Equivariant Neural Networks for Spin Dynamics Simulations of Itinerant Magnets

Abstract

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.

21 pages, 7 figures

References in corpus (13)

• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• Towards Universal Neural Network Potential for Material Discovery Applicable to Arbitrary Combination of 45 Elements
• Nanometric square skyrmion lattice in a centrosymmetric tetragonal magnet
• Imaging the coupling between itinerant electrons and localised moments in the centrosymmetric skyrmion magnet GdRu2Si2
• Square skyrmion crystal in centrosymmetric itinerant magnets
• e3nn: Euclidean Neural Networks
• Nanometric skyrmion lattice from anisotropic exchange interactions in a centrosymmetric host
• High-field depinned phase and planar Hall effect in skyrmion-host Gd${}_2$PdSi${}_3$
• Locking of skyrmion cores on a centrosymmetric discrete lattice: onsite versus offsite
• Low-Energy Excitations of Skyrmion Crystals in a Centrosymmetric Kondo-Lattice Magnet: Decoupled Spin-Charge Excitations and Nonreciprocity
• Arrested phase separation in double-exchange models: machine-learning enabled large-scale simulation
• Temperature-driven transition from skyrmion to bubble crystals in centrosymmetric itinerant magnets
• Machine learning nonequilibrium electron forces for adiabatic spin dynamics

Cited by in corpus (2)

• Machine learning for structure-property relationships: Scalability and limitations
• Deep learning methods for Hamiltonian parameter estimation and magnetic domain image generation in twisted van der Waals magnets