W- Structured type-ii superlattice based long and very-long wavelength infrared photodiodes.

Transcription

1 Invited Paper W- Structured type-ii superlattice based long and very-long wavelength infrared photodiodes. E. H. Aifer, J. G. Tischler, J. H. Warner, I. Vurgaftman, J. C. Kim, J. R. Meyer, B. R. Bennett, and L. J. Whitman Naval Research Laboratory, Washington, DC 2375 E. M. Jackson and J. R. Lorentzen SFA Inc., Largo, MD 2774 ABSTRACT W-structured type-ii superlattices (W-SLs) were initially developed to increase the gain of mid-wave infrared (MWIR) lasers. The design addressed the reduced optical transition matrix elements due to the spatial displacement between valence and conduction band wavefunctions in the type-ii superlattice (T2SL), and further improved the differential optical gain by providing a mostly two-dimensional density of states. As a result, W-SL and W interband cascade lasers have lower thresholds and higher pulsed and cw operating temperatures than any other III-V interband MWIR lasers. These same features give W-SLs desirable properties for IR detectors, and here we report for the first time on characteristics of W-SLs used for long-wave and very long-wave IR photodiodes. IR transmission measurements of W and conventional T2SL photodiodes revealed absorption characteristics that are well described by theory, including line shape and peak absorption coefficient values which are about a factor of 2 greater in the W-SLs. Similarly, the low temperature photoluminescence shows much higher and sharper emission intensity in the W-SLs. While the W-SLs have demonstrated superior optical properties, as predicted, additional work is needed to achieve higher detector quantum efficiency. Results suggest that the excess carrier collection in the W-structures is reduced with respect to similar T2SL structures, especially for the lowest energy state. Possible mechanisms of excess carrier loss, as well as new designs to improve charge collection, in the W-SL, will be discussed. KEYWORDS: type-ii superlattice, strained layer superlattice, GaSb, III-V, IR detector Quantum Sensing and Nanophotonic Devices II, edited by Manijeh Razeghi, Gail J. Brown, Proceedings of SPIE Vol (SPIE, Bellingham, WA, 25) X/5/$15 doi:.1117/

2 1. INTRODUCTION 1.1 Motivation Mercury cadmium telluride (MCT) and blocked impurity band silicon (Si-BIB) are currently the leading material systems for use in missile defense (MD) applications of long-wave infrared (LWIR) and very long-wave infrared (VLWIR) focal plane arrays (FPAs). While they possess many desirable performance characteristics, MCT and Si-BIB also require substantial cooling, with operation at 4-6K of VLWIR MCT 1 and 12K for Si-BIBs. 2 Since FPAs for MD applications are destined for operation in satellite or missile-borne systems, the added weight associated with cooling the FPA to its required operating temperature is a critical issue. Type-II superlattices (T2SLs) are a relatively new material system with the potential to meet MD performance needs, but with reduced cooling requirements, and higher pixel-to-pixel uniformity. T2SLs can thus have a huge impact on MD- FPAs by reducing both payload weight and system complexity Å material system T2SLs are composed of III-V binary and ternary compound semiconductors with lattice constants in the neighborhood of 6.1 Å, as indicated in Fig. 1. Structures are grown by molecular beam epitaxy (MBE) in compositions that are lattice matched to GaSb substrates. 5 GaAs Family InP Family Antimonides Cut-Off Wavelength (µm) InAs 1-x Sb x InAs Mid-IR Ga 1-x In x Sb In 1-x Ga x As GaSb GaAs InP Ga 1-x Al x Sb Al 1-x In x Sb AlAs In 1-x Al x As AlSb 6.1 Lattice Constant (Å) FIG Å material system The 6.1 Å material family offers high performance with enormous design flexibility. It includes wide, medium and narrow gap components, with insulating AlSb, and high mobility InAs with 3K mobilities of 2-3x 4 cm 2 /V s. Individual components have direct gaps with strong optical absorption and emission in the short and mid wave IR. Most importantly, this material family has constituents with both type-i and type-ii band alignments, as illustrated in Fig. 2. CB GaSb.7 ev 1.62 ev.36 ev VB AlSb InAs FIG. 2. Type I and type II band alignment in the 6.1 Å material family at 3K. 26 Proc. of SPIE Vol. 5732

