You are on page 1of 13

Fluid Phase Equilibria, 94 ( 1994) 115- 127 115

Elsevier Science B.V.

A gE model for single and mixed solvent electrolyte


systems
2. Results and comparison with other models

Hans-Martin Polka a, Jiding Li b and Ji.irgen Gmehling a**


a Lehrstuhl fiir Technische Chemie (FB 9), Universitiit Oldenburg, Postfach 2503, D-261 11
Oldenburg (Germany)
bDepartment of Chemical Engineering, Tsinghua University, Peking (China)
(Received May 24, 1993; accepted in final form August 24, 1993)

ABSTRACT

Polka, H.-M., Li, J. and Gmehling, J., 1994. A gE model for single and mixed solvent
electrolyte systems. 2. Results and comparison with other models. Fluid Phase Equilibria,
94: 115-127.

A new gE model describing the behaviour of single and mixed solvent electrolyte systems
is used to calculate vapour-liquid equilibria of 185 ternary systems (two solvents one salt),
and vapour pressure, osmotic coefficients and activity coefficients of 362 binary systems (one
solvent one salt). The results are compared with the models of Sander, Macedo, Pitzer,
Bromley and Chen. For the large data base used, the new model improves the results not
only for the description of single solvent data but also for the calculation of VLE for mixed
solvent systems. Despite the relative improvement compared to other models, in certain cases
serious deviations are obtained.

Keywords: theory, excess functions, vapour-liquid equilibria, electrolytes.

INTRODUCTION

Numerous models are available for the calculation of activity coefficients


and osmotic coefficients of single solvent electrolyte systems. Starting with
the basic work of Debye and Hiickel (1923), an extensive amount of
research work has been performed to develop equations suitable for the
calculation of more highly concentrated systems. Guggenheim (1935a, b),
based on earlier works of Brernstedt (1922), suggested an empirical exten-
sion introducing a coefficient specific for the interactions between cation and

* Corresponding author.

0378-3812/94/$07.00 0 1994 - Elsevier Science B.V. All rights reserved


SSDI 0378-3812(93)02424-L
116 H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127

anion. Using the pressure equation of statistical thermodynamics, Pitzer


(1973a, 1979) pointed out that the interaction coefficient is a function of the
ionic strength. Taking into account the ionic strength dependence a semiem-
pirical equation was proposed, which could represent the experimental
properties with high accuracy up to a ionic strength of around 6. Bromley
(1973) suggested a one-parameter equation found empirically from fitting to
experimental data. The solute specific parameters could successfully be
approximated as the sum of contributions from cation and anion. Recently,
different equations were proposed based on a local-composition model like
UNIQUAC or NRTL. Cruz and Renon (1978) combined the Debye-
Htickel expression with the NRTL model and a Born model contribution,
taking into consideration the change of the dielectric constant with salt
concentration. Four adjustable parameters are necessary to describe a
system containing one solvent and one salt. A similar approach was
suggested by Ball et al. (1985). In contrast to the model of Cruz and Renon,
the calculation of the Born contribution was carried out in a less empirical
way, which reduced the number of adjustable parameters to two for one
binary system. Chen et al. (1982) combined the electrostatic function of the
Pitzer model with a local composition term, which is an extension of the
NRTL equation to electrolyte solutions. The local mole fractions are
calculated based on two assumptions: (1) repulsion of ions carrying charges
of the same sign; (2) local electroneutrality around solvent molecules. Two
interaction parameters are necessary to describe a binary system. Haghtalab
and Vera (1988) adopted the general ideas of Chen using the original
Debye-Htickel expression instead of the Pitzer-Debye-Htickel formula,
and expressed the local composition through nonrandom factors. Liu et al.
(1989) introduced a modified Debye-Hiickel term, which covers only the
long-range interactions between the central ion and all ions outside the first
coordination shell. The electrostatic interactions between ions within the
first coordination shell are included in the short-range contribution, which
is based on a previous derivation of the three-parameter Wilson equation
(Renon and Prausnitz, 1969). The model parameters are ion specific and
were simultaneously fitted to a data base containing 55 data sets of single
solvent electrolyte systems (Liu and Gren, 1991).
The above-mentioned electrolyte models have at least three serious disad-
vantages:
(1) extension to mixed solvent systems is not taken into account and is in
some cases simply impossible;
(2) the model parameters are very often only available for aqueous
systems;
(3) the temperature is mostly limited to 25C which reduces the range of
applicability considerably.
H.-M. Polka et al. 1 Fluid Phase Equilibria 94 (1994) 115- 127 117

