frequency field is $W_{\tilde{\psi}}(\nu^{\prime},x^{\prime})=\int_{-\infty}^{+\infty}\frac{ds}{2\pi}e^{-ix^{\prime}s}\tilde{\psi}^{\ast}\left(\nu^{\prime}+\frac{s}{2}\right)\tilde{\psi}\left(\nu^{\prime}-\frac{s}{2}\right).$ (15) In space-frequency domain the Radon transform is along an infinitely thin phase space strip denoted by the real parameters $A,C$, $R\left(\nu\right)\equiv\iint\limits_{-\infty}^{\infty}dx^{\prime}dp^{\prime}\delta\left({\nu}-A{\nu}^{\prime}+Cx^{\prime}\right)W_{\tilde{\psi}}({\nu}^{\prime},x^{\prime})$ (16) rather than the parameters $D,B$ of the Radon transform (4) in spatial-domain. We want to examine if the above conclusion still holds in the space-frequency domain. For this propose, we rewrite (15) as $\displaystyle W_{\tilde{\psi}}(\nu^{\prime},x^{\prime})$ $\displaystyle=$ $\displaystyle\iint\limits_{-\infty}^{\infty}d{\nu}^{\prime\prime}d{\nu}^{\prime\prime\prime}\int_{-\infty}^{+\infty}\frac{ds}{2\pi}e^{-ix^{\prime}s}\tilde{\psi}^{\ast}\left({\nu}^{\prime\prime}\right)\tilde{\psi}\left({\nu}^{\prime\prime\prime}\right)\delta\left(\nu^{\prime\prime}-\nu^{\prime}-\frac{s}{2}\right)\delta\left(\nu^{\prime}-\frac{s}{2}-\nu^{\prime\prime\prime}\right)$ (17) $\displaystyle=$ $\displaystyle\iint\limits_{-\infty}^{\infty}d{\nu}^{\prime\prime}d{\nu}^{\prime\prime\prime}\tilde{\psi}^{\ast}\left({\nu}^{\prime\prime}\right)\tilde{\psi}\left({\nu}^{\prime\prime\prime}\right)e^{-i2x^{\prime}\left({\nu}^{\prime\prime}-{\nu}^{\prime}\right)}\delta\left(2{\nu}^{\prime}-{\nu}^{\prime\prime}-{\nu}^{\prime\prime\prime}\right)$ Then we substitute (17) into (16), the result is $\displaystyle R\left(\nu\right)$ $\displaystyle=$ $\displaystyle\frac{1}{\pi}\iint\limits_{-\infty}^{\infty}dx^{\prime}d{\nu}^{\prime}\delta\left({\nu}-A{\nu}^{\prime}+Cx^{\prime}\right)\iint\limits_{-\infty}^{\infty}d{\nu}^{\prime\prime}d{\nu}^{\prime\prime\prime}\tilde{\psi}^{\ast}\left({\nu}^{\prime\prime}\right)\tilde{\psi}\left({\nu}^{\prime\prime\prime}\right)e^{-i2x^{\prime}\left({\nu}^{\prime\prime}-{\nu}^{\prime}\right)}\delta\left(2{\nu}^{\prime}-{\nu}^{\prime\prime}-{\nu}^{\prime\prime\prime}\right)$ (18) $\displaystyle=$ $\displaystyle\frac{1}{2\pi C}\iint\limits_{-\infty}^{\infty}d{\nu}^{\prime\prime}d{\nu}^{\prime\prime\prime}\tilde{\psi}^{\ast}\left({\nu}^{\prime\prime}\right)\tilde{\psi}\left({\nu}^{\prime\prime\prime}\right)\exp\left\\{-\frac{i}{2C}\left[A\left({\nu}^{\prime\prime 2}-{\nu}^{\prime\prime\prime 2}\right)-2{\nu}\left({\nu}^{\prime\prime}-{\nu}^{\prime\prime\prime}\right)\right]\right\\}.$ On the other hand, from (13) we calculate $\displaystyle\|\tilde{\phi}\left({\nu}\right)\|^{2}$ $\displaystyle=$ $\displaystyle\frac{1}{2\pi C}\int_{-\infty}^{+\infty}\tilde{\psi}\left({\nu}^{\prime}\right)\exp\left[\frac{i}{2C}\left(D{\nu}^{2}-2{\nu}^{\prime}{\nu}+A{\nu}^{\prime 2}\right)\right]d{\nu}^{\prime}$ (19) $\displaystyle\times\int_{-\infty}^{+\infty}\tilde{\psi}^{\ast}\left({\nu}^{\prime\prime}\right)\exp\left[\frac{-i}{2C}\left(D{\nu}^{2}-2{\nu}^{\prime\prime}{\nu}+A{\nu}^{\prime\prime 2}\right)\right]d{\nu}^{\prime\prime}$ $\displaystyle=$ $\displaystyle\frac{1}{2\pi C}\int\int_{-\infty}^{+\infty}d{\nu}^{\prime}d{\nu}^{\prime\prime}\tilde{\psi}\left({\nu}^{\prime}\right)\tilde{\psi}^{\ast}\left({\nu}^{\prime\prime}\right)\exp\left[-\frac{i}{2C}\left(A\left({\nu}^{\prime\prime 2}-{\nu}^{\prime 2}\right)-2{\nu}\left({\nu}^{\prime\prime}-{\nu}^{\prime}\right)\right)\right].$ Comparing Eq. (19) with Eq. (18) and using (13)-(14) we see $\left\|\frac{1}{\sqrt{2\pi iC}}\int\exp\left[\frac{i}{2C}\left(D{\nu}^{2}-2{\nu}^{\prime}{\nu}+A{\nu}^{\prime 2}\right)\right]\tilde{\psi}\left({\nu}^{\prime}\right)d{\nu}^{\prime}\right\|^{2}=\iint\limits_{-\infty}^{\infty}dx^{\prime}dp^{\prime}\delta\left({\nu}-A{\nu}^{\prime}+Cx^{\prime}\right)W_{\tilde{\psi}}({\nu}^{\prime},x^{\prime}),$ (20) this is the theorem expressed in space-frequency domain. We now take an example to confirm the theorem. When the input field is described by a Gaussian function (Gaussian chirplet) $\psi_{0}\left(x^{\prime}\right)\equiv\sqrt[4]{\frac{\epsilon}{\pi}}\exp\left\\{-\left(\epsilon-i\beta\right)\frac{x^{\prime 2}}{2}\right\\},\text{ }\epsilon>0,\text{\ }$ (21) which depicts a Gaussian windowed linear chirp signal [13]. Using Eq.(1) we obtain its Wigner function $\displaystyle W_{\psi_{0}}(\nu^{\prime},x^{\prime})$ $\displaystyle=$ $\displaystyle e^{-\epsilon x^{\prime}{}^{2}}\sqrt{\frac{\epsilon}{\pi}}\int_{-\infty}^{+\infty}\frac{du}{2\pi}e^{-\frac{1}{4}\epsilon u^{2}+i\left(\nu^{\prime}-\beta x^{\prime}\right)u}$ (22) $\displaystyle=$ $\displaystyle\frac{1}{\pi}\exp\left\\{-\left[\epsilon x^{\prime}{}^{2}+\frac{1}{\epsilon}\left(\nu^{\prime}-\beta x^{\prime}\right)^{2}\right]\right\\},$ which shows that the energy of Gaussian chirplet is concentrated at $\nu^{\prime}=\beta x^{\prime}.$ According to Eq. (4), the Radon transform of (22) with the parameters $D,B$ is $\displaystyle\iint\limits_{-\infty}^{\infty}dx^{\prime}d\nu^{\prime}\delta\left(x-Dx^{\prime}+B\nu^{\prime}\right)W_{\psi_{0}}(\nu^{\prime},x^{\prime})$ (23) $\displaystyle=$ $\displaystyle\frac{1}{\pi D}\int_{-\infty}^{+\infty}d\nu^{\prime}\exp\left\\{-\frac{\epsilon}{D^{2}}\left(x+B\nu^{\prime}\right){}^{2}-\frac{1}{\epsilon D^{2}}\left[\left(D-\beta B\right)\nu^{\prime}-\beta x\right]^{2}\right\\}$ $\displaystyle=$ $\displaystyle\sqrt{\frac{\epsilon/\pi}{\left(D-B\beta\right)^{2}+B^{2}\epsilon^{2}}}\exp\left[\frac{-\epsilon x^{2}}{\left(D-B\beta\right)^{2}+B^{2}\epsilon^{2}}\right]\equiv R_{0}\left(x\right).$ On the other hand, according to (7) and (8), the Fresnel transform of input field $W_{\psi_{0}}(\nu^{\prime},x^{\prime})$ through a $\left[D\left(-B\right)\left(-C\right)A\right]$ optical system is given by $\displaystyle\frac{1}{\sqrt{-2\pi iB}}\sqrt[4]{\frac{\epsilon}{\pi}}\int_{-\infty}^{\infty}dx^{\prime}\exp\left[\frac{-i}{2B}\left(Dx^{\prime 2}-2x^{\prime}x+Ax^{2}\right)-\left(\epsilon-i\beta\right)\frac{x^{\prime 2}}{2}\right]$ (24) $\displaystyle=$ $\displaystyle\sqrt[4]{\frac{\epsilon}{\pi}}\frac{1}{\sqrt{-i}}e^{-\frac{iA}{2B}x^{2}}\sqrt{\frac{1}{\epsilon B-i\left(\beta B-D\right)}}\exp\left[\frac{-x^{2}}{2B\left(B\epsilon+iD-i\beta B\right)}\right]\left.\equiv\right.\phi_{0}\left(x\right).$ Comparing (24) with (23) and noticing $\frac{1}{2B\left(B\epsilon+iD-i\beta B\right)}+\frac{1}{2B\left(B\epsilon- iD+i\beta B\right)}=\allowbreak\frac{\epsilon}{\left(D-B\beta\right)^{2}+B^{2}\epsilon^{2}}$ (25) we see $\left\|\phi_{0}\left(x\right)\right\|^{2}=R_{0}\left(x\right),$ (26) as expected. In summary, we have derived a new theorem governing the connection between optical field’s Fresnel transform and its Wigner function’s Radon transformation, since both the Wigner function and the Fresnel transform are widely used in optical propagation, we hope this theorem would have new applications in the analysis of optical communication and optical tomography. For the application of Wigner function in deriving the quantum-mechanical photocount formula, we refer to [14]. This work was supported by the National Natural Science Foundation of China under grant 10775097. ## References [1] E. P. Wigner, ”On the quantum correction for thermodynamic equilibrium”, Phys. Rev. 40, 749 (1932). * [2] M. j. Bastiaans, ”The Wigner distribution function applied to optical signals and systems”, Opt. Commun. 25, 26 (1978). * [3] M. j. Bastiaans, ”Wigner distribution function and its application to first-order optics”, J. Opt. Soc. Am. 69, 1710 (1979). * [4] A. W. Lohmann, “Image rotation, Wigner rotation, and the fractional Fourier transform,” J. Opt. Soc. Am. A 10, 2181–2186 (1993). * [5] J. Radon, ”Uber die Bestimmung von Funktionen Durch Ihre Integralwerte Langs Gewisser Mannigfaltigkeiten”, Ber. Verh. Saechs. Akad. Wiss. Leipzig Math.Phys.K1. 69, 262-267, (1917). * [6] Y. Zhang, B. Gu, B. Dong, and G. Yang, ”Optical implementations of the Radon–Wigner display for one-dimensional signals,” Opt. Lett. 23, 1126-1128 (1998). * [7] For quantum states’ reconstruction, see e.g., Wolfgang P. Schleich, Quantum Optics in Phase Space, (Wiley-VCH, Birlin, 2001) and references therein * [8] S. A. Collins, ”Lens-system diffraction integral written in terms of matrix optic”, J. Opt. Soc. Am. A 60, 1168-1177 (1970). * [9] J. A. Arnaud, ”Mode Coupling in First-Order Optics,” J. Opt. Soc. Am. 61, 751-758 (1971). * [10] H. Fan and H. Lu, ”Collins diffraction formula studied in quantum optics,” Opt. Lett. 31, 2622-2624 (2006). * [11] J. W. Goodman, Introduction to Fourier Optics, (Mc Graw-Hill, 1968). * [12] Z. Liu, X. Xiu and D. Fan, ”Collins formula in frequency-domain and fractional Fourier transforms”, Opt. Commun. 155, 7 (1998). * [13] A. Papoulis, Signal Analysis, (McGraw-Hill, New York, 1977). * [14] Hong-yi Fan and Li-yun Hu, ”Two quantum-mechanical photocount formulas”, Opt. Lett. 33, 443-445 (2008).	arxiv-papers	2008-08-21T13:14:06	2024-09-04T02:48:57.375353	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Hong-yi Fan and Li-yun Hu", "submitter": "Liyun Hu", "url": "https://arxiv.org/abs/0808.2926" }
0808.2933	# Rare decay $\pi^{0}\to e^{+}e^{-}$ constraints on the light CP-odd Higgs in NMSSM Qin Changa,b, Ya-Dong Yanga,c aInstitute of Particle Physics, Huazhong Normal University, Wuhan, Hubei 430079, P. R. China bDepartment of Physics, Henan Normal University, Xinxiang, Henan 453007, P. R. China cKey Laboratory of Quark $\&$ Lepton Physics, Ministry of Education, P.R. China ###### Abstract We constrain the light CP-odd Higgs $A_{1}^{0}$ in NMSSM via the rare decay $\pi^{0}\to e^{+}e^{-}$. It is shown that the possible $3\sigma$ discrepancy between theoretical predictions and the recent KTeV measurement of ${\cal B}({\pi}^{0}\to e^{+}e^{-})$ cannot be resolved when the constraints from $\Upsilon\to\gamma A_{1}^{0}$, $a_{\mu}$ and $\pi^{0}\to\gamma\gamma$ are combined. Furthermore, the combined constraints also exclude the scenario involving $m_{A_{1}^{0}}=214.3$ MeV, which is invoked to explain the anomaly in the $\Sigma^{+}\to p\mu^{+}\mu^{-}$ decay found by the HyperCP Collaboration. PACS Numbers: 13.25.Cq, 14.80.Cp ## 1 Introduction Theoretically, the rare decay $\pi^{0}\to e^{+}e^{-}$ starts at the one loop level in the standard model (SM), which has been extensively studied [1, 2, 3, 4, 5, 6, 7, 8, 9, 10] since the first investigation in QED by Drell [1]. It is nontrivial to make precise predictions of the branching ratio ${\cal B}_{SM}(\pi^{0}\to e^{+}e^{-})$ because its sub-process involves the $\pi^{0}\to\gamma^{}\gamma^{}$ transition form factor. In Refs.[2, 3, 4, 5], the decay was studied via the Vector-Meson Dominance (VMD) approach, where the results are in good agreement with each other and converge in ${\cal B}(\pi^{0}\to e^{+}e^{-})\sim 6.2-6.4\times 10^{-8}$. By using the measured value of ${\cal B}(\eta\to\mu^{+}\mu^{-})$ to fix the counterterms of the chiral amplitude in Chiral Perturbation Theory (ChPT), Savage et al. predicted ${\cal B}(\pi^{0}\to e^{+}e^{-})=(7\pm 1)\times 10^{-8}$ [6]. Using a procedure similar to that used in Ref.[6] (although with an updated measurement of ${\cal B}(\eta\to\mu^{+}\mu^{-})$), Dumm and Pich predicted $(8.3\pm 0.4)\times 10^{-8}$ [7]. Alternatively, using the lowest meson dominance (LMD) approximation to the large-$N_{c}$ spectrum of vector meson resonances to fix the counterterms, Knecht et al. predicted $(6.2\pm 0.3)\times 10^{-8}$ [8], which is about $4\sigma$ lower than the value predicted by Ref.[7] but which agrees with the others. Most recently, using a dispersive approach to the amplitude and the experimental results of the CELLO [11] and CLEO [12] Collaborations for the pion transition form factor, Dorokhov and Ivanov [9] have found that ${\cal B}_{SM}(\pi^{0}\to e^{+}e^{-})\,=\,(6.23\,\pm\,0.09)\times 10^{-8},$ (1) which is consistent with most theoretical predictions of ${\cal B}_{SM}(\pi^{0}\to e^{+}e^{-})$ in the literature. Moreover, their prediction that ${\cal B}(\eta\to\mu^{+}\mu^{-})=(5.11\pm 0.2)\times 10^{-6}$ agrees with the experimental data (which gives a value of $(5.8\pm 0.8)\times 10^{-6}$ [13]). Experimentally, the accuracy of the measurements of the decay has increased significantly since the first $\pi^{0}\to e^{+}e^{-}$ evidence was observed by the Geneva-Saclay group [14] in 1978 with ${\cal B}_{SM}(\pi^{0}\to e^{+}e^{-})=(22^{+24}_{-11})\times 10^{-8}$. A detailed summary of the experimental situation can be found in Ref.[15]. Recently, using the complete data set from KTeV E799-II at Fermilab, the KTeV Collaboration has made a precise measurement of the $\pi^{0}\to e^{+}e^{-}$ branching ratio [16] ${\cal B}^{no-rad}_{KTeV}(\pi^{0}\to e^{+}e^{-})\,=\,(7.48\,\pm\,0.29\,\pm\,0.25)\times 10^{-8},$ (2) after extrapolating the full radiative tail beyond $(m_{e^{+}e^{-}}/m_{\pi^{0}})^{2}>0.95$ and scaling their result back up by the overall radiative correction of $3.4\%$. As was already noted in Ref. [9], the SM prediction given in Eq.(1) is $3.3\sigma$ lower than the KTeV data. The authors have also compared their result with estimations made by various approaches in the literature and found good agreements. Further analyses have found that QED radiative contributions [17] and mass corrections [18] are at the level of a few percent and are therefore unable to reduce the discrepancy. Although the discrepancy might be due to hadronic dynamics that are as of yet unknown, it is equally possible that this discrepancy is caused by the effects of new physics (NP). In this Letter we will study the latter possibility. As is known that leptonic decays of pseudoscalar mesons are sensitive to pseudoscalar weak interactions beyond the SM. Precise measurements and calculations of these decays will offer sensitive probes for NP effects at the low energy scale. Of particular interest to us is the rare decay $\pi^{0}\to e^{+}e^{-}$, which could proceed at tree level via a flavor-conserving process induced by a light pseudoscalar Higgs boson $A^{0}_{1}$ in the next-to-minimal supersymmetric standard model (NMSSM) [19]. We will look for a region of the parameter space of NMSSM that could resolve the aforementioned discrepancy of ${\cal B}(\pi^{0}\to e^{+}e^{-})$ at $1\sigma$. Then, we combine constraints from $a_{\mu}$ and the recent searches for $\Upsilon(1S),(3S)\to\gamma A^{0}_{1}$ by CLEO [20] and BaBar [21], respectively. ## 2 The amplitude of $\pi^{0}\to e^{+}e^{-}$ in the SM and the NMSSM The NMSSM has generated considerable interest in the literature, which extends the minimal supersymmetric SM (MSSM) by introducing a new Higgs singlet chiral superfield $\hat{S}$ to solve the known $\mu$ problem in MSSM. The superpotential in the model is [19] $W_{NMSSM}=\hat{Q}\hat{H}_{u}h_{u}\hat{U}^{C}+\hat{H}_{d}\hat{Q}h_{d}\hat{D}^{C}+\hat{H}_{d}\hat{L}h_{e}\hat{E}^{C}+\lambda\hat{S}\hat{H}_{u}\hat{H}_{d}+\frac{1}{3}\kappa\hat{S}^{3},$ (3) where $\kappa$ is a dimensionless constant and measures the size of Peccei- Quinn (PQ) symmetry breaking. In addition to the two charged Higgs bosons, $H^{\pm}$, the physical NMSSM Higgs sector consists of three scalars $h^{0},\,H_{1,2}^{0}$ and two pseudoscalars $A_{1,2}^{0}$. As in the MSSM, $\tan\beta=v_{u}/v_{d}$ is the ratio of the Higgs doublet vacuum expectation values $v_{u}=\langle H_{u}^{0}\rangle=v\sin\beta$ and $v_{d}=\langle H_{d}^{0}\rangle=v\cos\beta$, where $v=\sqrt{v_{d}^{2}+v_{u}^{2}}=\sqrt{2}m_{W}/g\simeq 174GeV$. Generally, the masses and singlet contents of the physical fields depend strongly on the parameters of the model (such as, in particular, how well the PQ symmetry is broken). If the PQ symmetry is slightly broken, then $A_{1}^{0}$ can be rather light, and its mass is given by $m^{2}_{A^{0}_{1}}=3\kappa xA_{k}+{\cal O}(\frac{1}{\tan\beta})$ (4) with the vacuum expectation value of the singlet $x=\langle S\rangle$; meanwhile, another pseudoscalar $A_{2}^{0}$ has a mass of order of $m_{H^{\pm}}$. For $\pi^{0}\to e^{+}e^{-}$ decay, the NMSSM contributions are dominated by $A_{1}^{0}$. The couplings of $A_{1}^{0}$ to fermions are [22] $\mathcal{L}_{A_{i}^{0}f\bar{f}}=-i\frac{g}{2m_{W}}\Big{(}X_{d}m_{d}\bar{d}\gamma_{5}d+X_{u}m_{u}\bar{u}\gamma_{5}u+X_{\ell}m_{\ell}\bar{\ell}\gamma_{5}\ell\Big{)}A_{1}^{0}$ (5) where $X_{d}=X_{\ell}=\frac{v}{x}\delta_{-}$ and $X_{u}=X_{d}/\tan^{2}\beta$; thus, the contribution of the $\bar{u}\gamma_{5}uA_{1}^{0}$ term in $\pi^{0}\to e^{+}e^{-}$ could be neglected in the large $\tan\beta$ approximation. Figure 1: Relevant Feynman diagram within NMSSM. To the leading order, the relevant Feynman diagram within NMSSM is shown in Fig. 1. We obtain its amplitude as $\mathcal{M}_{A_{1}^{0}}=-\frac{G_{F}}{\sqrt{2}}m_{e}m_{\pi^{0}}^{3}f_{\pi^{0}}\frac{1}{m_{\pi^{0}}^{2}-m_{A_{1}^{0}}^{2}}X_{d}^{2},$ (6) which is independent of $m_{d}$, since $m_{d}$ in the coupling of $A^{0}_{1}\bar{d}\gamma_{5}d$ is canceled by the $m_{d}$ term of the hadronic matrix $\langle\,0\,\|\,\bar{d}\gamma_{5}d\,\|\,\pi^{0}\rangle\,=-\frac{i}{\sqrt{2}}f_{\pi^{0}}\frac{m_{\pi^{0}}^{2}}{2m_{d}}\,.$ (7) In the SM, the normalized branching ratio of $\pi^{0}\to e^{+}e^{-}$ is given by [9] $R(\pi^{0}\to e^{+}e^{-})=\frac{{\cal B}(\pi^{0}\to e^{+}e^{-})}{{\cal B}(\pi^{0}\to\gamma\gamma)}=2\left(\frac{\alpha_{e}}{\pi}\frac{m_{e}}{m_{\pi^{0}}}\right)^{2}\beta_{e}(m^{2}_{\pi^{0}})\|{\cal A}(m^{2}_{\pi^{0}})\|^{2}$ (8) where $\beta_{e}(m^{2}_{\pi^{0}})=\sqrt{1-4\frac{m_{e}^{2}}{m_{\pi^{0}}^{2}}}$ and ${\cal A}(m^{2}_{\pi^{0}})$ is the reduced amplitude. To add the NMSSM amplitude to the above amplitudes consistently, we rederive the SM amplitude to look into possible differences between the conventions used in our Letter and the ones used in Ref. [9]. Figure 2: Triangle diagram for $\pi^{0}\to e^{+}e^{-}$ process. The Feynman diagram that proceeds via two photon intermediate states is shown in Fig. 2. We start with the $\pi^{0}\gamma^{}\gamma^{}$ vertex $H_{\mu\nu}=-i\,e^{2}\,\epsilon_{\mu\nu\alpha\beta}\,k^{\alpha}\,(q-k)^{\beta}f_{\gamma^{\ast}\gamma^{\ast}}\,F_{\pi^{0}\gamma^{\ast}\gamma^{\ast}}(k^{2},(q-k)^{2})$ (9) where $k$ and $q-k$ are the momenta of the two photons, $f_{\gamma^{\ast}\gamma^{\ast}}=\frac{\sqrt{2}}{4\pi^{2}\,andf_{\pi^{0}}}$ is the coupling constant of $\pi^{0}$ to two real photons. $F_{\pi^{0}\gamma^{\ast}\gamma^{\ast}}(k^{2},(q-k)^{2})$ is the transition form factor $\pi^{0}\to\gamma^{\ast}\gamma^{\ast}$, which is normalized to $F_{\pi^{0}\gamma^{\ast}\gamma^{\ast}}(0,0)=1$. The amplitude of Fig. 2 is written as $\mathcal{M}_{SM}(\pi^{0}\to e^{+}e^{-})=ie^{2}\int\frac{d^{4}k}{(2\pi)^{4}}\frac{L^{\mu\nu}H_{\mu\nu}}{(k^{2}+i\varepsilon)\big{(}(k-q)^{2}+i\varepsilon\big{)}\big{(}(k-p)^{2}-m_{e}+i\varepsilon\big{)}},$ (10) with $L^{\mu\nu}\,=\,\bar{u}(p,s)\gamma^{\mu}(\not\\!p\,-\not\\!k\,+m_{e})\gamma^{\nu}v(q-p,s^{\prime}).$ (11) There is a known, convenient way to calculate $L^{\mu\nu}$ with the projection operator for the outgoing $e^{+}e^{-}$ pair system[23] $\displaystyle\mathcal{P}(q-p,p)$ $\displaystyle=$ $\displaystyle\frac{1}{\sqrt{2}}\big{[}v(q-p,+)\otimes\bar{u}(p,-)+v(q-p,-)\otimes\bar{u}(p,+)\big{]}$ (12) $\displaystyle=$ $\displaystyle\frac{1}{2\sqrt{2t}}\big{[}-2m_{e}q_{\mu}\gamma^{\mu}\gamma^{5}+\frac{1}{2}\epsilon_{\mu\nu\sigma\tau}\big{(}p^{\sigma}(q-p)^{\tau}-(q-p)^{\sigma}p^{\tau}\big{)}\sigma^{\mu\nu}+t\gamma^{5}\big{]}$ where $t=q^{2}=m_{\pi^{0}}^{2}$. After some calculations, we get $\mathcal{M}_{SM}(\pi^{0}\to e^{+}e^{-})=2\sqrt{2}\,\alpha^{2}\,m_{e}\,m_{\pi^{0}}\,f_{\gamma^{\ast}\gamma^{\ast}}\,A(m^{2}_{\pi})$ (13) where the reduced amplitude $A(q^{2})$ is ${\cal A}(q^{2})=\frac{2\,i}{q^{2}}\int\frac{d^{4}k}{\pi^{2}}\,\frac{k^{2}\,q^{2}\,-\,(q\cdot k)^{2}}{(k^{2}+i\varepsilon)\big{(}(k-q)^{2}+i\varepsilon\big{)}\big{(}(k-p)^{2}-m_{e}+i\varepsilon\big{)}}F_{\pi^{0}\gamma^{\ast}\gamma^{\ast}}(k^{2},(q-k)^{2}).$ (14) We note that the ${\cal A}(q^{2})$ derived here is in agreement with Ref. [9]. Further evaluation of the integrals of ${\cal A}(q^{2})$ is quite subtle and lengthy [2, 24], and only the imaginary part of ${\cal A}(m_{\pi^{0}}^{2})$ can be obtained model-independently[1, 2]. In the following calculations, we quote the result of Ref. [9], ${\cal A}(m_{\pi}^{2})=(10.0\pm 0.3)-i17.5.$ (15) With Eq. (6) and Eq. (13), we get the total amplitude $\mathcal{M}=2\sqrt{2}\,\alpha^{2}\,m_{e}\,m_{\pi^{0}}\,f_{\gamma^{\ast}\gamma^{\ast}}\,A(m^{2}_{\pi})-\frac{G_{F}}{\sqrt{2}}m_{e}m_{\pi^{0}}^{3}f_{\pi^{0}}\frac{1}{m_{\pi^{0}}^{2}-m_{A_{1}^{0}}^{2}}X_{d}^{2}.$ (16) ## 3 Numerical analysis and discussion Now, we are ready to discuss the effects of $A_{1}^{0}$ numerically, with a focus on the $m_{A^{0}_{1}}<2m_{b}$ scenarios. The dependence of ${\cal B}(\pi^{0}\to e^{+}e^{-})$ on the parameter $\|X_{d}\|$ is shown in Fig. 3 with $m_{A_{1}^{0}}=m_{\pi}/2,\,214.3{\rm MeV},\,3{\rm GeV}$ as benchmarks. We have used the input parameters ${\cal B}(\pi^{0}\to\gamma\gamma)=0.988$ and $f_{\pi^{0}}=(130.7\pm 0.4)~{}{\rm MeV}$ [13]. As shown in Fig. 3, ${\cal B}(\pi^{0}\to e^{+}e^{-})$ is very sensitive to the parameter $\|X_{d}\|$ and $m_{A_{1}^{0}}$. For $m_{A_{1}^{0}}<m_{\pi^{0}}$, the NMSSM contribution is deconstructive and reduces ${\cal B}(\pi^{0}\to e^{+}e^{-})$ at small $\|X_{d}\|$ region. For $m_{A_{1}^{0}}>m_{\pi^{0}}$, the NMSSM contribution is constructive and could enhance ${\cal B}(\pi^{0}\to e^{+}e^{-})$ to be consistent with the KTeV measurement ${\cal B}^{no-rad}_{KTeV}(\pi^{0}\to e^{+}e^{-})=(7.48\pm 0.38)\times 10^{-8}$ (where $\|X_{d}\|$ strongly depends on $m_{A^{0}_{1}}$). Figure 3: The dependence of ${\cal B}(\pi^{0}\,{\to}\,e^{+}e^{-})$ on the parameter $\|X_{d}\|$ with $m_{A_{1}^{0}}=m_{\pi^{0}}/2$, $214.3{\rm~{}MeV}$ and $\rm 3~{}GeV$, respectively. The horizontal lines are the KTeV data, where the solid line is the central value and the dashed ones are the error bars ($1\sigma$). I. Constraint on the scenario of $m_{A^{0}_{1}}=214.3{\rm MeV}$ It is interesting to note that the HyperCP Collaboration [25] has observed three events for the decay $\Sigma^{+}\to p\mu^{+}\mu^{-}$ with a narrow range of dimuon masses. This may indicate that the decay proceeds via a neutral intermediate state, $\Sigma^{+}\to pP^{0},P^{0}\to\mu^{+}\mu^{-}$, with a $P^{0}$ mass of $214.3\pm 0.5{\rm MeV}$. The possibility of $P^{0}$ has been explored in the literature [26, 28, 29, 30]. The authors have proposed $A_{1}^{0}$ as a candidate for the $P^{0}$, and have also shown that their explanation could be consistent with the constraints provided by K and B meson decays [26, 27]. It would be worthwhile to check on whether the explanation could be consistent with the $\pi^{0}\to e^{+}e^{-}$ decay. Taking $m_{A_{1}^{0}}=214.3{\rm~{}MeV}$, we find that ${\cal B}(\pi^{0}\to e^{+}e^{-})$ is enhanced rapidly and could be consistent with the KTeV data within $1\sigma$ for $\|X_{d}\|=14.0\pm 2.4.$ (17) However, the upper bound $\|X_{d}\|<1.2$ from the $a_{\mu}$ constraint has been derived and used in the calculations of Ref. [26, 29]. So, with the assumption that $m_{A_{1}^{0}}=214.3{\rm~{}MeV}$, our result of $\|X_{d}\|$ violates the upper bound with a significance of $5\sigma$. Recently, CLEO [20] and BaBar [21] have searched for the CP-odd Higgs boson in radiative decays of $\Upsilon(1S)\to\gamma A^{0}_{1}$ and $\Upsilon(3S)\to\gamma A^{0}_{1}$, respectively. For $m_{A_{1}^{0}}=214{\rm MeV}$, CLEO gives the upper limit ${\cal B}(\Upsilon(1S)\to\gamma A_{1}^{0})<2.3\times 10^{-6}~{}~{}~{}~{}(90\%~{}C.L.)$ (18) which constrains $\|X_{d}\|<0.16$. The BaBar Collaboration has searched for $A_{1}^{0}$ through $\Upsilon(3S)\to\gamma A_{1}^{0}$, $A_{1}^{0}\to\textrm{invisible}$ in the mass range $m_{A_{1}^{0}}\leq 7.8{\rm~{}GeV}$ [21]. From Fig. 5 of Ref. [21], we read ${\cal B}(\Upsilon(3S)\to\gamma A_{1}^{0})\times{\cal B}(A_{1}^{0}\to\textrm{invisible})\lesssim 3.5\times 10^{-6}~{}~{}~{}~{}(90\%~{}C.L.)$ (19) for $m_{A_{1}^{0}}=214{\rm~{}MeV}$. Assuming ${\cal B}(A_{1}^{0}\to\textrm{invisible})\sim 1$, we get the conservative upper limit $\|X_{d}\|<0.19$ All of these upper limits are much lower than the limit of Eq.17 set by $\pi^{0}\to e^{+}e^{-}$; therefore, the scenario where $m_{A_{1}^{0}}\simeq 214{\rm~{}MeV}$ in NMSSM could be excluded by combining the constraints from $\pi^{0}\to e^{+}e^{-}$ and the direct searches for $\Upsilon$ radiative decays. II. Constraints on the parameter space of $m_{A^{0}_{1}}-\|X_{d}\|$ To show the constraints on NMSSM parameter space from $\pi^{0}\to e^{+}e^{-}$, we present a scan of $m_{A^{0}_{1}}-\|X_{d}\|$ space, as shown in Fig. 4. In order to scan the region of $m_{A_{1}^{0}}\sim m_{\pi^{0}}$, the amplitude of the $A_{1}^{0}$ contribution in Eq. (6) is replaced by the Breit-Wigner formula $\mathcal{M}_{A_{1}^{0}}\,=-\,\frac{G_{F}}{\sqrt{2}}m_{e}m_{\pi^{0}}^{3}f_{\pi^{0}}\frac{1}{m_{\pi^{0}}^{2}-m_{A_{1}^{0}}^{2}+i\Gamma(A_{1}^{0})m_{A_{1}^{0}}}X_{d}^{2}\,.$ (20) With the assumption that $A_{1}^{0}$ just decays to electron and photon pairs for $m_{A_{1}^{0}}\sim m_{\pi^{0}}$, the decay width of $A_{1}^{0}$ could be written as $\Gamma(A_{1}^{0})=\Gamma(A_{1}^{0}\to e^{+}e^{-})+\Gamma(A_{1}^{0}\to\gamma\gamma)$ (21) with $\displaystyle\Gamma(A_{1}^{0}\to e^{+}e^{-})$ $\displaystyle=$ $\displaystyle\frac{\sqrt{2}G_{F}}{8\pi}m_{e}^{2}m_{A_{1}^{0}}X_{d}^{2}\sqrt{1-4\frac{m_{e}^{2}}{m_{A_{1}^{0}}^{2}}}\,,$ $\displaystyle\Gamma(A_{1}^{0}\to\gamma\gamma)$ $\displaystyle=$ $\displaystyle\frac{G_{F}\alpha^{2}}{8\sqrt{2}\pi^{3}}m_{A_{1}^{0}}^{3}X_{d}^{2}\|\sum_{i}rQ_{i}^{2}k_{i}F(k_{i})\|^{2},$ (22) where $r=1$ for leptons and $r=N_{c}$ for quarks, $k_{i}=m_{i}^{2}/m_{A_{1}^{0}}^{2}$ and $Q_{i}$ is the charge of the fermion in the loop. The loop function $F(k_{i})$ reads [31] $F(k_{i})=\left\\{\begin{array}[]{ll}-2\big{(}\arcsin\frac{1}{2\sqrt{k_{i}}}\big{)}^{2}\,~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}~{}\textrm{for}\,k_{i}\geq\frac{1}{4}\,,\\\ \frac{1}{2}\Big{[}\ln\big{(}\frac{1+\sqrt{1-4k_{i}}}{1-\sqrt{1-4k_{i}}}\big{)}+i\pi\Big{]}^{2}\,~{}~{}~{}~{}~{}~{}\textrm{for}\,k_{i}<\frac{1}{4}\,.\end{array}\right.$ Figure 4: Constraints on the NMSSM parameter space through ${\cal B}(\pi^{0}{\to}e^{+}e^{-})$, ${\cal B}(\Upsilon(1S)\to\gamma A_{1}^{0})$, ${\cal B}(\Upsilon(3S)\to\gamma A_{1}^{0})$ and $a_{\mu}$ respectively. The shaded regions are allowed by the labeled processes. As shown in Fig. 4, only two narrow connected bands of the $\|X_{d}\|-m_{A^{0}_{1}}$ space survive after the KTeV measurement of ${\cal B}(\pi^{0}{\to}e^{+}e^{-})$, which show that $\pi^{0}{\to}e^{+}e^{-}$ is very sensitive to NP scenarios with a light pseudoscalar neutral boson. In the following, we will determine which part of the remaining parameter space could satisfy the constraints enforced by radiative $\Upsilon$ decays and $a_{\mu}$ simultaneously. To include the $a_{\mu}$ constraint, we use the experimental result that [32] $a_{\mu}(Exp)=(11659208.0\pm 6.3)\times 10^{-10}$ and the SM prediction [33] $a_{\mu}(SM)=(11659177.8\pm 6.1)\times 10^{-10}$. The discrepancy is $\triangle\,a_{\mu}=a_{\mu}(Exp)-a_{\mu}(SM)=(30.2\pm 8.8)\times 10^{-10}(3.4\sigma)$ (23) which is established at a $3.4\sigma$ level of significance. The contributions of $A_{1}^{0}$ to $a_{\mu}$ are given by [34] $\displaystyle{\delta}a_{\mu}({A_{1}^{0}})$ $\displaystyle=$ $\displaystyle{\delta}a_{\mu}^{1-loop}({A_{1}^{0}})+{\delta}a_{\mu}^{2-loop}({A_{1}^{0}})\,,$ (24) $\displaystyle{\delta}a_{\mu}^{1-loop}({A_{1}^{0}})\,$ $\displaystyle=$ $\displaystyle\,-\sqrt{2}G_{F}\frac{m_{\mu}^{2}}{8\pi^{2}}\|X_{d}\|^{2}f_{1}\big{(}\frac{m_{A_{1}^{0}}^{2}}{m_{\mu}^{2}}\big{)}\,,$ $\displaystyle{\delta}a_{\mu}^{2-loop}({A_{1}^{0}})\,$ $\displaystyle=$ $\displaystyle\,\sqrt{2}G_{F}\alpha\frac{m_{\mu}^{2}}{8\pi^{3}}\|X_{d}\|^{2}\Big{[}\frac{4}{3}\frac{1}{\tan^{2}\beta}f_{2}\big{(}\frac{m_{t}^{2}}{m_{A_{1}^{0}}^{2}}\big{)}+\frac{1}{3}f_{2}\big{(}\frac{m_{b}^{2}}{m_{A_{1}^{0}}^{2}}\big{)}+f_{2}\big{(}\frac{m_{\tau}^{2}}{m_{A_{1}^{0}}^{2}}\big{)}\Big{]}$ with $\displaystyle f_{1}(z)$ $\displaystyle=$ $\displaystyle\int_{0}^{1}dx\frac{x^{3}}{z(1-x)+x^{2}}\,,$ (25) $\displaystyle f_{2}(z)$ $\displaystyle=$ $\displaystyle z\int_{0}^{1}dx\frac{1}{x(1-x)-z}\ln\frac{x(1-x)}{z}.$ It has been found that the $A^{0}_{1}$ contribution is always negative at the one loop level and worsens the discrepancy in $a_{\mu}$; however, it could be positive and dominated by the two loop contribution for $A^{0}_{1}>3{\rm GeV}$[34]. One should note that there are other contributions to $a_{\mu}$ in NMSSM; for instance, the chargino/sneutino and neutralino/smuon loops. Moreover, the discrepancy $\triangle a_{\mu}$ could be resolved without pseudoscalars [34]. So, putting a constraint on $\|X_{d}\|$ via $a_{\mu}$ is a rather model-dependent process. There are two approximations with different emphases on the role of $A^{0}_{1}$; namely, (i) assuming that $\triangle a_{\mu}$ is resolved by other contributions and requiring that $A_{1}^{0}$ contributions are smaller than the $1\sigma$ error-bar of the experimental measurement, and (ii) assuming that the $A_{1}^{0}$ contributions are solely responsible for $\triangle a_{\mu}$. In Ref. [26], approximation (i) has been used to derive an upper bound of $\|X_{d}\|<1.2$. We present the $a_{\mu}$ constraints with the two approximations which are shown in Figs. 4(a) and (b), respectively. From Fig. 4(a), we can find that there are two narrow overlaps between the constraints provided by $a_{\mu}$ and ${\cal B}(\pi^{0}{\to}e^{+}e^{-})$: one is for $m_{A_{1}^{0}}\sim 3{\rm~{}GeV}$ with $\|X_{d}\|>150$ and another one is for $m_{A_{1}^{0}}\sim 135{\rm~{}MeV}$ with $\|X_{d}\|<1$. In the searches for $\Upsilon\to\gamma A^{0}_{1}$ decays, CLEO [20] obtains the upper limits for the product of ${\cal B}(\Upsilon(1S)\to\gamma A^{0}_{1})$ and ${\cal B}(A^{0}_{1}\to\tau^{+}\tau^{-})$ or ${\cal B}(A^{0}_{1}\to\mu^{+}\mu^{-})$, while BaBar presents upper limits on ${\cal B}(\Upsilon(3S)\to\gamma A^{0}_{1})\times{\cal B}(A^{0}_{1}\to invisible)$. All these limits fluctuate with the mass of $A^{0}_{1}$ frequently. For simplicity, we take the loosest upper limit ${\cal B}(\Upsilon(1S)\to\gamma A^{0}_{1})\times{\cal B}(A^{0}_{1}\to\tau^{+}\tau^{-})<6\times 10^{-5}$ of CLEO and assume ${\cal B}(A^{0}_{1}\to\tau^{+}\tau^{-})=1$. Similarly, we also use the loosest upper limits on ${\cal B}(\Upsilon(3S)\to\gamma A^{0}_{1})\times{\cal B}(A^{0}_{1}\to invisible)<3.1\times 10^{-5}$ of BaBar [21] and assume ${\cal B}(A^{0}_{1}\to invisible)=1$. With the loosest upper limits, we get their bounds on the $\|X_{d}\|-m_{A^{0}_{1}}$ space, which are shown in Fig. 4. From the figure, we can see the bounds (excluding the parameter space $X_{d}>1$) for $0<m_{A^{0}_{1}}<7.8{\rm~{}GeV}$. Fig. 4(b) shows that there is no region of parameter space satisfying all the aforementioned constraints if the contribution of $A^{0}_{1}$ is required to solely resolve the $a_{\mu}$ discrepancy. Of particular interest, as shown in Fig. 4(a), is the parameter space around $m_{A_{1}^{0}}\sim 135{\rm~{}MeV}$ with $\|X_{d}\|<1$ (which is still allowed with approximation (i)). To make a thorough investigation of the space, we read off the upper limits of BaBar [21] from Fig. 5 for the value $m_{A_{1}^{0}}\sim 135{\rm~{}MeV}$: ${\cal B}(\Upsilon(3S)\to\gamma A_{1}^{0})\times{\cal B}(A_{1}^{0}\to invisible)\lesssim 3.3\times 10^{-6}$. With the assumption that ${\cal B}(A_{1}^{0}\to invisible)\simeq 1$ and the constraints from ${\cal B}(\pi^{0}{\to}e^{+}e^{-})$, we get $\|X_{d}\|=0.10\pm 0.08,~{}~{}~{}~{}m_{A_{1}^{0}}=134.99\pm 0.01\,{\rm MeV},$ (26) where the constraint on $m_{A_{1}^{0}}$ is dominated by ${\cal B}(\pi^{0}{\to}e^{+}e^{-})$ and the limit of $\|X_{d}\|$ is dominated by ${\cal B}(\Upsilon(3S)\to\gamma A_{1}^{0})$. At first sight, the uncertainties in the above mentioned two parameters are too different. We find that the difference arises from our assumption $\Gamma(A_{1}^{0})\simeq\Gamma(A_{1}^{0}\to e^{+}e^{-})+\Gamma(A_{1}^{0}\to\gamma\gamma)$. From Eqs. (20) and (21), one can see that the $X_{d}^{2}$ factor in ${\cal M}_{A_{1}^{0}}$ could be canceled out by the one in $\Gamma(A_{1}^{0})$ when $m_{A_{1}^{0}}$ approaches $m_{\pi^{0}}$, which results in a very sharp peak for position of $m_{A_{1}^{0}}$. Thus, with the well measured quantities given in Eq. (20) and the sensitivity of the peak, $m_{A_{1}^{0}}$ turns out to be well-constrained. Furthermore, if we take $m_{A_{1}^{0}}=m_{\pi^{0}}$, we find that $X_{d}^{2}$ is canceled out exactly, so there is no parameter to tune; however, we have ${\cal B}(\pi^{0}\to e^{+}e^{-})\gg 1$, which violates the unitary bound and is thus excluded. From the results of Eq.26, we obtain ${\delta}a_{\mu}({A_{1}^{0}})=(-9.2\pm 8.9)\times 10^{-12}$ with $\tan\beta=30$ as a benchmark, which is small enough to be smeared by the chargino/sneutrino and neutralino/smuon contributions. Moreover, we have $\Gamma(A_{1}^{0})=(5.7\,\pm\,5.5)\times 10^{-13}\,{\rm MeV},$ (27) which corresponds to $\tau(A_{1}^{0})\sim 1.2\times 10^{-9}~{}s$ ($c\tau\sim 36{\rm~{}cm}$). Figure 5: $\sin 2\theta$ versus the mass difference of the unmixed states with $\|X_{d}\|=0.05$ and $0.18$. The solid and the dashed lines denote the real and the imaginary parts of $\sin 2\theta$, respectively. For the case where $A_{1}^{0}$ decays mostly to invisible particles, we take the width of $A_{1}^{0}$ as a free parameter and get $\Gamma(A_{1}^{0})\leq 8.24\times\,10^{-6}{\rm GeV}$, $m_{A_{1}^{0}}=134.99\pm 0.02{\rm~{}MeV}$ and $\|X_{d}\|\leq 0.18$. In this case, $m_{A_{1}^{0}}$ can equal $m_{\pi^{0}}$, and it is found that $\Gamma(A_{1}^{0})\leq 3.3\times 10^{-6}{\rm~{}GeV}$ and $\|X_{d}\|\leq 0.18$. III. The resonant effects of $m_{A^{0}_{1}}\sim m_{\pi^{0}}$ So far we have included only the width effects of $A^{0}_{1}$ with the Breit- Wigner formula for the propagator of $A^{0}_{1}$. When the masses of $A_{1}^{0}$ and $\pi^{0}$ are very close, the mixing between the two states could modify the parton level $\pi^{0}-A^{0}_{1}$ coupling. In a manner analogous to Ref.[35], the mixing can be described by introducing off-diagonal elements in the $A_{1}^{0}-\pi^{0}$ mass matrix $\displaystyle{\cal M}^{2}=\left(\begin{array}[]{cc}m_{A_{1}^{0}}^{2}-im_{A_{1}^{0}}\Gamma_{A_{1}^{0}}&\delta{m^{2}}\\\ \delta{m^{2}}&m_{\pi^{0}}^{2}-im_{\pi^{0}}\Gamma_{\pi^{0}}\end{array}\right)$ (30) with $\delta{m^{2}}=\sqrt{G_{F}/4\sqrt{2}}f_{\pi^{0}}m_{\pi^{0}}^{2}X_{d}$. The complex mixing angle $\theta$ between the states is given by $\sin^{2}2\theta=\frac{(\delta m^{2})^{2}}{\frac{1}{4}(m_{A_{1}^{0}}^{2}-m_{\pi^{0}}^{2}-im_{A_{1}^{0}}\Gamma_{A_{1}^{0}}+im_{\pi^{0}}\Gamma_{\pi^{0}})^{2}+(\delta m^{2})^{2}}.$ (31) The mass eigenstates $A_{1}^{\prime 0}$ and $\pi^{\prime 0}$ are obtained as $\displaystyle A_{1}^{\prime 0}$ $\displaystyle=$ $\displaystyle\frac{1}{N}(A_{1}^{0}\cos\theta+\pi^{0}\sin\theta),$ (32) $\displaystyle\pi^{\prime 0}$ $\displaystyle=$ $\displaystyle\frac{1}{N}(-A_{1}^{0}\sin\theta+\pi^{0}\cos\theta),$ (33) where $N=\sqrt{\|\sin\theta\|^{2}+\|\cos\theta\|^{2}}$. Then, we can write the decay amplitude of the “physical” state $\pi^{\prime 0}$ as $\|{\cal M}(\pi^{\prime 0}\to e^{+}e^{-})\|^{2}\,=\,\frac{1}{N^{2}}\Big{(}\|\cos\theta\|^{2}\|{\cal M}(\pi^{0}\to e^{+}e^{-})\|^{2}+\|\sin\theta\|^{2}\|{\cal M}(A_{1}^{0}\to e^{+}e^{-})\|^{2}\Big{)}.$ (34) Obviously, we obtain the SM result when $\theta$ is small. With $\|X_{d}\|=0.05$ and $0.18$, Fig. 5 shows $\sin 2\theta$ as a function of the difference between $m_{A_{1}^{0}}$ and $m_{\pi^{0}}$. We note that the imaginary part of $\sin 2\theta$ is negligibly small, since $\Gamma_{A^{0}_{1}}m_{A^{0}_{1}}+\Gamma_{\pi^{0}}m_{\pi^{0}}\ll\delta m^{2}$. So, the normalization parameter $N$ of the mixing states is nearly unity. Combining the constraints from ${\cal B}(\Upsilon(3S)\to\gamma A_{1}^{0})$ and ${\cal B}(\pi^{\prime 0}\to e^{+}e^{-})$, we get $\|X_{d}\|=0.17\pm 0.01,~{}~{}~{}~{}m_{A_{1}^{0}}\simeq m_{\pi^{0}}\,.$ (35) This confirms the results of our straightforward calculation from Eq. (26), but gives a somewhat stronger constraint on $\|X_{d}\|$. With this constraint, we get $\Gamma(A_{1}^{0})=(9.8\,\pm\,1.1)\times 10^{-13}\,{\rm MeV},$ (36) which is also in agreement with Eq. (27). Furthermore, we get $\|\sin\theta\|^{2}=0.31\pm 0.19$. It is well known that the decay width of $\pi^{0}\to\gamma\gamma$ agrees perfectly with the SM prediction, so it is doubtful that that $\pi^{0}\to\gamma\gamma$ would be compatible with Higgs with a degenerate mass $m_{\pi^{0}}$. Using the fitted result $\|\sin\theta\|^{2}=0.31\pm 0.19$ and $\|{\cal M}(\pi^{\prime 0}\to\gamma\gamma)\|^{2}\,=\,\frac{1}{N^{2}}\Big{(}\|\cos\theta\|^{2}\|{\cal M}(\pi^{0}\to\gamma\gamma)\|^{2}+\|\sin\theta\|^{2}\|{\cal M}(A_{1}^{0}\to\gamma\gamma)\|^{2}\Big{)},$ (37) one can easily observe that $\|{\cal M}(A_{1}^{0}\to\gamma\gamma)\|^{2}\simeq\|{\cal M}(\pi^{0}\to\gamma\gamma)\|^{2}$ (38) is needed to give $\Gamma(\pi^{\prime}\to\gamma\gamma)\simeq\Gamma(\pi^{0}\to\gamma\gamma)$. However, it would require a too large value of $\|X_{d}\|\simeq 10^{3}$; therefore, the degenerate case is excluded. ## 4 Conclusion We have studied the decay $\pi^{0}\to e^{+}e^{-}$ in the NMSSM and shown that it is sensitive to the light CP-odd Higgs boson $A^{0}_{1}$ predicted in the model. The possible discrepancy between the KTeV Collaboration measurement [16] and the theoretical prediction of ${\cal B}(\pi^{0}\,{\to}\,e^{+}e^{-})$ could be resolved in NMSSM by the effects of $A^{0}_{1}$ at the tree level. However, it excludes a large fraction of the parameter space of $m_{A^{0}_{1}}-\|X_{d}\|$. To further constrain the parameter space, we have included bounds from muon $g-2$ and the recent searches for $A^{0}_{1}$ from radiative $\Upsilon$ decays performed at CLEO [20] and BaBar [21]. Combining all these constraints, we have found that * • ${\cal B}(\pi^{0}\,{\to}\,e^{+}e^{-})$ and ${\cal B}(\Upsilon\to\gamma A_{1}^{0})$ put strong constraints on the NMSSM parameter $X_{d}$ and $m_{A_{1}^{0}}$. Due to their different dependences on the two parameters, the interesting scenario where $m_{A^{0}_{1}}=214.3{\rm~{}MeV}$ is excluded, which would invalidate the $A_{1}^{0}$ hypothesis for the three HyperCP events [25]. * • Although these constraints point to a pseudoscalar with $m_{A_{1}^{0}}\sim m_{\pi^{0}}$ and $\|X_{d}\|=0.10\pm 0.08$ ($0.17\pm 0.01$, $\pi^{0}-A_{1}^{0}$ mixing included) in the NMSSM, such an $m_{A_{1}^{0}}$ is excluded by $\pi^{0}\to\gamma\gamma$ decay. In this Letter, we have worked in the limit of $X_{d}\gg X_{u}$, i.e., the large $\tan\beta$ limit. If we relax the limit and take Eq.5 as a general parameterization of the couplings between a pseudoscalar and fermions, the $\bar{u}-u-A_{1}^{0}$ coupling should be included. However, its contribution is deconstructive to the contributions from $X_{d}$, since the $\pi^{0}$ flavor structure is $(u{\bar{u}}-d{\bar{d}})$. To give a result in agreement with the KTeV Collaboration measurement [16], $X_{u}\gg X_{d}$ would be needed, which would imply possible large effects in $\Psi(1S)$ radiative decays. Detailed discussion of this issue would be beyond the main scope of our present study. In summary, we could not find a region of parameter space of NMSSM with $m_{A_{1}^{0}}<7.8{\rm GeV}$ in the large $\tan\beta$ limit that is consistent with the experimental constraints. The HyperCP $214.3{\rm MeV}$ resonance and the possible $3.3\sigma$ discrepancy in $\pi^{0}\to e^{+}e^{-}$ decay are still unsolved. Finally, further theoretical investigation is also needed to confirm the discrepancy between the KTeV measurements and SM predications of $\pi^{0}\to e^{+}e^{-}$ decay. If the discrepancy still persists, it would be an important testing ground for NP scenarios with a light pseudoscalar boson. The work is supported by the National Science Foundation under contract Nos.10675039 and 10735080. ## References * [1] S. Drell, Nuov. Cim. XI (1959) 693. * [2] L. Bergström, Zeit. Phys. C 14 (1982) 129. * [3] K.S. Babu and E. Ma, Phys. Lett. B119 (1982) 449. * [4] L. Bergström, E. Masso, L. Ametller and A. Bramon, Phys. Lett. B 126 (1983) 117. * [5] Ll. Ametller, A. Bramon and E. Massó Phys. Rev. D 48 (1993) 3388 [hep-ph/9302304]. * [6] M.J. Savage, M. Luke, and M.B. Wise, Phys. Lett. B291 (1992) 481. * [7] D. Gómez Dumm and A. Pich Phys. Rev. Lett 80 (1998) 4633 [hep-ph/9801298]. * [8] M. Knecht, S. Peris and E. de Rafael Phys. Rev. Lett 83 (1999) 5230 [hep-ph/9908283]. * [9] A. E. Dorokhov and M. A. Ivanov, Phys. Rev. D 75 (2007) 114007, arXiv:0704.3498 [hep-ph]. * [10] M. D. Scadron and M. Visinescu, Phys. Rev. D 29, (1984) 911; A. N. Kamal and L. C. Huah, Phys. Rev. D 32, (1985) 1744; G. Triantaphyllou, Mod. Phys. Lett. A 8, (1993) 1691 [hep-ph/0301214]. * [11] H. J. Behrend et al. (CELLO Collaboration), Zeit. Phys. C 49 (1991) 401. * [12] J. Gronberg et al. (CLEO Collaboration), Phys. Rev. D 57 (1998) 33 [hep-ex/9707031]. * [13] C. Amsler et al. (Particle Data Group), Phys. Lett. B 667 (2008) 1. * [14] J. Fischer et al., Phys. Lett. B 73 (1978) 364. * [15] R. Niclasen, Ph.D Thesis, FERMILAB-THESIS-2006-12, UMI-32-07725. http:// lss. fnal. gov/ archive/ thesis/ fermilab-thesis -2006-12. shtml. * [16] E. Abouzaid et al., KTeV Collaboration, Phys. Rev. D 75 (2007) 012004 [hep-ex/0610072]. * [17] A. E. Dorokhov and M. A. Ivanov, Eur. Phys. J. C 55 (2008) 193, arXiv:0801.2028 [hep-ph]. * [18] A. E. Dorokhov and M. A. Ivanov, JETP Lett. 87 (2008) 531 arXiv:0803.4493 [hep-ph]. * [19] H. P. Nilles, M. Srednicki and D. Wyler, Phys. Lett. B 120 (1983) 346; J. M. Frere, D. R. T. Jones and S. Raby, Nucl. Phys. B 222, (1983) 11; J. P. Derendinger and C. A. Savoy, Nucl. Phys. B 237, (1984) 307; J. R. Ellis, J. F. Gunion, H. E. Haber, L. Roszkowski and F. Zwirner, Phys. Rev. D 39 (1989) 844; M. Drees, Int. J. Mod. Phys. A 4 (1989) 3635. * [20] W. Love, et al (CLEO Collaboration), arXiv:0807.2695 [hep-ex]. * [21] B. Aubert, et al (BaBar Collaboration), arXiv:0808.0017 [hep-ex]. * [22] G. Hiller, Phys. Rev. D 70 (2004) 034018 [hep-ph/0404220]. * [23] B. R. Martin, E. De Rafael and J. Smith, Phys. Rev. D 2, (1970) 179. * [24] M. Pratap and J. Smith, Phys. Rev. D 5, (1972) 2020; Z. K. Silagadze, Phys. Rev. D 74, (2006) 054003 [hep-ph/0606284]. * [25] H. K. Park, et al. (HyperCP Collaboration), Phy. Rev. Lett. 94 (2005) 021801 [hep-ex/0501014]. * [26] X. G. He and J. Tandean, Phy. Rev. Lett. 98 (2007) 081802 [hep-ph/0610362]. * [27] X. G. He, J. Tandean and G. Valencia, JHEP 0806 (2008) 002, arXiv:0803.4330 [hep-ph]. * [28] G. Valencia, arXiv:0805.3285 [hep-ph]. * [29] X. G. He, J. Tandean and G. Valencia, Phys. Lett. B 631 (2005) 100 [hep-ph/0509041]. * [30] N. G. Deshpande, G. Eilam and J. Jiang, Phys. Lett. B 632 (2006) 212 [hep-ph/0509081]. * [31] J. F. Gunion, G. Gamberini and S. F. Novaes, Phys. Rev. D 38, (1988) 3481. * [32] G. W. Bennett et al. (Muon ($g-2$) Collaboration), Phys. Rev. D 73 (2006) 072003; * [33] M. Passera, W. J. Marciano and A. Sirlin, arXiv:0804.1142. * [34] F. Domingo and U. Ellwanger, JHEP 0807 (2008) 079 arXiv:0806.0733 [hep-ph]. * [35] P. J. Franzini and F J. Gilman, Phys. Rev. D 32, (1985) 237; M. Drees and K. i. Hikasa, Phys. Rev. D 41, (1990) 1547; E. Fullana, M. A. Sanchis-Lozano, Phys. Lett. B 653 (2007) 67 [hep-ph/0702190].	arxiv-papers	2008-08-21T14:03:23	2024-09-04T02:48:57.379825	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Qin Chang, Ya-Dong Yang", "submitter": "Yadong Yang", "url": "https://arxiv.org/abs/0808.2933" }
0808.2978	# A Coherence-Based Approach for Tracking Waves in the Solar Corona Scott W. McIntosh1,3, Bart De Pontieu2 & Steven Tomczyk1 1: High Altitude Observatory, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307, USA. 2: Lockheed Martin Solar and Astrophysics Lab, 3251 Hanover St., Org. ADBS, Bldg. 252, Palo Alto, CA 94304, USA. 3: mailto:mscott@ucar.edu ###### Abstract We consider the problem of automatically (and robustly) isolating and extracting information about waves and oscillations observed in EUV image sequences of the solar corona with a view to near real-time application to data from the Atmospheric Imaging Array (AIA) on the Solar Dynamics Observatory (SDO). We find that a simple coherence / travel-time based approach detects and provides a wealth of information on transverse and longitudinal wave phenomena in the test sequences provided by the Transition Region and Coronal Explorer (TRACE). The results of the search are pruned (based on diagnostic errors) to minimize false-detections such that the remainder provides robust measurements of waves in the solar corona, with the calculated propagation speed allowing automated distinction between various wave modes. In this paper we discuss the technique, present results on the TRACE test sequences, and describe how our method can be used to automatically process the enormous flow of data ($\approx$1Tb/day) that will be provided by SDO/AIA after launch in late 2008. ††slugcomment: In Press Solar Physics Since the launch of the Transition Region and Coronal Explorer (TRACE; Handy et al. 1999) the community has invested a great deal of effort on the identification and analysis of longitudinal and transverse wave phenomena in loop structures seen in its EUV images of the corona (see Nakariakov and Verwichte 2005, for a good overview). The great interest in finding and characterizing coronal waves is driven by the promise of coronal seismology which could lead to the determination of otherwise inaccessible physical properties of the solar atmosphere by studying phase speeds, amplitudes, dissipation, etc. of the observed waves (Nakariakov and Verwichte 2005). There are many unresolved issues in coronal seismology: for example, it is unclear why only a subset of coronal loops show transverse oscillations in the wake of a flare or CME (e.g., Aschwanden et al. 2002), which physical processes dominate the damping of these transverse oscillations (e.g., Ofman and Aschwanden 2002; Arregui et al. 2007; Terradas, Andries, and Goossens 2007), why only a subset of coronal loops show propagating slow-mode oscillations (e.g., de Moortel, Ireland, and Walsh 2000; de Moortel and Rosner 2007), how the propagation of slow-mode waves at different temperatures can be used to probe the thermal structure and loop scale heights (King et al. 2003), or how lower-atmospheric oscillations generally leak into the corona (e.g., De Pontieu, Erdélyi, and De Moortel 2005; Jefferies et al. 2006; Khomenko et al. 2008)? Many of these issues require more extensive study with much larger statistical samples. It is surprising that the number of wave and oscillation detections in the above papers is only of order several dozen for both longitudinal and transverse oscillations. This is a very small number compared to the volume of TRACE data given that flares occur often and “wave-leakage” from the solar interior to the lower portions of the outer solar atmosphere is apparently quite abundant (De Pontieu, Erdélyi, and De Moortel 2005; Jefferies et al. 2006; Khomenko et al. 2008). The discrepancy between the plethora of observed wave driven phenomena in the chromosphere and transition region, and the relative paucity of waves seen in the corona partly motivates the effort discussed herein. Is the small number of coronal wave observations caused by the fact that all of the TRACE observations of waves have been found as a result of a “manual” search, i.e., limited by human patience and detection skills? Or are there actually very few locations in the corona where waves can be observed? One of our aims is to develop an automated wave-detection algorithm to help expand the number of known oscillations, and build up significant statistics on how frequent coronal oscillations occur. Such larger statistical samples are crucial to address many of the unresolved issues in coronal seismology, and will be within reach with the Atmospheric Imaging Assembly (AIA) after launch of the Solar Dynamics Observatory (SDO) by 2009. This instrument will provide full-disk coronal imaging at significantly increased signal-to-noise in eight coronal wavelengths at ten-second cadence continuously, producing an enormous data flow of $\approx$1Tb day-1. The potential for significantly increased detection of coronal oscillations and waves with such a data rate and quality are very high. However, AIA’s enormous data rate renders business- as-usual approaches that involve manually looking through data sets and individually flagging events unfeasible. Since it is unpractical to transfer the full volume of AIA data to remote users, it is critical that automated algorithms be developed that can search for wave-like phenomena and automatically flag temporal and spatial regions of interest for later, more detailed analysis. There have been a few papers recently that have started addressing some of the issues that need to be resolved to enable full optimal use of the AIA data, such as automated detection of locations with significant oscillatory power (Nakariakov and King 2007), or semi-automated detection of propagating and standing waves (Sych and Nakariakov 2008). The approach presented in this paper may be the first that can reliably and automatically distinguish between longitudinal and transverse oscillations at the same time as rejecting most false positive detections of oscillations. Our approach is based on the technique used to analyze and detect coronal Alfvén waves with the CoMP instrument (Tomczyk et al. 2007). In the following section we discuss the method developed, illustrating its application on example TRACE datasets (see Sect. 2) which are known to show a host of transverse and longitudinal wave phenomena and are therefore good for testing our approach. In Sect. 3 we explore the results of the analysis on the test data and discuss the method employed to minimize false detections. In Sect. 4 we compare some of our results with previous analyses of the TRACE datasets we have used, and discuss how our method could be used in a practical manner for SDO/AIA data. ## 1 Method We adapt the phase travel-time analysis (e.g., Jefferies et al. 1994, 1997; Finsterle et al. 2004; McIntosh, Fleck, and Tarbell 2004) to isolate the propagation characteristics of the wave modes observed in the TRACE image sequences time-series as a function of frequency. This technique is performed as a Fourier comparison of time-series in neighboring image pixels to extract information about the cross-correlation, coherence, cross-power and phase dependence of the neighboring signals. As demonstrated in Tomczyk et al. (2007) this is a particularly simple, but powerful technique to track coronal waves, albeit in the plane-of-the-sky. The technique has the inherent flexibility to be employed rapidly in the spectral domain producing robust results. The first step in the travel-time analysis involves computing the multi dimensional Fourier transform of the EUV image cube ($I(x,y,t)$) in the third dimension which gives a complex cube $I^{\ast}_{f}(x,y,f)$ that now has frequency $f$ [units: mHz] as its third dimension. Before calculating the Fourier transform we apply a Hanning window on the timeseries to reduce the impact of temporal trends in the short data sequences. The analysis involves comparisons of the intensity timeseries of our pixel of interest $(x,y)$ and that of a pixel within a square region (of dimensions $dx\times dy$) around the central pixel. At this point the analysis diverges into the spectral (Sect. 1.1) and temporal domains (Sect. 1.2) where the former has been developed by Tomczyk, and McIntosh (2008) to supersede the analysis of Tomczyk et al. (2007). We have chosen the range to be $\pm$10 pixels in both spatial dimensions. ### 1.1 Spectral Domain Analysis For the range of pixels $[x\pm dx:y\pm dy]$ we compute the cross-spectrum $CS_{RB}(f)=I^{\ast}_{R}(f)I^{\ast}_{B}(f)^{\dagger}$ (1) between the Fourier transformed timeseries at the reference pixel $I^{\ast}_{R}(f)$ for any other pixel in the analysis box $I^{\ast}_{B}(f)$, with $I(f)^{\dagger}$ being the complex conjugate of $I(f)$. We then smooth the cross-spectrum with a five-point boxcar smoothing function in frequency to reduce the contribution of noise in the computation. In the spectral domain we wish to rapidly compute wave diagnostics in several characteristic frequency filters simultaneously and for the sake of simplicity we will use Gaussian filters of the form $G(f_{0},\delta f)$ where $f_{0}$ is the central frequency and $\delta f$ is the width of the Gaussian filter chosen which, for this paper, is $f_{0}$/10. So, for each filter selected we can simultaneously compute the weighted signal cross-power [$WCP_{RB}(f_{0})$], phase [$\phi_{RB}(f_{0})$] , coherence [$C_{RB}(f_{0})$] and phase travel-time [$T_{RB}(f_{0})$] for the pixel of interest that are given by the following expressions: $WCP_{RB}(f_{0})=G(f_{0},\delta f)\\|CS_{RB}(f)\\|$ (2) $\phi_{RB}(f_{0})=G(f_{0},\delta f)\tan^{-1}\left(\frac{Im\\{C_{RB}(f)\\}}{Re\\{C_{RB}(f)\\}}\right)$ (3) $C_{RB}(f_{0})=G(f_{0},\delta f)\left(\frac{CS_{RB}(f)^{2}}{CS_{RR}(f)CS_{BB}(f)}\right)$ (4) $T_{RB}(f_{0})=\frac{\phi_{RB}(f_{0})}{2\pi f_{0}}$ (5) Repeating this step for all of the neighboring pixels results in the construction of 10$\times$10 pixel maps of the three quantities for each filter frequency, $f_{0}$. The panels in Figure 1 show coherence, cross-power, phase and phase travel- time maps from the spectral analysis of one pixel (coordinates 181, 97) in the TRACE 171Å dataset from 1 July 1998 12:03, - 1 July 1998 13:02 UTC (see below) at a filter frequency of $f_{0}$ = 5.5mHz. In particular, we see that the coherence map (panel A), like that shown in the approach of Tomczyk et al. (2007), has a region, or “island”, where the spectral coherence is large ($>$0.6) at this frequency across numerous neighboring pixels. It is the properties of the strong coherence island and its mappings in phase-time and cross-power that form the cornerstone of the analysis presented. They can provide a quick and reliable estimate of the propagation direction and phase- speed of the disturbance present, from which reasonable errors can be determined. After isolating the relevant high coherence pixels in the phase-time diagnostic map, we compute a distance-minimizing linear fit111In this case we seek the best fitting straight line through the cluster of points in the island that also minimizes the distance between the points and that straight line (see, e.g. Bevington and Robinson 2003). We have found that this approach is not subject to singularities for clusters of points that are orthogonal, or nearly orthogonal, to the horizontal as is the case in standard least-squares fits when the dependent and independent variables are interchangeable or both have measurement errors. to the cluster of points in the island and calculate a distance d along the resulting straight line as well as an angle (and the error on the angle) at which the disturbance is moving (relative to North - South in the coherence map). We also use the cluster of pixels to extract a measure of two coherent length-scales, one that is parallel to the measured propagating direction (the coherence length), and one that is perpendicular to it (the coherence width). These last diagnostics may contain physical information about the process exciting the disturbance and so we retain them at little computational cost. We then compute the phase-speed of the disturbance and its associated error by forming a scatter plot of distance $d$ (from the reference pixel) versus measured phase-time (see, e.g., Figure 2). The gradient of the least-squares linear fit to these data provides an estimate of the phase speed of the disturbance, errors in the gradient are determined by assessing a range of possible values that can achieve a similar quality of fit to the data (see, e.g. Sect. 6 of Bevington and Robinson 2003). We repeat the previous step for each filter required unless there are fewer than 6 pixels in the island. This process is then repeated for all of the other pixels in the original image with the result that we develop images of the apparent phase-speed, angle, (their respective errors) and additional measures of the island-mean coherence and cross-power (see, e.g, Figure 5 and Sect. 3). ### 1.2 Temporal Domain Analysis At the cost of a considerable increase in computation time, we can substitute the phase-time and phase-speed measurements by performing their calculation in the temporal domain. This step involves the cross-correlation of the Fourier filtered timeseries for all of the pixels in the high coherence island relative to the reference pixel. The cross-correlation function at each neighboring pixel is a Gabor wavelet that, when fitted, yields information about the phase (and energy carrying group) travel-times of the disturbance (Finsterle et al. 2004). In practice we cannot execute the temporal domain analysis for SDO/AIA as the computational cost is prohibitive for a method that is essentially mathematically equivalent to the one presented above. However, we do note that the process is illustrative of the general technique, the quantities and measures developed for the analysis and so we have chosen to provide an example. An example of two-step 222Typically we would fit both phase and group information simultaneously, as is done in Finsterle et al. (2004) and the other references provided above, but for the data at hand, where often the group envelope is not well defined, we fit them separately. cross-correlation fitting is shown in Figure 3. First, we fit the envelope of the function to extract group information, after which we fit the peak of the cross- correlation nearest to the lag origin to obtain phase information. The solid black line shows the cross-correlation of the reference timeseries used to create Figure 1 and that from another pixel (coordinate [181, 91] in the figure) as a function of lag in seconds. The group behavior is determined by fitting a Gaussian (dot-dashed line) to the envelope of the cross-correlation function that is outlined by the absolute values of the function maxima (red triangles) - the center of the fitted Gaussian measures the group travel-time (or energy transport time). The phase travel-time can then be extracted by fitting the peak of the cross-correlation function nearest to the origin. Using either a parabolic of Gaussian fit we can extract phase-times (the center of the Gaussian or parabolic fit) down to an accuracy of about one tenth of the image cadence. For the case considered the vertical dashed lines in panel C of Figure 3 show the positions of the estimated group travel-time (red; -4.17 minutes) and phase travel-time (green; -0.527 minutes. For good measure, we note that the latter value is consistent with the values in Figures 1 and demonstrates that the disturbance is propagating left to right which is also consistent with the movie of Figure 4) respectively. Repeating the steps used in the spectral analysis to compute the phase-speed we perform a least-squares linear-fit to the scatter plot of pixel distance and measured phase-travel times. We note that it is also possible to investigate the group-speed of the disturbance - by simply replacing the phase travel-times with the measured group travel-times and repeating the fit to the scatter. ### 1.3 Improving Algorithm Efficiency The algorithm described in Sects. 1.1 and 1.2 is quite slow, since it calculates cross-correlations or cross-spectral properties for 100 neighboring points at each location of the datacube. We have improved the efficiency of the algorithm by a large factor (typically of order 20) in the following way. For each location, we define a map of locations for which cross-correlations with the central pixel [$x_{0},y_{0}$] need to be calculated. At the first iteration this map contains only the 4 immediate neighbors of the central pixel. We then determine which of these locations has a coherence larger than the cutoff (0.6 in our case). If none of these locations has a coherence larger then the cutoff, the algorithm ends the calculations for pixel [$x_{0},y_{0}$], and moves on to [$x_{0}+1,y_{0}$]. If one or more locations shows a coherence larger than the cutoff, we repeat the iteration by calculating a new map of points for which cross-correlations need to be calculated. This map will now contain the immediate neighbors of the high coherence locations calculated in the previous round. We exclude from this map all locations that have already been calculated. After this the algorithm again checks whether any of the new points has a coherence larger than the cutoff, continuing until no new locations are above the cutoff. In this fashion we can usually reduce the number of calculations from 100 to around 5, since many locations show little coherence with their neighbors, and only a select few locations show significant coherence with their neighbors. This increase in efficiency has a critical impact on the usefulness of the algorithm for searching through large datasets, such as those that are expected from SDO/AIA. ## 2 Test Data To illustrate the algorithm we have applied it on two different TRACE datasets. Dataset I was taken on 1 July 1998 from 12:03 to 13:02 UTC in 171Å, with a cadence of 31.5s. Dataset II was taken on 14 July 1998 from 12:45 to 13:42 UTC in 171Å during the so-called Bastille day flare and has a cadence of 72.3 seconds. The first dataset contains an active region with some slow-mode running waves in one of the loops associated with a sunspot. The second dataset has been the subject of several papers (see, e.g. Aschwanden et al. 1999, 2002) and contains the “classical example” of transverse oscillations. Before applying the algorithm, we performed the calibration steps included in trace_prep.pro (as part of the TRACE tree of solarsoft IDL), which includes the subtraction of dark current/pedestal, correction for the ccd gain and lumogen degradation, correction of bad pixels and normalization of all images to remove differences in exposure time. In addition, we performed several iterations of cosmic ray removal using the trace_unspike.pro routine which removes excessively bright spikes by comparing the data with previous and following timesteps and replacing the spike with a temporal and spatial average of the surrounding pixels. As a final step we derotate the data by using poly_2d.pro to perform sub-pixels shifts that are calculated using diff_rot.pro. We also perform a 2$\times$2 rebinning to reduce the TRACE data to a spatial resolution of 1 arcsecond. We have tested the impact of the rebinning extensively and found that rebinning to the spatial resolution does not lead to any loss of useful information about the oscillations, and in fact improves the signal to noise of the sometimes weak oscillatory signals. Of course, rebinning the data naturally speeds up the algorithm by a factor of 4 compared to running it on the data at its original resolution. Figure 4 show snapshots of the raw and Fourier filtered timeseries of the 1 July 1998 (left) and July 14 1998 (right) datasets respectively. In each case, the upper left panel shows a snapshot of the intensity timeseries while other panels show the filtered timeseries with a Gaussian filter centered around 1.5 mHz, 3.5 mHz and 5.5 mHz respectively. In the left panel we see that the active region contains several sets of loops, with the right-most fan associated with a sunspot being highlighted in a box. The loops in the southern part of the active region are associated with plage regions. We see that the sunspot fan contains a clear 180s signal which is visible as a dark spot in the 5.5 mHz snapshot (lower right) and an oscillatory signal in the same panel of the provided movie. These kinds of oscillations have been described extensively in a series of papers by de Moortel et al. (2002a, b). In addition, the moss regions towards the north and middle of the active region show a predominance of 1.5 and 3.5 mHz power. This is not surprising and has been noted before by De Pontieu, Tarbell, and Erdélyi (2003) who showed that the periodic obscuration of upper transition region emission by chromospheric jets plays a dominant role in the EUV (semi-)periodicities. A very intriguing finding is that there is significant power in the 1.5 mHz filtered movies and snapshot in the plage-related loops towards the south of the active region. We will return to these low frequency oscillations in Sects. 3 and 4. As a cautionary note we see that generally, these filtered movies show not only periodic signals, but any signal that shows power at the chosen frequency. Because of the nature of the Fourier transform, this includes signals that are caused by abrupt changes in the brightness, such as cosmic rays or sudden loop brightenings (e.g., in the case of emerging flux reconnecting with overlying field, or in post-flare loop arcades). To exclude some of those signals when we run our algorithm on the data, we take care to use a Hanning window (reducing the influence of trends), and apply a mask that excludes from our calculations all pixel locations whose raw time series show deviations that are more than 3.5 times the standard deviation of the time series. This practically excludes many of the locations with cosmic rays and sudden brightenings. This approach is especially useful in the case of a flare where many locations show significant and sudden brightenings, such as is the case in dataset II, and shown in the right panel. The filtered movies and snapshots show many locations with significant oscillatory power, which is borne out by the unfiltered time series. The flare is associated with a coronal dimming, which leads to significant “wiggling” of the coronal loop structures, especially on the 5-10 minute time scales. Very prominent in these snapshots are also the post-flare loops in the middle of the active region. Most interesting is that the transversely oscillating loops show up very clearly in the movies and snapshots, especially at 3.5 mHz (lower left) and to a lesser degree at 5.5 mHz (lower right). ## 3 Results ### 3.1 Interpretation of Raw Results We present results of the spectral algorithm for both datasets. As we have stated above the results for the temporal code are very similar but run at a much lower speed so that, as such, it is not adequate for our ultimate goal of applying this technique to SDO/AIA. In principle our algorithm produces for each location, and for each frequency we choose to filter on, the following average properties of the island of high coherence around each location: cross-spectral power, average coherence, phase speed of propagating signal, error on the phase speed, the angle of propagation of the disturbance, the correlation length and the correlation width. To reduce the amount of noise in the resulting maps of the active region, we reject locations for which the island of high coherence consists of fewer than 6 neighbors. The resulting maps of the properties described above show only locations where the signal is significantly correlated with at least 5 neighbors (Figures 5, 6 and 7 for the frequencies 1.5, 3.5 and 5.5 mHz for dataset I, and Figures 8, 9 and 10 for dataset II). What is striking in these maps is the large amount of locations that have (oscillatory) correlation with their neighbors. However, since our algorithm is so sensitive, many of these locations actually show “false positives” in which a very small (mostly instrumental or spurious) oscillatory signal is correlated with a few neighboring pixels. Nevertheless, some interesting physical results immediately show up that confirm the findings of our cursory analysis of the Fourier-filtered time series in Section 3. For example, the 3.5 mHz map of phase speeds of dataset I (Figure 6) is dominated by locations where moss occurs, as can be expected from previous analyses (e.g., De Pontieu, Tarbell, and Erdélyi 2003). The phase speeds are quite variable and of order 10-50 km s-1 in the moss regions, which is perhaps not surprising since the correlations with neighboring pixels are most likely caused by chromospheric jets (with similar velocities to those we find; (e.g., De Pontieu, Tarbell, and Erdélyi 2003; Hansteen et al. 2006) moving across TRACE pixels. The 3.5 mHz map also shows evidence of oscillatory power in the bottom of some coronal loops (e.g., around 110, 90) with phase speeds of order 50 - 100 km s-1. The 5.5 mHz map of dataset I (Figure 7) is much more sparsely populated with only a few major concentrations of significant oscillatory power towards the lower right of the map. These locations are exactly where we find (through visual examination) oscillations in the original and Fourier-filtered time series: our algorithm nicely locates the sunspot oscillations at 180 second periods (around 180, 90). The phase speeds here are of order 50-150 km s-1, which is within the range that is expected for slow-mode propagating disturbances (de Moortel et al. 2002a, b). The 1.5 mHz results (Figure 5) show large parts of the southern coronal loops with significant oscillatory disturbances that propagate along the loops with phase speeds that are similar and of order 50 - 150 km s-1. If these oscillations are real (see Sect. 5), they are interesting, since they are of much longer periods than have been previously reported (typically five minutes in plage-related loops). We have verified that the signals are not introduced by slow pointing drifts on timescales of ten minutes by performing the same analysis using data that is co-aligned with various co-alignment strategies (see Sect. 5). In addition, co-alignment issues would lead to similar signals in similarly oriented loops, e.g., the loops at the northern edge of the sunspot fan. Our algorithm does not find any sign of significant propagating and oscillatory disturbances at 1.5 mHz in those loops. The results of dataset II are similarly interesting. First of all, the phase speeds shown in all frequencies are systematically higher in this dataset, with most being of order 150 - 500 km s-1. This is compatible with the fact that this dataset is dominated by a flare, coronal dimming and the associated (mostly fast magneto-acoustic mode) waves and/or standing waves (transverse oscillations) that have been well documented in previous work focusing on the Bastille day flare (e.g., Aschwanden et al. 1999, 2002). The 3.5 mHz maps (Figure 9) of phase speed are dominated by locations of significant oscillatory power at the same locations where previous analysis (based on visual inspection) had found evidence of transversely oscillating loops (e.g., around 170, 200 or 180, 70 or 240, 110). In addition, our algorithm finds a range of locations with oscillatory power that have not been discussed previously, such as the western side of the active region (the general area from x $>$ 200 and y $>$ 200). The phase speeds in the 3.5 mHz maps are of the order of several hundred km s-1, which is to be expected for standing waves or propagating fast-mode waves in the corona. The 5.5 mHz maps (Figure 10) are again much more sparsely populated, with only a small subset of locations showing significant oscillatory power. This is perhaps not surprising since previous analyses have shown that this particular flare led to transverse oscillations with periods of order five minutes. However, the fact that we do find power at 5.5 mHz ($\approx$three minute periods), and that it occurs at mostly the same locations as in the 3.5 mHz maps may suggest that the transverse oscillations are not dominated by a single peak in the Fourier spectrum, but rather have a range of periods. The 1.5 mHz maps (Figure 8) are very intriguing with most of the dimming regions showing significant oscillatory power that propagates at phase speeds of order 100-500 km s-1. However, in this case it is more difficult to clearly distinguish whether the resulting phase speeds and locations are uniquely the sites of an oscillatory disturbance propagating, or whether the sudden coronal dimming triggered by the flare is partially responsible for the results. However, visual inspection of the Fourier-filtered time series suggests that the coronal dimming is indeed accompanied by transient and quasi-periodic “wiggling” of at least some of the coronal structures in these regions. The observed range of phase speeds may then well be the propagation speed of the fast mode waves that are associated with such a drastic reorganization of the corona. In the above we have focused on the locations of significant oscillatory power and the associated phase speeds of the observed oscillatory disturbances. Our algorithm also naturally provides the direction in which the propagation occurs, as well as the coherence length and width of the islands of high coherence. The observed propagation direction usually is aligned with the dominant intensity structure (coronal loop or fan) that carries the disturbance. The coherence width and length are measures of how large the coherence island is, with the width by definition shorter than the length, and the difference between both larger as the island of high coherence is more oblong in shape. Typical coherence lengths are of order three - six Mm in dataset I, with longer coherence lengths for lower frequencies. For dataset II coherence lengths are of similar size, except for the low frequency results which have coherence lengths that are up to 10 - 35 Mm in the regions where the coronal dimming dominates. This is not surprising, since the dimming impacts a large area of the active region. ### 3.2 Reducing False Positives As we mentioned in the above, our algorithm is so sensitive that it finds a significant number of false positives - locations where instrumental or cosmic ray artifacts lead to spurious coherent signals. In addition, many of the locations have values for the phase speed that are based on poor fits between the distance along the propagation direction and the travel time, which can lead to significant errors on the parameters that are calculated. While the value of the maps we calculate is clear, the required human interpretation of what constitutes a false positive would render an automated classification (which is crucial for SDO/AIA) difficult. Fortunately we have found a method (which we describe in the following) to significantly reduce the number of false positives. We take several steps to prune the results down to locations where the observed signal is most likely caused by a real coronal signal, as opposed to an instrumental or detection artifact. We start with a map that contains locations that have enough neighbors (as defined in Sect. 1) with high coherence. As a first step to winnow down the number of false positives, we use the error on the determination of the phase speed (from the least-squares fit of distance versus travel-times). We reject all locations for which the relative error on the phase speed is larger than 0.3 (i.e., the error divided by the phase speed). This removes from our map a large number of locations where the travel times are not well correlated with distance along the island of high coherence. Such a poor correlation leads to a poor linear fit and a large error on the phase speed. To reduce the large number of locations caused by instrumental artifacts we then reject all locations in which the average brightness is below a certain threshold (1 DN s-1). This removes a large number of artifacts that are caused by read-out noise (which can dominate coherent signals in the dark areas of the field of view), as well as artifacts from JPEG compression around cosmic rays. While the despiker usually removes the cosmic rays, the compression artifacts around them can sometimes remain in the data. Most of these artifacts are removed as a result of the brightness thresholding. As a next step we use the map of locations and label each contiguous region in the map using the IDL routine label_region.pro. We then reject all regions that contain fewer than, say, eigth pixels (superpixels of 1 $\times$ 1 arcsecond2 in our case). This removes most of the moss regions and many of the instrumental artifacts that usually show coherence over only a few pixels. Note that this step bases the rejection on the number of contiguous pixels in our map that have a significant number of neighbors that are highly coherent at the frequency studied. This a different from the step in Sect. 1 where we rejected a location from the map if it had fewer than six neighboring pixels in the island of high coherence. While these two measures are related, they are not identical. The end result of these three steps is usually a cleaner map of locations with significant oscillatory power that are coronal in nature (as opposed to instrumental). We now have a map of regions (as opposed to pixels), in which adjacent pixels have been clustered together into one region. We then calculate the average phase speed, the average error on the phase speed, the average coherence length and width, and the average angle for each of these regions. The resulting figures show maps of oscillatory “events” with their averaged properties (see Figure 11 and 14 for dataset 1 and 2 respectively). The pruning method reduces the number of “pixels” from 850, 409, and 129 respectively for 1.5, 3.5, and 5.5 mHz and combines them to 46, 14, and 6 regions in dataset I. Dataset II contains many more pixels with significant oscillatory power, but the pruning method reduces the number of pixels from 950, 1103, and 1150 respectively down to 91, 94, and 55 events or regions. Such a significant reduction of information in an automated fashion renders the combination of our wavetracking and pruning algorithms into an interesting option to automatically mine the SDO/AIA data. The rejection of false positives is usually good enough so that the standard deviation of properties in each region is quite small (and is not particularly useful as a further method of rejecting artifacts). The rejection of false positives is generally a complicated task that is quite dependent on the quality of the data (exposure times, number of cosmic ray hits, efficiency of cosmic ray removal, etc.). As a result it usually requires some fine tuning of parameters. Fortunately the quality of the SDO/AIA data is expected to be higher and of a more constant quality than that of the TRACE data used here and so we will leave much of the “fine tuning” of the pruning thresholds, etc. until we have AIA data. This should greatly facilitate full and reliable automation of our data processing pipeline on the SDO/AIA data. ## 4 Discussion It is clear from our results that our method reliably finds locations of significant oscillatory power in the corona. The two-step approach involves wavetracking and pruning out false-positives that can reliably and automatically detect events that are worthy of further detailed analysis. These events can be characterized by the dominant frequency, the average phase speed of the observed periodic disturbance, the average angle of propagation and the average coherence width and length (on a pixel by pixel basis), as well as the size of the cluster of coherent pixels. We have demonstrated that the phase speed of the propagating disturbance can be reliably used to distinguish the various wave modes that are expected to appear in the corona, with low values ($<$150 km s-1) of the phase speed associated with propagating slow-mode waves, and higher values ($>$150 km s-1) associated with either propagating fast-mode waves or standing transverse oscillations. We note however that an accurate determination of the phase speed with the available data (and its relatively slow, and non-fixed cadence) is often difficult, so that the determined phase speeds are only indicative of the real value of the phase speed. Fixed high cadence, high S/N data such as that expected from SDO/AIA will alleviate much of this problem. In addition, we have experimented with the threshold of minimum coherence that defines the island of high coherence and find that higher thresholds generally of course reduce the size of the coherent islands, but can also lead to better determination of the phase speed. An additional problem that impacts the determination of the phase speed is the fact that we basically assume the island of coherence to be a roughly linear feature, such as a coronal loop. However, in certain cases, especially in fans associated with sunspots, or the large coronal dimming regions such an assumption may not be the most representative of how the propagation actually occurs. In such cases the source of the wave is relatively compact and the waves actually propagate outward in a more spherical manner, and not in a linear fashion, i.e., along a straight line which is what our method assumes. That means that our definition of distance travelled (which we define as the distance along the long axis of the island of high coherence) does not represent a good measure of distance along the propagation direction. This can clearly lead to a less accurate determination of the phase speed, and may be why some of the measured phase speeds of the slow-mode waves in the sunspot fan are on the low side (50 km s-1). This problem also directly affects the determination of the other parameters, especially the propagation angle and to a lesser extent the coherence width and length. Despite these issues, the phase speeds are different enough to distinguish propagating slow-mode waves from transverse oscillations. Our algorithm routinely finds phase speeds of order several hundred km s-1 for the latter, which is to be expected for partially standing waves. We note that previous analyses of these transverse oscillations have indicated that there is often a significant propagating component as well as a standing component to the oscillations. This may be one of the reasons why the phase speeds we obtain are not infinite. Other reasons, of course, are related to the inherent uncertainties in determining travel times and travelled distances, as described above. In its current form, our algorithm shows great potential for automated processing of SDO/AIA data. It is crucial for the SDO/AIA data that any kind of automated flagging of locations with significant oscillatory power contains so few false positives that the resulting list of events does not overwhelm the so-called “Heliophysics Knowledge Base” (http://www.lmsal.com/helio- informatics/hpkb/), and at the same time significantly reduces the amount of information the end-user needs to sift through before finding the subset of data that contains the oscillations of interest. As we have shown in this work, our algorithm does exactly that. It provides of order a few to a few dozen events per hour of EUV data and allows the users to immediately home in on the regions of interest that deserve more detailed study. Our algorithm takes one – two minutes (depending on the amount of oscillations present – we note that this does not include the time taken to derotate and coalign the image cubes) for a 256$\times$256$\times$110 datacube on a Mac Pro computer, without any effort to optimizing the FFT and correlation calculations for higher speeds. AIA data will be 2048$\times$2048 at this spatial resolution (i.e., two $\times$ two rebinning), so that a single processor would require of order one - two hours to process one hours worth of data in a single passband. This means that for all coronal AIA data, a pipeline containing of order twenty CPUs would be enough to keep pace with the data flow. The removal of false positives will require some finetuning of parameters and pruning approaches (e.g., using the cross-spectral power instead of the intensity to reject false positives) that will be investigated in the lead-up to the launch of SDO, and since much will depend on the quality of the data, during the commissioning phase. ## 5 Summary Our algorithm was applied to two datasets that contain a diverse range of oscillation or wave-like features. We were able to identify moss oscillations mostly at 1.5-3.5 mHz, propagating slow-mode waves at 3.5-5.5 mHz, partially standing transverse oscillations at 1.5-5.5 mHz, as well as a large number of previously unreported apparently propagating waves at 1.5 mHz in coronal loops associated with plage (dataset I). Our algorithm has the potential to perform automated studies of cross-wavelength correlations of oscillatory power in various EUV wavelengths (171 and 195Å for example, see, e.g., King et al. 2003; Robbrecht et al. 2001). This will be the subject of a future paper. Our algorithm can also be run on a large number of datasets to help determine how common propagating slow-mode oscillations are in coronal loops associated with sunspots and plage. Especially the latter have been the subject of speculation since they involve the leakage of a potentially significant amount of p-mode power (De Pontieu, Erdélyi, and De Moortel 2005; Jefferies et al. 2006) which could impact the amplitude of the p-mode spectrum (de Moortel and Rosner 2007). The presence of significant oscillatory power in the 1.5 mHz passband in non- flaring active region coronal loops provides us with some interesting scientific return. Previously some oscillatory power at ten minutes had been detected in polar plumes (Deforest and Gurman 1998; Ofman, Nakariakov, and Deforest 1999), but to our knowledge such power had not been reported in coronal loops in active regions. The signal is quite weak in these quiescent active-region loops, but most likely real. The fact that not all coronal loops show this propagating signal strongly suggests that slow co-alignment drifts are not the cause. We have tested this hypothesis by performing the same kind of calculations on dataset I using a variety of methods to remove the slow pointing drift that is introduced into TRACE datasets from thermal flexing of the telescope tube (Aschwanden et al. 2000). The measured propagation speeds of these 1.5 mHz waves are of order 50 km s-1, similar to that of the well- known 3.5 mHz propagating slow-mode oscillations. The presence of these 1.5 mHz waves is intriguing in part because their source in the lower atmosphere may well be different from that of the more well-known 3.5 mHz waves. The leakage of p-modes through the lower atmosphere has been suggested as a source for the prevalence of five minute period waves in the corona by a variety of authors (De Pontieu, Erdélyi, and De Moortel 2005; Jefferies et al. 2006). However, p-modes lack significant power at 1.5 mHz to drive what we see. This suggests that a different source may be working to produce these waves with ten minute periods in active region coronal loops. Recent work (Straus et al. 2008), has seen the identification of horizontally propagating internal gravity waves at what is thought to be the boundary between the photosphere and chromosphere. Long-standing theory (e.g., Lighthill 1967) has these waves coupling strongly to those that propagate along magnetic field lines into the outer atmosphere. A detailed statistical study of these 1.5 mHz waves may well reveal more details about what the real source of these disturbances is and whether these acoustic gravity waves couple into the outer atmosphere. We should note that at least in this one dataset the 1.5 mHz waves are absent from the sunspot fan, and appear only in a subset of plage-associated loops. We should note that it is also possible that these 1.5 mHz waves are actually changes in the coronal loop emission itself as a result of changes in heating or cooling of the loops. Detailed visual inspection of the unfiltered and filtered movies does not allow for an unequivocal identification of the dominant process underlying the signal our algorithm finds. We intend to investigate these 1.5 mHz signals further in follow-up work by studying a wide range of different TRACE datasets. Another new scientific finding of our numerical experiments is the presence of a significant amount of low-frequency oscillatory power in the wake of a coronal dimming. More detailed studies will be necessary to determine why these low frequencies are so prevalent in the dimming regions and whether this can be used to determine physical parameters of the dimming region itself. ## 6 Conclusion We have considered the problem of automatically (and robustly) isolating and extracting information about waves and oscillations observed in EUV image sequences of the solar corona with a view to near real-time application to data from the Atmospheric Imaging Array (AIA) on the Solar Dynamics Observatory (SDO). We have found that a simple coherence / travel-time based approach detects and provides a wealth of information on transverse and longitudinal wave phenomena in the test sequences provided by the Transition Region and Coronal Explorer (TRACE). The results of the search can be robustly “pruned” (based on diagnostic errors) to minimize false-detections such that the remainder provides reliable measurements of waves in the solar corona, with the calculated propagation speed allowing automated distinction between various wave modes and can be automatically applied to the enormous flow of data ($\approx$1Tb day-1) that will be provided by SDO/AIA. In addition to demonstrating the applicability of this approach we have found signatures of some interesting low frequency (1.5 mHz) oscillatory phenomena in the TRACE test data that motivate further study. Figure 1: Examples of the coherence, weighted cross-power, phase difference and phase travel-time for for the 1 July 1998 (panels A-D) and 14 July 1998 (panels E – H) datasets for 5.5 mHz and 1.5 mHz Fourier filter samples respectively. In each case the reference pixel is at 0,0, the solid black contour outlines regions of highly coherent signal and the fitted angle (solid white lines in panels C and G) to central coherence “island” is shown as a solid white line in panels C (14.6∘) and G (-30.4∘). Figure 2: Examples of phase speed determination for the cases shown in Figure 1. In each case the green triangles mark the positions from the reference pixel and measured phase travel-time at that pixel. The gradient of the least- squares linear fit (red dot-dashed line) yields the phase-speed of the propagating signal. Figure 3: Example pixel-to-pixel travel-time analysis in the temporal domain. In panels A and B we show the TRACE 171Å lightcurve (black solid line) and its 5.5 mHz filtered counterpart (blue solid line) between the reference pixel [181, 97] and pixel [179, 97] for the 1 July 1998 dataset (cf. Figure 1). In panel C we show the derived cross-correlation function (black solid line), envelope maxima (red triangles) and fit to the central peak (green solid line). The vertical dashed lines in panel C show the positions of the estimated group travel-time (red; -4.17 minutes) and phase travel-time (green; -0.527 minutes, i.e., consistent with the values in Figures 1 and propagating left to right which is consistent with the movie of Figure 4) respectively. Figure 4: Illustrations of the two TRACE 171Å timeseries used in the presented analysis; 1 July 1998 (left) and 14 July 1998 (right). In each panel we show (clockwise from the top left) snapshots of the TRACE 171Å intensity, 1.5 mHz, 5.5 mHz, and 3.5 mHz filtered timeseries respectively. The boxes marked on the panels correspond to the regions used to construct Figure 1 and the subsequent analyses. See the online edition of the Journal for animations of these figures. Figure 5: Results of the coronal wave detection algorithm for the 1.5 mHz filtered timeseries for the 1 July 1998 dataset. We show the average TRACE 171Å intensity image, weighted signal coherence, length, width, angle, phase speed, and relative error of the phase speed computed from the large coherence island. Figure 6: Results of the coronal wave detection algorithm for the 3.5 mHz filtered timeseries for the 1 July 1998 dataset. We show the average TRACE 171Å intensity image, weighted signal coherence, length, width, angle, phase speed, and relative error of the phase speed computed from the large coherence island. Figure 7: Results of the coronal wave detection algorithm for the 5.5 mHz filtered timeseries for the 1 July 1998 dataset. We show the average TRACE 171Å intensity image, weighted signal coherence, length, width, angle, phase speed, and relative error of the phase speed computed from the large coherence island. 9 Figure 8: Results of the coronal wave detection algorithm for the 1.5 mHz filtered timeseries for the 1 July 1998 dataset. We show the average TRACE 171Å intensity image, weighted signal coherence, length, width, angle, phase speed, and relative error of the phase speed computed from the large coherence island. Figure 9: Results of the coronal wave detection algorithm for the 3.5 mHz filtered timeseries for the 14 July 1998 dataset. We show the average TRACE 171Å intensity image, weighted signal coherence, length, width, angle, phase speed, and relative error of the phase speed computed from the large coherence island. Figure 10: Results of the coronal wave detection algorithm for the 5.5 mHz filtered timeseries for the 14 July 1998 dataset. We show the average TRACE 171Å intensity image, weighted signal coherence, length, width, angle, phase speed, and relative error of the phase speed computed from the large coherence island. Figure 11: Final results of the coronal wave detection algorithm for the 1.5 mHz filtered timeseries for the 14 July 1998 dataset that have been “pruned” to reduce the appearance of false positive detections using the technique discussed in Sect. 3.2. We show the average TRACE 171Å intensity image and the region averaged signal coherence, length, angle and phase speed, and the errors in the latter. Figure 12: Final results of the coronal wave detection algorithm for the 3.5 mHz filtered timeseries for the 1 July 1998 dataset that have been “pruned” to reduce the appearance of false positive detections using the technique discussed in Sect. 3.2. We show the average TRACE 171Å intensity image and the region averaged signal coherence, length, angle and phase speed, and the errors in the latter. Figure 13: Final results of the coronal wave detection algorithm for the 5.5 mHz filtered timeseries for the 1 July 1998 dataset that have been “pruned” to reduce the appearance of false positive detections using the technique discussed in Sect. 3.2. We show the average TRACE 171Å intensity image and the region averaged signal coherence, length, angle and phase speed, and the errors in the latter. Figure 14: Final results of the coronal wave detection algorithm for the 1.5 mHz filtered timeseries for the 14 July 1998 dataset that have been “pruned” to reduce the appearance of false positive detections using the technique discussed in Sect. 3.2. We show the average TRACE 171Å intensity image and the region averaged signal coherence, length, angle and phase speed, and the errors in the latter. Figure 15: Final results of the coronal wave detection algorithm for the 3.5 mHz filtered timeseries for the 14 July 1998 dataset that have been “pruned” to reduce the appearance of false positive detections using the technique discussed in Sect. 3.2. We show the average TRACE 171Å intensity image and the region averaged signal coherence, length, angle and phase speed, and the errors in the latter. Figure 16: Final results of the coronal wave detection algorithm for the 5.5 mHz filtered timeseries for the 14 July 1998 dataset that have been “pruned” to reduce the appearance of false positive detections using the technique discussed in Sect. 3.2. We show the average TRACE 171Å intensity image and the region averaged signal coherence, length, angle and phase speed, and the errors in the latter. SWM is supported by NSF ATM-0541567, NASA NNG06GC89G; BDP by NASA grants NAS5-38099 (TRACE), NNM07AA01C (Hinode) and NNG06GG79G. SWM and BDP are jointly supported by NASA grants NNX08AL22G and NNX08AH45G. ## References * Arregui et al. (2007) Arregui, I., Andries, J., van Doorsselaere, T., Goossens, M., Poedts, S.: 2007, MHD seismology of coronal loops using the period and damping of quasi-mode kink oscillations. A&A 463, 333 – 338. doi:10.1051/0004-6361:20065863. * Aschwanden et al. (2002) Aschwanden, M.J., de Pontieu, B., Schrijver, C.J., Title, A.M.: 2002, Transverse Oscillations in Coronal Loops Observed with TRACE II. Measurements of Geometric and Physical Parameters. Sol. Phys. 206, 99 – 132. doi:10.1023/A:1014916701283. * Aschwanden et al. (1999) Aschwanden, M.J., Fletcher, L., Schrijver, C.J., Alexander, D.: 1999, Coronal Loop Oscillations Observed with the Transition Region and Coronal Explorer. ApJ 520, 880 – 894. doi:10.1086/307502. * Aschwanden et al. (2000) Aschwanden, M.J., Nightingale, R.W., Tarbell, T.D., Wolfson, C.J.: 2000, Time Variability of the “Quiet” Sun Observed with TRACE. I. Instrumental Effects, Event Detection, and Discrimination of Extreme-Ultraviolet Microflares. ApJ 535, 1027 – 1046. doi:10.1086/308866. * Bevington and Robinson (2003) Bevington P.R., Robinson D.K.: 2003, Data reduction and error analysis for the physical sciences. Data reduction and error analysis for the physical sciences, 3rd ed., by Philip R. Bevington, and Keith D. Robinson. Boston, MA: McGraw-Hill, ISBN 0-07-247227-8, 2003.. * de Moortel, Ireland, and Walsh (2000) de Moortel, I., Ireland, J., Walsh, R.W.: 2000, Observation of oscillations in coronal loops. A&A 355, L23 – L26. * de Moortel et al. (2002a) de Moortel, I., Ireland, J., Walsh, R.W., Hood, A.W.: 2002a, Longitudinal intensity oscillations in coronal loops observed with TRACE I. Overview of Measured Parameters. Sol. Phys. 209, 61 – 88. doi:10.1023/A:1020956421063. * de Moortel et al. (2002b) de Moortel, I., Hood, A.W., Ireland, J., Walsh, R.W.: 2002b, Longitudinal intensity oscillations in coronal loops observed with TRACE II. Discussion of Measured Parameters. Sol. Phys. 209, 89 – 108. doi:10.1023/A:1020960505133. * de Moortel and Rosner (2007) de Moortel, I., Rosner, R.: 2007, An Estimate of P-Mode Damping by Wave Leakage. Sol. Phys. 246, 53 – 63. doi:10.1007/s11207-007-0392-6. * De Pontieu, Erdélyi, and De Moortel (2005) De Pontieu, B., Erdélyi, R., De Moortel, I.: 2005, How to Channel Photospheric Oscillations into the Corona. ApJ 624, L61 – L64. doi:10.1086/430345. * De Pontieu, Tarbell, and Erdélyi (2003) De Pontieu, B., Tarbell, T., Erdélyi, R.: 2003, Correlations on Arcsecond Scales between Chromospheric and Transition Region Emission in Active Regions. ApJ 590, 502 – 518. doi:10.1086/374928. * Deforest and Gurman (1998) Deforest, C.E., Gurman, J.B.: 1998, Observation of Quasi-periodic Compressive Waves in Solar Polar Plumes. ApJ 501, L217 – L220. doi:10.1086/311460. * Finsterle et al. (2004) Finsterle, W., Jefferies, S.M., Cacciani, A., Rapex, P., McIntosh, S.W.: 2004, Helioseismic Mapping of the Magnetic Canopy in the Solar Chromosphere. ApJ 613, L185 – L188. doi:10.1086/424996. * Handy et al. (1999) Handy, B.N., Acton, L.W., Kankelborg, C.C., Wolfson, C.J., Akin, D.J., Bruner, M.E., Caravalho, R., Catura, R.C., Chevalier, R., Duncan, D.W., Edwards, C.G., Feinstein, C.N., Freeland, S.L., Friedlaender, F.M., Hoffmann, C.H., Hurlburt, N.E., Jurcevich, B.K., Katz, N.L., Kelly, G.A., Lemen, J.R., Levay, M., Lindgren, R.W., Mathur, D.P., Meyer, S.B., Morrison, S.J., Morrison, M.D., Nightingale, R.W., Pope, T.P., Rehse, R.A., Schrijver, C.J., Shine, R.A., Shing, L., Strong, K.T., Tarbell, T.D., Title, A.M., Torgerson, D.D., Golub, L., Bookbinder, J.A., Caldwell, D., Cheimets, P.N., Davis, W.N., Deluca, E.E., McMullen, R.A., Warren, H.P., Amato, D., Fisher, R., Maldonado, H., Parkinson, C.: 1999, The transition region and coronal explorer. Sol. Phys. 187, 229 – 260. doi:10.1023/A:1005166902804. * Hansteen et al. (2006) Hansteen, V.H., De Pontieu, B., Rouppe van der Voort, L., van Noort, M., Carlsson, M.: 2006, Dynamic Fibrils Are Driven by Magnetoacoustic Shocks. ApJ 647, L73 – L76. doi:10.1086/507452. * Jefferies et al. (2006) Jefferies, S.M., McIntosh, S.W., Armstrong, J.D., Bogdan, T.J., Cacciani, A., Fleck, B.: 2006, Magnetoacoustic Portals and the Basal Heating of the Solar Chromosphere. ApJ 648, L151 – L155. doi:10.1086/508165. * Jefferies et al. (1994) Jefferies, S.M., Osaki, Y., Shibahashi, H., Duvall, T.L. Jr., Harvey, J.W., Pomerantz, M.A.: 1994, Use of acoustic wave travel-time measurements to probe the near-surface layers of the Sun. ApJ 434, 795 – 800. doi:10.1086/174782. * Jefferies et al. (1997) Jefferies, S.M., Osaki, Y., Shibahashi, H., Harvey, J.W., D’Silva, S., Duvall, T.L. Jr.: 1997, Sounding the Sun’s Chromosphere. ApJ 485, L49 – L52. doi:10.1086/310805. * Khomenko et al. (2008) Khomenko, E., Centeno, R., Collados, M., Trujillo Bueno, J.: 2008, Channeling 5 Minute Photospheric Oscillations into the Solar Outer Atmosphere through Small-Scale Vertical Magnetic Flux Tubes. ApJ 676, L85 – L88. doi:10.1086/587057. * King et al. (2003) King, D.B., Nakariakov, V.M., Deluca, E.E., Golub, L., McClements, K.G.: 2003, Propagating EUV disturbances in the Solar corona: Two-wavelength observations. A&A 404, L1 – L4. doi:10.1051/0004-6361:20030763. * Lighthill (1967) Lighthill M.F.: 1967, Predictions on the Velocity Field Coming from Acoustic Noise and a Generalized Turbulence in a Layer Overlying a Convectively Unstable Atmospheric Region. In: Thomas, R.N. (ed.) Aerodynamic Phenomena in Stellar Atmospheres, IAU Symposium, 28, 429 – 469. * McIntosh, Fleck, and Tarbell (2004) McIntosh, S.W., Fleck, B., Tarbell, T.D.: 2004, Chromospheric Oscillations in an Equatorial Coronal Hole. ApJ 609, L95 – L98. doi:10.1086/422748. * Nakariakov and King (2007) Nakariakov, V.M., King, D.B.: 2007, Coronal Periodmaps. Sol. Phys. 241, 397 – 409. doi:10.1007/s11207-007-0348-x. * Nakariakov and Verwichte (2005) Nakariakov, V.M., Verwichte, E.: 2005, Coronal Waves and Oscillations. Living Reviews in Solar Physics 2. http://solarphysics.livingreviews.org/Articles/lrsp-2005-3/ * Ofman and Aschwanden (2002) Ofman, L., Aschwanden, M.J.: 2002, Damping Time Scaling of Coronal Loop Oscillations Deduced from Transition Region and Coronal Explorer Observations. ApJ 576, L153 – L156. doi:10.1086/343886. * Ofman, Nakariakov, and Deforest (1999) Ofman, L., Nakariakov, V.M., Deforest, C.E.: 1999, Slow Magnetosonic Waves in Coronal Plumes. ApJ 514, 441 – 447. doi:10.1086/306944. * Robbrecht et al. (2001) Robbrecht, E., Verwichte, E., Berghmans, D., Hochedez, J.F., Poedts, S., Nakariakov, V.M.: 2001, Slow magnetoacoustic waves in coronal loops: EIT and TRACE. A&A 370, 591 – 601. doi:10.1051/0004-6361:20010226. * Straus et al. (2008) Straus T., Fleck B., Jefferies S.M., Cauzzi G., McIntosh S.W., Reardon K., Severino G., Steffen M.: 2008, The Energy Flux of Internal Gravity Waves in the Lower Solar Atmosphere. ApJ 681, L125 – L128. doi:10.1086/590495. * Sych and Nakariakov (2008) Sych R.A., Nakariakov V.M.: 2008, The Pixelised Wavelet Filtering Method to Study Waves and Oscillations in Time Sequences of Solar Atmospheric Images. Sol. Phys. 248, 395 – 408. 10.1007/s11207-007-9005-7. * Terradas, Andries, and Goossens (2007) Terradas, J., Andries, J., Goossens, M.: 2007, On the Excitation of Leaky Modes in Cylindrical Loops. Sol. Phys. 246, 231 – 242. doi:10.1007/s11207-007-9067-6. * Tomczyk et al. (2007) Tomczyk, S., McIntosh, S.W., Keil, S.L., Judge, P.G., Schad, T., Seeley, D.H., Edmondson, J.: 2007, Alfvén Waves in the Solar Corona. Science 317, 1192 – 1196. doi:10.1126/science.1143304. * Tomczyk, and McIntosh (2008) Tomczyk, S., McIntosh, S.W.: 2008, Time-Distance Coronal Seismology, ApJLett. in preparation	arxiv-papers	2008-08-21T19:06:29	2024-09-04T02:48:57.385812	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Scott W. McIntosh, Bart De Pontieu, Steven Tomczyk", "submitter": "Scott McIntosh", "url": "https://arxiv.org/abs/0808.2978" }
0808.3015	# Phase Coherence Effects in the Vortex Transport Entropy G. Bridoux bridoux@cabbat1.cnea.gov.ar G. Nieva F. de la Cruz Centro Atómico Bariloche and Instituto Balseiro, Comisión Nacional de Energía Atómica, Av. E. Bustillo 9500, R84002AGP S. C. de Bariloche, Argentina ###### Abstract Nernst and electrical resistivity measurements in superconducting YBa2Cu3O7-δ (YBCO) and Bi2Sr2CaCu2O8+δ(BSCCO) with and without columnar defects show a distinctive thermodynamics of the respective liquid vortex matter. At a field dependent high temperature region in the $H-T$ phase diagram the Nernst signal is independent of structural defects in both materials. At lower temperatures, in YBCO, defects contribute only to the vortex mobility and the transport entropy is that of a system of vortex lines. The transition to lower temperatures in BSCCO has a different origin, the maximum in the Nernst signal when decreasing temperature is not associated with transport properties but with the entropy behavior of pancake vortices in the presence of structural defects. ###### pacs: 74.25.Op, 74.72.-h, 74.25.Fy, 74.40.+k ††preprint: APS/123-QED The Nernst signal in superconductors is associated with the displacement of vortices induced by the presence of a temperature gradient, $\nabla$$T$, perpendicular to the internal magnetic field, B. Its detection is made measuring the Josephson voltage induced by vortices crossing electrical contacts aligned in a direction perpendicular to both, B and $\nabla$$T$. For an ideal superconductor and straight vortex lines if $\nabla$$T$ is in the x̂ direction, ($\nabla$$T$)x, and B in the ẑ direction the Nernst electric field is proportional to ($\nabla$$T$)x CampbellEvets . Within this regime it was shown CampbellEvets that the thermal force per unit vortex length is $F_{T}(H)=S_{\phi}(H,T)(\textbf{$\nabla$}$T$)_{x}$. Here $S_{\phi}(H,T)$ is the transport entropy per unit vortex length. In this limit the impedance to vortex displacement is the flux flow vortex viscosity, associated with the flux flow electrical resistivity, $\rho_{f}$. Thus, the Nernst signal is found Ri $e_{N}(H,T)=\frac{E_{y}}{(\nabla T)_{x}}=\frac{\rho_{f}S_{\phi}(H,T)}{\phi_{0}}$ (1) where $E_{y}$ is the Nernst electrical field and $\phi_{0}$ the flux quantum. In the normal state $e_{N}(T,H)$ is essentially zero Ong and since $S_{\phi}(H,T)=0$ at either $T=0$ or $H$ less than the lower critical field $H_{c1}(T)$ we see that $e_{N}$ should show a maximum as a function of $H$ or $T$. In real materials vortex pinning inhibits vortex displacements within an ohmic regime FisherErnesto . Therefore, the maximum of $e_{N}(T,H)$ is usually determined by the field or temperature where pinning reduces vortex mobility to zero. In low $T_{c}$ materials this imposes strong limitations on the range of fields and temperatures where $e_{N}(T,H)$ can be used to determine $S_{\phi}(H,T)$. In high $T_{c}$ superconductors the short coherence length, strong material anisotropy and thermal energy are responsible for a transition from a solid to a liquid vortex state in a wide region of the $H-T$ phase diagram. This phase diagram for a superconductor opened a renewed interest in the study of vortex physics: solid-liquid phase transitions, vortex cutting and reconnection, low vortex dimensionality, vortex decoupling and the contribution of thermal fluctuations play an important roll Blatter . The pioneering work by Ri et al. Ri showed that the Nernst effect in films of YBCO and BSCCO responds, qualitatively, to the behavior expected from Ginzburg-Landau theory. Recent work by Ong and collaborators Ong triggered intensive experimental and theoretical activity. Previous work on YBCO and BSCCO showed Phylos SafarBush that using the dc flux transformer contact configuration in twinned YBCO crystals the $c$ axis vortex phase correlation across the sample was established at a well defined sample thickness dependent temperature, $T_{th}(d,H)$. At $T_{th}(d,H)$ the resistivity in the $c$ direction, $\rho_{c}(H,T)$, drops abruptly and the resistivity in the $ab$ plane, $\rho_{ab}(H,T)$, of phase correlated vortices across the sample decreases Phylos and becomes zero at a thickness independent temperature $T_{i}(H)$, where the vortex liquid-solid transition takes place. In BSCCO no vortex phase correlation SafarBush across the sample in the $c$ direction was established in the liquid region of the $H-T$ phase diagram. From these transport properties of the two paramount high $T_{c}$ superconductors we expect that if the maximum of the Nernst signal is due to the decrease of vortex mobility it should take place at the proximity of $T_{th}(d,H)$ in YBCO and of $T_{i}(H)$ in BSCCO. To study the relevance of the vortex phase correlation in the field direction we measured $e_{N}(T,H)$ and the resistivity of optimally doped single crystals of YBCO and BSCCO with and without columnar defects, CD, for $\textbf{B}\\|\textbf{c}$ axis. In twinned YBCO crystals the sample thickness dependent flux cutting and reconnection induces the field and temperature dependent maximum of $e_{N}(T,H)$ but it does not contribute to the vortex transport entropy. We show that the maximum in $e_{N}(T,H)$ in YBCO below $T_{c}$ is induced by the size dependent vortex mobility. While in YBCO the vortices in the liquid state are accepted to respond as a three dimensional system, in BSCCO the vortices are considered to be BulaKosh uncoupled pancakes nucleated on Cu-O planes. We have found that the maximum in the Nernst voltage below $T_{c}$ in BSCCO is not associated with the vortex mobility in the liquid state but to an intrinsic change of the temperature dependence of $S_{\phi}$ of vortices in this phase. The YBCO and BSCCO single crystals were grown as described in references Phylos and Kaul . The columnar defects, nearly parallel to the $c$ axis, with a dose equivalent field of $B_{\phi}=3\,$T were created by irradiation with 278 MeV Sb24+ for BSCCO and 309 MeV Au26+ ions for YBCO at TANDAR-Argentina. The ions tracks are of the order of the coherence length of both materials (5-10 nm in diameter). The $T_{c}$’s of the non-irradiated samples used in the Nernst and the electrical resistance experiment were $T_{c}=$ 91.0 K and 91.3 K respectively for BSCCO and 93.6 K for YBCO. The irradiated samples have $T_{c}=$ 90.2 K for BSCCO and $T_{c}=$ 91.7 K for YBCO. The samples thickness were 20-40 ${\mu}$m for the non-irradiated samples and 8-15 ${\mu}$m for the irradiated ones. The Nernst angulo and the electrical resistance Phylos measurement setup were described elsewhere. In Fig.1(b) we show $e_{N}(T,H)$ and the electrical resistivity of YBCO crystals for $H=5$, 6 and 8 T with and without CD. Once the vortex system melts the mobility is finite, $e_{N}(T,H)$ grows fast with temperature up to a maximum at $T_{max}(H)$. At higher temperatures the Nernst voltage decreases towards the almost zero normal state value at temperatures well above $T_{c}$. The mobility edge of the irradiated sample moves to higher $T$ and so does the corresponding $T_{max}(H)$. The data suggest that $e_{N}(T,H)$ between the mobility edge and $T_{max}(H)$ is strongly determined by pinning. Figure 1: (a) $H-T$ diagram. The lines $T_{id}$ (up triangles), $T_{c2}$ (down triangles and linear fit) and the irreversibility lines $T_{i}$ for the irradiated (circles) and non-irradiated samples (squares) are shown. (b) Nernst signal $e_{N}$ and electrical resistivity $\rho_{ab}$ vs. $T/T_{c}$ in YBCO at $H=5,\,6$ and $8\,$T for the irradiated (circles) and non -irradiated samples (squares). $T_{th}$ of both samples are indicated with arrows following the references Phylos Righi . Gray regions in both panels are equivalent. More important, $e_{N}(T,H)$ of samples with and without CD coincides, within experimental uncertainty, at temperatures equal and above $T_{id}(H)$ as shown by a tilted arrow in Fig.1(b). It is found that the resistivities of the samples with and without CD become equal in the same range of temperatures, see Fig.1(b). This is observed in the whole range of field investigated, from 2 to 10 T, suggesting that for $T\geq T_{id}(H)$ the columnar defects and, possibly, all other type of pinning centers, are irrelevant for the vortex response to gradient of temperatures or electrical currents. Thus, $\rho_{ab}$, $e_{N}$ and consequently the entropy of samples with and without CD coincide for $T\geq T_{id}(H)$. From the previous discussion and experimental observation we define $T_{id}(H)$ as the highest temperature where vortex pinning is effective, as determined by $e_{N}$ measurements. Above $T_{id}(H)$ the pinning potential becomes irrelevant, independently of the type or strength of the pinning potential. As we demonstrate below for $T<T_{id}(H)$ not only the pinning potential of the columnar defects is switched on but also that associated with the defects in the non-irradiated sample as made evident by the behavior of $\rho_{ab}(H,T)$, see Fig.1(b), and the corresponding transport entropy of the vortices. This makes $T_{id}(H)$ a relevant line in the phase diagram of the liquid state of vortex lines. The gray region in Fig.1(b) indicates the experimental uncertainty for $T_{id}(H)$. In Fig.1(a) the line $T_{id}(H)$ is plotted together with the pinning potential dependent solid-liquid phase transition lines $T_{i}(H)$. From the definition of $T_{id}(H)$ we recognize that for $T\geq T_{id}(H)$, $U_{\phi}(T,H)=TS_{\phi}(H,T)$ should be the same for the YBCO samples investigated, independently of the type of pinning potential characterizing the transport properties at lower temperatures. Therefore, $U_{\phi}(T,H)$ becomes an intrinsic property of the ideal vortex system for $T\geq T_{id}(H)$. In Fig.2 we plotted $U_{\phi}(T,H)$ of samples with and without CD as a function of temperature at 8 T. The data are representative of $U_{\phi}(T,H)$ in the whole range of fields investigated. We see that both $U_{\phi}(T,H)$ coincide not only for $T\geq T_{id}(H)$ but also in a broader range of temperatures below $T_{id}(H)$. This demonstrates that the maximum of $e_{N}(H,T)$ is only induced by the effective pinning potential acting on electrical transport properties, see inset in Fig.2 for the results in the irradiated sample. Somewhat below its $T_{th}(H)$, $U_{\phi}(T,H)$ shows an anomalous increase when decreasing temperature as compared to that of the non- irradiated sample. The analysis of this feature helped to understand the relative contribution of $\rho_{ab}$ and $S_{\phi}$ to $e_{N}$ in YBCO. Figure 2: Uϕ vs. $T/T_{c}$ in YBCO at $H=8\,$T for the irradiated (red triangles plus line) and the non-irradiated sample (black squares, interpolation with small symbols). Inset: Measured $e_{N}$ and $\rho_{ab}$ (red circles and black squares respectively) and recalculated $e_{N}$ (small red circles) for the irradiated crystal. We see from Fig.2 that $U_{\phi}(T,H)$ of the non-irradiated sample increases linearly when decreasing $T$ well below the $T_{th}(H)$ of the irradiated sample. This linear dependence is expected in a mean field description CampbellEvets , its extrapolation to $U_{\phi}(T,H)=0$ provides a mean field $T_{c2}(H)$, also plotted in Fig.1(a). The results support the idea that the measured $U_{\phi}(T,H)$ is that of an intrinsic bulk superconductor, independently of the amount and type of pinning centers. Assuming this we calculated the $e_{N}(H,T)$ of the sample with CD using its measured $\rho_{ab}$ and the $U_{\phi}(H,T)$ from the sample without CD, as plotted in the inset of Fig.2. The results are important and revealing. We see that the calculated $e_{N}(H,T)$ reproduces the experimental data within the experimental error from temperatures well above $T_{c}$ to somewhat below the temperature at the maximum. More important, the calculated data extrapolates to the same $T_{i}(H)$, detected by electrical transport. This result makes evident the origin of the anomalous raise of $U_{\phi}(T,H)$ at temperatures below the maximum, in both samples. The thermal gradient used to determine the Nernst effect induces a force that exceeds the extremely small linear response regime close to the solid-liquid transition Grigera . The thermal force in this range of temperatures is calculated to be Bridoux up to one order of magnitude larger than that used in electrical measurements. The results show that the maximum of $e_{N}(H,T)$ in YBCO is a size effect that reflects the growth of the vortex phase coherence in the $c$ direction. The maximum takes place close to $T_{th}(d,H)$ where the $c$ axis vortex phase correlation length coincides with the sample thickness. Figure 3: (a) Nernst signal $e_{N}$ at $H=3$ and $5\,$T vs. $T/T_{c}$ in BSCCO . Red circles and black squares curves correspond to the irradiated ($B_{\phi}=3\,$T) and non-irradiated sample respectively. The dashed curves correspond to the Nernst signal of the irradiated sample normalized at the corresponding maximum of the non-irradiated one. The shading emphasizes the maximum region. (b) Entropy per unit vortex length $S_{\phi}$ ($S_{\phi}=e_{N}{\phi}_{0}/\rho_{ab}$) vs. $T/T_{c}$ in BSCCO at $H=5\,$T for the irradiated (red circles) and the non-irradiated samples (black squares). The dashed curve corresponds to $S_{\phi}$ of the irradiated sample normalized at the corresponding maximum of the non-irradiated one. The $\rho_{c}$ vs. $T/T_{c}$ at $H=5\,$T (up triangles) and $8\,$T (dash-dotted line). The inset shows $\rho_{ab}$ at $H=5\,$T for the samples with (red circles) and without CD (black squares). While the maximum occurs at a thickness dependent temperature the inflection of the Nernst signal occurs at the thickness independent $T_{id}(H)$. A remarkable result associated with $T_{id}(H)$ is that in this line the linear temperature dependence characteristic of a mean field behavior of $U_{\phi}$ is lost, see Fig.2. At higher temperatures the curvature of $U_{\phi}(T,H)$ indicates the dominant contribution of thermal fluctuations, insensitive to the presence of structural defects. In this sense $T_{id}(H)$ might well be an experimental verification of that proposed by Nguyen and SudbøSubdo where thermally induced vortex loops proliferation causes the loss of vortex line tension. The Nernst signal and electrical resistivity of BSCCO samples with and without CD for fields from 1T to 16 T were measured. Fig.3(a) shows typical results, in this case for $3$ and $5\,$T. In BSCCO we found that the absolute value of $e_{N}(H,T)$ of the sample with CD is systematically smaller than that of the non-irradiated sample. It is remarkable that for $T\geq T_{max}(H)=T_{id}(H)$ the ratio of $e_{N}(H,T)$ between the irradiated and non-irradiated samples is only a function of $H$ and not of temperature befi , see Fig. 3(a). At first glance some similarities are found between the results of BSCCO and YBCO: For $T\geq T_{id}(H)$, the $e_{N}(H,T)$ of YBCO as well as the normalized $e_{N}(H,T)$ of BSCCO in samples with and without CD coincide up to temperatures well above their respective $T_{c}$; for $T\leq T_{max}(H)$, $e_{N}(H,T)$ goes to zero at the corresponding $T_{i}(H)$ in each material. On the other hand differences are evident. Contrary to what is measured in YBCO, $T_{max}(H)$ of BSCCO is the same for samples with and without CD, as well as for samples of different thicknesses Ri ; in YBCO the rapid decrease of $\rho_{ab}(H,T)$ in each type of sample determines the maximum of $e_{N}(H,T)$; in BSCCO $\rho_{ab}(H,T)$ is independent of the presence of CD to temperatures well below $T_{max}(H)$, see inset of Fig. 3(b). The analysis of $e_{N}(T,H)$ and $\rho_{ab}(H,T)$ for samples with and without CD allows the detection of magnitudes associated with the equilibrium thermodynamic state of the vortex system. In YBCO, $T_{max}(H)$ is determined by the behavior of $\rho_{ab}(H,T)$ approaching $T_{i}(H)$. Thus, the maximum of $e_{N}(H,T)$ is not associated with an equilibrium state. On the contrary, that maximum in BSCCO is determined by the maximum of an equilibrium property $S_{\phi}(H,T)$, Fig.3(b), for all fields and temperatures studied Bridoux . This is an interesting and puzzling result: While the vortex mobility, $\rho_{ab}(H,T)$ as well as the temperature $T_{max}(H)=T_{id}(H)$ remain unaffected by the nature of the pinning potential for temperatures in the neighborhood of the maximum of $e_{N}(H,T)$, it is the thermodynamic equilibrium property $S_{\phi}(H,T)$ that becomes dependent on the type of structural defects for $T<T_{id}(H)$. The liquid vortex state in BSCCO is accepted to be BulaKosh that of 2D uncoupled vortices nucleated in Cu-O planes. On the other hand, the data in this work show that below the temperature $T_{id}(H)$ (identical for samples with and without CD) $S_{\phi}(H,T)$ becomes dependent on the type of structural defects. Thus, the order induced by correlated defects below $T_{id}(H)$ requires a change of the nature of the vortex system at this temperature. This is supported by the incipient decrease of $\rho_{c}(T,H)$ from the non-metallic normal state at $T_{id}(H)$ in both samples, see Fig. 3(b). This points out that for $T<T_{id}(H)$ the interaction between pancakes vortices in the neighbor Cu-O planes should not be disregarded. This interaction could induce a change of the effective dimensionality of the vortex system in BSCCO, decreasing the configurational entropy contribution Huse in the presence of columnar defects. In conclusion, transport and Nernst measurements in samples with and without correlated defects reveal the origin of the maximum of the Nernst signal in the liquid state of dissimilar vortices. In YBCO, vortex cutting and reconnection Blatter induces a sample size dependent $T_{max}(H)$, well below the corresponding $T_{id}(H)$. In this case the vortex transport entropy remains independent of the nature of defects. On the contrary, in BSCCO the maximum is due to the intrinsic maximum of the vortex transport entropy, suggesting a change of the effective vortex dimensionality below the well defined temperature where the resistance in the $c$ direction starts to deviate from its normal state value. We acknowledge correspondence with L. Bulaevskii and D. Huse. We thank L. Civale and H. Lanza for the irradiation, E. E. Kaul and N. Saenger for help with sample preparation and measurements. G. B. and G. N. acknowledge financial support from CONICET-Argentina. ## References * (1) A. M. Campbell and J. E. Evetts, Advances in Physics 50, 1249 (2001), and references there in. * (2) H.-C. Ri, R. Gross, F. Gollnik, A. Beck, R. P. Huebener, P. Wagner and H. Adrian. Phys. Rev. B 50, 3312 (1994). * (3) Y. Wang, L. Li, and N. P. Ong, Phys. Rev. B 73, 024510 (2006) and references there in. * (4) D. S. Fisher, Phys. Rev. B 31, 1396 (1985); F. de la Cruz, J. Luzuriaga, E. N. Martinez, and E. J. Osquiguil. Phys. Rev. B 36, 6850 (1987). * (5) G. Blatter, M. V. Feigel’man, V. B. Geshkenbein, A. I. Larkin, and V. M. Vinokur, Rev. Mod. Phys. 66, 1125 (1994). * (6) F. de la Cruz, D. López, and G. Nieva, Philos. Mag. B 70, 773 (1994). * (7) H. Safar, E. Rodriguez, F. de la Cruz, P. L. Gammel, L. F. Schneemeyer, and D. J. Bishop. Phys. Rev. B 46, 14238 (1992); R. Busch, G. Ries, H. Werthner, G. Kreiselmeyer, and G. Saemann-Ischenko, Phys. Rev. Lett. 69, 522 (1992). * (8) L. N. Bulaevskii, A. E. Koshelev, V. M. Vinokur, and M. P. Maley, Phys. Rev. B 61, R3819 (2000). * (9) E. E. Kaul, and G. Nieva, Physica C 341-348, 1343 (2000). * (10) G. Bridoux, P. Pedrazzini, F. de la Cruz, and G. Nieva, Physica C 460-462, 841 (2007). * (11) E. F. Righi, S. A. Grigera, D. López, G. Nieva, F. de la Cruz, L. Civale, G. Pasquini, and P. Levy. Phys. Rev. B 55, 5663 (1997). * (12) S. A. Grigera, E. Morré, E. Osquiguil, C. Balseiro, G. Nieva, and F. de la Cruz, Phys. Rev. Lett. 81, 2348 (1998). * (13) G. Bridoux, Ph.D. thesis, Instituto Balseiro, 2008 (unpublished). * (14) A. K. Nguyen and A. Sudbø, Phys. Rev. B 60, 15307 (1999). * (15) The ratio of the absolute values of $S_{\phi}$ for samples without and with CD above $T_{max}(H)$ in BSCCO varies as $B_{\phi}/H+1$. * (16) S. Mukerjee and D. A. Huse, Phys. Rev. B 70, 014506 (2004).	arxiv-papers	2008-08-22T00:27:34	2024-09-04T02:48:57.393196	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "G. Bridoux, G. Nieva, and F. de la Cruz", "submitter": "German Bridoux", "url": "https://arxiv.org/abs/0808.3015" }
0808.3087	# Noise in random Boolean networks Tiago P. Peixoto tiago@fkp.tu-darmstadt.de Barbara Drossel drossel@fkp.tu- darmstadt.de Institut für Festkörperphysik, TU Darmstadt, Hochschulstrasse 6, 64289 Darmstadt, Germany ###### Abstract We investigate the effect of noise on Random Boolean Networks. Noise is implemented as a probability $p$ that a node does not obey its deterministic update rule. We define two order parameters, the long-time average of the Hamming distance between a network with and without noise, and the average frozenness, which is a measure of the extent to which a node prefers one of the two Boolean states. We evaluate both order parameters as function of the noise strength, finding a smooth transition from deterministic ($p=0$) to fully stochastic ($p=1/2$) dynamics for networks with $K\leq 2$, and a first order transition at $p=0$ for $K>2$. Most of the results obtained by computer simulation are also derived analytically. The average Hamming distance can be evaluated using the annealed approximation. In order to obtain the distribution of frozenness as function of the noise strength, more sophisticated self-consistent calculations had to be performed. This distribution is a collection of delta peaks for $K=1$, and it has a fractal sructure for $K>1$, approaching a continuous distribution in the limit $K\gg 1$. ###### pacs: 89.75.Da,05.65.+b,91.30.Dk,91.30.Px ## I Introduction Random Boolean Networks (RBNs) Kauffman (1969); Drossel (2008) have been used as a simple model for a variety of dynamical systems consisting of interacting units, such as neural networks Rosen-Zvi et al. (2001), social networks Moreira et al. (2004) and, more prominently, gene regulatory networks Lagomarsino et al. (2005); Kauffman (1969). RBNs are composed of Boolean nodes that are coupled to each other. In the case of gene regulatory networks, the Boolean state is a step-function approach to the expression level of a particular gene. Despite this loss of detail, the most important features of gene regulatory processes are still captured in many cases, since they should not depend on biochemical details, but on the desired sequence of events in the cell Bornholdt (2005). So far, the dynamics of RBNs have mostly been studied using deterministic update rules. The dynamics of such models are non-ergodic, with periodic attractor trajectories in state space. Once the system has reached an attractor, it remains there. Another important property of RBNs is a phase transition, which occurs when the number $K$ of inputs per node is changed. For unbiased networks, the dynamics exhibit a frozen phase at $K=1$, where local perturbations die out quickly and most attractors are fixed points, and a “chaotic” phase at $K>2$, where perturbations increase exponentially fast and attractors have very long periods. At the boundary $K=2$ between those two phases are the so-called “critical” networks, where perturbations increase algebraically with time. Originally, it was suggested by Kauffmann Kauffman (1969) that such critical networks are best suited to model real systems, which are supposedly poised “at the edge of chaos”. In the meantime, there is agreement that RBNs of all three types have only limited validity when applied to real systems. Real networks usually have some level of stochastic behaviour, and for this reason several authors have investigated RBNs under the influence of stochasticity. For instance, in Greil and Drossel (2005) the nodes of RBNs were updated in a completely random order. This update method preserves the non-ergodicity of the system, and it is still possible to identify distinct attractors. Attractors are in this case defined as sets of states all of which are visited for a non-vanishing proportion of time during the same trajectory. The stochastic update sequence vastly reduces the number of attractors of critical RBNs, which becomes a power law as function of the network size. Similar results are obtained when the update sequence deviates only slightly from a synchronous update Klemm and Bornholdt (2005). Such a power law was for a long time falsely believed to occur in deterministic RBNs Kauffman (1969); Samuelsson and Troein (2003); Drossel (2005); Drossel et al. (2005). Instead of introducing stochasticity into the update times, other authors introduce it into the update functions. In Shmulevich et al. (2002), probabilistic Boolean functions are used, where a set of several Boolean functions is assigned to each node, and at each time step one of these is chosen randomly with a given probability. According to Shmulevich et al. (2002), this model is more realistic than models with a purely deterministic update scheme. However, the most important way of introducing noise into a RBN is in form of a “temperature”, leading often to ergodic behavior. The effect of thermal noise on Ising spins on a network was studied in Indekeu (2004); Aleksiejuk et al. (2002), where a “ferromagnetic” transition from the ordered to the disordered phases was observed at a critical noise strength value. In the language of gene regulatory networks, a temperature manifests itself as fluctuations in the protein concentrations, so that a gene may not always be turned on or off, given the same expression state of the other genes McAdams and Arkin (1997). This effect can be included into models by allowing a deviation from the deterministic update rule with a certain probability. In Fretter and Drossel (2008), for instance, a subset of nodes were perturbed in this way (this corresponds to turning on the temperature for a short time interval), and the response of the dynamics to this perturbation was evaluated, giving information about the basin structure of the system. Miranda et al. Miranda and Parga (1989) studied the effect of a permanently acting temperature by introducing a fixed probability $p$ that the state of a node becomes the opposite of what it should be according to the deterministic update rule. They evaluated the average crossing time between trajectories in state space which started from different initial states as function of noise strength Miranda and Parga (1989); Golinelli and Derrida (1989); Qu et al. (2002). By sampling the entire state space of small networks ($N\leq 20$), it was found that the “barriers”, which correspond to the attractor basin boundaries, can be crossed with non-vanishing probability when $p>0$, although the characteristic times may be large. This means that the system is always ergodic. This type of noise has also been studied for Boolean networks with threshold functions, corresponding to a majority update rule Huepe and Aldana- González (2002). This system undergoes a second order phase transition at a critical noise strength from an ordered dynamical phase, where all nodes assume the same value for the majority of time, to a disordered phase where nodes assume both states equally often. In this work, we investigate the effect of ongoing stochastic noise on RBNs. Following Miranda and Parga (1989); Golinelli and Derrida (1989); Qu et al. (2002), noise strength is tuned via a probability $p$ that a node does not obey its deterministic update rule. We monitor the transition from fully deterministic dynamics ($p=0$) to purely stochastic dynamics ($p=1/2$) as the noise strength is varied. Differently from Miranda and Parga (1989); Golinelli and Derrida (1989); Qu et al. (2002), we are interested in the behaviour of the networks in the limit of large system size, where it is impossible to explore large parts of the state space. In order to characterize the transition from zero to infinite temperature, we define two order parameters: the long-time average of the Hamming distance between a network with and without noise, and the average frozenness, which is a measure of the extent to which a node prefers one of the two Boolean states. We find, both analytically and numerically, that this transition for the Hamming distance is continuous for $K\leq 2$, and discontinuous at $p=0$ for $K>2$, when the Hamming distance is considered. This distinction is a direct consequence of the phase transition from frozen to chaotic dynamics in the deterministic model. The frozenness shows a smooth transition for all values of $K$. The distribution of frozenness shows a surprising richness in structure, as revealed by computer simulations. For $K\leq 2$ and for $K\gg 1$, we succeeded in reproducing this structure as function of $p$ by analytical considerations. The remainder of this paper is divided into the following parts: In Sec. II we define the RBN model and the type of noise used for our study. In Sec. III, we define the first order parameter, the Hamming distance, and evaluate it numerically and analytically. In Sec. IV, we define the second order parameter, the frozenness, and evaluate it using computer simulations and analytical considerations. Finally, we summarize and discuss our findings in Sec. V. ## II Model A Boolean network is defined as a directed network of $N$ nodes representing Boolean variables $\bm{\sigma}\in\\{1,0\\}^{N}$, which are subject to a dynamical update rule, $\bm{\sigma}(t+1)=\bm{f}\left(\bm{\sigma}(t)\right),$ (1) where $f_{i}$ is a function assigned to node $i$ that depends exclusively on the states of its inputs. We introduce noise into the system through a probability $p$ that a node does not obey its deterministic update rule, $\bm{\sigma}(t+1)=\bm{f}\left(\bm{\sigma}(t)\right)\veebar\bm{n},$ (2) where $\bm{n}$ is a random vector, with elements $n_{i}$ being $1$ with probability $p$ and $0$ otherwise. The symbol $\veebar$ represents the “exclusive or” Boolean operation. Hence, for $p=0$ the deterministic behaviour is recovered, and for $p=1/2$ the dynamics is completely stochastic. RBNs are a special case of Boolean networks, where all possible Boolean functions are assigned randomly to each node with the same probability, and where the nodes are randomly connected. The number of inputs of each node is fixed at a value $K$. The random wiring leads to a Poisson distribution with mean $K$ for the number of outputs. When updated deterministically, RBNs are in the frozen phase for $K=1$. After a transient time, they reach an attractor where all nodes (or all nodes apart from a small number) are permanently frozen in one of the two Boolean states. Networks with larger $K$ have also a _frozen core_ of nodes for $p=0$, and the nodes belonging to it become frozen after a transient time. For $K=2$, all but of the order of $N^{2/3}$ nodes belong to the frozen core. With increasing $K$, the frozen core contains an ever smaller proportion of nodes. For $K=2$, the nonfrozen part of the network consists of several independent components. Each of these components contains a set of _relevant nodes_ , which are connected such that there is at least one feedback loop among them, and “trees” of nonfrozen nodes which are rooted in the relevant nodes and which are slaved to the dynamics of the relevant nodes. ## III Average Hamming distance ### III.1 Definition We use the average in time of the Hamming distance between the states of two copies of a network in order to quantify the effect of noise on the dynamics. Consider a given network in the initial state $\bm{\sigma}(t=0)$, and an exact replica, which is initially in the same state, $\bm{\sigma}^{\prime}(t=0)=\bm{\sigma}(t=0)$. The dynamics of both networks are evolved in parallel, but noise is applied only to $\bm{\sigma}^{\prime}(t)$, as in Eq. (2). The mean Hamming distance $h(t)$ between the two networks is defined as $h(t)=\frac{1}{N}\langle\left\|\sigma_{i}(t)-\sigma_{i}^{\prime}(t)\right\|\rangle\,,$ (3) where $\langle\dots\rangle$ denotes the average over the noise. The long-time average $h$ of the Hamming distance is defined as $h=\lim_{T\to\infty}\frac{1}{T}\frac{1}{N}\sum_{t,i}\left\|\sigma_{i}(t)-\sigma_{i}^{\prime}(t)\right\|.$ (4) If the trajectories become completely uncorrelated after some time, we have $h=1/2$. If the trajectories remain closer in state space, we have $h<1/2$. The case $h>1/2$ does not occur in our model and is therefore not considered in this paper. ### III.2 Annealed approximation We will first evaluate analytically the Hamming distance by using the so- called annealed approximation Derrida and Pomeau (1986). This is a mean field theory, which neglects correlations between nodes and the finite size of the network. The annealed approximation corresponds to the behaviour of a (infinitely large) network where all the edges are randomly rewired at each time step. Within the annealed approximation, the dynamics of a RBN without noise (i.e. for $p=0$) is fully specified by the parameter $\lambda$, which is $K$ times the probability that a node changes its state when one (or more) of its inputs is flipped. For RBNs, we have $\lambda=K/2$, since for any input combination, there is an equal probability that the output of a function will be either $0$ or $1$. When considering two replicas of a network, $\lambda$ is identical to the mean number of nodes that assume a different state in the two networks at time $t=1$ when at time $t=0$ the state of only one node was different. At any time, the Hamming distance between a network with noise and its twin noiseless counterpart, as described by Eq. (3), is simply the fraction of nodes which were changed by noise or by the effect of previously changed nodes. The time evolution of $h(t)$ can then be described as the evolution of the population of flipped nodes. (A node in the replica with noise is called “flipped” it its state deviates from the state it has in the replica without noise.) Let $q(h(t))=1-(1-h(t))^{K}$ denote the probability that a node has at least one flipped input. Then the probability $h(t+1)$ that a node is flipped at time $t+1$ can be written as $\begin{split}h(t+1)&=\frac{\lambda}{K}q(1-p)+\left(1-\frac{\lambda}{K}\right)pq+(1-q)p\\\ &=\frac{\lambda(1-2p)}{K}\left[1-\left(1-h(t)\right)^{K}\right]+p,\\\ \end{split}$ (5) where the first term in the first line corresponds to the proportion of nodes that are flipped by previously flipped nodes (and are not flipped back by noise), and the second and third term are the proportion of nodes that are flipped by noise (with or without inputs being flipped). The fixed point of Eq. (5) determines the order parameter $h$, for given $K$ and $p$. We evaluated this fixed point numerically. Fig. 1 shows $h$ as function of the noise strength $p$ for several values of $K$. The solid lines are the fixed point solutions of Eq. (5), the symbols represent the result of computer simulations of quenched RBNs. The agreement between the annealed approximation and the real networks is very good. The most striking feature of Fig. 1 is the existence of a first-order transition at $p=0$ for $K>2$. This is due to the phase transition to “chaotic” behaviour for $K>2$. In chaotic networks, even the smallest local perturbations have a global effect. ### III.3 The Hamming distance on subsets of nodes We next evaluate separately the Hamming distance for the frozen core and for the nonfrozen part of the network. Fig. 2 shows the long-time Hamming distance, evaluated only for the nodes that belong to the frozen core. These curves can be fitted using the annealed approximation Eq. (5) under the condition that the factor $\lambda/K$ on the right-hand side of Eq. (5) (representing the probability that a node is flipped when at least one input is flipped) is replaced with $\lambda_{\text{eff}}/K$, with $\lambda_{\text{eff}}$ being used as a fit parameter. For $K=1$ and $2$, the frozen core is virtually indistinguishable from the rest of the network, and $\lambda_{\text{eff}}=\lambda$, but for $K>2$, $\lambda_{\text{eff}}$ decreases with increasing $K$. The reason is that the frozen core becomes composed mainly of nodes with constant functions which have $\lambda_{\text{eff}}=0$. Figure 1: (Color online) Average Hamming distance as function of noise strength for RBNs of size $N=10^{4}$ for different values of $K$. Each point was obtained by averaging the results over 3 different network realizations. The solid lines are the respective steady-state solutions of Eq. (5). Before evaluating $h$ for the nonfrozen nodes, let us consider the simplest possible connected set of nonfrozen nodes, which is a simple loop. For $K=1$ and $K=2$, such simple loops of nonfrozen nodes play an important role at determining the attractors with periods larger than 1 Kaufman and Drossel (2005), however, a considerable fraction of $K=2$ networks also have more complex relevant components. The effect of noise on such loops is very different from its effect on the frozen core, since if one of its nodes is flipped, this flip propagates indefinitely around the loop. One can calculate the accumulation of flips on such loops by considering the average Hamming distance at a given time between a loop without and with noise, $\begin{split}h(t)&=\sum_{i=0}^{\lfloor\frac{t-1}{2}\rfloor}{t\choose 2i+1}p^{2i+1}(1-p)^{t-(2i+1)}\\\ &\approx\sum_{i=0}^{\lfloor\frac{t-1}{2}\rfloor}\frac{(tp)^{2i+1}e^{-tp}}{(2i+1)!}\\\ &\approx\frac{1}{2}\left(1-e^{-2tp}\right)\,.\end{split}$ (6) The first equation evaluates the probability that a node has been flipped an odd number of times, and the subsequent transformations are valid for $t\gg 1$. The Hamming distance approaches the value $1/2$ with an exponential decay, and with an characteristic time $\tau=1/(2p)$. Figure 2: (Color online) Average Hamming distance as function of noise for RBN of size $N=10^{4}$ for different values of $K$. Only nodes belonging to the frozen core of the network (without noise) were considered. Each point on each curve was obtained by averaging the results for 3 different network realizations. The solid lines are the respective solutions of Eq. (5), with $\lambda_{\text{eff}}$ being used as a fit parameter. We evaluated how fast a trajectory leaves an attractor in the presence of noise by first letting the system approach an attractor and by then turning on the noise and measuring the Hamming distance $h(l)$ to the initial state after one attractor period $l$. This is identical to the distance from the state of the noiseless replica, which returns to the initial state at time $l$. Fig. 3 shows the values of $h(l)$ for RBNs with different values of $K$. For $K=1$, the data match Eq. (6) very well, since the nonfrozen part of the network in this case can only be composed of simple loops. For $K\geq 2$, the data points are considerably above this exponential curve because a node can become flipped via many different paths. The data are better fitted using Eq. (5), in particular for long periods (i.e. large times). Just as for the case of the frozen core, $\lambda_{\text{eff}}$ was used as a fit parameter. For smaller values of the attractor period, the data are considerably below the fitted line. The reason is that these attractor periods are much smaller than typical attractor periods, and networks with such short attractors are not characteristic of the ensemble, but have a state-space structure with a smaller set of recurrent states. Consequently trajectories diverge less fast than in typical networks. Figure 3: (Color online) Average Hamming distance $h(l)$ of the relevant components of RBNs with different values of $K$, after a full period $l$, with $p=0.01$. The curves were obtained by sampling at least $2\times 10^{4}$ attractors of several distinct RBNs of sizes $N=10^{2}$, $50$ and $25$ for $K\leq 2$, $3$ and $4$, respectively. The solid lines are given by Eq. 6 for $K>1$ and Eq. 5 for $K\geq 2$. ## IV Frozenness ### IV.1 Definition The “frozenness” of a network measures the extent to which the nodes spend more time in one of the two Boolean states. It is zero, when the nodes spend the same time in both states, and it is 1 when the network is frozen. The frozenness of node $i$ is defined by the expression $\Omega_{i}=\left(q_{0}^{(i)}-q_{1}^{(i)}\right)^{2},$ (7) where $q_{\sigma}^{(i)}$ is the proportion of time node $i$ is in state $\sigma$, $q^{(i)}_{\sigma}=\lim_{T\to\infty}\frac{1}{T}\sum_{t=0}^{T}\delta_{\sigma_{i}(t),\sigma}.$ (8) By eliminating one of the two probabilities from Eq. (7), we obtain $\Omega_{i}=\left(2q_{\sigma}^{(i)}-1\right)^{2},$ (9) where $\sigma$ is either $0$ or $1$. The frozenness $\langle\Omega\rangle$ of the network is obtained by averaging over the nodes. Figure 4 shows the frozenness $\langle\Omega\rangle$ as function of the noise strength $p$ obtained by computer simulations of networks of size $10^{4}$, for different $K$. Figure 4: (Color online) Frozenness as function of noise strength for RBNs of size $N=10^{4}$ for different values of $K$. Each curve was obtained by averaging the results for 100 different network realizations. The solid lines correspond to the averages of $\rho_{\Omega}(\Omega\|K,p)$, obtained in Secs. IV.2, IV.3 and IV.4, for $K=1$, $2$ and $\geq 3$, respectively. At $p=0$, the frozenness corresponds obviously to the size of the frozen core, and it decreases towards 0 as the noise strength approaches the value 0.5. In order to derive these curves analytically, the annealed approximation is of no use, since a network that is rewired during the course of time has very small frozenness, which is due uniquely to the constant functions. Therefore, a simple analytical calculation, which does not require the consideration of correlations between nodes, can only be performed for nodes with constant functions: for such nodes the frozenness is given by $\Omega_{i}=(1-2p)^{2}\,.$ (10) In the following, we will present more advanced analytical evaluations and further computer simulations for RBNs with different values of $K$. ### IV.2 $K=1$ For $K=1$ there are only $2^{2^{1}}=4$ possible Boolean functions, two of which are constant ($1$ or $0$), and the remaining ones are the copy ($f(\sigma)=\sigma$) and invert ($f(\sigma)=\lnot\sigma$) functions. As far as the analysis of frozenness is concerned, there are only two distinct functions, constant and non-constant, since the output value is not relevant, but only how often it changes. Since each of the two types of functions occurs equally often in a $K=1$ RBN, we have $\lambda=1/2$, but $K=1$ networks with other values of $\lambda$ can also be constructed. In a network with $K=1$, each node has only one input, and this input node also has one input, etc. In order to evaluate the probability that a node is flipped, one only needs to consider the chain of those nodes that can have an influence on the considered node. Nodes with constant functions present a barrier to the propagation of a perturbation, since they do not respond to a change in their inputs, and therefore the chain ends (or, more precisely, begins) at a node with a constant function. Without loss of generality, we define $q^{(i)}$ as being the proportion of time node $i$ assumes its most frequent value, $q^{(i)}\equiv\max(q^{(i)}_{\sigma},1-q^{(i)}_{\sigma})\;\in\left[1/2,1\right].$ (11) Since the value of $q^{(i)}$ is fully determined by the distance of node $i$ to a node with a constant function, we choose the label $i$ in the remainder of this subsection to signify this distance. If the node itself has a constant function, we have $i=0$, if the node has a non-constant function, but its input has a constant function, we have $i=1$, etc. The value of $q^{(0)}$, i.e. for nodes with constant functions, is simply $q^{(0)}(p)=1-p.$ (12) For larger values of $i$, we have the recursion relation $\begin{split}q^{(i)}&=(1-p)q^{(i-1)}+(1-q^{(i-1)})p\\\ &=\frac{1}{2}(1-2p)^{i+1}+\frac{1}{2},\end{split}$ (13) where the solution of the recursion relation was obtained using Eq. (12). The probability of finding a given $q^{(i)}$ in the network is $p_{q}(q^{(i)}\|p)=(1-\lambda)\lambda^{i}=\left(\frac{1}{2}\right)^{i+1}\,.$ (14) The frozenness of the network is thus given by $\left<\Omega\right>(p)=\sum_{i}p_{q}(q^{(i)}\|p)(1-2q^{(i)})^{2}\,,$ (15) which is plotted in Figure 4 and fits the curve for $K=1$ well. Although networks with $K=1$ have only a discrete set of possible $q$-values, the distribution of $q$ values appears as a continuum when determined by computer simulations. There are two reasons for this. First, the data points for $q$ close to 1/2 (i.e. for $\Omega$ close to 0) are so close to each other that they cannot be resolved, since a computer simulation uses a non-vanishing bin size. Solving Eq. (13) for $i$, inserting the result in (14), and using the relation $\rho_{q}(q\|K=1,p)dq=p_{q}(q\|K=1,p)$ with $dq=q^{(i)}-q^{(i+1)}$, we obtain $\rho_{q}(q\|K=1,p)\propto(2q-1)^{\frac{\ln\lambda}{\ln(1-2p)}-1}\,,$ (16) with $\lambda=1/2$ for RBNs. Thus, in the limit $q\to 1/2$ the distribution of $2q-1$ follows as a power-law with an exponent given by the above expression. The probability density of $\Omega$ also decays as a power law, in the limit $\Omega\to 0$, but with a different exponent, since $\begin{split}\rho_{\Omega}(\Omega\|K=1,p)=&\;\rho_{q}\left(\left.\frac{\sqrt{\Omega}+1}{2}\right\|K=1,p\right)\frac{dq}{d\Omega}\\\ \sim&\;\Omega^{\frac{\ln\lambda}{\ln(1-2p)}-3/2}\,.\end{split}$ (17) Second, a computer simulation averages only over a finite amount of time, $T$, and therefore the measured values $q^{\prime}$ are Gaussian distributed around the exact value $q$, $\begin{split}\tilde{\rho}_{q^{\prime}}(q^{\prime}\|K=1,T,q)\approx\frac{1}{\sqrt{2\pi q/T}}e^{-\frac{(q^{\prime}-q)^{2}}{2q/T}},\end{split}$ (18) This dependence on $T$ can be included in Eq. (14) to obtain the probability density function for $q$ $\begin{split}\rho_{q}(q\|K=1,p,T)=&\,(1-\lambda)\sum_{i=0}^{\infty}\lambda^{i}\tilde{\rho}_{q^{\prime}}(q\|T,q^{(i)}).\end{split}$ (19) For the distribution of $\Omega$ values, we obtain $\rho_{\Omega}(\Omega\|K,p,T)=\rho_{q}\left(\left.\frac{\sqrt{\Omega}+1}{2}\right\|K,p,T\right)\frac{1}{2\sqrt{\Omega}}.$ (20) Fig. 5 shows the distribution of the frozenness for a quenched network with $K=1$, for several values of $p$. It can be seen that there is very good agreement with Eq. (20). The presence of fluctuations significantly deviates some of the distributions from the expected power-law decay. For $p=0.01$, the small values of frozenness, which are not in agreement with the theoretical result, are due to the existence of loops, which are omitted in the analysis above. The probability that a node is part of a nonfrozen loop tends to zero as the network becomes larger, and therefore these points vanish in the limit of infinite system size. Figure 5: (Color online) Distribution of the frozenness for different values of the noise strength for RBNs of size $N=10^{5}$ for $K=1$ and $T=10^{4}$. Each curve was obtained by averaging over 100 different network realizations. The solid lines are given by Eq. (20). ### IV.3 $K=2$ As $K$ becomes larger, the number of possible functions grows very fast as $2^{2^{K}}$, and a detailed analysis of the frozenness, as was done for $K=1$, becomes more complicated. The values of $q$ are still discontinuously distributed, but their number increases fast with $K$, due to the numerous combinations of Boolean functions that can determine the $q$-values of the $K$ inputs of a node and thus, in combination with the node’s Boolean function, the $q$-value of its output. Here we will lay out the basic considerations needed to obtain the distribution of $q$ for all $K>1$, and we will obtain by numerical iteration the distribution for $K=2$. Without loss of generality, we redefine $q$ as $q\equiv q_{\sigma=1}$, i.e., the fraction of time a given node has the value 1 (as opposed to Eq. (11), which simplified the case $K=1$). In general, the probability density function $\rho_{q}(q\|K=1,p)$ needs to account for all possible recursive combinations of output functions and their inputs. We can thus write the following self-consistent expression, $\rho_{q}(q\|K,p)=\int_{0}^{1}\dotsi\int_{0}^{1}\sum_{f}p_{f}\delta(q^{(f)}-q)\prod_{i=1}^{K}\rho_{q}(q^{(i)}\|K,p)dq^{(i)},$ (21) where the sum is taken over all Boolean functions; $p_{f}$ is the probability of the $f$th Boolean function ($p_{f}=2^{-2^{k}}$ ), and $q^{(f)}=(1-2p)q^{(f)}(\\{q^{(i)}\\})+p,$ (22) where $q^{(f)}(\\{q^{(i)}\\})$ is the value of $q$ for a specific function $f$, given the values $\\{q^{(i)}\\}$ of its inputs, for $i=1,\dots,K$. Since Eq. (21) involves an expression $q^{(f)}(\\{q^{(i)}\\})$ for all Boolean functions, a general closed solution becomes unfeasible. However, for $K=2$ Eq. (21) can at least be solved numerically, since there are only $16$ possible functions, given in Table 1. Eq. (21) is then solved by iteration, until convergence to a self-consistent $q$ distribution is obtained. We started with the initial distribution $\rho^{0}_{q}(q\|K,p)=\frac{1}{2}\delta(q-p)+\frac{1}{2}\delta(q-(1-p))\,.$ (23) In the end, we determined the final distribution $\rho_{\Omega}(\Omega\|K=2,p)$ by using Eq. (20). $f$ \| $f_{i}(\sigma_{1},\sigma_{2})$ \| $q^{(f)}(q_{1},q_{2})$ ---\|---\|--- $0$ \| $0$ \| $0$ $1$ \| $\sigma_{1}\land\sigma_{2}$ \| $q_{1}q_{2}$ $2$ \| $\sigma_{1}\land\lnot\sigma_{2}$ \| $q_{1}(1-q_{2})$ $3$ \| $\sigma_{1}$ \| $q_{1}$ $4$ \| $\lnot\sigma_{1}\land\sigma_{2}$ \| $(1-q_{1})q_{2}$ $5$ \| $\sigma_{2}$ \| $q_{2}$ $6$ \| $\sigma_{1}\veebar\sigma_{2}$ \| $q_{1}+q_{2}-2q_{1}q_{2}$ $7$ \| $\sigma_{1}\lor\sigma_{2}$ \| $q_{1}+q_{2}-q_{1}q_{2}$ $8$ \| $\lnot(\sigma_{1}\lor\sigma_{2})$ \| $1-q^{(7)}(q_{1},q_{2})$ $9$ \| $\lnot(\sigma_{1}\veebar\sigma_{2})$ \| $1-q^{(6)}(q_{1},q_{2})$ $10$ \| $\lnot\sigma_{2}$ \| $1-q_{2}$ $11$ \| $\sigma_{1}\lor\lnot\sigma_{2}$ \| $q^{(7)}(q_{1},1-q_{2})$ $12$ \| $\lnot\sigma_{1}$ \| $1-q_{1}$ $13$ \| $\lnot\sigma_{1}\lor\sigma_{2}$ \| $q^{(7)}(1-q_{1},q_{2})$ $14$ \| $\lnot(\sigma_{1}\land\sigma_{2})$ \| $1-q_{1}q_{2}$ $15$ \| $1$ \| $1$ Table 1: Expressions of $q^{(f)}(q_{1},q_{2})$ for all Boolean functions for $K=2$. The Boolean expressions of each function is also given for reference. Fig. 6 shows the distribution of $\Omega$ for simulated quenched RBNs with $K=2$ for different values of $p$, compared with the result of the numerical evaluation of Eq. (21) as described above. There is a very good agreement between the two types of results. The peaks correspond to prominent values of the frozenness. The rightmost peak is always due to the constant functions, but large frozenness values are also obtained for other functions. For instance, $f_{1}$ assumes the value 0 whenever both inputs are different. If both inputs have $q=1/2$ (i.e., $\Omega=0$), the value of $q_{f=1}$ is $(1-2p)1/4+p$ and $\Omega=((1-2p)/2+2p-1)^{2}$. This is the second main peak of $\rho_{\Omega}(\Omega\|2,0.4,10^{4})$ (counted from the right end). For smaller values of $p$, the peaks are not discernible, and a broad continuum appears, with a distribution that follows a power-law with an exponent $\simeq 0.6$ as $\Omega\to 0$. As $T$ becomes larger, it is expected that the continuous regions become more and more discontinuous, as can be seen in Fig. 7, which shows the theoretical prediction for larger times $T$. Moreover, when the resolution is increased (see inset), it can be seen that peak-like regions which appear like fluctuations around a single value of $\Omega$, are in fact composed of sharper peaks, which themselves are composed of other peaks, building a fractal structure. Another distinguishing feature seen in Fig. 6 is a sharp transition at $p=0$, where the only two possible values of $q$ are $0$ and $1$, both of which amount to $\Omega=1$, leading to the variance $\sigma_{\Omega}^{2}=0$. For $p>0$ this abruptly changes, and a wide range of values of $q$ are possible, which discontinuously leads to $\sigma_{\Omega}^{2}>0$. There is no such discontinuous transition for other $K$ values, but a continuous one (see following section), which makes the case $K=2$ special. Figure 6: (Color online) Distribution of the frozenness of nodes for different noise strength for RBNs of size $N=10^{5}$ and for $K=2$ and $T=10^{4}$. Each curve was obtained by averaging over 100 different network realizations. The solid lines show the values of $\rho_{\Omega}(\Omega\|K=2,p,10^{4})$ according to Eq. (20). The line segment corresponds to a power law with an exponent $0.6$. Figure 7: (Color online) Expected distribution of frozenness of each node as function of noise for RBNs with $K=2$ and different values of $T$. The curves represent Eq. (20). The inset shows a zoom into $\rho_{\Omega}(\Omega\|K=2,0.4,T)$ for different values of $T$. The function $\langle\Omega\rangle$ is obtained by performing the integral $\int\Omega\cdot\rho_{\Omega}(\Omega\|K=2,p,T)d\Omega$. As can be seen in Figure 4, our calculation of this function agrees well with the results of computer simulations. ### IV.4 $K>2$ For larger values of $K$, numerical solutions of Eq. (21) become progressively more elaborate. We did not pursue the task of writing the expressions of 256 function $q_{f}(\\{q^{(i)}\\})$ for $K=3$ or of $65,536$ functions for $K=4$. Instead, we perform in the following an approximation that is good for a large number of inputs per node. When $K$ is large, the vast majority of Boolean functions have the output $1$ for approximately half the input combinations. This means that almost all nodes have at their inputs $q$ values close to $1/2$. We therefore make the assumption that the input values to each node are $1$ and $0$ with probability $1/2$, independently from each other. This means that for any given function all input combinations are equally probable. It then follows immediately that the possible $q$ values are identical to the possible fractions of output values 1 in the truth table of a Boolean function, and that the probability for a given $q$ value is $p_{q}(q\|K,p=0)=2^{-M}{M\choose qM}\,.$ (24) Here, we have defined $M=2^{K}$, and the possible $q$ values are thus multiples of $1/M$. In the presence of noise, each output value is inverted with probability $p$, implying that $q$ is changed to $q^{\prime}=q(1-p)+p(1-q)=(1-2p)q+p$. We therefore have $\rho_{q}(q\|K,p)=\frac{M}{1-2p}p_{q}\left(\left.\frac{q-p}{1-2p}\right\|K,p=0\right)\,.$ (25) For the frozenness $\Omega$ we obtain the distribution $\rho_{\Omega}(\Omega\|K,p)=\frac{M}{2(1-2p)\sqrt{\Omega}}p_{q}\left(\left.\frac{\sqrt{\Omega}+1-p}{2(1-2p)}\right\|K,p\right)\,.$ (26) Finally, one needs to take into account the effect of fluctuations, exactly as was done for the previous cases, $\rho_{q}(q\|p,K,T)=\int_{0}^{1}\rho_{q}(q^{\prime}\|p,K)\tilde{\rho}_{q}(q\|T,q^{\prime})dq^{\prime},$ (27) where $\tilde{\rho}_{q}(q\|T,q^{\prime})$ is given by Eq. (18). Fig. 8 shows the distributions of frozenness for $K=4$ and $5$. In contrast to the cases $K=1$ and $K=2$, the peaks are less pronounced, and are hardly visible. For $K=3$ (not shown) there are some peaks which are still visible, specially for high values of $p$. Therefore, the high-$K$ approximation Eq (27) is very good already for $K=4$. Both distributions show the same power- law decay $\rho_{\Omega}(\Omega\|K,p)\sim\Omega^{-1/2}$. This is simply due to the fact that for $\Omega\to 0$ ($q\to 1/2$) the shape of $\rho_{q}(q\|K,p)$ is essentially flat, and thus $\rho_{\Omega}(\Omega\|K,p)\sim dq/d\Omega=\Omega^{-1/2}/2$. Figure 8: (Color online) Distribution of the frozenness of nodes as function of noise strength for RBN of size $N=10^{4}$ for different values of $K$. Each curve was obtained by averaging the results for 100 different network realizations. The solid lines correspond to Eq. 27. The quality of our approximation can also be assessed by comparing the analytical prediction for $\left<\Omega\right>$ with the computer simulations (Fig. 4). We expect that the approximation becomes even better for larger $K$. ## V Conclusion We have investigated the effect of thermal noise on RBNs by evaluating two order parameters, the long time average of the Hamming distance between two networks and the average frozenness of the network. While for $K=1$ and $K=2$ the average Hamming distance increases continuously from 0 to 1/2 as $p$ increases from 0 to 1/2, it has a jump at $p=0$ for $K>2$. These findings are well reproduced by the annealed approximation, and they are a consequence of the transition from a frozen to a “chaotic” phase in the deterministic system. In the chaotic phase (occuring for $K>2$), initially nearby trajectories become eventually uncorrelated. The smooth increase of the Hamming distance towards the value 1/2 is compatible with what was found in Miranda and Parga (1989); Qu et al. (2002) for small networks. The analysis of the average frozenness of the network required more sophisticated calculations than the annealed approximation, and revealed intricate details of the network dynamics. For all values of $K$ the probability distribution of frozenness is a set of delta peaks. For $K=1$, these peaks can be obtained by considering the distance of nodes to nodes with constant functions. For $K>1$, the analysis becomes a lot more elaborate, due to the large number of Boolean functions and the resulting vast number of possible combinations of frozenness values for the inputs of each function. We explained the general method, and performed the actual numerical evaluation for the case $K=2$. The delta peaks show a fractal structure, which emerges from the iterated recursion relation for the possible frozenness values. The variance of the frozenness distribution changes continuously with $p$ for all $K\neq 2$, but for $K=2$ it has a jump at $p=0$, where the variance changes discontinuously from $0$ to a value larger than zero. For larger values of $K$, the delta peaks are so close to each other that the frozenness distribution appears continuous, and in this limit we succeeded in performing an approximate analytical calculation. We do not find a phase transition at finite noise strength, in contrast to Huepe and Aldana-González (2002), where Boolean networks with threshold functions following a majority rule were used. Such a system undergoes a second order phase transition from an ordered “ferromagnetic” phase, where all nodes assume the same value for the majority of time, to a disordered phase, where the nodes assume both states equally often. The presence of an ordered phase is a direct consequence of the majority rule, and this transition is similar to that in a network of Ising spins Indekeu (2004); Aleksiejuk et al. (2002). The order parameter in Huepe and Aldana-González (2002) was defined as the average “alignment” $s=\left\|\left<1-2\sigma\right>\right\|$, which is $1$ if all nodes are in the same state. The order parameter $s$ is only meaningful in systems where the system is ordered in the absence of noise, and where the symmetry between the states with values $0$ and $1$ is broken, as in ferromagnetic spin systems. Otherwise, $\left<\Omega\right>$ is a better order parameter, because it captures disordered frozen phases, such as for $K=1$ in RBNs. Of course, a phase transition in the value of $s$ is always accompanied by a phase transition in the value of $\left<\Omega\right>$. The opposite is not always true. It is to be expected that real networks show some kind of robustness to noise, since they must be able to carry out their function in a noisy environment. As the results of this work show, only for RBNs with $K=1$ do the order parameters change slowly as noise is switched on. RBNs with $K>1$ fail to exhibit robustness to noise, which is hardly surprising given the random wiring of the system and the random choice of functions. It will therefore be interesting to extend the present study to networks with a more restricted set of functions with more biological relevance, such as threshold Szejka et al. (2008) or canalizing functions Moreira and Amaral (2005); Kauffman et al. (2004). At least for some sets of functions, one should expect a phase transition at a finite noise strength, similar to the transition seen in Huepe and Aldana-González (2002). The survival of the “ordered” phase up to a certain noise strength can be viewed as a certain type of robustness. It remains to be seen how other network topologies Iguchi et al. (2007) and the incorporation of redundancy Gershenson et al. (2006) change a network’s response to noise. In Iguchi et al. (2007), it was shown that a scale-free input distribution changes the average number and length of attractors. In Gershenson et al. (2006), redundancy was introduced as functional duplications of nodes in the network, which resulted in greater robustness against random mutations of the update functions. In both papers, only deterministic dynamics were considered. The effects of these (or other more general) topological and functional characteristics may strongly alter the response of a network to thermal noise. Finding the general conditions required for reliable dynamics in a stochastic environment will be an important step towards a deeper understanding of the dynamical features of real networks. We acknowledge the support of this work by the Humboldt Foundation. ## References * Kauffman (1969) S. A. Kauffman, J. Theor. Biol. 22, 437 (1969). * Drossel (2008) B. Drossel, _Reviews of Nonlinear Dynamics and Complexity_ (Wiley, 2008), vol. 1, ISBN 3527407294. * Rosen-Zvi et al. (2001) M. Rosen-Zvi, A. Engel, and I. Kanter, Phys. Rev. Lett. 87, 078101 (2001). * Moreira et al. (2004) A. A. Moreira, A. Mathur, D. Diermeier, and L. A. N. Amaral, Proc. Nat. Ac. Sci. 101, 12085 (2004). * Lagomarsino et al. (2005) M. C. Lagomarsino, P. Jona, and B. Bassetti, Phys. Rev. Lett. 95, 158701 (2005). * Bornholdt (2005) S. Bornholdt, Science 310, 449 (2005). * Greil and Drossel (2005) F. Greil and B. Drossel, Phys. Rev. Lett. 95, 048701 (2005). * Klemm and Bornholdt (2005) K. Klemm and S. Bornholdt, Phys. Rev. E 72, 055101 (2005). * Samuelsson and Troein (2003) B. Samuelsson and C. Troein, Phys. Rev. Lett. 90, 098701 (2003). * Drossel (2005) B. Drossel, Phys. Rev. E 72, 016110 (2005). * Drossel et al. (2005) B. Drossel, T. Mihaljev, and F. Greil, Phys. Rev. Lett. 94, 088701 (2005). * Shmulevich et al. (2002) I. Shmulevich, E. R. Dougherty, S. Kim, and W. Zhang, Bioinformatics 18, 261 (2002). * Indekeu (2004) J. O. Indekeu, Physica A 333, 461 (2004). * Aleksiejuk et al. (2002) A. Aleksiejuk, J. A. Holyst, and D. Stauffer, Physica A 310, 260 (2002). * McAdams and Arkin (1997) H. H. McAdams and A. Arkin, Proc. Nat. Ac. Sci. 94, 814 (1997). * Fretter and Drossel (2008) C. Fretter and B. Drossel, Eur. Phys. J. B 62, 365 (2008). * Miranda and Parga (1989) E. N. Miranda and N. Parga, Europhys. Lett. 10, 293 (1989). * Golinelli and Derrida (1989) O. Golinelli and B. Derrida, J. Phys 50, 1587 (1989). * Qu et al. (2002) X. Qu, M. Aldana, and L. P. Kadanoff, J. Stat. Phys. 109, 967 (2002). * Huepe and Aldana-González (2002) Huepe and Aldana-González, J. Stat. Phys. 108, 527 (2002). * Derrida and Pomeau (1986) B. Derrida and Y. Pomeau, Europhys. Lett. 1, 45 (1986), ISSN 0295-5075. * Kaufman and Drossel (2005) Kaufman and Drossel, Eur. Phys. J. B 43, 115 (2005). * Szejka et al. (2008) A. Szejka, T. Mihaljev, and B. Drossel, New J. Phys. 10, 063009 (2008). * Moreira and Amaral (2005) A. A. Moreira and L. A. N. Amaral, Phys. Rev. Lett. 94, 218702 (2005). * Kauffman et al. (2004) S. Kauffman, C. Peterson, B. Samuelsson, and C. Troein, Proc. Nat. Ac. Sci. 101, 17102 (2004). * Iguchi et al. (2007) K. Iguchi, S. ichi Kinoshita, and H. S. Yamada, J. Theor. Biol. 247, 138 (2007). * Gershenson et al. (2006) C. Gershenson, S. A. Kauffman, and I. Shmulevich, _Artificial Life X: Proceedings of the Tenth International Conference on the Simulation and Synthesis of Living Systems_ (The MIT Press, 2006), ISBN 0262681625\.	arxiv-papers	2008-08-22T15:35:51	2024-09-04T02:48:57.399043	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Tiago P. Peixoto, Barbara Drossel", "submitter": "Tiago Peixoto", "url": "https://arxiv.org/abs/0808.3087" }
0808.3155	# On independent sets in purely atomic probability spaces with geometric distribution Eugen J. Ionascu and Alin A. Stancu Department of Mathematics, Columbus State University Columbus, GA 31907, USA ionascu_eugen@colstate.edu, stancu_alin1@colstate.edu (August 12th, 2008) ###### Abstract We are interested in constructing concrete independent events in purely atomic probability spaces with geometric distribution. Among other facts we prove that there are uncountable many sequences of independent events. †† Mathematics Subject Classification: Primary 60A10Key Words: Independence, purely atomic probability, geometric distribution ## 1 Introduction Let us assume a fixed ratio $r$ is given, $r\in(0,1)$. In what follows we will work with the discrete probability space $\mathbb{N}_{0}=\\{0,1,2,3,...\\}$ and the usual geometric probability on $\cal A$ (all subsets of $\mathbb{N}_{0}$) defined by $P_{r}(E):=\frac{1-r}{r}\sum_{k\in E\setminus\\{0\\}}r^{k}\ for\ every\ set\ E\in\cal A.$ We are interested to study the class of independent sets in this probability space. We are going to follow [2] and define: $A,B\in\Omega$ are called independent if $P(A\cap B)=P(A)P(B)$. With this definition for every $E\subset\Omega$, $\Omega$ and $E$ are independent and $\emptyset$ and $E$ are also independent. These are clearly trivial examples. Three or more subsets of $\Omega$, $A_{1}$, …, $A_{n}$ are called mutually independent or simply independent if for every choice of $k$ ($n\geq k\geq 2$) such sets, say $A_{i_{1}}$,…,$A_{i_{k}}$, we have $P(\bigcap_{j=1}^{k}A_{i_{j}})=\prod_{j=1}^{k}P(A_{i_{j}}).$ (1) So, for $n$ ($n\geq 2$) independent sets one needs to have $2^{n}-n-1$ relations as in (1) to be satisfied. An infinite family of subsets is called independent if each finite collection of these subsets is independent. Events are called trivial if their probability is 0 or 1. If $n\in\mathbb{N}$ then $\Omega(n)$ usually denotes the number of primes dividing $n$ counting their multiplicities (see [8]). In [1] and [6], independent families of events have been studied for finite probability spaces with uniform distribution. Eisenberg and Ghosh [6] show that the number of nontrivial independent events in such spaces cannot be more than $\Omega(m)$ where $m$ is the cardinality of the space. This result should be seen in view of the known fact (see Problem 50, Section 4.1 in [7]) that if $A_{1}$, $A_{2}$, …., $A_{n}$ are independent non-trivial events of a sample space $X$ then $\|X\|\geq 2^{n}$. One can observe that in general $\Omega(m)$ is considerably smaller than $\log_{2}m$. It is worth mentioning that according to [5] the first paper to deal to this problem in uniform finite probability spaces is [9]. In their paper, Shiflett and Shultz [9] raise the question of the existence of spaces with no non-trivial independent pairs, called dependent probability spaces. A space containing non-trivial independent events is called independent. For uniform distributed probability spaces $X$, as a result of the work in [6] and [1], $X$ is dependent if $\|X\|$ a prime number and independent if $\|X\|$ is composite. For denumerable sets $X$ one can see the construction given in [5] or look at the Example 1.1 in [10]. For our spaces, the Example 1.1 does not apply and in fact, we will construct explicitly lots of independent sets. For every $n\in\mathbb{N}$ one can consider the following space of geometric probability distribution, denoted here by ${\cal G}_{n}:=([n],{\cal P}([n]),P)$ where $P(k)=q^{k}$ with $k\in[n]:=\\{1,2,3,...,n\\}$ and of course $q$ is the positive solution of the equation $\sum_{k=1}^{n}q^{k}=1.$ This space is independent for every $n\geq 4$ with $n$ composite. Indeed, if $n=st$ with $s,t\in\mathbb{N}$ $s,t\geq 2$ on can check that the sets $A:=\\{1,2,3,...,s\\}$, $B:=\\{1,s+1,2s+1...,(t-1)s+1\\}$ represent non- trivial independent events. To match the uniform distribution situation, it would be interesting if ${\cal G}_{n}$ was a dependent space for every $n$ prime. The class of independent sets is important in probability theory for various reasons. Philosophically speaking, the concept of independence is at the heart of the axiomatic system of modern probability theory introduced by A. N. Kolmogorov in 1933. More recently, it was shown in [3] that two probability measures on the same space which have the same independent (pairs of) events must be equal if at least one of them is atomless. This was in fact a result of A. P. Yurachkivsky from 1989 as the same authors of [3] point out in the addendum to their paper that appeared in [4]. On the other hand, Szekely and Mori [10] show that if the probability space is atomic then there may be no independent sets or one may have a sequence of such sets. The following result that appeared in [10] is a sufficient condition for the existence of a sequence of independent events in the probability space. ###### Theorem 1. If the range of a purely atomic probability measure contains and interval of the form $[0,\epsilon)$ for some $\epsilon>0$ then there are infinitely many independent sets in the underlying probability space. Let us observe that, if $r=1/2$ the probability space $(\mathbb{N}_{0},{\cal A},P_{1/2})$ does satisfy the hypothesis of the above theorem with $\epsilon=1$ because every number in $[0,1]$ has a representation in base $2$. On the other hand if, let us say $r=1/3$, then the range of $P_{1/3}$ is the usual Cantor set which has Lebesgue measure zero, so Theorem 1 does not apply to $(\mathbb{N}_{0},{\cal A},P_{1/3})$. However, we will show that there are uncountably many pairs of sets that are independent in $(\mathbb{N}_{0},{\cal A},P_{r})$ for every $0<r<1$ (these sets do not depend of $r$). ## 2 Independent pairs of events for denumerable spaces The first result we would like to include is in fact a characterization, under some restrictions of $r$, of all pairs of independent events $(A,B)$, in which one of them, say $B$, is fixed and of a certain form. This will show in particular that there are uncountably many such pairs. In order to state this theorem we need to start with a preliminary ingredient. ###### Lemma 1. For $m\geq 1$, consider the function given by $f(x)=(2x-1)(1+x^{m})-x^{m}\ for\ all\ x\in[0,1].$ The function $f$ is strictly increasing and it has unique zero in $[0,1]$ denoted by $t_{m}$. Moreover, for all $m$ we have $t_{m}>1/2$, the sequence $\\{t_{m}\\}$ is decreasing and $\displaystyle\lim_{m\to\infty}t_{m}=\frac{1}{2}.$ Having $t_{m}$ defined as above we can state our first theorem. ###### Theorem 2. For every natural number $n\geq 2$, we define the events $E:=\\{0,n-1\\}$ and $\begin{array}[]{l}B:=\\{\underset{n-1}{\underbrace{1,2,...,n-1}},\underset{n-1}{\underbrace{2n-1,2n,...,3n-3}},\\\ \\\ \underset{n-1}{\underbrace{4n-3,4n-2,...,5n-5},...}\\}.\end{array}$ (2) Also, for $T\subset B$ an arbitrary nonempty subset we set $A:=E+T$ with the usual definition of addition of two sets in a semigroup. Then $A$ and $B$ are independent events in $(\mathbb{N}_{0},{\cal A},P_{r})$. Conversely, if $r<t_{m}$ (where $m=n-1$ and $t_{m}$ as in Lemma 1), $B$ is given as in (2) and $A$ forms an independent pair with $B$, then $A$ must be of the above form, i.e. $A=E+T$ for some $T\subset B$. Proof of Lemma 1. The function $f$ has derivative $f^{\prime}(x)=2(1+x^{m})-2m(1-x)x^{m-1}$, $x\in(0,1]$. For $m\geq 2$, using the Geometric-Arithmetic Mean inequality we have $(m-1)(1-x)x^{m-1}\leq\left[\frac{(m-1)(1-x)+\underset{m-1}{\underbrace{x+x+...+x}}}{m}\right]^{m}=\left(\frac{m-1}{m}\right)^{m}$ and so $m(1-x)x^{m-1}\leq(\frac{m-1}{m})^{m-1}\leq 1$ which implies $m(1-x)x^{m-1}\leq 1$. This last inequality is true for $m=1$ too. This implies that $f^{\prime}(x)=2(1+x^{m})-2m(1-x)x^{m-1}\geq 2x^{m}>0$ for all $x\in(0,1]$. Therefore the function $f$ is strictly increasing and because $f(1/2)=-\frac{1}{2^{m}}<0$ and $f(1)=1>0$, by the Intermediate Values Theorem there must be an unique solution $x=t_{m}$, of the equation $f(x)=0$ in the interval $(1/2,1)$. Because $f(t_{m-1})=\left(\frac{1-t_{m-1}}{1+t_{m-1}^{m-1}}\right)t_{m-1}^{m-1}>0$ we see that $t_{m}<t_{m-1}$ for all $m\geq 2$. Since $(2t_{m}-1)(1+t_{m}^{m})=t_{m}^{m}$ we can let $m$ go to infinity in this equality and obtain $t_{m}\to 1/2$.$\hfill\blacksquare$ Using Maple, we got some numerical values for the sequence $t_{m}$: $t_{1}=\frac{1}{\sqrt{2}}\approx 0.707$, $t_{2}\approx 0.648$, $t_{3}\approx 0.583$, $t_{4}\approx 0.539$ and for instance $t_{10}\approx 0.5005$. Proof of Theorem 2. First let us check that $E_{1}=E+1=\\{1,n\\}$ and $B$ are independent. Since $E_{1}\cap B=\\{1\\}$, $P(\\{1\\})=\frac{1-r}{r}r=1-r$ and $P_{r}(E_{1})=\frac{1-r}{r}(r+r^{n})=(1-r)(1+r^{n-1})$, we have to show that $P_{r}(B)=\frac{1}{1+r^{n-1}}$. We have $P_{r}(B)=\frac{1-r}{r}\left(\sum_{j=1}^{m}r^{j}\right)\left(\sum_{i=0}^{\infty}r^{2mi}\right)=\frac{r-r^{m+1}}{r}\frac{1}{1-r^{2m}}=\frac{1}{1+r^{m}}$ which is what we needed. Now, suppose $b\in B$ and consider $E_{b}=E+b=\\{b,b+n-1\\}$. We notice that by the definition of $B$, the intersection $B\cap E_{b}$ is $\\{b\\}$. Hence, $P_{r}(B\cap E_{b})=\frac{1-r}{r}r^{b}=(1-r)r^{c}$ (with $c=b-1$) and $P_{r}(B)P_{r}(E_{b})=\frac{1}{1+r^{m}}\frac{1-r}{r}\left(r^{b}+r^{b+m}\right)=(1-r)r^{c}.$ Hence, $B$ and $E_{b}$ are independent for every $b\in B$. > Next we would like to observe that if $(F_{1},B)$ and $(F_{2},B)$ are > independent pairs of events and $F_{1}\cap F_{2}=\emptyset$, then $F_{1}\cup > F_{2}$ and $B$ are independent events as well. Indeed, by the given assumption we can write $\begin{array}[]{l}P_{r}(B\cap(F_{1}\cup F_{2}))=P_{r}((B\cap F_{1})\cup(B\cap F_{2}))=P_{r}(B\cap F_{1})+P_{r}(B\cap F_{2})=\\\ \\\ P_{r}(B)P_{r}(F_{1})+P_{r}(B)P_{r}(F_{2})=P_{r}(B)(P_{r}(F_{1})+P_{r}(F_{2}))=P_{r}(B)P_{r}(F_{1}\cup F_{2}).\end{array}$ In fact, the above statement can be generalized to a sequence of sets $F_{k}$ which are pairwise disjoint, due to the fact that $P_{r}$ is a genuine finite measure and so it is continuous (from below and above). Then if $T\subset B$ is nonempty, $A=E+T=\bigcup_{b\in B}E_{b}$ is countable union and since $E_{b}\cap E_{b^{\prime}}=\emptyset$ for all $b,b^{\prime}\in B$ ($b\not=b^{\prime}$) the above observation can be applied to $\\{E_{b}\\}_{b\in T}$. So, we get that $B$ and $A$ are independent. For the converse, we need the following lemma. ###### Lemma 2. If $L\subset\mathbb{N}_{0}\setminus B$ and the smallest element of $L$ is $s=(2i-1)m+j$, where $i,j\in\mathbb{N}$, $j\leq m$, then $P_{r}(L)\leq r^{s-1}-\frac{r^{2im}}{1+r^{m}}.$ Proof of Lemma 2 Indeed, we have $\begin{array}[]{c}P_{r}(L)\leq\frac{1-r}{r}[(r^{s}+r^{s+1}+...+r^{2im})+(r^{(2i+1)m+1}+....)]=\\\ \\\ r^{s-1}-r^{2im}+r^{2im}P_{r}(\Omega\setminus B)=r^{s-1}-r^{2im}+r^{2im}(1-\frac{1}{1+r^{m}})=r^{s-1}-\frac{r^{2im}}{1+r^{m}}.\end{array}$ $\hfill\blacksquare$ So, let us assume that $r<t_{m}$, $B$ is as in (2) and $A$ is independent of $B$. We let $T$ be the intersection of $A$ and $B$ and we put $\alpha:=P_{r}(T)/P_{r}(B)$. Also, define $A^{\prime}:=T+\\{0,n-1\\}$, $L=A\setminus A^{\prime}$ and $L^{\prime}=A^{\prime}\setminus A$. We have clearly $L,L^{\prime}\subset\Omega\setminus B$. By the first part of our theorem $P_{r}(A^{\prime})=\alpha$. Because $A$ and $B$ are independent $P_{r}(A)$ must be equal to $\alpha$ as well. Hence $P_{r}(A)=P_{r}(A^{\prime})$ which attracts $\sum_{k\in L^{\prime}}r^{k}=\sum_{k\in L}r^{k}\Leftrightarrow\sum_{k\in L\cup L^{\prime}}r^{k}=2\sum_{k\in L^{\prime}}r^{k}.$ (3) From (3), it is clear that $L^{\prime}=\emptyset$ if an only if $L=\emptyset$ and so if $L^{\prime}$ is empty then $A=A^{\prime}$, which is what we need in order to conclude our proof. By way of contradiction, suppose $L^{\prime}\not=\emptyset$ (or equivalently $L\not=\emptyset$) we can assume without loss of generality that $L^{\prime}$ contains the smallest number of $L^{\prime}\cup L$, say $s$ which is written as in Lemma 2. Thus from equality (3) we have $P_{r}(L\cup L^{\prime})\geq 2P_{r}(L^{\prime})$ and then by Lemma 2 we get $r^{s-1}-\frac{r^{2im}}{1+r^{m}}\geq 2(1-r)r^{s-1}\Leftrightarrow 2r\geq 1+\frac{r^{2im+1-s}}{1+r^{m}}\Leftrightarrow 2r\geq 1+\frac{r^{n-j}}{1+r^{m}}.$ Therefore for every $n$ and $1\leq j\leq m$, $2r\geq 1+\frac{r^{n-j}}{1+r^{m}}\geq 1+\frac{r^{m}}{1+r^{m}}\Rightarrow f(r)=(2r-1)(1+r^{m})-r^{m}\geq 0.$ By Lemma 1 we see that $r\geq t_{m}$ which is a contradiction. It remains that $L$ and $L^{\prime}$ must be empty and so $A=A^{\prime}$. $\hfill\blacksquare$ In the previous theorem, since $T$ was an arbitrary subset of an infinite set we obtain an uncountable family of pairs of independent sets. Remark 1: If $r=\sqrt{\frac{1}{\phi}}$ where $\phi$ stands for the classical notation of the golden ratio (i.e. $\phi=\frac{\sqrt{5}+1}{2}$), $n=2$, $B=\\{1,3,5,7,...\\}$ as in (2), and $A=\\{1,4,6\\}$, then one can check that $P_{r}(B)=\frac{1}{1+r}$, $P_{r}(A\cap B)=1-r$, $P_{r}(A)=(1-r)(1+r^{3}+r^{5})$. So the equality $P_{r}(A\cap B)=P_{r}(A)P_{r}(B)$ is equivalent to $1+r=1+r^{3}+r^{5}$ which is the same as $r^{4}+r^{2}-1=0$. One can easily see that this last equation is satisfied by $r=\sqrt{\frac{1}{\phi}}$. Hence $A$ and $B$ are independent but clearly $A$ is not a translation of $\\{0,1\\}$ with a subset of $B$. Therefore the converse part in Theorem 1 cannot be extended to numbers $r\geq t_{m}$ such as $r=\sqrt{\frac{1}{\phi}}$. In fact, we believe that the constants $t_{m}$ are sharp, in the sense that for all $r>t_{m}$ the converse part is false, but an argument for showing this is beyond the scope of this paper. Remark 2: Another family of independent events which seems to have no connection with the ones constructed so far is given by $A=\\{1,2,3,4,...,n-1,n\\}$ and $B=\\{n,2n,3n,...\\}$, with $n\in\mathbb{N}$. A natural question arises as a result of this wealth of independent events: can one characterize all pairs $(A,B)$ which are independent regardless the value of the parameter $r$? ## 3 Three independent events The next theorem deals with the situation in which two sets as in the construction of Theorem 2 form with $B$ given by (2), a triple of independent sets. Let us observe that if $A_{1}$, $A_{2}$, and $B$ are mutually independent then by Theorem 2 (at least if $r\in(0,t_{m})$), $A_{1}$ and $A_{2}$ must be given by $A_{i}=T_{i}+E$ with $T_{i}\subset B$, $i=1,2$. Therefore $A_{1}\cap A_{2}=(T_{1}\cap T_{2})+E$. Also, we note that $P_{r}(A_{i})=P_{r}(T_{i})(1+r^{n-1})$, $i=1,2$, and $P_{r}(A_{1}\cap A_{2})=P_{r}(T_{1}\cap T_{2})(1+r^{n-1})$. This means that the equality $P_{r}(A_{1}\cap A_{2})=P_{r}(A_{1})P(A_{2})$ is equivalent to $P_{r}(T_{1}\cap T_{2})=P_{r}(T_{1})P_{r}(T_{2})(1+r^{n-1}).$ (4) On the other hand the condition $P_{r}(A_{1}\cap A_{2}\cap B)=P_{r}(A_{1})P_{r}(A_{2})P(B)$ reduces to $P_{r}(T_{1}\cap T_{2})=P_{r}(T_{1})P_{r}(T_{2})(1+r^{n-1})^{2}P_{r}(B),$ which is the same as (4). So, three sets $A_{1}$, $A_{2}$ and $B$ are independent if and only if (4) is satisfied. Let us notice that the condition (4) may be interpreted as a conditional probability independence relation: $P_{r}(T_{1}\cap T_{2}\|B)=P_{r}(T_{1}\|B)P_{r}(T_{2}\|B).$ (5) At this point the construction we have in Theorem 2 can be repeated. As a result, regardless of what $r$ is, we obtain an uncountable family of there events which are mutually independent in $(\mathbb{N}_{0},{\cal A},P_{r})$. ###### Theorem 3. For a fixed $n\geq 3$, we consider $B$ as in (2), and pick $b\in\\{2,...,n-1\\}$ such that $2(b-1)$ divides $m=n-1$ ($m=2(b-1)k$). For $F:=\\{0,b-1\\}$, we let $\begin{array}[]{c}B_{1}^{\prime}:=\\{\underset{b-1}{\underbrace{1,2,...,b-1}},\underset{b-1}{\underbrace{2b-1,2b,...,3b-3}},\underset{b-1}{\underbrace{4b-3,4b-2,...,5b-5}},\\\ \\\ ...,\underset{b-1}{\underbrace{(2k-2)(b-1)+1,...,(2k-1)(b-1)}}\\},\end{array}$ (6) $B_{1}:=B_{1}^{\prime}\cup(B_{1}^{\prime}+2m)\cup(B_{1}^{\prime}+4m)\cup(B_{1}^{\prime}+6m)\cup...$ and $T$ a subset of $B_{1}$. Then $T_{1}:=F+T$ and $B_{1}$ are independent sets relative to the induced probability measure on $B$. Moreover, $A_{1}:=T_{1}+\\{0,n-1\\}$, $A_{2}:=B_{1}+\\{0,n-1\\}$ and $B$ form a triple of mutually independent sets in $(\mathbb{N}_{0},{\cal A},P_{r})$ for all $r$. ###### Proof. The second part of the theorem follows from the considerations we made before the theorem and from the first part. To show the first part we need to check (4) for $T_{1}$ and $T_{2}=B_{1}$. Let us remember that $\begin{array}[]{l}B=\\{\underset{n-1}{\underbrace{1,2,...,n-1}},\underset{n-1}{\underbrace{2n-1,2n,...,3n-3}},\\\ \\\ \underset{n-1}{\underbrace{4n-3,4n-2,...,5n-5},...}\\},\ and\ \ P_{r}(B)=\frac{1}{1+r^{m}}.\end{array}$ We observe that $B_{1}^{\prime}\subset\\{1,2,...,n-1\\}$ and so $B_{1}\subset B$. Let us first take into consideration the case $T=\\{1\\}$. Since $T_{1}=\\{1,b\\}$ we get $T_{1}\cap T_{2}=\\{1\\}$, $P_{r}(T_{1})=(1-r)(1+r^{b-1})$, and $P_{r}(B_{1})=P_{r}(B_{1}^{\prime})(1+r^{2m}+r^{4m}+r^{6m}+...)=\frac{P_{r}(B_{1}^{\prime})}{1-r^{2m}}.$ So, it remains to calculate $P_{r}(B_{1}^{\prime})$: $\begin{array}[]{l}P_{r}(B_{1}^{\prime})=\frac{1-r}{r}(r+r^{2}+...r^{b-1})(1+r^{2(b-1)}+r^{4(b-1)}+...+r^{2(k-1)(b-1)})\\\ \\\ \displaystyle=(1-r^{b-1})\frac{1-r^{2k(b-1)}}{1-r^{2(b-1)}}=\frac{1-r^{m}}{1+r^{b-1}}\Rightarrow P_{r}(B_{1})=\frac{1}{(1+r^{b-1})(1+r^{m}).}\end{array}$ This shows that (4) is satisfied. In the general case, i.e. $T$ an arbitrary subset of $B_{1}$, we proceed as in the proof of Theorem 1. $\hfill\blacksquare$ ## 4 Uncountable sequences of independent events In [10], Szekely and Mori give an example of an infinite sequence of independent sets in $(\mathbb{N}_{0},{\cal A},P_{1/2})$. Given an infinite sequence of independent sets $\\{A_{n}\\}_{n}$ we may assume that $P_{r}(A_{k})\leq\frac{1}{2}$ and so by Proposition 1.1 in [10] we must have $\sum_{k=1}^{\infty}P_{r}(A_{k})<\infty.$ Let us observe that Theorem 2 can be applied to a different space now that can be constructed within $B$ given by (2)in terms of classes: $\widehat{\mathbb{N}}_{0}=\\{\hat{0},\hat{1},\hat{2},...\\}$ where $\hat{0}=\emptyset$, $\hat{1}:=\\{1,2,...,n-1\\}$, $\hat{2}:=\\{2n-1,2n,...,3n-3\\}$, $\hat{3}:=\\{4n-3,4n-2,...,5n-5\\}$, …, and the probability on this space is the conditional probability as subsets of $B$. Hence for $k\in\mathbb{N}$, one can check that $P(\hat{k})=\frac{1-r^{2m}}{r^{2m}}r^{2km},\ with\ m=n-1.$ This shows that this space is isomorphic to $(\mathbb{N}_{0},{\cal A},P_{s})$ with $s=r^{2m}$. One can check by induction the following proposition. ###### Proposition 1. Let $n\in\mathbb{N}$, $n\geq 2$. If $A_{1}$,…,$A_{n}$ are independent in $\widehat{\mathbb{N}}_{0}$ then $A_{1}+T$, $A_{2}+T$,…, $A_{n}+T$ and $B$ are indepenedent in $(\mathbb{N}_{0},{\cal A},P_{r})$. This construction can be then iterated indefinitely giving rise of a sequence $B$, $B_{1}$, $B_{2}$,…, which is going to be independent and its construction is in terms of a sequence $(n,n_{1},n_{2},...)$ with $n_{k}\geq 2$. As a result, we have a countable way of constructing sequences of independent sets. This construction coincides with the one in [10] if $n_{k}=2$ for all $k\in\mathbb{N}$. ## References * [1] Y. M. Baryshnikov and B. Eisenberg, Independent events and independent experiments, Proc. Amer. Math. Soc., vol. 118, No. 2, 1993, pp. 615-617 * [2] P. Billingsley, Probability and Measure, 3rd ed. J. Wiley $\&$ Sons, New York, 1995 * [3] Z. Chen, H. Rubin and R. A. Vitale, Independence and determination of probabilities, Proc. Amer. Math. Soc. 125, No. 12, (1997) * [4] Z. Chen, H. Rubin and R. A. Vitale, Addendum to “Independence and determination of probabilities”, Proc. Amer. Math. Soc. 129, No 9, (2001) * [5] W. Edwards, R. Shiflett, H. Shultz, Dependent probability spaces, Collega Mathematics Journal 39, No. 3, (2008), pp. 221-226 * [6] B. Eisenberg and B. K. Ghosh, Independent events in a discrete uniform probability space, Amer. Statist. 41 (1987), 52-56 * [7] C. M. Grinstead and J. L. Snell, Introduction to Probability, AMS, 1997, 510 pp * [8] I. Niven, H. S. Zuckerman and H.L. Montgomery, An introduction to the theory of numbers, 5th ed. J. Wiley $\&$ Sons, New York, 1991 * [9] R. C. Shiflett and H. S. Shultz, An approach to independent sets, Mathematical Spectrum 12 (1997/80), pp. 11-16. * [10] G. J. Skekely and T. F. Mori, Independence and atoms, Proc. Amer. Math. Soc. 130, No 1, pp 213-216, 2001	arxiv-papers	2008-08-22T22:52:02	2024-09-04T02:48:57.407160	{ "license": "Public Domain", "authors": "Eugen J. Ionascu and Alin A. Stancu", "submitter": "Eugen Ionascu Dr", "url": "https://arxiv.org/abs/0808.3155" }
