Multi-objective global optimization for stellarator plasma shape design

Stellarators are gaining popularity as fusion power plant (FPP) candidates due to recent advances in optimizing their three-dimensional plasma shape, enabling them to meet all FPP physics requirements simultaneously [1]. However, finding the best balance between physics and engineering goals remains challenging because of the non-convex, high-dimensional nature of the objective function, making trade-offs hard to analyze. In this work, we present a multi-objective global optimization framework that explores the global landscape in a subregion of interest of the input space. Recognizing that global optimization in the Fourier space commonly used for plasma shape optimization is unsuccessful due to its high-dimensionality and numerous failures of the objective function evaluations, we evolve the degrees of freedom in a lower dimensional space that is smoother and virtually free of evaluation failures. We draw Pareto fronts and highlight trade-off decisions between competing physics and engineering performance metrics, such as proxies for neoclassical and energetic particles confinement, reduced turbulence heat flux, ideal MHD stability, as well as engineering coils feasibility.

[1] C. C. Hegna et al., J. Plasma Phys. 2025