JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 7, NO.

3, MARCH 1989 473

Multiplication Noise of Wide-Bandwidth

InP/InGaAsP/InGaAs Avalanche
Photodiodes
J. C. CAMPBELL, MEMBER,IEEE, S . CHANDRASEKHAR, W. T. TSANG, G. J. QUA, MEMBER,IEEE,
AND B. C. JOHNSON

Abstract-Recently, high gain-bandwidth products have been electron and hole ionization coefficients in InP tend to
achieved with InP/InGaAsP/InGaAs avalanche photodiodes (APD’s) converge at high electric fields. As a consequence, the
having thin ( < 1 pm), heavily doped ( N d 2 2 X 10’6/cm3) multipli-
cation regions. This doping profile leads to very high electric fields near
wide-bandwidth SAGM-APD’s would be expected to ex-
the p-n junction. Since the electron and hole ionization coefficients of hibit high multiplication noise [161-[ 181. In this paper we
InP tend to converge at high fields, these APD’s should exhibit high report measurements of the multiplication noise of these
multiplication noise. We have measured the multiplication noise in wide-bandwidth SAGM-APD’s. We have found that the
APD’s of this type and we find that it is lower than would be predicted excess noise factors are lower than would be predicted by
by the ratio of the weighted ionization coefficients. We have also deter-
mined the gain-bandwidth product from the frequency response of the
averaging the ionization coefficients over the electric field
multiplication noise. The results confirm the gain-bandwidth products profile in the InP multiplication region. We show that this
obtained directly from bandwidth measurements. apparent suppression of the noise is consistent with a
model in which the average number of collisions per pri-
mary carrier is small. We have also determined the gain-
INTRODUCTION bandwidth product from the frequency response of the
T HE PROGRESSION of lightwave systems toward
multigigabit-per-second transmission rates has stim-
ulated research on wide-bandwidth avalanche photo-
multiplication noise [ 191-[21]. We find that this tech-
nique yields gain-bandwidth products close to those ob-
tained directly from bandwidth measurements.
diodes (APD’s) with high gain-bandwidth products [ 11.
To date, the structure that has demonstrated the best fre- DISCUSSION
quency response is the separate absorption, grading, and The design of the CBE system and the growth kinetics
multiplication (SAGM) APD [2]-[8]. In its most common have been described in [22]. For the SAGM-APD wafers,
configuration the SAGM-APD consists of an InP multi- arsine (AsH3, 100 percent) and phosphine (PH3, 100 per-
plication region and an Ino 53G%,47A~ absorbing layer sep- cent) were used as the arsenic and phosphorus sources.
arated by a transition region comprised of one or more An arsine and phosphine cracker with a reduced input
,
thin intermediate-bandgap Inl - Ga, As, PI - layers. It pressure in the range 40 to 200 torr was maintained on the
has recently been demonstrated with planar and mesa high-pressure side of the electronic mass flow controller;
structures grown by vapor-phase epitaxy [9], liquid-phase the cracking temperature was = 920°C. The gallium and
epitaxy [lo]-[12], and chemical beam epitaxy [l], [13] indium sources were triethylgallium (TEGa) and tn-
that the gain-bandwidth product can be improved by using methylindium (TMIn) maintained at 30°C and 37”C, re-
thin multiplication regions to minimize the effects of car- spectively. Hydrogen was used as the carrier gas for both
rier feedback. For InP /InGaAsP /InGaAs SAGM-APD’s TEGa and TMIn. The TEGa and TMIn flows were com-
this must be accompanied by an increase in the carrier bined to form a single beam impinging onto the substrate.
concentration in the multiplication region. Otherwise, the At the growth temperature of 550°C the growth rates of
electric field in the narrow-bandgap absorbing layer will InP, InGaAsP, and InGaAs were 2.0, 3.65, and 3.6
be too high ( 2 2 X lo5 V/cm) [14] and tunneling will pm/h, respectively. Elemental Be and Sn evaporated
lead to excessive dark current. A potential drawback to from conventional MBE effusion cells were used as the p-
this approach is that the maximum electric field at the and n-type dopants.
p-njunction is very high. Bulman [15] has shown that the A schematic cross section of the back-illuminated mesa-
structure SAGM-APD is shown in Fig. l . The structure
Manuscript received March 29, 1988; revised June 14, 1988. consists of 7 epitaxial layers grown on a ( 100 )-oriented,
J. C. Campbell, W. T . Tsang, G. J. Qua, and B. C. Johnson are with Zn-doped ( N , = 1 x 10i9/cm3) InP substrate. The first
AT&T Bell Laboratories, Holmdel, NJ 07733.
S. Chandrasekhar, was with AT&T Bell Laboratories, on leave from the
layer grown was a 0.25-pm-thick, Be-doped ( N o = 1 x
Tata Institute of Fundamental Research, Bombay, India. 1018/cm3)InP buffer layer. This was followed by the Sn-
IEEE Log Number 8823577. doped ( N d = 5.5 X 1016/cm3) InP multiplication region.