3 1.3 T2SL IR photodetectors 6.1 Å T2SLs, first proposed by Smith and Mailhiot, 3 are composed of alternating layers of type-ii aligned 6.1 Å materials. They provide a direct gap (in k-space) bandstructure with cutoff wavelengths demonstrated from 3-3 µm. 4-7 Electron and hole wavefunctions for an InAs/GaSb superlattice are shown in Fig. 3, along with the band alignments and ground-state energy levels at k=. Ψ n Ψ p H1 InAs (3 Å) E1 λ g = 6.5 µm CB VB GaSb (3 Å) FIG. 3 type II superlattice wavefunctions and bandstructure The negative gap between the InAs conduction band and the GaSb valence bands is overcome by quantum confinement effects induced by the forbidden regions of the neighboring layers. Electron states in the InAs layers are raised in energy while hole states in the GaSb layer are lowered, opening up a positive IR gap. From the wavefunctons it can be seen that spatially, the interband transition involves the displacement of probability density across the InAs/GaSb interface. In k- space however, the dispersion is parabolic and the gap is direct. Since the gap is controlled by the thicknesses of the epitaxially grown layers, T2SLs can provide a very high degree of pixel-to-pixel spectral uniformity. T2SLs are well suited for the realization of infrared photodiodes by virtue of in-growth doping and layer-thickness controlled energy gaps. In-growth doping allows for controlling doping profiles with monolayer precision, independently of layer depth. Also, since bandgap control is largely independent of growth parameters other than layer thickness, the energy gap may be arbitrarily tailored to optimize photodiode performance. Finally, and perhaps most importantly, the independence of doping and energy gap control makes for straightforward implementation of multi-band detectors. 1.4 Intrinsic material advantages There are two basic material advantages that give T2SLs a potential edge over the current leading IRFPA material, MCT. First, the band-to-band tunneling current depends exponentially on mass and bandgap, becoming increasingly severe at very long wavelengths for a system like MCT with a very light electron effective mass. 1/ 2 1/ 2 JTUNNEL = C1m exp( C2m EGap) (1) In T2SLs, band-to-band tunneling is suppressed because the electron mass is at least 3x larger than in MCT (m* MCT ~.1m vs. m* T2SL ~.3m ) resulting in much lower tunneling current in the VLWIR. The effective mass is also largely independent of wavelength, unlike MCT where it decreases with wavelength (m* MCT =.12(12 µm),.7(2 µm)). 8 The second advantage is that Auger recombination may also be suppressed in T2SLs relative to MCT when the superlattice is designed so that the strain induced splitting of the light and heavy-hole bands exceeds the interband gap. Auger processes are then limited to intervalence resonances far below the Fermi level where they are strongly reduced due to low hole-occupancy (Fig. 4). Auger lifetimes in T2SLs have been measured to be a factor of times longer than those found in MCT. 9 Proc. of SPIE Vol

