Multi-dimensional optical imaging on a chip (2505.00408v1)

Abstract: Light inherently consists of multiple dimensions beyond intensity, including spectrum, polarization, etc. The coupling among these high-dimensional optical features provides a compressive characterization of intrinsic material properties. Because multiple optical dimensions are intrinsically coupled rather than independent, analyzing their inter-relationships and achieving their simultaneous acquisition is essential. Despite the existing optical techniques to obtain different-dimensional data with cumbersome systems, joint acquisition of multi-dimensional optical information on a chip is still a serious challenge, limited by intensity-only photoelectric detection, single-dimensional optical elements, and finite bandwidth. In this work, we report a multi-dimensional on-chip optical imaging (MOCI) architecture, which is functionally composed of three layers, including a multi-dimensional encoding layer to simultaneously encode different dimensions of incident light, an image acquisition layer to collect coupled intensity data, and a computational reconstruction layer to recover multi-dimensional images from a single frame of coupled measurement. Following the MOCI architecture, we for the first time fabricated a real-time (74 FPS) on-chip polarization-hyperspectral imaging (PHI) sensor, with 2048$\times$2448 pixels at 61 spectral channels covering the VIS-NIR range and 4 polarization states. We applied the PHI sensor for simultaneously resolving hyperspectral and polarization information of complex scenes, and for the first time demonstrated new applications including hyperspectral 3D modeling with normal and height maps, and hyperspectral sensing against strong reflection and glare...