Symmetry-guided prediction of magnetic-ordered ground states
Abstract
Given the scarcity of experimentally confirmed magnetic structures, the prediction of magnetic ground states is crucial yet remains a long-sought challenge due to the complex potential energy landscape. Here, we propose a symmetry-guided framework that systematically generates magnetic configurations without requiring any experimental input or prior assumptions. Within a symmetry-breaking scenario, we incorporate the recently developed oriented spin space group formalism, which captures symmetry-breaking induced by both magnetic ordering and spin-orbit coupling. By performing nonrelativistic and relativistic first-principles calculations, we establish the energy ladder of the generated magnetic configurations. Exemplified by three prominent unconventional magnets, we demonstrate that only a few dozen calculations are sufficient to identify the ground-state magnetic structure. To demonstrate the universality and robustness of our approach, we conduct large-scale benchmark tests on the MAGNDATA database. Our framework successfully reproduces experimentally reported magnetic geometries for 78% of the surveyed materials, among which 93% have their spin orientations successfully generated when considering SOC. Furthermore, in a large-scale first-principles benchmark involving 305 compounds, 82% of experimentally reported magnetic structures are accurately captured within an energy tolerance of 5 meV per magnetic atom. Beyond reproducing known magnetic configurations, our framework further predicts a variety of low-energy metastable phases, including altermagnets, spin-orbit magnets, and noncollinear antiferromagnets with spin splitting or geometric Hall effect. Our work establishes a general and efficient route toward large-scale prediction of magnetic structures and unconventional magnets, and offers insight into the origins of magnetic interactions across diverse material systems.
Turn this paper into a full lesson
ArcXiv compiles a staged curriculum from this paper: 8-12 lessons across beginner → advanced, synthesised section guides, visuals, flashcards, a quiz, exercises, and on-demand deep dives per section. Grounded in the abstract, never invented.