Combining neural networks and metamaterials concept to compute temporal derivatives

This study introduces a multilayer metamaterial (MTM) for performing temporal differentiation. By using a neural network (NN), we efficiently determined optimal layer thicknesses for a 13-layer ITO and TiO2 multilayer MTM. The designed structure successfully mimics the temporal differential operator of a differentiator in the frequency domain, as evidenced by close agreement between theoretical, predicted, and simulated reflection spectra. The presented results demonstrate the potential of MTMs for wave-based computing of temporal operations, with implications for signal processing applications.