Verification of techniques for accelerated and stable SOLPS-ITER simulations including plasma drifts

Abstract SOLPS-ITER is a widely employed transport code for Scrape-Off Layer (SOL) modeling, commonly used to assist experimental interpretation in current tokamak experiments, and for the design of future devices. An aspect of particular importance for SOL transport is particle drifts. When drifts are included in SOLPS-ITER simulations, two types of numerical instabilities are frequently encountered. The first one is a core instability, that limits the maximum allowed time-step. For a typical simulation of an Ohmic L-mode plasma in the TCV tokamak, this leads to a convergence time in the order of 10 3 days when running on parallelized modern computer clusters. Different techniques to reduce the convergence time are verified, mostly based on the original work from Kaveeva et al . These comprise artificially increasing the maximum allowed time-step or reducing the longest physical time-scale in simulations, and reducing the CPU cost of a single iteration. For each technique, different speed-up methods and parameters are tested, and an optimal choice is identified. These optimal methods provide a significant speed-up and have no influence on the stationary solution. The three optimal methods are employed simultaneously in a large set of simulations, scanning different divertor geometries, divertor conditions, plasma currents, transport coefficients, and toroidal field direction. They consistently obtain a speed-up of approximately 3 orders of magnitude, while having no impact on the stationary solution. The second instability is a radial boundary instability, that limits the parameter range of attainable divertor conditions. Three factors affecting the radial boundary instability are presented: mesh setup, number of ‘second level’ internal iterations in the numerical scheme, and boundary conditions at the radial boundaries; when these are properly modified, a robust solution to the boundary instability can be obtained.