title
stringlengths 9
24
| content
stringlengths 0
1.31M
|
|---|---|
1.1140879.pdf | Undulator engineering for synchrotron radiation applications
J. M. Slater, S. C. Gottschalk, F. E. James, D. C. Quimby, K. E. Robinson, and A. S. Valla
Citation: Review of Scientific Instruments 60, 1881 (1989); doi: 10.1063/1.1140879
View online: http://dx.doi.org/10.1063/1.1140879
View Table of Contents: http://scitation.aip.org/content/aip/journal/rsi/60/7?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Progresses of synchrotron radiation applications at the NSRL
Rev. Sci. Instrum. 66, 1836 (1995); 10.1063/1.1145798
Infrared synchrotron radiation instrumentation and applications
Rev. Sci. Instrum. 63, 1535 (1992); 10.1063/1.1143014
Circularly polarized synchrotron radiation from the crossed undulator at BESSY
Rev. Sci. Instrum. 63, 339 (1992); 10.1063/1.1142750
New undulator and conventional lines at the Wisconsin Synchrotron Radiation Center (invited)
Rev. Sci. Instrum. 60, 1441 (1989); 10.1063/1.1140959
Perspectives on micropole undulators in synchrotron radiation technology
Rev. Sci. Instrum. 60, 1796 (1989); 10.1063/1.1140907
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
129.120.242.61 On: Sat, 22 Nov 2014 07:35:19Undulator engineering for synchrotron radiation applications
J. M. Slater, S. C. Gottschalk, F. E. James, D. C. Quimby, K. E Robinson, and
A.S. Valla
Spectra Technology, Inc .• 2755 Northup Way. Bellevue. Washington 98004-1495
(Presented on 29 August 1988)
Six undulators have been designed and built by STI since 1980 for synchrotron and FEL
applications. Several design concepts, producing successively higher fields, have been developed
during this period. A wedged-pole hybrid design has been demonstrated to yield the highest field
to date for a given gap-to-wavelength ratio. A simple method of reducing fieid errors has been
demonstrated on the wedged-pole hybrid, and it may lead to significant cost reduction through
relaxation of mechanical and magnet tolerances.
INTRODUCTION
Spectra Technology, Inc. (STI) has been actively involved
with FEL technology since 1979, when a major U.S. pro
gram series began in Seattle, Washington. These programs
have been directed toward the development of efficient visi
ble and IR PELs in a series of technology demonstration
experiments. During the course of this work, extensive capa
bility has been developed in FEL physics, systems engineer
ing, undulators and optical cavities. There has been special
emphasis on undulator (or wiggler) engineering. Six undu
Iators have been delivered to various customers with one
more currently in construction. These devices range from 50
em to 10 m in length, have periods from 2 to 8 em, and fields
to 10 kG. They are used in both FEL and synchrotron emis
sion applications.
During the continual improvement of the undulator
over these nine years, STI has concentrated on obtaining the
highest possible magnetic field strength with simultaneous
high field quality. This has lead from development of pure
permanent magnet systems, \ to hybrids of permanent mag
net with vanadium permendur poles,2 to a new wedged-pole
hybrid.3 The wedged-pole hybrid produces the highest fields
to date for a given gap-to-wavelength ratio. Both radiation
resistant samarium cobalt and the new higher-strength neo
dymium-iron-boron magnets have been used.
The measurement capability necessary to certify undu
lator coherence over the full device length has been devel
oped. Coherence is a strict requirement for FELs and is de
sirable for synchrotron emission when low-emittance beams
are used, but it is not easily achieved due to material and
mechanical imperfections. Typically, field adjustment after
assembly is necessary to achieve full coherence, and an inex
pensive, but accurate, tuning technique for adjusting the
magnetic field to the ideal values under each pole has been
developed.
This article highlights the high field strength wedged
pole design and a tuning method, called shim tuning, to sub
stantially reduce field errors.
I. HIGH FIELD STRENGTH WEDGED~POLE CONCEPT
The rare-earth permanent magnet (REPM) hybrid un
dulator (or wiggler) was originally proposed by Halbach4 as
a means for achieving high-quality high-strength periodic magnetic fields. This concept is gaining widespread ac
ceptance both as an insertion device for synchrotron radi
ation generation and for use in free-electron lasers. The prin
cipal advantages of the REPM-steel hybrid relative to the
pure-REPM undulator include higher magnetic field
strength at small gap-to-period ratio and higher field quality
by making the field distribution less sensitive to magnet in
homogeneities.
The use of wedged poles has now been demonstrated as a
means for increasing the field strength of the hybrid. The
wedged-pole3 configuration can cause the magnet surface
which faces the gap to be driven to the full magnet coercivity
He' thus resulting in higher on-axis field strength. Pole satu
ration is avoided by increasing the cross-sectional areas of
the pole tip without sacrificing magnet volume. Thus, the
design concept has the potential for both higher on-axis field
strength and improved field uniformity by operating the
poles farther from saturation. In addition, widening the pole
tips reduces the harmonic content of the field distribution. It
should be noted that wedged poles have previously been put
to use,5 but the geometric configuration of the permanent
magnets was not modified to exploit the advantages of the
wedged-pole shape.
The geometry of the wedged-pole concept and its field is
compared with the more conventional pure-REPM and hy
brid undu!ator concepts in Fig. 1. The pure-REPM undula
tor, used as the reference, consists of an array of permanent
magnet blocks, whereas the magnets are sandwiched
between highly permeable steel poles of rectangular cross
section in the conventional hybrid geometry. In the pure
REPM device, the field distribution is determined by the
strength and magnetic orientation of the magnet blocks.
The wedged-pole concept shown is an improvement
which is intended to alleviate some of the limitations that
occur in the basic hybrid geometry. In the conventional hy
brid, the on-axis field strength is maximized when the poles
are considerably narrower than the magnets. This not only
leads to considerable higher-order harmonic content in the
field distribution, but also implies that the achievable field
strength is limited by pole tip saturation.
In Fig. 1, the pure-REPM reference system is assumed
to have square blocks with unity fill factors. For both hy
brids the average magnet operating point is taken to be ap
proximately O.2B, (see Ref. 3 for additional detail). For
each full gap (g) to wavelength ().) ratio, the relative advan-
1581 Rev. Sci. Instrum. 60 (7), July 1989 0034-6748/89/071 8tU -04$01.30 @ 1989 American Institute of Physics 1881
." .-.. ,." ""·.·.".7.-•.•.• , ••.•.• :.~.:,:.:.~ •••• ' •• .:.:-:,:.;.:.;.: •.• ,";'.:.:.:.:,:.;: .•• ' •••.• ~.:;:.;.:-;.; •..••••• > ....... :.;.:.: •••••••••• ;.:.;.:.:.;.; ••••• ,..'.:.:.:.:.:.: •••••••• ~.~.:.;.:.:-:., •••••••• :.:.:.;.:.;-:.:, ••••• '.~.:.:.:.:.;.;.;0.', •.•.. ,.;0 ..• ;.: . .'..... ..;-; .... _._ ',' .,"'" •. , .....•.. "._ •...... ' .•....... _ ..... ; ..... -; .....•...• "._._ .• ! .•...•.....
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
129.120.242.61 On: Sat, 22 Nov 2014 07:35:19---
o ELECTROMAGNETIC
{19BSt PAlADiN
I Pure-REPM
Conventional Hybrid
Wedged-Pole Hybrid
I) ___ L __ i ___ -" ___ L __ -'---.--'
0.2 0.4 0,6
9/AW
tage of the hybrid and wedged hybrid is shown, The STI
Nos. I and 2 undulators indicated are the pure-REPM ge
ometry with the latter using oversize blocKs. THUNDER
and NISUS are STI undulators with conventional and
wedged-hybrid geometries, respectively. Also shown is an
electromagnetic undulator of the Paladin experiment. 6 At a
typicalg/ A ratio of 0.35, the conventional hybrid has a 28%
advantage over the reference, and the wedged hybrid has a
45% advantage over the reference.
The reason for the advantage of the wedged pole is
shown in the calculated plots of Fig. 2, exploiting the
quarter-period boundary conditions. The field in the con
ventional hybrid is limited by pole tip saturation. This prob
lem is reduced with the wide pole tip ofthe wedged geometry
while the magnet thickness is increased at the opposite end.
An additional benefit of the wider pole tip is a reduction of
third harmonic content of the field.
FIG. 2. Field plots show pole
tip saturation is reduced with
wedged pole.
1882 Rev. ScLlnstrum., Vol. 60, No.7, July 1989 EUI1ElwEl
8w8m8
FIG. L Comparison ofundulator geome
tries and relative field strengths. The
pure-REPM with square blocks and unit
fill factor is taken as a reference for each
full gap (g) to wavelength (It) ratio.
II. SHIM TUNING FOR FIELD ERROR REDUCTION
In practice, the undulator field quality is limited by the
presence of several undesirable factors, most notably trajec~
tory (steering) errors, phase-shift errors and higher-order
moment errors, such as improper quadruple moments or
excessive sextupole. These imperfections are caused in part
by inhomogeneities in the permanent magnets, imperfect
poles and mechanical misplacements. The errors become
more critical for longer systems, leading disproportionately
higher costs.
Discussions of the allowable magnetic field error toler
ances can be found in Refs. 7 and 8. In those papers, dipole
errors that lead to trajectory errors are considered; in Ref. 7,
these errors are shown to have different tolerances depend
ing on whether the undulator radiation is required to be co
herent or incoherent. For the FEL application, coherence is
required, whereas in synchrotron applications, the electron
emittance in some cases precludes coherence, independent of
the undulator errors. For the coherent case, there is the gen
eral requirement that the wiggler errors be sufficiently small
so that the phase space occupied by the electron beam is less
than the phase space occupied by a diffraction-limited pho
ton beam. Also in Ref. 7, it is shown that the dipole error
tolerance, if expressed in terms of the error of the integrated
dipole field, is dependent on the number of wiggler periods
and in many cases independent of the photon wavelength
and e-beam energy.
These considerations are for dipole errors which lead to
trajectory errors oflow spatial frequencies, that is, for orbit
errors that occur over a substantial fraction of the wiggler
length. A separate consideration is required for high spatial
frequency errors. An example of such an error would be the
errors remaining after the overall electron trajectory is cor
rected at several points along the wiggler length. In the limit
that the trajectory is corrected at very frequent intervals,
every few periods, for example, these remaining errors are
essentially phase errors (or time-of~flight errors for the elec
tron) rather than trajectory errors. Separate ca1culations9
have shown the RMS errors at each pole as small as several
tenths of a percent can be important even when the trajector
ies are otherwise perfect.
High Power beamlines 1882
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
129.120.242.61 On: Sat, 22 Nov 2014 07:35:19There is particular emphasis on reducing the errors of
the wedged-pole hybrid wiggler, since it produces roughly at
15% higher field than the more conventional straight-pole
hybrid and 40% higher than the samarium cobalt systems
without poles. The higher-field strength is important since
the FEL gain-extraction productlO scales as B2.
Up to now, there has been no simple scheme for elimi
nating, or tuning out, the hybrid's field errors, partly because
individual errors are not easily traced to specific magnets or
poles. If specific errors can be identified, then magnets and
poles can be relocated compensating locations. Such musi
cal-chair tuning schemes are poody suited to high precision
assemblies with special fix turing for the large forces involved
and are labor intensive.
Methods are dearly needed for achieving substantially
lower error levels than what has been demonstrated to date.
The techniques used to get from the first lO-ttm devices to
the long-undulator 0.5-and I-flm FELs (Ref. 2) are not
suitable for further extrapolation. These methods consisted
of ( 1) use of more stringent mechanical tolerances through
precision grinding and thermal control, and (2) a narrowing
of the acceptance criteria for the permanent magnets. Me
chanical tolerances are already in the O.OOl-in. range for
these large structures, and further magnet selection will be
come prohibitively expensive due to decreased yield. What is
needed is a simple, inexpensive method of tuning a wiggler to
the desired fields after it has been assembled.
A promising candidate for tuning is the newly demon
strated field shimming technique. Thus far, it has been ap
plied only on a short wiggler prototype to tune out dipole
errors in the plane of the primary field, although with devel
opment, any moment in either plane might be corrected. The
basic concept is that thin iron shims are used to selectively
shunt a small fraction of the field lines from regions where
the field is higher than desired. Proper placement of the
shims results in a unifonn field of slightly lower strength,
about 1 %, than the average initial field.
The geometry is shown in Fig. 3 using one wavelength of
the wedged-pole configuration, although the concept has
general applicability to aU wiggler types. For this geometry,
the shims are placed in the shallow recess on the flat tips of
the magnets, shunting field lines from one pole to another as
indicated in the figure. Clearly the effect of shunting field
lines between poles is to reduce the field on axis. Depending
on the local field, the shi.ms vary in thickness from 0 to ap
proximately 0.5 mm. With the large scalar potential differ
ence between the poles, the shims are completdy saturated
and the number of field Hnes shunted is determined simply
by their thickness. The field signature from a single pair of
shims (Le., top and bottom), away from the ends ofa iong
FtG. 3. Shim placement in
wedged-pole hybrid wiggler. The
primary field component (hol
low arrows) can be controlled
with the shunted field (solid ar
rows).
1883 Rev. SCi.lnstrum., Vol. 60, No.7, July 1989 300. · lPolQ fIo~a · t t " • 200. ~
" .,...+~ ..,
0 100. ;;;
!i
;Ji .300
-too. i
-4.00 -2.00 .000 2.00 4.0n
Z (em)
FlG. 4. Shim signature for single pair of shims (as in Fig. 3) in a long wiggler
assembly.
wiggler, is shown in Fig. 4. The effect is confined largely
between two poles, and it has been shown experimentally
that this signature is approximately linear in the shim thick
ness and additive with that of shims on neighboring poles.
Given that the effect is predictable, one can clearly gen
erate alogrithms that modify the field in some predeter
mined way. Thus far, the shims have been used successfully
to modify the RMS level of field errors in a short wedged
pole undulator with a 3.9-cm period, 1.4-cm fun gap, and
5.6-kG on-axis peak field. That data is used here as an exam
ple. It was desired to reduce the level of kick errors, defined
as the error in half period field integrals under each pole, so
that their RMS deviation could be reduced from the initial
1.3% to a much lower value.
A computer alogrithm was devised to use the measured,
uncorrected field and then identify the proper location and
thickness of shims to counteract the measured errors. The
shims are easily hand placed and self-attaching in the loca
tions indicated in Fig. 3. Afier one shim set plus one iter
ation, the result is shown in Fig. 5 for the central 18 poles of
the 26-pole undulator. The initial kick errors are shown as
the points connected by the dashed lines. The large sinusoi
dal field and any offset has been taken out and only the resid
ual errors are shown. The corrected field is shown by the
solid line connected by the solid line. In this case, the kick
error went from an initialleve1 of 1.3% to a value of 0.11 %.
It is interesting to note, from the shim symmetry of Fig.
3, that there will be no net dipole movement created by the
200 RESIDUAL E~ROAS
After SI'i!r.lmlng
/
" ,..../
I
200 /
/ ,\ A
f \ I '
'.00 W.O fI. ! \
I /
Inlilal Error .. t.3% RMS
ShlIfifOOIJ Error,. O.IV. RMS
t4.0
HAU' • P~J;IOO NUMS£R 18.0
FIG. 5. Measured comparison of field errors before and after shimmmg of
IS·pole section ncar the center of a NISUS prototype module. Crosses are
half period field integrals before shimming, solid line are after shimming.
Solid and dashed lines are for visual reference only.
High Power beamlines 1883
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
129.120.242.61 On: Sat, 22 Nov 2014 07:35:19CROSS SECTION
shims, and as a consequence one wonders if the shims can be
used to correct dipole errors. The answer is yes, and the
explanation makes a constructive point concerning various
length scales, or spatial frequencies, of errors and various
error correction schemes. The shimming algorithm can be
used to redistribute any arbitrary distribution of kick errors
to a new distribution, In the simplest case, this may be move
ment of a dipole error, associated with a single pole, to a
location of a dipole correction coil, so that the correction and
error then occur at the same location. In some systems, it is
convenient for the correction coils to have large axial extent,
and in that case the shims are used to redistribute the errors
to a new distribution, which is simply a constant (position
independent) error, This constant error is then removed
with an externally applied field oflarge axial extent, Le., low
spatial frequency.
Thus, the shims are a means of converting the high spa
tial frequency errors to lower spatial frequencies, where they
can be dealt with by other methods, In the case of steering
errors, the other method would be a long steering coil and, in
the case of phase errors, the other method might be an ad
justment of taper or gap in each module of the undulator;
these modules being perhaps I-m long.
With this in mind, we recall that for both curves of Fig.
5, the constant offset, i.e., the average dipole error, has been
removed before the RMS errors were calculated. The shim
algorithm adjusted whatever dipole errors are present to be a
constant, axially uniform. This can easily be cancelled by
constant bias field, and the RMS errors were calculated in
this spirit. One demonstrated method of applying this bias
field is the use of correction coils lying between the vacuum
system and poles as shown in Fig. 6, To allow for e-beam
diagnostics in the particular vacuum system considered, it is
convenient that these steering coils be restricted to approxi
mately 0.5 m in length and be repeated at meter intervals.
1884 Rev. SCi.lnstrum., Vol. 60, No.7, July 1989 FIG. 6. Vacuum system with imbedded
steering correction.
Under these circumstances, the shim algorithm would be
adjusted to move all of the dipole errors to the locations
under the correction wires, and leave no errors in the areas
that have no correctors. The short undulator section for the
measurements reported here was treated as ifit were entirely
within a portion of constant correction field.
III. SUMMARY
The wedged-pole hybrid geometry has been demon
strated to produce higher fields than the conventional
straight-pole hybrid, having an advantage of approximately
15% at ag/ A of 0.35, as well as lower harmonic content. An
inexpensive tuning method for the hybrid systems has also
been demonstrated.
'J. M. Slater, J. Adamski, D. C. Quimby, T, L. Churchill. L. Y. Nelson, and
R. E. Center, IEEEJ. Quantum Electron. QE-19, 374 (1983).
2K. E. Robinson, D, C. Quimby, J. M. Slater, T. L. Churchill, and A. Valla,
ill Proceedings of the Eighth International Free Electron Laser Conference,
Glasgow, UK, September 1986 [Nue!. lnstrum. Methods A 259, 62
(1987)].
'D. C. Quimby and A, L. Pindroh, Rev. Sci. lnstrum. 58, 339 (1987),
4K, Halbach. J. Phys. (Paris) 44, Cl (1183).
5G. A. Kornyukkin. G. N. Kulipanov, V. N. Utvinenko, N. A. Mesentsev,
A. N. Skrinsky, N. A, Vinokurov, and P. D, Voblyi, Nuc!. lnstrum. Meth
ods A 237, 281 (1985).
"G. A. Dies, in Proceedings of the Ninth International Free Electron Laser
Conference, Williamsburg, VA, September 1987 (to be published).
7J. M. Slater, in Proceedings afthe 1987 IEEE Particle Accelerator Confer
ence, Washington, D.C., March 1987 [IEEE Catalog No, 87CH2387-9, p.
479 (1987)].
"B. M. Kincaid. J. Opt. Soc, Am. B 2,1294 (l9SS).
9S. C. Gottschalk, Spectra Technology, Inc. and others (unpublished cal
culations) .
lOJ. M. Slater, AlP Conf. Proc. 130,505 (I985).
