Search

The use of silicon as an alternative substrate to bulk CdZnTe for epitaxial growth of HgCdTe for infrared detector applications is attractive because of potential cost savings as a result of the large available sizes and the relatively low cost of silicon substrates. However, the potential benefits of silicon as a substrate have been difficult to realize because of the technical challenges of growing low-defect-density HgCdTe on silicon where the lattice mismatch is ∼19%. This is especially true for long-wavelength infrared (LWIR) HgCdTe detectors where the performance can be limited by the high (∼5 × 106 cm−2) dislocation density typically found in HgCdTe grown on silicon. The current status of LWIR (9 μm to 11 μm at 78 K) HgCdTe on silicon focal-plane arrays (FPAs) is reviewed. Recent progress is covered including improvements in noise equivalent differential temperature (NEDT) and array operability. NEDT of 99% are highlighted for 640 × 480 pixel, 20-μm-pitch FPAs.

We have fabricated a series of 256 pixel×256 pixel, 40 µm pitch LWIR focal plane arrays (FPAs) with HgCdTe grown on (211) silicon substrates using MBE grown CdTe and CdSeTe buffer layers. The detector arrays were fabricated using Rockwell Scientific’s double layer planar heterostructure (DLPH) diode architecture. The 78 K detector and focal plane array (FPA) performance are discussed in terms of quantum efficiency (QE), diode dark current and dark current operability. The FPA dark current and the tail in the FPA dark current operability histograms are discussed in terms of the HgCdTe epitaxial layer defect density and the dislocation density of the individual diode junctions. Individual diode zero bias impedance and reverse bias current-voltage (I-V) characteristics vs. temperature are discussed in terms of the dislocation density of the epitaxial layer, and the misfit stress in the epitaxial multilayer structure, and the thermal expansion mismatch in the composite substrate. The fundamental FPA performance limitations and possible FPA performance improvements are discussed in terms of basic device physics and material properties.

Growth of Hg1−xCdxTe by molecular beam epitaxy (MBE) has been under development since the early 1980s at Rockwell Scientific Company (RSC), formerly the Rockwell Science Center; and we have shown that high-performance and highly reproducible MBE HgCdTe double heterostructure planar p-on-n devices can be produced with high throughput for various single- and multiplecolor infrared applications. In this paper, we present data on Hg1−xCdxTe epitaxial layers grown in a ten-inch production MBE system. For growth of HgCdTe, standard effusion cells containing CdTe and Te were used, in addition to a Hg source. The system is equipped with reflection high energy electron diffraction (RHEED) and spectral ellipsometry in addition to other fully automated electrical and optical monitoring systems. The HgCdTe heterostructures grown in our large ten-inch Riber 49 MBE system have outstanding structural characteristics with etch-pit densities (EPDs) in the low 104 cm−2 range, Hall carrier concentration in low 1014 cm−3, and void density

The use of silicon as a substrate alternative to bulk CdZnTe for epitaxial growth of HgCdTe for infrared (IR) detector applications is attractive because of potential cost savings as a result of the large available sizes and the relatively low cost of silicon substrates. However, the potential benefits of silicon as a substrate have been difficult to realize because of the technical challenges of growing low defect density HgCdTe on silicon where the lattice mismatch is ∼19%. This is especially true for LWIR HgCdTe detectors where the performance can be limited by the high (∼5×106 cm−2) dislocation density typically found in HgCdTe grown on silicon. We have fabricated a series of long wavelength infrared (LWIR) HgCdTe diodes and several LWIR focal plane arrays (FPAs) with HgCdTe grown on silicon substrates using MBE grown CdTe and CdSeTe buffer layers. The detector arrays were fabricated using Rockwell Scientific’s planar diode architecture. The diode and FPA and results at 78 K will be discussed in terms of the high dislocation density (∼5×106 cm2) typically measured when HgCdTe is grown on silicon substrates.

