Full text of DOI:10.1007/BF02665860

HgCdTe/CdTe/Si Infrared Photodetectors Grown
Journal of ELECTRONIC MATERIALS, Vol. 30, No. 6, 2001
Special Issue P7a1pe5r
by MBE for Near-Room Temperature Operation
HgCdTe/CdTe/Si Infrared Photodetectors Grown by
MBE for Near-Room Temperature Operation
S. VELICU,1,5 G. BADANO,1 Y. SELAMET,1 C.H. GREIN,1 J.P. FAURIE,1 S.
SIVANANTHAN,1 P. BOIERIU,2 DON RAFOL,3 and R. ASHOKAN4
1.—University of Illinois at Chicago, Microphysics Laboratory, Department of Physics, Chicago,
IL 60607. 2.—EPIR Ltd., Bolingbrook, IL 60440. 3.—Jet Propulsion Laboratory, Pasadena, CA
91109. 4.—Smart Pixel Inc., Bolingbrook, IL 60440. 5.—e-mail: silviu@uic.edu
Conventional HgCdTe infrared detectors need significant cooling in order to
reduce noise and leakage currents resulting from thermal generation and
recombination processes. Although the need for cooling has long been thought to
be fundamental and inevitable, it has been recently suggested that Auger
recombination and generation rates can be reduced by using the phenomena of
exclusion and extraction to produce nonequilibrium carrier distributions. The
devices with Auger suppressed operation requires precise control over the
composition, and donor and acceptor doping. The successful development of the
molecularbeamepitaxy(MBE)growthtechniqueformulti-layerHgCdTemakes
it possible to grow these device structures. Theoretical calculations suggest that
the ppn+ layer sequence is preferable for near-room temperature operation due
to longer minority carrier lifetime in lightly doped p-HgCdTe absorber layers.
However, because the low doping required for absorption and nonequilibrium
operation is easier to achieve in n-type materials, and because Shockley-Read
centersshouldbeminimizedinordertoobtainthebenefitsofAugersuppression,
we have focused on p+nn structures. Planar photodiodes were formed on CdTe/
Si (211) composite substrates by As implantation followed by a three step
annealing sequence. Three inch diameter Si substrates were employed since
they are of high quality, low cost, and available in large areas. Due to this
development,largeareafocalplanearrays(FPAs)operatedatroomtemperature
arepossibleinthenearfuture. ThestructureswerecharacterizedbyFTIR, x-ray
diffraction, temperature dependent Hall measurements, minority carrier life-
times by photoconductive decay, and in-situ ellipsometry. To study the relative
influence of bulk and surface effects, devices with active areas from 1.6¥ 10–5 cm2
to 10–3 cm2 were fabricated. The smaller area devices show better performance
in terms of reverse bias characteristics indicating that the bulk quality could be
further improved. At 80 K, the zero bias leakage current for a 40 mm ¥ 40 mm
diode with 3.2 mm cutoff wavelength is 1 pA, the R0A product is 1.1 ¥ 104 W-cm2
andthebreakdownvoltageisinexcessof500mV.Thedeviceshowsaresponsivity
of 1.3 ¥ 107 V/W and a 80 K detectivity of 1.9 ¥ 1011 cm-Hz1/2/W. At 200 K, the zero
bias leakage current is 5 nA and the R0A product 2.03 W-cm2, while the
breakdown voltage decreases to 40 mV.
Key words: HgCdTe, infrared detectors, MBE
INTRODUCTION
high sensitivities are required for many strategic,
space, and spectroscopic applications. It would be
highly desirable to preserve the near-BLIP perfor-
mance at 77 K of HgCdTe-based infrared (IR) detec-
tors but increase the operating temperature. The
main obstacle in achieving this goal is the large level
Present day near-room-temperature infrared pho-
ton detectors display sub-background limited perfor-
mance (BLIP). Background limited performance and
(ReceivedNovember25, 2000;acceptedDecember19, 2000)
711

Velicu, Badano, Selamet, Grein, Faurie,
Sivananthan, Boieriu, Rafol, and Ashokan
716
Fig. 1. MBE grown HgCdTe nnn+ heterostructure.
of noise due to Auger generation and recombination
near room temperature. Ashley and Elliott1 proposed
the suppression of the Auger process by reducing the
electron and hole concentrations below their equilib-
rium values. Two configurations—pnn+ and npp+—
are possible to achieve such nonequilibrium devices.
