Modeling solvation dynamics of transition metal redox ion through on-the-fly multi-objective Bayesian-optimized force...

Area of Science:

• Computational chemistry
• Materials science
• Physical chemistry

Background:

• Accurate modeling of solvation dynamics and properties is essential for designing advanced electrolytes for energy storage and conversion.
• Current force field parameterization methods can be computationally intensive and may not fully capture complex solvation behaviors.

Purpose of the Study:

• To develop and validate an on-the-fly multi-objective Bayesian optimization (OTF-MOBO) method for efficient and accurate force field parameterization.
• To improve the prediction of ionic solvation structures, thermodynamics, and transport properties using molecular dynamics simulations.

Main Methods:

• Implemented an OTF-MOBO algorithm to optimize force field parameters using solvation-free energy and solvation radii as training data.
• Employed molecular dynamics simulations to evaluate modeling accuracy, targeting the Pareto front for minimized prediction errors.
• Utilized simple force fields (Lenard-Jones and Coulombic potentials) for modeling aqueous solutions of transition metal redox ions.

Main Results:

• Achieved relative errors below 2% in predicting both solvation-free energy and solvation radii for transition metal ions.
• Demonstrated the effectiveness of the OTF-MOBO method in optimizing force fields for accurate solvation modeling.
• Validated the optimized force fields by successfully predicting solvation entropy and diffusivities with relative errors below 10% compared to experimental data.

Conclusions:

• The developed OTF-MOBO method provides an efficient and accurate approach for parameterizing force fields for ionic solvation.
• Simple force fields, when optimized using this data-driven method, can achieve high predictive power for electrolyte properties.
• This work advances the development of electrolytes for electrochemical devices by improving the reliability of molecular simulations.