Recent progress in InSb based quantum detectors in Israel

Infrared Physics & Technology 59 (2013) 172–181

Contents lists available at SciVerse ScienceDirect

Infrared Physics & Technology journal homepage: www.elsevier.com/locate/infrared

Recent progress in InSb based quantum detectors in Israel Philip Klipstein a,⇑, Daniel Aronov a, Michael ben Ezra a,b, Itzik Barkai a, Eyal Berkowicz a, Maya Brumer a, Rami Fraenkel a, Alex Glozman a, Steve Grossman a, Eli Jacobsohn a, Olga Klin a, Inna Lukomsky a, Lior Shkedy a, Itay Shtrichman a, Noam Snapi a, Michael Yassen a, Eliezer Weiss a a b

SemiConductor Devices, P.O. Box 2250, Haifa 31021, Israel Israel MOD

a r t i c l e

i n f o

Article history: Available online 29 December 2012 Keywords: Infrared detector Focal plane array InSb Bariode InAsSb XBn

a b s t r a c t InSb is a III–V binary semiconductor material with a bandgap wavelength of 5.4 lm at 77 K, well matched to the 3–5 lm MWIR atmospheric transmission window. When configured as a Focal Plane Array (FPA) detector, InSb photodiodes offer a large quantum efficiency, combined with excellent uniformity and high pixel operability. As such, InSb arrays exhibit good scalability and are an excellent choice for large format FPAs at a reasonable cost. The dark current is caused by Generation–Recombination (G–R) centres in the diode depletion region, and this leads to a typical operating temperature of 80 K in detectors with a planar implanted p–n junction. Over the last 15 years SCD has developed and manufactured a number of different 2-dimensional planar FPA formats, with pitches in the range of 15–30 lm. In recent years a new epi-InSb technology has been developed at SCD, in which the G–R contribution to the dark current is reduced. This enables InSb detector operation at 95–100 K, with equivalent performance to standard InSb at 80 K. In addition, using a new patented XBnn device architecture in which the G–R current is totally suppressed, epitaxial InAsSb detectors have been developed with a bandgap wavelength of 4.2 lm, which can operate in the 150–170 K range. In this short review of the past two decades, a number of key achievements in SCD’s InSb based detector development program are described. These include High Operating Temperature (HOT) epi-InSb FPAs, large format megapixel FPAs with high functionality using a digital Read Out Integrated Circuit (ROIC), and ultra low Size, Weight and Power (SWaP) FPAs based on the HOT XBnn architecture. Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction The Mid Wave Infra-Red (MWIR) ‘‘window’’ in the atmosphere spans optical wavelengths from 3 to 5 lm. High performance MWIR photodiode Focal Plane Array (FPA) detectors that operate in this window are usually made from Mercury Cadmium Telluride (MCT) or Indium Antimonide (InSb). The best MCT FPAs are grown on expensive Cadmium Zinc Telluride substrates and can exhibit a dark current that is close to the theoretical ‘‘diffusion limit’’ [1]. In contrast, even the best InSb FPAs are generation-recombination (G–R) limited [2], whereby mid-bandgap Shockley–Read–Hall traps [3] act as ‘‘stepping stones’’ for the creation of electron–hole pairs in the diode depletion region. The dark current in InSb FPAs is therefore significantly higher. Nevertheless, both InSb and MCT detectors must be cooled cryogenically, typically with a miniature Joule Thomson or Stirling Cycle refrigerator, in order to achieve a good signal to noise ratio (SNR). The higher dark current in InSb is usually suppressed by operating the FPA some tens of Kelvin de⇑ Corresponding author. E-mail address: [email protected] (P.C. Klipstein). 1350-4495/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.infrared.2012.12.035

grees colder than the equivalent diffusion limited MCT device. This apparent disadvantage is offset by the higher uniformity and significantly lower cost per pixel of InSb, making it the current favourite for large scale staring arrays [4]. At SCD we have been manufacturing two dimensional InSb FPAs for more than 10 years. Our standard base technology involves beryllium implanted p–n junctions bonded to a silicon Read Out Integrated Circuit (ROIC) which is designed in-house. This technology has proved to be very robust and reliable. We currently offer a range of formats with both analogue or digital ROICs containing up to 1.3 MegaPixels, and operating in the range 80–85 K. Digital ROICs are ideal for large format small pitch applications, where the well capacity is small, because they offer some of the lowest noise solutions at the system level. In 2002, SCD was one of the first companies to introduce a digital ROIC [5,6]. We have adopted two approaches for increasing the operating temperature in InSb like FPAs, in order to close the gap with MCT while maintaining all of the advantages of InSb, including scalability to large areas at affordable cost. In order to do this, the dark current must be reduced. The first approach involves reducing the concentration of G–R centres at the p–n junction of our InSb diodes, while

