Fluid Phase Equilibria 432 (2017) 28e44

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

Recent advances in modeling the vapor-liquid equilibrium of mixed

working fluids
Wen Su, Li Zhao*, Shuai Deng
Key Laboratory of Efficient Utilization of Low and Medium Grade Energy, Tianjin University, MOE, Tianjin, 300072, China

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

Article history: Vapor-liquid equilibrium (VLE) properties of mixed working fluids are essential for the study of organic
Received 26 July 2016 fluids for thermodynamic cycles such as heat pumps or organic Rankine cycles. Thus, the typical and
Received in revised form recent models for the VLE calculation of mixed working fluids are reviewed in this paper. The most
7 October 2016
popular cubic equations of state (EOSs) and two mixing rules, namely the Van der Waals and the excess
Accepted 18 October 2016
free energy, are summarized, respectively. VLE data of 13 alternative working fluids at different tem-
Available online 18 October 2016
peratures are employed to exhibit the capacities and limitations of three predictive models. The calcu-
lated VLE results show that the model with Van der Waals mixing rules is only valid for some types of
Keywords:
Vapor-liquid equilibrium
mixture, while the accuracy of the model with excess free energy mixing rules is greatly influenced by
Mixed working fluids the UNIFAC group parameters, which are usually fitted from the VLE data instead of the activity coef-
Cubic equations of state ficient data. Thus, a generalized model for mixed working fluids should be derived from the molecular
Van der Waals mixing rules theory or the group contribution method, in which the group parameters are fitted from the large VLE
Excess free energy mixing rules data.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction Although many researchers have proposed suitable pure fluids

Organic working fluids have been widely used in several
application domains, such as refrigeration, heat pump and organic
Rankine cycle. The physical properties of working fluids directly
determine the thermodynamic efficiency of system, the design of
system components, the stability and safety of system. Thus, a high-
efficient, safe and economical operation of thermodynamic cycle is
highly dependent on working fluids with appropriate properties. In
addition to the thermodynamic properties of working fluids,
environmental properties, such as ozone depletion potential (ODP)
and global warming potential (GWP), have to be considered in the
screening of working fluids with growing concerns regarding the
depletion of the ozone layer and the greenhouse effect [1]. Recently,
due to high ODP and high GWP, the traditional organic working
fluids chlorofluorocarbon (CFC) have been phased out in accor-
dance with the protocol of Montreal and the hydrochlorofluoro-
carbon (HCFC) will also be phased out in 2040. Therefore, it's urgent
to search the high-efficient fluids which are friendly to the
environment.
* Corresponding author. Tel./fax: þ86 022 27404188.
E-mail address: jons@tju.edu.cn (L. Zhao).
for different applications, few pure fluids, which are free of all
environmental and safety concerns, chemically and thermally sta-
ble, and perform efficiently, have been found [2e4]. Thus, mixtures,
whose individual drawbacks of the components can be compen-
sated by each other, are recommended as candidates of working
fluids. Basically, the mixed working fluids are preferable to pure
working fluids on account of energy saving and the flexibility of
operation [5e7].
A reliable thermodynamic model is required for evaluating the
performance of a given mixture in a thermodynamic cycle. In
general, the experimental data of vapor-liquid equilibrium (VLE) for
mixed working fluids are used to establish the highly accurate
Helmholtz energy equation, which is widely applied to calculate
the thermodynamic properties of corresponding mixture [8].
Furthermore, accurate VLE data are essential for designing and
modeling thermodynamic processes, which include the evapora-
tion and condensation of working fluids, the constituent separation
of mixtures [9]. For the mixed working fluids being in VLE, they
may produce different types of phase behavior, because of the
different molecular interactions and phase conditions. Although
experiments can provide accurate data at specific phase conditions
for mixtures, such data are limited and can't be expected to meet
the expanding industrial needs for thermodynamic design and

http://dx.doi.org/10.1016/j.fluid.2016.10.016
0378-3812/© 2016 Elsevier B.V. All rights reserved.
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44 29

development [10]. Besides, given a huge number of mixed working condition is that the temperature, pressure and fugacity of each
fluids, it's impossible to provide VLE information for every mixture species should be the same in liquid and vapor phases [15]. That is,
by the time-consuming phase equilibrium experiment with high
costs. Therefore, many VLE models have been proposed to provide TiL ¼ TiV ¼ T (1)
predictions of VLE for a wide variety of mixtures exhibiting varied
phase behaviors, based on the limited experimental data [11e14].
PiL ¼ PiV ¼ P (2)
Such calculation models can be categorized into two ap-
proaches. The first approach utilizes an equation of state (EOS) to
describe the phase behavior for both the liquid and vapor phases. fiL ðT; P; zi Þ ¼ fiV ðT; P; yi Þ (3)
The second approach utilizes an activity coefficient model to
describe the liquid phase, while a different model, usually an EOS, is Where superscripts L and V represent liquid and vapor, respectively.
used to describe the vapor phase [10]. Existing studies have zi, yi sequentially denote the molar fraction of species i in the liquid
demonstrated that the EOS model can be successfully applied to and vapor phases. The key of VLE calculation is to express the
correlate the VLE data of mixed working fluids over broad ranges of fugacity as a function of temperature, pressure and composition. An
temperature and pressure [15e18]. In contrast, the activity coeffi- EOS is generally used to obtain the fugacity of species i in the vapor
cient model has been applied successfully at low temperatures phase, and the generalized form is
where the liquid phase is relatively incompressible. However, this ! Z∞"  #
model is deficient in describing the phase behavior at high pres- f V ðT; P; yi Þ 1 vðPV=RTÞ dV
sures [10,15]. Thus, for the above reasons, this paper is focused on ln i ¼ 1  ln Z V
yi P RT vNi T;V;Njsi V
the EOS model. V
For the EOS model, cubic EOSs, such as SoaveeRedlicheKwong (4)
(SRK) and PengeRobinson (PR), are widely used to predict the VLE
of mixed working fluids, due to their inherent simplicity and effi- Where V is the total volume, ZV is the vapor compressibility factor
ciency [15]. A compositional dependence has to be introduced to computed from the EOS. However, there are two different methods
account for the molecular interactions in mixtures, in order to to describe the liquid phase: either the same EOS used for the vapor
predict the phase behavior using cubic EOSs. All extensions of cubic phase is also used for the liquid phase or the activity coefficient
EOSs to mixtures are, at least partially, empirical in nature, because method is employed.
there is no exact statistical mechanical solution relating the prop- For the activity coefficient method, the fugacity of species i in
erties of dense fluids to their intermolecular potentials, and the the liquid phase is obtained by
detailed information on such intermolecular potentials is unavai-
lable [19]. Therefore, a variety of mixing rules have been proposed fiL ðT; P; zi Þ ¼ zi gi fiPL ðT; PÞ (5)
to determine the parameters of cubic EOS for mixtures over the
years [15,19]. Generally speaking, the proposed mixing rules can be Where fiL ðT; PÞ is the fugacity of pure component i as a liquid at the
classified into two categories, namely, Van der Waals (VDW) mixing temperature and pressure of the mixture, and can be calculated
rules and excess free energy (GE) mixing rules. The VDW mixing from EOSs. At low temperature and pressure, fiL ðT; PÞ is approxi-
rules have been used to describe the VLE of mixed working fluids mately equal to the vapor pressure of pure component. For the
for decades, because of their simplification and relative accuracy for activity coefficient gi, many local composition models, such as
many mixture systems [20e22]. A route for developing excess free UNIFAC, NRTL and Wilson, have been proposed to relate gi to
energy mixing rules is to combine a cubic EOS with an activity composition zi [19]. However, these models are only valid at low
coefficient model, often a local composition model like Wilson, pressure and temperature where the liquid phase is relatively
NRTL, UNIQUAC or UNIFAC [12,23e25]. The excess free energy incompressible. By contrast, the EOS model can be applied to the
mixing rules have enhanced dramatically the range of cubic EOS's liquid phase over a wide range of temperatures and pressures. With
applicability and can give a good prediction for high pressure VLE of the same EOS used for the vapor phase, the fugacity of a species in a
mixtures. Furthermore, the group contribution activity coefficient liquid mixture is computed from
model UNIFAC has been widely used to develop completely pre- !
dictive models for mixed working fluids [25]. Z∞"  #
fiL ðT; P; zi Þ 1 vðPV=RTÞ dV
The models, which can be utilized to predict the VLE of mix- ln ¼ 1  ln Z L
zi P RT vNi T;V;Njsi V
tures, usually consist of two parts including cubic EOSs and mixing V
rules. This paper presents a state-of-the-art review of these models (6)
and discusses the capabilities and limitations of three predictive
models in exhibiting the phase behaviors of mixed working fluids. Where the liquid compressibility factor ZL is derived from the EOS.
The methodology of VLE is presented to demonstrate the process of This equation differs from Eq. (4) only in that the smallest volume,
VLE calculations in the next section. In Section 3, commonly-used which is the liquid phase solution to the EOS, is used to calculate
cubic EOSs are summarized and discussed. Two mixing rules the fugacity. Due to the applicability of Eq. (6) over a wide range of
involved the van der Waals and the excess free energy are reviewed temperatures and pressures in calculating the liquid fugacity, the
respectively in Section 4. In Section 5, the performances of three EOS model has been extensively used to predict the VLE of mixed
completely predictive models in predicting the VLE of mixed working fluids.
working fluids are discussed by comparing the calculated results Numerous EOSs have been proposed to establish the pressure-
with the experimental data. The knowledge gaps and development volume- temperature relationship of organic fluids. In general,
directions of VLE models are presented in Section 6. Conclusions of these proposed EOSs can be categorized into the “analytical” and
this review on modeling the VLE of mixtures are given in Section 7. “non-analytic” types [15]. The former means that when T and P are
specified, V can be found analytically. This type of EOSs is repre-
2. Methodology of VLE sented by the cubic EOSs, such as SRK and PR. However, V is
calculated numerically in the latter approach. In order to describe
For the mixture being in VLE, the required thermodynamic the complexity of property behavior, a large number of fitted
30 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44
parameters are introduced into the non-analytic EOSs, such as the
Benedict-Webb-Rubin (BWR) EOS and the Wagner formulation
[15]. Due to the wide spread use and simple form of cubic EOSs, the
review on modeling the VLE of mixed working fluids is limited to
cubic EOSs. Fig. 1 shows the flow diagram of the VLE calculation
with the cubic EOS. It can be seen that when given the specified
value of phase equilibrium temperature T and liquid mole fraction
zi, the phase equilibrium pressure P and vapor mole fraction yi are
iteratively calculated based on the ratio of fugacity between the
liquid and the vapor. The EOS is used to calculate the fugacity of
each species for liquid and vapor phases in the process of VLE
calculation. The cubic EOS parameters of mixture are obtained from
the fluid parameters and molar fraction of species based on the
mixing rule, in order to improve the accuracy of prediction on the
VLE of mixed working fluids. So far, two types of mixing rules,
namely Van der Waals and excess free energy, are frequently used
for the VLE prediction of working fluids.
Fig. 1. Flow diagram of the VLE calculation.
3. Cubic EOSs
Many cubic EOSs are available in the literature. A recent
comprehensive review on the cubic EOSs is presented by Valder-
rama [26]. For cubic EOSs, they are of third degree when solved
with respect to volume. It means that the volume can be found
analytically rather than only numerically at specified values of T and
P. To the best of author's knowledge, almost all cubic EOSs express
pressure as the sum of two terms, a repulsion pressure PR and an
attraction pressure PA.
P ¼ PR þ PA (7)
The equations of Van der Waals (VDW), SRK and PR are exam-
ples. For the repulsion pressure, it's always obtained from the Van
der Waals hard sphere equation. That is
RT
PR ¼ (8)
vb
Where v is the molar volume, the co-volume parameter b is related
to the size of the hard sphere. As for the attraction pressure, it can
be generally expressed as [27].
a
PA ¼  (9)
gðvÞ
Where g(v) is a function of the molar volume and the co-volume
parameter. The energy parameter a, which can be regarded as a
measure of the intermolecular attraction force, is a function of
temperature.
Ever since the appearance of the VDW equation in 1873, many
authors have successfully developed different cubic EOSs to
establish the relation between P, v and T. The most well-known
cubic EOSs together with the expressions often used for esti-
mating their parameters are shown in Table 1. Many of these EOSs
employ the repulsive term of VDW, while various expressions for
the attractive term. Furthermore, considering that any given two-
parameter EOS is intrinsically unable to quantitatively describe
the P-v-T properties of complex fluids and their asymmetric mix-
tures, the third parameter c has been introduced into cubic EOSs,
such as RK-PR, volume-translated PengRobinson (VTPR), in order
to improve the accuracy of the prediction in the liquid phase. For
the parameters involved in the cubic EOSs, two approaches, namely
corresponding states principle and data regression, are usually used
to get the values [15]. For the corresponding states principle, the
energy and co-volume parameters are usually determined by the
following two relationships, so that the EOSs can predict the critical
temperature and pressure of fluids accurately.
 