0808.3172	# Reconstruction of the unitary symmetry in super-relativity Peter Leifer (Cathedra of Informatics, Crimea State Engineering and Pedagogical University, 21 Sevastopolskaya st., 95015 Simferopol, Crimea, Ukraine; Hermon Laboratories, Ltd. Binyamina, 30500 Israel leifer@bezeqint.net, peter@hermonlabs.com ) ###### Abstract The reconstruction of the unitary symmetry [1] under non-linear dynamical mapping Hilbert space of action amplitudes $C^{N}$ onto projective Hilbert space $CP(N-1)$ [2] has been applied here to the quantum dynamics of elementary vacuum excitations. The “vacuum manifold of virtual action states” is represented here by $CP(N-1)$ whereas its tangent vectors define local dynamical variables (LDV’s) describing “matter”. The conservation laws of LDV’s express self-conservation of the “material particles” during continuous evolution being expressed as the affine parallel transport agrees with Fubuni- Study metric, create the “affine gauge potential” as the solution of the partial differential equations. Such procedure embeds the quantum dynamics into dynamical space-time whose state-dependent coordinates arose due to encoding results of quantum measurement by the qubit spinor whose components subjected to Lorentz transformations of “quantum boosts” and “quantum rotations”. Thereby, in the framework of super-relativity, the objective character of the quantum measurement is inherently related to the dynamical space-time structure that replaces the notion of “observer”. PACS 03.65.Ca; 03.65.Ta; 04.20.Cv ## 1 Introduction Space-time being macroscopically observable as global pseudo-Riemannian manifold may emerge due to newly defined objective quantum measurement [3]. It means that (objective) quantum interaction may be used as an operational procedure for “marking” non-local quantum “lump” used as a “pointer”. I would like to recall that the localization and “marking” classical events by means of classical electromagnetic field is based on the distinguishability (separability), i.e. individualization of pointwise material objects. However we loss the possibility to distinguish (non-local) quantum objects by mean of quantum fields in space-time and, hence, it is impossible directly to identify them with space-time points. Therefore obviously non-local quantum particles are too complicated objects in order to be used as fundamental _primordial_ elements of quantum theory. The pure states of quantum motion with quantized finite action used here as a fundamental element that should be localized in some manifold. The projective Hilbert state space $CP(N-1)$ takes the place of such manifold; dynamics of non-local particles, evolution and measurement procedure formulated in fibre bundle over this base manifold. The space-time notion may be introduced only as a manifold of some coordinates prescribed by self-consistent manner to “lump”. We have, therefore, twofold aim: to find energy distribution of a quantum particle as a function of “coordinates in space-time” and to show how it emerges due to the procedure of the objective quantum measurement. Generally, it is important to understand that the problem of identification of physical objects is the root problem even in classical physics and that its recognition gave to Einstein the key to formalization of the relativistic kinematics and dynamics. Indeed, only assuming the possibility to detect locally an approximate coincidence of two pointwise events of a different nature it is possible to build full kinematic scheme and the physical geometry of space-time [4, 5]. As such the “state” of the local clock gives us local coordinates - the “state” of the incoming train. In the classical case the notions of the “clock” and the “train” are intuitively clear and approximately may be identified with material points or even with space-time points. This supports the illusion that material bodies present in space-time (Einstein emphasized that it is not so!). Furthermore, Einstein especially notes that he did not discuss the inaccuracy of the simultaneity of two approximately coinciding events that should be overcame by some abstraction [4]. This abstraction is of course the neglect of finite sizes (and all internal degrees of freedom) of the both real clock and train. It gives the representation of these “states” by mathematical points in space-time. Thereby the local identification of two events is the formal source of the classical relativistic theory. The world line is the second essential component of the relativistic kinematics and dynamics. Namely, we hope (and our belief is based on macroscopic experience) that the evolution of material point obeys some dynamical law and even without intermediate measurements of coordinates we know its space-time position (determinism). The world line is a mathematical expression of the “fate” of some material point and two points have two different world lines. In that sense we have identification of each material point with mathematical points of the world line. However quantum object requires especial embedding in space-time and its the identification with space-time point is impossible since the localization of quantum particles is state-dependent. Hence the identification of quantum objects requires a physically motivated operational procedure with corresponding mathematical description. The quantum problem of localization in space-time is rooted in the linearity of the fundamental wave equations. Solution of such equations can not keep stable coherent superposition. This leads to impossibility to define the trajectory of quantum particles and their identification. Nevertheless, trajectories are clearly seen on the photo plates, in Wilson, or in bubble chamber. What the commonly used expression “we see” really means? What we see and how? We see the result of billion interactions of a moving particle with ionizing atoms. The droplets of mist or bubbles are condensing in the vicinity of ionized atoms and they (droplets or bubbles) shape the trace of the particle in the medium that we identify as a “trajectory”. Subjectively, the quantum measurement is a human observation (identification plus comparison) and a numerical encoding of the result of the comparison, i.e. in fact the reply on quantum question: “yes” or “no”. The chain of such quantum questions takes the place in the example given above. What is the objective content of the quantum measurement, if any? The objective quantum measurement is in fact invariant quantum geometry without any mention of human presence. Two notions comprising the basis of the general measurement scheme have been used: velocity of the GCS deformation and its qubit encoding. The quantum measurement problem should be invariantly formulated as a comparison of quantum LDV’s inherently connected with GCS (instead of the comparison of quantum states themselves) with help of their affine parallel transport in $CP(N-1)$ and encoding the result of this comparison by the qubit spinor whose components in infinitesimally close points of the state space define the dynamical space-time [3, 6]. In order to reach objective character of the whole procedure the qubit spinor components should be given by the “invariant state reduction procedure” too. It is possible to say that objective character of the quantum measurement is inherently related to the space-time structure that replaces the notion of “observer” in the framework of super-relativity. This scheme requires an essential reconstruction of all formal quantum apparatus. ## 2 Super-relativity The concept of super-relativity [2] arose as a development of the Fock’s idea of “relativity to measuring device”. This idea may be treated originally as generalization of the relativity concept in space-time to the some “functional relativity” in the state space. However the power of this program is very limited in comparison with power of Einstein’s concepts of special or general relativity. The main reason is that the notion of the “measuring device” could not be correctly formulated in the own framework of the standard quantum theory. Some additional and, in fact, outlandish classical ingredients should be involved. It definitely related to well known “measurement problem” in quantum theory. In order to avoid these difficulties, the pure quantum construction of dynamics in the projective Hilbert space has been proposed instead of “realistic” usage of “classical analogy method”. This choice is dictated by well known properties of pure quantum states but now the role of these states should be quite different in comparison with original Schrödinger scheme. I use these states initially in pure abstract manner in order to develop unitary classification of quantum motions: this leads to the important notions of the local dynamical variables (LDV’s) represented by tangent vector fields to $CP(N-1)$. Two kinds of these fields (generators of coset and isotropy sub- group transformations) are similar to Goldstone’s and Higgs fields, therefore there is some possibility to identify the geometry of the unitary group and quantum forces. Thus, in clear reason, arose an ambitious program that has been called “super-relativity”. Definitely, the abstract classification of unitary motions is not sufficient for physical theory since it should be connected with measuring in space-time. Ordinary approach of the relativistic QFT assumes that Poincaré group should be linearly represented in some Hilbert space by dynamical variables (linear operators) acting in this Hilbert space. This approach leads to a lot of the conceptual and technical problems [7]. Technically, super-relativity [2, 3, 6, 8] uses an “inverse representation”, namely: unitary dynamical group $SU(N)$ should be non-linearly represented by non-local soliton-like field object associated with quantum particle. In other words, I assume that “initial” state of quantum system may be represented by vectors the Hilbert space of action states (AS). The action states are used here as states of hidden stationary “elementary” quantum motions without any reference to its placement or momentum (background independent formulation). The rays $\\{\|F>\\}$ of AS serve merely for “functional” localization of quantum system. The classification of unitary motions of the rays is based on pure geometric structure of the $SU(N)$, its isotropy sub-group $H[\|F>]=U(1)\times U(N-1)$ of some AS vector $\|F>$, and the coset structure $G/H[\|F>]=SU(N)/S[U(1)\times U(N-1)]=CP(N-1)$ of the unitary transformations taking the place of “quantum force” [2, 3, 6, 8]. The dynamical space-time (DST) arises as the section of the tangent fiber bundle over $CP(N-1)$ and it has a “granular structure that respects Lorentz symmetry” only locally [9]. This DST is realized as coordinates $x^{\mu}$ manifold for energy distribution due to single process of quantum evolution in two infinitesimally close points of trajectory in $CP(N-1)$ (say, due to a measurement of some local dynamical variable). I used the affine parallel transport of local Hamiltonian $CP(N-1)$ along this trajectory in order to be sure that at two different states one has the “same” quantum system (self- conservation or self-identification as the reference to infinitesimally close previous state in the Cartan’s sense). Quantum measurement is encoded by the qubit spinor (formal two-level system with eigen-states $\|yes>,\|no>$ of the quantum question). One of them will be associated with tangent vector to $CP(N-1)$ shows the direction and the speed of evolution from one general coherent state (GCS) to another under the coset transformations, the second one will be associated with the normal vector to $CP(N-1)$ representing the transformations of isotropy group of some GCS. It is well known that two infinitesimally close spinors may be formally connected by infinitesimal Lorentz transformation. I assumed that this relation may have not merely mathematical sense, but, being applied to the qubit spinor encoding result of quantum interaction (self-interaction) used for a “measurement”, as the real reason of the four-dimension nature of dynamical space-time manifold. ## 3 The universality of “corpuscule-wave duality” and second quantization The statistical analysis of the energy distribution is the base of the black body radiation [10] and the Einstein’s theory of the light emission and absorption [11]. This conceptual line was logically finished by Dirac in his method of the second quantization [12]. This approach is perfectly fits to many-body weakly interacting quantum systems and it was assumed that the “corpuscule-wave duality” is universal. However the application of this method to single quantum “elementary” particles destroys this harmony. Physically it is clear why: quantum particle is self-interacting system and this interaction is at least of the order of its rest mass. Since the nature of the mass is the open problem we do not know the energy distribution in quantum particles up to now. Notice, Einstein [13] and Schrödinger [14] treated the statistical fundament of quantum theory as a perishable and temporal. A long time it is was assumed that the dynamical model may be found in the framework of the string theory, but the epitaph to string theory [7] only subscribes the deep crisis in particle physics as whole. One of the aim of this article to show how it is possible to find energy distribution in single non-local quantum particle - “lump”. It is remarkable that Blochintzev more then 50 years before discussed the universality of wave - particle “duality” connected with the method of second quantization [15, 16]. It was shown that such universality is generally broken for interacting quantum systems. Namely, attempt to represent two interacting boson fields as the set of free quantum oscillators leads to two types of oscillators: quantized and non-quantized. The second one arises under simple relation $g>\frac{mMc^{2}}{h^{2}}$ between coupling constant $g$ and masses $m$ and $M$ of two scalar fields. For such intensity of coupling we obtain a field state without any counterparting “particles”. For self-interacting scalar field of mass $m$ the intensity of self-interaction $g$ leads to breakdown of the universality of the wave - particle “duality” if it is larger than the inverse square of the Compton wavelength: $g>\frac{m^{2}c^{2}}{h^{2}}=\frac{1}{\lambda^{2}_{C}}$. In order to build dynamical in lieu of statistical quantum theory let me assume that there exist some ideal “elementary” quantum states of internal motion with quantized finite action without any connection to environment, i.e. without interaction (background independent quantum states) [3]. Namely, the action (not energy) of some “elementary” quantum motion is assumed to be primary quantized but energy distribution should be established during dynamics in state space. The states of these “elementary” quantum motion (action states) play the role similar to the role of inertial frames in classical physics. POSTULATE 1. There are elementary quantum states $\|\hbar a>,a=0,1,...$ belonging to the Fock space of an abstract Planck oscillator whose states correspond to the quantum motions with given number of Planck action quanta. One may image some “elementary quantum states” (EAS) $\|\hbar a>$ as a quantum motions with entire number $a$ of the action quanta. These $a,b,c,...=0,1,2,...$ take the place of the “principle quantum number” serving as discrete indices $0\leq a,b,c...<~{}\infty$. Each action state of quantum system is a quantum motion in some “dynamical order” defined, say, be some Lagrangian or action functional. The AS vector of this quantum motion may be represented by the generalized coherent state (GCS) $\displaystyle\|F>=\sum_{a=0}^{\infty}f^{a}\|\hbar a>,$ (1) where $\|\hbar a>=(a!)^{-1/2}({\hat{\eta}^{+}})^{a}\|\hbar 0>$ and, thereby, may be treated as “order parameter” belonging to Hilbert space $\cal{H}$ \- “the space of the order parameter”. These quantum AS of motion do not gravitate since they are “pre-matter” and don’t posses such fundamental physical attributes like position, momentum and mass/energy by itself. Therefore their linear superposition is robust and the rays of these GCS will be the main building blocks of the model. Probably quark’s multiplete is one of the such kind of non-observable quantum states. Only velocities of variation of these states given by local dynamical variables correspond to “materialized” quantum states. In order to analyze the quantum dynamics of such states we should use unitary kinematics for classification of the GCS motions [2]. Such excitations create quantum excitations like particles, solitons, unparticles, etc., under some conditions that should be especially established and studied. The introduction of the cosmic potential $\Phi_{U}=c^{2}$ is the simplest way to “materialization” of quantum motions. It forms some global vacuum $\|\Phi_{U}>=\|\hbar 0>$ whose perturbation by the Hamiltonian $\displaystyle\hat{H}=\hbar\omega{\hat{\eta}^{+}}{\hat{\eta}}=mc^{2}{\hat{\eta}^{+}}{\hat{\eta}}=\omega\hat{S_{P}}=\frac{mc^{2}}{\hbar}\hat{S_{P}},$ (2) where the Planck’s action quanta operator $\hat{S_{P}}=\hbar{\hat{\eta}^{+}}{\hat{\eta}}$ with the spectrum $S_{a}=\hbar a$ in the separable Hilbert space $\cal{H}$ is merely the simplest case of more general expression of the action operator $\displaystyle\hat{S}=\hbar A({\hat{\eta}^{+}}{\hat{\eta}}).$ (3) where $A$ is some analytic function of ordinary Bose operators of creation- annihilation $\hat{\eta}^{+}$ and $\hat{\eta}$. Formally these oscillations may be represented by the superposition in infinite dimension manifold of the _Planck’s oscillators of action_. I discuss here $N$-level model of finite quantum action in some system whose states $\|F>$ correspond to extremals of some least action problem and describe stationary quantum motion. This relative (local) vacuum of some problem is not necessarily the state with minimal energy, it is a state with an extremal of some action functional. Thus $G=SU(N)$ is dynamical group of the order parameter and its geometry is the base for the classification of GCS motions [2]. ## 4 Geometry of the quantum evolution and/or measurement Quantum evolution has generally unitary and/or non-unitary character. The transition from the unitary regime to the non-unitary one is used frequently for the connection of micro- to macrophysics in relation with the measurement problem in quantum theory; particulary with the definition of so-called “quantum measurement machine” [17]. I, however, think that we should recognize after all the rights of quantum system to have the objective sense without any reference to necessity of “observer” or “quantum measurement machine”. I propose the natural geometric mechanism of the unitary breakdown, it is in fact the reconstruction of the unitary symmetry under non-linear dynamical mapping [1]. This approach leads to following result: objective character of the quantum theory related to self-consistent introduction of the dynamical space-time. The internal unitary classification of quantum motions has been already proposed [2]. Physical motivation given 12 years before may be reinforced now by appeal to dynamical effects of Lorentz transformations as follows. One may assume that approximately the following equation for “free” electrons is correct $\displaystyle\|electron,x>=\|electron,y>=\|F(x)>=\|F(y)>or$ (4) $\displaystyle\|F(x)>=\|F(Lx+a)>=\|U_{H[\|F(x)>]}\|F(x)>,$ (5) where $U_{H[\|F(x)>]}$ unitary transformations from the isotropy subgroup of $H[\|F(x)>]=U(1)\times U(N-1)$. Since realization of the this subgroup is state-dependent the embedding (parametrization) of subgroup $H[F(x)]$ and the coset transformations $G/H[\|F(x)>]=SU(N)/S[U(1)\times U(N-1)]=CP(N-1)[\|F(x)>]$ is state-dependent too [6]. It has been shown that dynamical reconstruction of the unitary symmetry $SU(N)$ leads to creation of non-local “lump” of surrounding field $\Omega^{\alpha}(x)$ belong to adjoint representation of $SU(N)$. Physically it is motivated by the fact that states of two electrons may be identical only asymptotically since they interact at finite distance and mutually perturb their quantum states. Moreover, quantum electron is self-interacting and therefore even single electron in point $x$ and in point $y$ is not generally “same”. Strictly speaking it means that naive quantization of interacting (and self-interacting) systems leads if not to contradictions, but at least to new “unparticle physics” predicted by Blochintzev more then 50 years before [15, 16]. Formally it means that the quantum state $\|F(Lx+a)>$ of electron in the point $y=Lx+a$ could not be obtained by action of the isotropy group of $\|F(x)>$ representing Poincaré transformation. Generally one has rather $\displaystyle\|F(x)>\rightarrow\|F(y)>=\|F(Lx+a)>=(G/H)\|F(x)>$ (6) $\displaystyle\neq U_{H[\|F(x)>]}\|F(x)>.$ (7) However the last equation could not be globally exact since the coset parametrization is state dependent. Only locally in $CP(N-1)$ and in the space-time it is may be correct. It is the consequence of the fact that an actual (not mental) quantum motion from the one space-time point to another leads to dynamical effects. In order to take into account such effect I assumed that two postulates should be used: 1\. “Super-relativity” : coset deformation of quantum state may be created and compensated by some physical fields and 2\. “Dynamical space-time structure emergence” :quantum measurement of the local dynamical variables may be encoded by the local Lorentz transformations of the qubit spinor representing the state of two-level detector into local dynamical space-time. Thereby local space-time structure accompanying quantum dynamics may be established. Therefore the differential form of the equivalence $\displaystyle\|F(x+dLx)>=\|F(x)>+d(G/H)\|F(x)>$ (8) should be used. I formulate the equivalence problem (8) with help field equations for the $SU(N)$ parameters $\Omega^{\alpha},(1\leq\alpha\leq N^{2}-1)$ providing the affine parallel transport of the Hamiltonian field $H^{i}=\hbar\Omega^{\alpha}\Phi^{i}_{\alpha}$ [3, 8, 18]. Furthermore, since the space-time coordinates do not have physical meaning by itself, they should be introduced in self-consistent manner, i.e, dynamical reconstruction of the global unitary symmetry which is broken due to non- linear dynamical mapping lead to infinitesimal transformations of the surrounding “fields shell” $\Omega(x)$ if one expresses the conservation law of the local Hamiltonian in the form of the affine parallel transport in $CP(N-1)$. Then the quantum measurement of the LDV being encoded with help infinitesimal Lorentz transformations of qubit spinor leads to the emergence of the dynamical space-time. Non-local soliton-like objects arising due to affine parallel transport of the LDV’s breaks global Lorentz symmetry but it may be restored locally in dynamical space-time. Let me assume that “ground state” $\|G>=\sum_{a=0}^{N-1}g^{a}\|\hbar a>$ is a solution of some the least action problem. Since any action state $\|G>$ has (in appropriate basis) the isotropy group $H=U(1)\times U(N)$, only the coset transformations $G/H=SU(N)/S[U(1)\times U(N-1)]=CP(N-1)$ effectively act in $\cal{H}$. Therefore the ray representation of $SU(N)$ in $C^{N}$, in particular, the embedding of $H$ and $G/H$ in $G$, is a state-dependent parametrization. As I wrote before, we should use the local from of the equivalence principle (6). Technically the local $SU(N)$ unitary classification of the quantum motions requires the transition from the matrices of Pauli $\hat{\sigma}_{\alpha},(\alpha=1,...,3)$, Gell-Mann $\hat{\lambda}_{\alpha},(\alpha=1,...,8)$, and in general $N\times N$ matrices $\hat{\Lambda}_{\alpha}(N),(\alpha=1,...,N^{2}-1)$ of $AlgSU(N)$ to the tangent vector fields to $CP(N-1)$ in local coordinates [2]. The transition to the local coordinates is in fact non-linear dynamical mapping onto $CP(N-1)$ [1]. Hence, there is a diffeomorphism between the space of the rays marked by the local coordinates in the map $U_{j}:\\{\|G>,\|g^{j}\|\neq 0\\},j>0$ $\pi^{i}_{(j)}=\cases{\frac{g^{i}}{g^{j}},&if $1\leq i<j$\cr\frac{g^{i+1}}{g^{j}}&if $j\leq i<N-1$}$ (9) and the group manifold of the coset transformations $G/H=SU(N)/S[U(1)\times U(N-1)]=CP(N-1)$ and the isotropy group of the corresponding ray with local coordinates (8). This diffeomorphism is provided by the coefficient functions $\Phi^{i}_{\alpha}$ $\Phi_{\sigma}^{i}=\lim_{\epsilon\to 0}\epsilon^{-1}\biggl{\\{}\frac{[\exp(i\epsilon\hat{\lambda}_{\sigma})]_{m}^{i}g^{m}}{[\exp(i\epsilon\hat{\lambda}_{\sigma})]_{m}^{j}g^{m}}-\frac{g^{i}}{g^{j}}\biggr{\\}}=\lim_{\epsilon\to 0}\epsilon^{-1}\\{\pi^{i}(\epsilon\hat{\lambda}_{\sigma})-\pi^{i}\\}$ (10) of the local generators $D_{\sigma}=\Phi_{\sigma}^{i}\frac{\partial}{\partial\pi^{i}}+c.c.$ (11) comprise of non-holonomic basis of $CP(N-1)$ [2]. In fact the definition (10) is equivalent to the first variation of the dynamical mapping (see (7.4.4) in [1]). This provides the local projection of the unitary group $SU(N)$ onto the base manifold $CP(N-1)$. ### 4.1 New definition of the state vector The general spirit of the newly defined quantum evolution requires a new construction of the state vector $\|\Psi>$. It will be associated now with a velocity of the GCS variation and, thereby, with vector field of differential operators in $\pi^{i}$ with coefficient functions given by $\Phi^{i}_{\alpha}$. In that sense this construction is similar to the second quantization scheme where state vector has an operator nature too. The coordinates of the “ground state” $\|G>=\sum_{a=0}^{N-1}g^{a}\|\hbar a>$ may be expressed in local coordinates as follows: for $a=0$ one has $\displaystyle g^{0}(\pi^{1}_{j(p)},...,\pi^{N-1}_{j(p)})=(1+\sum_{s=1}^{N-1}\|\pi^{s}_{j(p)}\|^{2})^{-1/2}$ (12) and for $a:1\leq a=i\leq N-1$ one has $\displaystyle g^{i}(\pi^{1}_{j(p)},...,\pi^{N-1}_{j(p)})=\pi^{i}_{j(p)}(1+\sum_{s=1}^{N-1}\|\pi^{s}_{j(p)}\|^{2})^{-1/2}.$ (13) Then the velocity of the ground state evolution relative the length parameter in $CP(N-1)$ playing the role of the “world time” $\tau$ is given by the formula $\displaystyle\|\Psi>\equiv\|T>=\frac{d\|G>}{d\tau}=\frac{\partial g^{a}}{\partial\pi^{i}}\frac{d\pi^{i}}{d\tau}\|\hbar a>+\frac{\partial g^{a}}{\partial\pi^{i}}\frac{d\pi^{i}}{d\tau}\|\hbar a>$ (14) $\displaystyle=\|T_{i}>\frac{d\pi^{i}}{d\tau}+\|T_{i}>\frac{d\pi^{i}}{d\tau}=H^{i}\|T_{i}>+H^{i}\|T_{i}>,$ (15) is the tangent vector to the evolution curve $\pi^{i}=\pi^{i}(\tau)$, where $\displaystyle\|T_{i}>=\frac{\partial g^{a}}{\partial\pi^{i}}\|\hbar a>=T^{a}_{i}\|\hbar a>,\quad\|T_{i}>=\frac{\partial g^{a}}{\partial\pi^{i}}\|\hbar a>=T^{a}_{i}\|\hbar a>.$ (16) Thereby state vector $\|\Psi>\equiv\|T>$ giving velocity of evolution, is represented by the tangent vector to the projective Hilbert space $CP(N-1)$ in local coordinates $\pi^{k}_{(j)}=\frac{g^{k}}{g^{j}}$ of the quantum states. The parallel transport of $\|\Psi>$ is required to be in agreement with the Fubini-Study metric $G_{ik^{}}=[(1+\sum\|\pi^{s}\|^{2})\delta_{ik}-\pi^{i^{}}\pi^{k}](1+\sum\|\pi^{s}\|^{2})^{-2}.$ (17) Then the affine connection $\displaystyle\Gamma^{i}_{mn}=\frac{1}{2}G^{ip^{}}(\frac{\partial G_{mp^{}}}{\partial\pi^{n}}+\frac{\partial G_{p^{}n}}{\partial\pi^{m}})=-\frac{\delta^{i}_{m}\pi^{n^{}}+\delta^{i}_{n}\pi^{m^{}}}{1+\sum\|\pi^{s}\|^{2}}$ (18) takes the place of the gauge potential of the non-Abelian type playing the role of the covariant instant renormalization of the dynamical variables during general transformations of the quantum self-reference frame [3]. Velocity of the $\|\Psi>$ variation is given by the equation $\displaystyle\|A>$ $\displaystyle=$ $\displaystyle\frac{d\|\Psi>}{d\tau}$ (19) $\displaystyle=$ $\displaystyle(B_{ik}H^{i}\frac{d\pi^{k}}{d\tau}+B_{ik^{}}H^{i}\frac{d\pi^{k}}{d\tau}+B_{i^{}k}H^{i^{}}\frac{d\pi^{k}}{d\tau}+B_{i^{}k^{}}H^{i^{}}\frac{d\pi^{k}}{d\tau})\|N>$ (20) $\displaystyle+$ $\displaystyle(\frac{dH^{s}}{d\tau}+\Gamma_{ik}^{s}H^{i}\frac{d\pi^{k}}{d\tau})\|T_{s}>+(\frac{dH^{s}}{d\tau}+\Gamma_{i^{}k^{}}^{s}H^{i}\frac{d\pi^{k}}{d\tau})\|T_{s}>,$ (21) where I introduce the matrix $\tilde{B}$ of the second quadratic form whose components are defined by following equations $\displaystyle B_{ik}\|N>=\frac{\partial\|T_{i}>}{\partial\pi^{k}}-\Gamma_{ik}^{s}\|T_{s}>,\quad B_{ik^{}}\|N>=\frac{\partial\|T_{i}>}{\partial\pi^{k}}$ (22) $\displaystyle B_{i^{}k}\|N>=\frac{\partial\|T_{i}>}{\partial\pi^{k}},\quad B_{i^{}k^{}}\|N>=\frac{\partial\|T_{i}>}{\partial\pi^{k}}-\Gamma_{i^{}k^{}}^{s}\|T_{s}>$ (23) through the state $\|N>$ normal to the “hypersurface” of the ground states. Assuming that the “acceleration” $\|A>$ is gotten by the action of some linear Hamiltonian $\hat{H}_{S}$ describing the evolution (say, during a measurement), one has the “Schrödinger equation of evolution” $\displaystyle\frac{d\|\Psi>}{d\tau}$ $\displaystyle=$ $\displaystyle-i\hat{H}_{S}\|\Psi>$ (24) $\displaystyle=$ $\displaystyle(B_{ik}H^{i}\frac{d\pi^{k}}{d\tau}+B_{ik^{}}H^{i}\frac{d\pi^{k}}{d\tau}+B_{i^{}k}H^{i^{}}\frac{d\pi^{k}}{d\tau}+B_{i^{}k^{}}H^{i^{}}\frac{d\pi^{k}}{d\tau})\|N>$ (25) $\displaystyle+$ $\displaystyle(\frac{dH^{s}}{d\tau}+\Gamma_{ik}^{s}H^{i}\frac{d\pi^{k}}{d\tau})\|T_{s}>+(\frac{dH^{s}}{d\tau}+\Gamma_{i^{}k^{}}^{s}H^{i}\frac{d\pi^{k}}{d\tau})\|T_{s}>.$ (26) I should emphasize that the “world time” here is non identical to the “world time” of Stueckelberg-Horwitz since it is the time of evolution from one generalized coherent state (GCS) to another. In fact it is proportional to the length of the trajectory in $CP(N-1)$ in the sense of Fubini-Study metric. Probably it is better to call it “omnipresent time” in the sense that under identical initial conditions the rate of the quantum evolution at any place of Universe is identical. Thereby the unitary evolution of the action amplitudes generated by $\displaystyle\hat{U}(\tau)=e^{i\tau\Omega^{\alpha}\hat{\lambda}_{\alpha}}=e^{i\tau\hat{H}}$ (27) leads in general to the non-unitary evolution of the tangent vector to $CP(N-1)$ associated with “state vector” $\|\Psi>$ since the Hamiltonian $\hat{H}_{S}$ is non-Hermitian and its expectation values are as follows: $\displaystyle<N\|\hat{H}_{S}\|\Psi>$ $\displaystyle=$ $\displaystyle i(B_{ik}H^{i}\frac{d\pi^{k}}{d\tau}+B_{ik^{}}H^{i}\frac{d\pi^{k}}{d\tau}+B_{i^{}k}H^{i^{}}\frac{d\pi^{k}}{d\tau}+B_{i^{}k^{}}H^{i^{}}\frac{d\pi^{k}}{d\tau}),$ (28) $\displaystyle<\Psi\|\hat{H}_{S}\|\Psi>$ $\displaystyle=$ $\displaystyle iG_{p^{}s}(\frac{dH^{s}}{d\tau}+\Gamma_{ik}^{s}H^{i}\frac{d\pi^{k}}{d\tau})H^{p}+iG_{ps^{}}(\frac{dH^{s}}{d\tau}+\Gamma_{i^{}k^{}}^{s}H^{i^{}}\frac{d\pi^{k}}{d\tau})H^{p}$ (29) $\displaystyle=$ $\displaystyle i<\Psi\|\frac{d}{d\tau}\|\Psi>.$ (30) The minimization of the $\|A>$ under the transition from point $\tau$ to $\tau+d\tau$ may be achieved by the annihilation of the tangent component $\frac{dH^{s}}{d\tau}+\Gamma_{ik}^{s}H^{i}\frac{d\pi^{k}}{d\tau}=0,\quad\frac{dH^{s}}{d\tau}+\Gamma_{i^{}k^{}}^{s}H^{i^{}}\frac{d\pi^{k}}{d\tau}=0$ (31) i.e. under the condition of the affine parallel transport of the Hamiltonian vector field. The last equations in (28) shows that the affine parallel transport of $H^{i}$ agrees with Fubini-Study metric (17) leads to Berry’s “parallel transport” of $\|\Psi>$. ## 5 Dynamical space-time instead of “observer” Functionally invariant construction should be used instead of “observer”. I have assumed that the quantum measurement of the LDV being encoded with help infinitesimal Lorentz transformations of qubit spinor leads to emergence of the dynamical space-time that takes the place of the objective “quantum measurement machine” formalizing the process of numerical encoding the results of comparisons of LDV’s. Two these procedures are described below. ### 5.1 LDV’s comparison Local coordinates $\pi^{i}$ of the GCS in $CP(N-1)$ give reliable geometric tool for the description of quantum dynamics during interaction or self- interaction. This leads to evolution of GCS and that may be used in measuring process. Two essential components of any measurement are identification and comparison. The Cartan’s idea of “self-identification” by the reference to the previous infinitesimally close GCS has been used. Thereby, LDV is now a new essential element of quantum dynamics. We should be able to compare some LDV at two infinitesimally close GCS represented by points of $CP(N-1)$. Since LDV’s are vector fields on $CP(N-1)$, the most natural mean of comparison of the LDV’s is affine parallel transport agrees with Fubini-Study metric [2]. This parallel transport being applied to the Hamiltonian vector field $H^{i}=\hbar\Omega^{\alpha}\Phi^{i}_{\alpha}$ $\displaystyle\frac{\delta H^{k}}{\delta\tau}$ $\displaystyle=$ $\displaystyle\hbar\frac{\delta(\Phi^{k}_{\alpha}\Omega^{\alpha})}{\delta\tau}=0,$ (32) leads to the equation for differentials of the field parameters $\Omega^{\alpha}$ of the $SU(N)$ that may be expressed as follows: $\displaystyle\delta\Omega^{\alpha}=-(\Gamma^{m}_{mn}\Phi_{\beta}^{n}+\frac{\partial\Phi_{\beta}^{n}}{\partial\pi^{n}})\Omega^{\alpha}\Omega^{\beta}\delta\tau.$ (33) ### 5.2 Encoding the results of comparison The results of the comparison of LDV’s should be formalized by numerical encoding. Thus one may say that “LDV has been measured”. The invariant encoding is based on the geometry of $CP(N-1)$ and LDV dynamics, say, dynamics of the local Hamiltonian field. Its affine parallel transport expresses the self-conservation of quantum lump associated with “particle”. In order to build the qubit spinor $\eta$ of the quantum question $\hat{Q}$ [18] two orthogonal vectors $\\{\|N>,\|\Psi>\\}$ have been used. I will use following equations $\displaystyle\eta=\left(\begin{array}[]{cc}\alpha_{(\pi^{1},...,\pi^{N-1})}\\\ \beta_{(\pi^{1},...,\pi^{N-1})}\\\ \end{array}\right)=\left(\begin{array}[]{cc}\frac{<N\|\hat{H}\|\Psi>}{<N\|N>}\\\ \frac{<\Psi\|\hat{H}\|\Psi>}{<\Psi\|\Psi>}\\\ \end{array}\right)$ (38) for the measurement of the Hamiltonian $\hat{H}$ at corresponding GCS. Then from the infinitesimally close GCS $(\pi^{1}+\delta^{1},...,\pi^{N-1}+\delta^{N-1})$, whose shift is induced by the interaction used for a measurement, one get a close spinor $\eta+\delta\eta$ with the components $\displaystyle\eta+\delta\eta=\left(\begin{array}[]{cc}\alpha_{(\pi^{1}+\delta^{1},...,\pi^{N-1}+\delta^{N-1})}\\\ \beta_{(\pi^{1}+\delta^{1},...,\pi^{N-1}+\delta^{N-1})}\\\ \end{array}\right)=\left(\begin{array}[]{cc}\frac{<N\|\hat{H^{\prime}}\|\Psi>}{<N\|N>}\\\ \frac{<\Psi\|\hat{H^{\prime}}\|\Psi>}{<\Psi\|\Psi>}\end{array}\right).$ (43) Here $\hat{H}=\hbar\Omega^{\alpha}\hat{\lambda}_{\alpha}$ is the lift of Hamiltonian tangent vector field $H^{i}=\hbar\Omega^{\alpha}\Phi^{i}_{\alpha}$ from $(\pi^{1},...,\pi^{N-1})$ and $\hat{H^{\prime}}=\hbar(\Omega^{\alpha}+\delta\Omega^{\alpha})\hat{\lambda}_{\alpha}$ is the lift of the same tangent vector field parallel transported from the infinitesimally close point $(\pi^{1}+\delta^{1},...,\pi^{N-1}+\delta^{N-1})$ back to the $(\pi^{1},...,\pi^{N-1})$ into the adjoint representation space. Then one finds $\displaystyle\delta\eta=\left(\begin{array}[]{cc}\alpha_{(\pi^{1}+\delta^{1},...,\pi^{N-1}+\delta^{N-1})}-\alpha_{(\pi^{1},...,\pi^{N-1})}\\\ \beta_{(\pi^{1}+\delta^{1},...,\pi^{N-1}+\delta^{N-1})}-\beta_{(\pi^{1},...,\pi^{N-1})}\\\ \end{array}\right)=\left(\begin{array}[]{cc}\frac{<N\|\delta\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<N\|N>}\\\ \frac{<\Psi\|\delta\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<\Psi\|\Psi>}\\\ \end{array}\right),$ (48) where one should find how the affine parallel transport connected with the variation of coefficients $\Omega^{\alpha}$ in the dynamical space-time associated with quantum question $\hat{Q}$. The covariance relative transition from one GCS to another $\displaystyle(\pi^{1}_{j(p)},...,\pi^{N-1}_{j(p)})\rightarrow(\pi^{1}_{j^{\prime}(q)},...,\pi^{N-1}_{j^{\prime}(q)})$ (49) and the covariant differentiation (relative Fubini-Study metric) of vector fields provides the objective character of the “quantum question” $\hat{Q}$ and, hence, the quantum measurement. This serves as a base for the construction of the dynamical space-time as it will be shown below. Each quantum measurement consists of the procedure of encoding of quantum dynamical variable into state of a “pointer” of “macroscopic measurement machine” [17]. Quantum lump takes the place of such extended “pointer”. This extended pointer may be mapped onto dynamical space-time if one assumes that transition from one GCS to another is accompanied by dynamical transition from one Lorentz frame to another attached to adjacent point of the “pointer”. Thereby, infinitesimal Lorentz transformations define small “dynamical space- time” coordinates variations. It is convenient to take Lorentz transformations in the following form $\displaystyle ct^{\prime}=ct+(\vec{x}\vec{a})\delta\tau$ (50) $\displaystyle\vec{x^{\prime}}=\vec{x}+ct\vec{a}\delta\tau+(\vec{\omega}\times\vec{x})\delta\tau$ (51) where I put $\vec{a}=(a_{1}/c,a_{2}/c,a_{3}/c),\quad\vec{\omega}=(\omega_{1},\omega_{2},\omega_{3})$ [19] in order to have for $\tau$ the physical dimension of time. The expression for the “4-velocity” $V^{\mu}$ is as follows $V^{\mu}=\frac{\delta x^{\mu}}{\delta\tau}=(\vec{x}\vec{a},ct\vec{a}+\vec{\omega}\times\vec{x}).$ (52) The coordinates $x^{\mu}$ of points in dynamical space-time serve here merely for the parametrization of the energy distribution in the “field shell” arising under its motion according to non-linear field equations [2, 6]. It is interesting to note that “4-velocity” $V^{\mu}$ and “4-acceleration” $A^{\mu}=\frac{\delta^{2}x^{\mu}}{\delta\tau^{2}}=(\vec{a}[ct\vec{a}+\vec{\omega}\times\vec{x}],\vec{a}(\vec{a}\vec{x})+\vec{\omega}[ct\vec{a}+\vec{\omega}\times\vec{x}]).$ (53) have zero value at the origin and increase in all space-time directions. Probably, it somehow connected with observable expansion of Universe, but this topic is outside of our envision. Any two infinitesimally close spinors $\eta$ and $\eta+\delta\eta$ may be formally connected with infinitesimal “Lorentz spin transformations matrix” [19] $\displaystyle L=\left(\begin{array}[]{cc}1-\frac{i}{2}\delta\tau(\omega_{3}+ia_{3})&-\frac{i}{2}\delta\tau(\omega_{1}+ia_{1}-i(\omega_{2}+ia_{2}))\cr-\frac{i}{2}\delta\tau(\omega_{1}+ia_{1}+i(\omega_{2}+ia_{2}))&1-\frac{i}{2}\delta\tau(-\omega_{3}-ia_{3})\end{array}\right).$ (56) I have assumed that there is not only formal but dynamical reason for such transition when Lorentz reference frame moves together with GCS. Then “quantum accelerations” $a_{1},a_{2},a_{3}$ and “quantum angle velocities” $\omega_{1},\omega_{2},\omega_{3}$ may be found in the linear approximation from the equation $\displaystyle\eta+\delta\eta=L\eta$ (57) as functions of the qubit spinor components of the quantum question depending on local coordinates $(\pi^{1},...,\pi^{N-1})$ involved in the $\delta\Omega^{\alpha}$ throughout field equations (89) $\displaystyle\delta\eta=L\eta-\eta$ (58) $\displaystyle=\left(\begin{array}[]{cc}-\frac{i}{2}\delta\tau(\omega_{3}+ia_{3})&-\frac{i}{2}\delta\tau(\omega_{1}+ia_{1}-i(\omega_{2}+ia_{2}))\cr-\frac{i}{2}\delta\tau(\omega_{1}+ia_{1}+i(\omega_{2}+ia_{2}))&-\frac{i}{2}\delta\tau(-\omega_{3}-ia_{3})\end{array}\right)$ (61) $\displaystyle\left(\begin{array}[]{cc}\frac{<N\|\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<N\|N>}\\\ \frac{<\Psi\|\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<\Psi\|\Psi>}\end{array}\right).$ (64) Two complex linear equations for the infinitesimal variation of the qubit spinor contains 6 parameters: three of quantum boosts $\vec{a}$ and three of quantum rotations $\vec{\omega}$. Notice, due to (33) the left part of (48) is proportional to the qubit spinor with the complex multiplier $\displaystyle C=-(\Gamma^{m}_{mn}\Phi_{\beta}^{n}+\frac{\partial\Phi_{\beta}^{n}}{\partial\pi^{n}})\Omega^{\beta}.$ (65) Let me write this equations follows $\displaystyle\delta\eta$ $\displaystyle=$ $\displaystyle\left(\begin{array}[]{cc}-\frac{i}{2}(\omega_{3}+ia_{3})&-\frac{i}{2}(\omega_{1}+ia_{1}-i(\omega_{2}+ia_{2}))\cr-\frac{i}{2}(\omega_{1}+ia_{1}+i(\omega_{2}+ia_{2}))&-\frac{i}{2}(-\omega_{3}-ia_{3})\end{array}\right)\left(\begin{array}[]{cc}\frac{<N\|\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<N\|N>}\cr\frac{<\Psi\|\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<\Psi\|\Psi>}\end{array}\right)$ (70) $\displaystyle=$ $\displaystyle C\left(\begin{array}[]{cc}\frac{<N\|\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<N\|N>}\cr\frac{<\Psi\|\Omega^{\alpha}\hat{\lambda}_{\alpha}\|\Psi>}{<\Psi\|\Psi>}\end{array}\right).$ (73) One has therefore the eigen-problem and it is easy to find that $\displaystyle C=-(\Gamma^{m}_{mn}\Phi_{\beta}^{n}+\frac{\partial\Phi_{\beta}^{n}}{\partial\pi^{n}})\Omega^{\beta}=\pm\frac{1}{2}\sqrt{\vec{a}^{2}-\vec{\omega}^{2}-2i(\vec{a}\vec{\omega})}.$ (74) Eigen-vectors for these complex eigen-values are as follows: $\displaystyle V_{1}=\left(\begin{array}[]{cc}-\frac{i(\omega_{1}+ia_{1}-i\omega_{2}+a_{2})}{\sqrt{\vec{a}^{2}-\vec{\omega}^{2}-2i(\vec{a}\vec{\omega})}+i\omega_{3}-a_{3}}\cr 1\end{array}\right),\quad V_{2}=\left(\begin{array}[]{cc}1\cr-\frac{i(\omega_{1}+ia_{1}-i\omega_{2}+a_{2})}{-\sqrt{\vec{a}^{2}-\vec{\omega}^{2}-2i(\vec{a}\vec{\omega})}+i\omega_{3}-a_{3}}\end{array}\right).$ (79) I put $A=\vec{a}^{2}-\vec{\omega}^{2}$ and $B=-2(\vec{a}\vec{\omega})$, then $4C^{2}=A+iB$ and four equations arising under square from eigen-problem $\displaystyle\left(\begin{array}[]{cc}-\frac{i}{2}(\omega_{3}+ia_{3})&-\frac{i}{2}(\omega_{1}+ia_{1}-i(\omega_{2}+ia_{2}))\cr-\frac{i}{2}(\omega_{1}+ia_{1}+i(\omega_{2}+ia_{2}))&-\frac{i}{2}(-\omega_{3}-ia_{3})\end{array}\right)\left(\begin{array}[]{cc}u_{1}+iv_{1}\\\ u_{2}+iv_{2}\end{array}\right)$ (84) $\displaystyle=C\left(\begin{array}[]{cc}u_{1}+iv_{1}\\\ u_{2}+iv_{2}\end{array}\right).$ (87) Thereby, one has two complex (four real) equations for six variables $\vec{\omega},\vec{a}$ as functions of four real variables $u_{1},v_{1},u_{2},v_{2}$ and two equations of two real $A,B$. Analytical solution of this system is not found. ## 6 Energy distribution in the “lump” It is clear that spatially non-local description of the extended state could not be relativistically invariant. It means that two vertices say, A and B, arising under interaction of some detecting field with spatially extended soliton-like object lie outside of mutual light cones. In this case quantum measurement at A (in the frame moving toward A) determines value of some dynamical variable at B and vice verse. A long time this fact was the main argument against non-local quantum field theory. However there are no natural reasons for requirements of conservation of the global causal relations and space-time locality in self-interacting extended systems. For such kind of quantum systems, the relativity should be accompanied by super-relativity to the choice of functional reference frame [2, 3, 6]. Namely, broken Lorentz symmetry widely discussed now (see, say, [20]), should be locally restored with help the affine parallel transport of the local Hamiltonian in the projective Hilbert state space that leads to extended soliton-like solutions [3]. It is defined by the velocity of variation of qubit spinor $\eta$ during parallel transport of local Hamiltonian. Moreover, there is some affine gauge field which in some sense restores global Lorentz invariance since the filed equations (23) for for the lump are relativistically invariant. In fact not any classical field in space-time correspond to the parallel transport in $CP(N-1)$, but in dynamical space-time permissible only fields corresponding to conservation laws in $CP(N-1)$. These conservation laws are expressed by the affine parallel transport. The parallel transport of the local Hamiltonian provides the “self-conservation” of extended object, i.e. the affine gauge fields couple the soliton-like system (89) discussed in [2, 6]. The field equations for the $SU(N)$ parameters $\Omega^{\alpha}$ dictated by the affine parallel transport of the Hamiltonian vector field $H^{i}=\hbar\Omega^{\alpha}\Phi^{i}_{\alpha}$ $\displaystyle\frac{\delta H^{k}}{\delta\tau}$ $\displaystyle=$ $\displaystyle\hbar\frac{\delta(\Phi^{k}_{\alpha}\Omega^{\alpha})}{\delta\tau}=0,$ (88) are quasi-linear PDE $\frac{\delta\Omega^{\alpha}}{\delta\tau}=V^{\mu}\frac{\partial\Omega^{\alpha}}{\partial x^{\mu}}=-(\Gamma^{m}_{mn}\Phi_{\beta}^{n}+\frac{\partial\Phi_{\beta}^{n}}{\partial\pi^{n}})\Omega^{\alpha}\Omega^{\beta},\quad\frac{d\pi^{k}}{d\tau}=\Phi_{\beta}^{k}\Omega^{\beta}.$ (89) These field equations describes energy distribution in the lump which does not exist a priori but is becoming during the self-interaction. The PDE equation obtained as a consequence of the parallel transport of the local Hamiltonian for two-level system living in $CP(1)$ has been shortly discussed [8, 21] $\frac{r}{c}\psi_{t}+ct\psi_{r}=F(u,v)\rho\cos\psi.$ (90) The one of the exact solutions of this quasi-linear PDE is $\displaystyle\psi_{exact}(t,r)=\arctan\frac{\exp(2c\rho F(u,v)f(r^{2}-c^{2}t^{2}))(ct+r)^{2F(u,v)}-1}{\exp(2c\rho F(u,v)f(r^{2}-c^{2}t^{2}))(ct+r)^{2F(u,v)}+1},$ (91) where $f(r^{2}-c^{2}t^{2})$ is an arbitrary function of the interval. It is interesting that this non-monotonic distribution of the force field describing “lump” [2, 3, 6, 8, 18, 21] that looks like a bubble in the dynamical space- time. The question about stability of this solution and whether such approach deletes the necessity of some additional stabilization forces should be studied carefully. ## 7 Summary Affine gauge field associated with parallel transport of the local Hamiltonian in $CP(N-1)$ compensates the breakdown of Lorentz symmetry arose due to non- locality of such “elementary” particle. Phenomenologically it may appear as new states or particles resulting “deformation” of the Hamiltonian during the parallel transport and the continuous “measurement” provided by “quantum boosts” and “quantum rotations” of local Lorentz reference frame. In other words: in order to avoid the contradiction with causality, the local Lorentz reference frame should be adapted during “scanning along lump”. Such local Lorentz reference frame has been built above whose “quantum boosts” and “quantum rotations” are defined by formulas (74) and (84). Here we have example of relativistic non-local solution (lump) arose due to restoration of the Lorentz symmetry. Extended lumps represent smooth transition from one GCS to another. The conservation laws of LDV’s lead to non-monotonic distribution of the force field describing “lump” that looks like a bubble in the dynamical space-time. Probably it deletes the necessity of some additional stabilization forces preventing flying apart the “elementary” particle. Objective quantum measurement is understood here as a comparison of the local dynamical variables and the dynamical space-time serves as “measurement machine” for the measurement encoding. Summarizing it is possible to say that there is an unification of such kind of quantum measurement and dynamical space-time structure, namely: space-time does not exists as a physical entity without self-interacting quantum “lump” used as a “pointer”. Therefore, the geometric formulation of QM being taken not as embellishment but as serious reconstruction, paves the way to new physical interpretation resolving old paradoxes (EPR, Schrödinger’s Cat), namely: standard QM is incomplete and non-local [6]. It requires reformulation in accordance with super-relativity like the classical mechanics was reformulated in accordance with Lorentz invariance of Maxwell equations. ## References [1] H. Umezava and H. Matsumoto, M. Tachiki, Thermo Field Dynamics and Condensed States, North-Holland Publishing Company; Amsterdam-New-York-Oxford (1982). * [2] P. Leifer, Found. Phys. 27, (2) 261 (1997). * [3] P. Leifer, Annales de la Fondation Louis de Broglie, 32, (1) 25 (2007). * [4] A. Einstein, Ann. Phys. 17, 891 (1905). * [5] A. Einstein, Ann. Phys. 49, 769 (1916). * [6] P. Leifer, arXiv:0804.1931v1. * [7] B. Schroer, 0805.1911v2 [hep-th]. * [8] P. Leifer, arXiv:gr-qc/0503083. * [9] Y. Bonder, arXiv:0801.2919v1 [gr-qc]. * [10] M. Planck, Ann. Phys., 4, 553 (1901). * [11] A. Einstein, Ann. Phys., 17, 132 (1905). * [12] P.A. Dirac, Proc. Royal. Soc. A 114, 243 (1927). * [13] A. Einstein, B. Podolsky and N. Rosen, Phys.Rev. 47, 777 (1935). * [14] E. Schrödinger, Naturewissenschafen 23: pp.807-812; 823-828; 844-849 (1935). * [15] D.I. Blochntzev, “Uspechy Phys. Nauk”, XLIV, No.1, 104 (1951). * [16] D.I. Blochntzev, “Uspechy Phys. Nauk”, XLII, No.1, 76 (1950). * [17] J. van Wezel, arXiv:0804.3026v2 [cond-mat.other]. * [18] P. Leifer, Found.Phys.Lett., 18, (2) 195 (2005). * [19] C.W. Misner, K.S. Thorne, J.A. Wheeler,Gravitation, W.H.Freeman and Company, San Francisco, 1973. * [20] V.A. Kostelecký, arXiv:0802.0581v1 [gr-qc]. * [21] P. Leifer, L.P. Horwitz, arXiv:gr-qc/0505051 v2.	arxiv-papers	2008-08-23T08:36:22	2024-09-04T02:48:57.412590	{ "license": "Public Domain", "authors": "Peter Leifer", "submitter": "Peter Leifer", "url": "https://arxiv.org/abs/0808.3172" }
0808.3241	FACULTY OF MATHEMATICS UNIVERSITY OF BELGRADE DOCTORAL DISSERTATION MODULI OF CONTINUITY OF QUASIREGULAR MAPPINGS CANDIDATE MENTOR COMENTOR VESNA MANOJLOVIĆ MIODRAG MATELJEVIĆ MATTI VUORINEN BELGRADE 2008 ###### Contents 1. 1 Quasiconformal Mappings 1. 1.1 Introduction 2. 1.2 The extremal problems of Grötzsch and Teichmüller 3. 1.3 Moduli of continuity 4. 1.4 Inclusion relations for balls 5. 1.5 Removing a point 6. 1.6 Uniform continuity on union of two domains 7. 1.7 Quasiconformal maps with identity boundary values 8. 1.8 Distortion of two point normalized quasiconformal mappings 2. 2 Harmonic Quasiregular Mappings 1. 2.1 Subharmonicity of $\|f\|^{p}$ 2. 2.2 Moduli of continuity in Euclidean metric 3. 2.3 Lipschitz continuity up to the boundary on $B^{n}$ 4. 2.4 Bilipschitz maps Acknowledgements I would like to express my gratitude to Prof. Arsenović, Prof. Mateljević and Prof. Pavlović whose lessons and lectures seminars at graduate studies at Faculty of Mathematics, University of Belgrade gave me a solid knowledge base for my further research. It is at one of these seminars in December of 2006 that I met Prof. Matti Vuorinen from the University of Turku, Finland at the time when I was looking for a topic of Master Thesis. Our cooperation led to my Master Thesis in June of 2007, my three visits to Turku, in the period from December 2007 to June 2008, to do research in beautiful area of Quasiconformal mappings, which resulted in our two joint papers and this Dissertation. It is difficult to overstate my appreciation for his invaluable and crucial guidance throughout this process. Most of the first chapter is an elaboration of his ideas on natural metrics. I wish to thank Prof. Arsenović for his continuous support and advice given in relation to harmonic functions. Also, many thanks to Prof. Mateljević for useful exchange of ideas related to harmonic maps. Special thanks to Prof. Pavlović whose contribution is strongly felt in the second chapter of this thesis. Belgrade, July 2008 Vesna Manojlović Summary This thesis consists of Chapters 1 and 2. The main results are contained in the two preprints and two published papers, listed below. Chapter 1 deals with conformal invariants in the euclidean space $\mathbb{R}^{n},n\geq 2,$ and their interrelation. In particular, conformally invariant metrics and balls of the respective metric spaces are studied. Another theme in Chapter 1 is the study of quasiconformal maps with identity boundary values in two different cases, the unit ball and the whole space minus two points. These results are based on the two preprints: R. Klén, V. Manojlović and M. Vuorinen: _Distortion of two point normalized quasiconformal mappings,_ arXiv:0808.1219[math.CV], 13 pp., V. Manojlović and M. Vuorinen: _On quasiconformal maps with identity boundary values,_ arXiv:0807.4418[math.CV], 16 pp. Chapter 2 deals with harmonic quasiregular maps. Topics studied are: Preservation of modulus of continuity, in particular Lipschitz continuity, from the boundary to the interior of domain in case of harmonic quasiregular maps and quasiisometry property of harmonic quasiconformal maps. Chapter 2 is based mainly on the two published papers: M. Arsenović, V. Kojić and M. Mateljević: On Lipschitz continuity of harmonic quasiregular maps on the unit ball in $\mathbb{R}^{n}$., Ann. Acad. Sci. Fenn. Math. 33 (2008), no. 1, 315–318. V. Kojić and M. Pavlović: Subharmonicity of $\|f\|^{p}$ for quasiregular harmonic functions, with applications, J. Math. Anal. Appl. 342 (2008) 742-746 ## Chapter 1 Quasiconformal Mappings ### 1.1. Introduction Conformal invariance has played a predominant role in the study of geometric function theory during the past century. Some of the landmarks are the pioneering contributions of Grötzsch and Teichmüller prior to the Second World War, and the paper of Ahlfors and Beurling [AhB] in 1950. These results lead to farreaching applications and have stimulated many later studies [K]. For instance, Gehring and Väisälä [G3], [V1] have built the theory of quasiconformal mappings in $\mathbb{R}^{n}$ based on the notion of the modulus of a curve family introduced in [AhB]. In the first chapter of this dissertation our goal is to study two kinds of conformally invariant extremal problems, which in special cases reduce to problems due to Grötzsch and Teichmüller, resp. These two classical extremal problems are extremal problems for moduli of ring domains. The Grötzsch and Teichmüller rings are the extremal rings for extremal problems of the following type, which were first posed for the case of the plane. Among all ring domains which separate two given closed sets $E_{1}$ and $E_{2}$, $E_{1}\cap E_{2}=\emptyset$, find one whose module has the greatest value. In the general case these extremal problems lead to conformal invariants $\lambda_{G}(x,y)$ and $\mu_{G}(x,y)$ defined for a domain $G\subset\mathbb{R}^{n}$ and $x,y\in G\,.$ A basic fact is that $\lambda_{G}(x,y)^{1/(1-n)}$ and $\mu_{G}(x,y)$ are metrics. Following closely the ideas developed in [Vu1] and [Vu2] we study three topics: (a) the geometry of the metric spaces $(G,d)$ when $d$ is $\lambda_{G}(x,y)^{1/(1-n)}$ or $\mu_{G}(x,y)$, (b) the relations of these two metrics to several other metrics and (c) the behavior of quasiconformal mappings with respect to several of these metrics. One of our main results is to present a revised version of the Chart on p. 86 of [Vu1], taking into account some later developments, such as [H], [HV], [Vu2]. Then we present an application to the geometry of balls in these metrics. As a special case we investigate $\lambda$ metric in $B^{2}\setminus\\{0\\}$, continuing work of [H]. Another question we address is: if $f:(G_{i},m_{G_{i}})\longrightarrow(G_{i}^{\prime},m_{G_{i}^{\prime}})$ is uniformly continuous ($i=1,2$), is $(G,m_{G})\longrightarrow(G^{\prime},m_{G^{\prime}})$ uniformly continuous ($G=G_{1}\cup G_{2}$, $G^{\prime}=G_{1}^{\prime}\cup G_{2}^{\prime}$)? Chapter 1 concludes with displacement estimates for $K$-qc mappings which are identity on the boundary of $G$. In the second chapter we explore what additional information on a $K$-qc mappings we get if we assume it is also harmonic. We call such mappings hqc- mappings. In case $n=2$ we show that hqc map has the same type of moduli of continuity on $\overline{D}$ as on $\partial D$. A similar, for the Lipschitz case, result is proved on $B^{n}$. Finally, we show, for $n=2$, that any hqc map is bilipschitz in quasihyperbolic metric. ### 1.2. The extremal problems of Grötzsch and Teichmüller In what follows, we adopt the standard definitions notions related of quasiconformal mappings from [V1]. We use notation $B^{n}(x,r)=\\{y\in\mathbb{R}^{n}\colon\|x-y\|<r\\}$, $S^{n-1}(x,r)=\\{y\in\mathbb{R}^{n}\colon\|x-y\|=r\\}$, $H^{n}=\\{(x_{1},\dots,x_{n})\in\mathbb{R}^{n}\colon x_{n}>0\\}$ and abbreviations $B^{n}(r)=B^{n}(0,r)$, $B^{n}=B^{n}(1)$, $S^{n-1}(r)=S^{n-1}(0,r)$ and $S^{n-1}=S^{n-1}(1)$. For the modulus $M(\Gamma)$ of a curve family $\Gamma$ and its basic properties we refer the reader to [V1]. Its basic property is conformal invariance. For $E,F,G\subset\overline{\mathbb{R}}^{n}$ let $\Delta(E,F,G)$ be the family of all closed curves joining $E$ to $F$ within $G$. More precisely, a path $\gamma:[a,b]\rightarrow\overline{\mathbb{R}}^{n}$ belongs to $\Delta(E,F,G)$ iff $\gamma(a)\in E$, $\gamma(b)\in F$ and $\gamma(t)\in G$ for $a<t<b$. If $G$ is a proper subdomain of $\overline{\mathbb{R}^{n}}$, then for $x,y\in G$ with $x\neq y$ we define (1.2.1) $\lambda_{G}(x,y)=\inf_{C_{x},C_{y}}M(\Delta(C_{x},C_{y};G))$ where $C_{z}=\gamma_{z}[0,1)$ and $\gamma_{z}:[0,1)\longrightarrow G$ is a curve such that $\gamma_{z}(0)=z$ and $\gamma_{z}(t)\rightarrow\partial G$ when $t\rightarrow 1$, $z=x,y$. This conformal invariant was introduced by J. Ferrand (see [Vu2]). For $x\in\mathbb{R}^{n}\setminus\\{0,e_{1}\\}$, $n\geqslant 2$, define (1.2.2) $p(x)=\inf_{E,F}M(\Delta(E,F)),$ where the infimum is taken over all pairs of continua $E$ and $F$ in $\overline{\mathbb{R}^{n}}$ with $0,e_{1}\in E$, $x,\infty\in F$. This extremal quantity was introduced by O. Teichmüller (see [Vu2], [HV]). For a proper subdomain $G$ of $\overline{\mathbb{R}^{n}}$ and for all $x,y\in G$ define (1.2.3) $\mu_{G}(x,y)=\inf_{C_{xy}}M(\Delta(C_{xy},\partial G;G))$ where the infimum is taken over all continua $C_{xy}$ such that $C_{xy}=\gamma[0,1]$ and $\gamma$ is a curve with $\gamma(0)=x$ and $\gamma(1)=y$. For the case $G=B^{n}$ the function $\mu_{B^{n}}(x,y)$ is the extremal quantity of H. Grötzsch (see [Vu2]). Let $(X,d_{1})$ and $(Y,d_{2})$ be metric spaces and let $f:X\to Y$ be a continuous mapping. Then we say that $f$ is uniformly continuous if there exists an increasing continuous function $\omega:[0,\infty)\to[0,\infty)$ with $\omega(0)=0$ and $d_{2}(f(x),f(y))\leq\omega(d_{1}(x,y))$ for all $x,y\in X\,.$ We call the function $\omega$ the modulus of continuity of $f\,.$ If there exist $C,\alpha>0$ such that $\omega(t)\leq Ct^{\alpha}$ for all $t>0\,,$ we say that $f$ is Hölder-continuous with Hölder exponent $\alpha\,.$ If $\alpha=1\,,$ we say that $f$ is Lipschitz with the Lipschitz constant $C$ or simply $C$-Lipschitz. If $f$ is a homeomorphism and both $f$ and $f^{-1}$ are $C$-Lipschitz, then $f$ is $C$-bilipschitz or $C$-quasiisometry and if $C=1$ we say that $f$ is an isometry. These conditions are said to hold locally, if they hold for each compact subset of $X\,.$ A very special case of these are isometries. Let $(X_{1},d_{1})$ and $(X_{2},d_{2})$ be metric spaces and let $f:X_{1}\rightarrow X_{2}$ be a homeomorphism. We call $f$ an isometry if $d_{2}(f(x),f(y))=d_{1}(x,y)$ for all $x,y\in X_{1}$. In this section we introduce five types of metrics: 1. (1) Spherical (chordal) metric $q$. 2. (2) Quasihyperbolic metric $k_{G}$ of a domain $G\subset\mathbb{R}^{n}$. 3. (3) A metric $j_{G}$ closely related to $k_{G}$. 4. (4) Seittenranta’s metric $\delta_{G}$. 5. (5) Apollonian metric $\alpha_{G}$. The first one is defined on $\overline{\mathbb{R}^{n}}=\mathbb{R}^{n}\cup\\{\infty\\}$. The second and the third ones are defined in any proper subdomain $G\subset\mathbb{R}^{n}$, both of them generalize hyperbolic metric (on $B^{n}$ or $H^{n}$) to arbitrary proper subdomain $G\subset\mathbb{R}^{n}$. Seittenranta’s metric is natural, Möbius invariant analogue of the $j_{G}$-metric. Apollonian metric is defined in any proper subdomain $G\subseteq\mathbb{R}^{n}$ which boundary is not a subset of a circle or a line. 2.4. The spherical metric. The metric $q$ is defined by (1.2.5) $q(x,y)=\left\\{\begin{array}[]{ll}\displaystyle\frac{\|x-y\|}{\sqrt{1+\|x\|^{2}}\sqrt{1+\|y\|^{2}}},&x\neq\infty\neq y,\vspace{0.5em}\\\ \displaystyle\frac{1}{\sqrt{1+\|x\|^{2}}},&y=\infty.\end{array}\right.$ Absolute (cross) ratio of an ordered quadruple $a,b,c,d$ of distinct points in $\overline{\mathbb{R}^{n}}$ is defined (1.2.6) $\displaystyle\|a,b,c,d\|=\frac{q(a,c)\,q(b,d)}{q(a,b)\,q(c,d)}.$ Now we introduce distance ratio metric or $j_{G}$-metric. For an open set $G\subset\mathbb{R}^{n}$, $G\neq\mathbb{R}^{n}$ we define $d(z)=d(z,\partial G)$ for $z\in G$ and (1.2.7) $j_{G}(x,y)=\log\left(1+\frac{\|x-y\|}{\min\\{d(x),d(y)\\}}\right)$ for $x,y\in G$. For a nonempty $A\subset G$ we define the $j_{G}$-diameter of $A$ by $j_{G}(A)=\sup\\{j_{G}(x,y)\,\|\,x,y\in A\\}.$ For an open set $G\subset\mathbb{R}^{n}$, $G\neq\mathbb{R}^{n}$, and a nonempty $A\subset G$ such that $d(A,\partial G)>0$ we define $r_{G}(A)=\frac{d(A)}{d(A,\partial G)}.$ If $\rho(x)>0$ for $x\in G$, $\rho$ is continuous and if $\gamma$ is a rectifiable curve in $G$, then we define $l_{\rho}(\gamma)=\int_{\gamma}\rho\,ds.$ The Euclidean length of a curve $\gamma$ is denoted by $l(\gamma)$. Also, for $x_{1},x_{2}\in G$ we define (1.2.8) $\displaystyle d_{\rho}(x,y)=\inf l_{\rho}(\gamma),$ where the infimum is taken over all rectifiable curves from $x_{1}$ to $x_{2}$. It is easy to show that $d_{\rho}$ is a metric in $G$. Now we take any proper domain $G\subset\mathbb{R}^{n}$ and set $\rho(x)=\frac{1}{d(x,\partial G)}$. The corresponding metric, denoted by $k_{G}$, is called the quasihyperbolic metric in $G$. Since, $\rho(\varphi(x))=\frac{1}{d(\varphi(x),\partial\,(\varphi G))}=\frac{1}{d(x,\partial G)}=\rho(x),$ for Euclidean isometry $\varphi$, $k_{G^{\prime}}(x^{\prime},y^{\prime})=k_{G}(x,y),\qquad\mbox{where }G^{\prime}=\varphi(G),\,\,\,x^{\prime}=\varphi(x),\,\,\,y^{\prime}=\varphi(y).$ Now we introduce Seittenranta’s metric $\delta_{G}$ [Se]. For more details on Möbius transformations in $\mathbb{R}^{n}$ see [B1]. For an open set $G\subset\mathbb{R}^{n}$ with ${\rm card}\partial G\geqslant 2$ we set $m_{G}(x,y)=\sup_{a,b\in\partial G}\|a,x,b,y\|$ and $\delta_{G}(x,y)=\log(1+m_{G}(x,y))$ for all $x,y\in G$. Consider now the case of an unbounded domain $G\subset{\mathbb{R}}^{n},\infty\in\partial G\,.$ Note that if $a$ or $b$ in the supremum equals infinity, then we get exactly $j_{G}$ metric. This implies that we always have $j_{G}\leqslant\delta_{G}$. We will also use Apollonian metric studied by Beardon [B2], (also see [AVV, 7.28 (2)]) defined in open proper subsets $G\subset\mathbb{R}^{n}$ by $\alpha_{G}(x,y)=\sup_{a,b\in\partial G}\log\|a,x,y,b\|\quad\mbox{for all }x,y\in G.$ This formula defines a metric iff $\mathbb{R}^{n}\setminus G$ is not contained in an $(n-1)$-dimensional sphere in $\mathbb{R}^{n}$. In general, the hyperbolic-type metrics can be divided into length-metrics, defined by means of integrating a weight function and point-distance metric. Another group may again be classified by the number of boundary points used in there’s definition. So for instance, the $j$ metric is one-point metric, while the Apollonian metric is two-point metric. ###### 1.2.9 Definition. A domain $A\subset\overline{\mathbb{R}^{n}}$ is a ring if $C(A)$ has exactly two components, where $C(A)$ denotes the complement of $A\subset\mathbb{R}^{n}$. If the components of $C(A)$ are $C_{0}$ and $C_{1}$, we denote $A=R(C_{0},C_{1})$, $B_{0}=C_{0}\cap\overline{A}$ and $B_{1}=C_{1}\cap\overline{A}$. To each ring $A=R(C_{0},C_{1})$, we associate the curve family $\Gamma_{A}=\Delta(B_{0},B_{1},A)$ and the modulus of $A$ is defined by $\mod(A)=M(\Gamma_{A})$. Next, the capacity of $A$ is by definition ${\rm cap}A=\omega_{n-1}(\mod A)^{1-n}$. The complementary components of the Grötzsch ring $R_{G,n}(s)$ in $\mathbb{R}^{n}$ are $\overline{B}^{n}$ and $[s\cdot e_{1},\infty]$, $s>1$, while those of the Teichmüller ring $R_{T,n}(t)$ are $[-e_{1},0]$ and $[t\,e_{1},\infty]$, $t>0$. We shall need two special functions $\gamma_{n}(s)$, $s>1$, and $\tau_{n}(t)$, $t>0$, to designate the moduli of the families of all those curves which connect the complementary components of the Grötzsch and Teichmüller rings in $\mathbb{R}^{n}$, respectively. $\displaystyle\gamma_{n}(s)=M(\Gamma_{s})=\gamma(s),\quad\Gamma_{s}=\Gamma_{R_{G,n}}(s),$ $\displaystyle\tau_{n}(t)=M(\Delta_{t})=\tau(t),\quad\Delta_{t}=\Gamma_{R_{T,n}}(t).$ These functions are related by a functional identity [G1, Lemma 6] (1.2.10) $\gamma_{n}(s)=2^{n-1}\tau_{n}(s^{2}-1).$ ###### 1.2.11 Definition. Given $r>0$, we let $R\Psi_{n}(r)$ be the set of all rings $A=R(C_{0},C_{1})$ in $\overline{\mathbb{R}^{n}}$ with the following properties: 1. (1) $C_{0}$ contains the origin and a point $a$ such that $\|a\|=1.$ 2. (2) $C_{1}$ contains $\infty$ and a point $b$ such that $\|b\|=r$. Teichmüller first considered the following quantity in the planar case ($n=2$): $\tau_{n}(r)=\inf M(\Gamma_{A})=\inf\\{p(x)\,\|\,\|x\|=r\\},$ where the infimum is taken over all rings $A\in R\Psi_{n}(r)$ and $p(x)$ is as in (1.2.2). For $n\geqslant 3$ it was studied in [G1] and in [HV]. ###### 1.2.12 Theorem. [V1, Theorem 11.7] The function $\tau_{n}:(0,\infty)\rightarrow\mathbb{(}0,\infty)$ has the following properties: 1. (1) $\tau_{n}$ is decreasing, 2. (2) $\lim_{r\rightarrow\infty}\tau_{n}(r)=0\,,$ 3. (3) $\lim_{r\rightarrow 0}\tau_{n}(r)=\infty\,,$ 4. (4) $\tau_{n}(r)>0$ for every $r>0$. Moreover, $\tau_{n}:(0,\infty)\rightarrow\mathbb{(}0,\infty)$ and $\gamma_{n}:(1,\infty)\rightarrow\mathbb{(}0,\infty)$ are homeomorphisms. From the definition of $\tau_{n}$ and from the conformal invariance of the modulus, we obtain the following estimate: ###### 1.2.13 Theorem. Suppose that $A=R(C_{0},C_{1})$ is a ring and that $a,b\in C_{0}$ and $c,\infty\in C_{1}$. Then $M(\Gamma_{A})\geqslant\tau_{n}\left(\frac{\|c-a\|}{\|b-a\|}\right).$ Here equality holds for the Teichmüller ring, when $a=0,b=-e_{1},c=te_{1},t>0$ and $C_{0}=[-e_{1},0],C_{1}=[te_{1},\infty).$ ###### 1.2.14 Theorem. Let $C\subset B^{n}$ be a connected compact set containing $0$ and $x$, where $\|x\|<1$. Then the capacity of a ring domain with components $C_{0}=C$, $C_{1}=\\{x:\|x\|\geqslant 1\\}$ is at least $\gamma_{n}(\frac{1}{\|x\|})$. Here equality holds for the ring with the complementary components $[0,\|x\|e_{1}]$ and ${\mathbb{R}}^{n}\setminus B^{n}.$ These theorems state the extremal properties of the Teichmüller and Grötzsch rings and their proofs are based on the symmetrization theorem in [G1, Theorem 1]. ### 1.3. Moduli of continuity In this section we investigate the moduli of continuity of the identity mappings $id_{G}:(G,\rho)\longrightarrow(G,d)$ where $\rho$ and $d$ are chosen from the set of interesting metrics defined on $G$ (like quasihyperbolic metric $k$, modulus metric $\mu$ etc.). Hence, we are interested in results of type (1.3.1) $d(x,y)\leqslant\zeta(\rho(x,y))=\zeta_{\rho}^{d}(\rho(x,y)),\quad x,y\in G.$ We give several estimates of this type, and then we collect these results in a charts at the end of this section. Note that in our charts we have $\lambda_{G}^{-1}$, as well as in the inequalities of type (1.3.1); however reader should be aware that in general $\lambda_{G}^{-1}$ is not a metric. In fact $\lambda_{G}^{1/{1-n}}$ is always a metric. For more details on this matter see [Vu4]. It is well known that $j_{G}(x,y)\leqslant k_{G}(x,y)$, so $\zeta_{k}^{j}(t)=t$. ###### 1.3.2 Lemma. For $x,y\in G$ $k_{G}(x,y)\geqslant\log\left(1+\frac{m(x,y)}{\min\\{d(x),d(y)\\}}\right)\geqslant j_{G}(x,y).$ where $m(x,y)=\inf\\{l(\gamma)\,\|\,\gamma\mbox{ is a curve joining }x\mbox{ and }y\mbox{ in }G\\}$. ###### Proof. We may assume $0<d(x)\leqslant d(y)$. Choose a rectifiable arc $\gamma:[0,s]\rightarrow G$ from $x$ to $y$, parametrized by arc length: $\gamma(0)=x,\qquad\gamma(s)=y;$ obviously $s\geqslant\|x-y\|$. For any $0\leqslant t\leqslant s$ we have $d(\gamma(t))\leqslant d(x)+t,\qquad\mbox{(a key observation)},$ so, $l_{\rho}(\gamma)\geqslant\int_{0}^{s}\frac{dt}{d(x)+t}=\log\frac{d(x)+s}{d(x)}\geqslant\log\frac{d(x)+\|x-y\|}{d(x)}=j_{G}(x,y).$ ∎ The reverse inequality is not true in general; domain $G$ such that there is a constant $c>0$ such that $k_{G}\leqslant c\,j_{G}$ is called uniform domain, so in that case $\zeta_{j}^{k}(t)=ct$. ###### 1.3.3 Lemma. [Vu1, Lemma 2.21] Let $G$ be a proper subdomain of $\mathbb{R}^{n}$. If $x\in G$, $d(x)=d(x,\partial G)$ and $y\in B^{n}(x,d(x))=B_{x}$, $x\neq y$, then (1.3.4) $\lambda_{G}(x,y)\geqslant\lambda_{B_{x}}(x,y)\geqslant c_{n}\log\left(\frac{d(x)}{\|x-y\|}\right)$ where $c_{n}$ is the positive number in [V1, (10.11)]. There exists a strictly increasing function $h_{1}:(0,+\infty)\longrightarrow(0,+\infty)$ with $\lim_{t\rightarrow 0_{+}}h_{1}(t)=0$ and $\lim_{t\rightarrow+\infty}h_{1}(t)=+\infty$, depending only on $n$, such that (1.3.5) $\lambda_{G}(x,y)\leqslant h_{1}\left(\frac{\min\\{d(x),d(y)\\}}{\|x-y\|}\right)$ for $x,y\in G$, $x\neq y$. If $x\in G$ and $y\in B^{n}(x,d(x))=B_{x}$, $x\neq y$, then (1.3.6) $\mu_{G}(x,y)\leqslant\mu_{B_{x}}(x,y)=capR_{G}\left(\frac{d(x)}{\|x-y\|}\right)\leqslant\omega_{n-1}\left(\log\left(\frac{d(x)}{\|x-y\|}\right)\right)^{1-n}.$ From (1.3.6) we get $\mu_{G}(x,y)\leqslant\gamma\left(\frac{d(x)}{\|x-y\|}\right)$ for $x\in G$ and $y\in B_{x}$. It is equivalent with $\mu_{G}(x,y)\leqslant\gamma\left(\frac{1}{r}\right)$ where $r=\frac{\|x-y\|}{d(x)}$. We can express $j_{G}(x,y)$ in terms of $r$: $r=e^{j}-1$ and obtain $\mu_{G}(x,y)\leqslant\gamma\left(\frac{1}{e^{j}-1}\right).$ This gives $\zeta_{j}^{\mu}(t)=\gamma\left(\frac{1}{e^{t}-1}\right)$ locally. ###### 1.3.7 Lemma. [Vu1, Lemma 2.39] For $n\geqslant 2$ there exists strictly increasing function $h_{2}:[0,+\infty)\longrightarrow[0,+\infty)$ with $h_{2}(0)=0$ and $\lim_{t\rightarrow+\infty}h_{2}(t)=+\infty$ with the following properties. If $E$ is closed and $F$ is compact in $\mathbb{R}^{n}$ then (1.3.8) $M(\Delta(E,F))\leqslant h_{2}(T);\quad T=\min\\{j_{\mathbb{R}^{n}\setminus E}(F),j_{\mathbb{R}^{n}\setminus F}(E)\\}.$ In particular, if $G$ is a proper subdomain of $\mathbb{R}^{n}$, then (1.3.9) $\mu_{G}(x,y)\leqslant h_{2}(3k_{G}(x,y))$ for all $x,y\in G$. Moreover, there are positive numbers $b_{1},b_{2}$ depending only on $n$ such that (1.3.10) $\mu_{G}(x,y)\leqslant b_{1}k_{G}(x,y)+b_{2}$ for all $x,y\in G$. From (1.3.9) we have $\zeta_{k}^{\mu}(t)=h_{2}(3t)$. ###### 1.3.11 Lemma. [Vu1, Lemma 2.44] If $E,F\subseteq\mathbb{R}^{n}$ are disjoint continua, then $M(\Delta(E,F))\geqslant\bar{c}_{n}\min\\{j_{\mathbb{R}^{n}\setminus E}(F),j_{\mathbb{R}^{n}\setminus F}(E)\\}$ where $\bar{c}_{n}$ is a positive number depending only on $n$. ###### 1.3.12 Corollary. [Vu1, Corollary 2.46] If $E$ and $F$ are disjoint continua in $\overline{\mathbb{R}^{n}}$ and $\infty\in F$, then $M(\Delta(E,F))\geqslant c_{n}j_{\mathbb{R}^{n}\setminus F}(E).$ ###### 1.3.13 Corollary. [Vu4, Lemma 6.23] Let $G\subseteq\mathbb{R}^{n}$ be a domain $G\neq\mathbb{R}^{n}$ and connected boundary $\partial G$. Then (1.3.14) $\mu_{G}(a,b)\geqslant c_{n}j_{G}(a,b)$ holds for $a,b\in G$. If, in addition, $G$ is uniform, then (1.3.15) $\mu_{G}(a,b)\geqslant B\,k_{G}(a,b)$ for all $a,b\in G$. The first part of this corollary gives $\zeta_{\mu}^{j}(t)=\frac{1}{c_{n}}\,t$ if $\partial G$ is connected. (1.3.15) gives $\zeta_{\mu}^{k}(t)=c\,t$ if $\partial G$ is connected and $G$ is uniform. ###### 1.3.16 Lemma. [AVV, Corollary 15.13]Let $G$ be a proper subdomain of $\mathbb{R}^{n}$, $x$ and $y$ distinct points in $G$ and $m(x,y)=\min\\{d(x),d(y)\\}$. Then (1.3.17) $\lambda_{G}(x,y)\leqslant\sqrt{2}\tau\left(\frac{\|x-y\|}{m(x,y)}\right).$ From (1.3.17) using again $r=e^{j}-1$, $r=\frac{\|x-y\|}{m(x,y)}$, we have $\sqrt{2}\tau(e^{j}-1)\geqslant\lambda_{G},$ and then, since $\tau$ is decreasing, $e^{j}\leq\tau^{-1}\left(\frac{\lambda_{G}}{\sqrt{2}}\right)$ and from here $j\leqslant\log\left(1+\tau^{-1}\left(\frac{1}{\sqrt{2}\,\lambda_{G}^{-1}}\right)\right).$ Finally we obtain $\zeta_{\lambda^{-1}}^{j}(t)=\log\left(1+\tau^{-1}\left(\frac{1}{\sqrt{2}\,t}\right)\right)$. ###### 1.3.18 Definition. A closed set $E$ in $\mathbb{R}^{n}$ is called a $c$-quasiextremal distance set or $c$-QED exceptional or $c$-QED set, $c\in(0,1]$, if for each pair of disjoint continua $F_{1},F_{2}\subseteq\overline{\mathbb{R}^{n}}\setminus E$ (1.3.19) $M(\Delta(F_{1},F_{2};\overline{\mathbb{R}^{n}}\setminus E))\geqslant cM(\Delta(F_{1},F_{2})).$ If $G$ is a domain in $\overline{\mathbb{R}^{n}}$ such that $\overline{\mathbb{R}^{n}}\setminus G$ is a $c$-QED set, then we call $G$ a $c$-QED domain. ###### 1.3.20 Theorem. [Vu3, Theorem 6.21] Let $G$ be a $c$-QED domain in $\mathbb{R}^{n}$. Then (1.3.21) $\lambda_{G}(x,y)\geqslant c\tau(s^{2}+2s)\geqslant 2^{1-n}c\tau(s)$ where $s=\frac{\|x-y\|}{min(d(x),d(y))}$. From the first inequality in (1.3.21), taking into account $s=e^{j}-1$, we obtain $\lambda^{-1}=\frac{1}{\lambda}\leqslant\frac{1}{c}\,\frac{1}{\tau((s+1)^{2}-1)}=\frac{1}{c}\,\frac{1}{\tau(e^{2j}-1)}.$ This gives $\displaystyle\zeta_{j}^{\lambda^{-1}}(t)=\frac{1}{c}\,\frac{1}{\tau(e^{2t}-1)}$ for $c$-QED domain $G$. Combining $\zeta_{k}^{j}$ and $\zeta_{j}^{\lambda^{-1}}$ we estimate $\lambda_{G}^{-1}$ in terms of $k_{G}$, so $\zeta_{k}^{\lambda^{-1}}=\zeta_{j}^{\lambda^{-1}}\circ\zeta_{k}^{j}=\zeta_{j}^{\lambda^{-1}}$. In fact, we have $\lambda_{G}^{-1}\leqslant\frac{1}{c}\frac{1}{\tau(e^{2j}-1)}\leqslant\frac{1}{c\tau(e^{2k}-1)}.$ The fields $\zeta_{\mu}^{\lambda^{-1}}$, $\zeta_{\lambda^{-1}}^{k}$, $\zeta_{\lambda^{-1}}^{\mu}$ are obtained in the same fashion as $\zeta_{k}^{\lambda^{-1}}$, namely as compositions of appropriate functions $\zeta_{\rho}^{d}$. We use following inequalities. For $\zeta_{\mu}^{\lambda^{-1}}$ we have $\lambda_{G}^{-1}\leqslant\frac{1}{c\tau(e^{2j}-1)}\leqslant\frac{1}{c\tau(e^{2\mu/c_{n}}-1)}=\frac{1}{c\tau(e^{b\mu}-1)},$ where the second inequality follows from (1.3.14) and where $b=\frac{2}{c_{n}}$. For $\zeta_{\lambda^{-1}}^{k}$ we have $k_{G}\leqslant c\,j_{G}\leqslant c\log\left(1+\tau^{-1}\left(\frac{1}{\sqrt{2}\,\lambda_{G}^{-1}}\right)\right)$ and for $\zeta_{\lambda^{-1}}^{\mu}$ we have $\mu_{G}\leqslant\gamma\left(\frac{1}{e^{j}-1}\right)\leqslant\gamma\left(\frac{1}{e^{\log\left(1+\tau^{-1}\left(\frac{1}{\sqrt{2}\,\lambda_{G}^{-1}}\right)\right)}-1}\right)=\gamma\left(\frac{1}{\tau^{-1}\left(\frac{1}{\sqrt{2}\,\lambda_{G}^{-1}}\right)}\right).$ ###### 1.3.22 Theorem. [Se, Theorem 3.4] The inequalities $j_{G}\leqslant\delta_{G}\leqslant 2j_{G}$ hold for every open set $G\subset\mathbb{R}^{n}$. So, we deduce that $\zeta_{\delta}^{j}(t)=t$ and $\zeta_{j}^{\delta}(t)=2t$. ###### 1.3.23 Theorem. [Se, Theorem 4.2] Let $G\subset\mathbb{R}^{n}$ be a convex domain, then $j_{G}\leqslant\alpha_{G}$. This means that $\zeta_{\alpha}^{j}(t)=t$ for convex domains. ###### 1.3.24 Theorem. [Se, Theorem 6.2] Let $G$ be a domain in $\mathbb{R}^{n}$, for which ${\rm card}\,\partial G\geqslant 2$ and $\partial G$ is connected. Then, for distinct points $x,y\in G$, $\mu_{G}(x,y)\geqslant\tau_{n}\left(\frac{1}{e^{\delta_{G}(x,y)}-1}\right).$ Solving for $\mu$ and using a fact that $\tau_{n}$ is decreasing we get: $\tau_{n}^{-1}(\mu_{G}(x,y))\leqslant\frac{1}{e^{\delta_{G}(x,y)-1}}$ and from here $\delta_{G}(x,y)\leqslant\log\left(1+\frac{1}{\tau_{n}^{-1}(\mu_{G}(x,y))}\right).$ Hence, $\zeta_{\mu}^{\delta}(t)=\log\left(1+\frac{1}{\tau_{n}^{-1}(t)}\right)$ if $\partial G$ is connected and has at least two points. ###### 1.3.25 Theorem. [Se, Theorem 6.5] Let $G\subset\overline{\mathbb{R}^{n}}$ be a domain with card $\partial G\geqslant 2$. Then $\lambda_{G}(x,y)\leqslant\tau_{n}\left(\frac{m_{G}(x,y)}{2}\right).$ Expressing $\mu_{G}$ in terms of $\delta_{G}$ we get: $\lambda_{G}(x,y)\leqslant\tau_{n}\left(\frac{e^{\delta_{G}}-1}{2}\right)$ and from here we obtain $\delta_{G}(x,y)\leqslant\log\left(1+2\tau_{n}^{-1}\left(\frac{1}{\lambda_{G}^{-1}(x,y)}\right)\right).$ This means that $\zeta_{\lambda^{-1}}^{\delta}(t)=\log\left(1+2\tau_{n}^{-1}\left(\frac{1}{t}\right)\right)$ for domains with $card(\partial G)\geq 2$. At first, we give a $4\times 4$ chart. \| $j_{G}$ \| $k_{G}$ \| $\mu_{G}$ \| $\lambda^{-1}_{G}$ ---\|---\|---\|---\|--- $j_{G}$ \| 1 \| 2 \| 3 \| 4 $\zeta_{j}^{j}(t)=t$ \| $\zeta_{j}^{k}(t)=ct$ $G$ – uniform $\zeta_{j}^{k}(t)=\varphi(t)$ $G$ – $\varphi$ domain \| $\displaystyle\zeta_{j}^{\mu}(t)=\gamma\left(\frac{1}{e^{t}-1}\right)$ locally \| $\displaystyle\zeta_{j}^{\lambda^{-1}}(t)=\frac{1}{c\tau(e^{2t}-1)}$ $G$ – $c$-QED domain $k_{G}$ \| 5 \| 6 \| 7 \| 8 $\zeta_{k}^{j}(t)=t$ \| $\zeta_{k}^{k}(t)=t$ \| $\zeta_{k}^{\mu}(t)=h_{2}(3t)$ \| $\zeta_{k}^{\lambda^{-1}}=\zeta_{j}^{\lambda^{-1}}$ $\mu_{G}$ \| 9 \| 10 \| 11 \| 12 $\displaystyle\zeta_{\mu}^{j}(t)=\frac{1}{c_{n}}\cdot t$ $\partial G$ connected \| $\zeta_{\mu}^{k}(t)=c\cdot t$ $G$ uniform $\partial G$ connected \| $\zeta_{\mu}^{\mu}(t)=t$ \| $\zeta_{\mu}^{\lambda^{-1}}=\zeta_{\mu}^{j}\circ\zeta_{j}^{\lambda^{-1}}$ $G$ – $c$-QED domain $\partial G$ connected $\lambda^{-1}_{G}$ \| 13 \| 14 \| 15 \| 16 $\zeta_{\lambda^{-1}}^{j}(t)=\log\left(1+\tau^{-1}\left(\frac{1}{\sqrt{2}\,t}\right)\right)$ \| $\zeta_{\lambda^{-1}}^{k}=\zeta_{\lambda^{-1}}^{j}\circ\zeta_{j}^{k}$ $G$ uniform \| $\zeta_{\lambda^{-1}}^{\mu}=\zeta_{\lambda^{-1}}^{j}\circ\zeta_{j}^{\mu}$ locally \| $\zeta_{\lambda^{-1}}^{\lambda^{-1}}(t)=t$ Function $\zeta_{j}^{\mu}$ can be written in a different form using the estimate of $\gamma$ function. We define functions $\Phi$ and $\Psi$ as in [Vu2, 7.19] by (1.3.26) $\gamma_{n}(s)=\omega_{n-1}(\log(\Phi(s)))^{n-1},\quad s>1$ (1.3.27) $\tau_{n}(t)=\omega_{n-1}(\log(\Psi(t)))^{n-1},\quad t>0.$ ###### 1.3.28 Lemma. [Vu2, Lemma 7.22] For each $n\geqslant 2$ there exists a number $\lambda_{n}\in[4,2\,e^{n-1})$, $\lambda_{2}=4$, such that (1.3.29) $t\leqslant\Phi(t)\leqslant\lambda_{n}t,\quad t>1$ (1.3.30) $t+1\leqslant\Psi(t)\leqslant\lambda_{n}^{2}(t+1),\quad t>0.$ From (1.3.27) we have that $\omega_{n-1}(\log(\lambda_{n}^{2}(t+1)))^{1-n}\leqslant\tau_{n}(t)\leqslant\omega_{n-1}(\log(t+1))^{1-n}$. From (1.3.26) we have $\omega_{n-1}\left(\log\lambda_{n}t\right)^{1-n}\leqslant\gamma_{n}(t)\leqslant\omega_{n-1}\left(\log t\right)^{1-n},\quad t>1.$ Using the right side of this inequality we have $\gamma\left(\frac{1}{e^{t}-1}\right)\leqslant\omega_{n-1}\left(\log\left(\frac{1}{e^{t}-1}\right)\right)^{1-n}\leqslant\omega_{n-1}\left(\log\left(\frac{1}{t}\right)\right)^{1-n}.$ This gives $\zeta_{j}^{\mu}(t)\leqslant\omega_{n-1}\left(\log\left(\frac{1}{t}\right)\right)^{1-n}$ locally. ### 1.4. Inclusion relations for balls Each statement on modulus of continuity has its counterpart stated in terms of inclusions of balls. Namely, if for some metrics $d_{1}$ and $d_{2}$ holds $d_{1}(x,y)<t\Rightarrow d_{2}(x,y)<\zeta(t),$ then $D_{d_{1}}(x,t)\subset D_{d_{2}}(x,\zeta(t)).$ A related question is to find, for a given $x\in G$ and $t>0$, minimal $\zeta(x,t)$ such that $D_{d_{1}}(x,t)\subset D_{d_{2}}(x,\zeta(x,t)),$ This is circumscribed ball problem for a fixed $x\in G\,.$ The quasihyperbolic ball $D_{k}(x,r)$ is the set $\\{z\in G\,\|\,k_{G}(x,z)<r\\}$, when $x\in G$ and $r>0$. By [Vu2, (3.9)], we have the inclusions (1.4.1) $B^{n}(x,r\,d(x))\subset D_{k}(x,M)\subset B^{n}(x,R\,d(x)),$ where $r=1-e^{-M}$ and $R=e^{M}-1$. It was proved in [AVV, 15.13] that if $G$ is a proper subdomain of $\mathbb{\mathbb{missing}}R^{n}$ and if $x,y\in G$ with $x\neq y$, then (1.4.2) $\lambda_{G}(x,y)\leqslant\inf_{z\in\partial G}(\lambda_{\mathbb{R}^{n}\setminus\\{z\\}}(x,y))\leqslant\sqrt{2}\tau_{n}\left(\frac{\|x-y\|}{\min\\{d(x),d(y)\\}}\right)$ ###### 1.4.3 Theorem. [H, Theorem 6.11] Let $G$ be a proper subdomain of $\mathbb{R}^{n}$ and let $t>0$. We denote $c_{1}=\frac{1}{(1+\tau_{n}^{-1}(t/\sqrt{2}))}$, $c_{2}=\sqrt{\frac{\tau_{n}^{-1}(2t)}{(1+\tau_{n}^{-1}(2t))}}$ and $c_{3}=\tau_{n}^{-1}(t/\sqrt{2})$, then the inclusions (1.4.4) $D_{\lambda^{-1}}(a,t)\subset\\{z\in G\,\|\,d(z)>c_{1}d(a)\\},$ (1.4.5) $D_{\lambda^{-1}}(a,t)\supset B^{n}(a,c_{2}d(a))\supset D_{k}(a,\log(c_{2}+1))$ and (1.4.6) $D_{\lambda^{-1}}(a,t)\subset B^{n}(a,c_{3}d(a))\cap G$ are valid for all $a\in G$. If, in addition, $t>\sqrt{2}\tau_{n}(1)$, we have that (1.4.7) $B^{n}(a,c_{3}d(a))\subset D_{k}(a,\log(1/(1-c_{3}))).$ To prove the inclusion (1.4.6), we apply (1.4.2) to obtain $\lambda_{G}(a,z)\leqslant\sqrt{2}\tau_{n}\left(\frac{\|z-a\|}{d(a)}\right).$ From here with the assumption $t\leqslant\lambda_{G}(a,z)$ we have $\|z-a\|<\tau_{n}^{-1}(t/\sqrt{2})d(a)$. Since $D_{\lambda^{-1}}\subset G$, the inclusion (1.4.6) holds. Inclusion (1.4.7) follows directly from (1.4.1) after we notice that the condition $t>\sqrt{2}\tau_{n}(1)$ implies that $c_{3}<1$ and hence that the ball $B^{n}(a,c_{3}d(a))$ is included in $G$. ###### 1.4.8 Theorem. [H, Theorem 6.18] Let $G$ be a proper subdomain od $\mathbb{R}^{n}$ and assume that $G$ has a connected, nondegenerate boundary. Let $t>0$ and denote $d_{1}=\tau_{n}^{-1}(t)/(1+\tau^{-1}_{n}(t))$, $d_{2}=1/\gamma^{-1}_{n}(t)$ and $d_{3}=1/\tau^{-1}_{n}(t)$. Then, for all $a\in G$, the following inclusions hold (1.4.9) $D_{\mu}(a,t)\subset\\{z\in G\,\|\,d(z)>d_{1}d(a)\\},$ (1.4.10) $D_{\mu}(a,t)\supset B^{n}(a,d_{2}d(a))\supset D_{k}(a,\log(d_{2}+1))$ (1.4.11) $D_{\mu}(a,t)\subset B^{n}(a,d_{3}d(a))\cap G.$ If in addition $t<\tau_{n}(1)$, then (1.4.12) $B^{n}(a,d_{3}d(a))\subset D_{k}(a,\log(1/(1-d_{3}))).$ The numbers $d_{1}$, $d_{2}$ and $d_{3}$ are best possible for these inclusions. We prove (1.4.10) only, because that part is used later on. We assume that $a,z\in G$ and that $\|z-a\|\leqslant d_{2}\,d(a)$. Then, since $\gamma_{n}^{-1}(t)>1$, we have $d(z,a)<d(a)$. We consider the following curve families. $\Gamma_{J}=\Delta(J_{az},\partial G;G),$ $\Gamma=\Delta(J_{az},S^{n-1}(a,d(a));\overline{B^{n}(a,d(a))}),$ and (1.4.13) $\tilde{\Gamma}=\Delta([z^{\prime},+\infty),S^{n-1};\mathbb{R}^{n}\setminus B^{n}),$ where $z^{\prime}=\frac{d(a)}{\|z-a\|}\,e_{1}$. Since $J_{az}$ is a continuum which joins $a$ and $z$, we have (1.4.14) $\mu_{G}(a,z)\leqslant M(\Gamma_{J})$ and since $\Gamma<\Gamma_{J}$, we have that $M(\Gamma_{J})<M(\Gamma)$. Using Möbius transformations, we get (1.4.15) $M(\Gamma)=M(\tilde{\Gamma})=\gamma_{n}\left(\frac{d(a)}{\|z-a\|}\right),$ and since $\|z-a\|<d_{2}\,d(a)$ and $\gamma_{n}$ is a strictly decreasing homeomorphism, it follows that (1.4.16) $\gamma_{n}\left(\frac{d(a)}{\|z-a\|}\right)<\gamma_{n}\left(\frac{1}{d_{2}}\right)=t.$ Combining all these inequalities, we get $\mu_{G}(a,z)<t,$ which proves the left side of (1.4.10). The right side inclusion follows from (1.4.1). Theorem 1.4.3 ((1.4.6) and (1.4.7)) gives ###### 1.4.17 Theorem. $\lambda^{-1}(a,b)<\frac{1}{t}\Rightarrow k(a,b)<\log\frac{1}{1-\tau_{2}^{-1}\big{(}\frac{t}{\sqrt{2}}\big{)}},\quad\mbox{for }t>\sqrt{2}\tau_{2}(1)$ $\lambda^{-1}(a,b)<s\Rightarrow k(a,b)<\log\frac{1}{1-\tau_{2}^{-1}\big{(}\frac{1}{\sqrt{2}s}\big{)}},$ $\zeta_{\lambda^{-1}}^{k}(s)=\log\frac{1}{1-\tau_{2}^{-1}\big{(}\frac{1}{\sqrt{2}s}\big{)}},\quad s<\frac{1}{\sqrt{2}\tau_{2}(1)}.$ Also we obtain $\lambda^{-1}(x,a)<\frac{1}{t}\Rightarrow\|x-a\|<c_{3}d(a)<{\rm diam}(G)\,c_{3}(1/t)$ and from here $\zeta_{\lambda^{-1}}^{\|\cdot\|}(t)=\tau_{n}^{-1}(1/(\sqrt{2}t))\,{\rm diam}(G)\,.$ From Theorem 1.4.8 we deduce ###### 1.4.18 Theorem. In a domain $G$ with connected nondegenerate boundary: (1.4.19) $D_{\mu}(a,t)\supset D_{k}(a,\log(d_{2}+1)),\quad d_{2}=\frac{1}{\gamma^{-1}(t)},$ and $\mu(a,b)<t$ if $k(a,b)<\log(d_{2}+1)$. Also, $\zeta_{k}^{\mu}(s)=\gamma(1/(e^{s}-1))$. If we put $s=\log\left(\frac{1}{\gamma^{-1}(t)}+1\right),\quad\mbox{ we have }e^{s}-1=\frac{1}{\gamma^{-1}(t)},\quad t=\gamma\left(\frac{1}{e^{s}-1}\right).$ ###### 1.4.20 Theorem. [Se, Theorem 3.8] If $G\subset\mathbb{R}^{n}$ is open, $x\in G$ and $t>0$ then $D_{j}(x,t)\subset B^{n}(x,R)$ where $R=(e^{t}-1)\,d(x)$. This formula for $R$ is the best possible expressed in terms of $t$ and $d(x)$ only. Therefore, using $d(x)\leqslant{\rm diam}(G)$, we get $\zeta_{j}^{\|\cdot\|}(t)=(e^{t}-1)\,{\rm diam}(G)$. ###### 1.4.21 Theorem. [Se, Theorem 3.10] If $G\subset\mathbb{R}^{n}$ is an open set, $x\in G$ and $t>0$ then $D_{\delta}(x,t)\subset B^{n}(x,R)$ where $R=(e^{t}-1)\,d(x)$. As above, we get $\zeta_{\delta}^{\|\cdot\|}(t)=(e^{t}-1)\,{\rm diam}(G)$. From Lemma 1.3.7, we have that $\zeta_{k}^{\mu}(t)=h(3t).$ Now, from [H, Lemma 2.30] we may choose (the case $n=2$) $h(t)=\frac{2\pi\alpha}{\log\frac{1}{2t}},\quad\mbox{for }t\leqslant\frac{1}{4}.$ From here we have that $\zeta_{k}^{\mu}(t)=\frac{2\pi\alpha}{\log\left(\frac{1}{6t}\right)},\mbox{for }\quad t\leqslant\frac{1}{12}\quad\mbox{(more important case)}$ $\alpha=\max\\{1,\gamma\\},\quad\gamma=\frac{9}{8}\log 2>1\,,\quad\alpha=\gamma\,.$ In the second case, where $h(t)=36\beta\pi t^{2},\quad\mbox{for }t>\frac{1}{4}.$ we have $h(3t)=324\beta\pi t^{2}$, $t>\frac{1}{12}$. $\beta=\max\Big{(}1,\frac{1}{\gamma}\Big{)}=1$ $\zeta_{k}^{\mu}(t)=324\pi t^{2},\quad\mbox{for }t>\frac{1}{12}.$ \| $j_{G}$ \| $k_{G}$ \| $\mu_{G}$ \| $\lambda^{-1}_{G}$ ---\|---\|---\|---\|--- $j_{G}$ \| 1 \| 2 \| 3 \| 4 $\zeta_{j}^{j}(t)=t$ \| $\zeta_{j}^{k}(t)=ct$ $G$ – uniform $\zeta_{j}^{k}(t)=\varphi(t)$ $G$ – $\varphi$ domain \| $\displaystyle\zeta_{j}^{\mu}(t)=\omega_{n-1}\left(\log\left(\frac{1}{t}\right)\right)^{1-n}$ locally \| $\displaystyle\zeta_{j}^{\lambda^{-1}}(t)=\frac{1}{c\tau(e^{2t}-1)}$ $G$ – $c$-QED domain $k_{G}$ \| 5 \| 6 \| 7 \| 8 $\zeta_{k}^{j}(t)=t$ \| $\zeta_{k}^{k}(t)=t$ \| $\zeta_{k}^{\mu}(t)=\gamma\left(\frac{1}{e^{t}-1}\right)$ $\partial G$ connected, nondegenerate \| $\zeta_{k}^{\lambda^{-1}}=\zeta_{j}^{\lambda^{-1}}$ $\mu_{G}$ \| 9 \| 10 \| 11 \| 12 $\zeta_{\mu}^{j}(t)=\frac{t}{c_{n}}$ $\partial G$ connected \| $\zeta_{\mu}^{k}(t)=c\cdot t$ $G$ uniform $\partial G$ connected \| $\zeta_{\mu}^{\mu}(t)=t$ \| $\zeta_{\mu}^{\lambda^{-1}}=\zeta_{\mu}^{j}\circ\zeta_{j}^{\lambda^{-1}}$ $G$ – $c$-QED domain $\partial G$ connected $\lambda^{-1}_{G}$ \| 13 \| 14 \| 15 \| 16 $\zeta_{\lambda^{-1}}^{j}(t)=\log\left(1+\tau^{-1}\left(\frac{1}{\sqrt{2}\,t}\right)\right)$ \| $\zeta_{\lambda^{-1}}^{k}(t)=\log{\frac{1}{1-\tau_{2}^{-1}(1/(\sqrt{2}t))}}$ $t<\frac{1}{\sqrt{2}\tau_{2}(1)}$ \| $\zeta_{\lambda^{-1}}^{\mu}=\zeta_{\lambda^{-1}}^{j}\circ\zeta_{j}^{\mu}$ locally \| $\zeta_{\lambda^{-1}}^{\lambda^{-1}}(t)=t$ This is improved $4\times 4$ chart. ###### 1.4.22 Example. For $G\subset\mathbb{R}^{n}$ we choose $z_{0}\in\partial G$, sequence $x_{k}\in G$ such $x_{k}\rightarrow z_{0}$ and sequence $y_{k}\in G$ such that (1.4.23) $\|y_{k}-z_{0}\|<\frac{\|x_{k}-z_{0}\|}{k}.$ Clearly $\|x_{k}-y_{k}\|\rightarrow 0$ and (1.4.24) $\|x_{k}-y_{k}\|>\|x_{k}-z_{0}\|-\|y_{k}-z_{0}\|>\|x_{k}-z_{0}\|\left(1-\frac{1}{k}\right).$ But $j_{G}(x_{k},y_{k})\geqslant\log\left(1+\frac{\|x_{k}-y_{k}\|}{\|y_{k}-z_{0}\|}\right)\geqslant\log\left(1+\frac{1-\frac{1}{k}}{\frac{1}{k}}\right)=\log(k)\rightarrow+\infty.$ Hence $id:(G,\|\cdot\|)\longrightarrow(G,j_{G})$ is not uniformly continuous. By this reason, adequate fields in the chart are empty. Also, for a fixed small $d>0$ we can find $x,y\in G$ such that $\|x-y\|=d$ and $d(x,\partial G)$ as small as we like. So we get $k_{G}(x,y)$ as large as we like and there is no estimate of $k_{G}(x,y)$ in terms of $\|x-y\|$. In other hand function $\zeta_{k}^{\|\cdot\|}$ is obtained from: $k_{G}(x,y)\geqslant\int_{0}^{\|x-y\|}\frac{ds}{{\rm diam}(G)}=\frac{\|x-y\|}{{\rm diam}(G)}.$ From here we get that modulus of continuity of $id:(G,k_{G})\longrightarrow(G,\|\cdot\|)$ is $\zeta_{k}^{\|\cdot\|}(t)=t\,{\rm diam}(G)$ (where $G$ is bounded). All the remaining items are obtained by composition of the above moduli of continuity. And finally we have following charts: Sharper results can be obtained for special domains, for example $G=\mathbb{R}^{n}\setminus\\{0\\}$ was studied by R. Klen [Kl] in relation to $j_{G}$ metrics. We return to the question of moduli of continuity, from a different viewpoint, in chapter 2, sections 2 and 3. ### 1.5. Removing a point Let $\mathcal{M}$ be a collection of metrics on a domain $G\subset\mathbb{R}^{n}$ and $B_{m}(x,T)=\\{z\in G\,:\,m(x,z)<T\\}$, $m\in\mathcal{M}$. Let $\begin{array}[]{l}r_{T}=\sup\\{r>0\,:\,S^{n-1}(x,r)\subset B_{m}(x,T)\\},\vspace{0.5em}\\\ R_{T}=\inf\\{r>0\,:\,S^{n-1}(x,r)\cap B_{m}(x,T)=\emptyset\\}.\end{array}$ The question is can we find lower bound for $r_{T}$ and upper bound for $R_{T}$. ###### 1.5.1 Problem. (Radius of circumscribed ball) It is evident from the definition of $\lambda_{G}$ that adding new points, even isolated ones, to the boundary of $G$ will affect the value of $\lambda_{G}(x,y)$ for fixed points $x,y\in G$. We study this phenomenon in the case when $G=\mathbb{R}^{2}\setminus\\{0\\}$. We find an upper bound for radius of circumscribed ball, where $m=\lambda_{G}^{-1}$. We use notation $B_{\lambda}(1,T)=\\{z\in\mathbb{C}\,:\,\lambda_{G}(z,1)\geqslant T^{-1}\\}.$ Let $h(z)=\frac{z}{\|z\|^{2}}$ be an inversion. Since $h:B_{\lambda}\longrightarrow B_{\lambda}$ ($h$ is an isometry for $\lambda$ metric) we have $\lambda_{G}(1,z)=\lambda_{G}(1,h(z)).$ Figure 1.1. Radius of circumscribed ball From [SolV, (3.3), (3.22)] we have (1.5.2) $p(z)=\frac{2\pi}{\log M(2z-1)},\quad z\in\mathbb{C}\setminus\\{0,1\\}\quad\mbox{and}$ (1.5.3) $\log M(2e^{i\theta}-1)=\frac{2\pi{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(\sin\frac{\theta}{4}){\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(\cos\frac{\theta}{4})}{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}^{2}(\sin\frac{\theta}{4})+{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}^{2}(\cos\frac{\theta}{4})}.$ If we put $z=e^{i\theta}$ we have $p(e^{i\theta})=\frac{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}^{2}(\sin\frac{\theta}{4})+{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}^{2}(\cos\frac{\theta}{4})}{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(\sin\frac{\theta}{4}){\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(\cos\frac{\theta}{4})}.$ For $\|z\|=1$ we obtain $\lambda_{G}(1,z)=p(z)$. Choose $\theta$ such that $\sin\frac{\theta}{2}=\frac{R_{T}}{2}$. From here $\theta=2\arcsin\frac{R_{T}}{2}$. Now if we put (1.5.4) $y=\frac{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(\sin\frac{\theta}{4})}{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(\cos\frac{\theta}{4})}=\frac{\displaystyle 2}{\displaystyle\pi}\mu(\cos\frac{\theta}{4})$ we have $p(e^{i\theta})=y+\frac{1}{y}=\frac{1}{T}.$ We are interested for solutions $y<1$ because we want $\theta<\pi$. From here $y=\frac{2T}{1+\sqrt{1-4T^{2}}}$. Since from (1.5.4) $\theta=4\arccos(\mu^{-1}(\frac{\pi y}{2}))$ now we have (1.5.5) $\theta=4\arccos(\mu^{-1}\big{(}\frac{\pi}{2}\frac{2T}{1+\sqrt{1-4T^{2}}}\big{)})=4\arccos(\mu^{-1}\big{(}\frac{\pi T}{1+\sqrt{1-4T^{2}}}\big{)}).$ Hence, the radius of the circumscribed sphere is $R_{T}=2\sin\frac{\theta}{2},\quad T\in(0,\frac{1}{2}),\quad\theta\mbox{ from (\ref{theta})}.$ ###### 1.5.6 Open question. 1. (1) Can we find $r_{T}$ in the case above? 2. (2) Can we estimate $R_{T}$, where $G$ is now bounded subset of $\mathbb{C}$ (instead of $\mathbb{R}^{2}\setminus\\{0\\}$)? 3. (3) Consider $\mu_{G}$-balls where $\partial G$ is connected, say $\partial G=[0,e_{1}]$. Can we find a lower bound for $r_{T}$ (upper bound for $R_{T}$) in this case? ###### 1.5.7 Problem. (Estimate for $\lambda_{B^{2}\setminus\\{0\\}}(x,y)$) Next we investigate the following situation: $G\subseteq\mathbb{R}^{n}$ is domain, $a\in G$, $G^{\prime}=G\setminus\\{a\\}$. Is $\lambda_{G}(x,y)=\lambda_{G^{\prime}}(x,y)$ true under some additional assumptions, like $x,y$ close to $\partial G$? We consider a special case where $G=B^{2}$ and $a=0$. Figure 1.2. In [LeVu, Lemma 2.8] is proven that if $\Gamma_{0}=\Delta([0,x],[\tilde{y},x/\|x\|];B)$, where $\tilde{y}=\frac{\|y\|}{\|x\|}\,x$ and if we put $\|x\|=r$, $\|\tilde{y}\|=s$, then we have (1.5.8) $M(\Gamma_{0})=\tau\left(\frac{(s-r)(1-rs)}{r(1-s)^{2}}\right).$ Further, from [Vu1, (2.6)] we have that if $\Delta_{0}=\Delta([-\frac{x}{\|x\|},-x],[x,\frac{x}{\|x\|}];B)$ and if $\|x\|=r$ as before, then Figure 1.3. $M(\Delta_{0})=\frac{1}{2}\,\tau\left(\frac{4r^{2}}{(1-r^{2})^{2}}\right).$ Also, using Möbius transformation $T_{r}:B^{2}\longrightarrow B^{2}$, $T(r)=0$ we can map family of curves $\Delta_{1}$ to family of curves $\Delta_{1}^{\prime}$, where $\Delta_{1}=\Delta([-\frac{x}{\|x\|},-\tilde{y}],[0,x];B)$ and $\Delta_{1}^{\prime}=\Delta([-\frac{x}{\|x\|},-\tilde{y}^{\prime}],[-x,0];B)$. We know that $\rho(-s,0)=\rho(-r,-t),$ where $r$ and $s$ are as before and $-t=T_{r}(-s)$. Further, this is equivalent to (1.5.9) $\log\frac{1+s}{1-s}=\log\frac{1+t}{1-t}\frac{1-r}{1+r}.$ Solving (1.5.9) in $t$ we obtain $t=\frac{s+r}{1+sr}$. Now we have $M(\Delta_{1})=M(\Delta_{1}^{\prime})=\tau\left(\frac{(t-r)(1-tr)}{r(1-t)^{2}}\right)=\tau\left(\frac{s(1+r)^{2}}{r(1-s)^{2}}\right).$ Figure 1.4. The first equality holds because $T_{r}$ is conformal map, the second one follows from (1.5.8) and the third one from the expression for $t$. Now, if we put in last term that $r=s$, we obtain $M(\Delta_{1})=\tau\left(\left(\frac{1+r}{1-r}\right)^{2}\right).$ The question is when is $M(\Delta_{1})\geqslant M(\Delta_{0})$. In other words, when is (1.5.10) $\tau\left(\left(\frac{1+r}{1-r}\right)^{2}\right)\geqslant\frac{1}{2}\tau\left(\frac{4r^{2}}{(1-r^{2})^{2}}\right)?$ Applying formula [AVV, 5,19 (5)]: $\frac{1}{2}\tau(t)\geqslant\tau((\sqrt{t}+\sqrt{t+1})^{4}-1)$ for $t=4r^{2}/(1-r^{2})^{2}$ we have $\frac{1}{2}\tau\left(\frac{4r^{2}}{(1-r^{2})^{2}}\right)=\tau\left(\frac{8r(r^{2}+1)}{(1-r)^{4}}\right).$ Then (1.5.10) is equivalent to $\left(\frac{1+r}{1-r}\right)^{2}\leqslant\frac{8r(r^{2}+1)}{(1-r)^{4}},$ since $\tau$ is decreasing. The last inequality is equivalent to $r^{4}-8r^{3}-2r^{2}-8r+1\leqslant 0.$ This inequality holds for $r\in[0.12,1)$. This gives the answer to the question: For which values of $\|x\|$ we have $\lambda_{A}(x,-x)=M(\Delta(E,-E;B^{2})),$ where $A=B^{2}\setminus\\{0\\}$, $E=[x,\frac{x}{\|x\|}]$? A related result can be found in Heikkala’s dissertation, [H, Theorem 7.3]. In fact, this theorem deals with the more general situation: If $x$ and $y$ are close to the boundary and far apart then $\lambda_{B^{n}\setminus\\{0\\}}(x,y)=\lambda_{B^{n}}(x,y)$. His theorem is: ###### 1.5.11 Theorem. Let $G=B^{n}\setminus\\{0\\}$ and let $x,y\in G$ with $\|x-y\|\geq\delta>0$. Then, if $\min\\{\|x\|,\|y\|\\}\in(r_{1},1)$ with $r_{1}=\frac{\sqrt{\delta^{4}+64}-\delta^{2}}{8}$, we have that $\lambda_{G}(x,y)=\lambda_{B^{n}}(x,y).$ However, we have in the special case $x=-y$, better constant (letting $\delta=2\|x\|$ and $r_{1}=\|x\|$ in Theorem 7.3 gives equation $r_{1}^{3}+r_{1}^{2}-1=0$, and its real root is larger than $0.75$, and consequently larger than $0.12$). ### 1.6. Uniform continuity on union of two domains ###### 1.6.1 Definition. Let $\\{m_{D}\,:\,D\subseteq\overline{\mathbb{R}^{n}}\\}$ be a family of metrics. We say that this family is monotone if $D_{1}\subseteq D_{2}$ implies $m_{D_{1}}(x,y)\geq m_{D_{2}}(x,y)$ for all $x,y\in D_{1}$. ###### 1.6.2 Lemma. [Vu5, 2.27] Let $G_{1},G_{2}$ be domains in $\mathbb{R}^{n}$ with $G_{1}\cap G_{2}\neq\emptyset$, $G_{1}\neq\mathbb{R}^{n}\neq G_{2}$ and assume that there exists $c\in(0,1)$ such that (1.6.3) $d(x,\partial G_{1})+d(x,\partial G_{2})\geq c\,d(x,\partial(G_{1}\cup G_{2})),$ for all $x\in G=G_{1}\cup G_{2}$. Suppose that $f:G\longrightarrow fG$ is continuous, $fG\subseteq\mathbb{R}^{n}$; that $\\{m_{D}\,:\,D\subseteq\mathbb{R}^{n}\\}$ is a monotone family of metrics; and that (1.6.4) $m_{fG_{j}}(f(x),f(y))\leq\omega_{j}(k_{G_{j}}(x,y))$ for $x,y\in G_{j}$ and $j=1,2$. Then there exists $\omega:[0,+\infty)\longrightarrow[0,+\infty)$ such that (1.6.5) $m_{fG}(f(x),f(y))\leq\omega(k_{G}(x,y))$ and $\lim_{t\rightarrow 0+}\omega(t)=0$ provided $\lim_{t\rightarrow 0+}\omega_{j}(t)=0$, $j=1,2$. Now we consider a similar, but local result, with $j$ metric replacing $k$ metric. We can no longer use geodesics as was done in the proof of the above lemma. ###### 1.6.6 Lemma. Let $G_{1},G_{2}$ be domains in $\mathbb{R}^{n}$ with $G_{1}\cap G_{2}\neq\emptyset$, $G_{1}\neq\mathbb{R}^{n}\neq G_{2}$ and assume that there exists $c\in(0,1)$ such that $d(x,\partial G_{1})+d(x,\partial G_{2})\geq c\,d(x,\partial(G_{1}\cup G_{2})),$ for all $x\in G=G_{1}\cup G_{2}$. Suppose that $f:G\longrightarrow fG$ is continuous, $fG\subseteq\mathbb{R}^{n}$; that $\\{m_{D}\,:\,D\subseteq\mathbb{R}^{n}\\}$ is a monotone family of metrics; and that $m_{fG_{j}}(f(x),f(y))\leq\omega_{j}(j_{G_{j}}(x,y))$ for $x,y\in G_{j}$ and $j=1,2$. Then there exists $\omega:[0,\delta)\longrightarrow[0,+\infty)$, where $\delta=\log\left(1+\frac{c}{4}\right)$ such that (1.6.7) $m_{fG}(f(x),f(y))\leq\omega(j_{G}(x,y))$ for $x,y\in G$, $j_{G}(x,y)\leq\delta$ and $\lim_{t\rightarrow 0+}\omega(t)=0$ provided $\lim_{t\rightarrow 0+}\omega_{j}(t)=0$, $j=1,2$. ###### Proof. Let $d(x)=d(x,\partial G)$ and $j_{G}(x,y)\leq\delta$. Then, we have $\|x-y\|\leq\frac{c}{4}\min\\{d(x),d(y)\\}$. We may assume $d(x)\leq d(y)$. By the hypothesis (1.6.3) of the lemma there exists $i\in\\{1,2\\}$ such that $d(x,\partial G_{i})\geq\frac{c}{2}d(x)$, i.e., $B^{n}(x,c\,d(x)/2)\subseteq G_{i}$. Without loss of generality, we may assume that $i=1$. Then $y\in B^{n}(x,c\,\min\\{d(x),d(y)\\}/4)\subseteq B^{n}(x,\frac{1}{2}d(x,\partial G_{1})).$ We have $j_{G_{1}}(x,y)=\log\left(1+\frac{\|x-y\|}{\min\\{d_{1}(x),d_{1}(y)\\}}\right)$ where $d_{1}(z)=d(z,\partial G_{1})$. By the above calculation, $\|x-y\|\leq\frac{1}{2}\,d_{1}(x)$ and hence $d_{1}(y)\geq\frac{1}{2}\,d_{1}(x)$. The last inequality now yields $\begin{array}[]{rcl}j_{G_{1}}(x,y)&\leq&\displaystyle\log\left(1+\frac{2\|x-y\|}{d_{1}(x)}\right)\vspace{0.5em}\\\ &\leq&\displaystyle\log\left(1+\frac{4\|x-y\|}{c\,d(x)}\right)\vspace{0.5em}\\\ &<&\displaystyle\frac{4}{c}\,j_{G}(x,y).\end{array}$ Conclusion: (1.6.8) $m_{fG}(f(x),f(y))\leq m_{fG_{1}}(f(x),f(y))\leq\omega_{1}(j_{G_{1}}(x,y))\leq\omega_{1}\left(\frac{4}{c}\,j_{G}(x,y)\right)$ where also monotone property of the family $\\{m_{D}\\}$ was applied. ∎ ###### 1.6.9 Example. We present an example, due to J. Ferrand, of two domains $G_{1},G_{2}\subseteq\mathbb{C}$, $G_{1},G_{2}\neq\emptyset$ and an analytic function $f:H\longrightarrow\mathbb{C}$, $H=G_{1}\cup G_{2}$ such that 1. (1) (1.6.4) holds in $G_{1}$ and $G_{2}$. 2. (2) (1.6.5) does not hold on $H$. We set $G_{1}=\mathbb{C}\setminus\\{p+iq\,:\,p,q\in\mathbb{Z}\\}$ and $G_{2}=\mathbb{C}\setminus(\\{0\\}\cup\\{p+1/2+iq\,:\,p,q\in\mathbb{Z}\\})$. Note that $G_{1}\cap G_{2}\neq\emptyset$ and $G_{1}\cup G_{2}=H=\mathbb{C}\setminus\\{0\\}$. We define $f(\xi)=e^{4\pi\xi}$. This is an entire function. It is easy to see that $f(G_{1})=f(G_{2})=f(H)=H$. In fact $f(\Omega_{k})=H$, where $\Omega_{k}=\\{x+iy\,:\,k<y<k+1\\}$. Quasihyperbolic distance in $H$ satisfies (1.6.10) $k_{H}(w_{1},w_{2})=\inf_{e^{z_{1}}=w_{1},e^{z_{2}}=w_{2}}\|z_{1}-z_{2}\|.$ Also, for $i=1,2$ holds $d(\xi,\partial G_{i})\leqslant 1/2$, so the metric density $\frac{1}{d(\xi,\partial G_{i})}$ exceeds $\sqrt{2}$ and therefore (by a line integration) (1.6.11) $k_{G_{i}}(\xi_{1},\xi_{2})\geqslant\sqrt{2}\|\xi_{1}-\xi_{2}\|.$ Now, (1.6.10) tells us that $f:G_{i}\longrightarrow H$ is Lipschitz with respect to euclidean metric in $G_{i}$ and quasihyperbolic metric in $H$, and by (1.6.11) it is also Lipschitz with respect to quasihyperbolic metric in $H$ and $G_{i}$. But $f$ is not uniformly continuous as a map $(H,k_{H})\longrightarrow(H,k_{H})$: in fact we have $\lim_{n\rightarrow\infty}k_{H}(n,n+1)=\log\frac{n+1}{n}=0,$ while $\lim_{n\rightarrow\infty}k_{H}(f(n),f(n+1))=\log\frac{e^{4\pi(n+1)}}{e^{4\pi n}}=4\pi.$ Note that our domains fail to meet condition (1.6.10) from [Vu5, Lemma 2.27 ]. Indeed for large $\|x\|$ we have $d(x,\partial(G_{1}\cup G_{2}))=\|x\|$ and $d(x,\partial G_{1})+d(x,\partial G_{2})\leqslant 2\frac{1}{\sqrt{2}}=\sqrt{2},$ so there is no $c\in(0,1)$ such that (1.6.10) is valid. ###### 1.6.12 Remark. 1. (1) It would be of interest to find a homeomorphism $f$ with properties as in Example 1.6.9 due to Ferrand. 2. (2) It is natural to expect that there is a counterpart of Lemma 1.6.6 for other metrics in place of $j$. 3. (3) Is the condition (1.6.3) invariant under the quasiconformal mappings? ### 1.7. Quasiconformal maps with identity boundary values For a domain $G\subset\mathbb{R}^{n}$, $n\geqslant 2$, let $Id(\partial G)=\\{f\,:\,\overline{\mathbb{R}^{n}}\to\overline{\mathbb{R}^{n}}\mbox{ homeomorphism }:\,f(x)=x,\quad\forall x\in\overline{\mathbb{R}^{n}}\setminus G\\}.$ Here $\overline{\mathbb{R}^{n}}$ stands for the Möbius space $\mathbb{R}^{n}\cup\\{\infty\\}\,.$ We shall always assume that $card\\{\overline{\mathbb{R}^{n}}\setminus G\\}\geq 3.$ If $K\geqslant 1$, then the class of $K$-quasiconformal maps in $Id(\partial G)$ is denoted by $Id_{K}(\partial G)$. Here we use notation and terminology from Väisälä’s book [V2]. In particular, $K$-quasiconformal maps are defined in terms of the maximal dilatation as in [V2, p. 42] if not otherwise stated. We will study the following well-known problem: ###### 1.7.1 Problem. 1. (1) Given $a,b\in G$ and $f\in Id(\partial G)$ with $f(a)=b,$ find a lower bound for $K(f)$. 2. (2) Given $a,b\in G,$ construct $f\in Id(\partial G)$ with $f(a)=b$ and give an upper bound for $K(f)$. O. Teichmüller studied this problem in the case when $G$ is a plane domain with $card(\overline{\mathbb{R}^{2}}\setminus G)=3$ and proved the following theorem with a sharp bound for $K(f)$. ###### 1.7.2 Theorem. Let $G=\mathbb{R}^{2}\setminus\\{0,1\\}$, $a,b\in G$. Then there exists $f\in Id_{K}(\partial G)$ with $f(a)=b$ iff $\log(K(f))\geqslant s_{G}(a,b),$ where $s_{G}(a,b)$ is the hyperbolic metric of $G$. ###### 1.7.3 Theorem. If $f\in Id_{K}(\partial B^{n})$, then for all $x\in B^{n}$ $\rho_{B^{n}}(f(x),x)\leqslant\log\frac{1-a}{a},\quad a=\varphi_{1/K,n}(1/\sqrt{2})^{2},$ where $\varphi_{K,n}$ is as in (1.7.15). ###### 1.7.4 Theorem. If $f\in Id_{K}(\partial B^{n})$, then for all $x\in B^{n},n\geq 2,$ and $K\in[1,17]$ (1.7.5) $\|f(x)-x\|\leq\frac{9}{2}(K-1)\,.$ For $n=2$ we have (1.7.6) $\|f(x)-x\|\leqslant\frac{b}{2}(K-1),\quad b\leqslant 4.38.$ The theory of $K$-quasiregular mappings in ${\mathbb{R}}^{n},n\geq 3,$ with maximal dilatation $K$ close to $1\,$ has been extensively studied by Yu. G. Reshetnyak [R] under the name "stability theory". By Liouville’s theorem we expect that when $n\geq 3$ is fixed and $K\to 1$ the $K$-quasiregular maps "stabilize", become more and more like Möbius transformations, and this is the content of the deep main results of [R] such as [R, p. 286]. We have been unable to decide whether Theorem 1.7.3 follows from Reshetnyak’s stability theory in a simple way. V. I. Semenov [S] has also made significant contributions to this theory. For the plane case P. P. Belinskii has found several sharp results in [Bel]. ###### 1.7.7 Problem. It seems possible that there is a new kind of stability behavior: If $K>1$ is fixed, do maps in $Id_{K}(\partial B^{n})$ approach identity when $n\to\infty$? Our results do not answer this question. This kind of behavior is anticipated in [AVV, Open problem 9, p. 478]. ###### 1.7.8 Lemma. For $x,y\in B^{n}$ let $t=\sqrt{(1-\|x\|^{2})(1-\|y\|^{2})}$. Then for $x,y\in B^{n}$ (1.7.9) $\tanh^{2}\frac{\rho_{B^{n}}(x,y)}{2}=\frac{\|x-y\|^{2}}{\|x-y\|^{2}+t^{2}}\,,$ (1.7.10) $\|x-y\|\leqslant 2\tanh\frac{\rho_{B^{n}}(x,y)}{4}=\frac{2\|x-y\|}{\sqrt{\|x-y\|^{2}+t^{2}}+t}\,,$ where equality holds for $x=-y$. Next, we consider a decreasing homeomorphism $\mu:(0,1)\longrightarrow(0,\infty)$ defined by (1.7.11) $\mu(r)=\frac{\pi}{2}\,\frac{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(r^{\prime})}{{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(r)},\quad{\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(r)=\int_{0}^{1}\frac{dx}{\sqrt{(1-x^{2})(1-r^{2}x^{2})}}\,,$ where ${\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(r)$ is Legendre’s complete elliptic integral of the first kind and $r^{\prime}=\sqrt{1-r^{2}},$ for all $r\in(0,1)$. The Hersch- Pfluger distortion function is an increasing homeomorphism $\varphi_{K}:(0,1)\longrightarrow(0,1)$ defined by (1.7.12) $\varphi_{K}(r)=\mu^{-1}(\mu(r)/K)$ for all $r\in(0,1)$, $K>0$. By continuity we set $\varphi_{K}(0)=0$, $\varphi_{K}(1)=1$. From (1.7.11) we see that $\mu(r)\mu(r^{\prime})=\left(\frac{\pi}{2}\right)^{2}$ and from this we are able to conclude a number of properties of $\varphi_{K}$. For instance, by [AVV, Thm 10.5, p. 204] (1.7.13) $\varphi_{K}(r)^{2}+\varphi_{1/K}(r^{\prime})^{2}=1,\quad r^{\prime}=\sqrt{1-r^{2}},$ holds for all $K>0$, $r\in(0,1)$. 7.14. Special function $\varphi_{K,n}$ We use the standard notation (1.7.15) $\varphi_{K,n}(r)=\frac{1}{\gamma_{n}^{-1}(K\gamma_{n}(1/r))}.$ Then $\varphi_{K,n}:(0,1)\longrightarrow(0,1)$ is an increasing homeomorphism, see [Vu2, (7.44)]. Because $\gamma_{2}(1/r)=2\pi/\mu(r)$ by [Vu2, (5.56)], it follows that $\varphi_{K,2}(r)$ is the same as the $\varphi_{K}(r)$ in (1.7.12). 7.16. The key constant. The special functions introduced above will have a crucial role in what follows. For the sake of easy reference we give here some well-known identities between them that can be found in [AVV]. First, the function (1.7.17) $\eta_{K,n}(t)=\tau_{n}^{-1}(\tau_{n}(t)/K)=\frac{1-\varphi_{1/K,n}(1/\sqrt{1+t})^{2}}{\varphi_{1/K,n}(1/\sqrt{1+t})^{2}},\,K>0\,,$ defines an increasing homeomorphism $\eta_{K,n}:(0,\infty)\to(0,\infty)\,$(cf. [AVV, p.193]). The constant $(1-a)/a,a=\varphi_{1/K,n}(1/\sqrt{2})^{2},$ in (1.7.5) can be expressed as follows for $K>1$ (1.7.18) $(1-a)/a=\eta_{K,n}(1)=\tau_{n}^{-1}(\tau_{n}(1)/K)\,.$ Furthermore, by (1.7.13) (1.7.19) $\eta_{K,2}(t)=\frac{s^{2}}{1-s^{2}},\quad s=\varphi_{K,2}(\sqrt{t/(1+t)})\,$ and (1.7.20) $\eta_{K,2}(1)\in(e^{\pi(K-1)},e^{b(K-1)})$ where $b=(4/\pi){\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(1/\sqrt{2})^{2}=4.376879...$ Note that the constant $\lambda(K)$ in [AVV, 10.33 p. 218.] is the same as $\eta_{K,2}(1)\,.$ For the proof of Lemma 1.7.29, we record a lower bound for $\varphi_{1/K,n}(r)\,.$ The constant $\lambda_{n}$ is the so called Grötzsch ring constant, see [AVV]. ###### 1.7.21 Lemma. ([Vu2, 7.47, 7.50]) For $n\geq 2,K\geq 1,$ and $0\leq r\leq 1$ (1.7.22) $\varphi_{1/K,n}(r)\geq\lambda_{n}^{1-\beta}r^{\beta},\,\,\beta=K^{1/(n-1)},$ (1.7.23) $\lambda_{n}^{1-\beta}\geq 2^{1-\beta}K^{-\beta}\geq 2^{1-K}K^{-K}\,.$ ###### 1.7.24 Lemma. 1. (1) For all $m,n\geqslant 1$ there is $M>1$ such that the inequality (1.7.25) $\log(2^{mx-m+1}x^{nx}-1)\leqslant(2m\log 2+2n)(x-1)$ holds for $x\in[1,M]$ with equality only for $x=1$. Moreover, with $t=(m\log 2-n)/(2n)\,,$ $M$ can be chosen as $M=\sqrt{\frac{(m-1)\log 2+\log\left(1+\frac{(n+m\log 2)^{2}}{n}\right)}{n}+t^{2}}-t.$ 2. (2) Let $p(x)=\log(2^{mx-m+1}x^{nx}-1)$, $q(x)=(2m\log 2+2n)(x-1)$ and let us use the above notation. Let $a_{0}=M$ and $a_{n+1}=p^{-1}(q(a_{n}))$ for $n\geqslant 1$. Then the sequence $a_{n}$ is increasing and bounded. If $a=\lim_{n\rightarrow}a_{n}$ then the inequality (1.7.25) holds for $x\in[1,a]$ with equality iff $x\in\\{1,a\\}$. For $m=3$ and $n=2$ we have $a>17$. ###### Proof. Let $u(x)=(mx-m+1)\log 2+nx\log x,\quad v(x)=\log(e^{u(x)}-1)=\log(2^{mx-m+1}x^{nx}-1).$ Then we have $\begin{array}[]{rcl}v^{\prime\prime}(x)&=&\displaystyle(\log(e^{u(x)}-1))^{\prime\prime}=\left(\frac{u^{\prime}(x)\,e^{u(x)}}{e^{u(x)}-1}\right)^{\prime}\vspace{1em}\\\ &=&\displaystyle\frac{(u^{\prime\prime}(x)e^{u(x)}+(u^{\prime}(x))^{2}e^{u(x)})(e^{u(x)}-1)-(u^{\prime}(x)\,e^{u(x)})^{2}}{(e^{u(x)}-1)^{2}}\vspace{1em}\\\ &=&\displaystyle\frac{e^{u(x)}}{(e^{u(x)}-1)^{2}}\cdot((u^{\prime\prime}(x)+(u^{\prime}(x))^{2})(e^{u(x)}-1)-(u^{\prime}(x))^{2}e^{u(x)})\vspace{1em}\\\ &=&\displaystyle\frac{e^{u(x)}}{(e^{u(x)}-1)^{2}}\cdot(u^{\prime\prime}(x)(e^{u(x)}-1)-(u^{\prime}(x))^{2}).\end{array}$ Thus $v^{\prime\prime}(x)\leqslant 0\,\,\,\Leftrightarrow\,\,\,u^{\prime\prime}(x)(e^{u(x)}-1)\leqslant(u^{\prime}(x))^{2}.$ Since $e^{u(x)}=2^{mx-m+1}x^{nx},\quad u^{\prime}(x)=n+m\log 2+n\log x,\quad u^{\prime\prime}(x)=\frac{n}{x},$ we have $v^{\prime\prime}(x)\leqslant 0\,\,\,\Leftrightarrow\,\,\,\frac{n}{x}(2^{mx-m+1}x^{nx}-1)\leqslant(n+m\log 2+n\log x)^{2},$ therefore $v^{\prime\prime}(x)\leqslant 0$ is for $x\geqslant 1$ equivalent to $2^{mx-m+1}x^{nx}-1\leqslant\frac{x}{n}(n+m\log 2+n\log x)^{2}.$ Let $f(x)=2^{mx-m+1}x^{nx}-1$ and $g(x)=\frac{x}{n}(n+m\log 2+n\log x)^{2}$. Both functions $f$ and $g$ are increasing on $[1,+\infty)$ and $f(1)<g(1)$ because $f(1)=1\leqslant n=\frac{1}{n}\cdot n^{2}<\frac{1}{n}(n+m\log 2)^{2}=g(1).$ By continuity of $f$ we can conclude that there is $M>1$ such that $f(M)\leqslant g(1)$. For such $M$ $f(x)\leqslant f(M)\leqslant g(1)\leqslant g(x),\quad x\in[1,M].$ This implies that $v$ is concave on $[1,M]$ and consequently $v(x)\leqslant v(1)+v^{\prime}(1)(x-1),\quad x\in[1,M]$ i.e. $\log(2^{mx-m+1}x^{nx}-1)\leqslant(2m\log 2+2n)(x-1),\quad x\in[1,M].$ The inequality $f(x)\leqslant g(1)$ is equivalent to (1.7.26) $(mx-m+1)\log 2+nx\log x\leqslant\log\left(1+\frac{(n+m\log 2)^{2}}{n}\right).$ Because (1.7.27) $(mx-m+1)\log 2+nx\log x\leqslant(mx-m+1)\log 2+nx(x-1)$ the inequality (1.7.26) is the consequence of the inequality (1.7.28) $(mx-m+1)\log 2+nx(x-1)\leqslant\log\left(1+\frac{(n+m\log 2)^{2}}{n}\right).$ In (1.7.27) equality holds only for $x=1$. Because $1+\frac{(n+m\log 2)^{2}}{n}>1+\frac{n^{2}}{n}=1+n\geqslant 2$ the inequality (1.7.28) is a strict inequality for $x=1$. By this reason, the greater root of the quadratic equation $(mx-m+1)\log 2+nx(x-1)=\log\left(1+\frac{(n+m\log 2)^{2}}{n}\right)$ is greater than $1$. If we denote this root with $M$ the inequality (1.7.26) holds for $x\in[1,M]$ with equality only for $x=1$. The first part of Lemma is proved. Now we prove the second part of the inequality. Both of functions $p(x)$ and $q(x)$ are continuous and increasing. Consequently $r(x)=p^{-1}(x)$ is continuous and increasing. Because $p(a_{1})=q(a_{0})>p(a_{0})$ using monotonicity of $p(x)$ we can conclude that $a_{1}>a_{0}$. Now, by induction and monotonicity of $r$ we can conclude that the sequence $a_{n}$ is increasing. Now for $x\in[a_{n},a_{n+1})$ we have $p(x)<p(a_{n+1})=q(a_{n})\leqslant q(x).$ So $p(x)<q(x)$ holds for $x\in\bigcup_{n=0}^{\infty}[a_{n},a_{n+1})=[a_{0},a)$ and using already proved inequality, $p(x)<q(x)$ holds for $1<x<a$. For $x\geqslant 1$ holds $mx-m+1>1$ and $x^{nx}\geqslant 1$ and consequently $p(x)=\log(2^{mx-m+1}x^{nx}-1)>\log(2\,x^{nx}-1)\geqslant nx\log x.$ Because $p(x)>nx\log x\geqslant(n\log x)(x-1)$ inequality $p(c)>q(c)$ holds for $c$ such that $n\log c\geqslant 2m\log 2+2n$. It is easy to see that it is true for $c=2^{\frac{2m}{n}}e^{2}$. It implies that $a$ is finite (for example $a<2^{\frac{2m}{n}}e^{2}$) and $a_{n}$ is bounded. Letting $n\rightarrow\infty$ in $p(a_{n+1})=q(a_{n})$ and using continuity of both functions we conclude that $p(a)=q(a)\,.$ ∎ ###### 1.7.29 Lemma. If $a=\varphi_{1/K,n}(1/\sqrt{2})^{2}$ is as in Theorem 1.7.3 then for $M>1$ and $\beta\in[1,M]$ (1.7.30) $\log\left(\frac{1-a}{a}\right)\leq\log(\lambda_{n}^{2(\beta-1)}2^{\beta}-1)\leq V(n)(\beta-1)$ with $V(n)=(2\log(2\lambda_{n}^{2}))(2\lambda_{n}^{2})^{M-1}$ and for $K\in[1,17]$, (1.7.31) $\log\left(\frac{1-a}{a}\right)\leqslant(K-1)(4+6\log 2)<9(K-1),\quad$ with equality only for $K=1$. For $n=2$ (1.7.32) $\log\left(\frac{1-a}{a}\right)=\log\left(\frac{\varphi_{K,2}(1/\sqrt{2})^{2}}{\varphi_{1/K,2}(1/\sqrt{2})^{2}}\right)\leqslant b(K-1)$ where $b=(4/\pi){\mathchoice{\hbox{\,\fFt K}}{\hbox{\,\fFt K}}{\hbox{\,\fFa K}}{\hbox{\,\fFp K}}}(1/\sqrt{2})^{2}\leq 4.38\,.$ ###### Proof. For $\beta\in[1,M]$ we have by (1.7.22) $\log\left(\frac{1-a}{a}\right)\leq\log(\lambda_{n}^{2(\beta-1)}2^{\beta}-1)\,.$ Further, we have $\frac{\log(\lambda_{n}^{2(\beta-1)}2^{\beta}-1)}{\beta-1}\leqslant 2\,\frac{(2\lambda_{n}^{2})^{\beta-1}-1}{\beta-1}\leqslant(2\log(2\lambda_{n}^{2}))(2\lambda_{n}^{2})^{M-1}.$ The second inequality follows from the inequality $\log(t)\leqslant t-1$ and the third one from Lagrange’s theorem and monotonicity of the function $(2\log(2\lambda_{n}^{2}))(2\lambda_{n}^{2})^{x-1}$. This proves (1.7.30). From (1.7.23) it follows that the constant $a$ satisfies the inequality $a\geq 2^{2(1-K)}K^{-2K}(1/\sqrt{2})^{2K}\,$ and also $1/a\leq 2^{3K-2}K^{2K}\,,\quad K>1.$ By Lemma 1.7.24 we have $\log(2^{3K-2}K^{2K}-1)\leqslant(4+6\log 2)(K-1)$ for $K\in[1,17]$ with equality only for $K=1$. Now, from $\frac{1-a}{a}<2^{3K-2}K^{2K}-1,\quad K>1$ we conclude that $\log\left(\frac{1-a}{a}\right)\leqslant(4+6\log 2)(K-1)<9(K-1)\,.$ For the case $n=2$ we can apply the identity (1.7.19) and the inequality in (1.7.20). ∎ Figure 1.5. 7.33. Proof of Theorem 1.7.3. Fix $x\in B^{n}$ and let $T_{x}$ denote a Möbius transformation of $\overline{\mathbb{R}^{n}}$ with $T_{x}(B^{n})=B^{n}$ and $T_{x}(x)=0$. Define $g:\mathbb{R}^{n}\longrightarrow\mathbb{R}^{n}$ by setting $g(z)=T_{x}\circ f\circ T^{-1}_{x}(z)$ for $z\in B^{n}$ and $g(z)=z$ for $z\in\mathbb{R}^{n}\setminus B^{n}$. Then $g\in Id_{K}(\partial B^{n})$with $g(0)=T_{x}(f(x))$. By the invariance of $\rho_{B^{n}}$ under the group ${\mathcal{GM}}(B^{n})$ of Möbius selfautomorphisms of $B^{n}$ we see that for $x\in B^{n}$ (1.7.34) $\rho_{B^{n}}(f(x),x)=\rho_{B^{n}}(T_{x}(f(x)),T_{x}(x))=\rho_{B^{n}}(g(0),0).$ Choose $z\in\partial B^{n}$ such that $g(0)\in[0,z]=\\{tz\,:\,0\leqslant t\leqslant 1\\}$. Let $E^{\prime}=\\{-sz\,:\,s\geqslant 1\\}$, $\Gamma^{\prime}=\Delta([g(0),z],E^{\prime};\mathbb{R}^{n})$ and $\Gamma=\Delta(g^{-1}[g(0),z],g^{-1}E^{\prime};\mathbb{R}^{n})$. The spherical symmetrization with center at $0$ yields by [AVV, Thm 8.44] $M(\Gamma)\geqslant\tau_{n}(1)\quad(=2^{1-n}\gamma_{n}(\sqrt{2}))$ because $g(x)=x$ for $x\in\mathbb{R}^{n}\setminus B^{n}$. Next, we see by the choice of $\Gamma^{\prime}$ that $M(\Gamma^{\prime})=\tau_{n}\left(\frac{1+\|g(0)\|}{1-\|g(0)\|}\right).$ By $K$-quasiconformality we have $M(\Gamma)\leqslant K\,M(\Gamma^{\prime})$ implying (1.7.35) $\exp(\rho_{B^{n}}(0,g(0)))=\frac{1+\|g(0)\|}{1-\|g(0)\|}\leqslant\tau_{n}^{-1}(\tau_{n}(1)/K)=\frac{1-a}{a}.$ The last equality follows from (1.7.18). Finally, (1.7.34) and (1.7.35) complete the proof. $\hfill\square$ 7.36. Proof of Theorem 1.7.4. We have $\begin{array}[]{rcl}\|f(x)-x\|&\leqslant&\displaystyle 2\tanh\left(\frac{\rho_{B^{n}}(f(x),x)}{4}\right)\leqslant 2\tanh\left(\frac{\log\left(\frac{1-a}{a}\right)}{4}\right)\vspace{1em}\\\ &\leqslant&\displaystyle 2\tanh\left(\frac{(K-1)(4+6\log 2)}{4}\right)\vspace{1em}\\\ &\leqslant&\displaystyle(K-1)(2+3\log 2)\leqslant\frac{9}{2}(K-1).\end{array}$ The first inequality follows from (1.7.10), the second one from Theorem 1.7.3, the third one from Lemma 1.7.29 and the last one from inequality $\tanh(t)\leqslant t$ for $t\geqslant 0$. For $n=2$ we use the same first two steps and planar case of Lemma 1.7.29 to derive inequality $\|f(x)-x\|\leqslant\frac{b}{2}(K-1).\quad\quad\hfill\square$ A lower bound corresponding to the upper bound in (1.7.5) is given in the next lemma. ###### 1.7.37 Lemma. For $f\in Id(\partial G)$ let $\delta(f)\equiv\sup\\{\|f(z)-z\|:z\in G\\}\,.$ Then for $f\in Id_{K}(\partial B^{n}),K>1,\alpha=K^{1/(1-n)}$ (1.7.38) $\delta(f)\geq(1-\alpha)\alpha^{\alpha/(1-\alpha)}>\frac{1}{e}(1-\alpha).$ ###### Proof. The radial stretching $f:B^{n}\to B^{n},n\geq 2,$ defined by $f(z)=\|z\|^{\alpha-1}\,z,z\in B^{n},$ ($0<\alpha<1$) is $K$-qc with $\alpha=K^{1/(1-n)}$ [V2, p. 49] and $f\in Id_{K}(\partial B^{n})\,.$ Now we have $\|f(z)-z\|=\|\|z\|^{\alpha-1}z-z\|=\|r^{\alpha}-r\|,\quad\|z\|=r.$ Further, we see that $\delta(f)=\sup_{0<r<1}(r^{\alpha}-r),$ where the supremum is attained for $r=r_{\alpha}=\left(\frac{1}{\alpha}\right)^{\frac{1}{\alpha-1}}$, so $\delta(f)=({1}-{\alpha})\alpha^{\alpha/(1-\alpha)}\,.$ A crude, but simple, estimate is $\delta(f)\geq(1/e)^{\alpha}-(1/e)=\frac{1}{e}\left(\frac{1}{e^{\alpha-1}}-1\right)=\frac{1}{e}\left(e^{1-\alpha}-1\right)\geqslant\frac{1}{e}(1-\alpha)\,.$ ∎ Figure 1.6. ###### 1.7.39 Theorem. Let $f:\overline{\mathbb{R}^{n}}\longrightarrow\overline{\mathbb{R}^{n}}$ be a $K$-qc homeomorphism with $f(\infty)=\infty$ and $B^{n}(m)\subset f(B^{n})\subset B^{n}(M)$ where $0<m\leq 1\leq M$. Then $\eta_{1/K,n}\left(\frac{1+\|x\|}{1-\|x\|}\right)\leqslant\frac{M+\|f(x)\|}{m-\|f(x)\|}$ and $\frac{m+\|f(x)\|}{M-\|f(x)\|}\leqslant\eta_{K,n}\left(\frac{1+\|x\|}{1-\|x\|}\right)$ for all $x\in B^{n}$ where $\eta_{K,n}(t)=\tau_{n}^{-1}(\tau_{n}(t)/K)$. In particular, if $m=1=M$, then we have $\eta_{1/K,n}\left(\frac{1+\|x\|}{1-\|x\|}\right)\leqslant\frac{1+\|f(x)\|}{1-\|f(x)\|}\leqslant\eta_{K,n}\left(\frac{1+\|x\|}{1-\|x\|}\right)\,.$ Figure 1.7. ###### Proof. The proof is similar to the proof of Theorem 1.7.3. Fix $x\in B^{n}$ and choose $z^{\prime}\in\partial f(B^{n})$ such that $f(x)\in[0,z^{\prime}]$ and $[f(x),z^{\prime})\subset f(B^{n})\,$ and fix $z"\in\partial f(B^{n})$ such that $z^{\prime},0,z"$ are on the same line, $0\in[z^{\prime},z"],$ and $\\{-sz"\,:\,s\geqslant 1\\}\subset{\mathbb{R}}^{n}\setminus f(B^{n})\,.$ Let $\Gamma^{\prime}=\Delta([f(x),z^{\prime}],E^{\prime};\mathbb{R}^{n})$, $E^{\prime}=\\{-sz"\,:\,s\geqslant 1\\}$ and $\Gamma=\Delta(f^{-1}[f(x),z^{\prime}],f^{-1}E^{\prime};\mathbb{R}^{n})$. Then $M(\Gamma^{\prime})\leq\tau_{n}\left(\frac{m+\|f(x)\|}{M-\|f(x)\|}\right)$ while applying a spherical symmetrization with center at the origin gives $M(\Gamma)\geqslant\tau_{n}\left(\frac{1+\|x\|}{1-\|x\|}\right)$ because $f^{-1}E^{\prime}$ connects $\partial B^{n}$ and $\infty$. Then the inequality $M(\Gamma)\leqslant K\,M(\Gamma^{\prime})$ yields $\tau_{n}\left(\frac{1+\|x\|}{1-\|x\|}\right)\leq K\tau_{n}\left(\frac{m+\|f(x)\|}{M-\|f(x)\|}\right),$ $\tau_{n}^{-1}(\frac{1}{K}\tau_{n}\left(\frac{1+\|x\|}{1-\|x\|}\right))\geq\frac{m+\|f(x)\|}{M-\|f(x)\|}$ (1.7.40) $\frac{m+\|f(x)\|}{M-\|f(x)\|}\leqslant\eta_{K,n}\left(\frac{1+\|x\|}{1-\|x\|}\right).$ The lower bound follows if we apply a similar argument to $f^{-1}$ and the lower bound $M(\Gamma^{\prime})\geq\tau_{n}\left(\frac{M+\|f(x)\|}{m-\|f(x)\|}\right)\,.$ ∎ 7.41. Remark. Putting $x=0,m=1=M$ in (1.7.40) we obtain by (1.7.18) for a $K$-qc homeomorphism $f:\overline{\mathbb{R}^{n}}\longrightarrow\overline{\mathbb{R}^{n}}$ with $f(\infty)=\infty$ and $f(B^{n})=B^{n}$ that $\|f(0)\|\leq 1-2a\,,a=\varphi_{1/K,n}(1/\sqrt{2})^{2}\,.$ Further, if we use the lower bound (1.7.23) from Lemma 1.7.21 we obtain $\|f(0)\|\leq 1-2^{1-\beta}4^{1-K}K^{-2K}\,.$ In the special case when $n=2$ we have $\|f(0)\|\leq 1-2^{3(1-K)}K^{-2K}\leq(2+3\log 2)(K-1)\,.$ Note that this last inequality does not suppose that $f\in Id_{K}(\partial B^{n})\,,$ only the hypotheses of Theorem 1.7.39 are needed. 7.42. Maps of cylinder We next consider the class $Id_{K}(\partial Z)$ for the case when the domain $Z$ is an infinite cylinder. ###### 1.7.43 Theorem. Let $Z=\\{(x,t)\in\mathbb{R}^{n}\,:\,\|x\|<1,\,t\in\mathbb{R}\\}$, $f\in Id_{K}(\partial Z)$. Then $k_{Z}(0,f(0))\leqslant c(K)$ where $c(K)\rightarrow 0$ when $K\rightarrow 1$. Figure 1.8. ###### Proof. Let $f(0)=(y,t)$, $E^{\prime}=[w,f(0)]$, $F^{\prime}=\\{\overline{w}+s(y,0)\,:\,s\leqslant 0\\}$ where $w=(y/\|y\|,t)$, $\overline{w}=(-y/\|y\|,t)$. Then $E^{\prime}$ and $F^{\prime}$ are the complementary components of a Teichmüller ring and therefore writing $\Gamma^{\prime}=\Delta(E^{\prime},F^{\prime};\mathbb{R}^{n})$ we have $M(\Gamma^{\prime})\leqslant\tau_{n}\left(\frac{1+\|y\|}{1-\|y\|}\right).$ The modulus of the family $\Gamma=\Delta(E,F;\mathbb{R}^{n})$, $E=f^{-1}E^{\prime}$, $F=f^{-1}F^{\prime}$ can be estimated by use of spherical symmetrization with the center at 0. Note that $E=E^{\prime}$ because $E^{\prime}\subset\mathbb{R}^{n}\setminus Z$ and $f\in Id_{K}(\partial Z)$. By [Vu2, 7.34] we have $M(\Gamma)\geqslant\tau_{n}(1).$ By $K$-quasiconformality $M(\Gamma)\leqslant K\,M(\Gamma^{\prime})$ implying $\exp(\rho_{B^{n-1}}(0,y))=\frac{1+\|y\|}{1-\|y\|}\leqslant\tau_{n}^{-1}\left(\frac{\tau_{n}(1)}{K}\right).$ Figure 1.9. Next we shall estimate $t$. Fix first $z$ in $\\{w\in\partial Z\,:\,w_{n}=0\\}$ such that $\|f(0)-z\|$ is maximal. Then choose a point $w$ on the line through $f(0)$ and $z$ such that $\|z-w\|=1$ and $[z,w]\subset\mathbb{R}^{n}\setminus Z$. Let $E^{\prime}=[z,w]$ and $F^{\prime}=\\{f(0)+t(f(0)-z)\,:\,t\geqslant 0\\}$. Then $E^{\prime}$ and $F^{\prime}$ are the complementary components of a Teichmüller ring and with $\Delta^{\prime}=\Delta(E^{\prime},F^{\prime};\mathbb{R}^{n})$ we have $M(\Delta^{\prime})=\tau_{n}(\|f(0)-z\|).$ Observing that $E^{\prime}=f^{-1}E^{\prime}$, because $f\in Id_{K}(\partial Z)$ and carrying out a spherical symmetrization with center at $z$ we see that if $E=f^{-1}E^{\prime}$, $F=f^{-1}F^{\prime}$ then $M(\Delta)\geqslant\tau_{n}(1),\quad\Delta=\Delta(E,F;\mathbb{R}^{n}).$ By $K$-quasiconformality we have $1+t^{2}\leqslant\|f(0)-z\|^{2}\leqslant\tau_{n}^{-1}\left(\frac{\tau_{n}(1)}{K}\right)^{2}.$ The triangle inequality for $k_{Z}$ yields $\begin{array}[]{rcl}k_{Z}(0,f(0))&\leqslant&k_{Z}(0,(0,t))+k_{Z}((0,t),(y,t))\vspace{0.5em}\\\ &=&t+k_{B^{n-1}}(0,y)\leqslant\|t\|+2\,\rho_{B^{n-1}}(0,y)\vspace{0.5em}\\\ &\leqslant&\sqrt{\tau_{n}^{-1}\left(\frac{\tau_{n}(1)}{K}\right)^{2}-1}+2\log\left(\tau_{n}^{-1}\left(\frac{\tau_{n}(1)}{K}\right)\right)\vspace{0.5em}\\\ &\leqslant&\displaystyle\sqrt{e^{18(K-1)}-1}+18(K-1).\end{array}$ The last inequality follows from (1.7.18) and Lemma 1.7.29. ∎ ### 1.8. Distortion of two point normalized quasiconformal mappings Let $\eta\colon[0,\infty)\to[0,\infty)$ be an increasing homeomorphism and $D,D^{\prime}\subset\mathbb{R}^{n}$. A homeomorphism $f\colon D\to D^{\prime}$ is _$\eta$ -quasisymmetric_ if (1.8.1) $\frac{\|f(a)-f(c)\|}{\|f(b)-f(c)\|}\leq\eta\left(\frac{\|a-c\|}{\|b-c\|}\right)$ for all $a,b,c\in D$ and $c\neq b$. By [V2] K-quasiconformal mapping of the whole $\mathbb{R}^{n}$ is $\eta_{K,n}$\- quasisymmetric with a control function $\eta_{K,n}$. Let us define the optimal control function by $\eta^{}_{K,n}(t)=\sup\\{\|f(x)\|\colon\|x\|\leq t,f\in QC_{K}(\mathbb{R}^{n}),f(y)=y\textnormal{ for }y\in\\{0,e_{1},\infty\\}\\}.$ Vuorinen [Vu2, Theorem 1.8] proved an upper bound for $\eta^{}_{K,n}(t)$, which was later refined by Prause [P, Theorem 2.7] for $K<4/3$ into the following form (1.8.2) $\eta^{}_{K,n}(t)\leq\left\\{\begin{array}[]{ll}\displaystyle\eta^{}_{K,n}(1)\varphi_{K,n}(t),&0<t<1,\\\ \displaystyle 1+600\left((K-1)\log\frac{1}{K-1}\right),&t=1,\\\ \displaystyle\eta^{}_{K,n}(1)\frac{1}{\varphi_{1/K,n}(1/t)},&t>1,\end{array}\right.$ where (1.8.3) $\eta^{}_{K,n}(1)\leq\exp((4\sqrt{2}-\log(K-1))(K^{2}-1)).$ We also introduce a simpler estimate of $\eta^{}_{K,n}(1)$ from [AVV, Theorem 14.8] (1.8.4) $\eta^{}_{K,n}(1)\leq\exp(4K(K+1)\sqrt{K-1}).$ A more rough upper bound for $\eta^{}_{K,n}(t)$ can [Vu1, Theorem 7.47] be written as (1.8.5) $\eta^{}_{K,n}(t)\leq\left\\{\begin{array}[]{ll}\displaystyle\eta^{}_{K,n}(1)\lambda_{n}^{1-\alpha}t^{\alpha},&0<t\leq 1,\\\ \displaystyle\eta^{}_{K,n}(1)\lambda_{n}^{1-\beta}t^{\beta},&t>1,\end{array}\right.$ where $\alpha=K^{1/(1-n)}$ and $\beta=1/\alpha$. Furthermore, we can [Vu1, Lemma 7.50] estimate (1.8.6) $\lambda_{n}^{1-\alpha}\leq 2^{1-1/K}K\quad\textnormal{and}\quad\lambda_{n}^{1-\beta}\leq 2^{1-K}K^{-K}.$ ###### 1.8.7 Lemma. Let $K\in(1,2]$, $f\in QC_{K}(\mathbb{R}^{n})$, $f(x)=x$ for $x\in\\{0,e_{1}\\}$, $\alpha=K^{1/(1-n)}$ and $\beta=1/\alpha$. Then $\begin{array}[]{ll}\displaystyle\frac{1}{c_{3}}\|x\|^{\beta}\leq\|f(x)\|\leq c_{3}\|x\|^{\alpha},&\textnormal{if }0<\|x\|\leq 1,\\\ \displaystyle\frac{1}{c_{3}}\|x\|^{\alpha}\leq\|f(x)\|\leq c_{3}\|x\|^{\beta},&\textnormal{if }\|x\|>1\end{array}$ for $c_{3}=\exp(60\sqrt{K-1})$. ###### Proof. Since $f$ is quasiconformal it is also $\eta^{}_{K,n}$-quasisymmetric and by choosing $a=x$, $b=0$ and $c=e_{1}$ in (1.8.1) we have $\|f(x)\|\leq\eta^{}_{K,n}(\|x\|)$. Similarly, selection $(a,b,c)=(e_{1},0,x)$ in (1.8.1) gives $\|f(x)\|\geq 1/\eta^{}_{K,n}(1/\|x\|)$. Therefore (1.8.8) $\frac{1}{\eta^{}_{K,n}(1/\|x\|)}\leq\|f(x)\|\leq\eta^{}_{K,n}(\|x\|)$ for all $x\in\overline{\mathbb{R}}^{n}\setminus\\{0\\}$. Therefore by (1.8.5) $\begin{array}[]{ll}\displaystyle\frac{1}{c_{2}}\|x\|^{\beta}\leq\|f(x)\|\leq c_{1}\|x\|^{\alpha},&\textnormal{if }0<\|x\|<1,\\\ \displaystyle\frac{1}{\eta^{}_{K,n}(1)}\leq\|f(x)\|\leq\eta^{}_{K,n}(1),&\textnormal{if }\|x\|=1,\\\ \displaystyle\frac{1}{c_{1}}\|x\|^{\alpha}\leq\|f(x)\|\leq c_{2}\|x\|^{\beta},&\textnormal{if }\|x\|>1,\\\ \end{array}$ for $c_{1}=\eta^{}_{K,n}(1)\lambda_{n}^{1-\alpha}$ and $c_{2}=\eta^{}_{K,n}(1)\lambda_{n}^{1-\beta}$. We can estimate $\max\\{c_{1},c_{2}\\}\leq c_{3}=\exp(60\sqrt{K-1})$ for $K\in(1,2]$. ∎ We will consider $K$\- quasiconformal mapping $f\colon\overline{\mathbb{R}}^{n}\to\overline{\mathbb{R}}^{n}$ with $f(y)=y$ for $y\in\\{0,e_{1},\infty\\}$ and our goal is to find an upper bound for $\|f(x)-x\|$ or similar quantities in terms of $K$ and $n$, when $\|x\|\leq 2$ and $K>1$ is small enough. Fix $x\in\mathbb{R}^{n}\setminus\\{0,e_{1}\\}$ and assume that $\|x\|-\varepsilon\leq\|f(x)\|\leq\|x\|+\varepsilon$ and $\|x-e_{1}\|-\varepsilon\leq\|f(x)-e_{1}\|\leq\|x-e_{1}\|+\varepsilon$ for $\varepsilon\in(0,\min\\{\|x\|,\|x-e_{1}\|\\})$. Now (1.8.9) $\|f(x)-x\|\leq\frac{\textnormal{diam}\,(A)}{2},$ where $A=A(0,\|x\|+\varepsilon,\|x\|-\varepsilon)\cap A(e_{1},\|x-e_{1}\|+\varepsilon,\|x-e_{1}\|-\varepsilon)\cap\\{z\in R^{3}\colon z_{3}=0\\}$ and $A(z,R,r)=B^{n}(z,R)\setminus\overline{B}^{n}(z,r).$ We will now find upper bounds for $\textnormal{diam}\,(A)$. Figure 1.10. The set $A$. ###### 1.8.10 Theorem. For $\varepsilon<1$ and $A$ and $x$ as in (1.8.9) $\textnormal{diam}\,(A)\leq\sqrt{\varepsilon}4(\min\\{\|x\|,\|x-e_{1}\|\\}+1).$ ###### Proof. Let us assume $\|x\|\leq\|x-e_{1}\|$. Now $\textnormal{diam}\,(A)$ is maximal when $\|x-e_{1}\|/\|x\|$ is maximal. Therefore we may assume $x=-s$, where $s>0$. Denote ${y}=S^{1}(0,\|x\|+\varepsilon)\cap S^{1}(e_{1},\|x-e_{1}\|-\varepsilon)$. Area of the triangle $\triangle e_{1}0y$ is $(\textnormal{Im}\,y)/2$ and by Heron’s formula (1.8.11) $\frac{\textnormal{Im}\,y}{2}=\sqrt{p(p-1)(p-\|x\|-\varepsilon)(p+\varepsilon-\|x\|-1)},$ where $p=\|x\|+1$. By (1.8.11) and the assumption $\varepsilon<1$ we have $\textnormal{Im}\,y=2\sqrt{(\|x\|+1)\|x\|(1-\varepsilon)\varepsilon}\leq 2\sqrt{\varepsilon}(\|x\|+1)$ and the assertion follows since $\textnormal{diam}\,(A)\leq 2\textnormal{Im}\,y$. ∎ ###### 1.8.12 Theorem. Let $A$ be as in (1.8.9), $\|x\|<2$, $\|x-e_{1}\|\leq\|x\|$ and $\measuredangle(1,0,x)\geq\omega>0$. Then $\textnormal{diam}\,(A)\leq\varepsilon\left(1+\frac{70}{\omega}\right)$ for $\varepsilon<\min\left\\{1,\frac{1+\|x-e_{1}\|-\|x\|}{2},\frac{\|x\|+\|x-e_{1}\|-1}{2}\right\\}.$ ###### Proof. Let us denote by $y$ the intersection of $S^{1}(\|x\|+\varepsilon)$ and $S^{1}(e_{1},\|x-e_{1}\|)$ in the first quadrant. The triangles $\triangle(0,1,x)$ and $\triangle(0,1,y)$ give by the Law of Cosines $\|x-1\|^{2}=\|x\|^{2}+1-2\|x\|\cos\gamma$ and $\|y-1\|^{2}=\|y\|^{2}+1-2\|y\|\cos\delta,$ where $\eta$ is the angle $\measuredangle(1,0,x)$ and $\xi$ is the angle $\measuredangle(1,0,y)$. Therefore (1.8.13) $\cos\gamma=\frac{\|x\|^{2}+1-\|x-1\|^{2}}{2\|x\|}$ and (1.8.14) $\cos\delta=\frac{\|y\|^{2}+1-\|y-1\|^{2}}{2\|y\|}=\frac{(\|x\|+\varepsilon)^{2}+1-\|x-1\|^{2}}{2(\|x\|+\varepsilon)}.$ By the Jordan inequality $\|\cos\gamma-\cos\delta\|=\cos\delta-\cos\gamma=2\sin\frac{\delta+\gamma}{2}\sin\frac{\gamma-\delta}{2}\geq\frac{2}{\pi^{2}}(\gamma+\delta)(\gamma-\delta)$ and by assumption (1.8.15) $\|\gamma-\delta\|\leq\frac{\pi^{2}}{2\omega}\|\cos\gamma-\cos\delta\|.$ By the triangle inequality, the Jordan inequality, (1.8.15), (1.8.13) and (1.8.14) $\displaystyle\|x-y\|$ $\displaystyle\leq$ $\displaystyle\varepsilon+2\|x\|\sin\frac{\|\gamma-\delta\|}{2}$ $\displaystyle\leq$ $\displaystyle\varepsilon+\frac{2\|x\|}{\pi}\|\gamma-\delta\|$ $\displaystyle\leq$ $\displaystyle\varepsilon+\frac{\|x\|\pi}{\omega}\|\cos\gamma-\cos\delta\|$ $\displaystyle=$ $\displaystyle\varepsilon+\frac{\|x\|\pi}{\omega}\frac{\varepsilon(1+\|x-1\|^{2}+\|x\|(\|x\|+\varepsilon))}{2\|x\|(\|x\|+\varepsilon)}$ $\displaystyle\leq$ $\displaystyle\varepsilon+\frac{\pi}{\omega}\frac{\varepsilon(1+3^{2}+2(2+1))}{2(1/2+0)}$ $\displaystyle=$ $\displaystyle\varepsilon+\frac{51\varepsilon}{\omega}.$ Let us denote by $z$ the intersection of $S^{1}(\|x\|+\varepsilon)$ and $S^{1}(e_{1},\|x-e_{1}\|+\varepsilon)$ in the first quadrant. If $\delta>\omega/2$, then we obtain $\|x-y\|\leq\frac{\|x\|\pi}{\omega}\frac{\varepsilon\|\|y-1\|^{2}-\|z-1\|^{2}\|}{2(\|x\|+\varepsilon)}\leq\frac{19\varepsilon}{\omega}.$ Now $\textnormal{diam}\,(A)\leq\|x-z\|\leq\|x-y\|+\|y-z\|\leq\varepsilon+\frac{70\varepsilon}{\omega}$ and the assertion follows. ∎ ###### 1.8.16 Lemma. Let $n\geq 2$, $K>1$, $\alpha=K^{1/(1-n)}$, $\beta=1/\alpha$ and $c_{3}=\exp(60\sqrt{K-1})$. For $t\in(0,1)$ (1.8.17) $c_{3}t^{\alpha}-t\geq t-\frac{t^{\beta}}{c_{3}}$ and for $t>1$ (1.8.18) $c_{3}t^{\beta}-t\geq t-\frac{t^{\alpha}}{c_{3}}.$ ###### Proof. To prove (1.8.17) it is sufficient to prove that $f(t)\geq 0$, where $f(t)=c_{3}t^{\alpha}+\frac{1}{c_{3}}t^{1/\alpha}-2t$ and $0<t<1$. Because $\lim_{t\rightarrow 0+}f(t)=0,$ it is sufficient to prove $f^{\prime}(t)\geq 0$ for $0<t<1$, i.e. (1.8.19) $\alpha c_{3}t^{\alpha-1}+\frac{1}{\alpha c_{3}}t^{(1/\alpha)-1}-2\geq 0.$ Using inequality between arithmetic and geometric means, we can conclude that $\alpha c_{3}+\frac{1}{\alpha c_{3}}\geq 2$ holds. In other words, $\lim_{t\rightarrow 1-}f^{\prime}(t)=\alpha c_{3}+\frac{1}{\alpha c_{3}}-2\geq 0.$ By this reason, to prove inequality (1.8.19) it is sufficient to prove that $f^{\prime\prime}(t)\leq 0$ for $0<t<1$ i.e. $\alpha(\alpha-1)c_{3}t^{\alpha-2}+\frac{\frac{1}{\alpha}-1}{\alpha c_{3}}t^{(1/\alpha)-2}\leq 0,$ or equivalently $t^{\frac{1}{\alpha}-\alpha}\leq\alpha^{3}c_{3}^{2}.$ The last inequality follows from $t^{\frac{1}{\alpha}-\alpha}<1\leq\alpha^{3}c_{3}^{2}.$ The first inequality holds because $0<t<1$ and $\frac{1}{\alpha}-\alpha>0$ (because $0<\alpha<1$). Now we prove $\alpha^{3}c_{3}^{2}\geq 1$ to complete proof. This is same inequality as $K^{3/(1-n)}e^{120\sqrt{K-1}}\geq 1,$ or equivalently $e^{40(n-1)u}\geq u^{2}+1$ for $u=\sqrt{K-1}$. Because $u\geq 0$, using Taylor series for $e^{x}$ we can conclude that $e^{40(n-1)u}\geq 1+40(n-1)u+\frac{(40(n-1))^{2}u^{2}}{2}\geq 1+u^{2}.$ The inequality (1.8.18) is equivalent to (1.8.20) $c_{3}t^{(1/\alpha)-1}+\frac{t^{\alpha-1}}{c_{3}}\geq 2.$ Inequality (1.8.20) holds for $t=1$. To prove inequality (1.8.20) for $t>1$ it is sufficient to prove that derivation of the left side of inequality is nonnegative. We have following sequence of equivalent formulas: $c_{3}\left(\frac{1}{\alpha}-1\right)t^{\frac{1}{\alpha}-2}+\frac{\alpha-1}{c_{3}}\,t^{\alpha-2}\geq 0$ $\frac{(1-\alpha)c_{3}}{\alpha}\,t^{\frac{1}{\alpha}-2}\geq\frac{1-\alpha}{c_{3}}\,t^{\alpha-2}$ $t^{\frac{1}{\alpha}-\alpha}\geq\frac{\alpha}{c_{3}^{2}}.$ The last inequality is true because $t^{\frac{1}{\alpha}-\alpha}\geq 1\geq\frac{\alpha}{c_{3}^{2}}$ and the assertion follows. ∎ ###### 1.8.21 Lemma. Let $\varepsilon>0$. Then $\|x\|-\varepsilon\leq\|f(x)\|\leq\|x\|+\varepsilon$ for $1<K\leq\max\left\\{\left(\frac{\log(\varepsilon+1)}{60}\right)^{2}+1,2\right\\}.$ ###### Proof. Let us denote $l(x)=c_{3}^{-1}\max\\{\|x\|^{\alpha},\|x\|^{\beta}\\}$ and $u(x)=c_{3}\max\\{\|x\|^{\alpha},\|x\|^{\beta}\\}$. We will first consider the case $0<\|x\|<1$. By Lemma 1.8.16 $\displaystyle\max\\{u(x)-\|x\|,\|x\|-l(x)\\}$ $\displaystyle=$ $\displaystyle\max\left\\{c_{3}\|x\|^{\alpha}-\|x\|,\|x\|-\frac{1}{c_{3}}\|x\|^{\beta}\right\\}$ $\displaystyle=$ $\displaystyle c_{3}\|x\|^{\alpha}-\|x\|$ $\displaystyle\leq$ $\displaystyle\exp(60\sqrt{K-1})\|x\|^{\alpha}-\|x\|$ $\displaystyle\leq$ $\displaystyle\exp(60\sqrt{K-1})\|x\|^{1/K}-\|x\|$ $\displaystyle\leq$ $\displaystyle\exp(60\sqrt{K-1})-1.$ Now $\exp(60\sqrt{K-1})-1\leq\varepsilon$ is equivalent to (1.8.22) $K\leq\left(\frac{\log(\varepsilon+1)}{60}\right)^{2}+1.$ If $\|x\|=1$, then $\max\\{u(x)-\|x\|,\|x\|-l(x)\\}=c_{3}-1$ and therefore we want $\exp(60\sqrt{K-1})-1\leq\varepsilon$ for $K\in(1,2]$, which is equivalent to (1.8.23) $K\leq\left(\frac{\log(\varepsilon+1)}{60}\right)^{2}+1.$ Let us first consider the case $1<\|x\|<2$. By Lemma 1.8.16 $\displaystyle\max\\{u(x)-\|x\|,\|x\|-l(x)\\}$ $\displaystyle=$ $\displaystyle\max\left\\{c_{3}\|x\|^{\beta}-\|x\|,\|x\|-\frac{1}{c_{3}}\|x\|^{\alpha}\right\\}$ $\displaystyle=$ $\displaystyle c_{3}\|x\|^{\beta}-\|x\|$ $\displaystyle\leq$ $\displaystyle\exp(60\sqrt{K-1})\|x\|^{\beta}-\|x\|$ $\displaystyle\leq$ $\displaystyle\exp(60\sqrt{K-1})\|x\|^{K}-\|x\|$ $\displaystyle\leq$ $\displaystyle\|x\|(\exp(60\sqrt{K-1})\|x\|^{K-1}-1)$ $\displaystyle\leq$ $\displaystyle 2(\exp(60\sqrt{K-1}+(K-1)\log\|x\|)-1)$ $\displaystyle\leq$ $\displaystyle 2(\exp(60(K-1)^{3/2}-1).$ Now $2(\exp(60(K-1)^{3/2}-1)\leq\varepsilon$ is equivalent to (1.8.24) $K\leq\left(\frac{\log(\varepsilon/2+1)}{60}\right)^{2/3}+1.$ By combining (1.8.22), (1.8.23) and (1.8.24) we have $\|x\|-\varepsilon\leq\|f(x)\|\leq\|x\|+\varepsilon$ for $K\leq\min\left\\{\left(\frac{\log(\varepsilon+1)}{60}\right)^{2}+1,2,\left(\frac{\log(\varepsilon/2+1)}{60}\right)^{2/3}+1\right\\}=\left(\frac{\log(\varepsilon+1)}{60}\right)^{2}+1$ and the assertion follows. ∎ ###### 1.8.25 Lemma. For $0<\alpha<1$, $c>1$ and $t>0$ $\log(1+c\max\\{t^{\alpha},t^{1/\alpha}\\})\leq\left\\{\begin{array}[]{ll}\frac{c}{\alpha}\log^{\alpha}(1+t),&0<t<1,\\\ \frac{c}{\alpha}\log(1+t),&t\geq 1.\end{array}\right.$ ###### Proof. Let us first assume $t\geq 1$. Then $\max\\{t^{\alpha},t^{1/\alpha}\\}=t^{1/\alpha}$ and by the generalized Bernoulli inequality $(1+t)^{c/\alpha}\geq(1+ct)^{1/\alpha}\geq 1+c^{1/\alpha}t^{1/\alpha}\geq 1+ct^{1/\alpha}$ implying $\log(1+ct^{1/\alpha})\leq c/\alpha\log(1+t)$. Let us then assume $0<t<1$. Now $\max\\{t^{\alpha},t^{1/\alpha}\\}=t^{\alpha}$, we will show that function $f(t)=\log(1+ct^{\alpha})-\frac{c}{\alpha}\log^{\alpha}(1+t)$ is nonpositive. We easily obtain (1.8.26) $f^{\prime}(t)=\frac{\alpha ct^{\alpha-1}}{1+ct^{\alpha}}-\frac{c\log^{\alpha-1}(1+t)}{1+t}.$ Since $\alpha-1<0$ and $\log(1+t)\leq t$ we have (1.8.27) $t^{\alpha-1}\leq\log^{\alpha-1}(1+t).$ By assumptions $c>1\geq t^{1-\alpha}$ and therefore (1.8.28) $ct^{\alpha}\geq t.$ By (1.8.26), (1.8.27) and (1.8.28) $f^{\prime}(t)\leq 0$ is equivalent to $(1-\alpha)(1+t)\geq 0$ and therefore $f(t)$ is increasing. Now we have $f(t)\leq f(0)=0$ and the assertion follows. ∎ ###### 1.8.29 Theorem. Let $G=\mathbb{R}^{n}\setminus\\{0\\}$, $f\in QC_{K}$ and $f(0)=0$. There exists $c(K)$ such that $j_{G}(f(x),f(y))\leq c(K)\max\\{j_{G}(x,y)^{\alpha},j_{G}(x,y)\\},$ where $\alpha=K^{1/(1-n)}$, and $c(K)\to 1$ as $K\to 1$. ###### Proof. By symmetry we may assume $x=e_{1}$ and $\|y\|\geq 1$. Now $\frac{\|f(y)-f(e_{1})\|}{\|e_{1}\|}=\frac{\|f(y)-f(x)\|}{\|f(0)-f(e_{1})\|}\leq\eta\left(\frac{\|x-y\|}{\|0-e_{1}\|}\right)=\eta(\|x-y\|)$ and $\frac{\|f(y)-f(e_{1})\|}{\|f(y)\|}=\frac{\|f(y)-f(x)\|}{\|f(y)-f(0)\|}\leq\eta\left(\frac{\|x-y\|}{\|y-0\|}\right)=\eta\left(\frac{\|x-y\|}{\|y\|}\right).$ Therefore by Lemma 1.8.25 $\displaystyle j(f(x),f(y))$ $\displaystyle=$ $\displaystyle\log\left(1+\frac{\|f(x)-f(y)\|}{\min\\{\|f(x)\|,\|f(y)\|\\}}\right)$ $\displaystyle=$ $\displaystyle\log\left(1+\max\left\\{\eta(\|y-e_{1}\|),\eta\left(\frac{\|x-y\|}{\|y\|}\right)\right\\}\right)$ $\displaystyle=$ $\displaystyle\log(1+\eta(\|y-e_{1}\|))$ $\displaystyle\leq$ $\displaystyle\log(1+c_{3}\max\\{\|y-e_{1}\|^{\alpha},\|y-e_{1}\|^{1/\alpha}\\})$ $\displaystyle\leq$ $\displaystyle\left\\{\begin{array}[]{ll}\frac{c_{3}}{\alpha}\log^{\alpha}(1+\|y-e_{1}\|),&0<\|y-e_{1}\|<1,\\\ \frac{c_{3}}{\alpha}\log(1+\|y-e_{1}\|),&\|y-e_{1}\|\geq 1.\end{array}\right.$ By choosing $c(K)=c_{3}/\alpha$ we have $c(K)\to 1$ as $K\to 1$ by Lemma 1.8.7 and the assertion follows. ∎ ###### 1.8.31 Problem. Is this Theorem true for $k_{G}$ instead of $j_{G}$? ## Chapter 2 Harmonic Quasiregular Mappings It is well known that if $f$ is a complex-valued harmonic function defined in a region $G$ of the complex plane $\mathbb{C},$ then $\|f\|^{p}$ is subharmonic for $p\geq 1,$ and that in the general case is not subharmonic for $p<1.$ However, if $f$ is holomorphic, then $\|f\|^{p}$ is subharmonic for every $p>0.$ Here we consider $k$-quasiregular harmonic functions $(0<k<1).$ We recall that a harmonic function is quasiregular if $\|\bar{\partial}f(z)\|\leq k\|\partial f(z)\|,\qquad z\in G,$ where $\bar{\partial}f(z)=\frac{1}{2}\left(\frac{\partial f}{\partial x}+i\frac{\partial f}{\partial y}\right)\quad\text{and}\quad\partial f(z)=\frac{1}{2}\left(\frac{\partial f}{\partial x}-i\frac{\partial f}{\partial y}\right),\qquad z=x+iy.$ We prove that $\|f\|^{p}$ is subharmonic for $p\geq 4k/(1+k)^{2}=:q$ as well as that the exponent $q$ $(<1)$ is the best possible (see Theorem 2.1.1). The fact that $q<1$ enables us to prove that if $f$ is quasiregular in the unit disk $\mathbb{D}$ and continuous on $\overline{D\mathstrut}$, then $\tilde{\omega}(f,\delta)\leq{\rm const.}\omega(f,\delta),$ where $\tilde{\omega}(f,\delta)$ (respectively $\omega(f,\delta)$) denotes the modulus of continuity of $f$ on $\mathbb{D}$ (respectively $\partial\mathbb{D}$); see Theorem 2.2.1. ### 2.1. Subharmonicity of $\|f\|^{p}$ ###### 2.1.1 Theorem. [KP] If $f$ is a complex-valued $k$-quasiregular harmonic function defined on a region $G\subset\mathbb{C},$ and $q=4k/(k+1)^{2},$ then $\|f\|^{q}$ is subharmonic. The exponent $q$ is optimal. Recall that a continuous function $u$ defined on a region $G\subset\mathbb{C}$ is subharmonic if for all $z_{0}\in G$ there exists $\varepsilon>0$ such that (2.1.2) $u(z_{0})\leq\frac{1}{2\pi}\int_{0}^{2\pi}u(z_{0}+re^{it})\,dt,\qquad 0<r<\varepsilon,$ If $u(z_{0})$ = $\|f(z_{0})\|^{2}=0,$ then (2.1.2) holds. If $u(z_{0})>0,$ then there exists a neighborhood $U$ of $z_{0}$ such that $u$ is of class $C^{2}(U)$ (because the zeroes of $u$ are isolated), and then we may prove that $\Delta u\geq 0$ on $U$. Thus the proof reduces to proving that $\Delta u(z)\geq 0$ whenever $u(z)>0.$ In order to do this we will calculate $\Delta u.$ It is easy to prove that If $u>0$ is a $C^{2}$ function defined on a region in $\mathbb{C},$ and $\alpha\in\mathbb{R},$ then next two statements holds (2.1.3) $\Delta(u^{\alpha})=\alpha u^{\alpha-1}\Delta u+\alpha(\alpha-1)u^{\alpha-2}\|\nabla u\|^{2},$ (2.1.4) $\|\nabla u\|^{2}=4\|\partial u\|^{2}\quad\text{and}\quad\Delta u=4\partial\bar{\partial}u.$ ###### 2.1.5 Lemma. If $f=g+\bar{h},$ where $g$ and $h$ are holomorphic functions, then (2.1.6) $\Delta(\|f\|^{2})=4(\|g^{\prime}\|^{2}+\|h^{\prime}\|^{2}).$ ###### Proof. Since $\|f\|^{2}=(g+\bar{h})(\bar{g}+h),$ we have $\displaystyle\Delta(\|f\|^{2})$ $\displaystyle=4\partial(\overline{h^{\prime}\mathstrut}(\bar{g}+h)+(g+\bar{h})\overline{g^{\prime}\mathstrut})$ $\displaystyle=4(\overline{h^{\prime}\mathstrut}h+g\overline{g^{\prime}\mathstrut})$ $\displaystyle=4(\|g^{\prime}\|^{2}+\|h^{\prime}\|^{2}).$ ∎ ###### 2.1.7 Lemma. If $f=g+\bar{h},$ where $g$ and $h$ are holomorphic functions, then (2.1.8) $\|\nabla(\|f\|^{2})\|^{2}=4(\|\overline{g^{\prime}\mathstrut}\|^{2}+\|\overline{h^{\prime}\mathstrut}\|^{2})\|f\|^{2}+8\,{\rm Re}(\overline{g^{\prime}\mathstrut}h^{\prime}f^{2}).$ ###### Proof. We have $\displaystyle\|\nabla(\|f\|^{2})\|^{2}$ $\displaystyle=4\|\partial(\|f\|^{2})\|^{2}$ $\displaystyle=4\|\partial((g+\bar{h})(\bar{g}+h))\|^{2}$ $\displaystyle=4\|g^{\prime}\bar{f}+fh^{\prime}\|^{2}$ $\displaystyle=4(\|g^{\prime}\|^{2}+\|h^{\prime}\|^{2})\|f\|^{2}+8\,{\rm Re}(\overline{g^{\prime}\mathstrut}h^{\prime}f^{2}).$ ∎ ###### 2.1.9 Lemma. If $f=g+\bar{h},$ where $g$ and $h$ are holomorphic functions, then (2.1.10) $\Delta(\|f\|^{p})=p^{2}(\|g^{\prime}\|^{2}+\|h^{\prime}\|^{2})\|f\|^{p-2}+2p(p-2)\|f\|^{p-4}\,{\rm Re}(\overline{g^{\prime}\mathstrut}h^{\prime}f^{2})$ whenever $f\neq 0.$ ###### Proof. We take $\alpha=p/2,$ $u=\|f\|^{2},$ and then use (2.1.3), (2.1.6) and (2.1.8) to get the result. ∎ 1.11. Proof of Theorem 2.1.1. We have to prove that $\Delta(\|f\|^{p})\geq 0,$ where $p=4k/(1+k)^{2}.$ Since $p-2<0,$ we get from (2.1.10) that $\displaystyle\Delta(\|f\|^{p})$ $\displaystyle\geq p^{2}(\|g^{\prime}\|^{2}+\|h^{\prime}\|^{2})\|f\|^{p-2}+2p(p-2)\|f\|^{p-4}\|g^{\prime}\|\cdot\|h^{\prime}\|\cdot\|f\|^{2}$ $\displaystyle=p^{2}\|g^{\prime}\|^{2}(m^{2}+1)\|f\|^{p-2}+2p(p-2)\|g^{\prime}\|^{2}\|f\|^{p-2}m$ $\displaystyle=p\|g^{\prime}\|^{2}\|f\|^{p-2}[p(1+m^{2})+2(p-2)m],$ where $m=\|h^{\prime}\|/\|g^{\prime}\|\leq k.$ The function $m\mapsto p(1+m^{2})+2(p-2)m$ has a negative derivative (because $p<1$ and $m<1$), which implies that $(1+m^{2})p+2(p-2)m\geq(1+k^{2})p+2(p-2)k.$ On the other hand $(1+k^{2})p+2(p-2)k\geq 0$ if and only if $p\geq 4k/(1+k)^{2},$ which proves that $\|f\|^{q}$ is subharmonic. To prove that the exponent $q$ is optimal we take $f(z)=z+k\bar{z}.$ By (2.1.10), $\Delta(\|f\|^{p})(1)=p^{2}(1+k^{2})(1+k)^{p-2}+2p(p-2)(1+k)^{p-2}k.$ Hence $\Delta(\|f\|^{p})(1)\geq 0$ if and only if $p(1+k^{2})+2(p-2)k\geq 0,$ which, as noted above, is equivalent to $p\geq q.$ This completes the proof of Theorem 2.1.1. $\square$ ### 2.2. Moduli of continuity in Euclidean metric For a continuous function $f:\overline{\mathbb{D}\mathstrut}\mapsto\mathbb{C}$ harmonic in $\mathbb{D}$ we define two moduli of continuity: $\omega(f,\delta)=\sup\\{\|f(e^{i\theta})-f(e^{it})\|:\|e^{i\theta}-e^{it}\|\leq\delta,\ t,\theta\in\mathbb{R}\\},\quad\delta\geq 0,$ and $\tilde{\omega}(f,\delta)=\sup\\{\|f(z)-f(w)\|:\|z-w\|\leq\delta,\ z,w\in\overline{\mathbb{D}\mathstrut}\\},\quad\delta\geq 0.$ Clearly $\omega(f,\delta)\leq\tilde{\omega}(f,\delta),$ but the reverse inequality need not hold. To see this consider the function $f(re^{i\theta})=\sum_{n=1}^{\infty}\frac{(-1)^{n}r^{n}\cos n\theta}{n^{2}},\qquad re^{i\theta}\in\overline{\mathbb{D}\mathstrut}.$ This function is harmonic in $\mathbb{D}$ and continuous on $\overline{\mathbb{D}\mathstrut}.$ The function $v(\theta)=f(e^{i\theta})$, $\|\theta\|<\pi,$ is differentiable, and $\displaystyle\frac{dv}{d\theta}$ $\displaystyle=\sum_{n=1}^{\infty}\frac{(-1)^{n-1}\sin n\theta}{n}$ $\displaystyle=\frac{\theta}{2},\qquad\|\theta\|<\pi.$ This formula is well known, and can be verified by calculating the Fourier coefficients of the function $\theta\mapsto\theta/2,\ \|\theta\|<\pi.$ It follows that $\|f(e^{i\theta})-f(e^{it})\|\leq(\pi/2)\|\theta-t\|,\qquad-\pi<\theta,\,t<\pi,$ and hence $\omega(f,\delta)\leq M\delta,\ \delta>0,$ where $M$ is an absolute constant. On the other hand, the inequality $\tilde{\omega}(f,\delta)\leq CM\delta$, $C={\rm const},$ does not hold because it implies that $\|\partial f/\partial r\|\leq CM,$ which is not true because $\frac{\partial}{\partial r}f(re^{i\theta})=\sum_{n=1}^{\infty}\frac{r^{n-1}}{n},\quad\text{for $\theta=\pi,\ 0<r<1$.}$ However, as was proved by Rubel, Shields and Taylor [RST], and Tamrazov [TA], if $f$ is a holomorphic function, then $\tilde{\omega}(f,\delta)\leq C\omega(f,\delta)$, where $C$ is independent of $f$ and $\delta.$ Here we extend that result to quasiregular harmonic functions. ###### 2.2.1 Theorem. [KP] Let $f$ be a $k$-quasiregular harmonic complex-valued function which has a continuous extension on $\overline{\mathbb{D}\mathstrut}$, then there is a constant $C$ depending only on $k$ such that $\tilde{\omega}(f,\delta)\leq C\omega(f,\delta)$. In order to deduce this fact from Theorem 2.1.1, we need some simple properties of the modulus $\omega(f,\delta).$ Let $\omega_{0}(f,\delta)=\sup\\{\|f(e^{i\theta})-f(e^{it})\|:\|\theta-t\|\leq\delta,\ t,\theta\in\mathbb{R}\\}.$ It is easy to check that (2.2.2) $C^{-1}\omega_{0}(f,\delta)\leq\omega(f,\delta)\leq C\omega_{0}(f,\delta),$ where $C$ is an absolute constant, and that $\omega_{0}(f,\delta_{1}+\delta_{2})\leq\omega_{0}(f,\delta_{1})+\omega_{0}(f,\delta_{2}),\quad\delta_{1},\,\delta_{2}\geq 0.$ Hence, $\omega_{0}(f,2^{n}\delta)\leq 2^{n}\omega_{0}(f,\delta),$ and hence $\omega_{0}(\lambda\delta)\leq 2\lambda\omega_{0}(\delta),$ for $\lambda\geq 1,\delta\geq 0.$ From these inequalities and (2.2.2) it follows that (2.2.3) $\omega(f,\lambda\delta)\leq 2C\lambda\omega(f,\delta),\quad\lambda\geq 1,\delta\geq 0,$ and (2.2.4) $\omega(f,\delta_{1}+\delta_{2})\leq C\omega(f,\delta_{1})+C\omega(f,\delta_{2}),\quad\delta_{1},\,\delta_{2}\geq 0,$ where $C$ is an absolute constant. As a consequence of (2.2.3) we have, for $0<p<1,$ (2.2.5) $\int_{x}^{\infty}\frac{\omega(f,t)^{p}}{t^{2}}\,dt\leq C\frac{\omega(f,x)^{p}}{x},\quad x>0,$ where $C$ depends only on $p.$ Finally we need the following consequence of the harmonic Schwarz lemma (see [ABR]). ###### 2.2.6 Lemma. If $h$ is a function harmonic and bounded in the unit disk, with $h(0)=0,$ the $\|h(\xi)\|\leq(4/\pi)\\|h\\|_{\infty}\|\xi\|,$ for $\xi\in\mathbb{D}.$ 2.7. Proof of Theorem 2.2.1. It is enough to prove that $\|f(z)-f(w)\|\leq C\omega(f,\|z-w\|)$ for all $z,\,w\in\overline{\mathbb{D}\mathstrut},$ where $C$ depends only on $k.$ Assume first that $z=r\in(0,1)$ and $\|w\|=1.$ Then, by Theorem 2.1.1, the function $\varphi(\xi)=\|f(w)-f(\xi)\|^{q},$ where $q=4k/(1+k)^{2}<1,$ is subharmonic in $\mathbb{D}$ and continuous on $\overline{\mathbb{D}\mathstrut},$ whence $\displaystyle\varphi(r)\leq\frac{1}{2\pi}\int_{\partial\mathbb{D}}\frac{(1-r^{2})\varphi(\zeta)}{\|\zeta-r\|^{2}}\,\|d\zeta\|.$ Since, by (2.2.4), $\displaystyle\varphi(\zeta)$ $\displaystyle\leq(\omega(f,\|w-r\|+\|r-\zeta\|))^{q}$ $\displaystyle\leq C^{q}\omega(f,\|w-r\|)^{q}+C^{q}\omega(f,\|r-\zeta\|)^{q},$ we have $\displaystyle\varphi(z)$ $\displaystyle\leq C^{q}\omega(f,\|w-r\|)^{q}+\frac{C^{q}}{2\pi}\int_{\partial\mathbb{D}}\frac{(1-r^{2})\omega(f,\|r-\zeta\|)^{q}}{\|\zeta-r\|^{2}}\,\|d\zeta\|$ $\displaystyle=C^{q}\omega(f,\|w-r\|)^{q}+\frac{C^{q}}{2\pi}\int_{-\pi}^{\pi}\frac{(1-r^{2})\,\omega(\|r-e^{it}\|)^{q}}{\|e^{it}-r\|^{2}}\,dt.$ But simple calculation shows that $\|r-e^{it}\|=\sqrt{(1-r)^{2}+4r\sin^{2}(t/2)\mathstrut}\asymp 1-r+\|t\|\quad(0<r<1,\ \|t\|\leq\pi).$ From this, (2.1.2), and (2.2.5) it follows that $\displaystyle\int_{-\pi}^{\pi}\frac{(1-r^{2})\,\omega(f,\|r-e^{it}\|)^{q}}{\|e^{it}-r\|^{2}}\,dt$ $\displaystyle\leq C_{1}\int_{0}^{\pi}\frac{(1-r)\,\omega(f,1-r+t)^{q}}{(1-r+t)^{2}}\,dt$ $\displaystyle=C_{1}\Big{(}\int_{0}^{1-r}+\int_{1-r}^{\pi}\Big{)}\frac{(1-r)\,\omega(f,1-r+t)^{q}}{(1-r+t)^{2}}\,dt$ $\displaystyle\leq C_{2}\,(\omega(1-r))^{q}+C_{2}\,(1-r)\int_{1-r}^{\infty}\frac{\omega(f,t)^{q}}{t^{2}}\,dt$ $\displaystyle\leq C_{3}\,(\omega(f,1-r))^{q}$ $\displaystyle\leq C_{4}\,(\omega(f,\|w-z\|))^{q}.$ Thus $\|f(w)-f(z)\|\leq C_{5}\omega(f,\|w-z\|)$ provided $w\in\partial\mathbb{D}$ and $z\in(0,1).$ By rotation and the continuity of $f$, we can extend this inequality to the case where $w\in\partial\mathbb{D}$ and $z\in\overline{\mathbb{D}\mathstrut}.$ If $0<\|w\|<1,$ we consider the function $h(\xi)=f(\xi w/\|w\|)-f(\xi z/\|w\|)$, $\|\xi\|\leq 1.$ This function is harmonic in $\mathbb{D},$ continuous on $\overline{\mathbb{D}\mathstrut},$ and $h(0)=0.$ Hence, by the harmonic Schwarz lemma, inequality (2.1.2), and the preceding case, $\displaystyle\|f(w)-f(z)\|$ $\displaystyle=\|h(\|w\|)\|$ $\displaystyle\leq(4/\pi)\|w\|\,\\|h\\|_{\infty}$ $\displaystyle\leq C_{6}\|w\|\,\omega(f,\|w/\|w\|-z/\|w\|\,\|)$ $\displaystyle\leq C_{7}\,\omega(f,\|w\|\,\|\,\|w/\|w\|-z/\|w\|\,\|)$ $\displaystyle=C_{7}\,\omega(f,\|w-z\|),$ which completes the proof. $\square$ ### 2.3. Lipschitz continuity up to the boundary on $B^{n}$ It is known, even for $n=2$, that Lipschitz continuity of $\phi:T\rightarrow C$, where $T=\\{z\in C:\|z\|=1\\}$, does not imply Lipschitz continuity of $u=P[\phi]$. Here, for any $n\geq 2$, $P[\phi](x)=\int_{S^{n-1}}P(x,\xi)\phi(\xi)d\sigma(\xi),\;\;x\in B^{n}$ where $P(x,\xi)=\frac{1-\|x\|^{2}}{\|x-\xi\|^{n}}$ is the Poisson kernel for the unit ball $B^{n}=\\{x\in R^{n}:\|x\|<1\\}$, $d\sigma$ is the normalized surface measure on the unit sphere $S^{n-1}$ and $\phi:S^{n-1}\rightarrow R^{n}$ is a continuous mapping. Our aim is to show that Lipschitz continuity is preserved by harmonic extension, if the extension is quasiregular. The analogous statement is true for Hölder continuity without assumption of quasiregularity. ###### 2.3.1 Theorem. [AKM] Assume $\phi:S^{n-1}\rightarrow R^{n}$ satisfies a Lipschitz condition: $\|\phi(\xi)-\phi(\eta)\|\leq L\|\xi-\eta\|,\;\;\xi,\eta\in S^{n-1}$ and assume $u=P[\phi]:B^{n}\rightarrow R^{n}$ is $K$-quasiregular. Then $\|u(x)-u(y)\|\leq C^{\prime}\|x-y\|,\;\;x,y\in B^{n}$ where $C^{\prime}$ depends on $L$, $K$ and $n$ only. D. Kalaj obtained a related result, but under additonal assumption of $C^{1,\alpha}$ regularity of $\phi$, (see [KA]). ###### Proof. The main part of the proof is the estimate of the tangential derivatives of $u$, and in that part quasiregularity plays no role. We choose $x_{0}=r\xi_{0}\in B^{n}$, $r=\|x\|$, $\xi_{0}\in S^{n-1}$. Let $T=T_{x_{0}}rS^{n-1}$ be the $n-1$ dimensional tangent plane at $x_{0}$ to the sphere $rS^{n-1}$. We want to prove that (2.3.2) $\\|D(u\|_{T})(x_{0})\\|\leq C(n)L.$ Without loss of generality we can assume $\xi_{0}=e_{n}$ and $x_{0}=re_{n}$. By a simple calculation $\frac{\partial}{\partial x_{j}}P(x,\xi)=\frac{-2x_{j}}{\|x-\xi\|^{n}}-n(1-\|x\|^{2})\frac{x_{j}-\xi_{j}}{\|x-\xi\|^{n+2}}.$ Hence, for $1\leq j<n$ we have $\frac{\partial}{\partial x_{j}}P(x_{0},\xi)=n(1-\|x_{0}\|^{2})\frac{\xi_{j}}{\|x_{0}-\xi\|^{n+2}}.$ It is important to note that this kernel is odd in $\xi$ (with respect to reflection $(\xi_{1},\ldots\xi_{j},\ldots,\xi_{n})\mapsto(\xi_{1},\ldots,-\xi_{j},\ldots,\xi_{n})$), a typical fact for kernels obtained by differentiation. This observation and differentiation under integral sign gives, for any $1\leq j<n$, $\displaystyle\frac{\partial u}{\partial x_{j}}(x_{0})$ $\displaystyle=$ $\displaystyle n(1-r^{2})\int_{S^{n-1}}\frac{\xi_{j}}{\|x_{0}-\xi\|^{n+2}}\phi(\xi)d\sigma(\xi)$ $\displaystyle=$ $\displaystyle n(1-r^{2})\int_{S^{n-1}}\frac{\xi_{j}}{\|x_{0}-\xi\|^{n+2}}(\phi(\xi)-\phi(\xi_{0}))d\sigma(\xi).$ Using the elementary inequality $\|\xi_{j}\|\leq\|\xi-\xi_{0}\|$, ($1\leq j<n$, $\xi\in S^{n-1}$) and Lipschitz continuity of $\phi$ we get $\displaystyle\left\|{\partial u\over\partial x_{j}}(x_{0})\right\|$ $\displaystyle\leq$ $\displaystyle Ln(1-r^{2})\int_{S^{n-1}}{\|\xi_{j}\|\|\xi-\xi_{0}\|\over\|x_{0}-\xi\|^{n+2}}d\sigma(\xi)$ $\displaystyle\leq$ $\displaystyle Ln(1-r^{2})\int_{S^{n-1}}{\|\xi-\xi_{0}\|^{2}\over\|x_{0}-\xi\|^{n+2}}d\sigma(\xi).$ In order to estimate the last integral, we split $S^{n-1}$ into two subsets $E=\\{\xi\in S^{n-1}:\|\xi-\xi_{0}\|\leq 1-r\\}$ and $F=\\{\xi\in S^{n-1}:\|\xi-\xi_{0}\|>1-r\\}$. Since $\|\xi-x_{0}\|\geq 1-\|x_{0}\|$ for all $\xi\in S^{n-1}$ we have $\displaystyle\int_{E}{\|\xi-\xi_{0}\|^{2}\over\|x_{0}-\xi\|^{n+2}}d\sigma(\xi)$ $\displaystyle\leq$ $\displaystyle(1-r^{2})^{-n-2}\int_{E}\|\xi-\xi_{0}\|^{2}d\sigma(\xi)$ $\displaystyle\leq$ $\displaystyle(1-r^{2})^{-n-2}\int_{0}^{1-r}\rho^{2}\rho^{n-2}d\rho$ $\displaystyle\leq$ $\displaystyle{2\over n+1}(1-r)^{-1}.$ On the other hand, $\|\xi-\xi_{0}\|\leq C_{n}\|\xi-x_{0}\|$ for every $\xi\in F$, so $\displaystyle\int_{F}{\|\xi-\xi_{0}\|^{2}\over\|x_{0}-\xi\|^{n+2}}d\sigma(\xi)$ $\displaystyle\leq$ $\displaystyle C_{n}^{n+2}\int_{F}\|\xi-\xi_{0}\|^{-n}d\sigma(\xi)$ $\displaystyle\leq$ $\displaystyle C_{n}^{\prime}\int_{1-r}^{2}\rho^{-n}\rho^{n-2}d\rho$ $\displaystyle\leq$ $\displaystyle C_{n}^{\prime}(1-r)^{-1}.$ Combining these two estimates we get $\left\|{\partial u\over\partial x_{j}}(x_{0})\right\|\leq LC(n)$ for $1\leq j<n$. Due to rotational symmetry, the same estimate holds for every derivative in any tangential direction. This establishes estimate (2.3.2). Finally, $K$-quasiregularity gives $\\|Du(x)\\|\leq LKC(n).$ Now the mean value theorem gives Lipschitz continuity of $u$. ∎ ###### 2.3.3 Problem. 1. (1) Can one prove similar result for other type of moduli of continuity, as was done in Section 2 in the planar case? 2. (2) The same questions can be posed in other smoothly bounded domains. ### 2.4. Bilipschitz maps Bilipschitz property of harmonic quasiconformal mappings on the unit disc was investigated in [MAT]. A different approach to the following theorem is given in [MAT1]. ###### 2.4.1 Theorem. Suppose $D$ and $D^{\prime}$ are proper domains in $\mathbb{R}^{2}$. If $f:D\longrightarrow D^{\prime}$ is $K$-qc and harmonic, then it is bilipschitz with respect to quasihyperbolic metrics on $D$ and $D^{\prime}$. ###### Proof. Since $f$ is harmonic we have locally, representation $f(z)=g(z)+\overline{h(z)},$ where $g$ and $h$ are analytic functions. Then Jacobian $J_{f}(z)=\|g^{\prime}(z)\|^{2}-\|h^{\prime}(z)\|^{2}>0$ (note that $g^{\prime}(z)\neq 0$). Futher, $J_{f}(z)=\|g^{\prime}(z)\|^{2}\left(1-\frac{\|h^{\prime}(z)\|^{2}}{\|g^{\prime}(z)\|^{2}}\right)=\|g^{\prime}(z)\|^{2}\left(1-\|\omega(z)\|^{2}\right),$ where $\omega(z)=\frac{h^{\prime}(z)}{g^{\prime}(z)}$ is analytic and $\|\omega\|<1$. Now we have $\log\frac{1}{J_{f}(z)}=-2\log\|g^{\prime}(z)\|-\log(1-\|\omega(z)\|^{2}).$ The first term is harmonic function (it is well known that logarithm of moduli of analytic function is harmonic everywhere except where that analytic function vanishes, but $g^{\prime}(z)\neq 0$ everywhere). The second term can be expanded in series $\sum_{k=1}^{\infty}\frac{\|\omega(z)\|^{2k}}{k},$ and each term is subharmonic (note that $\omega$ is analytic). So, $-\log(1-\|\omega(z)\|^{2})$ is a continuous function represented as a locally uniform sum of subharmonic functions. Thus it is also subharmonic. Hence (2.4.2) $\log\frac{1}{J_{f}(z)}\mbox{ is a subharmonic function}.$ Note that representation $f(z)=g(z)+\overline{h(z)}$ is local, but that suffices for our conclusion (2.4.2). By the definition from [AG, Definition 1.5] $\alpha_{f}(z)=\exp\left(\frac{1}{n}(\log J_{f})_{B_{z}}\right),$ where $(\log J_{f})_{B_{z}}=\frac{1}{m(B_{z})}\int_{B_{z}}\log J_{f}\,dm,\quad B_{z}=B(z,d(z,\partial D)).$ In the case $n=2$ we have (2.4.3) $\frac{1}{\alpha_{f}(z)}=\exp\left(\frac{1}{2}\frac{1}{m(B_{z})}\int_{B_{z}}\log\frac{1}{J_{f}(w)}\,dm(w)\right).$ From (2.4.2) we have $\frac{1}{m(B_{z})}\int_{B_{z}}\log\frac{1}{J_{f}(w)}\,dm(w)\geq\log\frac{1}{J_{f}(z)}.$ Combining this with (2.4.3) we have $\frac{1}{\alpha_{f}(z)}\geq\exp\left(\frac{1}{2}\log\frac{1}{J_{f}(z)}\right)=\frac{1}{\sqrt{J_{f}(z)}}$ and therefore $\sqrt{J_{f}(z)}\geqslant\alpha_{f}(z).$ On the other hand, we have Theorem [AG, Theorem 1.8]: Suppose that $D$ and $D^{\prime}$ are domains in $\mathbb{R}^{n}$ if $f:D\longrightarrow D^{\prime}$ is $K$-qc, then $\frac{1}{c}\frac{d(f(x),\partial D^{\prime})}{d(x,\partial D)}\leq\alpha_{f}(z)\leq c\,\frac{d(f(x),\partial D^{\prime})}{d(x,\partial D)}$ for $x\in D$, where $c$ is a constant wich depends only on $K$ and $n$. From first inequality of this theorem we have (2.4.4) $\sqrt{J_{f}(z)}\geq\frac{1}{c}\frac{d(f(x),\partial D^{\prime})}{d(x,\partial D)}.$ Note that $J_{f}(z)=\|g^{\prime}(z)\|^{2}-\|h^{\prime}(z)\|^{2}\leq\|g^{\prime}(z)\|^{2}$ and by $K$-qclity of $f$, $\|h^{\prime}\|\leq k\|g^{\prime}\|$, $0\leq k<1$, where $K=\frac{1+k}{1-k}$. This gives $J_{f}\geq(1-k^{2})\|g^{\prime}\|^{2}$. Hence, $\sqrt{J_{f}}\asymp\|g^{\prime}\|\asymp\|g^{\prime}\|+\|h^{\prime}\|=\|\|f^{\prime}(z)\|\|.$ Finally (2.4.4) and the above asymptotic relation give $\|\|f^{\prime}(z)\|\|\geq\frac{1}{c}\frac{d(f(x),\partial D^{\prime})}{d(x,\partial D)},\quad c=c(k).$ For the reversed inequality we again use $J_{f}(z)\geq(1-k^{2})\|g^{\prime}(z)\|^{2}$, i.e. (2.4.5) $\sqrt{J_{f}(z)}\geq\sqrt{1-k^{2}}\|g^{\prime}(z)\|$ Further, we know that for $n=2$ $\alpha_{f}(z)=\exp\left(\frac{1}{m(B_{z})}\int_{B_{z}}\log\sqrt{J_{f}(x)}\,dm(w)\right).$ Using (2.4.5) $\begin{array}[]{rcl}\displaystyle\frac{1}{m(B_{z})}\int_{B_{z}}\log\sqrt{J_{f}(x)}\,dm(w)&\geq&\displaystyle\frac{1}{m(B_{z})}\int_{B_{z}}\log\sqrt{1-k^{2}}+\log\|g^{\prime}(w)\|\,dm(w)\vspace{0.5em}\\\ &=&\displaystyle\log\sqrt{1-k^{2}}+\frac{1}{m(B_{z})}\int_{B_{z}}\log\|g^{\prime}(w)\|\,dm(w)\vspace{0.5em}\\\ &=&\log\sqrt{1-k^{2}}+\log\|g^{\prime}(z)\|.\end{array}$ Now we have by harmonicity of $\log\|g^{\prime}\|$ $\begin{array}[]{rcl}\displaystyle\alpha_{f}(z)&=&\displaystyle\exp\left(\frac{1}{m(B_{z})}\int_{B_{z}}\log\sqrt{J_{f}(x)}\,dm(w)\right)\vspace{0.5em}\\\ &\geq&\displaystyle\exp(\log\sqrt{1-k^{2}}+\log\|g^{\prime}(z)\|)\vspace{0.5em}\\\ &=&\displaystyle\sqrt{1-k^{2}}\|g^{\prime}(z)\|\vspace{0.5em}\\\ &\geq&\displaystyle\frac{1}{2}\sqrt{1-k^{2}}(\|g^{\prime}\|+\|h^{\prime}\|)\vspace{0.5em}\\\ &=&\displaystyle\frac{\sqrt{1-k^{2}}}{2}\|\|f^{\prime}\|\|.\end{array}$ Again using the second inequality in [AG, Theorem 1.8] $\|\|f^{\prime}\|\|\leqslant c\sqrt{J_{f}(z)}\leqslant c\,\alpha_{f}(z)\leqslant c\,\frac{d(f(z),\partial D^{\prime})}{d(z,\partial D)},\quad c=c(k).$ Summarizing $\|\|f^{\prime}(z)\|\|\asymp\frac{d(f(z),\partial D^{\prime})}{d(z,\partial D)}.$ This pointwise result, via integration along curves, easily gives $k_{D^{\prime}}(f(z_{1}),f(z_{2}))\asymp k_{D}(z_{1},z_{2}).$ ∎ ###### 2.4.6 Problem. Is Theorem 2.4.1 true in dimensions $n\geq 3$? ## References [AhB] L. Ahlfors and A. Beurling: Conformal invariants and function-theoretic null-sets, Acta Math. 83 (1950), 101–129. * [AVV] G. D. Anderson, M. K. Vamanamurthy, and M. Vuorinen: Conformal invariants, inequalities and quasiconformal mappings, J. Wiley, 1997, 505 pp. * [AKM] M. Arsenović, V. Kojić and M. Mateljević: On Lipschitz continuity of harmonic quasiregular maps on the unit ball in $\mathbb{R}^{n}$., Ann. Acad. Sci. Fenn. Math. 33 (2008), no. 1, 315–318. * [AG] K. Astala and F. W. Gehring: _Quasiconformal analogues of theorems of Koebe and Hardy-Littlewood,_ Michigan Math. J. 32 (1985), 99-107. * [ABR] S. Axler, P. Bourdon, and W. Ramey: _Harmonic function theory,_ Graduate Texts in Mathematics, vol. 137, Springer-Verlag, New York, 1992. * [B1] A. F. Beardon: The geometry of discrete groups, Graduate Texts in Math. Vol 91, Springer Verlag, Berlin – Heidelberg – New York, 1982. * [B2] A. F. Beardon: The Apollonian metric of a domain in $\mathbb{R}^{n}$, Quasiconformal mappings and analysis – a collection of papers honoring F. W. Gehring, ed. by P. L. Duren, J. M. Heinonen, B. G. Osgood and B. P. Palka, Springer Verlag 1998, 91-108. * [Bel] P. P. Belinskii: General properties of quasiconformal mappings (Russian), Izd. Nauka, Novosibirsk, 1974. * [G1] F.W. Gehring: _Symmetrization of rings in space,_ Trans. Amer. Math. Soc. 101 (1961), 499–519. * [G2] F.W. Gehring: _Rings and quasiconformal mappings in space,_ Trans. Amer. Math. Soc. 103 (1962), 353–393. * [G3] F.W. Gehring:_Quasiconformal mappings in Euclidean spaces,_ Handbook of complex analysis: geometric function theory, Vol. 2, ed. by R. Kühnau, 1-29, Elsevier, Amsterdam, 2005. * [GO] F.W. Gehring and B.G. Osgood: Uniform domains and the quasi-hyperbolic metric, J. Anal. Math. 36 (1979), 50–74. * [H] V. Heikkala: Inequalities for conformal capacity, modulus, and conformal invariants, Ann. Acad. Sci. Fenn. Math. Diss. No. 132 (2002), 62 pp. * [HV] V. Heikkala and M. Vuorinen: Teichmüller’s extremal ring problem, Math. Z. 254 (2006), no. 3, 509–529. * [KA] D. Kalaj: _On harmonic quasiconformal self-mappings of the unit ball,_ Ann. Acad. Sci. Fenn. Math. 33, 2008, 261-271. * [Kl] R. Klén: _Local convexity properties of $j$-metric balls,_ Ann. Acad. Sci. Fenn. Math. 33, 2008, no. 1, 281-293. * [KMV] R. Klén, V. Manojlović and M. Vuorinen: _Distortion of two point normalized quasiconformal mappings,_ arXiv:0808.1219[math.CV], 13 pp. * [KM] M. Knežević and M. Mateljević: On the quasi-isometries of harmonic quasiconformal mappings, J. Math. Anal. Appl. 334 (2007), 404-413. * [KP] V. Kojić and M. Pavlović: Subharmonicity of $\|f\|^{p}$ for quasiregular harmonic functions, with applications, J. Math. Anal. Appl. 342 (2008) 742-746 * [K] R. Kühnau, ed.: _Handbook of complex analysis: geometric function theory,_ Vol. 1 and Vol. 2, Elsevier Science B.V., Amsterdam, 2002 and 2005. * [LeVu] M. Lehtinen and M. Vuorinen: On Teichmüller’s modulus problem in the plane, Rev. Roumaine Math. Pures Appl. 33 (1988), 97–106. * [MV] V. Manojlović and M. Vuorinen: _On quasiconformal maps with identity boundary values,_ arXiv:0807.4418[math.CV], 16 pp. * [MAT] M. Mateljević: Distorsion of harmonic functions and harmonic quasiconformal quasi-isometry, Rev. Roumaine Math. Pures Appl. 51:5-6, 2006, 711-722. * [MAT1] M. Mateljević: Distorsion of harmonic functions and harmonic quasiconformal quasi-isometry 2, Manuscript 2008, 43 pp. * [MatV] M. Mateljević and M. Vuorinen: _On harmonic quasiconformal quasi-isometries,_ arXiv:0709.454[math.CV], 11 pp. * [P] I. Prause: _Flatness properties of quasispheres,_ Comput. Methods Funct. Theory 7 (2007), no. 2, 527–541. * [R] Yu. G. Reshetnyak: _Stability theorems in geometry and analysis,_ Translated from the 1982 Russian original by N. S. Dairbekov and V. N. Dyatlov, and revised by the author. Translation edited and with a foreword by S. S. Kutateladze. Mathematics and its Applications, 304\. Kluwer Academic Publishers Group, Dordrecht, 1994. xii+394 pp. ISBN: 0-7923-3118-4. * [RST] L. A. Rubel, A. L. Shields, and B. A. Taylor: _Mergelyan sets and the modulus of continuity of analytic functions_ , J. Approximation Theory 15 (1975), no. 1, 23–40. * [Se] P. Seittenranta: Möbius-invariant metrics, Math. Proc. Cambridge Philos. Soc. 125 (1999), no. 3, 511–533. * [S] V. I. Semenov: Estimate of stability, distortion theorems and topological properties of quasiregular mappings, Mat. Zametki 51 (1992), 109-113. * [SolV] A. Yu. Solynin and M. Vuorinen: Extremal problems and symmetrization for plane ring domains, Trans. Amer. Math. Soc. 348 (1996), 4095–4112. * [TA] P. M. Tamrazov: _Contour and solid structural properties of holomorphic functions of a complex variable_ (Russian), Uspehi Mat. Nauk 28 (1973), 131–161. English translation in Russian Math. Surveys 28 (1973), 141–173. * [V1] J. Väisälä: Lectures on n-Dimensional Quasiconformal Mappings, Lecture Notes in Math. 229, Springer-Verlag, Berlin, 1971. * [V2] J. Väisälä: Quasisymmetric embeddings in Euclidean spaces, Trans. Amer. Math. Soc. 264 (1981), no. 1, 191-204. * [Vu1] M. Vuorinen: Conformal invariants and quasiregular mappings, J. Anal. Math. 45 (1985), 69–115. * [Vu2] M. Vuorinen: Conformal geometry and quasiregular mappings, Lecture Notes in Math., 1319, Springer, Berlin, 1988. * [Vu3] M. Vuorinen: Conformally invariant extremal problems and quasiconformal maps, Quart. J. Math. Oxford Ser. (2), 43 (1992), 501–514. * [Vu4] M. Vuorinen: _Metrics and quasiregular mappings,_ Proc. Int. Workshop on Quasiconformal Mappings and their Applications, IIT Madras, Dec 27, 2005 - Jan 1, 2006, ed. by S. Ponnusamy, T. Sugawa and M. Vuorinen, _Quasiconformal Mappings and their Applications_ , Narosa Publishing House, 291–325, New Delhi, India, 2007. * [Vu5] M. Vuorinen: On Picard’s theorem for entire quasiregular mappings, Proc. Amer. Math. Soc. 107 (1989), no. 2, 383–394. * [Vu6] M. Vuorinen: A remark on the maximal dilatation of a quasiconformal mapping, Proc. Amer. Math. Soc. 92 (1984), no 4, 505–508. * [Vu7] M. Vuorinen: Quadruples and spatial quasiconformal mappings, Math. Z. 205 (1990), no 4, 617–628.	arxiv-papers	2008-08-24T10:57:11	2024-09-04T02:48:57.421346	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Vesna Manojlovic", "submitter": "Vesna Manojlovic", "url": "https://arxiv.org/abs/0808.3241" }
0808.3288	# First-order quantum correction to the ground-state energy density of two- dimensional hard-sphere Bose atoms Sang-Hoon Kima,b shkim@mmu.ac.kr Mukunda P. Dasb mpd105@rsphysse.anu.au a Division of Liberal Arts $\&$ Sciences, Mokpo National Maritime University, Mokpo 530-729, Republic of Korea b Department of Theoretical Physics, RSPhysSE, Institute of Advanced Studies, The Australian National University, Canberra, ACT 0200, Australia ###### Abstract Divergence exponents of the first-order quantum correction of a two- dimensional hard-sphere Bose atoms are obtained by an effective field theory method. The first-order correction to the ground-state energy density with respect to the zeroth-order is given by ${\cal E}_{1}/{\cal E}_{0}\sim\|D-2\|^{-\alpha}\|\ln\gamma\|^{-\alpha^{\prime}}$, where $D$ is the spatial dimension, and $\gamma$ is the gas parameter ($\gamma=na^{D}$). As $D\rightarrow 2$, $\alpha=\alpha^{\prime}=1$. We show that the first-order quantum correction of the energy density is not perturbative in low dimensions of $D<2.2$ regardless of any gas parameter which is much less that 1. Boson systems, interacting boson model, quantum field theory ###### pacs: 05.30.Jp, 21.60.Fw, 03.70.+k ## I Introduction Weakly interacting uniform Bose gas is a fundamental topic of many-body Bose systems and has been studied for more than five decades. Bose-Einstein condensation(BEC) is known to be the only phase transition that does not require interaction. However, in the real system an interaction exists even in very dilute Bose gases, and often the diluteness is explained by a D-dimensional gas parameter: $\gamma(D)=na^{D}\ll 1$, where $n$ is the D-dimensional density and $a$ is the s-wave scattering length. At zero temperature the chemical potential, energy density, number density, and speed of sound are written as an expansion in powers of the gas parameter $\gamma$. A fundamental approximation method of a weakly interacting Bose system was introduced by Bogoliubov in 1947 bogoliubov . In three-dimensions(3D) the ground-state energy density is given by $\small{\cal E}_{0}+{\cal E}_{1}+\cdots={\cal E}_{0}\left[1+c_{1}\sqrt{\gamma}+c_{2}\gamma\ln\gamma+\cdots\right].$ (1) Similarly, the number density in 3D is given by $\small n_{0}+n_{1}+\cdots=n_{0}\left[1+\frac{8}{3\sqrt{\pi}}\sqrt{\gamma}+\cdots\right].$ (2) ${\cal E}_{0}=2\pi\hbar^{2}an^{2}/m$ is the zeroth-order or mean-field energy density. The first-order quantum correction $c_{1}=128/15\sqrt{\pi}$ was first obtained by Lee and Yang in 1957 for a hard-sphere Bose gas lee . Later, it has been shown by Brueckner et al. brueckner ; beliaev that above results are generally true for any short-range potential with scattering length $a$. The second-order quantum correction $c_{2}=8(4\pi-3\sqrt{3})/3\sqrt{\pi}$ that was originated from a three-body interaction was obtained by Wu et al. in 1959 wu ; hugenholtz ; sawada . Recently, an effective field theory(EFT) method has been suggested to obtain the coefficients(Braaten and Nieto braaten ). This method was introduced by Georgi et al. in 1990s georgi ; kaplan ; manohar . In two-dimensions(2D) an uniform Bose gas is governed by strong long-range fluctuations. The fluctuations inhibit the formation of a true long-range order. Therefore, 2D uniform Bose gas does not undergo BEC transition at finite temperatures ketterle . However, this 2D system turns superfluid below a certain temperature while under confinement. The ground-state energy density of hard-sphere bosons in 2D has been studied by Schick in 1971 schick . $\small{\cal E}_{0}+{\cal E}_{1}+\cdots={\cal E}_{0}\left[1+{{\cal O}}(1/\ln\gamma)\right],$ (3) and number density is $\small n_{0}+n_{1}+\cdots=n_{0}\left[1+\frac{1}{\ln(1/\gamma)}+\cdots\right].$ (4) ${\cal E}_{0}=2\pi\hbar^{2}n^{2}/(m\|\ln\gamma\|)$ is the mean-field energy density. The first-order number density in 2D is finite as $n_{1}/n_{0}=1/\ln(1/\gamma)\ll 1$, but the first-order quantum correction of the ground-state energy density has not been obtained yet because an infrared divergence of an integral prevents a direct calculation. It has the divergence form of $\displaystyle\left\|\frac{{\cal E}_{1}}{{\cal E}_{0}}\right\|$ $\displaystyle\sim$ $\displaystyle\frac{1}{\|D-2\|^{\alpha}\|\ln\gamma\|^{\alpha^{\prime}}},$ (5) where the exponents $\alpha$ and $\alpha^{\prime}$ are positive constants. In this paper, we will calculate the two exponents of divergence $\alpha$ and $\alpha^{\prime}$ analytically. The EFT has an advantage that it works even in non-integer dimensions. Therefore, we can rewrite the EFT in non-integer dimensions between two and three. Then, taking the limit of $D=\lim_{\epsilon\rightarrow o^{+}}2(1+\epsilon)$, we will obtain the two exponents of divergence $\alpha$ and $\alpha^{\prime}$. Furthermore, since the first-order quantum correction of the ground-state energy density converges in D=3 but diverges in D=2, we may find the dimensional border of the perturbative method which is effective between two and three. The paper is organized as follows: In Section II, we review the basic structure of EFT of a dilute Bose system in non-integer dimensions between two and three braaten ; andersen ; rakimov . In Section III, we reproduce the zeroth-order or mean-field results in D-dimensions. In section VI, we obtain the first-order quantum correction in 2D and two exponents $\alpha$ and $\alpha^{\prime}$. Also we interpolate the first-order quantum depletion between 2D and 3D. Finally, in section V we summarize the results. ## II Effective field theory by an effective Lagrangian EFT is a general approach that can be used to analyze the low energy behavior of a physical system and takes advantage of the separation of scales to make model-independent predictions georgi ; kaplan ; manohar . The effective lagrangian that describes the low energy physics is written in terms of only the long-wavelength degrees of freedom that includes every non-renormalizable interactions, but to a certain order in a low energy expansion, only a finite number of operators contribute to a physical quantity andersen . An uniform Bose gas can be described by field theoretic method with Hamiltonian ${{\cal H}}-\mu N$, where $\mu$ is the chemical potential and $N$ is the number operator. The number of atoms are conserved by the phase transformation: $\psi({\bf r},t)\rightarrow e^{i\theta}\psi({\bf r},t)$. The free energy in ground-state of the system is $\small{\cal F}(\mu)=\langle{{\cal H}}-\mu N\rangle_{\mu},$ (6) where $\langle\cdots\rangle_{\mu}$ denotes the expectation value in the ground state. Therefore, $\langle{\cal H}\rangle_{\mu}={\cal E}(\mu)$ and $\langle N\rangle_{\mu}=n(\mu)$ are the energy and number in the ground state. The free energy ${\cal F}$ is given by all connected vacuum diagrams that are Feynman diagrams with no external legs braaten . The number density $n$ is given by the expectation value of $(\psi^{\dagger}\psi)$ in the ground-state (we set $\hbar=1$ for convenience) $\displaystyle n(\mu)$ $\displaystyle=$ $\displaystyle\frac{1}{Z}\int{\cal{D}\psi^{\dagger}\cal{D}\psi(\psi^{\dagger}\psi)}\,e^{iS[\psi^{\dagger},\psi]}$ (7) where $\psi^{\dagger}$ and $\psi$ are complex field operators of bosons, $S$ is the action from the Lagrangian braaten , and $Z$ is the grand-canonical partition function given by $\small Z=\int{\cal{D}\psi^{\dagger}\cal{D}\psi}\,e^{iS[\psi^{\dagger},\psi]}.$ (8) The grand-canonical ensemble of uniform Bose particles with a two-body interaction is governed by the action $\small S[\psi^{\dagger},\psi]=\int dt\int d^{D}x\left[\psi^{\dagger}\left(i\frac{\partial}{\partial t}+\frac{\mbox{\boldmath$\nabla$}^{2}}{2m}+\mu\right)\psi-\frac{1}{2}g\left(\psi^{\dagger}\psi\right)^{2}\right],$ (9) where $m$ is the atomic mass, and $g$ is the D-dimensional coupling constant which contains pairwise interaction between atoms. In D-dimensions $g$ is given by nogueira $\small g(D)=\frac{4\pi^{D/2}a^{D-2}}{m}\frac{1}{2^{2-D}\Gamma(1-\frac{D}{2})\gamma^{D/2-1}+\Gamma(\frac{D}{2}-1)},$ (10) where $\Gamma$ is the Gamma function. It satisfies the two limiting cases. In 3D, it has the well-know form $\small g(3)=\frac{4\pi a}{m}.\\\ $ (11) In 2D, it has the logarithmic form kim ; lieb $\small g(2)=\frac{4\pi}{m}\frac{1}{\ln(1/\gamma)}.$ (12) When the temperature of a Bose system falls below the condensation temperature $T_{c}$, we can write the quantum field $\psi$ in terms of a time-independent condensate $\phi$ and a quantum fluctuation field $\tilde{\psi}$ $\small\psi=\phi+\tilde{\psi}.$ (13) The fluctuation field $\tilde{\psi}$ can be conveniently written in terms of two real fields $\psi_{1}$ and $\psi_{2}$ $\small\tilde{\psi}=\frac{\psi_{1}+i\psi_{2}}{\sqrt{2}}.$ (14) In the uniform system $\phi$ is the condensate order parameter and a real constant. It corresponds to a breaking of the global $U(1)$ symmetry and takes into account the Bogoluibov shift of the field operator. The conservation of particle numbers requires that $\tilde{\psi}$ has non-zero momentum component so that $\langle\psi\rangle=\phi$ and $\langle\tilde{\psi}\rangle=0$. Therefore, the condensate order parameter $\phi$ defines the density of condensed particles while $\tilde{\psi}$ defines the density of uncondensed particles. Then, the zeroth- and first-order quantum correction to the density are given by $\small n_{0}=\phi^{2},\quad n_{1}=\langle{\tilde{\psi}}^{\dagger}\tilde{\psi}\rangle.$ (15) Substituting Eqs. (13) and (14) into Eq. (9) of the action $S$, we decompose the real part of the action into three parts braaten ; andersen ; rakimov : $\small S[\phi,\psi_{1},\psi_{2}]=S_{clas}[\phi]+S_{free}[\phi,\psi_{1},\psi_{2}]+S_{int}[\phi,\psi_{1},\psi_{2}].$ (16) $S_{clas}$ is the classical part of the action. It does not contain any filed operator: $\small S_{clas}[\phi]=\int dt\int d^{D}x\left(\mu\phi^{2}-\frac{1}{2}g\phi^{4}\right).$ (17) $S_{free}$ is the free part of the action. It is quadratic in $\psi_{1}$ and $\psi_{2}$: $\small S_{free}[\phi,\psi_{1},\psi_{2}]=\int dt\int d^{D}x\left[\frac{1}{2}(\dot{\psi}_{1}\psi_{2}-\psi_{1}\dot{\psi}_{2})+\frac{1}{2}\psi_{1}\left(\frac{\mbox{\boldmath$\nabla$}^{2}}{2m}+X\right)\psi_{1}+\frac{1}{2}\psi_{2}\left(\frac{\mbox{\boldmath$\nabla$}^{2}}{2m}+Y\right)\psi_{2}\right],$ (18) where $\dot{\psi}\equiv\partial\psi/\partial t$, and the two new variable $X$ and $Y$ are $\displaystyle X$ $\displaystyle=$ $\displaystyle\mu-3g\phi^{2},$ (19) $\displaystyle Y$ $\displaystyle=$ $\displaystyle\mu-g\phi^{2}.$ (20) The terms $3g\phi^{2}$ and $g\phi^{2}$ in the $X$ and $Y$ are mean-field self- energies of the system. $S_{int}$ is the interaction part of the action. It is the remaining terms: $\small S_{int}[\phi,\psi_{1},\psi_{2}]=\int dt\int d^{D}x\left[\sqrt{2}J\psi_{1}+\frac{K}{\sqrt{2}}\psi_{1}(\psi_{1}^{2}+\psi_{2}^{2})-\frac{g}{8}(\psi_{1}^{2}+\psi_{2}^{2})^{2}\right],$ (21) where $\small J=\phi(\mu-g\phi^{2})=\phi Y,$ (22) and $\small K=-g\phi.$ (23) The free part of the action in Eq. (18) gives rise to a propagator, which can be used in perturbative framework. If we take the Fourier Transform in a momentum space, the fluctuating part is given by $\small\tilde{\psi}(t,{\bf r})=\displaystyle\frac{1}{\sqrt{V}}\sum_{n=-\infty}^{\infty}\sum_{p}\tilde{\psi}(\omega_{n},{\bf p})\exp\\{-i\omega_{n}t+i{\bf p}\cdot{\bf r}\\},$ (24) where $V^{-1}\displaystyle{\sum_{p}}=\int d^{D}p/(2\pi)^{D}$, and $\omega_{n}=2\pi nT$ is the Matsubara frequency. Therefore, the propagator of the free action $S_{free}$ is written as $\displaystyle G(\omega,p)=\displaystyle\frac{1}{\omega^{2}-E_{p}^{2}+iE_{p}}\left(\begin{array}[]{cc}\frac{p^{2}}{2m}-Y&-i\omega\\\ i\omega&\frac{p^{2}}{2m}-X\end{array}\right),$ (27) with the dispersion relation $\small E_{p}=\sqrt{\left(\frac{p^{2}}{2m}-X\right)\left(\frac{p^{2}}{2m}-Y\right)}.$ (28) This is a general form of the dispersion and includes every information of the energy spectrum in the two self-terms $X$ and $Y$. It is possible to diagonalize the propagator matrix in Eq. (27) by a field redefinition or renormalization, which is equivalent to the Bogoliubov transformation in the operator method. However, such a redefinition makes the interaction terms in the action more complicated and increases the number of diagrams. Therefore, we use the propagator matrix with off-diagonal elements to minimize the number of diagrams. ## III Thermodynamic potential and zeroth-order results It is convenient to introduce the thermodynamic potential $\Omega(\mu,\phi)$. The thermodynamic potential contains the information required to determine all of the thermodynamic functions. The free energy ${\cal F}(\mu)$ can be obtained by evaluating $\Omega$ at a particular value of $\phi$. The sum of the vacuum graphs is independent of the arbitrary background condensate $\phi$. Thus, the sum of connected vacuum diagram reduces to the sum of one-particle irreducible vacuum diagrams braaten ; andersen ; rakimov : $\small{\cal F}(\mu)=\Omega(\mu,\phi(\mu))$ (29) From Eq. (6) the ground-state energy density is written by $\small{\cal E}(\mu)={\cal F}(\mu)+n\mu.$ (30) By differentiating the free energy, we obtain the density and chemical potential, too. $\displaystyle n(\mu)=-\frac{d{\cal F}}{d\mu},\quad\mu(n)=\frac{d{\cal E}}{dn}.$ (31) The $n$-loop contribution to the $\Omega$ is denoted by $\Omega_{n}$ in FIG. 1. It is given by all one-particle irreducible vacuum diagrams and can be expanded in the number of loops: $\small\Omega(\mu,\phi)=\Omega_{0}(\mu,\phi)+\Omega_{1}(\mu,\phi)+\Omega_{2}(\mu,\phi)+\cdots.$ (32) If $\Omega$ is evaluated at a value of the condensate, all one-particle reducible diagrams vanish. Then, the free energy density in Eq. (29) is $\small{\cal F}(\mu)=\Omega_{0}(\mu,\phi)+\Omega_{1}(\mu,\phi)+\Omega_{2}(\mu,\phi)+\cdots.$ (33) Figure 1: Loop diagram of the two-body interaction. The zeroth, first, and second loop from the left to right Using Eqs. (13) and (14), the condition of the condensate reduces to $\langle\psi_{1}\rangle=\langle\psi_{2}\rangle=0$. This condition is equivalent to $\small\frac{\partial\Omega(\mu,\phi)}{\partial\phi}=\frac{\partial\Omega_{0}(\mu,\phi)}{\partial\phi}+\frac{\partial\Omega_{1}(\mu,\phi)}{\partial\phi}+\frac{\partial\Omega_{2}(\mu,\phi)}{\partial\phi}+\cdots=0$ (34) The number density in Eq. (31) is obtained from the thermodynamic potential, too. $\small n(\mu)=-\frac{\partial\Omega}{\partial\mu}.$ (35) The free energy can also be expanded in powers of quantum corrections around the mean-field value ${\cal F}_{0}(\mu)$: $\small{\cal F}(\mu)={\cal F}_{0}(\mu)+{\cal F}_{1}(\mu)+{\cal F}_{2}(\mu)+\cdots.$ (36) The loop expansion in Eq. (33) does not coincide with the above expansion of ${\cal F}(\mu)$ in powers of quantum corrections because of its independence of $\phi$. To obtain the expansion of ${\cal F}$ in powers of quantum corrections, we must expand the condensate $\phi$ around its classical minimum $\phi_{0}$, which satisfies $\small\frac{\partial\Omega_{0}(\mu,\phi_{0})}{\partial\psi}=0.$ (37) By expanding Eq. (34) in powers of $\phi-\phi_{0}$, and solving for $\phi$, we obtain the quantum expansion for the condensate: $\small\phi=\phi_{0}+\phi_{1}+\phi_{2}+\cdots,$ (38) where $\phi_{n}$ is the $n$th-order quantum correction. For instance, the first-order quantum correction for the condensate $\phi_{1}$ is obtained from the expanding of Eq. (37) around $\phi_{0}$. The nonzero terms are $\small\frac{\partial\Omega_{0}(\mu,\phi)}{\partial\phi}=\frac{\partial^{2}\Omega_{0}(\mu,\phi_{0})}{\partial\phi^{2}}(\phi-\phi_{0})+\frac{\partial\Omega_{1}(\mu,\phi_{0})}{\partial\phi}+\cdots=0.$ (39) Then, $\phi_{1}$ in Eq. (38) is written as $\small\phi_{1}=-\left[\frac{\partial\Omega_{1}(\mu,\phi_{0})}{\partial\phi}\right]\left[\frac{\partial^{2}\Omega_{0}(\mu,\phi_{0})}{\partial\phi^{2}}\right]^{-1}.$ (40) Therefore, keeping up to the terms of the second order and $\phi-\phi_{0}\approx\phi_{1}$, the first three terms of the free energy density is $\displaystyle{\cal F}_{0}(\mu)$ $\displaystyle=$ $\displaystyle\Omega_{0}(\mu,\phi_{0}),$ (41) $\displaystyle{\cal F}_{1}(\mu)$ $\displaystyle=$ $\displaystyle\Omega_{1}(\mu,\phi_{0}),$ (42) $\displaystyle{\cal F}_{2}(\mu)$ $\displaystyle=$ $\displaystyle\Omega_{2}(\mu,\phi_{0})+\frac{\partial\Omega_{1}(\mu,\phi_{0})}{\partial\phi}\phi_{1}+\frac{1}{2}\frac{\partial^{2}\Omega_{0}(\mu,\phi_{0})}{\partial\phi^{2}}\phi_{1}^{2}.$ (43) The mean-field thermodynamic potential is given by the terms in the classical action in Eq. (17) as $\small\Omega_{0}(\mu,\phi)=-\frac{1}{V}S_{clas}=-\mu\phi^{2}+\frac{1}{2}g\phi^{4}.$ (44) From Eq. (37) we obtain the classical minimum $\phi_{0}$ as $\small\phi_{0}=\sqrt{\frac{\mu}{g}}$ (45) Therefore, the mean-field free energy density in D-dimensions is obtained from Eq. (41) $\small{\cal F}_{0}(\mu)=\Omega_{0}(\mu,\phi_{0})=-\frac{\mu^{2}}{2g}.$ (46) The mean-field number density is obtained from Eq. (7) as $\small n(\mu_{0})=-\frac{\partial{\cal F}(\mu_{0})}{\partial\mu}=\frac{\mu_{0}}{g}$ (47) The mean-field ground-state energy density in D-dimensions is obtained from Eq. (30) $\small{\cal E}_{0}^{(D)}=-\frac{\mu_{0}^{2}}{2g}+n\mu_{0}=\frac{1}{2}g(D)n^{2}.$ (48) In 3D, since $g(3)=4\pi a/m$ from Eq. (11), the mean-field energy density is $\small{\cal E}_{0}^{(3)}=\frac{2\pi an^{2}}{m}.$ (49) In 2D, since $g(2)=4\pi/m\ln(1/\gamma)$ from Eq. (12), the mean-field energy density is $\small{\cal E}_{0}^{(2)}=\frac{2\pi an^{2}}{m\ln(1/\gamma)}.$ (50) Therefore, this method reproduced the well-known mean-field energy densities in Eqs. (1) and (3) successfully. ## IV First-order quantum correction The first-order quantum correction of the ground-state energy is obtained in the following way. Substituting Eq. (45) into Eqs. (19) and (20), we obtain the two variables $X$ and $Y$ at the minimum of $\phi_{0}$. $\displaystyle X$ $\displaystyle=$ $\displaystyle\mu-3g\phi_{0}^{2}=-2\mu,$ (51) $\displaystyle Y$ $\displaystyle=$ $\displaystyle\mu-g\phi_{0}^{2}=0.$ (52) The propagator and the dispersion relation in Eqs. (27) and (28) becomes $\displaystyle G(\omega,p)=\displaystyle\frac{1}{\omega^{2}-E_{p}^{2}+iE_{p}}\left(\begin{array}[]{cc}\frac{p^{2}}{2m}&-i\omega\\\ i\omega&\frac{p^{2}}{2m}+2\mu\end{array}\right),$ (55) and from Eq. (47) $\small E_{p}=\sqrt{\frac{p^{2}}{2m}\left(\frac{p^{2}}{2m}+2\mu\right)}.$ (56) These are the original Bogoliubov results. $E_{p}$ is gapless and is linear for small wave-vectors. For large wave-vectors, the dispersion relation becomes $E_{p}\simeq p^{2}/2m+2\mu\simeq p^{2}/2m+2g(D)n.$ Therefore, the $2g(D)n$ represents the mean-field energy in D-dimensions due to interaction with the condensed particles. In the Bogoliubov approximation one makes a pair approximation to the Hamiltonian by neglecting terms with three and four operators bogoliubov . On the other hand, in the Beliaev approximation, one goes one step further by calculating the leading quantum corrections to the quasi-particle spectrum beliaev . This is done by including all one-loop diagrams. The free energy can be written with the inverse of the propagator. From Eq. (8) $\small Z={\rm exp}\left\\{i\int dt\int d^{D}x\,{\rm det}G^{-1}(\omega,p)\right\\}.$ (57) Therefore, $\small{\cal F}={\rm Tr}\ln G^{-1}(\omega,p).$ (58) Note that $\ln\rm{det}{\bf A}={\rm Tr}\ln{\bf A}$ for any matrix ${\bf A}$. The one-loop contribution to the thermodynamic potential in D-dimensions, $\Omega_{1}=i\ln Z$, is obtained by the following way. $\displaystyle\Omega_{1}(\mu,\phi)$ $\displaystyle=$ $\displaystyle-\frac{i}{2}\int\frac{d\omega}{2\pi}\int\frac{d^{D}p}{(2\pi)^{D}}\ln\det G(\omega,p)^{-1}$ (59) $\displaystyle=$ $\displaystyle\frac{1}{2}\int\frac{d^{D}p}{(2\pi)^{D}}E_{p}.$ From Eq. (42), we have the free energy component $\small{\cal F}_{1}(\mu)=\Omega_{1}(\mu,\phi_{0})=\frac{1}{4m}I_{0,-1}(\kappa),$ (60) where $\kappa=4m\mu=4mgn$ and $I_{i,j}$ are loop integral function in $D$-dimensions, which is defined as braaten ; andersen $\displaystyle I_{i,j}^{(D)}(\kappa)$ $\displaystyle=$ $\displaystyle\int\frac{d^{D}p}{(2\pi)^{D}}\frac{p^{2i}}{p^{j}(p^{2}+\kappa)^{j/2}},$ (61) where $i$ and $j$ are integers. This $I_{i,J}$ is expressed by Gamma function, too. $\small I_{i,j}^{(D)}=\frac{1}{(4\pi)^{D/2}}\frac{\Gamma\left(\frac{D+2i-j}{2}\right)\Gamma\left(-\frac{D+2i-2j}{2}\right)}{\Gamma\left(\frac{D}{2}\right)\Gamma\left(\frac{j}{2}\right)}\kappa^{\frac{D+2i-2j}{2}}.$ (62) Note that in D=3, $I_{0,-1}^{(3)}(\kappa)=\kappa^{5/2}/15\pi^{2}$, $I_{1,1}^{(3)}(\kappa)=\kappa^{3/2}/3\pi^{2}$, and $I_{-1,-1}^{(3)}(\kappa)=-\kappa^{3/2}/6\pi^{2}$. It has the useful relation of the first derivative $\small\frac{d}{d\kappa}I_{i,j}^{(D)}(\kappa)=-\frac{j}{2}I_{i+1,j+1}^{(D)}(\kappa).$ (63) Ground-state energy density is expressed as $\displaystyle{\cal E}_{0}+{\cal E}_{1}+\cdots=\frac{1}{2}gn^{2}+\frac{1}{4m}I^{(D)}_{0,-1}(\kappa)+\cdots.$ (64) Therefore, the contribution of the first-order quantum correction in D-dimensions is obtained as $\displaystyle\frac{{\cal E}_{1}}{{\cal E}_{0}}$ $\displaystyle=$ $\displaystyle-\frac{\Gamma\left(-\frac{D+2}{2}\right)\Gamma\left(\frac{D+1}{2}\right)}{\pi^{(D+1)/2}\Gamma\left(\frac{D}{2}\right)}\left\\{mg(D)\right\\}^{D/2}n^{D/2-1}$ (65) $\displaystyle=$ $\displaystyle-\frac{2^{D}\pi^{(D^{2}-2D-2)/4}\Gamma\left(-\frac{D+2}{2}\right)\Gamma\left(\frac{D+1}{2}\right)\gamma^{D/2-1}}{\Gamma\left(\frac{D}{2}\right)\left\\{2^{2-D}\Gamma\left(1-\frac{D}{2}\right)\gamma^{D/2-1}+\Gamma\left(\frac{D}{2}-1\right)\right\\}^{D/2}}.$ Note that $\gamma=(n_{0}+n_{1}+\cdots)a^{D}$. Figure 2: The first-order quantum correction of the ground-state energy density in D-dimensions in the logarithmic scale. The divergence in D=2 is clear. $\gamma$ itself is a function of the dimensions. Substituting D=3, we obtain the well-known first-order quantum correction of the ground-state energy density in 3D as $\small\frac{{\cal E}_{1}}{{\cal E}_{0}}=\frac{128}{15\sqrt{\pi}}\sqrt{\gamma}.$ (66) Taking the limit of $D=\lim_{\varepsilon\to 0^{+}}2(1+\varepsilon)$, we obtain the first-order quantum correction of the ground-state energy density in 2D as $\small\frac{{\cal E}_{1}}{{\cal E}_{0}}\simeq\frac{\gamma^{\varepsilon}}{-\gamma^{\varepsilon}+1}\simeq\frac{1}{\varepsilon\ln(1/\gamma)}\gg 1.$ (67) For the singularities of the $\Gamma$ function at non-positive argument, we used the following expansions $\small\Gamma(-n+\varepsilon)=\frac{(-1)^{n}}{n}\left[\frac{1}{\varepsilon}+h(n+1)+\cal{O}(\varepsilon)\right],$ (68) where $h(n+1)=1+1/2+1/3+\cdots+1/n-\gamma_{E}$ and $\gamma_{E}=0.5772...$ is the Euler constant. In particular, $\Gamma(\varepsilon)=1/\varepsilon-\gamma_{E}$. Also, we used the approximations: $\gamma^{\varepsilon}\simeq 1+\varepsilon\ln\gamma$. In 2D there is a divergence, and we can obtain the two exponents of the divergence in Eq. (5) as $\alpha=\alpha^{\prime}=1$. We plotted ${\cal E}_{1}/{\cal E}_{0}$ in FIG. 2 as the functions of the dimensions and gas parameter in the logarithmic scale. From the FIG. 2 it is clear that the expansion in Eq. (3) satisfy the perturbative condition of ${\cal E}_{1}/{\cal E}_{0}\ll 1$ only when $D>2.2$. In low dimensions of $D<2.2$, the first-order ground-state energy density is not perturbative regardless of the magnitude of the gas parameter. The number density is obtained from Eqs. (13)-(15) as $\small n=\phi^{2}+\frac{1}{2}\langle\psi_{1}^{2}+\psi_{2}^{2}\rangle.$ (69) Therefore, taking the one-loop effects into account $\displaystyle n_{0}+n_{1}$ $\displaystyle=$ $\displaystyle n_{0}+\frac{i}{2}\int\frac{d\omega}{2\pi}\frac{d^{D}p}{(2\pi)^{D}}\left[\frac{p^{2}/2m+2mE_{p}^{2}/p^{2}}{\omega^{2}-E_{p}^{2}+iE_{p}}\right]$ (70) $\displaystyle=$ $\displaystyle n_{0}+\frac{1}{4}I_{1,1}(4mgn)+\frac{1}{4}I_{-1,-1}(4mgn)$ $\displaystyle=$ $\displaystyle n_{0}\left[1+\frac{\Gamma\left(-\frac{D}{2}\right)\left\\{2\Gamma\left(\frac{D+1}{2}\right)-\Gamma\left(\frac{D-1}{2}\right)\right\\}}{2^{3}\pi^{(D+2)/2}\Gamma\left(\frac{D}{2}\right)}(mg)^{D/2}n_{0}^{(D-2)/2}\right].$ Then, the first-order quantum correction of the number density in D-dimensions is obtained as $\displaystyle\frac{n_{1}}{n_{0}}$ $\displaystyle=$ $\displaystyle\frac{\pi^{(D^{2}-2D-2)/4}}{2^{3-D}}\frac{\Gamma\left(-\frac{D}{2}\right)\left\\{2\Gamma\left(\frac{D+1}{2}\right)-\Gamma\left(\frac{D-1}{2}\right)\right\\}}{\Gamma\left(\frac{D}{2}\right)\left\\{2^{2-D}\Gamma\left(1-\frac{D}{2}\right)\gamma^{D/2-1}+\Gamma\left(\frac{D}{2}-1\right)\right\\}^{D/2}}\gamma^{D/2-1}.$ (71) In 3D we obtain the well-known first-order quantum depletion as $\small\frac{n_{1}}{n_{0}}=\frac{8}{3\sqrt{\pi}}\sqrt{\gamma}.$ (72) In 2D taking the limit of $D=\lim_{\varepsilon\to 0^{+}}2(1+\varepsilon)$, we obtain the first-order quantum depletion as $\displaystyle\frac{n_{1}}{n_{0}}$ $\displaystyle\simeq$ $\displaystyle\frac{\gamma^{\varepsilon}}{-\frac{1}{\varepsilon}\gamma^{\varepsilon}+\frac{1}{\varepsilon}}.$ (73) $\displaystyle\simeq$ $\displaystyle\frac{1}{\ln(1/\gamma)}.$ We used the approximations: $\Gamma(-\frac{D}{2})\sim 1/\varepsilon$ and $2\Gamma(\frac{D+1}{2})-\Gamma(\frac{D-1}{2})\sim 2\sqrt{\pi}\varepsilon$ . This result is the same as the Schick’s value in Eq. (4). We plotted $n_{1}/n_{0}$ as a function of dimensions and gas parameter in Fig. 3. It is increasing monotonically as the dimension is decreasing from 3 to 2. The depletion is stronger in low dimensions. Figure 3: The first-order quantum correction of the number density at $\gamma=10^{-4}$ as a function the dimensions. ## V Summary EFT is a theory of symmetries. Once the symmetries have been identified, one writes down the most general local effective Lagrangian consistent with these symmetries. At zero temperature, the symmetries are Galilean invariance, time- reversal symmetries, and the global phase symmetry. These symmetries restricts the possible terms in the effective action. The first-order quantum correction to the ground-state energy density of a 2D uniform Bose atoms has the form of the divergence: ${\cal E}_{1}/{\cal E}_{0}\sim\|D-2\|^{-\alpha}\|\ln\gamma\|^{-\alpha^{\prime}}$. Applying an effective field theory method to the hard-sphere boson system in D-dimensions, we obtained the D-dimensional free energy and thermodynamic potential up to the one-loop result. This general form in D-dimensions reproduced every well- known result in 3D and 2D. Then, taking the limit of $D=\lim_{\varepsilon\to 0^{+}}2(1+\varepsilon)$, we obtained the zeroth- and first-order quantum correction in 2D. Finally, we obtained the two exponents of divergence $\alpha$ and $\alpha^{\prime}$ analytically as $\alpha=\alpha^{\prime}=1$. We also showed that the perturbative expression of the ground-state energy density is effective only in low dimensions of $D<2.2$. ###### Acknowledgements. This work is supported by Korea Science and Engineering Foundation(KOSEF) and Australian Academy of Science(AAS) scientific exchange program in 2008. ## References * (1) N. N. Bogoliubov, J. Phys. (U.S.S.R.) 11, pp. 23-32 (1947). * (2) T. D. Lee and C. N. Yang, Phys. Rev. 105, pp. 1119-1120 (1957). * (3) K. A. Brueckner, K. Sawada. Phys. Rev. 106, pp. 1117-1127 (1957). * (4) S. T. Beliaev, Sov. J. Phys. 7, pp. 289-299 (1958). * (5) T. T. Wu, Phys. Rev. 115, pp. 1390-1404 (1959). * (6) N.M. Hugenholtz and D. Pines, Phys. Rev. 116, pp. 489-506 (1959). * (7) K. Sawada, Phys. Rev. 116, 1344-1358 (1959). * (8) E. Braaten and A. Nieto, Euro. Phys. J. B 11, pp. 143-159 (1999). * (9) H. Georgi, Ann. Rev. Nucl. Part. Sci. 43, pp. 209-252 (1993). * (10) D. B. Kaplan, nucl-th/9506035 (1995). * (11) A. V. Manohar, Effective field theories, in Perturbative and Nonperturbative Aspects of Quantum Field Theory, Ed. by H. Latal and W. Schweiger (Springer-Verlag, 1997). * (12) A. Görlitz, J. M. Vogels, A. E. Leanhardt, C. Raman, T. L. Gustavson, J. R. Abo-Shaeer, A. P. Chikkatur, S. Gupta, S. Inouye, T. Rosenband, and W. Ketterle, Phys. Rev. Lett. 87, pp. 130402:1-4 (2001). * (13) M. Schick, Phys. Rev. A, 3, pp. 1067-1073 (1971). * (14) J. O. Andersen, Rev. Mod. Phys. 76, pp. 599-639 (2004). * (15) A. Rakhimov, C. K. Kim, S.-H. Kim, and J. H. Yee, Phys. Rev. A77, pp. 033626:1-9 (2008). * (16) F. S. Nogueira and H. Kleinert, Phys. Rev. B73, pp. 104515:1-9 (2006). There is a factor of 2 difference in the exponent of $\gamma$ of $g(D)$ in Eq. (10). We took $pa\rightarrow\sqrt{na^{2}}$ instead Nogueira’s Choice of $pa\rightarrow na^{2}$, where $p$ is the momentum. * (17) S.-H. Kim, C. Won, S. D. Oh, and W. Jhe, cond-mat/9904087 (1999). * (18) E. H. Lieb, R. Seiringer, and J. Yngvason, Commun. Math. Phys. 224, pp. 17-31 (2001).	arxiv-papers	2008-08-25T01:34:59	2024-09-04T02:48:57.431684	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Sang-Hoon Kim, Mukunda P. Das", "submitter": "Sang-Hoon Kim", "url": "https://arxiv.org/abs/0808.3288" }
