Assessment of HgCdTe photodiodes and quantum well infrared photoconductors for long wavelength focal plane arrays

Infrared Physics & Technology 40 Ž1999. 279–294 www.elsevier.comrlocaterinfrared

Assessment of HgCdTe photodiodes and quantum well infrared photoconductors for long wavelength focal plane arrays A. Rogalski

)

Institute of Applied Physics, Military UniÕersity of Technology, 2 Kaliskiego Street, 00-908 Warsaw 49, Poland Received 10 November 1998

Abstract Recent trends in infrared detectors are towards large, electronically addressed two-dimensional arrays. In the long wavelength infrared ŽLWIR. spectral range HgCdTe focal plane arrays ŽFPAs. occupy a dominant position. However, the slow progress in the development of large LWIR photovoltaic HgCdTe infrared imaging arrays and the rapid achievements of novel semiconductor heterostructure systems have made it necessary to foresee the future development of different material technologies in fabrication large FPAs. Among the competing technologies in LWIR are the quantum well infrared photoconductors ŽQWIPs. based on lattice matched GaAsrAlGaAs and strained layer InGaAsrAlGaAs material systems. This paper compares the technical merits of two IR detector arrays technologies; photovoltaic HgCdTe and QWIPs. It is clearly shown that LWIR QWIP cannot compete with HgCdTe photodiode as the single device especially at higher temperature operation Ž) 70 K. due to fundamental limitations associated with intersubband transitions. However, the advantage of HgCdTe is less distinct in temperature range below 50 K due to problems involved in HgCdTe material Žp-type doping, Shockley–Read recombination, trap-assisted tunnelling, surface and interface instabilities.. Even though the QWIP is a photoconductor, several of its properties such as high impedance, fast response time, long integration time, and low power consumption, well satisfy the requirements of fabrication of large FPAs. Due to the high material quality at low temperature, QWIP has potential advantages over HgCdTe for very LWIR ŽVLWIR. FPA applications in terms of the array size, uniformity, yield and cost of the systems. q 1999 Elsevier Science B.V. All rights reserved. Keywords: HgCdTe photodiodes; Quantum well infrared photoconductors; Focal plane arrays

1. Introduction Currently Hg 1y x Cd xTe ŽHgCdTe. is the most important semiconductor alloy system for infrared ŽIR. detectors. However, the slow progress in the development of large photovoltaic HgCdTe infrared imaging arrays and the rapid achievements of novel semiconductor heterostructure systems have made it more

)

Tel.: q48-22-685-9109; e-mail: [email protected]

difficult to predict what types of arrays will be readily available for future systems applications. For spaceborne surveillance systems, low background IR seekerrtracker systems, reliable and affordable sensors with long life are needed which can function effectively at temperatures higher than the 20–30 K currently required by bulk photon detectors. The only alternative to HgCdTe that had been available so far was extrinsic Si, which operates at much lower temperatures where a problematic three-stage cryocooler would be required. Improvement in surveil-

1350-4495r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 1 3 5 0 - 4 4 9 5 Ž 9 9 . 0 0 0 0 3 - 1

280

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

lance sensors and interceptor seekers require large area, highly uniform and multicolor Žor multispectral. IR focal plane arrays ŽFPAs. involving long wavelength IR ŽLWIR. and very long wavelength IR ŽVLWIR. regions. Among the competing technologies are the quantum well infrared photoconductors ŽQWIPs. based on lattice matched GaAsrAlGaAs and strained layer InGaAsrAlGaAs material systems. In this paper, we discuss the development and state of the art of LWIR and VLWIR IR detectors based on HgCdTe ternary alloy system and GaAsr AlGaAs QWIPs. In this context, the paper completed results presented in Refs. w1–4x. We compare the performance of both types of devices as single element devices. Next such FPA issues as array size, uniformity, operability, multicolour capability and cost of systems, will be discussed.

2. Device performance In this section the performance of HgCdTe photodiodes and QWIPs as the single devices are discussed. Details will be given considering the quantum efficiency ŽQE., dark current, detectivity and BLIP temperature. Table 1 compares the essential properties of both types of devices at 77 K. 2.1. Quantum efficiency HgCdTe has a direct band-gap energy that translates to high radiative efficiency and can be tailored to absorb IR radiation at the wavelength of interest. When operated in the photovoltaic mode, the optical

gain Žsimply defined as the ratio of the photoelectron lifetime to the transit time. is close to one and the responsivity is directly proportional to the QE of the photodiode. In the present paper QE is defined as the ratio of the number of photoexcited electrons collected by readout integrated circuit ŽROIC. to the number of photons incident at the detector surface. HgCdTe has large optical absorption and wide absorption band irrespective of the light polarisation, which greatly simplifies the detector array design. Quantum efficiency is routinely produced around 70% without antireflection ŽAR. coating and in excess of 90% with AR coating and are spectrally constant from less than 1 mm out to near the cutoff of the detector. The wide-band spectral sensitivity with near perfect QE enables greater system collection efficiency Žsmaller aperture. making the FPA useful for imaging, spectral radiometry, and longrange target acquisition. It should be noticed however, that the current LWIR staring array performance is mostly limited by the charge handling capacity on the ROIC and the background Žwarm optics.. Due to intersubband transitions in the conduction band, the n-type QWIP detection mechanism requires photons with non-normal angle of incidence to provide proper polarisation for photon absorption. The absorption QE is relatively small, about 25% using two-dimensional Ž2-D. grating. Since the QWIP is a photoconductive detector, the responsivity is proportional to the conversion efficiency, which is the product of the absorption QE times the optical gain. The optical gain of QWIP structures vary from 0.2 Žin the case of 30 to 50 wells. to larger than 1. From the above consideration results that the QE is

Table 1 Essential properties of LWIR HgCdTe photodiodes and QWIPs at T s 77 K Parameter

HgCdTe

QWIP

IR absorption

Normal incidence

Quantum efficiency Spectral sensitivity Optical gain Thermal generation lifetime R o A product Ž lc s 10 mm. Detectivity Ž lc s 10 mm, FOV s 0.

