LineShine exascale simulates 1.34 trillion atom magnetic skyrmions

Researchers have achieved a breakthrough in atomistic simulation of magnetic skyrmions—topological spin textures promising for next-generation spintronics. The team, led by Pin Chen, combined a spin-constrained density-functional-theory-trained neuro-evolution potential with a structure-preserving spin-lattice integrator within a machine-learned framework. They applied architecture-specific optimizations including kernel fusion, SVE2 vectorization, and NUMA-aware data layout, delivering a seven orders-of-magnitude speedup over prior spin-aware methods. This allows modeling of the coupled lattice and spin evolution that governs real-temperature topological magnetic dynamics in materials like FeGe, where thermal effects drive helix-to-skyrmion transitions.

Deployed on the LineShine exascale supercomputer, the simulation scales to 12.45 million CPU cores with 89.7% weak-scaling efficiency, enabling the simulation of 1.34 trillion atoms and an equal number of spins while reaching 48.5 PFLOPS in double precision. For the first time, researchers can directly resolve real-temperature skyrmion nucleation and reorganization at micrometer-scale domains—critical for device-relevant simulation. This work establishes a new regime for predictive simulation of coupled spin-lattice topological magnetic dynamics, paving the way for designing skyrmion-based memory and logic devices.