Metasurface-Based Optical Neural Network and its Application in Next-Generation Optical Communications and Networks
Chenguang Rong, Lin Wu, Jin Tao, Yongzhi Cheng, Kai Wang, Lin Chen, Hui Luo, Fu Chen, Xiangcheng Li
Abstract
Optical neural network (ONN), integrating deep learning algorithms and optical computing (OC) hardware, are gradually emerging as a powerful tool for artificial intelligence (AI) computing. Metasurfaces (MSs) have infused new vitality into the development of ONN by leveraging their nanoscale structures to precisely manipulate optical field parameters. In recent years, MS-based optical neural networks (M-ONNs) have sparked an innovative trend in the OC field, significantly impacting the development of next-generation optical communication and networking technologies. This paper presents an overview of M-ONN, beginning with an introduction to its theoretical framework and various design approaches. Then, the latest research progress in the practical applications of M-ONN is reviewed, emphasizing the crucial role of direct laser writing (DLW) in the manufacturing process. Further discussion covers the potential applications of M-ONN in next-generation optical communication and networking. Lastly, the focus shifts to the industrial progress of M-ONN. With technology maturing and the industrial chain solidifying, M-ONN is transitioning from the laboratory to real-world applications, poised to become a catalyst for transformative developments in the next-generation optical communication and networking field.