vP
¼0 (10)
vv Tc
!
v2 P
¼0 (11)
vv2
Tc
Applying a specified cubic EOS at the critical point where the
first and second derivatives of pressure with respect to volume, one
can obtain expressions for a and b in terms of the critical properties.
The VDW and RK EOSs are examples. For the data regression, the
parameter values are frequently obtained by fitting the data of
vapor pressure and saturated liquid density. The acentric factor ⍵,
which represents a measure of the acentricity of the molecule, is
usually employed to establish the relationship between the energy
parameter and the temperature in this approach, as shown in
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44 31

Table 1
The most important cubic EOSs for VLE calculation.

EOS Equation Parameters of cubic EOSs

VdW [28] P¼ RT  a ðRTc Þ2 1 RTc

vb v2 a ¼ 27
64 Pc ; b ¼ 8 Pc
RK [29] P¼ RT  a R2 Tc2:5
vb T 1=2 vðvþbÞ a¼ 0:42748 Pc ; b ¼ 0:08664 RT
Pc
c

RT  a
pffiffiffiffiffi 2
SRK [30] P ¼ vb a ¼ ac ½1 þ mð1  Tr Þ
vðvþbÞ

m ¼ 0:48 þ 1:574u  0:176u2

R2 Tc2 RTc
ac ¼ 0:42748 ; b ¼ 0:08664
Pc Pc
RT  a
pffiffiffiffiffi 2
PR [27] P ¼ vb a ¼ ac ½1 þ mð1  Tr Þ
vðvþbÞþbðvbÞ

m ¼ 0:37464 þ 1:54226u  0:26992u2

R2 Tc2 RTc
ac ¼ 0:45724 ; b ¼ 0:07780
Pc Pc
RT  a
pffiffiffiffiffi
PRSV [19] P ¼ vb a ¼ ac ½1 þ mð1  Tr Þ2
vðvþbÞþbðvbÞ

m ¼ m0 þ m1 ð1 þ Tr0:5 Þð0:7  Tr Þ
m0 ¼ 0:378893 þ 1:4897153u  0:1713184u2 þ 0:0196554u3
R2 Tc2 RTc
ac ¼ 0:45724 ; b ¼ 0:07780
Pc Pc
 
CSD [31] RT 1 þ y þ y2  y3 a ðRTc Þ2
P¼  a ¼ 4a ; 4a ¼ a0 expða1 Tr þ a2 Tr2 Þ
v 1  y3 vðv þ bÞ Pc
b RTc
y¼ b ¼ 4b ; 4 ¼ b0 þ b1 Tr þ b2 Tr2
4v Pc b
PT [32] RT 
P ¼ vb a 2
a ¼ Ua ðRTPcc Þ ½1 þ Fð1  Tr Þ2 ; b ¼ Ub RT
1=2
vðvbÞþcðvþbÞ Pc ;
c
c ¼ Uc RT
Pc
c

RK-PR [33] RT  a   k
P ¼ vb 3 ðRTc Þ2
RTc
ðvþcbÞþ vþ1c
a ¼ Ua ; b ¼ Ub ;
1þc
b 2 þ Tr Pc Pc

c ¼ d1 þ d2 ðd3  Zc Þd4 þ d5 ðd3  Zc Þd6

VTPR [34] RT 
P ¼ vþcb a
R2 Tc2:5 RT
ðvþcÞðvþcþbÞþbðvþcbÞ a ¼ 0:45724 aðTÞ; b ¼ 0:0778 c
Pc Pc
aðTÞ ¼ TrNðM1Þ expðLð1  TrNM ÞÞ; c ¼ vpR  vexp at Tr ¼ 0:7
Note: some variables used in the parameter expression of cubic EOSs are not specified. Their values are related with the fluids or involved with complex calculation. Please
consult the corresponding reference for these variables.

Table 1. Meanwhile, owing to the fact that two-parameter EOSs are

widely used in engineering calculation of VLE for mixed working X
N X
N

fluids, mixing rules for energy and co-volume parameters are Bm ¼ xi xj Bij (12)
i¼1 j¼1
reviewed in the next section.
Since Eq. (12) is exact, it can be considered low-density
boundary conditions that should be satisfied for mixtures by
4. Mixing rules other EOSs when expanded into the virial form [15]. Thus, in order
to satisfy the quadratic mixing rule for the second virial coefficient,
The concept of mixing rule is that the mixture can be assumed as the quadratic dependence on mole fractions of mixture parameters
a one fluid. For fixed compositions, the mixture properties and their is commonly used in the VDW mixing rules [15].
variations with T and P are the same as the pure fluid with appro-
priate parameter values. To describe all pure components as well as X
N X
N
mixtures, the mixture parameters must vary with composition so am ¼ xi xj aij (13)
that if the composition is actually for a pure component, the model i¼1 j¼1
describes that substance [15]. In general, there are two kinds of
widely used mixing rules, namely Van der Waals and excess free X
N X
N
energy, for energy and co-volume parameters of cubic EOSs in the bm ¼ xi xj bij (14)
VLE calculation of mixed working fluids. Thus, comprehensive re- i¼1 j¼1
views are presented for these mixing rules.
The following combining rules, namely, the geometric mean
rule for the cross energy and the arithmetic mean rule for the cross
co-volume parameter, are frequently used to obtain the cross co-
4.1. Van der Waals
efficients aij and bij from the corresponding pure component
parameters:
The first successful method of generalizing a pure fluid EOS to
mixtures is the one-fluid model proposed by van der Waals. The pffiffiffiffiffiffiffiffi 
aij ¼ ai aj 1  kij (15)
underlying assumption of this model is that the same EOS used for
pure fluids can be used for mixtures if a satisfactory way is found of
 
obtaining the mixture EOS parameters. One of the exact mixing bi þ bj  
rules is derived from statistical mechanics by virial EOS. For mix- bij ¼ 1  lij (16)
2
tures, the only composition dependence of the second virial co-
efficients Bm is given by Where kij denotes the binary interaction parameter for aij, lij is the
32 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44
binary interaction parameter for bij. Generally, for a specified
mixture, these interaction parameters can be accurately obtained
by fitting the measured phase equilibrium and volumetric data
[22]. Furthermore, lij is usually set to zero in the VLE calculation
for most mixtures, in order to simplify the binary interaction
parameters fitting process. As a result of that, the co-volume
parameter of mixtures can be derived by
X
N calculation results of mixtures, compared with the effect of kij.
bm ¼ x i bi (17)
i¼1
However, when the mixture is in high pressure phase equilib-
rium, the accuracy of VLE calculation can be improved, if lij is
determined by the experimental data [35].
For the binary interaction parameter kij, it's generally thought kij
is symmetric, i.e., kij equals kji. Furthermore, the value of kij has a
decisive effect on the phase behavior of mixtures. When fitting VLE
data with the VDW mixing rules, it is found that kij is approximately
zero for mixtures, which have small differences between the mo-
lecular structures of species, such as Hydrocarbon (HC) mixtures,
Hydrofluoro carbon (HFC) mixtures, whereas for some other mix-
tures such as HC with HFC, it is not only nonzero but also changes
with the compositions [29,30,35]. On the other hand, a theoretical
value of kij can be determined, owing to the fact that the geometric
mean rule employed for the cross-energy parameter can be derived
from the theory of Mie potential energy and London dispersion
forces [36]. However, the interaction parameters calculated via this
theoretical approach are often much higher than those actually
from a cubic EOS-based data fitting of the experimental data,
because there is no direct relation between the attractive part of the
intermolecular potential and the energy parameter of the cubic
EOS. Thus, such theoretical approach is usually used by adding
adjustable parameters. For example, a semi-empirical correlation of
kij, which is derived from the molecular theory, was proposed for
non-polar systems [36]. So far, according to the geometric mean
rule presented by Eq. (15), a variety of expressions have been
proposed to get the interaction parameter from molecular theories
and VLE experiments. Table 2 summarizes the proposed equations
for kij. It can be seen that the factors associated with kij involve the
temperature, the critical parameters, the composition, the acentric
factor, the ionization potential and the infinite dilution activity
coefficients. Some of these methods are applicable for various kinds
of mixtures, such as the linear relationship between kij and tem-
perature [37], the Margules-type and Van Laar-type expressions of
kij relating with composition [19,38]. However, more expressions of
kij are aimed at some types of mixtures. For instance, kij was pro-
posed as a function only of critical temperature and compressibility
factor for binary light hydrocarbon systems [39]. With regard to
HFC and HC mixtures, Chen et al. [40,41] supposed that each pure
component in a mixture has a mixing factor and kij can be calcu-
lated out by a simple operation of these two mixing factors.
Thereafter, the correlation was modified with the introduction of
acentric factors and the critical properties [42]. In 2016, Zhang et al.
[43] introduced the contribution of mixing factor to kij for mixtures
containing HCs, HFCs, PFCs, HFOs, CO2, RE170 and R13I1. Empirical
correlations were developed in terms of the absolute difference
between the acentric factors of the components and the critical
compressibility factor ratio, based on the fitted kij values of mix-
tures [44,45]. In addition, since 2004, a group contribution model,
called PPR78, has been proposed to determine kij from the mere
knowledge of chemical structures of molecules within the mixture
by Jaubert and Mutelet [11,46e48], in order to realize the accurate
prediction of interaction parameters for different mixtures. In the
model PPR78, the PR EOS is used and kij is a function of the tem-
perature, the pure components' critical temperatures, critical
pressures and acentric factors. So far, the PPR78 model is able to
represent the phase behavior of any fluid containing alkanes, al-
kenes, aromatic compounds, cycloalkanes, permanent gases, mer-
captans and water.
For the binary interaction parameter lij, it has the same sym-
metry characteristic as kij. However, lij has little effect on the VLE
Thus, lij is generally set to zero for mixtures and less attention is
paid to the research of lij. Yet, for non-ideal and high pressure phase
equilibrium, additional adjustable parameter lij can achieve a good
agreement between the calculation of VLE model and the experi-
mental data [15]. On the other hand, on the basis of Mie potential
energy and London dispersion theory, the cross co-volume can be
related to the hard sphere diameter [36]. Thus, many equations,
which are different from Eq. (16), have been proposed for cross co-
volume of mixtures, according to the hard sphere mixing rules. A
summary of cross co-volume bij is presented in Table 3. It can be
seen that most of the cross co-volume expressions are derived from
hard sphere mixing rules, such as the geometric mean of co-volume
[51], the method of Coutinho et al. [36] and the approach proposed
by Lee and Sandler [52]. However, real molecules are not hard
spheres. Therefore, some authors corrected these expressions with
temperature, composition and shape factor. For example, Weber
and Brunner [37] thought that the interaction parameter lij is a
linear function of temperature. Wang and Zhong [53] proposed a
composition dependent cross parameter for VLE calculations of
hydrogen containing systems at different temperatures. Han et al.
[54] introduced molecular shape factors to correct bij and derived a
theoretical formula, which could reflect the actual situations of
mixture systems.
4.2. Excess free energy
As mentioned above, the liquid behavior of highly non-ideal
system can be described by activity coefficient models. Mean-
while, considering the poor capacities of VDW mixing rules in
the VLE prediction of these systems, the excess free energy
mixing rules, which combine the strengths of the cubic EOSs and
the activity coefficient models, have been widely investigated,
since the publication of the Huron-Vidal (HV) mixing rule [12,29].
Many different models, called EOS/GE, have been available in a
very large number of literatures. In these models, the energy
parameter of cubic EOS is often derived by incorporating the
excess free energy into the EOS. Thus the cubic EOS with excess
free energy mixing rules can be applied to the VLE calculation of
polar and non-polar mixtures at both low and high pressures.
The starting point for deriving most EOS/GE models is the
equality of the excess Gibbs energy from an EOS and from an
explicit activity coefficient model at a suitable reference pressure
[55].
 EOS  Model
GE GE
¼ (18)
RT P RT P
Where a suitable reference pressure P can be the infinite pressure
or the zero pressure. The excess Gibbs energy from a cubic EOS is
calculated by
Table 2
Binary interaction parameter kij for cross-energy parameter.

