Monte Carlo studies on Fully Bayesian Unfolding for a multi-layer silicon beta-ray spectrometer

A compact multi-layer silicon beta-ray spectrometer (SBS) has been developed for beta spectrometry and dosimetry in the beta-gamma mixed fields at Canada Deuterium Uranium (CANDU) nuclear plants. Accurate determination of beta fluence spectra from SBS pulse-height data is challenging due to partial energy deposition within each detector. In this work, we present a simulation-based application of Fully Bayesian Unfolding (FBU) to SBS, leveraging Monte Carlo-derived response matrices under anti-coincidence/coincidence conditions. The FBU approach, implemented in Python using PyMC, provides full posterior distributions and credible intervals for unfolded spectra. Performance was evaluated using Geant4 simulations for beta-only and beta-gamma mixed fields with beta-to-gamma source particle ratios from 1 to 0.01. For ratios ⩾ 0.1, the mean unfolded-to-truth ratio for beta fluence above 0.1 MeV was within 20% of unity. The results demonstrate promising fluence unfolding for the SBS, enabling reliable beta fluence spectrum measurements in the beta-gamma mixed radiation fields while providing valuable insights into the gamma component of the field, even at low count rates and low beta-to-gamma ratios. These results demonstrate the feasibility of FBU for SBS, enabling improved beta-gamma discrimination and uncertainty quantification. Future work will focus on optimization of the algorithm to improve the accuracy and efficiency to achieve real-time unfolding for practical measurements.