Gibbs Energy Analysis of Phase Equilibria

Lee E. Baker, SPE. Amoco Production Co.

Alan C. Pierce, SPE, Amoco Production Co.
Kraemer D. Luks, SPE, U. of Tulsa

Abstract
Equations of state are used to predict or to match
equilibrium fluid phase behavior for systems as diverse
as distillation columns and miscible gas floods of oil
reservoirs. The success of such simulations depends on
correct predictions of the number and the compositions
of phases present at a given temperature, pressure, and
overall fluid composition. For example, recent research
has shown that three or more phases may exist in
equilibrium in CO 2 floods.
This paper shows why an equation of state can predict
the incorrect number of phases or incorrect phase compositions. The incorrect phase descriptions still satisfy
the usual restrictions on equality of chemical potentials
of components in each phase and conservation of moles
in the system. A new method and its mathematical proof
are presented for determining when a phase equilibrium
solution is incorrect.
Examples of instances where incorrect predictions
may be made are described. These include a binary
system in which a two-phase solution may be predicted
for a single-phase fluid and a multi component
CO 2 /reservoir oil system in which three or more phases
may coexist.

Introduction
Advances in reservoir oil recovery methods have
necessitated advances in methods for prediction of phase
equilibria associated with those methods. It was long
considered sufficient to approximate the reservoir
behavior of oil and gas systems with models in which
compositions of the phases in equilibrium were unimportant. In such a model, the amounts and properties of the
phases are dependent on pressure and temperature only.
Later, experience in production from condensate and
volatile oil reservoirs showed that models incorporating
compositional effects were required to simulate the
phase equilibria adequately. This led to the use of con0197-7520/82/0010-9806$00.25
Copyright 1982 Society of Petroleum Engineers of AIME

OCTOBER 1982

vergence pressure correlations and subsequently to the

development of more sophisticated equation of state
methods for modeling and predicting phase equilibria.
For adequate description of the compositional effects
that occur in enhanced oil recovery processes such as
CO 2 and rich gas flooding, an equation-of-state approach is a virtual necessity.
The use of equations of state for phase equilibrium
prediction is not limited to the petroleum industry. Such
equations also find wide use in basic chemical and
physical research, and in the refining and chemical processing industries.
Solution techniques for phase equilibrium problems
are varied and depend to some extent on the application
and equation of state used; however, there are three
restrictions that all phase equilibrium solutions must
satisfy.
First, material balance must be preserved. Second, for
phases in equilibrium there must be no driving force to
cause a net movement of any component from one phase
to any other phase. In thermodynamic parlance, the
chemical potentials for each component must be the
same in all phases. Third, the system of predicted phases
at the equilibrium state must have the lowest possible
Gibbs energy at the system temperature and pressure.
The requirement that the Gibbs energy of a system, at
a given temperature and pressure, must be a minimum is
a statement of the second law of thermodynamics,
equivalent to the more common version requiring the entropy of an isolated system to be a maximum. The
equivalence is demonstrated formally in Ref. I, for example. If the Gibbs energy of a predicted equilibrium
state is greater than that of another state that also satisfies
Requirements I and 2, the state with the greater Gibbs
energy is not thermodynamically stable.
Requirements I and 2, material balance and equality
of chemical potentials, are used commonly as the sole
criteria for solution of phase equilibrium problems. In
many cases, there is no problem with this usage;
731

PB
91

P2

4
a..

9z

P 3fl

""

PI
PA
0

XB--

Fig. 1-Pressure/composition diagram for AS system.

however, there are important instances when the Gibbs

energy minimization requirement also must be considered. As discussed later, equality of chemical potentials is necessary but not sufficient for minimization of
the Gibbs energy.
The iterative solution techniques usually used in finding equilibrium states may lead to a trivial situation (all
phases present having the same properties) that always
will satisfy the first two requirements but not necessarily
the third. Of more importance perhaps, particularly in
enhanced oil recovery, is the possibility of predicting
false two-phase states in or near a three-phase region.
These states can satisfy material balance and chemical
potential restrictions while failing to minimize the Gibbs
energy. This leads to predictions of incorrect phase
volumes and properties (density, viscosity, etc.), which
can adversely affect the results of an oil recovery
simulation.
Several authors 2-7 have reported the occurrence of
multiple (three or more) phases in rich gas/oil and
CO 2 /oil reservoir systems. Others S- IO have shown that
when there are multiple phases, it is possible to predict
equilibria at states that do not minimize the Gibbs energy
of a system.
This paper describes such failures in solving the phase
equilibria problem in terms of Gibbs energy analysis. It
also presents a method, and a mathematical proof supportive of the method, for identifying false solutions.
Examples of the technique are given. For an elegant
treatment of some properties of a Gibbs energy surface,
the reader is directed to Coleman, II Dunn and
Fosdick, 12 and references in those papers.

Discussion
Equations of state often are used to predict phase
equilibrium solutions. Gibbs energy analysis is the use of
an equation of state to calculate a Gibbs energy surface
and to determine whether a predicted equilibrium state
has the lowest possible Gibbs energy. We first illustrate
the use of a Gibbs energy surface to determine phase
equilibria and show how false solutions can be obtained.
732

I
12
1---

I
I

I
I
I
I
I

I
I

I
I

o XLl

FHO
XB--

Xv

Fig. 2-Gibbs energy diagram for A-S system at p 1 .

Then we describe a method for detecting false solutions

and illustrate why it succeeds when other methods fail.
Finally, we give specific examples of the utility of the
method, using a version of the Redlich-Kwong equation
of state.

