Current metasurface designs for analog image processing suffer from trade-offs between resolution, throughput, polarization asymmetry, bandwidth, and isotropy. This paper demonstrates that dispersion engineering optimizes these metrics simultaneously. Silicon metasurfaces are experimentally shown to perform isotropic, dual-polarization edge detection with numerical apertures above 0.35 and 35 nm bandwidths around 1500 nm. Quantitative metrics reveal high throughput efficiencies, approaching theoretical limits. These results advance low-loss, high-efficiency broadband optical computing and image processing.

Michele Cotrufo, Akshaj Arora, Sahitya Singh, Andrea Alù