Abstract:In recent years, significant progress has been made in the development of machine learning potentials (MLPs) for atomistic simulations with applications in many fields from chemistry to materials science. While most current MLPs are based on environment-dependent atomic energies, the limitations of this locality approximation can be overcome, e.g., in fourth-generation MLPs, which incorporate long-range electrostatic interactions based on an equilibrated global charge distribution. Apart from the considered interactions, the quality of MLPs crucially depends on the information available about the system, i.e., the descriptors. In this work we show that including -- in addition to structural information -- the electrostatic potential arising from the charge distribution in the atomic environments significantly improves the quality and transferability of the potentials. Moreover, the extended descriptor allows to overcome current limitations of two- and three-body based feature vectors regarding artificially degenerate atomic environments. The capabilities of such an electrostatically embedded fourth-generation high-dimensional neural network potential (ee4G-HDNNP), which is further augmented by pairwise interactions, are demonstrated for NaCl as a benchmark system. Employing a data set containing only neutral and negatively charged NaCl clusters, even small energy differences between different cluster geometries can be resolved, and the potential shows an impressive transferability to positively charged clusters as well as the melt.

What problem does this paper attempt to address?

The main goal of this paper is to improve the performance of machine learning potentials (MLPs) in atomic-scale simulations, particularly addressing the limitations of fourth-generation machine learning potentials in describing long-range electrostatic interactions and enhancing the quality and transferability of potential energy surfaces. To achieve this, the research team proposed two improvement methods: 1. **Introduction of empirical two-body interaction terms**: By incorporating empirical two-body interaction terms based on the Tosi-Fumi model, the potential is improved to ensure stability when the structure significantly differs from the training data and to enhance the stability of the potential in cases of close atomic contact. 2. **Proposing the electrostatically embedded fourth-generation high-dimensional neural network potential (ee4G-HDNNP)**: By extending the input layer of the atomic neural network to include element-specific electrostatic potentials as additional inputs, this method aims to capture electronic structure information in the atomic environment. This approach addresses the inadequacy in describing atomic environments, thereby reducing accuracy degradation due to conflicting training data. Specifically, ee4G-HDNNP not only considers the local geometric environment of atoms but also utilizes local potentials derived from the global charge distribution, which helps better describe atomic interactions under different chemical bonding scenarios, including non-local effects caused by distant changes in the system, such as long-range charge transfer. Additionally, ee4G-HDNNP incorporates the long-range electrostatic interactions from the previously proposed fourth-generation potentials and the newly introduced empirical two-body interaction terms. The paper demonstrates that ee4G-HDNNP can accurately resolve the subtle energy differences between different cluster geometries using sodium chloride (NaCl) clusters as a benchmark system. It also shows good transferability for positively charged clusters and molten states. Furthermore, the performance of ee4G-HDNNP is validated through minimum hopping simulations of clusters with different sizes and total charges. Finally, the study explores the transferability of ee4G-HDNNP from clusters to periodic systems and compares the simulation results with density functional theory (DFT) calculations.

Accurate Fourth-Generation Machine Learning Potentials by Electrostatic Embedding

Accurate Fourth-Generation Machine Learning Potentials by Electrostatic Embedding

A Fourth-Generation High-Dimensional Neural Network Potential with Accurate Electrostatics Including Non-local Charge Transfer

Accelerating Fourth-Generation Machine Learning Potentials Using Quasi-Linear Scaling Particle Mesh Charge Equilibration

Accelerating fourth-generation machine learning potentials by quasi-linear scaling particle mesh charge equilibration

Embedded Atom Neural Network Potentials: Efficient and Accurate Machine Learning with a Physically Inspired Representation

Incorporating long-range electrostatics in neural network potentials via variational charge equilibration from shortsighted ingredients

Global Neural Network Potential with Explicit Many-Body Functions for Improved Descriptions of Complex Potential Energy Surface.

Assessment of Embedding Schemes in a Hybrid Machine Learning/Classical Potentials (ML/MM) Approach

A Hessian-Based Assessment of Atomic Forces for Training Machine Learning Interatomic Potentials

Density-Based Long-Range Electrostatic Descriptors for Machine Learning Force Fields

Data-Efficient Construction of High-Fidelity Graph Deep Learning Interatomic Potentials

High-Dimensional Neural Network Potential for Liquid Electrolyte Simulations.

Learning Atomic Multipoles: Prediction of the Electrostatic Potential with Equivariant Graph Neural Networks

Machine Learning Interatomic Potentials and Long-Range Physics

A deep potential model with long-range electrostatic interactions

Assessment of embedding schemes in a hybrid machine learning/classical potentials (ML/MM) scheme

Self-consistent Coulomb interactions for machine learning interatomic potentials

Synergistic integration of physical embedding and machine learning enabling precise and reliable force field