High-Dimensional Neural Network Potentials for Magnetic Systems Using Spin-Dependent Atom-Centered Symmetry Functions

TL;DR

This paper introduces spin-dependent atom-centered symmetry functions as descriptors for neural network potentials, enabling accurate modeling of magnetic systems and their different spin states, demonstrated on MnO.

Contribution

The paper presents a novel spin-dependent descriptor for neural network potentials, allowing accurate simulation of magnetic states in materials.

Findings

Accurately predicts magnetic structures of MnO

Reproduces experimental and DFT data for MnO

Enables calculation of Néel temperature considering fluctuations

Abstract

Machine learning potentials have emerged as a powerful tool to extend the time and length scales of first principles-quality simulations. Still, most machine learning potentials cannot distinguish different electronic spin orientations and thus are not applicable to materials in different magnetic states. Here, we propose spin-dependent atom-centered symmetry functions as a new type of descriptor taking the atomic spin degrees of freedom into account. When used as input for a high-dimensional neural network potential (HDNNP), accurate potential energy surfaces of multicomponent systems describing multiple magnetic states can be constructed. We demonstrate the performance of these magnetic HDNNPs for the case of manganese oxide, MnO. We show that the method predicts the magnetically distorted rhombohedral structure in excellent agreement with density functional theory and experiment. Its efficiency allows to determine the N\'{e}el temperature considering structural fluctuations, entropic effects, and defects. The method is general and is expected to be useful also for other types of systems like oligonuclear transition metal complexes.