Machine Learning Approaches for CdZnTe/CdTe Radiation Detectors

Room temperature semiconductor radiation detectors (RTSDs) such as CdTe and CdZnTe are vital tools for X-ray and γ-ray applications in medical imaging, homeland security, astrophysics, and others. The imaging algorithms for these systems can be improved by detailed material characterization at the micron level. Characterization of charge transport and material defects for these sensors is vital. Several such RTSDs are typically used in a detector module. Current state-of-art characterization is done either as a whole or at most pixel by pixel in a labor-intensive and time-consuming process requiring sophisticated instruments. To overcome these limitations, we present a novel physics-based machine learning (PBML) model to characterize the RTSD crystals at micron length scales faster than the current methods. These models have been developed for 1D and 3D volumetric discretization of the RTSD. The model is based on physical transport of charges (electrons and holes) with electron and hole trapping centers as in an actual RTSD crystal. Each discretized volume or voxel contains different material properties such as charge drift, trapping, detrapping, and recombination, which are modeled as learnable parameters. The PBML model replicates the behavior of an actual physical RTSD. The high-energy photon (X-ray or γ-ray)-induced electron-hole pairs is input to the model at voxel locations, and the signals at the electrodes at either end of the RTSD and distributed charges (electrons and holes) in the voxels are output of the model. Reduced order models using fewer output data such as using only the charges and/or the signals at the electrodes can characterize the RTSD. Additionally, it is observed that in our model, just specifying the general charge transport defects is sufficient for the model to converge to the defects some of which may be present or absent in the actual RTSD crystal. This PBML method is applicable not only to RTSDs such as CdZnTe but also to semiconductor detectors such as high-purity Ge (HPGe) and others.