4 C n p EF P p HH p LH FIG 4. Suppression of CHHL Auger process in p-type T2SLs due to strain splitting of hole bands 2. W-STRUCTURED TYPE-II SUPERLATTICES (W-SLs) 2.1 Ternary type-ii superlattices (TSLs) A drawback of the simplest T2SLs, the InAs/GaSb binary type-ii superlattice (BSL), is that the SL layers must be increased in thickness to provide longer cutoff wavelengths, with a resulting loss of quantum efficiency. This occurs because as the electron and hole wells become thicker, the electron and hole wavefunctions become almost completely localized in separate InAs and GaSb layers respectively, reducing overlap and thus lowering the quantum efficiency (Fig. 3). This problem can be addressed by alloying the GaSb hole-wells with indium to form ternary In x Ga (1-x) Sb layers. The introduction of indium causes a proportional increase in the strain between the layers, and thus an increase in the type-ii band offset. The increase in offset means that more confinement is needed to produce the same miniband gap, allowing narrower wells to be used for a given cutoff wavelength, and increasing the optical matrix element (Fig. 5(a)). 3 The trend with increasing In-alloying is shown in Fig. 5(b), where the calculated square of the optical matrix element, M 2 CV (proportional to QE), is plotted as a function of alloy fraction x, for TSLs with fixed cutoff and InAs well thickness. a) 3% In E1 H1 InGaSb Ψ n InAs Ψ p CB VB b) M 2 cv [ev2 -Ang. 2 ] In x Ga 1-x Sb/InAs Superlattice L InAs =15 ML(46 Å) simulation: cutoff=9.4 microns x- Alloy fraction FIG 5. (a) TSL with same energy gap as BSL but with narrower wells and increased overlap. (b) Calculated dependence of M 2 CV on In alloy fraction x. 2.2 W-SL concept W-SLs are well known in mid-ir lasing applications, where they were first proposed by Meyer et al. to enhance gain. In the W-SL, shown in Fig. 6, thin AlSb barriers surround dual InAs electron wells that are symmetrically located on either side of a single In x Ga 1-x Sb hole-well. 262 Proc. of SPIE Vol. 5732

5 CB AlSb GaInSb InAs e1 hh1 VB L1 H2 H3 ψ p ψ n FIG. 6. W-SL bandstructure and wavefunctions. W-SLs can improve the optical matrix elements by increasing the electron overlap within the hole well, although the main advantage comes from the nearly 2-dimensional electron density of states arising from strong confinement by the AlSb barriers. This gives the W-SL a much sharper absorption edge than materials whose dispersion relations are more 3-dimensional, including other T2SLs. A key result is that W-SLs can have higher peak quantum efficiency at longer wavelengths than other structures having the same or even substantially lower energy gaps. To illustrate this, calculated absorption spectra for a W-SL and a TSL with the same energy gap of 62 mev (2µm) are compared against the spectral emission of a 23K black body (Fig. 7(a)). Not only is the absorption peak higher at longer wavelengths, but the convolution of the W-SL absorption with the black body emission (Fig. 7(b)) is more than twice as large as for the TSL. 3 Wavelength (µm) Wavelength (µm) 2 a) Absorption Coeff. (cm -1 ) 2 λ gap = 2µm TSL W-SL 23 K Black Body Photon Energy (mev) b) 23 K BB Absorp. (Arb.) λ gap = 2 µm W-SL TSL Energy (mev) FIG. 7. (a) Calculated absorption spectra for a W-SL (blue) and a TSL (red) with the same energy gap of 62 mev, compared with emission from a 23K black body. (b) Convolution of W-SL (blue) and TSL (red) absorption with the 23K black body. 2.3 W-SL interface configurations Interfaces (IFs) play a large role in W-SLs, where there are twice as many per period than in normal SLs, and the total thickness of the interface bond layers is nearly 5% of that of a typical AlSb barrier (5 ML). They also provide more design options to balance strain. Taking only the four simplest interface types, InSb, GaAs, AlAs and free (assumed lattice-matched), there are at least 12 practical W-SL interface configurations one may use, as illustrated in Fig. 8. Proc. of SPIE Vol

