Fluid Phase Equilibria 270 (2008) 87–96

Contents lists available at ScienceDirect

Fluid Phase Equilibria

journal homepage: www.elsevier.com/locate/fluid

Modeling aqueous electrolyte solutions

Part 1. Fully dissociated electrolytes
Christoph Held, Luca F. Cameretti, Gabriele Sadowski ∗
Laboratory of Thermodynamics, Faculty of Biochemical and Chemical Engineering, Dortmund University of Technology,
Emil-Figge-Str. 70, 44227 Dortmund, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Liquid densities (pvT), vapor pressures (VLE), and mean ionic activity coefficients (MIAC) at 25 ◦ C of 115
Received 12 March 2008 single-salt electrolyte solutions containing univalent up to trivalent ions are modeled with the ePC-SAFT
Received in revised form 17 June 2008 equation of state proposed by Cameretti et al. [L.F. Cameretti, G. Sadowski, J.M. Mollerup, Ind. Eng. Chem.
Accepted 19 June 2008
Res. 44 (2005) 3355–3362; ibid., 8944]. For each ion, only two model parameters were adjusted to exper-
Available online 28 June 2008
imental density and MIAC data. Without using any additional binary parameters, ePC-SAFT is able to
reproduce experimental data of the respective salt solutions up to high electrolyte molalities. Moreover,
Keywords:
it is even able to describe the reversed MIAC series for alkali hydroxides and fluorides.
Thermodynamic properties
Density © 2008 Elsevier B.V. All rights reserved.
Vapor–liquid equilibria
Mean ionic activity coefficient
Equation of state
PC-SAFT
Debye–Hückel
Aqueous solutions
Binary mixtures
Primitive model

1. Introduction the non-ionic short-range interactions resulting from repulsive as

well as attractive forces, also the long-range Coulombic interactions
Electrolyte solutions play an important role in biological and of the charged species need to be accounted for. The first ones can be
chemical engineering. Applications of these systems can be found described by using either gE models or equations of state (EOS). The
in waste and drinking water treatment, fertilizer production, or Coulombic interactions can be modeled, e.g. by the Debye–Hückel
enhanced oil recovery [2]. Other important technical processes theory (DH) developed in 1923 [4] or by the mean spherical approx-
based on thermodynamic properties of electrolytes are electrolysis imation (MSA) introduced by Waismann and Lebowitz [5] in 1970
[3], wet flue-gas scrubbing, osmosis and reverse osmosis of aque- who solved the Ornstein–Zernike equation for a fluid of charged
ous solutions, as well as reactive distillation with an electrolyte spheres of equal size.
serving as entrainer. Furthermore, electrolytes are used to stabilize Examples for electrolyte gE models are the electrolyte NRTL
biomolecules or to salt them out. model [6] or the Pitzer model [7]. Nasirzadeh et al. [8] used a MSA-
However, the first step towards modeling of complex multi- NRTL model [9] as well as an extended Pitzer model of Archer [10] to
component biological solutions is the ability to accurately describe describe osmotic coefficients of lithium hydroxide solutions. Both,
quasi-binary aqueous electrolyte systems. For that purpose, besides MSA-NRTL and the Pitzer–Archer model turned out to be excel-
lent models for the description of activity coefficients in electrolyte
solutions. However, nine adjustable parameters are needed for the
Abbreviations: AnCat, electrolyte of the form anion-cation; AAD, absolute aver- Pitzer–Archer model, five of them being temperature dependent.
age deviation, defined in Eq. (11); ARD, absolute average relative deviation, defined Six parameters have to be adjusted using the MSA-NRTL, one of
in Eq. (11); DH, Debye–Hückel model; EOS, equation of state; gE , excess Gibbs them even being a function of concentration.
energy; MIAC, mean ionic activity coefficient; MSA, mean spherical approximation;
The second group of models is represented by equations of
OF, objective function; pvT, pressure–volume–temperature behavior (density data);
VLE, vapor–liquid equilibrium.
state. Myers et al. [11] developed an electrolyte model based on
∗ Corresponding author. Tel.: +49 231 7552635; fax: +49 231 7552572. the Peng–Robinson EOS (PREOS). They used a Born-energy term
E-mail address: [email protected] (G. Sadowski). for charging up the uncharged reference system in a continuous

0378-3812/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.fluid.2008.06.010
88 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96

medium as well as a restricted version of the MSA. The resulting can be written as
EOS requires three adjustable salt parameters. Osmotic coefficients
A res
and salt activity coefficients of 138 aqueous single-salt solutions = ares = ahc + adisp + aassoc + aion (1)
N
were modeled with an overall average relative deviation (ARD) of
1.95%. However, the PREOS is known to fail when predicting liquid where N is the total number of molecules. ahc represents the
densities [1,11]. Fürst and Renon [12] proposed the combination of a hard-chain repulsion of the reference system. adisp , aassoc , and aion
modified Redlich–Kwong–Soave EOS with the MSA accounting for account for the Helmholtz-energy contributions due to disper-
the ionic part. Using six correlation parameters, the osmotic coef- sive, associative, and Coulomb interactions, respectively. Whereas
ficients of 28 halide systems and 35 non-halide systems could be expressions for adisp and aassoc are used as in the original PC-
calculated with a root mean square relative deviation of 2.9% and SAFT model [20], Cameretti et al. [1] used a Debye–Hückel term
2.2%, respectively. to describe the Helmholtz energy contribution aion to a system that
Since its development in 1989, the Statistical Associating Fluid is caused by charging the species j:
Theory (SAFT) has been applied to many different systems, includ-
aion  
ing electrolyte solutions. Paricaud et al. [13] gave an overview of =− × xj q2j j (2)
the developments using the SAFT in order to examine the effect kB T 12kB Tε
j
of added salt on the vapor–liquid equilibria (vapor pressure and
density) of aqueous solutions. Liu and co-workers [14] combined Here, xj and qj are the mole fraction and the charge of ion j, respec-
SAFT with the MSA primitive model. They could describe mean tively. kB represents the Boltzmann constant and T the temperature.
ionic activity coefficients (MIAC) and solution densities (pvT) for The ions are treated as spherical species in a uniform dielectric con-
30 electrolytes with an overall ARD of 1.6%. Galindo et al. [15] tinuum characterized by a dielectric constant ε. They can approach
successfully extended the SAFT-VR EOS to electrolyte solutions by each other to the distance aj which is equivalent to the ion diameter
using an additive Coulombic (MSA as well as DH) contribution.  j . In contrast to some other groups [21,22], we use a concentration-
Their model yields good results for vapor pressures and liquid den- independent dielectric constant ε for water. The quantity j in Eq.
sities of aqueous solutions of univalent ions. However, they only (2) is defined as
marginally considered the MIAC. Radosz et al. recently published
3
3 1

