Discovering Superhard B-N-O Compounds by Iterative Machine Learning and Evolutionary Structure Predictions.

Area of Science:

• Materials Science
• Computational Chemistry
• Solid-State Physics

Background:

• The search for superhard materials is crucial for technological advancements.
• Boron-nitrogen-oxygen (B-N-O) compounds represent a promising class of materials.
• Predictive modeling can accelerate the discovery of novel materials with desired properties.

Purpose of the Study:

• To identify new superhard B-N-O compounds using an iterative machine learning (ML) approach.
• To evaluate the thermodynamic stability and mechanical properties of predicted B-N-O compositions.
• To explore the electronic properties of promising superhard B-N-O candidates.

Main Methods:

• Iterative machine learning (ML) models trained on crystal structures from an evolutionary algorithm.
• Cohesive energy calculations to assess thermodynamic stability.
• Meta-GGA density functional theory (DFT) for electronic structure and bandgap analysis.

Main Results:

• ML models rapidly converged to identify potentially superhard and thermodynamically favorable B-N-O compositions (e.g., B5N3O3, B6N4O3) with x ≥ 3.
• Predicted B-N-O compounds are wide bandgap (≥4.4 eV) insulators.
• Valence band maximum is associated with nitrogen p-orbitals near vacant sites.

Conclusions:

• The combination of iterative ML and ab initio simulations is a powerful strategy for novel material discovery.
• Identified B-N-O compounds show significant potential for superhard applications.
• This study highlights the predictive power of computational methods in materials science.