High Image Quality Type-II Superlattice Detector for 3.3 µm
Detection of Volatile Organic Compounds
Hedda Malmi / Anders Gamfeldt, Rickard Marcks von Würtemberg, Dan Lantz,
Carl Asplund and Henk Martijn
IRnova AB, Electrum 236, SE-164 40 Kista, Sweden
Abstract
Recent improvements in material quality, structure design and processing have made type-II
superlattice a competing high end detector technology. This has also made it an attractive material
of choice to meet the industrial need of high end gas detection, as for example detection of methane
and other volatile organic compounds (VOC). A heterojunction structure with a cut off at 5 µm but
intended for detection of VOC at 3.3 µm will be presented. The detector format is 320x256 pixels
with 30 µm pitch using the ISC9705 read out circuit. The detector operability is 99.8 % and NETD 12
mK (7 ms integration time, object temperature 30°C and F#2.6, no cold filter used). The uniformity is
at least on par with QWIP detectors. Anti-reflective coating is used and the substrate is fully removed.
High quality imaging at operating temperature 110 K will be presented.
Classification codes (PACS and/or MSC) and keywords
Infrared detectors, strained layer superlattice, mid wave infrared, barrier structure, InAs/GaSb,
volatile organic compounds (VOC)
Introduction
Over the last ten years there has been a steady improvement of the detector technology type-II
superlattice (T2SL), consisting of epitaxially grown layers of InAs and GaSb. This detector technology
has the theoretical possibility of high quantum efficiency and low dark current, due in part to
reduced interband tunneling and suppressed Auger generation, and in part to the many degrees of
freedom in layer composition and thickness, which allows for efficient barrier designs by bandgap
engineering. Progress in GaSb substrate quality [1, 2], epitaxial structure design [3-6] and passivation
[7- 9] have made it possible to approach the performance of the competing detector technology
Mercury Cadmium Telluride (MCT) [10]. This has also made it attractive for industrial applications.
An emerging industrial application for high end infrared detectors is gas detection. Many gases are
transparent in the visible and most infrared wavelengths, but some gases have absorption in the
infrared regions. These gases also radiate in the absorbed wavelengths region. By using a filter, the
detection is limited to the region of interest and the contrast is increased. This method can be used
for visualization and documentation in real time of gas leaks. In the long wave infrared range sulfur
hexafluoride (SF6) can be detected with a peak at 10.55 µm [11]. In the mid wave infrared range
there is a large group of volatile organic compounds (VOC), for example methane, butane and
propane with peaks around 3.3 µm that can be detected.
Methane makes up a considerable part of human greenhouse gas emission and about a third of the
total methane emissions come from the production and distribution of natural gas and petroleum
[12]. This makes the detection of methane leaks in this industry an important factor for reducing
greenhouse emissions. Methane is also highly flammable and explosive if mixed with air, and
undiscovered gas leaks have led to several lethal accidents.
Another very important reason for minimizing the release of VOCs into the air is the ability of VOCs
to form ground level ozone when reacting with other pollutants such as nitrous oxide and carbon
monoxide. Ground level ozone is a fundamental component in the formation of smog and is a huge
problem for many of the world’s largest cities. Although emission regulations can help, locating the
sources of VOC emissions around and in cities can be very difficult limiting the enforceability of such
regulations. Besides the environmental and safety aspects of gas leak detection, there is also an
economic incentive to reduce the gas lost in leaks in for example natural gas distribution.
The use of infrared imaging for gas leak detection has several advantages over traditional “sniffing”
type instruments (where the instrument needs to be in direct contact with the gas to be detected).
The technical principle behind this technique is to make use of the absorption that many gases have
in certain infrared spectral bands. By fine tuning the detection band of an infrared imager the
otherwise invisible gas appears clearly in the infrared image.
Such a device gives the operator an actual live video of the equipment to be surveyed, with any leaks
appearing as dark or light clouds in the image. Obviously, this means large areas or big equipment
can much more easily be scanned for leaks than if using a sniffing device. When scanning for
explosive or toxic gases, the possibility to do the survey from a safe distance is also an important
advantage.
To match existing read out circuits (ROIC), mainly p-on-n detector structures are used to make T2SL
FPAs. That is also the case for the VOC detector described in this paper. However, a comparison was
made to an n-on-p structure in order to evaluate the detector material properties.
