Molecular beam epitaxy growth of high quantum efficiency InAs/GaSb superlattice detectors

Transcription

1 Molecular beam epitaxy growth of high quantum efficiency InAs/GaSb superlattice detectors G. J. Sullivan, a A. Ikhlassi, J. Bergman, R. E. DeWames, and J. R. Waldrop Rockwell Scientific Company, 1049 Camino Dos Rios, Thousand Oaks, California C. Grein Department of Physics, University of Illinois at Chicago, Chicago, Illinois M. Flatté OSTC and Department of Physics and Astronomy, University of Iowa, Iowa City, Iowa K. Mahalingam Air Force Research Laboratory, Wright-Patterson AFB, Ohio H. Yang, M. Zhong, and M. Weimer Department of Physics, Texas A&M University, College Station, Texas Received 25 March 2005; accepted 18 April 2005 InAs/GaSb superlattices are leading candidates for next generation long-wave infrared and very-long-wave infrared photodetectors. These heterostructures are expected to hold important advantages over existing materials systems, primarily bulk HgCdTe alloys. To realize their inherent potential, however, superlattice materials with low defect density and improved device characteristics must be demonstrated. Here, we report on the molecular beam epitaxy growth and characterization of an 11 µm cutoff wavelength InAs/GaSb superlattice detector with a state-of-the-art single-pass, internal quantum efficiency of 36%. The shutter sequencing used to form the GaSb-on-InAs and InAs-on-GaSb superlattice heterojunctions is described in detail, and the latter specifically identified as a source of morphological defects in these devices American Vacuum Society. DOI: / I. INTRODUCTION InAs/GaSb strained layer superlattices SLSs are the leading candidate materials for next generation very-longwave infrared VLWIR photodetectors. 1 3 HgCdTe alloys, which are the established material system for long-wave infrared applications, are widely anticipated to encounter fundamental difficulties in the VLWIR. Alloy fluctuations, for example, are expected to introduce a proportionally larger perturbation in the detector cutoff wavelength with decreasing band gap, making precise control of optical properties at these longer wavelengths in principle more difficult. Bandto-band tunneling is likewise expected to present a source of excess current for sufficiently narrow band gaps, although the available data show that HgCdTe diodes with cutoff wavelengths of up to twenty microns are not yet limited by such tunneling. 4 The ability to directly manipulate the band gap of a molecular beam epitaxy MBE -grown InAs/GaSb superlattice through precise tailoring of the well-to-barrier thickness ratio offers hope of circumventing the problems anticipated with bulk HgCdTe alloy fluctuations in the VLWIR regime. Furthermore, the much larger carrier effective masses associated with these structures should substantially suppress the bandto-band tunneling expected in HgCdTe. Finally, there is both theoretical and experimental evidence that, through suitable II. MATERIALS GROWTH, DEVICE FABRICATION, AND SLS LAYER DESIGN To grow the optically thick, low dislocation density layers needed for high quantum efficiency devices, the average lata Electronic mail: gsullivan@rwsc.com adjustment of the elastic strain within the individual layers of a SLS, Auger recombination can also be effectively reduced. 5,6 Silicon blocked-impurity-band detectors, the-incumbent VLWIR technology, have proven reliable near their inherent cutoff wavelength 20 µm for Si:Ga, 7 but are not optimal for shorter wavelengths because this cutoff cannot be conveniently adjusted. InAs/Ga In Sb SLS detectors, on the other hand, may be tuned to operate anywhere within the shortwave 3.7 µm 8 to very-long-wave 31 µm 9,10 infrared regimes through appropriate control of their respective layer thicknesses and alloy compositions. High quantum efficiencies and low excess currents must nevertheless be simultaneously achieved in order to meet the performance requirements for a useful technology. Though several quantum efficiency measurements have recently appeared, 11,12 reported values do not yet approach unity. Here, we describe 11 µm cutoff wavelength superlattice detectors with a stateof-the-art, single-pass internal quantum efficiency of 36%. The epitaxial growth and subsequent optical, as well as structural, characterization of these detectors is detailed, and a class of MBE shutter-sequence-related morphological defects that might ultimately bear on efforts to likewise achieve low excess currents identified. 1 J. Vac. Sci. Technol. B 23 3, May/Jun X/2005/23 3 /1/0/$ American Vacuum Society 1