G 70% Wide-band 1 f 1 ms 300 V cm2 2 = 10 12 cm Hz 1r2 Wy1

Eoptical H plane of well required Normal incidence: no absorption F 10% Narrow-band ŽFWHMf 1 % 2 mm. 0.2 Ž30–50 wells. f 10 ps 10 4 V cm2 2 = 10 10 cm Hz 1r2 Wy1

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

Fig. 1. Quantum efficiency versus wavelength for a HgCdTe photodiode and GaAsrAlGaAs QWIP detector with similar cutoff.

typically below 10% at the maximum response, rapidly rolling off on both the short and long wavelength sides off the peak. Fig. 1 compares the spectral QE of a HgCdTe photodiode to a QWIP. Higher bias voltage are used to boost QE. However, increase of the bias voltage also causes the leakage current to increase with reverse bias, which in consequence limits any potential system performance improvement. New grating designs are under study to improve the QE, such as antenna grating, and corrugated grating w3x. It should be stressed that so far, most efforts on QWIPS are for tactical applications, where the structure designs and the doping are optimised to increase the operating temperature and suit the readout charge handling capacity. It is well known that using a smaller number of quantum wells and bound to continuum structures, increase of the optical gain and improvement of the detector performance for low temperature applications are possible w5,6x. Tidrow has presented w7x a high performance QWIP consisting of only three quantum wells with conversion efficiencies up to 29% at a bias voltage y0.8 V and peak wavelength 8.5 mm.

281

current is less than 10y5 Arcm2 at 77 K. The bias-independent leakage current aids in achieving FPA uniformity as well as reducing detector biascontrol requirements during changes in photocurrent. Usually for Hg 1y x Cd xTe photodiodes with x f 0.22, in the zero-bias and low-bias region, diffusion current is the dominant current down to 60 K w8,9x. For medium reverse bias, trap-assisted tunnelling produces the dark current, and also dominates the dark current at zero bias below 50 K. For a high reverse bias, bulk band-to-band tunnelling dominates. At low temperature, such as 40 K, large spreads in R o A product distributions are typically observed due to onset of tunnelling currents associated with localised defects. Moreover, the HgCdTe photodiodes often have additional dark current, particularly at low temperature, which is related to the surface. The average value of R o A product at 77 K for a 10-mm cutoff HgCdTe photodiodes at 77 K is around 300 V cm2 and drops to 30 V cm2 at 12 mm w10,11x. At 40 K, the R o A product varies between 10 5 and 10 8 V cm2 with 90% above 10 5 V cm2 at 11.2 mm w11x. The dependence of the base region diffusion limited R o A product on the long wavelength cutoff for p-on-n HgCdTe photodiodes at temperatures F 77 K is shown in Fig. 3. The solid lines are calculated assuming that the performance of photodiodes are due to thermal generation governed by the Auger mechanism in the base n-type region of photodiodes with thickness t s 10 mm and doping Nd s 5 = 10 14

2.2. Dark current and R o A product Ideal photodiodes have very low leakage current that is insensitive to detector bias. Leakage current is the primary contribution to unwanted noise. Fig. 2 shows typical current–voltage characteristics of HgCdTe photodiode at temperatures between 40 and 90 K for a 12 mm cutoff detector at 40 K. Leakage

Fig. 2. Current–voltage characteristics at various temperatures for a 12 mm cutoff HgCdTe photodiode Žafter Ref. w4x..

282

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

Fig. 3. Dependence of the R o A product on the long wavelength cutoff for LWIR p-on-n HgCdTe photodiodes at temperatures F 77 K. The solid lines are calculated assuming that the performance of photodiodes are due to thermal generation governed by the Auger mechanism in the base n-type region of photodiodes with thickness t s10 mm and doping Nd s 5=10 14 cmy3 . The experimental values are taken from different papers Žafter Ref. w8x..

cmy3 . At 77 K the upper experimental data are situated about a half an order below ultimate theoretical predictions. With a lowering of the operation temperature of photodiodes, the discrepancy between the theoretical curves and experimental data increases, which is probably due to additional currents in the junctions Žsuch as tunnelling current or surface leakage current. that were not considered. In comparison with HgCdTe photodiodes, the behaviour of the dark current of QWIPs is better understood w3x. Fig. 4 shows the current mechanisms in a QWIP. At low temperatures ŽT - 40 K for lc s 10 mm., the dark current is mostly caused by defect related direct tunnelling. In the medium operating range between 40 and 70 K Žfor lc s 10 mm., the thermally assisted tunnelling dominates. In this case, electrons are thermally excited and tunnel

through the barriers with assistance from the defects and the triangle part of the barrier at high bias. At high temperature Ž) 70 K for l c s 10 mm., thermally excited electrons are thermionically emitted and transport above the barriers. It is difficult to block this dark current without sacrificing the photoelectrons Žtransport mechanisms of thermionically emitted current and photocurrent are similar.. Minimising thermionically emitted current is critical to the commercial success of the QWIP, as it allows the highly desirable high-temperature camera operation. Dropping the first excited state to the top theoretically causes the dark current to drop by a factor of f 6 at a temperature of 70 K. This compares well with the fourfold drop experimentally observed for 9-mm cutoff QWIPs w12x. The value of the QWIP dark current could be adjusted using different device structures, doping densities, and bias conditions. Fig. 5 shows the I–V characteristics for a range of temperature between 35 and 77 K measured on a device with 9.6 mm spectral peak. Typical operation at 2 V applied bias in the slowly-varying region of current with bias between the initial rise in current at low voltage and the later rise at high bias. Typical LWIR QWIP dark current at 77 K is about 10y4 Arcm2 , which is in the nanoamper range for 24 = 24 mm2 pixel. w3x Comparing Figs. 2 and 5 we can see show that a 9.6-mm QWIP must be cooled to 60 K to achieve leakage current comparable to a 12 mm HgCdTe photodiode operating 25 degrees warmer. Additional insight into difference in temperature dependence of the dark currents gives Fig. 6, which shows the current density versus temperature for

Fig. 4. The current mechanisms of QWIP: DT is the direct tunnelling, TAT is the thermally assisted tunnelling, and TE is the thermionic emission. The dashed lines show the photocurrent mechanism Žafter Ref. w3x..

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

Fig. 5. Current–voltage characteristics of a QWIP detector having a peak response of 9.6 mm at various temperatures, along with the 300 K background window current measured at 30 K with a 1808 FOV Žafter Ref. w13x..

GaAsrAlGaAs QWIP and a HgCdTe photodiode, both with lc s 10 mm w14x. The current densities of both detectors at temperatures below 40 K are tunnelling limited Žtemperature independent., however the current density of QWIP is lower than that of HgCdTe photodiode. The thermionic emission regime for QWIP ŽG 40 K. is highly temperature dependent, and ‘cuts on’ very rapidly. At 77 K, the QWIP has a dark current, which is approximately two orders of magnitude higher than that of the HgCdTe photodiode. QWIP operates at a bias voltage from 1 to 3 V depending on the structure and periods of the devices. Using the voltage divided by the dark current density, the RA products are usually larger than 10 7 V cm2 and 10 4 V cm2 when operated at 40 K and 77 K, respectively w3x. These values indicate very high impedance.

