Reprint of: Simulation based ionic liquid screeningfor benzene–cyclohexane extractive separation$

Zhaoxian Lyu a, Teng Zhou b, Lifang Chen a, Yinmei Ye a, Kai Sundmacher b,c, Zhiwen Qi a,n

a Max Planck Partner Group at the State Key Laboratory of Chemical Engineering, East China University of Science and Technology, Shanghai 200237, Chinab Max Planck Institute for Dynamics of Complex Technical Systems, 39106 Magdeburg, Germanyc Process Systems Engineering, Otto-von-Guericke University, 39106 Magdeburg, Germany

H I G H L I G H T S

� [C4mim][AlCl4] is selected as proper solvent to separate benzene and cyclohexane.� COSMO-RS is applied to screen solvent from combinations of 12 cations and 22 anions.� Interaction between ionic liquids and benzene is investigated by DFT calculation.� Continuous extraction processes are simulated using [C4mim][AlCl4] and sulfolane as solvent.

a r t i c l e i n f o

Article history:Received 8 June 2013Received in revised form4 April 2014Accepted 5 April 2014Available online 28 May 2014

Keywords:Ionic liquid [C4mim][AlCl4]Benzene–cyclohexane separationCOSMO-RSQuantum chemical calculationProcess simulation

a b s t r a c t

In order to screen ionic liquids (IL) as suitable solvents for the separation of benzene and cyclohexane,the extraction efficiency of ILs (12 cations and 22 anions) was estimated based on COSMO-RS predictionsof infinite dilution activity coefficients of benzene and cyclohexane in different ILs..[C4mim][AlCl4] wasfound to be the most promising solvent. To provide deep insight on how the IL structure influences theextraction efficiency, molecular interactions between IL ions and benzene were determined from DFTcalculations. Moreover, liquid–liquid equilibrium data of the ternary system benzene–cyclohexane—[C4mim][AlCl4] were experimentally determined and used to fit parameters of the NRTL activitycoefficient model. Based on the NRTL model the continuous extraction process was simulated andcompared with a reference process using sulfolane as solvent. For the extraction process using [C4mim][AlCl4], high cyclohexane product purity (99.65 wt%) and high benzene recovery efficiency (98.03%) canbe reached with at much lower energy consumption and higher product yield compared to conventionalextraction solvents. In conclusion, the ionic liquid [C4mim][AlCl4] is a promising solvent for theextractive separation of benzene and cyclohexane.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

As an important industrial chemical, cyclohexane can be producedby catalytic hydrogenation of benzene. Due to the limitation of thereaction equilibrium, non-converted benzene is entrained into thecyclohexane product and must be removed. Since benzene andcyclohexane have very close boiling points and can even form anazeotrope, it is difficult to separate this mixture. In industry, the

separation is commonly achieved by liquid–liquid extraction usingorganic compounds, such as sulfolane, as the solvent. In currentprocesses, the organic solvent is normally withdrawn from the topof the regenerator as vapor stream and returned to the bottom of theextractor as liquid stream. The vaporization of solvent results in highregeneration cost (Schneider, 2004). In addition, the volatility oforganic solvents can lead to serious environmental problems.

Ionic liquids (ILs) are innovative solvents entirely composed of ions.Their negligible vapor pressure makes the regeneration much easier,namely by evaporation or pervaporation of the other mixture compo-nents (Seddon, 1997; Huddleston et al., 1998). Due to much lowerenergy consumption and investment costs, extraction processes withILs as solvents are economically feasible (Meindersma and de Haan,2008). Moreover, the large number of possible cation and anioncombinations makes it possible to tailor a highly efficient ionic liquidsolvent for a specific separation task (Huddleston et al., 2001). Over

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/ces

Chemical Engineering Science

http://dx.doi.org/10.1016/j.ces.2014.05.0320009-2509/& 2014 Elsevier Ltd. All rights reserved.

DOI of original article: http://dx.doi.org/10.1016/j.ces.2014.04.011☆A publisher's error resulted in this article appearing in the wrong issue. The

article is reprinted here for the reader's convenience and for the continuity of thespecial issue. For citation purposes, please use the original publication details;Chem. Eng. Sci, 113, pp. 45-53.

n Corresponding author.E-mail address: zwqi@ecust.edu.cn (Z. Qi).

Chemical Engineering Science 115 (2014) 186–194

the past decade, ILs have beenwidely studied as solvents for aromatic/aliphatic liquid–liquid separations (Arce et al., 2007a, 2007b, 2008a,2008b, 2009, 2010; Ferreira et al., 2012; Meindersma and De Haan,2012; Zhang et al., 2007). According to Meindersma et al. (2010), morethan 120 ILs have been identified as possible candidates aromatic/aliphatic separation since the first application reported by John et al.(1982). It was found that ILs composed of anions with delocalizedelectron (e.g., [RSO4]� and [AlCl4]�) show much better extractionefficiency than other ILs. In several experimental studies, the separa-tion of benzene and cyclohexane was chosen as a benchmark case(Wang et al., 2008a, 2008b; González et al., 2010a, 2010b; Zhou et al.,2012a, 2012b).

Due to the huge number of cations and anions, it is necessary toapply reliable theoretical methods to guide the screening of ILs forseparation processes. So far, various predictive methods have beendeveloped for modeling the thermodynamic properties of IL-containing systems, e.g., the Perturbed-Chain Statistical AssociatingFluid Theory (PC-SAFT) (Paduszynski and Domanska, 2012; Domanskaet al., 2012) and the Conductor-like Screening Model for Real Solvents(COSMO-RS) model (Klamt and Eckert, 2000). As a quantum chemistrybased predictive method, the COSMO-RS model has been proven to bea powerful tool for fast IL solvent screening, including the separation ofaliphatic and aromatic hydrocarbon mixtures (Anantharaj andBanerjee, 2010, 2011a; Banerjee and Khanna, 2006; Burghoff et al.,2008; Gutierrez et al., 2010, 2012). Importantly, the available COSMOdatabase covers a large number of common cations and anions, whichmakes it very efficient for screening of ILs as solvents (Klamt andEckert, 2000; Diedenhofen et al., 2003).

In order to explain why a selected IL has a pronounced extractionefficiency, it is necessary to determine the specific molecular interac-tions between the solute molecules and IL ion pairs (Rezabal andSchäfer, 2013). Quantum chemical calculations are the best approachfor studying the cation–anion interactions, and interactions betweenILs and solute molecules (Dong et al., 2006, 2012; Dong and Zhang,2012; Nockemann et al., 2006; Cabaço et al., 2011; Lü et al., 2012). Onthe basis of the well-founded Hohenberg–Kohn theorem, densityfunctional theory (DFT) provides a sound method for obtaininginformation of energetics, structures, and properties of atoms andmolecules at high enough accuracy and at much lower computationalcosts than traditional ab initio wave function techniques (Geerlings etal., 2003).

