Stability of Polarons and Oxygen Vacancies in BiVO4 Revealed by Machine-Learning-Driven Simulations

This repository presents the input files and NEP model presented in the article "Operando Stability of Polarons and Oxygen Vacancies in BiVO4 Revealed by Machine-Learning Driven Simulations". The repository contains control.in : Input file for DFT calculations using FHI-AIMS. nep.in : Input file used to train the neuroevolution potential (NEP). train.xyz : Atomic structures and reference energies/forces to train the NEP. nep.txt : Final NEP model used for production runs. nep_inv.txt : Inverted NEP model, that is marked atoms (Ti) are treated as "normal" vanadiums. model.xyz : Example atomic structure for the oxygen vacancy in the neutral charge state. run_TI_nep.in : Input file to run thermodynamic integration between two charge states. run_vertical_transition.in : Input file to obtain vertical transition energies between two charge states. Important note : The atoms labeled as Ti are not titanium, but vanadium with localized electrons. Labeling atoms is necessary to model electron polarons and oxygen vacancies in charges states +2, +1 and 0.