Abstract:Coarse-graining is a molecular modeling technique in which an atomistic system is represented in a simplified fashion that retains the most significant system features that contribute to a target output, while removing the degrees of freedom that are less relevant. This reduction in model complexity allows coarse-grained molecular simulations to reach increased spatial and temporal scales compared to corresponding all-atom models. A core challenge in coarse-graining is to construct a force field that represents the interactions in the new representation in a way that preserves the atomistic-level properties. Many approaches to building coarse-grained force fields have limited transferability between different thermodynamic conditions as a result of averaging over internal fluctuations at a specific thermodynamic state point. Here, we use a graph-convolutional neural network architecture, the Hierarchically Interacting Particle Neural Network with Tensor Sensitivity (HIP-NN-TS), to develop a highly automated training pipeline for coarse grained force fields which allows for studying the transferability of coarse-grained models based on the force-matching approach. We show that this approach not only yields highly accurate force fields, but also that these force fields are more transferable through a variety of thermodynamic conditions. These results illustrate the potential of machine learning techniques such as graph neural networks to improve the construction of transferable coarse-grained force fields.

What problem does this paper attempt to address?

This paper discusses the thermodynamic transferability issue of coarse-grained (Coarse-Graining) force fields in molecular modeling. Coarse-graining is a method of simplifying atomic systems by retaining key characteristics and removing less relevant degrees of freedom to achieve simulations at larger spatial and temporal scales. The paper proposes a novel approach based on Graph Convolutional Neural Network (GCNN) called Hierarchical Interaction Particle Neural Network with Tensor Sensitivity (HIP-NN-TS) for constructing a highly automated coarse-grained force field training pipeline. This method not only generates high-precision force fields but also exhibits better transferability under various thermodynamic conditions. The research found that the transferability of force fields constructed by traditional coarse-graining techniques is limited under different thermodynamic conditions due to the neglect of internal dynamics' dependence on thermodynamic states. The paper improves the construction of transferable coarse-grained force fields through machine learning techniques, especially GCNN. Experimental results demonstrate that this method enhances the transferability of force fields during temperature variations while maintaining accuracy, and it can handle different molecular systems. The research methodology includes data generation, model architecture (using HIP-NN-TS), training process, and evaluation methods. By conducting comparative experiments, the paper compares the proposed machine learning coarse-graining (ML CG) model with the model established by open-source software OpenMSCG, demonstrating the superiority of ML CG and its better cross-condition transferability in certain conditions. Additionally, the performance of single-state point models and multi-state point models under temperature and density variations is studied. It is found that these models still possess considerable transferability even when the system structure undergoes significant changes.

Thermodynamic Transferability in Coarse-Grained Force Fields using Graph Neural Networks

A Thermodynamics-Informed Neural Network for Elastoplastic Constitutive Modeling of Granular Materials

Transferable Coarse Graining Via Contrastive Learning of Graph Neural Networks

Temperature-transferable coarse-graining of ionic liquids with dual graph convolutional neural networks

A Scalable Graph Neural Network Method For Developing An Accurate Force Field Of Large Flexible Organic Molecules

Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture

Simulate Time-integrated Coarse-grained Molecular Dynamics with Multi-Scale Graph Networks

Accurate and scalable multi-element graph neural network force field and molecular dynamics with direct force architecture

Evaluating the Transferability of Machine-Learned Force Fields for Material Property Modeling

Generalizability of Graph Neural Network Force Fields for Predicting Solid-State Properties

Machine Learning of coarse-grained Molecular Dynamics Force Fields

Transfer learning for atomistic simulations using GNNs and kernel mean embeddings

Top-down machine learning of coarse-grained protein force-fields

PhyNEO: A Neural-Network-Enhanced Physics-Driven Force Field Development Workflow for Bulk Organic Molecule and Polymer Simulations.

Coarse-Graining with Equivariant Neural Networks: A Path Towards Accurate and Data-Efficient Models

Graph neural network coarse-grain force field for the molecular crystal RDX

Transferable Implicit Solvation Via Contrastive Learning of Graph Neural Networks

Multi-Fidelity Physics-Constrained Neural Network and Its Application in Materials Modeling

Symmetry-adapted graph neural networks for constructing molecular dynamics force fields

Transferability of datasets between Machine-Learning Interaction Potentials