The Jiangmen Underground Neutrino Observatory (JUNO) is a 20 kton multipurpose liquid-scintillator detector whose main goal is the determination of the neutrino mass ordering using the measurement of the vacuum dominated oscillation pattern of reactor anti-neutrinos from eight reactor cores. The sensitivity of JUNO to the neutrino mass ordering can be enhanced via a combined analysis of reactor anti-neutrinos with atmospheric neutrinos, in which the matter-dominated oscillation depends on the mass ordering. Such an analysis requires a precise reconstruction of the energy and the direction of atmospheric neutrinos. As the largest liquid-scintillator detector
ever built, JUNO will also be able to measure the atmospheric neutrino flux down to lower energies than the current large Cherenkov detectors.
This poster presents the reconstruction of the energy of atmospheric neutrinos with a machine learning approach and the direction reconstruction with a novel approach. While the machine learning approach relies on the geometrical representation of the detector with a Graph Convolutional Neural Network, the latter focuses on the reconstruction of the photon emission topology in the JUNO detector. The results presented are based on Monte-Carlo simulations, including for the first time the full electronics response, calibration and waveform reconstruction.