Convolutional Neural Networks to Classify Single and Double Electron Events for the Majorana Demonstrator

In the search for neutrinoless double beta decay (0ν2β), single electron events in the Ge-76 detectors of the MAJORANA DEMONSTRATOR and future LEGEND project degrade our ability to measure the decay rate of 0ν2β decay. If we can find a way to filter out single electron events from our dataset, we can remove a significant portion of the background noise in our system. We demonstrate that through the use of convolutional neural networks (CNNs), simulated single and double electron events in P-Type point contact (PPC) Ge detectors can be classified with an area under the curve (AUC) of 0.822, meaning 82.2% of the time the CNN can differentiate between a single and double electron event based solely on the normalized simulated waveform.