INTRODUCTION
Distillation is one of the most important unit operations in chemical engineering, owing to its frequent use in separation processes and the cost associated with them in the chemical industry (^{Gmehling, 2009}). Sometimes, the separation of liquid mixtures through distillation is hindered by the existence of an azeotropic point. As is well known, conventional distillations are not able to purify binary liquid mixtures beyond the concentration of this azeotropic point. Nonetheless, distillation of mixtures that present an azeotropic point can be improved by using a separating agent, sometimes called an entrainer (^{Kiss & Suszwalak, 2012}; ^{Matsuda et al., 2011}; ^{Bastos et al., 2015}; ^{Zhang et al., 2015}). The entrainer is added in order to change the volatility of the components, and depending on its effects on the mixture, it leads mainly to two different distillations: azeotropic or extractive distillation. In the first case, the entrainer induces minimumboiling azeotropes and liquidliquid immiscibilities to modify the behavior of the mixture and allow the distillation. In the second case, the entrainer is a relatively nonvolatile component that allows the distillation by shifting or even eliminating the binary azeotrope, as a result of its differential affinity with the components of the original binary liquid mixture. In comparison with azeotropic distillation, extractive distillation is simpler, can use more entrainers, and is less expensive than azeotropic distillation (^{Tassios, 1972}). The use of highboiling point solvents or nonvolatile components in extractive distillations leads to lower energy consumptions in comparison to azeotropic distillation, because the entrainer does not need to be vaporized (Kiss & Suszwalak, 2012). In addition, the third component added in extractive distillations is often safer than the organic solvents commonly used in azeotropic distillation; this is the case, for example, of saline extractive distillations, where the entrainer is a salt, which presents very low toxicity and flammability. Although not as safe as inorganic salts as a consequence of a higher toxicity, the use of ionic liquids has recently become an interesting alternative as entrainer in SED because of their recognized advantages: in addition to their low vapor pressure, which allows an efficient recovery, ionic liquids are less corrosive and have higher solubility than inorganic salts (Bastos et al., 2015; ^{Calvar et al., 2007}; ^{Pereiro et al., 2012}; ^{Vicent Orchillés et al., 2008}; Vicent Orchillés et al., 2011, Zhang et al., 2015).
The addition of small quantities of dissolved salts or ionic liquids (^{Pereiro et al., 2012}) can shift or even eliminate the presence of the azeotropic point, thus making it possible to obtain high purity products from the top of the distillation column. The effect of the separating agent on the mixed solvent equilibrium is known as the salt effect. This salt effect of dissolved salts on liquid mixtures had already been studied and reported at the end of the nineteenth century (^{Kablukov, 1891}; ^{Miller, 1897}) , then followed by systematic investigations on the subject from the 1950’s (^{Cardoso and O’Connell, 1987}; ^{Jaques and Furter, 1974}; ^{Loehe and Donohue, 1997}; ^{Robinson and Stokes, 1959}). The appearance of new materials that better withstand the highly corrosive medium associated with salt systems, the fact that the addition of salts to azeotropic mixtures avoids the requirement of a third solvent, which is usually harmful to the environment, and the increasing necessity of processes with low energy consumption, leads to a renewed interest in saline extractive distillation (SED). In this direction, several papers (^{LlanoRestrepo and AguilarArias, 2003}; ^{Muzzio and Timmermann, 2014}; ^{Pinto et al., 2000}) have been recently published in order to better understand and optimize the SED process through the use of computer simulators. Computer simulation arose as one of the principal tools to improve the knowledge of the SED process, but this tool requires reliable models that allow the calculation of the phase equilibrium behavior of the system to be separated.