Experiment
Two heterojunction p-i-n-diodes were epitaxially grown on 3-inch n-type (Te-doped) GaSb (100)
substrates using solid source molecular beam epitaxy (MBE).
Structure A consists of a 0.3 µm lattice matched InAs0.91Sb0.09 etch-stop layer followed by 0.7 µm
heavily Si-doped n-type ‘‘M-superlattice’’ (M-SL) bottom contact, 0.5 µm lightly doped n-type M-SL,
3.9 µm weakly doped p-type SL absorber, 0.1 µm of heavily Be-doped p-type SL, and finally a 0.1 µm
thick heavily Be-doped p-type GaSb contact layer. The M-SL periods consist of InAs/GaSb/AlSb/GaSb
in the proportions 10/1/4/1 monolayers (ML), whereas the absorber SL consists of 10 ML InAs/ 11.5
ML GaSb. The top and bottom contacts are doped to 2 x 1018 cm-3 and 6 x 1017 cm-3 respectively, and
the doping concentration in the absorber is 3 x 1016 cm-3 (p-type). This is a slightly modified and
improved version of the epitaxial design reported in [13], as explained in detail in [15]. A calculated
band edge diagram for this structure is shown in Figure 1.
Structure B was also grown on n-type (Te-doped) GaSb (100) substrates using MBE with the aim of
producing an inverted version of Structure A describe above. Structure B consists of a 0.3 µm lattice
matched InAs0.91Sb0.09 etch-stop layer, heavily Si-doped to 2 x 1018 cm-3 followed by a 1 µm thick p-type
bulk GaSb contact and a 0.1 µm p-SL contact, a 3.9 µm weakly doped p-type SL absorber, 0.5 µm
lightly doped n-type M-SL barrier and finally a n-SL top contact. The M-SL period design and the
absorber SL design are both equal to the ones in Structure A, including doping concentrations. The
top and bottom contacts are doped to 6 x 1017 cm-3 and 2 x 1018 cm-3 respectively.
Single pixel detectors of sizes from 30x30 µm2 to 190x190 µm2 were produced using standard III/V
processing techniques. The pixels were formed by a combination of dry and wet etching [14] and
passivated using a dielectric passivation [15]. Mirror and contact metal was evaporated onto the
pixels and the chip was glued to a PCB.
Using the same processing methods, 320x256 pixels detector arrays with 30 µm pixel pitch were
produced out of Structure A. Stepper lithography was used to define the pixels in this case. Mirror,
contact metal and indium bumps were evaporated onto the pixels before dicing. The arrays were
then hybridized to the read out circuit ISC9705, underfill was deposited and finally the GaSb
substrate was fully removed. The focal plane arrays (FPA) are using an antireflective coating
optimized for detection at 3.3 µm.
The FPAs were then mounted on a ceramic carrier, wire-bonded and put in a cooled test dewar with
F#2.6 for tests of imaging performance. When applicable, a cold passband filter at 3.1-3.575 µm was
used.
Figure 1 Calculated band edges for Structure A under zero bias. Light enters from the left. The etch stop layer is not
shown.
Results and discussion
The external quantum efficiency of Structure A at 100 K with antireflection coating optimized for
3.3 µm wavelength is shown in Figure 2. The complete absence of bias dependence shows that
minority electrons flow unobstructed from absorber into the M-SL barrier – i.e. there is no barrier in
the SL conduction band.
1
0% substr. loss
0.9
External quantum efficiency
0.8
0.7
0.6
-0.800 V
-0.500 V
-0.100 V
+0.000 V
0.5
0.4
0.3
0.2
0.1
0
3
3.5
4
4.5
5
Wavelength (µm)
5.5
6
Figure 2 External quantum efficiency of Structure A at 100 K with antireflection coating optimized for 3.3 µm wavelength.
The effect of the antireflection coating is illustrated in Figure 3 where the external quantum
efficiency of Structure A was measured on focal plane arrays with light entering from the backside.
The cut-off wavelength is dependent on temperature but there is very little temperature
dependence below 4.5 µm. The standing wave pattern is due to optical interference between the
air/semiconductor interface and the semiconductor/metal interface on the bump side.
1
0% substr. loss
0.9
0% substr. loss
a
b
External quantum efficiency
0.8
T=
T=
T=
T=
T=
T=
T=
0.7
0.6
0.5
0.4
0.3
140 K
130 K
120 K
110 K
100 K
90 K
80 K
30oC photon radiance (a.u.)