High Power beamlines 1884
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
129.120.242.61 On: Sat, 22 Nov 2014 07:35:19 |
1.101524.pdf | Microstructure of epitaxial ErBa2Cu3O7−x thin films grown on MgO(100) substrates by
rf magnetron sputtering
J. Chang, M. Nakajima, K. Yamamoto, and A. Sayama
Citation: Applied Physics Letters 54, 2349 (1989); doi: 10.1063/1.101524
View online: http://dx.doi.org/10.1063/1.101524
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/54/23?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Superconducting YBa2Cu3O7− x thin films on metallic substrates prepared by RF magnetron sputtering using
BaTiO3 as a buffer layer
AIP Conf. Proc. 251, 96 (1992); 10.1063/1.42061
The effects of secondary particle bombardment on ion beam sputtered thin films of Y1Ba2Cu3O x deposited on
MgO (100)
AIP Conf. Proc. 200, 102 (1990); 10.1063/1.39062
Microstructure of epitaxially oriented superconducting YBa2Cu3O7−x films grown on (100)MgO by metalorganic
decomposition
Appl. Phys. Lett. 55, 286 (1989); 10.1063/1.102406
Superlattice modulation and epitaxy of Tl2Ba2Ca2Cu3O1 0 thin films grown on MgO and SrTiO3 substrates
Appl. Phys. Lett. 54, 1579 (1989); 10.1063/1.101387
Microstructures of YBa2Cu3O7−x superconducting thin films grown on a SrTiO3(100) substrate
Appl. Phys. Lett. 52, 841 (1988); 10.1063/1.99302
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.193.242.161 On: Tue, 09 Dec 2014 21:02:42Microstructure of epitaxial ErBa2CUa07_X thin films grown on MgO (100)
substrates by rf magnetron sputtering
J. Chang, M. Nakajima, K. Yamamoto, and A. Sayama
Yokohama R&D Laboratories, The Furukawa Electric Co., Ltd .. 2-4-3, Okano, Nishi-ku. Yokohama 220,
Japan
(Received 21 February 1989; accepted for publication 4 April 1989)
The microstructural properties of superconducting ErBa2Cuj07 _ x films on single-crystal
MgO substrates are studied by transmission electron microscopy. The as-grown films are
single-crystal-like and are composed of subgrains of 0.1-0.2 pm in size. Due to annealing, the
dislocations at the sub grain boundaries disappeared. The annealed films are epitaxial with
either the a or the b axis of the ErRa2Cu,07 _ x unit ceil along < 100) directions of the MgO
substrate. The stress caused by lattice mismatch is relaxed by the formation of misfit
dislocations at the film/substrate interface.
Various techniques have been applied to the preparation
of epitaxial high-temperature superconducting oxide films.
The best results for superconducting YRa2Cu307 x films
have been obtained using single-crystal SrTi03 substrates
due to the good lattice match between the substrate and
film, H and consequently, most of the microstructure studies
ofYBa1Cu307 x mms were made on SrTi03 substrates. S-7
However, despite the fact that MgO (100) substrates have a
rather large misfit (-8 %) with RE Ba2Cul07 _, [rare
earth metal (RE) 1 films, they are used because the films
produced are suitable for real applications. 4 In this letter, we
report the successful epitaxial growth of ErBa1Cu}07 _ x
(hereafter referred to as ErBCO) films on MgO (100) sub
strates by using rf magnetron sputtering. The annealed films
have a zero resistivity temperature 1:. of 82 K and a critical
current density Je > 105 A/cm] at 77 K. Furthermore, we
studied the microstructure of these as-grown and annealed
films by transmission electron microscopy. The as-grown
films are single-crystal-like and are composed of subgrains of
about 0.1-0.2 pm in size. After the 900 °C heat treatment,
most of the dislocations at the subgrain boundaries disap
peared. The films are epitaxially grown with either the a or b
axis of the ErRCO cell parallel to the (100) of MgO sub
strates, without an in-plane rotation ';,9 ofthe (001) planes of
the ErRCO unit cell. The stress caused by lattice mismatch is
relaxed by the formation of misfit dislocations at the film/
substrate interface.
We have grown almost completely c-axis oriented
ErRCO films on MgO (100) substrates by using rf magne
tron sputtering, In order to reduce the res puttering effect,
the sputtering was carried out under high pressure
(Ar/02 = 111,80-100 mTorr). The substrate was heated to
around 650 "C during the deposition and the deposition rate
was about 2 nm/min. From x-ray 2e diffraction analysis, as
grown films are oriented with the c axis perpendicular to the
suhstrate surface. The c-axis lattice parameter was measured
to be 11.75 A( ± 0.02 A). After annealing at 900°C for 2 h
in oxygen flow, the c-axis lattice parameter of a 0.3-0.4 /-lm
thick film became smaller and reached 11.68 A. By using a
standard dc four-probe transport method, films with T:
= 82 K and.le > 105 A/cm2 (with a best value 4X 105A/
em2) at 77 K were obtained. In order to further investigate the microstructure of
these c-axis oriented films, transmission electron micro
scopy (TEM) observations for both the as-grown films and
the annealed films were carried out. Figure 1 shows the plan
view image of an as-grown film. Figures 1 (a) and 1 ( c) arc
the bright field image and the selected area diffraction pat
tern from the area shown in Fig. 1 (b), respectively. In Fig.
1 (a) a granular structure is observed. However, the diffrac
tion pattern shows dear diffraction spots, corresponding to
(100) and (010) planes of the ErBCO unit cell. No ring
patterns characteristic of polycrystalline films are observed.
Furthermore, bright field/dark field TEM images showed
that dislocations occurred at the boundaries between the
granular structure of Fig. 1. However, twins did not occur in
the as-grown film. Figure 2 shows the cross-sectional image
of the as-grown film. Boundaries between c-axis oriented
FIG, 1. TEM plan-view images of an as-grown film: (a) bright field image
(RF.I): (b) B.EI with selector aperture: (c) ,elected area ditrraction pat
tern from (b).
2349 Appl. Phys. Lett. 54 (23), 5 June 1989 0003-6951/89/232349-03$01.00 @ 1989 American Institute of Physics 2349
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.193.242.161 On: Tue, 09 Dec 2014 21:02:42FIG. 2. TEM cross-sectional image of an as-grown film. The inset shows the
selected area diffraction pattern from an area which includes the MgO sub
strate.
domains can be observed. The diffraction pattern of the inset
of Fig. 2 shows that neighboring grains are oriented parallel
to the c axis. From these TEM observations, we can con
clude that the as-grown films are single-crystal-like and are
composed of small subgrains with an average size of 0.1--0.2
pm. The c-axis orientation spreads within a small deviation
angle. The standard deviation angle of the c axis measured
by the x-ray rocking curve is sharp and reaches a constant
value of 0.45° (resolution angle ~0.3°) for films thicker than
50um.
However, Tc of the as-grown films is below 77 K. In
order to provide a proper amount of oxygen into the films,
post-annealing was employed. Figure 3 shows the cross-sec
tional image of the annealed film. Comparing Figs. 2 and 3,
the boundaries of c-axis domains are seen to have disap
peared due to annealing. The lattice fringes parallel to the
interface correspond to the (002) planes ofthc ErRCO film.
Furthermore, bright field/dark field TEM images showed
that the twins of 20 to 30 nm in width occurred due to the
tetragonal-orthorhombic phase transformation as the film
was cooling down from 900 °C to room temperature. The
inset of this figure shows the selected area diffraction pattern
from the area which includes the MgO substrate; the inci-
FIG. 3. TEM cross-sectional image of an annealed film. The inset shows the
selected area diffraction pattern from an area which includes the MgO sub
strate.
2350 Appl. Phys. Lett., Vol. 54, No. 23, 5 June 1989 FIG. 4. High-resolution image of the film/substrate interface.
dent electron beam is parallel to (010) direction ofthe MgO
substrate. The diffraction spots of Er BCO (100) can be ob
served clearly. This result means that instead of an in-plane
rotation of the (00l) planes of ErBCO to fit the lattice pa
rameter ofMgO (100), the film grew epitaxially with either
the a or b axis along the MgO < 100) direction. Figure 4
shows a TEM high-resolution image of the interface. It indi
cates that the ErBCO film was epitaxially grown aligned
with the MgO {IOO} lattice. Arrows point to the misfit dislo
cations which occurred regularly along the interface. The
stress caused by lattice mismatch has been relaxed, presum
ably by the formation of these misfit dislocations. However,
an amorphous layer was not formed at the film-substrate
interface. Before annealing, we found that most of the sub
strate surface was rather rough with hills and valleys of
depth about 0,1 pm. However, apart from the roughn~ss at
the interface with depth ~ 12 A as shown in Fig. 4, most of
the interface of the annealed film became smooth. We believe
that the formation of a smooth interface is attributed to a
reaction that may have been taken place at the interface
between the MgO substrate and ErBCO film during anneal
ing.
In summary, epitaxial ErBCO films have been success
fully grown on MgO (100) substrates by rfmagnetron sput
tering. As-grown films are single-crystal-like and are com
posed of small subgrains. With proper annealing, films with
1~ = 82 K and Je > 105 A/cm2 at 77 K were obtained. These
films were epitaxially grown with their a or b axis along the
MgO < 100) direction. The stress caused by lattice mismatch
is relaxed by the formation of misfit dislocations at the inter
face.
The authors wish to express great appreciation for fruit
ful technical discussions and funding bestowed by a group of
Japanese electric power companies-Tokyo Electric Power
Co., Tohoku Electric Power Co., and Hokkaido Electric
Power Co.
Iy' Enomoto, T. Murakami, M. Suzuki, ane! K. Moriwaki, Jpn. 1. AppL
Phys. 26, L1248 (1987).
2T, Tcrashima and Y. Bando, Appl. Phys. Lett. 53, 2232 (1988).
.Ip. Chaudhari, R. H. Koch, R. B. Laibowitz, T. R. McGuirt:. and R. J.
Gambino. Phys. Rev. Lett. 58. 2684 (19X7).
Chang eta/. 2350
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.193.242.161 On: Tue, 09 Dec 2014 21:02:424J. K wo, M. Hong, D. J. Trevor, R. M. Fleming, A. E. White, R. C. Farrow,
A. R. K01'tan, and K. T. Short, App!. Phys. Lett. 53, 26R3 (1988).
'C. II. Chen. H. S. Chen, and S. H. Liou, App!. I'hys. Lett. 53, 2339 ( 198R).
"D. M. Hwang, 1.. Nazar, T. Vcnkatesan, and X. D. Wu, App!. Phys. Lett.
52, 1834 (1988).
2351 Appi. Phys. Lett., Vol. 54, No. 23, 5 June 1989
.•.••••••• X"; •••••••••••••••••• :.':':.:.:.~.:.:.:.:.:.:;;;-.: O;.:.:.:.: ••• ;.:.;';".O;-".· ••• ·.·;·;>.·.·.O;~.v.·.·.·.·.·.· ................................. ,. •. ;-0:.:.-;0:.:.:-;.;.;.; •••• ' ••••••••••••••••••••••••• ' ••••••• ' •.••••• -; •••••• 'lB. M. Clemens, C. W. Nieh, J. A. Kittl, W. L. Johnson, J. Y. Josefowicz,
and A. T. Hunter, App!. Phys. Lett. 53, 187l (1988).
"I. Bloch, M. Hciblum, and Y. Komem, Appl. Phys. Lett. 46,1092 (1985).
"M. Eizenberg, D. A. Smith, M. Heiblum, and A. Segmuller, App!. Phys.
Lett. 49, 422 (1986).
Chang etal. 235i
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.193.242.161 On: Tue, 09 Dec 2014 21:02:42 |
1.343793.pdf | Rate equation analysis of microcavity lasers
H. Yokoyama and S. D. Brorson
Citation: J. Appl. Phys. 66, 4801 (1989); doi: 10.1063/1.343793
View online: http://dx.doi.org/10.1063/1.343793
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v66/i10
Published by the American Institute of Physics.
Related Articles
Emitter injection in terahertz quantum cascade lasers: Simulation of an open system
Appl. Phys. Lett. 100, 102102 (2012)
Random laser action in dielectric-metal-dielectric surface plasmon waveguides
Appl. Phys. Lett. 95, 231114 (2009)
Modal characteristics of terahertz surface-emitting distributed-feedback lasers with a second-order concentric-
circular metal grating
J. Appl. Phys. 106, 053103 (2009)
Dynamic modeling of a midinfrared quantum cascade laser
J. Appl. Phys. 105, 093116 (2009)
Heisenberg algebra, umbral calculus and orthogonal polynomials
J. Math. Phys. 49, 053509 (2008)
Additional information on J. Appl. Phys.
Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors
Downloaded 07 Oct 2012 to 152.3.102.242. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissionsRate equation analysis of microcavity lasers
H, Yokoyama8) and S. Do 8rorson
Department 0/ Electrical Engineering and Computer Science and Research Laboratory afElectronics,
Massachusetts Institute a/Technology, Cambridge, Massachusetts 02139
(Received 21 October 1988; accepted for publication 24 July 1989)
We describe the light output properties of single mode lasers having cavity dimensions on the
order of the emitted wavelength. A simple rate equation formula is derived for a four-level
laser assuming enhanced spontaneous emission into the cavity. These rate equation analyses
show that increasing the coupling of spontaneous emission into the cavity mode causes the
lasing properties to become quite different from those of usual lasers having cavity dimensions
much larger than a wavelength. We find that the lasing threshold disappears, the light emission
efficiency increases, relaxation oscillations do not occur, and the dynamic response speed is
improved. It is shown that the spontaneous emission rate alteration caused by the cavity plays
an essentially important role for these characteristics.
L INTRODUCTION
The alteration of a material's spontaneous emission rate
in a cavity 1,2 has recently attracted much attention as a fun
damental means of studying the interaction of matters with
vacuum field fluctuations. To date, many experiments have
demonstrated such effects, using Rydberg atoms,3-9 a solid
state laser material,1O organic dyes,11.12 and semiconduc
tors. 13, 14 Altering the spontaneous emission, however, is also
interesting from the device point of view. For example, Yab
lonovitch has proposed the utilization ofinhibited spontane
ous emission in semiconductor lasers for extremely low cur
rent operation.15 On the other hand, Kobayashi et al.
proposed the concept of thresholdless lasers with the full
confinement of spontaneously emitted photons in closed mi~
ero-optical cavities (microcavities) .16 Although the concept
of spontaneous emission rate alternation has not been taken
into account in his idea, enhanced, rather than inhibited,
spontaneous emission should occur in that situation. For
recent surface emitting semiconductor lasers, very short cav
ity structures have been fabricated. 17,18 Changes in sponta
neous emission properties could play an important role in
these devices.
In this paper we describe an analysis for light output
properties of microcavity lasers, based on rate equations
which are simply derived taking into account the enhanced
spontaneous emission caused by a microcavity. It is shown
that, if the coupling ratio of spontaneous emission into the
one cavity mode is sufficiently high, the laser oscillation
characteristics are greatly changed, including the threshold
behavior, the influence of nonradiative processes on input
output conversion efficiency, and the dynamic modulation
response.
iI. EMISSION RATE ENHANCEMENT
First, we discuss the enhancement of spontaneous and
stimulated emission rate in a closed microcavity. Here, we
assume that only one resonant cavity mode overlaps the gain
bandwidth (free space transition width) ofthe laser medium
3) Presently on leave from Opto-Electronics Basic Research Laboratory,
NEC Corporation, Miyukigaoka, Tsukuba 305, Japan. because of the very small cavity volume (of wavelength di
mension) ; thi.s is the origin of the spontaneous emission rate
alteration. Two cases of microcavity operation exist. In the
first case, the gain bandwidth is much less than the cavity
mode band width. According to Fermi's golden fule, the
spontaneous emission rate Ac in a cavity is represented in
this case by2
(1)
with
(2)
(3)
where A is the spontaneous emission rate in free space (here
after, we use the word "free space" as the meaning of "with
out cavity"), Pc (vo) [PI (Va) ] is the mode density for a final
photon state in a cavity (in free space) at transition frequen
cy Va. Qis the cavity quality factor, cis the velocity oflight, V
is the mode volume (in this case, cavity volume), H is an
interaction hamiltonian, Ii) is the initial state without pho
tons, and I f) is the final state with one photon. In (1), 1/ is
the enhancement of the spontaneous emission caused by the
cavity, Although recent interest has been focused on sponta
neous emission, (1) is also valid for stimulated emission.
This becomes obvious with the quantization of electromag
netic field. In this procedure, the overall photon emission
rate Rc for an atom (or a molecule) in a cavity is expressed
as
(4)
where s represents the number of photons in the cavity mode
in the initial state.
The second case of micro cavity operation occurs when
the cavity resonance peak is sharper than the gain band
width. This often occurs in atomic systems and causes the
"golden rule" to break down. In this situation, coherent ef
fects, such as Rabi oscillations,S or "one atom maser" oper
ation6 occur.
Expressions (1) and (4) may also be adapted to such
broad transition linewidth systems as organic dyes, certain
4801 J. Appl. Phys. 66 (10), 15 November 1989 0021-8979/39/224601-05$02.40 @ 1989 American Institute of PhySiCS 4801
Downloaded 07 Oct 2012 to 152.3.102.242. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissionssolid-state laser materials, and semiconductors, as long as
the cavity mode separation width is much broader than the
transition linewidth in free space, and the cavity resonance
width is broader than the inverse of the radiative lifetime.
For example, in the case of a semiconductor material, based
on the discussion in Ref. 19, the spontaneous emission rate
enhancement ratio 11 can be expressed as
1/ = f" fc(v)dv 1100
ff(v)dv
== l"'Pccv)P(v)dv /1'0'0 Pf(v)P(v)dv, (5)
where fc (v) [rf (v)] is the spontaneous emission rate for
emitting photons of energy hv with (without) a cavity,
Pc (v) [PI (v) 1 is the mode density for photons, and P( v)
represents the transition rate per mode. In (5), as a practical
approximation, the free-space emission rate is given by
PfCvo)P(vo)t:.P, where Vo is the photon frequency at the
emission peak, and AP is the FWHM of P( l'). In a micro
cavity, Peev) is more sharply peaked than P(v), and the
spectrally integrated emission rate can be expressed as
Pc (v~) )P( vb )¥c> where vb is the photon energy at a reso
nancepeak, APe is the FWHM ofPe (v). However, it should
be noted that when absorption loss is negligible Pc (v)!::.pc is
nearly equal to PI (v)1.\1', where t:.v is the cavity mode sepa
ration width. Thus, the ratio ?J is roughly expressed by
(6)
If v,; = vo, the ratio is ~ av/ AP. This shows that the spec
trally integrated emission enhancement depends on the cav
ity mode separation. Note that if PC vo)/ P( vb) > /::"v/ /::,.p
(i,e., off resonance cavity), 1/ becomes less than 1; thus the
spontaneous emission is inhibited instead of enhanced. The
discussion based on (5) is also applicable to a homogeneous
ly broadened two-level system, if the phase coherence time is
much shorter than the population lifetime.
