Strained InGaAs/GaPAsSb heterostructures grown on GaAs (001) for optoelectronic
applications in the 1100–1550 nm range
W. Braun, P. Dowd, C.-Z. Guo, S.-L. Chen, C. M. Ryu, U. Koelle, S. R. Johnson, Y.-H. Zhang, J. W. Tomm, T.
Elsässer, and D. J. Smith
Citation: Journal of Applied Physics 88, 3004 (2000); doi: 10.1063/1.1287233
View online: http://dx.doi.org/10.1063/1.1287233
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/88/5?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Highly tensile-strained, type-II, Ga 1 − x In x As / GaSb quantum wells
Appl. Phys. Lett. 96, 062109 (2010); 10.1063/1.3303821
Comparison of GaNAsSb and GaNAs as quantum-well barriers for GaInNAsSb optoelectronic devices operating
at 1.3 – 1.55 μ m
J. Appl. Phys. 96, 6375 (2004); 10.1063/1.1807028
Molecular beam epitaxial growth and characterization of strain-compensated Al 0.3 In 0.7 P/InP/Al 0.3 In 0.7 P
metamorphic–pseudomorphic high electron mobility transistors on GaAs substrates
J. Vac. Sci. Technol. B 19, 1519 (2001); 10.1116/1.1374629
InGaAsNSb/GaAs quantum wells for 1.55 μm lasers grown by molecular-beam epitaxy
Appl. Phys. Lett. 78, 4068 (2001); 10.1063/1.1379787
Molecular beam epitaxial growth of InAs/AlGaAsSb deep quantum well structures on GaAs substrates
J. Vac. Sci. Technol. B 16, 2644 (1998); 10.1116/1.590249

[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

JOURNAL OF APPLIED PHYSICS

VOLUME 88, NUMBER 5

1 SEPTEMBER 2000

Strained InGaAsÕGaPAsSb heterostructures grown on GaAs „001…
for optoelectronic applications in the 1100–1550 nm range
W. Braun,a) P. Dowd,b) C.-Z. Guo,c) S.-L. Chen,d)C. M. Ryu,e) U. Koelle,f)
S. R. Johnson, and Y.-H. Zhang
Department of Electrical Engineering and Center for Solid State Electronics Research, Arizona State
University, Tempe, Arizona 85287-6206

J. W. Tomm and T. Elsässer
Max-Born-Institute for Nonlinear Optics and Short Pulse Spectroscopy, Max-Born-Str. 2 A, D-12489 Berlin,
Germany

D. J. Smith
Center for Solid State Science and Department of Physics and Astronomy, Arizona State University, Tempe,
Arizona 85287-1704

共Received 20 March 2000; accepted for publication 26 May 2000兲
We present a novel semiconductor quantum well 共QW兲 structure consisting of alternating 共In,Ga兲As
and Ga共P,As,Sb兲 layers grown pseudomorphically on a GaAs substrate by all-solid-source
molecular beam epitaxy. The band gap of the QW is determined by the thickness and composition
of both types of layers and can be varied from 1.1 to 1.55 ␮m. Calculations show that the observed
strong room-temperature photoluminescence in this wavelength range can be explained by a type-II
transition in the QW. Structural investigations by reflection high-energy electron diffraction,
transmission electron microscopy, and secondary ion mass spectroscopy confirm a triple layer
structure with laterally modulated composition. Photoluminescence measurements reveal a
linewidth of 50 meV at 1.3 ␮m and a luminescence decay time of 240 ps. Our investigations
demonstrate the feasibility of this materials system for vertical cavity surface-emitting lasers and
other optoelectronic devices on GaAs.
© 2000 American Institute of Physics.
关S0021-8979共00兲03917-7兴

I. INTRODUCTION

Vertical-cavity surface-emitting lasers 共VCSELs兲 operating at 1.3 and 1.55 ␮m are potentially attractive devices for
a wide range of applications, such as optical communication,
data links, and optical interconnection. In addition to their
wavelength matching the dissipation and dispersion minima
of optical fibers, such devices would offer a lower operating
voltage, allowing their direct integration with low-voltage
Si-based integrated circuits.
For the fabrication of such long-wavelength VCSELs,
materials need to be found that combine a strong radiative
transition at the desired wavelength for the active region with
a large difference in refractive index for the Bragg mirrors.
Several approaches to this problem are currently being investigated. Whereas the InP/In共Ga,As,P兲 materials system is the
system of choice for edge emitters in this wavelength range,
a兲

Now at Paul–Drude Institute for Solid State Electronics, Hausvogteiplatz
5-7, D-10117 Berlin, Germany; electronic mail: braun@pdi-berlin.de
b兲
Now at Nanyang Technological University, Department of Electrical and
Electronic Engineering, Nanyang Avenue, Singapore 639798.
c兲
Permanent address: Department of Physics, Peking University, Beijing
100871, China.
d兲
Permanent address: Department of Applied Mathematics, Tsing Hua University, Beijing, 100084, China.
e兲
Now at Motorola, 1303 E. Algonquin Rd. IL01/ANX2, Schaumburg, IL
60196.
f兲
Now at Agilent Technologies, 350 W. Trimble Rd./M.S. 90UB, San Jose,
CA 95131.

