Development of a Parallel Electrostatic PIC Code for Modeling Electric Propulsion
This thesis presents the parallel version of Coliseum, the Air Force Research Laboratory plasma simulation framework. The parallel code was designed to run large simulations on the world fastest supercomputers as well as home mode clusters. Plasma simulations are extremely computationally intensive as they require tracking millions of particles and solving field equations over large domains. This new parallel version will allow Coliseum to run simulations of spacecraft-plasma interactions in domain large enough to reproduce space conditions. The parallel code ran on two of the world fastest supercomputers, the NASA JPL Cosmos supercomputer ranked 37th on the TOP500 list and Virginia Tech's System X, ranked 7th. DRACO, the Virginia Tech PIC module to Coliseum, was modified with parallel algorithms to create a full parallel PIC code. A parallel solver was added to DRACO. It uses a Gauss-Seidel method with SOR acceleration on a Red-Black checkerboard scheme. Timing results were obtained on JPL Cosmos supercomputer to determine the efficiency of the parallel code. Although the communication overhead limits the code's parallel efficiency, the speed up obtained greatly decreases the time required to run the simulations. A speed up of 51 was reached on 128 processors. The parallel code was also used to simulate the plume expansion of an ion thruster array composed of three NSTAR thrusters. Results showed that the multiple beams merge to form a single plume similar to the plume created by a single ion thruster.