0808.3383	# The Ara OB 1a Association Scott J. Wolk Fernando Comerón Tyler Bourke ###### Abstract The Ara OB1a association is one of the closest sites where triggered star formation is visible for multiple generations of massive stars. At about 1.3 kpc distance, it contains complex environments including cleared young clusters, embedded infrared clusters, CO clouds with no evidence of star formation, and clouds with evidence of ongoing star formation. In this review we discuss the research on this region spanning the last half-century. It has been proposed that the current configuration is the result of an expanding wave of neutral gas set in motion between 10–40 million years ago in combination with photoionization from the current epoch. Harvard–Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138; USA European Southern Observatory,Karl-Schwarzschild-Strasse 2, 85748 Garching, Germany Harvard–Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138; USA ## 1\. Historical Summary (1960–1985) Ara OB1a, while somewhat enigmatic, may be one of the best examples of triggered star formation in the local Galaxy. The triggering in the most active portion is easily imagined from images of the $<$ 3 Myr NGC 6193/RCW 108-IR complex such as Figure 1. However the full association may be as much as 50 Myr in age and cover several degrees on the sky. Ara OB1 was first studied in detail by Whiteoak (1963). He used photographic photometry and objective prism spectroscopy to identify about 35 O and B star members. Whiteoak’s work was a follow up to an H$\alpha$ survey by Rodgers, Campbell & Whiteoak (RCW; 1960), which cataloged 181 H$\alpha$ emission regions in the southern sky, including RCW 108. NGC 6193 is an open cluster, discovered by James Dunlop in 1826, that is dominated by a pair of O stars, HD 150135 and HD 150136. These are the brightest optically revealed O stars in the association and are thought to be responsible for ionizing the bright rim of emission to the west (NGC 6188, discovered by John Herschel in 1836) which separates NGC 6193 from RCW 108–IR. The youth of the region is clear in sky survey plates and Figure 1 which show concurrent regions of ionized gas and dust lanes. Whiteoak noted two additional clusters in the region, NGC 6204 – about 2 degrees to the northeast – and NGC 6167 – about 1 degree to the southwest. Ara OB1a is a compact association covering about 1 sq. degree around a central cluster – NGC 6193. It is generally thought to be about 1.3 kpc away and can be equated with RCW 108. There is some confusion in the literature as to what is actually meant by RCW 108. The original definition of Rodgers, Campbell & Whiteoak (1960) refers to all the region where H$\alpha$ nebulosity is detected, for which they give a size of 210′x 120′centered at (l,b = 336.49, –1.48). Straw et al. (1987) used RCW 108–IR to refer to the embedded IR cluster about 15′ to the west of the O stars in NGC 6193 and which is identified with IRAS 16362–4845. The confusion arises when RCW 108-IR is abbreviated by dropping the ”–IR”. For the remainder of this paper we will refer to the embedded cluster as RCW 108–IR or IRAS 16362–4845. Moffat & Vogt (1973) measured photometry for 13 stars within 4′ of the center of NGC 6193 and found EB-V= 0.4 and a distance of 1360 pc. Herbst & Havlen (1977) describe Ara OB1 as having a “diamond ring” appearance, with RCW 108–IR as the “diamond” and a thin circular dust ring making up almost half of the “ring”. They performed photoelectric and photographic photometry on 702 stars in the region. Herbst (1974, 1975) identified parts of this region as an “R association” as there were three early type stars associated with reflection nebulosity. These stars may mark a separate site of star formation within Ara OB1a. We will discuss this further in Section 6. Mid–infrared MSX & IRAS maps of the region show several condensations. The brightest is coincident with RCW 108–IR and two apparently related peaks IRAS 16379–4856 and IRAS 16348–4849. One of the earliest radio studies of the region was in the survey by Shaver & Goss (1970). They made a 5 GHz and 408 MHz survey of over 250 Galactic radio source including RCW 108. RCW 108 was remarkable for having one of the smallest emission regions – unresolved at 3′ and having a relatively high density and temperature of the electron population. In addition to Ara OB1a, Whiteoak (1963) identified a background O star cluster coincident with NGC 6193 but with a different distance modulus. While the foreground Ara OB1a has a distance modulus of 10.5$\pm 0.6$ the second group of about 13 O and B stars (Ara OB1b) has a distance modulus of 12.7 $\pm 0.5$ or about 3500 pc. The central O stars of the two associations are offset by about 2o along the Galactic plane (Humphreys 1978), so this is a concern for membership determination. Figure 1.: Salient features of the most active portion of Ara OB1a. This image is a B,V H$\alpha$ (blue, green and red respectively) from the wide field imager on the MPI-ESO 2.2m telescope. The field of view is 32′x 32′ or about 12 pc on a side. ## 2\. NGC 6193 NGC 6193 is the open cluster to the east of the bright emission rim NGC 6188. In this region, Herbst & Havlen (1977) measured over 700 stars photometrically. Of these, 59 had photometric distances and reddening consistent with cluster membership. This appears to be complete to earlier than A0. This survey covered an area about 40′ on a side. They found a slightly steeper than usual reddening law. Their best-fit photometric distance was 1320$\pm 120$ pc. Several other groups have estimated the distance to the cluster using photometric parallax techniques and obtain results ranging from $\approx$1100 to $\approx$ 1400 pc (Moffat & Vogt 1973, Fitzgerald 1987, Kalcheva & Georgiev 1992). Vazquez & Feinstein (1992) obtained an age of about 3 Myr by fitting the upper main sequence, they also determined a distance of about 1410 $\pm 120$ pc using an R = 3.1 reddening law. They noted that the steep reddening found by Herbst & Havlen (1977) was due to a binarity-induced color shift. Extinction is low (A${}_{V}\sim 0.5$). It has been difficult to obtain a good catalog of stellar members. The 59 stars in Herbst & Havlen (1977) are only identified on a star chart. Similar practices were used by Moffat & Vogt (1973), Arnal et al. (1988) and Vazquez & Feinstein (1992). The deepest published study of NGC 6193 is the $Chandra$ zero order image of the $Chandra$ gratings observation of HD 150135. HD 150136 is a remarkable O+O spectroscopic binary with a short 2.66 day orbital period (Niemela & Gamen 2005). HD 150136 also has unusual radio properties, being a non-thermal radio emitter (Benaglia, Cappa, & Koribalski 2001) which made it an interesting $Chandra$ gratings target in its own right (Skinner et al. 2005). This deep spectrum produces a useful zero-order image. The zero-order data include the central 2′$\times$ 2′ region of the cluster. This observation, which should be fairly complete to 1 solar mass, reveals 43 X-ray sources within 4 square arcminutes (Skinner et al 2005). Only 11 of these had previous optical identifications, however all 43 were detected at H band and are likely cluster members. In addition to those sources cataloged by Skinner et al. there are about 30 X-ray point sources visible outside of the central 2′$\times$ 2′ region in this data set. Most of these have I-band magnitudes consistent with cluster membership and are between 2′ and 6.8′ from HD 150136. Wolk et al. (2008) obtained a 100 ks $Chandra$ observation of RCW 108–IR. Due to the orientation of the telescope pointing, about 64 square arcminutes of the NGC 6193 field was covered, mostly to the south and west of HD 150136. These data were a little less than 10 times deeper than the observation of Skinner et al. An additional 99 X-ray sources associated with NGC 6193 were detected. Most of these have infrared counterparts, which supports membership in Ara OB1a. This brings the total identified membership of NGC 6193 up to about 200\. But we emphasize the non-uniformity of the data. This is especially true of the Chandra pointings which cover perhaps 20% of the area of NGC 6193. A recent shallow $Spitzer/IRAC$ map of the region detected about 185 Class II sources, associated with both NGC 6193 and RCW 108–IR, and 13 Class I sources associated with RCW 108–IR alone (Wolk et al. 2008). But this observation was centered on RCW 108–IR and so again does not provide a uniform sample of NGC 6193. ## 3\. IRAS 16362–4845/RCW 108–IR In addition to the NGC 6193 cluster, the nearby source IRAS 16362–4845 harbors the extremely young stellar cluster RCW 108–IR, which has been recently studied (Comerón et al. 2005, Comerón & Schneider 2007, Wolk et al. 2008). The bright rim NGC 6188 marks the border between the HII region and a dense dark nebula containing RCW 108–IR. The bright rim is produced by the ionization of the molecular cloud hosting IRAS 16362–4845 by the two central stars of NGC 6193. The dark cloud contains a high emission measure knot (Wilson et al. 1970), which appears to be a compact HII region (Frogel & Persson 1974). The comprehensive UBVRI study by Herbst & Havlen (1977) proposed this was a site of very recent star formation. RCW 108–IR is a young, compact cluster partially embedded in its parent molecular cloud. Straw et al. (1987) used photometry at IR wavelengths (1.2-100 $\mu$m) to perform the first spatially complete survey of the cluster. They report on 55 objects; $<$ 20 of these have optical counterparts. In addition to the point sources, there is diffuse IR emission. The total luminosity of the cluster is about 1.8$\times 10^{5}$ L⊙, dominated by at least two early O stars. The full aggregate of O and B stars appears responsible for the ionization of the diffuse emission (Comerón et al. 2005). The primary exciting source star is identified as IRS 29 (Straw et al. 1987) – probably an O8 star. Straw et al. also identify at least one protostar. They put forward a physical model of the region in which photoionization of the older stars in the Ara OB1a association ionized the bright rim (NGC 6188) which lies at the edge of a finger of molecular material which thickens from east to west (Figure 2). Figure 2.: A simple model of the interface between RCW 108 and NGC 6193 from Straw et al. (1987). It now appears that the older stars lie somewhat behind the molecular cloud since none are seen superimposed over it. Several studies at radio wavelengths have been completed to understand the physical properties of the dark cloud. Goss & Shaver (1970) detected continuum emission at 5 GHz, Whiteoak & Gardner (1974) detected H2CO absorption at 4830 MHz, and CO emission at 115 GHz was detected by Whiteoak & Otrupcek (1982). The hydrogen recombination lines H109$\alpha$ and H166$\alpha$ were also measured (Wilson et al. 1970; Cersosimo 1982). RCW 108 IRS 29 is the second strongest water source seen by SWAS (second to the BN object in Orion; Gary Melnick-private communication) with a peak flux of over 7000 Jy in the 557 GHz line. A follow up study suggests that gaseous H2O is largely restricted to a thin layer of gas near the cloud surface. There have been two recent near-IR through mm studies of RCW 108–IR. Urquhart et al. (2004) compared the 2MASS and MSX observations of the cloud with radio recombination lines and radio continuum. They find that RCW 108–IR is an ultracompact HII region, less than 0.1 pc in scale with a core–halo morphology. They discuss 8 new sources within the UCHII region and report 3 sources detected at 25 $\mu$m and set back about 2.5′ from the bright rim. A more detailed study has been recently published by Comerón et al. (2005). From their high resolution CO map they estimate a total mass of the cloud at 8000 M⊙ with $<$ 200 M⊙ in the molecular concentration harboring the compact HII region. They produced a very high resolution JHKs map of the region (Figure 3). They identify 25 stars whose luminosities suggest spectral types earlier than A0 under the assumption that there is no significant circumstellar contribution to the K-band flux and conclude that the ionization of the UCHII region is provided by this aggregate. They suggest that low–mass star formation has yet to commence here. IRS 29 is found to be an O9 star from the visible spectrum of the compact HII nebula, which is in agreement with its infrared photometry. Out of 4365 stars brighter than Ks=14.5 in the whole 13′$\times$ 13′ field, 87 are found to have strong disk signatures, most of them located in the molecular cloud that contains RCW108–IR. Figure 3.: A JHKS (blue, green and red respectively) image of RCW 108–IR from the NTT. The field of view is about 13′ on a side, corresponding to about 6 pc on a side. Five-sigma detection limits in the least-exposed areas of the image are typically J=18.6, H=17.7, K=17.2 mag. The cloud is very opaque immediately north and south of the emission nebula. Most of the stars with optically thick disks are either within this emission or along the eastern edge of the cloud. Inset: A J (blue), H (green), and K (red) color composite of RCW 108-IR obtained using adaptive optics at the VLT. The frames are centered on the brightest source in the near-infrared, at RA(2000) = 16:40:00.2, Dec(2000) =-48:51:40. The field is 52′′ $\times$ 57′′. In their 100 ks $Chandra$ pointing centered on RCW 108–IR, Wolk et al. (2008) detect 32 point sources in the central arcminute, and 65 within the central 2′(Figure 4). They find a sharp rise in the absorption column of log NH $>$ 21.2 (A${}_{V}>10$) for these stars. By contrast X-ray sources associated with NGC 6193 have inferred extinctions between 0 and 5 AV. About 10 of the 65 sources in the core region do not have counterparts in the deep survey of Comerón et al. Figure 4.: $Chandra$/ACIS-I X-ray map of a somewhat wider region than shown in Figure 3. The colors indicate X-ray energy – red for soft X-rays ($<1$ keV), blue for hard X-rays ($>$ 2.4 keV) and green in between. Hard X-rays penetrate gas better than soft X-rays leading to X-ray ”blueing” equivalent to optical/infrared reddening. Thus the stars associated with RCW 108–IR appear blue and the stars to the east and south associated with NGC 6193 appear yellow or red. The region is about 20′ across. From Wolk et al. (2008). Recent JHKsL′ observations of the stellar content of RCW 108-IR have been carried out by Comerón & Schneider (2007) using adaptive optics-assisted imaging at the VLT. The superior resolution attained with adaptive optics provides a much more detailed and deep view than those available in previous studies. The deep imaging reveals faint members of the cluster, probably including massive brown dwarfs. This proves that, unlike suggested by Comerón et al.(2005), the cluster does contain low-mass stars. However, such members are present in numbers far below those expected when the bright end of the K-band luminosity function is extrapolated to fainter magnitudes (see e.g. Muench et al. 2002). This is interpreted as evidence for a top-heavy initial mass function, perhaps related to special circumstances that triggered star formation in the cluster (Sect. 6). The revised mass of the cluster based on these new observations is estimated to be $\sim 370$ M⊙. Comerón & Schneider (2007) also report the existence of a point-like Class I source detected only at 3.8 $\mu$m, as well as by Spitzer, projected on the border between the HII region and its surrounding molecular cloud. ## 4\. NGC 6204 and NGC 6167 NGC 6204 is an open cluster about 2 degrees northeast of NGC 6193. Whiteoak (1963) identified 78 possible members. He noted a wide variation in extinction, which prevented a good distance estimate. Hogg (1965) identified this clump as two distinct clusters. One cluster is still known as NGC 6204 and is estimated to be at 1180 $\pm 50$ pc. The stars in the other cluster, Hogg 22, are about twice as distant and hence not part of Ara OB1a or Ara OB1b. Forbes & Short (1996) performed photoelectric photometry on 36 member stars of NGC 6204 and find an age of about 125 Myr. This is much older than the other clusters in this region. Thus, these stars are probably not associated unless winds or shocks from extinct stars in this cluster triggered all that followed. The age difference between NGC 6204 and NGC 6193/RCW108–IR seems to safely exclude any role of the cluster in triggering star formation. Even if the velocity difference between the cluster and NGC 6193 were just a fraction of the typical velocity dispersion in a star forming complex, that would already imply that NGC 6204 formed far from its current location and that it was benign (with no stars with strong winds or able to explode as supernovae) by the time it got close to NGC 6193 or RCW 108–IR. Nonetheless, Waldhausen et al. (1999) argue that NGC 6204 is associated with NGC 6193 because their polarization angles are close (about 15o). NGC 6167 about 1 degree southwest of NGC 6193 was also poorly defined in the original work by Whiteoak (1963) due to highly variable absorption. Brück & Smyth (1967) identified this as a real cluster at 1200 $\pm 200$pc and 40 Myr old. Moffat & Vogt (1975) determined a distance of 600 pc, however their argument against the 1.2 kpc distance rests on a single, perhaps foreground, star. Rizzo & Bajaja (1994) assert the age to be closer to 10 Myr. Waldhausen et al. (1999) suggest that this is two overlain clusters since there appear to be two polarizations (offset by 15o). ## 5\. IRAS 16379–4856 and IRAS 16348–4849 Figure 5.: $Spitzer$/IRAC image of RCW 108–IR (IRAC bands 1, 2, 3 and 4 are blue, yellow, orange and red respectively.) Data are from a $Spitzer$ early release observation – Melnick-PI. The field of view is about 22′ on a side, corresponding to about 10 pc on a side. The red is dominated by PAH emission that fills much of the space and shows a deficit of stars in Figure 3. There are slight enhancements in the PAH emission on the southeastern and western portions of the field associated with IRAS 16379–4856 & IRAS 16348–4849. Most of the stars with optically thick disks are either within the core PAH emission or along the eastern edge. The two IRAS/CO peaks closest to RCW 108–IR are briefly discussed here. To the east is a small diffuse optical nebula near IRAS 16379–4856. Wolk et al. (2008) detect a complex of bright unresolved X–ray emission coincident with this region. A search of the 2MASS catalog for K-band excess sources in this region reveals the strongest concentration of such sources to be associated with this emission, at least 25 sources with K-band excesses in a 2′ $\times$ 1′ region. These sources all lie on the northwestern portion of a circular ring of 8 $\mu$m emission about 10′ in diameter. There are two 21 $\mu$m MSX point sources in the region. One centered among the 25 K-band excess sources, the other located near the inner edge of the ring coincident with the centroid of the IRAS source. This cloud is coincident with region H of the CO survey by Arnal et al. (2003). It is one of three CO clumps coincident with an IRAS source (RCW 108–IR and IRAS 16348–4849 - see below – are the other two). They estimate the mass of the cloud to be 5800 M⊙, second only to the core region of RCW 108–IR. MSX data indicate a mid-IR cloud about 10′ west of RCW 108–IR and coincident with IRAS 16348–4849 with a peak 8 $\mu$m flux of about 20% that of RCW 108. The overall extent of the cloud is about 10′ on a side. Tendrils of warm dust connecting the cloud and RCW 108 are evident in the MSX and $Spitzer$ images. Two sources from the recent 12CO observations by Arnal et al. (2005) lie within this dark cloud. Five B and A stars have been identified in this cloud; the most massive is the B2IV star HD 149658. 2MASS data show about 20 stars in the cloud with K${}_{S}<$ 15 and K-band excesses indicative of disks. Most of these stars lie within a circular region of 20 $\mu$m emission at about 1% the level of RCW 108–IR (Figure 5). These two regions, like RCW 108–IR, lie between NGC 6193 and the putative center of expansion. IRAS 16379–4856 is almost the same projected distance as RCW 108 from HD 150135/6 at the center of NGC 6193. So, the nominal cluster near IRAS 16379–4856 appears to be a sibling of RCW 108. Likewise formed by photoionization from the OB stars in NGC 6193 striking a cloud which appears to have been swept out by an event related to NGC 6167. The stars near IRAS 16357–4832 appear to compose a sibling cluster as well. The 8 $\mu$m morphology suggests that it was torn off of RCW 108 itself. This cluster is intermediate in mass between the former two. Additional bright CO and 8 $\mu$m emission is visible for a degree north of RCW 108. While the details as described here are speculative at this point, this is clearly a dynamic and active association. ## 6\. Triggered Star Formation It is the morphology of the region that makes Ara OB1a fascinating. NGC 6193 appears in relation to NGC 6188/IRAS 16362–4845 in much the same way that the $\sigma$ Orionis cluster (Walter et al. 1997, see also chapter by Walter et al.) relates to the Horsehead nebula. The emission in the open cluster is radiation bounded by the molecular cloud containing the embedded cluster RCW 108–IR, and is spectacularly revealed in the figures as a clear ridge between them (NGC 6188; Figure 1). This appears as one of the clearest examples of triggered star formation in the sky. But which mechanism and how it works is still an open issue. Herbst & Havlen (1977) originally put forward the hypothesis that Ara OB1a was an example of triggered star formation. Their favored model was that the trigger was a supernova event. The $prima-facie$ evidence for this was the circular shape of the dust ring upon which RCW 108–IR sits as a “diamond.” Straw et al. (1987) countered with the fact that RCW 108–IR is surrounded by some more evolved objects and point to the bright rim NGC 6188 as evidence that photoionization is very strong in the region. Thus, they favor the model originally detailed by Elmegreen & Lada (1977) in which successive sites of star formation are triggered by photoionization from the previous generation. Radio observations have much to contribute to this portion of the discussion since the radio data trace the gas and dust before the onset of protostellar collapse. Phillips et al. (1986) mapped the region in CO and find a CO cloud centered 2–3′ southeast of the bright stars of NGC 6193. They hypothesize that the cloud is behind NGC 6193 and is being compressed by the winds of the OB stars. However, there is no evidence for embedded sources within this cloud. The fragmented structure of the cloud supports its interpretation as the heavily eroded remnant of a now dispersed, larger cloud, possibly the progenitor of NGC 6193 (Comerón et al. 2005). Arnal et al. (1987) find a connection between the stellar population of Ara OB1a and an expanding neutral hydrogen ring centered near NGC 6167. The connection is that they derive the same distance to the gas as has been independently determined for the stars of $\approx 1400pc$. At that distance the HI ring is about 30 pc in radius and 20 pc in width expanding at 10 km s-1. Thus, Arnal et al. derive an upper age of this HI ring of 2$\times 10^{6}$ yr and an age closer to 1 Myr assuming a wind blown bubble (Weaver et al. 1977). While this is somewhat younger than current age estimates of NGC 6193, the age of NGC 6193 is based on very few stars. If the formation of NGC 6193 were the result of a supernova in NGC 6167 the problem of the age is more severe and in the opposite sense as NGC 6193 should be quite young, $\sim 0.1$ Myr, in this case (Straw et al. 1987). Figure 6.: A wide view of Ara OB1a. The base image is a DSS2 red plate about 2 degrees on a side. The top left quadrant is the region shown in Figure 1. The clusters NGC 6193 and NGC 6167 are indicated as well as RCW 108 and several IRAS point sources. The three IRAS sources nearest NGC 6193 all have 8 $\mu$m counterparts in MSX data. IRAS 60 $\mu$m contours of 300, 450, 600, 1200 and 2400 Jy/Sr are indicated. The large circle is used to connect the various features. There is a chain of CO clouds following along the eastern side of the circle. Rizzo & Bajaja (1994) follow up on the Arnal et al. finding with a more detailed HI study. They confirm the initial result. In their model, the 10 Myr stars of NGC 6167 formed the HI shell. This triggered the 3 Myr stars in NGC 6193. In their model the HI shell continues on to trigger RCW 108–IR. However, RCW 108–IR is misplaced in their Figure 6. It seems more likely that photoionization from the stars in NGC 6193 is triggering RCW 108 in the reverse direction relative to the outward propagating HI ring. Supporting evidence for external triggering in the clouds that host RCW 108–IR is given by Comerón et al. (2005) who found the cluster members they could identify are not symmetrically distributed throughout the cloud, but clumped along the ionization front as well as the core. At a smaller scale, Comerón & Schneider (2007) consider that their findings on the stellar population of RCW 108-IR provide evidence that its embedded cluster formed as a result of external triggering. In particular, a top-heavy initial mass function like that derived by those authors was suggested by Elmegreen & Lada (1977) as a possible byproduct of sequential star formation. This may be a consequence of the warm temperature of the gas in the shocked layer located ahead of the ionization front driven by the hot stars in NGC 6193, as well as of turbulence induced by instabilities in that layer and of coalescence of unstable fragments. Moreover, Comerón & Schneider (2007) also point out that the spatial distribution of massive stars in RCW 108-IR differs from that commonly found in other clusters containing high-mass members, which tend to concentrate towards the cluster center as a result of the formation conditions and early dynamical evolution (Bonnell et al. 2007). This is not expected for externally triggered star formation, and indeed is not observed in RCW 108-IR. ## 7\. The Future of Ara OB 1a Will more star formation take place in Ara OB1a? We believe this is the case and that it is already underway. A recent CO survey of the region finds over a dozen CO clouds in the region covering an area almost 3 degrees on a side (Arnal et al. 2003). While several groups of clouds are kinematically coherent, there is a spread in velocities of almost 20 km s-1. Three of the concentrations are adjacent to the bright rim separating RCW 108 and NGC 6193. IRAS data (see Figure 6) show three concentrations of cool dust in the vicinity of RCW 108–IR spaced about 5 pc apart. These concentrations are bright from 8 $\mu$m to 100 $\mu$m. The brightest of these is RCW 108–IR itself. The IRAS peaks are coincident with 13CO peaks found by Yamaguchi et al. (1999). The 13CO peaks then make a chain to the south. Arnal et al. (2003) identify five separate 12CO features in this cloud. The chain extends almost 1.5o south of RCW 108–IR. In fact, the CO peaks taken in concert with the 5 IRAS peaks (including IRAS 16379–4856 and IRAS 16348–4849 adjacent to RCW 108–IR) cover almost exactly 180 degrees of arc in a near perfectly circular alignment 1.03o from $l$=335:29:12 $b$=$-$1:42:23, about 8 pc from the center of NGC 6167. More specifically, Comerón et al. (2005) estimate a current star formation efficiency in the main cloud hosting IRAS 16362–4845/RCW 108–IR to be well below 10%, suggesting that most of star formation in the cloud still has to take place. They also speculate that its future evolution may lead to a complex similar to the Orion Nebula Cluster, with the aggregate of IRAS 16362–4845 being the equivalent to the Trapezium cluster. Acknowledgments. We acknowledge many useful comments from the anonymous referee, one of which was worked directly into the text. We gratefully acknowledge the financial support of NASA grant GO4-5013X (Chandra) and from NASA contract NAS8-39073 (CXC). Archival data was obtained from the Two Micron All Sky Survey (2MASS), a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation and additional data were obtained with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France, and of NASA’s Astrophysics Data System Bibliographic Services. ## References * Arnal et al. (1987) Arnal, E. M., Cersosimo, J. C., May, J., & Bronfman, L. 1987, A&A, 174, 78 * Arnal et al. (2003) Arnal, E. M., May, J., & Romero, G. A. 2003, A&A, 412, 431 * Arnal et al. (2005) Arnal, E. M., Romero, G. A., May, J., & Minniti, D. 2005, ASP Conf. Ser. 344: The Cool Universe: Observing Cosmic Dawn, ed. C. Lidman & D. Alloin, 173 * Bonnell et al. (2007) Bonnell, I.A., Larson, R.B., & Zinnecker, H., 2007, in Protostars and Planets V, eds. B. Reipurth, D. Jewitt, K. Keil, Univ. of Arizona Press, 149 * Benaglia et al. (2001) Benaglia, P., Cappa, C. E., & Koribalski, B. S. 2001, A&A, 372, 952 * Bruck & Smyth (1967) Brück, M. T. & Smyth, M. J. 1967, MNRAS, 136, 431 * Cersosimo (1982) Cersosimo, J. C. 1982, ApJ, 22, L157 * Comerón et al. (2005) Comerón, F., Schneider, N., & Russeil, D. 2005, A&A, 433, 955 * Comerón et al. (2007) Comerón, F. & Schneider, N., 2007, A&A, 2007, 473, 149 * Elmegreen & Lada (1977) Elmegreen, B. G. & Lada, C. J. 1977, ApJ, 214, 725 * Fitzgerald (1987) Fitzgerald, M. P. 1987, MNRAS, 229, 227 * Forbes & Short (1996) Forbes, D. & Short, S. 1996, AJ, 111, 1609 * Frogel & Persson (1974) Frogel, J. A. & Persson, S. E. 1974, ApJ, 192, 351 * Herbst & Havlen (1977) Herbst, W. & Havlen, R. J. 1977, A&AS, 30, 279 * Herbst (1975) Herbst, W. 1975, AJ, 80, 498 * Hogg (1965) Hogg, A. R. 1965, PASP, 77, 440 * Kaltcheva & Georgiev (1992) Kaltcheva, N. T. & Georgiev, L. N. 1992, MNRAS, 259, 166 * Moffat & Vogt (1973) Moffat, A. F. J. & Vogt, N. 1973, A&AS, 10, 135 * Moffat & Vogt (1975) Moffat, A. F. J. & Vogt, N. 1975, A&AS, 20, 155 * Muench et al. (2002) Muench, A.A., Lada, E.A., Lada, C.J., Alves, J., 2002, ApJ, 573, 366 * Niemela & Gamen (2005) Niemela, V. S. & Gamen, R. C. 2005, MNRAS, 356, 974 * Phillips et al. (1986) Phillips, J. P., de Vries, C. P., & de Graauw, T. 1986, A&AS, 65, 465 * Rizzo & Bajaja (1994) Rizzo, J. R. & Bajaja, E. 1994, A&A, 289, 922 * Rodgers et al. (1960) Rodgers, A. W., Campbell, C. T., & Whiteoak, J. B. 1960, MNRAS, 121, 103 * Shaver & Goss (1970) Shaver, P. A. & Goss, W. M. 1970, Australian Journal of Physics Astrophysical Supplement, 14, 133 * Skinner et al. (2005) Skinner, S. L., Zhekov, S. A., Palla, F., & Barbosa, C. L. D. R. 2005, MNRAS, 361, 191 * Straw et al. (1987) Straw, S., Hyland, A. R., Jones, T. J., Harvey, P. M., Wilking, B. A., & Joy, M. 1987, ApJ, 314, 283 * Urquhart et al. (2004) Urquhart, J. S., Thompson, M. A., Morgan, L. K., & White, G. J. 2004, A&A, 428, 723 * Vazquez & Feinstein (1992) Vazquez, R. A. & Feinstein, A. 1992, A&AS, 92, 863 * Waldhausen et al. (1999) Waldhausen, S., Martínez, R. E., & Feinstein, C. 1999, AJ, 117, 2882 * Walter et al. (1997) Walter, F. M., Wolk,S. J., Freyberg, M., & Schmitt, J. H. M. M. 1997, Memorie della Societa Astronomica Italiana, 68, 1081 * Weaver et al. (1977) Weaver, R., McCray, R., Castor, J., Shapiro, P., & Moore, R. 1977, ApJ, 218, 377 * Whiteoak et al. (1982) Whiteoak, J. B., Otrupcek, R. E., & Rennie, C. J. 1982, Proceedings of the Astronomical Society of Australia, 4, 434 * Whiteoak & Gardner (1974) Whiteoak, J. B. & Gardner, F. F. 1974, A&A, 37, 389 * Whiteoak (1963) Whiteoak, J. B. 1963, MNRAS, 125, 105 * Wilson et al. (1970) Wilson, T. L., Mezger, P. G., Gardner, F. F., & Milne, D. K. 1970, A&A, 6, 364 * n (0) Wolk, S. J., Spitzbart, B. D., Bradley, D., Bourke, T. L., Gutermuth, R. A., Vigil, M., & Comerón, F. 2008, AJ, 135, 693 * Yamaguchi et al. (1999) Yamaguchi, R., Saito, H., Mizuno, N., Mine, Y., Mizuno, A., Ogawa, H., & Fukui, Y. 1999, PASJ, 51, 791	arxiv-papers	2008-08-25T16:49:18	2024-09-04T02:48:57.437386	{ "license": "Public Domain", "authors": "Scott J. Wolk, Fernando Comeron and Tyler Bourke", "submitter": "Scott J. Wolk", "url": "https://arxiv.org/abs/0808.3383" }
0808.3385	# The Embedded Massive Star Forming Region RCW 38 Scott J. Wolk Tyler L. Bourke Miquela Vigil ###### Abstract RCW 38 is a uniquely young ($<$1 Myr), embedded ($A_{V}\sim 10$) stellar cluster surrounding a pair of early O stars ($\sim$O5.5) and is one of the few regions within 2 kpc other than Orion to contain over 1000 members. X-ray and deep near-infrared observations reveal a dense cluster with over 200 X-ray sources and 400 infrared sources embedded in a diffuse hot plasma within a 1 pc diameter. The central O star has evacuated its immediate surroundings of dust, creating a wind bubble $\sim$0.1 pc in radius that is confined by the surrounding molecular cloud, as traced by millimeter continuum and molecular line emission. The interface between the bubble and cloud is a region of warm dust and ionized gas, which shows evidence for ongoing star formation. Extended warm dust is found throughout a 2–3 pc region and coincides with extended X-ray plasma. This is evidence that the influence of the massive stars reaches beyond the confines of the O star bubble. RCW 38 appears similar in structure to RCW 49 and M 20 but is at an earlier evolutionary phase. RCW 38 appears to be a blister compact HII region lying just inside the edge of a giant molecular cloud. Harvard–Smithsonian Center for Astrophysics, Cambridge MA 02138, USA Harvard–Smithsonian Center for Astrophysics, Cambridge MA 02138, USA Lincoln Laboratory, Massachusetts Institute of Technology, Lexington MA 02420, USA Figure 1.: An optical image of RCW 38, about 20′ ($\sim$10 pc) on a side, based on images from the digitized sky survey (DSS). Blue plates are printed as blue, red plates are printed as yellow, near-infrared data are printed in red. The region in the green box is shown at mid-infrared wavelengths in Figure 2. Figure 2.: Mid-infrared image of RCW 38, about 11.5′ (5.7 pc) on a side, from $Spitzer$/IRAC observations. IRAC Band 1 (3.6 $\mu$m) is assigned to blue, Band 2 (4.5 $\mu$m) to green, and Band 3 (5.6 $\mu$m) to red. [3.6] band = blue, [4.5] band = green and [5.6] band = red. Voids in the 5.6 $\mu$m image tend to be filled by 4.5 $\mu$m emission. The yellow box is 2.5′ on a side and is shown in more detail in Figures 3 and 5. ## 1\. Introduction The evolution of high mass clustered star forming regions is complex and poorly understood. Only the nearby ($\sim$400 pc), optically revealed, Orion Nebula Cluster (ONC) is well studied (see Muench et al. and O’Dell et al. in this Handbook). Yet, a wide variety of high mass embedded clusters is found within 2 kpc of the Sun. Within this limit, the young cluster RCW 38 (08h59m47.2s -47∘31′57′′ (J2000), l,b = $268.03^{\circ},-0.98^{\circ}$) is one of the few regions other than the ONC to contain over 1000 members (Lada & Lada 2003; Wolk et al. 2006). RCW 38 has an embedded and dense stellar population, comparable to other regions that have been studied recently with Spitzer (e.g., M 20 and RCW 49; Rho et al. 2004, Whitney et al. 2004, Churchwell et al. 2004). RCW 38 provides a unique opportunity to study the evolution of a rich cluster during the phase where its most massive members, a pair of O5.5 stars (DeRose et al. 2008) have just completed their ultracompact HII region (UCHII) phase and are now greatly influencing its natal environment and the evolution of its low mass members. ## 2\. Overview RCW 38 was catalogued in both the H$\alpha$ survey by Rodgers, Campbell & Whiteoak (1960) and the earlier survey of southern HII regions (Gum 1955) as a moderately bright region of emission about 40′ on a side (Figure 1). Radio surveys during the 1960’s indicated that it was one of the brightest HII regions at radio wavelengths (e.g., Wilson et al. 1970). The radio brightness made it an early candidate supernova remnant until its spectrum was shown to be thermal. A complicating factor in the study of RCW 38 is that it is adjacent to the Vela Molecular Ridge (see the chapter by Pettersson). In fact, there is a faint ring shape structure – perhaps a supernova remnant (SNR) – in apparent contact with the star forming cloud (RX J0852.0-4622). However, this seems to be a chance superposition as the SNR is about a factor of two closer than RCW 38 based on absorption arguments (Aschenbach et al. 1999). The 5 GHz survey by Shaver & Goss (1969) shows RCW 38 as both a bright continuum source at multiple wavelengths (Shaver & Goss 1970) and broad in its extent of greater than 30′ in diameter. It was not until the early 1970’s that this region was clearly associated with massive young stars (Johnson 1973). The next year, with improving resolution and detectors, H90$\alpha$ observations indicated that the radio peak of the region had a ring–like shape about 2′ across (Huchtmeier 1974). Figure 1 shows an optical view of the RCW 38 region. Note the central region is opaque due to high extinction. This has confined the study of the central region mostly to infrared and radio wavelengths. Several infrared studies indicated a “ring–like” or “horseshoe” structure around IRS 2 (Figures 2,3), which is also seen in high resolution radio and millimeter wavelength studies (Vigil 2004; Wolk et al. 2006). These structures are shown in detail in Figure 3. There are two defining infrared sources in RCW 38 (Frogel & Persson 1974) – the brightest at 2 $\mu$m is labeled IRS 2 and the brightest at 10 $\mu$m is referred to as IRS 1. ## 3\. IRS 1 Frogel & Persson (1974) studied RCW 38 at 10 $\mu$m with moderate spatial resolution (14.5′′) and found that the warm dust emission follows a horse-shoe shape across a $1.8\times 1.8$ arcmin2 area; they labeled the brightest unresolved peak in its western side IRS 1. Epchtein & Turon (1979) mapped the central region of RCW 38, at 10 $\mu$m, with about twice the resolution of Frogel & Persson and resolved IRS 1 into several discrete sources; they suggested that this may be a cluster of embedded stars, younger than those associated with IRS 2. In MSX observations at 8.3 $\mu$m the central region of RCW 38 exhibits a circular emission structure, perhaps slightly elongated about 2′ across, in agreement with the earlier mid-infrared studies. This is resolved by $Spitzer$/IRAC observations at 3.6-8.0 $\mu$m (Wolk PI) to show a roughly heart shaped region of polycyclic aromatic hydrocarbon (PAH) emission which includes numerous bubbles filled with 4.5 $\mu$m diffuse emission (see Figure 2). The central region appears to lie within a large bubble about 0.4 pc across. The O5.5 binary IRS 2 appears to lie in the center of the bubble as seen in the mid- and near-infrared data (Sect. 4 and Figures 2 and 3). The inner face of the eastern part of the central bubble appears to be illuminated, giving the impression that we are not looking straight down the opening but rather offset by between 30 and 45 degrees. Smith et al. (1999) made high resolution mid-infrared maps of the core region of RCW 38. They concentrated on the IRS 1 region that has the peak radio and 10 $\mu$m emission. They found that IRS 1 is a dust ridge extending 0.1-0.2 parsec predominantly in the north–south direction less than 0.1 pc west of IRS 2. The dust ridge has a color temperature of about 175 K and has several condensations within it. They identify IRS 1 as the brightest peak among them, lying interior to all the other peaks about 12′′ west of IRS 2. Figure 3.: A close-up of the central 2.5′ ($\sim$1.2 pc) of RCW 38 (Wolk et al. 2006). In this VLT image, Z band data are printed as blue, H band data are green and K band are red. IRS 2 is the brightest point source in the field. IRS 1 is associated with the bright ridge of emission to the west of IRS 2, near the center of the image. The diffuse radiation is a mixture of starlight scattered by the dust and gas in the area, and atomic and molecular hydrogen line emission. Essentially every star visible in this image is a cluster member (Wolk et al. 2006; DeRose et al. 2008). ## 4\. IRS 2 From its near-infrared (NIR) colors and assuming a distance of 1.5 kpc, Frogel & Persson (1974) suggested that the brightest 2 $\mu$m source, IRS 2, is an early O-star (specifically O4) having about 12.8 magnitudes of visual extinction, or a group of later spectral type O stars. This is sufficient to account for the observed continuum emission at 5 GHz (Shaver & Goss 1970). Furniss, Jennings, & Moorwood (1975) combined broad-band 40-350 $\mu$m measurements with the previous work of Frogel & Persson, deriving a total luminosity of 7$\times 10^{5}$ L⊙ and arguing that an O4 star would supply more than enough luminosity to account for all the infrared luminosity and emit enough Ly photons to yield twice the observed radio flux, unless significant absorption by dust is occurring. They suggest however that IRS 2 is an O5 star with little or no dust absorption of the continuum photons in the ionized region. The higher resolution imaging of Ligori et al. (1994) confirmed the presence of a cluster of at least 5 optically invisible young stellar objects embedded in nebular emission in the central 1.5′ near IRS 2. Frogel & Persson (1974) produced detailed 1.25 – 20 $\mu$m mapping of the region. At 10 $\mu$m they confirm a ring-like structure with the bright 2 $\mu$m source – IRS 2 – very near the center and a dominating 10 $\mu$m source (IRS 1) about 10′′ to the west. A third 10 $\mu$m source about 40′′ to the east of IRS 2 was also noted. Persson et al. (1976) revisited this region using narrow band photometry and found no evidence of silicates. Other narrow line infrared imaging such as Br-$\gamma$ at 2.16 $\mu$m also indicated an open center morphology centered on IRS 2 with enhanced emission toward IRS 1 (Mizutani et al. 1987). Mid-infrared imaging and spectroscopy by Smith et al. (1999) provide further evidence that the ionizing source of the region is at least O5 or earlier. Through modeling of the ratio of emission lines at IRS1, the ionizing radiation field constrains the ionizing source to be between O4.5 and O5.5. Smith et al. find that the region surrounding IRS2 is depleted of dust, suggesting the region can be explained in terms of a wind-blown cavity, where winds from IRS2 have blown out the surrounding material. DeRose et al. (2008) used adaptive optics imaging to directly show that IRS2 is a binary with equal mass members and a projected separation of $\sim$ 500 AU. The presence of two O stars of equal mass combined with the earlier data on luminosity and ionization imply that they have spectral types of O5.5. ## 5\. Massive Stars Muzzio (1979) and Muzzio & Celotti de Frecha (1979) identified OB stars throughout the southern sky using photometry and objective prism photography. They identified about 20 OB star candidates in the vicinity of RCW 38. All have absolute V magnitudes less than zero but are concentrated about 10–20′ southwest of IRS 2. Wolk et al. (2006) use the absolute K-band magnitudes of X–ray sources to determine the existence of 31 candidate OB stars over a 5 pc2 region centered on IRS 2. More than 20 of these are found to be in the central $\sim$1 pc. No unusual clustering to the southeast was noted. To make mass estimates they estimated the age of the cluster at 0.5 Myr and noted that if they assumed an age of 1 Myr, the number of OB stars would have more than doubled to a somewhat less reasonable value. However, the Wolk et al. sample was X-ray based and hence biased against finding soft OB stars behind more than 20 AV of extinction111Recent X-ray observations indicate that O stars are capable of producing very hard X-rays (e.g. Gagné et al. 2005). However, most B stars and about half of the observed O stars are still soft sources and hence observations in dusty regions are biased against detecting them.. The OB candidates remain to be confirmed. ## 6\. Extinction NIR imaging and photometry of the region by Storey & Bailey (1982) showed a number of highly reddened point sources and they deduced an AV of order 60 to the cluster. The 2MASS data show that the nebulosity associated with RCW 38 is extensive across a large area, with dust lanes and patches running throughout. Wolk et al. (2006) measure the extinction and hydrogen column to over a score of sources with good X-ray and NIR data and find AV ranging from 3 to 20 with strong evidence of NIR sources embedded in more than 30 magnitudes of visual extinction. Smith et al. (1999) find that the region around IRS 2 has a gas to dust ratio much lower than 100 to 1\. Thus, this area is a true cavity and not just an extinction effect. Millimeter radio observations indicate that several hundred magnitudes of extinguishing material may lie behind the observed stars (see Section 11). Combining these results with the near–IR extinction observations suggests that RCW 38 is an embedded blister HII region lying just inside the front edge of a giant molecular cloud. Further, it appears that the HII region is compacted by the overlying material and beginning to break out in some locations. ## 7\. The Embedded Stellar Cluster Even in the remarkable $Spitzer$/IRAC image (Figure 2) the scope and extent of the cluster is hard to ascertain. Only a few dozen sources are identifiable within the bright nebular emission. Outside the IR nebula cluster members are difficult to distinguish from background stars. We have found that 116 sources in the 2MASS catalog, corresponding to the region shown in Figure 3, have IR- excesses consistent with an optically thick disk at K–band. A number of the cluster members are directly exposed to IRS 2 in the cavity, and could be undergoing the same fate as the proplyds in the ONC (DeRose et al. 2008). Further, about 250 sources in the regions are Class I or Class II objects based on their IRAC colors. This group includes the 2MASS selected candidates. Most of these are clearly associated with RCW 38. However, at least 2 are associated with a background star forming region – BRAN 231A. Coincidences among the X-ray detected sources and NIR excess sources are found over 10′ from IRS 2 indicating that the cluster extends at least 5 pc. The first hint of the extent of the low mass star population is provided by the spectacular NIR image of the heart of the region (Figure 3). Despite the small (2.5′ on a side) field of view, one can immediately ascertain several important features of the system from this image. First, the blue to red gradient indicates a steep dust extinction gradient increasing from northwest to southeast. Second, two regions are cleared of dust, one about 0.1 pc in diameter centered on IRS 2, and another of similar size, just west of IRS 1. Both of these cleared regions are visible in the optical plates indicating that extinction is not particularly high for these regions. The bright ridge separating the two cleared regions is the IRS 1 ridge which was mapped in detail by Smith et al. (1999). Third, there is a plethora of stars in this image. Over 480 stars are visible in this image and the vast majority of these are likely to be cluster members (Wolk et al. 2006, DeRose et al. 2008). The data are complete to the brown dwarf limit for extinctions less than an AV of 20 (Wolk et al. 2006). About 130 of the NIR sources have X–ray counterparts – there are fewer than a dozen X-ray sources without NIR counterparts in the central 2.5′. This leaves about 350 NIR detected cluster members not detected in X-rays – an incompleteness of over 70%. The X-ray counterparts demonstrate both membership in the cluster and masses of more than about 0.5 M⊙. Overall about 345 X-ray sources detected by $Chandra$ in a 17′x 17′ field centered here are associated with RCW 38 (Wolk et al. 2006). Completeness arguments based on the cluster distance, obscuration, and X-ray sensitivity lead to a total estimated cluster size of between 1500 and 2400 stars. This makes RCW 38 the largest embedded cluster by membership within 2 kpc of the Sun, after the ONC. ## 8\. Diffuse X-ray Emission As shown in Figure 4, RCW 38 hosts very bright diffuse X-ray emission (Wolk et al. 2002). This X-ray emission differs from that of most other massive star forming regions because of its power-law spectrum indicative of synchrotron, not thermal, emission. Since the spectral signature is not that of stellar coronae this emission cannot be the sum of unresolved stars. Further, it implies the existence of a strong magnetic field. Synchrotron emission requires an electron population to be driven along a magnetic field. Indeed a significant magnetic field has been measured in the gas associated with RCW 38 (38$\pm$3 $\mu$G; Bourke et al. 2001). Astrophysically, this is most common along the expanding shock front of a supernova remnant or during violent accretion onto a compact object with a large magnetic field. We noted earlier the existence of a possible supernova remnant in the field, it is offset from RCW 38 by about 6′ and there is known X-ray emission associated with this location. Further, the SNR is foreground to RCW 38 and the diffuse emission becomes more absorbed towards the southeast. This is the same pattern seen among the stars and nebulosity associated with RCW 38 – indicating that the diffuse X-ray emission is co–located with the stars and dust of RCW 38. RCW 38 is the first and currently only known source of diffuse synchrotron X-ray emission from a very young star forming region (Feigelson et al. 2007). The morphology of the diffuse X-ray emission is striking. It is strongest in the central region near IRS 2 where radio and IR-nebular emission appears cleared. Further it is confined on the southeast along a ridge that also confines the mm continuum (cool dust) emission. There is also a distinctive kink where the diffuse X-ray emission is deflected around the dust ridge near IRS 1. It also appears that the plasma is breaking out of some form of confinement to the northwest where it traces around the diffuse NIR emission – akin to the way a stream moves around a stone. This break out to the northwest is visible as reflection nebulosity in Figure 1 and corresponding bubble-like structures in Figure 2. Figure 4.: A combined X-ray and IR view of the inner 11.5′ (5.7 pc.) of RCW 38. The X-ray data have been adaptively smoothed to about 10′′ resolution. Soft X-rays (0.5 – 1.5 keV) are shown in green and harder X-rays (1.5 – 8 keV) in blue. The various structures in the diffuse X-ray emission seem to trace gaps in the polycyclic aromatic hydrocarbon (PAH) structures apparent in the 5.8 $\mu$m $Spitzer/IRAC$ data (red). The lack of X-ray emission to the southeast is likely due to high extinction. ## 9\. The Distance to RCW 38 The first reasonable estimate for the distance was from Radhakrishnan et al. (1972). They estimate a kinematic distance of 1.5 kpc. Muzzio (1979) identified about 10 OB star candidates in the vicinity of RCW 38 and derived a distance of about 1.7 kpc by placing the stars on the main sequence of an HR diagram. Murphy (1985) examined the CO structures in the Vela ridge in addition to the photometric data and derived a distance of 1.6 $\pm 0.8$ kpc. Avedisova & Palous (1989) performed photometric measurements of 218 star forming regions and also achieve a result of 1.7 kpc. The X-ray luminosity function is consistent with the 1.7 kpc distance (Wolk et al. 2006). At this distance, 1 arcmin corresponds to 0.495 pc. ## 10\. Radio Observations Early low resolution radio observations of continuum emission and recombination lines indicate typical HII region conditions, i.e., an electron temperature of $\sim$8000 K and an emission measure of $\sim 10^{6}$ pc cm-6 (e.g., McGee & Newton 1981, Caswell & Haynes 1987). A relatively high electron density of $\sim 10^{4}$ cm-3 supports the view that the HII region is confined by the surrounding molecular cloud. Resolved radio studies reveal a clumpy ring-like structure around IRS 2 showing the general morphology seen in the infrared (Huchtmeier 1974; Vigil 2004). As in the mid-infrared, the brightest radio peak coincides with IRS 1 (Figure 5). High resolution observations indicate the ring is mostly optically thick at centimeter wavelengths, with no evidence for synchrotron emission (Vigil 2004). Line widths observed in recombination lines are typically 30 km s-1 or greater. Sensitive high resolution 18 cm (1.6 GHz) observations show weak emission extending to the NW-SE outside of the ring, suggesting that, like the X-ray emission, some diffuse radio emission has escaped the compact HII region. Figure 2 shows a bubble-like feature associated with this emission in the NW. The radio ring is not well centered on IRS 2, likely indicating that the surrounding gas is not homogeneous. ## 11\. Molecular Gas and Dust Continuum Detailed molecular line observations of RCW 38 are rare. Gillespie et al. (1979) mapped a large region (10 $\times$ 5 pc) in CO 1-0 with a 3.2′ beam. They found two clouds separated by a bridge of emission. Yamaguchi et al. (1999) also mapped CO 1-0 with similar resolution over a larger region, confirming the earlier result, and estimated a mass of $1.5\times 10^{4}$ M⊙ for the two clouds as a whole. The RCW 38 cluster is directly associated with the eastern cloud. Zinchenko et al. (1995) mapped the eastern cloud over a 5 arcmin2 region in CS 2-1 with 1′ resolution. Their maps show that CS, like the radio, follows a ring-like structure, but close comparison indicates that the CS is located exterior to the radio, tracing the dense molecular gas in which the compact HII region is embedded. The mass traced by CS is $\sim 9000$ M⊙ (correcting for their incorrectly assumed distance of only 0.7 kpc) with a mean density of $10^{4}$ cm-3. A number of other high density tracers have been observed with 1′ or better resolution but maps are lacking. Deep absorption is observed in the main lines of the lowest rotational levels of OH at 18 cm, and these have been used to measure the magnetic field strength of 38$\pm$3 $\mu$G (Bourke et al. 2001). High resolution ($\sim 10\hbox{${}^{\prime\prime}$}$) observations of OH show that the absorption is seen against the radio ring, and is directly associated with the gas surrounding the HII region, sharing the same velocity as the recombination lines (T.Bourke, private communication). No maser emission is observed in OH. Water maser emission at 22 GHz is observed from a position near to IRS 1, and it is time variable (Kaufmann et al. 1977; Batchelor et al. 1980; Caswell et al. 1989). Water maser emission is a sure sign of active star formation, and high resolution observations are needed to precisely located this emission. The dust continuum has been mapped at wavelengths near to 1 mm (Cheung et al. 1980; Vigil 2004). Low resolution observations revealed only one clump centered near IRS 1 (Cheung et al. 1980), with a peak density of $6\times 10^{5}$ cm-3, a total mass of $3\times 10^{4}$ M⊙, and a visual extinction of 800 mag. Higher resolution observations reveal the ring-structure seen at other wavelengths (Vigil 2004), with a central hole around IRS 2, but mostly peaking exterior to the radio ring. In particular, the position of peak intensity is west of the IRS 1 ridge and associated with the extincted region seen in Figure 3. The mass calculated by Vigil (2004) is an order of magnitude less than that of Cheung et al. (1980), from similar fluxes. As the morphology of the dust emission observed by Vigil suggests that it is associated with the radio ring, the extinction determined by Cheung et al. (1980) may be an overestimate. ## 12\. Interpretation The ring-like shape of the two dimensional radio and infrared continuum image suggests the possibility of a shell–like structure surrounding IRS2. The bulk of the emission seen in the cm images is concentrated in a ring which would be the thick walls of a three dimensional shell. The center cavity would contain only the emission from the cover of the shell perpendicular to the direction of observation (Figure 5). Figure 5.: 4800 MHz (6 cm) radio continuum contours from the ATCA overlain on the VLT K-band image shown in Figure 3 (30′′ = 0.25 pc). The white triangle indicates the location of the peak at 10 $\mu$m, IRS 1, and the white star indicates the position of the O-stars IRS 2. The contour levels are 15, 30, 45, 60, 75 and 90 percent of the peak emission of 3.5 Jy/beam. The flux in the center region is low, between 15–30% of the peak flux level that is located near the IR ridge and IRS 1. The synthesised beam of 10′′ is shown at lower left. Adapted from Vigil 2004. We see two possible interpretations of the relationship among the observational data – one is wind driven (Smith et al. 1999), the other supernova driven (Wolk et al. 2002). The supernova scenario has, as its key piece of evidence, a symmetrical ring of emission in the radio continuum. A second line of evidence for a supernova in this region is the hot X-ray plasma with a power–law spectrum, which appears to be getting harder as it moves away from the cluster center. This is as one would expect with synchrotron emission being accelerated from a compact object near the cluster center. The total X-ray luminosity is about $3\times 10^{32}$ ergs sec-1. In these attributes it evokes the image of a shell supernova remnant such as SN 1006. Further, if one assumes equipartition, we derive a very reasonable magnetic field of about 4.9 $\mu$G and a reservoir of $2.2\times 10^{44}$ ergs of accelerated electrons, giving a lifetime of about 20,000 years. But, the radio continuum of RCW 38 reported here does not show the brightening at the edges seen in SN 1006 (Reynolds & Gilmore 1986). The positive spectral index of the radio continuum also favors thermal emission. Further the X-ray emission seen in shell supernovae tends to follow the form of the radio emission (c.f. Allen et al. 2001). Here, the plasma seems confined by the continuum radio emission with the exception of the breakout to the northwest. This confinement naturally lends itself to a wind driven scenario. Smith et al. (1999) note that the central region around IRS 2 is dominated by Ly$\alpha$ emission and is dust free. This suggests that RCW 38 is a compact HII region expanding into and confined by a non–homogeneous molecular cloud. The wind from the massive O-star (IRS 2) can give rise to collisional shocks near to the star as described by Cantó et al. (2000). Such shocks centered near IRS 2 would fill the cavity with a limb darkened X-ray emission as seen. But the derived thermal spectrum for the plasma is much warmer than expected by these models ($\sim$10 keV ($\sim$ 100 MK) vs. $\sim$ 1 keV ($\sim$ 10 MK)). This high temperature may have to do with confinement due to the surrounding molecular cloud. Heated plasma will tend to move outward and cool, but in the case of RCW 38 the molecular cocoon prevents meaningful cooling and causes a “greenhouse effect” on a parsec scale. Smith et al. (1999) derived a similar result using Ly$\alpha$ data. Analysis of the smoothed X-ray data supports the hypothesis that the molecular material is trapping the plasma that appears to be partially heated by the trapped Ly$\alpha$ photons. There appears to be a break out of the plasma from the ring of the molecular gas in the northwest corner while elsewhere the gas is confined much closer to the O star. Further evidence for this interpretation is found in other regions of massive star formation. Several massive star forming regions have been seen to have much cooler diffuse X-ray emission. Townsley et al. (2003) report thermal spectra of below 1 keV for the Rosette Nebula and M17. Other massive star forming regions, including the Arches cluster (Yusef-Zadeh et al. 2002) and NGC 3603 (Moffat et al. 2002) have thermal spectra above 3 keV. While a commonality among these clusters is a massive O star (O6 or earlier) the key difference between the former group and the latter group seems to be the presence of absorbing (confining) material. All the clusters with high temperature plasma have $N_{H}>5\times 10^{21}$ cm-2 while the clusters with cooler plasma have less absorption. The wind driven interpretation involves winds blowing from IRS 2 that are excavating the material in its near vicinity creating this expanding shell, seen as a radio continuum ring of emission. IRS 2 is near the center of the cavity and is likely the driving source. However, IRS 2 is displaced from the geometric center of the ring that is offset by about 18′′ to the northeast. As noted earlier, this appears due to our viewing angle at the system not being face on and the non-homogeneous surroundings. From inspection of the K–band image (Figure 5), an arc-like feature, reminiscent of a bow shock, coincident with the bright radio ridge, can be seen in the southwest of the ring. It suggests either a massive star forming region $\sim 10$′′ to the west, which is creating a bow shock by deflecting the winds from IRS 2 around its edge, a wind source from the southwest which could be counteracting the wind from IRS 2 causing a compression ridge to build up along the edge of the ridge – IRS 1, or a dense globule being ablated (Bertoldi & McKee 1997). Indeed, there are about a half dozen X-ray sources within this arc, and they all are found to have column densities in the range of 1.9 - 2.9$\times 10^{22}$ cm-2, equivalent to about 10-15 AV. The southwest source, be it a massive star driving a wind or simply a dense core, is preventing the expansion in the southwest direction, causing the shell to expand non-uniformly and mostly toward the northeast, thus explaining the offset of the center to the northeast. From close inspection of the VLT K-band images, an arc-like feature outlines the bright radio ridge (Figure 5). This specific feature implies dense gas that is likely photo-evaporating and could suggest that this is a site of triggered star formation. Additional evidence of active star formation occurring just west of IRS 1 is a possible protostar found embedded in the hydrogen gas and molecular jets eminating from this region (DeRose et al. 2008), as well as the previously mentioned water masers. The winds from IRS 2 appear to be confined by the surrounding molecular cloud. Assuming a typical O-star wind of 1000 km/s (Cantó et al. 2000) and an age for IRS 2 of about 500,000 years suggests that a freely expanding wind would have traveled about 500 pc since the star was formed. The inner width of the shell is close to a tenth of a parsec, much smaller than expected from a freely expanding wind. Also, in the smooth X-ray data, plasma is observed as far as 3.5 pc from IRS 2, far short of the 500 pc predicted for a freely expanding wind. ## 13\. Summary RCW 38 appears to be a blister compact HII region with winds originating from the central O-star IRS 2. These winds are excavating the mass in the immediate vicinity of IRS 2 creating a shell–like structure, detected as a radio continuum ring and also evident at infrared and millimeter wavelengths. A few hundred young low mass stars are found in the immediate vicinty of IRS 2, and may be directly exposed to its winds and ionizing radiation. The region as a whole is estimated to contain around 2000 young stars, with 30 OB star candidates. The region of the ring to the west of IRS 2, containing IRS 1, appears to be particularly active and is likely the site of ongoing (triggered) star formation. The IRS 1 ridge appears to be the interface between the IRS 2 wind and either a similar wind from a high mass star further to the west, or a dense clump of gas that is being ablated by IRS 2. Further observations are needed to test these scenarios. In many ways, RCW 38 appears to be a younger more embedded version of the ONC and deserves further study at higher angular resolution and across the spectrum. ### Acknowledgments. We acknowledge many useful comments from the referee Leisa Townsley. João Alves openly shared his unpublished data and ideas as part of our ongoing collaboration on this region. We thank Tom Megeath for fruitful discussions. We gratefully acknowledge the financial support from NASA contract NAS8-39073 (CXC). This work is based in part on observations made with the Spitzer Space Telescope, operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. Support for this work was provided by NASA. ## References * Allen et al. (2001) Allen, G. E., Petre, R., & Gotthelf, E. V. 2001, ApJ, 558, 739 * Aschenbach et al. (1999) Aschenbach, B., Iyudin, A. F., & Schönfelder, V. 1999, A&A, 350, 997 * Avedisova & Palous (1989) Avedisova, V. S., & Palous, J. 1989, Bull. Astron. Inst. Czechoslovakia, 40, 42 * Batchelor et al. (1980) Batchelor, R. A., Caswell, J. L., Haynes, R. F., Wellington, K. J., Goss, W. M., & Knowles, S. H. 1980, Australian Journal of Physics, 33, 139 * Bertoldi & McKee (1997) Bertoldi, F., & McKee, C. F. 1997, Rev. Mex. Astron. Astrofis. Conf. Series, 6, 195 * Bourke et al. (2001) Bourke, T. L., Myers, P. C., Robinson, G., & Hyland, A. R. 2001, ApJ, 554, 916 * Cantó et al. (2000) Cantó, J., Raga, A. C., & Rodríguez, L. F. 2000, ApJ, 536, 896 * Caswell & Haynes (1987) Caswell, J. L., & Haynes, R. F. 1987, Australian Journal of Physics, 40, 215 * Caswell et al. (1989) Caswell, J. L., Batchelor, R. A., Forster, J. R., & Wellington, K. J. 1989, Australian Journal of Physics, 42, 331 * Cheung et al. (1980) Cheung, L. H., Frogel, J. A., Hauser, M. G., & Gezari, D. Y. 1980, ApJ, 240, 74 * Churchwell et al. (2004) Churchwell, E., et al. 2004, ApJS, 154, 322 * DeRose etal. (2008) DeRose, K.L, Bourke, T. L., Gutermuth, R.A., Wolk, S. J., Megeath, S.T., Alves, J., Nürnberger, D. 2008 AJ, submitted * Epchtein & Turon (1979) Epchtein, N., & Turon, P. 1979, A&A, 72, L4 * Feigelson et al. (2007) Feigelson, E., Townsley, L., Güdel, M., & Stassun, K. 2007, Protostars and Planets V, eds. B. Reipurth, D. Jewitt, & K. Keil, 313 * Frogel & Persson (1974) Frogel, J. A. & Persson, S. E. 1974, ApJ, 192, 351 * Furniss et al. (1975) Furniss, I., Jennings, R. E., & Moorwood, A. F. M. 1975, ApJ, 202, 400 * Gagné et al. (2005) Gagné, M., Oksala, M.E., Cohen, D.H., Tonnesen, S.K., ud-Doula, A. et al. 2005, ApJ, 628, 986 * Gillespie et al. (1979) Gillespie, A. R., White, G. J., & Watt, G. D. 1979, MNRAS, 186, 383 * Gum (1955) Gum, C. S. 1955, MmRAS, 67, 155 * Huchtmeier (1974) Huchtmeier, W. 1974, A&A, 32, 335 * Johnson (1973) Johnson, H. M. 1973, PASP, 85, 586 * Kaufmann et al. (1977) Kaufmann, P., Scalise, E., Jr., Schaal, R. E., Gammon, R. H., & Zisk, S. 1977, AJ, 82, 577 * Lada & Lada (2003) Lada, C. J. & Lada, E. A. 2003, ARA&A, 41, 57 * Ligori et al. (1994) Ligori S., Moneti A., Robberto M., Guarnieri M. D., Zinnecker H., 1994, Mem. Soc. Astron. Ital., 303, 815 * McGee & Newton (1981) McGee, R. X., & Newton, L. M. 1981, MNRAS, 196, 889 * Moffat _et al._ (2002) Moffat, A.F.J., Corcoran, M.F., Stevens, I.R., Skalkowski, G., Marchenko, S.V. et al. 2002, ApJ, 573, 191. * Mizutani et al. (1987) Mizutani, K., Suto, H., Takami, H., Maihara, T., Sood, R. K., Thomas, J. A., Shibai, H., & Okuda, H. 1987, MNRAS, 228, 721 * Murphy (1985) Murphy, D. C. 1985, Ph.D. Thesis, Massachusetts Institute of Technology * Muzzio (1979) Muzzio, J. C. 1979, AJ, 84, 639 * Muzzio & Celotti de Frecha (1979) Muzzio, J. C. & Celotti de Frecha, M. B. 1979, MNRAS, 189, 159 * Persson et al. (1976) Persson, S. E., Frogel, J. A., & Aaronson, M. 1976, ApJ, 208, 753 * Radhakrishnan et al. (1972) Radhakrishnan, V., Goss, W. M., Murray, J. D., & Brooks, J. W. 1972, ApJS, 24, 49 * Reynolds & Gilmore (1986) Reynolds, S. P., & Gilmore, D. M. 1986, AJ, 92, 1138 * Rho et al. (2004) Rho, J., Ramírez, S. V., Corcoran, M. F., Hamaguchi, K., & Lefloch, B. 2004, ApJ, 607, 904 * Rodgers et al. (1960) Rodgers, A. W., Campbell, C. T., & Whiteoak, J. B. 1960, MNRAS, 121, 103 * Shaver & Goss (1969) Shaver, P. A. & Goss, W. M. 1969, Proceedings Astron. Soc. Australia, 1, 280 * Shaver & Goss (1970) Shaver, P. A. & Goss, W. M. 1970, Australian Journal of Physics, Astrophys. Suppl., 14, 77 * Smith et al. (1999) Smith, C. H., Bourke, T.L., Wright, C.M., Spoon, H.W.W.S., Aitken, D.K., et al. 1999, MNRAS, 303, 367 * Storey & Bailey (1982) Storey, J. W. V. & Bailey, J. 1982, Proceedings Astron. Soc. Australia, 4, 429 * Townsley et al. (2003) Townsley, L. K., Feigelson, E. D., Montmerle, T., Broos, P. S., Chu, Y.-H., & Garmire, G. P. 2003, ApJ, 593, 874 * Vigil (2004) Vigil M., 2004, M.Sc. Thesis, Massachusetts Institute of Technology * Whitney et al. (2004) Whitney, B. A., et al. 2004, ApJS, 154, 315 * Wilson et al. (1970) Wilson, T. L., Mezger, P. G., Gardner, F. F., & Milne, D. K. 1970, A&A, 6, 364 * Wolk et al. (2002) Wolk, S. J., Bourke, T. L., Smith, R. K., Spitzbart, B., & Alves, J. 2002, ApJ, 580, L161 * Wolk et al. (2006) Wolk, S. J., Spitzbart, B. D., Bourke, T. L., & Alves, J. 2006, AJ, 132, 1100 * Yamaguchi et al. (1999) Yamaguchi, R., Saito, H., Mizuno, N., Mine, Y., Mizuno, A., Ogawa, H., & Fukui, Y. 1999, PASJ, 51, 791 * Yusef-Zadeh et al. (2002) Yusef-Zadeh, F., Law, C., Wardle, M., Wang, Q. D., Fruscione, A., Lang, C. C., & Cotera, A. 2002, ApJ, 570, 665 * Zinchenko et al. (1995) Zinchenko, I., Mattila, K., & Toriseva, M. 1995, A&AS, 111, 95	arxiv-papers	2008-08-25T17:08:33	2024-09-04T02:48:57.442481	{ "license": "Public Domain", "authors": "Scott J. Wolk, Tyler L. Bourke and Miquela Vigil", "submitter": "Scott J. Wolk", "url": "https://arxiv.org/abs/0808.3385" }
0808.3417	# Self-healing diffusion quantum Monte Carlo algorithms: methods for direct reduction of the fermion sign error in electronic structure calculations F. A. Reboredo Materials Science and Technology Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA R. Q. Hood Lawrence Livermore National Laboratory, Livermore, CA 94550, USA P. R. C. Kent Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA ###### Abstract We develop a formalism and present an algorithm for optimization of the trial wave-function used in fixed-node diffusion quantum Monte Carlo (DMC) methods. The formalism is based on the DMC mixed estimator of the ground-state probability density. We take advantage of a basic property of the walker configuration distribution generated in a DMC calculation, to (i) project-out a multi-determinant expansion of the fixed-node ground-state wave function and (ii) to define a cost function that relates the fixed-node ground-state and the non-interacting trial wave functions. We show that (a) locally smoothing out the kink of the fixed-node ground-state wave-function at the node generates a new trial wave-function with better nodal structure and (b) we argue that the noise in the fixed-node wave-function resulting from finite sampling plays a beneficial role, allowing the nodes to adjust towards the ones of the exact many-body ground state in a simulated annealing-like process. Based on these principles, we propose a method to improve both single determinant and multi-determinant expansions of the trial wave-function. The method can be generalized to other wave-function forms such as pfaffians. We test the method in a model system where benchmark configuration interaction calculations can be performed and most components of the Hamiltonian are evaluated analytically. Comparing the DMC calculations with the exact solutions, we find that the trial wave-function is systematically improved. The overlap of the optimized trial wave function and the exact ground state converges to 100% even starting from wave-functions orthogonal to the exact ground state. Similarly, the DMC total energy and density converges to the exact solutions for the model. In the optimization process we find an optimal non-interacting nodal potential of density-functional-like form whose existence was predicted in a previous publication [Phys. Rev. B 77, 245110 (2008)]. Tests of the method are extended to a model system with a conventional Coulomb interaction where we show we can obtain the exact Kohn- Sham effective potential from the DMC data. ## I Introduction In diffusion quantum Monte Carlo (DMC) a trial wave-function is used to enforce both the antisymmetry of the electronic many-body wave- functionanderson79 ; ceperley80 ; reynolds82 and the nodal structure of the solution. In highly correlated materials, the accuracy of the trial wave- function becomes increasingly important and determines the success or failure of the method. Indeed, concerns about the fixed-node accuracy have tended to limit applications of DMC to pre-transition metal elements. The discovery and development of new methods to improve the trial wave-functions, ideally without great computational expense, is consequently highly desirable for almost all DMC calculations. In DMC calculations the trial wave-function $\Psi_{T}({\bf R})$ is commonly a product of an antisymmetric function $\Phi_{T}({\bf R})$ and a Jastrow factor $e^{J({\bf R})}$. Usually $\Phi_{T}({\bf R})$ is a Slater determinant constructed with single particle Kohn-Sham orbitals from density functional theory (DFT) or from other mean field approaches such as Hartree-Fock. The Jastrow factor does not change the nodes, but accelerates convergence and improves the algorithm’s numerical stability. The Jastrow factor is optimized in a previous variational Monte Carlo (VMC) calculation. The DMC algorithm finds the lowest energy of the set of all wave-functions that share the nodes of $\Psi_{T}({\bf R})$. The exact ground-state energy will be obtained only if the exact nodes are provided. Since any change to an antisymmetric wave- function must result in a higher energy than the antisymmetric ground state, the energy obtained with arbitrary nodes is an upper bound to the exact ground-state energy.anderson79 ; reynolds82 Only in small systems is it currently possible to improve the nodes bajdich05 ; filippi00 ; umrigar07 ; rios06 ; luchow07 or even avoid the trial wave-function approach altogether.ceperley84 ; kalos00 ; zhang91 For small or weakly correlated systems, where other numerical approaches can compete, the utility of DMC as a method depends crucially on the accuracy of the trial wave-function. Multiple determinant, pfaffian,bajdich05 and back-flowrios06 wave-functions and geminal productsbeaudet are increasingly popular due to the improved accuracy. To improve the DMC energy one must improve the nodal surface of the trial wave-function. However, to our knowledge, all algorithms for wave-function optimization are based on the VMC approach, with any improvement in the DMC energy occurring only as a side-effect. The use of VMC might be a limitation since VMC samples more frequently the regions of the wave-function that have larger probability density and are thus far from the nodes.luchow07 Accordingly, VMC based optimization methods improve first the wave-function at regions which are far from the nodes, while the nodes are only improved indirectly. It has been found, however, that VMC based optimization methods, in general, also improve the DMC energy.umrigar07 ; toulouse08 Nevertheless, a direct optimization of the DMC energy is desirable, and might have improved convergence properties compared to current indirect approaches. While it has been shown by us and others that, within the single Slater determinant approach, the computational cost of an electronic update step in the DMC algorithm can have an almost linear scaling with the number of electrons,williamson ; reboredo05 ; alfe04 the use of these methods is limited if we do not find a better source of trial wave-functions than those obtained from mean-field approaches such as DFT. We recently showedrosetta that Kohn-Sham DFT wave-functions cannot be expected to yield good nodes in general. As correlations increase, Kohn-Sham DFT wave-functions can be bad sources of nodal surfaces.rosetta Indeed, we also found that as the size of the system increases, the nodal error of DFT wave-functions might be of the order of the triplet excitation energies, precluding the prediction of accurate optical propertiestiago08jpc even for simple carbon fullerenes. Accordingly, it is highly desirable to find a method to (i) obtain trial wave- functions with accurate nodal structures, (ii) retain the simplicity of a mean field approach, or (iii) use a minimum number of Slater determinants i.e., the wave-functions are compact and easily evaluated, (iv) directly optimize the nodes in DMC, and (v) improve the nodal structure systematically independently of the starting point. In this contribution we provide such a method. In order to use DMC to find the best trial wave-function we overcome two major obstacles: (i) obtain a representation of the fixed-node ground-state DMC wave-function suitable for optimization of the nodes, and (ii) find a method to keep the trial wave-function compact in large systems by minimizing the number of determinants. This work is the natural continuation of a recent article (Ref. rosetta, ) where we proved the existence of an optimal effective nodal potential for generating the orbitals in the determinants in the trial wave-function used in DMC. While some details are rederived here, we recommend reading Ref. rosetta, before this article. We previously provedrosetta that specific properties of the interacting ground state can be retained via minimization of cost functions in the set of pure-state non-interacting densities. Each cost function defines the gradient of an effective non-interacting potential which is optimized in a Newton-Raphson-like approach until the cost function reaches a minimum. In this paper we take the next step: we use known properties of the walker distribution function generated in a DMC run to define a cost function relating the non-interacting wave-functions with the fixed-node ground-state wave-function. This allows us to obtain, for example, the Kohn-Sham potential or an effective nodal potential from the DMC calculation. The method appears to be limited by the quality of the fit, the statistics that one can collect in DMC and the representability of the nodal surface, which becomes increasingly more demanding as the number of electrons in the system increases. Although this might limit the applicability of the method to systems with small electron counts, we note that DMC is readily parallelized with excellent scaling on modern computers. We also expect that improved sampling and optimization schemes can be constructed using the initial ideas and methods presented here and in Ref. rosetta, . The remainder of this paper is organized as follows. In Section II we demonstrate that the nodes can be improved by locally removing the kinks in the fixed-node ground state. In Section III we derive a formalism and a method to obtain a multi-determinant expansion of the fixed-node ground-state wave- function directly from a DMC run. For many applications, this expansion may already be sufficient. In Section IV we present a cost function that allows the optimization of more compact trial wave-functions that match the fixed- node ground state. A formalism for wave-function optimization based on an effective DFT-like nodal potential is given. In Section V we apply and compare these methods to a model system that can be solved nearly analytically and demonstrate its convergence properties. In Section VI we propose a general algorithm based on the experience gathered solving the model. Finally in Section VII we summarize and discuss the prospects of this method for application in large systems. ## II Systematic reduction of the nodal error within DMC The importance sampling DMC algorithm, in the fixed-node approximation, finds the lowest energy fn:energy $E^{DMC}_{T}$ among the set of all wave-functions that share the nodal surface $S_{T}({\bf R})$ where the trial wave-function $\Psi_{T}({\bf R})=0$ and changes sign. The symbol ${\bf R}$ denotes a point in the many-body $3N$ dimensional space of electron coordinates. We denote this wave-function $\Psi_{FN}({\bf R})$ as the fixed-node ground state. It can be shown that $\Psi_{FN}({\bf R})$ corresponds to the ground state of the interacting Hamiltonian containing an additional infinite external potential located at the nodes of $\Psi_{T}({\bf R})$. Figure 1: a) (Color online) Schematic representation of trial wave-function ($\Psi_{T}$, blue dots), fixed-node ground-state ($\Psi_{FN}$, purple continuous), ground-state ($\Psi$, black dash and dots), and new trial wave- function ($\tilde{\Psi}_{T}$, red dashed line) in the direction perpendicular to the nodal surface ($x$). We show that smoothing the kink in the fixed-node wave-function $\Psi_{FN}$ moves the nodes of $\tilde{\Psi}_{T}$ towards the nodes of the ground state $\Psi$ . b) Schematic representation of how the nodal surface evolves, shown with increasing purple line thickness, after each iteration in the algorithm. The noise introduced in the nodes by random fluctuations of the walkers is assumed to correct itself if the statistics is increased from one iteration to the next. The gradient of the fixed-node ground-state wave-function $\Psi_{FN}({\bf R})$ can be discontinuous at the nodal surface $S_{T}({\bf R})$.reynolds82 Indeed, if the nodes of the trial wave-function do not correspond exactly to the nodes $S({\bf R})$ of an eigenstate of the Hamiltonian, the Laplacian of the fixed- node ground-state wave-function must have a $\delta({\bf R})$ contribution at least on part of $S_{T}({\bf R})$. Otherwise, since the time independent Schrödinger equation is satisfied elsewhere by $\Psi_{FN}({\bf R})$ with an energy $E^{DMC}_{T}$, without this delta in the Laplacian at the nodal surface, $\Psi_{FN}({\bf R})$ would be an eigenstate of the Hamiltonian. This implies that the gradient of $\Psi_{FN}({\bf R})$ must be discontinuous at least at one point of $S_{T}({\bf R})$ if the nodal surface $S_{T}({\bf R})\neq S({\bf R})$. In Fig. 1a we show a schematic representation of the trial wave-function $\Psi_{T}({\bf R})$, the ground-state wave-function $\Psi({\bf R})$ and the fixed-node ground-state $\Psi_{FN}({\bf R})$. In this section we show that when this kink in $\Psi_{FN}({\bf R})$ is locally smoothed away as $\displaystyle\tilde{\Psi}_{T}({\bf R})$ $\displaystyle=$ $\displaystyle\int{\bf dR^{\prime}}\Psi_{FN}({\bf R^{\prime}})\tilde{\delta}\left({\bf R^{\prime}-R}\right)$ $\displaystyle=$ $\displaystyle\int{\bf dR^{\prime}}\Psi_{FN}({\bf R+R^{\prime}})\tilde{\delta}\left({\bf R^{\prime}}\right),\;$ the nodes of the resulting functions improve for a broad class of $\tilde{\delta}\left({\bf R-R^{\prime}}\right)$. Provided that $\Psi_{FN}$ is an antisymmetric function with finite projection on the ground state $\Psi_{0}$, it has been shownanderson79 ; HLRbook that $\Psi$, and its nodes converge to the exact ground state $\Psi=\lim_{t\rightarrow\infty}e^{-t(\hat{\mathcal{H}}-E_{T})}\Psi_{FN}$ (2) where $\hat{\mathcal{H}}$ is the Hamiltonian and $E_{T}$ is an estimate for the ground-state energy. Setting $t=M\tau$ in Eq. (2) yields the equivalent equation $\Psi=\lim_{M\rightarrow\infty}\left(e^{-\tau(\mathcal{H}-E_{T})}\right)^{M}\Psi_{FN}\;.$ (3) In the limit of small $\tau$ a real-space linear-order expansion of $e^{-\tau(\mathcal{H}-E_{T})}$ takes the form $\displaystyle\tilde{\delta}\left({\bf R^{\prime}-R}\right)$ $\displaystyle=$ $\displaystyle(2\pi\tau)^{\frac{-3N}{2}}e^{-\tau\left(V({\bf R})-E_{T}\right)}e^{\frac{-({\bf R-R^{\prime}})^{2}}{2\tau}}$ (4) $\displaystyle\simeq$ $\displaystyle\sum_{n}e^{-\tau(E_{n}-E_{T})}\left\|\Psi_{n}\right>\left<\Psi_{n}\right\|$ where $V({\bf R})$ is the potential energy term (including interactions) in the Hamiltonian and the $E_{n}$ are eigenvalues of the eigenvectors $\Psi_{n}$. Replacing the first line in Eq. (4) in Eq. (II) we obtain a function $\tilde{\Psi}_{T}({\bf R})$ that has, by construction [see Eq. (4) second line], an energy less than or equal to the energy of $\Psi_{FN}({\bf R^{\prime}})$ (being equal for $S_{T}({\bf R})=S({\bf R})$). This form of trial wave-function is similar to a shadow wave-function.shadow ; fn:purediff If we could evaluate Eq. (II) analyticallyfn:bosons and use the result $\tilde{\Psi}_{T}({\bf R})$ in a new DMC run, we would obtain a new fixed-node ground-state wave-function with an even lower DMC energy. This implies that the nodes of $\tilde{\Psi}_{T}({\bf R})$ are better than the ones of $\Psi_{T}({\bf R})$. Note that Eq. (4) tends to the Dirac $\delta$ function as $\delta({\bf R})=(2\pi\tau)^{-3N/2}e^{-({\bf R-R^{\prime}})^{2}/2\tau}$ for $\tau\rightarrow 0$. The factor $e^{-\tau\left(V({\bf R})-E_{T}\right)}$ in Eq. (4) does not alter the nodes: it is a positive scalar function (only acts as a branching term in a one time step simulation). Accordingly, to linear order in $\tau$, only the Gaussian is required to improve the nodes. In turn the Gaussian factor can be replaced by any other approximation of the $\delta$ function as long as it does the same to the nodes of $\Psi_{FN}({\bf R})$ as some Gaussian for small $\tau$. In order to determine the class of smoothing functions that move the node in Eq. (II) as a Gaussian, we consider a patch $dS({\bf R})$ of the nodal surface $S_{T}({\bf R})$ centered at ${\bf R_{S}}$ with a diameter small enough (so that it can be considered a flat hyper-plane) but much larger than $\sqrt{\tau}$. The integration of the 3N dimensional Gaussian in the directions of the hyper-plane leads to a one dimensional Gaussian $G(x/\sqrt{\tau})=e^{-x^{2}/2\tau}/\sqrt{2\pi\tau}$. Any approximation of $\delta({\bf R})$ after integration in $3N-1$ coordinates should result in a function $d(x)$ that can be rescaled and translated to satisfy the following properties: $\int d(x)dx=1\;\;\int xd(x)dx=0\;\text{and}\;\int x^{2}d(x)dx=1\;.$ (5) In the immediate vicinity of ${\bf R_{S}}$, the function $\Psi_{FN}({\bf R})$ depends only on the coordinate in the direction normal to the surface ${\bf n_{S}}$ defined as $x={\bf(R-R_{S})\cdot n_{S}}={\bf\Delta R\cdot n_{S}}$. For $x\rightarrow 0$ we can approximate $\displaystyle\Psi_{FN}({\bf R_{S}+\Delta R})\simeq\Psi_{FN}({\bf R_{S}})+c_{1}x+k_{1}\|x\|$ $\displaystyle+$ $\displaystyle+c_{2}x^{2}+k_{2}(x-\|x\|)^{2}+O(x^{3})$ (6) and $\displaystyle\frac{d}{dx}\Psi_{FN}({\bf R_{S}+\Delta R})\simeq c_{1}+k_{1}\mathrm{sign}(x)+$ (7) $\displaystyle+2c_{2}x+4\;k_{2}(x-\|x\|)+O(x^{2}).$ In Eq. (II) the wave-function is expanded as a combination of a smooth function (with coefficients $c_{1}$ and $c_{2}$) plus a kink ($k_{1}$ and $k_{2}$). Replacing Eq. (II) and Eq. (7) into Eq. (II) and replacing the Gaussian by a generic approximation of $\delta(x)=d(x/\sqrt{\tau})/\sqrt{\tau}$ we get: $\tilde{\Psi}_{T}({\bf R_{S}})=k_{1}A[d]\sqrt{\tau}+O(\tau)$ (8) and the first derivative $\frac{d}{dx}\tilde{\Psi}_{T}({\bf R_{S}})=c_{1}+k_{1}S[d]+4k_{2}A[d]\sqrt{\tau}+O(\tau)$ (9) where $A[d]=\int\|x\|d(x)dx$ and $S[d]=\int\mathrm{sign}(x)d(x)dx$. Note that if $d(x)$ has the Gaussian form $A[G]=\sqrt{2/\pi}>0$ and $S[G]=0$. Using Eqs. (8) and (9) we can estimate the displacement of the node to be $\Delta x\simeq-\frac{k_{1}A[d]}{c_{1}+k_{1}S[d]}\sqrt{\tau}+O(\tau).$ (10) Therefore, for any symmetric approximation of the $\delta$ function $S[d]=0$, provided that $A[d]>0$, one can obtain the same displacement in the node as a Gaussian with $\tau^{\prime}=\pi A[d]^{2}\tau/2$. For a non-symmetric $d(x)$, the node will move in the same direction as long as the sign in the denominator of Eq. (10) does not change. However, a uniform rescaling of $\tau$ to match the Gaussian form will no longer be possible. That means that the node will move faster towards the exact node in some regions of the surface than in others. Thus, as long as the approximation of the delta used for smoothing is a function of the distance only, with $A[d]>0$, one can find some Gaussian that moves the node in the same way for every patch $dS({\bf R})$. This movement corresponds to a better node. The restrictions in $d(x)$ can be alleviated by using a repeated convolution. Using the central limit theorem it can be shown that a recursive convolution of any approximation of $d(x)$ tends to a Gaussian as long as the Taylor expansion of its Fourier transform exists. Thus if the shape of $d(x)$ is not known the method would be more stable if it is applied sequentially. In section III we will use a smoothing function of the form $\tilde{\delta}\left({\bf R,R^{\prime}}\right)=\sum_{n}^{\sim}\Phi_{n}({\bf R})\Phi_{n}^{}({\bf R^{\prime}}),$ (11) where the $\Phi_{n}({\bf R})$ are continuous functions without kinks forming a complete basis and the “$\sim$” in $\sum_{n}^{\sim}$ means that only some elements are included in the sum (with a criterion described below). If the $\Phi_{n}({\bf R})$ in Eq. (11) are obtained from a non-interacting problem and the criterion for truncation is an energy cutoff, it can be shown that the resulting function is only a function of the distance ${\bf(R-R^{\prime})}^{2}$. Since in that limit only plane waves of large energy are added to Eq. (11) and all the lower plane waves are included in the lower energy components, the basis can be transformed with a unitary transformation into a plane-wave basis with a spherical cutoff in reciprocal space. If there is the same number of plane waves in any direction the results of Eq. (11) only depend on the distance which implies that $S[d]=0$. Since we restrict the sum in Eq. (11) to fermionic antisymmetric $\Phi_{n}({\bf R})$, Eq. (11) expands an antisymmetrized deltafn:bosons . This form projects out any non-fermionic component introduced in the wave-function along the DMC algorithm as in the A-function approach used by Bianchi and collaborators.bianchi93 In Section IV we propose a simple interpolation scheme to smooth the node where the expansion used in Eq. (11) is not taken to the high energy cutoff limit. The fact that these smoothing methods work in practice suggests that the conditions to improve the nodes are extended beyond the exact equivalence to a Gaussian form. Note that a discontinuity of the gradient of the fixed-node wave-function $\Psi_{FN}({\bf R})$ at the node impliesfn:purediff that, if walkers are distributed according to $\Psi_{FN}({\bf R})$ with the sign (or phase) of $\Psi_{FN}({\bf R})$, there will be more walkers in the vicinity of one side of the nodal surface than on the other. Accordingly, if these walkers are released in a pure diffusion algorithm,HLRbook for $\tau\rightarrow 0$ they will cross, on average, more from one side of the nodal surface than from the other. The nodes defined by the population of these signed walkersHLRbook would move in the same direction that would result from smoothing the kink in $\Psi_{FN}({\bf R})$ provided the time step is short enough and kinetic energy term in the Green’s function [Eq. (4)] is dominant. Consequently, the nodes can be improved by moving them in the direction of least “walker pressure” within a pure diffusion approach. Any method to obtain $\Psi_{FN}({\bf R})$ from the walker distribution in a DMC runbianchi96 will carry the error of statistical fluctuations from using a finite sample of walkers. Even if the $\Psi_{FN}({\bf R})$ is forced to remain antisymmetric,bianchi96 the nodes might move in the wrong direction because of these fluctuations. We assume the method is robust against these random fluctuations when applied recursively, and can form the basis of an optimization process to improve the trial wave-function. Note that if incorrect fluctuations increase the kink in $\Psi_{FN}({\bf R})$ at the node, the probability to sample the correct fixed-node wave-function will remain higher and also the probability to move the node in the correct direction in successive iterations. Conversely, fluctuations that correctly improve the nodes will be reinforcedfn:reinforced in successive iterations. Since these fluctuations are reduced when the statistical sampling is improved, the nodal surfaces will converge to the true nodes if the statistics is improved from one iteration to the next (Fig. 1b). Note that we do not claim that this process is necessarily the most efficient optimization approach: more sophisticated iterative methods and optimization algorithms are clearly possible. Summarizing, we should be able to improve the nodes systematically provided we can obtain the anti-symmetric function $\Psi_{FN}({\bf R})$ from the walker configurations (probability distribution) of a DMC calculation after convolution with a smoothing function.fn:relnode ## III Determination of the fixed-node ground-state wave-function from the DMC probability distribution ### III.1 Sampling the fixed-node ground-state wave-function The distribution function of the walkers in an importance sampling DMC algorithm is given by:ceperley80 $\displaystyle f({\bf R})$ $\displaystyle=$ $\displaystyle\Psi_{T}^{}({\bf R})\Psi_{FN}({\bf R})$ $\displaystyle=$ $\displaystyle\lim_{N_{c}\rightarrow\infty}\frac{1}{N_{c}}\sum_{i}^{N_{c}}\delta\left({\bf R-R}_{i}\right)$ where $\Psi_{T}({\bf R})$ typically has the Slater-Jastrow form $\Psi_{T}({\bf R})=e^{J({\bf R})}\Phi_{T}({\bf R});$ (13) in which $\Phi_{T}({\bf R})$ consists of a single determinant for each electronic spin component composed of single-particle orbitals. The results of this paper are also valid if $\Phi_{T}({\bf R})$ has a more general form such as consisting of multi-determinant expansions for each spin component and/or containing back-flow or two-particle pfaffians. The ${\bf R}_{i}$ in Eq. (III.1) correspond to the positions of an equilibrated ensemble of $N_{c}$ configurations in a DMC algorithm (we have set the weights equal to one for simplicity). We note that $\Psi_{FN}({\bf R})$ in Eq. (III.1) can be rewritten as an antisymmetric function times the Jastrow factor $e^{J({\bf R})}$ as $\displaystyle\Psi_{FN}({\bf R})$ $\displaystyle=$ $\displaystyle e^{J({\bf R})}e^{-J({\bf R})}\Psi_{FN}({\bf R})$ (14) $\displaystyle=$ $\displaystyle e^{J({\bf R})}\sum_{n}\lambda_{n}\langle{\bf R}\|(\prod c^{{\dagger}}\prod c)\|\Phi_{T}\rangle$ $\displaystyle=$ $\displaystyle e^{J({\bf R})}\sum_{n}\lambda_{n}\Phi_{n}({\bf R})$ where $\sum\lambda_{n}(\prod c^{{\dagger}}\prod c)\|\Phi_{T}\rangle$ is a complete configuration interaction (CI) expansion in the basis of electron- hole pairs . Accordingly, in Eq. (14) the $\Phi_{n}({\bf R})$ are Slater determinants or pfaffiansbajdich05 obtained from replacing in $\Phi_{T}({\bf R})$ some of the occupied $\phi_{\nu}$ single particle functions by unoccupied $\phi_{n}$ functions, accordingly $\int{\bf dR}\Phi_{n}^{}({\bf R})\Phi_{m}({\bf R})=\delta_{n,m}$. In practice, the CI expansion can be truncated retaining, for example, only the $\Phi_{m}({\bf R})$ with a non-interacting energy below a given energy cutoff. The CI expansion in principle consists of all single, double, triple, quadruple and higher-order excitations. By analogy with conventional CI calculations, the higher-order excitations are expected to contribute less to the wave-function than low-order excitations. As the kinetic energy of higher- order excitations increases as compared with the interaction, their contribution to the ground-state wave-function decreases. While a Jastrow factor $e^{J({\bf R})}$ is not formally required in a complete expansion of the wave-function in Eq. (14), it is believed that the introduction of a Jastrow factor limits the number of coefficients required in the multi-determinant expansion, due in part to the more efficient description of the electron-electron cusp. For some applications it may be desirable to not employ a Jastrow factor, since the extracted wave-function may be more easily used in later analysis. Replacing Eq. (14) and Eq. (13) in Eq. (III.1) we obtain $f({\bf R})=e^{2J({\bf R})}\Phi_{T}^{}({\bf R})\sum_{n}\lambda_{n}\Phi_{n}({\bf R}).$ (15) Borrowing a method from Optimized Effective Potentials (OEP) we define the following projectors:us ; note $\xi_{n}({\bf R})=e^{-2J({\bf R})}\frac{\Phi_{n}({\bf R})}{\Phi_{T}({\bf R})}.$ (16) Note that the projectors $\xi_{n}({\bf R})$ are symmetric (bosonic) functions.fn:bosons Replacing $f({\bf R})$ by Eq. (15), using the definition of $\xi_{n}({\bf R})$ [Eq. (16)] and the orthogonality condition it can be demonstrated that $\int{\bf dR}f({\bf R})\xi_{n}^{}({\bf R})=\lambda_{n}\;.$ (17) Thus, the coefficients of the multi-determinant expansion Eq. (14) of the fixed-node DMC ground-state wave-function can be estimated directly as a sum over the total number of walkers $N_{c}$ along the DMC random walk, using the second line of Eq. (III.1), as $\langle\lambda_{n}\rangle=\frac{1}{N_{c}}\sum_{i=1}^{N_{c}}\xi_{n}^{}({\bf R}_{i})\;\gamma({\bf R}_{i})$ (18) where $\gamma({\bf R}_{i})=\frac{-1+\sqrt{1+2\|{\bf v}\|^{2}\tau}}{\|{\bf v}\|^{2}\tau}\text{ with }{\bf v}=\frac{\nabla\Psi_{T}({\bf R}_{i})}{\Psi_{T}({\bf R}_{i})}.$ (19) For convenience we divided by the number of walkers $N_{c}$ in Eqs. (III.1) and (18) since the normalization constant of $\Psi_{FN}({\bf R})$ and the corresponding coefficients $\lambda_{n}$ is arbitrary. The factor $\gamma({\bf R}_{i})$ in Eq. (18) is a time step, $\tau$, correction derived following Ref. umrigar93, that corrects the divergences of the projectors $\xi_{n}({\bf R}_{i})$ at the nodes. This correction is not always applied to estimators (e.g. the local energy) but we find that it reduces the error of the wave- function coefficients. For an uncorrelated sample of walker configurations the error bar of the multi-determinant expansion can be determined from $\displaystyle\langle\lambda_{n}^{2}\rangle$ $\displaystyle=$ $\displaystyle\frac{1}{N_{c}}\sum_{i=1}^{N_{c}}\|\xi_{n}({\bf R}_{i})\|^{2}\gamma({\bf R}_{i})^{2}$ (20) $\displaystyle\langle{\sigma_{n}}\rangle$ $\displaystyle=$ $\displaystyle\sqrt{\frac{\langle\lambda_{n}\rangle^{2}-\langle\lambda_{n}^{2}\rangle}{N_{c}}}$ $\displaystyle\lambda_{n}$ $\displaystyle\simeq$ $\displaystyle\langle\lambda_{n}\rangle\pm\frac{\langle{\sigma_{n}}\rangle}{\sqrt{N_{c}-1}}.$ As $N_{c}\rightarrow\infty$ in Eq. (20) the error bar in the multi-determinant coefficients $\lambda_{n}$ goes to zero. As usual, the error bars can be used to monitor convergence of the calculation. While the eventual goal is to obtain small error bars, we found in practice it is better to start with $N_{c}$ small and then to slowly increase it with each iteration as the trial wave-function improves (see below). By substituting Eqs. (III.1), (13), and (16) into Eq. (17) and defining the fixed-node function $\Phi_{FN}$ in terms of the trial function Jastrow and the fixed-node wave-function $\Psi_{FN}$ $\Psi_{FN}({\bf R})=e^{J({\bf R})}\Phi_{FN}({\bf R})$ (21) one can obtain this expression $\lambda_{n}=\int{\bf dR}\Phi_{n}^{}({\bf R})\Phi_{FN}({\bf R})$ (22) for $\lambda_{n}$. We define $\tilde{\Psi}_{T}({\bf R})$ to be the truncated expansion (denoted using $\sim$) of Eq. (14) $\tilde{\Psi}_{T}({\bf R})=e^{J({\bf R})}\sum_{n}^{\sim}\lambda_{n}\Phi_{n}({\bf R})\;.$ (23) Substituting Eq. (22) into Eq. (23) yields the equation $\tilde{\Psi}_{T}({\bf R})=e^{J({\bf R})}\int{\bf dR^{\prime}}\left[\sum_{n}^{\sim}\Phi_{n}({\bf R})\Phi_{n}^{}({\bf R}^{\prime})\right]\Phi_{FN}({\bf R}^{\prime})\;.$ (24) In Section II we showed that the appearance of a smoothing function of the form of Eq. (11) as in the term in brackets in Eq. (24) will smooth the nodes of $\Phi_{FN}({\bf R}^{\prime})$ yielding better nodes for $\tilde{\Psi}_{T}({\bf R})$. Since the $\Phi_{n}({\bf R})$ are selected to be eigenvectors of a non-interacting problem, highly localized features of $\Phi_{FN}({\bf R})$ would require components with high eigenvalues. At the same time, resolving those details would require a large number of configurations to improve the statistics. Accordingly, we truncate the expansion in Eq. (23) to the coefficients with relative errors smaller than 25%. Note that as the statistics is improved, the error bars diminishes, the number of functions retained in Eq. (11) increases and so does the localization of $\tilde{\delta}\left({\bf R,R^{\prime}}\right)$. Thus the conditions to improve the nodes systematically as described in Section II are reached as the statistics improves. ### III.2 Sampling the Jastrow factor Instead of expressing $\Psi_{FN}({\bf R})$ as a product of the same Jastrow factor used in $\Psi_{T}({\bf R})$ times a different multi-determinant expansion, one can choose to optimize the Jastrow factor while using the same antisymmetric function $\Phi_{T}({\bf R})$. It is easy to show that there is a symmetric bosonic factor that turns $\Phi_{T}({\bf R})$ into $\Psi_{FN}({\bf R})$ which is formally given by $e^{\tilde{J}({\bf R})}=\frac{\Psi_{FN}({\bf R})}{\Phi_{T}({\bf R})}\;.$ (25) Replacing Eq. (14) in Eq. (25) we find $\displaystyle e^{\tilde{J}({\bf R})}$ $\displaystyle=$ $\displaystyle e^{J({\bf R})}\sum_{n}\lambda_{n}\frac{\Phi_{n}({\bf R})}{\Phi_{T}({\bf R})}$ (26) $\displaystyle=$ $\displaystyle e^{3J({\bf R})}\sum_{n}\lambda_{n}\xi_{n}({\bf R})\;.$ Note that the product $e^{\tilde{J}({\bf R})}\Phi_{T}({\bf R})$ yields Eq. (14). While this shows that the projectors $\xi_{n}({\bf R})$ could be used to improve the Jastrow factor, since they diverge for $\Phi_{T}({\bf R})\rightarrow 0$, it is necessary to fit instead a continuous functional form using values away from the nodes where truncation and sampling errors play a dominant role (see Section IV). Updating the multi-determinant expansion of the antisymmetric part of the new trial wave-function, see Eq. (23), alters the nodes because (i) the expansion is truncated and (ii) the coefficients of the multi-determinant expansion have a random error due to finite sampling in Eq. (18). On the other hand, updating the Jastrow factor, see Eq. (26), keeps the nodes fixed but reduces the number of determinants required and the overall computational cost. There is a compromise between accuracy and speed.reynolds82 A very good wave-function might have a very small variance in the local energy, but if it is expensive to evaluate one might obtain the same statistical error in less wall-clock time with a faster lower quality wave-function. In an ideal case, if the nodes are $v$-representable (see below and Ref. rosetta, ) only a single determinant is required to describe the fixed-node ground-state wave-function to sufficient accuracy. In practice, the form of the Jastrow factor $e^{\tilde{J}({\bf R})}$ is unknown, while an infinite multi-determinant expansion is infeasible. This implies that both the factors in Eq. (14) are required in general; an efficient scheme will optimize both the Jastrow factor and determinantal part of the wave-function. Particularly for the case of a metallic system, the cost of a multi-determinant expansion might be prohibitive due to the large number of low-energy excitations. In this case it might be preferable to concentrate on an optimized Jastrow factor.wood06 ### III.3 A simple self-healing DMC algorithm We have formulated, for small systems, a working iterative algorithm based on a multi-determinant or multi-pfaffian expansion of the fixed-node ground-state wave-function. In this algorithm the calculated coefficients Eq. (18) of the expansion are used to form a new trial wave-function defined by Eq. (23). Initially the statistical errors present in $\lambda_{n}$ due to finite sampling appear to have a beneficial role, particularly when the initial trial wave-function has poor nodes. Note that in the limit of an infinite number of determinants in Eq. (23) with no statistical sampling errors in $\lambda_{n}$ the trial wave-function would exactly reproduce the fixed-node wave-function, and an iterative improvement of the nodes would not be possible. Statistical fluctuations in the coefficients $\lambda_{n}$ allow the nodes to move. In the next iteration regions near beneficial fluctuations are revisited by walkers while bad statistically insignificant fluctuations tend not to propagate or grow. This stability against random noise appears to be valid in practice. Thus, the statistical error in the coefficients plays the role of a random thermal fluctuation in a simulated annealing algorithm.correa05 It is ironic and remarkable that random errors can be used to eliminate systematic errors. While it is relatively economical to calculate a large number of multi- determinants every autocorrelation length, as more determinants are included in the trial wave-function each time step of the DMC calculation becomes more demanding. Accordingly, for large or continuum systems a method to minimize the number of determinants used to represent a given nodal surface is required. This is described in the next section. ## IV Derivation of the best nodal-effective potential from DMC While a working multi-determinant algorithm can be constructed on the basis of the multi-determinant expansion of the previous section, a significant step forward can be taken using the theory developed in Ref. rosetta, and taking advantage of Eq. (III.1) to construct a new trial wave-function that can be evaluated more efficiently than the multi-determinant expansion Eq. (23). This method will be most effective when the initial single particle orbitals involved in $\Phi_{T}({\bf R})$ are poor, e.g. if the system is strongly correlated. ### IV.1 A cost function for the DMC algorithm Given a probability density $p({\bf R})$ and a binned statistical sample of $N_{c}$ configurations of the random variable ${\bf R}$, we can define a new random variable $\chi^{2}=\sum_{i}^{M}\frac{\left[n_{i}-N_{c}\Omega_{i}\bar{p}({\bf R}_{i})\right]^{2}}{N_{c}\Omega_{i}\bar{p}({\bf R}_{i})}$ (27) which is distributed by the Chi-squared distribution function.HLRbook In Eq. (27) $\Omega_{i}$ is the volume of the bin $i$, with $n_{i}$ configuration counts, $\bar{p}({\bf R}_{i})$ is the average of $p({\bf R})$ in $\Omega_{i}$ and $M$ is the number of bins. Each term in Eq. (27) is the square deviation of $n_{i}$ divided by the expectation value of the mean. In the limit of large counts the square of the mean is expected to be equal to the square deviation for the Poisson distribution of counts in a bin. Accordingly, in $\chi^{2}$ relative deviations from the mean have the same impact independently of the absolute value of the probability density. We will take advantage of this property to replace a wave-function difficult to evaluate Eq. (III.1) by a simpler approximate one that retains key properties. Setting $n_{i}=N_{c}\Omega_{i}\bar{q}({\bf R}_{i})$ in Eq. (27), dividing by $N_{c}$ taking the limit $M\rightarrow\infty$, and using the mean value theorem, we find a cost function to compare two continuous distribution functions: $K_{pq}=\int{\bf dR}\frac{\left[q({\bf R})-p({\bf R})\right]^{2}}{p({\bf R})}\;.$ (28) We showed in Ref. rosetta, that if we wish to preserve properties, other than the density, cost functions can be defined relating the many-body ground-state $\Psi({\bf R})$ with a non-interacting wave-function $\Phi_{T}({\bf R})$. The walker distribution functionceperley80 given by Eq. (III.1) allows us to construct several cost functions relating the wave-function to optimize with the exact fixed-node ground-state $\Psi_{FN}({\bf R})$. Using Eq. (28) as a guide, we propose the following expression: $\displaystyle K_{DMC}$ $\displaystyle=$ $\displaystyle\int{\bf dR}\frac{\left\|\mu\;\Psi_{T}({\bf R})\tilde{\Psi}_{T}({\bf R})-f({\bf R})\right\|^{2}}{\left\|\mu\;\Psi_{T}({\bf R})\tilde{\Psi}_{T}({\bf R})\right\|}\times$ (29) $\displaystyle\theta\left(f({\bf R})-\eta\right),$ where $\tilde{\Psi}_{T}({\bf R})$ is a trial wave-function to be optimized, $\mu=\left[\int\Psi_{T}({\bf R})\tilde{\Psi}_{T}({\bf R}){\bf dR}\right]^{-1}$, $f({\bf R})$ is given by Eq. (15) with coefficients obtained from a previous DMC run using Eq. (18), $\theta(x)$ is the Heaviside function, and $\eta$ is a small positive n umber. Note in Eq. (29) that the first factor vanishes when $\tilde{\Psi}_{T}({\bf R})\rightarrow\Psi_{FN}({\bf R})$. Indeed, if $\tilde{\Psi}_{T}({\bf R})$ is constrained to have the nodal surface $S_{T}({\bf R})$ and the sign (or phase) of $\Psi_{T}({\bf R})$, the integral of the first factor in Eq. (29) measures the probability that the distribution of a given ensemble of walkers $f({\bf R})$ corresponds to the distributionHLRbook $\alpha({\bf R})=\mu\tilde{\Psi}_{T}({\bf R})\Psi_{T}({\bf R})\;.$ (30) In Eq. (29), we add an absolute value function in the denominator of the first factor and a Heaviside function in order to extend the set of $\tilde{\Psi}_{T}({\bf R})$ where the cost function can be evaluated beyond the fixed-node space. Note that, since $f({\bf R})>0$, while negative values for $\alpha({\bf R})$ are allowed, they are penalized in the numerator more than positive values. In Eq. (29), we add $\mu$ to enforce $\int\alpha({\bf R}){\bf dR}=\int f({\bf R}){\bf dR}$ for any $\tilde{\Psi}_{T}({\bf R})$. In Eq. (29) the nodes of $\tilde{\Psi}_{T}({\bf R})$ can move within a distance [which depends on $\eta$ and $f({\bf R}$)] around $S_{T}({\bf R})$. Otherwise, if the zeros of the numerator and denominator of Eq. (29) do not match, the value of the cost function would rise to infinity. An additional effect of $\theta$ is that any kink of $\Psi_{FN}({\bf R})$ at the node is not enforced by the cost function in $\tilde{\Psi}_{T}({\bf R})$. Since $\tilde{\Psi}_{T}({\bf R})$ will be obtained from the minimum energy solution of a non-interacting problemrosetta and departures at the node are not penalized, it will interpolate smoothly avoiding a kink. Note that we can choose alternative cost function forms. For example, we can replace the denominator in Eq. (29) by $f({\bf R}$). This choice would simplify the derivatives of the cost function but it has a couple of disadvantages: First $f({\bf R})$ is expected to be a very noisy function when its magnitude is small, while the product of non-interacting $v$-representable wave-functions $\alpha({\bf R})=\mu\;\tilde{\Psi}_{T}({\bf R})\Psi_{T}({\bf R})$ is expected to be smooth (see IV.2) . We choose not to amplify the noise of $f({\bf R})$ in the denominator. Second, in Eq. (29) a small number for $\alpha({\bf R})$ outside the window defined by the Heaviside function is highly penalized which confines the node of $\tilde{\Psi}_{T}({\bf R})$ to remain inside the window where the Heaviside function is zero. ### IV.2 Representability of the nodal surface Given an interaction in a many-body system, the Hohenberg-Kohn theoremhohenberg establishes a functional correspondence between electronic densities $\rho({\bf r})$, external potentials $V({\bf r})$, and ground-state wave-functions $\Psi({\bf R})$. The subset of densities $\rho({\bf r})$ corresponding to a ground state of an interacting system under an external potential $V({\bf r})$ are denoted as pure state $v$-representable.parr A non-interacting pure state $v$-representable density is given instead by $\bar{\rho}({\bf r})=\sum_{\nu}\|\phi_{\nu}\left({\bf r}\right)\|^{2}$ where $\phi_{\nu}\left({\bf r}\right)$ are Kohn-Sham-likekohn single particle orbitals, or eigenvectors, of the single-particle Hamiltonian: $\left[-\frac{1}{2}{\bf\nabla}^{2}+\bar{V}\left({\bf r}\right)\right]\phi_{\nu}\left({\bf r}\right)=\varepsilon_{\nu}\phi_{\nu}\left({\bf r}\right),$ (31) where $\bar{V}\left({\bf r}\right)$ is an effective single particle potential. The lowest energy Slater determinant constructed with the solution of Eq. (31) is a many-body non-interacting ground state. For simplicity we denote those quantities that are simultaneously interacting and non-interacting $v$-representable as simply $v$-representable. In addition, certain quantities can be multi-determinant $v$-representable, meaning that they can be represented by a finite multi-determinant expansion constructed with the solutions of Eq. (31). Since, the ground-state density $\rho({\bf r})$ determines the ground-state wave-function $\Psi({\bf R})$,hohenberg $\rho({\bf r})$ defines also the points ${\bf R}$ of the nodal surface $S({\bf R})$ where $\Psi({\bf R})=0$. The nodes of the trial wave-function, instead, are by construction those of $\Phi_{T}({\bf R})$ (non-interacting $v$-representable in the single determinant case). The exact nodes $S({\bf R})$ may or may not be representable in this manner.rosetta ### IV.3 Optimization of the effective nodal potential The trial wave-function is often constructed with non-interacting orbitals derived from an effective potential [see Eq. (31)], e.g. from Kohn-Sham DFT. For the moment we will assume that $\tilde{\Psi}_{T}({\bf R})$ is given in the single determinant Slater-Jastrow form: $\tilde{\Psi}_{T}({\bf R})=e^{\tilde{J}({\bf R})}\tilde{\Phi}_{T}({\bf R})$ (this derivation is extended to multiple determinants or pfaffians in IV.6). However, for now, we assume that the node can move within all the non-interacting $v$-representable set, which is a less restrictive condition than the fixed-node approximation but implies accepting an error if $S({\bf R})$ is not $v$-representable. In Ref. rosetta, we showed that, if the trial wave function depends on non- interacting orbitals in an effective potential [as in Eq. (31)], the effective potential $\bar{V}\left({\bf r}\right)$ required to retain a given property is a function of the cost function $K$. To simplify formulae, discussion and notation we assume here that all wave-functions are real. The potential can be obtained by adding recursively the following correction: $dV_{K}({\bf r})=-\epsilon\sum_{\nu}^{o}\int\\!\\!{\bf dr^{\prime}}\\!\\!\frac{\delta K}{\delta\phi_{\nu}\left({\bf r^{\prime}}\right)}\frac{\delta\phi_{\nu}\left({\bf r^{\prime}}\right)}{\delta V_{K}\left({\bf r}\right)}.$ (32) where $\epsilon$ is adjusted during the optimization. Replacing $K$ by $K_{DMC}$ we get $\frac{\delta K_{DMC}}{\delta\phi_{\nu}\left({\bf r^{\prime}}\right)}=\int{\bf dR}W({\bf R})e^{\tilde{J}(\bf R)}\frac{\delta\tilde{\Phi}_{T}({\bf R})}{\delta\phi_{\nu}\left({\bf r^{\prime}}\right)}$ (33) where $W({\bf R})=\frac{\delta K_{DMC}}{\delta\tilde{\Psi}_{T}({\bf R})}\;,$ for which we obtain $\displaystyle W({\bf R})$ $\displaystyle=$ $\displaystyle\frac{2\Delta({\bf R})\alpha({\bf R})-\Delta({\bf R})^{2}}{\|\alpha({\bf R})\|^{2}}\mathrm{sign}(\alpha({\bf R}))$ (34) $\displaystyle\times$ $\displaystyle\left[1-\alpha({\bf R})\right]\mu\;\Psi_{T}({\bf R})\theta\left(f({\bf R})-\eta\right),$ with $\Delta({\bf R})=f({\bf R})-\alpha({\bf R})$. Within first order perturbation theory $\frac{\delta\phi_{\nu}\left({\bf r^{\prime}}\right)}{\delta V_{K}\left({\bf r}\right)}=\sum_{n}^{u}\frac{\phi_{n}({\bf r})\phi_{\nu}({\bf r})}{\varepsilon_{\nu}-\varepsilon_{n}}\phi_{n}\left({\bf r^{\prime}}\right)\;.$ (35) Replacing Eq. (33) and Eq. (35) in Eq. (32), we find $\displaystyle dV_{K_{DMC}}({\bf r})$ $\displaystyle=$ $\displaystyle\epsilon\sum_{\nu}^{o}\sum_{n}^{u}\frac{\phi_{n}({\bf r})\phi_{\nu}({\bf r})}{\varepsilon_{\nu}-\varepsilon_{n}}\beta^{n}_{\nu}\;,$ (36) $\displaystyle\beta^{n}_{\nu}$ $\displaystyle=$ $\displaystyle\int{\bf dR}W({\bf R})e^{\tilde{J}(\bf R)}\tilde{\Phi}_{T,\nu}^{n}({\bf R})\;.$ (37) In Eqs. (32), (35), and (36) we used $\sum_{\nu}^{o}$ ( $\sum_{n}^{u}$ ) to define sums over occupied (unoccupied) states. In turn in Eq. (37) $\tilde{\Phi}_{T,\nu}^{n}({\bf R})$ means replacing the occupied state $\phi_{\nu}$ by $\phi_{n}$ which results from combining the cofactors of $\phi_{\nu}\left({\bf r^{\prime}}\right)$ [ $\frac{\delta\tilde{\Phi}({\bf R})}{\delta\phi_{\nu}\left({\bf r^{\prime}}\right)}$ ] in Eq. (33) with $\phi_{n}\left({\bf r^{\prime}}\right)$ in Eq. (35). The first factor in function $W({\bf R})$ [Eq. (34)] is obtained from the derivative of the cost function Eq. (29) with respect to $\alpha({\bf R})$ [ignoring contributions coming from the discontinuities of $\|x\|$ since the Heaviside function in Eq. (29) is zero near the nodes]. The second factor in $W({\bf R})$ results from the derivative of $\alpha({\bf R})$ with respect to $\tilde{\Psi}_{T}({\bf R})$. [note that $\mu$ is also dependent on $\tilde{\Psi}_{T}({\bf R})$ ] ### IV.4 Optimization of the Jastrow factor within DMC We argued in the previous section that an optimal Jastrow factor can be used to reduce the number of determinants in the multi-determinant expansion. Optimizing the Jastrow factor is important to limit the exponential cost of the CI expansion because, while the Jastrow factor cannot influence the nodes, it can reduce the burden of correcting the probability density from any value given by a Slater determinant (see Eq. (25)). Accordingly, if the Jastrow factor is optimized, the antisymmetric part of the wave-function is free to search for the nodes. Often the $\tilde{J}({\bf R})$ is dependent on a set of parameters $\gamma_{n}$. The value of the cost function (Eq. 29) is also affected by the Jastrow factor $e^{\tilde{J}(\bf R)}$. Thus the gradient of the cost function with respect to an arbitrary change in $e^{\tilde{J}(\bf R)}$ can be obtained within DMC via $\frac{dK_{DMC}}{d\gamma_{n}}=\int{\bf dR}W({\bf R})e^{\tilde{J}(\bf R)}\tilde{\Phi}_{T}({\bf R})\frac{d\tilde{J}({\bf R)}}{d\gamma_{n}}.$ (38) ### IV.5 Discussion Note at this point that (1) both the coefficients $\beta_{\nu}^{n}$ and $\gamma_{n}$ are integrals of the function $W({\bf R})$ which is only dependent on the particular form of the cost function selected in Eq. (29) and a representation of the walkers distribution $f({\bf R})$. (2) The function $f({\bf R})$ is an essential component of $W({\bf R})$ that can be obtained from the DMC run using Eqs. (15) and (18) or sampled directly by binning.binnote (3) Provided that $f({\bf R})$ is known, a distribution of configurations ${\bf R}_{i}$ with probability $\|W({\bf R})\|$ can be generated with the Metropolis algorithm. All integrals of the form $\int{{\bf dR}}g({\bf R})W({\bf R})$ involved in Eqs. (36) and (38) can be evaluated in a single correlated sampling step as $\sum_{i}\mathrm{sign}[W({\bf R}_{i})]g({\bf R}_{i})$ using points ${\bf R}_{i}$ drawn from the probability distribution defined by the absolute value of $W({\bf R})$. (4) In most methods, the Jastrow parameters $\gamma_{n}$ are optimized within a variational Monte Carlo approach (either minimizing the total energy or the energy variance). Here we optimize them within a DMC run. The role of the Jastrow factor within this approach, is different. Its role instead is to correct the trial wave-function $\tilde{\Phi}_{T}({\bf R})$ to match $\Phi_{FN}({\bf R})$. The optimization of the Jastrow parameters with Eq. (38) only ensures that the cost-function Eq. (29) is minimum. Optimization of the Jastrow factor is required to allow the antisymmetric part of the wave- function to move the nodes while the Jastrow factor takes care of the symmetric contribution. However, if the variational freedom of the Jastrow factor or the statistics are limited, the minimization of Eq. (29) does not necessarily imply a minimum in the VMC energy or its variance: the variance of the local energy might rise. In those cases the Jastrow factor must be optimized twice: first when the potential is optimized and second during a VMC variance minimization before a collection DMC run. Finally, (5) note that $\tilde{\Psi}_{T}({\bf R})$ and $\Psi_{T}({\bf R})$ have different Jastrow factors ($\Psi_{T}({\bf R})$ is kept fixed during the cost function optimization steps). ### IV.6 Optimization of multi-determinant wave-functions The multi-determinant expansion obtained in this subsection is different from the one obtained in Section III. In Section III we found a multi-determinant expression of $\Psi_{FN}({\bf R})$ in a given non-interacting orbital basis set for a given fixed Jastrow factor. Here we optimize the Jastrow factor and the non-interacting basis to match $\Psi_{FN}({\bf R})$ within a prescribed small number of determinants. If we restrict the search to pure-state non-interacting $v$-representable nodes, the minimum energy $E_{DMC}$ will be larger than the true ground-state energy $E[\rho({\bf r})]$, because of the upper-bound theorem, unless $S({\bf R})$ is $v$-representable. In DMC the $v$-representability constraint is not required and can be partially removed by including multi-determinants in $\Phi_{T}({\bf R})$ giving more variational freedom to the nodes. Note that if we express $\tilde{\Psi}_{T}({\bf R})$ as a multi-determinant expansion of the form $\tilde{\Psi}_{T}({\bf R})=e^{J({\bf R})}\sum_{k}\alpha_{k}\tilde{\Phi}_{k}({\bf R}),$ (39) an equivalent expression for wave-function optimization can be found. The sum over occupied (unoccupied) levels in Eq. (32) must be extended to every orbital that is occupied (unoccupied) in $\tilde{\Phi}_{k}({\bf R})$. Also, it is easy to prove that the only change in Eq. (36) required is in the values of the $\beta^{n}_{\nu}$ which must be replaced by $\displaystyle\beta^{n}_{\nu}$ $\displaystyle=$ $\displaystyle\int{\bf dR}W({\bf R})e^{\tilde{J}(\bf R)}\sum_{k}\alpha_{k}c_{n}^{{\dagger}}c_{\nu}\tilde{\Phi}_{k}({\bf R})\;,$ (40) where the operators $c_{n}^{{\dagger}}$ and $c_{\nu}$ change, when possible, the single particle state $\phi_{\nu}$ with $\phi_{n}$ in the Slater determinant $\tilde{\Phi}_{k}({\bf R})$; and give zero if $\phi_{\nu}$ is not included or $\phi_{n}$ is already occupied. The function $W({\bf R})$ is still given by Eq. (34). The coefficients $\alpha_{k}$ can be optimized using the following expression $\frac{dK_{DMC}}{d\alpha_{k}}=\int{\bf dR}W({\bf R})e^{\tilde{J}(\bf R)}\tilde{\Phi}_{k}({\bf R})\;.$ (41) ## V Model system tests In this section, to demonstrate the methods described above, we solve a simple yet non-trivial interacting model as a function of the interacting potential strength and shape. We then test a simple version of the algorithm described in Section III. Subsequently, we replace the model interaction by a realistic Coulomb interaction. Finally, in subsection V.4 we optimize the wave-functions by obtaining the effective nodal potential, as described in Section IV. ### V.1 A model interacting ground state For illustrative purposes we choose the same problem studied in Ref. rosetta, where we derived the existence of an effective potential for the wave-function nodes. Briefly, we solve the ground state of two spin-less electrons moving in a two dimensional square of side length $1$ with a repulsive interaction potential of the formunits $V({\bf r},{\bf r^{\prime}})=8\pi^{2}\gamma\cos{[\alpha\pi(x-x^{\prime})]}\cos{[\alpha\pi(y-y^{\prime})]}$. In this paper we show results for $\alpha=1/\pi$ and $\gamma=4$. With this choice of parameters the system is in the highly correlated regime, because the matrix element of the interaction potential between the non-interacting ground state and first excited state is larger than the non-interacting energy difference. We expand the many-body wave-function in a full CI expansion of Slater determinants with the same symmetry as the ground state. The ground state is degenerate because there are only two electrons. We choose one of the ground-state wave-functions according to the $D_{2}$ subgroup of the $D_{4}$ symmetry of the Hamiltonian. For more details see Ref. rosetta, . Figure 2: (Color online) Self-healed DMC run obtained using the method described in Section III. Black points denote the average value of the local energy for each DMC step. Green points mark the reference energy used for population control. Orange lines mark the average energy of the trial wave- function. The horizontal blue line marks the energy of the ground state in the full CI calculation. Vertical lines mark the steps when the coefficients of wave-function are updated. Inset: Detail of the DMC run for the first 10000 steps (same conventions as in the main figure). From the full CI calculation we obtain a nearly exact expression of the ground state $\Psi({\bf R})=\sum_{n}a_{n}\Phi_{n}({\bf R})$. ### V.2 Projection of the DMC fixed-node wave-function on a multi-determinant expansion In order to facilitate the comparison with the full CI results, we sample the mixed-estimator density with the projectors $\xi_{n}({\bf R})$ constructed using the same basis functions of the CI expansion. For the same reason, we utilized no Jastrow function ($J=0$ in Eq. 16). An initial trial wave-function must be selected. While the non-interacting solution has very good nodes,rosetta we intentionally chose a poor initial trial wave-function in order to test the strength of the multi-determinant method described in Section III. The worst case scenario is when the trial wave-function is orthogonal to the exact ground state. If the exact ground state is not included in the trial wave-function, a projector method such as the standard DMC algorithm cannot yield the exact ground-state energy. Accordingly, to test the method, we chose for this example $\lambda_{1}=a_{3}$, $\lambda_{3}=-a_{1}$ and $\lambda_{n}=0$ for all remaining $n$.basisnote Expanding, $\Psi_{FN}({\bf R})$ with these $\lambda_{n}$ and replacing it in Eq. (16) we obtain the projectors $\xi_{n}({\bf R})$. Next we obtained new values $\lambda_{n}$ sampling Eq. (18) every autocorrelation time. After many configurations are sampled, we construct a new trial wave-function with the new $\lambda_{n}$. We only include in the wave-function the coefficients that satisfied the condition $\|\bar{\lambda}_{n}\|>4\frac{\langle\tilde{\sigma}_{n}\rangle}{\sqrt{N_{c}-1}}$, i.e. that the coefficients are well determined according to this empirical threshold. Note that because the multi-determinants are solutions of a non- interacting problem, they will tend to have more nodes as their energy increases. Accordingly, high energy components of the wave-function will have smaller coefficients (${\lambda}_{n}$) in absolute value as compared with the error ($\tilde{\sigma}_{n}$). As a consequence, this acceptance threshold removes the contribution of the high energy components which implies that the resulting wave-function will be smoother than $\Psi_{FN}({\bf R})$ without the kinks at the nodes. This process is the core of a more complex algorithm we propose for larger systems that is explained in Section VI (see steps 3 and 4). The result of this iterative approach is summarized in Figs. 2, 3, 4 and 5. In Fig. 2 we show the average of the local energy $E_{L}$ (black dots) and the best estimator for the energy $E_{best}$ (green dots)umrigar93 as a function of the number of DMC steps. The average energy of the trial wave-function $\bar{E}=\langle\Psi_{T}\|H\|\Psi_{T}\rangle/\langle\Psi_{T}\|\Psi_{T}\rangle$ (orange) is also given for comparison. The run was carried out for a targeted population of $200$ walkers. The exact full CI result is given by the blue line. There is a dramatic decrease of $E_{L}$, $E_{best}$ and $\bar{E}$ as the trial wave-function is updated, and all these values converge to the full CI result. Similar results are obtained with different starting points and interaction strengths. The only limiting factor to reaching the exact CI results appears to be the iteration time. The reduction in the energy variance can be seen in Fig. 2 where the fluctuations in the local energy decrease as the run continues. Figure 3: (Color online) Values of the coefficients of the multi-determinant expansion (small green circles) as compared with a full CI calculation (large black circles). The DMC statistical errors of the coefficients is equal to the radius of the green circles. In Fig. 3 we show a plot of the values of the full CI coefficients as a function of the coefficient index compared with the average values obtained from the optimized trial wave-function and a final DMC run using Eq. (18). The coefficients are ordered with increasing non-interacting energy. The error bars of the coefficient are also given. The figure shows that a wave-function expansion with the quality of a CI expansion can be obtained with DMC. Note that (i) knowledge of the ground-state wave-function allows for the calculation of any other observable with an error bar that can be obtained from the error bars of the expansion coefficients. (ii) The same wave-function could be expressed with a smaller number of determinants if a Jastrow factor had been used. Figure 4: (Color online) Change in the values of the multi-determinant expansion as the DMC self-healing algorithm progresses. Light gray colors denote older coefficients while darker ones denote more converged results. The initial non-zero coefficients are highlighted in red squares. In Fig. 4 we show the evolution of the values of the full CI coefficients as a the algorithm progresses starting from a trial wave-function orthogonal to the ground state. Figure 5: (Color online) Logarithm of the residual projection $R_{P}$ [see Eq. (42)] as a function of the total weighted number of configurations along the complete run $N_{w}$. The lines are a guide to the eye. Inset: projection of the DMC self-healed wave-function onto the full CI ground state as function of the logarithm of $N_{w}$. The improved quality of the DMC optimized trial wave-function is also evident in Fig. 5. We plot the logarithm of the residual projection $R_{P}=\log{[1-\langle\Psi\|\Psi_{T}\rangle/(\|\Psi\|\|\Psi_{T}\|)]}$ (42) on the “exact” CI ground state as a function of the logarithm of the total weighted number of configurations along the complete run $N_{w}$. Remarkably, the error of the wave-function projection has decreased to $e^{-8}$ starting from 1. By noting that $\|\Psi_{T}\rangle=\|\Psi\rangle+\|\delta\Psi_{\perp}\rangle$, where $\|\delta\Psi_{\perp}\rangle$ is the difference between the ground-state $\|\Psi\rangle$ and the trial wave function $\|\Psi_{T}\rangle$ we get $R_{P}\simeq 2\log\left[\|\delta\Psi_{\perp}\|/\sqrt{2}\right]\;.$ (43) We can see that for a significant section of the run $R_{P}\sim 1/N_{w}^{3}$, where $N_{w}$ is the total number of weighted configurations of the run. This means that the magnitude of the error in the trial function decays with a faster exponent than $1/\sqrt{N_{w}}$ ($3/2$). This is surprising because if we had provided the exact ground state as trial wave-function, the error after finite sampling would have scaled as $\|\delta\Psi_{\perp}\|\sim 1/\sqrt{N_{w}}$, which replaced in Eq. (43) gives $R_{P}\simeq 1/\sqrt{N_{w}}$. This faster exponent, in a section of the plot, is a direct consequence of the fact that both the quality of the trial wave-function and the statistics have improved. This is another indication that the nodes continue to improve along the run. For the final part of the graph (the last three points), however, $R_{P}$ scales as $1/\sqrt{N_{w}}$. This possibly signals that after the nodal structure is improved to a critical distance from the exact ground state, the statistical error in the determination of the coefficients and not a small fluctuation in the nodal structure, is the limiting factor for this algorithm. We believe that a final $1/\sqrt{N_{w}}$ scaling of $R_{P}$ signals also that the overall nodal structure of the solution is correct and only small fluctuations of the coefficients are responsible for the small fluctuations from the exact node. Since a direct sampling of the fixed-node wave-function (Eq. (18)) aims to reproduce the fixed-node solution, a single DMC run cannot improve the nodes. Only by iterating with different trial wave-functions can the nodes be improved. In particular, if an infinite number of configurations were used, the nodes would not change. In practice however, we find that for a finite sample, the error in the wave-function coefficients plays a positive role. Errors act as random fluctuations in a simulated annealing algorithm. These fluctuations are reinforcedfn:reinforced or discarded in subsequent iterations. This allows the nodal error to be systematically reduced to the point that trial wave-functions with 0.9995 projections on the full CI ground state can be found starting from a trial wave-function initially orthogonal to the ground state. Since poor nodes are associated with discontinuities in the derivative of $\Psi_{FN}({\bf R})$ at the nodal surface, and consequently an increase in the kinetic energy, it is also convenient at first to initially limit the number of configurations sampled (including first the ones that cost less non-interacting energy). We recognize that the current work does not address the suitability and convergence of this method of relying on random fluctuations for systems with large numbers of electrons; this will be the subject of later studies. ### V.3 Coulomb potential results and discussion The use of a simplified electron-electron interaction facilitates the CI calculations and the validation of the optimization method described in Section III. However, it is also important to test the convergence and stability of the method with a realistic Coulomb interaction. Note that in two dimensions (2D) the correlations are enhanced as compared with three dimensions (3D) while the nodal surface remains non-trivial. We tested the stability of the algorithm by replacing the interaction potential with:units $V({\bf r},{\bf r^{\prime}})=20\pi^{2}/\|{\bf r-r\prime}\|$. Since the length of the square box side is 1, the difference in kinetic energy between the non-interacting ground state and the first-excited state is $3\pi^{2}$. This choice of parameters for the Coulomb potential placed the system in a strongly interacting regime. To further increase the role of correlations and the difficulties that the algorithm must overcome we did not included a Jastrow term, i.e. $J=0$. We also increased the chances of failure by setting the initial trial wave-function equal to the first excited state of the non-interacting system. Figure 6: (Color online) Energy of the DMC run as a function of the number of DMC steps used to gather statistical data of the wave-function in the previous block. The statistical error bars for the first three points on the left were not calculated. The statistical error bars of the points on the right were smaller than the size of the symbols. Blue squares denote calculations starting from a bad trial wave-function, while the red circles mark the results obtained from an initial trial wave-function corresponding to the best blue square on the right (see text). Green diamonds were generated starting from the best red circle. In Fig. 6 we show the evolution of the average of the local energy for each DMC optimization block as a function of the number of DMC steps in each optimization block $N_{DMC}$. Data for Eq. (18) is accumulated every $100$ DMC steps. As in the case of the model Hamiltonian, we increase $N_{DMC}$ in each optimization as $N_{DMC}=200\times 2^{n_{b}/2}$ where $n_{b}$ is the total number of blocks. With this choice we can expect the error bar in the energy and in the coefficient $\lambda_{n}$ of the multi-determinant expansion Eq. (14) to be reduced a factor $1/2$ after four successive blocks. Note that during each DMC run not only the local energy is sampled but also the values of the projectors $\xi({\bf R})$ used to construct the expansion of the trial wave-function of the next point on the right with Eq. (18). The blue squares in Fig. 6 show the progression in average DMC energy starting from the first excited state. The initial energy is above 420 compared with the fully converged energy of 402.718 $\pm$ 0.008. Even starting from such a bad initial trial wave function, our method is able to improve in the second block after only accumulating $\approx 400$ configurations. In contrast, the red circles in Fig. 6 denote the results obtained with an initial trial wave- function constructed with data collected with the right most blue square, a very good initial trial wave-function. As the optimization process is repeated, the average DMC energy fluctuates. Since the coefficients carry a statistical error, the wave-function is not the same from one block to the other and neither is the nodal error. There is a shift from one iteration to the next which is sometimes larger than the error bar in the energy. The energy and the variance can fluctuate and locally increase. However, as the statistics improve, fluctuations in the coefficients decrease. The statistical errors play the role of a thermal noise in the coefficient expansion. Improved statistics correspond to reduced temperatures in simulated annealing. Note that, initially, the average DMC energy from the very poor trial wave-function decreases (blue squares) as the algorithm progresses, while the energy of the average DMC energy from the good trial wave-function (red circles) actually increases. This is because when the statistics are poor the errors in the coefficient expansion allows improvement of a bad trial wave function but spoil a good quality one. Figure 6 shows that, as the algorithm progresses and improved statistics are obtained, the quality of the solution becomes independent the initial trial wave-function. Note that for intermediate blocks the DMC energy becomes flat, signaling that the the statistics are not enough to reduce the nodal error, but are sufficient to stop deterioration of the wave-function. Repeating the algorithm iteratively leads to an incremental improvement in the statistics which results in a clear reduction of the DMC energy beyond the error bar of the preceding calculations. The DMC energy and the energy variance are reduced systematically which is a clear indication of the reduction of nodal errors and improvement in the overall quality of the wave- function. The ground-state energy obtained after 240000 accumulated DMC iterations is 402.718 $\pm$ 0.008. Figure 7: (Color online) Values of the coefficients of the multi-determinant expansion (small green circles) obtained from the DMC run for two electrons in a square box with a Coulomb interaction in the highly correlated limit. The statistical errors in the values of the coefficients are equal to the size of the red bar. In Fig. 7 we show the values of the coefficients of the multi-determinant expansion as obtained with Eq. (18) corresponding to the right-most blue point in Fig. 6. Note that since no Jastrow factor is used and the interaction potential includes a singularity at ${\bf r}={\bf r^{\prime}}$, the number of coefficients with significant value is much larger the model interaction described earlier. The final reduction of nodal errors shown in the final steps of Fig. 6 is associated with subtle variations of the coefficients. If the Jastrow factor is set to one, the density takes a simple form (Eq. (44)) in terms of the single-particle orbitals $\phi_{n}({\bf r})$. Knowledge of this density allows the calculation of the Kohn-Sham potential as explained in Ref. rosetta, (see below) and suggests an alternative route for calculation of forces by applyingpayne92 the Hellmann-Feynman theorem directly to the Kohn-Sham total energy instead of the usual statistical sampling.badinski08 ; filippi00prbr The DMC density can be obtained in terms of the single-particle orbitals with the following equation:QTS $\rho({\bf r})=\sum_{n,\nu}\phi_{n}^{}({\bf r})\phi_{\nu}({\bf r})\sum_{k,l}\lambda_{k}^{}\lambda_{l}\langle\Phi_{k}\|c_{n}^{{\dagger}}c_{\nu}\|\Phi_{l}\rangle\;.$ (44) Note in Eq. (44) that all the matrix elements $\langle\Phi_{k}\|c_{n}^{{\dagger}}c_{\nu}\|\Phi_{l}\rangle$ corresponding to states that differ in more than one electron hole pair, do not contribute to the ground-state density. Figure 8: Density of the ground state of two spin-less electrons with Coulomb interaction in a square box. We choose one of the two degenerate ground states, reducing the symmetry of the density to $D_{2}$. a) Left side of the density of the many-body ground state constructed with the converged coefficients shown in Fig. 7. b) Kohn-Sham non-interacting density constructed as explained in Ref. rosetta, . Figure 9: Kohn-Sham potential for two spin- less electrons in a square box corresponding to the ground state of Figs. 7 and 8. The potential was constructed using the methods explained in Ref. rosetta, . In Fig. 8a we show the density corresponding to the coefficients of Fig. 7 and in Fig. 8b the non-interacting Kohn-Sham density constructed using the methods explained in Ref. rosetta, . In Fig. 9 we show the Kohn-Sham potential obtained using the methods described in Ref. rosetta, . We minimized the cost function in Eq. (2) of Ref. rosetta, using 14 Fourier components in the potential expansion. We believe that the sampled oscillations in the Kohn-Sham potential carry some physical meaning. Indeed, these oscillations are required in order to match the non-interacting density in Fig. 8b to the interacting self-healed DMC density in Fig. 8a. However, since the density $\rho({\bf r})$ has an error $\sigma_{\rho}({\bf r})$, there is also an error in the Kohn-Sham potential. In linear response,rosetta the error bar in the potential $\sigma_{KS}({\bf r})$ (not shown) can be obtained in terms of $\sigma_{\rho}({\bf r^{\prime}})$ and the inverse susceptibility as $\sigma_{KS}({\bf r})=\int{\bf dr^{\prime}}\sigma_{\rho}({\bf r^{\prime}})\frac{\delta V\left({\bf r^{\prime}}\right)}{\delta\rho\left({\bf r}\right)}\;\;\;.$ (45) Since, we have removed degeneracies in the ground state by restricting the symmetry of the wave-function, two potentials that give the same density can only differ by a constant. We have obtained from DMC not only the approximated DMC energy but also the derivative of the total energy with respect to local fluctuations of the density. Figures 8 and 9 show that this method can provide accurate benchmarks for the validation of DFT approximations in the highly correlated regime. ### V.4 Model system effective nodal potential and Jastrow factor To demonstrate that the effective nodal potential and Jastrow factor can be obtained through sampling in DMC, in this section we determine these quantities for a model corresponding to two electrons in a square box with Coulomb interactions. An additional goal is to show that a complex (multi- determinant) wave-function can potentially be replaced by a simpler one while retaining the same nodal structure. The results below correspond to a trial wave-function represented using the multi-determinant expansion shown in Fig. 7. While for larger dimensional systems the integrals can be performed more efficiently using a stochastic approach, in this case the probability densities were binned numerically over a grid of fifteen bins in all four dimensions. Approximately, $7.2\times 10^{5}$ weightedweighnote configurations were collected. The one-body and two-body Jastrow factors were simply written as a Fourier expansion and their coefficients were minimized with an accelerated steepest decent algorithm using Eq. (38). The antisymmetric part of the wave-function was given by a single determinant corresponding to the ground-state solution of a non-interacting effective potential. The effective interactive potential was expressed as a sum of cosine functions and optimized as explained in Ref. rosetta, . The Jastrow factors and the potentials can be optimized at the same time. However, since we wanted the Jastrow factor to carry most of the load in the optimization of the symmetric corrections to the probability density, the potential was optimized only every third iteration that the Jastrow factor was optimized. The resulting potential, and Jastrow factors are shown in Fig. 10. The value of the cost-function was reduced an order of magnitude from starting with the non-interacting ground state with zero effective potential. The effective potential resulting from this minimization procedure is an example of the nodal potential predicted in Ref. rosetta, . We also performed tests of this optimization algorithm using the model interaction discussed in Subsection V.2. In this case the nodal structure of the wave-function was also improved (as signaled by a reduction of the average DMC energy below the error bar of the preceding calculation). Figure 10: a) Effective nodal potential, b) one-body Jastrow, and c) two-body Jastrow factors obtained by minimizing Eq. (29), in which the multi- determinant expansion of Fig. 7 has been replaced by a single determinant function. ## VI Summary of improved self-healing DMC algorithm It is clear from previous sections that an effective wave-function optimization algorithm can be constructed solely on the basis of iteratively updating $\Psi_{T}$ by the multi-determinant expansion of $\Psi_{FN}$. An example of this algorithm applied to a soluble model is presented in Subsection V.2. However, multi-determinant expansions in DMC are computationally very expensive in large or continuum system, since the required number of determinants to reach a given accuracy will in general grow combinatorially. The method developed in Section IV to optimize a single Slater determinant becomes very attractive. (Results of the application of this method were shown in Subsection V.4). For large systems, the number of multi-determinants must be kept to a minimum and the two methods combined. Experimentation in small systems allows us to suggest an algorithm that will be efficient in larger systems: 1. 1. An initial trial trial wave-function $\Phi_{T}({\bf R})$ is generated using any fast method, e.g. an empirical screened pseudopotentialwang95 or a Thomas-Fermi theory. 2. 2. The Jastrow factor $J({\bf R})$ is optimized within VMC. 3. 3. A DMC run is performed. The number of configurations $N_{c}$ sampled is increased as this step is repeated. Statistically uncorrelated values of $\xi_{n}({\bf R})$ and $\xi_{n}({\bf R})^{2}$ are accumulated. 4. 4. The multi-determinant expansion of $f({\bf R})$ is constructed. Only the terms that are significantly non-zero are included in the expansion. 5. 5. A distribution of configurations ${\bf R}_{i}$ with probability $\|W({\bf R})\|$ is generated. The gradients of $K_{DMC}$ with respect to the effective nodal potential and the gradients of the Jastrow factor coefficients are evaluated with Eqs. (36), and (41). (Eventually the multi-determinant expansion coefficients $\alpha_{k}$ can be included, see Subsection IV.6.) 6. 6. The effective potentials $V\left({\bf r}\right)$ and $\tilde{J}({\bf R})$ are updated (eventually also the $\alpha_{k}$). New single particle orbitals are constructed using Eq. (31). Therefore the single particle orbitals used to construct the Slater determinants in the trial wave-function are now determined solely within DMC. 7. 7. A new $\tilde{\Psi}_{T}({\bf R})$ is constructed. Steps (5-7) are repeated until $\tilde{\Psi}_{T}({\bf R})$ does not change. 8. 8. At this step we can choose to improve the scaling in large systems. The single-particle orbitals $\phi_{n}({\bf r})$ shared by all determinants in the expansion $\tilde{\Phi}_{T}({\bf R})$ can be transformed to non-orthogonal localized orbitals.reboredo05 ; alfe04 9. 9. The trial wave-function $\Phi_{T}({\bf R})$ is updated to $\tilde{\Phi}_{T}({\bf R})$. Steps (2-9) are repeated until $\tilde{\Psi}_{T}({\bf R})$ and $E_{DMC}$ do not change.fn:change Note that (i) the methods in Sections III and IV are complementary. In Section III, we find a representation of the fixed-node ground state in a given basis. In Section IV, instead, we optimize and change the basis of the wave-functions so as to reproduce the fixed-node ground-state wave-function with a minimum number of Slater determinants. (ii) Only single configurations are included in Eq. (36) but multiple configurations are included in Eq. (14). (iii) We include a Jastrow function in Eq. (14) to minimize the number of Slater determinants required in the expansion. However, a final run with no Jastrow factor included with the configuration interaction expansion might be useful in order to obtain a pure expression of the ground-state density in terms of the single particle orbitals. Atomic forces could be obtained from this density. Finally (iv) the method is, in principle, self reliant: no DFT or HF are required. ## VII Summary We have presented an algorithm for sampling the fixed-node many-body wave- function in a single or multi-determinant expansion from a diffusion quantum Monte Carlo (DMC) calculation within the importance sampling technique. By combining this algorithm with a previously developed method for constructing effective potentials targeted at reproducing specific properties of the many- body wave-function,rosetta we presented an iterative algorithm that improves the nodes of the trial/fixed-node wave-functions used in DMC. Tests on a simple two electron model system confirm that this method is able to improve the nodes and that, at least in the case of the tested system, we find wave- functions and energies that exactly match fully converged configuration interaction calculations. We have proven that the nodes of the fixed-node wave-function improve as compared with the trial wave-function if the kinks at the nodes are locally smoothed out. The algorithms presented take advantage of this proof. We have argued that if the kink at the node increases with the “distance” from the exact ground-state node to the trial wave-function node, the algorithm should be stable against random statistical fluctuations. Proving this property in general might be difficult and is beyond the scope of this article. Clearly, in the absence of a proof, experimentation in larger systems is required. While in the past, methods were used to obtain the fixed-node wave-function (e.g Ref. bianchi96, ), to our knowledge this is the first time the fixed-node wave-function has been obtained through importance sampling. The availability of the fixed-node wave-function provides routes to determine the exact Kohn- Sham potential, allowing benchmark tests of density functionals in highly non- trivial and inhomogeneous systems. It also seems likely that many of the wave function optimization approaches (e.g. Refs. filippi00, ; umrigar07, ; rios06, ; luchow07, ) currently applied within variational Monte Carlo can be recast in the present scheme, making direct use of the fixed-node wave-function, and likely obtaining improved results. In ongoing work, we are continuing to develop these methods. Applications to larger and more complex electronic systems will be reported elsewhere. ## Acknowledgments Research performed at the Materials Science and Technology Division and the Center of Nanophase Material Sciences at Oak Ridge National Laboratory sponsored the Division of Materials Sciences and the Division of Scientific User Facilities U.S. Department of Energy. This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. The authors would like thank J. Kim for discussions and C. Umrigar for clarifications related to the use of Eq. (19). ## References * (1) J. B. Anderson, Int. J. of Quantum Chem. 15, 109 (1979). * (2) D. M. Ceperley and B. J. Alder, Phys. Rev. Lett. 45, 566 (1980). * (3) The energy of any wave-function $\Phi$ being $\langle\Phi\|\hat{\mathcal{H}}\|\Phi\rangle/\langle\Phi\|\Phi\rangle$ for a given Hamiltonian $\hat{\mathcal{H}}$. * (4) P. J. Reynolds, D. M. Ceperley, B. J. Alder, and W. A. Lester, J. Chem. Phys. 77, 5593 (1982). * (5) M. Bajdich, L. Mitas, G. Drobny, and L. K. Wagner, Phys. Rev. B 72, 075131 (2005); L. Mitas, Phys. Rev. Lett. 96, 240402 (2006). * (6) C. Filippi and S. Fahy, J. Chem. Phys. 112, 3523 (2000). * (7) C. J. Umrigar, J. Toulouse, C. Filippi, S. Sorella, and R. G. Hennig, Phys. Rev. Lett. 98, 110201 (2007). * (8) P. Lòpez-Rios, A. Ma, N. D. Drummond, M. D. Towler, and R. J. Needs, Phys. Rev. E 74, 066701 (2006). * (9) A. Lüchow, et. al., J. Chem. Phys. 126, 144110 (2007). * (10) M. H. Kalos and F. Pederiva, Phys Rev. Lett. 85, 3547 (2000). * (11) D. M. Ceperley and B. J. Alder, J. Chem. Phys. 81, 5833 (1984). * (12) S. W. Zhang and M. H. Kalos, Phys. Rev. Lett. 67, 3074 (1991). * (13) T. D. Beaudet, M. Casula, J. Kim, S. Sorella, and R. M. Martin, J. Chem. Phys. 129, 164711 (2008). * (14) J. Toulouse and C. J. Umrigar, J. Chem. Phys. 128, 174101 (2008). * (15) A. J. Williamson, R. Q. Hood, and J. C. Grossman, Phys. Rev. Lett. 87, 246406 (2001). * (16) F. A. Reboredo and A. J. Williamson, Phys. Rev. B 71, 121105(R) (2005). * (17) D. Alfe and M. J. Gillan, Phys. Rev. B 70, 161101(R) (2004). * (18) F. A. Reboredo and P. R. C. Kent, Phys. Rev. B. 77, 245110 (2008). * (19) M. L. Tiago, P. R. C. Kent, R. Q. Hood, and F. A. Reboredo, J. Chem. Phys. 129, 084311 (2008). * (20) B. L. Hammond, W. A. Lester, Jr., and P. J. Reynolds Monte Carlo Methods in Ab Inition Quantum Chemistry (World Scientific, Singapore-New Jersey-London-Hong Kong, 1994). * (21) S. Vitiello, K. Runge, and M. H. Kalos, Phys. Rev. Lett. 60, 1970 (1988). * (22) Equation (II) can be thought of as a one-step ($M=1$ in Eq. (3)) in a pure-diffusion DMC run using the small $\tau$ propagator of Eq. (4), in which the positive and negative walkers have the initial distribution $\Psi_{FN}({\bf R^{\prime}})$ and are not constrained by the nodes of $\Psi_{T}({\bf R})$. * (23) Because of elemental group theory theorems, i) the product of two anti-symmetric functions is symmetric, ii) the product of any symmetric operator with any anti-symmetric function results in an anti-symmetric function (with no projection on any symmetric bosonic function or any other representation of the symmetric group). Thus our analytic derivation is true. Since we use an anti-symmetric representation of the wave-function, our numerical method will not find bosonic states since they are excluded from the search.bianchi93 * (24) The nodes continue to move statistically in the same direction in successive iterations. * (25) The “release-node” method ceperley80 ; HLRbook involves two fixed trial wave-functions and two populations of walkers. Our method is very different since, instead, involves only one trial wave-function that is updated and a single population of walkers. In our method the walkers never cross the node: we move the node instead. In the so called release-node method the node is removed. * (26) R. Bianchi, D. Bressanini, P. Cremaschi, M. Mella, and G. Morosi, J. Chem. Phys. 98, 7204 (1993). * (27) R. Bianchi, D. Bressanini, P. Cremaschi, M. Mella, and G. Morosi, Int. J. of. Quant. Chem. 57, 321 (1996). * (28) F. A. Reboredo and C. R. Proetto, Phys. Rev. B 67, 115325 (2003). * (29) In Ref. us, the projectors $\xi_{n}({\bf R})$ are denoted as $\psi_{i}^{\sigma}(z)$. * (30) C. J. Umrigar, M. P. Nightingale, and K. J. Runge, J. Chem. Phys. 99, 2865 (1993). * (31) B. Wood and W. M. C. Foulkes, J. of Phys. C 18, 2305 (2006). * (32) A. A. Correa, F. A. Reboredo, and C. A. Balseiro, Phys. Rev. B 71, 035418 (2005). * (33) P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964). * (34) R. G. Parr and W. Yang, Density-Functional Theory of Atoms and Molecules (Oxford Science Publications, Oxford, 1989). * (35) W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 (1965). * (36) In section V.4 we solve a case where it is more efficient to bin $f({\bf R})$. * (37) We define the energy unit to be $\hbar^{2}/(2m)$. * (38) The basis functions in the CI expansion are ordered with increasing non-interacting energy. * (39) M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, and J.D. Joannopoulos Rev. Mod. Phys. 64, 1045 (1992). * (40) A. Badinski and R. J. Needs, Phys. Rev. B 78, 035134 (2008). * (41) C. Filippi and C. J. Umrigar, Phys. Rev. B 61, R16291 (2000). * (42) C. Kittel Quantum Theory of Solids (John Wiley & Sons, New York, 1987). * (43) The weight of the walkers goes from 2 to 1/2 and is multiplied by the correction given in Eq. (19). * (44) L.-W. Wang and A. Zunger, Phys. Rev. B 51, 17398 (1995). * (45) $\tilde{\Psi}$ must be converged twice: first in order to fit the fixed-node solution which allows the node to move, and second when the node no longer moves.	arxiv-papers	2008-08-26T01:45:21	2024-09-04T02:48:57.449274	{ "license": "Public Domain", "authors": "Fernando A. Reboredo, Randolph Q. Hood, Paul R. C. Kent", "submitter": "Fernando Reboredo", "url": "https://arxiv.org/abs/0808.3417" }
0808.3438	# A mathematical proof that the transition to a superconducting state is a second-order phase transition Shuji Watanabe Division of Mathematical Sciences Graduate School of Engineering, Gunma University 4-2 Aramaki-machi, Maebashi 371-8510, Japan Email: watanabe@fs.aramaki.gunma-u.ac.jp ###### Abstract We deal with the gap function and the thermodynamical potential in the BCS- Bogoliubov theory of superconductivity, where the gap function is a function of the temperature $T$ only. We show that the squared gap function is of class $C^{2}$ on the closed interval $[\,0,\,T_{c}\,]$ and point out some more properties of the gap function. Here, $T_{c}$ stands for the transition temperature. On the basis of this study we then give, examining the thermodynamical potential, a mathematical proof that the transition to a superconducting state is a second-order phase transition. Furthermore, we obtain a new and more precise form of the gap in the specific heat at constant volume from a mathematical point of view. Mathematics Subject Classification (2000): 45G10, 82D55 Keywords: Second-order phase transition, superconductivity, gap function, thermodynamical potential ## 1 Introduction Let $\varepsilon>0$ be small enough and let us fix it unless otherwise stated. Let $k_{B}>0$ and $\omega_{D}>0$ stand for the Boltzmann constant and for the Debye frequency, respectively. We denote Planck’s constant by $h\;(>0)$ and set $\hslash=h/(2\pi)$. Let $\mu>0$ stand for the chemical potential. Let $N(\xi)\geq 0$ stand for the density of states per unit energy at the energy $\xi$ $(-\mu\leq\xi<\infty)$ and let $N_{0}=N(0)>0$. Here, $N_{0}$ stands for the density of states per unit energy at the Fermi surface $(\xi=0)$. Let $U_{0}>0$ be a constant. It is well known that superconductivity occurs at temperatures below the temperature $T_{c}>0$ called the transition temperature. We now define it. ###### Definition 1.1. The transition temperature is the temperature $T_{c}>0$ satisfying $\frac{1}{\,U_{0}N_{0}\,}=\int_{\displaystyle{\varepsilon}}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}\frac{\,\tanh\eta\,}{\eta}\,d\eta\,.$ Generally speaking, the gap function is a function both of the temperature $T$ and of wave vector. In this paper we however regard the gap function as a function of the temperature $T$ only, and denoted it by $\Delta(T)$ $(\geq 0)$. Such a situation is considered in the BCS-Bogoliubov theory [1, 3], and is accepted widely in condensed matter physics (see e.g. [6, (7.118), p. 250], [12, (11.45), p. 392]). See also [9] and [10] for related material. The gap function satisfies the following nonlinear integral equation called the gap equation (c.f. [1]): For $0\leq T\leq T_{c}$, (1.1) $1=U_{0}N_{0}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\frac{1}{\,\sqrt{\,\xi^{2}+f(T)\,}\,}\tanh\frac{\,\sqrt{\,\xi^{2}+f(T)\,}\,}{2k_{B}T}\,d\xi.$ Here, for later convenience, the squared gap function is denoted by $f$, i.e., $f(T)=\Delta(T)^{2}$. ###### Remark 1.2. We introduce the cutoff $\varepsilon$ in Definition 1.1 and in the gap equation (1.1). When $\varepsilon=0$, Definition 1.1 and the gap equation (1.1) reduce to those in the BCS-Bogoliubov theory [1, 3]. Furthermore, when $\varepsilon=0$, the thermodynamical potential $\Omega$ in Definition 1.4 below reduces to that in the BCS-Bogoliubov theory (see also (1) and (1) below). See e.g. Niwa [6, sec. 7.7.3, p. 255]. The gap equation (1.1) is a simplified one, and the gap equation with a more general potential is studied extensively. Odeh [7] and Billard and Fano [2] established the existence and uniqueness of the positive solution to the gap equation with a more general potential in the case $T=0$. In the case $T\geq 0$, Vansevenant [8] and Yang [11] determined the transition temperature and showed that there is a unique positive solution to the gap equation with a more general potential. Recently Hainzl, Hamza, Seiringer and Solovej [4], and Hainzl and Seiringer [5] proved that the existence of a positive solution to the gap equation with a more general potential is equivalent to the existence of a negative eigenvalue of a certain linear operator to show the existence of a transition temperature. Let $f(T)$ be as in (1.1) and set $\displaystyle\quad\qquad\Omega_{S}(T)$ $\displaystyle=$ $\displaystyle\Omega_{N}(T)+\delta(T),$ $\displaystyle\quad\qquad\Omega_{N}(T)$ $\displaystyle=$ $\displaystyle-2N_{0}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\xi\,d\xi-4N_{0}k_{B}T\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\ln\left(1+e^{\displaystyle{-\xi/(k_{B}T)}}\right)\,d\xi$ $\displaystyle+V(T),\qquad T>0,$ $\displaystyle\delta(T)$ $\displaystyle=$ $\displaystyle\frac{\,f(T)\,}{U_{0}}-2N_{0}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\left\\{\sqrt{\xi^{2}+f(T)}-\xi\right\\}\,d\xi$ $\displaystyle-4N_{0}k_{B}T\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\ln\frac{1+e^{-\displaystyle{\sqrt{\xi^{2}+f(T)}/(k_{B}T)}}}{1+e^{-\displaystyle{\xi/(k_{B}T)}}}\,d\xi,\quad 0<T\leq T_{c},$ $\displaystyle\,\quad V(T)$ $\displaystyle=$ $\displaystyle 2\int_{\displaystyle{-\mu}}^{\displaystyle{-\hslash\omega_{D}}}\xi\,N(\xi)\,d\xi-2k_{B}T\int_{\displaystyle{-\mu}}^{\displaystyle{-\hslash\omega_{D}}}N(\xi)\ln\left(1+e^{\displaystyle{\,\xi/(k_{B}T)}}\right)\,d\xi$ $\displaystyle-2k_{B}T\int_{\displaystyle{\hslash\omega_{D}}}^{\infty}N(\xi)\ln\left(1+e^{\displaystyle{-\xi/(k_{B}T)}}\right)\,d\xi,\qquad T>0.$ ###### Remark 1.3. Since $N(\xi)=O(\sqrt{\xi})$ as $\xi\to\infty$, the integral on the right side of (1) $\int_{\displaystyle{\hslash\omega_{D}}}^{\infty}N(\xi)\ln\left(1+e^{\displaystyle{-\xi/(k_{B}T)}}\right)\,d\xi$ is well defined for $T>0$. ###### Definition 1.4. Let $\Omega_{S}(T)$ and $\Omega_{N}(T)$ be as above. The thermodynamical potential $\Omega$ is defined by $\Omega(T)=\left\\{\begin{array}[]{ll}\displaystyle{\Omega_{S}(T)}&(0<T\leq T_{c}),\\\ \vskip 8.5359pt\cr\displaystyle{\Omega_{N}(T)}&(T>T_{c}).\end{array}\right.$ ###### Remark 1.5. Generally speaking, the thermodynamical potential $\Omega$ is a function of the temperature $T$, the chemical potential $\mu$ and the volume of our physical system. Fixing the values of $\mu$ and of the volume of our physical system, we deal with the dependence of $\Omega$ on the temperature $T$ only. ###### Remark 1.6. Hainzl, Hamza, Seiringer and Solovej [4], and Hainzl and Seiringer [5] studied the gap equation with a more general potential examining the thermodynamic pressure. ###### Definition 1.7. We say that the transition to a superconducting state at the transition temperature $T_{c}$ is a second-order phase transition if the following conditions are fulfilled: (a) The thermodynamical potential $\Omega$, regarded as a function of $T$, is of class $C^{1}$ on $(0,\,\infty)$. (b) The second-order derivative $\left(\partial^{2}\Omega/\partial T^{2}\right)$ is continuous on $(0,\,\infty)\setminus\\{T_{c}\\}$ and is discontinuous at $T=T_{c}$. ###### Remark 1.8. Condition (a) implies that the entropy $\displaystyle{S=-\left(\partial\Omega/\partial T\right)}$ is continuous on $(0,\,\infty)$ and that, as a result, no latent heat is observed at $T=T_{c}$. On the other hand, (b) implies that the specific heat at constant volume, $\displaystyle{C_{V}=-T\left(\partial^{2}\Omega/\partial T^{2}\right)}$, is discontinuous at $T=T_{c}$. See Proposition 2.4 below, which gives a new and more precise form of the gap $\Delta C_{V}$ in the specific heat at constant volume at $T=T_{c}$ from a mathematical point of view. From a physical point of view, it is pointed out that the transition from a normal state to a superconducting state is a second-order phase transition. But a mathematical proof of this statement has not been given yet as far as we know. In this paper we first show that there is a unique solution: $T\mapsto f(T)$ of class $C^{2}$ on the closed interval $[\,0,\,T_{c}\,]$ to the gap equation (1.1) and point out some more properties of the gap function. Examining the thermodynamical potential $\Omega$, we then give a mathematical proof that the transition to a superconducting state at the transition temperature $T_{c}$ is a second-order phase transition. Furthermore, we obtain a new and more precise form of the gap in the specific heat at constant volume from a mathematical point of view. The paper proceeds as follows. In section 2 we state our main results without proof. In sections 3 and 4 we study some properties of the function $F$ defined by (2.1) below. On the basis of this study, in sections 5 and 6, we prove our main results in a sequence of lemmas. ## 2 Main results Let $h(T,\,Y,\,\xi)=\left\\{\begin{array}[]{ll}\displaystyle{\frac{1}{\,\sqrt{\,\xi^{2}+Y\,}\,}\tanh\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}}&(0<T\leq T_{c}\,,\quad Y\geq 0),\\\ \vskip 8.5359pt\cr\displaystyle{\frac{1}{\,\sqrt{\,\xi^{2}+Y\,}\,}}&(T=0,\quad Y>0)\end{array}\right.$ and set (2.1) $F(T,\,Y)=\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}h(T,\,Y,\,\xi)\,d\xi-\frac{1}{\,U_{0}N_{0}\,}\,.$ Set also (2.2) $\Delta_{0}=\frac{\hslash\omega_{D}}{\,\sinh\frac{1}{\,U_{0}N_{0}\,}\,},\quad\Delta=\frac{\,\sqrt{\left\\{\hslash\omega_{D}-2k_{B}T_{c}\,\varepsilon\,e^{1/(U_{0}N_{0})}\right\\}\left\\{\hslash\omega_{D}-2k_{B}T_{c}\,\varepsilon\,e^{-1/(U_{0}N_{0})}\right\\}}\,}{\,\sinh\frac{1}{\,U_{0}N_{0}\,}\,}.$ Since $\varepsilon>0$ is small enough, it follows that $\Delta_{0}>\Delta$. We consider the function $F$ on the following domain $W\subset\mathbb{R}^{2}$: $W=W_{1}\cup W_{2}\cup W_{3}\cup W_{4}\,,$ where $\displaystyle W_{1}$ $\displaystyle=$ $\displaystyle\left\\{(T,\,Y)\in\mathbb{R}^{2}:\;0<T<T_{c}\,,\;0<Y<2\,\Delta_{0}^{2}\right\\},$ $\displaystyle W_{2}$ $\displaystyle=$ $\displaystyle\left\\{(0,\,Y)\in\mathbb{R}^{2}:\;0<Y<2\,\Delta_{0}^{2}\right\\},$ $\displaystyle W_{3}$ $\displaystyle=$ $\displaystyle\left\\{(T,\,0)\in\mathbb{R}^{2}:\;0<T\leq T_{c}\right\\},$ $\displaystyle W_{4}$ $\displaystyle=$ $\displaystyle\left\\{(T_{c}\,,\,Y)\in\mathbb{R}^{2}:\;0<Y<2\,\Delta_{0}^{2}\right\\}.$ ###### Remark 2.1. The gap equation (1.1) is rewritten as $F(T,\,Y)=0$, where $Y$ corresponds to $f(T)$ $\left(=\Delta(T)^{2}\right)$. Let $g$ be given by (2.3) $g(\eta)=\left\\{\begin{array}[]{ll}\displaystyle{\frac{1}{\,\eta^{2}\,}\left(\frac{1}{\,\cosh^{2}\eta\,}-\frac{\,\tanh\eta\,}{\eta}\right)}&(\eta>0),\\\ \vskip 8.5359pt\cr\displaystyle{-\frac{\,2\,}{\,3\,}}&(\eta=0).\end{array}\right.$ Note that $g(\eta)<0$. Our main results are the following. ###### Proposition 2.2. Let $F$ be as in (2.1) and $\Delta$ as in (2.2). Then there is a unique solution: $T\mapsto Y=f(T)$ of class $C^{2}$ on the closed interval $[\,0,\,T_{c}\,]$ to the gap equation $F(T,\,Y)=0$ such that the function $f$ satisfies $f(0)=\Delta^{2}$ and $f(T_{c})=0$, and is monotonically decreasing on $[0,\,T_{c}]$: $f(0)=\Delta^{2}>f(T_{1})>f(T_{2})>f(T_{c})=0,\qquad 0<T_{1}<T_{2}<T_{c}\,.$ Furthermore the value of the derivative $f^{\prime}$ at $T=T_{c}$ is given by $f^{\prime}(T_{c})=8\,k_{B}^{2}T_{c}\,\frac{\,\displaystyle{\int_{\displaystyle{\varepsilon}}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}\frac{d\eta}{\,\cosh^{2}\eta\,}}\,}{\,\displaystyle{\int_{\displaystyle{\varepsilon}}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}g(\eta)\,d\eta}\,}<0\,.$ ###### Theorem 2.3. The transition to a superconducting state at the transition temperature $T_{c}$ is a second-order phase transition, and the following relation holds at the transition temperature $T_{c}$: $\lim_{T\uparrow T_{c}}\frac{\,\partial^{2}\Omega\,}{\,\partial T^{2}\,}(T)-\lim_{T\downarrow T_{c}}\frac{\,\partial^{2}\Omega\,}{\,\partial T^{2}\,}(T)=\frac{\,2N_{0}f^{\prime}(T_{c})\,}{T_{c}}\left(\frac{1}{\,1+e^{\displaystyle{2\varepsilon}}\,}-\frac{1}{\,1+e^{\displaystyle{\hslash\omega_{D}/(k_{B}T_{c})}}\,}\right),$ where $f^{\prime}(T_{c})$ is given by Proposition 2.2. Setting $\varepsilon=0$ in the results of Proposition 2.2 and Theorem 2.3 immediately yields the following. ###### Proposition 2.4. Let $T_{c}$ satisfy $\frac{1}{\,U_{0}N_{0}\,}=\int_{0}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}\frac{\,\tanh\eta\,}{\eta}\,d\eta$ and let $f^{\prime}(T_{c})$ be given by $f^{\prime}(T_{c})=8\,k_{B}^{2}T_{c}\,\frac{\,\displaystyle{\int_{0}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}\frac{d\eta}{\,\cosh^{2}\eta\,}}\,}{\,\displaystyle{\int_{0}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}g(\eta)\,d\eta}\,}<0\,.$ Then the gap $\Delta C_{V}$ in the specific heat at constant volume, $\displaystyle{C_{V}=-T\left(\partial^{2}\Omega/\partial T^{2}\right)}$, at the transition temperature $T_{c}$ is given by the form (2.4) $\Delta C_{V}=-N_{0}f^{\prime}(T_{c})\tanh\frac{\hslash\omega_{D}}{\,2k_{B}T_{c}\,}>0.$ ###### Remark 2.5. A form similar to (2.4) has already been obtained by a different method in the context of theoretical condensed matter physics, but it is an approximate one. However the form (2.4) is a more precise one obtained in the context of mathematics. ## 3 The first-order partial derivatives of the function $F$ In this section we deal with the first-order partial derivatives of the function $F$ and show that $F$ is of class $C^{1}$ on $W$. A straightforward calculation gives the following. ###### Lemma 3.1. Let $g$ be as in (2.3). Then the function $g$ is of class $C^{1}$ on $[0,\,\infty)$ and satisfies $g(\eta)<0,\qquad g^{\prime}(0)=0,\qquad\lim_{\eta\to\infty}g(\eta)=\lim_{\eta\to\infty}g^{\prime}(\eta)=0.$ ###### Lemma 3.2. The partial derivatives $\displaystyle{\frac{\,\partial F\,}{\,\partial T\,}}$ and $\displaystyle{\frac{\,\partial F\,}{\,\partial Y\,}}$ exist on $W$, and are given as follows. At $(T,\,Y)\in W\setminus W_{2}$ , $\left\\{\begin{array}[]{ll}\displaystyle{\frac{\,\partial F\,}{\,\partial T\,}(T,\,Y)=-\frac{1}{\,2k_{B}T^{2}\,}\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\frac{d\xi}{\,\cosh^{2}\displaystyle{\frac{\,\sqrt{\xi^{2}+Y}\,}{\,2k_{B}T\,}}\,}\,,}&\\\ \vskip 8.5359pt\cr\displaystyle{\frac{\,\partial F\,}{\,\partial Y\,}(T,\,Y)=\frac{1}{\,2(2k_{B}T)^{3}\,}\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}g\left(\frac{\,\sqrt{\xi^{2}+Y}\,}{\,2k_{B}T\,}\right)\,d\xi}&\end{array}\right.$ and at $(0,\,Y)\in W_{2}$ , $\left\\{\begin{array}[]{ll}\displaystyle{\frac{\,\partial F\,}{\,\partial T\,}(0,\,Y)=0,}&\\\ \vskip 8.5359pt\cr\displaystyle{\frac{\,\partial F\,}{\,\partial Y\,}(0,\,Y)=-\frac{1}{\,2\,}\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\frac{d\xi}{\,(\sqrt{\xi^{2}+Y})^{3}\,}\,.}&\end{array}\right.$ Lemmas 3.1 and 3.2 immediately give the following. ###### Lemma 3.3. At $(T,\,Y)\in W\setminus W_{2}$, $\frac{\,\partial F\,}{\,\partial T\,}(T,\,Y)<0,\qquad\frac{\,\partial F\,}{\,\partial Y\,}(T,\,Y)<0.$ We now study the continuity of the functions $F$, $(\partial F/\partial T)$ and $(\partial F/\partial Y)$ on $W$. ###### Lemma 3.4. The partial derivatives $\displaystyle{\frac{\,\partial F\,}{\,\partial T\,}}$ and $\displaystyle{\frac{\,\partial F\,}{\,\partial Y\,}}$ are continuous on $W_{1}$. Consequently, the function $F$ is of class $C^{1}$ on $W_{1}$. ###### Proof. It is enough to show that the functions: $(T,\,Y)\mapsto I_{1}(T,\,Y)$ and $(T,\,Y)\mapsto I_{2}(T,\,Y)$ (see Lemma 3.2) are continuous at $(T_{0},\,Y_{0})\in W_{1}$. Here, (3.1) $I_{1}(T,\,Y)=\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\frac{d\xi}{\,\cosh^{2}\eta\,}\,,\quad I_{2}(T,\,Y)=\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}g(\eta)\,d\xi\,,\quad\eta=\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\,.$ Set $\displaystyle{\eta_{0}=\frac{\,\sqrt{\,\xi^{2}+Y_{0}\,}\,}{2k_{B}T_{0}}}$. Since $(T,\,Y)\in W_{1}$ is close to $(T_{0},\,Y_{0})\in W_{1}$, it follows that $T>T_{0}/2$. Then $\displaystyle\left\|I_{1}(T,\,Y)-I_{1}(T_{0},\,Y_{0})\right\|$ $\displaystyle\leq$ $\displaystyle\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\left\|\left(\frac{1}{\,\cosh\eta\,}+\frac{1}{\,\cosh\eta_{0}\,}\right)\frac{\,\cosh\eta-\cosh\eta_{0}\,}{\,\cosh\eta\,\cosh\eta_{0}\,}\right\|\,d\xi$ $\displaystyle\leq$ $\displaystyle 2\hslash\omega_{D}\sinh\frac{\,\sqrt{\,\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}\,}\,}{k_{B}T_{0}}\left(\frac{\,\sqrt{\,\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}\,}\,}{k_{B}T_{0}^{2}}\|T-T_{0}\|+\frac{\|Y-Y_{0}\|}{\,k_{B}T_{0}\sqrt{Y_{0}}\,}\right),$ $\displaystyle\left\|I_{2}(T,\,Y)-I_{2}(T_{0},\,Y_{0})\right\|$ $\displaystyle\leq$ $\displaystyle\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\left\|g(\eta)-g(\eta_{0})\right\|\,d\xi$ $\displaystyle\leq$ $\displaystyle\hslash\omega_{D}\,\displaystyle{\max_{\eta\geq 0}\|g^{\prime}(\eta)\|}\left(\frac{\,\sqrt{\,\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}\,}\,}{k_{B}T_{0}^{2}}\|T-T_{0}\|+\frac{\|Y-Y_{0}\|}{\,k_{B}T_{0}\sqrt{Y_{0}}\,}\right).$ Thus the functions: $(T,\,Y)\mapsto I_{1}(T,\,Y)$ and $(T,\,Y)\mapsto I_{2}(T,\,Y)$, and hence $(\partial F/\partial T)$ and $(\partial F/\partial Y)$ are continuous at $(T_{0},\,Y_{0})\in W_{1}$. ∎ ###### Lemma 3.5. The function $F$ is continuous on $W$. ###### Proof. Note that $F$ is continuous on $W_{1}$ by Lemma 3.4. We then show that $F$ is continuous on $W_{2}$. Let $(0,\,Y_{0})\in W_{2}$ and let $(T,\,Y)\in W_{1}\cup W_{2}$. Since $(T,\,Y)$ is close to $(0,\,Y_{0})$, it follows that $Y>Y_{0}/2$. Then, by (2.1), $\displaystyle\left\|F(T,\,Y)-F(0,\,Y_{0})\right\|$ $\displaystyle\leq$ $\displaystyle\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\left\\{\frac{\,1-\tanh\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\,}{\sqrt{\,\xi^{2}+Y_{0}\,}}+\left\|\frac{1}{\,\sqrt{\,\xi^{2}+Y\,}\,}-\frac{1}{\,\sqrt{\,\xi^{2}+Y_{0}\,}\,}\right\|\right\\}\,d\xi$ $\displaystyle\leq$ $\displaystyle\hslash\omega_{D}\left\\{\frac{1}{\,\sqrt{Y_{0}}\,}\left(1-\tanh\frac{\,\sqrt{\,Y_{0}/2\,}\,}{2k_{B}T}\right)+\frac{\,2\,\|Y-Y_{0}\|\,}{\,(\sqrt{2}+1)Y_{0}^{3/2}\,}\right\\}.$ Thus $F$ is continuous on $W_{2}$. Similarly we can show the continuity of $F$ on $W_{3}$, and on $W_{4}$. ∎ ###### Lemma 3.6. The partial derivatives $\displaystyle{\frac{\,\partial F\,}{\,\partial T\,}}$ and $\displaystyle{\frac{\,\partial F\,}{\,\partial Y\,}}$ are continuous on $W$. Consequently, the function $F$ is of class $C^{1}$ on $W$. ###### Proof. Note that $(\partial F/\partial T)$ and $(\partial F/\partial Y)$ are continuous on $W_{1}$ by Lemma 3.4. We then show that $(\partial F/\partial T)$ and $(\partial F/\partial Y)$ are continuous at $(T_{c}\,,\,0)\in W_{3}$. We can show the continuity of those functions at other points in $W$ similarly. Step 1. Let $(T,\,Y)\in W_{1}$. We show $\frac{\,\partial F\,}{\,\partial T\,}(T,\,Y)\to\frac{\,\partial F\,}{\,\partial T\,}(T_{c}\,,\,0),\;\;\frac{\,\partial F\,}{\,\partial Y\,}(T,\,Y)\to\frac{\,\partial F\,}{\,\partial Y\,}(T_{c}\,,\,0)\quad\mbox{as}\;(T,\,Y)\to(T_{c}\,,\,0).$ Since $(T,\,Y)$ is close to $(T_{c}\,,\,0)$, it then follows that $T_{c}/2<T<T_{c}$. Set $\eta_{0}=\frac{\,\sqrt{\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}}\,}{k_{B}T_{c}}$. Then $\displaystyle\left\|\frac{1}{\,T^{2}\cosh^{2}\frac{\,\sqrt{\xi^{2}+Y}\,}{\,2k_{B}T\,}\,}-\frac{1}{\,T_{c}^{2}\cosh^{2}\frac{\,\xi\,}{\,2k_{B}T_{c}\,}\,}\right\|$ $\displaystyle\leq$ $\displaystyle\frac{\,8\cosh\eta_{0}\,}{T_{c}^{3}}\left\\{\left\|T-T_{c}\right\|\left(\cosh\eta_{0}+\eta_{0}\sinh\eta_{0}\right)+\frac{\,\sqrt{Y}\,}{4k_{B}}\sinh\eta_{0}\right\\},$ and hence $\displaystyle{(\partial F/\partial T)(T,\,Y)-(\partial F/\partial T)(T_{c}\,,\,0)\to 0\quad\mbox{as}\quad(T,\,Y)\to(T_{c}\,,\,0)}$. Since $\left\|g\left(\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\right)-g\left(\frac{\,\xi\,}{\,2k_{B}T_{c}\,}\right)\right\|\leq\max_{\eta\geq 0}\|g^{\prime}(\eta)\|\left(\frac{\,\hslash\omega_{D}\left\|T-T_{c}\right\|\,}{k_{B}T_{c}^{2}}+\frac{\,\sqrt{Y}\,}{\,k_{B}T_{c}\,}\right),$ it follows that $\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\left\\{g\left(\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\right)-g\left(\frac{\,\xi\,}{\,2k_{B}T_{c}\,}\right)\right\\}\,d\xi\to 0\quad\mbox{as}\quad(T,\,Y)\to(T_{c}\,,\,0),$ and hence $\displaystyle{(\partial F/\partial Y)(T,\,Y)-(\partial F/\partial Y)(T_{c}\,,\,0)\to 0\quad\mbox{as}\quad(T,\,Y)\to(T_{c}\,,\,0)}$. Step 2. When $(T,\,Y)=(T,\,0)\in W_{3}$ and $(T,\,Y)=(T_{c}\,,\,Y)\in W_{4}$, an argument similar to that in Step 1 gives $\frac{\,\partial F\,}{\,\partial T\,}(T,\,0)\to\frac{\,\partial F\,}{\,\partial T\,}(T_{c}\,,\,0),\quad\frac{\,\partial F\,}{\,\partial Y\,}(T,\,0)\to\frac{\,\partial F\,}{\,\partial Y\,}(T_{c}\,,\,0)\quad\mbox{as}\quad(T,\,0)\to(T_{c}\,,\,0)$ and $\frac{\,\partial F\,}{\,\partial T\,}(T_{c}\,,\,Y)\to\frac{\,\partial F\,}{\,\partial T\,}(T_{c}\,,\,0),\quad\frac{\,\partial F\,}{\,\partial Y\,}(T_{c}\,,\,Y)\to\frac{\,\partial F\,}{\,\partial Y\,}(T_{c}\,,\,0)$ as $(T_{c}\,,\,Y)\to(T_{c}\,,\,0)$. The result follows. ∎ ## 4 The second-order partial derivatives of the function $F$ In this section we deal with the second-order partial derivatives of the function $F$ and show that $F$ is of class $C^{2}$ on $W_{1}$. Let $G$ be given by (4.1) $G(\eta)=\left\\{\begin{array}[]{ll}\displaystyle{\frac{1}{\,\eta^{2}\,}\left\\{3\,g(\eta)+2\,\frac{\tanh\eta}{\,\eta\,\cosh^{2}\eta\,}\right\\}}&(\eta>0),\\\ \vskip 8.5359pt\cr\displaystyle{-\frac{\,16\,}{\,15\,}}&(\eta=0).\end{array}\right.$ A straightforward calculation gives the following. ###### Lemma 4.1. Let $G$ be as in (4.1) and $g$ as in (2.3). Then the function $G$ is of class $C^{1}$ on $[0,\,\infty)$ and satisfies $g^{\prime}(\eta)=-\eta\,G(\eta),\qquad G^{\prime}(0)=0,\qquad\lim_{\eta\to\infty}G(\eta)=\lim_{\eta\to\infty}G^{\prime}(\eta)=0.$ ###### Lemma 4.2. The values of the partial derivatives $\displaystyle{\frac{\,\partial^{2}F\,}{\,\partial T^{2}\,}}$, $\displaystyle{\frac{\partial}{\,\partial Y\,}\left(\frac{\,\partial F\,}{\,\partial T\,}\right)}$, $\displaystyle{\frac{\partial}{\,\partial T\,}\left(\frac{\,\partial F\,}{\,\partial Y\,}\right)}$ and $\displaystyle{\frac{\,\partial^{2}F\,}{\,\partial Y^{2}\,}}$ exist at each point in $W_{1}$. Furthermore, $\frac{\partial}{\,\partial Y\,}\left(\frac{\,\partial F\,}{\,\partial T\,}\right)=\frac{\partial}{\,\partial T\,}\left(\frac{\,\partial F\,}{\,\partial Y\,}\right)\qquad\mbox{on}\quad W_{1}\,.$ ###### Proof. Let $(T,\,Y)\in W_{1}$. Then there is a $\theta$ $(0<\theta<1)$ satisfying $\theta T_{c}<T<T_{c}$. Set $\displaystyle{\eta=\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}}$. Then $\left\|\frac{\partial}{\,\partial T\,}\frac{1}{\,\cosh^{2}\eta\,}\right\|\leq\frac{\,\sqrt{\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}}\,}{k_{B}\theta^{2}T_{c}^{2}},$ where the right side is integrable on $[2k_{B}T_{c}\,\varepsilon,\,\hslash\omega_{D}]$. So the function: $(T,\,Y)\mapsto I_{1}(T,\,Y)$ (see (3.1)), and hence $(\partial F/\partial T)$ (see Lemma 3.2) is differentiable with respect to $T$ on $W_{1}$, and the second-order partial derivative is given by $\frac{\,\partial^{2}F\,}{\,\partial T^{2}\,}(T,\,Y)=\frac{1}{\,k_{B}T^{3}\,}\left\\{I_{1}(T,\,Y)-\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\frac{\eta\,\tanh\eta}{\,\cosh^{2}\eta\,}\,d\xi\right\\},\qquad\eta=\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\,.$ Similarly we can show that $(\partial F/\partial T)$ is differentiable with respect to $Y$ on $W_{1}$, that $(\partial F/\partial Y)$ is differentiable with respect to $T$ on $W_{1}$, and that $(\partial F/\partial Y)$ is differentiable with respect to $Y$ on $W_{1}$. The corresponding second-order partial derivatives are given as follows: $\frac{\partial}{\,\partial Y\,}\left(\frac{\,\partial F\,}{\,\partial T\,}\right)(T,\,Y)=\frac{\partial}{\,\partial T\,}\left(\frac{\,\partial F\,}{\,\partial Y\,}\right)(T,\,Y)=\frac{1}{\,(2k_{B}T)^{3}T\,}\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\frac{\tanh\eta}{\,\eta\,\cosh^{2}\eta\,}\,d\xi,$ (4.2) $\frac{\,\partial^{2}F\,}{\,\partial Y^{2}\,}(T,\,Y)=-\,\frac{1}{\,4\,(2k_{B}T)^{5}\,}\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}G(\eta)\,d\xi\,,\qquad\eta=\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\,.$ Here, $G$ is that in Lemma 4.1 (see also (4.1)). ∎ ###### Lemma 4.3. The partial derivatives $\displaystyle{\frac{\,\partial^{2}F\,}{\,\partial T^{2}\,}}$, $\displaystyle{\frac{\partial}{\,\partial Y\,}\left(\frac{\,\partial F\,}{\,\partial T\,}\right)}$ and $\displaystyle{\frac{\,\partial^{2}F\,}{\,\partial Y^{2}\,}}$ are continuous on $W_{1}$. Consequently, $F$ is of class $C^{2}$ on $W_{1}$. ###### Proof. We show that $(\partial^{2}F/\partial Y^{2})$ is continuous on $W_{1}$. Similarly we can show the continuity of other second-order partial derivatives. By (4.2), it suffices to show that the function: $(T,\,Y)\mapsto I_{3}(T,\,Y)$ is continuous at $(T_{0}\,,\,Y_{0})\in W_{1}$. Here, $I_{3}(T,\,Y)=\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}G(\eta)\,d\xi\,,\qquad\eta=\frac{\,\sqrt{\,\xi^{2}+Y\,}\,}{2k_{B}T}\,.$ Since $(T,\,Y)$ is close to $(T_{0}\,,\,Y_{0})$, it then follows that $T_{0}/2<T$. A straightforward calculation then gives $\displaystyle\left\|\,I_{3}(T,\,Y)-I_{3}(T_{0}\,,\,Y_{0})\,\right\|$ $\displaystyle\leq$ $\displaystyle\hslash\omega_{D}\,\displaystyle{\max_{\eta\geq 0}\|G^{\prime}(\eta)\|}\left(\frac{\,\sqrt{\,\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}\,}\,}{k_{B}T_{0}^{2}}\|T-T_{0}\|+\frac{\|Y-Y_{0}\|}{\,k_{B}T_{0}\sqrt{Y_{0}}\,}\right).$ Hence the function: $(T,\,Y)\mapsto I_{3}(T,\,Y)$ is continuous at $(T_{0}\,,\,Y_{0})\in W_{1}$. ∎ ## 5 Proof of Proposition 2.2 In this section we prove Proposition 2.2 in a sequence of lemmas. ###### Remark 5.1. One may prove Proposition 2.2 on the basis of the implicit function theorem. In this case, an interior point $(T_{0},\,Y_{0})$ of the domain $W$ satisfying $\displaystyle{F(T_{0},\,Y_{0})=0}$ need to exist. But there are the two points $(0,\,\Delta^{2})$ and $(T_{c}\,,\,0)$ in the boundary of $W$ satisfying (5.1) $F(0,\,\Delta^{2})=F(T_{c}\,,\,0)=0.$ So one can not apply the implicit function theorem in its present form. ###### Lemma 5.2. There is a unique solution: $T\mapsto Y=f(T)$ to the gap equation $\displaystyle{F(T,\,Y)=0}$ such that the function $f$ is continuous on the closed interval $[0,\,T_{c}]$ and satisfies $f(0)=\Delta^{2}$ and $f(T_{c})=0$. ###### Proof. By Lemmas 3.3, 3.6 and (5.1), the function: $Y\mapsto F(T_{c}\,,\,Y)$ is monotonically decreasing and there is a $Y_{1}$ $(0<Y_{1}<2\Delta_{0}^{2})$ satisfying $F(T_{c}\,,\,Y_{1})<0$. Note that $Y_{1}$ is arbitrary as long as $0<Y_{1}<2\Delta_{0}^{2}$. Hence, by Lemma 3.6, there is a $T_{1}$ $(0<T_{1}<T_{c})$ satisfying $F(T_{1}\,,\,Y_{1})<0$. Hence, $F(T,\,Y_{1})<0$ for $T_{1}\leq T\leq T_{c}$. On the other hand, by Lemmas 3.3, 3.6 and (5.1), the function: $T\mapsto F(T,\,0)$ is monotonically decreasing and there is a $T_{2}$ $(0<T_{2}<T_{c})$ satisfying $F(T_{2}\,,\,0)>0$. Note that $T_{2}$ is arbitrary as long as $0<T_{2}<T_{c}$. Hence, $F(T,\,0)>0$ for $T_{2}\leq T<T_{c}$. Let $\max(T_{1}\,,\,T_{2})\leq T<T_{c}$ and fix $T$. It then follows from Lemmas 3.3 and 3.6 that the function: $Y\mapsto F(T,\,Y)$ with $T$ fixed is monotonically decreasing on $[0,\,Y_{1}]$. Since $F(T,\,0)>0$ and $F(T,\,Y_{1})<0$, there is a unique $Y$ $(0<Y<Y_{1})$ satisfying $F(T,\,Y)=0$. When $T=T_{c}$, there is a unique value $Y=0$ satisfying $F(T_{c}\,,\,Y)=0$ (see (5.1)). Since $F$ is continuous on $W$ by Lemma 3.6, there is a unique solution: $T\mapsto Y=f(T)$ to the gap equation $\displaystyle{F(T,\,Y)=0}$ such that the function $f$ is continuous on $[\max(T_{1}\,,\,T_{2}),\,T_{c}]$ and $f(T_{c})=0$. Since $(\partial F/\partial Y)(0,\,Y)<0$ $(0<Y<2\Delta_{0}^{2})$ by Lemma 3.2, there is a unique value $Y=\Delta^{2}$ satisfying $F(0,\,Y)=0$. Combining Lemma 3.6 with Lemma 3.3 therefore implies that the function $f$ is continuous on $[0,\,T_{c}]$ and that $f(0)=\Delta^{2}$ and $f(T_{c})=0$. ∎ ###### Lemma 5.3. The function $f$ given by Lemma 5.2 is of class $C^{1}$ on $[0,\,T_{c}]$, and the derivative $f^{\prime}$ satisfies $f^{\prime}(0)=0,\qquad f^{\prime}(T_{c})=8\,k_{B}^{2}T_{c}\,\frac{\,\displaystyle{\int_{\displaystyle{\varepsilon}}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}\frac{d\eta}{\,\cosh^{2}\eta\,}}\,}{\,\displaystyle{\int_{\displaystyle{\varepsilon}}^{\displaystyle{\hslash\omega_{D}/(2k_{B}T_{c})}}g(\eta)\,d\eta}\,}\,.$ ###### Proof. Lemma 3.6 immediately implies that the function $f$ is of class $C^{1}$ on the interval $[0,\,T_{c}]$ and that its derivative is given by (5.2) $f^{\prime}(T)=-\frac{\,F_{T}(T,\,f(T))\,}{\,F_{Y}(T,\,f(T))\,}\,.$ The values of $f^{\prime}(0)$ and $f^{\prime}(T_{c})$ are derived from (5.2). ∎ Combining (5.2) with Lemma 3.3 immediately yields the following. ###### Lemma 5.4. The function $f$ given by Lemma 5.2 is monotonically decreasing on $[0,\,T_{c}]$: $f(0)=\Delta^{2}>f(T_{1})>f(T_{2})>f(T_{c})=0,\qquad 0<T_{1}<T_{2}<T_{c}\,.$ Let $\phi$ be a function of $\eta$ and let $\eta$ be a function of $\xi$. Set (5.3) $I\left[\,\phi(\eta)\,\right]=\int_{2k_{B}T_{c}\,\varepsilon}^{\hslash\omega_{D}}\phi(\eta)\,d\xi\,.$ ###### Lemma 5.5. Let $f$ be given by Lemma 5.2 and let $I\left[\cdot\right]$ be as in (5.3). Then the function $f$ is of class $C^{2}$ on $[0,\,T_{c}]$, and the second derivative $f^{\prime\prime}$ satisfies $\displaystyle{f^{\prime\prime}(0)=0}$ and $\displaystyle f^{\prime\prime}(T_{c})$ $\displaystyle=$ $\displaystyle 16\,k_{B}^{2}\,\frac{\,\displaystyle{I\left[\,\frac{\,\eta_{0}\tanh\eta_{0}-1\,}{\cosh^{2}\eta_{0}}\,\right]}\,}{\,\displaystyle{I\left[\,g(\eta_{0})\,\right]}\,}-32\,k_{B}^{2}\,\frac{\,\displaystyle{I\left[\,\frac{1}{\,\cosh^{2}\eta_{0}\,}\,\right]I\left[\,\frac{\,\tanh\eta_{0}\,}{\,\eta_{0}\cosh^{2}\eta_{0}\,}\,\right]}\,}{\,\displaystyle{\left\\{\,I\left[\,g(\eta_{0})\,\right]\,\right\\}^{2}}\,}$ $\displaystyle\quad+8\,k_{B}^{2}\,\frac{\,\displaystyle{\left\\{I\left[\,\frac{1}{\,\cosh^{2}\eta_{0}\,}\,\right]\right\\}^{2}I\left[\,G(\eta_{0})\,\right]}\,}{\,\displaystyle{\left\\{\,I\left[\,g(\eta_{0})\,\right]\,\right\\}^{3}}\,}\,,\qquad\eta_{0}=\frac{\xi}{\,2k_{B}T_{c}\,}\,.$ ###### Proof. Lemma 4.3 implies that $f$ is of class $C^{2}$ on the open interval $(0,\,T_{c})$ and that (5.4) $f^{\prime\prime}(T)=\frac{\,-F_{TT}F_{Y}^{2}+2F_{TY}F_{T}F_{Y}-F_{YY}F_{T}^{2}\,}{F_{Y}^{3}}\,,\qquad 0<T<T_{c}\,.$ So we have only to deal with $f$ and its derivatives at $T=0$ and at $T=T_{c}$. Step 1. We show that $f^{\prime}$ is differentiable at $T=0$ and that $f^{\prime\prime}$ is continuous at $T=0$. Note that $f^{\prime}(0)=0$ by Lemma 5.3. Since $T$ is close to $T=0$, the inequality $f(T)>\Delta_{0}^{2}\,/2$ holds. It then follows from (5.2) and Lemma 3.2 that $\left\|\frac{\,f^{\prime}(T)-f^{\prime}(0)\,}{T}\right\|\leq 4\,\frac{\,\displaystyle{{\;\sqrt{\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}}\;}^{3}\exp\left(-\frac{\,\sqrt{\Delta_{0}^{2}/2}\,}{k_{B}T}\right)}\,}{\displaystyle{k_{B}T^{3}\left(\tanh\eta_{1}-\frac{\eta_{1}}{\,\cosh^{2}\eta_{1}\,}\right)}}\to 0\qquad(T\downarrow 0).$ Here, $\displaystyle{\eta_{1}=\frac{\,\sqrt{\,\xi_{1}^{2}+f(T)\,}\,}{2k_{B}T}\to\infty}$ as $T\downarrow 0$ $(2k_{B}T_{c}\,\varepsilon<\xi_{1}<\hslash\omega_{D})$. Hence $f^{\prime}$ is differentiable at $T=0$ and $f^{\prime\prime}(0)=0$. By (5.4), a similar argument gives $\displaystyle{\lim_{T\downarrow 0}f^{\prime\prime}(T)=0}$. Hence $f^{\prime\prime}$ is continuous at $T=0$. Step 2. We show that $f^{\prime}$ is differentiable at $T=T_{c}$ and that $f^{\prime\prime}$ is continuous at $T=T_{c}$. Note that $f^{\prime}(T_{c})=8\,k_{B}^{2}T_{c}\,\frac{\,\displaystyle{I\left[\,\frac{1}{\,\cosh^{2}\eta_{0}\,}\,\right]}\,}{\,\displaystyle{I\left[\,g(\eta_{0})\,\right]}\,}\,,\qquad\eta_{0}=\frac{\xi}{\,2k_{B}T_{c}\,}$ by Lemma 5.3. It follows from (5.2) and Lemma 3.2 that $f^{\prime}(T)=8\,k_{B}^{2}T\,\frac{\,\displaystyle{I\left[\,\frac{1}{\,\cosh^{2}\eta\,}\,\right]}\,}{\,\displaystyle{I\left[\,g(\eta)\,\right]}\,}\,,\qquad\eta=\frac{\,\sqrt{\,\xi^{2}+f(T)\,}\,}{2k_{B}T}\,.$ Hence $\displaystyle\frac{\,f^{\prime}(T_{c})-f^{\prime}(T)\,}{T_{c}-T}$ $\displaystyle=$ $\displaystyle 8\,k_{B}^{2}\,\frac{\,\displaystyle{I\left[\,\frac{1}{\,\cosh^{2}\eta_{0}\,}\,\right]}\,}{\,\displaystyle{I\left[\,g(\eta_{0})\,\right]}\,}+\frac{8\,k_{B}^{2}T}{\,T_{c}-T\,}\frac{I\left[\,\frac{1}{\,\cosh^{2}\eta_{0}\,}\,\right]\left\\{I\left[\,g(\eta)\,\right]-I\left[\,g(\eta_{0})\,\right]\right\\}}{I\left[\,g(\eta_{0})\,\right]I\left[\,g(\eta)\,\right]}$ $\displaystyle\quad+\frac{8\,k_{B}^{2}T}{\,T_{c}-T\,}\,\frac{\,I\left[\,g(\eta_{0})\,\right]\left\\{I\left[\,\frac{1}{\,\cosh^{2}\eta_{0}\,}\,\right]-I\left[\,\frac{1}{\,\cosh^{2}\eta\,}\,\right]\right\\}\,}{I\left[\,g(\eta_{0})\,\right]\,I\left[\,g(\eta)\,\right]}\,.$ Note that $g(\eta)-g(\eta_{0})=(\eta-\eta_{0})g^{\prime}(\eta_{1})$ and $\cosh\eta-\cosh\eta_{0}=(\eta-\eta_{0})\sinh\eta_{2}$. Here, $\eta_{0}=\frac{\xi}{\,2k_{B}T_{c}\,}<\eta_{i}<\eta=\frac{\,\sqrt{\,\xi^{2}+f(T)\,}\,}{2k_{B}T}\,,\qquad i=1,\,2$ and $\eta-\eta_{0}=\frac{1}{\,2k_{B}T\,}\left\\{\frac{f(T)}{\,\sqrt{\,\xi^{2}+f(T)\,}+\xi\,}+\xi\frac{\,T_{c}-T\,}{T_{c}}\right\\}.$ Since $T$ is close to $T_{c}$, the inequality $T>T_{c}\,/2$ holds. Therefore, by Lemma 4.1, $\left\|\frac{g^{\prime}(\eta_{1})}{\,\sqrt{\,\xi^{2}+f(T)\,}+\xi\,}\right\|\leq\frac{1}{\,k_{B}T_{c}\,}\,\max_{\eta\geq 0}\left\|G(\eta)\right\|$ and $\left\|\frac{\sinh\eta_{2}}{\,\sqrt{\,\xi^{2}+f(T)\,}+\xi\,}\right\|\leq\frac{1}{\,k_{B}T_{c}\,}\,\max_{0\leq\eta\leq M}\left\|\frac{\,\sinh\eta\,}{\eta}\right\|,\qquad M=\frac{\,\sqrt{\hslash^{2}\omega_{D}^{2}+2\,\Delta_{0}^{2}}\,}{k_{B}T_{c}}\,.$ So $f^{\prime}$ is differentiable at $T=T_{c}$, and it is easy to see that the form of $f^{\prime\prime}(T_{c})$ is exactly the same as that mentioned just above. Furthermore, it follows from (5.4) that $f^{\prime\prime}$ is continuous at $T=T_{c}$. ∎ ## 6 Proof of Theorem 2.3 In this section we prove Theorem 2.3 in a sequence of lemmas. Fixing the values of the chemical potential $\mu$ and of the volume of our physical system, we deal with the dependence of the thermodynamical potential $\Omega$ on the temperature $T$ only. ###### Lemma 6.1. Let $V$ be as in (1). Then $V$ is of class $C^{2}$ on $(0,\,\infty)$. ###### Proof. For each $T>0$, there are a $\theta_{1}$ $(0<\theta_{1}<1)$ and a $\theta_{2}$ $(\theta_{2}>1)$ satisfying $\theta_{1}T_{c}<T<\theta_{2}T_{c}$. Then $\left\|\frac{\partial}{\,\partial T\,}\ln\left(1+e^{\displaystyle{-\|\xi\|/(k_{B}T)}}\right)\right\|\leq\frac{\,\|\xi\|\,e^{\displaystyle{-\|\xi\|/(k_{B}\theta_{2}T_{c})}}\,}{k_{B}\theta_{1}^{2}T_{c}^{2}},$ where the right side is integrable on $[-\mu,\,-\hslash\omega_{D}]$ and on $[\hslash\omega_{D},\,\infty)$ since $N(\xi)=O(\sqrt{\xi})$ as $\xi\to\infty$ (see Remark 1.3). Hence $V$ is differentiable on $(0,\,\infty)$ and $\displaystyle\frac{\,\partial V\,}{\partial T}(T)$ $\displaystyle=$ $\displaystyle-2k_{B}\int_{[-\mu,\,-\hslash\omega_{D}]\,\cup\,[\hslash\omega_{D},\,\infty)}N(\xi)\ln\left(1+e^{\displaystyle{-\|\xi\|/(k_{B}T)}}\right)\,d\xi$ $\displaystyle\quad-\frac{2}{\,T\,}\int_{[-\mu,\,-\hslash\omega_{D}]\,\cup\,[\hslash\omega_{D},\,\infty)}N(\xi)\,\frac{\|\xi\|}{\,1+e^{\displaystyle{\|\xi\|/(k_{B}T)}}\,}\,d\xi.$ A similar argument gives $\left\|\frac{\partial}{\,\partial T\,}\,\frac{1}{\,1+e^{\displaystyle{\|\xi\|/(k_{B}T)}}\,}\right\|\leq\frac{\,\|\xi\|\,e^{\displaystyle{-\|\xi\|/(k_{B}\theta_{2}T_{c})}}\,}{k_{B}\theta_{1}^{2}T_{c}^{2}},$ where the right side is integrable on $[-\mu,\,-\hslash\omega_{D}]$ and on $[\hslash\omega_{D},\,\infty)$. Therefore, $(\partial V/\partial T)$ is again differentiable on $(0,\,\infty)$ and $\frac{\,\partial^{2}V\,}{\,\partial T^{2}\,}(T)=-\frac{2}{\,k_{B}T^{3}\,}\int_{[-\mu,\,-\hslash\omega_{D}]\,\cup\,[\hslash\omega_{D},\,\infty)}N(\xi)\,\frac{\|\xi\|^{2}\,e^{\displaystyle{\|\xi\|/(k_{B}T)}}}{\,\left(1+e^{\displaystyle{\|\xi\|/(k_{B}T)}}\right)^{2}\,}\,d\xi.$ Clearly, $(\partial^{2}V/\partial T^{2})$ is continuous on $(0,\,\infty)$. ∎ ###### Lemma 6.2. Let $\Omega_{N}$ be as in (1). Then $\Omega_{N}$ is of class $C^{2}$ on $(0,\,\infty)$. ###### Proof. An argument similar to that in the proof of Lemma 6.1 gives that $\Omega_{N}$ is of class $C^{2}$ on $(0,\,\infty)$ and that the derivatives are given by $\displaystyle\frac{\,\partial\Omega_{N}\,}{\,\partial T\,}(T)$ $\displaystyle=$ $\displaystyle-4N_{0}k_{B}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\ln\left(1+e^{\displaystyle{-\xi/(k_{B}T)}}\right)\,d\xi$ $\displaystyle\quad-\frac{\,4N_{0}\,}{T}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\frac{\xi}{\,1+e^{\displaystyle{\xi/(k_{B}T)}}\,}\,d\xi+\frac{\,\partial V\,}{\partial T}(T),$ $\displaystyle\frac{\,\partial^{2}\Omega_{N}\,}{\,\partial T^{2}\,}(T)$ $\displaystyle=$ $\displaystyle-\frac{\,4N_{0}\,}{\,k_{B}T^{3}\,}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\frac{\xi^{2}\,e^{\displaystyle{\xi/(k_{B}T)}}}{\,\left(1+e^{\displaystyle{\xi/(k_{B}T)}}\right)^{2}\,}\,d\xi+\frac{\,\partial^{2}V\,}{\,\partial T^{2}\,}(T).$ ∎ ###### Lemma 6.3. Let $\delta$ be as in (1). Then $\delta$ is of class $C^{2}$ on $(0,\,T_{c}]$. ###### Proof. Note that the squared gap function $f$ is of class $C^{2}$ on $[0,\,T_{c}]$ by Lemma 5.5 and that (6.1) $\delta(T_{c})=0$ since $f(T_{c})=0$ (see Lemma 5.2). A straightforward calculation gives that $\delta$ is continuous on $(0,\,T_{c}]$ and that $\left\|\frac{\partial}{\,\partial T\,}\sqrt{\xi^{2}+f(T)}\right\|\leq\frac{\,\displaystyle{\max_{0\leq T\leq T_{c}}\left\|f^{\prime}(T)\right\|}\,}{\,4k_{B}T_{c}\,\varepsilon\,},$ where the right side is integrable on $[2k_{B}T_{c}\,\varepsilon,\,\hslash\omega_{D}]$. By an argument similar to that in the proof of Lemma 6.1, $\delta$ is differentiable on $(0,\,T_{c}]$ and the derivative is given by $\displaystyle\frac{\,\partial\delta\,}{\,\partial T\,}(T)$ $\displaystyle=$ $\displaystyle f^{\prime}(T)\left\\{\frac{1}{\,U_{0}\,}-N_{0}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\frac{1}{\,\sqrt{\,\xi^{2}+f(T)\,}\,}\tanh\frac{\,\sqrt{\,\xi^{2}+f(T)\,}\,}{2k_{B}T}\,d\xi\right\\}$ $\displaystyle\quad-4N_{0}k_{B}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\ln\frac{\,1+e^{-\displaystyle{\sqrt{\xi^{2}+f(T)}/(k_{B}T)}}\,}{1+e^{-\displaystyle{\xi/(k_{B}T)}}}\,d\xi$ $\displaystyle\quad+\frac{\,4N_{0}\,}{T}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\left\\{\frac{\xi}{\,1+e^{\displaystyle{\xi/(k_{B}T)}}\,}-\frac{\sqrt{\xi^{2}+f(T)}}{\,1+e^{\displaystyle{\sqrt{\xi^{2}+f(T)}/(k_{B}T)}}\,}\right\\}\,d\xi,$ where the first term on the right side is equal to 0 by the gap equation (1.1). Note that (6.2) $\frac{\,\partial\delta\,}{\,\partial T\,}(T_{c})=0.$ An argument similar to that in the proof of Lemma 6.1 gives that $(\partial\delta/\partial T)$ is again differentiable on $(0,\,T_{c}]$ and the second-order derivative is given by $\displaystyle\frac{\,\partial^{2}\delta\,}{\,\partial T^{2}\,}(T)$ $\displaystyle=$ $\displaystyle\frac{4N_{0}}{\,k_{B}T^{3}\,}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\frac{\xi^{2}\,e^{\displaystyle{\xi/(k_{B}T)}}}{\,\left(1+e^{\displaystyle{\xi/(k_{B}T)}}\right)^{2}\,}\,d\xi$ $\displaystyle-\frac{4N_{0}}{\,k_{B}T^{3}\,}\int_{\displaystyle{2k_{B}T_{c}\,\varepsilon}}^{\displaystyle{\hslash\omega_{D}}}\frac{e^{\displaystyle{\sqrt{\xi^{2}+f(T)}/(k_{B}T)}}}{\,\left(1+e^{\displaystyle{\sqrt{\xi^{2}+f(T)}/(k_{B}T)}}\right)^{2}\,}\left\\{\xi^{2}+f(T)-\frac{\,Tf^{\prime}(T)\,}{2}\right\\}\,d\xi,$ which is also continuous on $(0,\,T_{c}]$. Thus $\delta$ is of class $C^{2}$ on $(0,\,T_{c}]$, and (6.3) $\frac{\,\partial^{2}\delta\,}{\,\partial T^{2}\,}(T_{c})=\frac{\,2N_{0}f^{\prime}(T_{c})\,}{T_{c}}\left(\frac{1}{\,1+e^{\displaystyle{2\varepsilon}}\,}-\frac{1}{\,1+e^{\displaystyle{\hslash\omega_{D}/(k_{B}T_{c})}}\,}\right).$ ∎ We now give a proof of Theorem 2.3. ###### Lemma 6.4. Let $f^{\prime}(T_{c})$ be given by Lemma 5.3 and let $\Omega$ be the thermodynamical potential given by Definition 1.4. (i) The thermodynamical potential $\Omega$, regarded as a function of $T$, is of class $C^{1}$ on $(0,\,\infty)$. (ii) The second-order derivative $\left(\partial^{2}\Omega/\partial T^{2}\right)$ is continuous on $(0,\,\infty)\setminus\\{T_{c}\\}$. (iii) $\lim_{T\uparrow T_{c}}\frac{\,\partial^{2}\Omega\,}{\,\partial T^{2}\,}(T)-\lim_{T\downarrow T_{c}}\frac{\,\partial^{2}\Omega\,}{\,\partial T^{2}\,}(T)=\frac{\,2N_{0}f^{\prime}(T_{c})\,}{T_{c}}\left(\frac{1}{\,1+e^{\displaystyle{2\varepsilon}}\,}-\frac{1}{\,1+e^{\displaystyle{\hslash\omega_{D}/(k_{B}T_{c})}}\,}\right).$ ###### Proof. Note that $\displaystyle{\delta(T_{c})=(\partial\delta/\partial T)(T_{c})=0}$ (see (6.1) and (6.2)). Hence both (i) and (ii) follow immediately from Lemmas 6.1, 6.2 and 6.3. Since $\lim_{T\uparrow T_{c}}\left(\partial^{2}\Omega/\partial T^{2}\right)(T)-\lim_{T\downarrow T_{c}}\left(\partial^{2}\Omega/\partial T^{2}\right)(T)=(\partial^{2}\delta/\partial T^{2})(T_{c}),$ (iii) follows immediately from (6.3). ∎ ###### Remark 6.5. This lemma implies that the transition to a superconducting state at the transition temperature $T_{c}$ is a second-order phase transition. ## References * [1] J. Bardeen, L. N. Cooper and J. R. Schrieffer, Theory of superconductivity, Phys. Rev., 108 (1957), pp. 1175–1204. * [2] P. Billard and G. Fano, An existence proof for the gap equation in the superconductivity theory, Commun. Math. Phys., 10 (1968), pp. 274–279. * [3] N. N. Bogoliubov, A new method in the theory of superconductivity I, Soviet Phys. JETP, 34 (1958), pp. 41–46. * [4] C. Hainzl, E. Hamza, R. Seiringer and J. P.Solovej, The BCS functional for general pair interactions, Commun. Math. Phys., in press. arXiv: 0703086. * [5] C. Hainzl and R. Seiringer, Spectral properties of the BCS gap equation of superfluidity, arXiv: 0802.0446. * [6] M. Niwa, Fundamentals of Superconductivity, Tokyo Denki University Press, Tokyo, 2002 (in Japanese). * [7] F. Odeh, An existence theorem for the BCS integral equation, IBM J. Res. Develop., 8 (1964), pp. 187–188. * [8] A. Vansevenant, The gap equation in the superconductivity theory, Physica, 17D (1985), pp. 339–344. * [9] S. Watanabe, Superconductivity and the BCS-Bogoliubov theory, JP Jour. Algebra, Number Theory and Appl., 11 (2008), pp. 137–158. * [10] S. Watanabe, Smoothness of the gap function in the BCS-Bogoliubov theory of superconductivity, preprint. * [11] Y. Yang, On the Bardeen-Cooper-Schrieffer integral equation in the theory of superconductivity, Lett. Math. Phys., 22 (1991), pp. 27–37. * [12] J. M. Ziman, Principles of the Theory of Solids, Cambridge University Press, Cambridge, 1972.	arxiv-papers	2008-08-26T03:26:28	2024-09-04T02:48:57.457952	{ "license": "Public Domain", "authors": "Shuji Watanabe", "submitter": "Shuji Watanabe", "url": "https://arxiv.org/abs/0808.3438" }