In the past several years, we have made significant progress in the growth of CdTe buffer layers on Si wafers using molecular beam epitaxy (MBE) as well as the growth of HgCdTe onto this substrate as an alternative to the growth of HgCdTe on bulk CdZnTe wafers. These developments have focused primarily on mid-wavelength infrared (MWIR) HgCdTe and have led to successful demonstrations of high-performance 1024×1024 focal plane arrays (FPAs) using Rockwell Scientific’s double-layer planar heterostructure (DLPH) architecture. We are currently attempting to extend the HgCdTe-on-Si technology to the long wavelength infrared (LWIR) and very long wavelength infrared (VLWIR) regimes. This is made difficult because the large lattice-parameter mismatch between Si and CdTe/HgCdTe results in a high density of threading dislocations (typically, >5E6 cm−2), and these dislocations act as conductive pathways for tunneling currents that reduce the RoA and increase the dark current of the diodes. To assess the current state of the LWIR art, we fabricated a set of test diodes from LWIR HgCdTe grown on Si. Silicon wafers with either CdTe or CdSeTe buffer layers were used. Test results at both 78 K and 40 K are presented and discussed in terms of threading dislocation density. Diode characteristics are compared with LWIR HgCdTe grown on bulk CdZnTe.

We report on Hg1−xCdxTe mid-wavelength infrared (MWIR) detectors grown by molecular-beam epitaxy (MBE) on CdZnTe substrates. Current-voltage (I-V) characteristics of HgCdTe-MWIR devices and temperature dependence of focal-plane array (FPA) dark current have been investigated and compared with the most recent InSb published data. These MWIR p-on-n Hg1−xCdxTe/CdZnTe heterostructure detectors give outstanding performance, and at 68 K, they are limited by diffusion currents. For temperatures lower than 68 K, in the near small-bias region, another current is dominant. This current has lower sensitivity to temperature and most likely is of tunneling origin. High-performance MWIR devices and arrays were fabricated with median RoA values of 3.96 × 1010 Ω-cm2 at 78 K and 1.27 × 1012 Ω-cm2 at 60 K; the quantum efficiency (QE) without an antireflection (AR) coating was 73% for a cutoff wavelength of 5.3 µm at 78 K. The QE measurement was performed with a narrow pass filter centered at 3.5 µm. Many large-format MWIR 1024 × 1024 FPAs were fabricated and tested as a function of temperature to confirm the ultra-low dark currents observed in individual devices. For these MWIR FPAs, dark current as low as 0.01 e−/pixel/sec at 58 K for 18 × 18 µm pixels was measured. The 1024 × 1024 array operability and AR-coated QE at 78 K were 99.48% and 88.3%, respectively. A comparison of these results with the state-of-the-art InSb-detector data suggests MWIR-HgCdTe devices have significantly higher performance in the 30–120 K temperature range. The InSb detectors are dominated by generation-recombination (G-R) currents in the 60–120 K temperature range because of a defect center in the energy gap, whereas MWIR-HgCdTe detectors do not exhibit G-R-type currents in this temperature range and are limited by diffusion currents.

Excess low frequency noise is investigated for the first time in infrared MBE grown LWIR Hg1−xCdxTe double layer planar heterostructure (DLPH) detectors grown on lattice matched substrates. LWIR detectors having R0Aopt values at 40K in the 101–107 Θ-cm2 range have been characterized as a function of temperature between 120 and 20K. Detectors with R0Aopt≥103Θ-cm2 at 40K have theoretical diffusion limited performance down to 78K and detectors with R0Aopt ≥105 Θ-cm2 at 40K are within a factor of two of theoretical diffusion limited performance for T>65K. Activation energies extracted from noise (Vd=−100 mV) and dark current (Vd=−100 mV) vs temperature measurements were detector dependent. The activation energy for detectors with R0Aopt≈106 Θ-cm2 at 40K is ∼0.90*Eg to 0.99*Eg. The noise measured between 78 and 105K in the intermediate performance (R0Aopt∼103–104 Θ-cm2 at 40K) detectors are higher than the noise measured in the higher performance (R0Aopt∼105–107 Θ-cm2) detectors. In addition, the excess low frequency noise and the dark current at −100 mV in the intermediate and poor (R0Aopt∼101 Θ-cm2) performance detectors are temperature independent. For each detector measured, the activation energy extracted from noise (Vd=−100 mV) vs temperature measurements is equal to the activation energy extracted from the total dark current (Vd=−100 mV) vs temperature measurements. For different dark current mechanisms, the excess low frequency noise varies with temperature and also with area within statistical accuracy in the same manner as the total dark current through the detector. At 78K, the Tobin14 expression holds in the general sense for equal area detectors dominated by different current mechanisms and also for detectors with a wide range of implant dimensions (Aimp=3.85×10−7 cm2 to Aimp=6.25×10−4 cm2). Following measurements, the detectors were stripped of the passivation and overlaying metal layers and dressed by a defect etch to reveal defects in each detector. A correlation among noise, leakage current and defect type has been determined for each detector.

