OPTO−ELECTRONICS REVIEW 22(2), 118–126
DOI: 10.2478/s11772−014−0186−y
MOCVD grown MWIR HgCdTe detectors for high operation
temperature conditions
P. MARTYNIUK1, A. KOŹNIEWSKI2, A. KĘBŁOWSKI2, W. GAWRON1, and A. ROGALSKI1
1Institute
of Applied Physics, Military University of Technology, 2 Kaliskiego Str., 00–908 Warsaw, Poland
2Vigo System S.A., 129/133 Poznańska Str., 05–850 Ożarów Mazowiecki, Poland
The paper reports on photoelectrical performance of the mid−wave infrared HgCdTe detector for high operating tempera−
ture condition. Detector structure was simulated with APSYS numerical platform by Crosslight Inc. The comprehensive
analysis of the detector performance such as dark current, detectivity, time response vs. device architecture and applied bias
has been performed. The N+pP+n+ HgCdTe heterostructure photodiode operating in room temperature at a wavelength
range of 2.6–3.6 μm enabled to reach: detectivity ~ 8.7×1010 cmHz1/2/W, responsivity ~ 1.72 A/W and time response ~ 145 ps
(V = 200 mV).
Keywords: MWIR, SWIR, HgCdTe heterostructures, HOT detectors.
1. Introduction
The short−wave and mid−wave infrared photodetectors
(SWIR, MWIR) operating at high operating temperature
conditions (HOT) are important in a variety of applications
from earth resources, astronomy, advanced optical applica−
tions (1.31 μm and 1.55 μm), military, medical to include
battlefield tracking lasers (1.06 μm−laser guided munitions;
1.55 μm−laser range finders) and identifying precancerous
cellular changes (1.3 μm−optical coherence tomography) [1].
Imaging in SWIR band is important because of the “night
glow”−light emitted by the sky between 1–2 μm providing
sufficient illumination enabling passive imaging even under
moonless overcast conditions. SWIR detectors principally
respond to reflected light from objects rather than the ther−
mal emission from them. The distinctive applications are
mostly related to the reduced scattering effects associated
with long−wave (LWIR) detection processes [2].
In order to meet SWIR and MWIR applications a variety
of materials can be used such as: Si, Ge, PbS, InGaAs, and
HgCdTe [3]. Each material system has some relative advan−
tages and drawbacks. Si image sensors have demonstrated
sensitivity to wavelengths from 0.4 to 1.2 μm (l » 0.8 μm;
absorption coefficient, a » 1000 cm–1). Broad−band res−
ponse from 0.4 to 1.65 μm for Ge detectors has been mea−
sured enabling to reach a » 5000 cm–1 for l = 1.6 μm and
quantum efficiency (QE) from 40% to 75% [4]. Si1–xGex
alloys could be easily formed of any Ge−composition up to
100%, however to be used with longer wavelength respon−
se, the Ge composition must be sufficiently high to achieve
proper quantum efficiency. Formerly photoconducting de−
*e−mail:
118
[email protected]
tectors and currently photodiodes based on PbS colloidal
quantum dots on Si substrates were reported by Heves et al.,
exhibiting QE = 32%, responsivity, Ri = 6 A/W and detecti−
vity, D* = 1011–1012 cmHz1/2/W for l = 1.45 μm at 250 K [5].
Of these, InGaAs detectors have shown high device perfor−
mance for material whose composition is nearly matched to
InP (l » 0.9–1.7 μm) and proved to be the most practical for
In composition to x = 0.82, where the wavelength response
of InGaAs can be extended to 2.6 μm. Single element
InGaAs detectors have been made with up to 2.6 μm cut−
−offs, while arrays have been demonstrated to 2.2 μm, how−
ever their performance decreases rapidly at longer wa−
velengths due to mismatch with the InP substrate [6,7].
HgCdTe is considered as the most important semicon−
ductor alloy system for infrared application. Modern advan−
ces in metal organic chemical vapour deposition (MOCVD)
of HgCdTe have created the opportunity for realising novel
detector designs through multi−layer in situ growth with
complete flexibility in the choice of alloy compositions and
doping concentrations. These advances have led to high per−
formance in both HgCdTe single elements and two−dimen−
sional detector arrays for remote sensing of IR radiation in
two atmospheric windows, MWIR (3–5 μm) and LWIR
(8–12 μm) spectral regions. But HgCdTe IR detectors and
their properties for applications for lPeak » 3.6 μm, are less
well−studied. Additionally, the published papers addressing
characteristics of HgCdTe detectors, which were grown by
MOCVD on GaAs substrates, is rather limited in this wave−
length range. Currently, HgCdTe has typically focused on
large array size with good performance at somewhat longer
wavelengths, where III–V materials do not perform well
[8–11].