0808.3497	# Thermally activated switching in the presence of non-Gaussian noise Lora Billings(1), Mark I. Dykman(2)∗, and Ira B. Schwartz(3) dykman@pa.msu.edu (1)Department of Mathematical Sciences, Montclair State University, Montclair, NJ 07043 (2) Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824 (3) US Naval Research Laboratory, Code 6792, Nonlinear System Dynamics Section, Plasma Physics Division, Washington, DC 20375 ###### Abstract We study the effect of a non-Gaussian noise on interstate switching activated primarily by Gaussian noise. Even weak non-Gaussian noise can strongly change the switching rate. The effect is determined by all moments of the noise distribution. The explicit analytical results are compared with the results of simulations for an overdamped system driven by white Gaussian noise and a Poisson noise. Switching induced by a purely Poisson noise is also discussed. ###### pacs: 05.40.-a, 72.70.+m, 05.70.Ln, 05.40.Ca Much progress has been made recently in the studies of switching between coexisting stable states, primarily because switching can be now investigated for a large variety of well-controlled micro- and mesoscopic systems ranging from trapped electrons and atoms to Josephson junctions and to nano- and micro-mechanical oscillators Lapidus et al. (1999); Siddiqi et al. (2005); Aldridge and Cleland (2005); Kim et al. (2005); Gommers et al. (2005); Stambaugh and Chan (2006); Abdo et al. (2007); Lupaşcu et al. (2007); Katz et al. (2007); Serban and Wilhelm (2007). Fluctuations in these systems are usually due to thermal or externally applied Gaussian noise. However, as the systems become smaller, an increasingly important role may be played also by non-Gaussian noise. It may come, for example, from one or a few two-state fluctuators hopping at random between the states, in which case the noise may be often described as a telegraph noise. The switching probability is sensitive to a non-Gaussian noise. This sensitivity attracted much attention after it was proposed Tobiska and Nazarov (2004) to use switching in Josephson junctions to measure the full counting statistics in electronic circuits Levitov and Lesovik (1992); Nazarov (2007). Several theoretical Pekola (2004); Ankerhold (2007); Sukhorukov and Jordan (2007); Grabert (2008) and experimental Timofeev et al. (2007); Huard et al. (2007) papers on measuring the 3rd moment of the current distribution from the switching rates were published recently, and different theoretical approaches were compared in Refs. Novotný, 2008; Sukhorukov and Jordan, 2008. In this paper we study switching induced by Gaussian noise in the presence of an additional non-Gaussian noise. Even where the latter has a smaller intensity than the Gaussian noise, its effect on the switching rate may be exponentially strong. We show that it can be described in a simple form in terms of the noise characteristic functional, thus accounting for all moments of the noise distribution. The analytical results are compared with simulations for an overdamped system driven by white Gaussian noise and a Poisson noise. We also consider switching induced by a Poisson noise alone; here, the result for the rate may be qualitatively different from that in the weakly non-Gaussian noise approximation. The potentially strong effect of an extra modulation, whether random or regular, on the rate of Gaussian noise induced switching can be understood from the well-established picture of the switching dynamics. Switching events result from large rare noise outbursts. For Gaussian noise, the switching rate is $W\propto\exp(-R/D)$, where $R$ is the activation energy and $D$ is the noise intensity Freidlin and Wentzell (1998). Even though switching happens at random, the system trajectories followed in switching form a narrow tube in the space of dynamical variables ${\bf q}=(q_{1},q_{2},\ldots)$ centered at the most probable (optimal) switching path ${\bf q}_{\rm opt}(t)$. One can think of the effect of an additional modulation in terms of a generalized work done by the modulation on the system moving along ${\bf q}_{\rm opt}(t)$ Smelyanskiy et al. (1997); Dykman et al. (1997). This work changes the activation barrier. The change $\delta R$ is proportional to the modulation amplitude. Therefore the overall change of the switching rate $\propto\exp(-\delta R/D)$ depends on the modulation amplitude exponentially. The switching rate gives the probability current from the occupied state Kramers (1940), making it an observable quantity. Thus, if the modulation is random, one has to simply average the factor $\exp(-\delta R/D)$ over realizations of the modulation. Since $\delta R$ is linear in the characteristic amplitude of random modulation, the ratio $\delta R/D$ does not have to be small even where the modulation intensity, which is quadratic in the amplitude, is smaller than $D$. However, the distribution of non-Gaussian modulation may decay slower than Gaussian on the tail. To determine whether the effect of a non-Gaussian noise on switching may be regarded as a perturbation one has to compare the probabilities of appropriate large fluctuations induced by the Gaussian and non-Gaussian noises, taking into account all moments of the distribution. We study switching for a system described by the Langevin equation $\dot{\bf q}={\bf K}({\bf q})+{\bf f}(t)+{\bm{\xi}}(t).$ (1) We assume that, in the absence of noise, the system has a stable stationary state ${\bf q}_{A}$ and a saddle point ${\bf q}_{\cal S}$ on the boundary of the basin of attraction to ${\bf q}_{A}$, with ${\bf K}({\bf q}_{A})={\bf K}({\bf q}_{\cal S})=0$. Switching from the stable state is due to the forces ${\bf f}(t)$ and ${\bm{\xi}}(t)$, which are the Gaussian and non-Gaussian noises, respectively. We separate them, since physically they often come from different sources. It is convenient to characterize ${\bf f}(t)$ by its probability density functional ${\cal P}_{{\bf f}}[{\bf f}(t)]=\exp\left(-{\cal R}_{{\bf f}}/D\right)$, ${\cal R}_{{\bf f}}[{\bf f}(t)]=\frac{1}{4}\int dt\,dt^{\prime}\,{\bf f}(t)\hat{{\cal F}}(t-t^{\prime}){\bf f}(t^{\prime}),$ (2) where $\hat{{\cal F}}(t-t^{\prime})/2D$ is the inverse of the pair correlator of ${\bf f}(t)$. The characteristic noise intensity $D$ is small, so that the switching rate $W\ll t_{\rm r}^{-1},t_{\rm c}^{-1}$, where $t_{\rm r}$ is the relaxation time of the system and $t_{\rm c}$ is the noise correlation time. The non-Gaussian noise is more conveniently described for our purpose by the characteristic functional $\tilde{\cal P}_{{\bm{\xi}}}[{\bf k}]=\left\langle\exp\left[i\int dt{\bf k}(t){\bm{\xi}}(t)\right]\right\rangle_{{\bm{\xi}}},$ (3) where $\langle\ldots\rangle_{{\bm{\xi}}}$ means averaging over ${\bm{\xi}}(t)$. We first consider the case where the intensity of the non-Gaussian noise ${\bm{\xi}}$ is smaller than $D$. We will disregard corrections proportional to this intensity, but the ratio of the characteristic amplitude $g_{0}$ of ${\bm{\xi}}$ to $D$ will not be assumed small. The switching rate can be written as $\displaystyle W=C\left\langle\exp[-{\cal R}[{\bm{\xi}}]/D]\right\rangle_{{\bm{\xi}}},$ (4) $\displaystyle{\cal R}[{\bm{\xi}}]=\min\left\\{{\cal R}_{{\bf f}}+i\int dt\,{\bf k}(t)\left[\dot{\bf q}-{\bf K}-{\bf f}(t)-{\bm{\xi}}(t)\right]\right\\},$ where $C$ is the prefactor that weakly depends on the noise intensity. The minimum is taken over trajectories ${\bf f}(t),{\bf q}(t),{\bf k}(t)$ that satisfy boundary conditions ${\bf f}(t),{\bf k}(t)\to 0$ for $t\to\pm\infty$, ${\bf q}_{t\to-\infty}\to{\bf q}_{A},{\bf q}_{t\to\infty}\to{\bf q}_{\cal S}$. This formulation was proposed in the weak-noise limit Dykman (1990); Dykman and Smelyanskiy (1998) for a time-periodic ${\bm{\xi}}(t)$, in which case ${\bf q}_{A,{\cal S}}$ are also periodic and there is no averaging over ${\bm{\xi}}$. The variational problem (4) describes coupled optimal trajectories ${\bf f}_{\rm opt}(t),{\bf q}_{\rm opt}(t),{\bf k}_{\rm opt}(t)$, with ${\bf f}_{\rm opt}(t)$ being the most probable noise realization that brings the system to the saddle on the basin boundary of the initially occupied state. It is known from variational calculus that, to first order in ${\bm{\xi}}$, the effect of ${\bm{\xi}}(t)$ on ${\cal R}$ can be calculated along the optimal trajectory unperturbed by ${\bm{\xi}}(t)$. Such a trajectory is an instanton. Its typical duration is $\sim\max(t_{\rm r},t_{\rm c})$. It is translation-invariant with respect to time and can be centered at any time $t_{0}$. If ${\bm{\xi}}(t)$ is periodic, it lifts time-translation symmetry and fixes $t_{0}$ (modulo the period) so as to minimize ${\cal R}[{\bm{\xi}}]$. If ${\bm{\xi}}(t)$ is a stationary noise, the switching rate $W$ is independent of time. In this case one can think not of the adjustment of the instanton center $t_{0}$ to ${\bm{\xi}}(t)$, but, equivalently, of the adjustment of ${\bm{\xi}}(t)$ to $t_{0}$ so as to maximize the overall probability of switching. This adjustment provides the major contribution to the value of $W$ when the averaging over realizations of ${\bm{\xi}}(t)$ is performed in Eq. (4) using a solution with a given $t_{0}$. From Eqs. (3), (4) one obtains a simple expression for the switching rate $\displaystyle W=W^{(0)}A_{\rm sw},\qquad A_{\rm sw}=\tilde{\cal P}_{{\bm{\xi}}}[i{\bm{\chi}}/D],$ (5) where $W^{(0)}$ is the switching rate in the absence of non-Gaussian noise. The factor $A_{\rm sw}$ describes the effect of non-Gaussian noise. It is expressed in a closed form in terms of the noise characteristic functional calculated for function ${\bm{\chi}}(t)=-i{\bf k}_{\rm opt}^{(0)}(t)$, where ${\bf k}_{\rm opt}^{(0)}(t)$ is the solution of the variational problem (4) for ${\bm{\xi}}={\bf 0}$. The real function ${\bm{\chi}}(t)$ is the logarithmic susceptibility which describes the linear response of $\log W$ to a perturbation Smelyanskiy et al. (1997); Dykman et al. (1997); Assaf et al. (2008). The structure of Eq. (5) resembles that of the expression for a large fluctuation probability in a birth-death system with non-Gaussian modulation of reaction rates Dykman et al. (2008). From Eq. (5), the effect of a non-Gaussian noise on the switching rate is determined by the ratio of the noise amplitude to the Gaussian noise intensity $D$. Equation (5) applies to both underdamped and overdamped systems. Examples of calculating ${\bf k}_{\rm opt}^{(0)}$ can be found in Refs. Smelyanskiy et al., 1997; Dykman et al., 1997; Assaf et al., 2008 and papers cited therein. As an illustration we will consider the case of a one-component $\delta$-correlated Poisson noise $\xi(t)$ with pulse area $g$ and mean frequency $\nu$. Using the explicit form of the noise characteristic functional Feynman and Hibbs (1965), we obtain $\displaystyle A_{\rm sw}=\exp\left\\{-\nu\int dt\,\left[1-\exp\left(-\chi(t)g/D\right)\right]\right\\},$ (6) where $\chi(t)$ is the corresponding component of the logarithmic susceptibility. If $g/D\ll 1$, $\log A_{\rm sw}$ is a series in $g/D$. The coefficients in this series describe the effects of the moments of the Poisson noise on the switching rate. In the opposite case, $g/D\gg 1$ (but the Poisson noise intensity $\nu g^{2}\ll D$), if $\chi(t)$ becomes negative, then $\log A_{\rm sw}\approx\nu\left[2\pi D/g\ddot{\chi}(t_{m})\right]^{1/2}\exp[-\chi(t_{m})g/D]$ where $t_{m}$ is the instant where $-\chi(t)$ is maximal. If $\chi(t)\geq 0$ for all $t$ and $g/D\gg 1$, the major contribution to $A_{\rm sw}$ comes from the region of small $\chi(t)$. If $\chi(t)$ is small only for $\|t\|\to\infty$, where it decays exponentially with $\|t\|$, then $\log A_{\rm sw}\propto\nu t_{\rm r}\log(g/D)$, to leading order in $g/D$. The Poisson noise distribution does not fall off as steeply as Gaussian. This imposes a limitation on the range of $g/D$ where Poisson noise may be treated as a perturbation and the above theory applies. To see the far-tail effect we consider switching due to a purely Poisson noise, where ${\bf f}={\bf 0}$ in equation of motion (1). We will use the method of optimal fluctuation, as for some other types of non-Gaussian noise McKane (1989). The switching rate is determined, to logarithmic accuracy, by the integral over trajectories ${\bf k}(t),{\bf q}(t)$ of the functional $\left\langle\exp\left\\{-i\int dt{\bf k}(t)[\dot{\bf q}-{\bf K}-{\bm{\xi}}(t)]\right\\}\right\rangle_{{\bm{\xi}}}$ Luciani and Verga (1987). We assume for brevity that different components of the Poisson noise are independent short pulses with areas ${\bf g}=(g_{1},g_{2},\ldots)$ and with average frequency $\nu$. It is convenient to consider a zero-mean noise, ${\bm{\xi}}(t)\to{\bm{\xi}}(t)-\nu{\bf g},\;{\bf K}\to{\bf K}+\nu{\bf g}$; we assume that ${\bf q}_{A},{\bf q}_{\cal S}$ are also appropriately shifted. Of interest for switching are trajectories that approach the saddle point Dykman (1990). In the spirit of the method of optimal fluctuation, for small $\|{\bf g}\|$ and for $\nu\lesssim t_{\rm r}^{-1}$ the integral over trajectories ${\bf q}(t),{\bf k}(t)$ can be calculated by steepest descent. This gives $\displaystyle W=C^{\prime}\exp[-R_{P}],\quad R_{P}=\min\int dt\left(i{\bf k}\dot{\bf q}-H\right),$ (7) $\displaystyle H\equiv H({\bf q},i{\bf k})=-\nu\left(1+i{\bf g}{\bf k}-e^{i{\bf g}{\bf k}}\right)+i{\bf k}{\bf K}({\bf q}).$ The variational problem (7) determines the optimal switching trajectory ${\bf q}_{\rm opt}(t),{\bf k}_{\rm opt}(t)$. It starts at $t\to-\infty$ at ${\bf q}\to{\bf q}_{A},{\bf k}\to{\bf 0}$ and goes to ${\bf q}\to{\bf q}_{\cal S},{\bf k}\to{\bf 0}$ for $t\to\infty$. On this trajectory $H=0$. As in systems driven by white Gaussian noise Freidlin and Wentzell (1998), the optimal trajectory can be described as a Hamiltonian trajectory of an auxiliary system with coordinate ${\bf q}$, momentum $i{\bf k}$, and Hamiltonian $H$. A similar approach was proposed in Ref. Sukhorukov and Jordan, 2007. However, the characteristic functional was not specified and the explicit analysis took into account only the 3rd moment of the noise distribution, which was considered a perturbation; therefore the results do not describe switching due to large Poisson fluctuations. The switching exponent in Eq. (7) is $R_{P}\gg 1$ for small $\|{\bf g}\|$. However, in contrast to the case of Gaussian noise, $R_{P}$ is not proportional to the reciprocal noise intensity $\nu{\bf g}^{2}$. Nor does it scale like reciprocal noise amplitude $\|{\bf g}\|^{-1}$, although $R_{P}\|{\bf g}\|$ appears to slowly vary with $\|{\bf g}\|$. An explicit dependence of the switching rate on the intensity $g$ of Poisson- distributed pulses can be found for a one-variable overdamped system with equation of motion $\dot{q}=-U^{\prime}(q)+f(t)+\xi(t).$ (8) Here, $U(q)$ is the effective potential. The stationary states $q_{A}$ and $q_{\cal S}$ correspond to the minimum and the barrier top of $U(q)$. If $f(t)$ is white Gaussian noise, $\langle f(t)f(t^{\prime})\rangle=2D\delta(t-t^{\prime})$, and the Poisson noise is weak, the Poisson-noise induced factor in the switching rate $A_{\rm sw}$ is described by Eq. (6) with $\chi(t)=-f_{\rm opt}^{(0)}(t)/2=-\dot{q}_{\rm opt}^{(0)}(t)$ and with $\dot{q}_{\rm opt}^{(0)}=U^{\prime}\left(q_{\rm opt}^{(0)}\right)$. In the opposite case where switching is due to purely Poisson noise, i.e., $f=0$ in Eq. (8), from Eq. (7) $\displaystyle R_{P}=\frac{1}{g}\int\nolimits_{\tilde{q}_{A}}^{\tilde{q}_{\cal S}}dq\kappa(q),\quad\kappa=\log\left\\{1+\left[\kappa U^{\prime}(q)/g\nu\right]\right\\}.$ (9) Here, $\tilde{q}_{A}$ and $\tilde{q}_{\cal S}$ are the shifted extrema of the potential given by equation $U^{\prime}(q)=g\nu$. From Eq. (9), $R_{P}\sim r_{P}\log(r_{P}/\nu t_{\rm r})$, with $r_{P}=(\tilde{q}_{\cal S}-\tilde{q}_{A})/g$. A qualitative feature of unipolar (pulses of one sign) Poisson noise is that, for an overdamped system, it causes switching only provided the noise pulses push the system from the stable state towards the saddle. In this case $r_{P}>0$. There is no switching for pulses of the opposite sign. The “one- sidedness” of fluctuations in overdamped systems has other manifestations, which includes the work fluctuation distribution Cohen . On the other hand, we expect that an underdamped system should be able to switch for Poisson pulses of any sign, in which case there should be a critical value of damping for which switching from the state is possible for a given sign of $g$. Equations (7), (9) apply if $r_{P},r_{P}/\nu t_{\rm r}\gg 1$. As a cause of switching, a Poisson noise is effectively weaker than a Gaussian noise if the switching exponent $R_{P}$ is larger than the switching exponent for the Gaussian noise. For white Gaussian noise of intensity $D$ in Eq. (8), the switching exponent is $\Delta U/D$ with $\Delta U=U(q_{\cal S})-U(q_{A})$ Kramers (1940). The condition $R_{P}>\Delta U/D$ effectively limits the range of applicability of Eq. (6). We saw that $\log A_{\rm sw}$ becomes large provided $-\chi(t_{m})g/D\gg 1$. An order of magnitude estimate shows that $-\chi(t_{m})g/D\sim(\Delta U/D)R_{P}^{-1}\log(r_{P}/\nu t_{\rm r})$, and therefore from Eq. (9) the large $\log A_{\rm sw}$ asymptotics applies only provided $\log(r_{P}/\nu t_{\rm r})\gg 1$. This condition is compatible with $-\chi(t_{m})g/D\gg 1$ only for very small $D$. We now apply the above results to an overdamped system (8) with a double-well potential $U(q)=-q^{2}/2+q^{4}/4,$ (10) which has been extensively studied in the context of white-noise driven systems. In the absence of Poisson noise, the escape rate in this case is $W^{(0)}=(\sqrt{2}/\pi)\exp(-1/4D)$, and the logarithmic susceptibility for escape from the negative-$q$ well ($q_{A}=-1$) is $\chi(t)=-\exp(t/2)(2\cosh t)^{-3/2}$ ($\chi(t)$ has opposite sign for switching from $q_{A}=1$). Figure 1: Poisson noise induced change of the switching exponent for an overdamped Brownian particle in potential (10). The dimensionless Poisson noise frequency is $\nu=0.5$. The data of numerical simulations are shown with circles and squares for the cases where in escape the particle moves along and opposite to Poisson pulses, respectively. The solid curves show the weak- Poisson noise theory (6) for these cases, and the dashed curves show the approximation where only the three moments of Poisson noise are taken into account. In Fig. 1 we present results of Monte Carlo simulations of switching of an overdamped Brownian particle described by Eqs. (8), (10). They are compared with the weak Poisson noise prediction, Eq. (6), and with the approximation where only terms up to $g^{3}$ are kept in Eq. (6). The Poisson noise intensity $\nu g^{2}<D$ in the whole range of studied $g/D$. For $g/D\lesssim 1$ the effect of Poisson noise is small and comes, primarily, to the change of the activation barrier, $\Delta U\to\Delta U\pm\nu g$, and the effective noise intensity, $D\to D+\nu g^{2}$. For larger $g/D\gtrsim 3$ the switching exponent changes significantly, as expected. For switching in the direction opposite to Poisson noise pulses, where $\log W/W^{(0)}<0$, the numerics agrees well with Eq. (6). This is to be expected, since the far tail of Poisson noise distribution is immaterial here; for $g/D\gtrsim 5$ the results differ noticeably from the three-moments approximation. For switching along the pulses, because of the far-tail effect, with increasing $g$ Poisson noise quickly becomes as important as white noise for chosen $D$. Therefore the perturbation theory fails and the dependence of the switching exponent on $g$ is much weaker than the exponential dependence expected from Eq. (6). Figure 2: Switching exponent $R_{P}=-\ln W$ for an overdamped particle in a potential (10) driven by a Poisson noise with dimensionless mean frequency $\nu=0.5$ and pulse intensity $g$. Squares show the results of Monte-Carlo simulations, the solid line is the asymptotic theory (9). Numerical simulations of the switching rate for purely Poisson noise are shown in Fig. 2. There is good agreement between the data and the asymptotic theory (9) for small $g$. In this range $R_{P}g$ slowly varies with $g$. In conclusion, we have considered switching in systems simultaneously driven by a Gaussian and a non-Gaussian noise. Even where the non-Gaussian noise has intensity smaller than that of the Gaussian, it may strongly change the switching rate. The effect is determined by the ratio of the non-Gaussian noise amplitude to the Gaussian noise intensity. It is described by the characteristic functional of the non-Gaussian noise calculated for a function determined by the system dynamics without this noise. A non-Gaussian tail of the noise distribution may strongly modify the switching rate even for a small noise intensity. We demonstrate this effect using Poisson noise as an example. Analytical results are compared with Monte Carlo simulations. MID acknowledges valuable discussions with M. Büttiker and A. Korotkov. LB is supported by ARO grant No. W911NF-06-1-0320. MID is supported by ARO grant No. W911NF-06-1-0324 and NSF grant DMR-0305746. IBS is supported by the Office of Naval Research. ## References * Lapidus et al. (1999) L. J. Lapidus, D. Enzer, and G. Gabrielse, Phys. Rev. Lett. 83, 899 (1999). * Siddiqi et al. (2005) I. Siddiqi et al., Phys. Rev. Lett. 94, 027005 (2005). * Aldridge and Cleland (2005) J. S. Aldridge and A. N. Cleland, Phys. Rev. Lett. 94, 156403 (2005). * Kim et al. (2005) K. Kim et al., Phys. Rev. A 72, 053402 (2005). * Gommers et al. (2005) R. Gommers et al., Phys. Rev. Lett. 94, 143001 (2005). * Stambaugh and Chan (2006) C. Stambaugh and H. B. Chan, Phys. Rev. B 73, 172302 (2006). * Abdo et al. (2007) B. Abdo, et al., J. Appl. Phys. 101, 083909 (2007). * Lupaşcu et al. (2007) A. Lupaşcu et al., Nature Physics 3, 119 (2007). * Katz et al. (2007) I. Katz et al., Phys. Rev. Lett. 99, 040404 (2007). * Serban and Wilhelm (2007) I. Serban and F. K. Wilhelm, Phys. Rev. Lett. 99, 137001 (2007). * Tobiska and Nazarov (2004) J. Tobiska and Y. Nazarov, Phys. Rev. Lett. 93, 106801 (2004). * Levitov and Lesovik (1992) L. S. Levitov and G. B. Lesovik, JETP Lett. 55, 555 (1992). * Nazarov (2007) Y. V. Nazarov, Ann. Phys. (Leipzig) 16, 720 (2007). * Pekola (2004) J. P. Pekola, Phys. Rev. Lett. 93, 206601 (2004). * Ankerhold (2007) J. Ankerhold, Phys. Rev. Lett. 98, 036601 (2007). * Sukhorukov and Jordan (2007) E. V. Sukhorukov and A. N. Jordan, Phys. Rev. Lett. 98, 136803 (2007). * Grabert (2008) H. Grabert, Phys. Rev. B 77, 205315 (2008). * Timofeev et al. (2007) A. V. Timofeev et al., Phys. Rev. Lett. 98, 207001 (2007). * Huard et al. (2007) B. Huard et al., Ann. Phys. (Leipzig) 16, 736 (2007). * Novotný (2008) T. Novotný, arxiv:0807.0387 (2008). * Sukhorukov and Jordan (2008) E. V. Sukhorukov and A. N. Jordan, arXiv.org:0807.2675 (2008). * Freidlin and Wentzell (1998) M. I. Freidlin and A. D. Wentzell, _Random Perturbations of Dynamical Systems_ (Springer-Verlag, New York, 1998), 2nd ed. * Smelyanskiy et al. (1997) V. N. Smelyanskiy et al., Phys. Rev. Lett. 79, 3113 (1997). * Dykman et al. (1997) M. I. Dykman et al., 79, 1178 (1997). * Kramers (1940) H. Kramers, Physica (Utrecht) 7, 284 (1940). * Dykman (1990) M. I. Dykman, Phys. Rev. A 42, 2020 (1990). * Dykman and Smelyanskiy (1998) M. I. Dykman and V. N. Smelyanskiy, Superlattices and Microstructures 23, 495 (1998). * Assaf et al. (2008) M. Assaf, A. Kamenev, and B. Meerson, arxiv:0807.4812 (2008). * Dykman et al. (2008) M. I. Dykman, I. B. Schwartz, and A. S. Landsman, Phys. Rev. Lett. 101, 078101 (2008). * Feynman and Hibbs (1965) R. P. Feynman and A. R. Hibbs, _Quantum Mechanics and Path Integrals_ (McGraw-Hill, New-York, 1965). * McKane (1989) A. J. McKane, Phys. Rev. A 40, 4050 (1989). * Luciani and Verga (1987) J. Luciani and A. Verga, Europhys. Lett. 4, 255 (1987). * (33) E. G. D. Cohen, private communication.	arxiv-papers	2008-08-26T15:40:25	2024-09-04T02:48:57.463596	{ "license": "Public Domain", "authors": "Lora Billings, Mark I. Dykman, and Ira B. Schwartz", "submitter": "Ira Schwartz", "url": "https://arxiv.org/abs/0808.3497" }
0808.3525	fbenadda@nyit.edu # Mass and Total Energy of Moving Bodies In a Quantified Expansion Fayçal Ben Adda New York Institute of Technology ###### Abstract In this paper, we present a new formulation of Lorentz transformations using a metric that quantifies the space expansion. As a consequence, we sort out that the limiting velocity of moving bodies is decreasing together with the space expansion. A new adjustment of relativistic laws is added to incorporate the non static nature of space-time. The conservation of the physical laws at each step of the quantified expansion allows the obtaining of new formalisms for the physical laws, in particular when an object starts moving under any force, its total energy, momentum and mass are directly affected by the expansion of the space. An example of inelastic collision is studied and several conclusions derived, specially the example of fission of atoms leads to clear correlation between liberated energy and universe expansion, it turns out that the liberated energy is increasing together with the universe expansion. ## 1 Introduction It is known that the special relativity theory is the physical theory of measurement in inertial frame of reference proposed by Albert Einstein in 1905, after substantial contributions of Hendrik Lorentz and Henri Poincaré ([6],[7]). This theory gave us $E=mc^{2}$ which led to the atomic bomb and unlocked the secret of the stars. At that time the expansion nature of the universe was not discovered yet, and this information was not included in the formulation of the special relativity theory. In 1915 Albert Einstein proposed the general relativity theory, in which the expansion of the universe was not the main objective. He proposed that gravity, as well as motion, can affect the intervals of time and of space. The general relativity theory gave us: The space warps, the Big Bang, and Black Holes. The metric $ds^{2}=c^{2}dt^{2}-a(t)(dx^{2}+dy^{2}+dz^{2})$ of the space-time proposed in general relativity does not allow the incorporation of the expansion in the the Lorentz transformation (since the increasing function of time $a(t)$, that characterizes the expansion nature of the space, omits the linearity in the considered transformations ), and then it does not permit to perceive if the laws of nature proposed by the special relativity are affected or not together with the universe expansion. Many essays have been achieved in this purpose in the last hundred years, which can be found easily in the literature. The only cited papers in this article are those who have a clear and direct relationship with this kind of approach. Given that the expansion was studied in the general relativity theory, it seems counter nature to discuss universe expansion in the special relativity theory. However, discussing the effect of the expansion on the laws of nature in the special relativity is different, and might entail some disturbing new concepts which will challenge deeply our deep-rooted understanding of the relativistic laws of nature. In this paper we will explore the incorporation of a special expansion in the special relativity theory that preserves the notion of inertial frame, and we will rewrite and interpret the relativistic laws of nature in the light of this special expansion. ### 1.1 Plan In this article we introduce a practical quantification of the space expansion in our universe using a metric with a numerical product that quantifies the expansion of the space step by step (we will call it universe expansion), and maintains the notion of inertial frame on it. In the second section we rewrite the Lorentz transformation equations using the new metric to obtain a family of Lorentz transformation equations $(T_{n})_{n\geq 0}$, where each transformation $T_{n}$ represents the Lorentz transformation equations at the step n of the space expansion. This family of transformation leads to a decreasing limiting velocity of moving bodies from one step to another, and then the physical laws are affected by the universe expansion, followed by conclusions derived concerning energy, masses, and momentum in an expanding space. In section three, the composition of velocity is derived using $T_{n}$-Lorentz transformation equations to discuss the invariance of the new limiting velocities of matter. The rest mass of photon is then discussed in section forth and new quantum formalism is introduced. We conclude this section by discussing an inelastic collision in which we find that the conversion of the rest-energy into kinetic energy is affected by the universe expansion as well as the fact that the total liberated energy by fission of atoms is increasing naturally together with the universe expansion. We conclude by some comments related to the notion of time and proper time. ### 1.2 Expansion quantification From the rate at which galaxies are receding from each other in our universe, it can be deduced that all galaxies must have been very close to each other at the same time in the past. Meanwhile galaxies are receding from each other what really happen to the physical laws, are they affected by the process of expansion or not? In the purpose to investigate the status of the physical laws together with the universe expansion we propose the following quantification of the universe movement: Suppose that we can quantify the whole universe movement, from the beginning of the universe expansion where matter was together until present time with current picture of matter distribution. Suppose that we can subdivide the whole process of expansion from the beginning until present time into n steps, in which we approximate the rate of change of the distance between any two separated events from one step to another together with the universe expansion as follow: $\bf{\mathcal{Q}}antification$ ${\mathcal{Q}}$: if the distance between two events at the step n is equal to $L_{n}$, and at the step (n+1) is equal to $L_{n+1}$ then $L_{n+1}=a_{n+1}L_{n}$ for all $n\geq 0$, where $a_{n}$ is a sequence such that $a_{0}=1$, $a_{n}>1$ $\forall n\geq 1$, and $\prod_{i=0}^{n}a_{i}$ converges. For this quantification the step 0 represents the beginning of the expansion of our universe (once $L_{0}\neq 0$) and the space-time is defined as the set of events together with the notion of squared interval $S_{0}^{2}$ defined between any two events. An event $A$ in a reference frame $\mathcal{S}$ is fixed by four coordinates $x_{0}$, $y_{0}$, $z_{0}$ and $t_{0}$, where $(x_{0},y_{0},z_{0})$ are the spatial coordinates of the point in space when the event in question occurs at the step 0, and where $t_{0}$ fixes the instant when this event occurs. Another event $B$ will be described within the same reference frame by four different coordinates $x_{0}+\Delta x_{0}$, $y_{0}+\Delta y_{0}$, $z_{0}+\Delta z_{0}$, and $t_{0}+\Delta t_{0}$. The points in space where these two events occur are separated by the distance $L_{0}$ given by $L_{0}^{2}=\Delta^{2}x_{0}+\Delta^{2}y_{0}+\Delta^{2}z_{0}$. The moments in time when these two events occur are separated at the step 0 by a time interval given by $\Delta t_{0}$. The squared interval $S_{0}^{2}$ between these two events is given as function of these quantities, by definition, through the generalization of Pythagorean Theorem $S_{0}^{2}=c^{2}\Delta^{2}t_{0}-L_{0}^{2}=c^{2}\Delta^{2}t_{0}-(\Delta^{2}x_{0}+\Delta^{2}y_{0}+\Delta^{2}z_{0})$ (1) where $c$ is the maximum speed of signal propagation at the beginning of the universe expansion (the maximum speed for the transmission of information). At the step 1 of the universe expansion, the event $A$ in the reference frame $\mathcal{S}$ is fixed by four new coordinates $(x_{1},y_{1},z_{1})$ and $t_{1}$, where $(x_{1},y_{1},z_{1})$ are the spatial coordinates of the point in space when the event in question occurs at the step 1, and where $t_{1}$ fixes the instant when this event occurs, meanwhile the event $B$ is described within the same reference frame by four different coordinates $x_{1}+\Delta x_{1}$, $y_{1}+\Delta y_{1}$, $z_{1}+\Delta z_{1}$, and $t_{1}+\Delta t_{1}$. The points in space where these two events occur are separated by the distance $L_{1}$ given by $L_{1}^{2}=\Delta^{2}x_{1}+\Delta^{2}y_{1}+\Delta^{2}z_{1}$. While the universe expands from the step 0 to the step 1 at a rate such that $\Delta x_{1}=a_{1}\Delta x_{0}$, $\Delta y_{1}=a_{1}\Delta y_{0}$ and $\Delta z_{1}=a_{1}\Delta z_{0}$, then we have $L_{1}=a_{1}L_{0}$. The moments in time when these two events occur are separated at the step 1 by a time interval given by $\Delta t_{1}$. The squared interval $S_{1}^{2}$ between these two events is given as function of these quantities, by definition, through the generalization of Pythagorean Theorem $S_{1}^{2}=c^{2}\Delta^{2}t_{1}-L_{1}^{2}=c^{2}\Delta^{2}t_{1}-a_{1}^{2}\Big{(}\Delta^{2}x_{0}+\Delta^{2}y_{0}+\Delta^{2}z_{0}\Big{)}$ (2) where $c$ is the maximum speed of signal propagation at the beginning of the universe expansion, etc.. At the step n of the universe expansion, the event $A$ in the reference frame $\mathcal{S}$ is fixed by four coordinates $x_{n}$, $y_{n}$, $z_{n}$ and $t_{n}$, where $(x_{n},y_{n},z_{n})$ are the spatial coordinates of the point in space when the event in question occurs at the step n, and where $t_{n}$ fixes the instant when this event occurs, meanwhile the event $B$ is described within the same reference frame by four different coordinates $x_{n}+\Delta x_{n}$, $y_{n}+\Delta y_{n}$, $z_{n}+\Delta z_{n}$, and $t_{n}+\Delta t_{n}$. The points in space where these two events occur are separated by the distance $L_{n}$ given by $L_{n}^{2}=\Delta^{2}x_{n}+\Delta^{2}y_{n}+\Delta^{2}z_{n}$. While the universe expands from the step (n-1) to the step n at a rate such that $\Delta x_{n}=a_{n}\Delta x_{n-1}$, $\Delta y_{n}=a_{n}\Delta y_{n-1}$ and $\Delta z_{n}=a_{n}\Delta z_{n-1}$, then we have $L_{n}=a_{n}L_{n-1}=a_{n}a_{n-1}L_{n-2}=\dots=(\prod_{i=1}^{n}a_{i})L_{0}.$ (3) The moments in time when these two events occur are separated at the step n by a time interval given by $\Delta t_{n}$. The squared interval $S_{n}^{2}$ between these two events is given as function of these quantities, by definition, through the generalization of Pythagorean Theorem $S_{n}^{2}=c^{2}\Delta^{2}t_{n}-L_{n}^{2}=c^{2}\Delta^{2}t_{n}-(\prod_{i=1}^{n}a_{i})^{2}L_{0}$ (4) which gives $S_{n}^{2}=c^{2}\Delta^{2}t_{n}-(\prod_{i=1}^{n}a_{i})^{2}\Big{(}\Delta^{2}x_{0}+\Delta^{2}y_{0}+\Delta^{2}z_{0}\Big{)}$ (5) where $c$ is the maximum speed of signal propagation at the beginning of the universe expansion. The above metrics, from the beginning of the universe expansion (step 0) until the present time, quantify the expansion of the space-time step by step. Indeed, a ball ${\mathcal{B}}_{0}(0,r)$ defined by $x_{0}^{2}+y_{0}^{2}+z_{0}^{2}\leq r^{2}$ (6) will increase if its radius increases, and we have for $a_{i}>1,\quad i=1,2,..,n$, $\forall n\in\hbox{\tenbb N}$, ${\mathcal{B}}_{0}\Big{(}0,r\Big{)}\subset{\mathcal{B}}_{1}\Big{(}0,(a_{1})r\Big{)}\subset\dots\subset{\mathcal{B}}_{n}\Big{(}0,(\prod_{i=0}^{n}a_{i})r\Big{)}.$ (7) However, for all $n\in\hbox{\tenbb N}$, the sphere ${\mathcal{S}}_{n}(0,(\prod_{i=0}^{n}a_{i})r)\subset{\mathcal{B}}_{n}(0,(\prod_{i=0}^{n}a_{i})r),$ is an expanding sphere given by $x_{n}^{2}+y_{n}^{2}+z_{n}^{2}=(\prod_{i=0}^{n}a_{i}^{2})r^{2}=(\prod_{i=0}^{n}a_{i}^{2})(x_{0}^{2}+y_{0}^{2}+z_{0}^{2})$ (8) where $\prod_{i=0}^{n}a_{i}>1$ is an increasing and bounded product. The infinitesimal space-time interval at the step n, $\forall n\in\hbox{\tenbb N}$, between two neighboring events within a locally inertial frame metric takes then the form $d\sigma^{2}_{n}=c^{2}dt_{n}^{2}-\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}(dx^{2}+dy^{2}+dz^{2}),$ (9) in which $n$ represents the expansion step of the physical universe. In such locally inertial reference frame the laws of the special relativity apply at each step. We propose to rewrite the physical laws in each space-time step by step to investigate any possible evolution in the laws of nature. The metric (9) can be written as $d\sigma^{2}_{n}=c^{2}dt_{n}^{2}-(dx_{n}^{2}+dy_{n}^{2}+dz_{n}^{2}),$ (10) with $dx_{n}=(\prod_{i=0}^{n}a_{i})dx_{0}$, $dy_{n}=(\prod_{i=0}^{n}a_{i})dy_{0}$, and $dz_{n}=(\prod_{i=0}^{n}a_{i})dz_{0}$. ## 2 Lorentz transformations equations Let us consider an expanding space $\mathcal{E}$ in which the proper distance between any distant points is increasing following the above quantification $\mathcal{Q}$ such that at small scale as well as at the subgalactic scale the fundamental forces of nature maintain the form of matter we have today. Let us consider that the expansion is totally modeled by an adequate numerical quantification represented by the product $(\prod_{i=0}^{n}a_{i})$ introduced above (that is to say if the position of a body at rest or the location of an event is expressed locally via three space coordinates $(x,y,z)$ in an inertial frame at the step 0, then at the step 1 its location at rest is expressed in the same frame by $(a_{1}x,a_{1}y,a_{1}z)$, at the step 2 is expressed in the same frame by $((a_{1}a_{2})x,(a_{1}a_{2})y,(a_{1}a_{2})z)$, and at the step n its location at rest is expressed by $((\prod_{i=0}^{n}a_{i})x,(\prod_{i=0}^{n}a_{i})y,(\prod_{i=0}^{n}a_{i})z)$, where n represents the expansion step of the space movement quantification). Let us consider two Cartesian frames $S(x,y,z,t)$ and $S^{\prime}(x^{\prime},y^{\prime},z^{\prime},t^{\prime})$ in $\mathcal{E}$ such that at each expansion step the equations of Newtonian mechanics hold good. Suppose that an event has Cartesian coordinates $(x,y,z,t)$ relative to $S$, and $(x^{\prime},y^{\prime},z^{\prime},t^{\prime})$ relative to $S^{\prime}$ at the same step of expansion. In general for a linear transformation of the Cartesian frames we have $\left\\{\begin{array}[]{ll}x^{\prime}=&a_{xx}x+a_{xy}y+a_{xz}z+a_{xt}t\\\ y^{\prime}=&a_{yx}x+a_{yy}y+a_{yz}z+a_{yt}t\\\ z^{\prime}=&a_{zx}x+a_{zy}y+a_{zz}z+a_{zt}t\\\ t^{\prime}=&a_{tx}x+a_{ty}y+a_{tz}z+a_{tt}t\end{array}\right.$ (11) where the coefficients $a_{ij}$ depend on the movement. The general properties of homogeneity and isotropy of the space with the choice that the frame $S^{\prime}$ moves in the x direction of $S$ with uniform velocity $v$ (without losing generality), such that the corresponding axes of $S$ and $S^{\prime}$ remain parallel throughout the motion having coincided at $t=t^{\prime}=0$ will give $\left\\{\begin{array}[]{ll}x^{\prime}=&a_{xx}(x-vt)\\\ y^{\prime}=&a_{yy}y\\\ z^{\prime}=&a_{zz}z\\\ t^{\prime}=&a_{tx}x+a_{tt}t\end{array}\right.$ (12) Suppose that a light signal is emitted from the common center of the two frames at $t=t^{\prime}=0$. Since the space $\mathcal{E}$ is expanding, then the observer in the frame $S$ will observe that the light travels in all direction in an expanding sphere, at the step n, given by $\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}(x^{2}+y^{2}+z^{2})=c^{2}t^{2}$ (13) and from the frame $S^{\prime}$, the observer perceives the same $\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}(x^{\prime 2}+y^{\prime 2}+z^{\prime 2})=c^{2}t^{\prime 2}$ (14) and we have $\displaystyle\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}(x^{2}+y^{2}+z^{2})-c^{2}t^{2}=\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}(x^{\prime 2}+y^{\prime 2}+z^{\prime 2})-c^{2}t^{\prime 2}.$ (15) Substituting formula (12) in (15) gives $\prod_{i=0}^{n}a_{i}^{2}(x^{2}+y^{2}+z^{2})-c^{2}t^{2}=\prod_{i=0}^{n}a_{i}^{2}\Big{(}(a_{xx}(x-vt))^{2}+(a_{yy}y)^{2}+(a_{zz}z)^{2}\Big{)}-c^{2}(a_{tx}x+a_{tt}t)^{2}$, which yields the following system $\left\\{\begin{array}[]{ll}\prod_{i=0}^{n}a_{i}^{2}\ =&(\prod_{i=0}^{n}a_{i}^{2})\ a_{xx}^{2}-c^{2}a_{tx}^{2}\\\ \quad 1\qquad=&a_{yy}^{2}\\\ \quad 1\qquad=&a_{zz}^{2}\\\ -c^{2}\qquad=&(\prod_{i=0}^{n}a_{i}^{2})\ a_{xx}^{2}v^{2}-c^{2}a_{tt}^{2}\\\ \quad 0\qquad=&(\prod_{i=0}^{n}a_{i}^{2})\ a_{xx}^{2}v+c^{2}a_{tx}a_{tt}.\end{array}\right.$ (16) The resolution of the system (16) leads to the Lorentz transformation equations $T_{n}$ with a new limiting velocity $T_{n}:\left\\{\begin{array}[]{ll}x^{\prime}=&\displaystyle{x-vt\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}\\\ y^{\prime}=&y\\\ z^{\prime}=&z\\\ t^{\prime}=&\displaystyle{t-(\prod_{i=0}^{n}a_{i}^{2}){vx\over c^{2}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}\end{array}\right.$ (17) ### 2.1 First consequence: limiting velocity The first immediate consequence from the Lorentz transformations in an expanding universe is the following: There exists a limiting velocity for moving bodies given by $v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}},$ (18) which depends on the fossil light velocity (velocity of light at the beginning of the universe expansion) and the expanding parameter of the universe. Indeed, $\prod_{i=1}^{n}a_{i}$ is an increasing and bounded product greater than one, then the limiting velocity $v_{l_{n}}$ is decreasing along with the space expansion of the universe. It turns out that the limiting velocity was maximal at the beginning of the expansion of the universe (step 0) and was equal to $c$, that is why we call it the fossil velocity of light. Thus the limiting velocity of moving body, including the light, will decrease in an expanding universe. However, the physical meaning of the equations (17) obtained in respect to moving rigid bodies and moving clocks remain the same as for the classical Lorentz transformations (deformability and lose of synchronization, time transformations, etc) except the fact that the expansion of the universe will be involved in all formulation of the physical laws as we will see later on. ### 2.2 Other Immediate consequences #### 2.2.1 Mass and Total Energy As a consequence of these new transformations and taking into account the dynamic of the universe, we propose to adjust the Einstein’s equation relative to the mass-energy as follow: The total energy of moving body at the step n of the space expansion is given by the law $E_{n}(v)=m_{n}(v)\ \Big{(}{c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2}$ (19) and the body’s rest-mass energy at the step n of the space expansion is given by the law $E_{n}(0)=m_{0}\ \Big{(}{c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2},$ (20) where $m_{0}=m_{n}(0)$ represents the mass of a body at rest relative to the observer in an expanding universe at the step n. The relative mass at the step n is given by the law $m_{n}(v)={m_{0}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}.$ (21) Indeed, using the Newtonian formula $\displaystyle{dE_{n}(v)\over dt}=F.v$, the equation (19), and the fact that the Newton’s law remains valid if it is written as $F={d\Big{(}m_{n}(v)\ v\Big{)}\over dt}$ (22) give the following: ${d\Big{(}m_{n}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}\over dt}={d(m_{n}\ v)\over dt}v={d(m_{n}\ v_{x})\over dt}\ \ v_{x}+{d(m_{n}\ v_{y})\over dt}\ \ v_{y}+{d(m_{n}\ v_{z})\over dt}\ \ v_{z},$ (23) on the other hand we have ${d\Big{(}m_{n}\ {c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2}\over dt}={2m_{n}\ c\over\prod_{i=0}^{n}a_{i}}\ {d\Big{(}m_{n}\ {c\over\prod_{i=0}^{n}a_{i}}\Big{)}\over dt}=2m_{n}\ {d\Big{(}m_{n}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}\over dt}$ (24) and ${d(m_{n}\ v_{s})^{2}\over dt}=2m_{n}\ v_{s}{d(m_{n}\ v_{s})\over dt}\qquad\hbox{for}\quad s=x,\ y,\ z,$ (25) which yields in (23) $\displaystyle{1\over 2m_{n}}{d\Big{(}m_{n}^{2}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}\over dt}={1\over 2m_{n}}{d(m_{n}\ v_{x})^{2}\over dt}+{1\over 2m_{n}}{d(m_{n}\ v_{y})^{2}\over dt}+{1\over 2m_{n}}{d(m_{n}\ v_{y})^{2}\over dt},$ (26) then ${d\Big{(}m_{n}^{2}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}\over dt}={d\Big{(}m_{n}^{2}(v_{x}^{2}+v_{y}^{2}+v_{z}^{2})\Big{)}\over dt}={d(m_{n}^{2}\ v^{2})\over dt}$ (27) with $v^{2}=v_{x}^{2}+v_{y}^{2}+v_{z}^{2}$. By integration we obtain $m_{n}^{2}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}=m_{n}^{2}\ v^{2}+k.$ (28) For $v=0$, and $m_{n}(0)=m_{0}$, we obtain $k=m^{2}_{0}\ \displaystyle{c^{2}\over\prod_{i=0}^{n}a_{i}^{2}},$ and then $m_{n}^{2}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}=m_{n}^{2}\ v^{2}+m^{2}_{0}\ {c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}$ (29) which gives equation (21). On the other hand the law (21) leads to the law (19), indeed, the work of displacing a body by a distance $dl$ at the step n is $Fdl={dP_{n}(v)\over dt}\ dl=dP_{n}(v)\ {dl\over dt}=v\ dP_{n}(v)$ (30) where $v$ is the velocity vector. If this work serves to increase the energy $E_{n}(v)$ of the body at the step n, then $dE_{n}(v)=v\ dP_{n}(v)=v\ d\Big{(}m_{n}(v)\ v\Big{)},$ (31) then $dE_{n}(v)=m_{n}(v)\ v\ dv+v^{2}\ dm_{n}(v).$ (32) Since $m_{n}(v)$ is given by (21), we have $dm_{n}(v)=\displaystyle{m_{n}(v)\ (\prod_{i=0}^{n}a_{i}^{2})\ v\ dv\over c^{2}\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}},$ (33) by substitution we obtain $dE_{n}(v)=m_{n}(v)\ v\ dv+\displaystyle{m_{n}(v)\ (\prod_{i=0}^{n}a_{i}^{2})\ v^{3}dv\over c^{2}\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}}.$ (34) then $dE_{n}(v)={m_{n}(v)\ v\ dv\ \over\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}}.$ (35) which gives together with formula (33) $dE_{n}(v)={c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}\ dm_{n}(v).$ (36) that is to say the law (19). The universe dynamic does not affect the mass of a given object at rest. However, object at rest has a rest energy $E_{n}(0)$ given by (20) in which the universe movement is manifested. When the object starts moving under any force, its total energy, momentum and mass are directly affected by the expansion parameters of the universe. If the speed $c$ is the fossil velocity of light, then the term $\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$ in laws (19), (20), and (21) represents the current experimental velocity of light, of course if we consider that the present time corresponds to the step n of the universe expansion, and it constitutes a limiting velocity for any motion or transfer of interaction at the step n of the universe expansion. #### 2.2.2 The relativistic kinetic energy For small values of $\displaystyle(\prod_{i=0}^{n}a_{i}){v\over c}$, the total energy (19) goes over into the classical expression of the kinetic energy as shown by $E_{n}(v)={m_{0}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}\ \Big{(}{c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2}$ (37) which gives $E_{n}(v)\simeq m_{0}\Big{(}{c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2}+{1\over 2}m_{0}v^{2}+\dots$ (38) The first term represents $E_{n}(0)$ the energy at rest, and the second term represents the Newtonian kinetic energy. The relativistic kinetic energy $E_{c_{n}}$ or the motion energy at the expansion step n can be defined by subtracting the rest energy $E_{n}(0)$ from the total energy $E_{n}(v)$ $E_{c_{n}}=E_{n}(v)-E_{n}(0)$ (39) the substitution of formula (19) and (20) in the last equality gives $E_{c_{n}}=m_{0}\Big{(}{c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2}\Big{(}{1\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}-1\Big{)}$ (40) and using relative mass (21) we obtain $E_{c_{n}}=(m(v)-m_{0})\Big{(}{c\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2}$ (41) The kinetic energy of a body is not only related to the increase of masses when they are moving as asserted by the special relativity but also related to the universe expansion. #### 2.2.3 The relativistic momentum and energy The use of equation (21) in the equation of motion (Force is equal to the rate of change of momentum) remains valid. However, here the momentum is now at the step n given by: $P_{n}(v)=m_{n}(v)\ v={m_{0}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}\ v.$ (42) The equation $F=\displaystyle{dP_{n}(v)\over dt}$ determines the motion of a body acted on by any force, but now correctly takes expanding effects into account. Clearly, the form of the momentum and force equation is very similar in relativity theory and in our approach; the effective mass $m_{n}(v)$ depends on the speed of motion of the body relative to the observer and the n-th step expansion parameter $\prod_{i=0}^{n}a_{i}$ according to the formula (21), while in relativity theory it is independent of the universe expansion. The crucial feature is that the effective mass $m_{n}(v)$ diverges as $v$ tends to the limiting velocity $v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$ and so the momentum $P_{n}(v)=m_{n}(v)v$ also diverges. As for relativity theory, the energy and momentum are related. Indeed, using formula (19) and (42) we obtain the following relation $P_{n}(v)\ c=E_{n}(v)\ {v\over c}\ \prod_{i=0}^{n}a_{i}^{2}.$ (43) ## 3 The composition of velocities Like for special relativity we will use the classical definition of velocity. If we call $\vec{u}$ the velocity of an object in the frame $S$, its components with respect to the three axes are $u_{x}={dx\over dt},\ u_{y}={dy\over dt},\ u_{z}={dz\over dt}$. The same velocity is measured in the frame $S^{\prime}$ as $\vec{u^{\prime}}$ with components given by $u^{\prime}_{x}={dx^{\prime}\over dt^{\prime}},\ u^{\prime}_{y}={dy^{\prime}\over dt^{\prime}},\ u^{\prime}_{z}={dz^{\prime}\over dt^{\prime}}.$ For simplicity and without losing generality, we will consider only that the velocity $\vec{u}$ is parallel to the x-axes. A simple differentiation of the equations of Lorentz transformations $T_{n}$ gives $\left\\{\begin{array}[]{ll}dx^{\prime}=&\displaystyle{dx- vdt\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}\\\ dy^{\prime}=&dy\\\ dz^{\prime}=&dz\\\ dt^{\prime}=&\displaystyle{dt-(\prod_{i=0}^{n}a_{i}^{2}){vdx\over c^{2}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}},\end{array}\right.$ (44) then we have $\left\\{\begin{array}[]{ll}\displaystyle{dx^{\prime}\over dt^{\prime}}=&\displaystyle{dx-vdt\over dt-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}dx}\\\ \displaystyle{dy^{\prime}\over dt^{\prime}}=&\displaystyle{dy\over dt-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}dx}\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}^{1\over 2}\\\ \displaystyle{dz^{\prime}\over dt^{\prime}}=&\displaystyle{dz\over dt-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}dx}\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}^{1\over 2}\\\ \end{array}\right.$ (45) which gives $\left\\{\begin{array}[]{ll}u^{\prime}_{x}=&\displaystyle{u_{x}-v\over 1-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}u_{x}}\\\ u^{\prime}_{y}=&\displaystyle{u_{y}\over 1-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}u_{x}}\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}^{1\over 2}\\\ u^{\prime}_{z}=&\displaystyle{u_{z}\over 1-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}u_{x}}\Big{(}1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{)}^{1\over 2}.\\\ \end{array}\right.$ (46) Since $\vec{u}$ is parallel to the x-axes, then the law of composition of velocity is given by $u^{\prime}={dx^{\prime}\over dt^{\prime}}={u-v\over 1-(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}u}$ (47) and inversely $u={dx\over dt}={u^{\prime}+v\over 1+(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}u^{\prime}}$ (48) which has the form of the relativist law of composition of velocity with the appearance of the term $(\prod_{i=0}^{n}a_{i})$ that characterizes the expansion of the universe at the step n. In the extreme case where an observer in the frame $S^{\prime}$ is emitting light signals in the direction of x-axis, since the frame is in an expanding universe at the step n, then the velocity of these signals in the frame $S^{\prime}$ is given by $u^{\prime}=v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$ which gives $u={v_{l_{n}}+v\over 1+(\prod_{i=0}^{n}a_{i}^{2}){v\over c^{2}}v_{l_{n}}}={{c\over\prod_{i=0}^{n}a_{i}}+v\over 1+(\prod_{i=0}^{n}a_{i}){v\over c}}$ (49) which gives $u={c\over\prod_{i=0}^{n}a_{i}}=v_{l_{n}}$ and then the limiting velocity of light $v_{l_{n}}$ at the step n is invariant, that is to say the velocity of light ”measured” in a moving frame at the step n, appears to be equal to $\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$ in any direction. The speed of light in vacuum, at the step n of the universe expansion, will be found the same (equal to $v_{l_{n}}$) by any two observers in uniform relative motion and this is true for the whole process of universe expansion, meanwhile this speed is decreasing together with the universe expansion. For $n=0$ the Lorentz transformation $T_{0}$ gives $c$ as an invariant limiting velocity of matter, meanwhile $T_{1}$ gives $\displaystyle{c\over a_{1}}$ as the invariant limiting velocity of matter, and so on, until the Lorentz transformation $T_{n}$ which gives $\displaystyle{c\over(\prod_{i=0}^{n}a_{i})}$ as the invariant limiting velocity of matter. It turns out that the limiting velocity of moving bodies is decreasing together with the universe expansion and that the velocity of light in empty space is independent of the velocity of its source at each step of the universe expansion despite of its decreasing nature together with the universe expansion. In the whole process of universe expansion the velocity of light remains independent of the reference frame of the observer. The formula (19) can be justified exactly as the famous formula $E=mc^{2}$ (Einstein [4]), by using the new composition of velocity provided by the Lorentz transformation $T_{n}$ and the momentum conservation at the step n. ## 4 Lorentz transformations and rest-mass of photon The quantification of the expansion introduced above leads us to set up different steps-Lorentz transformations $(T_{n})_{n\geq 0}$ that permit to quantify the physical laws evolution together with the universe expansion, and derive a relationship between the space-time views obtained by any two observers in an expanding universe. The result obtained unifies the fundamental special-relativity results of time dilation, length contraction, and the relativity of simultaneity, in a single relation for different steps of quantified expansion. However, from these different Lorentz transformations, we sort out that the limiting velocity $v_{l_{n}}$ of matter is decreasing from one step to another including the velocity of light. The velocity of light at the beginning of the expansion of the universe (step 0) was constant and equal to $c$ an unknown value, and the $T_{0}$-Lorentz transformation equations are given by $(T_{0})\left\\{\begin{array}[]{ll}x^{\prime}=&{x-vt\over\sqrt{1-{v^{2}\over c^{2}}}}\\\ y^{\prime}=&y\\\ z^{\prime}=&z\\\ t^{\prime}=&{t-{vx\over c^{2}}\over\sqrt{1-{v^{2}\over c^{2}}}}\end{array}\right.$ (50) this transformation $(T_{0})$ leads to the relativistic formulation of the physical laws $\left\\{\begin{array}[]{ll}E_{0}(v)&\hbox{the total energy at the step 0}\\\ E_{0}(0)&\hbox{the rest-energy at the step 0}\\\ E_{c_{o}}(v)&\hbox{the kinetic energy at the step 0}\\\ P_{0}(v)&\hbox{the momentum at the step 0 }\\\ m_{0}(v)&\hbox{the relative mass at the step 0 }\\\ m_{0}(0)=m_{0}&\hbox{the rest-mass at the step 0 }\end{array}\right.$ (51) On the other hand, at the step 1 the $T_{1}$-Lorentz transformation equations become $(T_{1})\left\\{\begin{array}[]{ll}x^{\prime}=&{x-vt\over\sqrt{1-a_{1}^{2}{v^{2}\over c^{2}}}}\\\ y^{\prime}=&y\\\ z^{\prime}=&z\\\ t^{\prime}=&{t-a_{1}^{2}{vx\over c^{2}}\over\sqrt{1-a_{1}^{2}{v^{2}\over c^{2}}}}\end{array}\right.$ (52) from which we sort out that the matter has a new limiting velocity $v_{l_{1}}=\displaystyle{c\over a_{1}}$. At this step all the equations of Newtonian mechanics hold good, and the formulations of the physical laws are obtained under some changes relative to the new limiting velocity. Similar formulations of the physical law equations (51) are introduced with the new limiting velocity and denoted by $\left\\{\begin{array}[]{ll}E_{1}(v)&\hbox{the total energy at the step 1}\\\ E_{1}(0)&\hbox{the rest-energy at the step 1}\\\ E_{c_{1}}(v)&\hbox{the kinetic energy at the step 1}\\\ P_{1}(v)&\hbox{the momentum at the step 1 }\\\ m_{1}(v)&\hbox{the relative mass at the step 1 }\\\ m_{1}(0)=m_{0}&\hbox{the rest-mass at the step 1 }\end{array}\right.$ (53) At the step n, the $T_{n}$-Lorentz transformation equations become $(T_{n})\left\\{\begin{array}[]{ll}x^{\prime}=&\displaystyle{x-vt\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}\\\ y^{\prime}=&y\\\ z^{\prime}=&z\\\ t^{\prime}=&\displaystyle{t-(\prod_{i=0}^{n}a_{i}^{2}){vx\over c^{2}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}},\end{array}\right.$ (54) in which the limiting velocity of matter reaches the value of $v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$, and where all formulations of the physical laws are possible to obtain under some changes with the hypothesis that the equations of Newtonian mechanics are valid at this step of the universe expansion (conservation of momentum, conservation of energy, etc). The new limiting velocity $v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$ at the step n induces some change in all formulations (as regards both mechanics and electro- magnetism in inertial systems), including the Maxwell equations, to obtain that the physical law remains the same under the $T_{n}$-Lorentz transformations. The formulation of the physical law equations at the step n with the new limiting velocity $v_{l_{n}}$ are denoted by $\left\\{\begin{array}[]{ll}E_{n}(v)&\hbox{the total energy at the step n}\\\ E_{n}(0)&\hbox{the rest-energy at the step n}\\\ E_{c_{n}}(v)&\hbox{the kinetic energy at the step n}\\\ P_{n}(v)&\hbox{the momentum at the step n }\\\ m_{n}(v)&\hbox{the relative mass at the step n }\\\ m_{n}(0)=m_{0}&\hbox{the rest-mass at the step n }\end{array}\right.$ (55) where the quantities $E_{n}(v)$, $E_{n}(0)$, $E_{c_{n}}(v)$, $P_{n}(v)$ and $m_{n}(v)$ take the form (19), (20), (40), (42), and (21) respectively under the $T_{n}$-Lorentz transformation for all $n\geq 0$. The Lorentz transformations $(T_{0})$, $(T_{1})$,$\dots$, and $(T_{n})$ have the same mathematical form, and all the physical quantities are affected by the space expansion from one step to another, except the rest-mass $m_{0}$ which is invariant under universe expansion. Since the light has no rest frame, and since the formalism must be the same for all matter, the quantity $m_{0}$ in the above formalism represents in reality the mass of matter which is invariant under universe expansion, we will call it the invariant mass. The family of transformation $(T_{n})_{n\geq 0}$ leads to fatal variation of the limiting velocity $v_{l_{n}}$ of moving bodies according to the universe expansion. This has interesting consequences. In the expansion process the rest energy at the step n can be evaluated compared to the rest energy at the beginning of the universe expansion. Indeed, the rest energy at the step n is given by the law $E_{n}(0)=m_{0}\Big{(}{c\over\prod_{i=}^{n}a_{i}}\Big{)}^{2}=m_{0}c^{2}\Big{(}{1\over\prod_{i=0}^{n}a_{i}}\Big{)}^{2},$ (56) and using the rest energy notation we obtain $E_{n}(0)=E_{0}(0){1\over\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}},$ (57) which means that the rest energy of matter is decreasing along with the universe expansion. However, the relative mass of matter is increasing together with the universe expansion, indeed, from the law (21) at the step n and the same law at the step n+1 we sort out that $m_{n+1}(v)\sqrt{1-(\prod_{i=0}^{n+1}a_{i}^{2}){v^{2}\over c^{2}}}=m_{n}(v)\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}$ (58) and we obtain for all $v<{c\over\prod_{i=0}^{n+1}a_{i}}$ ${m_{n+1}(v)\over m_{n}(v)}={\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n+1}a_{i}^{2}){v^{2}\over c^{2}}}}>1$ (59) which gives $m_{n+1}(v)>m_{n}(v)>\dots>m_{0}(v),$ (60) this means that the relative mass of matter in our universe is increasing along with the universe expansion as well as the momentum, meanwhile the rest energy is decreasing. It is possible to determine the evolution of energy, momentum and relative mass together with the universe expansion by using the invariant mass $m_{0}$, indeed, it is not difficult to write all these quantities at the step n function of their values at the step 0 (beginning of the expansion) and we have the law of evolution of the relative mass together with the universe expansion given by $m_{n}(v)=m_{0}(v){\sqrt{1-{v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}},$ (61) meanwhile the evolution of momentum together with the universe expansion is given by the law $P_{n}(v)=P_{0}(v){\sqrt{1-{v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}$ (62) and the total energy evolution together with the universe expansion is given by the law $E_{n}(v)=E_{0}(v){1\over\prod_{i=0}^{n}a_{i}^{2}}{\sqrt{1-{v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}.$ (63) However, the evolution of the physical laws from one step to another are given by the following formulas: $m_{n}(v)=m_{n-1}(v){\sqrt{1-(\prod_{i=0}^{n-1}a_{i}^{2}){v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}},$ (64) the evolution of the momentum from one step to another is given by $P_{n}(v)=P_{n-1}(v){\sqrt{1-(\prod_{i=0}^{n-1}a_{i}^{2}){v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}$ (65) and the evolution of the total energy from one step to another is given by the law $E_{n}(v)=E_{n-1}(v){1\over\ a_{n}^{2}}{\sqrt{1-(\prod_{i=0}^{n-1}a_{i}^{2}){v^{2}\over c^{2}}}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}.$ (66) It turns out that the concept of expansion is a real source of energy, and this will be analyzed later on with a concrete example. ### 4.1 Quantum formalism The velocity of light was constant at the beginning of the expansion of our universe and equal to $c$ (an unknown fossil value). What we measure today in our experiment and what we call the velocity of light, corresponds in reality to the velocity of light at the expansion step n, where its value is given by $\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$. The light is described in quantum mechanic as a quanta of zero mass where the relation between energy and frequency is given by the Planck-Einstein’s relation $E=h\nu$ and where the momentum is given by the relation $P={E\over c}={h\nu\over c}={h\over\lambda_{0}}$ (67) where $\lambda_{0}$ the wave length given by $\lambda_{0}=cT_{0}={c\over\nu_{0}}.$ (68) The rules of translation from corpuscular terminology to wave terminology, and vice versa are based on the fact that an electromagnetic wave of length $\lambda_{0}$ and intensity $I$ is considered as a stream of photons of frequency $\nu_{0}$ and intensity $I=Nh\nu_{0}$ where $N$ is the number of photons passing per unit time through unit areas, and the direction of motion of the wave front is the direction of motion of the photons. In the purpose to integrate the new formalism of expansion in quantum formulation we introduce the following: We denote $\lambda_{0}$ the fossil wave length of the light at the beginning of the universe expansion where the celerity of photon was $c$, with $\nu_{0}=\displaystyle{c\over\lambda_{0}}$ the fossil frequency of the light at that time. We denote $\lambda_{n}=v_{l_{n}}T_{n}=\displaystyle{{c\over\prod_{i=0}^{n}a_{i}}\over\nu_{n}}=\displaystyle{c\over\nu_{n}\prod_{i=0}^{n}a_{i}}$ (69) the current wave length of the light at the step n of the universe expansion, where the celerity of photon is $\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$, with $\nu_{n}=\displaystyle{c\over\lambda_{n}\prod_{i=0}^{n}a_{i}}$ the light frequency at the expansion step n. Hence the wave length represents a distance, therefore by using the property of the expansion of the space the wave length expands and we have $\lambda_{n}=\lambda_{0}\prod_{i=0}^{n}a_{i}$ which leads to the following relation between the fossil light frequency and the light frequency at the expansion step n $\nu_{n}={c\over\lambda_{n}\prod_{i=0}^{n}a_{i}}={c\over\lambda_{0}\prod_{i=0}^{n}a_{i}^{2}}$ (70) and then $\nu_{n}=\nu_{0}{1\over\prod_{i=0}^{n}a_{i}^{2}}.$ (71) From the relativistic relation between energy and momentum $P_{o}(v)c=E_{0}(v){v\over c}$, we sort out the momentum of photon with velocity $v=v_{l_{n}}$ given by $P_{0}(v_{l_{n}})={E_{0}(v_{l_{n}})\over c}{1\over\prod_{i=0}^{n}a_{i}}\neq 0,$ (72) thus $E_{0}(v_{l_{n}})\neq 0$ and well defined for $v=v_{l_{n}}$, then $E_{0}(v_{l_{n}})={E_{0}(0)\over\sqrt{1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}}\neq 0,$ (73) which gives that the rest energy of a photon $E_{0}(0)$ is not equal to zero. The substitution of (73) in the formula (72) gives $P_{0}(v_{l_{n}})={E_{0}(0)\over c\prod_{i=0}^{n}a_{i}}{1\over\sqrt{1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}}.$ (74) The precedent equality can be expressed using the rest energy (57) at the step n and $v_{l_{n}}$ as $P_{0}(v_{l_{n}})=\Big{(}{\prod_{i=0}^{n}a_{i}\over c}\Big{)}\Big{(}{E_{0}(0)\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}{1\over\sqrt{1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}},$ (75) therefore the energy and momentum of photon are related by the law $P_{0}(v_{l_{n}})={E_{n}(0)\over v_{l_{n}}}{1\over\sqrt{1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}}.$ (76) The law (76) leads to the following approximation $P_{0}(v_{l_{n}})\simeq{E_{n}(0)\over v_{l_{n}}}$ (77) where $v_{l_{n}}={c\over\prod_{i=0}^{n}a_{i}}$ is the limiting velocity of light measured at the step n. The relation (77) was introduced experimentally by A. H. compton in 1923, [3]. ### 4.2 Photon mass Using the formulation given by Lorentz transformation $(T_{0})$, the relative mass of moving body is given by $m_{0}(v)={m_{0}\over\sqrt{1-{v^{2}\over c^{2}}}}\qquad\hbox{for}\quad\forall v<c$ (78) which gives $m_{0}=m_{0}(v)\sqrt{1-{v^{2}\over c^{2}}}.$ (79) then $m_{0}^{2}=m_{0}^{2}(v)(1-{v^{2}\over c^{2}})\Longleftrightarrow m_{0}^{2}c^{2}=m_{0}^{2}(v)(c^{2}-v^{2}),$ (80) which gives $m_{0}^{2}c^{4}=m_{0}^{2}(v)c^{4}-m_{0}^{2}(v)c^{2}v^{2},$ (81) then for all $v<c$ we have $m_{0}^{2}c^{4}=E_{0}^{2}(v)-P_{0}^{2}(v)c^{2}.$ (82) Nevertheless, if the invariant mass $m_{0}$ of a photon was zero at the origin of the expansion when its velocity was $c$, it will remain zero at the step n when the light move at speed $v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$ because of its natural invariance under the universe expansion. If we suppose that the invariant mass of a photon $m_{0}=0$ at the step n, $\forall n>0$, (since for $n=0$ the formula (82) is not defined for $v_{l_{0}}=c$) from the universe expansion, then the speed of light is given by $v=v_{l_{n}}$, and by substitution in the the formula (82) we obtain $0=E_{0}^{2}(v_{l_{n}})-P_{0}^{2}(v_{l_{n}})c^{2}.$ (83) Using the value of $P_{0}(v_{l_{n}})$ given by formula (74), the equality (83) becomes $0=E_{0}^{2}(v_{l_{n}})-{E_{0}^{2}(0)\over{\prod_{i=0}^{n}a_{i}^{2}}}({1\over 1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}).$ (84) On the other hand we have (73), and then equality (84) gives $E_{0}^{2}(0)({1\over{1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}})={E_{0}^{2}(0)\over{\prod_{i=0}^{n}a_{i}^{2}}}({1\over 1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}),$ (85) which is equivalent to $E_{0}^{2}(0)={E_{0}^{2}(0)\over{\prod_{i=0}^{n}a_{i}^{2}}}.$ (86) Since the rest energy of a photon at the beginning of the universe expansion (step 0) was not null, then formula (86) leads to the contradiction $1={1\over{\prod_{i=0}^{n}a_{i}^{2}}}\qquad\forall n>0\quad\hbox{which is impossible},$ (87) then the invariant mass of photon $m_{0}\not=0$. Along the universe expansion the invariant mass $m_{0}$ of matter remains the same, meanwhile the associated rest energy decreases ($E_{0}(0)>E_{1}(0)>\dots>E_{n}(0)$), and the momentum of matter increases. If the rest energy of photon was zero, it can not decrease. Since $\forall v<c$, $E_{0}(v)\sqrt{1-{v^{2}\over c^{2}}}=m_{0}c^{2}$, then we can evaluate the invariant mass $m_{0}$ for a photon of velocity $v=v_{l_{n}}$. Indeed, we have $m_{0}={{E_{0}(v_{l_{n}})\over c^{2}}{\sqrt{1-{1\over\prod_{i=0}^{n}a_{i}^{2}}}}},$ (88) the use of the formula (73) in the precedent equality gives $m_{0}={E_{0}(0)\over c^{2}}$ (89) and we have $m_{0}={E_{0}(0)\over c^{2}}=\Big{(}{\prod_{i=0}^{n}a_{i}^{2}\over c^{2}}\Big{)}\Big{(}{E_{0}(0)\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}$ (90) using (57) and the velocity of light $v_{l_{n}}$ at the step n to obtain $m_{0}={E_{0}(0)\over c^{2}}={E_{n}(0)\over v^{2}_{l_{n}}},$ (91) which means that the ratio between rest energy and velocity of light is invariant under universe expansion. The formula (89) means that the light has non zero invariant mass. However, the invariant mass of a photon depends on the fossil rest energy (the rest energy at the step n is impossible to determine since the light is never at rest). We know only that the invariant mass and the rest energy of a photon are not null, they might be small, very small, but not null. ### 4.3 The conversion of the rest-energy into kinetic energy In the purpose to sort out the real nature of energy gained by matter in an expanding universe we introduce the following example of inelastic collision: Consider two identical particles which move toward each other along straight line, with equal speeds, and equal invariant mass $m_{0}$, they collide and stick together in an expanding space at the step n. The conservation of momentum at the step n gives: $m_{n}(v)\ v-m_{n}(-v)\ v=M_{n}(V)\ V$ (92) thus $0=M_{n}(V)\ V$ (93) which gives $V=0$, so the final object is at rest with mass $M_{n}(0)$. The conservation of total energy at the step n from the universe expansion gives $m_{n}(v){c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}+m_{n}(-v){c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}={M_{n}(V)c^{2}\over\prod_{i=0}^{n}a_{i}^{2}},$ (94) since the final object is at rest (that is to say $V=0$), then it follows that $m_{n}(v)+m_{n}(-v)=2m_{n}(v)=M_{n}(0),$ (95) then we obtain $M_{n}(0)={2m_{0}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}},$ (96) the mass of final object is larger than the sum of the original masses. To evaluate the nature of energy created with the increase of mass together with the universe expansion from the step 0 to the step n we have to compare the same collision at the origin of the expansion (step 0) and at the step n. Indeed the universe expansion creates the increase of moving masses as $M_{n}(0)-M_{0}(0)=\Big{(}{2m_{o}\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}-{2m_{o}\over\sqrt{1-{v^{2}\over c^{2}}}}\Big{)}.$ (97) For a small value of $(\prod_{i=0}^{n}a_{i}){v\over c}$ we can approximate (97) as $M_{n}(0)-M_{0}(0)\simeq 2m_{o}\ \Big{(}1+{1\over 2}(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}-1-{1\over 2}{v^{2}\over c^{2}}\Big{)}$ (98) which gives $M_{n}(0)-M_{0}(0)\simeq{1\over 2}(2m_{o})(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}\Big{(}1-{1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}.$ (99) The gain of masses represents a rest energy at the step n given by $\Big{(}M_{n}(0)-M_{0}(0)\Big{)}{c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}\simeq{1\over 2}(2m_{o})v^{2}\Big{(}1-{1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}.$ (100) This means that the rest energy gained by a moving particles together with the universe expansion is equal approximatively to the sum of the classical kinetic energy of the two particles factor the rate of increase of rest energy due to the universe expansion, that is to say $\Big{(}1-{1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}$. Indeed, from formula (57) it is easy to evaluate the variation of the rest energy between the beginning of the universe expansion (step 0) and the universe expansion at the step n, we obtain $E_{0}(0)-E_{n}(0)=E_{0}(0)\Big{(}1-{1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}$ (101) from which it turns out that $\Big{(}1-{1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{)}$ represents the rate of decrease of the rest energy together with the universe expansion. Since the mass of final object is larger than the sum of the original masses, we can approximate this excess of mass for a small value of $(\prod_{i=0}^{n}a_{i}){v\over c}$, and compare it to the same value at the beginning of the universe expansion. Indeed the approximation of the law (96), for all $n\geq 0$, gives $M_{n}(0)\simeq 2m_{0}+{1\over 2}(2m_{o})(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}$ (102) which gives $\Big{(}M_{n}(0)-2m_{0}\Big{)}\simeq{1\over 2}(2m_{o})(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}},$ (103) and then we obtain $\Big{(}M_{n}(0)-2m_{0}\Big{)}\simeq\Big{(}M_{0}(0)-2m_{0}\Big{)}\prod_{i=0}^{n}a_{i}^{2}.$ (104) It turns out that the excess mass of the composite object at the step n of the universe expansion compared to the mass of the same composite object at the beginning of the expansion (step 0) is increasing together with the universe expansion and then the kinetic energy brought in also increases. The lost Kinetic energy has been converted to rest energy (mass). The classical explanation for the loss of Kinetic energy attributes it to conversion into thermal energy (heat): the final object will have a higher temperature or a large internal energy. Now since the mass of the final object (after collision formula (96) will have an increasing total mass, then this means that the temperature of the final object will increase together with the universe expansion. Since many evidences prove that early earth was molten from bombardment of rocks and mini planets and others, then we can say that there must exist a natural factor that contributes to the global warming of our planet due to the universe expansion (the global warming of our planet is not only attributed to human activities!). ### 4.4 Fission of atoms in an expanding space Suppose that we have an atom of uranium with rest mass $M_{n}(0)=M$ measured in an expanding universe at the step n, and suppose that something happens so that the atom flies into two equal pieces moving with speed $v$, so that each part has a relative mass $m_{n}(v)$ at the step n . Suppose that these pieces encounter enough material to slow them up until they stop and then each part will have an invariant mass $m_{n}(0)=m_{0}$. To reach its rest position each part will give an amount of energy left in the material in some form, whatever. The left energy for one part is then given by $E_{left}=\Big{(}m_{n}(v)-m_{n}(0)\Big{)}{c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}.$ (105) The conservation of total energy in an expanding universe at the step n gives $E_{n}(0)=2E_{n}(v)\Longleftrightarrow M=2m_{n}(v),$ (106) so the total liberated energy at the step n is given by $E_{L_{n}}=2E_{left}=2\Big{(}m_{n}(v)-m_{0}\Big{)}{c^{2}\over\prod_{i=0}^{n}a_{i}^{2}},$ (107) then $E_{L_{n}}=\Big{(}M-2m_{0}\Big{)}{c^{2}\over\prod_{i=0}^{n}a_{i}^{2}},$ (108) which can be written as $E_{L_{n}}=E_{n}(v)-E_{n}(0),$ (109) where $\forall v<\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$, $E_{n}(v)=2m_{n}(v){c^{2}\over\prod_{i=0}^{n}a_{i}^{2}}$. The law given by (108) for $n=0$ was used to estimate how much energy would be liberated under fission in the atomic bomb since the mass of uranium atom was known as well as the atoms into which it splits (iodine, xenon, and so on). However, it turns out from the law (108) that the released energy when an atom of uranium undergoes fission is correlated to the universe expansion which has interesting consequences. As the universe expands we have the law (60), which gives $2(m_{n}(v)-m_{0})>2(m_{n-1}(v)-m_{0})>\dots>2(m_{0}(v)-m_{0}),$ (110) and from (108) we have $E_{L_{n}}\prod_{i=0}^{n}a_{i}^{2}>E_{L_{n-1}}\prod_{i=0}^{n-1}a_{i}^{2}>\dots>E_{L_{o}}.$ (111) Using the law (101) we obtain ${E_{L_{n}}\over E_{n}(0)}>{E_{L_{n-1}}\over E_{n-1}(0)}>\dots>{E_{L_{o}}\over E_{0}(0)},$ (112) this means that the ratio between the total liberated energy by an atom of uranium and its rest energy is increasing together with the universe expansion. The total liberated energy by an atom of uranium (109) can be approximated, indeed, we have $E_{L_{n}}=E_{n}(v)-E_{n}(0)=E_{n}(0){1\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}-E_{n}(0),$ (113) then we have $E_{L_{n}}=E_{n}(0)\Big{(}{1\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}-1\Big{)},$ (114) by using the law (57) that expresses the correlation between the rest energy at the step n and the rest energy at the step 0, we obtain $E_{L_{n}}=E_{0}(0){1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{(}{1\over\sqrt{1-(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}}-1\Big{)}.$ (115) For a small value of $(\prod_{i=0}^{n}a_{i}){v\over c}$, we obtain the following approximation $E_{L_{n}}\simeq E_{0}(0){1\over\prod_{i=0}^{n}a_{i}^{2}}\Big{(}{1\over 1-{1\over 2}(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}}-1\Big{)}.$ (116) which gives $E_{L_{n}}\simeq{E_{0}(0)\over\prod_{i=0}^{n}a_{i}^{2}}\Big{(}1+{1\over 2}(\prod_{i=0}^{n}a_{i}^{2}){v^{2}\over c^{2}}+({1\over 2})^{2}(\prod_{i=0}^{n}a_{i}^{2})^{2}{v^{4}\over c^{4}}-1\Big{)},$ (117) then $E_{L_{n}}\simeq E_{0}(0)\Big{(}{1\over 2}{v^{2}\over c^{2}}+({1\over 2})^{2}(\prod_{i=0}^{n}a_{i}^{2}){v^{4}\over c^{4}}\Big{)},$ (118) it turns out that the total liberated energy by an atom of uranium is increasing together with the universe expansion and we have for one atom of uranium the following evolution of liberated energy from one step to another $E_{L_{n}}>E_{L_{n-1}}>\dots>E_{L_{o}},$ (119) This has interesting consequences that will change our understanding of gravity, the missing masses, or the estimation of stars energy and others. The energy in our universe increases together with the universe expansion, and this excess of energy is naturally due to the recession of galaxies from each other. The increase of energy and mass in our universe compensates the increase of distance between matters and might explain why the gravity exists at long distance. Our estimation of stars energy, or galaxies masses are erroneous if we omit to take into account the movement of matter due to the universe expansion. We conclude by raising the fatal fact that there is no stationary safe distance when an atom of uranium undergoes fission in an expanding universe, since the liberated energy increases along with the universe expansion, the security of nuclear reactor must be endlessly reconsidered. It turns out that the tame of this energy is strewed with risk and peril together with the universe expansion. ### 4.5 The experimental data and the new quantification The family of transformations $(T_{0})$, $(T_{1})$, $(T_{2})$, $\dots$, $(T_{n})$ is a mathematical formulation of the Lorentz transformation that incorporates the given quantification of the space expansion. This family of transformations $(T_{i})_{i\geq 0}$ is defined for all $v<v_{l_{n}}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$, where $v_{l_{n}}$ represents the limiting velocity of moving bodies at the expansion step n. From this limiting velocity it turns out that the velocity of light is decreasing as the universe expands, and the limiting velocity $v_{l_{n}}$ was maximal only at the step 0 (in the beginning of the universe expansion $v_{l_{0}}=c$), however, at each step the velocity of light is constant in the sense that it is independent of the reference frame of the observer. All formulations and laws based on the assumption that the velocity of light in empty space is independent of the velocity of its source remain valid for matter and related physical phenomenon during the universe expansion step by step, and the relativistic formulation of the physical laws remains valid experimentally since the velocity of light at the step n is experimentally given by $\mathcal{C}_{n}=2.99792458\times 10^{8}$ which represents in this formalism $\displaystyle{c\over\prod_{i=0}^{n}a_{i}}$, that is to say $\mathcal{C}_{n}=\displaystyle{c\over\prod_{i=0}^{n}a_{i}},$ (120) where c is the fossil velocity of light (The fossil velocity of light could be detected in our space). At each step of the universe expansion the measured velocity of light remains invariant from one frame to another in uniform relative motion. The meaning of this approximation based on the universe quantification $\mathcal{Q}$ is that the velocity of light is constant locally and variable globally with respect to the universe age, and what we mean by constant locally is that the variation of the velocity of light is too small to be detected in short time with respect to the universe age. All our experimental tests and applications of the special relativity laws at present time are and remain valid since the substitution of the experimental measure of the light speed in the $T_{0}$-Lorentz transformation will represent $T_{n}$-Lorentz transformations. Indeed, The velocity of light is locally constant at the scale of normal time (using a short interval of time as unit) and globally variable at the scale of cosmical time (using a large interval of time as unit). This local and global behavior can be derived straight forward from the quantification ${\mathcal{Q}}$. Indeed, the local behavior is reached in the quantification ${\mathcal{Q}}$ if we use a big number of subdivisions: The bigger the used number $n$ of steps is the shorter the time interval of steps we obtain. Thus, as $n$ tends to infinity, we have $a_{n+1}\approx 1$ as a consequence of the convergence of the product $\prod_{i=0}^{n}a_{i}$, then $\prod_{i=0}^{n+1}a_{i}=(\prod_{i=0}^{n}a_{i})\ a_{n+1}\approx\prod_{i=0}^{n}a_{i},\qquad\forall n>A,$ (121) where $A$ is a large positive real number. Hence the equations (121) and (120) give $\mathcal{C}_{n}\approx\mathcal{C}_{n+1}\qquad\forall n>A,$ (122) therefore the velocity of light is almost constant for short period of time (locally). However, the equation (122) is not valid anymore for the large period of time. Indeed, the smaller the used number $n$ of steps is the bigger the time interval of steps we obtain, and in that case we have $a_{n+1}>1$ for all $n<A$ and $\prod_{i=0}^{n+1}a_{i}=(\prod_{i=0}^{n}a_{i})\ a_{n+1}>\prod_{i=0}^{n}a_{i},\qquad\forall n<A,$ (123) hence the equations (120) and (123) lead to the following inequality: $\mathcal{C}_{n}>\mathcal{C}_{n+1}\qquad\forall n<A,$ (124) that is to say the velocity of light is decreasing globally together with the universe expansion from one step to another. The formalism obtained within this work using a special expansion notion in the special relativity theory to study the uniform motion of the observer relative to the source is based on the validity of the special relativity theory step by step from the big bang to nowadays. which leads to the conclusion that this new formalism is an extension of the special relativity theory. #### 4.5.1 Time Space and Expansion We close this article by some comments on the notion of time. The notion of relative time as for special relativity remains valid with some adjustment, from the metric (9) is clear that time is a required coordinate to specify a physical event and that this quantity can not be defined without the notion of space and the notion of expansion, the dynamic of the universe is involved in the definition of time. The expansion of the universe affects the time and the space. If we refer to the proper time $\tau_{n}$ as the time which is measured by an observer in a reference frame at the step n of the universe expansion in which the events occur at the same spatial point, then $dx=dy=dz=0$ and the metric (9) gives $d\sigma^{2}_{n}=c^{2}d\tau_{n}^{2},$ (125) meanwhile in another inertial reference frame the same events will verify (9), which gives $c^{2}d\tau_{n}^{2}=c^{2}dt_{n}^{2}-\Big{(}\prod_{i=0}^{n}a_{i}\Big{)}^{2}(dx^{2}+dy^{2}+dz^{2}),$ (126) from which we obtain $({d\tau_{n}\over dt_{n}})^{2}=1-\prod_{i=0}^{n}a_{i}^{2}{v^{2}\over c^{2}}$ (127) where $v^{2}=({dx\over dt_{n}})^{2}+({dy\over dt_{n}})^{2}+({dz\over dt_{n}})^{2}$. It turns out that the rate of proper time for a system varies with motion through space as for special relativity, and what is new here is that the rate of the proper time varies together with the universe expansion even if $v$ is constant. If we refer to the universe proper time $\tau$ as the time which is measured by an observer in a reference frame at rest at the step n of the universe expansion in which the events occur at the same spatial point, it is not difficult to extract from the metric (9) the relationship between proper time and coordinate time $d\tau=dt_{n}\sqrt{1-\prod_{i=0}^{n}a_{i}^{2}{v^{2}\over c^{2}}}.$ (128) The time depends on the observer and the universe expanding step. What is new here is to associate space, time and expansion in the definition of event. Any change of reference or steps affects all of them. If the proper time $\tau$ of the universe is the same from one step to another and if the velocity $v$ and $c$ are constant from one step to another then it turns out from the step $n-1$ to the step $n$ that $1={dt_{n}\over dt_{n-1}}{\sqrt{1-\prod_{i=0}^{n}a_{i}^{2}{v^{2}\over c^{2}}}\over\sqrt{1-\prod_{i=0}^{n-1}a_{i}^{2}{v^{2}\over c^{2}}}}$ (129) and then ${dt_{n}\over dt_{n-1}}={\sqrt{1-\prod_{i=0}^{n-1}a_{i}^{2}{v^{2}\over c^{2}}}\over\sqrt{1-\prod_{i=0}^{n}a_{i}^{2}{v^{2}\over c^{2}}}}>1$ (130) which allows us to compare the coordinate time between successive steps as $dt_{n}>dt_{n-1}$ (131) which means that the clocks at the step n will run slower than an identical type of clocks at the previous step, the time will run slower together with the universe expansion. One more thing, the product $\prod_{i=0}^{n}a_{i}$ which allows the quantification of the universe expansion was extracted from the fractal manifold model [1]. More details about the natural construction of the expanding parameter $\prod_{i=0}^{n}a_{i}$ and the universe expansion as well as the universe shape can be found in [2]. ## References * [1] F. Ben Adda, Mathematical Model for Fractal Manifold., International journal of Pure and Applied Mathematics, 38, N 2, 159-190, (2007). * [2] F. Ben Adda, New Understanding of the Dark Energy in the Light of a New Space Time, in Proceeding of the Invisible Universe, AIP Conference Proceedings, Vol 1241, pp 487-496, (2010). * [3] A. Compton, A quantum theory of the scattering of X-rays by light elements, Physical Review, 21, 483-502,(1923). * [4] A. Einstein, Does the inertia of body depend upon its energy content?,translated from ”Ist die tr$\ddot{a}$gheit eines k$\ddot{o}$rpers von seinem energiegehalt abh$\ddot{a}$ngin?”, Annalen der physik, 17, (1905). * [5] A. Einstein, Kosmologishe betrachtungen zur allgemeinen relativit$\ddot{a}$tstheorie, Preussische Akademie Der Wissenschaften, Sitzungsberichte, 142-152, (1917). * [6] H. Poincar$\acute{e}$, Sur la dynamique de l’$\acute{e}$lectron, Comptes rendus de l’Acad$\acute{e}$mie des sciences de Paris 140, 1504-1508, (1905). * [7] H. Poincar$\acute{e}$, Sur la dynamique de l’$\acute{e}$lectron, Rendiconti del Circolo Mathematico di Palermo 21, 127-175, (1906).	arxiv-papers	2008-08-26T14:30:21	2024-09-04T02:48:57.468854	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/4.0/", "authors": "Faycal Ben Adda", "submitter": "Faycal Ben Adda", "url": "https://arxiv.org/abs/0808.3525" }
0808.3612	# DIFFICULTIES OF PRESERVING THE LEAP SECOND Serb. Astron. J. ˝ 177 (2008), 1 - 5 S. Šegan1,222Send comments to Stevo Segan, e-mail: ssegan@matf.bg.ac.yu, B. Šurlan1 and S. Vidojević1 1Department of Astronomy, Faculty of Mathematics, University of BelgradeStudentski trg 16, 11000 Belgrade, Serbia (Received: 2007; Accepted: … ) SUMMARY: In this article we examine the possibility to extend leap second extrapolation for a near future based on some periodic terms in the Earth’s rotation changes. The IERS data, covering the interval from 1962.15 to 2006.95, are analyzed. The difference $\Delta T$ is extrapolated till to 2035 and compared with the IERS extrapolated values to the 2012. It can be seen that for the interval from 2006 to 2024 only 1 leap seconds (negative) will be operated. Key words. Earth, Time ## 1 1\. The leap second The recommendations of the IAU (International Astronomical Union) were formalized by resolutions of their Commissions that the name $UTC$ (abbreviation is compromise between English Coordinated Universal Time and French Temps Universel Coordonné) was retained. $UTC$ was recommended as the basis of standard time in all countries, the time in common (civil) use. The limit of $[UT1-UTC]$ ($UT1$, Universal time) was set at $+0.950$ s, as this is the maximum difference that can be accommodated by the code format. The maximum deviation of $UTl$ from $[UTC+DUT1]$ (time correction $DUT1=UT1-UTC$) was set at $+0.100$ s. In 1974, the CCIR (Consultative Committee on International Radio, En. or Comité consultatif international pour la radio, Fr.), increased the tolerance for $[UT1-UTC]$ from $0.7$ s to $0.9$ s. The present $UTC$ system is defined by ITU-R (International Telecommunication Union – Radio) (formerly CCIR) Recommendation ITU–R TF.460–5: ”$UTC$ is the time scale maintained by the BIPM (Bureau International des Poids et Mesures, Fr. or International Bureau of Weights and Measures, En.), with assistance from the IERS (International Earth Rotation and Reference Systems Service), which forms the basis of a coordinated dissemination of standard frequencies and time signals. It corresponds exactly in rate with $TAI$ (International Atomic Time, En., Temps Atomique International, Fr.) but differs from it by an integral number of seconds. The $UTC$ scale is adjusted by the insertion or deletion of seconds (positive or negative leap seconds) to ensure approximate agreement with $UT1$.” Because the Earth’s rotation is gradually slowing down, and in addition it has both random and periodic fluctuations, it is not a uniform measure of time. The time difference $\Delta T=[ET-UT1]=[TT-UT1]$ represents the difference between the uniform scale of Ephemeris Time ($ET$) or Terrestrial Time ($TT$) and the variable scale of Universal Time ($UT1$). Before 1955, the values are given by $\Delta T=[ET-UT1]$ based on observations of the Moon. After 1955, values are given by $\Delta T=[TT-UT1]=[TAI+32^{s}.184-UT1]$ from measurements by atomic clocks as published by the BIH (Bureau International de l’Heure, Fr. or International Time Bureau, En.) and the BIPM. According to Stephenson and Morrison (1995), over the past 2700 years $\Delta T$ can be represented by a parabola of approximately the form $\Delta T=(31^{s}/\mathrm{cy}^{2})(T-1820)^{2}/(100)^{2}-20^{s}.$ where $\Delta T$ is expressed in seconds and $T$ is the year. The minimum is at the year 1820 and passes through 0 at the year 1900. Actual values of $\Delta T$ based on astronomical data may differ somewhat from this smoothed fit. For example, the value of $\Delta T$ is $32^{s}.184$ at 1958.0, the origin of $TAI$. However, no single parabola can satisfactorily represent all the observational data. The derivative of $\Delta T$ is $\Delta LOD=(0.0017\,\mathrm{s}\,\mathrm{d}^{-1}\mathrm{cy}^{-1})\frac{(T-1820)}{100}.$ which represent the change of the length of day ($LOD$) in SI seconds. Different studies implies different values. The actual value of the $LOD$ will depart from some long-term trend with short-term fluctuations (periodicity) between $\pm 3\;$ms on time scale of decades. Similarly, the insertion of leap seconds is due to the fact that the present length of the mean solar day is about $2.5\;$ms longer than a day of precisely 86400 SI seconds, as a consequence of the long-term trend, so that the Earth’s rotation runs slow with respect to atomic time. ## 2 2\. The data and model $UTC$ is kept within $0.9$ s of $UT1$ by the occasional insertion of a leap second adjustment. When the present $UTC$ system was established in 1972, the time difference $\Delta T=[TT-UT1]=[TAI+32^{s}.184-UT1]$ was equal to $42.23$ s. Thus the difference between $TAI$ and $UT1$ in 1972 was approximately 10 s. To maintain continuity with $UT1$, $UTC$ was initially set behind $TAI$ by this amount. As of January 1st, 2006 the 23 positive leap seconds have been added. Thus $UTC$ is presently behind $TAI$ by $33$ s. Figure 1 illustrates the relationships between $TAI$, $UTC$ and $UT1$ (IERS data). A least-squares fit of the difference $[UTC-TAI]$ since 1972, shown in Figure 1, implies a nearly linear increase with a slope of $(2.10+0.05)\;$ms per day. This value represents the average excess in the length of day during the past three decades and is in approximate agreement with the value computed on the basis of the long-term trend. Recent global weather conditions have contributed to a short-term change in the length of day. Decade fluctuations due to the interaction between the Earth’s core and mantle and global ocean circulation may also contribute. The model characterized by triaxial Earth’s structure, its fluid core, visco- elastic mantle and equilibrium ocean was proposed by the Vondrák (1987). Figure 1. $UTC-TAI$ as extrapolated by IERS at 2001 and 2002 compared with observed values of $UTC-TAI$ till 2006. As a contrary Yoshido and Hamano (1993) have proposed that one of the main causes of the secular variation of the geomagnetic field is length-of-day ($LOD$) variation, namely, the variation of the rotation velocity of the mantle. They developed an analytical model, in which fluctuations of the rotation velocity of the mantle induce flow in the outer core through topographic coupling at the core-mantle boundary (CMB). The flow in turn bends the toroidal field to produce a poloidal field. At periods longer than few years, extending to many tens of years, the so- called decadal variations, the sources of excitation for both $LOD$ and PM are more enigmatic. The difficulty is that at these periods other effects may be important, including visco-elastic behavior such as post-glacial rebound, and exchange of angular momentum with the fluid core. In particular, it has become common to invoke the core as the major cause of decadal $LOD$ changes. Although some climatic forcing of long period $LOD$ has been recognized (Salstein and Rosen, 1986; Eubanks, 1993), it is uncertain at what time scale air and water become less important than the core. Unfortunately, the role of the core remains largely unquantified because it is too remote to be easily observed. A further difficulty in assessing the air/water role at long periods is that the torques required to cause decadal $LOD$ variations are utterly insignificant when compared with those applied by the atmosphere at shorter time scales (Hide and Dickey, 1991). This means that atmospheric/oceanic torques of geodetic significance are of second-order importance in general circulation studies. Quantification of momentum budgets among Earth, air, and water reservoirs is thus lacking at long periods. The requirements for progress in this field coincide completely with the central problems of global climate change. The five Earth Orientation Parameters (EOP) - two components of polar motion $x,y$, two components of celestial pole offsets $\Delta\psi,\Delta\epsilon$, and universal time $UT1$ (that is nothing else but the angle of Earth’s rotation around its spin axis) - are now analyzed and routinely derived from the observations at several world’s centers, combined and regularly published in IERS bulletins. The most recent system of constants and algorithms (McCarthy, 1996, Vondrák, 1998) are used . The past solutions based on optical astrometry were merged with the combined solutions from the modern techniques (Vondrák, 2001). The existing ERP (Earth Orientation Parameters) series have been analyzed by many scientists. The most extensive reviews, with historical meaning, were given in well-known monographs by Munk & MacDonald (1960) and Lambeck (1980). Most recent of these analyses is that by Zharkov et al. (1996). A very detailed review, mostly concentrated on modern space techniques and discussion of short-periodic effects, was published by Eubanks (1993). Moreover, in our previous paper (Šegan et al., 2003) we have introduced the relax period as good explanation of the residuals in the $(UT1-UTC)_{\mathrm{BLI}}$ data. The motivation for the leap second, therefore, is due to the fact that the second as presently defined is ”too short” to keep in step with the Earth. However, had the second been defined to be exactly equal to a mean solar second at the origin of $TAI$ in 1958, the discrepancy would not have been removed; the agreement between the SI second and the mean solar second would have only been temporary and their difference would simply have become gradually more apparent over this century. Continuing use of a non-uniform time scale including leap seconds in the face of these considerations could lead to the necessity to proliferate an effective method for extrapolation of the future values of $\Delta T$. To try that, according to our analysis of the IERS $\Delta T$ data from 1962.15 to 2006.95, existence of some periodic terms is discovered. By using the harmonic analysis 17 components plus a linear term of equation (1) are determined. $\begin{split}M\Delta T_{i}&=C_{1}+C_{2}\times t_{i}+\\\ &+\sum_{j=1}^{17}A_{j}^{c}\cos\left(\frac{2\pi}{P_{j}}\cdot t_{i}\right)+A_{j}^{s}\sin\left(\frac{2\pi}{P_{j}}\cdot t_{i}\right).\end{split}$ (1) In equation (1) $M\Delta T_{i}$ is a modified $\Delta T_{i}$. In order to obtain the $\Delta T_{i}$ values one needs to perform a translation by a constant value of $32.184+10$ seconds: $\Delta T_{i}=32.184+10-M\Delta T_{i}.$ Table 1. The coefficients for 17 components obtained from the harmonic analysis. $j$ \| $A_{j}^{c}$ \| $A_{j}^{s}$ \| $\sigma^{2}_{j}$ \| $P_{j}$ ---\|---\|---\|---\|--- 1 \| $5.634826$ \| $-46.439608$ \| $.20$ \| $222.28$ 2 \| $-.531603$ \| $.283347$ \| $.027$ \| $19.54$ 3 \| $-.051990$ \| $.172145$ \| $.011$ \| $12.54$ 4 \| $-.001363$ \| $-.117686$ \| $.0039$ \| $46.04$ 5 \| $-.037362$ \| $.009689$ \| $.0021$ \| $22.22$ 6 \| $-.029915$ \| $.004394$ \| $.0016$ \| $5.84$ 7 \| $-.017975$ \| $.018877$ \| $.0013$ \| $7.90$ 8 \| $-.002987$ \| $-.020452$ \| $.0011$ \| $1.00$ 9 \| $-.007429$ \| $.017729$ \| $.00090$ \| $6.50$ 10 \| $-.004306$ \| $.016504$ \| $.00076$ \| $4.63$ 11 \| $-.000759$ \| $.016891$ \| $.00062$ \| $9.26$ 12 \| $.002933$ \| $-.013173$ \| $.00052$ \| $3.57$ 13 \| $-.010059$ \| $-.006987$ \| $.00045$ \| $5.28$ 14 \| $-.001567$ \| $-.010152$ \| $.00040$ \| $4.08$ 15 \| $.006573$ \| $.007869$ \| $.00034$ \| $.50$ 16 \| $-.002887$ \| $.004019$ \| $.00033$ \| $1.09$ 17 \| $.000452$ \| $-.000316$ \| $.00033$ \| $.17$ Figure 2. The dashed line (blue) represents IERS data and the solid one (red) represents our approximation (prediction) between 1962.15 and 2006.95. The coefficients of the linear function (in equation (1)) are: free term $C_{1}=3.408457$ and slope coefficient $C_{2}=0.357042$. The coefficients of the harmonic components are given in Table 1. $\sigma^{2}_{j}$ is the upper limit for the coefficients of the harmonic components (it is the same for both coefficients, $A_{j}^{c}$ and $A_{j}^{s}$) and it is given with two significant digits. The periods $P_{j}$ are given in years. Because of the secular and decade variations all terms corresponding to insignificant periods and amplitudes (smaller than few milliseconds) are on the noise level so that they are not determined. In Fig. 2 the dashed line (blue) represents the IERS data from 1962.15 to 2006.95, whereas the solid one (red) represents our approximation for the period between 1962.0 and 2007.0. One can notice an excellent agreement. According to our approximation till 2024 there is practically no need to introduce leap seconds because an accumulation of 0.9 seconds is reached as late as at 2014.0, i. e. an accumulation of one second in 2015. The local maximum in our approximation occurs at 2016.4 and its value is 65.944 which exceeds the difference of one second by 0.1 seconds only (Fig. 3). In view of the further negative trend of our approximation the next leap second (negative) should be introduced in 2024 if the difference of 0.1 seconds at the moment of approximation maximum were neglected. Figure 3. The approximation from 1990 to 2035 and IERS data till 2006.95. The indicated points are: the moment when the leap second was introduced for the last time (2006), the next introducing (2009), the point at which according to our approximation the difference of 0.9 seconds between $UT1$ and $UTC$ is reached (2014.0), the local maximum of our approximation (2016.4), the point for which $\Delta T$ is approximately the same as in 2006 when the leap second was introduced for the last time (2021.95) and the point when, according to our approximation, the next leap second (negative) should be introduced (2024.0). Note that this means to introduce only one negative leap second within 18 years (2006–2024). Figure 4. The historical observations covering the period 1620–2006.95 are represented by the dashed line (blue) and our approximation concerning the interval 1620–2035 by the solid line (red). ## 3 4\. CONCLUSIONS As the day at present is actually closer to 86400 SI seconds the leap seconds have not been required regularly. However, this condition cannot persist and the long or mid-term trend will be eventually restored. The asymptotic behaviour of the $\Delta T$ in the Stephenson–Morrison approximation (1995) is not natural. As can be seen from the analysis (Fig. 4) applied within a relatively short time interval, 1962.15–2006.95, the extrapolation of the $\Delta T$ value agrees well with both the historical values (1620-1962) and the real measurements from 1962.15 to 2006.95. The discrepancy in the amplitude is higher for the former period, but the agreement for the phase is very good which indicates that the approximation could be improved if a bigger set of measurements were available. Our conclusion, that the first (additional) leap second should be introduced in 2014 only, opposes to all the extrapolations proposed till nowadays by both international institutions and individual experts. The majority of them predicts 2008 as the year when the additional correction should be introduced. The values attained by $\Delta T$ are of such order that all the physical factors unambiguously recognized till nowadays cannot cause such a phenomenon and in this sense the present analysis means a strictly mathematical approximation only. We believe that the years to come will show the correctness of the obtained results. REFERENCES Eubanks, T.M., 1993, Variations in the orientation of the Earth. In: Smith, D. E., Turcotte, D.L. (Eds.), Contributions of Space Geodesy to Geodynamics: Earth Dynamics. Geodyn. Ser. vol. 24, pp. 1 54. Lambeck K. 1980 The Earth’s Variable Rotation: Geophysical Causes and Consequences (Cambridge Univ. Press). McCarthy, D. D. (ed.), 1996, IERS Conventions, IERS Technical Note 21, Observatoire de Paris, Paris. Munk W. H., MacDonald G. J. F. 1960 The Rotation of the Earth: A Geophysical Discussion (Cambridge Univ. Press). Salstein, D.A. and R.D. Rosen, 1989: Regional contributions to the atmospheric excitation of rapid polar motions. J. Geophys. Res., 94, 9971–9978. Stephenson, F. R., and Morrison L. V., 1995, Long-Term Fluctuations in the Earth’s Rotation: 700 BC to AD 1990, Phil. Trans. R. Soc. Lond., Vol. 351, No. 1695., pp. 165–202. Šegan, S., Damjanov, I. and Šurlan, B., 2003, Earth’s rotation irregularities derived from UTIBLI by method of multi-composing of ordinates, Serbian Astronomical Journal, Vol. 167, 53. Vondrák J., 1987 in Holota P. (Ed.) Proc. Internat. Symp. Figure and Dynamics of the Earth, Moon and Planets (Astron. Inst. and Res. Inst. Geod. Praha), 1039. Vondrák, J., Pešek, I., Ron, C., Čepek, A.: 1998, Earth orientation parameters 1899.7–1992.0 in the ICRS based on the HIPPARCOS reference frame, Astron. Inst. of the Academy of Sciences of the Czech R., Publ. No. 87. Zharkov, V.N., Molodensky, S.M., Brzezinski, A., Groten, E., Varga, P., 1996, The Earth and its rotation. Wichmann, Heidelberg, vol. XIII, 501. TEXKOΥE U OQUVANJU PRESTUPNE SEKUNDE S. Xegan1, B. Xurlan1, S. Vidojeviﬄ1 1Katedra za astronomiju, Matematiqki fakultet, Univerzitet u BeograduStudentski trg 16, 11000 Beograd, Srbija UDK … ……………………… Ispitana je moguﬄnost da se prognozira broj prestupnih sekundi u bliskoj buduﬄnosti na osnovi poznavanja nekoliko harmonijskih komponenti u Zemljinoj rotaciji. Analizirani su podaci IERS-a koji obuhvataju interval od 1962,15 do 2006,95. Razlika $\Delta T$ je ekstrapolirana do 2035. godine i uporedjena sa predvidjanjima IERS-a do 2012. godine. Mozhe se videti da je u intervalu od 2006. do 2024. godine potrebno uvesti samo jednu (negativnu) prestupnu sekundu.	arxiv-papers	2008-08-27T01:12:58	2024-09-04T02:48:57.476773	{ "license": "Public Domain", "authors": "S. Segan, B. Surlan, S. Vidojevic", "submitter": "Sonja Vidojevic", "url": "https://arxiv.org/abs/0808.3612" }
0808.3645	# Geometric cross sections of rotating strings and black holes Toshihiro Matsuo∗ and Kin-ya Oda† ∗Graduate School of Pure and Applied Sciences, University of Tsukuba, Ibaraki 305-8571, Japan tmatsuo@het.ph.tsukuba.ac.jp †Department of Physics, Osaka University, Osaka 560-0043, Japan odakin@phys.sci.osaka-u.ac.jp (September 24, 2008) ###### Abstract We study the production cross section of a highly excited string with fixed angular momentum from an ultra-high energy collision of two light strings. We find that the cross section exhibits geometric behavior in a certain region of angular-momentum/impact-parameter space. This geometric behavior is common to the differential cross sections of a black hole production with fixed angular momentum and thus we see another correspondence between strings and black holes. OU-HET 610/2008 ## 1 Introduction String theory not only cures the ultraviolet divergences from graviton loops in S-matrices but also has revealed nontrivial consequences of black hole physics in which both gravity and quantum mechanics play pivotal roles together, adding credibility as a candidate for the theory of quantum gravity. Especially it is claimed that there is a correspondence: a black hole should correspond to an ensemble of highly excited single string states at a certain threshold $g_{s}^{2}M\sim 1$, namely the correspondence point, with $g_{s}$ and $M$ being the (closed) string coupling constant and the mass, respectively.111In this paper, we shall work in string natural unit ${\alpha^{\prime}}=1,4$ etc. We will specify when necessary. The string and black hole pictures should be valid for $g_{s}^{2}M\ll 1$ and $g_{s}^{2}M\gg 1$, respectively. This is based on the observation that both the string and black hole descriptions give the same order of entropy at the correspondence point, as one varies adiabatically the string coupling constant $g_{s}$ for a fixed mass $M$ [1, 2]. When one takes into account the self interactions, the string gets entangled in itself and consequently its typical size is reduced to the order of the string length scale which coincides with the horizon radius of the black hole at the threshold, providing another support for the correspondence [3, 4]. It should be interesting to consider the correspondence between a formation process of a string and that of a black hole both from an ultra high energy scattering, since the latter involves non-perturbative dynamics of the gravitational interactions which is not successfully formulated within string theory so far.222We also note that in a scenario with the string scale around TeV [5, 6], this process is not only theoretically important but also directly testable at the CERN Large Hadron Collider and beyond [7, 8]. See also [9]. At low energies, the string picture should be valid while at high energies, the black hole picture would prevail. For a classical gravitational scattering with center of mass energy $\sqrt{s}$, the black hole production cross section has been proven to be of the order of the black disk with its size being the Schwarzshild radius $r_{S}\sim(g_{s}^{2}M)^{1/(D-3)}$ of the black hole mass $M\sim\sqrt{s}$, where $D=d+1$ is the number of (large) spacetime dimensions [10, 11, 12]. Physical interpretation is that when initial particles are wrapped within the horizon scale, a black hole forms [13]. Dimopoulos and Emparan [14] have investigated the production of a single highly excited string as a black hole progenitor, in view of the correspondence principle with a fixed string coupling $g_{s}\ll 1$ and with a varying mass $M$, or equivalently with a varying center of mass energy $\sqrt{s}\sim M$. At the tree level, they obtained a linearly raising total cross section with respect to $s$ (for closed string). The tree level cross section does not match the black hole one at the correspondence point $\sqrt{s}\sim M_{C}\equiv g_{s}^{-2}$. Actually the resultant string cross section hits the unitarity bound around $\sqrt{s}\sim g_{s}^{-1}$ which is below $M_{C}$. Historically, when string theory was applied to the strong interactions, it was conjectured, based on experimental observations, that the theory would provide a constant total cross section above the unitarity bound if one managed to take all the higher order loop corrections into account [15]. If it is the case, the constant cross section, that is to be of order one, matches the black hole one at the correspondence point. Currently one cannot prove the validity of this argument given the status of string theory where the cross section is computed only up to few loops and the relative phase of each loop cannot be fixed a priori. Also, the total cross section of open string turns out to be constant of order $g_{s}$ at tree level, as we will see below, and again it does not agree with that of the black hole at the correspondence point.333The correspondence principle is blind to the end points of strings. There does not seem to be a strong reason to exclude the gravitating string ball made of a long open string when self interactions including closed string exchanges are taken into account at higher orders. However, it is just a first step to study a total cross section that is a sum of infinitely many partial wave cross sections. By decomposing a total cross section into partial wave cross sections we are able to investigate the correspondence in more detail. When a black hole is produced from a high energy scattering, two initial particles have a finite impact parameter $b$ and therefore finite angular momentum $J\sim b\|\vec{p}\|\sim bM$. If one assumes that the initial angular momentum is not lost much during the black hole formation process, the differential cross section $d\sigma_{\text{BH}}/dJ$ increases geometrically with the angular momentum as $J^{D-3}$ until a threshold value $J_{\text{max}}$ [16]. On the other hand, a string partial wave cross section is considered to give exponential damping with respect to $J$ due to the softness of string at high energy. Thus at first sight there does not seem a correspondence to hold even at the partial wave level. In this paper, we revisit the partial wave string amplitude in the ultrahigh energy limit, and show that the softness of string is seen at relatively large angular momentum or impact parameter444See also Refs. [17, 18] for related arguments on string side with a finite impact parameter. while in a certain region of the angular momentum space the partial wave cross section indeed shows the geometric behavior for both closed and open strings thus we find an universal behavior that is comparable with that of black hole. The organization of the paper is as follows: In section 2, we present our computation of the production cross section of an excited string with fixed total angular momentum $J$. We find soft behavior characteristic to a string in the Regge limit at large angular momentum region as well as a geometric behavior of the partial wave cross section for an angular momentum which is rather large but less than a certain energy scale in string unit. Then we discuss the energy region where the partial wave unitarity condition is satisfied, which is the necessary condition for the perturbative expansion to be valid. We also argue open string case. In section 3, we briefly review the correspondence principle and discuss the production cross section of a rotating black hole. Then we apply the correspondence principle to the black hole/heavy string production processes. Section 4 provides the summary and discussions for possible future directions. In Appendix A, we review an alternative way to obtain cross sections through residue computation. In Appendix B, we collect formulae useful for our computations. In Appendix C, we review the derivation of the partial wave unitarity condition in $D$ dimensions. ## 2 Rotating string production In order to get the production cross section of a string with fixed total angular momentum $J$, we shall employ four tachyon tree amplitude at high energy. For closed strings, the imaginary part of the amplitude provides the fixed angle total cross section whereas the real part dominantly governs the unitarity condition. ### 2.1 Total cross section We consider the four tachyon tree amplitude in closed bosonic string theory. The amplitude is given by the Virasoro-Shapiro amplitude: $\displaystyle{\cal{A}}^{\text{closed}}(s,t)$ $\displaystyle=2\pi g_{s}^{2}{\Gamma(-1-s)\Gamma(-1-t)\Gamma(-1-u)\over\Gamma(2+s)\Gamma(2+t)\Gamma(2+u)},$ (1) where $s+t+u=-4$ with $\alpha^{\prime}=4$. The high energy process in our concern is controlled by the Regge limit $s\gg 1$ with a fixed $t$. The amplitude in the Regge limit is computed by Stirling’s formula $\Gamma(n+1)\simeq n^{n}e^{-n}\sqrt{2\pi n}$ and become555 This is obtained by the prescription which simulates the expected quantum corrections to the sharp tree level resonances on the physical sheet (along real axis in the complex $s$ plane). The Regge limit is taken on the second sheet avoiding the poles on the real axis. See also footnote 10. $\displaystyle{\cal{A}}^{\text{closed}}(s,t)$ $\displaystyle\to 2\pi g_{s}^{2}\,s^{2+2t}\left(-{1\over t}+i\pi\right)\quad\text{for $s\gg 1$}.$ (2) Once the asymptotic form of high energy amplitude is found it is straightforward to see the total cross section for production of a heavy string state. The optical theorem provides the total cross section from the imaginary part of the forward scattering:666In [14] the authors obtained (3) by computing the residues of the $s$-channel resonances and averaging the delta function (see also Appendix A) $\displaystyle\sigma^{\text{closed}}(s)\simeq-\frac{\pi}{s}\operatorname{Res}{\cal{A}}^{\text{closed}}(s,t=0)=2\pi^{2}g_{s}^{2}s.$ $\displaystyle\sigma^{\text{closed}}(s)=\frac{1}{s}\operatorname{Im}{\cal{A}}^{\text{closed}}(s,t=0)=2\pi^{2}g_{s}^{2}s.$ (3) The total cross section raises linearly with $s$, as opposed to the field theory cases in which total cross sections decrease with energy due to the uncertainty relation: the higher the energy of particle, the smaller its wave length, i.e. the smaller effective size of the scatterer. On the other hand, the stringy uncertainty relation [19] indicates that the higher the scattering energy, the bigger the area of string becomes, resulting in the growing cross section. Intuitively, one may think that the raising cross section implies a breakdown of unitarity. However string theory (and gravity) contains massless modes and thus has a long range potential. Therefore there is no known restriction for total cross sections with respect to the unitarity such as the Froissart bound, thus the linearly raising behavior (3) is not necessary a contradiction to any physical requirement.777 In [20, 21], the genus $G$ four point amplitude in the Regge limit is computed as $\displaystyle{\cal A}_{G}(s,t)\simeq g_{s}^{2G+2}(\ln s)^{-12G}s^{G+2},$ where the $t$ dependence is not determined and more importantly the relative phase is not also known. One might suppose the relative phase of the leading correction to the tree amplitude is pure imaginary and then $\displaystyle\sigma(s)\simeq g_{s}^{2}s-g_{s}^{4}(\ln s)^{-12}s^{2}+O(g_{s}^{6}),$ which seems to satisfy the unitarity condition at high energy. Although it would be interesting to consider such higher order corrections, we do not pursue them and shall push forward the tree level analysis in this paper. To argue the unitarity we need to decompose into partial waves which we shall investigate in the following. ### 2.2 Partial wave cross section—rotating string production Let us compute the production cross section of a heavy string state with a fixed total angular momentum $J$. To this end we shall consider the partial wave expansion of the amplitude. In $D$ dimensions, the spherical harmonics is given by the Gegenbauer polynomials and the partial wave expansion of the amplitude is given in terms of them $\displaystyle{\cal{A}}(s,t)$ $\displaystyle=\sum_{J=0}^{\infty}{\cal{A}}_{J}(s){C_{J}^{\nu}(\cos\theta)\over C_{J}^{\nu}(1)},$ (4) where $\nu=(D-3)/2$ and $\cos\theta=1+2t/s$ with $\theta$ being the scattering angle. In (4), we put the factor $C_{J}^{\nu}(1)={\left(2\nu\right)_{J}/\Gamma(J+1)}$ to yield ${\cal{A}}(s,0)=\sum_{J=0}^{\infty}{\cal{A}}_{J}(s)$ so that the simple normalization $\displaystyle\sigma_{J}(s)={1\over s}\operatorname{Im}{\cal{A}}_{J}(s)$ (5) leads to the required formula $\sigma(s)=\sum_{J=0}^{\infty}\sigma_{J}(s)$, where $\left(a\right)_{n}=\Gamma(a+n)/\Gamma(a)$ is the Pochhammer symbol. Using the orthogonality condition (48) with the normalization factor (49) in Appendix C, the partial wave amplitude is given by888 When $D=4$, the expansion reduces to the better-known formula with the Legendre polynomial $\displaystyle{\cal{A}}(s,t)$ $\displaystyle=\sum_{J=0}^{\infty}{\cal{A}}_{J}(s)\,P_{J}\left(1+2t/s\right),$ $\displaystyle{\cal{A}}_{J}(s)$ $\displaystyle={2J+1\over 2}\int_{-1}^{1}d\cos\theta\,P_{J}(\cos\theta)\,{\cal{A}}\left(s,-s{1-\cos\theta\over 2}\right).$ $\displaystyle{\cal{A}}_{J}(s)$ $\displaystyle={C_{J}^{\nu}(1)\over N_{J}^{\nu}}\int_{-1}^{1}dz\left(1-z^{2}\right)^{\nu-{1\over 2}}C_{J}^{\nu}(z)\,{\cal{A}}(s,t),$ (6) where $t=-s(1-z)/2$ (and $z=\cos\theta$). In the high energy limit $s\gg 1$, the integral (6) is dominated by the forward scattering region $\|t\|/s\ll 1$ or $1-z\ll 1$. Thus the Regge limit (2) yield good approximations for the integral (6) and we have: $\displaystyle{\cal{A}}_{J}^{\text{closed}}(s)$ $\displaystyle\simeq 2\pi g_{s}^{2}\,s^{2}{C_{J}^{\nu}(1)\over N_{J}^{\nu}}\int_{-1}^{1}dz\left(1-z^{2}\right)^{\nu-{1\over 2}}C_{J}^{\nu}(z)\left(-{1\over t}+i\pi\right)s^{2t}.$ (7) The production cross section of a highly excited string with total angular momentum $J$ is given by the partial wave cross section through the optical theorem $\displaystyle\sigma_{J}^{\text{closed}}(s)={1\over s}\operatorname{Im}{\cal{A}}_{J}^{\text{closed}}(s)=2^{\nu+1}\pi^{5\over 2}g_{s}^{2}\Gamma\\!\left(\nu+{1\over 2}\right){\left[C_{J}^{\nu}(1)\right]^{2}\over N_{J}^{\nu}}{s\,e^{-\lambda}\over\lambda^{\nu}}I_{J+\nu}(\lambda),$ (8) where $\lambda\equiv s\ln s$ and we have utilized Eq. (51) in Appendix B to perform the integration. Further from the limit (52) in $\lambda\gg 1$ for the modified Bessel function, we get999 In [22] similar expression is obtained. However the weight function that appears in the partial wave expansion is not specified there, thus they obtained the partial wave cross section up to the $J$-dependent coefficient which is fixed in the present paper. Actually this factor is the origin of the geometric behavior of the cross section. $\displaystyle\sigma_{J}^{\text{closed}}(s)$ $\displaystyle\simeq 2^{\nu+{1\over 2}}\pi^{2}g_{s}^{2}\Gamma\\!\left(\nu+{1\over 2}\right){\left[C_{J}^{\nu}(1)\right]^{2}\over N_{J}^{\nu}}{s\over\lambda^{\nu+{1\over 2}}}{e^{-{(J+\nu)^{2}\over 2\lambda}}}.$ (9) The exponential factor shows the softness of the cross section that is a characteristic feature of string in the high energy processes. We can introduce an impact parameter $b$ through the total angular momentum with a fixed center of mass energy as $b=J/\sqrt{s}$. We find the effective size of string is about $\sqrt{\ln s}$. This is consistent with the well-known argument that the Fourier transform of the Regge amplitude with respect to the transverse momentum $p_{\perp}^{2}\sim-t$ gives the effective size of the string in the transverse space resulting the gaussian profile of width $\sqrt{\ln s}$: $\displaystyle{1\over s}\int{d^{D-2}p_{\perp}\over(2\pi)^{D-2}}A_{\text{Regge}}e^{ip_{\perp}\cdot x}$ $\displaystyle=s(4\pi\ln s)^{-(D-2)/2}e^{-x^{2}/(4\ln s)},$ $\displaystyle A_{\text{Regge}}$ $\displaystyle\sim s^{2+2t}.$ (10) Note that this argument is only valid at large $x$ because $t$ is small in the Regge amplitude and that it is not reliable at small $x$ region. On the other hand, Eq. (9) is valid for all $J$ and thus we are able to investigate the small $b$ ($=J/\sqrt{s}$) region. Therefore it is interesting to see the behavior in Eq. (9) especially at $J\lesssim\sqrt{s\ln s}$ where the gaussian damping factor can be neglected. Recalling the normalization (49) in Appendix B and $C_{J}^{\nu}(1)=(2\nu)_{J}/\Gamma(J+1)$, one may immediately read from Stirling’s formula for large $J\gg 1$ but $J\lesssim\sqrt{s\ln s}$ that the cross section behaves geometrically $\displaystyle\sigma_{J}(s)\propto J^{D-3}.$ (11) Thus we conclude that at $1\ll b\lesssim\sqrt{s}$ the cross section is described by a black disc. As we will argue, this is the characteristic behavior of differential cross sections of black hole and thus we have a correspondence. Before proceeding, several comments are in order: • So far we have considered the partial amplitude decomposed by using a basis of the Gegenbauer polynomials that is the highest spherical harmonics in $D$ dimensions. This means that we have obtained partial wave amplitudes with respect to the total angular momentum $J$. However we may use the “lowest spherical harmonics,” namely a plane wave $e^{iJ_{12}\theta}$, which is an eigenfunction of the angular momentum $J_{12}$. In this case one has $\sigma_{J_{12}}={1\over s}\int_{0}^{2\pi}{d\theta\over 2\pi}\operatorname{Im}A(s,t)e^{-iJ_{12}\theta},\qquad\text{with}\quad t=-{s\over 2}(1-\cos\theta).$ (12) In other words, the $J_{12}$ partial wave cross section is obtained from the Fourier transform of the imaginary part of the amplitude [22]. Although this is the same form as Eq. (10) with $D=3$, these are different Fourier transformations. As before, we may introduce an impact parameter (projected onto 1-2 plane): $b_{12}=J_{12}/\sqrt{s}$ and may focus on the small $\theta$ region (as a consequence of the Regge amplitude $\|t\|\ll 1$) by defining $\tilde{\theta}\equiv\sqrt{s}\theta$: $\displaystyle\sigma_{J_{12}}$ $\displaystyle\sim{1\over s}\int_{0}^{2\pi\sqrt{s}}{d\tilde{\theta}\over 2\pi}e^{-{\tilde{\theta}^{2}\over 2}\ln s-ib_{12}\tilde{\theta}}=\sqrt{s\over 2\pi\ln s}e^{-{b_{12}^{2}\over 2\ln s}}.$ (13) * • The cross section (9) has been obtained by applying the Regge limit to the integrand. On the other hand, as we explain in Appendix A, one can estimate its production cross section with fixed angular momentum $J$ by reading off the residue $\displaystyle\sigma_{J}(s)$ $\displaystyle\simeq-{\pi\over s}\operatorname{Res}_{s=N}{\cal{A}}_{J}(s)$ $\displaystyle={2\pi^{2}g_{s}^{2}\over s}{C_{J}^{\nu}(1)\over N_{J}^{\nu}}\int_{-1}^{1}dz\,(1-z^{2})^{\nu-{1\over 2}}C_{J}^{\nu}(z)\left(\Gamma(s+t+3)\over\Gamma(s+2)\Gamma(t+2)\right)^{2}.$ (14) Note that the Regge limit has not been taken here. The integral can be evaluated numerically and we can compare (8) and (14) to check the validity of the Regge limit in the integrand for a given $s$. As an illustration, the result for $s=100$ is shown in Fig. 1. Figure 1: Cross section $\sigma_{J}/\sigma_{\text{tot}}$ for $s=100$. $\sigma_{J}$ is obtained by Regge limit (8) and by reading off the residues (14) (dashed and solid lines, respectively). $\sigma_{\text{tot}}$ is given by (3). ### 2.3 Unitarity bound For completeness, we discuss how large $s$ the cross section can be trusted via the unitarity argument. We will study the partial wave unitarity by analyzing the four point amplitude from which we read off the production cross section of a string with fixed angular momentum. The partial wave unitarity at tree level has been investigated in [23] (and beyond tree level with eikonal approximation in [24]), see also [25]. It has been shown that when the angular momentum is less than $\sqrt{s\ln s}$, the partial wave amplitude might break the unitarity bound at high energy. Here we check whether our geometric behavior of the amplitude is within the unitarity bound. Due to the $t$-channel exchange of the massless ($t=0$) graviton, we see from Eq. (2) that the real part will be larger than the imaginary part in Eq. (6) in the high energy limit $s\gg 1$. In the limit, the real part can be written as $\displaystyle\operatorname{Re}{\cal{A}}_{J}^{\text{closed}}(s)$ $\displaystyle=4\pi g_{s}^{2}{C_{J}^{\nu}(1)\over N_{J}^{\nu}}s\,e^{-s\ln s}\int_{0}^{\infty}dw\,e^{-w}\int_{-1}^{1}dz\left(1-z^{2}\right)^{\nu-{1\over 2}}C_{J}^{\nu}(z)e^{(s\ln s+w)z}.$ (15) Using the formula (51) in Appendix B, we get $\displaystyle\operatorname{Re}{\cal{A}}_{J}^{\text{closed}}(s)$ $\displaystyle=2^{\nu+2}\pi^{3\over 2}g_{s}^{2}\,\Gamma\\!\left(\nu+{1\over 2}\right){{C_{J}^{\nu}(1)}^{2}\over N_{J}^{\nu}}s\,e^{-s\ln s}\int_{0}^{\infty}dw\,e^{-w}{I_{J+\nu}(s\ln s+w)\over\left(s\ln s+w\right)^{\nu}}.$ (16) The modified Bessel function takes the limit $I_{J+\nu}(\lambda)\to e^{\lambda}/\sqrt{2\pi\lambda}$ for $\lambda\gg 1$ and we get $\displaystyle\operatorname{Re}{\cal{A}}_{J}^{\text{closed}}(s)$ $\displaystyle=2^{\nu+{3\over 2}}\pi g_{s}^{2}\,\Gamma\\!\left(\nu-{1\over 2}\right){{C_{J}^{\nu}(1)}^{2}\over N_{J}^{\nu}}{s\over(s\ln s)^{\nu-{1\over 2}}}.$ (17) With the normalization of Eq. (57), we have $\displaystyle\operatorname{Re}a_{J}^{\text{closed}}(s)={g_{s}^{2}s\over\left(2\nu-1\right)2^{3\nu+{1\over 2}}\pi^{\nu-{1\over 2}}\left(\ln s\right)^{\nu-{1\over 2}}}.$ (18) On the other hand the imaginary part is also written with the normalization of Eq. (57) as $\displaystyle\operatorname{Im}a_{J}^{\text{closed}}(s)={g_{s}^{2}s\over 2^{3\nu+{5\over 2}}\pi^{\nu-{3\over 2}}\left(\ln s\right)^{\nu+{1\over 2}}}.$ (19) Plugging these into the condition for the partial wave unitarity (58), we have ${g_{s}^{2}s\over\left(\nu-{1\over 2}\right)^{2}2^{3\nu+{1\over 2}}\pi^{\nu+{1\over 2}}\left(\ln s\right)^{\nu-{3\over 2}}}\lesssim 1\qquad\text{for}\quad\nu>{1\over 2}\quad(D>4),$ (20) where we have used the fact that the real part dominates over the imaginary part in the right hand side of (58) in Appendix C. Indeed the real part is larger than the imaginary part by a factor $\ln s$. This shows that no matter how small the string coupling is, the unitarity bound will be violated at sufficiently high energy. It is interesting to notice that up to the $\ln s$ factor the partial wave unitarity bound is hit at $s\simeq 1/g_{s}^{2}$ which is precisely at the total cross section (3) becomes of order one. When one includes the $\ln s$ factor, the total cross section is reliable up to larger value of $s$ for a fixed $g_{s}$. ### 2.4 Open string case So far we have considered the closed string scattering. We can repeat the above argument for the open string case. We start from the Veneziano amplitude: ${\cal{A}}^{\text{open}}(s,t)=g_{s}\left[{\Gamma(-\alpha(s))\Gamma(-\alpha(t))\over\Gamma(-\alpha(s)-\alpha(t))}+{\Gamma(-\alpha(t))\Gamma(-\alpha(u))\over\Gamma(-\alpha(t)-\alpha(u))}+{\Gamma(-\alpha(u))\Gamma(-\alpha(s))\over\Gamma(-\alpha(u)-\alpha(s))}\right],$ (21) where $\alpha(x)=1+{\alpha^{\prime}}x$ and $s+t+u=-4$ with $\alpha^{\prime}=1$. The Regge limit reads101010 We take the large $s$ limit off the real axis $s\to(1+i\epsilon)\infty$ so that $\left(\sin\pi\alpha(s)\right)^{-1}\to 0$ and $\left(\tan\pi\alpha(s)\right)^{-1}\to-i$ exponentially. Physically, this limit corresponds to the assumption that the $s$-channel resonances have decay widths that increase at least linearly with the pole masses, which is natural given the exponentially growing number of decay modes in string theory. See also footnote 5. $\displaystyle{\cal{A}}^{\text{open}}$ $\displaystyle\to{-g_{s}\pi\over\Gamma\left(1+\alpha(t)\right)\sin\pi\alpha(t)}\left(1+e^{-i\pi\alpha(t)}\right)\left(\alpha(s)\right)^{\alpha(t)}$ $\displaystyle(s\gg 1)$ $\displaystyle\simeq\pi g_{s}s\left({\pi\over 2}t+i\right)$ $\displaystyle(t\simeq 0),$ (22) where the last line is the case in which the amplitude is dominated over by the $t$-channel exchange of massless modes of the open string. Note that the real part is small compared with the imaginary part contrary to the closed string case. We find immediately the total cross section: $\displaystyle\sigma^{\text{open}}(s)=\frac{1}{s}\operatorname{Im}{\cal{A}}^{\text{open}}(s,t=0)=\pi g_{s}.$ (23) The partial wave cross section is obtained quite similarly as in the case of closed string. The imaginary part gives $\displaystyle\operatorname{Im}{\cal{A}}_{J}^{\text{open}}(s)=2^{\nu}\pi^{{3\over 2}}\Gamma\\!\left(\nu+{1\over 2}\right)g_{s}{\left[C_{J}^{\nu}(1)\right]^{2}\over N_{J}^{\nu}}{s\,e^{-\lambda^{\prime}}\over\lambda^{\prime\nu}}I_{J+\nu}(\lambda^{\prime}),$ (24) where $\lambda^{\prime}=(s\ln s)/2$. Again we find a geometric cross section at $1\ll J\leq\sqrt{s\ln s}$ $\displaystyle\sigma_{J}^{\text{open}}\simeq J^{D-3}.$ (25) We may also check the partial wave unitarity. The real part is negligible to the imaginary part and thus the partial wave unitarity condition $\operatorname{Im}a_{J}\geq\|a_{J}\|^{2}$ becomes $\displaystyle\operatorname{Im}a_{J}^{\text{open}}\leq 1.$ (26) From Eq. (24) we have $\displaystyle\operatorname{Im}a_{J}^{\text{open}}={g_{s}\over 2^{2\nu+3}\pi^{\nu+{1\over 2}}(\ln s)^{\nu+{1\over 2}}}$ (27) thus the partial wave unitarity condition is satisfied at any high energy for open strings. ## 3 Correspondence and black hole cross section Let us briefly review the correspondence principle for black hole and string [1, 2]. In the perturbative formulation of string theory, the string coupling constant $g_{s}$ is a free parameter, which should be fixed by a dilaton vacuum expectation value supposedly fixed by non-perturbative effects. When we vary $g_{s}$ and/or the mass $M$, it is observed that various physical quantities are smoothly transited between string and black hole pictures. Especially as we vary the coupling adiabatically, adiabatic invariants would not change drastically at the transition point. The order parameter for the transition is $\mu\equiv g_{s}^{2}M$. The larger the coupling and/or mass is, the stronger the gravitational interactions are. Therefore we will see that the black hole (string) picture becomes valid for $\mu\gg 1$ ($\ll 1$) and the transition point is given at $\mu\sim 1$. The gravitational constant is given by $G\sim g_{s}^{2}$ in any dimensions. When $D=d+1$ spacetime dimensions are large, that is, uncompactified or compactified with a length scale much larger than our region of interactions, the $D$-dimensional Planck mass (length) is given by $M_{D}\sim g_{s}^{-2/(D-2)}$ ($\sim\ell_{D}^{-1}$). Note that, for a string coupling being fixed at a small value $g_{s}<1$, the Planck scale is always smaller than the correspondence scale $M_{D}\lesssim M_{C}=g_{s}^{-2}$ and that the black hole picture should be valid at the correspondence point $M\sim M_{C}$. The Schwarzschild radius for a black hole with mass $M$ is given by $r_{S}\sim(GM)^{1/(D-3)}\sim\mu^{1/(D-3)}$ and its entropy is given by the horizon area $S_{\text{BH}}\sim r_{S}^{D-2}/G\sim(g_{s}^{2}M^{D-2})^{1/(D-3)}$. For free string states with a fixed mass $M$ (being equal with its length in string unit), the entropy becomes $S_{\text{string}}\sim M$. Therefore the entropy becomes the same order in both pictures at the correspondence point $\mu\sim 1$ for any number of (large) dimensions.111111We note that the self interactions of the string are neglected to obtain $S_{\text{string}}\sim M$. At this level, given the form $S_{\text{BH}}\propto M^{a}$ with $a$ being some constant, the agreement of the temperature $T=(\partial S/\partial M)^{-1}$ at the correspondence point is trivially derived from that of the entropy. See also [26]. We note that the string states are treated within a micro/grand canonical ensemble and that a black hole corresponds to the ensemble for the correspondence to hold. The compared quantities are averaged values within the ensemble. If we neglect the self interactions of the string, typical size would be that of the random walk $R_{\text{RW}}\sim\sqrt{M}$, which is much larger than the black hole horizon radius at the correspondence point for $g_{s}<1$. This discrepancy can be solved by properly taking into account the self interactions [3, 4]. If correspondence is valid and a black hole can be viewed as an ensemble of the corresponding string states, black hole production cross sections must be connected to the production cross sections of string states. As we have seen in section two, the tree level total production cross section of a closed string is obtained from the four point amplitude and the result is $\sigma_{\text{tot}}(s)\sim g_{s}^{2}s$. When $s\gtrsim g_{s}^{-2}$ the partial wave cross sections start to exceed the unitarity bound (20). We note that this is the scale where the D-brane interactions become also significant. Now let us consider the trans-Planckian collision of two light particles. The production cross section of a black hole can be computed by the assumption that the colliding target looks like a black disk with its radius being of order the Schwarzschild radius $r_{S}\sim(g_{s}^{2}M)^{1/(D-3)}$. Physically, when the initial particles are wrapped within the horizon, a black hole forms. It has been proven in four dimensions [10] and in higher dimensions [11] that a classical gravitational collision of two massless particles, whose gravitational fields are simulated by the Aichelburg-Sexl solution [27], leads to a formation of a trapped surface, outside which there must be an event horizon [28]. This argument has been generalized to the collision of two wave packets [29]. The resultant black hole production cross section turns out to be geometrical $\displaystyle\sigma_{\text{BH}}\sim R_{S}^{D-2}\sim(g_{s}^{2}M)^{D-2\over D-3}.$ (28) At the correspondence point, the black hole production cross section (28) becomes unity $\sigma_{\text{BH}}\sim 1$. Clearly, the tree level total cross section of string (3) and black hole do not coincide at the correspondence point $s\sim g_{s}^{-4}$. We note that the classical treatment of the gravitational interactions is valid when the black hole mass is larger than the $D$ dimensional Planck scale $M\gtrsim M_{D}\sim g_{s}^{-2/(D-2)}$ and that $M_{D}$ is smaller than the correspondence scale $M_{C}\sim g_{s}^{-2}$. Therefore the black hole production cross section can be trusted at $M\gtrsim M_{C}$ for $g_{s}<1$. The meaning of the correspondence point is that the (classical) stringy corrections cannot be neglected for $M\lesssim M_{C}$ though the classical treatment of the gravity is still valid. A cartoon is shown in Fig. 2 to help understanding. Figure 2: Schematic log-log plot for the production cross section of a string and black hole. As is emphasized in Introduction, it is important to consider the partial wave cross sections to see more detailed information. For a black hole, the partial wave cross section is simply given by the differential cross section $d\sigma_{\text{BH}}/dJ$. Let us consider a Kerr black hole with mass $M=\sqrt{s}$. There are ${}_{d}C_{2}=d(d-1)/2$ components of angular momenta according to the choice of planes in $d$ spatial directions. The number of independent components are $d/2$ for $d$ even and $(d-1)/2$ for $d$ odd. Let us review the argument presented in [16]. Initial two light particles collide with an impact parameter $b$ in $D=d+1$ dimensions. By choosing the scattering plane as 1–2 one, the initial system has only single non-zero angular momentum component $J=J_{12}=b\sqrt{s}/2$, with $\sqrt{s}$ being the center of mass energy and $b$ the impact parameter. If angular momentum is conserved during the black hole formation process, it is sufficient to consider a black hole having only single non-zero angular momentum component. For such a higher dimensional rotating black hole, the horizon radius $r_{h}$ is determined by $\displaystyle r_{h}^{D-5}\left(r_{h}^{2}+{(D-2)^{2}J^{2}\over 4M^{2}}\right)={16\pi GM\over(D-2)\Omega_{D-2}}.$ (29) Utilizing the rescaled mass and angular momentum $\displaystyle\tilde{\mu}\equiv{16\pi GM\over(D-2)\Omega_{D-2}},\qquad a_{}\equiv{(D-2)J\over 2Mr_{h}},$ (30) the horizon radius can be written as $\displaystyle r_{h}$ $\displaystyle=\left({\tilde{\mu}\over 1+a_{}^{2}}\right)^{1\over D-3}.$ (31) Note that the horizon radius $r_{h}$ is now given in terms of the mass $M=\sqrt{s}$ and the angular momentum $J=b\sqrt{s}/2$. For a fixed $s$, one can show that $r_{h}$ is a decreasing function of $b$. Following from the hoop conjecture, a black hole would be produced when the impact parameter of the collision is less than the diameter of the black hole121212However, this is not a coordinate invariant description. The diameter used here is provided by the Schwarzschild radius itself. See Appendix D for a related discussion on the radius. $\displaystyle b\leq 2r_{h}(b).$ (32) For a given $M=\sqrt{s}$, the right hand side of Eq. (32) can be shown to be a decreasing function of $b$, and hence there is a maximum impact parameter that saturates the inequality (32). Noting that $b=2J/M=4r_{h}a_{}/(D-2)$, the condition (32) leads to $a_{}\leq(D-2)/2$, whose equality gives the maximum impact parameter $\displaystyle b_{\text{max}}$ $\displaystyle=2\left[{\tilde{\mu}\over 1+a_{\text{max}}^{2}}\right]^{{1\over D-3}}\qquad\text{with}\quad a_{\text{max}}={D-2\over 2}.$ (33) It is amusing that $b_{\text{max}}$ exactly coincides with the naive Schwarzschild estimation $r_{S}=\tilde{\mu}^{1/(D-3)}$ in $D=4$ spacetime dimensions and that $b_{\text{max}}>r_{S}$ for $D\geq 5$. The more dimensions we have, the bigger the increase of the $b_{\text{max}}$, the cross section. This tendency agrees with the numerical analysis and the numerical values agrees within order ten percent accuracy in an appropriate number of (large) dimensions for string theory [11, 12].131313In Ref. [30], quite a similar condition $b\leq r_{h}(b)$ was examined and it was concluded that the cross section would decrease from $r_{S}$ (corresponding to different $a_{\text{max}}$). Therefore we conclude that our assumption that the initial angular momentum and energy are almost all packed into the black hole is justified unless $b$ is not very close to $b_{\text{max}}$.141414When $b\sim b_{\text{max}}$, the reduced mass, which gives the lower bound for the final black hole mass, is sizably reduced from the “all-packed” assumption [11]. The effect of the angular momentum appear only through $a_{}$ that is less than or of order one as explained above. Therefore one can drop the $a_{}$-dependent factors when considering the total cross section up to a numerical factor of order one.151515It is also true in the arugument of the entropy correspondence of rotating string/black hole. Neglecting all such numerical coefficients, we get $\displaystyle b_{\text{max}}\sim\left(G\sqrt{s}\right)^{1\over D-3}$ (34) and the black hole cross section is $\displaystyle\sigma_{\text{BH}}$ $\displaystyle\sim b_{\text{max}}^{D-2}\sim\left(G\sqrt{s}\right)^{{D-2\over D-3}}.$ (35) With the above assumptions we also get the differential cross section of black hole with given angular momentum $J$ as $\displaystyle{d\sigma_{\text{BH}}\over dJ}\simeq\frac{J^{D-3}}{s^{D-2\over 2}}\qquad\text{for}\quad J\lesssim J_{\text{max}}\sim\left(Gs^{D-2\over 2}\right)^{1\over D-3}.$ (36) where we have used $d\sigma_{\text{BH}}\simeq b^{D-3}db$.161616It is interesting to notice that the maximal angular momentum $J_{\text{max}}$ and the entropy have the same form $(Gs^{(D-2)/2})^{1/(D-3)}$. Finally, comparing Eqs. (11), (25) and (36) we have shown the correspondence of the partial wave cross sections between closed/open strings and black holes $\displaystyle\sigma_{J}^{\text{string}}(s)\leftrightarrow{d\sigma_{\text{BH}}\over dJ}$ (37) for $1\ll J\leq\sqrt{s\ln s}$. ## 4 Summary and discussion We have studied the production cross section of a highly excited string with a fixed angular momentum from a high energy collision of two light open or closed strings at the tree level. We have also re-derived the partial wave unitarity condition at high energy. We find that the cross sections exhibit, in addition to the softness in the large angular momentum/impact parameter space, geometric behaviors in the large but less than $\sqrt{\ln s}$ region of the impact parameter space. The geometric behavior is characteristic to differential cross sections of black hole and thus we see a correspondence between string and black hole. The total cross sections do not coincide at the correspondence point at the tree level analysis. These cross sections of closed string and black hole share the growing behavior, both break the unitarity at high energy, while that of the open string stays constant. On the other hand, we have presented that the behaviors of the partial wave cross sections are common between closed/open strings and black holes. Although our tree level amplitude for closed string cannot be trusted at the correspondence point because of the unitarity violation, the agreement of the geometric behaviors in the finite lower energy range is an incarnation of the correspondence principle. Generically, the geometric behavior of the partial cross sections is a consequence of black disc total cross section. Interestingly enough, high energy hadron collisions are also well described by the black disc approximation in which the radius of the disc is given by a range of strong interaction typically provided by a mass gap. The black disc provides Color Glass Condensation (CGC) in which gluon density saturates. The CGC is set as an initial condition to simulate the quark-gluon plasma generated in high energy hadron collisions, see e.g. [32, 33]. Our string geometric cross sections might well be a model for the hadron and the CGC. We are focusing on the $s$-channel single heavy string production, which is equivalent to the $t$-channel single graviton exchange. If the correspondence holds at all, production process of the single string, dressed by the graviton cloud that accounts for the self interactions, would be more or less smoothly connected to that of the black hole. However, Amati, Ciafaloni and Veneziano (ACV) showed long before [34, 35] that a string feels as if it is propagating under Aichelburg-Sexl background [27] after summing over the eikonal graviton exchanges with the other string, see also Ref. [17] for more recent discussion. Note that, as we already stated, a ‘collision’ of two Aichelburg- Sexl solutions is proven to lead to a black hole production in four dimensions [10] and in higher dimensions [11]. In this sense, the infinitely many eikonal $t$-channel graviton exchanges appear to provide the correspondence here. Furthermore, the dominant contribution in the path integral is from the configuration where all the internal strings share equal amount of energy and hence there are no special one to be picked out [36]. Therefore, it is somewhat puzzling how to reconcile the former correspondence picture with the latter ACV one. To repeat, the former involves a single $t$-channel massless string exchange (though with graviton clouds) while the latter is a sum over such exchanges, yet somehow to re-emerge a single-string-like behavior for the correspondence picture (a black hole being a long string at the threshold) to be recovered. We have studied the production processes in which the initial and final objects in the collision are same, that is, $\text{open}+\text{open}$ to an open string and $\text{closed}+\text{closed}$ to a closed string. In order to understand differences and/or common features (universality), it would be interesting to study the production cross section of a rotating closed string from two light open strings on D$p$-brane, which is also more realistic to describe a scattering in the brane world scenario [37], by performing similar analysis on the annulus amplitude with each pair of open strings attached on the inner and outer boundaries. We hope to report investigations on the points mentioned above elsewhere. ## Acknowledgments We are grateful to Hikaru Kawai and Tsunehide Kuroki for useful discussions presented in [22], which provided basis for this work. We thank Yoichiro Nambu for valuable comments. We thank Roberto Emparan, Yoshitaka Hatta, Feng-Li Lin and Yuji Satoh for useful discussions. T.M. also acknowledges the hospitality of Albert Einstein Institute, Niles Bohr Institute and theoretical high energy physics group at Pisa while this work was in progress. We thank the RIKEN Theory Group where a large fraction of this work is done. The work of K.O. is partially supported by Scientific Grant by Ministry of Education and Science, Nos. 19740171, 20244028, and 20025004. ## Appendix ## Appendix A Total cross section In this Appendix we briefly review the tree level computation of the production cross section of the level $N$ resonance [14]. We consider the Virasoro-shapiro amplitude (1) which can be written around an $s$-channel pole at $s=N$ as, $\displaystyle{\cal{A}}(s,t)$ $\displaystyle={1\over s-N+i\epsilon}\operatorname{Res}_{s=N}{\cal{A}}(s,t)+\text{(terms analytic at $s=N$)},$ (38) where the infinitesimal $\epsilon$ gives $\displaystyle\operatorname{Im}{\cal{A}}(s,t)$ $\displaystyle=-\pi\delta(s-N)\operatorname{Res}_{s=N}{\cal{A}}(s,t)+\text{(terms analytic at $s=N$)}.$ (39) Combining Eq. (39) with the optical theorem, the cross section around the $N$th resonance is given as $\displaystyle\sigma_{N}(s)$ $\displaystyle=-{\pi\over s}\delta(s-N)\operatorname{Res}_{s=N}{\cal{A}}(s,0)+\text{(terms analytic at $s=N$)}.$ (40) Noting that the $s=N$ residue resides only in $\operatorname{Res}_{s=N}\Gamma(-1-s)={(-1)^{N}\over\Gamma(N+2)}$ in the amplitude (1), it is straightforward to compute $\displaystyle\operatorname{Res}_{s=N}{\cal{A}}(s,t)$ $\displaystyle=-2\pi g_{s}^{2}\left(\Gamma(N+t+3)\over\Gamma(N+2)\Gamma(t+2)\right)^{2},$ (41) where we have also utilized $\Gamma(x)\Gamma(1-x)=\pi/\sin\pi x$. Therefore, we obtain $\displaystyle\sigma_{N}(s)$ $\displaystyle=2\pi^{2}g_{s}^{2}{(N+2)^{2}\over s}\delta(s-N)+\text{(terms analytic at $s=N$)}.$ (42) Finally the total cross section becomes [14] $\displaystyle\sigma_{\text{tot}}(s)$ $\displaystyle=\sum_{N}\sigma_{N}(s)\simeq\sum_{N}2\pi^{2}g_{s}^{2}N\,\delta(s-N),$ (43) where the $N\gg 1$ limit is taken in the last step. The total cross section (43), which has been given at the tree level, becomes zero and infinity off and at the resonances, respectively. When we take the loop corrections into account, the $N$th resonance will have a finite decay width $\Gamma_{N}$ that corresponds to a finite $\epsilon_{N}\simeq 2\sqrt{N}\Gamma_{N}$: $\displaystyle{1\over s-N+i\epsilon_{N}}$ $\displaystyle={s-N\over(s-N)^{2}+\epsilon_{N}^{2}}-i{\epsilon_{N}\over(s-N)^{2}+\epsilon_{N}^{2}}.$ (44) When the decay width for the $N$th resonance is small enough, we can recover $\displaystyle\lim_{\epsilon_{N}\to 0}{1\over s-N+i\epsilon_{N}}$ $\displaystyle\simeq{\cal{P}}{1\over s-N}-i\pi\delta(s-N).$ (45) Otherwise, the delta function is meant to give the correct value when integrated over the width of the peak.171717Recall that $\int_{-\infty}^{\infty}ds{\epsilon\over(s-N)^{2}+\epsilon^{2}}=\pi$ regardless of the magnitude of $\epsilon$. In other words, at the large $N$ the spacing of the delta function is close and we may have $\displaystyle\sigma_{\text{tot}}(s)\simeq 2\pi^{2}g_{s}^{2}s.$ (46) It should be kept in mind that, in reality, the decay width for the higher resonance $N\gg 1$ can be large due to the exponentially growing number of the decay modes. The pole at the tree level amplitude (1) does not have an imaginary part corresponding to the decay width of the resonance (other than the elastic one), which will be served by the higher loop corrections including many body final states. The sharp $s$-channel resonance will eventually be smeared out due to the exponential grow of the number of such decay channels. Note also that when one takes the Regge limit of a stringy amplitude, such a width is implicitly taken into account. ## Appendix B Gegenbauer polynomial In this Appendix we collect some useful formulas for our computations. For the gamma function: $\displaystyle\Gamma\\!\left(\nu+{1\over 2}\right)$ $\displaystyle={\sqrt{\pi}\over 2^{2\nu-1}}{\Gamma(2\nu)\over\Gamma(\nu)}={\sqrt{\pi}\over 2^{2\nu-1}}\left(\nu\right)_{\nu}.$ (47) The orthogonality condition for the Gegenbauer polynomial: $\displaystyle\int_{-1}^{1}dz\left(1-z^{2}\right)^{\nu-{1\over 2}}C_{J}^{\nu}(z)\,C_{J^{\prime}}^{\nu}(z)$ $\displaystyle=N_{J}^{\nu}\,\delta_{JJ^{\prime}},$ (48) where the normalization factor is given by $\displaystyle N_{J}^{\nu}$ $\displaystyle={\pi\Gamma(J+2\nu)\over 2^{2\nu-1}(J+\nu)\,\Gamma(J+1)\,\Gamma(\nu)^{2}}={\sqrt{\pi}\left(\nu\right)_{1/2}\over J+\nu}C_{J}^{\nu}(1).$ (49) The angular integral formula for the Gegenbauer polynomial: $\displaystyle\int d\Omega_{n}C_{J}^{\nu}(\cos\theta_{fn})C_{J^{\prime}}^{\nu}(\cos\theta_{in})$ $\displaystyle={(4\pi)^{\nu}N_{J}^{\nu}\over\left(\nu\right)_{\nu}C_{J}^{\nu}(1)}C_{J}^{\nu}(\cos\theta_{if})\delta_{JJ^{\prime}}.$ (50) Another useful integral formula for $\operatorname{Re}\nu>-1/2$: $\displaystyle\int_{-1}^{1}dz\left(1-z^{2}\right)^{\nu-{1\over 2}}C_{J}^{\nu}(z)e^{\lambda z}$ $\displaystyle={2^{\nu}\sqrt{\pi}\Gamma\\!\left(\nu+{1\over 2}\right)C_{J}^{\nu}(1)I_{J+\nu}(\lambda)\over\lambda^{\nu}}.$ (51) For large $\lambda$, the modified Bessel function has the asymptotic form [22] $\displaystyle I_{n}(\lambda)$ $\displaystyle\simeq\frac{1}{\sqrt{2\pi\lambda}}\exp\\!\left(\lambda-{n^{2}\over 2\lambda}\right).$ (52) ## Appendix C Partial-wave unitarity bound We spell out the unitarity condition in our notation basically following Ref. [23]. Set the $S$-matrix $S=1+iT$ and write generically $\braket{j}{T}{k}=(2\pi)^{D}\delta^{D}(P_{j}-P_{k})\,{\cal{A}}(k\to j)$. Let $\Ket{i}$ and $\Ket{f}$ be two-body states of the same particle contents so that $\braket{f}{T}{i}$ describes the corresponding elastic scattering. The unitarity condition $S^{\dagger}S=1$ implies $-i(T-T^{\dagger})=T^{\dagger}T$, that is, $\displaystyle 2\operatorname{Im}{\cal{A}}^{\text{el}}(i\to f)$ $\displaystyle=\sum_{n}(2\pi)^{D}\delta^{D}(P_{n}-P_{i})\,\left[{\cal{A}}(f\to n)\right]^{}\,{\cal{A}}(i\to n),$ (53) where the summation over $n$ includes momentum integrals. We can separate the sum into elastic and other parts $\sum_{n}=\sum_{n}^{\text{el}}+\sum_{n}^{\text{others}}$. In the first elastic sum, $\ket{n}$ has the same particle contents as $\ket{i}$ and $\ket{j}$, while the second sum includes inelastic scattering, many body final states, etc. From now on, we work in the center of mass frame unless otherwise stated. The phase space integral reads $\displaystyle\sum_{n}^{\text{elastic}}(2\pi)^{D}\delta^{D}(P_{n}-P_{i})$ $\displaystyle=\int{d^{d}\mathbf{p}_{n1}\over(2\pi)^{d}2E_{n1}}\int{d^{d}\mathbf{p}_{n2}\over(2\pi)^{d}2E_{n2}}(2\pi)^{D}\delta^{D}(p_{n1}+p_{n2}-P_{i})$ $\displaystyle={\|\mathbf{p}_{n1}\|^{D-3}\over(2\pi)^{D-2}4P_{i}^{0}}\int d\Omega_{n1}\to{s^{\nu-{1\over 2}}\over 2(4\pi)^{2\nu+1}}\int d\Omega_{n1},$ (54) where we used $P_{i}^{0}=\sqrt{s}$ and also $\|\mathbf{p}_{n1}\|\to\sqrt{s}/2$ for $s$ much larger than the corresponding mass squared. We expand Eq. (53) into partial waves by Eq. (4) $\displaystyle 2\sum_{J=0}^{\infty}\operatorname{Im}{\cal{A}}_{J}^{\text{el}}(i\to f){C_{J}^{\nu}(\cos\theta_{if})\over C_{J}^{\nu}(1)}$ $\displaystyle\quad=\sum_{n}(2\pi)^{D}\delta^{D}(P_{i}-P_{n})\sum_{J=0}^{\infty}\left[{\cal{A}}_{J}(f\to n)\right]^{}\frac{C_{J}^{\nu}(\cos\theta_{fn})}{C_{J}^{\nu}(1)}\sum_{J^{\prime}=0}^{\infty}{\cal{A}}_{J^{\prime}}(i\to n)\frac{C_{J^{\prime}}^{\nu}(\cos\theta_{in})}{C_{J^{\prime}}^{\nu}(1)}$ $\displaystyle\quad={s^{\nu-{1\over 2}}\over 2(4\pi)^{2\nu+1}}\sum_{J,J^{\prime}=0}^{\infty}\left[{\cal{A}}_{J}^{\text{el}}(f\to n)\right]^{}{\cal{A}}_{J^{\prime}}^{\text{el}}(i\to n)\int d\Omega_{n}{C_{J}^{\nu}(\cos\theta_{fn})C_{J^{\prime}}^{\nu}(\cos\theta_{in})\over C_{J}^{\nu}(1)C_{J^{\prime}}^{\nu}(1)}+\text{others}$ $\displaystyle\quad={s^{\nu-{1\over 2}}\over 2(4\pi)^{\nu+1}}{\Gamma(\nu)\over\Gamma(2\nu)}\sum_{J=0}^{\infty}\left[{\cal{A}}_{J}^{\text{el}}(f\to n)\right]^{}{\cal{A}}_{J}^{\text{el}}(i\to n){N_{J}^{\nu}\over{C_{J}^{\nu}(1)}^{2}}{C_{J}^{\nu}(\cos\theta_{if})\over C_{J}^{\nu}(1)}+\text{others},$ (55) where “others” denotes $\sum_{n}^{\text{others}}\left[{\cal{A}}(f\to n)\right]^{}{\cal{A}}(i\to n)\,(2\pi)^{D}\delta^{D}(p_{n}-p_{i})$ and we have utilized Eq. (54) and (50). In the forward scattering limit $\ket{f}\to\ket{i}$, each term in the sum $\sum_{n}$ in Eq. (55) goes to $\left\|{\cal{A}}(i\to n)\right\|^{2}$ and becomes positive. Noting that the elastic matrix elements depend only on $s$ and $J$, namely ${\cal{A}}^{\text{el}}_{J}(i\to n)={\cal{A}}_{J}^{\text{el}}(s)$ for any $n$, we get a sufficient condition for each $J$ in order to satisfy the unitarity of the S-matrix: $\displaystyle 2\operatorname{Im}{\cal{A}}_{J}^{\text{el}}(s)$ $\displaystyle\geq{s^{\nu-{1\over 2}}\over 2(4\pi)^{\nu+1}}{\Gamma(\nu)\over\Gamma(2\nu)}{N_{J}^{\nu}\over{C_{J}^{\nu}(1)}^{2}}\left\|{\cal{A}}_{J}^{\text{el}}(s)\right\|^{2}.$ (56) Defining $\displaystyle a_{J}(s)\equiv{s^{\nu-{1\over 2}}\over 4(4\pi)^{\nu+1}}{\Gamma(\nu)\over\Gamma(2\nu)}{N_{J}^{\nu}\over{C_{J}^{\nu}(1)}^{2}}{\cal{A}}_{J}^{\text{el}}(s),$ (57) the unitarity condition (56) reads $\displaystyle\operatorname{Im}a_{J}(s)\geq\|a_{J}(s)\|^{2}.$ (58) As an immediate corollary $\sqrt{\left(\operatorname{Re}a_{J}\right)^{2}+\left(\operatorname{Im}a_{J}-{1\over 2}\right)^{2}}\leq{1\over 2}$ (59) or $\displaystyle\|a_{J}(s)\|\leq 1.$ (60) ## Appendix D Various radii and ultra-spinning black disk formation In this Section we comment on some subtleties on the choice of the horizon radius which is to be used in the naive estimation of the cross section. Generically one should not take the numerical coincidence of the total cross section between the naive one computed from Eq. (33) and the exact lower bound by Refs. [11, 12] too literally, because the definition of the radius (31) is not coordinate invariant.181818 We thank Roberto Emparan for the discussion on which this section is based. Our metric for the relevant Myers-Perry black hole with a single angular momentum is given by $\displaystyle ds^{2}$ $\displaystyle=-dt^{2}+{\mu\over r^{D-5}\rho^{2}}\left(dt+a\sin^{2}\theta d\phi\right)^{2}+{\rho^{2}\over\Delta}dr^{2}+\rho^{2}d\theta^{2}+(r^{2}+a^{2})\sin^{2}\theta d\phi^{2}+r^{2}\cos^{2}\theta d\Omega^{2}_{D-4},$ (61) where $\displaystyle\rho^{2}(r,\theta)$ $\displaystyle=r^{2}+a^{2}\cos^{2}\theta,$ $\displaystyle\Delta(r)$ $\displaystyle=r^{2}+a^{2}-{\mu\over r^{D-5}}.$ (62) We can define the following proper radii for a rotating black hole in $D$ dimensions $\displaystyle r_{\text{tot}}\equiv\left(A_{\text{tot}}/\Omega_{D-2}\right)^{1/(D-2)}$ $\displaystyle=\left(1+a_{}^{2}\right)^{1/(D-2)}r_{h},$ (63) $\displaystyle r_{\parallel}\equiv\left(A_{\parallel}/4\pi\right)^{1/2}$ $\displaystyle=\left(1+a_{}^{2}\right)^{1/2}r_{h},$ (64) $\displaystyle r_{\text{eq}}\equiv\ell_{\text{eq}}/2\pi$ $\displaystyle=\left(1+a_{}^{2}\right)r_{h},$ (65) where $A_{\text{tot}}$ is the total horizon area, $A_{\parallel}$ is the two dimensional area along $\theta$ and $\phi$ directions,191919Note that $A_{\parallel}$ is not the area of the section of the horizon that lies on the brane. and $\ell_{\text{eq}}$ is the length of the equator.202020More explicitly, the length $\ell_{\text{eq}}$ is the integral of $\sqrt{g_{\phi\phi}}$ over $\phi=0$ to $2\pi$ with fixed $r=r_{h}$ and $\theta=\pi/2$ in the metric (61), etc. For a very large angular momentum $a_{}\gg 1$, the resultant ultra-spinning black hole takes the shape of a thin pancake whose thickness being of order $r_{h}$, defined in Eq. (31), while the pancake radius being of order $a_{}r_{h}\gg r_{h}$ [31]. In this regard, the pancake radius should correspond to (64). The estimation (33) that uses the shortest definition of the horizon size (31) provides the most conservative lower bound on $b_{\text{max}}$ within this naive estimation framework. If we require the impact parameter of the collision to be smaller than the diameter with respect to this pancake radius $\displaystyle b<2r_{\text{eq}}(b)$ (66) instead of (32), the condition (66) can be satisfied by an arbitrary large $b$ (and hence $a_{}$) without leading to an upper bound on $b$ and therefore the cross section diverges naively. In such an extreme, the upper bound on $b$ and hence on $a_{}$ would be put by requiring the pancake thickness to be longer than the Planck length $r_{h}\gtrsim M_{D}^{-1}$ for a fixed $M$ ($\sim\sqrt{s}$), instead of the classical naive consideration (66). We note that such an ultra-spinning black hole with $a_{}\gg 1$ will suffer from classical gravitational instabilities [31], which might lead to a formation of a black ring [16] that will also suffer from the black string instabilities.212121 In Ref. [16] and its series, terminology of “large angular momentum” is used to indicate that the Hawking radiation is greatly distorted from the Schwarzschild one, which is generically the case even for $a_{}\lesssim 1$. For example, the maximum angular parameter from the conservative estimate (32) takes values $a_{\text{max}}=1.5$ to 4 for $D=5$ to 10. In this paper, we constrain ourselves within more conservative range $a\leq a_{\text{max}}=(D-2)/2$, leaving a stringy consideration of the possible case $a_{}\gg 1$ for future research. ## References * [1] L. Susskind, “Some speculations about black hole entropy in string theory,” arXiv:hep-th/9309145. * [2] G. T. Horowitz and J. Polchinski, “A correspondence principle for black holes and strings,” Phys. Rev. D 55 (1997) 6189 [arXiv:hep-th/9612146]. * [3] G. T. Horowitz and J. Polchinski, “Self gravitating fundamental strings,” Phys. Rev. D 57 (1998) 2557 [arXiv:hep-th/9707170]. * [4] T. Damour and G. Veneziano, “Self-gravitating fundamental strings and black holes,” Nucl. Phys. B 568 (2000) 93 [arXiv:hep-th/9907030]. * [5] I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos and G. R. Dvali, “New dimensions at a millimeter to a Fermi and superstrings at a TeV,” Phys. Lett. B 436 (1998) 257 [arXiv:hep-ph/9804398]. * [6] L. Randall and R. Sundrum, “A large mass hierarchy from a small extra dimension,” Phys. Rev. Lett. 83 (1999) 3370 [arXiv:hep-ph/9905221]. * [7] S. B. Giddings and S. D. Thomas, “High energy colliders as black hole factories: The end of short distance physics,” Phys. Rev. D 65 (2002) 056010 [arXiv:hep-ph/0106219]. * [8] S. Dimopoulos and G. L. Landsberg, “Black holes at the LHC,” Phys. Rev. Lett. 87 (2001) 161602 [arXiv:hep-ph/0106295]. * [9] K. Oda and N. Okada, “Alternative signature of TeV strings,” Phys. Rev. D 66, 095005 (2002) [arXiv:hep-ph/0111298]. * [10] D. M. Eardley and S. B. Giddings, “Classical black hole production in high-energy collisions,” Phys. Rev. D 66 (2002) 044011 [arXiv:gr-qc/0201034]. * [11] H. Yoshino and Y. Nambu, “Black hole formation in the grazing collision of high-energy particles,” Phys. Rev. D 67 (2003) 024009 [arXiv:gr-qc/0209003]. * [12] H. Yoshino and V. S. Rychkov, “Improved analysis of black hole formation in high-energy particle collisions,” Phys. Rev. D 71 (2005) 104028 [arXiv:hep-th/0503171]. * [13] G. ’t Hooft, “Graviton Dominance in Ultrahigh-Energy Scattering,” Phys. Lett. B 198 (1987) 61. * [14] S. Dimopoulos and R. Emparan, “String balls at the LHC and beyond,” Phys. Lett. B 526 (2002) 393 [arXiv:hep-ph/0108060]. * [15] Y. Nambu, private communication. * [16] D. Ida, K. Oda and S. C. Park, “Rotating black holes at future colliders: Greybody factors for brane fields,” Phys. Rev. D 67, 064025 (2003) [Erratum-ibid. D 69, 049901 (2004)] [arXiv:hep-th/0212108]. * [17] G. Veneziano, “String-theoretic unitary S-matrix at the threshold of black-hole production,” JHEP 0411, 001 (2004) [arXiv:hep-th/0410166]. * [18] G. Veneziano, “Towards a unitary S-matrix description of black-hole formation and decay in string theory,” AIP Conf. Proc. 861, 39 (2006). * [19] T. Yoneya, “String theory and space-time uncertainty principle,” Prog. Theor. Phys. 103 (2000) 1081 [arXiv:hep-th/0004074]. * [20] D. J. Gross and P. F. Mende, “The High-Energy Behavior of String Scattering Amplitudes,” Phys. Lett. B 197 (1987) 129. * [21] D. J. Gross and P. F. Mende, “String Theory Beyond the Planck Scale,” Nucl. Phys. B 303 (1988) 407. * [22] T. Kuroki and T. Matsuo, “Production cross section of rotating string,” Nucl. Phys. B 798 (2008) 291 [arXiv:0712.4062 [hep-th]]. * [23] M. Soldate, “Partial Wave Unitarity and Closed String Amplitudes,” Phys. Lett. B 186 (1987) 321. * [24] I. J. Muzinich and M. Soldate, “High-Energy Unitarity of Gravitation and Strings,” Phys. Rev. D 37 (1988) 359. * [25] Y. Nambu and P. Frampton, “Asymptotic behavior of partial widths in the Veneziano model of scattering amplitudes,” Published in Quanta: Essays in Theoretical Physics Dedicated to Gregor Wentzel. Chicago, Ill., Univ. of Chicago Press, 1970. pp. 403-414. * [26] F. L. Lin, T. Matsuo and D. Tomino, “Hagedorn Strings and Correspondence Principle in AdS(3),” JHEP 0709 (2007) 042 [arXiv:0705.4514 [hep-th]]. * [27] P. C. Aichelburg and R. U. Sexl, “On the Gravitational field of a massless particle,” Gen. Rel. Grav. 2 (1971) 303. * [28] S. W. Hawking and G. F. R. Ellis, “The Large scale structure of space-time,” Cambridge University Press, Cambridge, 1973 * [29] S. B. Giddings and V. S. Rychkov, “Black holes from colliding wavepackets,” Phys. Rev. D 70 (2004) 104026 [arXiv:hep-th/0409131]. * [30] L. A. Anchordoqui, J. L. Feng, H. Goldberg and A. D. Shapere, “Black holes from cosmic rays: Probes of extra dimensions and new limits on TeV-scale gravity,” Phys. Rev. D 65, 124027 (2002) [arXiv:hep-ph/0112247]. * [31] R. Emparan and R. C. Myers, “Instability of ultra-spinning black holes,” JHEP 0309 (2003) 025 [arXiv:hep-th/0308056]. * [32] L. D. McLerran, “The color glass condensate and small x physics: 4 lectures,” Lect. Notes Phys. 583 (2002) 291 [arXiv:hep-ph/0104285]. * [33] E. Iancu and R. Venugopalan, “The color glass condensate and high energy scattering in QCD,” arXiv:hep-ph/0303204. * [34] D. Amati, M. Ciafaloni and G. Veneziano, “Superstring Collisions at Planckian Energies,” Phys. Lett. B 197 (1987) 81. * [35] D. Amati, M. Ciafaloni and G. Veneziano, “Classical and Quantum Gravity Effects from Planckian Energy Superstring Collisions,” Int. J. Mod. Phys. A 3 (1988) 1615. * [36] S. B. Giddings, D. J. Gross and A. Maharana, “Gravitational effects in ultrahigh-energy string scattering,” Phys. Rev. D 77 (2008) 046001 [arXiv:0705.1816 [hep-th]]. * [37] D. Chialva, R. Iengo and J. G. Russo, “Cross sections for production of closed superstrings at high energy colliders in brane world models,” Phys. Rev. D 71 (2005) 106009 [arXiv:hep-ph/0503125].	arxiv-papers	2008-08-27T07:56:02	2024-09-04T02:48:57.481486	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Toshihiro Matsuo and Kin-ya Oda", "submitter": "Kin-ya Oda", "url": "https://arxiv.org/abs/0808.3645" }
0808.3646	# Finite type invariants of words and Arnold’s invariants Mikihiro Fujiwara Department of Mathematical and Computing Sciences, Tokyo Institute of Technology, Tokyo 152-8552, Japan fujiwar5@is.titech.ac.jp ###### Abstract. We define a new finite type invariant for stably homeomorphic class of curves on compact oriented surfaces without boundaries and extend to a regular homotopy invariant for spherical curves. ###### Key words and phrases: immersed curves, finite type invariant, words ## 1\. Introduction This paper is motivated by N. Ito’s work in [3] which is based on the following three existing theories. First of all, V. I. Vassiliev developed in [8] a theory of finite type invariants of knots, also known as Vassiliev invariants, which is conjectured to classify knots. Secondly, V. Turaev suggested that words and their topology could be an effective coding for virtual knots in [6]. And finally, V. I. Arnold developed a theory of invariants for generic curves on the plane in [1, 2] which obeys some rules under regularly homotopic change. We should note that M. Polyak reconstructed Arnold’s invariants in terms of finite type invariants of plane curves. Ito defined finite type invariants of words called $SCI_{n}$ in [3], and showed that they become a complete invariant for stable homeomorphism classes and also that Arnold’s invariants are of first order. The purpose of this paper is to provide a very simple frame to define finite type invariants of words by introducing a new type of intersection, called singular intersection, which plays intermediate role between an actual and virtual intersections, and show that our finite type invariants contain all the information Ito’s finite type invariants have, and provide a clearer relation to Arnold’s invariant. The papers consists of 4 parts. The first part (Sections 2 - 4) is devoted to the basics. We define curves, signed words and study the relation between them. In the second part (Section 5) we introduce the new type of intersection, called singular intersection. In the third part (Sections 6 - 7) we define a new type of finite type invariants of words and discuss the structure. In the fourth part (Sections 8) we show how to extend finite type invariants of spherical curves and study relation to Arnold’s invariants. ## 2\. Curves A curve will be a generic immersion from an oriented circle to a closed oriented surface. Here generic means that the curve has only finite set of self-intersections, which are all transversally double points. Triple points and self-tangencies are not allowed. Every curve has a regular neighborhood. This is a narrow immersed band whose core is the curves in the surface. A pointed curve is a curve endowed with a base point, which is not on the singular points. Two curves are stably homeomorphic if there is a homeomorphism between their regular neighborhoods mapping the first curve onto the second one preserving the orientation of the curves and the surfaces. Similarly, two pointed curves are pointed stably homeomorphic if two curves are stably homeomorphic preserving their base points. ## 3\. Signed words Let $\mathcal{A}$ be a set of letters. A word of length $n$ in $\mathcal{A}$ is a mapping from ${\hat{n}}=\\{1,2,3,...,n\\}$ to $\mathcal{A}$ . A Gauss word is a word where each letter of $\mathcal{A}$ appears in the word exactly two times. A Gauss word is a signed word if the letters in the associated word naturally acquire signs $\pm$. To simplify notation, we omit signs and put bars on the letters of minuses. Thus instead of writing $A^{+}B^{-}A^{+}B^{-}$ we write $A\bar{B}A\bar{B}$ . By convention, there is one empty word $\phi$ of length $0$. Two signed words $w$ in $\mathcal{A}$ , $w^{\prime}$ in $\mathcal{A}^{\prime}$ are isomorphic if there is a bijection $f:\mathcal{A}\to\mathcal{A}^{\prime}$ such that $w^{\prime}=fw$ . Two signed words are isomorphic if two words are isomorphic preserving sings of letters. We denote by $\mathcal{W}_{n}$ the set generated by all the isomorphism classes of signed words of length $2n$. Put $\mathcal{W}=\cup_{n\geq 0}\mathcal{W}_{n}$. The isomorphism classes $\mathcal{W}$ of signed word bijectively correspond to stable homeomorphism classes of pointed curves on a surface [6]. Sub-words of a signed word are given by eliminating some letters from it. Sub- words of signed words on $\mathcal{A}$ given by eliminating letters $\mathcal{B}\subset\mathcal{A}$ are singed words on $\mathcal{A}-\mathcal{B}$. ###### Definition 3.1. The cyclic equivalence $\sim$ of signed words is defined by the relation by generated by $A^{\pm}xA^{\pm}y\sim xA^{\mp}yA^{\mp}$ where $A$ is a letter, $xy$ is a sub-word. Let us denote cyclic equivalence class of signed word by $W_{n}/\sim$, which bijectively corresponds to stable homeomorphism class of curves on surface [6]. ## 4\. Curves as signed words Consider a pointed curve on a surface. Label its crossings in an arbitrary way by different letters $A_{1},A_{2},...,A_{n}$ where $n$ is the number of the crossings. The Gauss word of the curve is obtained by moving along the curve starting at the base point and writing down the letters as we encounter them, finishing when we get back to the base point. The resulting word, $w$ on the alphabet $\mathcal{A}=\\{A_{1},A_{2},...,A_{n}\\}$ contains every letter twice. To make $w$ a signed word, consider the crossing of the curve labeled $A_{i}$. If, when moving along the curve as above, we first traverse this crossing from the bottom-left to the top-right, then $A_{i}^{-}$, otherwise $A_{i}^{+}$, see Figure 1, where the orientation of the ambient surface is counterclockwise. The trivial curve corresponds to the signed word $\phi$. We assign to our curve the class of this signed word in $\mathcal{W}_{n}$. A different choice of the labeling of the crossing gives an isomorphic signed word. If the curve is changed by a stable homeomorphism, then the associated signed word does not change, since it is defined by the behavior of the curve in its regular neightborhood. Figure 1. On the left $A_{i}^{-}$ and on the right $A_{i}^{+}$ ## 5\. Singular curves ### 5.1. Virtual crossings We may allow virtual (or unlabeled) crossings, and denote it by a crossing with a small circle around it. Such crossings do not contribute at all to the associated signed word. The idea is not that a virtual crossing is just an ordinary graphical vertex, but rather that the virtual crossing is not actual crossings. If we draw a non-planar curve in the plane, it necessarily acquires virtual crossings. These crossings are not a part of the structure of the curve itself. They are artifacts of the drawn of the graph in the plane. Let $c$ be a curve on surface, and $w$ be a signed word corresponding to $c$. Getting sub-words of $w$ is equal to changing double points of $c$ to virtual crossings. ### 5.2. Singular crossings We introduce the crossings called singular crossings as intermediate crossings between actual crossings and virtual crossings, denote as a crossing with a small box around it. A singular curve is a curve with singular crossings. We call a signed word corresponding to a singular curve a singular signed word. To get signed words corresponding to singular curves, label a singular crossing by the letter $A_{i}^{}$ instead of $A_{i}$. The letters such as $A^{}$(or $\bar{A^{}}$) are called singular letters. For a signed word $w$ without singular letter, denote $w^{}$ the word changing all letters of $w$ to singular letters saving signs. Figure 2. The singular curve corresponding to the signed word $\bar{A}^{}B\bar{A}^{}B$ ## 6\. Definition of finite type invariants Let $\mathcal{X}_{n}$ denote the set of isomorphism class of singular signed words with $n$ singular letters. By definition, $\mathcal{X}_{0}=\mathcal{W}$ is the set of non-singular signed words. Put $\mathcal{X}=\cup_{n\geq 0}\mathcal{X}_{n}$. ###### Definition 6.1. A cyclic equivalence invariant $v$ of signed words is a function of a signed word in a field $\mathbb{F}$, e.g. $\mathbb{Q}$, $\mathbb{R}$ or $\mathbb{C}$, which takes equal values on cyclically equivalent signed words. In other words, it is a mapping $v$ : $\mathcal{W}/\sim\to\mathbb{F}$ from $\mathcal{W}$ to $\mathbb{F}$. ###### Definition 6.2. Given a cyclic equivalence invariant $v$, we define its prolongation $\tilde{v}$ by the rule $\displaystyle\tilde{v}\mid_{\mathcal{X}_{0}}$ $\displaystyle=$ $\displaystyle v$ $\displaystyle\tilde{v}(A^{}xA^{}y)$ $\displaystyle=$ $\displaystyle\tilde{v}(AxAy)-\tilde{v}(xy).$ ###### Definition 6.3. A function $v$ is a finite type invariant of order $n$ if its prolongation vanishes on all singular words with more than $n$ singular letters: $v(\mathcal{X}^{n+1})=0$ (and hence $v(\mathcal{X}^{m})=0$ for all $m>n$). Notation: $\mathcal{V}_{n}$ stands for the set of $\mathbb{F}$-valued invariants of order $\geq n$, it will be vector space over $\mathbb{F}$. ## 7\. Structure of finite type invariant of words ###### Proposition 7.1. $\mathcal{V}_{0}=\\{{\rm consts}\\}$ and hence dim $\mathcal{V}_{0}=1$. ###### Proof. By definition, $v\in\mathcal{V}_{0}$ means that the value of $v$ on any signed word with at least one singular letter is $0$. Due to skein relation $v(AxAy)-v(xy)=v(A^{}xA^{}y)=0$ which means that the value of $v$ does not change when any letters adds (or removes). Therefore, $v$ is a constant function. ∎ ###### Proposition 7.2. dim $\mathcal{V}_{1}=2$. ###### Proof. By a similar argument as above, among singular signed word $w\in\mathcal{X}_{1}$ the value is invariant under adding (or removing) letters, and depends only on the value of $(AA)^{}$ in this case. Ont the other hand, once we choose the value of $(AA)^{}$, we get a finite type invariant of degree $1$. It follows from an easy argument that a finite type invariant $v$ of degree $1$ on any signed word $w$ can be presented by linear sum, $v(w)=a_{0}v(\phi)+a_{1}(w)v((AA)^{}),$ where $a_{0}$ is a constant and $a_{1}(w)$ is the number of letters, which is a finite type invariant of degree $1$. ∎ ###### Lemma 7.3. Let $w_{n}$ be a signed words of length $2n$ ($n=0,1,2,...$), and $v$ is a finite type invariant of degree $1$. Then, $v(w_{n})=v(\phi)+nv((AA)^{}).$ ###### Proof. The statement is true for $n=0$, since $v(w_{0})=v(\phi)$. Assume the statement is true for $n=m$, i.e., $v(w_{m})=v(\phi)+mv((AA)^{}).$ Then by the equation a letter of $w_{m+1}$, $v(A^{}xA^{}y)=v(AxAy)-v(xy)$,we have $v(w_{m+1})=v((AA)^{})+v(w_{m})=v((AA)^{})+mv((AA)^{})=(m+1)v((AA)^{}).$ Thus we have $v(w_{n})=v(\phi)+nv((AA)^{}).$ ∎ ###### Proposition 7.4. dim $\mathcal{V}_{2}=6$. ###### Proof. By a similar argument as above, among singular signed word $w\in\mathcal{X}_{2}$ the value is invariant under adding (or removing) letters; there is four possible values, $v((AABB)^{})$,$v((AA\bar{B}\bar{B})^{})$,$v((A\bar{B}\bar{B}A)^{})$,$v((ABAB)^{})$ in this case, since there is no relation. The value of a finite type invariant of degree $2$ on any signed word $w$ can be presented by linear sum, $\displaystyle v(w)=a_{0}v(\phi)+a_{1}(w)v((AA)^{})+b_{1}(w)v((AABB)^{})+b_{2}(w)v((AA\bar{B}\bar{B})^{})$ $\displaystyle+b_{3}(w)v((A\bar{B}\bar{B}A)^{})+b_{4}(w)v((ABAB)^{}).$ ∎ ###### Theorem 7.5. A finite type invariant is a complete invariant for cyclic equivalence class of signed word. ###### Proof. Let $w$ be a signed word of length $2n$, and $v$ be a finite type invariant of degree $n$. The finite type invariant $v(w)$ has the information of all sub- words of $w$. ∎ ###### Theorem 7.6. A finite type invariant is a complete invariant for stable homeomorphism classes of curves on compact oriented surfaces, without boundaries. ###### Proof. The facts that each cyclic equivalence class of signed words is corresponding to a stably homeomorphism class of curves and Theorem 7.5 imply this theorem. ∎ Also, following lemma derives from the definition of invariant $SCI_{n}$ in [3], where $SCI_{n}$ is a invariant for immersed curves. ###### Lemma 7.7. A finite type invariant of degree $n$ is equivalent to $\sum_{k=0}^{n}SCI_{k}$. ###### Proof. Let $\Gamma$ be a immersed curve with $n$ crossings, and $w$ be a signed word corresponding to $\Gamma$. $SCI_{m}(\Gamma)$ has the information of all sub- words of $w$ of length $2m$. $\sum_{k=0}^{n}SCI_{n}(\Gamma)$ has the information of all sub-words of $w$, so does a finite type invariant $v(w)$. They are complete invariants. ∎ ## 8\. Finite type inveriants for spherical curves Let $\mathcal{W}_{s}$ be a set $\\{w\in\mathcal{W}\mid w$ represents a curve on the sphere$\\}$ and $\mathcal{W}_{s}^{}$ be $\mathcal{W}_{s}/\sim\cup(\mathcal{W}-\mathcal{W}_{s})$. ###### Definition 8.1. A signed word invariant for $\mathcal{W}_{s}^{}$ is a function of a signed word which takes equal values on cyclically equivalent signed words corresponding to spherical curves. In other words, it is a mapping $v:\mathcal{W}_{s}^{}\to\mathbb{F}$ from $\mathcal{W}_{s}^{}$ to a field $\mathbb{F}$. Let $v$ be a finite type invariant of degree 2. For a signed word $w$ for curves on the sphere, $v$ can be presented as a linear combination as follows, $\displaystyle v(w)$ $\displaystyle=$ $\displaystyle a_{0}v(\phi)+a_{1}(w)v((AA)^{})+b_{1}(w)v((AABB)^{})$ $\displaystyle+b_{2}(w)v((AA\bar{B}\bar{B})^{})+b_{3}(w)v((A\bar{B}\bar{B}A)^{})+c_{1}(w)v((ABAB)^{})$ $\displaystyle+c_{2}(w)v((A\bar{B}A\bar{B})^{})+c_{3}(w)v((\bar{A}B\bar{A}B)^{})+c_{4}(w)v((\bar{A}\bar{B}\bar{A}\bar{B})^{}).$ ###### Definition 8.2 (Regular homotopy of signed words). Three regular homotopy moves on signed words are defined as follows. 1. (1) $xy\to A^{\pm}B^{\mp}xA^{\pm}B^{\mp}y$ . $x$,$y$ are words without letters $A$,$B$. Note that if $xy$ is a signed word on $\mathcal{A}$, then $A^{\pm}B^{\mp}xA^{\pm}B^{\mp}y$ is a signed word on $\mathcal{A}\cup\\{A,B\\}$. 2. (2) $xy\to A^{\pm}B^{\mp}xB^{\mp}A^{\pm}y$. $x$,$y$ are words without letters $A$,$B$. 3. (3) $A^{\pm}B^{\pm}xA^{\pm}C^{\pm}yB^{\pm}C^{\pm}z\to B^{\pm}A^{\pm}xC^{\pm}A^{\pm}yC^{\pm}B^{\pm}z$. $x$,$y$,$z$ are words without letters $A$,$B$,$C$. These moves correspond to moves of curves illustrated in Figure 3. Figure 3. Regular homotopy moves The invariants $J_{s}^{+}$,$J_{s}^{-}$,$St_{s}$ of generic spherical curves were introduced in [4]. These invariants are additive with respect to connected sum of curves and independent of the choice of orientation for the curve. $J_{s}^{\pm}$ and $St_{s}$ are defined by their behavior the moves and inverse moves in Figure 3. ###### Definition 8.3 (Arnold’s basic invariants). $J_{s}^{+}$ does not change under the second and third regular homotopy moves or inverse moves but increases (resp. decrease) by $2$ under the first regular homotopy move (resp. inverse move). $J_{s}^{-}$ does not change under the first and third regular homotopy moves or inverse moves but decrease (resp. increase) by $2$ under the second regular homotopy move (resp. inverse move). $St_{s}$ does not change under the first and second regular homotopy moves or inverse moves but increase (resp. decrease) $1$ under the third regular homotopy move (resp. inverse move). Normalizing conditions for Arnold fs invariants are given as follow. Let $w_{i}$ be a signed word $A_{1}A_{1}A_{2}A_{2}...A_{i}A_{i}$ (or $\bar{A_{1}}\bar{A_{1}}\bar{A_{2}}\bar{A_{2}}...\bar{A_{i}}\bar{A_{i}}$) corresponding to a plane curve $K_{i}$ , $i=1,2,...$, $\displaystyle J_{s}^{+}(K_{i})$ $\displaystyle=$ $\displaystyle\frac{(i-1)^{2}}{2},$ $\displaystyle J_{s}^{-}(K_{i})$ $\displaystyle=$ $\displaystyle\frac{(i-2)^{2}}{2}-\frac{3}{2},$ $\displaystyle St_{s}(K_{i})$ $\displaystyle=$ $\displaystyle-\frac{(i-1)^{2}}{4}.$ ###### Theorem 8.4. Arnold’s basic invariants are finite type invariants of degree $2$. ###### Proof. Let $v_{s}$ be a finite type invariant of signed words for spherical curves of degree $2$, and $v_{s}(\phi)=0,\ v_{s}((AA)^{})=p,$ $v_{s}((AABB)^{})=v_{s}((A\bar{B}\bar{B}A)^{})=-v_{s}((AA\bar{B}\bar{B})^{})=q,$ $v_{s}((ABAB)^{})=v_{s}((\bar{A}\bar{B}\bar{A}\bar{B})^{})=-v_{s}((A\bar{B}A\bar{B})^{})=-v_{s}((\bar{A}B\bar{A}B)^{})=r,$ Let $\Gamma$ be a curve corresponding to a signed word $w$. By using Polyak’s formulation of the Arnold’s basic invariants [4], we represent the Arnold’s basic invariants as $\displaystyle J_{s}^{+}(\Gamma)$ $\displaystyle=$ $\displaystyle v_{s}(w)\hskip 10.00002pt(p=-\frac{1}{2},\ q=1,\ r=-3),$ $\displaystyle J_{s}^{-}(\Gamma)$ $\displaystyle=$ $\displaystyle v_{s}(w)\hskip 10.00002pt(p=-\frac{3}{2},\ q=1,\ r=-3),$ $\displaystyle St_{s}(\Gamma)$ $\displaystyle=$ $\displaystyle v_{s}(w)\hskip 10.00002pt(p=\frac{1}{4},\ q=-\frac{1}{2},\ r=\frac{1}{2}).$ That is to say, the finite type invariants $v_{s}$ parameterized by three specific vectors $(p,q,r)=$ $(-\frac{1}{2},1,-3)$,$(-\frac{3}{2},1,-3)$,$(\frac{1}{4},-\frac{1}{2},\frac{1}{2})$ are equivalent to the Arnold’s basic invariants $J_{s}^{+}$, $J_{s}^{-}$ and $St_{s}$. ∎ The invariant can be extended to a invariant $J^{\pm}$,$St$ for plane curves if rotation number $i$ of curves is defined, by $J^{\pm}=J_{s}^{\pm}-\frac{1}{2}i^{2}$ , $St=St_{s}+\frac{1}{4}i^{2}$. ## References [1] V. I. Arnold, Plane curves, their invariants, perestroikas and classifications, Adv. Soviet Math. 21, Amer. Math. Soc., Providence, RI (1994), 33–91. * [2] V. I. Arnold, Topological invariants of plane curves and caustics, Univ. Lecture Ser. 5, Amer. Math. Soc., Providence, RI (1994). * [3] N. Ito, On a finite type invariant giving complete classification of curves on surface, Math.GT/0803.2082. * [4] M. Polyak, Invariants of curves and fronts via Gauss diagrams, Topology 37 (1998), 989–1009. * [5] V. Turaev, Curves on surface, charts, and words, Geom. Dedicata 116 (2005), 203–236. * [6] V. Turaev, Knots and words, Int. Math. Res. Not. (2006), Art. ID 84098, 1–23. * [7] V. Turaev, Lecture on topology of words, Jpn. J. Math. 2 (2007) 1–39 * [8] V. A. Vassilliev, Cohomology of knot spaces, Theory of singularities and its applications, Adv. Soviet Math. 1, Amer. Math. Soc. , Providence, RI (1990), 23–69.	arxiv-papers	2008-08-27T07:56:52	2024-09-04T02:48:57.488916	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Fujiwara", "submitter": "Mikihiro Fujiwara", "url": "https://arxiv.org/abs/0808.3646" }
