Dynamically polarisable force-fields for surface simulations via multi-output classification Neural Networks

TL;DR

This paper introduces a neural network-based method to incorporate electronic polarization into classical molecular dynamics force-fields, improving surface simulation accuracy especially for systems like graphene electrodes in electrolytes.

Contribution

The paper presents a novel multi-output neural network approach that treats surface polarization as a classification problem, enabling versatile and accurate polarization modeling in MD simulations.

Findings

Achieved high accuracy in polarization prediction for surface systems.

Validated model against quantum mechanical simulations with good agreement.

Demonstrated effectiveness in simulating graphene-electrolyte interfaces.

Abstract

We present a general procedure to introduce electronic polarization into classical Molecular Dynamics (MD) force-fields using a Neural Network (NN) model. We apply this framework to the simulation of a solid-liquid interface where the polarization of the surface is essential to correctly capture the main features of the system. By introducing a multi-input, multi-output NN and treating the surface polarization as a discrete classification problem, for which NNs are known to excel, we are able to obtain very good accuracy in terms of quality of predictions. Through the definition of a custom loss function we are able to impose a physically motivated constraint within the NN itself making this model extremely versatile, especially in the modelling of different surface charge states. The NN is validated considering the redistribution of electronic charge density within a graphene based electrode in contact with aqueous electrolyte solution, a system highly relevant to the development of next generation low-cost supercapacitors. We compare the performances of our NN/MD model against Quantum Mechanics/Molecular dynamics simulations where we obtain a most satisfactorily agreement.