Flexible Polarizable Water Model Parameterized via Gaussian Process Regression.

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Physical chemistry

Background:

• Water is a fundamental component in molecular dynamics (MD) simulations.
• Accurate water models are essential for studying chemical processes.
• Existing models may lack flexibility or polarization for certain applications.

Purpose of the Study:

• To develop a novel, flexible, and polarizable water model for MD simulations.
• To improve the efficiency and accuracy of simulating water properties.
• To enable direct calculation of vibrational spectra and modeling of chemical reactions.

Main Methods:

• Gaussian process regression was employed to predict water model properties.
• An extended Lagrangian scheme with Drude oscillators was utilized for efficiency.
• The SWM4/Fw model was developed and validated against reference properties.

Main Results:

• The SWM4/Fw model demonstrated high accuracy in reproducing reference water properties.
• The model's flexibility is critical for simulating bond stretching/breaking in reactions.
• The model allows for direct calculation of vibrational spectra.

Conclusions:

• SWM4/Fw is an efficient, flexible, and polarizable water model suitable for MD simulations.
• The model's performance is comparable to or exceeds existing models like SWM4-NDP.
• This development advances the simulation of complex chemical systems involving water.