Particle Identification at MeV Energies in JUNO

TL;DR

This paper explores particle identification techniques in JUNO, a large liquid scintillator neutrino detector, demonstrating effective discrimination methods for various particle pairs using neural networks and topological reconstruction, enhancing background control.

Contribution

It introduces advanced particle identification methods tailored for JUNO, comparing neural networks and topological techniques, and evaluates their performance for different particle pairings.

Findings

Excellent discrimination for α/β and p/β with Gatti and neural networks.

Partial separation of e+/e− and e−/γ using neural networks and topology.

Topological reconstruction shows high success in particle discrimination.

Abstract

JUNO is a multi-purpose neutrino experiment currently under construction in Jiangmen, China. It is primary aiming to determine the neutrino mass ordering. Moreover, its 20\,kt target mass makes it an ideal detector to study neutrinos from various sources, including nuclear reactors, the Earth and its atmosphere, the Sun, and even supernovae. Due to the small cross section of neutrino interactions, the event rate of neutrino experiments is limited. In order to maximize the signal-to-noise ratio, it is extremely important to control the background levels. In this paper we discuss the potential of particle identification in JUNO, its underlying principles and possible areas of application in the experiment. While the presented concepts can be transferred to any large liquid scintillator detector, our methods are evaluated specifically for JUNO and the results are mainly driven by its high optical photon yield of 1,200 photo electrons per MeV of deposited energy. In order to investigate the potential of event discrimination, several event pairings are analysed, i.e. $\alpha/\beta$, $p/\beta$, $e^+/e^-$, and $e^-/\gamma$. We compare the discrimination performance of advanced analytical techniques based on neural networks and on the topological event reconstruction keeping the standard Gatti filter as a reference. We use the Monte Carlo samples generated in the physically motivated energy intervals. We study the dependence of our cuts on energy, radial position, PMT time resolution, and dark noise. The results show an excellent performance for $\alpha/\beta$ and $p/\beta$ with the Gatti method and the neural network. Furthermore, $e^+/e^-$ and $e^-/\gamma$ can partly be distinguished by means of neural network and topological reconstruction on a statistical basis. Especially in the latter case, the topological method proved very successful.