Terahertz Holographic Image Encryption via Thermally Tunable Directional Janus Metasurface

ABSTRACT Terahertz (THz) holography demonstrates immense potential in biomedical detection, virtual reality, and information encryption security. Janus metasurfaces, which can independently control the amplitude and phase of THz waves from opposite directions, offer a promising platform for directional multiplexing in THz holography. However, most existing designs rely on conventional forward design methods, limiting the system to just two degrees of freedom (DOFs) and significantly restricting both channel count and holography imaging quality. In this study, we introduce a bidirectional deep neural network (Bi‐DNN) inverse design method, combined with thermally responsive phase‐change material vanadium dioxide (VO 2 ), for the development of Janus metasurfaces. The proposed design enables independent control of incident direction, frequency, and temperature as three distinct DOFs. The meta‐atoms selected through the Bi‐DNN exhibit a transmittance exceeding 90%, effectively generating low‐crosstalk, eight‐channel THz holographic images with an imaging efficiency of 78%. Furthermore, by combining the metasurface with the inherent superposition properties of Chinese characters to achieve high‐intensity and high‐density holographic encryption in the THz band. This design offers a novel strategy and technical foundation for applications in high‐capacity data storage, holographic encryption, and secure wireless communication.