The current-voltage characteristics and quantum efficiencies of double layer planar heterostructure photodiodes were investigated. Results are reported on devices with cutoff wavelengths of 1.8, 2.4, and 3.3 µm. For these respective devices, the dominant currents for temperatures >250,>200,>150K are diffusion currents limited by shallow Shockley-Hall-Read (SHR) processes. The remarkable result is that the electrical and optoelectronic properties of these devices of diverse cut-off wavelength can be explained by simple models using independently measured layer parameters such as the minority carrier lifetimes. For all three cases, the analysis suggests that the same shallow (SHR) centers located at 78% of the energy gap are causing the observed effects. These traps located in then-type base of the device are not influenced by the magnitude of n-type doping and this observation was used to significantly improve the performance of the devices and validate the predictive capability of the models used in the analysis. The shallow centers appear to be process induced rather than grown-in. This assertion is based on the observation that changes in the annealing process led to an order of magnitude improvement in the minority carrier lifetime.

The capability of growing state-of-the-art middle wavelength infrared (MWIR)-HgCdTe layers by molecular beam epitaxy (MBE) on large area silicon substrates has been demonstrated. We have obtained excellent compositional uniformity with standard deviation of 0.001 with mean composition of 0.321 across 1.5″ radii. R0A as high as 5 × 107 ω-cm2 with a mean value of 7 × 106 Θ-cm2 was measured for cut-off wavelength of 4.8 µm at 77K. Devices exhibit diffusion limited performance for temperatures above 95K. Quantum efficiencies up to 63% were observed (with no anti-reflection coating) for cut-off wavelength (4.8–5.4) µm @ 77K. Excellent performance of the fabricated photodiodes on MBE HgCdTe/CdTe/Si reflects on the overall quality of the grown material in the MWIR region.

Results are presented for minority carrier lifetime in n-type molecular beam epitaxy Hg1−xCdxTe with x ranging from 0.2 to 0.6. It was found that the lifetime was unintentionally degraded by post-growth annealing under Hg saturated conditions in a H2 atmosphere that was both time and temperature dependent. This effect was minimal or non-existent for x∼0.2 material, but very strong for x ≥ 0.3. Hydrogen was identified as responsible for this degradation. Identical annealing in a He atmosphere avoids this degradation and results in neartheoretical lifetime values for carrier concentrations as low 1 × 1015 cm−3 in ≥0.3 material. Modeling was carried out for x∼0.2 and x∼0.4 material that shows the extent to which lifetime is reduced by Shockley-Real-Hall recombination for carrier concentrations below 1 × 1015 cm−3, as well as for layers annealed in H2. It appears that annealing in H2 results in a deep recombination center in wider bandgap HgCdTe that lowers the lifetime without affecting the majority carrier concentration and mobility.

The molecular beam epitaxy (MBE) growth technology is inherently flexible in its ability to change the Hg1−xCdxTe material’s bandgap within a growth run and from growth run to growth run. This bandgap engineering flexibility permits tailoring the device architecture to the various specific requirements. Material with active layer x values ranging from ∼0.198 to 0.570 have been grown and processed into detectors. This wide range in x values is perfectly suited for remote sensing applications, specifically the National Polar Orbiting Environmental Satellite System (NPOESS) program that requires imaging in a multitude of infrared spectral bands, ranging from the 1.58 to 1.64 µm VSWIR (very short wave infrared) band to the 11.5 to 12.5 µm LWIR (longwave infrared) band and beyond. These diverse spectral bands require high performance detectors, operating at two temperatures; detectors for the VSWIR band operate near room temperature while the SWIR, MWIR (mid wave infra red), LWIR and VLWIR (very long wave infrared) detectors operate near 100K, because of constraints imposed by the cooler for the NPOESS program. This paper uses material parameters to calculate theoretical detector performance for a range of x values. This theoretical detector performance is compared with median measured detector optical and electrical data. Measured detector optical and electrical data, combined with noise model estimates of ROIC performance are used to calculate signal to noise ratio (SNR), for each spectral band. The SNR are compared with respect to the meteorological NPOESS system derived focal plane. The derived system focal plane requirements for NPOESS are met in all the spectral bands.

