J. Riemann

W7-X edge modelling with the 3D SOL fluid code BoRiS

BoRiS is a 3D scrape-off layer (SOL) transport code under development which is to solve a system of plasma fluid equations. BoRiS is currently extended towards a physics model including continuity, parallel momentum and energy equations for both electrons and ions. In addition the code requires the implementation of adequate solvers and the generation of high precision metric coefficients throughout the entire computational domain.

Benchmark of the 3-dimensional plasma transport codes E3D and BoRiS

The next generation of experiments - both for tokamaks and stellarators - requires the development of appropriate theoretical models. One important aspect here is the plasma edge physics description. Fluid transport codes extending beyond the standard 2-D code packages like B2-Eirene or UEDGE are under development. In the case of tokamaks, an interesting alternative line is the concept of an ergodic edge (necessary e.g. for ergodic divertors in TORE SUPRA or TEXTOR-94) creating a 3-D edge structure. To study these effects, a 3-D code E3D based upon a Multiple Coordinate Systems Approach is being developed. Presently, we are extending the program towards stellarator applications. A few new options are made available: single-island geometry and new formulation of boundary conditions. For the new stellarator W7-X, a 3-D finite volume code BoRiS is being developed using magnetic (Boozer) coordinates. In this paper, we present a benchmark of both codes, for a test geometry (one single magnetic island in W7-X) including full 3-D metric variations, solving for the strongly anisotropic electron heat conduction equation.

Modeling of localized neutral particle sources in 3D edge plasmas

A 3D fluid neutral model is added to the 3D plasma transport code BoRiS. The neutral model includes equations for parallel momentum and collisional perpendicular diffusion. This makes BoRiS an integrated plasma-neutral model suitable for a variety of applications. Results are presented for the distribution of neutrals from a localized gas source in the National Compact Stellarator Experiment geometry.

Hierarchy tests of edge transport models (BoRiS, UEDGE)

Abstract BoRiS is a 3D scrape-off layer (SOL) transport code under development to solve a system of plasma fluid equations. Using a simplified SOL model including particle continuity, parallel momentum and energy equations for both electrons and ions, BoRiS is tested in different geometries. To verify its proper operation in 1D and 2D cases, BoRiS solutions are compared to the results obtained with the established UEDGE code. In addition to these benchmarks some results for 3D problems are obtained.

Concept and Status of a 3D SOL Fluid Code

A new 3D scrape-off layer transport code is under development using the same plasma fluid approach as the well-known B2 code. The equations are solved in magnetic coordinates in order to deal with the complex 3D geometry. Starting from a simple temperature equation the code is successively extended towards a full physics model as in B2. Results for the solution of the coupled anisotropic Laplace equations for both electron and ion temperatures in a single magnetic island flux tube of W7-X are given.

Numerics in BoRis

Abstract We develop a new 3D scrape-off layer (SOL) transport code for the modelling of a fusion device. For the description of the complex 3D geometry we use magnetic coordinates. With this coordinate transform we are able to use standard discretization methods. The different plasma regions have different magnetic coordinates. All regions can be hold in a single grid using our grid generation. The Finite Volume Method leads to large sparse linear systems of equations. Different iterative solvers (Krylov subspace projection methods including matrix-free variants) and preconditioners are compared with respect to their computational efficiency, grid size scaling and memory requirements. Different schemes for interpolation of fluxes and gradients on scattered 3D data for general convection-diffusion problems can be used.

Linux Cluster Computing for Stellarator Studies

Publisher Summary This chapter discusses application of a Linux PC cluster for developing parallel programs that solve large systems of partial differential equations. By introducing 1 GBit network adapters and splitting the network tasks to multiple adapters, an improvement of the scaling of codes was achieved. The 1 GBit network cards were only used for the MPI communication of the parallel processes. These cards in conjunction with the 1 GBit switches have a low latency time, which is good for many small network packets. One of the 100 MBit adapters was used only for the parallel file system and another one was for all the other network communications. With these changes, a better parallel performance for the codes was achieved. The code scaled nearly linearly with the number of processors when using the 1 GBit network cards. For problems with moderate network traffic, linear scaling was nearly achieved. Results observed indicated that the scaling is sensitive to communication effort.