It has long been realized2 that the lifetime in Auger-
dominated HgCdTe devices is larger in p-type mate-
rial than in n-type material. Hence, the ppn+ struc-
ture would be preferable for the near-room tempera-
ture operation of non-equilibrium detectors from a
thermalgenerationratepointofview. Theppn+ struc-
tures have been grown by metalorganic vapor phase
epitaxy (MOVPE) and the Auger generation in the p
layer was found to be the limiting mechanism.3 In a
recent study, Elliott et al.4 have shown that the
doping required for background limited behavior in
the mid-wavelength infrared (MWIR) spectral range
in a detector with a 2p field of view is in mid 1014 cm–
3 range, difficult to achieve with the today’s MOVPE
technology.Themolecularbeamepitaxy(MBE)growth
technique offers the capability for n-type doping in
the low 1014 cm–3 range. Also, the quality of indium
doped n-HgCdTe layers is well known to be better
than As-doped p-HgCdTe layers mainly because
Shockley-Readcentersdegradep-typematerialmuch
more than n-type material. This leads to increased
mobilitiesandincreasedminoritycarrierlifetimesfor
n-typematerial.TorealizeAuger-suppresseddevices,
an epitaxial process with independent control of com-
position and doping is critical. The use of Si, a high
quality and cost-effective substrate that is available in
large areas, creates the possibility for large format
focal plane arrays (FPAs) operated at room tempera-
ture in the near future. Preliminary results on the
application of the MBE technique to grow HgCdTe
device structures on Si substrates, particularly to
satisfy the low donor concentration requirements, and
the potential for near-room temperature operation of
HgCdTe-based devices are discussed in this paper.
Fig. 2. Theoretical R0A product with and without radiative recombina-
tion mechanism included.
from which thermally generated carriers are col-
lected on the low-doped side of the junction is limited
by a low surface recombination velocity.
We performed preliminary calculations based on
the model developed by Tennant and Cabelli.5 The
modelassumesthatthedominantcurrentisdiffusion
fromtheabsorptionregion. Theinfluenceoftheinter-
faces, ofthedepletionregion, andoftheheavilydoped
side of the junction are neglected. Both radiative and
Auger mechanisms are used for the computation of
the minority carrier lifetimes. However, according to
Humphrey’s6 theory, radiative recombination does
not limit the performance of the diode due to the
reabsorption of the photons generated by recombina-
tion within the detector. Because the reabsorption
time is very small (<1 ps), the associated process is
found to be noiseless. We calculate the R0A product
according to both Humphrey’s and van Roesbroeck’s7
theories.WealsoslightlymodifiedtheTennantmodel
to allow for the computation of the R0A product in n+p
diodes and to compare n+p and p+n structures. The
ratio of the intrinsic Auger 7 to Auger 1 lifetime is
taken from Ref. 2:
6(1- 5Eg / 4kT)
1- 3Eg / 2kT
(1)
g =
where Eg is the bandgap of the absorber layer, k is the
Boltzman’s constant, and T is the temperature.
We computed the dependence of the R0A product on
the doping concentration in both n- and p-type ab-
sorber layers. The results are presented in Fig. 2 with
and without radiative recombination mechanisms
included. We assumed total carrier extraction so that
the Auger and radiative rates are determined only by
the extrinsic doping. It can be seen that a low 1014 cm–3
donor doping gives a better R0A product than a high
1015 acceptor doping. As we will show later, mid and
DEVICE STRUCTURE AND MODELING
Based on the above facts, we focused on p+nn device
structures. A schematic of such a non-equilibrium
photodiode is shown in Fig. 1. This device structure
involves both exclusion at the nn contactand extrac-
tion at the np+ contact. It offers an improvement over
the conventional two-layer type of junction even in an
equilibrium mode of operation because the volume

HgCdTe/CdTe/Si Infrared Photodetectors Grown
by MBE for Near-Room Temperature Operation
717
a
b
Fig. 3. Temperature dependence of the (a) carrier concentration and (b) mobility after annealing for the first batch of MWIR samples.