their SAFT1 [16,17] and SAFT2 [18,19] EOS yielding excellent results j = × + ln(1 + aj ) − 2(1 + aj ) + (1 + aj )2 (3)
concerning the properties of aqueous single-salt and multi-salt (aj )3 2 2
solutions. SAFT1 and SAFT2 are hybrid models that treat a salt as one
with  being the inverse Debye screening length given by
molecule consisting of two segments, the cation and the anion. Con-
 
sequently, three individual-ion parameters as well as one additional
NA  N e2  2
salt parameter have to be adjusted for each electrolyte solution. = × q2j cj = × zj xj (4)
kB Tε kB Tε
SAFT1 was used to describe six aqueous alkali halide solutions. It j j
was modified in SAFT2 by a new dispersion term which was applied
to single-salt aqueous systems. They could be modeled with abso- cj is the molar concentration, N the number density of the system
lute ARDs of 0.63% for the MIAC and 0.45% for the liquid densities, and NA is Avogadro’s constant, respectively.
respectively. The repulsive interactions of the ions and the attractive inter-
In this work we use the ePC-SAFT model developed by Cameretti action with water (hydration) are accounted for in ahc and adisp ,
et al. [1]. It is a combination of the PC-SAFT EOS by Gross and Sad- respectively.
owski [20] and the Debye–Hückel contribution [4] which accounts Using ePC-SAFT, three pure-component parameters are used to
for the Coulomb interactions. It considers the ionic species inde- describe the molecular properties of a molecule: the segment num-
pendent of the salt they are part of. Only two parameters are used ber mseg , the segment diameter , and the dispersion energy u/kB .
to characterize each ion. The first one is the ionic diameter  j which For associating components like water two additional parameters
is actually the diameter of the hydrated ion. As the DH contribution are required, namely the association energy εA1B1
hb
/kB and the asso-
only accounts for Coulombic forces among the ions, we describe the ciation volume hbA1B1 . The association scheme used here for water

interactions between ion and water (hydration) by means of disper- is the two-site 2B approach [23].
sion. This yields to a second ionic parameter: the dispersive-energy The segment number of ions is always set to one (mseg,j = 1)
parameter uj /kB , which reflects the strength of ionic hydration. yielding finally to two parameters for each ion: the diameter  j
The parameters in the previous work [1] were obtained by simul- and the dispersion energy uj /kB of the hydrated ion. Since uj /kB was
taneous fitting to density data and vapor pressures (VLE) of salt determined from aqueous electrolyte solution data, it gives a direct
solutions. Calculations for 12 salts could be performed with an hint to which extent the ion interacts with water.
overall ARD of 0.7% (pvT) and 2.4% (VLE), respectively. However, To describe salt solutions, the conventional Berthelot–Lorenz
using this parameter set, the MIAC can only be described in poor combining rules are used:
agreement with experimental data. Therefore, in this work we
1
apply a new approach for parameter estimation and present a ij = ( + j ) (5)
new parameter set that is able to reasonably describe the solu- 2 i

tion densities as well as the VLE and MIAC of about 115 electrolyte uij = ui uj (6)
systems.
Eq. (6) is applied for interactions between water and ions
only. Dispersion between two ions is neglected in this work.
2. ePC-SAFT equation of state Often, a binary interaction parameter kij is introduced for the
dispersive-energy parameter to account for deviations from the
The ePC-SAFT EOS is based on a perturbation theory where geometric-mean rule in the mixture. This parameter is usually fit-
the hard-chain system is used as the reference system. All other ted to the respective binary mixture’s properties. However, since
contributions are considered as additive contributions that can be the parameter uj /kB had already been fitted to salt/water systems,
considered independently. Thus, the residual Helmholtz energy ares a kij is not applied here.
 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96 89

Table 1
PC-SAFT parameters for water used in this work

Parameter Unit Abbreviation Value

Segment number – mseg,1 1.204659

Segment diameter Å 1 2.792700

Å Tdep,1 10.1100
1/K Tdep,2 −0.01775
Å Tdep,3 −1.41700
1/K Tdep,4 −0.01146

Dispersion energy K u1 /kB 353.9449

Association sites – N1 2
Association energy K εA1B1
hb
/kB 2425.6714
Association volume – hb
A1B1
0.0450989

3. Parameter estimation strong electrolytes that fully dissociate into their respective cations
and anions. Dispersive interactions were only considered between
The first step in calculating properties of aqueous electrolyte water–water and water–ion pairs. Two ions were assumed to inter-
solutions is to model the solvent water as accurately as possible. act only by repulsive (ahc ) and Coulomb forces (aion ).
The parameters of water described by the 2B association model [23] The two ion parameters – diameter  j and dispersion energy
are resumed from Cameretti et al. [1] and are given in Table 1. To uj /kB – were determined from fitting them to experimental data
account for the density anomaly of water, this parameter set was of aqueous salt solutions. In the previous work [1] saturated vapor
readjusted between 0 and 100 ◦ C by introducing a temperature- pressures and liquid densities were used for that purpose. However,
dependent segment diameter  T,W for water. The latter quantity is applying the so-determined parameters leads to poor results when
related to the temperature-independent one ( W ) by calculating MIAC values. Therefore, in this work in addition to den-
sity data (between 20 and 30 ◦ C) also MIAC at 25 ◦ C were used for
T,W = W + Tdep,1 × exp{Tdep,2 T } + Tdep,3 × exp{Tdep,4 T } (7)
the parameter estimation, which are much more sensitive to the
The respective coefficients are given in Table 1. Using these ionic parameters than VLE data.
parameters, the absolute average mean deviation for a temperature The MIAC ±∗,x of an electrolyte is defined as the geometrical
range between 0 and 100 ◦ C is 0.06% for the densities and 0.52% mean of the mole-fraction-based rational activity coefficients of
for the vapor pressures of water, respectively. Note that without the ions in solution [24]:
introducing the temperature-dependence of  w in Eq. (7) it is not + − 1/( + + − )
±∗,x = ((+∗,x ) (−∗,x ) ) (8)
possible to accurately model the water density between 0 and 25 ◦ C,
respectively (compare Ref. [1]). Here, + and − are the stoichiometric coefficients of cation and
For the calculation of solution densities, VLE, and MIAC of elec- anion in the salt [16] which add to . The rational activity coeffi-
trolyte solutions, several assumptions have to be made. First, a cients j∗,x of ions can be obtained by ePC-SAFT as the ion fugacity
reasonable approximation within the temperature range of this coefficient ϕj at the actual concentration related to the one at infi-
work is that the vapor phase above the solution consists of pure nite diluted state ϕj∞ :
water only. Secondly, the considered electrolytes were regarded as
ϕj (T, p, xj )
j∗,x = (9)
Table 2 ϕj∞ (T, p, xj → 0)
ePC-SAFT parameters used in this work; only valid with parameter set of water in
Table 1 ±∗ can be determined directly by potentiometric methods or
indirectly by isopiestic measurements and is available for many
Univalent cations Univalent anions
electrolytes in aqueous solutions at 25 ◦ C. In the literature [25] MIAC
Ion  j (Å) uj /kB (K) Ion  j (Å) uj /kB (K) are often given on a molal basis, whereas ePC-SAFT is mole-fraction
H+
2.2740 1616.4939 F−
1.6132 648.3127 based. Thus, the following conversion from mole fraction (x)-based
Li+ 1.8177 2697.2795 Cl− 3.0575 47.2878 to molality (m)-based values is applied:
Na+ 2.4122 646.0504 Br− 3.4573 60.2216
K+ 2.9698 271.0518 I− 3.9319 80.4347 ±∗,x
Rb+ 3.1443 215.3697 OH− 1.6401 2444.7555 ±∗,m =
(1 + 0.001 MW m± )
Cs+ SCN−
NH4 +
3.5606
3.4755
175.9357
212.3632 ClO4 −
4.0715
4.0731
69.6806
58.423
 
