Motivation: Understanding the atomic nucleus, and the reactions that create and destroy nuclei in extreme environments (e.g., nuclear and fusion reactors, thermonuclear weapons, and stellar processes), requires increasingly complex experimental tools and analysis pipelines.
Approach: Algorithms were developed to predict and subtract backgrounds at nuclear physics facilities including ATLAS, FRIB, and SECAR. Bayesian normalization of CMOS images in the recoil separator at FRIB and neural networks for rare event searches in FRIB Decay Station data were used.
Impacts: 20% reduction in experiment time for determination of neutron-induced cross sections and measurements of charge state distributions—results relevant to discovery science, stockpile stewardship, and nuclear energy—corresponding to ~$200k savings in facility operational costs per experiment.
S&T Challenge:
Aligns with “Unifying Physics from Quarks to Cosmos,” “Strengthening Deterrence Through Attribution of Nuclear and Radiological Signatures,” and “Achieving AI-Driven Autonomous Laboratories”