Oxygen Vacancy Diffusion in Rutile TiO₂: Insight from Deep Neural Network Potential Simulations

Area of Science:

• Materials Science
• Computational Chemistry
• Surface Science

Background:

• Defects significantly influence titanium dioxide (TiO2) surface properties and reactivity.
• Understanding defect behavior is crucial for electronic and catalytic applications of TiO2.

Purpose of the Study:

• To develop and validate machine learning potentials for simulating defective TiO2 surfaces.
• To investigate the stability and diffusion dynamics of oxygen vacancies in TiO2.

Main Methods:

• Trained deep neural network potentials (DPs) using active learning on ab initio data.
• Validated DPs against density functional theory (DFT) for accuracy.
• Performed nanosecond-scale molecular dynamics simulations using DPs.

Main Results:

• DPs demonstrated good consistency with DFT results.
• Oxygen vacancies were found to be stable at 330 K but converted to more favorable sites at 500 K.
• DP-predicted oxygen vacancy diffusion barriers matched DFT predictions.

Conclusions:

• Machine learning potentials offer a DFT-accurate and accelerated approach for molecular dynamics simulations.
• This method enhances the understanding of microscopic mechanisms in fundamental TiO2 surface reactions.