0733-8724/89/0300-0473$01.OO O 1989 IEEE

414 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 7, NO. 3, MARCH 1989

\
n+:InP
Au-Sn CONTACT
\
and
r PXm 1

3 x n : InGaAsP n-: I n G o A s L J+ J

with x, being the width of the multiplication region. In

addition, Naqvi [24] has shown that at high frequencies
the avalanche buildup time must be included. The result-

1 hu
Au-In-Zn
CONTACT

Fig. 1 . Schematic cross section of InP/InGaAsP/InGaAs back-illumi-

ing representation for the noise power spectral density of
an avalanche photodiode can be written as

(Papd =
pp-i-nM;F( M O )
(6)
nated mesa-structure-APD. { 1 + [U(& - 1)/2rGBI2)
This layer was approximately 0.55 pm thick, The transi- where GB is the gain-bandwidth product.
tion region consisted of three thin ( 5700 A ) InGaAsP Fig. 2 shows a block diagram of the apparatus used to
( E , = 1.13eV,0.95eV,and0.80eV)layerswithcamer measure the multiplication noise. The optical source was
concentrations of approximately 2 x 1016/cm3. The a light emitting diode (LED) that had an emission peak
Ino,53Gao.47As absorption layer was unintentionally doped near 1.3 pm. An LED was used instead of an injection
n-type ( N d = 2 x 10i5/cm3) and its thickness was 1.8 laser because the noise of the laser can be comparable to
pm. The wafers were capped with an n+ :InP (Nd 2 2 X that of the APD's. The incident light was coupled through
i~'~/cm contact
~ ) layer. a spatial filter with a 10-pm-diameter pinhole to insure
After crystal growth, p ( Au-In-Zn)- and n ( Au-Sn)-type that the illuminated area was much smaller than that of
contacts were deposited by pulse electroplating. The the APD's, thus eliminating edge effects as well as re-
mesas were etched in a dilute solution of Br-methanol. ducing the effect of nonuniformities in the gain profile
The mesas were typically 40 to 50 pm in diameter. A win- [25]. The noise was measured at frequencies from 30 MHz
dow was etched in the p-type contact to permit illumina- to 1.4 GHz with an HP8970A noise power meter. The
tion through the substrate. The chips were mounted on a instrument was first calibrated with a commercial noise
modified Wiltron connector for evaluation. source to eliminate the noise contribution of the amplifier.
It is well known that the shot-noise power density of a Then the shot noise of a p-i-n was measured as a function
p-i-n photodiode is given by [23] of the photocurrent. This provided the proportionality
factors in (1). Following this calibration procedure, which
(Pp-i-n = 2@phR(u) (1) had to be performed at each measurement frequency, the
where 4 is the electronic charge, o is the radian fre- noise power of the APD was measured under illumination
quency, Zph is the primary photocurrent, and R(w) is the and in the dark as a function of bias voltage. The readings
impedence of the p-i-n and the measurement circuit. If the of the noise power meter were in decibels relative to the
multiplication process were noiseless, the low frequency thermal noise of a 50-Qresistor in parallel with the pho-
shot-noise power of an APD would be that of a p-i-n mul- todetector. As described below the gain was inferred from
tiplied by Mi,the square of the dc avalanche gain. The the photocurrent at each bias level.
statistical nature of the multiplication process, however, The multiplication noise of an APD is usually presented
contributes an additional component to the noise power. as a plot of the excess noise factor F ( M o ) versus the dc
McIntyre [ 171, [181 has demonstrated that this multipli- avalanche gain MO. It is clear from (6) that F ( M o ) can
cation noise can be incorporated into (1) as an excess noise easily be obtained from the measured noise power pro-
factor F ( M o ) . For pure hole injection in the absence of vided MO is known. There are two phenomena which af-
absorption in the multiplication layer the excess noise fac- fect the determination of the gain. 1) The responsivity is
tor is the composite of the quantum efficiency and the gain and
F ( M 0 ) = keffM0 + [2 - 1/M0](1 - keff)* (2) it can be difficult to separate them. 2) The primary pho-
tocurrent increases slightly with bias. Frequently, the gain
The effective ratio of the ionization coefficients keRcan of an SAGM APD can be decoupled from the quantum
written as [ 181 efficiency by measuring the responsivity at punchthrough,
the voltage at which the depletion region first reaches the
(3) edge of the absorbing layer. If the gain at punchthrough
is unity, then the photocurrent will be relatively indepen-
dent of bias over several volts and the gain at higher bias
voltages can be determined from the ratio of the photo-
current to its value at punchthrough. For the present wide-
bandwidth APD's the gain at punchthrough is slightly
greater than 1. We have, however, been able to estimate
(4) the gain at punchthrough by comparing the measured shot
 415
CAMPBELL et al.: MULTIPLICATION NOISE OF WIDE-BANDWIDTH PHOTODIODES