173

the second uses a new patented InAs1xSbx n-type ‘‘bariode’’ technology to eliminate the G–R current entirely [7,8]. Bariodes (‘‘Barrier-diodes’’) can provide the same low diffusion limited dark current as MCT photodiodes whose composition corresponds to the same bandgap. However, they utilize III–V semiconductor materials and also offer processing advantages. The new n-type bariode detector, also known as an XBnn detector [9], has a cut-off wavelength of approximately 4.1 lm and is based on a heterostructure design that can be grown with high quality on commercially available substrates. The XBnn detector contains an n- or p-type contact layer made from InAs0.91Sb0.09 or GaSb (X), a barrier layer made from n-type AlSb0.91As0.09 (Bn), and an active layer made from n-type InAs0.91Sb0.09 (n). These compositions are all closely lattice matched to GaSb, which can be used as the substrate. Fig. 1 shows a schematic band diagram for an nBnn device (X = n-type InAs0.91Sb0.09) at operating bias, where no depletion exists in the active layer. Note that all members of the n-type bariode family contain the same n-type barrier and active layer unit (‘‘Bnn’’). The barrier blocks the flow of majority carriers, while minority carriers have a clear path through the structure. Both of our new technologies involve MBE growth. In the case of InSb diodes, a high quality homostructure is grown on an InSb substrate and diodes are isolated by etching a mesa structure through the p–n junction. Because implantation damage is avoided, they have a much lower concentration of G–R centres than our standard planar p–n junctions. The dark current is thus reduced according to the ratio of concentrations of G–R centres in the standard and MBE grown structures. This is shown schematically by the grey curves in Fig. 2, which display typical plots of the logarithm of the dark current vs. the reciprocal of the temperature. The steeper part of each curve is Diffusion limited dark current and the less steep part is G–R limited. For a given dark current the operating temperature is increased, according to the grey horizontal arrow drawn in Fig. 2. Note that for epi-InSb the cut-off wavelength is 5.4 lm, as for standard planar InSb. In an XBnn detector, the bulk G–R current is totally suppressed by excluding the depletion electric field from the narrow bandgap photon absorbing active layer material (see Fig. 1). The black lines in Fig. 2 show a comparison between an ideal standard InAs0.91Sb0.09 diode and an InAs0.91Sb0.09 n-type bariode. Note that the crossover temperature T0 between Diffusion and G–R limited dark current is higher in InAs0.91Sb0.09 because it has a larger bandgap than InSb [9]. Since the operating point in the bariode is only limited by diffusion, the operating temperature is increased relative to the diode, according to the horizontal black arrow drawn in Fig. 2. In this way we have achieved InAs0.91Sb0.09 bariode detector operation at F/3 with background limited performance (BLIP) at

Log I

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

Diffusion

InAsSb Bariode

Epi-InSb diode

InAsSb diode

T0-1

Implanted planar InSb diode

G-R

1/T

Fig. 2. Schematic Arrhenius plot of the temperature dependence of the dark current in InSb diodes in which the p–n junction is made by ion-implantation or grown by MBE (Grey), and InAs0.91Sb0.09 diodes and bariodes grown by MBE (Black). The diffusion and G–R limited portions of the InAs0.91Sb0.09 curves are labeled.

temperatures exceeding 160 K, more than doubling the traditional InSb operating temperature. Note that in both epi-InSb and XBnn InAs0.91Sb0.09, the devices can be operated at a lower temperature, more typical of planar InSb, in which case they both have an extremely small dark current. This is good for applications where the incident photon intensity is very low and the dark current would otherwise limit the SNR. The reductions in dark current and the corresponding increases in operating temperature that can be achieved with our MBE based epi-InSb diode and InAs0.91Sb0.09 bariode technologies have important consequences on the size (S), weight (W) and required cooling power (P), or SWaP, of the Integrated Dewar Cooler Assemblies (IDCAs) made with these FPA detectors. Table 1 compares some of the key IDCA parameters related to closed cycle Stirling cooler performance, for the standard planar and MBE based FPAs. The successive increases in operating temperature on going from planar to epi-InSb diodes to InAs0.91Sb0.09 bariodes deliver a considerable reduction in required cooling power. Size and weight are also reduced substantially (for example using the Ricor K562S cooler instead of K561), leading to an IDCA with a much reduced SWaP. Table 1 also shows that the cooler (and IDCA) reliability increases significantly as the operating temperature is raised. In the remainder of this paper we begin by analyzing the relative system advantages of choosing a detector with a cut-off wavelength appropriate to either InSb or InAs0.91Sb0.09. We then present some key radiometric parameters for our implanted or MBE based detectors, based on each of these sensing materials and with pitches down to 15 lm. The system analysis is presented in Section 2. The radiometric performance of a planar Megapixel

Table 1 Values of the IDCA parameters related to closed cycle Stirling cooler performance at an ambient temperature of 23 °C, for planar and epi-InSb diodes and InAs0.91Sb0.09 bariodes. The similar cool down times for planar and epi-InSb reflect the trade-off between a larger cooler for a smaller operating temperature and vice versa.

V

n

Bn

(Contact (Barrier layer) layer)

n (Active layer)

Fig. 1. Schematic band diagram of an nBnn bariode at operating bias.

Parameter

Planar InSb

Epi-InSb

InAs0.91Sb0.09 bariode

Ricor cooler type (max heat load) Weight FPA operating temperature Cooling power Cool down time Reliability (MTTF)

K561 (0.3 W)

K562S (0.25 W)

K562S (0.25 W)

290 gm 77 K

185 gm 95 K

185 gm 150 K

5W 6 min 5000 h

4W 6 min 5800 h

2W 3 min >10,000 h

174

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

(Hercules) detector is discussed in Section 3. This is followed by a section on the design of our digital ROICs. The performance of epiInSb FPAs appears in Section 5. Next, the performance of InAs0.91Sb0.09 bariode test structures and FPAs is presented in Section 6. Our conclusions are summarized in Section 7. 2. System level performance vs. cut-off wavelength When going to shorter MWIR cut-off wavelengths, the photon flux is reduced significantly. Combined with the small pixel area, it would seem at first sight that a higher Noise Equivalent Temperature Difference (NETD) would result and the radiometric performance would be degraded at the system level. However, the short cutoff wavelength turns out to have major advantages for applications related to long range observation. In standard atmospheric conditions, the optical transmission at wavelengths shorter than that of the CO2 absorption line (4.2–4.4 lm) is much higher than the transmission at longer wavelengths. For example, in the ninth line of Table 2, a comparison is made between two detectors, one having a cut-off wavelength of 4.9 lm and a calculated NETD of 23 mK, and the other with a cutoff wavelength of 4.1 lm and a calculated NETD of 63 mK. Due to the better atmospheric transmission at shorter cut-off wavelengths, the detection ranges of both detectors are similar. It should be noted that shorter wavelengths result in smaller diffraction effects, and the resulting improvement in spatial resolution also contributes to an increased detection range. In the rest of this section, simulations of system performance are presented for two spectral windows, 3.4–4.1 lm and 3.4– 4.9 lm, corresponding to the sensitivity ranges for XBnn and InSb detectors, respectively (for InSb, an external cold filter is used to cut-off the radiation at 4.9 lm). In the simulations, the detector dark current in the first window is that typical of an XBnn detector operating at 150 K, while in the second it is typical of an epi-InSb detector operating at 95 K or a planar InSb detector operating at 80 K. At low environmental temperatures (20 °C), however, the XBnn temperature is lowered to 140 K, in order to reduce the dark current further and improve the SNR. The dark current behaviour of the two detector families will be discussed below, in Sections 5 and 6. For the system level performance, we first calculate the SNR as a function of the cut-off wavelength. This calculation takes atmospheric absorption into account, which reduces the signal especially

