I1

Aa

b

a

ARRAA

KILAAS

1

rtiavttbontbmwp

hwwb

oU

0d

Fluid Phase Equilibria 294 (2010) 180–186

Contents lists available at ScienceDirect

Fluid Phase Equilibria

journa l homepage: www.e lsev ier .com/ locate / f lu id

somer effect in the separation of octane and xylenes using the ionic liquid-ethyl-3-methylimidazolium bis{(trifluoromethyl)sulfonyl}amide

lberto Arcea, Martyn J. Earleb,∗, Héctor Rodrígueza,b, Kenneth R. Seddonb, Ana Sotoa

Department of Chemical Engineering, University of Santiago de Compostela, E-15782 Santiago de Compostela, SpainThe QUILL Research Centre, The Queen’s University of Belfast, Belfast BT9 5AG, UK

r t i c l e i n f o

rticle history:eceived 23 November 2009eceived in revised form 26 January 2010ccepted 28 January 2010

a b s t r a c t

The liquid–liquid equilibria of three ternary systems comprising the aliphatic hydrocarbon octane, anaromatic hydrocarbon 1,n-(CH3)2C6H4 (where n takes the values 2, 3, or 4; that is, o-xylene, m-xylene,or p-xylene), and the ionic liquid 1-ethyl-3-methylimidazolium bis{(trifluoromethyl)sulfonyl}amide([C2mim][NTf2]), were experimentally determined at 25 ◦C. The results show a more favourable dis-

vailable online 4 February 2010

eywords:onic liquidiquid–liquid equilibrium

tribution ratio towards the ionic liquid-rich phase for 1,2-(CH3)2C6H4 than for 1,3-(CH3)2C6H4 or1,4-(CH3)2C6H4. Although selectivity values prove that it is feasible to separate the aliphatic and aro-matic hydrocarbons by using [C2mim][NTf2] as solvent, no significant differences are observed in theselectivity among systems. Thus, the proposed ionic liquid does not look suitable to perform the separa-tion of the different xylene isomers in a mixture. The experimental data were satisfactorily correlated by

n-ran

romatic hydrocarbonliphatic hydrocarbonolvent extraction

means of the classical ‘no

. Introduction

Extraction is a fundamental technique in industrial separations,anking second only after distillation [1]. The largest-volume indus-rial application of solvent extraction to date is in the petrochemicalndustry, where heat-sensitive liquid feeds need to be separatedccording to chemical type rather than by molecular weight orapour pressure [2]. The key to an effective extraction process ishe selection of an appropriate solvent. A revisited list of criteriahat are important to consider during the selection of a solvent haseen recently published by de Haan and Bosch [2]. The potentialf extraction as a separation technique is somehow limited by theumber of conventional solvents with the adequate characteristicso be used in the processes of interest. Ionic liquids, which haveeen the object of intensive research over the last decade [3], offerillions of combinations of cations and anions to generate solventsith different and interesting sets of properties [3,4], and thereforelenty of renewed possibilities in solvent extraction.

The separation of aromatic and aliphatic hydrocarbons

as become a paradigm of application of solvent extractionithin the field of petrochemistry [5,6]. In previous work [7],e showed that the ionic liquid 1-ethyl-3-methylimidazolium

is{(trifluoromethyl)sulfonyl}amide ([C2mim][NTf2]) could per-

∗ Corresponding author at: The QUILL Research Centre, The Queen’s Universityf Belfast, David Keir Building, Stranmillis Road, Belfast BT9 5AG, Northern Ireland,K. Tel.: +44 2890975591/2890975420; fax: +44 2890665297/2890665297.

E-mail address: quill@qub.ac.uk (M.J. Earle).

378-3812/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.fluid.2010.01.029

dom two-liquid’ (NRTL) model.© 2010 Elsevier B.V. All rights reserved.

form comparably well or better than the best conventional solvent(sulfolane) currently used to accomplish the aforementioned sepa-ration, exemplified with hexane and benzene. We also investigatedthe ability of this ionic liquid for the separation of the pairs hep-tane/toluene and octane/ethylbenzene, through the determinationof the liquid–liquid equilibrium (LLE) of the corresponding ternarysystems [8]. A decay in selectivity was observed as the molecularweight of the hydrocarbons was increased. Herein, this research iscomplemented by studying the phase equilibria of liquid systemsincluding [C2mim][NTf2] and other C8-fraction hydrocarbon pairs,namely octane and xylenes, at 25 ◦C. As previously investigated,only a weak influence of the temperature is expected for these kindof systems [7], so no other temperatures will be investigated.

In order to enhance their usefulness, and to facilitate theircomputerised treatment, the experimental LLE data have been cor-related by means of the classical ‘non-random two-liquid’ (NRTL)equation [9]. Although not conceptually intended for systemsinvolving electrolytes, the classical NRTL model has proven inrecent publications to adequately correlate this type of data in anal-ogous ternary systems involving ionic liquids and aromatic andaliphatic hydrocarbons [7,8,10–21].