NOISE
430 I I I I I
1

!“k-
a loo

Fig. 2 . Noise measurement apparatus.

noise at punchthrough with that of a p-i-n photodiode. At

low frequencies the higher noise of the APD is simply due
to the factor M i F ( M o ) . For MO < 2, F ( M o ) = MO and
i:
85
80 1
0.00 0.40 0.80
W (pm)
1.20 1.60 2.00

is relatively independent of the ratio of the ionization coef- Fig. 3. Primary photocurrent versus the fractional portion of the absorbing
ficients. This leads to an estimate of MO = 1.21 0.01 layer that is depleted.
for the gain at punchthrough.
The primary photocurrent increases with bias because 60
the collection efficiency of the photogenerated carriers
improves as the depletion region extends across the ab- 40

sorbing layer [26]. Essentially all of the carriers generated

in the depleted portion of the absorbing layer will even-
tually be collected. The carriers generated in the unde- 20

pleted portion of the absorbing layer, on the other hand,

must diffuse across the edge of the depletion region and
40
some of these carriers recombine before reaching the de-
F e
pletion region. As the bias voltage increases and the width
of the undepleted portion of the absorbing layer shrinks, 6

the probability that the carriers will be collected im- 4

proves. The photocurrent can be written in terms of the
drift and diffusion components as follows:
2

I I I l l I
1 2 4 6 810 20 40 60
M
Fig. 4. Excess noise factor F versus the dc gain M O .The dashed and heavy
where 11 is the quantum efficiency, Po is the incident solid lines are the calculated excess noise factors using continuum and
power, hu is the photon energy, CY is the absorption coef- discrete theories, respectively.
ficient, Lp is the hole diffusion length, and W/x,is the
fractional portion of the absorbing region that is depleted. approximately 0.6 near MO = 30. These results are con-
Fig. 3 shows the measured variation of I with W/x,for sistent with previously published noise measurements of
ph
SAGM APD’s with very thin multiplication regions SAGM APD’s having lower electric fields in the multi-
( < 0.1 pm). The values of W/xawere obtained from ca- plication region [21], [25], [27]-[31]. Using calculated
pacitance-voltage profiling. Calculations show, and mea- field profiles and the ionization coefficient data of Bulman
surements of the shot noise confirm, that the gain of these [15], (3) predicts that 0.57 Ikeff 5 0.60 for 5 IMO I
devices is unity well beyond the point where the absorb- 30 (dashed line in Fig. 4). The measured multiplication
ing layer is fully depleted. The solid line is a fit to (7). noise approaches the k = 0.6 line at the highest gain val-
From these results we have determined that for the wide- ues but is significantly lower at lower gains. There are
bandwidth InP/InGaAsP/InGaAs SAGM APD’s the pri- three possible explanations for this discrepancy. 1) If the
mary photocurrent increases approximately 13 percent p-n junction is not abrupt, the maximum electric field will
from punchthrough to full depletion of the absorbing re- be lower than we have calculated it to be. This would
gion. result in lower values of keff. This appears unlikely, how-
The excess noise data shown in Fig. 4 has been ob- ever, since capacitance-voltage measurements show that
tained by including the gain at punchthrough and the bias the voltage is proportional to the inverse square of the
dependence of the primary photocurrent. The different capacitance which is characteristic of abrupt p-n junc-
types of data points are different devices. We see that the tions. 2) It is possible that inaccuracies in the ionization
k , factor increases from 0.4 at low gain ( M O 5 10) to coefficient data of Bulman et al. [15] contribute to an
476 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 7, NO. 3, MARCH 1989

