TListSpectrum: an accessible gamma-spectroscopy toolkit that utilizes machine learning

The TListSpectrum ROOT package provides a gamma-spectroscopy toolkit that uses machine learning to characterize nuclear and radioactive materials. This is accomplished through 2D histogram feature classification using convolutional neural networks. Additionally, TListSpectrum institutes a non-traditional data sort method rather than event building for data pre-processing, which decreases the total time to analysis completion through the utilization of accessible memory, post processing, and PNNL’s high power computing resources. The Super X-ray Gamma Alpha Beta Radio-Isotope EvaLuator (SXGABRIEL) detector, a system comprised of eight HPGe detectors and a low-profile liquid scintillator cell designed to measure liquid radioactive samples, collected 123 GB of list mode data in twenty-four hours that was able to be completely analyzed with the TListSpectrum SXGABRIEL analysis suite in four hours. In addition, data taken with the Germanium Gamma-ray Imaging (PHDs GeGI) portable HPGe spectrometer, which is electronically segmented with sixteen channels on the front side and sixteen orthogonal channels on the backside, has been analyzed through TListSpectrum. Aspects of the TListSpectrum GeGI analysis suite will be displayed, including radial gamma-ray addback and variable geometric data parsing. Future development plans for the TListSpectrum GeGI analysis suite and TListSpectrum as a whole will be presented.