Due to their higher molecular weight, most of the reported ILscannot provide a high mass-based extraction efficiency, which mayobstruct their applications (Meindersma et al., 2010). Therefore, apotential IL identified with the help of COSMO-RS might not bequalified for industrial application. Hence, the extraction performanceof ILs as solvents needs to be further evaluated through extractionprocess simulations. For simulating a continuous extraction process,ternary liquid–liquid equilibrium (LLE) data need to be determinedand correlated with a thermodynamic activity coefficient model, suchas the NRTL model (Renon and Prausnitz, 1968).

The objective of this work is to screen ILs as potential solvents toextract benzene frommixtures with cyclohexane based on simulationson different scales, i.e., COSMO-RS predictions, DFT calculations, andprocess simulations. First of all, infinite dilution extraction efficienciesof ILs composed of 12 cations and 22 anions were calculated by usingthe COSMO-RS model. The interaction energy and the natural bondorbital (NBO) were analyzed by quantum chemical calculations toinvestigate the suitability of selected ILs from the molecule point ofview. For further confirming the extraction performance of theselected IL and exploring its potential for industrial application,process simulations were carried out using Aspen Plus 12.1. Sincethe molar content of the unreacted benzene in mixtures withcyclohexane in practice is lower that 20%, special attention was paidto the removal of the reactant at low-concentrations of benzene (Zhouet al., 2012b).

2. Ionic liquid screening

2.1. COSMO-RS model

COSMO-RS is a quantum chemistry based statistical thermo-dynamics model for the prediction of thermodynamic propertiesof fluids (Klamt and Eckert, 2000). There are generally two steps inthe COSMO-RS prediction procedure, namely (1) the quantumchemical COSMO computation for the molecular species involved,and (2) the COSMO-RS statistical thermodynamic treatment. Astandard COSMO-RS prediction only requires the screening chargedensity (SCD) information of the interested compounds. The SCDdistribution of a compound is normally obtained from quantumchemical calculation using the DFT approximation. The SCD needsto be calculated only once and afterwards it can be stored in aCOSMO file. Nowadays, the available COSMO database alreadycontains COSMO files of a huge number of common solvents and ILcations and anions. This makes the COSMO approach very fast andefficient when predicting thermodynamic properties, e.g., activitycoefficients, of various systems (Klamt and Eckert, 2000, 2004).

In this work, infinite dilution activity coefficients of benzene andcyclohexane in ionic liquids, composed of 12 common cations and 22common anions, were calculated by using the COSMOthermX soft-ware package (COSMOlogic GmbH & Co. KG, Version C3.0, Release13.01) based on the COSMO-RS method. Among the 12 cations, fourmain types of cations with different side chain length were selectedto evaluate the impact of cation classes and side chain lengths onsolvent extraction efficiency. For the 22 anions, F-containing and Cl-containing anions, sulfate- and phosphate-anions with different alkylsubstituents were chosen to investigate the influence of delocalizedelectron of anions on the extraction efficiency. The detailed informa-tion about the considered 12 cations and 22 anions is provided inTables S1 and S2, respectively.

Among the studied cations and anions, [AlCl4]� is not in thecurrent COSMO database. Therefore, the COSMO file of [AlCl4]�

was determined from quantum chemical calculation at the BP-TZVP level by Gaussian 03W (Frisch et al., 2004). COSMO files ofother ions, benzene, and cyclohexane were taken directly from thelatest COSMO database (version: BP_TZVP_C21_0111). An IL can betreated as either a single compound or a mixture of ions. In thiswork, cations and anions are treated as individual species withequal molar fractions (Diedenhofen and Klamt, 2010).

2.2. Extraction efficiency at infinite dilution

The activity coefficient of a solute i in a solvent S is calculatedfrom the difference between the chemical potentials of the solutein the solvent μi

S and in the pure solute μii

γiS ¼ expμiS�μi

i

RT

!ð1Þ

The solute distribution coefficient at infinite dilution β1i , which

indicates the extraction capacity of the solvent, can be expressedin terms of the infinite dilution activity coefficient of the solute inthe solvent

β1i ¼ 1

γ1i

� �ILphase

ð2Þ

where the subscript i represents benzene or cyclohexane. Theextraction selectivity is defined as the ratio of the composition ofbenzene in the extract phase to that of benzene in the raffinatephase. The separation selectivity at infinite dilution ðS123Þ is defined

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194 187

as Lei et al. (2007)

S123 ¼γ13γ12

� �ILphase

ð3Þ

In Eq. (3), γ12 and γ13 stand for the infinite dilution activitycoefficients of benzene and cyclohexane in the IL solvent, respec-tively. They can be predicted by the COSMO-RS method. Theperformance index (PIp) is defined as the product of the solutedistribution coefficient and the separation selevtivity, whichindicates the overall extraction efficiency of a solvent at infinitedilution (Anantharaj and Banerjee, 2010).

PI1 ¼ β12 � S123 ð4Þ

In order to validate the COSMO-RS predictions, the experi-mental and calculated infinite dilution benzene/cyclohexane selec-tivities of four ILs ([C2mim][BF4], [C4mim][BF4], [C6mim][BF4],[C8mim][BF4]) at different temperatures are compared in Table 1.The predicted selectivity decreases as the temperature increasesand as the length of the cation side chain grows. This tendency isin full agreement with the experimental findings. The estimatedroot mean square deviation (rmsd) between the experimental andcalculated selectivities is 15.6%. This deviation is comparable withthe predictions (rmsd¼11.0%) of Anantharaj and Banerjee (2011a)where the COSMO-RS model was successfully applied for thescreening of ILs as solvent for the aliphatic/thiophene separation.

After the benchmarking of COSMO-RS predictions, the screening ofionic liquids was subsequently carried out. As shown in Fig. 1a, formost of the investigated cations, the benzene distribution coefficientsof ILs with anions Cl� , [AlCl4]� , and [TOS]� are generally higher thanthat of ILs with F-containing anions, such as [BF4]� , [CF3SO3]� . This isbecause anions containing a Cl atom or an aromatic ring have moredelocalized electrons than F-containing anions. Anions with moredelocalized electrons show weaker interactions with cations, whichpromotes the interaction of the IL with benzene.

For [HSO4]� and [H2PO4]� , their benzene distribution coeffi-cients are lower than that of other anions. However, as the alkylchain length increases, e.g., from [H2PO4]� to [DBP]� (anionnumber: 18–21), the distribution coefficient increases substan-tially. This is because the electron-donating ability of alkyl groupsis stronger than that of the hydrogen atom. Thus the electrons ofanions with alkyl chain are more delocalized, which finally resultsin a stronger interaction between ILs and benzene. But, it wasfound that anions with highly delocalized electrons can not onlyform π-complextion with aromatics (Zhang et al., 2007), but alsofacilitate aromatics to share the positive charge of the cation(Meindersma et al., 2010). The major factor dominating theinteraction between IL and benzene was further studied viaquantum chemical calculations as discussed in the next section.