Binary interaction parameter Comments References

kij ¼ k0ij þ k1ij 1000

T kij is a linear function of temperature. k0ij , k1ij are determined by the experimental data. [37]
 
kij is set to be an exponential function of temperature, in order to model the phase behavior of asymmetric alkane þ alkane [18]
TT
kij ¼ k0ij e
c1 binary systems.Tc1 is the critical temperature of the more volatile componen.
 
 CN  The function is recommended in the literature for methane(i) with higher n-alkanes (j). kij increases monotonically and tends [49]
 20j
kij ¼ 0:12 1  e asymptotically to a maximum of 0.12 for an infinite carbon number in the second compound.

0 1Zcij For binary light hydrocarbon systems. kij was proposed as a function only of critical temperature and compressibility factor. [39]
pffiffiffiffiffiffiffiffiffi
2 TCi TCj A Zci þZcj
kij ¼ 1  @ TCi þTCj Zcij ¼ 2
Thus, cross-energy parameter can be obtained without any adjustable parameter.

kij ¼ xi Kij þ xj Kji The Margules-type expression provides a good correlation of VLE data of highly non-ideal systems by adding an additional [19]

W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44

composition dependence and parameters to kij.
kij ¼ Kij Kji =ðxi Kij þ xj Kji Þ kij is dependent on the composition. Their relationship is expressed as Van Laar-type [38]
kij ¼ ki  kj kij is assumed to be the difference between the mixing factors of the pure components. ki is the mixing factor of the component [40]
ki ¼ 0:30ui þ 0:031n0:1
F =ui
with a higher vapor pressure and kj is the mixing factor of the component with a lower vapor pressure. The mixing factor of pure
HFC and HC can be determined by the acentric factor and the number of F.
! !
uj Pc;j
0:5
ui Pc;i
0:5 The interaction parameter kij was modified with the introduction of the acentric factors and the critical properties of the [42]
kij ¼ Tc;j ki  Tc;i kj components. In order to avoid the mistake of kij ¼ -kji, all involved components are arranged in a certain sequence.

kij ¼ ðki þ kj Þ=2 kii ¼ ki The binary interaction parameters of HFC mixtures are calculated by the average of mixing factors. [41]
kij ¼ wj ki þ wi kj Constant wi is the contribution coefficient of component j mixing with component i. kij can be determined by summing the [43]
contribution coefficient of each pure component times its contribution.
0:45 On the basis of the fitted kij values for refrigerant mixtures, the empirical function is correlated, in terms of the absolute
kij ¼ 0:6858 þ 0:002713 ui  uj  0:6842ðZci =Zcj Þ [45]
difference between the acentric factors of the components and the critical compressibility factor ratio.
kij ¼ 0:16339 þ 0:30504ðui  uj Þ  0:14921ðZci =Zcj Þ Based on the interaction parameters of HFC mixtures, the new correlation for kij can be developed. [44]
!
qffiffiffiffiffiffiffiffiffiffiffiffi The correlation relates kij to the ionization potentials of the component species according to the low temperature VLE data for [50]
kij ¼ 0:171 Ii  Ij ln IIij binary systems containing hydrogen, helium, and neon.
0 1n3 The theoretical combining rule is derived from interaction potential function and London-Mie theory. The adjustable parameter [36]
pffiffiffiffiffiffi 3
kij ¼ 1  A@
bi bj A A, which is determined by experimental data, is introduced to improve the prediction accuracy.
bij

qffiffiffi qffiffiffi
kij ¼ 1  12
bj ai
 12 bi aj b RT
þ 12 pjffiffiffiffiffiffi qi The equation is derived based on the HuroneVidal mixing rules. Adjustable parameters q1q2,q3 are obtained by fitting [17]
bi aj bj ai ai aj q
Tri2 Prj3
q
experimental data of each mixture.
kij ¼ Kij xi þ Kji xj kij is calculated by the composition, the pure component EOS parameters and infinite dilution activity coefficients [19]
vap
ai ½bj =bi  ðln g∞ ∞
ji þ ln fj ðT; Pi Þ  li Þ=εi 
Kij ¼ 1 
2ðai aj Þ1=2
2 0pffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffi12 3,2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 Group contribution model PPR78, is used to determine kij from the mere knowledge of chemical structures of molecules within [11]
ai ðTÞ aj ðTÞ ai ðTÞaj ðTÞ
6 A 7 the mixture. So far, the PPR78 model is able to represent the fluid phase behavior of any fluid containing alkanes, alkenes,
kij ¼ 4Eij ðTÞ  @  5 42 5
bi bj bi bj aromatic compounds, cycloalkanes, permanent gases, mercaptans and water.
 
  Bkl
1X
Ng X
Ng Akl 1
298:15
Eij ðTÞ ¼  ðaik  ajk Þðail  ajl ÞAkl
2 T
k¼1 l¼1

33
34 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44

Table 3
Cross co-volume parameter bij.

Cross co-volume parameter Comments References

qffiffiffiffiffiffiffiffi
bij ¼ bi bj The cross parameter is a geometric mean of co-volumes of species. [51]

bij ¼ ðbii
1=3 1=3
þ bij Þ3 =8 Owing to the fact that the co-volume is about proportional to the cubic of hard sphere [36]
diameter s, i.e.bfs3, the equation can be derived from the hard spherediameter additive.
bij ¼ ððbii
2=3 2=3
þ bij Þ=2Þ3=2 The expression is obtained based on the relation between the hard sphere diameter and co- [52]
volume.
 
ðb þb Þ l0ij , l1ij are determined by the experimental data. [37]
bij ¼ i 2 j 1  l0ij  l1ij 1000
T

bij ¼ ½1  lij þ ðlij  lji Þxi ððbii

1=3 1=3
þ bij Þ=2Þ3 The composition dependent cross parameter for the co-volume is recommended for VLE [53]
calculations of hydrogen containing systems at different temperatures.
0 13 The theoretical combining rule is derived from interaction potential function and London- [36]
1=3 1=3
ðbi þbj Þ bi þbj
bij ¼ 2 ð1  C1 Þ þ C2 @ 2
A Mie theory. The adjustable parameters C1, C2, which are determined by experimental data,
are introduced to improve the prediction accuracy.
!
4j 4i qj qi
qffiffiffiffiffiffiffiffi The molecular shape factors (q,4) are used to correct the geometric mean of co-volume [54]
4ij bij ¼ 4ij q bi bj parameters.
ij

PN * *
bm ¼ i¼1 ½xi xj ð1  mij Þbij þ mij bij bij ¼ ðbi þ bj Þ=2; b*ij ¼ f ðbi ; bj Þ bij is any one of the non-classical mixing rules for co-volume and mij can be considered a [51]
binary mixing parameter used to proportion the overall mixing rule. Furthermore, five
forms of parameter mij have been proposed

! parameters of the GE expression must be re-estimated using an EOS

GE ðT; p; xi Þ X with the HV mixing rule, because the excess Gibbs energy is very
¼ Zm ðT; p; xi Þ  xi Zi ðT; pÞ
RT
i
! can be very different from the value at the experimental pressure.
X
 ln Zm ðT; p; xi Þ  xi ln Zi ðT; pÞ
i the use of the pressure-dependent excess Gibbs energy in the EOS
0 1
vm ðT;p;x
Z iÞ vi ðT;p;x
Z iÞ
B Zm  1 X Zi  1 C
@ dv  xi dvA
vm i
vi
∞ ∞
Meanwhile, there exists a rigorous thermodynamic relation
between the excess Gibbs energy GE and the activity coefficient gi.
Their relationship is expressed by
X Where AE is the excess Helmholtz free energy. The first of these
GE ¼ RT xi ln gi
i
Thus, an explicit activity coefficient model, such as UNIFAC,
Wilson or NRTL can be used to get the value of GE at a suitable
reference pressure. Generally, a linear mixing rule, given by Eq. (17),
is employed to get the co-volume parameter of cubic EOSs in EOS/
GE models. When submitting Eq. (17) into Eq. (19), the energy
parameter can be derived in terms of the excess free energy GE.
Meanwhile, the constants involved in the expression of energy
parameter are always related with the used EOS [15].
Many GE mixing rules have been proposed since the HV mixing
rule. Meanwhile, a large number of literatures have been published
to summarize the existing GE mixing rules and discuss the capac-
ities and limitations of these rules [56]. Thus, only the most suc-
cessful mixing rules for mixed working fluids are presented in this
paper. They are listed in Table 4. In general, these mixing rules can
be categorized as those following the infinite or zero reference
pressure assumption and those not having a specific reference
pressure. As the first successful combination of an EOS and excess
free energy, the basic assumption of HV mixing rule is the use of the
infinite pressure as the reference pressure [57]. For non-ideal
mixtures, HV shows good correlation capabilities but is not satis-
factory for extrapolation over a range of temperatures. Further-
more, it was observed that HV is superior to the VDW for
correlating the VLE data of highly non-ideal mixtures. However, it's
possible to reduce the HV to the VDW by choosing specific algebraic
forms for GE. On the other hand, one can't use the local composition
model parameters obtained at low pressure, and the interaction
pressure dependent. The excess Gibbs energy at infinite pressure
Consequently, a fundamental shortcoming of the HV mixing rule is
rather than the excess Helmholtz energy, which is much less
pressure dependent.
The Wong-Sandler (WS) mixing rule is also based on the infinite
reference pressure but with significant modifications compared to
(19) that in HV [58]. It's assumed that the following equation is
approximately true:
AE ðT; P ¼ ∞; xÞzAE ðT; low P; xÞzGE ðT; low P; xÞ (21)
(20)
equalities is based on the relative insensitivity of the excess
Helmholtz energy with pressure. The second of these equalities
results from the relation:
GE ¼ AE þ PvE (22)
E
At low pressure, the term Pv is neglected so that the second
equality of Eq. (21) can be derived. As GE is highly dependent on
pressure, the WS mixing rule is derived by equating the excess
Helmholtz energy AE to that of an EOS at infinite pressure. Ac-
cording to the equalities given by Eq. (21), the WS mixing rule uses
VLE information only at low pressure, and it can also be used to
predict the VLE of mixtures at high pressure. The low-pressure
activity coefficient model parameters reported in data banks,
such as the Dortmund Data Bank (DDB) [59], could be used directly
with good accuracy in the WS mixing rule. Thus, it is no need to refit
any experimental data for the activity coefficient model parame-
ters. The second advantage of WS over HV is that the proper
composition dependence of the second virial coefficient is assured
by introducing a second virial coefficient binary interaction
parameter kij. The value of kij is usually obtained by fitting experi-
mental VLE data. Meanwhile, various approaches have also been
proposed to predict the extra interaction parameter introduced in
the WS mixing rule. For instance, Orbey and Sandler [19] proposed
a slightly reformulated version of the WS mixing rule, which can go
smoothly to the VDW mixing rule. So that, the long term accu-
mulated binary interaction parameters for VDW mixing rule can be
used in this modified mixing rule. Thereafter, Valderrama and
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44 35

