Pulse shape simulation and discrimination using machine learning techniques

The backbone of any particle-detection experiment is an efficient signal-background discrimination. Pulse shape discrimination (PSD) is a basic method for this purpose in many nuclear, high-energy and rare-event search experiments where scintillator detectors are used. Conventional techniques exploit the difference between decay times of the pulse of signal and background events or pulse signals caused by different types of radiation quanta to achieve good discrimination. However, such techniques give good results only when the total light emission is sufficient to get a proper pulse profile. This is possible when there is significant energy deposition in the detector. Also, if the signal rate is comparable to the background rate, one can choose to have a tighter cut on the discriminating parameters so as to have a high signal purity. But, rare-event search experiments like neutrino or dark-matter direct search experiments cannot afford tighter cuts, since the signal rate and magnitude is low to begin with, compared to the background. Hence, it becomes imperative to have a method that can deliver a very efficient discrimination. The network architecture used here is a basic one and the results look promising with a discrimination accuracy of about 95%. There is a plan to test few other network architectures and an improved result is expected