Erlangen 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 7: Data, AI, Computing, Electronics I

T 7.5: Talk

Monday, March 16, 2026, 17:15–17:30, KH 00.024

ML-based LAr classification in LEGEND-200 — •Jonas Schlegel and Christoph Wiesinger for the LEGEND collaboration — Max-Planck-Institut für Kernphysik, Germany

LEGEND-200 uses the scintillation properties of liquid argon (LAr) to suppress backgrounds in the search for neutrinoless double beta decay of ⁷⁶Ge. The LAr is instrumented with wavelength-shifting fibers coupled to arrays of silicon photomultipliers. The current veto implementation relies on a global threshold and is limited by random coincidences, as it does not exploit spatial information. We implement a machine-learning (ML) based topology-aware LAr veto that combines the photoelectron pattern with the relative angular position of the triggering high-purity germanium detector. The network is trained on samples of true and random coincidences and outputs an event-by-event veto probability. This implementation achieves improved background discrimination and serves as a proof-of-concept for further improvements, including timing information and alternative neural network architectures.

Keywords: LEGEND-200; Neutrinoless Double Beta Decay; Liquid Argon Veto; Machine Learning; Low-Background Techniques