Modeling Vapor-Liquid Equilibria

4.2. The Huron-Vidal (HVO) Model

Vidal (1978) and later Huron and Vidal (1979) proposed the first successful combi-
nation of an EOS and activity coefficient models by requiring that the mixture EOS
at liquid densities should behave like an activity coefficient model. To ensure a liquid
density at all temperatures, this equality was made using eqn. (4.1.7) and the relation

G?(7\ P = oo, x) = GeExos(r, P = oo, x) (4.2.1)

-r
where G**{T, P = ocyjf) and Crgos(T.. P — oo, x) are the excess Gibbs free energies
-y
at infinite pressure (i.e., at liquid-like densities) calculated from an activity coefficient
model and the EOS, respectively. Because Gex = £* + P Ves, with Aex in the liquid
state being almost independent of pressure, to use eqn. (4.1.7) it is necessary that

y« = v _ J^mVj = b - J^**** = ° ^ 4 - 2 - 2 )
i i

since at infinite pressure from an equation of state V_,- = ht.

That is, to keep Gfos finite in order to use eqn. (4.1.7), eqn. (4.2.2) must also be
used. Equations (4.1.7 and 4.2.2) provide the two equations necessary to determine
the two EOS constants. The resulting mixing rule for the a parameter is

^xw
a = b (4.2.3)

where Gex is an excess free energy of mixing expression appropriate for the mixture
of interest. Equations (4.2.3 and 3.3.8) constitute the original Huron-Vidal mixing
rule. We will refer to this mixing rule as the HVO model in this monograph.
Some results obtained with the HVO mixing rule are presented in the following
paragraphs. These results were obtained with the computer program HV provided on
the accompanying diskette; the computational details are presented in Section D.4 of
Appendix D.
The correlation of data for the methane and pentane binary system is shown in
Figure 4.2.1. In this case the van Laar excess Gibbs free-energy model has been used
in the HVO model; the two model parameters were fitted to VLE data on the 277 K
isotherm, and the vapor-liquid equilibria at higher temperatures were predicted with
the same temperature-independent parameters. The results are very good in this case
and similar to those obtained with the 1PVDW and 2PVDW models.
The results for the carbon dioxide and propane binary system, shown as dashed
lines in Figure 4.2.2, on the other hand are not as good. When compared with the
performance of the 1PVDW model (solid lines in Figure 4.2.2), the use of the same
parameters for all isotherms leads to inferior results at higher temperatures despite
the use of an extra parameter in the Huron-Vidal model. This indicates that, for the
mixtures containing supercritical components, the HVO mixing rule, when combined
Mixing Rules that Combine an Equation of State with an Activity Coefficient Model

180

CO
-Q

a
V)
w
0>

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
mole fraction of methane
Figure 4.2.1. VLE correlation of the methane and
pentane binary system at 310, 377, and 444 K with
the Huron-Vidal original (HVO) mixing rule with the
van Laar excess free-energy model and the PRSV equa-
tion of state. The van Laar model parameters used are
fc/jfl = A12/A21 =0.1201/0.1430. Points are experi-
mental data from the DECHEMA Chemistry Data Se-
ries, Gmehling and Onken 1977, Vol. 6, p. 445; data
files for this system on the accompanying disk are
C1C5310.DAT, C1C5377.DAT andClC5444.DAT.

with the conventional excess free-energy models (such as that of van Laar) is inferior
to the van der Waals one-fluid model. This was observed earlier by Shibata and Sandler
(1989). It should be noted, however, that by choosing specific algebraic forms for the
excess free-energy term it is possible to reduce the HVO model to the 1PVDW model.
See, for example, Huron and Vidal (1979) and Orbey and Sandler (1995a,c) for the
details of such modified excess free-energy models.
For more nonideal mixtures, the HVO model shows good con-elation capabilities
but is not satisfactory for extrapolation over a range of temperatures. See, for example,
the results presented in Figures 4.2.3 and 4.2.4 for the acetone and water binary system
and those in Figures 4.2.5 and 4.2.6 for the 2-propanol and water binary system. In
these figures, the dashed lines are direct correlations of the isothermal VLE data with
the HVO mixing rule. The solid lines are predictions with model parameters obtained
from the DECHEMA Chemistry Series at or near room temperature for the excess
free-energy model. In each case the model was observed to be superior to both the
one-parameter (1 PVDW) and the two-parameter (2PVDW) van der Waals models for
the correlation of VLE data. On the other hand, poor predictive performance observed
in these figures indicates that, even though the Huron-Vidal approach allows the
use of G es models with an EOS, one cannot use excess Gibbs free-energy model
parameters obtained from the y-cfi method at low pressure (for example those in the

