Recently, under the guidance of Academician Zhuang Songlin from University of Shanghai for Science and Technology, Professors Zang Xiaofei and Zhu Yiming from the Terahertz Technology Innovation Team, in collaboration with Professor Ding Fei's research group from Eastern University of Technology, proposed a new method for achieving polarization-selectable diffraction neural networks (PS-DNNs). Through the combined phase and polarization multi-degree-of-freedom regulation, the unidirectional and bidirectional network function switching with controllable polarization has been achieved, thereby simultaneously realizing the purpose of information encryption and sharing. The research results are presented as "All-Optical Information Encryption and Sharing Achieved by metasurface diffraction neural networks" (" Polarization-selective unidirectional and bidirectional diffractive neural networks for The title "information security and sharing" was published in Nature Communications (IF=14.7, ranked in the first zone of the Chinese Academy of Sciences). Doctoral student Guo Ziqing is the first author, and Dr. Tan Zhiyu is the co-first author of the paper. Professors Zang Xiaofei, Zhu Yiming and Ding Fei are the corresponding authors.

This study addresses the contradiction between information security and data sharing in traditional full-capacity information processing and proposes a new architecture of a cascaded metasurface polarization-controllable diffraction neural network. By integrating phase control, polarization conversion, and direction/polarization selection functions, it realizes the switching between unidirectional information encryption and bidirectional information sharing functions of the neural network. To address the issue of limited degrees of freedom on traditional 3D printed diffraction surfaces, the research team innovatively cascaded a quarter-wave plate (QWP) metasurface atomic array with a metal grating to achieve a unidirectional and bidirectional diffraction neural network with polarization-selectable features for digital classification and non-destructive imaging. Furthermore, the half-wave plate (HWP) metasuric array and the metal grating were cascaded to achieve a more comprehensive unidirectional - bidirectional - unidirectional DNNs with functional switching. Finally, a new architecture for information transmission and data exchange with high security performance was designed based on PS-DNNs. The experimental results show that this architecture can achieve massive data transmission on the same physical platform while encrypting and sharing specific data. The relevant research provides new ideas for all-optical computing, all-optical information processing and communication security.

A high-security information transmission and data exchange system based on PS-DNNs

The article links: https://doi.org/10.1038/s41467-025-59763-6