Several models were published as part of the aforementioned investigations, e.g., ENRTL (^{Chen and Song, 2004}; ^{Mock et al., 1986}), ^{LiQUACE (Li et al., 1994}; ^{Mohs and Gmehling, 2013}) and the Extended UNIQUAC (^{Iliuta et al., 2000}; ^{Sander et al., 1986}). These models are capable of very accurate VLE data regression by considering the Gibbs energy as the combination of three different range contributions: longrange, middlerange and shortrange. Since some contributions are mainly theoretical (for example DebyeHückel) and other ones are semiempirical, with several parameters that have to be regressed from VLE data, these are limited to the range, quantity and precision of the correlated data; any extrapolation may lead to significant deviations and should be avoided. In addition, the regression of parameters is not always clear (^{Muzzio and Timmermann, 2014}) and, often, the parameter values required for a proper data regression lack physical meaning.
On the other hand, a very useful and simple relationship was proposed by ^{Johnson and Furter (1960}):
This equation relates the ratio of relative volatilities with ( α s ) and without ( 𝛼 0 ) salt present to the salt concentration in the liquid phase ( 𝑥 3 ). Owing to its simplicity and capacity to correlate experimental data, this model has been studied and extended by different authors (^{Hashitani and Hirata, 1969}; ^{Wu et al., 1999}; Wu et al., 2001). In spite of not having a deep thermodynamic basis, the fact that only one constant is needed for a binary liquid mixture + solute system gives this equation an advantage over other semiempirical methods like ENRTL, which requires the regression of nine parameters for a binary liquid mixture + salt system. Nevertheless, as stated by ^{Meranda and Furter (1971}), the value of k is not expected to remain constant when the solvent compositions are varied owing to the interactions between the different components in the liquid phase (including volatile and nonvolatile). Understanding this limitation in the Furter equation, Hashitani and Hirata (1969) reported an empirical model which takes into account these interactions within the salt system, related not only to the electrolyte composition but also to the proportion of components in the binary liquid mixture:
where Z_{+}is the mole fraction of the more volatile component in the liquid phase (salt free basis).
The current interest in extending the Furter equation throughout the entire composition range of the two volatile components leads to more recent developments (^{Wu et al., 1999}; Wu et al., 2001). For example, the interesting model proposed by Wu et al. (1999) includes a quadratic term and a second constant, i.e.:
The additional term improves the correlation of the model, as expected. Surprisingly, the proposed model does not include a term related to the proportion of the components in the binary liquid mixture.
In the present work, the Furter equation was analyzed from the consideration of activity models in the ratio of solvent volatilities. The analysis not only increased the knowledge on the Furter equation, but also set a basis for systematic extensions of this model. As a result of this approach, two new extensions of the Furter equation were proposed. One of these extensions was incorporated into a SED column simulator (GKTM, ^{Muzzio and Timmermann, 2014}) in order to improve the stability and rate of convergence of the previous algorithm, by obtaining better initial estimates of the vapor compositions at each stage. The simulation results proved that the inclusion of Furter extended models into the simulator software led to a more efficient algorithm.
FURTER EQUATION ANALYSIS
The expressions of fugacity and relative volatility of the volatile components in a binary liquid mixture + salt systems can be combined as following:
The ratio of relative volatilities (binary mixture and binary mixturesalt systems) is defined as:
where z_{ i } is the mole fraction of component i on a saltfree basis.
At moderate pressures, Eq. (5) can be simplified with negligible error considering an ideal gas in the vapour phase and replacing the fugacity of pure liquid by the vapour pressure of the pure liquid at the considered temperature:
Replacing in (6):
If the temperature variation in the system after adding the salt is low enough, or if the relationships between pressures for both volatile components after adding the salt are similar, the effect related with vapour pressures of the pure liquids can be neglected in comparison to the effect related with activity coefficients. As a result, the Furter equation may be expressed as:
It is convenient to introduce some additional useful definitions:
where 𝑧 1 and 𝑧 3 are the mole fractions of component 1 and salt respectively, on a salt free basis. The usual mole fractions of component 1 and salt are designed x_{1} and x_{2,} respectively.
Regular solution model
The simplest expression for the activity coefficient is the regular solution model. For the components in binary liquid mixtures, the expressions are the following (^{Prigogine and Defay, 1954}):
For ternary mixtures (two volatile components and a salt), if the electrolyte is regarded as a solute, the regular solution model can be expressed as:
where 𝐴≡ 𝐴 12 = 𝐴 21 , Δ𝐴≡ 𝐴 13 − 𝐴 23 , the subscripts 1 and 2 are reserved for the volatile components, and the subscript 3 for the salt.
