Machine-Learning Interatomic Potentials for Long-Range Systems

2025-02-07

AI Summary

Traditional machine learning interatomic potentials (MLIPs) predominantly model only local atomic environments, failing to accurately capture long-range interactions—thus limiting their ability to achieve quantum-mechanical accuracy in molecular simulations. To address this, we propose SOG-Net, a novel Sum-of-Gaussians neural network architecture that jointly models short- and long-range contributions via latent-variable coupling and employs adaptive Gaussian superposition to represent diverse long-range decay behaviors. Furthermore, SOG-Net integrates Fourier convolutional layers with non-uniform fast Fourier transforms (NUFFT) to efficiently enforce physical long-range constraints. Validated across multiple systems exhibiting significant long-range effects—including ionic, dipolar, and dispersion-dominated materials—SOG-Net achieves quantum-mechanical accuracy in energy and force predictions while maintaining near-linear computational scaling. This work represents the first MLIP framework to unify high long-range fidelity with practical scalability.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Search and Optimization: Learning to Search
• Natural Language Processing: (Large) Language Models

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Machine-learning interatomic potentials have emerged as a revolutionary class of force-field models in molecular simulations, delivering quantum-mechanical accuracy at a fraction of the computational cost and enabling the simulation of large-scale systems over extended timescales. However, they often focus on modeling local environments, neglecting crucial long-range interactions. We propose a Sum-of-Gaussians Neural Network (SOG-Net), a lightweight and versatile framework for integrating long-range interactions into machine learning force field. The SOG-Net employs a latent-variable learning network that seamlessly bridges short-range and long-range components, coupled with an efficient Fourier convolution layer that incorporates long-range effects. By learning sum-of-Gaussian multipliers across different convolution layers, the SOG-Net adaptively captures diverse long-range decay behaviors while maintaining close-to-linear computational complexity during training and simulation via non-uniform fast Fourier transforms. The method is demonstrated effective for a broad range of long-range systems.

Problem

• Addressing neglect of long-range interactions in ML potentials.
• Proposing SOG-Net for integrating long-range effects efficiently.
• Capturing diverse long-range decay behaviors adaptively.

Innovation

• SOG-Net integrates long-range interactions
• Employs latent-variable learning network
• Uses efficient Fourier convolution layer