0.2
0.1
0
3
3.5
4
4.5
5
Wavelength (µm)
5.5
63
3.5
4
4.5
5
Wavelength (µm)
5.5
6
Figure 3 External quantum efficiency of Structure A without (left) and with (right) antireflection coating optimized for
3.3 µm wavelength. The lowest temperature corresponds to the lowest quantum efficiency.
The solid line of Figure 4 shows the calculated background limited spectral detectivity (D*) of
Structure A (with an AR-coating optimized for 3-5 µm) at 100 K based on the measured quantum
efficiency. The dashed line indicates the peak spectral D* for an ideal photo voltaic detector as a
reference. Note that the peak value of 1.1x1011 cm*Hz1/2/W occurs at a shorter wavelength than the
cut-off wavelength λco,90 %=5.3 µm of the device. This is due to the lower quantum efficiency towards
the cut-off. At 100 K, the background photon noise is the predominant noise source and Johnson
noise or other noise sources do not play a significant role. The numbers are given for a background
temperature of 300 K and with 2π field of view.
spectral specific detectivity (2π)
12
10
D* (cm sqrt(Hz)/W)
D* detector
D* max
11
10
10
10
1
10
Wavelength (µm)
100
Figure 4 Calculated background limited spectral detectivity (D*) at 100 K, based on the measured quantum
efficiency. The numbers are given for a 300 K black body with 2π field of view.
After hybridization to read-out circuit ISC9705 and successful removal of the full substrate, pictures
at 120 K were taken with the FPA demonstrator, see Figure 5. The pictures are corrected for nonoperating pixels and offset and gain (two point correction). No cold filter is used.
Figure 5 120 K pictures taken with an FPA demonstrator. No cold filter is used.
The detector operability of this particular FPA is 99.77 %, close to the average of 99.84 % of all
detectors complying with specification of the first production run. A pixel is declared non-operating
when its response is deviating more than 7.5 % from its neighbors or if the NETD is larger than 40 mK.
A histogram of noise equivalent temperature difference (NETD) of an FPA at 85 K is shown in Figure 6.
F#2.6, integration time 6.5 ms and about 50 % well fill is used. The NETD is calculated based on a
series of 16 images taken at 30⁰C. The distribution is close to Gaussian and does not exhibit a
significant tail upwards. The mean NETD is 12 mK, very close to the theoretical limit for a background
limited detector.
The spatial and temporal noise at the same premises are shown in Figure 7. The spatial noise is a
measure of how much pixels statistically deviate from their neighbors. This graph is a good
description of how the detector performs at different temperatures. Between 70 K and 110 K the
performance does not change significantly. Above 110 K the Johnson noise and dark current start to
play a role, leading to an increase in temporal NETD. Even the spatial noise increases significantly.
Still it is possible to make good imaging at 120 K as seen in Figure 5; the increase in spatial noise is
not yet detectable in the image.
Histogram NETD IRnova320 MW
14000
12000
10000
Count
8000
6000
4000
2000
0
0
5
10
15
20
NETD (mK)
25
30
35
40
Figure 6 NETD histogram at 85 K, F#2.6, Tint 6.5 ms and ~ 50 % well fill.
Figure 7 Temporal and spatial noise for different temperatures, F#2.6, Tint 6.5 ms and ~ 50 % well fill.
Residual Fixed Pattern Noise
7000
6000
5000
count
4000
3000
2000
1000
0
-50
-40
-30
-20
-10
0
10
RFPN (mK)
20
30
40
50
Figure 8 Residual fixed pattern noise distribution (RFPN) at a blackbody temperature of 35⁰C with the two-point
correction temperatures at 30⁰C and 40⁰C and a maximum scene temperature of 50⁰C.
Figure 8 shows the residual fixed pattern noise distribution (RFPN) at a blackbody temperature of
35⁰C with the two-point correction temperatures at 30⁰C and 40⁰C and a maximum scene
temperature of 50⁰C. Dead pixels are excluded. The detector temperature is 82.5 K. The RFPN
defined as the standard deviation of these values is 6.8 mK, close to only 50 % of the temporal NETD.
The two-point correction is done based on the mean value of only 16 images at the correction points.
Some broadening of the distribution is caused by some residual temporal noise and hence the
underlying real RFPN is even lower.
The dark current components and the passivation of structure A has been discussed elsewhere [14].