Table 4
GE mixing rules for cubic EOSs.

Authors Mixing Equations Comments

rules
X 
Huron et al. HV xi ai GE The mixing rule is derived by supposing that the excess volume be zero at
[57] am ¼ bm þ HV infinite pressure. It violates the imposed by statistical thermodynamics
bi C
X quadratic composition dependency of the second virial coefficient.
bm ¼ xi bi


 
Wong and WS a a  aj The value of GE can be calculated by the existing activity coefficient model
Bij ¼ b  ¼ bi  i þ bj  1  kij Þ=2
Sandler RT ij RT RT parameters. In addition, the theoretically correct quadratic composition
[58] X  dependence of the second virial coefficient is satisfied in this mixing rule.
a GE
am ¼ bm xi i þ WS The binary interaction parameter kij in the cross second virial coefficient is
bi C
XX   usually obtained from fitting experimental VLE data.

a GE X a
bm ¼ xi xj b  1 þ WS  xi i
RT ij C RT RTbi

pffiffiffiffiffiffiffiffi
Orbey and MWS1 a 1 ai aj The equation of second virial coefficient introduces the binary interaction
Sandler Bij ¼ b  ¼ ðbi þ bj Þ  ð1  kij Þ parameter in a manner similar to that of the van der Waals mixing rule.
RT ij 2 RT
[19] X E  Thus, the interaction parameters used in van der Waals can be employed in
a G
am ¼ bm xi i þ WS this rule.
bi C
XX
 
a GE X a
bm ¼ xi xj b  1 þ WS  xi i
RT ij C RT RTbi


 
Han et al. MWS2 a a  aj kij is simplified as a function of composition-dependent and excess Gibbs
Bij ¼ b  ¼ bi  i þ bj  1  kij Þ=2
[61] RT ij RT RT energy-dependent. The VLE calculation shows that the correlated accuracy
P is as good as or better than that of the original WS mixing rule.
x B þ x2 B2 þ ða  1Þe xi ln bi GE X
kij ¼ 1 1 a¼ þ xi ai
x1 x2 ðB1 þ B2 Þ RTC WS i
X 
ai GE
am ¼ bm þ xi
bi C WS
XX
 
a GE X a
bm ¼ xi xj b  1 þ WS  xi i
RT ij C RT RTbi

pffiffiffiffiffiffiffiffi
Valderrama MWS3 a 1 ai aj According to the fitted value of kij for mixtures containing n-alkanols and
and Bij ¼ b  ¼ ðbi þ bj Þ  ð1  kij Þ carbon dioxide, the interaction parameter is correlated as a function of the
RT ij 2 RT
Zavaleta     reduced temperature TR and the polar parameter m.
0:1730 11:39 36:76 2
[60] kij ¼ 0:4680  þ  20:51 þ j þ 74:08  j
TR TR TR
X 
T m2 a GE
TR ¼ j ¼ 2:83 am ¼ bm xi i þ WS
Tc Tc vc bi C
XX
 
a GE X a
bm ¼ xi xj b  1 þ WS  xi i
RT ij C RT RTbi

"  #
Michelsen MHV1 am X a 1 GE X b An assumption of the MHV1 model is that the ratio of the zero-pressure
[63] ¼ xi i þ MHV1 þ xi ln liquid molar volume to the co-volume parameter b is the same for the
bm RT bi RT C RT i
bi
X mixture and for each of its pure components. The mixing rule is derived
bm ¼ xi bi from the linear approximation of the q function.

X X  
Michelsen MHV2 GE X b Quadratic extrapolation parameters were used to ensure continuity of the
[64] q1 ða  xi ai Þ þ q2 ða2  xi a2i Þ ¼ þ xi ln function q and its derivatives. In this mixing rule, the relation between the
i i
RT i
bi
X excess Gibbs free energy from an EOS and from an activity coefficient model
am a
a¼ a ¼ i bm ¼ xi bi takes the quadratic form.
bm RT i bi RT

Holderbaum PSRK SRK þ MHV1 þ UNIFAC The method combines the SRK cubic EOS with the group contribution model
and UNIFAC via MHV1 mixing rules. In order to predict thermodynamic
Gmehling properties over a large temperature and pressure range, the UNIFAC
[65] parameters are refitted to vaporeliquid equilibrium data around
atmospheric pressure

X 
Ahlers and VTPR xi ai GE VTPR EOS is combined with the excess Gibbs energy. However, Only the
Gmehling am ¼ bm  residual part of the excess Gibbs energy, which is calculated by the modified
bi 0:53087
[66] 0 1 UNIFAC, is used in the mixing rule for the parameter a. The exponent 3/4 in
3=4 3=4 4=3
XN X N bii þ bij the nonlinear combination rule for the cross-parameter bij, and the linear
bm ¼ xi xj bij bij ¼ @ A
2 mixing rule for constant c. This model leads to a significantly better
i¼1 j¼1
X description in the case of asymmetric systems
cm ¼ xi ci

  X
Orbey and HVOS am GE 1 X b a This model is developed at the limit of infinite pressure, and it is assumed
¼ þ x ln þ xi i
Sandler bm RT C HVOS RT C HVOS i i bi i
bi RT that there is a universal linear algebraic function that relates the liquid
[68] X molar volumes to their close packed hard-core volumes
bm ¼ xi bi
(continued on next page)
36 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44

Table 4 (continued )

Authors Mixing Equations Comments

rules

X   X
Coniglio et al. GCVM am GE b a Depending on the choice of values for C1 and C2, the generalized expression
[69] ¼ C1 þ C2 xi ln þ xi i can yield different mixing rules/methods
bm RT RT i
bi i
bi RT
X
bm ¼ xi bi

  E   X
Boukouvalas LCVM am l 1l G 1l X b a It is a hybrid model and lacks a firm theoretical basis because it combines GE
et al. [70] ¼ þ þ xi ln þ xi i of the HV model, which is evaluated at infinite pressure, and GE of MHV1,
bm RT C HV C MHV1 RT C MHV1 i bi i
bi RT
X which is evaluated at zero pressure. Experimental value l¼0.36 is obtained
bm ¼ xi bi from the fitting of experimental data.

  X
Orbey and CHV am GE 1d X b a The mixing rule has an infinite reference pressure and performs equally well
¼ þ x ln þ xi i
Sandler bm RT C CHV RT C CHV i i bi i
bi RT in comparison to the other models for VLE correlation-prediction of
[71] X mixtures of molecules with large size differences.
bm ¼ xi bi

Zavaleta [60] correlated kij as a function of the reduced temperature
and the polar parameter for the modified mixing rule, based on the
experimental data of mixtures containing n-alkanols and carbon
dioxide. Han et al. [61] proposed a completely predictable model
for kij based on the equality of GE calculated from a cubic EOS at zero
pressure and from a solution model. Only compositions and activity
coefficient parameters are required in the calculation of kij. The
three modified WS mixing rules are also presented in Table 4.
The problem with the HV model resulting from the pressure
dependence of the excess Gibbs energy led a number of in-
vestigators to propose EOS mixing rules based on the idea of
combining activity coefficient models and EOSs at low or zero
pressures, as originally suggested by Mollerup [62]. The exact
mixing rule obtained at zero pressure is
    X  
GE X bm am ai
þ xi ln ¼q  xi q
RT bi bm RT bi RT
i i
The function q requires solving for the liquid density at zero
pressure from the cubic EOS for each species in the mixture.
However, there may not be a liquid density solution for one or more
of the pure components at the temperature of the mixture and zero
pressure. Therefore, mixing rules are developed by taking some
sort of approximate extrapolation technique. Most of them are
based on a linear or a quadratic approximation of the q function.
The most well-known approximate zero reference pressure mixing
rules, namely MHV1 and MHV2 were proposed by Michelsen and
co-workers [63,64]. The activity coefficient models, such as UNIFAC
and UNIQUAC can be directly used in these mixing rules. Owing to
the fact that the linear mixing rule is used for the co-volume
parameter, the proposed GE mixing rules at low pressure do not
satisfy the theoretically justified mixing rule for the second virial
coefficient. Furthermore, Gmehling and co-workers [65] derived a
group contribution EOS for VLE of mixtures, namely predictive
Soave-RedlicheKwong (PSRK), based on the MHV1 mixing rule. In
PSRK, the modified group contribution model UNIFAC is employed
to calculate the activity coefficients. This model gives very good
results for thermodynamic properties, especially for VLE at high-
pressures. However, the large deviations between the calculations
and the experimental data occur for asymmetric systems. Thus,
some modifications of PSRK model were proposed later, such as the
VTPR. In VTPR, the logarithmic term of co-volume in the mixing
rule for the energy parameter and the combinatorial part in the
UNIFAC model are skipped simultaneously. A nonlinear mixing rule
for the co-volume parameter, instead of the linear mixing rule, is
employed [66,67]. With these two modifications, better results are
obtained for VLE of asymmetric systems. On the other hand, in
order to avoid calculations that a liquid phase solution to the EOS
may not exist for one or more of the pure components at the
temperature of interest, Orbey and Sandler [68] proposed Huron-
Vidal-Orbey-Sandler (HVOS) mixing rule based on the equality
given by Eq. (21). It's assumed that the ratio between the liquid
molar volumes and hard-core volumes is unity for pure fluids and
their mixtures at infinite pressure. Thus, this model is in agreement
with the spirit of the Van der Waals hard core concept. Meanwhile,
from the HVOS expression presented in Table 4, it can be seen that
HVOS mixing rule is algebraically similar to the commonly used
zero-pressure models, such as MHV1.
Considering that the mixing rules at infinite or zero pressure
have a weak performance on calculating the VLE of size-
asymmetric mixtures, a new type of mixing rules is suggested.
Coniglioet al [69] proposed a generalized expression for this type. It
can be written by
(23)
  X
am GE X bm a
¼ C1 þ C2 xi ln þ xi i (24)
bm RT RT i
b i i
b i RT
For the co-volume parameter, such models use the linear mixing
rule. The first successful mixing rule of this type is called LCVM,
which was proposed by Tassios and co-workers [70]. As indicated in
Table 4. LCVM combines the HV with MHV1 mixing rules linearly
and can't be derived from Eq. (18), owing to the unknown reference
pressure. Thus, it is uncertain whether the UNIFAC parameters
estimated from low pressure data can be used to calculate GE. Pa-
rameters of this mixing rule must be refitted based on the VLE data.
Although this model lacks a firm theoretical basis, the authors have
shown that the LCVM performs very well for VLE predictions
involving a wide variety of systems and is particularly suitable for
systems consisting of molecules of very different sizes. The reason
for this success was analyzed comprehensively by Georgios and
Philippos [12]. Subsequently, it was found that in the proposed
mixing rules, the size difference of the species appears twice, first
in the excess free energy term and again in the logarithmic term of
co-volume. Thus, a new parameter was introduced into the HVOS
mixing rule by Orbey and Sandler [71], in order to reduce the effect
of this double counting of the size difference. The new mixing rule
is called CHV, as listed in Table 4. Even though the CHV and LCVM
models look quite similar, there are two differences between them.
First, unlike the LCVM model, the CHV model has a well-defined
reference pressure. Second, in the CHV model, the value of
adjustable parameter only influences the logarithmic term of co-
volume as it is not a multiplier of the excess energy term. In
contrast, in the LCVM model, the value of adjustable parameter
affects both terms.
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44 37