∗,m
Choline+ 5.9216 220.4883 ClO3 − 3.8212 15.4978 1 ϕ±
= × ∗,m (10)
H2 PO4 − 3.7026 0.0000 (1 + 0.001 MW m± ) ϕ±,x
BrO3 − 3.5765 0.0000 solute →0

NO3 − 3.3805 0.0000

MW and m± are the molecular weight of water and the molality
Bi-/trivalent cations Bivalent anions of the electrolyte in moles of salt per kg of water, respectively.
Ion  j (Å) uj /kB (K) Ion  j (Å) uj /kB (K)

Mg2+ 2.3229 8145.3298 SO4 2− 2.4538 0.0000 Table 3

Ca2+ 2.8889 2146.9794 HPO4 2− 4.4608 0.0000 Comparison of hydrated cation sizes: experimental values from X-ray and neutron
Sr2+ 2.9882 1677.6061 diffraction measurements [26] versus ePC-SAFT parameters
Ba2+ 3.0982 1475.9880
Ion  exp (Å)  ePC-SAFT (Å)
Co2+ 2.4387 5837.6334
Cu2+ 2.6955 2545.1445 Li+ 1.86 1.81
Fe2+ 2.4828 5495.6986 Na+ 2.40 2.41
Cr3+ 3.2380 6905.3450 K+ 3.10 2.97
90 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96

Table 4
Number of data points (NP), maximum molality, and deviations of ePC-SAFT from experimental density, vapor pressure, and MIAC data

Electrolyte Density Vapor pressure Activity coefficient

N mmax ARD (%) AAD N mmax ARD (%) AAD (kPa) Reference N mmax ARD (%) AAD
(mol/kg) (kg/m3 ) (mol/kg) (mol/kg)

Fluorides
LiF 15 0.04 0.01 0.06 – – – –
NaF 18 0.93 0.24 2.44 – – – – 17 1 2.38 0.02
KF 18 8.59 0.43 4.83 – – – – 17 5 3.26 0.02
RbF 9 6.38 0.30 3.89 – – – – 24 3.5 6.70 0.05
CsF 10 5.39 0.21 2.60 – – – – 10 3.5 6.39 0.05

Chlorides
HCl 15 6.86 0.16 1.70 – – – – 17 5 9.80 0.14
LiCl 18 15.73 1.83 21.46 13 12.69 5.87 0.76 [31,38] 20 4.5 9.79 0.1
NaCl 18 5.82 0.74 8.31 6 6.22 0.77 0.06 [39] 16 3.2 3.43 0.02
KCl 17 3.93 0.49 5.45 8 4.29 1.13 0.04 [30,40] 20 4.5 2.38 0.01
RbCl 33 8.27 0.29 3.73 8 6.95 1.55 0.07 [30,39] 32 7.8 1.35 0.01
CsCl 18 3.96 0.22 2.79 40 8.59 3.52 0.43 [30] 18 5 1.91 0.01
NH4 Cl 9 6.23 0.97 10.16 14 5.32 0.62 0.01 [38] 22 7 0.95 0.01
ChCl 7 0.24 0.48 4.90 – – – – 23 6 15.48 0.08
MgCl2 18 5.66 1.69 20.77 40 4.8 1.9 0.20 [30] 20 4.5 11.08 0.19
CaCl2 19 7.37 2.79 35.38 40 7.89 12.52 1.02 [30] 19 5.5 26.11 0.8
SrCl2 24 3.4 2.20 27.46 40 3.2 0.95 0.10 [30,41] 38 3.5 11.05 0.08
BaCl2 18 1.6 0.99 11.58 25 1.39 0.74 0.12 [30,42] 36 1.79 5.68 0.03
FeCl2 – – – – – – – – 32 2 7.13 0.04
CuCl2 10 1.86 1.98 22.49 25 3.8 2.52 0.04 [43] 20 2.8 19.56 0.09
CoCl2 – – – – – – – – 40 4 12.10 0.13
CrCl3 13 2.48 1.42 16.24 – – – – 11 1.2 13.60 0.05

Bromides
HBr 20 10.11 0.38 5.01 – – – – 20 5 9.28 0.19
LiBr 18 7.68 0.86 10.83 45 15.97 30.27 0.92 [31] 20 4.5 3.52 0.04
NaBr 18 6.48 1.03 12.98 14 7.98 1.38 0.10 [30,38] 20 4.5 1.75 0.01
KBr 18 5.6 0.56 6.96 35 4.35 0.68 0.09 [30,39,44] 20 4.5 1.78 0.01
RbBr 36 7.39 0.31 4.38 – – – – 27 5 1.89 0.01
CsBr 18 3.13 0.17 2.16 30 5.89 2.41 0.60 [30] 18 5 2.56 0.01
NH4 Br 10 6.81 0.98 11.72 – – – – 19 7.99 6.56 0.04
ChBr – – – – – – – – 24 7 17.39 0.07
MgBr2 20 4.44 0.89 11.6 – – – – 20 4.5 13.43 0.35
CaBr2 22 5 2.06 28.62 40 4.6 6.25 0.84 [30,45] 22 3.5 18.31 0.24
SrBr2 21 1.73 1.09 13.14 40 3.34 1.69 0.20 [30] 40 2.12 11.91 0.08
BaBr2 27 2.24 0.70 9.03 40 3.4 1.33 0.19 [30] 42 2.32 6.47 0.04
CuBr2 – – – – – – – – 39 3.61 14.32 0.11
CoBr2 – – – – – – – – 44 5 8.55 0.18

