Fluid Phase Equilibria 158 160 1999 151 163 Prediction of phase equilibria in waterralcoholralkane systems Epaminondas C. Voutsas ), Iakovos V. Yakoumis, Dimitrios P. Tassios Laboratory of Thermodynamics and Transport Phenomena, Department of Chemical Engineering, National Technical UniÕersity of Athens, 9, Heroon Polytechniou Str., Zographou Campus, 15780, Athens, Greece Received 22 March 1998; accepted 9 October 1998 Abstract The Cubic Plus Association Equation of State CPA EoS. is applied to the prediction of Vapor Liquid Equilibrium VLE. and Liquid Liquid Equilibrium LLE. in ternary associating mixtures containing water, alcohols and alkanes. The appropriate set of combining rules for the cross-association energy and volume parameters of the EoS is identified by evaluating the performance of four such sets in the correlation of VLE and LLE for waterralcohol mixtures. Using only one interaction parameter per binary mixture, in the physical part of the EoS, CPA gives very satisfactory predictions of ternary VLE and LLE. q 1999 Elsevier Science B.V. All rights reserved. Keywords: Associating fluids; Equation of state; Vapor liquid equilibrium; Liquid liquid equilibrium 1. Introduction Knowledge of phase equilibria VLE and LLE. in waterralcoholralkane mixtures is very important for many practical applications such as the extraction of alcohols from aqueous solutions using near critical light hydrocarbon as solvents w1,2 x; or in the petroleum industry, since alcohols are widely used as fuel additives or to inhibit hydrate formation. Mixtures containing associating compounds are also of great interest from the theoretical point of view for testing molecular as well as statistical models. Species that form hydrogen bonds often exhibit unusual thermodynamic behavior. In pure fluids, strong attractive interactions between like molecules result in the formation of molecular clusters that considerably affect their thermodynamic properties. In mixtures, hydrogen bonding interactions can occur between molecules of the same species self-association or between molecules of different species cross-association or solvation. ) Corresponding author. Tel.: q30-1-7723232; fax: q30-1-7723155; e-mail: evoutsas@orfeas.chemeng.ntua.gr 0378-3812r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S0378-3812 99 00131-4
152 ( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 Classical engineering models, such as UNIFAC or cubic equations of state, cannot accurately represent phase equilibria in mixtures containing associating compounds w3,4x because they do not take explicitly into account the special interactions encountered in them. In the last few years, many equations of state that take explicitly into account hydrogen bonding have been developed w2,5 7 x. One of these is the Cubic Plus Association CPA., which combines the SRK EoS with the perturbation theory of Wertheim which is also used in SAFT wx 6. In a series of papers, CPA has been successfully applied to pure associating fluids as well as mixtures of associating fluids with inert compounds w3,8 10 x. In this study, we extend CPA to systems involving cross-association and proceed to investigate its predictive capabilities in ternary mixtures containing water, alcohols and alkanes. Only binary interaction parameters are used, which are determined by fitting to binary VLE and LLE data. 2. The CPA EoS The pressure equation of the CPA EoS is given by the following expression: RT a RT 1 1 E X A j Ps y q ÝxiÝrjÝ y, 1 A j. Vyb V Vqb. V i j A j X 2 Eri where the physical term is that of the SRK EoS and the association term is taken from SAFT wx 6. The A summation is over all association sites and the mole fraction of molecules not bonded at site A X i. is defined by: ž / y1 Ai Bj AiBj X s 1qrÝÝxX j D. 2 j B j In the above equation, D A i B j where g AB i j AB i j AB i j ž / RT is the association strength given by: D sg exp y1 b b, 3. is the radial distribution function defined by: b 2y 4V gs 3. 4. b 21y 4V ž / Finally, the energy parameter of the EoS a is defined using a Soave-type temperature dependency: 2 0 1 ( r ž ž // asa 1qc 1y T, 5 and the co-volume parameter b. is assumed to be temperature-independent.
( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 153 Table 1 CPA parameters for the components examined in this work Component b lrmol. a bar l rmol. c K. b 2 AB AB 0 1 water 0.01453 0.9894 1.0669 162.01 0.0787 methanol 0.0321 4.0220 0.4623 244.76 0.0154 ethanol 0.0491 8.5164 0.7502 215.00 0.0080 i-propanol 0.0641 10.387 0.9621 210.00 0.0092 n-butanol 0.0797 15.1250 1.0140 210.00 0.0090 2-butanol 0.0797 15.6063 0.9239 210.00 0.0041 propane 0.0578 9.1187 0.6307 i-butane 0.0747 12.9094 0.7021 n-hexane 0.1079 23.6810 0.8313 The CPA EoS uses three pure-compound parameters for nonassociating inert. compounds, a 0, c1 and b, while for associating compounds, two additional parameters are needed: the association energy AB. AB of interaction between sites A and B on a molecule and the parameter b, which is an association volume parameter. Pure-compound parameters are determined by fitting experimental saturated vapor pressure and liquid volume data. For inert compounds, these parameters can be also estimated using their critical properties and acentric factor as in the original SRK EoS. The latter represents an advantage of CPA over SAFT in cases where supercritical compounds are involved. The pure-compound parameters used in this study are presented in Table 1. 3. Phase equilibria calculations in binary systems 3.1. Systems containing only one self-associating compound and one inert compound Such systems involved in this study are the alcoholralkane and waterralkane binary systems. The alcohol molecules are treated as two-site molecules, while water as a four-site one. The extension of the CPA EoS to such mixtures requires explicit mixing rules only for the parameters a and b of the physical part, while the extension of the association term is straightforward. The classical van der Waals one-fluid mixing rules are used for the parameters a and b with a single binary interaction coefficient k. in the a parameter. A detailed study of the performance of CPA in the correlation and prediction of phase equilibria in alcoholralkane and waterralkane binary mixtures has been presented elsewhere w3,9,10 x. Here we consider only typical systems pertinent to the objective of this study. Thus, Table 2 and Figs. 1 and 2 present results for mixtures containing alcohols and light hydrocarbons at temperatures near the critical ones of the hydrocarbons, where the latter have been proved to be very promising solvents for the dehydration and recovery of alcohols wx 1. As it can be seen, CPA gives excellent correlation of the VLE data and it also accurately predicts the azeotropic behavior of the mixture ethanolri-butane Fig. 2., which hinders the application of i-butane as solvent for the supercritical extraction of ethanol from aqueous solutions. Results, finally, for a waterrlight hydrocarbon system is shown in Fig. 3 for the waterrpropane system.
154 ( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 Table 2 Correlation performance of the CPA EoS in alcoholrlight hydrocarbon binaries a Alcohol Light hydrocarbon Ref. T K k D P % D y 100 Ethanol propane w13x 325.1 y0.0015 1.06 0.54 350.1 y0.0049 2.04 0.82 375.1 y0.0030 2.12 0.94 i-butane w13x 308.6 0.0126 1.50 0.40 318.4 0.0204 1.34 0.84 363.5 0.0365 1.77 2.87 i-propanol propane w14x 313.1 0.0412 2.61 333.1 0.0290 4.13 0.49 353.1 0.0285 3.05 1.28 i-butane wx 1 331.5 0.0162 2.71 0.24 363.6 0.0086 3.78 1.03 2-Butanol propane w15x 328.1 0.0216 1.33 0.04 348.1 0.0201 1.40 0.06 368.1 y0.0131 1.33 0.11 a D P %. is average absolute percentage error in bubble point pressure. b D y is the average absolute deviation in vapor phase mole fraction. U b 3.2. Systems containing two self-associating compounds In such systems, like the waterralcohol systems, cross-association occurs. Since the type of association we used for the four-sitertwo-site waterralcohol systems is discussed in details in Refs. Fig. 1. Correlation of VLE of the binary ethanolrpropane system. Experimental data from Ref. w15 x.
( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 155 Fig. 2. Correlation of VLE of the ethanolr i-butane system. Experimental data from Ref. w15 x. w6,11 x, we will concentrate here to the combining rules for the cross-association energy and volume A i B j A i B parameters and b j Ai and Bj denote two different sites on two different molecules, e.g., one on the oxygen atom of the water molecule and one on the hydrogen atom on the alcohol molecule.. Different sets of combining rules have been used by different authors working with equations of state based on Wertheim s theory w7,11,12 x. In this study, four different sets of Fig. 3. VLLE calculations for the system waterrpropane. Experimental data from Ref. w25 x.