5. Discussion Table 6
Experimental data sets of the considered mixtures.

Recently, a large amount of experimental studies have been Mixed working fluids Year N T/K Data sources
conducted to measure VLE data for mixed working fluids R290 þ R600 2005 16 270.00e310.00 [82]
[9,72e79]. In these related literatures, four kinds of mixtures, R290 þ R600a 2005 34 260.00e320.00 [82]
namely HCs þ HCs binaries, HCs þ HFCs binaries, HFCs þ HFCs R152a þ R245fa 2015 36 323.15e353.15 [74]
binaries and trifluoroiodomethane (R13I1) þ HCs/HFCs binaries R134a þ R245fa 2001 31 293.15e313.15 [83]
R125 þ R245fa 2000 21 298.18, 313.22 [84]
are mainly considered. The reason is that these mixtures are
R134a þ R290 2011 36 255.00e285.00 [85]
considered to be alternative working fluids for thermodynamic R134a þ R600a 1999 32 303.15e323.15 [86]
cycles, and VLE data for these mixtures can provide useful infor- R290 þ R245fa 2000 32 298.18,313.22 [84]
mation for calculation of thermodynamics properties. Further- R600 þ R245fa 2016 51 303.15e373.15 [72]
R600a þ R245fa 2001 40 293.15e313.15 [83]
more, these related literatures also show that various VLE models,
R13I1þR290 2014 32 258.15e283.15 [78]
which combine cubic EOSs, like PR and SRK, and mixing rules, like R13I1þR600a 2013 36 263.15e293.15 [79]
VDW and MHV1, have been used to correlate the experimental R13I1þR152a 2015 36 258.15e283.15 [75]
data of mixtures. For the cubic EOSs, PR has been widely used to
describe the phase equilibrium of working fluids. Furthermore,
when the excess free energy mixing rules are used in the VLE fugacity in VLE calculation, the fugacity expressions based on the
calculation, the activity coefficient model NRTL is always three models are also given in Table 5. In order to exhibit the
employed to calculate GE. In order to predict the VLE accurately, capacities and limitations of these predictive models in predicting
the parameters of NRTL are fitted based on the measurements of the VLE of alternative mixed working fluids, the published VLE
VLE data. However, the fitted interaction parameters are only valid data sets of 13 working fluids at different temperatures are used.
for the corresponding mixtures. Therefore, predictive models are As indicated in Table 6, these mixed working fluids can be cate-
needed to calculate the great number of alternative mixtures, gorized into HCs þ HCs binaries, HFCs þ HFCs binaries, HCs þ HFCs
considering that VLE experiments are very time-consuming and binaries and R13I1þ HCs/HFCs binaries. For some working fluids,
expensive. In fact, when the group contribution method UNIFAC is VLE data were measured in recent years, and their data haven't
used to predict the activity coefficient model, any excess free been utilized to get the parameters of the predictive models. Thus,
energy mixing rule can be predictable. Meanwhile, some predic- performances of different models for the four types of mixtures
tive VDW mixing rules have also been proposed for mixed are discussed and compared. Table 7 presents the molecular
working fluids, as stated in Table 2. In order to compare the per- structures, critical parameters and acentric factors of pure work-
formance of VLE models with different mixing rules, three pre- ing fluids involved in these mixtures. The molecular structures
dictive models, namely PR þ VDW, PR þ MWS2þUNIFAC and and thermodynamic properties of pure fluids are essential to
PR þ MHV1þUNIFAC, are applied to predict the VLE behavior of predict the VLE of mixed working fluids for the predictive models.
alternative mixed working fluids, as listed in Table 5. These The calculated results from three predictive models are pre-
models are also called VDW, MWS2 and MHV1 methods respec- sented in Figs. 2e14 for 13 mixed working fluids respectively. In
tively. The VDW is selected on account of the completely predic- Figs. 2e14, the points (,) are the experimental data of liquid
tive ability for HFC and HC mixtures. The MWS2 and MHV1 have phase, the points (B) are the experimental data of vapor phase, the
different reference pressures. The reason for selecting MWS2 and red dot lines are the results of the VDW model, cyan dash lines are
MHV1 is that the UNIFAC parameters fitted by VLE data can be the results from the MWS2 model, and the black dash-dot lines are
directly used in these mixing rules. The group interaction pa- from the MHV1 model. In general, the results from the VDW,
rameters required for the VLE prediction of mixed working fluids MWS2, and MHV1 models are all quite similar, especially for the
can be obtained from Refs. [80] and [81]. Due to the importance of MWS2 and MHV1. Due to the fact that the same values of UNIFAC

Table 5
Three predictive models for the VLE calculation of mixed working fluids.

Models Fugacity expressions Remarks

0 P 1 0 P
PR þ VDW   2 xj aij pffiffiffi 1 kij¼kikjki ¼ 0:30ui þ 0:031n0:1
F =ui bm ¼ xi bi
fi ðT; P; xi Þ bi am B j b C Z þ ð1 þ 2ÞBA
ln ¼ ðZ  1Þ  lnðZ  BÞ  pffiffiffi @  i Aln@ pffiffiffi
xi P bm 2 2bm RT am bm Z þ ð1  2ÞB

B ¼ bm P=RT

0 pffiffiffi 1
PR þ MWS2þUNIFAC     C WS ¼ 0:62323
f ðT; P; xi Þ C am C2 C Z þ ð1 þ 2ÞBA
ln i ¼ 1 ðZ  1Þ  lnðZ  BÞ  pffiffiffi  1 ln@ pffiffiffi
xi P bm 2 2bm RT am bm Z þ ð1  2ÞB
   
2 X
a Q a ln g ai ln gi
C1 ¼ xj b   1  i  WSi C2 ¼ RTDC1 þ RTbm þ
1D j RT ij ð1  DÞ2 bi RT C bi RT C WS

X ai GE XX
a
B ¼ bm P=RT D¼ xj þ Q¼ xi xj b 
j
bi RT C WS RT i j
RT ij

0 pffiffiffi 1
PR þ MHV1þUNIFAC     CMHV1¼0.53
f ðT; P; xi Þ b 1 ai A Z þ ð1 þ 2ÞBA
ln i ¼ i ðZ  1Þ  lnðZ  BÞ  pffiffiffi þ MHV1 ln@ pffiffiffi
xi P bm 2 2 bi RT C Z þ ð1  2ÞB
 
bm b
A ¼ ln gi þ ln þ i 1 B ¼ bm P=RT
bi b
38 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44

Table 7
Molecular structures, critical parameters and acentric factors of pure working fluids.

ASHRAE number Refrigerant name Molecular structure Tc/K Pc/MPa u

R290 Propane CH3CH2CH3 369.89 4.25 0.15
R600a Isobutane CH(CH3)3 407.81 3.63 0.18
R600 Butane CH3CH2CH2CH3 425.13 3.80 0.20
R125 Pentafluoroethane CHF2CF3 339.17 3.62 0.31
R134a Norflurane CF3CH2F 374.21 4.06 0.33
R152a Difluoroethane CHF2CH3 386.41 4.52 0.28
R245fa Pentafluoropropane CF3CH2CHF2 427.16 3.65 0.38
R13I1 Trifluoroiodomethane CF3I 396.44 3.95 0.18

1.0
2.4 VDW
VDW 300K
MWS2 MWS2
0.8 MHV1 MHV1
2.0 353.15K

0.6 1.6

P/MPa
P/MPa

1.2
0.4

0.8
0.2 270K 323.15K

0.4

0.0
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Molar fraction of R290 Molar fraction of R152a

Fig. 2. VLE prediction results for the R290 þ R600 system. Experimental data: liquid Fig. 4. VLE prediction results for the R152a þ R245fa system. Experimental data: liquid
phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.) colour in this figure legend, the reader is referred to the web version of this article.)

interaction parameters are used for MWS2 and MHV1, the results of have obviously azeotropic behaviors, which mean that the first
these two models overlap on the scale of the figures in some cases. derivative of the pressure with respect to the composition equals to
From these figures, it can be seen that some mixed working fluids zero at the azeotropic point for the binary system. The phase

1.0 VDW
1.6 VDW 320K
MWS2
MWS2 313.15K
MHV1
MHV1
0.8
1.2

300K 0.6
P/MPa
P/MPa

0.8
0.4
280K
293.15K
0.4 0.2

260K
0.0
0.0
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Molar fraction of R290 Molar fraction of R134a

Fig. 3. VLE prediction results for the R290 þ R600a system. Experimental data: liquid Fig. 5. VLE prediction results for the R134a þ R245fa system. Experimental data: liquid
phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.) colour in this figure legend, the reader is referred to the web version of this article.)
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44 39

1.5
2.0 VDW
MWS2
MHV1 VDW
1.6 1.2 MWS2
MHV1
323.2K
313.22K
1.2
P/MPa

P/MPa
0.9

0.8

303.2K
0.6
0.4 298.18K

0.0 0.3
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Molar fraction of R125 Molar fraction of R134a
Fig. 6. VLE prediction results for the R125 þ R245fa system. Experimental data: liquid Fig. 8. VLE prediction results for the R134a þ R600a system. Experimental data: liquid
phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.) colour in this figure legend, the reader is referred to the web version of this article.)