Iodides
HI 20 6.4 0.12 1.43 – – – – 19 3.5 2.56 0.03
LiI 18 4.98 0.85 10.77 21 10.13 6.06 0.53 [31] 17 3 4.49 0.05
NaI 18 4.45 0.52 6.55 36 8.4 2.9 0.37 [30] 19 4.5 1.37 0.01
KI 18 4.02 0.16 1.94 27 5.65 1.26 0.09 [30,41] 20 4.5 1.44 0.01
RbI 42 8.74 0.18 2.92 – – – – 27 5 3.30 0.02
CsI 18 2.57 0.06 0.81 25 2.6 0.46 0.06 [30] 18 3 4.66 0.02
NH4 I 9 3.72 0.38 4.74 20 13.89 2.12 0.12 [41] 17 1.1 2.36 0.02
MgI2 18 2.4 0.76 9.96 – – – – 20 4.5 47.18 3.29
CaI2 18 2.27 1.36 17.73 25 2.92 3.66 0.57 [30] 15 2 10.15 0.09
SrI2 30 2.4 1.30 17.64 40 4.16 3.21 0.33 [30] 37 1.9 11.26 0.08
BaI2 39 3.84 0.88 14 – – – – 38 2 9.46 0.07
CoI2 – – – – – – – – 44 5 10.62 0.76

Hydroxides
LiOH 15 3.79 0.20 2.17 115 4.77 0.95 0.34 [8] 19 5 3.01 0.02
NaOH 15 10.88 0.44 5.07 18 12.19 3.54 0.58 [46] 20 5.5 6.78 0.05
KOH 15 15.51 0.7 9.44 – – – – 14 3 3.03 0.02
RbOH 26 3.25 0.36 4.53 – – – – – – – –
CsOH 12 1.18 0.21 2.30 – – – – 12 1.2 2.72 0.02
BaOH2 6 0.31 0.38 3.98 – – – – 6 0.2 2.26 0.01

Nitrates
HNO3 16 8.55 0.50 5.75 – – – – 17 5 6.87 0.06
LiNO3 20 3.63 0.75 8.03 16 12.86 1.56 0.03 [40] 22 5.5 14.25 0.14
NaNO3 20 2.94 0.33 3.54 20 13.35 2.71 0.10 [38,39,45] 24 7 1.99 0.01
KNO3 20 2.47 0.19 1.99 18 8.31 2.3 0.14 [41] 18 3.5 2.28 0.01
NH4 NO3 22 12.49 1.85 21.74 – – – – 22 9 29.15 0.1
RbNO3 19 3.65 0.20 2.53 – – – – 26 4.5 2.92 0.01
CsNO3 15 1.28 0.03 0.27 – – – – 14 1.5 3.68 0.02
Mg(NO3 )2 18 2.25 0.62 7.04 14 5.13 2.36 0.02 [41,42] 20 4.5 13.14 0.09
Ca(NO3 )2 22 6.09 1.38 17.72 12 10.42 2.60 0.05 [42,45] 20 4.5 13.56 0.05
Sr(NO3 )2 21 1.58 0.91 10.97 – – – – 19 4 11.95 0.03
 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96 91

Table 4 (Continued )

Electrolyte Density Vapor pressure Activity coefficient

N mmax ARD (%) AAD N mmax ARD (%) AAD (kPa) Reference N mmax ARD (%) AAD
(mol/kg) (kg/m3 ) (mol/kg) (mol/kg)

Ba(NO3 )2 5 0.2 0.23 2.43 – – – – 10 0.4 3.55 0.01

Fe(NO3 )2 6 3 5.12 66.74 – – – – – – – –
Cu(NO3 )2 15 4.19 1.32 16.39 – – – – 44 5 32.96 0.32
Co(NO3 )2 12 2.34 0.93 10.87 – – – – 44 5 15.92 0.19
Cr(NO3 )3 11 1.8 1.52 17.5 – – – – 12 1.4 15.48 0.05

Thiocyanates
NaSCN 10 9.63 0.10 1.11 – – – – 14 5 2.76 0.02
KSCN 12 3.3 0.21 2.34 – – – – 14 5 1.07 0.01
NH4 SCN – – – – – – – – 12 2 2.48 0.02

Chlorates
HClO3 6 3.74 0.03 0.30 – – – – – – – –
LiClO3 18 33.19 0.83 11.85 – – – – 14 10.27 11.60 0.17
NaClO3 7 4.03 0.31 3.51 – – – – 23 3 2.14 0.01
KClO3 7 0.43 0.06 0.59 – – – – 7 0.7 0.96 0.01
RbClO3 – – – – – – – – 6 0.3 0.98 0.01
CsClO3 – – – – – – – – 6 0.3 0.51 0
Mg(ClO3 )2 15 2.24 0.59 6.80 – – – – – – – –
Ba(ClO3 )2 12 1.04 0.47 5.45 – – – – – – – –

Perchlorates
HClO4 30 23.23 1.05 13.99 16 11.91 11.64 0.12 [47] 27 5 9.02 0.12
LiClO4 9 5.52 1.88 22.43 – – – – 26 4.5 4.13 0.06
NaClO4 26 9.98 0.50 6.63 – – – – 29 6 14.14 0.09
KClO4 – – – – – – – – 6 0.3 3.71 0.03
RbClO4 – – – – – – – – 6 0.3 5.10 0.03
CsClO4 – – – – – – – – 6 0.3 5.62 0.04
NH4 ClO4 9 1.50 0.19 2.00 – – – – 22 2.5 7.97 0.04
Mg(ClO4 )2 22 4.39 0.90 12.09 – – – – 16 1 9.53 0.07
Ca(ClO4 )2 12 4.53 0.64 8.07 – – – – 31 3 17.61 0.24
Sr(ClO4 )2 37 5.24 0.26 3.12 – – – – 39 4 15.91 0.28
Ba(ClO4 )2 13 4.46 0.42 5.81 – – – – 22 5.5 6.60 0.06
Cu(ClO4 )2 – – – – – – – – 36 3 25.66 0.63
Fe(ClO4 )2 10 0.15 0.04 0.38 – – – – – – – –
Co(ClO4 )2 – – – – – – – – 40 4 12.65 1.68

Bromates
LiBrO3 12 0.82 0.19 2.03 – – – – – – – –
NaBrO3 6 2.21 0.08 0.93 – – – – 23 2.62 3.01 0.02
KBrO3 9 0.32 0.05 0.45 – – – – 11 0.5 1.32 0.01
RbBrO3 – – – – – – – – 6 0.3 1.95 0.01
CsBrO3 – – – – – – – – 6 0.3 1.72 0.01
Mg(BrO3 )2 15 1.53 0.15 1.74 – – – – – – – –

