Abstract

This talk describes the high performance plasma simulation framework developed at the University of Michigan and discusses its application for fusion plasmas.

Tamas Gombosi
University of Michigan

The Space Weather Modeling Framework (SWMF) provides a flexible plug-and-play type, parallel, high performance framework for plasma simulations. All components can be replaced with alternatives, and one can use only a subset of the components. The components are coupled to the control module via standardized interfaces, and an efficient parallel coupling toolkit is used for the pairwise coupling of the components. The execution and parallel layout of the components is controlled by SWMF. Both sequential and concurrent execution models are supported. The configuration, compilation and execution of the framework can be done with a user friendly Graphical User Interface.

We also describe the 3D MHD code at the heart of SWMF, called BATS-R-US. Modern approaches to numerical solutions of the MHD equations are extensively based on fundamental mathematical properties of the governing equations. While the different formulations of the MHD equations describe the same physics at the differential equation level, there are important practical differences when one solves discretized forms of the various formulations. In the last decade several developments have led to a new paradigm for numerical simulations of physical systems governed by partial differential equations. They are: (i) the availability of massively parallel computers; (ii) the evolution of solution-adaptive techniques from research topic to practical tool; (iii) advances in basic numerical methods, particularly for hyperbolic conservation laws. Methods that have been developed by the aerodynamics, applied mathematics and DoE communities have recently begun to be implemented in space-physics codes, which solve the governing equations for a compressible magnetized plasma. These techniques include high-resolution upwind schemes, block-based solution-adaptive grids and domain decomposition for parallelization. While some of these techniques carry over relatively straightforwardly to plasma physics, space physics and fusion simulations pose some new challenges.