Event-selection technique for the multi-layer Si−CdTe Compton camera onboard Hitomi

The soft gamma-ray detector (SGD) onboard Hitomi, which has a unique design concept, based on the combination of a ”narrow-field multi-layer semi-conductor Compton camera”, and an active shielding, realizes astronomical observations in the 60−600 keV energy band with a high sensitivity. Development of optimum event selection criteria is essential for deriving the best observational performance of the SGD, but it is challenging because many parameters such as the detected photon energy, the Compton-scattering angle, and distance of each hit, among others, are non-linearly correlated. In this study, we propose a new method for distinguishing the signal from the background in the multi-parameter space utilizing a machine-learning approach. Our preliminary result, which uses both on-ground experimental data with good photon statistics and flight data with real in-orbit background and signal information, suggests that this approach might a good guide for an optimal event selection by the Compton camera.