behavior at the azeotropic point satisfies the following thermody-

namic criterion: N pexp  pcal
1 X i i
AARD P ¼  100 (26)
N i¼1 pexp
i

N
1 X exp
  AAD y ¼ yi  ycal
i (27)
dP N i¼1
¼0 (25)
dx T
Where N is the number of experimental data, the subscripts exp
Furthermore, Table 8 represents the deviations between the and cal mean experimental and calculated values respectively.
data calculated by these three models and the experimental data. For the VLE predictions of HCs þ HCs binaries, the results for
Two statistical parameters, namely the average absolute relative R290 þ R600, R290 þ R600a at different temperatures are shown in
deviation of pressure (AARD P) and the average absolute deviation Figs. 2 and 3 respectively. The two mixtures are the most symmetric
of vapor composition (AAD y) are used for every mixture. The pa- fluids in the studied mixtures. As can be seen, two excess Gibbs
rameters are defined as follows:

1.4
VDW VDW
MWS2 313.22K
0.8 1.2 MWS2
MHV1 MHV1
1.0

0.6 285K
0.8
P/MPa
P/MPa

275K 0.6
0.4
265K 0.4
298.18K
255K 0.2
0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Molar fraction of R134a Molar fraction of R290

Fig. 7. VLE prediction results for the R134a þ R290 system. Experimental data: liquid Fig. 9. VLE prediction results for the R290 þ R245fa system. Experimental data: liquid
phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.) colour in this figure legend, the reader is referred to the web version of this article.)
40 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44

0.7
VDW VDW
2.0 373.15K 283.15K
MWS2 MWS2
MHV1 0.6 MHV1

1.6
0.5 273.15K
1.2
P/MPa

P/MPa
343.15K 0.4
263.15K
0.8
0.3

0.4
0.2
303.15K
0.0
0.1
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Molar fraction of R600 Molar fraction of R13I1
Fig. 10. VLE prediction results for the R600 þ R245fa system. Experimental data: liquid Fig. 12. VLE prediction results for the R13I1þR290 system. Experimental data: liquid
phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.) colour in this figure legend, the reader is referred to the web version of this article.)

energy models, MWS2 and MHV1 give similar results for the
HCs þ HCs binaries at different temperatures. Compared with the
excess Gibbs energy models, the VDW model gives less accurate
results, and slightly underestimates the phase equilibrium pressure
for the VLE system R290 þ R600. However, for the mixture
R290 þ R600a, VDW presents a bit better correlation between the
calculated results and the experimental data. The reason for these
phenomena is that the binary interaction parameter of the used
VDW is assumed to be the difference between the mixing factors of
the pure components. However, this assumption has no founda-
tions of theory and the correlation is developed only based on the
experimental data of HCs/HFCs mixtures. Thus, the accuracy of VLE
predictions for mixed working fluids depends on the number of
experimental data used in the fitting process of VDW parameters.
The predictions for the VLE of R152a þ R245fa, R134a þ R245fa
and R125 þ R245fa at different temperatures are shown in Figs. 4, 5
and 6 respectively. The best results are obtained by the VDW model,
which gives good agreements with the experimental data for all the
HFCs þ HFCs binaries. The MWS2, MHV1 models give good results
only for the mixtures R134a þ R245fa, R125 þ R245fa. Although the
three mixtures are homogeneous zeotropes, the results from the
excess free energy models deviate far from the experimental data of
R152a þ R245fa, and with the increase of temperature, the de-
viations increase obviously. This is caused by the fitted UNIFAC
parameters. The early experimental data of R134a þ R245fa and
R125 þ R245fa were used to fit the interaction parameters of
UNIFAC groups in Ref. [80]. Thus, it's reasonable that the accurate
VLE predictions are obtained for R134a þ R245fa and

0.5
0.6
VDW
MWS2 VDW
MHV1 313.15K MWS2 293.15K
MHV1
0.5 0.4

283.15K
0.4
P/MPa

0.3
P/MPa

0.3 273.15K
293.15K
0.2
263.15K
0.2

0.1
0.1
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Molar fraction of R600a
Molar fraction of R13I1
Fig. 11. VLE prediction results for the R600a þ R245fa system. Experimental data:
liquid phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan Fig. 13. VLE prediction results for the R13I1þR600a system. Experimental data: liquid
dash lines for MWS2; the black dash-dot lines for MHV1. (For interpretation of the phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
references to colour in this figure legend, the reader is referred to the web version of for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to
this article.) colour in this figure legend, the reader is referred to the web version of this article.)
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44 41

azeotropic point accurately, that's why the UNIFAC parameters

VDW
MWS2 obtained from ref [80] are still widely used to get the azeotropic
0.4 MHV1 point of mixed working fluids, even though large deviations exist
283.15K between the calculated results and the VLE data.
The results for the correlation of the R13I1 þ HCs/HFCs binaries
at different temperatures are given in Figs. 12e14. For the three
mixtures R13I1þR290, R13I1þR600a and R13I1þR152a, the MHV1
0.3
model gives the most accurate results. When presenting the phase
P/MPa

273.15K behaviors of R13I1þR290 and R13I1þR600a, another EOS/GE

model, MWS2 has the same capacity as MHV1. However, for the
binary R13I1þR152a, MWS2 slightly underestimates the phase
0.2 pressure. The reason for the differences between the results of
263.15K these two models is that MHV1 was used to correlate the
R13I1 þ HCs/HFCs binaries in Ref. [81], in order to obtain the
UNIFAC group interaction parameters. Thus, best results are ob-
tained with the MHV1 model. Although the same parameters are
0.1 used for the activity coefficient in the MWS2, the good agreement
0.0 0.2 0.4 0.6 0.8 1.0 with the experimental data for the mixtures can't be assured by
Molar fraction of R13I1 the MWS2 model. For the VDW model, large deviations between
the predicted results and experimental data exist in these three
Fig. 14. VLE prediction results for the R13I1þR152asystem. Experimental data: liquid
mixtures. The reason is that the VDW model was proposed only
phase (,); vapor phase (B). The model: the red dot lines for VDW; the cyan dash lines
for MWS2; the black dash-dot lines for MHV1. (For interpretation of the references to based on the experimental data of HCs/HFCs. Thus, it's not
colour in this figure legend, the reader is referred to the web version of this article.) appropriate to use VDW model in the VLE prediction of
R13I1 þ HCs/HFCs binaries.
R125 þ R245fa. However, the experimental data of R152a þ R245fa
were measured in 2015 and not used in Ref. [80]. So, for the fitted
parameters of groups, the phase behavior of the new mixtures may
6. Knowledge gaps and development directions
not be described well by the excess Gibbs energy mixing rules. For
the mixture R152a þ R245fa, MWS2, MHV1 overestimates the
Many cubic EOSs and mixing rules have been widely combined
phase pressure greatly.
to predict the phase behavior of mixed working fluids. The accuracy
The prediction results for more asymmetric binary systems are
of a model depends on the used cubic EOS, the chosen mixing rule,
presented in Figs. 7e11 for R134a þ R290, R134a þ R600a,
and the mixture studied. So far, the most accurate descriptions on
R290 þ R245fa, R600 þ R245fa and R600a þ R245fa respectively.
the VLE of mixture are always obtained by the models, whose
These mixtures belong to HCs þ HFCs binaries. For the considered
interaction parameters are fitted based on the mixture's experi-
mixed working fluids, the VDW model correlates the data well. In
mental data. However, considering the huge number of alternative
some cases, the accuracy of the MHV1 can be comparable with the
mixed working fluids, predictive models are needed urgently to
VDW model. However, for most mixtures, the MHV1 un-
calculate the thermodynamic properties of working fluids. In gen-
derestimates the pressure. As for the MWS2, it is the least accurate.
eral, two-parameter cubic EOSs are accurate enough to describe the
Perhaps, the reason is that when fitting the UNIFAC group inter-
VLE behavior of working fluids. Thus, the key for VLE prediction of
action parameters, VLE data with a specified GE mixing rule, instead
mixed working fluids lies in the mixing rules for parameters of
of the activity coefficient data, are used. Therefore, for some mixing
EOSs. For the Van der Waals mixing rules, a variety of expressions
rules, such as MWS2, MHV1, the fitted parameters may be not
have been proposed to obtain the binary interaction parameter kij.
applicable. In addition, most of the HCs þ HFCs binaries have
But, most of these correlations are only valid for some types of
obviously azeotropic behaviors, such as R134a þ R290,
mixture, and the empirical equations from the VLE data can't
R134a þ R600a. As can be seen, all the models can predict the
guarantee the VLE accuracy of mixed working fluids. Although the
cross co-volume parameter can be correlated with the hard sphere
Table 8 diameter theoretically, the interaction parameter for cross co-
VLE calculated results from the three predictive models: PR þ VDW, volume has little effect on the VLE calculation in comparison with
PR þ MWS2þUNIFAC, PR þ MHV1þUNIFAC. that of the cross energy parameter. Thus, a generalized VDW model
Mixed working fluids AARD P AAD y for mixed working fluids should be derived from the molecular
theory or group contribution methods. The group contribution
VDW MWS2 MHV1 VDW MWS2 MHV1
values can be obtained from the large VLE data fitting. As for the
R290 þ R600 3.39 1.67 1.77 0.0099 0.0092 0.0089 excess free energy mixing rules, the energy parameter is expressed
R290 þ R600a 0.72 2.52 2.29 0.0091 0.0084 0.0083
R152a þ R245fa 0.56 14.32 16.96 0.0033 0.0323 0.0365
by the excess free energy GE, the co-volume parameter is generally
R134a þ R245fa 0.74 0.77 0.71 0.0022 0.0033 0.0031 obtained using a linear mixing rule. When the group contribution
R125 þ R245fa 0.83 0.80 0.83 0.0029 0.0032 0.0028 method UNIFAC is used for the activity coefficient, any model with
R134a þ R290 0.91 6.67 2.54 0.0077 0.0167 0.0062 GE mixing rules can be predictable. However, the UNIFAC group
R134a þ R600a 0.78 2.70 1.17 0.0054 0.0139 0.0044
interaction parameters fitted from VLE data by a GE mixing rule may
R290 þ R245fa 0.93 8.83 7.25 0.0025 0.0304 0.0245
R600 þ R245fa 1.76 10.95 8.62 0.0056 0.0308 0.0242 not be suitable for other EOS/GE models. The reason for this is that
R600a þ R245fa 0.71 11.47 8.78 0.0048 0.0470 0.0348 the interaction parameters aren't derived from the activity coeffi-
R13I1þR290 2.52 0.40 0.47 0.0077 0.0022 0.0029 cient data. Therefore, a predictive model, which consists of a cubic
R13I1þR600a 4.17 0.77 0.29 0.0129 0.0029 0.0043 EOS, a GE mixing rule and corresponding UNIFAC group parameters,
R13I1þR152a 3.67 1.18 0.53 0.0434 0.0024 0.0037
should be developed in the future, given the unavailability of ac-
tivity coefficient data for most of the mixed working fluids.
42 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44

