Gas Solubility in Dilute Solutions: A Novel Molecular Thermodynamic Perspective

Gas solubility in dilute solutions: A novel molecular thermodynamic perspective
Ariel A. Chialvo
Citation: The Journal of Chemical Physics 148, 174502 (2018); doi: 10.1063/1.5023893
View online: https://doi.org/10.1063/1.5023893
View Table of Contents: http://aip.scitation.org/toc/jcp/148/17
Published by the American Institute of Physics
Articles you may be interested in

Bulk viscosity of molecular fluids
The Journal of Chemical Physics 148, 174504 (2018); 10.1063/1.5022752
Time correlation functions of simple liquids: A new insight on the underlying dynamical processes
Comprehensive representation of the Lennard-Jones equation of state based on molecular dynamics

simulation data
Communication: Glass transition and melting lines of an ionic liquid

High-density equation of state for a two-dimensional Lennard-Jones solid

Orientational ordering of lamellar structures on closed surfaces

THE JOURNAL OF CHEMICAL PHYSICS 148, 174502 (2018)
Gas solubility in dilute solutions: A novel molecular

thermodynamic perspective
Ariel A. Chialvoa)
Knoxville, Tennessee 37922-3108, USA
(Received 29 January 2018; accepted 9 April 2018; published online 2 May 2018)
We present an explicit molecular-based interpretation of the thermodynamic phase equilibrium under-

lying gas solubility in liquids, through rigorous links between the microstructure of the dilute systems
and the relevant macroscopic quantities that characterize their solution thermodynamics. We apply the
formal analysis to unravel and highlight the molecular-level nature of the approximations behind the
widely used Krichevsky-Kasarnovsky [J. Am. Chem. Soc. 57, 2168 (1935)] and Krichevsky-Ilinskaya
[Acta Physicochim. 20, 327 (1945)] equations for the modeling of gas solubility. Then, we imple-
ment a general molecular-based approach to gas solubility and illustrate it by studying Lennard-Jones
binary systems whose microstructure and thermodynamic properties were consistently generated via
integral equation calculations. Furthermore, guided by the molecular-based analysis, we propose a
novel macroscopic modeling approach to gas solubility, emphasize some usually overlook modeling
subtleties, and identify novel interdependences among relevant solubility quantities that can be used
as either handy modeling constraints or tools for consistency tests. Published by AIP Publishing.
https://doi.org/10.1063/1.5023893
I. INTRODUCTION Since the study of gas solubility entails dealing with

the isothermal-isobaric composition of binary systems involv-
Thermodynamics of dilute solutions has been the subject ing a gaseous solute at equilibrium with its dilute coun-
of considerable theoretical and practical interest 1–4 closely terpart in a condensed solvent phase, we will analyze the
linked to the molecular theory of fluid mixtures and conse- conditions of vapor-liquid equilibrium in terms of both the
quently to the development of regression approaches for the species residual and the resulting system excess properties.
macroscopic modeling of experimental data associated with The rationale motivating this approach hinges around the
chemical engineering,5–7 bio-chemical,8,9 and geo-chemical advantages these two types of quantities encompass in the
processes.10 Among this wealth of theoretical development interpretation of the link between the (microscopic) inter-
and experimental data, there has been (understandably) room molecular interactions and the resulting (macroscopic) ther-
for confusion regarding the meaning of solvation quantities modynamic behavior.14,18 In fact, as lucidly discussed by
and processes, or as Ben-Naim argued, “between solvation Abbott and Nass,19 it is typically not practical to deal
thermodynamics and conventional standard thermodynamics directly with a solution molar property M (TPx) in the ther-
of solution” when proposing a rigorous analysis to identify modynamic description of a system, but rather its deviation
and define precisely the standard free energy of solution molar counterpart ∆M (TPx) ≡ Mreal (TPx) − Mmodel (TPx)
according to the so-called C (mole fraction, molarity, and defined as the difference between the actual (real) and a pre-
molality)-process of solvation.11 cisely defined model behavior at identical state conditions
In the quest for accurate engineering modeling of dilute (e.g., temperature and pressure) and composition (e.g., mole
solutions, we typically must confront two interconnected real- fractions).
ities, namely, (i) the required highly precise experimental To make the microscopic-macroscopic link both explicit
determination of thermodynamic data12 and (ii) the need for and unambiguous, we invoke here two types of molar devi-
fundamentally based and molecularly inspired thermodynamic ation quantity ∆M (TPx), namely, the residual M R (TPx)
models.2,13–15 In this work, we will focus on the second real- ≡ Mreal (TPx) − MIG (TPx) and the excess M E (TPx)
ity; specifically, in pursuit of unambiguous molecular-based ≡ Mreal (TPx) − MIS (TPx) properties. The first of these rep-
description of the phase equilibrium equations for the sol- resents the deviation from the ideal gas (i.e., null interactions
ubility of sparingly soluble gases in liquids, we provide a among species) behavior, while the second is the deviation
microscopic interpretation of and highlight the molecular- from an ideal solution reference (i.e., described by a precisely
level nature behind the approximations underlying the widely defined model containing the ideal gas plus assimilation con-
used Krichevsky-Kasarnovsky16 and Krichevsky-Ilinskaya17 tributions due to species distinguishability20 ). Thus, M R (TPx)
gas solubility correlations. measures the contribution of the actual interaction forces to
the property M (TPx), while M E (TPx) assesses the contri-
bution from the differences between species intermolecular
a)Author to whom correspondence should be addressed: [email protected] interactions (i.e., molecular asymmetry) in solution.
0021-9606/2018/148(17)/174502/16/$30.00 148, 174502-1 Published by AIP Publishing.

174502-2 Ariel A. Chialvo J. Chem. Phys. 148, 174502 (2018)
While the choice for the model reference might be some- (OZ) equation. Supported by the IE results of Sec. IV, we pro-
what arbitrary, the Lewis-Randall ideal solution21,22 offers us pose in Sec. V a novel macroscopic modeling approach to gas
two practical advantages: it allows us (i) to describe the ther- solubility guided by the molecular-based analysis developed
modynamic excess properties as differences of residual proper- in this work. Furthermore, in Sec. VI, we identify novel inter-
ties between the actual solution and each of its pure component dependences among relevant solubility quantities and discuss
counterparts (as well as their temperature derivatives23 ) and their modeling implications. Finally, we close our work with
consequently (ii) to identify the explicit connections between a summary of findings and future outlook.
(the magnitude and sign of) the molar excess properties and
the differences (in nature and strength18,24 ) of the intermolec- II. THERMODYNAMICS OF GAS SOLUBILITY
ular forces between species in solution. These two features are
A. Phase equilibrium fundamentals
essential for the versatile and successful modeling of solva-
tion phenomena either over the entire composition range25 or Here we focus our attention on the solubility of a pure
involving dilute multicomponent systems regardless of their gas in a pure liquid at a given temperature T and pressure
aggregation state and including highly compressible (near P and study the thermodynamic quantities that characterize
critical) media.3,14,26 the equilibrium solubility of the gaseous solute in the pres-
As we have discussed extensively elsewhere,14,15 novel ence of the vapor–liquid coexisting phases. After invoking the
separation processes on the one hand depend on our ability to general criterion of physicochemical phase equilibrium, i.e.,
tune the solvation behavior of species in solution according to µLi (TP {x}) = µVi (TP {y}), where µαi (· · · ) denotes the chemi-
required applications including food and pharmaceutical pro- cal potential of the i-solute species in the α-phase characterized
cessing,6,8,27 capture of anthropogenic gases,7,28 water reform- by the thermodynamic conditions (· · · ), we can rewrite it fol-
ing and chemical synthesis,29 and novel materials preparation lowing the (φ − γ) method21,31 to describe the vapor (V ) and
and characterization.5 On the other hand, all of these processes liquid (L) phases, respectively, i.e.,
either take place in or involve media with simultaneous solva-
f̂iV (TPyi ) = f̂iL (TPxi ) ,
tion of gases, non-polar and ionic species whose behavior can (1)
only be interpreted, and consequently described, by macro- yi φ̂Vi (TPyi ) = xi φ̂Li (TPxi ) = xi fio (TP) γiLR,L (TPxi )/P.
scopic correlations through a fundamental understanding of
The variables yi and xi in Eq. (1) represent the mole fraction of
the solvation processes in terms of microstructural changes
the i-solute species in the vapor and liquid phase, respectively,
undergone by the fluid environment.14
and φ̂Vi (TPyi ) and φ̂Li (TPxi ) are the corresponding phase equi-
Within this context, our first goal is the explicit molecular-
based interpretation of the thermodynamic phase equilibrium librium partial molar fugacity coefficients, while γiLR,L (TPxi )
equations underlying gas solubility in liquids obtained by denotes the Lewis-Randall activity coefficient at the prevailing
drawing unambiguous links between the microstructure30 of state conditions and composition.
the system and the relevant macroscopic quantities that char- Typically, the behavior of the i-solute in the vapor phase
acterize the resulting solution thermodynamics. As a con- is accurately characterized by an equation of state (EoS) such
sequence of this analysis, we will highlight the molecular- as Peng-Robinson,32 while that for the liquid phase involves a
level nature of the approximations used in the Krichevsky- solution model based on Henry’s law33 for the ideal solution
Kasarnovsky16 and Krichevsky-Ilinskaya17 equations for the reference, i.e.,
modeling of gas solubility. Then, our second goal is to tackle fiHL (TPxi ) = xi HIS
i,j (TP) , (2)
the implementation of a general molecular-based approach to
gas solubility for real systems and its illustration by modeling where HIS i,j (TP) represents Henry’s law constant of the i-
a binary system with interactions described by the Lennard- solute in the j-solvent at the (TP) state conditions, with the
Jones pair potential, whose microstructure and thermodynamic superscripts IS and HL denoting the condition of ideal solu-
properties are consistently generated via integral equation (IE) tion under the reference state given by Henry’s law. Thus,
calculations. the actual behavior of the i-solute in the j-solvent becomes
To pursue these goals, we first discuss in Sec. II the fun- described by the deviation from the behavior represented by
damental equations underlying the isothermal-isobaric vapor- Eq. (2) and accounted for the activity coefficient based on the
liquid equilibrium associated with gas solubility in conjunction corresponding IS reference as follows:
with the rigorous molecular-based solvation formalism for f̂iL (TPxi ) = γiHL (TPxi ) xi HIS
i,j (TP) , (3)
dilute solutions proposed previously26 to provide microscopic
where we identify explicitly Henry’s law ideal solution through
understanding of the macroscopic counterparts. In Sec. III,
the superscript HL, though the literature usually involves the
we invoke the proposed gas solubility analysis to perform
notation γiHL (TPxi ) ≡ γi∗ (TPxi ) for that purpose. Conse-
the microscopic interpretation of the Krichevsky-Kasarnovsky
quently, from Eq. (1), we have that
and Krichevsky-Ilinskaya equations for the modeling of gas f g f g
solubility, highlighting the molecular meaning of their approx- lim f̂iL (TPxi )/xi = lim f̂iV (TPyi )/xi
xi →0 T xi →0 T
imations. Moreover, to assess the relative contributions of the
individual quantities affecting the phase equilibrium, we study = HIS
i,j (TP) . (4)
in Sec. IV a well-characterized model, the Lennard-Jones sys- Considering that lim P (T ) = Ps (T ) = Ps,j (T ) ≡ Ps and that
tem, via IE calculations according to the Percus-Yevick (PY) xi →0
approximation for the solution of the atomic Ornstein-Zernike lim γiHL (T Ps xi ) = 1, then for tabulation purposes it becomes
xi →0
convenient to define Henry’s law constant at a reference pres- origin, though we should recognize that, because the tempera-
sure, whose natural choice is that of the j-solvent saturation ture dependence of the solvent saturation pressure,Ps (T ), we
pressure Ps (T ) at the (subcritical) temperature of the system; actually have φ̂∞ i (T ) and kij (T ) along the orthobaric curve.
in other words, Therefore, from Eqs. (9)–(11), the equilibrium partial molar
P f g ! fugacity of the dilute i-solute in the liquid phase becomes
HIS
i,j (TP) = HIS
i,j (T Ps ) exp υ̂ ∞,L
(TP)/RT dP , (5) P f g !
i
∞,L
Ps (T ) f̂iL (TPxi ) = xi HIS
i,j (T P s ) exp υ̂i
(TP)/RT dP
where the exponential term is usually known as the Poynting Ps (T )
" ( P )
correction and (Ref. 4)
× exp − ∂kij /∂P dP + kij (T Ps )
P (f Ps (T )
T
γi (TPxi ) = γi (T Ps xi ) exp
HL HL
υ̂iL (TPxi ) #
Ps (T )
! × xi − 0.5xi2 . (12)
g )
− υ̂i∞,L (TP) /RT dP . (6) Alternatively, by noting that the {· · · }-term in the second
exponential of Eq. (12) reduces to
Then, from Eqs. (3)–(6), we can rewrite Eq. (3) as a function of ( P )
the saturation pressure and the resulting Poynting correction
∂kij /∂P dP + kij (T Ps ) = kij (TP) , (13)
term as follows: T
Ps (T )
f̂iL (TPxi ) = xi γiHL (TPxi ) HIS

