(237b) Prediction of Vapor/Liquid Characteristics By Quantum Mechanics Data

A large number of chemical processes involve the purification of mixtures in industry, such as liquid extraction, distillation, and recrystallization. Infinite dilution coefficients can be used to characterize the vapor-liquid, solid liquid, and liquid-liquid phase behavior that underlies these unit operations. The largest deviation from ideality occurs in the dilute solution system. The ability to predict infinite dilution coefficients based on simple guidelines of hydrogen bonding could provide valuable design formulations and useful thermodynamic insights. The MOSCED model is sufficiently simple to fill this need. Unfortunately, a lack of characterized parameters for many compounds limits the capabilities of the MOSCED model. The goal of this paper is to develop a simple method to estimate the missing parameters based on readily available quantum mechanical results.

Sigma profiles have been calculated using quantum mechanics and tabulated at a website hosted by Virginia Tech. The charge density of the molecule over the molecular surface is described by the values where the projected surface charge is opposite of that of the molecule. Hydrogen bonding can occur when the charge density exceeds a threshold value. These observations can be used to calculate acidity and basicity parameters for the MOSCED model. Given that the bonding starts at a specific point it was assumed the rest of the contributions could be interpreted as another parameter, polarizability factor.

The acidity and basicity parameters were characterized by fitting the data from the area beneath the sigma profile at a certain distance and correlating the values to pre-determined values from Lazzaroni et al. The fit for the polarizability factor was determined by using a weighted function to give greater emphasis on values further from the center of the sigma profile.

Correlations of acidity and basicity were characterized based on experimental data for roughly 4500 binary solutions. A comparison of the results for the fit vs. the MOSCED with Lazzaroni parameter values and UNIFAC model for the solutions was used to determine the accuracy of the model. Several specific solutions were poorly characterized in comparison to the MOSCED model, but the correlation provides qualitative estimates of activity coefficients with in roughly 0.280 log units of the experimental data, compared to 0.106 log units for the original MOSCED method and 0.183 log units for UNIFAC.