VCSELs grown monolithically with these materials exhibit
poor performance due to the high thermal sensitivity and
refractive index properties of the phosphides.1 One solution
is to bond 共In,Ga兲共As,P兲 quantum well 共QW兲 active regions
to GaAs2 using wafer fusion techniques. This method has
been successfully applied to VCSEL structures,3,4 resulting
in an improved device performance. The main problem with
this technique is the wafer fusion process, which is technologically rather difficult and results in high cost per unit.
A single epitaxial growth process on a GaAs substrate is
therefore highly desirable. One approach uses quantum dot
共QD兲 structures produced by Stranski–Krastanov growth of
共In,Ga兲As on GaAs. 共In,Ga兲As QD structures have shown
photoluminescence 共PL兲 at up to 1.35 ␮m.5,6 An edgeemitting QD laser has been demonstrated using this active
region material.7 The emission wavelength was 1.31 ␮m and
in order to achieve lasing, a long 共2.1 mm兲 cavity and high
reflectivity facet coatings were required. Continuous wave
共cw兲 room temperature operation of an 共In,Ga兲As QD-based
VCSEL has also been achieved,8 but the lasing wavelength
was only 1.15 ␮m.
Room temperature PL at 1.3 ␮m has been observed using strained Ga共As,Sb兲 quantum wells 共QWs兲 and lasing has
been reported in an edge-emitting device at 1.27 ␮m.9 Using
共Ga,In兲共N,As兲 as the active layer material, cw operation of an
edge emitting laser at 1.3 ␮m was achieved in a device
where the nitrogen content of the QW was increased to 1%.10
So far, the approach able to achieve the longest emission

0021-8979/2000/88(5)/3004/11/$17.00
3004
© 2000 American Institute of Physics
[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

wavelength in a GaAs-based VCSEL structure used a single
共Ga,In兲共N,As兲 QW. Room-temperature 共RT兲 pulsed operation was achieved with an emission wavelength of 1.18 ␮m
and a threshold current density of 3.1 kA/cm2 . 10
PL up to 1.332 ␮m has been observed in Ga共As,Sb兲/
共In,Ga兲As bilayer QW samples, with a type-II band-edge
alignment.11,12 Among other advantages,13 such a type-II
band alignment allows more freedom in obtaining a certain
operation wavelength of an optoelectronic device, since the
composition and strain of two materials can be tailored,
thereby increasing the degrees of freedom in the device design. Both edge- and surface-emitting lasers have been demonstrated using type-II transitions around 2.2 ␮m14,15 and 4
␮m.16
In this work, we demonstrate a similar approach for the
communication wavelength range using 共In,Ga兲As/
Ga共P,As,Sb兲 QWs grown on GaAs. The samples show strong
room-temperature PL in the range of 1100–1550 ␮m, making these structures ideal candidates for GaAs-based optoelectronic devices for this entire wavelength range.
In the following, we first discuss theoretical considerations that demonstrate the feasibility and high efficiency of
such type-II structures. We then continue to describe the molecular beam epitaxy 共MBE兲 growth process using an all
solid-source growth chamber, together with the in situ and
postgrowth structural characterization of the layer structures.
The third section covers photoluminescence and electroluminescence measurements, including time-resolved data.
II. ENERGY LEVELS AND GAIN SPECTRA

In order to assess the feasibility of the proposed type-II
QW structure and to compare it with matured 1.3 ␮m
共In,Ga兲AsP/InP lasers, a well established empirical two-band
theoretical model taking into account nonparabolicities in
both the conduction and valence bands has been used17 to
calculate the quantum confined energy levels and their corresponding normalized wave functions. Without losing generality, the calculations presented here focus on typical structures consisting of In0.4Ga0.6As and GaP0.12As0.40Sb0.48 layers
共labeled layers A and B, respectively兲. The critical thickness
for the coupled double quantum well 共CDQW兲 structure
made of this material system is calculated to be 7 nm based
on a single-kink mechanism18 by the energy balance method.
Figure 1 shows the calculated band-edge diagram of a typical
CDQW structure with a thickness of 2 and 3 nm for the A
and B layers, respectively. Since the band offset between
these materials is not well known, the band-edge alignment
of the CDQW has been calculated using the model solid
theory19 taking into account strain effects. The band-edge
discontinuities and the energy levels of the electron confined
states in the conduction band and those of the heavy-hole
confined states in the valence band with no injection are
plotted to scale. The material parameters used for the theoretical calculations, together with the calculated band offsets
are listed in Table I. These calculations indicate that the
band-edge alignment of the CDQW structure is type II. The
electron barrier is 196 meV at the In0.4Ga0.6As/
GaP0.12As0.4Sb0.48 heterointerface.

Braun et al.

3005

FIG. 1. The band-edge diagram and confined states of electrons and heavyholes for an In0.4Ga0.6As/GaP0.12As0.4Sb0.48/In0.4Ga0.6As CDQW structure
with layer thicknesses dA ⫽2 nm and dB ⫽3 nm, embedded in GaAs. The
lowest energy-state electron and heavy-hole wave functions are also shown.

The calculated wave functions for both the electron and
the hole states in the CDQW are also plotted in Fig. 1. The
heavy-hole wave function is strongly localized in layer B due
to the heavy-hole effective mass. Since layer B is very thin,
has a low electron barrier, and possesses small electron effective mass, the electron wave function can penetrate
through it, resulting in strong coupling between the two A
关共In,Ga兲As兴 wells. Therefore, there is a substantial part of the
electron wave function present in the B 关Ga共P,As,Sb兲兴 barrier
layer. As a result, the electron- and hole-wave functions have
a reasonably large overlap.
For VCSELs, the injection level in the active region can
be very high. It is therefore important to take into account
band gap shrinkage 共band gap renormalization兲 due to
electron–electron, electron–hole, and hole-hole interactions.
To obtain a good estimate without going through complicated calculations of these many-body effects, a theoretical
model20 is used to estimate the band gap shrinkage. The
additional potential variation across the well region due to
the spatially separated charges of electrons and holes is expected to be small for the proposed narrow quantum well and
has not been included in our calculation.
To calculate the gain spectra of the strained type-II
CDQW structure, an equivalent type-I mixed QW model has
been developed. In this model, it is assumed that the coupling between the wells splits every energy level in an isolated well into two, with the total density of states being
conserved. The density of states for a coupled energy level in
the CDQW is equivalent to that for an energy level in a
single quantum well which has the same effective mass and
well width as the CDQW structure. This allows the optical
transition spectra to be calculated for an equivalent type-I
QW in which the electron and hole energy levels have been
determined by consideration of the type-II band-edge alignment. As it is found quantitatively that the injected carriers
are mainly 共more than 90%兲 populated in the lowest electron
and heavy-hole confined states, the densities of states for a
symmetrical CDQW can be defined as

[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

3006

Braun et al.