'•9
 Modeling Vapof-Liqutd Equilibria

70

60

50
«
B
<u 40
i_
3
OJ

S 30
Q.

10

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

mole fraction of carbon dioxide

Figure 4.2.2. VLE correlation (dashed lines) of
the carbon dioxide and propane binary system
at 277, 310, and 344 K with the Huion-Vidal
original (HVO) mixing rule with the van Laar
excess free-energy model and the PRSV equation
of state. The van Laar model parameters used are
%/P = Al2/A2i =1.1816/1.6901. Solid lines repre-
sent the 1PVDW model correlations presented earlier
in Figure 3.4.2. (Points are experimental data from the
DECHEMA Chemistry Data Series, Gmehling and
Onken 1977, Vol. 6, p. 589; data files for this system
on the accompanying disk are C02C3277.DAT,
CO2C3310.DAT and C02C3344.DAT.)

DECHEMA Chemistry Data Series) with this EOS model. This is because the excess
Gibbs free energy of mixing from experiment and as calculated from an EOS is
very pressure dependent; therefore, the excess Gibbs free energies at infinite pressure
and at the pressure at which experimental data were obtained can be very different.
Consequently, a fundamental shortcoming of the Huron-Vidal approach is the use of
the pressure-dependent Gibbs excess free-energy in the EOS rather than Helmholtz
excess free-energy, which is much less pressure dependent. This shortcoming was
corrected by Wong and Sandler (1992), and their work is discussed next.

4.3. The Wong-Sandier (WS) Mode!

Wong and Sandler (1992) have developed a mixing rule that combines an EOS with
a free-energy model but produces the desired EOS behavior at both low and high
densities without being density dependent, uses the existing table of Gex parameters,
allows extrapolation over wide ranges of temperature and pressure, and provides a

50
Mixing Rules that Combine an Equation of State with an Activity Coefficient Mode

1—' 1—i—r 1—i- • i

0.3 - o«VLE date al 298 K

•
_ * - • • ^ ® /

m
i \
0.2
/ s -
• / //
//
/
• / /
f /of
!
0.1 r / ¥//
4 / /7
• / i ^
i /
:/
, i i

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
mole fraction of acetone
Figure 4,2.3. VLE correlation of the acetone and wa-
ter binary system at 298 K with the Huron-Vidat orig-
inal (HVO) mixing rule with the van Laar excess free-
energy model and the PRSV equation of state. The
dashed lines denote results calculated with van Laar
model parameters^ = A| 2 /A 2 i = 3.5121/2.2227
obtained from fitting the experimental data, and the
solid lines represent the results obtained with model
parameters £//3 = A 12 /A 2) = 1.9399/1.8022 ob-
tained at the same temperature from the DECHEMA
Chemistry Series (Gmehling and Onken 1977, Vol. 1,
Pt. 1, p. 238).

conceptually simple method of accurately extending UNIFAC or other low-pressure

VLE prediction methods to high temperature and pressure. This mixing rule is based
on several observations. The first is that, although eqns. (3.3.4 and 3.3.5) are sufficient
conditions to ensure the proper composition dependence of the second virial coef-
ficient, they are not necessary conditions. In particular, the van der Waals one-fluid
mixing rules of eqns. (3.3.4 and 3.3.5) place constraints on the two functions a and b
to satisfy the single relation
a (4.3.1)
B(x,r) = JJJJx^B, 7 (r) RT
= h-
RT
i i S
The original version of the Wong-Sandier mixing rule (Wong and Sandler 1992) uses
the last equality of eqn. (4.3.1) as one of the restrictions on the parameters together
with the combining rule
a a, (4.3.2)
h- h (i-%)
~RT ~RT RT
which introduces a second virial coefficient binary interaction parameter k:j. Note that
eqn. (4,3.1) does not provide relations for the parameters a and b separately but only

