Speaker
Description
The EXO-200 experiment, a 100-kg-class liquid xenon time projection chamber, operated from 2011 to 2018, in searching for neutrino-less double beta decay of 136Xe. This is a process whose observation would establish the Majorana nature of neutrinos and help constrain their absolute mass scale. Precise reconstruction of event energy and position is essential for this search. Spatial information is used to define fiducial volumes, suppress external backgrounds, and discriminate between signal-like single-site events and background multi-site interactions.
In this work, we apply and investigate use of deep learning techniques to reconstruct event position of simulated electron events from EXO-200. A deep neural network (DNN) is developed to perform reconstruction directly from electronics waveforms, i.e. raw detector quantities. While previous studies have primarily relied on gamma-ray calibration events, this analysis focuses on electron events, which more closely match the topology of double beta decay signals. The DNN performance is compared to the standard EXO-200 reconstruction methods, showing at least a factor of two improvement in spatial resolution. The results demonstrate the potential of deep learning–based reconstruction for future liquid xenon experiments such as nEXO and motivate further development for application to experimental data.
| Keyword-1 | Deep Learning |
|---|---|
| Keyword-2 | Neutrinoless double beta decay |
| Keyword-3 | Neutrino |