Gain saturation and carrier distribution effects in molecular beam epitaxy grown Ga As
Sb ∕ Ga As quantum well lasers
S.-Q. Yu, X. Jin, S. R. Johnson, and Y.-H. Zhang
Citation: Journal of Vacuum Science & Technology B 24, 1617 (2006); doi: 10.1116/1.2192534
View online: http://dx.doi.org/10.1116/1.2192534
View Table of Contents: http://scitation.aip.org/content/avs/journal/jvstb/24/3?ver=pdfcov
Published by the AVS: Science & Technology of Materials, Interfaces, and Processing
Articles you may be interested in
GaAs-based room-temperature continuous-wave 1.59 μ m GaInNAsSb single-quantum-well laser diode grown
by molecular-beam epitaxy
Appl. Phys. Lett. 87, 231121 (2005); 10.1063/1.2140614
Thermal excitation effects of photoluminescence of annealed Ga In N As ∕ Ga As quantum-well laser structures
grown by plasma-assisted molecular-beam epitaxy
J. Vac. Sci. Technol. B 23, 1434 (2005); 10.1116/1.1935533
Use of transmission electron microscopy in the characterization of GaInNAs(Sb) quantum well structures grown
by molecular beam epitaxy
J. Vac. Sci. Technol. B 22, 1588 (2004); 10.1116/1.1650853
InGaAsNSb/GaAs quantum wells for 1.55 μm lasers grown by molecular-beam epitaxy
Appl. Phys. Lett. 78, 4068 (2001); 10.1063/1.1379787
High performance 1.3 μm InGaAsN:Sb/GaAs quantum well lasers grown by molecular beam epitaxy
J. Vac. Sci. Technol. B 18, 1484 (2000); 10.1116/1.591409

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 209.147.144.20 On: Sun, 08 Feb 2015 00:35:31

Gain saturation and carrier distribution effects in molecular beam epitaxy
grown GaAsSb/ GaAs quantum well lasers
S.-Q. Yu, X. Jin, S. R. Johnson,a兲 and Y.-H. Zhang
Center for Solid State Electronics Research and Department of Electrical Engineering,
Arizona State University, Tempe, Arizona 85287-6206, Tempe, Arizona 85287-6206

共Received 14 September 2005; accepted 5 December 2005; published 31 May 2006兲
GaAsSb/ GaAs quantum well 共QW兲 lasers grown by solid source molecular beam epitaxy are
fabricated into ridge lasers and tested. These devices have a lasing wavelength around 1.2 ␮m that
is substantially blueshifted relative to the electroluminescence peak. The magnitude of the blueshift
increases as the cavity length is shortened, indicating that the blueshift increases with injection level.
This blueshift is attributed to material gain saturation and band filling effects. The internal quantum
efficiency is ⬃75%, the transparency current density is ⬃120 A / cm2, and the threshold
characteristic temperature is ⬃60 K, all typical for GaAsSb/ GaAs based edge emitting lasers. The
extracted gain constant is ⬃800 cm−1 for single QW active regions and approximately half that
amount for double QWs. This discrepancy is attributed to nonuniform carrier distribution in double
QW structures. © 2006 American Vacuum Society. 关DOI: 10.1116/1.2192534兴

I. INTRODUCTION
Vertical-cavity surface-emitting lasers 共VCSELs兲 operating at 1.3 ␮m are of great interest for low-cost data transmission applications such as fiber to home, local area networks, and free-space optical interconnects. GaAs is the
preferred substrate for 1.3 ␮m VCSELs because it is compatible with the growth of near lattice-matched GaAs/
AlGaAs distributed Bragg reflectors 共DBRs兲 which have superior optical and thermal properties compared to other III-V
DBRs. Furthermore, the fabrication of GaAs based 1.3 ␮m
VCSELs can take full advantage of the current industry standard 850 nm VCSEL fabrication technology, which is attractive from a manufacturing point of view. GaAsSb/ GaAs
quantum wells 共QWs兲 have been shown to be one of the
most suitable candidates for 1.3 ␮m active regions on the
GaAs substrate.1–11
To further improve VCSEL performance, it is necessary to
understand the limitations of this material system and the
interaction of the various parameters that impact overall device performance. For example, the gain of this material system is restricted by a less than ideal electron-hole wave function overlap caused by a combination of strongly confined
holes and weakly confined electrons, which is a result of a
nearly flat conduction band alignment between GaAs and
GaAsSb.12 When multiple QWs are used to increase modal
gain, this strong hole confinement can influence the uniformity of the carrier distribution in each well. Furthermore,
GaAsSb/ GaAs QWs are highly strained 共⬃2.7% 兲 at the
composition necessary for 1.3 ␮m emission. This not only
limits the maximum QW number that can be grown without
misfit dislocations, but also results in strain-driven in-plane
composition fluctuations, which can reduce quantum efficiency and increase inhomogeneous linewidth broadening.
The quality of highly strained GaAsSb layers can be ima兲

