Accurately simulating gain and clock-induced charge production in the EMCCD gain register

Abstract

An electron-multiplying charge-coupled device (EMCCD) is capable of precise detections in low-signal environments, able to detect a single photon through electron multiplication. It has many applications, such as faint-target astronomy, quantum optics, molecule tracing, and others, and it will be used for faint companion detection in the Roman Telescope's coronagraph instrument. In an EMCCD, photons hit the pixels, and photo-electrons are created; these are multiplied via impact ionization as they travel through the gain register from one gain stage to the next. A high gain means a high multiplication factor, and this is achieved through a high voltage difference across a gain stage. If the gain is high enough, the chance of clock-induced charge (CIC) production in the gain register increases. The probability distribution function governing the gain process typically used only accounts for charge multiplication if one or more electrons enter the gain register. I discuss my implementation of the simulation of this effect and its customization in emccd_detect, the EMCCD detector simulator used for the Roman Telescope. In addition, the simulator has been updated to use the exact binomial distribution for EM gain instead of the approximate Gamma distribution usually used in the literature, which is only valid for large counts. I also examine some EMCCD data and show through maximum likelihood estimation with CIC_gain_register that the data conform better to the binomial distribution versus the approximate Gamma/Erlang distribution. The use of the modified distribution would in principle improve the fidelity of Roman's testing and lead to better EMCCD calibration and more accurate signal extraction from a frame.