156 ( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 combining rules are compared with respect to their performance in the correlation of VLE and LLE in alcoholrwater mixtures: Ai q B j b A i qb B j. AB i j AB i s, b i j s, which will be referred as CR-1 set. 2 2 A i B q j AB i j AB i j A i B ii. s, b s( b b j, which will be referred as CR-2 set. 2 AB i j ' Ai Bj AB i j Ai B ( j iii s, b s b b, which will be referred as CR-3 set. AB ' i j A i B iv. D s D D j. In this approach, which has been proposed by Suresh and Elliott wx 7, no combining rules are needed for the Ai B j and b A i B j parameters. Furthermore, unlike the other three sets, it gives an analytical solution for the mole fraction of molecules not bonded at site A of the A molecule i X i.. This combining rule will be referred as CR-4. 3.2.1. VLE in waterr alcohol binary systems Table 3 presents VLE correlation results using the four sets of combining rules for some waterralcohol systems using a single binary interaction parameter in the physical part of the EoS. It is A i B apparent that: a the combining rule for the b j parameter is the crucial one, since the arithmetic mean CR-1. gives poor results, while the geometric one CR-2 and CR-3. gives very good ones; b. the arithmetic mean and the geometric mean combining rules for the Ai B j parameter are equivalent A i B when coupled with the geometric mean for the b j parameter CR-2 and CR-3.; and c. the CR-4 combining rule gives very good results similar to the ones obtained by the CR-2 and CR-3 sets. 3.2.2. LLE in waterr alcohol binary systems Fig. 4 presents LLE correlation results with the CPA and the CR-2, CR-3 and CR-4 combining rules for the waterrn-butanol mixture. The CR-2 and CR-3 sets give similar satisfactory results, eventhough they tend to overestimate the upper critical solution temperature probably due to the use of a temperature independent k. On the other hand, the Elliott s combining rule CR-4. strongly overestimates the n-butanol solubilities in the water-rich phase. 3.2.3. A theoretical explanation of the CR-2 set of combining rules As suggested by Prausnitz w16 x, the standard Gibbs energy of forming cross-associates can be approximated by: 1 o. o. o. Dg s ž Dgii q Dg jj /. 6. 2 It follows from Eq. 6 that the cross-association constant K can be approximated by: ( K s K K. 7. ii jj On the other hand: DH DS ln K sy q, 8. RT R where DH and DS are the enthalpy and entropy of forming cross-associates, respectively.
Table 3 VLE correlation results for waterralcohol systems with the CPA EoS System Ref. NP T K CR-1 CR-2 CR-3 CR-4 k D P %. D y 100 k D P %. D y 100 k D P %. D y 100 k D P %. D y 100 U U U U Waterr w18x 12 298.2 0.0432 6.64 2.63 y0.0801 0.57 0.27 y0.1495 2.86 0.91 y0.1093 1.16 0.32 methanol w19x 16 373.2 0.0842 2.93 2.37 y0.0522 2.24 0.62 y0.1151 4.19 1.16 y0.0855 3.29 0.80 w19x 14 423.2 0.1359 3.17 1.85 y0.0264 0.71 0.75 y0.0926 1.88 1.41 y0.0666 1.38 1.11 w19x 15 473.2 0.1543 3.10 1.82 y0.0240 0.65 0.43 y0.0872 1.03 0.55 y0.0676 0.86 0.44 Waterr w18x 22 298.2 0.0709 14.58 9.49 y0.0560 3.10 1.94 y0.0750 1.57 1.09 y0.1004 1.31 0.59 ethanol w20x 13 328.2 0.1082 13.92 8.19 y0.0359 2.02 1.23 y0.0552 0.63 0.69 y0.0856 1.80 1.11 w20x 13 343.2 0.1332 12.46 7.64 y0.0286 1.59 0.98 y0.0490 0.46 0.62 y0.0834 1.86 1.19 w18x 24 363.2 0.1248 12.09 7.73 y0.0287 1.44 1.23 y0.0482 0.39 0.56 y0.0851 1.92 0.78 w21x 17 423.2 0.1765 9.36 7.09 y0.0406 1.15 1.57 y0.0633 1.50 1.29 y0.1157 2.76 1.45 w21x 17 473.2 0.1876 8.43 5.68 y0.0451 1.23 1.82 y0.0659 1.19 1.70 y0.1216 2.08 1.61 Waterr w22x 9 328.2 0.0363 18.03 13.73 y0.0708 5.34 5.05 y0.0845 3.80 4.48 y0.1339 1.23 2.93 i-propanol w20x 8 338.2 0.0551 16.96 11.15 y0.0636 4.95 3.09 y0.0781 3.50 2.27 y0.1332 1.60 0.82 w21x 19 423.2 0.1097 12.60 7.29 y0.0386 1.91 2.55 y0.0539 1.25 2.31 y0.1312 3.41 1.98 w21x 18 473.2 0.1334 11.09 4.96 y0.0476 1.77 1.90 y0.0634 1.18 1.76 y0.1504 2.46 1.45 OÕerall 217 13.40 8.56 2.65 2.32 2.32 2.12 2.49 1.65 errors E.C. Voutsas et al.rfluid Phase Equilibria 158 160 ( 1999 ) 151 163 157
