471395 an Extension of COSMO-Based Methodologies for Computer-Aided Mixture Design

Tuesday, November 15, 2016: 9:50 AM
Union Square 3 & 4 (Hilton San Francisco Union Square)
Nick Austin, Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA, Nick Sahinidis, Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA and Daniel W. Trahan, Engineering & Process Sciences, The Dow Chemical Company, Freeport, TX

Group contribution methods like UNIFAC have been successful in capturing thermodynamics in mixture design problems while remaining amenable to a mathematical optimization framework. Despite being accurate in many domains, these methods have an inherent limitation: because they are often regressed over large datasets of neutral, ground state structures, they can only consider such structures when applied to predict mixture thermodynamics. This means that UNIFAC-like methods cannot be extended to arbitrary systems as they would fail to capture the electronic complexities of species like transition states, reactive intermediates, and ionic liquids. In this work, we propose the use of quantum chemistry calculations via the COSMO solvation model as a means to incorporate arbitrary systems into mixture design problems. Every component of these systems that is not in the design space can be modelled at an ab initio level of accuracy and then mapped to a charge density histogram called a sigma profile. In COSMO post-processing methodologies like COSMO-RS1 and –SAC2, these sigma profiles represent the main information required to estimate mixture thermodynamics properties. To incorporate these techniques into a mixture design problem, we develop group contribution methods to estimate the sigma profiles of molecules in the design space. That is, we only rely on group contribution estimates for every species that is not fixed in the mixture. We also include several extensions to our existing COSMO-based mixture design methodology3 to improve the accuracy of the estimates.

We address the optimization of the mixture design problem using derivative-free optimization (DFO) methods. By projecting the problem onto a lower-dimensional space, we exploit the efficiency of DFO techniques at solving problems with few degrees of freedom. We apply these techniques to three case studies: (1) a reaction rates optimization problem, (2) control of a halogen-metal exchange reaction’s selectivity, and (3) maximizing the selectivity of an intramolecular nucleophilic aromatic substitution reaction. It is important to note that all of these problems cannot be modeled directly with UNIFAC-like approaches as they either involve species which do not have available UNIFAC groups or exhibit complex electronic properties. We consider all of these problems first from the perspective of unrestricted solvent design, meaning we consider mixtures of solvents which may not exist. Next, we consider all of these problems from a more practical point of view, designing an optimal blend of common laboratory and industrial solvents. Overall, these case studies demonstrate the ability of COSMO-based approaches to incorporate highly accurate quantum chemical information directly into mixture design applications.

[1] F. Eckert and A. Klamt. Fast solvent screening via quantum chemistry: COSMO‐RS approach.. AIChE Journal. 2002. 48(2), 369-385.

[2] S. T. Lin and S. I. Sandler. A priori phase equilibrium prediction from a segment contribution solvation model. Industrial & engineering chemistry research. 2002. 41(5), 899-913.

[3] N. D. Austin, N. V. Sahinidis, and D. W. Trahan . A COSMO-based approach to computer-aided mixture design. Chemical Engineering Science. 2016. submitted


Extended Abstract: File Not Uploaded
See more of this Session: Physical Properties for Chemical Process and Product Design
See more of this Group/Topical: Process Development Division