Search for high-capacity oxygen storage materials by materials informatics.

Area of Science:

• Materials Science
• Computational Chemistry
• Catalysis

Background:

• Oxygen storage materials (OSMs) like CeO2-ZrO2 (p-CZ) are crucial catalyst supports in automotive emission control.
• OSMs possess oxygen storage capacity (OSC), enabling reversible oxygen release/uptake based on atmospheric conditions.

Purpose of the Study:

• To accelerate the discovery of high-capacity OSMs using a materials informatics (MI) approach.
• To combine experimental, first-principles calculations, and machine learning (ML) for predicting OSC.

Main Methods:

• Measured OSC for 60 metal oxides under varying gas conditions and temperatures (973, 773, 573 K).
• Computed descriptors based on atomic properties and first-principles calculations.
• Trained a support vector machine regression model, selecting key features related to crystal stability and structure.

Main Results:

• Identified features like cohesive energy as highly correlated with OSC.
• The predictive model successfully screened 1300 existing oxides.
• Cu3Nb2O8 was synthesized and demonstrated higher OSC than p-CZ.

Conclusions:

• The developed MI scheme significantly accelerates the identification of novel OSMs.
• Cu3Nb2O8 shows promise as a next-generation oxygen storage material.