To overcome these difficulties a new expression for the excess Gibbs


energy was proposed, which has the capability of describing both single and
mixed solvent electrolyte systems. The expression consists of three contribu-
tions: a Debye-Hfickel term to account for long-range electrostatic interac-
tions, the UNIQUAC equation for the description of short-range
interactions between all particles, and a middle-range contribution to in-
clude all indirect effects of the charge interactions. Theoretical details,
equations and parameters are given elsewhere (Li et al., 1993). The object of
this paper is to use the new model for the calculation of single and mixed
solvent systems using a large data base and to compare the results with
those of other well-known models. For single solvent systems the equations
of Pitzer, Chen and Bromley were chosen, mainly because parameters from
these models are given for most aqueous systems in the literature.
Only a few published models are suitable for the calculation of both
single and mixed solvent systems. Mock et al. (1986) extended the single
solvent model of Chen to mixed solvent systems neglecting the long-range
contribution. Nine parameters are necessary to describe a system containing
two solvents and one salt. Sander et al. (1986) reported about an extension
of the UNIQUAC equation to mixed solvent systems. Introducing a con-
centration dependence into the UNIQUAC parameter for ion-solvent
interactions and adding a simplified Debye-Hiickel term made it possible to
describe successfully different solvent mixture systems. Macedo et al. (1990)
has used a theoretically more stringent expression for the electrostatic part
based on a work of Cardoso and OConnell (1987) but received only minor
improvement. Kikic et al. (1991) substituted the UNIQUAC equation by
the original UNIFAC group contribution model without further modifica-
tions. The deviations between ex~rimental and calculated vapour phase
mole fractions and pressures are in general poor compared to the model of
Sander and Macedo.
Because of their similarity to the approach discussed here and the greater
range of applicability, the models of Sander and Macedo were chosen to
compare the results of our model.

SINGLE SOLVENT SYSTEMS

Most of the data for single solvent systems available in the literature deals
with aqueous systems at room temperature. Useful data compilations
were published by Robinson and Stokes (1948) and Hamer and Wu
(1972). In most cases the data are givne for the osmotic coefficient, the
mean ionic activity coefficient or the vapour pressure of the solution
at constant temperature. In addition we have distinguished between a
high and low concentration range (low concentration range < 6 mol/kg),
118 H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127

TABLE 1
Mean percentage deviation for osmotic coefficients and mean ionic activity coefficients for 28
aqueous electrolyte systems at 25C. Maximum molality (max. m) up to 6 mol kg-

