Ions in solution play a fundamental role in many physical, chemical, and biological processes. For industrial applications these systems are usually described using simple analytical models which are fitted to reproduce the available experimental data. In this work, we propose a multi-scale coarse graining procedure to derive such models from atomistic descriptions. First, parameters for classical force-fields of ions in solution are extracted from ab-initio calculations. Effective (McMillan-Mayer) ion-ion potentials are then derived from radial distribution functions measured in classical molecular dynamics simulations, allowing us to define an implicit solvent model of electrolytes. Finally, perturbation calculations are performed to define the best possible representation for these systems, in terms of charged hard-sphere models. Our final model is analytical and contains no free "fitting" parameters. It shows good agreement with the exact results obtained from Monte-Carlo simulations for the thermodynamic and structural properties. Development of a similar model for the electrolyte viscosity, from information derived from atomistic descriptions, is also introduced.