283

noise is neglected in most cases, especially at high temperature operation. But at lower temperature and when the array pixel size is smaller, Johnson noise becomes comparable to dark noise. Owing to stable surface properties, there is very little 1rf noise observed in QWIPs. At FPA level, the pattern noise Žwhich results from local variation of the dark current, photoresponse, and cutoff wavelength. is the major limitation to the array performance, especially at low temperature. This type of noise is a nonuniformity appearing across the array, which does not vary with time and reflects the intrinsic properties of a FPA. The fixed pattern noise is smaller for QWIP arrays than that of HgCdTe arrays due to their material quality and better-controlled cutoff wavelength. Fig. 7 compares the detectivities of p-on-n HgCdTe photodiodes with GaAsrAlGaAs QWIPs w15x. The theoretical curve for HgCdTe photodiodes are calculated assuming constant cutoff wavelengths 10 mm and 11 mm. The VLWIR results for HgCdTe Ž14.8 mm at 80 K and 16.2 mm at 40 K. and the QWIP at 16 mm show the intrinsic superiority of the HgCdTe photodiodes. HgCdTe has roughly an order of magnitude higher detectivity, though the advantage decreases as the temperature is reduced. Perhaps the best example of where the QWIP could have a performance advantage is at low temperature. As we see from Fig. 7, the QWIP at 7.7 mm peak wavelength offers superior performance relative to a f 10.6 mm HgCdTe at temperature F 45 K.

2.3. DetectiÕity We can distinguish two types of detector noise: radiation noise and intrinsic noise. Radiation noise includes signal fluctuation noise and background fluctuation noise. For infrared detectors, background fluctuation noise is higher compared to the signal fluctuation noise. Usually for photodiodes, shot noise is the major noise. In the case of QWIPs, the major source of noise is the dark current. Due to high dark current, Johnson

Fig. 6. Current density versus temperature for HgCdTe photodiode and GaAsrAlGaAs QWIP with l c s10 mm Žafter Ref. w14x..

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

284

Fig. 7. LWIR detector detectivity versus temperature for GaAsrAlGaAs QWIPs and p-on-n HgCdTe photodiodes Žafter Refs. w5,15x..

generation rate in QWIP is considerably larger than in HgCdTe. The intersubband lifetime in QWIP is inherently short Ž- 100 ps. which results in low quantum efficiency, while the carrier lifetime in LWIR HgCdTe is in microsecond range. In 1989 Kinch and Yariv w1x have estimated that for HgCdTe the thermal generation rate is approximately five orders of magnitude smaller than for the corresponding AlGaAsrGaAs superlattice. Improved material growth, device design and optimised doping made thermal generation rate in QWIP’s much smaller, only 10 times larger than in HgCdTe at 77 K w17x. Fig. 9 shows evolution of the performance of very long wavelength GaAsrAlGaAs QWIP. As can be seen, rapid progress has been made in detectivity of long wavelength QWIPs, starting with bound-tobound QWIPs, which had relatively poor sensitivity, and culminating in high performance bound-to-quasibound QWIPs with random reflectors. All the QWIP data is clustered between 10 10 and 10 11

The above performance comparison of QWIPs with HgCdTe in the low temperature range is less profitable for photodiodes in the case of nq-p structures. At low temperatures the p-type HgCdTe material suffers from same non-fundamental limitations Žcontacts, surface, Schockley–Read processes. more than the n-type one. Including the influence of tunnelling, the comparison of detectivity is more advantageous for GaAsrAlGaAs QWIPs in spectral region below 14 mm and at temperature below 50 K Žsee Fig. 8.. In general, the HgCdTe photodiodes are fundamentally superior to the various QWIPs representatively sampled in this database. Detectivity as the main parameter characterising normalised signal to noise performance of detectors and can be described as w16x DU s 0.31

l

a

ž /

hc G

1r2

Ž 1.

where a is the absorption coefficient and G is the generation rate. The ratio of absorption coefficient to the thermal generation rate, arG, is the fundamental figure of merit of any material for IR photodetectors, which directly determines the detectivity of optimised devices. While the absorption coefficients for HgCdTe and QWIP are comparable, the thermal

Fig. 8. Dependence of detectivity on the long wavelength cutoff for GaAsrAlGaAs QWIPs and nq-p HgCdTe photodiodes at temperatures F 77 K. The dashed lines are calculated for n-doped QWIPs. The theoretical lines for photodiodes are calculated assuming Auger limited diffusion current mechanism Ždashed lines. and diffusion and tunnelling mechanisms Ždash–dotted lines. Žafter Ref. w2x..

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

Fig. 9. Evolution of the performance of very long wavelength GaAsrAlGaAs QWIP. All the data is normalised to wavelength l s15.4 mm at temperature T s 55 K Žafter Ref. w6x..

cm Hz 1r2rW at about 77 K operating temperature. Gunapala and Bandara w6x assert that the highest possible DU using random gratings, for example, is about 2 = 10 11 cm Hz 1r2rW; this appears consistent with the data including the enhanced QWIP formed by both patterning the multiple quantum wells and reducing the number of wells. 2.4. Background limited performance When background-photon noise is the dominant noise mechanism, the detector is operating in an

Fig. 10. Estimation of the temperature required for background limited operation of different types of photon detectors. In the calculations FOVs 308 and TB s 300 K are assumed Žafter Ref. w2x..

285

ideal mode, and is said to exhibit background limited performance ŽBLIP.. BLIP temperature is defined that the device is operating at a temperature at which the dark current equals to the background photocurrent, given a field of view ŽFOV., and a background temperature. In Fig. 10 plots of the calculated temperature required for background limited operation in 308 FOV are shown as a function of cutoff wavelength. We can see that the operating temperature of ‘bulk’ intrinsic IR detectors ŽHgCdTe and PbSnTe. is higher than for other types of photon detectors. HgCdTe detectors with background limited performance operate with thermoelectric coolers in the MWIR range, instead the LWIR detectors Ž8 F lc F 12 mm. operate at f 100 K. HgCdTe photodiodes exhibit higher operating temperature compared to extrinsic detectors, silicide Schottky barriers and QWIPs. However, the cooling requirements for QWIPs with cutoff wavelengths below 10 mm are less stringent in comparison with extrinsic detectors and Schottky barrier devices. Next figure ŽFig. 11. shows both photon and dark current noise voltage for an 8.5 mm cutoff QWIP and a 12 mm cutoff HgCdTe FPA. The measurements are normalised to the same dark current levels by adjusting the readout gain for comparison. Due to the lower QE and higher dark current, the QWIP detectors become BLIP only at photons fluxes in excess of 2 = 10 13 photonsrcm2 s. By comparison, LWIR HgCdTe photodiodes are BLIP down to a photon flux of 1 = 10 10 photonsrcm2 s what indicate on their privilege position for strategic and

Fig. 11. Noise performance of typical HgCdTe photodiodes and QWIP Žafter Ref. w4x..