System Max. Our model Pitzer Chen Bromley


m
Ayi A4 Ayi W AY+ A+ Ayi A&

H,O NaF 0.98 0.04 0.06 0.28 0.09 0.19 0.11 0.21 0.06
H,O NaCl 6.14 0.10 0.09 0.40 0.08 1.36 0.93 2.06 0.91
Hz0 KC1 4.5 0.60 0.35 0.26 0.06 0.26 0.20 0.12 0.10
H,O M&l, 5.0 2.11 1.24 0.96 0.16 16.35 8.97 10.01 3.94
H,O SrCl, 4.0 0.46 0.32 0.72 0.22 7.18 4.95 4.09 1.82
H,O FeCI, 2.0 0.22 0.18 0.54 0.14 2.30 2.09 0.89 0.68
H,O CoCl, 4.0 3.68 2.89 1.59 0.55 4.64 3.31 2.95 1.63
H,O NiCI, 5.0 2.58 1.82 5.01 1.67 7.93 4.83 3.53 1.74
Hz0 CuCl, 6.0 0.41 0.50 10.44 5.15 3.11 1.98 35.42 11.28
H,I MgBr, 5.0 1.17 0.55 1.87 0.22 19.36 9.30 9.13 3.23
H,O SrBr, 2.0 0.20 0.22 0.43 0.12 2.93 2.60 1.53 0.89
H,O BaBr, 2.0 0.52 0.52 0.16 0.13 2.04 1.94 1.69 0.54
H,O Lil 3.0 0.61 0.51 1.51 0.52 1.63 1.06 1.29 0.74
H,O KI 4.5 0.28 0.44 0.422 0.13 0.37 0.22 4.82 1.99
Hz0 MgI, 5.0 2.24 1.25 4.72 0.53 25.13 Il.44 8.29 2.62
H,O Cal, 2.0 2.36 2.19 2.70 0.10 3.89 2.94 3.10 1.60
H,O Sri, 2.0 1.70 1.46 1.82 0.12 3.73 3.02 0.76 0.85
H,O Bal, 2.0 0.24 0.39 1.59 0.42 3.10 2.41 10.88 4.09
H,O KNO, 3.5 0.16 0.38 1.04 0.44 0.54 0.62 1.46 1.71
H, 0 Mg(NO, )z 2.0 0.25 0.45 0.61 0.24 8.96 5.65 17.31 7.57
H,O Sr(NO,), 4.0 0.20 0.18 3.14 3.43 2.56 2.53 20.28 12.81
H,O Co(NWz 5.0 0.86 0.76 0.55 0.20 8.83 5.50 10.23 3.25
Hz0 Zn(NW, 6.0 2.83 3.30 5.29 1.85 12.03 6.56 8.39 2.64
H,O LiClO, 4.5 0.23 0.48 0.70 0.15 1.84 1.14 2.77 1.33
H,O NaClO, 6.0 0.11 0.11 0.43 0.08 0.76 0.52 4.25 1.55
H,O Mg(ClWz 4.0 1.35 0.66 4.24 0.19 16.76 8.35 4.98 2.33
H,O Zn( C10J2 4.0 1.13 3.29 4.46 0.26 16.97 8.39 9.31 3.59
H,O NaClO, 3.0 0.12 0.07 0.82 0.15 0.45 0.29 1.08 0.96

Total mean % dev. 0.95 0.88 2.02 0.62 6.26 3.63 6.46 2.73

because most of the above-mentioned electrolyte models fail at higher


concentrations. In Table 1 the mean absolute and percentage deviations
between experimental and calculated values for the osmotic and mean
activity coefficient of 28 aqueous systems at 25C are listed covering a
concentration range up to a molality of 6. The model parameters were
taken from the literature (Pitzer, 1973b, Bromley, 1973, Chen, 1982).
Similar results are obtained for our model and Pitzers equation. For the
H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127 119

TABLE 2
Mean percentage deviation for osmotic coefficients and mean ionic activity coefficients for 15
aqueous electrolyte systems at 25C. Maximum molality up to 26 mol kg-

System Max. Our model Pitzer Chen Bromley


m
AYE A4 AY, A# AYE A4 &i A+

H,O KF 17.5 1.75 1.03 7.46 2.68 14.16 7.77 2.90 1.44
H,O CaClz 10.0 9.52 4.83 6.73 6.43 19.62 12.44 7.86 5.17
H,O NaBr 9.9 0.60 0.43 1.01 0.51 3.96 2.38 0.34 0.23
H,O NaI 12.0 0.92 0.52 3.47 1.23 10.26 5.47 2.94 0.89
H1O LiNO, 20.0 1.37 1.06 5.99 3.36 9.81 4.98 26.16 8.40
H,O NaNO, 10.83 1.67 1.56 4.36 3.67 3.94 3.45 6.59 5.43
H,O AgNO, 13.0 0.41 0.53 3.73 6.84 2.16 3.97 12.21 28.74
H,O NH,NO, 25.95 0.95 0.78 5.36 7.77 5.27 5.50 39.65 84.69
H,O KOH 20.0 1.93 0.91 84.21 8.87 20.75 9.98 5.82 2.14
H,O LiCl 19.22 5.19 2.23 81.04 8.02 23.18 10.75 8.78 3.03
H,O NH,Cl 7.41 0.26 0.23 0.53 0.10 0.31 0.13 1.74 1.28
H,O MnCl, 8.0 1.63 1.37 3.10 1.82 5.11 3.32 60.74 12.25
H,O CaBr, 8.0 6.12 2.48 12.54 3.21 32.15 14.39 15.60 5.04
H,O CatNO,), 8.0 1.12 1.07 12.89 11.14 6.76 4.99 19.69 7.84
H,O Cu(NO,l, 8.0 2.53 1.71 14.86 8.04 11.96 6.82 9.96 3.35

