type ii superlattice technology for lwir detectors · 2, the al and bl t2sls are doped p-type which...
TRANSCRIPT
Invited Talk at Infrared Technology and Applications XLII (Baltimore, 2016)
Type II superlattice technology for LWIR detectors
P.C. Klipstein, E. Avnon, D. Azulai, Y. Benny, R. Fraenkel, A. Glozman, E. Hojman, O. Klin,
L. Krasovitsky, L. Langof, I. Lukomsky, M. Nitzani, I. Shtrichman,
N. Rappaport, N. Snapi, E. Weiss and A. Tuito†.
SemiConductor Devices P.O. Box 2250, Haifa 31021, Israel
†Israel MOD
ABSTRACT
SCD has developed a range of advanced infrared detectors based on III-V semiconductor
heterostructures grown on GaSb. The XBn/XBp family of barrier detectors enables diffusion limited
dark currents, comparable with MCT Rule-07, and high quantum efficiencies. This work describes
some of the technical challenges that were overcome, and the ultimate performance that was finally
achieved, for SCD’s new 15 m pitch “Pelican-D LW” type II superlattice (T2SL) XBp array detector.
This detector is the first of SCD’s line of high performance two dimensional arrays working in the
LWIR spectral range, and was designed with a ~9.3 micron cut-off wavelength and a format of 640
512 pixels. It contains InAs/GaSb and InAs/AlSb T2SLs, engineered using k p modeling of the
energy bands and photo-response. The wafers are grown by molecular beam epitaxy and are fabricated
into Focal Plane Array (FPA) detectors using standard FPA processes, including wet and dry etching,
indium bump hybridization, under-fill, and back-side polishing. The FPA has a quantum efficiency of
nearly 50%, and operates at 77 K and F/2.7 with background limited performance. The pixel
operability of the FPA is above 99% and it exhibits a stable residual non uniformity (RNU) of better
than 0.04% of the dynamic range. The FPA uses a new digital read-out integrated circuit (ROIC), and
the complete detector closely follows the interfaces of SCD’s MWIR Pelican-D detector. The Pelican-
D LW detector is now in the final stages of qualification and transfer to production, with first
prototypes already integrated into new electro-optical systems.
Keywords: Infrared Detector, Focal Plane Array, Type II superlattice, XBn, XBp, pBp, LWIR.
1. INTRODUCTION
In a pioneering article1 entitled “a new semiconductor superlattice” published in 1977, Sai-Halasz, Tsu and Esaki
proposed that a superlattice based on alternating layers of InAs and GaSb would exhibit rather unique properties,
including a zero bandgap at a critical value of the layer thicknesses. In this respect, the new superlattice bore a close
relationship with the alloy, HgxCd1-xTe, where the bandgap vanishes at a critical value of the composition parameter, x
(See Figure 1). HgxCd1-xTe has become one of the most widely used tunable infrared detector materials, because it is a
versatile technology that can match the characteristic photon wavelength of most infrared applications. On the other
hand, InAs/GaSb superlattices, also known as type II superlattices (T2SLs), have only recently been considered to be a
2
viable alternative technology to HgxCd1-xTe, due to the more challenging crystal growth that is required. T2SLs must be
grown by molecular beam epitaxy (MBE), which has taken longer to mature to a reliable commercial production tool
than the liquid phase epitaxy technique that is used most widely in the production of HgxCd1-xTe detector arrays. An
important factor driving T2SL development, however, is that commercial 3” or 4” GaSb substrates can be used for their
growth, while smaller and more expensive CdZnTe substrates must be used for the highest quality HgxCd1-xTe.
