Divertor design through adjoint approaches and efficient code simulation strategies

Abstract

Divertor power exhaust is one of the main issues to be addressed on the way to reliable fusion power plants. It is generally accepted that at least partially detached operation is essential to mitigate the heat and particle loads to the divertor target plates. These highly dissipative regimes pose serious challenges to the plasma edge codes used for the design of these components, leading to exacerbated runtimes and convergence problems due to the presence of statistical noise in the code system. Moreover, the large number of design variables precludes investigating a wide range of designs and operating points. Building on a framework of code speed-up based on the minimization of numerical errors in coupled finite-volume Monte Carlo codes, we develop an efficient discrete adjoint technique for the automated solution of divertor design problems. We are able to compute the sensitivity information with very low variance, despite the statistical noise. By integrating the sensitivity analysis into a one-shot optimization algorithm, robust convergence of the optimization problem is achieved at a cost independent of the number of design variables. The results are encouraging for the development of adjoint-based optimization techniques for plasma edge codes such as SOLPS-ITER.