Design calculations of an extractor for aromatic and aliphatic ...

Chemical Papers 67 (12) 1548–1559 (2013) DOI: 10.2478/s11696-012-0289-1

ORIGINAL PAPER

Design calculations of an extractor for aromatic and aliphatic hydrocarbons separation using ionic liquids Elena Graczová*, Pavol Steltenpohl, Martin Šoltýs, Tomáš Katriňák Institute of Chemical and Environmental Engineering, Faculty of Chemical and Food Technology, Slovak University of Technology in Bratislava, Radlinského 9, 812 37 Bratislava, Slovakia Received 6 July 2012; Revised 25 October 2012; Accepted 29 October 2012

The study concentrates on the separation of aromatic hydrocarbons from aliphatic hydrocarbon mixtures using ionic liquids as a new alternative of extraction solvents. Influence of the phase equilibrium description accuracy on the separation equipment design using different thermodynamic models was investigated. As a model system, a heptane–toluene binary mixture was chosen, employing 1-ethyl-3-methylimidazolium ethyl sulfate (EMIES) ionic liquid as an extractive solvent. Liquid–liquid equilibrium (LLE) data of the ternary system were calculated using NRTL equations with different quality model parameters. Model 1 corresponds to the NRTL equation with the original binary parameters evaluated independently from the respective binary equilibrium data. Model 2 is represented by an NRTL equation extended by the ternary correction term (with the original binary parameters and ternary correction term parameters evaluated from the ternary tie-lines). Model 3, i.e. the NRTL equation with binary model parameters determined via ternary LLE data regression using ASPEN Plus, was taken from Meindersma et al. (2006). Continuous-flow liquidphase extraction was simulated considering a cascade of mixer–settler type extractors according to the Hunter–Nash scheme (Hunter & Nash, 1934). Based on the simulation results, for a preset separation efficiency criterium, different accuracies of the equilibrium description caused serious discrepancies in the separation equipment design, e.g. in the number of theoretical stages, solvent to feed ratio, and product purity. c 2012 Institute of Chemistry, Slovak Academy of Sciences
Keywords: ionic liquid, extraction, theoretical stage, liquid–liquid equilibria, NRTL model, ternary contribution

Introduction Fossil materials (coal, crude oil, natural gas) are a common source of hydrocarbons. In past, aromatics, namely benzene, were obtained as by-products of coal coking. Nowadays, most of the aromatic hydrocarbons (over 90 % of the total consumption) are produced from crude oil, only a small portion is produced from coal (Wauquier, 2000). Practically all liquid fractions obtained by fractional distillation of naphtha contain aromatics; however, due to their low content, their separation is not profitable. In order to increase the aromatics’ content, primary fractions obtained by crude oil distillation are submitted to aromatization and the *Corresponding author, e-mail: [email protected]

obtained aromatics are separated from the product. Aromatic hydrocarbons are typically obtained from the product of catalytic reforming of primary gasoline or from pyrolysis of gasoline after selective hydrogenation of dienes (Blažek & Rábl, 2006). Feed of ethylene producing units in most refineries contains about 10 mass % to 25 mass % of aromatics depending on its source. This stream serves as a source of aromatic hydrocarbons. In praxis, there are several procedures for selective separation of aromatics from hydrocarbon mixtures: liquid extraction for aromatic contents from 20 mass % to 65 mass %, extractive distillation for 65 mass % to 90 mass %, and azeotropic distillation for aromatic content higher

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

than 90 mass %. Consequently, there is no suitable technology for the separation of aromatic compounds comprising less than 20 mass % of the feed (Weissermel & Arpe, 2003; Perreiro et al., 2012). Typical solvents used for the separation of aromatic–aliphatic mixtures are polar components such as sulfolane (Choi et al., 2002; Yorulmaz & Karpuczu, 1985; Schnieder, 2004), N-methylpyrrolidone (NMP), N-formylmorpholine (NFM), ethylene glycols, or propylene carbonate (Krishna et al., 1987; Yorulmaz & Karpuczu, 1985; Ali et al., 2003; Perreiro et al., 2012). When these conventional solvents are used, additional distillation steps for the separation of the solvent from both the extract and the raffinate phases and for the purification of the solvent are required, which increases the separation costs (Perreiro et al., 2012; Meindersma & de Haan, 2008). A new alternative to extraction solvents are ionic liquids (ILs), i.e. salts composed of an organic cation and inorganic anion or inorganic anion with organic functional groups. Most interesting are ILs that are liquid already at ambient temperature (room temperature ionic liquids, RTIL (Huddleston et al., 1998)). The huge number of possible combinations of cations and anions allows designing specific ILs suitable for the process. In this way, it is possible to tune the properties (selectivity and capacity) of the extractive solvent for the system to be separated. Moreover, thermodynamics and kinetics of the extraction process in the presence of ILs differ from those of systems comprising conventional solvents. Further advantages of ILs over conventional solvents are their very low volativity at room-temperatue, accelerated thermostability, low toxicity, non-flammability, liquid form in a wide interval of temperatures, unique solvatation properties, and recyclability (Hanusek, 2005; Seiler et al., 2004). Therefore, ILs represent a green alternative to conventional organic solvents with lower social and individual risk to humans. Compared to conventional organic solvents, several ionic liquids show higher aromatic/aliphatic selectivity. Detailed overview of ILs’ selectivity in aromatics separation from hydrocarbon mixtures can be found in studies (Gmehling & Krummen, 2003; Krummen et al., 2002; Perreiro et al., 2012). Thanks to their low volatility, regeneration of ILs is much simpler than the regeneration of conventional extraction solvents. Taking into account annual costs (Meindersma & de Haan, 2008), this fact is crucial in diminishing the overall costs of the extraction process. Although the use of ILs in separation processes seems promising, there is a lack of information regarding phase equilibria and physical properties of systems comprising ionic liquids, even the available data obtained from different sources are confusing. Taking into account that the quality of thermodynamic data plays an important role in cost-effective design and operation of a separation unit (Hendricks

1549

et al., 2010), obtaining reliable equilibrium data is crucial.

