Fluid Phase Equilibria 295 (2010) 26–37

Contents lists available at ScienceDirect

Fluid Phase Equilibria

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

Relationship between the binary interaction parameters (kij ) of the

Peng–Robinson and those of the Soave–Redlich–Kwong equations of state:
Application to the definition of the PR2SRK model
Jean-Noël Jaubert ∗ , Romain Privat
Nancy-Université, École Nationale Supérieure des Industries Chimiques, Laboratoire Réactions et Génie des Procédés, 1 rue Grandville, BP 20451, Nancy cedex 9, France

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

Article history: The purpose of this paper is to establish a relationship between the binary interaction parameters of the
Received 14 February 2010 Peng–Robinson (PR) and those of the Soave–Redlich–Kwong (SRK) equations of state (EoS). This objective
Received in revised form 20 March 2010 could be reached thanks to the rigorous equivalence between the classical mixing rules with temperature-
Accepted 22 March 2010
dependent kij and the combination at constant packing fraction of a Van Laar-type excess Gibbs energy
Available online 30 March 2010
model with a cubic EoS. This equivalence makes it possible to find out a relationship between the Eij (T)
parameters issued from the Van Laar function and the kij (T) of the classical mixing rules. Our key idea
Keywords: E,∞
was to make the hypothesis that the infinite pressure residual molar excess Gibbs energy (gres ) was
Cubic equation of state
Binary interaction parameters
independent of the used EoS. Doing so, a simple relationship between the Eij suitable for the PR-EoS (EijPR )
Predictive model and those suitable for the SRK EoS (EijSRK ) can be obtained. Using this relationship and the one linking the
Peng–Robinson kij and the Eij , it was possible to find out a simple and general equation connecting the kij of a given EoS to
Soave–Redlich–Kwong the kij of any other EoS. This approach was then used to deduce kijSRK from a known kijPR . In a second step,
Van Laar Gibbs excess energy model
Constant packing fraction mixing rules using the previously mentioned mathematical equation relating kijPR to kijSRK , the PPR78 model which is a
group contribution method for the estimation of the temperature-dependent BIPs of the PR-EoS was used
to generate kij (T) for the SRK EoS. It is shown how the group interaction parameters initially determined
for the PR-EoS can be simply used to predict the temperature-dependent BIPs of the SRK EoS. This new
predictive model has been called PR2SRK. As discussed in this paper, the accuracy of this model is similar
to the accuracy of the PPR78 approach.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction parameters of the pure components i. Parameter a is taken to be

Chemical engineers from petroleum companies widely use
cubic equations of state (EoS) and especially the ones proposed by
Peng and Robinson (Peng–Robinson EoS noted PR-EoS afterwards)
[1] and by Soave (Soave–Redlich–Kwong EoS, noted SRK EoS after-
wards) [2]. These equations of state directly stem from the Van der
Waals theory and can be written for a given mixture under the
general form:
RT a parameter.
P= −
v−b Q (v)
where P is the pressure, T, the absolute temperature and v, the molar
volume of the fluid. Q (v) is a second order polynomial in v. Classical
Van der Waals mixing rules are frequently used to relate the a and
b parameters of a mixture to the composition and to the ai and bi
∗ Corresponding author. Fax: +33 3 83 17 51 52.
E-mail address: jean-noel.jaubert@ensic.inpl-nancy.fr (J.-N. Jaubert).
a quadratic function of the mole fractions and parameter b, a lin-
ear function. A binary interaction parameter (BIP), classically noted
kij , is usually involved in the a-parameter expression to provide
more flexibility to the EoS. This kij parameter can be either chosen
constant or temperature-dependent (a mole fraction dependence
is more scarcely introduced in the kij expression). It is however
important to note that numerical values of BIPs are specific (i) to
the considered EoS, (ii) to the alpha-function (Soave, Twu, Mathias-
Copeman, etc.) involved in the mathematical expression of the ai
The key point when using such cubic EoS to describe complex
mixtures like petroleum fluids is thus to give appropriate values to
the binary interaction parameters. We however know by experi-
ence that the BIPs, suitable for the PR-EoS (kijPR ) cannot be used for
the SRK EoS (kijSRK =
/ kijPR ). This assessment makes it impossible for
petroleum engineers to mix the use of these two equations of state.
Indeed, they usually have tables containing the numerical values
of the BIPs only for the most widely used EoS in their company. To
overcome this limitation, a relationship between the SRK BIP (kijSRK )

0378-3812/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.fluid.2010.03.037
 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37 27

and the PR BIP (kijPR ) is established in this paper. This relationship in Eq. (1) and we thus obtain:
is then extended to relate the BIP of a given cubic EoS to the BIP of E,∞
gres = GE (3)
any other cubic EoS.
In addition, using the previously mentioned mathematical equa- In this section, a Van Laar-type expression i.e. a solely energetic
E,∞
tion relating kijPR and kijSRK , the PPR78 model [3–16] which may model is chosen for gres /CEoS . Using Eq. (3), we thus can write:
be seen as a group contribution method for the estimation of the 
N 
N
temperature-dependent BIPs of the widely spread PR-EoS, was used xi · xj · bi · bj · Eij (T )
to generate kij for the SRK EoS. It is shown how the group interaction E,∞
gres GE 1 i=1 j=1
parameters initially determined for the PR-EoS [3–9] can be sim- = Van Laar = (4)
CEoS CEoS 2 
N
ply used to predict the temperature-dependent BIPs for the SRK
bj xj
EoS. This new predictive model has been called PR2SRK (here PR
j=1
means predictive but also means Peng–Robinson since all the used
parameters come from the PPR78 model). It now becomes possible to show that such mixing rules (com-
bination at constant packing fraction of a Van Laar-type GE model
2. Relationship between the binary interaction parameters and a cubic EoS) are rigorously equivalent to the use of classical
of the Peng–Robinson and those of the SRK equations of mixing rules with temperature-dependent kij , that is:
state ⎧
⎪ 
N 
N
 
⎪
⎪ a(T, = xi xj ai aj 1 − kij (T )
2.1. Mathematical derivation ⎪
⎨
x)
i=1 j=1
(5)
Following the zeroth-order approximation of Guggenheim’s ⎪
⎪ 
N
⎪
⎪ b(x) = xi bi
quasi-reticular model [17], Péneloux et al. [18] proposed to com- ⎩
bine at constant packing fraction a cubic EoS and an expression for i=1

the molar excess Gibbs energy model (i.e. an activity coefficient Indeed, by sending Eq. (5) in Eq. (1), one has:
model). The resulting mixing rules [11,18] are:
E,∞ 
N

