You are on page 1of 35

Previous Page

46 Chapter 2. THERMODYNAMICS: PHASE Ei2UILlBRlA

(2.32)

m=l

The interaction parameters between groups, Y,,,, depend on the tempera-


ture:

v,,, = exp (- +) (2.33)

Equations 2.30 to 2.33 above can of course be applied to the calculation of


r, and r,,i.
Two categories of groups are defined. The "main" groups, currently 50 in
number, correspond to one type of interaction, and therefore to one particular
pair of parameters uk,,.The same is true for the "CH," and "CH2CO" groups for
example. For some of the main groups subgroups have been defined which dif-
fer in the values of the Bondi parameters, R, and Q,: for example the sub-
groups CH3, CH2, CH and C for the principal group "CH2" and CH,CO and CH2C0
for the principal group "CH2COn.There are a total of 108 subgroups.
Since the method's first definition it has been regularly developed (Skjold-
Jorgensen et al., 1979; Cmehling et al., 1982 Weidlich et al., 1983; Tiegs et al.,
1987; Hansen et al., 1991) and the matrix of interaction parameters determined
by experimental data correlation has been considerably extended as shown in
Table 2.8.
In addition to these extensions, numerous variations on the method have
been proposed (Fredenslund and Rasmussen, 1985). In particular, Larsen et al.
(1987) modified the combinatorial term and detailed the variations in parame
ters with temperature by taking heat of mixing into account in the data bases
that are used to determine them. More recently it has been shown that this
method could be incorporated in equations of state for the definition of mix-
ing rules. We will come back to this point later on.

2.1.4 "Homogeneous" Methods Applied


to the Calculation of the K Value
Contrary to heterogeneous methods which apply different models to the liq-
uid and vapor phases, emphasizing the deviations from ideality, homogeneous
methods represent the phases at equilibrium by the same equation of state.
This equation of state can be used to calculate the fugacity values in order to
apply the equilibrium condition. However, the values of the volume occupied
by the system, its enthalpy, etc. can also be deduced, thereby ensuring perfect
consistency among all of the properties calculated. This approach is therefore
attractive in its principle.
Most of the equations of state used are derived from the van der Waals the-
ory (1873) and must represent the relevant properties in a very wide range of
density, since they will be applied indifferently t o the liquid and vapor phases.
Chapter 2. THERMODYNAMICS: PHASE ~QUlLlSRlA 47
48 Chapter 2. THERMODYNAMICS: PHASE EQUILIBRIA

Because of their very design they comply with the critical coordinates (criti-
cal temperature, pressure) of pure substances and they can also be used to
calculate the vapor-liquid equilibria at high pressure and the critical points of
mixtures, at least in a semiquantitative manner. The continuity between the
liquid and vapor states is ensured by the very principle of these methods.
The success of the method proposed by Soave (1972) has given rise to
renewed interest in research in this field and in the applications that are made
of it. Rightly or wrongly, these homogeneous methods have become the
panacea of chemical engineering calculations, but their weaknesses must in no
way be ignored.

2.1.4.1 The Equations of State Derived


from the van der Waals Theory
a. The van der Waals Equation of State
We give the equation of state proposed by van der Waals (1873) below as a
reminder:
p = - -RT
- a
(2.34)
u-b o2
A repulsion term can be recognized in the equation which accounts for the
volume of the molecules per se by means of the parameter b,or covolume, as
well as an attraction term, or internal pressure, which depends on the param-
eter a.
For pure substances, the values of these parameters are determined by
applying the critical constraints and we would like to discuss the conse-
quences of this type of choice. There are three constraints: the equation of
state must be satisfied at the critical point, along with the equations below:

($)T=(g)T=O for T = T c and P=Pc (2.35)

In order to solve the system formed by these three equations which con-
nect five variables (the critical coordinates T,, P, and u,, and the two parame-
ters a and b of the equation of state), the values of two and only two of these
variables must therefore be fixed. As a general rule, these will be the critical
temperature and pressure, T, and P,. The other variables (parameters a and b,
critical volume u, or critical compressibility factor Z,) will be calculated. As a
result, for equations of state with two parameters, the critical coordinates cor-
responding to the equation of state will coincide with the experimental data
only for the values of pressure and temperature, and the volume (or the com-
pressibility factor) is usually overestimated. This leads to the equations:
R2T:
a=Qa- (2.36)
PC
Chapter 2. THERMODYNAMICS: PHASE fQUlLl6RlA 49

The numerical values of the nondimensional parameters, Q,, fib and the
critical compressibility factor are:
27 1 3
Q,=- Qb -- and Zc=-
64 8
If we assume that the parameters a and b are independent of the tempera-
ture, then the set of isotherms P ( u ) obtained has the right configuration so
that the liquid and vapor states are properly represented, at least qualitatively
speaking (the subcritical isotherm exhibits a maximum and a minimum), or
"supercritical" (the isotherm is monotonic). The essential qualities of the work
of a pioneer should be emphasized: representation of the different fluid states,
and prediction of the law of the corresponding states (Abbott, 1989).
The simplicity of the equation should also be noted: at a given temperature
and pressure, it is of the third degree in volume and can be solved readily with-
out resorting to an iterative processus. In addition, the roots corresponding to
the liquid and vapor phases are easily identifiable. The equations of state
derived from the van der Waals equation are often called "cubic equations of
state".
However, as van der Waals himself mentioned, this equation of state is not
of satisfactory accuracy. The critical compressibility factor corresponding to
it has a value (3/8) that is too high, and the calculation of molar volume in the
liquid phase entails systematic errors by unacceptable excesses. Additionally,
if the vapor pressure is calculated by this equation of state, the results are
very different from experimental values. These factors explain in part the con-
siderable volume of research that has extended van der Waals's work and
resulted in the equations of state that are widely applied today.

b. The Soave-Redlich-KwongEquation of State


In order to better represent the molar volumes of fluids, Redlich and Kwong
(1949) proposed an empirical modification of the van der Waals attraction
term. They also introduced a variation of the parameter a with temperature:

(2.38)

with:

(2.39)

and, the same as for the van der Waals equation:


50 Chapter 2. THERMODYNAMCS:
PHASE ~ Q U f L f S R f A

By applying the critical constraints, the values of the parameters R, and f i b


and of the critical compressibility factor 2, are determined, with the following
result:
1 2 113 - 1 1
R, = 0.42748 and L-2, = - % 0.086640 z, = -
%

