The COSMO-RS (COnductor like Screening MOdel for Realistic Solvents) was developed by Klamt and coworkers [1-3]. There are different implementations of COSMO-RS or derivatives, and different parametrizations. The method used in ADF is the one developed by Klamt et al., which is described in detail in Ref. [2], and is called here COSMO-RS. The implementation of COSMO-RS in ADF is described in Ref. [4]. In ADF it is possible to use a thermodynamically consistent combinatorial contribution to the chemical potential as is used in Ref. [3], and a temperature dependent hydrogen bond interaction, also described in Ref. [3]. The parameters in the paper [2] were reparametrized for ADF, see Ref. [4] for details. The ADF COSMO-RS command line program is called crs. The main authors of this program are Cory Pye (Saint Mary's University, Halifax NS Canada) and Jaap Louwen (Albemarle Corporation). The COSMO-RS GUI ADFcrs contains an input builder for COSMO-RS and can visualize results, see the COSMO-RS GUI tutorial and the COSMO-RS GUI reference manual.
COSMO-RS uses the intermediate results from quantum mechanical (QM) calculations on individual molecules to predict thermodynamic properties of mixtures of these molecules, for example, solubility. There are a fair number of reports of accurate prediction by COSMO-RS of thermodynamic properties in general in the literature. Many of these have been written by Klamt and co-workers, see Ref. [3] and references therein.
There are also empirical methods like UNIFAC that can predict thermodynamic properties (including the activity coefficients). These methods contain group specific parameters and are parametrized against a large data base. They will often do better than COSMO-RS methods (especially, of course, if the system of interest was part of the data base used for parameter estimation). However, these methods cannot handle every type of molecule. In particular when unusual combinations of functional groups occur (such as in drug molecules), no parametrization is available. COSMO-RS methods, on the other hand, only feature general parameters not specific to chemical groups or functionalities. All that is required is that a quantum mechanical calculation can be done on the molecule. Therefore, COSMO-RS can be a valuable tool for the prediction of chemical engineering thermodynamical properties, like, for example, partial vapor pressures, solubilities, and partition coefficients. An additional advantage of COSMO-RS over empirical methods is that the molecules dissolved may in fact be transition states of a chemical reaction. This follows from the fact that all that is required is that one can do a QM calculation on the solute and QM on a transition state has become standard in the last two decades. This affords a unique opportunity to predict the thermodynamics of a reaction including, for instance, the balance between kinetically and thermodynamically favored reaction pathways as a function of the solvent used.