For most of the anions, similar effects of the cation side chainlength on the solvent extraction capacity were found for bothimidazolium and ammonium ILs. Benzene distribution coefficientsgradually increase with the cation side chain length, i.e., [C2mim]þo[C4mim]þo[C6mim]þo[C8mim]þo[C10mim]þ . This is consistent

with the reported experimental results where [Cnmim][PF6] (n¼4, 5,6) was studied as solvent for the separation of benzene fromcyclohexane (Zhou et al., 2012a) and [Cnmim][NTf2] (n¼2, 4, 8, 10)for the separation of benzene from hexane (Arce et al., 2007b).

In order to evaluate the overall extraction efficiency of ILs, theirperformance index at infinite dilution was analyzed. As seen inFig. 1b, for all studied cations, ILs containing [AlCl4]� show veryhigh PIp values. Among them, [C2mim][AlCl4] was found to be thebest solvent candidate; but, it is a solid at room temperature, thuscannot be applied as solvent in an extraction process. Taking intoaccount the overall extraction efficiency, [C4mim][AlCl4] and[C4py][AlCl4] were found to be promising solvents. But, as therelatively high viscosity of [C4py][AlCl4] make it unsuitable forindustrial operation (Meindersma and De Haan, 2012), [C4mim][AlCl4] is finally proposed as solvent for the separation of benzenefrom cyclohexane via liquid–liquid extraction.

3. Quantum chemical calculations

3.1. Calculation methods

It was found out by COSMO-RS predictions that ILs with[AlCl4]� have an excellent extraction efficiency. It would be very

Table 1The comparision of experimental and COSMO-RS predicated benzene/cyclohexaneselectivities of IL solvents at infinite dilution condition.

Ionic liquids Experimental dataa Calculated data

303 K 323 K 343 K 303 K 323 K 343 K

[C2mim][BF4] 22.20 18.39 15.74 20.61 17.51 14.94[C4mim][BF4] 18.80 15.79 13.22 13.86 11.72 10.03[C6mim][BF4] 11.44 10.01 9.05 10.11 8.66 7.50[C8mim][BF4] 6.66 6.19 5.71 7.63 6.63 5.82

a From Foco et al. (2006).

0 2 4 6 8 10 12 14 16 18 20 22 240.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

[H2PO4]-[HSO4]

-[TOS]-[AlCl4]

-

Ben

zene

dis

trib

utio

n co

effic

ient

Anions

[C2mim]+ [C4mpyro]+

[C4mim]+ [C4mpip]+

[C6mim]+ [P4444]+

[C8mim]+ [N2222]+

[C10mim]+ [N3333]+

[C4py]+ [N4444]+

Cl-

0 2 4 6 8 10 12 14 16 18 20 22 241

2

3

4

5

6

7

8

9

10[AlCl4]

-

Anions

Perf

orm

ance

Inde

x

[C2mim]+ [C4mpyro]+

[C4mim]+ [C4mpip]+

[C6mim]+ [P4444]+

[C8mim]+ [N2222]+

[C10mim]+ [N3333]+

[C4py]+ [N4444]+

Fig. 1. Benzene distribution coefficient (a) and performance index (b) of ILs composedof 12 different cations and 22 different anions (Tables S1 and S2) predicted by theCOSMO-RS method at the infinite dilution condition (1. [Br]� , 2. [Cl]� , 3. [NO3]� , 4.[BF4]� , 5. [PF6]� , 6. [AlCl4]� , 7. [CH3COO]� , 8. [ClO4]� , 9. [SCN]� , 10. [CF3COO]� , 11.[CH3SO3]� , 12. [CF3SO3]� , 13. [TOS]� , 14. [HSO4]� , 15. [CH3SO4]� , 16. [EtSO4]� , 17.[BuSO4]� , 18. [H2PO4]� , 19. [DMP]� , 20. [DEP]� , 21. [DBP]� , 22. [Tf2N]�).

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194188

helpful to explain the suitability of [C4mim][AlCl4], if the contribu-tions of specific interactions, such as hydrogen bonding orπ-complextion, is analyzed in detail. As suggested, the impact ofCH-π and π–π interactions between the IL-cation and benzene onthe extraction efficiency cannot be neglected (Anantharaj andBanerjee, 2011b; Lü et al., 2012; Zhou et al., 2008). However, thereare still very few theoretical studies about the influence of anionson the extraction efficiency of ILs for benzene.

In this work, [C4mim][BF4] was taken as reference solvent to[C4mim][AlCl4] in the investigation of the interactions between ILsand benzene, due to their similar structure but differently deloca-lized electrons of anions. All quantum chemical calculations wereperformed with the Gaussian 03 software package (Frisch et al.,2004), where the structures were optimized at the B3LYP/6-31þþG** level. The radii of elements used are the Bondi Vander Waals radius (Bondi, 1964). By means of optimizing severalinitial configurations and comparing their final energies, thechosen geometry can be approximately regarded as global mini-mum. After the geometry optimization, the characteristics of thestationary points were checked and zero-point energies wereevaluated by frequency analysis at the same DFT level. Theinteraction energy was calculated via Eq. (5)

ΔE¼ EAB�EA�EB ð5Þwhere EAB is the energy of the A–B complex, and in the case of thiswork is IL-benzene or the cation–anion complex. Furthermore,second-order perturbation delocalization energies E(2) wereobtained by NBO analysis at the same level. E(2) is associatedwith the delocalization of i-j, which was is estimated from Eq. (6)

Eð2Þ ¼ΔEij ¼niðFijÞ2εj�εi

ð6Þ

where ni is the donor orbital occupancy, εi and εj are the diagonalelements, and Fij is the off-diagonal NBO Fock matrix element.

3.2. Interaction between ionic liquids and benzene

Fig. 2 shows the most stable configurations of the two studiedILs with benzene. As seen, anions are located in the middle of thecation and benzene molecule. Moreover, benzene is closer to thebutyl chain of the cation, and is parallel with the imidazole ring.

The interaction energy between the solvent and benzeneindicates the extraction ability of the solvent to some extent (Lüet al., 2012). From Fig. 2, the extraction ability of [C4mim][AlCl4] tobenzene (ΔE¼�2.04 kcal/mol) is stronger than that of [C4mim][BF4] to benzene (ΔE¼�1.49 kcal/mol). This is in accordancewith benzene distribution coefficients calculated by COSMO-RS(Fig. 1a). As a possible interpretation, the interaction between[AlCl4]� and [C4mim]þ (ΔEIL¼�68.27 kcal/mol) is weaker thanthat between [BF4]� and [C4mim]þ (ΔEIL¼�81.46 kcal/mol).Therefore, it is easier for the benzene molecule to penetrate thebond network in [C4mim][AlCl4] and interact with the anion or thecation (Ferreira et al., 2012).

Moreover, for the configuration of [C4mim][AlCl4] and benzene(Fig. 2a), the distance Cl5…H14 is 3.05 Å and Cl4…H13 is 3.20 Å.Both distances are less than the sum of van der Waals radii of theH atoms and the neighboring Cl atoms (the differences Δr are�0.20 Å and �0.05 Å, respectively), which indicates two H-bondsbetween the anion and benzene. For [C4mim][BF4] and benzene(Fig. 2b), the Δr between F27 and H41 is �0.28 Å. Normally, theH-bond is strengthened with the decreasing of Δr. Therefore, theH-bond strength between [BF4]� and benzene is stronger thanthat between [AlCl4]� and benzene. That is to say, the higherextraction efficiency of [C4mim][AlCl4] may not be directlyexplained by the H-bond interactions between benzene andthe anion.

