You are on page 1of 9

Ind. Eng. Chem. Res.

2005, 44, 3289-3297 3289

Determination of Ion-Specific NRTL Parameters for Predicting


Phase Equilibria in Aqueous Multielectrolyte Solutions
Hidetoshi Kuramochi,*,† Masahiro Osako,† Akiko Kida,† Kazuyuki Nishimura,†
Katsuya Kawamoto,† Yusuke Asakuma,‡ Keisuke Fukui,‡ and Kouji Maeda‡
Research Center for Material Cycles and Waste Management, National Institute for Environmental Studies,
16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan, and Department of Chemical Engineering, University of
Hyogo, 2167 Shosha, Himeji, Hyogo 671-2201, Japan

To develop an activity coefficient model for aqueous multielectrolyte systems, ion-specific NRTL
parameters were applied to the electrolyte-NRTL model. In total, 132 ion-specific NRTL
parameters for 21 ionic species (K+, Na+, Mg2+, H+, Ca2+, Zn2+, Al3+, NH4+, Cd2+, Co2+, Cu2+,
Pb2+, Mn2+, Ni2+, Cr3+, Cl-, SO42-, NO3-, OH-, ClO4-, CrO42-) were determined from the binary
activity coefficient data of aqueous electrolytes and used to predict the activity coefficients of
electrolytes in aqueous multieletrolyte solutions without any additional parameters. Furthermore,
solubilities of inorganic salts including heavy metals in aqueous mixed electrolyte systems and
the distribution ratio of Cu in a solvent extraction process containing HCl and CuCl2 were
predicted using the activity coefficients derived from our ion-specific electrolyte-NRTL model.
The results showed that the NRTL parameters obtained in this study accurately predict phase
equilibria of systems involving inorganic salts such that this model can be used in the design of
recovery processes such as crystallization and solvent extraction.

Introduction ficients in that as the concentration increases, the


greater the effect on the phase equilibrium. In the case
The industrial production and treatment processes of crystallization, highly concentrated electrolyte solu-
dealing with inorganic compounds usually involve elec- tions are necessary. In addition, for solvent extraction
trolyte solutions. Mineral drinks, foods, seawater, and and ion exchange, highly concentrated solutions of other
wastewater all involve electrolyte solution chemistry. electrolytes must be used even if the concentration of
The recovery or removal of inorganic salts including the electrolyte of interest is low or dilute.4-6 From this
heavy metals from wastewater and wastes is an emerg- point of view, the desired activity coefficient model
ing problem for the construction of sustainable societies would be capable of representing the activity coefficient
in terms of environmental protection and resource behavior of a given ion in not only a wide concentration
recovery. These recovery and removal processes include range but also in multielectrolyte solutions.
crystallization or precipitation, solvent extraction, sorp-
tion, and ion-exchange. These processes are currently Conventionally, in organic chemical and petroleum
the topics of many studies.1-8 Generally speaking, these industries, local composition models that express the
separation processes are predicated on the phase equi- excess Gibbs free energy due to short-range inter-
librium between an aqueous phase and one other molecular interactions such as the Wilson, NRTL, and
additional phase. For example, solvent extraction is UNIQUAC models10 have been used as activity coef-
based on the dissociation equilibrium of each salt in a ficient models and contributed to the design process
homogeneous phase and the distribution equilibrium of regarding distillation, extraction, and absorption. More-
individual ions between an aqueous phase and an over, activity coefficient prediction models such as
organic phase. Crystallization is based on solid-liquid UNIFAC11 have been proposed by combining the group
equilibrium, or the solubility product defined by the contribution concept with the local composition model.
activity coefficient and molality of aqueous ions. A In terms of inorganic chemical industries, activity
review of the various thermodynamic models of elec- coefficient models are still being developed for waste-
trolyte solutions from an engineering approach was water treatment, hydrometallurgy, desalination, and
reported by Loehe and Donohue.9 crystallization processes, as well as for the understand-
The activity coefficient represents the phase equilib- ing of electrolyte solutions. Generally, activity coef-
rium properties of the aqueous phase. However, the ficients in a dilute electrolyte solution can be expressed
activity coefficient data of binary aqueous electrolyte by the Debye-Hückel equation,12 which describes the
solutions are not sufficient in calculating the phase excess Gibbs energy due to long-range (electrostatic)
equilibrium in an aqueous multielectrolyte solution, as interaction. In early studies on concentrated solutions,
the addition of other electrolytes affects the phase some equations were proposed and these are briefly
equilibrium of the electrolyte of interest. There is also described below. In the 1970s, Pitzer13-16 suggested the
a concentration dependence regarding activity coef- Debye-Hückel equation be followed by a virial expan-
sion of terms. In 1978, Cruz and Renon17 first considered
* To whom correspondence should be addressed. Tel.:
that the NRTL model18 was combined with the Debye-
+81-(0)29-850-2841. Fax: +81-(0)29-850-2091. E-mail: Hückel model. The former and latter models represent
kuramochi.hidetoshi@nies.go.jp. short-range and long-range interactions, respectively.

