Uncooled low-voltage AlGaAsSb/InGaAsSb/GaSb

avalanche photodetectors
O.V. Sulima, M.G. Mauk, Z.A. Shellenbarger, J.A. Cox, J.V. Li, P.E. Sims, S. Datta and S.B. Rafol
Abstract: Low-voltage AlGaAsSb/InGaAsSb/GaSb separate absorption and multiplication
avalanche photodiodes (SAM-APD), as well as InGaAsSb/GaSb APDs with an AlGaAsSb
passivating layer, are fabricated using liquid phase epitaxy. Formation of the p n junction is
performed either epitaxially or through diffusion of Zn from the vapour phase. Responsivity as high
as 43 A/W at wavelength of 2100 nm is achieved at room temperature in AlGaAsSb/InGaAsSb/
GaSb SAM-APDs reverse biased at 6.7 V. Relatively high responsivity (8.9 A/W at 2000 nm) was
also measured in an InGaAsSb/GaSb APD reverse biased at 7.5 V.

Introduction

AlGaAsSb-InGaAsSb heterostructures lattice matched to

GaSb are of great interest for use in high-performance
photodetectors, LEDs and lasers in the mid-infrared
wavelength region as well as for thermophotovoltaic
devices. AstroPower has accumulated considerable
experience in the development of inexpensive liquid phase
epitaxy (LPE) for fabrication of AlGaAsSb=InGaAsSb=
GaSb photodetectors [1 3].
LPE inherently provides high material quality, both with
respect to defect density and purity. This may be attributed to
the preferential segregation of impurities to the liquid phase,
the high mobility of atoms in the liquid phase (as compared to
surface diffusion upon which vapour phase techniques
depend), and near-equilibrium growth conditions.
In this work, we continued to study LPE growth of
AlGaAsSb-InGaAsSb heterostructures. Additionally, a
technically uncomplicated and productive method of Zn
diffusion from the vapour phase was used to simplify and
improve the fabrication technology of photodetectors
including low-voltage SAM-APDs and APDs.
APDs are widely used for optical fibre communication.
However, commercially available APDs require relatively
high voltage: hundreds of volts for Si APDs l  1000 nm
and tens of volts for InGaAs=InP APDs l  1700 nm:
In this work, low-voltage (6 8 V) SAM-APDs and
conventional APDs were developed. Moreover, in comparison with Si or InGaAs=InP APDs; the long wavelength
sensitivity of these devices was extended to 2000 2100 nm.
Also, the high ratio of the hole ionisation coefficient b to
the electron ionisation coefficient a (115 in AlGaSb close to

q IEE, 2004
IEE Proceedings online no. 20040142
doi: 10.1049/ip-opt:20040142
Paper received 14th July 2003
O.V. Sulima, M.G. Mauk, J.A. Cox, J.V. Li, P.E. Sims and S. Datta are with
AstroPower Inc., Solar Park, Newark, DE 19716, USA. J.V. Li is on leave
from the University of Illinois at Urbana-Champaign
Z.A. Shellenbarger was with Astropower Inc., and is now with the Sarnoff
Corporation, Princeton, NJ 08453, USA
S.B. Rafol is with the Jet Propulsion Laboratory, 4800 Oak Grove Drive,
Pasadena, CA 91109, USA
IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004

GaSb at 77 K [4], 60 in Al0:4 Ga0:96 Sb at room temperature

[5], and 4 7 in In0:20 Ga0:80 As0:17 Sb0:83 at 230 K [6])
can minimise noise in AlGaAsSb=InGaAsSb APDs or
SAM-APDs.
2

