AI-accelerate discovery of altermagetic materials

Magnetic materials are the cornerstone of the modern information society, and their magnetism is mainly divided into two major categories: ferromagnetism and antiferromagnetism. In recent years, based on the theory of spin space groups, a new magnetic phase—altermagnetism—has been proposed. Altermagnetic materials possess the duality of antiferromagnetic ordering in real space and anisotropic spin splitting in reciprocal space. This endows altermagnetic materials with the advantages of both ferromagnetic and antiferromagnetic materials, thus opening up a new direction for the development of next-generation information technology. Owing to the anisotropic spin splitting, altermagnetic materials exhibit giant magnetoresistance and piezomagnetic effects. Under spin-orbit coupling, due to the breaking of effective time-reversal symmetry, altermagnetic materials can realize various anomalous phenomena similar to ferromagnetic materials, including the quantum anomalous Hall effect, anomalous Hall effect, and anomalous Kerr effect, among which the anomalous Hall effect has been experimentally confirmed. Recently, by combining spin group symmetry with known magnetic structures, researchers have identified 141 altermagnetic materials. However, the number of altermagnetic materials remains very limited, so there is an urgent need to discover more altermagnetic materials to deeply investigate their novel physical properties. Traditional methods for discovering altermagnetism rely on the magnetic structures of experimentally determined magnetic materials and the analysis of spin space group symmetry, meaning that traditional methods are ineffective in the absence of magnetic structure information. On the other hand, since determining the magnetic structure of materials requires costly neutron scattering experiments, the magnetic structures of less than 2% of magnetic materials have been experimentally confirmed to date. Therefore, developing a new method that can break through the limitations of magnetic structure information and discover new altermagnetic materials without relying on prior knowledge has become an urgent priority in current research.

Recently, the team of intelligent computational physics led by Ze-Feng Gao (Associate Researcher), Peng-Jie Guo (Associate Professor), Zhong-Yi Lu (Professor) from the School of Physics, and Hao Sun (Associate Professor with Tenure) from the Gaoling School of Artificial Intelligence at Renmin University of China, proposed an AI-based search engine method called MatAltMag to accelerate the discovery of altermagnetic materials. As shown in Figure 1, this method integrates a variety of advanced AI technologies (such as pre-training and fine-tuning techniques and graph neural networks) with physical methods (symmetry analysis and first-principles electronic structure calculations). It has identified 50 new altermagnetic materials under conditions with very limited labeled samples, including the first discovery of four i-wave altermagnetic materials. Meanwhile, the research team also investigated the physical properties of all the newly discovered altermagnetic materials. This work lays the foundation for AI-driven approaches to finding specific functional new materials. The achievement has been published in the National Science Review, 2025, titled “AI-accelerated Discovery of Altermagnetic Materials”. Ze-Feng Gao (Associate Researcher), Shuai Qu (PhD student), and Bocheng Zeng (PhD student) are the co-first authors, while Hao Sun (Associate Professor with Tenure), Peng-Jie Guo (Associate Professor), Zhong-Yi Lu (Professor) are the co-corresponding authors.

The 50 altermagnetic materials predicted by the research team possess a wealth of novel physical properties. Among them, Nb2FeB2 is a g-wave altermagnetic metal (as shown in Figure 2). In the absence of spin-orbit coupling, Nb2FeB2 features Weyl-type nodal rings protected by mirror symmetry, Weyl-type nodal lines protected by spin symmetry in the general regions of the Brillouin zone, and Dirac points protected by nontrivial spin symmetry. Under spin-orbit coupling, Nb2FeB2 transforms into an altermagnetic Weyl semimetal phase, with its magnetic moments oriented along the x-axis. As a result, Nb2FeB2 also exhibits a topologically enhanced large intrinsic anomalous Hall effect.

It is worth mentioning that MatAltMag is an efficient method for searching functional materials in databases. Based on MatAltMag, the research team has further developed an AI-driven inverse design workflow for materials, named InvDesFlow, using diffusion models. This method aims to discover functional new materials that are not present in existing databases. Utilizing this approach, the team has predicted that Li2AuH6 has a superconducting transition temperature of around 140 K under ambient pressure. Additionally, they have identified several ternary and quaternary hydrogen-rich compounds with high critical superconducting temperatures that are not found in known databases.

---

---