Then, Eq. (15a) and (15b) can be related by:
Similarly, Eq. (17a) and (17b) can be related by:
Finally, by considering equations (619), and expressing the logarithm of the ratio of relative volatilities with and without salt present, the following expressions can be derived:
From Eq. (13), β may be approximated by 1 when z_{3} is low enough and, as a consequence, Eq. (21) can be reduced to the usual Furter expression, where k = ΔA. Nonetheless, if this approximation is not considered, and taking into account that (β  1) =  z_{ 3 } β, the salt parameter can be expressed from Eq. (21) as
Equation (22) shows the functionality of the 𝑘 parameter with the composition of the volatile components. A more useful expression of Eq. (21) and Eq. (22) is the FurterRegular Solution (FRS) equation
where 𝛽 has been considered constant, since it is mostly between 0.9 and 1.
In comparison with the model proposed in ^{Wu’s work (Wu et al., 1999}), Eq. (23) includes a term of interaction between solvent and solute that could explain its better capacity of regression.
Solvation model
More accurate models can be introduced in Eq. (9). If the excess Gibbs energy is calculated as the sum of longrange (LR) and shortrange (SR) contributions,
the expression for calculation of the activity coefficient of solvent i is:
In this approach, the longrange contribution is related to the interactions between ions, while the shortrange (SR) contribution is related to interactions between all the species. Therefore, in a mixture without ions, only the shortrange contribution should be considered. Replacing Eq. (25) in Eq. (9), the following equation is obtained:
In this equation, γ_{ i } ,_{ 0 } was replaced with γ_{ i } ,_{ 0 } ^{ SR } because there are no ions in the system.
The DebyeHückel (DH) model is considered to take into account the longrange (LR) contribution in mixed solvent systems. If the partial molar volume of the solvent in the solution is approximated as the molar volume of the pure solvent, the LR logarithm can be expressed as (^{Li et al., 1994})
where M_{ i } is the molecular weight of the solvent, d_{ i } is the molar density of the pure solvent i, d is the mixedsolvent molar density; I, A and b are DH parameters, where I = 0.5∑_{ i } m_{ i } z_{ i } ^{ 2 } , m_{ i } , is the molality of ion i, v_{ i } is the salt stoichiometric number, A = 1.3278.10^{5} d^{0.5} / (DT)^{1.5}, and b = 6.360.d^{0.5}/(DT)^{0.5}.
Nevertheless, if the DH expression is considered to have negligible effect on the phase equilibrium behaviour, as stated by Chen and Evans (1986), especially in systems with high salt concentration (m>0.1), Eq. (26) can be simplified to:
As a result, the objective is to find the effect of the salt on the activity coefficient of each solvent for the SR contribution. An interesting approach to consider in Eq. (28) is related to the solvation of ions, in which the effect of the salt is taken into account from the decrease of free solvent molecules by the solvated molecules (^{Robinson and Stokes, 1959}). In this approach, the number of moles (n_{ i,F } ) of free solvent i is given by:
and the mole fraction (x_{ i,F } ) of free solvent i can be expressed as:
wherein h_{ i } is the solvation number of the salt with pure component of the mixed solvent system i and n_{ 3 } is the number of moles of salt. For binary liquid mixtures + salt systems, Eq. (30) yields:
For the resolution of this solvation model, the following general thermodynamic rule is applied (^{Nothnagel et al., 1973}; ^{Prigogine and Defay, 1954}). At equilibrium, when a fluid exists in several associated forms, the chemical potential (μ_{
i,F