9 (2 1’3 - 1) 3 3
As far as calculating density is concerned, the results are certainly
improved, in particular in the vapor phase, but still remain poor for the vapor
pressures. The Redlich-Kwong equation of state has, however, been one of the
most widely used and an is one of the component parts of the method pro-
posed by Chao and Seader (see Section 2.1.3.3.a) to calculate vapor-liquid
equilibria.
Cubic equations of state with two parameters can not in fact represent both
molar volumes and vapor pressures with satisfactory accuracy, and a priority
must be chosen. This is what Soave (1972) understood when he applied the
Redlich-Kwong equation of state to the calculation of hydrocarbon vapor pres-
sures. By considering that the covolume keeps the value corresponding to the
critical point (Eq. 2.37), any vapor pressure can be used to determine the value
of the attraction parameter a at this temperature by applying the equation of
state to the phases in equilibrium, (liquid and vapor) and by taking the equi-
librium condition (equality of fugacities) into account. Soave proposed the
expression below to represent the parameter variation with temperature
obtained in this way:
a(7-,)= [ l + m(1 - *)I2 (2.40)
The parameter m is specific to the component under consideration but its
value has been correlated versus the acentric factor:
m=M0+M,o+M202 (2.41)
with:
Mo = 0.48 Mi = 1.574 M2 = -0.176

For temperatures higher than the critical temperature, the preceding equa-
tions are assumed to apply without any modification.
The value of L-2, is of course the same as that involved in the Redlich-Kwong
equation of state (0.42748), so that the critical constraints are complied with.
The same is true for the critical compressibility factor which stems from the
equation of state: its value is equal to 1/3,i.e. much lower than the one in the
van der Waals equation of state, but still too high compared to experimental
values, which range between 0.25 and 0.29. It therefore comes as no surprise
that the Soave-Redlich-Kwong method should lead to a systematic deviation
by excess in the calculation of molar volumes, for the liquid phase in particu-
lar. For hydrocarbons, this deviation is particularly significant when the molec-
ular weight increases. In contrast, the vapor pressures are fairly well
represented as can be predicted by the criterion chosen to define a(n.
Between the boiling point at atmospheric pressure and the critical point, the
deviations are approximately 1 to 2% (Table 2.9). In contrast, extrapolation to
Chapter 2. THERMODYNAMICS: PHASE ~QUlLlSRlA 51

low reduced temperatures is usually poor. The numerous modifications that


have been suggested since Soave's work have mainly attempted to improve
the calculation of vapor pressures and render acceptable liquid phase molar
volumes.

Vapor pressure Density


Component (0.55 < Tr < 1) Vr =0-7)
vdW* SRK PR vdW* SRK PR

Methane 1.47 1.29 0.72 -34.5 -1.7 10.2


Ethane 1.11 0.92 0.56 -40.0 -5.3 1.5
Propane 1.34 0.92 0.42 -41.7 -6.3 6.1
n-Butane 1.26 1.05 0.37 -44.0 -7.8 4.8
n-Pentane 1.48 1.26 0.43 -47.0 -9.9 3.0
n-Hexane 1.70 1.46 0.82 -50.9 -13.8 0.6
n-Heptane 1.02 0.79 0.66 -52.9 -13.8 -0.6
n-Octane 1.55 1.30 0.71 -55.7 -15.6 -2.2
n-Nonane 1.83 1.56 0.60 -57.0 -16.5 -3.0
n-Decane 2.25 1.97 1.01 -55.9 -15.4 -2.1
* The van der Waals equation has been treated here with a method similar to the one applied
by Soave (relations 2.39 to 2.41).

Mean relative deviations PA) in the calculation of vapor pressures and densi-
ties by the modified van der Waals (vdW*), Soave-Redlich-Kwong(SRK) and
PeneRobinson (PR) equations of state (Rauzy, 1982).

c. m e Peng-Robinson Equation of State


The first and best known of these modifications is the result of the research by
Peng and Robinson (1976) who changed the attraction term in the equation of
state which becomes:
RT a
p = -- (2.42)
0-b v2+2bu-b2
The way of calculating the parameters is unchanged and Eqs. 2.37, 2.39,
2.40 and 2.41 are applicable. Of course the numerical values of the constants
Q,, a,, Mo, MI and M2 are different, they are listed in Table 2.10.
Due to the size of the data base, the vapor pressures are more precise
c a b l e 2.9), but improvement mainly resides in the calculation of densities in
the liquid phase.

2.1.4.2 Calculation of Thermodynamic Properties


with Equations of State
If we consider the general cubic equation:
RT a
p = -- (2.43)
u - b (U - brJ (U - brd
52 ChaDter 2. THERMODYNAMICS: PHASE EOUlLlBRlA

which, depending on the value of the parameters rl and r,, represents the van
der Waals, Soave-Redlich-Kwong or Peng-Robinson equations of state, it can be
shown (Vidal, 1997, p. 120) that the expression of the fugacity coefficient is:

P(v - b ) a
Incp=-In - RT
+z-1+-
bRT
U(v, b, rl, rd (2.44)

Table 2.10 lists the values of the main parameters that appear in Eq. 2.43,
in the expression of the covolume b (Eq. 2.37), of the attraction parameter a
(Eqs. 2.39 to 2.41), and the expression of the function U for these equations of
state. The method recommended by Soave to calculate the attraction parame-
ter (Eqs. 2.39 to 2.41) has been applied to the van der Waals equation of state.

Equation of state
Parameter
van der Waals have-Redlich-Kwong Peng-Robinson

0 -1 -lh
-1 -1 +v3
27/64 119 (2 1/3 - 1) 0.45724
118 (2 - 1)/3 0.07780
0.5000 0.48 0.374 64
1.5883 1.574 1.54226
-0.1757 -0.176 -0.26992
318 113 0.307
--
b 1 /v-br, \

General expression of the van der Waals, SoaveRedlich-Kwong and Peng-


Robinson equations.

By proposing a new method of determining the attraction parameter a in


the equations of state derived from the van der Waals theory, Soave enabled
vapor-liquid equilibria to be calculated for mixtures of apolar compounds, and
in particular for oil and gas fluids. Application of this type of equation of state
to a pure substance was a necessary step, which will complete the definition
of mixing rules (Section 2.1.4.3). These rules relate the values of the parame-
ters to the composition and play an important part in the practical value of an
equation of state.
It should be noted in particular that, if only the vapor pressure or the den-
sity of a pure (apolar) substance is to be calculated, there are simpler and
more accurate methods (Paradowski, Vol. 1, Chapter 4).
In any case, the cubic equations of state that we have just presented pro-
vide only a poor representation of the critical region.
Chapter 2. THERMODYNAMICS: PHASE fOUlLlBRlA 53