2 2 Sullivan et al.: MBE growth of high quantum efficiency InAs/GaSb 2 TABLE I. SLS layer sequence, along with nominal thicknesses and intended doping. Indium-to-gallium ratios, deduced from the x-ray spectrum in Fig. 1, are for the as-grown structure. Thickness m Description Doping cm 3 buffer 1.00 GaSb:Be 1 e 18 p-type SLS ML In/7.35 ML Ga 4 e 17 undoped SLS ML In/ 7.35 ML Ga n-type SLS ML In/6.90 ML Ga 6 e 17 cap 0.05 InAs:Si 1 e 18 tice constant of a SLS must match that of its substrate as closely as possible. The ideally binary GaSb layers in an InAs/GaSb SLS precisely lattice-matched to its GaSb substrate will be strain-free, whereas the corresponding InAs layers will be in tension. This tension may be balanced either with compressively strained InSb-like bonds at the respective superlattice heterojunctions, or by incorporation of bulkcoordinated InSb into the GaSb and/or InAs layers. As shown in Sec. IV, later, indium is unavoidably introduced into the GaSb layers by cation segregation across the GaSbon-InAs interface, and antimony likewise introduced into the InAs layers by anion segregation together with background incorporation across the InAs-on-GaSb interface. Achieving the sought after lattice match in spite of this uncontrolled alloying requires the growth of one or more test layers, whose respective residual strains are determined with x-ray diffraction, and then appropriately compensating for that strain in subsequent growths. Once such a lattice-matched recipe is established, it can be used to reproducibly fabricate high-structural-quality superlattices of the desired thickness. The layers reported on here were grown using elemental gallium and indium sources, together with valved crackers to produce arsenic and antimony dimers. The V/III flux ratios for InAs and GaSb were each about 1.5, with growth rates subsequently determined from high-resolution x-ray diffraction of 0.83±0.01 ML/s for GaSb and 0.218±0.001 ML/s for InAs. Silicon and beryllium were employed as n- and p-type dopants, with silicon exposure confined to the InAs layers because it acts as an acceptor in GaSb. The superlattices were grown on 001 -oriented 0.1 GaSb substrates, at a temperature of 396 C as measured with a pyrometer calibrated against the GaSb surface reconstruction transition. 13 This growth temperature was selected on the basis of a reported minimum in background carrier concentration 14 under these conditions, but such a minimum was not independently verified. Individual devices were fabricated by creating mesas with a citric acid and hydrogen peroxide-based etch that terminated in the GaSb buffer, and subsequently evaporating Ti/Au contacts to the top InAs and bottom GaSb layers. No intentional passivation of the mesa sidewalls was attempted. The layer sequence whose salient structural and optical characteristics are presented later Sec. III is described in Table I. The target superlattice was a binary binary structure with InSb-like bonding across both heterojunctions. The FIG x-ray spectrum Cu K 1 for the SLS in Table I. Diffraction angle is referenced to the substrate. n-type region was grown with a period slightly shorter 8.33 s of gallium and 66 s of indium than that of the corresponding p-type and undoped regions 8.82 s of gallium and 66 s of indium, each. This small period difference introduces the doubling of superlattice satellite peaks evident from the x-ray spectrum presented in Fig. 1. III. STRUCTURAL AND OPTICAL CHARACTERISTICS A plot of 2 sin / x-ray versus satellite peak order, where is the absolute diffraction angle associated with a given satellite peak, yields a straight line