⎧ gres
= xi
ai (T )
−
a(T, x)
⎪ a(T, x)  ai (T )
N CEoS bi b(x)
⎪
⎪ g E,∞
⎪
⎪ = xi − res i=1
 
⎨ b(x) bi CEoS N N
 
i=1 xi · xj ai · aj · 1 − kij (T )
(1)
⎪
⎪ 
N  N
ai (T )
⎪
⎪ = xi · −
i=1 j=1

⎪
⎩ b(x) = xi bi bi 
N
i=1
i=1 xj · bj (6)
j=1
E,∞ E (T, P, x)
gres (T, x) = lim gres
P→+∞
is the residual part of the molar 
N
ai (T )

N

N N
 
xi · × xj · bj − xi · xj ai · aj · 1 − kij (T )
excess Gibbs energy calculated from the EoS when the pressure bi
i=1 j=1 i=1 j=1
goes to infinity. a and b are EoS parameters. xi represents the mole =
fraction of component i and N the number of components in the 
N

xj · bj
mixture. CEoS , defined in the next section (see Eq. (29)), is a char-
acteristic constant of the used EoS. For the
√ PR and the√SRK EoS, the
j=1

following values are obtained: CPR = ( 2/2) ln(1 + 2) ≈ 0.6232 We can write:
and CSRK = ln(2) ≈ 0.6931. In Eq. (1), in order to allow calculation of
the a(T,x) parameter, a mathematical expression must be assigned 
N
ai (T ) 
N

E,∞ xi · × xj · bj
to gres (T, x). It may be chosen among the classical forms of molar bi
excess Gibbs energy functions (Redlich-Kister, Margules, Wilson, i=1 j=1

Van Laar, NRTL, UNIQUAC, UNIFAC, . . .) which do not depend on


N N
ai · bj 1 
N
ai · bj 1 
N
aj · bi
N N
pressure. In order to avoid confusion, we will write with a special = xi · xj = xi · xj + xi · xj
font (GE ) the selected activity coefficient model. We thus can write: bi 2 bi 2 bj
i=1 j=1 i=1 j=1 i=1 j=1
E,∞ (7)
gres = GEres (2)
Thus,



N

N
1 ai ·bj 1 aj ·bi
   
N

N
1 ai 1 aj
√ √
ai aj 
xi · xj 2 bi
+ 2 bj
− ai · aj · 1 − kij (T ) xi · xj · bi · bj 2 b2
+ 2 b2
− bi bj
· 1 − kij (T )
E,∞ i j
gres i=1 j=1 i=1 j=1
= = (8)
CEoS 
N

N

xj · bj xj · bj
j=1 j=1

At this step, we introduce for clarity:

When the GE model per se contains a combinatorial (or ather- √
ai
mal) part, essentially due to differences in size and shape of the ıi = (9)
bi
molecules in the mixture, and a residual part, essentially due to
energy interactions (e.g. the UNIQUAC and UNIFAC models), only which has the Scatchard–Hildebrand solubility parameter fea-
the residual term (GEres ) of the GE model must be introduced in Eq. ture, and we define the parameter Eij by:
(1). For other molar excess Gibbs energy models (Redlich–Kister, 
Eij = ı2i + ı2j − 2ıi ıj 1 − kij (T ) (10)
Margules, Wilson, Van Laar, NRTL), the complete GE function is used
28 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37

Eq. (8) thus writes: For a binary system (N = 2), remembering that E11 = E22 = 0 and
E21 = E12 , Eq. (18) writes:

N 
N

xi · xj · bi · bj · Eij (T ) 2x1 x2 bSRK
1 b2
SRK PR
 · E12 SRK
(T ) − E12 (T ) = 0 (19)
E,∞
gres 1 i=1 j=1
That is:
= · (11)
CEoS 2 
N
SRK PR
xj · bj E12 (T ) =  · E12 (T ) (20)
j=1 E,∞
For a binary system, equating the gres of the PR-EoS with that
Eq. (11) is strictly identical to Eq. (4). We thus demonstrated that of the SRK EoS thus leads to a unique linear relationship between
PR (T ) and E SRK (T ).
E12
it is rigorously equivalent to combine at constant packing fraction a 12
Van Laar-type GE model with a cubic EoS and to use classical mixing For a multi-component system (N > 2), it would be possible to
rules with temperature-dependent kij . From Eq. (10) we have: find many different mathematical relationships between EijPR (T )
2 and EijSRK (T ) so that Eq. (18) remains true. Among all of them, the
Eij (T ) − (ıi − ıj )
kij (T ) = (12) one derived for a binary system, that is: EijSRK (T ) =  · EijPR (T ) can
2ıi ıj
still apply when N > 2 (it verifies Eq. (18) whatever the number of
Eq. (12) thus establishes a connection between Eij of the Van Laar- components in the mixture). We however here work with binary
type GE model and kij of the classical mixing rules. interaction parameters: it means that a multi-component system
In order to ascertain a relationship between the BIPs of the must be considered as a system composed of different binary sys-
PR-EoS and those of the SRK EoS, our key idea was to make the tems. It is thus essential that the relationship between EijPR (T ) and
E,∞
hypothesis that gres was independent of the used EoS. A similar
EijSRK (T ) which was established for binary systems, still persists for
idea was originally used by Fransson et al. [19]. From Eq. (4), we
ternary systems, quaternary systems and so on (the binary relation-
thus obtain:
ship has to remain the same for any value of N). This is the reason

N 
N
why, the solution derived for a binary system which is acceptable
xi xj bPR
i
bPR
j
EijPR (T )
whatever the number of components is the only one to be con-
E,∞ 1 i=1 j=1 sidered afterwards. The universal relationship between EijPR (T ) and
gres = CPR
2 
N
EijSRK (T ) finally writes:
bPR
j
xj
j=1 EijSRK (T ) =  · EijPR (T ) (21)


