Enhancing the Efficiency of Complex Systems Crystal Structure Prediction by Active Learning Guided Machine Learning Potential

2025-05-13

AI Summary

Multicomponent crystal structure prediction (CSP) faces computational intractability due to exponential growth of configurational and stoichiometric spaces. Method: We propose an active-learning-driven machine learning potential (MLP) framework with high generalizability, featuring the first fully automated MLP training pipeline for complex multicomponent systems—integrating graph neural networks, active learning, high-throughput structural relaxation, and phase stability assessment. Contribution/Results: The framework achieves high-fidelity potential energy surface modeling from limited data, overcoming traditional CSP limitations for quaternary and higher-order systems. Applied to the Mg–Ca–H ternary and Be–P–N–O quaternary systems, it accelerates structural relaxation by over two orders of magnitude and discovers multiple thermodynamically stable novel compounds. This significantly enhances data efficiency and scalability of CSP for complex materials, enabling previously infeasible exploration of multicomponent chemical space.

Technology Category

• Machine Learning: Multimodal Learning
• Multiagent Systems: Multiagent Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Machine learning and data science for the Web

Abstract

Understanding multicomponent complex material systems is essential for design of advanced materials for a wide range of technological applications. While state-of-the-art crystal structure prediction (CSP) methods effectively identify new structures and assess phase stability, they face fundamental limitations when applied to complex systems. This challenge stems from the combinatorial explosion of atomic configurations and the vast stoichiometric space, both of which contribute to computational demands that rapidly exceed practical feasibility. In this work, we propose a flexible and automated workflow to build a highly generalizable and data-efficient machine learning potential (MLP), effectively unlocking the full potential of CSP algorithms. The workflow is validated on both Mg-Ca-H ternary and Be-P-N-O quaternary systems, demonstrating substantial machine learning acceleration in high-throughput structural optimization and enabling the efficient identification of promising compounds. These results underscore the effectiveness of our approach in exploring complex material systems and accelerating the discovery of new multicomponent materials.

Problem

• Overcoming combinatorial explosion in complex crystal structure prediction
• Reducing computational demands for multicomponent material systems
• Enhancing efficiency of machine learning potentials for CSP

Innovation

• Active learning guided machine learning potential
• Automated workflow for generalizable MLP
• High-throughput structural optimization acceleration