6 AlSb GaSb InGaSb > (I) InSb-IF (F) Free-IF (A) AlAs-IF (G) GaAs-IF W-IIII W-IFFI W-FIIF W-FFFF W-IGGI W-FGGF W-FIGF W-IIGI W-AIIA W-AFFA W-AGGA W-AIGA FIG 8. W-SL layer and interface-bond lattice constants, and 12 practical W-SL interface configurations. All of the W-SLs depicted in Fig. 8 are practical, in that they can be lattice matched to GaSb substrates while keeping the indium alloy concentration in the InGaSb hole wells to 35% or less, indicated by the blue-to-purple gradient. In Fig. 9, calculated W-SL bandgap wavelength is plotted vs. InAs layer thickness for 4 of the interface configurations illustrated in Fig. 8, IIII, FIIF, FFFF and AFFA. Each data point is labeled with the indium alloy fraction x required to balance the strain. Bangap wavelength (µm) W-Interface config. IIII FIIF FFFF AFFA data labels = In fraction x InAs Thickness (ML) FIG. 9. Calculated bandgap energy vs. InAs well thickness for W-SLs lattice matched to GaSb, with 4 of the interface configurations defined in FIG. 8. The data points are labeled with the indium alloy fraction required to strain balance the structure with basic period composition 5 ML AlSb / N-ML InAs / 9 ML InGaSb / N-ML InAs Proc. of SPIE Vol. 5732

7 The plot shows that over the range from to 2 µm there are at least 3 different W-SLs for every wavelength. At 14 µm for example, W-SLs with 14, 15, 16, and 17 ML InAs wells may be used. This design flexibility can be an important advantage over basic SL structures. 2.4 W-SL PIN photodiode Once the heterostructure is designed, band parameters calculated using a 16 band k p method are used to optimize the W-SL PIN photodiode structure. At present, device optimization is constrained by p-type background doping levels in the high 15 cm -3 range, and minority carrier lifetimes and vertical diffusion lengths that are 1 n-sec and <.5 µm, respectively (recently measured from 5.3 to K using electron beam induced current on T2SL diode cross-sections 11 ). The W-SL PIN structure shown in Fig. is arranged with a thin n+ InAs cap, and a p+ GaSb bottom contact, doped to 4x 17 cm -3 and 18 cm -3, respectively. Short, heavily doped n and p-sls are adjacent to the respective contacts, with the lightly doped ISL region in between. n-sl i-sl Parameter Value Units Temperature 77 K Energy gap/λg 118/.5 mev/µm absorption coefficient 2 1/cm electron mobility cm2/v-s p-sl hole mobility cm2/v-s electron lifetime.2 - nano-sec hole lifetime.2 - nano-sec a) p-gasb Substrate b) electron mass.4 Mo hole mass.4 Mo FIG.. (a) W-SL PIN structure. (b) Parameter values used in the simulations The PIN structures were simulated using SILVACO, 12 a numerical 2-dimensional transport solver, and results are shown in Fig. 11. Additional parameters used in the simulation are given in the table in Fig. b, including electron and hole mobilities of and cm 2 /V-sec, that were roughly approximated from Hall measurements of T2SLs grown on insulating,.5-µm-thick AlSb buffers on p-gasb substrates, 13 and the increased electron mass in the W-SL. The absorption coefficient of 2 cm -1, which comes from transmission measurements, is typical of the peak value near cutoff. In Fig. 11(a), photodiode conduction and valence bands, and the electric field along the growth direction, are plotted for various ISL doping levels. Since the vertical diffusion lengths in T2SLs are short, 11 one would like to maximize the space charge region (SCR) in the ISL, to maximize drift-transport of the photo-excited carriers. The electric field along the growth direction is plotted at the bottom of Fig. 11(a) for the same doping levels used to calculate the bands. For doping levels below 15 cm -3, the electric field exceeds V/cm across the entire 1 µm ISL. Internal (no reflection losses) quantum efficiency (QE) vs. ISL doping is plotted in blue in Fig. 11(b), showing modest improvement with reduced doping. Similarly, an order of magnitude increase in minority carrier lifetime also yields only a marginal improvement in the QE. This indicates that the ISL should be extended to take advantage of the additional diffusion current resulting from increased lifetimes. On the other hand, the dark current performance improves strongly with increasing ISL doping, as can be seen in the plot of dynamic impedance-area product (R A) vs. ISL doping in red in Fig 11(b). Since the simulation does not take into account Auger or trap assisted tunneling processes, however, the actual increase in R A with ISL doping will probably saturate rapidly at doping levels above 16 cm -3. It appears the overall detectivity will be highest ( * ) for doping levels from 5x cm -3. D QE R A Proc. of SPIE Vol