Total mean % dev. 2.40 1.38 16.49 4.91 11.29 6.42 14.73 11.33

osmotic coefficients and mean activity coefficients the overall deviations do


not exceed one percent in most cases. The Pitzer equation shows larger
deviations for the mean activity coefficient in a few cases (CuCl,,
Zn(NO,),). The results for the models of Chen and Bromley are clearly
worse. In some cases only a qualitative agreement is obtained with a
percentage deviation exceeding 10%.
Results for systems at high electrolyte concentration are given in Table 2.
Here the maximum molality reaches up to 26 mol kg-. As can be clearly
seen using the model parameters from the literature, in many cases serious
deviations are obtained for concentrated solutions. This is because data
points at high concentration were excluded to yield a better fit for the dilute
region. However, for many practical applications a reasonable description
of the concentrated region is necessary. The parameters for the present
model were therefore estimated using also data for solutions of molality
greater than 6. The deviations obtained are still in a range, which also
makes the model useful for practical calculations on highly concentrated
systems. A few typical examples are given in Fig. 1 for the mean ionic
activity coefficient. The model is able to represent both the dilute and
concentrated region with reasonable accuracy.
120 H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127

Fig. 1. Experimental ( LI) and calculated (-, - - -, - - -. -. -. -) mean ionic activity


coefficients for aqueous electrolyte solutions at 25C. (A) KF, (B) Ca(NO,),, (C) KOH.
Data from Stokes (1948) and Hamer and Wu (1972).

Up until now, only calculations at 25C have been presented; this is


because most of the data were measured at this temperature. Although the
thermodynamic behaviour of electrolyte solutions can be strongly affected
by a temperature change, parameters for the models of Pitzer, Chen and
Bromley were estimated using data for 25C only. The methods proposed by
the authors to obtain the parameters for a different temperature are mainly
empirical (Zemaitis et al., 1986) and require additional parameters, which
are not available for most systems. Therefore, the models of Sander and
Macedo were used as a comparison to our own results for the calculation of
systems at different temperatures, because the published model parameters
are valid for a larger temperature range. Table 3 lists the results for the
vapour pressure data. of 41 binary systems. Only data sets for which all
model parameters are available in the literature were selected. The results
from the models of Sander and Macedo are quite similar, which is reason-
able since the two models differ only in the electrostatic contribution. Here
the total mean percentage deviations amount to 3.1 and 2.9% respectively.
In some cases the deviations exceed lo%, which is mainly caused by the high
electrolyte concentration. The deviations obtained for our model are smaller
for 31 of the 41 systems listed. For the high concentration range in
particular, much better results are obtained (See Fig. 2).
As a whole, a data base of 239 binary data sets was used to check the
capabilities of our model, The mean absolute deviation for the osmotic
coefficient is 0.021 (82 data sets) and for the mean ionic activity coefficients
0.059 (57 data sets), corresponding to relative deviations of 1.44 and 2.22%
respectively. For the low concentration range up to a salt molality of
6 mol kg- the deviations are smaller than 1%. For most of the binary
H.-M. Polka et al. 1 Fluid Phase Equilibria 94 (1994) 115-127 121

TABLE 3
Mean absolute (abs.) and percentage (%) deviation between experimental and calculated
vapour pressures for 41 aqueous electrolyte systems at different temperatures

System 9 (C) Max. molality AP (kPa)

