Monte Carlo Simulation of Runaway Electrons

Abstract

Runaway electrons appear in tokamak plasmas during thermal quenches - disruptions that change the plasma conducti-vity. This gives rise to an accelerating electric field, which if, higher than the decelerating Coulomb friction force, can give electrons unlimited acceleration, resul-ting in relativistic particles, which may damage the first wall of the tokamak. In this thesis, I am discussing the relevance and application of computer simulation to model runaway electrons. The code that is used is called ARENA (Avalanche of Runaway Electrons Numerical Analysis), which utilises a Monte Carlo approach to solve the three-dimensional bounce averaged Fokker Planck equation. I also compare with the LUKE finite difference solver for primary runaway generation, as well as numerical and theoretical data [1, 2]. The majority of the thesis deals with perfor-mance and structural updates to the decade-old ARENA code which has resulted in a new ARENA 90-code which is written in the Fortran 90 language and is under active development by EFDA-ITM (European Fusion Development Agreement - Integrated Tokamak Modelling) task force. I have also made a proof of concept of a parallelised collision operator running on a GPU (Graphics Processing Unit) using the OpenCL API (Application Programming Interface) stan-dard, which demonstrates the exibility of the new ARENA code. The thesis is primarily centred around two bench-marks, the preservation of a Maxwellian distribution with no outer electric field, and primary runaway generation under a constant electric field. In addition, there is an in-depth discussion of the simulation parameter space and design solutions of the ARENA code. Secondary generation from a seed of runaway electrons is discussed, but not implemented in the new version of the code. The final results show a good match of published data for all test cases considered. The speed of simulations is greatly increased compared to the old ARENA code and the porta-bility and usability of the new code should help contri-bute to future works. Together, this offers valuable insight on the possible applications and limitations of runaway simulation and the ARENA code in particular.