Theoretical In liquid–liquid extraction, at least three components are involved. Proper description of ternary or multicomponent liquid–liquid equilibria (LLE) is necessary for reliable design of extraction equipment as equilibrium conditions are assumed taking into account both equilibrium (equilibrium stage) and nonequilibrium (phase interphase) separation equipment models. Multicomponent LLE calculations are usually based on the knowledge of binary equilibrium data of the pairs of components forming the studied multicomponent system. However, this information is usually insuficient, causing serious discrepances between experimental and calculated compositions of the equilibrium phases of such a system. As a consequence, design and simulation of the separation unit are inaccurate. One of the possibilities to overcome this problem is to consider experimental ternary LLE data. Sørensen and Arlt (1980a, 1980b) used direct fitting of experimental ternary LLE data for the determination of binary parameters of excess Gibbs energy (GE ) equations (NRTL and UNIQUAC). A disadvantage of this approach is that the validity of the obtained binary parameters of GE equations is strictly limited, applicable only in the concentration range for which the parameters were evaluated. Moreover, these parameters are unsuitable for extrapolation and they do not correctly describe equilibria of the binary subsystems forming the multicomponent system. Another approach improving the ternary LLE equilibrium description, extension of the original GE equations by the ternary correction term, was proposed by Surový et al. (1982). The authors assumed that the difference between the measured value of the excess Gibbs energy in a ternary mixture, ∆g E , and the value calculated on basis of binary equilibrium data, (∆g E )b , is caused by the ternary intermolecular interactions. Considering this fact, the authors added a simple ternary term to the common expression used to calculate the excess Gibbs energy of the system, ∆g E . In case of a ternary mixture, the following equation is used:
E ∆g ∆g E = + xi xj xk (Ei xi + Ej xj + Ek xk ) RT RT b i, j, k = A, B, C

(1)

where R, T, and x represent the gas constant, thermodynamic temperature, and the liquid phase mole fraction, respectively. Ternary parameters, Ei , Ej , Ek , quantify the ternary intermolecular interactions in a ternary system. A, B, and C denote the mixture components. The first term on the right hand side of Eq. (1) is the excess molar Gibbs energy of the system calculated

1550

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

using the original form of the GE equation; the second one is the ternary correction term. In case that the NRTL equation (Renon & Prausnitz, 1968) is used to calculate the phase equilibrium, variation of the activity coefficients with the liquid phase composition is expressed as follows:  τji Gji xj j ln γi =  + Gli xl l

 ⎛ ⎞ τkj Gkj xk  Gij xj ⎝τij − k  ⎠ + ∆t ln γi  + G x G x lj l lj l j l

l

i, j, k, l = A, B, C

(2)

fair accuracy although this GE model was developed for non-electrolyte systems. In 1982, the extension of the NRTL equation, electrolyte-NRTL (e-NRTL) for the description of phase equilibria of aqueous systems containing fully soluble inorganic electrolytes was sugested (Chen et al., 1982). Recently, a modified e-NRTL equation (Chan & Song, 2004) was applied to correlate the experimental equlibrium data of systems including ionic liquids (Simoni et al., 2008). The results showed only negligible improvement of the correlation of experimental data compared to that obtained by the original NRTL equation. Therefore, in the present study, the original NRTL equation and NRTL equation extended by the ternary correction term were used to describe LLE of the ternary system.

where the original model parameters τ ij and Gij are defined by the following expressions: gij − gjj Gij = exp (−αij τij ) RT i, j = A, B, C i
= j

τij =

(3)

whereby τ ij
= τ ji , τ ii = τ jj = 0, and αij is the nonrandomness parameter of the NRTL equation. Ternary correction term in Eq. (2), ∆t ln γi , is defined as follows:

Model development Continuous countercurrent flow of “lighter” and “heavier” liquids is commonly used in a cascade of mixer–settler contactors. Material balance of the n-th contactor (equilibrium stage) can be written as follows: nE,n+1 + nR,n−1 = nE,n + nR,n

n = 1, 2, ..., N (5)

∆t ln γi = xj xk [Ei xi (2 − 3xi ) + Ej xj (1 − 3xi ) + + Ek xk (1 − 3xi )] i, j, k = A, B, C (4)

yi,n+1 nE,n+1 + xi,n−1 nR,n−1 = yi,n nE,n + yi,n nR,n n = 1, 2, ..., N i = A, B, C (6)

If the ternary correction term parameters are set to be equal to zero (Ei = Ej = Ek = 0), Eq. (2) automatically becomes the original NRTL equation for the calculation of the excess Gibbs energy of multicomponent systems. The advantage of this approach dwels in the fact that the binary GE equation parameters can be evaluated independently from the binary equilibrium data and the ternary correction term parameters from the ternary equilibrium data using previously obtained binary parameters in the original form. Therefore, binary parameters preserve their physical meaning and suit the binary equilibrium description. Furthermore, in combination with the ternary correction term, the extended GE equation provides an improved ternary equilibrium correlation within the range of experimental data and reliable prediction outside its limits, over the whole concentration range. Moreover, the ternary correction term of the extended GE equation can be used to predict phase equilibrium of quaternary and more component systems comprising components of the previous ternary system. Some authors (Aznar, 2007; Arce et al., 2007; González et al., 2010; Hansmeier et al., 2010; Pereirotal., 2010; Varma et al., 2011) confirmed the capability of the original NRTL equation to correlate experimental LLE data of systems containing ionic liquids with

where xi and yi represent the mole fractions of the i-th component in raffinate and in extract, respectively, A, B, and C denote standard components in the extraction (A – original solvent, B – extracted component, C – extraction solvent), n is the contactor number in the cascade, N is the total number of contactors in the cascade, nR,n and nE,n correspond to molar flows of raffinate and extract from the n-th contactor, respectively, and nF and nS represent the feed and extractive solvent molar flows, respectively, entering the cascade of contactors countercurrently. Equilibrium at each contactor (stage) is expressed in a form of isoactivity condition: γiI xIi = γiII xII i

