Spin-informed universal graph neural networks for simulating magnetic ordering

• 成果类型: Article
• 署名作者: Xu, Wenbin; Sanspeur, Rohan Yuri; Kolluru, Adeesh; Deng, Bowen; Harrington, Peter; Farrell, Steven; Reuter, Karsten; Kitchin, John R.
• 署名单位: United States Department of Energy (DOE); Lawrence Berkeley National Laboratory; Carnegie Mellon University; Max Planck Society; Fritz Haber Institute of the Max Planck Society
• 刊物名称: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
• ISSN/ISSBN: 0027-12439
• DOI: 10.1073/pnas.2422973122
• 发表日期: 2025-07-08
• 关键词: total-energy calculations; wave; TRANSITION

摘要

The screening and discovery of magnetic materials are hindered by the computational cost of first-principles density-functional theory (DFT) calculations required to find the ground state magnetic ordering. Although universal machine-learning interatomic potentials (uMLIPs), also known as atomistic foundation models, offer high-fidelity models of many atomistic systems with significant speedup, they currently lack the inputs required for predicting magnetic ordering. In this work, we present a data-efficient, spin-informed graph neural network framework that incorporates spin degrees of freedom as inputs and preserves physical symmetries, extending the functionality of uMLIPs to simulate magnetic orderings. This framework speeds up DFT calculations through better initial guesses for magnetic moments, determines the ground-state ordering of bulk materials and even generalizes to magnetic ordering in surfaces. Furthermore, we implement a closed-loop anomaly detection approach that effectively addresses the classic chicken-and-egg problem of creating a high-quality dataset while developing a uMLIP, unearthing anomalies in large benchmark datasets and boosting model accuracy. Significance The development of universal machine-learning interatomic potentials capable of simulating magnetic ordering is vital for the in silico discovery of indispensable magnetic materials across vast chemical spaces. To date, progress has been hindered by challenges in model design and the availability of high-quality datasets. Here, we introduce a general spin-informed graph neural network framework, coupled with an anomaly detection approach, that achieves state-of-the-art performance in simulating magnetic ordering and enhances the quality of large benchmark datasets. These developments broaden the capabilities of atomistic foundation models and advance the evolution of data-centric AI.