even low 1014 cm–3 donor levels are experimentally
achievable using the MBE technique. For vacancy or
extrinsicdopedHgCdTe, theacceptorlevelinap-type
absorber is difficult to control at these low levels. So,
even if theoretically higher lifetimes are expected
because g is computed to be greater than unity in
lightly doped p-type HgCdTe, we believe that a prac-
tical way to achieve high temperature performance is
to use an n-type absorber layer. At near-room tem-
perature, the absorber is intrinsic and the minority
carrier extraction occurs at the reverse biased np+
junction.Theminoritycarrierholewillnotbeinjected
into the absorber layer when the np+ junction is
reverse biased because the p+ layer is of a wider band
gap. As a result, the net hole concentration in the
absorber layer will fall by orders of magnitude. To
maintain charge neutrality, the majority carrier con-
centrationdecreasestotheextrinsicdopinglevel.The
isotype junction, nn, forms an excluding hi-lo type of
contact to the central absorber region, thus prevent-
ing the injection of carriers to replace those extracted
at the diode junction. The n contact is a low recombi-
nation velocity interface and does not contribute to
the dark current due to both band bending and to
band filling to such a degree that Auger 1 generation
is quenched in that volume. It is important to note
that the electric fields are sufficiently low to avoid
carrier heating since the extraction process is princi-
pally a diffusion process. The reverse bias should be
chosen to exceed the critical field necessary for the
extraction process, but kept below a certain level to
avoid the carrier heating. Calculations by Ashley and
Elliott indicate that the required exclusion field can
be achieved in the MWIR HgCdTe diodes over a wide
range of doping levels without exceeding the esti-
mated field for carrier heating.
energy electron diffraction (RHEED) gun and an
ellipsometer to achieve real-time monitoring of the
crystallinestructureandthecomposition.CdTe(211)/
Si substrates, with x-ray rocking curve full width at
half maximum (FWHM) lower than 100 arcsec and
etch pit densities below 105 cm–2, were used for the
HgCdTe growth.
The targeted structure for this growth consists of
three HgCdTe layers capped with a CdTe layer, as
indicated in Fig. 1. The first HgCdTe layer from the
top is short wave infrared (SWIR), n-type doped with
indium in-situ during the growth. The second layer is
a n-type doped MWIR absorber layer, with targeted
CdTemolefractionofabout0.3andcarrierconcentra-
tion below 2 ¥ 1015 cm–3. The third layer is a heavily
doped MWIR layer. The planar p-n junction is formed
by selective ion implantation with 75 As++. A three-
step post-implant anneal places the electrical junc-
tion on the narrow gap side of the junction and also
reduces the residual defects in the n-MCT layer. The
samples were characterized by Fourier transform
infrared spectroscopy (FTIR), x-ray rocking curves,
Hall and lifetime measurements to determine the
crystalline, optical and electrical properties.
FTIRmeasurementshavebeenperformedtodeter-
mine the thickness of the layers and the composition
of the absorber. The data were fit to a multi-layer
structure model8 that allows the determination of the
thicknessofeachlayer.Thedataforfoursuchsamples
are presented in Table I. The x-ray rocking curve data
were fitted to obtain the FWHM of the peaks. From
the fitting, the FWHM obtained was 194 arcsec for
CdTe, and 106 for HgCdTe. The latter value is close to
the FWHM of the substrate, and is reasonably good
for HgCdTe layers grown on CdTe/Si substrates indi-
cating good structural quality of the material.
Temperature dependent Hall effect measurements
were carried out after etching away the top CdTe and
widerbandgapHgCdTelayersandannealingat235∞C
for 12 hours for Hg vacancy filling. The measured
carrier concentration at 77 K in the first batch
EXPERIMENT
The structures, as shown in Fig. 1, were grown on
CdTe/Si(211) substrates by MBE. The MBE growth
chamber is equipped with an in-situ reflection high-

Velicu, Badano, Selamet, Grein, Faurie,
Sivananthan, Boieriu, Rafol, and Ashokan
718
Fig. 4. I-V characteristics of a 40 ¥ 40 mm2 test diode (lc = 3.2 mm) at
three different temperatures.
Fig. 5. I-V characteristics of three diodes (lc = 3.2 mm) of different area
at 80 K.
a
b
Fig. 6. Temperature dependence of the (a) carrier concentration and (b) mobility after annealing for an optimized batch of MWIR samples.
(HCT1798, HCT1799) was in the 1–4 ¥ 1016 range,
while the mobility was around 10000 cm2/V-sec, re-
flecting the fact that, at this depth, an average be-
tweenthemobilityofthelightlydopedandtheheavily
doped deeper layers is measured, as shown in Fig. 3.
