We present an extensive characterization of the minority carrier transport properties in an nBn mid-wave infrared detector incorporating a Ga-free InAs/InAsSb type-II superlattice as the absorbing region. Using a modified electron beam induced current technique in conjunction with time-resolved photoluminescence, we were able to determine several important transport parameters of the absorber region in the device, which uses a barrier layer to reduce dark current. For a device at liquid He temperatures, we report a minority carrier diffusion length of 750 nm and a minority carrier lifetime of 200 ns, with a vertical diffusivity of 3 × 10−2 cm2/s. We also report on the device's optical response characteristics at 78 K.
Type-II superlattices (T2SLs), primarily utilizing antimonide-based III-V materials, have been the subject of a large amount of interest for both their mid-wave and long-wave infrared (MWIR and LWIR, respectively) detection capabilities.1 Progress in III-V materials engineering has led to practical demonstrations of photodiode detectors utilizing T2SL absorber regions,2,3 whose performances are promising, but ultimately far from the predicted theoretical ideal.4 More recently, novel device structures have been developed to improve T2SL photodetector performance to be competitive5 with HgCdTe (MCT) alloy-based detectors, the current favored material system for commercial mid-IR detectors. Recent innovations have explored methods of engineering the device band structure to mitigate harmful leakage currents that arise in junctions formed with narrow bandgap T2SLs. One such structure is the nBn detector,6 a unipolar device comprised of an n-type absorber layer, a thin barrier layer, and a top n-type contact layer, as depicted schematically in Fig. 1. The thin majority carrier barrier layer suppresses dark current by utilizing a high band gap blocking layer to block the flow of majority carriers while allowing optically generated minority carriers through due to a minimal valence band offset. This suppresses Shockley-Read-Hall recombination by keeping the depletion of the absorber region to a minimum.
While most initial nBn detector studies have focused on InAs/GaInSb T2SLs for MWIR and LWIR applications, there has been growing interest in Ga-free InAs/InAsSb T2SL structures. Such devices have demonstrated reduced dark currents7 and longer minority carrier lifetimes, hypothesized to result from the absence of native defects associated with GaSb.8 As we have demonstrated previously for p-i-n design T2SL detectors,9 electron beam induced current (EBIC)10 is a useful technique for characterizing carrier transport, particularly in samples intended for use in detector elements or arrays where excited carriers can be collected via electrical contacts. In the EBIC technique, a high energy electron beam is focused on the surface of a sample to generate excess carriers, which can be measured as an electrical current. Fig. 1 contains a simple illustration of the EBIC experiment. For a more detailed discussion of the fundamentals of EBIC, we refer readers to previous work by Bonard and Ganière.11
In the present work, we demonstrate the utility of the EBIC technique for characterization of an nBn detector based on a Ga-free, InAs/InAsSb T2SL absorber region. We also demonstrate that the EBIC analysis can be supplemented with time-resolved photoluminescence (TRPL) characterization of the same sample. This allows the minority carrier lifetime to be measured independently from the minority carrier diffusion length. By combining both measurements on the same device, the minority carrier diffusion coefficient in the growth direction (i.e., the vertical diffusivity) is obtained. The detector structures are characterized by photoluminescence (PL) spectroscopy and TRPL, photoresponse spectroscopy, and beam energy dependent EBIC, providing a comprehensive experimental characterization of both the devices' optical properties and the nonequilibrium carrier dynamics in the T2SL absorber.
The sample used for measurements was grown via molecular beam epitaxy on an undoped GaSb substrate. The epitaxial layers consist of a 500 nm GaSb buffer, a 950 nm n-type T2SL bottom contact layer, a 2.4 μm (256 periods) InAs/InAsSb T2SL absorber layer, the wide-bandgap superlattice barrier layer, and a T2SL top n-type contact layer. The T2SL absorber structure used was 49 Å InAs/45 Å InAs0.81Sb0.19, resulting in a superlattice bandgap of 5.5 μm. The bottom contact was created using 950 nm (101 periods) of the same T2SL structure as the absorber, doped n-type using Si to 1 × 1018 cm−3. Similarly, the top contact was formed with 96 nm (10 periods) of the absorber T2SL structure followed by 30 nm of bulk InAs. The top contact was uniformly doped n-type at 1 × 1018 cm−3 also using Si. The barrier layer consisted of 20 periods of a 29.3 Å InAs/21.7 Å AlGaSb superlattice. Devices were formed by etching mesas down to the bottom contact layer using a standard lithography and chemical wet etch process, followed by deposition of Ti/Pt/Au metal contacts.