Yet two more technological hurdles still had to be overcome, before T2SL technology began to look truly
competitive. The first was the suppression of generation-recombination (G-R) limited dark current2, which is absent in
high quality HgxCd1-xTe photodiodes, and the second was device passivation. A novel barrier device, known as an XBp
detector,3,4
was developed at SCD, which contains two types of T2SL: InAs/GaSb for the photon absorbing or “active”
layer (AL) and the contact layer (CL), and InAs/AlSb for the barrier
layer (BL). As shown in the “pBpp” version of this device in Figure
2, the AL and BL T2SLs are doped p-type which ensures that there
is no depletion layer in the narrow bandgap AL5. The figure shows
how G-R current is generated in the depleted BL through mid-gap
Shockley-Read-Hall (SRH) traps with an activation energy close to
half of its bandgap. Since this bandgap is more than twice that of the
AL, the G-R current is negligible compared to the diffusion current
coming out of the AL. For this current, the activation energy is equal
to the full AL bandgap, since all SRH traps are occupied5. In this
work it will be shown that the dark current in a T2SL XBp device is
quite close to the Rule 07 value, which is the state of the art metric
for HgxCd1-xTe photodiodes6.
The second technological hurdle that had to be overcome was a
reliable passivation treatment of the processed XBp devices, which
is compatible with all of the standard FPA processes, including wet
and dry etching, indium bump hybridization, glue under-fill, and
back-side polishing. The development of this passivation at SCD is
the final step that has enabled the fabrication of long wave infrared
(LWIR) T2SL FPAs with a performance comparable to high quality
HgxCd1-xTe FPAs. The design and performance of these FPAs is the
subject of this paper.
In section 2, the design and performance of T2SL barrier
devices is discussed, with particular emphasis on device simulation
and surface passivation. The next section introduces the key
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1
Ener
gy (
eV)
x
Hg1-xCdxTe
S
P
HgTe CdTe
Hg1-xCdxTe
S
P
InAs
GaSb
Long period T2SL Short period T2SL
(a) (b)
Figure 1
The zero bandgap state occurs when the bands with s- and p-like orbital symmetry cross at a critical value of (a) the composition
parameter, x, in Hg1-xCdxTe and (b) the InAs and GaSb layer thicknesses in a type II superlattice
InAs/GaSb
InAs/GaSb
InAs/AlSb
Figure 2
The profiles of the conduction and valence band
edges in a T2SL pBpp device. The AL (left) and CL
(right) are made from an InAs/GaSb T2SL and the
BL, from an InAs/AlSb T2SL. The broken wavy
arrows represent thermal generation processes in the
AL and BL, and the asterisk, a mid-gap SRH trap.
3
parameters of SCD’s new 15 μm pitch, 640 512 read-out integrated circuit (ROIC) designed to operate with XBp
devices (equivalent polarity to n-on-p photo-diodes). The radiometric characteristics of the new FPA are presented in
section 4, using data from over 200 FPAs from our new Pelican-D LW production line. The conclusions are
summarized in section 5.
2. DESIGN AND PERFORMANCE OF T2SL BARRIER
DEVICES
The T2SL structures used in this work were grown in-house in
a Veeco GenIII MBE system, on 3 inch “epi-ready” GaSb(100)
substrates. The InAs/GaSb (AL and CL) and the InAs/AlSb (BL)
superlattices had “InSb-like” interfaces in order to achieve a good
lattice match with the substrate. Further growth details may be
found in Ref. 7. The good reproducibility of the MBE growths
enabled optimizations of both the device structure and the wafer
process, which are further elucidated below.
In order to achieve a smooth conduction band profile but a
large barrier in the valence band as depicted in Figure 2, a method
of band structure and optical field simulation was developed at
SCD, based on the k p theory and optical transfer matrix (OTM)
techniques. The band structure simulation introduces a number of
novel features including the use of an interface matrix, and a way of
calculating the Luttinger parameters from standard reference
values.8, 9
Figure 3(a) shows a plot of the bandgap energy as
measured from the position of the photoluminescence (PL) peak at
10K vs. the calculated bandgap for more than 30 T2SL samples.