Sulfates
Li2 SO4 6 3.03 2.23 26.26 20 2.99 2.70 0.08 [41] 20 3.17 20.29 0.06
Na2 SO4 6 2.35 0.41 4.86 6 3.18 2.55 0.23 [45] 23 4.25 32.63 0.05
K2 SO4 5 0.64 0.08 0.86 18 0.97 0.39 0.03 [48] 13 0.69 1.61 0.01
Rb2 SO4 21 1.61 0.68 8.35 – – – – 14 1.8 11.94 0.03
Cs2 SO4 16 1.18 0.40 5.21 – – – – 16 1.63 12.31 0.03
(NH4 )2 SO4 19 7.57 2.16 26.59 12 6.01 2.49 0.07 [44] 19 4 17.44 0.03
MgSO4 18 2.77 0.54 6.55 – – – – 19 3.6 19.36 0.01
FeSO4 10 1.65 0.47 5.45 – – – – – – – –
CuSO4 9 1.38 0.66 7.48 – – – – 13 1.75 12.76 0.01
CoSO4 5 0.56 0.57 6.10 – – – – 15 1.51 17.58 0.03
Cr2 (SO4 )3 13 1.70 1.26 16.91 – – – – 11 1.2 35.18 0.01

Phosphates
Na2 HPO4 8 0.79 1.81 19.20 – – – – 10 1 11.72 0.03
K2 HPO4 – – – – – – – – 27 0.87 5.86 0.02
(NH4 )2 HPO4 – – – – – – – – 37 3.11 13.91 0.03
NaH2 PO4 7 0.58 0.59 6.00 – – – – 27 5 3.02 0.01
KH2 PO4 12 2.45 0.31 3.41 – – – – 20 1.8 2.47 0.01

Nsalts 97 97 37 37 106 106

Average 0.75 9.14 3.29 0.24 9.17 0.12

Experimental activity coefficients as well as solution densities are taken from Lobo et al. [24]. If not available there, the latter quantity is found in Ref. [25]. Vapor-pressure
data is taken from Refs. [8,31,33–44].

As mentioned before, ion-specific instead of salt-specific param- Obtaining such a universal set of parameters requires a simulta-
eters are used in ePC-SAFT. Thus, the ionic parameters determined, neous regression of several electrolyte solutions. For that purpose,
e.g. for Li+ are applicable to all electrolytes containing this ion. 14 electrolytes AnCat were selected with Cat+ = {Na+ , Li+ , K+ } and
Hence,  Li+ and uLi+ /kB have the same values in LiCl, LiBr, LiOH, etc. An− = {F− , Cl− , Br− , I− , OH− }. The ion parameters were adjusted to
92 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96

the respective electrolyte solutions except to LiF where no values

for activity coefficients were available. After having adjusted the
segment diameter and dispersion energy of these eight ions, the
parameters of other ions (e.g. Mg2+ ) were determined from fitting
to the respective solution data (e.g. MgCl2 , MgBr2 , and MgI2 ). All
parameters obtained this way are listed in Table 2. (Note that due
to the parameter estimation using aqueous-solution data, these ion
parameters are linked to the water parameters as listed in Table 1.)
For the cations Li+ , Na+ , and K+ Collins et al. [26] reported X-ray
and neutron diffraction measurements of aqueous salt solutions
providing the distance between a central cation and the nearest
water oxygen. This directly compares to the ePC-SAFT parameter
 j , since this in fact is the size of the hydrated cation. Comparing
the experimental values to the  j parameters given in Table 2 shows
an excellent agreement of experimental data and fitted parameters.
This can be seen in Table 3.
It becomes apparent from Table 2 that anions containing three
Fig. 1. Liquid densities of aqueous solutions of six cesium salts related to the density
or more oxygen atoms (e.g. the nitrate or bromate anion) can
of pure water at 20 ◦ C as function of salt molality. The lines represent calculations by
be modeled with uj /kB = 0, i.e. without accounting for dispersion ePC-SAFT. The circles represent experimental data [29]. Parameters and deviations
interactions. This can be explained by their structure: they are are shown in Tables 2 and 3.
sufficiently ‘padded’ with oxygen atoms to prevent an extensive
hydration (Ref. [24], p. 122). Robinson based this suggestion on con-
4.1. Solution densities and vapor pressures
ductivity measurements from which he concluded a high mobility
of these ions compared to other anions and thus only weak hydra-
As a typical example, the liquid densities of six cesium-salt solu-
tion. Although bivalent anions are in general strongly hydrated,
tions are shown in Fig. 1. All solution densities are presented as
Robinson observed the described effect also for the anion SO4 2−
density differences  between the density of a salt solution and
which agrees also to the zero SO4 2− dispersion parameter in Table 2.
pure water. As to be seen, the experimental data are described with
On the other hand sulfate salts (e.g. ammonium sulfate) have a
high accuracy even at high salt concentrations of up to 6 mol/kg.
salting-out effect on amino acids which militates in favor of strong
Although the cesium parameters have been adjusted to cesium-
water–sulfate interactions. Additionally, Cannon et al. [27] investi-
salt solution densities at 20 ◦ C, ePC-SAFT correctly predicts the
gated the structure of water and stated that a huge amount (12–13)
respective densities also at other temperatures. This is illustrated
of water molecules can be found in the first hydration shell of sulfate
in Fig. 2 which shows predicted liquid densities of cesium-salts
anions. These opposing conclusions drawn based on interpreting
solutions at 40 ◦ C. No experimental density data were available for
ePC-SAFT parameters as well as on analyzing different experimen-
CsF.
tal data obviously need further investigations. Furthermore, the
The ePC-SAFT parameters can be used to predict vapor pres-
appearance of ion paring in sulfate systems makes an analysis even
sures of salt solutions which is exemplarily shown in Fig. 3 for
more difficult.
some cesium salts at concentrations of up to 10 mol/kg at differ-
ent temperatures. The predicted vapor pressures agree very well
with the experimental data at different temperatures. In the case
4. Modeling results of CsCl solutions, applying the universal parameters leads to slight
model inaccuracies for higher system temperatures at concentra-
Using the parameters summarized in Table 2 liquid densities, tions above 4 mol/kg. At these conditions, model deviations might
vapor pressures (not included in the parameter fitting), and solute be due to the fact that CsCl cannot be assumed as fully dissociated
activity coefficients (MIAC) were modeled. Table 4 summarizes the any more. Therefore, the strength of Cs+ hydration might be over-
absolute average deviations (AAD) and absolute relative deviations
(ARD) of solution densities (pvT), vapor pressures (VLE) and MIAC
for 115 aqueous electrolyte solutions. AAD and ARD are calculated
by