for long target distances, and also atmospheric emission and the emission of the optics, which both contribute to the noise. In the following calculations, two different systems are considered. One is a hand held system with F/4, 200 mm optics, operated at 60 Hz and an integration time of 16 ms. The other system is for long range observation applications with F/5.5, 900 mm optics. This system works at a slower frame rate of 25 Hz with an integration time of 40 ms. In Fig. 3, the SNR is presented versus cut-off wavelength (a) for the hand held detector system looking at scenery 5 km away, and (b) for the long range observation system looking at scenery 15 km away. For each system, three environmental temperatures are considered, 20 °C, 0 °C and 27 °C, and a temperature contrast of 1 °C is assumed in the landscape. In the simulation the same temperature is used for the landscape, the atmosphere and the optics. US Standard atmospheric absorption spectra are used for hot scenery at 27 °C and a sub-arctic atmosphere is used for cold environmental temperatures of 0 °C and 20 °C. It is possible to see from Fig. 3, that there is an improvement in the SNR at first, as the cut-off wavelength increases, but then the SNR decreases near the CO2 absorption line, due to the combined effects of stronger absorption and stray light from the atmosphere that increases the noise without contributing to the signal. As the cut-off wavelength is increased further, the SNR starts to increase again. At a high environmental temperature, it is apparent from Fig. 3, that the SNR in the 3.4–4.1 lm window is better than SNR in the full MWIR window. At lower scene temperatures however (0 °C and 20 °C), the detrimental effect of atmospheric absorption above the CO2 line is reduced, and total flux also plays a dominant role, so the resulting SNR is higher in the full MWIR window. Nevertheless, it is clear that the SNR in the 3.4–4.1 lm band is greater than four in all cases, which is sufficient to allow a good quality image even in low temperature conditions. We believe that SNR = 2 is the threshold for getting a landscape image in the system. In Table 2, the detection and identification ranges at 90% probability are compared for two detectors in the two spectral windows (3.4–4.1 lm and 3.4–4.9 lm). The comparison is made for the two systems mentioned above and for two standard targets: a NATO panel of size 2.3 m  2.3 m and a human target 1.5 m  0.5 m, with temperature differences relative to their surroundings of 1.25 °C and 5 °C, respectively. The same environmental temperature range is considered as above (20 °C, 0 °C and 27 °C). It is apparent from the Table that in most scenarios both spectral bands have similar

Table 2 A comparison between the detection and identification ranges at 90% probability for two detectors (short and long range) in two spectral windows (3.4–4.1 lm and 3.4–4.9 lm). The comparison is made for three environmental temperatures (20 °C, 0 °C and 27 °C) and two standard targets: a NATO panel of size 2.3 m  2.3 m and a human target, 1.5 m  0.5 m, with temperature differences relative to surroundings of 1.25 °C and 5 °C, respectively. System-target

3.4–4.1 lm

3.4–4.9 lm

Detector NETD (mK)

Detection (90%) (km)

Identification (90%) (km)

Detector NETD (mK)

Detection (90%) (km)

Identification (90%) (km)

27 °C 1. Long range – NATO 2. Hand held – NATO 3. Long range – human 4. Hand held – human

26 29 26 29

27.8 10.2 15.6 4.6

6.7 1.9 2.8 0.8

19 18 19 18

23 9.3 15.1 4.5

6.4 1.9 2.8 0.8

0 °C 5. Long range – NATO 6. Hand held – NATO 7. Long range – human 8. Hand held – human

40 45 40 45

34 10.5 15.9 4.5

6.7 1.8 2.7 0.8

17 18 17 18

32.3 10.9 15.7 4.6

6.8 1.9 2.8 0.8

20 °C 9. Long range – NATO 10. Hand held – NATO 11. Long range – human 12. Hand held – human

63 74 63 74

31.4 9.9 15.6 4.3

6.5 1.7 2.7 0.7

23 26 23 26

31 10.2 15.5 4.6

6.6 1.9 2.8 0.8

175

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

14

25

5Km, short range, 27C 5Km, short range, 0C 5Km, short range, -20C SNR=2

20

15Km, long range, 27C 15Km, long range, 0C 15Km, long range, -20C SNR=2

12 10

SNR

SNR

15

8 6

10 4

5

0 3.6

2

3.8

4

4.2

4.4

4.6

4.8

5

0 3.6 3.8

4

4.2 4.4 4.6 4.8

Cut-off Wavelength (μm)

Cut-off Wavelength (μm)

(a)

(b)

5

Fig. 3. SNR versus cutoff for (a) a short range hand held system looking at scenery 5 km away and (b) a long range observation system looking at scenery 15 km away. For each system, three environmental temperatures are tested (20 °C, 0 °C and 27 °C) where a temperature contrast of 1 °C in the landscape is assumed. In each plot, the dashed line indicates SNR = 2, which is taken to be the threshold for a meaningful image.

performance, except for the high temperature and long range application, where the spectral band of 3.4–4.1 lm has an advantage due to relatively high atmospheric absorption above the CO2 line in a standard atmosphere. 3. Digital ROICs for InSb based FPAs Before discussing the performance and sensitivity of our InSb based detector arrays, we first present some details of the digital ROICs that we have been developing to register their signal. This development is important because by passing functionality from the system to the silicon chip, it is possible to reduce overall system power consumption and size, while ensuring signal integrity and stability to environmental conditions. In 2002, SCD introduced the first digital detector, which had 640  512 pixels and a 20 lm pitch [5,6]. A block diagram of the

