Accelerated Discovery of Topological Spin Phase Diagrams in CrSBr Magnets Via High‐Throughput Optimization

ABSTRACT Efficiently resolving the high‐dimensional energy landscapes of topological spin textures remains a major challenge in computational magnetism. Here, we develop a high‐throughput algorithm that integrates orthogonal‐spin optimization with the Limited‐Memory‐Broyden– Fletcher–Goldfarb–Shanno (M‐LBFGS) method to accelerate the elucidation of phase‐diagrams of spin textures on‐demand. By enforcing strict orthogonality constraints during optimization and exploiting the quasi‐Newton efficiency of M‐LBFGS, our algorithm achieves rapid, robust convergence to local minima while requiring orders of magnitude fewer iterations than conventional techniques. This enables high‐throughput screening across thousands of parameter combinations using only modest computational resources. Applied to the van der Waals magnet CrSBr, our framework resolves more than 2000 atomistic configurations and uncovers a hierarchy of previously unknown metastable phases, including domain‐wall bimerons, bimeron chains, Néel skyrmions, and hybrid spiral states. Minimum energy‐path techniques coupled with our solver reveal their stability, transition mechanisms, and thermal lifetimes, establishing a direct link between the emergence of topological spin textures and the noncritical anomaly observed experimentally in the susceptibility of CrSBr below 40 K. This optimization algorithm defines a generalizable and scalable paradigmatic for predictive mapping of magnetic phase diagrams, enabling the discovery of intricate spin textures and emergent topological states across a wide spectrum of magnetic materials.