Screening of Material Defects using Universal Machine-Learning Interatomic Potentials

Area of Science:

• Materials Science
• Computational Chemistry
• Condensed Matter Physics

Background:

• Computational modeling accelerates the discovery of new materials with desired properties.
• Machine-learning interatomic potentials offer high accuracy at reduced computational cost.
• Previous applications of these potentials have not explored screening defective materials.

Purpose of the Study:

• To assess the accuracy of universal machine-learning interatomic potentials for large-scale screening of defective materials.
• To explore the application of these potentials in discovering new stable compounds and simulating material properties.
• To analyze the formation energies of vacancies in relation to oxidation numbers.

Main Methods:

• Performed vacancy calculations for 86,259 materials using the Materials Project database.
• Utilized universal machine-learning interatomic potentials for accurate and efficient simulations.
• Analyzed formation energies and oxidation states to identify promising material candidates.

Main Results:

• Demonstrated that machine-learning interatomic potentials are sufficiently accurate for large-scale defective material screening.
• Identified new materials at or below the convex hull, indicating potential stability.
• Successfully simulated the etching of low-dimensional materials, showcasing model versatility.

Conclusions:

• Universal machine-learning interatomic potentials are a powerful tool for accelerating materials discovery, particularly for defective systems.
• The study validates the use of these potentials for identifying novel stable compounds and understanding defect properties.
• This approach opens new avenues for computational screening in materials science and engineering.