1  calc
NP
exp
AAD = |(yk − yk )|
NP
k=1

 (11)
1 
NP
ykcalc
ARD = 100 × 1 −
NP yk
exp
k=1

The above-mentioned thermodynamic properties of 115 aque-

ous electrolyte systems can be modeled reasonably by ePC-SAFT
with overall ARDs of 0.75% (pvT), 3.29% (VLE), and 9.17% (MIAC).
Considerably high deviations of MIAC values were found for some
sulfates, nitrates, and fluorides. One reason is that ion pairing,
which is expected and experimentally proven for these systems, is
not accounted for at the moment. This will be subject of the second Fig. 2. Liquid densities of aqueous solutions of six cesium salts at 40 ◦ C as function
part of this work [28]. of salt molality. No data available for CsF. Same notation as in Fig. 1.
 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96 93

Table 5
Comparison of relative bare Pauling ionic diameters

Cation  (Å) Anion  (Å)

+ −
Li 100% Cl 100%
Na+ 160% Br− 110%
K+ 220% I− 120%

The smallest cation as well as the smallest anion correspond to the 100% value.

As it can be seen from Table 2, also the ePC-SAFT diameters for

Fig. 3. Vapor pressures of aqueous solutions of three cesium salts at 60 ◦ C (squares the halide anions as well as for the alkali cations follow the same
and lines), 50 ◦ C (crosses and dashed lines), and 30 ◦ C (circles and dashed dotted
trend as the Pauling diameters. The size of H+ is found between
lines) as function of salt molality. Experimental data from Ref. [30].
the lithium and sodium ion. This can be explained by the fact that
H+ always exists as hydronium ion in water. As already mentioned
estimated at these concentrations and temperatures by ePC-SAFT above, the size parameter  j represent the diameter of the hydrated
leading to overestimated vapor pressures. ions rather than the diameter of the bare ions. That’s why all cation
Fig. 4a and b show the influence of halide anions (Fig. 4a) and parameters are considerably larger than the Pauling radii. In con-
alkali cations (Fig. 4b) on the water activity coefficients in aqueous trast, the rather weakly hydrated anions possess more or less the
solutions of alkali halides. As it can be seen, the predictions with same fitted diameter as the Pauling ones. Besides by their lower
ePC-SAFT are again in good agreement with the experimental data. charge density, this can be explained by the structure of the water
For all salt solutions, at very low salt concentrations first a slight molecules: the hydrogen atoms of water can approach the anions
increase of the experimental water activity coefficients is observed. about 0.85 Å more closely than the oxygen atoms approaching the
After passing a maximum, the values decrease continuously. This cations [26].
behavior holds true for every strong electrolyte in aqueous solution.
Moreover, the experimental data show decreasing water activity 4.2. Mean ionic activity coefficients
coefficients for increasing sizes of the anion but for decreasing
sizes of the cations. This means that the ion-water interactions A very important property characterizing electrolyte solutions
increase in the order Cl− < Br− < I− for the anions, but in the order is the mean ionic activity coefficient. As it deviates much more
K+ < Na+ < Li+ for the cations. Since the smallest cation but the largest from unity than the activity coefficient of water, it is a much more
anion causes the lowest water activity coefficients, there are obvi- sensitive quantity. The MIAC first decreases with increasing concen-
ously two opposing effects in aqueous electrolyte solutions. The tration (Fig. 5). After reaching a minimum it often increases with
first effect is a sterical effect that allows more water molecules to increasing salt concentration reaching in some cases very high val-
be placed around the larger ions (Ref. [32], p. 209 and Ref. [33], ues. The latter phenomenon might be an evidence for an extensive
pp. 32–33 and 55–57). This seems to be the important one when hydration [24].
comparing the influence of different anions. The second one is an Using the parameters from Table 2, ePC-SAFT performs well
electrostatic effect: at the same charge, the smallest ion causes the in modeling the MIAC for most electrolytes. Fig. 5 shows the
strongest electrical field which results in the strongest interaction modeled MIAC of some cesium electrolytes at 25 ◦ C. Only CsF is
with the surrounding water molecules (Ref. [32], p. 209 and Ref. calculated less accurately. At concentrations of above 0.2 mol/kg,
[33], pp. 32–33 and 55–57). The effect obviously dominates the the estimated value of ±∗,m for CsF is lower than that observed
alkali cation hydration. It can be further observed from Fig. 4 that experimentally (Fig. 5). This is observed also for rubidium fluo-
the influence of the halide anions (Fig. 4a) on the water activity ride, whereas sodium and potassium fluorides are modeled with
coefficient is much weaker than that of the alkali cations (Fig. 4b). considerably higher precision. Several authors report that fluoride
This is due to the fact that the relative size difference in Pauling salts form ion pairs even at room temperature [35,36]. In systems
diameters [34] (Table 5) is much bigger for the alkali cations than where extensive ion pairing is expected (e.g. fluorides) the model
it is for the halide anions.

Fig. 4. Influence of salts on the activity coefficient of water at 30 ◦ C. The symbols

represent experimental data [25,30,31] from isopiestic or vapor pressure measure-
ments. The lines are predictions with ePC-SAFT. (a) Influence of anions on the activity
coefficient of water. Circles: LiI, squares: NaI, triangles: KI. (b) Influence of cations on Fig. 5. Mean ionic activity coefficients of aqueous solutions of six cesium salts at
the activity coefficient of water. Circles: NaI, squares: NaBr, triangles: NaCl. Largest 25 ◦ C as function of salt molality. Experimental data from Lobo [25]. Same notation
anion and smallest cation cause the lowest water activity coefficient. as in Fig. 1.
94 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96