[T, P ]

i = A, B, C

(7)

where superscritpts I and II denote the equilibrium liquid phases. Eqs. (5)–(7) together with the summation equations for both equilibrium phases leaving each mem ber of the cascade of contactors ( xi,n = 1 and i  yi,n = 1) representing the equilibrium model of i

a cascade of extractors were solved numerically according to the Hunter–Nash method (Hunter & Nash, 1934) using a program developed in the MATLAB software (MathWorks, Natick, MA, USA).

1551

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

Fig. 1. Binary LLE data for system heptane (A)–EMIES (C): experimental data presented in Bendová and Wagner (2009) ( ), García et al. (2010) ( ), Meindersma et al. (2006) (
). Mole fraction of IL in the heptane-rich phase (left) and in the IL-rich phase (right).

•

Fig. 2. Binary LLE data for system toluene (B)–EMIES (C): experimental data presented in Bendová and Wagner (2009) ( ), nska et al. (2008) ( ), Meindersma et al. (2006) (
). Mole fraction of IL in the toluene-rich García et al. (2010) ( ), Doma´ phase (left) and in the IL-rich phase (right).

•

Results and discussion In the present study, influence of the quality of LLE prediction/description on the design of separation equipment for liquid extraction was investigated. As a model system, a mixture of aliphatic (heptane) and aromatic (toluene) hydrocarbons with low content of aromatic hydrocarbons was chosen assuming ionic liquid as the extraction solvent. Taking into account the available equilibrium data as well as the information on its basic properties (selectivity and capacity), 1-ethyl-3-methylimidazolium ethyl sulfate ([emim]C2 H5 SO4 , EMIES) was selected as the extractive solvent. Parameters of the excess molar Gibbs energy equation As stated above, variation of the components’ activity coefficients with the liquid phase mixture composition was expressed by the original NRTL equation (Renon & Prausnitz, 1968) or the NRTL equation extended by the ternary correction term according to

Surový et al. (1982). Neglecting the mixing heats, extraction can be assumed to be an isothermal process. Thus, the NRTL model parameters used in the simulation calculations were constant, independent on temperature. Binary parameters of the NRTL equation for partially miscible subsystems, i.e. heptane–EMIES and toluene–EMIES, were evaluated from the available binary LLE data; the value of the non-randomness parameter αij was fitted to the activity coefficient at infinite dilution of the hydrocarbon in IL, γ ∞ i , at the given temperature. While fitting the experimental data, simultaneous solution of the following equations was carried out: N  I

2 ai − aII =0 i

i = A, B, C

(8)

i

2 ∞ ∞ =0 γi,exp − γi,calc

i = A, B, C

(9)

Equilibrium data for the binary systems heptane– EMIES and toluene–EMIES are presented in Figs. 1 and 2, respectively. Sets of equilibrium data were col-

1552

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

Fig. 3. Variation of activity coefficients at infinite dilution of heptane (a) and toluene (b) in EMIES with temperature: experimental data presented in Krummen et al. (2002) ( ) and Sumartschenkowa et al. (2006) ( ).

•

lected from available literature as presented in the respective figure captions. Variations of the activity coefficients at infinite dilution of both hydrocarbons in EMIES are given in Fig. 3. Considering the large differences in the experimental data presented in Fig. 2, the data published by Meindersma et al. (2006) were excluded from the data set used for the evaluation of the NRTL binary parameters. Moreover, in case of the activity coefficient at infinite dilution of toluene in EMIES (Fig. 3b), contradictory tendencies in the temperature variation of γB∞ were found (Krummen et al., 2002; Sumartschenkowa et al., 2006). Mutual solubilities of hydrocarbons and EMIES at temperature T = 313.2 K were evaluated by fitting the experimental data. In case of the binary system heptane (A)–EMIES (C), solubility of IL in the heptanerich phase was xIC = 0.0053 and that of heptane in the IL-rich phase was xII A = 0.0058 (mole fraction of heptane in the IL-rich phase) with the activity coefficient ∞ of heptane in IL at infinite dilution being γA = 195. Similarly, solubilities and activity coefficients at infinite dilution were computed for the binary system toluene (B)–EMIES (C): xIC = 0.0050, xII B = 0.2729, and γB∞ = 5.3. Binary NRTL parameters for the fully miscible system heptane (A)–toluene (B) were taken from literature (Steltenpohl & Graczová, 2010). They were evaluated from the complete isothermal vapor– liquid equilibrium data measured at T = 313.2 K, i.e. the temperature at which the extraction was carried out. Parameters of the ternary correction term were obtained by fitting LLE data of the ternary system heptane (A)–toluene (B)–EMIES (C) at T = 313.2 K (Meindersma et al., 2006) by minimizing the sum of square differences between the experimental and the computed components’ mole fractions. A summary of the NRTL equation parameters is given in Table 1.