2.1.4.3 Equations of State Involving Three Parameters


and Hard Sphere Equations of State
The generalized application of Soave-Redlich-Kwong equations of state to
chemical engineering calculations and to the study of oil and gas fluids (natu-
ral gases, crude oils) has given rise to a large amount of research aiming to cor-
rect the equations’ defects. Without attempting to give an exhaustive review of
the modifications that have been proposed, we will present some examples.
The use of a third parameter provides highly advantageous flexibility when
both vapor-liquid equilibria and molar volumes are to be calculated.
Schmidt and Wenzel (1980), Harmens and Knapp (1980), Heyen (1980),
Pate1 and Teja (1982) present an equation of state fitting the general equation
below:
RT a
p = -- (2.45)
0-b v2+ubv+Wb2
with:
u+w=l
The basic unsuitability of cubic equations of state in representing both the
critical zone and the saturated liquid properly does not, however, allow the
third parameter to be chosen so that the three critical coordinates are calcu-
lated exactly. Even though the rationale used by the various authors in calcu-
lating the parameters differs, the results are not fundamentally different.
Translation of volume and “generalized cubic equation”
The variations on equations of state that we have just mentioned can be
pulled together by the concept of “translation of volume”. It was introduced by
Martin (1979) and the full extent of its value has been demonstrated by
PCnCloux et al. (1982).
We assume that we have two equations of state e(T, P, v) = 0 and e ’(T , P,
0’) = 0, such that for any value of pressure they provide the values u and v’ for
the volume respectively. In addition, these values differ only by a value c
which depends (possibly) only on the temperature. If this is true then their
graphic representation in the volume pressure plane comprises two curves
translated one from the other. It can be shown that if the Maxwell condition
(equality of areas located between the curve P(v) and the isobar P “) is com-
plied with for one it is also complied with for the other. These two equations
of state will therefore lead to the same results for the calculation of the vapor
pressure. Accordingly, the calculation of volume and of phase equilibria are
independent to a certain degree, and we can correct one equation of state
e’(T, P, 0’)= 0, which is satisfactory for calculating vapor-liquid equilibria but
”faulty” for calculating volumes, by a simple translation of volume:
U=V’-C (2.46)
This method is valid of course for any equation of state, but it is particu-
larly effective in providing the equations derived from the van der Waals the-
ory with the flexibility they were lacking.
54 Chapter 2. THERMODYNAMICS: PHASE EOU/L/BR/A

The value of the translation parameter can be adapted to the substance


under consideration. It is in fact frequent that we have data such as the den-
sity of the liquid at ordinary temperature or at the normal boiling point for low
molecular weight compounds. The parameter c can then be calculated in such
a way that the equation of state represents such data exactly. The translation
operated on the van der Waals, Soave-Redlich-Kwong, and Peng-Robinson
equations evens out the considerable differences that the equations exhibited
in calculating molar volumes.
The differences between the thermodynamic properties of the liquid and
vapor phases in equilibrium are naturally not modified by translation, for
example the vaporization enthalpy is unchanged.
The innumerable variations on the van der Waals equation of state that
have been proposed contain a large proportion of empiricism. The satisfac-
tory results obtained are in fact due to a fortunate balance between the inex-
actitudes in the expression of the repulsion term and the different forms
proposed for the attraction term.
The “hard spheres” theory and then the “hard chains” theory (see for
example Donohue and Vilmachand, 1988) have been used to propose new
equations of state that deserve lengthy development. However, they are not
commonly found in chemical engineering calculations and should be used
with caution.

2.1.4.4 M i x i n g Rules Associated with Equations of State


If we now consider mixtures, the “composition” variable will have to be taken
into account in formulating equations of state, which are written as follows:

and involve temperature, pressure, volume and the number of moles of each
component, or for one mole of mixture:
e(T, P, u, zl, z2,z~..)
=0

In the latter equation, ziis the mole fraction of component i, generally des-
ignated by xi if the mixture is homogeneous and liquid, and by yi if it is in the
vapor state.
These expressions should in fact also include the list of parameters of the
equation of state under consideration: covolume, attraction parameter, etc.,
already encountered for pure substances. It is usually in these parameters that
the influence of the composition is considered by mixing rules.
Mixing rules normally comply with the structure of the equation of state
under consideration, so that the implicit assumption can be made that a mix-
ture of specified composition behaves like a fictitious pure substance. The
terms “one fluid model” is also used. This is an assumption with no justifica-
tion except for its simplicity and the results obtained.
Chapter 2. THERMODYNAMICS: PHASE EOUlLlBRlA 55

a. Classical Mixing Rules


Generally speaking, equations of state derived from the van der Waals theory
are associated with what are termed "classical" mixing rules:
a= cn

i=l j=1
n
C ai,jzizj (2.47)
n
b = C bit; (2.48)
i=1

a..
131
= v E (1 - kU)
51 I,] with kj,i = k;,, (2.49)
In the preceding expressions, the terms a;,;,bi represent the parameters of
pure substances. The calculation of the binary terms involves a parameter
ki,,, called the interaction parameter which is determined from experimental
phase equilibrium data. We will come back to this term.
The composition of the mixture is designated by tiabove. When the system
is biphasic, the equation of state itself (Eq. 2.38 or 2.42 or 2.43) and Eqs. 2.47
to 2.49 are applied to one then the other phase, and the corresponding com-
position, xi for the liquid phase and yi for the vapor phase, will be involved in
these equations. At a specified pressure and temperature, the equation of state
can then be solved and the suitable root assigned to the phase under consid-
eration. The thermodynamic properties will then be calculated for one then
the other phase by application of the general equations.
For the calculation of the mixture fugacity in particular, it has been shown
(Vidal, 1997, p. 289) that:
P(u - b) b;
In 'pi= -In -+ - (Z- 1)
RT b

a
+ bRT(rl - r,J
In -
u - br,
(2.50)

b. Mixing Rules Applied to the Translation Parameter


We have seen (see Section 2.1.4.3) that any equation of state can be corrected
by "translation". To this end we considered that the result of the equation of
state before correction was a calculation instrument, a "fictitious" volume
noted u', and that the volume was obtained by means of the equation below:
u = u'-c (2.46)
The value of the translation, c, is independent of the pressure. In these con-
ditions the calculation of the vapor-liquid equilibrium, i.e. the vapor pressure,
is not modified for pure substances.
We must define a mixing rule for this parameter as well in such a way that,
for mixtures as well as for pure substances, translation does not modify the
calculation of phase equilibria. As a result, we will state:
i=n
c= c c;ti
i= 1
(2.5 1)
56 Chapter 2. THERMOOYNAMICS: PHASE ~CJUlLlBRlA

c. Area of Application of Conventional Mixing Rules


and Examples of Results
Theoretically at least, the van der Waals equation of state can be applied only
to non-polar compounds, all the more so when the conventional mixing rules
we have defined are applied to the equation. However, this leaves out a large
area, oil and gas fluids in particular. Both in producing oil and gas fields and in
the refining industry the Soave-Redlich-Kwong and the Peng-Robinson equa-
tions of state are commonly used. Many modifications have been made which
we have briefly mentioned in the first part of this discussion, but as long as we
confine ourselves to non-polar compounds, it is the conventional mixing rules
that are adopted.
Now the main problem that arises is that of determining the interaction
parameter ku. We will come back to this point.
The results are generally good. For example Figure 2.7 shows the equilib-
rium diagram of the COZ-ethanesystem. For this calculation, the value of the
interaction parameter k, is taken as equal to 0.13 and independent of the tem-
perature. It can be seen that the experimental values are quite well correlated
and that the same is true for the azeotropic behavior and the vicinity of the
critical region. Likewise, Figure 2.8 (Huron et al., 1977) represents the loci of
the critical points for mixtures of carbon dioxide and paraffinic hydrocarbons.
It should be noted that at 293 K the mixtures of COz and ethane exhibit two
critical points, as shown in Figure 2.7.