7. Conclusions VLE Vapor-liquid equilibrium

x Molar fraction
Recent advances in modeling the vapor-liquid equilibrium of y Molar fraction of vapor phase
mixed working fluids have been reviewed. Based on the method- z Molar fraction of liquid phase
ology of VLE calculation for mixed working fluids, the most Z Compressibility factor
frequently-used cubic EOSs has to combine with the mixing rules,
so that the phase behavior of mixtures can be described by the EOSs Greeks
with parameters obtained from the mixing rules. The most suc- g Activity coefficient
cessful cubic EOSs in expression of the VLE of mixtures have been
presented in this review. Thereafter, two types of mixing rules, Subscripts
namely Van der Waals and excess free energy are summarized m Mixture
respectively. For the Van der Waals mixing rules, extensive re- i, j Species i, j
searches focus on the cross energy parameter of the EOS, which has
a significant effect on the VLE prediction. Although quite a few Superscripts
semi-empirical correlations have been proposed to obtain the cross E Excess properties
co-volume parameter, a linear mixing rule is always used in the VLE L Liquid phase
calculation of mixed working fluids for simplicity. In the excess free V Vapor phase
energy mixing rules, the energy parameter is expressed by the exp Experimental values
excess free energy. When the group contribution activity coefficient cal Calculated values
method UNIFAC is used to calculate the excess free energy, the
formed models can predict the phase behavior of mixed working References
fluids without experimental data. So, experimental data of 13
mixed working fluids at different temperatures are used to evaluate [1] J.M. Calm, The next generation of refrigerants e historical review, consider-
the performances of three predictive models VDW, MWS2, MHV1 ations, and outlook, Int. J. Refrig. 31 (2008) 1123e1133.
[2] A. Desideri, S. Gusev, M. van den Broek, V. Lemort, S. Quoilin, Experimental
in this review. The results show that no predictive models can give comparison of organic fluids for low temperature ORC (organic Rankine cycle)
good agreements with the experimental data for all the mixed systems for waste heat recovery applications, Energy 97 (2016) 460e469.
working fluids. Therefore, it's thought that a generalized model [3] D. Roskosch, B. Atakan, Reverse engineering of fluid selection for thermody-
namic cycles with cubic equations of state, using a compression heat pump as
should be derived from the molecular theory or the group contri-
example, Energy 81 (2015) 202e212.
bution method, in which the group parameters are fitted from the [4] J. Wang, L. Zhao, X. Wang, An experimental study on the recuperative low
large VLE data. temperature solar Rankine cycle using R245fa, Appl. Energy 94 (2012) 34e40.
[5] P. Mavrou, A.I. Papadopoulos, P. Seferlis, P. Linke, S. Voutetakis, Selection of
working fluid mixtures for flexible Organic Rankine Cycles under operating
Acknowledgements variability through a systematic nonlinear sensitivity analysis approach, Appl.
Therm. Eng. 89 (2015) 1054e1067.
This work is sponsored by the National Nature Science Foun- [6] J. Wang, L. Zhao, X. Wang, A comparative study of pure and zeotropic mixtures
in low-temperature solar Rankine cycle, Appl. Energy 87 (2010) 3366e3373.
dation of China (51476110). [7] X. Wang, L. Zhao, Analysis of zeotropic mixtures used in low-temperature
solar Rankine cycles for power generation, Sol Energy 83 (2009) 605e613.
Nomenclature [8] R. Akasaka, Thermodynamic property models for the difluoromethane (R-
32) þ trans-1,3,3,3-tetrafluoropropene (R-1234ze(E)) and difluoromethane þ
2,3,3,3-tetrafluoropropene (R-1234yf) mixtures, Fluid Phase Equilib. 358
(2013) 98e104.
Symbols [9] Z. Yang, M. Gong, H. Guo, X. Dong, J. Wu, Phase equilibrium for the binary
mixture of {1,1-difluoroethane (R152a)þtrans-1,3,3,3-tetrafluoropropene
a Energy parameter
(R1234ze (E))} at various temperatures from 258.150 to 288.150K, Fluid
A Excess Helmholtz free energy Phase Equilib. 355 (2013) 99e103.
AARD P Average absolute relative deviation of pressure [10] A.M. Abudour, S.A. Mohammad, R.L. Robinson Jr., K.A.M. Gasem, Generalized
binary interaction parameters for the PengeRobinson equation of state, Fluid
AAD y Average absolute deviation of vapor composition
Phase Equilib. 383 (2014) 156e173.
b Co-volume parameter [11] R. Privat, J.-N. Jaubert, PPR78, a thermodynamic model for the prediction of
B Second virial coefficients petroleum fluid-phase behaviour, in: XXXVII JEEPe37th Conference on Phase
CFC Chlorofluorocarbon Equilibria, EDP Sciences, 2011.
[12] Philippos Coutsikos, Thirty years with EoS/GModels - what have We Learned?
DDB Dortmund Data Bank Ind. Eng. Chem. Res. 51 (2012) 4119e4142.
EOS Equation of state [13] S. Mahato, A. Kumar, T. Banerjee, PC-SAFT predictions on mixtures of 1-alkyl-
f Fugacity 3methylimidazolium-bis(trifluroromethanesulphonyl)amide with hydrocar-
bons, alcohols and aqueous systems using a correlative based binary inter-
G Free energy action parameter, J. Taiwan Inst. Chem. E 59 (2016) 69e78.
GWP Global warming potential [14] D. Rowland, T.J. Hughes, E.F. May, Extending the GERG-2008 equation of state:
HC Hydrocarbon improved departure function and interaction parameters for (methane þ
butane), J. Chem. Thermodyn. 97 (2016) 206e213.
HFC Hydrofluorocarbon [15] B.E. Poling, J.M. Prausnitz, J.P. O'connell, The Properties of Gases and Liquids,
HCFC Hydrochlorofluorocarbon 5th. ed, vol. 5, 2004.
k Binary interaction parameter for energy parameter [16] J. Escandell, E. Neau, C. Nicolas, A new formulation of the predictive NRTL-PR
model in terms of kij mixing rules. Extension of the group contributions for
l Binary interaction parameter for co-volume parameter the modeling of hydrocarbons in the presence of associating compounds,
N Number of experimental data Fluid Phase Equilib. 301 (2011) 80e97.
ODP Ozone depletion potential [17] S.-E.K. Fateen, M.M. Khalil, A.O. Elnabawy, Semi-empirical correlation for bi-
nary interaction parameters of the PengeRobinson equation of state with the
P Pressure
van der Waals mixing rules for the prediction of high-pressure vaporeliquid
PR PengeRobinson equilibrium, JARE 4 (2013) 137e145.
SRK SoaveeRedlicheKwong [18] M. Cismondi Duarte, M.V. Galdo, M.J. Gomez, N.G. Tassin, M. Yanes, High
T Temperature pressure phase behavior modeling of asymmetric alkane þ alkane binary
systems with the RKPR EOS, Fluid Phase Equilib. 362 (2014) 125e135.
v Molar volume [19] H. Orbey, S.I. Sandler, Modeling Vapor-liquid Equilibria: Cubic Equations of
VDW Van der Waals State and Their Mixing Rules, vol. 1, Cambridge University Press, 1998.
 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44
[20] J.-W. Qian, J.-N. Jaubert, R. Privat, Phase equilibria in hydrogen-containing
binary systems modeled with the PengeRobinson equation of state and
temperature-dependent binary interaction parameters calculated through a
group-contribution method, J. Supercrit. Fluid 75 (2013) 58e71.
[21] M. Cismondi Duarte, J. Cruz Doblas, M.J. Gomez, G.F. Montoya, Modelling the
phase behavior of alkane mixtures in wide ranges of conditions: new
parameterization and predictive correlations of binary interactions for the
RKPR EOS, Fluid Phase Equilib. 403 (2015) 49e59.
[22] X. Xu, J.-N. Jaubert, R. Privat, P. Duchet-Suchaux, F. Bran ~ a-Mulero, Predicting
binary-interaction parameters of cubic equations of state for petroleum fluids
containing pseudo-components, Ind. Eng. Chem. Res. 54 (2015) 2816e2824.
[23] H.R. Radfarnia, G. Bogdani c, V. Taghikhnai, C. Ghotbi, The UNIQUAC-NRF
segmental interaction model for vapor-liquid equilibrium calculations for
polymer solutions, Polim. (Zagreb) 26 (2005) 115e120.
[24] K. Tochigi, K. Kurihara, K. Kojima, Prediction of VLE using asog and zero
pressure GE mixing rule consistent the second virial coefficient condition,
J. Chem. Eng. Jpn. 28 (1995) 617e620.
[25] S. Skjold-Jørgensen, Group contribution equation of state (GC-EOS): a pre-
dictive method for phase equilibrium computations over wide ranges of
temperature and pressures up to 30 MPa, Ind. Eng. Chem. Res. 27 (1988)
110e118.
[26] J.O. Valderrama, The state of the cubic equations of state, Ind. Eng. Chem. Res.
42 (2003) 1603e1618.
[27] D.-Y. Peng, D.B. Robinson, A new two-constant equation of state, Ind. Eng.
Chem. Res. 15 (1976) 59e64.
[28] G.M. Kontogeorgis, G.K. Folas, Thermodynamic Models for Industrial Appli-
cations: from Classical and Advanced Mixing Rules to Association Theories,
Wiley, 2010.
[29] Han Xiaohong, Study of Mixing Rules for Equation of State and Vapor-liquid
Equilibrium of Mixture Refrigerants, Zhejiang University, 2005.
[30] Luo Mingjian, The Modification of SRK Equation of State and Phase Equilib-
rium Calculation, Tianjin University, 2005.
[31] R. De Santis, F. Gironi, L. Marrelli, Vapor-liquid equilibrium from a hard-
sphere equation of state, Ind. Eng. Chem. Fundam. 15 (1976) 183e189.
[32] N.C. Patel, A.S. Teja, A new cubic equation of state for fluids and fluid mixtures,
Chem. Eng. Sci. 37 (1982) 463e473.
[33] M. Cismondi, J. Mollerup, Development and application of a three-parameter
RKePR equation of state, Fluid Phase Equilib. 232 (2005) 74e89.
[34] B. Schmid, A. Schedemann, J. Gmehling, Extension of the VTPR group contri-
bution equation of state: group interaction parameters for additional 192
group combinations and typical results, Ind. Eng. Chem. Res. 53 (2014)
3393e3405.
[35] Chen Jianxin, Prediction of Thermophysical Properties and PVTx Measure-
ment of Refrigerant Mixtures, University of Science and Technology of china,
2007.
[36] J.A.P. Coutinho, G.M. Kontogeorgis, E.H. Stenby, Binary interaction parameters
for nonpolar systems with cubic equations of state: a theoretical approach 1.
CO2/hydrocarbons using SRK equation of state, Fluid Phase Equilib. 102
(1994) 31e60.
[37] W. Weber, S. Petkov, G. Brunner, Vapoureliquid-equilibria and calculations
