Stellarator Optimization with DESC

Stellarators must be optimized for both physics and engineering objectives, including good particle confinement and simple construction. The design space for these non-axisymmetric external magnetic fields is very large, which makes this high-dimensional optimization problem computationally challenging. Existing stellarator optimization codes wrap around other equilibrium solvers and rely on either gradient-free algorithms, finite differences, or adjoint methods to search the phase space. Each of these traditional methods has its disadvantages: slow convergence, expensive and inaccurate computations, or labor-intensive coding, respectively. The DESC equilibrium solver utilizes automatic differentiation, and it has been extended into a stellarator optimization code that takes advantage of this capability to quickly and exactly differentiate arbitrary objectives to any order. This presentation showcases the ability to efficiently explore the phase space of stellarator equilibria through the example of fixed-boundary optimization for quasi-symmetry. An overview of the optimization approach is presented along with example results and benchmarks against other codes. It is also demonstrated how the code can easily be extended to accommodate other optimization objectives.