Opto−Electron. Rev., 22, no. 2, 2014
Unauthenticated
Download Date | 6/19/17 3:11 AM
Typically, most of HgCdTe infrared photodiodes are
based on heterostructures to prevent thermal generation at
contacts and improve detectivity and frequency response.
In the solution proposed by Ashley and Elliott, the N+pP+
architecture was used, in which extraction and exclusion of
thermally generated charge carriers from the active region
under reverse bias was applied [12,13]. Currently, complex
multi−layer structures have been used with great success for
MWIR and LWIR range operating at near room tempera−
tures [14]. The main modification in comparison with the
standard three−layer N+pP+ structure is programmed grading
of band gap and doping level at interfaces, where the trans−
port of majority and minority carriers is determined by bar−
riers and Auger generation−recombination (GR) suppression
[15,16].
In this paper we present theoretical modelling of the
N+pP+n+ HgCdTe photodetector based on epitaxial graded
gap structures, where a narrow−gap absorber is sandwiched
between wider gap minority and majority carrier contacts.
The main layers are interfaced with thin graded gap and
doping level transition layer (interfaces), that are introduced
with programmed growth. The voltage and structural depen−
dences of the dark current (JDARK), photocurrent (JPHOTO),
responsivity (Ri), detectivity (D*) and time response (t) are
analyzed including both band−to−band (BTB) and trap−assis−
ted (TAT) tunnelling processes at the heterojunctions. Part
of theoretical performance’s predictions are compared with
experimental data. Simulated N+pP+n+ HgCdTe hetero−
structure enabled to reach: D* ~ 8.7×1010 cmHz1/2/W, Ri ~
1.72 A/W and t ~ 145 ps (V = 200 mV, NA = 6×1016 cm–3) at
T = 300 K in the wavelength range of lPeak = 2.6–3.6 μm.
2. Simulation procedure and experimental
results
The nine layers heterostructure with the absorber composi−
tion of x = 0.28, its thickness of t = 3 μm, and p−type doping
(NA = 5×1016 cm–3) is shown in Fig. 1. N+ and P+ barriers
prevent from thermal generation at contact and should sup−
press Auger 7 GR by non−equilibrium conditions. P+ layer is
covered with heavily doped p+ layer to reduce contact resis−
tance. The p+−n+ junction bas been applied for extra impro−
vement of electrical contact between P+ region and metal−
lization. RSeries was artificially added to the detector’s struc−
ture in order to fit to experimental results related to time
response modelling.
The detector presented in this paper was fabricated in
a joint laboratory run by VIGO System and Military Univer−
sity of Technology (MUT). The HgCdTe multi−layer struc−
tures were grown on 2 ¢¢ inch semi−insulating, slightly dis−
oriented (100) GaAs substrates in a horizontal MOCVD
AIX 200 reactor. The interdiffused multilayer process
(IMP) technique was applied for the HgCdTe layer deposi−
tion. The detailed description of the implemented MOCVD
growth procedure is presented in Ref. 17.
Opto−Electron. Rev., 22, no. 2, 2014
Fig. 1. Simulated N+pP+n+ HgCdTe heterostructure. The layer
number, type of doping, composition grading, doping×1016 cm–3,
and thickness of the layers in μm are marked. Red arrow presents
composition grading.
It must be noted that HgCdTe ternary alloy, as a nar−
row−gap semiconductor exhibits a non−parabolic conduc−
tion band and high carrier degeneracy. These conditions are
very difficult to manage because of numerical problems
with computation of the Fermi−Dirac integral for non−para−
bolic model [18,19]. Quan et al. [20] and Wang et al. [21]
have proposed simple approximations to this expression,
however mentioned solution have only been validated for
T < 120 K. Since HgCdTe device modelled in this paper
operates at room temperatures and proposed approxima−
tions have not been fully validated for room temperature
conditions yet, the computations were performed using the
Fermi−Dirac statistics for a non−degenerate semiconductor
model with parabolic energy bands [20–23]. According to
Wenus et al. such simplification gives quite good results in
a broad range of doping concentrations [24].
Theoretical modelling of the HgCdTe heterostructures
has been performed by numerical solving of Poisson’s and
the electron/hole current continuity equations. The APSYS
platform (Crosslight Inc.) was implemented in our simula−
tion procedure. APSYS simulator uses the Newton−Richar−
dson numerical method of nonlinear iterations. The spe−
cific equations describing drift−diffusion (DD) model are
presented in detail in APSYS manual [25]. In the case of
ohmic contacts, simple Dirichlet boundary conditions are
applied. The electron and hole quasi−Fermi levels are equal
and set to the applied bias of that electrode, i.e., Efn = Efp = V.
