nγ separation using deep learning method

We are measuring photo nuclear decomposition reaction using Laser Compton scattering gamma ray source at NewSUBARU. The purpose of this research is to develop a measuring device for separating neutrons and γ rays with higher precision in order to improve the measurement accuracy. Organic scintillators are often used to measure neutrons of several MeV to several hundred MeV. The main background is a signal caused by Compton scattering of γ rays incident on the scintillator. The time constant of the light emission time of the scintillation light by the protons recoiled by the neutron and the scintillation light by the electrons recoiled by the γ rays is different.When using the data of the whole waveform measured by FADC, information volume is large and it is difficult to be affected by noise. We applied deep learning as a method to use FADC data for particle identification. In order to improve the accuracy of separation by deep learning, it is to use higher purity data as learning data. If noise is mixed in the learning data, accurate learning cannot be used and expected results may not be obtained. Therefore, when neutron data was measured, shielding was carefully considered, and high purity data was obtained.