whose slope is inversely proportional to the corresponding superlattice periodicity. Comparison of the results so obtained for the combined p-type and undoped regions ±0.006 Å with those similarly derived for the n-type region 64.93±0.01 Å permits accurate determination of the respective numbers of gallium and indium monolayers deposited per period Table I. This presumes, of course, that the corresponding source fluxes have remained fixed throughout growth, but selfconsistently accounts for the small period perturbation introduced by segregation and background incorporation. The respective intercepts of these same plots yield the corresponding mismatch relative to the substrate. The residual strains thus found for the p-type plus undoped and n-type epitaxial regions were 0.185% and 0.180% ± 0.001%, with both sets of layers in compression. The presence of numerous, high-order satellite peaks in the x-ray spectrum of Fig. 1 suggests good structural quality. A 002 dark-field transmission electron micrograph TEM from a nominally identical growth, shown in Fig. 2, directly confirms the regular periodicity and overall structural quality inferred from x-ray diffraction. Figure 3 shows the normalized 40 K spectral response of a typical diode fabricated from the superlattice growth described in Table I. The corresponding detector cutoff, defined by the extremum in differential absorbance versus wavelength near threshold, is 11.1 µm; the more commonly J. Vac. Sci. Technol. B, Vol. 23, No. 3, May/Jun 2005

3 3 Sullivan et al.: MBE growth of high quantum efficiency InAs/GaSb 3 FIG dark-field TEM image from a SLS nominally identical with the one in Table I. quoted wavelength at 50% of peak responsivity judged here as 8 µm yields a similar 10.9 µm value. The detector s quantum efficiency was determined by focusing a narrowbandpass filtered spot of known intensity and 9.0 µm wavelength onto the top of a diode s 400 µm diameter mesa, and then scanning this spot across the entire mesa to ensure all incident radiation was captured while simultaneously recording the diode s electrical signal. The external quantum efficiency measured this way was found to be 25%. Following correction for an expected 30% reflection from the top InAs surface which was without an antireflection coating, this translates into a single-pass, internal quantum efficiency of 36%, the current state of the art for SLS photodiodes. It is worth pointing out that in a backside illuminated configuration, where reflection from the metal contact at the top of the mesa permits a second pass of incident radiation through the detector, which is common in focal plane arrays, this value could be made nearly twice as large. IV. COMPOSITIONAL GRADING The surprising number of structural and compositional degrees of freedom presented by these superlattices makes accurate knowledge of the as-grown versus intended structure essential for detector optimization. Cross-sectional scanning tunneling microscopy STM is aptly suited to this challenge, because it permits the layer-by-layer composition of such structures to be ascertained with atomic-scale precision. 15 Figure 4 shows a typical anion sublattice image from a closely related, though somewhat longer period 15 ML In / 17 ML Ga structure, grown under conditions similar to but not entirely identical with those described earlier. The image illustrates two, thus far universally observed, features of this nominally binary binary material system: spatially graded antimony-for-arsenic replacement due to a combination of anion segregation and background incorporation across the InAs-on-GaSb interface, 16 together with spatially graded indium-for-gallium replacement due to cation segregation across the GaSb-on-InAs interface. 