Towards a converged strategy for including microsolvation in reaction mechanism calculations.

A major part of chemical conversions is carried out in the fluid phase, where an accurate modeling of the involved reactions requires to also take into account solvation effects. Implicit solvation models often cover these effects with sufficient accuracy but can fail drastically when specific solvent-solute interactions are important. In those cases, microsolvation, i.e., the explicit inclusion of one or more solvent molecules, is a commonly used strategy. Nevertheless, microsolvation also introduces new challenges-a consistent workflow as well as strategies how to systematically improve prediction performance are not evident. For the COSMO and COSMO-RS solvation models, this work proposes a simple protocol to decide if microsolvation is needed and how the corresponding molecular model has to look like. To demonstrate the improved accuracy of the approach, specific application examples are presented and discussed, i.e., the computation of aqueous pKa values and a mechanistic study of the methanol mediated Morita-Baylis-Hillman reaction.