Correlation between Visual Defects and Increased Dark Current in Large-Area Hg^sub 1-x^Cd^sub x^Te Photodiodes

Journal of Electronic Materials, Jun 2005 by D'Souza, A I, Stapelbroek, M G, Willis, R, Masterjohn, S, Et al

The Cross-Track Infrared Sounder (CrIS) program [an instrument on the National Polar-Orbiting Operational Environmental Satellite System (NPOESS)] requires photodiodes with spectral cutoffs denoted by short-wavelength infrared [λ^sub c^(98 K) ∼ 5.1 μm], midwavelength infrared [λ^sub c^(98 K) ∼ 9.1 μm], and long-wavelength infrared (LWIR) [λ^sub c^(81 K) ∼ 15.5 μm]. The CrIS instrument also requires large-area (850-μm-diameter) photodiodes with state-of-art performance. Molecular beam epitaxy (MBE) is used to grow n-type shortwavelength infrared, midwavelength infrared, or LWIR Hg^sub 1-x^Cd^sub x^Te on latticematched CdZnTe. Detectors with p-type implants 7 μm in diameter are used to constitute the 850-μm-diameter lateral collection diodes (LCDs). The photodiode architecture is the double-layer planar heterostructure architecture.

Quantum efficiency, I-V, R^sub d^-V, and 1/f noise in photovoltaic Hg^sub 1-x^Cd^sub x^Te detectors are critical parameters that limit the sensitivity of infrared sounders. These are some of the parameters used to select photodiodes that will be part of the CrIS focal plane module (FPM). During fabrication of the FPM, the photodiodes are subject to a significant amount of handling while transitioning from part of newly processed Hg^sub 1-x^Cd^sub x^Te wafers to individual photodiodes mounted in a CrIS FPM ready to be flown on NPOESS. Quantum efficiency, I-V, noise, and visual inspections are performed at several steps in the detector's journey. Initial I-V and visual inspections are conducted at the wafer level followed by I-V, noise, and quantum efficiency after dicing and mounting the photodiodes in leadless chip carriers (LCCs). A visual inspection is performed following removal of the detectors from the LCCs. Finally, the individual photodiodes are precision mounted on an FPM base, and I-V, noise, quantum efficiency, and visual inspections are performed again. Each step in the FPM fabrication process requires handling and environmental conditioning that can result in detector dark current and noise increase. Some photodiodes on the first flightlike FPMs fabricated exhibited an increase in dark current and noise characteristics at the FPM level as compared to the measurements performed when the photodiodes were in LCCs prior to integration into the FPM. The degradation observed resulted in an investigation to discern the cause of the performance degradation (baking at elevated temperatures, mechanical handling, electrical stress, etc.). This paper outlines the results of the study and the corrective actions that led to the successful manufacture of LWIR large detectors from material growth to insertion into flight FPMs for the CrIS program.

Key words: HgCdTe, photodiodes, visual defects, dark current

INTRODUCTION

The National Polar-Orbiting Operational Environmental Satellite System (NPOESS) contains the Cross-Track Infrared Sounder (CrIS) instrument. CrIS is a Fourier transform infrared (FTIR) sensor used to measure earth radiance at high spectral resolution to derive pressure, temperature, and moisture profiles of the atmosphere. Each CrIS instrument has three focal plane modules (FPMs) that address different spectral bands: short-wavelength infrared [λ^sub c^(98 K) ∼ 5.1 μm], midwavelength infrared [λ^sub c^(98 K) ∼ 9.1 μm], and long-wavelength infrared (LWIR) [λ^sub c^(81 K) ∼ 15.5 μm]. There are nine 850-μm-diameter photodiodes per FPM,1 arranged in a 3 X 3 array, as shown in Fig. 1. The placement of detectors with respect to each other is highly precise with tight tolerances in the X, Y, and Z dimensions. Earlier developmental detectors for the CrIS instrument were 1,000 μm in diameter.

Following the fabrication of a set of engineering FPMs, photodiode dark currents and noise had increased from those that had been measured at the leadless chip carrier (LCC) level for some of the photodiodes, particularly in the LWIR module. The LWIR module is perceived as the greatest challenge for fabrication because of the very long cutoff wavelength at a relatively high temperature of 81 K. In this paper, we will specifically discuss results from studies on the LWIR detectors. Roughly similar, but perhaps not quite as severe, effects were observed for the other spectral band detectors.

An example of degraded photodiode performance is shown in Figs. 2 and 3. Figure 2 shows the I^sub d^-V^sub d^ data and the calculated Ra-Vd for the same photodiode measured in an LCC prior to mounting in an FPM and the postmounting measurement of the photodiode in the FPM. The I^sub d^-V^sub d^ curve from the LCC measurement was diffusion current limited down to 50 mV reverse bias, whereas the I^sub d^-V^sub d^ curve of the same detector mounted in the FPM was limited by a shunt current for reverse bias values greater than ∼30 mV. The magnitude of the current at V^sub d^ = -60 mV (the operating bias) increased by a factor of 4, and the noise measured at the operating bias, displayed in Fig. 3, increased by about an order of magnitude. Because the dynamic impedance at V^sub d^ = -60 mV decreased by almost two orders of magnitude, it is perhaps not surprising that the noise increased dramatically, as shown in Fig. 3. (The roll-on at 300 Hz for the module level noise measurement is due to a high-pass filter in the amplifier.) The observed performance degradation led to an extensive investigation to determine the cause of the changes occurring between LCC and FPM testing. Potential degradation causes examined were (1) contamination from adhesives or cleaning solvents used; (2) ultraviolet light exposure; (3) inadvertent excess bias applied to the photodiodes; (4) thermal stress via baking at elevated temperatures; (5) mechanical stress due to photodiode removal or attachment, dicing, wire bonding; or (6) measurement inaccuracy. Results of some of the experiments leading to determination of degradation cause are presented in the rest of this paper.