Fluid Phase Equilibria, 60 (1990) 47-58

47

Elsevier Science Publishers B.V., Amsterdam

A METHOD FOR INCORPORATING EXCESS GIBBS ENERGY

MODELS IN EQUATIONS OF STATE
MICHAEL L. MICHELSEN
Instituttet for Kemiteknik, Bygning 229, DTH, DK-2800 Lyngby (Denmark)

(Received September 12,1989; accepted in final form April 14, 1990)

ABSTRACT
Michelsen, M.L., 1990. A method for incorporating
of state. Fluid Phase Equilibria, 60: 47-58..

excess Gibbs energy models in equations

A procedure for incorporating information from an excess Gibbs energy model into an
equation of state is described. The gE model is used to define mixing rules for an equation of
state mixture parameter in such a manner that the equation of state reproduces the
vapour-liquid equilibrium predicted by the gE model at low temperatures.
The resulting equation of state is fully consistent, performs reasonably at elevated
temperatures and is computationally efficient. Typically, the cost of evaluating fugacity
coefficients exceeds that of evaluating activity coefficients from the excess Gibbs energy
model by 25-5048.

INTRODUCTION

The suggestion of incorporating excess Gibbs energy model information

into equations of state originates from Huron and Vidal(1979), who derived
a mixing rule for an equation of state mixture parameter by equating gE at
infinite pressure. The resulting model was capable of correlating high
temperature phase equilibrium but could not directly utilize parameter
tables for existing gE models, correlated from low temperature data.
GuptC et al. (1986) and later Gani et al. (1989) formulated a principle for
calculating the mixture parameter based on a condition of equality of the
excess Gibbs energy at the system temperature and pressure. Although this
approach enables direct utilization of low temperature parameters, several
drawbacks can be noted. Most important, the unspecified mixture parameter
is volume dependent, and pressure and volume thus become related through
a differential equation rather than through the conventional algebraic equation. Analytic treatment of the equation of state is difficult, and fugacity
coefficients are cumbersome to evaluate.
0378-3812/90/$03.50

0 1990 - Elsevier Science Publishers B.V.

48

The suggestion by Mollerup (1986) of equating the excess Gibbs energy at

zero pressure also permits direct utilization of existing low temperature
correlations at conditions where liquid-like properties for all the components of the mixture can be calculated at zero pressure. The present
procedure, like the very recent proposal by Heidemann and Kokal (1990),
combines the suggestion of Mollerup with an extrapolation procedure for
higher temperatures.
DEVELOPMENT

OF MODEL

The basis for our analysis is a model for the residual Helmholtz energy of
a mixture, of the general form
areS/RT=f(u)

- cre(r.4)

(1)

with u = u/b, where b is a mixture parameter with prespecified composition

dependence and LXis a volume-independent
mixture parameter, for which
the composition dependence has to be determined.
The corresponding pressure equations are:
P/RT

= ( - i3/ih)( des/~T)

+ l/u

or
Pb/RT

= -f,

+ (Ye,, + l/u

(2)

where f, and e, are the partial derivatives of f and e, respectively

constant temperature.
Familiar examples are the equations based on

at

f= ln[ u/( u - l)]

e = ln[(u + cA/(u

+ cr>]/(c~ - 4

for which
P=RT/(u-b)-a/[(u+c,b)(u+c,b)]

(3)

with
a = abRT

(4)

yielding the Soave-Redlich-Kwong

equation (Soave, 1972) ( c1 = 0, c2 = l),
c,=l+&)
(Pengand Robinthe Peng-Robinson equation ( c1 = l-a,
son, 1976), or the van der Waals equation (cr = 0, c2 = 0).
Conventionally, these equations employ a linear mixing rule for b and a
quadratic mixture rule for a.
The excess Gibbs energy gE(EOS) is related to ares through
gE/RT

= ures/RT + Pu/RT

- 1 - ln( u) - Cizi ln[ bL( T, P)/RT]

(5)

