Deep learning-enhanced prediction of terahertz response of metasurfaces

Abstract

Listen

Translate article

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.