Furthermore, if we assume that there are several cavity
modes within bandwidth, it can be easily found by carrying
out the integration of (5) that the spontaneous emission rate
does not change. Therefore, from the mode density point of
view, it is understood that we do not have to take into ac
count the spontaneous emission rate alternation for a con
ventionallargc-size (compared to the wavelength) cavity
laser. However, even in that case, it can be seen that spectral
ly partial emission enhancement occurs within each cavity
mode resonance width, and emission inhibition takes place
between cavity resonance peaks.
Classically, the spontaneous emission rate alteration
can be understood as caused by the change in radiation resis
tance experienced by a classical dipole when inserted in a
cavity. A complementary view is that it is the result of reso
nant enhancement of the electromagnetic field by multiple
reflections in the cavity, when the roundtrip time oflight is
much shorter than the phase coherence time of a dipole. This
effectively increases the coupling of the dipole to the field.
Drexhage adopted this method to calculate the spontaneous
emission rate modification of thin dye films. I I Furthermore,
semiclassical laser equations may be able to describe the be
havior of microcavity lasers if spontaneous emission pro-
4802 J, Appl. Phys., Vol. 66, No. 10, 15 November 1989 cesses are properly involved.20 However, as outlined by ex
pressions (1)-(6), a description based on mode density
alteration simplifies the discussion, and consistently treats
both spontaneous and stimulated emission in laser rate equa
tions, as long as we are not concerned with the laser's fre
quency and phase.
m. RATE EQUATIONS
To use rate equations based on Fermi's golden rule, we
must insure that an adiabatic approximation is valid. That is,
no transient coherent effects occur. The phase coherence
time of organic dyes and semiconductors are in the femto
second range, while the inverse of the Rabi frequency in a
cavity will be on the order of 1-10 ps for usual optical pump
ing rates ( < 1 MW cm-2). Thus, such transient coherent
phenomena as superradiance, optical nutation, etc., will not
occur for these materials, and a simple rate equation ap
proach is valid.
To begin, we may study the rate equations of a single
mode microcavity laser, which is completely enclosed by the
reflector. For such a device, the spontaneous emission rate is
given by (4). Assuming an ideal four-level laser material
(the decay rates of the highest state to the upper laser state,
and of the lower laser state to the lowest state are extremely
fast), with no nonradiative processes and no inversion satu
ration, the rate equations can be written as
dn -=p~Ac(s+ On, dt
ds ~ = Ac (s + l)n ~ ys,
elt (7)
(8)
where n is the number of excited atoms (molecules) in the
cavity of volume V, p represents the pumping rate, and y is
the damping rate for photons from the passive cavity. The
static solution of these equations is simple but noteworthy:
s=p/yand n=yp/[Ac(p+r)].
We see that the light output increases linearly with increas
ing pumping for all pumping rates. In other words, this de
vice works as a "thresholdless laser." As we will show, this
occurs because all photons are emitted into the one single
cavity mode. Note that n does not proportionally increase
with pumping increase, and this behavior is different from
that of ordinary spontaneous emission, in which the excited
state population n linearly increases with pumping increase.
This thresholdless nature is not necessarily the same as the
concept of one atom maser,5 in which at most only one atom
exists in the cavity at a time and whose behavior is not simply
described by an argument based on the golden rule.
Although enhanced spontaneous emission CAe >A), is
not the necessary condition for the lack of a threshold, the
consequent increase in the spontaneous emission rate has
some great advantages from the device point of view. For one
thing, the response speed of the device to dynamic modula
tion will be improved, as a result of the increased spontane
ous emission rate. Furthermore, the influence of nonradia
tive depopulation processes will be decreased since the
spontaneous emission lifetime will be much shorter than the
nonradiative lifetime. Another interesting feature of the
H. Yokoyama and S. D. Brorson 4802
Downloaded 07 Oct 2012 to 152.3.102.242. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissionsthresholdless laser is that relaxation oscillations will not oc-4.-----.-----.------.------:::.
cur. This happens because there is no threshold, so the
pumping energy is always immediately converted to laser
output. Thus, there is no mechanism for storing energy in
the laser medium, which is necessary for relaxation oscilla
tions. This is confirmed by a standard small-signal analysis,
which reveals that there is no resonance frequency for relax
ation oscillations.
So far, we have considered the case of a completely
closed cavity resonator. Now we would like to generalize to
the case of an open resonator. We assume there is still one
cavity mode, but now other modes exist which correspond to
photons leaving the open cavity. We assume that the sponta
neous emission into the cavity mode is stilI enhanced, but the
free-space modes have the free-space spontaneous emission
rate. This corresponds to the case discussed in Ref. 8. We
take the ratio of the solid angle subtended by the cavity mode
to the free space modes to be po Thus, /3 is proportional to the
inverse of the mode volume V; from another view point, it is
the light-material interaction strength, If a concentric cav
ity9 is assumed, the value of /3 simply corresponds to the
solid angle which an atom sees the cavity mirrors at the cav
ity center. Also taking into account nonradiative depopula-
tion processes, the rate equations can be represented as
dn -=p -(I-f3)An +/3A;O +s)n -rn, (9)
dt
ds - = tU ; (1 + s) n -ys. (10) dt
Here, s is the number of photons coupled to the cavity mede,
and r is the nonradiative depopulation rate. In this case, A ;
represents the enhanced spontaneous emission rate for the
cavity mode, and is related to the free space rate by A ; = FA,
where the enhancement factor is F. In the limit /3 -> I, we
have a closed cavity, and A ; reduces to Ac in Eq. (1).8 Note
that in a broad bandwidth material, F depends on the cavity
mode separation width as discussed in Sec. II. Thus, F de
pends on the cavity size, as does/3. Therefore, to get an large
/3A ~ value, the cavity should be quite small, and to avoid the
photon lifetime (lIy) decrease, the reflectivity of cavity
mirrors should be quite high. For an open microcavity with
wavelength dimensions. although a plane mirror Fabry
Perot configuration could provide a rather large value for f3
( ~ 0.1), the achievement of a microscopic confocal or con
centric Fabry-Perot configuration would improve the value
of /3. Full confinement of spontaneous emission into the cav
ity mode might be realized with microsphere or microcube
cavity structures.
IV. NUMERICAL RESULTS AND DISCUSSIONS
We have carried out numerical analysis using (9) and
(10). Steady-state solutions of (9) and (10), for an ideal
four level laser (r = 0), are shown in Fig. 1. In the figure, to
compare the output properties for cavities with different /3
(I.e., different mode volume), light (photon) output SOUl
excited state population N, and pumping P are, respectively,
normalized as
4803 J. AppL Physo, Vol. 66, No.1 0, 15 November 1989 3
0 4
z 2
z 0
!-(3-000001
<:I: 0,01 -l
::J 001 13
CL -~ 0 2 :3 4
PUMPING P
FIG. L Light output SO,,! and population inversion Nvs pumping Pofmi
cTOcavity four-lcvellaserso A co, 109 S -', F = 10, r = 10'2 s -', and r = o.
r /3A;
Sout = ---s = (JF5, A Y /3A' NT C =--11,
Y
and
1 (J A; /3F P=---p=-p.
A Y r
It is seen that as (J increases, the threshold disappearso In the
mode point of view, for the (J.( 1 open cavity case, even
though there is only one cavity mode, excited atoms are
mostly coupled with free-space modes, and the cavity mode
photons can only increase rapidly above "threshold" be
cause of intensive stimulated emission. Thus, the phase tran
sition (threshold) appears in the cavity mode output. (Note
that in actual semiconductor laser devices, the spontaneous
emission coupling ratio (J is 10-5_10-6 per cavity mode.)
On the other hand, in the case of /3 = 1 (closed micro
cavity), all the photons emitted couple into the single micro
cavity resonance mode. Therefore, the emission process
gradually changes from the spontaneous emission dominant
one to the stimulated emission dominant one without a
phase transition (threshold) 0 Although it may not be mean
ingful to distinguish spontaneous emission and stimulated
emission if there is no threshold, for convenience, we distin
guish the emission rate proportional to s in the equation as
stimulated emission. Therefore, for the pumping level shown
in Fig. 1, the light emission process is dominated by the "en
hanced spontaneous emission," because the unnormalized
number of photon s = 0.4 at P = 4 is less than 1. The behav
ior of the excited state population fl is also notable. For the
case of an ordinary laser, with increasing pumping, n in
creases until the lasing threshold level and then is clamped
there. On the other hand, 11 of a thresholdless laser very slow
ly increases with a pumping increase, and it reaches a con
stant value at infinitely large pumping (the condition for
s> 1 ) . As is shown in Fig .. 2, another noteworthy feature of a
Ho Yokoyama and S. D. 8rorson 4803
Downloaded 07 Oct 2012 to 152.3.102.242. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions4~----~--------'-------r-----~
... " <> (f) :3
~ 2
a...
~
::> o
o
PUMPING P
FIG. 2. Light output Sout vs pumping P of microcavity four-level lasers
involving nonradiativc processes. A = 109 s-', F= 10, Y = 10" s-', and
r = 1098-'.
thresholdless laser is to maintain high output conversion ef
ficiency, even if the nonradiative population lifetime is com
parable or less than the free-space spontaneous emission life
time. This occurs even while the lasing threshold of small /3
case markedly increases. This is easily understood, since in a
thresholdless laser, the ratio of radiative depopulation rate
to nonradiative one is greatly increased because of the en
hanced spontaneous emission.
It has also been found that for fixed p, increasing F also
gradually removes the threshold. This is because of the sub
stantial increase in the amount of spontaneous emission cou
pled into the cavity mode.
Concerning the dynamic properties of microcavity la
sers, as discussed in Sec. III, higher frequency response is
expected in thresholdless laser, because of the enhancement
of spontaneous emission rates. Figure 3 shows a calculated
result for microcavity four leve11asers with sinusoidal pump-
3r-----------------------------~
2
~ /3 " I
(J)
I-0
;:) :3
a... I-
;:)
0
:2
,ez 0.0001
o 4
TIME (ns)
FIG. 3. Dynamic light output properties of microcavity fom-levellasers.
A = 1095-', F= 10, r = 5 X 10" s-', f' = 0, Po O~ 2. The modulation fre
quency of pumping is taken as! = 2 X 10'" s '.
4804 J. Appl. Pl1ys., Vol. 66, No. 10, 15 November 1989 ing modulation of P = Po( 1 -cos 211'ft). To clearly extract
the effect of spontaneous emission coupling, the nonradia
tive depopulation rate r is taken to be zero in this calcula
tion. The enhancement factor is taken to be F = 50, since
this value can be realized by a halfwavdength size cavity for
semiconductors. In the small /3 case, a time delay for lasing
and relaxation oscillations are observed. On the other hand,
when f3 = 1, there i.s no relaxation oscillation (this is pre
dicted from standard small-signal analysis), and the modu
lation depth in steady state is much larger than that when/3
is small. It is noted here that the unnormalized average pho
ton number in the cavity s = Soutl/3F is much larger for
{3= 0.0001 case (s = 200) thanfor,8 = 1 case (s = 0.04). It
is emphasized, therefore, that the response speed improve
ment in the case of p = 1 is dominantly due to the decrease
in spontaneous emission lifetime by a factor of F for the
pumping level shown in Fig. 3.
Although (9) and (10) are valid forfour-levella.';er sys
tem, they are also approximately applicable to intrinsic
semiconductors (with bimolecular radiative recombina
tion). There, the spontaneous emission rate is represented by
A = B r n (under the Boltzmann carrier distribution approxi
mation), where Br is the bimolecular carrier recombination
coefficient. When the calculation is performed using this
expression for A, the features are qualitatively the same as
for the case offour-levellasers.
V. CONCLUSION
In summary, rate equation analyses have been imple
mented on static and dynamic output properties of micro
cavity lasers, based on the concept of spontaneous emission
rate enhancement in a cavity. Although our simple rate
equation analyses can bring information only about output
power, some attractive features of microcavity lasers have
been predicted. Among these are the lack of threshold, the
efficiency increase, and the high-speed response improve
ment, when the coupling ratio of spontaneous emission into
the cavity is sufficiently large. In these characteristics, the
increase in spontaneous emission rate plays an essentially
important role, Thus, it should be noted that the operational
properties of single-mode microcavity laser are not correctly
explained by simply counting the number of cavity modes,
without taking into account the spontaneous emission rate
alteration. Other aspects of microcavity lasers (oscillation
frequency, linewidth, etc.) are also interesting subjects to
study, but they should be discussed in the context of a semi
dassical or a fully quantum mechanical analysis.
ACKNOWLEDGMENTS
The authors are grateful to Professor T. Kobayashi of
Osaka University, Professor E. P. Ippen, Professor H. A.
Haus, Professor D. Kleppner, and Dr. J. Wang of Massa
chusetts Institute of Technology for their stimulating discus
sions. The valuable comments of Dr. R. Lang, Y. Nambu,
M. Suzuki, K. Nishi, T. Hiroshima, and T. Anan of NEC
Corporation are also gratefully acknowledged. This work
was supported in part at M.I.T. by the Joint Services Elec
tronics Program Contract No. DAAL03-86-K-0002.
H. Yokoyama and S. D. Brorson 4804
Downloaded 07 Oct 2012 to 152.3.102.242. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions'E. M. Purcell, Phys. Rev. 69, 681 (1946).
2D. Kleppner, Phys. Rev. Lett. 47,233 (1981).
3A. G. Vaidyanathan, W. P. Spencer, and D. Klcppner, Phys. Rev. Lett.
47,1592 (1981).
4p. Goy, J. M. Raimond. M. Gross, and S. Haroche, Phys. Rev. Lett. 50,
1903 (1983).
'D. Meschede. H. Walther. and G. Muller, Phys. Rev. Lett. 54, 551
(1985).
6G. Rempe and H. Walther, Phys. Rev. Lett. 58, 353 (1987).
7W. Jhe, A. Anderson, E. A. Hinds, D. Meschede, L Moi, and S. Haroche,
Phys. Rev. Lett. 58, 666 (1987).
'D. J. Heinzen, J. J. Childs, 1. E. Thomas, and M. S. Feld, Phys. Rev. Lett.
58,1320 (1987).
9D. J. Heinzen and M. S. Feld. Phys. Rev. Lett. 59, 2623 (1987).
lOp. DeMartini, Phys. Lett. 115, 421 (1986).
"K. H. Drexhage, in Progress in Optics, edited by E. Wolf (North Holland,
Amsterdam, 1974), VoL XII, p. 165.
'2F. DeMartini, G. Innocenti, G. R. Jacobovitz, and P. Mataloni, Phys.
Rev. Lett. 59, 2995 (1987).
"E. Yablonovitch, Phys. Rev. Lett. 51!. 2059 (I9!l7).
I4H. Yokoyama, K. Nishi, T. Anan, and H. Yamada, Tech. Dig. of Topical
4805 J. Appl. Phys., Vol. 66, NO.10,15 November 1989 Meeting on Quantum Wells for Optics and Optoelectronics, Salt Lake
City, March 1989, paper MD4 (unpublished).
"E. Yablol1ovitch, T. J. Gmitter, and R. Bila!, Phys. Rev. Lett. 61, 2546
(1988).
J('T. Kobayashi. T. Segawa, A Morimoto, and T. Sueta, Tech. Dig. of 43rd
Fal! Meeting of Japanese Applied Physics Society, paper 29a-B-S, Sep
tember 1982 (unpuhlished); T. Kobayashi, A Morimoto, and T. Sueta,
Tech. Dig. of 46th Fall Meeting of Japanese Applied Physics Society, pa
per 4a-N-l, October 1985 (unpublished) (both ill Japanese).
17J. L. Jewell, K. F. Huang, K. Tai, Y. H. Lee, R. Fischer, S. L. McCall, and
A. Y. Cho, App!. Phys. Lett. 55, 424 (1989).
"s. W. Corzine, R. S. Geels, R. H. Yan, J. W. Scott, L. A. Coldren, and P.
L. Gourly, IEEE Photonics Tech. I,etL 1,52 (1989).
'''n. c. Casey, Ir. and M. B. Panish, Heterostructure Lasers (Academic,
New York, 1978), Chap. 3.
20 A semiclassical description of spontaneous emission in a cavity has been
done in the following paper: J. J. Childs, D. J. Heinzen. J. T. Hutton, and
M. S. Feld (unpublished).
21M. Sargent III, M. O. Scully, and W. E Lamb, Jr., Laser Physics (Ad
dison-Wesley, Boston, MA, 1974), Chap. 8.
H. Yokoyama and S. D. Brorson 4805
Downloaded 07 Oct 2012 to 152.3.102.242. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions |
1.343077.pdf | Experimental vortex transitional nondestructive readout Josephson memory cell
Shuichi Tahara, Ichiro Ishida, Yumi Ajisawa, and Yoshifusa Wada
Citation: Journal of Applied Physics 65, 851 (1989); doi: 10.1063/1.343077
View online: http://dx.doi.org/10.1063/1.343077
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/65/2?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
3 ns single-shot read-out in a quantum dot-based memory structure
Appl. Phys. Lett. 104, 053111 (2014); 10.1063/1.4864281
Experimental and theoretical response of distributed read-out imaging devices with imperfect charge
confinement
J. Appl. Phys. 107, 083917 (2010); 10.1063/1.3327412
Fundamental criteria for the design of highperformance Josephson nondestructive readout random access
memory cells and experimental confirmation
J. Appl. Phys. 50, 8143 (1979); 10.1063/1.325955
Meter ReadOut for Vibroscopes
Rev. Sci. Instrum. 35, 232 (1964); 10.1063/1.1718787
Fast ReadOut Chronotron System
Rev. Sci. Instrum. 28, 1010 (1957); 10.1063/1.1715790
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15Experimental vortex transitional nondestructive read .. out Josephson
memory ceU
Shuich! Tahara, ichiro Ishida, Yumi Ajisawa, and Yoshifusa Wada
l.ficroelectronics Research Laboratories, NEC Corporation, 4-1-1, Miyazaki, Miyamae-ku, Kawasaki,
Kallagawa 213, Japan
(Received 7 April 1988; accepted for publication 20 September 1988)
A proposal vortex transitional nondestructive read-out Josephson memory cell is successfully
fabricated and tested. The memory cell consists of two superconducting loops in which a single
flux quantum is stored and a two-junction interferometer gate as a sense gate. The memory cell
employs vortex transitions in the superconducting loops for writing and reading data. The
vortex transitional memory operation of the cell contributes to improving its sense
discrimination and operating margin. The memory cell is activated by two control signals
without timing control signals, Memory cell chips have been fabricated using a niobium
planarization process. A ± 2i % address signal current margin and a ± 33% sense gate
current margin have been obtah1cd experimentally. Successful memory operations of a cell
driven by two-junction interferometer gates has been demonstrated. The single flux quantum
operations of this memory cell makes it an attractive basic element for a high-speed cache
memory.
I. INTRODUCTION
Josephson devices, with their high intrinsic switching
speed and low-power dissipation, are promising circuit ele
ments for future ultrahigh performance computer applica
tions. In the Josephson computer, a high-speed cache mem
ory is indispensable to complement the Josephson logic
circuits which have picosecond-switching characteristics. In
general, signal delay time through a memory array line is
first-order proportional to the amount of stored flux quanta
in the memory cell. I A single flux quantum memory cell,
therefore, is an attractive basic element for a high-speed
cache memory. Various kinds of single flux quantum mem
ory cells have been proposed and examined experimental
ly.2-4
A quantum loop cell proposed by Henkels et al.2•5 has
been the object of particular study for a high-speed memory.