Illustration of the Problem

Fig. 1 shows the pressure/composition diagram for a
system composed of hypothetical Components A and B
at a fixed temperature. The system exhibits two-phase
(L I + V) behavior at low pressures; a three-phase region
(L I + L2 + V) limited to a single pressure, P 3, for a
binary system; and regions of L I + L2 and L2 + V twophase equilibrium at pressures greater than P 3' In some
pressure and composition ranges, the system is singlephase L I , L 2 , or V.
The purpose of using an equation of state is to predict
accurately the phase behavior of such a system. The first
step in solving a phase equilibria problem can be described mathematically as finding a plane tangent to the
Gibbs energy surface (g surface), with material balance
restrictions. (See Appendix for a more detailed description.) The slope of the tangent plane corresponds to the
component chemical potentials (related to the
fugacities). The points oftangency of the plane and the g
surface correspond to the compositions of the predicted
equilibrium phases. For binary systems, as illustrated
here, the g surface is a curve and the tangent plane is a
straight line. For multicomponent systems, the g surface
is a hypersurface and the tangent plane is a hyperplane.
The material balance restrictions require the overall composition of a multiphase system to lie within the region
bounded by the points of tangency of the g surface and
the tangent plane.
These concepts are illustrated in some detail in Figs. 2
through 5, which represent the Gibbs energy surface
(curve) for the A-B system at increasing pressures.
Fig. 2 shows the Gibbs energy diagram (Gibbs energy
vs. composition) for the A-B system at pressure
PI <P3' The Gibbs energy surface (curve) is calculated
SOCIETY OF PETROLEUM ENGINEERS JOURNAL

t
""

91

~-

93

-J.--I
I
I

I
I

I
I
I
I
I
I

0 Xu

FEED

XL2

Xv

XB-Fig. 3-Gibbs energy diagram for A-8 system at p', .

for a hypothetical homogeneous (single) phase at all

compositions and does not represent the actual Gibbs
energy except in the single-phase region and at the
equilibrium phase compositions. The g surface is concave upward except in the two-phase composition
region. A tangent to the g surface indicates the
equilibrium liquid and vapor compositions (points of
tangency). For a given feed composition (as indicated in
Fig. 2), the liquid- and vapor-phase fractions can be
determined by application of the lever rule. The Gibbs
energy, g 2, is the molar sum of the phase Gibbs energies
(g's at the point of tangency) and'lies on the tangent line
at the feed composition. The two-phase Gibbs energy,
g 2, is less than the Gibbs energy, g 1 , of the hypothetical
homogeneous phase of this composition, indicating that
the two-phase system is more stable than a single phase.
For all feed compositions outside of the two-phase
region, a tangent to the g surface does not intersect the
surface at any other point; however, for any composition
within the two-phase region, a tangent to the
hypothetical g surface at that point does intersect (and lie
above) the g surface at some point in the overall composition range. In the next section and in the Appendix
we show that this is a necessary and sufficient condition
for stability of a predicted phase equilibrium solution.
For a stable solution, the tangent corresponding to the
solution does not lie above the g surface at any
composition.
Fig. 3 shows the Gibbs energy surface for A-B at a
higher pressure, P'I , which is still less than P 3</>. In this
case, the g surface resembles that at PI, but an additional
lobe (the incipient L2 phase) has appeared at a composition between that of the liquid and vapor phases. A
tangent to the g surface locates equilibrium phase compositions xLI and Xv for any feed composition in the
two-phase region. An additional tangent to the g surface
at XLI 'and xv' locates a false solution. For the feed indicated, either one of the tangent lines corresponds to a
OCTOBER 1982

Fig. 4-Gibbs energy diagram for A-8 system at p 3<1>'

pair of phases satisfying material balance and equality of

chemical potentials, but only the lower tangent gives a
minimum Gibbs energy (g 2 < g 2 ' < g 1 ). Consistent with
our earlier observation, only the lower tangent does not
lie above the g surface at any point.
Fig. 4 shows the g surface at the three-phase pressure,
P 3</> The phase equilibrium solution for the multiphase
region (XLI to xv) is given by a tangent at XLI, XL2, and
Xv. As in Figs. 2 and 3, in the composition range XLI to
Xv the upper curve is the (hypothetical) Gibbs energy for
a homogeneous phase (g 1 at the indicated feed), while
the actual Gibbs energy g3 is a mole-weighted sum of
the phase molar Gibbs energies at the points of tangency,
XLI,xL2, andx v
The relative amounts of phases L I, L 2 , and V are not
fixed by the compositional lever rule in this binary case,
but they are bounded in that as pressure is increased, L2
is formed from L 1 and V until one or the other of these
phases is consumed.
Fig. 5 presents the Gibbs energy surface atp2 >P3</>'
There are two two-phase regions: XLI -x L2, and X L2' -Xv
For the indicated feed, three possible states satisfying
material balance and equality of chemical potentials are
indicated. Only one of these (xLI-XL2, Tangent Line 2)
has the minimum Gibbs energy, The solutions corresponding to 2' and 2" both have Gibbs energies less
than g 1 of the hypothetical single phase but greater than
the Gibbs energy of the true solution.
A feed with composition between XL2 and xL2' will be
single-phase at the temperature and pressure specified.
However, Tangent Line 2' corresponds to a two-phase
(false) solution for a feed in this composition range, This
is an exception to the heuristic rule of thumb that, for a
given feed, the more phases present at equilibrium, the
lower the Gibbs energy will be. 13
An observation may be made here concerning the utility of second derivative tests 9 for determining stability of
a phase equilibrium solution. Such a test indicates the
733

~.-----.------.-----.------r-----.