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

TABLE I. Material parameters used for the calculations of an In0.4Ga0.6As/GaP0.12As0.4Sb0.48/In0.4Ga0.6As CDQW structure embedded in GaAs.
Layer

Substrate

Material
Electron effective mass, m c /m 0
Heavy-hole effective mass, m h /m 0
Light-hole effective mass, m l /m 0
Spin-orbital splitting energy, ⌬ s 共eV兲
Refractive index, n̄ r
Lattice constant, a 共nm兲
Strain, ␧ 共%兲
Poisson ratio, ␯
Electron-heavy-hole band gap, Egh␧ 共eV兲
Electron-light-hole band gap, Egl␧ 共eV兲

A

GaAs
0.067
0.45
0.074
0.34
3.409
0.5653
0
0.312
1.424
1.424

B

In0.4Ga0.6As
0.049
0.434
0.056
0.356
3.459
0.5815
⫺2.867
0.325
1.017
1.212

GaP0.12As0.4Sb0.48
0.06
0.439
0.065
0.539
3.557
0.5841
⫺3.304
0.302
0.983
1.201

Band offsets for different heterointerfaces
Heterointerfaces
Conduction band offset, V c␧ 共eV兲
Heavy-hole band offset, V h␧ 共eV兲
Light-hole band offset, V l␧ 共eV兲

S-A
0.237
0.171
-0.025

A-B
0.196
0.229
0.207

Energy gap between A and B bandedges, Ech0 共eV兲
Split between heavy- and light-hole band edges, ⌬Ehl 共eV兲

QW
␳ cn
⫽

␳ QW
vn ⫽

0.787
0.217

1 m cA
1 m cA
m cA
⫹
⫽
,
2
2
2 ␲ ប d A 2 ␲ ប d A ␲ ប 2d A

␲ប dB
2

n̄ r ⫽
共2.1兲

m vB

v ⫽h,l

,

for the conduction band and the valence bands, respectively,
where dA and dB are the well 共layer A兲 and barrier 共layer B兲
thicknesses, and mcA and mv B are the electron and hole 共including heavy- or light-hole兲 effective masses in the A and B
layers, respectively. The gain spectrum with intraband scattering broadening is given by
g 共 ប ␻ 兲 ⫽C g 共 ប ␻ 兲

兺
兺 冕E
n,m v ⫽h,l

⬁

⫺ f v m 共 E t 兲兴

qvn

r sp 共 ប ␻ 兲 ⫽C r 共 ប ␻ 兲

兺
兺 冕E
n,m v ⫽h,l

⬁

冉 冊

2 dE c v ,

共2.2兲

␰ ⫽C m

qvn

共 E c v ⫺ប ␻ 兲 2 ⫹

FM⫽

冉 冊
ប
␶ in

2 dE c v

共2.3兲

in which the prefactors are given by
C g⫽

C r⫽

␲ q 2e ប

1

c 0 ␧ 0 m 20 n̄ r •ប ␻
q 2e

␲ c 30 ␧ 0 m 20 ប 2

,

n̄ r •ប ␻ ,

2
兩 M 兩 2 ⫽ 兩 M b 兩 2 F M 兩 I vcnA
mB 兩 ␦ k ctA ,k v tB ,

兩 M 兩 2 ␳ QW
r v n f cn 共 E t 兲

1 ប
␲ ␶ in

⫻ 关 1⫺ f v m 共 E t 兲兴

where q e , c 0 , ␧ 0 , m 0 , and n̄ r are the electron charge, light
velocity, permittivity in vacuum, free electron mass, and refractive index, respectively. f cn and f v m are the Fermi–Dirac
distribution functions in the conduction and valence bands,
respectively. The optical momentum matrix element will be
given approximately by

兩 M b 兩 2 ⫽ ␰ m 0 E AB
g ,

ប
共 E c v ⫺ប ␻ 兲 ⫹
␶ in
2

2n̄ rA ⫹n̄ rB
,
3

共2.5兲

I vcnA
mB ⫽C c v n ␦ nm ,

兩 M 兩 2 ␳ QW
r v n 关 f cn 共 E t 兲

1 ប
␲ ␶ in

S-B
0.041
0.400
0.182

共2.4兲

3
4

E AB
g ⫹⌬ sB
2
E AB
g ⫹ ⌬ sB
3

冉

1⫹

共2.6兲
m cA
m0
,
m cA
6
m0

1⫺

冊

E hmB
,
m rh
E hmB ⫹
E
m hB t

where C m ⬇1.27, 21 and I vcnA
mB is the interband interlayer electron and hole wave function overlap integral with value C c v n
for m⫽n. Here, E AB
g is the bandgap including the effect of
shrinkage. F M is the polarization-dependent factor for the TE
mode. It is important to emphasize that the larger the electron and hole wave function overlap I vcnA
mB , the larger the
momentum matrix element and the material gain, as shown
in Eqs. 共2.2兲 and 共2.6兲. It is therefore desirable to choose the
structure with the largest wave function overlap.

[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

FIG. 2. The calculated TE mode material gain spectra for a 2 nm-3 nm-2 nm
structure as a function of carrier concentration.

