Neutrino Physics at a Research Reactor: Backgrounds and Analysis

The field of neutrino physics has a rich history and is currently the subject of much active research. The discovery of neutrino oscillations led to the conclusion that neutrinos have mass which was in contradiction to the Standard Model. Now researchers are investigating a number of open questions regarding neutrino properties such as their mass values or the existence of CP violation in the weak interaction. In order to answer these questions experimental and analytical techniques of neutrino detection are becoming more advanced, entering into an era of precision neutrino detection.

Nuclear reactors as a source of antineutrinos have played a historically important role in the field, and provide important knowledge at the interface of particle and nuclear physics. Recent and future planned experiments at reactors must deal with intense sources of gamma and neutron radiation while also discriminating neutrino events from cosmic rays. For this reason it is important to have a detailed understanding of the background radiation at the experiment site. This allows for effective shielding design for background mitigation and aids in understanding the requirements of analysis methods for background rejection. To that end, this dissertation provides the first high resolution spatial characterization of the gamma radiation field at the neutrino experiment hall at the High Flux Isotope Reactor (HFIR) in Oak Ridge, TN.

In the second part of the dissertation I discuss the \pspt~experiment which measured the antineutrino flux at HFIR in 2018. This was the first high precision measurement of the \el{U}{235} antineutrino energy spectrum. Sophisticated analysis techniques to select the antineutrino events were employed based on a detailed understanding of the detector response and the cosmic and reactor backgrounds present in the detector. Nowadays computational resources have grown to the point where it is feasible to utilize computationally expensive machine learning methods for such analysis tasks. This work includes an in depth comparison between classical analysis techniques and novel applications of machine learning, the latter of which are shown to improve the effective statistics of the experiment.