Because of the high extrinsic dopant concentration in
the absorption layer, no carrier extraction was ob-
served in the processed devices. However, as we
mentioned before, a reasonably high R0A product is
obtained because of the low recombination velocity of
the n+n contact. Also, the measured detectivity (dis-
cussedlater)indicatesthefeasibilityofHgCdTebased
infrared detectors for near room temperature opera-
tion.
AutoCAD software was used to prepare the layout
of the masks needed for the fabrication of these
devices. An automatic pattern generator was used for
the fabrication of high quality chrome/iron oxide
masks. The devices were fabricated using a MJB-3
contact mask aligner, indium, titanium, gold metalli-
zation systems, and other common device fabrication
facilities. Inthefirstlithographystep, windowsforAs
implantation are opened. Thick photoresist was used
as the mask layer. After ion implantation (350 KeV
energy, 1014 cm–2 dose), the samples were subjected to
athree-steptemperatureannealundermercurysatu-
rated conditions for the activation of dopants and the
removal of Hg vacancies. The P-HgCdTe contacts
using a gold chloride solution and the n-type contacts
using In deposition were made. After that, Ti/Au
contact/bonding pads are deposited. The distance
between different diodes is larger than the minority
carrier diffusion length so we can assume indepen-
dent operation.
RESULTS AND DISCUSSION
I-V characteristics were measured in a low tem-
perature probe station with 300 K background. To
avoid any damage during testing, the pads were
placed away from the active area. The temperature
dependence of the I-V curves of these devices was
studied and the results are presented in Fig. 4 at

HgCdTe/CdTe/Si Infrared Photodetectors Grown
by MBE for Near-Room Temperature Operation
719
Table I. Thickness and Composition of the Layers from FTIR Data Fitting
HCT1799
HCT1810
HCT1811
HCT1812
x*
t**
0.5
5
x*
t**
0.8
9
x*
t**
x*
t**
SWIR cap
layer
MWIR absorber
layer
MWIR buffer
Layer
0.45
0.37
0.37
0.42
0.33
0.33
0.49
0.39
0.39
0.47
6.10
2.30
0.39
0.33
0.33
0.74
9.75
3.75
2
2
*—Cd comp; **—thickness (mm)
Table II. Hall Concentration and Minority Carrier Lifetime at 80 K
Carrier
Concentration
(cm–3)
Measured
Lifetime
(ms)
Theoretical
Mobility
Lifetime
Sample
(cm–2/V-sec)
(Rad+Auger)
Hct1810
Hct1811
Hct1812
9 ¥ 1014
2.5 ¥ 1014
6.9 ¥ 1014
17900
14290
24300
1.45
2.45
2.50
5.4
17.5
7.3
80 K, 150 K, and 200 K for a 40 ¥ 40 mm2 device. The
forward turn-on voltage decreases with increasing
temperature as expected from the band gap depen-
dence on temperature and the breakdown voltage
decreases to approximately 400 mV at 200K, in com-
parison with >500 mV at 80 K. For the 40 ¥ 40 mm
diodes the zero bias leakage current is about 1 pA and
the R0A is 1.1 ¥ 104 ohm-cm2 at 80 K. Tunneling
currents become dominant at voltages larger than
500 mV. We show in Fig. 5 the I-V characteristics for
diodes of different area. It can be seen that the device
qualityimproveswhenthediodeareaisdecreased.As
we have shown in a previous study,9 this is an indica-
tion that the bulk has an important contribution in
degradingdiodeperformance. Largeareadiodeshave
more probability to circumvent killer defects and
their performance reflects the quality of the bulk
material. The surface contribution to the dark cur-
rent is expected to further reduce by removing the
damaged CdTe cap layer after post-implant anneal-
ing and re-passivating the surface. In this batch of
devices we have used the original grown CdTe layer
for passivation. Substantial damage occurs in this
layerduetothehighenergyimplantationandanneal-
ing processes. The R0A product is consequently lower
than the best achieved in these type of devices, result-
ing in lower detectivities than expected theoretically.
