Computer modeling and its application to problems in fluid phase equilibria

Models for predicting fluid phase equilibria are an integral part of separations process design. They allow the engineer to examine the performance of a proposed separation process, the effects of changes in process conditions on the system, and they provide a basis for understanding the equilibrium behavior of the system's chemical components. In order to improve the separation design process, it is important to continue to advance the capabilities of the models. The goal of this study is to explore and improve the modeling tools needed for correlating and predicting fluid phase equilibria. This goal was achieved by evaluating the performance of the existing activity coefficient models, developing a method for solving the Ornstein-Zernike equations for three components, examining the origin of the n {dollar}to{dollar} {dollar}pisp*{dollar} UV shift of benzophenone in supercritical ethane/trifluoroethanol in the context of radial distribution functions, and developing two LSER's: one for correlating partial molar heats of transfer and one for correlating infinite dilution activity coefficients of nonelectrolyte solutes in water. The first subject involves modeling multicomponent phase equilibria from binary data for a reacting system using the NRTL and UNIQUAC models. While multicomponent phase-equilibrium data are desired for design of separation equipment, such measurements often cannot be made for reacting systems due to chemical reaction. The second and third subjects involve the development of a stepwise direct iteration method for solving the Ornstein-Zernike equations for a ternary system of Lennard-Jones molecules and its application to modeling the environment around a benzophenone molecule in supercritical ethane/trifluoroethanol. The method allows the calculation of local densities, compositions, and thermodynamic properties a priori. The last two subjects describe multiple linear regression models for correlating {dollar}overline{lcub}hsb{lcub}TR{rcub}{rcub}{dollar} and {dollar}gammaspinfty{dollar} of solutes in water. The models use solvatochromic parameters {dollar}alpha{dollar}, {dollar}beta{dollar}, {dollar}pisp*{dollar}, molar volume, and other solute properties as regressors. The former LSER correlates {dollar}overline{lcub}hsb{lcub}TR{rcub}{rcub}{dollar} to within 4% which corresponds to a 28% error in {dollar}overline{lcub}hsp{lcub}Einfty{rcub}{rcub}{dollar}. The latter LSER correlates the solute {dollar}gammaspinfty{dollar} values to within an average absolute error of 0.294 {dollar}lngammaspinfty{dollar}.