Deep learning bandgaps of topologically doped graphene

Abstract

Manipulation of physical and chemical properties of materials via precise doping affords an extensive range of tunable phenomena to explore. Recent advance shows that in the atomic and nano scales topological states of dopants play crucial roles in determining their properties. However, such determination is largely unknown. Meanwhile, with the development of computer's power of calculation, the possibility that we can use deep learning method: Neural Network which has a large amount of parameters in its algorithm to train and predict the data become truth now. We incorporate these state-of-art concepts and developed deep learning algorithms to study the property of them. First we use neural network to predict bandgaps of boron-nitrogen pair doped graphene. The bandgaps were calculated by the ab initio calculations, and together with the structural information they were fed to convolutional neuron networks (CNNs). These trained CNNs afford great prediction accuracy, showing square of the coefficient of correlation (R2) of > 90% and root-mean-square errors of [about]0.1 eV. The transfer learning was further performed by leveraging data generated from smaller systems to improve the prediction for large systems. Success of this work provides a cornerstone for future investigation of other properties of topologically doped graphene and other 2D materials. Second we also do the reverse direction research. We Use Generative Adversarial Network (GANs) to predict the structure of doped graphene based on the bandgaps value we want and it gets decent performance. Meanwhile we build a web to show our project performance to others.