Temporal Signal Processing with Nonlocal Optical Metasurfaces

Abstract: Nonlocal metasurfaces have recently enabled an ultra-compact, low-power and high-speed platform to perform analog image processing. While several computational tasks have been demonstrated based on this platform, most of the previous studies have focused only on spatial operations, such as spatial differentiation and edge detection. Here, we demonstrate that metasurfaces with temporal nonlocalities - that is, with a tailored dispersive response - can be used to implement time-domain signal processing in deeply subwavelength footprints. In particular, we show a passive metasurface performing first-order differentiation of an input signal with high-fidelity and high-efficiency. We also demonstrate that this approach is prone to scalability and cascaded computation. Our work paves the way to a new generation of ultra-compact, passive devices for all-optical computation, with applications in neural networks and neuromorphic computing.