Low operating bias InAs/GaSb strain layer superlattice LWIR detector

Highlights

• Low dark current LWIR detectors via proper junction alignment.

• Proper Junction alignment via reduction of AlSb layer thickness in widegap layer.

• Arrhenius activation energy plots indicate diffusion as the dominant dark current mechanism.

Abstract

Minimization of operating bias and generation–recombination dark current in long wavelength infrared (LWIR) strained layer superlattice (SLS) detectors, consisting of a lightly doped p-type absorber layer and a wide band gap hole barrier, are investigated with respect to the band alignment between the wide band gap barrier and absorber layers. Dark current vs. bias, photoresponse, quantum efficiency, lifetime, and modeling are used to correlate device performance with the wide gap barrier composition. Decreases in dark current density and operating bias were observed as the conduction band of the wide gap barrier was lowered with respect to the absorber layer. The device achieved 95% of its maximum quantum efficiency at 0 V bias, and 100% by 0.05 V. This study demonstrates key device design parameters responsible for optimal performance of heterojunction based SLS LWIR detectors.

Introduction

Strain layer superlattice (SLS) devices, based on the III–V antimony material system, have demonstrated tremendous design flexibility with desired semiconductor properties such as band gap, band alignment, and suppression of various noise components. Significant numbers of reports demonstrating various SLS materials and device architectures, including infrared detection ranging from short through very long wavelength, have been published. The driving force behind these reports is the ability to engineer the properties of the SLS giving great flexibility to device design. However, this expansive flexibility can also be a hindrance to the advancement of SLS based devices as focus on device optimization has been spread thin across the many possible design variations. Each of these designs capitalizes upon specific features for improved performance. In order to determine optimal device structures, some general guidelines for device design and operation would be useful.

A common component to a majority of the device architectures is a wide gap “unipolar” barrier layer to block majority charge carriers [1], [2], [3], [4], [5], [6], [7], [8], as seen in the device structure illustrated in Fig. 1. This is because implementation of a wide gap layer is effective at reducing dark current contributions of the Shockley–Read–Hall SRH generation–recombination (g–r) process in the space-charge (depletion) region of the absorber, as well as providing a higher resistance to surface leakage current. Homojunction devices that do not possess this wide gap layer are plagued with excessively high dark current due to both surface leakage currents and g–r current. When properly designed and grown, an effective wide gap layer increases the resistance on the surface of the mesa sidewalls, and reduces the depletion in the absorber layer.

Theoretical modeling is advantageous for SLS device design: however, variation in growth, as well as challenges in the direct translation of a modeled design to an MBE growth chamber, often introduce small variances into the as-grown material. These variances in thickness and/or composition can result in slight shifting of the band gaps and band edge alignments at the interfaces between layers. When possible the best designs will incorporate some room for error due to these growth variations. In a previous attempt to design and grow a heterojunction structure it was apparent from the performance of test devices that a band offset was present [1]. The presence of this band offset greatly diminishes the performance of the device, and reduces the benefit of the wide gap layer.

In this study a device architecture including a p-type bottom contact, a p-type LWIR absorber layer, a wide gap lightly p-type heterojunction barrier, and a wide gap n-type top contact was investigated, Fig. 1. The band gap of the barrier layer is more than double that of the absorber layer. Because the conduction bands of the two layers are at the same level, the Fermi level of the p-type wide gap layer will fall below the valence band of the absorber. This will create a potential barrier when the Fermi levels align under the flat band condition for the device. Operation of the device will require the application of sufficient external bias to deplete through the wide gap layer to overcome the potential barrier and collect minority charge carrier electrons. The relationship between the applied bias, V, and the depletion widths within the absorber, (X_a), wide gap barrier, (d), and N+-contact, (X_n), layers have been derived from the solutions to Poisson’s equations:

$$V=\frac{{\mathit{qN}}_a{X}_a^2}{2{\epsilon }_a}+\frac{{\mathit{qN}}_b{d}^2}{2{\epsilon }_b}+\frac{{\mathit{qN}}_a{X}_{a}d}{{\epsilon }_b}+\frac{q{(N_a{X}_a+N_{b}d)}^2}{2{\epsilon }_n{N}_n}-V_0$$

where q is the elementary charge, ε, N, and X are the permittivity, carrier concentration, and depletion widths respectively, and the subscripts a, b, and n represent the absorber, wide gap, and N+-contact layers respectively. V₀ is the built-in potential of the device representing the energy difference of the Fermi levels for the absorber and N+-contact layers with respect to vacuum.

The minimum operating bias (V_min) is defined as the minimum bias voltage required to saturate the photoresponse of the detector. At this bias the depletion width will be fully contained within the wide gap material if the conduction band offset is less than or equal to zero. At V_min, when the built in bias has just been overcome, the depletion width in the absorber layer (X_a) will equal zero. Because of this, at V_min, there will be minimal g–r contribution to the dark current noise. However, as the bias is further increased beyond V_min the depletion will begin to extend into the absorber and g–r current will be collected. When a band offset is present, e.g., ΔE_c > 0, the depletion width in the absorber at V_min, X_a0, will be determined by equating the bias drop in the absorber to ΔE_c yielding:

$$X_{a0}=\sqrt{\frac{2{\epsilon }_a\mathrm{\Delta }{E}_c}{q^2{N}_a}}$$

Band diagrams in Fig. 2 below illustrate the impact of an offset between the absorber and wide gap barrier layers. Initially the offset acts as a barrier blocking current flow; as the bias is increased this potential barrier is lowered. Charge collection will begin when the peak of the barrier is at the same height as the absorber conduction band, Fig. 2(c). Notice that there is already appreciable depletion in the absorber at this point which will lead to the collection of g–r current. In an attempt to avoid this band offset we investigated device structures with band alignments that are intentionally designed to have a “negative offset.” That is, the conduction band for the wide gap barrier was designed to sit at an energy just below the conduction band of the narrow gap absorber. This was accomplished by redesigning the superlattice of the wide gap layer to incorporate thinner AlSb layers. The target for the conduction band of the redesigned wide gap layer is 15 meV below the conduction band edge of the absorber. As will be discussed, we found this was sufficient to lower the conduction band of the wide gap layer below that of the absorber and avoid the formation of a band offset.

A device designed to have a negative offset will accumulate charge in the absorber until sufficient bias is applied to deplete the wide gap barrier and allow charge collection from the absorber, Fig. 3. At V_min, the absorber will not have a depletion region and diffusion is expected to be the dominant current mechanism, Fig. 3(c). This is in contrast with the device having a positive band offset, shown in Fig. 2, where a depletion region is present in the absorber at V_min. However, as the bias is further increased a depletion region will form and the GR component of the dark current will increase approximately as the square root of the bias. Since diffusion current is not a function of bias the g–r current may eventually grow larger than the diffusion current, and at some higher bias value tunneling currents will dominate. A device with a ‘negative offset’ is expected to have a bias range where diffusion current dominates. The extent of this range will depend on the doping of the barrier and absorber layers, as well as the density and energy levels of g–r active defects in the absorber. V_min is dependent on the doping of the barrier and absorber layers, as well as the thickness of the barrier.