Permanent-Magnet Optimization for Stellarators as Sparse Regression

Abstract

Though permanent magnets are ubiquitous in science and everyday society, we still lack a systematic analysis of how to optimally place and orient a large set of them in a (much larger) set of possible locations. This study reformulates the problem in terms of sparse regression, and offers an algorithm that can effectively solve the problem for nonconvex systems with over 10${}^{6}$ optimizable variables and constraints. The authors then obtain high-performance designs for a class of familiar fusion-power devices called stellarators. Because their algorithm addresses problems that appear across many scientific domains, it should prove extremely impactful.