2. Experimental

2.1. Materials

Octane was purchased from Merck KGaA, with a nominal purityof 99%. The aromatic hydrocarbons o-xylene (1,2-(CH3)2C6H4),

Equilibria 294 (2010) 180–186 181

mstca

eoNdtno

2

(4pTtTatpdta

cboincfswTbpazgawptaBl

3

3

(s(sbte

Table 1Composition of the experimental tie-line ends, and values of the solutedistribution ratio (ˇ) and selectivity (S), for the ternary systems (octane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]), where n = 2, 3, or 4, at 25 ◦C. The mole fractionsof octane, 1,n-(CH3)2C6H4 and [C2mim][NTf2] are represented by x1 , x2 and x3 ,respectively.

Upper phase Lower phase ˇ S

x1 x2 x3 x1 x2 x3

Octane + 1,2-(CH3)2C6H4 + [C2mim][NTf2]1.000 0.000 0.000 0.015 0.000 0.985 – –0.893 0.107 0.000 0.015 0.063 0.922 0.59 35.10.796 0.204 0.000 0.016 0.120 0.864 0.59 29.30.688 0.312 0.000 0.015 0.173 0.812 0.55 25.40.562 0.438 0.000 0.014 0.229 0.757 0.52 21.00.464 0.536 0.000 0.017 0.278 0.705 0.52 14.20.362 0.638 0.000 0.014 0.326 0.660 0.51 13.20.246 0.754 0.000 0.012 0.383 0.605 0.51 10.40.127 0.873 0.000 0.007 0.445 0.548 0.51 9.20.000 1.000 0.000 0.000 0.523 0.477 0.52 –

Octane + 1,3-(CH3)2C6H4 + [C2mim][NTf2]1.000 0.000 0.000 0.015 0.000 0.985 – –0.928 0.072 0.000 0.014 0.034 0.952 0.47 31.30.837 0.163 0.000 0.014 0.078 0.908 0.48 28.50.754 0.246 0.000 0.014 0.115 0.871 0.47 25.20.631 0.369 0.000 0.012 0.167 0.821 0.45 23.80.499 0.501 0.000 0.014 0.227 0.759 0.45 16.10.381 0.619 0.000 0.013 0.280 0.707 0.45 13.30.337 0.663 0.000 0.015 0.300 0.685 0.45 10.20.231 0.769 0.000 0.012 0.352 0.636 0.46 8.80.114 0.886 0.000 0.006 0.408 0.586 0.46 8.70.000 1.000 0.000 0.000 0.481 0.519 0.46 –

Octane + 1,4-(CH3)2C6H4 + [C2mim][NTf2]1.000 0.000 0.000 0.015 0.000 0.985 – –0.938 0.062 0.000 0.014 0.030 0.956 0.48 32.40.858 0.142 0.000 0.014 0.066 0.920 0.46 28.50.776 0.224 0.000 0.015 0.106 0.879 0.47 24.50.683 0.317 0.000 0.016 0.145 0.839 0.46 19.50.568 0.432 0.000 0.013 0.196 0.791 0.45 19.80.472 0.528 0.000 0.013 0.235 0.752 0.45 16.20.361 0.639 0.000 0.012 0.284 0.704 0.44 13.40.279 0.721 0.000 0.010 0.322 0.668 0.45 12.5

A. Arce et al. / Fluid Phase

-xylene (1,3-(CH3)2C6H4), and p-xylene (1,4-(CH3)2C6H4), wereupplied by Fluka, with nominal purities greater than 99.5%. Allhese hydrocarbons were used as received, without further purifi-ation. No trace of isomer mixture/contamination was detected forny of the xylenes by 1H NMR.

The ionic liquid [C2mim][NTf2] was synthesised as reportedlsewhere [7], by a metathetic reaction between the chloride saltf the cation and the lithium salt of the anion. The 1H NMR and 13CMR spectra of the slightly yellowish final product confirmed theesired structure, and its water content was measured and foundo be below 200 ppm by Karl-Fischer titration. The absence of a sig-ificant amount of residual chloride ions was confirmed by meansf the silver(I) nitrate test.

.2. LLE experiments

For the determination of the LLE of the ternary systemsoctane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]), with n being 2, 3, or, multiple heterogeneous mixtures involving the three com-ounds (or just two, for the immiscible pairs) were prepared.hese mixtures were placed inside especially designed glass cells,hermostatted to 25.0 ◦C by means of a circulating water bath.he mixtures were vigorously stirred for more than 1 h, and thenllowed to settle for not less than 4 h, which was found to be enoughime for equilibrium to be reached in the system with completehase separation. After that, a sample of each phase was taken andissolved in deuteriated methanol (Aldrich, 99.8+ atom% D, con-aining 0.03% (v/v) TMS), inside NMR tubes, immediately sealednd stored for compositional analysis.

The compositional analysis of the phases in equilibrium wasarried out by 1H NMR spectroscopy. This technique has alreadyeen described in the literature for calculating the compositionsf tie-line ends in LLE experiments of analogous systems involvingonic liquids and hydrocarbons [7,8,15,16,20,22]. Briefly, the tech-ique consists of the selection of a set of peaks of the differentompounds involved in the 1H NMR spectra, from which the molarractions can be calculated. For the appropriate selection of peaks,everal homogeneous mixtures close to the immiscibility domainere prepared by weight, and their 1H NMR spectra were recorded.

he set of peaks selected was that which led to the lowest deviationsetween actual and calculated compositions for these testing sam-les. Specifically, the signals chosen were: those of the methylenend methyl groups attached to the nitrogen atoms of the imida-olium ring, for the ionic liquid; the peak of the terminal methylroups, for the octane; and both the signals of the protons in theromatic ring and the methyl groups for the xylenes. The deviationsere found to be lower than 0.017 (lower phase) or 0.007 (upperhase) in molar fraction, in any proportion of the compounds nearhe immiscibility curve. All the 1H NMR spectra for quantitativenalysis were run in either a Bruker AMX-500 spectrometer or aruker DRX-500 spectrometer, with relaxation time of 20 s and at

east 32 scans.