Calculated gain spectra under different injection levels
are shown in Fig. 2 for a structure with d A⫽2.0 nm and
d B⫽3.0 nm. In this case, the transparency carrier concentration is calculated to be approximately 5.3⫻1018 cm⫺3 . The
maximum value of the gain peak is 8 008 cm⫺1 under a
carrier injection of 1019 cm⫺3 and occurs at a wavelength of
1.33 ␮m. This carrier concentration is not unreasonable for
the operation of VCSELs.
III. SAMPLE PREPARATION

Based on these calculations, we designed, grew, and
characterized samples consisting of typically three successive QW layers embedded in GaAs. The layer sequence was
共In,Ga兲As–Ga共P,As,Sb兲–共In,Ga兲As with individual QW
thicknesses of approximately 3 nm. The target In composition was around ⬇30%, so that the thickness of the combined structure was below the calculated critical thickness of
10–20 nm, depending on the Sb content. The maximum
strain per layer was calculated to be less than 3%. Both the
In content in the electron confinement layer and the P:As:Sb
ratio in the hole confinement layer were varied systematically to obtain different transition wavelengths.
A. MBE growth

All samples were grown by molecular beam epitaxy
共MBE兲 on 2 in. GaAs substrates. We used the doublechamber system at Arizona State University in which one of
the growth chambers 共B兲 was equipped with solid source
valved cracker cells for As, Sb, and P. The second growth
chamber A, accessible from chamber B through a ultrahigh
vacuum 共UHV兲 transfer tunnel, was equipped for highuniformity growth using Al, Ga, In, and As with Be and Si as
dopants.
The growth temperature for the active region layer stack
was 505 °C, the remaining GaAs/AlGaAs growth being done
at typical growth temperatures around 580 °C. We did not
observe any significant difference between overgrowth at
505 or 580 °C as long as the initial capping of the active
layer was done at 505 °C. The samples for PL measurements

Braun et al.

3007

were deposited on top of a 20-period AlAs–GaAs superlattice followed by a 220 nm GaAs buffer layer, and they were
capped by a 10-nm-thick GaAs layer. A special sample with
a thinner cap layer was grown for high-resolution secondary
ion mass spectroscopy 共SIMS兲 measurements. Additionally,
an edge emitting light emitting diode 共LED兲 was fabricated
using the UHV transfer capabilities of the MBE machine.
Both the top and bottom AlGaAs cladding layers were deposited in chamber A while the QW region was grown in
chamber B with a total of two intergrowth UHV transfers.
Chamber B was equipped with all solid sources for Ga,
Al, In, As, Sb, and P. Additional dopant cells were not used
in this study. For As, Sb, and P, valved cracker sources were
used with the cracking zones set at 900 共As兲, 1000 共Sb兲, and
900 °C 共P兲. This resulted in a large fraction of dimers in the
molecular beams of these group V cells. The Ga growth rate
was determined from reflection high-energy electron diffraction 共RHEED兲 intensity oscillations during substrate
rotation.22 共In,Ga兲As growth rates were inferred from
RHEED oscillations on InAs, which are usually less accurate
than in the GaAs case.
The group V source fluxes could not be quantitatively
determined in the growth chamber since the flux measurement gauge of chamber B was not operational at the time of
the growths. Typical background pressures measured
roughly opposite the P cell and at a right angle to the As cell
with the Sb cell at an intermediate position were 2.3⫻10⫺8
mbar during 共In,Ga兲As deposition and 9⫻10⫺9 mbar during
growth of the Ga共P,As,Sb兲 layer.
To grow the mixed group V structure, shutters in front of
the Sb and P cells were used in addition to changing the
valve positions. A timing diagram for the growth of the
trilayer stack is shown in Fig. 3. The Ga共P,As,Sb兲 growth
was separated from the 共In,Ga兲As by 32 and 34 s growth
interruptions, in which the surface was soaked with the
group V fluxes of the following layer. Sb is known to preferentially incorporate into a nongrowing surface and replace
the As.23 This behavior may be modified by the additional
presence of a P flux in the present case, but we have to take
into account that the central layer may be thicker than nominally expected.
The valve positions during Ga共P,As,Sb兲 growth for the
samples discussed in this work are given in Table II. The As
data refer to the As valve position during growth of the Ga共P,As,Sb兲 layer, all 共In,Ga兲As layers were grown with a setting of 130. By varying the fluxes of In in the outer layers
and the group V fluxes of the central layer, we were able to
vary the resulting peak PL wavelength from 1.15 to 1.55 ␮m.
The dependence of PL wavelength on the Sb:P valve position ratio is shown in Fig. 4 for the values of Table II. All
data marked by solid lines were taken with the same Sb
valve setting, which means that the As:Sb ratio and the sum
of the As and Sb fluxes were constant, and only the P flux
was varied. For the data marked by thinner lines, the Sb
setting was lower, resulting in a higher relative As content of
the central layer. These two samples had nominally identical
Ga共P,As,Sb兲 layers and only the In content of the outer layers was varied, resulting in the observed wavelength shift.
Generally, increasing the Sb:P ratio led to longer wave-

[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

3008

Braun et al.

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

FIG. 3. Timing sequence for the shutters and valves during growth of the
active layer. Shutters in front of the
valved cells are used in addition to the
valves to increase the dynamic range
of the molecular fluxes.