i,j (TP)
we then get the following equivalent expression to Eq. (12):
P f
= xi γiHL (T Ps xi ) HIS
i,j (T Ps )
g
f̂iL (TPxi ) = xi HIS (T Ps ) exp υ̂i∞,L (TP)/RT dP
P f g !
i,j
Ps (T )
× exp υ̂iL (TPxi )/RT dP , (7) !
Ps (T ) − kij (TP) xi − 0.5xi2 . (14)
where we( highlight the ) cancellation of the term
exp ∫ PPs (T ) υ̂i∞,L (TP)/RT dP between the isothermal pres- The alternative expressions Eqs. (12) and (14) for the liquid-
phase partial molar fugacity of the dilute i-solute in vapor-
sure dependence of γiHL (TPxi ) and HIS i,j (TP) in Eqs. (6) and liquid equilibrium are based on the self-consistent second-
(3), respectively.
order composition dependence of the thermodynamic par-
At this point, we invoke the relation between the com-
tial molar quantities derived and discussed extensively else-
position dependence of the activity coefficients for the i-
where.26,34
solute describing the deviations from Henry’s law and the
Lewis-Randall rule,31 i.e., B. Molecular-based interpretation of relevant gas
solvation quantities
γiHL (TPxi ) = γiLR (TPxi )/γiLR (TPxi = 0)
= γiLR (TPxi )/γiLR, ∞ (TP). (8) To improve our understanding of the modeling capabil-
ities of the above macroscopic expressions, we now invoke
Now, by recalling the definition of the Lewis-Randall activ- Kirkwood-Buff’s fluctuation formalism of mixtures35 to make
ity coefficient as the ratio between the partial molar fugacity the interpretation of the relevant quantities associated with gas
coefficient of the i-solute at the prevailing state conditions and solvation by linking them with the behavior of the evolution
composition and the corresponding pure component counter- of the microstructure of the resulting dilute solutions.14,15
part, i.e., γiLR (TPxi ) = φ̂i (TPxi )/φoi (TP), we can recast Eq. (7) The quantities of interest in this context are kij (TP),
g its
after introducing Eq. (8) in the following alternative form:
f
pressure and temperature derivatives, i.e., ∂kij (TP)/∂P and
f g f g T
f̂iL (TPxi ) = xi φ̂i (T Ps xi )/φ̂∞
i (T Ps ) HIS
i,j (T Ps ) ∂kij (TP)/∂T , as well as υ̂i∞ (TP). We have already shown
P
P f g ! that the coefficient kij (TP) in the second-order truncated com-
× exp υ̂i (TPxi )/RT dP .
L
(9) position expansion of the partial molar fugacity coefficients
Ps (T ) of the species in a dilute binary mixture can be identified
The obvious appeal of Eq. (9) lies in the ratio microscopically by the following expression:3,26
φ̂i (T Ps xi )/φ̂∞
i (T Ps ) = f (xi ) where, because we are dealing

kij (TP) = G∞ ii + Gjj − 2Gij /υj ,
o ∞ o
(15)
with dilute solutions, f (xi ) can be expressed as a second-
order isothermal-isobaric composition expansion of φ̂i (T Ps xi ) where υjo ≡ ρoj (TP)−1 describes the molar volume of the pure
around the infinite dilution condition φ̂∞ i (T Ps ),
26,34 i.e.,
solvent at the prevailing state conditions and Gαβ ⊗
denotes the
Kirkwood-Buff total correlation function integral, TCFI,35 for
f g
f (xi ) = exp −kij (T Ps ) xi − 0.5xi2 , (10)
the α β-interactions at the composition condition of either pure
whose consequent fixed-composition isothermal-pressure component, ⊗ = o, or infinite dilution, ⊗ = ∞, i.e.,
derivative becomes26 ∞
Gαβ (TPx) = 4π
⊗ ⊗
hαβ (TPx, r)r 2 dr. (16)

υ̂i (TPxi ) = υ̂i∞ (TP) − RT ∂kij /∂P xi − 0.5xi2 . (11) 0
T
Note that we kept the notation (T Ps ) as the variable for The integrand in Eq. (16) comprises the radial total pair cor-
φ̂∞
i (T Ps ) and kij (T Ps ) in Eqs. (10) and (11) to highlight their
⊗
relation function hαβ ⊗
(TPx, r) ≡ gαβ (TPx, r) − 1, where the
corresponding radial pair distribution function gαβ ⊗

(TPx, r) where κ oIG
j = υjo /kT is the ideal gas solvent compressibility.
characterizes the (short- plus long-range) microstructure of the For all practical purposes, we can recast kijsolvation (TP) and
system, i.e., the environment of β-species around α-species at kijκ (TP) in terms of measurable macroscopic quantities, e.g.,
the state conditions and composition (TPx).
By definition, the total pair correlation function 2
∞ (TP, r) of an infinitely dilute mixture includes direct pair
hαβ kijsolvation (TP) = κ oj κ oIG
j /υ o
j (∂P/∂x ) ∞
i Tυ o , (20)
j
correlations associated with the perturbation of the α-solvent kijκ (TP) = − ∂ ln φ̂i /∂xj
∞
− κ oj κ oIG

j /υj
o
structure caused by the presence of the infinitely dilute β- P,T
2
solute as well as indirect pair correlations associated with
× (∂P/∂xi )∞ Tυ o , (21)
the propagation of the structural perturbation. This feature, j
extensively discussed elsewhere,3,14,36,37 allows the segrega- where (∂P/∂xi )∞

⊗ Tυjo —the finite isochoric pressure perturbation
tion between two sources of contributions to Gαβ and conse-
upon addition of an infinitely dilute i-solute into an other-
quently to kij (T Ps ), namely, the solvation process (short-range
wise pure j-solvent—can also be unambiguously described
direct pair correlations) and the propagation through the sol-
and calculated as either an isothermal perturbation of the solute
vent of the structural perturbation (a distance given by its cor-
partial molar volume or solvent molecular population around
relation length) that scales with the solvent’s isothermal com- ∞
pressibility. The short-/long-range split becomes rather handy the solute,13,37,39 with the limiting slope ∂ ln φ̂i /∂xj
∞ P,T
especially when dealing with highly compressible media (e.g., = ∂ ln γiLR /∂xj available from accurate methods.40
P,T
dilute near critical solutions) because it allows the unambigu- The partial molar volume of the solute species at infinite
ous interpretation of the solvation phenomena, while avoiding dilution, υ̂i∞ (TP), is typically assumed pressure independent
the interference of diverging compressibility-driven quanti- in the Poynting correction during the evaluation of the solute
ties. In fact, we have shown that we can split the TCFI’s of an fugacity, e.g., Eq. (7),31,41,42 a conjecture that can introduce
infinitely dilute binary solution into the direct correlation func- significant uncertainties on the calculated quantities in the
tion integrals (DCFI’s), comprising the short-range interaction sequential approach.31 In order to assess the accuracy of such
contributions, and the indirect correlation function integrals an approximation, we must analyze the molecular origin of
(ICFI’s), measuring the long-range interaction contributions, the underlying pressure dependence of υ̂i∞ (TP), and for that
i.e., purpose, we invoke its expression from Kirkwood-Buff’s fluc-
Gojj = Cjjo + κ oj kTCjjo2 /υjo2 , tuation formalism in terms of either TCFI’s or DCFI’s as
follows:3,39
ij = Cij + κ j kTCij Cjj /υj ,
G∞ ∞ o ∞ o o2
(17)
G∞
ii = Cii∞ + κ oj kTCij∞2 /υjo2 , υ̂i∞ (TP) = υ̂jo + Gojj − G∞ij , (22a)

where j-species denotes the solvent, κ oj ≡ βυjo2 / υjo − Cjjo
υ̂i (TP) = υj − Cij / 1 − ρoj Cjjo ,
∞ o ∞
(22b)

is the isothermal compressibility of the pure solvent, and υ̂i∞ (TP) = υ̂jo + Cjjo − Cij∞ + κ oj kTCjjo Cjjo − Cij∞ /υjo2 , (22c)
Cij⊗ represents the integral of the direct correlation function | {z } | {z }
υ̂i∞ (SR) υ̂i∞ (LR)
cij⊗ (TPx, r), i.e.,
∞ where we can distinguish in principle two sources of
Cij (TPx) = 4π
⊗
cij⊗ (TPx, r)r 2 dr, i, j = 1, 2, (18) pressure dependence, i.e., one coming from the pure j-
0 solvent and the other from the infinite dilute i-solute
which is the short-range counterpart to Gij⊗ according to the interacting with its surrounding j-solvent environment. In
Eq. (22c), SR and LR indicate the short-range (solvation)
Ornstein-Zernike equation.38 and long-range (compressibility-driven) contributions to the
According to Eqs. (16)–(18), we can now provide a quantity.3,36
microstructural interpretation for the coefficient kij (TP) in The other relevant solvation quantity is the finite iso-
Eq. (15) through the fundamental link between the finite local choric pressure perturbation upon addition of an infinite dilute
solvent density perturbation caused by the presence of the i-solute into an otherwise pure j-solvent,3 i.e.,
solute—expressed as an isochoric-isothermal pressure change
(∂P/∂xi )∞Tυjo
37,39 —and the range of its propagation that scales
(∂P/∂xi )∞ o
Tυj = ρo2
j kT G o
jj − G ∞
ij / 1 + ρo o
G
j jj
as the isothermal compressibility of the pure solvent, κ oj .
= ρj kT Cjj − Cij ,
o2 o ∞
(23)
Therefore, by invoking Eqs. (17), we can also split kij (TP),
Eq. (15), into kijsolvation (TP) and compressibility driven kijκ (TP) which can also be unambiguously described as an isothermal
contributions, i.e., perturbation of either the solute partial molar volume or sol-

∞ 2
vent molecular population around the solute.13,37,39 In fact,
kij (TP) = Cjjo + Cii∞ − 2Cij∞ /υjo + κ oj /κ oIG
j υ o
j C o
jj − Cij Eq. (23) provides a venue to identify three reference solute-
solvent interaction asymmetries and assess the behavior of

= Cjjo + Cii∞ − 2Cij∞ /υjo + κ oj κ oIG
j /υ o
j
2 (∂P/∂xi )∞Tυjo involving an infinitely dilute solute with decreas-
× (∂P/∂xi )∞ Tυ o ing solute-solvent interaction asymmetry—namely, (a) a real
j
solute in a Lewis-Randall ideal solution, (b) a real solute behav-
= kijsolvation (TP) + kijκ (TP) , (19) ing identically to the solvent, and (c) an ideal gas solute in
a real solution as discussed in detail in Appendix A. As we

will discuss in Sec. II C, (∂P/∂xi )∞
Tυjo plays a central role in
the microscopic interpretation of the sources of the pressure
dependence of υ̂i∞ (TP).
C. Molecular-based inspired macroscopic correlations

The macroscopic thermodynamic counterpart of the
molecular interpretation of υ̂i∞ (TP) and (∂P/∂xi )∞
Tυ o can bej
written in a revealing form as follows:3,39