The model incorporates both electrical and optical proper−
ties to include influence of radiative (RAD), Auger
(AUG), SRH GR at any location within the device and
BTB, as well as TAT tunnelling mechanisms at hetero−
junctions. After Casselman et al., we incorporated AUG
GR mechanisms using parabolic bands and non−degener−
ate statistics, which are considered to be suitable approxi−
mations for modelled device [26]. In TAT simulation, the
Hurkx et al. model was implemented [27]. APSYS plat−
form requires the input of HgCdTe material parameters
119
P. Martyniuk
Unauthenticated
Download Date | 6/19/17 3:11 AM
MOCVD grown MWIR HgCdTe detectors for high operation temperature conditions
(bandgap, electron affinity, dielectric constant, electron
and hole mobility, electron and hole effective mass, and
absorption coefficient), which were taken from published
models. In particular, the bandgap was obtained from
Hansen et al. [28]. The low−field electron mobility was
taken from the empirical formula based on Scott’s paper,
while hole mobility was basically taken as 0.01 of the elec−
tron mobility [29]. The absorption was only assumed in
active layer region and the absorption coefficient (a) was
estimated according to Kane model including its composi−
tion, doping and temperature dependence (a = 4710 cm–1,
l = 3.3 μm, T = 300 K) [30]. Analysis of high frequency
behaviour of a semiconductor device was performed using
Li et al. model [31]. The noise current was simulated using
the expression including thermal Johnson−Nyquist
component and electrical shot contributions
in (V ) = ( 4k BT / RA + 2qJ DARK ) A,
sent the real structure which profile is shaped by interdiffu−
sion processes during HgCdTe growth. The doping profiles
were simulated by applying gauss tail model, where doping
concentration falls off with a gaussian tail on the edges of
the polygon, dx; (see Table 1).
The secondary ion mass spectroscopy (SIMS) measure−
ments were used to verify the composition and doping pro−
files of the simulated structure at the beginning stage of the
simulations. SIMS profile shown in Fig. 2 fully confirms
implemented simulation procedure. Nominal pre−growth
absorber composition was assumed at the level of x = 0.33
corresponding to lc = 3.75 μm at T = 300 K, while SIMS
measurements showed x = 0.34. n+ and interface 3 layers
[In. 3−Fig. 2(a–c)] exhibit the biggest discrepancy, where
significant difference by nearly two orders of magnitude
between nominal and measured concentrations were ob−
served for n−type doping. Similar conditions are visible for
N+−contact layers [refer to Fig. 2(b)], while for p−type dop−
ing, assumed pre−growth nominal concentrations are much
higher in comparison with measured value in absorber re−
gion [refer to Fig. 2(c)] and these values were picked to fit to
the experimental results. Interface 3 grading composition
was nominally assumed to be within the range 0–0.25 in
order to create proper contact to P+−layer and facilitate car−
rier transport to n+−layer, while measured value indicates
grading within 0.14–0.22.
The composition of the active layer was also determined
from room temperature spectroscopic measurements. Figure
3 shows the transmittance spectrum of N+pP+n+ HgCdTe
heterostructure used in this study. It can be found that the
cut−off wavelength is about lc » 4.42–5.34 μm at room tem−
perature. The absorber composition of HgCdTe is about x »
0.257–0.293 depending on assumed absorber width, which
was calculated from the tangent method (l -c 1 » 1941 cm–1)
proposed by Finkmann and Schacham [32]. Presented
results fully confirm the idea of assuming lower active layer
composition in comparison with SIMS measurements and
nominal pre−growth compositions.
(1)
where A is the detector’s area, RA dynamic resistance area
product, JDARK is the dark current density, respectively, and
kB is the Boltzmann constant.
The quantum efficiency is a function of the incident
radiation wavelength and current responsivity, Ri, (without
electro−optical gain), while detectivity is defined by the fol−
lowing expressions
h ( l) = 1.24
Ri
,
l
D* =
Ri
n2 A .
in (V )
(2)
It must be stressed that the performance of generation−
−recombination noise limited photodetectors may be effecti−
vely increased by limitation of active volume of the detector
which is achieved by optical immersion to a high refractive
index (n) immersion lens.
Table 1 shows parameters taken in modelling of MWIR
N+pP+n+ HgCdTe heterostructure photodiode. Three inter−
face layers were assumed to be x−graded regions and repre−
Table 1. Parameters taken in modelling of MWIR N+pP+n+ HgCdTe heterostructure photodiode.
