Speaker
Description
The interaction of neutral and charged particles with the plasma-facing wall sets the boundary conditions for the confined plasma and is thus essential for important aspects like L-H power transition thresholds or impurity radiation. For example the transition from carbon to tungsten as wall material in ASDEX Upgrade reduced the L-H power threshold by about 25% which has been traced to the altered reflection properties of the surface [1]. Thus a proper description of the sputtering and reflection properties of plasma-facing walls is crucial. However, in most of the modelling of plasma-wall interaction in codes like EIRENE data tables computed for the case of atomistically flat surfaces are being used to generate the Monte Carlo (MC) samples – although in practice the plasma exposed surfaces exhibit a non-smooth morphology and roughness. While the simplification of a flat surface has been shown to be acceptable in many circumstances for the qualitative or even semi-quantitative assessment of sputtering yields (see e.g. [3]) the situation is different for the reflection properties. While for atomistically smooth samples and non-perpendicular particle impact forward reflection is always dominating sample roughness can cause pronounced backward reflection together with a sign-reversal of the integrated tangential momentum [4] – thus drastically altering some assumptions presently used in codes like SOLEDGE3X-EIRENE or SOLPS-ITER. Unfortunately due to their size a straightforward inclusion of the species, energy and angle dependent reflection distributions in EIRENE is challenging and a more efficient approach is desirable. In the contribution we present a practical method to generate simultaneously an arbitrary number of Monte Carlo samples following the reflection distribution. The key idea is the use of an adaptive recursive quadtree to assign samples according to the underlying probability density and this can be made efficient by exploiting recurrence relations of hemispherical harmonics. The performance is being demonstrated by applying the code to hydrogen, helium and neon-reflection distributions computed by SDTrimSP-7 for amorphous and crystalline ((100) and (111)) tungsten samples. More than one million MC-samples per second can now routinely be obtained – thus paving the way to include realistic reflection distributions in PWI-codes like EIRENE.
[1] L. M. Shao et al, Plasma Physics and Controlled Fusion 58 p. 025004, 2016
[2] R. Arredondo et al, Nuclear Materials and Energy 18 p. 72, 2019
[3] U. von Toussaint, R. Preuss, Nucl. Mat. Energy 41, p. 101817, 2024