A multi-purpose reconstruction method based on machine learning for atmospheric neutrinos at JUNO

Abstract

The Jiangmen Underground Neutrino Observatory (JUNO) experiment is designed to measure the neutrino mass ordering (NMO) using a 20-kton liquid scintillator (LS) detector. Besides the precise measurement of the reactor neutrino’s oscillation spectrum, an atmospheric neutrino oscillation measurement in JUNO offers independent sensitivity for NMO, which can potentially increase JUNO’s total sensitivity in a joint analysis. In this contribution, we present a novel multi-purpose reconstruction method for atmospheric neutrinos in JUNO at few-GeV based on a machine learning technique. This method extracts features related to event topology from PMT waveforms and uses them as inputs to machine learning models. A preliminary study based on the JUNO simulation shows good performances for event directionality reconstruction and neutrino flavor identification. This method also has a great application potential for similar LS detectors.