Overcoming Information Sparsity in Metasurfaces for Full-Color Holography via End-to-End Design

We propose an end-to-end (E2E) system for RGB meta-hologram generation that efficiently determines the optimal material and geometry for target holograms, eliminating the need for exhaustive simulations of every possible meta-atom configuration. A neural network is developed to accurately map the material and structural parameters to the transmission spectra of the single-layered metasurfaces. The E2E system leverages the network to identify suitable meta-atom candidates, aiming to (i) enhance the hologram efficiency and (ii) minimize cross-talk between images at different channels. By effectively navigating the sparse RGB phase distribution, the system enables the fabrication of an optimized metasurface that projects three distinct images at different wavelengths. We anticipate that our method will contribute to the development of multiplexed holographic displays and also pave the way for integrated metasurface design and optimization of metasurfaces.