The calculated value was based on InAs and GaSb layer widths
100
140
180
220
260
300
100 140 180 220 260 300
Calc
ula
ted
Ban
dga
p (
me
V)
PL Peak Energy (meV)
InAs/GaSb
340
390
440
490
540
340 380 420 460 500 540
PL Peak Energy (meV)
Calc
ula
ted B
andgap (
meV
)
3.5 3.0 2.5 (m)
InAs/AlSb
PL Peak Energy (meV)
Ca
lcu
late
d B
an
dg
ap
(m
eV
)
(a) (b) Figure 3
Comparison between calculated band gaps and PL peak energies measured at 10K for (a) more than 30 InAs/GaSb T2SLs spanning
the MWIR to LWIR wavelength range, and (b) 14 InAs/AlSb T2SLs. The thin lines show deviation by kBT at
77 K from ideal behavior (thick line)
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
6 8 10 12 14
J (
A/c
m2)
1000/T (K-1)
pBp
n-p diode
Bias = 0.6V
Figure 4 Log Jdark vs. 1000/T in pBpp barrier device (bias =
0.6V) and n-p diode (bias = 0.1V), each with an
InAs/GaSb active layer bandgap wavelength of G~10
m (mesa area = 100100 m2)
4
determined to an accuracy of 0.2 monolayer (ML) from X-ray diffraction and MBE beam flux measurements, as
described in Ref. 8. It can be seen in Figure 3(a) that the measured and calculated energies match the ideal behavior
(thick line) to within an accuracy of kBT at 77 K (thin lines), for bandgap energies which span the full range from
LWIR to MWIR. Figure 3(b) shows equivalent results for 14 InAs/AlSb superlattices. In this case, error bars are shown
for those samples where the layer width accuracy was less than 0.2 ML.
The advantage of the barrier device architecture over that of a simple diode is demonstrated in Figure 4. This
figure compares a standard LWIR n-on-p diode based solely on an InAs/GaSb T2SL and operating at a bias of 0.1V,
with a LWIR pBpp device based on the design in Figure 2, and operating at a bias of 0.6V. Both devices have an active
layer band gap wavelength close to 10 μm. The barrier device (blue line) is diffusion limited down to 77 K, while the
diode (red line) is G-R limited at this temperature, with a dark current over 20 larger. The dark current in the barrier
device of Figure 4 and in all of our XBp test devices is only about one order of magnitude greater than the HgxCd1-xTe
Rule-07 value10
. This is discussed further below, where the results on test devices are compared directly with
measurements made on FPAs. Although higher than the HgxCd1-xTe value, it is shown that the dark current is in a range
that allows good and stable background limited performance (BLIP) at 77 K.
In order to understand the role of surface passivation in our barrier devices, we have performed a series of
experiments on pBpp gate controlled device (GCD) structures, with mesas etched into the BL. Figure 5(a) shows the
design of such a GCD, which was used to monitor changes in the barrier surface potential caused by the glue under-fill
process. A small array of standard devices and GCDs was bonded to a silicon fan-out circuit with indium bumps, and
the electrical characteristics of the devices were measured as a function of temperature, before and after injection of a
glue under-fill, and polishing of the GaSb substrate down to a final thickness of about 10 m. Substrate thinning is
required in order to relieve stress on cooling. It was found that after performing the under-fill process, all of the
standard devices without gates showed very large leakage currents. These currents were several orders of magnitude
greater than the true device dark current that could be measured before under-fill. The results were similar for a number
of different glues that were tested. In contrast, the true dark current could be restored in the GCDs, by applying a
suitable negative bias to the gate. Figure 5(b) shows the device dark current at 77 K and a constant operating bias (VCL =
0.8V) as a function of the gate bias. Before under-fill, a plateau of minimum dark current can be seen which
corresponds to the true device current. Over the range of the plateau (-4V < gate bias < 0V) the I-VCL characteristic of
the device shows true diffusion limited behavior, where the dark current near VCL = 0.8V is virtually bias independent
and has an activation energy close to the bandgap of the AL. Outside this range, the dark current rises steeply,
CLAL BL
Gate
VCL
Comm Le
aka
ge
ch
an
nel
Le
aka
ge
ch
an
nel
1.0E-07
1.0E-06
1.0E-05
1.0E-04
1.0E-03
1.0E-02
1.0E-01
1.0E+00
-30 -20 -10 0 10
Jd
ark
(A/c
m2)
Gate Bias (V)
78K
I(0.8V), no glue
I(0.8V), glue + polish
Figure 5 GCD measurements at 77 K., when bonded with indium bumps to a silicon fan-out. A depiction of the III-V part of the device is
shown schematically in (a) with a metal gate (grey) separated by a dielectric layer (blue) from the etched device surface and in (b) the
dark current density at a device operating bias of 0.8V is plotted as a function of gate bias before (green) and after (blue) the injection
of a glue underfill. The CL is etched down to the BL with a mesa size of 7575 m2. The device bandgap wavelength at 77 K is
9.3m. The gate bias required for minimum dark current is shifted from 0V before glue injection to -12V after, demonstrating that
the underfill causes strong surface accumulation, opening a leakage channel between devices and to common.