Author to whom correspondence should be addressed; electronic mail:
shane.johnson@asu.edu

1617

J. Vac. Sci. Technol. B 24„3…, May/Jun 2006

proved by adding GaAsP strain compensation layers2,13–15
near the active region. To date the best performing
GaAsSb/ GaAs based edge-emitting lasers 共EELs兲 and
VCSELs have been demonstrated10,11 using GaAsP strain
compensating layers.
In this article, three typical EELs grown by molecular
beam epitaxy 共MBE兲 are studied; one with a single
GaAsSb/ GaAs QW, one with a strain compensated single
GaAsSb/ GaAs/ GaAsP QW, and one with two strain compensated GaAsSb/ GaAs/ GaAsP QWs. The internal quantum
efficiency, the internal loss, the transparency current density,
the material gain constant, and the characteristic temperatures of the threshold current density and the slope efficiency
are measured and compared for these three devices.
II. EXPERIMENTAL RESULTS
The EELs studied in this work were grown on 共100兲 n+
GaAs substrates using solid source MBE. Device A has an
active region containing one GaAs/ GaAs0.7Sb0.3 / GaAs
共5 nm/ 7 nm/ 5 nm兲 QW grown at 490 ° C, device B has an
active region containing one GaAs0.9P0.1 / GaAs/ GaAs0.7
Sb0.3 / GaAs/ GaAs0.9P0.1
共8 nm/ 3 nm/ 7 nm/ 3 nm/ 8 nm兲
QW grown at 500 ° C, and device C has an active region
containing two QWs of the same structure as device B also
grown at 500 ° C. The nominal Sb concentration is 30%, a
value estimated from photoluminescence 共PL兲 measurements
and modeling. The active region in device A is sandwiched
between two 30 nm Al0.25Ga0.75As layers, followed by two
150 nm thick AlGaAs layers with a linearly graded Al mole
fraction from 25% to 65% to form a graded-index 共GRIN兲
waveguide, followed by Si-doped 共2 ⫻ 1018 cm−3兲, 1.8 ␮m
thick, n-type Al0.65Ga0.35As cladding and 500 nm thick GaAs
buffer layers on the substrate side and Be-doped
共2 ⫻ 1018 cm−3兲, 1.8 ␮m thick, p-type Al0.65Ga0.35As cladding and 100 nm thick GaAs contact layers on the top side.
The doping concentration is decreased from 2 ⫻ 1018 to
1 ⫻ 1017 in both the p and n GRIN layers and is increased to

1071-1023/2006/24„3…/1617/5/$23.00

©2006 American Vacuum Society

1617

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 209.147.144.20 On: Sun, 08 Feb 2015 00:35:31

1618

Yu et al.: Gain saturation and carrier distribution effects

1618
TABLE I. Internal quantum efficiency ␩i internal loss ␣i, transparency current
density Jtr, and gain constant G0.
Device A

Device B

Device C

␩i 共%兲
␣i 共cm−1兲
Jtr 共A / cm2兲
G0 共cm−1兲

73
13
127
787

76
12
119
801

78
8
131
421

冋

␣i · L

1

␩e

FIG. 1. Inverse external quantum efficiency vs cavity length for a single QW
laser without strain compensation 共solid squares兲, a single QW laser with
strain compensation 共solid circles兲, and a double QW laser with strain compensation 共solid triangles兲. The solid line is a linear fit to the data, where
internal quantum efficiency and internal loss are fitting parameters.

