Enhanced deep-learning approach to spatiotemporal multi-hit reconstruction with delay-line detectors
Phys. Rev. Applied 25, 034062 – Published 18 March, 2026
DOI: https://doi.org/10.1103/4rqm-y1l9
Abstract
Delay-line detectors (DLDs) are essential instruments for precise spatiotemporal particle detection. Traditional DLD event reconstruction methods struggle with overlapping signals from particles that arrive close together in space and time, limiting their multi-hit capabilities. Here, we report substantial updates to machine learning-based multi-hit reconstruction in DLDs, with a focus on the double-hit events and the first results for triple-hit events. Our enhanced neural-network pipeline includes improved data generation, cross-channel peak detection, and peak matching models. Using a spatial grid measurement, as well as a measurement with a discrete temporal structure, as benchmarks, we demonstrate significant improvements in both the spatial and energy domains. These advancements enable improved temporal and spatially resolved multi-hit detection in particle correlation experiments, with potential applications in attosecond physics, cold target recoil ion momentum spectroscopy, ultracold chemistry, and other fields.
