Equivariant Neural Networks for Spin Dynamics Simulations of Itinerant Magnets
This work addresses the challenge of simulating complex magnetic systems like the Kondo lattice model, which is important for condensed matter physics and materials science, representing an incremental improvement with specific domain gains.
The authors tackled the problem of large-scale spin dynamics simulation for itinerant magnets by developing a novel equivariant neural network architecture that ensures lattice translation and spin rotation equivariances, achieving a validation error reduction to less than one-third compared to invariant descriptor-based models on a square lattice and successfully reproducing skyrmion crystal phase transitions on a triangular lattice.
I present a novel equivariant neural network architecture for the large-scale spin dynamics simulation of the Kondo lattice model. This neural network mainly consists of tensor-product-based convolution layers and ensures two equivariances: translations of the lattice and rotations of the spins. I implement equivariant neural networks for two Kondo lattice models on two-dimensional square and triangular lattices, and perform training and validation. In the equivariant model for the square lattice, the validation error (based on root mean squared error) is reduced to less than one-third compared to a model using invariant descriptors as inputs. Furthermore, I demonstrate the ability to reproduce phase transitions of skyrmion crystals in the triangular lattice, by performing dynamics simulations using the trained model.