Searching for Double-Beta Decay to Excited States of Barium in KamLAND-Zen 800 with KamNet

KamLAND-Zen 800 has set the world-leading limit on the Majorana neutrino mass with a search for neutrinoless double beta decay (0νββ) of Xe-136. This limit is an important stepping stone before the next generation of 0νββ experiments. However, competing nuclear theories predict vastly different rates for neutrinoless double beta decay, thereby adding large systematic uncertainties in our Majorana neutrino mass limits. Standard two-neutrino double-beta decay (2νββ) to excited states of the daughter nucleus (Barium-136 for Xe-136) can inform nuclear models. In this work, we present efforts to search for these double-beta decays to excited states of Ba-136. We show how a spherical convolutional deep learning model, KamNet, is improving our sensitivity to this rare process by rejecting backgrounds, as it did for KamLAND-Zen 800's 0νββ analysis. In particular, we discuss how insights gained from machine learning interpretability studies help us apply KamNet more effectively to this specific analysis.