Description
Using a high-intensity beam and the imaging capabilities of the Liquid Argon Time-Projection Chamber technology, the Deep Underground Neutrino Experiment (DUNE) will perform high-precision measurements of the parameters governing neutrino oscillations. The reduction of background events in nue appearance analyses is of primary importance, with a major contribution coming from pi0 decays to photon showers that may be misidentified as electrons. This poster discusses ongoing work to optimise electromagnetic shower reconstruction in neutrino interactions at the DUNE Far Detector, using Pandora's multi-algorithm approach to pattern recognition. The reconstruction of complex neutrino interaction topologies with multiple overlapping showers is targeted, and the pi0 invariant mass reconstruction is used as a sensitive performance metric that bridges the gap between reconstruction and physics analysis. First, the impact of a novel, deep learning-based approach to reduce contamination between reconstructed tracks and showers is trialled. Issues involving the merging of neighbouring showers are then investigated, and techniques to address these issues utilising calorimetric shower profiles are presented.
| Collaboration | DUNE |
|---|