An FPA was thermally cycled 60 times by submerging it in liquid nitrogen and holding it there until
the detector was of the same temperature as of the nitrogen and then bringing it back to room
temperature again by heating it with a flow of nitrogen gas. No visual deviations of the detector
surface (with AR-coating) could be detected, neither any changes in NETD, operability or response.
One FPA was mounted into a test dewar with F#2 and a cold passband filter at 3.1-3.575 µm.
Although using a hand-held mismatched lens (F#2.3), it was possible to make a proof of concept of
the VOC gas detection. In Figure 9 lighter gas (butane/propane) is clearly visible. The picture is
corrected for offset and gain and some non-operating pixels are removed.
Figure 9 A person lets gas escape from a lighter and the gas can be seen. The image is part of a film sequence made at
operating temperature 85 K and integration time 8 ms.
The local response non-uniformity (RNU) is a measure of the detector uniformity and is calculated as
the standard deviation of the difference between one pixel and the median value of the surrounding
15x15 pixels when looking at an extended blackbody. The median RNU before correction is 0.39 %
over a series of 13 VOC FPAs. This can be compared to a standard QWIP product of the same format,
measured at the same condition as the VOC except for operating temperature. The typical RNU
before correction for the QWIP is 0.64 % and hence the T2SL VOC detector uniformity is on par with
QWIPs or even better than QWIPs.
Comparison n-on-p to p-on-n
Figure 10 shows the dark current measurements of fully etched large mesas from structure A (a) and
the structure B (b). The bulk properties are very similar. These structures are diffusion limited
at >120 K due to the excellent material quality and an efficient hole barrier design, despite the rather
long cutoff wavelength λco,90 %=5.3 µm.
0
10
-1
10
-1
a
-2
-2
-3
-3
-4
-4
-5
-5
-6
-6
-7
-7
Current density (A/cm2)
10
10
10
10
10
10
-8
10
-0.25
b
T=
T=
T=
T=
T=
T=
T=
T=
T=
T=
T=
170 K
160 K
150 K
140 K
130 K
120 K
110 K
100 K
90 K
80 K
70 K
-8
-0.2
-0.15
-0.1 -0.05
Bias (V)
0
0.05
-0.25
-0.2
-0.15
-0.1 -0.05
Bias (V)
0
0.05
Figure 10 Dark current measurements of fully etched large mesas from structure A (a) and the structure B (b). The lowest
dark current density corresponds to the lowest temperature.
Both structures are passivated the same way. Mesa sidewall currents are effectively, although not
completely, suppressed on both structures. The bulk dark current at 120 K and bias -50 mV is 8 x 10 -7
A/cm2 for Structure A and for Structure B and the RA product is 2 x 105 Ωcm2 for structure A and 1.5 x
105 Ωcm2 for structure B.
The quantum efficiency of structure B is comparable with that of structure A, including a complete
absence of bias dependence up to at least -1.5 V bias.
To conclude, the performance of n-on-p is similar to that of p-on-n and so far no outstanding
advantage for one of them has appeared. However, n-on-p gives freedom of choice on the etch
depth, either to etch just through the pn-junction or to etch all the way into the p contact. In the pon-n case, there is no choice.
Summary and conclusion
High quality imaging at 120 K using a heterojunction p-on-n superlattice detector with a cutoff at
5.3 µm has been shown. Several system level performance measures have been presented and we
conclude that up to 100 K these devices show very good uniformity and operability and close to
theoretical NETD.
Proof of concept of VOC gas detection has been presented. Future work will be to improve the
detector performance and especially the operating temperature. Calculations indicate that by
changing the absorber cut-off wavelength from 5.3 to 3.8 µm the operating temperature can be
increased to 140 K. Another planned improvement is to increase the fillfactor. The present detector
performance is good even with a fillfactor of 75 %. Close to 90 % has been successfully implemented
on other T2SL products. Testing of the long term stability will continue.
The dark current of an n-on-p structure has been compared to the p-on-n structure used for imaging.
Both structures are diffusion limited above ~120 K. Their similarity gives freedom to use new
available ROICs [16] without being limited by performance of the detector material.
Acknowledgements
We would like to thank the European Commission for sponsoring part of this work through the EU
FP7 project MINERVA. We would also like to thank the Swedish Innovation Agency for sponsoring
part of this work through the Eurostars project MILES.