49

where zi is the mixture mole fraction of component i, and f; the pure

component fugacity at the mixture temperature and pressure.
Our objective is to determine a mixture rule for the mixture parameter (Y
in such a manner that the excess Gibbs energy given by eqn. (5) is in close
agreement with the excess Gibbs energy calculated from an independently
specified gE correlation, gE*(T, z), under conditions where such a correlation is valid.
Thus, our match requires that
areS/RT + Pu/RT - 1 - ln( u) = gE*/RT

+ Cizi ln[ bfi^(T, P)/RT]

(6)

It is immediately evident that eqn. (6) cannot be exactly satisfied over a

range of pressures if we require that a must be volume independent.
However, gE correlations are usually based on experimental data at low
pressures, with the implicit assumption that the pure component fugacities
refer to liquid phases. Typically, the reduced temperatures for the mixture
components are in the range 0.6-0.7, and the reduced pressures are below
0.1. Under such conditions, gE(EOS), as determined from eqn. (5), will be
insensitive to the actual pressure.
It is therefore reasonable to base our match on eqn. (6) with the additional specification of P = 0, i.e.
f(u)-cue(u)-l-ln(u)=gE*/RT+XiZiln[bf;,,(T)/RT]

(7)

where fiO is fy(T, P = 0) (liquid).

At zero pressure, the pressure equation (3) yields
Pb/RT

= 0 = -f,

+ ae, + l/u

(8)

Define s by:
s = gE*/RT + Cizi ln[ b&( T)/RT]

(9)

Then eqns. (7) and (8) become

f(u)-ae(z4)-l-ln(u)=s

(10)

and
-f,(u)+ae,(z4)+1/24=0

(11)
These equations establish a relation between the variables (Y,s and U. For
the cubic equations of state given by eqn. (3),
f=ln[u/(u-l)],

f,=l/u-l/(24-1)

(12)

and

e = ln[b -t 4/b + cdl/(c2- CA

e,= --l/[(u+cJ~+c~)]

03)

50
16

-3

Plot
for

of

the

vs.

SRK-

equation

8-

41
-10

-6

-8

-4

-2

Fig. 1. Plot of a vs. s for the SRK equation of state.

or, inserting these expressions in eqns. (10) and (11)

-~I(u-l)-aln[(u+~~)/(~+cr)]/(c~-cl)-l==

(14)

V(u - 1) - a/[( U + c*)( U + CJ] = 0

(15)

A cross-plot of a vs. s, a = h(s), can be obtained by specifying the value of

U, calculating the corresponding value of a from eqn. (ll), and subsequently
s from eqn. (10). Such a plot is shown for the SRK equation in Fig. 1.
We observe that a is not a single-valued function of s, and that a is
undefined when s exceeds a limiting value, s = slim (= - 3.37 for the SRK
equation). The minimum value of a, ati (= 4.82), corresponds to s = slim.
The upper branch of the curve h(s) corresponds to the unstable root of the
pressure equation and is without interest.
From eqns. (10) and (11) the following general results can be derived:
da/ds

= -e-l

(16)

and
d2a/ds2 = - ee3e,/(da/du)

(17)

The first derivative, da/ds, is thus finite everywhere, whereas the second
derivative, d2a/ds2, tends towards negative infinity when s approaches slim.
The relation between a and s defines the mixing rule for our equation of
state. When the equation of state is applied to pure components, the

51

corresponding pure component parameters ai are known functions of temperature, usually in the form of a generalized correlation obtained by
matching the parameter to vapour pressure information.
If eqn. (9) is used for a pure phase, we obtain
si = s( zi = 1) = ln( bfi,/RT)

(18)

and eqn. (9) can be written

s = gE*/RT + Cizisi + Czi ln( b/b,)

09)

To apply the equation of state to a mixture (z, T, P) under conditions

where our gE correlation is applicable, we proceed as follows:
(i) Given aj, calculate the corresponding ui (the smaller root) from eqn.
(11). For cubic equations of state, this is a quadratic equation in u.
(ii) Substitute ui and (Yiinto eqn. (10) to obtain si.
(iii) Evaluate gE*(z, T) and s, using the si value from step (ii).
(iv) Determine the mixture a value corresponding to the resulting s.
At low temperatures and pressures the ai values will be larger than a,.,_+,,and
the resulting s value will be well below s,.
The equation of state will
therefore exactly match the gE model at zero pressure.
EXTRAPOLATION