} ) of the monomer (free) molecule is equal to the stoichiometric (or apparent) chemical potential (μ_{
i,
} ),
In the case under study, the chemical potential of the free solvent molecule is equal to the chemical potential of solvent on a stoichiometric basis, an expression which yields:
If the free solvent nonideality is considered to be only due to electrostatic forces (DH) with ions and the interaction with the other solvent, then the activity coefficient can be rearranged from Eq. (33) as
For a binary system, Eq. (34) can be written as:
In the present work, h_{ i } is calculated as a function of z_{ i } , i.e. h_{ i } = h_{ i } ^{∞} z_{ i } ^{2}, where h_{ i } ^{∞} is the solvation number of salt with component i at infinite dilution of the salt. Replacing Eq. (34) in Eq. (28), the FurterSolvation (FS) model is obtained:
EVALUATION OF EXTENDED MODELS
The results of applying Eq. (1), Eq. (3), Eq. (23) and Eq. (38) are reported in Table 1. Correlation of experimental data (^{Calvar et al., 2007}; ^{Peña et al., 1994}; Peña et al., 1995; Peña et al., 1996; ^{Vercher et al., 1994}; Vercher et al., 1995; Vercher et al., 1996a; Vercher et al., 1996b; Vercher et al., 1999; Vercher et al., 2001; Vercher et al., 2002; Vercher et al., 2003; Vercher et al., 2004a; Vercher et al., 2004b; Vercher et al., 2005; Vercher et al., 2006; ^{Vicent Orchillés et al., 2008}; Vicent Orchillés et al., 2011; ^{Zemp and Francesconi, 1992}) through the four models was carried out by seeking the parameters that minimized the objective function, defined by:
where N is the number of experimental points, y_{ 1,s } ^{exp} is the experimental vapour phase mole fraction, and y_{ 1,s } ^{ cal } is the vapour phase mole fraction calculated from y_{ 1,s } ^{ cal } = z_{1}/[z_{1} + (1  z_{1})/α_{ s } ]. In order to obtain the best parameters from experimental data, the Solver tool of the Microsoft Excel spreadsheet was used to minimize the objective function.
System  Condition  x_{3} ^{a}  Δy_{ 1,s }  

Furter^{b}  Wu^{c}  FRS^{d}  FS^{e}  
PropanolwaterLiNO_{3}  100.0kPa  0.020.121  0.0151  0.0140  0.0081  0.0080 
Acetic acidwaterCH_{3}COONa  100.0kPa  0.00380.1185  0.0116  0.0111  0.0094  0.0090 
PropanolwaterLiCl  100.0kPa  0.0210.126  0.0150  0.0150  0.0091  0.0094 
PropanolwaterCuCl_{2}  100.0kPa  0.0210.118  0.0120  0.0098  0.0113  0.0122 
PropanolwaterCa(NO_{3})_{2}  100.0kPa  0.0090.144  0.0188  0.0154  0.0164  0.0118 
EthanolwaterCH3COOK  101.33kPa  0.0250.15  0.0260  0.0258  0.0217  0.0223 
AcetonemethanolLiNO_{3}  100.0kPa  0.0210.16  0.1319  0.1250  0.0477  0.0273 
Ethanolwater[C6mim] [Cl]  101.3kPa  0.0020.325  0.0534  0.0406  0.0533  0.0342 
EthanolwaterKNO_{3}  100.0kPa  0.0010.032  0.0123  0.0123  0.0104  0.0110 
EthanolwaterNaNO_{3}  100.0kPa  0.0030.153  0.0202  0.0196  0.0188  0.0212 
EthanolwaterCuCl_{2}  100.0kPa  0.0070.097  0.0167  0.0167  0.0166  0.0279 
EthanolwaterCoCl_{2}  100.0kPa  0.0070.072  0.0146  0.0123  0.0142  0.0154 
EthanolwaterSrBr_{2}  100.6kPa  0.0070.064  0.0197  0.0197  0.0162  0.0157 
EthanolwaterSrCl_{2}  100.0kPa  0.0040.045  0.0196  0.0196  0.0194  0.0191 
EthanolwaterSrNO_{3}  100.0kPa  0.0020.053  0.0192  0.0174  0.0160  0.0156 
Acetic acidwaterCH_{3}COOK  100.0kPa  0.00990.1199  0.0162  0.0138  0.0114  0.0109 
Acetic acidwaterCH_{3}COOLi  100.0kPa  0.0210.127  0.0134  0.0087  0.0112  0.0108 
Propanolwater[emim] [triflate]  100.0kPa  0.05330.3165  0.0823  0.0796  0.0343  0.0294 
Propanolwater[beim] [triflate]  100.0kPa  0.04250.3239  0.0917  0.0904  0.0334  0.0294 
Propanolwater[bmpyr] [triflate]  100.0kPa  0.05270.3274  0.0877  0.0867  0.0346  0.0307 
Average  0.035  0.033  0.021  0.019 
^{a.}Range of salt concentration for the selected data.