150

z
h

e
?!?
3
v)

$ 100
a

50

-
Figure
2.8 Loci of critical points for mixtures o f CO, and paraffins. The curves are calcu-
lated by the SRK method with k, = 0.13. The points are experimental.
Chapter 2 THERMODYNAMICS:
PHASE ~QUILIBRIA 57

It is tempting to extend this method to other families of compounds and


Asselineau et al. (1978) have shown that the Soave-Redlich-Kwongmethod can
be applied to mixtures of chlorofluorinated compounds, after the interaction
parameter has been determined of course.
In any case, a recommendation must be made: vapor-liquid equilibria of a
mixture can not be correctly correlated or predicted if the vapor pressure
itself of the pure substances is not correctly predicted. We have seen that the
application of the Soave-Redlich-Kwong and Peng-Robinson equations of state
in their original form leads to a mean error of approximately 1 to 2% in the tem-
perature interval between the boiling point at atmospheric pressure and the
critical temperature. These results deteriorate considerably at lower tempera-
tures and care is required at this point in dealing with systems containing com-
pounds with a low vapor pressure and, of course, compounds other than
hydrocarbons.

d. The Binary Interaction Parameter


Theoretically, the interaction parameter kY which is involved in the calcula-
tion of the binary terms ai,i:

a.hJ. = q K (1 - ku) with ki,i = ku (2.49)


must be determined from experimental vapor-liquid equilibrium data. It can be
disregarded in the case of paraffin mixtures, for example the loci of critical
points concerning ethane/paraffin mixtures C, to C,, can be calculated with
k, = 0. Additionally, in technical literature there are a large number of publi-
cations where laboratory results are correlated by Soave-Redlich-Kwong or
Peng-Robinson equations of state, thereby determining optimum values of this
parameter. Particular mention should be made of the research by Grabovski
and Daubert (1978a, 1978b, 1978c) and the compilation by Knapp et al. (1982)
By way of an example, Table 2.1 1 shows the interaction parameters recom-
mended for the binaries containing nitrogen, carbon dioxide or hydrogen sul-
fide.
The binaries containing hydrogen, which are often encountered in the refin-
ing industry, also pose a difficult problem which has been the subject of a large
number of studies (Grabovski and Daubert, 1978a, 1987b, 1978c; Moysan et al.,
1986). First of all, the whole equilibrium diagram can not be represented sat-
isfactorily. If we confine ourselves to pressures lower than 300 bar, then it is
possible to obtain valid results, but the values of the interaction parameter are
particularly high. For example, ki,i = 0.742 for the application of the Soave-
Redlich-Kwong equation to the hydrogenln-butane binary at 120°C. The inter-
action parameters of hydrogen with hydrocarbons are in addition not very
sensitive to the nature of the hydrocarbon component, but they vary with tem-
perature. Finally and more important, contrary to what is observed usually,
the results are not very sensitive to the value of the parameter k,. All these
observations can be explained fairly well if we take into account the very high
value of the reduced temperature of hydrogen (T,= 33 0. The application
of Eqs. 2.39 to 2.41 proposed by Soave results in a very low value for the
58 Chapter 2. THERMODYNAMICS: PHASE EQUILIBRIA

attraction parameter a for hydrogen. The same will therefore be true of the
binary parameter ai,, obtained by applying the classical mixing rule (Eq. 2.49),
if the values of k, are not particularly high themselves.

Component N2

N2 0 0 0
CO2 0 0 0.12
H2S 0 0.12 0
C1 0.02 0.12 0.08
c2 0.06 0.15 0.07
c3 0.08 0.15 0.07
i-C4 0.08 0.15 0.06
n-C4 0.08 0.15 0.06
i-C, 0.08 0.15 0.06
n-C, 0.08 0.15 0.06
n-C, 0.08 0.15 0.05
{Tablei-
Interaction parameters kij applied to the Soave-Redlich-Kwong equation of
state (h-om "Contributions in Petroleum Geology and Engineering", Vol. 5:
Properties of oils and natural gases, p. 84 by K.S. Pedersen, Aa. Fredenslund
and P Xbomassen. Copyright 0 1989 by Gulf Publishing Company, Houston,
Texas. Used with permission. All rights reserved).

Another problem involves the variation of the interaction parameter ki,.


with temperature. This variation is often disregarded, in any case it usually
remains moderate. It should, however, be pointed out that the variation can
have a not insignificant impact on the calculation of mixture enthalpy.
Another proposal is to determine the interaction parameter from the Henry
constant, i.e. from gas solubilities in hydrocarbon solvents. The Henry con-
stant is defined by the following equation:

(2.52)

If x1+ 0, then, if the solvent is a pure substance, x2 -+ 1, and in the mea-


surement conditions of the Henry constant, the pressure approaches the
vapor pressure, P ; of the solvent. Eq. 2.50 then becomes:
Chapter 2. THERMODYNAMICS: PHASE ~CJUlLlBR/A 59

the values of PF and of uk0 verifying the equation of state and the vapor-liquid
equilibrium equation for the pure solvent.
As a result, it can be seen that the Henry constant is explicitly related to
the interaction parameter k , and to the properties of the solvent at saturation.
This method is naturally excellent for predicting the part of the equilibrium
diagram corresponding to the dilute areas, but it is not usually possible to pre-
dict the whole diagram or in particular the critical area in this way.

2.1.5 Relations between .'Homogeneous"


and "Heterogeneous" Methods
We have seen that heterogeneous methods are basically characterized by the
definition of a model of deviations from ideality in the liquid phase. They offer
undeniable qualities of flexibility, and the choice of the method can (and must)
be adapted to the nature of the mixture under study. The concept of group
contribution gives them an appreciable predictive feature. However, heteroge-
neous methods can not be applied to the calculation of phase equilibria at
high pressure, particularly in the vicinity of the critical area.
In contrast, homogeneous methods comply with the continuity between
the liquid and vapor states. However, in practice they are limited to equations
of state and to the mixing rules that we have just described and can not be
applied to mixtures containing polar compounds in these conditions.
Can the qualities of the two sorts of methods be exploited at the same
time? The first research on this subject (Huron and Vidal, 1979) is based on the
simple relationship that exists between excess Gibbs energy and the fugacity
coefficients:
n
gE= RTQn cp - C tiIn 9): (2.53)
i= 1

In this relationship, cp represents the fugacity coefficient of the mixture hav-


ing a composition tiand cpf the fugacity of the components in the pure state
under the same conditions of temperature and pressure. After applying the
equation of state as well as the expression of fugacity coefficients (Eq. 2.44) to
the mixture and then to the components, the excess Gibbs energy can be cal-
culated, but the expression of this parameter is not explicit.
However, it can be shown that if the pressure approaches infinity, given the
linear mixing rule on the covolume (Eq. 2.48), a very simple relationship is
obtained:

(2.54)

where g ; ~ stands for the limit that the excess Gibbs energy approaches
when it is calculated by the equation of state and the pressure approaches
infinity. A is a numerical constant specific to the equation of state under
60 ChapferZ. THERMODYNAMICS:
PHASEEQUILIBRIA

consideration (for example, equal to In 2 for the Soave-Redlich-Kwongequa-


tion of state). The relationship can also be written as:
a;,; g pE- t r n
!&.--- (2.55)
b ;=I I bi A

The choice of a model applied to g; therefore generates a mixing


~

rule.The mixing rules derived from the NRTL, UNIQUAC and UNIFAC models
can be defined in this way and the concept of group contributions can be uti-
lized for homogeneous models.
We will not go into detail on the research that followed this proposal, but
the work by Abdoul et al. (1991), Michelsen (1990), Dahl and Michelsen (1990),
Lermite and Vidal(1992), Wong and Sandler (1992), and Wong et al. (1992) can
be mentioned.