N N
From Eq. (10), we can write:
xi xj bSRK bSRK EijSRK (T )
i j  2
1 i=1 j=1 Eij = 2kij ıi ıj + ıi − ıj (22)
= CSRK (13)
2  N
By combining Eqs. (21) and (22), we obtain:
bSRK xj
j  2  2
j=1 2kijSRK ıSRK
i
ıSRK
j
+ ıSRK
i
− ıSRK
j
=2kijPR ıPR
i
ıPR
j
+ ıPR
i
− ıPR
j
(23)
With:
⎧ The relation between kijSRK and kijPR is thus:
⎪ RTc,i X
⎪
⎪ bPR = ˝bPR ; ˝bPR = ≈ 0.07780 if
⎪
⎪
i Pc,i X +3
⎪
⎨  √  √
3 3
−1 + 6 2+8− 6 2−8 (14)
X=
⎪
⎪ 3
⎪
⎪ √
⎪
⎪ RTc,i 3
2−1 (24)
⎩ bi = ˝b
SRK SRK ; ˝bSRK = ≈ 0.08664
Pc,i 3 or conversely,
 2  2
Replacing in Eq. (13) bPR by:
i 2kijSRK ıSRK
i
ıSRK
j
+ ıSRK
i
− ıSRK
j
−  ıPR
i
− ıPR
j
˝bPR kijPR (T ) = (25)
bPR = bSRK (15) 2ıPR
i
ıPR
j
i SRK i
˝b
At this stage, two remarks can be made:
one obtains:

N

N
(1) From Eqs. (17) and (21), we can write:
CPR · ˝bPR · xi xj bSRK
i
bSRK
j
EijPR (T )
i=1 j=1 CSRK · ˝bSRK · EijSRK (T ) = CPR · ˝bPR · EijPR (T ) (26)


N

N Introducing Eq. (22), we thus can say that by assuming that
E,∞
= CSRK · ˝bSRK · xi xj bSRK
i
bSRK
j
EijSRK (T ) (16) gres is independent of the used EoS, we state that the quantity:
i=1 j=1
 2 
C · ˝b · Eij = C · ˝b 2kij ıi ıj + ıi − ıj (27)
CPR · ˝bPR
Introducing : = ≈ 0.807341 (17)
CSRK · ˝bSRK is independent of the used EoS.
Eq. (16) writes: (2) Using Eqs. (24) or (25), temperature-dependent kij will always
be obtained. As an example, let us assume that a constant kij

N

N
 (independent of the temperature) is known for a given binary
xi xj bSRK
i
bSRK
j
 · EijPR (T ) − EijSRK (T ) = 0 (18)
system and for the PR-EoS. By using Eq. (24), temperature-
i=1 j=1 dependent kij , suitable for the SRK EoS, will be obtained.
 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37 29

Table 1 Let us consider two cubic EoS, noted EoS1 and EoS2. Starting
Values of r1 and r2 for well-known cubic equations of state.
from Eq. (26), we can write:
Equation r1 r2
CEoS1 · ˝bEoS1 · EijEoS1 (T ) = CEoS2 · ˝bEoS2 · EijEoS2 (T ) (30)
Van der Waals (VdW) 0 0
Soave–Redlich–Kwong (SRK) 0 √ –1 √ That is:
Peng–Robinson (PR) −1 + 2 −1 − 2
CEoS1 · ˝bEoS1
EijEoS2 (T ) = 1→2 · EijEoS1 (T ) with 1→2 = (31)
CEoS2 · ˝bEoS2

2.2. Generalization of the previous relationship to any couple of By combining Eqs. (31) and (22), we obtain:
cubic equations of state
 2  2
21→2 kijEoS1 ıEoS1
i
ıEoS1
j
+ 1→2 ıEoS1
i
− ıEoS1
j
− ıEoS2
i
− ıEoS2
j
kijEoS2 = (32)
2ıEoS2
i
ıEoS2
j
Cubic equations of state deriving from the Van der Waals equa-
tion take the general following form: Using the previous equation, it is thus possible to calculate the
kij for a desired cubic EoS (EoS2) knowing the corresponding value
RT ai (T ) for another cubic EoS (EoS1).
P(T, v) = −
v − bi (v − r1 bi )(v − r2 bi ) Eq. (32) can also be used if we work with the same EoS (let us say
⎧ the PR-EoS) but if we decide to only change the ai (T) function (e.g.
⎪ RTc,i
⎪
⎨ bi = ˝b Pc,i we initially work with a Soave-type function [2] for which the kij
with (28) are known and we decide to work with a Mathias and Copeman [20]
⎪
⎪
2
R2 Tc,i function for which the kij are unknown). In this latter case, 1→2 = 1,
⎩ ai (T ) = ˝a · ˛i (T, Tc,i , ωi ) but Eq. (32) in which the ı parameters depend on the a(T) function
Pc,i
will lead to a relationship between the kij to be used with the first
a(T) function and those to be used with the second one.
The parameters r1 and r2 are specific to each equation. Table 1
reminds their values in the case of the most well-know models. The E,∞
2.3. Check on an example of the validity of the hypothesis: gres
CEoS constant defined by Eq. (1) can be calculated from r1 and r2 by: independent of the used EoS
1
1 − r 
2
CEoS = · ln (29) Eq. (24) which connects the kij for the SRK EoS (kijSRK ) and those
r1 − r2 1 − r1 E,∞
for the PR-EoS (kijPR ) relies on the hypothesis that gres is indepen-
In the particular case where: r1 = r2 = 0, then CEoS = 1. dent of the used EoS. In order to check whether this hypothesis

Fig. 1. System methane(1) + CO2 (2). (a) kijPR values used to generate the data. (b) +: pseudo experimental bubble points, *: pseudo experimental dew points. Isothermal dew
and bubble curves calculated with the fitted kijSRK . (c) +: pseudo experimental bubble points, *: pseudo experimental dew points. Isothermal dew and bubble curves calculated
with the kijSRK estimated from Eq. (24).
30 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37

Table 2 with:
Comparison between the kijSRK fitted on pseudo experimental VLE data and the kijSRK
calculated from Eq. (24). ⎧ −1
⎪ R = 8.314472 J mol K−1
⎪
⎪  
Temperature/K Fitted kijSRK kijSRK from Eq. (24) ⎪
⎪ √ √
⎪
⎪ −1 +
3
6 2+8−
3
6 2−8
⎪
⎪ Noting : X = ≈ 0.253076587, one has :
185 0.089 0.092
⎪
⎪
3
200 0.092 0.094 ⎪
⎪ bPR = ˝PR c,i with ˝PR =
RT X
220 0.097 0.098 ⎪
⎪ ≈ 0.0777960739
⎨ i b P
c,i
b X +3
240
260
0.103
0.109
0.102
0.107 2 2