Digital Focal Plane Processor is shown in Fig. 4. Since then a variety of other digital detector formats were developed with the Sebastian 20 lm pitch technology: 640  512, 480  384 and 320  256. The Megapixel Hercules detector which is discussed in the next section maintains the same functionality as previously, but in a smaller pitch (15 lm), by using 0.18 lm CMOS technology [10]. Along with this 1280  1024 pixel detector, we have also developed the smaller Pelican-D ROIC [11], which has a format of 640  512, and the SNIR (Short/Near IR) [12] which includes asynchronous laser pulse detection and two dimensional range finding capabilities performed in the ROIC. The 15 lm pitch Hercules digital ROIC has 2560 A/D converters, which allows full frame readout at a rate of 100 Hz. Despite the array size and the high data rate of the ROIC, the resultant total power consumption of the ROIC is quite modest: 80 mW for 60 Hz and 130 mW for 100 Hz frame rates. In the next section

Fig. 4. Block diagram of the Sebastian 20 lm pitch 640  512 DFPP.

176

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

Fig. 5. Hercules NETD map at 50% well-fill, F/4 and 77 K. The colour scale is in mK. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4. Hercules megapixel InSb FPA An important trend in the current development of infrared (IR) detector arrays is to increase the number of pixels and reduce their size. In this way it is possible to increase both the spatial resolution of the image and the field of view (FOV). Alternatively it is possible to reduce the overall system size by maintaining a similar resolution with smaller and lighter optics (shorter focal length and smaller optical diameter). In this section we present some of the novel features of SCD’s mega-pixel InSb detector, Hercules, which has a 1280  1024 pixel format and a 15 lm pitch [10, 13]. When the Hercules ROIC described in the previous section performs pixel readout, the introduction of additional noise components is suppressed because the data is extracted in digital format. Fig. 5 shows an FPA map of the NETD in each pixel at 50% well-fill. Due to the low readout noise and the low pixel dark current, this NETD corresponds to background limited performance (BLIP), i.e. due only to the shot noise of the arriving photons. As can be seen in the NETD map, there are no spatial features in the temporal noise, indicating no significant additional noise mechanisms. Hercules also demonstrates excellent linearity over almost its full dynamic range. The deviation from linearity is less than 0.06% of the full dynamic range from 5% to 90% capacitor well-fill. One should note that this level of performance is actually achieved at the system level, since Hercules is a digital ROIC. In order to withstand harsh environmental conditions while maintaining a high level of performance, great care was taken over the mechanical design of the complete IDCA. In spite of its large size, the Hercules FPA is integrated into a standard Dewar, based on a rugged external envelope and rugged supporting struts connecting to the cold finger. The struts are made from a material with a high stiffness and a low heat conductivity. The structure and geometry are optimized for a high natural frequency and low heat conduction. When the Dewar is assembled with the Ricor K508 half Watt standard cryo-cooler, the natural frequency of the Dewar is

about 1500 Hz. This results in a lateral movement of the FPA of less than one third of a pixel, when subjected to rough vibrations in the frequency range of 5–2000 Hz. In addition, the radiation cold shield was designed using ray tracing simulation so that it would stand up well to typical environmental temperature changes. Testing of the IDCA is performed in an environmental chamber. The detector is cooled down and the Non-Uniformity Correction (NUC) coefficients are calculated at an ambient temperature of 25 °C, using a standard two point correction procedure [14]. The temperature in the chamber is then varied. The Residual Non-Uniformity (RNU) determined at different temperatures using the same 25 °C NUC coefficients is plotted versus the pixel signal capacitor well-fill in Fig. 6. We find that when the ambient temperature is increased by as much as 20 °C, the RNU only degrades slightly, and only in the region of the two NUC correction points. We relate this minor RNU degradation to IR radiation emission from the window of the Dewar, which is in the FOV of the FPA. Stray light from other parts of the Dewar is thought to be negligible. The pixel operability (percentage of working pixels) at the FPA operating temperature of 80 K is above 99.5%. This is based on a 0.10

RNU ['% of Dynamic Range]

we demonstrate some of the radiometric properties of this detector.

25C 35C 45C

0.08

0.06

0.04

0.02

0.00

0

20

40

60

80

100

Well Fill [%] Fig. 6. RNU versus well-fill at different ambient temperatures using the same correction coefficients that were determined at an ambient temperature of 25 °C.

177

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

0.2

1.E+03

Epi-InSb

Idark (T) /I dark (epi, 95K)

Epi-InSb Planar InSb

RNU (Std/DR) (%)

Planar InSb 1.E+02

1.E+01

0.1

0

1.E+00

75

80

85

90

100

95

Temperature (K) 1.E-01 9

10

11

12

13

14

Fig. 9. V-curves for implanted planar InSb (open) and epi-InSb (solid) Pelican FPAs, calibrated by a standard two point correction at either 80 K or 95 K.

1000/T (K -1) 80

100

70

Operability

5. Performance of epi-InSb FPA

NETD (mK)

60

standard set of criteria for the identification of defective pixels that is used on our production line.

98

NETD

96

50 94 40 92 30 90

20

88

10 0 70

Normalised Pixel Count

100

110

86 120

1.0 0.9

Epi-InSb

0.8

Planar InSb

0.7 0.6 0.5 0.4

30 μm

0.2 Epi InSb (95K)

90

Fig. 10. Temperature dependence of the NETD and the pixel operability of an epiInSb Pelican FPA, measured with an aperture of F/4.1 in front of a Black Body at a temperature of 30 °C.