This memory cell, however, has several problems. Interfer
ometer gates with two control signals are used as write and
sense gates in the memory cell. Since tolerances on all of a
gate current and two control currents are equal for a maxi
mum margin in these gates, it is difficult to approach a theo
reticallimit of the operating margin. In the memory array,
memory cell selection results from a coincidence of X and Y
address signal currents. U nselected cells in the X and Y lines
suffer from half-selected disturbance. Therefore, a large op
erating tolerance in the memory cell is very important for
increasing the discrimination between selected and unselect
ed cells. In addition, a second problem for this cell is the fact
that the necessity of timing sequence for address signals
makes high-speed memory operations difficult. Supply of a
timing signal requires a large timing margin, which prevents
the circuits from reducing its cycle time. AdditionaHy, cell
driving current levels are not equal in the memory cell, re
quiring a different current level for signals such as address,
data, and read/write conditions, This often reduces an inter
connection margin between the cells and peripheral circuits. It is an important problem when memory circuits with large
operating margins are constructed.
In this paper, we discuss a single flux quantum memory
cell, called a vortex transitional memory ceU.6 Its main fea
tures are a large operating margin, a memory operation with
no timing control signals and an almost equal current level
for control signals. This memory cell is activated by X, Y
address signals and a sense signal, and employs vortex transi
tions in the superconducting loops for writing and reading
data. The vortex transition in the superconducting loops
coupled with the sense gate permits that operating margins
for address signals are almost independent of a sense current
margin. Therefore, the operating margin for the memory cell
can be optimally designed to be its theoretical Hmit. High
speed memory operations are possible because timing con
trol signals are not necessary. The memory cell contributes
to improve a margin for a total memory circuit since the
applied control current levels are almost equal.
The memory cell consists of two superconducting loops
each of which stores a persistent circulating current corre
sponding to a single flux quantum. The sense gate couples
with one of the superconducting toops. The cell is fabricated
by using Nb/ AIOJNb junctions and the niobium planari
zation technique.7 The surface flatness for each layer results
in high reliability. The basic circuit configuration is de
scribed in Sec. II. The fabrication process and experimental
results are presented in Sec. III. Conclusions are finally giv
en in Sec. IV.
II. CIRCUIT CONFIGURATIONS
An equivalent circuit for the vortex transitional nondes
tructive read-out (NDRO) Josephson memory ce!l is shown
in Fig. 1. The cell consists of two superconducting loops
(loop 1 and loop 2), each of which stores a single flux quan
tum. The superconducting loop contains a Josephson june-
851 J. Appl. Phys. 65 (2), 15 January 1989 0021-8979/89/020851-06$02.40 @ 1 9S8 American Institute of Physics 851
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15ly
Is
IDC -----'l'""-4-...ro1fP----
I )( Mi ._~( '-' ""","'--.J-""""""""",,,
LI
JI RI
FIG. 1. Equivalent circuit of a vortex transitional NDRD memory cell.
L, = 5 pH, L2 = 4 pH, L, = 7 pH, L4 = 1 pH, I, = 0.2 rnA, 12 = 0.1 rnA,
and I, ,= 14 = 0.1 rnA. (/,-1.,: critical currents of J,-J •. )
tion and inductance elements. Damping resistors R I and R2
are connected in parallel to junctions JI and J2, respectively,
and provide suitable damping conditions.6 The sense gate is
a two-junction interferometer gate, magnetically coupled
with loop 2. Loop 1 stores information in the foml of a single
flux quantum. The Josephsonjunction, J" included in loop 1
has a function in which a single flux quantum is caused to
enter loop 1 when X and Y address signals are coincident.
Reading of stored data can be accomplished by the loop 2
vortex transition, which depends on the stored data in loop
1, and the switching of the selected sense gate caused by the
transition. Optimum design parameters are listed in Fig. 1.
Here let us investigate the variation of the quantum
phase differences B\ and 82 of the Josephson junctions J1 and
J2 in the memory loops against the address signal currents Ix
coupled with inductances L I and L2, and Iy injected into the
loop 1 in Fig. 1. The flux quantization condition and the
current continuation condition yield equations for I" [I and
the quantum phase difference BI, 82 as
(tPo/21T)( 8J + 2m1T -82 + 2n1T)
= L2Iy + (Lj + L2) (lx -I, sin 8,) + L3I2 sin 82 ,
(1)
(<1>01211')(82 -2mr)
= L4ly -L4Ij sin 81 -(L3 + L4)I2 sin 82 , (2)
where II and 12 are the Josephson critical currents of junc
tion J1 and J2, respectively, <Po is the magnetic flux quantum,
L1, L2• L3, and L4 are inductances in Fig. 1, and m and n are
integers for the quantities of magnetic flux in loop 1 and loop
2, respectively. In these equations, we assume mutual induc
tances MJ and M2 equal L I and L2, respectively. The stability
condition produced by potential energy minimum is
A JlIz cos ()\ cos 82 + A2I( cos Bj
+ A312 cos B2 + A4>0, (3)
where
Al = (L(L3 + LjL4 + L2L3 + L2L4 + L3L4)/L2L4,
A2 = (<J>o/21T) (LI + L2 + L4)/(L2L4) ,
A3 = (tPo/21T)(L3 + L4)/(L2L4) ,
A4 = (<P0I21T)2/(LzL4) .
From Egs. (1 )-( 3), we can obtain the threshold charaeter-
852 J. Appl. Phys .• Vol. 65, No.2. 15 January 1989 noise
band
-0.4 0.8
FIG. 2. Threshold curves of the memory loops on the (0,0), (1,0),
and (0, 1) modes, along with hypothetical ± 10% variation of critical cur
rents and inductance values. The dotted areas indicate the operating margin
of Ix and I,.
istics (Fig. 2). Figure 2 shows several parts of the threshold
curves for the memory loops in the memory ceil. The hori
zontal and vertical axes represent the address signal cur
rents, Ix and Iy' respectively. The numbers in parentheses
correspond to flux quanta in the memory loops. That is,
(m,n) means m flux quanta in loop 1 and n flux quanta in
loop 2. Point "0" is the memory operating origin and is
defined by de powered current Ide' The (0,0) mode and
(1,0) mode in the memory loops are respectively correspon
dent to data" 1" and "0".
Cell operations and the stability of the dynamics were
established in Ref. 6. As shown in Fig. 2, the operating point
moves from "0 "to "A" or"B" according to data on writing.
On reading the stored information, the operating point
moves from "0" to "e." The vortex state for the memory
loop changes into the (0,1) mode only when the data" 1" is
stored, and then the sense gate switches into a voltage state.
After that, the vortex mode can return to the (0,0) mode at
point "0" under the suitable damping conditions. In Fig. 2,
the dotted areas indicate the operating regions for I~ and Iy
for" 1", "0," writing and reading. The shaded areas illustrate
thermal noise bands in a fashion similar to that described in
Ref. 6. Minimum operating margins Ix = 0.16 rnA ± 14%
and Iv = 0.18 rnA ± 14% are achieved, along with a hypo
thetical ± 10% variation in the Josephson critical current
and in the inductance value, while the optimally designed
cells have address signal current margins Ix = 0.17 rnA
± 33% and Iv = 0.2 mA ± 33%.
On reading the stored data, a vortex transition in the
loop 2 causes the sense gate to switch. Figure 3(a) shows
calculated characteristics in terms of the external current Ie
ofloop 2 and the quantum phase difference 82 of junction J2•
Figure 3(b) shows the threshold characteristics of the sense
gate, along with hypothetical ± 10% variations in Joseph
son critical current and inductance values, The shaded areas
in Figs. 3(a) and 3(b) illustrate the thermal noise bands. In
these calculations, we assumed a coupling factor between
loop 2 and the sense gate is approximately 0.5. When the
memory cell conserves data "1" ("0"), the operating point
Tahara eta!. 852
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15Ie (mAl (a)
I
noise K /
0.5 band ,..-.. , L~
I "~-
f: f..~l /1 ' H ll' J '--'
82
-411"
Is (mAl (b)
o
FlG. 3. (a) Characteristics of the external current and the phase difference
in loop 2. (b) Threshold curves of the sense gate. :±: 10% fabrication toler
ances of circuit parameters were assumed.
stays at "D" ("E") in Fig. 3 (a). The operating point moves
from "D" to "H" through "G" with the supply of positive
current Ix and negative current [y. On the other hand, when
data "0" is conserved, the operating point moves from "E"
to "F", and vortex transition does not occur.
The operating point for the sense gate coupled with loop
2 moves to "M" or "N" depending upon the applied magnet
ic field, as illustrated in Fig. 3 (b). The sense gate margin is
determined by the characteristics of the sense gate and the
input magnetic field at points "G" and "R" of Fig. 3(a).
The designed sense gate has the product LI of ~o/6, where L
is the total inductance value for the sense gate and I is the
critical current for one junction (J3 = J4 in Fig. 1). A sense
gate current margin has been designed Is = 0.12 rnA
+ 42% nominally, and Is = 0.13 rnA ± 13% assuming ± 10% parameter variations and thermal noise distur
bance. Since the fiux mode in loop 2 changes only for read-
ing, the sense gate current Is can be applied as a dock-pulse
like gate current.
As mentioned above, the operating margins for Ix and
I is almost independent of the sense gate margin, because
the sense gate detects the vortex transition in loop 2. There
fore, the operating margins of ±: 33% for Ix and Iy are nom
inally designed. And then the memory cell has capability for
high-speed memory operation, since the cell is activated by
the address signals and the sense signal without a timing
sequence. Moreover, the designed memory cell contributes
to improving the operating margin for a total memory cir
cuit because the applied control current level for Ix and ly
are designed to be nearly equal.
853 J. Appl. Phys., Vol. 65, No.2, 15 January 1989 TABLE 1. Layem for ,he vortex transitional memory cell.
Layer Material
GP Nb
GIl Nb2O,
GI2 Sial
IL, 1\0
II SiO,
RS Mo
ILl Nb
JJ Nb/AIO./Nb
1L3 Nb
III. EXPERiMENT
A. Fabrication Thickness
(nm) Function
200 Ground plane
30 Ground insulation
300 Ground insuiation
200 Interconnection layer I
200 Interconnection insulation 1
70 Resistor layer
200 Interconnection layer 2
200/6/200 Junction trilayer
200 Interconnection layer 3
A test chip with four-level interconnections having less
than about 50 nm planarity wa~ fabricated by a lift-off plan
arization technique,7 using sputtered Nb films for all metalli
zations except for Mo resIstors, and sputtered SiOz films for
insulation layers. Table I describes the layers for the de
signed cell circuits. A cross section of the cell is illustrated in
Fig. 4. It is composed of a ground plane, three interconnec
tion layers, Nbl AlO x Ir-..o Josephson tunnel junctions, resis
tors, insulation between interconnections, and contact holes
between interconnections. To heighten the reliability, the
lift-off planarization technique was applied to each level in
the cell structure because of its high layer-thickness control
lability, low-temperature process ability, and pattern-size
adaptability.
The fundamentallift-offplanarization process flow con
sisted of the foHowing five basic steps: (a) A sputtered lower
film was patterned by a reactive ion etching technique using
a CF4 gas plasma. (b) A second film was deposited over the
entire surface, including resist masks. (c) The second film
on the side waH of the etching mask is etched away selective
ly with a slight wet etching. (d) The resist and the second
film on the resist are removed with a solvent in an ultrasonic
treatment. (e) An upper layer is sputtered over the planar
ized surface.
An of the interconnection layers and contact holes are
planarized with the above technique. A microphotograph of
the memory cel! is presented in Fig. 5. The cell size was
49X49 pm2, and the minimum line width, minimum layer
to-layer registration, and minimum junction dimensions
RESISTOR
I~TERCCNNECT :ON :3)
''1"----~COUNTE.Ri
,,-JUNCTION
fT?+-,¥v?jt>--"-..L...-'-1:;;..;r'--'''17~_ s _1~~~~~?N"ECTiON (2)
~~~~~~=r~~$Lij~~· CONTACT KOLE
--INTERCONNECT:ON (\)
H"';"444J.".-r~++r-~ CONTAC; HOLE
'$"-GROUND-PLANE
\¥,-rr-r>-rrrrrn-rrr.rrn.,.,-r,,,..,,rrr.r-rn77T,'777, Tl-.s--gUeST RATE
FIG. 4. Vertical structure of the vortex transitional memory cell.
Tahara et al. 853
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15FiG. 5. Microphotograph of the 2 X 2 bits vortex transitional memory edt
were 1.5, 1.0, and 3.0 pm, respectively. In this designed cell,
the coupling of two control lines to loop 1 consumes a large
area. However, these two control lines can be changed into
one control line by improving driver circuits because one of
the two control lines is dc current line. Therefore, the mem
ory cell has the ability to be designed with a smaller size. A
Mo sheet resistance of 1.5 010 was achieved, almost the
same as the designed value. The critical current of the 3,um
Josephson junction was 0.08 mA, which was 20% smaller
than the optimally designed value.
B. Results and discussions
Low-frequency measurements were carried out to
evaluate the operation of the memory cello Figure 6 shows a
properly executed quasistatic test pattern, including
NDRO, for fuIl-and half-selected conditions. Current nota
tions are the same as those in Fig. 1. Sense signal Vou, is the
voltage across the sense gate. The sense gate current is ap
plied on both writing and reading. The first half of the pat
tern indicates the corresponding operation for writing and
nondestructive reading of data "I," indicated by voltage
v.,U( across the sense gate. The second half shows the same
for data "0," indicated by zero voltage across the sense gate.
Iy
Is
Vout U >( H H H H "0" H H H H WRRSRSRSRSRWRRSRSRSRSR
-0
-0
-0
-0
FIG. 6. Quasistatic fUllction test patterns demonstrating successful NDRO
memory operations. (W: write operation, R: read operation, and lIS: half ..
selected disturhance;1,: 0.2 mA/div,I,,:0.2mA/div,I,:O.15 mA/div. ~'"':
4mV/div.)
854 J. Appl. Phys., Vol. 65, No.2, 15 January 1989 Iy(mA)
"O·W(:!:24%)
I){(mAl
-0.3
i W ~p....=oc~~
(:!:33%) R(!21%)
-0.5
FIG. 7. Measured operating region (shaded areas) for "0" writing, ''1''
writing, and reading. Circle points are presented the threshold curve oftlle
vortex transitional memory cell.
As shown in Fig. 6, the memory cell operates successfuHy
even after encountering half-selected disturbance. In the
memory cell, the necessary information such as address,
data, and read/write conditions is transmitted only by I,
and Iy-The sense current is simply applied as a clock-pulse
like gate current. It improves the construction for the peri
pheral circuits.
The memory cell switching threshold. deduced from the
function test, is plotted in Fig. 7. A 0,2 mA dc powered
current was applied to the cell to set up an operating origin.
The circles in Fig. 7 illustrate the threshold values for "1"
writing, "0" writing, and reading. The operating regions,
shaded in Fig. 7, show the address current margins Ix = 0.14
rnA ± 24%, 0.21 mA ± 33%, and 0.14 rnA ± 21 %, and
1v=0.15 mA ±24%, 0.17 rnA ±33%, and 0.18 rnA ± 21 %, corresponding to "0," "1" writing, and reading,
respectively. The sense gate margin was measured Is = 0.14
rnA ± 33%. When the data was read, a single flux quantum
entered loop 2 of the cell. This vortex transition was detected
by the sense gate. Therefore, the sense gate margin is almost
independent of the address currents I, and Iv' Each of the
operating current margins is smalier than its designed value
because the Josephson critical current and inductance values
in fabricated chips were, respectively, 20% smaller and 10%
larger than their designed values.
The characteristics of the sense gate for loop 2 are ex
perimentally measured to examine the vortex transitions of
FrG. 8. Threshold characteristics oCthe sense gate measured on an isolated
monitor gate. Vertical axes: the sense gate current (0.1 mA/div). Horizon
tal axes: the external current (0.2 mA/div).
Tahara et al. 854
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15FIG. 9. Estimated circuit parameter by measuring stray and mutual induc
tallce value on isolated gates.
loop 2. In order to measure these characteristics, a monitor
gate consisting of the sense gate and loop 2 is fabricated. This
gate has two gate lines: one is loop 2 external gate line and the
other is the sense gate line. Figure 8 shows the characteristics
in terms of the sense gate current I, and loop 2 external
current Ie' The external gate current values at points "A ',"
"B '," "e '," and "D '" in this figure correspond to those at
Points "G" "I" "J" a d "K'" F' 3 ( ) Th b , , ,n In ·lg. _ a. e a rupt
changes in the sense gate current values at these points indi
cate that the magnetic field entering the sense gate increases
abruptly; that is to say, there is a vortex transition occurring
in loop 2. From estimates of the characteristics of the moni
tor gate, the circuit parameters were determined as fonows:
L] = 7.5 pH, L4 = 1.5 pH, and 12 = 0<08 rnA. One of the
reasons for the differences between these parameters and
their designed values is the existence of stray inductance at
contact holes, Josephsonjunctions, and so o~. Stray and mu
tual inductance measurements at isolated monitor gates pro
duced the circuit parameter estimates shown in Fig. 9< The
calculated threshold characteristic curves of the memory
loop for the experimental parameters are in good agreement
with the experimental data, as may be seen in Fig. 10.
Figure 11 illustrates the quasistatic results of the test for
the memory operation with timing sequence" The patterns
(a)-(d) in Fig. 11 show the writing and nondestructive
reading operations for the four combinations of the se
quences of setting and resetting for Ix and Iv' In pattern (a),
for example, proper operations are demonstrated in the case
Iy (rnA)
0.5
FIG< 10< Threshold curves (circle points) deduced from the quasistatic
function tests. Solid lines show calculated threshold characteristic curves of
the memory loops for the experimental parameters.
655 J. Appl. Phys .• VoL 65. No.2, 15 January 1989 Iv I,
I,
Vout (.) (b)
"1" "cr W 11 R A R
- -~- -, ---
--- --- ~ -- -- ---- ----","",- ~-----
{d)
FIG. I!. Execution of quasistatic pulse test pattern demonstrating propel'
operation with timing sequellce of a designed cell. (I,: 0.2 rnA/div, I,.: 0.4
mA/div, I,: 0.2 mA/div, V;",,: 4 m V /div.) .
of setting Ix earlier than ly and resetting Ix earlier than Iy.
The memory cell is successfully worked regardless of se
quence for Ix and Iy" These results show its capability of
reducing a cycle time of the memory operation.
In the actual memory circuit, the address signal cur
rents to the memory cell are applied from driver gates. In
order to test the operation of the cell with driver gates, the
test circuit illustrated in Fig. 12 was examined. The driver
gate switching time is estimated to be approximately 20 ps
from digital simulations. In this test, the dynamic stability of
the cell is also measured. Test elements consisted of two in
terferometer gates, the cell, and reset gates" In this circuit,
the gate currents and the input current for the interferometer
gates, the sense gate current and the reset gate currents were
supplied from room-temperature pulse generators. The two
interferometer gates drove the cell for X and Yaddress cur
rents. The reset gates returned the applied currents from the
cell to the interferometer gates" The quick pulses from the
driver gates apply to the celL Figure 13 shows the results of a
successful test of the test circuit, including nondestructive
read-out operations and half-selected conditions. The mem
ory ceH dynamicaHy operated propedy<
IV. CONCLUSION
A single flux quantum memory cell, caned a vortex tran
sitional nondestructive read-out Josephson memory cell,
was experimentally tested, using an isolated cell. The cell
employs vortex transitions in the superconducting loops, for
writing and reading information to improve operating toler
ance. Test chips were fabricated using a lift-onplanarization
techniqu.e with Nbl AIOxlNb junctions. Sputtered Nb,
linx Memory
Cell Ir
Is
2 JJ Interferometer Vout
FIG. 12. Test circuit diagram for the vortex transitiollal memory cell with
driver gates.
