ABSTRACT Two strategies are investigated to reduce the computational cost of steady‐state plasma boundary simulations with kinetic neutrals: underrelaxation and automatic local timestepping. The models are developed and tested in SOLPS‐ITER. The underrelaxation scheme allows reducing the number of Monte Carlo particles per iteration up to two orders of magnitude for a given accuracy. It is shown that using a cell‐based local timestep computed at runtime is a valid approach in SOLPS‐ITER and allows stable convergence to correct results. With this approach, the number of iterations is decreased by a factor of 4.8 for the studied JT‐60SA case, while being highly compatible with the underrelaxation scheme.
Uytven et al. (2026) studied this question.