Table 1. Binary NRTL and ternary correction term parameters for the system heptane (A)–toluene (B)–EMIES (C) at T = 313.2 K heptane Component

toluene

EMIES

Aa ij /K

heptane toluene EMIES

0 297.7628 1205.9298

–73.4919 0 72.8109

1231.2846 1563.4833 0

αij heptane toluene EMIES

0 0.3000 0.2581

0.3000 0 0.2485

0.2581 0.2485 0

Ei –7.8430

–0.4952

7.3292



a) Aij = ∆gij R.

Comparison of the phase equilibrium description using different NRTL models Fig. 4 presents a comparison of experimental and computed tie-lines and binodal curves (concentration region corresponding to simulation conditions) of the ternary system heptane (A)–toluene (B)–EMIES (C) at T = 313.2 K. To calculate the equilibrium, three different models were used: i) Model 1: original NRTL equation (binary parameters evaluated independently from binary equilibrium data); ii) Model 2: NRTL equation extended by the ternary correction term (besides the original binary parameters, ternary correction term parameters were evaluated from the ternary tie-lines);

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

1553

Fig. 4. Ternary diagram of system heptane (A)–toluene (B)–EMIES (C) at T = 313.2 K. Experimental tie-lines ( — ) (Meindersma et al., 2006); binodal curve (—) and tie-lines (– – – –) calculated by the NRTL equation: Model 1 (a), Model 2 (b), and Model 3 (c).

iii) Model 3: NRTL equation with binary model parameters obtained by direct fitting of ternary LLE data employing ASPEN Plus (Meindersma et al., 2006; Meindersma, 2005). For low toluene concentrations (from 0 mole % to 20 mole %), the accuracy of the equilibrium description by the original NRTL model (Model 1) characterized by the root mean square deviation is RMSD = 0.0087, while the description using Model 3 offers RMSD = 0.0133. Both these models present larger area of the two-phase region compared to the experimentally estimated one. The calculated raffinate branch of the binodal curve is practically the equilibrium triangle side AB. Moreover, it can be clearly seen that the slopes of the tie-lines are incorrect, especially

for Model 3. This model presents inaccurate binary solubilities and activity coefficients at infinite dilution of both hydrocarbons in EMIES: for the constituent binary system heptane (A)–EMIES (C): xIC = 0.000 ∞ and γA > 250, and for the binary subsystem toluene ∞ (B)–EMIES (C): xII B = 0.200 and γA = 6.4. One of the possible reasons of the inaccuracy of Model 3 are the proper experimental ternary and binary LLE data from which the binary NRTL parameters were evaluated by Meindersma et al. (2006) and Meindersma (2005). The solubility of toluene in EMIES as published by Meindersma et al. (2006) does not correspond with the data found by other authors (Bendová & Wagner, 2009; García et al., 2010; Doma´ nska et al., 2008) (Fig. 2).

1554

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

Influence of the chosen equilibrium model on the separation equipment design

Fig. 5. Variation of aromatic/aliphatic hydrocarbon selectivity, β BA , with the toluene content in raffinate. Experimental data (Meindersma et al., 2006) ( ) and those calculated by the NRTL equation: Model 1 (– · – · –), Model 2 (– – – –), and Model 3 (- - - - -).

•

Fig. 6. Relation between the number of theoretical stages, N, and the solvent to feed mole ratio, nS /nF , for the maximum toluene content in the final raffinate of 0.5 mole %, calculated by the NRTL equation: Model 1 (
), Model 2 ( ), and Model 3 ( ).

A better fit of experimental data was obtained employing Model 2, i.e. the NRTL equation extended by the ternary correction term. This model provided a more accurate description of the two-phase region, slopes of the tie-lines (for the given region RMSD = 0.0050), and the variation of the aromatic/aliphatic selectivity with the heptane-rich phase composition (Fig. 5). In the region of low toluene concentrations (up to 0.20 mole fraction), selectivity computed using Model 2 varied within the values of 53 and 48, which corresponds well with the experimental data published in Meindersma et al. (2006) and Meindersma (2005). Using the original NRTL equation (Model 1), a steeper dependency was observed and the results obtained with Model 3 (NRTL model with binary parameters evaluated from the ternary LLE data) were far from the experimental data presented by these authors (Meindersma, 2005; Meindersma et al., 2006).

Simulation runs were directed to assess the influence of the thermodynamic model accuracy on the extractor design. Simulations were carried out using the MATLAB program, which corresponds with the Hunter–Nash method (Hunter & Nash, 1934). The mixture to be separated was composed of heptane (90 mole %) and toluene (10 mole %). For extraction, EMIES was chosen as the extractive solvent, and the simulations were carried out at the temperature of 313.2 K. As a purity criterium, the maximum toluene concentration of 0.5 mole % in the final raffinate was considered. Firstly, the relation between the number of theoretical extraction stages (contactors), N, and the mole ratio of solvent to feed, nS /nF , was investigated taking into account the required purity of final raffinate. For different NRTL models, these dependences differ substantially (Fig. 6). The minimum solvent specific consumption, nSmin /nF , computed with the original NRTL equation (Model 1) was close to 3.0, while the value obtained by Model 2 was 3.3; Model 3 provided a value close to 2.2. Fig. 6 clearly shows that the NRTL models assumed resulted in different numbers of theoretical stages required to achieve the desired purity of final raffinate. For the solvent to feed mole ratio nS /nF = 3.5, nine equilibrium stages were predicted by Model 1, 14 by Model 2, and only five by Model 3. Figs. 7 and 8 present the concentration profiles of toluene in the raffinate and extract phases calculated for the predicted numbers of theoretical stages and the chosen solvent to feed mole ratio nS /nF = 3.5. Models 1 and 2 were used to calculate these profiles; computed yields/recoveries of the mixture components in the final streams are presented in Table 2. Employing the original NRTL equation (Model 1), the required purity of the final raffinate, xB,N ≤ 0.005, was reached at the solvent to feed ratio nS /nF = 4.5 using five theoretical stages; for nS /nF = 3.5, ten theoretical stages were sufficient. In case of Model 2, the separation conditions were met for nS /nF = 3.5 using 15 theoretical stages. The results obtained using the NRTL equation extended by the ternary correction term (Model 2, i.e. the model better describing the phase equilibrium) show that a higher number of theoretical stages is necessary to reach the required purity of the raffinate compared to the results obtained by the original NRTL equation. Moreover, the yield of toluene in extracts and the heptane loss in extracts obtained by Model 1 are higher than those calculated by Model 2, independently of the number of theoretical stages assumed. It was shown that for a preset separation efficiency criterium, different accuracy of the equilibrium description caused serious discrepancies in the separa-