National Institute for Environmental Studies. In the 1980s, Chen et al.19,20 developed an electrolyte-

University of Hyogo. NRTL model similar to the above model that resulted
10.1021/ie049377u CCC: $30.25 © 2005 American Chemical Society
Published on Web 03/12/2005
3290 Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005

/
Figure 1. Behavior of mean ionic activity coefficients (γ(,i ) of aqueous electrolytes as a function of molality (m). The percentage in
parentheses refers to the distribution in all the electrolytes used in this study.

in a good correlation of the activity coefficient between on unsymmetric convention, while that of organic
measured and calculated values over a wide concentra- compounds is symmetric.
tion range of salt in aqueous solution. Recently, Jaretun Activity coefficients of inorganic compounds show a
and Aly21,22 applied additional empirical parameters to negative or positive deviation from one with an increase
the electrolyte-NRTL model for better representation of in salt molality. These behaviors are categorized into
aqueous activity coefficient of electrolytes. In particular, three different patterns, as shown in Figure 1. The
the local composition model plays an important role in activity coefficient of Type I significantly decreases with
representing activity coefficients of electrolytes over an increase in molality. In the case of Type II, the
wide concentration ranges. Similarly, using the UNI- activity coefficient approaches a constant value much
QUAC model23 as the local composition term, Lu and lower than one at moderate to high molalities. Contrary
Maurer,24 and Thomsen’s group25,26 developed the elec- to Type I and II, Type III significantly increases above
trolyte-UNIQUAC model and the extended UNIQUAC moderate molalities. The distribution of the electrolytes
model, respectively. Among these models including the in this study among the three types was 17%, 27%, and
local composition model, Chen’s electrolyte-NRTL model 56% for Types I, II, and III, respectively.
is the simplest and requires fewer parameters in the
calculation. Unfortunately, the model’s extension to Activity Coefficient Models for Aqueous
multielectrolyte solutions requires some additional pa- Electrolytes
rameters besides binary NRTL parameters between an
electrolyte and water. However, this disadvantage can In a dilute solution (less than 0.1 mol kg-1), it is well-
be solved by replacing the electrolyte-specific NRTL known that the activity coefficient of an electrolyte can
binary interaction parameters used in the model with be expressed by the Debye-Hückel equation:12
ion-specific parameters.
The objective of this study is to determine ion-specific A|zi ‚zi|xI
NRTL parameters via correlating the activity coefficient ln γ/i ) (3)
1 + BrxI
data of inorganic compounds containing heavy metals
with the electrolyte-NRTL model having these param- where z and r denote the ion charge and distance of the
eters. These results will then be used to predict the closest approach, respectively. A and B are the Debye-
activity coefficients and solubilities of inorganic salts Hückel constants calculated from the dielectric constant
in mixed electrolyte solutions, and also to predict the of the solvent and temperature. The ionic strength is
distribution ratio of Cu between xylene and an aqueous defined by
HCl solution.
N
1
Activity Coefficients of Inorganic Salts in Water I) ∑
2 i
z2mi (4)
In this study, we assumed that an aqueous electrolyte
solution consists of water, anions, and cations. This Up to an ionic strength of a few tenths molal, the Davies
complete dissociation assumption was extended to aque- equation27 is considered to be accurate. In the Davies
ous multielectrolyte systems. Since the activity coef- equation, an additional term is introduced into the
ficient of an individual ion i (γ/i ) cannot be measured, Debye-Hückel equation as follows:
the activity coefficients of aqueous ions are usually
expressed as the mean ionic activity coefficient of A|zi ‚zi|xI
neutral electrolytes a-c. The mean activity coefficient ln γ/i ) + 0.1|zi ‚zi|I (5)
/
(γ(,ac ) is given by
1 + BrxI
Neither equation is able to express the decreasing and
/
νa ln γ/a + νc ln γ/c increasing behavior of an activity coefficient over a
ln γ(,ac ) (1) moderate concentration range like those seen in the
νac
activity coefficient profiles of Types I and III. In addi-
tion, eqs 3 and 5 give similar values using the same
where νa is the stoichiometric coefficient of the anion, concentration and electric charge when only ionic
or the number of moles of the anion produced by the strength is taken into account. This indicates that these
dissociation of one mole of a-c; νc denotes that of the equations cannot distinguish the significant differences
cation, and νac is in the behavior of activity coefficients among ionic
species.
νac ) νa + νc (2) To describe all types of activity coefficients, local
composition models such as the NRTL, Wilson, and
The activity coefficient of an electrolyte or ion is based UNIQUAC models are added as a correction term to the
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3291
/ ∆gij or τij denotes the NRTL binary interaction param-
Debye-Hückel term (ln γi,D-H ) as follows:
eter of the i-j combination for any species, respectively.
/ /
ln γ*i ) ln γi,D-H + ln γi,local•composition (6) ∆gij is defined as the difference in the interaction energy
between the i-j and j-j combinations. The NRTL-
These models can predict the increase or decrease in derived activity coefficient (γi, NRTL) is based on sym-
the values of activity coefficients, depending on the metric convention. This convention is different from that
values of the binary interaction parameters between of the activity coefficient of aqueous salts. Therefore,
water and ion or ion and ion. The variance in this the activity coefficient must be transformed to
/
increase or decrease is proportional to electrolyte mo- the unsymmetric activity coefficient (γi,NRTL ) using the
lality. Among these models, the NRTL model is consid- NRTL-derived infinite dilution activity coefficient

