Highly-Configurable Multi-Objective Optimization for Physical Parameter Extraction using Terahertz Time-Domain Spectroscopy

The use of terahertz time-domain spectroscopy provides one of the most versatile and promising techniques for the robust determination of optical parameters, which is needed to enable identification of materials for quality control, materials science advancement, tamper prevention, drug enforcement, and hidden explosives detection. Previously, the state-of-the-art relied on legacy error measures for minimization of simulation error and the standard practice was to use a single unique measurement for each unknown material in a sample. Successful optical parameter extraction for uniformly varying optical property materials is correlated with low variation in extracted optical properties. This work advances the state-of-the-art in optimization-based physical parameter extraction using terahertz time-domain spectroscopy. This is achieved by standardizing the signal processing methodology, clearly defining the best optimization formulation to yield low simulation error and optical property variation, and leveraging multiple measurements to reduce the impact of system-dependent artifacts on extracted optical properties. A thorough analysis of alternative error measures across numerous objective function formulations demonstrates that a 28% reduction in the Fabry-Perot etalon effect in the optical property of materials is achievable, compared with legacy approaches. The research conclusively demonstrates that time-domain objective function formulations yields simulation error that is 83% less than frequency-domain objective function formulations. Furthermore, the research shows that multi-measurement optimizations reduce oscillations in optical properties caused by the Fabry-Perot etalon effect by as much as 92%, compared with single-measurement optimizations. The research validates the numerical solutions to less than 6% error compared with analytical solutions, for uniform and non-uniform optical property materials. Importantly, the research extends the state-of-the-art by demonstrating the ability to simultaneously determine the effective sample thickness and orientation for high absorption samples comprised of solid and granular materials with uniform and non-uniformly varying optical properties. The outcomes of the research include a novel and comprehensive suite of methodologies that address fundamental complexities associated with exploitation of time-domain terahertz spectroscopic data