6000
L

""

is:! 4000
::>

Vl
Vl

.....
Q::
0..

2000

OL-____

____

____

______L -____

Fig. 5-Gibbs energy diagram for A-B system at P2'

Fig. 6-Pressure/composition diagram for C0 2 /toluene at

38.1C.

direction of curvature of the g surface. For a stable solution, the g surface must be concave upward at the points
of tangency. Downward curvature indicates definite instability. The solution (tangent) labeled as 2" is unstable
by this criterion because the g surface at the point of
tangency near Xv is concave downward. The g surface is
concave upward for both tangency points of Solution 2',
however, and this solution would be considered stable by
the second-derivative criterion.
At pressure P3 shown in Fig. I, the lobe of the g surface corresponding to the V phase will have vanished.
The Gibbs energy curve will resemble that of Fig. 2,
with L I and L2 phases forming from a feed.
We have shown schematically, and for very simple
systems, that false solutions (not corresponding to the
minimum Gibbs energy) to phase equilibrium problems
can be generated for compositions in and around
multi phase regions. The nature of the g surface and

associated tangents must be examined concurrently with

the solution of mass flow equilibria equations (material
balance and equality of chemical potentials) to ensure
that the minimum Gibbs energy has been found.
We also have indicated that the tangent corresponding
to an equilibrium solution qnnot lie above the Gibbs
energy surface for any composition. This is described in
more detail and extended to multicomponent systems in
the next section and in the Appendix.

The Gibbs Energy Analysis

The general phase equilibria problem for a feed of I components consists of finding a stationary state that is an
equilibrium state. A stationary state corresponds to an
extremum or saddle point in the total Gibbs energy of a
system (Appendix, Definition 2), where the system is a
set of phase compositions satisfying the mass balance requirement. An equilibrium state is a stationary state cor-

TABLE 1-EQUATION OF STATE AND PARAMETERS

nRT

An2

V -nB

T'h V(V +nB)

P=-----

Fluid Properties

Pc

(K)

(kPa)

(0C)

na

nb

CO 2

304.16

7398.1

toluene
n-decane
iodobenzene

591.77
617.55
721.11

4067.9
2096.0
4529.9

-25.0
35.0
38.1
38.1
-25.0
35.0

0.42564
0.43776
0.43776
0.44485
0.46432
0.41931

0.08003
0.08886
0.08886
0.07986
0.07042
0.08213

Component
I

A=

L; L;
1= 1

j= 1

B=

L;
i= 1

X,XJCl'.;j'

x/3;.

Tc

Interaction Parameters, C Ii
C0 2 /toluene
C0 2 /n-decane
CO 2liodobenzene

734

0.080
0.095
0.1365

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

2000 , - - - - - , - - - - - - - , - - - - y - - - , - - - - - - - ,

1750 r------r---..,.,.-----,-----r---,

1500

1700

:::; 1000

~ 1650

"0
E

>-<.:>

a:::

... -----

UJ

UJ

500

I
1
1 FALSE
1- L1 +L2 ------..,~;

Vl
<0
<0

<.:>

o
--

--

---

--I

I
1

-----+1- L1 +V-------t---

L+V

-500L-_ _ _ _ _ _ _ _ _ _ _
O. 2
O. 4
O. 6
O. 8
o
MOLE FRACTI ON CO
2
~

1550

V/

1500L-~_~_ _ _-u_ _ _~_~L_~

D. 6

D. 7

D. 8

D. 9

MOLE FRACTION CO

Fig. 7-Gibbs energy of mixing of CO 2 /toluene at 38.1C,

1379 kPa.

Fig. 8-Pressure/composition diagram for CO 2 /n-decane at

-25C.

responding to the global minimum in the total Gibbs

energy of a system. The mathematical criteria for determining the equilibrium state are developed in detail in
the Appendix. Briefly, it is shown that the equations of
mass flow equilibria are satisfied at a stationary state
(Theorem I). Solution of these equations-i.e., finding
the compositions of the coexisting phases-is equivalent
to finding the hyperplane that is tangent to the Gibbs
energy surface at the J points denoting the composition
of the phases in equilibria (Theorem 2). However, this
condition of tangency does not guarantee that the stationary state is an equilibrium state.
Theorems 3 and 4, with their associated lemmata,
show that the equilibrium state-i.e., the state in which
the Gibbs energy is a global minimum under the conservation of mass constraints imposed by the feed-is the
condition in which the Gibbs energy surface is
everywhere either on (tangent to) or above the tangent
plane.
The implication is, once the mass flow equilibrium
problem for I components is solved and the J phases are
identified, that one must verify that this (stationary) state
is the actual equilibrium state. This can be done by determining whether the tangent hyperplane lies above the
Gibbs surface at any composition. The tangent plane corresponding to the equilibrium state lies entirely below or
tangent to the Gibbs energy surface. The numerical application of this technique is not described here. In principle, any technique for finding roots of the equation
G( r) - L(T) = 0 < a should suffice to find points where
the Gibbs energy surface lies below the tangent plane
corresponding to a predicted phase equilibrium solution,
if any such points exist.

In the following examples, Gibbs energies of mixing

are calculated by using a version 14 of the RedlichKwong equation of state. Other equations of state will
produce similar results. The equations of state, mixing
rules, and the component parameters used for the binary
examples to follow are given in Table 1.
Calculation of the Gibbs energy for a system is based
on the relation

Examples of Gibbs Energy Analysis

For ease of illustration, we chose binary systems tc
demonstrate the range of possibilities of phase equilibria
solutions. As shown later in this section, the problem of
false phase equilibrium solutions is encountered also in
more complex systems (such as CO 2 ireservoir oil). The
principles described by the theorems in the Appendix apply equally well to these more complex mixtures.
OCTOBER 1982

G=pV+

I
J(lO (nRT)
p - - - dV-RT ~

+ ~

n;(U;'-TSn.

