Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pulse Shape Simulation and Discrimination using Machine-Learning Techniques

Shubham Dutta, Sayan Ghosh, Satyaki Bhattacharya, Satyajit Saha

TL;DR

The paper tackles the challenge of pulse shape discrimination in scintillator detectors at low light levels, where traditional methods struggle in rare-event searches. It compares two neural-network approaches, a Dense Neural Network (DNN) and a Recurrent Neural Network (RNN), against conventional PSD techniques using GEANT4-based detector simulations and experimental data from CsI(Tl) and BGO. The networks are trained on labeled pulse shapes derived from simulations and validated against real measurements, showing that ML-based PSD offers higher discrimination (larger ROC AUC) at energies down to tens of keV, with the RNN often providing a modest edge over the DNN at the lowest energies. The results imply that ML PSD can improve background rejection in dark matter and other rare-event experiments, motivating further real-world deployment and optimization.

Abstract

An essential metric for the quality of a particle-identification experiment is its statistical power to discriminate between signal and background. Pulse shape discrimination (PSD) is a basic method for this purpose in many nuclear, high-energy and rare-event search experiments where scintillation detectors are used. Conventional techniques exploit the difference between decay-times of the pulses from signal and background events or pulse signals caused by different types of radiation quanta to achieve good discrimination. However, such techniques are efficient only when the total light-emission is sufficient to get a proper pulse profile. This is only possible when adequate amount of energy is deposited from recoil of the electrons or the nuclei of the scintillator materials caused by the incident particle on the detector. But, rare-event search experiments like direct search for dark matter do not always satisfy these conditions. Hence, it becomes imperative to have a method that can deliver a very efficient discrimination in these scenarios. Neural network based machine-learning algorithms have been used for classification problems in many areas of physics especially in high-energy experiments and have given better results compared to conventional techniques. We present the results of our investigations of two network based methods \viz Dense Neural Network and Recurrent Neural Network, for pulse shape discrimination and compare the same with conventional methods.

Pulse Shape Simulation and Discrimination using Machine-Learning Techniques

TL;DR

The paper tackles the challenge of pulse shape discrimination in scintillator detectors at low light levels, where traditional methods struggle in rare-event searches. It compares two neural-network approaches, a Dense Neural Network (DNN) and a Recurrent Neural Network (RNN), against conventional PSD techniques using GEANT4-based detector simulations and experimental data from CsI(Tl) and BGO. The networks are trained on labeled pulse shapes derived from simulations and validated against real measurements, showing that ML-based PSD offers higher discrimination (larger ROC AUC) at energies down to tens of keV, with the RNN often providing a modest edge over the DNN at the lowest energies. The results imply that ML PSD can improve background rejection in dark matter and other rare-event experiments, motivating further real-world deployment and optimization.

Abstract

An essential metric for the quality of a particle-identification experiment is its statistical power to discriminate between signal and background. Pulse shape discrimination (PSD) is a basic method for this purpose in many nuclear, high-energy and rare-event search experiments where scintillation detectors are used. Conventional techniques exploit the difference between decay-times of the pulses from signal and background events or pulse signals caused by different types of radiation quanta to achieve good discrimination. However, such techniques are efficient only when the total light-emission is sufficient to get a proper pulse profile. This is only possible when adequate amount of energy is deposited from recoil of the electrons or the nuclei of the scintillator materials caused by the incident particle on the detector. But, rare-event search experiments like direct search for dark matter do not always satisfy these conditions. Hence, it becomes imperative to have a method that can deliver a very efficient discrimination in these scenarios. Neural network based machine-learning algorithms have been used for classification problems in many areas of physics especially in high-energy experiments and have given better results compared to conventional techniques. We present the results of our investigations of two network based methods \viz Dense Neural Network and Recurrent Neural Network, for pulse shape discrimination and compare the same with conventional methods.
Paper Structure (8 sections, 2 equations, 21 figures, 2 tables)

This paper contains 8 sections, 2 equations, 21 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Machine Learning Methods
  3. Detector Simulation
  4. Experimental and Training Data
  5. Analysis
  6. Conclusion
  7. Variation of Input with Simulation Parameters
  8. Loss Curves

Figures (21)

  • Figure 1: Dense Network for PSD
  • Figure 2: Unified model of reflection in GEANT4 (Adopted from:unifiedModel_ref)
  • Figure 3: Simulation for BGO. a) Pulse shape comparison by taking average of 100 pulses. b) $\gamma$-spectra with $662$ keV $^{137}$Cs source measured using BGO scintillator. Both the plots show good agreement with experimental data.
  • Figure 4: Simulation for CsI(Tl). a) Pulse shape comparison by taking average of 100 pulses. b) $\gamma$-spectra with $662$ keV $^{137}$Cs source measured using CsI(Tl) scintillator. The experimentally obtained spectra is broader than the simulated one. This can be attributed to the presence of electronic noise in the experimental data.
  • Figure 5: Score distributions of ML methods for 20 keV recoil energy. a) BDT. b) DNN. c) RNN.
  • ...and 16 more figures