overestimation of the noise. On the other hand, validation

of their data is provided by the close agreement between
calculated breakdown voltages and experimental values.
3) It is most probable that the lower-than-expected noise
is attributable to the very narrow width of the multipli-
cation region. Since the electric field profile is approxi-
mately triangular and drops to < 2 X lo5 V/cm at the
edge of the multiplication region, the effective avalanch-
ing length is even less than the 0.5-pm width of the InP
multiplication layer. Conventional APD noise theories
[17],[18],[32]-[34]assume that the avalanching region
is very long compared to 1/p (or 1/a). This is equiva-
lent to assuming that the number of ionizing collisions per
primary carrier is large. These assumptions may not be
valid for the narrow multiplication regions of these
APD’s. Van Vliet et al. [35] have shown that the multi-
plication noise is suppressed if the number of collisions
per primary carrier is small. In their formulation, the gain
for n ionizations per primary carrier is given by the
expression
3
- 4001

FREQUENCY ( G H z )
10

(b)
where k = X/p with p and X .being the a priori probabil- Fig. 5. (a) qaN/qp.,.”M: versus the dc gain MO measured at 30 MHz and
ities for ionization by a hole and an electron, respectively. 1.4 GHz. The solid line has been calculated using a gain-bandwidth
The k factor corresponds roughly with keRin (3).The ex- product of 60 GHz. (b) Frequency response at MO = 44.
cess noise factor can be expressed as
o2 predicted by the conventional continuum theory. How-
F(Mn) = 1 - + (9) ever, at lower gains the noise is suppressed. This is prob-
(Mn> 2 ably an artifact of the very thin multiplication layers which
where u2 is the variance of the distribution of the gain have been used to increase the gain-bandwidth product.
u2 = ( M : ) - ( M n ) 2 . From the frequency response of the noise power we have
The dark solid line in Fig. 4 is a plot of (9) for k = 0.65 deduced a gain-bandwidth product of 60 GHz which is
and p = 0.4.The close fit to the experimental data is consistent with the value of 57 GHz obtained directly from
probably more of an indication that the conventional con- bandwidth measurements.
tinuum theory should be modified for APD’s of this type ACKNOWLEDGMENT
than a confirmation of this specific model.
The frequency dependence of the multiplication noise The authors thank M. A. Pollack for his continued sup-
is illustrated in Fig. 5(a), where cpapd/pp-i-nM~ is plotted port of this work and R. J. McIntyre and P. P. Webb for
as a function of the avalanche gain at 30 MHz and 1.4 many useful discussions about noise in APD’s. They are
GHz. For MO > 10 the effective excess noise factors de- grateful to J. N. Hollenhorst, D. T. Ekholm, and D. J.
rived from the l .4GHz data are less than those obtained Muehlner for help with the noise measurement apparatus
at 30 MHz. This is due to the finite time required for the and for supplying reference APD’s. They also thank J. C.
avalanche process to build up and is the origin of the gain- Centanni for assistance with device fabrication and T. H.
bandwidth term in (4).The solid line in Fig. 5 was cal- Chiu and J. E. Cunningham for help with crystal growth.
culated using a gain-bandwidth product of 60 GHz. The REFERENCES
gain-bandwidth product of this APD has also been deter-
B. L. Kasper and J . C. Campbell, “Multigigabit-per-second ava-
mined directly from bandwidth measurements (Fig. 5(b)). lanche photodiode lightwave receivers,” J . Lightwave Techno[., vol.
For MO = 44 the bandwidth is 1.3 GHz which gives a LT-5,pp. 1351-1364, 1987.
gain-bandwidth product of 57 GHz in good agreement S. R. Forrest, 0. K. Kim, and R. G . Smith, “Optical response time
with the value determined from the frequency dependence of Ino.s,G~b.47Asavalanche photodiodes,” Appl. Phys. L e r r . , vol. 41,
pp. 95-98, 1982.
of the shot noise. J . C. Campbell, A. G. Dentai, W. S. Holden, and B . L. Kasper,
“High-performance avalanche photodiodes with separate absorption,
CONCLUSION grading, and multiplication regions,” Electron. Lett., vol. 18, pp.
In conclusion, we have measured the multiplication of 818-820, 1983.
Y . Matsushima, K. Sakai, and Y. Noda, “New type InGaAs/InP
wide-bandwidth InP/InGaAsP/InGaAs SAGM APD’s. heterostmcture avalanche photodiode with buffer layer,” IEEE Elec-
At high gains the excess noise factors approach values tron. Dev. Let., vol. EDL-2, pp. 179-181, 1981.
 471
CAMPBELL er al.: MULTIPLICATION NOISE OF WIDE-BANDWIDTH PHOTODIODES

