Styrene and ethylbenzene absorption in ionic liquids: comparing DFT affinity calculations with experimental data

Styrene is a widely used bulk chemical produced by dehydrogenation of ethylbenzene (EB). Purification of styrene to contain < 100 ppm EB is not cost-effective by conventional separation methods. One separation method is extractive distillation with an ionic liquid (IL) as a binding agent for one of the components, thereby lowering the vapour pressure of this component. In this study, using quantum density functional theory (DFT), we have simulated 22 IL anion–cation pairs, styrene and EB affinities to them, and ion-pair dimer affinities of the ILs. These are compared with experimental liquid–liquid equilibrium studies of M.T.G. Jongmans, B. Schuur, and A.B. de Haan, Ind. Eng. Chem. Res. 50 (2011), pp. 10800–10810. It is shown that experimental selectivity and distribution coefficients of styrene and EB in the ILs are related to computed gas phase anion–cation stabilisation energies and ion-pair–ion-pair dimer affinities. The inverse of molar volume is found to strongly correlate with the selectivity. The computational results also qualitatively correlate with molar volume, and consequently, it is possible to use DFT calculations as a qualitative prediction tool in screening of ILs for this separation process. This tool does not account for effects caused by long alkyl chains, as the length does not seem to affect dimer stabilisation energy beyond ethyl group.     

On the prediction of transport properties of ionic liquid using 1-n-butylmethylpyridinium tetrafluoroborate as an example

The Green–Kubo and Einstein–Helfand approaches are examined for calculating diffusion coefficient, electronic conductivity and shear viscosity of ionic liquid using 1-n-butylmethylpyridinium tetrafluoroborate [C₄PY][BF₄] as an example. Both methods suffer numerical errors accumulated at long simulation time, resulting divergences in the integrated autocorrelation time (IAT) and nonlinearity in the mean square displacement (MSD). The numerical errors can be reduced using smaller time step in the simulation. By identifying a converged plateau in IAT and a linear segment in MSD both approaches yield consistent predictions. Using a validated force field, the predicted diffusion coefficient and electrical conductivity agree reasonably well with the experimental data. However, the shear viscosity is significantly underestimated. Analysis of the simulation data indicates that a much slow relaxation in the pressure tensor must be considered, which is unfortunately infeasible due to the accumulated numerical error. Alternatively, the non-equilibrium periodic perturbation method shows promising improvement in the prediction.