1=1

(V)

n;ln - . -

nlRT

. .................. (1)

;=1

Insertion of the equation of state to be used into Eq. I

(as illustrated by Baker and Luks 15 or by Peng and
Robinson 16) leads to an expression that can be evaluated
at the pressure, temperature, and composition of a given
fluid phase.
Throughout this discussion, we refer to the Gibbs
energy of mixing, which is the phase Gibbs energy less
the mole-weighted sum of the pure component Gibbs
energies. This treatment makes the choice of the pure
component standard states immaterial. For convenience,
we may choose U? and S? to be zero for all components.

Example I: CO 2 /Toluene System at IOO.6F

(38.I O C). The pressure/composition diagram 17 for
CO 2 /toluene is shown in Fig. 6; the Gibbs energy of
mixing at 200 psia (1379 kPa) (noted by the dashed line)
is plotted as a function of CO 2 concentration in Fig. 7.
In this case, there are no three-phase phenomena. The
two-phase region at 200 psia (1379 kPa) is spanned by
the tangent to the Gibbs energy curve, namely, the range
of 12 to 99% CO 2 , The regions outside this two-phase
region are one-phase. In the region of 65 to 87 % CO 2 ,
there is more than one possible Gibbs energy at each
composition because there are three real compressibility
solutions for the equation of state. In calculating a Gibbs
735

0.2
0.4
0.6
0.8
MOU FRACTIOO CO
2

Fig. 9-Gibbs energy of mixing of CO 2 In-decane at - 25C,

1610.6 kPa.

Fig. 11-Gibbs energy of mixing of CO 2/n-decane at - 25C,

1620.3 kPa.

- JIm !:--:'-=---::'-:_-:!-:_::'-::-----:
o 0.2 0.4
0.6
0.8
MOU FRACTIOO CO

Fig. 10-Gibbs energy of mixing of C0 2 /n-decane at - 25C,

1613.4 kPa.

Fig. 12-Gibbs energy of mixing of CO 2 In-decane at - 25C,

1723.7 kPa.

energy curve, where there are multiple real compressibility solutions of the equation of state, only the
solution giving the lowest Gibbs energy should be considered, and that is the one shown in Fig. 7. We emphasize again that the solid Gibbs energy curves plotted
in the two-phase region (12 to 99% CO 2 ) do not represent the equilibrium Gibbs energy of the system; the
dashed line tangent to the Gibbs energy surface
represents the true equilibrium Gibbs energy in the twophase region. The solid curves are the Gibbs energies
calculated as if the system were a single homogeneous
phase at the pressure, temperature, and composition
specified.

possible false solutions, an L, ' + L 2 ' and an L2 " + V

solution, as evidenced by the dashed lines tangent to the
Gibbs energy surface at L, ' and L 2 ' and at L 2 " and V.
These solutions are shown in Fig. 8 as dashed line extensions (into the L, + V region) of the L, + L2 and L2 + V
solutions that exist at pressures greater than P 3</>
For each of the possible solutions shown in Fig. 9, the
Gibbs energy curve is concave upward at the points of
tangency of the solution lines. Thus, tests of the second
derivatives of G, (a2Glanianj), as recommended by
Sorensen et al., '0 could not distinguish the true solution
from the false solutions. However, for each of the false
solutions, the tangent line (solution line) lies above the
Gibbs energy surface at some point. The tangent line for
the true solution does not lie above the Gibbs energy
curve at any point, in accordance with the theorems
proved in the Appendix.
Fig. 10 shows the Gibbs energy curve at the threephase pressure, p=234 psia (1613.4 kPa). In this case, it
would be possible to determine L, + L 2 , L2 + V,
L, +V, or L, +L2 +V solutions; however, the twophase solutions simply correspond to the case of absence
of a third phase, which (if it existed) would be in
equilibrium with the other two. This may occur physicalIy. Depending on factors such as the system enthalpy,
there may be L, +L 2 , L, +V, L2 +V, <1r L, +L2 +V.
An increase in system enthalpy will drive the equilibrium
to L, + L2 or L2 + V, depending on the composition,
while a decrease in enthalpy will drive the equilibrium to

Example 2: CO 2 /n-decane at -l3F (-25C). The

pressure/composition diagram for this system is shown
in Fig. 8. Both CO 2 and decane are subcritical in this
case. In the pressure range considered, the system exhibits one-phase, two-phase (L, + V, L, + L 2 , L2 + V)
and three-phase (L, + L2 + V) behavior. The three-phase
region for a binary is limited to a single pressure (at fixed
temperature), in this case 234 psia (1613.4 kPa). The
solid lines are the equation of state predictions, and they
closely mimic experimental data. '8 Gibbs energy of
mixing predictions are shown in Figs. 9 through 12. In
Fig. 9, atp=233.6 psia (1610.6 kPa) (P<P3</, the twophase liquid and vapor equilibrium compositions are
shown as L, and V and are connected by a line tangent
to the Gibbs energy surface. In this case, there are two
736