[5j Y. Matsushima, A. Akiba, K. Sakai, K. Kushiro, Y. Noda, and K. alanche photodiode fabrication by vapor phase epitaxy,” Electron.
Utaka, “High-speed planar-structure InP/InGaAsP/InGaAs het- Lett., vol. 19, pp. 543-545, 1983.
erostructure avalanche photodiode with InGaAsP buffer layers,” [29] T. Shirai, T . Mikawa, and T . Kaneda, “InGaAs photodiodes for 1
Electron. Lett., vol. 18, pp. 945-946, 1982. pm wavelength region,” Electron. Lett., vol. 19, pp. 534-536, 1983.
[6] Y. Sugimoto, T. Torikai, K. Makita, H. Ishihara, K. Minemura, and [30] T. Shirai, S. Yamasaki, F. Osaka, K. Nakajima, and T . Kaneda,
K. Taguchi, “High-speed planar-structure InP/InGaAsP/InGaAs “Multiplication noise in planar InP/InGaAsP heterostructure ava-
avalanche photodiode grown by VPE,” Electron. Lett., vol. 20, pp. lanche photodiodes,” Appl. Phys. L e f t . , vol. 40, pp. 532-533, 1982.
653-654, 1984. [31] B . L. Kasper, J . C. Campbell, and A. G. Dentai, “Measurements of
[7] K. Yasuda, T . Mikawa, Y. Kishi, and T. Kaneda, “Multiplication- the statistics of excess noise in separate absorption grading, and mul-
dependent frequency responses of InP/InGaAs avalanche photo- tiplication (SAGM) avalanche photodiodes,” Electron. Lett., vol. 20,
diode,” Electron. Lett., vol. 20, pp. 373-374, 1984. pp. 796-798, 1984.
[8] R. Trommer and L. Hoffman, “Time-response of InGaAs/InP ava- [32] A. S. Tager, “Current fluctuations in a semiconductor (dielectric)
lanche photodiodes,” in Proc. 11th Int. Symp. GaAs and Related under the conditions of impact ionization and avalanche breakdown, ”
Compounds (Biaritz, France), 1984. Sov. Phys.-Solid State, vol. 6 , pp. 1919-1925, 1965.
[9] Y . Sugimoto, T. Torikai, K. Makita, H. Ishihara, and K. Taguchi, 1331 S . D. Personick, “New results on avalanche multiplication statistics
“Gain-bandwidth improvement in planar InP/InGaAsP/InGaAs av- with applications to optical detection,” Bell Sysr. Tech. J . , vol. 50,
alanche photodiodes,” in Proc. 5th Int. Conf. Integrated Opt. and pp. 167-190, 1971.
Opt. Fiber Commun. (Venice, Italy), 1985. [34] S. D. Personick, “Statistics of a general class of avalanche detectors
[ I O ] W. S. Holden, J. C. Campbell, J . F. Ferguson, A. G. Dentai, and with applications to optical communication,” Bell. Sysr. Tech. J . ,
Y, K. Jhee, “Improved frequency response of InP/InGaAsP/InGaAs vol. 50, 3075-3094, 1971.
avalanche photodiodes with separate absorption, grading, and multi- [35] K. M. Van Vliet, A. Friedman and L. M. Rucker, “Theory of carrier
plication regions,” Electron. Lett., vol. 21, pp. 886-887, 1985. multiplication and noise in avalanche devices-Part 11: Two-carrier
[ I l l J . C. Campbell, W. S . Holden, J. F. Ferguson, A. G. Dentai, and processes,” IEEE Trans. Electron D e v . , vol. Ed-26, pp. 752-764,
Y. K. Jhee, “A high-speed InP/InGaAsP/InGaAs avalanche pho- 1979.
todiode exhibiting a gain-bandwidth product of 60 GHz,” in Proc.
5th Int. Con$ Integrated Opt. and Opt. Fiber Commun. (Venice, It-
aly), 1985 (postdeadline paps.).
M. Kobayashi, H. Machida, T. Shirai, Y. Kishi, N. Takagi, and T.
Kaneda, “Optimized GaInAs avalanche photodiode with low noise
*
and large gain-bandwidth product,” in Proc. 6th Conf. Integrated
Opt. and Opt. Fiber Commun. (Reno, NV), 1987.
5 . C. Campbell, W. T. Tsang, G . J . Qua, and J . E. Bowers,
“InP/InGaAsP/InGaAs avalanche photodiodes with 70 GHz gain- Joe C. Campbell (S’73-M’74) was born in Gor-
bandwidth product,” Appl. Phys. Lett., vol. 51, pp. 1454-1456, man, TX, on January 11, 1947. He received the
1987. B.S. degree in physics from the University of
[I41 S . R. Forrest, R. G . Smith, and 0. K. Kim, “Performance of Texas at Austin in 1969, and the M.S. and Ph.D.
In, J i a , ,,As/InP avalanche photodiodes,” IEEE J . Quanrum Elec- degree in physics from the University of Illinois