l at&TOPHANESI
51
 Modeling Vapor-Liquid Equilibria

60

1 50

• o VLE data at 523 K

40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
mole fraction of acetone
Figure 4.2.4. VLE correlation of the acetone and wa-
ter binary system at 523 K with the Huron-Vidal orig-
inal (HVO) mixing rule combined with the van Laar
excess free-energy mode] and the PRSV equation of
state. The dashed lines represent results calculated
with van Laar model parameters f//} = A12/A21 =
4.2206/1.7264 obtained from luting the experimental
data, and the solid lines denote the results obtained with
model parameters £/£ = A12/A21 = 2.1700/1.7264
from the DECHEMA Chemistry Series at 353 K
(Gmehling and Onken 1977, Vol. 1, Pt. 1. p. 334).

for the sum {b — a/RT), and thus an additional equation is needed. Also, by using
eqn. (4.3.1) as one of the relations to determine the EOS parameters, the proper com-
position dependence of the second virial coefficient is assured, regardless of which
additional equation is used.
The second equation in their mixing rule is based on the observation that the excess
Heimholtz free-energy of mixing calculated from a cubic EOS is much less sensitive
to pressure than the Gibbs free-energy, as can be seen in Figure 4.3.1. Consequently.
to an excellent approximation:

CftX, P = 1 bar, x) = A ex (7\ P = 1 bar, x) = A^T, high pressure, x) (4.3.3)

The first of these equalities results from the relation (7K = Aex + P Vex and that the
pyex t e r m a t j o w p r e s s u r e s xhe second of these equalities is a result of the essential
pressure independence of Aex at liquid densities. Therefore, the second equation for
the a and b parameters comes from eqn. (4.3.3) in the form of

AeE*os(7\ p = oo,x)= g*{T, P = oo, x) = A ; X ( T , low P, x)

= Ge;(7\ low P, x) (4.3.4)

52
Mixing Rules that Combine an Equation of State with an Activity Coefficient Model

1.1 1
—i—'—r

1.0

0.9
* * *

J3
0.S -
f
i

01 *
S °-7 / // /
! // / /
0.6 /® Xs 6* VLE data at 353 K
f /
0.5 1 /
1 / s

1 1 I . I . I .
0.4
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
mole fraction 2-propanol
Figure 4.2.5. VLE correlation of the 2-propanol and
water binary system at 353 K with the Huron-Vidal
original (HVO) mixing rule combined with the NRTL
excess free-energy model and the PRSV equation
of state. The dashed lines represent results calcu-
lated with a =0.2893 and T | 2 / T 2 1 =0.7882/3.9479
obtained from fitting the experimental data, and
the solid lines denote results calculated with a =
0.2893 and T 1 2 / T 2 , =0.1509/1.8051 obtained from
the DECHEMA Chemistry Series at 303 K (Gmehling
and Onken 1977, Vol. 1, Pt. 1, p. 325).

Combining eqns. (4.3.1, 4.3.4, and 4.1.6) gives the following mixing rules:

a a
= J212x'xi (4.3.5)

and

CRT bRT £* a,
biRT
(4.3.6)

where the cross term in eqn. (4.3.5) is obtained from eqn. (4.3.2). Any excess Gibbs
free-energy model may be used in eqn. (4.3.6). The model parameters are the parameter
k,j in eqn. (4.3.2) and the parameters of the excess Gibbs free-energy model used in
eqn. (4.3.6). This mixing rule is referred to as the WS mixing rule in the remainder
of this monograph.
The WS mixing rule satisfies the low-density boundary condition that the second
virial coefficient be quadratic in composition and the high-density condition that ex-
cess free energy be produced like that of currently used activity coefficient models,
whereas the mixing rule itself is independent of density. This model provides a cor-
rect alternative to the earlier ad hoc density-dependent mixing rales (Copeman and

53
 Modeling Vapor-Liquid Equilibria

75

c**0 • * • . .

