Tripolarization and multimode holography via a neural metasurface for high-capacity security encryption

Metasurfaces have emerged as a promising platform for multifunctional integration through precise manipulation of diverse physical properties of light, exhibiting significant potential for high-capacity information technologies. As the escalating demands for data processing capabilities, deep neural networks have been extensively investigated in high-capacity optical computing. In this work, we propose a novel dual-dimensional multiplexing holography which can fully synergize polarization states and spatial modes as information channels, using a single-layer and non-interleaved metasurface with assistance of a neural network. As a proof, we present an 8-channel neural metasurface using two orthogonal polarization states and four spatial modes, in addition to a 12-channel neural metasurface exploiting three orthogonal polarization bases and four spatial modes, enabling reconstructions of various holographic images. By leveraging the dual-dimensional multiplexing neural metasurface, we provide an optical encryption scheme that offers benefits of construction compatibility, transmission capacity, and function scalability through a low-complexity design and ultra-compacity device. Combining metasurfaces, neural networks and holographic displays, our framework provides a promising path for intelligent metasurface-driven applications in optical information classification and encryption.