tron., vol. QE-18, pp. 2040-2048, 1982. at Urbana-Champaign in 1971 and 1973, respec-
1151 G . E. Bulman, “The experimental determination of impact ionization tively.
coefficients in gallium arsenide and indium phosphide,” Ph. D. dis- From 1974 to 1976 he was employed by Texas
sertation, Univ. of Illinois, Urbana, IL, 1983. Instruments where he worked on integrated optics
[ 161 E. B. Emmons, “Avalanche-photodiode frequency response,” J . in GaAs/AlGaAs. Since 1976 he has been at the
Appl. Phys., vol. 38, pp. 3705-3714, 1967. Crawford Hill Laboratory of Bell Laboratories in
[ 171 R. J. McIntyre, “Multiplication noise in uniform avalanche diodes,” Holmdel, NJ, where he has worked on electrooptic devices for fiber optics
IEEE Trans. Electron D e v . , vol. ED-13, pp. 164-168, 1966. applications.
1181 R. J . McIntyre, “The distribution of gains in uniformly multiplying Dr. Campbell is a member of Sigma Xi, the Amencan Physical Society,
avalanche photodiodes: Theory,” IEEE Trans. Electron D e v . , vol. and the Optical Society.
ED-19, pp. 703-713, 1972.
1191 T. Kaneda and H. Takanashi, “Avalanche build-up time of the ger-
manium avalanche photodiode,” Japan J . Appl. Phys., vol. 12, pp.
1091-1092, 1973.
1201 T . Kaneda and H. Takanashi, “Avalanche build-up time of silicon *
avalanche photodiodes,” Appl. Phys. Lett., vol. 26, pp. 642-644,
1975.
[21] Y. K. Jhee, J . C. Campbell, J . F. Ferguson, A. G . Dentai, and W.
S. Holden, “Avalanche buildup time of InP/InGaAsP/InGaAs av-
alanche photodiodes with separate absorption, grading, and multipli- S. Chandrasekhar was born in Tiruchirapalli,
cation regions,” IEEE 1. Quantum Electron., vol. QE-22, pp. 753- Tamil Nadu, India, on January 26, 1952. He re-
755, 1986. ceived the B.Sc., M.Sc., and Ph.D. degrees in
[221 W. T . Tsang, “Chemical beam epitaxy of InGaAs,” J . Appl. Phys., physics from the University of Bombay, Bombay,
vol. 58, pp. 1415-1418, 1985. India, in 1973, and 1985, respectively. He was a
1231 B. M. Oliver, “Thermal and quantum noise,” Proc. IEEE, vol. 53, Government of India National Science Talent
pp.436-454, 1965. Scholar from 1971 to 1979. His Ph.D. disserta-
1241 I . M. Naqvi, “Effects of time dependence of multiplication process tion was on laser recrystallized polycrystalline sil-
on avalanche noise,” Solid State Electron., vol. 16, pp. 19-28, 1973. icon on insulators.
1251 Y. K. Jhee, J. C. Campbell, W. S. Holden, A. G. Dentai, and J. K . He ioined the Tata Institute of Fundamental
Plourde, “The effect of nonuniform gain on the multiplication noise Research, Bombay, India, as a Research Scholar
of InP/InGaAsP/InGaAs avalanche photodiodes with separate ab- in 1975 and was subsequently appointed a Permanent Member in 1979. He
sorption grading, and multiplication regions,’’ IEEE J . Quanmm has been engaged in research on Si-SiO, interface studies, CMOS inte-
Electron., QE-21, pp. 1858-1861, 1985. grated circuits, ion implantation, and charge coupled devices. In 1986, he
[26] C. W. Kao and C. R. Crowell, “Impact ionization by electrons and joined AT&T Bell Laboratories, Crawford Hill Laboratories, Holmdel, NJ.
holes in InP,” Solid State Electron., vol. 23, pp. 881-891, 1980. as a post-doctoral Member of the Technical Staff, on study leave from
1271 S. R. Forrest, G. F. Williams, 0 . K. Kim, and R. G. Smith, “Excess- T.I.F.R., Bombay. Since then, he has been working on 111-V compound
noise and receiver sensitivity measurements of In, &ao ,,As/InP semiconductor devices for optoelectronic applications, primarily wave-
avalanche photodiodes,” Electron. Lett., vol. 17, pp. 917-919, 1981. guides, photodetectors, heterojunction phototransistors, and bipolar tran-
1281 H. Ando, Y. Yamauchi, and N.Susa, “High-speed InP/InCaAs av- sistors.
47 8 JOURNAL OF LIGHTWAVE TECHNOLOGY. VOL. 7. NO. 3. MARCH 1989