286

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

space applications, where longer-range detection at low backgrounds and with longer wavelength sensitivity is required.

3. Focal plane array performance The combination of existing high-performance arrays and highly developed silicon integrated circuits in the hybrid IR arrayrsilicon planes proved to be useful for thermal imaging systems with the thermal and spatial resolution unmatched by any competing technologies at present. In hybrid FPAs detectors and multiplexers are fabricated on different substrates and mated with each other by the flip-chip bonding. In this case we can optimise the detector material and multiplexer independently. Other advantages of the hybrid FPAs are near 100% fill factor and increased signal-processing area on the multiplexer chip. The detector array can be illuminated from either the frontside Žwith the photons passing through the transparent silicon multiplexer. or backside Žwith photons passing through the transparent detector array substrate.. In general, the latter approach is most advantageous, as the multiplexer will typically have areas of metallizations and other opaque regions, which can reduce the effective optical area of the structure. In HgCdTe hybrid FPAs, photovoltaic detectors are formed on thin HgCdTe epitaxial layer on transparent CdTe or ZnCdTe substrates. For HgCdTe flip-chip hybrid technology, the maximum chip size

is on the order of 10-mm square. In order to overcome this problem, the PACE ŽProducible Alternative to CdTe for Epitaxy. technology is being developed with sapphire or silicon as the substrate of HgCdTe detectors. When using opaque materials, substrates must be thinned to 10–20 mm in order to obtain sufficient quantum efficiencies and reduce the crosstalk. The progress in the LWIR HgCdTe photodiodes has been relatively slow until the recent development of molecular beam epitaxy ŽMBE. and metal-organic chemical vapour deposition ŽMOCVD. growth technologies. In October 1992 in USA, a consortium was assembled and supported by DARPA to develop p-on-n double layer heterostructure HgCdTe photodiode FPAs and two colour arrays w18x. The MBE technology gives HgCdTe more potential to produce high quality FPAs w19x. Up till now, 128 = 128 and 256 = 256 LWIR arrays are commercially available Žsee Table 2., while such large formats as 640 = 480 on silicon substrate have been demonstrated w20x. Also a prototype two-colour assembly with 128 = 128 pixel format has been produced to acquire IR imagery w21x. It should be insisted however, that the array size, uniformity, reproducibility, and yield are still difficult issues especially for low temperature and low background operation. The QWIP technology is relatively new that has been developed very quickly in the last decade. Large size LWIR GaAsrAlGaAs FPAs with up to 640 = 480 pixels have been demonstrated w22,23x,

Table 2 Performance specifications for LWIR HgCdTe FPAs Žafter SOFRADIR and SBRC data sheets. Parameter

SOFRADIR

SBRC

Configuration Waveband Žmm. Operating temperature ŽK. Detector pitch Žmm. Fill factor Ž%. Charge handling capacity Frame rate ŽHz. DU peak RMSrTintrpitch Žaverage. Žcm Hz 1r2 Wy1 . Pixel NETD Žaverage. NEI Žphotonsrcm2 s. Žmax. Fixed pattern noise Crosstalk Žoptical and electrical. Ž%. Array operability Ž%.

128 = 128 7.7–10.0 80 50 ) 70 ) 121 = 10 6 ey Ž19 pC. to 300 1.1 = 10 11 10 mK for 275 Hz

256 = 256 8.5–11.0 77 Žup to 100. 30 8 = 10 6 ey Žmin. to 120

1.52 = 10 12 7% RMS 2 99

99

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

287

Table 3 Properties of JPL 9-mm and 15-mm GaAsrAlGaAs QWIP FPAs Žafter Refs. w12,24x. Parameter

lc s 9 mm

lc s 15 mm

Array size Pixel pitch Žmm. Pixel size Žmm. Optical coupling Peak wavelength Žmm. Cutoff wavelength, 50% Žmm. Operability Ž%. Uncorrected nonuniformity Ž%. Corrected uniformity, 17–278C Ž%. Quantum efficiency Ž%. DU Žcm Hz 1r2 Wy1 . NETD ŽmK.

256 = 256 38 = 38 28 = 28 Aperiodic grating 8.5 8.9 99.98 6.8 0.05 6.9 2.3 = 10 11 Ž70 K. 40 Ž70 K.

128 = 128 50 = 50 38 = 38 Aperiodic grating 14.2 14.9 ) 99.9 2.4 0.2 3 1.6 = 10 10 Ž55 K. 30 Ž45 K.

with excellent uniformity and operability. Among the cooled IR detector systems, only HgCdTe and QWIP offer wavelength flexibility and multicolour capabilities. Tables 2 and 3 summarise the present status of the performance of both detector systems w12,24x. 3.1. Uniformity The nonuniformity value is usually calculated using the standard deviation over mean, counting the number of operable pixels in an array. For a system operating in the LWIR band, the scene contrast is about 2%rK of change in scene temperature. Therefore, to obtain a pixel to pixel variation in apparent temperature to less than, e.g., 20 mK, the nonuniformity in response must be less than 0.04%. This is nearly impossible to obtain in the uncorrected response of the FPA so a two-point correction is typically used. FPA uniformity influences an IR system complexity. The uniformity is important for accurate temperature measurements, background subtraction, and threshold testing. Nonuniformities require elaborate compensation algorithms to correct the image but by consuming a number of analog-to-digital bits, they also reduce the system dynamic range. Tactical IR FPAs usually require operating in the LWIR window with a small number of applications in the 3–5-mm window. Ranges from the sensor to the target are typically short allowing use of imaging sensors with large FPAs where precise radiometry is not critical. Imaging arrays can usually tolerate some percentage of dead or degraded pixels without jeop-

ardising mission performance. Tactical backgrounds in the IR windows are relatively high with about 10 16 photonsrcm2 s reaching the detector. Strategic IR FPAs have much greater restrictions on pixel uniformity since dead or degraded pixels could completely miss a target, and absolute radiometry, not imagery, is the method by which most target information is obtained. The space-based FPAs are usually designed with larger pixels, roughly equal to the blur spot of the optics. The nonuniformity can be different depending on the specification of operability; e.g., a higher requirements on the operability usually leads to a lower uniformity and vice verse. An example of this dependence, the corrected response nonuniformity as a function of operability for the centre 64 = 64 section of a Lockheed Martin 256 = 256 bound-tominiband FPA, is shown in Fig. 12. Note that the corrected response nonuniformity is less than 0.04% with only 10 pixels excluded, which means a 99.75% operability. If only 4 pixels are excluded, the operability is 99.90%. Typical uncorrected response nonuniformity in QWIP FPAs is 1–3% with an operability Žthe fraction of good pixels. greater than 99.5% w22x. For the 128 = 128 15-mm array fabricated by Jet Propulsion Laboratory Žsee Table 3., the uncorrected standard deviation is 2.4% and the corrected nonuniformity 0.2%. For recently described large 640 = 486 9-mm FPA the uncorrected noise nonuniformity is about 10%, and after two-point correction improves to an impressive 0.1% w23x. For the same format FPA recently demonstrated by

