Thermodynamic properties predictions using the COSMO-SAC solvation method

The estimation of the thermodynamic properties of pure substances and their mixtures is essential in the field of chemical engineering. There are simple empirical or semi-empirical classical thermodynamic models for prediction of physical properties, such as cubic equations of state, activity coefficient models, and group contribution methods. These classical thermodynamic models are popular as a result of their simple implementation and fast calculations. The limitations are that these models depend on the availability of experimental data to obtain parameter values, so that their accuracy relies on the range and accuracy of the underlying experimental data and their accuracy. There is a need to develop more fundamental molecular-based theories. A complete description of the molecule and its environment is one of the challenges in thermodynamics. The subject of this dissertation is to make a contribution to the connection between fundamental molecular theories and the macroscopic physical properties of chemicals and their mixtures.;The approach we use in this dissertation is examining the solvation process by calculating the free energy, which is referred to as the transfer of a molecule from the gas phase to a solvent. In the solvation process, the solvent is treated as a continuum homogeneous medium, which has the advantage of fewer degrees of freedom and much faster computational time than using an explicit solvent model. The conductor-like screening model (COSMO) is a novel continuum solvation model to calculate the solvation free energy and therefore the macroscopic thermodynamic properties. The COSMO-SAC (segment activity coefficient) model is a variation of the COSMO-RS (Real Solvent) model that has been used by Lin and Sandler to calculate activity coefficients and to successfully predict vapor-liquid phase diagrams. Here we first extend the original COSMO-SAC model to calculate properties such as vapor pressures, heats of vaporization and normal boiling temperatures for pure substances ranging from simple compounds to complex multi-functional chemicals. The results of the COSMO-SAC model with a small number of fixed parameters are comparable in terms of accuracy and computational time with the results of group contribution methods that require many more parameters.;Determining the correct conformation of a molecule in the COSMO solvation theories is important, especially for flexible molecules. A straightforward way to select the structure of a solute molecule is by choosing the geometry with lowest energy, though this can be difficult if there are many conformations with different structures but similar energies. Another important issue is the essential approximation in COSMO theory that the structure of the solute is unchanged in going from the gas phase to the perfect conductor phase, and then to the real solvent. Here we have tested this assumption using COSMO-SAC for the ensembles of liquid structures from Monte Carlo simulation to see how changes in the phase (and therefore the local environment) in going from a gas to a liquid result in differences in molecular conformations, and how these affect thermodynamic property predictions.;Finally, we have unified the original and extended COSMO-SAC model into a single universal model with re-optimized parameters. This new COSMO-SAC then is used to predict thermodynamic properties for more than 1000 pure substances and more than 500 mixtures, including compounds that hydrogen bond. Only four universal parameters are used in the mixture property predictions, compared with hundreds of fitting parameters in the UNIFAC models. This unified model is then applied to new property predictions as an a priori method, that is, without fitting data for these new applications. First, octanol/water partition coefficients are predicted and compared with experimental data and with the results of simulations and regression for nearly a hundred commercial pharmaceutical compounds. In addition, the method is successfully used for ionic liquids although the COSMO-SAC model had been developed and optimized only for neutral, non-ionizing compounds. The Henry's law constants for a series of gas solutes in one ionic liquid are predicted using the COSMO-SAC model and compared with experimental data. The reasonably accurate results for these two new applications indicate the great potential of COSMO-SAC as a powerful a priori prediction method for thermodynamic property estimation.