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

1555

Fig. 7. Concentration profile of toluene in the raffinate, mole fraction xB , for extractors with different numbers of theoretical stages, N: 15 ( ), 10 (
), and 5 ( ). Simulations carried out for the solvent to feed mole ratio nS /nF = 3.5 using the NRTL equation: Model 1 (a) and Model 2 (b).

Fig. 8. Concentration profile of toluene in the extract, mole fraction yB , for extractors with different numbers of theoretical stages, N: 15 ( ), 10 (
), and 5 ( ). Simulations carried out for the solvent to feed mole ratio nS /nF = 3.5 using the NRTL equation: Model 1 (a) and Model 2 (b).

tion equipment design, e.g. in the number of theoretical stages, solvent to feed ratio, and product purity. Extractor simulation for N = 12, nS /nF = 3.5 Meindersma in his study (Meindersma, 2005) investigated the properties of three ILs, namely 4-methyl-N-butylpyridinium tetrafluoroborate ([mebupy]BF4 ), 1-methyl-3-methylimidazolium methyl sulfate ([mmim]CH3 SO4 ), and 1-ethyl-3-methylimidazolium ethyl sulfate ([emim]C2 H5 SO4 , EMIES), for the alkane/aromatics separation. The results were compared with those obtained with a commonly used extraction solvent, sulfolane. It was stated that among these ILs, [mebupy]BF4 was the best alternative to the common solvent. Compared to sulfolane, all ILs studied showed higher selectivity and [mebupy]BF4 offered the highest extraction capacity among the considered

ILs. EMIES, the solvent selected for this study, shows selectivity comparable to that of [mebupy]BF4 in the heptane–toluene mixture separation; its capacity is, however, a little lower than that of sulfolane. Meindersma and de Haan (2008) carried out a set of simulations for the separation of a heptane–toluene mixture (containing 10 mass % of toluene) in the presence of [mebupy]BF4 . Simulation results showed that for the required yield of toluene (98 % in the final extract) and the recovery of heptane (98 % in the final raffinate), an extractor with twelve theoretical stages is necessary considering the IL to feed mass ratio mS /mF = 5.5 (nS /nF = 2.31). Calculated loss of IL in the raffinate was 0.00 % (i.e., total immiscibility of [mebupy]BF4 in heptane was assumed). Simulations were carried out using the ASPEN Plus simulation software. Here, the results of simulations carried out for

1556

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

Table 2. Comparison of component yields/recoveries in raffinate and extract; data obtained using the NRTL equation: Model 1 and Model 2 Number of stages, N Solvent to feed mole ratio, nS /nF

5

5

10

10

15

3.5

4.5

3.5

4.5

3.5

0.005 96.78 95.56 3.22 0.10

0.004 97.42 96.44 2.58 0.13

0.0005 96.76 99.55 3.24 0.10

0.002 97.41 98.67 2.59 0.13

0.007a 97.27 93.68 2.73 0.11

0.008a 97.91 93.06 2.09 0.14

0.001 97.28 99.01 2.72 0.10

0.005 97.91 95.88 2.09 0.14

Model 1 Toluene content in raffinate, xB,N Recovery of heptane in raffinate/% Yield of toluene in extract/% Heptane loss in extract/% EMIES loss in raffinate/%

0.012a 97.44 89.16 2.56 0.13 Model 2

Toluene content in raffinate, xB,N Recovery of heptane in raffinate/% Yield of toluene in extract/% Heptane loss in extract/% EMIES loss in raffinate/%

0.016a 97.90 85.48 2.10 0.15

a) Requirement on the raffinate purity not met.

Fig. 9. Comparison of toluene concentration profiles in raffinates (a) and extracts (b) for N = 12 and nS /nF = 3.5. Results computed using the NRTL equation: Model 1 (
), Model 2 ( ), and Model 3 ( ).

the EMIES extractive solvent and an extractor with twelve theoretical stages are given. The required separation efficiency was defined as the maximum content of toluene in the final raffinate, xB,N ≤ 0.005. A feed composed of 90 mole % of heptane and 10 mole % of toluene was assumed and the solvent to feed mole ratio of 3.5 (i.e., mS /mF = 8.3) was taken from our previous computations. Higher consumption of EMIES compared to [mebupy]BF4 (mS /mF = 5.5) corresponds to the respective values of the toluene distribution coefficient in the region of low toluene concentrations. In Fig. 9, a comparison of the concentration profiles obtained for the raffinate and extraxt phases at individual theoretical stages is presented. Simulation based on Model 3 (NRTL equation with binary parameters evaluated from the ternary LLE data) predicts that the required separation efficiency is reached alredy at the 6th stage. This can be explained by

the overestimated selectivity of the extraction solvent, β BA , as shown in Fig. 10. These values, however, are in disagreement with the experimental data (Meindersma et al., 2006). Figs. 11 and 12 present the variation of the heptane recovery in the final raffinate, YAR , and the yield of toluene in the final extract, YBE , with the solvent to feed mole ratio for a separation unit with 12 theoretical stages. According to the simulation results of Model 3, heptane recovery higher than 98 % was obtained already for the solvent to feed ratio nS /nF = 2.31. On the other hand, toluene yield in the final extract is on the level of 93.3 %. By increasing the solvent to feed mole ratio to 3, both the toluene yield in extract and the heptane recovery in raffinate, higher than 98 % (YBE = 99.6 %, YAR = 98.7 %) were obtained. Further increase of the solvent consumption, nS /nF = 3.5, caused an increase of the