Tahara et al. 855
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15FIG. 13. Resu1tsofth~success
ful measurements of the cdl
with driver gates, including
NDRO operation and half-se
lected condition (W: write op
eration, R: read operation, and
HS: half-selected disturbance).
Si02, and Mo films were used for interconnections, insula
tions, and resistors, respectively. Successful quasistatic test
patterns were obtained. Cell switching threshold character
istics were deduced from a function test. There was good
agreement with calculated threshold characteristics using
experimental circuit parameters. The experimentally ob
tained current margins were Ix = 0.14 rnA ± 21 %,
ly = 0.16 mA ± 21%, and Is = 0.14 mA ± 33%, in spite
of Josephson critical currents 20% smaner and indictance
values 10% larger than their designed values. It was experi
mentally shown that timing sequence is not necessary in the
memory cell operation. The cell driven by interferometer
856 J. Appl. Phys., Vol. 65, No.2, 15 January 1989 gates was successfully operated, and the dynamic stability of
the cell was evaluated also.
ACKNOWLEDGMENTS
The authors would like to thank H. Abe for his contin
uous encouragement during this work, and J. S. Tsai, H.
Tsuge, M. Hidaka, and S. Nagasawa for their helpful techni
cal comments. The present research effort is part of the Na
tional Research and Development Program on "Scientific
Computing System," conducted under a program set by the
Agency of Industrial Science and Technology, Ministry of
International Trade and Industry.
'w. H. Henkels, J. AppJ. Phys. 50, 8143 (1979).
"W. H. Henkels and J. H. Greiner, IEEEJ. Solid-State Circuits SC-14, 794
(1979).
3K, Kojima, T. Noguchi. and K. Hamanaka, IEEE Electron Devices Lett
EDL-4. 264 ( 1983).
4H. Bena, IEEE Trans. Magn. MAG·IS, 424 (1979).
-'w. H. Henkels, L. M. Gappcrt, J. Kadlec, P. W. Epperlein, W. H. Chang,
and H. Jaeckel, J. App!. Phys. 58, 2379 (1985).
"5. Tahara and Y. Wada, lpn. J. App!. Phys. 26,1463 (1987).
71. Ishida, S. Tahara, Y. Ajisawa, and Y. Wada, Extended Abstract.s of the
19th Conferellce 011 Solid State Device alld Materials. Tokyo 1987 (The
Japan Society of Applied Physics, Tokyo, 1987), p. 443.
Tahara et al. 856
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
129.120.242.61 On: Tue, 25 Nov 2014 15:00:15 |
1.1140379.pdf | Lowtemperature bolometer array
M. Boninsegni, C. Boragno, P. Ottonello, and U. Valbusa
Citation: Review of Scientific Instruments 60, 661 (1989); doi: 10.1063/1.1140379
View online: http://dx.doi.org/10.1063/1.1140379
View Table of Contents: http://scitation.aip.org/content/aip/journal/rsi/60/4?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
The lowtemperature energy calibration system for the CUORE bolometer array
AIP Conf. Proc. 1185, 677 (2009); 10.1063/1.3292432
Rapid temperature variation of hopping conduction in GaAs lowtemperature bolometers
Appl. Phys. Lett. 42, 685 (1983); 10.1063/1.94072
Low Temperature Silicon Thermometer and Bolometer
Rev. Sci. Instrum. 41, 547 (1970); 10.1063/1.1684573
LowTemperature Lab
Phys. Today 17, 70 (1964); 10.1063/1.3051382
LowTemperature Conference
Phys. Today 9, 65 (1956); 10.1063/1.3059842
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
138.51.164.120 On: Thu, 27 Nov 2014 22:26:33Low .. temperature bolometer array
M. Boninsegni,a) C. Boragno, P. Ottonello, and U. Valbusa
Dipartimento di Fisica, Universita' di Genova, Via Dodecaneso 33. 16146 Genova. ltafy
(Received 14 June 1988; accepted for publication 27 December 1988)
The implementation of a 16-channel, low-temperature bolometer linear detector array is
described. The detectors are silicon samples, whose surfaces are doped with phosphorus by the
technique oHonic implantation. A single digital processor implemented on a common PC both
provides the scanning of the array and performs synchronous signal detection from the different
bolometers. The actual system has been tested with a broad infrared source, and some possible
improvements are indicated.
INTRODUCTION
Thermal detectors are widely used in detecting infrared
OR) radiation. In this class of sensors, the energy of the
absorbed radiation raises the temperature of the detecting
element and, as a result of it, changes the properties of the
detector. Bolometers belong to this class of detectors. They
are resistive elements fabricated with a material with a large
temperature coefficient so that the absorbed radiation
changes the value of its electrical resistance.
In order to obtain the ultimate performance from this
class of detectors, Low! was the first to develop a bolometer
operating in liquid helium. Cryogenic bolometers are made
of superconducting materials2 which have a large tempera
ture coefficient. However, semiconducting bolometers are
more widely used than the superconducting ones because
they do not require critical temperature control. Recently,
cryogenic bolometers have been used for detecting IR radi
ation,2 molecular beams,3 ballistic phonons,4 and single par
ticles5; they have been largely used in astrophysics, laser
spectroscopy, surface science, atomic and molecular phys
ics, and solid state physics. In all these applications, either
the position of the detector is fixed with respect to the source
or the detector itself is mechanically displaced through
successive angular positions. Most current far-infrared and
millimeter imaging systems, for instance, depend on a single
detector with mechanically scanned optics, whereas in mo
lecular-beam scattering experiments, the bolometer can ro
tate around the target in order to record the angular distribu
tion of the scattered molecules? For many applications,
however, this approach is inadequate. The required integra
tion time may be in some cases too long, the events can occur
too quickly, or the construction is too complicated. There
fore, the development of a bolometer array becomes particu
larly important, for instance, in constructing ground-based,
airborne, and balloon-borne telescopes6 for infrared astron
omy or in the field of surface science for imaging the diffrac
tion pattern of a molecular beam from a crystal surface as
done in a similar way by a low-energy electron diffraction
(LEED) screen.
Multichannel bolometers 7,8 have been recently used in
plasma diagnostic, In this case the bolometers work at room
temperature and this simplifies the design of the array.
In the present paper we describe a cryogenic array made
of 16 phosphorus-implanted silicon bolometers driven by a microcomputer-controlled system which allows the collec
tion of data. Section I describes the experimental setup with
emphasis on the construction of the array (Sec. I A), on the
calibration procedure (Sec. I B), and on the electronic sys
tem controlling the imaging procedure (Sec. Ie). Section II
reports the results. The array is used to detect the angular
distribution of the radiant intensity of an IR light-emitting
diode (LED) located in front of the array. A discussion on
the performance of the array concludes the paper.
I. EXPERIMENT
A. Bolometer
Each bolometer of the array is realized by ion implanta
tion of phosphorus in a n-Si (100) wafer 300 pm thick and
with a resistivity p = 103 n cm. The implant doses and ener
gies are reported in Table L
This procedure allows one to obtain a surface region
uniformly doped for a depth of = 5600 ;"',9 as resulting from
the Lindhard-Scharff-Schiott (LSS) method; this region has
a net donor concentration of n = lAX 1018 cm-3, close to
the critical value n" = 3.74X 1018 cm-3 for the metal-insu
lator transition. lO Each bolometer of 4 X 2 X 0.3 mm 3, has
been cut out from the wafer and provided with electrical
contacts. The resulting detector is sketched in Fig. 1 (a).
Two gold wires (150 f-lm in diameter) are soldered to
the device by using the following procedure: two gold pads
500 A thick are first realized at both sides of the bolometer
by thermal evaporation; the device is next maintained at a
temperature of about 150"C and then, by flowing current
through the wires, the temperature is locally raised up to the
eutectic temperature of the Au-Si alloy (370 ·C) to produce
the soldering. The procedure is carried out in inert and
slightly reducent atmosphere (90% Nz + 10% Hz) in order
to avoid formation of oxides at the Au-Si interface. The anal-
TABLE L Implant doses and energies of phosphorus in the
.'I-Si( 100) wafer."
Ion energy (keV) 65
Doses (1013cm2) 0.53 lOS
0.83 160
1.26 265
1.99 370
3.32
'The silicon wafers, after the ion-implantation procedure, have been an
nealed for l5 min in N 2 gas at 920 "C and immediately after, for 15 min, in
O2 gas at 920 "Co
661 Rev. Sci. Instrum. 60 (4), April 1989 0034-6748/89/040661-05$01.30 @ 1989 American institute of Physics 661
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
138.51.164.120 On: Thu, 27 Nov 2014 22:26:33COPPEH DISK
FIG. 1. (a) Schematic view ofa single bolometer. The sensitive area is I X2
mm2• (b) Schematic view of the bolometer array. The bolometers are locat
ed close together in order to form a strip of 32 mm in length and ~ m~ m
height. The G-IOeR substrate is 40X IOX2 mm3. The cop~er dls.k IS m
good thermal contact with the bath. The electrical ~onnectlOns Wlt~ the
external electronic system arc thermally connected with the copper disk.
ysis of the interface, carried out with Auger spectroscopy,
did not reveal oxides within the sensitivity of the method.
After calibration (see next section), each bolometer is
attached by General Electric 7031 (G E) varnish onto a
glass-cloth/epoxy-Iaminate (G-lOCR) substrate 2 m~
thick to form a linear array of 16 bolometers; the substrate IS
fixed with a silicon grease onto a copper disk in thermal
contact with the liquid-helium bath. The complete arrange
ment is shown in Fig. 1 (b).
The gold wires on each bolometer are soldered by indi
um alloy onto the copper pads evaporated on the substrate.
These 32 pads are electrically connected to as many copper
wires of 0.5 mm in diameter which link the array to the
external electronic system. Care has been taken in reducing
the input of heat through the copper wires by thermally con
necting them to the copper disk.
The array is inserted in a cryostat schematically shown
in Fig. 2; the working temperature is fixed at 1.2 K by pump
ing onto the liquid-helium bath. In front of the array, at a
distance of5.5 cm, is located an infrared LED (Texas Instru
ment, TIL 903) which is used to test the capability of the
device in detecting the angular distribution of the emitted
radiation.
B. Calibration
The single component of the array has been tested by
measuring the R-T curve in the 4.2/1.2-K temperature
range. For all bolometers we found that the a( T} coeffi-
662 Rev. SCi.lnstrum., Vol. 60, No.4, April 1989 LED light
Li<\N"
FIG. 2. Schematic view ofthe experimental setup. The bolometer array and
the radiation source are maintained at a pressure of 10-5 mbar. The LED
array distane is 5.S cm. The tube diameter is 16 mm. The LED is located
along the tube axis.
cient, defined as [ 1/ R (T ) ] [dR ( T ) I dT ], is the same with
in 1 % and its value at T= 1.2 K is a = -2.9 K-1• After
the realization of the array, we measured, for each bolo
meter, the responsivity S, the response time T, and the noise
equivalent power (NEP).
The responsivity has been determined from the load
curve, as suggested in Ref. 1. The measured values of S are
reported in Table II. The current at the working point is
fixed at 18 /lA.
The responsivity can be also calculatedl by the follow
ing equation:
S = aiR I ( G -ai 2 R), (1 )
once the thermal conductance G between bolometer and
thermostat has been evaluated. A simple model of the bolo
meter has been made by assuming that the sensitive element
can exchange heat with the copper pads across the gold wires
and the copper disk of Fig. 1 (b) across the substrate (Gen
eral Electric varnish, G-IOCR-silicon grease). To calculate
G we took into account both these contributions, Gwires and
G to the thermal conductance. G was calculated by substrate' .
using the values of thermal conductivity and dimenslOns re-
ported in Table III, considering that (a) Gwires and GS~bstrate
are two conductance in "parallel," and (b) Gsubstrate IS the
"series" of G2, G3, and G4•
With these considerations and by using the values of
Table III, we obtained for each bolometer a value of
G = 3.1 X 10-5 W IK. This value is predicted on the base of
the values of Table III and is only a first approximation of
the real situation. Each bolometer, in fact, differs from the
others for several reasons (length of the wires, goodness of
the thermal contact, etc.), and therefore the values of Table
III can vary up to 20% from one bolometer to the other.
By using the values of G determined with this model,
Eq. (1), and the values of resistance (see Table II), we ob
tained the values ofresponsivity S * reported in Table II. The
agreement between the experimental values S and the pre
dicted ones S * confirms the goodness of the model.
Figure 3 reports the response of a bolometer (No.8 of
Table II) to an impinging modulated (20 Hz) square-wave
Bolometer array 662
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
138.51.164.120 On: Thu, 27 Nov 2014 22:26:33TABLE H. Responsivity S and resistance R as measured from the load curve. a
Bolometer 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
S (104V/W) 5.3 5.7 5.7 5.7 5.8 5.9 5.3 5.3 5.1 5.7 5.1 5.9 5.5 5.0 5.9 S.7
R (kH) 18 19 19 19 18 18 14 18 15 19 15 18 20 24 18 19
S* b (10' V /W) 6.7 7.5 7.5 7.5 6.7 6.7 4.1 6.7 4.6 7.5 4.6 6.7 8.5 14.9 6.7 7.5
"The working point is fixed at i= 18 11A. The temperature is 1.2 K.
b S * is the responsivity calculated by using Eq. (l) and assuming G = 3.1 X 10-5 W /K.
radiation. The response time r is 5 ms. According to Ref. 1, r
is given by
r= CI(G-ai2R), (2)
where C is the thermal capacity of the bolometeL C can be
calculated as
This formula takes into account the contribution which
arises from the thermal1inks of the sensor element. 13 Assum
ing for the specific heat (at 1.2 K) ofthe materials constitu
ent the different parts of the detector the following values:
CSi = 4.5 X 10-7 11K g,14 CAu = 8.3 X 10-6 11K g,14 and
CG _ IOCR = 2 X 10' 6 J/K g,15 C results equal to 3 X 10'8 JI
K. By using Eq. (2) and the values of G and C previously
calculated, one obtains for the bolometer No.8, r = 2 ms.
which is in close agreement with the observed one. '
The NEP for all the bolometers is of "'" 10-12 W Imi.
No care has been taken in minimizing it since it is lower than
the noise input equivalent power of the electronic-acquisi
tion system developed in the present work.
c. Electronics
When no radiation is impinging on the detector surface,
a constant voltage Vo = Roi appears across the bolometer,
where Ro is its resistance at the working temperature and i is
the current supplied by a suitable constant-current gener
ator. A change t:..R in the bolometer resistance occurs when
ever radiation is absorbed, and the corresponding change in
voltage, 11 V, is a measurement of the intensity of the imping
ing radiation.
The array is controlled by the electronics shown in Fig.
4, which allows the selection of each bolometer as well as
low-noise amplification of the signal from the detectors. The
TABLE III. Values of thermal conductivity gi' length Li' area A" and ther
rna] conductance Gi of the materials forming the bolometer array."
gi (W/cmK) L, (em) Ai (cml) G, (W/K)
Gold + 0.07% Fe 5 X 10-3 b
single wire
GE 7031 varnish 5X 10" c
G-lOCR 10-4 d
Silicon grease 10-5 e
"The temperature is 1.2 K.
h From Ref. 11, extrapolated at 1.2 K.
C From Ref. 11, extrapolated at 1.2 K,
d From Ref. 12, extrapolated at 1.2 K.
e From Ref. 11. 0.5
0.01
0.2
0.01 8XlO--1 4XIO I
8XlO'2 4XlO'
8Xl0 2 8X1O'
663 Rev. SCi.lnstrl.lm., Vol. 60, No.4, April 1989 IBM PC is equipped with an analog (12-bit resolution) and
digital I/O interface board (LabMaster TM-30). The over
all gain is distributed along the amplification chain in order
not to saturate the analog-to-digital converter whose input
range is switched from unipolar (0/5 V) to bipolar ( ± 2.5
V) passing from Il;) to ~ V measurement. Throughout a mea
surement run, the chopper supplies an interrupt each time
the incoming radiation is turned off. At these instants the
constant current is switched from a detector to the next one
along the array and a measurement is performed within a
period of the modulation. Two successive steps can be out
lined corresponding to the different halves of this period: 0)
in absence of radiation, the dc component Vo is analog-to
digital converted and stored in the main memory of the PC,
and (ii) when the radiation is on, the voltage across the de
tector, Il;) -fj. V, appears at one input of the instrumentation
amplifier, while the previously stored Vo value is fed, via the
DI A converter, to the other input. The large dc component
being removed by subtraction, the small Ll V can be amplified
with a gain factor G2 = 500 to a level suitable for AID con
version. Also all voltage contributions from the on resistance
of the different switches making up the analog multiplexers
are strongly reduced. The subtracting procedure is, in fact,
very effective because passing from step (i) to step (ii) noth
ing in the system changes but impinging radiation. As a re
sult of the two steps, a sample (i.e., a term in the sums yield
ing the averaged values) of both Vo and 11 V is obtained for
the selected bolometer.
It is worth noting that the stored value of b.. Vis formed,
as well as the Il;) value dl,ring the first step, by numerical
FIG. 3. Picture of the oscilloscope output for a square-wave radiation irn
pinging on bolometer No.8. Horizontal axis scale i, !O ms/cm; vertical axis
scale is 50 m V / ern.
Bolometer array 663
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
138.51.164.120 On: Thu, 27 Nov 2014 22:26:33r--------------,
out
61 G = 10 I LAB MASTER
I
~ FIG. 4. Block diagram of the control and
data-acquisition electronics. Both tem
perature and long-term drift ofthe current
generator are actually limited to 5 ppml"C
and 50 ppm/lOoo Hr, respectively, by us
ing high-precision voltage sources and
operational amplifiers,
L ______________ J
averaging over 32 readings performed within the corre
sponding halves of the modulation period. This operation
reduces the input noise level, although the gain in the signal
to-noise ratio is lower than that possible with fully indepen
dent events. A test carried out with the actual instrument
when no radiation is impinging on the array gave a noise
growth factor of about 1.8.jn, where n is the number of sam
ples.
A complete run consists of a preselected number N of
measurement cycles whose length is equal to the number of
detectors ( 16) times the period r of the modulated incoming
radiation.
All the bolometers are cyclically selected and the ratio
LlV(j)IVo(j) is measured for each one (j= 1,2, ... ,16)
during the measurement cycle. The intensity pattern is ob
tained after having averaged over N cycles.
The back~up procedure employed to remove the dc
component assures the measurement of the comparatively
large values Vo(j ) as well as the small Ll V(j ) without in
troducing any time constant. In fact, the alternative simpler
way, based on ac coupling between the detector and amplifi
er chain, causes memory effects, preventing the fast scanning
ofthe array.
Also the 16 time sequences Va [j ; (16k + j) r]
(j = 1,2, ... ,16; k = measurement cycle index
= O,1,2, ... ,N -1) can be stored for a later check of the sta-
bility of individual bolometers and of the system in its entire
ty. Currently, no further exploiting ofthese large amount of
data is foreseen.
