Quantum efficiency and quantum yield of an HgCdTe infrared sensor array

We measure the quantum efficiency and quantum yield of an H2RG HgCdTe near-infrared sensor array. Using a blackbody, narrowband filters, and a pinhole camera to provide a calibrated irradiance on the HgCdTe sensor, we determine a 2 s lower limit of 85% for the quantum efficiency averaged over the array and over three observed wavelengths, 3.38, 3.9, and 4.5 mu m. The peak quantum efficiency occurs near the center of the sensor and is unity within measurement uncertainty (5%). By substituting a calibrated PbSe diode for the H2RG sensor, we verify the methods and equipment at the three wavelengths, 3.38, 3.9, and 4.5 mu m. Accurate measurement of the quantum efficiency requires a well-calibrated system gain, the determination of which is complicated by interpixel capacitance, which correlates noise between adjacent pixels. If unaccounted for, the correlation induced by interpixel capacitance would cause the system gain and hence the quantum efficiency to be overestimated by similar to 20%. We accurately measure the interpixel capacitance using the autocorrelation method of Brown, Schubnell, and Tarle and by a method described in this paper that uses post-readout binning. The two methods yield consistent results, but the binning method is more robust than the autocorrelation method in the presence of electromagnetic interference. The interpixel capacitance we measure for the H2RG-S010 sensor is similar to that of another epoxy-backfilled H2RG sensor tested recently by Brown, Schubnell, and Tarle. Using the same PbSe diode to calibrate the quantum efficiency of the 5.5-mu m-cutoff H2RG sensor from 0.7 mu m to 6.0 mu m, and correcting for the interpixel capacitance, we derive implausible quantum efficiencies greater than unity between 1.4 mu m and 2.4 mu m that we cannot explain. The apparent system gain, measured at many monochromatic wavelengths with the usual variance-versus-mean photontransfer analysis of flat-field images, is a function of wavelength for wavelengths less than 2 mu m, which we interpret as evidence for quantum yield larger than unity, i.e., more than one electron being produced by each photon detected by the sensor. We derive the effects of Fano noise and quantum yield on the photon-transfer curve.