NA, ND (cm–3)
Gauss tail, dx (μm)
1
2
3
4
5
6
7
8
9
N+
n+
n
p
p+
P+
p+
n+
n+
1017–1018
1017
5×1017
5×1017
2×1017
5×1017
1017–1018
0.05
0.1
0.05
0.05
0.05
0.05
0.05
Composition, x
0.46
Geometry, t (μm)
11.79
5.8×1016 5×1016
0.1
0.46®0.4 0.4®0.28
0.62
0.61
0.1
0.28 0.28®0.39 0.39 0.39®0.14 0.14®0.22 0.22®0.23
3.06
0.63
Diameter, d (μm)
260
Overlap matrix, F1F2
0.15
Trap energy level, ETrap
0.33×Eg
Trap concentration, NTrap (cm–3)
s
Capture cross section SRH n (cm–2)
sp
8×1014
0.82
0.54
0.82
5 ×10 -15
2.5 ×10 -15
Incident power density, F (W/cm2)
120
1.86
400
Opto−Electron. Rev., 22, no. 2, 2014
© 2014 SEP, Warsaw
Unauthenticated
Download Date | 6/19/17 3:11 AM
Fig. 3. Infrared transmission spectra of N+pP+n+ HgCdTe hetero−
structure at T = 300 K. A tangent (blue dashed line) is drawn to the
nearly linear part of the transmission characteristics to obtain the
zero−intercept cut−off l–1.
The measured dark current and differential resistance
area product, RA, vs. voltage at T = 300 K are presented
in Fig. 4. In the range of high voltages, JDARK is mostly
driven by TAT and BTB tunnelling at interface 1 and inter−
face 3. The maximum RA » 1.8 Wcm2 is estimated at
V = 600 mV. The detector’s electrical area is estimated as
A = 0.053 cm2.
The measured spectral response and detectivity charac−
teristics for V = 0.1 mV and T = 300 K are presented in
Fig. 5. The 50% cut−off wavelength assumes lc = 3.65 μm at
T = 300 K (x = 0.28). Both Ri and D* keeps almost constant
within range of l = 2.6–3.6 μm. The maximum responsivity
of Ri » 1.72 A/W is estimated for l = 3.3 μm, while maxi−
mum D* with optical immersion lens (D* is proportional to
~ n2 ~ 10 for GaAs substrate; n−refractive index) is esti−
mated ~ 8.7×1010 cmHz1/2/W. Field of view (FOV) was
assumed to be 36°.
Fig. 2. SIMS profile of the N+pP+n+ HgCdTe heterostructure: com−
position, xCd (a); donor concentration, ND (b); acceptor concentra−
tion, NA (c).
Opto−Electron. Rev., 22, no. 2, 2014
Fig. 4. JDARK and RA vs. volatge of MWIR N+pP+n+ HgCdTe
heterostructure.
121
P. Martyniuk
Unauthenticated
Download Date | 6/19/17 3:11 AM
MOCVD grown MWIR HgCdTe detectors for high operation temperature conditions
Fig. 5. Ri and D* vs. l of MWIR N+pP+n+ HgCdTe heterostructure.
3. Results and discussion
3.1. Band diagrams
The modelled energy band diagram profile of the MWIR
structure is presented in Fig. 6. The calculations were per−
formed for HOT conditions, T = 300 K, for zero and the
reverse bias voltage polarization (V = 1200 mV). Assuming
abrupt heterojunctions in simulated N+pP+n+ heterostruc−
ture, the discontinuities of both conduction and valence
bands may be visible. This could have adverse effect on
device properties contributing to spikes in charge carrier
concentration, thermal generation rates and electric fields.
The consequence of this effect is a large JDARK due to the
thermal generation and tunnelling mechanisms. As men−
tioned, the possible efficient solution to circumvent this
problem is programmed grading of band gap and doping
level at heterojunctions (interfaces 1–3).
Under reverse voltage polarization, the electrons are ex−
tracted from the absorber region by positive electrode con−
nected to N+−contact layer. The electrons are also excluded
from the absorber because they cannot be injected from neg−
ative electrode into P+−barrier layer. The energy barrier bet−
ween n+ and P+ regions enhances exclusion of electrons
from the absorber region and, as a consequence, they cannot
be replenished due to the low concentration of the electrons
in P+−barrier region. Applying reverse bias, it is necessary to
have N+−p (extraction) and p−P+ (exclusion) heterojunctions
close enough to keep the carrier concentration below intrin−
sic level.
It must be noted that interface 3 (n+−P+ heterojunction) is
forwardly biased and assuming x = 0.14, the conduction and
valence band should coincide (bandgap enrgy, Eg = 0.07 eV)
enabling tunnelling mechanisms in this region, which in
turn should improve detector’s frequency response. Electric
field drops mostly on interface 1, shaping performance of
N+pP+n+ structure.