R Tc,i
  2 (34)
T
⎪
⎪ aPR = ˝aPR 1 + mPR 1−
280 0.115 0.112
⎪
⎪
i Pc,i i Tc,i
290 0.118 0.115 ⎪
⎪
⎪
⎪ with ˝aPR = 8(5X + 1) ≈ 0.457235529
300 0.122 0.118
⎪
⎪ 49 − 37X
⎪
⎪ if ω ≤ 0.491 mPR = 0.37464 + 1.54226ω − 0.26992ω2
⎪
⎪
⎪
⎩
i i i i
is valid, it was decided to generate pseudo experimental VLE data if ωi > 0.491 mi PR
= 0.379642 + 1.48503ωi − 0.164423ωi + 0.016666ωi
2 3

using the PR-EoS and temperature-dependent kij for the system

methane(1) + carbon dioxide(2). By this way, the numerical val- P is the pressure, R the gas constant, T the temperature, a and b
ues of the kijPR are perfectly known. In a second step, these pseudo are EoS parameters, v is the molar volume, Tc the critical tempera-
experimental data were correlated with the SRK EoS. For each ture, Pc the critical pressure and ω the acentric factor. In the PPR78
temperature, the kijSRK was fitted in order to minimize the devi- approach, classical mixing rules with kij [see Eq. (5)] or equiva-
ations between calculated and pseudo experimental data. Such lently combination at constant packing fraction of a Van Laar-type
kij can thus be compared to those calculated using Eq. (24). In GE model with the PR78 EoS [see Eqs. (1) and (4)] are used. In order
this example, 181 pseudo experimental data (87 bubble points, to obtain a predictive model, and to completely define the PPR78
87 dew points, and 7 mixture critical points) were generated model, we developed, following the previous works of Redlich et
at eight temperatures ranging from 185 K to 300 K (T1 /K = 185, al. [22], Kehiaian et al. [23] and of Abdoul et al. [24], a group con-
T2 /K = 200, T3 /K = 220, T4 /K = 240, T5 /K = 260, T6 /K = 280, T7 /K = 290 tribution method (GCM) to estimate indifferently the interaction
and T8 /K = 300). In Fig. 1a, the kijPR used to generate the data are parameters Eij (T) in Eq. (4) or the kij (T) in Eq. (5). The following
plotted with respect to the temperature. Such values originate from equations were used:
Vitu et al. [6]. The fitted kijSRK , and the kijSRK calculated from Eq. (24)  BPR 
Ng Ng
1 
 298.15  kl −1
APR
EijPR (T ) = − (˛ik − ˛jk )(˛il − ˛jl )APR
kl
· kl (35)
2 T/K
k=1 l=1

⎡  PR ⎤
Ng Ng
   BklPR −1   PR  PR 2
a (T ) a (T )
− 12 ⎣ ⎦−
A i j
(˛ik − ˛jk )(˛il − ˛jl )APR · 298.15 kl
kl
T/KPR − PR b b
i j
k=1 l=1
kijPR (T ) =  (36)
aPR (T )·aPR (T )
i j
2
bPR ·bPR
i j

are compared in Table 2. In Fig. 1b and c, the corresponding isother- Let us recall that Eq. (36) is simply the coupling of Eqs. (9) (12)
mal dew and bubble curves calculated either with the fitted kijSRK and (35).
In Eqs. (35) and (36), T is the temperature. Ng is the number
or with the kijSRK calculated from Eq. (24) are plotted. From Table 2,
of different groups defined by the method. ˛ik is the fraction of
it clearly appears that the kijSRK calculated using Eq. (24) are very molecule i occupied by group k (occurrence of group k in molecule
close to the fitted kijSRK (the average deviation between the two sets i divided by the total number of groups present in molecule i).
of kijSRK values is only 2%). As a consequence, Fig. 1b and c are more or Akl = Alk and Bkl = Blk (where k and l are two different groups) are
constant parameters determined in our previous studies [3–9]
less undistinguishable. We can thus conclude (on this example) that
E,∞ (Akk = Bkk = 0) by minimizing the deviations between calculated
the molar residual excess Gibbs energy at infinite pressure (gres )
and experimental VLE data from an extended binary-system
is independent of the used EoS. The validity of this hypothesis is
database containing 65,700 experimental data points (36,535
better evidenced in the next section.
bubble points, 28,051 dew points, 58 azeotropic points and 1,056
mixture critical points). Most of the binary experimental data
3. From the PPR78 model to the PR2SRK model available in the open literature were collected. The detailed list of
these experimental data is given in our previous papers [3–9]. For
3.1. Some reminders on the PPR78 model the 15 groups available, a total of 204 parameters (expressed in
MPa) were determined. They are recalled in Table 3. With these
In 1976, Peng and Robinson [21] published their well-known
groups, it is possible to predict the BIPs for any mixture containing
equation of state, called in this paper PR76. In 1978, the same
alkanes, aromatics, naphthenes, CO2 , N2 , H2 S and mercaptans. As
authors published an improved version of their equation of state
recently shown [11], such a model is perfectly able to predict the
[1], which yields more accurate vapor pressure predictions for
behavior of petroleum fluids. A GCM has been chosen to estimate
the heavy hydrocarbons than those obtained by using PR76. This
the BIPs because we were aware that the group contribution
improved equation, used in the PPR78 approach is noted PR78 in
concept could be useful to model complex processes like those
this paper. For a pure component, the PR78 EoS is:
involving supercritical fluids [25–27] and because the number of
aPR binary systems for which phase equilibrium data are available is at
RT i
(T )
P= − (33) most several thousands while the number of the compounds used
v − bPR
i
v(v + bi ) + bPR
PR
i
(v − bPR
i
) now by industry is estimated at around 100,000.
Table 3
Group interaction parameters: (APR
kl
= APR
lk
PR
)/MPa and (Bkl = Blk
PR
)/MPa.

CH3 CH2 CH C CH4 C2 H6 CHaro Caro Cfused aromatic CH2,cyclic CHcyclic or CO2 N2 H2 S SH
(group 1) (group 2) (group 3) (group 4) (group 5) (group 6) (group 7) (group 8) rings (group 10) Ccyclic (group 12) (group 13) (group 14) (group 15)
(group 9) (group 11)
CH3 0 – – – – – – – – – – – – – –
(group 1)

