No abstract

The present work examines the accuracy of the SPEADMD molecular simulation methodology in correlating experimental data relative to a standard low-pressure database for testing VLE models. The database contains 104 binary systems categorized according to polarity and ideality. Although the database is somewhat small, it covers a broad range of chemical functionality, including halocarbons and carboxylic acids as well as hydrocarbons and alcohols. Six models were tested and compared for their characterization of these mixtures. Four standard models were evaluated to establish a basis for comparison: the Margules, NRTL, PR, and PRWS models. The SPEADMD model was evaluated in three forms. In its elementary form, the SPEADMD model includes ∼10% deviations in vapor pressure because of the application of transferable potential functions in the molecular model. An alternative model is developed on the basis of SPEADMD combined with corrected vapor pressures and customized self-interaction parameter for pure compounds. This alternative is referred to as the SPEADCI model, in which CI stands for customized interactions. Results show that SPEADCI model provides accuracy similar to the NRTL and PRWS models, even though it includes only one adjustable parameter per binary system, whereas the NRTL model includes two and the PRWS models include three. Deviations in correlated bubble point pressure are roughly 1-2% for these models. The SPEADMD models have the advantage that transferable potentials can be applied for solvation interactions that are similar to the Kamlet-Taft interaction parameters.

No abstract

Vapor pressure and liquid density are used to characterize step potentials for fluorinated, chlorinated, brominated, and iodinated hydrocarbons, along with a variety other compounds, bringing the transferable database to 339 training compounds, 112 of which are added in this manuscript, and 25 "validation" compounds. The potentials were characterized by four-step potentials consistent with those of previous studies for the SPEADMD model. Vapor pressure deviations average near 10 % for most compounds in the training set and near 40 % for the validation set. Higher deviations appear in the validation set for compounds in which multiple functionalities are located in close proximity, indicating sensitivity to the transferability assumption in these cases. Deviations in liquid density approach 4 %, despite the large shifts in density caused by the relatively heavy halogenated atoms. The availability of transferable potentials for so many compounds sets the stage for systematic studies of phase behavior over a broad range of molecular types. In the context of this study, several key elements were identified for organizing the physical property database, simulation results, and analytical tools to infer optimal characterizations of the molecular interactions. The physical property database must be critically evaluated to eliminate extrapolations and ambiguous data. The directory structure must be flexible and extensible to accommodate continuous improvement as more data and more compounds are incorporated into the analysis. Finally, an efficient methodology must be implemented to permit optimal characterization of the molecular interactions in a reasonable time on a continuing basis. The methodology presented in this paper permits a fresh optimization of the entire database in roughly 12 h.