Our model Sander Macedo

Abs. % Abs. % Abs. %

H,O LiCl 30 18.46 0.03 2.1 0.17 14.1 0.16 11.8


H,O LiCl 40 18.46 0.05 2.0 0.28 13.0 0.26 10.7
H,O LiCl 50 18.46 0.11 3.0 0.44 12.1 0.40 9.7
H,O LiCl 60 18.46 0.25 4.2 0.7 11.29 0.60 9.1
H,O LiCl 70 18.46 0.53 5.7 0.97 10.2 0.86 8.2
H,O LiCl 75 3.20 0.22 0.6 0.44 1.3 0.54 1.6
H,O LiCl 100 3.20 0.87 0.9 0.67 0.7 0.69 0.1
H,O LiCl 125 3.20 1.01 0.5 0.89 0.4 0.86 0.4
H,O LiCl 150 3.21 1.43 0.3 0.96 0.2 1.51 0.4
H,O NaCl 30 5.16 0.03 0.8 0.05 1.3 0.02 0.5
H,O LiCl 18 5.66 0.01 0.8 0.03 2.0 0.03 1.5
H,O NaCl 25 5.66 0.01 0.3 0.04 1.3 0.03 1.2
H,O NaCl 50 5.66 0.05 0.5 0.08 0.7 0.10 0.9
H,O KC1 30 4.29 0.04 1.1 0.04 1.1 0.07 1.7
Hz0 KC1 40 4.29 0.04 0.7 0.03 0.4 0.07 1.0
H,O KC1 50 4.29 0.05 0.4 0.03 0.3 0.07 0.6
H,O KC1 60 4.29 0.06 0.3 0.07 0.4 0.07 0.4
H,O Kc1 70 4.29 0.15 0.5 0.14 0.5 0.12 0.4
H,O CaCl, 30 7.89 0.05 2.4 0.12 4.9 0.14 5.4
H,O CaCl, 40 7.89 0.08 2.2 0.18 4.1 0.20 4.6
Hz0 CaCl, 50 7.89 0.12 2.1 0.25 3.4 0.29 3.9
H,-0 CaCl, 40 11.61 0.17 4.9 0.30 7.3 0.32 7.7
H,O CaCl, 50 11.61 0.17 2.9 0.36 4.9 0.39 5.3
H,O CaCl, 60 11.51 0.33 3.5 0.54 4.7 0.59 5.0
H,O CaCl, 70 12.40 0.56 4.0 0.82 4.9 0.90 5.3
H,O CaCll 80 13.80 1.03 5.6 1.21 5.3 1.33 5.6
H,O NaBr 30 7.98 0.03 1.1 0.07 2.5 0.06 2.2
Hz-o NaBr 40 7.98 0.04 0.7 0.11 2.0 0.09 1.8
H,O NaBr 50 7.98 0.04 0.4 0.14 1.6 0.11 1.4
H,O NaBr 60 7.98 0.04 0.3 0.17 1.3 0.15 1.2
H,O NaBr 70 7.98 0.10 0.4 0.18 0.8 0.22 1.1
H,O KBr 30 4.35 0.02 0.6 0.04 1.0 0.03 0.7
H,-0 KBr 40 4.35 0.03 0.5 0.07 1.1 0.04 0.5
H,O KBr 50 4.35 0.03 0.3 0.10 0.9 0.04 0.4
H,O KBr 60 4.35 0.04 0.2 0.15 0.9 0.07 0.4
H,O KBr 70 4.35 0.10 0.4 0.15 0.5 0.14 0.5
H,O KOAc 85 4.10 0.15 0.3 0.34 0.7 0.55 1.1
H,O KOAc 90 5.26 0.29 0.5 0.84 1.4 1.24 2.1
H,O KOAc 9.5 4.53 0.10 0.1 0.43 0.6 0.77 1.0
H,O KOAc 100 3.97 0.98 1.1 0.94 1.0 0.85 0.9
H,O KOAc 105 4.51 0.92 0.9 0.53 0.5 0.64 0.6
Total mean dev. 0.25 1.5 0.34 3.1 0.38 2.9
122 H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127

m m m

Fig. 2. Experimental (A) and calculated (-) vapour pressures for H,O-LiCl at different
temperatures. (A) Our model, (B) Sander, (C) Macedo. Data from Patil et al. ( 1990).

data - especially for nonaqueous solvent systems - the vapour pressure


of the salt solution is given as a function of salt concentration at constant
pressure. Here, the mean absolute deviation between calculated and experi-
mental pressure is 0.2 kPa (223 data sets) corresponding to a 1.4% relative
deviation.

