Abstract:We add the magnetic degrees of freedom to the widely used Gaussian Approximation Potential of machine learning (ML) and present a model that describes the potential energy surface of a crystal based on the atomic coordinates as well as their noncollinear magnetic moments. Assuming an adiabatic approximation for the spin directions and magnitudes, the ML model depends solely on spin coordinates and orientation, resulting in computational efffciency and enabling ab initio spin dynamics. Leveraging rotational symmetries of magnetic interactions, the ML model can incorporate various magnetic interactions, expanding into two-body, three-body terms, etc., following the spirit of cluster expansion. For simplicity, we implement the ML model with a two-body form for the exchange interaction. Comparing total energies and local fields predicted by the model for noncollinear spin arrangements with explicit results of constrained noncollinear density functional calculations for bcc Fe yields excellent results, within 1 meV/spin for the total energy. Further optimization, including three-body and other terms, is expected to encompass diverse magnetic interactions and enhance the model's accuracy. This will extend the model's applicability to a wide range of materials and facilitate the machine learning ab initio spin dynamics.

What problem does this paper attempt to address?

The main objective of this paper is to develop a machine learning method to simulate the ab initio process of spin dynamics in magnetic materials. Specifically, the authors propose a machine learning framework to describe the potential energy surface of non-collinear magnetic moments in crystals by extending the existing Gaussian Approximation Potential (GAP) model to include magnetic degrees of freedom. The paper addresses the following key issues: 1. **Extending the GAP model to include magnetic information**: By introducing a Machine Learning Exchange Field (ML-EF), the direction and magnitude of magnetic moments are incorporated into the GAP model, which originally only considered atomic positions. This enables the model to handle non-collinear magnetic structures. 2. **Improving computational efficiency**: By adopting the adiabatic approximation, which assumes that changes in spin direction are much slower than changes in spin magnitude, the model is simplified and computational efficiency is enhanced. 3. **Constructing descriptors**: A new descriptor, called "Smooth Overlap of Spin Orientations" (SOSO), is designed to capture spin orientations and their relationship with the surrounding environment. 4. **Validating model accuracy**: The model's accuracy is validated by comparing the total energy predicted by the model with the results of constrained non-collinear calculations based on Density Functional Theory (DFT). The results show that the prediction error in bcc iron is less than 1 meV/spin. In summary, this study aims to provide an efficient and accurate computational tool for the microscopic properties of magnetic materials by combining machine learning techniques with quantum mechanical principles. This approach is expected to enhance the understanding of complex magnetic phenomena in a wide range of material systems and accelerate the design and discovery of new materials.

Machine Learning Exchange Fields for Ab-initio Spin Dynamics

Machine learning force-field models for metallic spin glass

Machine learning assisted derivation of minimal low-energy models for metallic magnets

Machine learning force-field model for kinetic Monte Carlo simulations of itinerant Ising magnets

Interatomic Interaction Models for Magnetic Materials: Recent Advances

Learning QM/MM Potential using Equivariant Multiscale Model

Non-perturbative Many-Body Treatment of Molecular Magnets

Coarsening of chiral domains in itinerant electron magnets: A machine learning force field approach

Ab-Initio Approach to Many-Body Quantum Spin Dynamics

Machine-learning interatomic potentials for materials science

Machine Learning in QM/MM Molecular Dynamics Simulations of Condensed-Phase Systems

Machine Learning Magnetic Parameters from Spin Configurations

Machine Learning Interatomic Potentials and Long-Range Physics

Machine learning assisted determination of electronic correlations from magnetic resonance

A Machine Learning Approach to Decipher the Origin of Magnetic Anisotropy in Three-Coordinate Cobalt Single-Ion Magnets

Ab Initio Multireference Configuration-Interaction Theoretical Study On The Low-Lying Spin States In Binuclear Transition-Metal Complex: Magnetic Exchange Of [(Nh3)(5)Cr(Mu-Oh)Cr(Nh3)(5)](5+) And [Cl3feofecl3](2-)

Machine learning magnetism classifiers from atomic coordinates

Accurate machine learning force fields via experimental and simulation data fusion

De novo exploration and self-guided learning of potential-energy surfaces