0.3 1.2

80

Temperature (K)

MTF

In the previous section, we presented the performance of a large format FPA, based on our planar implanted InSb technology. In this section we show some of the advantages of our epi-InSb technology, particularly for applications where a higher operating temperature is desirable, and compare detector performance with that typical of similar planar arrays. The temperature dependence of the dark current averaged over all pixels of 640  512 Pelican FPAs with a 15 lm pitch is compared in Fig. 7 for planar- and epi-InSb detectors. The epi-InSb FPA was fabricated by mesa etching an InSb epitaxial p–n structure grown on a 300 InSb substrate in a Veeco Gen 200 MBE machine. For ease of comparison, the dark current has been normalized to that at the epi-InSb FPA operating temperature, T = 95 K (1000/ T = 10.53). The lines through the points are fits to the data based on the G–R formula: I = I0T3/2 exp(DE/k T) where k = Boltzman’s constant. In both cases, DE = 0.12 eV, which corresponds to approximately half the bandgap of InSb at low temperatures, exactly as expected for G–R limited behaviour [15]. The prefactors, I0, are in the ratio 17.3, which represents the factor by which the

Operability (%)

Fig. 7. Temperature dependence of the mean dark current of Pelican FPAs (15 lm pitch), with planar (open points) and epi-InSb (solid points) diodes. Fits to the G–R formula discussed in the text are shown as dashed and solid lines, respectively.

20 μm

15 μm

0.1

Planar InSb (80K)

1

0 0.8

10

20

30

40

50

60

70

80

90

Spatial Frequency (mm-1)

0.6

Fig. 11. Modulation Transfer Function curves for implanted planar InSb and epiInSb, measured on FPAs with pitches of 15, 20 and 30 lm.

0.4 0.2 0 0

10

20

30

NETD (mK) Fig. 8. Comparison of the NETD distributions in Pelican FPAs, with implanted planar InSb diodes at 80 K and epi-InSb diodes at 95 K. The Black Body temperature was 45 °C and the measurement was performed with an aperture of F/4.1.

dark current has been decreased by going to the MBE based technology. Note that the same dark current is achieved at 80 K in planar InSb and 95 K in epi-InSb. Since the quantum efficiency (QE) is comparable (>80%) for both technologies, the reduction of the dark current in epi-InSb allows the FPA operating temperature to be increased to 95 K with no loss of performance. The equivalence of performance is demonstrated in the next two figures, which compare the NETD distributions

178

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

50

95K

50

100

100

150

150

200

200

250

250

300

300

350

350

400

400

450

450

500

110K

500 100

200

300

400

500

600

100

200

300

400

500

600

Fig. 12. Images registered with a 640  512 15 lm pitch Pelican FPA at temperatures of 95 K and 110 K with an aperture of F/4.1.

4

2.5

x 10

Number of pixels

150K 2 1.5 1 0.5 0

0

0.1

0.2

0.3

0.4

0.5

Dark Current (pA) Fig. 13. Dark current distribution in Pelican InAs0.91Sb0.09 n-type bariode FPA at 150 K.

80

100

70

50

98

NETD Operability

96

40 94

30 20

Operability (%)

60

NETD (mK)

and the stability of the residual non-uniformity (RNU) to temperature drift (so called V-curves), respectively, for Pelican planar and epi-InSb FPAs. In Fig. 8, the NETD distributions are shown for planar InSb at 80 K, and epi-InSb at 95 K, measured in front of a 45 °C black body with an F/4.1 aperture and a 60% well fill. They are virtually identical. In Fig. 9, the V-curve for epi-InSb at 95 K is similar to that of planar InSb at 80 K. The V-curve is a plot of the RNU (at a well fill of 45%) as the FPA temperature is varied from its calibration temperature. The calibration temperature is the temperature at which the minimum of the V-curve is located. The calibration was performed using a standard two point non-uniformity correction [14], at well fill values of 30% and 60%, respectively. The V-curve is a measure of the stability of the FPA image quality with respect to temperature drift, where a wider curve corresponds to a greater stability. Epi-InSb at 95 K has a width of 7 K at RNU = 0.1% and is actually more stable than planar InSb at 80 K, whose width is 4 K. At 80 K the V-curve for epi-InSb is much wider than that for planar InSb, indicating the robustness of its RNU to temperature drift. In Fig. 10 we show the temperature dependence of the NETD at F/4.1, and the pixel operability, for an epi-InSb Pelican FPA. It may be seen that the NETD hardly changes over a range of temperature up to about 100 K. Similarly, the number of defects, determined by SCD’s standard production line criteria, essentially does not change in this temperature range. The pixel operability remains above 99.8% up to 100 K. This shows that the epi-InSb Pelican detector has a useful operating range up to, and even beyond, 100 K. An important issue in FPA image quality is pixel cross talk. This is the fraction of the light signal falling on a given pixel that is detected by one of its neighbours. It is most conveniently quantified by the Modulation Transfer Function (MTF). The MTF is the amplitude of a spatially periodic signal detected by the FPA as a function of the signal’s spatial frequency. Details of the method of its measurement are given in Ref. [16]. In Fig. 11, MTF curves are compared for both planar and epi-InSb FPAs with three different pitches: 30, 20 and 15 lm, corresponding to SCDs Blue Fairy, Sebastian and Pelican ROICs, respectively. The epi-InSb MTF curves are wider than the corresponding curves for planar InSb for the two largest pitches, and are essentially identical for the smallest pitch. Together Figs. 8–11 demonstrate that the higher operating temperature in epi-InSb has been achieved without any degradation of the image quality, even at the smallest pitch of 15 lm. The final proof is in the image itself, examples of which are given in Fig. 12 for a Pelican epi-InSb FPA with an aperture of F/4.1, at temperatures of 95 K and 110 K, respectively. Even at 110 K, the image quality is very good, and only begins to degrade noticeably above 120 K.

92 10 0 120

90 140

160

180

200

Temperature (K) Fig. 14. Temperature dependence of the NETD and the pixel operability for an InAsSb n-type Pelican-D FPA, measured with an aperture of F/3.2 in front of a Black Body at a temperature of 50 °C.