2.1.6 Calculation Algorithms


Generally speaking, the calculation of vapor-liquid equilibria is iterative. It
includes in particular the verification of equilibrium relations (equalities of
chemical potentials or of fugacities), which are usually complex expressions of
temperature, pressure and composition. We will not go into a detailed presen-
tation here of the algorithms that have been proposed, the reviews by
Heidemann (1983) and Michelsen (1992) can be consulted in this connection.
The application of heterogeneous methods does not normally present any
major difficulties.Examples of algorithms have been published by Renon et al.
(1971) and Prausnitz et al. (1980).
Looking at the application of a homogeneous method, we will review first
of all the calculation principle and the relations that are associated with the
model and must simultaneously be verified.
At each step of the iterative processus we have values of temperature T, of
pressure P and of composition of the phases in equilibrium, xi and yi.These
values can be the problem data, the result of the previous iteration or of the
initialization.
The parameters of the mixture components can be calculated: a;,;, bi
(Eqs. 2.37, 2.39 and 2.41) along with the binary terms a;,, (Eq. 2.49) from the
temperature.
The mixing rules (2.47 and 2.48) are then applied to the liquid phase of
composition xi:
aL = C.
I
C.J a.Y.x.x.
I J

b L= Cibixi

and the equation of state is solved (Eq. 2.43):


RT aL
p=--
UL - bL (uL - bLr,)(oL - bLrJ
Chapter 2 THERMODYNAMICS: PHASE ~OUlLlBRlA 61

If this equation has three volume roots, then only the one with the lowest
value will be considered. If only one root is found, we must be sure that it really
corresponds to a liquid state.
The mixture fugacities are then calculated (Eq. 2.50):
P(uL-bL) b;
In cpf = -In + 7(ZL - 1)
RT b
aL 2x,ai,,xi bi In uL- bLrl
+
bLRT(rl- rd (- c)
7 u L - bLr2
The preceding steps are resumed to calculate the fugacity coefficients in
the vapor phase: application of mixing rules with the yi compositions:
xi
a V = xI. al,lYlYl
. .. .
bV= Cibiyi
solution of the equation of state:
RT av
p=--
u v - bV (uv - bvrJ (uv - bvra
If three volume roots are found, the one with the highest value is chosen; if
there is only one root, we must be sure it really corresponds to a vapor state.
The expression of fugacity coefficients is then:
P(uV-b") b;
In cpy = -In +- (ZV - 1)
RT bV
av 2zjai,jy, bi In u v - bvrl
+
bVRT(rl- r.J (7s) - 0"- bVr2

The equilibrium condition is finally written in the following form:


cp,"x;= cpyy; (2.56)
The series of operations seems to be simple. It is based mainly on the ini-
tialization which is made up of the missing variables and on the process that
links up the iterations.
In fact, the calculation of vapor-liquid equilibria under pressure involves a
number of difficulties due to:
the neighborhood of the critical region,
the presence of the retrograde condensation region (see Section 2.1.1.2),
the possible continuity between "liquid-liquid" equilibria and "vapor-
liquid" equilibria.
For example, a calculation may happen to result in the "trivial" solution,
according to which the two phases in equilibrium are identical, thereby mis-
takenly situating the system outside the region of two-phase coexistence. This
difficulty can be fairly well understood if the solution of the equation of state
for one phase or the other, or for both, generated only one root.
62 Chapter 2. THERMODYNAMICS:
PHASE ~QUlLl8RlA

The habitual numerical methods, called substitution methods, which are


widely applied in calculating vapor-liquid equilibria at low pressure can be
used in many cases. Each step is rapid, but convergence can, however, be very
slow.
Calculation algorithms have also been the subject of numerous studies. A
review has been made of them by Heidemann (1983) and more recently by
Michelsen (1992) and PCneloux (1992). We will not enter into detail about
these studies and will limit our comments to giving the principle of the most
widely used methods.

2.1.6.1 Newton’s Numerical Method


If the quantity of each of the components present in a system is known, then
all the thermodynamic properties of the system (volume, enthalpy, etc.) can
be determined, if the temperature, pressure, quantity (total number of moles)
of each of the existing phases and composition (mole fractions) of these
phases are also determined. For a twephase state, there are 2n + 4 variables
(see Section 2.1.2.2). In addition, for each component there are material bal-
ance equation and equilibrium conditions (Eq. 2.56). The sum of the mole frac-
tions in each of the phases is equal to one, i.e. 2n + 2 equations. The problem
is therefore defined as soon as two variables are fixed as data, to be chosen
among those that have been listed or among the other thermodynamic p r o p
erties (total volume of both the phases, enthalpy, etc.). For example if these
data are the temperature, T,,, and the total volume vda&
T = Gat, (2.57)
and
NLuL+ NVuV= vd&a (2.58)
where uL and uv are the values of the molar volume in the liquid and vapor
phase, which can be calculated by applying the equation of state to the two
phases.
These two constraint equations thus complete the problem, consisting of a
system of 2n + 4 (non-linear) equations with 2n + 4 unknowns. Asselineau et al.
(1979) proposed applying Newton’s method of solution. In this connection, it
is necessary to know the Jacobian made up of the partial derivatives of each
of the functions defined by the equations listed above in relation to the vari-
ables, and particularly of the fugacities in relation to temperature, pressure
and composition. From one type of problem to the other, depending on the
problem data, only the last two lines of the system change (Eqs. 2.57and 2.58).
The resulting system can of course be simplified in many instances.
It is a well known fact that Newton’s method is characterized by a high
convergence speed, but also by particular fragility with respect to initializa-
tion. The expression of K values proposed by Wilson is usually applied for
initialization:
Pc,i
In Ki= In - + 5.373 (1 + mi) (2.59)
P
Chapter 2. THERMODYNAMICS: PHASE ~QUlLlERlA 63

The calculation generally converges in a satisfactory way.


It is important to note that the equality of fugacities corresponds to a
extremum of Gibbs energy, whereas the stability condition demands a mini-
mum of the function. The calculation may result in a false solution. More
recent and more efficient algorithms have been proposed based on the princi-
ple of what is called the “tangent plane”.

