Kinematic Reconstruction in DarkLight@ARIEL: A Machine Learning Approach

DarkLight is an experiment at TRIUMF’s ARIEL facility aimed at searching for light dark sector particles, including the hypothesized 17 MeV X17 boson. The setup uses a 30–50 MeV electron beam incident on a tantalum target to produce electron-positron pairs. These are detected in two arms positioned at asymmetric angles relative to the beam axis, each equipped with Gas Electron Multiplier detectors and trigger scintillators.

This work focuses on reconstructing the kinematics of these particles from GEM detector hits. Using simulated data, we train a machine learning model to predict momentum and angular information at the reaction point. These predictions are then used to reconstruct the invariant mass of the e⁺e⁻ system - a key observable for identifying potential dark sector signals.