0808.3650	# Study of intermediate two-body decays in $\bar{B}^{0}\to\Sigma_{c}(2455)^{0}\bar{p}\pi^{+}$ H. O. Kim hokim@knu.ac.kr H. Kichimi kichimi@post.kek.jp Corresponding author. I. Adachi H. Aihara K. Arinstein V. Aulchenko T. Aushev A. M. Bakich V. Balagura I. Bedny A. Bondar A. Bozek M. Bračko T. E. Browder P. Chang Y. Chao B. G. Cheon R. Chistov I.-S. Cho Y. Choi J. Dalseno M. Dash S. Eidelman D. Epifanov N. Gabyshev H. Ha K. Hayasaka M. Hazumi D. Heffernan Y. Hoshi W.-S. Hou H. J. Hyun T. Iijima K. Inami A. Ishikawa R. Itoh M. Iwasaki Y. Iwasaki D. H. Kah J. H. Kang N. Katayama H. Kawai T. Kawasaki H. J. Kim Y. I. Kim Y. J. Kim S. Korpar P. Križan P. Krokovny R. Kumar A. Kuzmin Y.-J. Kwon J. S. Lee M. J. Lee T. Lesiak S.-W. Lin D. Liventsev A. Matyja S. McOnie K. Miyabayashi H. Miyata Y. Miyazaki M. Nakao H. Nakazawa Z. Natkaniec S. Nishida O. Nitoh S. Ogawa T. Ohshima S. Okuno C. W. Park H. Park H. K. Park K. S. Park L. S. Peak R. Pestotnik L. E. Piilonen A. Poluektov H. Sahoo Y. Sakai O. Schneider K. Senyo M. E. Sevior M. Shapkin V. Shebalin J.-G. Shiu B. Shwartz J. B. Singh A. Somov S. Stanič M. Starič T. Sumiyoshi M. Tanaka G. N. Taylor Y. Teramoto S. Uehara Y. Unno S. Uno P. Urquijo Y. Usov G. Varner A. Vinokurova C. H. Wang M.-Z. Wang P. Wang X. L. Wang Y. Watanabe E. Won Y. Yamashita M. Yamauchi Z. P. Zhang V. Zhilich V. Zhulanov T. Zivko A. Zupanc O. Zyukova Budker Institute of Nuclear Physics, Novosibirsk, Russia Chiba University, Chiba, Japan University of Cincinnati, Cincinnati, OH, USA T. Kościuszko Cracow University of Technology, Krakow, Poland The Graduate University for Advanced Studies, Hayama, Japan Hanyang University, Seoul, South Korea University of Hawaii, Honolulu, HI, USA High Energy Accelerator Research Organization (KEK), Tsukuba, Japan Institute of High Energy Physics, Chinese Academy of Sciences, Beijing, PR China Institute for High Energy Physics, Protvino, Russia Institute for Theoretical and Experimental Physics, Moscow, Russia J. Stefan Institute, Ljubljana, Slovenia Kanagawa University, Yokohama, Japan Korea University, Seoul, South Korea Kyungpook National University, Taegu, South Korea École Polytechnique Fédérale de Lausanne, EPFL, Lausanne, Switzerland Faculty of Mathematics and Physics, University of Ljubljana, Ljubljana, Slovenia University of Maribor, Maribor, Slovenia University of Melbourne, Victoria, Australia Nagoya University, Nagoya, Japan Nara Women’s University, Nara, Japan National Central University, Chung-li, Taiwan National United University, Miao Li, Taiwan Department of Physics, National Taiwan University, Taipei, Taiwan H. Niewodniczanski Institute of Nuclear Physics, Krakow, Poland Nippon Dental University, Niigata, Japan Niigata University, Niigata, Japan University of Nova Gorica, Nova Gorica, Slovenia Osaka City University, Osaka, Japan Osaka University, Osaka, Japan Panjab University, Chandigarh, India Saga University, Saga, Japan University of Science and Technology of China, Hefei, PR China Seoul National University, Seoul, South Korea Sungkyunkwan University, Suwon, South Korea University of Sydney, Sydney, NSW, Australia Toho University, Funabashi, Japan Tohoku Gakuin University, Tagajo, Japan Department of Physics, University of Tokyo, Tokyo, Japan Tokyo Metropolitan University, Tokyo, Japan Tokyo University of Agriculture and Technology, Tokyo, Japan Virginia Polytechnic Institute and State University, Blacksburg, VA, USA Yonsei University, Seoul, South Korea ###### Abstract We present results of a detailed study of the three-body $\bar{B}^{0}\to\Sigma_{c}(2455)^{0}\bar{p}\pi^{+}$ decay. A significant enhancement of signal events is observed in the $\bar{p}\pi^{+}$ mass system near $1.5\,{\rm GeV}/c^{2}$ that is consistent with the presence of an intermediate baryonic resonance $\bar{N}^{0}$, where $\bar{N}^{0}$ is the $\bar{N}(1440)^{0}P_{11}$ or $\bar{N}(1535)^{0}S_{11}$ state, or an admixture of the two states. We measure the product ${\cal B}(\bar{B}^{0}\to\Sigma_{c}(2455)^{0}\bar{N}^{0})\times{\cal B}(\bar{N}^{0}\to\bar{p}\pi^{+})$=($0.80\pm 0.15\rm{(stat.)}\pm 0.14\rm{(syst.)}\pm 0.21$)$\times 10^{-4}$, where the last error is due to the uncertainty in ${\cal B}(\Lambda_{c}^{+}\to pK^{-}\pi^{+})$. The significance of the signal is 6.1 standard deviations. This analysis is based on a data sample of 357 fb-1, accumulated at the $\Upsilon(4S)$ resonance with the Belle detector at the KEKB asymmetric-energy $e^{+}e^{-}$ collider. ###### keywords: B-meson, Charmed baryon, $N^{0}$ $P_{11}$ and $S_{11}$ resonaces. ###### PACS: 13.20.H ††journal: Physics Letters B Belle Preprint 2008-23 KEK Preprint 2008-23 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , and , Various charmed baryonic $B$ decays into four-, three- and two-body final states have been reported [1, 2, 3, 4, 5, 6, 7], and the measured branching fractions show clearly that the branching fraction increases with the multiplicity of the final state [8, 3]. To understand this hierarchy, it is interesting to study decays of $\bar{B}^{0}\to\Lambda_{c}^{+}{\bar{p}}\pi^{+}\pi^{-}$ into three- and two- body final states. The branching fractions are predicted from CKM matrix elements [9], while the form factors of the decay vertices depend on the decay mechanism. Experimental studies provide stringent constraints on the theoretical models [10, 11, 12]. In this report, we perform a detailed study of the intermediate three-body decay $\bar{B}^{0}\to\Sigma_{c}(2455)^{0}\bar{p}\pi^{+}$ observed in the previous analysis of $\bar{B}^{0}\to\Lambda_{c}^{+}\bar{p}\pi^{+}\pi^{-}$ [1], using a data sample of $388\times 10^{6}$ $B\bar{B}$ events, corresponding to 357 fb-1 accumulated at the $\Upsilon(4S)$ resonance with the Belle detector at the KEKB asymmetric-energy $e^{+}e^{-}$ collider [13]. The Belle detector is a large-solid-angle spectrometer based on a 1.5 Tesla superconducting solenoid magnet. It consists of a silicon vertex detector (SVD) (a three-layer SVD for the first sample of $(152.0\pm 1.2)\times 10^{6}$ $B\bar{B}$ events and a four-layer SVD for the latter $(235.8\pm 3.6)\times 10^{6}$ $B\bar{B}$ events), a 50-layer central drift chamber (CDC), an array of aerogel threshold Cherenkov counters (ACC), a barrel-like arrangement of time-of-flight scintillation counters (TOF), and an electromagnetic calorimeter (ECL) comprised of CsI (Tl) crystals located inside the superconducting solenoid coil. An iron flux return located outside the coil is instrumented to detect $K_{L}^{0}$ mesons and to identify muons (KLM). The detector is described in detail elsewhere [14]. We simulate the detector response and estimate the efficiency for signal reconstruction by Monte Carlo simulation (MC). We use the EvtGen program [15] for signal event generation and a GEANT-based [16] detector simulation program to model the Belle detector response for the signal. We first describe briefly the previous analysis of ${\bar{B}}^{0}\to\Lambda_{c}^{+}{\bar{p}}\pi^{+}\pi^{-}$ [1], We select $\bar{B}^{0}\to\Lambda_{c}^{+}\bar{p}\pi^{+}\pi^{-}$ events by reconstructing $\Lambda_{c}^{+}\to pK^{-}\pi^{+}$ decays, using charged tracks reconstructed by the SVD and CDC, and hadron identification information (such as protons, kaons and pions) provided from the CDC $dE/dx$, TOF and ACC (PID) [17], and ECL and KLM information to veto electron and $\mu$ tracks. Charge-conjugate modes are implicitly included throughout this paper unless noted otherwise. After the event selection, we fit the $\Delta{E}$ distribution for the $B$ candidate events with $5.27\,{\rm GeV}/c^{2}<M_{\rm bc}<5.29\,{\rm GeV}/c^{2}$, with a double Gaussian fixed to the signal MC shape ($\sigma_{\rm core}=7\,{\rm MeV}/c^{2}$, $\sigma_{\rm tail}=16\,{\rm MeV}/c^{2}$) plus a linear background. The variable $\Delta{E}=E_{B}-E_{\rm{beam}}$ is the difference between the reconstructed $B$ meson energy ($E_{B}$) and the beam energy ($E_{\rm beam}$) evaluated in the center-of-mass system (CMS), while $M_{\rm bc}=\sqrt{E_{\rm beam}^{2}-P_{B}^{2}}$ is the beam-energy-constrained $B$ meson mass and ${P_{B}}$ is the momentum of the $B$ meson also evaluated in the CMS. We obtain a $B$ signal of $1400\pm 49$ events for ${\bar{B}}^{0}\to\Lambda_{c}^{+}{\bar{p}}\pi^{+}\pi^{-}$. Figure 1 shows the $\Lambda_{c}^{+}\pi^{+}$ and $\Lambda_{c}^{+}\pi^{-}$ mass distributions for the events in the $B$ signal region $\|\Delta{E}\|<0.03{\,\rm GeV}$ ($\pm 4\sigma$) and $5.27{\,\rm GeV}/c^{2}<M_{\rm bc}<5.29{\,\rm GeV}/c^{2}$ ($\pm 4\sigma$). We focus our discussion on the $\Sigma_{c}(2455)^{++}$ and $\Sigma_{c}(2455)^{0}$ resonances clearly observed in figure [1]; $(182\pm 15)$ events for $\bar{B}^{0}\rightarrow\Sigma_{c}(2455)^{++}\bar{p}\pi^{-}$ and $(122\pm 14)$ events for $\bar{B}^{0}\rightarrow\Sigma_{c}(2455)^{0}\bar{p}\pi^{+}$, corresponding to branching fractions of ($2.1\pm 0.2\rm{(stat.)}\pm 0.3\rm{(syst.)}\pm 0.5$)$\times 10^{-4}$ and ($1.4\pm 0.2\rm{(stat.)}\pm 0.2\rm{(syst.)}\pm 0.4$)$\times 10^{-4}$, respectively. Hereafter, we denote $\Sigma_{c}(2455)$ as $\Sigma_{c}$. Figure 1: The mass distributions of (a) $\Lambda_{\mathrm{c}}^{+}\pi^{+}$ and (b) $\Lambda_{\mathrm{c}}^{+}\pi^{-}$ in $\bar{B}^{0}\rightarrow\Lambda_{\mathrm{c}}^{+}\bar{p}\pi^{+}\pi^{-}$. The points with error bars show the mass distribution for the events in the $B$ signal box, and the shaded histogram indicates that for the background. The solid and dashed curves represent the signal and the background, respectively, obtained from a simultaneous binned likelihood fit. Figure 2 shows (a) the Dalitz plot and (b) the $M^{2}(\bar{p}\pi^{-})$ distribution for the $\bar{B}^{0}\to\Sigma_{c}^{++}\bar{p}\pi^{-}$ events, and (c) the Dalitz plot and (d) the $M^{2}(\bar{p}\pi^{+})$ distribution for the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ events. Here we require the $\Sigma_{c}$ candidates satisfy the invariant mass requirement $2.447\,{\rm GeV}/c^{2}<M(\Lambda_{c}^{+}\pi^{\pm})<2.461\,{\rm GeV}/c^{2}$ ($\pm 2\sigma$). We find that the $M^{2}(\bar{p}\pi^{-})$ distribution for $\bar{B}^{0}\rightarrow\Sigma_{c}^{++}{\bar{p}}\pi^{-}$ is consistent with three-body phase space, while the $M^{2}(\bar{p}\pi^{+})$ distribution for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ has a significant peak. In what follows, we present a detailed study of the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ decay. Figure 3 shows the $\Delta{E}$ distribution for the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ events, which are selected from the $\bar{B}^{0}\to\Lambda_{c}^{+}\bar{p}\pi^{+}\pi^{-}$ sample with the additional requirement that the $\Lambda_{c}^{+}\pi^{-}$ mass be consistent with the $\Sigma_{c}^{0}$. The curves show fits to the data with a double Gaussian function with shape parameters fixed to the values from signal MC and a linear background. We obtain a $B$ signal yield of $(102\pm 11)$ events and a background of $(17\pm 3)$ events. The signal reduction of 16% is consistent with the MC estimation of the effect due to the ${\Sigma_{c}^{0}}$ mass requirement. We estimate a non-$\Sigma_{c}^{0}$ background of $(8\pm 4)$ events from a fit to the $\Delta{E}$ distribution in the $\Sigma_{c}$ mass sideband $2.435{\,\rm GeV}/c^{2}<M(\Lambda_{c}^{+}\pi^{-})<2.442{\,\rm GeV}/c^{2}$ and $2.466{\,\rm GeV}/c^{2}<M(\Lambda_{c}^{+}\pi^{-})<2.473{\,\rm GeV}/c^{2}$. This can be compared with $(2.5\pm 0.5)$ events estimated from MC simulation of $\bar{B}^{0}\rightarrow\Lambda_{c}^{+}\bar{p}\pi^{+}\pi^{-}$ decay with four-body phase space normalized to the total of 1400 events [1]. Here the error is due to the statistics of the simulation. We do not take into account the non-$\Sigma_{c}^{0}$ background in the analysis that follows. Figure 2: (a) Dalitz plot and (b) $M^{2}(\bar{p}\pi^{-})$ distribution for $\bar{B}^{0}\to\Sigma_{c}^{++}\bar{p}\pi^{-}$. (c) Dalitz plot and (d) $M^{2}(\bar{p}\pi^{+})$ distribution for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$. Points with error bars indicate the data, and histograms are the decays simulated according to three-body phase space. Figure 3: $\Delta{E}$ distribution for the $\bar{B}^{0}\to\Lambda_{c}^{+}\bar{p}\pi^{+}\pi^{-}$ events in the $M_{\rm bc}$ signal region with $2.447\,\rm{GeV}/c^{2}<M(\Lambda_{c}^{+}\pi^{\pm})<2.461\,\rm{GeV}/c^{2}$. The curves indicate the fit with a double Gaussian for the signal and a linear background. Figure 4 shows (a) the $\bar{p}\pi^{+}$ mass, (b) $\cos\theta_{p}$ and (c) $\Sigma_{c}^{0}\pi^{+}$ mass distributions for the selected $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ events. Here, $\cos\theta_{p}$ is the cosine of the angle between the $\bar{p}$ momentum and the direction opposite to the $B$ momentum in the $\bar{p}\pi^{+}$ rest frame. The shaded histograms indicate the distributions for the background discussed above. The background shapes are obtained by fits to the data in the sideband region $\|\Delta{E}\|<0.1\,{\rm GeV}$ and $5.26\,{\rm GeV}/c^{2}<M_{\rm bc}<5.29\,{\rm GeV}/c^{2}$ outside the $B$ signal region, and the yield is fixed to 17 events. Here, the $M(\bar{p}\pi^{+})$ distribution is parameterized by the function $P_{\rm bkg}(M)=c\cdot\sqrt{t_{\rm min}\cdot t_{\rm max}}\cdot(1+c_{1}\cdot t_{\rm min})(1+c_{2}\cdot t_{\rm max})$ with $t_{\rm min}=(M^{2}-M^{2}_{\rm min})$ and $t_{\rm max}=(M^{2}_{\rm max}-M^{2})$. $M_{\rm min}$ and $M_{\rm max}$ are the minimum and maximum masses. The variable $c$ is a normalization constant, and $c_{1}$ and $c_{2}$ are shape parameters. The $\cos\theta_{p}$ distribution is modeled by a second-order Chebyshev polynomial. Figure 4: Data distributions for (a) $M(\bar{p}\pi^{+})$, (b) $\cos\theta_{p}$ and (c) $M(\Sigma_{c}^{0}\pi^{+})$. The shaded histograms indicate the normalized background. We find a significant structure in the $\bar{p}\pi^{+}$ mass distribution, and a forward peak in the $\cos\theta_{p}$ distribution, and a low mass $\Sigma_{c}^{0}\pi^{+}$ enhancement, denoted as $(\Sigma_{c}^{0}\pi^{+})_{\rm X}$. The $\bar{p}\pi^{+}$ mass structure has a mass near $1.5\,{\rm GeV}/c^{2}$ and a width of about $0.3\,{\rm GeV}$. We denote it as $\bar{N}^{0}$, and investigate its characteristics in detail. In order to describe the $\bar{p}\pi^{+}$ mass structure, which is not explained by a simple phase space non-resonant $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ decay, we consider an intermediate two-body decay $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ with a resonant state ${\bar{N}}^{0}\to{\bar{p}}\pi^{+}$. However, we still cannot reproduce the forward $\cos\theta_{p}$ peak and the $\Sigma_{c}^{0}\pi^{+}$ low mass structure with these two modes only. Therefore, we introduce one additional mode $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ to account for the observed features. As the low mass $\Sigma_{c}^{0}\pi^{+}$ structure is close to threshold, it produces a forward peak in the $\cos\theta_{p}$ distribution. In the low $\Sigma_{c}^{0}\pi^{+}$ mass region, we search for known $\Sigma_{c}^{0}\pi^{+}$ resonant states [18] in finer mass bins, but find no signals. So far, there is no good candidate to interpret this broad structure as a resonance. Therefore we assume that there is a threshold mass enhancement with a mass of $2800\,{\rm MeV/}c^{2}$ and a width of $350\,{\rm MeV}$ obtained from a fit to the $\Sigma_{c}^{0}\pi^{+}$ mass distribution using a relativistic Breit-Wigner (S-wave) function. Figure 5 compares binned Probability Density Functions (PDF) of the MC simulated events for the three assumed decay modes. The histograms show the PDFs for (a) the $M(\bar{p}\pi^{+})$, (b) $\cos\theta_{p}$ and (c) the $M(\Sigma_{c}^{0}\pi^{+})$ distributions. The solid histograms show the distributions of the mode $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$, assuming a P-wave relativistic Breit-Wigner amplitude with a mass of $1530\,{\rm MeV/}c^{2}$ and a width of $340\,{\rm MeV}$. Figure 5: Binned probability distributions of (a) $M(\bar{p}\pi^{+})$, (b) $\cos\theta_{p}$ and (c) $M(\Sigma_{c}^{0}\pi^{+})$, where we compare MC simulated distributions for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ (solid lines), $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ (dashed lines), and $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ (dotted lines). We make a simultaneous fit to the distributions in (a) and (b). To determine the $N^{0}$ mass, width and the yields of the three modes, we perform a maximum likelihood fit to the observed $M({\bar{p}}\pi^{+})$ and $\cos\theta_{p}$ distributions shown in Fig. 4. These two distributions are sufficient to fully describe the three-body decay $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$. To model the observed distribution, we construct a function $F(M(\bar{p}\pi^{+}),\cos\theta_{p})$ from the sum of PDFs of the three decay modes and the background. $\begin{split}F=\nu_{1}P_{\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}}+\nu_{2}P_{\bar{B}^{0}\to(\Sigma_{c}^{0}\pi^{+})_{\rm X}\,\bar{p}}\\\ +\nu_{3}P_{\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}}+\nu_{4}P_{\rm bkg},\end{split}$ where $P_{i}$ denotes a product of the normalized PDFs, $Q_{i}(M(\bar{p}\pi^{+}))$ (20 bins) and $R_{i}(\cos\theta_{p})$ (16 bins), and $\nu_{i}$ stands for the yield of the $i$-th mode. We plot $Q(M(\bar{p}\pi^{+}))$ and $R(\cos\theta_{p})$ distributions from the detector MC simulation for $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ and $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ modes, as shown in Figs 5(a) and (b), respectively. For the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ mode, we use the MC simulated $R(\cos\theta_{p})$ distribution and a Breit-Wigner for $Q(M({\bar{p}}\pi^{+}))$ with the ${\bar{N}}^{0}$ mass and width ($m_{R}$, $\Gamma$) as free parameters. A small systematic error due to the use of the $Q(M({\bar{p}}\pi^{+}))$ distribution without the MC detector simulation is discussed later. We use a P-wave (S-wave) relativistic Breit-Wigner shape. $BW_{P}(m^{2})=\frac{p^{2}}{(m^{2}-m_{R}^{2})^{2}+m_{R}^{2}{\Gamma^{2}(m)}}{\left[\frac{B(p)}{B(p_{0})}\right]}^{2},$ $BW_{S}(m^{2})=\frac{m_{R}\Gamma(m)}{[(m^{2}-{m_{R}}^{2})^{2}+{m_{R}}^{2}\Gamma^{2}(m)]},$ $\Gamma(m)=\left(\frac{p}{p_{0}}\right)^{2L+1}\left(\frac{m_{R}}{m}\right)\Gamma_{0}{\left[\frac{B(p)}{B(p_{0})}\right]^{2}}.$ Here $m$ is the mass of the $\bar{p}\pi^{+}$ system, and $m_{R}$ is the nominal $\bar{N}^{0}$ mass, and $\Gamma(m)$ is the width. The variable $p$ is the momentum of a daughter particle in the $\bar{N}^{0}$ rest frame, and $p_{0}$ is that for the nominal $\bar{N}^{0}$ mass. $B(p)=1/{\sqrt{1+(Rp)^{2}}}$ is the Blatt-Weisskopf form factor [19]. The value $R$, called the centrifugal barrier penetration factor, is set to $3\,{\rm(GeV}/c)^{-1}$ for a P-wave, and is zero for an S-wave. $L$ indicates the orbital angular momentum. For the S-wave Breit-Wigner amplitude [20] we use $\Gamma(m)$ with $m=m_{R}$ and $L=0$ to parameterize the smooth shape near the mass threshold. We define an extended unbinned likelihood with coarse bins, and carry out a maximum likelihood fit. $L=\frac{e^{-(\nu_{1}+\nu_{2}+\nu_{3}+\nu_{4})}}{N!}\prod F(m_{R},\Gamma,\nu_{1},\nu_{2},\nu_{3},\nu_{4})$ We fit the ${\bar{N}}^{0}$ mass and width $(m_{R},\Gamma)$ and the yields $\nu_{1},\nu_{2}$ and $\nu_{3}$ as free parameters, while the background yield $\nu_{4}$ is fixed to 17 events. Table 1 summarizes the fit results with various model assumptions. We calculate the statistical significance from the quality -2${\rm ln}({L}_{0}/{L}_{\rm max})$, where $L_{\max}$ is the maximum likelihood returned from the fit, and $L_{0}$ is the likelihood with the signal yield fixed to zero, and taking into account the reduction of the degrees of freedom. We obtain a significance of 7.0$\,\sigma$ for the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ contribution. The signal in the mode $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ has a statistical significance of 4.6$\,\sigma$, while that for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ is not significant (0.8 $\sigma$). Here, we calculate the goodness-of-fit from the likelihood ratio $\lambda$ [18], $\begin{split}\chi^{2}\approx-2{\rm ln}\lambda=2\sum_{j=1}^{36}[{\cal F}_{j}(m_{R},\Gamma,\nu_{1},\nu_{2},\nu_{3},\nu_{4})-F_{j}\\\ +F_{j}{\rm ln}(\frac{F_{j}}{{\cal F}_{j}(m_{R},\Gamma,\nu_{1},\nu_{2},\nu_{3},\nu_{4})})],\end{split}$ where $F_{j}$ and ${\cal F}_{j}$ are the observed and the fitted yields, respectively, in the $j$-th bin: $j=1,20$ for 20 bins in $Q(M(\bar{p}\pi^{+}))$ and $j=21,36$ for 16 bins in $R(cos\theta_{p})$. The small contribution ($\nu_{3}=-11\pm 10$) can be understood from Fig. 5. The mode $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ has a broad $\bar{p}\pi^{+}$ mass distribution similar to $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$, while it does not reproduce the forward $\cos\theta_{\bar{p}}$ peak. On the other hand, the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ mode can reproduce the ${\bar{p}}\pi^{+}$ mass bump structure and the uniform $\cos\theta_{\bar{p}}$ distribution. Hence, we fix $\nu_{3}=0$ in the subsequent fit and the uncertainty of this contribution is taken into account in the systematic error. Table 1: Summary of the simultaneous fits to the $M(\bar{p}\pi^{+})$ and $\cos\theta_{\bar{p}}$ distributions with the three decay modes $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$, $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ and $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$. (a) - (d) represent fits with various assumed contributions. Here, we show the fit results with the P-wave assumption, as we find no significant difference from the S-wave assumption. Decay mode \| (a) \| (b) \| (c) \| (d) \| Signif. \| ---\|---\|---\|---\|---\|---\|--- $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ \| free \| free \| free \| 0 \| 7.0 \| $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ \| free \| free \| 0 \| free \| 4.6 \| $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ \| free \| 0 \| free \| free \| 0.8 \| $\chi^{2}$/ndf \| 31.7/31 \| 32.4/32 \| 52.8/32 \| 88.4/34 \| \| Figure 6: Simultaneous fit to (a) $M({\bar{p}}\pi^{+})$ and (b) $\cos\theta_{p}$ distributions with a P-wave Breit-Wigner. The points with error bars are the data, and the curves are the contributions from $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ (dashed), $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ (dotted), the background (shaded) and their sum (solid). (c) $M(\Sigma_{c}^{0}\pi^{+})$ distribution, where the curves represent their contributions obtained by the fit to (a) and (b). Figure 6 shows the results of a fit to (a) the $M({\bar{p}}\pi^{+})$ and (b) $\cos\theta_{\bar{p}}$ distributions under the P-wave assumption. The data are the points with error bars. The curves are the contributions from $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ (dashed), $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ (dotted), the background (shaded) and their sum (solid). We obtain yields of ($70\pm 11$) and ($32\pm 9$) for the modes $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ and $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$, respectively. Figure 6(c) shows that the $M(\Sigma_{c}^{0}\pi^{+})$ distribution is consistently represented by the fitted parameters even though the $M(\Sigma_{c}^{0}\pi^{+})$ distribution is not included in the fit. Table 2: The fitted ${\bar{N}}^{0}$ mass and width with relativistic S-wave and P-wave Breit-Wigners. The first errors are statistical and the second are systematic including the uncertainties in the yields of $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ and the background, and the assumption of a low mass $(\Sigma_{c}^{0}\pi^{+})_{\rm X}$ structure. Item \| Yield \| Mass \| $\Gamma$ \| $\chi^{2}$/ndf ---\|---\|---\|---\|--- \| Events \| $\rm{MeV}/c^{2}$ \| $\rm{MeV}$ \| S-wave \| $71\pm 11\pm 10$ \| $1473\pm 31\pm 2$ \| $315\pm 72\pm 53$ \| 32.9/32 P-wave \| $70\pm 11\pm 10$ \| $1516\pm 29\pm 14$ \| $365\pm 97\pm 90$ \| 32.4/32 $\Sigma_{c}^{0}\pi^{+}$ sys. \| $\pm 5$ \| $\pm 8$ \| $\pm 50$ \| Table 2 compares the fit results for the ${\bar{N}}^{0}$ yield, mass and width with P-wave and S-wave assumptions. The fitted yields are found to be comparable with each other, while the mass and width show a systematic difference. We estimate systematic errors by varying the fitted yields by $\pm\sigma$ for the background ($\pm 3$) and $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ ($\pm 11$), and by taking into account the uncertainty in modeling the low mass $(\Sigma_{c}^{0}\pi^{+})_{\rm X}$ structure as discussed in the following. The simulated $R(\cos\theta_{p})$ distribution for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ is almost flat, as the generated $\cos\theta_{p}$ distribution is uniform for $P$\- and $S$-waves. However, the $R(cos\theta_{p})$ distribution is slightly affected by the assumed BW parameters due to efficiency changes in $\cos\theta_{p}$. We study the systematics of the fitted ${\bar{N}}^{0}$ mass and width due to the assumption on $R(\cos\theta_{p})$, by changing the mass and the width in EvtGen in ranges between $1400\,{\rm MeV}/c^{2}$ and $1620\,{\rm MeV}/c^{2}$, and between $200\,{\rm MeV}$ and $450\,{\rm MeV}$, respectively. We find variations of $\pm 1\,{\rm MeV}/c^{2}$ in the fitted mass and $\pm 5\,{\rm MeV}/c^{2}$ in the width. We also study the systematic errors due to the parameterization of the $(\Sigma_{c}^{0}\pi^{+})$ low mass structure. Instead of assuming a model with a single Breit-Wigner $(\Sigma_{c}^{0}\pi^{+})_{\rm X}$, we consider a combination of known states $\Lambda_{c}^{}\to\Sigma_{c}(2455)^{0}\pi^{+}$; $\Lambda_{c}^{+}(2625)$ ($\Gamma_{\rm total}<1.9\,{\rm MeV}$) [18], $\Lambda_{c}^{+}(2765)\to\Lambda_{c}^{+}\pi^{+}\pi^{-}$ ($\Gamma\sim 50\,{\rm MeV}/c^{2}$), $\Lambda_{c}^{+}(2880)$ ($\Gamma=5.8\pm 1.3\,{\rm MeV}$), and $\Lambda_{c}^{+}(2940)$ ($\Gamma=13^{+28}_{-9}\,{\rm MeV}$). Here we use the partial widths for $\Sigma_{c}^{0}\pi^{+}$ decay of the last three states given by Ref. [21]. We make a fit to the $\bar{N}^{0}$ mass, width and the yield of $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ with the individual $\Lambda_{c}^{+}$ yields floated and with the background fixed as mentioned previously. We obtain $N({\bar{p}}\pi^{+})$ mass and width values in good agreement with those obtained by the fit with the $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ model. The branching fraction product ${\cal B}$($\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$)$\times{\cal B}(\bar{N}^{0}\to\bar{p}\pi^{+})$ is calculated as $N_{s}/(N_{B\bar{B}}\cdot\epsilon\cdot CF\cdot{\cal B}_{\Lambda_{c}^{+}\to pK^{-}\pi^{+}})$ assuming $N_{B^{+}B^{-}}=N_{B^{0}\bar{B}^{0}}$. For $N_{s}$ we use the P-wave yield in Table 2 as it gives a better confidence level than an S-wave fit. We use $N_{B\bar{B}}=(387.7\pm 4.8)\times 10^{6}$ for the integrated luminosity of 357 fb-1, and the signal efficiency $\epsilon=(5.18\pm 0.13$)% from the MC simulation of $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$. We apply a correction factor $CF=(86.7\pm 7.9)$%, which takes into account the systematic difference in particle identification (PID) between data and MC simulation. Correction factors for proton, kaon and pion tracks are determined from a comparison of data and MC simulation for large samples of $D^{+}\to D^{0}(K\pi)\pi^{+}$ and $\Lambda\to p\pi^{-}$ decays. The overall PID correction factor is then calculated as a linear sum over the six tracks for the selected $B$ signal events. We assign an error of 7.2% due to track reconstruction efficiency for the six charged tracks in the final state. The systematic error on the branching fraction arising from a quadratic sum of the uncertainties on $N_{B\bar{B}}$, the signal efficiency $\epsilon$, and particle identification $CF$ and track reconstruction, is found to be 12%. Including the systematic error in the $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ yield, we arrive at the total systematic uncertainty in the branching fraction of 17.6 %. Thus, we obtain the branching fraction product of ${\cal B}(\bar{B}^{0}\to\Sigma_{c}(2455)^{0}\bar{N}^{0})\times{\cal B}(\bar{N}^{0}\to\bar{p}\pi^{+})$=($0.80\pm 0.15\rm{(stat.)}\pm 0.14\rm{(syst.)}\pm 0.21$)$\times 10^{-4}$, and a significance of 6.1 standard deviations including systematics. The last error is due to an uncertainty in ${\cal B}(\Lambda_{c}^{+}\to pK^{-}\pi^{+})=(5.0\pm 1.3)\%$ [18]. Next, we investigate goodness-of-fits with the masses and widths fixed to representative values for $N({\bar{p}}\pi^{+})$ states [18], and by floating the yields for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$ and $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$. The fit results are summarized in Table 3. Here $L_{2I,2S}$ stands for a resonance of isospin $I$ and spin $S$ with an orbital angular momentum of $L$=S, P and D for $L$=0, 1 and 2, respectively. We exclude $\Delta$ states such as $\Delta(1600)P_{33}$ and $\Delta(1620)S_{31}$, as we have no significant structure in the $\bar{p}\pi^{-}$ mass distribution in $\bar{B}^{0}\to\Sigma_{c}^{++}\bar{p}\pi^{-}$ decay. The fits favor $N(1440)P_{11}$ and $N(1535)S_{11}$, while they disfavor $N(1520)D_{13}$ and $N(1650)S_{11}$. In the decay $\bar{B}^{0}\to\Sigma^{0}_{c}\bar{N}^{0}$ (assuming $S(\Sigma^{0}_{c})=\frac{1}{2}$), one expects a uniform $\cos\theta_{p}$ distribution for the $N(1440)P_{11}$ state, $N(1535)S_{11}$ state and $N(1650)S_{11}$ state, and a $(1+3\cos^{2}\theta)$ distribution for the $N(1520)D_{13}$ state. As shown in Fig. 6(b), the distribution has a peak only in the forward direction, which is well reproduced by the mode $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$. The remaining uniform distribution is due to $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$. Thus, the observed $\cos\theta_{\bar{p}}$ distribution is consistent with both, $\bar{N}(1440)P_{11}$ and $\bar{N}(1535)S_{11}$ states, with a preference for the former due to the width of the state. Table 3: Results of the fits using the parameters of known $N({\bar{p}}\pi^{+})$ resonances [18]. States \| \| Mass \| $\Gamma$ \| $\Sigma_{c}^{0}\bar{N}^{0}$ \| $(\Sigma_{c}^{0}\pi^{+})_{X}\bar{p}$ \| $\chi^{2}$/ndf ---\|---\|---\|---\|---\|---\|--- \| \| ${\rm MeV}/c^{2}$ \| $~{}{\rm MeV}$ \| Events \| Events \| N(1440) \| $P_{11}$ \| 1440 \| 300 \| $65\pm 10$ \| $39\pm 9$ \| 37.6/34 N(1520) \| $D_{13}$ \| 1520 \| 115 \| $46\pm 9$ \| $53\pm 10$ \| 53.5/34 N(1535) \| $S_{11}$ \| 1535 \| 150 \| $58\pm 10$ \| $43\pm 10$ \| 40.1/34 N(1650) \| $S_{11}$ \| 1655 \| 165 \| $44\pm 10$ \| $55\pm 11$ \| 74.2/34 Finally, we try to perform a fit with an incoherent sum of the two Breit- Wigners, as we find that the fit results favor $N(1440)P_{11}$ and $N(1535)S_{11}$, and both give a distribution uniform in $\cos\theta_{p}$. Figure 7 shows the result of a fit to (a) the $M(\bar{p}\pi^{+})$ and (b) $\cos{\theta}_{\bar{p}}$ distributions, where the $N$ masses and widths are fixed to the values in Ref. [18], and the individual yields are floated. The histograms show the contributions from $N(1440)$ (solid), $N(1535)$ (dashed) states, $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ (dotted), and the background (shaded). The yields are $(37\pm 12)$ for the $N(1440)$ and $(30\pm 11)$ for the $N(1535)$, while the $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ yield is $(35\pm 9)$. We obtain the goodness of fit $\chi^{2}$/ndf=30.3/33, which indicates a slight preference (by 2.7 $\sigma$) for a mixed state of $N(1440)$ and $N(1535)$ [18]. Figure 7: Simultaneous fit to the $M(\bar{p}\pi^{+})$ and $\cos{\theta}_{\bar{p}}$ distributions with the $N(1440)P_{11}$ and $N(1535)S_{11}$ Breit-Wigners. The histograms indicate the contributions from the $P_{11}$ (solid) and $S_{11}$ (dashed) states, $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ (dotted), and the background (shaded). In summary, we study the three-body decay $\bar{B}^{0}\rightarrow\Sigma_{c}(2455)^{0}\bar{p}\pi^{+}$ with the same data set used for the analysis of the four-body decay $\bar{B}^{0}\rightarrow\Lambda_{c}^{+}\bar{p}\pi^{+}\pi^{-}$ [1]. We observe a broad $\bar{p}\pi^{+}$ mass structure near $1.5\,{\rm GeV}/c^{2}$, and a uniform $\cos{\theta}_{\bar{p}}$ distribution with a sharp forward peak. To explain these structures, we find that contributions from an intermediate two- body decay $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$, non-resonant three-body decay $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{p}\pi^{+}$ and a low mass structure near threshold $\bar{B}^{0}\to(\Sigma^{0}_{c}\pi^{+})_{\rm X}\,\bar{p}$ are needed. We perform a simultaneous fit to the $M(\bar{p}\pi^{+})$ and $\cos{\theta}_{\bar{p}}$ distributions with those three modes, and determine the yield and the relativistic Breit-Wigner parameters of the ${\bar{N}}^{0}$ state for $\bar{B}^{0}\to\Sigma_{c}^{0}\bar{N}^{0}$. We obtain the branching fraction product of ${\cal B}(\bar{B}^{0}\to\Sigma_{c}(2455)^{0}\bar{N}^{0})\times{\cal B}(\bar{N}^{0}\to\bar{p}\pi^{+})$=($0.80\pm 0.15\rm{(stat.)}\pm 0.14\rm{(syst.)}\pm 0.21$)$\times 10^{-4}$ with a signal significance of 6.1 standard deviations including systematics. The fitted mass and width are consistent with $\bar{N}(1440)$$P_{11}$ and $\bar{N}(1535)S_{11}$; both states also produce a uniform helicity distribution that is in good agreement with the data. The structure is also consistent with an interpretation in terms of an admixture of these two states. ACKNOWLEDGMENT We thank the KEKB group for excellent operation of the accelerator, the KEK cryogenics group for efficient solenoid operations, and the KEK computer group and the NII for valuable computing and Super-SINET network support. We acknowledge support from MEXT and JSPS (Japan); ARC and DEST (Australia); NSFC (China); DST (India); MOEHRD, KOSEF and KRF (Korea); KBN (Poland); MES and RFAAE (Russia); ARRS (Slovenia); SNSF (Switzerland); NSC and MOE (Taiwan); and DOE (USA). ## References [1] K.S. Park et al. (Belle Collaboration), Phys. Rev. D 75, 011101 (2007). * [2] N. Gabyshev et al. (Belle Collaboration), Phys. Rev. D 66, 091102(R) (2002). * [3] N. Gabyshev et al. (Belle Collaboration), Phys. Rev. Lett. 90, 121802 (2003). * [4] N. Gabyshev et al. (Belle Collaboration), Phys. Rev. Lett. 97, 202003 (2006). * [5] N. Gabyshev et al. (Belle Collaboration), Phys. Rev. Lett. 97, 242001 (2006). * [6] X. Fu et al. (CLEO Collaboration), Phys. Rev. Lett. 79, 3125 (1997). * [7] S.A. Dytman et al. (CLEO Collaboration), Phys. Rev. D 66, 091101(R) (2002). * [8] H. Kichimi, Nucl. Phys. B Proc. Suppl. 142, 197 (2005). * [9] M. Kobayashi and K. Maskawa, Prog. Theor. Phys. 49, 652 (1973). * [10] M. Jarfi et al., Phys. Lett. B 237, 513 (1990); M. Jarfi et al., Phys. Rev. D 43, 1599 (1991); N. Deshpande, J. Trampetic and A. Soni, Mod. Phys. Lett. 3A, 749 (1988). * [11] V. Chernyak and I. Zhitnisky, Nucl. Phys. B 345, 137 (1990). * [12] H.Y. Cheng and K.C. Yang, Phys. Rev. D 67, 034008 (2003). * [13] S. Kurokawa and E. Kikutani, Nucl. Instr. Meth. A 499, 1 (2003), and other papers included in this Volume. * [14] A. Abashian et al. (Belle Collaboration), Nucl. Instr. and Meth. A 479, 117 (2002). * [15] D.J. Lange, Nucl. Instr. and Meth. A 462, 152 (2001). * [16] The detector response is simulated using GEANT: R. Brun et al., GEANT 3.21, CERN Report DD/EE/84-1 (1984). * [17] E. Nakano, Nucl. Instr. and Meth. A 494, 402 (2002). * [18] W.-M. Yao et al. (Particle Data Group), J. Phys. G 33, 1 (2006) (URL:http://pdg.lbl.gov). * [19] J. Blatt and V. Weisskopf, Theoretical Nuclear Physics, New York: John Wiley & Sons (1952). * [20] H.M. Pilkuhn, The Interactions of Hadrons, Amsterdam: North-Holand Pub. Co. (1967). * [21] R. Mizuk et al. (Belle collaboration), Phys. Rev. Lett. 98, 262001 (2007).	arxiv-papers	2008-08-27T08:26:24	2024-09-04T02:48:57.493260	{ "license": "Public Domain", "authors": "Belle collaboration: H.O.Kim, et al.", "submitter": "Hiromichi Kichimi", "url": "https://arxiv.org/abs/0808.3650" }
0808.3741	# Geometric Approach to the Weil-Petersson Symplectic Form Reynir Axelsson Háskóli Íslands, HjarÐarhaga 6, IS-107 Reykjavík, Ísland reynir@raunvis.hi.is and Georg Schumacher Fachbereich Mathematik und Informatik, Philipps-Universität Marburg, Lahnberge, Hans-Meerwein-Strasse, D-35032 Marburg,Germany schumac@mathematik.uni-marburg.de ###### Abstract. In a family of compact, canonically polarized, complex manifolds the first variation of the lengths of closed geodesics is computed. As an application, we show the coincidence of the Fenchel-Nielsen and Weil-Petersson symplectic forms on the Teichmüller spaces of compact Riemann surfaces in a purely geometric way. The method can also be applied to situations like moduli spaces of weighted punctured Riemann surfaces, where the methods of Kleinian groups are not available. ## 1\. Introduction In 1958 André Weil introduced the methods of deformation theory of compact complex manifolds to the study of Teichmüller spaces [WE]. In particular, he used the $L^{2}$-inner product, induced by a metric of constant negative curvature, to the harmonic representatives of Kodaira-Spencer classes. He conjectured that the resulting Hermitian inner product on the Teichmüller space would be of Kähler type and of negative curvature. The resulting Kähler form is since known as the Weil-Petersson form. Beginning in 1982, Wolpert analyzed the Fenchel-Nielsen structure of the Teichmüller space and showed that the standard symplectic form $\omega^{FN}=-\sum d\tau^{i}\wedge d\ell^{i}$, defined in terms of Fenchel- Nielsen coordinates, actually coincides with the Weil-Petersson form $\omega^{WP}$. This was achieved in a sequence of papers [WO2, WO3, WO4], which also contain important arithmetic results. Most of the arguments use in an essential way the theory of Kleinian groups. For example, one key idea in Wolpert’s approach is to solve Beltrami equations on the upper half plane, to compute derivatives of cross-ratios of the solutions, and then use Poincaré series, in order to descend to the Riemann surfaces. Our aim is to extend these results to cases where the original methods do not apply, i.e. to cases where the hyperbolic structure is not necessarily induced by the universal covering, such as weighted punctured Riemann surfaces. Our methods also provide a purely differential geometric approach to a key result in Wolpert’s approach, namely the Weil-Petersson duality of the twist vector fields $\partial/\partial\tau^{j}$ and the differentials $d\ell^{j}$ of the corresponding length functions. We use path integrals along closed geodesics of harmonic Beltrami differentials and holomorphic quadratic differentials. More generally, the first variation of the length of a closed geodesic in a holomorphic family of Kähler-Einstein manifolds of negative curvature can be expressed as the integral of the harmonic Kodaira-Spencer tensor along a geodesic; this is our Theorem 3.2. We consider the Teichmüller space of (possibly weighted) punctured Riemann surfaces of genus $p$ with $n$ punctures. The existence of pants decompositions and Fenchel-Nielsen coordinates was shown by Zhang in [ZH] under the natural assumption that the cone angles do not exceed the value of $\pi$, i.e. the weights are in the interval $[1/2,1]$ (c.f. Section 8). Let $s^{\kappa};\kappa=1,\ldots,N$ be local holomorphic coordinates on the Teichmüller space near a fixed point $s_{0}$ corresponding to a Riemann surface $X$; then the tangent vectors $\partial/\partial s^{\kappa}\|_{s_{0}}$ correspond to harmonic Beltrami differentials $\mu_{\kappa}$ on $X$. An immediate consequence of Theorem 3.2 is the following result: ###### Theorem 1.1. The complex derivative of a length coordinate with respect to a holomorphic coordinate equals half the integral of the corresponding harmonic Beltrami differential along the closed geodesic: (1) $\left.\frac{\partial\ell^{i}(s)}{\partial s^{\kappa}}\right\|_{s_{0}}=\frac{1}{2}\int_{\gamma_{i}}\mu_{\kappa}.$ The above formula is independent of the choice of a set of Fenchel-Nielsen coordinates and holds for the length function $\ell^{\gamma}(s)$ of a differentiable family $\gamma(s)$ of closed geodesics. In the classical case, closely related results, where the first variation of the length of a closed geodesic is expressed in terms of integrals, can be found in [WO1, WO2]. The shearing technique for twists along closed geodesics plays an important role in Teichmüller theory. We handle the twist parameters $\tau^{i}$ in the context of deformation theory and cocycles of vertical automorphisms. This reflects only one aspect of the complex earthquake map associated to a given closed geodesic in the sense of McMullen [Mc] (cf. also [E-M]). The tangent vector $\partial/\partial\tau^{i}\|_{s_{0}}$ of the Teichmüller space is uniquely determined by its action on the cotangent space, i.e. the space of holomorphic quadratic differentials. (Here we use the type decomposition of the real tangent space of the Teichmüller space as embedded into its complexification.) ###### Theorem 1.2. The tangent vector $\partial/\partial\tau^{i}\|_{s_{0}}$ applied to a quadratic holomorphic differential $\varphi=\varphi(z)dz^{2}$ is given by the integration of the quadratic differential along the corresponding closed geodesic. (2) $\left(\left.\frac{\partial}{\partial\tau^{i}}\right\|_{s_{0}}\\!\\!,\varphi(z)dz^{2}\right)=\frac{\sqrt{-1}}{2}\int_{\gamma_{i}}\varphi.$ Note that also the above formula is independent of the choice of Fenchel- Nielsen coordinates. It holds for the twist vector $\partial/\partial\tau^{\gamma}$ for any given closed geodesic $\gamma$. Observe that the above inner product is actually a product of the $(1,0)$-part of $(\partial/\partial\tau^{i})$ and the quadratic holomorphic differential $\varphi$. As a corollary we obtain a key result proved by Wolpert for the classical case in [WO2]. ###### Theorem 1.3. The Fenchel-Nielsen symplectic form on the Teichmüller space is equal to the Weil-Petersson form: (3) $\omega^{FN}=\omega^{WP}\text{, i.e. }-\sum_{i=1}^{N}d\tau^{i}\wedge d\ell^{i}=\sqrt{-1}\;G_{\kappa{\overline{\lambda}}}^{WP}\;ds^{\kappa}\wedge ds^{\overline{\lambda}}.$ In fact the formulas (1), (2) and (3) are intimately related in the sense that any two of them imply the third one. In the situation of Fuchsian groups, the above geodesic integral (2) seems to be related to the monodromy integral of the quotient of the given quadratic differential by a certain Abelian differential in [WO2] (cf. Section 6). Our approach can also be used to compute second variations – our computations and applications will appear elsewhere. Acknowledgements. The second named author would like to thank Inkang Kim for important discussions. The authors would like to thank the referee for pointing out the reference [E-M]. This paper was written at the University of Iceland. The second named author would like to express his thanks for the kind hospitality and support. ## 2\. Families of Kähler-Einstein Manifolds Let $\\{{\mathcal{X}}_{s}\\}_{s\in S}$ be a holomorphic family of canonically polarized compact complex manifolds parameterized by a (connected) complex space $S$. It is given by a proper, smooth, holomorphic mapping $f:{\mathcal{X}}\to S$ such that ${\mathcal{X}}_{s}=f^{-1}(s)$ for all $s\in S$. For simplicity we will assume that the base $S$ is smooth, although our results can also be given a meaning for possibly non-reduced singular base spaces. We denote by $X={\mathcal{X}}_{s_{0}}$; $s_{0}\in S$ a distinguished fiber. Let $n=\dim_{\mathbb{C}}X$, and denote by $z^{\alpha}$; $\alpha=1,\ldots,n$ local holomorphic coordinates on $X$. A Kähler form on $X$ will be denoted by $\omega_{X}=\sqrt{-1}{g_{\alpha{\overline{\beta}}}}dz^{\alpha}\wedge dz^{\overline{\beta}}$ On the fiber $X$ we are using the summation convention together with the $\nabla$-notation for covariant derivatives. A $\|$–symbol will denote an ordinary derivative. Also, $\partial_{\alpha}$ and $\partial_{\overline{\beta}}$ will stand for $\partial/\partial z^{\alpha}$ and $\partial/\partial z^{\overline{\beta}}$ respectively. The raising and lowering of indices is defined as usual in terms of covariant derivatives. For the Ricci tensor $R_{\alpha{\overline{\beta}}}$ of $X$ we will use the sign convention $R_{\alpha{\overline{\beta}}}=-\log(g(z))_{\|{\alpha{\overline{\beta}}}},$ where $g(z)=\det({g_{\alpha{\overline{\beta}}}}(z))$. Accordingly we set $g(z,s)=\det(g_{\alpha{\overline{\beta}}}(z,s))$ for the family $f:{\mathcal{X}}\to S$, where we equip the fibers ${\mathcal{X}}_{s}$ with Kähler-Einstein forms $\omega_{{\mathcal{X}}_{s}}=\sqrt{-1}g_{\alpha{\overline{\beta}}}(z,s)dz^{\alpha}\wedge dz^{\overline{\beta}}$ of constant negative curvature $-1$, i.e. $R_{\alpha{\overline{\beta}}}(z,s)=-g_{\alpha{\overline{\beta}}}(z,s)$. Let $\rho:T_{s_{0}}S\to H^{1}(X,{\mathcal{T}}_{X})$ be the Kodaira-Spencer map for the corresponding deformation of $X$ over $S$ at the point $s_{0}\in S$. A natural inner product on the space $H^{1}(X,{\mathcal{T}}_{X})$ of infinitesimal deformations of $X$, the Weil-Petersson Hermitian inner product on $T_{s_{0}}S$, is induced by the Kähler-Einstein metric $\omega_{X}$ on $X$. Namely, given tangent vectors $u,v\in T_{s_{0}}S$, we denote by $A_{u}=A_{u{\overline{\beta}}}^{\alpha}\partial_{\alpha}dz^{\overline{\beta}}$ and $A_{v}$ the harmonic representatives of $\rho(u)$ and $\rho(v)$ respectively. Then the inner product of $u$ and $v$ equals $\langle u,v\rangle_{WP}=\int_{X}A_{u{\overline{\beta}}}^{\alpha}A_{{\overline{v}}\gamma}^{\overline{\delta}}g_{\alpha{\overline{\delta}}}g^{{\overline{\beta}}\gamma}g\/dV,$ where $A_{\overline{v}}$ denotes the adjoint (conjugate) tensor of $A_{v}$, and $g\/dV$ the volume element. We note that the Weil-Petersson inner product is positive definite at a given point of the base, if the induced deformation is effective. Let $s_{j}$; $j=1,\ldots,N$ be local coordinates on $S$, and set $A_{j}=A_{\partial/\partial s_{j}}$. Then the Weil-Petersson form on $S$ equals $\omega^{WP}=\sqrt{-1}G_{i{\overline{\jmath}}}^{WP}(s)ds^{i}\wedge ds^{\overline{\jmath}},$ where we use the notation $G_{i{\overline{\jmath}}}^{WP}(s)=\langle\partial/\partial s^{i},\partial/\partial s^{j}\rangle_{WP}=\int_{X}A_{i{\overline{\beta}}}^{\alpha}A_{{\overline{\jmath}}\gamma}^{\overline{\delta}}g_{\alpha{\overline{\delta}}}g^{{\overline{\beta}}\gamma}g\/dV.$ A closed real $(1,1)$-form on the total space ${\mathcal{X}}$ of the given family is defined by (4) $\omega_{\mathcal{X}}=-\sqrt{-1}\partial{\overline{\partial}}\log g(z,s).$ The Kähler-Einstein condition for the fibers implies that the restrictions of $\omega_{\mathcal{X}}$ to the fibers are the given Kähler-Einstein metrics: $\omega_{\mathcal{X}}\|_{{\mathcal{X}}_{s}}=\omega_{{\mathcal{X}}_{s}}$ for all $s\in S$. An important fact is that the family of Kähler-Einstein metric tensors contains the harmonic representatives of Kodaira-Spencer classes. The short exact sequence $0\to{\mathcal{T}}_{{\mathcal{X}}/S}\to{\mathcal{T}}_{{\mathcal{X}}}\to f^{}{\mathcal{T}}_{S}\to 0$ induces the Kodaira-Spencer map via the edge homomorphism for direct images. Again, we use $\nabla$-notation for covariant derivatives as well as raising and lowering of indices in fiber direction for families. A lift of a tangent vector $\partial/\partial s^{i}$ of $S$ at $s_{0}$ as a differentiable vector field of ${\mathcal{X}}$ on $X$ is of the form $\partial/\partial s^{i}+b_{i}^{\alpha}\partial_{\alpha}.$ Its exterior ${\overline{\partial}}$-derivative $B_{i{\overline{\beta}}}^{\alpha}\partial_{\alpha}dz^{\overline{\beta}}$, where $B_{i{\overline{\beta}}}^{\alpha}=\nabla_{\overline{\beta}}b_{i}^{\alpha}$, is interpreted as a ${\overline{\partial}}$-closed $(0,1)$-form on $X$ with values in the tangent bundle of $X$. It represents the obstruction against the existence of a holomorphic lift of the given tangent vector, i.e. the infinitesimal triviality of the deformation. Moreover, (5) $\rho(\partial/\partial s^{i})=[B_{i{\overline{\beta}}}^{\alpha}\partial_{\alpha}dz^{\overline{\beta}}]\in H^{1}(X,{\mathcal{T}}_{X}).$ Our argument is based upon the notion of canonical lifts in the sense of Siu (cf. [SIU]), which turned out to be $\omega_{\mathcal{X}}$-horizontal: The form $\omega_{\mathcal{X}}$ is positive definite in fiber directions but need not be positive on ${\mathcal{X}}$. Its components in the directions of $z^{\alpha}$ and $s^{j}$ are denoted by $g_{\alpha{\overline{\jmath}}}$ etc. However, horizontal lifts of tangent vectors are well-defined and of the form (6) $\partial/\partial s^{i}+a_{i}^{\alpha}\partial_{\alpha},$ where (7) $a_{i}^{\alpha}=-{g^{{\overline{\beta}}\alpha}}g_{i{\overline{\beta}}}.$ We set $A_{i{\overline{\beta}}}^{\alpha}=\nabla_{\overline{\beta}}a_{i}^{\alpha}$ and obtain (cf. [SCH2, SCH3]) the following fact: ###### Proposition 2.1. The horizontal lifts with respect to $\omega_{\mathcal{X}}$ of tangent vectors induce harmonic representatives of Kodaira-Spencer classes: The harmonic representative of $\rho(\partial/\partial s^{i})$ equals $A_{i{\overline{\beta}}}^{\alpha}\partial_{\alpha}dz^{\overline{\beta}}$. These satisfy the following properties (i) $\nabla_{\overline{\delta}}A^{\alpha}_{i{\overline{\beta}}}=\nabla_{\overline{\beta}}A^{\alpha}_{i{\overline{\delta}}},$ * (ii) $\nabla_{\gamma}A_{i{\overline{\beta}}}^{\alpha}g^{{\overline{\beta}}\gamma}=0,$ * (iii) $A_{i{\overline{\beta}}{\overline{\delta}}}=A_{i{\overline{\delta}}{\overline{\beta}}}.$ The conditions (i) and (ii) above correspond to harmonicity, whereas condition (iii) reflects the relationship with the metric tensor. ## 3\. Families of closed geodesics Let $(f:{\mathcal{X}}\to S,\omega_{\mathcal{X}})$ be a family of Kähler- Einstein manifolds with constant negative Ricci curvature equal to $-1$, where $\omega_{\mathcal{X}}$ is given by (4). We denote by $\gamma_{s}$ a differentiable family of closed geodesics contained in the fibers ${\mathcal{X}}_{s}$. In order to describe the variation of the length of closed geodesics in a family, we use the notion of integrating a tensor along a geodesic. Exemplarily we define: ###### Definition 3.1. Let $C=C_{{\overline{\beta}}{\overline{\delta}}}$ be a tensor on the Kähler manifold $X$, and $\gamma$ be a geodesic of length $\ell$, parameterized by $u(t)=(u^{1}(t),\ldots,u^{n}(t))$, such that $\\|\dot{u}(t)\\|=1$. Then $\int_{\gamma}C=\int_{\gamma}C_{{\overline{\beta}}{\overline{\delta}}}dz^{\overline{\beta}}dz^{\overline{\delta}}:=\int_{0}^{\ell}C_{{\overline{\beta}}{\overline{\delta}}}(u(t))\dot{u}^{\overline{\beta}}\dot{u}^{\overline{\delta}}dt.$ For contravariant tensors of order one this notation coincides with the integration of a differential form along the curve $\gamma$. The main result of this section is the following theorem: ###### Theorem 3.2. Let ${\mathcal{X}}\to S$ be a family of canonically polarized, compact, complex manifolds equipped with Kähler-Einstein metrics. Let $\gamma_{s}$ be a differentiable family of closed geodesics on the fibers ${\mathcal{X}}_{s}$, having the lengths $\ell(s)$. Let $\partial/\partial s^{i}\|_{s=s_{0}}\in T_{s_{0}}S$ be a tangent vector. Then the first variation of the length of the closed geodesic with respect to the holomorphic parameter of the family is half the integral of the symmetric, harmonic Kodaira-Spencer tensor along the geodesic: $\left.\frac{\partial}{\partial s^{i}}\right\|_{s=s_{0}}\\!\\!\\!\ell(s)=\frac{1}{2}\int_{\gamma(s_{0})}A_{i{\overline{\beta}}{\overline{\delta}}}dz^{\overline{\beta}}dz^{\overline{\delta}}.$ In order to prove the theorem, we may assume that $S=\\{s\\}$ is a disk in $\mathbb{C}$, and $s_{0}=0$. The closed geodesic curves $\gamma_{s}$ are solutions of the equation for geodesics (8) $\ddot{u}^{\alpha}(t,s)+\Gamma^{\alpha}_{\gamma\sigma}(u(t,s))\dot{u}^{\gamma}(t,s)\dot{u}^{\sigma}(t,s)=0.$ The solution is unique up to the constant value of the speed $\\|\dot{u}(t,s)\\|^{2}={{g_{\alpha{\overline{\beta}}}}(u(t,s),s)\dot{u}^{\alpha}(t,s)\dot{u}^{\overline{\beta}}(t,s)}.$ For $s=0$ we chose $\\|\dot{u}\\|=1$, for the remaining values of $s$ the value of $\\|\dot{u}\\|$ will be determined by the fact that the parameter $t$ assumes values in the interval $[0,\ell_{0}]$, where $\ell_{0}$ is the length of $\gamma_{0}$. Now $\ell(s)=\int_{0}^{\ell_{0}}\\|\dot{u}(t,s)\\|dt$ so that (9) $\left.\frac{d\ell(s)}{ds}\right\|_{s=0}=\frac{1}{2}\int_{0}^{\ell_{0}}\frac{d}{ds}\\|\dot{u}(t,s)\\|^{2}dt.$ We need to study various tensors on ${\mathcal{X}}$ along the geodesics. In particular, we consider the vector field $U_{s}(t,s)=\left.\left(\frac{\partial}{\partial s}+u_{s}^{\alpha}(t,s)\frac{\partial}{\partial z^{\alpha}}+u_{s}^{\overline{\beta}}(t,s)\frac{\partial}{\partial z^{\overline{\beta}}}\right)\right\|_{u(t,s)},$ where $u_{s}^{\alpha}(t,s)=\frac{\partial u^{\alpha}(t,s)}{\partial s}\text{ and }u_{s}^{\overline{\beta}}(t,s)=\frac{\partial u^{\overline{\beta}}(t,s)}{\partial s}\,.$ The function $\chi(t,s)=\langle U_{s},\dot{u}\rangle$ is defined as the inner product of vector fields along the geodesics with values on the total space ${\mathcal{X}}$. Concerning $\omega_{\mathcal{X}}$ we indicate the direction of the one-dimensional space $S$ by $s$. We have $\chi(t,s)=g_{s{\overline{\beta}}}\dot{u}^{\overline{\beta}}+{g_{\alpha{\overline{\beta}}}}u_{s}^{\alpha}\dot{u}^{\overline{\beta}}+{g_{\alpha{\overline{\beta}}}}\dot{u}^{\alpha}u_{s}^{\overline{\beta}}.$ Using (8) we simplify the expression for the derivative $\dot{\chi}$ and get: (10) $\begin{split}\dot{\chi}(t,s)=g_{s{\overline{\beta}}\|\alpha}\dot{u}^{\alpha}\dot{u}^{\overline{\beta}}+g_{s{\overline{\beta}}\|{\overline{\delta}}}\dot{u}^{\overline{\beta}}\dot{u}^{\overline{\delta}}+g_{s{\overline{\beta}}}\ddot{u}^{\overline{\beta}}+g_{\alpha{\overline{\beta}}\|\gamma}\dot{u}^{\gamma}u_{s}^{\alpha}\dot{u}^{\overline{\beta}}+\\\ {g_{\alpha{\overline{\beta}}}}\dot{u}_{s}^{\alpha}\dot{u}^{\overline{\beta}}+g_{\alpha{\overline{\beta}}\|{\overline{\delta}}}\dot{u}^{\alpha}u_{s}^{\overline{\beta}}\dot{u}^{\overline{\delta}}+{g_{\alpha{\overline{\beta}}}}\dot{u}^{\alpha}\dot{u}_{s}^{\overline{\beta}}.\end{split}$ ###### Lemma 3.3. $\frac{\partial}{\partial s}\\|\dot{u}\\|^{2}-\dot{\chi}=A_{s{\overline{\beta}}{\overline{\delta}}}\dot{u}^{\overline{\beta}}\dot{u}^{\overline{\delta}}.$ ###### Proof. We use (8) and (10) together with (6) and (7) and apply Proposition 2.1. The statement now follows from a calculation. ∎ The proof of Theorem 3.2 concludes with (9) and Lemma 3.3. ## 4\. Cocycles of vertical automorphisms Let $X$ be a compact complex manifold, and $(S,s_{0})$ a germ of a complex space. We follow the approach of Forster and Knorr [F-K] and assign to any deformation of $X$ over a $(S,s_{0})$ an element of the first cohomology $H^{1}(X,\mathfrak{G}_{S})$ of $X$ with values in the sheaf $\mathfrak{G}_{S}$ of vertical automorphisms. Its derivative with respect to the base space yields the Kodaira-Spencer map. For any deformation of $X$ given by a cartesian diagram $\textstyle{X\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{{\mathcal{X}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{f}$$\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{(S,s_{0})}$ there exist open coverings $\\{U_{i}\\}$ of $X$ and $\\{Z_{i}\\}$ of ${\mathcal{X}}$ together with isomorphisms $\psi_{i}:U_{i}\times S\to Z_{i}$, compatible with the cartesian diagram, and equal to the identity for $s=0$. The actual cocycles $\gamma_{ij}$ are defined on $(U_{i}\cap U_{j})\times S$ by $\psi_{j}^{-1}\circ\psi_{i}$. We look at the derivatives of the $\gamma_{ij}$ with respect to tangent vectors of the base and values in the sheaf of holomorphic vector fields. These define the Kodaira-Spencer classes and will be compared with suitable $(0,1)$-forms representing the same classes. We take a differentiable trivialization $\zeta:{\mathcal{X}}\to X\times S$ of the family, which equals the identity for $s=s_{0}$. Then, according to (5), the class $\rho(\partial/\partial s^{\kappa})\|_{s=s_{0}}$ is represented by (11) $\left.\frac{\partial^{2}\zeta^{\alpha}(z,s)}{\partial s^{\kappa}\partial z^{\overline{\beta}}}\right\|_{s=s_{0}}\\!\frac{\partial}{\partial z^{\alpha}}dz^{\overline{\beta}}.$ We now consider the situation where the covering that is involved in the deformation only consists of only two elements, $U_{1}$ and $U_{2}$, say. A differentiable trivialization of the holomorphic family can be chosen to be equal to the identity on the complement of any given compact neighborhood of $Z_{1}\cap Z_{2}\subset{\mathcal{X}}$. ## 5\. Application to Riemann surfaces For families of Riemann surfaces (with notation from Section 1) the above Proposition 2.1 states that harmonic Beltrami differentials are of the form (12) $\mu_{\kappa}=\mu_{\kappa}(z)\frac{\partial}{\partial z}{\overline{dz}}=-\frac{\partial}{{\overline{\partial z}}}\left(\frac{1}{g}\frac{\partial^{2}\log g}{\partial s^{\kappa}{\overline{\partial z}}}\right)\frac{\partial}{\partial z}{\overline{dz}}.$ We write these also in the form (13) $\mu_{\kappa}(z)\frac{\partial}{\partial z}{\overline{dz}}=\frac{\varphi_{\overline{\kappa}}(z)}{g(z)}\frac{\partial}{\partial z}{\overline{dz}}$ for certain quadratic holomorphic differentials $\varphi_{\kappa}=\varphi_{\kappa}(z)dz^{2}$, with $\varphi_{\overline{\kappa}}:={\overline{\varphi_{\kappa}}}$. Now Theorem 1.1 is an immediate consequence of Theorem 3.2. ## 6\. Twist coordinates vs. holomorphic coordinates In this section we will prove Theorem 1.2. We will first treat the case of Riemann surfaces $X$ equipped with the hyperbolic metrics $g=g(z)\|dz\|^{2}$. Given the context, we normalize a metric of constant curvature $-1$ so that the Ricci curvature equals $-1$, i.e. $\partial^{2}\log g(z)/\partial z{\overline{\partial z}}=g(z)$. Our point is to give a direct proof, which only uses the hyperbolic geometry of Riemann surfaces. We consider a closed geodesic $\gamma$ on $X$ parameterized by $u(t)$ with $\\|\dot{u}(t)\\|=1$. Let the length of $\gamma$ be $\ell_{0}=2\pi r_{0}$. The twist along $\gamma$ gives rise to a real, one parameter family of Riemann surfaces, in particular it gives rise to a real (analytic) curve in the appropriate Teichmüller space. As local coordinate set we use an annulus $W=\\{z\in\mathbb{C};r_{1}<\|z\|<r_{2}\\}$ embedded into $X$, such that $\gamma$ corresponds to a circle $\\{\|z\|=r_{0}\\}$ for $r_{1}<r_{0}<r_{2}$, with parameterization $u(t)=r_{0}e^{\sqrt{-1}t/r_{0}}$ for $0\leq t\leq\ell_{0}$. For $\tau\in{\mathbb{R}}$ a family of holomorphic automorphisms of $W$ is given by the twist $z\mapsto e^{\sqrt{-1}\tau/r_{0}}z$. In the sense of Section 4 we interpret these as a family of vertical automorphisms with respect to an open covering of $X$ consisting of two elements. Now a differentiable trivialization (which is equal to the identity on the complement of $W$) is given by (14) $\zeta(z,\tau)=e^{\sqrt{-1}\frac{\tau}{r_{0}}h(r)}z,$ where $r=\|z\|$ and $h\in C^{\infty}((r_{1},r_{2}),{\mathbb{R}})$ denotes a differentiable function which is identically zero near $r_{1}$, and identically equal to one near $r_{2}$. The value $\tau=\ell_{0}$ represents a Dehn twist. We take a deformation theoretic standpoint, and we use in this context the shearing technique [WO2, Section 1]: A Beltrami differential will again stand for an infinitesimal deformation. If we replace in (11) a tangent vector $\partial/\partial s^{\kappa}$ by a real tangent vector $\partial/\partial\tau$, we obtain the image of its $(1,0)$-component under the Kodaira-Spencer map. The resulting Beltrami differential $\mu=\mu(z)(\partial/\partial z){\overline{dz}}$ is identically zero on the complement of $W$. For $z\in W$ it follows immediately that (15) $\mu(z)=\frac{\sqrt{-1}}{2}\frac{z^{2}}{rr_{0}}h^{\prime}(r).$ Now we can prove Theorem 1.2. The idea is to determine a Beltrami differential as a linear form on the cotangent space of the Teichmüller space. The duality $H^{1}(X,{\mathcal{T}}_{X})\times H^{0}(X,{\mathcal{O}}_{X}(2K_{X}))\to\mathbb{C}$ of tangent and cotangent spaces of ${\mathcal{T}}$ is given by $\left(\mu(z)\frac{\partial}{\partial z}{\overline{dz}},\varphi(z)dz^{2}\right)\mapsto\int_{X}\mu(z)\varphi(z)dA,$ where $dA=\sqrt{-1}dz\wedge{\overline{dz}}$ is the Euclidean area element. Observe that $\mu$ need not be harmonic in this expression, in fact we get the same value for any function $h(r)$ with the above properties. The claim of the theorem follows from (15) in principle by Stokes’ theorem – we can also let $h$ tend to the function $h_{0}$, which is defined to be equal to zero for $r<r_{0}$ and equal to one for $r\geq r_{0}$. Then $h^{\prime}(r)$ converges to a delta function, and integration over $r$ implies that that for any quadratic holomorphic differential $\varphi$ $(\mu,\varphi)=\frac{\sqrt{-1}}{2}\;r_{0}\int_{0}^{2\pi}e^{2\sqrt{-1}\vartheta}\varphi(r_{0}e^{\sqrt{-1}\vartheta})\/\/d\vartheta.$ Now the necessary reparameterization $\vartheta=t/r_{0}$ yields the claim of the theorem. ∎ We mention the monodromy integrals from [WO2]. If $A$ is a hyperbolic element in $PSL(2,{\mathbb{R}})$, lifted to $\widehat{A}\in SL(2,{\mathbb{R}})$, the monodromy integral is ${\mathcal{P}}(\psi,\widehat{A})=\frac{1}{2}\int_{s}^{As}\psi\Omega_{\widehat{A}}$ where $\psi$ is a quadratic holomorphic differential, and $\Omega_{\widehat{A}}^{-1}$ a suitable Abelian differential. Then by [WO2, Lemma 4.2] the following formula holds for the twist tangent vector $\mu_{0}$ applied to a holomorphic quadratic differential $\int_{\Delta}\mu_{0}\psi=-\sqrt{-1}\sigma_{\widehat{A}}{\eta(\widehat{A})}^{-1}{\mathcal{P}}(\psi,\widehat{A}),$ where $\Delta$ is a fundamental domain, $\sigma_{\widehat{A}}=\pm 1$, and $\eta(\widehat{A})=(({\mathrm{\,tr\,}}\widehat{A})^{2}-4)^{1/2}$. ## 7\. Computation of the Fenchel-Nielsen symplectic form The results of this section are valid for the Teichmüller space of (possibly weighted) punctured Riemann surfaces. Combining our geodesic integral formulas (1) and (2) we immediately obtain a result, which was proved in the classical case by Wolpert ([WO2, Theorem 2.10]). ###### Corollary 7.1. The vector fields $\partial/\partial\tau^{j}$ are Hamiltonian with respect to the Weil-Petersson form: (16) $\displaystyle\frac{\partial\ell^{j}}{\partial s^{\overline{\kappa}}}$ $\displaystyle=$ $\displaystyle-\sqrt{-1}\left\langle\frac{\partial}{\partial\tau^{j}},\mu_{\kappa}\right\rangle,$ (17) $\displaystyle\frac{\partial\ell^{j}}{\partial s^{\kappa}}$ $\displaystyle=$ $\displaystyle\sqrt{-1}\left\langle\mu_{\kappa},\frac{\partial}{\partial\tau^{j}}\right\rangle.$ Here, the Weil-Petersson inner product, which is defined initially on the complex tangent space, is extended in a natural way to the complexified real tangent space (which also contains the complex tangent vectors of type $(0,1)$). Next, we prove the coincidence (3) of the Fenchel-Nielsen and Weil-Petersson symplectic forms using (1) and (2). We will give a short computational argument: Proof of Theorem 1.3. We write $\omega^{WP}$ in terms of Fenchel-Nielsen coordinates and begin with the contribution of the twist parameter involving $d\tau^{j}\wedge d\tau^{k}$. Since we are dealing with real forms, we need to show: Claim. The real part $\text{\rm Re}\left(\sqrt{-1}\;\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial s^{\beta}}\right\rangle\frac{\partial s^{\alpha}}{\partial\tau^{j}}\frac{\partial s^{\overline{\beta}}}{\partial\tau^{k}}\right)$ vanishes, in particular, it is symmetric in $j$ and $k$. Proof of the claim: $\displaystyle\sqrt{-1}\;\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial s^{\beta}}\right\rangle\frac{\partial s^{\alpha}}{\partial\tau^{j}}\frac{\partial s^{\overline{\beta}}}{\partial\tau^{k}}$ $\displaystyle=$ $\displaystyle\sqrt{-1}\;\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial s^{\beta}}{\partial\tau^{k}}\frac{\partial}{\partial s^{\beta}}\right\rangle\frac{\partial s^{\alpha}}{\partial\tau^{j}}$ $\displaystyle=$ $\displaystyle\sqrt{-1}\;\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial\tau^{k}}\right\rangle\frac{\partial s^{\alpha}}{\partial\tau^{j}}\text{\footnotesize\quad(type consideration) }$ $\displaystyle=\atop\eqref{eq:ltaudual2}$ $\displaystyle\frac{\partial\ell^{k}}{\partial s^{\alpha}}\cdot\frac{\partial s^{\alpha}}{\partial\tau^{j}}.$ The real part of this quantity equals $None$ $\frac{1}{2}\frac{\partial\ell^{k}}{\partial\tau^{j}}=0.$ We conclude the proof of Theorem 1.3. The vanishing of the coefficients of $d\ell^{k}\wedge d\ell^{m}$ is an immediate consequence of the “dual” equation (18), which may not be available at this point. However, we can revert here to Wolpert’s purely (differential) geometric argument [WO4, Corollary 1.2], which involves orientation reversing (anti-holomorphic) symmetries. Finally, the coefficient of $d\tau^{j}\wedge d\ell^{m}$ in $\omega^{WP}$ equals $\displaystyle\sqrt{-1}\;\left(\left\langle\frac{\partial s^{\alpha}}{\partial\tau^{j}}\frac{\partial}{\partial s^{\alpha}},\frac{\partial s^{\beta}}{\partial\ell^{m}}\frac{\partial}{\partial s^{\beta}}\right\rangle-\left\langle\frac{\partial s^{\alpha}}{\partial\ell^{m}}\frac{\partial}{\partial s^{\alpha}},\frac{\partial s^{\beta}}{\partial\tau^{j}}\frac{\partial}{\partial s^{\beta}}\right\rangle\right)$ $\displaystyle=$ $\displaystyle\sqrt{-1}\;\left(\left\langle\frac{\partial}{\partial\tau^{j}},\frac{\partial s^{\beta}}{\partial\ell^{m}}\frac{\partial}{\partial s^{\beta}}\right\rangle-\left\langle\frac{\partial s^{\alpha}}{\partial\ell^{m}}\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial\tau^{j}}\right\rangle\right)$ $\displaystyle=\atop{\eqref{eq:ltaudual1}\eqref{eq:ltaudual2}}$ $\displaystyle-\left(\frac{\partial\ell^{j}}{\partial s^{\overline{\beta}}}\frac{\partial s^{\overline{\beta}}}{\partial\ell^{m}}+\frac{\partial\ell^{j}}{\partial s^{\alpha}}\frac{\partial s^{\alpha}}{\partial\ell^{m}}\right)$ $\displaystyle=$ $\displaystyle-\delta_{\ell m}.\hskip 170.71652pt\qed$ Now we obtain the following formula of Wolpert from the equality $\omega^{FN}=\omega^{WP}$: ###### Corollary 7.2. The length coordinates $\ell^{j}$ give rise to Hamiltonian vector fields $\partial/\partial\ell^{j}$: (18) $\frac{\partial\tau^{j}}{\partial s^{\alpha}}=-\sqrt{-1}\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial\ell^{j}}\right\rangle.$ The above equation (17) also implies the following corollary, whose classical version is Theorem 2.11 from [WO2]: We first observe that (17) holds in a slightly more general situation. Namely, as usual, we pick local holomorphic coordinates $(s_{1},\ldots,s_{N})$ on the Teichmüller space near a point $s_{0}$ and denote by $\mu_{\alpha}$ the harmonic Beltrami differentials corresponding to the $\partial/\partial s^{\alpha}\|_{s_{0}}$. Then for any closed geodesic $\gamma$ on $X$ we have the length function $\ell(\gamma)$, and also $\partial/\partial\tau(\gamma)$ is defined at the given point $s_{0}$ without a choice of a system of Fenchel-Nielsen coordinates. Now (17) reads (19) $\frac{\partial\ell(\gamma)}{\partial s^{\alpha}}=\sqrt{-1}\;\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial\tau(\gamma)}\right\rangle.$ for all $\alpha=1,\ldots,N$. ###### Corollary 7.3. Let $\gamma$ and $\delta$ be closed geodesics on $X$. Then (20) $\frac{\partial\ell(\gamma)}{\partial\tau(\delta)}=-\frac{\partial\ell(\delta)}{\partial\tau(\gamma)}.$ ###### Proof. We show that the left-hand side is skew symmetric in $\gamma$ and $\delta$. $\displaystyle\frac{\partial\ell(\gamma)}{\partial\tau(\delta)}$ $\displaystyle=$ $\displaystyle 2\text{\rm Re}\left(\frac{\partial\ell(\gamma)}{\partial s^{\alpha}}\cdot\frac{\partial s^{\alpha}}{\partial\tau(\delta)}\right)$ $\displaystyle=\atop\eqref{eq:ltaudualvar}$ $\displaystyle 2\text{\rm Re}\left(\sqrt{-1}\;\frac{\partial s^{\alpha}}{\partial\tau(\delta)}\cdot\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial}{\partial\tau(\gamma)}\right\rangle\right)$ $\displaystyle=$ $\displaystyle 2\text{\rm Re}\left(\sqrt{-1}\;\frac{\partial s^{\alpha}}{\partial\tau(\delta)}\cdot\left\langle\frac{\partial}{\partial s^{\alpha}},\frac{\partial s^{\beta}}{\partial\tau(\gamma)}\frac{\partial}{\partial s^{\beta}}\right\rangle\right)\text{\footnotesize\quad(type consideration) }$ $\displaystyle=$ $\displaystyle 2\text{\rm Re}\left(\sqrt{-1}\;G^{WP}_{\alpha{\overline{\beta}}}\frac{\partial s^{\alpha}}{\partial\tau(\delta)}\frac{\partial s^{\overline{\beta}}}{\partial\tau(\gamma)}\right)$ $\displaystyle=$ $\displaystyle\sqrt{-1}\;G^{WP}_{\alpha{\overline{\beta}}}\left(\frac{\partial s^{\alpha}}{\partial\tau(\delta)}\frac{\partial s^{\overline{\beta}}}{\partial\tau(\gamma)}-\frac{\partial s^{\alpha}}{\partial\tau(\gamma)}\frac{\partial s^{\overline{\beta}}}{\partial\tau(\delta)}\right).$ ∎ ## 8\. Weighted punctured Riemann surfaces and conical metrics Let $X$ be a compact Riemann surface of genus $p$ and ${\bf a}=\sum_{i=1}^{n}a_{i}p_{i}$ be an ${\mathbb{R}}$-divisor, with $0<a_{i}\leq 1$ for all $i$. Then $(X,\bf a)$ is called a weighted punctured Riemann surface. We always assume that the degree of the ${\mathbb{R}}$-divisor $K_{(X,{\bf a})}:=K_{X}+{\bf a}$, is positive. It was shown by Hulin, McOwen and Troyanov [H-T, McO, T] that such surfaces possess unique hyperbolic conical metrics. These are characterized as follows: * (i) For $a_{j}=1$ the metric $g$ satisfies a Poincaré growth condition at $p_{j}$, i.e. $g=\frac{\rho(z)}{\|z\|^{2}\log^{2}(\|z\|^{2})}\|dz\|^{2}$ in some local holomorphic coordinate system, where $p_{j}$ corresponds to $z=0$, and $\rho$ is a continuous positive function. * (ii) For $0<a_{j}<1$ the metric is of the form $g=\frac{\rho(z)}{\|z\|^{2a_{j}}}\|dz\|^{2}$, where $\rho$ is continuous. In a holomorphic family, in particular regarding Teichmüller and moduli spaces, punctures are labeled. The complex structure of the resulting Teichmüller space ${\mathcal{T}}_{\bf a}$ of weighted punctured surfaces (with weights being fixed) can be identified with ${\mathcal{T}}_{p,n}$. However, weights give rise to a hierarchy of compactifications of the corresponding moduli spaces, which were extensively investigated by Hassett [HA]. On the other hand, hyperbolic conical metrics induce different Kähler structures on Teichmüller and moduli spaces. Generalized Weil-Petersson metrics were introduced and studied in [S-T1, S-T2]. From the viewpoint of hyperbolic geometry, the Teichmüller space of weighted punctured Riemann surfaces were studied by Zhang in [ZH]. Under the seemingly necessary assumption that the weights are between $1/2$ and $1$, Zhang showed the existence of Fenchel-Nielsen coordinates so that the Fenchel-Nielsen symplectic form $\omega^{FN}$ becomes meaningful. The above range of weights contains the interesting weights of the form $1-1/k$ that arise from finite quotients. Any pants decomposition of a Riemann surface with conical singularities with the above restriction avoids the punctures. So Theorems 1.1 and 1.2 are still valid. Concerning the notion of harmonic Beltrami differentials, Hilbert space theory is not available in general. Instead we use equation (13) as a definition for the space $H^{1}(X,\bf a)$ of such differentials on a weighted punctured Riemann surface $(X,\bf a)$. In [S-T1] the existence of a duality (21) $H^{1}(X,{\bf a})\times H^{0}(X,{\mathcal{O}}_{X}(2K_{X}+D))\to\mathbb{C},$ was shown, where $D=\sum_{i}p_{i}$, is the underlying divisor to $\bf a$ with weights one. Also (12) holds. The $L^{2}$ inner product on $H^{1}(X,{\bf a})$ induced by the conical hyperbolic metric gives rise to a generalized Weil- Petersson metric, which is of Kähler type [S-T1, S-T2]. The statement and proof of the Theorem 1.3 can be carried over litterally to this case. So all necessary conditions for the coincidence of $\omega^{FN}$ and $\omega^{WP}$ are satisfied for weighted punctured Riemann surfaces. ## References * [E-M] Epstein, D.B.A., Marden, A.: Convex hulls in hyperbolic space, a theorem of Sullivan, and measured pleated surfaces. Analytical and geometric aspects of hyperbolic space, Symp. Warwick and Durham/Engl. 1984, Lond. Math. Soc. Lect. Note Ser. 111 113–253 (1987). * [F-K] Forster, O.; Knorr, K.: Ein neuer Beweis des Satzes von Kodaira-Nirenberg-Spencer. Math. Z. 139 257–291 (1974). * [HA] Hassett, B.: Moduli space of weighted pointed stable curves. Adv. math. 173, 316–352 (2003). * [H-T] Hulin, D., Troyanov, M.: Prescribing curvature on open surfaces. Math. Ann. 293, 277–315 (1992). * [Mc] McMullen, C.: Complex earthquakes and Teichmüller theory. J. Am. Math. Soc. 11, 283–320 (1998). * [McO] McOwen, R.: Prescribed curvature and singularities of conformal metrics on Riemann surfaces. J. Math. Analays. Appl. 177, 287–298 (1993). * [SCH1] Schumacher, G.: Harmonic maps of the moduli space of compact Riemann surfaces. Math. Ann. 275, 455–466 (1986). * [SCH2] Schumacher, G.: The curvature of the Petersson-Weil metric on the moduli space of Kähler-Einstein manifolds. Ancona, V. (ed.) et al., Complex analysis and geometry. New York: Plenum Press. The University Series in Mathematics. 339–354 (1993). * [SCH3] Schumacher, G.: The theory of Teichmüller spaces. A view towards moduli spaces of Kähler manifolds. Several complex variables VI. Complex manifolds, Encycl. Math. Sci. 69, 251–310 (1990). * [S-T1] Schumacher, G.; Trapani, S.: Variation of cone metrics on Riemann surfaces. J. Math. Anal. Appl. 311, 218–230 (2005). * [S-T2] Schumacher, G., Trapani, St.: The Weil-Petersson structure for moduli of weighted punctured Riemann surfaces, in preparation. * [SIU] Siu, Y.-T.: Curvature of the Weil-Petersson metric in the moduli space of compact Kähler-Einstein manifolds of negative first Chern class. Contributions to several complex variables, Hon. W. Stoll, Proc. Conf. Complex Analysis, Notre Dame/Indiana 1984, Aspects Math. E9, 261–298 (1986). * [T] Troyanov, M.: Prescibing curvature on compact surfaces with conical singularities. Trans. Am. Math. Soc. 324, 793–821 (1991). * [WE] Weil. A.: Final report on contract AF 18(603)-57, Coll. Works (1958). * [WO1] Wolpert, S.: An elementary formula for the Fenchel-Nielsen twist. Comment. Math. Helvetici 56, 132–135 (1981). * [WO2] Wolpert, S.: The Fenchel-Nielsen deformation. Ann. Math. 115, 501–528 (1982). * [WO3] Wolpert, S.: On the symplectic geometry of deformations of a hyperbolic surface. Ann. Math. 117, 207–234 (1983). * [WO4] Wolpert, S.: On the Weil-Petersson Geometry of the Moduli Space of Curves, Am. J. Math. 107 969–997 (1985). * [ZH] Zhang, Y.: Hyperbolic cone-surfaces, generalized Markoff maps, Schottky groups and McShane’s identity, Doctoral Thesis, Department of Mathematics, National University of Singapore, 2004.	arxiv-papers	2008-08-27T16:32:33	2024-09-04T02:48:57.499578	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Reynir Axelsson, Georg Schumacher", "submitter": "Georg Schumacher", "url": "https://arxiv.org/abs/0808.3741" }
0808.3781	# Radiation from relativistic jets ,a Y. Mizunob, P. Hardeec, H. Sold, M. Medvedeve, B. Zhangf, Å. Nordlundg, J. T. Frederikseng, G. J. Fishmanh, R. Preecea, a National Space Science and Technology Center, 320 Sparkman Drive, VP 62, Huntsville, AL 35805, USA b National Space Science and Technology Center, 320 Sparkman Drive, VP 62, Huntsville, AL 35805, USA, (visiting Univ. of Nevada, Las Vegas) c Department of Physics and Astronomy, The University of Alabama Tuscaloosa, AL 35487, USA d LUTH, Observatore de Paris-Meudon, 5 place Jules Jansen, 92195 Meudon Cedex, France e Department of Physics and Astronomy, University of Kansas, KS 66045, USA f Department of Physics, University of Nevada, Las Vegas, NV 89154, USA g Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, 2100 Copenhagen Ø, Denmark h NASA-Marshall Space Flight Center, National Space Science and Technology Center 320 Sparkman Drive, VP 62, Huntsville, AL 35805, USA E-mail ###### Abstract: Nonthermal radiation observed from astrophysical systems containing relativistic jets and shocks, e.g., gamma-ray bursts (GRBs), active galactic nuclei (AGNs), and Galactic microquasar systems usually have power-law emission spectra. Recent PIC simulations of relativistic electron-ion (electron-positron) jets injected into a stationary medium show that particle acceleration occurs within the downstream jet. In the presence of relativistic jets, instabilities such as the Buneman instability, other two-streaming instability, and the Weibel (filamentation) instability create collisionless shocks, which are responsible for particle (electron, positron, and ion) acceleration. The simulation results show that the Weibel instability is responsible for generating and amplifying highly nonuniform, small-scale magnetic fields. These magnetic fields contribute to the electron’s transverse deflection behind the jet head. The “jitter” radiation from deflected electrons in small-scale magnetic fields has different properties than synchrotron radiation which is calculated in a uniform magnetic field. This jitter radiation, a case of diffusive synchrotron radiation, may be important to understand the complex time evolution and/or spectral structure in gamma- ray bursts, relativistic jets, and supernova remnants. ## 1 Introduction Shocks are believed to be responsible for prompt emission from gamma-ray bursts (GRBs) and their afterglows, for variable emission from blazars, and for particle acceleration processes in jets from active galactic nuclei (AGN) and supernova remnants (SNRs). The predominant contribution to the observed emission spectra is often assumed to be synchrotron- and inverse Compton radiation from these accelerated particles for gamma-ray bursts [1-7] and for AGN jets [8-13]. It is assumed that turbulent magnetic fields in the shock region lead to Fermi acceleration, producing higher energy particles [14, 15]. To make progress in understanding emission from these object classes, it is essential to place modeling efforts on a firm physical basis. This requires studies of the microphysics of the shock process in a self-consistent manner [16, 17]. ## 2 Method of calculation Three-dimensional relativistic particle-in-cell (RPIC) simulations have been used to study the microphysical processes in relativistic shocks. Such PIC simulations show that rapid acceleration takes place in situ in the downstream jet [18-33]. Three independent simulation studies confirm that relativistic counter-streaming jets do excite the Weibel instability [34], which generates current filaments and associated magnetic fields [35], and accelerates electrons [18-22]. In order to determine the luminosity and spectral energy density (SED) of synchrotron radiation, it is general practise to simply assume that a certain fraction $\epsilon_{\rm B}$ of the post-shock thermal energy density is carried by the magnetic field, that a fraction $\epsilon_{\rm e}$ is carried by electrons, and that the energy distribution of the electrons is a power- law, $d\log n_{\rm e}/d\log\varepsilon=p$ (above some minimum energy $E_{\rm m}$ which is determined by $\epsilon_{\rm e},\epsilon_{\rm B}$ and $p$). In this approach, $\epsilon_{\rm B}$, $\epsilon_{\rm e}$, and $p$ are treated as free parameters, to be determined by observations. However, more constraining data now require additional free parameters such as the introduction of broken power-law to reproduce the spectral energy distributions of TeV blazars for instance [11]. Due to the lack of a first principle theory of collisionless shocks, a purely phenomenological approach to modeling radiation is applied, but one must recognize that emission is then calculated without a full understanding of the processes responsible for particle acceleration and magnetic field generation [17]. It is important to clarify that the constraints imposed on these parameters by the observations are independent of any assumptions regarding the nature of the shocks and the processes responsible for particle acceleration or magnetic field generation. Any model proposed for the actual shock micro-physics must be consistent with these phenomenological constraints. Since magnetic fields are generated by the current structures produced in the filamentation (Weibel) instability, it is possible that “jitter” radiation [36-44] is an important emission process in GRB and AGN jets. It should be noted that synchrotron- and ‘jitter’-radiation are fundamentally the same physical processes (emission of accelerated charges in a magnetic field), but the relative importance of the two regimes depends on the comparison of the deflection angle and the emission angle of the charges [36]. Emission via synchrotron- or “jitter”-radiation from relativistic shocks is determined by the magnetic field strength and structure and the electron energy distribution behind the shock, which can be computed self-consistently with RPIC simulations. The full RPIC simulations may actually help to determine whether the emission is more synchrotron-like or jitter-like. The characteristic differences between Synchrotron- and jitter radiation are relevant for a more fundamental understanding of the complex time evolution and/or spectral propertis of GRBs (prompt and afterglows) [45]. For example, jitter radiation has been proposed as a solution of the puzzle that below their peak frequency GRB spectra are sometimes steeper than the “line of death” spectral index associated with synchrotron emission [35-38], i.e., the observed SED scales as $F_{\nu}\propto\nu^{2/3}$, whereas synchrotron SEDs should follow $F_{\nu}\propto\nu^{1/3}$, or even more shallow (i.e. $F_{\nu}\propto\nu^{\alpha}$ where $\alpha\leq 1/3$, e.g., [37]). Thus, it is crucial to calculate the emerging radiation by tracing electrons (positrons) in self-consistently evolved electromagnetic fields. This highly complex analytical and computational task requires sophisticated tools, such as multi- dimensional, relativistic, PIC methods. ### 2.1 New Computing Method of Calculating Synchrotron and Jitter Emission from Electron Trajectories in Self-consistently Generated Magnetic Fields Consider a particle at position ${\bf{r}_{0}}(t)$ at time $t$ (Fig. 1). At the same time, we observe the associated electric field from position $\bf{r}$. Because of the finite propagation velocity of light, we actually observe the particle at an earlier position $\bf{r}_{0}(\rm{t}^{{}^{\prime}})$ along its trajectory, labeled with the retarded time $t^{{}^{\prime}}=t-\delta t^{{}^{\prime}}=t-\bf{R}(\rm{t}^{{}^{\prime}})/c$. Here $\bf{R}(\rm{t}^{{}^{\prime}})=\|\bf{r}-\bf{r}_{0}(\rm{t}^{{}^{\prime}})\|$ is the distance from the charge (at the retarded time $t^{{}^{\prime}}$) to the observer’s position. Figure 1: Definition of the retardation effect. From an observers point, r, one sees the particle at position $\bf{r}_{0}(\rm{t}^{{}^{\prime}})$ where it was at retarded time t’ (from Figure 2.2 in [46]). The retarded electric field from a charged particle moving with instantaneous velocity $\boldsymbol{\beta}$ under acceleration $\boldsymbol{\dot{\beta}}$ is expressed as [48], $\displaystyle\bf{E}$ $\displaystyle=$ $\displaystyle\frac{q}{4\pi\epsilon_{0}}\left[\frac{\bf{n}-\boldsymbol{\beta}}{\gamma^{2}(1-\bf{n}\cdot\boldsymbol{\beta})^{\rm{3}}\rm{R}^{2}}\right]_{\rm ret}+\frac{q}{4\pi\epsilon_{0}c}\left[\frac{\bf{n}\times\\{(\bf{n}-\boldsymbol{\beta})\times\boldsymbol{\dot{\beta}}\\}}{(1-\bf{n}\cdot\boldsymbol{\beta})^{\rm{3}}\rm{R}}\right]_{\rm ret}$ (1) Here, $\bf{n}\equiv\bf{R}(\rm{t}^{{}^{\prime}})/\|\bf{R}(\rm{t}^{{}^{\prime}})\|$ is a unit vector that points from the particle’s retarded position towards the observer. The first term on the right hand side, containing the velocity field, is the Coulomb field from a charge moving without influence from external forces. The second term is a correction term that arises when the charge is subject to acceleration. Since the velocity-dependent field falls off in distance as $R^{-2}$, while the acceleration-dependent field scales as $R^{-1}$, the latter becomes dominant when observing the charge at large distances ($R\gg 1$). The choice of unit vector $\bf{n}$ along the direction of propagation of the jet (hereafter taken to be the $Z$-axis) corresponds to head-on emission. For any other choice of $\bf{n}$ (e.g., $\theta\lesssim 1/\gamma$), off-axis emission is seen by the observer. The observer’s viewing angle is set by the choice of $\bf{n}$ ($n_{\rm x}^{2}+n_{\rm y}^{2}+n_{\rm z}^{2}=1$). After some calculation and simplifying assumptions (for detailed derivation see [46]) the total energy $W$ radiated per unit solid angle per unit frequency can be expressed as $\displaystyle\frac{d^{2}W}{d\Omega d\omega}$ $\displaystyle=$ $\displaystyle\frac{\mu_{0}cq^{2}}{16\pi^{3}}\left\|\int^{\infty}_{-\infty}\frac{\bf{n}\times[(\bf{n}-\boldsymbol{\beta})\times\boldsymbol{\dot{\beta}}]}{(1-\boldsymbol{\beta}\cdot\bf{n})^{2}}e^{i\omega(t^{{}^{\prime}}-\bf{n}\cdot\bf{r}_{0}({\rm t}^{{}^{\prime}})/{\rm c})}dt^{{}^{\prime}}\right\|^{2}$ (2) This equation contains the retarded electric field from a charged particle moving with instantaneous velocity $\boldsymbol{\beta}$ under acceleration $\boldsymbol{\dot{\beta}}$, and only the acceleration field is kept since the velocity field decreases rapidly as $1/R^{2}$. The distribution over frequencies of the emitted radiation depends on the particle energy, radius of curvature, and acceleration. These quantities are readily obtained from the trajectory of each charged particle. Since the jet plasma has a large velocity $Z$-component in the simulation frame, the radiation from the particles (electrons and positrons) is heavily beamed along the $Z$-axis as jitter radiation [36, 37, 38, 44]. ## 3 Radiation from relativistic electrons: a simple case to test computing method Here we have calculated the radiation from two electrons with Lorentz factor ($\gamma=15.8,40.8$) [29, 30]. The electrons gyrate in the $x-z$ plane with the uniform magnetic field ($B_{\rm y}$) and the results are shown in Figures 2 & 3. Figure 2: The paths of two charged particles moving in a fixed homogenous magnetic field (left panel) ($\gamma=15.8,40.8$). The particles produce a time dependent electric field. An observer situated at great distance along the n-vector sees the retarded electric field from the gyrating particles (right panel). As a result of relativistic beaming, the field is seen as pulses peaking when the particles move directly towards the observer. Figure 3: The observed power spectrum from two charged particles, gyrating in a magnetic field at different viewing angles. The viewing angles are 0∘, 1∘, 2∘, 3∘, 4∘, 5∘, and 6∘ ($n_{\rm y}\neq 0$). With larger angles the frequencies above the Nyquist frequency should be strongly damped, however they increase due to aliasing [46]. The units on both axes are arbitrary. The theoretical synchrotron spectrum for a viewing angle equal to 0∘ is plotted for comparison as a red curve for the electron with $\gamma=40.8$ (multiplied by 2 for clarity). The spectra observed far from the electron at angles with respect to the $z$ direction are shown in Fig. 3. The higher frequencies ($>f_{\rm c}$) are strongly damped with increasing angles as $e^{(-f/f_{\rm c})}$, see [48]. Since the critical frequency $f_{\rm c}=\frac{3}{2}\gamma^{3}\left(\frac{c}{\rho}\right)=2309$, where $\rho=11.03$ for the electron with $\gamma=40.8$ is larger than that for $\gamma=15.8$, the radiation from the electron $\gamma=40.8$ is dominant [30]. The electron with $\gamma=15.8$ gyrates about three times in this period, the ripples in the spectrum shows the electron cyclotron frequency. However, in order to resolve it much longer time is required [46]. We have very good agreement between the spectrum obtained from the simulation and the theoretical synchrotron spectrum expectation (red curve) from eq. 3 (eq. 7.10 [46]). Synchrotron radiation with the full angular dependency for the parallel polarization component is given by [48], $\displaystyle\frac{d^{2}W_{\|\|}}{d\omega d\Omega}$ $\displaystyle=$ $\displaystyle\frac{\mu_{0}cq^{2}\omega^{2}}{12\pi}\left(\frac{r_{\rm L}\theta^{2}_{\beta}\beta^{2}}{c}\right)^{2}\frac{\|K_{\frac{2}{3}}(\chi/\sqrt{\cos\theta\beta^{3}})\|^{2}}{(\cos\theta\beta^{3})^{2}},$ (3) where $\theta$ is the angle between n and the orbital plane $\theta^{2}_{\beta}\equiv 2(1-\beta\cos\theta)$, $\chi=\omega r_{\rm L}\theta^{3}_{\beta}/3c$ and $r_{\rm L}$ the gyro-radius $\gamma mv/(qB)$. For $\beta\rightarrow 1$ and $\theta\rightarrow 0$ this expression converges toward the solution one normally finds in text books [48, 47]. It should be noted that the method based on the integration of the retarded electric fields calculated by tracing many electrons described in this section can provide a proper spectrum in turbulent electromagnetic fields. On the other hand, if the formula for the frequency spectrum of radiation emitted by a relativistic charged particle in instantaneous circular motion is used [47, 48], the complex particle accelerations and trajectories are not properly accounted for and the jitter radiation spectrum is not properly obtained (for details see [46, 49]). The results described above validate the technique used in our code as described previous section [29, 30, 46, 49]. ## 4 Discussion We have started to calculate emission directly from our simulations using the same method described in the previous section. In order to calculate the (jitter-like) synchrotron radiation from the particles in the electromagnetic fields generated by the filamentation instability, the retarded electric field from a single particle is Fourier-transformed to give the individual particle spectrum as described in the previous section. The individual particle spectra are added together to produce a total spectrum over a particular simulation time span [46, 49]. It should be noted that for this calculation very large simulations over a long time ($t_{\rm s}$) are required using a small time step ($\Delta t$) in order to increase the upper frequency limit to the spectrum (Nyquist frequency $\omega_{\rm N}=1/2\Delta t$). Frequency resolution is limited by the time span ($\Delta\omega=1/t_{\rm s}$) [46, 49]. For a case with the time step $\Delta t=0.01/\omega_{\rm pe}$ and the time span $t_{\rm s}=50/\omega_{\rm pe}$, a calculated spectrum will have the highest frequency, $50\omega_{\rm pe}$ and the frequency resolution (the lowest frequency), $0.02\omega_{\rm pe}$. $\omega_{\rm pe}$ is calculated with an appropriate plasma density. Simulations over a long time allow us to obtain multiple spectra at sequential time spans so the spectral evolution can be calculated. Synthetic spectra obtained in the way we have described above should be compared with GRB prompt and afterglow observations. In the case of AGN jets, diffusive synchrotron radiation has already been invoked by several works [39, 41, 43] to reproduce spectra of 3C273, M87 and Cen A knots from radio to X-rays. For TeV blazars, taking into account the relative importance of the energy densities contained in the small-scale and large-scale magnetic fields may be an elegant alternative to the choice of a broken power-law for the energy distribution of radiating electrons. ###### Acknowledgments. K.I.N., Y.M., and P.H. are supported by NSF-AST-0506719, AST-0506666, NASA- NNG05GK73G, NNX07AJ88G, and NNX08AG83G. M.M has been supported by NSF- AST-0708213, NASA-NX07AJ50G, and NNX08AL39G, and DoE-DE-FG02-07ER54940. Simulations were performed at the Columbia facility at the NASA Advanced Supercomputing (NAS) and IBM p690 (Copper) at the National Center for Supercomputing Applications (NCSA) which is supported by the NSF. Part of this work was done while K.-I. N. was visiting the Niels Bohr Institute. He thanks the director of the institution for generous hospitality. K.-I. N. prepared this proceeding during his visit at Meudon Observatory, and thanks the director of the institution for generous hospitality. ## References * [1] T. Piran, _Gamma-ray bursts and the fireball model_ , _Phys. Rep._ 316 (575) 1999. * [2] T. Piran, _Gamma-ray bursts - a puzzle being resolved_ , _Phys. Rep._ 333 (529) 2000. * [3] T. Piran, _The physics of gamma-ray bursts_ , _Rev. Mod. Phys._ 76 (1143) 2005 * [4] B. Zhang, and P. Meszaros, _Gamma-Ray Bursts: progress, problems & prospects_, _Int. J. Mod. Phys._ A19 (2385) 2004. * [5] P. Meszaros, _Theories of Gamma-Ray Bursts_ , _ARAA_ 40 (137) 2002. * [6] P. Meszaros, _Gamma-Ray Bursts_ , _Rept. Prog. Phys._ 69 (2259) 2006. * [7] B. Zhang, _Gamma-Ray Bursts in the Swift Era_ , _Chin. J. Astron. Astrophys._ 7 (1) 2007. * [8] M. Chiaberge, and G. Ghisellini, _Rapid variability in the synchrotron self-Compton model for blazars_ , _MNRAS_ 306 (551) 1999. * [9] S. Inoue, and F. Takahara, _Electron Acceleration and Gamma-Ray Emission from Blazars_ , _ApJ_ 463 (555) 1996. * [10] S.D. Bloom, and A.P. Marscher, _An Analysis of the Synchrotron Self-Compton Model for the Multi–Wave Band Spectra of Blazars_ , _ApJ_ 461 (657) 1996. * [11] K. Katarzynski, H. Sol, and A. Kus, _The multifrequency emission of Mrk 501. From radio to TeV gamma-rays_ , _A &A_ 367 (809) 2001. * [12] F. Aharonian, et al (HESS collaboration), _Discovery of VHE $\gamma$-rays from the high-frequency-peaked BL Lacertae object RGB J0152$+$017_, _A &A_ 481 (L103) 2008. * [13] F. Aharonian, et al (HESS collaboration), _Discovery of very high energy $\gamma$-ray emission from the BL Lacertae object H 2356-309 with the HESS Cherenkov telescopes_, _A &A_ 455 (461) 2006. * [14] E. Fermi, _On the Origin of the Cosmic Radiation_ , _Phys. Rev._ 75 (1169) 1949. * [15] R. Blandford, and D. Eichler, _Particle Acceleration at Astrophysical Shocks - a Theory of Cosmic-Ray Origin_ , _Phys. Rep._ 154 (1) 1987. * [16] T. Piran, _Magnetic Fields in Gamma-Ray Bursts: A Short Overview_ , in the proceedings of _Magnetic Fields in the Universe_ , _AIPC_ 784 (164) 2005. * [17] E. Waxman, _Gamma-ray bursts and collisionless shocks_ , _Plasma Phys. Control. Fusion_ 48 (B137) 2006. * [18] L.O. Silva, R.A. Fonseca, J.W. Tonge, J.M. Dawson, W.B. Mori, and M.V. Medvedev, _Interpenetrating Plasma Shells: Near-equipartition Magnetic Field Generation and Nonthermal Particle Acceleration_ , _ApJ_ 596 (L121) 2003. * [19] J.T. Frederiksen, C.B. Hededal, T. Haugbølle, and Å. Nordlund, _Collisionless Shocks - Magnetic Field Generation and Particle Acceleration_ , in proceedings of _1st NBSI on Beams and Jets in Gamma Ray Bursts_ , held at NBIfAFG/NORDITA, Copenhagen, Denmark, August, 2002, [arXiv:astro-ph/0303360] * [20] J.T. Frederiksen, C.B. Hededal, T. Haugbølle, and Å. Nordlund, _Magnetic Field Generation in Collisionless Shocks: Pattern Growth and Transport_ , _ApJ_ 608 (L13) 2004. * [21] C.B. Hededal, T. Haugbølle, J.T. Frederiksen, and Å. Nordlund, _Non-Fermi Power-Law Acceleration in Astrophysical Plasma Shocks_ , _ApJ_ 617 (L107) 2004. * [22] C.B. Hededal, and K.-I. Nishikawa, _The Influence of an Ambient Magnetic Field on Relativistic collisionless Plasma Shocks_ , _ApJ_ 623 (L89) 2005. * [23] K.-I. Nishikawa, P. Hardee, G. Richardson, R. Preece, H. Sol, and G.J Fishman, _Particle Acceleration in Relativistic Jets due to Weibel Instability_ , _ApJ_ 595 (555) 2003. * [24] K.-I. Nishikawa, P. Hardee, G. Richardson, R. Preece, H. Sol, and G.J Fishman, _Particle Acceleration and Magnetic Field Generation in Electron-Positron Relativistic Shocks_ , _ApJ_ 622 (927) 2005. * [25] Nishikawa, K.-I. Nishikawa, P. Hardee, G. Richardson, R. Preece, H. Sol, and G.J Fishman, __,_High Energy Gamma-Ray Astronomy_ , _AIPC_ 745 (534) 2005. * [26] K.-I. Nishikawa, P. Hardee, C.B. Hededal, and G.J. Fishman, _Acceleration Mechanics in Relativistic Shocks by the Weibel Instability_ , _ApJ_ 642 (1267) 2006 * [27] K.-I. Nishikawa, C.B. Hededal, P. Hardee, C. Kouveliotou, G.J. Fishman, and Y. Mizuno, _3-D Rpic Simulations of Relativistic Jets: Particle Acceleration, Magnetic Field Generation, and Emission_ , _Ap &SS_ 307 (319) (DOI 10.1007/s10509-006-9234-5) 2007\. * [28] E. Ramirez-Ruiz, K.-I. Nishikawa, and C.B. Hededal, _e ± Pair Loading and the Origin of the Upstream Magnetic Field in GRB Shocks_, _ApJ_ 671 (1877) 2007. * [29] K.-I. Nishikawa, Y. Mizuno, G.J Fishman, and P. Hardee, _Particle Acceleration, Magnetic Field Generation, and Associated Emission in Collisionless Relativistic Jets_ , in proceedings of _The workshop: High Energy Phenomena in Relativistic Outflows (HEPRO)_ , Dublin, 24-28 September 2007 [arXiv:0801.4390]. * [30] K.-I. Nishikawa, P. Hardee, Y. Mizuno, M. Medvedev, B. Zhang, D. Hartmann, and G.J Fishman, _Relativistic Particle-In-Cell Simulation Studies of Prompt and Early Afterglows from GRBs_ , in proceedings of _Seventh European Workshop on Collisionless Shocks_ , Paris, 7- 9 November 2007 [arXiv:0802.2558]. * [31] P. Chang, A. Spitkovsky, and J. Arons, _Long Term Evolution of Magnetic Turbulence in Relativistic Collisionless Shocks: Electron-Positron Plasmas_ , _ApJ_ 674 (378) 2008. * [32] A. Spitkovsky, _On the Structure of Relativistic Collisionless Shocks in Electron-Ion Plasmas_ , _ApJ_ 673 (L39) 2008. * [33] A. Spitkovsky, _Particle acceleration in relativistic collisionless shocks: Fermi process at last?_ , _ApJ_ submitted to ApJL, 2008. * [34] E.S. Weibel,_Spontaneously Growing Transverse Waves in a Plasma Due to an Anisotropic Velocity Distribution_ , _Phys. Rev. Lett._ 2 (83) 1959. * [35] M.V. Medvedev, and A. Loeb, _Generation of Magnetic Fields in the Relativistic Shock of Gamma-Ray Burst Sources_ , _ApJ_ 526 (697) 1999. * [36] M.V. Medvedev, _Theory of “jitter” radiation from small-scale random magnetic fields and prompt emission from gamma-ray burst shocks_ , _ApJ_ 540 (704) 2000. * [37] M.V. Medvedev, _The theory of spectral evolution of the GRB prompt emission_ , _ApJ_ 637 (869) 2006. * [38] G.D. Fleishman, _Diffusive synchrotron radiation from relativistic shocks in gamma-ray burst sources_ , _ApJ_ 638 (348) 2006 * [39] G.D. Fleishman, _Diffusive synchrotron radiation from extragalactic jets_ _MNRAS_ 365 (L11) 2006. * [40] G.D. Fleishman, and I.N. Toptygin, _Diffusive Radiation in One-dimensional Langmuir Turbulence_ , _PRE_ 76 (017401) 2007. * [41] G.D. Fleishman, and I.N. Toptygin, _Diffusive radiation in Langmuir turbulence produced by jet shocks_ , _MNRAS_ 381 (1473) 2007. * [42] G.D. Fleishman, and M.F. Bietenholz, _Diffusive synchrotron radiation from pulsar wind nebulae_ _MNRAS_ 376 (625) 2007. * [43] J. Mao, , and J. Wang, _Knot in Centaurus A: A Stochastic Magnetic Field for Diffusive Synchrotron Radiation?_ _ApJ_ 669 (13) 2007. * [44] J.C. Workman, B.J. Morsony, D. Lazzati, and M.V. Medvedev, _Jitter radiation in gamma-ray bursts and their afterglows: emission and self-absorption_ , _MNRAS_ 386 (199) 2008. * [45] R.D. Preece,M.S. Briggs, R.S. Mallozzi, G.N. Pendleton, W.S. Paciesas, and D.L. Band, _The Synchrotron Shock Model Confronts a “Line of Death” in the BATSE Gamma-Ray Burst Data_ , _ApJ_ 506 (L23) 1998. * [46] C.B. Hededal, _Gamma-Ray Bursts, Collisionless Shocks and Synthetic Spectra_ , PhD thesis, 2005 [arXiv:astro-ph/0506559]. * [47] G.B. Rybicki, and A.P. Lightman, _Radiative Processes in Astrophysics_ , John Wiley & Sons, New York, 1979. * [48] J.D. Jackson, _Classical Electrodynamics_ , Interscience, 1999. * [49] C.B. Hededal, and Å. Nordlund, _Gamma-Ray Burst Synthetic Spectra from Collisionless Shock PIC Simulations_ , submitted to ApJL, 2005, [arXiv:astro-ph/0511662].	arxiv-papers	2008-08-27T20:30:55	2024-09-04T02:48:57.505477	{ "license": "Public Domain", "authors": "K.-I. Nishikawa, Y. Mizuno, P. Hardee, H. Sol, M. Medvedev, B. Zhang,\n A. Nordlund, J. T. Frederiksen, G. J. Fishman, R. Preece", "submitter": "Ken-Ichi Nishikawa", "url": "https://arxiv.org/abs/0808.3781" }
0808.3948	# The doping-driven evolution of the superconducting state of a doped Mott insulator: a key for the high temperature superconductivity. M. Civelli1 1 Theory Group, Institute Laue Langevin, Grenoble, France ###### Abstract High-temperature superconductors at zero doping can be considered strongly correlated two-dimensional Mott insulators. The understanding of the connection between the superconductor and the Mott insulator hits at the heart of the high-temperature superconducting mechanism. In this paper we investigate the zero-temperature doping-driven evolution of a superconductor towards the Mott insulator in a two dimensional electron model, relevant for high temperature superconductivity. To this purpose we use a cluster extension of dynamical mean field theory. Our results show that a standard (BCS) d-wave superconductor, realized at high doping, is driven into the Mott insulator via an intermediate state displaying non-standard physical properties. By restoring the translational invariance of the lattice, we give an interpretation of these findings in momentum space. In particular, we show that at a finite doping a strong momentum-space differentiation takes place: non-Fermi liquid and insulating-like (pseudogap) character rises in some regions (anti-nodes), while Fermi liquid quasiparticles survive in other regions (nodes) of momentum space. We describe the consequence of these happenings on the spectral properties, stressing in particular the behavior of the superconducting gap, which reveals two distinct nodal and antinodal energy scales as a function of doping. We propose a description of the evolution of the electronic structure while approaching the Mott transition and compare our results with tunneling experiments, photoemission and magnetotransport on cuprate materials. ###### pacs: 71.10.-w,71.10.Fd,74.20.-z,74.72.-h ## I Introduction Since the discovery in 1986 of the high temperature (H-TC) superconductivity in copper oxide based materialsbednorz86 , much effort has been devoted by the scientific community to understand the physics of this phenomenon, but its key-ingredients remain still unknown. On the experimental side, the complexity of these materials, which present a rich phase diagram with many competing instabilities, has made impossible to perform resolutive experiments. On the theoretical side, many theories have been proposed, but a general consensus has not been achieved yet because of the lack of tools capable of perform reliable calculations. The strongly-interacting many-body nature of these systems makes in fact standard techniques hardly applicable. The recent discovery of high-temperature superconductivity in a new family of materialskamihara08 (for a perspective see e.g. ref.day08 ), with a general composition LaFeAsO1-xFx, different from the typical Cu-O plane structure of all known H-TC compounds, has revitalized the attention on the origin of the H-TC mechanism. In early times, P.W. Andersonanderson suggested that H-TC superconductivity is the result of doping a Mott insulator with a small number of carriers. The understanding of the way a superconductor can transform into a Mott insulator has been a longstanding open problem and its connection with a strongly- interacting superconductivity has been long sought. The aim of this article is to investigate the doping-driven evolution of the superconducting state of a doped Mott insulator in a two dimensional lattice model of strongly correlated electrons. To this purpose we employ a cluster extension of Dynamical Mean Field Theory (DMFT) (for a review see ref.revmodmft ). #### I.0.1 Dynamical Mean Field Theory and its Cluster extensions to study H-TC superconductivity DMFT has in recent years proved very successful in revealing the physics of the Mott transition in three dimensional compounds, like e.g. Vanadium Oxide. DMFT is a self-consistent mean field method, which maps a lattice problem onto a single-impurity embedded in a bath of free electrons. It is exact in the limit of infinite dimension, where the physics is purely local and DMFT virtually provides the complete solution. Its strength stands in the ability of treating on the same footing high and low energy physics, capturing in one framework both the Mott insulating state as well as the metallic state, and hence allowing a non-perturbative investigations. A good summary of the physics captured by DMFT is described in the DMFT phase diagram. For example, single-site DMFT in conjunction with electronic structure methods has provided results in excellent agreement with experiments on numerous three dimensional compounds (with valence electrons in the sp shellzein05 , 3d shell, 4d shelllich02 , 4f shellheld01 ; haule05 and 5f shellnature01 ). For some recent reviews of this field see e.g. ref.rmp06 . In spite of these successes in genuine strongly correlated materials, DMFT has also well understood limitations, mainly due to its purely local character. It does not capture for example the feedback of collective modes, such as magnetic fluctuations on the single particle quantities. One consequence of this shortcoming is that the effective mass diverges as the Mott transition is approached, while in finite dimensions exchange effects should remove this enhancement in the region where the renormalized kinetic energy is of the order of the super-exchange-interaction or the temperature. These effects are negligible whenever strong frustration, finite temperature or orbital degeneracy help justifying a local approximation. The DMFT approach, however, breaks down in the cuprate-based superconductors, whose physics is two- dimensional and which have a large super-exchange. As the Mott transition is approached in fact, experimental evidence shows that the effective mass remains finite and that physical properties are strongly momentum dependent (see for example Angle Resolved Photoemission Spectroscopy ARPES resultsdamascelli ; campuzano ). Being a local theory, DMFT is not able by construction to describe a strong momentum dependence, hence it is not suitable to study real finite-dimensional H-TC superconductors. Recently, extensions of DMFT capable to go beyond the local approximation have been developed. For a review of different cluster extensions of DMFT see e.g. ref.maier00 ; lich00 ; jarrell-rmp04 ; tremblay06 ; biroli04 . The Cellular Dynamical Mean Field Theory (CDMFTcdmft ) is a cluster method which retains the real space interpretation of single-site DMFT but incorporates non- perturbative momentum-dependence. It replaces the single site of DMFT by a cluster of impurities, offering the possibility of well describing short- ranged spatial correlation and providing a first correction to momentum- dependent properties. It also allows to naturally describe phases characterized by a spatially dependent order parameter, like for example antiferromagnetism (AF) or d-wave superconductivity (dSC). The first cluster- DMFT studies of the Hubbard model in two dimensions were carried out on a 2 $\times$ 2 plaquette in ref.lich00 ; maier00 . These groups have shown the existence of antiferromagnetism, pseudogap and superconductivity, focusing on relatively small values of the on-site interaction (mostly U =4t), for which the Fye-Hirsch Quantum Monte Carlo (QMC) fye-hirsch86 ; fye-hirsch89 method is applicable in implementing the cluster DMFT. One school of thought, for example in ref.jarrell05 ; jarrell-scalapino06 ; senechal04 , advocates the study of large clusters, at present possible for relatively small interactions. In this latter case, however, it is not yet known what aspects of the Mott physics are captured. Cluster DMFT implemented on the 2$\times$2 plaquette for stronger values of U (the Mott regime) were the subject of several publicationsbpk ; marce05 ; bumsoo06 ; venky05 on the Hubbard Model. #### I.0.2 A 2$\times$2 plaquette Cellular DMFT study Following the cluster DMFT studies mentioned above, we focus in this paper on a two-dimensional 2$\times$2 cluster-plaquette. According to our viewpoint, elucidating the physical content of the mean field theory on a plaquette is a very important step to accomplish before proceeding to realistic studies of the actual instabilities that govern the phase diagram of the model. The work in referencessenechal05 ; massimo06 ; aichhorn06 ; aichhorn06b on small clusters have in fact shown that there are several competing phases and possible phase-separation, which could lead to complicate patterns in real space depending on the boundary conditions or on various terms not explicitly included in the Hamiltonian. A proper investigation of those states in the framework of DMFT-based methods, requires therefore an understanding of the pure phases of the simple cluster-plaquette. Implementation of CDMFT on a 2$\times$2 plaquette for large values of the interaction has already revealed several interesting effects not present in single-site DMFT, indicative of a very rich physics. For example, publicationsbpk ; marce05 reveal that the approach to the Mott transition as a function of doping occur in a very anisotropic way in momentum space. In the Hubbard model (relevant for the description of hole-doped cuprates), the spectral weight disappears much more rapidly in the antinodal than in the nodal region of momentum space, where quasiparticles survive close to the Mott transition at a fixed temperature. Evidence for the formation of a pseudogap in the one electron spectra can be seen in other cluster-DMFT studieshuscroft- jarrell01 ; stanescu03 ; senechal05 . CDMFT allows a natural interpolation of the nodal and antinodal spectral function, and it opens the way to study the $k$-dependence of the one-electron spectra in the Mott insultorbumsoo06 . The origin of the pseudogap can be traced to the growth of the self energy in certain parts of $k$-space, where eventually at zero temperature lines of poles of the self-energy (i.e. zeroes of the one-particle Green’s function) appeartsvelik ; tudor ; berthod . In this latter case, the Fermi arcs observed in ARPES are interpreted as the result of a binding of segments of Fermi surface and of a lines of zeros of the Green’s function. Results along those lines beyond mean field theory can be found in ref.stanescu-2007-75 . Within CDMFT not only the under-doped regime is anomalous. At optimal doping, where the maximum of the critical temperature occurs, a maximum in the one electron scattering rate and the presence of power laws in the optical conductivity are foundhaule06 . In this paper we study the superconducting state which arises upon doping a Mott insulator by implementing the 2$\times$2 plaquette CDMFT with exact diagonalizationkrauth-caffarel (ED-CDMFT) at zero temperature (Lanczos method). Our work is complementary to the 2$\times$2 plaquette CDMFT study of ref.haule-ctqmc , where the continuous time quantum Monte Carlo (CTQMC) at finite temperature was used as impurity solver. As compared with QMC methods, ED allows to extract in an unbiased way real frequency quantities, which can be more easily physically interpreted. This will turns out to be fundamental in interpreting our CDMFT results and making important connection with experiments (as we will widely present throughout the paper). The Lanczos method, on the other hand, is limited by the finite size of the system used to describe an effective impurity model, associated with the original lattice model (see the following discussion on the ED-CDMFT procedure). If rightly implemented, however, it is able to well capture the difficult physics of the finite dimensional Mott transition (see for example referencesmarce ; rmp06 ; civelli-2007 ). We will in fact show that the physics described by our results well compares (at least at the qualitative level) with the CTQMC-CDMFT results of referencehaule-ctqmc , where finite size limitations are absent. Earlier studiesvenky05 with ED-CDMFT have compared the phase diagram of the two dimensional Hubbard Model with the one of cuprate-based materials, studying the interplay between the superconducting and antiferromagnetic instabilities, either in the hole and electron doped sides. Here we complete the work presented in a previous short publicationmarce08 , and we focus on doping-driven evolution of the superconductor towards the Mott insulator. In particular we show that, in a small region around optimal doping, novel properties appear in the superconducting state which are not ascribable to the standard Bardeen Cooper Schrieffer (BCS) theory of superconductivity. The most striking phenomenon, as widely presented in publicationmarce08 , is the rising of two distinct doping-dependent energy scales, which do not fit neither into the framework of BCS approaches nor within the most popular theories of H-TC superconductivity, like for instance the resonating valence bond theories (RVBanderson , for a recent review see e.g.patrickrmp ). We interpret our cluster results extracting the momentum-dependence (i.e. restoring the translational invariance of the lattice) of the electronic structure as a function of doping. According to our physical picture, the Fermi liquid description holds at low energies in the nodal region of momentum space, where the d-wave superconducting gap is zero. We show that this fact is best described in momentum space introducing a periodizing scheme based on the local (within cluster) self-energybiroli02 ; bpk . On the other hand, in the antinodal region, where the d-wave gap is maximal, besides the superconducting contribution, a further contribution to the one-particle gap appears also in the normal component at the finite critical doping. In this case we show that the periodization of another local quantity, the irreducible two-point cumulant, offers a more adequate descriptiontudor ; tudor06 . We introduce therefore a mixed-periodization scheme, which was foreshadowed in a phenomenological approach to the transport properties of cuprate materials in the normal statePSK and in a previous CDMFT publicationmarce08 , and confront the resulting picture of the electronic evolution as a function doping with photo-emission, scanning tunneling experiments and magnetotransport on cuprate materials. #### I.0.3 Set-up of the paper The paper consists of two main parts. The first comprises sections I-IV. In section I, after this introduction, we present the model and briefly explain the ED-CDMFT method. In section II we present the raw cluster DMFT results, mainly stressing their evolution with doping and showing the appearance (at small doping) of two distinct energy scales. It is not however straightforward to interpret these cluster quantities in terms of physical observables. Therefore, in section III, we cope with the problem of extracting lattice quantities (which can be more easily compared with experimental results) from the corresponding cluster ones, i.e. we present and justify different periodization-methods. We show that nodal and antinodal regions of momentum space turn out having different physical properties which require different periodization schemes to be rightly described. In section IV we characterize the different properties in the nodal and antinodal regions, according to the periodization schemes introduced, clearly linking them with spectroscopy experiments. The second part starts from section V, where, in order to be able to compare our results with experiments resolved in momentum space, we introduce a more general mixed-periodization scheme. This latter, interlacing the nodal and antinodal properties in all the momentum space, allows us to propose a description of the evolution of quasiparticle spectra in approaching the Mott transition. Spectra, which are directly comparable with photo-emission on the cuprates, are presented in detail in section VI. In section VII we present peculiar characteristics of the spectra, the so called ”kink” feature, also comparing with photoemission results. Finally, in section VIII, by applying a phenomenological Boltzmann approach on our CDMFT mixed-periodization result, we derive the Hall resistivity, which is a direct probe of the charge carries in the system. We compare its evolution as a function of doping with magnetotransport experiments and draw conclusions on a topological phase transition of the Fermi surface, which takes place at low doping within the mixed-periodization scheme introduced in section V. We finally derive our conclusions in section IX. ### I.1 The Model We consider the one-band two dimensional Hubbard Model on a square lattice: $\mathcal{H}=-\sum_{\langle i,j\rangle,\sigma}t_{ij}\,(c^{\dagger}_{i,\sigma}c_{j,\sigma}+h.c.)+U\sum_{i}n_{i\uparrow}n_{i\downarrow}-\mu\sum_{i}n_{i},$ (1) which is universally considered a minimal description of cuprate-based materialshubbard63 . Here $c_{i,\sigma}$ ($c^{\dagger}_{i,\sigma}$) are destruction (creation) operators for electrons of spin $\sigma$, $n_{i\sigma}=c^{\dagger}_{i\sigma}c_{i\sigma}$ is the density of electrons, $\mu$ is the chemical potential tuning doping and $t_{ij}$ are the orbital hopping integrals. For convenience’s sake we consider only the nearest neighbor amplitude $t=1$, and a next nearest neighbor hopping $t^{\prime}=-0.3t$. We set the on-site repulsion $U=12t$, larger than the band-width $8t$, to be in the Mott regime. ### I.2 ED-CDMFT procedure Similarly to single-site DMFT revmodmft , in CDMFT the original model (eq. 1) is described in terms of an effective action containing a Weiss dynamical field $\hat{\cal G}_{0}(\tau)$ describing the degrees of freedom outside the cluster (in the bath) as a time dependent hopping within the cluster $S_{\rm eff}=\int_{0}^{\beta}d\tau d\tau^{\prime}\Psi_{\tau}^{\dagger}\left[\hat{\cal G}_{0\tau-\tau^{\prime}}^{-1}\right]\Psi_{\tau^{\prime}}+U\sum_{\mu}\int_{0}^{\beta}n_{\mu\uparrow}n_{\mu\downarrow}d\tau.$ (2) $\mu=1..N_{c}$ ($N_{c}=4$ in the 2$\times$2 plaquette) labels the degrees of freedom inside the cluster. For the case of a $2\times 2$ plaquette considered in this paper, a convenient Nambu-spinor notation has been introduced: $\Psi^{\dagger}\equiv(c_{1\uparrow}^{\dagger},\dots,c_{4\uparrow}^{\dagger},c_{1\downarrow},\dots,c_{4\downarrow})$ (3) With this notation the Weiss field $\hat{\cal G}_{0}$ is a $8\times 8$ matrix with both normal (particle-hole) and anomalous (particle-particle) components venky05 . Physically, this action describes a cluster embedded in a self- consist bath of free electrons with dSC correlations. In the CDMFT procedure, a starting guess of the Weiss field $\hat{\cal G}_{0}$ is given as input. Then the cluster single-particle propagator $\hat{G}_{c}$ is computed trough the effective action eq. (2) and the cluster self-energy is determined through the Dyson’s equation $\hat{\Sigma}_{c}=\hat{\cal G}_{0}^{-1}-\hat{G}_{c}^{-1}$ (4) Here, $\hat{G}_{c}\left(\tau,\tau^{\prime}\right)=\left(\begin{array}[]{cc}\hat{G}_{\uparrow}\left(\tau,\tau^{\prime}\right)&\hat{F}\left(\tau,\tau^{\prime}\right)\\\ \hat{F}^{\dagger}(\tau,\tau^{\prime})&-{\hat{G}}_{\downarrow}\left(\tau^{\prime},\tau\right)\end{array}\right)$ (5) is an 8 X 8 matrix, $G_{\mu\nu,\sigma}\equiv\langle- Tc_{\mu\sigma}(\tau)c^{\dagger}_{\nu\sigma}(0)\rangle$ and $F_{\mu\nu}\equiv\langle-Tc_{\mu\downarrow}(\tau)c_{\nu\uparrow}(0)\rangle$, ($\mu,\nu=1...N_{c}$ label sites of the cluster) are the normal and anomalous cluster-Green’s functions respectively. From the cluster self-energy $\hat{\Sigma}_{c}$, we use the CDMFT self-consistency condition to re-compute the local cluster Green’s function $\hat{G}_{loc}(i\omega_{n})=\sum_{K}\,\hat{G}(K,i\omega_{n})$, where $\hat{G}(K,i\omega_{n})=\left[i\omega_{n}+\mu-\hat{t}(K)-{\hat{\Sigma}_{c}}(i\omega_{n})\right]^{-1}$ (6) In eq.(6) $\hat{t}(K)$ is the Fourier transform of the hopping matrix defined on the lattice formed by the clusters and the sum over $K$ is therefore performed over the Brillouin zone reduced by the partition in clusters of the latticevenky . We finally re-derive a new Weiss field $\hat{\mathcal{G}}^{new}_{0}(i\omega_{n})^{-1}=\hat{G}^{-1}_{loc}(i\omega_{n})+\hat{\Sigma}_{c}(i\omega_{n})$ (7) and iterate until convergence is reached. In practice, as we mentioned in the introduction, in order to solve the cluster impurity problem, in this work (as in ref.venky05 ; marce08 ) we use the Exact Digonalization method krauth-caffarel . A parametrized Anderson- impurity Hamiltonian describes the action eq.(2) and couples the cluster impurity with a discrete number $N_{b}$ of bath orbitals (we have fix throughout this work $N_{b}=8$, which is the limit in practice accessible with standard computational resources) : $\displaystyle\mathcal{H}_{\rm imp}$ $\displaystyle=$ $\displaystyle\sum_{\mu\nu\sigma}E_{\mu\nu\sigma}c_{\mu\sigma}^{\dagger}c_{\nu\sigma}+U\sum_{\mu}n_{\mu\uparrow}n_{\mu\downarrow}+$ (8) $\displaystyle+\sum_{k\sigma}\epsilon_{k\sigma}a_{k\sigma}^{\dagger}a_{k\sigma}+\sum_{k\mu\sigma}V_{k\mu\sigma}a_{k\sigma}^{\dagger}c_{\mu\sigma}+{\rm h.c.}+$ $\displaystyle+\sum_{k\mu\sigma}V^{\rm sup}_{k\mu\sigma}a_{k\sigma}^{\dagger}c^{\dagger}_{\mu\bar{\sigma}}+\sum_{k\mu\sigma}V^{\dagger\rm sup}_{k\mu\sigma}c_{\mu\bar{\sigma}}a_{k\sigma}$ Here $E_{\mu\nu\sigma}=-\mu\delta_{\mu\nu}$, $E_{\mu\nu\sigma}=-t\delta_{\mu,\nu\pm 1}$. Under the self-consistency constrain eq.(6) and (7), the bath-parameters $\epsilon_{k\sigma}$, $V_{k\mu\sigma}$ and $V^{\rm sup}_{k\mu\sigma}$ are determined at each CDMFT- iteration by fitting the Anderson-impurity Weiss field (eq. 7) with a $N_{b}$-pole bath function $\hat{\mathcal{G}}^{new}_{N_{b}}(i\omega_{n})=\imath\mathbf{1}\omega-\hat{E}-\hat{\Delta}$ $\displaystyle\hat{\Delta}=\sum_{k}^{N_{b}}\,\mathbf{V}_{k}^{\dagger}(\imath\mathbf{1}\omega-\hat{\varepsilon}_{k})^{-1}\mathbf{V}_{k}$ (9) $\displaystyle\mathbf{V}_{k}(2\times 8)=\left(\begin{array}[]{cc}V_{k\mu\uparrow}&V^{sup}_{k\mu\uparrow}\\\ -V^{sup}_{k\mu\downarrow}&-V_{k\mu\downarrow}\\\ \end{array}\right)_{\mu=1,N_{c}}$ (12) $\displaystyle\mathbf{E}_{k}(2\times 2)=\left(\begin{array}[]{cc}\varepsilon_{k\uparrow}&0\\\ 0&-\varepsilon_{k\downarrow}\\\ \end{array}\right)$ (15) The fitting is obtained via a conjugate gradient minimization algorithm, which uses a distance functionmarce $f=\sum_{\mu\nu}\,\left\|\hat{\mathcal{G}}^{new}_{0}(i\omega_{n})-\hat{\mathcal{G}}^{new}_{N_{b}}(i\omega_{n})\right\|_{\mu\nu}/\omega_{n}$ (16) that emphasizes the lowest frequencies and it is computed on the imaginary frequency axis $\omega_{n}=(2n-1)\pi/\beta$. This introduces an effective inverse temperature, which is a fitting parameter (it is not the real temperature which is $T=0$ in our study) and is set $\beta=300t$ (much higher than the $\beta=50t$ used in publication of ref.venky05 ) throughout the whole paper. This parameter determines the energy resolution accessed in this work (see also appendix A). On a practical level, to start the ED-CDMFT procedure it is most useful to introduce a reduced parameterization of the bath, which enlightens the symmetries of the input-guess $\hat{\mathcal{G}}_{0}(i\omega_{n})$. The constrain on the bath parameters can be then relaxed in a second step. Further details are given in appendix B. ## II Cluster Results We start this section by considering raw cluster quantities, which directly output from the cluster-impurity solution. ### II.1 A d-wave Superconducting state Figure 1: (Color online). Left panel: doping $\delta=1-<n>$ versus chemical potential $\mu$. Right panel: cluster d-wave order parameter dOP= $<<c_{\mu\uparrow}c_{\mu+1\downarrow}>>_{(\tau=0)}$ as a function of doping $\delta$. As we explained above, within CDMFT a superconducting state can be studied by allowing particle-particle pairing terms $V^{\rm sup}_{k\mu\sigma}a_{k\uparrow}c_{q\downarrow}$ in the impurity Hamiltonian eq.(8) (or equivalently upon a unitary rotation, pairing bath terms $a_{\mu\uparrow}a_{k\downarrow}$, see Appendix B). CDMFT-self-consistency condition may accept non-zero values of the $V^{\rm sup}_{k\mu\sigma}$ terms, resulting in a non-zero superconducting-pair Green’s function $F_{\mu,\nu}(\tau)=T_{\tau}<c_{\mu\uparrow}(\tau)c_{\nu\downarrow}>$. In drawing a parallel with the classical mean field example of the Ising Modelrevmodmft , this is equivalent to assume a non-zero on-site magnetization $m=<S^{z}_{i}>$ as starting hypothesis in building an effective Hamiltonian, and to determine then $m$ self-consistently. As reported in previous work venky05 , CDMFT supports indeed a d-wave superconducting state in a region of doping that precedes the Mott transition. This is shown in Fig. 1, where we show the doping $\delta=1-<n_{i}>$ versus the chemical potential $\mu$ and the d-wave order parameter (dOP), defined as $F_{\mu,\mu\pm 1}(\tau=0)$, also as a function of $\delta$. As expected, $\delta$ versus $\mu$ is monotonically decreasing. The dOP has a dome-like shape and its sign is alternating by exchange of the $x$-$y$ bonds on the cluster plaquette. We can locate the maximum only with some degree of uncertainty around $\delta_{p}\sim 0.06<\delta<\delta_{c}\sim 0.08$. The two doping values $\delta_{p}\sim 0.06$ and $\delta_{c}\sim 0.08$ present as two special points, which mark changes in the physical properties of the system. We will come back on these two points in throughout the paper. Here we stress that at $\delta_{c}$ in particular, our dynamical mean field result seems to branch two different lines of solution. This is evident either in the $\delta-\mu$ and the dOP plots. In mean field approaches this behavior may be the signature of a phase transition. It is therefore intriguing that we find such a behavior close to the maximum of the dOP dome. Following the nomenclature typical for cuprate materials, we will call hereafter the region $\delta<\delta_{p}$ under-doped, $\delta>\delta_{c}$ over-doped and $\delta_{p}\sim 0.06<\delta<\delta_{c}\sim 0.08$ optimal doping region. It is clear we do not intend to draw a quantitative parallel with cuprate-based system, where the typical optimal doping (unambiguously defined as the maximum of critical temperature $T_{C}$) is around $\delta_{c}\sim 0.15$, but we rather follow the qualitative aspects of the physics of these material, marking a correspondences with our CDMFT results on the Hubbard Model in two dimensions. ### II.2 Cluster self-energies It is worth to investigate in detail the cluster-outputs. The typical output of the CDMFT-scheme is a cluster-self-energy $\Sigma_{\mu\nu}$ (eq. 4), which can be expressed as a 2$\times$ $N_{c}\times N_{c}$ ($N_{c}=4$ for the 2$\times$2 plaquette) matrix with normal and anomalous components in the Nambu notation introduced in eq. (3). We first look at the normal components, which are shown on the Matsubara-frequency axis in Fig. 2. Figure 2: (Color online). Normal components of the cluster self-energy $\Sigma_{\mu\nu}$ vs the Matsubara frequency $\omega_{n}$ as a function of doping $\delta$. In the left column the real parts are displayed, in the right column the imaginary parts. From the top row to the bottom we show the local $\Sigma_{11}$, the next neighbor $\Sigma_{12}$, and nearest next neighbor $\Sigma_{13}$ self-energies. Figure 3: (Color online). Eigenvalues of the normal component of the cluster self-energy $\Sigma_{\mu\nu}$ vs the Matsubara frequency $\omega_{n}$ as a function of doping $\delta$. In the left column the real parts are displayed, in the right column the imaginary parts. From the top row to the bottom we show the self-energy corresponding to the points $k=(0,0)$, $(0,\pi)$ and $(\pi,\pi)$ in the first quadrant of the Brillouin Zone (see section III). The real parts are displayed on the left and the imaginary parts on the right column from high (in our solution) doping $\delta=0.14$ (the over-doped side) until close to the Mott transition for $\delta=0.02$ (in the under-doped side). From the top row to the bottom we have the local self-energy $\Sigma_{11}$, the next neighbor $\Sigma_{12}$ and the nearest next neighbor $\Sigma_{13}$. At high doping, the local components is dominant. We expect in this region that the single site DMFT is already a good approximation. By decreasing $\delta$, however, also the non-local components $\Sigma_{12}$ and $\Sigma_{13}$ grow considerably, becoming comparable with the local one. This effect is totally missed by single site DMFT and it is only captured by using a cluster extension. The growing of the non local components of $\Sigma_{\mu\nu}$ determines the physical properties of the system in approaching the Mott transition in a fundamental way, as we explain in the following. The imaginary parts (right column) display a low-energy Fermi- liquid-like behaviour Im$\Sigma\sim\alpha\omega_{n}$. The slope $\alpha$ is connected with the quasiparticle residuum (see the following). Here we notice that, with respect to the other components, the energy range of linear behavior in the nearest next neighbor Im$\Sigma_{13}$ is narrower and that its slope $\alpha$ also changes behavior in passing through the doping $\delta_{c}\sim 0.08$ (it grows for $\delta<\delta_{c}$, decreases until becoming negative for $\delta>\delta_{c}$), while it also changes sign at $\delta_{c}\sim 0.06$. This behavior is the only mark we find in the cluster self-energy of the special nature of the points $\delta_{c}$ and $\delta_{p}$ that we identified in the Fig. 1. The local and next neighbor components show instead a smooth continuous behavior as a function of doping $\delta$. As in ref.marce05 ; tudor06 , it is convenient to look at the eigenvalues of the cluster-self-energy matrix, which can be directly related to the corner points of the quadrant of the Brillouin Zone (as we will explain in more detail in the following section). A convenient way is to re-express the Nambu- spinor notation eq.(3) by grouping two by two the up and down spin construction/destruction operators on each cluster-site: $\Psi^{\dagger}\equiv\left[(c_{1\uparrow}^{\dagger},c_{1\downarrow})\dots,(c_{4\uparrow}^{\dagger},c_{4\downarrow})\right]$ (17) The cluster self-energy matrix assumes the form: $\mathbf{\Sigma}_{c}=\,\left(\begin{array}[]{cccc}\hat{\Sigma}_{0}&\hat{\Sigma}_{1x}&\hat{\Sigma}_{2}&\hat{\Sigma}_{1y}\\\ \hat{\Sigma}_{1x}&\hat{\Sigma}_{0}&\hat{\Sigma}_{1y}&\hat{\Sigma}_{2}\\\ \hat{\Sigma}_{2}&\hat{\Sigma}_{1y}&\hat{\Sigma}_{0}&\hat{\Sigma}_{1x}\\\ \hat{\Sigma}_{1y}&\hat{\Sigma}_{2}&\hat{\Sigma}_{1x}&\hat{\Sigma}_{0}\\\ \end{array}\right)\\\ $ (18) where the $2\times 2$ matrices are function of cluster elements $\Sigma_{\mu\nu,\sigma}=\Sigma_{\|\mu-\nu\|,\sigma}$: $\displaystyle\hat{\Sigma}_{0}=\,\left(\begin{array}[]{cc}\Sigma_{0\uparrow}&0\\\ 0&-\Sigma_{0\downarrow}\\\ \end{array}\right)$ $\displaystyle\hat{\Sigma}_{1x}=\,\left(\begin{array}[]{cc}\Sigma_{1\uparrow}&\Sigma_{ano}\\\ \Sigma_{ano}&-\Sigma_{1\downarrow}\\\ \end{array}\right)$ (23) $\displaystyle\hat{\Sigma}_{2}=\,\left(\begin{array}[]{cc}\Sigma_{2\uparrow}&0\\\ 0&-\Sigma_{2\downarrow}\\\ \end{array}\right)$ $\displaystyle\hat{\Sigma}_{1y}=\,\left(\begin{array}[]{cc}\Sigma_{1\uparrow}&-\Sigma_{ano}\\\ -\Sigma_{ano}&-\Sigma_{1\downarrow}\\\ \end{array}\right)$ (28) We follow the procedure used to diagonalize the cluster-self-energy matrix in ref.haule06b : $\mathbf{\Sigma}_{c}=\,\left(\begin{array}[]{cccc}\hat{\Sigma}_{00}&0&0&0\\\ 0&\hat{\Sigma}_{\pi\pi}&0&0\\\ 0&0&\hat{\Sigma}_{0\pi}&0\\\ 0&0&0&\hat{\Sigma}_{\pi 0}\\\ \end{array}\right)$ (29) The diagonal elements are linear combination of the original cluster self- energies matrices: $\displaystyle\hat{\Sigma}_{00}$ $\displaystyle=$ $\displaystyle\,\hat{\Sigma}_{0}+\tilde{\Sigma}_{1}+\hat{\Sigma_{2}}$ $\displaystyle\hat{\Sigma}_{0\pi}$ $\displaystyle=$ $\displaystyle\,\hat{\Sigma}_{0}+\tilde{\Sigma}_{a}-\hat{\Sigma_{2}}$ $\displaystyle\hat{\Sigma}_{\pi 0}$ $\displaystyle=$ $\displaystyle\,\hat{\Sigma}_{0}-\tilde{\Sigma}_{a}-\hat{\Sigma_{2}}$ $\displaystyle\hat{\Sigma}_{\pi\pi}$ $\displaystyle=$ $\displaystyle\,\hat{\Sigma}_{0}-\tilde{\Sigma}_{1}-\hat{\Sigma_{2}}$ (30) For convenience’s sake we have defined : $\displaystyle\tilde{\Sigma}_{1}=\,2\left(\begin{array}[]{cc}\Sigma_{1\uparrow}&0\\\ 0&-\Sigma_{1\downarrow}\end{array}\right)$ $\displaystyle\tilde{\Sigma}_{a}=\,2\left(\begin{array}[]{cc}0&\Sigma_{ano}\\\ \Sigma_{ano}&0\end{array}\right)$ (35) Notice that from eq. (30) the anomalous self-energy appears only in the $(0,\pi)$ and $(\pi,0)$ components, where the dSC gap is expected to open. The $2\times 2$ eigenvalue-matrices of the cluster-self-energy are interpreted as describing the four momentum-space points $(0,0)$, $(0,\pi)$, $(\pi,0)$ and $(\pi,\pi)$ (hence the choice of the labels), as we will explain in detail in the following section III. Figure 4: (Color online). Cluster quasiparticle residua, associated with the eigenvalues of the cluster self-energy matrix, as a function of doping $\delta$. The normal components of the eigenvalues of $\hat{\Sigma}_{c}$ are shown in Fig. 3, in a fashion similar to Fig. 2. The first striking difference with respect to the Fig. 2 is that the eigenvalue $\Sigma_{00}$ changes little as a function of doping (either the real and the imaginary parts). $\Sigma_{0\pi}$ and $\Sigma_{\pi\pi}$ both grow in reducing doping $\delta$, but the growth in $\Sigma_{\pi\pi}$ is on order of magnitude bigger. This eigenvalue is indeed the one that drives the system into the Mott insulating state as $\delta\to 0$, as evident from the big values of both the real part and of the slope of the imaginary part at small doping. This behavior are very similar (i.e. it appears as a smooth continuation) to the one displayed by the self-energy of the normal state ED-CDMFT study published in ref.marce05 . With respect to this latter study however, here we display the normal components of a superconducting solution (while in ref.marce05 no superconductivity is allowed) and the energy resolution achieved ($\beta t=300$) is an order of magnitude smaller than in ref.marce05 ($\beta t=32$). This allows us to extract the very low-energy properties. In particular, as expected by the fact that $\hat{\Sigma}$ has to be negative in order to respect casuality, we can observe in Fig. 3 that the slope $\alpha$ of the Im$\Sigma\sim\alpha\omega_{n}$ is always negative. The fact then the $\omega_{n}\to 0$ behavior of Im$\Sigma$ is linear indicates that these eigenvalues components have Fermi liquid properties. Following a standard Fermi liquid approach, it is instructive to define cluster quasiparticle residua $Z_{X}=\left(1-\partial\Sigma_{X}/\partial\omega_{n}\right)$ with $X=(0,0),(0,\pi),(\pi,\pi)$ , even if these quantities have a real physical meaning only in correspondence of a real Fermi surface (i.e. $Z_{X}$ has to be interpreted here as renormalized cluster quantity). In the standard picture of the Mott transitionrevmodmft , the quasiparticle residuum $Z\to 0$ as doping is reduced $\delta\to 0$. From Fig. 4 we see that only $Z_{(\pi,\pi)}$ appears to display this behavior, while $Z_{(0,0)}$ and $Z_{(0,\pi)}$ clearly extrapolate to a non-zero value. A last remark concerns once again the special critical doping $\delta_{c}$, where remarkably the $Z_{(0,\pi)}$ shows a clear change in behavior. For $\delta>\delta_{c}$ it closely follows $Z_{(\pi,\pi)}$, decreasing with doping as expected in the standard Mott transition picture. At $\delta_{c}$, however, $Z_{(0,\pi)}$ departs from $Z_{(\pi,\pi)}$, and shows a behavior more similarly to $Z_{(0,0)}$. This once again indicates a quick change in the physical properties of the system in correspondence of $\delta_{c}$. We will come back later to discussing the physical interpretation of these observations on the cluster quasiparticle residua (see section IV). Figure 5: (Color online). Left panel: real part of the anomalous component of the cluster-self-energy $\Sigma_{ano}$ (the imaginary part is negligible) vs Matsubara frequency $\omega_{n}$ for different doping $\delta$. Right panel: Re$\Sigma_{ano}(\omega_{n}\to 0)$ versus doping $\delta$. We now turn to the anomalous component of $\hat{\Sigma}_{c}$, which we show in the left side of Fig. 5 on the Matsubara frequency $\omega_{n}$. As resulting output of our CDMFT solution, only the real part of the nearest-neighbor component is appreciably non-zero on the Matsubara axis, and it assumes a d-wave sign bond-alternating value on the cluster-plaquette. Re$\Sigma_{ano}$ starts assuming appreciably non-zero values in the over-doped side $\delta=0.14$, and grows for decreasing doping, until $\delta_{c}\sim 0.08$ is reached. In the under-doped side however at $\delta<\delta_{p}\sim 0.06$ the curves change tendency, and they decrease by further decreasing doping. The qualitative change in the behavior is better enlightened by looking at the $\omega\to 0$ limit: in the under-doped side the curves reach $\omega=0$ with a finite slope, while in the under-doped side (after $\delta\leq 0.08$) the slope has a smaller value. In right side of Fig. 5 we show the $\omega_{n}\rightarrow 0$ extrapolated value of Re$\Sigma_{ano}$ as a function of doping $\delta$. In first approximation this value can be related to the superconducting gap (as we will explain in detail in the section III, see eq. 49). The behavior of the Re$\Sigma_{ano}$ as a function of doping shows therefore to be non-monotonic and roughly tracks the behavior of the order parameter (see Fig. 1), similarly to standard BCS theory. This result is fundamental and it is in striking contrast with slave-boson resonating valence bond theoriesanderson ; liu ; patrickrmp , where the amplitude associated to a particle-particle paring channel is a monotonically decreasing function with doping and it has its maximum close to the insulating transition. In section IV we will better discuss the physical consequences of these results, making connection with recent spectroscopy experiments on cuprate-based materials. With the ED-CDMFT procedure it is straightforward to analytically continue on the real axisrevmodmft (differently from other computational methods like e.g. Quantum Monte Carlo, which require further approximate methods). The cluster Green’s function $\mathbf{G_{c}}(\omega_{n})$ and the cluster self- energy $\mathbf{\Sigma_{c}}(\omega_{n})$ (via the Dyson’s equation 4), are expressed in a pole expansion form containing terms of the form $b/(\omega_{n}-a)$. To go from the Matsubara to the real axis it is then enough to substitute $i\omega_{n}\to\omega+i\eta$, where $\eta$ is a small broadening introduced to display poles (for more details see appendix A). The price to pay comes from the discreteness introduced by truncating the bath in the impurity model (eq. 8) with a finite number $N_{b}$ of orbitals. This reproduces continuous functions, like $\mathbf{G_{c}}(\omega_{n})$ and $\mathbf{\Sigma_{c}}(\omega_{n})$, through a finite number of poles. In Fig. 6 we display on the real axes the local density of states $N(\omega)=\frac{1}{\pi}\left[G_{c}(\omega)\right]_{11}$, obtained from the cluster-impurity output (see eq. 4), for different doping $\delta$ (from top to bottom), using a small imaginary broadening i$\eta=7\hbox{i}\times 10^{-3}$. As we will explain in detail in the following section, in a d-wave superconducting state a linear in $\omega$ ”V” shaped density of states is expected for $\omega\to 0$. This behavior cannot be captured by the discreteness of our ED-CDMFT solution for small $\omega$, hence we have rather a ”U” shape for $\omega\to 0$ given by the broadening $\eta$. The shape of $N(\omega)$ and the energy scales of the superconducting gap can however be estimated by the location and intensities of the peaks. So, in looking at $N(\omega)$ from the over-doped side (top row, left panel for $\delta=0.16$) to the under-doped side (bottom row, left panel for $\delta=0.02$) of the phase diagram, we can make interesting observations: 1. 1. $N(\omega)$ is asymmetric function around $\omega=0$ at low doping ($\delta<\delta_{c}$). A rather symmetric shape is instead observed around optimal doping $\delta_{c}\sim 0.08>\delta>\delta\sim 0.06$ (in agreement with previous cluster DMFT results of ref.haule-ctqmc ) . 2. 2. the total superconducting gap, which can be evaluated by measuring the distance of the spectral peaks from $\omega=0$, is increasing by reducing doping $\delta$. Figure 6: (Color online). Local density of state $-\frac{1}{\pi}G_{11}(\omega)$ obtained from the cluster-impurity solution (eq. 8) for various doping $\delta$ (from top to bottom). The broadening used for display is $\eta=7\times 10^{-3}$. A measure of the energy scale associated with the one-particle gap (always present in the superconducting state) is given by the left (red circle) and right (green square) distances of the ED poles from the Fermi level $\omega=0$. These are displayed as a function of doping $\delta$ in the right bottom panel. To better elucidate these observations, we have measured the distance from the Fermi level of the left (red circle) and right peaks (green square), and displayed them as a function of doping $\delta$ (bottom row, right panel). In the over-doped side ($\delta>\delta_{c}$) the peaks are equally distant from $\omega=0$, i.e. the superconducting d-wave gap is symmetric at low frequency, as expected in a standard d-wave BCS theory. At optimal doping $\delta_{p}<\delta<\delta_{c}$, however, the asymmetry in the gap strikes in, and it is present in all the under-doped side. This is in agreement with the cluster DMFT results obtained in ref.haule-ctqmc and it is in nicely agreement with Scanning Tunneling experiments (STMdavis05 ) on cuprate materials. The fact that $N(\omega)$ is most symmetric at optimal doping $\delta_{p}<\delta<\delta_{c}$ (as evident from the red circles which display a change in the slope as a function of $\delta$) is in line with the observation carried in Fig. 1 on the special nature of this region (consistently e.g with the avoided-quantum-critical-point scenario proposed in the CDMFT study of ref.haule06 ). The black dashed line is the average of the right and left gaps, and the fact that it is increasing by reducing doping (similarly to the predictions of resonating valence bond slave boson theoriesanderson ; liu ; patrickrmp ) shows the important result on the presence in our solution of another energy scale, different from the energy scale marked by the behavior of Re$\Sigma_{ano}(\omega_{n}\to 0)$ (which is instead decreasing at low doping ($\delta<\delta_{p}$), see right panel Fig. 5). The interpretation of these two energy scales brings to interesting physical insights, which can be put in relation with recent experiments on the cuprate materials tacon06 ; tanaka06 ; kondo07 ; gomes2007 ; deutscher99 ; millis06 ; cho06 ; huefner-2008-71 . This has been presented in a previous short publicationmarce08 and it will be re-proposed more in detail in the following sections. ## III Momentum dependent quantities: periodization procedures We want now to interpret the cluster results we presented in the previous section in terms of physical observables, which could be possible related to experiments. The relevant information is typically embodied in the one- particle Green’s function, which in a superconducting state can be conveniently written in a Nambu-matrix notation: $\begin{array}[]{l}\mathbf{G}^{-1}_{\sigma}(k,\omega)=\\\ \\\ \,\left(\begin{array}[]{cc}\omega-\xi_{k}-\Sigma_{k}(\omega)&-\Sigma_{ano}(k,\omega)\\\ -\Sigma_{ano}(k,\omega)&\omega+\xi_{k}+\Sigma^{\ast}_{k}(-\omega)\end{array}\right)\end{array}$ (36) Here $\xi_{k}=t_{k}-\mu$ is the free band dispersion of our model (see equation 1). In order to determine the Green’s function we need therefore to determine the momentum dependent self-energy from the cluster solution, i.e. we need a periodization scheme. In previous workbpk ; marce05 ; bumsoo06 various periodization schemes have been proposed. The idea consists in determining the most local quantity $W_{r}$, which can be captured within the dimension of the cluster-impurity, and construct its truncated Fourier expansionbiroli04 : $W_{\sigma}(k)=\,\frac{1}{N_{c}}\,\sum_{\mu\nu}e^{-\imath k\mu}\,W_{\sigma}(\|\mu-\nu\|)\,e^{\imath k\nu}$ (37) Smaller the neglected Fourier coefficients are (for $\|\mu-\nu\|>\sqrt{N_{c}}$), compared to the cluster $W_{\sigma}(\|\mu-\nu\|)$, more the $k$-dependent quantity $W_{\sigma}(k)$ is well approximated. In the following we present two possible cluster-quantities that can be adopted to construct the $k$-dependent self-energy, showing in which cases they can be considered good local quantities. ### III.1 Self-energy $\Sigma$-periodization The cluster self-energy (eq. 18) is a natural candidate. It is convenient, for the discussions in the following sections, to recast formula (37) in terms of the cluster eigenvalues (eq. 29): $\hat{\Sigma}_{\sigma}(k)=\sum_{X}\hat{\Sigma}_{X\sigma}\,\gamma_{X}(k)$ (38) where $\hat{\Sigma}_{\sigma}(k)$ is a 2 $\times$ 2 matrix containing normal and anomalous components: $\displaystyle\hat{\Sigma}_{\sigma}(k)$ $\displaystyle=$ $\displaystyle\left(\begin{array}[]{cc}\Sigma^{nor}_{k\uparrow}(\omega)&\Sigma_{ano}(\omega)\\\ \Sigma_{ano}(\omega)&-\Sigma^{nor}_{k\downarrow}(-\omega)\end{array}\right)$ (41) We notice that with this formula the cluster eigenvalues $\hat{\Sigma}_{X}$ are directly related to the corner points in first quadrant of the Brillouin Zone $X=$$(0,0)$, $(0,\pi)$, $(\pi,0)$ and $(\pi,\pi)$ (and we justify the notation introduced in eq. 29). $\gamma_{X}(k)$ are positive functions, such $\sum_{X}\gamma_{X}(k)=1\,\forall k\cite[cite]{\@@bibref{Authors Phrase1YearPhrase2}{tudor,tudor06}{\@@citephrase{(}}{\@@citephrase{)}}}$: $\displaystyle\gamma_{00}$ $\displaystyle=$ $\displaystyle\,\frac{1}{4}\left(1+\cos k_{x}+\cos k_{y}+\cos k_{x}\cos k_{y}\right)$ $\displaystyle\gamma_{\pi\pi}$ $\displaystyle=$ $\displaystyle\,\frac{1}{4}\left(1-\cos k_{x}-\cos k_{y}+\cos k_{x}\cos k_{y}\right)$ $\displaystyle\gamma_{0\pi}$ $\displaystyle=$ $\displaystyle\,\frac{1}{4}\left(1+\cos k_{x}-\cos k_{y}-\cos k_{x}\cos k_{y}\right)$ $\displaystyle\gamma_{\pi 0}$ $\displaystyle=$ $\displaystyle\,\frac{1}{4}\left(1-\cos k_{x}+\cos k_{y}-\cos k_{x}\cos k_{y}\right)$ We remark that, by construction, with this procedure we assume that the system is a simple Fermi liquid. We have in fact shown in Fig. 2 that the normal components of the eigenvalues of the cluster self-energy have Fermi-liquid behavior (Im$\Sigma_{X}\rightarrow 0$ for $\omega\rightarrow 0$). Their simple linear combination extends this property to all the $k$-space. The anomalous component of the lattice self-energy $\Sigma_{ano}(k)$ turns out to have a d-wave shape $\Sigma_{ano}(k)=2\Sigma_{ano}\,\left(\cos k_{x}-\cos k_{y}\right)$ (43) in agreement with the symmetry of the superconductive gap measured in experiments on cupratesdamascelli ; campuzano . ### III.2 Cumulant $\mathcal{M}$-periodization In a normal state study of the two-dimensional Hubbard Model of ref.tudor ; tudor06 , it has been shown that a more suitable local quantity to describe the Mott transition is the two point irreducible cumulant $\mathcal{M}$, which arises from the atomic limit by perturbatively expanding the hopping term $t$ in Hamiltonian (1). It is simply related to the normal-component lattice self- energy $\Sigma^{nor}_{k}$ by: $\mathcal{M}^{nor}_{k}(\omega)=\frac{1}{\imath\omega+\mu-\Sigma^{nor}_{k}}$ (44) In the cluster-impurity we have $2\,N_{c}\times N_{c}$ cumulant relations (2 is for the spin degeneracy), conveniently represented by a $N_{c}\times N_{c}$ cumulant-cluster-matrix $\hat{\mathcal{M}}_{\sigma c}$: $\displaystyle\mathbf{\hat{\mathcal{M}}^{nor}_{\sigma c}}(\omega)$ $\displaystyle=$ $\displaystyle\left[(\imath\omega+\mu)\mathbf{1}-\mathbf{\hat{\Sigma}^{nor}_{\sigma c}}\right]^{-1}$ (45) where $\mathbf{1}$ is the $N_{c}\times N_{c}$ identity matrix. The eigenvalue of the cumulant matrices are straightforwardly related to the eigenvalues of the self-energy matrices (eq. 29): $\displaystyle\mathcal{M}^{nor}_{X}=\,(\,\imath\omega+\mu-\Sigma^{nor}_{X}\,)^{-1}$ (46) where the notation is again $X=$$(0,0)$, $(0,\pi)$, $(\pi,0)$ and $(\pi,\pi)$. As in the eq. (38), we obtain the lattice cumulant $\mathcal{M}^{nor}_{k}$ by periodizing the eigenvalues of the cluster cumulants: $\mathcal{M}^{nor}_{k}=\sum_{X}\,\mathcal{M}^{nor}_{X}\,\gamma_{X}(k)$ (47) Inverting eq. (44), we finally obtained the normal-component lattice self- energy $\Sigma^{nor}_{k}$ in the $\mathcal{M}$-periodization. ### III.3 Nodal and Antinodal dichotomy: $\Sigma$-vs $\mathcal{M}$-periodization Figure 7: (Color online). We compare the nodal and antinodal quasiparticle peaks obtained from the normal component of the superconductive ED-CDMFT- solution at zero temperature and the QMC-CDMFT normal state solution at much higher temperature ($1/T=75$). The functions on the real axis with QMC are obtained with the maximum entropy method. To display the ED-CDMFT curves the slightly $\omega$-dependent broadening $\eta(\omega)=0.075+\omega^{2}$ was used (see discussion in appendix A). It is not trivial to decide whether it is better to periodize the self-energy $\Sigma$ or the cumulant $\mathcal{M}$. The choice could be strongly dependent on the physical properties of the system, which are not a priori well known. And it is not trivial either understanding a priori to which extend the two approximated schemes could be able to describe such properties. In this case, we rely on experimental results on cuprate materials, like e.g. the already mentioned ARPESdamascelli ; campuzano , to fix a physically reasonable starting hypothesis. It is a well established experimental fact that approaching the Mott insulator the normal state Fermi surface (measured at temperatures above TC) breaks up, displaying well defined quasiparticles in the regions (nodes) close to the centers of the quadrants of the Brillouin Zone $k\sim(\pm\frac{\pi}{2},\pm\frac{\pi}{2})$, while quasiparticle disappear in the strong scattering regions (anti-nodes) close to the corners of the quadrants $k\sim(0,\pm\pi)$ and $(\pm\pi,0)$. Besides photo-emission experiments, measures of transport properties show anomalous (non Fermi liquid) power-low exponents in the temperature-dependence, especially in the under-doped regime of cuprate materials. In this case, a series of phenomenological approaches, which try to explain these experimental observations in the framework of Boltzmann theoryPSK ; Millis1998 ; Zheleznyak1998 ; Hlubina1995 , have been based on dividing the momentum space in regions of high quasiparticle scattering rate, hot spots around the antinodal points, and regions of low quasiparticle scattering rate, cold spots around the nodal point, where quasiparticles have a much longer lifetime. The simple idea underlying this choice is that the nodal region behaves as a standard Fermi liquid, while the system in the antinodal region is like an insulating state. This nodal/antinodal dichotomy fits into the CDMFT frame as the natural path taken by the system to approach the doping-driven Mott transition marce05 ; bumsoo06 ; tudor ; tudor06 . This is a general property independent of the periodizing scheme adopted. Some important differences however hold between the $\Sigma$ and the cumulant $\mathcal{M}$ periodizations, which we will illustrate in the following. In Fig. 7 we show for example the spectral functions $A(k,\omega)=\,-\frac{1}{\pi}\hbox{Im}G^{nor}_{k}(\omega)$ in the nodal $k\sim(\frac{\pi}{2},\frac{\pi}{2})$ (black continuous line) and antinodal $k\sim(0,\pi)$ (red dashed line) points of momentum space, obtained via $\mathcal{M}$-periodization for the case $U/t=12$, $t^{\prime}=-0.3t$ and $5\%$ doping. We confront our zero temperature ED-CDMFT result with a QMC- CDMFTbpk ; olivier-note at much higher temperature ($T=1/75t$), where the system is in the normal state. In the ED case we have extracted the normal part of the superconducting solution (setting $\Sigma_{ano}=0$ in eq. 36), interpreting it as a low-temperature normal state parent of the high- temperature QMC-solution. This is far from being a trivial statement, as the normal component of a superconducting solution is not generally a normal state solution. Our aim in this picture is however to present a qualitative comparison between two very different impurity-solving methods in two very different regimes to show the generality and solidity of the nodal/antinodal dichotomy concept. This figure serves also to compare results of maxent analytic continuation of Fye-Hirsch QMC scheme with the ED results at low temperatures. The qualitative agreement is reasonably good, supporting the observation we made above on experimental results and the dichotomy nature of the physical properties in this system. We observe in particular in Fig. 7 that the antinodal quasiparticle peak is very broad in the high-temperature QMC solution, denoting a short lifetime of quasiparticles, while it is sharper in the low-temperature ED solution, much more than the narrowing due to the different temperature (we have scaled the heights of the peaks taking the high of the nodal quasiparticle peak as reference to fix the scale between the ED and QMC curves). This goes in the direction of ARPES experimental observations, which show a sharpening of the antinodal quasiparticle in going from the normal to the superconductive stateding by decreasing temperature. We finally point out that with the $\mathcal{M}$-periodization used in Fig. 7, in the antinodal point the quasiparticle peak shifts to negative energies, opening a pseudogap. This is also in line with the above mentioned ARPES observations. On the contrary, by using the $\Sigma$-periodization the antinodal peak remains always (for all dopings) at the Fermi level, but with reduced spectral weight compared to the nodal pointmarce05 . All these observations, either from experimental facts and from the output of our method, point towards a dichotomy of the nodal and antinodal regions of momentum space, which show different coherence energies. The nodal region has the highest coherence scale and sharp quasiparticles. In the antinodal region a pseudogap opens in the spectrum, and quasiparticles, if present at the gap- edge, are more broad and incoherent. It is therefore natural to assume that the Fermi-liquid nodal region is better portrayed by the $\Sigma$-periodization (which as we mentioned above describes a Fermi liquid by construction), while the antinodal insulating region is better described by the $\mathcal{M}$-periodization (as shown in Fig. 7). Figure 8: (Color online). Top panel: reconstructed local density of states $-\frac{1}{\pi}\sum_{k}G^{nor}_{k}(\omega)$ for the normal component ($\Sigma_{ano}=0$) of the Green’s function at small doping ($\delta=0.05$). $G^{nor}_{k}(\omega)$ has been obtained by either the cumulant $\mathcal{M}$-periodization (red point-dash line labeled $M$-scheme) and by the $\Sigma$-periodization (green dash line labeled $\Sigma$-scheme), and confronted with the local density of states of the cluster-impurity solution (black continuous line). Middle panel: close-up at the Fermi level. Bottom panel: the local density of states of full superconducting solution ($\Sigma_{ano}\neq 0$) selects in the $k$-summation only the nodal points. A low-energy confront with the cluster-impurity solution shows a different matching of the cumulant and self-energy periodizations with the cluster- impurity result. The display on the real axis of the Green’s function has been obtained by introducing a $\omega$-dependent broadening i$\eta(\omega)$ (see appendix A for details). Taking this last assumption as starting point, we compare the results of the $\Sigma$ and $\mathcal{M}$-periodization, stressing virtues and defaults in the frame of the physical observations carried over above. A stringent test is given by reconstructing the local density of states $N(\omega)=\,-\frac{1}{\pi}\sum_{k}\hbox{Im}G_{k}(\omega)_{11}$ (where $G_{k}(\omega)$ is obtained by eq. 36), which can be compared with the local Green’s function $G_{imp}=(\hat{G}_{c})_{11}$, obtained directly in the cluster-impurity solution. This test is presented in Fig. 8. First, it is once again instructive to separately study the normal component of the superconducting state by setting $\Sigma_{ano}=0$ in eq. (36). We focus the attention on a small-doping case $\delta=0.05$ close to the Mott transition point. The top panel displays a full energy range which includes the lower and upper Hubbard bands. The cluster-impurity density of states $N(\omega)$ is represented by the continuous black line, the $\mathcal{M}$-periodization result by a red dot-dash line and the $\Sigma$-periodization result by a green dash line. Already at small doping, in the region labeled (A), the formation of a Mott gap is visible in the cluster-solution. The $\mathcal{M}$-periodization is capable to describe this part of the spectrum quite well (matching the continuous black line of the impurity result). On the contrary, the $\Sigma$-periodization creates artificial states in the Mott gap. This is true also at low energy (region labeled B). A close up on the Fermi level ($-t<\omega<t$) is displayed in the middle panel (with again $\Sigma_{ano}=0$). The $\mathcal{M}$-periodization reproduces the formation of a low-energy pseudogap in the normal component of the spectra, as already remarked in previous normal-state studiestudor ; tudor06 , while, once again, the $\Sigma$-periodization introduces artificial states that fill the pseudogap. This simple test therefore evidences the failure of the $\Sigma$-periodization in well describing the high and low energy normal- component of Green’s function, in the regions of momentum space where a pseudogap is present, like in the antinodal $k$-point of Fig. 7. The scenario is different if we look at the low-energy superconducting density of states $N(\omega)$ by restoring $\Sigma_{ano}\neq 0$ in eq. (36), as presented in the bottom panel of Fig. 8. Here it is the $\Sigma$-periodization line (green dash line) which better portrays the cluster spectrum (black continuous line). The $\mathcal{M}$-periodization this time introduces spurious states close to the Fermi level ($\omega=0$). This is not contradictory with respects to the result presented above on the normal component of the system (top and middle panels). By adding the d-wave superconducting gap in fact, we selects at low energies ($\omega\rightarrow 0$) only the region of momentum space close to the nodal points, where the gap is zero. These points dominate in fact the sum $-\frac{1}{\pi}\sum_{k}\hbox{Im}G_{k}(\omega)_{11}$. $N(\omega)$ for $\omega\to 0$ is therefore a direct probe of the nodal point behavior only (while without superconducting gap, i.e. $\Sigma_{ano}=0$, all $k$-points eventually contribute in the summation at low energies). We clarify this last statement. Let’s assume as starting point that at the nodes quasiparticles are well defined on the full range of doping. We can extract from eq. (36) the low energy ($\omega_{n}\rightarrow 0$) Green’s function: $\displaystyle\hbox{Re}\Sigma_{k}(\omega_{n})$ $\displaystyle\sim$ $\displaystyle\hbox{Re}\Sigma_{k}(0)$ $\displaystyle\hbox{Im}\Sigma_{k}(\omega_{n})$ $\displaystyle\sim$ $\displaystyle(1-Z^{-1}_{k})\omega_{n}$ and $\displaystyle G(k,\omega_{n})$ $\displaystyle\sim$ $\displaystyle\frac{\frac{Z_{k}}{2}(1-\tilde{\xi}_{k}/E_{k})}{\imath\omega_{n}+E_{k}}+\frac{\frac{Z_{k}}{2}(1+\tilde{\xi}_{k}/E_{k})}{\imath\omega_{n}-E_{k}}$ (48) For convenience’s sake, we have enlighten the quasiparticle dispersion $\displaystyle\tilde{\xi}_{k}$ $\displaystyle=$ $\displaystyle Z_{k}\,\left[\xi_{k}+\hbox{Re}\Sigma_{k}(0)\right]$ $\displaystyle E_{k}$ $\displaystyle=$ $\displaystyle\,\sqrt{\|\tilde{\xi}_{k}\|^{2}+\|Z_{k}\,\Sigma_{ano}(k,0)\|^{2}}$ (49) The nodal point is the only one gapless at low energies and the quasiparticle spectrum can be linearized (eq. 49): $\displaystyle E_{k_{nod}}$ $\displaystyle=$ $\displaystyle\,\sqrt{\|\tilde{\xi}_{k_{nod}}\|^{2}+\|Z_{k_{nod}}\,\Sigma_{ano}(k_{nod})\|^{2}}$ (50) $\displaystyle=$ $\displaystyle\,\sqrt{v_{nod}^{2}k^{2}_{\bot}+v_{ano}^{2}k^{2}_{\\|}}$ $v_{nod}=\|\nabla_{k}\tilde{\xi}_{k}\|$ is the quasiparticle Fermi velocity perpendicular to the Fermi surface, $v_{ano}=Z_{nod}\|\nabla_{k}\Sigma_{ano}(k)\|$ is parallel to the Fermi surface. After analytic continuation $\omega_{n}\rightarrow\omega+\imath\delta$, it is now easy to calculate the low energy behavior of the one-particle density $N(\omega)=\frac{1}{\pi}\,\sum_{k}\,\hbox{Im}G_{k}(\omega)$ $N(\omega)\sim\frac{1}{\pi}\sum_{j}^{\tiny\hbox{nodes}}\,\frac{Z_{nod_{j}}}{v_{nod_{j}}v_{ano_{j}}}\,\omega$ (51) i.e. $N(\omega)$ is linear in frequency close to the Fermi energy and the slope is uniquely determined by the quasiparticle nodal velocity $v_{nod}$, by the nodal derivative of the superconducting gap $v_{ano}$ and by the nodal quasiparticle residuum $Z_{k_{nod}}$. In other words, the low-energy density of states in the d-wave superconductor is a direct measure of the spectra at the nodes. This result explains the low-energy spectrum of the cluster density of states (black-continuous line in Fig. 8), which is roughly linear for $\omega\to 0$, in the limit of the energy-resolution given by ED impurity solver. The fact that here the $\Sigma$-periodization better portrays the cluster result is consistent with the hypothesis of Fermi-liquid behaviour at the nodal points of momentum space, well described by the self-energy. The results of the normal component of the system suggest instead that in other regions of momentum space (at least close to the Mott transition point), a pseudogap opens in the spectrum and quasiparticles die in the anti-nodes. In the latter regions a cumulant $\mathcal{M}$-periodization is more appropriate than the $\Sigma$-periodization. ## IV Properties of the nodal and antinodal points In this section we follow the observations presented above on the nodal/antinodal dichotomy and study the physical properties of the system in the nodal and antinodal points, completing in detail the work presented in ref.marce08 . In order to be able to make contact with the experiments, we need to extract the fully momentum dependent Green’s function $\hat{G}_{k}(\omega)$ (see equation 36), which can be generally related to the response functions. We employ therefore a periodizing procedure to extract the $k$-dependent normal and anomalous components $\Sigma_{k}(\omega)$ and $\Sigma_{ano}(k,\omega)$ of the $k$-dependent self-energy. As discussed above, it is reasonable to start by assuming a d-wave shape of the superconducting gap, which is naturally obtained in our scheme by periodizing the anomalous component of the cluster self-energy via eq. (43). At the nodes, we periodize also the normal component of the self-energy via eq. (38). This guarantees in particular Fermi liquid properties in the nodal region of momentum space. At the anti-nodes instead the insulating properties of the normal component of the system (like the formation of a pseudogap at low energy and of the Mott gap at higher energies, see Fig. 8) are better portrayed by periodizing the cumulant $\mathcal{M}^{nor}$, according to eq. (47). Figure 9: (Color online). Panel A: nodal $k_{nod}$ and antinodal $k_{anod}$ positions of the quasiparticle peaks in the first quadrant of the Brillouin Zone as a function of doping $\delta$. Panel B: Nodal peaks are found always at the Fermi level and decrease with doping $\delta$ in approaching the insulator. Panel C: At high doping ($\delta>\delta_{c}\sim 0.08$) antinodal peaks are also found at the Fermi level in the normal component of the system (set $\Sigma_{ano}=0$ in eq. 36). For ($\delta<\delta_{p}\sim 0.06$) however they shifts to negative energy opening a pseudogap. Panel D: The actual antinodal spectra ($\Sigma_{ano}\neq 0$ in eq. 36) show to become asymmetric for $\delta>\delta_{c}$, when the pseudogap starts opening in the normal component (see panel C, and bottom right panel of Fig. 6 ). By adopting this procedure we can first obtain quasi-particle spectra in the nodal and antinodal points of momentum space, as it is shown in Fig. 9. In order to define the $k$ vectors in the nodes and in the anti-nodes, we follow an operative definition, similar to the one used in the ARPES experiments of ref.tanaka06 . The nodal point $k_{nod}$ [antinodal point $k_{anod}$] is chosen as the one having the sharpest quasiparticle peak in moving on the path $(0,0)\to(\pi,\pi)$ [$(0,\pi)\to(\pi,\pi)$] of momentum space. The vectors $k_{nod}$ and $k_{anod}$ as a function of doping $\delta$ are shown in panel A of Fig. 9. We notice that $k_{anod}$ is a monotonic decreasing function of the $\delta$. This is intuitively expected in a standard Fermi liquid, where the approach to the Mott transition at $\delta=0$ is accompanied by an increase of the volume enclosed by the Fermi surface in momentum space, which is proportional to the density $n=1-\delta$ of the system (Luttinger theoremLuttinger60 ). The $k_{nod}$ vector on the contrary displays a non- monotonic behavior, showing to decrease by reducing doping in correspondence of the special doping $\delta_{c}\sim 0.08$, which appears in our study as a critical point marking a change in the physical properties of the system. In the following section we will show that we can relate this behavior of the nodal and antinodal $k$ points to a topology change in the Fermi surface. In the remaining panels of Fig. 9, we show the spectral function Im$[\hat{G}_{k}(\omega)]_{11}$ (from eq. 36). As previously explained (see also appendix A), within the ED-CDMFT method it is possible to display Green’s functions on the real frequency axis $\omega$ by expressing them in a pole expansion, displayed by adding in the denominator a small imaginary part i$\eta$ (here we adopted $\eta=0.03t$). At the nodal point (panel B) the d-wave superconducting gap is zero, and a Fermi liquid quasi-particle peak is always found at the Fermi level ($\omega=0$) for different doping $\delta$. The approach to the Mott insulator ($\delta\to 0$) is marked by a progressive reduction of the quasi-particle peaks. This behavior is reminiscent of the Mott transition described in the standard infinite dimensional Hubbard Model. At the antinodal point the superconducting gap is maximal, a gap in the spectra is therefore expected. Before looking at the full antinodal spectrum, however, it is instructive to look at the contribution coming from the normal component, which can provide information on the physical properties of the liquid underlying the superconducting state. The normal-component spectra can be simply obtained in our scheme by zeroing the anomalous component of the self-energy $\Sigma_{ano}(k,\omega)$ in eq. 36. These spectra are displayed in panel C of Fig. 9. At doping $\delta>\delta_{c}$, a Fermi liquid quasi- particle peak is also found at the Fermi level ($\omega=0$). In this region of the phase diagram, the normal properties of the system are therefore Fermi- liquid-like (and also $k_{nod}$ and $k_{anod}$ are monotonically decreasing with $\delta$, as described above). In correspondence of the critical doping $\delta_{c}$ however, a pseudogap opens and a quasi-particle peak is found at the gap edge at negative energy ($\omega<0$). The pseudogap increases in approaching the Mott transition ($\delta\to 0$), while, differently from the nodal point, the peaks show a roughly constant height. This behavior at the anti-nodes of the normal component of this superconducting solution can be smoothly connected to results previously obtained in CDMFT studies of the normal statetudor ; tudor06 . For $\delta<\delta_{c}$ therefore, the normal component of the system is not a Fermi liquid in the strict sense, at least in the region of momentum space close to the antinodal points (but a behavior unusual for a Fermi liquid is also detected by the decreasing value of $k_{nod}$ in panel A). This behavior appears also in the total antinodal spectra (upon restoring the superconducting gap $\Sigma_{ano}\neq 0$ in eq. 36), which we show in panel D. At doping $\delta>\delta_{c}$, the quasi- particle peaks present in the normal component are parted into two bands by the opening of a superconducting gap, resulting in the typical BCS symmetric spectra. For $\delta<\delta_{c}$, however, the pseudogap already present in the normal component super-impose to the superconducting gap, resulting in asymmetric spectra. This antinodal spectra nicely explain Fig. 6, where the local density of states $N(\omega)$, directly obtained from the cluster- impurity solution, is displayed. The appearance of the asymmetry for $\delta<\delta_{c}$ is therefore interpreted by our $k$-momentum analysis as the appearance of the pseudogap phase, which marks a departure from a Fermi liquid based BCS superconductor, once again at the critical doping $\delta_{c}$. And these observations must be directly linked to experimental spectra, either in the anti-nodes of momentum space, obtained for example with angle resolved photo-emissiondamascelli ; campuzano , and locally in real space, with for example scanning tunneling spectroscopydavis05 . Figure 10: (Color online). Quasiparticle residuum at the nodal point $Z_{nod}$ and antinodal points $Z_{anod}$ as a function of the doping $\delta$, evaluated as area of the quasiparticles peaks shown in Fig. 9 (as a crosscheck, $Z_{nod}$ is also evaluated as the slope of the imaginary part of the self-energy on the Matsubara axis, green cross). While $Z_{nod}$ monotonically decrease with doping, as in the standard picture of the Mott transition, the antinodal $Z_{anod}$ shows to stay constant upon opening of the pseudogap ($\delta>\delta_{c}\sim 0.08$). For comparison’s sake, we display the nodal $Z_{nod}$ obtained by CTQMChaule-ctqmc on the two dimensional Hubbard Model with similar parameters ($U=12t,t^{\prime}=0.0,\beta=200$). These quasiparticle residua should be compared with their cluster counter-parts in Fig. 4. In order to characterize the Mott transition, it is useful to extract the quasi-particle residuum, which is defined as $Z_{k}=\left(1-\frac{\partial\hbox{Re}\Sigma^{nor}(k,\omega)}{\partial\omega}\right)^{-1}_{\omega\to 0,k=k_{F}}$ (52) with $k_{F}=k_{nod}$ or $k_{anod}$ in our case. This quantity corresponds to the area of the quasiparticle peaks (e.g. panel A of Fig. 9), it is unity in the non-interacting case, less than unity in an interacting Fermi liquid. In the standard description of the Mott transition (i.e. the infinite dimensional Hubbard Model) $Z\to 0$ linearly as $\delta\to 0$ (see ref. revmodmft ). It is interesting therefore to observe how this quantity behave in the nodal and antinodal $k$-points by varying doping $\delta$ , as we show in Fig. 10. We can give in this way a momentum space interpretation of the corresponding cluster $Z_{X}$ that we have discussed in Fig. 4. As we already said, at the nodes well defined quasiparticle peaks are observed at every doping $\delta$ at the Fermi level. The residuum $Z_{nod}$ is therefore well defined according to the expression given above (which is strictly valid at $\omega=0$), as in a typical Fermi liquid. This is confirmed by the good numerical accord between the blue circles (calculated by integrating the area of the peaks, which are displayed by introducing the artificial broadening parameter $\eta=0.03t$) and the green crosses (calculated more precise within our method by using eq. 52 on the Matsubara axis). For comparison’s sake, we present also the nodal quasi-particle residuum extracted in a CDMFT-study implemented with a different impurity solver, the CTQMChaule-ctqmc , in the two dimensional Hubbard model with Hamiltonian parameters $U=12t$ and $t^{\prime}=0t$. The comparison is only qualitative and it is aimed to get insight into the physical trends. We indeed observe a monotonically decreasing $Z_{nod}$ as a function of the doping $\delta$, similarly to the standard infinite dimensional Hubbard Model. We cannot however state within our numerical resolution if $Z_{nod}\to 0$ exactly at the Mott point $\delta=0$ (as it seems also to suggest the CTQMC result), or rather it extrapolates to a finite but very small ($Z_{nod}<0.02t$) value. In the antinodal point more attention has to be paid in defining a quasi-particle residuum $Z_{anod}$. As stressed above, for $\delta>\delta_{c}\sim 0.08$ quasiparticle peaks are present at the Fermi level (panel C of Fig. 9), and $Z_{anod}$ can be well defined by eq. 52. For $\delta<\delta_{c}\sim 0.08$ however a pseudogap opens. Even if a peak can be identified at the gap-edge, it is not strictly speaking a Landau-Fermi liquid quasiparticle, as the imaginary part of the self-energy is non-zero (even if small), i.e. the quasiparticle has a finite lifetime. Formula 52 cannot be directly employed. However we can still calculate the area of the peak, and display its behavior as a function of the doping $\delta$. We find in our result that for $\delta<\delta_{c}$, once the pseudogap opens and the peaks move to negative frequency, the weight $Z_{anod}$ stays constant up to the Mott transition point. The behavior of $Z_{nod}$ and $Z_{anod}$ here presented has to be connected with the effective cluster correspondents $Z_{X}$ (Fig. 4). According to eq. LABEL:gamma-k, these are interpreted in momentum space as the quasiparticle residua in the corner points of the first quadrant of the Brillouin Zone (therefore far from the Fermi surface), while $Z_{nod}$ and $Z_{anod}$, which are instead calculated on the Fermi surface, have a real physical meaning. In spite of this however, the cluster $Z$s already embody the physical properties (a $Z$ going to zero and another non decreasing for $\delta\to 0$) characteristic of the Mott transition in this two-dimensional system. We stress that the description of the Mott transition we find in this study of the two dimensional Hubbard Model is very different from the standard Mott transition picture in infinite dimensionrevmodmft . In our system different regions of momentum space behave very differently in approaching the transition point. In this way we can go from a Fermi-liquid-based superconductor (realized for $\delta>\delta_{c}$) into the Mott insulator (at $\delta=0$) by passing through a phase $0<\delta<\delta_{p}$, where the system is at the same time insulating in the antinodal region and Fermi liquid in the nodal region of momentum space. This latter appears to approach the Mott point in the standard (infinite dimensional Hubbard Model) way, with a quasiparticle residuum $Z_{nod}\to 0$ (at least within our numerical precision). In the antinodal region instead the quasiparticle peak (which underlies the superconducting gap) stops reducing at $\delta=\delta_{c}$ and shifts to negative energies opening a pseudogap. This behavior is reminiscent of the orbital selective Mott transition, found e.g. in two band Hubbard-like models deleo-2008 ; ferrero-2005-72 ; medici-2005-72 , where the spectral weight is not transferred from the low energy ($\omega=0$) to the Hubbard bands (located at a energy scale of order $\sim U$), as in the standard Mott transition, but rather onto a smaller energy scale of the order of an exchange coupling $J\sim 1/U^{2}$, inside the Mott gap. In spite our model is a one band one, different regions of momentum space appear to behave as different bands. By decreasing doping $\delta$, a first ”orbital selective Mott transition” takes place at $\delta_{c}$ in the antinodal regions, and a full Mott transition takes finally place at $\delta=0$. Figure 11: (Color online). Top: Nodal velocity as a function of doping measured from the quasiparticle dispersion of different materials (figure taken from ref.Shen-Nature03 ). To be confronted with the nodal component of the velocity $v_{nod}$ calculated in our work and displayed in the bottom. Bottom: Nodal anomalous velocity $v_{ano}=Z\Sigma_{ano}(k_{nod})$ (tangent component of the nodal velocity) and nodal quasiparticle velocity $v_{nod}=Z\|\nabla_{k}\xi_{k}\|$ (component of the nodal velocity perpendicular to the Fermi surface) as a function of doping $\delta$ (units are $a_{o}t$, where $a_{o}$ is the square lattice spacing). The trends are compared with CTQMChaule-ctqmc calculations on the two dimensional Hubbard Model with similar parameters ($U=12t,t^{\prime}=0.0,\beta=200$). Useful information on the nodal point can be extract by performing a low energy expansion of the Green’s function (see eq. 48 and 50), taking advantage of the Fermi liquid properties of the nodal point. In particular the nodal velocity has two components, one coming from the normal part $v_{nod}=\|\nabla_{k}\tilde{\xi}_{k}\|$ parallel to the Fermi surface, and the other related to the superconducting gap $v_{nod}=Z_{nod}\|\nabla_{k}\Sigma_{ano}(k)\|$, perpendicular to the Fermi surface. $v_{nod}$ can be experimentally extracted, e.g. from the ARPES quasiparticle dispersion at the nodeShen-Nature03 , while $v_{ano}$ can be determined e.g. as the slope of the superconducting gap at the nodetanaka06 . In Fig. 11 $v_{nod}$ and $v_{ano}$ are displayed as a function of doping $\delta$. For comparison’s sake, in the top panel we show experimental nodal velocity extracted from different materials (the figure has been taken from the supplementary material of ref.Shen-Nature03 ). In the bottom panel we show our result, and also insert for a qualitative comparison the CTQMC-CDMFT result of ref.haule-ctqmc . $v_{nod}$ shows to be greater than $v_{ano}$, in agreement with experimental observation (see e.g. ref.damascelli ). In ED- CDMFT it slightly oscillates around a constant value from the over-doped to the under-doped ($\delta<\delta_{c}$) side of the phase diagram, while in the CTQMC case it is slightly decreasing with decreasing doping. This behavior is in good qualitative agreement with experimental results reported in the top panel (a quantitative comparison would roughly held, for a lattice spacing $a_{o}\sim 4\AA$ and a $t\sim 0.5$ eV, $v_{nod}\sim 1$ eV$\AA$, which is of the order of magnitude of experiments). Remarkably, either with ED and with CTQMC, $v_{ano}$ displays a dome-like shape, with a maximum around optimal doping $\delta_{p}<\delta<\delta_{c}$ (notice that in the ED case $t^{\prime}=-0.3t$ $\delta_{c}\sim 0.08$ while in the CTQMC case $t^{\prime}=0$ $\delta_{c}\sim 0.12$). This is an important result: as shown in formula 49 and discussed in ref.marce08 , the anomalous velocity $v_{ano}=\sqrt{2}Z_{nod}\Sigma_{ano}(\omega=0)\sin k_{nod}$ can be interpreted as a direct measure of the superconducting gap in the nodal region, which reveals non-monotonic, in agreement with some recent experimental spectroscopy resultstanaka06 ; tacon06 . The agreement of the trend of $v_{ano}$ between the ED and CTQMC solutions shows that this result is solid from the theoretical side too. We will come back to discuss the nodal and antinodal gap more in detail at the end of this section. At the nodal point, the combination of the quasiparticle residuum $Z_{nod}$ and nodal velocities can give further information, which can be confronted with experimental observable quantities and which can further support the physical description drawn from our CDMFT result. Basing on the Fermi liquid assumption at the nodes, in eq. (51) we have for example extracted the low energy ($\omega\to 0$) behavior of the local density of states $N(\omega)\sim\frac{Z_{nod}}{v_{nod}v_{ano}}\,\omega$. This value is displayed as a function of doping $\delta$ in Fig. 12, comparing once again the ED-CDMFT results of this work (with $t^{\prime}=-0.3t$) with the CTQMC-CDMFT results of ref.haule-ctqmc (with $t^{\prime}=0$). Once again it is the trend we want to compare rather than the quantitative values. Starting from the over-doped side, the slope of $N(\omega)$ is decreasing monotonically by decreasing doping, until showing a slight up-turn close to the Mott transition (observable both in the ED and CTQMC cases). While the linearity of $N(\omega)$ it is well established in scanning tunneling experimentsdavis05 , the behavior of the slope as a function of doping is at the moment very difficult to extract (as it is not possible to obtain absolute values for different densities). Our result (in particular the up-turn tendency at small doping) has to be therefore considered a theoretical prediction. A further ratio of experimental relevance can be connected at first order to the low energy ($\omega\to 0$) linear behavior of the $B_{2g}$ Raman response function $\chi_{B2g}\sim\frac{Z_{nod}^{2}}{v_{nod}v_{ano}}\,\omega$ (see ref.tacon06 ) and the low temperature ($T\to 0$) behavior of the superfluid stiffness $\rho_{s}(T)-\rho_{s}(0)\sim\frac{Z_{nod}^{2}}{v_{nod}v_{ano}}\,\omega$, which can be extracted from measures of the penetration depth Bonn96 ; Panagopoulos98 . Our results (together with the CTQMC results of Ref.haule- ctqmc ) are shown on the bottom panel of Fig. 13 and compared with the aforementioned Raman and penetration depth data presented in the top panel (the figure has been taken from ref.tacon06 ). The remarkable feature, found in experiments and supported in our calculation, is the constant value displayed by this ratio in the under-doped region ($\delta<\delta_{c}$). While in the experimental results a sum rule is assumed, in order to being able to compare measures from different doping/samples, our theoretical results are derived from a bare strongly correlated electron model, where other kind of assumptions and approximations (as explained in the previous sections) are implied. The convergence of experimental and theoretical results therefore strongly supports these findings, presenting them as distinctive feature of the cuprate superconductor nodal dispersion. Figure 12: (Color online). The slope of the low energy local density of states $N(\omega)\sim\frac{Z_{nod}}{v_{\Delta}v_{nod}}\omega$ is displayed as a function of doping $\delta$. Trends are compared with CTQMChaule-ctqmc calculations on the two dimensional Hubbard Model with similar parameters ($U=12t,t^{\prime}=0.0,\beta=200$). Figure 13: (Color online). Top: the linear coefficient of the low energy Raman response $\chi_{B2g}(\omega)\sim\alpha\omega$ and the low temperature superfluid stiffness (from penetration depth) $\rho_{s}(T)-\rho_{s}(0)\sim\beta T$, extracted from various experiments (figure taken from ref.tacon06 ). Quantities are normalized at the optimal doping value. Bottom: The linear coefficient of the Raman response and superfluid stiffness (see top panel) as extracted from our calculation $\frac{Z_{nod}^{2}}{v_{\Delta}v_{nod}}$ and displayed as a function of doping $\delta$. Trends are compared with CTQMChaule-ctqmc calculations on the two dimensional Hubbard Model with similar parameters ($U=12t,t^{\prime}=0.0,\beta=200$). The behavior of the spectra presented in Fig. 9 can be directly connected to spectroscopy experiments. In particular, recently a lively debate has risen on the momentum resolved structure of the superconducting gapdeutscher99 ; kyung02 ; yang-2006-73 ; millis06 ; cho06 ; huefner-2008-71 ; valenzuela07 ; aichhorn-2007-99 . We proceed as in ARPES experiments (see e.g. tanaka06 ), and extract the quasiparticle gap in the nodal and antinodal points of momentum space, taking advantage of the periodizing scheme we have introduced. To this purpose, it is convenient to use the low energy ($\omega\to 0$) expansion carried out in eq. 48. In particular we have seen that quasiparticles peaks are always found in our result, even if not in a strict sense (in the pseudogap region $\delta<\delta_{c}$ quasiparticle peaks are located at the gap-edge and have finite lifetime), and therefore we expect the expansion to be reasonably good at small frequency (i.e. $\omega\leq$ the superconducting gap). We can in this case write the total gap $\Delta_{tot}(k,\omega)$ as the quadratic sum of two contributions (see eq. 49): $\Delta^{2}_{tot}(k,\omega)=\Delta^{2}_{nor}(k,\omega)+\Delta^{2}_{sc}(k,\omega)$ (53) where $\Delta_{sc}(k,\omega)=Z_{k}\Sigma_{ano}(k,\omega)$ is the usual d-wave superconducting gap (notice that it is directly connected, except for constant factors, to the anomalous component of the nodal velocity $v_{ano}\sim Z_{k_{nod}}\Sigma_{ano}(\omega=0)$ discussed in Fig. 11), $\Delta_{nor}(k,\omega)=\tilde{\xi}_{k}(\omega)=Z_{k}\,\left[\xi_{k}+\hbox{Re}\Sigma_{k}(\omega)\right]$ is a normal contribution to the gap which can arise only if, for some $k$ and $\omega\to 0$, the normal component of the self-energy $\Sigma_{k}(\omega)$ grows enough so that the band equation $\tilde{\xi}_{k}=0$ cannot be satisfied, i.e. there is not Fermi surface. Now, this does not take place at the nodes, where quasiparticles are found for all dopings (panel B of Fig. 9), $\tilde{\xi}_{k}=0$, the system presents standard Fermi liquid properties, and the total spectral gap coincides with the superconducting gap $\Delta_{tot}\equiv\Delta_{sc}$. (This already suggests that in looking for the “real” superconducting gap of the system one should look at the nodal region gap. Recent theoreticalhaule-ctqmc and experimental Raman spectroscopy studiesguyard-2008 point at this direction.) In the antinodal region, instead, we have observed that a pseudogap opens in the normal component for $\delta<0.08$ (panel C of Fig. 9), and this fact is associated in our calculation with the appearance of lines in $k$-space close to the antinodal region where the self-energy is diverging (see ref.tudor ; tudor06 ). It is not possible therefore to satisfy the equation $\tilde{\xi}_{k}=0$, and a Fermi surface does not exists anymore. The normal contribution $\Delta_{nor}$ kicks in, and determines the properties of the total antinodal gap $\Delta_{tot}$, originating the asymmetric spectra we already described in panel D of Fig. 9. Figure 14: (Color online). Top: Nodal ($B_{1g}$) and antinodal ($B_{2g}$) quasiparticle gap extracted by different spectroscopy experiments (in particular, the figure is taken from the Raman spectroscopy results of ref.tacon06 ) as a function of doping. Bottom: The antinodal gap $\Delta_{tot}$ and the nodal $\Delta_{Sc}$ extracted from the spectra of Fig. 9, as a function of doping $\delta$. $\Delta_{nor}$ is the pseudogap in the normal component extracted in panel C of Fig. 9. We can extract, like in ARPES experiments (e.g. ref.tanaka06 ), the antinodal gap $\Delta_{tot}$ from the spectra of panel D of Fig. 9 by measuring the distance of the quasiparticle peaks at the gap-edge from the Fermi level $\omega=0$. In the same way, from panel C of Fig. 9, we can measure the normal contribution $\Delta_{nor}$, and display them as a function of doping $\delta$. We can compare directly with experimental results, which are shown in the top panel of Fig. 14 (for convenience’s sake we extract the picture from the Raman results of ref.tacon06 , but ARPES points are also displayed), while in the bottom panel we show the results of our calculation. The antinodal gap $\Delta_{tot}$ is a monotonic decreasing function of doping (curve labeled “Anti Nodal B1g” in the top panel of Fig. 14), as it has been known from experiments since a long time (see e.g. ref.damascelli ; campuzano ) and predicted in the most popular theories of high temperature superconductivity ( e.g. resonating valence bond theoryanderson , for a recent general review see e.g.patrickrmp ). In our calculation we show that at small doping an important contribution comes indeed from the normal component $\Delta_{nor}$, which appears at $\delta_{c}$ and it is also monotonic. From the experimental side, the novelty comes from precise measures of the nodal gap, recently obtained by Raman spectroscopytacon06 and ARPEStanaka06 , which show surprisingly that in this region of momentum space the spectral gap is non-monotonic with doping $\delta$, tracking instead the behavior of the critical temperature $T_{c}$ (curve labeled “Nodal B2g” in the top panel of Fig. 14). In our calculation the nodal gap corresponds to the anomalous component of the nodal velocity $v_{ano}$, which we have discussed in Fig. 11, and which shows indeed a behavior strongly similar to these experimental results. We clarify now how it is possible that the nodal component of the gap tracks $T_{c}$ as a function of doping $\delta$, while at the same time the antinodal component is monotonic, by disentangling he superconducting contribution $\Delta_{sc}=\sqrt{\Delta^{2}_{tot}-\Delta_{nor}^{2}}$, which we also display in the bottom panel of Fig. 14. To check the validity of our formula 53, and making connection with $v_{ano}\sim Z_{k_{nod}}\Sigma_{ano}$, which has been evaluated at the nodal point, we also evaluate and display $\Delta_{sc}=Z_{anod}\Sigma_{ano}(k_{anod},\omega=\Delta_{nor})$, finding numerical agreement. This shows indeed that the total antinodal gap $\Delta_{tot}$ has indeed two distinct contributions, displaying opposite trends with dopings. In the under-doped side $\Delta_{nor}$ dominates in the antinodes, and creates a monotonically increasing total gap. In the nodes, instead, $\Delta_{nor}$ is zero and only the superconducting gap is detectable. In our results we connect therefore the experimentally observed two-gap phenomenon with the opening of the pseudogap at the anti-nodes at a finite critical doping $\delta_{c}$. And, according to our point of view, this is an effect arising in a strongly correlated electron system that approaches the doping-driven Mott transition in a two dimensional lattice. The fact that $\Delta_{sc}$ tracks $T_{c}$guyard-2008 is remarkably similar to the standard BCS superconductivity. These results put strong constrains on the theories of cuprate-based superconductivity, which have to consider the presence of these two distinct components in the spectral gap and its interplay with the dome- like shape of the order parameter, the rising of the pseudogap phase and the approaching to the Mott insulator. Our cluster results, interpreted via a periodization procedure, well fit the experimentally observations and give a simple interpretation in terms of a combination of all these effects. ## V A mixed-periodization scheme The discussion presented in the previous subsections is valid close to the nodal and antinodal points of momentum space, where we have shown that periodizing the self-energy or the cumulant its a reasonable approximation. Obtaining information in intermediate region of momentum space, between the nodal and antinodal ones, is beyond the limits of a pure $2\times 2$-plaquette study. In this region in fact the Fermi liquid physical properties of the nodal points have to interlace in some non-trivial way with the insulating- like properties of the antinodal point. A detailed description of this phenomenon can be taken into account by studying bigger cluster (i.e. obtaining a better truncation of the Fourier expansion in eq. 37), at present not possible with ED-CDMFT. Nevertheless we want to keep in this section a low profile, and introduce a first order description of the spectral properties in all momentum space, which could be compared to experimental results (like e.g. ARPES). We want therefore to introduce a periodizing scheme able to describe 1. 1. a Fermi liquid quasiparticle in correspondence of the nodal point 2. 2. the opening of a pseudogap in the antinodal points in the under-doped regime 3. 