The NBO analysis was performed to further determine otherpossible orbital interactions, such as CH-π and π–π interactionsbetween ILs and benzene. The results are listed in Table 2.Generally, the higher the E(2) is, the stronger the interaction inthe orbits will be (Zhao et al., 2012). As seen, E(2) between[AlCl4]� and benzene (1.31 kcal/mol) is lower than that between[BF4]� and benzene (2.11 kcal/mol), which indicates that the inter-action between [AlCl4]� and benzene is weaker than that between[BF4]� and benzene. Moreover, the interaction of π (C6–C7)-πn

(C24–H34) and π (C8–C9)-πn (C21–C22) indicates the existence ofCH-π and π–π interactions between [C4mim]þ and benzene,

rij= 3.20 Å r= -0.05 Å

= 158.8

rij= 3.05 Å r= -0.20 Å

= 158.8

IL= -68.27 kcal/mol= -2.04 kcal/mol

rij = 2.85 Å r = +0.18 Å

= 123.1

rij = 2.39 Å r= -0.28 Å

= 153.8

IL= -81.46 kal/mol= -1.49 kal/mol

Fig. 2. Optimized spatial configurations of IL-benzene complexes ((a) [C4mim][AlCl4], (b) [C4mim][BF4]) (Notes: ΔEIL is the interaction energy between cations andanions, ΔE is the interaction energy between IL and benzene).

Table 2NBO analysis and second-order perturbation stabilization energy E(2) (kcal/mol)between ILs and benzene.

Solvent Cation–benzene Anion–benzene

Donor Acceptor E(2) Donor Acceptor E(2)

[C4mim][BF4] π (C33–C34) πn (C7–H17) 0.24 LP (B26) sn (C35–H41) 0.14π (C35–C36) πn (C4–C5) 0.10 LP (F27) sn (C35–H41) 1.83

LP (F30) sn (C34–H40) 0.14

[C4mim][AlCl4] π (C6–C7) πn (C24–H34) 0.48 LP (Cl4) sn C7–H13 0.39π (C8–C9) πn (C21–C22) 0.17 LP (Cl5) sn C8–H14 0.92

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194 189

respectively. Importantly, E(2) between benzene and the cation[C4mim]þ in [C4mim][AlCl4] (0.65 kcal/mol) is higher than thatbetween benzene and the cation in [C4mim][BF4] (0.35 kcal/mol).It may indicate that the weaker bond between [AlCl4]� and[C4mim]þ can facilitate the association of benzene and the cation.

In conclusion, there are not only strong H-bonds betweenanion and benzene but also comparable CH-π and π–π interactionsbetween the cation and benzene. Compared to [BF4]� , [AlCl4]�

shows a weaker interaction with [C4mim]þ . This makes it mucheasier for benzene to disrupt the connection between cation andanion. As a result, the CH-π and π–π interactions between[C4mim]þ and benzene are promoted, which finally resultsin a higher affinity of [C4mim][AlCl4] towards benzene. This maybe the main reason why [C4mim][AlCl4] has the better extractionefficiency.

4. Liquid–liquid equilibrium

4.1. Experimental

In order to preliminarily evaluate the extraction efficiency ofthe selected solvent [C4mim][AlCl4] in practice, LLE experimentswere conducted. Benzene and cyclohexane were purchased fromLingfeng Co., Ltd. China, with purity higher than 99.5 wt%. Theionic liquid [C4mim][AlCl4] was purchased from Chengjie ChemicalCo., Ltd. Shanghai, with a purity higher than 99.0 wt% at a watercontent below 0.03 wt%. All chemicals were used directly withoutfurther purification.

The LLE experiments were carried out at isothermal conditionscontrolled by the oil bath with temperature fluctuations below70.1 K (Huber Ministat 230, Germany). The compositions of theliquid mixture were determined by the Sartorius BSA224S-CWbalance with a precision of 70.0001 g. A certain amount (3.0 g) ofthe liquid mixture was sealed in a bottle and stirred with amagnetic stirrer for at least 3 h and then settled overnight. After-wards, the upper and lower phases were carefully separated withsyringes and weighted. Considering the instability of the ionicliquid [C4mim][AlCl4] in humid air, the whole experiment wascarried out under dry nitrogen atmosphere in a glove box. Thesamples of both phases were analyzed by gas chromatograph(Agilent 7890 GC, USA) equipped with a flame ionization detector(FID) and a PEG-20M column. The GC was equipped with a pre-column as the ionic liquid cannot be analyzed due to its negligiblevapor pressure. First, the GC relative response factor of benzene tocyclohexane was obtained by using a standard mixture of purecomponents. An area normalization method was then applied todetermine the mass ratio of benzene to cyclohexane in each phase.With the initial weights of benzene and cyclohexane in themixtures and the weight of both phases, the content of benzeneand cyclohexane in each phase can be determined. Finally, thecontent of IL in each phase can be obtained by subtracting the sumof the measured mass fractions of benzene and cyclohexanefrom unity.

All the LLE experiments were carried out twice and eachsample was analyzed three times to check the reproducibilityand allow for statistical data analysis. The analysis of the twoliquid phases with the syringe causes a variance of 70.02 g. Theabsolute deviation of the compositions was estimated to be lessthan 1.5 wt%.

4.2. Evaluation of solvent extraction performance

The mass-based LLE data for the ternary system benzene–cyclohexane—[C4mim][AlCl4] at 298.15 K and atmospheric pres-sure are displayed in Table S3 and illustrated in Fig. 3. From the LLE

diagram, it is clear that benzene shows a much higher solubility inthe extract phase than cyclohexane. Moreover, the miscibility gapis very broad at the low-benzene-content region, which indicatesthat [C4mim][AlCl4] is very suitable to separate benzene at lowconcentration from cyclohexane.

The solvent usage for the extraction process is determined by thesolute distribution coefficient and the product purity mainly dependson the solvent selectivity. For the benzene and cyclohexane separation,a good solvent should exhibit a high distribution coefficient forbenzene and meanwhile a reasonable separation selectivity. Themass-based benzene distribution coefficient (β) and solvent selectivity(S) are expressed by Eqs. (7) and (8), respectively

β2 ¼wE

2

wR2

ð7Þ

S23 ¼wE

2=wR2

wE3=w

R3

ð8Þ

where subscripts 2 and 3 represent benzene and cyclohexane,respectively; superscripts E and R represent the extraction andraffinate phases, respectively. To evaluate the overall extractionefficiency of a solvent, the performance index (PI) was calculated byEq. (9).