2.1.6.2 Stability of a Mixture: the Tangent Plane Criterion


The method termed ”tangent plane” was proposed by Michelsen (1982a,
1982b) to verify the stability of a mixture. At specified temperature and pres-
sure, Gibbs energy defines a surface g(z) depending on the compositions.
Given a mixture with composition zi, Gibbs energy corresponding to it must be
minimal according to the second principle. There is a plane Po tangent to the
surface g(z) for this composition. There may be other planes tangent to the
surface g(z) and parallel to Po. If one of them is located at a “height”lower than
Po with respect to the Gibbs energy scale, then the mixture of composition zi
is unstable. If there is one that is the same as Po, then the corresponding mix-
ture is in equilibrium with the mixture of composition zi.
Figure 2.9 concerning the ethane/propane system at 40°C and 25 bar helps
better understand the method. The variation in the Gibbs energy of the sys-
tem, calculated by means of the Soave-Redlich-Kwong equation of state, has
been plotted versus the composition. The values corresponding to pure sub-
stances have been chosen as the origin of the Gibbs energy in the same con-
ditions of temperature and pressure. For compositions ranging from 22 to 65%
ethane, the equation of state has two roots corresponding to the liquid and
vapor states, to which two values of Gibbs energy correspond, represented by
the curves gLand gv respectively. If we consider a liquid mixture of composi-
tion x1 = 0.4, and for this composition draw the tangent to the curve gL,it can
be seen that curve gv has a parallel tangent for a composition of approxi-
mately y1 = 0.6 and that this tangent is located above the first one. The liquid
mixture of composition 0.4 is therefore not stable. It can also be noted that for
compositions 0.34 and 0.53, the tangents to curves gL and gv are the same:
there is a vapor-liquid equilibrium.
It may be desirable to proceed to a stability test before any equilibrium cal-
culation. This will avoid making useless calculations if the mixture is stable
and more important will provide an excellent initialization for the equilibrium
calculation if the mixture is identified as unstable.

2.2 Liquid#Liquid Equilibria


Partial miscibility in the liquid phase is a commonly observed fact, stemming
from molecular interactions which are exerted among the components in a
dense phase. The differences in polarity play an important part in mixtures
between hydrocarbons and solvents such as acetonitrile, dimethylformamide
64 Chapter 2. THERMODYNAMICS:PHASE fQUlLl6RlA

-0.25
h
9
I
0,

-0.5

V
Y
-0.75 I I I I
0 0.2 0.4 0.6 0.8 1
Mole fraction of ethane
-
Figure
2.9 Determination of the phase equilibrium by the "tangent plane" method.
Mixture of ethane and propane 0 = 40°C; P = 25 bac gM: Cibbs energy per mole

and dimethylsulfoxide for example. Self-association by means of hydrogen


bonds must also be mentioned. It is responsible for the almost total immisci-
bility of water and hydrocarbons at near ordinary temperature and for the par-
tial miscibility of methanol with saturated hydrocarbons.
Liquid-liquid equilibria are selective: in a ternary mixture that separates
out into two liquid phases, the distribution of each component bears witness
to the molecular interactions that we have just mentioned. Liquid-liquid
extraction processes utilize this selectivity, already illustrated by the values of
the activity coefficients at infinite dilution of hexane and benzene in polar sol-
vents listed earlier (Table 2.4).
Liquid-liquid equilibria can of course be associated with vapor-liquid equi-
libria and the implementation of azeotropic distillation processes is based on
these two types of equilibria between phases.
A relatively recent review (Sprrensen, 1979), has examined liquid-liquid
equilibria from the standpoint of bibliographical data and presented the main
calculation methods.
Chapter 2. THERMODYNAMICS: PHASE EOUlLlERlA 65

2.2.1 General Description

2.2.1.1 Binary Systems

Partial miscibility in the liquid phase is often highly sensitive to temperature.


Figures 2.10A to 2.10C show the variation in mutual solubility on the abscissas
versus temperature on the ordinates for a specified pressure. A diagram like
the one in Figure 2.10A is the most frequent; it concerns the aniline (l)/hexane
(2) system. At 25°C a system containing 25% aniline separates out into two
phases. One of them (that we will call the "raffinate") is rich in hexane
(xl = 0.08) and the other (the "extract") is rich in aniline (xl = 0.90). The pro-
portions of the two phases are obtained from the material balance which mate-
rializes the rule of levers. It can be seen that if the temperature rises, the
composition of the phases in equilibrium varies. At approximately 57°C the
composition of the raffinate reaches the value of 25% and the heterogenous
mixture considered above becomes homogeneous. The composition of the
extract and the raffinate are the same at the upper critical solution tempera-
ture (UCST), here approximately 60°C. It is then a "critical point" similar to the
vapor-liquid critical points. The composition is fixed, usually close to the
equimolar mixture.
The lines representing the variation in composition of the phases in equi-
librium versus temperature delimit the two-phase domain. Any system whose
representative point is inside this domain is heterogeneous.
An increase in temperature may be associated with a decrease in mutual
solubility (Fig. 2.10B). This is the case for example of the water/dipropylamine
system and quite often of polymer solutions. The system then exhibits a lower
critical solution temperature (LCST). The two-phase area can also be closed
(Fig. 2.10C) as in the case of the tetrahydrofuran/water system.
Liquid-liquid equilibria are usually not very sensitive to pressure as shown
by the equilibrium relation.

2.2.1.2 Ternary Systems


Figure 2.11A and Table 2.12 represent the liquid-liquid equilibria of the
n-hexane/benzene/dimethylsulfoxidesystem at 20°C and atmospheric pres-
sure. Hexane and benzene are soluble in any proportion and the same is true
for benzene and dimethylsulfoxide. In contrast, hexane and dimethylsulfoxide
exhibit a huge miscibility gap. A mixture containing for example 27.5% hexane,
14.8%benzene and 57.7%dimethylsulfoxide by weight (point M> separates out
into two phases whose representative points are at the ends of the equilib-
rium straight line that passes through M, and whose composition is provided
by the third line in Table 2.12. If the overall composition of the mixture is var-
ied, the representative points of the phases in equilibrium will describe the
two branches (extract and raffinate) of the two-phase envelope. The two
branches will come together at the critical point. Such a diagram is said to be
"closed".
66 Chapter 2. THERMODYNAMICS: PHASE EQUILIBRIA
0 0
0 8 t N
(30)0
(I)
.-
-C
.-
I
I
- us-
I %
I W-
I %
I
I
I
-
I
(I)
0 m o I
t C U N
Chapter 2 THERMODYNAMICS: PHASE ~QUlLlSRlA 67

Benzene

Hexane DMSO

Hexane Glycol

Figure
2.11 Examples of ternary liquid-liquidequilibrium diagrams.
A. “Closed”system. B. “Open’’system.
68 Chapter 2. PHASEEOUILIBRIA
THERMODYNAMICS:

Raffinate Extract
lquilibrium
line
n-Hexane Benzene DMSO n-Hexane Benzene DMSO
~

1 92.2 7.6 0.2 3.8 4 92.2


2 85.5 13.9 0.6 4 8 88
3 80.3 18.7 1 4.8 13.1 82.1
4 74 24.9 1.1 5.3 18.3 76.4
5 68.5 30 1.5 5.9 22.4 71.7
6 62 36 2 7 27.4 65.6
7 56 41.2 2.8 8.8 33 58.2
8 48 46.8 5.2 10.5 38.1 51.4
9 40 50 10 13.6 42.7 43.7
10 33 51 16 16.3 45.3 38.4

Liquid-Liquid equilibrium of the n-hexane/benzene/dimethylsulfoxidesystem


at 20°C. Compositions are given in wt %.