If we try to apply the procedure described above at higher temperatures,

the following problems may arise:
(a) the pure component ai falls below amin and does not allow solution of
eqn. (11);
(b) the mixture s exceeds s,, and does not enable determination of an a
value.
Under such conditions, attempts to match gE from an excess Gibbs energy
model are, however, meaningless, since the reference for gE, i.e. the pure
component liquid, may be non-existent. Clearly, any extrapolation procedure becomes highly empirical.
We can, however, formulate a set of requirements that any extrapolation
procedure must satisfy in order to be acceptable.
(9 It must blend continuously with the well-defined low temperature
calculation. As a minimum requirement, a and its first and second
temperature and composition derivatives must be continuous everywhere.
(ii) For pure phases, the procedure must recover the proper pure component parameter.
(iii) Computationally, as well as conceptually, the procedure should be as
simple as possible.

52

An extrapolation scheme satisfying these requirements

Replace (Y= h(s) by a = hE(s), where
h(s) fors<s_
US)

h (smax) + h( s,)(

is the following.

(with S,<.slim)
s - s,,)

+ :h( smax)( s - s,,,,)~

for s > s,,

(20)
Equation (20) establishes a one-to-one correspondence between CYand s,
permitting solution for (Y,given any s, and solution for s, given any CLIt is
important to realize that s,, should be chosen well below Slim, otherwise
the second derivative of (Ywith respect to s becomes large (cf. eqn. (17)),
which might well result in intrinsic instability of the phase for which
properties are evaluated.
For the SRK and Peng-Robinson
equations, the maximum u value
yielding liquid roots is about 2.41, and a value of s,, corresponding to
u = 1.4-1.8 appears reasonable, being at a fair distance from the limiting
value and at the same time well above the u values found under the
conditions applicable for the equilibrium data on which the gE correlations
are based.
The computational procedure at high temperature thus becomes completely identical to that used at lower temperatures. Equation (17) is used,
first for determining the pure component si (given the ai), and subsequently
for determining a from the mixture s. Note that when si, as calculated in
this manner, exceeds s,_, the formal relation between si and fi,,, eqn. (18),
no longer holds. Near or above the pure component critical temperature it is
not even possible to calculate a liquid phase with the equation of state
applied to the pure component.
The present suggestion is identical to that of Mollerup (1986) and of
Heidemann and Kokal (1990) at low temperatures. Mollerup does not
attempt an extrapolation, whereas Heidemann and Kokal extend their
procedure to high temperature by redefining the reference states used in
their definition of the equation of state gE. The procedure of Heidemann
and Kokal, like the present, satisfies the requirements posed above for a
sound extrapolation method.
COMPUTATIONAL

ASPECTS

If s exceeds s_, the calculation of CYand its derivatives from eqn. (20) is
trivial. When s is below s,_, (Ycan be eliminated from eqns. (10) and (ll),
and the resulting equation is solved for u using Newtons method. Equation
(11) subsequently yields a, and the derivatives da/ds
and d*a/ds* are
found from eqns. (16) and (17).

53

The calculation of fugacity coefficients requires the evaluation of the

partial composition derivatives of a. These are determined as follows.
From
(a/&2,)( na) = (Y+ &/an,

= (Y+ (da/ds)(&/%z,)

(21)

we notice that the (Yderivatives can be found from the corresponding

derivatives. These, in turn, are now calculated.
Define S = ns; then

as/an, = s + n ib/an,

(22)

with

as/an, = In yi + si + m( bIbi) - (n/b) ab/ani

(23)

