From density functional theory to machine learning predictive models for electrical properties of spinel oxides

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• Spinel oxides are promising for energy storage and catalysis due to abundant transition metals.
• Predicting electronic conductivity is crucial for optimizing these materials in devices.

Purpose of the Study:

• To develop machine learning models for predicting electronic conductivity and band gaps of spinel oxides.
• To establish a new database of spinel oxide properties from first-principles calculations.

Main Methods:

• Created a database of spinel oxide band structure and conductivity properties.
• Trained machine learning algorithms on this database to predict electronic conductivity and band gaps.
• Scaled models from quantum level to continuum conductivity.

Main Results:

• Accurate predictions of band gap and conductivity were achieved using a relatively small database.
• The model predicted high conductivity for nickel-rich spinels.
• Model predictions aligned with experimental trends for manganese cobalt spinels.

Conclusions:

• Machine learning models can accurately predict electronic conductivity and band gaps of spinel oxides based on composition.
• This approach facilitates the discovery and optimization of materials for energy conversion devices like batteries and supercapacitors.