ISBN 978-3-8439-4261-4

978-3-8439-4261-4, Reihe Ingenieurwissenschaften

Stephen Copplestone
Particle-Based Numerical Methods for the Simulation of Electromagnetic Plasma Interactions

230 Seiten, Dissertation Universität Stuttgart (2019), Hardcover, A5

Zusammenfassung / Abstract

This work addresses the kinetic simulation of plasma flows interacting with electromagnetic and electrodynamic forces, for which particle-based methods are utilized.

A fundamental equation that can be used for describing partially ionized plasma flows is Boltzmann's equation, which can be split into a collisionless and a collisional part by an operator splitting approach.

The kinetic description of the collisionless part, which is referred to as Vlasov's equation, is handled by Particle-In-Cell methods, for which two different approaches are developed in this work, an electrostatic solver based on Poisson's equation and an electrodynamic solver based on Maxwell's equations. For both solvers, high-order discontinuous Galerkin methods on unstructured curvilinear hexahedral meshes are employed, which offer superior parallelization approaches as compared with other high-order methods.

For the collisional part, the Direct Simulation Monte Carlo method is utilized, which also accounts for chemical reactions within the plasma flow, such as dissociation and ionization processes, which are relevant to this work.

Different practical applications, for which the Particle-In-Cell method is coupled with the Direct Simulation Monte Carlo method, are simulated during the course of this work that are related to gyrotron resonator devices and laser-plasma interaction with thin foils.

Furthermore, high-performance computing aspects are shown to be highly critical when efficient simulations are to be performed. Here, the influence of static and dynamic workload-balancing approaches is investigated, for which improved algorithms become necessary. Different workload measurement techniques and workload distribution approaches are developed and tested and these schemes continue to play a crucial role in the future, when simulations with strong load imbalances are encountered.