Band gap and band alignment prediction of nitride-based semiconductors using machine learning

Abstract

Nitride has been drawing much attention owing to its wide range of applications in optoelectronics and there remains plenty of room for materials design and discovery. Here, a large set of nitrides has been designed, with their band gap and alignment being studied by first-principles calculations combined with machine learning. The band gap and band offset against wurtzite GaN accurately calculated by the combination of the screened hybrid functionals of HSE and DFT-PBE were used to train and test machine learning models. After comparison among different machine learning techniques, when elemental properties are taken as features, support vector regression (SVR) with radial kernel performs best for predicting both the band gap and band offset with a prediction root mean square error (RMSE) of 0.298 eV and 0.183 eV, respectively. The former is within the HSE calculation uncertainty and the latter is small enough to provide reliable predictions. Additionally, when the band gap calculated by DFT-PBE was added into the feature space, the band gap prediction RMSE decreased to 0.099 eV. Through a feature engineering algorithm, the elemental feature space-based band gap prediction RMSE further drops by around 0.005 eV and the relative importance of elemental properties for band gap prediction was revealed. Finally, the band gap and band offset of all designed nitrides were predicted and two trends were noticed: as the number of cation types increases, the band gap tends to narrow while the band offset tends to increase. The predicted results will provide useful guidance for precise investigation of nitride engineering.