High-order implicit solver in conservative formulation for tokamak plasma transport equations

TL;DR

This paper introduces a high-order implicit numerical scheme for solving stiff plasma transport equations in tokamaks, emphasizing efficiency, conservation, and improved accuracy over traditional methods.

Contribution

The paper presents a novel high-order implicit solver combining a 4th order spatial discretisation with a 2nd order implicit time-stepping scheme, optimized for plasma transport simulations.

Findings

Achieves accurate, stable, and fast non-linear convergence.

Enforces conservation in spatial coordinate up to machine precision.

Improves transient process simulation accuracy with second-order time stepping.

Abstract

An efficient numerical scheme for solving transport equations for tokamak plasmas within an integrated modelling framework is presented. The plasma transport equations are formulated as diffusion-advection equations in two coordinates (a temporal and a spatial) featuring stiff non-linearities. The presented numerical scheme aims to minimise computational costs, which are associated with repeated calls of numerically expensive physical models in a processes of time stepping and non-linear convergence within an integrated modelling framework. The spatial discretisation is based on the 4th order accurate Interpolated Differential Operator in Conservative Formulation, the time-stepping method is the 2nd order accurate implicit Runge-Kutta scheme, and an under-relaxed Picard iteration is used for accelerating non-linear convergence. Temporal and spatial accuracies of the scheme allow for coarse grids, and the implicit time-stepping method together with the non-linear convergence approach contributes to robust and fast non-linear convergence. The spatial discretisation method enforces conservation in spatial coordinate up to the machine precision. The numerical scheme demonstrates accurate, stable and fast non-linear convergence in numerical tests using analytical stiff transport model. In particular, the 2nd order accuracy in time stepping significantly improves the overall convergence properties and the accuracy of simulating transient processes in comparison to the 1st order schemes.