Estimation of e-NRTL binary interaction parameters and its impact on the prediction of thermodynamic properties of multicomponent electrolyte systems

The main objective of the current work is to obtain and validate the binary interaction parameters of the e-NRTL model for different salts in water, using the data of mean ionic activity coefficients vs. molality compiled by Robinson and Stokes (Electrolyte Solutions, Dover Publications, 2012). This way, the interaction parameters tau(salt),(water) and tau(water),(salt) of the e-NRTL thermodynamic model for 171 binary (salt + water) electrolyte systems have been calculated by non-linear regression of the data available in the literature. The fitting was performed by the Marquardt-Levenberg algorithm which was computationally implemented in an Excel VBA homemade program. The e-NRTL equation was solved by using two different procedures: (i) included in the Excel VBA code or, (ii) implemented in Aspen Plus by linking the process simulator and the Excel VBA code via automation as the third-party software. The interaction parameters obtained by both approaches were able to adequately fit the experimental results. They were successfully validated by the thermodynamic prediction of multicomponent systems reported by other authors. Additionally, the current predictions were compared with those obtained with the parameters contained in the Aspen Properties database showing a significant improvement for several systems. Finally, this work demonstrates that it is possible enhancing the quality of the prediction of multicomponent systems with proper values of the binary electrolyte interaction parameters, without considering electrolyte-electrolyte interactions foreseen in the original e-NRTL model, which makes possible the simulation of processes containing a variety of chemical species regardless their nature. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

The study aims to calculate and validate binary interaction parameters for different salts in water using regression analysis, showing that the calculated parameters obtained through two different procedures can effectively fit experimental results and improve the prediction quality of multicomponent systems.