υ̂i∞ (TP) = υjo 1 + κ oj (∂P/∂xi )∞
Tυ o , (24)
j
so that after invoking Eq. (23) and the first term of Eq. (22c)
FIG. 1. Isothermal-pressure dependence of the thermodynamic quantities
we obtain β ∫ PP υ̂i∞ dP ∞ (P−P )
(∂P/∂xi )∞ , υ̂ ∞ , υjo , and κjo (left axis) as well as e
T υo i
s and e β υ̂i s
Tυ o = ρj kT υ̂i (SR) − υj ,
o2 o
(∂P/∂xi )∞ ∞
(25) j
(right color-coordinated axes) for the H2 (i)–Ar(j) system as described by the
j
integral equation calculations involving the Lennard-Jones model along the
where the link between υ̂i∞ (TP) and (∂P/∂xi )∞ Tυjo can be T = 100 K isotherm.
pursued in terms of the difference of partial molar volumes
between the i-solute at infinite dilution, υ̂i∞ (SR), and the pure highlight the small negative departure of the actual Poynt-
j-solvent, υ̂jo .3,36 ing correction from the approximated value when assuming
Equations (24) and (25) highlight immediately three fac- υ̂i∞ (TP) = υ̂i∞ (T Ps ). For a common solvent, the observed
tors contributing to the isothermal pressure dependence of compensating behavior will depend squarely on the mag-
υ̂i∞ (TP), where two of them are properties of the pure solvent. nitude of the (∂P/∂xi )∞ Tυjo , and consequently, the accuracy
We note that, according to the semi-empirical Tait equation,43 of the conjectured incompressible υ̂i∞ (TP) in the Poynt-
κ oj (P)T exhibits an inverse proportionality with pressure so that ing correction will be a direct manifestation of the balance
f g of the above two competing contributions. For example,
κ oj (TP) = κ oj (T Po )/ 1 + nκ oj (T Po ) P , (26) after substituting Eqs. (26), (27), and (29) (based on Tait’s
g 1/n
equation) into Eq. (24), fwe obtain the resulting
g Poynting
f
υjo (TP) = υjo (T Po )/ 1 + nκ oj (T Po ) P , (27) P ∞,L
integral PC = exp ∫ Ps (T ) υ̂i (TP)Tait /RT dP for repre-
where P = P − Po , Po is a reference pressure [e.g., Po sentative values of υjo (T Po ) and κ oj (T Po ) whose behav-
= Ps,j (T )], and 3 . n . 15. Thus, according to Eqs. (26) ior is illustrated in Fig. 2. This plot provides clear evi-
and (27), the pure j-solvent quantities contribute significantly dence of the significant contribution of the Poynting cor-
to the pressure dependence υ̂i∞ (TP). In addition, after invok- rection to Henry’s law constant as (P − Ps )T increases
ing O’Connell et al.’s correlation for simple solutes13 for the not only from the magnitude of υ̂i∞ (T Ps ) but also
difference of DCFI’s in the right-hand side of Eq. (23) as from the actual isothermal pressure dependence υ̂i∞ (TP).
follows: f g In fact, the resulting PC (P)T for 39.6 ≤ υ̂i∞ (T Ps ) cm3 /mol
Cjjo − Cij∞ = a + b exp c ρoj − 1 , (28) ≤ 84.6 exhibits a quadratic rather than exponential pres-
we can piece together the sought isothermal pressure depen- sure, dependence in the Ps (T ) ≤ P (atm) ≤ 110 interval,
dence associated with the solute-solvent interactions, i.e.,
f g 2/n
Tυjo = kT ρj (T Po ) 1 + nκ j (T Po ) P
o2 o
(∂P/∂xi )∞
f
× a + b exp c ρoj (T Po )
f g 1/n
× 1 + nκ oj (T Po ) P − 1 . (29)
The analysis based on the IE calculations (vide infra

Sec. IV) of the isothermal pressure dependence of the three
factors in Eq. (24), i.e., Eqs. (26)–(28) for the H2 (i) − Ar(j)
system, result in n 9.056, a 5.4606, b 0.9831, and
c 0.78 306 indicating that there are two competing contri-
butions to the isothermal-pressure dependence of υ̂i∞ (TP).
The first contribution originates in the solute-solvent inter-
action asymmetry, i.e., encapsulated in (∂P/∂xi )∞ Tυjo , while FIG.f 2. Isothermal-pressure dependence of the Poynting correction
the other comes from the pure solvent and results in an PC TP, υ̂i∞,L (TP)Tait with representative values of υjo (T Po ) and κjo (T Po )
g
opposite pressure trend (i.e., they scale as either P(−1/n) for infinitely dilute Lennard-Jones systems as described by integral equation
or P[−(1+n)/n] ) as clearly depicted in Fig. 1, where we can calculations along the T = 100 K isotherm.
i.e., an indication of the non-negligible pressure rigorous result only for systems obeying the Lewis-Randall
effect on υ̂i (TP), while PC P = 100, υ̂i (T Ps ) increases
∞ ∞ ideality (vide infra Appendix A), i.e., those for which
T
by a factor of 1.25. Under these conditions,
P f̂i (TPxi ) = fio (TP) xi → γiLR (TPxi ) = 1, ∀xi ,
the departure ∆PC = exp ∫ Ps (T ) υ̂i∞,L (TP)Tait dP/RT
f g
− exp υ̂i∞,L (T Ps )Tait (P − Ps )/RT will amount to 7 < |∆PC | υ̂i (TPxi ) = υ (TPxi ) + xj Gjj − Gij → υ̂i∞ (TP)
(%) < 12.
IS (33)
= υjo (TP) + Gojj − G∞ij ,
f g
Hi,j (T Ps ) = lim f̂i (TPxi )/xi = fi (T Ps ) .
IS o
III. MACROSCOPIC MODELING OF GAS SOLUBILITY xi →0
AS CONVENTIONALLY INTERPRETED
AND IMPLEMENTED
Obviously, this assumption is a rather restrictive condition for
the gas solubility behavior in real systems, one that constrains
At this point, we have a rigorous thermodynamic frame- the potential use of the Krichevsky-Kasarnovsky equation to
work, complemented with the statistical mechanical interpre- systems whose solute exhibits extremely small deviation from
tation of the relevant solvation quantities, for the rational the Lewis-Randall ideality, i.e., to those whose solute-solvent
analysis of existing correlations widely used in the study interaction asymmetry is negligibly small.
of gas solubility, i.e., the Krichevsky-Kasarnovsky and the
Krichevsky-Ilinskaya equations and variations of them.21,31,41 B. Special case 2: Krichevsky-Ilinskaya equation
In what follows, we identify the embedded assumptions As highlighted by the derived equation (12), the full and
associated with these modeling expressions, discuss their general description of the partial molar fugacity of the i-
microscopic consequences, and highlight some overlooked solute in the condensed phase requires the knowledge of the
modeling subtleties. pressure and composition dependences of the i-solute partial
molar fugacity, φ̂i (TPxi ), or Henry’s law activity coefficient,
A. Special case 1: Krichevsky-Kasarnovsky equation
γiHL (T Ps xi ), to account for the system pressure and composi-
The composition and pressure dependence of the par- tion dependences in the second exponential term of Eq. (12).
tial molar fugacity coefficient (or its activity counterpart) Unfortunately, such information is rarely available; hence, the
becomes frequently blurred by the inaccuracies of the exper- regression of solubility measurements frequently proceeds via
imental measurements following the condition of high dilu- conjectured approximations such as υ̂iL (T Ps xi ) υ̂i∞,L (T Ps )
tion of the system under consideration. Consequently, it is in the Krichevsky-Kasarnovsky equation as already discussed
frequently assumed that γiLR (T Ps xi ) γiLR,∞ (T Ps ), i.e., above. The additional factor in the second exponential terms
γiHL (T Ps xi = 0) = 1 with υ̂iL (T Ps xi ) υ̂i∞,L (T Ps ) indepen- of Eq. (12) encompasses the composition dependence of the
dent of the system pressure so that Eq. (7) reduces to isobaric-isothermal partial molar fugacity (or its associated
f g Henry’s law activity) coefficient for the i-solute in the con-
∞,L
i,j (T Ps ) exp (P − Ps ) υ̂i
f̂iL (TPxi ) xi HIS (T Ps )/RT , (30) densed phase. In principle, according to the specialized litera-
ture including Refs. 41, 42, and 44, the Krichevsky-Ilinskaya
which is known in the literature as the Krichevsky-
equation accounts for the non-ideality originated in the solute-
Kasarnovsky equation of gas solubility.16
solvent interactions (vide infra Appendix B) via a simple
According to the microscopic interpretation of Eq. (12)
two-suffix Margules (or Porter), ln γiHL = A (TP) xj2 − 1 , i.e.,
derived above, the validity of the Krichevsky-Kasarnovsky
Eq. (12) simplifies to the familiar expression
equation as a modeling tool hinges around the condition
kij (T Ps ) = ∂kij /∂P = 0 within its second exponential,
f ∞,L
T
f̂iL (TPxi ) = xi HIS
i,j (T Ps ) exp (P − Ps ) υ̂i (T Ps )/RT
whose microstructural manifestation according to Eq. (15) is
2
g
represented by the following two conditions: + A (T Ps ) xj − 1 (34)
known in the literature as the Krichevsky-Ilinskaya equa-
ii + Gjj − 2Gij = 0,
o
∆ij (T Ps ) ≡ G∞ ∞
(31)
tion for the ij-binary system.17 If we express the Mar-
∂∆ij /∂P = 0. (32) gules/Porter equation in terms of the i-solute
mole fraction, i.e.,
T
A (TP) xj2 − 1 = −2A (TP) xi − 0.5xi2 , where A ≡ ln γiLR,∞
These microscopic conditions indicate that the solubility
(vide infra Appendix B), then the direct comparison between
of a gas in a binary solution would obey the Krichevsky-
Eqs. (34) and (12) allows us to identify 2A (T Ps ) = kij (T Ps )
Kasarnovsky equation if the system behaved as a Lewis- and provide a meaningful molecular-based interpretation for
Randall ideal solution,15 i.e., G∞ ij = 0.5 G ∞ + Go , not only at
ii jj the Margules/Porter coefficient A (T Ps ) in this type of dilute
the solvent’s saturation pressure Ps (T ) but also at any system system as follows:
pressure along the chosen isotherm. f g
ii + Gjj − 2Gij /υj
o o
Note that the microscopic expression Eq. (31) represents A (T Ps ) ≡ 0.5 G∞ ∞
T Ps
even a more restrictive constraint than that in the tradition-
= 0.5kij (T Ps ) . (35)
ally invoked condition γiHL (T Ps xi ) 1, in the conventional
derivation of the Krichevsky-Kasarnovsky equation; that is, Note that the Krichevsky-Ilinskaya equation as conventionally
Eq. (31) tells us that not only γiHL (T Ps xi = 0) = 1 but also written, Eq. (34), invokes also a pressure and composition inde-
γiLR (T Ps xi = 0) = γiLR,∞ (T Ps ) = 1. In other words, we should pendent partial molar volume for the i-solute, i.e.,υ̂iL (T Ps xi )
stress that the Krichevsky-Kasarnovsky equation is a υ̂i∞,L (T Ps ). Considering the fundamental pressure dependence
for Henry’s law activity coefficients,4 or their partial molar molecular dynamics simulation results.50,51 For the model sol-
fugacity counterparts, i.e., vent argon, we use the critical conditions of the model as
f g f g predicted by the PY-IE following the method used elsewhere,52
∂ ln γiHL (TPxi )/∂P = υ̂iL (TPxi ) − υ̂i∞L (TP) /RT , (36) resulting in ε jj /k = 114.2 K and σjj = 3.367 Å, while for the
T xi
solute hydrogen, we adopted the parameters from Lotfi and
the presumed υ̂iL (T Ps xi ) υ̂i∞,L (T Ps ) condition means that Fischer,49 i.e., ε ii /k = 32.1 K and σii = 2.892 Å. Finally,
the Margules/Porter parameter must also be pressure indepen- for the unlike-pair interaction size and energy parameters, we
dent, ∂A (TP)/∂P = 0, for the Krichevsky-Ilinskaya equation determined their deviations from the conventional Lorentz-
(34) to be self-consistent according to our derived equation Berthelot combining rules24,53 η = σij /σijLorentz ≈ 1.07 584
(12). fIn
other words, Eq. (35) should actually read A (T ) and ξ = ε ij /ε Berthelot ≈ 0.96 398 required to describe rea-
g
o − 2G∞ /υ o , and from a microscopic view- ij
≡ 0.5 G∞ + G
ii jj ij j T sonably and simultaneously Bij (T )54 and Hi,j (T Ps )55 around
point, its pressure independence would signify an imposed arti- T = 100 K.
ficial compensation between the pressure effects on the system In order to determine the isothermal-pressure dependence
microstructure as described by G∞ + G o − 2G∞ and the sys-
ii jj ij of the solubility of the i-gas solute in the j-condensed sol-
tem’s saturation density. However, we will argue that Eq. (34) vent, we apply the alluded IE-PY approach to calculate the
does not characterize properly the actual Krichevsky-Ilinskaya macroscopic quantities in Eq. (12) or (14), from the resulting
gas solubility equation (vide infra Sec. VI). ∞,L
system microstructure, i.e., HISi,j (T Ps ), υ̂i (TP), kij (TP), and
V (TPy )
f̂i i (see Appendix C for details), as a venue to solve the
i = f̂i (TPxi ) required in
IV. GAS SOLUBILITY BEHAVIOR OF MODEL V (TPy ) L
phase equilibrium condition f̂i
BINARY SOLUTIONS the macroscopic (φ − γ) modeling of gas solubility discussed
Because we recognize the challenges behind obtaining in Sec. V.
accurate gas solubility measurements,45 and to avoid any
ambiguity originated in experimental data uncertainties, here
V. MODELING OF ISOTHERMAL GAS SOLUBILITY
we provide a set of thought experiments to generate accu-
rate gas solubility data for a simple model system to illus- The self-consistent microstructural information gener-
trate Eqs. (12)–(14). Our choice of the model system is a ated in Sec. IV and the associated macroscopic solvation
binary mixture interacting via Lennard-Jones pair potentials, quantities along the chosen solubility isotherm become the
whose microstructure and thermodynamic properties are con- accurate sources to find unambiguous answers to the rele-
sistently determined via integral equation (IE) calculations. vant gas solubility modeling issues, including (i) how the
In these thought experiments,46 we have full control of the assumed γiHL (T Ps xi ) γiHL,∞ (T Ps ) = 1 and its conse-
species interaction strength and hence a complete knowledge quent behavior described by Eq. (33) compare against the
of the links between the solute-solvent molecular interac- actual behavior of the model system, (ii) how large the
tion asymmetry, the resulting microstructural manifestation, impact of the assumed pressure independence of υ̂i∞,L (TP)
and ultimately the corresponding thermodynamic properties ' υ̂i∞,L (T Ps ) might be in the calculation of the Poynting cor-
of interest to gas solvation. rection, and (iii) how accurate the approximation υ̂iL (T Ps xi )
The rationale behind the illustration and the choice of the υ̂i∞,L (T Ps ) might be to describe the pressure dependence of
model is manifold, namely, (a) the model’s well-characterized gas solubility.
orthobaric phase envelope, (b) the suitable solution of the Toward that end, except for issue (iii) that will be
atomic Ornstein-Zernike integral equations38 according to addressed in Sec. VI for reasons that will become clear below,
the Percus-Yevick (PY) approximation,47 which provides the we invoke the derived molecular-based expressions, Eqs. (12)–
required complete microstructural information consistent with (14), for the description of isothermal gas solubility in binary
the thermodynamic state and the infinite dilution conditions, aqueous systems, where the modeling proceeds according to
and (c) the accurate access, resulting from the internal consis- a sequential approach. The raw data from the actual solubility
tency, to the magnitude and sign of the individual quantities experiments must comprise the isothermal pressure depen-
affecting the phase equilibrium that affords the direct assess- dence of the vapor- and liquid-phase compositions, yi and xi ,
ment of the impact of the missing terms in the Krichevsky- the second virial coefficient Bij (T ) for the ij-pair interactions,
Kasarnovsky and Krichevsky-Ilinskaya equations and, for
and a value for υ̂i∞,L (T Ps ) from an independent experimen-
that matter, any other equations for the correlation of gas f g
solubility. tal source. After defining =α xi − 0.5xi2 = ln f̂iV (TPyi )/xi
For that purpose, we invoked the approach of McGuigan − Pi υ̂i∞,L (T Ps )/RT based on Eq. (12), its plot allows
and Monson48 to determine the pair correlation functions, their us to extract simultaneously ln HIS i,j (T Ps ) and kij (T Ps ) as
integrals, and the thermodynamic properties of infinitely dilute follows:
binary Lennard-Jones systems, a methodology already used
successfully for this purpose.3 The illustration involves the lim =α xi − 0.5xi2 = ln HISi,j (T Ps ) , (37a)
xi →0
Lennard-Jones parameterization of Lotfi and Fischer 49 as a f g
crude, yet surprisingly acceptable representation of the second (−) lim ∂=α xi − 0.5xi2 /∂ xi − 0.5xi2 = kij (T Ps ) .
xi →0
virial coefficient and Henry’s law constant for the H2 (i)–Ar(j)
system according to the available perturbation theory and (37b)
FIG. 3. Isothermal-composition dependence of the modeling function