288

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

ating temperature. Due to lower quantum well barriers, the dark current of thermionic emission dominates at a lower temperature. In order to achieve equivalent performance of a 10-mm cutoff QWIP at 77 K, the temperature needs to be cooled down to 55 K for a 15-mm cutoff and 35 K for a 19-mm cutoff w4x Žsee Table 3 and Fig. 7.. 3.2. NEDT

Fig. 12. Corrected response uniformity versus operability of centre 64=64 section of a 256=256 bound-to-miniband FPA ŽFr1.7 optics.. Calibration is at scene temperatures of 20 and 408C. Nonuniformity is shown at the midpoint of the calibration interval Žafter Ref. w22x..

Lockheed Martin, the operability of greater than 99.98% was described w22x. It is very hard for HgCdTe to compete with QWIP for high uniformity and operability with large array format, especially at low temperature and VLWIR. High uniformity and high operability as shown in the above examples demonstrate the mature GaAs growth and processing technology. In this context, the nonuniformity and operability have been an issue for HgCdTe, although recently published values for Sofradir and SBRC arrays are as high as 99%. As we compare the performance of both types of FPAs Žsee Tables 2 and 3., the array operability is higher for QWIPs and is over 99.9%. The nonuniformity is a serious problem in the case of VLWIR HgCdTe detectors. The variation of x across the Hg 1y x Cd xTe wafer causes a much larger spectral nonuniformity Že.g., at 77 K, a variation of D x s 0.2% gives a D lc s 0.064 mm at lc s 5 mm, but D lc s 0.51 mm at 14 mm., which cannot be fully corrected by the two or three point corrections w3x. Therefore, the required composition control is much more stringent for VLWIR than for MWIR. In the case of QWIPs, extending of cut-off wavelength to VLWIR is relatively easier since there is little change in material properties, growth and processing. However, a serious requirement for maintaining the device performance is to lower the oper-

For FPAs the relevant figure of merit is the noise equivalent temperature difference ŽNEDT., the temperature change of a scene required to produce a signal equal to the rms noise. As was discussed by Levine w5x, when the detectivity is approaching value above 10 10 cm Hz 1r2rW Žsee Fig. 13., the FPA performance is uniformity limited and thus essentially independent of the detectivity. An improvement of nonuniformity from 0.1% to 0.01% after correction could lower the NEDT from 63 to 6.3 mK. For strategic FPAs the performance figure of merit is the noise equivalent irradiance ŽNEI., the radiant flux density necessary to produce a signal equal to the rms noise. The relationship between the NEI and NEDT is simple w3x NEDTs NEI

dF b

y1

ž /

Ž 2. dT where F b is the background photon flux. Usually NEI is limited by the temporal noise in which the

Fig. 13. Noise equivalent difference temperature as a function of detectivity. The effects of nonuniformity are included for us10y3 and 10y4 Žafter Ref. w5x..

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

dark current noise nonunifomity plays an important role w25x. Comparing the values of NEDT parameters for both type of FPAs Žsee Tables 2 and 3., we can see that the performance of LWIR HgCdTe arrays is better. 3.3. Charge handling capacity and integration time One of the simplest and most popular readout circuit for IR FPAs is the direct injection ŽDI. input, where dark current and photocurrent are integrated into a storage capacity. In this case, the bias varies across the array by about "Ž5–10. mV due to variations in the transistor thresholds. At 80 K, HgCdTe diodes show very little dependence of leakage current with small changes in reverse-bias near zero volts Žsee Fig. 2.. The goal is to fit as large a capacitor as possible into the unit cell, particularly for high tactical applications were signal-to-noise rations can be obtained through longer integration times. For high injection efficiency, the resistance of the FET should be small compared to the diode resistance at its operating point. Generally, is not problem to fulfil this inequality for MWIR HgCdTe staring designs where diode resistance is large Žthe R o A product is in the range above 10 6 V cm2 ., but it can be very important for LWIR designs where R diode is small Žthe R o A product is about 10 2 V cm2 .. LWIR HgCdTe photodiodes operate with a small reverse-bias near zero volts. In this case a large bias is desirable, but it strongly depends on the material quality of the array. For very high quality LWIR HgCdTe array, y1 V bias is possible. The bias on a QWIP array is larger than that of a HgCdTe array and usually is around 2 to 3 V. However, the readout power consumption is similar for QWIP and HgCdTe and is negligible compared with the readout electronics. For example, power dissipation for an imaging 640 = 480 QWIP FPA is - 150 mW w22x. It is result of very high impedance of QWIP, at the GV range at 77 K for a pixel size of 24 = 24 mm2 w3x. This high impedance makes the readout design very easy in achieving high efficiency; e.g., the injection efficiency of a 9-mm cutoff 640 = 486 QWIP array is 99.5% w22x. However, recently Singh and Cardimona w26x have uncovered possible problem connected with offsets Žup to f 1

289

V. in the low-temperature current–voltage characteristics Žnon-zero current at zero bias.. This observation indicates the presence of a time-constant effect due to the charging of a capacitance in the QWIP detector. The presence of large offsets, or large dynamic resistance, could create difficulties in matching the impedances of the FPA and ROIC especially at low biases and certainly at low temperatures. The DI input circuit is not generally used for low backgrounds due to injection efficiency issues. The strategic applications many times have low backgrounds and require low noise multiplexers interfaced to high resistance detectors. A commonly used input circuits for strategic applications is the capacitative transimpedance amplification ŽCTIA. input circuit w27x. Increasing of the injection efficiency is also possible using buffered DI. The latest two input circuits also accentuate the 1rf noise and the operability, but require higher power to operate. It can be shown that w28x y1

( /

NEDTs t o C Tlh BLIP NC

ž

Ž 3.

where t o is the optics transmission, C Tl is the thermal contrast, h BLIP is the percentage of BLIP, and NC is the number of photogenerated carriers integrated for one integration time t int at incident background flux F B as follows: NC s h A d t intF B

Ž 4.