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

L I + V. The case is analogous to that of a pure liquid at

its boiling point, where liquid and vapor may coexist in
any proportion.
Fig. 11 shows the calculated Gibbs energy curve for
p=235 psia (1620.3 kPa). There is an LI +L2 region,
an L 2 ' + V region, and an intervening single phase (L 2 )
region (as seen in Fig. 8). The tangent lines for the correct solutions are shown. Also shown is a false L I + V
prediction obtained for a feed composition in the singlephase region. At the points of tangency of this unstable
solution, the Gibbs energy curve is concave upward; second derivative tests could not determine stability of the
solution. However, the tangent line intersects a lobe of
the Gibbs energy curve, as theory predicts it should for
an unstable solution. This is also a prime example contradicting the heuristic rule proposed by Gautam and
Seider 13 that a P-phase solution, if it exists, has a lower
Gibbs energy than a P-I phase solution.
Fig. 12 presents the Gibbs energy curve for CO 2 /ndecane atp=250 psia (1723.7 kPa) (p> >P3q,). In this
case, the vapor lobe of the curve has disappeared, leaving only an L I + L2 two-phase region and L I and L2
single-phase regions.
Example 3: CO 2 /Iodobenzene at 95F (35C). The
CO 2 /iodobenzene system is interesting because it
demonstrates the possibility of falsely predicting a
critical point. Experimental data on this system are not
available, so the specific description of the phase
behavior may be regarded as somewhat conjectural.
However, we also have observed the illustrated
phenomenon in other, more complex, systems in which
the presence of a predicted critical point was not supported by experimental data.
The predicted pressure/composition diagram for
CO 2 /iodobenzene at 95F (35C) is shown in Fig. 13.
There is a liquid/vapor (L I + V) region, a liquid/liquid
(L I + L 2 ) region, a liquid/vapor (L2 + V) region bounded above by a critical point, and a three-phase pressure.
At a slightly higher temperature [98.4F (36.9C)], the
L2 + V region disappears, with a two-phase K-point
(L I + L2 = V critical fluid) solution.
The most interesting aspect of this diagram is the
critical point solution indicated at 96.8% CO 2 , 993.1
psia (6847.2 kPa).
The Gibbs energy curve in the region of this critical
point is shown in Fig. 14. The mathematical criteria for a
binary system critical point,

10000

9500

900l L1

T,p.n2

(:~~)

~-

L1 + L2

8500
CRITICAL POINT-

L1+~+V,\

6500
6000

iii

~L1+V

5500
0.4

0.5

0.6

0.7

MOLE FRACTION CO

0.9

0.8
2

Fig. 13-Pressure!composition diagram for CO 2!iodobenzene at 35C.

FALSE CRITICAL

o __f!'..L..S._~~~!!~~~_~'?'!l!_!~~!-- ___P_~I~~_-

-~~~~~~~--~~

0.2

0.4
0.6
MOLE FRACTION CO

0.8

Fig. 14-Gibbs energy of mixing of CO 2!iodobenzene at

35C, 6847.2 kPa.

26000
24500

21000
;;:, 17500

l5!' 14000

L1

:::>

T.p.n2

=0, ......... (2)

'"
~

10500

UPPER

c..
4

are satisfied. However, the inequality (a G/an I 4h.P.n,

> 0 that ensures stability is not satisfied. Consequently:
the dashed line tangent to the curve at the predicted
critical point lies above the curve almost everywhere outside the immediate vicinity of the point. The curvature of
the Gibbs energy surface at this point is caused by the incipient L2 lobe, which grows downward as pressure is
increased. The true liquid/vapor phase equilibrium is indicated by the dashed line tangent to the g surface at L I
and V of Fig. 14.
OCTOBER 1982

FALSE CRITICAL POINT .--/"

"'"

(::~)

BO~D

OF L + V SOLUTIONS
J

LJ + ~ + V
LOWER BO~D OF L
J

3500

SOLUTIONS

O~~L-~L-~__~__~

0.2

0.4

0.6

MOLE FRACTION CO

O. 8
2

Fig. 1S-C0 2!Leveliand oil phase behavior at 41.1 C.

737

Similar false critical points for this system can be

calculated (with the equation of state and parameters
shown in Table I) over a temperature range of 70 to
116F (21 to 47C).

Example 4: CO 2 /Levelland Oil at 106F (41.1C).

The pressure/composition diagram for the CO 2 /
Levelland oil system 7 is shown in Fig. 15. We cannot
represent rigorously either the pressure/composition surface or the Gibbs energy surface for this system because
such a representation would require N dimensions,
where N is the number of components in the system.
However, the regions of multiphase (L I + V, L I + L 2 ,
L I + L2 + V) solutions are shown, with an indication of
the regions where unstable solutions also may be found.
In general, with complex systems such as these, L I + V
solutions may be obtained not only in the actual L I + V
region but also in the L I + L2 region and the three-phase
(L I + L z + V) region. Similarly, L I + L z solutions also
may be found in all three regions. Three-phase solutions
are found only in the region marked L I + L z + V. Gibbs
energy analysis is as successful in detecting false solutions for this system as it is in the binaries of Examples I
through 3.
It is usually possible to obtain L I + V solutions in the
L I + L z region for only a (relatively) small pressure interval above the three-phase region, while L I + L2 solutions often may be found in the L I + V region at
pressures far below the three-phase region. This results
from the variation in shape of the Gibbs energy curve,
with an L2 lobe persisting at quite low pressures, while
the vapor lobe disappears rapidly with pressure increasing above the three-phase pressure. This is not surprising, because the properties of a liquid phase are more insensitive to pressure variation than the properties of a
vapor phase.
Since false phase equilibrium predictions can be obtained over a wide range of pressure and composition, it
is important to confirm that the correct solution is found.
For the CO 2 /Levelland oil system, the properties (density, viscosity) of the L I liquid phase are similar in the
L I + V and the L I + L2 regions. However, the properties of the vapor (V) and the L2 liquid phases are strikingly different. For example, the L z phase density is
commonly two or more times the density of the vapor
phase. The use of a false L I + V solution for a fluid in
the L I + L2 region of the phase diagram could lead to a
much different performance in a reservoir model of a
COz flood. Similarly, the third phase present in the
L I + L2 + V region could have a marked effect on the
fluid flow behavior and relative permeabilities. Unless
the model recognizes the possibility of a third phase, a
two-phase (L I + V or L 1 + L 2 ) equilibrium would be
predicted, with possibly adverse consequences.

