Phase-to-pattern inverse design paradigm for fast realization of functional metasurfaces via transfer learning

• Published in: Nature communications 2021-05, Vol.12 (1), p.2974-2974, Article 2974
• Main Authors: Zhu, Ruichao, Qiu, Tianshuo, Wang, Jiafu, Sui, Sai, Hao, Chenglong, Liu, Tonghao, Li, Yongfeng, Feng, Mingde, Zhang, Anxue, Qiu, Cheng-Wei, Qu, Shaobo
• Format: Article
• Language: English
• Subjects: 639/301/1034/1037; 639/301/930/543; 639/624/399/1015; 639/705/1042; Accuracy; Algorithms; Big Data; Computer networks; Deep learning; Design optimization; Design techniques; Dictionaries; Electromagnetic radiation; Engineering; Humanities and Social Sciences; Information storage; Interdisciplinary subjects; Inverse design; Inverse method; Learning; Libraries; Machine learning; Metasurfaces; multidisciplinary; Neural networks; Science; Science (multidisciplinary); Transfer learning
• ISSN: 2041-1723; 2041-1723
• DOI: 10.1038/s41467-021-23087-y

Metasurfaces have provided unprecedented freedom for manipulating electromagnetic waves. In metasurface design, massive meta-atoms have to be optimized to produce the desired phase profiles, which is time-consuming and sometimes prohibitive. In this paper, we propose a fast accurate inverse method of designing functional metasurfaces based on transfer learning, which can generate metasurface patterns monolithically from input phase profiles for specific functions. A transfer learning network based on GoogLeNet-Inception-V3 can predict the phases of 2 8×8 meta-atoms with an accuracy of around 90%. This method is validated via functional metasurface design using the trained network. Metasurface patterns are generated monolithically for achieving two typical functionals, 2D focusing and abnormal reflection. Both simulation and experiment verify the high design accuracy. This method provides an inverse design paradigm for fast functional metasurface design, and can be readily used to establish a meta-atom library with full phase span. The design and optimization of a metasurface is a computationally- and time-consuming effort. Here, the authors propose a neural network-based algorithm for functional metasurface design, and demonstrate it for some functional metasurfaces.