17,18 By way of comparison, there is virtually no arsenic-for-antimony replacement FIG. 3. Normalized 40 K spectral response of a typical 400 µm diameter photodiode patterned from the SLS in Table I. Features at 2.5, 4, and 5 7 µm are due to atmospheric absorption in the spectrometer. due to arsenic cross incorporation, 19 and comparatively little 0.3% gallium-for-indium replacement due to gallium cross incorporation. The length scales and impurity fractions that characterize these compositionally graded regions may be ascertained from average impurity profiles Fig. 5 reconstructed via a statistical analysis of the STM images from a representative ensemble of heterojunctions. Because the superlattice period in a representative device is often comparable with the decay lengths for anion and cation segregation, this compositional grading can strongly influence the average alloy composition and resulting mismatch of the epitaxial layers. The average layer compositions required to account for the residual compressive strain exhibited by the superlattice in Table I, for example, are Ga 0.94 In 0.06 Sb and InAs 0.94 Sb 0.06, respectively, when typical GaAs-like fractions are adopted at the normal and inverted interfaces. 20 V. MBE SHUTTER SEQUENCING AND MORPHOLOGICAL DEFECTS The MBE shutter sequencing initially chosen to promote InSb-like bonding across both heterojunctions of the superlattice described in Table I was as follows: beginning with the completion of a specified GaSb layer, the gallium shutter was closed and the static GaSb surface soaked for 5 s with antimony. The antimony shutter and cracker valve were next closed, and the indium shutter then opened to deposit a single monolayer of indium without additional antimony or arsenic. During the course of this indium soak, the arsenic cracker valve was opened, and only once the intended cation dose was completed, was the arsenic shutter opened to commence growth of InAs. Upon completion of a specified InAs layer, the indium shutter was first closed, and the static InAs surface continually soaked with arsenic for another 5 s while the antimony valve opened. The arsenic shutter and valve were next closed, and the antimony and indium shutters subsequently JVST B-Microelectronics and Nanometer Structures

4 4 Sullivan et al.: MBE growth of high quantum efficiency InAs/GaSb 4 FIG. 4. Atomic-resolution STM image anion sublattice from a closely related SLS in 110 cross section. Sb As and In Ga denote anion and cation substitutional impurities associated with segregation or cross incorporation across the InAs-on-GaSb and GaSbon-InAs interfaces, respectively. Growth direction is toward the top-left corner of the image. opened to codeposit a single monolayer of InSb. Following closure of the indium shutter, this static surface was again soaked for 5 s with antimony, and the gallium shutter then finally opened to begin another round of GaSb growth. The obvious asymmetry between indium deposition without arsenic at the InAs-on-GaSb heterojunction, and indium deposition in the presence of antimony at the GaSbon-InAs heterojunctions, was adopted to suppress the unwanted displacement of antimony by arsenic with attendant GaAs bond formation at the GaSb surface by way of otherwise thermodynamically favored anion exchange reactions Strained-layer superlattices fabricated with this shutter sequence and lattice matched within the resolution of the x-ray spectrometer nevertheless regularly displayed the small circular surface defects seen in Fig. 6 with a typical density of cm 2. Atomic force microscopy AFM imaging Fig. 7 revealed that these defects, while typically protruding only some 2.5 nm or so above the growth surface, often possessed crevices whose depth exceeded 25 nm. Although the observed size uniformity suggested nucleation early on, like examination of identically prepared GaSb buffer layers demonstrated those growths were virtually free of similar features. Turning next to a possible connection between these defects and indium deposition without a group V flux at the InAs-on-GaSb heterojunction, comparison was then made FIG. 5. Average impurity profiles, reconstructed from an ensemble of heterojunctions, illustrate the typical length scales and impurity fractions associated with cation and anion segregation, respectively. Layer indices are referenced with respect to the underlying superlattice heterojunction. J. Vac. Sci. Technol. B, Vol. 23, No. 3, May/Jun 2005