158 ( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 Fig. 4. LLE correlation of the waterr n-butanol mixture. Experimental data from Ref. w26 x. Combination of Eqs. 7. and 8. leads to the following set of combining rules for DH and DS : DHiiqDHjj DSiiqDSjj DHs and DSs. 9. 2 2 Economou and Donohue w17x and Kontogeorgis et al. w8x have shown that the chemical and perturbation theories yield essentially the same expressions for the compressibility factor and the mole fraction of monomers. They have concluded that KRTAD, which after some easy algebra, leads to the following relationships: AB DS i j R AB i j DH A and e Ab. 10. RT RT Eq. 10. suggests that the arithmetic mean combining rule for the cross-association enthalpy DH. A i B is equivalent to the arithmetic mean one for the cross-association energy parameter j. and that the arithmetic mean combining rule for the cross-association entropy DS. is equivalent to the A i B geometric mean one for the cross-association volume parameter b j.. Note that the similar results Table 4 VLE prediction results in ternary mixtures with the CPA EoS U U System Ref. NP T K. D P %. D y 100 D y 100 1 2 Waterr ethanolr propane w23x 10 333.0 4.43 0.16 0.20 w23x 7 363.0 4.87 0.13 0.65 w23x 4 383.0 4.54 0.05 0.63 Waterr i-propanolr i-butane wx 1 3 319.3 0.76 0.47 0.26 wx 1 3 338.3 0.99 0.57 0.36 wx 1 3 363.6 1.20 0.78 0.22 Waterr i-propanolr propane w24x 5 353.0 3.06 0.30 0.76
( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 159 Fig. 5. Waterrpropane relative volatilities in mixtures with ethanol. Experimental data from Ref. w23 x. A i B j between arithmetic and geometric mean for CR-2 and CR-3 are due to the fact that the corresponding values differ by less than 1%. 4. Prediction of ternary phase equilibrium Phase equilibrium predictions with the CPA have been performed for ternary waterralcoholralkane mixtures using the CR-2 set of combining rules for the waterralcohol cross-associating wx Fig. 6. i-propanol distribution coefficients between water and i-butane. Experimental data from Ref. 1.
160 ( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 Fig. 7. Binodal curve predictions for the mixture waterrmethanolrpropane at 293 K. Experimental from Ref. w27 x. mixtures. All k values are obtained from the corresponding binary data except for waterri-butane where, in the absence of data, the value for the waterrbutane was used. Table 4 presents VLE prediction results with the CPA for the waterrethanolrpropane, waterripropanolri-butane and the waterri-propanolrpropane mixtures and typical ones in Figs. 5 and 6. Very satisfactory predictions are obtained for the three systems. Thus, Figs. 5 and 6 indicate that CPA correctly predicts the ability of the two light hydrocarbons to act as dehydrating agents: the larger than Fig. 8. Prediction of methanol distribution between water and propane. Experimental data from Ref. w27 x.
( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 161 Fig. 9. Prediction of methanol distribution coefficients between water and n-hexane at 293 K. Experimental data from Ref. w28 x. 1 relative volatilities of water for high propane concentrations in Fig. 5; and the low distribution coefficients of i-propanol in Fig. 6 combined with the correctly predicted high values for water. Very satisfactory, finally, LLE predictions are obtained as suggested by the results for the ternary waterrmethanolrpropane, waterrmethanolrn-hexane and waterrethanolrn-hexane mixtures presented in Figs. 7 10. Fig. 10. Prediction of ethanol distribution coefficients between water and n-hexane at 298 K. Experimental data from Ref. w29 x.
