Employing Machine Learning Techniques to Optimize SECAR Separator Optics for Improved Detection of ⁵⁶Co(p,n)⁵⁶Ni Recoil Products.

The SEparator for CApture Reactions (SECAR) is a recoil mass separator system at the Facility for Rare Isotope Beams (FRIB) at Michigan State University. SECAR is designed to study nuclear astrophysical reactions. To study each reaction, up to 21 electro-magnetic elements of SECAR need to be tuned so the recoil products produced in the reaction at the SECAR target are physically separated from unreacted beam while they are transported to the silicon detector at the end of the system. In order to separate the recoil products from the unreacted beam, desirable solutions should achieve multiple objectives: maximum physical separation between the unreacted beam and recoil products along the SECAR system in at least two locations, optimal transmission of the recoils, and a final recoil beam spot size within the size of the silicon detector. To find solutions that meet all objectives, a Multi Objective Evolutionary Algorithm (MOEA) optimization is applied. This machine learning algorithm uses COSY Infinity, a computer code that performs extensive ion-optical calculations and optimizations. COSY Infinity is employed to explore the recoil product trajectories of different magnet settings. The MOEA algorithm learns from its initial results to find optimal solutions (calculated with COSY) that dominate in each objective. In this work, I will show results from the optimization and analysis of simulated SECAR settings for the ⁵⁶Co(p, n)⁵⁶Ni reaction.