6. Performance of XBnn InAsSb FPA In this section we discuss n-type bariode structures which operate at even higher temperatures than the epi-InSb diodes discussed in the previous section. Except where stated otherwise, the bariode structures were grown latticed matched to 200 or 300 GaSb substrates, in a Veeco Gen III MBE machine. The principal layers in the bariode structures were a thick (1.5–3 lm) n-type

179

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

50

103K

50

116K

50

127K

50

100

100

100

100

150

150

150

150

200

200

200

200

250

250

250

250

300

300

300

300

350

350

350

350

400

400

400

450

450

150K

500 100

50

200

300

500

400

500

600

150K

100

50

400

450

161K

161K

200

300

400

500

500 600 50

450

172K 100

172K

200

300

400

500

500 600 50

100

100

100

100

150

150

150

150

200

200

200

200

250

250

250

250

300

300

300

300

350

350

350

350

400

400

400

400

450

450

450

193K

500 100

200

300

204K

500 400

500

193K

50

100

600

200

300

500

600

100

200

300

400

500

500 600

50

50

100

100

100

150

150

150

150

200

200

200

200

250

250

250

250

300

300

300

300

350

350

350

350

400

400

400

400

450

450

450

450

500

500

100

200

300

400

500

600

204K

214K

500

100

200

300

400

500

600

182K 100

200

300

100

200

300

182K

400

500

600

400

500

600

450

214K

500

400

139K

225K

50

225K

100

500

100

200

300

400

500

600

100

200

300

400

500

600

Fig. 15. Images at FPA temperatures between 103 and 225 K, for an n-type bariode FPA bonded to SCD’s Digital Pelican D signal processor. The F/number was F/3.2.

n-type bariode FPA. The operability was determined with similar criteria to those used for the epi-InSb FPA shown in Fig. 10. The NETD and operability only begin to change above 170 K, consistent with the BLIP temperature of 175 K estimated above. Examples of some images registered with a similar Pelican D bariode FPA to that shown in Fig. 14 are presented in Fig. 15, for the FPA operating at temperatures between 103 K and 225 K, and an aperture of F/3.2. The image does not show any obvious signs of degradation until temperatures above 180–190 K. These temperatures are consistent with the estimate made above for a BLIP temperature of 175 K. In this detector, we have demonstrated a NETD of 17.5 mK, as shown in Fig. 16. Using a standard two point non-uniformity correction and standard criteria for the identification of defective pixels, similar to those used in SCD’s matured InSb detectors, the pixel operability at 150 K was 99.85%. The mean spectral response of the bariodes of an nBnn FPA bonded to SCD’s Blue Fairy ROIC (30 lm pitch) was measured at 150 K. The spectrum is shown as a dotted line in Fig. 17a, where it is compared with the result of an optical transfer matrix simulation (solid line) [18]. The bariode had a 2.6 lm thick AL and an ARC with an optical thickness of 1.14 lm. The metallic contact on the top of the mesa had a low optical reflectivity, so the device was

2.5

Counts (x104)

InAs 0.91 Sb 0.09 photon absorbing active layer (AL), a thin (0.2–0.35 lm) n-type AlSb0.91As0.09 barrier layer (BrL), and a thin (0.2–0.5 lm) n-type InAs0.91Sb0.09 contact layer (CL). The grown wafers were processed into 640  512 pixel arrays, with a 15 lm pitch. In all cases, the mesas were etched to a depth greater than the thickness of the CL, and a common contact was made to the AL outside the active device area. The arrays were flip-chip bonded with indium bumps to a Pelican or Pelican-D ROIC. (Pelican is the analogue version, and Pelican-D, the digital version, of SCD’s 15 lm pitch ROIC). The substrate was usually etched away, and an antireflection coating (ARC) applied. The mean dark current in our FPAs shows identical behaviour to that reported previously for single devices [15,17]. It can be fitted to the standard dependence for Diffusion limited behaviour of the form J a T3 eDE/kT, yielding an activation energy of DE  336 meV. In Refs. [15,18,19], it was shown that a correction of 36 meV must be subtracted in order to take into account the dependence of the bandgap on temperature and thereby deduce the bandgap energy at 150 K. This procedure yields DE (150 K)  300 meV, which corresponds very well to the expected bandgap energy of 302 meV (bandgap wavelength 4.1 lm) for lattice matched InAs1xSbx [19]. It was also shown in Refs. 15 and 19 that the diffusion current can be used to deduce a value of 700 ns for the minority carrier lifetime, and a value of 50 lm for the bulk diffusion length. At 150 K, the typical operating dark current density is 2– 3  107 A/cm2, corresponding to a pixel dark current of just a few tenths of a pico-ampere for an FPA with a 15 lm pitch. This value is demonstrated in the distribution shown in Fig. 13 for a InAs0.91Sb0.09 Pelican n-type bariode FPA at 150 K. It has a mean value of 0.2 pA and a FWHM of 15%. This mean value is about 50 times lower than the photocurrent at F/3 for a QE of 70% between 3 and 4 lm. Such a QE value is quite realistic, and will be demonstrated below. It suggests that we can achieve very good BLIP operation at 150 K, and that the detector should remain in BLIP operation at F/3 up to 175 K, when the photocurrent and dark current become equal. Fig. 14 shows the temperature dependence of the NETD and pixel operability at F/3.2, for a 15 lm pitch Pelican-D InAs0.91Sb0.09

2 1.5 1 0.5 0 12 14 16 18 20 22

NETD (mK) Fig. 16. NETD distribution for the FPA in Fig. 15 at a black body temperature of 50 °C (Peak = 17.5 mK, standard deviation = 1.3 mK).