PI ¼ β2 � S23 ð9ÞAs [C4mim][PF6] and [C4mim][BF4] were already reported as

promising IL solvents in the benzene–cyclohexane separation (Zhouet al., 2012b), they were used as reference solvents for comparisonwith the here selected [C4mim][AlCl4]. In addition, the organic solventsulfolane which is commonly used in industry was also considered. βand PI values of these four solvents are plotted together with thebenzene mass fraction in the raffinate phase (see Fig. 4).

From Fig. 4a, [C4mim][AlCl4] has a much higher benzenedistribution coefficient than the other three solvents. Moreover,its benzene distribution coefficient increases at decreasing ben-zene concentration in the raffinate phase. β41.0 when thebenzene concentration is lower than 0.2.

From Fig. 4b, the PI of all four solvents decreases as the benzeneconcentration in the raffinate phase increases. [C4mim][AlCl4] showsthe largest PI value over the whole benzene concentration region, andit is about 4–7 times higher than the PI value of sulfolane.

Furthermore, the solvent selectivity can be calculated by Eq. (8)based on the data in Fig. 4a and b. As a result, in the low benzeneconcentration region, the selectivity of [C4mim][AlCl4] is above 20and much higher than that of sulfolane.

0.00 0.25 0.50 0.75 1.00

0.00

0.25

0.50

0.75

1.000.00

0.25

0.50

0.75

1.00

Benzene

Exp.

Cyclo

hexa

ne

[C4mim][AlCl4]

NRTL fitting

Fig. 3. Phase diagram for the experimental and NRTL predicted LLE data (massbased) of the ternary system {[C4mim][AlCl4]þbenzeneþcyclohexane} at 298.15 K.

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194190

The LLE experiment data presented and discussed above,confirm that the computationally preselected solvent [C4mim][AlCl4] in fact has a high extraction efficiency at low benzeneconcentration, indicating its great potential for the deep separa-tion of benzene from cyclohexane. These findings suggest that theCOSMO-RS model is an effective tool for pre-screening of solventswithout the need for extensive experiments.

5. Process simulation and analysis

5.1. NRTL model regression

In order to investigate the solvent extraction efficiency undercontinuous process operation conditions, first the experimentallydetermined phase equilibrium data were regressed with the NRTLmodel. The activity coefficient γi for component i is given by theNRTL model (Renon and Prausnitz, 1968)

ln γi ¼∑jτjixjGji

∑kxkGkiþ∑

j

xjGij

∑kxkGkjτij�

∑mτmjxmGmj

∑kxkGkj

� �ð10Þ

τij ¼ΔgijRT

ð11Þ

Gij ¼ expð�αijτijÞ ð12Þ

where Δgij is an energy parameter that characterizes the interac-tion between species i and j, xi is the mole fraction of component i,αij is a non-randomness parameter related to the mixture, R is thegas constant, and T is the absolute temperature.

The regression was carried out using Aspen Plus and the IL wastreated as a non-associating component (Seiler et al., 2004). Thenon-randomness parameter αij was set to 0.2 or 0.3 according tothe characteristics of the component in the mixture, as recom-mended by Renon and Prausnitz (1968). The NRTL model para-meters were regressed by least squares minimization of thedifferences between the calculated and the experimental molefraction of the components in the two liquid phases.

In order to compare the extraction processes with [C4mim][AlCl4] and sulfolane as alternative solvents, the LLE experimentaldata for both systems were fitted (LLE data of the systemsulfolane–benzene–cyclohexane were taken from Chen et al.(2000)). The identified NRTL model parameters for the studiedsystems are listed in Table 3. The mass-based rmsd between theexperimental and calculated LLE compositions was evaluated toprovide a measure of the correlation accuracy (Domańska et al.,2007). The calculated rmsd are 0.0044 and 0.0084, respectively,indicating high regression quality for both systems.

5.2. Extraction process analysis

Using the NRTL model, a continuous extraction process forseparating benzene from cyclohexane was preliminarily simulatedin Aspen Plus 12.1. The extractor, evaporator, and distillationcolumn are modeled by the Extract block, Flash block, and RadFracblock, respectively. The simulated feed contained 14 wt% of ben-zene and was added to the bottom of the extractor. The extractphase and the raffinate phase were collected separately anddelivered to the flash or distillation column to remove the solvent.Recycled solvents were added to the top of the extractor, togetherwith the fresh solvent. The process flowsheet is depicted in Fig. 7.

In order to achieve a high cyclohexane product purity of at least99.50 wt% and meanwhile to ensure a high benzene recovery ratioof about 98.0%, the required stage number of the extractor and thecorresponding solvent-to-feed ratio (S/F) for the studied solventswere calculated and plotted in Fig. 5. As can be seen, the requiredsolvent amount decreases at increasing number of stages. For thesame separation task, the required amount of [C4mim][AlCl4] isonly about 30% of [C4mim][BF4] (Zhou et al., 2012b) and also ofsulfolane with the same number of stages. For example, when thenumber of stages is n¼6, the required solvent-to-feed ratio areS/F¼1.49, 5.33, and 5.29 for [C4mim][AlCl4], [C4mim][BF4], andsulfolane, respectively.

For [C4mim][AlCl4] as extracting solvent, the purity of the cyclo-hexane product and the benzene recovery ratio as functions of the S/Fratio at different number of stages are given in Fig. 6. Obviously, when

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4 [C4mim][PF6] (Zhou et al., 2012b) [C4mim][BF4] (Zhou et al., 2012b) [C4mim][AlCl4] (This work) Sulfolane (Chen et al., 2000)

wben in the raffinate phase (mass)

Ben

zene

dis

trib

utio

n co

effic

ient

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00

5

10

15

20

25

30 [C4mim][PF6] (Zhou et al., 2012b) [C4mim][BF4] (Zhou et al., 2012b) [C4mim][AlCl4] (This work) Sulfolane (Chen et al., 2000)

wben in the raffinate phase (mass)

Perf

orm

ance

Inde

x

Fig. 4. Experimental benzene distribution coefficient (a) and solvent performanceindex (b) for the ternary systems {solventþbenzeneþcyclohexane} at 298.15 K.Error bars represent the standard deviations.

Table 3NRTL parameters regressed from the experimental LLE data for the ternary system{[C4mim][AlCl4] or sulfolane (1)þbenzene (2)þcyclohexane (3)} at 298.15 K.

Component NRTL parameters rmsd

i–j (Δgij/R)/K (Δgji/R)/K αij

{[C4mim][AlCl4] (1)þbenzene (2)þcyclohexane (3)}1–2 �1613.92 184.48 0.3 0.00441–3 344.13 1127.42 0.32–3 1635.88 �1259.25 0.3{sulfolane (1)þbenzene (2)þcyclohexane (3)}a

1–2 81.45 �79.28 0.31–3 636.79 1857.73 0.2 0.00842–3 1063.45 �707.30 0.3

a The LLE data are from Chen et al. (2000).