J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37

CH2 A12 = 74.81 0 – – – – – – – – – – – – –
(group 2) B12 = 165.7
CH (group 3) A13 = 261.5 A23 = 51.47 0 – – – – – – – – – – – –
B13 = 388.8 B23 = 79.61
C A14 = 396.7 A24 = 88.53 A34 = −305.7 0 – – – – – – – – – – –
(group 4) B14 = 804.3 B24 = 315.0 B34 = −250.8
CH4 A15 = 32.94 A25 = 36.72 A35 = 145.2 A45 = 263.9 0 – – – – – – – – – –
(group 5) B15 = −35.00 B25 = 108.4 B35 = 301.6 B45 = 531.5
C 2 H6 A16 = 8.579 A26 = 31.23 A36 = 174.3 A46 = 333.2 A56 = 13.04 0 – – – – – – – – –
(group 6) B16 = −29.51 B26 = 84.76 B36 = 352.1 B46 = 203.8 B56 = 6.863
CHaro A17 = 90.25 A27 = 29.78 A37 = 103.3 A47 = 158.9 A57 = 67.26 A67 = 41.18 0 – – – – – – – –
(group 7) B17 = 146.1 B27 = 58.17 B37 = 191.8 B47 = 613.2 B57 = 167.5 B67 = 50.79
Caro A18 = 62.80 A28 = 3.775 A38 = 6.177 A48 = 79.61 A58 = 139.3 A68 = −3.088 A78 = −13.38 0 – – – – – – –
(group 8) B18 = 41.86 B28 = 144.8 B38 = −33.97 B48 = −326.0 B58 = 464.3 B68 = 13.04 B78 = 20.25
Cfused aromatic A19 = 62.80 A29 = 3.775 A39 = 6.177 A49 = 79.61 A59 = 139.3 A69 = −3.088 A79 = −13.38 A89 = 0.0 0 – – – – – –
rings B19 = 41.86 B29 = 144.8 B39 = −33.97 B39 = −326.0 B59 = 464.3 B69 = 13.04 B79 = 20.25 B89 = 0.0
(group 9)
CH2,cyclic A1–10 = 40.38 A2–10 = 12.78 A3–10 = 101.9 A4–10 = 177.1 A5–10 = 36.37 A6–10 = 8.579 A7–10 = 29.17 A8–10 = 34.31 A9–10 = 34.31 0 – – – – –
(group 10) B1–10 = 95.90 B2–10 = 28.37 B3–10 = −90.93 B4–10 = 601.9 B5–10 = 26.42 B6–10 = 76.86 B7–10 = 69.32 B8–10 = 95.39 B9–10 = 95.39
CHcyclic or A1–11 = 98.48 A2–11 = −54.90 A3–11 = −226.5 A4–11 = 17.84 A5–11 = 40.15 A6–11 = 10.29 A7–11 = −26.42 A8–11 = −105.7 A9–11 = −105.7 A10–11 = −50.10 0 – – – –
Ccyclic B1–11 = 231.6 B2–11 = −319.5 B3–11 = −51.47 B4–11 = −109.5 B5–11 = 255.3 B6–11 = −52.84 B7–11 = −789.2 B8–11 = −286.5 B9–11 = −286.5 B10–11 = −891.1
(group 11)
CO2 A1–12 = 164.0 A2–12 = 136.9 A3–12 = 184.3 A4–12 = 287.9 A5–12 = 137.3 A6–12 = 135.5 A7–12 = 102.6 A8–12 = 110.1 A9–12 = 267.3 A10–12 = 130.1 A11–12 = 91.28 0 – – –
(group 12) B1–12 = 269.0 B2–12 = 254.6 B3–12 = 762.1 B4–12 = 346.2 B5–12 = 194.2 B6–12 = 239.5 B7–12 = 161.3 B8–12 = 637.6 B9–12 = 444.4 B10–12 = 225.8 B11–12 = 82.01
N2 A1–13 = 52.74 A2–13 = 82.28 A3–13 = 365.4 A4–13 = 263.9 A5–13 = 37.90 A6–13 = 61.59 A7–13 = 185.2 A8–13 = 284.0 A9–13 = 718.1 A10–13 = 179.5 A11–13 = 100.9 A12–13 = 98.42 0 – –
(group 13) B1–13 = 87.19 B2–13 = 202.8 B3–13 = 521.9 B4–13 = 772.6 B5–13 = 37.20 B6–13 = 84.92 B7–13 = 490.6 B8–13 = 1892 B9–13 = 1892 B10–13 = 546.6 B11–13 = 249.8 B12–13 = 221.4
H2 S A1–14 = 158.4 A2–14 = 134.6 A3–14 = 193.9 A4–14 = 305.1 A5–14 = 181.2 A6–14 = 157.2 A7–14 = 21.96 A8–14 = 1.029 A9–14 = 1.029 A10–14 = 120.8 A11–14 = −16.13 A12–14 = 134.9 A13–14 = 319.5 0 –
(group 14) B1–14 = 241.2 B2–14 = 138.3 B3–14 = 307.8 B4–14 = −143.1 B5–14 = 288.9 B6–14 = 217.1 B7–14 = 13.04 B8–14 = −8.579 B9–14 = −8.579 B10–14 = 163.0 B11–14 = −147.6 B12–14 = 201.4 B13–14 = 550.1
SH A1–15 = 799.9 A2–15 = 459.5 A3–15 = 425.5 A4–15 = 682.9 A5–15 = 706.0 N.A.(*) A7–15 = 285.5 A8–15 = 1072 A9–15 = 1072 A10–15 = 446.1 A11–15 = 411.18 N.A.(*) N.A.(*) A14–15 = −77.21 0
(group 15) B1–15 = 2109 B2–15 = 627.3 B3–15 = 514.7 B4–15 = 1544 B5–15 = 1483 B7–15 = 392.0 B8–15 = 1094 B9–15 = 1094 B10–15 = 549.0 B11–15 = −308.8 B14–15 = 156.1
(*)
N.A. = not available

31
32 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37

Whatever the used method, an equation similar to Eq. (36) is

obtained (it is enough to change the superscript PR in SRK):
⎡  SRK ⎤
Ng Ng
   BklSRK −1   SRK  SRK 2
a (T ) a (T )
1 ⎣ SRK 298.15 A
⎦ i j
−2 (˛ik − ˛jk )(˛il − ˛jl )A · kl
kl −
T/K SRK − SRK b b
i j
k=1 l=1
kijSRK (T ) =  (41)
aSRK (T )·aSRK (T )
i j
2
bSRK ·bSRK
i j

