Identification of muon-electron elastic events using Graph Neural Networks for precision measurements

¹ University of Pisa, Italy
² INFN Pisa, Italy
³ Institute of Nuclear Physics, Polish Academy of Sciences, Kraków, Poland
⁴ Cracow University of Technology, Kraków, Poland

^* e-mail: emma.hess@df.unipi.it

Published online: 7 October 2025

Abstract

High Energy Physics experiments require robust particle identification and event classification capabilities, often achievable through machine learning techniques. A Graph Neural Network (GNN) technique is employed, tailored to identifying processes occuring when a muon beam interacts with the atomic electrons of thin, low-Z targets in a series of tracking stations of the MUonE experiment [1], which aims to precisely measure the leading hadronic contribution to the muon magnetic moment anomaly. The application of developed technique has been tested in a case study utilizing simulated data from a reduced geometrical configuration of the MUonE experiment, focusing on µ⁺ e− elastic scattering signal and e⁺ e− pair production events. The proposed GNN classifier achieves a classification accuracy of 97 % in distinguishing signal events from pair-production background, thereby laying the groundwork for an even more precise determination of the leading-order hadronic contribution to the muon’s anomalous magnetic moment.