ered to be the simplest and most convenient from a (γi,NRTL ) as follows:
calculation perspective because only the interaction

parameters are required for the activity coefficient ln γi,/ NRTL ) ln γi,NRTL - ln γi,NRTL (11)
calculation. In the other models, such as the UNIQAUC
model, size parameters of ionic or molecular species
According to the original NRTL equation,18 the following
except for the interaction parameters are required. In
six NRTL parameters (∆gij) are required for the water-
this study, therefore, the electrolyte-NRTL model pro-
single electrolyte system: ∆gwa for water-anion, ∆gaw
posed by Chen et al.19 was used for predicting the
for anion-water, ∆gwc for water-cation, ∆gcw for cation-
activity coefficients of electrolytes in single and mixed
water, ∆gac for anion-cation, and ∆gca for cation-anion.
salt aqueous systems.
In all the electrolyte-NRTL-type models, the use of the
Correlation Method local electric neutralization concept led to the reduction
Chen’s Electrolyte-NRTL Model. In the electrolyte- of the six parameters to the following two parameters:
NRTL model proposed by Chen et al.,19 the activity ∆gacw or τacw for electrolyte (ac)-water and ∆gwac or τwac
coefficient of an aqueous ion i is represented by the for water-electrolyte (ac). These NRTL parameters are
combination of the Pitzer-Debye-Hückel equation and referred to as the electrolyte-specific NRTL parameters,
NRTL equation as follows: and make the NRTL term much simpler. Unfortunately,
these parameters, which are determined from activity
ln γ/i ) ln γi,pdh
/ /
+ ln γi,NRTL (7) coefficient data of binary aqueous systems, are some-
times difficult to predict activity coefficients in aqueous
The Pitzer-Debye-Hückel equation is a function of multielectolyte solutions because the electrolyte-NRTL
ionic strength only, and is given by model is a correlative model for aqueous multielectrolyte
solutions and thus two additional NRTL parameters
/ between an electrolyte ac and another electrolyte a′c′
ln γi,pdh )

( ) {( ( )}
∆gaca′c′ and ∆ga′c′ac are required for a ternary solution of

) zi2Iz1/2 - 2Iz3/2
1/2
2zi2 multielectrolytes.
1000
- Aφ ln (1 + FIz1/2) + Modifications of Chen’s Electrolyte-NRTL Model.
Ms F 1 + FIz1/2 The disadvantage of Chen’s model is the electrolyte-
(8) specific parameters, which require some additional
parameters and mixing rules to predict phase equilib-
where Ms, Aφ, and F refer to the molecular weight of rium in water-multielectrolyte systems such as waste-
the solvent, the Pitzer-Debye-Hückel parameter, and water and wet wastes. The parameters and rules cannot
the closest approach parameter () 14.9), respectively. be determined from only binary activity coefficient data.
Aφ is calculated according to Chen’s paper.19 Ix is the The ion-specific parameter proposed by Liu et al.28 is
ionic strength based on the mole fraction x defined by considered to be a convenient and adaptable modifica-
tion for overcoming this disadvantage. The difference
1 N between electrolyte-specific and ion-specific parameters
Ix )
2
∑i z2xi (9) can be briefly explained as follows. For a water-
electrolyte binary system, in the former ∆gaw ) ∆gcw )
∆gacw, while in the latter ∆gaw * ∆gcw. In the present
The NRTL equation is generally expressed as study, two modifications were performed to introduce

( )
the ion-specific parameters to Chen’s electrolyte-NRTL
ln γi,NRTL ) model. First, the local electric neutralization was ex-
N N cluded. Although this exclusion results in a more
∑j zjGjiτji
N zjGij
∑k zkGkjτkj complicated calculation, it makes the model more
extensible to multielectrolyte solutions. Second, ∆gcc or
N
+ ∑j N τij -
N
(10) ∆gaa, for cation-cation or anion-anion interaction
∑k zkGki ∑k zkGkj ∑k zkGkj parameters in the multielectrolyte system, was assumed
to be zero on the basis of Lu’s assumption24 for interac-
tions between ions carrying a charge of the same sign.
In their assumption the cation-cation and anion-anion
where
interactions gcc and gaa are set to 16628 (2000 × R)
J‚mol-1 due to a large repulsive force. Hence, either ∆gcc
Gij ) exp(-Rτij)
or ∆gaa for any ion species becomes zero by the definition
∆gij gij - gjj of ∆gij () gii - gjj). This assumption is useful for
τij ) ) reducing the number of the NRTL parameters required
RT RT for activity coefficient calculation for multielectrolyte
R ) nonrandom factor () 0.2) systems. Consequently, six NRTL parameters (∆gwa,
3292 Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005

