Reactor Neutrino Energy Reconstruction with Machine Learning Techniques for the JUNO Experiment

Abstract

The Jiangmen Underground Neutrino Observatory (JUNO) with its satellite Taishan Antineutrino Observatory (TAO) is a next-generation neutrino experiment with a broad physics program. Currently under construction, JUNO is expected to start data-taking in 2024. The central detector of JUNO is an acrylic sphere filled with 20 kt of liquid-scintillator (LS) surrounded by 43212 photomultiplier tubes (PMTs). The primary goals of the experiment are to determine the neutrino mass ordering (NMO) within 3-4𝜎 in 6 years and to measure neutrino oscillation parameters sin 2 𝜃 12 , Δ𝑚 2 21 , Δ𝑚 2 31 with subpercent precision. To achieve the goals, JUNO will study reactor antineutrino emitted from two nuclear power plants located 52.5 km away from the detector. The main requirement for JUNO is a high energy resolution. The detector is constructed to provide an energy resolution of 3% at 1 MeV. In this study, neutrino energy reconstruction with machine learning techniques is presented. The reconstruction techniques are based on aggregated information collected by PMTs. Two models are considered: Boosted Decision Trees and Fully Connected Deep Neural Network. Moreover, the transferability of the approach is shown with an example of JUNO's satellite detector TAO.