Constructing the Pareto Front for the SECAR Recoil Mass Separator

The SEparator for CApture Reactions (SECAR) is a next-generation recoil mass separator being commissioned at the Facility for Rare Isotope Beams (FRIB). SECAR uses 8 dipole magnets, 15 quadrupoles, and 2 Wien filters to make precise, direct measurements of astrophysically relevant nuclear reaction rates. With such a complex system, the code COSY Infinity is used to simulate the ion optics of SECAR for given magnet settings. Here, we use these simulations in conjunction with a multiobjective evolutionary algorithm based on decomposition (MOEA/D), to construct a phase space of the magnet parameters and the corresponding objective space. The objective space consists of several properties of the beam, evaluated at multiple points along the beamline. These include the mass resolution and the energy/angular acceptances of the separator. When evolved, the algorithm finds the full Pareto front of the SECAR optimization problem, giving a selection of non-dominated points which perform comparatively better than all other points in at least one objective value. The user can then select a point from the space that performs suitably across the objectives for their purpose. Evolving this algorithm over a large number of distinct iterations more completely maps the entire phase space.