High performance MWIR type-ii superlattice detectors

Transcription

1 High erformance MWIR tye-ii suerlattice detectors Henk Martijn, Carl Aslund, Rickard Marcks von Würtemberg, Hedda Malm IRnova AB Electrum 236, SE Kista, Sweden ABSTRACT A VGA format tye-ii suerlattice focal lane array (FPA) for the mid-wave infrared (MWIR) atmosheric window has been designed, manufactured and characterized. The detector material is based on a heterojunction structure with a barrier that effectively decreases the Shockley-Read-Hall based comonent of the dark current. A very effective assivation method has been used which successfully inhibits all surface leakage currents. The barrier structure has a 50 % cutoff at 5 µm and 65 % quantum efficiency without antireflective coating. The dark current density is A/cm 2 at V bias and 120 K. The otical cavity of the detector has been otimized for maximum cature of available light in the MWIR window. A focal lane array with 640 by 512 ixels and 15 µm itch was rocessed based on this barrier structure. High-quality imagery in a system with high F-number will be resented. Keywords: tye II suerlattice, heterostructure, MWIR, hotodetector, InAs/GaSb, M-structure 1. INTRODUCTION In the 1980s, Smith and Maihiot [1] roosed tye-ii InAs/GaSb based strained layer suerlattices (SLS) for infrared (IR) detection. SLS or also called T2SL (Tye II SuerLattice), is a material / technology that can be used to make high quality cooled infrared hoton detectors with a cut-off wavelength ranging from 2 μm to 30 μm. This covers the SWIR, MWIR, LWIR and VLWIR wavelength bands, loosely defined as 2-3 μm, 3-5 μm, 8-12 μm and >12 μm, resectively. Commercially available IR detectors in the MWIR region are mercury cadmium telluride (MCT) and bulk InSb hoton detectors. However Tye-II suerlattices have interesting roerties comared to current technologies. T2SL benefits from a larger carrier effective mass, comared to the state-of-the-art MCT technology, leading to reduced tunneling currents [1] [2] and significant reduction of Auger recombination rate [3]. This in its turn leads to higher oerating temeratures. These advantages however, have not yet exerimentally been shown in T2SL, mainly because other dark current mechanisms related to defects in the material still dominate. Therefore one of the challenges for T2SL to reach its full otential is to either eliminate the defects or work around them. Fortunately the InAs/GaSb/AlSb material system gives a very large design freedom that can be utilized to realize more comlex device structures with extra features intended to reduce the undesirable dark current mechanisms. In simle homojunction devices a lightly doed intrinsic region (i) is sandwiched between heavier doed - and n-tye contact layers. The i absorber region is referably doed as holes have lower mobility than electrons and it is therefore beneficial to have the latter as minority carriers. Under light reverse bias the sace charge region arises almost comletely in the lower doed i layer. Exerimental measurements show that the dark current is u to high temeratures dominated by g-r current in this tye of structures [4] [5]. The source of g-r current is Shockley-Read-Hall (SRH) centers in the deletion region; tyically caused by defects, imurities or interface states. In the general case the net generation rate can be described by the following classical exression [6] [7]: 2 G R = n n) /(( n + n ) τ + ( + ) τ ), (1) ( i n0 where G is the generation rate, R is the recombination rate, and n 1 and 1 are the electron and hole densities in the material when the Fermi level is equal to the energy level of the SRH centers in the bandga. τ 0 and τ n0 are the electron and hole lifetime resectively. For a homojunction device the SRH generation currents are directly roortional to the width of the sace charge region, W. The relation to the doing levels is then as follows: J SRH = ( G R) qw, where W ( N A + N D ) / N AN D (2) Infrared Technology and Alications XXXIX, edited by Bjørn F. Andresen, Gabor F. Fulo, Charles M. Hanson, Paul R. Norton, Proc. of SPIE Vol. 8704, 87040Z 2013 SPIE CCC code: X/13/$18 doi: / Proc. of SPIE Vol Z-1

