Publicaciones

CIEMAT Publications

Back to Search

Cover image

ISSN: 2695-8864 | NIPO: 152-26-008-6

PUBLISHED IN 2026

Informes técnicos CIEMAT

Neural network-based reconstruction of neutrino interaction coordinates using the SBND photon detection system

Crespo Anadón, J.I.; Aceituno García, A; Domínguez Vidales, S.; Álvarez Garrote, R.

Free

Abstract

This report presents a machine learning-based approach using only scintillation light signals from the SBND Photon Detection System to reconstruct key neutrino interaction parameters. To reconstruct the neutrino interaction time, optical hit sequences are processed with LSTM and Transformer architectures, achieving nanosecond-level resolution (~2-3 ns). Using reduced ResNet-18 models to process 2D photomultiplier tube (PMT) light patterns, the position model achieves a 5-7 cm resolution for the energy barycenter, while the energy-deposited direction achieves an angular error of 15.3° in the YZ plane. The methods described in this report could be used to enhance the reconstruction of neutrino interaction in the SBND