Table 1. Aqueous Binary Systems for Single Electrolytes determined, since these ions were the primary constitu-
Used to Determine the Ion-Specific NRTL Parametersa ents of the salts shown in Figure 2 and the nine salts
activity with these ions exhibited all types of activity coefficient
coefficient behavior. If the determined parameters were not able
behavior electrolytes to accurately represent the activity coefficient behavior,
type I CuSO4, Pb(NO3)2, (NH4)2SO4, NH4NO3, Na2SO4, the use of the ion-specific parameters would be discred-
K2SO4, KNO3, CdCl2 ited as a good approach for representing and predicting
type II NH4Cl, Al2(SO4)3, ZnSO4, MgSO4, KCl, NiSO4, the activity coefficients for all of the salts. The deter-
MnSO4, CdSO4, K2CrO4, Na2CrO4, Cr2(SO4)3,
H2SO4
mined NRTL parameters are given in Table 2. The
type III CuCl3, Cu(NO3)2, Pb(ClO3)2, Co(NO3)2, CoCl2, initial results for the nine binary systems accurately
AlCl3, Zn(NO3)2, Zn(ClO4)2, Ca(NO3)2, CaCl2, represented the experimental data, as shown in Figure
Ca(ClO4)2, HClO4, HCl, HNO3, KOH, NaOH, 3. On the basis of these reliable parameters, the other
Mg(ClO4)2, MgCl2, Mg(NO3)2, NaNO3, NaCl, NRTL parameters for the other 15 ionic species were
NiCl2, MnCl2, CrCl3, Cr(NO3)3 determined from the binary activity coefficient data for
a Data sources: Robison and Stokes,27 Mullin,29 and Majima and the 36 electrolytes. All the determined parameters are
Awakura30. summarized in Table 2. The two-step determination
successfully correlated the binary activity coefficient
data with the ion-specific NRTL parameters. To evalu-
ate the reliability of the obtained parameters, a com-
parison with the results obtained by Chen et al.19 was
carried out in respect to the standard deviation of the
logarithmic mean activity coefficient (σlnγ*(,i). The re-
sults are summarized in Table 3. The average σlnγ*(,i in
our calculations with 132 NRTL parameters (0.083) was
almost equal to that in Chen’s model with 90 param-
eters (0.079). This comparison demonstrated that this
study can provide reliable and reasonable ion-specific
NRTL parameter sets. Since the parameters can only
predict the activity coefficients in aqueous multielec-
trolyte solutions without any additional parameters,
this study suggests the first universal NRTL parameters
for aqueous electrolyte solutions. As further works, the
parameter values for other ions will be determined.
Future work should add the parameters of other ions,
Figure 2. Ion-specific NRTL parameter matrix for 66 pairs.
particularly AsO43- and PO43-, to the list for the
purposes of environmental protection and resource
∆gaw, ∆gwc, ∆gcw, ∆gac, and ∆gca) required for the recovery.
calculation of one binary system can be determined from
Prediction of Activity Coefficients in Multielec-
the experimental binary activity coefficient data of
trolyte Solutions. The activity coefficients for multiple
aqueous electrolytes.
components in aqueous solutions were predicted based
on the ion-specific NRTL parameters obtained above.
Results and Discussion Figure 4 shows the predicted and experimental mean
Correlation of Binary Activity Coefficients of activity coefficients30-32 of various solutes at different
Aqueous Electrolytes. The binary activity coefficient molalities of cosolutes in an aqueous solution. In terms
data for single electrolytes in water were collected from of the water-HCl-CuCl2 system, the predicted results
two books27,29 and the literature30 and summarized in were in good agreement with the experimental data.
Table 1. In total, 45 binary systems were available to Similarly, in the water-MgSO4-MnSO4 system and the
determine 66 pairs of the ion-specific NRTL parameters. water-H2SO4-(NH4)2SO4 system, the ion-specific NRTL
In Figure 2, the parameters to be determined are parameters predicted complicated behaviors of activity
categorized in a matrix, where dark, hatched, and gray coefficients in two-salt systems. The deviations between
areas indicate the newly determined parameters and calculated and experimental results are given in Table
parameters set to be zero, respectively. In the parameter 4. The accuracy of the predictions was around the
determination, a nonlinear mean square (Marquardt) deviations in the correlated results of binary systems
method was used as the optimization algorithm. To (Table 3). These results demonstrate that the ion-
determine a pair of the parameters between an ion and specific NRTL parameters are useful for predicting the
water, the activity coefficient data for all salts with the multicomponent activity coefficients of different cations
same ion are required. Hence, all the parameters were and anions in water, and thus enable us to design
simultaneously determined using all of the experimental equilibrium separation and recovery processes dealing
data to minimize the absolute deviation between ex- with aqueous mutlielectrolyte solutions. In addition,
perimental and calculated values. However, the use of these results indicate that the assumption concerning
such a large number of parameters led to the diversion the interaction parameters between ions with the same
of the objective function and a huge calculation time, sign does not cause large deviations in accuracy with
making it impossible to simultaneously correlate all of respect to the prediction of multielectrolyte systems.
the ion/water parameters. In this study, therefore, the However, in the water-H2SO4-CuSO4 system, the
parameter determination process was divided into two predicted curves did not quantitatively represent the
steps. First, the ion-specific parameters for 3 cations and experimental behavior, despite relatively accurate re-
anions (Na+, K+, Mg2+, Cl-, NO3-, and SO42-) were sults for the water-H2SO4-(NH4)2SO4 system. This
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3293