8 Z(um) n+sl I-SL p+sl p+gasb (a) E[eV] Electric field Z [V/cm] conduction band E F valence band 1E14 5E14 1E15 ISL Na(cm -3 ) 5E15 1E16 E-field (z-comp) ISL Doping 1E14 5E14 1E15 5E15 1E Z[um] (b) Internal Quantum Efficiency Simulated QE and RA at 77K RA QE 1E-8 5E-9 SRH lifetime = 1E-9 SRH lifetime = 1E-9 {2E-} 1E-8 5E-9 2E- increasing carrier lifetimes 1E14 1E15 1E16 I-Layer Doping (cm -3 ) 1.1 RA (Ohm-cm 2 ) FIG. 11(a). SILVACO simulation of W-SL PIN bandstructure and electric field along the growth direction. (b) Simulated QE and R A as a function of ISL doping and minority carrier lifetime. 3. W-SL RESULTS 3.1 Structural properties For this study we grew a set of 5 W-SL photodiodes designed with IIII, FFFF, IFFI and FIIF interface configurations (following the conventions of Fig. 8). The sample structure parameters are given in the table in Fig. 12. WAFER IF ISL thckns ISL p-dopng InAs In(x)GaSb AlSb InSb x % Cutoff UNITS microns cm^(-3) ML ML ML ML % microns T FFFF 1 5.E %.5 T FIIF 1 5.E % 9.9 T FFFF 2 6.E %.5 T IFFI 2 5.E %.3 T b IIII 2 5.E %.8 FIG. 12. Design values of structure parameters for InAs/In(x)Ga(1-x)Sb/InAs/AlSb W-SL photodiodes. The wafers were grown in a Riber compact 21T molecular beam epitaxy (MBE) system on epi-ready, undoped GaSb substrates from Wafer Technology with p-type background doping levels of about 2x 17 cm -3. Diatomic antimony and arsenic fluxes were provided by sources equipped with valved crackers, and a second indium source was used for alloying in the InGaSb layers. InSb interfaces were forced using the migration enhanced epitaxy technique (MEE), 14 where a monolayer of In is deposited on a just-completed layer after a brief stabilizing soak in its group-v element, followed by the growth of the next III-V layer. Unforced interfaces were produced simply by closing one set of III-V shutters (and arsenic valve for the InAs layers) and then opening the next set. 266 Proc. of SPIE Vol. 5732

9 While one of the samples (T W-IFFI) exhibited a high density of mound defects that precluded device fabrication, x-ray diffraction (XRD) revealed in general, highly uniform growths with numerous satellites and narrow peak widths, as indicated by the XRD spectra of sample T W-FFFF in Fig a_perp=6.743 a_sl=6.851 PCT.STRN=-.177 SLPERIOD=131.3 ML.SL=43.2 FWHM.AVG=3 inset 5 3 SL[-1] GaSb SL[] SL[+1] 42 a-sec % = -.18% Counts Avg. SL FWHM = 3 a-sec - 2 δφ[a-sec] δφ[a-sec] FIG. 13 XRD spectra of sample T W-FFFF. The XRD spectra also revealed however, that the structures had more strain than anticipated in the design, with SL lattice constants typically.15% shorter than expected. The cause of this excess strain became apparent under crosssectional scanning tunneling microscopy (XSTM). 15 In Fig. 14, the group-v sub-lattice of the W-FFFF and W-FIIF samples are shown in filled-state grey-scale images. T W-FFFF (AFFA) T W-FIIF (AIIA) AlAs IF InAs AlSb InGaSb InSb IF AlAs IF FIG. 14 Filled-state XSTM images of the W-FFFF (left) and W-FIIF (right) group-v sub-lattices. The (1) cross-sectional surface is displayed in the 36x36 nm 2 scans, with the substrate towards the upper left. Proc. of SPIE Vol

