多路复用
计算机科学
可扩展性
光子学
全息术
制作
电子工程
光学工程
极化(电化学)
波分复用
软件
光纤
简单
稳健性(进化)
光通信
光开关
集成光学
三维光学数据存储
光学性能监测
寄主(生物学)
可微函数
频分复用
波长
光学
信号处理
计算机硬件
作者
Jie Wang,Feilong Yu,Jin Chen,Jiuxu Wang,Rongsheng Chen,S. R. Wayne Chen,Tie Geng,Yiming Xiao,Ziying Li,Yue Zuo,Huaizhong Xing,Cheng Guo,Guanhai Li,Xiaoshuang Chen,Wei Lu
标识
DOI:10.1002/lpor.202502573
摘要
ABSTRACT Metasurfaces provide powerful control over the amplitude, phase, polarization, and wavelength of light, enabling compact and multifunctional photonic systems. However, designing metasurfaces with high‐dimensional multiplexing capabilities often requires composite or interleaved meta‐atom configurations, which can introduce fabrication complexity, near‐field coupling, and limited scalability. Here, we present an end‐to‐end differentiable design framework that integrates meta‐atom geometry modeling with global optical performance optimization. By leveraging a neural network–based surrogate model, we establish a continuous mapping between geometric parameters and polarization‐dependent optical responses, allowing gradient‐based co‐optimization of metasurface functionality using only single‐type, fabrication‐friendly meta‐atoms. We experimentally demonstrate a single‐layer metasurface that achieves eight‐channel polarization multiplexing at a single operating wavelength, simultaneously generating distinct nanoprinting patterns and 3D holograms with minimal inter‐channel crosstalk. Extending this approach across two wavelengths enables 16 independent optical channels, validating the scalability of the framework. The proposed strategy unifies structural simplicity with functional versatility, offering a robust and generalizable platform for advanced applications in high‐capacity optical displays, secure data encoding, and integrated photonic systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI