Fluid Phase Equilibria 427 (2016) 231e245

Contents lists available at ScienceDirect

Fluid Phase Equilibria

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / fl u i d

Pseudolattice Theory of electrolyte solutions: Consistency analysis of

the Quasi-Random Lattice model at innite dilution
Elsa Moggia
Department of Naval, Electric, Electronic and Telecommunications Engineering (DITEN), University of Genoa, Via Opera Pia 11A, 16145 Genoa, Italy

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

Article history: In recent years, the study of the electrolyte solutions has signicantly drawn advantage from the
Received 9 April 2016 Pseudolattice Theory, developed through various approaches and successfully applied to systems of
Received in revised form technological and scientic interest such as ionic liquids and rare-earth uids. However, promising
30 June 2016
potentialities from the applicative point of view are counterbalanced by a limited investigation about
Accepted 6 July 2016
general consistency of pseudolattice models with fundamentals of Solution Theory. This article focuses
Available online 9 July 2016
on the Quasi-Random Lattice approach and discusses, in particular, the theoretical consistency at innite
dilution, since convergence to the Debye-Hckel Limiting Law is a notoriously difcult task for lattice
Keywords:
Pseudolattice Theory
models not developed within the Debye-Hckel-Poisson-Boltzmann frame. The discussion throws a new
Effective-carrier atmosphere light on the pseudolattice treatment of electrolyte solutions, and denitely states in what sense an ionic
Ionic atmosphere lattice is included in the QRL model at strong, and even innite, dilution. Present developments
Debye-Hckel Limiting Law generalize previous QRL formalism and allow for advancing toward a unied pseudolattice approach.
Mean activity coefcient 2016 Elsevier B.V. All rights reserved.

1. Introduction equations suitable for engineering applications under very general

During the extensive study of the electrolytic solutions, re-
searchers have often resorted to the idea of some lattice-like
ordering of ions and molecules in the solution that would
explain, in particular, the formation of local structures experi-
mentally observed in medium-high concentrated solutions.
Numerous theories have been therefore developed around a lattice
concept, many of them with focus on the activity-coefcient
modeling.
In most local-composition models, notably based on quasi-
chemical and/or molecular thermodynamic approaches (Wilson
Equation, UNIQUAC, UNIFAC, NRTL, plus numerous renements and
hybrid treatments, see, e.g., Ref. [1] for a detailed review), a lattice-
like behavior is considered through the presence of non-
randomness factors that account for the tendency of molecules to
show some preference in choosing their immediate neighbors.
Originally developed for non-electrolytes, local-composition
models were proposed for electrolyte solutions by inclusion of a
modied Debye-Hckel (DH) term to account for long-range ion-
ion interactions. On the whole, these models deal with most ther-
modynamic functions and properties, and offer exible sets of
E-mail addresses: elsa.moggia@unige.it, elsa@biosafetyengineering.it.
conditions in terms of chemical composition, pressure, and tem-
perature ranges [1]. However, a relevant drawback is represented
by the number and the variety of required adjustable parameters,
often empirical or semi-empirical in character, which must account
for incompleteness in the available information about size and
shape of particles, ionic and molecular interactions, and thermo-
dynamic excess functions. A corollary difculty is that even pa-
rameters with a very similar denition from one model to another
can yield very dissimilar results. One should also state that, despite
their huge parameterization, sometimes these models present
inaccurate [2], and are forced to undergo an almost continuous
upgrading [2e4].
Some theories, adopting nearest-neighbor interaction princi-
ples, put a particular emphasis on the electrostatic character of
ionic and dipolar interactions, and in this connection it is to quote
the Lattice Restricted Primitive Model (LRPM) [5,6], developed with
main application to critical phenomena and phase transitions
(topics generally treated by the Lattice Density Functional Theory
[7]). Grand canonical Monte-Carlo simulations and ne-lattice
discretization methods have supported LRPM results [5], however
direct applicability of LRPM is thus far quite limited and proved
useful under very specic conditions (e.g., for modeling defects in
ionic crystals). In consideration of the innite-dilution analysis
presented in this article, it is also outlined that the LRPM equation

http://dx.doi.org/10.1016/j.uid.2016.07.005
0378-3812/ 2016 Elsevier B.V. All rights reserved.
232 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245
combines the linearized Poisson-Boltzmann (PB) equation with a
lattice Laplacian represented by a sum of nearest-neighbor differ-
ence potentials [6]. However, since the LRPM equation is built on
the DH equation, it does not allow for an independent derivation of
the DH Limiting-Law (DHLL) based on a pseudolattice approach.
Pseudolattice approaches developed almost independently of
the DH-PB theory rst appeared in literature in the early twentieth
century [8]. Most of them include the so called cube-root law,
originally motivated by the incompatibility between the ne-
grainedness of the ionic atmosphere, implied in the DH treat-
ment, and the coarse-grainedness of actual solutions [9]. The law
describes ion-ion long-range interactions as depending on the cu-
bic root of the concentration, that is, on the mean distance between
neighboring ions within an ionic crystal. The formal similarity with
ionic crystals justies the presence of the Madelung constant [10]
in the electrostatic contribution to the mean activity coefcient
g(molal scale), the latter typically provided by an equation such as
below.
 

 n n mM1
ln g Ac1=3 Bc Dc2  ln 1 (1)
1000
In Eq. (1), c is the molar concentration (mol/dm3); the last term
of Eq. (1) accounts for conversion from rational to molal scale; m is
the molal concentration (mol/kg), M1 the molecular weight of
solvent, n and n the stoichiometric numbers. In the long-range
electrostatic term Ac1/3, A is proportional to the Madelung con-
stant of a suitable ionic crystal. The term Bc arises from the
modeling of ion-solvent interactions that, in general, reect into
changes in the dielectric permittivity of the solvent from bulk to
local values around an ion. These concepts are also present in local
composition models [1] and Specic Ion Theory [8]. Theoretical
expressions for B were proposed based on hydration salting-out
effects [11], local ion-dielectric gradient interactions [12], and
dispersion energy [13]. In Eq. (1), the Dc2 term rst appeared in
Ref. [14] and was derived from the ion-ion short-range interaction
potential, which includes both the hard-core and attractive tail
(Van der Waals) parts. However, thus far the parameters B and D
(that depend on pressure and temperature) have been used as
adjustable parameters and their values only obtained by tting Eq.
(1) to experimental activity coefcients. Although quasi-ionic lat-
tice structures are presently observed in non-dilute solutions by
means of various experimental techniques (X Ray Diffraction,
Raman Spectroscopy and Neutron Scattering [15]) and are sup-
ported by conductivity studies [16e18], pseudolattice approaches
(Eq. (1)) have been subject to controversial attention over the last
few decades. Competitive potentialities of Eq. (1) pertaining to
highly concentrated solutions and ionic liquids [19], while using a
limited number of adjustable parameters, are counterbalanced by
the dependence on the cube-root law that causes problems in using
the equation with dilute solutions (the electrostatic Ac1/3 contri-
bution becomes predominant and suggests unrealistic, crystal-like
distribution functions). The main drawback of the cube-root law is
the evident inconsistency with DHLL that predicts a square-root
dependence on concentration. Many efforts were spent in the
past [8] in order to circumvent this problem and to establish which
starting concentration should be used such that the cubic-root law
is indeed preferable. Glueckauf [20] proposed an extended version
of the DH equation that approximated a cube-root law at inter-
mediate concentrations, Pitzer [21] tested the equation by
Glueckauf [20] versus the Virial Coefcient expansion, whereas
Frank and Thompson [9], and Rasaiah [22] explored concentration
ranges where a cubic-root law would collapse approximately into a
square-root law. However, no theory was derived and only empir-
ical rules, electrolyte-specic, were suggested.
The appealing potentialities of models depending on a limited
number of adjustable parameters are notably appreciable in the
Quasi-Random Lattice (QRL) model [23] developed independently
of Eq. (1). In QRL the parameterization is generally given by one
electrolyte-specic concentration parameter (at given temperature
and pressure), which also sets the upper limit of applicability (in
terms of concentration range) for the model. An undoubted
advantage is that the concentration parameter is experimentally
known [23e26] in a relevant number of cases including several
saturated and supersaturated solutions of both symmetric and
asymmetric electrolytes, for which no tting procedure is needed
for obtaining mean-activity and osmotic coefcients. In addition,
previous comparison [23e26] made over common aqueous 1:1, 1:2
and 1:3 electrolytes with most familiar one-parameter theories
developed within the PM frame (notably DH extended equations,
MSA and HNC) has shown that QRL usually performs better in
terms of agreement with experimental mean-activity and osmotic
coefcients, and this is remarkable noting that, in PM de-
velopments, the agreement is often conditioned by use of ionic
diameters adjusted through data regression techniques (in this
connection, a further comparison will be proposed with a recent
DH extended theory [27]). However, thus far the theoretical
agreement of the QRL model with fundamentals of Solution Theory
has been little investigated and needs to be analyzed in detail. To
the aim, the present article discusses, in particular, the thermody-
namic consistency of QRL at the innite dilution. It will be shown
that QRL represents an independent pseudolattice approach from
which DHLL can formally be obtained (the existence of a relation-
ship between QRL and DHLL was only qualitatively intuited in
previous work [25]). The QRL-to-DHLL derivation will allow for
advancing toward a unied pseudolattice approach, applicable
from the innite dilution, and able to enlarge scientic consensus
and research efforts in regard to pseudolattice models that are
already under active consideration [18,19,26,28].
In the QRL model an ionic lattice represents the reference
electrostatic conguration which is, however, continually disor-
dered by thermal forces and molecular collisions occurring in the
solution. Lattice sites are conveniently used as points from which
observing and evaluating (in a statistical sense) the dynamic
behavior of carriers of charge, that are solute ions at strong dilution,
groupings of solute ions and solvent molecules at higher concen-
trations. At each lattice point (and related cell volume) one can then
associate an effective carrier, representing the average behavior of
all carriers pertaining to the cell, and described by an effective
(average) density of charge. The standard deviation (from the lat-
tice point) associated with the effective density represents the
extent of the effective carrier, and arises from combining effects
due to size and stochastic movement of carriers. Standard de-
viations are generally large at strong dilution, causing the mean
interaction energy (which yields the mean activity coefcient) to
deviate signicantly from the Madelung energy of the reference
ionic-lattice.
The convenience in considering an ordered ionic lattice as the
reference electrostatic conguration persists until the number of
ions becomes too small even at microscopic scale, and charges
orderly disposed according to their signs no longer represent the
preferential conguration from the electrostatic point of view.
Therefore, the key step in going from QRL to DHLL is to relax the
reference conguration by replacing the ordered lattice with an
almost randomized lattice, characterized by the minimum corre-
lation in terms of site occupancy, i.e., correlation concerns only
neighboring sites, while the average interaction involves the min-
imum number of ions, that is, two ions of opposite sign.
A milestone in the study of electrolyte solutions from the
statistical-mechanical point of view is represented by the
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245 233

Fluctuation Solution Theory (FST), rst developed by Kirkwood and

co-workers [29,30] upon the grand canonical ensemble of inde-
pendent variables, and reconsidered later through numerous
reformulations, among them, those by O Connell and De Glance
[31], Kusalik and Patey [32]. DHLL was obtained as a by-product of
FST under specic limit procedures, that is, by the rst-order
approximation of a charging parameter [30], and by evaluation of
limiting forms of correlation functions and their integrals through
algebraic manipulations [31] and Fourier Transform analysis [32],
with emphasis on partial derivatives of the mean activity coef-
cient and related limiting laws expressed at the various levels of
representation (McMillan-Mayer, Lewis-Randall, and Kirkwood-
Buff [31e33]). Besides the quoted works, further derivations were
proposed, e.g., in the Field Theory Representation [34] DHLL was
obtained as the Gaussian limit of a power-expansion of the local
electrostatic potential of charged systems. In comparison with FST,
the QRL-to-DHLL derivation is a simplied route because the study
upon the reduced set of dynamic variables pertaining to carriers
(the remaining observables from the full set at molecular level [35]
being ascribed to the environmental particles forming the thermal
bath) is converted into the study upon a system of equivalent
dielectric bodies (effective carriers). Consequently, effective den-
sities of charge are not straightforwardly readable in terms of the
usual distribution functions. However, the QRL-to-DHLL procedure
is simpler per se, in addition it is stressed that many models are
forced to be consistent, a priori, with DHLL, yet with some artice
[36], by inclusion of modied DH terms or by parameterization of
properties [1,8,37e40] often leading to theoretical corrections of
the DH Screening Length.
Although this article is mainly concerned with the innite Fig. 1. QRL approach. Strong dilution. (a) Reference electrostatic conguration (ionic
dilution, some aspects relevant to medium-high concentrations lattice) and reference cell for a charge Q. Red (blue) circles: positive (negative) charges.
will also be discussed, and application of QRL to systems of interest (b, c, d) Displacements of Q from its reference lattice point, due to the environmental
in geochemical context will be presented and compared with thermal bath. Numbers refer to rounded probabilities (%) of nding Q within spheres
centered on the lattice point (aqueous 1:1 electrolytes, 104 103 mol/dm3, 298.15 K).
recent results from other approaches based on lattice concepts.
(For interpretation of the references to color in this gure legend, the reader is referred
to the web version of this article.)
2. Theoretical methods

In what follows, the innite dilution is dealt with starting from
reconsidering the two-step QRL procedure [23]: rstly, the refer-
ence electrostatic conguration and secondly, the perturbing ef-
fects due to Brownian Motion in a thermal bath. In this paragraph,
only highly dilute solutions are concerned, therefore as far as large
separation distances among carriers are considered, carrier-size
niteness and short-range forces can be neglected, and only long-
range Coulomb interactions taken into account. Correspondingly,
carriers are simply solute ions modeled as point charges and the
solvent is a continuous medium with dielectric permittivity . Fig. 1
visualizes main ideas of the two-step QRL procedure for the strong-
dilution case.
In a volume V, N solute molecules give rise to n N carriers with
charge Q, and n N carriers with charge Q. The mean inter-ionic
distance is R 1/(nr)1/3, where n n n and r N/V. The
reference conguration corresponds to ions simultaneously
disposed at the reticular points (sites) in their nal state, to be
intended as the most observed conguration after multiple repe-
titions of the same process, starting from motionless ions, innitely
distant from each other. Clearly, the conguration representing the
initial state is a completely random lattice, whose potential energy
must be zero because ions are located at reticular points irre-
spective of their charges, meaning that, in practice, no role is played
by electrostatic forces. And in fact, the electrostatic consistency of
this conguration is guaranteed by the Coulomb interaction being
zero when the inter-ionic distance is innite. Note the analogy with
the Debye charging process [8], where ions are endowed with their
nominal charges, however, they are initially not interacting as if
they were all uncharged. The total (potential kinetic) energy of
the nN-particle system is consequently null. In the nal state, the
potential energy corresponds to the Madelung energy of an ionic
crystal provided that N is large enough (on the microscopic scale).
The Madelung energy arises from the sum of N identical contri-
butions, one per each molecule and each of the form aQQ/
(4pR), where a is the Madelung constant [26] (the term molecule
indicates, hereinafter, a generic grouping of n positive ions and n
negative ions; with QRL, Madelung constants refer to R and not to
the minimum distance between ions of opposite sign [10]). The
charge correlation in site occupancy decreases when N decreases,
hence at the innite dilution the minimum correlation is to be
considered. In such situation, the only correlation effect is that of
favoring, among ions, some formation of groupings albeit statisti-
cally independent from each other. Each grouping, observable over
a few process realizations, is constrained by electro-neutrality
considerations to arrange n (neighboring) sites involving n posi-
tive ions and n negative ions (once again, a molecule). Since each
molecule is statistically independent of the others, and ions
belonging to the same molecule are in there randomly located, on
the whole, an almost randomized lattice is observed. Within a
molecule at any site A, the average charge <QA> is given by
el Q  Q  , where n =n is the probability (in presence
el el
of electrostatic forces only) of nding a positive or a negative ion
and, since nQ nQ 0 in virtue of electro-neutrality, it is
immediately <QA> 0. For any pair of sites A and B not belonging
to the same molecule, the pertinent interaction energy Wel AB is
averaged to account for the lattice randomization, leading to
234 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245

AB > hQ A Q B i=4pRAB hQ A ihQ B i=4pRAB 0,

< Wel where In summary, since the total energy of the nN-particle system is
RAB |RARB| (RA and RB are the space coordinates of sites A and B). null, for any molecule one has:
Conversely, if A and B belong to the same molecule, then
n M v2
X
    i el;i aQ Q 
el
Q Q n 1 Q  n 0 Strong Dilution (4)
<Wel
AB > A;B 2 4pR
4pRAB n 1 n 1 A;B i1

  
el Q Q n Q n 1 and
   A;B
4pRAB n 1 A;B n 1
n M v2
X i el;i Q Q 
el Q fQ n 1Q  n gel
 Q  fQ n Q  n 1g
0 Infinite  Dilution Limit (5)
2 4pR
4pRAB n 1 i1

In Eqs. (4) and (5), Mi and vel,i are the mass and velocity (in
Q fQ g Q  fQ  g n Q fQ g n Q  fQ  g
el el
presence of electrostatic forces only) of i-th ion (for brevity, the
4pRAB n 1 4pRAB nn 1
subscript i will be omitted in the following). It is stressed again
n Q  fQ g n Q fQ  g Q Q that R is very large at the scale of short-range interactions, and in

4pRAB nn 1 4pRAB n 1 addition, Eqs. (4) and (5) state that velocities (in modulus) are of 1/
(2) R1/2 order, which allows for considering the 1/R1/2 regime as
innitesimal so that dynamic properties change very slowly.
For symmetric electrolytes, since n n 1 there is one pair of Accordingly, in spite of the minimum potential energy at R 0,
sites within each molecule, at distance RAB R; with asymmetric from Eq. (4) the ionic lattice will be kept with negligible modi-
electrolytes, the conclusion is less immediate because we should cations due to almost-zero velocities, while from Eq. (5), charges at
rigorously consider electrolyte order and space geometry. However, neighboring sites will more or less keep their relative positions.
since R is very large we can neglect any RAB > R and assume a very The second step of the QRL procedure was to include the envi-
general arrangement of lattice sites (and cells), conceptually valid ronmental thermal bath by observing that the probability for an
for any n and n (in some way expressing the principle that, at the ion QA to remain within its own reference cell has to be small
innite dilution, all lattice geometries must lose their specicity), because an ion hypothetically located at point RA would be widely
that is, cells forming an open chain-like structure, where there are displaced far from RA due to thermal uctuations. So, being
exactly n1 pairs of neighboring sites for which RAB R (Fig. 2). small, was estimated by the rate dVel/dVth, where the volume
Therefore, the average interaction energy <Wel>, associated to a dVel |dxdydz| |vel.xvel,yvel,z|dt3 is that accessible (in a time in-
single molecule, is equivalent to one (Q,Q) interaction at distance terval dt) by means of vel of Eq. (4). Concerning dVth <|vx,th|><|
R, as shown below. vy,th|><|vz,th|>dt3, it is a statistical measure of the volume where
X X deviations from purely electrostatic paths occur due to Brownian
Q Q Q Q
< Wel > z Motion and dissipative forces (molecular collisions and momentum
RAB
4pRAB n  1 R R 4pRn  1 transfer within the thermal bath [35]), being, known as, <|vx,th|
AB
AsB
> <|vy,th|> <|vz,th|>(2kBT)1/2/(pM)1/2, where kB is the
Q Q Q Q
n  1  (3)
4pRn  1 4pR

Fig. 3. QRL approach. Strong dilution. Probability density PA (multiplied by R3 for

Fig. 2. Almost-randomized lattices. Pictorial representation. (a) Symmetric elec- readability), Eq. (7), for an aqueous 1:1 electrolyte at 298.15 K; FCC (Face Cubic
trolyte; (b) asymmetric electrolyte (exemplied by an uni-trivalent system). Red (blue) Centered) reference lattice [23]; Madelung Constant [10] 1.747. Black line,
circles: positive (negative) charges located at lattice points (sites) (gure (b) refers to c 105 mol/dm3; red line, c 104 mol/dm3; blue line, c 103 mol/dm3. The green
Q 3Q, n 1,n 3); yellow shapes: open-chain molecules. (For interpretation arrow symbolizes a Dirac distribution centered on the lattice point RA, for a compar-
of the references to color in this gure legend, the reader is referred to the web version ison with a crystal-like distribution. (For interpretation of the references to color in this
of this article.) gure legend, the reader is referred to the web version of this article.)
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245
Fig. 4. QRL approach. Strong dilution. Probability density PA (multiplied by R3 for
readability), Eq. (7), for an aqueous 1:1 electrolyte at 298.15 K. Black grid: plane view of
the reference lattice (FCC), R 203 (c 104 mol/dm3). The lattice point RA belongs
to the z 0 plane.
Boltzmann constant and T the absolute temperature. In the 1/R1/2
regime, mean and peak velocities can be approximately inter-
changed, and the probability (product of three axial probabili-
ties) evaluated at the most favorable (maximizing) condition for
each axial probability, that is, as if vel were wholly developed along
with that axial direction while keeping the remaining ions
motionless. In virtue of Eq. (4), the maximum axial velocity in
modulus equates (a|QQ|)1/2/(2pMR)1/2, yielding below.
s !3 0s 13 respect to RA-RB, and, by comparison with Eq. (8), we write:
ajQ Q  j ajQ Q  jrnR2 A
@
4kB TR 4kB T
1 R3
p3 !3 Strong Dilution (6)
2p p
p2
UD
pa
In Eq. (6), the DH Screening Length UD (kBT)1/2/(nr|QQ|)1/2
is evidenced. As expected, since R/UD is proportional to r1/6, is
very small in the strong-dilution range, which also allows for
approximating the (unknown) underlying probability density PA by
means of a Gaussian distribution centered on RA, that must lead to
after integrating over a cell volume R3 (centered on RA) in the
approximation R << UD (from Eq. (6), it is easily seen that the linear
standard deviation of the approximating Gaussian density must be
21/2UD/(pa)1/2). In conclusion, a charge QA located at RA in the
reference lattice, can only be assigned via a probability density PA,
largely spreading out of the cell volume (because of the thermal
environment) and leading to the effective charge density QAPA.
!
Q jr  RA j2
Q A PA r pA 3 exp  (7)
2p U 2U 2
In Eq. (7) r is a generic position with respect to a xed origin in
the space frame. By denition, 3U2 <|rA  RA|2>, where rA de-
notes the stochastic position of QA, and U 21/2UD/(pa)1/2, from Eq.
(6) (note, U being proportional to UD, the extent of the effective
carriers is large at strong dilution, in spite of the assumption of
235
zero-size carriers). Figs. 3 and 4 show PA densities for various
concentrations of an exemplicative aqueous electrolyte at
298.15 K (the relative dielectric constant of water is 78.45).
In actual fact, PA PA(|r  RA|2) can be derived from a joint
probability density PAB0 B00 ..(r, r0, r00,..) written in the form PAB0 B00 ..(|
r  RA  r RB0  r00 RB00 ..|2), which leads to Eq. (7) if r0 RB0 ,
r00 RB00 .. with certainty (Eq. (7) implies that any QB other than QA is
kept at its lattice position while displacing QA along with its elec-
trostatic path, because of the almost-zero velocity regime, Eq. (4)).
At the innite-dilution limit, for an ion QA we can derive an
expression for analogous to Eq. (6) just by considering (|QQ|)1/2/
(2pMR)1/2 in place of (a|QQ|)1/2/(2pMR)1/2, from Eq. (5).
s!  3 3
j Q Q j 3 1 R
4kB TR 2 UD3
1 R3
p3 ! Infinite  Dilution Limit (8)
2p p 3
D 2
Up
p
In a given molecule, without loss of generality, we consider QA in
a cell A and an ion QB in a neighboring cell B. From the statistical
independence among molecules the joint probability density now
reduces to PAB(|rr0 (RA-RB)|2), that provides information about
the probability of nding QA at a relative position r-r0 from QB (in
some way reminding us of the statistical uncertainty that con-
strains to rather deal with RA-RB differences albeit in the reference
lattice). Integration of the Gaussian density PAB, yielding of Eq.
(8) in the approximation R << UD, is performed with respect to r-r0
over a volume R3 such that the r-r0 component (denoted
with xRB RA ) along with the direction of RB-RA is taken within the
range xRB RA R  R=2, the x-component of RB-RA being either R
or R (B either precedes or follows A), and analogously for the
remaining components. Accordingly, the linear standard deviation
in PAB now concerns the stochastic relative position rA-rB with
p !2
UD 2
3 p < jrA  rB  R A RB j 2 >
p
< j rA  RA j2 > < j rB  RB j2 >  2 < rA
 RA $rB  R B >
(9)
At the innite-dilution limit, due to Brownian Motion, carries
are moving almost independently from each other such that the
cross term can be neglected and Eq. (9) approximated by
p !2
UD 2
3 p z < j rA  RA j2 > < j rB  RB j 2 > 3U 2 3U 2
p
6U 2
(10)
Eq. (10) yields the innite-dilution expression for U.
U
U pD U Infinite  Dilution Limit (11)
p
The Gaussian character of a joint density also reects onto
marginal densities, therefore Eq. (7) is still formally valid (despite
the uncertainty in determining which RA pertains to QA) provided
that U U is used.
The calculation of ln(g) proceeded [23] with that of the mean
236 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245

energy EA of an effective carrier A due to its interactions with the

others. Calculations, reported in detail in previous works [23,26],
rst required solving a Poisson equation to nd the potentials {VB}
generated by the effective densities {QBPB}.

 
j rR
p B j
Q B erf
2U
VB r (12)
4pj r  RB j

Fig. 6. Mean potential from ionic atmosphere (DH) and from effective-carrier at-
mosphere (QRL). Red line: DH theory; blue and black lines: QRL, Eq. (22). Aqueous 1:1
electrolyte at 298.15 K. Mean potentials (QA > 0) are all multiplied by UD (DH Screening
Length), for readability. (For interpretation of the references to color in this gure
legend, the reader is referred to the web version of this article.)

In Eq. (12), the symbol erf indicates the Error function. Then,
the overall contribution from any B, excluding A, was space-
averaged with respect to PA, leading to:
Z X  
QA X QB R
EA Q A PA r0 VB r0 dr0 erf AB
B
4p B RAB 2U
BsA BsA

(13)
Finally, ln(g) followed from the relationship ln(g) E/kBT,
P
where E 1=2 EA =n N n N (as in the Robinson-Stokes
charging process A[8], the factor avoids interactions counted
twice). The equation below was found useful for theoretical in-
vestigations [26], while fast numerical computation of ln(g) was
proposed via integral equations containing elliptic functions [41].
Helpful manipulations of lattice sums were also performed ac-
cording to the space-group geometry.
 

 1 1 1 X Q AQ B R
ln g erf AB (14)
kB T 2nN 4p A;B RAB 2U
AsB

Eq. (14) outlines that with QRL the reference lattice does not
represent short-range order in structures [12e14], but there is also
little in common with previous long-range order models [42].

Fig. 5. Ionic atmosphere (DH) and effective-carrier atmosphere (QRL). Strong

dilution, c 104 mol/dm3, aqueous 1:1 electrolyte at T 298.15 K. Dashed grid: plane
view of the reference lattice (FCC). (a) Ionic atmosphere, rDH, surrounding the central
ion QA. (b) Effective-carrier atmosphere, Eq. (19), surrounding dQA (belonging to
the effective carrier A with overall charge QA). (c) Average effective-carrier atmo-
sphere, Eq. (20). In all sub-gures, densities are multiplied by the square distances of
points from the frame origin, for readability.
Fig. 7. QRL model from 0 to clim. Eqs. 25 and 31 versus c/clim for aqueous electrolytes
at 298.15 K. All reported clim (M mol/dm3) are experimental [45]. Lattice structures
are detailed in Ref. [23](NaCl-like), [25](Fluorite-like) and [26](LaCl3-like). For the
Fluorite-like structure [25] simulations here refer to U U U with Madelung
Constant 1.679 (with respect to R; conventional values of Madelung constants are
given in Ref. [10]).
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245 237

completely independent of the analytical expression of U, Eq. (14)

can also be used at the innite-dilution limit. The largest distance
RAB is proportional to L and, since U is proportional to r1/2
(through UD, Eq. (11)), the greatest rate RAB/U is proportional to Lr1/
2
LN1/2/L3/2 (N/L)1/2 which /0 when L / . This allows using a
known approximation, that is, erf(x) z 2x/p1/2 (x small). Eq. (14)
becomes

 1 1 1 X Q A Q B 2 RAB
ln g z p
kB T 2nN 4p A;B RAB p 2U
AsB

1 1 1 1 X
p Q AQ B (15)
pU kB T 2nN 4p A;B
AsB

One also notes that:

P
 Q 2A
1 X A Q 2 Nn  Q 2 Nn
Q AQ B
nN A;B nN nN
AsB
 
n Q n Q  1 1
Q Q  (16)
n n n
Throughout Eq. (16) the electro-neutrality condition has been
applied. After substituting Eq. (16) into Eq. (15) one has:

 1 Q Q 1
ln g z p (17)
kB T 8p pU
At the innite-dilution limit, the DHLL is straightforwardly ob-
tained by putting U U from Eq. (11).


ln g / 1 Q  Q 1 1 Q Q 1
p DHLL (18)
r/0 kB T 8p pU kB T 8p UD
Within the QRL context it is possible to build analogies with the
most familiar concepts of the DH theory for dilute solutions: the
ionic atmosphere (excess charge density) surrounding the central
ion, and the corresponding mean potential. In the QRL space frame,
a small volume with charge dQA (belonging to an effective carrier
A), located at a point r0, is surrounded by a charge density rQRL at
any point r r0 due to all the effective carriers excluding A:

dQ A r0 Q A PA r0 dr0 ; rQRL dQ A r0 ; r r0
X
Q B PB r r0 (19)
B
BsA

The average effective-carrier atmosphere that can be associ-

Fig. 8. Pseudolattice Theory of electrolyte solutions. Strong-dilution. Plots refer to ated to A is obtained from Eq. (19) as follows:
aqueous electrolytes at 298.15 K, experimental data in Ref. [45]. Pseudolattice ap-
proaches [13,14]: Eq. (1), with A, B, D from the indicated references. Lattice geometries Z
 X

used by all: (a, b) NaCl-like structure, (c) Fluorite-like structure. < r > A
QRL r PA r Q B PB r r dr
B;BsA
!
1 X j r RA  R B j2
However, long-range order seems more consistent than short- p3 Q B exp
range order with use of Madelung constants that arise from 8 p U 3 B;BsA 4U 2
Coulomb sums slowly converging and only conditionally [43,44].
(20)

3. Results and discussion
3.1. Derivation of DHLL from QRL
Consider the limit r N/V / 0. For N nite, V / is taken as a
cubic volume, that is, V L3 / as L / , so the limit r N/
L3 / 0 implies N/L / 0. Since the derivation of Eq. (14) was
(Eq. (20) is derived in Appendix A). For the concentrations
considered in this article, Eq. (20) can be interpreted by stating that
a carrier (ion) QA is surrounded by an average charge
density < r > A
QRL r, r being taken from QA and the average done
over all possible positions for QA observed in the QRL frame (ionic
lattice) and not in the DH frame (QA itself). Integration of Eq. (20)
with respect to r yields:
238 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245

effective-carrier atmosphere is obtained after generalizing the QRL

procedure that yielded Eq. (13), preliminarily noting that
EA Q A < j > AQRL 0.
Z
 X

<j>A
QRL r PA r VB r r dr
B
BsA
 
j rR A RB j
X Q B erf 2U
(22)
B
4pjr RA  RB j
BsA

(Eq. (22) is derived in Appendix A). Fig. 6 compares < j > A

QRL r
with the classical DH functionjatm
DH (r) QA(exp(r/UD)1)/(4pr),
r |r|, for an exemplicative aqueous electrolyte. At the innite-
dilution limit, it is lim < j > A
QRL r lim jDH r, that can easily
atm
r/0 r/0
be proven from Eq. (22) after putting U U and after applying the
usual rst-order approximation of the Error function.

3.2. Unied Pseudolattice approach

It is best to gather results derived separately for the innite and

strong dilution to facilitate easier handling of the whole available
concentration range with QRL, that is, the range [0, clim], where the
molar concentration denoted with clim was dened [23] as the
(upper) concentration, at given pressure P and temperature T (that
is, clim clim(P,T)), at which the electrolyte reaches g 1, either
experimentally or hypothetically. To this aim, we aim for a positive
function f(c/clim) such that f(0) 1 (innite dilution), and we start
by generalizing U by writing:
 
U c
U pD f ; 0  c  clim exp  k (23)
p clim
The constant k appearing in Eq. (23) solves [25]:

Fig. 9. QRL approach. From strong dilution to low dilution. Pictorial representa-
tion. (a) Carrier composition in different concentration ranges. (b) Spherical volumes
lled by 20% of charge QA pertaining to an effective-carrier A in different concentration
ranges (reference cells are also shown for a better visual impression).

Z
<r>A
QRL rdr

Z !,
P j r RA  R B j2 p
QB exp 2
2U 2p 3 dr (21)
BsA 4U
P X
QB Q B  Q A Q A
BsA B

Eq. (21) states that the overall charge of the effective-carrier

atmosphere compensates that of QA (that is, electro-neutrality is
a condition intrinsic to the QRL approach). Fig. 5 shows the formal
process from the DH theory to the QRL approach, and presents
some numerical results for an exemplicative electrolyte at strong
dilution.
The mean potential < j > A QRL r that can be associated to the
Fig. 10. Pseudolattice Theory and local composition models. For case (a) and
Fluorite-like structure implementation details see also Fig. 7. Cases (b, c) refer to
saturated water vapor pressure (dielectric permittivities and densities: Ref. [49]).
Experimental data: Refs. [45,50]. Experimental clim (no tting procedure): 3.4 mol/dm3
at 423.15 K; 4.48 mol/dm3 at 473.15 K (MgCl2 densities are from Ref. [51]).
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245 239

The concentration climexp(k) was assumed as the highest

concentration at which the strong-dilution formulae should be
rigorously implemented, substantiated by a sufciently large k
(with the symmetric FCC, a 1.747 and k 10.65 from Eq. (24);
higher k values were obtained for asymmetric systems [25,26]). The
experimental clim usually lies between 1 mol/dm3 and 9 mol/dm3
for many aqueous electrolytes [23e26,45], thus the concentration
climexp(k) within 2.3  105 2  104 mol/dm3 for symmetric
systems (and within smaller ranges for asymmetric systems).
Hence, since climexp(k) is very small, we add a negligible error by
letting U 21/2UD/(pa)1/2 be exactly satised only at
c climexp(k). We then consider a power expansion for f(c/clim),
c  climexp(k).

   2  3
c c c c
f z1 A1 A2 A3
clim clim clim clim (26)
c  clim exp  k
The expansion coefcients A1, A2, and A3 (higher terms are
neglected here) are found by imposing continuity on partial de-
rivatives of U at c climexp(k) (in the mPT frame). From Eqs. 23
and 26, the continuity condition, U 21/2UD/(pa)1/2 at
c climexp(k), now reads:
r
k 2k 3k 2
Fig. 11. Pseudolattice Theory and local composition models. Cases (b, c) refer to 1 A1 e A2 e A3 e (27)
saturated water vapor pressure (dielectric permittivities and densities: Ref. [49]). a
Experimental clim 5.38 mol/dm3 (298.15 K). Estimates of clim (10.2 mol/dm3, 473.15 K;
Continuity of rst-order partial derivatives of U yields Eq. (28).
15 mol/dm3, 573.15 K) are obtained from the mean activity coefcient measured at the
highest available concentration. NaCl densities: Ref. [52]; experimental data:
Refs. [45,52,53]. A1 2A2 ek 3A3 e2k 0 (28)

r Continuity of second-order partial derivatives of U yields Eq.

k pa (29).
(24)
e1 lnk 2 r
1 2
Eq. (24) represents the continuity constraint imposed at 2A2 e2k 6A3 e3k (29)
c climexp(k) in going from strong dilution (U 21/2 UD/(pa)1/2) k2 a
to medium-high concentrations [25]: Eqs. 28 and 29 are derived in Appendix B (a proof of positiveness
of f is also given). Solving the system of Eqs. 27e29, one yields A1,
UD A2, A3 below.
U clim expk  c  clim (25)

 1=k
logk clim c r
!
c clim
ek 2 
A1 6k2 1  6k2
2k2 a
r !
e2k 2

A2 3k2 1  3k2 (30)
k2 a
r !
e3k 2
2

2
A3 2k 1  2k
2k2 a
In summary, Eq. (23) becomes

8 r ! 9
> ek c 2
2 
2 >
>
> 1 6k 1  6k >
>
>
>
> 2k2 clim a >
>
>
>
> >
>
>
> >
>
>
> r
! >
>
UD < e2k c2 2  =
U p  2 3k2 1  3k2
p> >
> k c2lim a >
>
>
>
> >
> (31)
>
> ! >
>
>
> r >
>
>
>
> e3k c3 2
2  >
>
>
: 2k 1  2k2 ;
2k2 c3lim a
0  c  clim exp  k
Fig. 12. QRL model and DH extended theory [27]. Plots refer to aqueous electrolytes
at 298.15 K. The mean activity coefcient is converted from molal to rational scale for a
comparison with results in Ref. [27]. Experimental clim [45]: 5.38 mol/dm3 (NaCl);
2.26 mol/dm3 (LiCl). In Ref. [27] ionic radii are set to their Pauling values. In spite of the presence of clim(P,T), limiting laws (limiting
240 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245
slopes) related to partial derivatives of DHLL with respect to P and T
are obtained by deriving Eq. (17) combining with the equation
below.
lim vnk U 1 lim vnk UD
p (32)
m/0 vP n vT k p m/0 vP n vT k
(A proof of Eq. (32) is given in Appendix C). Fig. 8 compares ln(g),
calculated using Eqs. (14), (25) and (31), with results from pseu-
dolattice approaches involving the cube-root law (Eq. (1)), with
emphasis on the strong-dilution range.
To deal with non-dilute solutions, QRL relies on the concept it-
self of effective carrier of charge that it indeed includes modica-
tions due to an increase of concentration through modications on
U, given in Eq. (25), while keeping the general formalism unaltered.
In Eq. (25), the factor(logk(clim/c)(c/clim)1/k)1 generalizes the quasi-
proportionality law with respect to UD derived at the strong dilu-
tion and states that local interactions among ions and molecules,
due to their granular nature, more and more predominate over
effects due to thermal bath (which implies a noisy but almost
continuous medium). Carriers were dened as groupings of solute
ions and solvent molecules, generally increasing with concentra-
tion, assembled by means of their short-range interactions, and
delimited by free solvent (Fig. 9).
Although full dissociation is assumed at strong dilution (other-
wise association constants are also required, e.g., the case of di-
divalent metal-sulfates [24]), at medium-high concentrations car-
riers are overlapping formations allowing, in principle, for ion as-
sociation and cluster-like behavior. In summary, with non-dilute
solutions, carriers enter into the reference lattice in the form of nN
dielectric bodies where charges are distributed and no longer
concentrated on zero-size particles. Dielectric bodies allow using
tools from Continuum Electrostatics, that is, using effective den-
sities into a Poisson equation and throughout all the equations up
to Eq. (14) even if short-range forces, due to structures in matter,
are strongly prevailing. The theoretical constraint for such an
approach is that one has ln(g)  0, from Eq. (14), with ln(g) 0 at
c clim (U / ). To achieve the aim of a unied pseudolattice
approach to the study of electrolyte solutions, clim establishes a
well-dened range of model applicability, connected to the
constraint ln(g)  0 in a transparent way. Hence, empirical or case-
by-case rules are avoided when switching to other existing models
to deal with concentrations > clim.
In the QRL approach the effect of short-range interactions is
primarily that of distributing charges over growing formations,
a point of view that outlines the electrical nature (ionic and
dipolar) of local forces. However, although carriers are distin-
guished from free solvent because of their non-zero net charges,
they were assigned the same dielectric permittivity [23]. This
was motivated by avoiding boundary problems due to disconti-
nuities in dielectric media, nevertheless it suggests relatively
weak or short-time modications in the solvent due to ionic
solvation, possibly allowing for molecules reversibly going from
bulk solvent to bound state (as far as c  clim). This said, the
dielectric behavior should be considered in detail, and various
points of view, from Mixture Theory to microscopic treatments
[46], should be evaluated. In addition, although thus far advan-
tageous because of simplicity and no restrictions, the absence of
any explicit modeling of local forces is an aspect demanding
improvement, and the mesoscopic nature of dielectric bodies
necessitates being linked with the underlying molecular nature
of solute and solvent.
Pseudolattice models, traditionally developed for modeling the
mean activity coefcient, are also being used to evaluate other
properties, in particular concerning ionic liquids and their apparent
volumes [47,48] and conductivities [18,19,28], with reporting of
experimental evidence to support the Bahe-Varela theory [14].
Perhaps QRL represents the most systematic application of a
pseudolattice approach to the calculation of mean activity co-
efcients and osmotic coefcients over a wide class of aqueous
electrolytes (1:1, 2:2, 1:2, 1:3), at 273e373 K and 0.1 MPa. The
osmotic coefcients were calculated [24] by integrating the
GibbseDuhem equation from the knowledge of ln(g). Figs. 10
and 11 show some results concerning the application of QRL (Eq.
(14)), with inclusion of developments presented in this article and
leading to Eq. (31), to electrolytic systems of practical interest
in chemical-geological context (geothermal brines), at high tem-
peratures (>400 K) and saturated water vapor pressures. Figs. 10
and 11 also show results from Bahe-Varela Theory [14] and from
Revised Extended UNIQUAC (REUNIQUAC) [2], in order to compare
model performances, in particular, in terms of tting procedures,
number of required adjustable parameters and explored concen-
tration ranges.
Fig. 12 compares QRL results with those from the DH extended
theory [27] concerning some 1:1 electrolytes at ambient condi-
tions. For the considered cases, in both theories all the involved
parameters have been assigned experimentally (a fundamental test
condition for any theory).
Fig. 12 outlines a remarkable capability of QRL in capturing some
essential aspects of the activity behavior.
4. Conclusions
The article has focused on important theoretical aspects of the
QRL (Quasi-Random Lattice) approach, in particular, the consis-
tency of the model at innite dilution has been discussed, and
convergence to DHLL (Debye-Hckel Limiting Law) proven. The
QRL-to-DHLL derivation has given occasion for clarifying in what
sense an ionic lattice is used within QRL at strong, and even innite,
dilution, the sense being that the lattice (with ions situated at the
reticular points more or less randomly depending on their number)
represents a reference electrostatic conguration while hypothet-
ically keeping the thermal bath off. In so doing, electrostatic in-
teractions are evaluated separately from thermal forces, the latter
acting as perturbations that are, of course, widely prevailing at the
considered concentrations.
In perspective, further research will be devoted to evaluating the
theoretical consistency of QRL with main paradigms of Solution
Theory, from the Moment conditions built upon the Ornstein-
Zernike equation [36], to the consistency analysis via thermody-
namic routes [32,36,39].
Presented results are conrming that QRL is a powerful
approach to the study of electrolyte solutions. This conclusion is
also supported by general agreement with experimental data as
documented in previous works, and also illustrated here by
some application to systems of practical interest. A wider explo-
ration of model performances, concerning volumetric and thermal
properties of electrolytes, is already under consideration, albeit
preliminarily, suggesting that results of remarkable accuracy
can be obtained by including relatively simple modications in
the model, that is, in accordance with the QRL philosophy,
without need for resorting to a prohibitive, or unphysical
parameterization.
Acknowledgment
The author wishes to thank Dr. Wilma Kirsten (London, UK) for
her proofreading support, and Prof. Bruno Bianco (University of
Genoa, Italy) for his scientic comments.
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245 241

Appendix A. Derivation of equations (20) and (22)

From Eq. (7) one has:

Z
0 X
 0
0
PA r VB r r dr
B;BsA

 r r0  R 

0 2 .  B
Z exp  r  R 2U 2 X Q B erf p
A 2U 0
p3 0 dr
2p U 3
BsA 4p j r r  R B

0 2 . 

exp  r  RA 2U 2
p3 
Z 2p U 3
2 !
Z1 h2 r r  RB
0
P QB 2 0
p exp  dh dr
BsA
4p p 2U 2
0

1 2 X
p3 p Q B
4p 2p U 3 p BsA
0 2 2 !
Z Z1 0
r  R A h2 r r  RB 0
exp  dhdr
2U 2
0
X (A.1)
1
p Q B
4 2p3 U 3 BsA

2 !
Z Z1 00 00
r 2
h2 r r RA  RB 00
exp  dhdr
2U 2
0

r00 r0  R A
1 X
p3 QB 
2p U BsA
X
Z1
h2 j r R  R j2 .
3=2 1
p3 QB 
exp 
A B 1 h2 dh 4 p U BsA
2U 2 1 h2
0

Z1 !
P QB u2 j r RA  RB j2
BsA p3  exp dusu=p2
2p U 4U 2
0
p
1= 2 !
Z
s2 j r R A  RB j2
exp dshs=1s2 1=2
2U 2
0
242 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245

Eq. (20) results from Eq. (A.1) after integrating over r00. c
From Eqs. 7 and 12, one has (using the integral representation of s (B.1)
clim
the Error function):
Partial derivatives are evaluated in the mPT frame so, after

Z
X

PA r VB r r dr
B;BsA

 r r  R 

2 .  B
Z exp  r  R 2U 2 X Q B erf p
A 2 U dr
p3
2p U 3
BsA 4p r r
 RB

2 . 

exp  r  RA 2U 2
p3 
Z 2p U 3
2 !
P QB 2
Z1 h2 r r  RB
p exp  dh dr
BsA
4p p 2U 2
0

1 2 X
p3 p Q B
4p 2p U 3 p BsA
2 2 !
Z Z1
r  RA h2 r r  RB
exp  dhdr
2U 2
0
X (A.2)
1
p Q B
4 2p3 U 3 BsA

2 !
Z Z1
r 2 h2 r r RA  RB
exp  dhdr
2U 2
0

r00 r  RA
1 X
p3 QB 
2p U BsA
X
Z1
h2 j r R  R j2 .
3=2 1
p3 QB 
exp 
A B 1 h2 dh 4 p U BsA
2U 2 1 h2
0

Z1 !
P QB u2 j r RA  RB j2
BsA p3  exp dusu=p2
2p U 4U 2
0
p
1= 2
Z !
s2 j r R A  RB j2
exp dshs=1s2 1=2
2U 2
0

denoting with d(m, P, T) the solution density, we use a standard

formula [45] for conversion from c to m.
Eq. (22) results after applying the integral representation of
erf(x) to Eq. (A.2), with x |r RA  RB|.
md
c (B.2)
1 m103 M2
Appendix B. Derivation of equations (28) and (29)
In Eq. (B.2) M2 is the molecular weight of solute. Continuity
For notational brevity we put: requirements are imposed at s1 exp(k).
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245 243

  v2 U 1 v 2 UD
vs 1 vd 1 vclim p f s
s  vT 2 p vT 2
vT d vT clim vT
  (B.3) 2 vUD vf vs
vs 1 vd 1 vclim p
s  p vT vs vT (B.11)
vP d vP clim vP
"  2 #
Since U U(r, P, T), the continuity of vU vU vr vc
vm vr vc vm is based on 1 v2 f vs vf v2 s
that of vU pUD
vr . Eqs. 23 and 25 give: p vs2 vT vs vT 2
s  s1
  
vU 1 U 1 v 1 vd vf vs
p D  f s UD
vP p 2 vP d vP vs vP (B.4) v2 U
lnk
s  s1 vT 2

v2 UD s1=k
 
vU U 1 v 1 vd 8 vT 2 lns 9
 >
> >
>
vP 2 vP d vP >
> >
>
>
> >
>
>
> >
>
 >
>  >
>
U lns 1 vd 1 vclim (B.5) >
> vUD s1=k1 1 1 vs >
>
 1  >
> >
>
>
> 2 >
>
lns k d vP clim vP >
> vT lns k lns vT >
>
>
> >
>
s  s1 >
> >
>
>
> >
>
>
> >
>
>
<  2 >
=
s1=k1 1 1 v s (B.12)
UD
   >
> lns k lns vT 2 >
>
vU 1 U 1 v 1 1 vd vf vs >
> >
>
p D  f s UD >
> >
>
vT p 2 vT T d vT vs vT >
> >
>
(B.6) >
> >
>
>
>  2 >
>
s  s1 >
> vs s1=k2 >
>
>
> >
>
>
> UD  >
>
>
> vT lns >
>
>
> >
>
>
> >
>
  >
> >
>
vU U 1 v 1 1 vd >
: >
;
    
vT 2 vT T d vT 1 1 1 1 1
 (B.7) 1
U lns 1 vd 1 vclim s  s1 k lns k lns lns2
 1 
lns k d vT clim vT
s  s1
Equating Eq. (B.11) with Eq. (B.12) at s s1 gives:
vU 1 U f s UD vf vs r
p D p s  s1 (B.8) v2 f 2 s2
vr p 2r p vs vr s s1 (B.13)
vs2 a k2
In Eq. (B.13), we have used Eq. (B.10) and Eq. (24). From Eq. (26)
  2
vU U U 1 1 one has vvs2f 2A2 6A3 s, which yields Eq. (29) after substitution
  s  s1 (B.9)
vr 2r r lns k into Eq. (B.13). Concerning the second-order partial derivative of U
with respect to P, continuity at s s1 again yields Eq. (B.13) (cal-
In Eqs. (B.4, B.6) the expansion coefcients of f are assumed not culations are omitted).
depending on P or T. This hypothesis is consistent with A1, A2, and A3 To evaluate the positiveness of f(s) over the interval [0, s1], we
given in Eq. (30), which denitely depend on the Madelung con- evaluate the stationary points of f(s) by solving
stant a of the reference lattice (note, k k(a) in virtue of Eq. (24)).  
2 vf
Rigorously, real electrolytes can modify their crystalline structure A1 2A2 s 3A3 s 0 vs 0 , which possesses a root at s1, as
with T and P, however the QRL reference lattice is set as the most

2
representative according to the electrolyte order, hence changes in already stated. Since vvs2f > 0 from Eq. (B.13), f(s1) is a local

the lattice geometry are here neglected. Since f(s1) (2/a)1/2, ss1
U(s1) 21/2UD/(pa)1/2, and (1 ln(s1)/k) 0, continuity between minimum of f. The second root is meaningful when lying within (0,
the various partial derivatives of U (Eqs. (B.4-B.5, B.6-B.7, B.8-B.9)) is s1), and in such case it gives a local maximum, thus we conclude
satised only if that, over the interval [0, s1], f(s)min (f(0), f(s1)) min(1, (2/a)1/
2
) > 0.

vf
0 (B.10)
vs ss1
Appendix C. Derivation of equation (32)
Eq. (B.10) gives Eq. (28) for f(s) of Eq. (26). Concerning the
second-order partial derivative with respect to T, from Eqs. 23 and From Eq. (23) and Eq. (B.1) we write, considering, e.g., de-
25 one has: rivatives with respect to T:
244 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245

T Absolute temperature [K] Eq. (6)

lim vk U 1 lim vk f sUD P Pressure [Pa]
p
m/0 vT k p m/0 vT k QA Charge of carrier (ion) A; total charge of effective carrier A
  [Coulomb] Eq. (7)
1 lim X k vp f s vkp UD
pk
p (C.1) RA Space coordinates of lattice point A [m] Eq. (7)
p m/0 p0 p vT p vT kp PA Effective charge density (divided per QA) Eq. (7)
kp U Linear standard deviation of PA [m] Eq. (7)
In the mPT frame, when m/0 (r / 0), each vvT kp UD
in Eq. (C.1) PAB0 B00 .. Joint probability density among ions A, B0 , B.
1/2kp
contains a diverging term proportional to r , which is the rA Stochastic position of carrier A [m] Eq. (9)
greatest innite-order, thus the only to retain in the limit m/0. U Linear standard deviation of PA at innite dilution [m] Eq.
p
Concerning vvTf s
p , it involves partial derivatives of s, which are ob- (11)
tained by iterative application of Eq. (B.3), so they can denitely be erf Error Function Eq. (12)
expressed by products of the form sh, where h(m, P, T) is some not- VB Mean potential of effective carrier B Eq. (12)
diverging function when m/0. In the limit m/0, it is s/0, and for EA Mean interaction energy of effective carrier A Eq. (13)
vp f s dQA Elemental charge of effective carrier A [Coulomb] Eq. (19)
p > 0 the only term to retain within vT p is that proportional to s,
rQRL Effective-carrier atmosphere around dQA Eq. (19)
that is, vfvss vp s
vT p . Consequently, in the limit s/0, for p > 0 the order of
p <r>A QRL Effective-carrier atmosphere around QA Eq. (20)
v f s v UD
kp k k

vT p vT kp is r11/2kp r1/2kp; for p 0 it is f 0 vvTUkD vvTUkD , <j>A

1/2k QRL Mean potential from effective-carrier atmosphere around
whose innite-order is r >r 1/2kp
(p > 0). In conclusion, Eq. QA Eq. (22)
(C.1) reduces to: jatm Mean potential from DH ionic atmosphere around QA
DH
clim Molar concentration corresponding to g 1 [mol/dm3]
lim vk U 1 lim v k UD 1 lim vk UD Eq. (23)
p f s p (C.2)
m/0 vT k p m/0 vT k p m/0 vT k k QRL lattice constant Eq. (23)
p
Generalization of Eq. (C.2) (omitted here for brevity) leads to Eq. f pU/UD Eq. (23)
(32). It is worthwhile noting that, c being proportional to r, A1, A2, A3 Expansion coefcients of f Eq. (26)
thenvvsr rs is not diverging in the limit r / 0 and Eq. (32) gives: s c/clim Eq. (B1)
d Absolute density of the solution (g/cm3) Eq. (B2)
vU 1 U f s UD vf s 1 U M2 Molecular weight of solute Eq. (B2)
p D p /  p D r/0 (C.3)
vr p 2r p vs r p 2r
References
In deriving Eq. (17) with respect to r, one applies Eq. (C.3) and
obtains the usual limiting law [32] for
vlng
j
vr T;P r/0 . [1] G.M. Kontogeorgis, G.K. Folas, Thermodynamic Models for Industrial Appli-
cations. From Classical and Advanced Mixing Rules to Association Theories, J.
Wiley and Sons, Chichester, UK, 2010.
[2] F.F. Hingerl, T. Wagner, D.A. Kulik, K. Thomsen, T. Driesner, A new aqueous
Nomenclature activity model for geothermal brines in the system
NaeKeCaeMgeHeCleSO4eH2O from 25 to 300  C, Chem. Geol. 381 (2014)
QRL Quasi-Random Lattice Model 78e93.
[3] S. Gebreyohannes, B.J. Neely, K.A.M. Gasem, Generalized interaction param-
DHLL Debye-Hckel Limiting Law eter for the modied Nonrandom Two-Liquid (NRTL) activity coefcient
g Mean activity coefcient (molal scale) Eq. (1) model, Ind. Eng. Chem. Res. 53 (2014) 20247e20257.
c, m Molar, molal concentration [mol/dm3, mol/kg] Eq. (1) [4] K. Parvaneh, R. Haghbakhsh, A. Shariati, C.J. Peters, A crossovereUNIQUAC
model for critical and noncritical LLE calculations, AIChE J. 61 (2015)
A, B, C Parameters of Eq. (1) Eq. (1)
3094e3103.
n,n Stoichiometric numbers Eq. (1) [5] A.Z.J. Panagiotopoulos, Critical parameters in the restricted primitive model,
M1 Molecular weight of solvent Eq. (1) J. Chem. Phys. 116 (2002) 3007e3011.
[6] V. Kobelev, A.B. Kolomeisky, M.E. Fisher, Lattice models of ionic systems,
V Solution volume [m3]
J. Chem. Phys. 116 (2002) 7589e7598.
N Number of solute molecules in a volume V [7] G.L. Aranovich, M.D. Donohue, New approximate solutions to the Ising
Q Ion () charge [Coulomb] Eq. (2) problem in three dimensions, Phys. A 242 (1997) 409e422.
RAB Distance between lattice points A and B [m] Eq. (2) [8] B.E. Conway, Ionic interactions and activity behavior, in: B.E. Conway,
J.O.M. Bockris, E. Yeager (Eds.), Comprehensive Treatise of Electrochemistry,
el Ion ()site-occupancy probability (electrostatic forces vol. 5, Plenum Press, New York, NY, 1983, pp. 111e222.
only) Eq. (2) [9] H.S. Frank, P.T. Thompson, Fluctuations and the limit of validity of the
Dielectric permittivity of solvent [F/m] Eq. (2) DebyeeHckel theory, J. Chem. Phys. 31 (1959) 1086e1094.
[10] D.F. Swinehart, The Madelung constant and crystal lattice energy, in:
<Wel AB > Average interaction energy, pertaining to sites A, B R.C. Weast, M.J. Astle (Eds.), CRC Handbook of Chemistry and Physics, CRC
(electrostatic forces only) [J] Eq. (2) Press, Boca Raton, FL, 1980. Fe242.
<Wel> Average interaction energy, single molecule (electrostatic [11] J.E. Desnoyers, B.E. Conway, Activity coefcients of electrolytes at interme-
diate concentrations and the cube-root law, J. Phys. Chem. 68 (1964)
forces only) [J] Eq. (3) 2305e2311.
R 1/(nN/V) 1/3, mean inter-ionic distance in a lattice-like [12] L.W. Bahe, Structure in concentrated solutions of electrolytes.
conguration [m] Eq. (3) Fieldedielectricegradient forces and energies, J. Phys. Chem. 76 (1972)
1062e1071.
a Madelung Constant Eq. (4) [13] J.F. Chen, G.R. Choppin, Activity coefcients of single electrolytes in concen-
M Ion mass [kg] Eq. (4) trated solutions derived from a quasi-lattice model, J. Sol. Chem. 24 (1995)
vel Ion deterministic velocity (electrostatic forces only) [m/s] 465e484.
[14] L.M. Varela, M. Garcia, F. Sarmiento, D. Attwood, V. Mosquera, Pseudo-lattice
Eq. (4)
theory of strong electrolyte solutions, J. Chem. Phys. 107 (1997) 6415e6419.
vth Ion thermal velocity (Brownian Motion) [m/s] [15] M.A. Marques, M.I.D. Marques, M.I. Cabao, A.M. Gaspar, M.P.M. Marques,
r N/V Eq. (6) A.M. Amado, A.M.A. da Costa, Evidence of a local order in concentrated
Ion cell-occupancy probability Eq. (6) aqueous solutions of salts constituted by ions of different valences. Xeray
diffraction and Raman spectroscopy experiments, J. Mol. Liq. 134 (2007)
UD Debye-Hckel Screening Length [m] Eq. (6) 142e150.
kB Boltzmann Constant Eq. (6) [16] A. Chagnes, B. Carre, P. Willmann, D. Lemordant, Ion transport theory of
 E. Moggia / Fluid Phase Equilibria 427 (2016) 231e245
nonaqueous electrolytes. LiClO4 in butyrolactone: the Quasi Lattice approach,
Electrochim. Acta 46 (2001) 1783e1791.
[17] A. Chagnes, B. Carre, P. Willmann, D. Lemordant, Modeling viscosity and
conductivity of lithium salts in butyrolactone, J. Power Sources 109 (2002)
203e213.
[18] E. Rilo, J. Vila, S. Garcia-Garabal, L.M. Varela, O. Cabeza, Electrical conductivity
of seven binary systems containing 1-ethyl-3-methyl imidazolium alkyl sul-
fate ionic liquids with water or ethanol at four temperatures, J. Phys. Chem. B
117 (2013) 1411e1418.
[19] T. Mendes-Morales, L.M. Varela, Ionic liquids: theory and simulations, in:
A. Perez De Los Rios, F.J. Hernandez-Fernandez (Eds.), Ionic Liquids in Sepa-
ration Technology, Elsevier, Amsterdam, Netherlands, 2014, pp. 70e81.
[20] E. Glueckauf, in: W.J. Hamer (Ed.), The Structure of Electrolyte Solutions,
Wiley, New York, NY, 1959, pp. 97e112.
[21] K.S. Pitzer, Thermodynamics of electrolytes. I. Theoretical basis and general
equations, J. Phys. Chem. 77 (1973) 268e277.
[22] J.C. Rasaiah, A view of electrolyte solutions, J. Sol. Chem. 2 (1973) 301e338.
[23] E. Moggia, B. Bianco, Mean activity coefcient of electrolyte solutions, J. Phys.
Chem. B 111 (2007) 3183e3191.
[24] E. Moggia, Osmotic coefcients of electrolyte solutions, J. Phys. Chem. B 112
(2008) 1212e1217.
[25] E. Moggia, Activity behaviour of electrolyte solutions: evaluation of temper-
ature effects by means of the Quasi-Random Lattice model, Fluid Phase Equil.
286 (2009) 144e153.
[26] E. Moggia, Application of the Quasi-Random Lattice model to rare-earth halide
solutions for the computation of their osmotic and mean activity coefcients,
J. Rare Earths 32 (2014) 979e988.
[27] I.Y. Shilov, A.K. Lyashchenko, The role of concentration dependent static
permittivity of electrolyte solutions in the DebyeeHckel Theory, J. Phys.
Chem. B 119 (2015) 10087e10095.
[28] T. Murphy, L.M. Varela, G.B. Webber, G.G. Warr, R. Atkin, Nanostructure-
thermal conductivity relationships in protic ionic liquids, J. Phys. Chem. B 118
(2014) 12017e12024.
[29] J.G. Kirkwood, F.P. Buff, The statistical mechanical theory of solutions, J. Chem.
Phys. 19 (1951) 774e782.
[30] J.G. Kirkwood, J.C. Poirier, The statistical mechanical basis of the
DebyeeHckel Theory of strong electrolytes, J. Phys. Chem. 58 (1954)
591e596.
[31] J.P. OConnell, A.E. De Glance, Thermodynamic properties of strong electrolyte
solutions from correlation functions, J. Sol. Chem. 4 (1975) 763e778.
[32] P.G. Kusalik, G.N. Patey, The thermodynamic properties of electrolyte solu-
tions: some formal results, J. Chem. Phys. 86 (1987) 5110e5116.
[33] J.P. OConnell, Application of uctuation solution theory to strong electrolyte
solutions, Fluid Phase Equil. 83 (1993) 233e242.
[34] R.R. Netz, H. Orland, Field theory for charged uids and colloids, Europhys.
Lett. 45 (1999) 726e732.
[35] B. Bagchi, Molecular Relaxation in Liquids, Oxford University Press, Oxford,
UK, 2012.
[36] P. Attard, Asymptotic analysis of primitive model electrolytes and the elec-
trical double layer, Phys. Rev. E 48 (1993) 3604e3621.
245
[37] R. Kjellander, R. Ramirez, Yukawa multipole electrostatics and nontrivial
coupling between electrostatic and dispersion interactions in electrolytes,
J. Phys. Cond. Matter 20 (2008) 1e20.
[38] D. Fraenkel, Simplied electrostatic model for the thermodynamic excess
potentials of binary strong electrolyte solutions with size-dissimilar ions, Mol.
Phys. 108 (2010) 1435e1466.
[39] B. MariboeMogensen, G.M. Kontogeorgis, K. Thomsen, Comparison of the
DebyeeHckel and the mean spherical approximation theories for electrolyte
solutions, Ind. Eng. Chem. Res. 51 (2012) 5353e5363.
[40] D. Rowland, P.M. May, Thermodynamics of strong aqueous electrolyte solu-
tions at T 25  C described by the Hckel equations, J. Chem. Eng. Data 59
(2014) 2030e2039.
[41] E. Moggia, B. Bianco, Closed-form expression for the potential within a face
centred cubic ionic crystal, J. Electrost. 61 (2004) 269e280.
[42] R.M. Pytkowicz, K. Johnson, C. Curtis, Long-range order model of aqueous
electrolyte solutions, Geochem. J. 11 (1977) 1e7.
[43] R.E. Crandall, New representations for the Madelung constant, Exp. Math. 8
(1999) 367e379.
[44] L. Glasser, Solid-state energetics and electrostatics: Madelung constants and
Madelung energies, Inorg. Chem. 51 (2012) 2420e2424.
[45] V.M.M. Lobo, Handbook of Electrolyte Solutions, Elsevier Science, New York,
NY, 1990.
[46] W.R. Fawcett, Liquids, Solutions, and Interfaces from Classical Macroscopic
Descriptions to Modern Microscopic Details, Oxford University Press, Oxford,
UK, 2004.
[47] I.B. Malham, P. Letellier, A. Mayaffre, M. Turmine, Part I: thermodynamic
analysis of volumetric properties of concentrated aqueous solutions of 1-
butyl-3-methylimidazolium tetrauoroborate, 1-butyl-2,3-dimethylimidazo-
lium tetrauoroborate, and ethylammonium nitrate based on pseudo-lattice
theory, J. Chem. Thermodyn. 39 (2007) 1132e1143.
[48] E. Rilo, M. Domnguez-Pe rez, L.M. Varela, O. Cabeza, Experimental evidence of
the validity of BaheeVarela theory to describe the volumetric properties of
ionic liquids, J. Chem. Thermodyn. 54 (2012) 68e75.
[49] D.P. Fernandez, A.R.H. Goodwin, E.W. Lemmon, J.M.H. Levelt-Segers,
R.C.A. Williams, Formulation for the static permittivity of water and steam at
temperatures from 238 K to 873 K at pressures up to 1200 MPa, including
derivatives and DebyeeHckel Coefcients, J. Phys. Chem. Ref. Data 25 (1997)
1125e1139.
[50] P. Wang, K.S. Pitzer, J.M. Simonson, Thermodynamic properties of aqueous
magnesium chloride solutions from 250 to 600 K and to 100 MPa, J. Phys.
Chem. Ref. Data 27 (1998) 971e991.
[51] S. Al Ghafri, G.C. Maitland, J.P.M. Trusler, Densities of aqueous MgCl2(aq),
CaCl2(aq), KI(aq), NaCl(aq), KCl(aq), AlCl3(aq), and (0.964 NaCl 0.136
KCl)(aq) at temperatures between 283 and 472 K, pressures up to 68.5 MPa,
and molalities up to 6 mol$kg-1, J. Chem. Eng. Data 57 (2012) 1288e1304.
[52] K.S. Pitzer, J.C. Peiper, R.H. Busey, Thermodynamic properties of aqueous so-
dium chloride solutions, J. Phys. Chem. Ref. Data 13 (1984) 1e102.
[53] D.G. Archer, Thermodynamic properties of the NaClH20 system. II. Ther-
modynamic properties of NaCl (aq), NaCl.2H20 (cr), and phase equilibria,
J. Phys. Chem. Ref. Data 21 (1992) 793e829.