3.2. Dark current photocurrent and responsivity
Fig. 6. Energy band diagram of MWIR N+pP+n+ HgCdTe hetero−
structure: (a) at equilibrium; (b) at reverse bias V = 1200 mV.
122
The very first step in simulation procedure is to fit JDARK
measured results. It was found that we were forced to lower
active layer composition by nearly x = 0.05 to fit to JDARK
and Ri characteristics in comparison to nominal pre−growth
value. The fitting procedure assumed active layer doping at
the level of NA = 5×1016 cm–3 and composition x = 0.28. The
simulation includes RAD, AUG, SRH and TAT/BTB con−
tributions at graded heterojunctions. The trap density, NTrap,
and trap energy level, ETrap, were assumed at NTrap = 8×1014
cm–3 and ETrap = 0.33×Eg, respectively (counted from con−
duction band). The series resistance influences the slope of
the JDARK –V characteristics at low voltage region < 120 mV
and was found to be at the level of RSeries = 50 W. The cap−
ture cross sections are assumed at level sn = 5×10–15 cm–2,
sp = 2.5×10–15 cm–2, which corresponds to the following
SRH carrier lifetimes in active region: tn = 3.5 ns and
tp = 35 ns, respectively. Since electric field drops mostly on
interface 1, its contribution is believed to be decisive as for
as photoelectrical performance is concerned.
In the low range of bias < 120 mV, JDARK is fitted by
SRH GR mechanism from active region. JDARK with no TAT
and BTB influence is presented in Fig. 7(a) (refer to pink
dashed line). At higher reverse polarization (> 120 V), TAT
Opto−Electron. Rev., 22, no. 2, 2014
© 2014 SEP, Warsaw
Unauthenticated
Download Date | 6/19/17 3:11 AM
Fig. 8. Simulated and measured Ri vs. l for MWIR N+pP+n+
HgCdTe heterostructure for selected temperatures.
Fig. 7. Simulated and measured JDARK of MWIR N+pP+n+ HgCdTe
structure vs. voltage for selected temperatures (a) and for selected
absorber’s doping (b).
mechanism dominates, which corresponds to RA–V charac−
teristic presented in Fig. 4. Above > 700 mV BTB mecha−
nism starts dominating leading to decreasing of RA. Proper
agreement to experimental results is reached up to 1100 mV.
JDARK vs. voltage for selected absorber’s doping is presented
in Fig 7(b). For NA = 5×1015 cm–3, JDARK –V characteristics
is SRH driven in an analysed voltage range, while for
5×1017 cm–3, up to 400 mV lower dark currents could be
achieved. Negative differential resistance region is not ob−
served in JDARK –V characteristics, what indicates that car−
rier concentration in absorber region is not reduced below
intrinsic concentration level (ni = 5×1015 cm–3).
Theoretically predicted and measured spectral response
characteristics of N+pP+n+ HgCdTe heterostructure operat−
ing at 230 and 300 K at reverse bias voltage V = 0.1 mV and
l = 3.3 μm and incident photon flux of 400 W/cm2 are pre−
sented in Fig 8. It is believed that discrepancy between sim−
ulation and measured results for l > 3.5 μm comes from an
assumption that photogeneration process takes place only in
active region (layer 4). Absorption coefficient vs. l is also
Opto−Electron. Rev., 22, no. 2, 2014
presented in Fig. 8 (pink dashed line). JPHOTO vs. voltage
and Ri vs. l for selected doping of p−type absorber are pre−
sented in Fig. 9. The higher doping, the lower Ri may be
reached in low range of bias.
Figure 10(a–c) presents JDARK –V and Ri–l characteris−
tics for selected N+−, n+−, P+−layer doping. Once doping of
N+−layer changes within the range ND = 5×1016–1017 cm–3,
JDARK increases for V > 200 mV, while below this value,
JDARK is found to be independent on ND. The higher electric
field drops on interface 1, which increases both TAT and
BTB contribution to the net current. Ri is not dependent on
N+−doping within analyzed range. It must be noted that,
when N+−contact layer composition becomes comparable to
p−type absorber, the extraction junction capability deterio−
rates, which in turn leads to dark current increase. n+−con−
tact doping mostly influences Ri for l < 3.7 μm, while JDARK
keeps nearly constant.
At the higher doping of P+−layer the lower JDARK current
could be reached. Doping of P+−layer influences Ri in wave−
length range l < 3.6 μm. Influence of P+−barrier composi−
Fig. 9. JPHOTO and Ri for MWIR N+pP+n+ HgCdTe heterostructure
vs. voltage and wavelength for selected p−type absorber doping.
123
P. Martyniuk
Unauthenticated
Download Date | 6/19/17 3:11 AM
MOCVD grown MWIR HgCdTe detectors for high operation temperature conditions
tion was not considered, but we believe that JDARK increases
sharply when P+−composition becomes comparable to the
both p−type absorber and n+−layer compositions, respec−
tively (exclusion junction capability worsens). The doping
of barrier has no influence on JDARK for V < 75 mV.