– based on the stoichiometric electrolyte concentration – under-

estimates the effects of non-ideality. The results in these systems
can be improved by implementing a theory dealing with ion pair
formation which will be pursued in the second part of this work
[28].
Analyzing experimental mean ionic activity coefficients, e.g. of
the alkali metal bromides (Fig. 6) again the hydration of ions can
be compared. The following sequence is observed: LiBr ∗ ∗
> NaBr >
∗
KBr ∗
> RbBr ∗
> CsBr . As already concluded from Fig. 4, the small-
est alkali cation (highest surface charge density) is most strongly
hydrated and strong hydration (low water activity coefficient) leads
to high MIAC values. The calculations for the MIAC series in Fig. 6
follow the same trend as the experimental data.
Additionally, it is generally known that 2:1 electrolytes (biva-
lent cation) like CaCl2 have much higher activity coefficients than
1:2 electrolytes (bivalent anion) such as, e.g. Na2 SO4 . This is again
based on the extent of hydration: bivalent cations (Ca2+ ) are much
more hydrated than bivalent anions because of their higher charge Fig. 7. Mean ionic activity coefficients of aqueous solutions of four hydroxides at
density. This is also reflected by the dispersive-energy parameter 25 ◦ C as function of salt molality. Experimental data: squares (LiOH), stars (NaOH),
circles (KOH), triangles (CsOH). The dotted lines represent ePC-SAFT calculations.
uj /kB (e.g. Ca2+ has a much higher value than SO4 2− ). Activity coefficients increase with increasing atomic size of the cation in the
Another interesting effect can be observed for hydroxides, flu- sequence Cs+ > K+ > Na+ > Li+ .
orides, and acetates (the latter are not shown here). Electrolytes
containing these anions show a reversed sequence of activity coef-
ficients compared to alkali salts shown in Fig. 6. Here the sequence
∗
is CsOH ∗
> RbOH ∗
> KOH ∗
> NaOH ∗
> LiOH which is illustrated for
the alkali hydroxides in Fig. 7.
A possible explanation for the reversal of the activity coefficients
is the so-call localized hydrolysis [37]. A hydrated cation Cat+ can
be regarded as a complex like H+ –OH− · · ·Cat+ · · ·OH− –H+ with the
cation placed in the center. A positive partial charge is built up at
the outer surface of the hydration shell. An anion which is a strong Fig. 8. Thermodynamic properties of aqueous RbCl solutions predicted by ePC-SAFT
proton acceptor (i.e. a derivative of a weak acid like F− , OH− , or the without parameter fitting to RbCl. (a) Vapor pressures at 40 ◦ C (squares) and at 30 ◦ C
acetate ion), will “dock” at the hydration shell to build an acid with (circles). Data from Ref. [30]. (b) Mean ionic activity coefficient at 25 ◦ C. Data from
Lobo et al. [25]. (c) Solution densities at 20 ◦ C. Data from Ref. [29].
the respective proton. Consequently, the water molecule is split up
to form the hydroxide of the cation and the acid of the anion:
the cation has a high charge density and the anion is a strong pro-
Cat+ · · ·OH− –H+ + F− → Cat+ · · ·OH− ||H+ · · ·F− ton acceptor. Thus, if the anion is a weak proton acceptor, as e.g.
the bromide ion, the lithium salt will be found to have the highest
As a result, the effective number of the ions in the solution is activity coefficient compared to the other bromides (Fig. 6) because
decreased which also leads to a decrease of the MIAC. The localized it is the most strongly hydrated cation. However, if the anion is a
hydrolysis depends at one side on how strong the cation polarizes strong proton acceptor, e.g., the hydroxyl ion, the extent of local-
the water molecules, and on the other-hand side on how strong ized hydrolysis increases in the order Rb+ < Cs+ < K+ < Na+ < Li+ . This
the anion acts as proton acceptor. It is strongest in systems where effect is large enough to reverse the order of the activity-coefficient
values. It is worth mentioning, that ePC-SAFT can even describe
this effect for the hydroxides and fluorides using the parameters
given in Table 2. The description of the acetate systems requires
accounting for ion pairing and will be considered in a subsequent
publication [28].
To demonstrate the predictive power of ePC-SAFT, the parame-
ters for the Rb+ ion were adjusted to six rubidium salts except to
RbCl. The parameters for the chloride ion had already been adjusted
before to LiCl, NaCl, and KCl. In Fig. 8, thermodynamic properties
of RbCl solutions are presented without explicitly having fitted the
Rb+ parameters to this salt. As it can be seen, vapor pressures, activ-
ity coefficients as well as solution densities of RbCl can be predicted
precisely although these experimental data were not included in
the parameter estimation.

5. Conclusions

In this study, the ePC-SAFT EOS proposed by Cameretti et al. [1]

Fig. 6. Mean ionic activity coefficients of five bromide salts in aqueous solu- was applied to describe thermodynamic properties of numerous
tion at 25 ◦ C as function of salt molality. Experimental data: squares (LiBr), stars
aqueous electrolyte solutions. Ion-specific parameters were used
(NaBr), circles (KBr), crosses (RbBr), triangles (CsBr). The dotted lines represent ePC-
SAFT calculations. Activity coefficients decrease with increasing size of the cation
which are independent of the electrolyte the ions are part of. Two
Li+ > Na+ > K+ > Rb+ > Cs+ . parameters, namely the diameter  j and the dispersion energy uj /kB
 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96 95

