Machine Learning Methods for IBD Analysis with PROSPECT Data

The Precision Reactor OSCillation and SPECTrum experiment (PROSPECT) is a segmented multi-ton scale 6Li doped liquid scintillator detector that was deployed at Oak Ridge's High Flux Isotope Reactor in 2018 to measure the reactor antineutrino spectrum and look for potential short baseline oscillation effects due to beyond standard model interactions. Each segment of the detector was outfitted with photomultiplier tubes (PMTs) on either end for light collection enabling centimeter scale position resolution along the length of the cell. The detection mechanism utilized the inverse beta decay (IBD) of the antineutrino on hydrogen within the scintillator, which is characterized by a prompt burst of light from the positron deposition followed by a well defined signal from neutron capture on 6Li. Throughout the duration of the experiment some PMTs became inoperative and were turned off for the initial published IBD analysis. There has since been an analysis effort to recover data from segments containing a single working PMT for the purposes of improving background rejection. This presentation goes over machine learning efforts utilizing convolutional neural networks and graph neural networks to reconstruct physics variables in single-ended segments such as the position, energy, and particle ID. It will also show the impact this has on the effective statistics for an IBD selection.