Multistep Explicit Solvation Protocol for Calculation of Redox Potentials

The calculation of molecular redox potentials in aqueous solution presents a challenge to quantum chemistry due to the need to calculate charged, open-shell species experiencing large solvent effects. Traditionally, redox potentials are calculated via the use of density functional theory and continuum solvation methods, but such protocols have been found to often suffer from large errors, particularly in the case of aqueous solution. While explicit solvation models hold promise of higher accuracy to describe solvent effects in general, their complicated use and lack of well-defined, reliable protocols has hindered their adoption. In this study, we present an explicit-solvation-based approach for the calculation of molecular redox potentials. We combine the use of affordable semiempirical QM/MM molecular dynamics (making use of the recently proposed GFN-xTB method by Grimme et al.) for both redox states and use the linear response approximation to relate vertical ionization energies to the adiabatic redox potential. Simulation length, averaging over snapshots, and accounting for bulk and polarization effects are systematically evaluated using phenol as a working example. We find that it is crucial to reliably account for bulk solvation effects in these calculations, as well as polarization effects which we divide up into short-range and long-range contributions. The short-range polarization contribution is accounted for via QM-region expansion, while the long-range contribution is accounted for via Drude-polarizable QM/MM. Our multistep protocol has been coded to be used in a fully automatic way in a local version of Chemshell. It has been evaluated on a test set of oxidation potentials of organic molecules and found to give gas-solution redox shifts with a mean absolute error of 0.13 eV with respect to experiment, compared to mean absolute errors of 0.26 and 0.21 eV with CPCM and SMD continuum models, respectively.