On-demand design of holographic metasurfaces and continuous phase and amplitude modulation method based on deep learning

Abstract

Metasurfaces have shown unique application value in the field of holography due to its outstanding ability to manipulate electromagnetic waves. However, improving the design efficiency and imaging quality remains a challenging task. In this work, we propose a deep learning method that can design holographic metasurface structures on demand, with the Mean Absolute Error (MAE) of 0.04 for both amplitude and phase. We utilize this method to inverse design all-silicon-based metasurfaces operating in the terahertz range, achieving a MAE of 0.015 for two target images. This method not only significantly enhances the design efficiency of holographic metasurfaces but also enables continuous modulation of both phase and amplitude. Consequently, it greatly improves both the design efficiency and imaging quality of holographic metasurfaces, providing a new direction for their design.