. Results and discussion

.1. LLE data

The experimental LLE data at 25 ◦C for the ternary systemsoctane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]), where n = 2, 3, or 4, arehown in Table 1. No ionic liquid was observed in the upper

hydrocarbon-rich) phase, for any global composition in any of theystems. Nevertheless, a very small amount of ionic liquid maye present in the lighter phase, in a concentration low enougho remain undetected. A previous observation in a larger scalexperiment for a similar system [7], which unveiled a concentra-

0.189 0.811 0.000 0.008 0.364 0.628 0.45 10.60.113 0.887 0.000 0.007 0.408 0.585 0.46 7.40.047 0.953 0.000 0.003 0.444 0.553 0.47 7.30.000 1.000 0.000 0.000 0.470 0.530 0.47 –

tion of the ionic liquid in the range of few parts per million inthe hydrocarbon-rich phase, seems to point in this direction. Thelower (ionic liquid-rich) phase, or extract phase, is composed ofa mixture of ionic liquid and hydrocarbons (mainly aromatics) indifferent proportions. Given the practically non-volatile characterof the ionic liquid at the typical conditions of operation in extrac-tion units, a clean separation of the ionic liquid from the extractedhydrocarbons would be rendered feasible by, e.g. distillation tech-niques. Thus, it would be possible to fully recycle the ionic liquidwith virtually no loss.

The corresponding triangular diagrams of the LLE of the ternarysystems are shown in Fig. 1, along with that of the system(octane + ethylbenzene + [C2mim][NTf2]) [8], also at 25 ◦C, for directvisual comparison of all four systems involving the same ionic liq-uid and a pair of aliphatic and aromatic hydrocarbons belongingto the C8-fraction. All these systems, according to the classificationproposed by Sørensen et al. [23], correspond to a Type 2 category,in which two of the pairs of compounds exhibit partial miscibility,and there is only one continuum immiscibility domain throughoutthe compositional spectrum. The tie-lines that lie on the edges of

the triangular diagrams graphically indicate how the xylenes aremuch more soluble in the ionic liquid than in octane, and that allsolubilities of the different xylene isomers in [C2mim][NTf2] arequite similar, both among them and compared to that of ethylben-zene. There are no tie-lines with positive slope, which would mean

182 A. Arce et al. / Fluid Phase Equilibria 294 (2010) 180–186

Fig. 1. Experimental (solid circles and solid lines) tie-lines for the LLE at 25 ◦C of the ternary systems (a) (octane + 1,2-(CH3)2C6H4 + [C2mim][NTf2]), (b) (octane + 1,3-(CH ) C H + [C mim][NTf ]), (c) (octane + 1,4-(CH ) C H + [C mim][NTf ]), and (d) (octane + ethylbenzene + [C mim][NTf ]) [8]. For the novel three systems reported in thep uares,

a(lep

rllafd

ˇ

S

w(sp

s

3 2 6 4 2 2 3 2 6 4 2 2

resent work, the correlated tie-lines (NRTL, with ˛ = 0.20) are also drawn (open sq

(desirable) greater concentration of xylene in the lower phaseextract) than in the upper phase (raffinate). All left ends of the tie-ines lie on the octane-xylene edge of the triangular diagrams, asxpected from the absence of ionic liquid in the hydrocarbon-richhases.

Solute distribution ratio (ˇ) and selectivity (S) are the mostelevant criteria in the evaluation of a solvent for a particulariquid–liquid extraction. These two parameters have been calcu-ated from the LLE data reported in this work, to evaluate thebility of [C2mim][NTf2] as solvent for the extraction of the dif-erent xylene isomers from their mixtures with octane. ˇ and S areefined according to Eqs. (1) and (2):

= xII2

xI2

(1)

= xII2/xI

2

xII1/xI

1

(2)

here x is mole fraction, subscripts 1 and 2 refer to the carrier

octane) and solute (1,n-(CH3)2C6H4) compounds respectively, anduperscripts I and II refer to the upper phase (raffinate) and lowerhase (extract) respectively.

The calculated values of ˇ and S for the three ternary systemstudied are shown in the columns on the right of Table 1. Consis-

2 2

dashed lines).

tent with the negative slope of the tie-lines in Fig. 1, the valuesof ˇ are lower than unity. They remain quite constant with vari-ation of the global composition of the system, and interestinglythey are higher in the system with 1,2-(CH3)2C6H4 (values in therange 0.51–0.59) than in the systems with 1,3-(CH3)2C6H4 or 1,4-(CH3)2C6H4, for which no significant difference is observed (valuesin the range 0.44–0.48). On the contrary, selectivity values clearlydecrease, for all three systems, as the concentration of aromatichydrocarbon is increased.