Table 2. Ion-Specific NRTL Parameters ∆gij/J‚mol-1


From H2O to CO2+
H2O K+ Na+ Mg2+ H+ Ca2+ Zn2+ Al3+ NH4+ Cd2+ Co2+
H2O 0 1100 7150 8170 8950 -15000 14500 -5530 -22500 18700 -13600
K+ 13900 0 0 0 0 0 0 0 0 0 0
Na+ -10700 0 0 0 0 0 0 0 0 0 0
Mg2+ -13600 0 0 0 0 0 0 0 0 0 0
H+ -9000 0 0 0 0 0 0 0 0 0 0
Ca2+ 1850 0 0 0 0 0 0 0 0 0 0
Zn2+ -13700 0 0 0 0 0 0 0 0 0 0
Al3+ -13400 0 0 0 0 0 0 0 0 0 0
NH4+ 2820 0 0 0 0 0 0 0 0 0 0
Cd2+ -11100 0 0 0 0 0 0 0 0 0 0
Co2+ -8960 0 0 0 0 0 0 0 0 0 0
Cu2+ -10900 0 0 0 0 0 0 0 0 0 0
Pb2+ -6090 0 0 0 0 0 0 0 0 0 0
Mn2+ -13500 0 0 0 0 0 0 0 0 0 0
Ni2+ -11200 0 0 0 0 0 0 0 0 0 0
Cr3+ -14900 0 0 0 0 0 0 0 0 0 0
Cl- -9280 -10600 -12200 -19100 -11300 414 0 -25600 -23300 -10300 -33800
SO42- -5100 14100 -14400 -18500 -15700 0 -12400 -12900 -7500 194 0
NO3- -13300 -16800 -25200 -13400 -11700 -31200 -6400 0 -41500 0 -39500
OH- -11700 -11200 -13700 0 0 0 0 0 0 0 0
ClO4- -11300 0 0 -20400 -15100 1620 -13900 0 0 0 0
CrO42- -6790 36800 -12600 0 0 0 0 0 0 0 0

From Cu2+ to CrO42-


Cu2+ Pb2+ Mn2+ Ni2+ Cr3+ Cl- SO42- NO3- OH- ClO4- CrO42-
H2O 19400 -92500 2820 11400 -56400 -13200 -7990 -1540 -17100 -16300 -12500
K+ 0 0 0 0 0 -12900 -18200 -20000 -1440 0 9220
Na+ 0 0 0 0 0 -31900 -25200 9960 -39900 0 -30300
Mg2+ 0 0 0 0 0 3360 5480 -31400 0 -510 0
H+ 0 0 0 0 0 1760 5860 -24200 0 3900 0
Ca2+ 0 0 0 0 0 2310 0 -25600 0 265 0
Zn2+ 0 0 0 0 0 0 -1220 -31700 0 -1310 0
Al3+ 0 0 0 0 0 64700 -18800 0 0 0 0
NH4+ 0 0 0 0 0 -27800 -20800 -19800 0 0 0
Cd2+ 0 0 0 0 0 -36800 5640 0 0 0 0
Co2+ 0 0 0 0 0 5830 0 -22300 0 0 0
Cu2+ 0 0 0 0 0 -33700 5220 -29800 0 0 0
Pb2+ 0 0 0 0 0 0 0 -30700 0 707 0
Mn2+ 0 0 0 0 0 -31900 7750 0 0 0 0
Ni2+ 0 0 0 0 0 49700 5640 0 0 0 0
Cr3+ 0 0 0 0 0 -34500 -21200 -33100 0 0 0
Cl- 32300 0 -18900 -11300 -66400 0 0 0 0 0 0
SO42- 801 0 -24700 -11000 -66600 0 0 0 0 0 0
NO3- 4330 -114000 0 0 -76300 0 0 0 0 0 0
OH- 0 0 0 0 0 0 0 0 0 0 0
ClO4- 0 -116000 0 0 0 0 0 0 0 0 0
CrO42- 0 0 0 0 0 0 0 0 0 0 0

indicates the obtained parameters are still not perfect where R is the gas constant. The constants A and B
for quantitative predictions of aqueous mutielectrolyte denote the thermal parameters of the pure salt, and are
solutions. For more accurate predictions of multlielec- independent of the concentration of the salt in the
trolyte systems, the ion-specific parameters will most aqueous solution. The relationship between the mole
likely need to be modified. fraction of the ion species and the molality of salt (m)
Solubility of Mixed Salts in Aqueous Solution. in solubility units is given by
A typical application of activity coefficient models is to
predict the solubility of a salt in an aqueous solution υ+m υ-m
containing other salts. The solubility is defined by the
x+ ) , x- )
1000 1000
solubility product Ks + m(υ+ + υ-) + m(υ+ + υ-)
Ms Ms
(14)
/ /
Ks ) aν+
+ a- ) (γ+ x+) + (γ- x-)
ν- ν ν-
(12)
The solubility data of mixed salts at the temperature
of interest are not easily obtained. Therefore, the
where a and x denote the activity and solubility in units thermal parameters A and B should be determined by
of mole fraction, respectively. The subscripts + and - correlating the experimental data29,34 with the electrolyte-
refer to cation and anion, respectively. In this study, NRTL model that includes the ion-specific parameters
the temperature dependence is assumed to be repre- prior to the solubility prediction in a multielectrolyte
sented by the van’t Hoff equation as follows: solution. If these constants are available, one can predict
solubilities in mixed salt systems at various tempera-
B tures. Thermal parameters determined in this way for
R ln Ks ) A + (13)
T five salts with major ions and the strength of their
3294 Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005

/
Figure 3. Correlation results of mean ionic activity coefficients (γ(,i ) for nine alkali and alkaline-earth metal salts.