1557

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

Fig. 10. Variation of the toluene/heptane selectivity along the extractor with N = 12 and nS /nF = 3.5. Results computed using the NRTL equation: Model 1 (
), Model 2 ( ), and Model 3 ( ).

Fig. 12. Variation of the toluene yield in the final extract with the solvent to feed mole ratio. Separation efficiency set as the maximum toluene content in the final raffinate, xB,N ≤ 0.005. Results of simulation employing the NRTL equation: Model 1 (
), Model 2 ( ), and Model 3 ( ).

Table 3. Comparison of the component yield/loss in the final streams calculated for an extractor with 12 theoretical stages and the solvent to feed mole ratio of 3.5 Parameter

Fig. 11. Variation of the heptane recovery in the final raffinate with the solvent to feed mole ratio. Separation efficiency set as the maximum toluene content in the final raffinate, xB,N ≤ 0.005. Results of simulation employing the NRTL equation: Model 1 (
), Model 2 ( ), and Model 3 ( ).

toluene yield in the final extract up to YBE = 99.9 % and a slight decrease of the heptane recovery in the final raffinate to 98.5 % (Table 3). Results of simulations employing Model 1 (original NRTL equation) show lower yield/recovery of hydrocarbons compared to the simulation results obtained using Model 3 (Table 3). Assuming the solvent to feed ratio nS /nF = 3.5, heptane recovery in the final raffinate was YAR = 97.4 %, and toluene yield in the final extract reached the value of 97.6 %. Using this model, the separation efficiency requirement, i.e. xB,N ≤ 0.005, was met. Under the same conditions, N = 12 and nS /nF = 3.5, the simulation based on Model 2 (NRTL equation extended by the ternary correction term, i.e. the model with the best phase equilibrium description) predicted the lowest recovery of toluene in the extract (YBE = 94.5 %), while the recovery of heptane was YAR = 97.9 %. Moreover, using Model 2, the re-

Toluene content in raffinate, xB,N EMIES content in raffinate, xC,N Heptane content in extract, yA,1 Toluene content in extract, yB,1 Heptane recovery in raffinate/% Toluene yield in extract/% Heptane loss in extract/% Toluene loss in raffinate/% EMIES loss in raffinate/%

Model 1 Model 2 Model 3 0.003 0.005 0.006 0.027 97.41 97.61 2.59 2.39 0.13

0.006a 7 × 10−4 0.006 0.000 0.005 0.004 0.026 0.028 97.91 98.48 94.45 99,94 2.09 1.52 5.55 0.06 0.14 0.00

a) Requirement on the raffinate purity not met.

quired purity of the final raffinate was not achieved (Table 3). According to the results of simulations using either the original NRTL equation (Model 1) or its form extended by the ternary correction term (Model 2) for the assumed separation equipment with N = 12 and nS /nF = 3.5 (Table 3), the calculated loss of the extractive solvent, EMIES, in the final raffinate was 0.13 % (Model 1) and 0.14 % (Model 2). These values agree with the IL loss observed experimentally (Meindersma & de Haan, 2008). The EMIES loss simulated using Model 3 was negligible (the result corresponds to the complete immiscibility of EMIES in heptane predicted by this model). In literature (Meindersma & de Haan, 2008), the loss of [mebupy]BF4 was measured experimentally in a disc extraction column with N = 3 and mS /mF = 2. The authors estimated the content of this IL in the raffinate to be 0.26 mass %, which corresponds to the ILs loss of about 0.13 %. The simulation using the ASPEN Plus simulation program

1558

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

resulted in a much lower value of 0.01 mass % (ILs loss on the level of 0.005 %) (Meindersma & de Haan, 2008). Again, the influence of the LLE description quality on the results of the separation equipment simulation was shown. In this case, design calculations of an extractor by the NRTL equation (Model 3) provided distorted results on the number of theoretical stages compared with the simulations using Model 1 and Model 2. Moreover, incorrect data on the extraction solvent consumption, toluene yield in extract and heptane recovery in raffinate were obtained.

Conclusions This study confirms the fact that the quality of the description of thermodynamic properties of a mixture plays an important role in the design and operation of separation units. In our case, three models for the description of LLE were used. Model 1 was based on a set of experimental binary data from which the binary parameters of the original NRTL equation were evaluated. Model 2 was the NRTL equation extended by the ternary correction term, where the ternary parameters were obtained by fitting experimental ternary LLE data while the binary NTRL parameters (evaluated from the binary equilibrium data) remained unchanged. Binary NRTL parameters of Model 3 were obtained by directly fitting the ternary LLE data according to Sørensen and Arlt (1980a, 1980b). Simulation results showed serious discrepancies between the basic design parameters of extractors obtained by the three models used for the ternary LLE calculation; especially, when considering the relatively narrow concentration range of the extracted component (toluene). Already a small difference in the equilibrium description can cause considerable differences in the number of theoretical stages, extraction solvent consumption, and the equipment separation efficiency. Model 3 predicted a larger immiscibility region compared to the experimental data (see Fig. 4, raffinate line) thus overestimating the equipment’s separation efficiency. As a consequence, much lower number of theoretical stages, lower extractive solvent consumption, and higher products’ purity resulted from the extractor simulation compared to those obtained using Models 1 and 2. It is worth noting that an IL recovery problem was identified when analyzing the results of extractor simulation based on the LLE description by the models used (Table 3). More realistic results were obtained when the LLE of the ternary system was calculated on basis of Models 1 and 2. However, relatively large differences were observed comparing the results obtained for these models due to the inaccuracy of the ternary phase equilibria prediction based solely on the binary model parameters. Quality of the ternary equilibrium description was improved employing the model that ac-

