Track the frontier, from condensed matter to cosmology.
Search papers, labs, and topics across Lattice.
FLARE replaces stepwise numerical integration of stochastic Landau–Lifshitz–Gilbert dynamics by directly predicting full-field magnetization endpoints across arbitrary physical horizons via conditional flow matching. By parameterizing transport over anchor-relative local axis-angle rotations, the framework strictly preserves pointwise unit-norm spherical manifold constraints without post-hoc projection or accumulation of integration error. Evaluated on multi-nanosecond trajectories, FLARE cuts fair energy score by 37.3% over competitive learned baselines while delivering up to a 3,062× batch speedup over standard GPU-accelerated numerical solvers like MuMax³.
Stochastic micromagnetics can bypass sequential PDE integration entirely: horizon-conditioned flow matching over local rotation manifolds directly predicts stochastic full-field states while matching exact geometric invariants at a 3,062× speedup over MuMax³.
Long-horizon micromagnetic simulation remains expensive because conventional and learned solvers typically propagate Landau--Lifshitz--Gilbert (LLG) dynamics step by step. Existing learned approaches generally retain stepwise integration or model deterministic evolution, leaving full-field, direct-horizon stochastic prediction largely unexplored. We propose FLARE, a flow-matching framework that recasts stochastic finite-time magnetization prediction as conditional transport over anchor-relative local axis-angle rotations. This rotation-space formulation respects the intrinsic geometry of magnetization dynamics and preserves pointwise unit norm by construction. By explicitly conditioning on the physical prediction horizon, FLARE directly generates full-field stochastic endpoints across multiple target times without stepwise integration. Against the strongest single-checkpoint external baseline on each metric, FLARE achieves 29.9% lower angular energy distance ($15.30^\circ$), and a 37.3% lower fair energy score (0.393). On a representative composed 5-ns two-segment protocol, FLARE achieves a $3{,}062\times$ best-batch speedup over the widely used GPU micromagnetic solver MuMax$^3$ on a single GPU.