We used Gaussian separation and receiver operating characteristic (ROC) curves to optimize the neutron sensitivity and gamma rejection of an ultra-thin ⁶LiF:ZnS(Ag)-scintillator-based neutron detector paired with a silicon photomultiplier (SiPM). We recorded the waveforms while operating the detector in a monochromatic cold neutron beam and in the presence of isotopic ¹³⁷Cs and ⁶⁰Co gamma sources. We used a two-window charge comparison (CC) pulse-shape discrimination (PSD) technique to distinguish the neutron capture events from other types of signals. By feeding the recorded waveforms through variants of this algorithm, it was possible to optimize the duration of the integration windows [(0-100 ns) for the prompt window and (100-2300 ns)] for the delayed window. We then computed the detector's ROC curve from waveform recordings and compared that with the experimental performance. We also used this procedure to compare a series of detector configurations to select the optimal bias voltage for the SiPM photosensor.

Keywords: Chromatic analysis neutron diffractometer or reflectometer (CANDOR); LiF:ZnS(Ag); charge comparison (CC); neutron detector; pattern recognition; pulse shape; radiation detector; receiver operating characteristic (ROC) curve; silicon photomultiplier (SiPM); support vector machine (SVM).