^{b.}Results obtained by means of Furter equation.
^{c.}Results obtained by means of Wu equation.
^{d.}Results obtained by means of FurterRegular Solution model.
^{e.}Results obtained by means of FurterSolvation model.
System  Furter^{a}  Wu^{b}  FRS^{c}  FS^{d}  

k  k  k'  k  k'  A  ΔA  h_{10}  h_{20}  
PropanolwaterLiNO_{3}  6.02  7.37  13.67  4.23  4.33  6.40  2.17  10.62  2.80 
Acetic acidwaterCH_{3}COONa  0.39  0.03  4.59  1.41  4.47  0.83  2.24  2.72  1.34 
PropanolwaterLiCl  7.81  7.74  0.63  6.33  4.22  8.44  2.11  13.86  4.03 
PropanolwaterCuCl_{2}  4.55  6.28  19.69  4.04  1.32  4.70  0.66  6.14  2.92 
PropanolwaterCa(NO_{3})_{2}  10.17  13.09  36.03  6.85  6.19  9.95  3.10  18.61  4.16 
EthanolwaterCH_{3}COOK  6.03  7.64  20.82  2.77  7.62  6.58  3.81  11.77  1.97 
AcetonemethanolLiNO_{3}  7.03  17.29  89.46  7.45  43.10  14.10  21.55  130.62  31.84 
Ethanolwater[C6mim] [Cl]  1.68  4.42  16.10  2.65  2.42  1.44  1.21  0.24  0.41 
EthanolwaterKNO_{3}  9.35  9.34  0.17  13.25  48.87  11.19  24.44  10.08  9.32 
EthanolwaterNaNO_{3}  7.12  5.70  13.55  7.98  11.93  2.01  5.96  5.31  4.38 
EthanolwaterCuCl_{2}  5.45  5.44  0.16  5.12  1.52  5.88  0.76  5.10  4.82 
EthanolwaterCoCl_{2}  10.85  14.53  81.85  12.71  4.69  10.36  2.35  8.83  9.52 
EthanolwaterSrBr_{2}  16.79  16.77  0.60  13.91  11.38  19.60  5.69  33.78  9.63 
EthanolwaterSrCl_{2}  17.28  17.28  0.02  16.04  7.20  19.64  3.60  29.44  12.01 
EthanolwaterSrNO_{3}  10.03  3.39  177.80  13.11  31.36  2.57  15.68  6.77  9.27 
Acetic acidwaterCH_{3}COOK  2.01  1.93  48.79  1.46  7.56  2.32  3.78  6.00  1.51 
Acetic acidwaterCH_{3}COOLi  1.69  4.60  31.41  3.15  3.19  1.55  1.59  0.04  3.33 
Propanolwater[emim] [triflate]  1.45  3.32  8.02  2.30  10.39  2.89  5.20  8.99  3.37 
Propanolwater[beim] [triflate]  0.03  1.44  5.03  3.69  10.33  1.47  5.16  7.34  6.57 
Propanolwater[bmpyr] [triflate]  0.48  1.49  4.01  3.10  9.79  1.80  4.90  7.32  5.15 
^{a.} Parameters obtained for Furter equation.
^{b.}Parameters obtained for Wu equation.
^{c.}Parameters obtained for FurterRegular Solution model.
^{d.}Parameters obtained for FurterSolvation model.
