Machine-learning-aided DFT-1/2 calculations for bandgaps of zinc oxide thin films

Abstract

Using the Density Functional Theory (DFT) -1/2 method, the electronic structure and bandgap of zinc oxide (ZnO) are calculated, providing the initial data sets for machine learning based on the genetic-algorithm-based artificial neural networks (GANNs) model. The predicted bandgaps from the well-trained GANNs model are close to (less than 5%) the calculated ones from the DFT-1/2 method along cutoff radius = 2.4 for Oxygen and cutoff radius = 1.2 for Zinc, agreeing with the experimental data. Our results show that combining the DFT-1/2 method with GANNs is an efficient way to correct the band gap of Zinc Oxide in DFT simulation.