MIXED SOLVENT SYSTEMS

Most of the VLE data available in the literature for mixed solvent
electrolyte systems cover alcohol-water systems. Mixtures containing only
nonaqueous solvents have received minor attention, which may be at-
tributed to the low salt solubihty. Using the present model, 185 systems in
total were studied. The data base contains 152 alcohol-water systems and
33 nonaqueous solvent systems. Among them, 157 were measured at
constant pressure and 28 at constant temperature. In the case of isobaric
data, the mean absolute deviations obtained are 0.018 for the vapour phase
mole fraction and 1.08 K for the temperature, as given in Table 4. For 28
isothermal data sets the mean pressure deviation was 0.78 kPa (see Table 5).
Again, the models of Sander and Macedo were used for comparison with
our results. In addition the NRTL extension proposed by Mock, Evans and
Chen was applied. Unfortunately, due to the gaps in the parameter tables of
the models mentioned above, it is not possible to obtain results for all 185
data sets. Otherwise, in order to compare two model equations, it seems to
be more objective to use the same data base for both equations. Therefore,
each equation was tested using the greatest possible data base, depending on
the number of parameters published. For each data base, results are
presented for the respective model and our new expression. As can be seen
H.-M. Polka et al. 1 Fluid Phase Equilibria 94 (1994) 115-127 123

TABLE 4
Mean absolute and percentage deviation of temperature and vapour phase mole fraction for
the electrolyte models of Macedo, Sander and Mock-Evans-Chen in comparison to our
model. All data sets measured at constant pressure

Model No. of AY Ay (%) AT(K) AT (%)


data sets

Our model 157 0.018 3.3 1.08 1.3

Macedo 116 0.028 5.0 1.55 1.8


Our model 116 0.017 3.2 1.22 1.4

Sander 91 0.022 4.1 1.36 1.6


Our model 91 0.018 3.3 1.21 1.4

Mock-Evans-Chen 63 0.020 3.8 0.90 1.1


Our model 63 0.015 3.0 0.96 1.1

TABLE 5
Mean absolute and percentage deviation of pressure and vapor phase mole fraction for the
electrolyte models of Macedo, Sander and Mock-Evans-Chen in comparison to our model.
All data sets measured at constant temperature

Model No. of AJ Ay (%) AP (kPa) AP (%)


data sets

Our model 28 0.016 3.6 0.78 2.9

Macedo 21 0.020 4.2 1.10 5.2


Our model 21 0.017 3.8 0.70 3.1

Sander 21 0.021 4.3 0.88 4.5


Our model 21 0.017 3.8 0.70 3.1
Mock-Evans-Chen 11 0.026 4.2 2.28 10.1
Our model 11 0.017 2.6 1.08 3.7

from Tables 4 and 5, on average the calculation for each data base gives
better results for our model. In the case of isobaric data the extended NRTL
model shows the smallest difference to our results. However, this is the
model with the lowest range of applicability, which is due to the fact that
the model parameters are specific for solvent-salt interactions instead of
solvent-ion interactions. Moreover, in some cases different parameter sets
must be used for the same system depending on the temperature and
pressure conditions, which indicates poor predictive capabilities.
124 H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127

Fig. 3. Experimental (0, A) and calculated (-) X, y diagram for ethanol-water-potas-


sium acetate at atmospheric pressure. (A) Our model, (B) Sander. (C) Macedo. Data from
Schmitt (1979) and Vercher et al. (1991).