5
indicating an additional leakage path. After the under-fill and polishing process, the plateau of minimum dark current
maintains the same value as before, but is shifted to negative gate bias by approximately -12V. This shows that at zero
gate bias, and in devices with no gate, the under-fill induces strong surface accumulation on the BL. This layer of
negative charge opens a leakage channel between devices (as depicted in Figure 5(a)) and can short all of the devices to
common through a single low resistance defect that shunts the BL. Only by applying a gate bias of between about -12
and -16V is the negative surface accumulation layer suppressed.
In a real FPA detector, it is not feasible to use a negatively biased gate electrode around each device without
introducing an extra contact per pixel and sacrificing device yield and fill factor. Instead, once the cause of the dark
current leakage had been identified, we were able to develop a suitable passivation process that suppresses the surface
accumulation even after under-fill and polishing.
3. 640 512/15 m DIGITAL ROIC
The new ROIC has an architecture that closely follows that of the mature and successful MWIR Pelican D ROIC11
.
This approach was chosen due to the excellent performance expected in terms of readout noise, Residual Non
Uniformity (RNU), power dissipation and frame rate. Another important incentive was to support our incumbent
customers with fast integration into systems and cameras. Nevertheless, the new T2SL LWIR technology poses two
important challenges for the ROIC design:
1. The device polarity requires a polarity inversion in the
ROIC circuitry.
2. The higher photon flux compared with the MWIR
imposes very short integration periods due to the limited
area available for the integration capacitors.
To overcome this problem we have implemented a high frame
rate (up to 360Hz) with frame averaging (up to 8 frames)
performed in the proximity electronics. This enables us to
achieve an NETD value @ F/2.7 and 30Hz of less than 15mK.
The ROIC was tested at room temperature and 77 K and the
results compared favorably with preliminary predictions. Table I
presents the measured performance.
4. PELICAN-D LW FPA
"Pelican-D LW" is the first in our new line of LWIR XBp FPAs manufactured at SCD. The detector operates at 77
K and is based on a 15 µm pitch, 640 512 FPA bonded to the new digital silicon ROIC, described above. The FPA
has a nominal T2SL cut-off of 9.5 m, but in the final Integrated-Detector-Cooler assembly (IDCA) we normally
include a cold filter with a cut-off wavelength of 9.3 m. More than 80 FPAs have been made with this nominal
wavelength, but in order to demonstrate the technology in more general terms, results are shown for a larger number of
FPAs spanning a useful T2SL cut-off wavelength range between 8.7 and 10.3 m. Unless an IDCA is stated explicitly,
measurements were performed in a laboratory test Dewar without a cut-off filter.
Figure 6(a) shows the FPA dark current distribution at 77 K for 5 FPAs with cut-off wavelengths spanning the
nominal value of 9.5 μm. The median dark current value for a cut-off of 9.4 μm is close to 100 pA and the distribution
is quite narrow with a full width at half maximum of only ~6% of the median (red curve in Figure 6(a)). The maximum
distribution width in Figure 6(a) is 12%, for the two FPAs with a 9.8μm cut-off, which is still quite narrow. Note that
these two FPAs have virtually identical distributions, showing the high reproducibility of our FPA process. None of the
distribution curves have a significant high current tail, and their narrow widths demonstrate a high degree of uniformity.