The FRS and FS models yielded better results than the original Furter equation, as was expected since the proposed equations comprise two parameters. In comparison to the twoparameter ^{Wu model (Wu et al., 1999}), the FRS and FS models have a better performance in most of the analyzed systems. The average errors for every model show to what extent the models developed in this work improved the correlation of experimental data. The parameters regressed for each model are summarized in Table 2; in particular, parameters for both forms of the FRS model are informed. As can be seen, the restriction of positive values for the solvation parameters at infinite dilution (h_{ i } ^{0}) has been relaxed in order to obtain the best OF result (Eq. 39). A comparison of the curves obtained from each model for the VLE data of the acetone + methanol + lithium nitrate system (^{Vercher et al., 2006}) is depicted in Fig. 1. Both the FRS and FS models yielded the better correlations.
PROCESS SIMULATION
The extended models proposed in this work can be readily incorporated in a simulator of SED columns. Taking advantage of their mathematical simplicity, the models developed as extensions of the Furter equation can offer a good starting point for simulators based on algorithms for simultaneous resolution of nonlinear equation systems, e.g. NewtonRaphson’s algorithm. As is well known, these algorithms are sensitive to the initial estimates, and a large error in the initial estimates can contribute to the convergence failure. On the other hand, as proved in a previous work (^{Muzzio and Timmermann, 2014}), the temperature profiles of SED columns have significant differences in comparison with conventional columns, hampering the estimation of compositions from usual models where temperature is an input. Fortunately, the extended models based on the Furter equation are not limited by this requirement, thus providing initial estimates from the approximate knowledge of compositions and flows of input streams. To obtain this approximate knowledge, a SED column can be simulated as a conventional column in a first approach, i.e., a column without electrolyte. From the results of liquid and vapor flows and compositions for each stage, and the knowledge of the inlet stream of salt, an estimate of the salt composition in the liquid phase for each stage can be calculated. Then, in the Furter estimationcorrection (FEC) approach, the vapor composition at each stage is corrected by means of, for example, the FRS model. These corrected compositions are the initial estimates that will be used to solve the complete nonlinear equation system (^{Naphtali and Sandholm, 1971}). The different steps of the FEC algorithm, in comparison with the stepwise heat addition (SHA) algorithm (Muzzio and Timmermann, 2014), are shown in Fig. 2. To prove the utility of the proposed approach, the FRS model was incorporated and tested in the GKTM simulator, where two systems were simulated: ethanolwaterpotassium acetate, and 1propanolwaterlithium nitrate. The properties and process conditions considered for these systems are tabulated in Table 3. The simulations carried out by using the FECFRS approach (Furter estimationcorrection algorithm, where estimations are carried out by means of the FurterRegular Solution model), were compared to the ones performed by the previous SHA algorithm. As can be seen from the results in Figures 36, the number of iterations and convergence time were significantly enhanced by using the FECFRS approach.
System  1PropanolWaterLiNO_{ 3 }  EtOHWaterKAc  

Algorithm  SHA^{a}  FECFRS^{b}  SHA^{a}  FECFRS^{b} 
Rigorous model  ENRTL  ENRTL  ENRTL  ENRTL 
Main feed  flow rate ("F")  100 kmol/h  100 kmol/h  100 kmol/h  100 kmol/h 
Main feed  mole fraction ("Fx1")  0.3  0.3  0.3  0.3 
Main feed  water mole fraction ("Fx2")  0.7  0.7  0.7  0.7 
Main feed  temperature ("TF")  362 K  362 K  357 K  357 K 
Main feed  stage  4  4  514  514 
Salt feed  flow rate ("Q")  1 kmol/h  1 kmol/h  3 kmol/h  3 kmol/h 
Salt feed  mole fraction ("Qx1")  0.80.95  0.80.95  0.95  0.95 
Salt feed  salt mole fraction ("Fx2")  0.050.2  0.050.2  0.05  0.05 
Salt feed  stage  1  1  1  1 
Salt feed  temperature ("TQ")  360 K  360 K  360 K  360 K 
Operating pressure ("P")  1 atm  1 atm  1 atm  1 atm 
Number of equilibrium stages ("N")  7  7  716  716 
Condenser duty ("Qc")  100000 kJ/h  100000 kJ/h  500000 kJ/h  500000 kJ/h 
Reboiler duty ("Qr")  100000 kJ/h  100000 kJ/h  100000 kJ/h  100000 kJ/h 
^{a.}Stepwise heat addition algorithm.