lengths. The trend cannot be extrapolated, however, since
with 共In,Ga兲As/GaAsSb/共In,Ga兲As samples 共without P兲, the
maximum wavelength we obtained was 1.24 ␮m.
During growth, we recorded RHEED snapshots from the
rotating sample, usually close to the 关 1̄10兴 azimuth. The results for two samples, B126 and B124, are displayed in Fig.
5. The letters indicate the position of each panel in the
growth sequence of Fig. 3. The two different samples differed in their Sb–to–P ratio 共see Table II兲, which was 5 for
B126 and 8.3 for B124. For good reproduction in print, the
intensity grayscale within each subpanel is normalized to its
maximum. This means that the contrast in each panel is an
indication of the relative peak intensity: low contrast panels
indicate low intensity of the diffraction peak, high contrast
panels mean high intensity of the diffraction peak. The high
contrast panels therefore correspond to a more ordered surface in the sense that the surface morphology is of regular
periodicity. The low contrast panels indicate irregular arrangements of surface atoms with weak correlations.
An additional indication for the varying surface morphology during growth is the position of the main diffraction
spot in the different panels. The lower position corresponds

to the specularly reflected spot, moving slightly up and down
during rotation due to substrate wobble. The upper position
is the 002 transmission diffraction spot, which is stationary
because it is locked to the direction of the primary beam by
the diffraction condition. Transmission spots appear if elevated surface features are present that have a lateral dimension smaller than the extinction length of the electrons. Comparisons between RHEED diffraction patterns and atomic
force microscopy 共AFM兲 images24 demonstrate that this transition can appear for as little as 2 monolayers of deposited
mismatched material with a modulation amplitude below 1
nm. The typical chevron shape of 共In,Ga兲As Stranski–
Krastanov growth is present in both the bottom and top
共In,Ga兲As layers of both structures 关Fig. 5, B126共c,d,i兲,
B124共c,d,e,i兲兴. This indicates the presence of a modulated
surface morphology at the corresponding moments in the
growth sequence.
Very weak diffraction spots, indicated by low contrast in
the figure, occur in panels Fig. 5共h兲 for B126 and Fig. 5共g兲
and 5共h兲 for B124. These are the points in the growth sequence where the group-V flux is switched back to arsenicstable conditions 共Fig. 3兲. The change in group-V flux seems
to cause a major rearrangement of the surface, accompanied
by strongly increased disorder. Another unusual phenom-

TABLE II. Room temperature PL wavelengths and selected growth parameters of the samples discussed in this work.
PValve
B101
B114
B115
B116
B117
B118
B119
B120
B121
B122
B123
B124
B125
B126
B127
B132
B133
B134
B140

0
50
50
50
50
50
50
70
70
70
40
30
20
40
40
40
0
40
40

AsValve

SbValve

100
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
130
40
40

0
250
250
250
250
250
250
250
250
250
250
250
250
200
200
250
0
200
250

InTemp
801
795
795
795
795
796
796
796
796
796
796
796
796
796
790
796
796
785
796

␭ ( ␮ m兲
1.14
1.29–1.32
1.31
1.24–1.25
1.27
1.31–1.33
1.20–1.26
1.24–1.25
1.15–1.16
1.36–1.41
1.37–1.41
1.41–1.55
1.49–1.52
1.33–1.40
1.44–1.49
1.12–1.22
1.33
1.30–1.32a

FIG. 4. Dependence of the PL wavelength on the flux ratios during growth
of the active layer. The thinner lines indicate samples grown with a different
Sb valve setting. The corresponding numerical values can be found in
a
Table II.
Electroluminescence.
[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:

209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

Braun et al.

3009

FIG. 5. RHEED diffraction patterns obtained from two samples during growth of the active region. The individual panels show the specular/transmission
diffraction spot on the 共00兲 streak at various times indicated by the corresponding letters in Fig. 3.

enon is the special shape of the diffraction spot in Fig. 5共f兲
and 5共g兲 for B126 and Fig. 5共f兲 for B124. In contrast to both
the two-dimensional 共2D兲 reflection and the 共In,Ga兲As layer
pattern, the diffraction spot is wider or even split parallel to
the surface and more narrow perpendicular to it. At the same
time, the contrast is high, indicating good correlation. The
diffraction spot, however, is still at the three-dimensional
共3D兲 position, which means that the surface is not flat. The
horizontally extended or split peak shape can be explained
by lateral domain formation.25 In such a model, the layer
material segregates into volumes with different structure or
composition without significant variation of the lattice parameter. The lateral spot size of roughly 1/50 of the Brillouin
zone corresponds to a minimum real-space period 共domain
size兲 of 50 lattice constants or 20 nm, which agrees well with
the average modulation period observed with transmission
electron microscopy 共TEM兲 共Fig. 8兲. A more detailed analysis of the layer structure at this point in the growth sequence
is difficult based on the present diffraction data. Spontaneous
ordering as the parameter distinguishing the different domains seems rather unlikely, since this usually takes place
along the 具111典 directions. The splitting or broadening, however, is parallel to the sample surface. This would rather
point to a compositional modulation parallel to the surface.
This may be due to the Ga共P,As,Sb兲 filling in the valleys of
the Stranski–Krastanov template, accumulating on the
Stranski–Krastanov mounds, or a more homogeneous deposition where mostly the composition varies. More detailed
studies will be needed to clarify this point.

Samples B123 and B132 were compared with a dedicated SIMS sample 共B134兲 that was grown with a very thin
cap layer to obtain maximum depth resolution. These highresolution B134 SIMS profiles are shown in Fig. 6. The
steepest slopes of all peaks in such a scan represent a lower
bound for the resolution of the SIMS. In the present case,
these are the surface side shoulders of the phosphorus and
antimony peaks. Any slopes that are less steep must therefore be due to structural broadening.
Both the P and the Sb peaks are well localized, indicating that the intended three-layer structure is indeed realized.
A slight asymmetry of the phosphorus peak towards the substrate indicates diffusion into the bulk material. In segregates
with a diffusion length of 2.2 nm, determined from the

B. SIMS analysis

Several samples were investigated using secondary ion
mass spectroscopy 共SIMS兲 to obtain compositional information of the as-grown structures.

FIG. 6. High-resolution SIMS scan of sample B134.
[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

3010

Braun et al.

