Stability-based stellarator optimization with derivatives

The field of stellarator optimization has seen resounding success optimizing for critically important quantities like neoclassical transport and fast particle confinement, making the optimized stellarator a viable magnetic confinement fusion concept.

One of the most important remaining optimization challenges deals with stability on both the macro- and microscale, such as MHD ballooning instabilities and kinetic drift waves.

While different reduced models and metrics for stability have been proposed, the availability of computing power makes it now possible to assess stability metric by directly computing the solution to plasma stability eigenproblems.

These solutions alone do not immediately inform optimization efforts, to fully harness modern optimization techniques one needs to have access to derivative information.

A new framework for performing stellarator optimization has been developed in the Julia programming language, enabling seamless integration to automatic differentiation tools.

Coupled with Julia defined stability metrics, this framework enables end-to-end calculation of eigenvalue and eigenfunction derivatives with respect to magnetic shaping.

Details of the implementation of this framework and a demonstration of robust optimization of stability-based metrics will be presented.