The light detection system of the ICARUS detector in the Short Baseline Neutrino program at Fermilab. (28th February 2023)

• Record Type: Journal Article
• Authors: Babicz, M.
• Is Part Of: Journal of instrumentation. Volume 18:Number 2(2023)
• Publication Date: 2023-02-28
• Subjects: Large detector systems for particle and astroparticle physics -- Neutrino detectors -- Particle identification methods -- Time projection chambers; Scientific apparatus and instruments -- Periodicals; 502.84
• DOI: 10.1088/1748-0221/18/02/C02056
• Languages: English
• ISSNs: 1748-0221

Abstract

Abstract: The Short Baseline Neutrino (SBN) program at Fermilab is an extensive experimental programme aiming at searching for sterile neutrino(s) [1], whose existence is one of the fundamental open questions of neutrino physics. It employs three Liquid Argon Time Projection Chamber (LArTPC) detectors, called ICARUS, MicroBooNE and SBND, sampling the Booster Neutrino Beam (BNB) at different locations from its target. The SBN detectors, working near the Earth's surface, are subjected to a substantial cosmic background, which can mimic genuine neutrino interactions. Thus, it is essential to distinguish the signals related to the neutrino beams from those induced by the cosmic rays. The light detection system is a vital part of LArTPCs, but its role is even more critical for surface-operating detectors like ICARUS. The ICARUS light detection system is based on 360 Photomultiplier Tubes (PMTs). Its main role is to provide an efficient trigger and contribute to the 3D reconstruction of detected events. The light detection system calibration and further trigger system improvements were performed for the final detector configuration during the detector commissioning at Fermilab. The trigger system effectively exploits the information given by the PMTs. To further increase that system's efficiency in event filtering, an alternative method based on Convolutional Neural Network (CNN) has been developed. Simulation-based results show that this technique can reduce the cosmic background by up to 76% with a neutrino selection efficiency of 99%. However, to filter the real data cases, which are usually not identical to the simulated ones, this method was improved by applying Domain Adversarial Neural Network (DANN). The results prove that adversarial training through a DANN can alleviate the simulation bias, demonstrating the first successful application of DANN for CNN as an event classifier for a LArTPC.