From the above formulas results that the charge handling capacity of the readout and the integration time becomes the major issues of IR FPAs. The NEDT is inversely proportional to the square root of the integrated charge and therefore the greater the charge, the higher the performance. At high backgrounds, it is often impossible to handle the large amount of carriers generated over frame times compatible with standard video frame rates. The user is forced to integrate for only a fraction of the full frame time. Off-FPA integration of subframes is used to attain a level of sensor sensitivity commensurate with the detector-limited FPA detectivity and not the charge-handling-limited sensor detectivity. Fig. 14 shows the theoretical NEDT versus charge handling capacity for HgCdTe FPAs assuming that

290

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

Fig. 14. NEDT versus charge handling capacity Žafter Ref. w29x..

the integration capacitor is filled to half the maximum capacity Žto preserve dynamic range. under nominal operating conditions in different spectral bandpasses: 3.4–4.2 mm, 4.4–4.8 mm. 3.4–4.8 mm, and 7.8–10 mm w29x. Also shown are measured data for both TCM2800 at 95 K, other PACE HgCdTe FPAs at 78 K and representative LWIR FPAs. We can see that the measured sensitivities agree with the expected values. The well charge capacity is the maximum amount of charge that can be stored on the storage capacitor of each cell. The size of the unit cell is limited to the dimensions of the detector element in the array Žin the case of large LWIR HgCdTe hybrid array, a mismatch in the coefficient of thermal expansion between detector array and the readout can force the cell pitch to 20 mm or less to minimise lateral displacement.. However, the development of heteroepitaxial growth techniques for HgCdTe on Si has opened up the possibility of cost-effectively producing significant quantities of large-area arrays through utilisation of large-diameter Si substrates. Fig. 15 shows the charge handling capacity versus cell pitch. We can see that for a 30 = 30 mm2 pixel size, the storage capacities are limited to 1 to 5 = 10 7 electrons Žit depends on design feature; at a 0.8-mm ˚ to recoup feature if the oxide is thinned to 150 A capacity, the readout yield and device reliability can be reduced due to oxide damage from hot carriers w28x.. For example, for a 5 = 10 7 electron storage capacity, the total current density of a detector with a 30 = 30 mm2 pixel size has to be smaller than 27 mArcm2 with a 33 ms integration time w3x. If the

Fig. 15. Charge handling capacity versus cell pitch Žafter Ref. w28x..

total current density is in 1 mArcm2 range, the integration time has to be reduced to 1 ms. For the LWIR HgCdTe FPAs the integration time is usually below 100 ms. Since the noise power bandwidth D f s 1r2 t int , a small integration time causes extra noise in integration w3x. Usually, LWIR QWIP FPAs using conventional ROIC have typically operated at 60–65 K. Due to a smaller quantum efficiency of QWIP, filling the charge capacitor is not a problem at high background application. QWIP allows a longer integration time, which gives a relatively lower NEDT. However at higher temperatures, the dark current of QWIP is high and fills the charge capacitor very quickly. The current subtraction and switched capacitor noise filtering capabilities of more recent Lockheed Martin ROICs permit low NEDT at higher operating temperatures, what is shown in Fig. 16. In this case

Fig. 16. NEDT measurements for Lockheed Martin 64=64 QWIP FPA with current-subtracting ROIC. Data was acquired with a fr1.97 optical system Žafter Ref. w22x..

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

291

however, the readout circuit is complicated which limits the size of array. 3.4. Multi-band FPAs As the IR technology continues to advance, there is a growing demand for multispectral detectors for advanced IR systems with better target discrimination and identification. So far, the multiple waveband measurements have been achieved using separate FPAs with a diachronic filter, a mechanical filter wheel, or a dithering system with a striped filter w3x. These approaches are expensive in terms of size, complexity, and cooling requirements. At present considerable efforts are directed to fabricate a single FPA with multicolour capability to eliminate the spatial alignment and temporal registration problems that exist whenever separate arrays are used, to simplify optical design, and reduce size, weight, and power consumption w3x. Both HgCdTe and QWIP detectors offer the multicolour capability in the MWIR and LWIR range. Considerable progress has been recently demonstrated by research groups at Hughes Research Laboratory w30,31x and Lockheed Martin w32x in multispectral HgCdTe detectors employing MBE and MOCVD for the growth of variety devices. Devices for the sequential and simultaneous detection of two closely spaced sub-bands in the MWIR, sequential detection of MWIR and LWIR radiation as well as those for the sequential detection of two sub-bands in the LWIR have been demonstrated w31x. One of such two-colour HgCdTe detector is the bias-selectable back-to-back photodiode. This device is formed by sequentially growing n-p- p- n four-layer or n- p- n three-layer double heterojunction structures Žthe underlined letters mean the materials with larger bandgap energy. with two HgCdTe photodiodes w31x. In one mode, the MW junction is reverse-biased, producing a signal for the MWIR; and the LW junction is forward-biased, producing no signal. Reversing the bias polarity activates the LW junction but not the MW junction. The relative spectral response of this kind of detector is shown in Fig. 17. The LW detectors were fabricated with R o A ) 100 V cm2 Ž lc s 10.5 mm. and the MW detectors with R o A ) 10 5 V cm2 Ž lc s 5.7 mm. at 78 K w30x. Mitra et al. w32x have developed a new two-colour MWrLW HgCdTe IR detector which overcomes the

Fig. 17. Relative spectral response versus wavelength for dual-band ŽMWIR and LWIR. detectors operating at reverse and forward-bias modes Žafter Ref. w30x..

problems of the bias-selectable device. The detectors with two bumps per unit cell for electrical, mechanical, and thermal interconnections have been implemented in 64 = 64 dual-band hybrid FPAs, but in general the device performance is not quite as good as in a single-colour HgCdTe devices. The successful demonstration of the prototype 128 = 128 array operated as back-to-back diodes for the simultaneous detection of two closely spaced sub-bands in the MWIR spectrum has been given by Rajavel et al. w31x. The device structures were delineated as mesa isolated structures and contacts were made to the top n-type layer and the intermediate p-type layer. Fill factors as high as 80% were achieved by using a single mesa structure to accommodate the two indium bump contacts required for each unit cell. The devices were characterised by R o A product in excess of 5 = 10 5 V cm2 at 78 K, at fr2 FOV and quantum efficiency greater than 70% in each band. Imagery was acquired at temperature as high as 180 K without any visible degradation in image quality. Also several approaches to achieving multicolour QWIP detection have been proposed including two stack, two colour MWrLW QWIPs at the single device level with either three terminal simultaneous registration w13,23x, or voltage tuneable between MWIR and LWIR w33x. Also asymmetrically coupled quantum well structures are used to achieve voltage tuneable three-colour detection within one atmospheric window w34x.

