Divertor power exhaust is one of the remaining challenges in the design of successful fusion power plants. Usually, one has to resort to plasma edge codes to extrapolate the current understanding of divertor exhaust physics towards the operational windows anticipated in reactors. It is generally accepted that partially detached divertor operation will be required to mitigate heat loads to the divertor targets. However, the speed and convergence of the edge codes become serious bottlenecks in these highly-collisional regimes. In combination with a large number of design variables and uncertain model parameters, this turns computational divertor design into a challenging, time-consuming task.
In this seminar, an efficient one-shot optimization approach for divertor design is presented, which can achieve optimal divertor designs at a cost of only a few plasma edge simulations. Special attention is paid to the speed-up of the simulations using advanced fluid and hybrid fluid-kinetic neutral models. Recent progress for the computation of accurate adjoint sensitivities in the presence of statistical Monte Carlo noise is presented.