Characterization of Infrared Arrays

Eric Bellm

Mentors: Peter Onaka and Alan Tokunaga

Modern infrared astronomy is made possible by the use of large detector arrays. Most of these arrays are "hybrids" which bond semiconducting pixels to a silicon multiplexer for amplification and readout. Arrays characteristically use pixels which are photodiodes, most simply two oppositely-doped semiconductors in contact. Incoming photons are absorbed, producing  electron-hole pairs which accelerate across the contact potential. The resulting photocurrent is then passed to the readout electronics through the multiplexer. Each photon which is absorbed discharges some of the capacitance of the photodiode, though, so detector behavior is distinctly nonlinear. With sufficient illumination, the detector becomes saturated and does not output additional signal until it is reset.

In order to utilize data obtained with infrared arrays, it is first necessary to characterize the performance of the array. A number of parameters are needed. The quantum efficiency (QE) is the percentage of incident photons which produce electrons in the detector. The gain specifies the number of electrons needed to produce one count, or ADU. Read noise and dark current are both sources of extraneous counts, the former due to properties of the readout electronics, the latter due to thermal effects and leakages. Finally, the nonlinearity of the detector response to illumination requires definition of an effective operational range.

Array readout utilizes some variant of correlated double sampling (CDS). In CDS, the uncertainty in the zeroing of the array due to reset noise sources is removed by reading out a pedestal image immediately after reset. This pedestal is then subtracted from the integrated signal. As discussed by Vacca, Cushing, and Rayner [PASP 116, 352], it is necessary to correct the signal and pedestal levels for nonlinearity separately to achieve best accuracy. However, generally only the subtracted values are available, so an iterative algorithm is required to estimate the pedestal and signal levels.

For my project, I tightened and expanded existing IDL characterization code. In addition to improving its generality, ease of use, and speed, I evaluated  modules that implemented the iterative linearization routine outlined above and generated fake data for testing. This code will facilitate characterization of future arrays as well as data linearization.

Additionally, I sought to understand transient events which appear in SpeX  data at a rate of about one per second. Clusters of hot pixels ranging in size from 1 to 6 pixels appear at random in the frames. The operating hypothesis was that these were cosmic rays, but there was some concern that they might be due to radioactive lens coatings in the SpeX enclosure itself. I found that the observed rate of occurrence was consistent within a factor of 2 with predicted rates for cosmic ray muons (1–2 cm^–2 min^–2) [Eidelman et al., Phys. Lett. B592,1], with the expected variance in rate with altitude. However,  the rays roughly double in rate and size when the detector array is placed in  the SpeX enclosure, suggesting that some of the larger events originate within  the instrument itself.

Finally, I developed a procedure for cycling arrays from room temperature to operating temperature (~30 K) while keeping the rate of temperature change below a specified threshold. Subjecting arrays to temperature gradients of more than about ± 0.3 K/min may result in destructive thermal stresses.