Neutron Lifetime Measurements: Machine Learning Methods in Performing Ultracold Neutron Coincidence Analysis

As part of the UCNtau project, which is dedicated to the ongoing investigation of neutron lifetime, our work focuses on refining these measurements. The precision of neutron properties plays a pivotal role in our understanding of the Standard Model, in building more advanced physics structures, and in exploring the early universe. We devised a setup to detect ultracold neutrons (UCNs) through their scintillation effects and the resultant light pulses, which were captured by detectors. The subsequent analysis was conducted by implementing a time window with specific algorithms acting as thresholds. These algorithms are based on pulse detection by two photomultipliers operating in tandem, a process we refer to as coincidence analysis. This analytical method treats each neutron within a given time interval as a distinct unit for measuring lifetime, effectively eliminating background noise. This approach necessitates the identification of each UCN during the interval. We have obtained unadulterated data that preclude pileup effects and have generated a Monte-Carlo simulation based on these datasets. The simulation can output labeled pulse patterns and neutron counts. We utilized machine learning methods to train these patterns in various classification models and ascertain the accuracy of the counts corresponding to the patterns. Moreover, we attempted to develop several regression models to calculate the number of photons and the time difference that gave rise to these UCN patterns.