Study of low-temperature plasma development utilizing a GPU-implemented 3D PIC/MCC simulation

Summary form only given. A GPU-accelerated 3-dimensional PIC/MCC simulation code was developed using the CUDA environment to study the physical processes involved in the development of a low-temperature plasma. The simulation results aid in quantifying transient plasma development as it is often inaccessible experimentally in detail even with modern noninvasive techniques such as non-linear laser spectroscopy or high-speed electrical diagnostics. Hence, computational methods, such as Particle-in-Cell (PIC) and Monte Carlo Collision (MCC), provide a complementary approach to determining the mechanisms leading to plasma development. However, fully modeling the physics of the plasma development is made difficult by the number of plasma processes that must be tracked simultaneously, and only recently have computing resources provided the capability to track tens of millions of particle interactions. Furthermore, the introduction of graphics processing unit (GPU) computing provides an attractive means for economical and efficient parallelization of scientific codes through a framework such as NVIDIA CUDA. As such, a GPU-accelerated 3-dimensional PIC/MCC simulation was developed using the CUDA environment to provide characteristics during the initial stage of plasma development in atmospheric pressure nitrogen. The simulation was run on a NVIDIA GTX 580 with 3 GB of memory and 512 CUDA cores. The simulated geometry consists of two paraboloid electrodes with a gap distance of 5 millimeters with Dirichlet boundary conditions, and 22 unique electron interactions with molecular nitrogen are considered. The electrodes are excited with a step voltage pulse of several thousand volts also assuming a uniformly distributed initial electron density of 10
⁴
cm
^-3
in the volume. For instance, results from a 5 nanosecond simulation reveal the development of positive ion space charge channels near the anode and cathode regions. These channels appear consistent with high-speed streamer photographs captured during plasma formation. The electron energy distribution function (EEDF) indicates a non-Maxwellian velocity distribution during the application of the high electric field. Furthermore, a typical electron density near the cathode is on the order of 7 × 10
⁸
cm
^-3
. The results from numerical simulation will be compared in detail to experimentally accessible parameters such as electron temperature and dissociation degree.