counts for both binary and ternary equilibrium data. Design calculations based on the most realistic Model 2 predicted the largest separation equipment and minor separation efficiency. Taking into account the above-mentioned results, it is clear that while evaluating the binary model parameters from the ternary LLE data, model parameters of the partially miscible binaries should be linked to the respective mutual binary solubilities. Binary model parameters obtained by this procedure improve the ternary equilibrium description in the whole concentration range, particularly the data lying close to the sides of the ternary LLE diagrams. These values may play a dominant role in the separation unit design calculations. As it was shown in the study, small diferences in equilibria description can have significant impact on the results of countercurrent extractor simulation. Acknowledgements. The authors acknowledge the Research and Development Agency APVV (Grant APVV-0353-06) for financial support.

References Ali, S. H., Lababidi, H. M. S., Merchant, S. Q., & Fahim, M. A. (2003). Extraction of aromatics from naphtha reformate using propylene carbonate. Fluid Phase Equilibria, 214, 25– 38. DOI: 10.1016/s0378-3812(03)00323-6. Arce, A., Earle, M. J., Katdare, S. P., Rodríguez, H., & Seddon, K. R. (2007). Phase equilibria of mixtures of mutually immiscible ionic liquids. Fluid Phase Equilibria, 261, 427–433. DOI: 10.1016/j.fluid.2007.06.017. Aznar, M. (2007). Correlation of (liquid + liquid) equilibrium of systems including ionic liquids. Brazilian Journal of Chemical Engineering, 24, 143–149. DOI: 10.1590/s010466322007000100013. Bendová, M., & Wagner, Z. (2009). Thermodynamic description of liquid–liquid equilibria in systems 1-ethyl3-methylimidazolium ethylsulfate + C7-hydrocarbons by polymer-solution models. Fluid Phase Equilibria, 284, 80–85. DOI: 10.1016/j.fluid.2009.06.014. Blažek, J., & Rábl, V. (2006). Základy zpracování a využití ropy (2nd ed.). Prague, Czech Republic: VŠCHT. Chan, C., & Song, Y. H. (2004). Generalized electrolyte-NRTL model for mixed-solvent electrolyte systems. AIChE Journal, 50, 1928–1941. DOI: 10.1002/aic.10151. Chen, C. C., Britt, H. I., Boston, J. F., & Evans, L. B. (1982). Local composition model for excess Gibbs energy of electrolyte systems. Part I: Single solvent, single completely dissociated electrolyte system. AIChE Journal, 28, 588–596. DOI: 10.1002/aic.690280410. Choi, Y. J., Cho, K. W., Cho, B. W., & Yeo, Y. K. (2002). Optimization of the sulfolane extraction plant based on modeling and simulation. Industrial & Engineering Chemistry Research, 41, 5504–5509. DOI: 10.1021/ie010435a. Doma´ nska, U., Laskowska, M., & Marciniak, A. (2008). Phase equilibria of (1-ethyl-3-methylimidazolium ethylsulfate + hydrocarbon, + ketone, and + ether) binary systems. Journal of Chemical & Engineering Data, 53, 498–502. DOI: 10.1021/je700591h. García, J., Torrecilla, J. S., Fernández, A., Oliet, M., & Rodríguez, F. (2010). (Liquid + liquid) equilibria in the binary systems (aliphatic, or aromatic hydrocarbons + 1-ethyl3-methylimidazolium ethylsulfate, or 1-butyl-3-methylimid-

