Author:
Yoo Changhyun,Goh Junghwan
Abstract
The JSNS2(J-PARC Sterile Neutrino Search at the J-PARC Spallation Neutron Source) experiment searches for neutrino oscillations at 24m baseline with the J-PARC’s 3 GeV 1 MW proton beam incident on a mercury target at the Materials and Life science experimental Facility (MLF). The JSNS2detector consists of three cylindrical layers, an innermost neutrino target, an intermediate gamma-catcher, and an outermost veto. The neutrino target is made of 17 tonnes of Gd-loaded LS (Gd-LS) stored in an acrylic vessel, 3.2m(D) 2.5m(H). The detector consists of a total of 120 photomultiplier tubes (PMTs), 96 PMTs for inner and 24 PMTs for veto. In JSNS2, a maximum likelihood method based on the PMT charges is used to reconstruct position and energy of the event. We introduce Static Graph Convolution Neural Network (SGCNN), which is a combined model of PointNet and Graph Neural Network (GNN). The model was trained by Monte Carlo (MC) samples, and the position and charge of 96 inner PMTs was used as the training feature.
Reference11 articles.
1. Ajimura et al., arXiv preprint arXiv:1705.08629 (2017)
2. The JSNS2 detector
3. Graph Neural Networks for low-energy event classification & reconstruction in IceCube
4. Vertex and energy reconstruction in JUNO with machine learning methods
5. Seong I., Hertel L., Collado J., Li L., Nayak N., Bian J., Baldi P., Convolutional neural networks for energy and vertex reconstruction in DUNE, in 33rd Conference on Neural Information Processing Systems (NeurIPS), Machine Learning and the Physical Sciences Workshop (2019)