Improvements to nuclear data in service of Intentional Forensics

The Intentional Forensics (IF) venture seeks to identify and perfect tagging technologies for nuclear fuel provenance and tracking. These should highlight materials outside of regulatory control, thereby identifying gaps in safeguards, assisting law enforcement, and serving as a deterrent to future trafficking. A major consideration is the complex interaction of neutrons and isotopes. Many particle transport codes used in this development rely on the data in the Evaluated Nuclear Data Format part B (ENDF/B) library, managed by the National Nuclear Data Center at Brookhaven National Laboratory with stakeholders from government, academia, and industry in the United States and global partners. This work describes the data products provided to the IF venture by the NNDC. The first is a compilation and review of ENDF/B-VIII.0 database. This includes a quick reference and new calculations, and the analysis comprises data quality, resonance evaluation, integral metrics, fission product yields, covariances, and accompanying documentation. The review and results will be summarized here. The second product is a recommendation of proposed remediation for identified deficiencies from global sources, supplemented with a machine learning (ML) effort. ML is being used to predict poorly understood n-capture cross sections using the full slate of nuclear data collected and hosted by the NNDC. In particular, we present a neural network (NN) model with demonstrated improvement on a common mathematical model, the liquid drop model.