‘Computing metasurfaces’ to perform mathematical operations

Exploring new methods to rapidly process information and compute large amounts of data has become of growing importance in modern technology. Here we propose a fully analog computational platform, based on metasurfaces that can perform mathematical operations on the incident electromagnetic field distribution, as waves propagate through them. We demonstrate the potential of this concept by designing thin stacks of frequency selective surfaces that perform convolution with rectangular kernels and operate the second-order spatial derivative, fundamental operations for real-time image processing and data manipulation. This may pave the way towards ultrafast wave-based fully analog signal processing.