A New Architecture for High‐Channel Color Router Using Optical Diffractive Neural Networks

In multispectral imaging, traditional color filter arrays have long been constrained by an inherent limitation: the difficulty in achieving both high spectral accuracy and high optical efficiency simultaneously. To overcome this challenge, this paper applies the optical diffractive neural network (ODNN) to the Multi‐channel color Router (MCR), proposing a novel MCR‐ODNN architecture. MCR‐ODNN uses multi‐layer diffractive surfaces to achieve efficient spectral splitting and spatial routing of incident light, significantly enhancing multi‐channel spectral separation capability while maintaining high optical energy efficiency. It supports a number of color routing channels far exceeding currently reported results. Research indicates that the multi‐layer diffractive structure can effectively suppress inter‐channel crosstalk, although its efficiency decays exponentially with an increasing number of layers. Furthermore, increasing the number of diffraction layers provides a greater performance improvement than increasing the neuron density within a single layer. This paper successfully fabricated and validated a 4‐channel MCR‐ODNN, with experimental measurements showing strong agreement with simulation data. This research provides an innovative technical pathway for high‐fidelity color imaging and holds broad application prospects.