3. the formation of the Mott gap at high energies in approaching the Mott insulator (see top panel of Fig. 8 region marked A). We can satisfy the first condition by using the self-energy $\Sigma$-periodization. The second and third points are instead obtained by using the cumulant $\mathcal{M}$-periodization, which is able to describe the formation of the antinodal. Figure 15: Left: the shaded region is the patch $\Phi_{k}$ used in the phenomenological Boltzmann approach of ref.PSK in the first quadrant of the Brillouin Zone. Right: the patch $\Phi_{k}$ used in this work (see formula 55). In order to combine all these requests, we take advantage from ideas introduced in phenomenological Fermi-liquid Boltzmann approaches to the normal-state transport properties of cuprate superconducting materialsPSK ; Millis1998 ; Zheleznyak1998 ; Hlubina1995 . We base in particular on the work of ref.PSK , where the division of momentum space in nodal Fermi-liquid-like regions, cold spots, and antinodal insulator-like regions, hot spots, was achieved by projecting the quasiparticle scattering operator $C_{kk^{\prime}}$ on a basis of patches $\\{\phi_{\alpha}\\}$ in momentum space with different temperature scattering-dependencies: $C_{kk^{\prime}}=\,\sum_{\alpha\beta}\,\phi_{\alpha}(k)C_{\alpha\beta}(T)\phi_{\beta}(k)$ (54) In this case, it is possible to solve exactly the Boltzmann equation for the simple case of two patches (one marking the nodal and the other the antinodal region). The shape and the scattering properties of the cold patch (modeled by a small set of parameters) was fixed by obtaining the best fit on few transport quantities (resistivity, Hall coefficient), and a systematic correspondence with others transport quantities (magnetoresistance and termoelectric power) was then obtained. For comparison, the nodal patch $\Phi_{k}$ used in ref.PSK is shown in the left hand side of Fig. 15 (the antinodal patch is simply defined as $1-\Phi_{k}$). In the same spirit, we introduce here a mixed periodization-scheme, by projecting the lattice self-energy $\Sigma_{k}(\omega)$ on a nodal $\Phi(k,\omega)$ and an antinodal $1-\Phi(k,\omega)$ patch in momentum space: $\Sigma_{k}(\omega)=\,\Phi(k,\omega)\Sigma_{k}(\omega)[\hat{\Sigma}]+[1-\Phi(k,\omega)]\Sigma_{k}(\omega)[\hat{\mathcal{M}}]$ where $\Sigma_{k}(\omega)[\hat{\Sigma}]$ is obtained by periodization of the cluster $\Sigma$ and $\Sigma_{k}(\omega)[\hat{\mathcal{M}}]$ by periodization of the cluster cumulant $\mathcal{M}$ (eq. 45). The patch $\Phi(k,\omega)$ separates the $\Sigma$-periodized from the $\mathcal{M}$-periodized regions (right hand side of Fig. 15). For convenience’s sake, we choose a form with $k$ and $\omega$ separable: $\displaystyle\Phi(k,\omega)=\,\Psi(k)\,\Psi(\omega)$ $\displaystyle\Psi(\alpha)=\frac{1}{2}\,\left[1-\tanh(\beta_{\alpha}r_{\alpha})\right]$ (55) where $r_{\alpha}$ is the distance from the a center in $k$ or $\omega$ space: $\displaystyle r_{k}$ $\displaystyle=$ $\displaystyle\,\sqrt{(k_{x}-k_{x_{c}})^{2}+(k_{y}-k_{y_{c}})^{2}}-r_{o}(\theta)$ $\displaystyle r_{\omega}$ $\displaystyle=$ $\displaystyle\,\|\omega\|-\omega_{o}$ (56) Here $(k_{x_{c}},k_{y_{c}})\sim(\frac{\pi}{2},\frac{\pi}{2})$, $r_{o}(\theta)=r_{o}\,e^{-(\theta-\frac{\pi}{4})/\sigma_{o}^{2}}$ and $\theta=\arctan(k_{y}/k_{x})$ can be chosen to properly shape the patch. We fix through all the rest of the paper $\omega_{o}\sim 0.5t$, $\beta_{k}\sim 100$ and $\beta_{\omega}\sim 15$, which smooth the boundaries of the patch. It remains to fix the parameters $\mathbf{k}_{c}$, $r_{o}$ and $\sigma_{o}$, which have to be chosen to mimic the phenomenological patch of Fig. 15. There is of course a good degree of arbitrariness in its form. We want however to have a first order qualitative description of the physical properties in momentum space, with the aim to address consideration that are only slightly dependent on the exact form of the patch. ### V.1 Evolution of the Fermi Surface Figure 16: (Color online). We show the evolution of the normal-component Fermi surface (blue continuous line in the $\mathcal{M}$-periodization scheme, green dashed line in the $\Sigma$-periodization scheme), defined has $\xi_{k}+\hbox{ Re }\Sigma_{k}=0$ in eq. 36. Panels span from the over-doped region ($\delta=0.13$ top-left panel) to the Mott insulator ($\delta=0.0$ bottom- right panel). By periodizing the cumulant $\mathcal{M}$, the Fermi surface disappears in the insulating state and it is replaced by a line of zeroes of $G_{k}$ (red lines). The way this takes place is via a topological phase transition around $\delta_{p}\sim 0.06<\delta<\delta_{c}\sim 0.08$, from a large hole-like Fermi surface to a pocket-like Fermi surface. The low doping region is marked by a phase where Fermi surface and lines of zeroes co-exists. This gives origin to the pseudogap and the fragmentation of the Fermi arcs in measures of spectra of the normal state system (see e.g. Fig. 18). By periodizing the self-energy $\Sigma$ instead, by construction, the lines of zeroes cannot appear. The pseudogap is realized only trough a modulation of the spectral intensity along the Fermi surface line, which extends always on all the Brillouin Zone (see also Fig. 18). In order to fix the dimension of the patch $\Psi(k)$ in $k$-space it is useful to follow the evolution of the Fermi surface in approaching the Mott insulator with the two periodization schemes $\Sigma$ and $\mathcal{M}$, as shown in the panels of Fig. 16. At high doping (in our case $\delta=0.13$) all the system is well described by a Fermi liquid, the self-energy is mostly local, the $\Sigma-$ (green dash line on the left top panel of Fig. 16) and $\mathcal{M}-$ (blue continues line on the left top panel of Fig. 16) periodizations give in practice the same result. We can choose to describe the system with the $\Sigma$-periodization and $\Psi(k)$ covering all the $k$-space (for example see the left panel of Fig. 17 for $\delta=0.13$, where the patch is marked by the gray region covering all the quadrant). By reducing doping, however, the off-diagonal components of the cluster self-energy $\hat{\Sigma}$ are not negligible anymore, and the two periodizations produce different results. In particular, as we stressed in the previous sections, doping $\delta_{c}\sim 0.08$ and $\delta_{p}\sim 0.06$ are special points. In the $\mathcal{M}-$periodization the Fermi surface shows a striking topological phase transition, produced by the appearance of lines of zeroes of the Green’s function $G_{k}(\omega\to 0)$ at the Fermi level (marked by a continuous red line in Fig. 16). The effect of the appearance of the lines of zeroes of the Green’s function is at the origin of the opening of a pseudogap in the spectral function close to the $(0,\pi)-(\pi,\pi)$ and $(\pi,0)-(\pi,\pi)$ sides of the first quadrant in the Brillouin zone (as shown for example in panel C of Fig. 9) . In the $\Sigma$-periodization instead a more continuous evolution of the original high-doping Fermi surface (green dashed line) takes place at all dopings up to the Mott insulating state, where it disappears. We notice however that the doping $\delta_{c}$ still marks a change in the curvature of the Fermi surface (even if with the $\Sigma$ periodization the effect is more difficult to be noticed at naked eye), as evidenced in the downturn of the $k_{nod}$ vector as a function of doping in panel A of Fig. 9. This effect was first noticed in ref.marce05 , where we show that the result of periodizing $\Sigma$ produces a Fermi surface which enhances its hole-like curvature while reducing doping (and the spectral weight reduces too with respect to the nodal point). This goes in the direction of forming a hole pocket, which however never arrives to be created within the $\Sigma-$periodization. Within the $\mathcal{M}$ scheme instead, the Fermi pocket forms at low doping and its progressive reduction in approaching the Mott transition describes the way the Fermi surface disappears. In the Mott insulating state Fermi lines have of course totally disappeared, but a line of zeroes of $G_{k}$ (i.e. a line of divergent self-energy $\Sigma^{nor}_{k}(\omega=0)$) remains (in red) close to the corner $k=(\pi,\pi)$ of the first quadrant of the Brillouin Zone. It is very difficult to state how close to the real solution one or the other of the two descriptions are. Particular intriguing is the $\mathcal{M}$ periodization result. Not only it well portrays the pseudogap in anti-nodes, it also produces, together with the Fermi surface, lines of zeroes of $G_{k}$. In this way it describes a continuity from the Fermi liquid at high doping (where only the Fermi surface is present) to the Mott insulating state (where only lines of zeroes are present). To this respect, it is clear that the $\Sigma$ periodization scheme fails in describing the Mott state, as it is unable by construction to build up lines of zeroes. Within our analysis however, we cannot claim that in the real physical system a Fermi pocket, together with lines of zeros, is actually present. Recent experiments on cuprate systems, where it has been possible to induce a low temperature normal state by the application of an external magnetic field, have actually observed de Haas-van Alphen oscillations compatible with a Fermi pocket pictureDoiron- Leyraud07 . Hall resistivity measures, extracted at low temperature by suppressing superconductivity with the application of an external magnetic fieldbalakirev-2003 ; balakirev-2007 ; daou-2008 , are also compatible with the scenario of a topological phase transition of the Fermi surface. Our results are also in strong resemblance with the theoretical study of ref.yang-2006-73 , where similar conclusions on the evolution of the Fermi surface and the appearance of lines of zeroes have been drown starting from an ad hoc model for the doping-driven Mott insulator transition in two dimensions. Other non-perturbative microscopic approaches have drawn conclusions in similar directions (see e.g. ref.Plakida07 ; avella07 ). There are however some caveats in concerning the hole pocket which have to be considered. In first place, according to the generalized Luttinger theoremLuttinger60 ; Dzyaloshinskii03 ; essler-2003-90 ; konik-2005 the volume enclosed between the Fermi surface and the line of zeroes (if present) should be equal to the particle density $n=1-\delta$. This theorem is quite respected (but not so much at low doping) by the Fermi surface derived with the $\Sigma$-periodization (green dashed line of Fig. 16). It is not clearly obeyed instead at low doping by $\mathcal{M}$-periodization, as evident e.g. by looking at the area enclosed between the Fermi pocket (blue line) and the lines of zeroes (red line) in the panel $\delta=0.05$ of Fig. 16 (close to half-filling the volume which gives the correct density should be half of the quadrant). Far from stating that the system is violating Luttinger theorem (which in strongly correlated system is actually a possibilityrosch-2007-59 ; stanescu-2007-75 ), this effect is likely an artifact coming from the truncated Fourier expansion in eq. 37, which can be improved only by increasing the cluster size. A possible scenario is that in real systems the line of zeroes is closer to the zone diagonal which goes form $(0,\pi)\to(\pi,0)$, as actually we find at half-filling for a chemical potential value in the middle of the Mott gap $\mu\sim U/2$ (the panel displayed in Fig. 16 has a chemical potential close to the Mott gap edge $\mu\sim U/2-\Delta_{M}/2$, where $\Delta_{M}$ is the total Mott gap). This is in fact what was propose in ref.yang-2006-73 as starting hypothesis. In this way, the side of the hole-pocket facing the $(\pi,\pi)$ corner point is ”cancelled” by the proximity of the lines of zeroes (see for example the spectral functions of Fig. 17), and the resulting picture is a Fermi arc in the nodal region, which is replaced by a lines of zeroes in the antinodal regions. In second place, the discussion we have previously carried out (see Fig. 8 in section III and the description of nodal properties in section IV) shows that the low energy nodal point is better portrayed by periodizing $\Sigma$, i.e. at the nodal point we rather have a Fermi arc more than a pocket. This is important if the nodal point properties (presented in the previous sections III and IV) have to be well portrayed. The real solution result at the nodes is likely to lay in between the $\mathcal{M}$ and $\Sigma$ periodization schemes. ### V.2 The choice of the patch shape Figure 17: (Color online). The dimension of the patch $\Psi_{k}$ in momentum space (gray shaded region) is determined in such a way that the $\mathcal{M}$-Fermi surface (black line) is cut away and replaced by the $\Sigma$-Fermi surface (dashed green line). The red line are where the self- energy $\Sigma_{k}$ is diverging and $G_{k}=0$. Figure 18: (Color online). The spectral function $-\frac{1}{\pi}\hbox{Im}G_{k}(\omega\to 0)$ in the first quadrant of the Brillouin Zone obtained with the mixed periodization (top row) is confronted with one obtained from the self-energy ($\Sigma$) and the cumulant ($\mathcal{M}$) periodizations (second and third row from the top). The broadening $\eta=0.052$, the color scale $x=0.5$ (see appendix A for details). Figure 19: (Color online). Comparison between the $\Sigma$-periodization scheme, the cumulant $\mathcal{M}$-periodization scheme and the mixed- periodization scheme for a low doped system $\delta=0.05$. On the top row we show the full energy range $-14<\omega<14$ covering the lower Hubbard and upper Hubbard bands. The gross features are very similar. Notice the $\Sigma$-scheme introduces fake density of states in the Mott Hubbard gap. In the bottom row we show a close up at low energy $-1<\omega<2$, where the methods are most different. The mixed scheme is constructed to retain a $\Sigma$-periodization character close to the nodal region $k\sim(\frac{\pi}{2},\frac{\pi}{2})$ and a $\mathcal{M}$-periodization character in the antinodal region $k\sim(0,\pi)$. Spectral functions are displayed by introducing a $\omega$ dependent broadening i$\eta(\omega)$ and the maximal color scale value are $x=0.20[0.25]$ for the top [bottom] panel (see appendix A). In line with the discussion above, we choose therefore to assign to the nodal region the path $\Psi_{k}$, which uses the $\Sigma$ periodization. The rest of the $k$-space (covered by $1-\Psi_{k}$) is described with the $\mathcal{M}$ periodization. By following this criterion we establish the evolution of $\Psi(k)$ with doping, as shown in Fig. 17. The size and shape of the patch $\Psi_{k}$ is determined so that the $\mathcal{M}$ Fermi surface is cut away, and the only piece of Fermi surface inside $\Psi_{k}$ is the one produced by periodizing $\Sigma$ (green dashed line). By reducing doping $\delta$ the size of the patch $\Psi_{k}$ progressively reduces, until possibly disappearing in the Mott insulating state ($\delta=0$) where only the $\mathcal{M}$-scheme reproduces well the gapped spectra. We have fixed the center of $\Psi_{k}$ by choosing $(k_{x_{c}},k_{y_{c}})=(k_{x_{nod}},k_{y_{nod}})$, the momentum coordinate of the nodal point described in panel A of Fig. 9. A complete description of the patch parameters $r_{o}$ and $\sigma_{o}$ (see formula 55) as a function of doping is given in the following table: $\delta$ \| $>$0.08 \| 0.08 \| 0.06 \| 0.055 \| 0.05 \| 0.04 \| 0.03 ---\|---\|---\|---\|---\|---\|---\|--- $r_{o}$ \| 3.00 \| 2.20 \| 2.20 \| 2.20 \| 2.20 \| 2.20 \| 2.20 $\sigma_{o}$ \| $\pi/2$ \| $\pi/2.4$ \| $\pi/2.8$ \| $\pi/5$ \| $\pi/8$ \| $\pi/12$ \| $\pi/16$ We notice that with the choice of this patch most of the lines of zeroes disappear from the quadrant the Fermi level. As mentioned above, we are not able to make definitive statement about the actual position of these lines of zeroes. If it were closer to the quadrant diagonal (as conjecture in workyang-2006-73 ), they would re-appear in the $1-\Psi_{k}$ region of momentum space, which is described by the cumulant. As stressed in the previous subsection, this would be important if one want to respect the Luttinger theorem on the particle counting. In this mixed periodization used for the $2\times 2$ plaquette result however this does not take place. We remark however that in the $1-\Psi_{k}$ region scattering rate (i.e. self- energy) is very high (similarly to the results in ref.avella07 ), and this fact is ultimately the reason for the appearance of a pseudogap, whether or not the lines zeroes are effectively present. The division of the $k$-space via the patch $\Psi_{k}$ is surely artificial, and it has not pretext of describing in detail the real properties of the system. It has however capable of capturing in a unique scheme either the virtues of the $\Sigma$-periodization (above all in the nodes) and of the $\mathcal{M}$-periodization (above all in the anti-nodes), which portray with some good degree of confidence the physical properties in different regions of momentum space (as we discussed in the previous sections). A confront between the $\Sigma$, $\mathcal{M}$ and mixed periodizations is presented in Fig. 18 and Fig. 19. In Fig. 18 we show the spectral density $A(k,\omega)=\,\frac{1}{\pi}$Im$G^{nor}_{k}(\omega\to 0)$ in the first quadrant of the Brillouin Zone for decreasing doping $\delta$ (from left to right). The mixed-periodization scheme (top row) is confronted with the $\Sigma$ and $\mathcal{M}$ periodizations (bottom rows), showing how the patch $\Psi_{k}$ is interlacing them. The well known phenomenon of the Fermi arc breakupdamascelli ; campuzano is reasonably well described by all methods, showing this is a solid result of CDMFT. Moreover, the similarity with spectra calculated in previous CDMFT work on the normal statetudor ; tudor06 shows the smooth continuity between the normal component spectra of this superconducting state solution with the spectra of a normal state result. In Fig. 19 we show the band spectrum $A(k,\omega)$ plotted as a function of the energy $\omega$ along the path $(0,0)\rightarrow(\pi,\pi)\rightarrow(0,\pi)\rightarrow(0,0)$ in the first quadrant of the Brillouin Zone. In the top panel we confront the full energy range $-14<\omega<14$ covering upper Hubbard band and lower Hubbard band. At this energy resolution the three schemes are qualitatively very similar. We just stress that the $\Sigma$-periodization artificially introduces spectral weight in the Mott gap (as discussed in Fig. 8 and evident in the figure around $\omega\sim 3-7t$). This justify the choice of cutting the patch $\Psi(\omega)$ at low energies $\omega\leq 0.5t$, using the $\mathcal{M}$-periodization in the remaining of the energy range. In the bottom panel of Fig.(16) we show a close up at low energy ($-1<\omega<2$). In this energy range the methods most differ, however the qualitative results are still very similar. In particular the mixed scheme has been designed to well describe a Fermi liquid linear dispersion at the node, and the right description of the Mott gap, especially in the antinodal region. In the analysis of the following sections we will therefore apply the mixed scheme introduced here through formula V to periodize the normal component of the self-energy (while the anomalous component $\Sigma_{ano}(\omega)$ is always obtained through formula 43, implying by construction a d-wave shape of the superconducting gap). ### V.3 Local density of state with the mixed scheme Figure 20: (Color online). Local density of states $N(\omega)$ as a function of the doping obtained with the mixed-periodization scheme. The V-shape at $\omega\rightarrow 0$ is expected from a d-wave superconductor with nodal Fermi liquid quasiparticles. For doping $\delta<\delta_{p}\sim 0.6$ spectra develop a marked asymmetry. This result interprets the cluster result of Fig. 6 and it is in good agreement with scanning tunneling experiments (see e.g. ref.davis05 ). The display on the real axis of the Green’s function has been obtained by introducing a $\omega$-dependent broadening i$\eta(\omega)$ (see appendix A for details). We use now the mixed periodization to re-calculate the local density of states $N(\omega)=-\frac{1}{\pi}\sum_{k}\,\hbox{Im}G_{k}(\omega)$ at low energy, beyond the cluster energy resolution (which is due to the finite dimension of the truncated Anderson impurity model used to implement CDMFT, see Fig. 6). As widely explained above, this relies on the implicit assumption that the superconducting gap has a d-wave form (see eq. 43) and at the node we have a well defined Fermi liquid arc. The result is shown in Fig. 20. A ”V-shaped” $N(\omega)$ is observed either in the over-doped and under-doped regions. Coming from the over-doped side towards the under-doped side, the slope is always decreasing, until reaching a saturating value at the small doping (until eventually showing a small up-turn for the smaller doping, see Fig. 12). The slopes well fit the analytical value extracted at low energy in formula 51. On a wide range of energy ($-1<\omega<1$), the V-shape is quite symmetric in the over-doped region $\delta\geq\delta_{c}\sim 0.08$ (as already remarked in the CTQMC-CDMFT study of ref.haule-ctqmc ). Spectra become strongly asymmetric in the under-doped region($\delta\leq\delta_{p}\sim 0.06$), when the pseudogap in the normal part of the system opens and super- impose to the superconducting gap (as commented in panel D of Fig. 9). The qualitative behavior of these curves should be compared with the raw cluster result of Fig. 6. The periodizing mixed scheme we introduced should be considered as the best fit we could achieve to our cluster DMFT results, basing on few reasonable solid assumptions on the physical properties of the system (like the d-wave superconducting gap, the pseudogap formation, the Fermi liquid properties at the node). This allows us to recover a momentum dependent Green’s function and access physical quantities comparable to experimental results on cuprates. The qualitative behaviour of $N(\omega)$ we determined well portrays in fact results of scanning tunneling experimentaldavis05 , and it supports a comparison in momentum space. ## VI Quasiparticle spectra from the superconductor to the insulator. In this section we use the mixed-periodization scheme and derive a detailed description of the connection between the spectra of the Mott insulator and the superconductor by varying doping. ### VI.1 Doping a Mott insulator Figure 21: (Color online). Band dispersion $A(k,\omega)$ along the path $(0,0)\rightarrow(\pi,\pi)\rightarrow(0,\pi)\rightarrow(0,0)$ in the first quadrant of the Brillouin Zone. The Mott-insulating state ($\delta=0$) is confronted with a small doping ($\delta=0.04$) state. Top-left corner: the full range of energies ($-15<\omega<15$) covered by the one-particle spectrum is displayed. The gross band-structure, displaying a lower Hubbard band (LHB) ($-6<\omega<0$) and a upper band (UHB) ($10<\omega<15$) separated by a Mott gap ($0<\omega<10$) remains un-changed in going from the insulator into the metal. Top-right corner: detail of the LHB, which rigidly shifts in going from the insulating to the metallic state. Bottom-left corner: detail of the low energy ($-1<\omega<2$) feature. In going into the metallic state, the shift of the bands is not rigid and spectral weight appears at positive energy as soon as doping is added into the system. Bottom-right corner: detail of the UHB. Upon doping, a rigid shift in the metallic state is accompanied by a transfer of spectral weight to low energies. The color scale maximum value is always $x=0.20$, except for the lower left panel, where it is set $x=0.25$. We refer to appendix A for the choice of the broadening i$\eta(\omega)$. To this purpose, we analyze in Fig. 21 the spectral functions, $A(k,\omega)=\,-\frac{1}{\pi}\hbox{Im}G(k,\omega)$ in the $\omega-k$ space, along the $k$-path $(0,0)\rightarrow(\pi,\pi)\rightarrow(0,\pi)\rightarrow(0,0)$ in the first quadrant of the Brillouin zone. We compare the insulating state ($\delta=0$) with a slightly doped state ($\delta=0.04$). In the left top panel we display a wide energy-range $-14<\omega<14$, which covers the lower (LHB) and upper (UHB) Hubbard bands. We notice that, upon adding a small doping, the gross structure of the lower Hubbard band ($-6<\omega<0$) and the upper band ($10<\omega<15$), which are separated by a Mott gap ($0<\omega<10$), remains substantially un-changed. A detail of the lower Hubbard band, a closeup of the Fermi level ($-1<\omega<2$) and a detail of the upper Hubbard band are presented in the right top, left bottom, right bottom panels respectively. The LHB (top right) rigidly shifts in going from the insulating to the metallic state, loosing spectral intensity, which goes to build up quasiparticles at the Fermi level $\omega=0$ (bottom left). The novel outcome from the CDMFT calculation, as compared with the single-site DMFT revmodmft , consists in the anisotropic fashion quasiparticles first occupy the Fermi level in momentum space. This is better seen in a closeup of the band dispersion around the Fermi level (bottom left). The region where firstly quasiparticles appear is close to $k=(\frac{\pi}{2},\frac{\pi}{2})$. In fact, already in the insulating state ($\delta=0$), we observe that around $k=(\frac{\pi}{2},\frac{\pi}{2})$ a heavy-particle hook-shaped band is closest to the $\omega=0$ level, while in the proximity of $k=(0,\pi)$ there is a ”pseudogap” and the band disperses at negative energies. This kind of band-structure survives upon doping, as we can observe in the $\delta=0.04$ panel. Contrary to the Hubbard bands, however, the low-energy band does not shift rigidly as doping is added to the insulator, rather it stays pinned at the Fermi level, and the shifts in frequency is only a small fraction of the changing in chemical potential. We also observe the appearance of spectral weight at positive energy, coming from both the LHB and UHB, which starts building up a full Fermi-liquid-like band, as we show more in detail in the following. In the bottom right corner of Fig. 21 we finally show the UHB. In this case a rigid shift in the metallic state is accompanied by a strong reduction of spectral weight. In fact, according to our result, the UHB narrows with respect to the insulating state. ### VI.2 Approaching the Mott insulator from the over-doped side We start now from the viewpoint of the highly doped system and observe how the approach to the Mott insulator affects the electronic structure of the superconducting state. The high-doping system offers the advantage of having more standard Fermi-liquid-like properties (the patch $\Psi(k)$ introduced in the previous section covers all the momentum space). Figure 22: (Color online). Evolution of the one-particle spectrum from the over-doped (left side, $\delta=0.13$) to the under-doped (right side, $\delta=0.04$) state. The typical Fermi-liquid band, well visible in a narrow energy range ($-1<\omega<1$) around the Fermi level at high doping $\delta=0.13$, is progressively destroyed by reducing doping, while the upper and lower Hubbard bands build up acquiring spectral weight . These features are shown in detail in Figures 23 and 24. The color scale maximum value is $x=0.20$ and we refer to appendix A for the choice of the broadening i$\eta(\omega)$. In Fig. 22 we show the electronic band $A(k,\omega)$ in the $k-\omega$ space, once again in the path $(0,0)\rightarrow(\pi,\pi)\rightarrow(0,\pi)\rightarrow(0,0)$ of the first quadrant of the Brillouin zone, on the full energy-range $10<\omega<15$ covering the LHB and UHB. Here we want to display the evolution from high doping (left) to small doping (right). At high a doping $\delta=0.13$ a Mott Hubbard gap which separates a LHB structure from the UHB is already visible. A narrow but Fermi Liquid-like band is however present, and it crosses the Fermi level either in the region of momentum-space around $k=(\frac{\pi}{2},\frac{\pi}{2})$ and $k=(0,\pi)$ (see also Fig. 24). Reducing the doping $\delta$, we see that this narrow band loses intensity more and more (follow the horizontal line at $k=(\pi,\pi)$ in $\omega\sim 0$) to the advantage of the Hubbard bands, which instead gain spectral weight in approaching the Mott insulator. To have a glance on how this is taking place we look in the following figures at the different energy-regions of the band in further detail. Figure 23: (Color online). Top panel: Evolution of the LHB (top) from the over-doped (left side, $\delta=0.13$) to the under-doped (right side, $\delta=0.04$) state. The LHB rigidly shift according to the change in chemical potential by reducing doping, acquiring structure (follow the horizontal line at $k\sim(0,\pi)$. Bottom panel: Evolution of the UHB from the over-doped (left side, $\delta=0.13$) to the under-doped (right side, $\delta=0.04$) state. The UHB rigidly shift according to the change in chemical potential by reducing doping and, as evident in the color intensity, it gains spectral weight. The color scale maximum value is $x=0.20$ and we refer to appendix A for the choice of the broadening i$\eta(\omega)$. Fig. 23 shows the UHB and the LHB in detail. The statements made above are confirmed: both Hubbard bands gain spectral weight in decreasing doping, mainly in the region of momentum space close to the anti-nodes $k\sim(0,\pi)$ (follow once again the horizontal line), while rigidly shifting with respect to the change of chemical potential $\Delta\mu$ (one can actually show that the shift of the bands in energy is equal to $\Delta\mu$). The behavior of the Hubbard bands is therefore in agreement with a picture describing the approach to the Mott insulator as a rigidly moving bands, which transfer part of their weight to low energy. As mentioned in the previous subsection, the novelty of our CDMFT result in finite dimension, with respect to the standard vision of the Mott transition given by single-site DMFT in infinite dimension, is that this transfer of spectral weight takes place in a very anisotropic fashion, with the antinodal regions $k\sim(0,\pi)$ getting insulating before the nodal ones. Figure 24: (Color online). Top panel: we show the evolution of the low-energy one-particle spectral band from the over-doped (left side, $\delta=0.13$) to the under-doped (right side, $\delta=0.04$) state. A Fermi-liquid-like band at high doping (left side, $\delta=0.13$) is progressively destroyed by decreasing doping. Around the region $k=(\frac{\pi}{2},\frac{\pi}{2})$ the dispersion presents a Fermi liquid quasiparticle crossing the Fermi Level. Around the region $k=(0,\pi)$, the d-wave superconducting-state opens a gap by removing spectral weight from the Fermi level to positive and negative energies. Bottom panel: we show the contribution to the low energy band coming from the normal component of the system (we set the anomalous self-energy $\Sigma_{ano}=0$). In particular we observe that in the under-doped side ($\delta\leq 0.08$) a gap (the pseudogap) opens in the $k=(0,\pi)$ region even if the superconductive term is absent. For convenience’s sake, the color scale maximum value is set $x=0.30$ for $\delta=0.13$ and $x=0.25$ for all the other doping values. We refer to appendix A for the choice of the broadening i$\eta(\omega)$. A closer look to the behavior at low energy, Fig. 24, reveals much richer phenomena taking place. In the top panel we show the spectra resulting from the one-particle Green’s function $G_{11}(k,\omega)$ (eq. 36) in the superconducting state. First of all, contrary to the behaviour of the Hubbard bands, the low energy part of the band does not shift proportionally to $\Delta\mu$, in agreement with the observation already made in Fig. 21. The Fermi-liquid-like band at high doping (left side, $\delta=0.13$) is progressively destroyed by decreasing doping (with progressive reduction of spectral weight in the arc at $0<\omega<2$). In the region close to $k=(0,\pi)$ instead, the d-wave superconducting-state opens a gap by removing spectral weight from positive and negative energies (a Bogoliubov band is formed at $\omega>0$). As widely discussed in the previous sections however, the presence of lines of zeroes in the Green’s function kicks in the under- doped region ($\delta<0.08$), opening a pseudogap in the normal component of $G_{11}(k,\omega)$ around $k=(0,\pi)$. This is shown in the bottom row of Fig. 24, where the same panels of the top row are reproduced by imposing the anomalous self-energy identically zero $\Sigma_{ano}=0$ in eq. 36. Switching off superconductivity has in general little effect on the quasiparticle bands, except indeed close to the antinodal point $k\sim(0,\pi)$ and close to the Fermi level $\omega\sim 0$ (the nodal points $k\sim(\frac{\pi}{2},\frac{\pi}{2})$) are practically unaffected). The differences are most evident in the over-doped region (see e.g. $\delta\sim 0.13-0.08$ close to $k\sim(0,\pi)$), where the superconducting gap has disappeared and the quasiparticle band is reconstructed. The weight in the Bogoliubov branch at $\omega>0$ also disappears. For $\delta>\delta_{c}\sim 0.8$ differences are much less evident (above all $\delta=0.04$). A gap is also present in the normal component solution, the structure of the quasiparticle dispersion around $k\sim(0,\pi)$ is very reminiscent of the superconducting solution in the corresponding top row. It seems that in the under-doped region the superconducting gap appears to complete the structure already present in the normal component solution. This indicates that the pseudo-gap appearing in the normal component of the system (and present at temperatures above TC) and the d-wave superconducting gap are possible answers of the system to the same instability, and they co-exist in the under-doped region (while in the over-doped only the superconducting gap is present). This instability is, in our view, connected to the approach to the Mott transition, and it reflects the anisotropic way chosen by the Hubbard system in two dimension to approach the insulating state: the regions around the anti-nodes $k\sim(0,\pi)$ become insulting (at $\delta=\delta_{c}\sim 0.08$) before the regions close to the nodes $k\sim(\frac{\pi}{2},\frac{\pi}{2})$ (which become finally insulating $\delta=0$). ## VII Low-energy kink in the quasiparticle spectra Figure 25: (Color online). The the low energy kink-feature in the spectral intensity at the nodal point for various values of doping. For convenience’s sake, the color scale maximum value is set $x=0.35$ for $\delta=0.13$ and $x=0.30$ for all the other doping values. We refer to appendix A for the choice of the broadening i$\eta(\omega)$. We can now compare the quasiparticle spectra presented in the previous section with the spectra measured, e.g. by photo-emission, in cuprate-based systems. Unlike the linear dispersion predicted by simple band calculations, in recent years a series of experiments on the electronic structure of many H-TC superconductor compounds has shown sharp breaks in the dispersion of spectra , ”kinks”, at binding energies of the order of 50-80 meVLanzara01 ; Z-XShen02 ; he-2001-86 ; Hwang04 ; cuk-2004-94 . Sudden changes in the quasiparticle velocity were reported by a factor two or more. This break in the dispersion is evident at and away from the $d$-wave node line, and its magnitude shows doping and temperature dependence. Kinks may provide useful information on the nature of the coupling between electrons and possible single-particle or many- body excitations, which are at the origin of strongly-correlated many-body properties of the system. In H-TC superconductors these feature have been associated either with phononsLanzara01 ; Z-XShen02 or spin-fluctuation basedhe-2001-86 ; Hwang04 pairing mechanisms. In Fig. 25 the dispersion we observe in the quasiparticle band of our result around $k=(\frac{\pi}{2},\frac{\pi}{2})$ is similar to the experimentally observed kink (see e.g. ref.kaminski ). A neat linearly dispersing quasiparticle crosses the Fermi level ($\omega=0$) in the $k$-$\omega$ plot, but around $\omega\sim-0.2t$, the dispersion suddenly changes in slope and gets more incoherent, as evidenced by the broadening spectra. If we set $t\sim 300$meV as order of magnitude, we have that the kink appears at $\omega\sim 60$meV, in good agreement with the observed experimental energy range. The kink present in a wide range of doping, from the under-doped to the over-doped regime, and its slope is increasing with decreasing doping. This new energy scale in H-TC superconductors arises the debate on the possible nature of the electron-electron coupling. In order for superconductivity to take place in metals, it is necessary that electrons bind into pairs, which condense in a phase-coherent quantum state. In standard BCS superconductivity, coupling between electrons and phonons (lattice vibrations) drives the formation of the pairs. The existence of the kink low-energy scale, not explicable in band theory calculation, may give an hint on the low-energy nature of the interaction between electrons, and hence help revealing the pairing mechanism of the unconventional H-TC superconductivity. The single- band Hubbard Model studied in this paper does not take into account phonons by construction. Therefore the presence of the kink supports the idea that the origin of these features are indeed purely electronic, in agreement with the DMFT and CDMFT studies of ref.byczuk-2007-3 ; chakraborty-2007 . In Fig. 26 we present the spectral function $A(k,\omega)=\,-\frac{1}{\pi}$Im$G(k,\omega)$ calculated in our theory, confronted with experimental ARPES data taken from ref.kaminski . Fig. 26 shows the spectral function $A(k,\omega)$ as a function of the energy $\omega$, along two vertical cuts in the first quadrant of the Brillouin Zone in correspondence of the nodal and antinodal region $(k_{nod},0)\sim(\frac{\pi}{2},0)\rightarrow(k_{nod},k_{nod})\sim(\frac{\pi}{2},\frac{\pi}{2})$ and $(k_{anod},0)\sim(\pi,0)\rightarrow(k_{anod},k_{anod})\sim(\pi,\frac{\pi}{2})$. The system is close to optimal doping $\delta_{c}\sim 0.08$. In the left column we display the CDMFT calculation, while the right column the experimental data, taken from ref.kaminski . The nodal quasiparticle clearly show a dispersion, that from the Fermi level propagates at negative frequencies until $\omega\sim-0.2t$, where it has a sudden broadening, indicating a strong incoherence. The antinodal quasiparticle shows instead a much flatter dispersion at energy corresponding to the superconductive gap, and the antinodal quasiparticle (which shows to have less weight than the nodal quasiparticle) does not lose much coherence. This plots have a good resemblance with the experimental data on the right column. The comparison between the energy scale of our results in the left $-t<\omega<0$ and of the experimental data $-300$eV$<\omega<0$ shows also that the esteem we have used $t\sim 300$eV is a reasonable order of magnitude. Figure 26: (Color online). Top panel: The low energy spectral intensity $A(k,\omega)$ vs $\omega$ calculated within CDMFT (left) is confronted with the experimental result of ref.kaminski in the nodal and antinodal regions. Vertical cuts are traced in $k$-space close to the node and the antinode, as traced on the figures. $\mathbf{k}$ is expressed in units of $(\pi,\pi)$. Bottom panel: Color-intensity plots of the spectral intensity $A(k,\omega)$ in the $k-\omega$ space in the nodal (left) and antinodal (right) cuts of momentum space (the same used in the top panel). The color scale is set $x=0.30$ (see appendix A). In the bottom row the experimental result of ref.kaminski are displayed for comparison. We refer to appendix A for the choice of the broadening i$\eta(\omega)$. The same comparison is also presented in the intensity color plot at the bottom of Fig. 26, where the nodal (left side) and the antinodal (right side) dispersion calculated with CDMFT are displayed in the top row, while experimental plots of ref.kaminski are in the bottom row. From these diagrams, it is more evident the quasiparticle-like dispersion in the nodal point, which compares very well to the data from photo-emission. In the photo- emission data however at $\omega\sim-0.075$eV the quasiparticle appears loosing coherence (effect marked by the loss of red color that gets yellow) and at the same time the slope of the dispersion changes substantially. In the CDMFT calculation instead, as soon as the quasiparticle loses coherence (around $\omega\sim-0.2t$), the dispersion stops (i.e. is $\omega(k)\sim$ constant), and no spectral weight is present in the region at smaller energy ($\omega<-0.2t$). The kink appears in our result as a ”gap” in the $\omega$ vs $k$ spectra. A similar phenomenon takes place at higher energies, as marked by an arrow in the top-left panel of Fig. 23, where an evident gap is present in between the low energy band and the lower Hubbard band. This could be possibly associated with the ”water-fall” dispersion features recently observed in many cuprate materials (see e.g. ref.graf-2007-98 ; valla-2007-98 ) at energies $\sim 350-600$ meV, where the same $k$-vector is marking the dispersion $\omega(k)$ in a rather extensive range of energy. In our calculation this phenomenon is described as a gap in the dispersion. The antinodal point, where the superconducting gap is present, is non-dispersing at low energy (right- bottom side in Fig. 26), and the resemblance of our result with the photo- emission data is rather good. ## VIII The Hall resistivity to detect the topological phase transition of the Fermi Surface We have shown in section V that in the under-doped region ($\delta<\delta_{p}\sim 0.06$) a different regime sets in our solution, marked by a topological transition from a large to a small Fermi surface (which reduces to an arc or pocket). The strong reduction of the Fermi surface area should correspond to a strong reduction of the carrier density too. Figure 27: Hall resistivity $R_{H}$ as a function of doping $\delta$. The topological phase transition of the Fermi Surface is efficiently detected by a rapid change of the behavior in the Hall resistivity RH, which indicates a change of the carrier density. Left: our theoretical result (units are in the unitary cell volume $v_{o}$ over electron charge $e$). Right: the experimental result for $LSCO$ extracted from ref.balakirev-2007 . This could be in principle experimentally detected by measuring the Hall resistivity $R_{H}=-\frac{1}{n_{c}e}$, which is directly related to the carrier density $n_{c}$, provided superconductivity can be suppressed and the underlying normal liquid extracted. This latter is far from being a trivial procedure, which has many caveats. A standard method is for example to apply a magnetic field, which suppresses superconductivity. Recent publications have addressed this problem in $BSLCO$balakirev-2003 ($Bi_{2}Sr_{1.51}La_{0.49}CuO_{6+\delta}$) and $LSCO$balakirev-2007 ($La_{2-\delta}Sr_{\delta}CuO_{4}$), where magnetic field up to 65 Tesla was applied, and in $Nd-LSCO$daou-2008 (La1.6-xNd0.4SrxCuO4), where more moderate magnetic fields (up to 35 Tesla) can totally suppress superconductivity. The results of these experiences on the Hall resistivity show in fact to be compatible with a reduction of the carrier density at low temperature in an under-doped sample (as we will discuss more in detail in the following). In our study we can easily estimate the Hall resistivity by using a Boltzmann approach, similarly for example to the phenomenological study of ref.PSK . The Hall resistivity $R_{H}$ can be expressed in terms of the in-plane conductivity (in the direction of an applied voltage) $\sigma_{xx}$ and the transverse conductivity $\sigma_{xy}$ (perpendicular to the applied voltage): $R_{H}=-\frac{\sigma_{xy}}{\sigma_{xx}^{2}-\sigma_{xy}^{2}}\frac{1}{H}$ (57) where $H$ is the applied field. We can use the low-energy limit (similarly to eq. 49) and extract the normal component of the system (by setting $\Sigma_{ano}=0$ in eq. 36). The Green’s function is conveniently written on the Matsubara axis as: $G(k,\omega_{n})=\,\frac{1}{\hbox{i}\frac{(2n-1)\pi}{Z_{k}\beta}-\zeta_{k}}$ (58) where a re-normalized dispersion $\zeta_{k}=\left[\xi_{k}+\hbox{Re}\Sigma_{k}(0)\right]$ and a inverse temperature $\beta^{-1}$, re-normalized by the quasiparticle residuum $Z_{k}$, are introduced. We consider this liquid as the base on which to start applying a Boltzmann theory. The conductivities can then be expressed in the first order in $H$ (see e.g. ref.PSK ): $\displaystyle\sigma_{xx}$ $\displaystyle=$ $\displaystyle-2e^{2}\tau\,\sum_{k}(v_{k}^{x})^{2}\,\frac{\partial f_{(Z_{k}\zeta_{k})}}{\partial\zeta_{k}}$ (59) $\displaystyle\sigma_{xy}$ $\displaystyle=$ $\displaystyle\frac{2e^{3}H\tau^{2}}{\hbar c}\,\sum_{k}v_{k}^{x}\,\left[\,v_{k}^{y}\partial_{k_{x}}v_{k}^{y}-\,v_{k}^{x}\partial_{k_{y}}v_{k}^{y}\,\right]\,\frac{\partial f_{(Z_{k}\zeta_{k})}}{\partial\zeta_{k}}$ Here, $v_{k}^{\nu}=\partial\zeta_{k}/\partial k_{\nu}$ is $\nu=x,y$ component of the normal velocity, $f(x)=1/(e^{\beta x}+1)$ is the Fermi function, $\tau$ is the scattering time, which at $T=0$ is, for example, given by the impurities, and that, for convenience’s sake, in a first order approach, we can hypothesize $k$-independent. In the pure model, like the Hubbard Model, this term is absent, so we have to add it in order to simulate the finite $T=0$ resistivity of a real material. As the Hall resistivity $R_{H}$ is anyway independent of a constant $\tau$, we do not need to make any special further assumption about it. The Hall resistivity resulting from our calculation implemented via a mixed periodization and plug into the Boltzmann expression is displayed in the left hand side of Fig. 27 as a function of doping $\delta$. The unit used to display $R_{H}$ is the ratio between the unitary cell volume $v_{o}$ and the electron charge $e$. To give an idea of the order of magnitude, taking for $LSCO$ (see e.g. ref.cieplak06 ) an average lattice spacing $a_{o}\sim 4\AA=4\times 10^{-7}mm$ in the $Cu-O_{2}$ planes, and $a_{1}\sim 3a_{o}$ in the $c$-axis $\frac{v_{o}}{e}\sim 1.2\,mm^{3}/C$, which well compares with the experimental results extracted from ref.balakirev-2007 (but similar results hold for $BSLCO$balakirev-2003 ), and presented on the right hand side. In the following, in drawing a parallel between our theoretical result and the experimental data, one should keep in mind that in our study the ”optimal doping” (which we can only identify with some degree of uncertainty as a maximum in the order parameter $\delta_{c}\sim 0.08<\delta<\delta_{p}\sim 0.06$, see also the discussion in the conclusions) is situated at a smaller doping than the experimental value $\delta_{opt}\sim 0.17$. $R_{H}$ is positive, as expected by the hole-like Fermi surface in the hole- doped system. As, in approaching the Mott insulator, the localization of particles reduces free carriers, $R_{H}$ is generally expected to monotonically decrease with doping. Two regimes are clearly separable in our result (left side of Fig. 27). At $\delta\sim\delta_{p}\sim 0.06$ we observe a discontinuity in the behavior of $R_{H}$ (which slightly decreases instead of increasing), and for $\delta<\delta_{p}$ a sudden increase of $R_{H}$, related to the topological phase transition of the Fermi surface presented in section V, marks the reduction in the carriers in approaching the Mott state. Such a discontinuity is present and more evident in the experimental data (right side of Fig. 27). The presence of a local minimum at optimal doping $\delta_{opt}\sim 0.17$ represents a crucial point in the results presented in referencesbalakirev-2003 ; balakirev-2007 . This behavior is interpreted in terms of a quantum critical point, associated with a change in the topology from a large hole-like Fermi surface (realized for $\delta>\delta_{opt}$) to a small (pocket-like?) Fermi surface, realized for $\delta<\delta_{opt}$). According to their point of view, the force driving the superconductivity is related to the fluctuations around a critical point. This attractive force could overcome the mutual Coulomb repulsion of electrons, delocalizing and freeing carriers right in proximity of the quantum critical point. This fact would originate the local minimum in the Hall resistivity $R_{H}$ (which is inversely proportional to the number of free carriers). The critical point would correspond then to the optimal doping $\delta_{opt}$, where the critical temperature $T_{C}$ is the highest. Our result shows qualitatively very similar trends, with a minimum (even if milder) just at doping $\delta_{p}$, where the topological transition of the Fermi surface takes place. This strongly supports the analysis of our results and the conclusion we have derived, in comparison with the experimental evidence on the Hall resistivity. In our 2$\times$2 plaquette study at $T=0$, however, we are not unfortunately able to state if $\delta_{p}$ is or not the optimal doping (which is defined at $T=T_{C}$). This point is as a matter of facts close to the maximum of the d-wave order parameter (as evident in Fig. 1). The study of ref.haule-ctqmc has pointed out that the exact determination of the optimal doping may depend on temperature. Also, we cannot demonstrate that $\delta_{p}\sim 0.06$ is a quantum critical point, as we have not clearly identified an order parameter or the divergence, to some order, in the free energy. Our study however finds good agreement with the experimental results of ref.balakirev-2003 ; balakirev-2007 ; daou-2008 , and it is not in contradiction with their conclusions. We can in fact associate the discontinuity of $R_{H}$ with a dramatic re-arrangement of the electronic structure in the system, corresponding to the topological transformation of the Fermi surface. This happening is marked by a fast crossover region (between $\delta_{c}\sim 0.08<\delta<\delta_{p}\sim 0.06$), where the maximum of order parameter is located, the pseudogap appears (at $\delta\sim\delta_{c}$), the superconducting gap start decreasing (at $\delta\sim\delta_{p}$). All these facts have striking consequences on spectra and transport properties, which we have presented throughout the paper. More developments beyond the 2$\times$2 plaquette CDMFT are needed, in order to be able to reveal if in reality $\delta_{c}\equiv\delta_{p}$, or rather they represent two different ”transition” points. It would be then important to clarify if any of these points are a quantum critical point, and their exact connection with the optimal doping $\delta_{opt}$ of the system. All these are important open questions left for the future studies. ## IX Conclusion ### IX.1 Comparison with the resonating valence bond mean field theory CDMFT can be viewed as a generalization of the earlier slave boson resonating valence bond mean field theorybza ; ruckestein87 . It is therefore useful to put our results in this context. RVB mean field theories had numerous early successes, like the prediction of d-wave superconductivity and a pseudogap phase having the same symmetry as the superconducting stateliu . Both RVB mean field theories and our CDMFT study are formulated in terms of variables defined in a plaquette. The main differences stem from the fact that in CDMFT these variables are frequency dependent, and thus they are able to properly describe coherence to incoherence crossover in the momentum space. In the slave boson RVB mean field theory there are two very important notions. The slave boson order parameter, which measures ”Fermi liquid coherence”, namely the emergence of a well defined quasiparticle peak in the spectral function, and the ”spinon condensation” order parameters, both in the particle-particle and particle-hole channel. The position of the quasiparticle is shifted by a Lagrange multiplier which, adding to the bare chemical potential, forms a quasiparticle chemical potential. The dynamical mean field picture of CDMFT develops these ideas, allowing them to acquire non trivial dependence in momentum space. To begin with, CDMFT describes naturally a high temperature state showing poor coherence. As discussed in referencemarce05 and Fig. 7, the peaks in the spectral functions in the normal state are broad and incoherent. It is only in the superconducting state that we can, for the first time, extract the Fermi liquid parameters discussed in this work. Coherence appears at low temperature in the nodal region more than in the antinodal region, as it is clearly seen for example Fig. 7. Furthermore the weight of the quasiparticle is different (and larger in most of the phase diagram) in the nodal than in the antinodal region, as seen in Fig. 10. At small doping quasiparticles in the antinodal region cannot be defined in a strict Landau-Fermi liquid sense, as a quasiparticle peak cannot be identified at the Fermi level anymore. As a matter of fact a pseudogap opens in the antinodal quasiparticle spectra. In the nodal region instead, quasiparticles are always well defined and their weight dramatically reduces as the doping is reduced. The slave boson RVB picture describes the pseudogap phase in terms of the formation of spin- singlets, parameterized by two order parameters, one describing correlations in the particle-particle channel and the second in the particle-hole channel. A similar but more complete description is achieved in terms of the dynamical anomalous and normal self-energies. Notice the remarkable fact that the anomalous self-energy at low temperatures and at low frequencies has a similar order of magnitude as the normal self-energy (Fig. 2 and Fig. 5), indicating that, as in the RVB mean field theory, they might both have a similar singlet- pairing origin deriving from the approach to the Mott insulator. On the other hand, the normal state self-energy has a more complicated angular dependence than the simple harmonic (d-wave) $k$-dependence of the anomalous self-energy. In this sense, the normal and the anomalous gaps can also be viewed as competing for the same electrons, suggesting that further refinements of CDMFT along the lines of ref.stanescu04 are worth being pursued. ### IX.2 The scenario presented by our CDMFT results While the dynamical frequency dependence introduced by DMFT well describes the low-to-high energy crossover, as well known from the infinite dimensional caserevmodmft , the momentum dependence introduced by a cluster DMFT uncovers a wider spectrum of $k$-dependent physical phenomena, which turn out fundamental in describing the approach to the Mott transition in two dimensions. In the first part of the paper (section II) we have presented raw cluster quantities, which are direct output of the CDMFT procedure, but which can be only partially interpreted in physical terms. Nevertheless we have shown that the doping-driven approach to the Mott transition takes place via an intermediate regime (rising after an ”optimal doping region” $\delta_{c}\sim 0.08>\delta>\delta_{p}\sim 0.06$), where physical properties depart from the standard picture of a BCS superconductor. In particular we were able to identify two distinct energy-scales, one associated with the anomalous component of the self-energy (see e.g. Fig. 5), the other better enlightened in the local density of states (see Fig. 6), showing different doping dependence. In the low doping regime the local spectra also show a strong asymmetry in $\omega$, as observed in experimentsdavis05 and contrary to the expectations from a BCS superconductor. In order to physically interpret the genuine cluster results, we have restored the lattice translational invariance (broken in the CDMFT procedure) by introducing a periodization scheme. In section III we have justified and compared two possible methods, based on periodization of the cluster self- energy and the cluster cumulant respectively. To this purpose, we have discussed the physical properties of our system, taking advantage from either the cluster results (supported also by the comparison with CDMFT results obtained with QMC impurity-solver methodsbpk ; olivier-note ; haule-ctqmc ) and experimental observation. We have performed a robust test on our approach by reconstructing the local quantities from the momentum-dependent Green’s function $G(k,\omega)$ (see Fig. 20), which show to be not far from the corresponding cluster quantities obtained directly in the impurity model. By introducing a periodization procedure in momentum-space we were able to analyze our result in terms of experimentally observable quantities in the nodal and antinodal points of momentum space (section IV), making contact with recent spectroscopy experimentstacon06 ; tanaka06 . In particular, we were able to interpret the two energy scale in terms of a pure superconducting gap (dominant in the nodes) co-existing with a normal component gap (related to the pseudogap of the normal state and dominant in the anti-nodal region of momentum space at low doping). We complete in this way the work presented in the short publication of ref.marce08 . We have then extended our procedure to describe in first approximation physical properties in all momentum space (section V). The scenario which results present an under-doped state where electronic structure undergoes a dramatic re-arrangement. We could associate this fact with the appearance of a topological phase transition of the Fermi surface (see e.g. Fig. 16), driven by the appearance in momentum space of lines of diverging self-energy, fingerprints of the Mott physics. While these results may be quantitatively different for different periodizing methods, in terms of a large enhancement of the real part of the self-energies they indicate the same qualitative trends. Enhancement of self-energy $\Sigma_{k}$ is most relevant in the cumulant scheme. The formation of a pseudogap in the antinodal region, however, only requires a large value of the self-energy and not a strict divergence (which in any case can only occur at $T=0$). Based on this physical idea, at finite doping the antinodal region is closer to the Mott insulating phase, while the nodal region is closer to the Fermi liquid state (see the band structure of Fig. 24 and the quasiparticle peaks Fig. 9). We achieve in this way a consistent picture of the evolution of the electronic structure with doping within our formalism. We were able to explain the strong asymmetry in the local density of states, observed e.g. in scanning tunneling experimentsdavis05 . We were able to give a complete description of the evolution of electronic spectra with varying doping, comparing the doped state with the Mott insulating one (see section VI). In particular, we have shown that the transfer of spectral weight in approaching the Mott transition takes place in a strongly anisotropic fashion. We have described the behavior of the low energy band close to the Fermi level, which is strongly renormalized by the interaction and progressively destroys by reducing doping, opening first a pseudogap in the antinodal region. This latter appears to be present already in the parent Mott insulator. We have stressed how this phenomenon is connected to the aforementioned lines of diverging self-energy. We have described how, from this underlying structure of the normal component, a d-wave superconducting gap rises. We have also shown (in section VII) how the combination of these effects results in spectra which show ”kink” features similar to the ones observed in the electronic dispersion of many photo-emission experiments. Finally we have shown, using a simplified Boltzmann approach, how the rising of the under-doped regime, associated in our study to a topological phase- transition of the Fermi surface, is experimentally detectable from the doping- dependent behavior of transport quantities, like for instance a singularity in the Hall resistivity. All these properties are comparable with experiments on cuprate materials, and the good agreement we have found supports our study of the evolution with doping of the superconducting state of the Hubbard Model, offering a self-consistent scenario of the approach to the Mott transition in two dimension. This is most relevant in connection with cuprate H-TC superconductors, in which the evolution of the electronic structure with doping (and its relation with H-TC superconductivity) remains a fundamental open question. In particular, in recent times, the issue of two (nodal and antinodal) gaps tacon06 ; tanaka06 or one pure d-wave gapcampuzano07 ; campuzano07b ; campuzano08 has been at the center of the experimental and theoretical debatemillis06 ; cho06 ; huefner-2008-71 . In this contest, our CDMFT results on the simplest electronic model for H-TC materials show a good agreement with the two gaps scenario and, at the same time, with many other electronic properties of cuprates. An important question left open in our study is to determine the exact nature of the small ”transition region” between $\delta_{p}\sim 0.06$ and $\delta_{c}\sim 0.08$, where fundamental changes take place in the physical properties of the system, in going from the Fermi-liquid of the over-doped side into the anomalous liquid of the under-doped side of phase diagram. In particular, we have shown that lines of diverging self-energy appear at the Fermi level, a pseudogap opens in the one-particle spectrum and a topological phase transition of the Fermi surface occurs (see Fig. 16). This small region of doping locates, within the numerical precision of our result, the ”optimal doping”, which we identify as a maximum in the d-wave order parameter (Fig. 1). All these observations point towards indicating a tight connection between the physical happening taking place in this optimal doping region and the H-TC mechanism. We are not able to state within our study if behind lays a quantum critical point (which could provide the binding force for a high critical temperature). This scenario is actually supported by many theoriessachdev92 ; perali96 ; kivelson98 ; varma99 and experimental studiesando95 ; boebinger96 ; balakirev-2003 , and it could be that further developments of CDMFT (i.e. increasing cluster size) could reveal that the two points $\delta_{p}$ and $\delta_{c}$ actually coincide. Or two distinct points could be actually present, and one (or both) could have the characteristic of a quantum critical point. To reveal this, the right divergence in any order of the free energy and an order parameter should be clearly identified. At present it is not possible to demonstrate that our solution will survive in the thermodynamic limit (i.e. in the infinite cluster size limit), in the sense that the real ground-state of the Hubbard Model in two dimensions could be another one of the possible competing instabilities, like, e.g. stripe ordering or antiferromagnetism. The latter, in particular, is expected to be the ground-state close to the Mott insulating state. Within a (dynamical) mean-field approach, however, we can study the pure paramagnetic phase, showing that it is a relevant phase even if it is not the true ground-state of the system. We leave open the physical question of what terms need eventually to be added in the Hamiltonian in order to make this state a real ground- state. In order to make our first order picture more rigorous, further developments are needed. These should involve the cumulants directly in the self-consistency condition, as proposed for example in ref.stanescu04 ; tudor , and should exploit the flexibility of CDMFT, which can be formulated in terms of a set of adaptive patches in momentum space (on this line see the recent workferrero-2008 ). These extensions, as well as the use of more powerful solvers, which could allow going beyond a 2$\times$2 cluster, are worth pursuing and are left for future studies. ## Appendix A Displaying function on the real axis with ED-CDMFT Once the ground-state $\|gs>$ of the associate Anderson Impurity Model (eq. 8) has been determined via the Lanczos procedure, it is possible to determined the zero-temperature Green’s function via a second Lanczos step (see e.g. the reviewrevmodmft ). To this purpose, one has to take as initial vector $c^{\dagger}_{\mu}\|gs>$ ($\mu$ denoting the generic cluster-site index), and write the ground-state Green’s function in a continued-fraction expansion, describing the ”particle” and ”hole” excitations: $G(\omega_{n})=G^{>}(\omega_{n})+G^{>}(\omega_{n})$ (60) with $G^{\alpha=>,<}(\omega_{n})=\frac{<gs\|c_{\mu}c^{\dagger}_{\mu}\|gs>}{\hbox{i}\omega_{n}-a_{0}^{\alpha}-\frac{b_{1}^{\alpha 2}}{\hbox{i}\omega_{n}-a_{1}^{\alpha}-\frac{b_{2}^{\alpha 2}}{\hbox{i}\omega_{n}-a_{2}^{\alpha}-...}}}$ (61) The parameters $a_{j}^{\alpha}$ and $b_{j}^{\alpha}$ ($j=1...$ number of Lanczos steps, $\alpha=>,<$) output directly from the second Lanczos stephaydock1975 . In the ED-CDMFT procedure the cluster Green’s function and the self-consistency equation 4 are evaluated on the Matsubara axis. This implies introducing a parameter $\beta$ which determines the grid of Matsubara points $\omega_{n}=(2n-1)\pi/\beta$, and which plays the role of a fictitious temperature (our solution is however at zero temperature). We can easily analytically continue the Green’s functions by replacing in the continued- fraction expansion (eq. 61) $\hbox{i}\omega\to\omega+\hbox{i}\eta$ (62) where $\omega$ is the real axis frequency and $\eta$ a small parameter used to display the poles. It is difficult to know a priori the smallest value we can assign to $\eta$. It depends on the physical problem considered, on the size of the associated Anderson Impurity Problem used in the Lanczos procedure and on the energy resolution imposed in satisfying the self-consistency condition on the Matsubara axis (i.e. the parameter $\beta$). A reasonable guess for a lower bound value would be for example $\eta\approx\pi/\beta$. Moreover we expect $\eta$ to be frequency dependent too. The Lanczos better determines the ground state of the system, and portrays better low-energy properties. The uncertainty in the ground-state vector $\|gs>$ propagates in the determination of the $a_{j}^{\alpha}$ and $b_{j}^{\alpha}$ coefficients (which is a further Lanczos step), and further propagates in the periodization procedures (section III). The error in the second Lanczos step turns out much bigger at higher frequency, and, if $\eta$ is chosen too small, this can create enormous errors, even breaking casuality (which is instead by construction always satisfied in the impurity solver output, i.e. in the cluster Green’s function of eq. 6). A too big value of $\eta$ results however in a poor resolution, which may hide important features, above all at small energy. Figure 28: Example of the broadening $\eta(\omega)$ used to display spectral function on the real axis. Left: the low energy range $-1<\omega<1$ example of Fig. 20. Right: the wide energy range $-14<\omega<14$ example of Fig. 19. For a complete list of $\eta(\omega)$ see table A In order to be able to display at the same time high and low energy features, we have introduced for convenience’s sake in sections III (with the only exception of Fig. 7, see the corresponding caption), V, VI and VII a $\omega$ dependent $\eta$, arbitrarily choosing a function-shape which quickly separates a low from a high energy range (see Fig. 28): $\eta(\omega)=\eta_{0}+\frac{\eta_{1}}{e^{-\|\omega-\omega_{0}\|/T_{w}}+1}$ (63) The branching between the low and high energy regimes takes place in a Fermi- function-like step at $\omega=\omega_{0}$ with rapidity $1/T_{w}$. In Fig. 28 we report as example the $\eta(\omega)$ adopted to display the local density of states of Fig. 20 (on the left), where the low-energy range is $-1<\omega<1$ , and (on the right) the $\eta(\omega)$ of the density plots of the Fig. 19 and 21, where the full energy scale $-14<\omega<14$, including the lower and upper Hubbard bands, is considered. In the following we report exactly the values of the parameters determining $\eta(\omega)$ for all the figures involved: Figures \| $\eta_{0}$ \| $\eta_{1}$ \| $\omega_{0}$ \| $T_{w}$ ---\|---\|---\|---\|--- 8, top panel \| 0.01 \| 0.15 \| 0.05 \| 0.01 8, middle and bottom panels \| 0.001 \| 0.05 \| 0.05 \| 0.01 19,21,22,23, 24,25 \| 0.01 \| 0.2 \| 0.4 \| 0.3 20 \| 0.001 \| 0.175 \| 0.2 \| 0.05 26 \| 0.025 \| 0.04 \| 0.2 \| 0.025 It is clear that spectral peaks displayed with different $\eta(\omega)$ at different frequency present uncomparable heights and widths. Our task is not however to make this kind of comparison, rather to focalize on the position of the spectral peaks and compare heights and widths at different doping (but at the same frequency). This has to be kept in mind when analyzing these figures. Figure 29: Color scale code adopted in displaying the color-density plots. $x$ is the maximum value of the scale. Finally, the color-code we adopted to display the density-plots (see e.g. figures 19, 21, 22, 23, 24, 25) is shown in Fig. 29. $x$ is the maximum value of the scale, that we have chosen according to the picture (see the corresponding caption). If the value of the function displayed is bigger than $x$, the color remains red. ## Appendix B Superconducting Bath-parameterization within CDMFT The general form of the associated cluster-Anderson impurity Hamiltonian $\mathcal{H}_{\rm imp}$ can be written: $\displaystyle\mathcal{H}_{\rm imp}$ $\displaystyle=$ $\displaystyle\sum_{\mu\nu\sigma}E_{\mu\nu\sigma}c_{\mu\sigma}^{\dagger}c_{\nu\sigma}+U\sum_{\mu}n_{\mu\uparrow}n_{\mu\downarrow}+$ (64) $\displaystyle+\sum_{k\sigma}\epsilon_{k\sigma}a_{k\sigma}^{\dagger}a_{k\sigma}+\sum_{k\mu\sigma}V_{k\mu\sigma}a_{k\sigma}^{\dagger}c_{\mu\sigma}+{\rm h.c.}+$ $\displaystyle+\sum_{k\mu\sigma}V^{\rm sup}_{k\mu\sigma}a_{k\sigma}^{\dagger}c^{\dagger}_{\mu\bar{\sigma}}+\sum_{k\mu\sigma}V^{\dagger\rm sup}_{k\mu\sigma}c_{\mu\bar{\sigma}}a_{k\sigma}$ Here $\mu,\nu=1,...,N_{c}$ label the sites in the cluster and $E_{\mu\nu\sigma}$ represents the hopping and the chemical potential within the cluster. $\epsilon_{k\sigma}$ is the energy level of the $(k,\sigma)$ orbital in the free electron bath, $V$s represent the hybridization hopping amplitude either for a particle-desctution/construction and for the singlet- destruction/destruction (construction/construction) between the impurity- cluster and the bath. In order to drive the solutions towards physically interesting regions of the bath-parameters space, we have introduced a reduced bath-parametrizationvenky05 which allows to exploit the symmetries in the square lattice to gain a better physical insight of the Green’s function symmetries. Moreover using fewer parameters the work required by the minimization procedure is faster and the result simpler to interpret: $\displaystyle\mathcal{H}^{R}_{\rm imp}$ $\displaystyle=$ $\displaystyle\sum_{\mu\nu\sigma}E_{\mu\nu\sigma}c_{\mu\sigma}^{\dagger}c_{\nu\sigma}+U\sum_{\mu}n_{\mu\uparrow}n_{\mu\downarrow}$ (65) $\displaystyle\sum_{mm^{\prime}\sigma}\epsilon_{mm^{\prime}\sigma}^{\alpha}a_{m^{\prime}\sigma}^{\dagger\alpha}a_{m\sigma}^{\alpha}+\sum_{m\mu\sigma}V_{m\mu\sigma}^{\alpha}a_{m\sigma}^{\dagger\alpha}c_{\mu\sigma}+{\rm h.c.}+$ $\displaystyle+\sum_{\alpha}\Delta^{\alpha}(a_{1\uparrow}^{\alpha}a_{2\downarrow}^{\alpha}-a_{2\uparrow}^{\alpha}a_{3\downarrow}^{\alpha}+a_{3\uparrow}^{\alpha}a_{4\downarrow}^{\alpha}-a_{4\uparrow}^{\alpha}a_{1\downarrow}^{\alpha}$ $\displaystyle+a_{2\uparrow}^{\alpha}a_{1\downarrow}^{\alpha}-a_{3\uparrow}^{\alpha}a_{2\downarrow}^{\alpha}+a_{4\uparrow}^{\alpha}a_{3\downarrow}^{\alpha}-a_{1\uparrow}^{\alpha}a_{4\downarrow}^{\alpha})+h.c.$ The energy levels in the bath are grouped into multiples of the cluster size ($N_{c}=4$) with the labels $m=1,\cdots,N_{c}$ and $\alpha=1,2$ such that we have 8 bath energy levels $\epsilon_{m\sigma}^{\alpha}$ coupled to the cluster via the hybridization matrix $V_{m\mu\sigma}^{\alpha}$. Using lattice symmetries we take $V_{m\mu\sigma}^{\alpha}\equiv V^{\alpha}\delta_{m\mu}$ and $\epsilon^{\alpha}_{m\sigma}\equiv\epsilon^{\alpha}$. $\Delta^{\alpha}$ represents the amplitude of superconducting correlations in the bath. No static mean-field order parameter acts directly on the cluster sites venky05 ; poilblanc . $\epsilon^{\alpha}$, $V^{\alpha}$ and $\Delta^{\alpha}$ are determined by imposing the self-consistency condition in eq. 4 using a conjugate gradient minimization algorithm with a distance function that emphasizes the lowest frequencies of the Weiss fieldmarce (see eq. 16). The reduced form eq. 65 is in fact a sub-case of the more general Hamiltonian eq. 64, and a canonical transformation connects eq. 65 to eq. 64. In order to see this, it is most convenient to express $\mathcal{H}$ in a Nambu’s form, introducing cluster-plaquette spinors: $\displaystyle\Psi_{c}^{\dagger}\equiv(c_{1\uparrow}^{\dagger},\dots,c_{4\uparrow}^{\dagger},c_{1\downarrow},\dots,c_{4\downarrow})$ $\displaystyle\Phi_{a}^{\dagger}\equiv(a_{1\uparrow}^{\dagger},\dots,a_{4\uparrow}^{\dagger},a_{1\downarrow},\dots,a_{4\downarrow})$ (66) and recast the Hamiltonian: $\displaystyle\mathcal{H}_{\rm imp}$ $\displaystyle=$ $\displaystyle\Psi^{\dagger}\,\mathbf{E}\,\Psi+U\sum_{\mu}n_{\mu\uparrow}n_{\mu\downarrow}+$ $\displaystyle+\sum_{\alpha}\,\Phi^{\dagger}_{\alpha}\,\mathbf{E^{\alpha}_{B}}\,\Phi_{\alpha}+\Phi^{\dagger}_{\alpha}\,\mathbf{V}_{\alpha}\,\Psi+\Psi^{\dagger}\,\mathbf{V}^{\dagger}_{\alpha}\,\Phi_{\alpha}$ where the Hamiltonian coupling constants are expressed my 8X8 matrices (leaving for convenience’s sake the multi-bath index $\alpha$ implicit): $\mathbf{E}=\,\left(\begin{array}[]{cc}E_{\mu,\nu,\uparrow}(4{\rm X}4)&0\\\ 0&E_{\mu,\nu,\downarrow}(4{\rm X}4)\end{array}\right)$ $\mathbf{E_{B}}=\,\left(\begin{array}[]{cc}\mathbf{E_{B}}_{\uparrow}(4{\rm X}4)&\mathbf{E^{S}_{B}}(4{\rm X}4)\\\ \mathbf{E^{S\dagger}_{B}}(4{\rm X}4)&-\mathbf{E_{B}}_{\downarrow}(4{\rm X}4)\end{array}\right)$ and $\mathbf{V}=\,\left(\begin{array}[]{cc}\mathbf{V}_{\uparrow}(4{\rm X}4)&\mathbf{V}^{S}_{\uparrow}(4{\rm X}4)\\\ -\mathbf{V}^{S\dagger}_{\uparrow}(4{\rm X}4)&-\mathbf{V}_{\downarrow}(4{\rm X}4)\end{array}\right)$ In the more general case $\mathcal{H}_{imp}^{\ref{Ham-imp-apx}}$, the bath matrix is diagonal, so that $\mathbf{E_{B\sigma}^{kk^{\prime}}}^{\ref{Ham-imp- apx}}=\delta_{kk^{\prime}}\varepsilon_{k\sigma}$, and $\mathbf{E^{S}_{B}}^{\ref{Ham-imp-apx}}=0$. $\mathbf{V}_{\sigma}$ and $\mathbf{V}_{\sigma}^{S}$ have generally non-zero elements. In the reduced parameterization case instead (eq. 65), the bath matrix is not diagonal but chosen to mimic a cluster-plaquette, by introducing in each multi-bath $\alpha$ the same bath-energy $\varepsilon_{\alpha\sigma}$ on every bath-site, a next-neighbor hopping $t^{\alpha}_{b}$ and a nearest next-neighbor hopping $t^{\prime\alpha}_{b}$: $\mathbf{E^{\alpha}_{B}}^{\ref{Ham-imp- red}}_{\sigma}=\,\left(\begin{array}[]{cccc}\varepsilon_{\alpha\sigma}&t^{\alpha}_{b}&t^{\prime\alpha}_{b}&t^{\alpha}_{b}\\\ t^{\alpha}_{b}&\varepsilon_{\alpha\sigma}&t^{\alpha}_{b}&t^{\prime\alpha}_{b}\\\ t^{\prime\alpha}_{b}&t^{\alpha}_{b}&\varepsilon_{\alpha\sigma}&t^{\alpha}_{b}\\\ t^{\alpha}_{b}&t^{\prime\alpha}_{b}&t^{\alpha}_{b}&\varepsilon_{\alpha\sigma}\end{array}\right)$ Moreover superconductive d-wave pairing terms $\Delta_{\alpha}$ appear in the bath: $\mathbf{E^{S\alpha}_{B}}^{\ref{Ham-imp- red}}=\,\left(\begin{array}[]{cccc}0&\Delta_{\alpha}&0&-\Delta_{\alpha}\\\ \Delta_{\alpha}&0&-\Delta_{\alpha}&0\\\ 0&-\Delta_{\alpha}&0&\Delta_{\alpha}\\\ -\Delta_{\alpha}&0&\Delta_{\alpha}&0\end{array}\right)$ while the hybridization between the cluster and the bath is only normal $\mathbf{V}_{\sigma}^{S}=0$ and simplified by connecting each site of the cluster $\mu$ with only one site $k_{\mu}^{\alpha}$ of the multi-bath $\alpha$ with the same coupling constant $V_{\sigma\alpha}$: $\mathbf{V}_{\sigma\alpha}^{\ref{Ham-imp- red}}=\,\left(\begin{array}[]{cccc}V_{\sigma\alpha}&0&0&0\\\ 0&V_{\sigma\alpha}&0&0\\\ 0&0&V_{\sigma\alpha}&0\\\ 0&0&0&V_{\sigma\alpha}\end{array}\right)$ To connect the reduced parameterization eq. 65 to the general eq. 64 it is sufficient to diagonalize the hermitian bath-matrix $\mathbf{E^{\alpha}_{B}}$ via a unitary transformationcivelli-2007 : $\displaystyle\mathbf{E^{\alpha}_{B}}^{\ref{Ham-imp- apx}}=\,\mathbf{S_{\alpha}^{\dagger}}\,\mathbf{E^{\alpha}_{B}}^{\ref{Ham-imp- red}}\,\mathbf{S_{\alpha}}$ (68) and $\mathbf{S_{\alpha}^{T}}\,\mathbf{S_{\alpha}}=\mathbf{1}$. This is in fact the requirement needed to have a canonical transformation which preserves the fermionic commutation-relations. In fact if we apply the transformation to the vectors: $\left\\{\begin{array}[]{c}\mathbf{\Phi_{\alpha}}^{\ref{Ham-imp- red}}=\,\mathbf{S_{\alpha}}\,\mathbf{\Phi_{\alpha}}^{\ref{Ham-imp-apx}}\\\ \mathbf{\Phi^{\dagger}_{\alpha}}^{\ref{Ham-imp- red}}=\,\mathbf{\Phi^{\dagger}_{\alpha}}^{\ref{Ham-imp- apx}}\,\mathbf{S^{\dagger}_{\alpha}}\end{array}\right.$ (69) and we have $\left\\{\mathbf{\Phi_{\alpha}}^{\ref{Ham-imp- red}},\mathbf{\Phi^{\dagger}_{\alpha}}^{\ref{Ham-imp- red}}\right\\}=\,\mathbf{1}$ (70) it is $\displaystyle\left\\{\Phi_{q\alpha}^{\ref{Ham-imp- red}},\Phi^{\dagger\ref{Ham-imp-red}}_{q\alpha}\right\\}=\,$ (71) $\displaystyle\sum_{kk^{\prime}}\,S^{\dagger}_{qk}S_{k^{\prime}q}\,\left\\{\Phi_{k\alpha}^{\ref{Ham- imp-red}},\Phi^{\dagger\ref{Ham-imp-red}}_{k^{\prime}\alpha}\right\\}=\,$ (72) $\displaystyle\sum_{k}\,S^{\dagger}_{qk}S_{kq}=\,1$ (73) Finally $\mathbf{V}_{\alpha}$ is subjected to the same transformation: $\mathbf{V}_{\alpha}^{\ref{Ham-imp- apx}}=\,\mathbf{S}^{\dagger}_{\alpha}\,\mathbf{V}_{\alpha}^{\ref{Ham-imp- red}}$ (74) ###### Acknowledgements. We thank G. Kotliar, who inspired many of the ideas in this work. We thank O. Parcollet and K. Haule for sharing their QMC results. We also acknowledge the fruitful discussion with A. Georges, M. Capone and T. D. Stanescu. We enjoyed the exchange of ideas and the hospitality of A. Sacuto and the SQUAP group at the University of Paris 7. We acknowledge A.-M. S. Tremblay, S. S. Kancharla, B. Kyung, I. Paul, A. Cano, E. Kats and P. Nozières for their useful comments. ## References * (1) J. G. Bednorz and K. A. Muller, Z. Phys. B 64 189 (1986). * (2) Y. Kamihara, T. Watanabe, M. Hirano and H. Hosono, J. Am. Chem. Soc. 130 3296 (2008). * (3) C. Day, Physics Today 61 11-12 (2008). * (4) P. W. Anderson, Science 235 1196 (1987). * (5) A. Georges, G. Kotliar, W. Krauth and M. J. Rozenberg, Rev. Mod. Phys. 68 13 (1996). * (6) N. E. Zein, S. Y. Savrasov and G. Kotliar, Phys. Rev. Lett. 96 226403 (2006). * (7) ”Ruthenate and Rutheno-Cuprate Materials: Unconventional Superconductivity, Magnetism and Quantum Phase Transitions” A. Lichteinstein and A. Liebsch, Springer-Verlag, Berlin, Germany (2002). * (8) K. Held, A. K. McMahan and R. T. Scalettar, Phys. Rev. Lett. 87 276404 (2001). * (9) K. Haule, V. Oudovenko, S. Y. Savrasov and G. Kotliar, Phys. Rev. Lett. 94 036401 (2005). * (10) S. Savrasov, G. Kotliar and E. Abrahams, Nature 410 793 (2001). * (11) G. Kotliar, S. Savrasov, K. Haule, V. Oudovenko, O. Parcollet and C. Marianetti, Rev. Mod. Phys. 78 000865 (2006). * (12) A. Damascelli, Z. X. Shen and Z. Hussain, Rev. Mod. Phys. 75 473 (2003). * (13) ”Physics of Superconductors II” J. C. Campuzano, M. R Norman and M. Randeria, K. H. Bennemann and J. B. Ketterson (2004) 167-273. * (14) Th. Maier, M. Jarrell, Th. Pruschke and M. Hettler, Rev. Mod. Phys. 77 1027-1080 (2005). * (15) G. Biroli, O. Parcollet and G. Kotliar, Phys. Rev. B 69 205108 (2004). * (16) A. I. Lichtenstein and M. I. Katsnelson, Phys. Rev. B 62 R9283 (2000). * (17) Th. Maier, M. Jarrell, Th. Pruschke and J. Keller, Phys. Rev. Lett. 85 1524 (2000). * (18) A.-M. S. Tremblay, B. Kyung and D. Sénéchal, Low Temperature Physics 32 424 (2006). * (19) G. Kotliar, S. Y. Savrasov, G. Palsson and G. Biroli, Phys. Rev. Lett. 87 186401 (2001). * (20) R. M. Fye and J. E. Hirsch, Phys. Rev. Lett. 56 2521 (1986). * (21) R. M. Fye and J. E. Hirsch, Phys. Rev. B 40 47804796 (1989). * (22) T. A. Maier, M. Jarrell, T. C. Schulthess, P. R. C. Kent and J. B. White, Phys. Rev. Lett. 95 237001 (2005). * (23) T. A. Maier, M. Jarrell and D. J. Scalapino, Phys. Rev. Lett. 96 047005 (2006). * (24) D. Sénéchal and A.-M. S. Tremblay, Phys. Rev. Lett. 92 126401 (2004). * (25) O. Parcollet, G. Biroli and G. Kotliar, Phys. Rev. Lett. 92 226402 (2004). * (26) M. Civelli, M. Capone, S. S. Kancharla, O. Parcollet and G. Kotliar, Phys. Rev. Lett. 95 106402 (2005). * (27) B. Kyung, S. S. Kancharla, D. Sénéchal, A.-M. S. Tremblay, M. Civelli and G. Kotliar, Phys. Rev. B 73 165114 (2006). * (28) S. S. Kancharla, B. Kyung, D. Sénéchal, M. Civelli, M. Capone, G. Kotliar and A.-M. S. Tremblay, Phys. Rev. B 77 184516 (2008). * (29) M. Capone and G. Kotliar, Phys. Rev. B 74 054513 (2006). * (30) M. Aichhorn, E. Arrigoni, M. Potthoff and W. Hanke, Phys. Rev. B 74 235117 (2006). * (31) M. Aichhorn, E. Arrigoni, M. Potthoff and W. Hanke, Phys. Rev. B 74 024508 (2006). * (32) D. Sénéchal, P.-L. Lavertu, M.-A. Marois and A.-M. S. Tremblay, Phys. Rev. Lett. 94 156404 (2005). * (33) C. Huscroft, M. Jarrell, Th. Maier, S. Moukouri and A. N. Tahvildarzadeh, Phys. Rev. Lett. 86 139 (2001). * (34) T. D. Stanescu and P. Phillips, Phys. Rev. Lett. 91 017002 (2003). * (35) F. H. L. Essler and A. M. Tsvelik, Phys. Rev. B 65 115117 (2002). * (36) T. D. Stanescu and G. Kotliar, Phys. Rev. B 74 125110 (2006). * (37) C. Berthod, T. Giamarchi, S. Biermann and A. Georges, Phys. Rev. Lett. 97 136401 (2006). * (38) T. D. Stanescu, P. W. Phillips and T.-P. Choy, Phys. Rev. B 75 104503 (2007). * (39) K. Haule and G. Kotliar, Phys. Rev. B 76 092503 (2007). * (40) M. Caffarel and W. Krauth, Phys. Rev. Lett. 72 1545 (1994). * (41) K. Haule and G. Kotliar, Phys. Rev. B 76 104509 (2007). * (42) M. Capone, M. Civelli, S. S. Kancharla, C. Castellani and G. Kotliar, Phys. Rev. B 69 195105 (2004). * (43) M. Civelli, PhD Thesis, arXiv.org:0710.2802 (2007). * (44) M. Civelli, M. Capone, A. Georges, K. Haule, O. Parcollet, T. D. Stanescu and G. Kotliar, Phys. Rev. Lett. 100 046402 (2008). * (45) P. A. Lee, N. Nagaosa and X.-G. Wen, Rev. Mod. Phys. 78 17 (2006). * (46) G. Biroli and G. Kotliar, Phys. Rev. B 65 155112 (2002). * (47) T. D. Stanescu, M. Civelli, K. Haule and G. Kotliar, An. of Phys. 321 1682 (2006). * (48) A. Perali, M. Sindel and G. Kotliar, Eur. Phys. J. B 24 487 (2002). * (49) J. Hubbard, Proc. Roy. Soc. A 276 238 (1963). * (50) C. J. Bolech, S. S. Kancharla and G. Kotliar, Phys. Rev. B 67 075110 (2003). * (51) K. Haule, Phys. Rev. B 75 155113 (2007). * (52) G. Kotliar and J. Liu, Phys. Rev. B 38 R5142 (1988). * (53) K. McElroy, D.-H. Lee, J. E. Hoffman, K. M. Lang, J. Lee, E. W. Hudson, H. Eisaki, S. Uchida and J. C. Davis, Phys. Rev. Lett. 94 197005 (2005). * (54) M. Le Tacon, A. Sacuto, A. Georges, G. Kotliar, Y. Gallais, D. Colson and A. Forget, Natur. Phys. 2 537 (2006). * (55) K. Tanaka, W. S. Lee, D. H. Lu, A. Fujimori, T. Fujii, Risdiana, I. Terasaki, D. J. Scalapino, T. P. Devereaux, Z. Hussain and Z.-X. Shen, Science 314 1910 (2006). * (56) A. Cho, Science 314 1072 (2006). * (57) A. J. Millis, Science 314 1888 (2006). * (58) G. Deutscher, Nature 397 410 (1999). * (59) S. Huefner, M. A. Hossain, A. Damascelli and G. A. Sawatzky, Rep. Prog. Phys. 71 062501 (2008). * (60) T. Kondo, T. Takeuchi, A. Kaminski, S. Tsuda and S. Shin, Phys. Rev. Lett. 98 267004 (2007). * (61) K. K. Gomes, A. N. Pasupathy, A. Pushp, S. Ono, Y. Ando and A. Yazdani, Nature 447 569 (2007). * (62) L. B. Ioffe and A. J. Millis, Phys. Rev. B 58 11631 (1998). * (63) A. T. Zheleznyak, V. M. Yakovenko, H. D. Drew and I. I. Mazin, Phys. Rev. B 57 3089 (1998). * (64) R. Hlubina and T. M. Rice, Phys. Rev. B 51 9253 (1995). * (65) O. Parcollet, private communication . * (66) H. Matsui, T. Sato, T. Takahahi, S.-C. Wang, H.-B. Yang, H. Ding, T. Fujii, T. Watanabe and A. Matsuda, Phys. Rev. Lett. 90 217002 (2003). * (67) J. M. Luttinger, Phys. Rev. 119 1153 (1960). * (68) L. De Leo, M. Civelli and G. Kotliar, arXiv.org:0804.3314 (2008). * (69) M. Ferrero, F. Becca, M. Fabrizio and M. Capone, Phys. Rev. B 72 205126 (2005). * (70) L. de’ Medici, A. Georges and S. Biermann, Phys. Rev. B 72 205124 (2005). * (71) X. J. Zhou, T. Yoshida, A. Lanzara, P. V. Bogdanov, S. A. Kellar, K. M. Shen, W. L. Yang, F. Ronning, T. Sasagawa, T. Kakeshita, T. Noda, H. Eisaki, S. Uchida, C. T. Lin, F. Zhou, J. W. Xiong, W. X. Ti, Z. X. Zhao, A. Fujimori, Z. Hussain and Z.-X. Shen, Nature 423 398 (2003). * (72) D. A. Bonn, Czech. J. Phys. 46 3195 (1996). * (73) C. Panagopoulos and T. Xiang, Phys. Rev. Lett. 81 2336 (1998). * (74) B. Kyung and A.-M. S. Tremblay, cond-mat/0204500 (2002). * (75) Kai-Yu Yang, T. M. Rice and Fu-Chun Zhang, Phys. Rev. B 73 174501 (2006). * (76) B. Valenzuela and E. Bascones, Phys. Rev. Lett. 98 227002 (2007). * (77) M. Aichhorn, E. Arrigoni, Z. B. Huang and W. Hanke, Phys. Rev. Lett. 99 257002 (2007). * (78) W. Guyard, A. Sacuto, M. Cazayous, Y. Gallais, M. Le Tacon, D. Colson and A. Forget, arXiv.org:0802.3166 (2008). * (79) N. Doiron-Leyraud, C. Proust, D. LeBoeuf, J. Levallois, J.-B. Bonnemaison, R. Liang, D.A. Bonn, W.N. Hardy and L. Taillefer, Nature 447 565 (2007). * (80) R. Daou, D. LeBoeuf, N. Doiron-Leyraud, S. Y. Li, F. Laliberte, O. Cyr-Choiniere, Y. J. Jo, L. Balicas, J. -Q. Yan, J. -S. Zhou, J. B. Goodenough and L. Taillefer, arXiv.org:0806.2881 (2008). * (81) F. F. Balakirev, J. B. Betts, A. Migliori, I. Tsukada, Yoichi Ando and G. S. Boebinger, arXiv.org:0710.4612 (2007). * (82) F. F. Balakirev, J. B. Betts, A. Migliori, S. Ono, Yoichi Ando and G. S. Boebinger, Nature 424 912 (2003). * (83) N. M. Plakida and V. S. Oudovenko, JETP 104 230 (2007). * (84) A. Avella and F. Mancini, Phys. Rev. B 75 134518 (2007). * (85) I. E. Dzyaloshinskii, Phys. Rev. B 68 085113 (2003). * (86) F. H. L. Essler and A. M. Tsvelik, Phys. Rev. Lett. 90 126401 (2003). * (87) R. M. Konik, T. M. Rice and A. M. Tsvelik, arXiv.org:cond-mat/0511268 (2005). * (88) A. Rosch, Eur. Phys. Jour. B 59 495 (2007). * (89) A. Lanzara, P. V. Bogdanov, X. J. Zhou, S. A. Kellar, D. L. Feng, E. D. Lu, T. Yoshida, H. Eisaki, A. Fujimori, K. Kishio, J.-I. Shimoyama, T. Noda, S. Uchida, Z. Hussain and Z.-X. Shen, Nature 412 510 (2001). * (90) Z.-X. Shen, A. Lanzara, S. Ishihara and N. Nagaosa, Philos. Mag. B 82 1349 (2002). * (91) H. He, Y. Sidis, P. Bourges, G. D. Gu, A. Ivanov, N. Koshizuka, B. Liang, C. T. Lin, L. P. Regnault, E. Schoenherr and B. Keimer, Phys. Rev. Lett. 86 1610 (2001). * (92) J. Hwang, T. Timusk and G. D. Gu, Nature 427 714 (2004). * (93) T. Cuk, F. Baumberger, D. H. Lu, N. Ingle, X. J. Zhou, H. Eisaki, N. Kaneko, Z. Hussain, T. P. Devereaux, N. Nagaosa and Z.-X. Shen, Phys. Rev. Lett. 94 117003 (2004). * (94) A. Kaminski, M. Randeira, J. C. Campuzano, M. R. Norman, H. Fretwell, J. Mesot, T. Sato, T. Takahasci and K. Kadowaki, Phys. Rev. Lett. 86 1070 (2001). * (95) K. Byczuk, M. Kollar, K. Held, Y. -F. Yang, I. A. Nekrasov, Th. Pruschke and D. Vollhardt, Nature Physics 3 168 (2007). * (96) S. Chakraborty, D. Galanakis and P. Phillips, arXiv.org:0712.2838 (2007). * (97) J. Graf, G.-H. Gweon, K. McElroy, S. Y. Zhou, C. Jozwiak, E. Rotenberg, A. Bill, T. Sasagawa, H. Eisaki, S. Uchida, H. Takagi, D.-H. Lee and A. Lanzara, Phys. Rev. Lett. 98 067004 (2007). * (98) T. Valla, T. E. Kidd, Z.-H. Pan, A. V. Fedorov, W.-G. Yin, G. D. Gu and P. D. Johnson, Phys. Rev. Lett. 98 167003 (2007). * (99) M. Z. Cieplak, A. Abal’oshev, I. Zaytseva, M. Berkowski, S. Guha and Q. Wu, Acta Physica Polonica A 109 573 (2006). * (100) G. Baskaran, Z. Zou and P. W. Anderson, Solid State Com. 63 973 (1987). * (101) A. E. Ruckenstein, P. J. Hirschfeld and J. Appel, Phys. Rev. B 36 857 (1987). * (102) T. D. Stanescu and G. Kotliar, Phys. Rev. B 70 205112 (2004). * (103) M. Shi, J. Chang, S. Pailhés, M. R. Norman, J. C. Campuzano, M. Mansson, T. Claesson, O. Tjernberg, A. Bendounan, L. Patthey, N. Momono, M. Oda, M. Ido, C. Mudry and J. Mesot, Phys. Rev. Lett. 101 047002 (2007). * (104) A. Kanigel, U. Chatterjee, M. Randeria, M. R. Norman, S. Souma, M. Shi, Z. Z. Li, H. Raffy and J. C. Campuzano, Phys. Rev. Lett. 99 157001 (2007). * (105) A. Kanigel, U. Chatterjee, M. Randeria, M. R. Norman, G. Koren, K. Kadowaki and J. C. Campuzano, arXiv:0803.3052v1 (2008). * (106) S. Sachdev and J. Ye, Phys. Rev. Lett. 69 2411 (1992). * (107) A. Perali, C. Castellani, C. Di Castro and M. Grilli, Phys. Rev. B 54 16216–16225 (1996). * (108) S. A. Kivelson, E. Fradkin and V. J. Emery, Nature 393 550 (1998). * (109) C. M. Varma, Phys. Rev. Lett. 83 3538 (1999). * (110) Yoichi Ando, G. S. Boebinger, A. Passner, T. Kimura and K. Kishio, Phys. Rev. Lett. 75 4662–4665 (1995). * (111) G. S. Boebinger, Yoichi Ando, A. Passner, T. Kimura, M. Okuya, J. Shimoyama, K. Kishio, K. Tamasaku, N. Ichikawa and S. Uchida, Phys. Rev. Lett. 77 5417–5420 (1996). * (112) M. Ferrero, P. S. Cornaglia, L. De Leo, O. Parcollet, G. Kotliar and A. Georges, arXiv.org:0806.4383 (2008). * (113) R. Haydock, V. Heine and M. J. Kelly, J. Phys. C 8 2591 (1975). * (114) D. Poilblanc and D. J. Scalapino, Phys. Rev. B 66 052513 (2002) .	arxiv-papers	2008-08-28T16:08:49	2024-09-04T02:48:57.516203	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "M. Civelli", "submitter": "Marcello Civelli", "url": "https://arxiv.org/abs/0808.3948" }
0808.3956	# Resonance and limit cycle in a noise driven Lorenz model Himadri S. Samanta h.s.samanta@sheffield.ac.uk Department of Physics and Astronomy, University of Sheffield, Sheffield, S3 7RH, UK J. K. Bhattacharjee jkb@bose.res.in S.N.Bose National Centre For Basic Sciences, JD-Block,Sector- III, Salt Lake City, Kolkata-700098, India ###### Abstract The effect of an external noise on the Lorenz model is investigated near the onset of convection and near the Hopf bifurcation. We show the existence of a diverging time scale near the onset of convection and a resonance near the Hopf bifurcation. Our calculation provides an understanding of the noise induced stabilization of the limit cycle that had been observed numerically. ###### pacs: 05.45.-a The noise induced phenomena have an intense interest in several disciplines ranging from physics, to chemistry, and to biology 9 . Physical systems are usually not isolated from their environment, and the environmental influences on the system often appear as fluctuations. It is clear that the fluctuations are an integral part of the evolution of physical, chemical and biological systems and must be understood if we are to accurately and quantitatively describe the world around us, particularly at the small scale. In the past few years, it has become clear that fluctuations can actually be used constructively, by us or by nature, to produce organised behavior that is not possible in the absence of noise. Examples where noise leads to organised behavior include stochastic resonance 1 ; 2 , noise induced phase transitions 3 ; 4 , noise induced pattern formation 5 ; 6 , and noise induced transport 8 ; 9 . The constructive role of noise is only possible in the non-linear non- equilibrium systems and is entirely the result of the intricate interplay of noise and non-linearity away from equilibrium. The effect of noise on dynamical systems include noise induced hopping between multiple stable attractors10 ; 11 and noise induced stabilization of the Lorenz attractor 7 ; omar near the threshold of its formation. We will focus on the second aspect here and as an example of the effect of noise on hydrodynamic instability. We consider the Lorenz model with an external noise source. Zippelius et. al.luke first studied the correlations in this model. We will supplement their work by focusing specially on the situations where the control parameter is in the vicinity of an instability. This will lead to a noise induced resonance near the Hopf bifurcation point. We then focus on the effect that the noise can have on the limit cycle beyond the Hopf bifurcation point. Ordinarily the limit cycle is unstable, the numerical work of Gao et. al. 7 shows it can be stabilized by noise and here we explicitly show how such a stabilization is possible by drawing on the technique of Mclaughlin and Martinmartin . We study the statistical properties of the following noise driven Lorenz model. $\displaystyle\dot{x}$ $\displaystyle=$ $\displaystyle\sigma(y-x)+\eta_{1}(t)$ (1) $\displaystyle\dot{y}$ $\displaystyle=$ $\displaystyle rx-y- xz+\eta_{2}(t)$ $\displaystyle\dot{z}$ $\displaystyle=$ $\displaystyle- bz+xy+\eta_{3}(t)$ where, $\sigma$ is the Prandtl number, $b$ is a geometric factor, and the Rayleigh number $r$ is the control parameter in this model. The parameter $\sigma=10$ and $b=8/3$ are held fixed at their standard values, as was originally done by Lorenz. $\eta_{i}(t)$ is an external white noise source with zero mean and $<\eta_{i}(t)\eta_{j}(t^{{}^{\prime}})>=\delta_{ij}2D\delta(t-t^{{}^{\prime}})$, $i,j=1,2,3$. $D$ is the noise strength. $D=0$ describes the unforced Lorenz model. When the noise is absent the Lorenz system lorenz ; row shows remarkable change in behavior depending on the control parameter $r$. In the conduction range, $r<1$, the trivial steady state solution is stable and loses stability to the other two describing steady convection through a bifurcation at $r=1$. Thus, for $r>1$, there is a pair of stable fixed points, $(\pm\sqrt{b(r-1)},\pm\sqrt{b(r-1),r-1})$ and these in turn lose stability at $r=r_{T}=\sigma(\sigma+b+3)/(\sigma-b-1)=24.74$ through a Hopf-bifurcation. For $r>r_{T}$, no stable steady state solution exists and the system has a strange attractor. Depending on initial conditions, the solution may settle down on any of these three attractors. The trajectories are non-periodic and wander around in the vicinity of strange set of attracting points for $r>r_{T}$ and turbulence sets in. Now we study the effect of noise on the Lorenz model in the conduction, convection and turbulent regime. We calculate the time dependent correlation functions and compare with the behavior of the unforced model. In frequency space equation (1) reads $\displaystyle(-i\omega+\sigma)x(\omega)-\sigma y(\omega)=\eta_{1}(\omega)$ (2) $\displaystyle- rx(\omega)+(-i\omega+1)y(\omega)+\sum_{\omega_{1}}x(\omega_{1})z(\omega-\omega_{1})=\eta_{2}(\omega)$ $\displaystyle(-i\omega+b)z(\omega)-\sum_{\omega_{2}}x(\omega_{2})y(\omega-\omega_{2})=\eta_{3}(\omega)$ Linearizing around the steady state $x=y=z=0$ i.e. neglecting the non-linear terms in (2), we can calculate the correlation function by solving the linearized equation. The correlation of the $x$ variable reads $\displaystyle C_{xx}(\omega)$ $\displaystyle=$ $\displaystyle<x(\omega)x(-\omega)>$ $\displaystyle=$ $\displaystyle\frac{(\omega^{2}+1)+\sigma^{2}}{([\omega^{2}+(r-1)\sigma]^{2}+(\sigma+1)^{2}\omega^{2})}$ The time dependent correlation function $C_{xx}(t)$ can be written as $C_{xx}(t)=\int_{-\infty}^{\infty}d\omega e^{-i\omega t}C_{xx}(\omega)$ (4) and is the sum of two exponential function with two inverse relaxation times which are independent of the strength $D$ of the fluctuating force. Clearly, the actual relaxation time of the non-linear system (1) does depend on the strength of the fluctuating force. The correlation function of $x$ gets damped out faster with increasing noise strength $D$. The correlation functions $C_{yy}(t)$ and $C_{zz}(t)$ follow the same behavior in the regime $r<1$. Near, $r=1$, the correlation time goes as $\tau\propto\frac{1}{r-1}$ (5) i.e. the relaxation time becomes infinitely big as $r=1$ is approached- a sign of critical slowing down. Now, for $r>1$, Fourier analysis of the Lorenz model breaks down in its present form. This is very similar to what happens when one enters a symmetry breaking phase in critical phenomena. Accordingly we need to go to the shifted variable $u_{1},u_{2},u_{3}$ defined as $u_{1}=x-x_{0}$, $u_{2}=y-y_{0}$ and $u_{3}=z-z_{0}$. The Lorenz equation takes the following form $\displaystyle(-i\omega+\sigma)u_{1}(\omega)-\sigma u_{2}(\omega)=\eta_{1}(\omega)$ (6) $\displaystyle- u_{1}(\omega)+(-i\omega+1)u_{2}(\omega)+x_{0}u_{3}(\omega)=\eta_{2}(\omega)$ $\displaystyle- y_{0}u_{1}(\omega)-x_{0}u_{2}(\omega)+(-i\omega+b)u_{3}(\omega)=\eta_{3}(\omega)$ Now, the correlation function takes the following form $\displaystyle<u_{1}(\omega)u_{1}(-\omega)>=$ (7) $\displaystyle\frac{(-\omega^{2}+br)^{2}+(b+1)^{2}\omega^{2}+\sigma^{2}(b^{2}+\omega^{2})+\sigma^{2}b(r-1)}{\omega^{2}[\omega^{2}-b(r+\sigma)]^{2}+[\omega^{2}(\sigma+1+b)-2\sigma b(r-1)]^{2}}$ The time dependent correlation function can be calculated from $C_{u_{1}u_{1}}(t)=\int_{-\infty}^{\infty}d\omega e^{-i\omega t}C_{u_{1}u_{1}}(\omega)$. Their time dependence is determined by the three poles in the complex frequency plane. One is purely imaginary reflecting the exponential decay in the correlation function, the other two have finite real parts of the opposite sign, reflecting spiral motion around one of the attractors ($\pm x_{0},\pm y_{0},z_{0}$). This gives rise to the oscillatory behavior of $C_{u_{1}u_{1}}(t)$ superimposed upon the exponential decay of the correlations caused by the crossing between stable fixed points. In the absence of random force, the system is attracted in general to one of the stable fixed points ($\pm x_{0},\pm y_{0},z_{0}$), depending on its initial condition. If the random force is applied, the trajectories are no longer confined to one of the steady state points. The exponential decay shows the motion from one fixed point to the other. The oscillatory motion slows down the decay of correlation function i.e. the memory effect of the initial state. At the Hopf-bifurcation point, i.e. $r=\frac{\sigma(b+\sigma+3)}{\sigma-b-1}$, the correlation function takes the following form $\displaystyle<u_{1}(\omega)u_{1}(-\omega)>=$ (8) $\displaystyle\frac{(-\omega^{2}+br)^{2}+(b+1)^{2}\omega^{2}+\sigma^{2}(b^{2}+\omega^{2})+\sigma^{2}b(r-1)}{[\omega^{2}-\omega_{0}^{2})]^{2}[\omega^{2}+(\sigma+1+b)^{2})]}$ where, $\omega_{0}=b(\sigma+\frac{\sigma(b+\sigma+3)}{\sigma-b-1})$. Now, the real time correlation function $C_{u_{1}u_{1}}(t)=\int_{-\infty}^{\infty}d\omega e^{-i\omega t}C_{u_{1}u_{1}}(\omega)$ exists as a principle value. Hence, the real time correlation function goes as $C_{u_{1}u_{1}}(t)\propto Re\ te^{i\omega_{0}t}$ (9) This is a consequence of the fluctuating force. It is similar in appearance to the resonance in a simple harmonic oscillator subjected to a sinusoidal force. Approach to this resonance is of the following $<u_{1}u_{1}>=\lim_{\epsilon->0}\frac{1}{\epsilon}e^{i\omega t}$ (10) when, $r=\frac{\sigma(b+\sigma+3)}{\sigma-b-1}-\epsilon$. In the absence of noise there is no periodic state above the Hopf bifurcation point. This is because the limit cycle is unstable.In this case however,a periodic state was observed when the stochastic force was turned on. We will try to understand this on the basis of perturbation theory in terms of small noise strength $\epsilon$. We return to Eq.(8). Writing $r=r_{0}+\Delta r\epsilon$, $x_{0}=x_{00}+\tilde{\epsilon}$, where $x_{00}=\sqrt{b(r-1)}$, $\tilde{\epsilon}=\frac{\epsilon\Delta r}{2}\sqrt{\frac{b}{r-1}}$, we find in real time $L\left(\begin{array}[]{c}u_{1}\\\ u_{2}\\\ u_{3}\end{array}\right)=\left(\begin{array}[]{c}\eta_{1}\\\ \eta_{2}\\\ \eta_{3}\end{array}\right)+\left(\begin{array}[]{c}0\\\ -u_{1}u_{3}\\\ u_{1}u_{2}\end{array}\right)+\tilde{\epsilon}\left(\begin{array}[]{c}0\\\ -u_{3}\\\ u_{1}+u_{2}\end{array}\right)$ (11) where the operator $L$ is given by $L=\left(\begin{array}[]{c}\begin{array}[]{cc}\begin{array}[]{ccc}\frac{\partial}{\partial t}+\sigma&-\sigma&0\\\ -1&\frac{\partial}{\partial t}+1&x_{00}\\\ -x_{00}&-x_{00}&\frac{\partial}{\partial t}+b\end{array}\end{array}\end{array}\right)$ (12) Consistent ordering in powers of $\epsilon$, leads to the expansion $\left(\begin{array}[]{c}u_{1}\\\ u_{2}\\\ u_{3}\end{array}\right)=\epsilon^{1/2}\left(\begin{array}[]{c}u_{10}\\\ u_{20}\\\ u_{30}\end{array}\right)+\epsilon\left(\begin{array}[]{c}u_{11}\\\ u_{21}\\\ u_{31}\end{array}\right)+\epsilon^{3/2}\left(\begin{array}[]{c}u_{12}\\\ u_{22}\\\ u_{32}\end{array}\right)$ (13) The zeroth order solution is found from $L\left(\begin{array}[]{c}u_{10}\\\ u_{20}\\\ u_{30}\end{array}\right)=\left(\begin{array}[]{c}\eta_{1}\\\ \eta_{2}\\\ \eta_{3}\end{array}\right)$ (14) as $u_{i0}=ReA_{i}e^{iwt}+\int G_{ij}(t-t^{\prime})\eta_{j}(t^{\prime})dt^{\prime\prime}$ (15) Figure 1: Time series for the Lorenz equations with r=24.72, D= 0.001 as obtained by Gao et. al. where, $A_{1}=A$, $A_{2}=\frac{A}{\sigma}(\sigma-iw)$, $A_{3}=\frac{A}{x_{00}}[1-\frac{(1-iw_{0})(\sigma-iw_{0})}{\sigma}]$ and $G_{ij}$ are found in frequency space from the particular integral of Eq.(14). At $\mathcal{O}(\epsilon)$, we have $L\left(\begin{array}[]{c}u_{11}\\\ u_{21}\\\ u_{31}\end{array}\right)=\left(\begin{array}[]{c}0\\\ -u_{10}u_{30}\\\ u_{10}u_{20}\end{array}\right)$ (16) and the driving term in the above equation has contributions which are; (i) a time independent term, (ii) a time periodic term with frequency $2w_{0}$, (iii) a term which is a product of a periodic term of frequency $w_{0}$ and a stochastic term $\eta$ and (iv) a term which is a product of two stochastic terms. The solutions for $u_{i1}$ is a contribution of these driving terms with suitable coefficients. Having obtained $u_{i1}$, we can now go to $\mathcal{O}(\epsilon^{3/2})$ and we have $L\left(\begin{array}[]{c}u_{12}\\\ u_{22}\\\ u_{32}\end{array}\right)=\left(\begin{array}[]{c}0\\\ -u_{11}u_{30}-u_{10}u_{31}\\\ u_{10}u_{21}+u_{11}u_{20}\end{array}\right)+\frac{\Delta r}{2}\sqrt{\frac{b}{r-1}}\left(\begin{array}[]{c}0\\\ -u_{30}\\\ u_{10}+u_{20}\end{array}\right)$ (17) The driving terms on the right hand side need to be analysed. The second term is time periodic with frequency $w_{0}$ and has an amplitude proportional to $(\Delta r)A$. In the first term, we need to analyse a typical term ($u_{11}u_{30}$ e.g.) to understand the different time dependent components of the drive. The term $u_{30}$ is periodic with frequency $w_{0}$. Two of the components of $u_{11}$ as shown above are constant in time and periodic with frequency $2w_{0}$. The product $u_{30}u_{11}$ then has terms with frequency $w_{0}$ and the amplitude of these terms are proportional to $A^{3}$. The operator $L$ has a zero mode at frequency $w_{0}$, consequently in order to allow a finite solution of Eq.(17), the right hand side has to be orthogonal to the left eigenvector of $L$ with zero eigenvalue. In the absence of the stochastic drive, this is the entire calculation and the result is that $\Delta r=-\beta^{2}A^{2}$, where $\beta$ is a constant. Thus there is no real value of $A$ if $\Delta r$ is positive and the limit cycle of the Lorenz model cannot be seen. In the presence of stochastic term, things change because $u_{i0}$ (Eq.(14)) has a term proportional to $\eta$ and $u_{i1}$ has the two terms which are listed in (ii) and (iv) above. If we consider $u_{i0}u_{j1}$ then there are two terms whose solution is $A\eta\eta e^{iw_{0}t}$. Since we are in the presence of stochastic terms we can only talk about averages over $\eta$. Consequently, we first need to average Eq.(17) over $\eta$ and only terms involving product of two $\eta$’s will survive. After the averaging the term with structure $A\eta\eta e^{iw_{0}t}$ will acquire the structure $Ae^{iw_{0}t}$ and will be a part of the dangerous term on the right hand side. The orthogonality condition now leads to an equation $\Delta r+N=-\beta^{2}A^{2}$, where $N$ is the extra contribution coming from the stochastic term. In this particular case $N$ is negative and hence for $\Delta r<N$, we can see the limit cycle stabilized and this is the mechanism which allows the time series to be periodic in a region of $r$, where there is no stable limit cycle. In closing, we have studied the statistical properties of noise driven Lorenz model. We have seen critical slowing down at stationary bifurcation point. At $r>1$, the Fourier analysis breaks down due to lack of time translational invariance in present form of Lorenz model. An interesting resonance appears at the Hopf-bifurcation point, which shows the the noise induced stability. We also analysed the noise induced stability of limit cycle in the Lorenz model. ## References * (1) P. Reimann, Phys. Rep. 361, 57 (2002). * (2) R. Benzi, A. Sutera, and A. Vulpiani, J. Phys. A 14, L453, (1981). * (3) L. Gammaitoni, P. Hanggi, P. Jung, and F. Marchesoni, Rev. Mod. Phys. 70, 223 (1998). * (4) C. Van den Broeck, J. M. R. Parrondo, and R. Toral, Phys. Rev. Lett. 73, 3395 (1994); C. Van den Broeck, J. M. R. Parrondo, J. Armero, and A. Hernández-Machado, Phys. Rev. E 49, 2639 (1994). * (5) G. Grinstein, M. A. Muñoz, and Y. Tu, Phys. Rev. Lett. 76, 4376 (1996). * (6) J. Buceta, M. Ibañes, J. M. Sancho, and Katja Lindenberg, Phys. Rev. E 67,0 021113 (2003). * (7) M.C. Cross and P.C. Hohenberg, Rev. Mod. Phys. 65, 851 (1993). * (8) J. B. Gao, Wen-wen Tung and Nageswara Rao, Phys. Rev. Lett. 89, 254101 (2002). * (9) F. Julicher, A. Ajdari and J. Prost Rev. Mod. Phys. 69, 1269 (1997). * (10) R. L. Kautz, J. Appl. Phys. 58, 424 (1985). * (11) F. Arecchi, R. Badii, and A. Politi, Phys. Rev. A 32, 402 (1985). * (12) Annette Zippelius and Manfred L$\ddot{u}$cke, J. Stat. Phys. 24, 345 (1981) * (13) McLaughlin, J. B. and Martin, P. C., Phys. Rev. A 12, 186 (1973). * (14) Lorenz, E. N., J. Atmos. Sci. 20, 130 (1962). * (15) Rowlands, G., J. Phys. A 16, 585 (1983) * (16) Omar Osenda, Carlos B. Briozzo, and Manuel O. Caceres, Phys. Rev. E 55, R3824 (1997).	arxiv-papers	2008-08-28T17:19:54	2024-09-04T02:48:57.529447	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Himadri S. Samanta and J. K. Bhattacharjee", "submitter": "Himadri Samanta", "url": "https://arxiv.org/abs/0808.3956" }
0808.3980	# Solar Coronal Structures and Stray Light in _TRACE_ C.E. DeForest(), P.C.H. Martens(+), and M.J. Wills-Davey(+) ()Southwest Research Insitute, 1050 Walnut Street Suite 300, Boulder, CO 80302 USA, deforest@boulder.swri.edu (+)Smithsonian Astrophysical Observatory ###### Abstract Using the 2004 Venus transit of the Sun to constrain a semi-empirical point- spread function for the _TRACE_ EUV solar telescope, we have measured the effect of stray light in that telescope. We find that 43% of 171Å EUV light that enters _TRACE_ is scattered, either through diffraction off the entrance filter grid or through other nonspecular effects. We carry this result forward, via known-PSF deconvolution of _TRACE_ images, to identify its effect on analysis of _TRACE_ data. Known-PSF deconvolution by this derived PSF greatly reduces the effect of visible haze in the _TRACE_ 171 Å images, enhances bright features, and reveals that the smooth background component of the corona is considerably less bright (and hence more rarefied) than might otherwise be supposed. Deconvolution reveals that some prior conclusions about the Sun appear to have been based on stray light in the images. In particular, the diffuse background “quiet corona” becomes consistent with hydrostatic support of the coronal plasma; feature contrast is greatly increased, possibly affecting derived parameters such as the form of the coronal heating function; and essentially all existing differential emission measure studies of small features appear to be affected by contamination from nearby features. We speculate on further implications of stray light for interpretation of EUV images from _TRACE_ and similar instruments, and advocate deconvolution as a standard tool for image analysis with future instruments such as SDO/AIA. Instrumentation: miscellaneous, Sun: corona, Sun: UV radiation ††slugcomment: _(In press, Astrophysical Journal, fall 2008)_ ## 1 Introduction All telescopes, including _TRACE_ (Handy et al., 1998), scatter light. The principal scattering mechanisms in a space-based telescope include diffraction through the aperture and any obscuration in the beam of the telescope, irregularities or dust on the mirrors themselves, and reflection or scattering in the detector at the focal plane. All these effects contribute to forming broad, shallow wings on the point-spread function (PSF) of the instrument, which describes the image produced by that instrument when viewing an ideal point source of light. These wings are typically 3-5 orders of magnitude fainter than the core of the PSF, but 3-4 orders of magnitude larger, so that a significant fraction of the light incident into the telescope is spread over a large portion of the image. For discrete scenes such as starfields, broad PSF wings are not greatly important except that they reduce the net efficiency of the telescope by reducing the apparent brightness of point sources. For continuous or near- continuous, high-contrast scenes such as viewed by solar telescopes, broad scattering wings are quite important: scattering from large distributed bright structures can overwhelm the emission from dark regions in the same image. These effects, while well known, have received little attention from the solar data analysis community when applied to data from normal-incidence EUV telescopes such as _TRACE,_ but they are present nonetheless and must be accounted for when interpreting images from _TRACE_(and all other solar telescopes). Not only quantitative analysis, but even some qualitative interpretations of _TRACE_ data may be compromised if stray light effects are not accounted for. X rays and ultraviolet light are particularly susceptible to scattering and defocus, and efforts have been made to account for point-spread function effects in previous instruments. For example, Maute & Elwert (1981) attempted blind iterative deconvolution on the X-ray data from _Skylab,_ Svestka et al. (1983) applied it to the _SMM/_ HXIS instrument; and Martens et al. (1995) determined a spatially variable PSF for the _Yohkoh_ /SXT instrument. These studies have largely focused on iterative methods to identify and remove blurring effects caused by a broad PSF core in the subject instruments, though stray light has also been an object of study. Stray light deconvolution was commonly used on SXT data in the later years of that mission (e.g. Foley et al. (1997); Gburek & Sylwester (2002); Schrijver et al. (2004)). The _TRACE_ EUV PSF has been studied by several groups. Lin et al. (2001) used compact, bright flares to study diffraction patterns on the _TRACE_ focal plane and concluded that diffraction from the aluminum filter grids used in _TRACE_ scatters 19% of incident EUV photons into a highly structured, broad diffraction pattern; they speculated that the scattering may be affecting imaging performance. Gburek et al. (2006) used that scattering pattern both to derive both a best-fit PSF core for _TRACE_ and also to determine a portion of the emitted EUV spectrum from particular flare events. In this report, we consider primarily the diffuse scattering wings of the _TRACE_ EUV PSF, and particularly their implications for interpreting coronal images. The wings are not readily measured using a point source such as a flare, because the local intensity of the PSF is quite small far from the core. Outside of diffraction maxima the weak scattered signal is overwhelmed by local emission even for bright events such as flares. Deriving the PSF thus requires analysis of occulted images, using an obstructing body such as the Moon or a planet. We examined _TRACE_ data collected near the times of several solar eclipses, but did not find a suitable EUV image set that contained a clear image of the lunar limb. On 2004 Jun 08, Venus passed in front of the Sun, and several 171 Å image sequences were collected as the planet traversed the disk of the Sun and the off-limb corona. We have used those images to derive a semi-empirical scattering PSF for the _TRACE_ 171 Å channel, and have tested the PSF for correctness by using it to deconvolve several representative __ images of interesting coronal structures. In §2 we briefly review deconvolution and how it is performed, in §3 we describe the forward modeling process and present our measured PSF, and in §4 we demonstrate deconvolution of some representative images. Finally, in §5 we discuss implications for interpretation of EUV coronal images and recommend deconvolution as a standard reduction pipeline component for future telescopes. ## 2 Review of Deconvolution Compensating for the effect of scatter within a telescope requires _deconvolution_ : the telescope convolves the scene with the instrument’s PSF; the effects of the PSF can then be removed by post-processing.__ Here we briefly review known-PSF deconvolution and how it is performed; the process is much simpler and more robust than “blind deconvolution”, which does not require a PSF that is known in advance. Telescopes in general respond to a point source of light by generating an image that has finite extent. This image is the PSF of the telescope, and generally varies at most slowly across the image plane; in this treatment, we consider it to be constant with respect to position on the image plane. Images from the telescope are best described as the convolution __ of the scene being viewed, with the PSF of the telescope. The convolution operation spreads out features by integration (summing) over portions of the source scene, weighted by the PSF. For simplicity, we consider only the post-sampling image plane and use discrete operations such as summing, rather than smooth operations such as integration. Convolving an $n_{x}\times n_{y}$ pixel image $I$ with a convolution kernel $K$ (the instrument’s PSF) involves taking a weighted sum at each location (i,j) in the source image: $\left(I\bigotimes K\right)_{i,j}\equiv\sum_{k=-n_{x}}^{n_{y}}\sum_{l=-n_{y}}^{n_{y}}I_{i-k,j-l}K_{k,l}.$ (1) By construction, it is clear that convolution is a linear operation, so it can be represented with matrix multiplication of $I$ (treated as a $n_{x}n_{y}$-dimensional column vector) by a $n_{x}n_{y}\times n_{x}n_{y}$ matrix $M_{K}$. Undoing the convolution simply requires inverting $M_{K}$. In general, matrix inversion of large matrices is a hard problem (e.g. Claerbout (2004)). Fortunately, convolution in real space is equivalent to elementwise multiplication in Fourier space; in other words, the Fourier basis diagonalizes $M_{K}$, so that finding its inverse is trivial. This is the well-known _convolution theorem_ ; a nice treatment and proof may be found in Bracewell (1999). $I\bigotimes K=\digamma^{-1}\left(\mathcal{I}\cdot\mathbb{\mathcal{K}}\right)$ (2) where $\digamma$ denotes Fourier transformation, the dot product represents elementwise multiplication, and curly vectors $\mathcal{I}$ and $\mathcal{K}$ are the Fourier doubles of their italic counterparts $I$ and $K$. The Fourier transform $\mathcal{K}$ of the PSF is the _optical transfer function_ of the telescope, and its magnitude $\mathcal{\left\|K\right\|}$ is the _modulation transfer function_. Inverting the convolution operation, then, just requires multiplying by the reciprocal of the optical transfer function: $I=\digamma^{-1}(\mathit{\mathcal{I}})=\digamma^{-1}\left(\mathit{\mathit{\mathcal{I\cdot K\cdot}}\mathfrak{R}(\mathit{\mathcal{K}})}\right)$ (3) where $\mathfrak{R}$ is the elementwise reciprocal operator. Because Fourier transformation is itself a linear operation, the components inside the Fourier transform in Eq. 2 can be pulled out, to write: $I=(I\otimes K)\otimes K^{inv}$ (4) where $I\otimes K$ is a source image, and $K^{inv}$ is the function whose Fourier transform is $\mathfrak{R}(\mathcal{K})$. Eq. 4 is useful because it shows that the entire deconvolution operation can be represented as a convolution by a single inverse PSF. If the optical transfer function is known, then $K^{inv}$ is trivial to find. It is just: $K^{inv}=\digamma^{-1}\left(\mathfrak{R}\left(\mathcal{K}\right)\right)$ (5) Because of the reciprocal operation, $K^{inv}$ exists only for kernels with no zero Fourier coefficients. Even when the PSF $K$ is known, deconvolution is not quite as simple in practice as Eq. 5 suggests, because of the presence of noise (which is generally a combination of additive uncorrelated noise, multiplicative uncorrelated noise, and calibration error). A typical image can be represented as a (convolved) true focal-plane image, plus a noise image (which is not wholly independent of the true image): $I=I_{true}\bigotimes K+N$ (6) where $N$ is the noise image. Dividing by $\mathcal{K}$ in the Fourier plane deconvolves the image but also increases the noise term: $\displaystyle I_{deconv.}$ $\displaystyle=$ $\displaystyle\digamma^{-1}\left(I\cdot\mathfrak{R}\left(\mathcal{K}\right)\right)$ (7) $\displaystyle=$ $\displaystyle I_{true}+\digamma^{-1}\left(\mathcal{N\cdot}\mathfrak{R}\left(\mathcal{K}\right)\right).$ (8) The noise term is left in Fourier space to emphasize that $\mathfrak{R}(\mathcal{K})$ is a multiplier on the noise level. Most real telescope PSFs attenuate high spatial frequencies; thus, $\mathfrak{R}(\mathcal{K})$ amplifies those frequencies in the noise by the same factor. The amplified noise term at the far right of Eq. 8 can easily overwhelm $I_{true}$. The theory of _Wiener filters_ develops the optimal balance between signal amplification and noise suppression for additive noise for linear filters such as deconvolution (e.g. Press et al. (1989)).__ Rather than develop the ideal Wiener filter for each image, it is convenient to prevent excessive noise amplification with a _normalized reciprocal_ that rolls over after a certain level of amplification. We used a simple approximation $\mathfrak{R}_{\alpha,\epsilon}$: $\mathfrak{R_{\alpha,\epsilon}}\left(z\right)=\left(\frac{z}{\left\|z\right\|}\right)\frac{\left\|z\right\|^{\alpha}}{\left\|z\right\|^{\alpha+1}+\epsilon^{\alpha+1}},$ (9) where $z$ is complex but $\alpha$ and $\epsilon$ are real. $\mathfrak{R}_{\alpha,\epsilon}$ converges to $z^{-1}$ for large values of $z$ and to $z\left\|z\right\|^{\alpha-1}\epsilon^{-\alpha-1}$ for small values of $\left\|z\right\|$ (when compared with $\epsilon$), and rises to a maximum value close to $\alpha\epsilon^{-1}$. Figure 1 demonstrates construction of an inverse kernel using $\mathfrak{R}_{0.5,0.05}$ for the reciprocal. Figure 1: _Inversion of a sample circularly symmetric kernel using the Fourier transform and regularized reciprocation. (A): original sample PSF, $K$; (B): true (dashed) and regularized (solid) reciprocal of the Fourier components of $K$; (C): resulting inverse kernel $K^{inv}$; (D): convolution of $K\bigotimes K^{inv}$ would be a delta function in the ideal case; it is much improved over the original in (A), which is overplotted as a dashed line._ ## 3 Constraint of the _TRACE_ PSF We generated a PSF using data from the Venus transit of 2004 June 08, assuming that Venus emits no EUV light. The images were prepared, cleaned, and aligned, and a forward model of the scattering was made by convolving the EUV solar images with a PSF to determine the effect of the emission on the center of the Venus image. We generated a forward model PSF that had a narrow core (because we were not interested in sharpening the images, only in reducing stray light) and included _a priori_ the known 171 Å diffraction pattern first described by Lin et al. (2001). Added to the core and the diffraction pattern was a broad truncated Lorentzian described by three parameters: the height, the width parameter, and the width of a Gaussian envelope that was used to truncate the Lorentzian. We convolve the parameterized PSF with the solar EUV images (with the portion inside the disk of Venus masked to black) and calculated a model intensity at each of twelve test loci within the disk of Venus: the center of each of six images of the planet, and two offset loci in each of three on-disk images. We compared these model intensities to the original image brightnesses at the test loci, and adjusted the parameters to find the best fit for all twelve loci. ### 3.1 Image preparation Two sets of _TRACE_ EUV images are present from the Venus occultation, both in the 171 Å passband. During the transit itself many images were taken with no binning and either 16 s, 30 s, or 90 s exposure (in the hour 09:00 - 10:00 UT); and shortly after the transit a series of images with 16 second exposure and 2x2 binning were collected off-limb (in the hour 11:30 - 12:30 UT). On- disk, we used the 90 s exposures to minimize background noise; off-limb, we median-filtered blocks of 5 images along the time axis, to reduce background noise in those less-well-exposed frames. Figure 2 shows a sample on-disk and off-limb image. We downloaded Level 1 data directly from the _TRACE_ web site. We further corrected the zero point by subtracting the average of the 30x30 pixel region in the lower-left corner of each _TRACE_ image (in the filter vignetted area), then scaled the images to 1” per pixel (thus binning 2x2 the on-disk images) and divided out the exposure time and binned pixel size to arrive at calibrated images in _TRACE_ digitizer numbers (DN) arcsec-2 s-1, so that the images were directly comparable. We despiked each image using a simple unsharp-mask + threshold algorithm (the “spikejones.pdl” routine in the PDL portion of the _Solarsoft_ (Freeland & Handy (1998)) software distribution), and replaced each spike value with the median of valid values in its neighborhood in the same image. We made a 5-image pixelwise median of the off-limb images, then spatially shifted each image to center Venus in the frame (Figure 3). Because _TRACE_ has a limited field of view, but the primary and secondary mirrors are exposed to the entire solar disk, it was necessary to extend the field of view to a good fraction of the solar disk to model the extended PSF and reduce the possibility that edge effects would affect the result. We used the closest-in-time full-disk 171Å image from _SOHO_ /EIT (Delaboudiniere et al. (1995)), collected at approximately 19:00 UT on the same day, to fill in missing values outside the _TRACE_ field of view. Although the EIT image was collected some 7-9 hours after the Venus occultation data, the portions of the solar image that are affected are far from test loci in the disk of Venus, and therefore only large spatial scales are important; brightness on these scales varies on timescales of hours to days. The EIT image was prepared using the instrument-supplied eit_prep software, scaled to 1” per pixel, derotated to the _TRACE_ time, and multiplied so that a 100x100 pixel sum (chosen to be far from the _TRACE_ vignetted regions) was equal between the EIT and _TRACE_ image. Then the dark (vignetted) portions of the _TRACE_ image were replaced with the corresponding portion of the scaled, corrected EIT image. Figure 3 shows all of the resulting cleaned, combined images and the loci within them that were used to constrain the fit. Figure 2: __TRACE_ EUV images of the 2004 Venus transit (left) and its immediate aftermath (right), showing the restricted size of the on-disk field of view. The images were extended by combining them with the closest EIT image at the same wavelength. _ Figure 3: _Regularized images from the 2004 June 08 Venus transit used for fitting the _TRACE_ scattering PSF. The EIT image from 19:00 has been used to fill in the missing regions outside the original field of view. The field of view is 800” across; the best-fit PSF has a scattering full-width of ~300”. The sample regions are marked: each image is sampled in a 12”x12” square at the center of the (58” diameter) disk of Venus, and disk images are also sampled in a 12”x6” rectangle offset up 18” from the center of the disk, and 6”x12” rectangle offset left 18” from the center of the disk._ ### 3.2 Forward Modeling of the PSF We forward modeled the scattering portion of the _TRACE_ PSF as the sum of the measured diffraction pattern (Lin et al., 2001; Gburek et al., 2006) and a circularly symmetric scattering profile produced by revolving a radial function about the origin. The revolved function was a sum of a narrow core, a “shoulder” Gaussian, and a truncated Lorentzian intended to represent the scattering wings. The central core width was chosen to have a value much less than the Gburek et al. (2006) width, because the intent is to remove scattering wings rather than to sharpen the core of the PSF by deconvolution. The diffraction pattern was convolved with the central core to avoid pixelization artifacts due to the delta functions in it. The analytic formula is: $K_{\alpha,w,\sigma}(r,\theta)=\gamma^{-1}\left(\left(e^{-4ln(2)r^{2}}/1.27\right)\otimes D(r,\theta)+\frac{\alpha}{r^{2}/w^{2}+1}e^{-4ln(2)r^{2}/\sigma_{t}^{2}}+\beta e^{-4ln(2)r^{2}/\sigma_{s}^{2}}\right),$ (10) where $r$ is distance in the image plane, measured in arcseconds; the 1.27 normalizes the integral under the first Gaussian to unity; $\otimes$ represents convolution; $D(r,\theta)$ is the diffraction pattern described below, including a central core; $\alpha$ is relative strength of the Lorentzian wings; $w$ is the width of the Lorentzian; $\sigma_{t}$ is the FWHM of the truncating Gaussian (in arcsec); $\beta$ is the strength of a Gaussian shoulder to the curve; $\sigma_{s}$ is the width of the shoulder; and $\gamma$ is a factor to normalize the integral under the 2-D convolution kernel to unity. The parameters were found by an iterative fit method: for each guess set of parameters, the kernel was calculated on an 800”x800” grid at 0.5” resolution, multiplied by each of the six composite images, and summed to find the expected scattered intensity at the center of each Venus image. Note that the _TRACE_ field of view is only about 500” across – the larger FOV (available with the EIT overlay) was used so that the _TRACE_ field of view, itself, wouldn’t constrain the fits. In addition, we used an offset kernel to calculate the intensity at an off- center locus in each of the three on- disk images, to constrain the shoulders of the curve a few arcseconds from the core. We did not use the off-center brightness in the off-limb Venus images, because pixels above the limb of the Sun probably contain proportionally more scattered light than do pixels on the solar disk. The geometry of each sample point is shown in Figure 3. We compared the intensities to the forward scattering model, and adjusted the parameters initially “by eye” to find a reasonable match with the twelve data points. Finally, we optimized the fit with an amoeba algorithm (e.g. Press et al. (2007)), holding $w_{core}$ at the conservative 0.5 $arcsec$ full-width at half-maximum and penalizing errors in the overcompensation direction (taking the image value below zero) a factor of 100x worse than errors in the undercompensation direction. The resulting parameters are given by: $\begin{array}[]{ccc}\alpha&=&2.06\times 10^{-5}\\\ w&=&57.7\,arcsec\\\ \sigma_{t}&=&68.4\,arcsec\\\ \beta&=&6\times 10^{-7}\\\ \sigma_{s}&=&15\,arcsec\end{array}$ (11) The fit is within $0.021\,ct\,arcsec^{-2}\,s^{-1}$ of the measured data value at each sample point and 0.019 $ct\,arcsec^{-2}\,s^{-1}$ RMS across the 12 data points, compared to absolute brightnesses of 0.5-1 $ct\,arcsec^{-2}\,s^{-1}$ in the interior of the disk of Venus. The fitted PSF and its inverse are plotted in Figure 4. There is no significant contribution to the total energy outside of a 100” radius. The maximum intensity in the PSF core is 0.5; hence, the modeled isotropic scattering function is down by 5 orders of magnitude from the center of the PSF. Nevertheless, its large cover compared to the core of the PSF yields a significant amount of scattering. Figure 4: _The best-fit point-spread function (and its inverse) for the 171Å channel of _TRACE_. The intensity is plotted on a log scale and is normalized to an integral of unity. _ The encircled energy is plotted versus distance in Figure 5. The encircled energy curve can be counter-intuitive at first: while the scattering wings start and remain small (at under $10^{-3}$ of the intensity at the core of the PSF), at each successive radius more area is available to contribute to the total integrated energy. Hence, in a nearly uniform scene most of the stray light at a given point in the image plane arises from features 20-50 $arcsec$ (40-100 _TRACE_ pixels) away. Approximately 43% of the energy in the derived PSF exists more than 2” from the center of the core. Figure 5: _Encircled energy vs. distance for the model PSF calculated with Equations 10-11. About 57% of the PSF’s total energy is contained in the central core, and 43% is contained in broad scattering wings that extend about 100” (200 _TRACE_ pixels) in all directions. The “wobbles” in the curve are due to successive diffraction maxima._ ## 4 Deconvolution of sample _TRACE_ images Scattering of 43% has a significant impact on images collected with _TRACE._ Here we present results of deconvolution, using the measured PSF from §3. A broad variety of 171Å images were deconvolved and tested for correctness. In no case did any of the pedestal-subtracted, deconvolved images have significant negative-flux regions, an indication that the fitted PSF is either correct or conservative compared to the real PSF of the instrument. Isolated pixels may be carried below zero, due to JPEG artifacts or photon counting noise, but smoothing the image with a $5\times 5$ pixel boxcar kernel eliminates the negative regions. In general, bright features get marginally brighter, and dark regions get much darker, after deconvolution. Figure 6 shows the results of deconvolution of a limb scene, a near-disk-center scene, and several dark prominences near the limb. All of the images have been cropped to the middle $500\times 500$ pixels of the _TRACE_ detector after deconvolution, to avoid edge effects. The images were despiked using a spatial spike finder (“spikejones.pdl” in the _Solarsoft_ distribution; Freeland & Handy (1998))__ In each case, the median value of the lower, left 15x15 pixel region from each despiked _TRACE_ image was used as a zero-point reference: because _TRACE_ is vignetted by the thin- foil filter ring on board, the lower-left corner serves as a reasonable dark reference value. No additional background subtraction was performed. In general, contrast is greatly enhanced throughout the images. For example, the lane in the disk-center active region (center row of Figure 6) is shown to be about a factor of 3 darker than might be expected from naïve analysis of the image (without deconvolution), and small features embedded in bright regions (such as the fan of threads on the right hand side of that image) can be as much as doubled in contrast relative to their local background. The prominences in the bottom row of Figure 6 demonstrate the effectiveness of deconvolution at removing nearby coronal brightness: the prominences (which protrude about one density scale height at 1MK, hence two intensity scale heights) are shown to be quite dark, presumably because they protrude above most of the quiet coronal emissions. The leftmost prominence is directly behind a loop base and is therefore not darkened nearly as much as the others; this forms a good check that the deconvolution is not simply darkening features arbitrarily. The second-from-left prominence is seen between two bright loop structures, and is hence quite dark despite the apparent bright foreground. Figure 6: _Deconvolution of sample _TRACE_ 171 images greatly increases contrast of dark features, eliminates ‘haze’. Top: limb scene. Center: disk AR. Bottom: Prominences near the limb. In each row, left is a level-1 processed _TRACE_ image; middle is the same image, deconvolved; and right is a plot of brightness along the indicated cut. Compact bright features (such as the small active region at the left hand side of the top panel) are increased by nearly 50% in brightness; dark features in the midst of bright regions (such as the dark lane near the center of the central image, or the dark prominences at bottom) are darkened by a factor of about 3\. Note that the prominence at far left of the lower panel is behind a large bright loop structure, while the others in that image are not._ ## 5 Discussion The result that _TRACE_ images contain a significant amount of scattered light is not, in itself, new. Most telescope PSFs include scattering wings, and _TRACE_ is no exception. The filter grid in the front of the telescope is known to scatter $\sim 20\%$ of incident light into a structured diffraction pattern with myriad local maxima (Lin et al. (2001); Gburek et al. (2006)). The present analysis is new in three important respects: it is (to our knowledge) the first analysis of scattered light using _TRACE_ occultation data to derive a PSF; much more scattered light is found than can be accounted for merely by diffraction; and the process is taken to its natural conclusion of deconvolving the original _TRACE_ images to show the effect of the stray light on scientific interpretation of the images. The _TRACE_ Venus data are not detailed enough to constrain a highly structured PSF model, but the simple empirical fit described here is sufficient to improve existing images via deconvolution, and passes the most basic of deconvolution tests, suggesting that it is not overcompensating for the scattering wings. Deconvolution with our scattering PSF has a similar effect to background subtraction on the interpretation of small features: for features small compared to the scattering wings, the effect of the surrounding bright features is approximately constant, so subtraction of a modeled or fixed background yields similar effects in particular local areas of a given _TRACE_ image. The principal advantages of deconvolution are that an approximation of absolute brightness is reproduced, rather than the offset relative brightness that may be extracted from simple background-subtracted images; and that moderate-scale features are treated correctly (they are not treated correctly by simple background or pedestal subtraction). Deconvolution not only darkens the faintest portions of the image, it both increases the relative contrast of small bright features embedded in a bright background and also affects photometric estimates of the relative density of any small bright feature seen with _TRACE_. This describes several features of interest in the _TRACE_ data, including active-region threads that are a subject of current debate (Watko & Klimchuk, 2000; Warren & Winebarger, 2003; Fuentes et al., 2006; DeForest, 2007). The greater coronal contrast we find in deconvolved _TRACE_ images gives indirect support to the idea that the corona is close to hydrostatic equilibrium despite the observed tallness of bright features such as active region loops. The coronal density scale height is about 50 Mm at 1 MK, so the emissivity scale height of Fe IX & Fe X emission line features (close to 1MK ionization temperature) is about 25 Mm (0.035 $R_{S}$) assuming local thermal equilibrium. Thus the EUV-visible corona might be expected to form a thin layer near the photosphere with no significant emission arising at altitudes higher than about 0.07 $R_{S}$. Essentially all _TRACE_ EUV images show significant background brightness high in the corona; the brightness is visible above the detector “pedestal”, because there is contrast between dark but “live” pixels that are part of the image, and corner pixels that are vignetted by the round filter mount at the back of the instrument. The current measured instrument PSF suggests that most or all of this background brightness is due to scattering within the telescope, because dark features (such as the prominences in the bottom row of Figure 6) are reduced nearly to zero brightness when deconvolved. This is the general behavior to be expected from a thin hydrostatic atmosphere at a particular temperature: tall features that are more than 1-2 scale heights tall should have little or no emission above them. Active region loops appear to have a large scale height compared to that expected for 1-2 MK plasma (Schrijver & McMullen, 2000; Aschwanden & Nitta, 2000; Winebarger et al., 2003; Fuentes et al., 2006; DeForest, 2007). Three explanations that have been advanced are resonant scattering of EUV (which varies as $n_{e}$ rather than $n_{e}^{2}$), support by non-hydrostatic momentum transport mechanisms such as siphon flows or wave motion, or geometric considerations that attenuate brightness at the bases of the loop. Our result that the quiet corona appears to be consistent with the expected hydrostatic scale height seems to eliminate resonant scattering as a mechanism for tallness, because it would imply a stronger haze in the foreground at high solar altitudes. Further, it seems to limit the functional form of anomalous support mechanisms that could lengthen active region loops’ scale height, because such mechanisms must act preferentially against active region loops and not quiet sun loops, to be consistent with the morphology of the deconvolved images. This can further be construed as circumstantial evidence for a geometric, rather than intrinsic, explanation for active region loops’ long apparent length (DeForest, 2007). Moore et al. (2008) have recently used image-processing techniques to separate the hazy and sharp components of active region loops viewed with _TRACE._ Such analyses rely on the sharp component of the corona as an indicator of stray light. With 43% scattering of the total light incident on the telescope, 57% is left to be focused; hence, we expect that the hazy portion of such separated image pairs derived from _TRACE_ 171Å data should contain about 75% as much total brightness as the sharp portion does, on the basis of stray light alone. In addition to morphological differences, corrections to the relative brightness of features such as active region threads and voids affect parameters such as the derived Alfvén speed, because of the $n_{e}^{2}$ dependence of EUV emission. Structures with spiky density profiles emit more EUV per electron than do smooth structures, and the inferred Alfvén speed depends both on the magnetic field and the derived electron density. Onset of some _TRACE-_ observed EIT waves appears to require high Alfvén speeds of up to $3\,Mm\,s^{-1}$ (Wills-Davey et al., 2007); this high speed is difficult to explain in the presence of a diffuse background corona around the source region of the EIT wave. If in fact active regions contain nearly evacuated regions (as in the center panel of Figure 6), then the variation in Alfvén speed is greatly increased and the region-wide average Alfvén speed may be significantly higher than would otherwise be inferred. As a final example of the impact of stray light in EUV images _,_ coronal heating properties have been derived (Schrijver et al., 2004) by examining the contrast between coronal holes and bright structures, and may be affected by scattering in _SOHO_ /EIT and/or _Yohkoh_ /SXT. Specifically, if coronal holes are significantly darker, and bright structures are significantly sharper and brigher, than is apparent in raw EUV and X-ray images, then the coronal heating mechanism may not be as distributed as might otherwise be inferred. The model PSF that we have derived for _TRACE_ is somewhat simplified: we have included, _a priori,_ detailed structure that is known from earlier studies (Lin et al., 2001), and parameterized an additional scattering term based on the empirical behavior of stray light on rough mirrors. We have not taken into account possible anisotropy or spatial variability of the scattering, attempted to gain physical understanding of the causes of the PSF, or modeled scattering phenomena that do not fit within the paradigm of a simple PSF. Based on measurements of the Venus transit in 2004, we have found that roughly 43% of incident energy is scattered by _TRACE_ , so that approximately half of the scattered energy may be ascribed to the diffraction pattern found by Lin et al. and approximately half to other mechanisms. Deconvolution greatly improves contrast in _TRACE_ images, raising concerns about the interpretation of those images. More generally, deconvolution to increase contrast in images with scattering wings is strongly recommended for observation from present and future EUV and X-ray telescopes. We have shown that deconvolution can greatly affect the contrast of observed features, and discussed how this effect may affect a broad variety of science questions. Further, deconvolution to remove broad scattering wings is in general not as hazardous to the data as is deconvolution to increase sharpness in the core of the telescope PSF. That is because there are high spatial frequencies present in the core of the kernel, even if it is added to a much broader distribution, so that noise is not increased as much from deconvolution of a scattering PSF as from a broad PSF core. Thanks to the _TRACE_ and _SOHO_ /EIT teams for making their data available to everyone. We also owe thanks to L. Golub, J. Cirtain, K. Schrijver, and H. Throop for illuminating discussions, and to the anonymous referee for several suggestions that improved the work. SOHO is a project of international cooperation between NASA and ESA. This work was funded under NASA’s SHP-SR&T program. ## References * Aschwanden & Nitta (2000) Aschwanden, M. J. & Nitta, N. 2000, ApJ, 535, L59 * Bracewell (1999) Bracewell, R. N. 1999, The Fourier Transform and Its Applications (McGraw-Hill) * Claerbout (2004) Claerbout, J. 2004, Image Estimation by Example (Stanford Exploration Project. Available online: http://sepwww.stanford.edu/sep/prof/gee) * DeForest (2007) DeForest, C. E. 2007, ApJ, 661, 532 * Delaboudiniere et al. (1995) Delaboudiniere, J.-P., Artzner, G. E., Brunaud, J., Gabriel, A. H., Hochedez, J. F., Millier, F., Song, X. Y., Au, B., Dere, K. P., Howard, R. A., Kreplin, R., Michels, D. J., Moses, J. D., Defise, J. M., Jamar, C., Rochus, P., Chauvineau, J. P., Marioge, J. P., Catura, R. C., Lemen, J. R., Shing, L., Stern, R. A., Gurman, J. B., Neupert, W. M., Maucherat, A., Clette, F., Cugnon, P., & van Dessel, E. L. 1995, Sol. Phys., 162, 291 * Foley et al. (1997) Foley, C. R., Culhane, J. L., & Acton, L. W. 1997, ApJ, 491, 933 * Freeland & Handy (1998) Freeland, S. L. & Handy, B. N. 1998, Sol. Phys., 182, 497 * Fuentes et al. (2006) Fuentes, M. C. L., Klimchuk, J. A., & Démoulin, P. 2006, ApJ, 639, 459 * Gburek & Sylwester (2002) Gburek, S. & Sylwester, J. 2002, Sol. Phys., 206, 273 * Gburek et al. (2006) Gburek, S., Sylwester, J., & Martens, P. 2006, Sol. Phys., 239, 531 * Handy et al. (1998) Handy, B. N., Bruner, M. E., Tarbell, T. D., Title, A. M., Wolfson, C. J., Laforge, M. J., & Oliver, J. J. 1998, Sol. Phys., 183, 29 * Lin et al. (2001) Lin, A. C., Nightingale, R. W., & Tarbell, T. D. 2001, Sol. Phys., 198, 385 * Martens et al. (1995) Martens, P. C., Acton, L. W., & Lemen, J. R. 1995, Sol. Phys., 157, 141 * Maute & Elwert (1981) Maute, K. & Elwert, G. 1981, Sol. Phys., 70, 273 * Moore et al. (2008) Moore, R. L., Cirtain, J. W., & Falconer, D. A. 2008, _Eos Trans. AGU_ , 89(23), Jt. Assem. Suppl, Abstract SP43C * Press et al. (1989) Press, W. H., Flannery, B. P., Teukolsky, S. A., & Vetterling, W. T. 1989, Numerical Recipes (Cambridge University Press) * Press et al. (2007) Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 2007, Numerical Recipes 3rd Edition: The Art of Scientific Computing (Cambridge University Press) * Schrijver & McMullen (2000) Schrijver, C. J. & McMullen, R. A. 2000, ApJ, 531, 1121 * Schrijver et al. (2004) Schrijver, C. J., Sandman, A. W., Aschwanden, M. J., & DeRosa, M. L. 2004, ApJ, 615, 512 * Svestka et al. (1983) Svestka, Z., Schrijver, J., Somov, B., Dennis, B. R., Woodgate, B. E., Fuerst, E., Hirth, W., Klein, L., & Raoult, A. 1983, Sol. Phys., 85, 313 * Warren & Winebarger (2003) Warren, H. P. & Winebarger, A. R. 2003, ApJ, 596, L113 * Watko & Klimchuk (2000) Watko, J. A. & Klimchuk, J. A. 2000, Sol. Phys., 193, 77 * Wills-Davey et al. (2007) Wills-Davey, M. J., Sechler, M., & McIntosh, S. W. 2007, AGU Fall Meeting Abstracts, A222+ * Winebarger et al. (2003) Winebarger, A. R., Warren, H. P., & Mariska, J. T. 2003, ApJ, 587, 439	arxiv-papers	2008-08-29T18:30:27	2024-09-04T02:48:57.534168	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "C.E. DeForest, P.C.H. Martens, M.J. Wills-Davey", "submitter": "Craig DeForest", "url": "https://arxiv.org/abs/0808.3980" }
0808.4066	# The Ground State Energy of Dilute Bose Gas in Potentials with Positive Scattering Length Jun Yin Department of Mathematics, Harvard University, Email: JYin@math.harvard.edu (August 29th, 2008) ###### Abstract The leading term of the ground state energy/particle of a dilute gas of bosons with mass $m$ in the thermodynamic limit is $2\pi\hbar^{2}a\varrho/m$ when the density of the gas is $\varrho$, the interaction potential is non-negative and the scattering length $a$ is positive. In this paper, we generalize the upper bound part of this result to any interaction potential with positive scattering length, i.e, $a>0$ and the lower bound part to some interaction potentials with shallow and/or narrow negative parts. ${\,}$${\,}$footnotetext: **${\,}$${\,}$footnotetext: © 2008 by the author. This paper may be reproduced, in its entirety, for non-commercial purposes. ## 1 Introduction and main theorems In Dyson’s work [9] and Lieb, Yngvason and Seiringer’s work [7, 6], it is rigorously proved that the leading term of the ground state energy/particle of a three dimensional dilute bose gas of mass $m$ in the thermodynamic limit with density $\varrho$ is $2\pi\hbar^{2}a\varrho/m$, i.e., $e(\varrho,m)=2\pi\hbar^{2}a\varrho/m(1+o(1))\,\,\,{\rm if}\,\,\,a^{3}\varrho\ll 1$ (1.1) where they assumed that the interaction potential is non-negative, the scattering length $a$ is positive. This result is generalized to a two dimensional dilute bose gas in [8]. In this paper, first, in Theorem 1, we generalize the upper bound part of (1.1) to general interaction potentials $v$ with positive scattering length. On the other hand, for the lower bound on the ground energy, it was conjectured in [7] that the lower bound part of (1.1) should hold if the scattering length is positive and $v$ has no $N$-body bound states for any $N$. Recently, it is proved in [11] that in some cases with partly shallow negative potential the lower bound part of (1.1) holds. In Theorem 2, we introduce a different method for the lower bound on (1.1) when $v$ can have shallow and/or narrow negative components and provide better(smaller) error term. We begin with describing the questions more precisely. We write the Hamiltonian of a system of $N$ interacting bosons which are restricted to a cubic box of volume $\Lambda=L^{3}$ in the following way (in units where $\hbar=2m=1$): $\displaystyle H_{N}\equiv\sum_{i=1}^{N}-\Delta_{i}+\sum_{1\leq i<j\leq N}v^{a}(x_{i}-x_{j})$ (1.2) Here $\Delta$ denotes the Laplacian on $\Lambda$ with periodic boundary condition and $v^{a}$ is a scaled interaction potential, i.e., $v^{a}(x)={a^{-2}}\cdot v(x/a),\,\,\,a>0$ (1.3) The pair interaction potential $v$ is spherically symmetric and supported on the set $\\{x\in{\mathbb{R}}^{3}:\|x\|\leq R_{0}\\}$ for some $R_{0}>0$. ###### DEFINITION 1 (Scattering Length). Assume that $w$ is a pair spherically symmetric interaction potential with compact support. Denote $E[\phi\,]$ as the energy of the complex-valued function $\phi$ on ${\mathbb{R}}^{3}$ as follows, $E[\phi\,]=\int_{{\mathbb{R}}^{3}}\|\nabla\phi(x)\|^{2}+\mbox{$\frac{1}{2}$}w(x)\|\phi(x)\|^{2}dx.$ (1.4) Define the scattering length $SL(w)$ of potential $w$ as the following minimum energy. $SL(w)\equiv\min_{\phi}\left\\{\frac{1}{4\pi}E[\phi\,]:\lim_{\|x\|\to\infty}\phi(x)=1\right\\}$ (1.5) Note: If $SL(w)>-\infty$, one can easily prove that the Hamiltonian $-\Delta+\mbox{$\frac{1}{2}$}w$ has no bound state. In particular, when $w\geq 0$, we have $SL(w)\geq 0$ and $w$ has no bound state. One can see that this definition is equivalent to the definition of scattering length in [5] when $w>0$. With the relation between $v$ and $v^{a}$ in (1.3), we can assume that $SL(v)=1,\,\,\,\,SL(v^{a})=a$ (1.6) Let $f_{1}(x)$ be the solution of the zero-energy scattering equation of $v$, i.e., $-\Delta f_{1}(x)+\mbox{$\frac{1}{2}$}v(x)f_{1}(x)=0,$ (1.7) then we have that $f_{a}(x)\equiv f_{1}(x/a)$ is the solution of the following zero-energy scattering equation of $v^{a}$. $-\Delta f_{a}(x)+\mbox{$\frac{1}{2}$}v^{a}(x)f_{a}(x)=0,$ (1.8) As in [5], one can prove that if $f_{a}$ is normalized as $\lim_{\|x\|\to\infty}f_{a}(x)=1$, then $f_{a}(x)=1-a/x,\,\,\,{\rm for}\,\,\,\|x\|>R_{0}a$ (1.9) In this paper, we are interested in the ground energy $E(N,\Lambda)$ of $H_{N}$ in the thermodynamic limit that $\Lambda\to\infty$, $N\to\infty$ and $N/\Lambda=\varrho$. Low density means that the average inter-particle distance $\varrho^{-1/3}$ is much larger than the scattering length $a$, i.e. $a^{3}\varrho\ll 1$. First, we state that for any fixed $v$, the upper bound on (1.1) holds for the dilute bose gas. ###### THEOREM 1. Fix $v$ with $SL[v]=1$ and $v^{a}$ satisfying (1.3). Let $f_{1}$ be the solution of zero-energy equation of $v$ and normalized as: $f_{1}(\infty)=1$. In the thermodynamic limit, $\lim_{N\to\infty}N/\Lambda=\varrho$, we have the following upper bound on $E(N,\Lambda)$, which is the ground energy of $H_{N}$ in (1.2), $\displaystyle\limsup_{N\to\infty}\frac{E(N,\Lambda)}{4\pi a\varrho N}\leq 1+{\rm const.\,}(a^{3}\varrho)^{1/4},$ (1.10) for some constant depending on $\\|f_{1}\\|_{\infty}$, provided that $\frac{4\pi}{3}a^{3}\varrho\leq 1$. Note: So far, the best proof of the error term on upper bound, when $v\geq 0$, is $O(a^{3}\varrho)^{1/3}$, as in [5]. On the other hand, for the lower bound in (1.1), we prove that as long as $v$ has a positive core and is bounded from below, (1.1) holds when the negative part is small enough (shallow and/or narrow). In the appendix, we show that if $v^{a}$ is a continuous function on ${\mathbb{R}}^{3}$ and $H_{N}$ has no bound state for any $N$, $v^{a}$ satisfies the above two requirements, i.e., $v^{a}(0)>0,\,\,\,\,\min v^{a}(r)>-\infty$ (1.11) The above two inequalities (1.11) also hold when $v^{a}$ is stable [1] (the stability of potential is assumed in [11]). ###### THEOREM 2. We assume that $v(x)=v_{+}(x)+v_{-}(x)$, $v_{+}(x)\geq 0$, $v_{-}(x)\geq-\lambda_{-}$, $\lambda_{-}>0$ and $v_{+}$ has a positive core, i.e. $\exists\,r_{1}$, such that $v_{+}(x)\geq\lambda_{+}>0$ for $\|x\|\leq r_{1}$. Here $v_{-}$ need not be negative. There exist $c_{1}(R_{0}/r_{1})$ and $c_{2}(R_{0}/r_{1})$, which are greater than one and only depend on $R_{0}/r_{1}$, such that the following holds. If there exists some positive number $t$ satisfying $SL[c_{1}(R_{0}/r_{1})\cdot(v+tv_{-})]\geq 0\,\,\,{\rm and}\,\,\,\lambda_{+}\geq(1+t^{-1})\,c_{2}(R_{0}/r_{1})\cdot\lambda_{-},$ (1.12) we have the following lower bound on $E(N,\Lambda)$, $\displaystyle\liminf_{N\to\infty}\frac{E(N,\Lambda)}{4\pi a\varrho N}\geq 1-{\rm const.\,}(a^{3}\varrho)^{1/17},$ (1.13) for some constant depending on $v_{+}$ and $v_{-}$, provided that $\frac{4\pi}{3}a^{3}\varrho$ is smaller than some constant depending on $v_{+}$, $v_{-}$ and $t$. Note: So far, the best estimation of the error term of the lower bound, when $v>0$, is also $O(a^{3}\varrho)^{1/17}$, as in [5]. This theorem implies the following two corollaries. ###### COROLLARY 1. Assume that $v(x)=v_{+}(x)+\lambda_{-}v_{-}(x),\,\,\,v_{+}(x)\geq 0,\,\,\,v_{-}(x)\geq-1$ (1.14) and $v_{+}$ has a positive core, i.e. $\exists r_{1}$ such that $v_{+}(x)\geq\lambda_{+}$ for $\|x\|\leq r_{1}$. There exists $\lambda_{0}(r_{1},R_{0},\lambda_{+},v_{-})$ such that, if $0\leq\lambda_{-}\leq\lambda_{0}$, i.e., the potential is shallow enough, we have the following lower bound on $E(N,\Lambda)$, $\displaystyle\liminf_{N\to\infty}\frac{E(N,\Lambda)}{4\pi a\varrho N}\geq 1-{\rm const.\,}(a^{3}\varrho)^{1/17}$ (1.15) provided that $\frac{4\pi}{3}a^{3}\varrho$ is smaller than some constant depending on $v_{+}$ and $v_{-}$. ###### Proof. For fixed $R_{0}$, $r_{1}$ and $\lambda_{+}$, when $\lambda_{-}$ is small enough, we have that $SL[c_{1}(R_{0}/r_{1})(v_{+}+2\lambda_{-}v_{-})]\geq 0\,\,\,{\rm and}\,\,\,\lambda_{+}\geq 2\,c_{2}(R_{0}/r_{1})\lambda_{-}.$ (1.16) Using Theorem 2, with the choice $t=1$, we arrive at the desired result. ∎ ###### COROLLARY 2. Assume that $v(x)=v_{+}(x)+v_{-}(x),\,\,\,\,v_{+}(x)\geq 0\geq v_{-}(x)\geq-\lambda_{-}$ (1.17) and $v_{+}$ has a positive core, i.e. $\exists r_{1}$ such that $v_{+}(x)\geq\lambda_{+}$ for $\|x\|\leq r_{1}$. There exist $\lambda_{0}(R_{0}/r_{1},\lambda_{-})$ and $\varepsilon\,(R_{0},r_{1},\lambda_{-})$ such that, if $\lambda_{+}\geq\lambda_{0}\,\,\,{\rm and}\,\,\,\int_{x\in{\mathbb{R}}^{3}}\|v_{-}(x)\|dx\leq\varepsilon\,(R_{0},r_{1},\lambda_{-}),$ (1.18) we have the following lower bound on $E(N,\Lambda)$, $\displaystyle\liminf_{N\to\infty}\frac{E(N,\Lambda)}{4\pi a\varrho N}\geq 1-{\rm const.\,}(a^{3}\varrho)^{1/17}$ (1.19) provided that $\frac{4\pi}{3}a^{3}\varrho$ is smaller than some constant depending on $v_{+}$ and $v_{-}$. ###### Proof. We choose $\lambda_{0}=\max\\{3,2\,c_{2}(R_{0}/r_{1})\\}\lambda_{-}$, then we have that $\lambda_{+}\geq\lambda_{0}\geq 3\lambda_{-}$, which implies that $[v+v_{-}](x)\geq\lambda_{+}/3\geq 0\,\,\,{\rm for}\,\,\,\|x\|\leq r_{1}$ (1.20) Then we claim that for any $n\geq 1$ and $\lambda_{+}\geq 3\lambda_{-}$, there exists $\xi(n)>0$, $\int_{{\mathbb{R}}^{3}}\|v_{-}(x)\|dx\leq\xi(n)\Rightarrow SL[n(v+v_{-})]\geq 0.$ (1.21) To prove (1.21), we shall prove that there exists $\xi(n)>0$, if $\int_{{\mathbb{R}}^{3}}\|v_{-}(x)\|dx\leq\xi(n)$, for any non-negative radial function $f$, $\int_{\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)+\frac{n}{2}(v+v_{-})f^{2}(x)dx\geq 0.$ (1.22) We can see, with (1.20), $\displaystyle\int_{\|x\|\leq R_{0}}\left[\|\nabla f\|^{2}(x)+\frac{n}{2}(v+v_{-})f^{2}(x)\right]dx$ (1.23) $\displaystyle\geq$ $\displaystyle\int_{r_{1}\leq\|x\|\leq R_{0}}\left[\|\nabla f\|^{2}(x)-\\|nv_{-}\\|_{\infty}f^{2}(x)\right]dx$ $\displaystyle\geq$ $\displaystyle\int_{r_{1}\leq\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)dx-n\lambda_{-}\int_{r_{1}\leq\|x\|\leq R_{0}}\|f\|^{2}(x)dx$ Hence, if (1.22) does not hold, the right side of (1.23) is less than 0. With Sobolev inequality and Schwarz’s Inequality, we obtain that there exists $\eta(n)$ such that $\left(\int_{r_{1}\leq\|x\|\leq R_{0}}\|f(x)\|^{4}dx\right)^{1/2}\leq\eta(n)\int_{r_{1}\leq\|x\|\leq R_{0}}\|f\|^{2}(x)dx$ (1.24) On the other hand, with (1.20), $v+v_{-}\geq 2v_{-}$ and Schwarz’s Inequality, we have that $\displaystyle\int_{\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)+\frac{n}{2}(v+v_{-})f^{2}(x)dx$ (1.25) $\displaystyle\geq$ $\displaystyle\int_{\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)+\int_{\|x\|\leq r_{1}}\frac{n\lambda_{+}}{6}f^{2}(x)-\left\|\int_{r_{1}\leq\|x\|\leq R_{0}}\\!\\!\\!\\!\\!\\!\\!\\!\\!\\!nv_{-}(x)\|f(x)\|^{2}dx\right\|$ $\displaystyle\geq$ $\displaystyle\int_{\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)+\int_{\|x\|\leq r_{1}}\frac{n\lambda_{+}}{6}f^{2}(x)$ $\displaystyle-n\eta(n)\left\|\int_{r_{1}\leq\|x\|\leq R_{0}}v^{2}_{-}(x)dx\right\|^{1/2}\int_{r_{1}\leq\|x\|\leq R_{0}}\|f\|^{2}(x)dx$ $\displaystyle\geq$ $\displaystyle\int_{\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)+\int_{\|x\|\leq r_{1}}\frac{n\lambda_{+}}{6}f^{2}(x)-n\eta(n)\lambda_{-}\\|v_{-}\\|_{1}^{1/2}\int_{\|x\|\leq R_{0}}\|f\|^{2}(x)dx$ Thus, for $n\geq 1$, if $\displaystyle\\|v_{-}\\|_{1}^{1/2}\leq\left(\xi(n)\right)^{1/2}\equiv\frac{1}{n\eta(n)}\cdot\min_{f}\frac{\int_{\|x\|\leq R_{0}}\|\nabla f\|^{2}(x)+\int_{\|x\|\leq r_{1}}\frac{n\lambda_{+}}{6}f^{2}(x)dx}{\int_{\|r\|\leq R_{0}}\|f\|^{2}(x)dx},$ the inequality (1.22) holds. We note that it is easy to see that $\xi(n)>0$. Hence we arrive at the desired result (1.21). At last, choosing $\varepsilon\,(R_{0},r_{1},\lambda_{-})=\xi\left(c_{1}(R_{0}/r_{1})\right)$ (1.26) and using the result of Theorem 2 with $t=1$, we arrive at the desired result (1.19). ∎ Remark: Compared with the result of [11], we improve the error term (It was $(a^{3}\varrho)^{1/31}$ in [11]) and generalize the shapes of potentials, i.e., the negative part of potential can be shallow and/or narrow. In particular, there is no restriction on the depth of the interaction potential $v$, i.e. for $\forall\lambda_{-}>0$, there $\exists v$ satisfying $\min_{x\in{\mathbb{R}}^{3}}v(x)<-\lambda_{-}$ and Theorem 2 holds. ## 2 Proofs ### 2.1 Proof of Theorem one ###### Proof. As usual, to prove the upper bound on the ground state energy, we only need to construct a sequence of trial states $\Psi_{N,\Lambda}$ satisfying $\limsup_{N\to\infty}\frac{\langle\Psi\|H_{N}\|\Psi\rangle}{N\langle\Psi\|\Psi\rangle}\leq 4\pi a\varrho(1+{\rm const.\,}Y)$ (2.1) for some constant that depends only on $\\|f_{1}\\|_{\infty}$. Here we denote $Y$ as $Y\equiv\left(\frac{4\pi}{3}a^{3}\varrho\right)^{1/4}$ (2.2) Following the ideas in [9, 6], we construct the trial state of the following form, $\Psi_{N}=\prod_{p=1}^{N}F_{p}$ (2.3) In [9], $F_{p}$ depends on the the nearest particle to the $x_{p}$ among all the $x_{i}$ with $i<p$, i.e., $F_{p}=f(t_{p}),\,\,\,t_{p}=\min_{i<p}\left\\{\left\|x_{i}-x_{p}\right\|\right\\}$ (2.4) via the function $f$ which is very close to the zero energy scattering solution and satisfies $0\leq f\leq 1,\,\,\,\,f\,^{\prime}\geq 0$ (2.5) Hence in [9], $F_{p}$ has the following property $F_{p,i}\cdot f(\|x_{p}-x_{i}\|)\leq F_{p}\leq F_{p,i}$ (2.6) Here $F_{p,i}$ is defined in [9] as the value that $F_{p}$ would take if the point $x_{i}$ were omitted from consideration. But in our case where the potential has a negative part, the zero energy scattering solution $f_{a}$ of $v^{a}$ may not be an increasing function or bounded by 1 (if it was, the proof would be much simpler). Hence we do not have the property (2.6). For this reason, our choice of $F_{p}$ will be more complicated. Our $F_{p}$ depends on all particles near the $x_{p}$, not just the nearest. We remark that the function $F_{p}$ should have following properties. 1. 1. $F_{p}$ is a continuous function of $x_{i}$ ($1\leq i\leq N$). 2. 2. When $\|x_{i}-x_{p}\|$ is large enough, the position of $x_{i}$ does not effect $F_{p}$, i.e., $\nabla_{x_{i}}F_{p}=0$. 3. 3. $F_{p}$ has a similar property as (2.6). First we define $\theta_{r}(x)$ as the characteristic function of the set $\\{x:\|x\|\leq r\\}$ and $\bar{\theta}_{r}\equiv 1-\theta_{r}$. Choosing $b=a/Y$, we have $a/b=\frac{4\pi}{3}b^{3}N/\Lambda=Y.$ (2.7) Without loss of generality, we assume that $b>\max\\{2R_{0}a,4a\\}$, as in [9, 5]. We define $f(x)$ as $f(x)=\left\\{\begin{array}[]{ll}f_{a}(x)/f_{a}(b)&b\geq\|x\|\geq 0\\\ 1&{\rm otherwise}\,,\end{array}\right.$ (2.8) Here $f_{a}$ is the zero energy scattering solution of $v^{a}$, as in (1.8). With the equation (1.9), we note that $f(x)=\frac{1-a/\|x\|}{1-a/b},\,\,{\rm for}\,\,\,b\geq\|x\|\geq R_{0}a.$ (2.9) Let ${\widetilde{R}}=\max\\{R_{0}a,2a\\}$, which implies that $f({\widetilde{R}})>\mbox{$\frac{1}{2}$}$. We define $\Theta^{in}_{p}$, $\Theta^{out}_{p}$ ($1<p\leq N$) as $\displaystyle\Theta^{in}_{p}\equiv\prod_{j<p}\theta_{\widetilde{R}}(x_{j}-x_{p}),\,\,\,\Theta^{out}_{p}\equiv\prod_{j<p}\bar{\theta}_{\widetilde{R}}(x_{j}-x_{p})$ (2.10) We can see that $\Theta^{in}_{p}=1$ when $\|x_{j}-x_{p}\|\leq{\widetilde{R}}$ for all $j<p$ and $\Theta^{out}_{p}=1$ when $\|x_{j}-x_{p}\|>{\widetilde{R}}$ for all $j<p$. With $\Theta^{in}_{p}$ and $\Theta^{out}_{p}$, we can define $r_{p}(x_{1},\cdots,x_{N})$ and $R_{p}(x_{1},\cdots,x_{N})$ as follows, ($x_{i}\in[0,L]^{3},i=1,\cdots N$) $\displaystyle r_{p}\equiv$ (2.11) $\displaystyle(1-\Theta^{out}_{p})\cdot\min_{i<p}\bigg{\\{}\|x_{i}-x_{p}\|:f(x_{i}-x_{p})=\min_{j<p}\big{\\{}f(x_{j}-x_{p}):\|x_{j}-x_{p}\|\leq{\widetilde{R}}\big{\\}}\bigg{\\}}$ $\displaystyle R_{p}\equiv{\widetilde{R}}\cdot\Theta^{in}_{p}+(1-\Theta^{in}_{p})\times\min_{i<p}\bigg{\\{}\|x_{i}-x_{p}\|:\|x_{i}-x_{p}\|>{\widetilde{R}}\bigg{\\}}$ With the definition of $R_{p}$ and (2.9), we have that 1. 1. $f(R_{p})\leq f(x_{j}-x_{p})$ for any $j<p$ satisfying $\|x_{j}-x_{p}\|>{\widetilde{R}}$, 2. 2. $R_{p}\leq\|x_{j}-x_{p}\|$ for any $j<p$ satisfying $\|x_{j}-x_{p}\|>{\widetilde{R}}$. 3. 3. $R_{p}\geq{\widetilde{R}}$ 4. 4. When $\Theta^{in}_{p}=0$, there exists $j_{p}$ such that $\|x_{j_{p}}-x_{p}\|=R_{p}$ Similarly, we have 1. 1. $f(r_{p})\leq f(x_{j}-x_{p})$ for any $j<p$ satisfying $\|x_{j}-x_{p}\|\leq{\widetilde{R}}$, 2. 2. $r_{p}\leq\|x_{j}-x_{p}\|$ for any $j<p$ satisfying $\|x_{j}-x_{p}\|\leq{\widetilde{R}}$ and $f(x_{j}-x_{p})=f(r_{p})$. 3. 3. $r_{p}\leq{\widetilde{R}}$ 4. 4. When $\Theta^{out}_{p}=0$, there exists $i_{p}$ such that $\|x_{i_{p}}-x_{p}\|=r_{p}$ Then, we define a continuous function $T$ on ${\mathbb{R}}$ as follows $T(\|x\|)=\left\\{\begin{array}[]{ll}1&2{\widetilde{R}}\geq\|x\|\\\ (\|x\|^{-1}-b^{-1})(2{\widetilde{R}}^{-1}-b^{-1})^{-1}&b\geq\|x\|\geq 2{\widetilde{R}}\\\ 0&{\|x\|\geq b}\,,\end{array}\right.$ (2.12) At last we define $F_{p}(x_{1},\cdots,x_{N})$ on $[0,L]^{3N}$ as follows ($1<p\leq N$), $\displaystyle F_{p}\equiv\left\\{\begin{array}[]{ll}f(r_{p})&\Theta^{in}_{p}=1\\\ f(R_{p})&\Theta^{out}_{p}=1\\\ f(r_{p})+T(R_{p})\left[f(R_{p})-f(r_{p})\right]_{-}&{otherwise}\,,\end{array}\right.$ (2.16) and $F_{1}=1$. Here $[\cdot]_{-}$ denotes the negative part, i.e., $[x]_{-}=x$ when $x<0$ and $[x]_{-}=0$ when $x\geq 0$. We note that for any $x$ $[x]_{-}\leq 0$ (2.17) Note: If $v\geq 0$, it is well known that $f$ is an increasing function, which implies the $F_{p}$ we defined is equal to the $F_{p}$ in [9]. One can prove that $F_{p}$ is a continuous function of $(x_{1},\cdots,x_{N})$ by checking that, for any $j\neq p>1$ and fixed $x_{1},\cdots,x_{j-1},x_{j+1},\cdots,x_{N}$, $F_{p}$ is a continuous function of $x_{j}$. First we can see that it is trivial for $j>p$, since $F_{p}$ is independent of $x_{j}$ when $j>p$. For $j<p$, it only remains to check that $F_{p}$ is continuous when $x_{j}$ moves from $\|x_{j}-x_{p}\|={\widetilde{R}}$ to $\|x_{j}-x_{p}\|={\widetilde{R}}+0^{+}$. One can see that when $\|x_{j}-x_{p}\|={\widetilde{R}}$, $f(R_{p})\geq f({\widetilde{R}})=f(x_{j}-x_{p})\geq f(r_{p})$, so $F_{p}=f(r_{p})$, i.e. $F_{p}=\min\bigg{\\{}\min_{k:k\neq j,k<p}\\{f(x_{k}-x_{p}):\|x_{k}-x_{p}\|\leq{\widetilde{R}}\\},f(x_{j}-x_{p})\bigg{\\}}$ (2.18) On the other hand, when $\|x_{j}-x_{p}\|={\widetilde{R}}+0^{+}\leq 2{\widetilde{R}}$, we can see that $R_{p}=\|x_{j}-x_{p}\|$, $T(R_{p})=1$ and $f(R_{p})=f({\widetilde{R}})+0^{+}$. Hence, $F_{p}=\min\bigg{\\{}\min_{k:k\neq j,k<p}\\{f(x_{k}-x_{p}):\|x_{k}-x_{p}\|\leq{\widetilde{R}}\\},f(x_{j}-x_{p})\bigg{\\}}$ (2.19) Hence we arrive at the desired result that $F_{p}$ is continuous function. We can also see that $F_{p}$ is non-negative and bounded as follows $M\equiv\\|F_{p}\\|_{\infty}=\\|f\\|_{\infty}\leq(1-a/b)^{-1}\\|f_{a}\\|_{\infty}=(1-a/b)^{-1}\\|f_{1}\\|_{\infty}\leq 2\\|f_{1}\\|_{\infty}.$ (2.20) Here we use the fact $f_{a}(x)=f_{1}(x/a)$. By the definition of $F_{p}$, one can see that $F_{p}=1$ when $\prod_{q<p}\bar{\theta}_{b}(x_{p}-x_{q})=1$ and $F_{p}\leq 1$ when $\prod_{q<p}\bar{\theta}_{\widetilde{R}}(x_{p}-x_{q})=1$, so $1-\sum_{q<p}\theta_{b}(x_{p}-x_{q})\leq F_{p}\leq 1+\sum_{q<p}(M-1)\theta_{\widetilde{R}}(x_{p}-x_{q})$ (2.21) We now construct the state functions $\Phi_{k}$ as follows ($1\leq k\leq N$) $\Phi_{k}=\prod_{p=1}^{k}F_{p}$ Note: all $\Phi$’s are functions on $[0,L]^{3N}$ and $\Phi_{k}$ is independent of $x_{l}$ for $l>k$. We will choose $\Psi=\Phi_{N}$ for (2.1). As in [7], for proving the upper bound on the total energy $\langle\Phi_{N}\|H_{N}\|\Phi_{N}\rangle\\|\Phi_{N}\\|^{-2}_{2}$, we shall estimate the upper bounds on $\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i}\|\nabla_{i}\Phi_{N}\|^{2}\prod_{j=1}^{N}dx_{j}\,\,\,\,\,\,{\rm and}\,\,\,\,\,\,\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i<j}v^{a}(x_{i}-x_{j})\|\Phi_{N}\|^{2}\prod_{k=1}^{N}dx_{k}$ (2.22) Since in our case $v^{a}$ has negative parts, our strategy is more complicated, i.e., we need to estimate the upper bounds on $\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i}\|\nabla_{i}\Phi_{N}\|^{2}\prod_{j=1}^{N}dx_{j}\,\,\,\,\,\,{\rm and}\,\,\,\,\,\,\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i<j}[v^{a}]_{+}(x_{i}-x_{j})\|\Phi_{N}\|^{2}\prod_{k=1}^{N}dx_{k}$ (2.23) and the lower bound on $\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i<j}\bigg{\|}[v^{a}]_{-}(x_{i}-x_{j})\bigg{\|}\cdot\|\Phi_{N}\|^{2}\prod_{k=1}^{N}dx_{k}$ (2.24) In the remainder of this section we are going to prove the following three inequalities • $\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i}\|\nabla_{i}\Phi_{N}\|^{2}\prod_{j=1}^{N}dx_{j}\leq(1+o(1))\frac{N^{2}}{\Lambda}\int_{{\mathbb{R}}^{3}}\|\nabla f(x)\|^{2}dx$ * • $\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i<j}[v^{a}]_{+}(x_{i}-x_{j})\|\Phi_{N}\|^{2}\prod_{k=1}^{N}dx_{k}\leq(1+o(1))\frac{N^{2}}{\Lambda}\int_{{\mathbb{R}}^{3}}\mbox{$\frac{1}{2}$}[v]_{+}\|f(x)\|^{2}dx$ * • $\\|\Phi_{N}\\|^{-2}_{2}\int\sum_{i<j}\bigg{\|}[v^{a}]_{-}(x_{i}-x_{j})\bigg{\|}\cdot\|\Phi_{N}\|^{2}\prod_{k=1}^{N}dx_{k}\geq(1-o(1))\frac{N^{2}}{\Lambda}\int_{{\mathbb{R}}^{3}}\mbox{$\frac{1}{2}$}\left\|[v]_{-}\right\|\cdot\|f(x)\|^{2}dx.$ To prove these inequalities, we begin with proving the following three inequalities (all $\Phi$’s are functions on $[0,L]^{3N}$): 1. 1. For any $m$-variable function $g_{m}(x_{i_{1}}\cdots x_{i_{m}})$, $m<k\leq N$, $i_{j}\neq i_{k}$ for $j\neq k$, we have $\displaystyle\\|\Phi_{k}F^{-1}_{i_{1}}\cdots F^{-1}_{i_{m}}g_{m}\\|^{2}_{2}\leq(2M)^{2m}\Lambda^{-m}\\|\Phi_{k-m}\\|^{2}_{2}\\|g_{m}\\|^{2}_{2}$ (2.25) 2. 2. For any two variable function $g_{2}(x_{i},x_{i^{\prime}})$ ($i<i^{\prime}$), we have $\\|\Phi_{N}F^{-1}_{i^{\prime}}g_{2}\\|^{2}_{2}\leq\\|g_{2}\\|^{2}_{2}\Lambda^{-2}\\|\Phi_{N}\\|^{2}_{2}(1+{\rm const.\,}Y)$ (2.26) 3. 3. Let $f_{i,i^{\prime}}=f(x_{i}-x_{i^{\prime}})$, for any two variable function $g_{2}(x_{i},x_{i^{\prime}})$ ($i<i^{\prime}$), we have $\displaystyle\\|\Phi_{N}f^{-1}_{i,i^{\prime}}g_{2}\\|^{2}_{2}\geq\\|g_{2}\\|^{2}_{2}\Lambda^{-2}\\|\Phi_{N}\\|^{2}_{2}(1-{\rm const.\,}Y)$ (2.27) Note: if $f_{i,i^{\prime}}=0$ and $i<i^{\prime}$, then $F_{i^{\prime}}=0$, so $\Phi_{N}f^{-1}_{i,i^{\prime}}$ is definable. We will use (2.25) for controlling the error terms. The inequalities (2.26) and (2.27) will be used in estimating the terms (2.23) and (2.24), respectively. We begin with deriving a lower bound on $\\|\Phi_{N}\\|^{2}_{2}$. For $i\neq p$, Let $F_{p,i}$ be the value that $F_{p}$ would take if changing the order of particles as follows, $F_{p,i}(x_{1}\cdots x_{N})\equiv F_{n(p,i)}(x_{1}\cdots x_{i-1},x_{i+1}\cdots x_{N},x_{i}\,)$ (2.28) Here $n(p,i)$ is defined as follows ($i\neq p$) $n(p,i)=\left\\{\begin{array}[]{ll}p&i>p\\\ p-1&i<p\,,\end{array}\right.$ (2.29) Similarly, we can define $F_{p,i,j}(x_{1}\cdots x_{N})$ as $F_{p,i,j}(x_{1}\cdots x_{N})\equiv F_{m(p,i,j)}(x_{1}\cdots x_{i-1},x_{i+1}\cdots x_{j-1},x_{j+1}\cdots x_{N},x_{i},x_{j}\,)\,\,\,for\,\,\,i<j$ (2.30) and $F_{p,i,j}=F_{p,j,i}$ for $j<i$. Here $m(p,i,j)$ is defined as the number of the elements of the set $\\{1,\cdots,p\\}\setminus\\{i,j\\}.$ Note: As we mentioned $F_{p}$ we defined is equal to the $F_{p}$ in [6] in the case when $v\geq 0$. Furthermore, one can see that our definitions of $F_{p,i}$ and $F_{p,i,j}$ are equivalent to those definitions in [6] when $v\geq 0$. With the definitions of $F_{p}$ and $F_{p,i}$, we obtain that $F_{p,i}$ is independent of $x_{i}$ and $F_{p}$ is bounded from below as follows $F_{p}(x_{1}\cdots x_{N})\geq\left\\{\begin{array}[]{ll}F_{p,i}&i>p\\\ F_{p,i}\bar{\theta}_{b}(x_{p}-x_{i})&i<p\,,\end{array}\right.$ (2.31) and $F_{p}\geq\prod_{i<p}\bar{\theta}_{b}(x_{p}-x_{i}).$ Then $\Phi^{2}_{N}$ is bounded from below, for any fixed $i$, by $\displaystyle\bigg{\|}F_{1}\cdot F_{2}\cdots F_{N}\bigg{\|}^{2}$ $\displaystyle\geq\bigg{\|}F_{1,\,i}\cdots F_{i-1,\,i}F_{i+1,\,i}\cdots F_{N,\,i}\bigg{\|}^{2}\times\prod_{j\neq i}\bar{\theta}_{b}(x_{i}-x_{j})$ $\displaystyle\geq\bigg{\|}F_{1,\,i}\cdots F_{i-1,\,i}F_{i+1,\,i}\cdots F_{N,\,i}\bigg{\|}^{2}\times\bigg{(}1-\sum_{j\neq i}\theta_{b}(x_{i}-x_{j})\bigg{)}$ Integrating both sides with $\int\prod_{j=1}^{N}dx_{j}$, we obtain that $\displaystyle\\|\Phi_{N}\\|_{2}^{2}\geq\\|\Phi_{N-1}\\|_{2}^{2}\left(1-\frac{4\pi b^{3}}{3}N/\Lambda\right)=\\|\Phi_{N-1}\\|_{2}^{2}(1-Y)$ (2.33) Here we used the fact that $\\|\Phi_{N-1}\\|^{2}_{2}=\int\bigg{\|}F_{1,\,i}\cdots F_{i-1,\,i}F_{i+1,\,i}\cdots F_{N,\,i}\bigg{\|}^{2}\prod_{j}dx_{j}.$ (2.34) Similarly, one can also prove that for $k\leq N$, $\displaystyle\\|\Phi_{k}\\|_{2}^{2}\geq\\|\Phi_{k-1}\\|_{2}^{2}(1-Y)$ (2.35) Next we are going to prove (2.25) in the case $m=1,k=N$, i.e., $\displaystyle\\|\Phi_{N}F_{i}^{-1}g_{1}\\|^{2}_{2}\leq 4M^{2}\Lambda^{-1}\\|\Phi_{N-1}\\|^{2}_{2}\,\\|g_{1}\\|^{2}_{2}$ (2.36) One can check that $F_{p}>F_{p,i}$ only when the following conditions are satisfied: 1. 1. $i<p$ 2. 2. $\|x_{i}-x_{p}\|\leq{\widetilde{R}}$, 3. 3. for any other $j<p$, $\|x_{j}-x_{p}\|$ is greater than ${\widetilde{R}}$, 4. 4. $T(R_{p})<1$, i.e. for any other $j<p$, $\|x_{j}-x_{p}\|>2{\widetilde{R}}$, i.e., $F_{p}>F_{p,i}\Rightarrow G_{p,i}\equiv\theta_{\widetilde{R}}(x_{p}-x_{i})\prod_{j<p,j\neq i}\bar{\theta}_{2{\widetilde{R}}}(x_{j}-x_{p})=1$ (2.37) On the other hand, using the fact that $f({\widetilde{R}})>\mbox{$\frac{1}{2}$}$, one obtains that if $F_{p}>F_{p,i}$, $F_{p,i}\prod_{j<p,j\neq i}\bar{\theta}_{2{\widetilde{R}}}(x_{j}-x_{p})>f({\widetilde{R}})\prod_{j<p,j\neq i}\bar{\theta}_{2{\widetilde{R}}}(x_{j}-x_{p})\geq\frac{1}{2}\prod_{j<p,j\neq i}\bar{\theta}_{2{\widetilde{R}}}(x_{j}-x_{p})$ Hence when $G_{p,i}=1$, we have $2MF_{p,i}G_{p,i}\geq M\geq F_{p}$, i.e., $F_{p}\leq\left\\{\begin{array}[]{ll}F_{p,i}&i>p\\\ F_{p,i}\bigg{(}1+(2M-1)G_{p,i}\bigg{)}&i<p\,.\end{array}\right.$ (2.38) By the definition of $G$’s, one can see that if $p,q>i$ and $p\neq q$, $G_{p,i}G_{q,i}=0.$ (2.39) Hence, we have that $\displaystyle\prod_{p>i}\bigg{(}1+(2M-1)G_{p,i}\bigg{)}\leq 2M$ (2.40) Combining (2.38) and (2.40), we have the upper bound on $\|\Phi_{N}F_{i}^{-1}\|$ as follows, $\displaystyle\bigg{\|}F_{1}\cdot F_{2}\cdots F_{N}F_{i}^{-1}\bigg{\|}^{2}\leq 4M^{2}\bigg{\|}F_{1,\,i}\cdots F_{i-1,\,i}F_{i+1,\,i}\cdots F_{N,\,i}\bigg{\|}^{2},$ (2.41) which implies the desired result (2.36) with (2.34). Furthermore, for any $m$-variable function $g_{m}(x_{i_{1}}\cdots x_{i_{m}})$ $\displaystyle\\|\Phi_{N}F^{-1}_{i_{1}}\cdots F^{-1}_{i_{m}}g_{m}\\|^{2}_{2}\leq(2M)^{2m}\Lambda^{-m}\\|\Phi_{N-m}\\|^{2}_{2}\\|g_{m}\\|^{2}_{2}$ (2.42) With the inequality (2.33) and the fact $F_{i}\leq M$ for any $i\leq N$, we get $\displaystyle\\|\Phi_{N}g_{m}\\|^{2}_{2}\leq$ $\displaystyle(2M^{2})^{2m}\\|\Phi_{N-m}\\|^{2}_{2}\\|g_{m}\\|^{2}_{2}$ (2.43) $\displaystyle\leq$ $\displaystyle(1-Y)^{-m}(2M^{2})^{2m}{\Lambda}^{-m}\\|\Phi_{N}\\|^{2}_{2}\\|g_{m}\\|^{2}_{2}$ Similarly, we can generalize this result to $m<k\leq N$ $\displaystyle\\|\Phi_{k}g_{m}\\|^{2}_{2}\leq$ $\displaystyle(2M^{2})^{2m}\\|\Phi_{k-m}\\|^{2}_{2}\\|g_{m}\\|^{2}_{2}$ (2.44) $\displaystyle\leq$ $\displaystyle(1-Y)^{-m}(2M^{2})^{2m}{\Lambda}^{-m}\\|\Phi_{k}\\|^{2}_{2}\\|g_{m}\\|^{2}_{2}$ Now we shall prove the upper bound on $\\|\Phi_{N}\\|^{2}_{2}$ with (2.44). Choosing $p=N$, with the bounds of $F_{p}$ in (2.21), we get that $\displaystyle\Phi_{N}^{2}\leq$ $\displaystyle F^{2}_{1}\cdot F^{2}_{2}\cdots F^{2}_{N-1}\left(1+\sum_{j<N}M^{2}\theta_{\widetilde{R}}(x_{j}-x_{N})\right)$ (2.45) $\displaystyle=$ $\displaystyle\Phi_{N-1}^{2}+\Phi_{N-1}^{2}\sum_{j<N}M^{2}\theta_{\widetilde{R}}(x_{j}-x_{N})$ Hence, using the inequalities (2.44)($m=1$) and (2.35), we obtain that $\displaystyle\\|\Phi_{N}\\|^{2}_{2}$ $\displaystyle\leq\\|\Phi_{N-1}\\|^{2}_{2}+(1-Y)^{-1}(2M^{2})^{2}Y\\|\Phi_{N-2}\\|^{2}_{2}$ $\displaystyle\leq\\|\Phi_{N-1}\\|^{2}_{2}(1+{\rm const.\,}Y)$ Putting (2.1) and (2.33) together, we obtain the relation between $\\|\Phi_{N}\\|$ and $\\|\Phi_{N-1}\\|$ $\\|\Phi_{N}\\|^{2}_{2}=\\|\Phi_{N-1}\\|^{2}_{2}(1+O(Y))$ (2.47) Similarly, for $k\leq N$ $\\|\Phi_{k}\\|^{2}_{2}=\\|\Phi_{k-1}\\|^{2}_{2}(1+O(Y))$ (2.48) Next, we shall prove (2.26), i.e., $\displaystyle\\|\Phi_{N}F^{-1}_{i^{\prime}}g_{2}\\|^{2}_{2}\leq\\|g_{2}\\|^{2}_{2}\Lambda^{-2}\\|\Phi_{N}\\|^{2}_{2}(1+{\rm const.\,}Y).$ (2.49) Using the inequalities $F_{i}\leq\left(1+\sum_{l<i}(M-1)\theta_{\widetilde{R}}(x_{l}-x_{i})\right)$ and (2.38), with the property of the $G$’s in (2.39), we get $\displaystyle\prod_{k\neq i^{\prime}}F_{k}$ $\displaystyle\leq\left(1+\sum_{l<i}(M-1)\theta_{\widetilde{R}}(x_{l}-x_{i})\right)\prod_{k\neq i^{\prime},i}F_{k\,,i}\cdot\bigg{(}1+\sum_{j>i}2MG_{j,i}\bigg{)}$ (2.50) Similarly, replacing $F_{p,i}$’s with $F_{p,i,i^{\prime}}$’s and using the fact that $G_{k,l}\leq\theta_{\widetilde{R}}(x_{k}-x_{l})$, we get $\displaystyle\Phi_{N}F^{-1}_{i^{\prime}}\leq$ $\displaystyle\prod_{k\neq i^{\prime},i}F_{k,i,i^{\prime}}\left(1+\sum_{l<i}M\theta_{\widetilde{R}}(x_{l}-x_{i})\right)$ $\displaystyle\times\bigg{(}1+\sum_{j>i}2M\theta_{\widetilde{R}}(x_{j}-x_{i})\bigg{)}\times\bigg{(}1+\sum_{j^{\prime}>i^{\prime}}2M\theta_{\widetilde{R}}(x_{j}^{\prime}-x_{i}^{\prime})\bigg{)}$ Expanding (2.1), multiplying $g_{2}(x_{i},x_{i^{\prime}})$ to each side and integrating them with $\prod_{k=1}^{N}dx_{k}$, with the result of (2.44, 2.48), we obtain that $\\|\Phi_{N}F^{-1}_{i^{\prime}}g_{2}\\|^{2}_{2}\leq\\|g_{2}\\|^{2}_{2}\Lambda^{-2}\\|\Phi_{N}\\|^{2}_{2}(1+{\rm const.\,}Y)$ (2.52) So far we proved some upper bounds of the expectation value of $\Phi_{N}$. Next we are going to prove the following lower bound on $\\|\Phi_{N}f^{-1}_{i,i^{\prime}}g_{2}\\|^{2}_{2}$: $\displaystyle\\|\Phi_{N}f^{-1}_{i,i^{\prime}}g_{2}\\|^{2}_{2}\geq\\|g_{2}\\|^{2}_{2}\Lambda^{-2}\\|\Phi_{N}\\|^{2}_{2}(1-{\rm const.\,}Y)$ (2.53) Here we denote $f_{i,i^{\prime}}=f(x_{i}-x_{i^{\prime}})$ ($i<i^{\prime}$). First, by the definition of $F_{i^{\prime}}$, one can see that $F_{i^{\prime}}=f(x_{i}-x_{i^{\prime}})$ when $\prod_{k<i^{\prime},k\neq i}\bar{\theta}_{b}(x_{k}-x_{i}^{\prime})=1$, i.e., $F^{2}_{i^{\prime}}\geq f(x_{i}-x_{i^{\prime}})^{2}\left(1-\sum_{k<i^{\prime},k\neq i}\theta_{b}(x_{k}-x_{i^{\prime}})\right)$ (2.54) Using this inequality and (2.42) with $m=3$, $i_{1}=i$, $i_{2}=i^{\prime}$ and $i_{3}=k$, we obtain that $\displaystyle\left\\|\Phi_{N}f^{-1}_{i,i^{\prime}}g_{2}\right\\|^{2}_{2}\geq\left\\|\Phi_{N}F^{-1}_{i^{\prime}}g_{2}\right\\|^{2}_{2}-{\rm const.\,}Y\left\\|g_{2}\right\\|^{2}_{2}\Lambda^{-2}\left\\|\Phi_{N}\right\\|^{2}_{2}$ (2.55) Then with the lower bound on $F_{i}^{2}$ in (2.21), i.e., $F^{2}_{i}\geq 1-\sum_{k<i}\theta_{b}(x_{k}-x_{i})$, we obtain that $\displaystyle\\|\Phi_{N}f^{-1}_{i,i^{\prime}}g_{2}\\|^{2}_{2}\geq\\|\Phi_{N}F^{-1}_{i}F^{-1}_{i^{\prime}}g_{2}\\|^{2}_{2}-{\rm const.\,}Y\\|g_{2}\\|^{2}_{2}\Lambda^{-2}\\|\Phi_{N}\\|^{2}_{2}$ (2.56) Again, using the bound on $F_{p}$ in (2.31), we see that $\displaystyle\Phi_{N}F^{-1}_{i}F^{-1}_{i^{\prime}}\geq$ $\displaystyle\prod_{k\neq i,i^{\prime}}F_{k,i,i^{\prime}}\left(1-\sum_{l<i}\theta_{b}(x_{l}-x_{i})\right)\times\left(1-\sum_{l^{\prime}<{i^{\prime}},\,\,l^{\prime}\neq i}\theta_{b}(x_{l^{\prime}}-x_{i^{\prime}})\right)$ Then using (2.44) and (2.48), we arrive at the desired result (2.53). So far, we have proved the inequalities we need for calculating the value of $\langle\Phi_{N}\|\sum_{i,j}v^{a}(x_{i}-x_{j})\|\Phi_{N}\rangle$. Then we need to calculate $\nabla_{i}\Phi_{N}$. We denote $i_{p}$ as the particle satisfying ${i_{p}}<p$ and $\|x_{i_{p}}-x_{p}\|=r_{p}$ and $n^{r}_{p}$ as the unit vector in the direction of $x_{p}-x_{i_{p}}$. Similarly, denote $j_{p}$ as the particle satisfying ${j_{p}}<p$ and $\|x_{j_{p}}-x_{p}\|=R_{p}$ and $n^{R}_{p}$ as the unit vector in the direction of $x_{p}-x_{j_{p}}$. We remark that such $i_{p}$ or $j_{p}$ may not exist in some cases, but we do define them as $0$. We denote $\nabla_{0}F_{p}=0$. Recall the definition of $F_{p}$ in (2.16). We have $\displaystyle-\nabla_{p}F_{p}$ $\displaystyle=\nabla_{i_{p}}F_{p}+\nabla_{j_{p}}F_{p}$ (2.57) $\displaystyle\nabla_{i_{p}}F_{p}$ $\displaystyle=-n^{r}_{p}f\,^{\prime}(r_{p})\bigg{(}\Theta^{in}_{p}+\Theta^{-}_{p}(1-T(R_{p}))+\Theta^{+}_{p}\bigg{)}$ $\displaystyle\nabla_{j_{p}}F_{p}$ $\displaystyle=-n^{R}_{p}\bigg{(}\Theta^{out}_{p}f\,^{\prime}(R_{p})+\Theta^{-}_{p}T(R_{p})f^{\prime}(R_{p})+\Theta^{-}_{p}T\,^{\prime}(R_{p})\big{(}f(R_{p})-f(r_{p})\big{)}\bigg{)}$ Here $\Theta^{+}_{p}$ is the function of $(x_{1}\cdots x_{N})$ which is defined as $\Theta^{+}_{p}\equiv\big{[}1-\Theta^{in}-\Theta^{out}\big{]}\cdot h\big{[}f(R_{p})-f(r_{p})\big{]}$ (2.58) and $\Theta^{-}_{p}$ is defined as $\Theta^{-}_{p}\equiv\big{[}1-\Theta^{in}-\Theta^{out}\big{]}\cdot h\big{[}f(r_{p})-f(R_{p})\big{]}$ (2.59) Here $h$ is the Heaviside step function. By the definition of $\Phi_{N}$, we obtain that $\displaystyle\frac{\|\nabla_{p}\Phi_{N}\|^{2}}{\|\Phi_{N}\|^{2}}=\left\|-F_{p}^{-1}\nabla_{i_{p}}F_{p}-F_{p}^{-1}\nabla_{j_{p}}F_{p}+\sum_{q,i_{q}=p}F_{q}^{-1}\nabla_{p}F_{q}+\sum_{q,j_{q}=p}F_{q}^{-1}\nabla_{p}F_{q}\right\|^{2}$ Then with (2.57), we have that $\displaystyle\sum_{p}\|\nabla_{p}\Phi_{N}\|^{2}\leq$ $\displaystyle 2\|\Phi_{N}\|^{2}\sum_{p}F_{p}^{-2}\bigg{(}\|f\,^{\prime}(r_{p})\|^{2}\big{(}\Theta^{in}_{p}+\Theta^{-}_{p}\|1-T(R_{p})\|^{2}+\Theta^{+}_{p}\big{)}$ $\displaystyle+\|T(R_{p})f\,^{\prime}(R_{p})\|^{2}\Theta^{-}_{p}+\|T\,^{\prime}(R_{p})\|^{2}\|f(R_{p})-f(r_{p})\|^{2}\Theta^{-}_{p}$ $\displaystyle+\|T(R_{p})\|\cdot\|1-T(R_{p})\|\cdot\|f\,^{\prime}(r_{p})\|\cdot\|f\,^{\prime}(R_{p})\|\Theta^{-}_{p}+\|f\,^{\prime}(R_{p})\|^{2}\Theta^{out}_{p}\bigg{)}$ $\displaystyle+2\|\Phi_{N}\|^{2}\sum_{k<p<q}F_{p}^{-1}F_{q}^{-1}\bigg{(}\|\nabla_{k}F_{p}\|\cdot\|\nabla_{p}F_{q}\|+\|\nabla_{k}F_{p}\|\cdot\|\nabla_{k}F_{q}\|\bigg{)}$ Because $0\leq T\leq 1$ and $i_{p}\neq j_{p}$, one can easily prove that for any fixed $p$, $\displaystyle\bigg{(}\|f\,^{\prime}(r_{p})\|^{2}\big{(}\Theta^{in}_{p}+\Theta^{-}_{p}\|1-T(R_{p})\|^{2}+\Theta^{+}_{p}\big{)}+\|f\,^{\prime}(R_{p})\|^{2}\Theta^{out}_{p}$ $\displaystyle+$ $\displaystyle\|T(R_{p})f\,^{\prime}(R_{p})\|^{2}\Theta^{-}_{p}+\|T(R_{p})\|\cdot\|1-T(R_{p})\|\cdot\|f\,^{\prime}(r_{p})\|\cdot\|f\,^{\prime}(R_{p})\|\Theta^{-}_{p}\bigg{)}$ $\displaystyle\leq$ $\displaystyle\sum_{k:k<p}f\,^{\prime}(\|x_{p}-x_{k}\|)^{2},$ and $\|T\,^{\prime}(R_{p})\|^{2}\|f(R_{p})-f(r_{p})\|^{2}\Theta^{-}_{p}\leq M^{2}\sum_{k:k<p}\left(T\,^{\prime}(\|x_{p}-x_{k}\|)^{2}\sum_{j:j\neq k,\,p}\theta_{\widetilde{R}}(x_{j}-x_{p})\right)$ (2.60) Hence, we obtain that $\displaystyle\langle\Phi_{N}\|H_{N}\|\Phi_{N}\rangle\leq 2\sum_{i<j}{\int\|\Phi_{N}\|^{2}F_{j}^{-2}\bigg{(}\mbox{$\frac{1}{2}$}f(x_{i}-x_{j})^{2}[v(x_{i}-x_{j})]_{+}+f\,^{\prime}(x_{i}-x_{j})^{2}\bigg{)}}$ $\displaystyle-2\sum_{i<j}{\int\|\Phi_{N}\|^{2}f^{-2}(x_{i}-x_{j})\bigg{(}\mbox{$\frac{1}{2}$}f^{2}(x_{i}-x_{j})\bigg{\|}[v(x_{i}-x_{j})]_{-}\bigg{\|}\bigg{)}}$ $\displaystyle+2\sum_{i<j}\int\|\Phi_{N}\|^{2}M^{2}\sum_{k<p}\left(T\,^{\prime}(\|x_{p}-x_{k}\|)^{2}\sum_{j:j\neq k,\,p}\theta_{\widetilde{R}}(x_{j}-x_{p})\right)$ (2.61) $\displaystyle+2\sum_{k<p<q}\int\|\Phi_{N}\|^{2}F_{p}^{-1}F_{q}^{-1}\bigg{(}\|\nabla_{k}F_{p}\|\cdot\|\nabla_{p}F_{q}\|+\|\nabla_{k}F_{p}\|\cdot\|\nabla_{k}F_{q}\|\bigg{)}$ Here $[\cdot]_{+}$ and $[\cdot]_{-}$ denote the positive and negative part, respectively and we used the fact that $F_{j}\leq f(x_{i}-x_{j})$ when $i<j$ and $\|x_{i}-x_{j}\|\leq{\widetilde{R}}$, which implies that $[v(x_{i}-x_{j})]_{+}\leq F_{j}^{-2}f(x_{i}-x_{j})^{2}[v(x_{i}-x_{j})]_{+}$ (2.62) With the results in (2.52) and (2.53), we can obtain the upper bound on the main part of $\langle\Phi_{N}\|H_{N}\|\Phi_{N}\rangle$, i.e., $\displaystyle 2\sum_{i<j}\int\|\Phi_{N}\|^{2}\times$ $\displaystyle\frac{1}{2}\bigg{(}F_{j}^{-2}\left[f^{2}(x_{i}-x_{j})[v(x_{i}-x_{j})]_{+}+f\,^{\prime}(x_{i}-x_{j})^{2}\right]-\frac{f^{2}(x_{i}-x_{j})}{f^{2}(x_{i}-x_{j})}\bigg{\|}[v(x_{i}-x_{j})]_{-}\bigg{\|}\bigg{)}$ $\displaystyle\leq 4\pi aN^{2}/\Lambda(1+{\rm const.\,}Y)\\|\Phi_{N}\\|^{2}_{2}$ (2.63) With the definition of $T$ in (2.12) and (2.43), we obtain that the third line of (2.61) is bounded as ${\rm const.\,}aN^{2}Y\\|\Phi_{N}\\|^{2}_{2}/\Lambda$. For the other terms, we have $\|\nabla_{i_{p}}F_{p}\|\leq\|f\,^{\prime}(\|x_{i_{p}}-x_{p}\|)\|,\,\,\,\,\|\nabla_{j_{p}}F_{p}\|\leq\|f\,^{\prime}(\|x_{j_{p}}-x_{p}\|)\|+MT\,^{\prime}(\|x_{j_{p}}-x_{p}\|)$ (2.64) Hence, with the inequality (2.42), we can prove that the last line in (2.61) are bounded as ${\rm const.\,}N^{3}\Lambda^{-2}(K+L)^{2}\\|\Phi_{N}\\|^{2}_{2}$ (2.65) Here $K$ and $L$ are defined as follows $\displaystyle K\equiv\int_{{\mathbb{R}}^{3}}\|f\,^{\prime}\left(\|x-y\|\right)\|dy\,\,\,\,\,\,\,L\equiv\int_{{\mathbb{R}}^{3}}T\,^{\prime}\left(\|x-y\|\right)dy.$ (2.66) Note that $K$ and $L$ are independent of $x$. By the definitions of $f$ in (2.8) and $T$ in (2.12), we get that $K=O(ab),\,\,\,\ L=O({\widetilde{R}}b)=O(ab)$ (2.67) Hence we obtain that the last line in (2.61) are bounded by ${\rm const.\,}aN^{2}Y^{2}$. Combining this result with (2.63), we get the following result, $\frac{\langle\Phi_{N}\|H_{N}\|\Phi_{N}\rangle}{\\|\Phi_{N}\\|^{2}_{2}}\leq 4\pi aN^{2}/\Lambda(1+{\rm const.\,}Y)$ (2.68) At last, by choosing $\Psi=\Phi_{N}$, we arrive at the desired result (2.1), which implies Theorem 1. ∎ ### 2.2 Proof of Theorem Two ###### Proof. Following the ideas in [7], we need to replace the hard potential by a soft potential at the expense of local kinetic energy. This method has been used in many papers on dilute bose or fermi gases [7, 8, 6, 2, 3]. But in this method the kinetic energy of particle $i$ only can be used for the hard-soft potential replacement between the particle $i$ and one other $j$ (the nearest particle [7]). In our case that $v^{a}$ is partly negative, we can not ignore the potential between $i$ and other $k$’s for the lower bound on the energy. To solve this problem, we begin with separating the whole Hamiltonian into two parts, (1) The Hamiltonian of the energy when two particles are close to each other and they are far away from the others. (2) The Hamiltonian of the remaining energy. In the remainder of this section, we prove that the first part is greater than $4\pi aN^{2}\Lambda^{-1}(1-O(a^{3}\varrho)^{1/17})$ and the second part is non-negative. Another important property Lieb and Yngvason used in [7] is the superadditivity of the ground energy $E(n,\ell)$ of $n$ particles in $[0,\ell]^{3}$ with Neumann boundary condition, i.e., $E(n+n^{\prime},\ell)\geq E(n,\ell)+E(n^{\prime},\ell)$ (2.69) This property is trivial in the case $v^{a}\geq 0$. In our proof, we are not going to prove any similar property, actually we only need the property (2.141) that for fixed $\ell$, when $n$ is larger than $4\varrho\ell^{3}$, the energy/particle is greater than $8\pi a\varrho$, as in (2.62) of [5], i.e., $E(n,\ell)/n\geq 8\pi a\varrho(1-{\rm const.\,}(a^{3}\varrho)^{1/17})$ (2.70) which will be proved in Lemma 1. Choosing $R=a(a^{3}\varrho)^{-5/17}\geq 2R_{0}a,$ (2.71) we define $F_{i,j}$ for $i\neq j$ as follows: $F_{i,j}=\theta_{R}(x_{i}-x_{j})\prod_{k\neq i,j}\bar{\theta}_{2R}(x_{i}-x_{k})$ (2.72) Here $\theta_{R}$ is the characteristic function of the open set $\|x\|<R$, and $\bar{\theta}_{R}=1-\theta_{R}$. We note that $F_{i,j}\neq F_{j,i}$ and $F_{i,j}$ is equal to $1$ only when $x_{j}$ is close to $x_{i}$, but the other $x_{k}$’s are not. It is easy to check that $\sum_{i:i\neq j}F_{i,j}\leq 1$, so $-\nabla_{j}\sum_{i:i\neq j}F_{i,j}\nabla_{j}\leq-\Delta_{j}$ (2.73) for any fixed $x_{1},\cdots x_{j-1},x_{j+1},\cdots,x_{N}$. Then we denote $v_{+}^{a}$ and $v_{-}^{a}$ as scaled potentials as follows, $v_{+}^{a}(r)=a^{-2}v_{+}(r/a),\,\,\,v_{-}^{a}(r)=a^{-2}v_{-}(r/a)$ (2.74) Choosing $Y=(a^{3}\varrho)^{1/17}$ (2.75) and $\varepsilon$ satisfying $3\cdot\bigg{(}\min\big{\\{}1,SL[v_{+}]\big{\\}}\bigg{)}^{-1}\cdot Y=\varepsilon<\frac{t}{2(1+t)},$ (2.76) with the definition $v_{\varepsilon}^{a}\equiv v^{a}-\varepsilon v^{a}_{+},$ (2.77) we separate the Hamiltonian $H_{N}$ as follows $\displaystyle H_{N}=$ (2.78) $\displaystyle(1-\varepsilon)\sum_{j}-\nabla_{j}\sum_{i}F_{i,j}\nabla_{j}+\sum_{i\neq j}F_{ij}\frac{v_{\varepsilon}^{a}}{2}(x_{i}-x_{j})-\varepsilon\sum_{j}\Delta_{j}+\varepsilon\sum_{i\neq j}\frac{v^{a}_{+}}{2}(x_{i}-x_{j})$ $\displaystyle+(1-\varepsilon)\sum_{j}-\nabla_{j}(1-\sum_{i}F_{i,j})\nabla_{j}+\sum_{i\neq j}(1-F_{ij})\frac{v_{\varepsilon}^{a}}{2}(x_{i}-x_{j})$ First, we claim the following Lemma 1, which will be proved in next section. ###### Lemma 1. Define $Y$, $F_{i,j}$, $\varepsilon$, $v^{a}_{\varepsilon}$ and $R$ as in (2.75), (2.72), (2.76), (2.77) and (2.71) respectively. There exists $C$ depending only on $v$ such that $\displaystyle H^{\prime\prime}\equiv$ (2.79) $\displaystyle(1-\varepsilon)\sum_{j}-\nabla_{j}\sum_{i}F_{i,j}\nabla_{j}+\sum_{i\neq j}F_{ij}\frac{v_{\varepsilon}^{a}}{2}(x_{i}-x_{j})-\varepsilon\sum_{j}\Delta_{j}+\varepsilon\sum_{i\neq j}\frac{v^{a}_{+}}{2}(x_{i}-x_{j})$ $\displaystyle\geq 4\pi aN^{2}/\Lambda(1-CY)$ Hence, to obtain Theorem 2, it only remains to prove that the last line of (2.78), as an operator, is bounded from below by zero, i.e., $(1-\varepsilon)\sum_{j}-\nabla_{j}(1-\sum_{i}F_{i,j})\nabla_{j}+\sum_{i\neq j}(1-F_{ij})\mbox{$\frac{1}{2}$}v^{a}_{\varepsilon}(x_{i}-x_{j})\geq 0$ By the assumptions $\varepsilon<\,t\,(2+2t)^{-1}$, we have $v_{\varepsilon}^{a}\geq\frac{2+t}{2+2\,t}v^{a}_{+}+v^{a}_{-}.$ Hence, it remains to prove that $0\leq H_{N}^{\prime}\equiv\frac{2+t}{2+2\,t}\sum_{j}-\nabla_{j}(1-\sum_{i}F_{i,j})\nabla_{j}+\frac{1}{2}\sum_{i\neq j}(1-F_{ij})\left(\frac{2+t}{2+2\,t}v^{a}_{+}+v^{a}_{-}\right)(x_{i}-x_{j})$ Because $\lim_{N\to\infty}E(N,\Lambda)/N$ exists, for proving Theorem 2, we can assume that $N$ is even, i.e., $N=2N_{1}$. Consider any partition $P=(\pi_{1},\pi_{2})$ of ${1,...,N}$ into two disjoint sets with $N_{1}$ integers in $\pi_{1}$ and $\pi_{2}$ respectively. For each $P$, we define that $\displaystyle H_{P}=H_{(\pi_{1},\pi_{2})}\equiv$ $\displaystyle\frac{2+t}{1+t}\sum_{j\in\pi_{1}}-\nabla_{j}(1-\sum_{i\neq j}F_{i,j})\nabla_{j}+\sum_{i,j\in\pi_{1}}(1-F_{i,j})\mbox{$\frac{1}{2}$}v_{1,1}^{a}(x_{i}-x_{j})$ $\displaystyle+$ $\displaystyle\sum_{i\in\pi_{2},j\in\pi_{1}}(1-F_{i,j})\mbox{$\frac{1}{2}$}v^{a}_{2,1}(x_{i}-x_{j})+\sum_{i,j\in\pi_{2}}(1-F_{i,j})\mbox{$\frac{1}{2}$}v^{a}_{2,2}(x_{i}-x_{j})$ Here we denote $v^{a}_{\alpha,\beta}$ as the interaction potential between particles in $\pi_{\alpha}$ and $\pi_{\beta}$, which are chosen as $v_{1,1}^{a}=v^{a}_{2,2}=\frac{t}{1+t}v^{a}_{+}\geq 0,\,\,\,\,v^{a}_{2,1}=\frac{4}{1+t}v^{a}_{+}+4v^{a}_{-},$ (2.80) so $\frac{1}{4}\big{(}v^{a}_{1,1,}+v^{a}_{2,1}+v^{a}_{2,2}\big{)}=\frac{2+t}{2+2\,t}v^{a}_{+}+v^{a}_{-}\leq v^{a}_{\varepsilon}.$ It is easily to check that $H^{\prime}_{N}=\sum_{P}H_{P}/\sum_{P}1$ (2.81) Hence, to obtain $H^{\prime}_{N}\geq 0$, it remains to prove that for $\forall P$, $H_{P}\geq 0$. Because there is no kinetic energy of particles in $\pi_{2}$, we can fix the configuration of $x_{i}$’s with $i\in\pi_{2}$. Since permutation of the labels in $\pi_{1}$ and $\pi_{2}$ is irrelevant, we assume that $\pi_{1}=\\{1,\cdots,N_{1}\\}$, $\pi_{2}=\\{N_{1}+1,\cdots,N\\}$. As we can see $v^{a}_{2,1}$ is the only partly negative component in $H_{P}$. For fixed $\pi_{2}$ particles, we can write $v^{a}_{2,1}(x_{j}-x_{i})$ as $v^{a}_{2,1}(x_{j}-x_{i})=v^{a}_{2,1}(x_{j}-x_{i})(1-\chi_{A}(x_{i}))+v^{a}_{2,1}(x_{j}-x_{i})\chi_{A}(x_{i})$ (2.82) Here $\chi_{A}$ is the characteristic function of $A$, which is a subset of $[0,L]^{3}$ (2.99). We shall show $A$ is the area where the density of $\pi_{2}$ particles is less than some fixed number. To obtain $H_{P}\geq 0$, our strategy is to prove that 1. 1. The total energy of the interaction potential $v_{1,1}^{a}$ and $v^{a}_{2,2}$ cancels out the negative part of $v^{a}_{2,1}(1-\chi_{A})$. 2. 2. The total kinetic energy and the positive part of $v^{a}_{2,1}$ cancels out the negative part of $v^{a}_{2,1}\chi_{A}$. To make the strategy more clear, we shall define $A$ where the density of $\pi_{2}$ particles is less than some fixed number. First we divide the cubic box $[0,L]^{3}$ into small cubes $B_{n}$ ($n\in{\mathbb{N}}$) of side length $\ell$, with $\ell=\mbox{$\frac{1}{2}$}r_{1}a.$ Then, with fixed $x_{i}$’s, $i\in\pi_{2}$, for any $x\in[0,L]^{3}$, we define the $G(x)$ as the set of $i$’s which satisfy $i\in\pi_{2}$ and $\|x_{i}-x\|\leq R_{0}a$, i.e., $\displaystyle G(x)\equiv\\{i\in\pi_{2}:\|x_{i}-x\|\leq R_{0}a\\}$ (2.83) We denote $\|G(x)\|$ as the number of the elements of $G(x)$. We denote $d(x,B_{n})$ as the distance between the cube $B_{n}\subset{\mathbb{R}}^{3}$ and $x\in{\mathbb{R}}^{3}$. Since $\|G(y)\|$ is uniformly bounded ($\|G(y)\|\leq N_{1}$), there must exist a point $X(B_{n})\in{\mathbb{R}}^{3}$ satisfying $d(X(B_{n}),B_{n})\leq 2R_{0}a$ and $\displaystyle\|G(X(B_{n}))\|=\max\\{\|G(y)\|:d(y,B_{n})\leq 2R_{0}a\\}$ (2.84) We define $G(B_{n})\equiv G(X(B_{n}))$. We are going to prove that there exists $n_{1}\in{\mathbb{N}}$ depending on $R_{0}/r_{1}$ such that 1. 1. The total energy of the interaction potential $v_{1,1}^{a}$ and $v^{a}_{2,2}$ cancels out the negative parts of $v^{a}_{2,1}(x_{j},x_{i})$’s when $x_{i}$ is in a cube $B_{n}$ such that $\|G(B_{n})\|>n_{1}$. 2. 2. The total kinetic energy and the positive part of $v^{a}_{2,1}$ cancel out the negative part of the remaining $v^{a}_{2,1}$’s. First, we derive the lower bound on the total energy of $v^{a}_{2,2}$, i.e. (2.89, 2.91). With the definition of $G(B_{n})=G(X(B_{n}))$, we know that the set $\\{x_{k}:k\in G(B_{n})\\}$ can be covered by a sphere of radius $R_{0}a$. So the number of the cubes which one need to cover this set is less than ${\rm const.\,}(R_{0}/r_{1})^{3}$. We denote these cubes as $B_{n_{1}}\cdots B_{n_{m}}$ $(m\leq{\rm const.\,}(R_{0}/r_{1})^{3})$ and assume the number of $i$’s satisfying $i\in G(B_{n})$ and $x_{i}\in B_{n_{k}}$ is $a_{n_{k}}$. Because the side length of $B_{n_{k}}$ is equal to $r_{1}a/2$, the distance between the two particles in the same cube is no more than $\frac{\sqrt{3}}{2}r_{1}a<r_{1}a$. Hence we have $\displaystyle\sum_{i,j\in G(B_{n})}\theta_{r_{1}a}(x_{i}-x_{j})\geq$ $\displaystyle\sum_{k=1}^{m}\left[(a_{n_{k}})^{2}-(a_{n_{k}})\right]$ (2.85) $\displaystyle\geq$ $\displaystyle\frac{(\sum_{k=1}^{m}a_{n_{k}})^{2}}{m}-(\sum_{k=1}^{m}a_{n_{k}})$ $\displaystyle\geq$ $\displaystyle{\rm const.\,}(R_{0}/r_{1})^{-3}\|G(B_{n})\|^{2}-\|G(B_{n})\|$ Hence, we obtain that there exist $n_{1}\geq 3$ and $n_{1},n_{2}={\rm const.\,}(R_{0}/r_{1})^{3}$ such that when $\|G(B_{n})\|\geq n_{1}$, $\sum_{i,j\in G(B_{n})}\theta_{r_{1}a}(x_{i}-x_{j})\geq\frac{1}{n_{2}}\|G(B_{n})\|^{2},$ (2.86) which implies $\sum_{i,j\in G(B_{n})}v^{a}_{2,2}(x_{i}-x_{j})\geq\frac{t\lambda_{+}a^{-2}}{(1+t)n_{2}}\|G(B_{n})\|^{2}$ (2.87) Here, we used (2.80) and (2.74), i.e., $v^{a}_{2,2}(r)=\frac{t}{(1+t)}v_{+}^{a}(r)=\frac{t}{(1+t)}a^{-2}v_{+}(r/a)$ (2.88) Again, with the fact that the set $\\{x_{k}:k\in G(B_{n})\\}$ can be covered with a sphere of diameter $2R_{0}a\leq R$, one can see that if $i\in G(B_{n})$ and $\|G(B_{n})\|\geq 3$, we have $F_{i,j}=0$ for any $j\neq i$. Hence we obtain that, for any fixed $B_{n}$ satisfying $\|G(B_{n})\|\geq n_{1}$, $\sum_{i,j\in G(B_{n})}(1-F_{i,j})v^{a}_{2,2}(x_{i}-x_{j})=\sum_{i,j\in G(B_{n})}v^{a}_{2,2}(x_{i}-x_{j})\geq\frac{t\lambda_{+}a^{-2}}{(1+t)n_{2}}\|G(B_{n})\|^{2}$ (2.89) Then, we are going to sum up all the cubes satisfying $\|G(B_{n})\|\geq n_{1}$. It is easy to see that $d(x_{i},B_{n})\leq 3R_{0}a,\,\,\,\,{\rm{for}}\,\,\,\,i\in G(B_{n}),$ (2.90) which implies that for any fixed $i\in{\pi_{2}}$, the number of cubes $B_{n}$’s satisfying $i\in G(B_{n})$ is less than some constant $n_{3}$, which is less than ${\rm const.\,}(R_{0}/r_{1})^{3}$. Hence, summing up all the blocks satisfying $\|G(B_{n})\|\geq n_{1}$, with the inequality (2.89), we get that $\displaystyle\sum_{i,j\in\pi_{2}}(1-F_{i,j})v^{a}_{2,2}(x_{i}-x_{j})$ $\displaystyle\geq\sum_{n:\|G(B_{n})\|\geq n_{1}}\sum_{i,j\in G(B_{n})}(1-F_{i,j})v^{a}_{2,2}(x_{i}-x_{j})$ (2.91) $\displaystyle\geq\sum_{n:\|G(B_{n})\|\geq n_{1}}\frac{t\lambda_{+}a^{-2}}{(1+t)n_{2}n_{3}}\|G(B_{n})\|^{2}$ Second, we derive the lower bound on the interaction potential between particles in $\pi_{1}$. Because the distance between any two points in the same cube is less than $r_{1}a$, we have $v_{1,1}^{a}(x_{i}-x_{j})\geq a^{-2}\lambda_{+}t(1+t)^{-1}$ when $i,j\in\pi_{1}$ and $x_{i},x_{j}\in B_{n}$, i.e., $\sum_{i,j\in\Pi_{1}(B_{n})}v_{1,1}^{a}(x_{i}-x_{j})\geq\frac{a^{-2}t\lambda_{+}}{1+t}\bigg{(}\|\Pi_{1}(B_{n})\|^{2}-\|\Pi_{1}(B_{n})\|\bigg{)}$ (2.92) Here $\Pi_{1}(B_{n})$ is defined as the set of $i$’s such that $i\in\pi_{1}$ and $x_{i}\in B_{n}$ and $\|\Pi_{1}(B_{n})\|$ is the number of the elements of $\Pi_{1}(B_{n})$. Furthermore, if $x_{i}\in B_{n}$ and $\|G(B_{n})\|\geq 1$, there must be a $k\in\pi_{2}$ satisfying $\|x_{i}-x_{k}\|\leq 4R_{0}a\leq 2R$, hence $F_{i,j}=0$ for any other $j\in\pi_{1}$. Using this result, for any $B_{n}$ satisfying $\|G(B_{n})\|\geq 1$, we have that $\sum_{i,j\in\Pi_{1}(B_{n})}(1-F_{i,j})v_{1,1}^{a}(x_{i}-x_{j})\geq\frac{t\lambda_{+}a^{-2}}{1+t}\bigg{(}\|\Pi_{1}(B_{n})\|^{2}-\|\Pi_{1}(B_{n})\|\bigg{)}$ (2.93) At last, we derive the lower bound on $v^{a}_{2,1}$. By the definitions of $\|G(B_{n})\|$ and $v^{a}_{2,1}$, we have that $\forall x\in B_{n}$, $\sum_{i\in\pi_{2}}[v^{a}_{2,1}]_{-}(x-x_{i})\geq-4\lambda_{-}a^{-2}\|G(B_{n})\|.$ Here we denote $[v^{a}_{2,1}]_{-}$ as the negative part of $v^{a}_{2,1}$ which is equal to $4[v^{a}]_{-}$. With the facts $0\geq 4[v^{a}]_{-}\geq-4\lambda_{-}a^{-2}$ and $0\leq F_{i,j}\leq 1$, we have the following inequality $\sum_{j\in\Pi_{1}(B_{n}),\,\,i\in\pi_{2}}(1-F_{i,j})[v^{a}_{2,1}]_{-}(x_{i}-x_{j})\geq-4\lambda_{-}a^{-2}\cdot\|\Pi_{1}(B_{n})\|\cdot\|G(B_{n})\|$ (2.94) One can check that if $\|G(B_{n})\|\geq n_{1}$ and $\lambda_{+}\geq(1+t^{-1})\lambda_{-}\cdot\max\\{2\sqrt{n_{2}n_{3}},\frac{n_{2}n_{3}}{4n_{1}}\\}\sim{\rm const.\,}(1+t^{-1})\lambda_{-}(R_{0}/r_{1})^{3},$ (2.95) the sum of the right sides of (2.93) and (2.94) is bounded from below as follows, $\displaystyle-4\lambda_{-}\cdot\|\Pi_{1}(B_{n})\|\cdot\|G(B_{n})\|+\frac{t\lambda_{+}}{1+t}\bigg{(}\|\Pi_{1}(B_{n})\|^{2}-\|\Pi_{1}(B_{n})\|\bigg{)}$ (2.96) $\displaystyle\geq$ $\displaystyle-\frac{t\lambda_{+}}{(1+t)n_{2}n_{3}}\|G(B_{n})\|^{2}$ Hence, with (2.93) and (2.94), we obtain that if (2.95) holds and $\|G(B_{n})\|\geq n_{1}$, $\displaystyle 0\leq$ $\displaystyle\frac{t}{1+t}\frac{\lambda_{+}a^{-2}}{n_{2}n_{3}}\|G(B_{n})\|^{2}+\sum_{i,j\in\Pi_{1}(B_{n})}(1-F_{i,j})v_{1,1}^{a}(x_{i}-x_{j})$ (2.97) $\displaystyle+$ $\displaystyle\sum_{j\in\Pi_{1}(B_{n}),\,\,i\in\pi_{2}}(1-F_{i,j})[v^{a}_{2,1}]_{-}(x_{i}-x_{j}),$ Then summing up all the $B_{n}$’s satisfying $\|G(B_{n})\|>n_{1}$, with (2.91) and $v_{11}\geq 0$, we obtain that as long as (2.95) holds, $\displaystyle 0\leq$ $\displaystyle\sum_{i,j\in\pi_{2}}(1-F_{i,j})\mbox{$\frac{1}{2}$}v^{a}_{2,2}(x_{i}-x_{j})+\sum_{i,j\in\pi_{1}}(1-F_{i,j})\mbox{$\frac{1}{2}$}v_{1,1}^{a}(x_{i}-x_{j})$ $\displaystyle+\sum_{j\in\pi_{1},i\in\pi_{2}}(1-F_{i,j})\mbox{$\frac{1}{2}$}[v^{a}_{2,1}]_{-}(x_{i}-x_{j})\big{(}1-\chi_{A}(x_{j})\big{)}$ Here $A$ is defined as the set $\cup_{\|G(B_{n})\|\leq n_{1}}B_{n}$. $A=\cup_{\|G(B_{n})\|\leq n_{1}}B_{n}$ (2.99) So far, we proved the interaction potential between particles of the same groups cancels out the negative part of the $v^{a}_{2,1}(1-\chi_{A})$ term in (2.82). We shall show that the kinetic energy and the positive part of $v^{a}_{2,1}$ cancel out the remaining negative part of $v^{a}_{2,1}$. For the other terms in the Hamiltonian $H_{P}$, we claim that as long as $SL[4\,n_{1}(v+tv_{-})]\geq 0$ (2.100) we have $\displaystyle 0\leq$ $\displaystyle\frac{1}{2}\sum_{j\in\pi_{1},i\in\pi_{2}}(1-F_{i,j})\bigg{(}[v^{a}_{2,1}]_{+}(x_{i}-x_{j})+[v^{a}_{2,1}]_{-}(x_{i}-x_{j})\chi_{A}(x_{j})\bigg{)}$ $\displaystyle+\frac{2+t}{1+t}\sum_{j\in\pi_{1}}-\nabla_{j}\left(1-\sum_{i}F_{i,j}\right)\nabla_{j}$ As we can see that (2.2) and (2.2) implies that $H_{P}\geq 0$ when $SL[4n_{1}(v+tv_{-})]\geq 0$ and $(\ref{lambda+-})$ holds, i.e., $\lambda_{+}\geq{\rm const.\,}(1+t^{-1})\lambda_{-}(R_{0}/r_{1})^{3}$, which completes the proof of Theorem 2. To prove (2.2), we only need to prove the following operator inequality, for any fixed $x_{2},\cdots,x_{N}$, $\displaystyle 0\leq$ $\displaystyle-\frac{2+t}{1+t}\nabla_{1}(1-\sum_{i=2}^{N}F_{i,1})\nabla_{1}$ $\displaystyle+\mbox{$\frac{1}{2}$}\sum_{j\in\pi_{2}}(1-F_{j,1})\bigg{(}[v^{a}_{2,1}]_{+}(x_{1}-x_{j})+[v^{a}_{2,1}]_{-}(x_{1}-x_{j})\chi_{A}(x_{1})\bigg{)}$ First, if $[v^{a}_{2,1}]_{-}(x_{1}-x_{j})\chi_{A}(x_{1})\neq 0$, then $d(B^{x_{1}}_{n},x_{j})\leq R_{0}a$, here the $B^{x_{1}}_{n}$ is the cube where $x_{1}$ is. We obtain that $j\in\pi_{2}^{\prime}\subset\pi_{2}$, here $\pi^{\prime}_{2}$ is defined as $\pi^{\prime}_{2}\equiv\\{j\,^{\prime}\in\pi_{2}:\exists B_{n},D(x_{j\,^{\prime}},B_{n})\leq R_{0}a,\|G(B_{n})\|\leq n_{1}\\}$ (2.103) Hence, it only remains to prove that $\displaystyle 0\leq$ $\displaystyle-\frac{2+t}{1+t}\nabla_{1}(1-\sum_{i=2}^{N}F_{i,1})\nabla_{1}+\mbox{$\frac{1}{2}$}\sum_{j\in\pi^{\prime}_{2}}(1-F_{j,1})v^{a}_{2,1}(x_{1}-x_{j})$ (2.104) Second, we claim the following inequality which will be proved later. $n_{1}\left(1-\sum_{i=2}^{N}F_{i,1}\right)\geq\sum_{j\in\pi^{\prime}_{2}}(1-F_{j,1})\theta_{(R_{0}a)}(x_{1}-x_{j})$ (2.105) which implies that $-\nabla_{1}\left(1-\sum_{i=2}^{N}F_{i,1}\right)\nabla_{1}\geq-\frac{1}{n_{1}}\nabla_{1}\sum_{j\in\pi^{\prime}_{2}}(1-F_{j,1})\theta_{(R_{0}a)}(x_{1}-x_{j})\nabla_{1}$ (2.106) With (2.106), we obtain that the right side of (2.104) is not less than $\displaystyle\sum_{j\in\pi^{\prime}_{2}}\bigg{(}-\frac{2+t}{n_{1}(1+t)}\nabla_{1}(1-F_{j,1})\theta_{(R_{0}a)}(x_{1}-x_{j})\nabla_{1}+\frac{1}{2}(1-F_{j,1})v^{a}_{2,1}(x_{j}-x_{1})\bigg{)}$ $\displaystyle\geq$ $\displaystyle\sum_{j\in\pi^{\prime}_{2}}\bigg{(}1-\prod_{k\neq 1\,or\,j}\bar{\theta}_{2R}(x_{k}-x_{j})\bigg{)}\times\frac{2}{n_{1}(1+t)}$ (2.107) $\displaystyle\times\bigg{(}-\nabla_{1}\theta_{(R_{0}a)}(x_{1}-x_{j})\nabla_{1}+2n_{1}(v^{a}+tv^{a}_{-})(x_{j}-x_{1})\bigg{)}$ Here we used the definition of $F_{j,1}$ and (2.80), i.e., $v^{a}_{2,1}=\frac{4}{1+t}[v^{a}+tv^{a}_{-}]$. With the assumption $SL[4\,n_{1}(v+v_{-})]\geq 0$, we obtain that (2.107)$\geq 0$, which implies inequality (2.104). Hence, it only remains to prove (2.105). For $x_{2},\cdots,x_{N}$ fixed, we define $\pi_{3}$ as following, $\pi_{3}=\left\\{2\leq j\leq N:\prod_{2\leq k\leq N,k\neq j}\bar{\theta}_{2R}(x_{j}-x_{k})=1\right\\}$ (2.108) With the definition of $\pi_{3}$, we obtain that $F_{j,1}=\left\\{\begin{array}[]{ll}\theta_{R}(x_{j}-x_{1})&j\in\pi_{3}\\\ 0&j\notin\pi_{3}\,,\end{array}\right.$ (2.109) Hence, it only remains to prove that $n_{1}\left(1-\sum_{i\in\pi_{3}}\theta_{R}(x_{1}-x_{i})\right)\geq\sum_{j\in\pi^{\prime}_{2},j\notin\pi_{3}}\theta_{(R_{0}a)}(x_{1}-x_{j})$ (2.110) or $\max_{x\in{\mathbb{R}}^{3}}\left(n_{1}\sum_{i\in\pi_{3}}\theta_{R}(x-x_{i})+\sum_{j\in\pi^{\prime}_{2},j\notin\pi_{3}}\theta_{(R_{0}a)}(x-x_{j})\right)\leq n_{1}$ (2.111) Because the distance between $x_{i}$ ($i\in\pi_{3}$) and $x_{j}$ ($2\leq j\leq N,j\neq i$) are not less than $2R$, we have that if $i\in\pi_{3}$ $\theta_{R}(x-x_{i})=1\Rightarrow\sum_{j\neq 1,j\neq i}\theta_{R}(x-x_{j})=0$ (2.112) So, it only remains to prove that $\max_{x}\left(\sum_{j\in\pi^{\prime}_{2},j\notin\pi_{3}}\theta_{(R_{0}a)}(x-x_{j})\right)\leq n_{1}$ (2.113) By the definition of $\pi^{\prime}_{2}$ in (2.103), if $j\in\pi^{\prime}_{2}$ and $\theta_{(R_{0}a)}(x-x_{j})=1$, there exist $B_{n}$ satisfying $\|G(B_{n})\|\leq n_{1}$ and $d(x,B_{n})\leq 2R_{0}a$. Hence by the definition of $G(B_{n})$ in (2.84) and (2.83), we obtain that, for $\forall x\in{\mathbb{R}}^{3}$ $\sum_{i\in\pi^{\prime}_{2}}\theta_{(R_{0}a)}(x-x_{i})=1\Rightarrow\sum_{i\in\pi_{2}}\theta_{(R_{0}a)}(x-x_{i})\leq n_{1}$ (2.114) With the fact that $\pi^{\prime}_{2}\subset\pi_{2}$, we arrive at the desired result (2.113) and complete the proof of Theorem 2. ∎ ### 2.3 Proof of Lemma 1 ###### Proof. Let $\delta\Omega$ be any infinitismal solid angle. With the definition of scattering length, we have that if $\phi$ is a complex-valued function such that $\phi(x)=1\,\,\,{\rm for}\,\,\,x\in\delta\Omega\otimes{\mathbb{R}}\,\,\,{\rm and}\,\,\,\|x\|=R^{\prime}\geq R_{0}a$ (2.115) then $\delta\Omega\cdot a\leq\int_{\delta\Omega\otimes[0,R\,^{\prime}]}\|\nabla\phi(x)\|^{2}+\mbox{$\frac{1}{2}$}v^{a}(x)\|\phi(x)\|^{2}dx$ (2.116) Hence we obtain that $a\int_{{\mathbb{R}}^{3}}\delta(\|x\|-R^{\prime})\|\phi(x)\|^{2}dx\leq\int_{\|x\|\leq R\,^{\prime}}\|\nabla\phi(x)\|^{2}+\mbox{$\frac{1}{2}$}v^{a}(x)\|\phi(x)\|^{2}dx$ (2.117) which says that for any non-negative radial function $U_{0}(x)$, supported in the annulus $R_{0}a\leq\|x\|\leq R$, with $\int_{{\mathbb{R}}^{3}}U_{0}(x)\,dx=4\pi$, we have $\displaystyle-\nabla\theta_{R}(x)\nabla+\mbox{$\frac{1}{2}$}v^{a}\geq aU_{0}$ (2.118) Note: The result of lemma 2.5 of [5] shows the $\theta_{R}(x)$ in above inequality can be replaced with the characteristic function of any star-shaped set when $v^{a}\geq 0$. Furthermore, one can easily prove that for fixed $v$ ($SL[v]=1$), $v_{+}$ and small enough $\varepsilon$, $SL[v-\varepsilon v_{+}]>0,\,\,\,\,\,\,\|f_{\varepsilon}\|_{\infty}\leq{\rm const.\,}$ (2.119) Here we denote $f_{\varepsilon}$ as the normalized solution ($\lim_{\|x\|\to\infty}f_{\varepsilon}(x)=1$) of the zero-energy scattering equation of $v-\varepsilon v_{+}$. Hence, by the definition of scattering length, using $f_{\varepsilon}$ as the trial function for $v$, we obtain that $SL[v]\leq SL[v-\varepsilon v_{+}]+\varepsilon\\|v_{+}\\|_{1}\cdot\|f_{\varepsilon}\|_{\infty}\leq SL[v-\varepsilon v_{+}]+{\rm const.\,}\varepsilon$ (2.120) Combining this result with (2.118) and the definition of $v^{a}_{\varepsilon}$, we have, $-(1-\varepsilon)\nabla\theta_{R}(x)\nabla+\mbox{$\frac{1}{2}$}v^{a}_{\varepsilon}\geq(1-{\rm const.\,}\varepsilon)aU_{0}$ (2.121) and $-(1-\varepsilon)\nabla_{j}F_{i,j}\nabla_{j}+\mbox{$\frac{1}{2}$}F_{i,j}v^{a}_{\varepsilon}(x_{i}-x_{j})\geq(1-{\rm const.\,}\varepsilon)aF_{i,j}U_{0}(x_{i}-x_{j})$ (2.122) Hence, we obtain the following lower bound on $H^{\prime\prime}$, which is defined in (2.79) $\displaystyle H^{\prime\prime}\geq\varepsilon\sum_{j}-\Delta_{j}+\varepsilon\sum_{i\neq j}\frac{v^{a}_{+}}{2}(x_{i}-x_{j})+(1-{\rm const.\,}\varepsilon)a\sum_{i\neq j}W_{i,j}$ (2.123) Here $W_{i,j}$ is defined as $W_{i,j}=F_{ij}U_{0}(x_{i}-x_{j})\geq 0$ (2.124) As in [4], we choose $\ell=aY^{-6}$ (2.125) and divide $\Lambda$ into small cubes with side length $\ell$. Then we have $H^{\prime\prime}\geq H^{(3)}\equiv\varepsilon\sum_{j}-\Delta_{j}+\varepsilon\sum_{i\neq j}\frac{v^{a}_{+}}{2}(x_{i}-x_{j})+(1-{\rm const.\,}\varepsilon)a\sum_{i\neq j}W^{\prime}_{i,j}$ (2.126) Here $W^{\prime}_{i,j}$ is defined as $W^{\prime}_{i,j}=G_{ij}U_{0}(x_{i}-x_{j}),\,\,\,\,G_{ij}\equiv F_{ij}\chi(x_{i})\chi(x_{j})\geq 0,$ (2.127) and $\chi(x)$ is equal to 1 when the distance between $x$ and the edges of the small cubes is greater than $2R$; otherwise it is equal to 0. As we can see the particles in different cubes don’t affect each other in $H^{(3)}$. We are going to estimate the ground energy $E^{(3)}(n,\ell)$ of $H^{(3)}$ for $n$ particles in $[0,\ell]^{3}$ with Neumann boundary condition. First, in the case that $n\leq\frac{8}{3}\varrho\ell^{3}Y^{-1}$, with the definition of $\varepsilon$ in (2.76), we have that $\displaystyle H^{(3)}$ $\displaystyle\geq 3Y\sum_{j}-\Delta_{j}+(1-{\rm const.\,}Y)a\sum_{i\neq j}W^{\prime}_{i,j}$ (2.128) Then with the Temple inequality in [10], as in [5] (Ineq. 2.60, 2.66), we have that $\displaystyle\frac{E^{\left(3\right)}\left(n,\ell\right)}{n}\geq$ $\displaystyle 4\pi\frac{an}{\ell^{3}}\left(1-\frac{1}{n}\right)\left(1-{\rm const.\,}Y\right)\left(1-\frac{2R}{\ell}\right)^{3}\left(1+\frac{4\pi n}{3}\left(\frac{2R}{\ell}\right)^{3}\right)^{-1}$ (2.129) $\displaystyle\left(1-\frac{3}{\pi}\frac{an}{\left(R^{3}-\left(aR_{0}\right)^{3}\right)\left(3\pi Y\ell^{-2}-4a\ell^{-3}n^{2}\right)}\right)$ $\displaystyle\geq$ $\displaystyle 4\pi\frac{an}{\ell^{3}}(1-\frac{1}{n})(1-{\rm const.\,}Y)(1-\frac{n}{6\ell^{3}\varrho}Y)$ Second, when $n\geq\frac{8}{3}\varrho\ell^{3}Y^{-1}$, using the fact $W^{\prime}\geq 0$, we obtain that $H^{(3)}\geq\varepsilon H^{(4)}\equiv\varepsilon\bigg{(}\sum_{j}-\Delta_{j}+\sum_{i\neq j}\frac{v^{a}_{+}}{2}(x_{i}-x_{j})\bigg{)}$ (2.130) Using superadditivity of the ground state energy of $H^{(4)}$, we obtain that the ground energy $E^{(4)}(n,\ell)$ of $H^{(4)}$ is bounded from below as follows, ($n\geq p$) $\displaystyle E^{(4)}(n,\ell)\geq\left[\frac{n}{p}\right]E^{(4)}(p,\ell)\geq\frac{n}{2p}E^{(4)}(p,\ell)$ (2.131) Here $[n/p\,]$ is the largest integer not greater than $n/p$. Actually, $H^{(4)}$ is just the Hamiltonian for the pure non-negative interaction potential, as in [7]. Denote $a_{+}$ as follows: $a_{+}=\min\\{SL(v^{a}),SL(v^{a}_{+})\\}\leq a$ (2.132) Replacing $v_{+}^{a}$ with soft potential, we obtain that, $H^{(4)}\geq 3Y\sum_{j}-\Delta_{j}+(1-{\rm const.\,}Y)a_{+}\sum_{i\neq j}W^{\prime}_{i,j}$ (2.133) As (2.129), we can prove that when $p=\frac{8}{3}\varrho\ell^{3}Y^{-1}$, $E^{(4)}(p,\ell)/p\geq\frac{32}{3}\pi a_{+}\varrho Y^{-1}\left(1-\frac{1}{2}\right)(1-{\rm const.\,}Y)\geq\frac{16}{3}\pi a_{+}\varrho Y^{-1}$ (2.134) Hence when $n\geq\frac{8}{3}\varrho\ell^{3}Y^{-1}$, we have the following lower bound on the ground energy $E^{(3)}(n,\ell)$ of $H^{(3)}$ $E^{(3)}(n,\ell)/n\geq\varepsilon E^{(4)}(n,\ell)/n\geq\frac{8}{3}\pi\varepsilon a_{+}\varrho Y^{-1}\geq 8\pi a\varrho$ (2.135) For the last inequality, we used the definition of $\varepsilon$ in (2.76). So far we proved that $\displaystyle\frac{E^{(3)}(n,\ell)}{n}\geq\left\\{\begin{array}[]{ll}4\pi\frac{an}{\ell^{3}}(1-\frac{1}{n})(1-{\rm const.\,}Y)(1-\frac{n}{6\ell^{3}\varrho}Y)&1\leq n\leq\frac{8}{3}\ell^{3}\varrho Y^{-1}\\\ 8\pi a\varrho&n\geq\frac{8}{3}\ell^{3}\varrho Y^{-1}\,,\end{array}\right.$ (2.138) which implies that when $Y$ is small enough, $\displaystyle\frac{E^{(3)}(n,\ell)}{n(1-{\rm const.\,}Y)}\geq\left\\{\begin{array}[]{ll}4\pi\frac{an}{\ell^{3}}(1-\frac{1}{n})&1\leq n\leq 4\ell^{3}\varrho\\\ 8\pi a\varrho&n\geq 4\ell^{3}\varrho\,,\end{array}\right.$ (2.141) Recall the following two facts, 1. 1. the interaction potential $W^{\prime}_{i,j}$ only depends on the particles in the same cubes as $i$ and $j$, 2. 2. the particles in different cubes have no interaction. Using the inequality above (2.141), with the method in [7] (Ineq. 2.55, 2.56), one can prove that the ground state $E^{(3)}(N,\Lambda)$ of $H^{(3)}$ of $N$ particles in big cubic $\Lambda$ is greater than $E^{(3)}(N,\Lambda)/N\geq 4\pi a\varrho(1-{\rm const.\,}Y).$ (2.142) Here $Y$ is defined in (2.75), which implies the desired result (2.79). ∎ ## 3 Appendix In this appendix, we show that if $v^{a}$ is a continuous function and $H_{N}$ has no bound state for any $N$, $v^{a}$ has a positive core and bounded from below, i.e., $v^{a}(0)>0,\,\,\,\,\min v^{a}(r)\neq-\infty$ (3.1) And these inequalities also hold when $v^{a}$ is stable [1] in the sense of (3.2). One can see that $\min v^{a}(r)\neq-\infty$ is trivial when $v^{a}$ is continuous. So it only remains to prove that $v^{a}(0)>0$. First, we prove the statement in the case when $v^{a}$ is stable, which is defined as follows: there exists constant $C$, for any $N$, $x_{1},\cdots,x_{N}$, $\sum_{1\leq i\neq j\leq N}v^{a}(x_{i}-x_{j})\geq-CN,$ (3.2) Inserting $x_{1}=x_{2}=\cdots=x_{[N/2]}=0,\,\,\,\,x_{[N/2]+1}=x_{[N/2]+2}=\cdots=x_{N}=x_{0}$ (3.3) into the left side of (3.2), for some $x_{0}\in{\mathbb{R}}^{3}$ satisfying $v^{a}(x_{0})<0$, we obtain that ${\rm const.\,}v^{a}(0)N^{2}-{\rm const.\,}v^{a}(x_{0})N^{2}\geq-CN,$ (3.4) which implies the desired result that $v^{a}(0)>0$. Next, we prove the statement in the case that $H_{N}$ has no bounded state for any $N$. Because $v^{a}$ is not pure non-negative, there exist $x_{0}\in{\mathbb{R}}^{3},r_{1},C\in R$ satisfying that $v^{a}(x)<-C,\,\,\,{\rm for}\,\,\,x\in B(x_{0},r_{1})\subset{\mathbb{R}}^{3}$ (3.5) Here $B(x_{0},r_{1})$ is the sphere of radius $r_{0}$ centered at $x_{0}$. If $v^{a}(0)\leq 0$, there exists $r_{2}<r_{1}/2$ satisfying that $v^{a}(x)<C/2,\,\,\,{\rm for}\,\,\,x\in B(0,r_{2})$ (3.6) We construct the trial state such that $x_{1},x_{2},\cdots,x_{[N/2]}$ are localized in $B(0,r_{2})$ with the Dirichelet boundary condition and $x_{[N/2]+1},x_{[N/2]+2},\cdots,x_{N}$ are localized in $B(x_{0},r_{2})$ with the same boundary condition. The energy of this state is less than $-\frac{C}{8}N^{2}+\frac{{\rm const.\,}}{r_{2}^{2}}N$ (3.7) Here the first term is potential energy and the second term is kinetic energy. When $N$ goes to infinity, the energy of this trial state is negative and hence there are bound states, which is a contradiction with our assumptions. So we arrive at the desired result that $v^{a}(0)>0$. ## References * [1] D. Ruelle, Statistical Mechanics: Rigorous Results (1989). * [2] E.H. Lieb and R. Seiringer, Derivation of the Gross-Pitaevskii Equation for Rotating Bose Gases, Comm. Math. Phys. Volume 264, Issue 2, pp.505-537 (2005). * [3] E.H. Lieb, R. Seiringer, J.P. Solovej, Ground State Energy of the Low Density Fermi Gas, in: Recent Advances in Differential Equations and Mathematical Physics, N. Chernov, Y. Karpeshina, I. Knowles, R. Lewis, R. Weikard, eds., Amer. Math. Soc. Contemporary Math. series 412, 239-248 (2006) * [4] E.H. Lieb, R. Seiringer, J. P. Solovej, J. Yngvason The Ground State of the Bose Gas Current Developments in Mathematics, (2001), International Press, Cambridge, (2002), pp. 131-178. * [5] E.H. Lieb, R. Seiringer, J. P. Solovej, J. Yngvason, The Mathematics of the Bose Gas and its Condensation, Oberwolfach Seminars Vol. 34, (2005). * [6] E.H. Lieb, R. Seiringer, J. Yngvason, Bosons in a Trap: A Rigorous Derivation of the Gross-Pitaevskii Energy Functional, Phys. Rev. A 61, 043602 (2000). * [7] E.H. Lieb, J. Yngvason, Ground State Energy of the Low Density Bose Gas, Phys. Rev. Lett. 80, 2504–2507 (1998). * [8] E.H. Lieb, J. Yngvason, The Ground State Energy of a Dilute Two-dimensional Bose Gas, J. Stat. Phys. 103, 509 (2001). arXiv math-ph/0002014, mp arc 00- 63\. * [9] F.J. Dyson, Ground State Energy of a Hard-Sphere Gas, Phys. Rev. 106, 20–26 (1957). * [10] G. Temple, The theory of Rayleigh’s Principle as Applied to Continuous Systems, Proc. Roy. Soc. London A 119, 276–293 (1928). * [11] L.J. Oon, Ground State Energy of Dilute Bose Gas in Small Negative Potential Case Arxiv: 0803.0533.	arxiv-papers	2008-08-29T10:14:38	2024-09-04T02:48:57.541703	{ "license": "Public Domain", "authors": "Jun Yin", "submitter": "Jun Yin", "url": "https://arxiv.org/abs/0808.4066" }