of the hydrated ions were adjusted to aqueous-solution densities W water

and MIAC data. Both parameters possess a physical meaning and ± mean ionic
show reasonable trends within the ion series. Using this approach, 0 pure substance
liquid densities, vapor pressures, and solute activity coefficients of 1 water
115 aqueous electrolyte systems were modeled reasonably with
overall ARDs of 0.75% (pvT), 3.29% (VLE), and 9.17% (MIAC). The Superscripts
predictability of the model has been proven by a quantitative assoc association
description of , p, and ±∗ of RbCl without having adjusted the calc calculated
parameters of Rb+ and Cl− to experimental data of that salt. disp dispersion
Electrolytes containing fluoride or hydroxyl anions show a exp experimental
reversed sequence of activity coefficients when compared to other hc hard chain
anion series. Even this phenomenon is correctly described by ePC- hs hard sphere
SAFT. ion ionic
m based on molality
List of symbols res residual
symm symmetrical (related to pure component)
x based on mole fraction
a Helmholtz free energy per number of particles (J) +, − positive or negative charge
aj distance at closest approach of two ions j (Å) ∞ infinitely diluted
A Helmholtz free energy (J)
ci molarity (moles solute per liter solution) (mol/l) Acknowledgements
e elementary charge, 1.6022 × 10−19 (C)
kB Boltzmann constant, 1.38065 × 10−23 (J/K) The authors gratefully acknowledge the financial support by the
kij binary interaction parameter German Society of Industrial Research (AiF) with Grant 14778N/3
m molality (moles solute i per kg solvent) (mol/kg) and the German Science Foundation (DFG) with Grant SA 700/7.
mseg number of segments We also wish to thank our student Markus Arndt for his help in the
M molecular weight (g/mol) parameter estimation.
N total number of particles; number of association sites
NP number of data points References
NA Avogadro’s constant, 6.023 × 10−23 (mol−1 )
[1] L.F. Cameretti, G. Sadowski, J.M. Mollerup, Ind. Eng. Chem. Res. 44 (2005)
p pressure (kPa, bar) 3355–3362, ibid., 8944.
qj charge of ion j (C) [2] R.M. Enick, S.M. Klara, SPE Reserv. Eng. 7 (1992) 253–258.
[3] M. Luckas, J. Krissmann, Thermodynamik der Elektrolytlösungen: Eine ein-
T temperature (K)
heitliche Darstellung der Berechnung komplexer Gleichgewichte, Springer,
Tdep,1.4 parameters for the temperature-dependent segment Berlin, 2001.
diameter (Å, 1/K) [4] P. Debye, E. Hückel, Phys. Z. 9 (1923) 185–206.
[5] E. Waisman, J.L. Lebowitz, J. Chem. Phys. 52 (1970) 4307–4311.
u/kB dispersion-energy parameter (K) [6] C.C. Chen, H.I. Britt, J.F. Boston, L.B. Evans, AIChE J. 28 (1982) 588–596.
x mole fraction [7] K.S. Pitzer, J. Phys. Chem. 77 (1973) 268–277.
z charge number [8] K. Nasirzadeh, R. Neueder, W. Kunz, Ind. Eng. Chem. Res. 44 (2005) 3807–3814.
[9] N. Papaiconomou, J.P. Simonin, O. Bernard, W. Kunz, Phys. Chem. Chem. Phys.
4 (2002) 4435–4443.
Greek letters [10] D.G. Archer, J. Phys. Chem. Ref. Data 20 (1991) 509–555.
i symmetrical activity coefficient of component i (related [11] J.A. Myers, S.I. Sandler, R.H. Wood, Ind. Eng. Chem. Res. 41 (2002) 3282–3297.
[12] W. Fürst, H. Renon, AIChE J. 39 (1993) 335–343.
to pure component) [13] P. Paricaud, A. Galindo, G. Jackson, Fluid Phase Equilib. 194 (2002) 87–96.
i∗ asymmetrical activity coefficient of component i (related [14] W.B. Liu, Y.G. Li, J.F. Lu, Fluid Phase Equilib. 160 (1999) 595–606.
to infinite dilution) [15] A. Galindo, A. Gil-Villegas, G. Jackson, A.N. Burgess, J. Phys. Chem. B. 103 (1999)
10272–10281.
ε dielectric constant of a medium, εr ε0 (C/Vm) [16] S.P. Tan, H. Adidharma, M. Radosz, Ind. Eng. Chem. Res. 44 (2005) 4442–4452.
εr relative permittivity [17] X.Y. Ji, S.P. Tan, H. Adidharma, M. Radosz, Ind. Eng. Chem. Res. 44 (2005)
ε0 permittivity in vacuum, 8.85416 × 10−12 (C/Vm) 7584–7590.
AiBi [18] S.P. Tan, X.Y. Ji, H. Adidharma, M. Radosz, J. Phys. Chem. B. 110 (2006)
εhb /kB association-energy parameter (K) 16694–16699.
 Debye length, defined in Eq. (4) (1/Å) [19] X.Y. Ji, H. Adidharma, Ind. Eng. Chem. Res. 45 (2006) 7719–7728.
AiBi /k
hb association-volume parameter [20] J. Gross, G. Sadowski, Ind. Eng. Chem. Res. 40 (2001) 1244–1260.
B
[21] R. Triolo, J.R. Grigera, L. Blum, J. Phys. Chem. 80 (1976) 1858–1861.
stoichiometric factor [22] F.X. Ball, H. Planche, W. Fürst, H. Renon, AIChE J. 31 (1985) 1233–1240.
 density (kg/m3 ) [23] S.H. Huang, M. Radosz, Ind. Eng. Chem. Res. 29 (1990) 2284–2294.
N number density (number of particles per volume) (1/Å3 ) [24] R.A. Robinson, R.H. Stokes, Electrolyte Solutions, 2nd ed., Butterworth, London,
1970.
i temperature-independent segment diameter of molecule [25] V.M.M. Lobo, J.L. Quaresma, Handbook of Electrolyte Solutions, Parts A and B,
i (Å) Elsevier, Amsterdam, 1989.
 T,W temperature-dependent segment diameter of water (Å) [26] K.D. Collins, G.W. Neilson, J.E. Enderby, Biophys. Chem. 128 (2007) 95–104.
[27] W.R. Cannon, B.M. Pettitt, J.A. Mccammon, J. Phys. Chem. 98 (1994) 6225–6230.
ϕi fugacity coefficient of component i [28] C. Held, G. Sadowski, Fluid Phase Equilib., in preparation.
 abbreviation for expression in Eq. (3) [29] J. D’Ans, H. Surawski, C. Synowietz, Landolt-Börnstein, Bd. IV/1b, Springer,
Berlin, 1977.
[30] K.R. Patil, A.D. Tripathi, G. Pathak, S.S. Katti, J. Chem. Eng. Data 36 (1991)
Subscripts 225–230.
An, − anion [31] K.R. Patil, A.D. Tripathi, G. Pathak, S.S. Katti, J. Chem. Eng. Data 35 (1990)
Cat, + cation 166–168.
[32] R. Steudel, Chemie der Nichtmetalle, 2. Auflage, W. de Gruyter, Berlin, 1998.
i, j, k component indexes [33] C.H. Hamann, A. Hamnett, W. Vielstich, Electrochemistry, 2nd ed., Wiley-VCH,
seg segment Weinheim, 2007.
96 C. Held et al. / Fluid Phase Equilibria 270 (2008) 87–96

[34] A.L. Horvath, Handbook of Aqueous Electrolyte Solutions: Physical Properties, [41] A. Apelblat, E. Korin, J. Chem. Therm. 30 (1998) 459–471.
Estimation and Correlation Methods, Ellis Horwood, Chichester, 1985. [42] A. Apelblat, J. Chem. Therm. 24 (1992) 619–626.
[35] R. Buchner, G.T. Hefter, J. Barthel, J. Chem. Soc. Faraday Trans. 90 (1994) [43] T. Ewan, W.R. Ormandy, B. Berkeley, J. Chem. Soc. T. 61 (1892) 769–781.
2475–2479. [44] A. Apelblat, J. Chem. Therm. 25 (1993) 1513–1520.
[36] A.D. Pethybridge, D.J. Spiers, J. Chem. Soc. Faraday Trans. 73 (1977) 768–775. [45] A. Apelblat, E. Korin, J. Chem. Therm. 34 (2002) 1621–1637.
[37] R.A. Robinson, H.S. Harned, Chem. Rev. 28 (1941) 419–476. [46] B. Behzadi, B.H. Patel, A. Galindo, C. Ghotbi, Fluid Phase Equilib. 236 (2005)
[38] A. Apelblat, J. Chem. Therm. 25 (1993) 63–71. 241–255.
[39] A. Apelblat, E. Korin, J. Chem. Therm. 30 (1998) 59–71. [47] J.N. Pearce, A.F. Nelson, J. Am. Chem. Soc. 55 (1933) 3075–3081.
[40] J.N. Pearce, A.F. Nelson, J. Am. Chem. Soc. 54 (1932) 3544–3555. [48] H.G. Leopold, J. Johnston, J. Am. Chem. Soc. 49 (1927) 1974–1988.