10 XSTM provides chemical identification through sensitivity to the local bond length and electron density of states. The dark bands or rows of atoms bordering the AlSb layers in both samples reflect the presence of As atoms bonded to Al atoms, with the shorter bond length (5.665 Å) reducing the As atom height. In the image of the W-FIIF sample on the right, the leading row of Sb atoms in the InGaSb layers appears brighter due to the increased height of the Sb atoms bonded to In ( Å). The images show that for both samples, the AlSb/InAs interfaces are predominantly AlAs bonded rather than neutral, despite the fact that these interfaces were unforced. On the other hand, the InGaSb/InAs interfaces in both samples appear to conform to the intended type. Treating the structures as W-AFFA and W-AIIA gives much better agreement with XRD results, although there still remains a small amount of strain to be accounted for. 3.2 Optical properties The optical absorption coefficient was determined from IR transmission of samples with substrates thinned to µm or less at a slight bevel of.25. The sample transmission was referenced to the transmission of an adjacent piece of the thinned wafer, in which all epi-layers had also been removed. As predicted, the peak absorption strength was found to be significantly stronger in the W-SLs than in ternary SLs. Transmission data for three of the five samples grown for this study are shown in Fig. 15(a), labeled by interface configuration. There is little difference in the peak absorption coefficients, either with regard to interface type or ISL thickness. The absorbtance of the W-FFFF sample, with the 2- µm-thick ISL, and a simulation are plotted in Fig. 15(b). The absorbtance of the W-FFFF structure is modulated by Fabry-Perot fringes of the epi-layer, and is also slightly higher than predicted. This is probably due to uncertainty in the relative thickness of the sample and its reference described above. The measured W-SL does, however, exhibit the expected lineshape characteristics, and is at or near the expected absorption strength. (a) Absorbtion coefficient (cm -1 ) K Binary Ternary WSL W-FFFF 2um ISL W-FFFF W-FIIF Wavelength (µm) (b) Absorption coefficeint (cm -1 ) measured W-F abs coeff F-P fringes 2.64 µm tk SL calculated abs. coeff. HH-E1S HH-E1A Photon energy (mev) FIG 15. (a) Peak absorption in the W-SLs vs. TSLs. (b) Measured vs. simulated absorption coefficient for the W-FFFF structure. The results are very different, however, for the photoluminescence (PL) intensity measurements shown in Fig. 16(a). Here there is a significant variation with respect to interface type, and the W-FFFF samples have two or more times the emission intensity of samples with forced InSb interfaces. Unlike absorption, which depends almost exclusively on the optical matrix elements and densities of states, the emission intensity also reflects defect properties like the non-radiative recombination rate. For this reason there is a strong correlation between quantum efficiency and PL, which is demonstrated in Fig. 16(b). LH-E1S 268 Proc. of SPIE Vol. 5732

11 (a) Relative PL Intensity (a.u.) Wavelength (µm) W-FFFF W-FFFF 2µm ISL W-IIII 2µm ISL W-FIIF W-IFFI 2 µm ISL Energy [mev] T=4K Si Filter (b) Normalized Peak QE B-F/W-F B-F/W-F W-FIIF W-FFFF 2um ISL W-FFFF Relative PL Intensity (arb.) FIG. 16. (a) PL for W-SLs with various interface configurations. (b) Strong correlation between QE and PL. 3.3 Carrier losses in transport One can get a measure of the loss of photo-excited carriers in transport by comparing the absorption coefficient obtained directly from transmission, α, with an effective absorption coefficient extracted from the QE spectrum, α QE 1 QE( λ) αqe = ln 1 (2) d (1 R) χeff where QE is the external quantum efficiency, d is the superlattice thickness, R is the reflection coefficient, and χ eff, the fraction of photo-generated carriers that contribute signal current, is used as a fitting parameter. α QE and α are plotted in Fig. 17 for the W-FFFF sample with 2µm ISL. Accordingly, only 29% of the photo-generated carriers in this structure are collected at the diode contacts. This compares with 45% for the W-FFFF with a 1µm ISL, and 32% for the W-FIIF structure with 1µm ISL. Absorbtion coeff. (cm -1 ) Absorp. coeff. from 8K χ eff =.29 WSL-FFFF 2µm ISL SL tk = 2.64µm index=3.3 absorp. coeff. from 8K (χ eff =1) Energy (mev) FIG. 17. Absorption coefficient from transmission (red/dash-dotted) compared with absorption extracted from QE with χ eff =1 (grey/solid), and χ eff =.29 (blue/dashed). Carrier loss during transport indicated by χ eff, the fraction of photo-generated carriers that contribute photo-current. One can define an effective minority carrier electron diffusion length L eff, by setting the fraction of collected carriers χ eff equal to the fraction of carriers that diffuse across the ISL of width d (neglecting drift in the depletion regions), Proc. of SPIE Vol