using the RedlicheKwong-Aspen-equation of state for tristearin, tripalmitin,
and triolein in CO2 and propane, Fluid Phase Equilib. 158e160 (1999)
695e706.
[38] R. Stryjek, J.H. Vera, PRSV: an improved Peng-Robinson equation of state with
new mixing rules for strongly nonideal mixtures, Can. J. Chem. Eng. 64 (1989)
334e340.
[39] G. Gao, J.L. Daridon, H. Saint-Guirons, P. Xans, F. Montel, A simple correlation
to evaluate binary interaction parameters of the Peng-Robinson equation of
state: binary light hydrocarbon systems, Fluid Phase Equilib. 74 (1992) 85e93.
[40] J.-X. Chen, P. Hu, Z.-S. Chen, Study on the interaction coefficients in PR
equation with vdW mixing rules for HFC and HC binary mixtures, Int. J.
Thermophy 29 (2008) 1945e1953.
[41] Jianxin Chen, Zeshao Chen, Peng Hu, The Calculation of Interaction Parameters
for the Binary HFC Mixtures, Jounal of University of Science and Technology of
China, 2006, pp. 656e659.
[42] P. Hu, L.-X. Chen, Z.-S. Chen, A modified differential-model for interaction
parameters in PR EoS with vdW mixing rules for mixtures containing HFCs
and HCs, Fluid Phase Equilib. 324 (2012) 64e69.
[43] H. Zhang, M. Gong, H. Li, Y. Zhao, Q. Zhong, X. Dong, J. Shen, J. Wu, A simple
model for temperature-independent kij of the PR-vdW model for mixtures
containing HCs, HFCs, PFCs, HFOs, CO 2, RE170 and R13I1, Fluid Phase Equilib.
425 (2016) 374e384.
[44] Duan Yuanyuan, Lin Hong, Zhang Qinling, Zhang QiuFang, Interaction co-
efficients for HFC mixtures, J. Eng. Therm. (2004) 197e200.
[45] M.J. Lee, H.C. Sun, Thermodynamic property predictions for refrigerant mix-
tures, Ind. Eng. Chem. Res. 31 (1992) 1212e1216.
[46] V. Plee, J.N. Jaubert, R. Privat, P. Arpentinier, Extension of the E-PPR78 equa-
tion of state to predict Fluid Phase Equilib of natural gases containing carbon
monoxide, helium-4 and argon, J. Petrol Sci. Eng. 133 (2015) 744e770.
[47] J.W. Qian, R. Privat, J.N. Jaubert, P. Duchet-Suchaux, Enthalpy and heat ca-
pacity changes on mixing: fundamental aspects and prediction by means of
the PPR78 cubic equation of state, Energ Fuel 27 (2013) 7150e7178.
[48] J.-N. Jaubert, F. Mutelet, VLE predictions with the PengeRobinson equation of
state and temperature dependent kij calculated through a group contribution
method, Fluid Phase Equilib. 224 (2004) 285e304.
43
[49] M. Cismondi Duarte, J. Cruz Doblas, M.J. Gomez, G.F. Montoya, Modelling the
phase behavior of alkane mixtures in wide ranges of conditions: new
parameterization and predictive correlations of binary interactions for the
RKPR EOS, Fluid Phase Equilib. 403 (2015) 49e59.
[50] M.J. Hiza, A.G. Duncan, A correlation for the prediction of interaction energy
parameters for mixtures of small molecules, Aiche J. 16 (1970) 733e738.
[51] R.A. Mcfarlane, M.R. Gray, J.M. Shaw, Evaluation of co-volume mixing rules for
bitumen liquid density and bubble pressure estimation, Fluid Phase Equilib.
293 (2010) 87e100.
[52] K.H. Lee, S.I. Sandler, The generalized van der waals partition functiondIV.
Local composition models for mixtures of unequal-size molecules, Fluid Phase
Equilib. 34 (1987) 113e147.
[53] W. Wang, C. Zhong, Mixing rules for hydrogen-containing systems, Fluid
Phase Equilib. 47 (1989) 103e114.
[54] Han Xiaohong, Wenjie Guan, Zanjun Gao, Xiaolong Cui, Xiangguo Xu,
Guangming Chen, Modification of co-volume parameter b in mixing rule for
calculating vapor liquid equilibrium, J. Chem. Ind. Eng. (2015) 489e496.
[55] Lan Xue, Li Xia, Shuguang Xiang, Research progress of GE-EOS models, Chem.
Ind. Eng. Prog. (2014) 304e308.
[56] H. Huang, S.I. Sandler, Prediction of vapor-liquid equilibria at high pressures
using activity coefficient parameters obtained from low-pressure data: a
comparison of two equation of state mixing rules, Ind. Eng. Chem. Res. 32
(1993) 7.
[57] M.J. Huron, J. Vidal, M.J. Huron, J. Vidal, New mixing rule in simple equations
of state for representing vapour-liquid equilibria of strongly non-ideal mix-
tures, Fluid Phase Equilib. 3 (1979) 255e271.
[58] D.S.H. Wong, S.I. Sandler, A theoretically correct mixing rule for cubic equa-
tions of state, Aiche J. 38 (1992) 671e680.
[59] Gmehling, The Dortmund Data Bank. http://www.ddbst.de/ [Accessed: 15
June 2016].
[60] J.O. Valderrama, J. Zavaleta, Generalized binary interaction parameters in the
WongeSandler mixing rules for mixtures containing n-alkanols and carbon
dioxide, Fluid Phase Equilib. 234 (2005) 136e143.
[61] Han Xiaohong, Guangming Chen, Qin Wang, Xiaolong Cui, vapor-liquid
equilibrium calculation with simplified WS mixing rule, J. Eng. Therm. 26
(2005) 721e724.
[62] E.W. Askew, D.E. Roberts, S.M. Wood, A note on the derivation of mixing rules
from excess Gibbs energy models, Fluid Phase Equilib. 25 (1986) 323e327.
[63] M.L. Michelsen, A method for incorporating excess Gibbs energy models in
equations of state, Fluid Phase Equilib. 60 (1990) 47e58.
[64] M.L. Michelsen, A modified huron-vidal mixing rule for cubic equation of
state, Fluid Phase Equilib. 60 (1990) 213e219.
[65] T. Holderbaum, J. Gmehling, PSRK: a group contribution equation of state
based on UNIFAC, Fluid Phase Equilib. 70 (1991) 251e265.
[66] J. Ahlers, J. Gmehling, Development of an universal group contribution
equation of state: I. Prediction of liquid densities for pure compounds with a
volume translated PengeRobinson equation of state, Fluid Phase Equilib. 191
(2001) 177e188.
[67] C. Jian, F. Kai, J. Gmehling, Modification of PSRK mixing rules and results for
vaporeliquid equilibria, enthalpy of mixing and activity coefficients at infinite
dilution, Fluid Phase Equilib. 200 (2002) 411e429.
[68] H. Orbey, S.I. Sandler, On the combination of equation of state and excess free
energy models, Fluid Phase Equilib. 111 (1995) 53e70.
[69] L. Coniglio, K. Knudsen, R. Gani, Prediction of supercritical fluid-liquid equi-
libria for carbon dioxide and fish oil related compounds through the equation
of state d excess function (EOS-g E) approach, Fluid Phase Equilib. 116 (1996)
510e517.
[70] C. Boukouvalas, N. Spiliotis, P. Coutsikos, N. Tzouvaras, D. Tassios, Prediction of
vapor-liquid equilibrium with the LCVM model: a linear combination of the
vidal and michelsen mixing rules coupled with the original UNIFAC and the t-
mPR equation of state, Fluid Phase Equilib. 92 (1994) 75e106.
[71] H. Orbey, S.I. Sandler, A comparison of Huron-Vidal type mixing rules of
mixtures of compounds with large size differences, and a new mixing rule,
Fluid Phase Equilib. 132 (1997) 1e14.
[72] L. Yang, M. Gong, H. Guo, X. Dong, J. Shen, J. Wu, Isothermal (vapour þ liquid)
equilibrium measurements and correlation for the {n-butane (R600) þ
1,1,1,3,3-pentafluoropropane (R245fa)} system at temperatures from
(303.150 to 373.150) K, J. Chem. Thermodyn. 95 (2016) 49e53.
[73] M. Gong, Y. Zhao, X. Dong, H. Guo, J. Shen, J. Wu, Measurements of isothermal
(vapor þ liquid) equilibrium for the (propane þ cis-1,3,3,3-
tetrafluoropropene) system at temperatures from (253.150 to 293.150) K,
J. Chem. Thermodyn. 98 (2016) 319e323.
[74] L. Yang, M. Gong, H. Guo, X. Dong, J. Wu, (Vapour þ liquid) equilibrium data
for the {1,1-difluoroethane (R152a) þ 1,1,1,3,3-pentafluoropropane (R245fa)}
system at temperatures from (323.150 to 353.150) K, J. Chem. Thermodyn. 91
(2015) 414e419.
[75] M. Gong, K. Cheng, X. Dong, H. Guo, Y. Zhao, J. Wu, Measurements of
isothermal (vapor þ liquid) phase equilibrium for {trifluoroiodomethane
(R13I1) þ 1,1-difluoroethane (R152a)} from T ¼ (258.150 to 283.150) K,
J. Chem. Thermodyn. 88 (2015) 90e95.
[76] X. Dong, Y. Zhao, M. Gong, G. Hao, J. Wu, (Vapour þ liquid þ liquid) equi-
librium measurements and correlation for the {1,1,1,2-tetrafluoroethane
(R134a) þ n-butane (R600)} system, J. Chem. Thermodyn. 84 (2015) 87e92.
[77] Y. Zhao, M. Gong, X. Dong, H. Guo, J. Wu, (Vapor þ liquid þ liquid) equilibrium
measurements and correlation for {1,1,2,2-tetrafluoroethane (R134) þ
44 W. Su et al. / Fluid Phase Equilibria 432 (2017) 28e44
isobutane (R600a)} system, J. Chem. Thermodyn. 78 (2014) 182e188.
[78] M. Gong, H. Guo, X. Dong, H. Li, J. Wu, (Vapor þ liquid) phase equilibrium
measurements for {trifluoroiodomethane (R13I1) þ propane (R290)} from T ¼
(258.150 to 283.150) K, J. Chem. Thermodyn. 79 (2014) 167e170.
[79] H. Guo, M. Gong, X. Dong, J. Wu, Measurements of (vapour þ liquid) equi-
librium data for {trifluoroiodomethane (R13I1) þ isobutane (R600a)} at
temperatures between (263.150 and 293.150) K, J. Chem. Thermodyn. 58
(2013) 428e431.
[80] Shuxin Hou, Duan Yuanyuan, X.D. Wang, VaporLiquid equilibria predictions
for new refrigerant mixtures based on group contribution theory, Ind. Eng.
Chem. Res. 46 (2007) 9274e9284.
[81] Zhao Yanxing, Jie Guo, Xueqiang Dong, Maoqiong Gong, Jianfeng Wu, Vapor-
liquid Equilibrium Description for (R13I1 þ HC/HFC) Systems Based on Group
Contribution Method, 2014.
[82] Y. Kayukawa, K. Fujii, Y. Higashi, Vapor-liquid equilibrium (VLE) properties for
the binary systems propane (1)þ n-butane (2) and propane (1)þ isobutane
(3), J. Chem. Eng. Data 50 (2005) 579e582.
[83] S. Bobbo, L. Fedele, M. Scattolini, R. Camporese, Isothermal VLE measurements
for the binary mixtures HFC-134a þ HFC-245fa and HC-600a þ HFC-245fa,
Fluid Phase Equilib. 185 (2001) 255e264.
[84] S. Bobbo, R. Camporese, G. Scalabrin, Isothermal vapour - liquid equilibrium
measurements for the binary mixtures HFC125 þ HFC245fa and HC290 þ
HFC245fa, High. Temp-High Press 32 (2000) 441e447.
[85] X. Dong, M. Gong, J. Liu, J. Wu, Experimental measurement of vapor pressures
and (vapor þ liquid) equilibrium for {1,1,1,2-tetrafluoroethane (R134a) þ
propane (R290)} by a recirculation apparatus with view windows, J. Chem.
Thermodyn. 43 (2011) 505e510.
[86] J.S. Lim, J.-Y. Park, B.-G. Lee, Y.-W. Lee, J.-D. Kim, Phase equilibria of CFC
alternative refrigerant mixtures: binary systems of isobutaneþ 1, 1, 1, 2-
tetrafluoroethane,þ 1, 1-difluoroethane, andþ difluoromethane, J. Chem.
Eng. Data 44 (1999) 1226e1230.