Active phase discovery in heterogeneous catalysis via topology-guided sampling and machine learning

Area of Science:

• Catalysis
• Materials Science
• Computational Chemistry

Background:

• Understanding active phases is crucial for advancing heterogeneous catalysis.
• Computational modeling of active phases faces challenges in configuration generation and calculation.
• Existing methods struggle with the vast array of atomic configurations and material morphologies.

Purpose of the Study:

• To develop an automatic and efficient framework for exploring active phases in catalysis.
• To systematically sample configurations across diverse coordination environments and material morphologies.
• To enable rapid computations using efficient machine learning force fields.

Main Methods:

• Utilizing a topology-based algorithm leveraging persistent homology for systematic configuration sampling.
• Employing efficient machine learning force fields for rapid computational analysis.
• Demonstrating the framework on hydrogen absorption in Palladium (Pd) and oxidation dynamics of Platinum (Pt) clusters.

Main Results:

• Successfully modeled hydrogen absorption in Pd, revealing a "hex" reconstruction critical for CO2 electroreduction.
• Investigated Pt cluster oxidation, showing oxygen incorporation reduces activity in oxygen reduction reactions.
• Predicted active phases and their catalytic impacts align closely with experimental observations.

Conclusions:

• The proposed framework effectively models complex catalytic systems.
• It enables the discovery of active phases under specific environmental conditions.
• This strategy advances the understanding and design of heterogeneous catalysts.