2 I In the case of much lower absorber doing N A, as comared to N D, this can be simlified further to: J SRH 1/ N A (3) In other words, in a homojunction device the SRH current varies slowly and redictably with absorber doing N A. As is clear from Eq. (1), for midga defect levels in the deletion region the g-r g current is roortional to t the intrinsic carrier concentration n i. The intrinsic carrier concentration is a strong function of o the bandga energy, so in this work we use a heterojunction design similar to [8] where the most of the deletion regionn resides in the higher bandga material. Due to the fact that only a minor art off the deletionn region extends into the absorber, the g-r current is much reduced, but the deendence on the absorber doing level is much stronger than in Eq. (3), as we shall see in Section 3.2. In this work the imrovements and otimizations thatt can be made on a heterojunction designn for the MWIR wavelength band are studied. We will look in moree detail at otical resonse, dark current imrovements and assivation methods. The target is a focal lane array with 640 by 512 ixels on 15µm itch oeratingg at 120 K; to be used in systems with F/4 aerture. 2. EXPERIMENTAL The baseline detector material used in this work, Design A, is grown on n-tye (Te-doed) GaSb (100) substratess using solid source molecular eitaxy (MBE). The material consists of a 0.3 µmm lattice matched InAs 0.91 Sb 0.09 etch-sto layer, followed by 0.7 µm heavily Si-doed n-tye M-suerlattice (M-SL) bottom contact, 0.5 µm lightly doed n-tye M-SL, 3 µm weakly doed -tye SL absorber, a 0.1 µm of an intermediate bandga heavilyy Be-doed -tye SL, and finally a 0.1 µm thick heavily Be-doedd -tye GaSbb contact layer. The M-SL eriods consist of InAs/GaSb/AlSb/GaSbb in the roortions 10/1/4/11 ML, whereas the absorberr SL consists of 9.5 ML InAs / 11.5 MLL GaSb. The to and bottom contacts are doed to and cm -3, resectively, whereas the doing concentration in the absorber is cm -3 (-tye); see Figure 1. nas/gasb: E T==L j - Mirror and :ontact ;I :r_. t I neta l Figure 1 Overview of the detector structure under zero bias condition Proc. of SPIE Vol Z-2

3 Design B and Design C are variants of the baseline design. They vary in absorber thickness, cut-off wavelength and absorber doing level, see Table 1. Table 1 Summary of the different designs used in this work with varying absorber thickness, cut-off wavelength and absorber doing level. Absorber thickness Absorber Suer Lattice Absorber doing (cm -2 ) Design A 3 µm 9.5 ML InAs / 11.5 ML GaSb Design B 4 µm 10 ML InAs / 11.5 ML GaSb Design C 4 µm 10 ML InAs / 11.5 ML GaSb Standard III/V-rocessing was erformed on Design A ei material to roduce a detector array with ixels and a ixel itch of 15 µm; mesas were formed using a combination of dry and wet etching [9], after which they were assivated using olymer based assivation. Mirror and contact metal was evaorated and after dicing the detector material was hybridized to an ISC0403 readout. Finally, the substrate was fully removed. A fill factor of 81% was achieved using this rocess. The same rocess was also used to fabricate devices with a number of single ixels of different sizes, both with olymer and dielectric assivation. These devices were used to evaluate dark current behavior and otical resonse. All results reorted in this aer were obtained for devices without antireflective (AR) coating. The quantum efficiency (QE) sectrum was measured using a disersive sectrometer with a calibrated hoton flux entering through the substrate, or in the case of removed substrate through the bottom contact material. A lock-in amlifier was used to suress noise and background signal. The generation-recombination rates were modeled by solving the standard semiconductor drift-diffusion equations in 1D. Convergence of self-consistent solutions for low temeratures was achieved by increasing the numerical recision beyond the 64 bits reresentation of a standard double recision floating oint number. An 8 8 k.p transfer-matrix enveloe function aroximation simulator was used to model the suerlattice band structure [10]. Material data, including band offsets were taken from ref. [11] Proc. of SPIE Vol Z-3

4 3. RESULTS 3.1 Otical otimization Reference design A has a cut-off at 4.8 µm. This is slightly too short to take full advantage of f all hotons available a in the MWIR band; esecially considering that the hotonn flux or Planck radiationn of a room temerature body increasess raidly for longer wavelengths. To increase the cut-off wavelength, the InAss suerlattice layer in the absorber was therefore increased by 0.5 ML to 10 ML. To also imrove the quantum efficiency the absorber thickness was increased to 4 µm. Total external qua ntuciefficieicy.caf bias Photo current frorn 30 C radiation, fl4l aerture, OV bias Lcl 0.5 Design A - Design B nt density (Alcm2 /m) N N G) [n in [' 1E {E 05 1E-05 ie-05 - DE DE.sign A.sign B Sectral hoto curre P o in in ie-05 L TJavrelerçti(. n! Wavelengt Figure 2 External quantum efficiency and sectral hotocurrent density of Design A and Design B. No antireflective coating was alied. The increase of the absorber thickness resulted in an increase of QE from 40 % to t 60 %; see Figure 2. The increase i of the cut-off wavelength to 5.1 µm corresonds very well with the exected shift as redicted by k.p simulations, validating the usefulness of the used k.p model. If I we look at the total hotocurrent integrated over the 3 to 5 µm interval for a tyical alication (30 C background F/ /4) we see a two-fold increase due to these imrovements in the detector design. 3.2 Dark Current In a correctly designed heterojunction device, with a lower doing level in the barrier than in the absorber, most of the sace charge region lies in the higher bandga material. In the high bandga barrier, the generation rate is i much lower than in the absorber region, hence only the tail of thee deletion region extendingg into the absorber region will contribute to the SRH current. Because the assumtions leading to eq. (3) do not hold, we have h to rely on simulationss to redict the generation rate. The extension of the deletion region into the absorber will decrease raidly with increasing absorber doing N A. The simulation results in Figure 3 to Figure 5 show this very clearly: a 3 fold increase of the absorber doing leads to a very raid decrease of the total net generation rate, by far more than the factor of 3 that would be exected from corresonding homojunction devices. The simulations show also clearly that t the generation rate in the barrier (to the left of the 1.2 µm mark) can be neglected. Proc. of SPIE Vol Z-4

