Reconstruction of Atmospheric Neutrino Events in JUNO using GCNs and Calibration of the OSIRIS Radiopurity Pre-Detector

Abstract

The Jiangmen Neutrino Underground Observatory (JUNO) is a liquid scintillator detector located in Jiangmen, China. With its planned completion in summer 2025, JUNO will be the largest of its kind with a 20 kT liquid scintillator (LS) volume. JUNO’s main physics program is to unveil the neutrino mass ordering (NMO) up to 3 σ within six years of data taking by observing reactor electron anti-neutrino disappearance from associated power plants. To resolve the NMO JUNO features an energy resolution of 3 % at 1 MeV with a double calorimetry system, an optical coverage of > 75 % and a low background level. Firstly, this thesis covers the reconstruction of atmospheric neutrinos events in NMO, which will provide a second independent analysis channel within JUNO. The atmospheric neutrino oscillation probabilities feature distinct patterns in the GeV range, which differ for normal and inverse NMO. With JUNO’s high energy resolution and size, these patterns will be distinguishable. To do so, the atmospheric neutrino events need to be reconstructed regarding energy, incoming direction and initial particle type. In this work Graph Convolutional Networks (GCNs) are used on simulated detector response data. For the energy reconstruction, the quenched deposited energy (qEdep) was reconstructed with dqEdepNMO = (0.01 ± 1.91) % in the NMO region. For particle identification the discrimination of neutral current (NC)/charged current (CC) electron neutrino/electron anti-neutrino/muon neutrino/muon anti-neutrino events is approached. This results in an efficiency on NC events of 68 %. For electron neutrino/electron anti-neutrino 75 %/74 % are classified as the electron flavor with an impurity of 24 %/32 % in the respective anti-particle. The muon flavor is correctly classified for muon neutrino/muon anti-neutrinoin 76 %/74 % with a mixing of 15 %/41 % between neutrino and anti-neutrino. This shows a good identification of NC, against CC electron and muon flavor events, but low discrimination power between ν/ν̄ on the used data. Secondly, the charge and energy calibration of the Online Scintillator Internal Radioactivity Investigation System (OSIRIS) are shown. It is a pre-detector of JUNO, monitoring the scintillator’s radio purity during filling. On the data of the first two calibration campaigns, the charge spectra are analyzed, and the calibration constants defined and monitored over time. The used multi-channel plate (MCP)-Photomultiplier tubes (PMTs) non-linear multiplication is investigated using different methods. For the energy calibration the multi-γ source spectra are fitted, to extract the effective light yield of (232.4 ± 7.9) p.e./MeV.