Predicting and synthesizing terahertz spoof surface plasmon polariton devices with a convolutional neural network model.

With the increasing global attention to deep learning and the advancements made in applying convolutional neural networks in electromagnetics, we have recently witnessed the utilization of deep learning-based networks for predicting the spectrum and electromagnetic properties of structures instead of traditional tools like fully numerical-based methods. In this study, a Convolutional Neural Network (CNN is proposed for predicting spoof surface plasmon polaritons, enabling the examination of the absorption spectrum of metallic multilevel-grating structures (MMGS) and designing various sensor devices and absorbers in the shortest time possible. To expedite the training process of this network, a semi-analytical method of rigorous coupled-wave analysis (RCWA) enhanced with the fast Fourier factorization (FFF) technique has been employed, significantly reducing the data generation time for training. This CNN enables the prediction of existing spoof surface plasmon polaritons (SSPPs) and their intensity within a 110% frequency range. In addition to the speed of dataset generation, the simple framework of this network has also facilitated the training of the network in a short time. By comparing the network in this study with previous works, it is apparent that in addition to structural and geometrical changes in the unit cell, the designer is afforded greater freedom in determining the material of the incident medium and, for the first time, specifying the angle of incidence of the source. Finally, for the validation of the suggested network, the predictive power of the absorption spectrum of various structures is compared with traditional methods. Three examples are provided for inversely designing several sensor devices and absorbers in the terahertz band using the proposed CNN and the genetic optimization algorithm.