FIG.
f 4. Isothermal-pressure dependence of the modeling function = β (P) =
=α xi − 0.5xi2 for the H2 (i)–Ar(j) system as described by the integral equa- g
ln f̂1V (TPy1 )/x1 and its limiting slope (left red) and the corresponding
tion calculations involving the Lennard-Jones model along the T = 100 K
isotherm. behavior of kij (TP) = G∞ii + Gjj − 2Gij /υj (right blue) for the H2 (i)–Ar(j)
o ∞ o
system as described by the integral equation calculations involving the

Lennard-Jones model along the T = 100 K isotherm.
In Fig. 3, we plotted the isothermal-composition
dependence
2
of the modeling function =α xi − 0.5xi for the H2 (i)–Ar(j) Consequently, Eqs. (38a) and (38b) reduce to the equivalent
system as described by the IE calculations discussed in Sec. IV. forms
This function exhibits an advantageous quasi-linear behavior
f g
lim ∂=β (TPxi )/∂P = υ̂i∞,L (T Ps )/RT − kij (T Ps )
that allows the reliable extraction of the two relevant solvation xi →0 T
f g
quantities associated with the limiting conditions, Eqs. (37a) × lim ∂ xi − 0.5xi2 /∂P
T
and (37b), i.e., ln HIS
i,j (T Ps ) 0.502 and kij (T Ps ) 10.02.
xi →0
These results provide an answer to issue (i) above, i.e., for a = υ̂i∞,L (T Ps )/RT − kij (T Ps )/HIS
i,j (T Ps ),
rather conservative low i-solute concentration we have, after
(42)
(B3), that 0.6 ≤ γiHL (T Ps xi ≤ 0.05) ≤ 1.0; therefore, the f g
assumption γiHL (T Ps xi ) γiHL,∞ (T Ps ) = 1 becomes less with =β (TPxi ) = ln f̂iV (TPyi )/xi as illustrated in Fig. 4 and
plausible as the system departs from the Lewis-Randall ideal-
lim ∂=α (Pi )/∂Pi T = −kij (T Ps )/HIS

ity, even for the small interaction asymmetries encountered in xi →0 i,j (T Ps ) (43)
real systems (e.g., vide infra CH4 − H2 O). from which we can determine kij (T Ps ) after substituting
Then, from the pressure dependence of either Eq. (12) or HIS
i,j (T Ps ) from Eqs. (38a) and (38b).
Eq. (14), we also have that In order to address issue (ii), weanalyze the behavior of
f g the Poynting correction as ∆PC = exp ∫ PPs (T ) υ̂i∞,L (TP) dP/RT
∂ ln f̂iV /xi /∂P f g
f T
g − exp υ̂i∞,L (T Ps ) (P − Ps )/RT with υ̂i∞,L (TP) generated by
= υ̂i∞,L (T Ps )/RT − ∂kij /∂P xi − 0.5xi2 the IE calculations along the T = 100 K isotherm and
T
( P ) P = 40 atm, i.e., slightly below the limit of stability for
− ∂kij /∂P dP+kij Ps
(T ) this mixture. On the right ordinate of Fig. 1, we display the
T
P (T ) two exponential terms of ∆PC (TP) fwhere we clearly
f s g g observe
× ∂ xi − 0.5xi /∂P ,2
(38a) a very small over-prediction PC TP, υ̂i∞,L (T Ps ) relative
T f g f g
to PC TP, υ̂i∞,L (TP) , i.e., ∆PC (TP)/PC TP, υ̂i∞,L (T Ps )
f g
∂ ln f̂iV /xi /∂P
T / 2%. Obviously, the departure ∆PC (TP) becomes monoton-
f ∞,L g
= υ̂i (T Ps )/RT − ∂kij /∂P xi − 0.5xi2 ically larger with increasing values of υ̂i∞,L (TP) and system
T
f g pressure.
− kij (TP) ∂ xi − 0.5xi2 /∂P . (38b)
T
VI. DISCUSSION AND OUTLOOK

In addition, after considering the limiting condition
lim P (T ) = Ps (T ) ≡ Ps , it follows that We frequently find in textbooks and the specialized lit-
xi →0
erature the Krichevsky-Ilinskaya equation written as either
f g (i)
∂ xi − 0.5xi2 /∂P = (1 − xi )/[∂ (P − Ps )/∂xi ]T , (39) f g ∞,L
T
ln f̂iL (TPxi )/xi = ln HIS
i,j (T Ps ) + (P − Ps ) υ̂i /RT
lim [∂ (P − Ps )/∂xi ]T = HIS
i,j (T Ps ) , (40)
xi →0
f g + A xj2 − 1 (44)
lim ∂ xi − 0.5xi2 /∂P = 1/HIS i,j (T Ps ). (41) under the clarification that “. . . Equation (8.3.9) assumes
xi →0 T
P→Ps that the partial molar volume of the solute is independent of
pressure and composition over the pressure and composition solution, i.e.,
ranges under consideration”41 or (ii)
υ̂i∞,IS (TP) = υjo − lim
∞
∆∞ ∞ ∞
ij + Gij − Gii
∆ij →0
f g ∞,L
i,j (T Ps ) (P − Ps ) υ̂i
ln f̂iV (TPyi )/xi HIS (T Ps )/RT = υio (TP) .

= A (T Ps ) xj2 − 1 , (45) An instructive example of these overlooked modeling sub-
tleties can be extracted from the recent work of Hou et al.57
with f̂iV (TPyi ) = yi P φ̂Vi (TPyi ), where the author explic- who illustrated the failure of the Krichevsky-Kasarnovsky
itly approximates the actual solute partial molar volume as equation for the solubility of CO2 in the CO2 − H2 O sys-
υ̂iL (TPxi ) υ̂i∞,L (T Ps ) whose state conditions are (T Ps ) tem along thef T ≈ 423 K isotherm
g by highlighting in their
Fig. 14 that ∂ ln f̂i /xi /∂Pi
L < 0, which according

≡ T Pσ,j in our notation.4,56 T ,xi →0
Here we argue that, as long as Henry’s law activity coef- to Eq. (12) would translate into υ̂i∞,L (T Ps ) < 0. Obviously
ficient is described
by the two-suffix Margules expression, this outcome becomes incompatible with the solvation behav-
ln γi = A 1 − xj , such an approximation is both unnec-
HL 2
ior of gases in water, one characterized as either weakly
essary and misleading; in fact, by a simple look at the rigorous attractive or repulsive (non-volatile)58,59 that must fulfill the
condition (∂P/∂xi )∞ Tυ > 0 and consequently
equations (11)–(14), it becomes clear that Eqs. (44) and (45) thermodynamic

might be considered as alternative interpretations of the orig- υ̂i /υj = 1 + κ oj (∂P/∂xi )∞
∞ o
Tυ > 0. In fact, in the context of
inal Krichevsky-Ilinskaya equation in that υ̂i∞,L (T Ps ) comes 57
the observation by Hou et al., this behavior provides a hint
from the first term of the composition dependence of the par-
into the likely effect of the missing term in the Krichevsky-
tial molar volume of the i-solute, υ̂iL (TPxi ) = υ̂i∞,L (TP) −
Kasarnovsky equation, later added to the Krichevsky-Ilinskaya
RT ∂kij /∂P xi − 0.5xi2 —Eq. (11)—where A = 0.5kij equation, e.g., vide supra Eq. (42),
T
(vide infra Appendix B). This argument is not only of aca- f g
demic interest but also of practical relevance in that (a) ∂ ln f̂iL /xi /∂Pi = υ̂i∞,L (T Ps )/RT
T xi →0
it identifies the microscopic nature and characterizes the
− 2A (T Ps )/HIS
i,j (T Ps ) ≶ 0, (46)
actual source of solution non-ideality through the param-
eter A and (b) it highlights the fact that the state con- with Pi = P − Ps . Clearly, Eq. (46) can be satisfied by
ditions (TP) in the last term fof the Krichevsky-Ilinskaya
g υ̂i∞,L (T Ps ) > 0 since A (T Ps ) ≡ 0.5ρoj ∆∞ij ≶ 0 (vide infra
equation, A (TP) xj2 − 1 ≡ ln γiLR,∞ (TP) xj2 − 1 , are Appendix B), i.e., a necessary condition for A (T Ps ) to be
the result of the pressure integration given by Eq. (13), able to describe positive/negative deviations from the Lewis-
not a result of the alleged approximation indicated in Randall ideal solution reference (e.g., see Fig. 8 and related
Eqs. (44) and (45). This implicit pressure dependence con- discussion in Ref. 25).
spires against the conventional linear regression =α (zi )T = ln A key feature underlying the sequential approach dis-
cussed in Sec. V is its exclusive reliance on the i-solute
f g ∞,L
f̂iV (TPyi )/xi HIS
i,j (T P s ) − υ̂i
(T Ps ) (P − Ps )/RT ≡ 2A (zi ),
phase equilibrium condition alone, i.e., no need for the cor-
where zi = xi − 0.5xi , to extract the Margules/Porter param-
2
responding j-solvent counterpart. This scenario precludes the
eter A (TP) from its composition slope;31 because the system
automatic compliance of thermodynamic constraints including
pressure is not constant and xi = =(P)T , the slope A (TP) is
the Gibbs-Duhem and Maxwell relations, and consequently, it
not constant along the plotted isotherm.
may lead to thermodynamic inconsistencies (see, e.g., Sec. 5
The Krichevsky-Kasarnovsky equation has been fre-
of Ref. 15). By contrast, the simultaneous approach provides
quently used to extract υ̂i∞,L (TP) from solubility measure-
an alternative avenue for the determination of the gas solubility
ments at elevated pressures, and the conventional wisdom sug-
parameters by applying readily available non-linear optimiza-
gests that under those state conditions the gas solubility might
tion methodologies.60 The success of these optimized regres-
be already large enough to invalidate the underlying assump-
sions will be greatly enhanced by invoking the i-solute and
tion γiHL (T Ps xi ) 1, rendering unreliable the υ̂i∞,L (TP)
corresponding j-solvent phase-equilibrium conditions, vide
values.4,41 Our molecular-based analysis, encapsulated in infra Appendix D, with the addition of suitable constraints
Eqs. (31)–(33), unravels an overlooked subtlety on the above to ensure thermodynamic as well as solvation parameter con-
scenario, namely, that the Krichevsky-Kasarnovsky equation sistency and avoid potential unphysical modeling results as we
becomes a better descriptor of gas solubility with isothermal discuss below.
decreasing solute-solvent molecular asymmetry, which in turn Beyond the required compliance with the Gibbs-Duhem
makes the i-solute more soluble in the j-solvent environment. and Maxwell relations for the second-order composition
In other words, the conjectured γiHL (T Ps xi ) 1 is not the expansion underlying the phase-equilibrium conditions,26 we
consequence of the xi → 0 limiting condition, but rather of recognize additional relations from the definition of HIS i,j (TP),
γiHL (TPxi ) ≡ γiLR (TPxi )/γiLR,∞ (TP) → 1 with γiLR,∞ (TP) → Eq. (4), where we identify two alternative expressions linking
1 as ∆∞ ij → 0 limiting conditions, i.e., when the strength of HISi,j (TP) to an infinite dilution coefficient with well-defined
the solute-solvent interactions approaches that of the average microstructural nature, i.e.,
between solvent-solvent and solute-solute interactions regard- f g
i,j (TP) = P lim φ̂i (TPxi )
HIS L
less of the system composition (vide infra, Appendix A). xi →0 T
Consequently, the regressed value of υ̂i∞,L (TP) will approach
that of the solute in the corresponding Lewis-Randall ideal = Pφo,L
i
(TP) γiLR,∞ (TP) , (47)
where Pφo,L
i
(TP) = fio,L (TP) denotes the pure distribution coefficient Ki∞ (T Ps ) as well as the corresponding
component fugacity of the i-solute, while Ostwald coefficient Li,j ∞ (T P ) to be discussed below.
f g s
γiLR,∞ (TP) = lim φ̂Li (TPxi )/φio,L (TP) , at the prevailing Regarding the alluded a priori assessment of conjec-
xi →0 T
tured behaviors, in items (iii-iv), we note that Dhima et al.’s
state conditions. Thus, by invoking Eq. (35) in conjunction
gas solubility study61 reported the regression of the quantity
with (B3), we have
f g γiHL (T Ps xi ) HIS IS
i,j (T Ps ) rather than the usual Hi,jf(T Ps ) in their
HISi,j (T Ps ) = fi (T Ps ) exp 0.5kij (T Ps ) ,
o
(48)
g
Eq. (5) and properly indicated that the plot of ln f̂iL (TPxi )/xi
as a function of (P − Ps ) will exhibit a linear behavior with
f g
kij (T Ps ) = 2 ln HIS o
i,j (T Ps )/fi (T Ps ) , (49)
slope υ̂i∞,L (T Ps )/RT with an ordinate at the origin equal to
where once again Ps is the saturation pressure of the j-solvent γiHL (T Ps xi ) HIS ∞,L
i,j (T Ps ) as long as both γi and υ̂i
HL behaved as
at the prevailing temperature. Consequently, we can now pro- independently of pressure and composition. The constancy of
vide additional insights into Eq. (46), by rewriting the ratio the product γiHL (T Ps xi ) HIS i,j (T Ps ), however, becomes a direct
2A (T Ps )/HIS
i,j (T Ps ) in two alternative ways as follows: consequence of Eqs. (48) and (52), i.e.,
f g f g
2A (T Ps )/HISi,j (T Ps ) = kij (T Ps ) exp −0.5kij (T Ps ) /fi (T Ps ) i,r (T Ps ) = fi (T Ps ) exp 0.5kij (T Ps ) xj
γiHL (T Ps xi ) HIS
o o 2
f g f g
= 2 ln HIS i,j (T Ps )/fi (T Ps ) /Hi,j (T Ps ).
o IS
fio (T Ps ) exp 0.5kij (T Ps ) , (53)
(50) since the j-solvent mole fraction xj ' 1, regardless of
any condition of pressure-composition independence for γiHL
Equation (50) reveals a (so far) hidden molecularly based
and υ̂i∞,L .
link between the coefficient A (T Ps ) and Henry’s law con-
The interdependence of the solvation parameters embod-
stant HIS i,j (T Ps ) and, consequently, highlights the fact that the
ied in Eqs. (48) and (49) provides a quick test of consistency
Krichevsky-Ilinskaya slope, identified by Eq. (46), comprises
for regressed quantities as suggested in item (v). In fact,
three macroscopic quantities but actually only two variables.
an instructive illustration on the alluded internal consistency
More importantly, Eqs. (48) and (49) provide a novel avenue
regards the study of the CH4 − H2 O binary system. On the
for a few relevant observations including (i) the reformulation
one hand, Bender et al.62 reported in their Fig. 9 the tem-
of the gas solubility, Eq. (14), in a pair of equivalent forms as
perature dependence of the Margules parameter ACH4 –H2 O (T )
follows:
f g f g according to its description from two equations of state (EoSs).
ln f̂iL (TPxi )/xi = xj2 ln HIS o o
i,j (T Ps )/fi (T Ps ) + ln fi (T Ps ) On the other hand, Dhima et al.61 analyzed this aqueous sys-
P f g tem and determined the corresponding Henry’s law constant,
+ υ̂i∞,L (TP)/RT dP CH4 ,H2 O (T = 344 K) = 5930 MPa. At the water satura-
i.e., HIS
Ps (T ) tion conditions, fCH o (T P ) P (T = 344 K) 0.0306 MPa,
4
s s
0.5xj2 kij (T Ps ) + ln fio (T Ps ) and then according to Eq. (49), we have that ACH4 –H2 O (T Ps )
12.2. The comparison between this value and the graphi-
+ υ̂i∞,L (T Ps ) (P − Ps )/RT , (51a)
f g cal estimations from Bender et al.’s Fig. 9, ACH4 –H2 O (T Ps )PR
ln f̂iL (TPxi )/xi = 0.5xj2 kij (T Ps ) + ln fio (T Ps ) ≈ 6.4 and ACH4 –H2 O (T Ps )RK ≈ 10.6 according to the Peng-
P f g Robinson and Redlick-Kwong EoS calculations indicates an
+ υ̂i∞,L (TP)/RT dP underestimation of the actual activity coefficient γCH∞ by a fac-
4
Ps (T )
tor of 66 and 5, respectively, and provides convincing evidence
0.5xj2 kij (T Ps ) + ln fio (T Ps ) for the lack of internal consistency of the thermodynamic
+ υ̂i∞,L (T Ps ) (P − Ps )/RT , (51b) results in Ref. 62, which has been previously implied.63
As mentioned in item (vi), Eqs. (48) and (49) allow the
(ii) the estimation of the challenging υ̂i∞,L (T Ps ) from the formal link between the descriptors of the microstructure of
independent knowledge of Henry’s law constant HIS the phases in vapor-liquid equilibrium and the i-solute distribu-
i,j (T Ps )
or kij (T Ps ), (iii) the a priori assessment of the conjectured tion coefficient Ki∞ (T Ps ),64 as derived in Appendix E, results
γiHL (TPxi ) 1 condition in either the Krichevsky-Ilinskaya from invoking the phase equilibrium condition Eq. (1) and the
or Krichevsky-Kasarnovsky equation, (iv) the consistent eval- alternative expression for Henry’s law constant, Eq. (47), i.e.,
f g (
uation of the corresponding Henry’s law activity coefficient Ki∞ (T Ps ) = fio (T Ps )/Ps (T ) exp 0.5kij (T Ps )
for the dilute i-solute in a j-solvent, i.e., f g)
− βPs δjj (T ) − Bii (T ) ,
ln γiHL (T Ps xi ) = −kij (T Ps ) xi − 0.5xi2
f g where we can also express fio (T Ps ) in terms of the microstruc-
= 2 ln fio (T Ps )/HIS 2
i,j (T Ps ) xi − 0.5xi , ture of the pure i-solute at the prevailing state conditions as
(52) depicted by Eqs. (E4a) and (E4b) in Appendix E. Likewise,
the corresponding Ostwald coefficient Li,j ∞ (T P ) becomes65
s
and consequently (v) the detection of potential inconsistencies
among solvation parameters in the solubility data regression as Ps (T ) ρo,L
i
(T Ps ) ( f g
∞
Li,j (T Ps ) = exp βPs δij (T ) − Bii (T )
well as (vi) the formal link between the microstructural behav- fio (T Ps ) ρo,V
i
(T Ps )
ior of the dilute solution under vapor-liquid equilibrium and )
additional relevant solvation quantities including the i-solute − 0.5kij (T Ps ) ,
where Bij (T ) = lim Gij⊗ (TPxi ) and δij (T ) = lim ∆∞

ij (TP). so that, for the i-solute at infinite dilution, the microscopic
ρ→0 ρ→0
manifestation of the Lewis-Randall ideal solution reads as
follows:
VII. SUMMARY ∞,IS
Gii∞,IS + Go,IS
jj − 2G ∞,IS
ij = 0 → G ∞,IS
ij = 0.5 G ii + G o,IS
jj , 0,
In this work, we have developed an explicit molecularly
(A2)
based interpretation of the thermodynamic phase equilibrium
underlying gas solubility through the identification of funda- and consequently, from Eq. (15), kijIS (TP) = 0. Thus from
mental links between the microstructure of the infinite dilute (A2), Eq. (23) becomes
reference systems and the relevant macroscopic quantities
IS
that characterize the resulting dilute solutions and phases ∞,IS

o = kT ρj / 1 + ρoj Gojj
o2
(∂P/∂xi )Tυ Gojj − G∞
ij
in equilibrium. The formalism hinges around the thermo- j
dynamic self-consistent second-order truncated composition ∞ IS

= −kT ρo2 ∞
j Gii − Gij / 1 + ρoj Gojj , 0. (A3)
expansion of the partial molar quantities of the species regard-
less of its aggregation state,15,26 which (a) provides the tools Likewise, from Eq. (22), we obtain
to identify and assess at a molecular level the nature of the IS
approximations behind the commonly used equations for the υ̂i∞,IS (TP) = υjo + Gojj − G∞
ij
modeling of gas solubility, (b) highlights the relevant sub-
∞ IS

tleties behind the conjectured approximations associated with = υjo − G∞ii − Gij
the macroscopic modeling, and (c) delivers a general funda- = υio (TP) . (A4)
mentally and molecularly based platform for gas solubility
modeling. The second scenario is a more restrictive condition for
We have illustrated the derived relationships by generating the Lewis-Randall ideal solution, one in which all TCFI’s
proxy data from IE calculations of a model system and unrav- are identically the same, Gojj = G∞ii = Gij , resulting in
∞
eled novel interdependences between the solvation parameters IS,Gojj =G∞

ij =Gii
∞ ∞,IS,Gojj =G∞
ij =Gii
∞
kij (TP) = 0, υ̂i = υ̂jo = ρo(−1)
j ,
relevant to gas solubility in dilute systems. This analysis results ∞,IS,Gojj =G∞
ij =Gii
∞
in, among other things, the reformulation of the gas solubility and therefore, (∂P/∂xi )Tυ o = 0, i.e., this scenario
j
problem leading to the a priori assessment of the validity of describes an infinite dilute solution where the i-species and
modeling assumptions, the detection of potential inconsisten- the j-species are only distinguishable by their label, yet, the
cies among gas solubility regressed quantities, and the formal strength of like- and unlike-pair interactions are all the same
links between the Ostwald coefficient, the solute distribution as that in the pure j-solvent.36
coefficient, and the microstructural behavior of the system at Then, the third scenario involves a non-interacting, or
vapor-liquid phase equilibrium. ideal gas, infinite dilute particle immersed in an otherwise
real pure solvent, whose microstructural signature is a uni-
form distribution of the solute-solvent interactions described
ACKNOWLEDGMENTS
ij (PT ) = 0. Thus, after invoking Eq. (23), we find that
by G∞ 67
The author expresses his gratitude to Professor Peter A.

i
Monson for providing years ago the original version of his IE- (∂P/∂xi )∞,IG
Tυjo
= kT ρo2
j Gjj / 1 + ρj Gjj
o o o
PY code used in the reported calculations and to Dr. Sebastian −1 −1

Chialvo for the critical reading of the manuscript. = κ o,IG
j − κ oj , (A5)
where IG i superscript denotes the ideal gas behavior of the

APPENDIX A: EFFECT OF SOLUTE-SOLVENT i-solute in solution, κ o,IG = kT ρoj is the ideal gas isothermal
INTERACTION ASYMMETRIES ON RELEVANT GAS j
compressibility of the j-solvent at the prevailing conditions,
SOLVATION QUANTITIES
and κ oj /κ o,IG
j = 1 + ρoj Gojj . In addition, from Eqs. (22) and (23),
From the microscopic definition of υ̂i∞ (TP) and its link to we arrive at the expression for the partial molar volume of the
(∂P/∂xi )∞
Tυ o , Eqs. (22a) and (23), and the molecular identifica- ideal gas solute at infinite dilution in the real solvent, i.e.,
j
tion of the limiting composition slope kij (TP), Eq. (15), we can
analyze the impact, and derive the expressions, of the strength υ̂i∞,IG i (TP) = υjo κ oj /κ o,IG
j = kT κ oj . (A6)
of the solute-solvent interaction for a few precisely defined
molecular asymmetry, namely, (a) a real solute in a Lewis- The expression (A6) highlights the fact that υ̂i∞,IG i (TP) > 0,
Randall ideal solution, (b) a real solute behaving identically to a behavior that characterizes an infinite dilute solute as either
the solvent, and (c) an ideal gas solute in a real solution volatile59 or weakly repulsive (i.e., at the border between
The first scenario defines the Lewis-Randall ideal solution repulsive and weakly attractive)58 species given the fact that
for a binary system, a reference for which the excess chemical υ̂i∞,IG > 0 simultaneously with G∞ ij = 0. Moreover, accord-
potential of an i-species µEi (TPxi ) = kT ln γiLR (TPxi ) ≡ 025,66 ing to Eq. (15), the coefficient of the composition
expansion

that translates into ∆IS (TPxi ) = 0, i.e., for this case becomes kijIG i (TP) = Gojj /υjo = κ oj /κ o,IG
j − 1,
ij
a negative quantity since κ oj < κ o,IG
j that translates into
∆IS
ij (TPxi ) ≡ GIS
ii + GIS
jj − 2GIS
ij = 0, ∀TPxi (A1) kijIG i (TP) < 0.
with ln γiHL (TPxi ) = ln γiLR (TPxi ) − ln γiLR,∞ (TP) and, con-

sequently, A = ln γiLR,∞ (TP), i.e.,

ln γiHL (TPxi ) = −2A (TP) xi − 0.5xi2 ,
(B3)
ln γjLR (TPxi ) = A (TP) xi2 .
We have also indicated that ρoj ∆∞ ij can be determined as

the limiting slope of the limiting composition dependence of
ln γi (TPxi ).15,25 Alternatively, we can express A (TP) in terms
of other thermodynamic properties of the infinite dilute solute
and the corresponding pure solvents as
A (TP) = − 0.5ρoj [2Bi⊗ (TP) + vjo − 2υ̂i∞ + kT κ oj ], (B4)
where Bi⊗ (TP) is the osmotic second virial coefficient 69 so

FIG. 5. Isothermal liquid-phase equilibrium composition x1 resulting from that, from Eq. (16),
the phase equilibrium condition f̂1V (TPy1 ) = f̂1L (TPx1 ) when f̂1L (TPx1 ) is ∞
represented by three distinct prescriptions around Henry’s law for the Bi⊗ (TP) = − 0.5G∞
ii = −2π hii∞ (TP)r 2 dr (B5)
H2 (i)–Ar(j) system as described by the integral equation calculations 0
involving the Lennard-Jones model along the T = 100 K isotherm.
and the pure solvent isothermal compressibility
From the definition of Henry’s law constant, κ oj = υjo + Gojj /kT , with ρoj = 1/υjo . In addition, after invok-
f g ing Eq. (22a), from (B4) and (B5), we recover Eq. (35),
Hi,j (T Ps ) = Ps exp β µR,∞ i
(T Ps ) i.e.,
f g
= kT ρoj (T Ps ) exp β µR,∞ T ρoj , (A7) f
A (TP) = − 0.5ρoj −G∞
g
ii + 2 υj − υ̂i
o
i ∞
+ Gojj
where we have invoked the link between µR,∞
f g
i
(TP) and = − 0.5ρoj −G∞ ii + 2Gij − Gjj = 0.5ρj ∆ij .
∞ o o ∞
(B6)
µR,∞
i
(T ρ), 19 and after the fact that for an ideal gas i-solute
Note that (B4) has been previously derived by Alvarez and
µR,∞,IG
i
i
T ρoj = 0,68 we obtain HIG i,j (T Ps ) = kT ρj (T Ps ).
i o
Fernandez-Prini,70 denoted b2 (TP) in their Eq. (7), and used
As an illustration of the effect of the solute-solvent molecular as the coefficient for the linear composition dependence
asymmetry on gas solubility, we plot in Fig. 5 the isothermal of Henry’s law activity coefficient for the i-solute, their
liquid-phase equilibrium composition x1 resulting from the Eq. (6), i.e.,
phase equilibrium condition f̂1V (TPy1 ) = f̂1L (TPx1 ). For that
ln γiHL = b2 (TP) xi . (B7)
purpose, f̂1L (TPx1 ) is represented by the prescriptions around
Henry’s law discussed above involving the H2 –Ar system However, we must highlight, as we have discussed extensively
along the T = 100 K isotherm according to the IE calcula- elsewhere,14,15,26 that the linear composition dependence in
tions of Sec. IV. This figure highlights the balance between (A5) as well as in Eq. (11) of Ref. 71 and Eq. (A1) of
the Poynting and the non-ideality corrections to Henry’s law Ref. 72 are thermodynamically inconsistent within the frame-
solubility, a feature that is typically overlooked and lost dur- work of composition perturbation expansion approaches, i.e.,
ing the regression of solubility quantities, whose simultaneous their consistency requires a quadratic composition term as
inaccurate representations for υ̂i∞ (TP) and γiHL (TPxi ) lead to a that given by (A1) and (A2) to satisfy simultaneously the
reasonable fit due to compensation of those corrections. This is Gibbs-Duhem and Maxwell relations.73 Moreover, we should
precisely one of the scenarios where we need to have available also note that b2 (TP) does not account just for the solute-
a tool for the detection of and test for internal thermodynamic solute interactions (or correlations at infinite dilution) as fre-
inconsistencies. quently claimed in the literature,71,72 but also for the solvent-
solvent and, most importantly, the solute-solvent interac-
tions. This is a crucial observation in that even when we
APPENDIX B: MICROSCOPIC INTERPRETATION
OF THE MARGULES/PORTER PARAMETER A(T , P)
are in the presence of a non-interaction ideal gas solute,
there is a contribution from the solute-solvent
interactions,
As we have already identified it, according to Eq. (12), the i.e., b2 (TP) = kijIG i (TP) = κ oj /κ o,IG
j − 1, resulting from
A(T , P) parameter in the Krichevsky-Iliinskaya equation can the solute-solvent interaction asymmetry as discussed in
be rigorously expressed as follows: Appendix A.
f g
ii + Gjj − 2Gij /υj
o o
A (TP) ≡ 0.5 G∞ ∞
TP
APPENDIX C: INTEGRAL EQUATION CALCULATIONS
= 0.5ρoj ∆∞
ij , (B1) OF THE GAS SOLUBILITY QUANTITIES
where the Margules/Porter equation reads21 The solution of the Ornstein-Zernike integral equation
provides the three TCFI’s—G∞ o ∞
g (TPxi )/RT = A (TP) xi xj ii , Gjj , Gij —and correspond-
E
∞ o ∞
ing DCFI’s—Cii , Cjj , Cij —for the infinite dilution system at
= xi ln γiHL (TPxi ) + xj ln γjLR (TPxi ) , (B2) the relevant state conditions, i.e., from the saturation condition
to the pressure of interest along the chosen isotherm. For each f̂jV (TPyi ) = f̂jL (TPxi ) ,
state condition, we determine the following quantities: (D1)
yj φ̂Vj (TPyi ) = xj φ̂Lj (TPxi ) ,
υ̂i∞,L (TP) ← Eqs. (22a) and (22b), (C1)
kij (TP) ← Eq. (15), (C2) where we kept the i-solute mole fraction as the descriptor of
the phase compositions. By invoking the second-order com-
Hi,j T , Pj,s = fio T , Pj,s γiLR,∞ T , Pj,s position expansion26 to both phases, the partial molar fugacity
of the j-solvent in each phase becomes

= Ps φ̂∞i T , Pj,s , (C3)

where the activity coefficient γiLR,∞ T , Pj,s is calculated as
( f g)
f̂jV (TPyi ) = P (1 − yi ) exp β Bjj P − Pj,s − Pδij yi2 ,
follows (see Appendix B for details):
(D2)
A (TP) = 0.5ρoj ∆∞
f g
ij f̂jL (TPxi ) = Pj,s (1 − xi ) exp βυjo P − Pj,s + 0.5kij xi2 ,
LR,∞
= ln γi (TP) , (C4) (D3)

while the fugacity coefficient fio T , Pj,s = Pj,s
f g where υjo is the molar volume of the pure j-solvent and
exp β µor T , P results from the density integral,
i j,s
γjLR (TPxi ) = exp 0.5kij xi2 as derived in Appendix B. Thus,
 ρ(Pj,s )  after solving the phase equilibrium
condition,
f (D1), we define
g
the plotting function =γ xi2 = ln f̂jV (TPyi )/Pj,s (1 − xi ) −

µor
i T , Pj,s = − β −1  Ciio d ρ + ln z T , Pj,s  , (C5a)
 0
βυjo P − Pj,s whose limiting composition slope becomes


 ρ(Pj,s ) f o
 g
µor
i T , Pj,s = − β −1 
 0 G ii / 1 + ρG o
ii dρ f g
 lim ∂=γ xi2 /∂xi2 = 0.5kij T Pj,s . (D4)
xi →0 T
 
+ ln z T , Pj,s  , (C5b)
 Note that if kij (TP) exhibits a weak pressure depen-
dence, the
limiting condition (D4) can be replaced by
where the compressibility z T , Pj,s can be written as (e.g., f g
∂=γ xi2 /∂xi2 0.5kij T Pj,s . In Fig. 6, we illustrate
see Appendix C of Ref. 3) T
the well-behaved nature of =γ xi2 and the limiting con-
ρ(Pj,s )
dition for the H2 (i)–Ar(j) system from the IE calculations
z T , Pj,s = −ρ−1 ρCiio d ρ, (C6a) reported in Sec.IV. However, we note that despite its linear
0
ρ(Pj,s ) f g appearance, =γ xi2 displays a very small curvature such that
z T , Pj,s = −ρ −1
ρGoii / 1 + ρGoii d ρ, (C6b)
f g
lim ∂=γ xi2 /∂xi2 5.01, while the average xi2 -slope of
0 xi →0 T

for the supercritical state conditions T ∗ = kT /ε ii and =γ xi2 is ∆=γ xi2 /∆xi2 4.86 in comparison with the true
P∗ = Pj,s σii3 /ε ii . Finally, the fugacity f̂iV (TPyi ) is determined

value given by (D4), i.e., 0.5kij T Pj,s = 5.04.
in terms of the truncated virial equation of state, the consistent
low-density counterpart of the second-order truncated com-
position expansion of the partial molar quantities for liquid
mixtures used throughout this analysis (see Sec. IV B of Ref. 26
for details), i.e.,
f̂iV (TPyi ) = yi φ̂Vi (TPyi ) P

f g
= P − Pj,s exp βP Bii − yj2 δij , (C7)
where δij (T ) = lim ∆∞

ij (TP) = Bii (T ) + Bjj (T ) − 2Bij (T ),
ρ→0
whose individual contributions can be determined from the
available tabulation, i.e., Bij∗ kT /ε ij = 3Bij /2πσij3 , for the
Lennard-Jones fluid.54
APPENDIX D: SIMULTANEOUS SOLUTION

OF THE GAS-SOLUBILITY PHASE EQUILIBRIUM
While we have focused almost exclusively on the i-solute
FIG.
f 6. Isothermal-pressure
g g the modeling function =γ =
fdependence of
ln f̂jV (TPyi )/Ps xj − ∫ PPs (T ) υjo (TP)/RT dP and its limiting slope (D4)
phase equilibrium equations for the current analysis, those for for the H2 (i)–Ar(j) system as described by the integral equation cal-
the j-solvent can also help in that task. In fact, the counterparts culations involving the Lennard-Jones model along the T = 100 K
of Eq. (1) for the solvent are isotherm.

APPENDIX E: DERIVATION OF THE NOVEL where κ o,IG
i T Pj,s is the ideal gas isothermal compressibility
EXPRESSIONS FOR Ki∞ (T Ps ), L∞
i,j (T Ps ), of the pure i-solute at the j-solvent saturation conditions.
AND THEIR LIMITING BEHAVIORS Consequently, the corresponding Ostwald coefficient
∞ (T P ) becomes65
Li,j
Starting with the phase equilibrium condition Eq. (1) and s
the alternative expression for Henry’s law constant, Eq. (47),
Ps (T ) ρo,L (T Ps )
we have that ∞
Li,j (T Ps ) = i
fio (T Ps ) ρo,V (T Ps )
Ki∞ (T Ps ) = lim (yi /xi ) ( i f g )
xi ,yi →0 × exp βPs Bjj (T ) − 2Bij (T ) − 0.5kij (T Ps )
P→Ps
= φ̂∞,L (T Ps )/φ̂∞,V (T Ps ) (E5a)

i i
∞,V
= HIS
i,j (T Ps )/Ps φ̂i (T Ps ). (E1) or in the equivalent form
Ps (T ) ρo,L (T Ps )
Moreover, for the vapor phase away from the f j-solvent
critical
g ∞
Li,j (T Ps ) = i
conditions, we have that φ̂∞,V
i
(T P s ) = exp βP s B ii − δ ij ,
26
fio (T Ps ) ρo,V (T Ps )
i.e., ( i f g )
( × exp βPs δij (T ) − Bii (T ) − 0.5kij (T Ps ) .
∞,V
f g)
φ̂i (T Ps ) = exp −0.5 βPs lim Gii (T Ps ) − ∆ij (T Ps )
∞ ∞
(E5b)
ρ→0
Note also that if the i-solute behaves as an ideal gas particle
( f g)
= exp 0.5 βPs lim 2Gij (T Ps ) − Gjj (T Ps ) ,
∞ o
then, from Eqs. (E1), (C5b), and (C6b), we have that the vapor-
ρ→0
liquid distribution coefficient of an ideal gas i-solute in a real
(E2)
j-solvent reads
and, after invoking Eqs. (48) and (C5)–(C7), Eqs. (E1) and Ki∞,IG i (T Ps ) = zjo,V (T Ps )/zjo,L (T Ps )
(E2) lead to the sought novel interpretation for Ki∞ (T Ps ),
ρjo,V (Pj,s ) f g
( ρo,L
j Pj,s ∫ 0 ρGoii / 1 + ρGoii d ρ
=
f g
Ki∞ (T Ps ) = fio (T Ps )/Ps (T ) exp 0.5kij (T Ps ) ρjo,L (Pj,s ) f g ,
o,V
ρj Pj,s ∫ 0 ρGii / 1 + ρGii d ρ
o o
f g)
− βPs lim G∞ ij (T Ps ) − 0.5G o
jj (T Ps ) (E6)
ρ→0
while the corresponding Ostwald coefficient becomes
f g (
= fio (T Ps )/Ps (T ) exp 0.5kij (T Ps )
∞,IG i
o,V
(T Ps ) = zjo,L (T Ps ) ρo,L Pj,s /zj (T Ps ) ρo,V
f g)
− βPs Bjj (T ) − 2Bij (T ) (E3a) Li,j j j Pj,s
or in the equivalent form, after invoking δij (T ) = Bii (T ) + = 1. (E7)

Bjj (T ) − 2Bij (T ), as follows: Here we note that the behavior of the vapor-liquid distribution
f g ( coefficient Ki∞ (T Ps ), Eqs. (E1) and (E6), obeys the following
Ki∞ (T Ps ) = fio (T Ps )/Ps (T ) exp 0.5kij (T Ps ) limiting condition:
f g)
− βPs δjj (T ) − Bii (T ) , (E3b)
lim Ki∞ (T Ps ) = lim Ki∞,IG i (T Ps ) = 1, (E8)
T →Tc,j T →Tc,j
where kij (T Ps ) = ρoj ∆∞
ij is defined by Eq. (15) and fio (T Ps ) Ps,j →Pc,j Ps,j →Pc,j
can be expressed by its molecular counterparts in terms of
i.e., regardless of the type of i-solute. Likewise, from Eq. (E5),
either the DCFI, Ciio (T Ps ), or the TCFI, Goii (T Ps ), for the pure
i-solute at the prevailing saturation conditions, i.e., lim Li∞ (T Ps ) = 1. (E9)
f g T →Tc,j
fio T Pj,s = Pj,s exp β µor i T Pj,s Ps,j →Pc,j
 ρ(Pj,s )  Finally, for the i-solute in the ideal solution

= Pj,s exp − Ciio d ρ − ln z T Pj,s  defined by the Lewis-Randall reference, we have that
γiLR (TPxi ) = φ̂i (TPxi )/φoi (TP) = 1; therefore, from (E1), we
 0 
ρ(Pj,s )
conclude that
= κ o,IG
i T P j,s
* ρCiio d ρ+
0 Ki∞,IS (T Ps ) = φo,L
i
(T Ps )/φo,V
i
(T Ps )
 ρ(Pj,s )
, -
= fio,L (T Ps )/fio,V (T Ps ),

(E10)
× exp − Ciio d ρ , (E4a)
 0 
where fio,L (T Ps ) is given by (E4b), while fio,V (T Ps ) can be
  ρ(Pj,s ) f g  obtained by invoking the density limiting behavior Goii (TP) →
fio T Pj,s = κ o,IG
i T P j,s
 ρGoii / 1 + ρGoii d ρ 
 0  −2Bii (T ) in (E4b). Obviously, as expected
 ρ(Pj,s ) f
 
