A Beam-Splitting Bianisotropic Metasurface Designed by Optimization and Machine Learning

Electromagnetic metasurfaces have attracted significant interest recently due to their low profile and advantageous applications. Practically, many metasurface designs start with a set of constraints for the radiated far-field, such as main-beam direction(s) side lobe levels, end non-uniform physical structure surface. This problem is quite challenging, since required tangential field transformations are not completely known when only placed on scattered fields. Hence, surface properties cannot be solved analytically. Moreover, translation desired unit cells can time-consuming difficult, it often one-to-many mapping in large solution space. Here, we divide inverse design process into two steps: macroscopic microscopic step. In former, use an iterative optimization find that radiate far-field pattern complies specified constraints. exploits non-radiating currents ensure passive lossless design. step, these optimized realized using machine learning surrogate models. The effectiveness this end-to-end synthesis demonstrated through measurement results beam-splitting prototype.