References
[1] L.P. Allen, J.P. Flint, G. Meshew, J. Trevethan, G. Dallas, D. Bakken, A. Khoshakhlagh, C. J. Hill,
Manufacturing of 100mm diameter GaSb and 150mm InSb substrates for advanced space based
applications, Proc. SPIE 8012, Infrared Technology and Applications XXXVII, 801215 (2011)
[2] M. Furlongen, R. Martinez, S. Amirhaghi, B. Smith and A. Mowbray, Multiwafer production of
epitaxy ready 4" GaSb substrates: requirements for epitaxially growth infrared detectors, Quantum
Sensing and Nanostructured Devices IX, Proc. SPIE 8268, 826818 (2012)
[3] I. Vurgaftman, E.H. Aifer, C.L. Canedy, J.G. Tischler, J.R. Meyer, J.H. Warner, E.M. Jackson, G.
Hildebrandt, G.J. Sullivan, Graded band gap for dark-current suppression in long-wave infrared Wstructured type-II superlattice photodiodes, Appl. Phys. Lett. 89 (2006) 121114.
[4] D. Z.-Y. Ting, C. J. Hill, A. Soibel, S. A. Keo, J. M. Mumolo, J. Nguyen, S. D. Gunapala, A highperformance long wavelength superlattice complementary barrier infrared detector, Appl. Phys. Lett.
95 (2009) 023508.
[5] B.-M. Nguyen, M. Razeghi, V. Nathan, G.J. Brown, Type-II ‘‘M’’ Structure Photodiodes: An
Alternative Material Design for Mid-Wave to Long Wavelength Infrared Regimes’’, Proc. SPIE 6479,
64790S, 2007.
[6] S. Abdollahi Pour, E.K. Huang ,G. Chen ,A. Haddadi, B.-M. Nguyen, M. Razeghi, High operating
temperature midwave infrared photodiodes and focal plane arrays based on type-II InAs/GaSb
superlattices, Appl. Phys. Lett. 98 (2011) 143501.
[7] G. Chen, B.-M. Nguyen, A.M. Hoang, E.K. Huang, S.R. Darvish, M. Razeghi, Elimination of surface
leakage in gate controlled type-II InAs/GaSb midinfrared photodetectors, Appl. Phys. Lett. 99 (2011)
183503.
[8] E. A. Plis, M. N. Kutty, and S. Krishna, Passivation techniques for InAs/GaSb strained layer
superlattice detectors, Laser Photonics Rev., 1–15 (2012).
[9] O. Salihoglu, A. Muti, K. Kutluer, T. Tansel, R. Turan, C. Kocabas, and A. Aydinli, Atomic layer
deposited Al2O3 passivation of type II InAs/GaSb superlattice photodetectors, Journal of Applied
Physics 111, 074509 (2012)
[10] M. Razeghi, A. Haddadi, A.M. Hoang, E.K. Huang, G. Chen, S. Bogdanov, S.R. Darvish, F.
Callewaert, R. McClintock, Advances in antimonide-based Type-II superlattices for infrared detection
and imaging at center for quantum devices, Infrared Physics & Technology 59 (2013) 41–52.
[11] http://www.ir-nova.se/applications/gas-detectionindustrial/
[12] "U.S. INVENTORY OF U.S. GREENHOUSE GAS EMISSIONS AND SINKS: 1990-2010 (April 2012)".
Greenhouse Gas Inventory Report. United States Environmental Protection Agency. April 2012.
Retrieved February 20, 2013.
[13] C. Asplund , R. Marcks von Würtemberg, D. Lantz, H. Malm, H. Martijn, E. Plis, N. Gautam and S.
Krishna, Performance of mid-wave T2SL detectors with heterojunction barriers, Infrared Physics &
Technology 59 (2013) 22–27.
[14] R. Chaghi, C. Cervera, H. Aït-Kaci, P. Grech, J. B. Rodriguez and P. Christol, Wet etching and
chemical polishing of InAs / GaSb superlattice photodiodes, Semicond. Sci. Technol. 24, p. 065010,
(2009).
[15] H. Martijn, C. Asplund, R. Marcks von Würtemberg and H. Malm, High performance MWIR typeII superlattice detectors, Infrared Technology and Applications XXXIX, Proc. of SPIE Vol. 8704, 87040Z
(2013).
[16] See e.g. the ISC0905 Readout Integrated Circuit: http://cvs.flir.com/isc0905, or the ISC0903:
http://cvs.flir.com/isc0903
i
Corresponding author: email: [email protected], phone: +46 8 793 66 14, fax: +46 8 519 02518