Back-substituting eqns. (23) and (22) into eqn. (21) yields the desired
derivatives of (Y.
Equation (23) can be used with any mixing rule for the b parameter. With
the conventional linear mixing rule, we obtain

as/an, = h yi + si + ln( b/hi) - hi/b + 1

(24

Evaluation of the temperature and composition derivatives of the fugacity

coefficients is equally straightforward. The calculation of these derivatives
requires evaluation of d20r/ds2 and of the temperature and composition
derivatives of the activity coefficients from the gE model.
The computation times for calculation of fugacity coefficients and the full
set of temperature, pressure and composition derivatives for the combination of the SRK equation of state and the Wilson gE model are shown in
Fig. 2. The computation times are compared with those used for the Wilson
equation alone (including derivatives). Analytical differentiation is used for
all derivatives. It should be noted that the overhead associated with
combining the Wilson model with an equation of state is only 25-50s for
mixtures containing more than five components.
Computations were made on an Acer 1100 personal computer (16 MHz
Intel 80386 + 80387) using the University of Salford FTN77/386 Fortran
compiler.
EXAMPLES
All examples are calculated by combining the Wilson gE model with the
SRK equation of state. The value of s,
in eqn. (20) was chosen as
s max= -3.855, corresponding to u/b = 1.60.
The first example illustrates the ability of the proposed method to
calculate phase envelopes, including critical points. The phase envelope for a

54

125

Computing

time

00

75

50

25

0
2

6
No.

of

10

12

14

components

Fig. 2. Computing time for evaluation of fugacity coefficients and their composition and
temperature derivatives, using the Wilson equation combined with the SRK equation of state,
compared with those for the Wilson equation only.

Phase
40

01
250

Envelope

c,

20%

c,

50%

300
Temperature,

--

Original

SRK

Wilson-

SRK

350

400
K

Fig. 3. Phase envelope for a mixture containing 20% ethane, 50% propane and 30% n-butane,
using the Wilson equation, combined with the SRK equation, and using the SRK equation
with conventional mixing rules.

55
125

Ethanol

- Water

0.

Wilson

0.2
Ethanol

0.4
mole

0.6

0.8

SRK

1.0

fractions

Fig. 4. Calculated (Wilson + SRK) phase diagrams for the ethanol-water

with experimental data.

system compared

containing
20% ethane, 50% propane and 30% butane is shown in
Fig. 3, (i) using the original SRK equation with the conventional quadratic
mixing rule for the a parameter and all interaction coefficients equal to
zero, and (ii) with the Wilson-SRK combination, where the Wilson volumes
are chosen proportional to the component b parameter, and the Wilson
interaction energies are all zero. The critical pressures calculated from the
two models are nearly identical and the critical temperatures differ by about
2 K.
The second example shows P-x-y diagrams for the mixture ethanol-water
at elevated temperatures (Fig. 4). Experimental data are from Barr-David
and Dodge (1959), and the Wilson parameters are those used by Gmehling
and Onken (1977a) for correlating the 760 mmHg data of Bloom et al.
(1961). Good agreement is observed at 200 C and at 220 C, and the
behaviour at 275 o C is qualitatively correct.
Figure 5 shows calculations for acetone-water mixtures at high temperature, together with data of Griswold and Wong (1952). The Wilson model
parameters are taken from Gmehling and Onken (1977a), based on the data
of Verhoye and Schepper (1973) at 760 mmHg. Again, qualitatively correct
behaviour is observed.
Model calculations for the methanol-benzene
system (data of Butcher
and Medani, 1968) are shown in Fig. 6, using Wilson parameters from
Gmehling and Onken (1977b), based on the 740 mmHg data of Strubl et al.
mixture

56

Acetone
Wilson

0.0

Water

SRK

0.4

0.2
Acetone

0.6
mole

0.8

1.0

fraction

Fig. 5. Calculated (Wilson+SRK)

with experimental data.

phase diagrams for the acetone-water

system compared

(1972). For this system, the quality of the predictions deteriorates at higher
temperatures, yielding separated branches of the P-x-y curve at 220 o C.
Finally, activity coefficients at saturation conditions and atmospheric

Methanol

Wilson

0.0

0.2
Methanol

Benzene
-SRK

0.4

0.6
mole

0.8

1.0

fraction

Fig. 6. Calculated (Wilson + SRK) phase diagrams for the methanol-benzene

pared with experimental data.

system com-

57

pressure were calculated for the systems shown in Figs. 4, 5 and 6, using the
equation of state. As expected, these activity coefficients are virtually
identical to the corresponding activity coefficients calculated from the
Wilson equation, deviations being everywhere below 0.1%.
CONCLUSION

