We present the theory and implementation of a Poisson-Boltzmann implicit solvation model for electrolyte solutions. This model can be combined with arbitrary electronic structure methods that provide an accurate charge density of the solute. A hierarchy of approximations for this model includes a linear approximation for weak electrostatic potentials, finite size of the mobile electrolyte ions and a Sternlayer correction. Recasting the Poisson-Boltzmann equations into Euler-Lagrange equations then significantly simplifies the derivation of the free energy of solvation for these approximate models. The parameters of the model are then either fit directly to experimental observables -e.g. the finite ion size -or optimized for agreement with experimental results. Experimental data for this optimization is available in the form of Sechenov coefficients that describe the linear dependence of the salting-out effect of solutes with respect to the electrolyte concentration. In the final part we rationalize the qualitative disagreement of the finite ion size modification to the Poisson-Boltzmann model with experimental observations by taking into account the electrolyte concentration dependence of the Stern layer. A route towards a revised model that captures the experimental observations while including the finite ion size effects is then outlined. This implementation paves the way for the study of electrochemical and electrocatalytic processes of molecules and cluster models with accurate electronic structure methods.
I. MotivationIn contrast to the ab initio calculation of chemical processes in the gas phase, the investigation of such processes in solution adds an additional layer of tremendous complexity. The electrostatic interaction between solvent and solute directly alters the enthalpic landscape whereas osmotic pressure and reorientation of the solvent molecules based on the electrostatic interactions has a profound influence on the free energy. Although these effects can in principle be calculated with molecular dynamics (MD) simulations, the amount of sampling required to converge the free energy combined with the large number of structures involved in chemical processes for which such a sampling would have to be performed, renders this approach unfeasible at present. Hence, implicit solvation models, rather than explicit ones, are usually the method of choice in computational quantum chemistry. Famous examples of the former are the polarizable continuum model (PCM) 1-4 or the conductor like screening model (COSMO), 5 that exist in various flavours and specializations, are available in most multi-purpose electronic structure packages and have proven suitable in countless applications. 6,7 For the description of electrochemical processes, however, the effect of the electrolyte ions needs to be taken into account. In the realm of electrocatalysis, the phenomenological a)