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194 191

S/F or the number of stage increases, both the cyclohexane productpurity (Fig. 6a) and the benzene recovery ratio (Fig. 6b) are enhanced.But this improvement is very limited when S/F or the number of stagesreaches a high value. For example, taking S/F¼1.2, when the numberof stages increases from 4 to 6, there are significant improvements ofthe cyclohexane product purity (from 98.92 wt% to 99.47 wt%) and thebenzene recovery ratio (from 93.79% to 97.00%), while with the stagenumber changing from 6 to 8, the improvement of the cyclohexaneproduct purity (from 99.47 wt% to 99.63 wt%) and benzene recoveryratio (from 97.00% to 97.88%) is almost negligible. Based on the aboveanalysis, the number of stages and the solvent-to-feed ratio for[C4mim][AlCl4] were finally chosen to be n¼6 and S/F¼1.49,respectively.

With the obtained operation parameters, the extraction processusing [C4mim][AlCl4] as solvent was simulated. The main unitconfigurations with the stream flow rate and component massfraction information as well as the energy consumptions are givenin Fig. 7a. The total mass flow rate of the cyclohexane/benzenefeed was set to 8310.0 kg/h (100 kmol/h). The finally achievedmass purity of the cyclohexane product was 99.65 wt% and therecovery ratio of benzene was 98.03%.

Before returning to the top of the extractor, the recycled ILstream from Flash B is used to preheat the extract phase stream toits bubble point in order to reduce the total energy consumption.The calculated energy of the heat exchanger is 161.0 kW. Aftersuch heat exchange, the total energy consumption is mainlycaused by the regeneration of solvent and condensation of theflashed vapor. From Fig. 7a, the total energy consumption forsolvent regeneration is 1686.2 kW, and for cooling of the recycledIL and the flashed vapor it is �1629.0 kW.

The extraction process was also simulated with sulfolane asthe solvent under the same feed and extractor conditions as inthe case of [C4mim][AlCl4]. The sulfolane usage is determined bythe separation requirements, as defined in the [C4mim][AlCl4]process. Similarly, the extract phase stream is preheated by therecycled solvent stream, and the exchanged energy is 723.4 kW.Instead of flash units, a distillation column is used for thesolvent regeneration, as shown in Fig. 7b. The energy requiredfor heating in the reboiler is 5572.6 kW and for cooling in thecondenser it is �5479.1 kW. These energy consumptions areabout 3.3 and 3.4 times of those in the [C4mim][AlCl4] process,respectively. The much higher energy consumption of thesulfolane process is mainly attributed to its larger amount ofrecycled solvent (43,998.5 kg/h), compared to 12,400.0 kg/h inthe [C4mim][AlCl4] process. Moreover, the cyclohexane recovery

in the sulfolane process (87.70%) is less than that in the [C4mim][AlCl4] process (90.10%), due to the lower extraction selectivityof sulfolane.

In summary, compared to conventional organic solvents, theionic liquid [C4mim][AlCl4] is demonstrated to be a suitablealternative solvent for the separation of benzene and cyclohexane.This solvent leads to much low energy consumption and a highproduct yield. Considering the stability problem of [C4mim][AlCl4]in the presence of water, an inert gas could be used to purgehumid air out of the process during the start-up phase of anindustrial extraction process.

6. Conclusions

For the screening of ILs as solvents for separating the mixture ofbenzene and cyclohexane, the benzene distribution coefficient andsolvent performance index at infinite dilution were calculated byusing the COSMO-RS model for 264 IL solvents (12 cations and 22anions). As a result, [C4mim][AlCl4] was identified as the bestsolvent candidate. The molecular interactions between IL ions andbenzene were analyzed by DFT calculations. The results revealthat, compared to [BF4]� , [AlCl4]� shows a weaker interactionwith [C4mim]þ , which contributes to the reinforcement of CH-πand π–π interactions between [C4mim]þ and benzene. This leadsto the relatively higher capacity of [C4mim][AlCl4] as extractionsolvent. Based on own experimental LLE data, NRTL parameters

2 4 6 8 10 12 14 16 18 20 220.40.60.81.01.21.41.61.82.0

3

4

5

6

7

[C4mim][AlCl4] (This work)

[C4mim][BF4] (Zhou et al., 2012b)

Sufolane (This work)

Mas

s rat

io o

f S/F

Number of stages

Fig. 5. Mass ratio of solvent-to-feed (S/F) plotted with the number of stages for theseparation demands of cyclohexane mass purity 499.50% and benzene recoveryratio 498%.

1.0 1.2 1.4 1.6 1.8 2.0 2.20.970

0.975

0.980

0.985

0.990

0.995

1.000

Mass ratio of S/F

NT=4 NT=6 NT=8 NT=10NT=12

Mas

s pur

ity o

f cyc

lohe

xane

pro

duct

1.0 1.2 1.4 1.6 1.8 2.0 2.20.82

0.84

0.86

0.88

0.90

0.92

0.94

0.96

0.98

1.00

NT=4 NT=6 NT=8 NT=10 NT=12

Mass ratio of S/F

Rec

over

y ra

tio o

f ben

zene

Fig. 6. Mass purity of cyclohexane product (a) and recovery ratio of benzene(b) plotted with the mass ratio of solvent-to-feed (S/F) at different number of stagesfor the solvent [C4mim][AlCl4].

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194192

were regressed and the extraction process with [C4mim][AlCl4] asthe solvent was simulated. The optimal number of stages andsolvent-to-feed mass ratio are n¼6 and S/F¼1.49, respectively. Forthese values, a high cyclohexane product purity of 99.65 wt% and ahigh benzene recovery ratio of 98.03% can be achieved. Moreover,compared to conventional extraction processes where sulfolane isused as the solvent, the proposed extraction process exhibits muchlower energy consumption and a higher product yield. This showsthat [C4mim][AlCl4] is a very attractive solvent for the separationof benzene–cyclohexane mixtures.

Acknowledgements

The authors gratefully acknowledge the support from the MaxPlanck Society in the framework of the Max Planck Partner Groupprogram, the Specialized Research Fund for the Doctoral Programof Higher Education (SRFDP 20120074110008), 863 Program

2012AA061601, Fundamental Research Funds for the CentralUniversities of China, and 111 Project (B08021).

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/110.1016/j.ces.2014.05.032.

References

Anantharaj, R., Banerjee, T., 2010. COSMO-RS-based screening of ionic liquids asgreen solvents in denitrification studies. Ind. Eng. Chem. Res. 49, 8705–8725.

Anantharaj, R., Banerjee, T., 2011a. COSMO-RS based predictions for the desulphur-ization of diesel oil using ionic liquids: effect of cation and anion combination.Fuel Process. Technol. 92, 39–52.

Anantharaj, R., Banerjee, T., 2011b. Quantum chemical studies on the simultaneousinteraction of thiophene and pyridine with ionic liquid. AlChE J. 57, 749–764.