The separation selectivity can be evaluated by the ratio of hexane (1) and
benzene (2) composition in the two phases:
w?/w! 0.803/0.187
-= = 11.7
wk/w 9 0.048/0.131
If two of the binary systems comprising the ternary system under consid-
eration exhibit partial miscibility, then the diagram looks like the one in
Figure 2.1 1B (hexane/benzene/ethylene glycol system). The diagram is said to
be “open”.

2.2.1.3 LfquideLiquidWapor Equilibria


We will comment on the isobaric liquid-liquid-vapor equilibrium diagrams
(Figs. 2.12A and 2.12B) in this section.
The way this type of diagram looks depends on both the amplitude of devi-
ations from ideality and the difference in vapor pressure between the two com-
ponents.
If the difference is not too large, the diagram will look like Figure 2.12A. The
border of the liquid-liquid immiscibility domain (dotted line) can be seen.
The condensation process of a mixture will go through different stages
depending on the overall composition.
Beginning with a totally vapor mixture (point MI)at constant pressure, the
temperature is lowered to the dew point (point A). The liquid that appears is
usually homogeneous and its representative point, B, is located in the total
miscibility area. Its composition is usually different from that of the vapor
(here x1 < y1 a s the difference in boiling point of the two components would
Chapter 2. THERMODYNAMICS: PHASE ~OlJlLlBRlA 69

XE, X R , Y
- Figure
2.12 Liquid-liquid-vaporequilibria at constant pressure.
A. Small difference in volatility between the two components.

suggest). The system is bivariant, as condensation proceeds, the temperature


decreases, the composition of the liquid and vapor phases varies, the repre-
sentative points describing two arcs of the bubble and dew point curves, BR
and AH respectively. The description of the change in phase remains identical
to the one for an entirely miscible system in the liquid phase.
However, when the composition of the liquid and the temperature are such
that the solubility limit 9 has been reached (intersection of the liquid-liquid
equilibrium curve and the bubble point curve, point R), a second liquid phase
appears of composition 9 (point E). The system is then monovariant and,
since the pressure is fixed, the temperature and the composition of the vapor
phase and of the two liquid phases will remain constant and equal to the Val-
ues reached when the second liquid phase appeared (figurative points R, E for
the two liquid phases, H for the vapor phase). Condensation continues with
the two phases deposited, the proportion of the phase of composition 9
increasing, so that the point representing the overall composition of the liquid
phases moves from R toward E. When the overall composition reaches a value
identical to that of the initial system (abscissa of point MI)condensation is
ended and the last trace of vapor is still represented by point H.
The condensation of a system represented by point M, would go through
similar stages. However, in the bivariant domain an inversion of volatility
would be noted, since the liquid phase is richer than the vapor phase in the
volatile component.
If the initial system (point M2) had a composition equal to that character-
izing point H, condensation would have immediately allowed two liquid
phases to appear of composition xE and $. Their overall composition would
be equal to that of the vapor and their proportions equal to the ratio of
segments HR and HE. Condensation would have occurred without any varia-
tion in temperature or composition. This particular point, H, is termed
"heteroazeotrope".
70 Chapter 2. THERMODYNAMICS: PHASE ~QUlLISRlA

Finally, in contrast condensation of systems represented by points M, or


M, would occur without three-phase equilibrium, but with an inversion of
volatility for the mixture rich in the volatile component.
If the volatility of the two components is very different, there is no het-
eroazeotrope (Fig. 2.12B). During the change in phase, the three-phase equi-
librium will still be found (points R, E, and H). Condensation of a system
represented by point M will occur in three stages: the first, bivariant, is repre-
sented by arcs BR (bubble point curve) and AH (dew point curve), then a
monovariant stage (points R, E, and H), and finally a last bivariant stage cor-
responding to arcs ED (bubble point curve) and HC (dew point curve).
A variation of the last diagram can be found when an azeotrope is present
in one of the total miscibility areas.

2.2.2Calculation of Liquid4.iquid
and Liquid4iquidWapor Equilibria
The same as for vapor-liquid equilibria, the equilibrium condition is expressed
by the equality of chemical potentials in all the phases present and for all the
components. Using once again the terms raffinate and extract for the two liq-
uid phases, the following can be written:
p: = py (2.60)
p? = py = pLy (2.61)

or in terms of fugacity:
f?= f? (2.62)
f?= fy =fy (2.63)
depending on whether the equilibria are two- or three-phase. The calculation
of fugacities will generally use a heterogeneous method, since the mixing rules
(at least in their classical form) applied to the equations of state can not be
employed with the systems involved in liquid-liquid equilibria due to their lim-
ited flexibility and the polarity of some of the components.
In these conditions and at low pressure, Eq. 2.1 1 that is used to calculate
fugacities in the liquid phase can be simplified:
f ? =pyx; y y f ? = pyx7 y y
and the liquid-liquid equilibrium condition is written:
x ; y y = xy y y (2.64)
It is thus possible to link up the "solvent capacity", the "partition coeffi-
cient" and the "selectivity" with the activity coefficients yf.
By "solvent capacity" we mean the solubility of a representative compound
of a given family in the solvents under consideration, for example hexane in
polar solvents. Considering Eq. 2.64 above, and designating hexane by the
Chapter 2. THERMODYNAMICS: PHASE ~OUILIBRIA 71

subscript 1, the solvent by the subscript 2, the hexane-rich phase by the expo-
nent R and the solvent-rich phase by the exponent E, the solubility of hexane
can be expressed as follows:

If in addition, the mutual solubility of hexane and of the solvent is low, then
the activity of hexane (equal to the product x F Y ~ in ~ )the hexane-rich phase
is very close to 1 (like the activity of the solvent in the solvent-rich phase) and
the following can be written:
1
XPE;: -
YVE
In the literature there are numerous data on activity coefficients at infinite
dilution, which correspond to the inverse of the solvent capacity when their
value is high.
The "partition coefficient" of a compound corresponds to the ratio of the
composition in one phase to that in the other phase. It is thus equal to the
ratio of the activity coefficients:
_
x; -
- -yb"
x; y y
Finally the "selectivity" of a solvent with respect to two solutes is equal to
the ratio of the partition coefficients, or to the ratio of the concentration of the
two solutes in the extract phase and in the raffinate phase:

