Cosmic Ray Jump Detection for the Roman Wide Field Instrument

Abstract

We investigate algorithms for detecting and correcting for jumps due to cosmic rays in infrared detectors, with emphasis on Roman telescope’s Wide Field Instrument. We use a statistic S based on the excess difference between adjacent resultants (average over a group of reads) normalized to the square root of the expected variance that accommodates the uneven resultants of Roman. We show that it is important to account for the covariance of excess difference with that of the estimated count rate. Due to averaging of reads, the ability to detect jumps is reduced, specially if the jump is in the first or the last resultant. Having the first and last resultants as a single-read resultants improves the ability to detect cosmic rays. The signal due to a jump is split across two resultant differences and this motivates small adjustments to the basic algorithm which improves the jump detection. Bias and false negative rate are investigated using Monte Carlo simulations for a few readout patterns. In order to investigate other scenarios approximate formulas for predicting the bias and the misclassification rate are presented. Using cosmic ray properties based on JWST darks, we show that for high count rates (1000 e−1 s−1) most of the cosmic rays remain undetected, with 65% missed identifications. When averaging over multiple exposures, due to low event rates of cosmic rays, the overall bias in estimated count rate due to undetected cosmic rays is negligible. However, for a single exposure with an undetected cosmic ray, one can have a bias of a few percent. This will manifest as abrupt changes in brightness of targets. Fortunately, the bias ranges between 1 and 5 the measurement uncertainty, and this fact can be used to screen out cosmic rays for scientific applications that seek to detect short duration time domain events.