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

TABLE III. Ratios of active layer compositions deduced from SIMS
measurements.
Sample
123
132
134

In
1
1.13
0.97

P
1
0.63
0.78

Sb

␭ ( ␮ m兲

1
1
1.25

1.36–1.41
1.44–1.49
1.33

the emission wavelength. A change of the P/Sb ratio with
slightly lower In content 共B134兲, however, results in a
slightly decreased emission wavelength. The basis of three
measurement points does not allow us to determine whether
strain or composition is the governing factor that defines the
emission wavelength.
FIG. 7. Standard resolution SIMS scan of samples B123 and B132.

shoulder on the surface side of the active layer 共4–9 nm
depth兲. A small amount of P seems to segregate along with
the In, as can be seen by the low-intensity P peak shoulder
between 5 and 10 nm depth. A second period of the active
layer was grown underneath the measured structure at ⬇40
nm depth, resulting in an In shoulder with the same slope
that is visible at 25–32 nm depth.
Of the two first-period In peaks, the right one has a
steeper slope towards the surface with only about 1 nm segregation length. This coincides with the tendency of P to
diffuse in the opposite direction at this interface. At the opposite interface, where 共In,Ga兲As is grown on Ga共P,As,Sb兲,
the behavior is reversed: whereas P does not segregate, In
shows enhanced diffusion back into the Ga共P,As,Sb兲. The
two interfaces within the active layer therefore show a distinct asymmetry in their formation behavior, causing the two
共In,Ga兲As layers to be different.
The two low-resolution SIMS measurements are shown
in Fig. 7. With the lower resolution, the In peaks can no
longer be separated and, apart from In segregation, slope
differences can no longer be resolved. All peak slopes except
for the left side of the indium peak are the same. The clear
broadening of the Sb and P peaks towards the sample surface
may be due to inhomogeneous segregation of In. This would
result in laterally different hardness and therefore different
sputtering rates of the cap layer, leading to a rough sputtering
front. The softer Ga共P,As,Sb兲 layer would then be sputtered
off the harder underlying 共In,Ga兲As layer before this layer is
attacked, resulting in a smoother sputtering front for the underlying 共In,Ga兲As.
No calibration standard is available for Ga共P,As,Sb兲.
The present SIMS results can therefore not be directly quantified. The count rates for Ga and As away from the active
region stack, however, happen to be the same for all three
samples to within the noise levels. This enables us to make
quantitative statements about the relative variations of each
constituent from sample to sample. The results are shown in
Table III, normalized to the composition of sample B123.
Increasing the strain from B123 to B132 by increasing the In
content and decreasing the P content leads to an increase in

C. TEM analysis

TEM observations of cross-sectional samples have been
carried out to characterize the layer structure and the strain
distribution. A cross-sectional view of the active region of
sample B114 is shown in Fig. 8. The growth direction was
upwards. The QW structure, which is approximately 9 nm
wide, can be seen running horizontally through the middle of
the image. Strain contrast is clearly visible extending into the
GaAs cladding layers on either side of the well. The strain
field variations are most probably due to the thickness and
compositional variations within the QW itself.
Lattice imaging showed no visible sign of structural defects. Since the image contrast obtained for different group V
ions is likely to be similar under high-resolution imaging
conditions, the composition within this QW cannot be determined easily. However, the undulating contrast within the
QW structure confirms the thickness and composition variations. These variations are consistent with the interpretation
of the RHEED oscillation measurements.
IV. OPTICAL CHARACTERIZATION

Temperature-dependent PL measurements have been
performed using the 514 nm line of an argon–ion laser at
excitation densities up to 1000 W/cm2 . The spectra of a
range of samples were obtained using a monochromator and
the spectrally resolved signal was detected using a cooled
Ge-detector connected to a lock-in amplifier.

FIG. 8. TEM micrograph of the active region 共B114兲.
[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

Braun et al.

3011

FIG. 10. Linewidth and integrated PL intensity as a function of temperature
for sample B114.

FIG. 9. Room temperature PL spectra of two different 共In,Ga兲As/
Ga共P,As,Sb兲/共In,Ga兲As QW structures embedded in GaAs. 共a兲 Sample B114
at different excitation intensities, 共b兲 sample B132 at an excitation intensity
of 1000 W/cm2 .

A. Photoluminescence

Samples with different compositions for the Ga共P,As,Sb兲
layer have shown PL emission over a broad wavelength
range, extending up to about 1.5 ␮m at room temperature
共RT兲.
Figure 9 shows the RT PL spectra obtained for two
samples. For sample B114 关Fig. 9共a兲兴, the peak of the PL
spectrum was observed at 1.3 ␮m and the linewidth, defined
as the full width at half maximum 共FWHM兲, was measured
to be 110 meV. As the excitation intensity increased, a blueshift of the PL spectrum of 13 meV was observed. For
sample B132 关Fig. 9共b兲兴, the peak wavelength was close to
1.5 ␮m, with a FWHM of 130 meV. Additionally it was
observed that the peak emission intensity and the integrated
emission was weaker than for samples emitting closer to 1.3
␮m. For this sample, the Sb:P valve ratio was higher and
SIMS also indicates an increased Sb content 共see Table III兲,
thus we conclude that more Sb is incorporated in the middle
layer of the CDQW. This could increase the valence band
offset of the Sb layer, resulting in a narrower effective band
gap. A higher Sb content in the layer also causes a larger
lattice mismatch between the layer and GaAs. For the increased Sb content, it is therefore possible that dislocations