180

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

100

External QE (%)

External QE (%)

80 60 40 20 0 1.5

2

2.5

3

3.5

4

4.5

80 60 40

AL=5 micron AL=4 micron AL=3 micron

20 0 3.0

3.5

4.0

λ (μm)

λ (μm)

(a)

(b)

4.5

Fig. 17. Optical transfer matrix simulation of QE as a function of wavelength (spectral response) at 150 K (a) for a one pass nBnn bariode detector with a 2.6 lm AL and an AR coating of optical thickness 1.14 lm (thick solid line), compared with the measured spectrum (thick dotted line) (b) for a fully optimized two pass structure with a 3, 4 or 5 lm AL. The internal QE values used in the simulations were 91% and 95% respectively.

essentially a one pass device. The agreement between measured and simulated spectra is very good when an internal quantum efficiency of 91% is used. By increasing the reflectivity of the top contact in the simulation to 95%, increasing the AL thickness to 5 lm and by adjusting the thickness of the ARC for optimum performance, the upper spectrum shown in Fig. 17b is obtained. The internal quantum efficiency was also increased to 95% since the lower value fitted above is smaller than typical. Such a detector has a cut-off wavelength of 4.15 lm at 50% of maximum response, and a QE of >70% between 3 and 4 lm. Fig. 17b also shows how the QE and cut-off wavelength of an optimized detector increase with AL thickness. In addition to the lattice matched InAs0.91Sb0.09 n-type bariode FPAs grown on GaSb substrates, discussed so far, we have also begun investigations of non-lattice matched bariodes grown on GaAs substrates [20]. High quality GaAs substrates can be supplied commercially at lower cost and with diameters greater than the maximum 400 currently available for GaSb. We have found that by first growing a suitable thick buffer layer we can achieve comparable performance to the lattice matched FPAs, with negligible loss of pixel operability. The solid line in Fig. 18 is a ‘‘Rule 07 Plot’’ due to Tennant showing the expected dark current for high performance MCT diode detectors [1], with the two dashed lines indicating an acceptable range of uncertainty (corresponding to a factor of 2.5 increase or decrease from the ideal Tennant value). Onto this

1.0E-03

nB nn Pelican-D FPA (GaAs)

1.0E-04

nB nn Pelican-D FPA (GaSb)

2

Jdark [A/cm ]

1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1

1.5

2

1000/(λe*T) [1000/(µm-K)] Fig. 18. Dark current comparison between MCT Rule 07 and the mean value in Pelican-D InAs0.91Sb0.09 n-type bariode FPAs operating at 150 K, one of which was grown on GaSb and the other on GaAs. The AL thickness was 3.0 lm in each case.

plot have been added points to indicate the typical mean dark current of our Pelican-D bariode FPAs grown on each type of substrate. The FPAs were made from wafers containing nominally identical bariode structures with a 3 lm thick AL. After growth, the bandgap wavelengths of the ALs were measured optically and found to be very similar. The points showing the n-type bariode results fall slightly below the solid line of the Tennant formula, demonstrating that the performance of both the lattice matched and the non-lattice matched n-type bariode FPAs are entirely comparable with that of high quality diode FPAs made from MCT with the same AL bandgap.

7. Summary and conclusions At SCD we have developed both planar and epitaxial technologies for the fabrication of MWIR FPAs based on InSb. In all cases, these FPA detectors can be bonded to analogue or digital ROICs. Our base technology involves beryllium implanted planar InSb FPAs, in pitches down to 15 lm and in formats from 320  256 up to 1280  1024 (1.3 MegaPixel). These FPAs operate in the temperature range 80–85 K with a pixel operability of above 99.5%. Over the past 10 years, we have worked on two strategies for raising the operating temperature without compromising imaging performance. Both technologies are based on MBE grown InAs1xSbx, in one case with x = 1 and the other with x  0.09. These two compositions correspond to cut-off wavelengths of around 5.4 and 4.1 lm, respectively. We have performed a system analysis which shows that the shorter cut-off wavelength actually confers some advantage on systems designed for long range detection and identification, while for shorter ranges, the system performance is comparable in most situations. There are some applications, however, where the full MWIR window is required in any range. In epi-InSb diodes (x = 1), the G–R current is reduced by a factor of about 17, relative to our standard planar FPA technology. EpiInSb FPAs operate in the temperature range 95–100 K with the same cut-off wavelength and performance as for planar InSb at 80 K. In XBnn InAs0.91Sb0.09 bariodes, the G–R current is totally suppressed and the bariode FPAs operate at 150–160 K. At F/3, they exhibit BLIP performance up to 175 K. By raising the FPA operating temperature significant reductions can be made in the SWaP figure of the complete IDCA. In epi InSb operating at 95 K, the required cooling power is reduced by about 20% and this reduction increases to 60% for InAs0.91Sb0.09 bariode detectors operating at 150 K. The result is an improved range of solutions for a wide range of applications, including faster cooldown time and mission readiness, lower cooling power, higher

P. Klipstein et al. / Infrared Physics & Technology 59 (2013) 172–181

cooler reliability, and more compact detectors for volume-critical applications. The bariode detector arrays obey MCT Rule 07 very well and exhibit dark currents similar to the best MCT diodes with the same active layer bandgap. Moreover, they can be grown on relatively low cost 300 or 400 GaSb substrates or even larger GaAs substrates, making them a particularly attractive prospect for large scale, low SWaP, staring arrays. Both epi-InSb and XBnn InAs0.91Sb0.09 FPAs may also be operated below their standard operating temperatures, in order to reduce the dark current to a negligible value. This is useful in applications where the incident photon flux is particularly low, or where the FPA temperature is liable to drift during operation. Epi-InSb 15 lm pitch Pelican or Pelican D FPAs for 95 K operation covering the full MWIR spectrum (kC = 5.4 lm) are now in production at SCD. 15 lm pitch XBnn InAs0.91Sb0.09 bariode FPAs for 150 K operation in the ‘‘blue’’ part of the MWIR spectrum (kC = 4.1 lm) will be in production by 2013. Acknowledgements We would like to acknowledge support for this work received from the Chief Scientist of the Israeli Ministry of Trade and Industry (MOITAL). The authors also acknowledge technical support from Mr. S. Greenberg, who was responsible for the smooth operation of the two MBE machines, and Ms. B. Yariv, Ms. H. Moshe, Ms. H Schanzer, Mr. I. Bogoslavsky, Mr. Y. Caracenti, Mr. D. Gur, Ms. L. Krivolapov, Mr. Y. Osmo and Mr. S. Weinstein who have all contributed to the successful processing, packaging or characterization of the devices. References [1] W.E. Tennant, Rule 07 Revisited. . .Still a Good Heuristic Predictor for HgCdTe Performance? 2009 US Workshop on the Physics and Chemistry of II–VI materials, Chicago, 2009. [2] A. Glozman, E. Harush, Eli Jacobsohn, O. Klin, P.C. Klipstein, T. Markovitz, V. Nahum, E. Saguy, J. Oiknine-Schlesinger, I. Shtrichman, M. Yassen, B. Yofis, E. Weiss, High performance InAlSb MWIR detectors operating at 100 K and beyond, Proc. SPIE 6206 (2006) (6206-0M). [3] W. Shockley, W.T. Read, Statistics of the recombinations of holes and electrons, Phys. Rev. 87 (1952) 835–842; R.N. Hall, Electron hole recombination in germanium, Phys. Rev. 87 (1952) 387.