Fig. 2a shows the solute distribution ratios represented asa function of the mole fraction of the aromatic hydrocar-bon in the raffinate. The values for the analogous system(octane + ethylbenzene + [C2mim][NTf2]), also at 25 ◦C [8], are alsoplotted, for comparison. As mentioned above, the values are higherfor the system with 1,2-(CH3)2C6H4 compared to the other twosystems with 1,3-(CH3)2C6H4 or 1,4-(CH3)2C6H4. This may be theresult of the higher dipole moment of 1,2-(CH3)2C6H4 (0.62 D)compared to those of 1,3-(CH3)2C6H4 and 1,4-(CH3)2C6H4 (0.33 Dand 0.00 D, respectively) [24]; thus making the partitioning of 1,2-

(CH3)2C6H4 to the ionic liquid-rich phase more favourable. This issomehow expressed as well in the slightly (but significantly) highersolubility of 1,2-(CH3)2C6H4 in [C2mim][NTf2] compared to theother two isomers, as can be seen from the mole fractions of xylenesin the lower phases of the last experimental tie-lines of the different

A. Arce et al. / Fluid Phase Equili

Fig. 2. Solute distribution ratio (a) and selectivity (b), as a function ofthe mole fraction of solute in the hydrocarbon-rich phase (x2), for theternary systems (octane + 1,2-(CH3)2C6H4 + [C2mim][NTf2]) (�), (octane + 1,3-(CH3)2C6H4 + [C2mim][NTf2]) (�), (octane + 1,4-(CH3)2C6H4 + [C2mim][NTf2]) (�),af(2

Llisarbtu1F

p1abstizaatbo

nd (octane + ethylbenzene + [C2mim][NTf2]) (�) [8], at 25 ◦C. For comparison, valuesrom data in the literature are included for the ternary systems (octane + 1,3-CH3)2C6H4 + sulfolane) at 50 ◦C (�) [25], (octane + 1,4-(CH3)2C6H4 + sulfolane) at5 ◦C (�) [26], and (octane + “xylene” + sulfolane) at 25 ◦C (�) [27].

LE data sets, which correspond to binary mixtures (xylene + ioniciquid). This is also observed for the data points plotted for x2 = 1.0n Fig. 2a. However, if strictly based on this rationale, the lack ofignificant difference in the distribution ratios of 1,3-(CH3)2C6H4nd 1,4-(CH3)2C6H4, in spite of their different dipole moments,emains an enigma. On the other hand, the system with ethyl-enzene, whose dipole moment (0.59 D) is intermediate betweenhose of 1,2-(CH3)2C6H4 and 1,3-(CH3)2C6H4 [24], leads to ˇ val-es which approximately lie in between those of the systems with,2-(CH3)2C6H4 and with 1,3-(CH3)2C6H4 or 1,4-(CH3)2C6H4 (seeig. 2a), in accordance with the dipole moment hypothesis.

Regarding their evolution with change in the system com-osition, the ˇ values for the systems with 1,3-(CH3)2C6H4 and,4-(CH3)2C6H4 remain remarkably constant. Although there is notstrong variation in the system with 1,2-(CH3)2C6H4, ˇ seems toe somewhat higher at low levels of aromatic hydrocarbon in theystem, while decreasing and approaching the values of the sys-ems with the other two xylene isomers as the content in xylenencreases. The data corresponding to the system with ethylben-ene look closer to the values for the systems with 1,3-(CH3)2C6H4

nd 1,4-(CH3)2C6H4 at low concentrations of aromatic hydrocarbonnd, as the latter increase, the values tend to match those of the sys-em with 1,2-(CH3)2C6H4. As a result of all these observations, it cane said that the higher dipole moment of 1,2-(CH3)2C6H4 over thether aromatic isomers seems to play a relevant role in leading to

bria 294 (2010) 180–186 183

higher values of ˇ at low concentrations of aromatic hydrocarbonin the system (when the extract phase comprises mainly ionic liq-uid and small amounts of hydrocarbons); whereas the differencesamong ˇ of the different systems are reduced as the aromatic con-tent of the system increases (when the extract phases contain alarge amount of hydrocarbons, and therefore its overall “ionic/polarcharacter” becomes diminished). Greater detail on the latter ratio-nale, for analogous systems, can be found in a previous work [22].

The selectivities of all three systems (octane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]) studied are represented in Fig. 2b,as a function of the mole fraction of the aromatic hydrocarbonin the raffinate. Again, the values corresponding to the system(octane + ethylbenzene + [C2mim][NTf2]), at the same tempera-ture, are included for comparison. No significant differences areobserved among these systems. Perhaps the S values for the sys-tem with ethylbenzene are a little lower, but in general the samedecreasing pattern is observed as the concentration of aromaticsincreases. The selectivities are quite acceptable when the presenceof aromatic hydrocarbons in the system is low, but they dropconsiderably as the system gets richer in aromatic content, inaccordance with the results previously obtained for other systemsconstituted by an aliphatic hydrocarbon, an aromatic hydrocarbonand the ionic liquid [C2mim][NTf2] [7,8].