5 0.5 A B C Band edges (ev) Figure 3 Band alignment of Design B under reverse bias, Region A is the N + contact layer, B The n - - barrier and C the - absorber. Carrier concentration (cm -3 ) n net recomb. rate net gen. rate Carrier concentration (cm -3 ) n net recomb. rate net gen. rate Figure 4 Design B, simulation of free carrier concentration in units of cm -3, and net generation rate of electron-hole airs in arbitrary units. The figure on the right is an enlargement of the interfacee between the barrier and the t absorber. Carrier concentration (cm -3 ) n net recomb. rate net gen. rate Carrier concentration (cm -3 ) n net recomb. rate net gen. rate Figure 5 The same kind of simulation ass in Figure 4, but for Design C. Please note that the y-axis scale is logarithmic, l so the volume with any significant SRH current c generation has decreased dramatically from that of Design B. Proc. of SPIE Vol Z-5

6 The simulations have been exerimentally verified: two samles with single ixels of different sizes have been reared. Design B based on material with nominal -tye absorber doing and Design C with 3 times higher doing. Current density (A/cm 2 ) 0Dark current density of E0027, conn. 3, area: um Design B SRH Current density (A/cm 2 ) 0Dark current density of E0028, conn. 3, area: um Design C Band-to- Band T = 170 K T = 160 K T = 150 K T = 140 K T = 130 K T = 120 K T = 110 K T = 100 K T = 90 K T = 80 K T = 70 K T = 60 K T = 50 K T = 40 K Bias (V) Figure 6 I-V characteristics for Design B (left) and Design C (right). At low bias and high temerature the SRH currents are dominating in Design B, see Figure 6. At high bias however, the band-to-band tunneling is larger in Design C than in Design B. This can be exlained by the higher field strength in the deletion area due to the higher -doing. The higher band to band tunneling is of no imortance because these devices are normally oerated in a low bias regime where this comonent of the dark current is negligible. The maximum RA roduct for these structures occur around -40 mv reverse bias and can be u to an order of magnitude larger than R 0 A. Table 2 RA and Dark current at 120 K oerating temerature and -40 mv reverse bias RA (Ωcm 2 ) Dark current Density (A/cm 2 ) Design B Design C Bias (V) By increasing the doing, the RA roduct has increased with almost a factor 7 while the dark current has imroved with a factor 10. The numbers for Design C suggest that it will be ossible to oerate this material with good erformance at 120 K. Proc. of SPIE Vol Z-6

7 3.3 Passivation To test the effect of different tyes of mesa etch and assivation rocesses, single ixel devices with different sizes and geometries were fabricated, starting with the ei material of Design A. As a reference we use a assivation rocess based on a olymer. The best results were obtained with a dielectric assivation Dark current density at -0.05V vs. P/A ratio of 8F Dark current density at -0.05V vs. P/A ratio of E V (A/cm 2 ) V (A/cm 2 ) T = 80 K T = 100 K T = 120 K T = 140 K T = 160 K T = 160 K T = 140 K T = 120 K T = 100 K T = 80 K Perimeter/Area (cm -1 ) Perimeter/Area (cm -1 ) Figure 7 Dark current density as function of erimeter over area ratio for mesas with different sizes (170 µm, 60 µm and 30 µm) and temeratures. Left grah shows results for the olymer assivated mesas, right grah shows the results for the dielectric assivated mesas. Figure 7 shows dark current density as function of erimeter over area. For a device exhibiting no surface leakage this relation should be a constant. As can be seen the olymer assivated samles showed a significant amount of edge currents while the dielectric assivated devices are comletely bulk current dominated. When these results are extraolated to even smaller 15 µm ixels, the imroved assivation will give two orders of magnitude decrease of dark current. That corresonds to an imrovement in oerating temerature of 30 K. The bulk dark current density of A/cm 2 at 120 K is significantly lower than for homojunction devices of similar cutoff wavelength. However, there is still room for imrovement. As shown by Chen et al. [12], with carefully otimized doing and band offsets the bulk dark current can reach the low 10-8 A/cm 2 range in similar heterojunction devices. Proc. of SPIE Vol Z-7