Fig. 10. JDARK and Ri of MWIR N+pP+n+ HgCdTe heterostructure
vs. voltage and wavelength for selected N+ (a); n+ (b); and P+ (c)
layer doping.
124
3.3. Time response and detectivity
Except detectivity, many SWIR and MWIR fibre communi−
cation applications require fast response devices (even in
picoseconds range). These two parameters stay in contradic−
tion at HOT conditions in terms of detector’s optimization.
Typically, the response time is determined by drift and dif−
fusion of photogenerated charge carriers to the contact re−
gion for higher voltage condition, while for weak reverse
bias recombination decay plays dominant role. Assuming
that, the depletion region occupies only small part of active
region and absorption occurs in neutral region of absorber,
the response time is conditioned by recombination decay
and diffusion of photogenerated carriers to the contacts. For
absorber thickness comparable or larger than diffusion
length, the response time is limited by the recombination
time. Typically, the AUG 7 HgCdTe GR times are quite
short in MWIR region, reaching 0.5 ns for absorber’s dop−
ing NA = 1018 cm–3. In considered detector structure, the
absorber’s AUG 7 carrier lifetimes are estimated at the level
of tA7 » 0.3 μs (x = 0.28, NA = 5×1016 cm–3), while tSRH » 3 ns
(NTrap = 8×1014 cm–3; sn = 5×10–15 cm–2; sp =2.5×10–15 cm–2),
which means that for unbiased and low voltage, high fre−
quency response (in terms of absorber layer and thermal GR
mechanisms) is SRH driven.
The ambipolar diffusion is considered to be the most
important mechanism in devices optimized for low voltage
conditions. In p−type absorber region the ambipolar diffu−
sion is conditioned by electron diffusion coefficient and its
characteristic time is found at the level of ~ 0.4 ns. Fre−
quency response could be improved by drift transport,
where drift time is estimated to be ~ 50 ps for 500 mV.
In the case of lower absorber thicknesses, the both ambi−
polar diffusion and drift time improves.
In theoretical calculations we assumed 25 ps pulses and
l = 3.3 μm, respectively. Time response is estimated from
exponential decay of the photocurrent versus time. The
reverse bias in the range of V = 0.1 to 1000 mV applied to
the considered structure, reduces time response by nearly
two times for analyzed RSeries. This behaviour is directly
connected with rapid improve of the drift time for higher
voltage. The experimental results presented in Fig. 11 are
fitted theoretically by assuming extra RSeries = 50 W attached
to the N+pP+n+ structure. This points out, that RSeries,
believed to be coming from the detector’s processing, have
not been overcome problem yet. Theoretical estimates for
no RSeries influence are more than one order of magnitude
better in comparison with the presented experimental
results.
The time response is also calculated vs. p−type doping of
absorber layer reaching t = 145 ps for NA = 6×1016 cm–3. For
V = 200 mV and T = 300 K the most decisive is doping of
active layer, while both contact n+– (ND > = 4×1016 cm–3)
and N+– (ND > = 6×1016 cm–3) layers have no influence on
time response with wide range of analyzed n−type doping.
The results are presented in Fig. 12.
Opto−Electron. Rev., 22, no. 2, 2014
© 2014 SEP, Warsaw
Unauthenticated
Download Date | 6/19/17 3:11 AM
Fig. 11. Simulated and measured response time versus voltage
for MWIR N+pP +n+ HgCdTe heterostructure for selected
RSeries.
Figure 13 shows measured and modelled D* vs. l. Simi−
larly to Ri–l, D* characteristics exhibit similar discrepancy
for l > 3.5 μm. Incorporation of immersion (GaAs
substarte) enables to reach D* ~ 8.7×1010 cmHz1/2/W at
T = 300 K.
The very last figure (Fig. 14) compares the D* (1 μm < l
< 4 μm, FOV = 36°) of selected SWIR and MWIR IR photo−
voltaic detector technologies including InGaAs with lPeak =
1.55 μm; HgCdTe with lPeak = 2 μm; InAs with lPeak = 3.25
μm and theoretically estimated N+pP+n+ HgCdTe (with
optical immersion) photodiode with lPeak = 3.3 μm. SWIR
HgCdTe photovoltaic detectors without optical immersion
are sub−background limited (BLIP) devices with perfor−
mance close to the generation−recombination limit. Situa−
tion is less favourable within range 2.6–3.6 mm for optically
immersed N+pP+n+ HgCdTe structure, where detectivity is
below the BLIP limit.