A formalism for incorporating low pressure gE information into an

equation of state has been developed. The resulting model accurately reproduces the low temperature behaviour of the gE model and enables smooth
extrapolation to higher temperatures without compromising consistency,
simplicity and computational convenience.
It would be inappropriate to associate the quality of the high temperature
results observed in the present examples with predictive properties of the
model. First, the Wilson parameters used in the calculation are based on
measurements over a very narrow temperature range and, second, the
extrapolation procedure has to be empirical. The essential aim of the present
development is to provide a means for incorporating an excess Gibbs energy
model within the framework of an equation of state in a reasonable
manner, satisfying conditions of mathematical consistency, continuity,
proper representation of pure components and computational convenience.
The examples serve to demonstrate that this aim can be accomplished.
ACKNOWLEDGEMENT

The author wishes to thank Jorgen Mollerup, Instituttet

for many helpful suggestions and discussions.
NOMENCLATURE

equation of state mixture parameter

res molar residual Hehnholtz energy
b
equation of state mixture parameter
pure component value of b parameter
bi
cl, c2 constants in equation of state
volume function in equation of state, eqn. (1)
volume function in equation of state, eqn. (2)
;
f;
pure component fugacity, component i
fro
pure component fugacity, component i, zero pressure
excess Gibbs energy, equation of state
gE
E* excess Gibbs energy from gE model
:
function relating s and (Y

for Kemiteknik,

58
h,

i
f,
R
s
si
S

T
u
V
=i

function defined by eqn. (20)

component index
total number of moles in mixture
pressure
gas constant
defined in eqn. (9)
s value, component i, eqn. (18)
ns
temperature
reduced volume, u = v/b
molar volume
mole fraction, component i

Greek letters
a
Yi

equation of state mixture parameter

activity coefficient, component i

REFERENCES
Barr-David, F. and Dodge, B.F., 1959. Vapor-liquid
equilibrium at high pressure-the
systems ethanol-water and 2-propanol-water. J. Chem. Eng. Data, 4: 107-121.
Butcher, K.L. and Medani, M.S., 1968. Thermodynamic properties of methanol-benzene
mixtures at elevated temperatures. J. Appl. Chem., 18: 100.
Gani, R., Tzouvaras, N., Rasmussen, P. and Fredenslund, Aa., 1989. Prediction of gas
solubility and VLE by group contribution. Fluid Phase Equilibria, 47: 133-152.
Gmehling, J. and Onken, U., 1977a. DECHEMA Chemistry Data Series, Vol. I, Part 1,
Aqueous-Organic Systems. DECHEMA, Frankfurt am Main.
Gmehling, J. and Onken, U., 1977b. DECHEMA Chemistry Data Series, Vol. I, Part 2a,
Organic Hydroxy Compounds: Alcohols. DECHEMA, Frankfurt am Main.
Griswold, J. and Wong, S.Y., 1952. Phase equilibria of the acetone-methanol-water
system
from 100 o C into the critical region. Chem. Eng. Prog. Symp. Ser., 48(3): 18-34.
Guptt, P.A., Rasmussen, P. and Fredenslund, Aa., 1986. A new group-contribution
equation
of state for vapour-liquid equilibria. Ind. Eng. Chem. Fundam., 25: 636-645.
Heidemann, R.A. and Kokal, S.L., 1990. Combined excess free energy models and equations
of state. 4th CODATA Symposium, September 4-6, 1989. Gradisca dIsonzo, Italy. Fluid
Phase Equilibria, 56: 17-37.
Huron, M.J. and Vidal, J., 1979. New mixing rules in simple equations of state for
representing vapor-liquid
equilibria of strongly non-ideal mixtures. Fluid Phase Equilibria, 3: 255-271.
Mollerup, J., 1986. A note on the derivation of mixing rules from excess Gibbs energy
models. Fluid Phase Equilibria, 25: 323-326.
Peng, D.Y. and Robinson, D.B., 1976. A new two-constant equation of state. Ind. Eng.
Chem. Fundam., 15: 59-64.
Soave, G., 1972. Equilibrium constants from a modified Redlich-Kwong equation of state.
Chem. Eng. Sci., 27: 1197-1903.