Experimental

2.1 Epitaxial growth

A standard horizontal slide boat technique is used for the
LPE growth of the InGaAsSb and AlGaAsSb layers. The
graphite slide boat is situated in a sealed quartz tube placed
in a microprocessor-controlled, programmable, three-zone
tube furnace. The growth ambient is palladium-diffused
hydrogen at atmospheric pressure with a flow rate of
300 ml=min: The substrates with an area of 14  14 mm2
are polished, (100) oriented, n-type GaSb wafers doped to
3 5  1017 cm3 with tellurium.
The basic growth solution for InGaAsSb epitaxy is
indium xIn 0:59; gallium xGa 0:21 and antimony
xSb 0:20; where xi is the atomic fraction of the element
i in the solution. High purity 99:9999% indium, gallium,
and antimony metals are used and no intentional doping of
the melt is applied. Prior to growth, the melt is baked out at
700  C for 15 hours under flowing hydrogen to de-oxidise
the metallic melt components and outgas residual impurities. After the baking and cooling, an undoped polycrystalline InAs float wafer is placed on top of the melt, and the
boat is heated to 530 C: In this way, arsenic from the float
InAs wafer is added to the melt through a partial dissolution
of the wafer at this temperature. The melt is equilibrated for
one hour at 530 C and then cooled to the growth
temperature. At 515 C; the substrate is contacted with the
melt for two minutes to grow a 5-micron thick
In0:15 Ga0:85 As0:17 Sb0:83 base layer with an electron carrier
concentration of 1 4  1016 cm3 : Unintentionally doped
In0:15 Ga0:85 As0:17 Sb0:83 layers could be both of n- and
p-type. Adding a tiny amount (less than 1 mg) of rare earth
elements such as Gd or Yb to the In Ga As Sb melt
results in a low doped n-type In0:15 Ga0:85 As0:17 Sb0:83 layers.
Afterwards, the melt is wiped from the substrate. Electron
probe microanalysis (EPMA) was used to determine
the solid composition of the InGaAsSb layer. The electron
concentration in InGaAsSb was calculated from C V
measurements performed after a p n junction was formed
in the InGaAsSb layer (see below).
1

The growth of the AlGaAsSb layer is carried out from the

growth solution having liquidus composition of gallium
xGa 0:959; antimony xSb 0:013 and aluminium
xAl 0:028: Arsenic is added to this melt from an
undoped polycrystalline float GaAs wafer through the
partial dissolution of the wafer at the growth temperature.
The growth of AlGaAsSb is performed either in the same
epitaxial run together with InGaAsSb, or in a separate
epitaxial process using an InGaAsSb=GaSb wafer as a
substrate. The reason to carry out separate processes is as
follows. We have found that in some experiments drops of
the first melt (In Ga As Sb) were left on the surface of
the wafer after wiping (especially at the edges). These drops
when mixed with the second melt (Al Ga As Sb), lead to
the local formation of unpredictable melt compositions and
hence irreproducible growth. A better reproducibility is
achieved when the wafer is taken out from the boat after the
first layer growth, and the drops of carryover metal (if any)
are removed. In the case of separate epitaxies, the abovedescribed procedure of melt baking is repeated for a Ga-Sb
fraction of the second melt prior to growth. Aluminium and
a GaAs float wafer are added to the melt after the baking
and cooling. The growth is performed at 509 C during a
90-second long contact of the substrate with the melt
resulting in a 2-micron thick Al0:28 Ga0:72 As0:014 Sb0:986
layer. AlGaAsSb layers grown without intentional doping
are p-type. As we need an n-type layer, some Te-doped
GaSb with a doping concentration of 2  1018 cm3 was
added and partly replaced Ga and Sb in the melt. Figure 1
shows the carrier concentration in n-type Al0:28 Ga0:72 As0:014
Sb0:986 against the GaSb:Te amount in the Al Ga As Sb
melt. Adding of GaSb:Te into the melt provided an accurate
control of the electron concentration down to n
6 7  1016 cm3 :
It is interesting to note that adding of a small amount of
Yb (less than 1 mg) extended the p-type area in Fig. 1
towards a higher amount of Te. In other words, more Te was
necessary to overcome the natural p-type background in
AlGaAsSb if Yb was present in the melt. This effect can be
explained by a chemical reaction between Te and Yb
leading to formation of YbTe and preventing Te from
entering AlGaAsSb during its growth. A similar decrease of
donor concentration was observed in LPE-grown InP doped
with Te when another rare-earth element, Dy, was added to
the melt [7].
The lattice matching Da=a between the GaSb substrate
and InGaAsSb or AlGaAsSb layers was determined through
X-ray diffraction measurements. The Da=a values were as
low as 0:1% and 0:3% for InGaAsSb and AlGaAsSb layers,
respectively.