8 3.4 FPA results Material from Design A was rocessed with olymer as assivation and hybridised to an ISC0403 readout circuit. The substrate was comletely removed. A relatively high and not imressive NETD of 41 mk (50% well fill) was measured at 90 K with F/4 and 21 ms integrationn time. This iss most likely due to the incomlete assiviation of this device and hence high surface leakagee current and related r noise on the mesa edges. 7- Figure 8 Images taken at 90 K with a 640 by 512 ixel, 15 µm itch focal lane array a rocessed from Design A material. The images in Figure 8 taken at 90 K were two-oint corrected and dead ixels have been removed. Thee oerability is 99 % if ixels are marked as non-oerating when the resonse deviates more than 7.5 % fromm its nearest neighbors or its NETD > 100 mk. 4. CONCLUSIONS In conclusion, heterojunction detectors have been otimized with resect to dark current, otical resonse and surface assivation. High quality imaging of a 640 by 512 ixels FPA detector has been demonstrated. The erformance is limited because not all imrovements on dark current and otical resonse as described d have been imlemented in the demonstrator r FPA. An otimized FPA should s reach a NETD < 20 mk at an integration time < 10 ms. This FPA will be used in an IDCA (Integrated Dewar Cooler Assembly) with low weight, size andd ower. ACKNOWLEDGEMENTS The authors would like to acknowledge the suort from the Swedish National Sace Board and VINNOVA; Sweden s Innovation Agency who has made this work ossiblee through the Center of excellence IMAGIC. The authors wish also to thank their colleagues at IRnova for rocessing of the samles. REFERENCES [1] Smith, D. L. and Mailhiot, C., "Proosal for strained layer tye II suerlattice infraredd detectors," J. J Al. Pyhs., vol. 62, , (1987). [2] Mailhiot, C. and Smith, D. L., "Long-wavelength infrared detectors based on suerlattices," Vac. Sci. Technol.7 A(2), , (1989). InAs-Ga1-xInxSb tye-ii [3] Mohseni, H., Litvinov, V. I. and Razeghi, M., "Interface-induced suression of the Auger recombination in tye-ii InAs/GaSb suerlattices," Physicall Review B, vol. 58 (23), , (1998). [4] Malm, H., Marcks von Würtemberg, R., Aslund, C., Martijn, H., Karim, A., Gustafsson, O., Plis, E. and Krishna, S., "Recent develoments in tye-ii suerlattice detectors at IRnova AB," Proc. P SPIE, vol. 8353, T, (2012). Proc. of SPIE Vol Z-8

9 [5] Aslund, C., Marcks von Würtemberg, R., Lantz, D., Malm, H., Martijn, H., Plis, E., Gautam, N. and Krishna, S., "Performance of mid-wave T2SL detectors with heterojunction barriers," Infrared Physics and Technology, in ress, htt://dx.doi.org/ /j.infrared , (2012). [6] Shockley, W. and Read, W. T., "Statistics of the Recombinations of Holes and Electrons," Phys. Rev., vol. 87, , (1952). [7] Hall, R. N., "Electron-Hole Recombination in Germanium," Phys. Rev., vol. 87,. 387, (1952). [8] Abdollahi Pour, S., Huang, E. K., Chen, G., Haddadi, A., Nguyen, M. and Razeghi, M., "High oerating temerature midwave infrared hotodiodes and focal lane arrays based on tye-ii InAs/GaSb suerlattices," Al. Phys. Lett. 98, , (2011). [9] Chaghi, R., Cervera, C., Aït-Kaci, H., Grech, P., Rodriguez, J. B. and Christol, P., "Wet etching and chemical olishing of InAs / GaSb suerlattice hotodiodes," Semicond. Sci. Technol. 24, , (2009). [10] Szmulowicz, F., "Numerically stable secular equation for suerlattices via transfer-matrix formalism and alication to InAs/In0.23Ga0.77Sb and InAs/In0.3Ga0.7Sb/GaSb suerlattices," Phys. Rev. B, vol. 57,. 9081, (1998). [11] Vurgaftman, I., Meyer, J. R. and Ram-Mohan, L. R., "Band arameters for III V comound semiconductors and their alloys," J. Al. Phys., vol. 89,. 5815, (2001). [12] Chen, G., B. Nguyen, M., Hoang, A. M., Huang, E. K., Darvish, S. R. and Razeghi, M., "Elimination of surface leakage in gate controlled tye-ii InAs/GaSb mid-infrared hotodetectors," Al. Phys. Lett., vol. 99, , (2011). Proc. of SPIE Vol Z-9