3.2. Construction of the PR2SRK model It is thus possible to use the GCM we developed for the PR-EoS
to predict the temperature-dependent BIPs of the SRK EoS without
For a pure component, the SRK EoS [2] is:
having to re-estimate the parameters. Eqs. (5), (37), (38), (39) and
RT aSRK
i
(T ) (41) thus define a new predictive model based on the SRK EoS. It was
P= − (37)
v − bSRK v(v + bSRK ) decided to call this new model PR2SRK because it is predictive and
i i
because the group interaction parameters are those we developed
with: for the PR-EoS.
⎧ R = 8.314472 J · mol−1 · K−1 3.3. Generalization of the previous methodology in order to
⎪
⎪ √
⎪
⎪ RTc,i 3
2−1 estimate by a GCM, the kij for any cubic EoS using the GC
⎨ bSRK
i
= ˝ SRK
b Pc,i
with ˝ SRK
b
=
3
≈ 0.08664
parameters of the PPR78 model
2 2
  2
⎪
⎪
R Tc,i T 1
⎪
⎪
aSRK =˝aSRK
Pc,i
1 + mSRK
i
1−
Tc,i
with ˝aSRK = √ ≈ 0.42748 Let us consider a cubic EoS as defined by Eq. (28) and noted EoS1.
⎩
i 3
9( 2 − 1)
mSRK = 0.480 + 1.574ωi − 0.176ωi2
Let us define:
i
(38) CPR · ˝bPR
PR→EoS1 = (42)
CEoS1 · ˝bEoS1

Starting from Eqs. (39) and (41), we can write:

⎡  PR ⎤
Ng Ng
   BklPR −1   EoS1  EoS1 2
a (T ) a (T )
− 12 ⎣ ⎦−
A i j
(˛ik − ˛jk )(˛il − ˛jl )PR→EoS1 APR 298.15 kl
kl T/K
EoS1 − EoS1 b b
i j
k=1 l=1
kijEoS1 (T ) =   (43)
aEoS1 (T )·aEoS1 (T )
i j
2
bEoS1 ·bEoS1
i j

Classical mixing rules as defined by Eq. (5) or mixing rules Using the previous equation, it is thus possible to calculate
defined by Eqs. (1) and (4) can obviously be used with the SRK EoS. by GC, the temperature-dependent kij for any desired cubic EoS
However remembering that EijSRK (T ) =  · EijPR (T ) with  ≈ 0.807341, (EoS1), with any desired ai (T) function, using the group contribu-
it is noticeable from Eq. (35) that it is enough to multiply by  the APR tion parameters (APR kl
PR ) we determined for the PPR78 model.
and Bkl
kl
and Bkl PR group interaction parameters to obtain the corresponding Eq. (43) can also be used if we work with the PR-EoS but with
a different ai (T) function than the one defined by Eq. (34). In this
EijSRK (T ) values for the SRK EoS. We thus can write:
latter case, PR→EoS1 = 1, but Eq. (43) will lead to kij values different
 of those obtained from the PPR78 model.
ASRK
kl
=  · APR
kl
(39)
SRK =  · BPR
Bkl kl 4. Results and discussion
and
 BSRK  The extended database, containing 65,700 data points, we pre-

Ng

Ng   kl −1 viously built to develop the PPR78 model, was used in order to
1 298.15 ASRK
EijSRK (T ) = − (˛ik − ˛jk )(˛il − ˛jl )ASRK
kl
· kl (40) determine the accuracy of the PR2SRK model and thus to check the
2 T/K E,∞
k=1 l=1 validity of the hypothesis we made (gres independent of the EoS).
The following objective function was calculated:
The kijSRK (T ) values can now be obtained either by cou- Fobj,bubble + Fobj,dew + Fobj,crit. comp + Fobj,crit. pressure
pling Eq. (36) and Eq. (24) or by sending Eq. (40) in Eq. (12). Fobj = (44)
nbubble + ndew + ncrit + ncrit
⎧    
⎪
n
bubble x x      
⎪
⎪ Fobj,bubble = 100 + with x = x1,exp − x1,cal  = x2,exp − x2,cal 
⎪
⎪
0.5
x1,exp x2,exp
⎪
⎪
⎪
⎪
i=1
    
i
⎪
⎪ ndew
   
⎪
⎪  y y      
⎪
⎪ Fobj,dew = 100 0.5 + with y = y1,exp − y1,cal  = y2,exp − y2,cal 
⎪
⎨ y1,exp y2,exp
i=1 i
    
⎪
⎪ ncrit
xc  xc       
⎪
⎪ Fobj,crit. comp = 100 0.5 with xc  = xc1,exp − xc1,cal  = xc2,exp − xc2,cal 
⎪
⎪ x
+
x
⎪
⎪ c1,exp c2,exp
⎪
⎪ i=1

i
⎪
⎪  
⎪
⎪
ncrit
  Pcm,exp − Pcm,cal 
⎪
⎪
⎩ Fobj,crit. pressure = 100 Pcm,exp
i=1 i
 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37 33

nbubble , ndew and ncrit are the number of bubble points, dew points
and mixture critical points respectively. x1 is the mole fraction in
the liquid phase of the most volatile component and x2 the mole
fraction of the heaviest component (with x2 = 1 − x1 ). Similarly, y1
is the mole fraction in the gas phase of the most volatile compo-
nent and y2 the mole fraction of the heaviest component (with
y2 = 1 − y1 ). xc1 is the critical mole fraction of the most volatile com-
ponent and xc2 the critical mole fraction of the heaviest component
(with xc2 = 1 − xc1 ). Pcm is the binary critical pressure.
For the 65,700 data points included in our database, the objec-
tive function defined by Eq. (44) is: Fobj = 7.9% (for the same
database, the PPR78 model leads to: Fobj PPR78 = 7.6%). The average

overall deviation on the liquid phase composition is:

⎧ nbubble  
⎪ x1,exp − x1,cal 
⎪
⎪ i=1 i
⎨ x1 = x2 = nbubble
= 0.022 Fig. 2. Plot of the objective function defined by Eq. (44) as a function of the 
parameter.
and .
⎪
⎪
⎪
⎩ x% = x1 % + x2 % = Fobj,bubble = 7.86 %
2 nbubble

Fig. 3. Temperature dependence of estimated binary interaction parameters (kij ) by means of Eqs. (36) and (41). Solid line: kij for the PR-EoS. Dashed line: kij for the SRK
EoS. (a) System methane + cyclohexane. (b) System n-undecane + n-eicosane. (c) System propane + benzene. (d) System ethane + hydrogen sulfide. (e) System 2,2-dimethyl
butane + 2,4-dimethyl pentane.
34 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37

Fig. 4. Prediction of isothermal or isobaric dew and bubble curves for 4 binary systems using the PR2SRK model. Solid line: predicted curves with the PR2SRK model. +: exper-
imental bubble points, *: experimental dew points. (a) system methane(1)/ethane(2) at 5 different temperatures: T1 = 130.37 K (kij = 0.00057), T2 = 172.04 K (kij = 0.0012),
T3 = 199.93 K (kij = 0.0025), T4 = 230.00 K (kij = 0.0040), T5 = 283.15 K (kij = 0.0069). (b) system methane(1)/decane(2) at 5 different temperatures: T1 = 410.93 K (kij = 0.046),
T2 = 477.59 K (kij = 0.059), T3 = 510.95 K (kij = 0.068), T4 = 563.25 K (kij = 0.086), T5 = 583.05 K (kij = 0.095). (c) system propane(1)/1,3,5-trimethyl benzene (2) at 3 different temper-
atures: T1 = 313.10 K (kij = 0.017), T2 = 403.10 K (kij = 0.023), T3 = 473.00 K (kij = 0.030). (d) system ethane(1)/cyclohexane(2) at 4 different pressures: P1 = 24.1 bar, P2 = 44.8 bar,
P3 = 58.6 bar, P4 = 82.7 bar.

