Conditional Generative Adversarial Networks for Inverse Design of Multifunctional Metasurfaces

Electromagnetic (EM) metasurfaces can present a versatile platform for realization of multiple diverse EM functionalities with incident wave frequency, polarization, or propagation direction through appropriate choice unit cells structures. However, the inverse design multifunctional relies on massive full‐wave numerical simulations to obtain an optimized solution. Herein, step‐by‐step procedure based conditional generative adversarial networks (cGANs) integrated Gramian angular fields (GAFs) reduce computational time required in microwave is proposed. The proposed initially implements GAFs encode desired multiobjective scattering parameters (SPs) images and then passes them cGAN model map three‐layer metasurfaces. study uses robust dataset, including 54 000 metasurface structures corresponding SPs train validate model. This article also presents two case examples using different independent full‐space coverage justify performance studies demonstrate that, despite random nature training data samples, reliably predicts SPs.