Prediction of THz Absorption and Inverse Design of Graphene-Based Metasurface Structure Using Deep Learning

This paper presents a new method based on the neural network for predicting the absorption of the THz Graphene-based metasurface structure and inversely designing the structure based on desired absorption spectrum. the absorption spectra have been computed for different values of structure's parameters. A neural network is trained to predict the absorption of the structure for any random value of the structure parameters with high accuracy and less time-consuming. Moreover, the well-trained neural network can perform the inverse design of the structure based on the desired absorption spectrum. Finally, the performances of various deep learning algorithms are demonstrated.