Rapid Design of Linear-to-Circular Polarization Conversion Metasurfaces Using Deep Learning

Based on deep learning, this study proposes a rapid design method of line-to-circular (LTC) polarization conversion metasurface (PCM) using ellipticity angle and pixelated metasurfaces. In this study, the ellipticity angle is used to characterize the LTC polarization conversion performance and the grouped-trained multi-convolutional neural network combination models is utilized for the rapid design. In the frequency range from 6 GHz to 14 GHz, the rapid designs of the narrowband LTC PCM are executed, which indicates the diversity of the designed metasurfaces, validating the effectiveness of the rapid design method.