0808.4120	# Magnetocapacitance of a graphene monolayer M. Tahir1∗, K. Sabeeh2†, A. MacKinnon1 1Department of Physics, The Blackett Laboratory, Imperial College London, SW7 2AZ London, United Kingdom 2Department of Physics, Quaid-i-Azam University, Islamabad 45320, Pakistan (date; date; date; date) ###### Abstract We present a theoretical study of magnetocapacitance in a graphene monolayer at finite temperature taking into account the effects of disorder. The density of states (DOS) and magnetocapacitance found for graphene are compared to those found in standard two dimensional electron gas (2DEG) systems. The magnetic oscillations in DOS and magnetocapacitance are found to be enhanced and much more robust with respect to temperature damping in monolayer graphene in comparison with a 2DEG. Furthermore, we find that there is a $\pi$ phase shift between magnetic 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: 72.80.Rj, 72.15.Gd ††preprint: LABEL:FirstPage01 LABEL:LastPage#102 ## I Introduction Monolayer graphene is a gapless semiconductor with conical touching of electron and hole bands. Since the quasiparticles in a graphene monolayer obey the massless Dirac equation they act as massless Dirac Fermions. 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 effect and a $\pi$ Berry phase1 ; 2 ; 3 . Besides the fundamental interest in understanding the electronic properties of graphene there are also serious suggestions that it can serve as the building block for nanoelectronic devices 4 . Among the most important tools in studying the electronic properties of solid state systems are capacitance measurements. Capacitance measurements can be effectively used to probe the thermodynamic density of states of an electron system. It has been successfully used to probe the density of states in a conventional 2DEG systems and references therein5 ; 6 ; 7 . In the present work, we determine the magnetocapacitance of a graphene monolayer. It is possible to carry out experiments, along the lines of those carried out on conventional 2DEG systems5 ; 6 ; 7 realized in semiconductor heterostructures, carbon nanotubes and graphene nanoribbons8 , that allow verification of the model presented here. If confirmed experimentally, our theoretical predictions provide another method of determining the density of states of a graphene monolayer in a magnetic field. In section II, we present the formulation of the problem. Section III contains the calculation of the temperature dependent density of states and magnetocapacitance of a graphene monolayer, including the comparison with 2DEG and discussion of numerical results. The conclusions are in section IV. ## II Formulation We consider two-dimensional Dirac electrons in graphene moving in the $x$-$y$-plane. The magnetic field ($B$) is applied along the $z$-direction perpendicular to the graphene plane. We employ the Landau gauge and write the vector potential as $A=(0,Bx,0)$. The two-dimensional Dirac-like Hamiltonian1 ; 2 ; 3 for an electron in the Landau gauge is $H_{0}=v_{D}\overleftrightarrow{\sigma}.(-i\hslash\overrightarrow{\nabla}+e\overrightarrow{A})$. Here $\sigma=\\{\sigma_{x},\sigma_{y}\\}$ are the Pauli matrices and $v_{D}$ characterizes the velocity of Dirac electron. The Landau level energy eigenvalues are given by $\varepsilon_{n}=\hslash\omega_{g}\sqrt{n},$ where $\omega_{g}=v_{D}\sqrt{\frac{2eB}{\hslash}}$ is the cyclotron frequency of Dirac electrons and $n$ is an integer. The Landau level spectrum for Dirac electrons is significantly different from the spectrum for electrons in a conventional 2DEG which is given as $\varepsilon(n)=\hbar\omega_{c}(n+1/2)$, where $\omega_{c}$ is the cyclotron frequency. This difference between the energy dispersion of Dirac electrons and conventional electrons is also reflected in the Density of States (DOS). The oscillations in the capacitance, in the presence of a magnetic field, are directly related to the density of states at the Fermi energy. In the presence of an external magnetic field, without taking into account any disorder effects, the DOS is represented as a series of delta functions. In all practical situations, there is always some degree of disorder present in the system. The DOS delta functions broaden due to the presence of scattering centers and the DOS can be represented as a Gaussian, or a Lorentazian function to take this broadening into account7 . Here, we first consider the Gaussian broadening of DOS denoted by $D_{G}(\varepsilon)$ given as $D_{G}(\varepsilon)=\frac{2}{2\pi l^{2}}\underset{n=-\infty}{\overset{\infty}{{\displaystyle\sum}}}\frac{1}{\Gamma\sqrt{2\pi}}\exp\left[-\frac{(\varepsilon-\varepsilon_{n})^{2}}{2\Gamma^{2}}\right]$, where $\Gamma$ is the Gaussian distribution broadening width of zero shift. If we use the Poisson summation formula9 : $\frac{1}{2}F(0)+\underset{k=1}{\overset{\infty}{{\displaystyle\sum}}}F(k)={\displaystyle\int\limits_{0}^{\infty}}F(n)dn+2\overset{\infty}{{\displaystyle\sum_{k=1}}}(-1)^{k}{\displaystyle\int\limits_{0}^{\infty}}F(n)\cos[2\pi kn]dn$, the summation over $n$ in the DOS expression can be carried out by changing the variable as $x=\frac{\varepsilon-\hslash\omega_{g}\sqrt{n}}{\sqrt{2}\Gamma}$ and applying the limit $\varepsilon\sim\varepsilon_{F}\gg\Gamma,$ where $\varepsilon_{F}=v_{D}\hslash k_{F}$ denotes the Fermi energy with Fermi wavenumber $k_{F}=\sqrt{2\pi n_{s}}$ and $n_{s}$ is the number density of the system. The first integral yields $D_{0}=\frac{\varepsilon}{\pi\hslash^{2}v_{D}^{2}}$ which is the DOS in the absence of magnetic field and the second integral becomes proportional to ${\displaystyle\int\limits_{-\infty}^{\infty}}dx\exp[-x^{2}]\cos\left[\frac{4\pi k\varepsilon\Gamma}{\hslash^{2}\omega_{g}^{2}}x\right]=\sqrt{\pi}\exp\left[-4\left(\frac{\pi k\varepsilon\Gamma}{\hslash^{2}\omega_{g}^{2}}\right)^{2}\right]$. The final result for the Gaussian broadened DOS can be written as $D_{G}(\varepsilon)=D_{0}\left\\{1-2\overset{\infty}{{\displaystyle\sum_{k=1}}}(-1)^{k}\exp\left[-4\left(\frac{\pi k\varepsilon\Gamma}{\hslash^{2}\omega_{g}^{2}}\right)^{2}\right]\cos\left[\frac{2\pi k\varepsilon^{2}}{\hslash^{2}\omega_{g}^{2}}\right]\right\\}$. Furthermore, following the same approach we can also determine Lorentazian broadened DOS $D_{L}(\varepsilon)=\frac{2}{2\pi l^{2}}\underset{n=-\infty}{\overset{\infty}{{\displaystyle\sum}}}\frac{1}{\pi}\frac{\Gamma}{(\varepsilon-\varepsilon_{n})^{2}+\Gamma^{2}}$ with the result $D_{L}(\varepsilon)=D_{0}\left\\{1-2\overset{\infty}{{\displaystyle\sum_{k=1}}}(-1)^{k}\exp\left[-\frac{2\pi k\varepsilon\Gamma}{\hslash^{2}\omega_{g}^{2}}\right]\cos\left[\frac{2\pi k\varepsilon^{2}}{\hslash^{2}\omega_{g}^{2}}\right]\right\\}.$ (1) We see that both forms of broadening yield the same numerical results10 in the regime $\varepsilon\sim\varepsilon_{F}\gg\Gamma$. For $\Gamma\gg\hbar\omega_{g}$, it is sufficient to retain only the first order term ($k=1$ term) since the contribution from the higher order terms is highly damped. Furthermore, the above result for the DOS is the same as obtained in Ref. 11 in the limit when energy gap $\Delta=0$ 11 . ## III Magnetocapacitance of a graphene monolayer We consider a top-gated graphene device in which the capacitor is formed between the top gate and the graphene sheet. In order to deposit charge on the capacitor formed by the graphene sheet and the gate, the voltage source has to do both electrostatic and chemical work on the system. The effective capacitance is the geometric capacitance in series with ”chemical capacitance”. In an external magnetic field, the effective capacitance $C(B)$ can be determined from the following relation5 ; 6 $\frac{1}{C(B)}-\frac{1}{C_{0}}=\frac{1}{e^{2}D_{0}}\left[\frac{D_{0}}{D_{T}(B)}-1\right],$ (2) where $C_{0}=1168\ast 10^{-6}$ F m-2 is the capacitance at zero magnetic field and $D_{T}(B)$ is the temperature dependent DOS at finite magnetic field that is determined from $D_{T}(B)=\frac{\partial n_{s}}{\partial\varepsilon_{F}}={\displaystyle\int\limits_{0}^{\infty}}d\varepsilon D_{L}(\varepsilon)\frac{\partial f(\varepsilon-\varepsilon_{F})}{\partial\varepsilon_{F}}$, where $n_{s}$ is the number density and $f(\frac{\varepsilon-\varepsilon_{F}}{k_{B}T})$ is the Fermi Dirac distribution function. In the above expression for $D_{T}(B),$ we introduce a change of variable $\beta(\varepsilon-\varepsilon_{F})=s$ and apply the low temperature limit such that $\beta\ll\varepsilon_{F}$ where $\beta=\frac{1}{k_{B}T}.$ As a result, it can be expressed as $D_{T}(B)=D_{0}\left\\{1+\exp\left[-\frac{2\pi\varepsilon_{F}\Gamma}{\hslash^{2}\omega_{g}^{2}}\right]\cos\left[\frac{2\pi\varepsilon_{F}^{2}}{\hslash^{2}\omega_{g}^{2}}\right]{\displaystyle\int\limits_{0}^{\infty}}ds\frac{\cos\left[\frac{4\pi\varepsilon_{F}}{\hslash^{2}\omega_{g}^{2}\beta}s\right]}{\cosh^{2}[s/2]}\right\\}.$ (3) The integration can be performed by using the following identity12 : ${\displaystyle\int\limits_{0}^{\infty}}dx\frac{\cos ax}{\cosh^{2}\beta x}=\frac{a\pi}{2\beta^{2}\sinh(a\pi/2\beta)}$, with the result that the temperature dependent DOS is $D_{T}(B)=D_{0}\left\\{1+2\frac{\alpha}{\sinh\alpha}\exp\left[-\frac{2\pi\varepsilon_{F}\Gamma}{\hslash^{2}\omega_{g}^{2}}\right]\cos\left[\frac{2\pi\varepsilon_{F}^{2}}{\hslash^{2}\omega_{g}^{2}}\right]\right\\},$ (4) where $\alpha=\frac{4\pi^{2}\varepsilon_{F}}{\hslash^{2}\omega_{g}^{2}\beta}$ is a dimensionless parameter which expresses the characteristic temperature for damping of oscillations in $D_{T}(B)$. Following the same approach as given above for graphene monolayer, we can obtain the temperature dependent DOS for conventional 2DEG5 ; 6 ; 7 ; 9 with the result $D_{T}(B)=D_{s}\left\\{1+2\frac{\alpha_{s}}{\sinh\alpha_{s}}\exp\left[-\frac{2\pi\Gamma}{\hslash\omega_{c}}\right]\cos\left[\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}-\pi\right]\right\\},$ (5) where $D_{s}=\frac{m^{\ast}}{2\pi\hslash^{2}}$ and $\alpha_{s}=\frac{2\pi^{2}}{\hslash\omega_{c}\beta}$. Now, we are in a position to compare the results for the temperature dependent DOS for monolayer, given by Eq.(4), with the result for 2DEG, shown in Eq. (5). The following differences in temperature dependent DOS in the two systems are found: i)-The argument of the oscillatory cosine term, $\cos[\frac{2\pi\varepsilon_{F}}{\hslash\omega_{c}}-\pi]$ responsible for the $\pi$-Berry phase shift, in 2DEG depends on the ratio of the Fermi energy to the cyclotron frequency while for monolayer graphene the argument of the cosine depends on the square of the ratio of the Fermi energy to the corresponding cyclotron frequency without an extra $\pi$. ii)-The exponential term, $\exp[-\frac{2\pi\Gamma}{\hslash\omega_{c}}],$ responsible for the exponential decay of the amplitude of the oscillatory DOS for 2DEG depends on the ratio of $\Gamma$ to the cyclotron frequency $\omega_{c}.$ From Eq.(4), in monolayer graphene the exponential term depends on the ratio of the product of $\Gamma$ and $\varepsilon_{F}$ to the square of the corresponding cyclotron frequency $\omega_{g}$. The magnitude of the argument of the exponential term is 1.84 in GaAs 2DEG while 0.22 in Graphene at fixed value of the magnetic field (1 Tesla) and disorder (0.5 meV). Therefore, we expect that for experimentally relevant parameters as considered here, disorder will cause less damping of DOS oscillations in a graphene monolayer compared to a standard 2DEG. iii)- the temperature dependence of the oscillation amplitude of DOS is given by the factor $\underset{\alpha->\infty}{\lim}\frac{\alpha}{\sinh(\alpha)}=2\alpha e^{-\alpha}$. The amplitude of the oscillatory DOS decays exponentially as the temperature is raised. From this temperature dependent coefficient the characteristic temperature for damping of oscillations in DOS, $T_{D},$ can be identified such that $k_{B}T_{D}=\frac{v_{D}eB}{2\pi^{2}k_{F}},$ with $k_{F}$ being the Fermi wavenumber, equal to $\sqrt{2\pi n_{s}}$ , while in a standard 2DEG5 ; 6 ; 7 ; 9 the corresponding expression is $k_{B}T_{s}=\frac{\hslash\omega_{c}}{2\pi^{2}}=\frac{\hslash eB}{2\pi^{2}m^{\ast}}$. Here $m^{\ast}=0.068$ $m_{e}$ in GaAs and $m_{e}$ is the usual electron mass. These characteristic temperatures determine the robustness with respect to temperature of magnetic oscillations in DOS. The relative magnitudes of the characteristic temperatures for damping in 2DEG and graphene monolayer is $\frac{T_{s}}{T_{D}}=\frac{v_{p}}{v_{D}}$, the ratio of the characteristic temperatures is equal to the ratio of the corresponding Fermi velocities in the two systems, where $v_{p}=\frac{\hbar k_{F}}{m^{\ast}}$ is the Fermi velocity in a 2DEG. Typically, the Fermi velocity of 2DEG systems ($\sim 10^{5}$m/s) is an order of magnitude smaller than that of Graphene ($\sim 10^{6}$m/s). An estimate of this ratio of temperatures is given by $\frac{T_{s}}{T_{D}}=\frac{\hbar k_{F}}{v_{D}m^{\ast}}=\frac{\hbar\sqrt{2\pi n_{s}}}{v_{D}m^{\ast}}=0.24$ for $GaAs$, For $Ge$ is $\simeq 0.03$ and $\simeq 0.015$ for $Si$ with a fixed value of the number density ( $n_{s}=3.2\times 10^{15}$ m-2). For comparison for an equal temperature scale the effective mass should be $\simeq 0.016$ $m_{e}$, which is smaller than that for most 2DEG systems such as Si, Ge, GaAs. From the discussion of the DOS expressions in the two systems given above, we expect that the oscillations in the magnetocapacitance will be affected less by temperature in a monolayer graphene compared to a 2DEG (GaAs is considered for comparison in the numerical results) due to the higher damping temperature in the monolayer. On varying the magnetic field, changes in the DOS are reflected by oscillations of the magnetocapacitance. Since magnetocapacitance is directly related to the DOS, the exponential decay in the amplitude of the magnetocapacitance oscillations with temperature and disorder is expected. Before we present results for the magnetocapacitance, the dimensionless DOS ($D_{T}(B)/D_{0}$) for the two systems is shown in Fig.(1). The DOS is plotted as function of the inverse magnetic field. The dimensionless magnetic field $B_{0}/B$ is introduced through the characteristic magnetic field $B_{0}=\hbar k_{F}^{2}$. $B_{0}=13.25$ T for $n_{s}=3.2\times 10^{15}$ m-2. The temperature is $1.2$ K and the disorder energy $\Gamma=0.5$ meV in Fig.(1). The magnetic oscillations, with $\pi$-Berry phase difference, are clearly visible in the DOS of the two systems . Furthermore, the oscillation in the DOS in the graphene monolayer are found to be enhanced and less damped with temperature compared to those in a 2DEG. In order to study the effects of temperature on the magnetocapacitance in the monolayer and 2DEG, we show in Figs. (2, 3) the temperature dependent magnetocapacitance as a function of the magnetic field at two different temperatures: $1.5$ K and $22$ K. In these figures, $\Gamma$ is held constant at $0.5$ meV. We observe, as expected from the above discussion, that oscillations in magnetocapacitance are less damped by temperature and there is a $\pi$-Berry phase difference in graphene monolayer compared with the same oscillations found in a 2DEG. For the parameters5 ; 6 ; 7 considered here and with constant disorder, the oscillations found in the magnetocapacitance in a 2DEG become completely damped at $22$ K whereas they are found to persist upto a temperature of $60$ K in a graphene monolayer. Therefore, we have found that magnetocapacitance oscillations are more robust with respect to temperature damping in a graphene monolayer compared to a 2DEG. Furthermore, these magnetic oscillations have a $\pi$-Berry phase difference in the two systems. ## IV Conclusions We have investigated the finite temperature density of states and magnetocapacitance in a graphene monolayer taking into account finite temperature as well as disorder. The results obtained are compared to those found for conventional 2DEG systems. We have found that magnetic oscillations in the DOS and magnetocapacitance in a graphene monolayer are enhanced, are more robust against temperature and have a $\pi$-Berry phase difference compared to a 2DEG. ∗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, Nature 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). * (3) 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).; J. B. Oostinga, H. B. Heersche, X. Liu, A. F. Morpurgo and L. M. K. Vandersypen, Nat. Mat. 7, 151 (2008).; T. Ohta, A. Bostwick, T. Seyller, K. Horn and E. Rotenberg, Science 313, 951 (2006). * (4) C. Berger,et.al, Science 312, 1191 (2006).; 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). * (5) T. P. Smith, B. B. Goldberg, P. J. Stiles, and M. Heiblum, Phys. Rev. B 32, 2696 (1985).; V. Mosser, D. Weiss, K. von Klitzing, K. Ploog, and G. Weimann, Solid State Commun. 58, 5 (1986).; H.-z. Zheng, A. Song, F.-h. Yang, Y.-z. Li, Phys. Rev. B 49, 1802 (1994).; T. Jungwirth, L. Smrcka, Phys. Rev. B 51, 10181 (1995).; D. Weiss,C. Zhang, R. R. Gerhardts, K. von Klitzing, G. Weimann, Phys. Rev. B 39, 13020 (1989).; T. Ando, Y. Uemura, J. Phys. Soc. Jpn 36, 959 (1974).; D. Weiss and K. von Klitzing in High magnetic fields in semiconductor physics, edited by G. Landwehr, Springer series in Solid State Sciences, Vol. 71 (Springer-Verlag, Berlin, 1987). * (6) V. S. Khrapai, A. A. Shashkin, M. G. Trokina, V. T. Dolgopolov, V. Pellegrini, F. Beltram, G. Biasiol, and L. Sorba, Phys. Rev. Lett. 99, 086802 (2007).; V. S. Khrapai, A. A. Shashkin, M. G. Trokina, V. T. Dolgopolov, V. Pellegrini, F. Beltram, G. Biasiol, and L. Sorba, Phy. Rev. Lett. 100, 196805 (2008). * (7) T. Ando, A. B. Fowler and F. Stern, Rev. Mod. Phys. 54, 437 (1982).; Akira Isihara and Ludvig Smrcka, J. Physics. C: Solid State Phys. 19, 6777 (1986).; A. AIsihara, Physica Scripta. Vol. 32, 26 (1985). * (8) S. Ilani, L. A. K. Donev, M. Kindermann and P. L. McEuen, Nat. Phys. 2, 687 - 691 (2006).; Jing Guo, Youngki Yoon, and Yijian Ouyang, Nano Lett., 7 (7), 1935 (2007). * (9) R. B. Dingle, Proc. R. Soc. A 211, 517(1952).; Akira Isihara and Ludvig Smrcka, J. Physics. C: Solid State Phys. 18, 4703(1985). * (10) Akira Endo and Yasuhiro IyE, J. Phys. Soc. Jpn 77, 064713 (2008), and references therein. * (11) V. P. Gusynin and S. G. Sharapov, Phys. Rev. Lett. 95 146801 (2005).; 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). * (12) I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products (Academic Press, New York, 1980).	arxiv-papers	2008-08-29T16:03:43	2024-09-04T02:48:57.549227	{ "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/0808.4120" }
0809.0062	# The Stochastic Logarithmic Norm for Stability Analysis of Stochastic Differential Equations Sk. Safique Ahmad safique@gmail.com Nagalinga Rajan rajan@rishi.serc.iisc.ernet.in Soumyendu Raha raha@serc.iisc.ernet.in ###### Abstract To analyze the stability of Itô stochastic differential equations with multiplicative noise, we introduce the stochastic logarithmic norm. The logarithmic norm was originally introduced by G. Dahlquist in 1958 as a tool to study the growth of solutions to ordinary differential equations and for estimating the error growth in discretization methods for their approximate solutions. We extend the concept to the stability analysis of Itô stochastic differential equations with multiplicative noise. Stability estimates for linear Itô SDEs using the one, two and $\infty$-norms in the $l$-th mean, where $1\leq l<\infty$, are derived and the application of the stochastic logarithmic norm is illustrated with examples. ###### keywords: Logarithmic norms, Stochastic differential equations. and and ††thanks: Scientific Computation Laboratory, Supercomputer Education and Research Centre, Indian Institute of Science, Bangalore 560012, India. The last author is the corresponding author. ††thanks: This work was partially supported by the Naval Research Board, DRDO, Government of India with grant number DNRD/05/4003/NRB/88. ## 1 Introduction For $A\in{\mathbb{C}}^{n\times n}$ and $X(t)\in{\mathbb{C}}^{n},$ we consider the ordinary differential equation (ODE) $dX_{t}=AX_{t}dt,X(0)=x_{0}.$ Then we have $\\|X(t)\\|\leq\\|x_{0}\\|e^{\mu(A)t}$ where $\mu(A)$ is the logarithmic norm of the matrix $A$ [4, 19, 20]. If $\mu(A)<0$, then the ODE is asymptotically stable. Also, $\mu(A)$ using the matrix $2$-norm gives an estimate [13] for the pseudospectrum [15] of $A$: $\max\Re\lambda_{\epsilon}(A)-\epsilon\leq\mu(A)$ where $1\gg\epsilon>0$. Since the pseudospectrum captures the stability of the numerical solution of the ODE over a finite number of time steps under the effect of local stiffness and nonnormality of $A$ (we shall refer to this as the numerical stability) [6, 7], having $\mu(A)<0$ implies numerical stability in addition to the asymptotic stability of the ODE. As is already shown in [6], transient numerical stability affects the computation and choice of methods for numerical integration of the ODE. In this paper we extend the classical logarithmic norm to the stability analysis of Itô stochastic differential equations (SDE) and introduce the stochastic logarithmic norm for estimating the numerical stability of an SDE in order to facilitate the selection of stiff and balanced stochastic numerical integration schemes. Letting $A,B\in{\mathbb{C}}^{n\times n},X(t)\in{\mathbb{C}}^{n},~{}B\in{\mathbb{C}}^{n\times n}$, the Itô SDE with a single channel of multiplicative noise is considered in the form of $dX_{t}=AX_{t}dt+BX_{t}dW_{t}$ given the initial condition that $X(0)=x_{0}$ with probability (w.p.) $1$, where $W$ is the one dimensional Wiener process such that $\int_{0}^{s}dW_{t}\sim N(0,s)$, i.e., is standard Gaussian distributed with mean $0$ and variance $s$. In our definition of the stochastic logarithmic norm we shall use the the matrix $p$-norm induced by the vector $p$-norm and the expectation of the $l$-th raw moment, i.e., the $l$-th mean of the solution to the linear multiplicative SDE. The stochastic logarithmic norm is computed over the sample paths as the expected logarithmic norm of the system in the sense of the existence of a generalized derivative of the Wiener process which itself is not obligatory differentiable with respect to time. Throughout the paper the following standard assumptions are made as in [9]. Let there be a common probability space $(\Omega,{\mathcal{A}},P)$ with index $t\in{\mathcal{T}}\subset{\mathbb{R}}$ on which the stochastic process $X(t)$ is a collection of random variables. The Wiener process $W=\\{W_{t},t\geq t_{0}\\}$ is associated with an increasing family of $\sigma$-algebras $\\{{\mathcal{A}}_{t},t\geq t_{0}\\}$. For the general case of multi- dimensional noise, each component of $\\{W_{t}^{(i)}\\}$ is ${\mathcal{A}}_{t}$-measurable with ${{\bf E}}(W(t_{0}))=0$ w.p. $1$,$~{}{{\bf E}}(W(t)\|{\mathcal{A}}_{t_{0}})=0$, $~{}~{}{{\bf E}}((W^{(i)}_{t}-W^{(i)}_{s})(W^{(j)}_{t}-W^{(j)}_{s})\|{\mathcal{A}}_{s})=\delta_{i,j}(t-s)$ for $t_{0}\leq s\leq t$ and $\Delta W=W(t_{n+1})-W(t_{n})$, the component wise increments of the multi-dimensional Wiener process, are independent of each other at all points in the partition of the time interval ${\mathcal{T}}$: $t_{0}\leq t_{1}\leq t_{2}\ldots\leq t_{r}\leq t_{r+1}\leq\ldots\leq t_{N}=t_{f}$. The initial value $X_{0}$ is assumed to be $\mathcal{A}_{t_{0}}$-measurable with $\left\\|X_{0}\right\\|_{p}<\infty$ w.p. $1$. All expectations on a function $\phi(X_{t})$ are evaluated as ${\bf E}(\phi(X_{t})\|{\mathcal{A}}_{t})$ unless otherwise stated. Inequalities and equalities involving random variables hold almost surely where applicable. The stability analysis in [17] uses test equations with scalar and $2$-by-$2$ matrix coefficients having multiplicative noise of dimension one [16]. Some of the analysis uses the classical logarithmic norm to establish the stability of the moment equations (derived from the SDE) which are deterministic. The present approach with stochastic logarithmic norm generalizes the stability analysis of SDEs as found in [16, 17]. ### 1.1 Classical Logarithmic Norm For $1\leq p\leq\infty,$ the $p$-norm on ${\mathbb{C}}^{n}$ is given by $\displaystyle\\|x\\|_{p}:=\left\\{\begin{array}[]{ll}\left(\sum^{n}_{j=1}\|x_{j}\|^{p}\right)^{1/p},&\mbox{ for }1\leq p<\infty,\\\ \max_{1\leq j\leq n}\|x_{j}\|,&\mbox{ for }p=\infty.\end{array}\right.$ Obviously, $\\|x\\|_{2}:=(x^{H}x)^{1/2}$ is the $2$-norm on ${\mathbb{C}}^{n}$. For $A\in{\mathbb{C}}^{n\times n},$ the spectrum $\Lambda(A)$ of $A$ is given by $\Lambda(A):=\\{\lambda\in{\mathbb{C}}:{\bf rank}(A-\lambda I)<n\\}.$ We denote a matrix $p$-norm on ${\mathbb{C}}^{n\times n}$ induced by the vector $p$-norm as $\\|\cdot\\|_{p}$ for $p=1,2,\infty$ and define these norms as $\\|A\\|_{2}:=\max_{j}\\{\sqrt{\lambda_{j}}:\lambda_{j}\ \in\Lambda(AA^{H})\\},$ $\;\\|A\\|_{1}:=\max_{j=1,\dots,n}\left(\sum_{i=1}^{n}\|a_{ij}\|\right)\,\mbox{and}\,\\|A\\|_{\infty}:=\max_{i=1,\dots,n}\left(\sum_{j=1}^{n}\|a_{ij}\|\right)$. Then the logarithmic norm for a single matrix is defined as $\mu_{p}(A):=\lim_{h\rightarrow 0^{+}}\frac{\\|I+hA\\|_{p}-1}{h}.$ For the $1,2$ and $\infty$-norms, the classical logarithmic norms, respectively, are computed ([5],Vol 1) as $\mu_{1}(A):=\max_{j}\left(\Re(a_{jj})+\sum_{j\not=i}^{n}\|a_{ij}\|\right),\,\mu_{2}(A):={\displaystyle{\frac{\lambda_{\max}(A+A^{H})}{2}}}\,$ and as $\mu_{\infty}(A):=\max_{i}\left(\Re(a_{ii})+\sum_{i\not=j}^{m}\|a_{ij}\|\right)$. ### 1.2 Stability of the SDE The stability of the vector SDE with single channel multiplicative noise is defined as follows. ###### Definition 1 [2, 3] The equilibrium solution $X_{t}\equiv 0$ to $dX_{t}=AX_{t}dt+BX_{t}dW_{t}$ is stochastically stable in the $l$-th mean ($l$ is a finite integer $\geq$ 1) using a vector $p$-norm if $\forall\epsilon>0$, $\exists\delta>0$ such that ${\bf E}(\\|X(t)\\|_{p}^{l})<\epsilon\quad\forall\,t\geq t_{0}\quad and\quad\\|X(t_{0})\\|_{p}<\delta~{}w.p.~{}1$ (2) and is asymptotically stable in $l$-th mean if in addition, $\exists\delta_{0}>0$ such that $\lim_{t\to\infty}{\bf E}(\\|X_{t}\\|_{p}^{l})=0\quad\forall\,\,\\|X(t_{0})\\|_{p}<\delta_{0}~{}w.p.~{}1.$ (3) ## 2 Background In the following items we review the existing stability analysis of linear stochastic differential equations with multiplicative noise. * $(a)$ In [16] scalar stochastic differential equations of the form $dX_{t}=\lambda X_{t}dt+\beta X_{t}dW_{t}$ with $X_{0}=1$ w.p. $1$, where $\lambda$ and $\beta$ are constants have been considered and it is shown that the above stochastic differential equation is mean square stable using the $2$-norm if $2\Re(\lambda)+\|\beta\|^{2}\leq 0$ when $\lambda,\beta$ are complex scalars. * $(b)$ In [17] vector stochastic differential equations, with single channel multiplicative noise, of the form $dX_{t}=DX_{t}dt+BX_{t}dW_{t},$ where $D=\left(\matrix{\lambda_{1}&0\cr 0&\lambda_{2}}\right)$ and $B=\left(\matrix{\alpha_{1}&\beta_{1}\cr\beta_{2}&\alpha_{2}}\right)$ have been analyzed for mean square stability. It is shown that the above stochastic differential equation is mean square stable in the $\infty$ norm if $\max\\{2\lambda_{1}+(\|\alpha_{1}\|+\|\beta_{1}\|)^{2},2\lambda_{2}+(\|\alpha_{2}\|+\|\beta_{2}\|)^{2}\\}<0.$ * $(b_{1})$ In [16] for the same stochastic differential equation as in $(a)$ the Mil’stein scheme [11] $X_{n+1}=X_{n}+\lambda X_{n}h+\mu X_{n}\Delta W+\frac{(\Delta W)^{2}-h}{2}\mu^{2}X_{n}$ with $O(h^{1.5})$ root mean square error is analyzed for stability in the mean square sense. The mean square stability function $R(h)=\|1+h\lambda\|^{2}+\|h\mu^{2}\|+\frac{1}{2}\|h^{2}\ \mu^{4}\|$ is derived and it is shown that the SDE is stochastically asymptotically stable in the mean square when $R(h)<1$. * $(b_{2})$ The reference [16] also considered the SDE as given in $(b)$ and applied the Euler-Maruyama numerical integration scheme as $X_{n+1}=X_{n}+hDX_{n}+BX_{n}\Delta W,$ where $h$ and $\Delta W$ stand for step-size and the increment of the Wiener process, respectively. Then the discretized SDE is shown to be stochastically asymptotically stable if $\max\\{(1+\lambda_{1}h)^{2}+(\|\alpha_{1}\|+\|\beta_{1}\|)^{2},(1+\lambda_{2}h)^{2}+(\|\alpha_{2}\|+\|\beta_{2}\|)^{2}\\}<1.$ Next we briefly review a few essential aspects of the logarithmic norm that are used in the stability analysis of ODEs. * $(c)$ In [4, 10] the logarithmic norm was introduced in order to derive error bounds for the solution of initial value ODE problems using differential inequalities that distinguish between forward and reverse time integration. This led to requirements for the stability of initial value and boundary value ODE problems. The classical analysis, using the vector norm $\\|\cdot\\|{}$ on ${\mathbb{C}}^{n}$ and the sub-ordinate matrix norm on ${\mathbb{C}}^{n\times n}$, defined the logarithmic norm of a matrix $A$ as $\mu_{p}(A):=\lim_{h\rightarrow 0^{+}}\frac{\\|I+hA\\|_{p}-1}{h}.$ * $(d)$ More recently Söderlind [20] considered $f:D\subset X\rightarrow X$ and defined least upper bounds (lub) and greatest lower bounds (glb) Lipschitz constants by $L[f]=\sup_{u\not=v}\frac{f(u)-f(v)}{\|u-v\|},~{}~{}l[f]=\inf_{u\not=v}\frac{f(u)-f(v)}{\|u-v\|},$ for $u$ and $v\in D,$ where the domain is path connected and $l[f]\|u-v\|\leq\|f(u)-f(v)\|\leq L[f]\|u-v\|.$ If $l[f]>0,$ then $f(u)\rightarrow f(v)$ implies that $u\rightarrow v.$ Then $f$ is an injection, with an inverse on $f(D),$ the same as a matrix $A\in{\mathbb{C}}^{n\times n}$ being invertible if its glb is strictly positive. Also then, $L[f^{-1}]=l[f^{-1}],$ where $L[f^{-1}]$ is defined over $f(D).$ If $f=A$ is a linear map then $L[A]=\\|A\\|.$ Hence $L[.]$ is left and right $G$-differentiable for the class of Lipschitz maps. This allows one to define the lub and glb logarithmic Lipschitz constant, by $M[f]=\lim_{h\rightarrow 0^{+}}\frac{L[I+hf]-1}{h},m[f]=\lim_{h\rightarrow 0^{-}}\frac{L[I+hf]-1}{h}.$ The lub logarithmic Lipschitz generalizes the classical logarithmic norm for every matrix $A,$ so that, $M[A]=\mu(A).$ In the following section we develop the stochastic logarithmic norm as an upper bound estimate of the rate of growth of the solution of a multiplicative noise linear SDE. The rate of growth is analyzed as a Dini derivative of the $l$-th mean in vector $p$-norm of the solution. This approach may be seen as a special case of the modern definition of logarithmic norm in [20] and as an extension of the classical definition as in [4, 10, 19]. Other works [8] deal with logarithmic norm of matrix pencils with one invertible matrix. However in our current treatment we do not use matrix pencils for defining the stochastic logarithmic norm. ## 3 Definition of the Stochastic Logarithmic Norm A linear Itô SDE with a single channel of multiplicative noise is written as $dX_{t}=AX_{t}dt+BX_{t}dW_{t},$ (4) where $A,B\in{\mathbb{C}}^{n\times n},$ are constant matrices. For the non- linear Itô SDE with a single channel multiplicative noise given by $dX_{t}=f(X,t)dt+g(X,t)dW_{t},$ (5) the strong order $1.0$ Itô-Taylor expansion (Mil’stein scheme when used in numerical integration) of $X_{t}$ at $t=t_{n+1}$ is given as $X_{n+1}=X_{n}+f(X_{n},t_{n})h+g(X_{n},t_{n})\Delta W+\left(g\frac{\partial{g}}{\partial{x}}\right)_{(X_{n},t_{n})}\frac{((\Delta W)^{2}-h)}{2}+R,$ (6) where $h=t_{n+1}-t_{n}$ is the step-size, $W(t_{n+1})-W(t_{n})=:\Delta W\sim N(0,h)$ and $R$ are the $O(h^{1.5})$ remainder terms in the root mean square sense. When linearized at $t=t_{n}$, $A:=\left(\frac{\partial f}{\partial x}\right)_{n}$ and $B:=\left(\frac{\partial g}{\partial x}\right)_{n}$. We shall define the stochastic logarithmic norm by studying the growth of the $l$-th mean of the solution $X$ of the linear SDE (4) (linearized SDE in case of SDE (5)) w. r. t. time. We seek an upper bound for the growth rate of ${\bf E}(\\|X\\|_{p}^{l})$ using the upper-right Dini derivative which for any function $\Xi(t)$ w.r.t $t$ is defined as $D_{+}\left(\Xi(t)\right)=\lim_{h\to 0+}\frac{\Xi(t+h)-\Xi(t)}{h}.$ (7) The strong order $1.0$ Itô-Taylor expansion (linearized at $t$ in case of the non-linear SDE (5)) applied to $X_{t}$ over the time interval $[t,t+h]$ gives $X(t+h)=X(t)+hAX(t)+\Delta WBX(t)+\frac{(\Delta W)^{2}-h}{2}B^{2}X(t)+R,$ (8) with the remainder terms $R$ being $O(h^{1.5})$ in the root mean square sense. In the above expansion (8) $\Delta W:=\left\\{\Delta W(t),t\geq 0\right\\}$ are the independent increments of a Wiener process over the interval $[t,t+h]$. Taking the $p$-norms and raising both sides of (8) to the power of $l$, we can write the following inequality: $\\|X(t+h)\\|_{p}^{l}\leq\\|I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}+R_{x}\\|_{p}^{l}\\|X(t)\\|_{p}^{l},$ (9) where $R_{x}$ are root mean square $O(h^{1.5})$ remainder terms. For the $l$-th mean using a $p$-norm, we apply expectation to both sides of the above inequality and get, almost surely (a.s.), ${\bf E}\\|X(t+h)\\|_{p}^{l}\leq{\bf E}\left(\\|I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}+R_{x}\\|_{p}^{l}\right){\bf E}\\|X(t)\\|_{p}^{l}.$ (10) observing that $X(t)$ is independent of the Wiener increment $\Delta W$ since the Wiener process is a non-anticipative process. Then, we estimate the expected rate of growth as ${\bf E}\left(D_{+}\\|X(t)\\|_{p}^{l}\right)\leq\lim_{h\to 0^{+}}\frac{{\bf E}(\\|I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}+R_{x}\\|_{p}^{l})-1}{h}{\bf E}(\\|X(t)\\|_{p}^{l}).$ Based on the limit term on the right hand side above we introduce the stochastic logarithmic norm. ###### Definition 2 The stochastic logarithmic norm of a square matrix pair of same dimensions, $(A,B)$, in the $l$-th mean using a matrix $p$-norm is defined as $\nu_{p}^{l}\left(A,B\right)=\lim_{h\to 0^{+}}\frac{{\bf E}\left(\\|I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}\\|_{p}^{l}\right)-1}{h},$ (11) where the limit is taken in the sense of the existence of the generalized derivative of the Wiener process and $\\|A\\|_{p},\\|B\\|_{p}$ are assumed to be finite. From the above definition the expected rate of growth of the solution can be estimated as ${\bf E}\left(D_{+}\\|X(t)\\|_{p}^{l}\right)\leq\nu_{p}^{l}\left(A,B\right){\bf E}(\\|X(t)\\|_{p}^{l})$ since $R_{x}$ are of root mean square $O(h^{1.5})$. Obviously the weaker estimate $D_{+}{\bf E}\\|X(t)\\|_{p}^{l}\leq\nu_{p}^{l}\left(A,B\right){\bf E}(\\|X(t)\\|_{p}^{l})$ also holds so that ${\bf E}\\|X(t)\\|_{p}^{l}\leq e^{\nu_{p}^{l}\left(A,B\right)t}\\|X(t_{0})\\|_{p}^{l}$ for the linear SDE (4). The linear SDE (4) is stochastically stable in the $l$-th mean using a $p$-norm when $\nu_{p}^{l}\left(A,B\right)\leq 0$ and is asymptotically stable in the $l$-th mean using a $p$-norm when $\nu_{p}^{l}\left(A,B\right)<0$. The linearized SDE (5) (and hence the linear SDE (4)) is numerically stochastically stable in the $l$-th mean using the $p$-norm if $\nu_{p}^{l}\leq 0$ since over $k~{}(k\ll\infty)$ time steps each of size sufficiently small $h_{i}>0$ almost surely we have ${\bf E}\\|X_{t+kh}\\|_{p}^{l}\leq e^{\sum_{i=1}^{k}\nu_{p}^{l}(A_{i},B_{i})h}{\bf E}\\|X_{t}\\|_{p}^{l}$ when each $\nu_{p}^{l}(A_{i},B_{i})\leq 0$. If ${\bf E}\\|X_{t}\\|_{p}^{l}$ is finitely bounded w.p. $1$, then ${\bf E}\\|X_{t+kh}\\|_{p}^{l}$ is almost surely finitely bounded. Later in the paper we relate the stochastic logarithmic norm to the pseudospectrum of the Ito stability matrix $A-\frac{1}{2}B^{2}$ for the linear SDEs with multiplicative noise. It may be remarked that for an SDE with additive noise and for an ODE, the stochastic logarithmic norm is given as $\nu_{p}^{1}(A,0)=\mu_{p}(A)$. ## 4 Some Estimates of the Stochastic Logarithmic Norm In this section we derive some estimates of the stochastic logarithmic norm using the matrix $p$-norm, where $p=1,2,\infty.$ The estimates show the incremental behavior of the stochastic logarithmic norm under perturbations and also the effect of noise on the deterministic ODE. Let $\lambda_{\max}(A)$ be the largest eigenvalue of any square matrix $A$. We state the following Lemma from pp.62, [1] which we shall use in estimating the stochastic logarithmic norm while using the matrix $2$-norm. ###### Lemma 1 Let $A,B\in{\mathbb{C}}^{n\times n}$ be Hermitian matrices. Then $\lambda_{\max}{(A)}+\lambda_{\min}{(B)}\leq\lambda_{\max}(A+B)\leq\lambda_{\max}(A)+\lambda_{\max}(B).$ The equality holds when $B=kI,$ where $k$ is a scalar constant and $I$ is an identity matrix. For the matrix $2$-norm and in the $l^{th}$ mean we can estimate the following from (11). ###### Theorem 5 The stochastic logarithmic norm in the $l^{th}$ mean, where $1\leq l<\infty$ using the matrix $2$-norm satisfies the following bounds in the sense of existence of a generalized derivative for the Wiener process as $\zeta dt=dW_{t},~{}\zeta\sim N(0,1)$. $\displaystyle\nu_{2}^{l}(A,B)$ $\displaystyle\leq$ $\displaystyle\frac{l}{2}\lambda_{\max}(A+A^{H})+\frac{l}{4}\left(\lambda_{\max}(B+B^{H})+\lambda_{\max}(-B-B^{H})\right)$ (12) $\displaystyle+\frac{l}{2}\lambda_{\max}(B^{H}B)+\frac{l(l-2)}{8}\lambda_{\max}^{2}(B+B^{H}),~{}~{}l>2,$ $\displaystyle\nu_{2}^{l}(A,B)$ $\displaystyle\leq$ $\displaystyle\frac{l}{2}\lambda_{\max}(A+A^{H})+\frac{l}{4}\left(\lambda_{\max}(B+B^{H})+\lambda_{\max}(-B-B^{H})\right)$ (13) $\displaystyle+\frac{l}{2}\lambda_{\max}(B^{H}B),~{}l\leq 2,$ $\displaystyle\nu_{2}^{l}(A,I)$ $\displaystyle=$ $\displaystyle\frac{l}{2}\lambda_{\max}(A+A^{H})+\frac{l}{2}+\frac{l(l-2)}{2},$ (14) where $A,B\in{\mathbb{R}}^{n\times n}$ are the (linearized) drift and diffusion coefficient matrices in a vector SDE with a single channel of Wiener process noise. Proof: For $1\leq l<\infty$ we have from (11) $\nu_{2}^{l}\left(A,B\right)=\lim_{h\to 0^{+}}\frac{{\bf E}\left(\\|I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}\\|_{2}^{l}\right)-1}{h}=$ $\lim_{h\to 0^{+}}\frac{{\bf E}\left(\sqrt{\lambda_{\max}\left((I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2})^{H}(I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2})\right)}\right)^{l}}{h}.$ Then we may write $\displaystyle(I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2})^{H}(I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2})$ $\displaystyle=(I+hA^{H}+\Delta WB^{H}+\frac{(\Delta W)^{2}-h}{2}(B^{H})^{2})(I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2})=$ $\displaystyle I+hA+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}+hA^{H}+\Delta WB^{H}+(\Delta W)^{2}B^{H}B$ $\displaystyle+\frac{((\Delta W)^{2}-h)(B^{H})^{2}}{2}+\cdots=I+h(A+A^{H})+\Delta W(B+B^{H})+(\Delta W)^{2}B^{H}B+$ $\displaystyle\frac{(\Delta W)^{2}-h}{2}(B^{2}+(B^{H})^{2})+\cdots.$ It is possible to write $C=h(A+A^{H})+\Delta W(B+B^{H})+(\Delta W)^{2}B^{H}B+\frac{(\Delta W)^{2}-h}{2}(B^{2}+(B^{H})^{2})$ in a series in the normalized time step size $h\ll 1$. We use the identities $\lambda_{\max}(cA)=\|c\|\lambda_{\max}(\mathrm{sign}(c)A)$ (where $c$ is a constant, $\mathrm{sign}(c)={\displaystyle{\frac{z}{\|z\|}}},z\not=0$) and $\lambda_{\max}^{2}(A)=\lambda_{\max}(A^{2})$ for estimating the terms in the series. Using the triangle inequality and the identity $[\lambda_{\max}(I+C)]^{\frac{l}{2}}=(1+\lambda_{\max}(C))^{\frac{l}{2}}$, the following can be written: $\displaystyle(1+\lambda_{\max}(C))^{\frac{l}{2}}=1+\frac{l}{2}\lambda_{\max}(C)+\frac{l(l-2)}{8}\lambda_{\max}(C)^{2}+\dots=1+$ $\displaystyle\frac{l}{2}\lambda_{\max}\left(h(A+A^{H})+\Delta W(B+B^{H})+(\Delta W)^{2}B^{H}B+\frac{(\Delta W)^{2}-h}{2}(B^{2}+(B^{H})^{2})\right)+$ $\frac{l(l-2)}{8}\lambda_{\max}\left(h(A+A^{H})+\Delta W(B+B^{H})+(\Delta W)^{2}B^{H}B+\frac{(\Delta W)^{2}-h}{2}(B^{2}+(B^{H})^{2})\right)^{2}$ $\displaystyle+\cdots\leq 1+\frac{l}{2}h\lambda_{\max}(A+A^{H})+\frac{l}{2}\lambda_{\max}(\Delta W(B+B^{H}))+\frac{l}{2}(\Delta W)^{2}\lambda_{\max}(B^{H}B)$ $\displaystyle+((\Delta W)^{2}-h)\frac{l}{4}\lambda_{\max}(B^{2}+(B^{H})^{2})+(\Delta W)^{2}\frac{l(l-2)}{8}\lambda_{\max}^{2}(B+B^{H})+\cdots.$ (15) The equality holds when $B=I$ and follows from Lemma 1. Taking expectation on both sides and in the sense of generalized derivative of the Wiener process which is the Gaussian white noise $\zeta$ we have ${\bf E}(\lambda_{\max}(I+C))^{\frac{l}{2}}={\bf E}(1+\lambda_{\max}(C))^{\frac{l}{2}}\leq 1+\frac{l}{2}h\lambda_{\max}(A+A^{H})+{\bf E}(\|\zeta\|)h\frac{l}{4}(\lambda_{\max}(B+B^{H})+\lambda_{\max}(-B-B^{H}))+{\bf E}((\Delta W)^{2})\frac{l}{2}\lambda_{\max}(B^{H}B)+({\bf E}((\Delta W)^{2})-h)\frac{l}{4}\lambda_{\max}(B^{2}+(B^{H})^{2})+{\bf E}((\Delta W)^{2})\frac{l(l-2)}{8}\lambda_{\max}^{2}(B+B^{H})+O(h^{1.5})=1+h\frac{l}{2}\lambda_{\max}(A+A^{H})+h\frac{l}{4}(\lambda_{\max}(B+B^{H})+\lambda_{\max}(-B-B^{H}))+\frac{l}{2}h\lambda_{\max}(B^{H}B)-h\frac{l(l-2)}{8}\lambda_{\max}^{2}(B+B^{H})+O(h^{1.5}).$ For $h\to 0^{+}$ we obtain $\lim_{h\rightarrow 0^{+}}\frac{{\bf E}(\lambda_{\max}(1+C))^{\frac{l}{2}}-1}{h}\leq\frac{l}{2}\lambda_{\max}(A+A^{H})+\frac{l}{4}(\lambda_{\max}(B+B^{H})+\lambda_{\max}(-B-B^{H}))$ $+\frac{l}{2}\lambda_{\max}(B^{H}B)+\frac{l(l-2)}{8}\lambda_{\max}^{2}(B+B^{H}).$ $(b)$ When $B=I,$ we have $\displaystyle\lim_{h\rightarrow 0^{+}}\frac{{\bf E}(\lambda_{\max}(I+C))^{\frac{l}{2}}-1}{h}$ $\displaystyle=$ $\displaystyle\frac{l}{2}\lambda_{\max}(A+A^{H})+\frac{l}{2}\lambda_{\max}(B^{H}B)+$ $\displaystyle\frac{l(l-2)}{8}\lambda_{\max}^{2}(B+B^{H})$ $\displaystyle\Rightarrow\nu_{2}^{l}(A,B)$ $\displaystyle=$ $\displaystyle\frac{l}{2}\lambda_{\max}(A+A^{H})+\frac{l}{2}+\frac{l(l-2)}{2}.\blacksquare$ ###### Corollary 1 For a positive integer $l>2$, $\displaystyle\nu_{2}^{l}(A,B)$ $\displaystyle\leq$ $\displaystyle l\left(\mu_{2}(A)+\frac{1}{2}\\|B\\|_{2}^{2}+\frac{1}{2}(\mu_{2}(B)+\mu_{2}(-B))+\frac{l-2}{2}(\mu_{2}(B))^{2}\right)$ (16) For $l\leq 2$, $\displaystyle\nu_{2}^{l}(A,B)$ $\displaystyle\leq$ $\displaystyle l\left(\mu_{2}(A)+\frac{1}{2}(\mu_{2}(B)+\mu_{2}(-B))+\frac{1}{2}\\|B\\|_{2}^{2}\right)$ (17) Proof: The results (16) and (17) follow directly from (15) in Theorem 5 by Lemma 1 and the Itô isometry for the expectation of the Itô integrals.$\blacksquare$ The inequality (16) has been used for mean square stability estimates in [17]. ###### Corollary 2 If $l=1,2$ and $p=2$, that is, in the mean and in the mean square and using the matrix two norm, the following results hold when $B$ is the identity matrix $I$. $\displaystyle\nu_{2}^{1}(A,I)=\lambda_{\max}\left(\frac{A+A^{H}}{2}\right)+\frac{1}{2}-\frac{4}{8}=\lambda_{\max}\left(\frac{A+A^{H}}{2}\right)=\mu_{2}(A),$ (18) $\displaystyle\nu_{2}^{2}(A,I)=\lambda_{\max}\left(A+A^{H}\right)+1=2\mu_{2}(A)+1,$ (19) Proof: The results follow from Theorem 5, by substituting $p=2,l=1$ and $p=2,l=2.\blacksquare$ Some properties of the stochastic logarithmic norm under perturbation is obtained in the following result. In the generalized sense of the Wiener process derivative a couple of useful inequalities showing how the Wiener process perturbs the deterministic ODE is also given. ###### Theorem 6 For any square matrices of same dimensions, $B,A,\Delta A,\Delta B$ and a real number $\alpha>0$ the stochastic logarithmic norm has the following properties. $\displaystyle\nu_{p}^{l}(\alpha A,\sqrt{\alpha}B)$ $\displaystyle=$ $\displaystyle\alpha\nu_{p}^{l}(A,B)$ (20) $\displaystyle\nu_{p}^{1}(A+\Delta A,B+\Delta B)$ $\displaystyle\leq$ $\displaystyle\nu_{p}^{1}(A,\sqrt{2}B)+\nu_{p}^{1}(\Delta A,\sqrt{2}\Delta B)+\frac{1}{\sqrt{2}}\\|(B-\Delta B)^{2}\\|_{p}$ (21) $\displaystyle\nu_{p}^{l}(A+\Delta A,B+\Delta B)$ $\displaystyle\leq$ $\displaystyle\nu_{p}^{l}(A,\frac{B+\Delta B}{\sqrt{2}})+\nu_{p}^{l}(\Delta A,\frac{B+\Delta B}{\sqrt{2}})$ (22) $\displaystyle\nu_{p}^{l}(A,0)$ $\displaystyle=$ $\displaystyle l\mu_{p}(A)$ (23) Further, in the sense of the existence of a generalized derivative of the Wiener process which is a Gaussian white noise, the following estimate holds for a positive integer $l$ and for any of the matrix $p$-norms. $\displaystyle\nu^{l}_{p}(A,B)$ $\displaystyle\leq$ $\displaystyle l\mu_{p}(A)+\frac{l}{2}\mu_{p}(-B^{2})+\frac{l(l+1)}{4}\\|B\\|^{2}_{p}+l\\|B\\|_{p}$ (24) $\displaystyle\leq$ $\displaystyle l\mu_{p}(A)+l\\|B\\|_{p}\left(1+\frac{l+3}{4}\\|B\\|_{p}\right)\leq l\mu_{p}(A)+l\left(1+\frac{l+3}{4}\\|B\\|_{p}\right)^{2}$ (25) Proof: We can scale the Wiener process as $\sqrt{\alpha}$ and write $\displaystyle\nu_{p}^{l}(\alpha A,\sqrt{\alpha}B)$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}\alpha\frac{{\bf E}\left(\\|I+(\alpha h)A+\Delta(\sqrt{\alpha}W)B+\frac{(\Delta\sqrt{\alpha}W)^{2}-\alpha h}{2}B^{2}\\|_{p}^{l}\right)-1}{\alpha h}$ $\displaystyle=$ $\displaystyle\alpha\nu_{p}^{l}(A,B)$ for any positive integer $l$ and any of the matrix $p$-norms. From the definition of the stochastic logarithmic norm and scaling the Wiener process, we can write $\nu_{p}^{1}(A+\Delta A,B+\Delta B)\leq\lim_{h\to 0^{+}}\bigg{(}\frac{{\bf E}(\\|I+(2h)A+(\sqrt{2}\Delta W)(\sqrt{2}B)+\frac{(\Delta\sqrt{2}W)^{2}-2h}{2}(\sqrt{2}B)^{2}\\|_{p})-1}{2h}$ $+\frac{{\bf E}(\\|I+(2h)\Delta A+(\sqrt{2}\Delta W)(\sqrt{2}\Delta B)+\frac{(\Delta\sqrt{2}W)^{2}-2h}{2}(\sqrt{2}\Delta B)^{2}\\|_{p})-1}{2h}+$ $\frac{{\bf E}\|\int_{0}^{h}W_{u}dW_{u}\|}{h}\\|B\Delta B+\Delta BB-B^{2}-(\Delta B)^{2}\\|_{p}\bigg{)}\\\ \leq\nu_{p}^{1}(A,\sqrt{2}B)+\nu_{p}^{1}(\Delta A,\sqrt{2}\Delta B)+\frac{1}{\sqrt{2}}\\|(B-\Delta B)^{2}\\|_{p}$ since $\lim_{h\to 0^{+}}\frac{{\bf E}\|\int_{0}^{h}W_{u}dW_{u}\|}{h}=h/(\sqrt{2}h)=\frac{1}{\sqrt{2}}$. For any positive integer $l$, the above can be re-written as $\displaystyle\nu_{p}^{l}(A+\Delta A,B+\Delta B)$ $\displaystyle\leq$ $\displaystyle\lim_{h\to 0^{+}}\bigg{(}\bigg{(}2^{l-1}\frac{1}{2^{l-1}}{\bf E}\\|I+(2h)A+(\sqrt{2}\Delta W)(\sqrt{2}\frac{B+\Delta B}{2})$ $\displaystyle+\frac{(\Delta\sqrt{2}W)^{2}-2h}{2}(\sqrt{2}\frac{B+\Delta B}{2})^{2}\\|_{p}^{l}-1\bigg{)}/(2h)$ $\displaystyle+\bigg{(}2^{l-1}\frac{1}{2^{l-1}}{\bf E}\\|I+(2h)\Delta A+(\sqrt{2}\Delta W)(\sqrt{2}\frac{B+\Delta B}{2})$ $\displaystyle+\frac{(\Delta\sqrt{2}W)^{2}-2h}{2}(\sqrt{2}\frac{B+\Delta B}{2})^{2}\\|_{p}^{l}-1\bigg{)}/(2h)\bigg{)}$ $\displaystyle\leq\nu_{p}^{l}(A,\frac{B+\Delta B}{\sqrt{2}})+\nu_{p}^{l}(\Delta A,\frac{B+\Delta B}{\sqrt{2}}).$ Applying (21) recursively to $\nu_{p}^{l}(A+0,B/2+B/2)$ and using (20), we obtain $\nu_{p}^{l}(A,B)\leq\lim_{n\to\infty}(\nu_{p}^{l}(A,\frac{1}{(\sqrt{2})^{n}}B)+(2^{n}-1)\nu_{p}^{l}(0,\frac{1}{(\sqrt{2})^{n}}B))=\nu_{p}^{l}(A,0)+\lim_{n\to\infty}(2^{n}/2^{n})\nu^{l}_{p}(0,B)+0=\nu_{p}^{l}(A,0)+\nu_{p}^{l}(0,B)$. For $l=1$, the inequality reduces to $\nu_{p}^{1}(A,B)\leq\mu_{p}(A)+\nu_{p}^{1}(0,B)$. For the deterministic case with no noise, we have $\displaystyle\nu_{p}^{l}(A,0)$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}\frac{\\|I+hA\\|^{l}_{p}-1}{h}$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}l\\|I+hA\\|^{l-1}_{p}D_{+,h}\\|I+hA\\|_{p}$ $\displaystyle=$ $\displaystyle\lim_{\epsilon\to 0^{+}}l\frac{\\|I+\epsilon A\\|_{p}-1}{\epsilon}$ $\displaystyle=$ $\displaystyle l\mu_{p}(A).$ For the stochastic logarithmic norm of the diffusion coefficient $B$ it is possible to write $\displaystyle\nu_{p}^{l}(0,B)$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}\frac{{\bf E}\\|I+\Delta WB+\frac{(\Delta W)^{2}-h}{2}B^{2}\\|_{p}^{l}-1}{h}$ $\displaystyle\leq$ $\displaystyle\lim_{h\to 0^{+}}\left(\frac{\\|I-(2h)\frac{1}{2}B^{2}\\|_{p}^{l}-1}{(2h)}+\frac{{\bf E}\\|I+2B\Delta W+B^{2}(\Delta W)^{2}\\|^{l}_{p}-1}{(2h)}\right)$ $\displaystyle\leq$ $\displaystyle\nu_{p}^{l}(-\frac{1}{2}B^{2},0)+\frac{l}{2}\\|B\\|^{2}_{p}+l\\|B\\|_{p}+\frac{l(l-1)}{4}\\|B\\|^{2}_{p}$ in the sense of the existence of a generalized derivative of the Wiener process so that $\int_{0}^{h}\xi dt=\int_{0}^{h}dW_{s}$ where $\xi\sim N(0,1)$ is a Gaussian white noise. Further we have $\nu_{p}^{l}(-\frac{1}{2}B^{2},0)\\\ =\frac{l}{2}\mu_{p}(-B^{2})$ and $\frac{l}{2}\|\mu_{p}(-B^{2})\|\leq\frac{l}{2}\\|B\\|_{p}^{2}$ from the properties of the logarithmic norm [5]. Hence the estimates (24) and (25) . $\blacksquare$ We remark that the stochastic logarithmic norm does not satisfy the triangle inequality property nor the that of the multiplication by a scalar in the same way as the (deterministic) logarithmic norm. However, it is consistent with Itô calculus in its property (20) of multiplication with a scalar. Again, Itô calculus makes the noise “redistribute“ for any additive perturbation to the drift and diffusion coefficients as found in (22). The logarithmic norm coincides with the stochastic logarithmic norm in special cases. One such case is obtained from Theorem 5 by setting $l=1$ and $B=I$ (and also by setting $B=0$ in the trivial case) for $p=2$. ###### Theorem 7 $\max_{\lambda}\Re(\lambda(A))\leq\frac{1}{2}\nu_{2}^{2}(A,I)-\frac{1}{2}$ where $I$ is the identity matrix. Proof: From (19) we have $\nu_{2}^{2}(A,I)=2\mu_{2}(A)+1$. The logarithmic norm has the lower bound property [5]: $\max_{\lambda}\Re(\lambda(A))\leq\mu_{2}(A)$. Hence the inequality. $\blacksquare$ ## 8 Conditions for Mean and Mean Square Stability We recall that the linear SDE (4) is stochastically stable in the $l$-th mean using a $p$-norm if $\nu_{p}^{l}(A,B)\leq 0.$ To estimate of the stability of a linear SDE one of the upper bounds derived in the last section may be used, especially, when it is needed to compute incrementally the effect of adding noise to an ODE or an SDE with known stability estimates. In practice, when using the upper bounds for estimating the stability of an SDE incrementally, a small positive number (determined by the stochastic stability region of the numerical integrator) is used as cut-off rather than a very small tolerance or zero so that the effect of non-normality and stiffness [6] in the SDE’s (both linear and locally linearized) transient numerical behavior in the stochastic logarithmic norm is taken into account. The following results are a couple of special cases of the stochastic logarithmic norm approach to the stability of SDEs with multiplicative noise. * • We consider the scalar case $A=\alpha\in{\mathbb{C}}$ and $B=\beta\in{\mathbb{C}}$ in (4). Then we apply Theorem 5$(a)$ and consider (16) for $l=1,2$ the condition $\nu_{p}^{l}(A,B)<0,$ so that for $p=2$, we get the condition for stochastic stability in the mean as $\displaystyle\Re(\alpha)+\frac{1}{2}\|\beta\|^{2}\leq 0$ (26) and in the mean square as $\displaystyle 2\Re(\alpha)+\|\beta\|^{2}\leq 0.$ (27) The above conditions are essentially the same but in practice the $0$ on the right hand side is replaced by $TOL$ which is a small positive real number [13] and thus $l$ becomes significant for estimates in higher moments. These results are the same as in review $(a)$ of Section 2. * • For the matrices $A=\left(\matrix{\lambda_{1}&0\cr 0&\lambda_{2}}\right)$ and $B=\left(\matrix{\alpha_{1}&\beta_{1}\cr\beta_{2}&\alpha_{2}}\right),$ the SDE (4) is mean square stable in the $\infty$-norm if $\displaystyle 2\max\\{\lambda_{1},\lambda_{2}\\}+2\left(\frac{5}{4}\max\\{\|\alpha_{1}\|+\|\beta_{1}\|,\|\alpha_{2}\|+\|\beta_{2}\|\\}+1\right)^{2}\leq 0.$ (28) The above condition is obtained using (25) in which $\displaystyle\nu^{2}_{\infty}(A,0)$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}\frac{\left(\max\\{\|1+h\lambda_{1}\|,\|1+h\lambda_{2}\|\\}\right)^{2}-1}{h}$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}2\max\\{\|1+h\lambda_{1}\|,\|1+h\lambda_{2}\|\\}\max\\{\lambda_{1},\lambda_{2}\\}$ $\displaystyle=$ $\displaystyle 2\max_{i=1,2}{\lambda_{i}}$ and $\\|B\\|_{\infty}=\max\\{\|\alpha_{1}\|+\|\beta_{1}\|,\|\alpha_{2}\|+\|\beta_{2}\|\\}.$ ## 9 Direct Computation of the Stochastic Logarithmic Norm and Sharper Bounds ###### Theorem 10 In the sense of the existence of a generalized derivative for the Wiener process, i.e., $\zeta dt=dW_{t},~{}\zeta\sim N(0,1)$, we can compute the stochastic logarithmic norm directly as follows. $\nu_{p}^{l}(A,B)=l{\bf E}\left(\mu_{p}\left(A-\frac{1}{2}B^{2}+B\zeta\right)\right)$ Proof: $\displaystyle\nu_{p}^{l}(A,B)$ $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}\frac{{\bf E}\\|I+hA-\frac{1}{2}B^{2}h+B\Delta W+\frac{1}{2}B^{2}(\Delta W)^{2}\\|_{p}^{l}-1}{h}$ (29) $\displaystyle=$ $\displaystyle\lim_{h\to 0^{+}}{\bf E}\bigg{(}l\\|I+hA-\frac{1}{2}B^{2}h+B\Delta W+\frac{1}{2}B^{2}(\Delta W)^{2}\\|_{p}^{l-1}$ $\displaystyle\times D_{+,h}\\|I+hA-\frac{1}{2}B^{2}h+B\Delta W+\frac{1}{2}B^{2}(\Delta W)^{2}\\|_{p}\bigg{)}$ $\displaystyle=$ $\displaystyle l\lim_{\epsilon\to 0^{+}}\frac{{\bf E}\\|I+(A-\frac{1}{2}B^{2})\epsilon+B\int_{0}^{\epsilon}\zeta ds+\frac{1}{2}B^{2}\left(\int_{0}^{\epsilon}\zeta ds\right)^{2}\\|-1}{\epsilon}$ $\displaystyle=$ $\displaystyle l{\bf E}\left(\mu_{p}\left(A-\frac{1}{2}B^{2}+B\zeta\right)\right),$ where the last equality follows from considering the Itô formula along with the generalized derivative of the Wiener process. $\blacksquare$ The above result shows that the stochastic logarithmic norm is the expected logarithmic norm behavior of the SDE. Consequently we can derive the following inequalities. ###### Corollary 3 $\displaystyle\nu_{p}^{l}(A,B)$ $\displaystyle\leq$ $\displaystyle l\mu_{p}(A)+\frac{l}{2}\left(\mu_{p}(-B^{2})+\mu_{p}(B)+\mu_{p}(-B)\right)$ (30) $\displaystyle\nu_{p}^{l}(A,B)$ $\displaystyle\geq$ $\displaystyle l\mu_{p}(A)-\frac{l}{2}\left(\mu_{p}(B^{2})+\mu_{p}(B)+\mu_{p}(-B)\right)$ (31) Proof: From (29) we have $\nu_{p}^{l}(A,B)\leq l\mu_{p}(A)+\frac{l}{2}\left(\mu_{p}(-B^{2})+{\bf E}(\mu_{p}(B\zeta))\right)$ in view of the triangular inequality of the logarithmic norm. Since $\zeta\sim N(0,1)$, we have ${\bf E}(\mu_{p}(B\zeta))={\bf E}\|\zeta\|(\mu_{p}(B)+\mu_{p}(-B))/2=\left(\mu_{p}(B)+\mu_{p}(-B)\right)/2$. Again, $\mu_{p}(A)=\mu_{p}(A-\frac{1}{2}B^{2}+B\zeta+\frac{1}{2}B^{2}-B\zeta)\leq\mu_{p}(A-\frac{1}{2}B^{2}+B\zeta)+\mu_{p}(\frac{1}{2}B^{2})+\mu_{p}(-B\zeta)$ so that $l\mu_{p}(A)-l\mu_{p}(\frac{1}{2}B^{2})-l{\bf E}\mu_{p}(-B\zeta)\leq\nu_{p}^{l}(A,B)$. Hence the inequalities. $\blacksquare$ ###### Corollary 4 The stochastic logarithmic norm can be bounded as follows. $\left\|\nu_{p}^{l}(A,B)\right\|\leq l\left\|\mu_{p}(A-\frac{1}{2}B^{2})\right\|+l{\bf E}\left\|\mu_{p}(B\zeta)\right\|\leq l\left\\|A-\frac{1}{2}B^{2}\right\\|_{p}+l\left\\|B\right\\|_{p}$ (32) Proof: By Theorem 10 we have $\left\|\nu_{p}^{l}(A,B)\right\|=l\left\|{\bf E}\left(\mu_{p}\left(A-\frac{1}{2}B^{2}+B\zeta\right)\right)\right\|$. Then, by Jensen inequality one obtains, a.s., $l\left\|{\bf E}\left(\mu_{p}\left(A-\frac{1}{2}B^{2}+B\zeta\right)\right)\right\|\leq l{\bf E}\left\|\left(\mu_{p}\left(A-\frac{1}{2}B^{2}+B\zeta\right)\right)\right\|.$ From the triangular inequality we have $\left\|\mu_{p}\left(A-\frac{1}{2}B^{2}+B\zeta\right)\right\|\leq\left\|\mu_{p}\left(A-\frac{1}{2}B^{2}\right)\right\|+\left\|\mu_{p}(B\zeta)\right\|$ and the bound property of the logarithmic norm leads to $\left\|\nu_{p}^{l}(A,B)\right\|\leq l\left\\|A-\frac{1}{2}B^{2}\right\\|_{p}+l{\bf E}\left\\|B\zeta\right\\|_{p}\leq l\left\\|A-\frac{1}{2}B^{2}\right\\|_{p}+l\left\\|B\right\\|_{p}.\blacksquare$ ## 11 Examples ### 11.1 Stabilization of an inverted pendulum It is well known [18, 14] that a vertical inverted pendulum can be stabilized around a mean vertical position by application of a suitable highly oscillatory excitation in the form of an appropriate noise. We compute the stochastic logarithmic norm of such a system to show how an appropriate noise stabilizes the system. The equation of the inverted pendulum may be written as $\displaystyle d\theta$ $\displaystyle=$ $\displaystyle v~{}dt+\epsilon v~{}dW_{t},~{}1\gg\epsilon>0$ (33) $\displaystyle dv$ $\displaystyle=$ $\displaystyle\frac{g}{l}\theta~{}dt+b\theta~{}dW_{t},$ (34) where $g$ is acceleration due to gravity, $l>0$ is the effective length of the pendulum and $\theta$ is the small angular displacement from the mean verticalposition, i.e, $\theta=0$, so that $A=\pmatrix{0&1\cr\frac{g}{l}&0}$ and $B=\pmatrix{0&\epsilon\cr b&0}$. It may be noted that $\mu_{2}(A)=\frac{1}{2}+\frac{g}{2l}$ and $\Re\lambda_{\max}(A)=\frac{g}{l}>0$. Thus the system without the Wiener process excitation is unstable. The stochastic logarithmic norm of the Wiener process excited system may be computed as $\nu_{2}^{2}(A,B)={\bf E}\left(\max\left\\{1+\frac{g}{l}+(b+\epsilon)\zeta-b\epsilon,-\left(1+\frac{g}{l}+(b+\epsilon)\zeta+b\epsilon\right)\right\\}\right),$ where $\zeta\sim N(0,1)$. For stabilizing the pendulum around the mean position $\theta=0$ we require $\nu_{2}^{2}(A,B)\leq 0$ and get the condition that $b\geq\frac{1}{\epsilon}\left(1+\frac{g}{l}\right)$ (35) in which $\epsilon/(2(1+g/l))$ can be interpreted physically as the amplitude of a very wide band vertical excitation (as an approximation to Wiener process) at the base of the pendulum. Since a Karhunen-Loeve expansion of the Wiener process [9] contains the high frequency terms, this also shows how the above result is consistent with the result (in [14]) that a small amplitude highly oscillatory wide band vertical excitation stabilizes a vertical inverted pendulum. ### 11.2 Nonnormality From [7] we take this linear SDE in the form of (4) in which $A:=\pmatrix{-1&b\cr 0&-1},B:=\pmatrix{0&\sigma\cr-\sigma&0},b\in{\mathbb{R}}$. Obviously the system without any Wiener process excitation, i.e, the deterministic ODE is asymptotically stable but $\mu_{2}(A)=\max\\{\frac{b}{2}-1,-\left(\frac{b}{2}+1\right)\\}$ due to non- normality of $A$ and the ODE system tends to be numerically unstable in its transient behavior when $\|b\|>2$. Formulating the system as a multiplicative noise SDE we compute the stochastic logarithmic norm directly from (29) in the mean square and using the $2$-norm as in [7]: $\nu_{2}^{2}=\max\\{\sigma^{2}-2\pm b\\}$ whence it is required that $\sigma^{2}\leq\min\\{2\mp b\\}$ so that $\nu_{2}^{2}\leq 0$ for the mean square stability of the SDE. If $\sigma\in{\mathbb{R}}$ and $\|b\|>2$, then it is not possible to numerically stabilize the SDE in the mean square by choosing an appropriate $\sigma$. For $\|b\|>2$ and $\sigma\in{\mathbb{C}}$, $\sigma^{2}\leq 2-\|b\|$ would be sufficient for exploiting the noise towards numerically stabilizing the SDE system in the mean square. If $\sigma,b\in{\mathbb{R}}$ and $\sigma:=b^{-\frac{1}{4}}$, then the SDE system is stochastically stable when $1\geq b\geq\frac{3-\sqrt{5}}{2}$. ### 11.3 Numerical examples ###### Example 1 We consider * $(a)$ $A=\left(\matrix{-100&0\cr 0&-200}\right),B=\left(\matrix{5&0\cr 0&6}\right).$ * $(b)$ $A=\left(\matrix{-100&0\cr 200&-200}\right),B=\left(\matrix{5&2\cr 0&6}\right).$ * $(c)$ $A=\left(\matrix{-100&20\cr 0&-200}\right),B=\left(\matrix{5&2\cr 0&6}\right).$ * $(d)$ $A=\left(\matrix{-100+20i&0\cr 2&-200+i}\right),B=\left(\matrix{5+i&0\cr 2i&-6-10i}\right).$ * $(e)$ $A=\left(\matrix{-100&20\cr 7&-200}\right),B=\left(\matrix{5&2\cr 4&6}\right).$ * $(f)$ $A=-100,B=10$ ([16]). * $(g)$ $A:=\left(\matrix{A_{1}&A_{12}\cr 0&A_{2}}\right);B:=\left(\matrix{B_{1}&B_{12}\cr 0&B_{2}}\right)$ where $A_{1}:=\left(\matrix{0.1&4&20\cr 0&0.1&5\cr 0&0&0.1}\right),\\\ A_{2}:=\left(\matrix{-0.2&3&100\cr 0&-0.2&50\cr 0&0&-0.2}\right),B_{1}:=\left(\matrix{2&30&10\cr 0&2&50\cr 0&0&2}\right),B_{2}:=\left(\matrix{4&6&20\cr 0&4&40\cr 0&0&4}\right),$ $A_{12}=\left(\matrix{2.2857\times 10^{-2}&-2.3547\times 10^{-2}&-6.8279\times 10^{-2}\cr 9.3914\times 10^{-2}&-9.6719\times 10^{-2}&-2.8049\times 10^{-1}\cr 2.8585\times 10^{-1}&-2.9443\times 10^{-1}&-8.5382\times 10^{-1}}\right),$ and $B_{12}=\left(\matrix{1.2606\times 10^{-1}&-4.6007\times 10^{-1}&7.0963\times 10^{-3}\cr 1.8156\times 10^{-1}&-6.6259\times 10^{-1}&1.0235\times 10^{-2}\cr 1.4481\times 10^{-1}&-5.2845\times 10^{-1}&8.1625\times 10^{-3}}\right).$ * $(h)$ $A=100\times C_{100\times 100},B=100\times D_{100\times 100};~{}C_{ij},D_{ij}\sim U(0,1)$. * $(i)$ $\left(\matrix{-100&0\cr 0&-1}\right).$ $B=\left(\matrix{0&2\cr 2&0}\right)$ [17] Example \| Lbound \| $\nu_{2}^{2}(A,B)$ \| Ubound ---\|---\|---\|--- $(a)$ \| $-1.1239\times 10^{2}$ \| $-1.0470\times 10^{2}$ \| $-4.0393\times 10^{1}$ $(b)$ \| $-1.1919\times 10^{2}$ \| $-1.1468\times 10^{2}$ \| $-3.1393\times 10^{1}$ $(c)$ \| $-2.4082\times 10^{2}$ \| $-2.2415\times 10^{2}$ \| $-1.5302\times 10^{2}$ $(d)$ \| $-2.2490\times 10^{2}$ \| $-2.2354\times 10^{2}$ \| $-5.9075\times 10^{1}$ $(e)$ \| $-2.6837\times 10^{2}$ \| $-2.3232\times 10^{2}$ \| $-1.21915\times 10^{2}$ $(f)$ \| $-3.0000\times 10^{2}$ \| $-3.0026\times 10^{2}$ \| $-1.0000\times 10^{2}$ $(g)$ \| $-9.1852\times 10^{2}$ \| $9.2453\times 10^{2}$ \| $4.8398\times 10^{3}$ $(h)$ \| $-2.5191\times 10^{7}$ \| $1.2369\times 10^{5}$ \| $2.5330\times 10^{7}$ $(i)$ \| $-6.0000$ \| $-5.91409$ \| $-2.0000$ Table 1: The table gives values $\nu_{2}^{2}(A,B)$ and compares them with the estimates $Ubound$ and $Lbound$. In Table 1 Ubound is computed using the right hand side in (30) with $p=2,~{}l=2$ and Lbound is computed using the right hand side in (31). The stochastic logarithmic norm is computed using (29). ## 12 Extension to Multiple Noise Channels The definition of the stochastic logarithmic norm can be extended to the vector SDE with multiple channels of multiplicative noise (i.e., with multi- dimensional Wiener process). Considering the Itô-Taylor strong order 1.0 (Mil’stein) scheme for the SDE $\displaystyle dX_{t}=AX_{t}dt+\sum_{j=1}^{m}B^{(j)}X_{t}dW_{t}^{(j)},$ (36) where each $W^{(j)}$ is an independent component Wiener process, we can define the following. ###### Definition 3 The stochastic logarithmic norm for a tuple of $m+1$ square matrices of same dimensions, $(A,B^{(1)},B^{(2)},\cdots,B^{(m)})$, i.e., $\left(A,B^{(1:m)}\right)$, which are the (linearized) drift and diffusion coefficients of a (non-linear) vector SDE with $m$ channels of multiplicative noise, is defined as $\nu_{p}^{l}\left(A,B^{(1:m)}\right)=\lim_{h\to 0^{+}}\frac{{\bf E}\\|I+hA+{\displaystyle{\sum_{j=1}^{m}B^{(j)}\Delta W^{(j)}+\sum_{i=1}^{m}\sum_{j=1}^{m}B^{(i)}B^{(j)}}}\int_{0}^{h}\int_{0}^{s}dW_{u}^{(i)}dW_{s}^{(j)}\\|^{l}_{p}-1}{h},$ where the limit is taken in the sense of the existence of the generalized derivative of the Wiener process and $\\|A\\|_{p}$ and each $\\|B^{(i)}\\|_{p}$ are assumed to be finite. For $p=2,l\geq 2$ it is easy to obtain the following estimate after proceeding as in the estimate in (16). $\displaystyle\nu_{2}^{l}(A,B^{(1:m)})$ $\displaystyle\leq$ $\displaystyle l\mu_{2}(A)+\frac{l}{2}\sum_{j=1}^{m}\\|B^{(j)}\\|_{2}^{2}$ (37) $\displaystyle+\frac{l}{2}\sum_{j=1}^{m}(\mu_{2}(B^{(j)})+\mu_{2}(-B^{(j)}))+\frac{l(l-2)}{2}\sum_{i=1}^{m}(\mu_{2}(B^{(i)}))^{2}$ In general, for any $p$, it is possible to estimate $\nu_{p}^{l}(0,B^{(1:m)})$ in the sense of the existence of a generalized derivative of the Wiener process such that $dW^{(i)}=\zeta^{(i)}dt,~{}\zeta^{(i)}\sim N(0,1),~{}{\bf E}(\zeta^{(i)}\zeta^{(j)})=0$ for $i\neq j$: $\nu_{p}^{l}(0,B^{(1:m)})=\lim_{h\to 0^{+}}\frac{{\bf E}\\|I+\sum_{j=1}^{m}B^{(j)}\Delta W^{(j)}+{\displaystyle{\sum_{i=1}^{m}\sum_{k=1}^{m}}}B^{(i)}B^{(k)}\int_{0}^{h}\int_{0}^{s}dW^{(i)}_{u}dW^{(k)}_{s}\\|_{p}^{l}-1}{h}$ $\leq\lim_{h\to 0^{+}}\bigg{(}\frac{\\|I-(2h){\displaystyle{\sum_{i=1}^{m}}}\frac{1}{2}{B^{(i)}}^{2}\\|_{p}^{l}-1}{(2h)}+$ $\frac{{\bf E}\\|I+2\Delta W{\displaystyle{\sum_{i=1}^{m}B^{(i)}+(\Delta W)^{2}\sum_{i=1}^{m}{B^{(i)}}^{2}+2\sum_{i=1}^{m}\sum_{j=1,~{}i\neq j}^{m}}}B^{(i)}B^{(j)}\int_{0}^{h}\int_{0}^{s}dW_{u}^{(i)}dW_{s}^{(j)}\\|^{l}_{p}-1}{(2h)}\bigg{)}$ $\leq\frac{l}{2}\mu_{p}(-\sum_{i=1}^{m}B^{(i)})+l\sum_{i=1}^{m}\\|B^{(i)}\\|_{p}+\frac{l}{2}\sum_{i=1}^{m}\\|B^{(i)}\\|^{2}_{p}+\frac{l}{\sqrt{2}}\sum_{i=1}^{m}\sum_{j=1,~{}i\neq j}^{m}\\|B^{(i)}B^{(j)}\\|_{p}.$ Then we can upper bound the stochastic logarithmic norm as $\displaystyle\nu_{p}^{l}(A,B^{(1:m)})$ $\displaystyle\leq$ $\displaystyle l\mu_{p}(A)-\frac{l}{2}\mu_{p}(\sum_{i=1}^{m}B^{(i)})+l\sum_{i=1}^{m}\\|B^{(i)}\\|_{p}+\frac{l}{2}\sum_{i=1}^{m}\\|B^{(i)}\\|^{2}_{p}+$ (38) $\displaystyle\frac{l}{\sqrt{2}}\sum_{i=1}^{m}\sum_{j=1,~{}i\neq j}^{m}\\|B^{(i)}B^{(j)}\\|_{p}.$ Similar to (29) we can compute the stochastic logarithmic norm for the multi- channel case with $\displaystyle\nu_{p}^{l}(A,B^{(1:m)})=l{\bf E}\left(\mu_{p}\left(A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}+\sum_{i=1}^{m}B^{(i)}\zeta^{(i)}\right)\right),$ (39) and bound it as in (32) as $\displaystyle\left\|\nu_{p}^{l}(A,B^{(1:m)}\right\|\leq l\left\\|A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}\right\\|_{p}+l\sum_{i=1}^{m}\left\\|B^{(i)}\right\\|_{p}.$ (40) From (39) and as in Corollary 3, the stochastic logarithmic norm for the multiplicative multiple channel noise can be bounded as $\displaystyle l\mu_{p}(A)-\frac{l}{2}\sum_{i=1}^{m}\left(\mu_{p}({B^{(i)}}^{2})+\mu_{p}(B^{(i)})+\mu_{p}(-B^{(i)})\right)\leq\nu_{p}^{l}(A,B^{(1:m)})$ $\displaystyle\leq l\mu_{p}(A)+\frac{l}{2}\sum_{i=1}^{m}\left(\mu_{p}(-{B^{(i)}}^{2})+\mu_{p}(B^{(i)})+\mu_{p}(-B^{(i)})\right).$ (41) ## 13 Relationship with Pseudospectrum We have mentioned in the introduction that the logarithmic norm as a bound on the pseudospectrum of the stability matrix provides an estimate of the finite time interval numerical stability of an ODE. The finite time interval numerical stability differs from the asymptotic stability in capturing the effect of nonnormality of the stability matrices and local stiffness that affect the computation of the numerical integration. In SDEs with multiplicative noise the diffusion coefficients may significantly affect this transient numerical stability of an SDE. It may be recalled that balanced methods have been designed [12] to overcome difficulties arising from stiffness in both drift and diffusion. The stochastic logarithmic norm relates to the pseudospectrum of the drift coefficients by way of diffusion coefficients acting as perturbations and thus captures the expected transient stability of the SDE. The estimate of stability based on the stochastic logarithmic norm can then be used for selecting an appropriate stiff stochastic integrator. Let $\\|-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}+\sum_{i=1}^{m}B^{(i)}\zeta^{(i)}\\|_{2}=\beta$. From the definition of pseudospectrum [15], we may write: $\displaystyle{\bf E}\max_{\lambda}\Re\lambda_{\beta}(A)$ $\displaystyle=$ $\displaystyle{\bf E}\max_{\lambda}\Re{\lambda\left(A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}+\sum_{i=1}^{m}B^{(i)}\zeta^{(i)}\right)}$ (42) $\displaystyle\leq{\bf E}\left(\mu_{2}\left(A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}+\sum_{i=1}^{m}B^{(i)}\zeta^{(i)}\right)\right)$ $\displaystyle~{}~{}=\frac{1}{2}\nu_{2}^{2}(A,B^{(1:m)})$ using (39). For small noise with $1\gg\beta>0$, obviously, the stochastic logarithmic norm gives an upper bound estimate of the mean transient numerical stability behavior of the deterministic ODE $dX_{t}=AX_{t}dt$. Denoting $\gamma=\\|\sum_{i=1}^{m}B^{(i)}\zeta^{(i)}\\|_{2}$, and proceeding in a similar fashion as in (42) one obtains $\displaystyle{\bf E}\max_{\lambda}\Re\lambda_{\gamma}\left(A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}\right)$ $\displaystyle\leq$ $\displaystyle{\bf E}\left(\mu_{2}\left(A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}+\sum_{i=1}^{m}B^{(i)}\zeta^{(i)}\right)\right)$ (43) $\displaystyle=\frac{1}{2}\nu_{2}^{2}(A,B^{(1:m)}).$ In the above inequalities the stochastic logarithmic norm appears as an upper bound on the mean maximum real part of the perturbed spectrum of $A-\frac{1}{2}\sum_{i=1}^{m}{B^{(i)}}^{2}$ which is significant for the stochastic asymptotic stability of the SDE (36). ## 14 Conclusion This paper has extended the classical logarithmic norm to define the stochastic logarithmic norm for the numerical stability analysis of vector Itô stochastic differential equations with multi-dimensional multiplicative noise. Incremental estimates of the stochastic logarithmic norm due to perturbations and bounds with respect to logarithmic norm of the drift and diffusion coefficient matrices have been studied. Further investigation relating the stochastic logarithmic norm to the pseudo-spectrum and stiffness, both in drift and diffusion, is needed since the stochastic logarithmic norm gives an upper bound on the mean maximum real part of the pseudospectrum of a matrix perturbed by the noise. This last property may be used in choosing stiff and balanced numerical integrators and a detailed study for various class of integrators in this respect remains to be done. Detailed study of application of the stochastic logarithmic norm to the stability analysis of non-linear SDE is necessary too. ## References * [1] R. Bhatia, Matrix Analysis, Springer-Verlag, 1997. * [2] K. Burrage, P. Burragea and T. Mitsui, Numerical solutions of stochastic differential equations – implementation and stability issues, Journal of Computational and Applied Mathematics, 125:1-2 (2000), pp. 171–182 * [3] K. Burrage, P. M. Burrage and T. Tian, Numerical methods for strong solutions of stochastic differential equations: an overview, Royal Society of London Proceedings Series A, 460:2041 (2004), pp.373–402 * [4] G. Dahlquist, Stability and the error bounds in the numerical integration of ordinary differential equations, Almqvist and Wiksells, Uppasala, 1958: Transactions of Royal Institute of Technology, Stockholm, 1959. * [5] E. Hairer and G. Wanner, Solving Ordinary Differential Equations I, II: Stiff and Differential- Algebraic Problems, Springer-Verlag,1996. * [6] D. J. Higham and L. N. Trefethen, Stiffness of ODEs, BIT, 33(1993) pp. 285. * [7] D. J. Higham and X. Mao, Nonnormality and stochastic differential equations, BIT, 46(2006), pp. 525–532. * [8] I. Higueras and B. G. Celayeta, Logarithmic norms for matrix pencils, SIAM J. Matrix Anal., 20 (1999), pp. 646–666. * [9] P. E. Klöden and E. Platen, The Numerical Solution of Stochastic Differential Equations, Springer-Verlag, Berlin, 1992. * [10] S. M. Lozinskii, Error estimates for the numerical integration of ordinary differential equations, part I, Izv. Vyss. Uceb. Zaved Mathematika, 6(1958), pp. 52–90. * [11] G. N. Mil’stein,Approximate integration of stochastic differential equations, Theory Probab. Appl., 19(1974), pp. 557–562. * [12] G. N. Mil’stein and M. V. Tretyakov, Stochastic Numerics for Mathematical Physics, Springer-Verlag, New York, 2004. * [13] S. Raha and L. R. Petzold, Constraint partitioning for stability in path-constrained dynamic optimization problems, SIAM J. Sci. Comput., 22 (2001), pp. 2051–2074. * [14] R. Sharp, Y.-H. Tsai and B. Engquist , Multiple Time Scale Numerical Methods for the Inverted Pendulum Problem, Lecture Notes in Computational Science and Engineering, Springer-Verlag, 44 (2005), pp. 241-262. * [15] L. N. Trefethen and M. Embree, Spectra and Pseudospectra: The Behavior of Nonnormal Matrices and Operators, Princeton University Press, 2005. * [16] Y. Saito and T. Mitsui, Stability analysis of numerical schemes for stochastic differential equations, SIAM J. Numer. Anal., 33(1996), pp. 2254–2267. * [17] Y. Saito and T. Mitsui, Mean square stability of numerical schemes for stochastic differential systems, Math. Sci., SIS/GSHI.,2(2002), Nagoya Univ. * [18] J. M. Sanz-Serna, Modulated Fourier expansions and heterogeneous multiscale methods, IMA Journal of Numerical Analysis, doi:10.1093/imanum/drn031 (2008) * [19] T. Strôm, On logarithmic norms, SIAM J. Numer. Anal., 12(1975), pp. 741–753. * [20] G. Söderlind, The logarithmic norm history and modern theory, BIT, 46 (2006), pp. 631–652.	arxiv-papers	2008-08-30T14:07:41	2024-09-04T02:48:57.557796	{ "license": "Public Domain", "authors": "Sk. Safique Ahmad and Nagalinga Rajan and Soumyendu Raha", "submitter": "Safique Ahmad Sk", "url": "https://arxiv.org/abs/0809.0062" }
0809.0068	# The $\ell$-adic Dualizing Complex on an Excellent Surface with Rational Singularities Ting Li Department of Mathematics Sichuan University Chengdu 610064 P. R. China moduli@live.cn ###### Abstract. In this article, we show that if $X$ is an excellent surface with rational singularities, the constant sheaf $\mathbb{Q}_{\ell}$ is a dualizing complex. In coefficient $\mathbb{Z}_{\ell}$, we also prove that the obstruction for $\mathbb{Z}_{\ell}$ to become a dualizing complex lying on the divisor class groups at singular points. As applications, we study the perverse sheaves and the weights of $\ell$-adic cohomology groups on such surfaces. ###### Key words and phrases: Poincaré Duality, rational singularity, $\ell$-adic cohomology ††Mathematics Subject Classification 2000: 14F20, 14J17††This work is supported by NSFC(10626036) ## Introduction In [9, I], the dualizing complexes in étale cohomology was considered by Grothendieck. O. Gabber made a breakthrough on this fields. He prove that every excellent schemes admits a dualizing complex; and on a regular excellent scheme, the constant sheaf $\mathbb{Z}_{\ell}$ is a dualizing complex. See [12] for the detail. In this paper, we study the properties of dualizing complexes on surfaces with rational singularities. In [11], J. Lipman prove that a two-dimensional normal local ring $R$ has a rational singularity if and only if $R$ has a finite divisor class group. This makes me think that rational singularities only affect the torsion part of the étale cohomology with coefficient $\mathbb{Z}_{\ell}$, but leave the free part invariant. As an evidence, we prove that on an excellent surface $X$ with at most rational singularities, the potential dualizing complex $\mathscr{K}_{X}$ (determinated by the dimension function $x\mapsto 2-\dim\mathcal{O}_{X,x}$) concentrates on the entries $-2$ and $-4$; in detail, $\mathrm{H}^{-4}(\mathscr{K}_{X})=\mathbb{Z}_{\ell}(2)$ and $\mathrm{H}^{-2}(\mathscr{K}_{X})(-1)$ is the $\ell$-torsion part of the divisor class groups at singular points (see Theorem 3.2). In particular, $\mathbb{Q}_{\ell}$ is a dualizing complex of such surface in coefficient $\mathbb{Q}_{\ell}$. The paper is organized as follows. In §1, we briefly review the theory of “potential dualizing complex” in [12]. Because the dualizing complexes on a scheme are not unique, they vary by tensor products with invertible complexes. So we prefer the _potential dualizing complex_ which is a dualizing complex unique determinated by the dimension function. Basing on this theory, we reiterate the étale homology in [10]. Since lacking of suitable references related to étale homology both in adic coefficients and over arbitrary base schemes, it is necessary to spend a little more space on rewriting it. In §2, we calculate the étale homology on curves. In §3, we prove the main results. First we calculate the étale homology on arbitrary surfaces. Basing on it, we obtain the main results about the dualizing complexes on an excellent surface with rational singularities; and prove the Poincaré Duality for such surfaces. In §4, we give two applications of above theory. First we study perverse sheaves on a surface with rational singularities. Basing on it, we prove that the cohomologies of smooth sheaves of punctually pure weights on such surfaces, are also punctually pure of weights. ## 1\. Review of dualizing sheaves and étale homology Let $\ell$ be a prime number, $\Lambda$ the integral closure of $\mathbb{Z}_{\ell}$ in a finite extension field of $\mathbb{Q}_{\ell}$. We consider the following conditions related to a scheme $X$: * (†) $X$ is Noetherian, excellent, of finite Krull dimension, $\ell$ is invertible on $X$ and $\mathrm{cd}_{\ell}(X)<\infty$. From the Gabber’s finitenes theorem for étale cohomology in [7], we may construct an adic formalism $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$ together with Grothendieck’s six operations on schemes satisfying (†). See also [4] for details. Moreover if $X$ is a scheme satisfying (†), then any scheme of finite type over $X$ satisfies (†). Next, we introduces the “potential dualizing complex” defined in [12]. ###### Definition 1.1 ([12], 2.2). Let $X$ be a scheme satisfying (†), $\delta\colon X\to\mathbb{Z}$ a dimension function. A _potential dualizing complex_ (in coefficient $\Lambda$) on $X$ consisting of 1. (1) an object $\mathscr{K}$ in $\mathbf{D}_{\mathrm{c}}^{\textbf{+}}(X_{\textup{\'{e}t}},\Lambda)$, 2. (2) an isomorphism $\mathbf{R}\varGamma_{x}(\mathscr{K})\xrightarrow{\,\sim\,}\Lambda\bigl{(}\delta(x)\bigr{)}\bigl{[}2\delta(x)\bigr{]}$ in $\mathbf{D}_{\mathrm{c}}^{\textbf{+}}(x_{\textup{\'{e}t}},\Lambda)$ for every point $x$ on $X$; and these data satisfy that: for every immediate specialization $\bar{y}\to\bar{x}$ on $X$, the following diagram commutes, $\textstyle{\mathbf{R}\varGamma_{\bar{y}}(\mathscr{K})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathrm{sp}^{X}_{\bar{y}\to\bar{x}}}$$\scriptstyle{\cong}$$\textstyle{\mathbf{R}\varGamma_{\bar{x}}(\mathscr{K})(1)[2]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\cong}$$\textstyle{R\bigl{(}\delta(y)\bigr{)}\bigl{[}2\delta(y)\bigr{]}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{R\bigl{(}\delta(x)+1\bigr{)}\bigl{[}2(\delta(x)+1)\bigr{]}}$ where $\mathrm{sp}^{X}_{\bar{y}\to\bar{x}}$ is the transition morphism of codimension $1$ defined in [12, §1]. The potential dualizing complexes have the following properties. ###### Proposition 1.2 ([12], 4.1 & 5.1). Every scheme $X$ satisfying (†) equipped with a dimension function $\delta$ has a potential dualizing complex, unique up to unique isomorphism, which we denoted by $\mathscr{K}_{X,\delta}$ or simply by $\mathscr{K}_{X}$. Moreover the potential dualizing complex is a dualizing complex. ###### Proposition 1.3 ([12], 2.8). Let $X$ be a regular scheme satisfying (†). Then $\delta(x)\mathrel{\mathop{:}}=-\dim\mathcal{O}_{X,x}$ is a dimension function on $X$ and we have $\mathscr{K}_{X,\delta}=\Lambda$. ###### Proposition 1.4 ([12], 4.3). Let $f\colon X\to Y$ be a compactifiable morphism of schemes satisfying (†). Let $\delta_{Y}$ be a dimension function on $Y$ and equip $X$ with the dimension function $\delta_{X}(x)\mathrel{\mathop{:}}=\delta_{Y}\bigl{(}f(x)\bigr{)}+\operatorname{tr.d.}k(x)\big{/}k\bigl{(}f(x)\bigr{)}\,.$ Then we have $\mathscr{K}_{X,\delta_{X}}=\mathbf{R}f^{!}\mathscr{K}_{Y,\delta_{Y}}$. ###### Proposition 1.5 ([12], 4.2). Let $f\colon X\to Y$ be a regular morphism of schemes satisfying (†). Let $\delta_{Y}$ be a dimension function on $Y$ and equip $X$ with the dimension function $\delta_{X}(x)\mathrel{\mathop{:}}=\delta_{Y}(y)-\dim\mathcal{O}_{X_{y},x}\qquad\bigl{(}y\mathrel{\mathop{:}}=f(x)\bigr{)}$ Then we have $\mathscr{K}_{X,\delta_{X}}=f^{\ast}\mathscr{K}_{Y,\delta_{Y}}$. Basing on above theory, we may generalize the étale homology in [10] to arbitrary schemes satisfying (†). Since most proof are almost the same with [10], we only give these which meed special care. ###### Definition 1.6. Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$. For each $n\in\mathbb{Z}$, we define $\mathrm{H}_{n}(X,\delta,\Lambda)\mathrel{\mathop{:}}=\mathrm{H}^{-n}(X_{\textup{\'{e}t}},\mathscr{K}_{X,\delta})=\operatorname{Hom}_{\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)}\bigl{(}\Lambda_{X},\mathscr{K}_{X,\delta}[-n]\bigr{)}\,.$ We also use $\mathrm{H}_{n}(X,\delta)$ or $\mathrm{H}_{n}(X,\Lambda)$ or $\mathrm{H}_{n}(X)$ to denote $\mathrm{H}_{n}(X,\delta,\Lambda)$, if no confusion arise. Let $f\colon X\to Y$ be a compactifiable morphism of schemes satisfying (†), $\delta$ a dimension function on $Y$, and $\delta^{\prime}$ the dimension function on $X$ induced by $\delta$ as in section 1. Then we also use $\mathrm{H}_{n}(X,\delta,\Lambda)$ or $\mathrm{H}_{n}(X,\delta)$ to denote $\mathrm{H}_{n}(X,\delta^{\prime},\Lambda)$. If $f\colon X\to Y$ is a proper morphism of schemes satisfying (†), $\delta$ a dimension function on $Y$. For each $n\in\mathbb{Z}$, we define a homomorphism of $\Lambda$-modules $f_{\ast}\colon\mathrm{H}_{n}(X,\delta)\to\mathrm{H}_{n}(Y,\delta)$ as follows. For each $\alpha\in\mathrm{H}_{n}(X,\delta)$, regarding $\alpha\colon\Lambda_{X}\to\mathscr{K}_{X,\delta}[-n]$ as a morphism in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$, then $f_{\ast}(\alpha)$ is defined to be the composition $\Lambda_{Y}\to\mathbf{R}f_{\ast}\Lambda_{X}\xrightarrow{\mathbf{R}f_{\ast}(\alpha)}\mathbf{R}f_{\ast}\mathscr{K}_{X,\delta}[-n]\xrightarrow{\,\sim\,}\mathbf{R}f_{\ast}\circ f^{\ast}\mathscr{K}_{Y,\delta}[-n]\to\mathscr{K}_{Y,\delta}[-n]\,,$ where the first and the last morphisms are induced by the adjunctions. It is easy to verify that if $f\colon X\to Y$ and $g\colon Y\to Z$ are two proper morphisms, then $(g\circ f)_{\ast}=g_{\ast}\circ f_{\ast}$. If $Y$ is the spectrum of a separably closed field, as $\mathrm{H}_{0}(Y)=\Lambda$, we may write $\deg\mathrel{\mathop{:}}=f_{\ast}\colon\mathrm{H}_{0}(X)\to\Lambda\,.$ ###### Proposition 1.7. Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$, $Y$ a closed subscheme of $X$ and $U\mathrel{\mathop{:}}=X\setminus Y$. Then there is a long exact sequence of $\Lambda$-modules: $\cdots\mathrm{H}_{n+1}(U,\delta)\to\mathrm{H}_{n}(Y,\delta)\to\mathrm{H}_{n}(X,\delta)\to\mathrm{H}_{n}(U,\delta)\to\mathrm{H}_{n-1}(Y,\delta)\to\cdots\,.$ ###### Proposition 1.8 (Mayer-Vietoris Sequence). Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$, $X_{1}$ and $X_{2}$ two closed subschemes of $X$ such that $X=X_{1}\cup X_{2}$ (as sets). Then we have a long exact sequence $\cdots\to\mathrm{H}_{n}(X_{1}\cap X_{2},\,\delta)\to\mathrm{H}_{n}(X_{1},\delta)\oplus\mathrm{H}_{n}(X_{2},\delta)\to\mathrm{H}_{n}(X,\delta)\to\mathrm{H}_{n-1}(X_{1}\cap X_{2},\,\delta)\to\cdots\,.$ ###### Notation 1.9. Let $X$ be a Noetherian scheme, $\delta$ a dimension function on $X$. Then we define $\dim_{\delta}(X)\mathrel{\mathop{:}}=\sup_{x\in X}\delta(x)\,.$ If $\xi_{1},\xi_{2},\ldots,\xi_{r}$ are all generic points of irreducible components of $X$, then $\dim_{\delta}(X)=\max_{1\leqslant i\leqslant r}\delta(\xi_{i})<+\infty\,.$ ###### Proposition 1.10 (Vanishing). Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$. Then $\mathrm{H}_{n}(X,\delta)=0$ for all $n>2\dim_{\delta}(X)$. ###### Corollary 1.11. Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$, $Y$ a closed subscheme of $X$, $X^{\prime}\mathrel{\mathop{:}}=X\setminus Y$. Then for each integer $n>2\dim_{\delta}(Y)+1$, there is a canonical isomorphism $\mathrm{H}_{n}(X,\delta)\xrightarrow{\,\sim\,}\mathrm{H}_{n}(X^{\prime},\delta)$ of $\Lambda$-modules. ###### Proof. We have only to apply section 1. ∎ We define the pull-back maps as follows. Let $Y$ be a scheme satisfying (†) equipped with a dimension function $\delta$, $f\colon X\to Y$ is a compactifiable flat morphism of relative dimension $d$. For each $n\in\mathbb{Z}$, we define a homomorphism of $\Lambda$-modules $f^{\ast}\colon\mathrm{H}_{n}(Y,\delta)\to\mathrm{H}_{n+2d}(X,\delta)(-d)$ as follows. For each $\beta\in\mathrm{H}_{n}(Y,\delta)$, $f^{\ast}(\beta)$ is defined to be the composition $\Lambda_{X}\xrightarrow{\mathrm{t}_{f}}\mathbf{R}f^{!}\Lambda_{X}[-2d](-d)\xrightarrow{\mathbf{R}f^{!}(\beta)}\mathbf{R}f^{!}\mathscr{K}_{Y}[-n-2d](-d)\xrightarrow{\,\sim\,}\mathscr{K}_{X}[-n-2d](-d)\,,$ where $\mathrm{t}_{f}$ is the morphism dual to the trace morphism $\mathrm{Tr}_{f}$ (see [8, XVIII (3.2.1.2)]). It is easy to verify that if $f\colon X\to Y$ and $g\colon Y\to Z$ be two compactifiable flat morphisms of relative dimension $d$ and $e$ respectively, then $(g\circ f)^{\ast}=f^{\ast}\circ g^{\ast}\colon\mathrm{H}_{n}(Z)\to\mathrm{H}_{n+2d+2e}(X)(-d-e)\,.$ Moreover the maps $f_{\ast}$ and $f^{\ast}$ commute in Cartesian squares. Now we define the cycle map for étale homology. ###### Notation 1.12. Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$, $d\mathrel{\mathop{:}}=\dim_{\delta}(X)$. 1. (1) First we assume that $X$ is regular and integral. Then we use $\mathrm{t}_{X}\colon\Lambda\xrightarrow{\,\sim\,}\mathscr{K}_{X,\delta}(-d)[-2d]$ to denote the canonical morphism defined by section 1, and use $\mathrm{cl}(X)$ to denote the element in $\mathrm{H}_{2d}(X,\delta)(-d)$ corresponding to $\mathrm{t}_{X}$. 2. (2) Second we only assume that $X$ is integral. As $X$ is excellent, $U\mathrel{\mathop{:}}=X_{\mathrm{reg}}$ is an open dense subset of $X$. Obviously $\dim_{\delta}(X\setminus U)<d$. By section 1, there is a canonical isomorphism $\mathrm{H}_{2d}(X,\delta)(-d)\xrightarrow{\,\sim\,}\mathrm{H}_{2d}(U,\delta)(-d)\,.$ Let $\mathrm{cl}(X)$ be the inverse image of $\mathrm{cl}(U)$ under above isomorphism. 3. (3) In general case, we let $X_{1},X_{2},\ldots,X_{r}$ be all irreducible components of $X$ with $\dim_{\delta}(X_{i})=d$. For each $i$, regard $X_{i}$ as a reduced subscheme of $X$, let $\iota_{i}\colon X_{i}\hookrightarrow X$ be the inclusion and let $\xi_{i}$ be the generic point of $X_{i}$. Then we define $\mathrm{cl}(X)\mathrel{\mathop{:}}=\sum^{r}_{i=1}\mathrm{length}(\mathcal{O}_{X,\xi_{i}})\cdot\iota_{i,\ast}\bigl{(}\mathrm{cl}(X_{i})\bigr{)}\in\mathrm{H}_{2d}(X,\delta)(-d)\,.$ We also use $\mathrm{t}_{X}\colon\Lambda\xrightarrow{\,\sim\,}\mathscr{K}_{X,\delta}(-d)[-2d]$ to denote the morphism in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X,\Lambda)$ corresponding to $\mathrm{cl}(X)$. ###### Notation 1.13. Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$, $Y$ a closed subscheme of $X$, $d\mathrel{\mathop{:}}=\dim_{\delta}(Y)$. Then we define $\mathrm{cl}_{X}(Y)\mathrel{\mathop{:}}=\iota_{\ast}\bigl{(}\mathrm{cl}(Y)\bigr{)}\in\mathrm{H}_{2d}(X,\delta)(-d)\,,$ where $\iota\colon Y\hookrightarrow X$ is the inclusion. Next, we study the étale homology under birational morphisms. First we need two lemmas. ###### Lemma 1.14. Let $X$ be a scheme satisfying (†), $Y$ a closed subscheme of $X$, $U\mathrel{\mathop{:}}=X\setminus Y$, $i\colon Y\hookrightarrow X$ and $j\colon U\hookrightarrow X$ the inclusions. Then for each object $\mathscr{F}$ in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$, there are two distinguished triangles in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$: 1. (1) $j_{!}j^{\ast}\mathscr{F}\to\mathscr{F}\to i_{\ast}i^{\ast}\mathscr{F}\to j_{!}j^{\ast}\mathscr{F}[1]$ , 2. (2) $i_{\ast}\mathbf{R}i^{!}\mathscr{F}\to\mathscr{F}\to\mathbf{R}j_{\ast}j^{\ast}\mathscr{F}\to i_{\ast}\mathbf{R}i^{!}\mathscr{F}[1]$ . ###### Lemma 1.15. Let $X$ be a scheme satisfying (†), $Y$ a closed subscheme of $X$, $U\mathrel{\mathop{:}}=X\setminus Y$, $i\colon Y\hookrightarrow X$ and $j\colon U\hookrightarrow X$ the inclusions, $\mathscr{F}^{\prime}\xrightarrow{\varphi}\mathscr{F}\xrightarrow{\psi}\mathscr{F}^{\prime\prime}$ a sequence of objects in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$. Assume that one of the following two conditions holds: 1. (1) $j^{\ast}(\varphi)\colon j^{\ast}\mathscr{F}^{\prime}\xrightarrow{\,\sim\,}j^{\ast}\mathscr{F}$ is an isomorphism, $j^{\ast}\mathscr{F}^{\prime\prime}=0$, and for all $n\in\mathbb{Z}$, $0\to\mathrm{H}^{n}(i^{\ast}\mathscr{F}^{\prime})\to\mathrm{H}^{n}(i^{\ast}\mathscr{F})\to\mathrm{H}^{n}(i^{\ast}\mathscr{F}^{\prime\prime})\to 0$ is a short exact sequence; 2. (2) $j^{\ast}(\psi)\colon j^{\ast}\mathscr{F}\xrightarrow{\,\sim\,}j^{\ast}\mathscr{F}^{\prime\prime}$ is an isomorphism, $j^{\ast}\mathscr{F}^{\prime}=0$, and for all $n\in\mathbb{Z}$, $0\to\mathrm{H}^{n}(\mathbf{R}i^{!}\mathscr{F}^{\prime})\to\mathrm{H}^{n}(\mathbf{R}i^{!}\mathscr{F})\to\mathrm{H}^{n}(\mathbf{R}i^{!}\mathscr{F}^{\prime\prime})\to 0$ is a short exact sequence. Then there exists a morphism $\delta\colon\mathscr{F}^{\prime\prime}\to\mathscr{F}^{\prime}[1]$ which makes $\mathscr{F}^{\prime}\xrightarrow{\varphi}\mathscr{F}\xrightarrow{\psi}\mathscr{F}^{\prime\prime}\xrightarrow{\delta}\mathscr{F}^{\prime}[1]$ a distinguished triangle in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$. ###### Proof. The morphism $\varphi$ extends to a distinguished triangle $\mathscr{F}^{\prime}\xrightarrow{\varphi}\mathscr{F}\xrightarrow{\phi}\mathscr{G}\xrightarrow{\rho}\mathscr{F}^{\prime}[1]$ in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$. Note that $0\to\mathscr{F}^{\prime\prime}\xrightarrow{\mathrm{id}}\mathscr{F}^{\prime\prime}\to 0[1]$ is also a distinguished triangle. Hence there exists a morphism $\alpha\colon\mathscr{G}\to\mathscr{F}^{\prime\prime}$ such that the triple $(0,\psi,\alpha)$ is a morphism of distinguished triangles, i.e., the following diagram commutes. $\textstyle{\mathscr{F}^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\varphi}$$\scriptstyle{0}$$\textstyle{\mathscr{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\phi}$$\scriptstyle{\psi}$$\textstyle{\mathscr{G}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\rho}$$\scriptstyle{\alpha}$$\textstyle{\mathscr{F}^{\prime}[1]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{0}$$\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathscr{F}^{\prime\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathrm{id}}$$\textstyle{\mathscr{F}^{\prime\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{0[1]}$ Now if (1) holds, then both $j^{\ast}(\alpha)$ and $i^{\ast}(\alpha)$ are isomorphic; and if (2) holds, then both $j^{\ast}(\alpha)$ and $\mathbf{R}i^{!}(\alpha)$ are isomorphic. In either case, by section 1, the morphism $\alpha\colon\mathscr{G}\xrightarrow{\,\sim\,}\mathscr{F}^{\prime\prime}$ is an isomorphism. Therefore we may let $\delta\mathrel{\mathop{:}}=\rho\circ\alpha^{-1}$. ∎ Basing on above lemma, we have the following two propositions. ###### Proposition 1.16. Let $\textstyle{Y^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{i^{\prime}}$$\scriptstyle{q}$$\textstyle{\square}$$\textstyle{X^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{p}$$\textstyle{Y\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{i}$$\textstyle{X}$ be a Cartesian square of schemes satisfying (†), $r\mathrel{\mathop{:}}=p\circ i^{\prime}=i\circ q$. Assume that $p$ is proper, $i$ is a closed immersion; and there exists an open subset $U$ of $X$ such that $i(Y)=X\setminus U$ and $p$ induces an isomorphism $p^{-1}(U)\xrightarrow{\,\sim\,}U$. Then for each object $\mathscr{F}$ in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\Lambda)$, there are two distinguished triangles: $\displaystyle\mathscr{F}\xrightarrow{\delta_{i}\oplus\delta_{p}}i_{\ast}\circ i^{\ast}\mathscr{F}\oplus\mathbf{R}p_{\ast}\circ p^{\ast}\mathscr{F}\xrightarrow{\delta_{q}\circ\mathrm{pr}_{1}-\delta_{i^{\prime}}\circ\mathrm{pr}_{2}}\mathbf{R}r_{\ast}\circ r^{\ast}\mathscr{F}\to\mathscr{F}[1]\,,$ $\displaystyle\mathbf{R}r_{\ast}\circ\mathbf{R}r^{!}\mathscr{F}\xrightarrow{\vartheta_{q}\oplus\vartheta_{i^{\prime}}}i_{\ast}\circ\mathbf{R}i^{!}\mathscr{F}\oplus\mathbf{R}p_{\ast}\circ\mathbf{R}p^{!}\mathscr{F}\xrightarrow{-\vartheta_{i}\circ\mathrm{pr}_{1}+\vartheta_{p}\circ\mathrm{pr}_{2}}\mathscr{F}\to(\mathbf{R}r_{\ast}\circ\mathbf{R}r^{!}\mathscr{F})[1]\,,$ in $\mathbf{D}_{\mathrm{c}}^{\textbf{+}}(X_{\textup{\'{e}t}},\Lambda)$, where $\delta$ and $\vartheta$ are induced by the adjunctions. Apply the second triangle in above proposition to $\mathscr{K}_{X}$, we obtain the following proposition. ###### Proposition 1.17. Let the assumptions and the notations be as in section 1; and let $\delta$ be a dimension function on $X$. Then there is a long exact sequence of $\Lambda$-modules: $\cdots\to\mathrm{H}_{n+1}(X,\delta)\to\mathrm{H}_{n}(Y^{\prime},\delta)\xrightarrow{q_{\ast}\oplus i^{\prime}_{\ast}}\mathrm{H}_{n}(Y,\delta)\oplus\mathrm{H}_{n}(X^{\prime},\delta)\xrightarrow{-i_{\ast}\circ\mathrm{pr}_{1}+p_{\ast}\circ\mathrm{pr}_{2}}\mathrm{H}_{n}(X,\delta)\to\cdots\,.$ ## 2\. Calculation of étale homology on the exceptional divisor In this section, we calculate the étale homology of exceptional curves. ###### Proposition 2.1. Let $C$ be a $1$-equidimensional proper algebraic scheme over a separably closed field $k$. Let $C^{(1)},C^{(2)},\ldots,C^{(r)}$ be all connected components of $C$; and let $C_{1},C_{2},\ldots,C_{n}$ be all irreducible components of $C$. Assume that $\mathrm{H}^{1}(C,\mathcal{O}_{C})=0$. Then we have 1. (1) As to homology, we have 1. (a) There is a canonical isomorphism: $\mathrm{H}_{0}(C)=\bigoplus^{r}_{i=1}\mathrm{H}_{0}(C^{(i)})\xrightarrow[\,\sim\,]{\oplus\deg}\Lambda^{\oplus r}\,.$ 2. (b) $\mathrm{H}_{1}(C)=0$. 3. (c) $\mathrm{H}_{2}(C)(-1)$ is a free $\Lambda$-module with basis $\mathrm{cl}(C_{1}),\ldots,\mathrm{cl}(C_{n})$. 2. (2) As to cohomology, we have 1. (a) There is a canonical isomorphism: $\mathrm{H}^{0}(C,\Lambda)=\bigoplus^{r}_{i=1}\mathrm{H}^{0}(C^{(i)})\xrightarrow{\,\sim\,}\Lambda^{\oplus r}\,.$ 2. (b) $\mathrm{H}^{1}(C,\Lambda)=0$. 3. (c) There is a canonical isomorphism $\mathrm{H}^{2}\bigl{(}C,\Lambda(1)\bigr{)}\xrightarrow{\,\sim\,}\Lambda^{\oplus n}$ which sends $c_{1}(\mathscr{L})$ to $\bigl{(}\deg(\mathscr{L}\|_{C_{1}}),\ldots,\deg(\mathscr{L}\|_{C_{r}})\bigr{)}$ for each invertible $\mathcal{O}_{C}$-module $\mathscr{L}$. ###### Proof. (1). Let $\bar{k}$ be the algebraic closure of $k$. After replacing $k$ with $\bar{k}$ and $X$ with $(X\otimes_{k}\bar{k})_{\mathrm{red}}$, we may assume that $k$ is algebraically closed and $X$ is reduced. Now we use induction on the number $n$ of irreducible components of $C$. First consider the case $n=1$. In this case, $C$ is integral. Let $\widetilde{C}$ be the normalization of $C$. Then there is a long exact sequence $0\to\varGamma(C,\mathcal{O}_{C})\to\varGamma(C,\mathcal{O}_{\widetilde{C}})\to\varGamma(C,\mathcal{O}_{\widetilde{C}}/\mathcal{O}_{C})\to\mathrm{H}^{1}(C,\mathcal{O}_{C})\,.$ Note that $\mathrm{H}^{1}(C,\mathcal{O}_{C})=0$, $\varGamma(C,\mathcal{O}_{C})=k$ and $\varGamma(C,\mathcal{O}_{\widetilde{C}})=\varGamma(\widetilde{C},\mathcal{O}_{\widetilde{C}})=k$. Hence $\varGamma(C,\mathcal{O}_{\widetilde{C}}/\mathcal{O}_{C})=0$. Since $\mathcal{O}_{\widetilde{C}}/\mathcal{O}_{C}$ is a coherent $\mathcal{O}_{C}$-module whose support is of dimension $\leqslant 0$, we have $\mathcal{O}_{\widetilde{C}}/\mathcal{O}_{C}=0$, i.e., $C=\widetilde{C}$ is normal. So we have $C=\mathrm{P}_{k}^{1}$. Assume that (1) is valid for all integers $<n$. Let $C^{\prime}$ be an irreducible component $C$ and let $C^{\prime\prime}$ be the union of all other irreducible components. Regard $C^{\prime}$ and $C^{\prime\prime}$ as reduced closed subschemes of $C$. Then there is a short exact sequence: $0\to\mathcal{O}_{C}\xrightarrow{c\mapsto(\bar{c},\bar{c})}\mathcal{O}_{C^{\prime}}\oplus\mathcal{O}_{C^{\prime\prime}}\xrightarrow{(a,b)\mapsto\bar{a}-\bar{b}}\mathcal{O}_{C^{\prime}\cap C^{\prime\prime}}\to 0\,,$ which induces a long exact sequence: $\displaystyle 0\to\varGamma(C,\mathcal{O}_{C})\to\varGamma(C,\mathcal{O}_{C^{\prime}})\oplus\varGamma(C,\mathcal{O}_{C^{\prime\prime}})\to\varGamma(C,\mathcal{O}_{C^{\prime}\cap C^{\prime}})$ $\displaystyle\to\mathrm{H}^{1}(C,\mathcal{O}_{C})\to\mathrm{H}^{1}(C,\mathcal{O}_{C^{\prime}})\oplus\mathrm{H}^{1}(C,\mathcal{O}_{C^{\prime\prime}})\to 0\,.$ Let $r$ and $s$ be the numbers of connected components of $C$ and $C^{\prime\prime}$ respectively; and let $Z_{1},Z_{2},\ldots,Z_{m}$ be all connected components of $C^{\prime\prime}$ which intersect with $C^{\prime}$. Then $m=s-r+1$. On the other hand, we have $\dim_{k}\varGamma(C,\mathcal{O}_{C})=r,\quad\dim_{k}\varGamma(C,\mathcal{O}_{C^{\prime}})=1,\quad\dim_{k}\varGamma(C,\mathcal{O}_{C^{\prime\prime}})=s.$ As $\mathrm{H}^{1}(C,\mathcal{O}_{C})=0$, we have $\mathrm{H}^{1}(C,\mathcal{O}_{C^{\prime}})=\mathrm{H}^{1}(C,\mathcal{O}_{C^{\prime\prime}})=0$, and $\dim_{k}\varGamma(C,\mathcal{O}_{C^{\prime}\cap C^{\prime}})=s-r+1=m\,.$ This shows that for each $i$, $Z_{i}\cap C^{\prime}$ contains only one point $P_{i}$, and $C^{\prime}\cap C^{\prime\prime}=\\{P_{1},P_{2},\ldots,P_{m}\\}$. Using induction on $Z_{i}$, $C^{\prime}$ and $C^{\prime\prime}$, we obtain that (1) is valid on all $Z_{i}$, $C^{\prime}$ and $C^{\prime\prime}$. Applying section 1, we obtain an isomorphism $\mathrm{H}_{2}(C^{\prime})\oplus\mathrm{H}_{2}(C^{\prime\prime})\xrightarrow{\,\sim\,}\mathrm{H}_{2}(C)$ and an exact sequence $0\to\mathrm{H}_{1}(C)\to\mathrm{H}_{0}(C^{\prime}\cap C^{\prime\prime})\xrightarrow{l}\mathrm{H}_{0}(C^{\prime})\oplus\mathrm{H}_{0}(C^{\prime\prime})\to\mathrm{H}_{0}(C)\to 0\,.$ Note that the map $l$ has an inverse which is defined by the composition $\mathrm{H}_{0}(C^{\prime})\oplus\mathrm{H}_{0}(C^{\prime\prime})\xrightarrow{\text{pr}}\mathrm{H}_{0}(C^{\prime\prime})\xrightarrow{\text{pr}}\bigoplus^{m}_{i=1}\mathrm{H}_{0}(Z_{i})\xrightarrow[\,\sim\,]{\oplus\varphi_{i}}\bigoplus^{m}_{i=1}\mathrm{H}_{0}(P_{i})=\mathrm{H}_{0}(C^{\prime}\cap C^{\prime\prime})\,,$ where the isomorphism $\varphi_{i}$, which maps $\mathrm{cl}_{Z_{i}}(P_{i})$ to $\mathrm{cl}(P_{i})$, is obtained by applying (a) on $Z_{i}$. Therefore (1) are valid for $C$. (2) Let $\pi\colon C\to\operatorname{Spec}k$ be the structural morphism. Note that $\mathrm{H}^{q}(C,\Lambda)=\mathrm{H}^{q}(\mathbf{R}\pi_{\ast}\Lambda_{C})=\mathrm{H}^{q}\bigl{(}\mathscr{D}(\mathbf{R}\pi_{\ast}\mathscr{D}(\Lambda_{C}))\bigr{)}=\mathrm{H}^{q}\bigl{(}\mathscr{D}(\mathbf{R}\pi_{\ast}\mathscr{K}_{C})\bigr{)}\,,$ where $\mathscr{D}\mathrel{\mathop{:}}=\mathscr{H}om(\makebox[8.00003pt][c]{\raisebox{0.86108pt}{\Tiny\textbullet}},\mathscr{K})$ is the dualizing functor. By (1), we have $\mathrm{H}^{q}(\mathbf{R}\pi_{\ast}\mathscr{K}_{C})=\mathrm{H}_{-q}(C)=\begin{cases}\Lambda^{\oplus r}&q=0,\\\ \Lambda(1)^{\oplus n}&q=-2,\\\ 0&\text{otherwise.}\end{cases}$ In other words, all entries of $\mathbf{R}\pi_{\ast}\mathscr{K}_{C}$ are free $\Lambda$-modules. Therefore we have $\mathrm{H}^{q}(C,\Lambda)=\begin{cases}\Lambda^{\oplus r}&q=0,\\\ \Lambda(-1)^{\oplus n}&q=2,\\\ 0&\text{otherwise.}\end{cases}$ So (2) is verifies. (c) is also by [8, IX, 4.7]. ∎ ###### Remark 2.2. In fact, every connected component of $(C\otimes_{k}\bar{k})_{\mathrm{red}}$ is a tree of nonsingular rational curves with normal crossings. ## 3\. Calculation of étale homology on surfaces with rational singularities In this paper, a surface means a two-dimensional integral normal scheme which satisfies (†). Note that for each surface $X$, $\delta(x)\mathrel{\mathop{:}}=2-\dim\mathcal{O}_{X,x}$ is a dimension function on $X$; and for all points $x\in X$, we have $\delta(x)=\dim\overline{\\{x\\}}$. From now on, all surfaces are tacitly equipped with this type of dimension function. We recall the definition of _rational singularity_ (cf. [1] & [11]). A normal local ring $R$ of dimension $2$ is said to have a _rational singularity_ if there exists a desingularization $f\colon X\to\operatorname{Spec}(R)$ such that $\mathrm{H}^{1}(X,\mathcal{O}_{X})=0$. By [11, Proposition (16.5)], for a normal local ring $R$ of dimension $2$, if one of the four rings $R$, $\widehat{R}$, $R^{\mathrm{sh}}$ and $\widehat{R^{\mathrm{sh}}}$ has a rational singularity, so have the other three rings. A surface $X$ is said to have a rational singularity at a closed point $x$, if the local ring $\mathcal{O}_{X,x}$ has a rational singularity. Now we calculate the étale homology on arbitrary surfaces. ###### Proposition 3.1. Let $X$ be a surface which is the spectrum of a strictly Henselian local ring, $P$ the closed point of $X$, $U\mathrel{\mathop{:}}=X\setminus P$. Assume that $U$ is regular, and $X$ admits a desingularization $\pi\colon\widetilde{X}\to X$ which induces an isomorphism $\pi^{-1}(U)\xrightarrow{\,\sim\,}U$. Let $E\mathrel{\mathop{:}}=\pi^{-1}(P)$ be the exceptional divisor. Then we have 1. (1) The map $\mathrm{H}^{0}(\mathrm{t}_{X})\colon\Lambda\xrightarrow{\,\sim\,}\mathrm{H}_{4}(X)(-2)$ is an isomorphism. 2. (2) $\mathrm{H}_{3}(X)\xrightarrow{\,\sim\,}\mathrm{H}^{1}\bigl{(}E,\Lambda(2)\bigr{)}$. 3. (3) There is an exact sequence of $\Lambda$-modules: $0\to\mathrm{H}_{2}(E)\to\mathrm{H}^{2}\bigl{(}E,\Lambda(2)\bigr{)}\to\mathrm{H}_{2}(X)\to\mathrm{H}_{1}(E)\to 0\,,$ Moreover the kernel of the map $\mathrm{H}_{2}(X)\to\mathrm{H}_{1}(E)$ is a finite group. 4. (4) There is an exact sequence of $\Lambda$-modules: $0\to\mathrm{H}_{1}(X)\to\mathrm{H}_{0}(E)\xrightarrow{\deg}\Lambda\to 0\,.$ 5. (5) $\mathrm{H}_{q}(X)=0$ for $q>4$ or $q<1$. ###### Proof. After applying section 1 to the following Cartesian square, $\textstyle{E\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\tilde{i}}$$\scriptstyle{\tau}$$\textstyle{\square}$$\textstyle{\widetilde{X}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\pi}$$\textstyle{P\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{i}$$\textstyle{X}$ (3.1) we obtain a long exact sequence $\cdots\to\mathrm{H}_{n+1}(X)\to\mathrm{H}_{n}(E)\to\mathrm{H}_{n}(P)\oplus\mathrm{H}_{n}(\widetilde{X})\to\mathrm{H}_{n}(X)\to\cdots\,.$ (3.2) For each $q\in\mathbb{Z}$, there is a canonical isomorphism of $\Lambda$-modules: $\mathrm{H}_{q}(\widetilde{X})\xrightarrow{\,\sim\,}\mathrm{H}^{4-q}\bigl{(}\widetilde{X},\Lambda(2)\bigr{)}\xrightarrow{\,\sim\,}\mathrm{H}^{4-q}\bigl{(}E,\Lambda(2)\bigr{)}\,,$ where the first isomorphism is induced by section 1 (since $\widetilde{X}$ is regular), and the last isomorphism is from the base change theorem via the square (3.1). Thus $\mathrm{H}_{q}(\widetilde{X})=0$ for $q<2$ and $q>4$. Also note that $\mathrm{H}_{q}(E)=0$ for $q<0$ and $q>2$; and $\mathrm{H}_{q}(P)=\begin{cases}0&q\neq 0,\\\ \Lambda&q=0.\end{cases}$ Hence we may divide (3.2) to three exact sequences $\displaystyle 0\to\mathrm{H}^{0}\bigl{(}E,\Lambda(2)\bigr{)}\to\mathrm{H}_{4}(X)\to 0,$ (3.3) $\displaystyle 0\to\mathrm{H}^{1}\bigl{(}E,\Lambda(2)\bigr{)}\to\mathrm{H}_{3}(X)\to\mathrm{H}_{2}(E)\to\mathrm{H}^{2}\bigl{(}E,\Lambda(2)\bigr{)}\to\mathrm{H}_{2}(X)\to\mathrm{H}_{1}(E)\to 0,$ (3.4) $\displaystyle 0\to\mathrm{H}_{1}(X)\to\mathrm{H}_{0}(E)\xrightarrow{\deg}\Lambda\to\mathrm{H}_{0}(X)\to 0.$ (3.5) Now (1) is by the exact sequence (3.3). From the sequence (3.5), to prove (4) & (5) we have only to prove the map $\mathrm{H}_{0}(E)\xrightarrow{\deg}\Lambda$ is epimorphic. Since $X$ is strictly local, the field $k\mathrel{\mathop{:}}=k(P)$ is separably closed. Let $Q$ be a closed point on $E$. Then $\deg\mathrm{cl}_{E}(Q)=\bigl{[}k(Q),k\bigr{]}$. Put $p\mathrel{\mathop{:}}=\operatorname{char}k$. Since $k$ is separably closed, the number $\bigl{[}k(Q),k\bigr{]}$ is either $1$ or a power of $p$; in both cases, it is invertible on $\Lambda$ (as $\ell\neq p$). Hence $\deg$ is surjective. From the sequence (3.4), to prove (2) & (3) we have only to prove that the homomorphism $\mathrm{H}_{2}(E)(-1)\to\mathrm{H}^{2}\bigl{(}E,\Lambda(1)\bigr{)}$ is injective and its cokernel is a torsion group. Let $E_{1},E_{2},\ldots,E_{n}$ be all irreducible components of $E$. Then $\mathrm{H}_{2}(E)(-1)$ is a free $\Lambda$-Module with a basis $\bigl{\\{}\mathrm{cl}_{E}(E_{i})\bigr{\\}}^{n}_{i=1}$. Note that the composite map $\mathrm{H}_{2}(E)(-1)\to\mathrm{H}_{2}(\widetilde{X})(-1)\xrightarrow{\,\sim\,}\mathrm{H}^{2}\bigl{(}\widetilde{X},\Lambda(1)\bigr{)}\xrightarrow{\,\sim\,}\mathrm{H}^{2}\bigl{(}E,\Lambda(1)\bigr{)}\xrightarrow[\,\sim\,]{\rho}\Lambda^{\oplus n}$ (3.6) sends each $\mathrm{cl}_{E}(E_{i})$ to $\sum\limits^{n}_{j=1}a_{ij}e_{j}$, where the isomorphism $\rho$ is defined in [8, IX, 4.7], $e_{1},e_{2},\ldots,e_{n}$ is the standard basis for $\Lambda^{\oplus n}$, and $a_{ij}=\deg\bigl{(}\mathscr{L}_{\widetilde{X}}(E_{i})\|_{E_{j}}\bigr{)}=(E_{i},E_{j})\,,$ where $\mathscr{L}_{\widetilde{X}}(E_{i})$ means the invertible sheaf on $\widetilde{X}$ associated to the divisor $E_{i}$. In other words, the map (3.6) is given by the intersection matrix $\bigl{(}(E_{i},E_{j})\bigr{)}$, which is negative-definite by [11, (14.1)]. Therefore the map (3.6) is injective with torsion cokernel. ∎ ###### Theorem 3.2. Let the assumptions and the notations be as in section 3. We further assume that $X$ has a rational singularity at $P$. Then we have 1. (1) Let $\mathrm{Cl}(X)$ denote the Weil divisor class group of $X$. Then there is a canonical isomorphism $\mathrm{cl}\colon\mathrm{Cl}(X)\otimes_{\mathbb{Z}}\Lambda\xrightarrow{\,\sim\,}\mathrm{H}_{2}(X)(-1)$ which sends each $[Y]\otimes 1$ to $\mathrm{cl}_{X}(Y)$. 2. (2) The map $\mathrm{H}^{0}(\mathrm{t}_{X})\colon\Lambda\xrightarrow{\,\sim\,}\mathrm{H}_{4}(X)(-2)$ is an isomorphism. 3. (3) $\mathrm{H}_{q}(X)=0$ for $q\neq 2,4$. ###### Proof. By [11, (4.1)], any desingularization of a surface with rational singularities is a product of quadratic transformations. Since $X$ is the spectrum of a local ring, the exceptional divisor $E$ is connected. Thus the map $\deg\colon\mathrm{H}_{0}(E)\to\Lambda$ is an isomorphism by section 2 (1a). Then (2) & (3) are by section 3 and section 2. So we have only to prove (1). We adopt the assumptions and notation in [11, §14]. Let $E_{1},E_{2},\ldots,E_{n}$ be all irreducible components of $E$. For each $i$, let $d_{i}>0$ be the greatest common divisor of all the degrees of invertible sheaves on $E_{i}$. Put $k\mathrel{\mathop{:}}=k(P)$ and $p\mathrel{\mathop{:}}=\operatorname{char}k$. For each $i$, let $Q_{i}$ be a closed point on $\mathrm{Reg}(E_{i})$. Then $Q_{i}$ defines an invertible sheaf of degree $\bigl{[}k(Q_{i}),k\bigr{]}$ on $E_{i}$. Since the number $\bigl{[}k(Q_{i}),k\bigr{]}$ is invertible on $\Lambda$, so is $d_{i}$. Let $\mathbf{E}$ be the additive group of divisors on $\widetilde{X}$ generated by $\\{E_{i}\\}^{n}_{i=1}$. Then $\mathbf{E}$ is a free abelian group with basis $E_{1},E_{2},\ldots,E_{n}$. Put $\mathbf{E}^{\ast}\mathrel{\mathop{:}}=\operatorname{Hom}_{\mathbb{Z}}(\mathbf{E},\mathbb{Z})$ and let $\varepsilon_{1},\varepsilon_{2},\ldots,\varepsilon_{n}$ be the dual basis. We define a homomorphism of groups: $\vartheta\colon\mathbf{E}\to\mathbf{E}^{\ast}\,,\qquad E_{i}\mapsto\sum^{n}_{j=1}\tfrac{1}{d_{j}}(E_{i},E_{j})\cdot\varepsilon_{j}\,.$ By [11, (14.3), (14.4) & (17.1)], there is an exact sequence $0\to\mathbf{E}\xrightarrow{\vartheta}\mathbf{E}^{\ast}\to\mathrm{Cl}(X)\to 0\,.$ Let $e_{1},e_{2},\ldots,e_{n}$ be the basis for the free $\Lambda$-module $\mathrm{H}^{2}\bigl{(}E,\Lambda(1)\bigr{)}$ defined by the isomorphism in section 2 (2c). Since each $d_{j}$ is invertible on $\Lambda$, the homomorphism $\delta\colon\mathbf{E}^{\ast}\otimes_{\mathbb{Z}}\Lambda\to\mathrm{H}^{2}\bigl{(}E,\Lambda(1)\bigr{)}\,,\qquad\varepsilon_{j}\otimes 1\mapsto d_{j}e_{j}$ is an isomorphism of $\Lambda$-modules. It is easy to check that the following diagram $\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathbf{E}\otimes_{\mathbb{Z}}\Lambda\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\vartheta\otimes\mathrm{id}}$(i)$\scriptstyle{\cong}$$\textstyle{\mathbf{E}^{\ast}\otimes_{\mathbb{Z}}\Lambda\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\delta}$$\scriptstyle{\cong}$$\textstyle{\mathrm{Cl}(X)\otimes_{\mathbb{Z}}\Lambda\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\scriptstyle{\mathrm{cl}}$$\textstyle{0}$$\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathrm{H}_{2}(E)(-1)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathrm{H}^{2}\bigl{(}E,\Lambda(1)\bigr{)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{\mathrm{H}_{2}(X)(-1)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$$\textstyle{0}$ is commutative with both rows exact, where the isomorphism (i) sends each $E_{i}\otimes 1$ to $\mathrm{cl}_{E}(E_{i})$. Therefore $\mathrm{cl}\colon\mathrm{Cl}(X)\otimes_{\mathbb{Z}}\Lambda\xrightarrow{\,\sim\,}\mathrm{H}_{2}(X)(-1)$ is an isomorphism. ∎ By [11, (17.1)], the divisor class group $\mathrm{Cl}(X)$ is a finite group. So we have: ###### Corollary 3.3. Let $X$ be a surface having only finitely many singular points, all of which are rational singularities. Then $\mathrm{t}_{X}\colon\mathbb{Q}_{\ell}(2)[4]\xrightarrow{\,\sim\,}\mathscr{K}_{X}$ is an isomorphism in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\mathbb{Q}_{\ell})$. In particular, $\mathbb{Q}_{\ell}$ is a dualizing complex on $X_{\textup{\'{e}t}}$ in coefficient $\mathbb{Q}_{\ell}$. ###### Corollary 3.4 (Poincaré Duality). Let $k$ be a separably closed field on which $\ell$ is invertible. And let $X$ be a surface over $k$. Assume that $X$ has only finitely many singular points, all of which are rational singularities. Then 1. (1) The trace map $\operatorname{Tr}_{X}\colon\mathrm{H}_{\mathrm{c}}^{4}\bigl{(}X,\mathbb{Q}_{\ell}(2)\bigr{)}\xrightarrow{\,\sim\,}\mathbb{Q}_{\ell}$ is an isomorphism. 2. (2) For every $0\leqslant r\leqslant 4$ and every object $\mathscr{F}$ in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\mathbb{Q}_{\ell})$, the pairing $\mathrm{H}_{\mathrm{c}}^{r}(X,\mathscr{F})\times\mathrm{Ext}^{4-r}_{X}\bigl{(}\mathscr{F},\mathbb{Q}_{\ell}(2)\bigr{)}\to\mathrm{H}_{\mathrm{c}}^{4}\bigl{(}X,\mathbb{Q}_{\ell}(2)\bigr{)}\xrightarrow[\,\sim\,]{\operatorname{Tr}_{X}}\mathbb{Q}_{\ell}$ is nondegenerate. 3. (3) For every $0\leqslant r\leqslant 4$, the pairing $\mathrm{H}_{\mathrm{c}}^{r}(X,\mathbb{Q}_{\ell})\times\mathrm{H}^{4-r}\bigl{(}X,\mathbb{Q}_{\ell}(2)\bigr{)}\to\mathrm{H}_{\mathrm{c}}^{4}\bigl{(}X,\mathbb{Q}_{\ell}(2)\bigr{)}\xrightarrow[\,\sim\,]{\operatorname{Tr}_{X}}\mathbb{Q}_{\ell}$ is nondegenerate. ###### Proof. (3) is a special case of (2). Let $p\colon X\to\operatorname{Spec}k$ be the structural morphism. By section 3 and section 1, we have isomorphisms $\mathbb{Q}_{\ell}(2)[4]\xrightarrow[\,\sim\,]{\mathrm{t}_{X}}\mathscr{K}_{X}\xrightarrow{\,\sim\,}\mathbf{R}p^{!}\mathbb{Q}_{\ell}\,.$ Hence we have $\displaystyle\mathrm{Ext}_{X}^{4-r}\bigl{(}\mathscr{F},\mathbb{Q}_{\ell}(2)\bigr{)}$ $\displaystyle\xrightarrow{\,\sim\,}\operatorname{Hom}_{\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\mathbb{Q}_{\ell})}\bigl{(}\mathscr{F}[r],\mathbb{Q}_{\ell}(2)[4]\bigr{)}$ $\displaystyle\xrightarrow{\,\sim\,}\operatorname{Hom}_{\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\mathbb{Q}_{\ell})}\bigl{(}\mathscr{F}[r],\mathbf{R}p^{!}\mathbb{Q}_{\ell}\bigr{)}$ $\displaystyle\xrightarrow{\,\sim\,}\operatorname{Hom}_{\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(\mathbb{Q}_{\ell})}\bigl{(}\mathbf{R}p_{!}\mathscr{F}[r],\mathbb{Q}_{\ell}\bigr{)}$ $\displaystyle\xrightarrow{\,\sim\,}\mathrm{H}_{\mathrm{c}}^{r}(X,\mathscr{F})^{\vee}\,.$ So (2) is proved. In above isomorphism, letting $r=4$ and $\mathscr{F}=\mathbb{Q}_{\ell}$, we obtain $\mathrm{H}_{\mathrm{c}}^{4}(X,\mathbb{Q}_{\ell})^{\vee}\xrightarrow{\,\sim\,}\mathrm{H}^{0}\bigl{(}X,\mathbb{Q}_{\ell}(2)\bigr{)}=\mathbb{Q}_{\ell}(2)\,.$ So (1) is proved. ∎ ###### Corollary 3.5. Let $X$ be a surface having only finitely many singular points, all of which are rational singularities. Assume further that for any singular point $P$, $\mathcal{O}_{X,P}^{\mathrm{sh}}$ is factorial. Then $\mathrm{t}_{X}\colon\mathbb{Z}_{\ell}(2)[4]\xrightarrow{\,\sim\,}\mathscr{K}_{X}$ is an isomorphism in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\mathbb{Z}_{\ell})$. In particular, $\mathbb{Z}_{\ell}$ is a dualizing complex on $X_{\textup{\'{e}t}}$ in coefficient $\mathbb{Z}_{\ell}$. ###### Remark 3.6. By [11, (17.2)], the ring $\mathcal{O}_{X,P}^{\mathrm{sh}}$ is factorial if and only if its completion $\widehat{\mathcal{O}}_{X,P}^{\mathrm{sh}}$ is factorial. See [11, §25] for the complete list of two-dimensional factorial complete local rings with rational singularities. ## 4\. Applications In this section, we study perverse sheaves and the weights of smooth sheaves on surfaces with rational singularities. First we briefly introduce the t-structures on arbitrary schemes. Let $X$ be a scheme satisfying (†) equipped with a dimension function $\delta$. For each point $x$ on $X$, the canonical morphism $i_{x}\colon\operatorname{Spec}k(x)\to X$ factor as $\operatorname{Spec}k(x)\xrightarrow{j}\overline{\\{x\\}}\xrightarrow{i}X\,.$ Then we define two functors: $\displaystyle i_{x}^{\ast}$ $\displaystyle\mathrel{\mathop{:}}=j^{\ast}\circ i^{\ast}\colon\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})\to\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}\bigl{(}x_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell}\bigr{)}\,,$ $\displaystyle i_{x}^{!}$ $\displaystyle\mathrel{\mathop{:}}=j^{\ast}\circ\mathbf{R}i^{!}\colon\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})\to\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}\bigl{(}x_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell}\bigr{)}\,.$ Now we define a pair of full subcategories $\bigl{(}\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\leqslant 0}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell}),\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\geqslant 0}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})\bigr{)}$ of $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$ as following: for every object $\mathscr{F}$ in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$, 1. (1) $\mathscr{F}\in\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\leqslant 0}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$ if and only if for any point $x$ on $X$, $i_{x}^{\ast}\mathscr{F}\in\mathbf{D}_{\mathrm{c}}^{\leqslant-\delta(x)}(x_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$; 2. (2) $\mathscr{F}\in\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\geqslant 0}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$ if and only if for any point $x$ on $X$, $i_{x}^{!}\mathscr{F}\in\mathbf{D}_{\mathrm{c}}^{\geqslant-\delta(x)}(x_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$. By [5] or [6], this pair defines a $t$-structure on $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$. If $f\colon X\to Y$ is a quasi-finite morphism, then the functors $\mathbf{R}f_{!}$ and $f^{\ast}$ are right t-exact, and the functors $\mathbf{R}f^{!}$ and $\mathbf{R}f_{\ast}$ are left t-exact. Thus we may define the functor $f_{!\ast}$. ###### Proposition 4.1. Let $X$ be a surface having only finitely many singular points, all of which are rational singularities; $U$ an open dense subset contained in $X_{\mathrm{reg}}$ and $j\colon U\hookrightarrow X$ the inclusion. Let $\mathscr{F}$ be a smooth $\overline{\mathbb{Q}}_{\ell}$-sheaf on $X_{\textup{\'{e}t}}$. Then $\mathscr{F}[2]$ is an object in $\mathrm{Perv}(X,\overline{\mathbb{Q}}_{\ell})$ and there is a canonical isomorphism $j_{!\ast}\circ j^{\ast}\mathscr{F}[2]\xrightarrow{\,\sim\,}\mathscr{F}[2]$. ###### Proof. Obviously we may assume that $U=X_{\mathrm{reg}}$. Then $Y\mathrel{\mathop{:}}=X\setminus U$ contains only finitely many singular points, all of which are rational singularities of $X$. Let $P$ be a point in $X\setminus U$. Then $i_{P}^{\ast}\mathscr{F}[2]=\mathscr{F}_{P}[2]\in\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\leqslant-2}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})\,.$ Next we calculate $i_{P}^{!}\mathscr{F}[2]$. Put $\overline{X}\mathrel{\mathop{:}}=\operatorname{Spec}\mathcal{O}_{X,P}^{\mathrm{sh}}$, and let $\rho\colon\overline{X}\to X$ and $\iota\colon P\hookrightarrow\overline{X}$ be the canonical morphisms. Since $\overline{X}$ is strictly local and $\mathscr{F}$ is a smooth sheaf, $\rho^{\ast}\mathscr{F}\xrightarrow{\,\sim\,}\overline{\mathbb{Q}}_{\ell}^{\oplus r}$ for some $r\in\mathbb{N}$. So we have $\displaystyle i_{P}^{!}\mathscr{F}[2]$ $\displaystyle=\mathbf{R}\iota^{!}\circ\rho^{\ast}\mathscr{F}[2]\xrightarrow{\,\sim\,}\mathbf{R}\iota^{!}\overline{\mathbb{Q}}_{\ell}^{\oplus r}[2]$ $\displaystyle\xrightarrow{\,\sim\,}\bigl{(}\mathbf{R}\iota^{!}\mathscr{K}_{\overline{X}}(-2)[-4]\bigr{)}^{\oplus r}[2]$ $\displaystyle\qquad(\text{\autoref{M-A21}})$ $\displaystyle=\mathbf{R}\iota^{!}\mathscr{K}_{\overline{X}}^{\oplus r}(-2)[-2]$ $\displaystyle\xrightarrow{\,\sim\,}\mathscr{K}_{P}^{\oplus r}(-2)[-2]$ $\displaystyle\qquad(\text{\autoref{P-78E}})$ $\displaystyle\xrightarrow{\,\sim\,}\overline{\mathbb{Q}}_{\ell}^{\oplus r}(-2)[-2]\,.$ Hence $i_{P}^{!}\mathscr{F}[2]\in\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\geqslant 2}(X_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$. This show that $\mathscr{F}[2]\in\mathrm{Perv}(X,\overline{\mathbb{Q}}_{\ell})$. Let $i\colon Y\hookrightarrow X$ and $j\colon U\hookrightarrow X$ be the inclusion. Then $i^{\ast}\mathscr{F}[2]\in\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\leqslant-2}(Y_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$ and $\mathbf{R}i^{!}\mathscr{F}[2]\in\vphantom{\mathbf{D}}^{\mathrm{P}}\mathbf{D}_{\mathrm{c}}^{\mathrm{b},\geqslant 2}(Y_{\textup{\'{e}t}},\overline{\mathbb{Q}}_{\ell})$. So both ${}^{\mathrm{P}}i^{\ast}\mathscr{F}[2]$ and ${}^{\mathrm{P}}i^{!}\mathscr{F}[2]$ are zero. By section 1, there are two long exact sequences in $\mathrm{Perv}(X,\overline{\mathbb{Q}}_{\ell})$: $\displaystyle\cdots\to{}^{\mathrm{P}}j_{!}\circ j^{\ast}\mathscr{F}[2]\to\mathscr{F}[2]\to i_{\ast}\circ{}^{\mathrm{P}}i^{\ast}\mathscr{F}[2]\to\cdots\,,$ $\displaystyle\cdots\to i_{\ast}\circ{}^{\mathrm{P}}i^{!}\mathscr{F}[2]\to\mathscr{F}[2]\to{}^{\mathrm{P}}j_{\ast}\circ j^{\ast}\mathscr{F}[2]\to\cdots\,.$ Thus ${}^{\mathrm{P}}j_{!}\circ j^{\ast}\mathscr{F}[2]\to\mathscr{F}[2]$ is an epimorphism and $\mathscr{F}[2]\to{}^{\mathrm{P}}j_{\ast}\circ j^{\ast}\mathscr{F}[2]$ is a monomorphism. Therefore $j_{!\ast}\circ j^{\ast}\mathscr{F}[2]=\mathscr{F}[2]$. ∎ Next we study the $\ell$-adic cohomology of surfaces over finite fields. Let $p\neq\ell$ be a prime number, $q\mathrel{\mathop{:}}=p^{e}$ and $k\mathrel{\mathop{:}}=\mathbb{F}_{q}$. If $X$ is an algebraic scheme over $k$, we write $\overline{X}\mathrel{\mathop{:}}=X\otimes_{k}\bar{k}$; and for each sheaf $\mathscr{F}$ (or complex of sheaves) on $X_{\textup{\'{e}t}}$, we use $\overline{\mathscr{F}}$ to denote the inverse image of $\mathscr{F}$ on $\overline{X}_{\textup{\'{e}t}}$. ###### Definition 4.2. Let $w\in\mathbb{Z}$. An element $\alpha\in\overline{\mathbb{Q}}_{\ell}^{\ast}$ is said to be _pure of weight $w$_ if for any $\mathbb{Q}$-embedding $\iota\colon\overline{\mathbb{Q}}_{\ell}\hookrightarrow\mathbb{C}$, we have $\|\iota(\alpha)\|=q^{r/2}$. ###### Definition 4.3. Let $X$ be an algebraic scheme over $k$. 1. (1) A constructible $\overline{\mathbb{Q}}_{\ell}$-sheaf $\mathscr{F}$ on $X$ is said to be * • _punctually pure of weight $n$_ if for every closed point $x$ on $X$, all eigenvalues of the geometric Frobenius $F_{x}$ acting on $\mathscr{F}_{\bar{x}}$ are pure of weight $w$; * • _mixed_ (_of weights $\leqslant w$_), if there exists a finite filtration $0=\mathscr{F}_{r}\subset\mathscr{F}_{r-1}\subset\cdots\subset\mathscr{F}_{1}\subset\mathscr{F}_{0}=\mathscr{F}$ consisting of constructible $\overline{\mathbb{Q}}_{\ell}$-sheaves, such that each $\mathscr{F}_{i}/\mathscr{F}_{i+1}$ is punctually pure (of weights $\leqslant n$). 2. (2) An object $\mathscr{F}$ in $\mathbf{D}_{\mathrm{c}}^{\mathrm{b}}(X_{\textup{\'{e}t}},\widetilde{\mathbb{Q}}_{\ell})$ is said to be * • _mixed of weights $\leqslant w$_ if for every $n\in\mathbb{Z}$, $\mathrm{H}^{n}(\mathscr{F})$ is mixed of weights $\leqslant n+w$; * • _mixed of weights $\geqslant w$_ if $\mathrm{D}(\mathscr{F})=\mathbf{R}\mathscr{H}\\!om(\mathscr{F},\mathscr{K}_{X})$ is mixed of weights $\leqslant-w$; * • _pure of weight $w$_ if it is mixed of weights $\leqslant w$ and $\geqslant w$. ###### Proposition 4.4. Let $X$ be a complete surface over $k$, $\mathscr{F}$ a smooth $\overline{\mathbb{Q}}_{\ell}$-sheaf on $X_{\textup{\'{e}t}}$. Assume that $X$ has only finitely many singular points, all of which are rational singularities; and $\mathscr{F}$ is punctually pure of weight $w$. Then for every $n\in\mathbb{N}$, the $\overline{\mathbb{Q}}_{\ell}$-module $\mathrm{H}^{n}(\overline{X}_{\textup{\'{e}t}},\overline{\mathscr{F}})$ is punctually pure of weight $n+w$. ###### Proof. Put $U\mathrel{\mathop{:}}=X_{\mathrm{reg}}$ and let $j\colon U\hookrightarrow X$ be the inclusion. Since $U$ is smooth and $j^{\ast}\mathscr{F}$ is a smooth $\overline{\mathbb{Q}}_{\ell}$-sheaf, $j^{\ast}\mathscr{F}[2]$ is pure of weight $w+2$. By section 4, $\mathscr{F}[2]\in\mathrm{Perv}(X,\overline{\mathbb{Q}}_{\ell})$ and $j_{!\ast}j^{\ast}\mathscr{F}[2]\cong\mathscr{F}[2]$. So after applying [2, Corollaire 5.4.3], we obtain that $j_{!\ast}j^{\ast}\mathscr{F}[2]\cong\mathscr{F}[2]$ is pure of weight $w+2$. Let $\pi\colon X\to\operatorname{Spec}k$ be the structural morphism. By [3, (3.3.1) & (6.2.3)], the functor $\mathbf{R}\pi_{\ast}=\mathbf{R}\pi_{!}$ sends perverse sheaves pure of weight $w+2$ on $X$ to perverse sheaves pure of weight $w+2$ on $\operatorname{Spec}k$. Hence we prove the proposition. ∎ ## References * [1] M. Artin. On isolated rational singularities of surfaces. Amer. J. Math., 88:129–136, 1966. * [2] A. A. Beĭlinson, J. Bernstein, and P. Deligne. Faisceaux pervers. In Analysis and topology on singular spaces, I (Luminy, 1981), volume 100 of Astérisque, pages 5–171. Soc. Math. France, Paris, 1982\. * [3] P. Deligne. La conjecture de Weil. II. Inst. Hautes Études Sci. Publ. Math., (52):137–252, 1980. * [4] T. Ekedahl. On the adic formalism. In The Grothendieck Festschrift, Vol. II, volume 87 of Progr. Math., pages 197–218. Birkhäuser Boston, Boston, MA, 1990. * [5] K. Fujiwara. A proof of the absolute purity conjecture (after Gabber). In Algebraic geometry 2000, Azumino (Hotaka), volume 36 of Adv. Stud. Pure Math., pages 153–183. Math. Soc. Japan, Tokyo, 2002. * [6] O. Gabber. Notes on some $t$-structures. In Geometric aspects of Dwork theory. Vol. I, II, pages 711–734. Walter de Gruyter GmbH & Co. KG, Berlin, 2004. * [7] O. Gabber. Finiteness theorems for ètale cohomology of excellent schemes. In Conference in honor of P. Deligne on the occasion of his 61st birthday, IAS, Princeton, October 2005. * [8] A. Grothendieck. Théorie des Topos et Cohomologie Étale des Schémas, SGA 4. Lecture Notes in Math. 269, 270, 305. Springer-Verlag, Berlin, 1972-73. * [9] A. Grothendieck. Cohomologie l-adique et fonctions L, SGA 5. Lecture Notes in Math. 589. Springer-Verlag, Berlin, 1977. * [10] G. Laumon. Homologie étale. In Séminaire de géométrie analytique (École Norm. Sup., Paris, 1974-75), pages 163–188. Astérisque, No. 36–37. Soc. Math. France, Paris, 1976. * [11] J. Lipman. Rational singularities, with applications to algebraic surfaces and unique factorization. Inst. Hautes Études Sci. Publ. Math., (36):195–279, 1969. * [12] J. Riou. Dualité (d’après ofer gabber). unpublished, Dec. 2007.	arxiv-papers	2008-08-30T15:32:11	2024-09-04T02:48:57.563982	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Ting Li", "submitter": "Ting Li", "url": "https://arxiv.org/abs/0809.0068" }
0809.0198	# An embedded-atom method model for liquid Co, Nb, Zr and supercooled binary alloys Pascal Thibaudeau Commissariat à l’Energie Atomique Le Ripault, BP 16, F-37260 Monts, France Julian D. Gale Nanochemistry Research Institute Department of Applied Chemistry, Curtin University of Technology, GPO Box U1987, Perth 6845, Western Australia ###### Abstract The parameters of many-body potentials for Co, Nb and Zr metals, based on the embedded-atom method, have been systematically derived. The analytical potential scheme allows us to reproduce correctly the cohesive energies and structural properties of the pure metals and selected alloys making use of a small set of parameters. With a pair potential going smoothly to zero for a sufficient cutoff radius, radial partial and bond angular distribution functions for Co, Nb, Zr and alloys are computed using molecular dynamics simulations that ensure good quantitative agreement with the available experimental data up to the melting point. Atomic short range order is analysed in the light of consecutive Gaussian function decomposition and Honeycutt-Andersen indices. ###### pacs: 61.25.Mv, 71.15.Mb ## I Introduction Today we are not surprised that a non-crystalline solid orders magnetically. It is known that, with few important exceptions, the amorphous and crystalline phases of the same material do not differ very much magnetically. In the last three decades, the discovery and large scale investigation of rapidly solidified alloys have made possible a new scenario in basic and applied magnetism. The unusual behaviour of the bulk metallic magnetic glasses (BMMGs) - formed by supercooling the liquid state of certain metallic magnetic alloys - usually occurs for systems containing atoms that exhibit a well-known sensitivity to the immediate neighbourhood Inoue:2003 ; Yavari:2007 . However, despite the large number of technological applications of BMMGs, the detailed origin of the links between structural and magnetic properties has yet to be established, but is strongly dependent on the atomic short-range order Miracle:2004 . Nowadays, atomic-scale simulations of solids based on interatomic potentials are routinely performed to explore the short-range order of BMMGs Rozler:2000 ; Wolf:2004 as a precursor to the study of the magnetic ordering, which is evidently beyond such methods. However, the construction of realistic $n$-body potentials is mandatory to any simulations. When applied to metals and alloys, several major methods and extensions of construction of such potentials have been established from density functional theory, i.e., the embedded-atom method (EAM) Daw:1984 or tight-binding second moment approximation, i.e., the Finnis-Sinclair (FS) model Finnis:1984 and related models Hausleitner:1992 ; Philips:1994 . The two models have very similar computational requirements, and the names are often used interchangably, however there are some distinctions which come to the fore when considering multicomponent alloys Finnis:2003 . Because of the parameters involved in these model potentials, the EAM method was first used to study simple metals in their relevant crystalline structures Oh:1988 ; Johnson:1989 . Extensions to binary alloys were formulated Johnson:1989a with special attention to hcp-fcc Cai:1996 and hcp-bcc Zhang:2005 systems. Because they are often based on spin-less approximations of electronic density, such effective model potentials traditionally neglect the magnetic ordering Finnis:1984 ; Finnis:2003 ; Ackland:1997 ; Mendelev:2003 . Notable exceptions have been published recently by Dudarev et al. Dudarev:2005 and Ackland et al. Ackland:2003 ; Ackland:2006 , but such methods are still in their infancy. Even in spin-less schemes, these effective potentials may be able to represent the compositional ordering which is, at least in the localised magnetism picture, a prerequisite for understanding the magnetic behaviour. Existing magnetic potentials concentrate on the one-site magnetism, for which the energetics reduces in form to a simple embedded-atom-type potential, explaining the success of standard EAM schemes on magnetic materials. Supercooled Co1-xNbx and Co1-xZrx magnetic metallic glasses in the cobalt-rich region represent test cases for both hcp-bcc and hcp-hcp metal sub-systems. These compounds have been experimentally studied because they are strong ferromagnets, as revealed by a high value of the exchange constant Suran:1990 . Consequently, large short-range compositional inhomogeneities should induce significant variations of the long-range magnetic order. The present work considers both the construction of embedded-atom potentials for such materials, as well as the application to molecular dynamics (MD) simulations in order to assess one of the issues in the large field of these BMMGs, namely short-range order quasicrystallinity. ## II Methodology for potentials In the FS method, the total internal energy $E$ of a $N$-atom system and the electron density, $\rho({\bf{R}}_{i})$, for an atom located at ${\bf{R}}_{i}$ due to all other atoms are given as; $E=\frac{1}{2}\sum_{i,j,(i\neq j)}^{N}\phi(r_{ij})-\sum_{i=1}^{N}F(\rho({\bf{R}}_{i})),$ (1) $\rho({\bf{R}}_{i})=\sum_{j\neq i}^{N}f(r_{ij}),$ (2) where $f(r_{ij})$ is the electron density at atom $i$ due to atom $j$ as a function of the distance between them, $r_{ij}=\\|{\bf{R}}_{i}-{\bf{R}}_{j}\\|$ is the separation distance between atoms $i$ and $j$, $F(\rho({\bf{R}}_{i}))$ is the energy to embed atom $i$ in an electron density $\rho({\bf{R}}_{i})$, and $\phi(r_{ij})$ is a two-body potential between atoms $i$ and $j$. As long as an angular independent formulation is considered, the electron density is a radial function only. For an alloy model, an embedding function, $F$, has to be specified for each atomic species supplemented by an atomic electron- density function, $f$, and a two-body potential, $\phi$, specified for each possible combination of atomic species. For uncompressed metals, Gupta Gupta:1981 and Tomànek et al. Tomanek:1985 have shown that the host electron density can be represented as an exponentially decreasing function of the distance to better account for atomic relaxation near impurities and surfaces. In this approximation, $f$ is given as; $f(r)=f_{e}\exp(-\chi(r/r_{e}-1)),$ (3) where $f_{e}$ is a scaling factor determined by the cohesive energy, $E_{c}$, and the atomic volume, $r_{e}$ is the nearest-neighbour distance in the relevant pair of atoms and $\chi$ is an adjustable parameter. Analysing the interatomic interactions in effective-medium theory, Jacobsen et al. Jacobsen:1987 have shown that if an exponential form is chosen for the density function, then the interatomic potential, $\phi$, should also be an exponential function of the distance. In this study, the interatomic potentials of all the pairs considered are defined by a potential of the form; $\phi(r)=\left\\{\begin{array}[]{ll}A\exp(-r/r_{0})&0\leq r\leq r_{1},\\\ \displaystyle{\sum_{i=0}^{5}a_{i}r^{i}}&r_{1}\leq r\leq r_{m},\\\ 0&r_{m}\leq r,\end{array}\right.$ (4) where the interaction is designed so as to go smoothly to zero at the distance $r_{m}$ according to a polynomial spline function. The potentials are constructed subject to the constraints that the radial functions and their first and second derivatives must be continuous at the boundary points, and also that the function must have a stationary point at $r_{m}$. Once $A$, $r_{0}$, $r_{1}$ and $r_{m}$ are fixed, this procedure ensures that the coefficients, $\\{a_{i}\\}$, are uniquely determined by solving a simple $6\times 6$ linear system of equations. These coefficients are reported in Table 1 for completeness. For all the pairs of atoms, $r_{1}=2.5$Å is kept fixed and corresponds to a typical radius where the stiff repulsive part of the potential ends in metals Dai:2006 ; Yifang:1996 ; Hausleitner:1990 . Because of the screening in metals, the stationary point is located at least between the second and third nearest neighbours for the lowest energy crystal phases as previously noted Bangwei:1993 . Since the embedding energy is assumed to be independent of the source of the electron density and the hopping integrals are a function only of a radial distance between atoms, the embedding functional $F$ is taken as $F[\rho(r)]=\sqrt{\rho(r)}.$ (5) This functional form gives a band energy proportional to the square root of the second moment of the electron density of states Gupta:1981 . However, moments of higher order cannot be expressed in such a simple analytical form and a more complex method must be applied Turchi:1985 . Johnson Johnson:1989a has considered that since the electron density at any location is taken as a linear superposition of atomic electron densities, this function should be taken directly from monoatomic models with a relative scaling factor between elements for an alloy model. On the other hand, Finnis and Sinclair Finnis:1984 and Cleri and Rosato Cleri:1993 have considered mixed pair electron-density functions not necessarily connected to the atomic ones, removing the alloy scaling factors. For hcp-Co and hcp-Zr, the parameters for the atomic electron-density and the interatomic potential are fitted in order to reproduce the experimental cohesive energies, the unit cell parameters and the five independent elastic constants of these systems as given by Cleri and Rosato Cleri:1993 . Moreover, cell parameters and elastic constants of fcc-Co and bcc-Zr are also included during the fitting procedure as taken from references Yoo:1998 ; Modak:2006 ; Willaime:1991 and references therein. Generally, for a small cutoff distance the largest number of interacting neighbours per atom in the crystalline structure leads to the more stable phase. Ducastelle Ducastelle:1972 has shown that in the second-moment approximation, with interactions restricted to the nearest neighbours, the cohesive energy for the hcp and fcc phases is the same and the $c/a$ ratio is equal to the ideal value. Hence, it is necessary to go up to at least the fourth-moment approximation to discriminate between the fcc and hcp phases and to give a value of $c/a$ different from $2\sqrt{2/3}$. In our case, since the $c/a$ ratios are not taken to be the ideal one and because the potentials have a very short range, the cutoff distance of atomic electron density should be larger than that of the potential. So for the electron density, the cutoff distance is taken to be $4.87$Å, thus including up to seven shells of neighbours within hcp-Co, and three for both hcp-Zr and bcc-Nb, which are all the stable phases for each pure element. A check is also performed to ensure that at least three shells of neighbours have been included for the high pressure/high temperature structures since this is necessary to keep the relative energies of each phases in the correct order Willaime:1991 . Moreover, it has been observed that in incorporating the elastic constants of the bcc-Zr phase in the database, the fitting of both the cell parameters and cohesive energy of the hcp-Zr is rather poor with this cutoff radius, so only the hcp-Co, fcc-Co, hcp-Zr and bcc-Nb elastic constants are included during the fit. For Nb, the cohesive energy, the lattice parameter and the three independent elastic constants of the bcc-Nb phase are taken from reference Pasianot:1991 and the theoretical fcc-Nb cell parameter and cohesive energy are also included Haglund:1993 . No elastic constants of the fcc-Nb were found to incorporate into the training set of observables. The selection of the functional form taken in Eq.(1)-(2) is extended to AB alloys based on the second-moment form that has been applied to Zr Willaime:1989 ; Willaime:1991 . The embedding function, atomic electron- density function and two-body potential are assumed to be of the same form as in Eq.(3)-(4), with $\phi_{AB}$ and $\phi_{BA}$ assumed to be equal. The alloy potentials and atomic electron-density functions are determined independently of the monoatomic counterparts if sufficient data are available. However, it is known that for equilibrium immiscible systems it is a challenging task to fit cross potentials, since there is often insufficient experimental data related to the respective alloy compounds. In order to circumnavigate this problem, density functional calculations have been performed on selected intermetallic structures using the Quantum-ESPRESSO package espresso . For these calculations, non-local ultrasoft pseudo-potentials are employed in combination with a plane-wave basis set. The generalized-gradient approximation, as parametrized by Perdew, Burke and Ernzerhof Perdew:1996 , is selected for the exchange and correlation term. For the Brillouin zone sampling, a 12$\times$12$\times$12 Monkhorst-Pack mesh is used for the $k$-point summation in the self-consistent calculations Monkhorst:1976 for all the primitive cells, which leads to converged structural parameters to within 0.1% of the cell parameter. Thus the lattice constants and cohesive energies of several Co-(Zr,Nb) and Zr-Nb crystalline structures reported in Table (3) are obtained and then included in the fitting procedure for the Co-(Zr,Nb) and Zr-Nb cross potentials and atomic electron-density functions. The parameters of the fitted terms are listed in Table 1. \| Co-Co \| Nb-Nb \| Zr-Zr \| Co-Nb \| Co-Zr \| Nb-Zr ---\|---\|---\|---\|---\|---\|--- $f_{e}$ (eV2) \| 1.6862 \| 11.0160 \| 3.4022 \| 37.6462 \| 18.2157 \| 39.0130 $\chi$ \| 3.4513 \| 5.9621 \| 3.6880 \| 3.8908 \| 2.8733 \| 5.5304 $r_{e}$(Å) \| 2.4968 \| 2.8579 \| 3.2133 \| 2.5879 \| 2.7985 \| 3.0591 $A$ (eV) \| 162418.11 \| 2758.20 \| 27693.54 \| 1353.63 \| 13168.79 \| 3092.13 $r_{0}$(Å) \| 0.1770 \| 0.3495 \| 0.2613 \| 0.3211 \| 0.2322 \| 0.3754 $r_{m}$(Å) \| 3.5269 \| 3.6450 \| 3.5798 \| 4.3151 \| 3.4794 \| 3.4939 $a_{0}$ (eV) \| 193.2032 \| 622.3713 \| 1066.3553 \| 107.1854 \| 399.1098 \| 1720.4867 $a_{1}$ (eV.Å-1) \| -301.8495 \| -937.6188 \| -1616.1544 \| -143.1635 \| -619.6420 \| -2777.0048 $a_{2}$ (eV.Å-2) \| 188.4275 \| 572.3560 \| 984.8131 \| 76.5117 \| 386.7728 \| 1810.9619 $a_{3}$ (eV.Å-3) \| -58.7182 \| -176.1599 \| -301.0688 \| -20.4017 \| -121.1298 \| -593.4294 $a_{4}$ (eV.Å-4) \| 9.1312 \| 27.2357 \| 46.1097 \| 2.7101 \| 19.0083 \| 97.3551 $a_{5}$ (eV.Å-5) \| -0.5668 \| -1.6874 \| -2.8270 \| -0.1434 \| -1.1944 \| -6.3808 Table 1: Potential and atomic electron-density parameters for (Co,Nb,Zr) systems. The overall fitting procedure is performed in two separate steps. First, the densities and potentials are derived for the simple metals. Once the corresponding parameters are obtained, the fit is applied to selected binary metals for cross-densities and potentials without altering the terms for the simple metals. The same cutoff radius is kept constant during all the steps and the calculations are performed within the GULP computer code GaleRohl:2003 . In Table 2, a list of some basic physical properties as computed by the present set of potentials and the corresponding experimental values are shown for Co, Nb and Zr. \| \| hcp-Co \| fcc-Co \| hcp-Zr \| bcc-Zr \| bcc-Nb \| fcc-Nb ---\|---\|---\|---\|---\|---\|---\|--- a(Å) \| Fitted \| 2.5065 \| 3.5414 \| 3.1865 \| 3.5358 \| 3.3138 \| 4.2198 Experiment(1) \| 2.507 \| 3.544 \| 3.2317 \| 3.574 \| 3.3 \| 4.23 c(Å) \| Fitted \| 4.0606 \| - \| 5.2035 \| - \| - \| - Experiment(1) \| 4.0689 \| - \| 5.1476 \| - \| - \| - $E_{c}$ \| Fitted \| -4.402 \| -4.360 \| -6.192 \| -5.970 \| -7.577 \| -7.383 (eV/atom) \| Experiment(1) \| -4.386 \| - \| -6.167 \| -6.13 \| -7.57 \| -7.39 $C_{11}$ (GPa) \| Fitted \| 315 \| 255 \| 150 \| 103 \| 245 \| 101 Experiment(2) \| 319 \| 242 \| 154 \| 104 \| 245 \| - $C_{12}$ (GPa) \| Fitted \| 155 \| 159 \| 80 \| 89 \| 132 \| 122 Experiment(2) \| 166 \| 160 \| 67 \| 93 \| 132 \| - $C_{44}$ (GPa) \| Fitted \| 78 \| 128 \| 33 \| 70 \| 28 \| 25 Experiment(2) \| 82 \| 128 \| 36 \| 38 \| 28 \| - $C_{13}$ (GPa) \| Fitted \| 111 \| 159 \| 52 \| 89 \| 132 \| 122 Experiment(2) \| 102 \| 160 \| 65 \| 93 \| 132 \| - $C_{33}$ (GPa) \| Fitted \| 373 \| 255 \| 177 \| 103 \| 245 \| 101 Experiment(2) \| 373 \| 242 \| 172 \| 104 \| 245 \| - $C_{66}$ (GPa) \| Fitted \| 80 \| 128 \| 35 \| 70 \| 28 \| 25 Experiment(2) \| 77 \| 128 \| 44 \| 38 \| 28 \| - Table 2: Physical properties for Co, Zr and Nb simple metals as fitted with a cutoff radius of 4.87Å. (1) Cohesive energies and lattice parameters are taken from Kittel Kittel:1966 , (2) elastic constants are taken from Simmons and Wang Simmons:1971 . The fit correctly reproduces the structures and properties of hcp-Co, fcc-Co, hcp-Zr and bcc-Nb. The absolute average percentage difference between calculation and experiment is found to be 0.6%, 0.7% and 8%, for the cell parameters, cohesive energies elastic constants, respectively. However, the elastic constants of fcc-Nb produced a negative Young’s modulus, as expected, in this excited locally unstable structural phase Craievich:1997 . The properties of bcc-Zr are reproduced with a sufficient accuracy, except for the $C_{44}$ elastic constant where the largest percentage error of 45% occurs. The potentials fitted by Willaime et al. Willaime:1991 also exhibit such a discrepancy though with a much larger error. This may be corrected by relaxing the constraint of the square-root form of the embedding functional and considering more neighbours Johnson:1989 or by including explicit angularly dependent terms in the potentials at a cost of additional parameters Baskes:1994 . In Table 3, a list of some basic physical properties fitted from these potentials and the corresponding ab initio calculated values for selected binary alloys is shown. \| a (Å) \| $E_{c}$ (eV/atom) ---\|---\|--- Structure \| Fitted \| ab initio \| Fitted \| ab initio CoZr (B2) \| 3.1612(1) \| 3.1753 \| -6.9368 \| -6.9850 Co2Zr (C15) \| 6.8944 \| 6.9040 \| -6.8659 \| -6.7946 Co3Zr (L12) \| 3.7058 \| 3.7189 \| -6.3484 \| -6.5120 CoNb (B2) \| 2.9502 \| 3.0523 \| -8.3599 \| -8.1750 Co2Nb (C15) \| 6.6038 \| 6.7357(2) \| -7.5813 \| -7.6012 Co3Nb (L12) \| 3.5233 \| 3.6289 \| -7.2423 \| -7.2425 ZrNb (B2) \| 3.4380 \| 3.4380 \| -8.6865 \| -8.6200 Zr2Nb (C15) \| 7.9961 \| 8.0208 \| -8.2544 \| -7.7621 Nb2Zr (C15) \| 7.8805 \| 7.8565 \| -9.1236 \| -9.1725 Table 3: Lattice parameters and cohesive energies for selected binary metals in their corresponding symmetry as computed with a cutoff radius of 4.87Å. (1) experimental value of 3.181Å Buschow:1982a , (2) experimental value of 6.774Å Pargeter:1967 . The B2, C15 and L12 classification is relative to the strukturbericht structural types classification Smithell:1983 . The B$2$, C$15$ and L$1_{2}$ denominations are relative to the strukturbericht structural types classification Smithell:1983 , such as B2 is the CsCl structure type, C15 is the MgCu2 structure type and L12 is the Cu3Au type. The fitting procedure appears to correctly reproduce the ab initio derived phase stability order and lattice parameters. However, the absolute cohesive energies are reproduced to a lesser extent. This may be also improved consistently by increasing the cutoff radius on Zr and Nb electron density terms, which includes more neighbours in the total energy sums of Eq.(1). ## III Application to simple metals The validity of this potential in describing the atomic interactions can be illustrated outside the original systems used for parametrisation by considering liquid cobalt. An MD simulation with 300 atoms, which allows an individual description up to the 13th nearest-neighbors in hcp-Co, and periodic boundary conditions was first performed within the isobaric, isothermal ensemble (NPT) at a temperature of 1670 K, which is slightly below the experimental melting point. The simulation was run for 200 ps with a time step of 0.1 fs to simulate the radial pair distribution function (RPDF). Once thermal equilibrium is reached, the RPDFs are sampled every 0.2 ps to produce an average. As these RPDFs are subject to statistical noise, a smoothing formula is used to replace each RPDF by a least-squares polynomial that fits a sub-range of several points. For all the calculated radial functions, a third- degree, five-point smoothing procedure is applied several times on the data until convergence Hildebrand:1965 . MD simulations are repeated on three different atomic configurations at the same temperature in order to sample more accurately the configurational space and an overall averaged RPDF is computed. This averaged RPDF compares very well to the experimental data Holland-Moritz:2002 as shown in Fig. 1. Figure 1: Simulated and experimental RPDF $g(r)$ of cobalt at a temperature of 1670 K, below the melting point. Experimental data from Holland-Moritz:2002 are shown by points. The insert shows the simulated $rg(r)$ at 1670 K and its analysis in Gaussian peaks (dashed curves). The agreement is similar to that presented by Bhuiyan et al. Bhuiyan:1995 and more recently by Han and coworkers Han:2004 using much more elaborate EAM potentials, but in these works no quantitative analysis of the RPDFs was performed. However, the first peak of the RPDF $g(r)$ appears to be asymmetric and might be composed of more than one atomic shell. Following Kita et al. Kita:1994 , a decomposition of $rg(r)$ in consecutive Gaussian functions is applied. The insert in Fig. 1 shows the decomposition of $rg(r)$ at 1670 K in six Gaussian functions where all 18 parameters are allowed to vary freely during the fit. The first coordination shell is defined by a cutoff distance $r_{c}$, which is taken to be the first minimum of $g(r)$. For this temperature, $r_{c}=3.46$Å. The first peak is composed of three Gaussian subpeaks located at $r_{1}=2.407$Å, $r_{2}=2.645$Å and $r_{3}=3.095$Å, respectively. The coordination number $N_{c}$ is calculated by integrating the Gaussian function according to $N_{c}=4\pi^{3/2}n_{0}A\sigma r_{i}$, where $n_{0}$ is the atomic density, $A$ is the amplitude of the Gaussian function, $\sigma$ is the square root of the variance, and $r_{i}$ is the maximum radius. For each subpeak, $N_{c}$ is equal to $3.46$, $4.71$ and $3.81$ with a sum of $11.98$. For a temperature of 1800K, which is slightly greater than the experimental melting point, the sum decreases to $11.57$. These coordination numbers are close to the experimental ones found in liquid Co Holland- Moritz:2002 ($12.5\pm 0.5$ at 1670 K and $12.1\pm 0.5$ at 1800 K) and consistent with those calculated for other metallic systems Kresse:1993 . Such a high value of the coordination number and the possibility of decomposing the first peak of $g(r)$ is an indication that the short-range order of the liquid Co is more complex than the one given by a simple icosahedral ordering as suggested by Holland et al. Holland-Moritz:2002 . Moreover, performing MD simulations for these two temperatures allows us to predict a variation of the density with the temperature of $d\rho/dT=-9.68$ $10^{-4}$gcm-3K-1 in good agreement with the experimentally reported value of $d\rho/dT=-9.88$ $10^{-4}$gcm-3K-1 value Smithell:1983 . MD simulations were repeated under the same conditions for pure Zr to a higher temperature of 2290 K, above the experimental melting point. The experimental RPDF is compared against the simulation of this liquid state, shown in Fig. 2. For this metal at that temperature, the first peak is composed of two Gaussian functions located at $r_{1}=3.091$Å and $r_{2}=3.574$Å, respectively. The ratio $r_{2}/r_{1}=1.156$ is close to that of the two first nearest-neighbour distances for a bcc lattice $2/\sqrt{3}=1.1547$. The coordination number $N_{c}$ for each subpeak is $6.10$ and $5.59$ with a sum of $11.69$. This value is close to the experimental one of $11.9\pm 0.5$ found in liquid Zr Schenk:2002 . Even if the ratio of the radii tends to favour a bcc lattice, the corresponding coordination is very different. Figure 2: Simulated and experimental RPDF $g(r)$ of liquid zirconium at a temperature of 2290 K. Experimental data from Schenk:2002 are shown by points. The insert shows the simulated $rg(r)$ at 2290 K and its analysis in Gaussian subpeaks (dashed curves). MD simulations are repeated under the same conditions for pure Nb to a higher temperature of 2750 K, which is the experimental melting point. The simulated RPDF of this liquid state is shown in Fig. 3. For this metal, the first peak of $rg(r)$ is strongly asymmetric and is composed of three Gaussian functions up to $r_{c}=4.0$Å, located at $r_{1}=2.691$Å, $r_{2}=3.141$Å and $r_{3}=3.840$Å. The ratio $r_{2}/r_{1}=1.167$ is greater than that of the two first nearest-neighbour distances for a bcc lattice. The coordination number $N_{c}$ for each subpeak is $2.95$, $7.06$ and $3.10$ with a sum of $13.11$. To our knowledge, no experimental data on the radial distribution function of pure Nb is available in order to compare with. Figure 3: Simulated RPDF $g(r)$ of liquid niobium at a temperature of 2750 K. The insert shows the simulated $rg(r)$ and its analysis in seven Gaussian subpeaks (dashed curves). The short-range order can also be examined by calculating the bond angle distribution functions $g(\theta)$ that represent the angle between the bonds connecting a central atom to two neighbouring atoms, as illustrated in Fig. 4 for liquid cobalt, zirconium and niobium. The angle is calculated for pairs of interatomic distances given by a cutoff corresponding to the first minimum of the RPDF (i.e. 3.5Å for Co, 4.3Å for Zr and 4.1Å for Nb). In the case of Zr, the calculated distribution exhibits a prominent peak near $\theta=57^{\circ}$ (close to an equilateral triangle), a broader maximum near $\theta=109^{\circ}$ and a rather flat maximum near $\theta=150^{\circ}$. In the case of Co, the first peak is broader and close to $55^{\circ}$ whereas the second peak enlarges but remains at $109^{\circ}$. For Nb, the situation is different with two broader peaks near $50^{\circ}$ and $99^{\circ}$. Viewing the structure in terms of dominant clusters, the bond angle with the highest density at the nearest neighbour distance are for a regular icosahedron $63.4^{\circ}$ and $116.4^{\circ}$, while for fcc the prominent angles are $60^{\circ}$, $90^{\circ}$ and $120^{\circ}$. For hcp, angles of $109.471^{\circ}$ and $146.443^{\circ}$ are added when the ratio $c/a=2\sqrt{2/3}$ but they are less frequent. For a bcc lattice, the prominent angles are $70.53^{\circ}$ and $109.471^{\circ}$. In the Zr structure, the first peak tends to favour the fcc and hcp structures while the second peak tends to favour the hcp and bcc. This means that the dominant structure should be hcp. However, the angles of $90^{\circ}$ and $120^{\circ}$ are not so strong whereas some defective icosahedron angles should be there too. This suggests a predominantly distorted icosahedral character. For the same reasons, the case of Co also favours defective icosahedron as well. For Nb as the first peak is much more located near $50^{\circ}$, this suggests much more intriguing short-range structures with less neighbours. Figure 4: Bond angle distribution at $T=2290$ K for liquid Zr (solid line), at $T=1800$ K for liquid Co (dashed line) and at $T=2750$ K for liquid Nb (dotted line). The peaks in the bond angle distribution for perfect icosahedral order are indicated by the vertical lines. To assess more quantitatively the local structures in amorphous alloys, Honeycutt and Andersen (HA) analysis has been proven to successfully differentiate face-centered cubic, hexagonal close packed, icosahedron and binary bcc structures Honeycutt:1987 ; Jonsson:1988 . To perform such analysis, a set of four indices is constructed for each pair: (i) the first index denotes to what peak of the RPDF, $g(r)$, the pair under consideration belongs; (ii) the second index represents the number of near neighbours shared by this pair; (iii) the third index counts the number of nearest-neigbour bonds among the shared neighbours; (iv) and a fourth index is used to differentiate configurations with the same three indices, but with a different topology. For instance, fcc crystals are fully described by four pairs, such as 1421, 2101, 2211 and 2441, whereas hcp crystals also contains the 1422 and 2331 pairs in addition. Moreover, the 1441, 1661, 2101, 2211, and 2441 are the only pairs in a perfect bcc crystal. Icosahedral order is described by Mackay icosahedra, composed of twinned, distorted fcc tetrahedra with an index of 1551, whereas the 1541 and 1431 indices are more characteristic of a distorted icosahedral local order. Up to the distance cutoff corresponding to first minimum of each $g(r)$, Table 4 reports the HA analysis in the liquid state of each simple elements. index \| Co \| Zr \| Nb ---\|---\|---\|--- 1311 \| 0.07$\pm$0.01 \| 0.06$\pm$0.01 \| 0.03$\pm$0.01 1321 \| 0.07$\pm$0.01 \| 0.06$\pm$0.01 \| 0.05$\pm$0.01 1421 \| 0.04$\pm$0.01 \| 0.03$\pm$0.01 \| 0.02$\pm$0.01 1422 \| 0.07$\pm$0.01 \| 0.06$\pm$0.01 \| 0.04$\pm$0.01 1431 \| 0.19$\pm$0.01 \| 0.17$\pm$0.01 \| 0.13$\pm$0.01 1441 \| 0.04$\pm$0.01 \| 0.05$\pm$0.01 \| 0.07$\pm$0.01 1541 \| 0.15$\pm$0.01 \| 0.15$\pm$0.01 \| 0.11$\pm$0.01 1551 \| 0.12$\pm$0.01 \| 0.14$\pm$0.01 \| 0.14$\pm$0.01 1661 \| 0.05$\pm$0.01 \| 0.06$\pm$0.01 \| 0.07$\pm$0.01 2101 \| 1.58$\pm$0.02 \| 1.55$\pm$0.01 \| 1.52$\pm$0.02 2211 \| 0.95$\pm$0.02 \| 0.94$\pm$0.02 \| 0.96$\pm$0.02 2321 \| 0.23$\pm$0.01 \| 0.22$\pm$0.01 \| 0.21$\pm$0.01 2331 \| 0.50$\pm$0.02 \| 0.53$\pm$0.02 \| 0.58$\pm$0.02 2441 \| 0.08$\pm$0.01 \| 0.08$\pm$0.01 \| 0.07$\pm$0.01 Table 4: Honeycutt and Andersen analysis of the simulations in the liquid state for Co (T=1670K), Zr (T=2290K) and Nb (T=2750K). The microscopic analysis emerging from the data of Table 4 indicates that the short-range order of the liquid state is dominated by distorted icosaheral and icosahedral structures since the 1541, 1431 and 1551 indices respectively are large as anticipated. The high value of the 2331 pairs is also an indication of the icosahedral order. The small distortion from perfect icosahedral order observed in the angular distributions of these liquids suggests that the local icosahedral order should dominate. However, small distortion form a perfect tetrahedron does not form different HA indices from those for a perfect icosahedron. Our HA analysis shows that the icosahedral distortion is larger than reported by the bond angle distribution curves. This result is in agreement with the experimental investigation on liquid Ti, Zr and Ni conducted by Kim and Kelton Kim:2007 . Using first-principles molecular dynamics simulations, Jakse and Pasturel Jakse:2003 have concluded there exists competition between a polyhedral and bcc-type short-range order in liquid and supercooled Zr, whereas Kim and Kelton Kim:2007 have reported no regular dominant cluster type that can describe the experimental liquid structure of transition metals, including Zr. This is supported by the values of the HA indices reported in Table 4 which are not very different from each element in the liquid state. The abundance of the 1661 pairs indicates that bcc order is very low, but slightly increases when going from Co to Nb. Interestingly, the lowest energy geometrical structures of magnetic cobalt clusters mainly follow an icosahedral growth pattern with some cubic-type structures at some particular sizes Rodriguez-Lopez:2004 . For liquid Zr, these low values have been reported both experimentally Kim:2007 and using first-principles molecular dynamics simulations Jakse:2003 . Furthermore for liquid Zr, Jakse and Pasturel have performed HA analysis on the inherent structures and found an abundance of the 1551 pairs in the liquid state that is twice the value found here. Such structures indicate the presence of perfect icosahedra as a local minima of the potential energy surface. For instance on liquid Nb, the 1551 index goes to 0.30$\pm$0.01 on inherent structure. ## IV Application to binary alloys To validate the quality and transferability of our potentials, lattice constants and atomic internal positions of several crystal structures not entering in the fit have been calculated minimising the free energy at T=300K, and compared with the experimental values Shurin:1965 ; Kuzma:1966 . The results are summarised in Table 5 for the varying degrees of freedom according to the corresponding space group and the agreement is generally good with an average absolute error of 1.76%. \| Co7Nb6 \| Co23Zr6 \| ---\|---\|---\|--- \| exp.Shurin:1965 \| EAM(300K) \| \| exp.Kuzma:1966 \| EAM(300K) a(Å) \| 5.01 \| 5.019(+0.18%) \| a(Å) \| 11.516 \| 11.484(-0.28%) c(Å) \| 26.5 \| 24.66(-6.95%) \| x${}_{Co_{3}}$ \| 0.378 \| 0.3791(+0.30%) x${}_{Co_{1}}$ \| 0.5 \| 0.5005(+0.10%) \| x${}_{Co_{4}}$ \| 0.178 \| 0.1732(-2.71%) y${}_{Co_{1}}$ \| 0.5 \| 0.4995(-0.10%) \| x${}_{Zr_{1}}$ \| 0.208 \| 0.2075(-0.26%) z${}_{Co_{1}}$ \| 0.59 \| 0.5801(-1.67%) \| \| \| z${}_{Nb_{1}}$ \| 0.167 \| 0.1664(-0.34%) \| \| \| z${}_{Nb_{2}}$ \| 0.346 \| 0.3211(-7.19%) \| \| \| z${}_{Nb_{3}}$ \| 0.448 \| 0.4359(-2.71%) \| \| \| Table 5: Lattice constants and atomic internal positions of Co7Nb6 and Co23Zr6 calculated with our potentials at 300K and compared to the experimental values Shurin:1965 ; Kuzma:1966 . To complete the validation, the elastic constants of CoNb and Co3Nb are calculated ab initio applying finite differences to the stress tensor and compared with those calculated analytically by our potentials. The results are shown in Table 6 and the agreement is satisfying. \| CoNb ($D_{2}$) \| Co3Nb ($L_{12}$) ---\|---\|--- \| ab initio \| EAM \| ab initio \| EAM C11 (GPa) \| 251 \| 242 \| 368 \| 357 C12 (GPa) \| 173 \| 143 \| 164 \| 194 C44 (GPa) \| 60 \| 71 \| 160 \| 131 Table 6: ab-initio elastic constants of selected binaries as compared to the present set of potentials. As for liquids, MD simulations are performed on Co0.9Zr0.1 for a 300 atom system with periodic boundary conditions. First, the atoms are placed randomly into the simulation cell using a hard sphere criterion based on their atomic radii and the cell volume is then adjusted according to the phenomenological Miedema theory Miedema:1980 , which gives good estimates for the experimental volume of glasses of these alloys. Then at constant pressure and a temperature of 1800 K (higher than the liquidus phase boundary for this alloy composition), MD has been run for 200 ps to simulate the liquid phase. The sample is then quenched at a rate of $7.5\times 10^{14}$Ks-1 and maintained at 300 K for at least a further 200 ps. In Fig. 5, the simulated RPDFs are computed and compared against experiment Babanov:1998 . Figure 5: Radial partial distribution functions of Co0.9Zr0.1 at 300 K compared with experiment (open circles) Babanov:1998 . For each atomic pair, the insert shows the simulated $rg(r)$ at 300 K and its analysis in Gaussian peaks (dashed curves). The simulated partial distribution functions for cobalt compare well to experiment including the position of the first and second peaks. However, the experimental Zr-Zr RPDF does not exhibit a structural trend, whereas our simulation does. Rößler and Teichler have reported similar results in their study of atomic mobilities and structural properties of supercooled amorphous Co1-xZrx using very different interatomic potentials Rozler:2000 . This suggests a possible lack of resolution both in X-ray diffraction and EXAFS because of a low contribution in the spectra of these minority atoms, as previously anticipated Rozler:2000 . However, our potentials seem to reproduced correctly the position of the first peak for Zr-Zr and the double peak character of the second peak in the Co-Co and Co-Zr distributions in comparison with ref.Rozler:2000 . The maximum of the Co-Co peak is simulated to occur at 2.44Å, compared with an experimental result of 2.42Å Babanov:1998 . Using the Gaussian subpeak analysis, this first peak is found to be composed of 3 shells at $2.43$Å, $2.55$Å and $2.88$Å with coordination numbers of $4.42$, $4.83$ and $2.30$. The total coordination number up to $r_{c}=3.12$Å is equal to $11.55$ in comparison with $10.90$ found experimentally. In our simulations a narrower scattering of the positions of the first 3 subpeaks is observed, in contrast to the liquid state, which is anticipated during the cooling. In the closest crystalline form, Co23Zr6, the highest coordination number is obtained for a pair of cobalt atoms located at $2.4334$Å in a cubic cluster and other local structures with $4$ and $3$ neighbours are also found at 2.36Å, 2.51Å, 2.51Å and 2.81Å Kuzma:1966 . This suggests tetrahedral clustering below $r_{c}$ or defective icosahedra up to $r_{c}$ in this supercooled alloy. On the other hand, the maximum of the first peak in Co-Zr is simulated to be at 2.79Å, in excellent agreement with the reported experimental distance of 2.79Å Babanov:1998 . The analysis through Gaussian functions reveals 2 subpeaks located at 2.71Å and 3.00Å with a coordination number of 8.68 and 8.40, respectively. This also suggests a clustering of the bcc-type at very short range. MD simulations have also been performed for Co0.9Nb0.1 under the same conditions. The reported RPDFs are shown in Fig. 6. In Co0.9Nb0.1 and Co0.9Zr0.1, the Co-Co first peak distance is calculated to be at 2.437Å and is not affected by the non-magnetic added atoms at such a low concentration. The situation changes for the second and third peaks with a much more distinct third shell in Co0.9Nb0.1 than in Co0.9Zr0.1. Interestingly, the Co-Nb (resp.Co-Zr) first neighbour equilibrium distance is 2.49Å (resp.2.79Å). This is lower than the simple prediction related to their corresponding atomic radii (2.70Å (resp.2.85Å)) Vainshtein:2000 . However, Jamet et al. Jamet:2000 have reported a Co-Nb distance of 2.58Å in studying cobalt nanoparticles embedded in a niobium matrix. Figure 6: Simulated radial partial distribution functions of Co0.9Nb0.1 quenched at 300K. The simulated correlation in the minority pairs is structured in the Co-Nb system, as for Co-Zr. However, in both cases the first shell of neighbours seems to be depleted of their atoms to fill the second or third shells. It is doubtful that these structures are an artefact of the low number of atoms considered in our simulations because Rößler and Teichler have simulated systems more than twice the size of ours and found the same behaviour. Up to $r_{c}=3.1$Å, the bond angular distributions of these two systems are calculated and shown in Fig. 7. These distributions exhibit well-structured peaks suggesting more crystalline environments including a $150^{\circ}$ distinct angle in the Co0.9Nb0.1. As the concentration of minority atoms is low, these distributions are dominated by hcp-like Co clusters upon cooling. This trend is more pronounced for added Nb atoms than Zr atoms probably because of their corresponding atomic radii. Figure 7: Bond angular distribution functions of Co0.9Zr0.1 and Co0.9Nb0.1 quenched at 300K. To assess such hypothesis, the HA indices are calculated and reported in Table 7. index \| Co0.9Zr0.1 \| Co0.9Nb0.1 ---\|---\|--- 1311 \| 0.02 \| 0.06 1321 \| 0.03 \| 0.03 1421 \| 0.02 \| 0.06 1422 \| 0.04 \| 0.11 1431 \| 0.14 \| 0.21 1441 \| 0.05 \| 0.01 1541 \| 0.15 \| 0.17 1551 \| 0.24 \| 0.16 1661 \| 0.08 \| 0.03 2101 \| 1.49 \| 1.57 2211 \| 0.81 \| 0.87 2321 \| 0.20 \| 0.20 2331 \| 0.70 \| 0.62 2441 \| 0.10 \| 0.13 Table 7: Honeycutt and Andersen analysis of the simulations for supercooled Co0.9Zr0.1 and Co0.9Nb0.1 at 300K. The absolute error bars of the abundances are 0.01. In Co0.9Zr0.1 the dominant pairs are 1551 and 1541 exhibiting more icosahedral than distorted icosahedral order. This tendency is inverted in Co0.9Nb0.1 system with a larger amount of 1431 pairs favouring distorted icosahedral order. This is explained by a smaller difference in atomic radii between Co and Nb than Co and Zr because an atomic size difference of approximately 10% can relieve spatial frustration and stabilise the icosahedral structure Nelson:1989 . In Co0.9Nb0.1 the 1422 pairs are abundant indicating an hcp order which is less present in Co0.9Zr0.1 system. Even if the value is rather low, the 1661 pairs are of importance indicating a slight bcc order, as anticipated. It is observed that while useful for gaining understanding of the evolution of the dominant short-range order, the single cluster model cannot capture the richness of the supercooled binary alloys structures. ## V Conclusion A fitting procedure has been performed to consistently derive a self- consistent set of many-body parameters for Co, Zr and Nb simple metals and selected alloys, including validation against first principles results where there are gaps in the experimental data. Combined with MD simulations, these parameters allow us to calculate RPDFs and bond angular distributions in the liquid phase for Co, Zr and Nb. Applied to supercooled binary alloys, clear short-range order is shown in agreement with available experiments for the majority pairs. The situation is different for the minority pairs within the Co-rich region. In this region, both simulated Zr-Zr and Nb-Nb RPDFs are correlated and exhibit transfers of atoms from the first shells of neighbours to the second and third. The Honeycutt-Andersen analysis exhibits mainly both distorted and pure icosahedral orders of various degrees in competition with other crystalline orders in the liquid phases and supercooled alloys. JDG would like to thank the Government of Western Australia for a Premier’s Research Fellowship. ## References * [1] A. Inoue, B. Shen, H. Koshiba, H. Kato, and A.R. Yavari. Nature Materials, 2:661, 2003. * [2] A.R. Yavari. Nature Materials, 6:181, 2007. * [3] D. B. Miracle. Nature Materials, 3:697, 2004. * [4] U.K. Rößler and H. Teichler. Phys. Rev. E, 61:394, 2000. * [5] D. Wolf, V. Yamanakov, S.R. Phillpot, A. Mukherjee, and H. Gleiter. Acta Materialia, 53:1, 2004. * [6] M.S. Daw and M.I. Baskes. Phys. Rev. B, 29:6443, 1984. * [7] M.W. Finnis and J.E. Sinclair. Phil. Mag. A, 50:45, 1984. * [8] Ch. Hausleitner and J. Hafner. Phys. Rev. B, 45:115, 1992. * [9] R Philips, J. Zou, A.E. Carlsson, and M. Widom. Phys. Rev. B, 49:9322, 1994. * [10] M.W. Finnis. Interatomic Forces in Condensed Matter. Oxford University Press, Oxford, 2003. * [11] D.J. Oh and R.A. Johnson. J. Mater. Res., 3:471, 1988. * [12] R.A. Johnson and D.J. Oh. J. Mater. Res., 4:1195, 1989. * [13] R.A. Johnson. Phys. Rev. B, 39:12554, 1989. * [14] J. Cai and Y.Y. Ye. Phys. Rev. B, 54:8398, 1996. * [15] R.F. Zhang, Y. Kong, and B.X. Liu. Phys. Rev. B, 71:214102, 2005. * [16] G.D. Ackland, D.J. Bacon, A.F. Calder, and T. Harry. Phil. Mag. A, 75:713, 1997. * [17] M.I. Mendelev, S. Han, D.J. Srolovitz, G.D. Ackland, D.Y. Sun, and M. Asta. Phil. Mag., 83:3977, 2003. * [18] S.L. Dudarev and P.M. Derlet. J. Phys.: Condens. Matter, 17:7097, 2005. * [19] G.D. Ackland and S.K. Reed. Phys. Rev. B, 67:174108, 2003. * [20] G.D. Ackland. Phys. Rev. Lett., 97:015502, 2006. * [21] G. Suran, M. Naili, M Rivoire, and J.C.S. Levy. J. Appl. Phys., 67:5649, 1990. * [22] R.P. Gupta. Phys. Rev. B, 23:6265, 1981. * [23] D. Tománek, A.A. Aligia, and C.A. Balseiro. Phys. Rev. B, 32:5051, 1985. * [24] K.W. Jacobsen, J.K. Nøskkov, and M.J. Puska. Phys. Rev. B, 35:7423, 1987. * [25] X.D. Dai, Y. Kong, J.H. Li, and B.X. Liu. J. Phys.: Condens. Matter, 18:4527, 2006. * [26] O. Yifang, Z. Bangwei, L. Shuzhi, and J. Zhanpeng. Z. Phys. B - Condensed Matter, 101:161, 1996. * [27] Ch. Hausleitner and J. Hafner. J. Phys.: Condens. Matter, 2:6651, 1990. * [28] Z. Bangwei and O. Yifang. Phys. Rev. B, 48:3022, 1993. * [29] P. Turchi and F. Ducastelle. The Recursion Method and Its Applications. Springer-Verlag, Berlin, 1985. * [30] F. Cleri and V. Rosato. Phys. Rev. B, 48:22, 1993. * [31] C.S Yoo, P. Söderlind, and H. Cynn. J. Phys.: Condens. Matter, 10:L311, 1998. * [32] P. Modak, A.K. Verma, R.S. Rao, B.K. Godwal, and R. Jeanloz. Phys. Rev. B, 74:012103, 2006. * [33] F. Willaime and C. Massobrio. Phys. Rev. B, 43:11653, 1991. * [34] F. Ducastelle. PhD thesis, Université de Paris-Sud, Orsay, France, 1972. * [35] R. Pasianot, D. Farkas, and E.J. Savino. Phys. Rev. B, 43:6952, 1991. * [36] J. Häglund, A. Fernandez-Guillermet, G Grimvall, and M. Körling. Phys. Rev. B, 48:11685, 1993. * [37] F. Willaime and C. Massobrio. Phys. Rev. Lett., 63:2244, 1989. * [38] S. Baroni, A. Dal Corso, S. de Gironcoli, P. Giannozzi, C. Cavazzoni, G. Ballabio, S. Scandolo, G. Chiarotti, P. Focher, A. Pasquarello, K. Laasonen, A. Trave, R. Car, N. Marzari, and A. Kokalj. http://www.pwscf.org. * [39] J.P. Perdew, K. Burke, and M. Ernzerhof. Phys. Rev. Lett., 78:1396, 1996. * [40] H.J. Monkhorst and J.D. Pack. Phys. Rev. B, 13:5188, 1976. * [41] J.D. Gale and A.L. Rohl. Molecular Simulation, 29:291, 2003. * [42] C. Kittel. Introduction to the Solid State Physics. Wiley, New York, 1966. * [43] G. Simmons and H. Wang. Single Crystal Elastic Constants and Calculated Aggregated Properties. MIT Press, Cambridge, 1971. * [44] P.J. Craievich, J.M. Sanchez, R.E. Watson, and M. Weinert. Phys. Rev. B, 55(2):787, 1997. * [45] M.I. Baskes and R.A. Johnson. Modelling Simul. Mater. Sci. Eng., 2:147, 1994. * [46] K. H. J. Buschow. Journal of the Less-Common Metals, 85:221, 1982. * [47] J.K. Pargeter and W. Hume-Rothery. Journal of the Less-Common Metals, 12:366, 1967. * [48] C.J. Smithell. Smithells Metals Reference Book. Butterworths, London, 1983. * [49] F.B. Hildebrand. Introduction to Numerial Analysis. McGraw-Hill, 1965. * [50] D. Holland-Moritz, T. Schenk, R. Bellissent, V. Simonet, K. Funakoshi, J.M. Merino, T. Buslaps, and S. Reutzel. J. Non-Cryst. Solids, 312-314:47, 2002. * [51] G.M. Bhuiyan, M. Silbert, and M.J. Stott. Phys. Rev. B, 53:636, 1995. * [52] X.J. Han, J.Z. Wang, M. Chen, and Z.Y. Guo. J. Phys.: Condens. Matter, 16:2565, 2004. * [53] Y Kita, J.B. Van Zydveld, Z. Morita, and T. Iida. J. Phys.: Condens. Matter, 6:811, 1994. * [54] G. Kresse and J. Hafner. Phys. Rev. B, 48:13115, 1993. * [55] T. Schenk, D. Holland-Moritz, V. Simonet, R. Bellissent, and D. M. Herlach. Phys. Rev. Lett., 89:075507, 2002. * [56] J.D. Honeycutt and H.C. Andersen. J. Phys. Chem., 91:4950, 1987. * [57] H. Jonsson and H.C. Andersen. Phys. Rev. Lett., 60:2295, 1988. * [58] T.H. Kim and K.F. Kelton. J. Chem. Phys., 126:054513, 2007. * [59] N Jakse and A Pasturel. Phys. Rev. Lett., 91:195501, 2003. * [60] J.L. Rodriguez-Lopez, F. Aguilera-Granja, K. Michaelian, and A. Vega. J. of Alloys and Compounds, 369:93, 2004. * [61] A.K. Shurin, P.I. Kripyakevich, and E.I. Gladyshevskii. Kristallografiya, 10(3):414, 1965. * [62] Yu.B. Kuz’ma, V.Ya. Markiv, Yu.V. Voroshilov, and R.V. Skolozdra. Izvestiya Akademii Nauk SSSR, Neorganicheskie Materialy, 2:259, 1966\. * [63] A.R. Miedema, P.F. de Châtel, and F.R. de Boer. Physica B, 100:1, 1980. * [64] Yu.A. Babanov, A.F. Sidorenko, A.V. Ryazhkin, V.R. Shvetsov, J. Moessinger, and H. Kronmueller. Nucl. Inst. and Meth. in Phys. Res. A, 405:400, 1998. * [65] B.K Vainshtein, V.M. Fridkin, and V.L. Indenbom. Structure of Crystals, volume 2 of Modern Crystallography. Springer, 3rd edition, 2000. * [66] M. Jamet, V Dupuis, P. Mélinon, G. Guiraud, A. Pérez, W. Wernsdorfer, A. Traverse, and B. Baguenard. Phys. Rev. B, 62:493, 2000. * [67] D.R. Nelson and F. Spaepen. volume 42, page 1. Academic, Boston, 1989.	arxiv-papers	2008-09-01T10:01:28	2024-09-04T02:48:57.573045	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Pascal Thibaudeau and Julian Gale", "submitter": "Pascal Thibaudeau", "url": "https://arxiv.org/abs/0809.0198" }