2.2 Diffusion of Zn
Pseudo-closed box diffusion of Zn from the vapor phase was
performed into n-AlGaAsSb layers in a H2 atmosphere
purified by a Pd cell. A specially designed graphite boat
similar to that described in [8] was used. Separate pure Zn
and Sb vapor sources were used in more than sufficient
quantities to provide the saturation of vapour pressures. The
design of the graphite boat ensures the uniformity of the Zn
vapour pressure across the wafer surface, and thus the
uniformity of the p-AlGaAsSb layer depth.
Figure 2 shows two as-diffused Zn profiles in
Al0:28 Ga0:72 As0:014 Sb0:986 obtained by means of a fourhour diffusion of Zn at 460  C and 470  C: The profiles were
measured by secondary ion mass spectroscopy (SIMS). The
part of the profile near the surface has a very high
concentration of Zn (about 1020 cm3 : This 300 400 nm
thick dead layer (shown hatched in Fig. 2) with a low
electron diffusion length is removed by means of anodic
oxidation and selective oxide etching. The constant Zn level
of 3  1016 cm3 beyond 1500 nm is caused by a limited
SIMS sensitivity and should not be considered.
If the AlGaAsSb layer is diffused all the way through, Zn
diffusion in InGaAsSb must be considered. To evaluate
Zn diffusion in InGaAsSb, data from [9] were used.

Fig. 2 Zn concentration profiles measured by SIMS in an

Al0:28 Ga0:72 As0:014 Sb0:986 layer
Hatched part of the profile shows a dead layer that is removed
later. The dotted line shows the electron concentration in the n-type
Al0:28 Ga0:72 As0:014 Sb0:986 layer determined by C V measurements

2.3 Device fabrication

Figure 3 shows a photodiode mesa structure fabricated and
studied in this work. It is composed of a 5-micron thick
n-type In0:15 Ga0:85 As0:17 Sb0:83 layer with n 1 4 
1016 cm3 and a 2-micron thick Al0:28 Ga0:72 As0:014 Sb0:986
layer. The n-part of the AlGaAsSb layer (several tenths of
micron thick) is doped at the level of n 6 7  1016 cm3 :
In this work we applied two methods to form a p n
junction: (i) epitaxial, i.e. both p- and n-type AlGaAsSb
layers were grown epitaxially, and (ii) hybrid (epitaxy and
diffusion), i.e. the p-AlGaAsSb layer was formed by
diffusion of Zn from the vapour phase into the epitaxially
grown n-AlGaAsSb layer. Several advantages of the hybrid
approach are:

Fig. 1 Carrier concentration in n-Al0:28 Ga0:72 As0:014 Sb0:986

layers against amount of tellurium doped GaSb in Al Ga As
Sb melt
2

. Technically, diffusion is simpler than epitaxy

. As only one AlGaAsSb layer is epitaxially grown, a

difference in chemical composition between matrices of

p- and n-layers is not possible
IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004

Fig. 3 Scheme of the fabricated AlGaAsSb=InGaAsSb=GaSb

SAM-APD
In the case of a pn-InGaAsSb photodiode or APD, the n-AlGaAsSb layer
was replaced by a p-InGaAsSb one, and the p-AlGaAsSb layer was used
only for the InGaAsSb surface passivation

Fig. 5 Dark I V curves of two AlGaAsSb=InGaAsSb=GaSb

diodes with epitaxial and diffused p n junctions in the
Al0:28 Ga0:72 As0:014 Sb0:986 layer at 25 C
Diameter of mesas: 400 mm

Fig. 4 Top-view of the fabricated photodiodes

. Diffusion of Zn is faster at the places of possible crystal
defects in the n-AlGaAsSb layer. Thus, electrical isolation
of these defects from the p n junction takes place. Such
isolation is not evident in the case of two epitaxial layers.