E. Graczová et al./Chemical Papers 67 (12) 1548–1559 (2013)

azolium methylsulfate ionic liquids). The Journal of Chemical Thermodynamics, 42, 144–150. DOI: 10.1016/j.jct.2009. 07.023. Gmehling, J., & Krummen, M. (2003). German Patent DE101 54052 DE. Munich, Germany: German Patent and Trade Mark Office. González, E. J., Calvar, N., Gómez, E., & Domínguez, Á. (2010). Separation of benzene from linear alkanes (C6 – C9 ) using 1-ethyl-3-methylimidazolium ethylsulfate at T = 298.15 K. Journal of Chemical & Engineering Data, 55, 3422–3427. DOI: 10.1021/je1001544. Hansmeier, A. R., Jongsmans, M., Meindersma, G. W., & de Haan, A. B. (2010). LLE data for the ionic liquid 3-methylN-butyl pyridinium dicyanamide with several aromatic and aliphatic hydrocarbons. The Journal of Chemical Thermodynamics, 42, 484–490. DOI: 10.1016/j.jct.2009.11.001. Hanusek, J. (2005). Iontové kapaliny – nový směr v “zelené” chemii. Chemické Listy, 99, 263–294. Hendricks, E., Kontogeorgis, G. M., Dohrn, R., de Hemptinne, J. C., Economou, I. G., Fele Žilnik, L., & Vesovic, V. (2010). Industrial requirements for thermodynamics and transport properties. Industrial & Engineering Chemistry Research, 49, 11131–11141. DOI: 10.1021/ie101231b. Huddleston, J. G., & Rogers, R. D. (1998). Room temperature ionic liquids as novel media for ‘clean’ liquid–liquid extraction. Chemical Communications, 1998, 1765–1766. DOI: 10.1039/a803999b. Hunter, T. G., & Nash, A. W. (1934). The application of physico-chemical principles to the design of liquid-liquid contact equipment. Part II: Application of phase-rule graphical method. Journal of the Society of Chemical Industry, 53, 95T–102T. DOI: 10.1002/jctb.5000531407. Krishna, R., Goswami, A. N., Nanoti, S. M., Rawat, B. S., Khanna, M. K., & Dobhal, J. (1987). Extraction of aromatics from 63-69 ◦C naphtha fraction for food grade hexane production using sulfolane and NMP as solvents. Indian Journal of Chemical Technology, 25, 602–606. Krummen, M., Wasserchied, P., & Gmehling, J. (2002). Measurement of activity coeffitients at infinite dilution in ionic liquids using the dilutor technique. Journal of Chemical & Engineering Data, 47, 1411–1417. DOI: 10.1021/je0200517. Meindersma, G. W. (2005). From solvent development to pilot RDC evaluation: Extraction of aromatics from naphtha with ionic liquids. PhD. thesis, University of Twente, Enschede, The Netherlands. Meindersma, G. W., Podt, A. J. G., & de Haan, A. B. (2006). Ternary liquid–liquid equilibria for mixtures of toluene + nheptane + an ionic liquid. Fluid Phase Equilibria, 247, 158– 168. DOI: 10.1016/j.fluid.2006.07.002. Meindersma, G. W., & de Haan, A. B. (2008). Conceptual process design for aromatic/aliphatic separation with ionic liquids. Chemical Engineering Research and Design, 86, 745– 752. DOI: 10.1016/j.cherd.2008.02.016. Pereiro, A. B., Deive, F. J., Esperan¸ca, J. M. S. S., & Rodríguez, A. (2010). Alkylsulfate-based ionic liquids to separate azeotropic mixtures. Fluid Phase Equilibria, 294, 49–53. DOI: 10.1016/j.fluid.2010.05.006.

1559

Perreiro, A. B., Araújo, J. M. M., Esperan¸ca, J. M. M. S., Marrucho, I. M., & Rebelo, L. P. N. (2012). Ionic liquids in separations of azeotropic systems – A review. The Journal of Chemical Thermodynamics, 46, 2–28. DOI: 10.1016/j.jct.2011.05.026. Renon, H., & Prausnitz, J. M. (1968). Local compositions in thermodynamic excess functions for liquid mixtures. AIChE Journal, 14, 135–144. DOI: 10.1002/aic.690140124. Schnieder, D. F. (2004). Avoid sulfolane regeneration problems. Chemical Engineering Progress, 100(7), 34–39. Seiler, M., Jork, C., Kavarnou, A., Arlt, W., & Hirsch, R. (2004). Separation of azeotropic mixtures using hyperbranched polymers or iononic liquids. AIChE Journal, 50, 2439–2454. DOI: 10.1002/aic.10249. Simoni, L. D., Lin, Y. D., Brennecke, J. F., & Stadtherr, M. A. (2008). Modeling liquid–liquid equilibrium of ionic liquid systems with NRTL, electrolyte-NRTL, and UNIQUAC. Industrial & Engineering Chemistry Research, 47, 256–272. DOI: 10.1021/ie070956j. Sørensen, J. M., & Arlt, W. (1980a). Liquid-liquid equilibrium data collection. (DECHEMA Chemistry data series, Vol. V, Part 2. Ternary systems). Frankfurt/Main, Germany: DECHEMA. Sørensen, J. M., & Arlt, W. (1980b). Liquid-liquid equilibrium data collection. (DECHEMA Chemistry data series, Vol. V, Part 3. Ternary and quaternary systems). Frankfurt/Main, Germany: DECHEMA. Steltenpohl, P., & Graczová, E. (2010). Application of extended NRTL equation for ternary liquid–liquid and vapor–liquid– liquid equilibria description. Chemical Papers, 64, 310–317. DOI: 10.2478/s11696-010-0006-x. Sumartschenkowa, I. A., Verevkin, S. P., Vasiltsova, T. V., Bich, E., Heintz, A., Shevelyova, M. P., & Kabo, G. J. (2006). Experimental study of thermodynamic properties of mixtures containing ionic liquid 1-ethyl-3-methylimidazolium ethyl sulfate using gas–liquid chromatography and transpiration method. Journal of Chemical & Engineering Data, 51, 2138–2144. DOI: 10.1021/je0602723. Surový, J., Dojčanský, J., & Bafrncová, S. (1982). The calculation of ternary liquid-liquid (L-L) equilibrium data using a ternary correction to the excess Gibbs free energy. Collection of Czechoslovak Chemical Communications, 47, 2380–2392. DOI: 10.1135/cccc19822380. Varma, N. R., Ramalingam, A., & Banerjee, T. (2011). Experiments, correlations and COSMO-RS predictions for the extraction of benzothiophene from n-hexane using imidazolium-based ionic liquids. Chemical Engineering Journal, 166, 30–39. DOI: 10.1016/j.cej.2010.09.015. Wauquier, J. P. (2000). Petroleum refining 2: Separation pro` cesses. Paris, France: Editions Technip. Weissermel, K., & Arpe, H. J. (2003). Industrial organic chemistry (4th ed., pp. 313–336). Weinheim, Germany: Wiley– VCH. Yorulmaz, Y., & Karpuczu, F. (1985). Sulfolane versus diethylene glycol in recovery of aromatics. Chemical Engineering Research and Design, 63, 184–190.