Accurate Simulations of Water and Aqueous Solutions through Fine-Tuned Dispersion-Corrected Density Functional Theory...

Area of Science:

• Computational Chemistry
• Materials Science
• Physical Chemistry

Background:

• Dispersion-corrected density functional theory (DFT-D) is crucial for modeling large molecular systems and developing machine-learning interatomic potentials (MLIPs).
• Selecting appropriate DFT-D models for high accuracy across various properties and conditions remains challenging.
• MLIPs enable reliable molecular dynamics (MD) simulations of condensed-phase systems.

Purpose of the Study:

• To develop an effective computational strategy for enhancing standard DFT-D models.
• To create high-fidelity MLIPs for accurate prediction of molecular system properties.
• To accurately model the behavior of water and aqueous solutions using the developed MLIPs.

Main Methods:

• Developed a novel computational strategy to improve DFT-D accuracy.
• Derived a new MLIP for water using the enhanced strategy.
• Applied the MLIP to simulate water in various forms (clusters, liquid, ice) and aqueous solutions (MgCl2 in water).

Main Results:

• The novel MLIP accurately predicts properties of water, including radial distribution functions, enthalpies, diffusion constants, and density isobars.
• The MLIP captures water's anomalous behavior, often missed by standard first-principle MD simulations.
• MLIPs for MgCl2 in water accurately predict metal ion hydration and water exchange dynamics, showing improved agreement with experimental data.

Conclusions:

• The proposed computational strategy effectively enhances DFT-D accuracy and enables the development of high-fidelity MLIPs.
• The derived MLIPs provide accurate predictions for water and aqueous solutions, outperforming standard DFT-D and classical force fields.
• This approach offers a reliable method for simulating complex condensed-phase systems and their properties.