Prediction of Energy Gaps in Graphene—Hexagonal Boron Nitride Nanoflakes Using Artificial Neural Networks

Machine learning methods are currently applied in conjunction with ab initio density functional theory (DFT) simulations in order to establish computationally efficient alternatives for high-throughput processing in atomistic computations. The proposed method, based on artificial neural networks (ANNs), was used to predict the HOMO-LUMO energy gap in quasi-0D graphene nanoflake systems with randomly generated boron nitride embedded regions. Several artificial neural network (ANN) algorithms were tested in order to optimize the network parameters for the problem at hand. The trained ANNs prove to be computationally efficient at determining the energy gap with good accuracy and show a significant speedup over the classical DFT approach.