g  lim Ki∞,IS (T Ps ) = 1. (E11)
× exp − Goii / 1 + ρGoii d ρ , (E4b) T →Tc,j
 0  Ps,j →Pc,j
1 R. Battino and H. L. Clever, Chem. Rev. 66, 395 (1966); I. R. Krichevsky and 32 D. Peng and D. B. Robinson, Ind. Eng. Chem. Fundam. 15(1), 59 (1976).
L. Makarevi, Dokl. Akad. Nauk SSSR 175(1), 117 (1967); P. G. Debenedetti 33 W. Henry, Philos. Trans. R. Soc. London 93, 29 (1803).
and S. K. Kumar, AIChE J. 32, 1253 (1986); P. C. Ho, H. Bianchi, D. 34 A. A. Chialvo, S. Chialvo, and J. M. Simonson, in Gas-Expanded Liquids
A. Palmer, and R. H. Wood, J. Solution Chem. 29, 217 (2000); A. Ben- and Near-Critical Media: Green Chemistry and Engineering, edited by K.
Naim, J. Solution Chem. 30(5), 475 (2001); A. D. Pelton, G. Eriksson, and W. Hutchenson, A. M. Scurto, and B. Subramaniam (American Chemical
C. W. Bale, CALPHAD: Comput. Coupling Phase Diagrams Thermochem. Society, 2009), Vol. 1006, p. 66.
33(4), 679 (2009). 35 J. G. Kirkwood and F. P. Buff, J. Chem. Phys. 19, 774 (1951).
2 D. A. Jonah, Adv. Chem. Ser. 204, 395 (1983); J. Sedlbauer and R. H. Wood, 36 A. A. Chialvo, P. T. Cummings, J. M. Simonson, and R. E. Mesmer, J.
J. Phys. Chem. B 108(31), 11838 (2004). Chem. Phys. 110, 1075 (1999).
3 A. A. Chialvo and P. T. Cummings, AIChE J. 40, 1558 (1994). 37 A. A. Chialvo, P. G. Kusalik, P. T. Cummings, J. M. Simonson, and R.
4 E. Wilhelm, J. Solution Chem. 1, 1004–1061 (2015).
E. Mesmer, J. Phys.: Condens. Matter 12(15), 3585 (2000).
5 Z. Knez and E. Weidner, Curr. Opin. Solid State Mater. Sci. 7(4-5), 353 38 J. P. Hansen and I. R. McDonald, Theory of Simple Liquids, 2nd ed.
(2003). (Academic Press, New York, 1986).
6 G. Ferrentino, D. Barletta, F. Donsı̀, G. Ferrari, and M. Poletto, Ind. Eng. 39 A. A. Chialvo and P. T. Cummings, Mol. Phys. 84, 41 (1995).
Chem. Res. 49(6), 2992 (2010). 40 D. A. Jonah, Chem. Eng. Sci. 41(9), 2261 (1986); A. C. Chialvo, Can. J.
7 K. S. Lackner, S. Brennan, J. M. Matter, A. H. A. Park, A. Wright, and
Chem.-Revue Canadienne De Chimie 70(6), 1645 (1992).
B. van der Zwaan, Proc. Natl. Acad. Sci. U. S. A. 109(33), 13156 (2012). 41 J. M. Prausnitz, R. N. Lichtenthaler, and E. G. de Azevedo, Molecular Ther-
8 Z. Knez and M. Skerget, J. Supercrit. Fluids 20(2), 131 (2001).
modynamics of Fluid Phase Equilibria, 2nd ed. (Prentice-Hall, Englewood
9 Y. M. Monroy, R. A. F. Rodrigues, A. Sartoratto, and F. A. Cabral, J.
Cliffs, 1986).
Supercrit. Fluids 107, 250 (2016). 42 E. Wilhelm, J. Therm. Anal. Calorim. 108(2), 547 (2012).
10 J. Sedlbauer, J. P. O’Connell, and R. H. Wood, Chem. Geol. 163(1-4), 43 43 J. R. Macdonald, Rev. Mod. Phys. 38(4), 669 (1966).
(2000); D. Koschel, J. Y. Coxam, and V. Majer, Ind. Eng. Chem. Res. 52(40), 44 J. M. Prausnitz, in Advances in Chemical Engineering, edited by T. B. Drew,
14483 (2013); T. Driesner, in Thermodynamics of Geothermal Fluids, edited G. R. Cokelet, J. W. Hoopes, and T. Vermeulen (Academic Press, 1968),
by A. Stefansson, T. Driesner, and P. Benezeth (2013), Vol. 76, pp. 5; A. Vol. 7, p. 139; R. E. Gibbs and H. C. van Ness, Ind. Eng. Chem. Fundam.
V. Plyasunov, Geochimica Et Cosmochimica Acta 168, 236 (2015). 10(2), 312 (1971); J. J. Carroll and A. E. Mather, J. Solution Chem. 21(7),
11 A. Ben-Naim and Y. Marcus, J. Chem. Phys. 81(4), 2016 (1984); A. Ben-
607 (1992); E. Wilhelm, in Developments and Applications in Solubility
Naim, Solvation Thermodynamics (Plenum Press, New York, 1987). (The Royal Society of Chemistry, 2007), p. 3; I. Tosun, in The Thermody-
12 K. Kojima, S. J. Zhang, and T. Hiaki, Fluid Phase Equilib. 131(1-2), 145
namics of Phase and Reaction Equilibria (Elsevier, Amsterdam, 2013), p.
(1997); E. Wilhelm, Thermochim. Acta 300(1-2), 159 (1997); E. Schuhfried, 447; E. Wilhelm and R. Battino, in Volume Properties: Liquids, Solutions
F. Biasioli, E. Aprea, L. Cappellin, C. Soukoulis, A. Ferrigno, T. D. Mark, and Vapours, edited by T. L. Emmerich Wilhelm (The Royal Society of
and F. Gasperi, Chemosphere 83(3), 311 (2011); Q. Q. Xu, B. G. Su, X. Chemistry, 2015), p. 273; F. Y. Jou, A. E. Mather, and K. A. G. Schmidt, J.
Y. Luo, H. B. Xing, Z. B. Bao, Q. W. Yang, Y. W. Yang, and Q. L. Ren, Chem. Eng. Data 62(9), 2761 (2017).
Anal. Chem. 84(21), 9109 (2012); W. Afzal, B. Yoo, and J. M. Prausnitz, 45 E. Wilhelm, Crit. Rev. Anal. Chem. 16(2), 129 (1985).
Ind. Eng. Chem. Res. 51 (11), 4433 (2012). 46 A. A. Chialvo and L. Vlcek, Pure Appl. Chem. 88(3), 163 (2016).
13 J. P. O’Connell, A. V. Sharygin, and R. H. Wood, Ind. Eng. Chem. Res. 35, 47 J. K. Percus and G. J. Yevick, Phys. Rev. 110, 1 (1958).
2808 (1996). 48 D. B. McGuigan and P. A. Monson, Fluid Phase Equilib. 57, 227 (1990).
14 A. A. Chialvo, in Fluctuation Theory of Solutions: Applications in Chem- 49 A. Lotfi and J. Fischer, Mol. Phys. 66, 199 (1989).
istry, Chemical Engineering and Biophysics, edited by E. Matteoli, J. 50 J. Fischer, M. Bohn, and M. Gebauer, Fluid Phase Equilib. 17(1), 131
P. O’Connell, and P. E. Smith (CRC Press, Boca Raton, 2013), p. 191. (1984); A. Lotfi, “Molekulardynamische Simulationen an Fluiden: Phasen-
15 A. A. Chialvo, Fluid Phase Equilib. (in press).
gleichgewicht und Verdampfung,” Ph.D. dissertation (Ruhr-Universität,
16 I. R. Krichevsky and J. S. Kasarnovsky, J. Am. Chem. Soc. 57, 2168 (1935).
1993).
17 I. R. Krichevsky and A. A. Ilinskaya, Acta Physicochim. 20, 327 (1945). 51 We should not expect the Lennard-Jones model to be an accurate repre-
18 K. P. Shukla, A. A. Chialvo, and J. M. Haile, Ind. Eng. Chem. Res. 27, 664
sentation of this binary mixture at low temperature as a result of likely
(1988); L. Vlcek, A. A. Chialvo, and D. R. Cole, J. Phys. Chem. B 115(27), significant quantum effects from the infinitely dilute molecular hydrogen
8775 (2011). [Q. U. Wang, J. K. Johnson, and J. Q. Broughton, Mol. Phys. 89(4), 1105
19 M. M. Abbott and K. K. Nass, ACS Symp. Ser. 300, 2 (1986).
(1996)].
20 A. Ben-Naim, Am. J. Phys. 55(8), 725 (1987). 52 A. A. Chialvo and J. Horita, J. Chem. Phys. 119(8), 4458 (2003).
21 J. P. O’Connell and J. M. Haile, Thermodynamics: Fundamentals for 53 J. Fischer, D. Möller, A. A. Chialvo, and J. M. Haile, Fluid Phase Equilib.
Applications (Cambridge University Press, New York, 2005). 48, 161 (1989).
22 While the Lewis-Randall characterization is the most frequently found in 54 J. H. Dymond and E. B. Smith, The Virial Coefficients of Pure Gases and
the chemical engineering literature, this standard state is also known as the Mixtures: A Critical Compilation (Clarendon Press, 1980).
“Symmetric Ideality” Ben-Naim, 1992 #14950; Ben-Naim, 2006 #386. 55 M. Orentlicher and J. M. Prausnitz, Chem. Eng. Sci. 19(10), 775 (1964).
23 J. M. Haile, Fluid Phase Equilib. 26, 103 (1986); A. A. Chialvo, J. Chem. 56 E. Wilhelm, in Experimental Thermodynamics, edited by R. D. Weir and T.
Phys. 92, 673 (1990). W. De Loos (Elsevier, 2005), Vol. 7, p. 137.

24 A. A. Chialvo, Mol. Phys. 73(1), 127 (1991). 57 S. X. Hou, G. C. Maitland, and J. P. M. Trusler, J. Supercrit. Fluids 73, 87
25 A. A. Chialvo, in Advances in Thermodynamics, edited by E. Matteoli and (2013).
58 P. G. Debenedetti and R. S. Mohamed, J. Chem. Phys. 90, 4528 (1989).
G. A. Mansoori (Taylor & Francis, New York, 1990), Vol. 2, p. 131.
26 A. A. Chialvo, S. Chialvo, J. M. Simonson, and Y. V. Kalyuzhnyi, J. Chem. 59 J. M. H. Levelt Sengers, J. Supercrit. Fluids 4, 215 (1991).
60 W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical
Phys. 128(21), 214512 (2008).
27 Y. M. Monroy, R. A. F. Rodrigues, A. Sartoratto, and F. A. Cabral, J. Super- Recipes in FORTRAN: The Art of Scientific Computing, 2nd ed. (Cambridge
crit. Fluids 118, 11 (2016); F. Chemat, N. Rombaut, A. Meullemiestre, University Press, New York, 1993).
61 A. Dhima, J.-C. de Hemptinne, and J. Jose, Ind. Eng. Chem. Res. 38(8),
M. Turk, S. Perino, A. S. Fabiano-Tixier, and M. Abert-Vian, Innovative
Food Sci. Emerging Technol. 41, 357 (2017). 3144 (1999).
28 L. Chahen, T. Huard, L. Cuccia, V. Cuzuel, J. Dugay, V. Pichon, J. Vial, 62 E. Bender, U. Klein, W. P. Schmitt, and J. M. Prausnitz, Fluid Phase Equilib.
C. Gouedard, L. Bonnard, N. Cellier, and P. L. Carrette, Int. J. Greenhouse 15(3), 241 (1984).
63 R. D. Deshmukh and A. E. Mather, Fluid Phase Equilib. 35(1-3), 313
Gas Control 51, 305 (2016); M. Lucquiaud and J. Gibbins, Chem. Eng. Res.
Des. 89(9), 1553 (2011). (1987).
29 L. J. Guo and H. Jin, Int. J. Hydrogen Energy 38(29), 12953 (2013); Z. Knez, 64 E. Wilhelm, Thermochim. Acta 162(1), 43 (1990).
65 E. Wilhelm, Pure Appl. Chem. 57(2), 303 (1985).
E. Markocic, M. K. Hrncic, M. Ravber, and M. Skerget, J. Supercrit. Fluids
66 A. Ben-Naim, Molecular Theory of Solutions (Oxford University Press,
96, 46 (2015).
30 In this context, the term microstructure actually denotes volume integrals Oxford, 2006).
67 This expression can also be derived from the virial theorem as done else-
over pair correlation functions defining the Kirkwood-Buff integrals.
31 H. C. Van Ness and M. M. Abbott, Classical Thermodynamics of Nonelec- where [A. A. Chialvo, P. T. Cummings, and Y. V. Kalyuzhnyi, AIChE J.
trolyte Solutions (McGraw Hill, New York, 1982). 44(3), 667 (1998)].
68 This is a direct consequence of Widom’s potential distribution theorem for 71 J.L. Alvarez and R. F. Prini, J. Solution Chem. 37(10), 1379 (2008).
a non-interacting particle [B. Widom, J. Phys. Chem. 86(6), 869 (1982)]. 72 R. Fernandez-Prini, J. L. Alvarez, and A. H. Harvey, J. Phys. Chem. Ref.
69 C. A. Cerdeirina and B. Widom, J. Phys. Chem. B 120(51), 13144 Data 32(2), 903 (2003).
(2016). 73 An alternative compelling argument is given in Sec. 3.22 of Ben-Naim’s
70 J. Alvarez and R. Fernandez-Prini, Fluid Phase Equilib. 66, 309 (1991). “Solvation Thermodynamics,” Ref. 11.

Gas Solubility in Dilute Solutions: A Novel Molecular Thermodynamic Perspective

Uploaded by

Copyright:

Available Formats

Gas Solubility in Dilute Solutions: A Novel Molecular Thermodynamic Perspective

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gas Solubility in Dilute Solutions: A Novel Molecular Thermodynamic Perspective

Uploaded by

Copyright:

Available Formats

Gas solubility in dilute solutions: A novel molecular thermodynamic perspective

Articles you may be interested in

Comprehensive representation of the Lennard-Jones equation of state based on molecular dynamics

Communication: Glass transition and melting lines of an ionic liquid

High-density equation of state for a two-dimensional Lennard-Jones solid

Orientational ordering of lamellar structures on closed surfaces

Gas solubility in dilute solutions: A novel molecular

We present an explicit molecular-based interpretation of the thermodynamic phase equilibrium under-

I. INTRODUCTION Since the study of gas solubility entails dealing with

0021-9606/2018/148(17)/174502/16/$30.00 148, 174502-1 Published by AIP Publishing.

f̂iL (TPxi ) = xi γiHL (TPxi ) HIS

corresponding radial pair distribution function gαβ ⊗

extensively discussed elsewhere,3,14,36,37 allows the segrega- where (∂P/∂xi )∞

a real solution as discussed in detail in Appendix A. As we

C. Molecular-based inspired macroscopic correlations

The analysis based on the IE calculations (vide infra

FIG. 3. Isothermal-composition dependence of the modeling function

system as described by the integral equation calculations involving the

VI. DISCUSSION AND OUTLOOK

where Bij (T ) = lim Gij⊗ (TPxi ) and δij (T ) = lim ∆∞

dynamic self-consistent second-order truncated composition ∞ IS

eled novel interdependences between the solvation parameters IS,Gojj =G∞

The author expresses his gratitude to Professor Peter A.  

PY code used in the reported calculations and to Dr. Sebastian   −1   −1

where IG i superscript denotes the ideal gas behavior of the

with ln γiHL (TPxi ) = ln γiLR (TPxi ) − ln γiLR,∞ (TP) and, con-

We have also indicated that ρoj ∆∞ ij can be determined as

where Bi⊗ (TP) is the osmotic second virial coefficient 69 so

f̂iV (TPyi ) = yi φ̂Vi (TPyi ) P

where δij (T ) = lim ∆∞

APPENDIX D: SIMULTANEOUS SOLUTION

= φ̂∞,L (T Ps )/φ̂∞,V (T Ps ) (E5a)

or in the equivalent form, after invoking δij (T ) = Bii (T ) + = 1. (E7)

 ρ(Pj,s )    Finally, for the i-solute in the ideal solution

Phys. 92, 673 (1990). W. De Loos (Elsevier, 2005), Vol. 7, p. 137.

You might also like

FIG. 3. Isothermal-composition dependence of the modeling function

The author expresses his gratitude to Professor Peter A.

PY code used in the reported calculations and to Dr. Sebastian −1 −1

 ρ(Pj,s )  Finally, for the i-solute in the ideal solution