^{b.}Furter estimationcorrection algorithm, where estimations are carried out by means of the FurterRegular Solution model.
The FECFRS algorithm proved to be faster than the SHA algorithm, as depicted in Fig. 3. The difference is more evident when the salt values are low because the corrected estimates in the FECFRS algorithm are close to the final solution, thus requiring few iterations in the NewtonRaphson step (Fig. 4). The rate of convergence for the FECFRS algorithm is also superior when applied to systems with high condenser and reboiler duties, because the SHA algorithm requires more iterations between NewtonRaphson and heat addition steps. As can be seen from Fig. 5 and Fig. 6, the FECFRS is less affected than the SHA algorithm by the addition of stages in the simulated column. Finally, when the salt composition in the column stages are high, the FECFRS algorithm proved to be more stable than the SHA algorithm.
CONCLUSIONS
In this work, the Furter equation has been studied and extended by means of the activity models approach. As a result of this analysis, two new models have been proposed as extensions of the Furter equation. These new models retain the simplicity and improve the average deviation of vapor phase mole fraction in comparison to previous models, including another twoparameter model proposed by ^{Wu et al. (1999}), even when applied to the entire salt/solvent concentration range. In addition, through the consideration of activity coefficients, the interactions between the different components (volatile and nonvolatile) were included in the model. As a consequence, the relationship between the Furter constant and the proportion of the volatile components arose naturally.
Although the new models developed in this work are an improvement over previous equations, more important is the possibility offered by the activity coefficient approach to be used as a tool to extend the Furter equation toward even more accurate models.
Finally, the Furter extended models have been integrated into a SED column simulator in order to enhance the stability and rate of convergence. Since these models are not affected by the uncertainty in the temperature profile of the distillation column, the vapor compositions at each stage can be corrected from a conventional distillation column to a SED column. As has been proved from the simulation results of two different systems, the inclusion of Furter extended models into the simulator software led to a more efficient algorithm.
NOMENCLATURE
Aij: 
parameter of the regular solution model. 
C,C',C'': 
constant of integration 
f _{ i }: 
fugacity of component i in the liquid phase 
GE: 
molar excess Gibbs free energy 
h1: 
solvation number of salt with component i. 
Hi: 
molar enthalpy of component “i” 
k, k': 
parameter of the salt effect. 
nα: 
number of moles “i” in phase. 
P: 
pressure. 
R: 
universal constant gas. 
Si: 
molar entropy of component “i”. 
T: 
temperature. 
Vi: 
molar volume of component “i”. 
xi 
liquid phase mole fraction of component “ ” based on undissociated species. 
xia: 
effective mole fraction of solvent “i” on a salt free basis. 
xia: 
vapor phase mol fraction of component “ ” in the saltfree system. 
: 

zi 
liquid phase mol fraction for component “ ”, on a salt free basis. 
. 
Greek letters 
α: 
relative volatility 
(: 
1/(1+z _{ 3 } ) 
ˆ: 
fugacity coefficient of the component i in the mixture. 
i γ: 
activity coefficient for the solvent i 
μα: 
chemical potencial of component “i” in the phase. 
μα: 
derivative of chemical potential of component “i” with respect to “j” component in the phase. 
v3: 
electrolyte stoichiometric coefficient. 
σ: 
saline coefficient. 
. 
Subscripts 
F: 
free molecule 
s: 
system with salt 
+: 
property of the more volatile component 
0: 
saltfree system 
1,2: 
volatile components 
3: 
nonvolatile component (salt or ionicliquid) 
. 
Superscripts 
LR: 
longrange. 
SR: 
shortrange. 
(: 
infinite dilution of the salt 
°: 
pure component 