have been introduced into sample B132 that act as nonradiative recombination centers. This could provide an explanation for the reduced emission intensity at this wavelength.
At lower temperatures, the FWHM of the samples decreased and the PL intensity increased. Figure 10 shows the
results for sample B114 at an excitation density of about 50
W/cm2 . For temperatures below about 120 K, the linewidth
remains roughly constant, with a minimum value of ⬇74
meV. Above this temperature, the linewidth can be seen to
broaden. Figure 10 also shows the integrated PL intensity for
sample B114.
On decreasing the temperature to 2 K, the intensity increases by a factor of 16. Most of the intensity increase occurs as the temperature is reduced to 120 K. Below this
point, the intensity change is considerably smaller.
Figure 11共a兲 shows the excitation-dependent linewidths
for three type-II samples 共B114, B116, B129兲 and one
共In,Ga兲As sample 共B101兲 at a temperature of 2 K. A minimum linewidth of 50 meV is obtained for sample B129,
where the RHEED measurements indicated that the
Stranski–Krastanov relaxation of the 共In,Ga兲As layers was
less pronounced. For the 共In,Ga兲As-only sample, B101, the
minimum linewidth is found to be 40 meV.
Figure 11共b兲 shows the peak emission wavelength of
four type-II samples 共B114, B116, B123, B129兲 as a function
of the excitation intensity. A large blue-shift of the wavelength of the two 共In,Ga兲As/Ga共P,As,Sb兲/共In,Ga兲As samples
is clearly visible as the pumping power increases.
As the excitation density is raised from 5 to 250 W/cm2 ,
a larger blue-shift of the peak wavelength corresponding to
30 meV is observed. This is primarily attributed to the band
filling of lower energy states which arise from compositional
fluctuations within the GaPAsSb layer. Such blue-shifts have
been observed in type-II structures.12,26 In contrast, the peak
wavelength of the 共In,Ga兲As-only sample is found to be independent of the excitation density for this measurement
range.
The broad linewidths of the samples are believed to be
dominated by fluctuations in the composition of the
Ga共P,As,Sb兲 layer and the widths of the QW layers,27 see
Secs. III B and III C. Additional linewidth broadening

[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

3012

Braun et al.

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

FIG. 12. Transient infrared PL from sample B114. Zero on the time axis
was chosen corresponding to the maximum of the excitation pulse.

FIG. 11. Linewidths and emission wavelengths as a function of excitation
power density for various samples at a temperature of 2 K.

mechanisms are believed to include 共In,Ga兲As QW asymmetry resulting from segregation in the growth direction and the
strong band-filling effect in type-II structures.28
B. Time-resolved photoluminescence

Room-temperature transient PL measurements have
been performed with sub-100 fs pulses (␭ exc⫽840 nm, 1.48
eV兲 at a repetition rate of 82 MHz that were generated by a
diode-laser-pumped Cr:LiSAF laser. A streak camera
equipped with a cooled charge-couple device 共CCD兲 camera
and an infrared-enhanced S1 streak tube 共300–1600 nm兲
were used for detection. The FWHM of the total system
response was better than 15 ps. Typical excitation photon
fluxes per pulse were on the order of 1012 – 1013 cm⫺2 . Si
and other edge filters were used for spectral discrimination.
Samples B114, B123, B126, and B133 were investigated. Out of this series, the longest wavelength sample
B126 共cf. Table II兲 produced the weakest signal. Since the
sensitivity of the streak tube is reduced by one order of magnitude from 1300 to 1500 nm, this is most likely due to the
tube characteristics.
Figure 12 shows a typical PL transient of the IR emission in the 1050–1500 nm spectral window from sample
B114. A biexponential decay with time constants of ␶ 1
⫽(240⫾30) ps and ␶ 2 ⫽(450⫾50) ps was observed. Within
the error limits of the fit, we found a similar behavior for
sample B123, whereas both time constants were substantially
reduced for the reference sample B133 that basically represents an In0.4Ga0.6As double QW. Here, ␶ 1 amounted to
about 150 ps, whereas ␶ 2 was roughly 300 ps. Excitation-

dependent measurements indicated an almost linear dependence for the ‘‘slow’’ ( ␶ 2 ) and a sublinear 共saturation-like兲
dependence for the fast ( ␶ 1 ) process.
Since we were not able to separate both carrier decay
processes with respect to the spectral position of their emission, a definite assignment is difficult. It is worth to note,
however, that all observed ␶ values are smaller than the 2.5
ns calculated in Sec. II for the spontaneous emission lifetime.
Probably, nonradiative processes decrease the PL decay
time.
We should mention that we did not observe any infrared
emission from the reverse side of the substrate or from uncoated front side regions near the wafer edge. Nonetheless,
very strong GaAs PL at about 870 nm ( ␶ ⬇500 ps兲 is present
in all time-resolved measurements. Despite the fact that PL
contributions from the GaAs substrate or cap were eliminated, indirect interference by photon recycling processes
cannot be excluded. Thus the slow ( ␶ 2 ) infrared emission
might be excited by reabsorbed GaAs PL.
C. Electroluminescence

In the case of the edge-emitting LED, its schematic
structure is shown in Fig. 13. Here, the 共In,Ga兲As/

FIG. 13. Schematic structure of the edge-emitting LED.
[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

Braun et al.

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

3013

FIG. 16. Temperature-dependent integrated electroluminescence as a function of the current density at various temperatures.
FIG. 14. Room temperature electroluminescence spectra as a function of
current density.

Ga共P,As,Sb兲/共In,Ga兲As QW region was sandwiched between
two doped and graded AlGaAs cladding layers. Devices with
a stripe width of 80 ␮m and a cavity length of approximately
750 ␮m have been investigated and characterized.
Figure 14 shows room-temperature optical spectra of
one of the electrically pumped devices operating at a range
of current densities between ⬇175 to ⬇700 A/cm2 , corresponding to injection currents of 100–400 mA. The pulse
duration was 10 ␮s, and the duty cycle was 10%. For a
maximum injection current of 400 mA (700 A/cm2 ), strong
EL at close to 1.3 ␮m was measured. As the injection current
was decreased to 200 (350 A/cm2 ) and 100 mA
(175 A/cm2 ), the room-temperature optical spectra shifted to
longer wavelengths 共approximately 1.325 and 1.35 ␮m, respectively兲. The spectral blue-shift at the higher bias currents
is primarily attributed to the band filling of lower energy
states which arise from thickness and compositional fluctuations within the 共In,Ga兲As and Ga共P,As,Sb兲 layers. Gain
saturation of the small volume active region may also occur.
Low-temperature measurements have also been carried
out and these have revealed a change of the peak wavelength