The optical responsivities of the fabricated detectors were measured via Fourier transform infrared spectroscopy (FTIR). Uncalibrated spectra of the device were taken using a Bomem DA-8 FTIR instrument with a globar source and CaF2 beamsplitter, and a Stanford Research Systems SR570 current preamplifier to measure the photocurrent of the device. Fig. 2 shows the resulting responsivity spectrum of a fabricated detector cooled with liquid nitrogen to 77 K at a bias of 0.35 V, with both the raw data (grey) and a smoothed interpolation (red) plotted. Note that forward bias on the device refers to a positive voltage applied to the bottom n-contact. Fig. 2 also shows PL spectrum taken at 77 K for comparison. The PL data were acquired using a Bruker V80v FTIR operating in amplitude modulation step-scan mode. For the PL measurement, the sample was mounted in a cryostat behind a ZnSe window and excited using a 980 nm diode laser pulsed at 45 kHz and incident at −45° through a quartz window. The excited PL was collected at +45°, collimated via a Ge lens (which blocks the pump laser) and fed to the FTIR. The response of the internal FTIR MCT detector is taken to an external lock-in amplifier and returned to the FTIR for processing. From the collected PL spectra, we observe strong emission from our sample at a wavelength just under 5 μm, in good agreement with the cut-off wavelength of our detectors observed in the responsivity data.
TRPL measurements were carried out using a 1064 nm pump laser with a 4 ns pulse width and 1 kHz repetition rate. The sample was housed in a closed-cycle He cryostat for temperature control and excited with a range of optical powers. The resulting emission was collected by a fast MCT detector and recorded with a high speed oscilloscope. The various transient curves are plotted in Fig. 3 for a sample temperature of 16 K. The minority carrier lifetime was obtained from these curves by fitting the 3 μW data with a single exponential decay function. The model fits the portion of the TRPL curve after t = 200 ns, corresponding to non-degenerate carrier concentrations expected in normal operation of the nBn photodetector. From this measurement, the hole lifetime in the absorber region was estimated to be 200 ns. As can be seen from the slopes of the higher excitation power data in Fig. 3, similar hole lifetimes are obtained for increasing pump pulse energies when fitted at low carrier concentrations (t ≥ 200 ns). While our measured lifetime is lower than that observed in studies of state-of-the-art InAs/InAsSb T2SLs,12 they are well within the bounds of reasonable performance observed in the literature.13
EBIC measurements were carried out as described in a previous publication.9 Measurements were taken at various electron beam energies, ranging from 5 keV to 25 keV at a probe current of 1 nA. The EBIC data for various beam energies at a fixed temperature of 6 K, as well as the corresponding modeled fits are shown in Fig. 4. The data exhibit a maximum in signal near the top contact of the device, defined as x = 0, with a drop-off in signal as the beam moves away from this point, through the absorber, and towards the substrate. With increasing beam energies, a shoulder appears in the EBIC data, beginning at beam positions corresponding to the absorber/barrier interface and decaying as the beam position moves towards the bottom contact. Using the standard Bonard and Ganière11 method to fit the data, we were unable to replicate the obtained EBIC signal without introducing a thin artificial region of unrealistically high surface recombination and low diffusion length. The unique shape of the nBn detector EBIC profiles (sharp peak with a shoulder appearing at increasing beam energies) suggests that the collection efficiency of minority carriers generated in the absorber layer is less than unity, since the rise in signal outside of the peak (in the shoulder) does not trend towards the global maximum. This is likely due to the presence of a relatively small minority carrier potential barrier due to valence band mismatch between the absorber and barrier layers. To account for this, we modeled our EBIC results using two separate contributions to EBIC current. Using Monte Carlo simulations,14 we extracted the distribution of energy absorbed from inelastic collisions of electrons (i) within the barrier layer and (ii) within the absorber region for each position and energy of the electron beam. These absorbed energy distributions represent electron-hole pairs (EHPs) generated (i) directly in the higher band gap barrier layer, which are swept away by the electric field in the barrier region to create drift current and (ii) holes generated in the absorber region, which must diffuse to the top contact, across the small potential barrier at the absorber/barrier interface. The first current contribution (drift) dominates the EBIC signal when the electron beam spot is within or near the barrier layer and leads to the peaks (at x = 0) observed in the data. The second current contribution (diffusion) is responsible for the shoulder observed in the absorber region of the nBn detector.
