Date of Award
12-24-2025
Date Published
January 2026
Degree Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Physics
Advisor(s)
Denver Whittington
Keywords
Neutrino physics
Subject Categories
Physical Sciences and Mathematics | Physics
Abstract
With five core-collapse supernova candidates within 1 kiloparsec from Earth, extracting information from the neutrino flux will be important to understanding the explosion mechanism of the star (and its composition). The Deep Underground Neutrino Experiment, (DUNE), will be capable of detecting neutrinos from supernovae, primarily from charged-current interactions due to the electron neutrino flavor. However, considering neutrino oscillations in the presence of matter (MSW effects), the muon neutrinos and tau neutrinos will be undetectable if not for the neutral current channel. About 5% of the neutral current interactions will cause the argon nucleus to excite and release a gamma ray at de-excitation. If this de-excitation gamma Compton scatters, the neutral current interaction can be detected. Detecting and classifying these events can give us an understanding of the total flux of the core-collapse supernova, as this channel is mostly invariant to uncertainties in MSW effects. This dissertation will focus on simulating these interactions as if they were occurring in a DUNE detector module and classifying these interactions among others, like the charged-current interactions and electron scattering interactions. Classifiers tested include convolutional neural networks, gradient-boosted decision trees, and traditional statistical methods. This dissertation serves as a proof-of-concept study for using the gamma-producing neutral current channel to study supernova-neutrino bursts. We estimate 1,129 gamma-producing neutral current events can be detectable in a 10 kiloton liquid argon detector, as a result of a 20 solar-mass progenitor, 1 kpc away. This analysis suggests that DUNE's existing 2x2 Demonstrator's threshold is sufficient to detect these events and classify them (to some extent) with out-of-the-box classifiers. The best classifier in this study offers a signal-to-background rate of one-to-ten, which improves the problem, and finally, we present suggestions for future improvement.
Access
Open Access
Recommended Citation
Thomas, Sierra, "Machine-Learning Classification of Gamma-Producing Neutral Current Interactions in Liquid Argon from a Supernova Neutrino Burst" (2025). Dissertations - ALL. 2249.
https://surface.syr.edu/etd/2249
