Abstract:Polymer simulation with both accuracy and efficiency is a challenging task. Machine learning (ML) forcefields have been developed to achieve both the accuracy of ab initio methods and the efficiency of empirical force fields. However, existing ML force fields are usually limited to single-molecule settings, and their simulations are not robust enough. In this paper, we present PolyGET, a new framework for Polymer Forcefields with Generalizable Equivariant Transformers. PolyGET is designed to capture complex quantum interactions between atoms and generalize across various polymer families, using a deep learning model called Equivariant Transformers. We propose a new training paradigm that focuses exclusively on optimizing forces, which is different from existing methods that jointly optimize forces and energy. This simple force-centric objective function avoids competing objectives between energy and forces, thereby allowing for learning a unified forcefield ML model over different polymer families. We evaluated PolyGET on a large-scale dataset of 24 distinct polymer types and demonstrated state-of-the-art performance in force accuracy and robust MD simulations. Furthermore, PolyGET can simulate large polymers with high fidelity to the reference ab initio DFT method while being able to generalize to unseen polymers.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenge of achieving both high - precision and high - efficiency simultaneously in polymer simulations. Traditional ab - initio methods (such as density functional theory DFT) can provide high - precision results, but the computational cost is very high, making it difficult to be applied to large - scale or long - time simulations. Empirical force fields, although computationally efficient, simplify the chemical interactions between atoms, resulting in limited precision and insufficient molecular dynamics (MD) simulation performance and chemical insight. To overcome these limitations, researchers have developed machine - learning (ML) force fields, aiming to combine the precision of ab - initio methods and the computational efficiency of empirical force fields. However, existing ML force fields are usually limited to single - molecule settings, and the simulations in multi - molecule settings are not robust enough. In addition, existing ML force fields usually optimize energy and force simultaneously during training, which may lead to conflicts between the targets and thus affect the generalization ability of the model. For this reason, this paper proposes PolyGET, a new framework based on the equivariant transformer for the generation of polymer force fields. The main contributions of PolyGET are as follows: 1. **Multi - molecule training and generalization**: PolyGET can be trained on diverse polymer families and can be generalized to larger - scale polymers that have not been seen before. 2. **Accurate and robust simulations**: PolyGET can generate reliable and accurate forces in molecular dynamics (MD) simulations, thereby improving the robustness of the simulations. Specifically, PolyGET achieves the above advantages through the following two innovative designs: 1. **Unified multi - molecule model**: Use the equivariant transformer as the backbone model to train a unified multi - molecule model, effectively capturing complex quantum interactions. 2. **New training paradigm**: Propose a new training paradigm that focuses on optimizing forces rather than optimizing energy and force simultaneously. This force - centered objective function avoids the conflict between the energy and force targets, enabling the model to be generalized among different polymer families. Through evaluation on Poly24, a large - scale dataset containing 24 different polymer types, PolyGET demonstrates state - of - the - art performance in force prediction and MD simulations and can be generalized to unseen large polymers.

PolyGET: Accelerating Polymer Simulations by Accurate and Generalizable Forcefield with Equivariant Transformer

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

Development of scalable and generalizable machine learned force field for polymers

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

A Transferrable Range-Separated Force Field for Water: Combining the Power of Both Physically-Motivated Models and Machine Learning Techniques

A Transferrable Coarse-Grained Force Field for Simulations of Polyethers and Polyether Blends

Efficient Machine Learning Force Field for Large-Scale Molecular Simulations of Organic Systems

From Peptides to Nanostructures: A Euclidean Transformer for Fast and Stable Machine Learned Force Fields

Parameterization of General Organic Polymers within the Open Force Field Framework

Data-Driven Parametrization of Molecular Mechanics Force Fields for Expansive Chemical Space Coverage

sGDML: Constructing Accurate and Data Efficient Molecular Force Fields Using Machine Learning

An Extendible, Graph-Neural-Network-Based Approach for Accurate Force Field Development of Large Flexible Organic Molecules

Accurate interatomic force field for molecular dynamics simulation by hybridizing classical and machine learning potentials

Machine-learned molecular mechanics force fields from large-scale quantum chemical data

Atomic permutationally invariant polynomials for fitting molecular force fields

Improving machine learning force fields for molecular dynamics simulations with fine-grained force metrics

On the design space between molecular mechanics and machine learning force fields

Force field optimization by end-to-end differentiable atomistic simulation

EL-MLFFs: Ensemble Learning of Machine Leaning Force Fields

Machine Learning of Accurate Energy-Conserving Molecular Force Fields

Accurate Machine Learned Quantum-Mechanical Force Fields for Biomolecular Simulations