Conclusions
For fluid systems that exhibit multiple phases, an equation of state may predict false phase equilibrium
solutions.
This paper presents a self-consistent method for determining whether a predicted equilibrium state is false.
The method makes use of the equation of state to
calculate the Gibbs energy surface and the tangent plane
corresponding to the predicted equilibrium solution. If
738

the tangent plane lies above the Gibbs energy surface at

any point, the predicted equilibrium solution is false.
Conversely, if the plane lies entirely below or tangent to
the Gibbs energy surface, the solution does describe the
equilibrium state.

Nomenclature
A = temperature-, pressure,- and compositiondependent phase parameter (Table I)
B = temperature-, pressure-, and compositiondependent phase parameter (Table I)
C ij = interaction parameter for component i with
component j (Table I)
D = G- L, the difference between the Gibbs
energy surface and a tangent plane
(function of composition)
F = function defined in Lemma 2 (Appendix)
G = Gibbs energy surface
g = molar Gibbs energy
I = number of components
J, K = number of phases

= hypothetical phase vector

L = liquid phase (subscripted)
L = plane tangent to Gibbs energy surface, G
m ij = composition variable, composition of
component i in phase j
mj = vector (mlj' m2j . .. mlj), composition of
phase j
m= state, set of phase composition vectors mj
n = number of moles
= state, set of phase composition vectors
p = pressure, psia (kPa)
r = phase composition vector (r I , r 2 . . r I)
R = gas law constant
S7 = standard state molar entropy of component i
T = temperature
U7 = standard state molar internal energy of
component i
V = vapor
x = mole fraction
IX ij = temperature-dependent component
parameter
{3 i = temperature-dependent component
parameter
o = negative number
Ej = positive number
T = variable parameterizing a curve in composition space
11- = chemical potential
= variable defined in Lemma 2 (Appendix)
na, nb = generalized component parameters, function of the reduced temperature and the
component Pitzer acentric factor

Subscripts
c = critical
i = component i
j = component j (Table I); phase j elsewhere
SOCIETY OF PETROLEUM ENGINEERS JOURNAL

phase k
total
v = component v
cf>
phase
1,2,3
phase identification

References
I. Callen, H.B.: Thermodynamics, John Wiley and Sons Inc., New
York City (1960).
2. Huang, E.T.S. and Tracht, J.H.: "The Displacement of Residual
Oil by Carbon Dioxide," paper SPE 4735 presented at the SPE
Third Symposium on Improved Oil Recovery, Tulsa, April 22-24,
1974.
3. Shelton, J.L. and Yarborough, L.: "Multiple Phase Behavior in
Porous Media During CO z or Rich-Gas Flooding," J. Pet. Tech.
(Sept. 1977) 1171-78.
4. Gardner, J.W., Orr, F.M., and Patel, P.D.: "The Effect of Phase
Behavior on CO) Flood Displacement Efficiency," J. Pet. Tech.
(Nov. 1981) 206-7-81.
5. Henry, R.L. and Metcalfe, R.S.: "Multiple Phase Generation
During CO 2 Flooding," paper SPE 8812, presented at the First
Joint SPE/DOE Symposium on Enhanced Oil Recovery, Tulsa,
April 20-23, 1980.
6. Orr, F.M., Yu, A.D., and Lien, c.L.: "Phase BehaviorofCO z
and Crude Oil in Low Temperature Reservoirs," Soc. Pet. Eng. J.
(Aug. 1981) 480-92.
7. Turek, E.A. et al.: "Phase Equilibria in Carbon Dioxide-Multicomponent Hydrocarbon Systems: Experimental Data and Improved Prediction Technique," paper SPE 9231 presented at the
SPE 1980 Technical Conference and Exhibition, Dallas, Sept.
21-24.
8. Heidemann, R.A.: "Three Phase Equilibria Using Equations of
State," AIChE J. (1974) 20,847-55.
9. Evelein, K.A., Moore, R.G., and Heidemann, R.A.: "Correlation of the Phase Behavior in the Systems Hydrogen Sulfide-Water
and Carbon Dioxide-Water," I&EC Proc. Des. Dev. (1976) 15,
423-28.
10. Sorensen, J.M. et al.: "Liquid-Liquid Equilibrium Data: Their
Retrieval, Correlation and Prediction," Fluid Phase Equilibria
(1979) 3, 47-82.
II. Coleman, B.D.: "On the Stability of Equilibrium States of
General Fluids," Arch. Rational Mech. Anal. (1979) 36, 1-32.
12. Dunn, J.E. and Fosdick, R.L.: "Morphology and Stability of
Phases," Arch. Rational Mech. Anal. (1980) 74, 1-99.
13. Gautam, R. and Seider, W.D.: "Computation of Phase and
Chemical Equilibrium: Part II-Phase Splitting," AIChE J.
(1979) 25, 999-1006.
14. Yarborough, L.: "Application of a Generalized Equation of State
to Petroleum Reservoir Fluids," Equations of State in Engineering
and Research, K.C. Chao and R.L. Robinson Jr. (eds.) ACS,
Washington, D.C. (1979) 385-439.
15. Baker, L.E. and Luks, K.D.: "Critical Point and Saturation
Pressure Calculations for Multicomponent Systems," Soc. Pet.
Eng. J. (Feb. 1980) 15-24.
16. Peng, D.Y. and Robinson, D.B.: "A Rigo~ous Method for
Predicting the Critical Properties of MUlticomponent Systems
from an Equation of State," AIChE 1. (1977) 23, 137-44.
17. Ng, H.-J. and Robinson, D.B.: "Equilibrium Phase Properties of
the Toluene-Carbon Dioxide System," 1. Chern. Eng. Data
(1978) 23, 325-27.
18. Kulkarni, A.A. et al.: "Phase Equilibrium Behavior of System
Carbon Dioxide-n-Decane at Low Temperatures," J. Chern. Eng.
Data (1974) 19, 92-94.

