Inverse Design of Graphene FET by Deep Neural Network

Abstract

We propose deep neural networks to uncover the relationship between the gate shape and the electrical response of graphene field effect transistors. A deep neural network is used to efficiently optimize a transport gap for a graphene field effect transistor that utilizes the pseudo-optic negative reflection at a p-n junction. Using the finite-difference-time-domain method for massless Dirac fermions, the electrical responses of graphene field effect transistors with arbitrary gate shapes were calculated, and the results were used to train a deep neural network. It turns out that the trained deep neural network was not only able to foresee the graphene pseudo-optic response for a specific gate shape but also to provide an optimized design for a desired electrical response by the inverse design.