Bayesian Optimization of the Coarse-Grained Force Field for Ions and Ionic Surfactants Based on a Polarizable Water Model

Developing precise and transferable coarse-grained (CG) force fields (FFs) for ions and ionic surfactants in solution is challenging because solutions always contain ions with opposite charges, necessitating the simultaneous determination of multiple interaction parameters. In this study, we introduce an efficient workflow that combines global Bayesian optimization with local meta -multilinear interpolation parametrization and latin hypercube sampling to optimize CG FFs for NaCl, NaBr, and surfactants sodium dodecyl sulfate (SDS) and dodecyl trimethylammonium chloride/bromide (DTAC/DTAB). This approach is based on a polarizable CG water model, where van der Waals interactions are described using a piecewise Morse potential. The developed force fields accurately reproduced properties such as density, surface tension, dielectric constant, and osmotic pressure of salt solutions across a wide concentration range. They also successfully captured the micelle structures formed by SDS and DTAC/DTAB, along with the surface tensions at the solution/air interfaces. This work lays the groundwork for developing force fields for more complex and charged molecules.