of roughly 3.5 Å /K. Furthermore, as can be seen in Fig. 15,
a long-wavelength spectral tail becomes clearly visible. This
provides further evidence of a broad range of lower-energy
‘‘tail’’ states, caused by the compositional fluctuations in the
GaPAsSb layer. A narrowing of the spectral width from
about 150 meV at room temperature to 120 meV at 20 K has
also been observed.
The integrated EL has been measured and is shown in
Fig. 16. For each temperature, the integrated intensity increases as the current density increases. However, it tends to
saturate at the higher current densities. On increasing the
pulse length or the duty cycle, the integrated output was
found to decrease. The saturation is attributed to the fact that
there are likely to be a large number of defects in the nonoptimized active region of the device. These cause nonradiative recombination to dominate the device performance.
V. CONCLUSION

In conclusion, we have theoretically and experimentally
demonstrated a novel layered semiconductor system that allows strong room-temperature emission at 1.1 to 1.55 ␮m
wavelength with devices based on GaAs substrates. The ease
of fabrication on GaAs together with the possibility to vary
the wavelength across the entire communications wavelength
range by only changing the composition of the active layers
make this system a promising candidate for the fabrication of
optoelectronic devices, and in particular GaAs-based
VCSELs. The main problem that needs to be addressed in
further research is the relatively broad linewidth of the emission line, which is most probably due to the lateral composition variations in the active region. One challenge will
therefore be to grow this structure without approaching the
Stranski–Krastanov transition to obtain a two-dimensional
geometry.

J. Piprek and S. J. B. Yoo, Electron. Lett. 30, 866 共1994兲.
T. Chung, N. Hosoda, T. Suga, and H. Takagi, Appl. Phys. Lett. 72, 1565
共1998兲.
3
J. J. Dudley, D. I. Babic, L. Yang, R. Mirin, B. I. Miller, R. J. Ram, T.
Reynolds, E. L. Hu, and J. E. Bowers, Lasers and Electro-Optics Society
Annual Meeting, 1993. LEOS ’93 Conference Proceedings, IEEE, 1993,
pp. 560–561.
4
J. Behrend et al., J. Cryst. Growth 188, 295 共1998兲.
5
R. P. Mirin et al., Appl. Phys. Lett. 67, 3795 共1995兲.
6
K. Nishi, H. Saito, S. Sugou, and J.-S. Lee, Appl. Phys. Lett. 74, 1111
FIG. 15. Temperature-dependent electroluminescence spectra at 400 mA as
共1999兲.
7
a function of temperature.
D. L. Huffaker et al., Appl. Phys. Lett. 73, 2564 共1998兲.
[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:
1
2

209.147.144.21 On: Tue, 10 Feb 2015 00:41:57

3014
8

Braun et al.

J. Appl. Phys., Vol. 88, No. 5, 1 September 2000

D. L. Huffaker, H. Deng, and D. G. Deppe, IEEE Photonics Technol. Lett.
10, 185 共1998兲.
9
M. Yamada, T. Anan, K. Tokutome, K. Nishi, A. Gomyo, S. Sugou,
Lasers and Electro-Optics Society Annual Meeting, 1998, LEOS’98,
IEEE, Vol. 1, pp. 149–150.
10
K. Nakahara et al., IEEE Photonics Technol. Lett. 10, 487 共1998兲.
11
M. Peter et al., Appl. Phys. Lett. 67, 2639 共1995兲.
12
M. Peter et al., J. Electron. Mater. 24, 1551 共1995兲.
13
J. Meyer, C. Hoffman, F. Bartoli, and L. Ram-Mohan, Appl. Phys. Lett.
67, 757 共1996兲.
14
A. N. Baranov et al., Electron. Lett. 32, 2279 共1996兲.
15
A. N. Baranov et al., Electron. Lett. 34, 281 共1998兲.
16
J. Malin et al., Appl. Phys. Lett. 68, 2976 共1996兲.
17
D. Nelson, R. C. Miller, and D. Kleinman, Phys. Rev. B 35, 7770 共1987兲.
18
G. Salviati et al., The Proceedings of the 17th Congress on Electron Mi-

croscopy, Italian Society, Lecce 2730, 1989 共unpublished兲.
C. van de Walle, Phys. Rev. B 39, 1871 共1989兲.
20
D. Kleinman and R. Miller, Phys. Rev. B 32, 2266 共1985兲.
21
S. Corzine, R. Yan, and L. Coldren, Appl. Phys. Lett. 57, 2835 共1990兲.
22
W. Braun, H. Möller, and Y.-H. Zhang, Appl. Phys. Lett. 74, 138 共1999兲.
23
F. Hatami et al., Appl. Phys. Lett. 67, 656 共1995兲.
24
B. J. Ohlsson and M. S. Miller, J. Cryst. Growth 188, 387 共1998兲.
25
H. Raether, in Encyclopedia of Physics, edited by S. Flügge 共Springer,
Berlin, 1957兲, pp. 512–515 共in German兲.
26
E. R. Glaser, B. R. Bennett, B. V. Shanabrook, and R. Magno, Appl. Phys.
Lett. 68, 3614 共1996兲.
27
A. Chin and W. J. Chen, Appl. Phys. Lett. 69, 443 共1996兲.
28
Y.-H. Zhang, R. Miles, and D. H. Chow, IEEE J. Sel. Top. Quantum
Electron. 1, 749 共1995兲.
19

[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:
209.147.144.21 On: Tue, 10 Feb 2015 00:41:57