12 χ eff d 1 = e d x / Leff dx. (3) This gives effective diffusion lengths of.53 and.6 µm for the W-FFFF structures with 1 and 2 µm ISLs respectively, and.34 µm for the W-FIIF 1 µm ISL device. The indication is that carrier recombination rates are about the same in the W-FFFF samples, and are significantly higher in the W-FIIF sample. 3.4 Device characteristics The normalized current/power responsivity showed good spectral uniformity over the 6 mm die, with a std dev/mean of only.16% about the mean µm cutoff. This is evident in the spectra of 34 devices on the W-FFFF 2µm ISL, shown in Fig. 18. Since the bandgap is primarily determined by the superlattice period, the data reflect a high degree of layer-thickness uniformity across the wafer. Normed Responsivity [a. u.] T=8K 5% Cutoff (µm) MEAN=13.11 SD=.21 MIN=13.8 MAX=13.17 N= Wavelength [µm] FIG. 18. Normalized current/power responsivity spectra for W-FFFF with 2µm ISL. The magnitudes of the responses for the same devices, however, are not as uniform. The data are plotted as QE spectra in Fig. 19(a). A histogram of the peak QE at 124meV is shown in Fig 19(b), where the std. dev./mean is 16%. Also, the QE clearly suffers from carrier losses as described above, resulting in a mean value of the peak QE of only 8.3% (or 9% excluding devices with QE 6%). For the 1µm ISL W-FFFF sample, the mean QE was 8.4%, and it was 6.5% for the W-FIIF sample. (a) Quantum Efficiency [%] 5 T C1(41214) BB BP=W692-9 T=8K Energy [mev] (b) Quantity WSL-F at 8K PEAK QE: λ PK = µm E PK =124meV MEAN=8.34 SD=1.34 MIN=4.68 MAX=9.52 SUM.N=34 <QE>=9% External Quantum Efficiency(%) FIG. 19. (a) QE spectra for the W-FFFF 2µm ISL sample. (b) Histogram of the peak QE at 124 mev. Dark current measurements were also performed on three of the W-SL samples. Current density vs. bias (J-V) plots for the W-FFFF 2µm ISL sample are shown in Fig. 2(a). While field dependence in reverse bias is clearly evident, dynamic 27 Proc. of SPIE Vol. 5732