5 5 Sullivan et al.: MBE growth of high quantum efficiency InAs/GaSb 5 VI. SUMMARY A single-pass internal quantum efficiency of 36% has been demonstrated in a 3-µm-thick InAs/GaSb strained-layer superlattice diode with 11 µm cutoff wavelength. This stateof-the-art performance was achieved using MBE material grown with a shutter sequence designed to suppress arsenicfor-antimony replacement, and thus GaAs-like bond formation, at the InAs-on-GaSb heterojunction. The procedure, which calls for indium deposition absent a simultaneous group V flux, introduces small morphological defects whose influence on superlattice detector properties is presently unknown. Detailed characterization with x-ray diffraction and transmission electron microscopy nevertheless confirm these superlattices possess good structural quality, whereas crosssectional scanning tunneling microscopy images demonstrate atomic-scale compositional grading of both anion and cation sublattices in closely related samples. FIG. 6. Digitally processed, differential interference contrast optical micrograph highlighting the small defects regularly present at the surface of SLSs fabricated with the MBE shutter sequencing described in Sec. V. with an otherwise identical growth in which the arsenic and indium shutters were simultaneously opened at each of these heterojunctions. That apparently small modification yielded a better than a hundredfold reduction in surface defect density cm 2, but simultaneously shifted the overall superlattice strain from compressive 0.2% to tensile 0.4%, as expected. Although this residual tension could easily be balanced by intentional alloying with additional indium, its presence signals the appearance of a substantial, and potentially undesirable, GaAs-like component at the InAs-on-GaSb heterojunction. 20 FIG. 7. Representative AFM scan over one of the defects in Fig. 6. ACKNOWLEDGMENTS The authors are pleased to acknowledge sponsorship of this work by AFRL and MDA, as well as the technical and programmatic support of Dr. s Gail Brown and Vaidya Nathan. 1 F. Fuchs, L. Buerkle, R. Hamid, N. Herres, W. Pletschen, R. E. Sah, R. Kiefer, and J. Schmitz, Proc. SPIE 4288, E. M. Jackson, G. I. Boishin, E. H. Aifer, B. R. Bennett, and L. J. Whitman, J. Cryst. Growth 270, G. J. Brown, F. Szmulowicz, K. Mahalingam, S. Houston, Y. Wei, A. Gin, and M. Razeghi, Proc. SPIE 4999, W. Tennant, Rockwell Scientific Co. private communication. 5 E. R. Youngdale et al., Appl. Phys. Lett. 64, C. H. Grein, P. M. Young, M. E. Flatté, and H. Ehrenreich, J. Appl. Phys. 78, H. H. Hogue, M. L. Mathew, D. Renolds, E. W. Atkins, E. W. Stapelbroek, and G. Maryn, Proc. SPIE 4850, Y. J. Wei, J. Bae, A. Gin, A. Hood, M. Razeghi, G. J. Brown, and M. Tidrow, J. Appl. Phys. 94, Y. Wei, A. Gin, M. Razeghi, and G. J. Brown, Appl. Phys. Lett. 81, E. H. Aifer, E. M. Jackson, G. Boishin, L. J. Whitman, I. Vurgaftman, J. R. Meyer, J. C. Culbertson, and B. R. Bennett, Appl. Phys. Lett. 82, F. Fuchs, L. Buerkle, R. Hamid, N. Herres, W. Pletschen, R. E. Sah, R. Kiefer, and J. Schmitz, Proc. SPIE 4288, Y. Wei, A. Gin, M. Razeghi, and G. J. Brown, Appl. Phys. Lett. 80, A. S. Bracker, M. J. Yang, B. R. Bennett, J. C. Culbertson, and W. J. Moore, J. Cryst. Growth 220, L. Burkle, F. Fuchs, J. Schmitz, and W. Pletschen, Appl. Phys. Lett. 77, J. Harper, M. Weimer, D. Zhang, C.-H. Lin, and S. S. Pei, Appl. Phys. Lett. 73, J. Steinshnider, J. Harper, M. Weimer, C.-H. Lin, S. S. Pei, and D. H. Chow, Phys. Rev. Lett. 85, J. Steinshnider, M. Weimer, R. Kaspi, and G. W. Turner, Phys. Rev. Lett. 85, C. Renard, X. Marcadet, J. Massies, I. Prévot, R. Bisaro, and P. Galtier, J. Cryst. Growth 259, J. Harper, M. Weimer, D. Zhang, C.-H. Lin, and S. S. Pei, J. Vac. Sci. Technol. B 16, M. Zhong, J. Steinshnider, M. Weimer, and R. Kaspi, J. Vac. Sci. Technol. B 22, Q. Xie, J. E. Van Nostrand, J. L. Brown, and C. E. Stutz, J. Appl. Phys. 86, B. Z. Nosho, B. R. Bennett, L. J. Whitman, and M. Goldenberg, J. Vac. Sci. Technol. B 19, R. Kaspi, J. Steinshnider, M. Weimer, C. Moeller, and A. Ongstad, J. Cryst. Growth 225, JVST B-Microelectronics and Nanometer Structures