A few parameters (number N of cycles, option of storing
the Vo time sequences, or their undersampled versions, etc.)
can be passed to the machine-coded, interrupt-driven rou
tine which, as already said, allows the synchronous scanning
664 Rev. Sci. (nstrum., Vol. 60, No.4, April 1989 of the array, the acquisition of data from the different chan
nels, and the updating, cycle after cycle, of the different aver
ages.
II. RESULTS AND DISCUSSION
Figure 5 reports the radiation-intensity pattern of the
LED source as measured by the linear array. This pattern
has been obtained accumulating 6400 samples (N = 200)
per bolometer. The LED source is supplied with a square
wave at a repetition rate of 10 Hz whose intensity level is set
>f-
(f)
Z
W f
Z
w >
f
<: .J w
cr: ~ '.
lli~ -ID D 10
DISPLACEMENT FROM OTPICAL AXIS (mm)
FIG. 5. Angular distribution of the radiation emitted by the LED source as
detected by the 16-bolometer array compared with the theoretical one
(squares).
Bolometer array 664
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
138.51.164.120 On: Thu, 27 Nov 2014 22:26:33in order to have a signal-to-noise ratio = 1 for the central
bolometer. In the same figure we report a simulation of the
pattern based on the LED characteristics and on the as
sumption that the detected radiation pattern is formed by
two contributions: the first stems directly from the LED
source and the second from a single reflection on the wall of
the tube.
Because of the low resolution of the employed AID and
D/ A converters, the quantization noise is the main limiting
factor for the sensitivity of the apparatus. In fact, the rms
equivalent noise voltage across the bolometer (referred to
the system bandwidth of 200 Hz) due to the independent
contributions ftom the AID and D/ A conversions is
11 D/A voltage output range _1 __ 1_=50 V.
212 ,-G /1
2~3 1
A first, considerable reduction (to = 3 fl V) is easily
achieved by using a 16-bit analog-to-digital interface board.
A further step towards lower noise level is possible because
any quantization noise is generally considered to be white
whenever the AID converter input (signal + analog noise)
changes by at least a few quantization levels between sam
plcs.16 That being the case in our operating conditions, the
input equivalent noise voltage, i.e., the input 110ise power,
can be reduced by lowering the system bandwidth, which
can be obtained by simple averaging. 17 For instance, with
reference to the above figure, a factor of 100 can be gained by
accumulating 10 000 independent samples for each bolo
meter (which requires a 25-min-long measuring run).
6S5 Rev. SCi.lnstrum., Vol. 60, No.4, April 1989 ACKNOWLEDGMENT
We are grateful to Dr. F. Mod who realized the ion
implants.
a) Present address: Physics Dept., Florida State University, Tallahasse, FL.
'F. J. Low. J. Opt. Soc. Am. 51,1300 (1961).
'E. H. PUlley, in Optical and In/rared Detectors. edited by R. J. Keyes
(Springer, Berlin, 1977), pp. 71-100.
IG. Scoles, Ed., Atomic and Molecular Beam Methods (Oxford University,
New York. 1988), Vol. 1.
4c. Boragno, U. Valbusa, and G. Pignatel, App!. Phys. Lett. 50, 583
(1987).
5S. M. Moseley, J. C. Mather, and D. Me Cammon, 1. App!. Phys. 56, 1257
(1984).
"F. J. Low, T. Nishimura, A. W. Davidson, and M. Alwardi, inProcredillgs
of Workshop all Ground-based Astronomical Observations with Illfrared
Array Defectors, Hila, Hawaii, 1987.
7B. Joyc, P. Marmillod, and S. Nowak, Rev. Sci. Instrum. 57, 2449 (1986).
gpo E. Young, D. P. Neikirk. P. P. Tong, D. B. Rutledge, and N. C. Luh
mann, Jr., Rev. Sci. Instrum. 56, 81 (1985).
"c. Boragno, U. Valbusa, G. Gallinaro, D. Bassi, S. Iannotta. and F. Mori,
Cryogenics 24, 6R I (1984).
"'T. F. Rosenbaum. R. F. Milligan. M. A. Paalanen, G. A. Thomas, R. N.
Bhatt, and W. Lin, Phys. Rev. B 27, 7509 (1983).
"G. E. Childs, L. J. Ericks, and R. L. Powell, Thermal cOllductiuityofsolids
at room temperature and below, NBS Monograph No. 131, 1973.
12M. B. Kasen, G. R. MacDonald, D. H. Beekman. and R. E. Schramm, in
Advances in Cryogenics Enginel'ring, edited by A. 1'. Clark and R. P. Reed
(Plenum, New York. 1980), Vo!. 26, p. 235.
I3G. Gallinaro, C. Salvo, and S. Terreni, Cryogenics 26, 9 (1986).
I4JJandhook of Physics and Chemistry, 64th ed. (CRC, Boca Raton, 1983).
151'. Fabbricatore (pl'ivate communication).
;6B. Allen Montijo, Hewlett Packard J. 1988, 70 (June 1988).
i7J. Max, it/ethodes et Techniques de Traitement du Signal et Applications
(lUX Mesures Physiques (Masson, Paris, 1981).
Bolometer array 665
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
138.51.164.120 On: Thu, 27 Nov 2014 22:26:33 |
1.1140027.pdf | Paralleled transconductance ultralownoise preamplifier
Robert B. Hallgren
Citation: Rev. Sci. Instrum. 59, 2070 (1988); doi: 10.1063/1.1140027
View online: http://dx.doi.org/10.1063/1.1140027
View Table of Contents: http://rsi.aip.org/resource/1/RSINAK/v59/i9
Published by the AIP Publishing LLC.
Additional information on Rev. Sci. Instrum.
Journal Homepage: http://rsi.aip.org
Journal Information: http://rsi.aip.org/about/about_the_journal
Top downloads: http://rsi.aip.org/features/most_downloaded
Information for Authors: http://rsi.aip.org/authors
Downloaded 05 Oct 2013 to 128.103.149.52. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://rsi.aip.org/about/rights_and_permissionsParalleled transconductance ultraiow .. noise preamplifier
Robert B. Hallgren
School o/Electrical Engineering, Cornell University, Ithaca, New York 14850
(Received 21 September 1987; accepted for publication 27 May 1988)
A simple NJFET preamplifier was constructed from commercial parts using parallel input
devices in a cascode configuration. The equivalent input noise resistance was 8.5 n (0.38
nV /~Hi) at 1 kHz, and 12 n (0.45 nV /,fHZ) at 100 Hz, measured at room temperature,
independent of the source resistance. For SO-H sources, a gain of29 dB was achieved from 3 Hz to
13 MHz. The input noise equivalent resistance is verified by measuring the thermal noise oflow
valued wire-wound resistors. Circuit utility is demonstrated by noise measurements performed on
GaAs Ohmic contacts at room temperature, under various bias conditions. Design considerations
for using parallel input devices, the bias criteria for them, and possible design extensions arc
discussed.
INTRODUCTION
Much research has recently been devoted to the noise behav
ior and noise mechanisms of various physical systems.] For
accurate measurements, the noise of the system of interest
should be greater than the residual noise of the amplifier
used for the measurement. Whatever the source ofthe noise,
it is limited by the thermal noise of the dc resistance present
in the system. Flicker noise, or any nonthermal sources pres
ent in the system, appear as excess noise above this thermal
background. For detailed measurements of any nonthermal,
or excess, noise in the system, it is necessary, first of all, to be
able to measure the thermal noise present. In this way, the
thermal component can be removed, leaving only the noise
of interest.
This measurement of the thermal noise is limited by the
residual noise of the instrumentation used. In those systems,
where the de resistance is quite low, the instrumentation
must have an input noise resistance (Reg) which is corre
spondingly lower. Such systems are gallium arsenide
MESFETs used in microwave circuits, where the mean
channel resistance is often less than 10 n. The excess noise in
these devices is flicker noise, and the point at which the
flicker component dominates the thermal noise ( 1/1 corner)
is usually at a frequency above 1 MHz.2 To study accurately
the excess noise in these systems, and the bias dependence of
it, the midband noise of the amplifier must be less than the
thermal noise of the channel, and the bandwidth must ex
ceed the 1/1 corner. The preamplifier presented here ad
dresses these needs by paralleling commercial JFET devices,
in a cascode configuration, operating at room temperature.
I. BACKGROUND
The input noise of an amplifier is usually dominated by
the device noise of the input stage. By careful selection of the
devices, the bias points, and the temperature of operation,
the input noise can be reduced. Silicon NJFETs are usually
used as the input devices,3 where the noise power generated
is inversely proportional to the transconductance of the
FET, gm' as given by4
0) where a is a constant,
gm = (2/1 Vp I) (lnss1ns ) 1/2,
and Vp is the pinch-off voltage of the channel.
Increasing gm reduces the noise, but necessitates in
creasing lDss (the saturated drain current at zero gate bias).
For maximum circuit gain, the FET must remain in satura
tion during the entire voltage swing, and this poses a lower
limit to the noise reduction possible. The minimum drain
source voltage for saturation, VnsAT, and the drain current
used, must generate less than the allowed power dissipation
of the device, thus limiting the amount by which loss can be
increased. The quantities loss, VOSAT' andgm are related to
the gate dimensions. By making the PET physically larger,
the noise can be reduced. This increase in the FET size is at
the expense of a larger gate capacitance, limiting the useful
frequency range of the device. What is needed is a way to
reduce the noise of an FET, while remaining within the max
imum power dissipation and allowing a sufficiently wide
bandwidth. Motchenbacher and Fitchen have done this5 by
paralleling separate devices at the input to an amplifier,
thereby creating a much larger FET with a greatly increased
power-handling capacity. This parallel connection has been
used in FET amplifiers6 and op-amps circuits,? all having
similar results, though for different applications.
II. CIRCUIT THEORY
A. Theory of operation
Consider an FET as shown in Fig. 1 (a), biased at some
current IDS' at a voltage greater than VosAr' The noise is
dependent upongm' as given by (2), and for this single PET,
the transconductance is
gm, = (2/j v," I) (loss los ) 1/2 (A/V). (2)
If N such identical devices were connected in parallel
lFig. 1 (b)], biased to a total current of los, each individual
FET would carry a current of Ins = los/No The transcon
j
ductance of an individual FET is thus
(3)
2070 Rev. SCLlnstrum. 59 (9), September 1988 0034-6748/88/092070-05$01.30 @ 1988 American instItute of Physics 2070
Downloaded 05 Oct 2013 to 128.103.149.52. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://rsi.aip.org/about/rights_and_permissions(a) (b)
~~ IDS
Making the transconductance of the parallel combina
tion to be the sum of the individual contributions gives
(A/V). (4)
In this connection, each FET is biased well below 1 DSS'
and contributes a transconductance that is less than the
maximum available for that device. The sum of the separate
contributions, however, is greater than the maximum trans
conductance of the single FET considered above, using the
same power dissipation in both cases. The ratio of the gain
increase is N 1/2, so connecting more devices can give less
noise. The input capacitance of the paranel combination in
creases with N, making the largest number of devices that
can be paralleled limited by the source resistance and the
desired bandwidth.
For wide-bandwidth circuits, a cascade connection is
generally employed to reduce the effects of the increased
feedback capacitance. The cascode FET then becomes the
limiting device, as it is biased to the total drain current bias
ing all of the input devices, and must itself remain in satura
tion. As the number of paralleled FETs increases, the total
bias current increases, so that the power dissipated by this
cascode FET eventually exceeds the rated power dissipation.
B. Circuit description
The circuit tested is shown in Fig. 2, Six input FETs
were used (Q l-Q6), with a single cascode FET (Q7) and a
resistor load (RL ). The input stage (Q 1-Q7) is connected to
a simple source follower (Q8) circuit, biased to give an ap
proximate SOon. output impedance. The circuit operates
ALL TRANSISTORS RRE 2N5434 as
790 UF I-
1329 UF -j 4.84K
10011
Rl-RS
FIG. 2, Preamplilier schematic,
2071 Rev. Sci. Instrum., Vol. 59, No.9, September 1988 FIG, L (a) Single FETbiased ailps' (b)
N paralleled FETs with total current of
los,
from a single 24-V supply obtained from commercial12-V
wet cells.
The input stage uses capacitive coupling from six paral
lel tantalum capacitors to reduce the series resistance of the
capacitors, and to allow adequate coupling to an 8-H source
for frequencies below 10 Hz. Each input FET uses a separate
resistor (RI-R6) for source degeneration, which allows for
variations in the input device characteristics. These resistors
use large bypass capacitors (CI-C6) for extending the low
er-frequency limit.
The bias point selected for the input devices was the
largest current that the cascode FET could handle, while
remaining at a drain-source voltage of 3 V, with a power
dissipation less than one-half the maximum rating (to mini
mize heating). This amounted to 10 rnA per input FET. At
this drain current, the input FETs needed a gate-source vol
tage of approximately -2 V, which was supplied by the
appropriate source resistors, allowing the gate to be biased
through a lOO-MH resistor to ground. The transconduc
tance of the input transistors at this operating point was over
30 mS each, making the total close to 200 mS.
C. Device description
An FETs used in this circuit are commercial parts, ob
tained as samples from Siliconix, and each device was tested
on an HP 4145B semiconductor parameter analyzer. The
2N5434 are low ON resistance, N-channeI JFETs used com
monly for switching applications, Other devices have been
tested, though not in the current design; among these devices
are 2N6550, U3 11, 2N4416, and 2SK1161. These transistors
have all been reported as having low noise,8 but the commer
cial availability and frequency response are unknown. The
saturated drain current was over 100 rnA for all devices,
giving a transconductance greater than 100 mS at a drain
current of 100 rnA. Breakdown voltages are specified at 25 V
and the maximum power dissipation is 300 m W ambient.
Device geometry is quite large, to give the low channel resis
tance for switching, and this results in a gate capacitance of
around 50 pF per device. The saturation voltage is over 3 V
at IDss' and the output conductance is rather large, even
when operating in saturation.
The 2N5434 devices are extremely rugged, due to a
maximum forward gate current of 100 rnA, and at no time
during construction or testing did a device faiL The reverse
gate current was measured to be below 100 pA at room tem
perature and 2-V reverse bias.
Low-noise preamplifier 2071
Downloaded 05 Oct 2013 to 128.103.149.52. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://rsi.aip.org/about/rights_and_permissionsIII. CIRCUIT PERFORMANCE
A. Amplifier measurements
The gain of the amplifier was tested by providing a
-70-dBm input signal from an HP 3336C signal generator,
and r-eading the output signal from an HP 3586C selective
level meter. Additional amplification was provided by a low
noise bipolar postamplifier, which had approximately 1.6
n V / JHi input noise and 33 dB of gain. The gain of the
preamplifier was measured to be 29 dB from 3 Hz to over 10
MHz. The upper frequency limit of the preamplifier depends
upon the source resistance and the total input capacitance.
The 300-pF total input capacitance is high for an rf circuit,
but using a 50-0 source, the circuit operated to 13 MHz, and
only for a source resistance of over 300 n was the bandwidth
reduced to a few MHz. The lower-frequency corner depends
upon the input source impedance/coupling capacitor and
the FET degeneration pole. The values used gave a cutoff of
3 Hz, which can be reduced by using larger bypass compo
nents.
B. Noise measurements
To quantify the noise performance of the preamplifier,
measurements were made of the thermal noise from 5%,
wire-wound resistors at room temperature. The noise tests
were made using an HP 3586C selective level meter, with the
bipolar postamplifier providing additional gain. Multiple
readings of the noise power at various frequencies were taken
and the average computed. The power spectral density was
calculated by dividing the average noise power by the band
width of the filter, which was set to 20 Hz. The noise spec
trum was referred to the input of the amplifier by subtracting
the total gain, and the resultant value expr~ssed as a voltage
spectrum in dBV (V2/Hz). The noise voltage spectra, as
obtained from the resistors and from a shorted input to the
preamplifier, are shown in Fig. 3. The noise from a lOon
resistor is seen to be more than 3 dB greater than the shorted
input noise, indicating that the amplifier equivalent input
noise resistance is less than 10 ·n.
N :r:
"> co
"
<Il The noise measured from the various resistors can be
-170.-------------------------------, ]30 OH~1
lSl OHM
56 OHM
~ -180
o >
<Il
" o
Z 10
.__---------+ 8. 5 OHMS
-190L-~~~L-~~~L-~~~L-~~~
! 02 103 I (] 4 105 ICE
Frequency (Hz)
FIG. 3. Measured noise of wire-wound resistors. The shorted input noise
floor of the preamplifier is shown along with the thermal noise expected
from 8.5 n at room temperature.
2072 Rev. Sci. Instrum., Vol. 59, No.9, September 1988 normalized by subtracting the shorted input noise at each
frequency of measurement. This normalized noise is plotted
in Fig. 4. The right ordinate shows the expected value of the
noise as calculated from the measured dc resistance of each
sample resistor. The agreement is excellent in all cases, and
only for resistances below 70 n can the instrumentation 1/1
noise be seen. Bandwidth reduction is evident only for the
measurement of the 330-H resistor.
C. Equivalent input noise resistance
The input noise resistance can be defined as the resis
tance value that contributes an amount of noise equal to the
amplifier residual. This value is determined by using the data
from the various resistor noise measurements. At each fre
quency, the noise from a resistor is plotted versus the value of
the resistance. These data are plotted in Fig. 5 in terms of
log[e~(/)/e6(/) -1] vs logR, (5)
where e~ ( I) is the noise of resistance R at frequency J, and
e6 (f) is the shorted input noise at!
The regression line from a least.squares fit for the data is
plotted through the data points at two sample frequencies:
150 Hz and 400 kHz. The log R-axis intercepts give the val
ue of the input noise resistance for these two frequencies. In a
similar manner, ReG for each frequency can be found, and
the values obtained are plotted versus frequency in Fig. 6,
where the mean value is around 8.5 fl. This value is very
close to the resistance equivalent of the shorted input noise,
which shows that the preamplifier noise characteristics are
not affected by the source resistance, as expected for FETs.
This fact is also seen from the linearity of the noise plots
yersus resistance value (Fig. 5). If the noise were dependent
upon the source resistance, some nonlinearity would be seen,
and the shorted input noise would be less than that generated
by the equivalent noise resistance. This manner of testing for
noise behavior is particularly useful, as it gives the noise
equivalence very accurately in terms of a resistance for dif
ferent frequencies. Included in the equivalent resistance is
the noise due to any parasitics, such as the series resistance of
the input coupling capacitance, or any frequency depen
dence due to this capacitor.
N :r:
" > rn
'0
" OJ
'" +'
o >
.-o
Z
-1~0L-~~~~~~~~~~~~~
i02 11,3 104 10" 1116
Frequency (Hz)
FIG. 4. Noise of resistors minus amplifier residual. The right ordinate labels
the resistor values and the dashed lines indicate the expected thermall).oise
of this resistance value at room temperature.
Low-noise preamplifier 2072
Downloaded 05 Oct 2013 to 128.103.149.52. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://rsi.aip.org/about/rights_and_permissions2.5r---------------------------~
7 2 * JS800flHZ
150HZ
L1 • ~j -o
..J
?,5
LOG (R,nputl
FIG. 5. Normalized resistor noise as a function of resistance value at two
sample frequencies. The R-axis intercept gives the equivalent input noise
resistance at the frequency plotted.