Responsivity measurements were carried out at 80 K
and 300 K on the 40 ¥ 40 mm2 diode at the background
temperature of 300 K using a blackbody at 500 K, an
aperture size of 0.2 inch, distance between the black-
body and detector of 5 inch, and the chopper frequency at
1000 Hz. The measured responsivity was 1.3 ¥ 107 V/W
and hence the detectivity was 1.9 ¥ 1011 cm-Hz1/2/W at
80 K and 1 ¥ 1010 cm-Hz1/2/W at 300 K. The theoreti-
cally expected values10 are in the range of high 1011
and mid 1010 cm-Hz1/2/W at 80 K and 300 K, respec-
tively. The noise measurement was performed with
zerobiasvoltageand10Hzbandwidth.Themeasured
knee frequency, fk, in this device is 100 Hz at 80 K, a
consequence of the lower cut-off wavelength and fairly
good electrical properties. For a comparative device11 at
80 K with Is = 14 nA and R0A = 10 W-cm2, the fk is 4 Hz
and increases to 6 MHz for uncooled operation. For this
reason, even though the detectivity (3 ¥ 109 Jones) is
superior to pyroelectric devices, these devices could
not be used for imaging applications. The knee fre-
quency depends mainly on leakage current density
and passivation conditions. To reduce the fk and 1/f
noise further, improvements in both bulk and surface
properties are needed.
In the subsequent growth, doping level in the
low 1014 cm-3 and mobility in the mid 104 cm2/V-sec
were obtained, as shown in the Fig. 6, following
the same procedure described previously. This
indicates a continuous improvement in the growth
conditions. The mobility values are in the same
range as those measured on samples grown on
CdZnTe substrate. The composition and thickness
obtained from FTIR measurements for three
samples grown under the optimized condition
(HCT1810, HCT1811, and HCT1812) are presented
in Table I. Minority carrier lifetime measurements
were also performed on these samples using the
photoconductive decay technique at 80K, using a
pulsed GaAs diode laser with a 90 ns pulse width
and a fall time of 70 ns. Close to theoretical Auger
and radiative lifetimes were obtained as shown in
Table II, indicating low density of Shockley-Read
centers. However, the relative independence of
doping suggests that both band-to-band and some
form of Shockley-Read mechanism limit these life-
times. A summary of the Hall and lifetime data is
presented in the Table II. Exclusion and extrac-
tion are expected to occur in devices fabricated
from these layers.

Velicu, Badano, Selamet, Grein, Faurie,
Sivananthan, Boieriu, Rafol, and Ashokan
720
303 (1986).
CONCLUSIONS
2. T.N. Casselman, J. Appl. Phys. 52, 848 (1981).
3. C.T. Elliott, N.T. Gordon, R.S. Hall, T.J. Phillips, and A.M.
White, J. Electron. Mater. 38, 1139 (1996).
4. C.T. Elliott, N.T. Gordon, and A.M. White, Appl. Phys. Lett.
78, 2881 (1999).
5. W.E. Tennant and C. Cabelli, Mater. Res. Soc. Symp. Proc.
484, 221 (1998).
6. R.G. Humphreys, Infrared Phys. 23, 171 (1983).
7. W. van Roosbroeck and W. Shockley, Phys. Rev. 94, 1558
(1954).
8. M. Daraselia, C.H. Grein, S. Rujirawat, B. Yang, S.
Sivananthan,F.Aqariden,andH.D.Shih,J.Electron.Mater.
28, 743 (1999).
9. R. Ashokan, N.K. Dhar, B. Yang, A. Akhiyat, T.S. Lee, S.
Yusuf, and S. Sivananthan, J. Electron. Mater. 29, 636
(2000).
10. T.S.Lee,R.Ashokan,C.H.Grein,S.S.Yoo,and S.Sivanathan,
Proc. of SPIE 374,4028 (2000).
11. C.L.Jones,N.E.Metcalfe,A.Best,R.Catchpole,C.D.Maxey,
N.T. Gordon, R.S. Hall, T. Colin, and T. Skauli, J. Electron.
Mater. 27, 733 (1998).
Triple layer pnn+ devices were grown by MBE on Si
substrates. Electrical and optical characterization of
these devices demonstrates that HgCdTe has the
potential for higher temperature operation. The
measured responsivity 1.3 ¥ 107 V/W and the calcu-
lated detectivity 1.9 ¥ 1011 cm-Hz1/2/W of a 40 ¥ 40 mm2
diode at 80 K are promising results. A room tempera-
ture detectivity of 1010 cm-Hz1/2/W is the best reported
result for a HgCdTe/Si pnn+ structure, to the best of
our knowledge. A low donor doping concentration in
the 1014 cm–3 range for the HgCdTe absorber layer by
MBE is demonstrated for the first time and creates
the possibility for exclusion-extraction devices in the
near future.
REFERENCES
1. T. Ashley, C.T. Elliott, and A.T. Harker, Infrared Phys. 26,