Fig. 5. Prediction of isothermal or isobaric dew and bubble curves for 4 binary systems using the PR2SRK model. Solid line: predicted curves with the PR2SRK model. +:
experimental bubble points, *: experimental dew points. (a) System carbon dioxide(1)/ethane(2) at 4 different temperatures: T1 = 250.00 K (kij = 0.14), T2 = 260.00 K (kij = 0.14),
T3 = 270.00 K (kij = 0.14), T4 = 283.15 K (kij = 0.14). (b) System nitrogen(1)/toluene(2) at 2 different temperatures: T1 = 498.15 K (kij = −0.23), T2 = 548.15 K (kij = −0.31). (c) System
2-methyl butane (1)/ethyl mercaptan (2) at P1 = 5.88 bar. (d) System hydrogen sulfide(1)/propane(2) at T = 243.2 K (kij = 0.062).
 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37 35

Fig. 6. Prediction of the critical locus of various binary mixtures using the PR2SRK model. Solid line: predicted critical locus. Dashed line: pure component vaporization
curve. +: pure component critical point. , , ×, , *,,♦: experimental data. (a) Critical locus of ethane + n-alkane binary mixtures. (b) Critical locus of 6 binary systems
containing aromatic compounds. (c) Critical locus of benzene + n-alkane binary mixtures. (d) Critical locus of 8 binary systems containing paraffinic, naphthenic and aromatic
compounds.

Fig. 7. Prediction of the critical locus of various binary mixtures using the PR2SRK model. Solid line: predicted critical locus. Dashed line: pure component vaporization curve.
+: pure component critical point. : experimental data. (a) Critical locus of 6 binary systems containing CO2 and a n-alkane. (b) critical locus of 2 binary systems containing
CO2 and a n-alkane. (c) Critical locus of the methane + H2 S system.
36 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37