Sulfolane, which offers good thermal and hydrolytic stability,high density and boiling point, and a good balance of solvent prop-erties [5], performs well and is commonly accepted as the bestsolvent for the separation of aromatic and aliphatic hydrocarbonsby liquid–liquid extraction. Thus, it can be considered a good bench-mark for alternative solvents proposed for such separation process,such as our ionic liquid. Some LLE were available in the literature forternary systems involving octane, xylene (sometimes not specify-ing which isomer, or if a mixture was used, just saying “xylene”) andsulfolane [25–27]. Those reported at 25 ◦C or the closest tempera-ture available were selected. The resulting ˇ and S are plotted inFig. 2 (open symbols) along with our original data. While no signif-icant differences in selectivity are observed, the solute distributionratios are clearly better in the case of the ionic liquid, throughoutthe composition range. In addition, from the original LLE data sets,it can be seen that sulfolane enters the raffinate phase in appre-ciable amounts, which might be a further advantage for the use of[C2mim][NTf2].

Also, some articles in the literature have already exploredthe separation of some aromatic and aliphatic hydrocarbonsof the C8-fraction by means of ionic liquids as alterna-tive solvents for liquid extraction [14,15,28]. In particular,LLE data sets were available for the system (octane + 1,3-(CH3)2C6H4 + ethyl(2-hydroxyethyl)dimethylammonium bis{(tri-fluoromethylsulfonyl)amide} ([N1 1 2 2OH][NTf2])) at 25 ◦C[15], and for the systems (octane + 1,3-(CH3)2C6H4 + 1-butyl-4-methylpyridinium tetrafluoroborate ([C4mpy][BF4])) and(octane + ethylbenzene + [C4mpy][BF4]) at 50 ◦C [14]. The ˇ and Svalues from these works are plotted in Fig. 3, along with the orig-inal data of this work for the equivalent systems. [C2mim][NTf2]exhibits higher solute distribution ratios than [N1 1 2 2OH][NTf2]and [C4mpy][BF4] in any comparable case. However, in termsof selectivity, the values for the systems with [C4mpy][BF4] arehigher, especially at low concentrations of aromatic hydrocarbon.Although these values correspond to systems at 50 ◦C and not at25 ◦C, the evaluating parameters would be expected to be onlya weak function of temperature (at least at that temperaturerange) [7]. In any case, an important risk associated with the use

of this tetrafluoroborate ionic liquid would be the potential hazardof generation of hydrogen fluoride as a result of its hydrolysis,which may occur in the presence of water, humidity or acid traces[29]. Therefore, [C2mim][NTf2] might still be preferable from anecological point of view.

184 A. Arce et al. / Fluid Phase Equilibria 294 (2010) 180–186

Fig. 3. Solute distribution ratio (a) and selectivity (b), as a func-tion of the mole fraction of solute in the hydrocarbon-rich phase(x2), for the ternary systems (octane + 1,3-(CH3)2C6H4 + [C2mim][NTf2])(�), (octane + ethylbenzene + [C mim][NTf ]) (�) [8], and (octane + 1,3-(s(

3

helaotritpttttbirtmact

Table 2Mean error of the solute distribution ratio (�ˇ) and residual function F forthe NRTL correlation of the LLE data of the ternary systems (octane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]), where n = 2, 3, or 4, at 25 ◦C, for different values ofthe non-randomness parameter ˛.

Xylene isomer in the system ˛ �ˇ F

1,2-(CH3)2C6H4 0.10 1.8 0.31320.20 2.1 0.25430.30 5.3 0.4310

1,3-(CH3)2C6H4 0.10 1.8 0.31540.20 2.2 0.24540.30 5.7 0.3572

The binary interaction parameters resulting from the NRTLcorrelation with ˛ = 0.20 are shown in Table 3, for the threesystems (octane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]). In theory, thebinary interaction parameters for the pair (octane + [C2mim][NTf2])should be the same for all three systems. Quite often, this is not the

Table 3Binary interaction parameters (�gij , �gji) of the NRTL model (˛ = 0.20) obtainedfrom the correlation of the LLE data of the ternary systems (octane (1) + 1,n-(CH3)2C6H4 (2) + [C2mim][NTf2] (3)), where n = 2, 3, or 4, at 25 ◦C.

Xylene isomer inthe system

Components i–j Binary interaction parameters

�gij/J mol−1 �gji/J mol−1

1,2-(CH3)2C6H4 1–2 −2292.3 2747.21–3 13408 6294.42–3 15138 −3215.2

1,3-(CH3)2C6H4 1–2 −2539.1 2784.71–3 13175 6329.3

2 2

CH3)2C6H4 + [N1 1 2 2OH][NTf2]) (�) [15], at 25 ◦C; and for the ternaryystems (octane + 1,3-(CH3)2C6H4 + [C4mpy][BF4]) (�) [14], andoctane + ethylbenzene + [C4mpy][BF4]) (♦) [14], at 50 ◦C.

