Spin-multiplexed point spread function engineering via dielectric metasurface for simultaneous optical differentiation and high-resolution imaging

Abstract In the era of artificial intelligence and machine learning, all-optical computing has garnered renewed attention for its intrinsic advantages in processing speed, energy efficiency, and parallelism over electronic systems. A key operation in such architecture is optical differentiation, which enables edge detection and feature extraction from target scenes. Metasurfaces—capable of controlling the amplitude, phase, and polarization of light at the nanoscale—have emerged as promising platforms for implementing compact optical differentiators. However, existing designs typically rely on auxiliary imaging optics to capture the processed outputs and are often limited in spatial resolution or differentiation order. Here, we introduce a new class of metasurface optical differentiators that simultaneously achieve arbitrary-order optical differentiation and high-resolution imaging in a standalone single-layer configuration, eliminating the need for additional optics. By precisely engineering multiple complex-valued point spread functions via spin multiplexing, our design enables on-demand, arbitrary-order differentiation directly across the light field and overcomes long-standing constraints in spatial resolution and operational flexibility. We experimentally demonstrate two devices capable of performing 0 th /1 st -order and 2 nd /3 rd -order differentiation with spatial resolution up to 228.0 lp/mm (line width of 2.19 μm). Their performance is further validated across diverse scenarios, including high-intensity illumination and real-time live cell imaging. Our results demonstrate that through rigorous point spread function engineering, metasurfaces offer a transformative platform for integrated, high-performance all-optical computing systems with broad potential in biological imaging, information processing, and material characterization.