Simulating HPGe Detector Signals for Machine Learning Applications (2024)

Undergraduate: Natalie Gray

Faculty Advisor: Julieta Gruszko
Department: Physics and Astronomy

---

The Large Enriched Germanium Experiment for Neutrinoless Double-Beta Decay (LEGEND) collaboration is currently conducting a search for neutrinoless double-beta decay using Germanium-76 detectors. Discovering evidence of such events could profoundly impact our understanding of neutrino masses, lepton number conservation, and the matter/antimatter asymmetry. A new tool is being developed that uses a CycleGAN machine learning algorithm to generate waveforms that accurately mimic real data. The goal of this work is to validate the performance of that tool by testing it with a library comprised of two sets of simulated waveforms: one that uses the idealized simulation output and the other replicating electronic responses and noise. Leveraging the simulation software Siggen, derived from the Majorana Demonstrator source code, we modeled signals resulting from energy depositions at specific locations within the detector. Building upon these simulations, we constructed a model replicating data obtained from the HPGe detectors by scaling the simulations to match energy levels of Full Energy Peaks (FEP) and Double Escape Peaks (DEP), while introducing random pink noise to simulate electronic response and background. These libraries are now being used to directly test the ML algorithm performance, aiming to enhance pulse shape discrimination and analysis.