.2. Data correlation

From a perspective of application, it is generally necessary toave the LLE behaviour described by a suitable correlation of thexperimental data, rather than just the discrete experimental tie-ines. This allows easy interpolation or extrapolation of new data,s well as computational treatment. The most important groupf equations for the correlation of LLE data are those based onhe modelling of the excess Gibbs energy. Among these, the ‘non-andom two-liquid’ (NRTL) model holds a relevant position sincet has given very satisfactory correlations and is probably one ofhe most widely used models for liquid–liquid systems [30], inarticular for those systems involving ionic liquids, as stated inhe Introduction section. Thus, the NRTL model has been used inhis work for the correlation of the three novel LLE data sets ofhe ternary systems studied. The binary interaction parameters ofhe model were calculated by using a computer program designedy Sørensen [31], whose objective functions have been described

n detail in a previous work [8]. The NRTL model includes a non-andomness parameter, ˛, which in a crude way can be identified ashe inverse of the number of nearest neighbours of a molecule in the

ixture, and would be expected to have a value in the range 0.1–0.3pproximately [30]. In principle, the non-randomness parameteran be specified according to a set of rules devised by the authors ofhe model [9], but a value fixed in an empirical way is usually given

1,4-(CH3)2C6H4 0.10 1.9 0.29170.20 2.4 0.23580.30 5.9 0.3805

to it, according to previous experience in the reduction of experi-mental data of other systems. Thus, the following correlation of thedifferent LLE data sets was carried out for three arbitrarily chosenvalues of ˛, covering the range of values that it commonly takes:0.10, 0.20 and 0.30.

The quality of the correlation in each case was evaluated via themean error of the solute distribution ratio, �ˇ, and the residualfunction F, defined as follows:

�ˇ = 100

[1M

∑k

(ˇk − ˆ

k

ˇk

)2]0.5

(3)

F = 100

⎡⎣∑

k

min∑

i

∑j

(xijk − xijk

)2

6M

⎤⎦

0.5

(4)

where x is mole fraction; subscripts and sum indices i, j and k refer tocomponents, phases and tie-lines, respectively; M is the total num-ber of experimental tie-lines; the circumflex symbol on top refersto a calculated value; and “min” corresponds to the minimum valueobtained by the Nelder–Mead method [32]. Table 2 summarises theresiduals obtained for the correlations with the different values of˛ and for the three ternary systems. The best results correspondto the correlations with ˛ = 0.20 for all three systems. Although the�ˇ values are somewhat lower for ˛ = 0.10, the marked decrease ofF for ˛ = 0.20 suggests the latter value as the optimum. For ˛ = 0.30,the quality of the correlation is clearly not so good.

2–3 14828 −2778.1

1,4-(CH3)2C6H4 1–2 −2514.6 2834.21–3 13506 6315.92–3 14995 −2714.0

Equili

cctplbcfat1tipoaraascta

lTscttlsfffTtp

4

uas(hootihpamf

bohttub

u

[

A. Arce et al. / Fluid Phase

ase when obtaining the binary interaction parameters by directorrelation of the LLE data of a ternary system, and the parame-ers obtained act as a mathematical fitting, rather than having anyhysical meaning, as they should according to the original formu-

ation of the NRTL model. However, this time not only qualitativeut also quantitative agreement is achieved for the aforementionedommon pair of compounds. The binary interaction parametersor the other pairs also look qualitatively meaningful, and the rel-tive quantitative differences found between the parameters forhe system with 1,2-(CH3)2C6H4 versus those for the systems with,3-(CH3)2C6H4 or 1,4-(CH3)2C6H4 seem to appropriately mirrorhe differences in the LLE, as discussed before. Obtaining “mean-ngful” binary interaction parameters would be of interest for therediction of LLE of other ternary systems involving the pairsf compounds for which the binary interaction parameters havelready been obtained. An interesting approach in this regard hasecently been presented by Simoni et al. [33], who suggest ansymmetric model based on the NRTL model and on one of itsdaptations to electrolyte systems to predict the LLE of ternaryystems on the basis of only binary interaction parameters thatan be obtained just from binary LLE data. Although the success ofheir predictions is so far limited, this strategy may open importantvenues in the prediction of LLE of ternary systems.

The correlated equivalent tie-lines of the experimental tie-ines, generated from the binary interaction parameters reported inable 3, are plotted in the triangular diagrams of Fig. 1 for all threeystems (octane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]), along with theorresponding experimental tie-lines. A very good match betweenhe experimental and the correlated sets of tie-lines is observed forhe different systems studied. Nevertheless, the absence of ioniciquid in the raffinate, as indicated by the experimental data, is notupported by the NRTL model. Compositions of up to 0.009 in moleraction of ionic liquid, although barely perceptible in Fig. 1, resultrom the correlated tie-lines sets. This issue was also previouslyound in our previous research with analogous systems [7,8,20].his is a critical aspect, since the presence or not of ionic liquid inhe raffinate may largely condition the economics of the extractionrocess in a real, scaled-up plant.

. Conclusions

A further body of information on the ability of the ionic liq-id [C2mim][NTf2] as solvent for the separation of aromatic andliphatic hydrocarbons by liquid–liquid extraction has been pre-ented in this work. In particular, the LLE of the ternary systemsoctane + 1,n-(CH3)2C6H4 + [C2mim][NTf2]), where n = 2, 3, or 4,ave been studied at 25 ◦C, allowing for an analysis of the effectf the isomers of the aromatic C8-fraction on the performancef the ionic liquid as extraction solvent for the desired separa-ion. Together with results published in previous works [7,8,20],n which other ternary systems involving aromatic and aliphaticydrocarbons of the C6- to C8-fractions were investigated, a globalicture can be drawn of the possibilities of [C2mim][NTf2] as anlternative solvent in extracting aromatic hydrocarbons from theirixtures with aliphatic hydrocarbons in naphthas and similar oil

ractions.A good correlation of the LLE data for these systems is possi-

le even with classical models such as NRTL. However, this modelften fails to adequately correlate the absence of ionic liquid in theydrocarbon-rich phase, as observed experimentally; although, onhe other hand, it is possible that some ionic liquid actually enters

hat phase, below the detection limit of the experimental techniquesed, thus partially conditioning the advantages of the ionic liquidased system.