13 impedance values at zero bias were not far from the values predicted by transport simulations that neglected tunneling current, ~.5-1 Ω-cm 2 at 8K. The activation plot of R A in Fig. 2(b), however, indicates that excess carrier recombination dominates the devices below about 77K, where the hard roll-over is consistent with trap assisted tunneling processes. 16 (a) J(A/cm 2 ) E-3 T=8K W-F 13µm cutoff V PN (Volts) (b) RA(Ohm-cm 2 ) E-3 T Diff ~77K K 77-2 K E ACT (mev): 91 5 λ ACT (µm): 13.6 >> W-FFFF: λ CO = 13µm W-FIIF: =.2 µm W-FFFF: =13µm (2µm ISL) /T(K) FIG 2. (a) Dark current density vs. bias for the W-FFFF 2µm ISL sample. (b) R A activation plot for 3 W-SLs. CONCLUSIONS We reported an initial study of W-SL LWIR photodiodes, focusing on the impact of interface configuration on the material and device properties. Unforced AlSb/InAs interfaces turned out to be predominantly AlAs bonded rather than neutral, resulting in additional strain. While the absorption strength may not be influenced by interface configuration, the PL intensity appears to degrade in samples with any forced InSb interfaces. Relative PL intensity shows strong correlation with QE, and the one InSb IF sample that QE was measured on is low relative to the free IF samples. The discrepancy between the absorption coefficient obtained from transmission and that extracted from QE was used to estimate the minority carrier losses and diffusion lengths. The diffusion length in the W-FIIF sample was about 4% shorter (.34µm) than the W-FFFF samples. Device results indicate good uniformity in cutoff wavelength, with nonuniformity (σ/mean) <.2%, but there was 16% non-uniformity in peak QE. Also, the peak QE levels were only in the 6-9% range, as a result of carrier losses rather than absorption strength. The dark current performance at 8K and above was scattered, but was relatively much better than the QE. Below 8K, the performance degraded sharply, most likely due to trap assisted tunneling. ACKNOWLEDGEMENTS We wish to express our gratitude to Gernot Hildebrandt of Rockwell Scientific for providing additional quantum efficiency measurements of samples in this study. We also thank the Office of Naval Research and the Missile Defense Agency for supporting this work. REFERENCES 1 M. B. Reine, Infrared Detectors and Emitters: Materials and Devices, (Kluer Academic, Boston 21), p M. A. Kinch, J. Elec. Mat. 29, 89 (2). 3 D. L. Smith and C. Mailhiot, J. Appl. Phys. 62, 2545 (1987). Proc. of SPIE Vol

14 4 W. W. Bewley, E. H. Aifer, C. L. Felix, I. Vurgaftman, J. R. Meyer, C. H. Lin, S. J. Murry, D. Zhang, and S. S. Pei, Appl. Phys. Lett. 71, 367 (1997). 5 F. Fuchs, U. Weimer, W. Pletschen, J. Schmitz, E. Ahlswede, M. Walther, J. Wagner, and P. Koidl, Appl. Phys. Lett. 71, 3251, (1997). 6 Y. Wei, A. Gin, M. Razeghi, and G. J. Brown, Appl. Phys. Lett. 81, 3675 (22) 7 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, 4411 (23). 8 R. Dornhaus and G. Nimtz, Springer-Verlag Tracts in Modern Physics, vol. 98 (Springer, Berlin 1983), p E. R. Youngdale, J. R. Meyer, C. A. Hoffman, F. J. Bartoli, C. H. Grein, P. M. Young, H. Ehrenreich, R. H. Miles and D. H. Chow, Appl. Phys. Lett. 64, 316 (1994). J. R. Meyer, C. A. Hoffman, F. J. Bartoli, and L. R. Ram-Mohan, Appl. Phys. Lett. 67, 757 (1995) 11 J. V. Li, S. L. Chuang, E. M. Jackson, and E. H. Aifer, Appl. Phys. Lett. 85, 1984 (24). 12 computer code SILVACO, (SILVACO International, Santa Clara, CA, 2). 13 E. H. Aifer, 24 (unpublished). 14 B. R. Bennett, B. V. Shanabrook, R. J. Wagner, J. L. Davis, and J. R. Waterman, Appl. Phys. Lett. 63, 949 (1993). 15 B. Z. Nosho, W. Barvosa-Carter, M. J. Yang, B. R. Bennett, and L. J. Whitman, Surf. Sci. 465, 361 (2). 16 Q. K. Yang, F. Fuchs, J. Schmitz, and W. Pletschen, Appl. Phys. Lett. 81, 4757 (22). 272 Proc. of SPIE Vol. 5732