/
Figure 4. Prediction results of mean ionic activity coefficients (γ(,i ) for four mixed electrolyte aqueous solutions at 298 K.

correlation with experimental values are given in Table is difficult for the electrolyte-NRTL model to predict the
5. Furthermore, the results for three heavy metallic solubility at higher concentrations of the unsaturated
salts in Table 5 are also given in Figure 5. Strong component if there is ion association, since molecular
correlations, with the exception of K2SO4, were ob- association is not considered in the NRTL model.
served. The ability to calculate the solubility of a salt Whereas, the concentration level of the unsaturated
in water at various temperatures allowed for the predic- component in the case of the heavy metallic salts was
tion of solubilities of different salts in an aqueous mixed much higher than that in the case of NaCl. As shown
salt solution at various concentrations. The predicted in Figure 6, the calculated values were consistently
results of the ternary solubilities are summarized in within 10% of the experimental values below an HCl
Table 6. The deviations for NaCl and the heavy metallic concentration of 6 mol kg-1. This indicates that a good
salts were relatively high, and increased with an prediction can be provided at relatively concentrated
increase in the concentration of the unsaturated com- solutions for the heavy metallic salts. As a whole, the
ponent. These deviations may be due to the fact that it present solubility predictions for the aqueous mixed salt
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3295

Table 3. Comparison of the Standard Deviation of the Table 5. Thermal Parameters for Equation 13 for Alkali
Logarithmic Mean Activity Coefficient Data of an and Alkaline-Earth Metal Salts, and Heavy Metal Salts
Electrolyte i in Water (σlnγ*(,i) between This Study and and the Relative Average Deviations
Chen’s Work19
relative average
electrolyte this study/- Chen’s work/- electrolytes A/J mol-1 K-1 B/J mol-1 deviationa/% ref
NaCl 0.018 0.024 alkali and alkaline metal salts
Na2SO4 0.043 0.024 K2SO4 114 -6460 10.6 29
NaNO3 0.126 0.002 NaCl 90.0 15900 2.72 29
KCl 0.003 0.003 KCl 41.4 -2610 2.18 29
K2SO4 0.017 0.008 Na2SO4 179 21800 0.81 29
KNO3 0.023 0.008 MgSO4 83.2 -5060 5.79 29
MgCl2 0.159 0.202 heavy metal salts
MgSO4 0.068 0.036 CoCl2 29.2 -17200 1.65 33
Mg(NO3)2 0.106 0.125 CuCl2 -132 22500 0.52 33
CuSO4 0.087 0.037 NiCl2 -41.5 9490 2.60 33
CuCl2 0.033 0.038
a ∑|(m
Cu(NO3)2 0.082 0.113 cal - mexp)/mexp|/N × 100 (m, solubility; N, number of data;
Pb(NO3)2 0.087 0.220 the subscript cal and exp denote calculation and experimental
Pb(ClO4)2 0.083 0.147 values, respectively).
Co(NO3)2 0.098 0.108
CoCl2 0.031 0.055
NH4Cl 0.003 0.001
(NH4)2SO4 0.052 0.017
NH4NO3 0.050 0.012
AlCl3 0.084 0.115
Al2(SO4)3 0.080 0.075
Zn(NO3)2 0.123 0.148
ZnSO4 0.061 0.038
Zn(ClO4)2 0.164 0.211
Ca(NO3)2 0.051 0.060
CaCl2 0.262 0.205
Ca(ClO4)2 0.337 0.272
HClO4 0.051 0.063
HCl 0.029 0.035
HNO3 0.045 0.008
KOH 0.062 0.023 Figure 5. Correlation results of the solubility (msol) of heavy
NaOH 0.034 0.029 metallic salts in water at various temperatures.
Mg(ClO4)2 0.159 0.208
NiCl2 0.053 0.092 Table 6. Comparison of the Predicted Results with the
NiSO4 0.097 0.031 Experimental Data for the Solubilities in Mixed
MnSO4 0.074 0.037 Electrolyte Solutions
MnCl2 0.036 0.047 relative
CdCl2 0.213 0.214 maximum of average
CdSO4 0.089 0.037 electrolyte maximum of total molality deviation of
K2CrO4 0.017 0.022 systems temp/ molality of of A1 and the solubility
(A1-A2) K A1/mol kg-1 A2/mol kg-1 of A2a/% ref
Na2CrO4 0.050 0.057
Cr2(SO4)3 0.178 0.129 K2SO4-MgO4 313 0.171 3.763 2.46 34
CrCl3 0.053 0.069 MgSO4-K2SO4 313 1.861 2.742 31.3 34
Cr(NO3)3 0.080 0.054 KCl-K2SO4 313 2.146 2.41 39.2 34
H2SO4 0.233 naa K2SO4-KCl 313 0.171 5.398 1.51 34
averageb 0.083 0.079 Na2SO4-K2SO4 313 1.718 2.27 161 34
K2SO4-Na2SO4 313 0.342 3.617 4.91 34
a na: not available. b denotes the average of σ
lnγ*(,i except for Na2SO4-MgSO4 313 1.374 4.717 2.86 34
H2SO4. MgSO4-Na2SO4 313 1.489 4.603 1.94 34
NaCl-KCl 313 2.146 6.278 11.7 34
Table 4. Standard Deviation of the Predicted KCl-NaCl 313 2.492 7.032 20.8 34
Logarithmic Mean Activity Coefficient Data (σlnγ*(,i) for NaCl-Na2SO4 313 2.491 4.626 9.29 34
Mixed Electrolyte Aqueous Solutions at 298 K Na2SO4-NaCl 313 1.374 6.305 10.7 34
HCl-CuCl2 298 7.802 10.849 22.6 33
molality molality σ lnγ/(,i HCl-NiCl2 298 8.342 9.415 10.9 33
electrolyte system of A1 of A2 of A2 HCl-CoCl2 298 8.428 9.077 24.2 33
(A1-A2) /mol kg-1 /mol kg-1 /- ref
a ∑|(m
cal - mexp)/mexp|/N × 100 (m, solubility; N, number of data;
CuCl2-HCl 0.1-4.0 0.4-2.4 0.055 30 the subscript cal and exp denote calculation and experimental
HCl-CuCl2 0.1-4.0 0.4-2.4 0.021 30 values, respectively).
(NH4)2SO4-H2SO4 0.240-4.188 0.5-2.0 0.118 31
H2SO4-(NH4)2SO4 0.252-2.889 0.5-2.0 0.068 31
MgSO4-MnSO4 0.1-2.8 0.1-2.4 0.057 32 the predicted values in the Na2SO4 system were opposite
MnSO4-MgSO4 0.1-2.4 0.1-2.8 0.051 32 of the experimental values. Consideration regarding the
CuSO4-H2SO4 0-1.2 0.4-1.6 0.184 30 hydration effect, similar to Lu’s model,24 may result in
H2SO4-CuSO4 0.1-1.6 0.4-0.8 0.393 30
a more successful prediction.
systems are considered to be successful because the ion- Distribution Ratio of Cu in a Solvent Extraction.
specific NRTL parameters are universal regarding all Another application of the model developed in this study
salts and all systems. Most studies reporting good is to predict the distribution equilibrium of a metallic
solubility predictions use binary parameters that lack ion between aqueous and organic phases in a solvent
this universality. Although good results were obtained extraction process. On the basis of the fact that the ion-
for the most part, the results for K2SO4 were difficult specific parameters accurately predicted ternary activity
and problematic, as shown in Figure 7. In particular, coefficients and solubilities in the water-HCl-CuCl2
3296 Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005

