Speaker
Description
Detection of trace concentrations of tritium using fieldable detectors remains a major need for environmental monitoring and nonproliferation applications, made difficult due to the low energy and short range of the tritium beta ray emission, less than 20 keV. High performance charge-coupled devices (CCDs) developed for astronomy and basic science applications are an attractive option for tritium detection, as they feature extremely low noise, trigger-less readout, fine pixelation, and can be produced with thin entrance windows suitable for few keV beta rays. CCDs provide rich information on each observed particle, ideal for exploiting with machine learning analysis techniques. We explore the sensitivity of CCD-based tritium detection strategies, in particular focusing on the enhancement attainable using advanced machine learning algorithms. Tritium detection in CCDs and the results of machine learning analysis on experimental CCD data are presented. Future perspectives on tritium detection are discussed.