By fixing the thickness of the n-AlGaAsSb layer and

changing the position of the p n junction within it, different
types of diodes can be fabricated. If the p n junction is in
AlGaAsSb and is far from the AlGaAsSb InGaAsSb
interface, a homojunction p n AlGaAsSb photodiode is
realised. The low bandgap InGaAsSb has no function in this
case. If the p n junction is in AlGaAsSb close to the
AlGaAsSb InGaAsSb interface, a SAM-APD diode is
fabricated. And finally, if the AlGaAsSb-layer is diffused
all the way through, and a p n junction is formed in the
InGaAsSb layer, a p n InGaAsSb APD with an AlGaAsSb
window is realised.
Mesa photodiodes with 400-mm diameter total area and
200-mm diameter active area were formed using photolithography and wet chemical etching (Fig. 4). Metallisation
for the back n-type contacts was planar Au=Sn e-beam
deposition while front p-type contacts were annular
Ti=Ni=Au: A spin-on, photosensitive polyimide layer was
deposited and patterned before contact deposition. This had
several functions including planarising the surface for the
front contact deposition, providing isolation of the junction,
and passivating the edges of the device area. After
photodiodes were formed, the substrate was diced into
1-mm2 pieces with a single device in the middle of each
square. These devices were mounted to TO-18 headers
using silver conducting epoxy and wire-bonded from the
bonding part of the front contact to the header post. No
antireflection coatings were applied purposely, however, a
visible though not homogenous oxide layer was formed on
the surface after processing. This layer can be considered as
an antireflection coating and should be taken into account by
analysing results of quantum efficiency measurements.
3

junctions in the Al0:28 Ga0:72 As0:014 Sb0:986 layer. One can see
that formation of a diffused p n junction provides a higher
breakdown voltage and essentially lower dark current at the
voltages close to the breakdown. This can be explained by
the electrical isolation of possible crystal defects in the
AlGaAsSb layer due to a higher diffusion rate of Zn in the
vicinity of such defects. If a random crystal defect is formed
close to an epitaxial p n junction, it increases dark current
or can even short-circuit the p n junction. In the case of
diffusion, Zn penetration into the AlGaAsSb layer (and thus
position of the p n junction) is deeper in the vicinity of the
defect. Thus, the p n junction interface is formed away
from the defect, effecting a kind of passivation.
Both curves in Fig. 5 (shown for room temperature)
exhibited positive temperature dependence of the breakdown voltage. This indicates that, although a mixture of
both avalanching and tunnelling breakdown mechanisms
may exist in these samples, the avalanching is dominant.
Figure 6 displays dark I V curves of AlGaAsSb=
InGaAsSb=GaSb diodes with diffused p n junctions in
InGaAsSb. For these samples, different breakdown mechanisms are quite obvious but the reason for such a difference
is not quite clear. We believe, that different breakdown
mechanisms can be explained by a slight variation of doping
in InGaAsSb. The C V measurements performed with the
tunnelling diode showed a doping level of n 3 4 
1016 cm3 in the n-InGaAsSb. Unfortunately, it was not
possible to make a definite conclusion about the
electron concentration from the C V measurements of

Results and discussion

Figure 5 shows the comparison of dark I V curves at room

temperature of two diodes with epitaxial and diffused p n
IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004