IV. CIRCUIT UTILITY
As an example of one possible application for this
preamplifier, the noise from a series of Ohmic contacts was
measured as a function of the voltage across them. The mea
surements were made on a test pattern from a production
gallium arsenide wafer. The pattern consisted of a series of
minimum-sized Ohmic contacts with metal interconnects.
The dc characteristics were measured on an HP 4145B semi
conductor parameter analyzer. The current-voltage plot
was linear, the intercept was through the origin, and the
slope was calculated to be 12.5 n. The noise voltage from
this sample was measured at bias voltages of 0, 10, and 50
m V across the sample. The results are plotted, in normalized
form (with the amplifier residual subtracted), in Fig. 7. In
dicated in the figure is the expected thermal noise floor from
the 12.5-0, resistance at room temperature, and the agree
ment is good. The low 1/ f noise corner of the preamplifier
allows the excess noise of the contacts to be seen increasing
as the bias voltage increases. At a bias of 10 mY, the 11/
comer of the amplifier is already below the 1// noise of the
sample.
V. CIRCUIT EXTENSIONS
Extensions of the present design are possible, allowing
some ability to tailor it to any specific application. The low
\2
:1
E ..c
0 10
ij) 9 -
u c 8 -(C
+'
(/) 7
(~
(j) b 0:::
5
102 103 \04
Frequency
FIG. 6. Equivalent input noise resistance of the preamplifier circuit as a
function of frequency as found from the intercept values. The mean value is
seen to he around 8.5 n.
2073 Rev. Sci. Instrum., Vol. 59, No.9, September 1988 N
I
"> rn
"1J --160 r--------------~
.;; -180
-o
Z ~""""'~F .... -...... __ .._!-12.5 OHliS
-I '30 ~J...-I..J...U.'_'_":_-J...-'-'-'-U-l.<L... ........ -J,.Wu.uL .......... --'-'..u.u.J ill 103 11')4 10"
Frequency (Hz)
F.IG. 7. No~se of Ohmic contact test pattern (minus instrument residual) at
different bIas voltages. The sample measured 12.5-H dc resistance, and this
value of thermal noise is indicated by a dashed line,
value of the total input gate current to the preamplifier al
lows direct coupling to the noise source, provided the sample
can supply the one-half of a nanoampere required by the
gates. If the bias voltage needed for the sample is small
enough, direct coupling would be a distinct advantage. The 2
V dropped across the source resistors (RI-R7) biases the
input PETs and allows for some deviation in the de input
voltage from the sample, without appreciable gain devia
tions in the preamplifier. Direct coupling places the low
frequency limit at the FETs' source resistorlbypass pole fre
quency, which is as low as the capacitor can be made large.
An additional FET gain stage can be added directly to
the follower output, and if the total gain were sufficient, no
additional amplifiers would be needed. The present design is
useful as a buffer between existing amplifiers and any smail
signal, low-impedance noise source. The gain of almost 30
dB, and noise floor ofO.44nV I~Hz, anow postamplifiers to
be used that have an input noise floor under 6 n V I[Hz, with
no degradation in the noise performance, Commercial, low
noise op-amps that typically have noise floors of around 4
n V I/Hz-would provide ample additional gain for lower
frequency designs.
Cooling the input devices would further reduce the
noise,9 with some additional concern for high~frequency sta
bility. The number of devices used at the input can be adjust
ed to effect a compromise among the bandwidth, sample
impedance, and the noise floor desired. To accommodate
additional input FETs, the cascode FET could itself consist
of two parallel devices, thereby doubling the power dissipa
tion allowed.
ACKNOWLEDGMENTS
. . The author would like to express gratitude to and appre
CIatIOn for the staff and all concerned at the Microwave
Technology Center of Hewlett-Packard, and to Siliconix
Inc. Special thanks are due to Nicole Bute! for patience and
assistance in preparing much of this effort.
'See, for example, Proceedings of the International Conferences on Noise
and Physical Systems (North-Holland, Amsterdam, 1983, 1984).
low-noIse preamplifier 2073
Downloaded 05 Oct 2013 to 128.103.149.52. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://rsi.aip.org/about/rights_and_permissions2B. Hughes, N. G. Fernandez, and J. M. Gladstone, IEEE Trans. Electron
Devices ED-34, 733 (1987).
3c. D. Motchenbacher and F. C. Fitchen, Low Noise Electronic Design
(Wiley, New York, 1973), Chap. 6.
4A. Van der Ziel, Noise in Measurements (Wiley, New York, 1973).
Sc. D. Motchenbacher and F. C. Fitchen, Low Noise Electronic Design
2074 Rev. Sci. Instrum., Vol. 59, No.9, September 1988 (Wiley, New York, 1973), Chap. 12.
fiP. Bardoni and G. V. Pallotino, Rev. Sci. lnstrum. 48, 757 (1977).
7B. Sundvquist and G. Back.strom, Rev. Sci. lnstrum. 46, 928 (1975).
8D. Bloyet and r. Lapaisant, Rev. Sci. lustrum. 56,1763 (1985).
9S. Klein, W. Innes, and r. Price, Rev. Sci. lnstrum. 56,1941 (1985).
Low-noise preamplifier 2074
Downloaded 05 Oct 2013 to 128.103.149.52. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://rsi.aip.org/about/rights_and_permissions |
1.342547.pdf | Electrical effects of atomic hydrogen incorporation in GaAsonSi
J. M. Zavada, S. J. Pearton, R. G. Wilson, C. S. Wu, Michael Stavola, F. Ren, J. Lopata, W. C. DautremontSmith
, and S. W. Novak
Citation: Journal of Applied Physics 65, 347 (1989); doi: 10.1063/1.342547
View online: http://dx.doi.org/10.1063/1.342547
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/65/1?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Depth dependence of hydrogenation using electron cyclotron plasma in GaAs-on-Si solar cell structures
J. Vac. Sci. Technol. A 17, 453 (1999); 10.1116/1.581605
New mechanism for Si incorporation in GaAsonSi heteroepitaxial layers grown by metalorganic chemical vapor
deposition
Appl. Phys. Lett. 55, 1674 (1989); 10.1063/1.102232
Heterointerface stability in GaAsonSi grown by metalorganic chemical vapor deposition
Appl. Phys. Lett. 51, 682 (1987); 10.1063/1.98333
Formation of the interface between GaAs and Si: Implications for GaAsonSi heteroepitaxy
Appl. Phys. Lett. 51, 523 (1987); 10.1063/1.98386
Activation characteristics and defect structure in Siimplanted GaAsonSi
Appl. Phys. Lett. 50, 1161 (1987); 10.1063/1.97949
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19Electrical effects of atomic hydrogen incorporation in GaAs",on .. Si
Jo Mo Zavada
u.s. Army European Research Office, London NWl 5TH, United Kingdom
S. J. Pearton
AT& T Bell Laboratories, Murray Hill, New Jersey 07974
R. G. Wilson
Hughes Research Laboratories, Malibu, California 90265
C. S. WU,a) Michael Stavola, F. Ren, J. Lopata, and W. C. Dautremont-Smlth
AT&T Bell Laboratories, Murray Hill, New Jersey 07974
S. W. Novak
Charles Evans and Associates, Redwood City, California 94063
(Received 11 July 1988; accepted for publication 13 September 1988)
~ e have introduced atomic hydrogen by two methods into GaAs layers epitaxially grown on
SI substrates, namely, by exposure to a hydrogen plasma or by proton implantation. In both
cases, when proper account is taken of shallow dopant passivation or compensation effects,
there is a significant improvement in the reverse breakdown voltage of simple TiPtAu Schottky
diodes. Proton implantation into un doped (n = 3 X 1016 em -3) GaAs-on-Si leads to an
increase in this breakdown voltage from 20 to 30 V, whereas plasma hydrogenation improves
the value from 2.5 to 6.5 V in n-type (2 X 1017 cm-3) GaAs-on-Si. Annealing above 550'C
removes the beneficial effects of the hydrogenation, coincident with extensive redistribution of
the hydrogen. This leaves an annealing temperature window of about 50·C in the H-implanted
material, in comparison to 150·C for the plasma-hydrogenated material. The hydrogen
migrates out of the GaAs to both the surface and heterointerface, where it shows no further
motion even at 700 ·C. Trapping in the GaAs close to the heterointerface is shown to occur at
stacking fautts and microtwins, in addition to extended dislocati.ons.
INTRODUCTION
There has been an extensive effort in recent years to
grow and characterize GaAs layers on Si substrates. 1-3 The
reasons for this interest are wen documented, but briefly
they relate to the advantages of replacing brittle, small-di
ameter GaAs substrates with larger-diameter Si substrates
of superior thermal and mechanical properties. At some
point i.n the future, it may also be possible to combine the
functions of GaAs-based photonic devices with those of
very-large-scale integration (VLSI) Si electrical circuits, all
on the same chip. At present this optoelectronic integration
is hampered by the fact that aU GaAs layers grown on 8i
substrates exhibit high densities of extended defects, in par
ticular threading dislocations.4 These defects result from the
4% lattice constant difference between GaAs and 8i and
appear in the initially coherently strained GaAs after a few
hundred angstroms of growth. Regardless of the lattice mis
match between the III-V layer (GaAs, GaP, InP) and the
group-IV element substrate (Si or Ge), there appears to be
an almost invariant defect density of 107_108 cm·2 at dis
tances of -1 /Lm from the heterointerface.4 This may well be
an interaction-distance argument in the sense that near the
interface an initially high density of defects can tangle and
terminate. This leads to a reduction in defect density with
distance from the heterointerface until the remaining defects
become far enough apart that their probability for interact
ing with each other becomes small. This appears to occur at a
a) Permanent address: Hughes Aircraft Co., Torrance, CA 90S09. distance of 1-10 p,m, corresponding to a defect density of
107_108 cm·--2•
The performance of electrical devices fabricated on
GaAs-on-Si is characterized by the presence of high reverse
bias voltage leakage currents, whose origin is clearly related
to the high defect density in the materiaLS The mechanism
for production of these excess leakage currents is not, how
ever, quite so clear. Intuitively, one might expect that recom
bination at the extended defects would be a major contribu
tor, although there is some evidence that defect-assisted
tunneling may in fact be the dominant mechanism for the
leakage current. (, The presence of the defects in the GaAs
layer is perhaps even more deleterious to the performance of
photonic devices, especially lasers. The defects tend to be
mobile under minority-carrier injection and agglomerate in
the active region of the laser, forming nonradiative areas.
This obviously degrades the light output from the device and
eventually leads to the termination of lasing action.7
It is clearly of interest to examine the effects of atomic
hydrogen incorporation into this highly defected material
system. Hydrogenation has previously been shown to passi
vate or neutralize the electrical activity of a wide range of
impurities and defects in semiconductors and might be ex
pected to reduce the defect-related leakage currents in
GaAs-on-Si diode structures. x We have previously reported
this effect for the case in which the hydrogen was i.ntroduced
by exposure of the GaAs-on-Si to a hydrogen plasma.9 In
some respects a more controlled method of incorporating
the hydrogen is by ion implantati.on, in which a known dose
can be placed at known depths in the GaAs. The disadvan-
347 J. Appl. Phys. 65 (1), 1 January 1989 0021-8979/89/010347 -07$02.40 © 1988 American Institute of Physics 347
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19tage of this technique, of course, is the introduction oflattice
damage by the implanted ions, which requires annealing at
elevated temperatures. The major question is whether this
damage can be annealed out without removing the beneficial
effects of the hydrogen. A related point of interest is the
extent of any redistribution of the hydrogen during the post
implant anneal. The GaAs grown on Si is highly strained and
defective, and the motion of hydrogen within it might be
expected to be somewhat different from conventional GaAs.
In this paper we compare hydrogenation of GaAs-on-Si
by both plasma exposure and ion implantation, compare the
annealing required to restore the original conductivity, and
examine the redistribution of the hydrogen during post-hy
drogenation annealing. We observe a correlation of the
amount of hydrogen incorporated during plasma exposures
with the amount of initial disorder in the GaAs layer and
note a somewhat surprising thermal stability of hydrogen
located around the heterointerface region. We have predom
inantly used leakage current-voltage (1-V) measurements as
a qualitative indication of the concentration of electrically
active defects. Deep-level transient spectroscopy is not sensi
tive to the types of defect in GaAs-on-Si, and so this is the
reason we use the somewhat indirect 1-V measurements.
EXPERIMENTAL DETAilS
The GaAs layers were deposited onto the Si substrates
using a three-step technique which is basicaUy standard
these days. We used 2-in.-diam, I-n em, p-type (B-doped)
Si cut 4· off (l00) toward the [011], which was solvent
cleaned and lightly etched before being loaded into a vertical
geometry metalorganic chemical vapor deposition
(MOCVD) reactor. 10 The Si substrates were then heated at
900 ·C for 10 min under AsH3 to thermally desorb native
oxide from their surfaces. The substrate temperature was
then lowered to 450°C to nucleate the growth ofGaAs, with
deposition of ~ 100 A of material. Following this, the wafer
temperature was raised to the growth temperature of
-650 ·Cfor deposition of the GaAs at a rate of -4,um h.-I
The final layer thicknesses varied frem 1.5 to 10 pm. Capaci
tance-voltage (C~ V) profiling showed that all of the un
doped GaAs layers were n type with net carrier densities in
the range 1-3 X 1016 cm--3. Companion samples were exam
ined by both plan-view and cross-sectional transmission
electron microscopy (TEM). The defect structures and den
sities observed in the material were similar to those reported
previously by many authors,4 and discussed earlier in this
paper.
TABLE I. Types of GaAs-on-Si investigated.
Structure
No. GaAs layer
sequence Doping We investigated hydrogenation in three basic layer
structures summarized in Table 1. The first was simply to
implant protons into the undoped GaAs. This was done both
at high doses (1016 cm -2 at 100 keY), for the purpose of
monitoring the redistribution of the implanted species upon
annealing, and at low doses (5X 1013 cm-2 at 100 keY) to
try to passivate the electrical activity of some of the defects in
the material. The second type of structure consisted of a 0.15
,um-thick n-type region (n~3 X 1017 cm-3) formed by im
plantation of 29Si ions at a dose of5 X 1012 cm -2 (100 keY of
energy), into the undoped GaAs. As we discussed in a pre
vious paper," this simulates the depletion region of field-ef
fect transistor structures, the most common electrical device
used in GaAs technology. The implanted Si was activated by
proximity rapid annealing at 900 ·C for 10 s. This annealing
treatment reduced the microtwin and stacking fault density
in the GaAs layer, but the threading dislocation density re
mained essentially unchanged.!! These n-implanted GaAs
structures on Si were hydrogenated by exposure to a 30-kHz,
O.08-W cm--2 plasma contained within a parallel plate, ca
pacitively coupled reactor. The samples were held at 250°C,
and the exposure time varied from 0.5 to 3 h. After each
plasma treatment, these samples were annealed at 400 ·C for
5 min in N2 to restore the electrical activity of the shallow
donor impurities in the material. This anneal is necessary to
ensure that we can make valid comparisons of hydrogenated
GaAs-on-Si with unhydrogenated material of the same dop
ing density. We have previously demonstrated that such an
anneal is sufficient to remove shallow-donor passivation in
GaAs grown on GaAs or Si. !2 The third type of structure
consisted of -2.um ofSi-doped n+ (2X 1018 cm-3) GaAs
grown on the Si, followed by 8 f.lm of undoped GaAs. Some
samples were given an in situ anneal under AsH3 at 750·C
for 10 min after deposition of the n -{-layer, in order to reduce
the interfacial disorder. Companion samples were grown
with a similar structure, but without the annealing step. The
doping concentration in the undoped GaAs was similar in all
samples (n-3X 1016 cm-3). The purpose of both types of
samples was to examine the effect of the presence of varying
degrees of lattice disorder on the total amount of hydrogen
incorporated into the GaAs-on-Si.
The electrical effects of hydrogen incorporation were
examined by current-voltage (I-V) measurements in
TiPtAu Schottky diode structures. The TiPtAu contacts
were deposited by electron-beam evaporation through a
shadow mask, to a total thickness of2S00 A. Ohmic contact
was made by a low-temperature (_325°C) anoy of In on
Hydrogenation method
1
2 3-,um undoped
0.15-,um n-type
2.85-,um undoped n~3x 1016 em-J
n-3 X 1017 ern -3
n-3X1016cm 3 H+ implant; sx 10"_10IE em 2,IOOkeV
H plasma 250 'c, 0.5-3 h
3
348 8-,um undoped
2-Pfi n+ GaAs
J. Appl. Phys., Vol. 65, No. i, 1 January 1989 n-3XlO '6cm'
n-2X 10'8 em-, H plasma 250 ·c, 0.5-3 h
Zavada et at. 348
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19lOp.m Gata-ON-Si
AS -GROWN lOf'!'I'l GaAs -ON-Si
IN -SITU ANNEALED
'? 1020 D PLASMA O.5h, 250"C 106
~ r-------~----, o PLASMA 0.5h, 250°C 106
FIG. 1. SIMS profiles of deuterium in
GaAs-on-Si samples grown either with or
without an in situ anneal and subsequently
exposed to it D plasma for 0.5 h at 250°C,
o 2 4 6 8 10 12
DEPTH (f\-m) o 2 4 6 8 10 12 14
DEPTH {fLml
the front face of the samples. The atomic profiles of hydro
gen or deuterium in the implanted or plasma-treated materi
al were obtained using negative secondary ion mass spec
trometry (SIMS) measurements with Cs-+ -ion
bombardment in a Cameca IMS 3fsystem.13 The concentra
tions obtained in this way were calibrated by comparison
with implanted standards and the depth scales established
by stylus measurements of the sputtered crater depths. The
former are usually quoted to be accurate to within a factor of
2, while the latter are generally accepted to be accurate to
±7%.
RESULTS AND DISCUSSION
The amount of hydrogen or deuterium incorporated
into semiconductors depends on a number off actors related
to the density of sites to which it can bond. These sites in
clude dopants, defective bonds, and regions of strain in the
material associated with line and point defects and certain
types of impurities.8 The high level of lattice disorder near
the heterointerface of GaAs-ou-Si might be expected to at
tract a significant density of hydrogen. To examine this we
exposed the to-,um-thick GaAs layers on Si (layer structure
3 in Table 1) to a deuterium plasma for 0.5 h at 250"C.
Figure 1 shows the SIMS profiles obtained from samples
that either had or had not received the 750 "C, lO-min an
neal. There are two components to the D profile in each
sample. The ben-shaped distribution within the first 5,um is
typical of that observed in plasma-exposed GaAs. It does not
correspond to a classical error-function profile for unimped
ed, one-species diffusion. Based on our current understand
ing of the permeation of hydrogen into semiconductors, it is
possible that the SIMS profile in this region represents deu
terium present in at least two forms. The first is deuterium
complexed with the shallow-donor impurities in the GaAs.
These are present at a concentration of only _1016 cm-3,
and therefore there must be at least one other form of deuter
ium present at a concentration ofS X 1017 cm-3• This almost
349 J. Appl. Phys .• Vol. 65, No.1. 1 January 1989 certainly includes some form of dusters of deuterium, possi
bly as simple as deuterium molecules, or may be larger asso
ciates such as the extended platelets observed in proton-im
planted GaAs which has been annealed above 200 QC. 14 The
spike in the distributions near 3.5 pm corresponds to a
growth-interruption step during the GaAs deposition and
probably represents deuterium accumulation at interfacial
defects or impurities.