In general it can be said that [C2mim][NTf2] leads to good val-es of selectivity, especially at low or medium concentrations of

[[[

[

bria 294 (2010) 180–186 185

aromatic hydrocarbons in the system. The latter is, in fact, thecase of most interest in industrial application of solvent extrac-tion for the separation of aromatic and aliphatic hydrocarbons [5].Regarding the solute distribution ratios, [C2mim][NTf2] works bet-ter for the C6-fraction than for the C7-fraction, and than for theC8-fraction; so the lower the average fraction of the feed in anextracting process, the lower amount of ionic liquid will be neededto reach a given level of separation. For the C6- and C7-fractions,the distribution ratio diminishes a little as the concentration ofaromatics in the feed increases, whereas it is more insensitiveto this factor in the case of the C8-fraction. A greater partitionof 1,2-(CH3)2C6H4 towards the ionic liquid-rich phase has beenobserved compared to 1,3-(CH3)2C6H4 and 1,4-(CH3)2C6H4, butthe difference is small in relative terms. Therefore, [C2mim][NTf2]can be considered as a good choice for the separation of aromaticand aliphatic hydrocarbons, but the body of LLE data developedon systems of aliphatic hydrocarbon, aromatic hydrocarbon and[C2mim][NTf2], suggests that this ionic liquid will not be ade-quate for the separation of different Cn-fractions or isomers withinmixtures of aromatic hydrocarbons or mixtures of aliphatic hydro-carbons, given the similarities in selectivity.

As analysed in terms of solute distribution ratio and selectivity,[C2mim][NTf2] seems to perform equally well or even better thansulfolane, a well-established solvent of reference for this process inindustry. A more detailed discussion on this point can be found else-where [7,8]. Nonetheless, deeper knowledge of the characteristicsof the ionic liquid, such as toxicity or biodegradability, as well as anoverall economic balance of an scaled-up extraction unit, would beneeded in order to definitely judge the advantageous character of[C2mim][NTf2] (or even other ionic liquids!) over the conventionalsolvents currently used in this sort of processes. It must be notedthat the relative high price of this and other ionic liquids is likelyto critically decrease as they are produced on a much larger scalefor implementation in industrial processes.

Acknowledgements

The authors are grateful to the Industrial Advisory Board of theQUILL Research Centre for their support. K.R.S. acknowledges theEPSRC (Portfolio Partnership Scheme, grant number EP/D029538/1)for support. A.A. and A.S. thank the Ministerio de Educación y Cien-cia (Spain) for financial support through project CTQ2006-07687.

References

[1] H.R.C. Pratt, G.W. Stevens, in: J.D. Thornton (Ed.), Science and Practice ofLiquid–Liquid Extraction, vol. 1, Clarendon Press, Oxford, 1992, pp. 343–344.

[2] A.B. de Haan, H. Bosch, Fundamentals of Industrial Separations, second edition,2007.

[3] A. Stark, K.R. Seddon, in: A. Seidel (Ed.), Kirk-Othmer Encyclopedia of ChemicalTechnology, vol. 26, John Wiley & Sons, Hoboken, 2007, pp. 836–920.

[4] K.R. Seddon, in: S. Boghosian, V. Dracopoulos, C.G. Kontoyannis, G.A. Voyi-atzis (Eds.), The International George Papatheodorou Symposium: Proceedings,Institute of Chemical Engineering and High Temperature Chemical Processes,Patras, 1999, pp. 131–135.

[5] K. Weissermel, H.-J. Arpe, Industrial Organic Chemistry, third edition, VCH,Weinheim, 1997.

[6] R. Wennersten, in: J. Rydberg, M. Cox, C. Musikas, G.R. Choppin (Eds.), SolventExtraction Principles and Practice, second edition, Marcel Dekker, New York,2004, pp. 415–453.

[7] A. Arce, M.J. Earle, H. Rodríguez, K.R. Seddon, Green Chem. 9 (2007) 70–74.[8] A. Arce, M.J. Earle, H. Rodríguez, K.R. Seddon, A. Soto, Green Chem. 10 (2008)

1294–1300.[9] H. Renon, J.M. Prausnitz, AIChE J. 14 (1968) 135–144.10] M.S. Selvan, M.D. McKinkey, R.H. Dubois, J.L. Atwood, J. Chem. Eng. Data 45

(2000) 841–845.

11] T.M. Letcher, N. Deenadayalu, J. Chem. Thermodyn. 35 (2003) 67–76.12] T.M. Letcher, P. Reddy, J. Chem. Thermodyn. 37 (2005) 415–421.13] G.W. Meindersma, A.J.G. Podt, A.B. de Haan, Fluid Phase Equilib. 247 (2006)

158–168.14] G.W. Meindersma, A. Podt, A.B. de Haan, J. Chem. Eng. Data 51 (2006)

1814–1819.

