The Jiangmen Underground Neutrino Observatory (JUNO) is a next-generation large (20 kton)
liquid-scintillator neutrino detector, which is designed to determine the neutrino mass ordering
from its precise reactor neutrino spectrum measurement. Moreover, high-energy (GeV-level)
atmospheric neutrino measurements could also improve its sensitivity to mass ordering via
matter effects on oscillations, which depend on the capability to identify electron
(anti-)neutrinos and muon (anti-)neutrinos against each other and against neutral current
background, as well as to identify neutrinos against antineutrinos. However, this flavor
identification task has never been attempted in large homogeneous liquid scintillator detectors
like JUNO.
This poster presents a machine-learning approach for the flavor identification of atmospheric
neutrinos in JUNO. In this method, several features relevant to event topology are extracted
from PMT waveforms and used as inputs to machine learning models. Moreover, the features
from captured neutrons provide additional capability of neutrinos versus anti-neutrinos
identification. Two independent strategies are developed to utilize the primary interaction and
neutron-capture information with different machine-learning models. Preliminary results based
on Monte Carlo simulations show promising potential for this approach.