2 ⫻ 1019 in the p contact layer. Devices B and C have the
same structure and doping profile as device A, except that the
QW in device B 共device C兲 is sandwiched between two
78 nm 共65 nm兲 thick Al0.25Ga0.75As layers. The active region
growth temperatures were optimized to give the strong PL
and minimal inhomogeneous linewidth broadening. Further
details of the growth of this material system can be found in
previous work.10–12,14,15
The devices were fabricated using photolithography and
inductance coupled plasma 共ICP兲 dry etching to define stripe
ridges, ranging from 4 to 32 ␮m wide. By etching down
through the p-GaAs contact layer and stopping about 0.1 ␮m
above the active region, these ridges provide current confinement as well as waveguiding. A photoresist mask is used to
define the ridges, which also serves as a lift-off mask for the
deposition of an Al2O3 isolation layer. This procedure ensures that a self-aligned contact window is exposed after liftoff. Next wide Ti/ Pt/ Au p-contact stripes are deposited using a second mask, after which the wafers are lapped down
to 100 ␮m and AuGe/ Ni/ Au n-metal contacts are deposited
on the backside of the substrate; this is followed by rapid
thermal annealing for both metal contacts. The wafers were
cleaved to form EEL devices with various cavity lengths.
The as-cleaved devices were mounted junction-side up on a
metal test stage and are driven by a pulsed current source
using a 0.5 ␮s wide pulse and a 0.1% duty cycle. The power
output was measured using a calibrated power meter
equipped with an InGaAs detector and an integration sphere.
Light power output versus current 共L-I兲 measurements
were done on 32 ␮m wide devices with different cavity
lengths. The inverse external quantum efficiency 1 / ␩e versus
the cavity length L is plotted in Fig. 1. The internal quantum
efficiency ␩i and the internal loss ␣i are extracted by fitting
the following relation to the data 共fit given by the solid
curve兲:

=

1

␩i

1+

ln共1/冑R1R2兲

册

共1兲

,

where R1 and R2 are the mirror reflectivity, both equal to
0.33 for uncoated mirrors. All device cavity lengths are
longer than 450 ␮m to minimize errors caused by gain saturation. The extracted internal quantum efficiency and loss for
devices A, B, and C are 73% and 13 cm−1, 76% and 12 cm−1,
and 78% and 8 cm−1, respectively. A summary of the extracted device characteristic is given in Table I. A substantial
amount of the internal loss 共⬃6 cm−1兲 is attributed to free
carrier absorption in the relatively heavily doped cladding
layer. The variation in the remaining portion of the internal
loss is attributed to processing variations. The internal quantum efficiency of these devices is higher than previous reported values5,6,9,13 for the same material system. We believe
this is due to improved material quality through optimization
of material growth and to slightly lower Sb concentrations.
The device temperature characteristics are determined
from temperature dependent pulsed L-I measurements over a
0 – 90 ° C range using a thermoelectric temperature stage.
The threshold current density characteristic temperature T0
and external quantum efficiency characteristic temperature
T1 were extracted by plotting the threshold current density
and the external quantum efficiency versus temperature, respectively. The extracted T0 and T1 for long cavity devices
are listed in Table II. All three devices give similar values
around 60 K for T0, which is in agreement with most published work for GaAsSb QWs.5,6,9,13,16,17 Considering the improved internal efficiency of these devices, we believe that
this value for T0 reflects the intrinsic temperature property of
this material system.
The material gain was evaluated by plotting threshold current density Jth versus the total loss ␣tot in Fig. 2. The material gain constant G0 and the transparency current density Jtr
for devices A, B, and C are extracted by fitting the following
equation to the data 共see solid curve兲:

TABLE II. Cavity length L, lasing threshold characteristic temperature T0,
and slope efficiency characteristic temperature T1

L 共␮m兲
T0 共K兲
T1 共K兲

Device A

Device B

Device C

1563
59
70

1344
66
110

1265
66
82

J. Vac. Sci. Technol. B, Vol. 24, No. 3, May/Jun 2006

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 209.147.144.20 On: Sun, 08 Feb 2015 00:35:31

1619

Yu et al.: Gain saturation and carrier distribution effects

1619

FIG. 2. Current density vs total loss for a single QW laser without strain
compensation 共solid squares兲, a single QW laser with strain compensation
共solid circles兲, and a double QW laser with strain compensation 共solid triangles兲. The solid line is an exponential fit to the data, where the gain
constant and the transparency current density are fitting parameters.