The average overall deviation on the gas phase composition is: behavior. For this reason, it has been decided to predict the critical
⎧ ndew
locus of various binary mixtures by using the PR2SRK equation of
⎪
⎪  state. A comparison between theory and experiment is illustrated
⎪
⎪ (|y1,exp − y1,cal |)i
⎪
⎪ in Figs. 6 and 7. The overall quality of predictions obtained from the
⎨ i=1 PR2SRK model are very good for type I phase diagrams in the classi-
y1 = y2 = = 0.014
ndew . fication scheme of Van Konynenburg and Scott [28]. Nevertheless,
⎪
⎪
⎪
⎪
and binary systems exhibiting type III phase behavior are more difficult
⎪
⎪ y1 % + y2 % Fobj,dew
⎩ y% = = = 8.09 %
to predict with accuracy.
2 ndew
5. Conclusion
The average overall deviation on the critical composition is:
⎧ ncrit
⎪
⎪    A group contribution method allowing the estimation of the
⎪
⎪ (xc1,exp − xc1,cal ) temperature-dependent binary interaction parameters (kij (T)) for
⎪
⎪ i
⎨ i=1
the widely used Soave–Redlich–Kwong equation of state has been
xc1 = xc2 = = 0.017 developed. This new model has been called PR2SRK. The group
ncrit .
⎪
⎪ interaction parameters (ASRK SRK ) used in this model are those
and Bkl
⎪
⎪
and kl
⎪
⎪ xc1 % + xc2 % Fobj,crit.comp previously determined for the PR-EoS multiplied by  = 0.807. It was
⎩ xc % = = = 6.57 % thus not necessary to re-estimate the parameters for the SRK EoS. A
2 ncrit
very long, tedious and difficult job has thus been saved. In a paper
The average overall deviation on the binary critical pressure is: to be published as soon as possible, it will be shown that by fit-
ting the group interactions of the SRK EoS (ASRKkl
and BklSRK ) directly
⎧ ncrit on the 65,700 VLE data points used to elaborate the PPR78 model,
⎪
⎪   
⎪
⎪ Pcm,exp − Pcm,cal  only a small improvement of the SRK model can be obtained in
⎪
⎪ i
⎨ i=1
comparison to the PR2SRK model developed in this study.
Pc = = 6.39 bar . The key idea used in this paper was to assume that the residual
n E,∞
⎪
⎪
crit
molar excess Gibbs energy at infinite pressure (gres ) was indepen-
⎪
⎪
and
⎪
⎪ F dent of the EoS when one combines at constant packing fraction a
⎩ Pc % = obj,crit. pressure = 3.57 % Van Laar-type GE function and a cubic EoS.
ncrit
We have also shown that the mathematical equations derived
These results clearly indicate that the PR2SRK model is really for the SRK EoS are general and applicable to any cubic EoS. It is
an accurate predictive model. The deviations are very similar to thus possible to calculate by GC, the temperature-dependent kij for
those obtained with the PPR78 approach: x1 and y1 are exactly any desired cubic EoS, with any desired ai (T) function, using the
the same for both these models. The critical compositions and group-contribution parameters (APR kl
PR ) we determined for
and Bkl
the binary critical pressures are however slightly better predicted the PR-EoS.
with the PPR78 model. We can thus conclude that the hypothesis The most important conclusion is that the results obtained with
E,∞
we made (gres independent of the EoS) is pertinent. In order to the PR2SRK model are in many cases very accurate in both the
strengthen this conclusion, it was decided to plot Fobj defined by sub-critical and critical regions. We can thus conclude that the
E,∞
Eq. (44) as a function of the  parameter. Our idea was to check hypothesis we made (gres independent of the EoS) is pertinent. The
whether the value of this parameter determined from Eq. (17), that equations established for the SRK EoS only relies on this hypoth-
is  =0.807, was the optimized value. In this study,  was varied esis. They are thus general and applicable to any cubic EoS. As a
between 0.4 and 1.3 and the resulting curve smoothed with a sixth consequence, whatever the considered EoS, we can trust in the kij
order polynomial may be seen in Fig. 2. From this figure, the ␰ value estimated by GCM from the PPR78 model.
minimizing the objective function is: optimized = 0.808 which is
extremely close to the theoretical value (theoretical = 0.807) defined List of symbols
by Eq. (17). This result undeniably reinforces our conviction that a(T) temperature-dependent function of the equation of state
E,∞
gres is independent of the EoS. Akl , Bkl group interaction parameters allowing the calculation of
At this step, it looks interesting to compare the numerical values the BIPs
of the kij for both the PR and the SRK EoS. Some typical illustra- b covolume
tions, where the temperature varies between 70 and 700 K can be Eij binary interaction parameter of the Van Laar excess func-
found in Fig. 3. From this figure it is possible to conclude that both tion
EoS (PR and SRK) need different kij values. The difference between GE expression for the molar excess Gibbs energy (i.e. activity
the two sets of kij is however not very large. Depending on the coefficient model)
system, the kij for the PR-EoS may be larger or smaller than the GEres residual term of an activity coefficient model
kij for the SRK EoS. We can also conclude that the PPR78 and the E
gres residual part of the molar excess Gibbs energy calculated
PR2SRK models are able to predict the different kij trends commonly from an EoS
E,∞
encountered. Indeed, the kij may increase, decrease or go through gres residual part of the molar excess Gibbs energy under infi-
with an extremum with respect to temperature. nite pressure calculated from an EoS
Figs. 4 and 5 illustrate for some typical binary systems the accu- kij binary interaction parameter
racy of the proposed PR2SRK model. From these figures, we can m shape parameter
conclude that accurate results are obtained whatever the tem- P pressure
perature and the pressure. However, as previously explained [7], Pc critical pressure
systems containing nitrogen are difficult to predict with a cubic R gas constant
EoS. T temperature
The prediction of phase equilibria of binary mixtures in the criti- Tc critical temperature
cal region is an important problem because the shape of the critical v volume
locus gives necessary basic information about the topology of phase xi , yi , zi mole fraction
 J.-N. Jaubert, R. Privat / Fluid Phase Equilibria 295 (2010) 26–37
Greek letters
ω acentric factor
˛ik fraction occupied by group k in the molecule i
References
[1] D.B. Robinson, D.Y. Peng, The characterization of the heptanes and heavier
fractions for the GPA Peng–Robinson programs, Gas processors association,
Research report RR-28, 1978, (booklet only sold by the GPA = gas processors
association).
[2] G. Soave, Chem. Eng. Sci. 27 (1972) 1197–1203.
[3] J.-N. Jaubert, F. Mutelet, Fluid Phase Equilib. 224 (2) (2004) 285–304.
[4] J.-N. Jaubert, S. Vitu, F. Mutelet, J.-P. Corriou, Fluid Phase Equilib. 237 (1/2)
(2005) 193–211.
[5] S. Vitu, J.-N. Jaubert, F. Mutelet, Fluid Phase Equilib. 243 (2006) 9–28.
[6] S. Vitu, R. Privat, J.-N. Jaubert, F. Mutelet, J. Supercrit. Fluids 45 (1) (2008) 1–26.
[7] R. Privat, J.-N. Jaubert, F. Mutelet, Ind. Eng. Chem. Res. 47 (6) (2008)
2033–2048.
[8] R. Privat, J.-N. Jaubert, F. Mutelet, J. Chem. Thermodyn. 40 (2008)
1331–1341.
[9] R. Privat, F. Mutelet, J.-N. Jaubert, Ind. Eng. Chem. Res. 47 (2008) 10041–10052.
[10] R. Privat, J.-N. Jaubert, F. Mutelet, Ind. Eng. Chem. Res. 47 (2008) 7483–7489.
[11] J.-N. Jaubert, R. Privat, F. Mutelet, AIChE J. (2010), doi:10.1002/aic.12232.
37
[12] F. Mutelet, S. Vitu, R. Privat, J.-N. Jaubert, Fluid Phase Equilib. 238 (2005)
157–168.
[13] S. Vitu, J.-N. Jaubert, J. Pauly, J.-L. Daridon, D. Barth, J. Chem. Eng. Data 52 (2007)
1851–1855.
[14] S. Vitu, J.-N. Jaubert, J. Pauly, J.-L. Daridon, D. Barth, J. Supercrit. Fluids 44 (2)
(2008) 155–163.
[15] S. Vitu, J.-N. Jaubert, J. Pauly, J.-L. Daridon, J. Chem. Thermodyn. 40 (2008)
1358–1363.
[16] R. Privat, F. Garcia, J.-N. Jaubert, M. Moliere, J. Eng. Gas Turbine Power (2010),
doi:10.1115/1.4000610.
[17] E.A. Guggenheim, Mixtures, Oxford University Press, Oxford, 1952.
[18] A. Péneloux, W. Abdoul, E. Rauzy, Fluid Phase Equilib. 47 (2–3) (1989) 115–132.
[19] E. Fransson, L. Vamling, J. Vidal, Chem. Eng. Sci. 48 (1993) 1753–1759.
[20] P.M. Mathias, T.W. Copeman, Fluid Phase Equilib. 13 (1983) 91–108.
[21] D.Y. Peng, D.B. Robinson, Ind. Eng. Chem. Fundam. 15 (1976) 59–64.
[22] O. Redlich, E.L. Derr, G.J. Pierotti, J. Am. Chem. Soc. 81 (1959) 2283–2285.
[23] H.V. Kehiaian, K. Sosnkowska-Kehiaian, R. Hryniewicz, J. Chim. Phys. Phys.
Chim. Biol. 68 (6) (1971) 922–934.
[24] W. Abdoul, E. Rauzy, A. Péneloux, Fluid Phase Equilib. 68 (1991) 47–102.
[25] J.-N. Jaubert, L. Coniglio, F. Denet, Ind. Eng. Chem. Res. 38 (1999) 3162–3171.
[26] J.-N. Jaubert, L. Coniglio, Ind. Eng. Chem. Res. 38 (1999) 5011–5018.
[27] J.-N. Jaubert, P. Borg, L. Coniglio, D. Barth, J. Supercrit. Fluids 20 (2001) 145–155.
[28] P.H. Van Konynenburg, R.L. Scott, Philos. Trans. R. Soc. Lond. Ser. A 298 (1980)
495–540.