Inverse Design of Metal-Insulator-Metal Metasurface Based on Equivalent Circuit Model

Metasurfaces formed of two-dimensional array of tailored meta-atoms has the capability to control the transmission and reflection electromagnetic waves in subwavelength scale. The flexibility in designing meta-atom makes the design process involve iterative full wave simulation, which is time-consuming and labor-intensive. To accelerate the design process, this paper proposes a method based on an equivalent circuit model and deep learning network to predict the forward response and perform inverse design of metal-insulator-metal (MIM) metasurfaces. We introduce an equivalent circuit modeling approach to analysis the response of MIM metasurfaces. The equivalent circuit model combines with deep learning networks to predict the optical response of metasurfaces and perform inverse design of the structural geometric parameters for desired responses. By comparing the prediction results with full-wave simulation results, we demonstrate that the proposed model achieves high prediction accuracy.