Fussy Inverse Design of Metamaterial Absorbers Assisted by a Generative Adversarial Network

The increasing demands for metasurfaces have led researchers to seek effective inverse design methods, which are counting on the developments in optimization theory and deep learning techniques. Early approaches of based established a unique mapping between device’s geometry parameters its designated EM characteristics. However, generated solution traditional method may not be applicable due practical fabrication conditions. designers sometimes want choose most one from multiple schemes can all meet requirements given indicators. A fuzzy is quite demand. In this study, we proposed metamaterial absorbers generative adversarial network (GAN). As data-driven method, self-built data sets constructed trained by GAN, contain absorber’s their corresponding spectral response. After training process finished, it generate possible satisfy customized absorptivity frequency bands absorbers. model include structure sizes impedance values, indicates that has ability learn variety features. effectiveness robustness been verified several examples both narrowband broadband Our work proves feasibility using methods break limits one-to-one method. This profound usage more complex device problems future.