Date of Award

May 2026

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mechanical Engineering

First Advisor

Chonglin Zhang

Abstract

This thesis presents the numerical verification of Comet, a GPU-accelerated Particle-in-Cell with Monte Carlo Collisions (PIC/MCC) code designed for low-temperature plasma simulations. To ensure efficient deployment across heterogeneous computing architectures, the framework leverages the Kokkos performance-portability layer for high-performance execution. Verification was conducted through a series of hierarchical stages with increasing physical complexity, benchmarking each component of the fundamental PIC/MCC simulation cycle. This structured approach establishes the numerical fidelity of the code's algorithmic building blocks prior to its application in predictive plasma research.

The first stage verifies field interpolation and particle pusher modules against canonical single-particle cases, focusing on ion gyration in a uniform magnetic field and electron oscillation within a purely electric field. The resulting trajectories and characteristic frequencies show excellent agreement with analytical solutions, confirming the accuracy of the force interpolation and numerical integration schemes.

The second stage evaluates the Monte Carlo Collisions (MCC) module by benchmarking electron impact ionization kinetics. Under field-free and loss-free conditions, the ionization rate coefficient (K_ion) was extracted using a logarithmic slope method, showing excellent agreement with theoretical values derived from direct cross-section integration. Furthermore, the simulated electron temperature evolution aligns with analytical energy balance predictions across multiple regimes without observable systematic drift.

In the final stage, a two-dimensional simulation of a radio-frequency capacitively coupled plasma (CCP) discharge in argon was performed, providing an integrated verification of all PIC/MCC computational cycle components through comparison with the reference code eduPIC. Spatial profiles of particle densities, electric potential, and power deposition exhibit consistent qualitative agreement between the two frameworks. The simulated sheath thickness aligns with the Child-Langmuir law, with observed deviations attributable to the inherent expansion of RF sheaths beyond time-averaged analytical estimates. Although a quantitative discrepancy of approximately 5.2% in peak electron density was observed, a subsequent sensitivity study identified insufficient initial particle statistics (N_PPC) as the primary source of the offset. These findings emphasize that statistically converged and grid-independent solutions are necessary prerequisites for cross-code verification, and establish systematic N_PPC convergence studies as an essential protocol for future high-resolution simulations.

Download

Share

COinS