A narrow dark current distribution is desirable for the stability of the FPA against temperature or bias fluctuations
Parameter Value
Well Capacity > 6Me-
Noise Floor < 1300e-
ROIC RNU < 0.025% of DR
Dynamic Range 5350
ADC Resolution 13 bit (> 7000 DL)
Maximum Frame Rate 360Hz
Input Clock Frequency 80MHz
Power Consumption 110mW @ 360Hz
Windowing Supported
Up/Down Readout Supported
Integration modes ITR, IWR
Table I
Pelican-D LW ROIC performance at 77 K
6
The median dark current values of all five FPAs are about one order of magnitude higher than the dark current
range predicted by HgxCd1-xTe Rule-07.6
Results for these and more FPAs are shown on a Rule 07 plot in Figure 6(b),
together with similar results from test devices spanning a similar wavelength range. In each case, the cut-off wavelength
(in m) is given in the legend, as measured by photoluminescence spectroscopy at 10K, and for the FPAs, the number
of FPAs measured is also stated. The standard deviation of the FPA dark current is plotted as positive and negative error
bars, and is generally quite small. It can be seen that the distributions of FPA and test device results fall in the same
range which is slightly more than one order of magnitude above the solid blue Rule-07 line.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 100 200 300
Coun
t (a
rb.
units)
Dark current (pA)
9.1 μm
9.4 μm
9.8 μm
9.8 μm
10.3 μm
1.00E-07
1.00E-06
1.00E-05
1.00E-04
1.00E-03
1.2 1.3 1.4 1.5
Jd
[A/c
m2]
1000/(e*T) [1000/(µm-K)]
Rule 07
Rule 079.2 (test dev)9.2 (test dev)9.2 (test dev)9.3 (test dev)9.3 (test dev)9.4 (test dev)9.4 (test dev)9.5 (test dev)9.6 (test dev)9.6 (test dev)9.7 (test dev)9.7 (test dev)9.7 (test dev)9.7 (test dev)9.9 (test dev)8.7 (10 FPAs)9 (6 FPAs)9.1 (19 FPAs)9.4 (56 FPAs)9.5 (18 FPAs)9.6 (12 FPAs)9.7 (23 FPAs)9.8 (30 FPAs)10.1 (19 FPAs)10.3 (7 FPAs)
10
0.4
20
(a) (b) Figure 6
77 K dark current: (a) normalized distribution of five Pelican-D LW FPAs with cut-off wavelengths between 9.1μm and 10.3μm
(b) comparison of test devices and FPAs with MCT Rule 07 (solid blue line) over a similar range of cut-off wavelengths
0
1
2
3
4
5
6
7
8
9
10
0%
20%
40%
60%
80%
100%
5 6 7 8 9 10 11
QE
Wavelength (m)
QE(k.p), λ₀ = 9.46 μm
QE (FPA)
0
10
20
30
40
50
60
70
80
1.5 2.5 3.5 4.5 5.5
<Q
E(%
)>
AL thickness (m)
3 μm (25 FPAs, 1 pass)
4.5 μm (144 FPAs, 1 pass)
(a) (b) Figure 7
Quantum efficiency at 77 K: (a) for a 1 pass detector with AL = 4.5 m as a function of wavelength, comparing measured
spectrum (blue) and simulation based on kp theory (purple) (b) simulated as a function of AL thickness for both one pass (solid
line) and two pass (broken line) detectors. Measured values, calculated by averaging over the 300 K black body spectrum from 8.0
m (filter cut-on value) to detector cut-off wavelength, are shown as points for the average of 144 FPAs with a 4.5 m AL and 25
FPAs with a 3 m AL. The error bars show the standard deviation of the FPA results.
