Leveraging Parallel Speedup for Intelligent Design of Metasurface
Hang Yu Ge,Yu Tong Jiang,Feng Liu,Wei E. Sha
DOI: https://doi.org/10.23919/inc-usnc-ursi61303.2024.10632435
2024-01-01
Abstract:Over the past few years metasurface has drawn enormous research interests due to its excellent capability of manipulating electromagnetic (EM) waves across a wide frequency range. By carefully designing the geometric parameters of each unit at subwavelength scale, metasurface can perform flexible control of EM responses including amplitude, phase, polarization and reflection/transmission angle, etc. Consequently, metasurface has been employed in many wireless communication scenes (e.g. frequency converter, beam steering and filter). However, the design of metasurface element is very challenging. On the one hand, traditional methods using massive full wave simulation based on commercial software entail high computational burden and significant time consumption. On the other hand, although rigorous analytical methods based on EM theory and optimization algorithms (like genetic algorithm and neural network) enable quick calculation of the EM response, apparent performance degradation occurs in multi-element scenario, because edge effect and mutual coupling which are usually hard to estimate are not considered. Therefore, building an efficient approach that combines the accuracy of full wave simulation and the intelligence of optimization algorithm is highly desirable for metasurface design.