Understanding the Dielectric Relaxation of Liquid Water Using Neural Network Potential and Classical Pairwise Potential

Jae Hyun Ryu,Ji Woong Yu,Tae Jun Yoon,Won Bo Lee

DOI: https://doi.org/10.26434/chemrxiv-2023-5jsc8-v3

2023-12-01

Abstract:Understanding the role of hydrogen bond networks in determining the relaxation dynamics is essential for understanding natural phenomena in liquid water. Classical pairwise additive models have been widely utilized for elaborating the underlying mechanism behind the relaxation phenomena. However, they have shown their limits due to either the absence or inaccurate descriptions of many-body and medium-to-long-range interactions. This work demonstrates that the Deep Potential Molecular Dynamics (DPMD) model trained with SCAN functional help calculate the dielectric constant at the accuracy of the first-principles simulations. The DPMD model outperforms the classical force fields (SPC/Fw and TIP4P/epsilon) in predicting dielectric spectra especially in replicating high-frequency excesses, attributed to its adeptness in simulating intricate hydrogen bond networks. Through a comprehensive analysis of the simulation results, it becomes evident that only the DPMD model effectively accommodates a wide range of hydrogen bond coordination scenarios thereby characterizing the intricate nature of the hydrogen bond network. This adaptability stems from the intricate interplay of many-body interactions and intramolecular dynamics. In addition, orientation defects within the DPMD model play a significant role in shaping the potential energy barrier due to the adaptability.

Chemistry

What problem does this paper attempt to address?

This paper mainly explores how to use Neural Network Potential and classical double potential to better understand the dielectric relaxation phenomenon of liquid water. The highlights of the research are as follows: 1. The dielectric spectra obtained from the simulation using Neural Network Potential are more consistent with experimental results, surpassing traditional force field methods. 2. The Neural Network Potential molecular dynamics simulation effectively captures the high-frequency excess in the dielectric spectra, which is related to the complexity of the hydrogen bond network. 3. The Neural Network Potential model exhibits significant differences in adapting to different hydrogen bond coordination environments, while the traditional force field performs poorly. 4. In the Neural Network Potential model, the influence of orientational defects on the energy barrier is particularly significant, which is related to the adaptability of the model. By comparing the simulation results of the Neural Network Potential model (DPMD, trained using the SCAN functional) with classical force fields (such as SPC/Fw and TIP4P/ε), the paper demonstrates that the DPMD model has higher accuracy in calculating the dielectric constant of liquid water, especially in reproducing high-frequency excess. In addition, the DPMD model can better simulate the dynamic behavior of hydrogen bond networks and adapt to various hydrogen bond coordination environments, thanks to its fine simulation of many-body interactions and molecular dynamics. The study also adopted the continuous relaxation time distribution method and discrete model regression to accurately calculate the dielectric constant.

Understanding the Dielectric Relaxation of Liquid Water Using Neural Network Potential and Classical Pairwise Potential

Understanding the Dielectric Relaxation of Liquid Water Using Neural Network Potential and Classical Pairwise Potential

Liquid water simulation using hydrogen bond corrected SCAN and neural network potentials.

Modeling Liquid Water by Climbing Up Jacob's Ladder in Density Functional Theory Facilitated by Using Deep Neural Network Potentials

Molecular dynamics simulation of dielectric properties of water

Molecular Dynamics Simulation of Zinc Ion in Water with an Ab Initio Based Neural Network Potential

Polarizable Water Model with Ab Initio Neural Network Dynamic Charges and Spontaneous Charge Transfer

Isotope effects in molecular structures and electronic properties of liquid water via deep potential molecular dynamics based on the SCAN functional

Dielectric Saturation in Water from a Long Range Machine Learning Model

Ab Initio Validation on the Connection Between Atomistic and Hydrodynamic Description to Unravel the Ion Dynamics of Warm Dense Matter

Isotope effects on molecular structures and electronic properties of liquid water via deep potential molecular dynamics based on SCAN functional

Condensed Phase Water Molecular Multipole Moments from Deep Neural Network Models Trained on Ab Initio Simulation Data

Many-body interactions and deep neural network potentials for water

Enhanced Deep Potential Model for Fast and Accurate Molecular Dynamics; Application to the Hydrated Electron

A deep potential model with long-range electrostatic interactions

Characterization of the Hydrogen-Bond Network in High-Pressure Water by Deep Potential Molecular Dynamics

Toward High-level Machine Learning Potential for Water Based on Quantum Fragmentation and Neural Networks

Raman Spectrum and Polarizability of Liquid Water from Deep Neural Networks

Enhanced deep potential model for fast and accurate molecular dynamics: application to the hydrated electron

Modelling infrared spectra of the O-H stretches in liquid H2O based on a deep learning potential, the importance of nuclear quantum effects