Fig. 12. Simulated response time of MWIR N + pP + n +
HgCdTe heterostructure vs. doping of the N+−layer, absorber and
n+−layer.
Opto−Electron. Rev., 22, no. 2, 2014
Fig. 13. Theoretically predicted and measured D* vs. l for MWIR
N+pP+n+ HgCdTe photodiode operating at room temperature.
4. Conclusions
In the paper we estimated and compared to the experimental
results, the performance of the MWIR HgCdTe N+pP+n+
heterostructure designed for HOT conditions. It is observed
that electron concentration in absorber region is not below the
intrinsic level and negative differential resistance region is not
visible. Dark current is analysed in detail as a function of the
structural parameters, in particular: N+−, n+−contacts and P+−
−barrier layers, respectively. The optimized structural parame−
ters are presented for selected voltages and doping range.
The theoretically predicted time response of the devices
are in the range of ~ 145 ps (with no RSeries influence). Using
GaAs immersion lenses, detectivity enables to reach
D* ~ 8.7×1010 cmHz1/2/W at room temperature for devices
with cut−off wavelength of 4.7 μm. Detectivity of optically
immersed MWIR N+pP+n+ HgCdTe structure operating at
T = 300 K, is below the BLIP limit.
Fig. 14. Comparison of room temperature spectral detectivities of
photovoltaic SWIR and MWIR detectors: N+pP+n+ HgCdTe, InAs
and InGaAs. FOV = 36°.
125
P. Martyniuk
Unauthenticated
Download Date | 6/19/17 3:11 AM
MOCVD grown MWIR HgCdTe detectors for high operation temperature conditions
Acknowledgements
This paper has been done under the financial support of the
Polish National Science Centre−the grant no. DEC 2011/01/
B/ST5/06283. We also acknowledge the support by Natio−
nal Centre of Research and Development−the grant no. PBS
1/B5/2/2012. Piotr Martyniuk wishes to thank Professor
James Harris and Tomas Sarmiento of Department of Elec−
trical Engineering, Stanford University for help given in his
visit to MBE Laboratory within confines of the Polish 500
Top Innovators Program, 2013.
References
1. M.H. Ettenberg and D. Malchow, “InGaAs SWIR detectors –
from military to medical applications”, www.photonicson−
line.com.
2. M.P. Hansen and D.S. Malchow, “Overwiew of SWIR detec−
tors, cameras, and applications”, Proc. SPIE 6939, 693901
(2008).
3. A. Rogalski, Infrared Detectors, CRC Press, Boca Raton,
2011.
4. C. Rafferty, C. King, B. Ackland, J. O’Neil, I. Aberg, T.S.
Sriram, A. Mackay, and R. Johnson, “Monolithic germanium
SWIR imaging array”, Proc. SPIE 6940, 69400N (2008).
5. E. Heves, C. Oztruk, V. Ozguz, and Y. Gurbuz, “Solution−
−based PbS photodiodes, integrable on ROIC, for SWIR detec−
tor applications”, IEEE Electron Device Lett. 34, 5 (2013).
6. A. Rogalski and R. Ciupa, “InGaAs vs. HgCdTe for short
wavelength infrared applications”, Proc. SPIE 3629,
0277–786X1 (1999).
7. J. Kaniewski and J. Piotrowski, “InGaAs for infrared photo−
detectors. Physics and technology”, Opto−Electron. Rev. 12,
139–148 (2004).
8. P. Norton, “HgCdTe infrared detectors”, Opto−Electron.
Rev. 10, 159–174 (2002).
9. J.−S. Kim, S.−Y. An, S.−H. Suh, “Characteristics of SWIR di−
odes of HgCdTe/CdTe/GaAs grown by metal organic vapor
phase epitaxy”, Phys. Stat. Sol. B229, 1089–1092 (2002).
10. A. Rogalski, “HgCdTe infrared detector material: history,
status and outlook”, Rep. Prog. Phys. 68, 2267–2336 (2005).
11. L.O. Bubulac, W.E. Tennant, J.G. Pasko, L.J. Kozlowski, M.
Zandian, M.E. Motamedi, R.E. DeWames, J. Bajaj, N.
Nayar, W.V. McLevige, N.S. Gluck, R. Melendes, and D.E.
Cooper, “High performance SWIR HgCdTe detector arrays”,
J. Electron. Mater. 26, 649–655 (1997).
12. T. Ashley and C.T. Elliott, “Non−equilibrium mode of opera−
tion for infrared detection”, Electron. Lett. 21, 451–452 (1985).
13. C.T. Elliott, “Non−equilibrium mode of operation of nar−
row−gap semiconductor devices”, Semicond. Sci. Technol. 5,
S30–S37 (1990).