§
3
h

am VLE data at 523 K

35
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

mole fraction of 2-propanol

Figure 4.2.6. VLE correlation of the 2-propanol and
water binary system at 523 K with the Huron-Vidai
original (H VO) mixing rule combined with the NRTL
excess free-energy model and the PRSV equation
of state. The dashed lines represent results calcu-
lated with a =0.2893 and T 12 /r 2i =0.3952/4.1518
obtained by fitting the experimental data, and the
solid lines denote results calculated with a — 0.2893
and r 12 /T 2 j =0.1019/1.2185 obtained from the
DECHEMA Chemistry Series (Gmehling and Onken
1977, Vol. l,Pt. l , p . 325).

0.450
G « at 1 bar, A «* at 141 000 bar

0.350

-0.050
0.0 0.2 0.4 0.6 0.8 1.0
mole fraction of methanol
Figure 4.3.1. The excess Gibbs and Helmholtz energies
of mixing for the methanol and benzene binary system at
373 K calculated with the Wong-Sandier (WS) mixing rule
and the PRSV equation of state at 1 and 1000 bar.

54
Mixing Rules that Combine an Equation of State with an Activity Coefficient Model

Mathias 1986; Michel, Hooper, and Prausnitz 1989; Sandler el al. 1986) and retains
an important feature of the one-fluid model: that the EOS for the pure fluids and the
mixture have the same density dependence.
This mixing rule has been successful in several ways. First, when combined with any
cubic EOS that gives the correct vapor pressure and an appropriate activity coefficient
model for the Gex term, it has been shown to lead to very good correlations of vapor-
liquid, liquid-liquid, and vapor-liquid-liquid equilibria, indeed generally comparable
to those obtained when the same activity coefficient models are used directly in the y-<p
method. Consequently, this mixing rule extends the range of application of equations
of state to mixtures that previously could be correlated only with activity coefficient
models.
Second, because low-pressure (7X information has been used in developing this
mixing rule, Wong, Orbey, and Sandler (1992) found that activity coefficient parame-
ters reported in data banks, such as the DECHEMA Data Series, could be used directly
and with good accuracy in the WS mixing rule without the need of refitting any exper-
imental data. In this case, however, one point about the binary interaction parameter
kjj needs to be made. The excess Helmholtz free-energy calculated from an EOS is
not necessarily independent of pressure for an arbitrary value of ktj. Therefore, the
binary interaction parameter k,j should be chosen so that the G" curve is reproduced
as closely as possible by A | Q S at the pressure at which the activity coefficient model
parameters are reported. Thus, the parameter ktj contains no further information than
that already included in G".
The correlative and predictive capabilities of the WS mixing rule are shown for
some systems in Figures 4.3.2 to 4.3.9. The results obtained were calculated using
the computer program WS on the diskette provided with this monograph, and the
programming details are given in Section D.5 of Appendix D. Results obtained for
the methane and M-pentane binary mixture are presented in Figure 4.3.2. In this case
the model parameters are fitted to VLE data at 277 K, and the VLE behavior at higher
temperatures is predicted with those values of parameters. The results are good but
slightly less accurate than those obtained by the 1PVDW model correlated to each
isotherm. In contrast, the correlation of the carbon dioxide and propane binary system
shown in Figure 4.3.3 is superior to both the HVO and the 1PVDW models. However,
the true advantage of the WS model is in predicting the vapor-liquid equilibria of
highly nonideal mixtures containing condensable components for which activity co-
efficient model parameters are available only at or near room temperature. In Figures
4.3.4 and 4.3.5, the results for the 2-propanol and water binary system are presented
at 353 and 523 K, respectively. In these figures the dashed lines are obtained from the
direct fit of the model parameters to the experimental data, whereas the solid lines are
predictions with Gex model parameters that have been obtained from the DECHEMA
correlation from data at 303 K. The results show that the correlations are excellent.
but more importantly the predictions at temperatures as much as 200 K above the cor-
relation temperature are almost as accurate as the correlations. Similar results, shown

55