7
Figure 7 (a) compares the spectral response of a Pelican-D LW
FPA measured at 77 K with the simulated spectrum based on the
k p and OTM calculation, described above. The measurement was
performed using a Bruker Equinox FTIR spectrometer and is quite
noisy due to a weak globar source and to some atmospheric
absorption along the measurement path. Nevertheless, it can be seen
that very good agreement is obtained with the simulation, for a
T2SL with a cut-off wavelength of 9.46 m. The fitted
inhomogeneous broadening8 at the band edge was 10 meV, which
agrees very well
with the
fluctuations in the
band edge energies
calculated from the
k p model due to
single monolayer
interface steps. The
maximum QE in the
measured spectrum is 60%. The Pelican-D LW FPA has an AL
thickness of 4.5 m, and the current version is a one pass device. The
detector QE, denoted <QE> in Figure 7 (b) is the value obtained when
the spectrum in Figure 7 (a) is averaged over the 300K black body
spectrum from 8.0 m (filter cut-on value) to the detector cut-off
wavelength. In this figure, the simulated <QE> is plotted as a function
of the AL thickness for one pass (solid line) and two pass (broken line)
detectors, respectively.
The one pass values are
compared with the
average of all the results
from the first 144
Pelican-D LW FPAs
from our new production line, where the result for a given FPA is the
mean of <QE> over all of its pixels. The standard deviation of the FPA
values is shown as positive and negative error bars. Results for 25 similar
FPAs with a 3m thick AL are also shown. In both cases the average
value falls very close to the simulated value. The simulations in Figure 7
(b) demonstrate that it should be possible to achieve a <QE> of about
60% in a two pass
version of the
Pelican-D LW
detector with the
same AL thickness
of 4.5 m.
The high
uniformity of our T2SL FPA technology is well demonstrated in
Figure 8, which shows a map of the raw signal registered (in
Digital Levels) when the Pelican-D LW FPA is placed in front of
a black body at 35C before any non-uniformity correction (NUC)
has been performed. The map is very uniform, with no stains,
large clusters or other variations across its area. This in turn leads
to a very low RNU after a two point NUC (2PC) is performed.
The magenta line in Figure 9 shows a plot of RNU vs. well fill
(WF) for an IDCA with this FPA, measured using different black
body temperatures and a constant integration time, with a 2PC
x axis pixels: 1 to 640
y a
xis
pix
els
: 1 t
o 5
12
100 200 300 400 500 600
50
100
150
200
250
300
350
400
450
500 3600
3800
4000
4200
4400
4600
4800
5000
5200
5400
Figure 8
Raw signal map (in Digital levels) registered by the
Pelican-D LW FPA, operating at 77 K in front of a
black body at 35 C.
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
40 60 80
RN
U (
%)
Well Fill (%)
Original 2PC
After 1HR, no correction
After OFF/ON, 1PC
Next day, 1PC
Figure 9
RNU of Pelican-D LW IDCA at 77 K after
performing a two point correction (2PC) at 52%
and 72% well fill (magenta); one hour later with
no correction (red); After turning off and on with
a 1PC; On next day with a 1PC.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 10 20 30
Norm
aliz
ed
pix
el co
un
t
NETD (mK)
Figure 10
NETD distribution of the Pelican-D LW FPA at
77 K and at a frame rate of 30 Hz
x axis pixels: 1 to 640
y a
xis
pix
els
: 1 t
o 5
12
QE
100 200 300 400 500 600
50
100
150
200
250
300
350
400
450
500
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
Figure 11
<QE> map of the Pelican-D LW FPA operating at 77 K
8
performed at WF values of 52% and 72% respectively. The RNU remains below 0.02% of the dynamic range for WF
values in the range 40 - 80%. Moreover, the image is very stable for many hours without any further corrections, as
shown by the red line, which was registered one hour after the 2PC was performed. The dashed lines show RNU curves
after switching the detector off and then immediately back on (black), or back on the next day (blue), and performing a
one point offset NUC (1PC). In all cases the RNU curves remain within ~0.01% of the original calibration curve.
At 65% well fill of its 6Me- capacitor and a frame rate of 240Hz, Pelican-D LW offers an NETD of 36mK when
configured with F/2.7 optics.
By averaging 8 frames at a
time, the detector operates at
an effective frame rate of 30
Hz. The NETD distribution
measured under these
conditions for the Pelican-D
LW FPA is shown in Figure
10. The distribution is narrow
and symmetric with no
pronounced tailing. The peak
value is 13 mK which is a
reduction of 8 relative to the
single frame value. This is as
expected for pure shot noise,
and shows that any noise
introduced by the averaging
procedure is negligible.
