Asadi Shad, Tannaz2026-04-282026-04-282026https://hdl.handle.net/1828/23741This work presents a hybrid design framework for terahertz (THz) filters that integrates genetic algorithm (GA) optimization with ABCD matrix modeling and full-wave validation using HFSS. The proposed approach enables efficient inverse design by combining fast circuit-level analysis with electromagnetic accuracy. The developed framework, based on the work of Ali Dehghanian (2025) [32], employs a GA to explore a binary design space representing metallic and dielectric pixel distributions. Each candidate geometry is evaluated using an analytical ABCD model, enabling rapid calculation of S-parameters during optimization. The final geometry is then validated through finite-element simulation in HFSS to ensure physical accuracy. Two filter types were designed and evaluated: a low-pass filter (LPF) and a band-stop filter (BSF), with performance analyzed across frequencies ranging from 0.25 THz to 2.0 THz. For the LPF, the results show that at lower frequencies (0.25 THz), the design achieved strong attenuation (–10.2 dB) but required higher structural complexity and slower convergence. As the frequency increased, the optimization became more stable and efficient, with consistent convergence behavior and improved transmission characteristics. At higher frequencies (1.5–2.0 THz), the LPF demonstrated faster convergence, reduced structural complexity (as low as 12 rows), and stable performance with fitness values around –6.3 dB. For the BSF, a similar trend was observed. Lower frequencies exhibited wider stopbands but slower convergence, while mid-range frequencies (0.75–1.0 THz) showed improved stability and faster convergence. At higher frequencies, the BSF achieved stronger notch characteristics and more efficient optimization, with the best performance observed at 2.0 THz (–4.844 dB), along with smooth convergence and reduced parameter sensitivity. Across both filter types, the results indicate that increasing frequency leads to improved optimization efficiency, reduced structural requirements, and more stable convergence behavior. A comparison between ABCD-based analytical results and HFSS simulations shows strong agreement in both magnitude and phase responses, validating the accuracy and reliability of the proposed GA–ABCD–HFSS framework. Overall, the proposed methodology provides a fast, consistent, and physically reliable approach for designing high-performance THz filters.enAvailable to the World Wide WebCPSHFSSABCDTHzgenetic algorithmproject