As a typical example for the VLE of mixed solvent systems under the
influence of salt components, Fig. 3 shows experimental and calculated vapour
phase compositions for the ethanol/water/potassium acetate system, using
the models of Sander, Macedo and the present approach. The addition of
salt results in an increase of alcohol concentration in the vapour phase
(salting-out). At sufficiently large salt concentration (x,,rt = 0.1) the
azeotropic point is completely removed. This behaviour is well described by
all three models used, but larger deviations occur for the data set at higher
salt concentration, which appears to be better correlated by the present model.
Despite the fact that the model discussed here gives better results on
average, it is important to note that serious deviations from the experimen-
tal data are sometimes obtained. Figure 4 gives an example of the calcula-

3 rn=.l 8 mole/kg
_A m=32 .,

ci m=zo

a rn=,?

< m=cle. ,)

0 ln=clI

LL_
04 06 08 02 0.4 0 6 0 8 02 04 0.6 08 1.0

x, (salt free) I, (saltfree) Y, (saltfree)


Fig. 4. Experimental and calculated x, y diagram for 2-propanol-water-calcium chloride at
six different salt concentrations. (A) Our model, (B) Sander, (C) Macedo. Data from Ohe et
al. (1969).
H.-M. Polka et al. 1 Fluid Phase Equilibria 94 (1994) 115-127 125

tion of VLE covering a wide range of salt concentrations from dilute to


almost saturated solution. The three models used only give a rough qualita-
tive description of the VLE behaviour although the present model seems to
be slightly superior in representing the high concentration range. From the
scattering of data points it may be assumed that the quality of the data sets
is rather low.
Poor results are also obtained for the methanol/ethanol/CaClz and 2-
propanol/water/LiCl systems. In these cases, for some data points the
deviation exceeds 10%. It may be assumed that this is at least partly due to
the underlying simplifications of the model approach discussed here, for
example, the assumption of complete dissociation and the rather empirical
estimation of the ionic strength dependence of the middle-range interaction
coefficients.

MODEL PARAMETERS

As discussed in detail in the first part of this paper, four parameters are
necessary for the present approach to describe the interaction between a
solvent-ion or ion-ion pair: two for the short-range interaction and two
concentration dependent interaction coefficients for the so called middle-
range contribution. In a simple electrolyte solution containing one solvent
and one electrolyte assumed to be fully dissociated, there are three binary
pairs to be considered, which results in a total number of 12 interaction
parameters. Although this appears to be rather high compared to the
models of Sander, Macedo (seven interaction parameters) and the equation
of Mock, Evans and Chen (two interaction parameters), it can be shown
that for a sufficiently large number of systems the present model requires the
smallest number of parameters. Dependent on the total number of cations
N,, anions N, and solvents N, included the total number of parameters N
necessary to describe each possible interaction between the species can be
calculated using the following formulas:
This model N = (N,N, + N,N, + N,N,) x N,(N, - 1)
Sander, Macedo N = (N,N, + N,N, + NcN,) x 2
+ N&V, - 1) + N,N,N,
Mock-Evans-Chen N = N,N,N, x 2 + N,(N, - 1)
In Table 7 the total number of parameters for a fixed number of 20 cations
and 10 anions (which is close to the data base used here) in dependence of
the total number of solvents are listed. For up to two solvents the models
of Sander, Macedo and Mock-Evans-Chen need less parameters, but for
126 H.-M. Polka et al. / Fluid Phase Equilibria 94 (1994) 115-127

TABLE 6
Required number of interaction parameters for a given data base containing 20 cations and
10 anions in relation to the number of solvents N,

NS Our Model Sander, Macedo Mock-Evans-Chen

1 920 660 400


2 1042 922 802
3 1166 1186 1206
4 1292 1452 1612
5 1420 1720 2020
6 1550 1990 2430
7 1682 2262 2842
8 1816 2536 3256
9 1952 2812 3672
10 2090 3090 4090

more than two solvents our model requires the smallest number. This is
because in our case the interaction parameters represent only binary interac-
tions, whereas Sander and Macedo have introduced an additional parameter
expressing the concentration dependence of the cation-anion interactions,
which is specific to a special solvent component, and the parameters of the
extended NRTL model represent salt-solvent pairs.