APPENDIX

Theorems Related to Gibbs Energy Analysis

The following terminology and definitions are needed in
the development of the theorems.
Let a hypothetical phase of I components be
represented by r=(rl, r2 ... r/), where ri is the
number of moles of species i in the phase. If a feed
OCTOBER 1982

stream is represented by (nl, n2'" n1), we say the

phase r is admissible if 0 < r i < n i' This discussion is
limited to systems of admissible phases, in which each
component is present (at least in a trace amount) in each
phase. The results are not strictly applicable to a system
in which any component is excluded from one or more
phases.
A possible state of a system of I species and J phases
resulting from a feed n is designated m=[mj] wheremj
is thejth phase. Conservation of mass requires
J

L: mij=ni ............................ (A-I)

j=1

for each i, I $. i $. I. Herein, it is assumed that one has a

nonreacting system.
At a given pressure and temperature, the Gibbs energy
function for a phase is G=G(r). The Gibbs energy is a
continuous first-order function in accordance with
classical thermodynamics (e.g., see Callen I). The total
Gibbs energy for a system of J phases is

L:

G(mj), ..................... (A-2)

j=1

which is the sum of the Gibbs energies of the constituent

phases.
An equilibrium state for a system will exhibit a global
minimum in the Gibbs energy. The following definitions
are made to distinguish between phase states corresponding to local and global minima.

Definition 1. A state of J admissible phases satisfying

conservation of mass is an equilibrium state if
Gt(n)=min Gr(m) where the minimum is taken over all
states mof K admissible phases satisfying conservation
of mass (K not necessarily equal to 1).

Definition 2. Let
be a state of J admissible phases
satisfying conservation of mass and let G be differentiable at each phase in n. Suppose that for every differentiable curve m(7)=[mj(7)], which is defined for 7 in
some open interval containing 0 and which satisfies conservation of mass-i.e.,
J

L: mij(7)=ni,

......................... (A-3)

j=1

and for which mjCO) = n j, I $.j $.J, we have

d
A
-G r [m(7)]I7=O =0.

d7

. .................... (A-4)

Then we say that n is a stationary state.

A stationary state corresponds to an extremum (local
or global minimum or maximum) or saddle point of the
total Gibbs energy. These two definitions suggest that
out of the stationary states found, the equilibrium state
739

must be identified. The purpose of the theorems

presented is to develop criteria for this determination.
We show in Theorem I that the conditions of mass
flow equilibria,

are equivalent to the condition of stationarity of the

system.

Thus, is a stationary state. Conversely, stationarity

of a state
implies that the conditions of mass flow
equilibria apply. This can be demonstrated by considering special curves, along each of which the only variation is in one component in two different phases.
Central to the development of the Gibbs energy
analysis is recognizing that can be determined by the
points of tangency of a hyperplane tangent to the surface
G.

Theorem 1. Let n be a state of J admissible phases [n)]

satisfying conservation of mass. The state
is a stationary state if and only if G is differentiable at each n)
and the chemical potentials, fLij, do not vary with the
phase-i.e., fLij =fLib as in Eq. A-5.

Theorem 2. Let
be a state of J admissible phases
[n)] satisfying conservation of mass. Then G r is stationary at n if and only if G is differentiable at each n)
and the surface G has the same tangent plane at each of
the points [n ) ] .

fLij=fLib l~i~I, l~(j, k)~J, ............ (A-5)

Proof. The plane L j tangent to Gat

Proof. Suppose that the chemical potentials do not vary
with phase. Consider nl(T) as in Definition 2. By the
chain rule,

nj is

Also,

...................... (A-6)

since G is a first-order function (Euler's theorem). It

follows that

Then

aG
-a (n)ri = ~ fLijri' ....... (A-l3)
i=1
ri
i=1
I

L)(r)=

~
LJ

~ aG[m)(O)] x dmij(O) ,
LJ
......... (A-7)

)=1

i=1

amij

dT

but
aG[m)(O)]
amu

aG(n)
=---=fLij, ................ (A-8)
ami}
.

But because fL ij does not depend on) for a stationary

state (Theorem I), L) (r) is the same tangent plane for
each of the} phases. Conversely, if all L j are the same,
differentiation shows that the fL U' s do not depend on}.
We now sho)V that a stationary state,
is an
equilibrium state if and only if the corresponding common tangent plane of Theorem 2 never lies above the
surface G at any point. Let D be the difference between
G and the tangent plane L :

n,

D(r)= G(r) -L(r). . .................... (A-14)

and fLij does not depend on the label}. Thus,

D(r)=G(r)-

~ fLijri
i=1
I

=G(r)- ~ fLilri ................ (A-15)

fLi} LJ

dT

)=1

i=1

i=1

dmij(O) --0

............... (A-9)
'

since ~ mij(T)=ni = fixed feed constraint, and so

for a stationary state

)=1

740

It is clear from this that

never negative.