The second component in each D profile occurs at
depths between 8 and 10 pm. In both samples this region is Si
doped to a level higher than that of the overlying 8 lim of
GaAs, and so one might expect more deuterium to accumu
late there. However, there is clearly less deuterium between
8 and 10 pm in the sample that received an in situ anneal.
This sample contained less disorder near the heterointerface
than the un annealed sample, as measured by He-ion chan
neling and cross-sectional TEM. There was a complete ab
sence of stacking faults and microtwins in the in situ an
nealed material, and the backscattering yield at a depth of 8
/-lm was 36%, compared with 49% in the unannealed GaAs
on-Si. This is consistent with the previously observed char
acteristics of in situ annealed material. ]5 The increased con
centration of deuterium near the heterointerface in the latter
sample therefore represents the combined influence of stack
ing faults, microtwins, and other defects which can bond
deuterium.
Capacitance-voltage profiling of the GaAs-on-Si after
hydrogenation showed reductions in the carner density in
the first -1 lim from the surface in all samples. Typically,
there was a reduction of approximately an order of magni
tude within this region, corresponding to passivation of the
shanow donors in the material. The initial doping levels were
restored by annealing at 400 ·C for 5 min but even after this
treatment we observed significant reductions in diode re
verse leakage current in hydrogenated material. Figure 2
shows reverse J-V characteristics from the Si-implanted
GaAs-on-Si material, hydrogenated for 3 h at 250 ·C, an
nealed at 400"C to restore the shallow-donor doping, and
Zavadaefal. 349
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19'10-10 OL.L.L-~2--!3~-..J,4----'5~--:!6'---±-7---!S
VR (VOLTS)
FIG. 2. Reverse-bias 1-V characteristics from TiPtAu diodes fabricated on
S1-implanted (n~ 3 X IO{7 em -3) GaAs-on-Si either untreated or plasma
hydrogenated (3 h, 250 'C), followed by annealing at 400 'C for 5 min to
restore the shallow-donor activtiy.
then processed into diode structures. These diodes show
breakdown voltages, defined as the reverse bias at which the
leakage current is 1 rnA, of -6.5 V, compared with 2.5 V for
unhydrogenated diodes. We emphasize that c-V measure
ments showed that doping concentrations were identical in
the two types of samples, with the only difference being that
hydrogen is still presumably bound at defect sites in the plas
ma-treated material, even after the anneal to restore the shal
low doping level. Diodes formed in exactly the same fashion
on homoepitaxial GaAs of the same doping density showed
reverse breakdown voltages of ~ 8 V and displayed no im
provement upon hydrogenation and annealing at 400 °e. We
varied the plasma exposure conditions for the GaAs-on-Si
over the temperature range 125-250°C, and from 30--180
min, but were unable to achieve diode breakdown voltages as
high as in the homoepitaxial diodes. This could be due to
several factors, including the possibility that some passivat
ed defects were reactivated by the 400°C anneal to restore
the shallow doping concentration, or that not all of the elec
trically active defects were passivated by hydrogen. We have
no way to distinguish these possibilities, although it is typical
of many hydrogenation experiments to observe only partial
passivation of defects or impurities. Passivation of the intrin
sic defect levels in molecular-beam cpitaxially grown
GaAs16 and of DX centers in AIGaAs, 17 all of which showed
complete passivation to the deep level transient spectroscopy
(DLTS) detection limit, are notable exceptions.
It is worth mentioning at this point that not only was the
reverse breakdown voltage in the GaAs-on-Si altered by hy
drogen-plasma exposure, but the Schottky barrier height de
termined from the J-V characteristics was also changed, as
shown in Table II. In untreated samples the barrier height
was measured to be 0.67 V, while after a 3 h, 250°C plasma
exposure, followed by 400 °C annealing and deposition of the
TiPtAu, the barrier height was reduced to 0.52 V.9 In sam
ples hydrogenated by proton implantation, we observed im-
350 J. Appl. Phys., Vol. 65, No.1, 1 January 1989 TABLE II. Average ideality factors (n), barrier heights (,pe). and break
down voltages (VB) in GaAs-on-Si diodes, obtained from 1-V measure-
ments.
Layer structure 2
Untreated
n ifJB (eV) VB (V)
1.35 ± 0.08 0.67 ± 0.02 2.44 ± 0.07
Hydrogenated
n ifJB (eV) VB (V)
1.32 ± 0.Q1 0.52 ± O.Ql 6.48 ± 0.27
Layer structure 3
Untreated
n ,pB (eV) VB (V)
1.28 ± 0.06 0.71 ± 0.02 19.54 ± 0.58
Hydrogenated
n o/B (eV) VB (V)
1.29 ± 0.04 o.n ± 0.Q3 30.30 ± 0.95
provements in reverse breakdown voltage, but no change in
barrier height, as also shown in Table II. This is consistent
with our previous assumption that the change in barrier
height in plasma-treated material is due to removal of free
As and its oxides from the surface as AsH) and water
vapor.9,tH
The improvement in reverse breakdown voltage was not
stable for annealing above 550°C, decreasing to 3.5 V after a
600°C, 5-min treatment. Figure 3 shows the atomic profiles
of deuterium in a plasma-treated sample after annealing at
400°C for 5 min. There was no motion of the deuterium up
to 400 °C, at which temperature some redistribution is evi
dent, with the onset of pileup at the heterointerface. After
600 ·C annealing there is diffusion of deuterium both toward
the surface and to the heterointerface where there is a sub-
1020 4p.m GaAs -ON -Si lOS D PLASMA O.5h, 250°C
..., 5 MIN ANNEALS
I
E
'"
1019 to5 m z I-
0 Z
I-::l
0 <1 AS -TREATED ~ u 0::: I-104 Z
Z Q lJ.!
(,) . >-Z 0:::
0 400QC ! <1 u 1017 103 0 'f 1 :'--Ga-+ z :ii: 0 ::l '...., e e U
a::: \I,...L! w
, I' m w : \ I-
::l 10i6 .. 102
W
0
600·e
1015 101
0 2 4 6 8
DEPTH (ftm)
FIG. 3. SIMS profiles of deuterium in undoped GaAs-on-Si treated in a
plasma for O.S h at 250'C and subsequently annealed for 5 min at 4OO·C or
600 'C.
Zavadaetal. 350
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19II')
\ e
Q 1021
Wi 9
>-
f-
Cf)
Z
W o
Z
W
(!) o
0:: o
)0-
J: 300"C
DEPTH (,urn)
FIG. 4. SIMS profiles of hydrogen in proton-implanted (10'6 em -2, 100-
ke V) GaAs-on-Si as a function of post-implant annealing temperature (20-
min anneals).
stantial accumulation. This is the region of maximum disor
der in the material and emphasizes once again that hydrogen
and deuterium are attracted to any site of strain in semicon
ductors. It is interesting that the hydrogen (or deuterium)
must be in an atomic state, since molecules show no evidence
of significant motion or trapping in any semiconductor.19
After trapping, however, the hydrogen is strongly bound
and upon annealing shows no ability to passivate dopants. It
is therefore in an apparently inactive state.
The accumulation of hydrogen at the heterointerface
upon annealing was even more evident in proton-implanted
material (structure 1 in Table I). Figure 4 shows SIMS pro
files of hydrogen in a sample implanted with IOO-keV H+
ions to a dose of 1 X 1011> cm-2, followed by annealing up to
700 ·C for 20-min periods. In this case there was little motion
at 200 ·C, but some slight redistribution at 300 ·C, especially
on the tail of the implanted profile. With increasing anneal
ing temperature, hydrogen is lost to the surface, but there is a
tremendous accumulation at the heterointerface. The sur
prising result is that this accumulation is stable to 700·C
annealing, and even in the original implanted region as hy
drogen is lost by diffusion, the remaining hydrogen retains
the profile shape of the implanted distribution. This indi
cates that there is still some remnant damage in the GaAs
even after 700°C annealing, and that the hydrogen is decor
ating this damage. The enhanced accumulation near the he
terointerface in the implanted material compared with the
plasma-treated GaAs-on-Si may be slightly misleading, be
cause it must be remembered that the implanted layer was
only 2.5 f..tm thick and therefore had poorer crystalline quali
ty than the 4-,um-thick sample that underwent plasma expo-
351 J. AppL Phys., Vol. 65, No. i, 1 January 1989 2,um GaAs-ON-GaAs
.. Ht 5)\ 1013 cm-2
.. 3He+3xIO '3cm-2 ....
I \
\
'" \
\
$
\
\\
~
\
lOT \
1O't \
: INITIAL SHEET RESISTANCE
1021 I I I ! I I
o 100 200 300 400 500 600 700 800
ANNEALING TEMPERATURE (OC)
FIG. 5. Sheet resistance of n-type (\0.7 em -3) GaAs layers implanted with
multiple-energy (30-, 100-, and 20G-keV) H+ or 'He ~ at doses of 5 X 10'3
or 3 X JO l:l em -2, respectively, as a function of post-implant annealing tem
perature.
sure. The anneals in the former case were also for 5 min only,
while in the latter case they were for 20 min, and deuterium
was used in the plasma exposure compared with implanted
hydrogen. Taking an these factors into account, there ap
pears to be a roughly similar rate of accumulation at the
interface for the two methods of hydrogen introduction.
The obvious problem with the use of conventional high
energy implantation as a technique for hydrogenating
GaAs-on-Si is the introduction of the lattice damage so evi
dent from the results in Fig. 4. The question is whether, for
the dose levels that might actually be used in device struc
tures, this damage can be annealed while still retaining the
beneficial effects of hydrogen. The first thing to determine is
the annealing temperature required to remove the hydrogen
implant damage. Figure 5 shows the sheet resistance of 2-
pm-thick n-type (_1017 cm-3) GaAs layers grown on
semi-insulating GaAs substrates, after proton implants at
multiple energies (30, 100, and 200 ke V) at a dose of 5 X 10 \3
em -2, and then annealed for 5 min at the indicated tempera
tures. The evolution of the sheet resistance with annealing
temperature can be explained by the introduction of dam
age-related deep levels which trap electrons in the GaAs,
increasing the resistivity of the material after implantation.
However, the damage sites are close enough that electrons
can hop from one to another, leading to a low-mobility con
duction. The hopping conductivity is reduced as some ofthe
damage is annealed out with increasing annealing tempera
ture, leading to an increase in the resistivity. At some tem
perature (around 300 ·C for proton implants) the deep-level
density falls below that of the donor concentration and eIec-
Zavada et at. 351
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:191018
H 5 x 1013 cm~2 100 keY
...... GaAs -ON-Si
---'--GaAS! Si
1017
", ,
E
t)
<{
Z
i
0
Z
1016
DEPTH (fLm)
FIG. 6. Carrier profiles in GaAs-on-Si implanted with tOO-keY H+ ions at a
dose of 5 X 10 13 em -2, and subsequently annealed for 5 min at either 400 aT
500 "C.
trons are returned to the conduction band, lowering the re
sistivity until eventually it reaches its unimplanted vaIue.20
This occurs at 500 ·C for this particular dose of protons into
GaAs. It is worth noting that even the use oeRe +-ions, also
shown in Fig. 5, shifts the annealing curve somewhat to
higher temperatures, and therefore the use of another defect
passivating species, such as Li, is probably precluded by the
extra annealing required for heavier ions.
As a further check that 500·C annealing restores the
initial condition of the GaAs lattice for proton implants at
doses around 5 X 1013 cm -2, we made electrochemical C-V
measurements on implanted GaAs-on-Si samples after sev
eral annealing treatments. Figure 6 shows the initial carrier
profile and after a 5X lOLl cm-2, tOO-keY H+ implant. In
the latter case the reduction in doping will be due predomi
nantly to the damage introduced with possibly a small com
ponent due to donor passivation by hydrogen. After anneal
ing a 400 ·C this latter effect will be removed, as will some of
the damage (refer to Fig. 5). Finally, we see that 500·C
annealing restores the carrier density to its initial value.
Based on this information, we can look for the beneficial
effects of hydrogen incorporation by implanting protons
into the GaAs-on-Si, annealing at 500 ·C to restore the initial
carrier density, and comparing the 1-V characteristics of a
diode structure with that of an unimplanted companion. The
reverse-bias J-V data from TiPtAu diodes fabricated on un
doped (n = 3X 1016cm-3) 2-3-.um~thickGaAslayersonSi
are shown in Fig. 7. The untreated sample had a reverse
breakdown voltage of ~ 19.5 V, whereas the hydrogenated
diode shows a value of -30 V. An unimplanted sample that
352 J. Appl. Phys., Vol. 65, No.1, i January 1989 10-3
10-4
10-5
10~r
'< It: 10-7
....
10~8
10-9
10-10
10-11
0 ;0 TIPIAu
A"2.08x10-:~cm-2
i5 20 25 30 35
VR (VOLTS) 40
FIG. 7. Reverse-biasl- Vcharacteristics from undoped (n-3 X 10'6 em-3)
GaAs-on-Si samples processed into TiPtAu Schottky diodes. One of the
samples was implanted with lOO-keY H+ ions (5 X 1013 cm-2) and an
nealed at 500·C for 5 min prior to metallization.
also underwent a 500 ·C anneal showed a similar breakdown
voltage as the control diode (i.e., 19.5 V). Therefore, the
proton implant appears to be effective in improving the di~
ode characteristics of GaAs-on~Si structures by passivating
the electrical activity of some of the defects in the material.
Once again, however, the breakdown voltage of even a hy~
drogenated diode was inferior to one fabricated on homoepi
taxiaI GaAs of the same doping density. In the latter case we
observed a diode breakdown voltage of -43 V. We note also
that the thermal stability of improvement in performance of
the implanted diodes was similar to that of plasma~exposed
samples. The only difference between the two methods of
hydrogen introduction was the fact that there was no lower
ing of the Schottky barrier height for proton-implanted sam
ples.
CONCLUSIONS AND SUMMARY
We have compared hydrogenation of GaAs-on-Si by
two different methods: Hz-plasma exposure and proton im
plantation. In both cases there is a significant improvement
in reverse breakdown voltage of TiPtAu Schottky diodes,
compared with unhydrogenated diodes. This improvement
is presumably due to a reduction in the number of electrical
ly active defects in the GaAs-on-Si upon hydrogenation.
There is extensive redistribution of hydrogen to the heteroin
terface at annealing temperatures above 500 °C, and the hy
drogen appears to be in a strongly bonded form when it is in
the interface region because of its subsequent thermal stabil
ity. It could indeed be in several forms, such as bound to
defects or dangling bonds, or in a clustered state. Since after
high-temperature annealing there is no apparent dopant pas
sivation during stimuli such as minority~carrier injection,
the hydrogen is apparently in an inactive state.
I t is worth emphasizing that the defect passivation in the
Zavada at al. 352
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19material is incomplete, a frequent feature of hydrogenation
experiments. Therefore, the incorporation of hydrogen is
not a panacea for the high defect density in GaAs-on-Si, but
rather it indicates the important role these defects play in
degrading the electrical quality of material. As is widely re
cognized, the future utility of GaAs-on-Si depends on mak
ing real progress in reducing the defect density from the cur
rent value of ~ 108 to 104 cm--2 or less.
ACKNOWLEDGMENTS
The authors acknowledge the supply of some of the
GaAs-on-Si material from So Mo Vernon and V. E. Haven
(Spire Corporation), and the interest of A S. Jordan. The
ion-channeling results were provided by K. T. Short (AT&T
Bell Laboratories).
'R. M. F1etcher, D. K. Wagner, and J. M. Ballantyne, App!. Phys. Lett. 44,
967 (1984).
2R, J. Fischer, W. F. Kopp, J. S. Gedymin, and H. Morko,<, IEEE Trans.
Electmn Devices ED-33, 1407 (1986).
3T. H. Wind hom, G. M. Metze, R-Y. Tsauf, and J. C. C. Fan, App!. Phys.
Lett. 45,309 (1984).
353 J. Appl. Phys., VoL 65, No.1, 1 January 1989 'See, for example, papers in Proc. Mater. Res. Soc. Syrup. 91 (1987).
5N. Chand, F. Ren. S. J. Pearton, N. J. Shah, and A. Y. Cho, IEEE Trans.
Electron. Device Lett. EDL .. S, 185 (1987).
6N. Chand, R. Fischer, A. M. Sergent, S. J. Pearton, D. V. Lang, and A. Y.
Cho, App!. Phys. Lett. 51, 1013 (1987).
7J. P. Van del' Ziel, R. D. Dupuis, R. A. Logan, R. M. Mikulyak, C. J.
Pinzone, and A. Savage, Appl. Phys. Lett. SO, 456 (1987).
"S. J. Pearlon, J. W. Corbett, and T. S. Shi, App!. Phys. A 43,153 (1987).
95. J. Pearton, C. S. Wu, M. Stavola, F. Ren, J. Lopata, W. C. Dautremont ..
Smith, S. M. Vernon, and V. E. Haven, App!. Phys. Lett. 51, 496 (1987).
IllS. M. Vernon, V. E. Haven, S. P. Tobin, and R. G. Wolfson, 1. Cryst.
Growth 77,530 (1986).
lIS. M. Vernon, S. J. Pearton, J. M. Gibson, K. T. Short, and V. E. Haven,
AppL Phys. Lett. 50, 1161 (1987).
12J. Chevallier, W. C. Dautremont-Smith, C. W. Tu, and S. J. Pearton,
App!. Phys. Lett. 47,108 (1985).
13Charles Evans & Associates, Redwood City, CA.
14H. C. Synman and J. H. NecthHng, Radiat. Eff. 69, 199 (1983).
"Y. Ito, Appl. Phys. Lett. 52, 1617 (1988).
I·W. C. Dautreruont-Smith, J. C. Nabity, V. Swaminathan, M. Stavola, J.
Chevallier, C. W. Tu, and S. J. Pearton, Appl. Phys. Lett 49, 1098
(1986).
17J. C. Nabity, M. Stavola, J. Lopata, W. C. Dautremont-Smith, C. W. Tu,
and S. J. Pearton, Appl. Phys, Lett. 50, 921 (1987).
'"F. Capasso and G. F. Williams, J. Electrochem. Soc. 124, 821 (1983).
19S. J, Pearton, J. W. C:orbett, and T. S. Shi, AppL Phys. A43, 153 (1987).
2OS. J. Peart on, J. M. Poate, F. Sette. J. M. Gibson, D. C. Jacobson, and J. S.
Williams, Nucl. lnstrum. Methods B 10/20, 369 (1987).
Zavada et al. 353
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
131.94.16.10 On: Sat, 20 Dec 2014 23:59:19 |
1.1140557.pdf | "High performance xray area detector suitable for smallangle scattering,\ncrystallographic, and kine(...TRUNCATED) |
1.576238.pdf | "Particle bombardment effects on thinfilm deposition: A review\nD. M. Mattox \n \nCitation: Journal (...TRUNCATED) |
1.101217.pdf | "Highfield perpendicular conduction in GaAs/AlAs superlattices\nA. Sibille, J. F. Palmier, C. Minot,(...TRUNCATED) |
End of preview. Expand
in Dataset Viewer.
README.md exists but content is empty.
Use the Edit dataset card button to edit it.
- Downloads last month
- 37