CONCLUSION

Using a data base containing 185 mixed and 362 single solvent systems,
a new gE model was tested and compared with six other models. The results
for binary data using the equations of Pitzer, Chen and Bromley as a
comparison show a considerable improvement for high electrolyte concen-
trations. For concentrations below a molality of 6 kg mol-, the deviations
are similar to the model of Pitzer and do not exceed one percent on average.
The results from the new model for mixed solvent systems (a mean
deviation of 0.018 for the vapour phase mole fraction, 1.04 K for the
temperature and 0.78 kPa for the pressure) are better than those from the
models of Sander, Macedo and the electrolyte NRTL model of Mock,
Evans and Chen. However, in some cases rather large deviations are
obtained, which make it necessary to improve the new model further.
Unfortunately, the present problem is the serious lack of good quality data
sets for mixed solvent systems, which are necessary to estimate reliable
parameters.
H.-M. Polka et al. 1 Fluid Phase Equilibria 94 (1994) 115-127 127

REFERENCES

Ball, F.X., Fiirst, W. and Renon. H., 1985, AIChE J., 31: 392.
Bromley, L.A., 1973, AIChE J., 19: 313.
Bronstedt, J.N., 1922, J. Am. Chem. Sot., 44: 938.
Cardoso, M. and OConnell, J., 1987, Fluid Phase Equilibria, 33: 315.
Chen, C.C., Britt, HI., Boston, J.F. and Evans, L.B., 1982, AIChE J., 28: 589.
Cruz, J.L. and Renon, H., 1978, AIChE J., 24: 817.
Debye, P. and Hiickel, E., 1923, Physik. Z., 24: 185.
Guggenheim, E.A., 1935a, Philos. Mag., (Series 7) 19: 588.
Guggenheim, E.A., and Turgeon, J.C., 1935b, Trans. Faraday Sot., 51: 747.
Haghtalab, A. and Vera, J.H., 1988, AIChE J., 34: 803.
Hamer, W.J. and Wu, Y.C., 1972, J. Phys. Chem. Ref. Data, 1: 1047.
Kikic, I., Fermeglia, M. and Rasmussen, P., 1991, Chem. Eng. Sci, 46: 2775.
Li, J., Polka, H.M. and Gmehling, J., 1994. Fluid Phase Equilibria, 94: 89-l 14.
Liu, Y., Harvey, A.H. and Prausnitz, J.M., 1989, Chem. Eng. Commun., 77: 43.
Liu, Y. and GrCn, U., 1991, Chem. Eng. Sci, 46: 1815.
Macedo, E.A., Skovborg, P. and Rasmussen, P., 1990, Chem. Eng. Sci., 45: 875.
Mock, B., Evans, L. B. and Chen, C. C., 1986, AIChE J., 32: 1655.
Ohe, S., Yokoyama, K. and Nakamura, S., 1969, Kogyo Kagaku Zasshi, 72: 313.
Patil, R., Tripathi, A.D., Pathak. G. and Katti, S.S., 1990, J. Chem. Eng. Data, 35: 166.
Pitzer, K.S., 1973a, J. Phys. Chem., 77: 268.
Pitzer, KS., 1973b. J. Phys. Chem., 77: 2300.
Pitzer, K.S., 1979. Theory: Ion Interaction Approach. In: Activity Coefficients in Electrolyte
Solutions, Vol. I, CRC Press, Boca Raton, FL, p. 157.
Renon, H. and Prausnitz, J.M., 1969, AIChE J., 15: 785.
Sander, B., Fredenslund, A. and Rasmussen, P.. 1986, Chem. Eng. Sci., 41: 1171.
Schmitt, D., Ph.D. Thesis, Karlsruhe, 1979.
Stokes, R.H.. 1948, Trans. Faraday Sot., 44: 295.
Vercher, E., Muiioz, R. and Martinez-Andreu, A., 1991, J. Chem. Eng. Data, 36: 274.
Zemaitis, J.F., Clark, D.M., Rafal, M. and Scrivner, N.C., 1986. Handbook of Aqueous
Electrolyte Thermodynamics, DIPPR, New York, p. 84.

You might also like