(427b) Combining Molecular Simulations and Theory for Predicting the Binary Interaction Parameters of the NRTL Model

Ravichandran, A., Texas Tech University
Khare, R., Texas Tech University
Chen, C. C., Texas Tech University
Tun, H., Texas Tech University
Predicting the thermodynamic properties of liquid mixtures is a fundamental step involved in the design of industrial processes. Classical thermodynamic models based on local composition ideas are widely applied for property calculations and have several advantages such as ease of applicability and reliable extrapolation. However, such models are not completely predictive, and rely on experimental data for correlating the binary molecular interaction parameters (the input to these models), which in turn, are further used for predicting the thermodynamic properties of multicomponent systems. Hence, the ability to calculate the molecular interaction parameters from statistical mechanical theories and simulations is of substantial interest. We describe a novel technique that combines two-fluid theory and molecular simulations for predicting these binary interaction parameters. The thermodynamic model of interest in this study is the non-random two-liquid (NRTL) model. We validate our technique by predicting the interaction parameters of a dozen of binary mixtures involving components such as water, methanol, n-hexane, and chloroform. These molecules vary widely in their characteristics (i.e. hydrophilic, hydrophobic, and polar) and are representative of hundreds of commonly used industrial solvents. The binary parameters are expressed in terms of the molecular sizes, the sizes of the neighbor shells, and the net interaction energies which are directly obtained from molecular simulations. In each case, the phase behavior of the binary mixture as predicted from the NRTL model using the interaction parameters obtained from molecular simulations is compared with that predicted by using the parameter values derived from data regression. We show that the proposed approach can predict the binary interaction parameters with a reasonable accuracy.