162 ( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 5. Conclusions The performance of the CPA EoS in the prediction of VLE and LLE in mixtures containing water, alcohols and alkanes is investigated. Four different sets of combining rules for the cross-association energy and volume parameters of the EoS were evaluated on the basis of their performance in the correlation of VLE and LLE in waterralcohol mixtures. It is concluded that the combining rule for the cross-association volume parameter is the crucial rule. The geometric mean gives good results, while the arithmetic mean gives poor ones. For the cross-association energy parameter, the arithmetic mean represents the appropriate choice. Finally, the Elliott s combining rule gives good results in the correlation of VLE, but poor ones in the correlation of LLE in waterralcohol mixtures. Very satisfactory VLE and LLE prediction results are obtained for waterralcoholralkane mixtures suggesting that CPA is a reliable tool for modelling both high-pressure separation methods such as the supercritical extraction as well as low-pressure ones such as liquid liquid extraction. 6. List of symbols a a0 b c1 g Dg DH k K P R DS T Tr V X A energy parameter attractive-term parameter co-volume parameter parameter in the attractive term radial distribution function standard Gibbs energy of forming cross associates enthalpy of forming cross associates binary interaction coefficient cross-association constant pressure gas constant entropy of forming cross associates temperature reduced temperature molar volume mole fraction of the component not bonded at site A Greek letters b AB parameter in the association term of the CPA EoS D AB strength of interaction between sites A and B AB association energy of interaction between sites A and B r molar density Subscripts A, B for site A, B, on the molecule i component i j component j
( ) E.C. Voutsas et al.rfluid Phase Equilibria 158 160 1999 151 163 163 References wx 1 M.S. Zabaloy, G.D.B. Mabe, S.B. Bottini, E.A. Brignole, Fluid Phase Equil. 83 1993. 159. wx 2 H.P. Gross, S. Bottini, E.A. Brignole, Fluid Phase Equil. 116 1996. 537. wx 3 I.V. Yakoumis, G.M. Kontogeorgis, E.C. Voutsas, D.P. Tassios, Fluid Phase Equil. 130 1997. 31. wx 4 G. Wagner, S.I. Sandler, J. Chem. Eng. Data 40 1995. 1119. wx 5 A. Anderko, Fluid Phase Equil. 45 1989. 39. wx 6 S.H. Huang, M. Radosz, Ind. Eng. Chem. Res. 29 1990. 2284. wx 7 S.J. Suresh, J.R. Elliott, Ind. Eng. Chem. Res. 31 1992. 2783. wx 8 G.M. Kontogeorgis, E.C. Voutsas, I.V. Yakoumis, D.P. Tassios, Ind. Eng. Chem. Res. 35 1996. 4310. wx 9 E.C. Voutsas, G.M. Kontogeorgis, I.V. Yakoumis, D.P. Tassios, Fluid Phase Equil. 132 1997. 61. w10x I.V. Yakoumis, G.M. Kontogeorgis, E.C. Voutsas, E.M. Hendriks, D.P. Tassios, Ind. Eng. Chem. Res. 37 1998. 4175. w11x Y.-H. Fu, S.I. Sandler, Ind. Eng. Chem. Res. 34 1995. 1897. w12x Suresh, E.J. Beckman, Fluid Phase Equil. 99 1994. 219 240. w13x M.S. Zabaloy, H.B. Gross, S.B. Bottini, E.A. Brignole, J. Chem. Eng. Data 39 1994. 214. w14x M.S. Zabaloy, G.D.B. Mabe, S.B. Bottini, E.A. Brignole, J. Chem. Eng. Data 38 1993. 40. w15x H.B. Gross, M.S. Zabaloy, E.A. Brignole, J. Chem. Eng. Data 41 1996. 335. w16x J.M. Prausnitz, Molecular Thermodynamics of Fluid Phase Equilibria, Prentice-Hall, Englewood Cliffs, NJ, 1969. w17x I.G. Economou, M.D. Donohue, AIChE J. 37 12. 1991. 1875. w18x J. Gmehling, U. Onken, W. Arlt, Vapor liquid equilibrium data collection, DECHEMA Chemistry Data Series: Frankfurt 1 1981. Part 1a. w19x J. Griswold, S.Y. Wong, Chem. Eng. Prog. Symp. Ser. 48 1952. 18. w20x J. Gmehling, U. Onken, Vapor liquid equilibrium data collection, DECHEMA Chemistry Data Series: Frankfurt 1 1977. Part 1. w21x F. Barr-David, B.F. Dodge, J. Chem. Eng. Data 4 1959. 107. w22x J. Gmehling, U. Onken, J.R. Rarey-Nies, Vapor liquid equilibrium data collection, DECHEMA Chemistry Data Series: Frankfurt 1 1988. Part 1b. w23x H. Horizoe, T. Tanimoto, I. Yamamoto, Y. Kano, Fluid Phase Equil. 84 1993. 297. w24x M. Zabaloy, G. Mabe, S.B. Bottini, E.A. Brignole, J. Supercritical Fluids 5 1992. 186. w25x R. Kobayashi, D.L. Katz, Ind. Eng. Chem. 45 1953. 440. w26x J.M. Sorensen, W. Arlt, Liquid liquid equilibrium data collection, DECHEMA Chemistry Data Series: Frankfurt 5 1979. Part 1. w27x Noda et al., J. Chem. Eng. Japan 8 1975. 492. w28x V.B. Kogan et al., Zh. Prikl. Khim. 29. 9 1956. 1387. w29x J.M. Sorensen, W. Arlt, Liquid liquid equilibrium data collection, DECHEMA Chemistry Data Series: Frankfurt 5 1980. Part 2.