Fig. 6 Dark I V curves of AlGaAsSb=InGaAsSb=GaSb diodes

with p n junctions in InGaAsSb with different doping at 25 C
3

Fig. 7

Spectral response measurement set-up

the avalanching diode. However, regarding the growth

conditions, one cannot expect a large difference in the
electron concentration of these two diodes. In this work, the
lowest measured n-value in In0:15 Ga0:85 As0:17 Sb0:83 Eg
0:55 eV at room temperature) was about 1  1016 cm3 :
In [5] avalanching in In0:20 Ga0:80 As0:17 Sb0:83 Eg 0:54 eV
at room temperature) with n-values 2  1016 cm3 was
observed. Thus one can surmise that avalanching in
InGaAsSb with Eg 0:54  0:55 eV can be observed
when n-values are  1  2  1016 cm3 :
A computer-controlled spectrometer was used to
measure spectral response of the fabricated devices
(Fig. 7). The light source is a 1000 W tungsten halogen
lamp. A calibrated detector with a calcium fluoride
window is used as a reference for relative measurements.
This detector has flat response in the wavelength range
from 200 nm to over 10 000 nm. To determine the
absolute spectral response, we used a reference
AlGaAsSb=InGaAsSb=GaSb photodiode, which was calibrated at the Jet Propulsion Laboratory (JPL) and had the
same design as the tested diodes.
Figure 8 shows the responsivity of an InGaAsSb=GaSb
APD: Both the I V curve of this diode with a sharp increase
of current at the breakdown voltage (Fig. 6), and
responsivity against reverse voltage (Fig. 8) clearly show
that this diode is behaving as an APD. Improvement of
carrier collection and separation due to reverse-bias
widened space charge region (SCR) with voltage cannot
explain the enhancement of responsivity up to 8:9 A=W at
7.5 V. This value corresponds to an external quantum

Fig. 9 Spectral response of a low-voltage AlGaAsSb=InGaAsSb=

GaSb SAM-APD at different reverse bias voltages at 25 C

efficiency of 546%; which can only be explained by

avalanche multiplication.
Figure 9 shows that an AlGaAsSb=InGaAsSb=GaSb
SAM-APD provides a much higher responsivity (up to
43 A=W) than an InGaAsSb=GaSb APD: However, for
array applications not only responsivity but also reproducibility of device parameters must be considered. From this
point of view, a simpler design of the InGaAsSb=GaSb APD
in which the exact position of the p n junction in the
InGaAsSb layer is not of crucial importance, can be
advantageous. In contrast, the thickness of the n-AlGaAsSb
layer in different AlGaAsSb=InGaAsSb=GaSb APDs
integrated into an array must be kept precisely the same
(about 0.3 microns) to obtain the same multiplication values
at one voltage.
In comparison with InGaAsSb=GaSb APDs with the
p n junction in the In0:15 Ga0:85 As0:17 Sb0:83 layer with a
bandgap of about 0.55 eV, much lower dark currents were
obtained in AlGaAsSb=InGaAsSb=GaSb SAM-APDs
with the p n junction in the Al0:28 Ga0:72 As0:014 Sb0:986
layer with a bandgap of about 1.2 eV. This shows that it
is advantageous to have a p n junction in a material
with a higher bandgap, in particular for noise reduction
in the diode. However, as it was shown in [5], absorption
layers with less Al in Al(Ga)AsSb might be also
beneficial for reduction of excess noise arising due to
the multiplication. As a band resonance condition
(bandgap is equal to the spin-orbit splitting of the
valence band) takes place in Al0:04 Ga0:96 Sb at 0.76 eV,
very high values of a ratio of the hole and electron
ionisation coefficients (up to 60) can be achieved. Thus,
an essentially unipolar multiplication by holes is
provided that reduces the excess noise problem in these
APDs. In future work we will compare what bandgap of
AlGaAsSb will result in the best parameters of
SAM-APDs.
4