Residual liquid

Cyclohexaneproduct

Feed

Solvent

Recycled ionic liquid

8310.0 kg/hrw1=0.0000w2=0.1400w3=0.8600

NT=6P=1 atmT=298.3K

12161.8 kg/hrw1=0.9959w2=0.0041w3=4 PPM

6461.7 kg/hrw1=0.0000w2=0.0035w3=0.9965

1848.3 kg/hrw1=0.0000w2=0.6171w3=0.3829

288.2 kg/hrw1=0.9988w2=0.0010w3=0.0002

P=0.01 atmT=356.15 K845.0 KW

P=0.01 atmT=523.15 K841.2 KW

-433.9 KW

Recycled ionic liquid

0.0 kg/hrw1=1.0000w2=0.0000w3=0.0000

161.0 KW

-844.7 KW

-350.4 KW

A

B

6749.9 kg/hrw1=0.0426w2=0.0034w3=0.9539

14010.1kg/hrw1=0.8645w2=0.0850w3=0.0505

Feed

Solvent

8310.0 kg/hrw1=0.0000w2=0.1400w3=0.8600

4.8 kg/hrw1=1.0000w2=0.0000w3=0.0000

46018.5 kg/hrw1=0.9560w2=0.0249w3=0.0191

6291.5 kg/hrw1=760 PPMw2=0.0030w3=0.9962

-4827.4 KW

NT=6P=1 atmT=300.0 K

4

R=1.5

43998.5 kg/hrw1=1.0000w2=70 PPMw3=16 PPB

2020.0 kg/hrw1=7 PPMw2=0.5650w3=0.4350

NT=5P=1 atm

T=349.3 K-651.7 KW

T=558.6 K5572.6 KW

Cyclohexaneproduct

Recycled solvent

723.4 KW

Fig. 7. Extraction processes for the separation of benzene and cyclohexane using [C4mim][AlCl4] (a) and sulfolane (b) as the solvent. w—mass fraction, 1—solvent,2—benzene, 3—cyclohexane.

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194 193

Arce, A., Earle, M.J., Rodríguez, H., Seddon, K.R., 2007a. Separation of aromatichydrocarbons from alkanes using the ionic liquid 1-ethyl-3-methylimidazoliumbis{(trifluoromethyl) sulfonyl}amide. Green Chem. 9, 70–74.

Arce, A., Earle, M.J., Rodríguez, H., Seddon, K.R., 2007b. Separation of benzene andhexane by solvent extraction with 1-alkyl-3-methylimidazolium bis{(trifluor-omethyl) sulfonyl}amide ionic liquids: effect of the alkyl-substituent length.J. Phys. Chem. B 111, 4732–4736.

Arce, A., Earle, M.J., Katdare, S.P., Rodríguez, H., Seddon, K.R., 2008a. Application ofmutually immiscible ionic liquids to the separation of aromatic and aliphatichydrocarbons by liquid extraction: a preliminary approach. Phys. Chem. Chem.Phys. 10, 2538–2542.

Arce, A., Earle, M.J., Rodríguez, H., Seddon, K.R., Soto, A., 2008b. 1-Ethyl-3-methylimidazolium bis{(trifluoromethyl) sulfonyl}amide as solvent for theseparation of aromatic and aliphatic hydrocarbons by liquid extraction—extension to C7- and C8-fractions. Green Chem. 10, 1294–1300.

Arce, A., Earle, M.J., Rodríguez, H., Seddon, K.R., Soto, A., 2009. Bis{(trifluoromethyl)sulfonyl}amide ionic liquids as solvents for the extraction of aromatic hydro-carbons from their mixtures with alkanes: effect of the nature of the cation.Green Chem. 11, 365–372.

Arce, A., Earle, M.J., Rodríguez, H., Seddon, K.R., Soto, A., 2010. Isomer effect inthe separation of octane and xylenes using the ionic liquid 1-ethyl-3-methylimidazolium bis{(trifluoromethyl) sulfonyl}amide. Fluid Phase Equilib.294, 180–186.

Banerjee, T., Khanna, A., 2006. Infinite dilution activity coefficients for trihexylte-tradecyl phosphonium ionic liquids: measurements and COSMO-RS prediction.J. Chem. Eng. Data 51, 2170–2177.

Burghoff, B., Goetheer, E.L., Haan, A.B.D., 2008. COSMO-RS-based extractant screen-ing for phenol extraction as model system. Ind. Eng. Chem. Res. 47, 4263–4269.

Bondi, A., 1964. Van der Waals volumes and radii. J. Phys. Chem. 68, 441–451.Cabaço, M.I., Besnard, M., Danten, Y., Coutinho, J.A.P., 2011. Solubility of CO2 in

1-butyl-3-methyl-imidazolium-trifluoro acetate ionic liquid studied by ramanspectroscopy and DFT investigations. J. Phys. Chem. B 115, 3538–3550.

Chen, J., Li, Z., Duan, L., 2000. Liquid–liquid equilibria of ternary and quaternarysystems including cyclohexane, 1-heptene, benzene, toluene, and sulfolane at298.15 K. J. Chem. Eng. Data 45, 689–692.

Diedenhofen, M., Eckert, F., Klamt, A., 2003. Prediction of infinite dilution activitycoefficients of organic compounds in ionic liquids using COSMO-RS. J. Chem.Eng. Data 48, 475–479.

Diedenhofen, M., Klamt, A., 2010. COSMO-RS as a tool for property prediction of ILmixtures—a review. Fluid Phase Equilib. 294, 31–38.

Domańska, U., Pobudkowska, A., Królikowski, M., 2007. Separation of aromatichydrocarbons from alkanes using ammonium ionic liquid C2NTf2 atT¼298.15 K. Fluid Phase Equilib. 259, 173–179.

Domanska, U., Zawadzki, M., Paduszynski, K., Krolikowsk, M., 2012. Perturbed-chainSAFT as a versatile tool for thermodynamic modeling of binary mixturescontaining isoquinolinium ionic liquids. J. Phys. Chem. B 116, 8191–8200.

Dong, K., Zhang, S., Wang, D., Yao, X., 2006. Hydrogen bonds in imidazolium ionicliquids. J. Phys. Chem. A 110, 9775–9782.

Dong, K., Song, Y., Liu, X., Cheng, W., Yao, X., Zhang, S., 2012. Understandingstructures and hydrogen bonds of ionic liquids at the electronic level. J. Phys.Chem. B 116, 1007–1017.

Dong, K., Zhang, S., 2012. Hydrogen bonds: a structural insight into ionic liquids.Chem. Eur. J. 18, 2748–2761.

Ferreira, A.R., Freire, M.G., Ribeiro, J.C., Lopes, F.M., Crespo, J.G., Coutinho, J.A., 2012.Overview of the liquid–liquid equilibria of ternary systems composed of ionicliquid and aromatic and aliphatic hydrocarbons, and their modeling by COSMO-RS. Ind. Eng. Chem. Res. 51, 3483–3507.

Foco, G.M., Bottini, S.B., Quezada, N., de la Fuente, J.C., Peters, C.J., 2006. Activitycoefficients at infinite dilution in 1-alkyl-3-methylimidazolium tetrafluorobo-rate ionic liquids. J. Chem. Eng. Data 51, 1088–1091.

Frisch, M.J., Trucks, G.W., Schlegel, H.B., Scuseria, G.E., Robb, M.A., Cheeseman, J.R.,et al., 2004. Gaussian 03, Revision C.02. Gaussian, Inc., Wallingford CT.

