Ind. Eng. Chem. Res. 2008, 47, 989
989
Reply to “Comments on ‘Performance of COSMO-RS with Sigma Profiles from Different Model Chemistries’” Tiancheng Mu,†,‡ Ju1 rgen Rarey,† and Ju1 rgen Gmehling*,† Lehrstuhl fu¨r Technische Chemie, Institut fu¨r Reine und Angewandte Chemie, Carl Von Ossietzky UniVersita¨t Oldenburg, D-26111, Oldenburg, Germany, and Department of Chemistry, Renmin UniVersity of China, 100872, Beijing, People’s Republic of China Sir: In his comments about our manuscript, Klamt discusses aspects of our work, and we feel that several of his comments require further clarification by the authors. It is correct that the COSMO dielectric continuum model replaces only the dielectric boundary condition by an ideal conductor and does not necessarily use an infinite dielectric constant in the dielectric continuum itself. We used the COSMO model with different dielectric constants ourselves in the past and should have mentioned this possible use besides the generation procedure for σ profiles. We used the two different COSMO-RS flavors in our work, together with various QC levels without re-parametrization. In Figure 3 of the original work, overall results are given also for COSMO-SAC, using the VT-2005 profiles and COSMO-RS(Ol) using the Turbomole BP/TZVP. These are the QC levels that were used for parametrization of the two models. To stress the importance of parametrization to a certain QC level, we included the respective statement from a publication of Klamt’s company, COSMOLogic: Different model chemistries usually lead to different results, and the COSMO-RS parameters should be adjusted to the model chemistry used: “COSMO-RS parameters are not specific of functional groups or molecule types. The parameters have to be adjusted for the QM-COSMO method that is used as a basis for the COSMO-RS calculations only.” (COSMOtherm Manual, C12 0103) The goal of parametrization to a certain QC level is to improve the calculation results of the COSMO-RS flavor. In our work, we observed comparable or slightly better results for B3LYP/6-311G(d,p), even without the required re-parametrization. We firmly believe that this allows the conclusion that B3LYP/6-311G(d,p) is at least slightly more suitable, as reparametrizing would result in even better, not worse, results. In the following paragraph, Klamt states that the generally small differences between the COSMO-RS flavors using the different QC levels do not allow one to draw the conclusion that one calculation path would have any advantages. This is not very different from our findings that advantages are observable but small. Klamt assumes an advantage of COSMORS(Ol), because of the training of the method to the same data that were used to test the models. In our opinion, this has very little effect, because only a very limited amount of data was used, due to the numerically intensive calculations required. The group contributions of UNIFAC and modified UNIFAC, on the * To whom correspondence should be addressed. Tel.: ++49-4417983831. Fax: ++49-441-7983330. E-mail address: gmehling@ tech.chem.uni-oldenburg.de. † Lehrstuhl fu ¨ r Technische Chemie, Institut fu¨r Reine und Angewandte Chemie, Carl von Ossietzky Universita¨t Oldenburg. ‡ Department of Chemistry, Renmin University of China.
other side, were regressed to a significant amount of the data (although many of the parameters are rather old and the amount of VLE data alone increased by more than 57% during the last 10 years). Due to the intensive training of a large number of parameters to experimental data possible in group contribution methods, we felt that a direct comparison would be unfair toward the much more predictive methods analyzed in this paper and mostly avoided this comparison. When analyzing the results of COSMO-SAC and COSMORS(Ol), in the case of activity coefficients at infinite dilution, we had hoped to find less-biased deviations than in the case of the UNIFAC and modified UNIFAC methods. Both group contribution methods are based on the UNIQUAC model, which is known to systematically underestimate these data. Our expectations were motivated by several publications and an argument from a book of Klamt about COSMO-RS. Unluckily, the sentence in the manuscript must have been changed and we did not notice it. In section 4.2.2.8, instead of “This also has been reported by other authors.37,39” it should read “This is in contradiction to the findings of other authors.37,39” In a further comment, Klamt argues that the self-consistent equation approach used in COSMO-RS would correctly model the association equilibria in mixtures with one or more hydrogen-bonding species. We are very skeptical and do not believe that COSMO-RS, in its current form, can reproduce the equilibria between distinct associates like dimers, chains, or cyclic tetramers observed, for example, in alkanol solutions. Currently, from our experience, we also do not expect that it could be possible to extract reliable knowledge about association and solvation constants and their temperature dependence from the charge density distribution given in the σ profiles but that this would require quantum mechanical ab initio calculations of associated molecular clusters. One of the reasons for this complication lies in the angle dependence of the hydrogenbonding energy. Strongest hydrogen bonding is achieved when the hydrogen, its electronegative bond partner, and the electronegative acceptor form a straight line. Similar angle dependence is not observed in the case of other polar (charge misfit) interactions. In a last comment, Klamt states that it would be misleading to classify ethers and esters as unpolar components and that our misclassification would result from our work with group contribution methods, which would be our main area of expertise. Two of the three authors are chemists and in chemistry, ethers and larger esters are generally, and without any doubt, regarded as unpolar solvents. Small ester molecules are considered slightly but not dominantly polar, and we, therefore, decided not to distinguish between small and large esters, for the sake of simplicity. IE071614G
10.1021/ie071614g CCC: $40.75 © 2008 American Chemical Society Published on Web 12/19/2007
