Deep learning-enhanced prediction of terahertz response of metasurfaces

Metasurfaces offer an exciting opportunity to manipulate electromagnetic waves, presenting vast potential across diverse applications. In this study, we introduce a novel deep learning approach that integrates an Autoencoder with a Multi-Layer Perceptron to effectively forecast the Terahertz (THz) spectral response of metasurfaces. By harnessing a large dataset of training examples, our model adeptly captures the intricate correlation between metasurface structures and their optical responses, circumventing the traditionally time-consuming analysis of complex patterns. This proposed methodology furnishes a valuable tool for examining the THz transmission response of metasurfaces and has the potential to expedite metasurface design processes.