.Won T. Tsang was born on October 24, 1950. He G. J. Qua, a citizen of Singapore, was born in
received the Ph.D. degree in electrical engineer- Malaysia on January 27, 1957. He received the
ing from the University of California, Berkeley, A.O.S. degree in electronic,circuits and systems
in 1976. from the Technical Career Institute, New York,
Since joining AT&T Bell Laboratories, Mur- NY, in 1980 and the B.S. degree in electrical en-
ray Hill, in the same year, his research areas have gineering technology from Fairleigh Dickinson
included semiconductor injection lasers, photo- University, Teaneck, NJ, in 1986.
detectors, molecular beam epitaxy, and chemical In September 1980 he joined the Photonic De-
beam epitaxy of 111-V compound semiconductors. vices Research Department at the Crawford Hill
His research in these areas resulted in the inven- Laboratory of AT&T Bell Laboratories in Holm-
tion of many important lasers among which are del, NJ. He is currently a Member of the Tech-
the stnp buried hetemstructure laser, the graded-index separate confine- nical Staff in the Lightwave Photodetectors Group, where he works on elec-
ment heterostructure laser, the electronically tunable single-frequency trooptic devices for lightwave applications.
cleaved-coupled cavity laser, and the demonstration of the first superlattice
avalanche photodetector. He has made significant contnbutions to the de-
velopment of molecular beam epitaxy and chemical beam epitaxy. In 1984, *
he was appointed head of the Electrophotonics Research Department,
Holmdel. In 1987, he transferred to become head of the Semiconductor
Electronics Research Department, Murray Hill. Bart C. Johnson was born in Minneapolis, MN,
Dr. Tsang has published over 310 papers and holds 30 patents. He is in 1956. He studied electrical engineering as an
the editor of five volumes of books on lightwave communications technol- undergraduate at the University of Minnesota, and
ogy in the treatise of semiconductors and semimetals. In 1971, he was the also as a graduate student at M.I.T. where he re-
recipient of the Phi Kappa Phi Award from Georgia Institute of Technol- ceived the Ph.D. degree in 1986. His thesis re-
ogy. In 1981, he received the Honorable Mention of Outstanding Young search, performed at M.I.T. Lincoln Laboratory,
Electncal Engineer from Eta Kappa Nu, the National Electncal Engineer was on picosecond studies of diode laser modu-
Honorary Society, and in 1982 the Adolph Lomb Medel from the Optical lation.
Society of Amenca. He is a member of Phi Eta Sigma, Tau Beta Pi, Eta Since then, he has been with the photonic de-
Kappa Nu, Phi Kappa Phi, Amencan Physical Society, Electrochemical vice research group at AT&T Bell Laboratories in
Society, Amencan Association for the Advancement of Science, the New Holmdel, NJ, where he has been involved in the
York Academy of Science, and a fellow of the Optical Society of America fabrication and characterization of semiconductor diode lasers and photo-
and of the American Physical Society. detectors.