Jth =

冉 冊 冋
nWJtr

␩i

exp

册

1
␣tot − 1 .
共nW⌫WG0兲

共2兲

Here nW is QW number and ⌫W is the optical confinement
factor for each QW, which is determined to be 0.018 for all
three devices using the slab waveguide theory. The extracted
gain constant and transparency current density are 787 cm−1
and 127 A / cm2, 801 cm−1 and 119 A / cm2, and 421 cm−1
and 131 A / cm2 for devices A, B, and C, respectively, which
are lower than the previously reported values.5 These results
are in agreement except for the gain constant of the double
QW laser 共device C兲 which is surprisingly low, since, for
example, devices B and C have similar structures 共other than
number of QWs兲 and were grown and fabricated back to
back using the same process. The number of QWs appears in
two places in Eq. 共2兲; once with Jtr and once with G0. Assuming that device C behaves like an nW = 2 laser in terms of
loss and an nW = 1 laser in terms of gain, the fit results return
a gain constant and transparency density of 842 cm−1 and
131 A / cm2, which agree with the single QW lasers. These
results indicate that the carriers are not uniformly distributed
in the double QW active region. Since the GaAsSb/ GaAs
heterojunction has a very large valence band offset, this nonuniformity is most likely a result of inadequate hole
transport.
The electroluminescence 共EL兲 and lasing spectra of devices A, B, and C were measured for different cavity lengths
at low injection and just above threshold, respectively, starting with the longest cavity. The same device is then cleaved
into several shorter cavity length devices and then measured.
This preparation process excludes any possible wavelength
variations due to wafer nonuniformity. The devices are
driven by a pulsed current source using a 0.5 ␮s wide pulse
and a 1% duty cycle. The signal is collected using a 200 ␮m
core diameter multimode fiber and is analyzed with an Ando

FIG. 3. Electroluminescence spectra under low injection and lasing spectra
at threshold for different cavity length devices, showing a blueshift in lasing
wavelength that increases with injection level. Plot 共a兲 is for a single QW
laser without strain compensation 共device A兲, plot 共b兲 is for a single QW
laser with strain compensation 共device B兲, and plot 共c兲 is for a double QW
laser with strain compensation 共device C兲.

6315A optical spectrum analyzer using a 1 nm resolution
setting. The measured spectra for various device lengths are
shown in Fig. 3; device A in plot 共a兲, device B in plot 共b兲,
and device C in plot 共c兲. There is a blueshift in the lasing
wavelength relative to the EL peak position that increases as
the threshold current density increases for the shorter cavity

JVST B - Microelectronics and Nanometer Structures

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 209.147.144.20 On: Sun, 08 Feb 2015 00:35:31

1620

Yu et al.: Gain saturation and carrier distribution effects

FIG. 4. Threshold lasing blueshift vs threshold current density for a single
QW laser without strain compensation 共solid squares兲, a single QW laser
with strain compensation 共solid circles兲, and a double QW laser with strain
compensation 共solid triangles兲.

devices, indicating gain saturation, which is typical for lower
gain materials.18 The EL peaks exhibit a small redshift as the
laser cavity gets shorter; this is attributed to a drop in active
region carrier density caused by an increase in leakage current density in the shorter devices.
The lasing wavelength blueshift versus the lasing threshold current density for the different cavity length devices is
summarized in Fig. 4. The origin of the blueshift is likely
due to gain saturation and band filling, which are exacerbated by a high density of localized tail states caused by
in-plane composition fluctuations in the GaAsSb QWs. The
low injection luminescence linewidth for all three devices is
about 55 meV, and is in the middle to low range of the
typical 40– 80 meV values observed, which vary depending
on the extent of strain-driven Sb segregation.14 Here Sb segregation was minimized by using strain compensation in devices B and C and by growing the active region 10 ° C lower
in device A. The growth of these highly strained QWs is a
trade off between excess inhomogeneous linewidth broadening and diminished internal quantum efficiency, where segregation can be reduced using lower growth temperatures,
with the penalty of increased material defects. The double
QW device unexpectedly shows a similar sized blueshift
trend to that of the single QW devices. Under the same injection current density, the injection per QW, and consequently any band filling related blueshift, is expected to be
smaller in a double QW device. Again these results point to
only one QW contributing to lasing in the double QW device, likely due to a nonuniform carrier density in the active
region.
III. DISCUSSION
The examination of gain saturation and gain blueshift under high injection is important for VCSEL design, since
modal gain and gain spectrum/cavity mode matching are
paramount in VCSEL performance. In the use of
GaAsSb/ GaAs active materials for VCSEL applications, two