292

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

Fig. 18. MWIR and LWIR response of dual-band FPA Žafter Ref. w22x..

The best quality two-colour QWIP FPAs has been fabricated by Lockheed Martin in a simple way—by stacking quantum well layers with the desired spectral responses. A near-infrared ŽNIR.rLWIR and MWIRrLWIR FPAs have been fabricated w22x. In the first case, a bound-to-miniband QWIP stack for LWIR detection and interband transitions in an epitaxial InGaAs layer for NIR detection were used. More recently, a proof of principle effort was undertaken which demonstrates the feasibility of a dual colour LWIRrMWIR QWIP FPA. The dual band detector array structure was designed to interface with the existing 512 = 512 LWIR design ROIC which is not optimised for MWIR or dual band operation. Integration time is extended during MWIR operation to compensate for the oversized charge well. Preliminary testing has verified spectral response in the 3–5 and 8–12-mm bands as shown in Fig. 18. Spectral cross-coupling between the LWIR and MWIR bands was measured at less than 0.3%. Preliminary NEDT data were acquired in both MWIR and LWIR bands with a fr3.4 wide band optical system and are as the following: in MWIR equal 0.015 K, and in LWIR equal 0.024 K. It should be marked that the research group at Jet Propulsion Laboratory is currently developing a 640 = 486 LWIR and VLWIR dual-band QWIP FPA camera w35x. Compared with HgCdTe, it seems that QWIP FPAs are more feasible to achieve multicolour detection in VLWIR region. 3.5. Cost The cost of a FPA depends strongly on the maturity of the technology and varies with production

quantity in different companies. So far, large size LWIR FPAs are developed in R & D laboratories without mass production experience w3x. According to Sofradir HgCdTe experience, the cost of making high performance cooled components can be broken down into three parts of about equal weight: the chip Ždetector and ROIC., the dewar, integration and tests w36x. In addition, the user must add the cryogenic machine cost that is not negligible compared to the component one’s. Even if the detection circuit is free of charge, the total cost would only be reduced by about 15–20%. This explains why the cost of PtSi and QWIP detectors is not markedly less than that of quantum detectors of the same complexity, even though the raw materials ŽSi or GaAs. is much less than for HgCdTe. Moreover, since PtSi requires a very wide optical aperture to obtain acceptable performance, and since QWIP requires a lower operating temperatures than other photon detectors, a possible reduction in the purchase price is counterbalance by a significant increase in operating cost. HgCdTe detectors have been the centre of a major industry by the last three decades. The technology is relatively mature at MWIR with size up to 640 = 480 w28x, but it does not fold over to LWIR. To make components with more pixels requires reducing the pitch or mastering the thinning operation needed to withstand the thermal cycling Ždifferential thermal expansion between CdZnTe and silicon.. In the future, more advantage approach seems to be using of Si substrates which offer many well-known advantages relative to bulk CdZnTe substrates Žmuch larger available size at lower cost, a thermal expansion match to Si readout chips, higher purity, and compatibility with automated wafer processingrhandling methodology due to their superior mechanical strength and flatness.. Development of LWIR and multicolour HgCdTe detectors are extremely difficult, especially for low background applications. In comparison with HgCdTe FPAs, the industrial experience in QWIP FPAs is lower and improvements can be expected because this technology is at a lower step of development. The major challenge is at the device and grating designs to improve the device performance. Because of the maturity of the GaAs growth technology and stability of the material system, no investment is needed for developing QWIP substrates, MBE growth, and processing tech-

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294

nology w3x. It means a lower cost in technology developing and production compared with HgCdTe. 3.6. Summary Comparing photovoltaic HgCdTe and QWIP technologies, we arrive at the following conclusions: Ø two major issues that impede the performance of QWIPs should be overcome: optical conversion efficiency and dark current, Ø in the case of HgCdTe, the improving of the array uniformity is necessary, Ø QWIP seems more potential to realise VLWIR FPA operation Žalso with multicolour detection.. In our discussion the reliability issue has been omitted due to fact that statistical data on this subject is not available. In several applications, especially military systems demand high reliability to ensure both the success of the mission and minimal risk to the user. Two reliability challenges affect both FPAs; survival in high temperature system storage environments and withstanding repetitive thermal cycles between ambient and cryogenic temperatures. In HgCdTe as well as in QWIP FPAs the indium bumps are used to hybridise both type of detectors with silicon multiplexer. However, certain problems can be expected in the case of QWIP arrays, since the indium bumps have many known alloys with III–V compounds.

293

Even though that QWIP is a photoconductor, several its properties such as high impedance, fast response time, long integration time, and low power consumption, well comply requirements of fabrication large FPAs. Due to the high material quality at low temperature, QWIP has potential advantages over HgCdTe for VLWIR FPA applications in terms of the array size, uniformity, yield and cost of the systems. QWIP FPAs combine the advantages of PtSi Schottky barrier arrays Žhigh uniformity, high yield, radiation hardness, large arrays, lower cost. with the advantages of HgCdTe Žhigh quantum efficiency and long wavelength response.. Up till now, the efforts in QWIP technology have been limited to increasing operation temperature for tactical applications. However, the discussion in this paper indicates that more efforts should be directed in order to develop QWIP toward VLWIR strategic applications.

Acknowledgements This work was partially supported by the KBN ŽPoland. under grant number PBZ 28.11rP6.

References 4. Conclusions The intention of this paper has been to provide information to compare two competitive technology, photovoltaic HgCdTe to QWIP, in LWIR and VLWIR spectral ranges with emphasise on the material properties, device structure, and their impact on FPA performance. From the above discussion results that LWIR QWIP cannot compete with HgCdTe photodiode as the single device especially at higher temperature operation Ž) 70 K. due to fundamental limitations associated with intersubband transitions. However, the advantage of HgCdTe is less distinct in temperature range below 50 K due to problems involved in a HgCdTe material Žp-type doping, Shockley–Read recombination, trap-assisted tunnelling, surface and interface instabilities..

w1x M.A. Kinch, A. Yariv, Appl. Phys. Lett. 55 Ž1989. 2093. w2x A. Rogalski, Infrared Phys. Technol. 38 Ž1997. 295. w3x M.Z. Tidrow, presented during 6th Annual AIAArBMDO Technology Readiness Conference and Exhibit, San Diego, CA, August 18–22, 1997. w4x P.R. Norton, Proc. SPIE 3379 Ž1998. 102. w5x B.F. Levine, J. Appl. Phys. 74 Ž1993. R1. w6x S.D. Gunapala, K.M.S.V. Bandara, in: M.H. Francombe, J.L. Vossen ŽEds.., Thin Films, Vol. 21, Academic Press, New York, 1995, pp. 113–237. w7x M.Z. Tidrow, Proc. SPIE 2999 Ž1996. 109. w8x A. Rogalski, R. Ciupa, J. Appl. Phys. 77 Ž1995. 3505. w9x A. Rogalski, R. Ciupa, J. Appl. Phys. 80 Ž1996. 2483. w10x M.B. Reine, K.R. Maschhoff, S.P. Tobin, P.W. Norton, J.A. Mroczkowski, E.E. Krueger, Semicond. Sci. Technol. 8 Ž1993. 788. w11x J. Bajaj, J.M. Arias, M. Zandian, J.G. Pasko, L.J. Kozlowski, R.E. De Wames, W.E. Tennant, J. Electron. Mater. 24 Ž1995. 1067. w12x S.D. Gunapala, J.K. Liu, J.S. Park, M. Sundaram, C.A.