h Cu C
C h CuR2
D) ) (16)
CCu C 2+
Cu

where CCu and C h Cu are the concentration of Cu2+ in the


aqueous phase (CCu2+) and the concentration of the
copper-extractant complex (CuR2) in the organic phase
h CuR2), respectively. If the activity coefficients of all
(C
components in the organic phase can be approximated
to unity, the distribution equilibrium constant Ke is
simplified as follows:

h CuR2 × (γH+ × CH+)2


C
Figure 6. Predicted solubility (msol) of heavy metallic electrolytes
in hydrogen chloride solution for the ternary systems at 298 K.
Ke ) (17)
h HR2
(γCu2+ × CCu2+) × C

Consequently, substituting eq 16 into eq 17, the distri-


bution was calculated using

CCuR2 γCu2+
log ) log + 2log CHR + log Ke
CCu2+ (γH+ × CH+)2
(18)
The literature value for Ke determined by Tanaka35 was
used in this calculation. By substituting the electrolyte-
NRTL model for activity coefficients in eq 18, the
distribution ratios of Cu were accurately predicted. The
Figure 7. Predicted solubility (mK2SO4) of K2SO4 in mixed elec- predicted ratios are plotted as a function of pH in Figure
trolyte solutions at 313 K. Added electrolyte (A1): MgSO4, Na2- 8. As shown in Figure 8, the predicted distribution ratios
SO4, and KCl.
decreased with a decrease in pH in good quantitative
system, a solvent extraction system consisting of HCl agreement with the experimental data.35 Comparing
and CuCl2 was selected. The complete system used was these results with those of the Davies equation in Figure
the water-xylene(-β-hydroxyoxime)-HCl-CuCl2 two- 8 indicates that the present electrolyte-NRTL model is
phase system,35 and the distribution ratio of Cu between more useful at relatively high concentrations of HCl.
the two phases was predicted. The extraction equilib- This level of accuracy is attributed to the reasonable
rium of Cu with β-hydroyoxime as the extractant (HR) values of the activity coefficients in aqueous multielec-
is described by the following equation: trolyte solutions achieved using this model.