)=1
J

n.

D(n )=0, 1s,}s,J. We now show in the next few

results that is an equilibrium state if and only if D is

dmij(O)

dni

dT

dT

----''--- = -

=0. . ................ (A-IO)

Lemma 1. Let be a stationary state with J admissible

phases [n]. Let liz be a collection of K admissible
phases [m) satisfying conservation of mass (K and J are
not necessarily equal). Then

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

Proof. Recall that

G{(m)-G{(n)=

D(mj)' .............. (A-16)

D(r)=G(r)-L(f)

j=i

Proof. From mass balance,

=G(r)-G(n j)K

nij=

j=i

mij' 1:51:51. .............. (A-17)

j=i

...................... (A-23)
Also, J1.ij =J1.iI for 1:5 i :5/, 1 :5}:51. Now,
for each}. Since G is differentiable at nj, the function

G{(n)=

GUi j)

j=i

D(r)/

b b
j=i ;=i

Irv-n'jl ..................... (A-24)

approaches zero as
approaches nj. Let

b
j=i

b
v=i

J1.iI n ij

bJ1.;i m ij

approaches O-i.e., as

;=i
f

r- nj

L(mj),

.................... (A-IS)

F;Cr - n j)=/ID(r)

II b

Irv

-n,jl ......... (A-25)

v=i

j=i

Then

so

ID(r)l=

:51 m.ax Ir;-n;;lID(r)11

ISlsf
v=i

Ir

"

- nVj

......................... (A-I9)

D(mj)'

=F/(f-n j ) m;lx Ir;-n;/1. ....... (A-26)

.
ISlsf
.

j=i

Theorem 3. Let be a stationary state and suppose that

G lies on or above the common tangent plane-i.e.,
D( r) ~ 0 for all admissible phases r. Then n is an
equilibrium state.
Proof. Consider any state
Then each D(mj) 2: 0, and

mcontaining K phases 1mj J.

.

Irv-nvjl
f

b
v=i

[G(mj)-L(mj)]

j=i
=

Ir;-nijIID(r)11

;=i

Theorem 4. Let n be an equilibrium state with J admissible phases-z at each of which G is differentiable, and
suppose that f is an admissible hypothetical phase.
Then D(f) 2: O.
Proof. Assume D( f) < O. Note that since n is an
equilibrium state, it is also a stationary state.
Choose the set of positive numbers [Ej J so that

D(mj)2:0 ......................... (A-20)

j=i

nij-E/;>O, ........................... (A-27)

or
G{(m)-G{(n)2:0 ....................... (A-2I)

by Lemma 1. By Definition 1, n is an equilibrium state.

Next we prove that D, as the difference between a
function and its tangent plane, is (roughly speaking) of
second order in its argument.

for 1 :5i:5/, 1 :5}:5J. Also, let these numbers be sufficiently small that, when Lemma 2 is applied to
r= n j -E?
- n)
j ' F(r
j
j satisfies
Fj(-Ejf)<ID(f)11 m;lx (f;) . . . . . . . . . . , .. (A-2S)
I Sisl

Construct a state mof J + 1 phases by

Lemma 2. Let
be a stationary state with J nonzero
phases [n jJ. Then, for 1 :5}:5J, there is a functi(),n Fj
which approaches 0 as its argument approaches 0 and
for which
ID(r)1 :5F/Cr-n/) m;lx Ir;-n;jl . ........ (A-22)
.
. ISlsf
.
OCTOBER 1982

and
J

mJ+i =

b EJ ........................ (A-30)
j=i
741

m,

This state,
satisfies conservation of mass with
respect to the feed composition. Furthermore, each mij
is positive and each mj is admissible. It follows that
J+!

E/) ..... (A-31)

j=!

j=!

D(n j -E/)+(

j=!
J

~ ~

Ej)D(f)

j=!
J

[D(n j-E/)[+(

j=!

[D()[Ej+(

j=!

~ D(mj)= ~ D(mj)+D( ~
j=!

< ~

~ Ej)D()
j=!

=( ~ Ej)[[D()[+D()J
j=!

=0, .................................. (A-33)

if D()<O. Thus,
Ej)D(l\

j=!

J+!

. . . . . . . . . . . . . . . . . . . . . . (A-32)

D(mj)<O .

j=!

By Lemma 1, GI(m) < GI(n), or n is not an equilibrium

state. However, we have specified that n~ is an
equilibrium state; therefore, the assumption D( e)< 0 is
incorrect.

Using Lemma 2,

SI Metric Conversion Factors

OF
(OF-32)/1.8
psi x 6.894 757

+( ~ Ej)D()]
j=!

E+OO

=oC
=kPa

SPEJ
J

~ F ( - E r)E ( m;lx ei ) + ( ~
LJ }
}
} 1$ [ $ /
LJ!

j=!

742

J=

Original manuscript received in Society of Petroleum Engineers office Dec. 17, 1980.

E )D(
J

Paper accepted for publication May 2, 1982. Revised manuscript received July 6,
1982. Paper (SPE 9806) first presented at the 1981 SPE/DOE Enhanced Oil Recovery
Symposium held in Tulsa April 5-8.

SOCIETY OF PETROLEUM ENGINEERS JOURNAL