181

[4] Donald A. Reago, Future Army Applications for IR Focal Plane Arrays, Keynote Address, Infrared Technology and Applications XXXIV Conference, Orlando, 2008. . [5] S. Elkind, A. Adin, I. Nevo, A.B. Marhasev, Focal plane processor with a digital video output for InSb detectors, SPIE 4820 (2002) 751. [6] O. Nesher, S. Elkind, A. Adin, I. Nevo, A.B. Yaakov, S. Raichshtain, A.B. Marhasev, A. Magner, M. Katz, T. Markovitz, D. Chen, M. Kenan, A. Ganany, J. Oiknine Schlesinger, Z. Calahorra, A digital cooled InSb detector for IR detection, SPIE 5074 (2003) 120. [7] P.C. Klipstein, Depletionless Photodiode with Suppressed Dark Current, US Patent 7,795,640, 2 July 2003. [8] P.C. Klipstein, Unipolar Semiconductor Photodetector with Suppressed Dark Current, US Patent 8,004,012, 6 April 2006. [9] P.C. Klipstein, XBn barrier photodetectors for high sensitivity and high operating temperature infrared sensors, Proc. SPIE 6940 (2008) (6940-2U). [10] O. Nesher, I. Pivnik, E. Ilan, Z. Calahorra, A. Koifman, I. Vaserman, J. Oiknie Schlesinger, R. Gazit, I. Hirsh, High resolution 12801024, 15 lm pitch compact InSb IR detector with on-chip ADC, SPIE 7298 (2009) (7298-3K). [11] T. Markovitz, I. Pivnik, Z. Calahorra, E. Ilan, I. Hirsch, E. Zeierman, M. Eylon, E. Kahanov, I. Kogan, N. Fishler, Digital 640  512/15 lm InSb detector for high frame rate, high sensitivity, low power applications, SPIE 8012 (2011) (80122Y). [12] L. Langof, D. Nussinson, E. Ilan, S. Elkind, R. Dobromislin, I. Nevo, F. Khinich, M. Labilov, Z. Calahorra, S. Vaserman, T. Markovitz, O. Manelaa, D. Elooza, A. Tuito, D. Oster, Advanced multi-function infrared detector with on-chip processing, SPIE 6542 (2007) 654231. [13] L. Shkedy, T. Markovitz, Z. Calahorra, I. Hirsh, I. Shtrichman, Megapixel digital InSb detector for midwave infrared imaging, Opt. Eng. 50 (2011) 061008. [14] L. Shkedy, O. Amir, Z. Calahorra, J. Oiknine-Schlesinger, I. Szafranek, Temperature dependence of spatial noise in InSb focal plane arrays, SPIE 4028 (2000) 481. [15] P.C. Klipstein, O. Klin, S. Grossman, N. Snapi, I. Lukomsky, M. Yassen, D. Aronov, E. Berkowicz, A. Glozman, T. Fishman, O. Magen, I. Shtrichman, E. Weiss, XBn barrier photodetectors based on InAsSb with high operating temperatures, J. Opt. Eng. 50 (2011) 061002. [16] I. Shtrichman, T. Fishman, U. Mizrahi, V. Nahum, Z. Calahorra, Y. Aron, Spatial resolution of SCD’s InSb 2D detector arrays, Proc. SPIE 6542 (2007). 6542-3M. [17] P.C. Klipstein, O. Klin, S. Grossman, N. Snapi, I. Lukomsky, M. Yassen, D. Aronov, E. Berkowicz, A. Glozman, O. Magen, I. Shtrichman, R. Frenkel, E. Weiss, High Operating temperature XBn-InAsSb bariode detectors, Proc. SPIE 8268 (2012). 8268-0U. [18] P.C. Klipstein, O. Klin, S. Grossman, N. Snapi, B. Yaakobovitz, M. Brumer, I. Lukomsky, D. Aronov, M. Yassen, B. Yofis, A. Glozman, T. Fishman, E. Berkowicz, O. Magen, I. Shtrichman, E. Weiss, XBn barrier detectors for high operating temperatures, Proc. SPIE 7608 (2010). 7608-1V. [19] P.C. Klipstein, O. Klin, S. Grossman, N. Snapi, B. Yaakobovitz, M. Brumer, I. Lukomsky, D. Aronov, M. Yassen, B. Yofis, A. Glozman, T. Fishman, E. Berkowicz, O. Magen, I. Shtrichman, E. Weiss, MWIR InAsSb XBn detectors for high operating temperatures, SPIE 7660 (2010). 7660-2Y. [20] E. Weiss, O. Klin, S. Grossmann, N. Snapi, I. Lukomsky, D. Aronov, M. Yassen, E. Berkowicz, A. Glozman, P.C. Klipstein, A. Fraenkel, I. Shtrichman, InAsSb-based XBnn bariodes grown by molecular beam epitaxy on GaAs, J. Cryst. Growth 339 (2012) 31.