Machine Learning on a Synergistic Transition Metal Dual-Atom Surface for Efficient Decomposition of Ammonia

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry
• Renewable Energy

Background:

• Ammonia (NH₃) decomposition is a promising route for clean hydrogen (H₂) production, crucial for transitioning away from fossil fuels.
• Developing efficient catalysts is essential for optimizing NH₃ decomposition technology.
• Dual-atom catalysts offer tunable properties for enhanced catalytic activity.

Purpose of the Study:

• To design and implement a computational framework for screening dual-atom catalysts for NH₃ decomposition.
• To integrate machine learning (ML) with high-throughput (HT) calculations for efficient catalyst discovery.
• To identify novel dual-atom catalyst candidates with superior performance.

Main Methods:

• Conducted first-principles-based high-throughput (HT) calculations on selected dual-atom systems.
• Employed feature engineering to identify important and low-correlation descriptors.
• Trained a machine learning model using HT calculation results to predict catalytic performance for a large number of structures.

Main Results:

• Successfully predicted the catalytic performance of 2187 dual-atom structures using the trained ML model.
• Identified several promising dual-atom catalysts, including RuMo─O─C, ScOs─N─C, and OsV─N─C, for NH₃ decomposition.
• Density of states and differential charge analyses confirmed synergistic catalytic effects in the identified catalysts.

Conclusions:

• The integrated ML-HT framework is effective for accelerating the discovery of advanced dual-atom catalysts.
• The identified catalysts (RuMo─O─C, ScOs─N─C, OsV─N─C) show significant potential for efficient NH₃ decomposition.
• This approach paves the way for designing next-generation catalysts for clean hydrogen energy.