In Figure 11 we show a
map of the pixel <QE>
measured in front of a black
body at 35C for the same FPA
as in Figure 10, in which none
of the defective pixels have
been removed. The total
number of both hard and soft
defects on the FPA, defined
according to SCD’s stringent production line criteria, is 1446 giving an FPA operability of 99.56% in this case. The
<QE> map in Figure 11 shows an almost constant value of ~ 48% across the whole FPA, which agrees quite well with
0 10 20 30 40 50 60 70 806
8
10
12
14
16
18
20
22
FPA number
NE
TD
[m
K]
0 10 20 30 40 50 60 70 80
80
82
84
86
88
90
92
94
96
98
100
FPA number
Ope
rab
ilit
y
(a) (b)
Figure 12
Distribution of (a) NETD values and (b) pixel operability values for 75 Pelican-D LW FPAs from the new SCD production line
Parameter Value
Format 640 512
Pitch 15 m
Cut-off wavelength 9.3m (filter)
Quantum efficiency > 50%
Operability > 99%
RNU < 0.04% STD/DR @ 10 -
90% Well Fill capacity
NETD
15mK @ 65% Well Fill
capacity, 30Hz
(by averaging 8 frames)
Response Uniformity < 2.5% (STD/DR)
Electronics Camera Link
Cooler Ricor K548
Weight 750 gm
Environmental
conditions -40 to +71C
Total Power at 23C 16 W
Cool down time 8 min
MTTF (depends on
mission profile) 15,000 hours
Table II
Appearance of Pelican-D LW IDCA and specification of performance at 77 K
9
the value expected for a one pass detector with
the same AL thickness (4.5 m) discussed
above. If we define the BLIP temperature as the
detector operating temperature at which the
dark current is equal to the photocurrent, the
QE and dark current results discussed in this
section yield a BLIP temperature of 87 K for a
single pass Pelican-D LW FPA with F/2.7
optics. This is significantly higher than the
operating temperature of 77 K. At 77 K, the
dark current is more than 10 smaller than the
photocurrent and as noted above, its distribution
is narrow, making for good image stability
against small temperature fluctuations.
Figure 12 shows the variation of NETD
and operability values for the first 75 Pelican-D
LW FPAs coming from our new production
line. Only 2 FPAs exhibit an NETD of more
than 15 mK, and more than 65% of the FPAs
have an operability above 99%.
Finally, in Figure 13, we show an image at
5 km registered with a demonstration camera
containing a Pelican-D LW FPA operating at 77
K, and in Table II we show both the appearance
of the IDCA itself and a listing of its current performance specification.
5. CONCLUSIONS
In this article we have presented Pelican-D LW, which is SCD’s new LWIR FPA detector. It has a format of 640
512 with a 15 m pitch, and operates at 77 K with a nominal detector cut-off wavelength, defined currently by a cold
filter, of 9.3 m. The pixel operability is above 99%, according to SCD’s standard production line criteria for the
definition of bad pixels. The present version is a single pass detector with a QE of ~48% and exhibits very high
uniformity and stability of its response and dark current. The FPA has a very stable RNU below 0.02% of the dynamic
range (DR) for well fills between 40 and 80%. Over the entire working range of the detector, the RNU is expected to be
below 0.04%. The active sensing material is based on a patented diffusion limited XBpp barrier architecture, where an
InAs/GaSb T2SL is used for the AL and an InAs/AlSb T2SL is used for the barrier layer. A robust passivation process
has been developed at SCD that allows glue under-fill and substrate thinning, after bonding the sensor array with
indium bumps to the custom designed silicon ROIC. We have demonstrated sophisticated simulation techniques which
can predict the detector cut-off wavelength and spectral response a-priori, according to the individual layer thicknesses
chosen for each superlattice period, and the overall active layer stack thickness. Pelican-D LW demonstrates the
versatility of InAs/GaSb as a tunable active layer detector material. The highly reproducible results from our new LWIR
detector production line show that this material can now be considered to be a realistic alternative to HgxCd1-xTe for
small pitch, high performance LWIR and dual-color FPA detectors
ACKNOWLEDGEMENTS
The authors acknowledge technical support from Mr. S. Greenberg, who was responsible for the smooth operation of
the MBE machine, and Ms. H Schanzer, Mr. Hanan Geva, Ms. H. Moshe, Mr. Y. Caracenti, Ms. N. Hazan, Mr. I.