Fig. 8 Spectral response of a low-voltage InGaAsSb=GaSb APD

at different reverse bias voltages at 25 C
4

Conclusions

A low-voltage AlGaAsSb=InGaAsSb=GaSb SAM-APD and

an InGaAsSb=GaSb-APD with an AlGaAsSb passivating
window were fabricated and studied by using inexpensive
liquid phase epitaxy and diffusion of Zn from the vapour
phase.
It was found that substitution of an epitaxial p n junction
in AlGaAsSb with a diffused junction, leads to a marked
decrease of the dark current.
IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004

Responsivities as high as 43 A=W at a wavelength of

2100 nm were achieved in AlGaAsSb=InGaAsSb=GaSb
SAM-APD biased at 6.7 V at room temperature. An
avalanching regime was achieved in InGaAsSb=GaSb
APDs and a responsivity of 8:9 A=W at a wavelength of
2000 nm was measured at 7.5 V at room temperature.
5

Acknowledgment

The authors wish to thank B. Ber from Ioffe Institute,

St. Petersburg, Russia for SIMS measurements.
6

References

1 Shellenbarger, Z.A., Mauk, M.G., Cox, J.A., South, J.D., Lesko, J.D.,
Bower, J.R., Sims, P.E., Jhabvala, M., and Kraut Fortin, M.:
Recent progress in GaInAsSb and InAsSbP photodetectors for midinfrared wavelengths, Proc. SPIEInt. Soc. Opt. Eng., 1998, 3287,
pp. 138 145
2 Shellenbarger, Z.A., Mauk, M.G., Sims, P.E., Cox, J.A., Lesko, J.D.,
Bower, J.R., South, J.D., and DiNetta, L.C.: Progress on GaInAsSb and
InAsSbP photodetectors for mid-infrared wavelengths, Mater Res. Soc.
Symp. Proc., 1998, 484, pp. 135140

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004

3 Shellenbarger, Z.A., Mauk, M.G., Gottfried, M.I., Lesko, J.D., and

DiNetta, L.C.: GaInAsSb and InAsSbP photodetectors for
mid-infrared wavelengths, Proc. SPIEInt. Soc. Opt. Eng., 1997,
2999, pp. 2533
4 Zhingarev, M.Z., Korolkov, V.I., Mikhailova, M.P., and Sazonov, V.V.:
Dependence of the electron and hole impact-ionization coefficients on
the orientation and composition in solid solutions, Sov. Tech. Phys. Lett.,
1981, 7, (12), pp. 637 638
5 Andreev, I.A., Afrailov, M.A., Baranov, A.N., Marinskaya, N.N.,
Mirsagatov, M.A., Mikhailova, M.P., and Yakovlev, Yu.P.: Low-noise
avalanche photodiodes with separated absorption and multiplication
regions for the spectral interval 1.6-2.4 microns, Sov. Tech. Phys. Lett.,
1989, 15, (9), pp. 692 694
6 Andreev, I.A., Mikhailova, M.P., Melnikov, S.V., Smorchkova, Yu.P.,
and Yakovlev, Yu.P.: Avalanche multiplication and ionisation
coefficients of GaInAsSb, Sov. Phys. Semicond., 1991, 25, (8),
pp. 861 865
7 Gorelenok, A.T., Kamanin, A.V., and Shmidt, N.M.: Rare-earth
elements in the technology of InP, InGaAsP and devices based on
these semiconductor compounds, Microelectron. J., 1995, 26, (7),
pp. 705 723
8 Bett, A.W., Keser, S., and Sulima, O.V.: Study of Zn diffusion into
GaSb from the vapour and liquid phase, J. Cryst. Growth, 1997, 181,
pp. 9 16
9 Sulima, O.V., Beckert, R., Bett, A.W., Cox, J.A., and Mauk, M.G.:
InGaAsSb photovoltaic cells with enhanced open-circuit voltage,
IEE Proc., Optoelectron., 2000, 147, (3), pp. 199204