1 Equili

[

[

[[

[

[

[[

[

[

[[[

[

[

[

[

[[

Ana Soto was born in Vigo (Spain) in 1967, and graduatedfrom the University of Santiago de Compostela, with bothBSc and PhD in industrial chemistry. She is currently a fullprofessor at the same university. She has published over100 papers and co-authored two books.

86 A. Arce et al. / Fluid Phase

15] U. Domanska, A. Pobudkowska, M. Królikowski, Fluid Phase Equilib. 259 (2007)173–179.

16] U. Domanska, A. Pobudkowska, Z. Zołek-Tryznowska, J. Chem. Eng. Data 52(2007) 2345–2349.

17] R.M. Maduro, M. Aznar, Fluid Phase Equilib. 265 (2008) 129–138.18] R. Wang, J. Wang, H. Meng, C. Li, Z. Wang, J. Chem. Eng. Data 53 (2008)

1159–1162.19] R. Wang, C. Li, H. Meng, J. Wang, Z. Wang, J. Chem. Eng. Data 53 (2008)

2170–2174.20] A. Arce, M.J. Earle, H. Rodríguez, K.R. Seddon, A. Soto, Green Chem. 11 (2009)

365–372.21] A. Pereiro, A. Rodríguez, J. Chem. Thermodyn. 41 (2009) 951–956.22] A. Arce, M.J. Earle, H. Rodríguez, K.R. Seddon, J. Phys. Chem. B 111 (2007)

4732–4736.23] J.M. Sørensen, T. Magnussen, P. Rasmussen, A. Fredenslund, Fluid Phase Equilib.

2 (1979) 297–309.24] J.G. Speight, Lange’s Handbook of Chemistry, sixteenth edition, McGraw-Hill,

New York, 2005.25] W.-C. Lin, N.-H. Kao, J. Chem. Eng. Data 47 (2002) 1007–1011.26] S. Lee, H. Kim, J. Chem. Eng. Data 40 (1995) 499–503.27] J. Chen, L.-P. Duan, J.-G. Mi, W.-Y. Fei, Z.-C. Li, Fluid Phase Equilib. 173 (2000)

109–119.28] C.C. Cassol, A.P. Umpierre, G. Ebeling, B. Ferrera, S.S.X. Chiaro, J. Dupont, Int. J.

Mol. Sci. 8 (2007) 593–605.29] D.G. Archer, J.A. Widegren, D.R. Kirklin, J.W. Magee, J. Chem. Eng. Data 50 (2005)

1484–1491.30] D.M.T. Newsham, in: J.D. Thornton (Ed.), Science and Practice of Liquid–Liquid

Extraction, vol. 1, Clarendon Press, Oxford, 1992, pp. 1–39.31] J.M. Sørensen, Ph.D. Thesis, Danmarks Tekniske Højskole, Lyngby, Denmark,

1980.32] J.A. Nelder, R. Mead, Comp. J. 7 (1965) 308–313.33] L.D. Simoni, A. Chapeaux, J.F. Brennecke, M.A. Stadtherr, Ind. Eng. Chem. Res.

48 (2009) 7257–7265.

Alberto Arce is a professor of chemical engineering atthe University of Santiago de Compostela (Spain). He hasbeen the head of the department, and leads a researchgroup devoted to separation processes and phase equilib-ria. In the last years he has worked on phase behaviour ofsystems involving tertiary ethers and ionic liquids. He isthe author of more than 150 publications in different SCRjournals.

Martyn John Earle received both his BSc and PhD (organicchemistry) from Loughborough University of Technology.After his PhD, he worked at the Ohio State University for 2years as a post-doctoral researcher before returning to theUK in 1995. He has been employed at the Queen’s Univer-

sity of Belfast since March 1995 and at QUILL ResearchCentre since 1999, where he is currently the assistantdirector. He has over 20 years experience of working in thefield of Chemistry, with specialist knowledge in chemicalreactions and separations in ionic liquids, over 75 papersand patents in this field.

bria 294 (2010) 180–186

Héctor Rodríguez got his PhD from the University of San-tiago de Compostela in December 2006, and right afterhe moved to the University of Alabama as a visiting post-doctoral researcher. For the last two years, he has beena research fellow at the QUILL Research Centre, in theQueen’s University of Belfast. In 2010 he has started anew post back in Santiago de Compostela. His work mainlyfocuses on the use of ionic liquids for the development ofimproved separation processes, with emphasis on ther-modynamics for process design.

Kenneth Richard Seddon was born in Liverpool in 1950,and graduated from Liverpool University with a first classBSc (Hons) and a PhD, whence he moved to a researchfellowship at St Catherine’s College, Oxford, and later toa lectureship in experimental chemistry at the Universityof Sussex. In 1993, he was appointed to the chair in inor-ganic chemistry at the Queen’s University of Belfast, wherehe is also a co-director of QUILL (Queen’s University IonicLiquids Laboratories), an industrial–academic consortiumwhich was awarded the 2006 Queen’s Anniversary Prizefor Higher and Further Education. He is a Professor Cate-drático Visitante at ITQB (New University of Lisbon), andholds a visiting professorship of the Chinese Academy of

Sciences. He has published over 350 papers and patents, co-authored four books,and co-edited eight books.