0809.0210	# An analytical formulation for $\phi^{4}$ field-potential dynamics Arash Ghahraman ar-gh51@stu-mail.um.ac.ir Kurosh Javidan javidan@um.ac.ir Department of physics, Ferdowsi university of Mashhad 91775-1436 Mashhad Iran ###### Abstract An analytical model for adding a space dependent potential to the $\phi^{4}$ field is presented, by constructing a collective coordinate for solitary solution of this model. Interaction of the $\phi^{4}$ field with a delta function potential barrier and also delta function potential well is investigated. Most of the characters of the interaction are derived analytically while they are calculated by other models numerically. We will find that the behaviour of the solitary solution is like a point particle ’living’ under the influence of a complicated potential which is a function of the field initial conditions and the potential parameters. ## 1 Introduction Dynamical evolution of a field in the presence of an external potential, in which case the parameters of the model are functions of space is an important phenomenon from the mathematical point of view and also because of its applications. An external potential can be added to the equation of motion as perturbative terms [1, 2]. These effects also can be taken into account by making some parameters of the equation of motion to be function of space or time [3, 4]. The external potential also can be added to the field through the metric of background space-time [5, 6]. This method is mainly suitable for nonlinear field theories contain solitonic solutions. As is well known, when waves scatter on a potential, they can be partly reflected and partly transmitted. For the fields with solitonic solutions, the situation is more complicated as solitons cannot split and thus must either bounce, pass through or become trapped inside the potential. This behaviour is very sensitive to the value of all the parameters of the model as well as to the initial conditions for the scattering. Most of the researches are in base of numerical studies because such these systems are generally non-integrable. So it is clear that we need suitable models with analytic solutions to test the validity of such phenomenon and predict their behaviour. In this paper an analytical model for the interaction of $\phi^{4}$ field with an external potential is presented. This method can be used for other field theories and from this viewpointin, we will present an example for explaining the method. However the results are perfectly valid for the $\phi^{4}$ field theory. So a model for the $\phi^{4}$ field in an space dependent potential is presented in section 2. The analytic model is introduced and will be solved in section 3. The results for the field-potential barrier system are presented in section 4. In section 5 field-potential well system is discussed. Some conclusion and remarks will be presented in section 6. ## 2 $\phi^{4}$ field and a space dependent potential The general form of the action in an arbitrary metric is: $I=\int{{\cal L}(\phi,\partial_{\mu}\phi)\sqrt{-g}d^{n}xdt}$ (1) where ”g” is the determinant of the metric $g^{\mu\nu}(x)$. Energy density of the ”field + potential” can be found by varying ”both” the field and the metric [7]. For the lagrangian of the form ${\cal L}=\frac{1}{2}\partial_{\mu}\phi\partial^{\mu}\phi-U(\phi)$ (2) the equation of motion becomes [7, 8] $\frac{1}{\sqrt{-g}}\left(\sqrt{-g}\partial_{\mu}\phi\partial^{\mu}\phi+\partial_{\mu}\phi\partial^{\mu}\sqrt{-g}\right)+\frac{\partial U(\phi)}{\partial\phi}=0$ (3) One can add a space dependent potential to the lagrangian of the system by introducing a suitable nontrivial metric for the back ground space-time, without missing the topological boundary conditions [6, 7]. In other words, the metric carries the introduced potential. The suitable metric in the presence of a weak potential $v(x)$ is [5, 6, 7]: $g_{\mu\nu}(x)\cong\left(\begin{array}[]{clrr}1+V(x)&0\\\ 0&-1\end{array}\right)$ (4) The equation of motion (3) (describes by Lagrangian (2)) in the background space-time (4) is $\left(1+V(x)\right)\frac{\partial^{2}\phi}{\partial t^{2}}-\frac{\partial^{2}\phi}{\partial x^{2}}-\frac{1}{2\left\|1+V(x)\right\|}\frac{\partial V(x)}{\partial x}\frac{\partial\phi}{\partial x}+\frac{\partial U(\phi)}{\partial\phi}=0$ (5) For the $\phi^{4}$ model, we have $U(\phi)=\left(\phi^{2}-1\right)^{2}$. The $\phi^{4}$ model has solitary wave solution $\phi(x,t)=tanh\left(\sqrt{2}\frac{x-X(t)}{\sqrt{1-v^{2}}}\right)$ (6) where $X(t)=x_{0}-vt$ and $v$ is the velocity of the solitary wave. By inserting the solution (6) in the lagrangian (2) and using the metric (4), with adiabatic approximation [1, 2] we have ${\cal L}=\sqrt{1+v(x)}\left(\left(1+v(x)\right)\dot{X}^{2}-2\right)sech^{4}\left(\sqrt{2}\left(x-X(t)\right)\right)$ (7) For the weak potential $v(x)$ (7) becomes ${\cal L}\approx\left(\left(1+\frac{3}{2}v(x)\right)\dot{X}^{2}-2\left(1+\frac{1}{2}v(x)\right)\right)sech^{4}\left(\sqrt{2}\left(x-X(t)\right)\right)$ (8) ## 3 Collective coordinate variable The center of a soliton can be considered as a particle, if we look at this variable as a collective coordinate. The collective coordinate could be related to the potential by using the above model. This model is able to give us an analytic solution for most of the features of the soliton-potential system. For the lagrangian (8) X(t) remains as a collective coordinate if we integrate (8) over variable x $L=\int{{\cal L}dx}=\frac{2\sqrt{2}}{3}\dot{X}^{2}-\frac{4\sqrt{2}}{3}+\left(\frac{3}{2}\dot{X}^{2}-1\right)\int{sech^{4}\left(\sqrt{2}\left(x-X(t)\right)\right)v(x)dx}$ (9) If we take the potential $v(x)=\epsilon\delta(x)$ then (9) becomes $L=\frac{2\sqrt{2}}{3}\dot{X}^{2}-\frac{4\sqrt{2}}{3}+\left(\frac{3}{2}\dot{X}^{2}-1\right)\epsilon sech^{4}\left(\sqrt{2}X(t)\right)$ (10) The equation of motion for the variable X(t) is derived from (10) $\left(\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}X\right)\right)\ddot{X}-\sqrt{2}\epsilon tanh(\sqrt{2}X)sech^{4}\left(\sqrt{2}X\right)\left(3\dot{X}^{2}+2\right)=0$ (11) The above equation shows that the peak of the soliton moves under the influence of a complicated force which is a function of its position and its velocity. If $\epsilon>0$ we have a barrier and $\epsilon<0$ creates a potential well. Fortunately equation (11) has an exact solution as follows $\frac{3\dot{X}^{2}+2}{3\dot{X}_{0}^{2}+2}=\frac{\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}\left(X_{0}\right)\right)}{\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}\left(X\right)\right)}$ (12) where $X_{0}$ and $\dot{X}_{0}$ are initial position and initial velocity respectively. The energy of the soliton in the presence of the potential $v(x)=\epsilon\delta(x)$ becomes $E=\left(\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}X\right)\right)\dot{X}^{2}+\frac{4\sqrt{2}}{3}+\epsilon sech^{4}\left(\sqrt{2}X\right)$ (13) Subtituting $\dot{X}$ in (13) from (12) shows that the energy is a function of initial conditions $X_{0}$ and $\dot{X}_{0}$ only. Therefore the energy of the sytem is conserved. Some features of the soliton-potential dynamics can be investigated using equations (12) and (13) analytically which are discussed in the following. ## 4 Potential barrier Suppose that a potential barrier of the height $\epsilon$ is located at the origin. When the soliton is far from the center of the potential ($X\rightarrow\infty$) (13) reduces to $E=\frac{2\sqrt{2}}{3}\dot{X}_{0}^{2}+\frac{4\sqrt{2}}{3}$. It is the energy of a particle with a mass of $\frac{4\sqrt{2}}{3}$ and a velocity of $\dot{X}_{0}$. A soliton with a low velocity reflects back from the barrier and a high energy soliton climbs over the barrier and passes over it. So we have a critical value for the velocity of the soliton which separates these two situations. The energy of a soliton in the origin (X=0) comes from (13)$E(X=0)=\left(\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}\right)\dot{X}_{0}^{2}+\frac{4\sqrt{2}}{3}+\epsilon$. The minimum energy for a soliton in this position is $E=\frac{4\sqrt{2}}{3}+\epsilon$. On the other hand, a soliton which comes from the infinity with initial velocity $v_{c}$ has the energy of $E\left(X=\infty\right)=\frac{2\sqrt{2}}{3}v_{c}^{2}+\frac{4\sqrt{2}}{3}$. Therefore it can pass through the barrier if $v_{c}>\sqrt{\frac{3\sqrt{2}}{4}\epsilon}$. The same result is derived by substituting $\dot{X}=0$, $\dot{X}_{0}=v_{c}$, $X_{0}=\infty$ and $X=0$ in (12). If the soliton is located at some position like $X_{0}$ (which is not necessary infinity) the critical velocity will not be $\sqrt{\frac{3\sqrt{2}}{4}\epsilon}$. The soliton can pass over the barrier if the soliton energy is greater than the energy of a static soliton at the top of the barrier. So a soliton in the initial position $X_{0}$ with initial velocity of $\dot{X}_{0}$ has the critical initial velocity if its velocity becomes zero at the top of the barrier $X=0$. Consider a soliton with initial conditions of $X_{0}$ and $\dot{X}_{0}$. If we set $X=0$ and $\dot{X}=0$ in equation (12) then $v_{c}=\dot{X}_{0}$. Therefore we have $v_{c}=\sqrt{\frac{\epsilon\left(1-sech^{4}\left(\sqrt{2}X_{0}\right)\right)}{\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}x_{0}\right)}}$ (14) When the initial velocity is less than the $v_{c}$ then there exists a return point in which the velocity of the soliton is zero. For this situation we have $\left(\frac{3}{2}\dot{X}_{0}^{2}+1\right)\left(\frac{2\sqrt{2}}{3}+\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}X_{0}\right)\right)-\frac{2\sqrt{2}}{3}=\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}X_{stop}\right)$ (15) Therefore this model predicts that $sech^{4}\left(\sqrt{2}X_{stop}\right)$ is proportional to $sech^{4}\left(\sqrt{2}X_{0}\right)$. Also (15) shows that $sech^{4}\left(\sqrt{2}X_{stop}\right)$ is proportional to $\frac{1}{\epsilon}$. A soliton with initial conditions $X_{0}$ and $\dot{X}_{0}$ will go to the infinity after the interaction with a potential barrier. The final velocity of the soliton after the interaction is $\dot{X}=\sqrt{\dot{X}_{0}^{2}+\frac{3\epsilon}{2\sqrt{2}}\left(\frac{3\dot{X}_{0}^{2}}{2}+1\right)sech^{4}\left(\sqrt{2}X_{0}\right)}$ (16) which is greater than the initial velocity $\dot{X}_{0}$. Equations (12) and (13) show that the soliton finds its initial velocity after the interaction when it reaches its initial position. This means that the interaction is completely elastic. Numerical simulation are in agreement with above analytical results [7, 9, 10]. ## 5 Soliton-well system The soliton-well system is very interesting problem. Suppose a particle moves toward a frictionless potential well. It falls in the well with increasing velocity and reaches the bottom of the well with its maximum speed. After that, it will climb the well with decreasing velocity and finally pass through the well. Its final velocity after the interaction is equal to its initial speed. Changing $\epsilon$ to $-\epsilon$ in (12) changes potential barrier to potential well. The solution for the system is $\frac{3\dot{X}^{2}+2}{3\dot{X}_{0}^{2}+2}=\frac{\frac{2\sqrt{2}}{3}-\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}\left(X_{0}\right)\right)}{\frac{2\sqrt{2}}{3}-\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}\left(X\right)\right)}$ (17) There is not a critical velocity for a soliton-well system, but an escape velocity can be defined. A soliton with initial position $X_{0}$ reaches the infinity with a zero final velocity if its initial velocity is $\dot{X}_{escape}=\sqrt{\frac{\epsilon sech^{4}\left(\sqrt{2}X_{0}\right)}{\frac{2\sqrt{2}}{3}-\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}X_{0}\right)}}$ (18) In other words, a soliton which is located in the initial position $X_{0}$ can escape to infinity if its initial velocity $\dot{X}_{0}$ is greater than the escape velocity $\dot{X}_{escape}$. Consider a potential well with the depth of $\epsilon$ and a soliton at the initial position $X_{0}$ which moves toward the well with the initial velocity $\dot{X_{0}}$ smaller than the $\dot{X}_{escape}$. The soliton interacts with the potential and reaches a maximum distance $X_{max}$ from the center potential with a zero velocity and then come back toward the well. The soliton oscillates around the well with the amplitude $X_{max}$. The required initial velocity to reach $X_{max}$ is found from (17) as $\dot{X}_{0}=\sqrt{\frac{\epsilon\left(sech^{4}\left(\sqrt{2}X_{max}\right)-sech^{4}\left(\sqrt{2}X_{0}\right)\right)}{\frac{2\sqrt{2}}{3}-\frac{3\epsilon}{2}sech^{4}\left(\sqrt{2}X_{max}\right)}}$ (19) If the initial velocity is lower than the escape velocity the soliton oscillates around the well. The period of oscillation can be calculated numerically using equation (17). ## 6 Conclusion and Remarks A model for the $\phi^{4}$ field-potential interaction has been presented. The model predicts a critical velocity for the soliton-barrier interaction as a function of initial conditions of the field and the potential characters. The model predicts specific relations between some functions of initial conditions and other functions of final state of the field during the interaction. An escape velocity has been derived for the field-well system. This model is able to explain most of the features of the system analytically. This model can be used for prediction the results of other field theories beside the $\phi^{4}$ model. ## References * [1] Kivshar Y S, Fei Z and Vasquez L 1991 Phys. Rev. Lett 67 1177 * [2] Fei Z, Kivshar Y S and Vasquez L 1992 Phys. Rev.A46 5214 * [3] Piette B, Zakrzewski W J and Brand J 2005 J. Phys. A: Math. Gen.38 10403-10412 * [4] Piette B. Zakrzewski W.J.2007 J. Phys. A: Math. Theor.40 No 2, 329-346 * [5] Kalbermann G 1999 Phys. Lett A252 37-42 * [6] Javidan K and Sarbishaei M 2001 Indian J. Physics B75 (5) 413-418 * [7] Javidan K. 2006 J. Phys. A: Math. Gen. 39 No 33 10565-10574 * [8] Felsager B. (1998) Geomrtry, Particles, and fields (New York: Springer-Verlag) pp 434-437 * [9] Al-Alawi Jassem H. , Zakrzewski W.J. (2007) J. Phys. A 40, 11319-1131 * [10] Javidan K arXiv:0807.1618	arxiv-papers	2008-09-01T12:01:52	2024-09-04T02:48:57.578867	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Arash Ghahraman and Kurosh Javidan", "submitter": "Kurosh Javidan", "url": "https://arxiv.org/abs/0809.0210" }
0809.0234	Present address: ]Westinghouse Electric Germany GmbH, Dudenstrasse 44, D-68167 Mannheim. Present address: ]Physik Department E12, Technische Universtät München, D-85748 Garching # Cross section measurements of $\alpha$-induced reactions on 92,94Mo and 112Sn for $p$-process studies W. Rapp [ I. Dillmann iris.dillmann@ph.tum.de [ F. Käppeler Institut für Kernphysik, Forschungszentrum Karlsruhe, Postfach 3640, D-76021 Karlsruhe, Germany U. Giesen H. Klein Physikalisch-Technische Bundesanstalt, Bundesallee 100, D-38116 Braunschweig, Germany T. Rauscher Departement für Physik, Universität Basel, Klingelbergstrasse 82, CH-4056 Basel D. Hentschel S. Hilpp Institut für Nukleare Entsorgung, Forschungszentrum Karlsruhe, Postfach 3640, D-76021 Karlsruhe, Germany ###### Abstract The 92Mo($\alpha,n$)95Ru, 94Mo($\alpha,n$)97Ru, and 112Sn($\alpha,\gamma$)116Te cross sections were measured at the upper end of the $p$-process Gamow window between 8.2 MeV and 11.1 MeV. Our results are slightly lower than global Hauser-Feshbach calculations from the code NON- SMOKER, but still within the uncertainty of the prediction. The 112Sn($\alpha,\gamma$)116Te cross section agrees well with a recently measured thick-target cross section in the same energy range. For the 92,94Mo($\alpha,n$) reactions the present data close to the reaction thresholds could eliminate previous uncertainties within a factor of 20, and we can present now useful comparisons to statistical model calculations with different optical potentials. nucleosynthesis, $p$ process, $\alpha$-induced cross sections ###### pacs: 12.40.Ee, 25.55.-e, 26.30.-k, 27.60.+j ††preprint: APS/123-QED ## I Introduction A “$p$ process” was postulated to produce 35 stable but rare isotopes between 74Se and 196Hg on the proton-rich side of the valley of stability. Unlike the remaining 99% of the heavy elements beyond iron these isotopes cannot be created by (slow or rapid) neutron captures Burbidge et al. (1957), and their solar and isotopic abundances are 1-2 orders of magnitude lower than the respective $s$\- and $r$-process nuclei Anders and Grevesse (1989); Rosman and Taylor (1998). However, so far it seems to be impossible to reproduce the solar abundances of all $p$ isotopes by one single process. In current understanding, several (independently operating) processes seem to contribute. The largest fraction of $p$ isotopes is created in the “$\gamma$ process” by sequences of photodissociations and $\beta^{+}$ decays Woosley and Howard (1978, 1990); Rayet et al. (1990); Arnould and Goriely (2003). This occurs in explosive O/Ne burning during SNII explosions and reproduces the solar abundances for the bulk of $p$ isotopes within a factor of $\approx$3 Rayet et al. (1990, 1995). The SN shock wave induces temperatures of 2-3 GK in the outer (C, Ne, O) layers, sufficient for triggering the required photodisintegrations. More massive stellar models (M$\geq$20 M⊙) seem to reach the required temperatures for efficient photodisintegration already at the end of hydrostatic O/Ne burning Rauscher et al. (2002). Historically, the $p$ process was thought to proceed via proton captures but today they are found to play no role since the required amount of free protons is not available in the Ne and O layers within the $p$-process timescale of a few seconds and proton captures are too slow for elements with large $Z$. Therefore, the term ”$\gamma$ process” is also often used synonymously because seed nuclei pre- existing in the stellar plasma are photodissociated and more proton-rich isotopes are produced, initially by $(\gamma,~{}n)$ reactions. When ($\gamma,~{}p$) and ($\gamma,~{}\alpha$) reactions become comparable or faster than neutron emission within an isotopic chain, the reaction path branches out and feeds nuclei with lower charge number $Z$. The decrease in temperature after passage of the shock leads to a freeze-out via neutron captures and mainly $\beta^{+}$ decays, resulting in the typical $p$-process abundance pattern with maxima at 92Mo ($N$=50) and 144Sm ($N$=82). However, the $\gamma$-process scenario suffers from a strong underproduction of the most abundant $p$ isotopes, 92,94Mo and 96,98Ru, due to lack of seed nuclei with $A$$>$90\. For these missing abundances, alternative processes and sites have been proposed, either related to strong neutrino fluxes in the deepest ejected layers of a SNII ($\nu p$ process Fröhlich et al. (2006)), or rapid proton-captures in proton-rich, hot matter accreted on the surface of a neutron star ($rp$ process Schatz et al. (1998, 2001)). Modern, self-consistent studies of the $\gamma$ process have problems to synthesize $p$ nuclei in the regions $A<124$ and $150\leq A\leq 165$ Rauscher et al. (2002). It is not yet clear whether the observed underproductions are only due to a problem with astrophysical models or also with the nuclear physics input, i.e. the reaction rates used. Thus, the reduction of uncertainties in nuclear data is strictly necessary for a consistent understanding of the $p$ process. Experimental data can improve the situation in two ways, either by directly replacing predictions with measured cross sections in the relevant energy range or by testing the reliability of predictions at other energies when the relevant energy range is not experimentally accessible. Accordingly, model calculations of the $p$ process require large nuclear reaction networks involving about 1800 isotopes and more than ten thousand reaction rates. Since the largest fraction of such a network contains proton- rich, unstable isotopes, which are not accessible for cross section measurements with present experimental techniques, almost all of these rates have to be determined theoretically by means of the statistical Hauser- Feshbach (HF) formalism Hauser and Feshbach (1952). Hence, there is no alternative to employing statistical model calculations, e.g. those performed with the codes NON-SMOKER Rauscher and Thielemann (2000, 2001) or MOST Goriely (2005). Comparison with experimental results has shown that theoretical HF predictions with global parameter sets reproduce the neutron- and proton- induced cross sections within a factor of two or even better. This deviation increases for $\alpha$-induced reactions. It has been noted that the difficulties in calculating these rates are due to the poor knowledge of the $\alpha$-optical potential at astrophysically relevant energies Mohr et al. (1997); Demetriou et al. (2002). The optical potential governs the transmission coefficient in the HF model. Additional input to this model are nuclear level densities and particle separation energies. The latter data are well defined since nuclear masses in the $p$-process network are experimentally known. It is only one decade ago that charged particle-induced reaction rates at $p$-process energies ($E_{p}$$\approx$1-10 MeV, $E_{\alpha}$$\approx$5-15 MeV) became subject of experimental efforts (for a list of reactions, see KADoNiS (2005)). For example, a series of $\alpha$-scattering experiments at energies below the Coulomb barrier helped to improve the optical $\alpha$-nucleus potential Demetriou et al. (2002); Avrigeanu et al. (2003); Fülöp et al. (2001); Rauscher (2003a, b). However, firm conclusions are still difficult to draw due to the lack of experimental cross sections. Nevertheless, extrapolations to the low astrophysical energies remain necessary since the Coulomb barrier hampers scattering experiments in the astrophysically relevant energy range. Scattering data within a confined energy range can be described equally well by different optical potentials with different extrapolation behavior. This ambiguity, which is called the family problem, makes extrapolations very uncertain McFadden and Satchler (1966); Fülöp et al. (1996); Mohr et al. (1997); Galaviz et al. (2005). These uncertainties can be avoided if experimental cross sections are available for comparison with HF predictions in or close to the astrophysical Gamow window. Up to now only five $(\alpha,\gamma)$ measurements with relevance for the $p$ process were performed within the Gamow window between 5 and 15 MeV: 70Ge Fülöp et al. (1996), 96Ru Rapp et al. (2002), 106Cd Gyürky et al. (2006), 112Sn Özkan et al. (2007), and 144Sm Somorjai et al. (1998). With exception of the latter, the first four experiments agree within a factor of three with the NON-SMOKER prediction using a global parameter set with the McFadden-Satchler $\alpha$ potential McFadden and Satchler (1966). For the $(\alpha,n)$ particle-exchange reactions a much larger amount of data exists, but almost all of these measurements are restricted to the upper end of the Gamow window. It was noted in Rapp et al. (2006); Rauscher (2006) that especially in the heavy mass region (beyond $N$=82) the knowledge of $(\alpha,x)$ cross sections is crucial, whereas at lower masses the $p$-induced reactions have more importance. However, measurements of $\alpha$-induced reactions are mainly limited – apart from some exceptions for $(\alpha,n)$ cross sections – to the lower mass region below 144Sm. In an attempt to constrain the $\alpha$+nucleus optical potential at low energy, it has been suggested to use ($n,\alpha$) reactions and some reactions around $A\approx 147-152$ have been studied in this way Gledenov et al. (2000); Koehler et al. (2000, 2004). Finally, it has to be emphasized that with a few exceptions ($\alpha,~{}\gamma$) and ($p,~{}\gamma$) data in the relevant energy range are more useful for a direct application in reaction networks than ($\gamma,~{}\alpha$) or ($\gamma,~{}p$) data despite of the fact that the $p$ process proceeds via photodisintegration reactions. This may seem puzzling at first sight but has two reasons. First, reactions in the stellar plasma involve thermally excited targets and not only targets in the ground state as in the laboratory. Measured rates have to be corrected for this effect. It can be shown that the thermal effects are much more pronounced in the photodisintegration than in the capture channel. Thus the measured rate is closer to the actual rate when considering capture reactions. The reverse rate can then be derived applying detailed balance. Second, in order to avoid inconsistencies, forward and reverse rates for a given reaction in a reaction network have to stem from the same source, related to each other by detailed balance. Because of the exponential factor $\exp(-Q_{\mathrm{forward}})$ in the expression for the relation of reverse to forward rate (see, e.g., Rauscher and Thielemann (2000)), it is preferrable to have data for the direction of positive $Q$ value to limit numerical errors in the simulation. In most cases this is the capture direction, exclusively in the case of proton captures close to stability. However, in the intermediate to heavy mass region the $p$ process also includes $\alpha$ emitters with negative $\alpha$ capture $Q$ values. In this work we study ($\alpha,n$) reactions on light Mo isotopes and $\alpha$ capture on 112Sn, motivated by the need to better understand the above described deficiencies in the $p$ process models. Previous investigations of astrophysical relevance were restricted to $\alpha$ scattering on 92Mo Fülöp et al. (2001), 106Cd Kiss et al. (2006), and on 112Sn Galaviz et al. (2005) attempting to derive an improved optical potential for these cases. The experimental setup for the measurement of $\alpha$-induced cross sections is described in Sec. II. The resulting cross sections and a comparison with previous experimental results for 92Mo($\alpha,n$)95Ru, 94Mo($\alpha,n$)97Ru, and 112Sn($\alpha,\gamma$)116Te are given in Sec. III. The impact of using different optical $\alpha$+nucleus potentials in theoretical calculations for these reactions is investigated in Sec. IV. ## II Cross section measurements All cross section measurements have been carried out at the cyclotron of the Physikalisch-Technische Bundesanstalt (PTB) in Braunschweig/ Germany Brede et al. (1980) with the activation technique by irradiation of thin sample layers and subsequent counting of the induced activity. A detailed description of the measurement technique and the data analysis can be found in Ref. Rapp et al. (2002). ### II.1 Preparation and characterization of samples The sample material was deposited on 0.4 mm thick Ta backings, which were thick enough to stop the $\alpha$-beam completely to ensure a reliable charge collection. Background reactions are efficiently suppressed by the high Coulomb barrier of Ta because $\alpha$ capture on 181Ta leads to the stable product nucleus 185Re. The 181Ta$(\alpha,n)$184Re channel opens at 9.86 MeV but does not interfere with our measurements. The contribution from the extremely rare isotope 180Tam (natural elemental abundance 0.012% Rosman and Taylor (1998)) could be neglected. The Mo samples were produced by sputtering natural metallic Mo in a pure Ar atmosphere. These samples were 10 mm in diameter and exhibited an area density between 5$\times$1017 and 3$\times$1018 Mo atoms/cm2, corresponding to layer thicknesses between 100 nm and 600 nm. The 112Sn samples were prepared from 10 mg of enriched metal powder containing 98.9% of 112Sn. This batch was dissolved in 30% hydrochloric acid and diluted with 90% acetone. The solution was then deposited on the Ta backings, where uniform SnCl2 layers with an area density between 6.2 and 9.8$\times$1017 Sn atoms/cm2 (layer thicknesses $\approx$300-450 nm) were obtained after evaporation of the acetone. The adhesion of the layers was improved by soft sand-blasting of the Ta backings before the deposition. The masses of the Mo and Sn samples were determined by X-ray fluorescence analysis using a crystal spectrometer (Siemens SRS 3000), which was operated with a LiF (100) crystal and a rhodium anode. The fluorescence yield was measured with a gas counter and a NaI detector. The efficiency of the spectrometer was calibrated by the radiation emitted from an empty backing and by nine well defined Mo and Sn reference samples. The latter were prepared from standard solutions of (NH4)2MoO4 and SnCl2, respectively. The sample mass was derived from the measured Kα1 and Kβ1 X-ray lines. In case of the Sn sample an additional Ta diaphragm 10 mm in diameter was placed on top of the sample to restrict the measurement exactly to the area covered in the irradiations with the $\alpha$ beam. While the statistical uncertainties were smaller than 0.3%, the definition of the sample mass was affected by systematic uncertainties of 4.0% and 4.6% for the Mo and the Sn samples, respectively. ### II.2 Irradiations and $\gamma$ counting A sketch of the beamline setup at the PTB is shown in Fig. 1. The activation chamber is designed as a Faraday cup, and the charge deposited on the sample was recorded in small time steps by a current integrator for later correction of beam fluctuations. Secondary electrons were suppressed by a negatively charged (US=-150 V) diaphragm at the entrance of the activation chamber. Figure 1: Experimental setup at the PTB Braunschweig. For each energy step the $\alpha$-beam spot was adjusted by means of a quartz window so that the sample was illuminated as complete and homogeneously as possible by wobbling of the $\alpha$-beam. The energy of the $\alpha$ beam was defined within $\pm$25 keV uncertainty by means of the field calibration of two analyzing magnets as well as by a time-of-flight measurement of the particle velocity Böttger (2002). The effective $\alpha$-energy $E^{\rm{eff}}$ was calculated from the expression 4.99 in Iliadis (2007) $E^{\rm{eff}}=E_{\rm{c.m.}}-\Delta+\Delta\cdot\left(-\frac{\sigma_{2}}{\sigma_{1}-\sigma_{2}}+\left[\frac{\sigma_{1}^{2}+\sigma_{2}^{2}}{2(\sigma_{1}-\sigma_{2})^{2}}\right]^{1/2}\right).$ (1) where $\Delta$ is the target thickness calculated with the Monte Carlo program SRIM 2003 Ziegler and Biersack (2003), $E_{\rm{c.m.}}$ the respective center- of-mass energy, and $\sigma_{1}$, $\sigma_{2}$ the measured cross sections of two neighboring points. As can be seen from Eq. 1 the error bars of $E^{\rm{eff}}$ become asymmetric, with the smaller component in positive direction due to the correction factor in brackets (see Tables 3, 4, and 5). Table 1: Decay properties of product nuclei and corrections for $\gamma$-ray counting Product \| $t_{1/2}$ \| $E_{\gamma}$ \| $I_{\gamma}$ \| Reference \| Summing corrections [%] ---\|---\|---\|---\|---\|--- nucleus \| \| [keV] \| per decay \| \| $\gamma$ Cascades \| X Rays 95Ru \| (1.643 $\pm$ 0.014) h \| 336.4 \| 0.702 $\pm$ 0.005 \| Burrows (1993) \| 7 $\pm$ 1 \| 11.2 $\pm$ 0.4 \| \| 626.8 \| 0.178 $\pm$ 0.005 \| \| 7 $\pm$ 1 \| “ 97Ru \| (2.9 $\pm$ 0.1) d \| 215.7 \| 0.856 $\pm$ 0.013 111No uncertainty given in literature. Assumed uncertainty 1.6%, as for the 324.4 keV line. \| Artna-Cohen (1993) \| $<$0.2 \| 12.4 $\pm$ 1.5 \| \| 324.4 \| 0.1079 $\pm$ 0.0017 \| \| “ \| “ 116Te \| (2.49 $\pm$ 0.04) h \| 93.7 \| 0.331 $\pm$ 0.021 \| Blachot (2001) \| $<$0.2 \| 15.6 $\pm$ 1.7 During the activations, the samples were irradiated with average $\alpha$-beam currents of 5 $\mu$A for periods between 20 minutes and 14 hours, depending on isotope and cross section. For the Sn samples the waiting time between the end of activation and the start of activity counting was extended to one hour to eliminate the 38K decay from the concomitant 35Cl($\alpha,n$)38K reaction ($t_{1/2}$= 7.5 min). The induced $\gamma$ activities were measured with a calibrated HPGe detector. Natural backgrounds were suppressed by a shielding consisting of an inner layer of 5 mm copper surrounded by 10 cm of lead. The samples were analyzed as described in Ref. Rapp et al. (2002) using the $\gamma$-ray energies and intensities of the characteristic lines listed in Table 1. In cases, where two $\gamma$-ray lines could be used in data analysis, the results were averaged by weighting with the respective transition probabilities. Summing corrections due to the coincident detection of $\gamma$-cascades were evaluated by means of simulations with the Monte Carlo code CASC Jaag (1993) using the decay schemes of the corresponding product nuclei 95Ru, 97Ru, and 116Te Burrows (1993); Artna-Cohen (1993); Blachot (2001). The resulting corrections for $\gamma$ summing are given in Table 1 together with the summing corrections due to the coincident detection of X rays. The analysis of the induced activity in the 112Sn samples via the 93.7 keV transition in the decay of 116Te was complicated by a constant background activity of (0.031$\pm$0.003) cts/s, corresponding presumably to the Po-Kβ-line, as well as by the characteristic 93.3 keV $\gamma$ line from the decay of 67Ga. The latter background was produced by 63Cu($\alpha,\gamma$) reactions due to a copper contamination of the samples. The background component from the 67Ga decay could be corrected by means of the associated 67Ga line at 184.6 keV. ### II.3 Uncertainties The systematic uncertainties (Table 2) are governed by the target thickness and the detector efficiency. For 97Ru, the uncertainty in the half-life gives rise to an additional contribution of 3.4%. The same holds for the $\gamma$ intensity of the 93.7 keV line from the decay of 116Te with an uncertainty of 6.3% Blachot (2001). The overall uncertainties are 5.3% and 6.3% for the 92Mo$(\alpha,n)$ and 94Mo$(\alpha,n)$, and 8.7% for the 112Sn$(\alpha,\gamma)$116Te cross sections. Table 2: Systematic uncertainties (in %). Source of uncertainty \| 92Mo($\alpha,n$) \| 94Mo($\alpha,n$) \| 112Sn($\alpha,\gamma$) ---\|---\|---\|--- Target thickness \| 4.0 \| 4.0 \| 4.6 Efficiency \| 3.0 \| 3.0 \| 3.0 Beam current \| 1.0 \| 1.0 \| 1.0 $\gamma$ intensity \| 0.7 \| 0.5 \| 6.3 Decay constants \| 0.9 \| 3.4 \| 1.6 Summing corrections \| 1.1 \| 1.5 \| 1.7 Overall uncertainty \| 5.3 \| 6.3 \| 8.722footnotemark: 2 11footnotetext: Does not include effect of background line (see text). ## III Results and comparison with previous data The results are summarized in Tables 3, 4, and 5, and in Fig. 2 in form of cross sections and astrophysical $S$ factors. The cross sections of all three reactions (Fig. 2) are systematically lower by factors of two to three compared to the global Hauser-Feshbach predictions from NON-SMOKER (dashed line). For 112Sn($\alpha,\gamma$)116Te our data points follow the Hauser-Feshbach dependence within the uncertainties, whereas the results from Özkan et al. Özkan et al. (2007), which were obtained with rather thick targets and represent averaged cross sections over energy bins of 440 to 565 keV, seem to exhibit a slightly different energy trend. Unfortunately, the present measurements at lower energies were severely hampered by an interfering background line from 63Cu$(\alpha,\gamma)$67Ga reactions on the Cu impurity of the samples. This background could be only normalized by a much weaker 67Ga line at 184.6 keV, resulting in additional systematic and statistical uncertainties. For the $(\alpha,n)$ cross sections some older measurements were available for comparison (Fig. 3). The data given by Levkovskij Levkovskij (1991) is reproducing the NON-SMOKER cross sections in both cases perfectly for energies above 13 MeV, but there is a systematic deviation (compared to all other data) below this energy, presumably due to an undetected background component. This problem is particularly apparent at the lowest data point at 8.4 MeV, whereas the 92Mo$(\alpha,n)$95Ru threshold is at 9.39 MeV. A similar problem exists for 94Mo$(\alpha,n)$. The values from Esterlund et al. Esterlund and Pate (1965) for 92Mo$(\alpha,n)$ and Graf et al. Graf and Muenzel (1974) for 92,94Mo$(\alpha,n)$ seem to be low compared to the energy trend below 15 MeV, but agree like the Levkovskij data rather well with the predictions at higher energies. The only data which follow the Hauser-Feshbach trend also at lower energies down to 11.8 MeV and at the same time fits well with our highest data point are the 92Mo$(\alpha,n)$ cross sections of Denzler et al. Denzler et al. (1995). This comparison shows the necessity to measure $(\alpha,n)$ cross sections – especially for $p$-process studies – down to the threshold. Table 3: Cross sections and $S$ factors for 92Mo$(\alpha,n)$95Ru. $E{\rm{}^{eff}_{c.m.}}$ \| Cross section \| $S$ factor ---\|---\|--- [MeV] \| [mbarn] \| $\times$1020[MeV barn] 10.989 ${}^{+0.038}_{-0.039}$ \| 9.26 $\pm$ 0.65 \| 2.13 $\pm$ 0.15 10.981 ${}^{+0.046}_{-0.047}$ \| 10.6 $\pm$ 0.9 \| 2.47 $\pm$ 0.21 10.543 ${}^{+0.033}_{-0.037}$ \| 3.88 $\pm$ 0.33 \| 2.38 $\pm$ 0.20 10.019 ${}^{+0.039}_{-0.043}$ \| 2.53 $\pm$ 0.26 \| 5.39 $\pm$ 0.55 9.638 ${}^{+0.066}_{-0.087}$ \| 1.02 $\pm$ 0.09 \| 5.71 $\pm$ 0.50 9.112 ${}^{+0.036}_{-0.047}$ \| 0.194 $\pm$ 0.015 \| 4.57 $\pm$ 0.35 Table 4: Cross sections and $S$ factors for 94Mo$(\alpha,n)$97Ru. $E{\rm{}^{eff}_{c.m.}}$ \| Cross section \| $S$ factor ---\|---\|--- [MeV] \| [mbarn] \| $\times$1020[MeV barn] 10.997 ${}^{+0.039}_{-0.040}$ \| 20.90 $\pm$ 2.00 \| 4.81 $\pm$ 0.46 10.966 ${}^{+0.070}_{-0.073}$ \| 18.50 $\pm$ 1.50 \| 4.56 $\pm$ 0.37 10.548 ${}^{+0.038}_{-0.044}$ \| 8.14 $\pm$ 0.77 \| 5.06 $\pm$ 0.48 10.034 ${}^{+0.034}_{-0.036}$ \| 5.47 $\pm$ 0.62 \| 11.50 $\pm$ 1.30 9.676 ${}^{+0.037}_{-0.043}$ \| 2.49 $\pm$ 0.25 \| 12.96 $\pm$ 1.30 9.114 ${}^{+0.042}_{-0.056}$ \| 0.58 $\pm$ 0.06 \| 13.92 $\pm$ 1.44 8.572 ${}^{+0.067}_{-0.102}$ \| 0.12 $\pm$ 0.01 \| 14.52 $\pm$ 1.21 8.136 ${}^{+0.061}_{-0.095}$ \| 0.021 $\pm$ 0.002 \| 10.49 $\pm$ 1.00 Table 5: Cross sections and $S$ factors for 112Sn$(\alpha,\gamma)$116Te. $E{\rm{}^{eff}_{c.m.}}$ \| Cross section \| $S$ factor ---\|---\|--- [MeV] \| [$\mu$barn] \| $\times$1023[MeV barn] 11.067 ${}^{+0.042}_{-0.038}$ \| 408 $\pm$ 57 \| 1.10 $\pm$ 0.15 10.615 ${}^{+0.040}_{-0.043}$ \| 295 $\pm$ 43 \| 2.61 $\pm$ 0.38 10.097 ${}^{+0.037}_{-0.045}$ \| 77 $\pm$ 15 \| 2.93 $\pm$ 0.57 9.180 ${}^{+0.036}_{-0.048}$ \| 9.52 $\pm$ 3.68 \| 6.51 $\pm$ 2.52 8.209 ${}^{+0.043}_{-0.064}$ \| 0.48 $\pm$ 0.38 \| 11.78 $\pm$ 9.33 Figure 2: Cross section and $S$ factor for 92Mo$(\alpha,n)$95Ru, 94Mo$(\alpha,n)$97Ru, and 112Sn$(\alpha,\gamma)$116Te in comparison with standard NON-SMOKER predictions (dashed line). The grey area shows the estimated uncertainty of the prediction (factor 10). Figure 3: (Color online) Comparison of previous experiments at higher energies (up to 20 MeV) with our results and with the standard NON-SMOKER predictions (dashed line). The grey area shows the uncertainty of the prediction (factor 10). ## IV Test of the optical $\alpha$+nucleus potential The present experimental $S$ factors were compared to NON-SMOKER Hauser- Feshbach calculations employing different optical $\alpha$+nucleus potentials while keeping the standard inputs for the other parameters. The standard optical potential used in NON-SMOKER Rauscher and Thielemann (2000) is the one by McFadden and Satchler McFadden and Satchler (1966), derived from scattering data above 60 MeV. This is also the potential used in the standard NON-SMOKER predictions shown in Figs. 2 and 3. Another global potential is the one by England et al. England et al. (1982), also fitted to data at higher energies. A more recent global potential, which also included reaction data, was given by Avrigeanu et al. Avrigeanu et al. (2003). Finally, Fröhlich et al. (see Rauscher (2003c), erratum in Rauscher (2003d)) used reaction data at $A\approx 150$ to derive a low-energy potential which also reproduces data at lower masses (e.g., Rauscher (2003c); Gyürky et al. (2006)). The reaction 92Mo$(\alpha$,n)95Ru is mainly sensitive to the optical $\alpha$ potential in the measured energy range. Only up to a few keV above the threshold, the averaged neutron widths are smaller than the $\alpha$ widths and therefore the sensitivity is higher to the neutron widths close to the threshold. As can be seen in Fig. 4, the theoretical results obtained with the different optical potentials display the same energy behavior but with different absolute values. This energy dependence describes that of the data reasonably well. The $S$ factors obtained with the McFadden and Satchler potential have to be divided by 1.8, whereas the ones obtained with the Fröhlich potential are a factor of 1.8 too low. All other potentials used lead to larger deviations. An optical potential for 92Mo+$\alpha$ was derived from scattering data and carefully extrapolated to low energy in Ref. Fülöp et al. (2001). We do not show the results with this potential because they coincide with the Avrigeanu et al. results at the low energy range shown here, whereas they coincide with the McFadden and Satchler $S$ factors at the high end of the energy range. Thus, its energy dependence is slightly different from that of the other potentials and the data. Also the reaction 94Mo$(\alpha,n)$97Ru is almost exclusively sensitive to the $\alpha$ widths and shows only a small dependence on the neutron widths, except close to the threshold. Inspection of Fig. 4 evidently shows that the energy dependence of the data cannot be reproduced by any of the potentials tried here. The peak $S$ factor is shifted to higher energy in the data as compared to the one found in the calculated $S$ factor. The cause of this remains unclear. The importance of the neutron widths does not extend beyond a few tens of keV above the ($\alpha$,n) threshold. The impact of width fluctuation corrections is also confined to a small window above the reaction threshold. Similar to the case above, multiplication of the results obtained with the Fröhlich potential with a factor 1.8 and division of the ones obtained with the McFadden and Satchler potential by 1.7 lead to an improved reproduction of the data. However, because of the distinctly different energy dependence none of these solutions are satisfactory. Figure 4: (Color online) Comparison of $S$ factors obtained with different optical potentials in the NON-SMOKER code for the reactions 92Mo$(\alpha,n)$95Ru and 94Mo$(\alpha,n)$97Ru (see text for details). Figure 5: (Color online) Comparison of $S$ factors obtained with different optical potentials in the NON-SMOKER code for the reaction 112Sn$(\alpha,\gamma)$116Te (see text for details). It may be surprising at first glance that the $S$ factors of the reaction 112Sn$(\alpha,\gamma)$116Te are much more sensitive to the $\alpha$ widths than to the $\gamma$ widths. However, due to the Coulomb barrier, the $\alpha$ widths are smaller by one to several orders of magnitude than the $\gamma$ widths in the energy range covered in the measurement. There are two data sets available, the one obtained in this work and the one by Özkan et al. Özkan et al. (2007). Both of them are compared to theoretical results in Fig. 5. Concerning the theoretical predictions using different optical $\alpha$ potentials it is not surprising that the Özkan et al. data at the high energy range are best described employing the potential I of Demetriou et al. Demetriou et al. (2002). A preliminary version of these data was part of the procedure to derive the Demetriou et al. potential, a global potential for heavy targets. Similar agreement was already found in a study of $\alpha$ scattering on 112Sn Galaviz et al. (2005). In that study, an optical potential was derived from scattering data and then used to compute the 112Sn$(\alpha,\gamma)$116Te $S$ factors. Peculiarly, however, it was found that the derived potential did not reproduce the Özkan et al. reaction data well. Much better agreement was achieved with the Demetriou et al. potential and best agreement was obtained when using the Fröhlich potential, although the latter did not describe the scattering data. Here, we confirm these findings Galaviz et al. (2005). We do not show the result with the optical potential derived in Galaviz et al. (2005) separately because it basically coincides with the one obtained with the McFadden and Satchler potential. Interestingly, our new experimental data seem to indicate a slightly steeper energy dependence than the Özkan data. This yields good agreement (within the error bars) with the prediction making use of the Fröhlich-Rauscher potential (with exception of the data point at 10.615 MeV). The calculation using the Demetriou potential Demetriou et al. (2002) exceeds our data (including error bar) at the three highest energies and is compatible to our data only at the two lowest energies. Comparing to the Özkan et al. data, the Fröhlich-Rauscher potential gives acceptable agreement only below 9.5 MeV and underestimates the $S$ factors above that energy. On the other hand, the Demetriou et al. potential excellently reproduces the Özkan et al. data above 8.5 MeV but strongly overestimates them below that energy. ## V Summary We have measured the $(\alpha,n)$ cross sections of the $p$ nuclei 92,94Mo and the $(\alpha,\gamma)$ cross section of the lightest tin isotope 112Sn. For 94Mo$(\alpha,n)$ no data for comparison at lower energies was available, but for 92Mo our energy trend seems to fit well with the results of Denzler et al. Denzler et al. (1995) starting at 12 MeV. For 112Sn$(\alpha,\gamma)$116Te we can essentially reproduce the thick-target cross sections from Özkan et al. Özkan et al. (2007) although we find slightly lower values above 10 MeV. Additionally, we studied the sensitivity of Hauser-Feshbach predictions performed with the code NON-SMOKER for these reactions. We find that the cross sections and $S$ factors are mostly sensitive to the $\alpha$ widths and thus to the optical $\alpha$+nucleus potential employed in the calculations. The standard, global predictions – utilizing the global potential by McFadden- Satchler McFadden and Satchler (1966) – systematically overestimate the data by a factor of two. The potential of Fröhlich and Rauscher Rauscher (2003c, d) describes well the reaction data of 112Sn$(\alpha,\gamma)$116Te but underestimates the 92,94Mo data by a factor of about two. The energy dependence of 94Mo($\alpha$,n)97Ru cannot be reproduced by the statistical model calculations. ###### Acknowledgements. We thank the operating team O. Döhr, H. Eggestein, T. Heldt, and M. Hoffmann for providing the $\alpha$-beam at the PTB cyclotron. This work was partially supported by the Swiss NSF (grant 2000-113984/1). ## References * Burbidge et al. (1957) E. Burbidge, G. Burbidge, W. Fowler, and F. Hoyle, Rev. Mod. Phys. 29, 547 (1957). * Anders and Grevesse (1989) E. Anders and N. Grevesse, Geochim. Cosmochim. Acta 53, 1997 (1989). * Rosman and Taylor (1998) K. Rosman and P. Taylor, Pure and Appl. Chem. 70, 217 (1998). * Woosley and Howard (1978) S. Woosley and W. Howard, Astrophys. J. Suppl. 36, 285 (1978). * Woosley and Howard (1990) S. Woosley and W. Howard, Astrophys. J. 354, L21 (1990). * Rayet et al. (1990) M. Rayet, N. Prantzos, and M. Arnould, Astron. Astrophys. 227, 271 (1990). * Arnould and Goriely (2003) M. Arnould and S. Goriely, Phys. Rep. 384, 1 (2003). * Rayet et al. (1995) M. Rayet, M. Arnould, M. Hashimoto, N. Prantzos, and K. Nomoto, Astron. Astrophys. 298, 517 (1995). * Rauscher et al. (2002) T. Rauscher, A. Heger, R. Hoffman, and S. Woosley, Astrophys. J. 576, 323 (2002). * Fröhlich et al. (2006) C. Fröhlich, G. Martínez-Pinedo, M. Liebendörfer, F.-K. Thielemann, E. Bravo, W. R. Hix, K. Langanke, and N. T. Zinner, Phys. Rev. Lett. 96, 142502 (2006). * Schatz et al. (1998) H. Schatz, A. Aprahamian, J. Görres, M. Wiescher, T. Rauscher, J. Rembges, F.-K. Thielemann, B. Pfeiffer, P. Möller, H. Herndl, et al., Phys. Rep. 294, 167 (1998). * Schatz et al. (2001) H. Schatz, A. Aprahamian, V. Barnard, L. Bildsten, A. Cumming, M. Ouellette, T. Rauscher, F.-K. Thielemann, and M. Wiescher, Phys. Rev. Lett. 86, 3471 (2001). * Hauser and Feshbach (1952) W. Hauser and H. Feshbach, Phys. Rev. 87, 366 (1952). * Rauscher and Thielemann (2000) T. Rauscher and F.-K. Thielemann, At. Data Nucl. Data Tables 75, 1 (2000). * Rauscher and Thielemann (2001) T. Rauscher and F.-K. Thielemann, At. Data Nucl. Data Tables 79, 47 (2001). * Goriely (2005) S. Goriely, Hauser-Feshbach rates for neutron capture reactions (version 08/26/05), http://www-astro.ulb.ac.be/Html/hfr.html (2005). * Mohr et al. (1997) P. Mohr, T. Rauscher, H. Oberhummer, Z. Máté, Z. Fülöp, E. Somorjai, M. Jaeger, and G. Staudt, Phys. Rev. C 55, 1523 (1997). * Demetriou et al. (2002) P. Demetriou, C. Grama, and S. Goriely, Nucl. Phys. A707, 253 (2002). * KADoNiS (2005) KADoNiS, The Karlsruhe Astrophysical Database of Nucleosynthesis in Stars, Online: http://nuclear-astrophysics.fzk.de/kadonis (2005). * Avrigeanu et al. (2003) M. Avrigeanu, W. von Oertzen, A. Plompen, and V. Avrigeanu, Nucl. Phys. A723, 104 (2003). * Fülöp et al. (2001) Z. Fülöp, G. Gyürky, Z. Máté, L. Somorjai, E. Zolnai, D. Galaviz, M. Babilon, P. Mohr, A. Zilges, T. Rauscher, H. Oberhummer, et al., Phys. Rev. C 64, 065805 (2001). * Rauscher (2003a) T. Rauscher, Nucl. Phys. A 719, 73 (2003a). * Rauscher (2003b) T. Rauscher, Erratum, Nucl. Phys. A 725, 295 (2003b). * McFadden and Satchler (1966) L. McFadden and G. Satchler, Nucl. Phys. 84, 177 (1966). * Fülöp et al. (1996) Z. Fülöp, A. Kiss, E. Somorjai, C. Rolfs, H.-P. Trautvetter, T. Rauscher, and H. Oberhummer, Z. Phys. A 355, 203 (1996). * Galaviz et al. (2005) D. Galaviz, Z. Fülöp, G. Gyürky, Z. Máté, P. Mohr, T. Rauscher, E. Somorjai, and A. Zilges, Phys. Rev. C 71, 065802 (2005). * Rapp et al. (2002) W. Rapp, M. Heil, D. Hentschel, F. Käppeler, R. Reifarth, H. J. Brede, H. Klein, and T. Rauscher, Phys. Rev. C 66, 015803 (2002). * Gyürky et al. (2006) G. Gyürky, G. Kiss, Z. Elekes, Z. Fülöp, E. Somorjai, A. Palumbo, J. Görres, H. Lee, W. Rapp, M. Wiescher, et al., Phys. Rev. C 74, 025805 (2006). * Özkan et al. (2007) N. Özkan, N. Efe, R. Güray, A. Palumbo, J. Görres, H. Lee, L. Lamm, W. Rapp, E. Stech, M. Wiescher, et al., Phys. Rev. C 75, 025801 (2007). * Somorjai et al. (1998) E. Somorjai, Z. Fülöp, A. Kiss, C. Rolfs, H.-P. Trautvetter, U. Greife, M. Junker, M. Arnould, M. Rayet, S. Goriely, et al., Proc. Nuclei in the Cosmos V, edts. N. Prantzos and S. Harissopulos, Editions Fronti$\grave{e}$res p. 459 (1998). * Rapp et al. (2006) W. Rapp, J. Görres, M. Wiescher, H. Schatz, and F. Käppeler, Ap. J. 653, 474 (2006). * Rauscher (2006) T. Rauscher, Phys. Rev. C 73, 015804 (2006). * Gledenov et al. (2000) Y. M. Gledenov, P. E. Koehler, J. Andrzejewski, K. H. Guber, and T. Rauscher, Phys. Rev. C 62, 042801(R) (2000). * Koehler et al. (2000) P. E. Koehler, Y. M. Gledenov, J. Andrzejewski, K. H. Guber, S. Raman, and T. Rauscher, Nucl. Phys. A688, 86c (2000). * Koehler et al. (2004) P. E. Koehler, Y. M. Gledenov, T. Rauscher, and C. Fröhlich, Phys. Rev. C 69, 015803 (2004). * Kiss et al. (2006) G. Kiss, Z. Fülöp, G. Gyürky, Z. Máté, E. Somorjai, D. Galaviz, A. Kretschmer, and K. Sonnabend, Eur. Phys. J. A 27, 197 (2006). * Brede et al. (1980) H. J. Brede, M. Cosack, G. Dietze, H. Gumpert, S. Guldbakke, R. Jahr, M. Kutscha, D. Schlegel-Bickmann, and H. Schölermann, Nucl. Instr. Meth. 169, 349 (1980). * Böttger (2002) R. Böttger, private communication (2002). * Iliadis (2007) C. Iliadis, _Nuclear Physics of Stars_ (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim/ Germany, 2007). * Ziegler and Biersack (2003) J. Ziegler and J. Biersack, ”The stopping and range of ions in matter”, SRIM-2003.26, http://www.srim.com (2003). * Burrows (1993) T. Burrows, Nucl. Data Sheets 68, 635 (1993). * Artna-Cohen (1993) A. Artna-Cohen, Nucl. Data Sheets 70, 85 (1993). * Blachot (2001) J. Blachot, Nucl. Data Sheets 92, 455 (2001). * Jaag (1993) S. Jaag, Computer code CASC, Technical report, Forschungszentrum Karlsruhe (1993). * Levkovskij (1991) V. Levkovskij, Activation Cross Sections By Protons and Alphas, Moscow/ USSR (1991). * Esterlund and Pate (1965) R. Esterlund and B. Pate, Nucl. Phys. 69, 401 (1965). * Graf and Muenzel (1974) H. Graf and H. Muenzel, Journal of Inorganic and Nuclear Chemistry 36, 3647 (1974). * Denzler et al. (1995) F.-O. Denzler, F. Rösch, and S. Qaim, Radiochimica Acta 68, 13 (1995). * England et al. (1982) J. B. A. England, S. Baird, D. H. Newton, T. Picazo, E. C. Pollacco, G. J. Pyle, P. M. Rolph, J. Alabau, E. Casal, and A. Garcia, Nucl. Phys. 388, 573 (1982). * Rauscher (2003c) T. Rauscher, Nucl. Phys. A719, 73c (2003c). * Rauscher (2003d) T. Rauscher, Nucl. Phys. A725, 295 (2003d).	arxiv-papers	2008-09-01T13:15:30	2024-09-04T02:48:57.584549	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "W. Rapp, I. Dillmann, F. K\\\"appeler, U. Giesen, H. Klein, T. Rauscher,\n D. Hentschel, and S. Hilpp", "submitter": "Iris Dillmann", "url": "https://arxiv.org/abs/0809.0234" }
0809.0277	# Structural stability of attractor-repellor endomorphisms with singularities Pierre Berger ###### Abstract We prove a theorem on structural stability of smooth attractor-repellor endomorphisms of compact manifolds, with singularities. By attractor-repellor, we mean that the non-wandering set of the dynamics $f$ is the disjoint union of a repulsive compact subset with a hyperbolic attractor on which $f$ acts bijectively. The statement of this result is both infinitesimal and dynamical. Up to our knowledge, this is the first in this hybrid direction. Our results generalize also a Mather’s theorem in singularity theory which states that infinitesimal stability implies structural stability for composed mappings, to the larger category of laminations. ###### Contents 1. 1 Motivations and statement of the main result 2. 2 Proof of the main result 3. 3 Partition of the manifold by invariant and persistent laminations 4. 4 Infinitesimal stability implies stability on the non-wandering set ## 1 Motivations and statement of the main result ### 1.1 Structural stability of Dynamical Systems Let $f$ be a _smooth endomorphism_ of a compact manifold $M$. This means that $f$ is a $C^{\infty}$ map from $M$ into itself which is not necessarily bijective and can have _singularities_ : its tangent map can be non surjective at some points. For $k\in\mathbb{N}\cup\\{\infty\\}$, the map $f$ is _$C^{k}$ -structurally stable_, if for every $f^{\prime}$ close to $f$ in the $C^{k}$-topology, there exists a homeomorphism $h$ of $M$ such that the following diagram commutes: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ h&\uparrow&&\uparrow&h\\\ &M&\rightarrow&M&\\\ &&f&&\\\ \end{array}$ The topology of the space of $C^{k}$ maps is the usual one for $k<\infty$, and for $k=\infty$ it is the union of the $C^{j}$ topologies for $j\in\mathbb{N}$. The combined work of many authors (Smale [Sma67], Palis [PS70], De Melo [dM73], Robin [Rob71], Robinson [Rob76], Mañe [Mañ88]) has led to a satisfactory characterization of $C^{1}$-structurally stable diffeomorphisms. Let us state their characterization: ###### Theorem 1.1. Let $M$ be a compact connected manifold and $f$ a $C^{1}$ diffeomorphism of $M$. Then the following conditions are equivalent: 1. 1. $f$ is $C^{1}$-structurally stable, 2. 2. $f$ satisfies Axiom $A$ and the strong transversality condition, 3. 3. $f$ is $C^{0}$-infinitesimally stable. Let us recall the definitions of statements 2 and 3. A diffeomorphism $f$ satisfies _Axiom $A$_ if its non-wandering set $\Omega$ is hyperbolic and equal to the closure of the set of periodic points. The diffeomorphism $f$ satisfies moreover the _strong transversality condition_ if the stable and unstable manifolds of points of $\Omega$ intersect each other transversally. The concept of infinitesimal stability will play a crucial role in this work. A $C^{r+1}$ endomorphism is _$C^{r}$ -infinitesimally stable_ if the following map is surjective: $\sigma\in\chi^{r}(M)\mapsto\sigma\circ f-Tf\circ\sigma\in\chi^{r}(f),$ with $\chi^{r}(M)$ the space of $C^{r}$ sections of the tangent bundle $TM$, and with $\chi^{r}(f)$ the space of $C^{r}$ sections of the push forward bundle $f^{}TM$ whose fiber at $x$ is $T_{f(x)}M$. In order to understand infinitesimal stability, let us regard a smooth endomorphism $f$ of the torus $\mathbb{T}=\mathbb{R}^{n}/\mathbb{Z}^{n}$. We notice that the following map is Fréchet differentiable: $\phi\;:\;Diff^{\infty}(\mathbb{T})\times Diff^{r}(\mathbb{T})\rightarrow C^{r}(\mathbb{T},\mathbb{T})$ $(f^{\prime},h)\mapsto h\circ f^{\prime}-f^{\prime}\circ h,$ where $Diff^{\infty}(\mathbb{T})$ and $Diff^{r}(\mathbb{T})$ are the spaces of $C^{\infty}$ and $C^{r}$ diffeomorphisms of $\mathbb{T}$ respectively. Moreover, its partial derivative at $(f,id)$ with respect to the second variable is $\sigma\in\chi^{r}(\mathbb{T})\mapsto\sigma\circ f-Tf\circ\sigma\in\chi^{r}(\mathbb{T}).$ Consequently the above partial derivative is surjective iff $f$ is $C^{r}$-infinitesimally stable. ### 1.2 Structural stability in Singularity Theory Meanwhile, the school initiated by Whitney and Thom was interested in the following problem. Let $f$ be a smooth map from a manifold $M_{1}$ into a manifold $M_{2}$. For $k>0$, the map $f$ is _$C^{k}$ -equivalently stable_111Usually, in singularity theory, we say $C^{k}$-structurally stable; however this conflicts the dynamical terminology. if there exists a neighborhood $U$ of $f$ in the $C^{\infty}$ topology such that for any $f^{\prime}\in U$, there exist $C^{k}$ automorphisms $h_{1}$ and $h_{2}$ of $M_{1}$ and $M_{2}$ respectively such that the following diagram commutes: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M_{1}&\rightarrow&M_{2}&\\\ h_{1}&\uparrow&&\uparrow&h_{2}\\\ &M_{1}&\rightarrow&M_{2}&\\\ &&f&&\\\ \end{array}.$ The problem is then to describe the $C^{k}$-equivalently stable maps which is usually simpler. This program was carried out by Mather who solved many conjectures of Thom. One of his results is the following: ###### Theorem 1.2 (Mather [Mat68a], [Mat69]). The $C^{\infty}$-equivalently stable, proper maps are the $C^{\infty}$-equivalent infinitesimal stable maps. Here _$C^{\infty}$ -equivalent infinitesimal stability_ means that the following map is onto: $(\sigma_{1},\sigma_{2})\in\chi^{\infty}(M_{1})\times\chi^{\infty}(M_{2})\mapsto\sigma_{2}\circ f-Tf\circ\sigma_{1}\in\chi^{\infty}(f).$ ### 1.3 Statement of the main result Concerning the structural stability of endomorphisms, there is not yet a criterion, neither a satisfactory description. For instance, Dufour [Duf77] showed that there is no $C^{\infty}$-infinitesimal stable endomorphism with a periodic point. On the other hand, the map $x\mapsto x^{2}$ on the one point compactification of $\mathbb{R}$, is $C^{0}$-infinitesimally stable but not $C^{\infty}$-structurally stable. There are very few old theorems stating sufficient conditions for the structural stability of endomorphisms that are not diffeomorphisms. Nowadays the subject regains of interest with new examples ([IPR08], [IPR97]). We recall below all these theorems. Our main theorem generalizes all the theorems implying the $C^{\infty}$-stability of endomorphisms that are not diffeomorphisms. A first old result was stated by Shub in his PhD thesis [Shu69], and requires the following definition: ###### Definition 1.3.1. Let $f$ be a $C^{1}$ endomorphism of a Riemannian manifold $(M,g)$, which sends a compact subset $K\subset M$ into itself. The compact subset $K$ is _repulsive_ if there exist $\lambda>1$ and $n\geq 1$ such that for every $x\in K$, the tangent map $T_{x}f^{n}$ is invertible with contractive inverse. The map $f$ is _expanding_ if $M$ is compact and $K=M$. ###### Theorem 1.3 (Shub 69). The expanding $C^{1}$-endomorphisms of a compact manifold are $C^{1}$-structurally stable. The following result is known since many times: ###### Theorem 1.4. Let $f$ be a rational function of the Riemann sphere. If the Julia set of $f$ is repulsive and if the critical points of $f$ are quadratic, with non- preperiodic and disjoint orbits, then $f$ is structurally stable for holomorphic perturbations. In the above example the singularities are not (equivalently) stable for smooth perturbations and so the dynamics is not $C^{\infty}$-structurally stable. However, since its critical points are necessarily in the attracting basin of periodic attracting orbits, it is a good example of attractor- reppelor dynamics. ###### Definition 1.3.2. Let $f$ be a smooth endomorphism of a compact, non necessarily connected manifold. The endomorphism $f$ is _attractor-repellor_ if its non-wandering set is the disjoint union of two compact subsets $R$ and $A$ such that: • the compact subset $R$ is repulsive, but non necessarily transitive, * • the compact subset $A$ is hyperbolic, the restriction of $f$ to $A$ is bijective, and the unstable manifolds of points of $A$ are contained in $A$. However $A$ is not necessarily transitive. From the above discussion we understand that our result on structural stability of endomorphisms needs to mix criteria from dynamical systems and singularity theory. ###### Theorem 1.5 (Main Result). Let $f$ be an attractor-repellor, smooth endomorphism of a compact, non necessarily connected manifold $M$. If the following conditions are satisfied, then $f$ is $C^{\infty}$-structurally stable: * (i) the periodic points of $f$ are dense in $A$, * (ii) the singularities $S$ of $f$ have their orbits that do not intersect the non- wandering set $\Omega$, * (iii) the restriction of $f$ to $M\setminus\hat{\Omega}$ is $C^{\infty}$-infinitesimally stable, with $\hat{\Omega}:=cl\big{(}\cup_{n\geq 0}f^{-n}(\Omega)\big{)}$, * (iv) $f$ is transverse to the stable manifold of $A$’s points: for any $y\in A$, for any point $z$ in a local stable manifold $W^{s}_{y}$ of $y$, for any $n\geq 0$, and for any $x\in f^{-n}(\\{z\\})$, we have: $Tf^{n}(T_{x}M)+T_{z}W_{y}^{s}=T_{z}M.$ We recall that the _singularities_ of a smooth endomorphism $f$ of $M$ are the points $x$ in $M$ for which the tangent map $T_{x}f$ is not surjective. ###### Remark 1.3.3. Actually, we prove that the conjugacy map $h$ between $f$ and its perturbation is a smooth immersion restricted to each stable manifold of $f$ without $\hat{A}:=\cup_{n\geq 0}f^{-n}(A)$. Moreover, we prove that the partial derivatives along the stable manifolds depend continuously on the base point over all $W^{s}(\Omega)\setminus\hat{\Omega}$. The following non-trivial remark follows easily from the proof of the main result: ###### Remark 1.3.4. The hypotheses of this theorem are open: any small smooth perturbation of $f$ also satisfies them. The main theorem generalizes a well known result: ###### Corollary 1.6. Let $f$ be a smooth endomorphism of the circle $\mathbb{S}^{1}$. If $f$ is an attractor-repellor endomorphism such that the critical points of $f$ are quadratic, with disjoint orbits in the basin of the attractor but non- preperiodic, then $f$ is $C^{\infty}$-structurally stable. ###### Remark 1.3.5. Actually by [KSvS07], the above corollary is maximal in dimension 1: there are no more structurally stable endomorphisms of the circle. Moreover a $C^{\infty}$-generic map of the circle satisfies this hypothesis. ###### Proof of Corollary 1.6. Only the infinitesimal condition hypothesis is not obvious. By Lemma 1.3.10 (see below), we only need to prove the $C^{\infty}$-equivalent infinitesimal stability of the map $f:\;x\in\mathbb{R}\mapsto x^{2}\in\mathbb{R}$. Let $\xi\in\chi(f)$. Let $\sigma_{2}$ be the constant function equal to $\xi(0)$ and: $\sigma_{1}(x):=\frac{\sigma_{2}(x^{2})-\xi(x)}{2x}=-\int_{0}^{1}\frac{\xi^{\prime}(tx)}{2}dt$ We notice that $\sigma_{1},\sigma_{2}\in\chi^{\infty}(\mathbb{S}^{1})$ satisfy: $\sigma_{2}(x^{2})-2x\sigma_{1}(x)=\xi(x),\quad\forall x\in\mathbb{S}^{1}.$ $\mathrm{i.e.}\;\sigma_{2}\circ f-Tf\circ\sigma_{1}=\xi.$ ∎ Equivalent infinitesimal stability is a $C^{\infty}$-generic property for maps from compact manifolds of dimension less than 9 (see [GG73] p163 when the differentiable map sends a manifold into one of different dimension). Conditions on the dimension of the manifold are given in [Nak89] when the singularities overlap. This provides other applications of the main result. ###### Example 1.3.6. Let $R$ be a rational function of the Riemann sphere such that all its critical points belong to the basin of attracting periodic orbits. Then the non-wandering set $\Omega$ of $R$ split into two sets: the union of the basin of a finite number of attracting periodic points, and the Julia set $J$ which is a repulsive compact subset. Thus $R$ is an attractor-repellor endomorphism of the sphere. But condition $(ii)$ is not satisfied since its singularities are not stable. Nevertheless, for $C^{\infty}$-generic perturbations $R^{\prime}$ of $R$ the singularities $\Sigma$ of $R$ are equivalently stable and do not overlap along their orbits ($f^{k}(\Sigma)\cap\Sigma=\emptyset$, $\forall k>0$). Thus $R^{\prime}$ satisfies the hypothesis of the main theorem, and hence is structurally stable. To have the shape of the singularities of $R^{\prime}$ see [AGZV85] p. 20. This result was recently shown in [IPR97]. ###### Example 1.3.7. When the endomorphism $f$ does not have singularities, hypothesis $(iii)$ is always satisfied, as we will see in Lemma 1.3.10. Consequently, our main result implies the structural stability of the following endomorphism: $f:\;\mathbb{S}^{2}\times\mathbb{S}^{1}\rightarrow\mathbb{S}^{2}\times\mathbb{S}^{1}$ $(z,z^{\prime})\mapsto\Big{(}\frac{z+z^{\prime}}{3},z^{\prime 2}\Big{)},$ where $\mathbb{S}^{2}$ is the Riemann sphere and $\mathbb{S}^{1}$ is the unit circle of the complex plane. We notice that the non-wandering set is the disjoint union of the Smale solenoid (inside the product of the unit disk with $\mathbb{S}^{1}$) with the repulsive circle $\\{\infty\\}\times\mathbb{S}^{1}$. This example is the first structurally stable endomorphism known which is neither a diffeomorphism nor expanding and was found in [IPR08]. ###### Example 1.3.8. Let us introduce some singularities in the previous example. Let $\mathbb{D}$ be the closed unit disk. We notice that $\mathbb{D}\times\mathbb{S}^{1}$ contains the solenoid, is sent into its interior by the dynamics and the restriction of $f$ to $\mathbb{D}\times\mathbb{S}^{1}$ is a diffeomorphism onto its image. Let us parametrize $\bar{\mathbb{T}}:=\mathbb{D}\times\mathbb{S}^{1}$ in polar coordinates $\\{(re^{i\theta},z^{\prime});\;r\in[0,1],\;\mathrm{and}\;z^{\prime}\in\mathbb{S}^{1}\\}$. Let $\rho\in C^{\infty}([0,1],[0,1])$ be a Morse function equal to the identity everywhere except a small subset $U$ close to $1$ but with closure disjoint from $\\{1\\}$. We suppose that $\rho$ has exactly two singularities and that $\rho$ sends $U$ into $U$. We notice that $\rho_{1}:\;re^{i\theta}\mapsto\rho(r)e^{i\theta}$ has for singularities two folds over two disjoint circles. For some perturbation $\rho_{2}$ of $\rho_{1}$, the diffeomorphism $\rho_{2}$ is still equal to the identity on the complement of $U\times\mathbb{S}^{1}$, its singularities consist of two folds over two disjoint circles transverse to those of $\rho_{1}$, and $\rho_{2}$ sends $U\times\mathbb{S}^{1}$ into itself. Let us now regard: $f:\;(z,z^{\prime})\mapsto\left\\{\begin{array}[]{cc}f(z,z^{\prime})&\mathrm{for}\;(z,z^{\prime})\in(\mathbb{S}^{2}\times\mathbb{S}^{1}\setminus\bar{\mathbb{T}})\cup f^{2}(\bar{\mathbb{T}})\\\ f(\rho_{1}(z),z^{\prime})&\mathrm{for}\;(z,z^{\prime})\in\bar{\mathbb{T}}\setminus f(\bar{\mathbb{T}})\\\ f^{2}\circ(\rho_{2}(\cdot,id)\circ f^{-1}_{\|\bar{\mathbb{T}}}(z,z^{\prime})&\mathrm{for}\;(z,z^{\prime})\in f(\bar{\mathbb{T}})\setminus f^{2}(\bar{\mathbb{T}})\end{array}\right.$ Let us show how that the main theorem implies that $f^{\prime}$ is $C^{\infty}$-structurally stable. First we notice that $f^{\prime}$ is a smooth endomorphism of $\mathbb{S}^{2}\times\mathbb{S}^{1}$, and is an attractor-repellor endomorphism: it has the same non-wandering set as $f$ $(i)$. Moreover $f^{\prime}$ is transverse to the stable manifolds of the solenoid $(iv)$. Its singularities are formed by four folds; only two of them intersect the other ones along their orbits. This intersection occurs only at the first iteration and is transverse. Since the singularity of $\rho$ are close to $\\{1\\}$, the singularities of $f^{\prime}$ are disjoint from the solenoid, but included in $\bar{\mathbb{T}}$. As $f$ embeds $\bar{\mathbb{T}}$ into itself, the orbits of the singularities do not intersect the non- wandering set $(ii)$. Let us prove Property $(iii)$. For this end we are going to show the $C^{\infty}$-equivalent infinitesimal stability of the diagram: $\begin{array}[]{ccccc}&f_{1}&&f_{2}&\\\ \mathbb{R}^{2}&\rightarrow&\mathbb{R}^{2}&\rightarrow&\mathbb{R}^{2}\\\ (x,y)&\mapsto&(x^{2},y)&&\\\ &&(x,y)&\mapsto&\big{(}(x+y)^{2},y\big{)}\end{array}$ The _$C^{\infty}$ -equivalent infinitesimal stability_ of this diagram means that the map $(\sigma_{1},\sigma_{2},\sigma_{3})\in\chi^{\infty}(\mathbb{R}^{2})^{3}\mapsto(\sigma_{3}\circ f_{1}-Tf_{2}\circ\sigma_{1},\sigma_{3}\circ f_{2}-Tf_{3}\circ\sigma_{1})\in\chi^{\infty}(f_{1})\times\chi^{\infty}(f_{2})$ is surjective. Such an infinitesimal stability will be sufficient for our purpose. The product of above maps with the identity of $\mathbb{R}$ provides a local model of the singularities of $f$. The equivalent infinitesimal stability of such a product follows easily from the one of the above diagram. Also by using a partition of the unity, we get the infinitesimal stability of $f$ restricted to a neighborhood of the singularities. Then Lemma 1.3.10 implies the infinitesimal stability Property $(iii)$. Let us prove the $C^{\infty}$-equivalent infinitesimal stability of the above diagram. Let $\xi\in\chi^{\infty}(f_{1})$ and $\zeta\in\chi^{\infty}(f_{2})$. We want to construct $\sigma_{1},\sigma_{2},\sigma_{3}\in\chi^{\infty}(\mathbb{R}^{2})$ such that: $\sigma_{2}\circ f_{1}-Tf_{1}(\sigma_{1})=\xi\quad\mathrm{and}\quad\sigma_{3}\circ f_{2}-Tf_{2}(\sigma_{2})=\zeta.$ By putting an exponent $1$ or $2$ to denote the first or second component respectively, the above system is equivalent to: $\displaystyle\sigma_{2}^{1}(x^{2},y)-2x\sigma_{1}^{1}(x,y)=\xi^{1}(x,y)$ $\displaystyle\sigma_{2}^{2}(x^{2},y)-\sigma_{1}^{2}(x,y)=\xi^{2}(x,y)$ $\displaystyle\sigma_{3}^{1}\big{(}(x+y)^{2},y\big{)}-2(x+y)\big{(}\sigma_{2}^{1}(x,y)+\sigma_{2}^{2}(x,y)\big{)}=\zeta^{1}(x,y)$ $\displaystyle\sigma_{3}^{2}\big{(}(x+y)^{2},y\big{)}-\sigma_{2}^{2}(x,y)=\zeta^{2}(x,y)$ Therefore it is sufficient to solve the following system with $X:=x+y$ and $Y:=y$: $\displaystyle\sigma_{1}^{1}(x,y)=\frac{\sigma_{2}^{1}(x^{2},y)-\xi^{1}(x,y)}{2x}$ (1) $\displaystyle\sigma_{2}^{1}(X,Y)+\sigma_{2}^{2}(X,Y)=\frac{\sigma_{3}^{1}\big{(}X^{2},Y)-\zeta^{1}(X,Y)}{2X}$ (2) $\displaystyle\sigma_{1}^{2}(x,y)=\sigma_{2}^{2}(x^{2},y)-\zeta^{2}(x^{2},y)$ (3) $\displaystyle\sigma_{2}^{2}(X,Y)=\sigma_{3}^{2}\big{(}X^{2},Y\big{)}-\zeta^{2}(X,Y)$ (4) We put $\sigma_{3}^{1}(X,Y):=\zeta(0,Y)$ which makes sense to $(2)$ when $X$ approaches $0$ and $\sigma_{3}^{2}(Y^{2},Y):=-\xi^{1}(Y,Y)+\frac{\zeta^{1}(0,Y)-\zeta^{1}(Y,Y)}{2Y}+\zeta^{2}(Y,Y)$ and smoothly extended off the parabola $X=Y^{2}$. Thus $\sigma_{2}^{2}$ is defined without ambiguity by $(4)$, $\sigma_{2}^{1}$ by $(2)$, and $\sigma_{1}^{2}$ by $(3)$. Finally we compute that $\sigma_{2}^{1}(0,y)=\xi^{1}(0,y)$ and so we can well define $\sigma_{1}^{1}$ by $(1)$. ###### Example 1.3.9. Let $M$ be the $2$-sphere that we identify with the one-point compactification of $\mathbb{R}^{2}$. Let $(r,\theta)$ be the polar coordinates of $\mathbb{R}^{2}$. $\mathrm{Let}\quad f:\;\hat{\mathbb{R}}^{2}\rightarrow\hat{\mathbb{R}}^{2}$ $(r,\theta)\mapsto\Big{(}\frac{r}{2+2r^{2}},\theta\Big{)}.$ We notice that $f$ is an attractor-repellor endomorphism with $\\{0\\}$ as attractor and with empty repellor. Also the singularities of $f$ are folds that do not overlap. Consequently the hypotheses of the main theorem are satisfied and so $f$ is structurally stable. Here is the lemma that we needed for all our computations. ###### Lemma 1.3.10. Let $f$ be an attractor-repellor endomorphism of a compact manifold $M$. Suppose that $f$ satisfies Properties $(i)$ and $(ii)$ of the main theorem. Let $\Omega$ be the non-wandering set of $f$ and $M^{\prime}:=M\setminus cl(\cup_{n\geq 0}f^{-n}(\Omega))$. Suppose moreover the existence of an open neighborhood $U\subset M^{\prime}$ of the singularities such that: * $(a)$ for every $x\in U$, if an iterate $f^{n}(x)$ belongs to $U$, then $(f^{k}(x))_{k=0}^{n}$ belongs to $U$, * $(b)$ The restriction of $f^{n}$ to $U\setminus f^{-1}(U)$ is injective with injective derivative, for every $n\geq 0$, * $(c)$ the map $\sigma\in\chi^{\infty}(M^{\prime})\mapsto(\sigma\circ f-Tf\circ\sigma)_{\|U}\in\chi^{\infty}(f_{\|U})$ is surjective, then the restriction of $f$ to $M^{\prime}$ is $C^{\infty}$-infinitesimally stable. This lemma will be shown at this end of in Subsection 4.6. ### 1.4 Links with structural stability of composed of mappings If we mentioned several times that the manifolds are not necessarily connected, it is because in this case, our main result is a complement to a theorem of Mather on structural stability of composed mapping. Moreover, contrarily to what happen for diffeomorphisms or local diffeomorphisms (such as expanding maps), the endomorphisms can send a connected component of the manifold into one of different dimension. As mentioned by Baas [Baa74], the problem of composed mapping was first stated by Thom, and have many applications in Biology (see [Tho71] and [Baa73]), in the study of network (see Baas [Baa74]) and in the study of the so-called Laudau singularity of Feynman integral (see [Pha67]). To state the problem of composed mapping, let us consider a finite oriented graph $G:=(V,A)$ with a manifold $M_{i}$ associated to each vertex $i\in V$, and with a smooth map $f_{ij}\in C^{\infty}(M_{i},M_{j})$ associated to each arrow $[i,j]\in A$ from $i$ to $j$. ###### Example 1.4.1. $\begin{array}[]{ccccc}&&f_{12}&\\\ &M_{1}&\rightarrow M_{2}&\\\ f_{12}&\uparrow&\nearrow f_{32}&\\\ &M_{3}&&\\\ &\circlearrowright&&\\\ &f_{33}&&\end{array}$ For $k\geq 0$, such a graph is _$C^{k}$ -equivalently stable_ if for every $(f^{\prime}_{ij})_{[i,j]\in A}$ close to $(f_{ij})_{[i,j]\in A}$ in $\prod_{[i,j]\in A}C^{\infty}(M_{i},M_{j})$, there exist diffeomorphisms $(h_{i})_{i}\in\prod_{i\in V}C^{k}(M_{i},M_{i})$ s.t. for every $[i,j]\in A$ the following diagram commutes: $\begin{array}[]{rcccl}&&f^{\prime}_{ij}&&\\\ &M_{i}&\rightarrow&M_{j}&\\\ h_{i}&\uparrow&&\uparrow&h_{j}\\\ &M_{i}&\rightarrow&M_{j}&\\\ &&f_{ij}&&\\\ \end{array}.$ The graph $G$ is _convergent_ if at most one arrow of $A$ starts from each vertex. The graph is _without cycle_ if for any family of arrows $([i_{k},i_{k+1}])_{k=1}^{N}\in A^{N}$, the vertices $i_{1}$ and $i_{N+1}$ are different. The following theorem was proved by Mather, and then written by Baas [Baa74]: ###### Theorem 1.7. Let $G$ be a graph of smooth proper maps, convergent and without cycle. The graph is $C^{\infty}$-equivalently structurally stable if the following map is surjective: $\prod_{i\in V}\chi^{\infty}(M_{i})\rightarrow\prod_{[i,j]\in A}\chi^{\infty}(f_{ij})$ $(\sigma_{i})_{i}\mapsto(Tf_{ij}\circ\sigma_{i}-\sigma_{j}\circ f_{ij})_{[ij]}.$ We will see that our main theorem generalizes the above result when the manifolds $(M_{i})_{i}$ are compact. Though this theorem has never been published, but it has been cited several times ([Nak89], [Duf77], [Buc77]…). Hence we will explain how to deduce the proof of this theorem in the non- compact case from this work (see Remark 4.2.8). But before, we shall notice that there is a canonical graph of maps associated to each smooth endomorphism $f$ of a compact manifold $M$. Let $(M_{i})_{i\in V}$ be the connected components of $M$. Let $A$ be the set of arrows $[i,j]$ such that $f$ sends $M_{i}$ into $M_{j}$. Let $f_{ij}$ be the restriction of $f$ to $M_{i}$. Therefore, $G:=(V,A)$ is a graph of smooth maps, convergent but _always with cycles_. Also, we notice that the $C^{\infty}$-structural stability of $f$ is the $C^{0}$-equivalent stability of this graph of maps. Conversely, let us show that our main theorem implies the Mather’s one on equivalent stability of graphs of maps, in the compact case. Let $G=(V,A)$ be a convergent and without cycle graph of maps of compact manifolds. Let $\hat{V}$ be the union of $V$ with the circle $\hat{\mathbb{R}}$ and with the trivial $0$-dimensional manifold $\\{0\\}$. We identify the circle $\hat{\mathbb{R}}$ to the one-point compactification of the real line. Let $f_{\hat{\mathbb{R}}\hat{\mathbb{R}}}:=x\in\hat{\mathbb{R}}\mapsto 2x$ and let $f_{0\hat{\mathbb{R}}}:=0\in\\{0\\}\mapsto 1\in\hat{\mathbb{R}}$. Let $V^{\prime}\subset V$ be the subset of vertices from which no arrow starts. For $i\in V^{\prime}$, let $f_{i0}$ be the constant map from $M_{i}$ onto $\\{0\\}$. Let finally $\hat{A}:=A\cup\\{[i,0]\\}_{i\in V^{\prime}}\cup\\{[0,\hat{\mathbb{R}}]\\}\cup\\{[\hat{\mathbb{R}},\hat{\mathbb{R}}]\\}$. One easily remarks that the hypotheses of Mather’s theorem for the graph $G$ implies those of the main result for the smooth endomorphism $f$ of the disjoint union $M=\coprod_{i\in\hat{V}}M_{i}$, whose restriction to $M_{i}$ is the map $f_{ij}$, with $i\in\hat{V}$ and $[i,j]\in\hat{A}$. Also for any smooth perturbation $(f^{\prime}_{ij})_{[i,j]\in A}$ of $(f_{ij})_{[i,j]\in A}$, we can use the above algorithm to associate a perturbation $f^{\prime}$ of $f$. By the main theorem, the endomorphism $f^{\prime}$ is conjugated to $f$. By Remark 1.3.3, the conjugacy is smooth along the stable manifold of $M$. As they contain each manifold $(M_{i})_{i\in V}$, this implies the Mather’s theorem, in the compact case. ### 1.5 Thanks I thank J. Milnor for his interest on this project. I am grateful to A. Rovella for the communication of his results. I am very thankful to J.N. Mather for fruitful conversations and for providing me the unpublished preprint of N. Baas on stability of composed mappings. This work have been done in the Institute for Mathematical Science, in State University of New York at Stony Brook, whose hospitality is gratefully acknowledged. ## 2 Proof of the main result The proof is split into two parts. The first is the construction of a geometry which is preserved by the dynamics and which is persistent for small perturbations. For this end, we will use the formalism of laminations and some equivalent of the Hirsch-Pugh-Shub theory. This part is mostly dynamic and geometric. The second part consists of showing the structural stability of the dynamics with respects to this geometry. More precisely we will show a generalization of the Mather’s theorem on equivalent stability of composing mappings in the larger context of laminations and by allowing these maps to overlap. This part uses mostly ingredients from singularity theory. As both parts use the formalism of laminations, we shall first define them rigorously. ### 2.1 Definition of the laminations and their morphisms #### 2.1.1 Definition of lamination Let us consider a locally compact and second-countable metric space $L$ covered by open subsets $(U_{i})_{i}$ called _distinguished open subset_ , endowed with homeomorphisms $h_{i}$ from $U_{i}$ onto $V_{i}\times T_{i}$, where $V_{i}$ is an open set of $\mathbb{R}^{d}$ and $T_{i}$ is a metric space. The _charts_ $(U_{i},h_{i})_{i}$ define a ($C^{\infty}$) _atlas_ of a lamination structure on $L$ if the _coordinate changes_ $h_{ij}=h_{j}\circ h_{i}^{-1}$ can be written in the form: $h_{ij}(x,t)=(\phi_{ij}(x,t),\psi_{ij}(x,t)),$ where $\phi_{ij}$ takes its values in $\mathbb{R}^{d}$, $\psi_{ij}(\cdot,t)$ is locally constant for any $t$, and the partial derivatives $(\partial_{x}^{s}\phi_{ij})_{s=1}^{\infty}$ exist and are continuous on _all_ the domain of $\phi_{ij}$. A _lamination_ is such a metric space $L$ endowed with a maximal $C^{\infty}$-atlas $\mathcal{L}$. A _plaque_ is a subset of $L$ of the form $h_{i}^{-1}(V_{i}^{0}\times\\{t\\})$, for a chart $h_{i}$, $t\in T_{i}$ and a connected component $V_{i}^{0}$ of $V_{i}$. The _leaves_ of $\mathcal{L}$ are the smallest subsets of $L$ which contain any plaques that intersect them. If $V$ is an open subset of $L$, the set of the charts $(U,\phi)\in\mathcal{L}$ such that $U$ is included in $V$, forms a lamination structure on $V$, which is denoted by $\mathcal{L}_{\|V}$. The reader might read [Ghy99] and [Ber08] for example of laminations. #### 2.1.2 Morphisms of laminations For $r\in[\\![0,\infty]\\!]$, a _$C^{r}$ -morphism (of laminations)_ from $(L,\mathcal{L})$ to $(L^{\prime},\mathcal{L}^{\prime})$ is a continuous map $f$ from $L$ to $L^{\prime}$ such that, seen via charts $h$ and $h^{\prime}$, it can be written in the form: $h^{\prime}\circ f\circ h^{-1}(x,t)=(\phi(x,t),\psi(x,t)),$ where $\psi(\cdot,t)$ is locally constant, $\phi$ takes its values in $\mathbb{R}^{d^{\prime}}$ and has its $r$-first derivatives with respect to $x$ that are continuous on _all_ the domain of $\phi$. When $r=\infty$, the morphism is _smooth_. If, moreover, the linear map $\partial_{x}\phi(x,t)$ is always one-to-one (resp. onto), we will say that $f$ is an _immersion (of laminations)_ (resp. _submersion_). An _isomorphism (of laminations)_ is a bijective morphism of laminations whose inverse is also a morphism of laminations. An _embedding (of laminations)_ is an immersion which is a homeomorphism onto its image. An _endomorphism of $(L,\mathcal{L})$_ is a morphism from $(L,\mathcal{L})$ into itself. We denote by $T\mathcal{L}$ the vector bundle over $L$ whose fiber at $x\in L$, denoted by $T_{x}\mathcal{L}$, is the tangent space at $x$ to its leaf. If $f$ is morphism from $\mathcal{L}$ to $\mathcal{L}^{\prime}$, we denote by $Tf$ the bundle morphism from $T\mathcal{L}$ to $T\mathcal{L}^{\prime}$ over $f$ induced by the differential of $f$ along the leaves of $\mathcal{L}$. #### 2.1.3 Equivalent Classes of morphisms and their topologies We will say that two morphisms $f$ and $f^{\prime}$ from a lamination $(L,\mathcal{L})$ into a lamination $(L^{\prime},\mathcal{L}^{\prime})$ are _equivalent_ if they send each leaf of $\mathcal{L}$ into the same leaf of $\mathcal{L}^{\prime}$. We will deal with two topologies on such an equivalent class. Let us describe a base of neighborhoods of some morphism $f$ from $\mathcal{L}$ to $\mathcal{L}^{\prime}$ in each of these topologies. For this end, let us fix a cover $(K_{i})_{i\in\mathbb{N}}$ of $L$ by compact subsets s.t.: * - the cover $(K_{i})_{i}$ is locally finite: any point of $L$ has a neighborhood which intersects finitely many subsets of this family, * - for every $i$, the compact subsets $K_{i}$ and $f(K_{i})$ are included in distinguished open subsets endowed with charts $(h_{i},U_{i})$ and $(h^{\prime}_{i},U^{\prime}_{i})$ respectively. We define $(\phi_{i},\psi_{i})$ by $h^{\prime}_{i}\circ f\circ h_{i}^{-1}=(\phi_{i},\psi_{i})$ on $h_{i}(K_{i})$. ###### Definition 2.1.1. In the _Whitney $C^{r}$-topology_, with $r<\infty$, a base of neighborhoods of $f$ in its equivalent class $Mor_{f}^{r}(\mathcal{L},\mathcal{L}^{\prime})$ is given by the following open subsets with $(\epsilon_{i})_{i}$ going all over the set of families of positive real numbers: $\Omega:=\Big{\\{}f^{\prime}\in Mor^{r}_{f}(\mathcal{L},\mathcal{L}^{\prime})\;:\;\forall i\geq 0\;f^{\prime}(K_{i})\subset U^{\prime}_{i}\;\mathrm{and\;with}\;\phi^{\prime}_{i}\;\mathrm{defined\;by}\;\quad$ $h^{\prime}_{i}\circ f^{\prime}\circ h_{i}^{-1}=(\phi^{\prime}_{i},\psi_{i}),\;\mathrm{we\;have}\;\max_{h_{i}(K_{i})}\Big{(}\sum_{s=1}^{r}\\|\partial_{x}^{s}\phi_{i}-\partial_{x}^{s}\phi^{\prime}_{i}\\|\Big{)}<\epsilon_{i}\Big{\\}}.$ The _Whitney $C^{\infty}$-topology_ is the union of all the Whitney $C^{r}$-topologies, for $r\geq 1$. The Whitney $C^{r}$-topology is denoted by $\mathcal{W}^{r}$. ###### Definition 2.1.2. In the _compact-open $C^{r}$-topology_ with $r<\infty$, a base of neighborhoods of $f$ in its equivalent class is given by the following open subsets with $\epsilon>0$ and $i\geq 0$: $\Omega:=\Big{\\{}f^{\prime}\in Mor^{r}_{f}(\mathcal{L},\mathcal{L}^{\prime})\;:\;f^{\prime}(K_{i})\subset U^{\prime}_{i}\;\mathrm{and\;with}\;\phi^{\prime}_{i}\;\mathrm{defined\;by}\;\quad$ $h^{\prime}_{i}\circ f^{\prime}\circ h_{i}^{-1}=(\phi^{\prime}_{i},\psi_{i}),\;\mathrm{we\;have}\;\max_{h_{i}(K_{i})}\Big{(}\sum_{s=1}^{r}\\|\partial_{x}^{s}\phi_{i}-\partial_{x}^{s}\phi^{\prime}_{i}\\|\Big{)}<\epsilon\Big{\\}}.$ The _$C^{\infty}$ -compact-open topology_ is the union of all the $C^{r}$-compact-open topologies, for $r\geq 1$. The compact-open $C^{r}$-topology is denoted by $CO^{r}$. We remark that when $L$ is compact, both topologies are the same. We notice that when the laminations $(L,\mathcal{L})$ and $(L^{\prime},\mathcal{L}^{\prime})$ are manifolds, both definitions are consistent with the usual ones. ### 2.2 Results of the geometric part The following theorem describes the geometry that the dynamics preserves: ###### Theorem 2.1. If $f$ is a $C^{\infty}$-endomorphism of a smooth manifold $M$ satisfying the hypotheses of the main result, then the stable manifolds of the points of the attractor $A$ form the leaves of a $C^{\infty}$-lamination $(L^{s},\mathcal{L}^{s})$. Moreover the compact set $\hat{R}:=cl\big{(}\cup_{n\geq 0}f^{-n}(R)\big{)}$ is canonically endowed with a lamination structure $\mathcal{R}$ whose leaves are backward images by $f$ of points of $R$. Moreover the lamination $(\hat{R},\mathcal{R})$ is $r$-normally expanded, for any $r\geq 1$. This means that there exist $C>0$ and $\lambda>1$ s.t. for every $x\in R$, for every $u\in T_{x}\mathcal{R}$, $v\in T_{x}\mathcal{R}^{\bot}$, $n\geq 0$: $\\|p\circ T_{x}f^{n}(v)\\|\geq C\cdot\lambda^{n}\max(1,\\|T_{x}f^{n}(u)\\|^{r}),$ with $p$ the orthogonal projection of $TM_{\|\hat{R}}$ onto $T\mathcal{R}^{\bot}$. Also $M$ is equal to the disjoint union of $L^{s}$ with $\hat{R}$. We will prove this theorem in Section 3. The following theorem states the persistence of this geometry: ###### Theorem 2.2. Let $f$ be a smooth endomorphism satisfying the hypotheses of the main result. Let $(R,\mathcal{R})$ and $(L^{s},\mathcal{L}^{s})$ be the laminations provided by Theorem 2.1. Let $L^{\prime s}$ be a precompact, open subset of $L^{s}$ whose closure is sent into $L^{s}$ by $f$. Then, for any endomorphism $f^{\prime}$ $C^{\infty}$-close to $f$, there exists two smooth embeddings $i_{R}$ and $i_{s}$ of respectively $(\hat{R},\mathcal{R})$ and $(L^{\prime s},\mathcal{L}^{s}_{\|L^{\prime s}})$ into $M$ and two smooth endomorphisms $f^{\prime}_{R}$ and $f^{\prime}_{s}$ of these laminations such that: * • $f^{\prime}_{R}$ and $f^{\prime}_{s}$ are equivalent and $CO^{\infty}$-close to the restrictions of $f$ to respectively $\hat{R}$ and $L^{\prime s}$. * • $i_{R}$ and $i_{s}$ are $CO^{\infty}$-close to the canonical inclusions of respectively $(\hat{R},\mathcal{R})$ and $(L^{\prime s},\mathcal{L}^{s}_{\|L^{\prime s}})$ into $M$, * • the following diagrams commute: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ i_{R}&\uparrow&&\uparrow&i_{R}\\\ &\hat{R}&\rightarrow&\hat{R}&\\\ &&f^{\prime}_{R}&&\end{array}\qquad\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M&\\\ i_{s}&\uparrow&&\uparrow&i_{s}\\\ &L^{\prime s}&\rightarrow&L^{\prime s}&\\\ &&f^{\prime}_{s}&&\end{array}.$ This theorem will be shown in Section 3. ###### Remark 2.2.1. Even if we do not need this to show the main theorem, by using [Ber07], we can easily show that the embeddings $i_{R}$ and $i_{S}$ are the restrictions of a homeomorphism of $M$ which respects smoothly all the laminations. ### 2.3 Result of the singularity theory part The main remaining difficulty of the proof of the main theorem is to conjugate $f^{\prime}_{s}$ to $f_{\|L^{\prime s}}$. Such a conjugation follows basically from techniques of singularity theory, and require the following definition to be stated. ###### Definition 2.3.1. A morphism $f$ from a lamination $(L^{\prime},\mathcal{L}^{\prime})$ into $(L,\mathcal{L})$ is _transversally bijective_ if for any point $x\in L^{\prime}$, there exist charts $(U^{\prime},\phi^{\prime})$ and $(U,\phi)$ of respectively $x$ and $f(x)$ s.t. : * • $f$ sends $U$ into $U^{\prime}$, * • $\phi^{-1}\circ f\circ\phi^{\prime}$ has its transverse component which is bijective. The following theorem generalizes the one of Mather on equivalent stability of composed mapping of compact manifolds. ###### Theorem 2.3. Let $(L,\mathcal{L})$ be a $C^{\infty}$-lamination. Let $C$ be a compact subset of $L$. Let $f$ be a smooth endomorphism of $(L,\mathcal{L})$ satisfying: * • $f$ is transversally bijective and proper, * • there exists $N>0$ such that for every $x\in C$, an iterate $f^{n}(x)$ does not belong to $C$, for some $n\leq N$, * • the following map is surjective: $\chi^{\infty}(\mathcal{L})\rightarrow\chi^{\infty}(f)$ $\sigma\mapsto\sigma\circ f-Tf\circ\sigma,$ with $\chi^{\infty}(\mathcal{L})$ the space of $C^{\infty}$-sections of $T\mathcal{L}$ and $\chi^{\infty}(f^{})$ the space of $C^{\infty}$-sections of $f^{}T\mathcal{L}\rightarrow L$. Then for every $f^{\prime}$ close to $f$ for the compact-open $C^{\infty}$-topology, there exists an isomorphism $I(f^{\prime})$ of $(L,\mathcal{L})$ such that for every $x\in C$: $f^{\prime}\circ I(f^{\prime})(x)=I(f^{\prime})\circ f(x)$ This theorem will be proved in the last section of this work. ### 2.4 Proof of the main theorem ###### Proof of Theorem 1.5. Le $f$ be a smooth endomorphism of a compact manifold $M$, satisfying the hypotheses of the main theorem. Let $(\hat{R},\mathcal{R})$ and $(L^{s},\mathcal{L}^{s})$ be the laminations provided by Theorem 2.1. Let $f^{\prime}$ be $C^{\infty}$-close to $f$ and let $i_{R}$, $i_{s}$ and $f^{\prime}_{s}$ be the maps provided by Theorem 2.2. We remind that $f^{\prime}_{s}$ is close the restriction of $f$ to the lamination $(L^{s},\mathcal{L}^{s})$ for the compact-open $C^{\infty}$-topology. We put $L:=L^{s}\setminus\cup_{n\geq 0}f^{-n}(A)$. We will show in Section 4.4 the following lemma: ###### Lemma 2.4.1. The morphism $f$ satisfies the hypotheses of Theorem 2.3. The difficulty of this lemma is to pass from infinitesimal stability for manifold to infinitesimal stability for lamination. This lemma together with Theorem 2.3 provide the structural stability of $f$ on any compact subset of $L$. Let us chose carefully this compact subset. As $A$ is an attractor on which $f$ acts bijectively, there exists a small open neighborhood $U$ of $A$ such that: * • the restriction of $f$ to $U$ is a diffeomorphism, * • $f$ sends $cl(U)$ into $U$, * • $A$ is the maximal invariant subset of $U$: $A=\cap_{n\geq 0}f^{n}(U)$. Let $U^{}:=\cup_{n>0}f^{-n}_{\|U^{c}}(f(U))$. For $U$ sufficiently small, the open subsets $U^{}$ and $U$ have their closures disjoint from the singularities $S$ and the critical values $f(S)$ of $f$. Let $N>0$ be such that $f^{-N}(U)$ contains $S$. Let $K$ be the compact subset: $K:=cl\Big{(}f^{-N}(U)\setminus(U^{}\cup U)\Big{)}.$ We notice that $K$ satisfies the following conditions: 1. (1) $K$ is a neighborhood of the singularities and their image included in $L$, 2. (2) for every $x\in K$, if some iterate $f^{k}(x)$ belongs to $K$ then all the iterates $(f^{j}(x))_{j=0}^{k}$ belong to $K$. In order to apply Theorem 2.3, let us associate a perturbation $f^{\\#}$ of $f_{\|L}$ such that the restrictions $f^{\\#}_{\|K}$ and $f^{\prime}_{s\|K}$ are equal. For we take a function $\rho\in Mor^{\infty}(\mathcal{L},\mathbb{R})$ with compact support and equal to 1 on $K$. Then given an isomorphism $\phi$ from $T\mathcal{L}$ to a neighborhood of the diagonal of $\mathcal{L}\times\mathcal{L}$ which sends a neighborhood of the zero section to the diagonal, we put: $f^{\\#}:\;x\in L\mapsto\phi\big{(}\rho(x)\phi^{-1}(f(x),f^{\prime}(x)\big{)}.$ We can now apply Theorem 2.3 with $C:=K$ and for $f_{\|L}$. Let $I(f^{\prime\\#})$ be the provided isomorphism of $(L,\mathcal{L})$ associated to $f^{\prime\\#}$. We define now: $h_{0}:\;x\in\hat{K}\cup\hat{R}\mapsto\left\\{\begin{array}[]{cl}i_{R}(x)&\mathrm{if}\;x\in\hat{R}\\\ i_{s}\circ I(f^{\prime\\#})(x)&\mathrm{if}\;x\in K\end{array}\right.$ We notice that $h_{0}$ respects the lamination $\mathcal{L}_{s}$ and that for $x\in K\cup\hat{R}$ we have: $f^{\prime}\circ h_{0}(x)=h_{0}\circ f(x).$ Let $M_{0}$ be the union of the connected components of $M$ which intersect the non-wandering set $\Omega$ of $f$. We denote by $K_{0}$ the intersection $K\cap M_{0}$, by $D^{n}:=f^{n}(U)\setminus f^{n+1}(U)$ for $n>0$ and by $D^{n}:=f^{n}_{\|M_{0}}(K_{0})\setminus f^{n+1}_{\|M_{0}}(K_{0})$ for any $n<0$. We notice that $A\cup\bigcup_{n>0}D^{n}$ is a neighborhood of $A$. Let $R_{0}:=M_{0}\cap\hat{R}$. Let us show that $M_{0}$ is contained in the disjoint union $A\cup K\cup R_{0}\cup U^{}\cup\bigcup_{n\not=0}D^{n}.$ Let $x\in M_{0}$. As the limit set of $x$ is included in $\Omega=A\cup R_{0}$, every large iterate $f^{n}(x)$ is either close to $A$ or close to $R_{0}$. If $f^{n}(x)$ is close to $A$, then $x$ is included in $A\cup\cup_{n\not=0}D^{n}\cup U^{}$. If the orbit of $x$ is never close to $A$, then $(f^{n}(x))_{n}$ belongs to a small neighborhood of $R_{0}$, for $n$ sufficiently large. As $R_{0}$ is repulsive and $f_{\|M_{0}}$-invariant ($f^{-1}(R_{0})\cap M_{0}=R_{0}$), $R_{0}$ is locally maximal: $R_{0}=\cap_{n\geq 0}f^{-n}(V),\quad\mathrm{for\;some\;neighborhood}\;V\mathrm{of}\;R_{0}.$ This implies that $f^{n}(x)$ belongs to $R_{0}$. This completes the proof of the following inclusion: $M_{0}\subset A\cup R_{0}\cup U^{}\cup K\cup\bigcup_{n\not=0}D^{n}.$ (5) We endow $M$ with an adapted metric to the hyperbolicity of $R_{0}$ and $A$. By restricting $U$ and taking $N$ sufficiently large, we can suppose that $f^{\prime}$ is uniformly expending on $\cup_{n<0}D^{n}\cup R_{0}$ and uniformly contracting along the leaves of $\mathcal{L}_{\|U}^{s}$. Thus we can suppose $\epsilon>0$ small enough and then $f^{\prime}$ close enough to $f$ such that: * • for every $x\in cl(\cup_{n<0}D^{n}\cup R^{0})$, the restriction of $f^{\prime}$ to any ball $B(x,\epsilon)$ centered at $x$ and with radius $\epsilon$ is an expanding diffeomorphism onto a neighborhood of $B(f(x),\epsilon)$, * • For every $x\in U$, the restriction of $f^{\prime}$ to $\mathcal{L}_{x}^{\prime}$ is contracting with precompact image in $\mathcal{L}^{\prime}_{f(x)}$, where $\mathcal{L}^{\prime}_{x}$ and $\mathcal{L}^{\prime}_{f(x)}$ are the image by $i_{s}$ of respectively the union of $\epsilon$-$\mathcal{L}^{s}$-plaque containing $x$ and $f(x)$ respectively. For $x\in A$, let $h_{A}(x)$ be the unique element of the intersection $\cap_{n\geq 0}f^{\prime n}\big{(}\mathcal{L}^{\prime}_{f_{\|U}^{-n}(x)}\big{)}$. $\mathrm{Let}\;h_{1}:=x\in M_{0}\setminus U^{}\mapsto\left\\{\begin{array}[]{cl}h_{0}(x)&\mathrm{if}\;x\in R_{0}\cup K\\\ f^{\prime n}\circ h_{0}\circ f^{-n}_{\|U}&\mathrm{if}\;x\in D^{n},\;n>0\\\ f^{\prime-1}_{\|B(x,\epsilon)}\circ\cdots\circ f^{\prime-1}_{\|B(f^{n-1}(x),\epsilon)}\circ h_{0}\circ f^{n}(x)&\mathrm{if}\;x\in D^{-n},n>0\\\ h_{A}(x)&\mathrm{if}\;x\in A\end{array}\right.$ We notice that $h_{1}$ is well defined since the union (5) is disjoint. Also since $f^{\prime}$ and $f$ are conjugated via $h_{0}$ on $K$, the map $h_{2}$ is continuous on $M_{0}\setminus(A\cup R_{0}\cup U^{})$. As $h_{0}$ is close to the identity, by commutativity of the diagram and expansion of $R_{0}$, for every $x\in R_{0}$, the intersection: $\cap_{n>0}f^{\prime-1}_{\|B(x,\epsilon)}\circ\cdots\circ f^{\prime-1}_{\|B(f^{n-1}(x),\epsilon)}\big{(}B(f^{n}(x),\epsilon\big{)}$ is exponentially decreasing to $\\{h_{0}(x)\\}$. Consequently, for every $x^{\prime}\notin R_{0}\cup U^{}$ close to $x$ and so in $D^{-n}$ with $n$ large, $h_{1}(x)$ belongs to $f^{\prime-1}_{\|B(x,\epsilon)}\circ\cdots\circ f^{\prime-1}_{\|B(f^{n-1}(x),\epsilon)}\big{(}B(f^{n}(x),\epsilon)\big{)}$ and so is close to $h_{0}(x)$. Thus $h$ is continuous at $R_{0}$. For $x\in A$, we recall that the intersection $\cap_{n\geq 0}f^{\prime n}(\mathcal{L}^{\prime\epsilon}_{f^{-n}_{\|U}(x)})$ is exponentially decreasing to $\\{h_{1}(x)\\}$. Also for every $x^{\prime}$ close to $x$ and so in $D^{n}$ with $n$ large, the point $h_{1}(x^{\prime})$ belongs to the intersection $\cap_{k=0}^{n^{\prime}}f^{\prime k}(\mathcal{L}^{\prime\epsilon}_{f^{-k}_{\|U}(x^{\prime})})$ whose diameter is small for $n^{\prime}<n$ large. As $(\mathcal{L}^{\prime}_{\|f^{-k}_{\|U}(x^{\prime})})_{k=0}^{n^{\prime}}$ and $(\mathcal{L}^{\prime}_{\|f^{-k}_{\|U}(x)})_{k=0}^{n^{\prime}}$ are close, $h_{1}(x^{\prime})$ and $h_{1}(x)$ are close. This proves that $h_{1}$ is continuous at $A$. Since $U^{}$ has its closure disjoint from the singularities of $f$, $f$ is a local diffeomorphism on a neighborhood of $cl(U^{})\cap M_{0}$. Consequently, for $\epsilon$-small enough and then $f^{\prime}$ close enough to $f^{\prime}$, the restriction $f^{\prime}_{\|B(x,\epsilon)}$ is a diffeomorphism onto its image, for every $x\in U^{}$. As the connected components of $U^{}$ accumulate on $R_{0}$, we can define similarly for $f^{\prime}$ close enough to $f$: $h_{2}:\;x\in M_{0}\mapsto\left\\{\begin{array}[]{cl}h_{1}(x)&\mathrm{if}\;x\in M_{0}\setminus U^{}\\\ f^{\prime-1}_{\|B(x,\epsilon)}\circ\cdots\circ f^{\prime-1}_{\|B(f^{k-1}(x),\epsilon)}\circ h_{1}\circ f^{k}(x)&\mathrm{if}\;x\in f^{-k}_{\|U^{c}}(U)\setminus f^{-k+1}(U)\end{array}\right.$ The continuity of $h_{2}$ on $M_{0}\setminus R_{0}$ follows from the conjugacy of $f^{\prime}$ and $f$ via $h_{1}$ on $M_{0}\setminus U^{}$. The continuity of $h_{2}$ at $R_{0}$ is proved similarly as we did for $h_{1}$. By expansiveness of $f_{\|A\cup R_{0}}$, $h_{1}$ is injective on $A\cup R^{0}$. Also by definition, $h_{1}$is injective on $M^{0}\setminus(A\cup R_{0}\cup U^{})$. By using the attraction-repulsion of $A-R_{0}$, we get that $h_{1}$ is injective on $M\setminus U^{}$. By local inversion of $f_{\|U^{}}$, we have the injectivity of $h_{2}$. Therefore $h_{2}$ is a continuous injective map from $M_{0}$ into itself, $C^{0}$-close to the identity. By compactness of $M_{0}$ this implies that $h_{0}$ is a homeomorphism of $M_{0}$. Let us finally construct $h$. Let $K^{\prime}$ be a neighborhood of the singularities, included in the interior of $K$ and satisfying properties $(1)$ and $(2)$ of $K$. Let $M_{s}:=M\setminus(M_{0}\cup K^{\prime})$. Since the restriction of $f$ to the $M_{s}$ is a submersion, we can foliate $M_{s}$ by the two following ways. For $x\in M_{s}$, let $m_{x}>0$ be minimal such that $y:=f^{m_{x}}(x)$ belongs to $M_{0}$. Let $F_{y}$ be the submanifold equal to $\cup_{k>0}f^{-k}(\\{y\\})\setminus(M_{0}\cup K^{\prime})$. Let $F^{\prime}_{y}$ be the submanifold equal to $\cup_{k>0}f^{\prime-k}(\\{y\\})\setminus(M_{0}\cup K^{\prime})$. We notice that $(F_{y})_{y}$ and $(F^{\prime}_{y})_{y}$ are the leaves of two foliations on $M_{s}$. We notice that the restriction of the exponential map associated to the metric of $M$: $\phi:\;TF^{\prime\bot}_{y}\rightarrow M$ $(x,u)\mapsto\exp_{x}(u)$ is a diffeomorphism from a neighborhood of the zero section of $TF^{\prime\bot}_{y}\rightarrow F^{\prime}_{y}$ onto a neighborhood $V$ of $F^{\prime}_{y}$. Let $\pi_{y}:\;V\rightarrow F^{\prime}_{y}$ be the composition of $\phi^{-1}$ with the projection $TF^{\prime\bot}_{y}\rightarrow F^{\prime}_{y}$. Let $K^{\prime\prime}$ be a compact neighborhood of $K^{\prime}$ in $int(K)$. We notice that the following map is well defined for $f^{\prime}$ close enough to $f$. $h_{2}^{\prime}:\;x\in M\setminus(M_{0}\cup K^{\prime\prime})\mapsto\pi_{h_{2}(y)}(x),\quad\mathrm{if}\;x\in F_{y}.$ We notice that $h_{0}$ and $h_{2}^{\prime}$ sends every point $x\in K\setminus K^{\prime\prime}$ into $F_{h_{2}(y)}^{\prime}$. Thus we may patch $h^{\prime}_{2}$ and $h_{2}$ to a map $h:\;M\rightarrow M$ satisfying that: * • $h$ is equal to $h_{2}$ on $M_{0}$, to $h_{0}$ on $K^{\prime\prime}$ and to $h_{2}^{\prime}$ on $M\setminus(K\cup M_{0})$, * • $h$ sends any points $x\in M\setminus K^{\prime\prime}$ to a points of $F^{\prime}_{h_{2}(y)}$, with $y:=f^{m_{x}}(x)$, * • The restriction of $h$ to $L$ is a smooth embedding of $\mathcal{L}$ into $M$, * • $h$ is a homeomorphism onto its image $C^{0}$-close to the canonical inclusion. We notice that $h$ is a homeomorphism of $M$ close to the identity. Moreover, it satisfies: $f^{\prime}\circ h=h\circ f.$ ∎ ## 3 Partition of the manifold by invariant and persistent laminations ### 3.1 Construction of the invariant laminations In this subsection we prove Theorem 2.1 which states the existence of a splitting of $M$ into two laminations $(L^{s},\mathcal{L}^{s})$ and $(\hat{R},\mathcal{R})$. #### 3.1.1 The disjoint union of $L$ with $\hat{R}$ is $M$ By the work of Przytycki [Prz77], there exists a neighborhood $V$ of $R$ s.t. $R$ is the maximal invariant of $V$: $R=\bigcap_{n\geq 0}f^{-n}(V).$ As the limit set is included in the non-wandering set, the last equality implies that the complement of the basin of $A$ is $\cup_{n\geq 0}f^{-n}(R)$. Thus the last union is closed and so equal to $\hat{R}$. To summary we have: $\hat{R}=\bigcup_{n\geq 0}f^{-n}(R)\quad\mathrm{and}\quad M=\hat{R}\cup L.$ #### 3.1.2 Construction of $\mathcal{R}$ As the forward orbit of the critical set does not intersect $R$, the backward images of point of $R$ form a partition of $\hat{R}$ by compact submanifolds. Moreover these submanifolds depend continuously on the point, and so are the leaves of a lamination $\mathcal{R}$ on $\hat{R}$. Let $M_{0}$ be the union of the connected components of $M$ which intersect $\Omega$. We notice that the restriction $\mathcal{R}_{\|M_{0}\cap\hat{R}}$ is a $0$-dimensional lamination: its leaves are points. Also since $M$ is compact, there exists $N\geq 0$ s.t. $f^{-N}(M_{0})=M$. Consequently to prove that $(\hat{R},\mathcal{R})$ is a normally expanded lamination, we only need to prove that $\hat{R}\cap M^{\prime}$ is repulsive; this is proved in [IPR08] Lemma 1. #### 3.1.3 Construction of $\mathcal{L}^{s}$ The existence of a laminar structure $\mathcal{L}^{s}$ on $L^{s}$ is a consequence of the following proposition: ###### Proposition 3.1. Let $r\in\\{1,\dots,\infty\\}$, $f$ a $C^{r}$-endomorphism of a manifold $M$ and $K$ a hyperbolic compact subset of $M$. We suppose that the singularities of $f$ have their orbits disjoint from $K$ and that for every $y\in K$, for every $n\geq 0$, the map $f^{n}$ is transverse to a local stable manifold of $y$. Then the union $W^{s}(K)$ of stable manifolds of points of $K$ is the image of a $C^{r}$-lamination $(L,\mathcal{L})$ immersed injectively. Moreover if every stable manifold does not accumulate on $K$, then $(L,\mathcal{L})$ is a $C^{r}$-embedded lamination. ###### Proof. We endow $M$ with an adapted metric $d_{M}$ to the hyperbolic compact subset $K$. For a small $\epsilon>0$, the _local stable manifold of diameter $\epsilon$ of $x\in K$_ is the set of points whose (forward) orbit is $\epsilon$-distant to the orbit of $x$. Let $W^{s}_{\epsilon}(K)$ be the union of stable manifolds of points in $K$ of diameter $\epsilon$. For $\epsilon$ small enough, the closure of $W^{s}_{\epsilon}(K)$ is sent by $f$ into $W^{s}_{\epsilon}(K)$ and supports a canonical $C^{r}$-lamination structure $\mathcal{L}_{0}$. Let $C$ be the subset $W^{s}_{\epsilon}(K)\setminus f^{2}\big{(}cl(W^{s}_{\epsilon}(K))\big{)}$. For $i>0$, we denote by $C_{i}$ the set $f^{-i}(C)$ and by $C_{0}$ the set $W^{s}_{\epsilon}(K)$. Consequently, the union $\cup_{n\geq 0}C_{n}$ is equal to $W^{s}(K)$. Moreover, for $k,l\geq 0$, if $C_{k}$ intersects $C_{l}$ then $\|k-l\|\leq 1$. Let us now construct a metric on $W^{s}(K)$ such that $(C_{n})_{n}$ is an open cover and such that the topology induced by this metric on $C_{n}$ is the same as the one of $M$. For $(x,y)\in W^{s}(K)^{2}$, we denote by $d(x,y)$: $\inf\Big{\\{}\sum_{i=1}^{n-1}d_{M}(x_{i},x_{i+1});\;n>0,\;(x_{i})_{i}\in W^{s}(K)^{n},\;\mathrm{such\;that}$ $x_{1}=x,\;x_{n}=y,\;\forall i\exists j:(x_{i},x_{i+1})\in C_{j}^{2}\Big{\\}}.$ We remark that $d$ is a distance with the announced properties. Let $L$ be the set $W^{s}(K)$ endowed with this distance. We notice that if every stable manifold does not accumulate on $K$, then the topology on $L$ induced by this metric and the metric of $M$ are the same. In other words $L$ is embedded. For $i>0$, we now construct on the open subset $C_{i}$ a $C^{r}$-lamination structure $\mathcal{L}_{i}$. Let $(U_{k},h_{k})_{k}$ be an atlas of $\mathcal{L}_{0\|C}$ of the form: $h_{k}:\;U_{k}\rightarrow\mathbb{R}^{d_{k}}\times T_{k}$ $x\mapsto\big{(}\phi_{k}(x),\psi_{k}(x)\big{)}$ Let $U_{k}^{\prime}:=f^{-i}(U_{k})$ and $\psi_{k}^{\prime}:\;U_{k}^{\prime}\rightarrow T_{k}$ $x\mapsto\psi_{k}\circ f^{i}(x)$ By shrinking a slice $U_{k}$ (and hence $U_{k}^{\prime}$) and by using the transversality of $f$, there exists a neighborhood $T_{k}^{\prime}$ of any $t\in T_{k}$, such that $(\psi_{k}^{\prime-1}(t^{\prime}))_{t^{\prime}\in T_{k}^{\prime}}$ is a family of manifolds that are all diffeormorphic to $M_{t}:=\psi_{k}^{\prime-1}(t)$ by a diffeomophism that depends $C^{r}$-continuously on $t^{\prime}\in T_{k}^{\prime}$. Let us denote by $\phi_{k}^{\prime}:U_{k}^{\prime\prime}\rightarrow M_{t}$ this continuous family of $C^{r}$ diffeormorphisms, with $U^{\prime\prime}_{k}:=\Psi_{k}^{\prime-1}(T_{k}^{\prime})$. Now let $(U_{\alpha},h_{\alpha})_{\alpha\in A}$ be a $C^{r}$-atlas of the manifold $M_{t}$. For each $\alpha$, let $U_{k}^{\alpha}$ be the open subset of $U_{k}^{\prime\prime}$ equal to $\phi_{k}^{\prime-1}(U_{\alpha})$. We notice that the map: $h_{k}^{\alpha}:\;U_{k}^{\alpha}\rightarrow\mathbb{R}^{d}_{\alpha}\times T^{\prime}_{k}$ $x\mapsto(h_{\alpha}\circ\phi_{k}^{\prime}(x),\psi^{\prime}_{k}(x))$ is a chart of an atlas of lamination $\mathcal{L}_{i}$ on $C_{i}$, for $k\geq 0$, $t\in T_{k}$, and $\alpha\in A$. Moreover, for $i,j$ consecutive, the restriction of $\mathcal{L}_{i}$ and $\mathcal{L}_{j}$ to $C_{i}\cap C_{j}$ are equivalent. Thus, the structures $(\mathcal{L}_{i})_{i\geq 0}$ span a $C^{r}$-lamination structure $\mathcal{L}$ on $L$. ∎ ### 3.2 Persistence of the laminations In this section we prove Theorem 2.2. The existence of $i_{R}$ and $f_{R}$ follows from the fact that $(\hat{R},\mathcal{R})$ is a compact, $r$-normally expanded lamination for any $r\geq 1$. Thus by Theorem 0.1 of [Ber08], the lamination $(\hat{R},\mathcal{R})$ is $C^{r}$-persistent. This means that for $f^{\prime}$ $C^{r}$-close to $f$, there exist a $C^{r}$-embedding $i_{R}^{\prime}$ of $(\hat{R},\mathcal{R})$ into $M$, and a $C^{r}$-endomorphism $f_{R}$ of $(R,\mathcal{R})$ such that the following diagram commutes: $\begin{array}[]{rcccl}&&f^{\prime}&&\\\ &M&\rightarrow&M_{0}&\\\ i^{\prime}_{R}&\uparrow&&\uparrow&i^{\prime}_{R}\\\ &\hat{R}&\rightarrow&\hat{R}&\\\ &&f_{R}&&\end{array}.$ Moreover $i^{\prime}_{R}$ is $C^{r}$-close to the canonical inclusion and $f_{R}$ is $C^{r}$-close to $f_{\|\hat{R}}$. By Theorem 2.1, the lamination $(\hat{R},\mathcal{R})$ embedded by $i_{R}$ is actually of class $C^{\infty}$. Thus to smooth $i^{\prime}_{R}$ and $f_{R}$ we consider a smooth tubular neighborhood (see Section 1.5 of [Ber08]). This is the data of: * • a smooth laminar structure $\mathcal{F}$ on $F:=TM_{\|\hat{R}}/T\mathcal{R}$ such that the leaves of $\mathcal{F}$ are the preimages by $\pi:\;F=TM_{\|\hat{R}}/T\mathcal{R}\rightarrow L$ of the leaves of $\mathcal{L}$ and such that $\pi$ is a smooth submersion, * • an immersion $I$ from the restriction of $\mathcal{F}$ to a neighborhood of the zero section $0_{F}$, such that $I\circ 0_{F}=id_{\hat{R}}$ By compactness of $\hat{R}$, there exists $\epsilon$ such that the restriction of the ball $\mathcal{F}_{x}^{\epsilon}$ centered at $x\in\hat{R}$ and with radius $\epsilon$ in the leaf of $0_{x}$ in $\mathcal{F}$ is sent diffeomorphically by $I$ onto an open subset of $M$. Let $F_{x}^{\epsilon}$ be the intersection of $\mathcal{F}_{x}^{\epsilon}$ with the fiber $F_{x}$ of $F\rightarrow\hat{R}$ at $x$. For $f^{\prime}$ close to $f$, the image by $I$ of $F_{x}^{\epsilon}$ intersects transversally at a unique point $i_{R}(x)$ the image by $i^{\prime}_{R}$ of a plaque $\mathcal{L}_{x}$ of $x$. We notice that $i_{R\|\mathcal{L}_{x}}$ is the composition of $i$ with the holonomy from the transverse section $i(\mathcal{L}_{x})$ to $i_{R}(\mathcal{L}_{x})$ along the foliation $(F_{x^{\prime}}^{\epsilon})_{x^{\prime}\in\mathcal{L}_{x}}$. As these submanifolds and foliations are smooth, $i_{R\|\mathcal{L}_{x}}$ is smooth. As these foliations and manifolds depend continuously on $x$, $i_{R}$ is a smooth morphism of $(\hat{R},\mathcal{R})$ into $M$. As $i_{R}^{\prime}(\mathcal{L}_{x})$ is $C^{r}$-close to $i_{R}(\mathcal{L}_{x})$, $i_{R}$ is $C^{r}$-close to $i$. In particular $i_{R}$ is an immersion. Also by construction $i_{R}$ is injective and so, by compactness of $\hat{R}$, $i_{R}$ is an embedding. Let finally $f^{\prime}_{R}:=x\in R\mapsto\pi\circ I^{-1}_{\|\mathcal{F}_{f(x)}^{\epsilon}}\circ f^{\prime}\circ i_{R}(x)$. The composition of a smooth morphisms $f^{\prime}_{R}$ is smooth. Also one easily notes that $f_{R}$ is equivalent and $C^{r}$ close to $f_{\|R}$. The persistence of $(L^{\prime s},\mathcal{L}^{s}_{\|L^{\prime s}})$ is showed similarly, by using this time Theorem 3.1 of [Ber08]. ## 4 Infinitesimal stability implies stability on the non-wandering set In this section, we prove Theorem 2.3. Throughout this section we denote by $(L,\mathcal{L})$ a $C^{\infty}$-lamination, $C$ a compact subset of $L$. Let $f$ be a proper, smooth endomorphism of the lamination $(L,\mathcal{L})$ s.t. for some $N\geq 0$, $x\in C$ there exists $n\leq 0$ s.t. $f^{n}(x)$ does not belong to $C$. ### 4.1 Stability under deformations We are going to prove that infinitesimal stability $(\mathcal{I})$ implies another property called _stability under deformations_ $(\mathcal{D})$. #### 4.1.1 Condition $\mathcal{D}$ ###### Definition 4.1.1. A _deformation of $f$_ is a smooth endomorphism $F$ of the product lamination $(L\times\mathbb{R},\mathcal{L}\times\mathbb{R})$, of the form: $F\;:\;(x,t)\in L\times\mathbb{R}\mapsto(f_{t}(x),t)\in L\times\mathbb{R},$ and such that $f_{0}=f$. We remind that the leaves of the lamination $\mathcal{L}\times\mathbb{R}$ are the product of the leaves of the lamination $\mathcal{L}$ with $\mathbb{R}$. ###### Definition 4.1.2. Let $B$ be a neighborhood of $0\in\mathbb{R}$. A deformation $F=(f_{t})_{t}$ of $f$ is _trivial relatively to $C\times B$_ if there exists a deformation $H$ of the identity of $(L,\mathcal{L})$: $H:\;(x,t)\in L\times\mathbb{R}\mapsto(h_{t}(x),t),$ such that for all $(x,t)\in C\times B$: $h_{t}\circ f(x)=f_{t}\circ h_{t}(x).$ The automorphism $H$ is a _trivialization_ of $F$ (relatively to $C\times B$). ###### Definition 4.1.3 (Condition $\mathcal{D}$). We say that $f$ is _stable under deformations relatively to $C$_ if for any bounded ball $B$ centered at $0$, for any deformation $F:\;(x,t)\mapsto(f_{t}(x),t)$ of $f$ $CO^{\infty}$-close enough to $F_{0}:=(x,t)\mapsto(f(x),t)$ is trivial relatively to $C\times B$. Also the following proposition is obvious: ###### Proposition 4.1. If $f$ is stable under $k$-deformations relatively to $C$, then $f$ is _structurally stable relatively to $C$_: for any $f^{\prime}$ $CO^{\infty}$-sufficiently close to $f$ there exists an isomorphism $h$ of $(L,\mathcal{L})$ such that for any $x\in C$, we have: $h\circ f(x)=f^{\prime}\circ h(x)$ #### 4.1.2 Sufficiency of the implication $I\rightarrow D$ The main remaining difficulty is to prove the following theorem: ###### Theorem 4.2. Let $f$ be a proper, $C^{\infty}$ endomorphism of a lamination $(L,\mathcal{L})$. Let $C$ be a compact subset of $L$ such that for some $N>0$ the intersection $\cap_{n=0}^{N}f^{-n}(C)$ is empty. If $f$ is transversally bijective and if: * $\mathcal{I}$) the following map is surjective: $\chi^{\infty}(\mathcal{L})\rightarrow\chi^{\infty}(f)$ $\sigma\mapsto\sigma\circ f-Tf\circ\sigma,$ Then: * $\mathcal{D}$) The morphism $f$ is stable under deformations relatively to $C$. By the previous proposition, this theorem implies Theorem 2.3. ### 4.2 Condition $\mathcal{I}\Rightarrow$ condition $\mathcal{D}$ Let $f$ be infinitesimally stable. We want to prove that $f$ satisfies condition $\mathcal{D}$. Let $\chi^{\infty}(\mathcal{L},\mathbb{R})^{0}$ be the space of smooth vector fields with $\mathbb{R}$ component equal to 0. Let $\frac{\partial}{\partial t}$ be the canonical unit vector field of the product $\mathcal{L}\times\mathbb{R}$ associated to $\mathbb{R}$. #### 4.2.1 Thom-Levin Theorem The following theorem transforms the problem of the existence of trivialization $H$ to a linear problem. This will allow us to solve this problem algebraically. ###### Theorem 4.3 (Thom-Levine theorem adapted). Let $B$ be a subset of $\mathbb{R}$ and let $W$ be a neighborhood of $C\times B$ in $L\times\mathbb{R}$. There exists a $\mathcal{W}^{\infty}$-neighborhood $V_{\xi}$ of $0\in\chi^{\infty}(\mathcal{L}\times\mathbb{R})^{0}$. A deformation $F$ of $f$ is trivial relatively to $C\times B$, if there exists $\xi\in V_{\xi}$ such that on $W$: $\tau_{F}:=TF\circ\frac{\partial}{\partial t}_{\|L\times\mathbb{R}}-\frac{\partial}{\partial t}\circ F=TF\circ\xi-\xi\circ F.$ ###### Remark 4.2.1. Actually the statement of the Thom-Levin theorem is for vector field of manifold and is interested in equivalence and not conjugacy as here. Nevertheless the proof of this theorem is an adaptation of the one written by Golubitsky-Guillemin [GG73] P123-127. ###### Remark 4.2.2. The above theorem remains true if $C$ is possibly non-compact but closed in $L$. For such a generalization, the proof bellow works as well. Before proving Theorem 4.3, we need a few lemmas. ###### Lemma 4.2.3. Let $\xi$ be a $\mathcal{W}^{\infty}$-small vector field on $\mathcal{L}\times\mathbb{R}$ with zero $\mathbb{R}$-component. Then there is an automorphism $H$ of $(L\times\mathbb{R},\mathcal{L}\times\mathbb{R})$, which is a deformation of $id_{L}$ satisfying: $TH\circ\frac{\partial}{\partial t}\circ H^{-1}=-\xi+\frac{\partial}{\partial t},$ Moreover, $H$ is $\mathcal{W}^{\infty}$-close to the identity. ###### Proof. See [GG73] Sublemma 3.4 p125. ∎ ###### Lemma 4.2.4. By using the same notations as in the above lemma, we have: * $(i)$ $\xi=-T\pi\circ TH\circ\frac{\partial}{\partial t}\circ H^{-1}$, * $(ii)$ $\xi=TH\circ T\pi\circ TH^{-1}\circ\frac{\partial}{\partial t}$, where $\pi\;:\;L\times\mathbb{R}\rightarrow L$ is the canonical projection. ###### Proof. The first statement of this lemma is obvious since $T\pi\circ\frac{\partial}{\partial t}=0$ and $T\pi\circ\xi=\xi$. Applying $TH^{-1}$ to both sides of the equation of Lemma 4.2.3, we get: $TH^{-1}\circ\big{(}-\xi+\frac{\partial}{\partial t}\big{)}=\frac{\partial}{\partial t}\circ H^{-1}.$ Applying $T\pi$ on both sides above, we have: $-T\pi\circ TH^{-1}\circ\xi+T\pi\circ TH^{-1}\circ\frac{\partial}{\partial t}=0.$ As the $\mathbb{R}$-component of $\xi$ is $0$, so is the $\mathbb{R}$ component of $TH^{-1}\circ\xi$. Thus: $TH^{-1}\circ\xi=T\pi\circ TH^{-1}\circ\frac{\partial}{\partial t}.$ By applying $TH$ on both sides of the above equation, we have $(ii)$. ∎ ###### Proof of Theorem 4.3. It is sufficient to show the existence of a deformation $H$ of the identity such that: $F^{\prime}=H^{-1}\circ F\circ H,$ is equal to the trivial deformation $F_{0}$ on $C\times B$. We notice that this holds if and only if the following vector field is $0$ on $C\times B$: $\tau_{F^{\prime}}:=T\pi\circ TF^{\prime}\circ\frac{\partial}{\partial t}.$ (6) By assumption, there exists a $\mathcal{W}^{\infty}$-small vector field $\xi$ on $(L\times\mathbb{R},\mathcal{L}\times\mathbb{R})$, whose $\mathbb{R}$ component is zero and such that: $\tau_{F}=TF\circ\xi-\xi\circ F,\quad\mathrm{on}\;W.$ Let us construct $H$ and so $F^{\prime}$ such that $\tau_{F^{\prime}}$ is zero on $C\times B$. By applying Lemmas 4.2.3 and 4.2.4, we get the existence of a deformation: $H:\;L\times\mathbb{R}\rightarrow L\times\mathbb{R}$ so that $\xi=-T\pi\circ TH\circ\frac{\partial}{\partial{t_{k}}}\circ H^{-1}.$ As $H$ is $\mathcal{W}^{\infty}$-close to the identity, it sends $C\times B$ into $W$. We recall that $F^{\prime}:=H^{-1}\circ F\circ H$. Let $p$ be in $C\times B$, $q:=H(p)$ and $r:=F\circ H(p)$. Using the fact that $F$ and $H$ are deformations, we have that: $TF^{\prime}\circ\frac{\partial}{\partial t}(p)=\frac{\partial}{\partial{t}}\circ F^{\prime}(p)+T\pi\circ TH^{-1}\circ\frac{\partial}{\partial t}(r)+TH^{-1}\Big{[}T\pi\circ TF\circ\frac{\partial}{\partial{t}}(q)-TF\circ\xi(q)\Big{]}$ One the other hand, by using statement $(ii)$ of Lemma 4.2.4, we have: $T\pi\circ TH^{-1}\circ\frac{\partial}{\partial{t}}(r)=TH^{-1}\circ\xi(r).$ The two last equations imply that $TF^{\prime}\circ\frac{\partial}{\partial t}(p)=\frac{\partial}{\partial{t}}\circ F^{\prime}(q)+TH^{-1}\Big{[}\xi(r)+T\pi\circ TF\circ\frac{\partial}{\partial{t}}(q)-TF\circ\xi(q)\Big{]}$ By assumption: $\tau_{F}=TF\circ\frac{\partial}{\partial{t}}-\frac{\partial}{\partial{t}}\circ F=TF\circ\xi-\xi\circ F,\quad\mathrm{on}\;W$ We notice that $\tau_{F}=T\pi\circ TF\circ\frac{\partial}{\partial t}$. Thus, we have: $\tau_{F^{\prime}}(p):=T\pi\circ TF^{\prime}\circ\frac{\partial}{\partial t}(p)=T\pi\circ TH^{-1}\Big{[}-\tau_{F}(q)+T\pi\circ TF\circ\frac{\partial}{\partial{t}}(q)\Big{]}=0$ ∎ #### 4.2.2 Proof of $\mathcal{I}\Rightarrow\mathcal{D}$ using the Thom-Levine theorem Let $f$, $C$ and $(L,\mathcal{L})$ be as stated in the theorem. In particular $f$ is infinitesimally stable. We want to prove that $f$ is stable by deformation, relatively to $C$. Let $B$ be a bounded ball centered at $0$ and $W$ a neighborhood of $C\times B$ in $L$. By the adaptation of the Thom-Levin theorem, it is sufficient to show, for any $CO^{\infty}$-small deformation $F:\;L\times\mathbb{R}\rightarrow L\times\mathbb{R}$ of $f$, the existence of a $\mathcal{W}^{\infty}$-small vector field $\xi\in\chi^{\infty}(\mathcal{L}\times\mathbb{R})^{0}$ with $\mathbb{R}$-component equal to 0 such that on $W$: $\tau_{F}:=TF\circ\frac{\partial}{\partial t}-\frac{\partial}{\partial t}\circ F=TF\circ\xi-\xi\circ F,$ with $\partial/\partial t\in\chi^{\infty}(L\times\mathbb{R})$ the canonical unit vector field associated to $\mathbb{R}$. ##### Local version We first prove the existence of $\xi$ locally: ###### Proposition 4.4. Let $\hat{P}_{1}$ be a small plaque of $\mathcal{L}$ such that $\hat{P}_{i+1}:=f^{-i}(\hat{P}_{1})$ is disjoint from $\hat{P}_{1}$ for every $i>0$. Let $P_{1}$ be a precompact plaque whose closure is included in $\hat{P}_{1}$. Let $P_{i+1}:=f^{-i}(P_{1})$, for every $i\geq 0$. Let $s>0$ be such that $\hat{P}_{s}$ is non-empty. Then $(\hat{P}_{i})_{i=1}^{s}$ consists of manifolds. Moreover, there exists a family of disjoint open neighborhoods $(U_{i})_{i=1}^{s}$ of $(P_{i})_{i=1}^{s}$ s.t. $f$ sends $\Omega^{\prime}:=\coprod_{i=2}^{s}U_{i}$ into $\Omega:=\coprod_{i=1}^{s}U_{i}$, and s.t. for any deformation $F$ $CO^{\infty}$-close to $F_{0}$, there exists $\xi\in\chi^{\infty}(\mathcal{L}\times\mathbb{R})^{0}$ which is $\mathcal{W}^{\infty}$-close to zero with $\mathbb{R}$-component equal to $0$ and such that : $\tau_{F}(x,t)=TF\circ\xi(x,t)-\xi\circ F(x,t),\quad\forall x\in\Omega^{\prime},\;t\in\mathbb{B}.$ ###### Proof. The fact that $(\hat{P}_{i})_{i}$ is a family of manifolds follows from the transverse bijectivity of $f$. The main interest of the proof is the usefulness of the algebraic tools developed by Mather ([Mat68a], [Mat69], [Mat68b]) and very well written by Tougeron [Tou95]. Let $R_{i}^{\prime}$ be the ring $C^{\infty}(\hat{P}_{i})$ of smooth real functions on $\hat{P}_{i}$. For $t\in\mathbb{R}$, let $R_{i}^{t}$ be the ring $C^{\infty}_{\hat{P}_{i}\times\\{t\\}}(\hat{P}_{i}\times\mathbb{R})$ of smooth germs at $\hat{P}_{i}\times\\{t\\}$ of smooth real functions on $\hat{P}_{i}\times\mathbb{R}$. We notice that $R_{i}^{\prime}$ is isomorphic to the quotient $R_{i}^{t}/I_{i}^{t}$, where $I_{i}^{t}$ is the ideal of $R_{i}^{t}$ formed by the germs equal to $0$ on $\hat{P}_{i}\times\\{t\\}$. For $i\in\\{1,\cdots,s-1\\}$, we notice that the map $\phi_{i}^{t}:\;R_{i}^{t}\rightarrow R_{i+1}^{t}$ $\rho\mapsto\rho\circ F_{0\|\hat{P}_{i+1}\times\mathbb{R}}$ is a ring morphism that satisfies: $\phi_{i}^{t}(I_{i}^{t})\subset I_{i+1}^{t}.$ Also, the morphism $\phi_{i}^{t}$ induces on the quotient $R^{\prime}_{i}\cong R_{i}^{t}/I_{i}^{t}\rightarrow R^{\prime}_{i+1}\cong R_{i+1}^{t}/I_{i+1}^{t}$ the following ring morphism: $\phi^{\prime}_{i}:\;R^{\prime}_{i}\rightarrow R^{\prime}_{i+1}$ $\rho\mapsto\rho\circ f.$ Let $\tilde{R}_{i}$ be the ring $C^{\infty}(\hat{P}_{i}\times\mathbb{R})$ formed by the smooth real functions on $\hat{P}_{i}\times\mathbb{R}$. Let $X$ be in the space of smooth deformations $F:\;\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R}\rightarrow\coprod_{i=1}^{s}\hat{P}_{i}\times\mathbb{R}$ of $f_{\|\coprod_{i=2}^{s}\hat{P}_{i}}$ endowed with the Whitney topology. Let $R_{i}^{X}$ be the ring $C^{0}_{F_{0}}(X,C^{\infty}(\hat{P}_{i}\times\mathbb{R}))$ formed by the germs at $F_{0\|\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R}}$ of continuous maps from $X$ into the space of smooth real functions on $\hat{P}_{i}\times\mathbb{R}$. We notice that $\tilde{R}_{i}$ is isomorphic to the quotient $R_{i}^{X}/I_{i}^{X}$ where $I_{i}^{X}$ denotes the ideal formed by the germs that vanish at $F_{0\|\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R}}$. For $i<s$, we notice that the map $\phi_{i}^{X}:\;R_{i}^{X}\rightarrow R_{i+1}^{X}$ $(\rho_{F})_{F\in X}\mapsto(\rho_{F}\circ F_{\|\hat{P}_{i+1}\times\mathbb{R}})_{F\in X}$ is a ring homomorphism that satisfies $\phi_{i}^{X}(I_{i}^{X})\subset I_{i+1}^{X}.$ Also, the homomorphism $\phi_{i}^{X}$ induces on the quotient $\tilde{R}_{i}\cong R_{i}^{X}/I_{i}^{X}\rightarrow\tilde{R}_{i+1}\cong R_{i+1}^{X}/I_{i+1}^{X}$ the following ring homomorphism: $\tilde{\phi}_{i}:\;\tilde{R}_{i}\rightarrow\tilde{R}_{i+1}$ $\rho\mapsto\rho\circ F_{0\|\hat{P}_{i+1}\times\mathbb{R}}.$ Let us give the formulation of the problem in these algebraic settings. For $i\in\\{1,\dots,s\\}$, we denote by: * • $N_{i}^{\prime}$ the $R_{i}^{\prime}$-module $\chi^{\infty}(\hat{P}_{i})$ of smooth vector fields on $\hat{P}_{i}$. * • $N_{i}^{t}$ the $R_{i}^{t}$-module $\chi_{\hat{P}_{i}\times\\{t\\}}^{\infty}(\hat{P}_{i}\times\mathbb{R})^{0}$ of smooth germs of vector fields on $\hat{P}_{i}\times\mathbb{R}$ at $\hat{P}_{i}\times\\{t\\}$ with $\mathbb{R}$-component equal to zero. * • $\tilde{N}_{i}$ the $\tilde{R}_{i}$-module $\chi^{\infty}(\hat{P}_{i}\times\mathbb{R})^{0}$ of smooth vector fields on $\hat{P}_{i}\times\mathbb{R}$ with $\mathbb{R}$-component equal to zero. * • $N_{i}^{X}$ the $R_{i}^{X}$-module $C^{0}_{F_{0}}(X,\chi^{\infty}(\hat{P}_{i}\times\mathbb{R})^{0})$ of germs at $F_{0\|\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R}}$ of continuous functions from $X$ into $\chi^{\infty}(\hat{P}_{i}\times\mathbb{R})^{0}$. For $i\in\\{2,\dots,s\\}$, let us denote by: * • $M_{i}^{\prime}$ the $R_{i}^{\prime}$-module $\chi^{\infty}(f_{\|\hat{P}_{i}})$ of smooth vector fields along $f_{\|\hat{P}_{i}}$ (i.e. of smooth sections of $f^{}_{\|\hat{P}_{i}}T\hat{P}_{i-1}$), • $M_{i}^{t}$ be the $R_{i}^{t}$-modules $\chi^{\infty}_{\hat{P}_{i}\times\\{t\\}}(F_{0\|\hat{P}_{i}\times\mathbb{R}})^{0}$ of germs at $\hat{P}_{i}\times\\{t\\}$ of vector fields along $F_{0\|\hat{P}_{i}\times\mathbb{R}}$ with $\mathbb{R}$-component equal to zero. * • $\tilde{M}_{i}$ be the $\tilde{R}_{i}$-module $\chi^{\infty}(F_{0\|\hat{P}_{i}\times\mathbb{R}})^{0}$ of smooth vector fields along $F_{0\|\hat{P}_{i}\times\mathbb{R}}$ with $\mathbb{R}$-component equal to zero. * • $M_{i}^{X}$ be the $R_{i}^{X}$-module of germs at $x_{0}$ of continuous sections $\sigma$ of the trivial bundle $X\times\chi^{\infty}(\hat{P}_{i}\times\hat{P}_{i-1}\times\mathbb{R})\rightarrow X$, such that $\sigma(F)$ belongs to $\chi^{\infty}(F_{\|\hat{P}_{i}\times\mathbb{R}})^{0}$. For $i\in\\{2,\dots,s\\}$, the following maps are homomorphisms of respectively $R_{i}^{\prime}$, $R_{i}^{t}$, $\tilde{R}_{i}$ and $R_{i}^{X}$-modules: $\begin{array}[]{cc}\alpha^{\prime}_{i,i}:\;N_{i}^{\prime}\rightarrow M_{i}^{\prime}&\alpha^{t}_{i,i}:\;N_{i}^{t}\rightarrow M_{i}^{t}\\\ \xi\mapsto Tf\circ\xi&\xi\mapsto TF^{0}\circ\xi\\\ &\\\ \tilde{\alpha}_{i,i}:\;\tilde{N}_{i}\rightarrow\tilde{M}_{i}&\alpha^{X}_{i,i}:\;N_{i}^{X}\rightarrow M_{i}^{X}\\\ \xi\mapsto TF^{0}\circ\xi&\quad(\xi_{F})_{F}\mapsto(TF\circ\xi_{F})_{F\in X}\end{array}$ Also for $i\in\\{1,\dots,s-1\\}$, the following maps are module homomorphisms over $\phi_{i}^{\prime}$, $\phi_{i}^{t}$, $\tilde{\phi}_{i}$ and $\phi_{i}^{X}$ respectively: $\begin{array}[]{cc}\alpha^{\prime}_{i,i+1}:\;N_{i}^{\prime}\rightarrow M_{i+1}^{\prime}&\alpha^{t}_{i,i+1}:\;N_{i}^{t}\rightarrow M_{i+1}^{t}\\\ \xi\mapsto\xi\circ f_{\|\hat{P}_{i+1}}&\xi\mapsto\xi\circ F^{0}_{\|\hat{P}_{i+1}\times\mathbb{R}}\\\ &\\\ \tilde{\alpha}_{i,i+1}:\;\tilde{N}_{i}\rightarrow\tilde{M}_{i+1}&\alpha^{X}_{i,i+1}:\;N_{i}^{X}\rightarrow M_{i+1}^{X}\\\ \xi\mapsto\xi\circ F^{0}_{\|\hat{P}_{i+1}\times\mathbb{R}}&\quad(\xi_{F})_{F}\mapsto(\xi_{F}\circ F_{\|\hat{P}_{i+1}})_{F\in X}\end{array}$ For $\delta\in\\{^{\prime},t,\tilde{\;},X\\}$, we denote by $M^{\delta}:=\bigoplus_{i=2}^{s}M_{i}^{\delta}\qquad\mathrm{and}\qquad N^{\delta}:=\bigoplus_{i=1}^{s}N_{i}^{\delta}$ the Abelian sum of the modules $(M_{i}^{\delta})_{i=2}^{s}$ and $(N_{i}^{\delta})_{i=1}^{s}$ respectively. The modules $M^{\delta}$ and $N^{\delta}$ are modules over the rings $A:=\bigoplus_{i=2}^{s}R_{i}^{\delta}$ and $B:=\bigoplus_{i=1}^{s}R_{i}^{\delta}$ respectively. Let $I\cdot M^{\delta}$ (resp. $I\cdot N^{\delta}$) be the submodule of $M^{\delta}$ (resp. $N^{\delta}$) spanned by elements of the form $i_{k}\cdot x_{k}$, with $i_{k}\in I_{k}^{\delta}$, $x_{k}\in M_{k}^{\delta}$ (resp. $x_{k}\in N_{k}^{\delta}$) and $k\in\\{2,\dots,s\\}$ (resp. $k\in\\{1,\dots,s\\}$). Let us consider the (additive) group morphism: $\alpha^{\delta}:\;N^{\delta}\rightarrow M^{\delta}$ $(\xi_{i})_{i=1}^{s}\mapsto\big{(}\alpha_{i,i}^{\delta}(\xi_{i})-\alpha_{i-1,i}^{\delta}(\xi_{i-1})\big{)}_{i=2}^{s}.$ Let us show that the infinitesimal hypothesis implies the surjectivity of the map $\alpha^{\prime}$: for every $\zeta\in\chi^{\infty}(f_{\|\coprod_{i=2}^{s}\hat{P}_{i}})$, we can find a smooth function $r:\;\coprod_{i=1}^{s}\hat{P}_{i}\rightarrow\mathbb{(}0,+\infty)$, which is $f$-invariant ($\forall x\in\coprod_{i=2}^{s}\hat{P}_{i}$, $r\circ f(x)=r(x)$) and such that $r\cdot\zeta$ can be extended to a smooth section $\zeta^{\prime}$ of $f^{}T\mathcal{L}$. Let $\xi^{\prime}$ be a vector field on $\mathcal{L}$ such that: $\xi^{\prime}\circ f-Tf\circ\xi^{\prime}=\zeta^{\prime}.$ Let $\xi^{\prime\prime}:=-\frac{1}{r}\cdot\xi^{\prime}_{\|\coprod_{i=1}^{s}\hat{P}_{i}}$. We notice that $\xi^{\prime\prime}$ is a smooth vector field on $\coprod_{i=1}^{s}\hat{P}_{i}$. Also: $Tf\circ\xi^{\prime\prime}-\xi^{\prime\prime}\circ f=\frac{1}{r}(\xi^{\prime}\circ f-Tf\circ\xi^{\prime})=\frac{\zeta^{\prime}}{r}=\zeta,\quad\mathrm{on\;}\coprod_{i=2}^{s}\hat{P}_{i}.$ Let us show that the proposition is proved if we show that $\alpha^{X}$ sends $I^{X}\cdot N^{X}$ onto $I^{X}\cdot M^{X}$. For $F$ in $X$, let $\tau_{F}:=T\pi\circ TF\circ\frac{\partial}{\partial t}$. We notice that $(\tau_{F})_{F\in X}$ is an element of $I^{X}\cdot M^{X}$. Thus, we have the existence of $(\xi_{F})_{F\in X}\in I^{X}\cdot N^{X}$ and of a neighborhood $X^{\prime}$ of $F_{0\|\coprod_{i=2}^{s}\hat{P}_{i}}$ in $X$ such that for every $F\in X^{\prime}$, we have on $\coprod_{i=2}^{s}\hat{P}_{i}$: $\tau_{F}:=TF\circ\xi_{F}-\xi_{F}\circ F.$ Let us construct $(U_{i})_{i}$. Let $\hat{T}$ be a locally compact metric space such that a small neighborhood $\hat{U}_{1}$ of $\hat{P}_{1}$ is isomorphic to the product $\hat{P}_{1}\times\hat{T}$. Let $\tau_{0}\in\hat{T}$ s.t. this isomorphism sends $\hat{P}_{1}$ to $\hat{P}_{1}\times\\{\tau_{0}\\}$. Since $f$ is transversally bijective, for $\hat{U}_{1}$ small enough, we notice that a neighborhood $\hat{U}_{i}$ of $\hat{P}_{i}$ is canonically isomorphic to $\hat{P}_{i}\times\hat{T}$, and this isomorphism sends $\hat{P}_{i}$ onto $\hat{P}_{i}\times\\{\tau_{0}\\}$. For $\hat{T}$ sufficiently small the open subsets $(\hat{U}_{i})_{i}$ are disjoint. Let $\hat{\Omega}:=\coprod_{i=1}^{s}\hat{U}_{i}$ and $\hat{\Omega}^{\prime}:=\coprod_{i=2}^{s}\hat{U}_{i}$. Let $F_{0}:L\times\mathbb{R}\rightarrow L\times\mathbb{R}$ be the trivial deformation of $f$. Let $\exp$ be the exponential map associated to a complete metric on $\coprod_{i=1}^{s}\hat{P}_{i}$. Let $r\in C^{\infty}(\coprod_{i=1}^{s}\hat{P}_{i}\times\mathbb{R},\mathbb{R})$ be a compactly supported function equal to 1 on a neighborhood of $\coprod_{i=1}^{s}P_{i}\times B$. There exists a $CO^{\infty}$-neighborhood $V_{F}$ of the trivial deformations $F^{0}$ of $f$ and a neighborhood $T$ of $\tau_{0}\in\hat{T}$ such that for every deformation $F\in V_{F}$ and $\tau\in T$ the following map is well defined: $F_{\tau}:\;\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R}\rightarrow\coprod_{i=1}^{s}\hat{P}_{i}\times\mathbb{R}$ $(x,t)\mapsto\left[\begin{array}[]{cl}\Big{(}\exp_{f(x,\tau_{0})}\big{(}r(x,t)\cdot\exp^{-1}_{f(x,\tau_{0})}\big{(}F(x,\tau,t)\big{)}\big{)},t\Big{)}&\mathrm{if}\;r(x,t)\not=0\\\ \big{(}f(x,\tau_{0}),t)&\mathrm{else}\end{array}\right.$ with in particular the restriction of $F$ to $\coprod_{i=2}^{s}P_{i}\times\\{\tau\\}\times\mathbb{R}$ canonically identitified to a map from $\coprod_{i=2}^{s}P_{i}\times\mathbb{R}$ into $\coprod_{i=1}^{s}\hat{P}_{i}\times\mathbb{R}$. Also, when $F$ is $CO^{\infty}$-close to $F_{0}$ and $\tau$ is close to $\tau_{0}$, then $F_{\tau}$ is $\mathcal{W}^{\infty}$-close to $F_{0\|\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R}}$. We suppose $V_{F}$ and $T$ sufficiently small such that $F_{\tau}$ belongs to $X^{\prime}$ and such that $r\circ F_{\tau}$ is equal to 1 on $\coprod_{i=2}^{s}P_{i}\times B$, for every $F\in V_{F}$ and $\tau\in T$. Let $\rho\in C^{0}(\hat{T},\mathbb{R})$ be a function equal to $1$ on a neighborhood $T^{\prime}$ of $\tau^{0}\in T$ and to 0 off $T$. For $F\in V_{F}$, let $U_{i}:=P_{i}\times T^{\prime}$, $\Omega:=\coprod_{i=1}^{s}U_{i}$, $\Omega^{\prime}:=\coprod_{i=2}^{s}U_{i}$ and: $\xi:=z\in L\times\mathbb{R}\mapsto\left\\{\begin{array}[]{cl}\rho(\tau)\cdot r(x,t)\cdot\xi_{F^{\tau}}(x,t)&\mathrm{if}\;z=(x,\tau,t)\in\coprod_{i=1}^{s}\hat{P}_{i}\times T\times\mathbb{R}\\\ 0&\mathrm{else}\end{array}\right..$ We notice that $\xi$ belongs to $\chi^{\infty}(\mathcal{L}\times\mathbb{R})^{0}$ and that we have on $\Omega^{\prime}$: $\tau_{F}:=T\pi\circ TF\circ\frac{\partial}{\partial t}=TF\circ\xi-\xi\circ F.$ Also when $F$ is $CO^{\infty}$-close to $F_{0}$, then $\xi$ is $\mathcal{W}^{\infty}$-small. Hence the proposition is shown. The proof that the surjectivity of all $(\alpha^{t})_{t\in\mathbb{R}}$, implies the surjectivity of $\tilde{\alpha}$ is easy. It will be done at the end. To show that the surjectivity of $\alpha^{\prime}$ implies the one of $\alpha^{t}$, and that the one of $\tilde{\alpha}$ implies the one of $\alpha^{X}$ and that $\alpha^{X}(I^{X}\cdot N^{X})=I^{X}\cdot M^{X}$, we shall use the following techniques of Mather. ##### The algebraic Machinery Let $R$ and $S$ be rings with units. Let $I$ and $J$ be _Jacobson ideals_ (this means that for every $z\in J$, the element $1+z$ is invertible) in $R$ and $S$ respectively. Let $\phi:\;R\rightarrow S$ be a ring homomorphism which sends $I$ into $J$. ###### Definition 4.2.5. The homomorphism $\phi:\;(R,I)\rightarrow(S,J)$ is _adequate_ if the following condition is satisfied: Let $A$ be a finitely generated $R$-module. Let $B$ and $C$ be $S$-modules, with $C$ finitely generated over $S$. Let $\beta:\;B\rightarrow C$ be a homomorphism of $S$-modules. Let $\alpha:\;A\rightarrow C$ be a homomorphism over $\phi$ (i.e. $\alpha(a+b)=\alpha(a)+\alpha(b)$ and $\alpha(r\cdot a)=\phi(r)\cdot\alpha(a)$, for $a,b\in A$ and $r\in R$). Suppose that: $\alpha(A)+\beta(B)+J\cdot C=C.$ Then we can conclude $\alpha(A)+\beta(B)=C\qquad\mathrm{and}\qquad\alpha(I\cdot A)+\beta(J\cdot B)=J\cdot C.$ Let us illustrate the above definition by the following non-trivial examples shown by Mather in [Mat68a]-[Mat69]-[Mat68b] and rewritten in this form by Tougeron [Tou95]: ###### Theorem 4.5. For any $i\in\\{2,\dots,s\\}$, the ring homomorphisms: $\phi_{i}^{t}:\;(R_{i}^{t},I_{i}^{t})\rightarrow(R_{i+1}^{t},I_{i+1}^{t})$ and $\phi_{i}^{X}:\;(R_{i}^{X},I_{i}^{X})\rightarrow(R_{i+1}^{X},I_{i+1}^{X})$ are adequate. This is the algebraic theorem of Mather: ###### Theorem 4.6 (Mather). Let $(R_{i},I_{i})$, $i=1,\dots,s$ be rings with units where $I_{i}$ is a Jacobson ideal for every $i$. Let $(R_{1},I_{1})\stackrel{{\scriptstyle\phi_{1}}}{{\rightarrow}}(R_{2},I_{2})\stackrel{{\scriptstyle\phi_{2}}}{{\rightarrow}}\cdots\stackrel{{\scriptstyle\phi_{s-1}}}{{\rightarrow}}(R_{s},I_{s})$ be a sequence of adequate homomorphisms. For every $i$, let $N_{i}$ and $M_{i}$ be $R_{i}$-modules, finitely generated (with possible exception for $N_{s}$). Put for $i<j$: $\phi_{ij}=\phi_{j}\circ\cdots\circ\phi_{i}$ and $\phi_{ii}$ the identity of $R_{i}$. For $j\geq i$, let $\alpha_{ij}:\;N_{i}\rightarrow M_{j}$ be a module-homomorphism over $\phi_{ij}$. Let $N:=\oplus_{i=1}^{s}N_{i}$ and $M:=\oplus_{i=1}^{s}M_{i}$ as the direct sums of Abelian groups and let $\alpha:\;N\rightarrow M$ be given by $\alpha(\xi_{1},\dots,\xi_{s})=(\alpha_{11}(\xi_{1}),\alpha_{12}(\xi_{1})+\alpha_{22}(\xi_{2}),\dots,\alpha_{1s}(\xi_{1})+\dots+\alpha_{ss}(\xi_{s}).$ Suppose that (I) $\alpha(N)+\sum_{i=1}^{s}I_{i}M_{i}=M.$ Then: * (II) $\alpha$ sends $N$ onto $M$. * (III) Moreover $\alpha$ sends $\sum_{i=1}^{s}I_{i}\cdot N_{i}$ onto $\sum_{i=1}^{s}I_{i}\cdot M_{i}$. ###### Remark 4.2.6. We can illustrate the morphism $\alpha$ by the following diagram: $\begin{array}[]{rcccccl}(R_{1},I_{1})&\stackrel{{\scriptstyle\phi_{1}}}{{\rightarrow}}&(R_{2},I_{2})&\stackrel{{\scriptstyle\phi_{2}}}{{\rightarrow}}&\cdots&\stackrel{{\scriptstyle\phi_{n-1}}}{{\rightarrow}}&(R_{s},I_{s})\\\ &&&&&&\\\ N_{1}&\oplus&N_{2}&\oplus&\cdots&\oplus&N_{s}\\\ &&&&&&\\\ \alpha_{11}\downarrow&\alpha_{12}\searrow&\alpha_{22}\downarrow&&\searrow\cdots\searrow&\searrow&\downarrow\alpha_{ss}\\\ &&&&&&\\\ M_{1}&\oplus&M_{2}&\oplus&\cdots&\oplus&M_{s}.\end{array}$ ###### Remark 4.2.7. For $s=1$ this theorem is the Nakayama’s lemma. For $s=3$ is was shown by F. Latour. As the proof is purely algebraic, we will prove this theorem at the end of this work. Let us conclude the proof of the proposition. If we omit the exponent $X,t,\;^{\prime}\;,\tilde{\;}$, and we put $M_{1}=0$ and $\alpha_{ji}=0$ when $j<i-1$, it follows from the last theorem that the surjectivity of $\alpha^{\prime}$ implies the one of $\alpha^{t}$, and that the surjectivity of $\tilde{\alpha}$ implies the one of $\alpha^{X}$ with $\alpha^{X}(I^{X}N^{X})=I^{X}M^{X}$. Thus it only remains to prove that the surjectivity of all $(\alpha^{t})_{t\in\mathbb{R}}$ implies the one of $\alpha^{\prime}$. Let $\tau\in\chi^{\infty}(F_{0})^{0}$. For $t\in\mathbb{R}$, as $\alpha^{t}$ is surjective, there exists a germ $\xi_{t}\in A_{t}$ such that $\tau_{t}:=\alpha_{t}(\xi_{t})$. The germs $\xi_{t}$ is defined on a neighborhood $V_{t}$ of $\coprod_{i=1}^{s}\hat{P}_{i}\times\\{t\\}$ in $\coprod_{i=1}^{s}\hat{P}_{i}\times\mathbb{R}$. By shrinking a slice $\hat{\Omega}$ and then by shrinking $V_{t}$ for every $t\in\mathbb{R}$, we may suppose that $V_{t}$ is of the form $(\coprod_{i=1}^{s}\hat{P}_{i})\times W_{t}$, where $W_{t}$ is a neighborhood of $t\in\mathbb{R}$. Let $(W_{j})_{j\in\mathbb{N}}$ be a locally finite subcovering of $(W_{t})_{t\in\mathbb{R}}$. By locally finite we mean that there exists for every point $t\in\mathbb{R}$ a finite number of integers $j\in\mathbb{N}$ such that $W_{j}$ intersects a neighborhood of $t$. By subcovering we mean that for every $j\geq 0$ there exists $t_{j}$ s.t. $W_{j}$ is included in $W_{t_{j}}$. Let $(\rho_{j})_{j}$ be a partition of the unity subordinate to $(W_{j})_{j}$. Let $\pi^{\prime}:\;\coprod_{i=1}^{s}\hat{P}_{i}\times\mathbb{R}\rightarrow\mathbb{R}$ be the projection on the second coordinate. We notice that: $\pi^{\prime}\circ F_{0}=\pi^{\prime}\quad\mathrm{on}\;\coprod_{i=2}^{s}\hat{P}_{i}\times\mathbb{R},$ since the map $F_{0}$ is a deformation of $f$. Let $\xi:=\sum_{j=1}^{\infty}\rho_{j}\circ\pi^{\prime}\cdot\xi_{t_{i}}$. We notice that $\xi$ is sent by $\alpha^{\prime}$ to $\tau$. This concludes the proof of the proposition. ∎ ###### Remark 4.2.8. We notice that one easily simplifies the above proof to show that under the hypotheses of Theorem 1.7 of Mather, for every deformation $(F_{ij})_{[i,j]\in A}$ which is $\mathcal{W}^{\infty}$-close to the trivial deformation of $(f_{ij})_{[i,j]\in A}$ there exists a $\mathcal{W}^{\infty}$-small vector vector field $\xi\in\chi^{\infty}(\coprod_{i\in V}M_{i}\times\mathbb{R})^{0}$ such that $(TF_{ij}\circ\xi-\xi\circ f_{ij})_{\|M_{i}}=TF_{ij}\circ\frac{\partial}{\partial t}-\frac{\partial}{\partial t}\circ F_{ij\|M_{i}},$ for every $[i,j]\in A$. Thus by using Remark 4.2.2 with $C$ the disjoint union of the manifold from which an arrow start, we get the proof of the Theorem 1.7. ### 4.3 From local to global: proof of Theorem 2.3 Let $f$ be an endomorphism of $\mathcal{L}$ and let $C$ be a compact subset of $L$ as in the statement of the theorem. We notice that by compactness of $C$, there exists a compact neighborhood $V$ of $K$ s.t. for every $x\in V$ some iterate $f^{n}(x)$ does not belong to $V$, for $n\in\mathbb{N}$. Let $W:=f(V)\setminus int(V)$. For every $p\in W$, we define $(p_{i})_{i\geq 1}$ inductively: $P:=\\{p\\}$; $P_{i+1}:=f^{-1}(P_{i})\cap V$. We notice that $P_{N}$ is empty. Let $s_{p}$ be maximal s.t. $P_{s_{p}}$ is not empty. We define $(\hat{P}_{i}^{p})_{i=1}^{s_{p}}$ inductively: $\hat{P}_{1}^{p}$ is a small plaque that contains $p$ and $\hat{P}_{i+1}^{p}:=f^{-1}(\hat{P}_{1}^{p})$. We notice that for $\hat{P}_{1}^{p}$ sufficiently small, $\hat{P}_{i}^{p}$ is disjoint from $\hat{P}_{1}^{p}$ (else we would have a cycle in $V$). Let $P_{1}^{p}$ be a precompact neighborhood of $P_{1}^{p}$ and let $(U_{i}^{p})_{i=1}^{s_{p}}$ be the open subsets provided by Proposition 4.4. Also, by shrinking if necessarily, we may suppose that the closure of $U_{1}^{p}$ does not intersect $C$. As $W$ is compact, we can extract from $(U_{1}^{p})_{p\in C}$ a finite subcover $(U^{p_{j}}_{1})_{j\geq 1}$ of $W$. As $W$ is compact and $\cap_{k=0}^{N}f^{-n}(V)$ is empty, we notice that for every $f^{\prime}$ $CO^{\infty}$-close enough to $f$, for every $x\in C$, there exists an integer $n\geq 1$, s.t.: $f^{n}(x)\in\Delta:=\cup_{j\geq 1}\breve{U}_{1}^{p_{j}}.$ For every $CO^{\infty}$-small deformation $F$ of $f$, for every $j$, there exists $\xi_{j}\in\chi^{\infty}(\cup_{i=1}^{s_{j}}U_{i}^{p_{j}}\times\mathbb{R})^{0}$, such that $\tau_{F}=TF\circ\xi_{j}-\xi_{j}\circ F,\quad\mathrm{on}\;\cup_{i=2}^{s_{j}}U_{i}^{p_{j}}\times\mathbb{R}$ Let $(\rho_{1}^{j})_{j=1}^{N}$ be a partition of the unity subordinate to $(U_{1}^{p_{j}})_{j}$. $\mathrm{Let}\;\rho_{2}^{j}=(x,t)\in L\times\mathbb{R}\mapsto\left\\{\begin{array}[]{cl}\rho_{1}^{j}\circ p_{1}\circ F^{i}(x,t)&(x,t)\in U_{i}^{p_{j}}\times\mathbb{R}\\\ 0&\mathrm{else}\end{array}\right.,$ with $p_{1}:\;L\times\mathbb{R}\rightarrow L$ the canonical projection. Let $R:=\sum_{j=1}^{N}\rho_{2}^{j}$. Since $(U_{i}^{p_{j}})_{i,j}$ is finite, the function $R$ is well defined and is a smooth morphism of $\mathcal{L}\times\mathbb{R}$ into $\mathbb{R}$. Let $\rho_{j}:=\frac{\rho_{2}^{j}}{R}$ and $\xi:=x\mapsto\sum_{j=1}^{N}\rho_{j}(x)\xi_{j}(x)$. As $C$ is disjoint from the closure of $\cup_{j=1}^{N}U_{1}^{p_{j}}$ we have for all $(x,t)\in C\times\mathbb{R}$, $\tau_{F}(x,t)=TF\circ\xi(x,t)-\xi\circ F(x,t).$ ### 4.4 Infinitesimal stability for manifold implies infinitesimal stability for embedded lamination We recall that to prove Theorem 2.3, we only used the surjectivity of the map: $\chi^{\infty}(\coprod_{i=1}^{s}P_{i})\rightarrow\chi^{\infty}(f_{\|\coprod_{i=2}^{s}P_{i}})$ $\sigma\mapsto\sigma\circ f-Tf\circ\sigma.$ But this surjectivity is an easy consequence of the infinitesimal stability of $f$ stated in hypothesis $(ii)$: for every $\tau\in\chi^{\infty}(f_{\|\coprod_{i=2}^{s}P_{i}})$, we can find a smooth function $r:\;\coprod_{i=1}^{s}P_{i}\rightarrow\mathbb{(}0,+\infty)$, which is $f$-invariant ($\forall x\in\coprod_{i=2}^{s}P_{i}$, $r\circ f(x)=r(x)$) and such that $r\cdot\tau$ can be extended to a smooth section $\tau^{\prime}$ of $f^{}TM$. Let $\xi^{\prime}$ be a vector field on $M\setminus\hat{\Omega}$ such that on this domain: $\xi^{\prime}\circ f-Tf\circ\xi^{\prime}=\tau^{\prime}$ Let $\xi^{\prime\prime}:=\frac{1}{r}\cdot\xi^{\prime}_{\|\coprod_{i=1}^{s}P_{i}}$. We notice that $\xi^{\prime\prime}$ is a smooth vector field on $\coprod_{i=1}^{s}P_{i}$. Also $\xi^{\prime\prime}\circ f-Tf\circ\xi^{\prime\prime}=\frac{1}{r}(\xi^{\prime}-Tf\circ\xi^{\prime})=\frac{\tau^{\prime}}{r}=\tau$ Now we have to transform $\xi^{\prime\prime}$ to a vector field tangent to $\coprod_{i=1}^{s}P_{i}$. Let $N_{1}\rightarrow P_{1}$ be the smooth vector bundle whose fiber at $x\in P_{1}$ is $T_{x}P_{1}^{\bot}$. Let $N_{i}\rightarrow P_{i}$ be the smooth vector bundle whose fiber at $x\in P_{i}$ is $T_{x}f^{-i+1}(T_{f^{i-1}(x)}P_{1}^{\bot})$. By transversality of $f$ to $\mathcal{L}$, $P_{1}$ is also a smooth vector bundle. Let $\pi$ be the projection of $TM_{\|\coprod_{i=1}^{s}P_{i}}$ onto $T(\coprod_{i=1}^{s}P_{i})$ parallelly to $\coprod_{i}N_{i}\rightarrow\coprod_{i=1}^{s}P_{i}$. We notice that $\xi:=\pi\circ\xi^{\prime\prime}$ satisfies the requested properties. ### 4.5 Proof of Mather’s algebraic theorem The proof of this theorem comes from an unpublished manuscript of Mather. It was then rewritten by Baas in a manuscript as well unpublished. Here we only copy the last proof. The first step is to show that it is sufficient to prove the theorem for $M_{1}=0$. Let us assume that this has been proved and from this prove the theorem. Put $M^{\prime}=\bigoplus_{i=2}^{s}M_{i}$ and $\alpha^{\prime}\;:\;N\rightarrow M^{\prime}$ is given by $\alpha^{\prime}(n_{1},\dots,n_{s})=(\alpha_{12}(n_{1})+\alpha_{22}(n_{2}),\dots,\alpha_{1s}(n_{1})+\cdots+\alpha_{ss}(n_{s})).$ The hypothesis $(I)$ clearly implies: (I’) $\alpha^{\prime}(N)+\sum_{i=2}^{s}I_{i}M_{i}=M^{\prime}.$ Then the special case of the theorem with $M_{1}=0$ which we assume gives: $\alpha^{\prime}(N)=M^{\prime}\quad\mathrm{and}\quad\alpha^{\prime}\Big{(}\sum_{i=1}^{s}I_{i}\cdot N_{i}\Big{)}=\sum_{i=2}^{s}I_{i}\cdot M_{i}.$ $\mathrm{Let}\quad N_{1}^{\prime}=(\alpha_{12}+\cdots+\alpha_{1,s})^{-1}\alpha\Big{(}\sum_{i=2}^{s}N_{i}\Big{)}.$ It is now sufficient to prove $\alpha_{11}(N^{\prime}_{1})=M_{1}.$ This would give $(I)$ and also imply: $\alpha_{11}(I_{1}N^{\prime}_{1})=I_{1}M_{1}$ which together with $(\alpha_{12}+\cdots+\alpha_{1s})(I_{1}\cdot N_{1}^{\prime})\subset\alpha\Big{(}\sum_{i=2}^{s}I_{i}\cdot N_{i}\Big{)}$ gives $(III)$. Let us prove that $\alpha_{11}(N_{1}^{\prime})=M_{1}$ follows from this implication of $(III^{\prime})$: $M=\alpha(N)+\alpha^{\prime}\Big{(}\sum_{i=1}^{s}I_{i}\cdot N_{i}\Big{)}+I_{1}\cdot M_{1}=\alpha\Big{(}\sum_{i=2}^{s}N_{i}\Big{)}+\alpha^{\prime}(I_{1}\cdot N_{1})+\alpha(N_{1})+I_{1}\cdot M_{1}.$ For $m_{1}\in M_{1}$, there exist $n_{1}^{\prime}\in I_{1}\cdot N_{1}$, $m_{1}^{\prime}\in I_{1}\cdot M_{1}$ and $n_{i}\in N_{i}$, for $i\in\\{1,\dots,s\\}$, such that: $m_{1}\oplus 0\oplus\cdots\oplus 0=\alpha(n_{2}+\cdots+n_{s})+\alpha^{\prime}(n_{1}^{\prime})+\alpha(n_{1})+(m_{1}^{\prime}\oplus 0\cdot\oplus 0).$ Componentwise this gives: $\displaystyle b_{1}=\alpha_{11}(n_{1})+m_{1}^{\prime}$ (7) $\displaystyle 0=(\alpha_{1s}+\cdots+\alpha_{2s})(n_{1})+(\alpha_{12}+\cdots+\alpha_{1s})(n_{1}^{\prime})+\alpha(n_{2}+\cdots+n_{s}).$ (8) From the second equation, it follows that the point $n_{1}+n_{1}^{\prime}$ belongs to $N_{1}^{\prime}$. Hence: $m_{1}=\alpha_{11}(n_{1}+n_{1}^{\prime})+(m^{\prime}_{1}-\alpha_{11}(n^{\prime}_{1}))\in\alpha_{11}(N^{\prime}_{1})+I_{1}M_{1}.$ So we have shown $M_{1}=\alpha_{11}(N^{\prime}_{1})+I_{1}M_{1}$ and by applying Nakayama’s Lemma, we get: $\alpha_{11}(N^{\prime}_{1})=M_{1}.$ And this finishes the proof of the theorem, assuming its holds for $M_{1}=0$. So the next step is to prove the theorem for $M_{1}=0$. This is done by induction on $s$. For $s=1$, the theorem is trivial. So assume the theorem inductively for $s-1$. Put $N_{1}^{}=N_{1}\bigotimes_{R_{1}}R_{2}$ where $R_{2}$ is regarded as an $R_{1}$-module via $\phi_{1}$. Let $\alpha_{1j}^{}:=\alpha_{1j}\otimes\phi_{2j}:\;N_{1}^{}\rightarrow M_{j},\;j\geq 2$ and $N^{}=N_{2}\oplus\cdots\oplus N_{s}\oplus N_{1}^{}$ $M=M_{2}\oplus\cdots\oplus M_{s}$ Define $\alpha^{}:\;N^{}\rightarrow M$ by $\alpha^{}(n_{1},\dots,n_{s})=\Big{(}\alpha_{22}(n_{2})+\alpha_{12}^{}(n_{1}),\dots,\alpha_{1s}^{}(n_{1})+\cdots+\alpha_{ss}(n_{s})\Big{)}$ Clearly $N_{1}^{}$ is a finitely generated $R_{2}$-module and $\alpha_{1j}^{}$ is a homomorphism over $\phi_{1j}$. Now $(I)$ implies $\alpha^{}(N^{})+\sum_{i=2}^{s}I_{i}M_{i}=M.$ And by the induction hypothesis we conclude * $(II^{})$ $\alpha^{}$ sends $N^{}$ onto $M$. $(III^{})$ Moreover $\alpha^{}$ sends $\sum_{i=2}^{s}I_{i}\cdot N_{i}+I_{2}N_{1}^{}$ onto $\sum_{i=2}^{s}I_{i}\cdot M_{i}$. Let $\beta=id\otimes\phi_{1}:\;N_{1}=N_{1}\bigotimes_{R_{1}}R_{1}\rightarrow N_{1}^{}=N_{1}\bigotimes_{R_{1}}R_{2}.$ Then $\alpha^{}_{1j}\circ\beta$ is equal to $\alpha_{1j}$, for every $j$. . Let $C:=(\alpha_{12}^{}+\cdots+\alpha_{1s}^{})^{-1}\cdot\alpha\Big{(}\sum_{i=2}^{s}N_{i}\Big{)}$ and this is an $R_{2}$-submodule of $N_{1}^{}$. Let $n^{}\in M_{1}^{}$ and $m=\alpha^{}(n^{})\in M$. Then by our assumption $(I)$: $m=\alpha(n_{1}+\cdots+n_{s})+m_{2}^{\prime}+\cdots+m_{s}^{\prime},$ where $n_{i}\in N_{i}$ and $m_{i}^{\prime}\in I_{i}\cdot M_{i}$. By $(III^{})$ $m_{2}^{\prime}+\cdots+m_{s}^{\prime}=\alpha^{}(n_{1}^{}+\cdots+n_{s}^{})$ where $n_{1}^{}\in I_{2}\cdot N_{1}^{}$, and $n_{i}^{}\in I_{i}N_{i}$, for $i\geq 2$. Hence $m=\alpha\Big{(}(n_{1}+(n_{2}+n_{2}^{})+\cdots+(n_{s}+n_{s}^{})\Big{)}+\alpha^{}(n_{1}^{}).$ Therefore $\alpha^{}(n^{}-\beta(n_{1})-n_{1}^{})=m-\alpha(n_{1})-\alpha^{}(n_{1}^{})=\alpha\big{(}(n_{2}+n_{2}^{})+\cdots+(n_{s}+n_{s}^{})).$ Put $c=n^{}-\beta(n_{1})-n^{}_{1}\in C$ Since $n^{}=c+\beta(n_{1})+n_{1}^{}\in C+\beta(N_{1})+I_{2}N_{1}^{}$ We get $N_{1}^{}=C+\beta(N_{1})+I_{2}N_{1}^{}.$ Now since $\phi_{2}$ is adequate we deduce $(II^{*})$ $N_{1}^{}=C+\beta(N_{1})$. * $(III^{*})$ $I_{2}N_{1}^{}=I_{2}C+\beta(I_{1}N_{1})$. Clearly $(II^{})$ and $(II^{})$ give $(II)$ and $(III^{})$. Also $(III^{})$ and $(III^{})$ give $(III)$. This finishes the proof of the theorem. $\square$ ### 4.6 Proof of Lemma 1.3.10 Let $\xi\in\chi^{\infty}(f_{\|M^{\prime}})$. By $(c)$ there exists $\sigma_{0}\in\chi(M^{\prime})$ such that: $(\sigma_{0}\circ f-Tf\circ\sigma_{0})_{\|U}=\xi_{\|U}.$ By restricting a slice $U$ s.t. $(a)$ and $(b)$ are still satisfied, we may suppose that $\sigma_{0}$ can be smoothly extended to $M$. We define inductively $(\sigma_{n})_{n\geq 0}\in\prod_{n}\chi(f_{\|\cup_{k=0}^{n}f^{k}(U)})$ by: $\sigma_{n+1}:x\mapsto\left\\{\begin{array}[]{cl}\sigma_{n}(x)&\mathrm{if}\;x\in\cup_{k=0}^{n}f^{k}(U)\\\ \xi\circ f^{-1}(x)+Tf\circ\sigma_{n}\circ f^{-1}(x)&\mathrm{if}\;x\in f^{n+1}(U)\setminus U\end{array}\right.$ We notice that $(\sigma_{n})_{n}$ is well defined and is locally eventually constant on the open subset $U^{+}:=\cup_{k\geq 0}f^{k}(U)$. Thus $(\sigma_{n})_{n}$ converges to some section $\sigma^{+}_{0}\in\chi(U^{+})$. Also $\sigma^{+}$ satisfies: $(\sigma^{+}_{0}\circ f-Tf\circ\sigma^{+}_{0})_{\|U^{+}}=\xi_{\|U^{+}}.$ Let us define $(\sigma_{n}^{+})_{n}\in\prod_{n\geq 0}\chi(f_{\|f^{-n}(U^{+})})$ by induction. Let $n\geq 0$ and suppose $\sigma_{n}^{+}$ constructed. For this end we notice that the forward orbit $O^{+}(\Sigma)$ of the singularities of $f$ is closed and that on $M^{\prime\prime}:=M^{\prime}\setminus O^{+}(\Sigma)$, the map $S:=x\mapsto\ker(T_{x}f)$ is a smooth section of the Grassmannian of $TM^{\prime\prime}$. Let $p$ be the orthogonal projection of $TM^{\prime\prime}$ onto $S$. Remember that $\sigma_{0}$ can be smoothly extended to $M^{\prime}$. This implies that $\sigma_{n}$ can be smoothly extended to $M^{\prime}$. Let $\sigma_{n}^{s}\in\chi(M^{\prime})$ be a smooth extension of $p\circ\sigma_{n\|M^{\prime\prime}}^{+}$ such that $\sigma_{n}^{s}(x)$ belongs to $S(x)$, for every $x\in M^{\prime}$. We can now define inductively for $n\geq 0$: $\sigma_{n+1}^{+}:=x\in f^{-n-1}(U^{+})\mapsto\left\\{\begin{array}[]{cl}\sigma_{n}^{+}(x)&\mathrm{if}\;x\in f^{-n}(U^{+})\\\ (Tf^{-1}_{\|S(x)^{\bot}}\big{(}\sigma_{n}^{+}\circ f(x)-\xi(x)\big{)}&\mathrm{if}\;x\in f^{-n-1}(x)\end{array}\right.$ We notice that $(\sigma_{n}^{+})_{n}$ is well defined and eventually constant on $\hat{U}:=\cup_{n\geq 0}f^{-n}(U^{+})$. Thus $(\sigma_{n}^{+})_{n}$ converges to some section $\hat{\sigma}\in\chi(\hat{U})$. Also $\hat{\sigma}$ satisfies: $(\hat{\sigma}\circ f-Tf\circ\hat{\sigma})_{\|\hat{U}}=\xi_{\|\hat{U}}.$ Let $U^{\prime}$ be an open neighborhood of the singularities satisfying Property $(b)$ of the lemma, such that $U$ contains the closure of $U^{\prime}$. Let $\hat{U}^{\prime}:=\cup_{n\geq 0}f^{-n}(\cup_{k\geq 0}f^{k}(U^{\prime}))$ and let $\breve{U}$ be the complement of $\hat{U}^{\prime}$ in $M^{\prime}$. We notice that $(\hat{U},\breve{U})$ is an open cover of $M^{\prime}$. Thus to finish the proof of the lemma, it is sufficient to find a partition of the unity $(r,1-r)$ subordinate to this cover, which is $f$-invariant ($r\circ f=r)$ and to find a section $\breve{\sigma}\in\chi(\breve{U})$ such that $(\breve{\sigma}\circ f-Tf\circ\breve{\sigma})_{\|\breve{U}}=\xi_{\|\breve{U}}.$ Then we notice that $\sigma:=r\cdot\hat{\sigma}+(1-r)\cdot\breve{\sigma}$ satisfies the requested property. Let $V$ be a manifold with boundary such that $A$ is the maximal invariant of $V$ (i.e $A=\cap_{n}f^{n}(V)$) and $f$ sends $V$ into its interior. Let us construct $r$. As we are here in the diffeomorphism case, the construction a partition of the unity $(r_{1},1-r_{1})$ subordinate to the cover $(V\cap\hat{U},V\cap\breve{U})$ and $f$-invariant is classic. Then we define $r:=x\in M^{\prime}\mapsto r\circ f^{n}(x)$, if $x\in f^{-n}(V)$ which is convenient for our purpose. Let us construct $\breve{\sigma}$. Let $D:=(V\setminus f(V))\cap\breve{U}$. Let $\partial^{+}D:=\partial D\cap f(V)\cap\breve{U}$ and $\partial^{-}D=\partial D\cap int(V)^{c}\cap\breve{U}$, with $\partial D$ the boundary of $D$. On a neighborhood of $\partial^{-}D$ we define $\breve{\sigma}_{+}=0$ and on a neighborhood of $\partial^{-}D$ we define $\breve{\sigma}_{-}=Tf^{-1}\circ(\breve{\sigma}_{+}\circ f-\xi)$. Then we chose a section $\breve{\sigma}_{0}$ of $\chi^{\infty}(M)$ equal to $\breve{\sigma}_{+}=0$ on a neighborhood of $\partial^{+}D$ and to $\breve{\sigma}_{-}$ on a neighborhood of $\partial^{-}D$. As for the construction of $\hat{\sigma}$, we define then $\breve{\sigma}$ on $\breve{U}^{+}:=\cup_{n\geq 0}f^{n}(D)$ and finally on $\breve{U}=\cup_{n\geq 0}f^{n}(\breve{U}^{+})$. $\square$ ## References [AGZV85] V. I. Arnol′d, S. M. Guseĭn-Zade, and A. N. Varchenko. Singularities of differentiable maps. Vol. I, volume 82 of Monographs in Mathematics. Birkhäuser Boston Inc., Boston, MA, 1985. * [Baa73] Nils Andreas Baas. On the models of Thom in biology and morphogenesis. Math. Biosci., 17:173–187, 1973. * [Baa74] N. A. Baas. Structural stability of composed mappings part. i. Preprint, 1974. * [Ber07] P. Berger. Persistence of stratification of normally expanded laminations. arXiv:math.DS, 2007. * [Ber08] P. Berger. Persistence of laminations. arXiv:math.DS, 2008. * [Buc77] M. A. Buchner. Stability of the cut locus in dimensions less than or equal to 6. inventionnes mathematicae, 43(3), 1977. * [dM73] W. de Melo. Structural stability of diffeomorphisms on two-manifolds. Invent. Math., 21:233–246, 1973. * [Duf77] J.-P. Dufour. Sur la stabilité des diagrammes d’applications différentiables. Ann. Sci. École Norm. Sup. (4), 10(2):153–174, 1977. * [GG73] M. Golubitsky and V. Guillemin. In Stable Mappings and Their Singularities, volume 464 of Gratuate Texts in Mathematics. Springer-Verlag, 1973. * [Ghy99] É. Ghys. Laminations par surfaces de Riemann. In Dynamique et géométrie complexes (Lyon, 1997), volume 8 of Panor. Synthèses, pages ix, xi, 49–95. Soc. Math. France, Paris, 1999\. * [IPR97] J. Iglesias, A. Portela, and A. Rovella. Structurally stable perturbations of polynomials in the riemann sphere. To appear in IHP, available at www.premat.fing.edu.uy, 2007/97. * [IPR08] J. Iglesias, A. Portela, and A. Rovella. $c^{1}$ stable maps: Example without saddles. 2008\. Preprint. * [KSvS07] O. Kozlovski, W. Shen, and S. van Strien. Density of hyperbolicity in dimension one. Ann. of Math. (2), 166(1):145–182, 2007. * [Mañ88] R. Mañé. A proof of the $C^{1}$ stability conjecture. Inst. Hautes Études Sci. Publ. Math., (66):161–210, 1988. * [Mat68a] John N. Mather. Stability of $C^{\infty}$ mappings. I. The division theorem. Ann. of Math. (2), 87:89–104, 1968. * [Mat68b] John N. Mather. Stability of $C^{\infty}$ mappings. III. Finitely determined mapgerms. Inst. Hautes Études Sci. Publ. Math., (35):279–308, 1968. * [Mat69] John N. Mather. Stability of $C^{\infty}$ mappings. II. Infinitesimal stability implies stability. Ann. of Math. (2), 89:254–291, 1969. * [Nak89] Isao Nakai. Topological stability theorem for composite mappings. Ann. Inst. Fourier (Grenoble), 39(2):459–500, 1989. * [Pha67] Frédéric Pham. Singularités des processus de diffusion multiple. Ann. Inst. H. Poincaré Sect. A (N.S.), 6:89–204, 1967. * [Prz77] Feliks Przytycki. On $U$-stability and structural stability of endomorphisms satisfying Axiom A. Studia Math., 60(1):61–77, 1977. * [PS70] J. Palis and S. Smale. Structural stability theorems. In Global Analysis (Proc. Sympos. Pure Math., Vol. XIV, Berkeley, Calif., 1968), pages 223–231. Amer. Math. Soc., Providence, R.I., 1970\. * [Rob71] J. W. Robbin. A structural stability theorem. Ann. of Math. (2), 94:447–493, 1971. * [Rob76] C. Robinson. Structural stability of $C^{1}$ diffeomorphisms. J. Differential Equations, 22(1):28–73, 1976. * [Shu69] M. Shub. Endomorphisms of compact differentiable manifolds. Amer. J. Math., 91:175–199, 1969. * [Sma67] S. Smale. Differentiable dynamical systems. Bull. Amer. Math. Soc., 73:747–817, 1967. * [Tho71] René Thom. Modèles mathématiques de la morphogénèse. Accademia Nazionale dei Lincei, 1971. Lezioni Fermiane, Classe di Scienze. * [Tou95] J.-C. Tougeron. Stabilité des applications différentiables (d’après J. Mather). In Séminaire Bourbaki, Vol. 10, pages Exp. No. 336, 375–390. Soc. Math. France, Paris, 1995. Pierre Berger, Institute for Mathematical Sciences, Stony Brook University, Stony Brook, NY 11794-3660 pierre.berger((at))normalesup.org http://www.math.sunysb.edu/~berger/	arxiv-papers	2008-09-01T16:46:05	2024-09-04T02:48:57.591584	{ "license": "Public Domain", "authors": "Pierre Berger", "submitter": "Pierre Berger", "url": "https://arxiv.org/abs/0809.0277" }
0809.0348	# Topological dynamics and dynamical scaling behavior of vortices in a two- dimensional XY model Wei-Kai Qi Institute of Theoretical Physics, Lanzhou University, Lanzhou $730000$, China Yong Chen Email: ychen@lzu.edu.cn Institute of Theoretical Physics, Lanzhou University, Lanzhou $730000$, China ###### Abstract By using topological current theory we study the inner topological structure of vortices a two-dimensional (2D) XY model and find the topological current relating to the order parameter field. A scalar field, $\psi$, is introduced through the topological current theory. By solving the scalar field, the interaction energy of vortices in a 2D XY model is revisited. We study the dynamical evolution of vortices and present the branch conditions for generating, annihilating, crossing, splitting and merging of vortices. During the growth or annihilation of vortices, the dynamical scaling law of relevant length in a 2D XY model, $\xi(t)\propto(t-t^{})^{1/z}$, is obtained in the neighborhood of the limit point, given the dynamic exponent $z=2$. This dynamical scaling behavior is consistent with renormalization group theory, numerical simulations, and experimental results. Furthermore, it is found that during the crossing, splitting and merging of vortices, the dynamical scaling law of relevant length is $\xi(t)\propto(t-t^{})$. However, if vortices are at rest during splitting or merging, the dynamical scaling law of relevant length is a constat. ###### pacs: 47.32.C-, 61.72.Cc, 64.70.qj ## I INTRODUCTION Vortices play an important role in understanding a variety of problems in physics. In the 1970s, Kosterlitz and Thouless constructed a detailed and complete theory of 2D systems Kt . The KT phase transition theory predicts that vortices pair unbinding will lead to a second-order transition in 2D systems, such as superfluid films, superconductors, and XY models two ; Dn . In a 2D XY model, there exist meta-stable states corresponding to vortices, which are closely bound in pairs below a critical temperature. Above the critical temperature, the paired vortices unbind and become free. Vortices in a 2D XY model disrupt the spin alignments even at large distances and correspond to singularities of the order-parameter field Cl . In 2D XY model, vortices are topologically stable configurations. It is found that the high temperature disorder phase with an exponential correlation is a result of the formation of vortices. The critical temperature at which the KT transition occurs is, in fact, that at which vortex generation becomes thermodynamically favorable. At temperatures below this, the system has a power law correlation. In a low temperature phase, vortices appear in a small density of tightly bound dipole pairs. With an increase in temperature, vortex pairs dissociate and become free in the disorder phase. The evolution of vortices plays an important role in the KT transition of the 2D XY model. There has been progress on the study of the defects associated with an n-component vector order parameter field, $\phi(\vec{r},t)$. For the scalar case (n=1), the defects are domain walls, which are points of the spatial dimensionality d=1, lines for d=2, planes for d=3, etc. More generally, for n=d, one has point defects. This leads to vortices for n=d=2. It is interesting to consider the appropriate form for the point defect densities when expressed in terms of the vector order parameter field. This has been carried out by Halperin Ha , and exploited by Liu and Mazenko LM ; however, their analyses are incomplete Duan00 . In a 2D system, a gauge field- theoretical formalism has been developed by Kleinert HK . Furthermore, the gauge theory of topological quantum melting in a $2+1$ dimensional Bose system was developed by Nussinov et al, and the superfluidity and superconductivity can arise in a strict quantum field-theoretical setting Z1 . A topological field theory for topological defects has been developed by Duan et al Duan . By using the $\phi$-mapping method and topological current theory, the evolution of the topological defect that relates to singularities of the order-parameter field, such as the vortex in BEC Duan04 and superconductivity Duan05 , was studied. In this paper, we will discuss the topological quantization and evolution of vortex in two-dimensional XY model. We introduce a scalar field, $\psi$, through the topological current theory. By solving the scalar field, the Hamiltonian of the interaction of vortices in the 2D XY model is revisited. More recently, the dynamical behavior of the 2D XY model following a quench of the temperature to below the KT critical temperature has been studied Br01 . During the quench of a dynamic 2D XY model, aging phenomena and dynamic scaling behavior become extremely important Bo . The assumption of dynamical scaling for the predicted asymptotic growth law the characteristic length, $\xi(t)\propto(t/lnt)^{1/2}$ Br02 , which is also characteristic of the spacing between defects. However, scaling violations were reported. According to an expansion in $\epsilon=4-d$ using standard field-theoretic renormalization group theory, it can be shown that the growing length is given as $\xi(t)\propto t^{1/z}$ for large t, where z is the critical exponent for equilibrium critical dynamics Jan . This standard theoretical approach also shows that the result $\xi(t)\propto t^{1/z}$ is independent of the initial conditions. However, this renormalization group method does not involve the effects of topological defects, such as vortices in the 2D XY model. In the recent work, it was shown that, for the specific case of the 2D XY model, $\xi(t)\propto(t/lnt)^{1/2}$ if free vortices are present, while $\xi(t)\propto t^{1/2}$ if there are no free vortices present in the initial state Brprl . Through our topological current theory for the 2D XY model, the dynamic of vortices is studied, and the branch conditions for generating, annihilating, crossing, splitting and merging of vortices are given. During the growth or annihilation of vortices, the dynamical scaling law of relevant length in the 2D XY model, which is $\xi(t)\propto(t-t^{})^{1/2}$, can be obtained in the neighborhood of the limit point. This indicates that vortices in the 2D XY model are the source of the scaling violations. This dynamical scaling behavior is consistent with numerical simulations LD and experimental results Pa . Furthermore, we have also found that during the crossing, splitting, and merging of vortices, the dynamical scaling law of relevant length in the 2D XY model, which is $\xi(t)\propto(t-t^{})$. It has been shown that why the relevant length of vortices scales with time as $t^{-1/2}$ for small values of times and then deviate toward a linear dependence in nematic liquid-crystal experimental observations Pa2 . However, if vortices are at rest during splitting and merging, the dynamical scaling law of relevant length is $\xi=const$. Moreover, it is worthwhile to note that the dynamical scaling law of vortices, which was deduced from the topological current theory, only depends on topological properties of the order parameter field. The organization of this paper is as follows. In Sec. II, we describe the 2D XY model and the $\phi$-mapping method, and topological current theory is discussed. The Hamiltonian of vortices in the 2D XY model is revisited in section III. The evolution of vortices and dynamical scaling discussed in section IV. Finally, in section V, we summarize our results. ## II Topological Current in the 2D XY model The 2D XY model is a system of spins confined to rotate in the plane of the lattice. The Hamiltonian of the system is give as: $H_{0}=-J\sum_{<ij>}\textbf{S}_{i}\textbf{S}_{j}=-J\sum_{<ij>}\cos(\theta_{i}-\theta_{j}),$ (1) where J is the strength of the nearest-neighbor interaction. $\theta_{i}$ denotes the angle of the spin on site i with respect to arbitrary polar direction in the 2D vector space containing spins. By using the continuum limit, we can approximate $\cos(\theta_{i}-\theta_{j})$ by the first two terms $1-(\theta_{i}-\theta_{j})^{2}/2$ of the Taylor expansion. We can express partial derivatives through $\theta_{i}-\theta_{j}=\partial\theta$ for the two sites i and j. This leads to the continuum Hamiltonian: $H=H_{0}-E_{0}=\frac{J}{2}\int d\textbf{r}(\nabla\theta)^{2},$ (2) where $E_{0}=2JN$ is the energy of the completely aligned ground state of $N$ spins. Kosterlitz and Thouless suggested that the disordering is caused by topological defects, such as vortices in two-dimensional XY model, which are characterized by a mapping from some loop $\Gamma$ in real space onto the order parameter space. For two-dimensional XY model, this implies $\oint d\theta=\oint\nabla\theta d\vec{s}=2\pi W,W=0,\pm 1,\ldots,n,$ (3) where $W$ is the winding number of the vortex. By introducing a 2D XY model, the local order parameter is defined as: $\Psi=\Psi_{0}e^{i\theta}=\phi^{1}+i\phi^{2}.$ (4) Quantized vortices are topological objects associated with topological properties of the order parameter $\Psi$. It is worth noting that the phase of the order parameter is undefined at the vortex core. In other words, vortices correspond to singularities of the order-parameter field. We define the unit vector field $\vec{n}$ as $n^{a}=\frac{\phi^{a}}{\|\|\phi\|\|},~{}\|\|\phi\|\|=\sqrt{\phi^{a}\phi^{a}},~{}a=1,2.$ (5) where $n^{a}n^{a}=1$. From the unit vector, $\vec{n}(x)$, we can construct a topological current of the order parameter field in the 2D XY model, which carries the topological information of $\vec{\phi}(x)$: $j^{k}=\frac{1}{2\pi}\epsilon^{ijk}\epsilon_{ab}\partial_{i}n^{a}\partial_{j}n^{b},\quad i,j,k=0,1,2.$ (6) We will see that this current does not vanish at the zero point of $\vec{\phi}(x)$ or the singularities of the unit field $\vec{n}(x)$. By using the 2D topological current theorem, Eq. (6) can be rewritten in the compact form; $j^{k}=\delta^{2}(\vec{\phi})J^{k}\left(\frac{\phi}{x}\right).$ (7) It is the important relation between the $\delta-$like topological current of vortices and the order parameter, $\Psi$, where $J^{k}(\phi/x)$ is the vector Jacobian of $\vec{\phi}$: $J^{k}\left(\frac{\phi}{x}\right)=\frac{1}{2}\epsilon^{ijk}\epsilon_{ab}\partial_{i}\phi^{a}\partial_{j}\phi^{b},$ (8) or $\displaystyle J^{0}\left(\frac{\phi}{x}\right)$ $\displaystyle=$ $\displaystyle\det\left(\begin{array}[]{cc}\partial_{x}\phi^{1}&\partial_{y}\phi^{1}\\\ \partial_{x}\phi^{2}&\partial_{y}\phi^{2}\end{array}\right),$ (11) $\displaystyle J^{1}\left(\frac{\phi}{x}\right)$ $\displaystyle=$ $\displaystyle\det\left(\begin{array}[]{cc}\partial_{y}\phi^{1}&\partial_{t}\phi^{1}\\\ \partial_{y}\phi^{2}&\partial_{t}\phi^{2}\end{array}\right),$ (14) $\displaystyle J^{2}\left(\frac{\phi}{x}\right)$ $\displaystyle=$ $\displaystyle\det\left(\begin{array}[]{cc}\partial_{t}\phi^{1}&\partial_{x}\phi^{1}\\\ \partial_{t}\phi^{2}&\partial_{x}\phi^{2}\end{array}\right).$ (17) According to the implicit function theorem, the Jacobian’s determinant can be given as: $J^{0}\left(\frac{\phi}{x}\right)=J\left(\frac{\phi}{x}\right)\neq 0,$ (18) The solutions of the zero point of $\vec{\phi}(x)$ can be generally expressed as: $x=x_{l}(t),\quad y=y_{l}(t),\quad l=1,2,\ldots,N,$ (19) which represent $N$ zero points, $\vec{z_{l}}(t)$ (l=1,2,…,N), or a world line of $N$ vertices in space-time. With the $\delta$-function theory, $\delta^{2}(\phi)$, can be expanded as: $\delta^{2}(\phi)=\sum_{l=1}^{N}\frac{\beta_{l}}{\|J(\phi/x)_{z_{l}}\|}\delta^{2}\left(\vec{r}-\vec{z_{l}}(t)\right)$ (20) where the positive integer, $\beta_{l}$, is called the Hopf index of map $x\rightarrow\vec{\phi}$. The meaning of $\beta_{l}$ is that when the point $\vec{r}$ covers the neighborhood of the zero, $\vec{z_{l}}$, once the vector field, $\vec{\phi}$, covers the corresponding region for $\beta_{l}$ times. Using the implicit function theorem and the definition of the vector Jacobian (Eq. 8), we can find the velocity of the $l$-th defect, $\displaystyle\vec{v_{l}}$ $\displaystyle=$ $\displaystyle\frac{d\vec{z_{l}}}{dt}=\left[\frac{\vec{J}(\phi/x)}{J(\phi/x)}\right]_{\vec{z_{l}}}$ $\displaystyle\vec{J}\left(\frac{\phi}{x}\right)$ $\displaystyle=$ $\displaystyle\left[J^{1}\big{(}\phi/x\big{)},J^{2}\left(\phi/x\right)\right]$ (21) The spatial and temporal components of the defect current, $j^{k}$, can be written as the form of the current and the density of a system of $N$ classical point particles moving in a $(2+1)$-dimensional space-time, $\vec{j}=\sum_{l=1}^{N}\beta_{l}\eta_{l}\vec{v_{l}}\delta^{2}(\vec{r}-\vec{z_{l}}(t))$ (22) $j^{0}=\rho=\sum_{l=1}^{N}\beta_{l}\eta_{l}\delta^{2}(\vec{r}-\vec{z_{l}}(t))$ (23) where $\eta_{l}$ is the Brouwer degree, $\eta_{l}=\frac{J(\phi/x)}{\|J(\phi/x)\|}\bigg{\|}_{\vec{z_{l}}}=\pm 1.$ (24) It can clearly be seen that Eq. (22) shows the movement of vertices. The topological charge of vertices in the 2D XY model are conserved: $\frac{\partial\rho}{\partial t}+\nabla\vec{j}=0.$ (25) In addition, there is a constraint of charge neutrality: $\int\rho d^{2}x=\frac{1}{2\pi}\sum_{l=1}^{N}\beta_{l}\eta_{l}=0,$ (26) which indicates that vortices in the 2D XY model appear in pairs. ## III The Hamiltonian of vortices in the 2D XY model revisited In analogy to the velocity field in a superfluid, the distortion field, $\vec{u}=\nabla\theta$, carries the topological information of vortices in the 2D XY model. By using Eq. (4) and (5), we can prove that $\vec{u}=\epsilon_{ab}n^{a}\nabla n^{b},$ and the vorticity is given as: $\nabla\times\vec{u}=\textbf{e}_{i}(\epsilon^{ijk}\epsilon_{ab}\partial_{j}n^{a}\partial_{k}n^{b})$ (27) where $\textbf{e}_{k}(k=1,2,3)$ are the base vectors in the Cartesian coordinate system. Comparing Eq. (27) and Eq. (6), we conclude that $\nabla\times\vec{u}=2\pi j^{0}\textbf{z}.$ (28) Therefore, in the 2D XY model, the vorticity of distortion field, $\nabla\times\vec{u}$, can be expressed in terms of the topological current of the order parameter field. By comparing the $\delta-$like topological current Eq. (7) and Eq. (28), we have the important relation between vorticity and the order parameter field in the 2D XY model: $\nabla\times\vec{u}=2\pi\delta^{2}(\phi)J\left(\frac{\phi}{x}\right)\textbf{z},$ (29) From Eq. (29), we see that the vorticity of distortion field, $\nabla\times\vec{u}$, does not vanish at the zero points of $\Psi$. The location and direction of the $i$th vortex are determined by the $i$th singular point, $\vec{z_{l}}(t)$, and the vector Jacobian, $J(\phi/x)$, on $\vec{z_{l}}(t)$, respectively. In the absence of vorticity, there are no zero values of the order parameter field, $\delta^{2}(\phi)$ is zero, and Eq. (29) becomes: $\nabla\times\vec{u}_{0}=0,$ (30) which is the condition of irrotationality. Thus, Eq. (29) describes both the vortex-state and the irrotationality-state. To describe a collection of vortices at locations $\vec{z_{l}}(t)$, we substitute Eq. (23) into Eq. (28), leading to: $\nabla\times\vec{u}=2\pi\sum_{l=1}^{N}\beta_{l}\eta_{l}\delta^{2}(\vec{r}-\vec{z_{l}}(t))\textbf{z}.$ (31) It can be see that Eq. (31) represents N vortices that are charged with the topological charge, $W$. $W$ is $\beta_{l}\eta_{l}$, which describes the inner topological structure of the vortex. To find the solution for Eq. (31), we can introduce a harmonious scalar field, $\psi$ defining by: $\displaystyle\nabla^{2}\psi$ $\displaystyle=\partial_{i}\partial_{j}\theta-\partial_{j}\partial_{i}\theta$ (32) $\displaystyle=\epsilon^{ij}\epsilon_{ab}\partial_{i}n^{a}\partial_{j}n^{b}.$ In the absence of a vortex in the 2D XY model, we can see that the condition, $\partial_{i}\partial_{j}\theta=\partial_{j}\partial_{i}\theta$, leads to $\nabla^{2}\psi=0$. Thus, the 2D distortion can be written as, $\vec{u}=\vec{u}_{0}-\nabla\times(\textbf{z}\psi)$, which contains two parts: the first part, $\vec{u}_{0}=\nabla\phi$, describe spin waves, while the second part is associated with vortices. By using 2D topological current theorem, it can be shown that: $\nabla^{2}\psi=2\pi\sum_{l=1}^{N}\beta_{l}\eta_{l}\delta^{2}(\vec{r}-\vec{z_{l}}(t)).$ (33) The scalar field, $\psi$, behaves like the potential due to a set of charged particles. The solution of Eq. (33) is given as: $\psi(x)=\sum_{l=1}\beta_{l}\eta_{l}ln(\|\vec{r}-\vec{z_{l}}(t)\|).$ (34) The continuum Hamiltonian of the vortices in Eq. (2) can be rewritten as: $H-\sum_{i}E_{ci}=\frac{J}{2}\int(\nabla\psi)^{2}d^{2}\textbf{x}=-\frac{J}{2}\int\psi\nabla^{2}\psi,d^{2}\textbf{x}$ (35) where $E_{ci}$ is the core energy of the vortex. Substituting Eq. (33) and (34) into the Hamiltonian (35), one can obtain: $\beta H=\beta E_{c}\sum_{i}W_{i}^{2}-\pi K\sum_{ij}W_{i}W_{j}ln(\vec{z_{i}}(t)-\vec{z_{j}}(t)),$ (36) which is identical to the Hamiltonian of a 2D Coulomb gas with point charges of charge $W_{i}=\beta_{i}\eta_{i}$. The constraint given by Eq. (26) is thus the constraint of charge neutrality. ## IV The evolution of vortices in the 2D XY model The zero point of the order parameter field, $\Psi$, which is associated with the locations of the core of vortices, plays an important role in describing the evolution of vortices in the 2D XY model. If the Jacobian determinant, $J^{0}(\phi/x)\neq 0$, we will have the isolated solution of the zeros of the order parameter field in $2+1$ dimensional space-time. But, when $J^{0}(\phi/x)=0$, the above results will change in some way and will lead to the branch process of vortices. We denote one of the vectors Jacobians at the zero points as $(t^{},\vec{z_{l}})$. According to the values of the vector Jacobian at the zero points of the order parameter, there are usually two kinds of branch points, namely, the limit points and bifurcation points KU . Each kind corresponds to different cases of branch processes. Figure 1: Generating and annihilating of vortices pairs. (a) The origin of two vortices. (b) Two vortices annihilate in collision at the limit point. ### IV.1 The generation and annihilation of vortices We explore what will happen to vortices at the limit point $(t^{},\vec{z_{l}})$. The limit points are determined by $\displaystyle J^{0}\left.\left(\frac{\phi}{x}\right)\right\|_{t^{},\vec{z_{l}}}=0,\quad J^{1}\left.\left(\frac{\phi}{x}\right)\right\|_{t^{},\vec{z_{l}}}\neq 0$ (37) $\displaystyle J^{0}\left.\left(\frac{\phi}{x}\right)\right\|_{t^{},\vec{z_{l}}}=0,\quad J^{2}\left.\left(\frac{\phi}{x}\right)\right\|_{t^{},\vec{z_{l}}}\neq 0.$ (38) Considering the condition given by Eq. (37) and making use of the implicit function theorem, the solution of the zero points of $\phi(x)$ in the neighborhood of the point ($t^{},\vec{z_{l}}$) is given as: $t=t(x),\quad y=y(x),$ (39) where $t^{}=t(z_{l}^{1})$. In this case, one can see that: $\left.\frac{dx}{dt}\right\|_{(t^{},\vec{z_{l}})}=\left.\frac{J^{1}(\phi/x)}{J(\phi/x)}\right\|_{(t^{},\vec{z_{l}})}=\infty,$ (40) or $\left.\frac{dt}{dx}\right\|_{(t^{},\vec{z_{l}}}=0.$ (41) The Taylor expansion of $t=t(x)$ at the limit points $(t^{},\vec{z_{l}})$ is given as: $t-t^{}=\left.\frac{1}{2}\frac{d^{2}t}{dx^{2}}\right\|_{t^{},\vec{z_{l}}}(x-z_{l}^{1})^{2},$ (42) which is a parabola in the x-t plane. From this equation, we can obtain two solutions $x_{1}(t)$ and $x_{2}(t)$, which give two branch solutions (world lines of the vortices). If $\left.\frac{d^{2}t}{dx^{2}}\right\|_{(t^{},\vec{z_{l}})}>0,$ we have the branch solutions for $t>t^{}$ which are related to the origin of a dipole pair. Otherwise, we have the branch solutions for $t<t^{}$, which related to the annihilation of a dipole pair (Fig. 1). Figure 2: Two vortices collide with different directions of motion at the bifurcation point in $(2+1)$-dimensional space-time. Two world lines intersect with different directions at the bifurcation point; i.e., two vortices encounter at the bifurcation point. Since the topological current is identically conserved, the topological charges of these two generated or annihilated vortices must be opposite of one another at the limit points, such as: $\beta_{1}\eta_{1}+\beta_{2}\eta_{2}=0.$ (43) This indicates that the vortices always generate and annihilate in pairs. From Eq. (42), one also obtains that the velocity of the vortices is infinite when they are annihilating, which agrees with the result introduced by Bray Bray . Furthermore, a new result can be obtained that shows that the velocity of the vortices is infinite when they are generating, which is gained only from the topology of the order parameter field. For a limit point, it is required that, $J^{1}(\phi/x)\|_{t^{},\vec{z_{l}}}\neq 0$. A bifurcation point, on the other hand, must satisfy a more complex condition. This case will be discussed in the following. ### IV.2 Encountering, Splitting and Merging of vortices Now, let us turn to consider the case in which the restrictions on the zero point $(t^{},\vec{z_{l}})$ are given by: $J^{k}\left.\left(\frac{\phi}{x}\right)\right\|_{(t^{},\vec{z_{l}})}=0,\quad k=0,1,2,$ (44) which imply an important fact that the function relationship between t and x or y is not unique in the neighborhood of the bifurcation point $(t^{},\vec{z_{l}})$. This fact is easily seen from $\frac{dx}{dt}=\left.\frac{J^{1}(\phi/x)}{J(\phi/x)}\right\|_{t^{},\vec{z_{l}}},~{}~{}\frac{dy}{dt}=\left.\frac{J^{2}(\phi/x)}{J(\phi/x)}\right\|_{t^{},\vec{z_{l}}},$ (45) which, under Eq. (44), directly shows the indefiniteness of the direction of the integral curve of Eq. (45) at $(t^{},\vec{z_{l}})$. For this reason, the point $(t^{},\vec{z_{l}})$ is called a bifurcation point of the orientation order parameter. Figure 3: Vortices with the same direction of motion. (a) Two world lines tangentially contact; i.e., two vortices tangentially encounter at the bifurcation point. (b) Two world lines merge into one world line; i.e., two vortices merge into one vortex at the bifurcation point. (c) One world line resolves into two world lines; i.e., one vortex splits into two vortices at the bifurcation point. As we know, at the bifurcation point, $(t^{},\vec{z_{l}})$, the rank of the Jacobian matrix, $\left[\partial\phi/\partial\textbf{x}\right]$, is $1$ in the 2D vector order parameter. With the aim of finding the different directions of all branch curves at the bifurcation point, we assume: $\left.\frac{\partial\phi^{1}}{\partial y}\right\|_{t^{},\vec{z_{l}}}\neq 0.$ (46) From the implicit function theorem, there is one function relation: $y=f(x,t).$ (47) According to the $\phi$-mapping theory, the Taylor expansion of the solution of the zeros of the order parameter field in the neighborhood of $(t^{},\vec{z_{l}})$ can be expressed as: Duan $A(x-z_{l}^{1})^{2}+2B(x-z_{l}^{1})(t-t^{})+C(t-t^{})^{2}+\cdots=0,$ (48) which leads to $A\left(\frac{dx}{dt}\right)^{2}+2B\frac{dx}{dt}+C=0,$ (49) and $C\left(\frac{dt}{dx}\right)^{2}+2B\frac{dt}{dx}+A=0,$ (50) where A, B, and C are constants determined by the order parameter. The solutions of Eq. (49) or Eq. (50) give different directions for the branch curves (world line of the vortices) at the bifurcation point. There are four possible cases which demonstrate the physical meaning of the bifurcation points. Figure 4: Two important cases of Eq. (53). (a) One world line resolves into three world lines; i.e., one vortex resolves into three vortices at the bifurcation point. (b) Three world lines merge into one world line; i.e., three vortices merge into one vortex at the bifurcation point. Case 1 $(A\neq 0)$: For $\Delta=4(B^{2}-AC)>0$, from Eq. (49) we get two different motion directions of the core of the vortex given as: $\left.\frac{dx}{dt}\right\|_{1,2}=\frac{-B\pm\sqrt{B^{2}-AC}}{A},$ (51) which is shown in Fig. 2, where two world lines of two vortices intersect with different directions at the bifurcation point. This shows that two vortices encounter and then depart at the bifurcation point. Case 2 $(A\neq 0)$: For $\Delta=4(B^{2}-AC)=0$, form Eq. (49), we obtain only one motion direction of the core of the vortex given as: $\left.\frac{dx}{dt}\right\|_{1,2}=-\frac{B}{A},$ (52) which includes three important cases (Fig. 3). These three cases are: (i) Two world lines tangentially contact; i.e., two vortices tangentially encounter at the bifurcation point. (ii) Two world lines merge into one world line; i.e., two vortices merge into one vortex at the bifurcation point. (iii) One world line resolves into two world lines; i.e., one vortex splits into two vortices at the bifurcation point. Case 3 $(A=0,C\neq 0)$: For $\Delta=4(B^{2}-AC)=0$, from Eq.(49) we have: $\left.\frac{dt}{dx}\right\|_{1,2}=\frac{-B\pm\sqrt{B^{2}-AC}}{C}=0,~{}-\frac{2B}{C}.$ (53) There are two important cases (Fig. 4): (i) One world line resolves into three world lines; i.e., one vortex splits into three vortices at the bifurcation point. (ii) Three world line merge into one world line; i.e., three vortices merge into one vortex at the bifurcation point. Case 4 (A=C=0): Equations (49) and (50) give respectively: $\frac{dx}{dt}=0,\quad\frac{dt}{dx}=0.$ (54) This case shows that two world lines intersect normally at the bifurcation point, which is similar to case 3. It is no surprise that both parts of Eq. (54) are correct because they give the slope coefficients of two different curves at the same point $(t^{},\vec{z_{l}})$. The remaining components of $dy/dt$ can be calculated from a function relation of Eq. (47): $\frac{dy}{dt}=\frac{\partial f}{\partial x}\frac{\partial x}{\partial t}+\frac{\partial f}{\partial t}.$ (55) The above solutions reveal the evolution of the vortices. The topological structure of the vortices is detailed in the neighborhood of the bifurcation points of the order parameter field. Besides the encountering of vortices, i.e., two vortices encountered at and then depart from the bifurcation point along different branch curves (Fig. 2 and Fig. 3(a)), splitting and merging of vortices are also included. When multicharged vortices pass the bifurcation point, it may split into several vortices along different branch curves (Fig. 3(c), Fig. 4(a)). On the other hand, several vortices can merge into one vortex at the bifurcation point (Fig. 3(b), Fig. 4(c)). As before, since the topological current of the vortices is identically conserved, the sum of topological charges of the final vortices must be equal to that of the initial vortices at the bifurcation point, which is given as (for fixed l): $\sum_{f}\beta_{lf}\eta_{lf}=\sum_{i}\beta_{li}\eta_{li}.$ (56) This indicates that vortices with a higher value of Burgers vector can evolve to the lower value of Burgers vector, or that vortices with a lower value of Burgers vector can evolve to a the higher value of Burgers vector through the bifurcation process. Furthermore, we see that the generation, annihilation, and bifurcation of vortices are not gradual changes, but that they start at a critical value of a parameter, i.e., a sudden change. It is important to note that further bifurcations are possible during the evolution of vortices besides the cases studied in this work. It is necessary need to assume that the terms in the Taylor series considered above vanish and that the expansion of the field is dominated by higher order terms. ### IV.3 Dynamical scaling law of vortices in the 2D XY model In the neighborhood of the limit point, we denote the scale length $l=\nabla x$. The growth velocity or annihilation velocity of vortices is $v=l/\nabla t$ Xu . From Eq. 42), one can obtain the scaling law as: $v\varpropto(t-t^{})^{-1/2}.$ (57) It can be seen that $E_{k}\varpropto(t-t^{})^{-1}$. In the 2D XY model, the growth or annihilation is parameterized in terms of a relevant characteristic length $\xi(t)$, which is also characteristic of the mean vortex-antivortex separation distance. From Eq. (57), we find that the relevant length, $\xi(t)$, obeys the following: $\xi(t)\sim(t-t^{})^{1/2}.$ (58) Equation (58) is the dynamic scaling law of the vortex-antivortex pairs. The low temperature equilibrium phase has essentially no vortex pairs, and $\xi(t)$ is infinite. The relationship between the relevant length and the metastable vortex density below the critical temperature is given as: $\xi(t)=\sqrt{\frac{1}{\rho_{v}}}.$ (59) then the number of vortices that satisfies the power law $N\propto t^{-1}$. The dynamic scaling law of the vortex-antivortex pairs is consistent with renormalization group theory and results from recent numerical simulations using Langevin equations and Monte Carlo methods Brprl ; LD . The specially prepared nematic liquid-crystal system Pa , which is developed, exhibits 2D XY behavior and can be utilized to obtains the dynamical scaling behavior of vortices (Fig. 5). This result is consistent with the expansion in $\epsilon=4-d$ using standard field-theoretic renormalization group theory. This standard theory shows that the growing length, $\xi(t)\propto t^{1/z}$ for large $t$, where $z$ is the critical exponent for equilibrium critical dynamics Jan . This standard theoretical approach also shows that the result $\xi(t)\propto t^{1/z}$ is independent of the initial conditions. However, this renormalization group method does not involve the effects of topological defects, such as vortices in the 2D XY model. Figure 5: Number of defects versus time. Experimental and simulation results are shown. The results of simulation were the average of 20 quenches LD , for which $T^{}=5.0$ initially and is suddenly reduced to $T^{}=0.8$ ($\square$) and $T^{}=0.9$ ($\triangle$). The experimental results, which are obtained by Pargellis et al Pa , were averaged over nine experimental runs ($\bullet$). Solid lines are the expected $N\propto t^{-1}$ scaling. From the assumptions of dynamical scaling law, it has been predicted that the asymptotic growth law of the characteristic length is given by $\xi(t)\propto(t/lnt)^{1/2}$ Br02 ; however, scaling violations were reported. In a recent work, it was shown that the specific case of the 2D XY model is given as $\xi(t)\propto(t/lnt)^{1/2}$ if free vortices are present, while $\xi(t)\propto t^{1/2}$ if there are no free vortices present in the initial state Brprl . Since the topological current is identically conserved, there are no free vortices during the growth or annihilation process. In the neighborhood of the bifurcation point, we denote scale length $\nabla x=l$. From Eqs. (51)-(53), we can then obtain the growth or annihilation velocity of the vortices, which is given as $v\propto const.$ (60) The approximation asymptotic relation of is then $\xi(t)\propto(t-t^{}).$ (61) It can be seen that this is the reason why the relevant length of the vortices scales with time as $t^{-1/2}$ for small values of $t$ and then deviates towards a linear dependence in nematic liquid-crystal experimental observations Pa2 . From Eq. (54), one can obtain $\xi(t)=const,\quad v=0.$ (62) It can be seen that that vortices are relatively at rest when $\xi(t)=const$. During the growth or annihilation of vortices, the dynamical scaling law of the relevant length in the 2D XY model is given as $\xi(t)\propto(t-t^{})^{1/2}$. This indicates that the vortices in the 2D XY model are the source of the scaling violations. During the crossing, splitting and merging of vortices, the dynamical scaling law of relevant length in the 2D XY model is given as $\xi(t)\propto(t-t^{})$. However, if vortices are at rest during merging or splitting, the dynamical scaling law of relevant length is $\xi(t)=const$. Moreover, it is worth noting that the dynamical scaling law of vortices depends only on topological properties of the order parameter field. ## V Conclusions In summary, we have studied the inner structure and evolution of vortices in the 2D XY model by making using of the $\phi$-mapping topological current theory. A scalar field, $\psi$, is introduced through the topological current theory. By solving the scalar field, the interaction energy of vortices in the 2D XY model is revisited. By using the $\phi$-mapping current theory, the densities of vortices in terms of the order parameter field in the 2D XY model are obtained directly from the definition of the topological charges of the vortices. The inner topological structure of the charge of the vortices, which is characterized by the Hopf index and the Brouwer degree, was obtained. By using the 2D topological current theorem, the Hamiltonian of the interaction vortices in the 2D XY model was revisited. This result is identical to the Hamiltonian of a 2D Coulomb gas with point charges of charge $W_{i}=\beta_{i}\eta_{i}$. Furthermore, we have studied the evolution of vortices in the 2D XY model and concluded that there are crucial cases of branch processes in the evolution of vortices when $J^{0}(\phi/x)=0$ and $\phi=0$, i.e., $\eta_{l}$ is indefinite. It is important to note that according to the Landau-Ginzburg approach, there is instability in the 2D XY model at the zero point of the order parameter fields in the low temperature phase. This indicates that vortices are unstable in the low temperature phase. In the high temperature phase, the zero point of the order parameter field is stable. Due to the branch condition, the vortices generate or annihilate at the limit points and encounter, split, or merge at the bifurcation points of the order parameter field. This result also indicates that the velocity of the vortices is infinite when they are being annihilated or generated, which is obtained only from the topological properties of the order parameter field in the 2D XY model. The scaling law of relevant length, which is given as $R(t)\propto(t-t^{})^{1/2}$, can be obtained in the neighborhood of the bifurcation point during the growth or annihilation of vortices. During the crossing, splitting and merging of vortices, the dynamical scaling law of relevant length is $\xi(t)\propto(t-t^{})$. However, if vortices are at rest during merging or splitting, the dynamical scaling law of relevant length is $\xi=const$. The dynamical scaling law of vortices only depends on topological properties of the order parameter field. ## Appendix A Topological Current theory We study a two-component vector order parameter, $\vec{\phi}=(\phi^{1},\phi^{2})$, over the base manifold, M (in this paper $M=R^{2}\bigotimes R$): $\Psi=\phi^{1}+i\phi^{2},\quad\phi^{a}=\phi^{a}(x,y,t),\quad a=1,2.$ (63) Quantized defects are topological objects associated with topological properties of the order parameter $\Psi$. It is worth noting that the phase of the order parameter is undefined at the vortex core. In other words, vortices correspond to singularities of the order-parameter field. Let us define the unit vector field, $\vec{n}$, as: $n^{a}=\frac{\phi^{a}}{\|\|\phi\|\|},\quad\|\|\phi\|\|=\sqrt{\phi^{a}\phi^{a}},\quad a=1,2.$ (64) where, $n^{a}n^{a}=1$. From the unit vector, $\vec{n}(x)$, we can construct a topological current of the order parameter field in the 2D XY model, which carries the topological information of $\vec{\phi}(x)$: $j^{k}=\frac{1}{2\pi}\epsilon^{ijk}\epsilon_{ab}\partial_{i}n^{a}\partial_{j}n^{b},\quad i,j,k=0,1,2.$ (65) It can be seen that this current does not vanish only at the zero point of $\vec{\phi}(x)$, or the singularities of the unit field $\vec{n}(x)$. Obviously, the current given by Eq.(65) is identically conserved: $\partial_{k}j^{k}=0.$ (66) Suppose there is a vortex located at $z_{i}$. The topological charge of the defect is defined by the Gauss map, $n:\partial\sum_{i}\rightarrow S^{1}$: $W(\phi,z_{i})=\frac{1}{2\pi}\oint_{\partial\sum_{i}}\epsilon_{ab}n^{a}dn^{b}.$ (67) Using Stokes’ theorem in the exterior differential form, one can deduce that: $W(\phi,z_{i})=\frac{1}{2\pi}\oint_{\sum_{i}}\epsilon_{ab}\epsilon^{ij}\partial_{i}n^{a}\partial_{j}n^{b}d^{2}x.$ (68) ## Appendix B Two-dimensional topological current Theorem By using the $\phi$-mapping method, we can prove that the topological current given by Eq. (65) is a $\delta$-like current: $j^{k}=\delta(\vec{\phi})J^{k}\left(\frac{\phi}{x}\right).$ (69) From the above formula, it can be seen that the topological current does not vanish only at the zero points of the vector field $\vec{\phi}(x)$. Substituting Eq. (64) into Eq. (65) and considering that: $\partial_{k}n^{a}=\frac{\partial_{k}\phi^{a}}{\|\|\phi\|\|}-\phi^{a}\partial_{k}\left(\frac{1}{\|\|\phi\|\|}\right).$ (70) we have $j^{k}=\frac{1}{2\pi}\epsilon^{ijk}\epsilon_{ab}\partial_{i}\phi^{a}\partial_{j}^{b}\frac{\partial}{\partial\phi^{a}}\frac{\partial}{\partial\phi^{b}}(\ln(\|\|\phi\|\|)),$ (71) If we define the Jacobian as: $\epsilon^{ab}J^{k}\left(\frac{\phi}{x}\right)=\epsilon^{ijk}\partial^{j}\phi^{a}\partial^{k}\phi^{b},$ (72) and, by virtue of the Laplacian relation in $\phi$ space GS $\triangle_{\phi}\left(\ln(\|\|\phi\|\|)\right)=2\pi\delta^{2}(\vec{\phi}),$ (73) where $\triangle_{\phi}=\frac{\partial}{\partial\phi^{a}}\frac{\partial}{\partial\phi^{a}}$ is the two-dimensional Laplacian operator in $\phi$ space, we obtain a $\delta$-function like current. This current is given by: $j^{k}=\delta(\vec{\phi})J^{k}\left(\frac{\phi}{x}\right).$ (74) From the above formula, one can see that the topological current does not vanish only at the zero points of the vector field $\vec{\phi}(x)$. Therefore, it is essential to investigate the solutions of $\vec{\phi}(x)=0$. ###### Acknowledgements. We would like to thank Y. X. Liu, T. Zhu, and Y. S. Duan for their helpful discussion. Y.C. was supported by the SRF for ROCS, SEM, and by the Fundamental Research Fund for Physics and Mathematics of Lanzhou University. ## References * (1) N. D. Mermin, Phys. Rev 176, 250 (1968); N. D. Mermin and H. Wagner, Phys. Rev. Lett. 22, 1133 (1966); V. L. Berezinskii, Sov. Phys.-JETP, 32, 493 (1970); H. E. Stanley and T. A. Kaplan, Phys. Rev. Lett 17, 913 (1966); J. M. Kosterlitz and D. J. Thouless, J. Phys. C 6, 1181 (1973). * (2) K. J. Strandburg, Rev. Mod. Phys. 60, 161 (1988). * (3) D. R. Nelson, Defects and Geometry in Condensed Matter Physics, (Cambridge University Press, Cambridge, 2002). * (4) P. M. Chaikin and T. C. Lubensky, Principles of condensed matter physics, (Cambridge University Press, Cambridge, 1995). * (5) B. I. Halperin, Physics of Defects, edited by R. Balian et al. (North-Holland, Amsterdam, 1981). * (6) F. Liu and G. F. Mazenko, Phys. Rev. B 46, 5963 (1992). * (7) Y. S. Duan, H. Zhang, and L. B. Fu, Phys. Rev. E 59, 528 (1999). * (8) H. Kleinert, Gauge Theory in Condensed Matter, (World Scientific, Singapore, 1989). * (9) J. Zaanen, Z. Nussinov and S. I. Mukhin, Ann. Phys(New York) 310, 181(2004); V. Cvetkovic, Z. Nussinov, S. Mukhin, and J. Zaanen, Europhys. Lett 81, 27001 (2007). * (10) Y. S. Duan and S. L. Zhang, Int. J. Eng. Sci. 29, 1593 (1991); G. H. Yang and Y. S. Duan, Int. J. Theor. Phys. 37, 2371 (1998); Y. S. Duan and H. Zhang, Phys. Rev. E 60, 2568 (1999). * (11) Y. S. Duan and H. Zhang, Eur. Phys. J. D 5, 47 (1999). * (12) Y. S. Duan, H. Zhang, and S. Li, Phys. Rev. B 58, 125 (1998). * (13) A. J. Bray, Adv. Phys. 43, 357 (1994). * (14) B. Zheng, Int. J. Mod. Phys. B 12, 1419 (1998). * (15) A. J. Bary and A. D. Rutenberg, Phys. Rev. E 49, R27 (1994). * (16) H. K. Janssen, B. Schaub, and B. Schmittmann, Z. Phys.B 73, 539 (1989). * (17) A. J. Bray, A. J. Briant, and D. K. Jervis. Phys. Rev. Lett 84, 1503 (2000). * (18) R. Loft and T. A. DeGrand, Phys. Rev. B 35, 8528 (1987). * (19) A. N. Pargellis, S. Green, and B. Yurke, Phys. Rev. E 49, 4250 (1994). * (20) A. N. Pargellis, P. Finn, J. W. Goodby, P. Panizza, B. Yurke and P. E. Cladis, Phys. Rev. A 46, 7765 (1992). * (21) M. Kubicek and M. Marek, Computational Methods in Bifurcation Theory and Dissipative Structures, (Springer, New York, 1983). * (22) A. J. Bray, Phys. Rev. E 55, 5297 (1997). * (23) T. Xu, Phys. Rev. E 72, 036303 (2005). * (24) I. M. Gel’fand and G. E. Shilov, Generalized Function, (Academic, New York, 1964) Vol. 1.	arxiv-papers	2008-09-02T04:40:45	2024-09-04T02:48:57.601825	{ "license": "Creative Commons - Attribution - https://creativecommons.org/licenses/by/3.0/", "authors": "Wei-Kai Qi and Yong Chen", "submitter": "Yong Chen", "url": "https://arxiv.org/abs/0809.0348" }

Subsets and Splits