Water plays a fundamental role in life on earth and is an active participant in many physical, chemical, and biological processes. Modeling water in condensed phases is an indispensable part of modern water research and rigid non-polarizable water models, such as TIP4P/2005, have been very popular in molecular simulations due to their high efficiency. Although these water models can reproduce many properties of water, they fail in predicting the dielectric properties of water, such as the dielectric constant and low-frequency infrared spectra. Machine learning (ML) techniques are widely used in recent chemistry research since it can significantly reduce the computational cost of quantum mechanics. We propose to improve these models by modifying their dipole moment surfaces based on ML with quantum chemistry accuracy. We have proposed a protocol that can assign point charges of water molecules in liquid phase properly. A neural network is trained with the charges obtained from the protocol. The dielectric constant and low-frequency infrared spectra are computed based the charges predicted by the neural network. Good agreement is achieved between available experimental data and the calculated results based on our ML model. The effects of polarization and charge transfer are also be discussed.

Machine learning based point charges predict the dielectric constant and low-frequency infrared spectra for liquid water

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.