Short wave infrared (SWIR) devices have been fabricated using Rockwell’s double layer planar heterostructure (DLPH) architecture with arsenic-ion implanted junctions. Molecular beam epitaxially grown HgCdTe/CdZnTe multilayer structures allowed the thin, tailored device geometries (typical active layer thickness was ∼3.5 µm and cap layer thickness was ∼0.4 µm) to be grown. A planar-mesa geometry that preserved the passivation advantages of the DLPH structure with enhanced optical collection improved the performance. Test detectors showed Band 7 detectors performing near the radiative limit (∼3-5X below theory). Band 5 detector performance was ∼4-50X lower than radiative limited performance, apparently due to Shockley-Hall-Read recombination. We have fabricated SWIR HgCdTe 256 × 12 × 2 arrays of 45 um × 45 µm detector on 45 µm × 60 µm centers and with cutoff wavelength which allows coverage of the Landsat Band 5 (1.5−1.75 µm) and Landsat Band 7 (2.08−2.35 µm) spectral regions. The hybridizable arrays have four subarrays, each having a different detector architecture. One of the Band 7 hybrids has demonstrated performance approaching the radiative theoretical limit for temperatures from 250 to 295K, consistent with test results. D* performance at 250K of the best subarray was high, with an operability of ∼99% at 1012 cm Hz1/2/W at a few mV bias. We have observed 1/f noise below 8E-17 AHz 1/2 at 1 Hz. Also for Band 7 test structures, Ge thin film diffractive microlenses fabricated directly on the back side of the CdZnTe substrate showed the ability to increase the effective collection area of small (nominally

This paper describes recent advances in MBE HgCdTe technology. A new 3 inch production molecular beam epitaxy (MBE) system, Riber Model 32P, was installed at Rockwell in 1994. The growth technology developed over the years at Rockwell using the Riber 2300 R&D system was transferred to the 32P system in less than six months. This short period of technology transfer attests to our understanding of the MBE HgCdTe growth dynamics and the key growth parameters. Device quality material is being grown routinely in this new system. Further advances have been made to achieve better growth control. One of the biggest challenges in the growth of MBE HgCdTe is the day-to-day control of the substrate surface temperature at nucleation and during growth. This paper describes techniques that have led to growth temperature reproducibility within + - 1°C, and a variation in temperature during substrate rotation within 0.5°C. The rotation of the substrate during growth has improved the uniformity of the grown layers. The measured uniformity data on composition for a typical 3 cm × 3 cm MBE HgCdTe/CdZnTe shows the average and standard deviation values of 0.229 and 0.0006, respectively. Similarly, the average and standard deviation for the layer thickness are 7.5 and 0.06 µm, respectively. P-on-n LWIR test structure photodiodes fabricated using material grown by the new system and using rotation during growth have resulted in high-performance (R0)A, quantum efficiency) devices at 77 and 40K. In addition, 128 × 28 focal plane arrays with excellent performance and operability have been demonstrated.

This work presents characterization of implanted and annealed double layer planar heterostructure HgCdTe for p-on-n photovoltaic devices. Our observation is that compositional redistribution in the structure during implantation/ annealing process differs from that expected from classical composition gradient driven interdiffusion and impacts the placement of the electrical junction with respect to the metallurgical heterointerface, which in turn affects quantum efficiency and RoA. The observed anomalous interdiffusion results in much wider cap layers with reduced composition difference between base and cap layer composition. The compositional redistribution can, however, be controlled by varying the material structure parameters and the implant/anneal conditions. Examples are presented for dose and implanted species variation. A model is proposed based on the fast diffusion in the irradiation induced damage region of the ion implantation. In addition, we demonstrate spatial uniformity obtained on molecular beam epitaxy (MBE) material of the compositional and implanted species profile. This reflects spatial uniformity of the ion implantation/annealing Processes and of the MBE material characteristics.

Characterization of defects in Hg1−xCdxTe compound semiconductor is essential to reduce intrinsic and the growth-induced extended defects which adversely affect the performance of devices fabricated in this material system. It is shown here that particulates at the substrate surface act as sites where void defects nucleate during Hg1−xCdxTe epitaxial growth by molecular beam epitaxy. In this study, we have investigated the effect of substrate surface preparation on formation of void defects and established a one-to-one correlation. A wafer cleaning procedure was developed to reduce the density of such defects to values below 200 cm−2. Focal plane arrays fabricated on low void density materials grown using this new substrate etching and cleaning procedure were found to have pixel operability above 98.0%.

We have carried out a study and identified that MBE HgCdTe growth-induced void defects are detrimental to long wavelength infrared photodiode performance. These defects were induced during nucleation by having surface growth conditions deficient in Hg. Precise control and reproducibility of the CdZnTe surface temperature and beam fluxes are required to minimize such defects. Device quality material with void defect concentration values in the low 102 cm2 range were demonstrated.