Bogoslavski, Mr. Y. Osmo, Mr. M. Keinan, Ms. L Krivolapov, and Ms. M. Menahem who have all contributed to the
successful processing, packaging or characterization of the materials and devices.
Figure 13
Image registered with a demonstration camera containing the 15m pitch,
640 512 Pelican-D LW FPA, operating with F/2.7 optics at 77 K (Scene
distance = 5 Km)
10
REFERENCES
1 G.A. Sai-Halasz, R. Tsu, and L. Esaki, “A new semiconductor superlattice” Appl. Phys. Lett. 30, 651 (1977)
2 A. Glozman, E. Harush, E. Jacobsohn, O. Klin, P.C. Klipstein, T. Markovitz, V. Nahum, E. Saguy, J. Oiknine-
Schlesinger, I. Shtrichman, M. Yassen, B. Yofis, and E. Weiss, “High performance InAlSb MWIR detectors operating at 100K and beyond” Proc. SPIE 6206, 6206-0M (2006)
3 P.C. Klipstein, Depletionless Photodiode with Suppressed Dark Current…, US Patent 7,795,640 (2 July 2003)
4 P.C. Klipstein, Unipolar semiconductor photodetector with Suppressed Dark Current…, US Patent 8,004,012 (6 April
2006) 5 P.C. Klipstein, “XBn Barrier Photodetectors for High Sensitivity and High Operating Temperature Infrared Sensors”,
Proc. SPIE 6940, 6940-2U (2008) 6 W.E. Tennant, “Rule 07 Revisited: Still a Good Heuristic Predictor of p/n HgCdTe Photodiode Performance?” Journal
of Electronic Materials 39, 1030 (2010) 7 O. Klin, N. Snapi, Y. Cohen, and E. Weiss, " A study of MBE growth-related defects in InAs/GaSb type-II superlattices
for long wavelength infrared detectors", J. Crystal Growth 425, 54 (2015) 8 Y.Livneh, P.C. Klipstein, O. Klin, N. Snapi, S. Grossman, A. Glozman and E. Weiss, “k · p model for the energy
dispersions and absorption spectra of InAs/GaSb type-II superlattices”, Phys. Rev. B 86, 235311 (2012); Erratum, Phys. Rev. B 90, 039903 (2014)
9 P.C. Klipstein, Y. Livneh, A. Glozman, S. Grossman, O. Klin, N. Snapi and E. Weiss, “Modeling InAs/GaSb and
InAs/InAsSb Superlattice Infrared Detectors”, Journal of Electronic Materials, 43, 2984 (2014) 10
P.C. Klipstein, E. Avnon, Y. Benny, R Fraenkel, , A. Glozman, S. Grossman, O. Klin, L. Langoff, Y. Livneh, I Lukomsky, M. Nitzani, L. Shkedy, I. Shtrichman, N. Snapi, A. Tuito and E. Weiss, “InAs/GaSb Type II superlattice barrier devices with a low dark current and a high-quantum efficiency”, Proc. SPIE 9070, 9070-0U (2014)
11 P.C. Klipstein, Y. Gross, D. Aronov, M. Ben Ezra, E. Berkowicz, Y. Cohen, R. Fraenkel, A. Glozman, S. Grossman, O.
Klin, I. Lukomsky, T. Markowitz, L. Shkedy, I. Shtrichman, N. Snapi, A. Tuito, M. Yassen, and E. Weiss “Low SWaP MWIR detector based on XBn focal plane array”, Proc. SPIE 8704, 8704-1S (2013)