Emulator-based Bayesian Inference on Non-Proportional Scintillation Models by Compton-Edge Probing

TL;DR

This paper introduces a Bayesian inference framework using emulator-based machine learning to accurately model non-proportional scintillation responses in inorganic scintillators, validated with gamma-ray measurements and Monte Carlo simulations.

Contribution

It presents a novel methodology combining Compton edge probing, Bayesian inference, and machine learning emulators to infer non-proportional scintillation models without extra electron response data.

Findings

Successfully constrains the NPSM and quantifies intrinsic resolution.

Predicts spectral Compton edge dynamics based on scintillation mechanisms.

Provides a general framework applicable to various inorganic scintillators.

Abstract

Scintillator detector response modelling has become an essential tool in various research fields such as particle and nuclear physics, astronomy or geophysics. Yet, due to the system complexity and the requirement for accurate electron response measurements, model inference and calibration remains a challenge. Here, we propose Compton edge probing to perform non-proportional scintillation model (NPSM) inference for inorganic scintillators. We use laboratory-based gamma-ray radiation measurements with a NaI(Tl) scintillator to perform Bayesian inference on a NPSM. Further, we apply machine learning to emulate the detector response obtained by Monte Carlo simulations. We show that the proposed methodology successfully constrains the NPSM and hereby quantifies the intrinsic resolution. Moreover, using the trained emulators, we can predict the spectral Compton edge dynamics as a function of the parameterized scintillation mechanisms. The presented framework offers a novel way to infer NPSMs for any inorganic scintillator without the need for additional electron response measurements.