Emerging computational and machine learning methodologies for proton-conducting oxides: materials discovery and...

Area of Science:

• Materials Science
• Computational Chemistry
• Solid-State Chemistry

Background:

• Proton-conducting oxides are crucial for electrochemical devices like fuel cells.
• Discovering and understanding these materials is complex.
• Advanced computational and machine learning (ML) approaches are needed.

Purpose of the Study:

• To develop and apply computational and ML methodologies for proton-conducting oxide discovery.
• To gain fundamental insights into proton transport mechanisms.
• To accelerate the identification of novel, high-performance proton conductors.

Main Methods:

• Development of computational and ML methodologies over a 5-year research project.
• Application of replica exchange Monte Carlo simulations for defect and hydration analysis.
• Integration of computational insights with experimental data for materials exploration ('Materials discovery through interpretation').

Main Results:

• Discovery of three new proton-conducting oxides (perovskite and non-perovskite structures).
• Identification of octahedral tilts/distortions and oxygen affinity as key factors influencing proton transport in doped barium zirconates.
• Revealed realistic defect configurations and hydration behavior using Monte Carlo simulations.
• Identified perovskites with proton conductivity >0.01 S/cm and high chemical stability at 300°C.

Conclusions:

• Computational and ML methodologies are effective tools for accelerating materials discovery in proton-conducting oxides.
• Understanding structure-property relationships, such as the role of octahedral distortions and oxygen affinity, is vital for designing efficient proton conductors.
• The 'Materials discovery through interpretation' approach successfully integrates theory and experiment to identify promising materials for energy applications.