Passivation of MWIR Heterostructure p-InAsSbP/n-InAs Photodiodes Using SiO₂ Layers for Near-Room-Temperature Operation ^†

Abstract

We examined the effect of SiO₂ passivation on the parameters of mesa heterostructure InAs/InAsSbP photodiodes with a spectral responsivity 50% cut off at 3.5 µm at 295 K, specific to the InAs absorber layer. The R₀A product was found to increase by 30% after passivation of the devices of 113 µm in diameter, up to 0.9 Ωcm², while for those with a diameter of 1.13 mm, R₀A of 2.2 Ωcm² was achieved, with a value of D* > 3 × 10⁹ cmHz^1/2/W at the peak of the spectrum, 1 kHz, 0 V bias, 295 K. To the best of our knowledge, this is the highest R₀A value at room temperature reported to date for a photodiode with an InAs absorber.

1. Introduction

Although the MWIR spectral region (3–5 µm, medium wavelength infrared radiation) can be covered by different IV-VI and II-VI photodetectors (e.g., based on PbS, PbSe, HgCdTe), InAs photodiodes (PDs) appear to be the best choice for MWIR detection, due to the mature technology of low-cost InAs substrates. The main advantage of InAs PDs is their ability to operate at near room temperature. The elimination of cryogenic cooling reduces cost and results in more reliable operation. InAs photodiodes exhibit excellent noise characteristics at low frequencies and a fast response for detecting rapid processes. Compared to II-VI and IV-VI infrared detectors, InAs photodiodes demonstrate greater parameter stability [1]; however, at room temperature, the shunt resistance of the InAs photodiode is relatively low, comparable to the series resistance, which affects its responsivity [2] (pp. 389–391) [3]. Consequently, uncooled InAs photodetectors do not achieve background-limited performance. This effect is less pronounced in small-area detectors, which have higher shunt resistance and a lower surface area. Their performance can also be optimized by operating at lower temperatures achievable with thermoelectric mini-coolers. However, the contribution of surface leakage to the dark current increases as the temperature decreases and the diode becomes smaller. Therefore, the development of chip processing techniques, including advanced surface passivation, is currently a priority to suppress surface leakage, achieve bulk-limited dark current, and reach the specific detectivity attainable with the InAs compound. This consideration is especially critical for mesa structure chip processing, a method still in common use. Etching that isolates the detector mesa disrupts the semiconductor crystal lattice along the mesa sidewalls, leaving dangling chemical bonds that introduce surface states within the bandgap and pin the surface Fermi level near mid-gap. This causes non-zero surface potential and band-bending that forms a space-charge region at the sidewall, induces near-surface conductive leakage channels where accumulation of majority carriers or their inversion occurs, and facilitates Shockley–Read–Hall generation-recombination (GR), trap-assisted tunneling (TAT) and inter-band tunneling surface currents. The dangling bonds, native oxides and impurities on the surface create the additional surface states within the bandgap and the fixed surface charge that modifies band-bending and the related effects. The exposed to ambient atmosphere InAs reacts with atmospheric oxygen, forming indium oxide with elemental arsenic distributed through the oxide. Native oxides should therefore be avoided and removed from the InAs surface before its passivation. The wide-gap unipolar barrier suppresses the GR current generated at the surface edge and within the volume of the p-n junction. However, surface TAT through the barrier remains possible [4], along with GR processes, outside the barrier, induced by surface band bending. Therefore, surface passivation of barrier detector mesa structures is necessary too. An ideal surface would be a wide-gap insulator, repelling carriers to the active volume of the device, having negligible density of fixed charge and surface states.

InAs photodiodes operating at zero bias with a useful sensitivity, preferably without cooling, are desired for DC, low-frequency, low-power applications, in laser-diode spectroscopy, laser range finding, medicine and environmental monitoring.

2. Materials and Methods

The material composition of the device, doping and thickness of the respective layers are given in the device cross-section diagram in Figure 1. The MWIR absorption region (absorber) of the photodetector (layer 2) is placed between the wide-gap unipolar barrier layer—the top contact for holes—and the heavily n-type doped bottom contact for electrons. The absorber is an InAs layer 1.5 µm thick with unintentional n-type doping 1–3 × 10¹⁶ cm⁻³. The barrier in the conduction band within the contact for holes blocks injection of electrons into the absorber. The highly degenerate n-type layer of the contact for electrons has very few holes so their injection into the absorber is reduced too; thus, the dark current is minimized.

The detector structure was designed using Crosslight Software’s SimuApsys software, a commercially available 2D/3D finite element analysis and modeling tool, which is particularly useful for simulating bulk semiconductor heterostructures. It solves the Poisson equation and continuity equations, taking into account dependence of mobility on the electric field, tunneling, avalanche multiplication, trap assisted mechanisms of carrier generation-recombination, heat transfer, material parameters dependent on temperature, optical phenomena (absorption, reflection, interference) and other factors. The package is equipped with a material database with data source references.

The device was grown, according to its design, in a horizontal MOVPE reactor at 50 hPa total pressure. The substrate used for the grown material was an Si-doped InAs epi-ready 2” wafer, with n-type doping 1.5 × 10¹⁸ cm⁻³, (100) oriented without deliberate surface miscut. The epitaxial growth began with a 100 nm thick n-doped InAs lower contact layer (#1 in Figure 1), followed by the next layers, as illustrated in Figure 1.

The as-grown multilayer films were characterized using optical microscopy, high-resolution X-ray diffraction, photoluminescence, and secondary ion mass spectrometry. The InAsSbP barrier layer was engineered to achieve a minimal valence band offset and an optimal lattice match with the InAs layers.

The passivation layer was SiO₂, grown by inductively coupled plasma chemical vapor deposition. A lower surface leakage and a higher R₀A were observed with a SiO₂ thickness of 50 nm compared to 500 nm. Particular attention was paid to removing photoresist and byproduct residues after both dry and wet etching steps to reduce the risk of increased surface leakage current.

The surface component of the dark current was estimated using variable area diode array test structures, using the following approximation [5,6]:

$${\displaystyle \frac{1}{R_{0}A}} ={\displaystyle \frac{1}{{{(R}_{0}A)}_V}} + {\displaystyle \frac{\mathit{\alpha }}{D}}$$

(1)

where (R₀A)_V approximates the R₀A product of an infinite area diode, D is the diode transverse linear size—diameter of the circular mesa, and α is a coefficient related to the surface generation of carriers at the diode perimeter.

3. Results

Figure 2 illustrates R₀A and DC dark current as a function of bias for the uncooled samples, showing reduced dark current and enhanced R₀A following passivation, for the best diodes across their respective size groups, ranging from the smallest (D = 113 µm) to the largest (D = 1128 µm), with the most significant improvement seen for the smallest devices. However, a several-fold variation in the measured dark current was noted within individual diode groups of a given size, including the forty-three smallest diodes. Additionally, the linear correlation between the reciprocals of R₀A and size D disappeared when the same devices were measured at 215 K.

4. Conclusions

The loss of correlation between diode size and the measured R₀A and dark current at 215 K may result from the dominance of surface carrier generation over volume generation at this temperature, as well as the observed variation in dark current intensity. This suggests that chip-processing technology, particularly cleaning procedures, still requires further optimization.

The obtained detector performance, also taking into account the measured spectral responsivity, surpasses that of commercially available InAs photodiodes [3,7,8]. To the best of our knowledge, we have produced uncooled InAs detectors with the highest R₀A value ever reported—even for barrier InAs diodes [9].