Reverse design and optimization research of digital terahertz bandpass filters

Abstract

In this paper, an ingenious reverse design method is applied to the design and optimization of terahertz bandpass filters in order to achieve standardized design of high-performance terahertz functional devices. An equivalent model of subwavelength metasurface mapped to digital space is established. Based on ideal objective functions and constraints, intelligent algorithms begin a bold journey to explore the vast potential structure in the solution space. Through iterative refinement, the algorithm reveals optimal structural patterns, unlocking areas of unparalleled performance. The Direct Binary Search (DBS) algorithm and the Binary Particle Swarm Optimization (BPSO) algorithm are compared in optimization prowess. When using the DBS algorithm to optimize the design area, it polls the logic states of all pixel units point by point, which is time-consuming and easy to get stuck in the local optimal value. However, BPSO algorithm has stronger global search capabilities, faster convergence speed, and higher accuracy. Through a comprehensive comparison of the device performance optimized by the two algorithms, the solution optimized by BPSO algorithm has better out-of-band suppression performance and a narrower full width at half peak, but slightly lower transmittance at the center frequency. The bandpass filter has a center frequency of 0.51 THz, a bandwidth of 41.5 GHz, and an insertion loss of -0.1071 dB. When considering computational efficiency, DBS algorithm lags behind, with a simulation time of 11550 seconds, while BPSO algorithm takes only 9750 seconds. Compared with the traditional forward design, the reverse design method can achieve the narrower band, lower insertion loss, better out-of-band suppression and polarization stability. The fine structural changes of the optimal results have a significant impact on spectral performance, demonstrating the superiority and uniqueness of reverse design. This technology contributes to the design and optimization of high-performance and novel functional devices.