1620

important observations are presented in this work: 共i兲 the
lasing wavelength blueshift with injection level and 共ii兲 the
lack of improvement in modal gain with the addition of a
second QW. The injection dependent blueshift indicates band
filling and gain saturation, both of which are aggravated by
limited material gain and material inhomogeneities that are
evident from the low injection linewidth. The lack of improvement in modal gain when a second QW is added is
evident from both the gain constant and the blueshift measurements, which points to a nonuniform carrier distribution
that results in one of the two QWs achieving gain while the
other is at best transparent. In the worst case, the additional
QW provides loss, which has been reported in three QW
GaN based blue lasers.19 Our results do not provide information on how carriers are distributed, though insufficient hole
transport seems to be the most likely reason.
The phenomena of material gain saturation in GaAsSb/
GaAs QWs and nonuniform carrier distribution in double
QWs present a dilemma for VCSEL active region design:
since single QWs do not provide sufficient modal gain, a
straightforward solution is to incorporate more QWs to increase modal gain; however, increasing the QW number does
not necessarily increase the modal gain when the carrier distribution is not uniform. In VCSELs there are intrinsic processes that can reduce carrier distribution nonuniformity,
such as reabsorption of stimulated emission and active region heating. In the case of reabsorption, an absorbing QW
will become transparent through optical pumping from the
high optical field in the VCSEL cavity. In the case of active
region heating, the probability of thermionic emission of
holes from one QW to the other increases with temperature.
IV. CONCLUSIONS
The performance characteristics for three edge-emitting
laser structures with a single GaAsSb/ GaAs QW, a single
strain compensated GaAsSb/ GaAs/ GaAsP QW, and a
double strain compensated GaAsSb/ GaAs/ GaAsP QW active region are compared. These devices laser at around
1.2 ␮m and exhibit internal quantum efficiencies of up to
78%, the highest value reported to date for this material system. The threshold current density characteristic temperature
is determined to be around 60 K and is thought to reflect the
intrinsic temperature property of this material system, since
the internal quantum efficiency is sufficiently high that defect related Shockley-Read-Hall recombination is not dominant. Two key observations are ascertained from gain constant and lasing blueshift measurements regarding the use of
GaAsSb/ GaAs active materials for laser applications: 共i兲
there is a substantial blueshift in the gain peak with injection
level, resulting from band filling and gain saturation, and 共ii兲
there is a lack of improvement in modal gain with the addition of a second QW, resulting from inadequate carrier injection into one of the two QWs.
1

T. Anan, K. Nishi, S. Sugou, M. Yamada, K. Tokutome, and A. Gomyo,
Electron. Lett. 34, 2127 共1998兲.
2
P. Dowd et al., Appl. Phys. Lett. 75, 1267 共1999兲.
3
T. Anan, M. Yamada, K. Tokutome, S. Sugou, K. Nishi, and A. Kamei,

J. Vac. Sci. Technol. B, Vol. 24, No. 3, May/Jun 2006

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 209.147.144.20 On: Sun, 08 Feb 2015 00:35:31

1621

Yu et al.: Gain saturation and carrier distribution effects

Electron. Lett. 35, 903 共1999兲.
T. Anan, M. Yamada, K. Nishi, K. Kurihara, K. Tokutome, A. Kamei, and
S. Sugou, Electron. Lett. 37, 566 共2001兲.
5
O. Blum and J. F. Klem, IEEE Photonics Technol. Lett. 12, 771 共2000兲.
6
S. Ryu and P. D. Dapkus, Electron. Lett. 36, 1387 共2000兲.
7
F. Quochi, D. C. Kilper, J. E. Cunningham, M. Dinu, and J. Shah, IEEE
Photonics Technol. Lett. 13, 921 共2001兲.
8
Z. B. Chen et al., Conference on Lasers and Electro-Optics Technical
Digest, 2001, Vol. 1, p. 209.
9
P.-W. Liu, M.-H. Lee, H.-H. Lin, and J.-R. Chen, Electron. Lett. 38, 1354
共2002兲.
10
P. Dowd et al., Electron. Lett. 39, 987 共2003兲.
11
S.-Q. Yu et al., Conference on Lasers and Electro-Optics Technical
Digest, 2004, Vol. 1, p. 2.
4

1621
J.-B. Wang et al., Phys. Rev. B 70, 195339 共2004兲.
M. S. Noh, R. D. Dupuis, D. P. Bour, G. Walter, and N. Holonyak, Appl.
Phys. Lett. 83, 2531 共2003兲.
14
S. R. Johnson et al., J. Vac. Sci. Technol. B 19, 1501 共2001兲.
15
X. H. Zheng, D. S. Jiang, S. R. Johnson, and Y.-H. Zhang, Appl. Phys.
Lett. 83, 4149 共2003兲.
16
M. Yamada, T. Anan, K. Tokutome, A. Kamei, K. Nishi, and S. Sugou,
IEEE Photonics Technol. Lett. 12, 774 共2000兲.
17
S. Ryu and P. D. Dapkus, Electron. Lett. 38, 564 共2002兲.
18
G. Park, O. Shchekin, and D. G. Deppe, IEEE J. Quantum Electron. 36,
1065 共2000兲.
19
K. Domen, R. Soejima, A. Kuramata, K. Horino, S. Kubota, and T.
Tanahashi, Appl. Phys. Lett. 73, 2775 共1998兲.
12
13

JVST B - Microelectronics and Nanometer Structures

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 209.147.144.20 On: Sun, 08 Feb 2015 00:35:31