Extending & Optimizing a Numerical Model of an Antenna in Plasma

Document Type

Project

Lead Author Type

CIS Masters Student

Advisors

Dr. Greg Wolffe, wolffe@gvsu.edu

Embargo Period

1-19-2016

Abstract

Solar energy constantly impacts the ionospheric plasma layer of the Earth’s atmosphere, especially during a solar event. These collisions result in an abnormal increase in entropy as electrons and ions are generated, potentially disrupting long-distance communications and unlocking GPS satellites at low altitudes. In order to better understand these phenomena, the plasma environment can be simulated using a full-wave, self-consistent, Finite-Difference Time-Domain (FDTD) model of an antenna interacting with collisional, magnetized, multi-species plasma. In this method, a 3D mathematical model is solved numerically and used to predict the behavior of the plasma over time and under varying conditions.

In current FDTD models of plasma regions surrounding objects, the simulation space must be very large in order to allow a full investigation of effects. A sizeable simulation space results in a correspondingly large computational demand and thus, suffers from prolonged execution times. Addressing this problem was the first goal of this project. Phase one utilized execution time / memory space profiling to identify bottlenecks and memory usage patterns, identifying the potential for code optimization. The next phase required both local and global data analyses, and was designed to expose loop iteration interdependencies. By eliminating these interdependencies, the code could be parallelized using multi-threading to fully exploit modern multi-core architectures. These optimizations dramatically reduced execution time, which allowed for the study of much larger simulation spaces.

Although plasma phenomena have been recorded and explored at multiple layers of the ionosphere, current analytical theory cannot fully explain nor account for the effects of plasma collision frequency and temperature on lower plasma frequency behaviors. To facilitate the exploration of the effect of these and other parameters, a distributed Pythonic testing suite was developed to comprehensively explore parameter ranges by automatically distributing workloads among a cluster of machines.

Finally, in an effort to better understand these effects, and those of collisions occurring at the boundary conditions of the simulation, several visualizations were developed. With longer recordings, lower frequency waves became measurable in the simulated results. The effect of the plasma frequency, collision frequency, and temperature environment variables on the simulated lower frequency-range recordings can now be tested to better understand their contribution to plasma behavior. Further optimizations of the boundary conditions will allow for more extensive parameter testing, which will ultimately lead to better agreement between simulation results and experimentally observed data.

This document is currently not available here.

Share

COinS