W-band frequency selective digital metasurface using active learning-based binary optimization

The W-band is essential for applications like high-resolution imaging and advanced monitoring systems, but high-frequency signal attenuation leads to poor signal-to-noise ratios, posing challenges for compact and multi-channel systems. This necessitates distinct frequency selective surfaces (FSS) on a single substrate, a complex task due to inherent substrate resonance modes. In this study, we use a digital metasurface platform to design W-band FSS on a glass substrate, optimized through binary optimization assisted by active learning. The digital metasurface is composed of a periodic array of sub-wavelength unit cells, each containing hundreds of metal or dielectric pixels that act as binary states. By utilizing a machine learning model, we apply active learning-aided binary optimization to determine the optimal binary state configurations for a given target FSS profile. Specifically, we identify optimal designs for distinct FSS on a conventional glass substrate, with transmittance peaks at 79.3 GHz and Q-factors of 32.7.