Conclusion
Cu2+ + 2HR ) CuR2 + 2H+ (15)
The introduction of ion-specific NRTL parameters into
where the bar above the compounds denotes the xylene the conventional electrolyte NRTL model proposed by
phase. The distribution ratio D for Cu between both Chen et al. led to the accurate prediction of activity
phases is defined by coefficients in aqueous multielectrolyte solutions with-

Figure 8. Measured and predicted distribution ratio (D) of Cu in solvent extractions (HCl-CuCl2-water-xylene-β-hydroxyoxime systems)
as a function of pH in aqueous phase. Symbols, measured values; solid lines, predicted values. (a) Prediction by the electrolyte-NRTL
model with ion specific parameters. (b) Prediction by the Davies equation.
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3297

out any additional parameters. This electrolyte-NRTL (14) Pitzer, K. S.; Mayorga, G. J. Phys. Chem. 1973, 77, 2300-
model was also used to predict the solubilities of 2308.
inorganic salts including heavy metals in multielectro- (15) Pitzer, K. S.; Kim, J. J. J. Am. Chem. Soc. 1974, 96, 5701-
5707.
lyte solutions and the distribution ratio of copper (16) Pitzer, K. S. J. Am. Chem. Soc. 1980, 102, 2902-2906.
between aqueous and organic phases with a high level (17) Cruz, J.-L.; Renon, H. AIChE J. 1978, 24, 817-830.
of accuracy. These predictions demonstrated that the (18) Renon, H.; Prausinitz, J. M. AIChE J. 1968, 14, 135-144.
obtained parameters have the potential to be very useful (19) Chen, C.-C.; Britt, H. I.; Boston, J. F.; Evans, L. B. AIChE
in the design of processes involving the removal and J. 1982, 28, 588-596.
recovery of hazardous heavy metallic salts and other (20) Chen, C.-C.; Evans, L. B. AIChE J. 1986, 32, 444-454.
industrial inorganic compounds. This model can also (21) Jaretun A.; Aly, G. Fluid Phase Equilib. 1999, 163, 175-
193.
significantly contribute not only to the solution of many (22) Jaretun A.; Aly, G. Fluid Phase Equilib. 2000, 175, 213-
environmental and resource problems, but also to the 228.
development of recycling technologies. The parameters (23) Abrams, D. S.; Prausinitz, J. M. AIChE J. 1975, 21, 116-
for other important ions such as PO43- should be 128.
determined in future work. (24) Lu, X.-H.; Maurer, G. AIChE J. 1993, 39, 1527-1538.
(25) Thomsen, K.; Rasmussen, P.; Gani, R. Chem. Eng. Sci.
1996, 51, 3675-3683.
Literature Cited (26) Iliuta, M. C.; Thomsen, K.; Rasmussen, P. AIChE J. 2002,
(1) Baltpurvins, K. A.; Burns, R. C.; Lawrance, G. A. Environ. 48, 2664-2689.
Sci. Technol. 1996, 30, 1493-1499. (27) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions, revised
(2) Tanaka, M.; Koyama K.; Shibata, J. Ind. Eng. Chem. Res. ed; Butterworth: London, 1965.
1998, 37, 1943-1949. (28) Liu, Y.; Harvey, A. H.; Prausnitz, J. M.; Chem. Eng. Comm.
(3) Esalah, J. O.; Weber, M. E.; Vera, J. H. Can. J. Chem. Eng. 1989, 77, 43-66.
2000, 78, 948-954. (29) Mullin, J. M. Crystallization, 3rd ed.; Butterworth-Heine-
(4) Shen, S. B.; Tyagi, R. R. D.; Blais, J. F. Environ. Technol. mann: Oxford, 1993.
2001, 22, 961-970. (30) Awakura, Y.; Majima, K. Metall. Mater. Trans. B 1988,
(5) Regel, M.; Sastre, A. M.; Szymanowski, J. Environ. Sci. 19B, 347-354.
Technol. 2001, 35, 630-635. (31) Park, S.-K.; Awakura, Y.; Majima, K. Metall. Mater. Trans
(6) Gupta, B.; Deep, A.; Tandon, S. N. Ind. Eng. Chem. Res. B 1989, 20B, 13-20.
2002, 41, 2948-2952. (32) El Guendouzi, M.; Mounir, A.; Dinane, A. Fluid Phase
(7) Juang, R. S.; Wang, Y. C. Ind. Eng. Chem. Res. 2002, 41, Equilib. 2002, 202, 221-231.
5558-5564. (33) Linke, W. F. Solubilities: Inorganic and Metal-Organic
(8) Vaessen, R.; Seckler, M.; Witkamp, G. P. Ind. Eng. Chem. Compounds; A Compilation of Solubility Data form the Periodical
Res. 2003, 42, 4874-4880. Literature, 4th ed.; American Chemical Society: Washington, DC,
(9) Loehe, J. R.; Donohue, M. C. AIChE J. 1997, 43, 180-195. 1965.
(10) Prausnitz, J. M.; Lichtenthaler, R. N.; de Azevedo, E. G. (34) Maeda, K.; Kuramochi, H.; Shinkawa, T.; Fukui, K. J.
Molecular Thermodynamics of Fluid Phase Equilibria, 2nd ed.; Chem. Eng. Data 2002, 47, 1472-1475.
Prentice Hall: Englewood Cliffs, NJ, 1986. (35) Tanaka, M. Hydrometallurgy 1990, 24, 317-331.
(11) Fredenslund, A.; Jones, R. L.; Prausnitz, J. M. AIChE J.
1975, 21, 1086-1099. Received for review July 16, 2004
(12) Lewis, G. N.; Randall, M. Thermodynamics, 2nd ed. Revised manuscript received January 19, 2005
(revised by K. S. Pitzer and L. Brewer); McGraw-Hill: New York, Accepted February 9, 2005
1961.
(13) Pizter, K. S. J. Phys. Chem. 1973, 77, 268-277. IE049377U

You might also like