294

w13x w14x w15x w16x w17x w18x

w19x

w20x

w21x

w22x w23x

A. Rogalskir Infrared Physics & Technology 40 (1999) 279–294 Shott, T. Hoelter, T.L. Lin, S.T. Massie, P.D. Maker, R.E. Muller, G. Sarusi, IEEE Trans. Electron Devices 44 Ž1997. 51. M.Z. Tidrow, J.C. Chiang, S.S. Li, K. Bacher, Appl. Phys. Lett. 70 Ž1997. 859. A. Singh, M.O. Manasreh, Proc. SPIE 2397 Ž1995. 193. L.J. Kozlowski, K. Vural, J.M. Arias, W.E. Tennant, R.E. DeWames, Proc. SPIE 3182 Ž1997. 2. J. Piotrowski, W. Gawron, Infrared Phys. Technol. 38 Ž1997. 63. K.K. Choi, C.Y. Lee, M.Z. Tidrow, W.H. Chang, S.D. Gunapala, Appl. Phys. Lett. 65 Ž1994. 1703. J.D. Benson, J.H. Dinan, J. Waterman, C.J. Summers, R.D. Benz, B.K. Wagner, S.D. Pearson, A. Parikh, J.J. Jensen, O.K. Wu, R.D. Rajavel, G.S. Kamath, K.A. Harris, S.R. Jost, J.M. Arias, L.J. Kozlowski, M. Zandian, J. Bajaj, K. Vural, R.E. DeWames, C.A. Cockrum, G.M. Venzor, S.M. Johnson, H.D. Shih, M.J. Bevan, J.A. Dodge, A. Simmons, Proc. SPIE 2744 Ž1996. 126. O.K. Wu, T.J. DeLyon, R.D. Rajavel, J.E. Jensen, in: P. Capper ŽEd.., Narrow-Gap II–VI Compounds for Optoelectronic and Electromagnetic Applications, Chapman & Hall, London, 1997, pp. 97–130. T.J. DeLyon, M.D. Gorwitz, J.E. Jensen, C.A. Cockrum, S.M. Johnson, G.M. Venzor, presented during The 1998 US Workshop on the Physics and Chemistry of II–VI Materials, Charleston, October 20–22, 1998. R.D. Rajavel, D.M. Jamba, J.E. Jensen, O.K. Wu, P.D. Brewer, J.A. Wilson, J.L. Johnson, E.A. Patten, K. Kosai, J.Y. Caulfield, P.M. Goetz, J. Electron. Mater. 27 Ž1998. 7471. W.A. Beck, T.S. Faska, Proc. SPIE 2744 Ž1996. 193. S.D. Gunapala, S.V. Bandara, J.K. Liu, W. Hong, E.M. Luong, J.M. Mumolo, M.J. McKelvey, D.K. Sengupta, A. Singh, C.A. Shott, R. Carralejo, P.D. Maker, J.J. Bock, M.E.

w24x

w25x

w26x w27x w28x

w29x w30x

w31x

w32x

w33x w34x w35x

w36x

Ressler, M.W. Werner, T.N. Krabach, Proc. SPIE 3379 Ž1998. 382. S.D. Gunapala, J.S. Park, G. Sarusi, T.L. Lin, J.K. Liu, P.D. Maker, R.E. Muller, C.A. Shott, T. Hoelter, IEEE Trans. Electron Devices 44 Ž1997. 45. R.L. Whitney, K.F. Cuff, F.W. Adams, Semiconductor Quantum Wells and Superlattices for Long-Wavelength Infrared Detectors, Artech House, Boston, 1993, pp. 55–108. A. Singh, D.A. Cardimona, Proc. SPIE 2999 Ž1997. 46. L.J. Kozlowski, S.A. Cabelli, D.E. Cooper, K. Vural, Proc. SPIE 1946 Ž1993. 199. L.J. Kozlowski, R.B. Balley, S.E. Cooper, I.S. Gergis, A.C. Chen, W.V. McLevige, G.L. Bostrup, K. Vural, W.E. Tennant, P.E. Howard, Opt. Eng. 33 Ž1994. 54. L.J. Kozlowski, Proc. SPIE 3179 Ž1997. 200. R.D. Rajavel, D.M. Jamba, O.K. Wu, J.E. Jensen, J.A. Wilson, E.A. Patten, K. Kosai, P. Goetz, G.R. Chapman, W.A. Radford, J. Crystal Growth 175 Ž1997. 653. R.D. Rajavel, D.M. Jamba, J.E. Jensen, O.K. Wu, P.D. Brewer, J.A. Wilson, J.L. Johnson, E.A. Patten, K. Kasai, J.T. Caulfield, P.M. Goetz, J. Electron. Mater. 27 Ž1998. 747. P. Mitra, S.L. Barnes, F.C. Case, M.B. Reine, P. O’Dette, R. Starr, A. Hairston, K. Kuhler, M.H. Weiler, B.L. Musicant, J. Electron. Mater. 26 Ž1997. 482. M.Z. Tidrow, K.K. Choi, A.J. DeAnni, W.H. Chang, S.P. Svensson, Appl. Phys. Lett. 67 Ž1995. 1800. J.C. Chiang, S.S. Li, M.Z. Tidrow, P. Ho, M. Tsai, C.P. Lee, Appl. Phys. Lett. 69 Ž1996. 2412. S.D. Gunapala, S.V. Bandara, A. Singh, J.K. Liu, E.M. Luong, J.M. Mumolo, M.J. McKelvey, Proc. SPIE 3379 Ž1998. 225. F. Bertrand, J.T. Tissot, G. Destefanis, in: V. Kumar, S.K. Agarwal ŽEds.., Physics of Semiconductor Devices, Vol. II, Narosa Publishing House, New Delhi, 1998, pp. 713–720.