We combined the modeled drift current signal with the traditional diffusion current from EHPs generated in the absorber layer, via a weighted sum, to reconstruct the total EBIC current. The ratio of diffusion current to drift was determined to be 0.35 via empirical comparison of the modeled signals' shoulder heights to the data. It is worthwhile to mention that in the typical application of an nBn detector, the wide gap of the barrier layer provides enough selectivity to ignore photogenerated carriers within it. In the EBIC measurements, where the electrons used to generate current have much higher energy and are focused at a very small beam spot, we cannot ignore such a contribution to current, even when we are primarily concerned with the behavior of carriers in the absorber region. For this reason, the barrier band gap is not observed in the optical data presented in Fig. 2.
The resulting modeled data, obtained from the Monte Carlo simulations combined with our drift/diffusion model of carrier collection, are shown in Fig. 4(b) alongside the experimental data (Fig. 4(a)). We note that the theoretical data follow the same trend as the experimental data, with a narrow peak followed by a shoulder, whose signal decays with increasing distance from the top contact. Additionally, by using the same diffusion length and surface recombination velocities for each beam energy in the model, we were able to replicate the beam energy dependence of the data, obtaining close fits to each set of experimental data by only changing the carrier generation profile due to beam energy (obtained by Monte Carlo simulations).
From the data in Fig. 4, we were able to extract a hole diffusion length in the absorber region of 750 nm and a surface recombination to diffusivity ratio of 106 cm−1 at T = 6 K. Based on the fit of the surface recombination to diffusivity ratio, we estimate a surface recombination velocity of 3 × 104 cm/s. The surface recombination parameter becomes increasingly less sensitive as its value grows larger and this value is an estimate of the lower bound for the true surface recombination velocity. Nonetheless, the large surface recombination velocity of our result compared to previous modeling done on comparable InAs/GaSb T2SL pn junction photodiodes15 suggests a strong contribution to the effective carrier lifetime by the unpassivated surface of the InAs/InAsSb T2SL. This suggests that device fabrication methods which avoid etching below the barrier layers should be used to optimize device performance. In conjunction with the lifetime measured via TRPL, we estimate a hole diffusivity of 3 × 10−2 cm2/s. Using Einstein's relation,16 a hole mobility of 60 cm2/V s is estimated. This value is an order of magnitude larger than the vertical hole transport reported recently for InAs/GaSb T2SLs,17 suggesting that the issue of hopping transport for holes in our Ga-free T2SL is less severe than for holes in an InAs/GaSb T2SL.
In conclusion, we have demonstrated minority carrier diffusion length and lifetime characterization using EBIC and TRPL for an nBn device structure utilizing a type-II InAs/InAsSb superlattice absorber layer. By studying the dependence of the EBIC data on the electron beam energy, we were able to characterize the sample's surface recombination characteristics as well as the minority carrier diffusion length. When combined with the lifetime via TRPL data, we were able to additionally determine the hole vertical mobility and diffusivity, providing a comprehensive picture of device performance and excited carrier dynamics in Ga-free nBn T2SL detectors.
This research was carried out as part of an Army Research Office Multi-Disciplinary Research Initiative under Grant No. W911NF-10-1-0524. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under Contract No. DE-AC04-94AL85000.