High-efficiency optical diffractive neural network color router with one-dimensional structure

We propose a high-efficiency optical diffractive neural network color router (ODNN-CR). This structure features four one-dimensional diffraction layers, in which the first three layers are responsible for efficient RGB spectral separation across the visible spectrum (400-700 nm), while the fourth layer performs orthogonal focusing to mitigate inter-pixel crosstalk. The ODNN-CR demonstrates outstanding performance in the visible spectrum, achieving peak color routing efficiencies of 61% (R), 64% (G), and 60% (B), an exceptionally high average total efficiency of 87%, and low inter-channel crosstalk ratios of 29% (R), 44% (G), and 36% (B). Experimental validation using LED sources shows that the measured results closely match simulations. This study offers a feasible technological solution for next-generation image sensors with high color fidelity.