Geerlings, P., de Proft, F., Langenaeker, W., 2003. Conceptual density functionaltheory. Chem. Rev. (Columbus) 103, 1793–1873.

González, E.J., Domínguez, I., González, B., Canosa, J., 2010a. Liquid–liquid equilibriafor ternary systems of {cyclohexaneþaromatic compoundsþ1-ethyl-3-methyl-pyridinium ethylsulfate}. Fluid Phase Equilib. 296, 213–218.

Gonza lez, E.J., Calvar, N., Gonza lez, B., Dominguez, A., 2010b. Liquid extraction ofbenzene from its mixtures using 1-ethyl-3-methylimidazolium ethylsulfate as asolvent. J. Chem. Eng. Data 55, 4931–4936.

Gutie rrez, J.P., Meindersma, W., de Haan, A.B., 2010. Ionic liquids selection andperformance evaluation for the separation of methylcyclohexane/toluene byextractive distillation. Distill. Absorpt., 473–478.

Gutierrez, J.P., Meindersma, G.W., de Haan, A.B., 2012. COSMO-RS-based ionic-liquidselection for extractive distillation processes. Ind. Eng. Chem. Res. 51,11518–11529.

Huddleston, J.G., Willauer, H.D., Swatloski, R.P., Visser, A.E., Rogers, R.D., 1998. Roomtemperature ionic liquids as novel media for ‘clean’ liquid–liquid extraction.Chem. Commun. 16, 1765–1766.

Huddleston, J.G., Visser, A.E., Reichert, W.M., Willauer, H.D., Broker, G.A., Rogers, R.D., 2001. Characterization and comparison of hydrophilic and hydrophobicroom temperature ionic liquids incorporating the imidazolium cation. GreenChem. 3, 156–164.

John, S.W., Joseph, A.L., Robert, A.W., Charles, L.H., 1982. Dialkylimidazoliumchloroaluminate melts: a new class of room-temperature ionic liquids forelectrochemistry, spectroscopy and synthesis. Inorg. Chem. 21, 1263–1264.

Klamt, A., Eckert, F., 2000. COSMO-RS: a novel and efficient method for the a prioriprediction of thermophysical data of liquids. Fluid Phase Equilib. 172, 43–72.

Klamt, A., Eckert, F., 2004. Prediction of vapor liquid equilibria using COSMOtherm.Fluid Phase Equilib.217, 53–57.

Lei, Z., Chen, B., Li, C., 2007. COSMO-RS modeling on the extraction of stimulantdrugs from urine sample by the double actions of supercritical carbon dioxideand ionic liquid. Chem. Eng. Sci. 62, 3940–3950.

Lü, R., Lin, J., Qu, Z., 2012. Theoretical study on interactions between thiophene/dibenzothiophene/cyclohexane/toluene and 1-methyl-3-octylimidazolium tet-rafluoroborate. Struct. Chem. 24, 1–9.

Meindersma, G.W., de Haan, A.B., 2008. Conceptual process design for aromatic/aliphatic separation with ionic liquids. Chem. Eng. Res. Des. 86, 745–752.

Meindersma, G.W., Hansmeier, A.R., de Haan, A.B., 2010. Ionic liquids for aromaticsextraction. Present status and future outlook. Ind. Eng. Chem. Res. 49,7530–7540.

Meindersma, G.W., de Haan, A.B., 2012. Cyano-containing ionic liquids for theextraction of aromatic hydrocarbons from an aromatic–aliphatic mixture. Sci.Chin. Chem. 55, 1488–1499.

Nockemann, P., Thijs, B., Pittois, S., Thoen, J., Glorieux, C., Van Hecke, K., Binnemans,K., 2006. Task-specific ionic liquid for solubilizing metal oxides. J. Phys. Chem. B110, 20978–20992.

Paduszynski, K., Domanska, U., 2012. Thermodynamic modeling of ionic liquidsystems: development and detailed overview of novel methodology based onthe PC-SAFT. J. Phys. Chem. B 116, 5002–5018.

Renon, H., Prausnitz, J.M., 1968. Local compositions in thermodynamic excessfunctions for liquid mixtures. AlChE J. 14, 135–144.

Rezabal, E., Schäfer, T., 2013. First principle approach to solvation bymethylimidazolium-based ionic liquids. J. Phys. Chem. B 117, 553–562.

Schneider, D.F., 2004. Reactions and separations avoid sulfolane regenerationproblems. Chem. Eng. Prog. 100, 34–39.

Seddon, K.R., 1997. Ionic liquids for clean technology. J. Chem. Technol. Biotechnol.68, 351–356.

Seiler, M., Jork, C., Kavarnou, A., Arlt, W., Hirsch, R., 2004. Separation of azeotropicmixtures using hyperbranched polymers or ionic liquids. AlChE J. 50,2439–2454.

Wang, R., Wang, J., Meng, H., Li, C., Wang, Z., 2008a. Liquid–liquid equilibria forbenzeneþcyclohexaneþ1-methyl-3-methylimidazolium dimethylphosphateor þ1-ethyl-3-methylimidazolium diethylphosphate. J. Chem. Eng. Data 3,1159–1162.

Wang, R., Li, C., Meng, H., Wang, J., Wang, Z., 2008b. Ternary liquid–liquid equilibriameasurement for benzeneþcyclohexaneþN-methylimidazole, or N-ethylimi-dazole, or N-methylimidazolium dibutylphosphate at 298.2 K and atmosphericpressure. J. Chem. Eng. Data 53, 2170–2174.

Zhang, J., Huang, C., Chen, B., Ren, P., Lei, Z., 2007. Extraction of aromatichydrocarbons from aromatic/aliphatic mixtures using chloroaluminate room-temperature ionic liquids as extractants. Energy Fuels 21, 1724–1730.

Zhao, X., Xing, H., Yang, Q., Li, R., Su, B., Bao, Z., Ren, Q., 2012. Differential solubilityof ethylene and acetylene in room-temperature ionic liquids: a theoreticalstudy. J. Phys. Chem. B 116, 3944–3953.

Zhou, J., Mao, J., Zhang, S., 2008. Ab initio calculations of the interaction betweenthiophene and ionic liquids. Fuel Process. Technol. 89, 1456–1460.

Zhou, T., Wang, Z.Y., Chen, L.F., Ye, Y.M., Qi, Z.W., Freund, H., Sundmacher, K., 2012a.Evaluation of the ionic liquids 1-alkyl-3-methylimidazolium hexafluoropho-sphate as a solvent for the extraction of benzene from cyclohexane:(liquidþ liquid) equilibria. J. Chem. Thermodyn. 48, 145–149.

Zhou, T., Wang, Z.Y., Ye, Y.M., Chen, L.F., Xu, J., Qi, Z.W., 2012b. Deep separation ofbenzene from cyclohexane by liquid extraction using ionic liquids as thesolvent. Ind. Eng. Chem. Res. 51, 5559–5564.

Z. Lyu et al. / Chemical Engineering Science 115 (2014) 186–194194