14. J. Piotrowski and A. Rogalski, High−Operating Temperature
Infrared Photodetectors, SPIE Press, Bellingham, 2007.
15. C. T. Elliot, C. T. Gordon, R. S. Hall, T. J. Philips, A. M.
White, C. L. Jones, C. D. Maxey, N. E. Metcalfe, “Recent re−
sults on MOVPE grown heterostructure devices”, J. Elec−
tron. Mater. 25, 1139–1145 (1996).
126
16. J. Piotrowski, W. Gawron, Z. Orman, J. Pawluczyk, K. Kłos,
D. Stępień, and A. Piotrowski, “Dark currents, responsivity
and response time in graded gap HgCdTe structures”, Proc.
SPIE 7660, 766031 (2010).
17. A. Piotrowski, P. Madejczyk, W. Gawron, K. Kłos, J. Paw−
luczyk, J. Rutkowski, J. Piotrowski, and A. Rogalski, “Prog−
ress in MOCVD growth of HgCdTe heterostructures for unco−
oled infrared photodetectors”, Infrared Phys. & Technol. 49,
173–182 (2007).
18. R.K. Bhan and V. Dhar, “Carrier density approximation for
non−parabolic and highly degenerate HgCdTe semiconduc−
tors”, Semicond. Sci. Technol. 19, 413–416 (2003).
19. S. Gupta, R.K. Bhan, and V. Dhar, “Unified carrier density
approximation for non−parabolic and highly degenerate
HgCdTe semiconductors covering SWIR, MWIR and LWIR
bands”, Infrared Phys. & Technol. 51, 259–262 (2008).
20. Z.J. Quan, G.B. Chen, L.Z. Sun, Z.H. Ye, Z.F. Li, and W. Lu,
“Effects of carrier degeneracy and conduction band non−
−parabolicity on the simulation of HgCdTe photovoltaic de−
vices”, Infrared Phys. & Technol. 50, 1–8 (2007).
21. J. Wang, X.S. Chen, Z.Q. Wang, W.D. Hu, W. Lu, and F.Q.
Xu, “The mechanism of the photoresponse blueshifts for the
n−type conversion region of n+−on−p Hg0.722Cd0.278Te infra−
red photodiode”, J. Appl. Phys. 107, 044513 (2010).
22. J. Wang, X. Chen, W. Hu, L. Wang, Y. Chen, W. Lu, and F.
Xu, “Different approximation for carrier statistic in non−par−
abolic MWIR HgCdTe photovoltaic devices”, Proc. SPIE
8012, 80123B (2011).
23. Z.J. Quan, X.S. Chen, W.D. Hu, Z.H. Ye, X.N. Hu, Z.F. Li,
and W. Lu, “Modelling of dark characteristics for long−wa−
velength HgCdTe photodiode”, Opt. Quant Electron. 38,
1107–1113 (2007).
24. J. Wenus, J. Rutkowski, and A. Rogalski, “Two−dimensional
analysis of double−layer heterojunction HgCdTe Photodio−
des”, IEEE Trans. Electron. Devices 48, 7 (2001).
25. APSYS Macro/User’s Manual ver. 2011. Crosslight Soft−
ware, Inc. (2011).
26. T.N. Casselman and P.E. Petersen, “A comparison of the
dominant Auger transitions in p−type (HgCd)Te”, Solid State
Commun. 33, 615–619 (1980).
27. G.A. Hurkx, D.B. M. Klaassen, and M.P.G. Knuvers,
“A new recombination model for device simulation includ−
ing tunnelling”, IEEE Trans. Electron. Devices 39, 2 (1992).
28. G.L. Hansen, J.L. Schmidt, and T.N. Casselman, “Energy
gap vs. alloy composition and temperature in Hg1−xCdxTe”,
J. Appl.Phys. 53, 7099 (1982).
29. W. Scott, “Electron mobility in Hg1−xCdxTe”, J. Appl. Phys.
43, 1055 (1972).
30. W.W. Anderson, “Absorption constant of Pb1−xSnxTe and
Hg1−xCdxTe alloys”, Infrared Phys. & Technol. 20, 363
(1980).
31. Q. Li and R.W. Dutton, “Numerical small−signal AC model−
ing of deep−level−trap related frequency−dependent output
conductance and capacitance for GaAs MESFET’s on semi−
−insulating substrates”, IEEE Trans. Electron. Devices 38,
1285–1288 (1991).
32. E. Finkman and S.E. Schacham, “The exponential optical ab−
sorption band tail of Hg1−xCdxTe”, J. Appl. Phys. 56, 10
(1984).
Opto−Electron. Rev., 22, no. 2, 2014
© 2014 SEP, Warsaw
Unauthenticated
Download Date | 6/19/17 3:11 AM