Modeling the Behavior of Complex Aqueous Electrolytes Using Machine Learning Interatomic Potentials: The Case of...

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Materials Science

Background:

• Accurate modeling of solvated ions is crucial for electrochemistry, water treatment, and energy storage.
• Ab initio molecular dynamics methods offer high accuracy but are limited by computational cost.
• Classical force fields lack the necessary accuracy for complex ion interactions.

Purpose of the Study:

• To explore the structure and thermodynamics of complex monovalent-divalent ion pairs.
• To investigate sodium-sulfate ion pairing in aqueous solutions using a novel computational approach.
• To determine hydration structures and potentials of mean force for ion pairs at various concentrations.

Main Methods:

• Development and application of a machine learning interatomic potential (MLIP) trained on density functional theory (DFT) data.
• Simulation of sodium sulfate aqueous solutions (Na2SO4(aq)) across a concentration range of 0.1-2 M.
• Analysis of bulk properties, hydration structures, potentials of mean force, and minimum energy pathways.

Main Results:

• The MLIP accurately reproduced key bulk properties of water, including density and radial distribution functions.
• Detailed hydration structures of sodium and sulfate ions were obtained, along with potentials of mean force for ion pairing at 0.1 M.
• At low concentrations, strong sulfate solvation favors solvent-separated ion pairs over contact ion pairs; ion coordination is a sequential process.

Conclusions:

• Machine learning interatomic potentials provide a powerful tool to achieve DFT-level accuracy for studying complex solvated ions.
• This method enables investigations into ion pair thermodynamics and structures at concentration and time scales previously inaccessible.
• The findings advance our understanding of ion-water interactions, with implications for electrochemical and separation technologies.