Zhengchang Liu, Pu Peng, Xiao He, Zhibo Dang, Yuchen Dai, Yuxiang Chen,Xinyuan Shao, Yu Li, Yijing Huang, Donglin Liu, Guangyi Tao, Yunhao Zhang,and Zheyu Fang*
Abstract:As planar optical elements, metasurfaces confer an unprecedented potentialto manipulate light, which benefits from the deep control of the interactionsbetween nanostructures and light. In the past decade, considerable progresshas been made in various metasurfaces for on-demand functions, drawinggreat interest from the scientific community. However, it is a great challengeto integrate different functions into a single metasurface, due to theincapability of manipulating light at different dimensions and the lack ofuniversal intelligent design strategy. Here, an intelligent design platformbased on deep neural networks is proposed, which can map betweenstructure parameters and optical response. The well-trained network modelcan intelligently retrieve nanostructures to meet multidimensional opticalrequirements of metasurfaces. Four metalenses for chiral focusing arerealized by the design platform and the simulation results are highlyconsistent with the design target. In addition, metalenses based on arbitrarypolarization at various working wavelength are also demonstrated, showingthat the method has powerful design ability. Various optical properties ofnanostructures, such as phase shift and polarization, are manipulated bydeep neural networks, which can greatly promote the development ofmultifunctional devices and further pave the way for optical display,communication, computing, sensing, and other applications
Advanced Physics Research 21 February 2023 
Advanced Physics Research - 2023 - Liu.pdf