Since the raffinate phase is made up basically of the two solutes and in this
phase their activity coefficients are close to one, the following can be written:
y y
ai,j = -
y;E
It is, however, only a convenient approximation, since in the case of an aro-
maticlparaffin separation for example, the hydrocarbon mixtures are nowhere
near ideal (see Table 2.3). It can be used along with the activity coefficients at
infinite dilution for a paraffin and an aromatic in polar solvents to establish a
selectivity scale that can help in choosing solvents.
The calculation of liquid-liquid equilibria is based on the estimation of
activity coefficients. In order to predict or correlate these coefficients, the
models presented earlier (NRTL, UNIQUAC, UNIFAC) will need to be used.
Theoretically, the models should allow the calculation of two- or three-phase
equilibria with the same set of parameters. In actual fact this is seldom the
case. For example the UNIFAC model has been proposed for these calculations,
but only after modification of the parameter values (Magnussen et al., 1981).
These difficulties can be explained by the imperfections in the models and the
72 Chapter 2. THERMODYNAMICS: PHASE EQUlLlBRlA

extreme sensitivity of liquid-liquid equilibria with respect to the values of


activity coefficients or the excess Gibbs energy. Figure 2.13 illustrates this sen-
sitivity. The upper part represents the variation in excess Gibbs energy versus
composition. Two models have been applied which lead to results that are
only slightly different. To calculate the Gibbs energy of mixing, the ideal mix-
ture term must be added:
g M =gE +RT (xl In x1 +x2 In x&
The two curves in the lower part of the figure are obtained. One of them
exhibits two minima and it is possible to have a large immiscibility domain cor-
respond to it by plotting a straight line tangent to the curve in two points, in
the same way as for the vapor-liquid equilibria (Fig. 2.9). As a result, these two
models will lead to opposite conclusions concerning miscibility in the liquid
phase.
Simultaneous calculation of liquid-liquid and vapor-liquid equilibria by the
same model with the same parameters must be considered an open question.

1
-
0.8-
-
0.6 -

t-
%b 0.4 -

0.2 -

0-

-0.2 I I I
0 0.25 0.5 0.75 1
Mole fraction of propane

Figure
2.13 Sensitivity of liquid-liquid equilibria to the values of excess Gibbs energy.
Propane/methanol system T = 313.15 K, P = 13 bar.
Chapter 2 THERMODYNAMICS:
PHASE fOUlLSRlA 73

In any case, liquid-liquid equilibria can usually be estimated only if the data
base that serves as a basis for the relevant correlation contains liquid-liquid
data.

2.3 Solid#LiquidEquilibria

2.3.1 General Description


The oil industry is not the same as metallurgy and the solid phases that are
found are not as well known as metal alloys. A few practical examples can,
however, be mentioned.
The formation of paraffin deposits by cooling crude oils or diesel oils has
practical consequences on the transportation of these fluids. Some of the
characteristics of this problem have been addressed by Zhou (1991), who will
serve as a reference for much of the development that follows.
Likewise, formulation of lube oils requires dewaxing which will be
described later on in this volume (see Chapter 9).
Purification of paraxylene, a precursor of many synthetic textiles, can not
be done by distillation due to the small differences in volatility of the isomers
present in the aromatic C8 cut. It can not be done by liquid-liquid extraction
either because the solvents are not sufficiently selective with respect to the
isomers. In contrast, the large difference in melting point between paraxylene
(0, = 14°C) and metaxylene (0, = -47"C), for example can lead to a separation
process by crystallization.
The formation of a pure substance in the solid phase often characterizes
solid-liquid equilibria. Figure 2.14a represents the variation versus tempera-
ture of a binary liquid mixture at constant pressure. Depending on its compo-
sition it will form a solid phase by crystallization, which comprises one of the
two components. The concentration in the liquid phase of the component that
is deposited therefore decreases and for crystallization to proceed, it requires
lower and lower temperatures. Depending on the initial composition, one of
the two arcs represented is described. In both cases, the composition of the
liquid phase reaches a value corresponding to the intersection of these two
arcs and crystallization then proceeds by simultaneous deposit of the two
solid phases. Since the system is monovariant, the temperature remains con-
stant: the "eutectic" point has been reached. This is the case of the metaxy-
lene/paraxylene system.
For a given compound it is frequent for the solid phase to exhibit different
crystallographic forms according to the temperature. For example hexatria-
contane (C,,H,J melts at 76.2"C, but has solid-solid transition points at 73.9"C
and 72.2"C. The equilibrium diagram bears witness to these transitions by
breaks in the slope of the arc representing crystallization, as shown in
Figure 2.14b. We have confined the figure to showing two crystalline forms so
that it would be clearer.
-l
Liquid
7 Liquid

Solid 2A
P

Liquid J

Solid 1 Solid 26
I

Sohd 1 + Solid 2 Soh$ 1 +Solid 2B


I I
I I I

0 x, (eutectic) x2 x, (eutectic) 5
a. Simple eutectic system. b. System with 1 eutectic and 1 crystalline transition point.

't

Solid solution

0 x2

c. System forming a solid solution. d. System with 1 eutectic and 2 solid solutions.
Chapter 2. THERMODYNAMICS:
PHASE EOUILIBRIA 75

The compounds under consideration may form a solid solution whatever


their proportions. The diagram then shows a two phase equilibrium region
(Fig. 2.14c), very similar to the one observed for vapor-liquid equilibria
(Fig. 2.1). This occurrence of course requires compatibility of crystalline forms
specific to the two compounds, which is found in the case of naphthalene/
anthracene mixtures and for some paraffin mixtures. For example, tetracosane
(C24H5J leads to a simple eutectic point with heptadecane (C,,H,a and the dia-
gram is similar to the one in Figure 2.14a. Meanwhile with eicosane (C29H42) it
forms a solid solution and the crystallization diagram exhibits a two phase
shape (Fig. 2.14~).
Finally, if the solubility of the solid phase is limited to a certain concentra-
tion domain, a diagram like the one in Figure 2.14d will be observed.
Depending on the initial composition in the liquid phase, crystallization starts
by the deposit of one of the solid solutions S, or S,, and finishes at a eutectic
point where two immiscible solid solutions are deposited at the same time.
When crystallization ends a decrease in temperature may cause a variation in
composition of the two solid solutions in equilibrium. This type of diagram is
absolutely analogous to the liquid-liquid-vapor equilibrium diagram in
Figure 2.12A.
Solid-liquid equilibrium diagrams can take on a number of other shapes, in
particular due to formation by association of definite compounds exhibiting
“congruent“ melting points.
An important point must be made: in equilibria involving the appearance of
a solid phase, “equilibrium delays” are frequent (supercooling, delayed equi-
libria between allotropic forms) and experimental data may in many cases be
reported for metastable states.

2.3.2 Calculation of Solid0Liquid Equilibria


If the mixture contains only low vapor pressure compounds, the pressure has
practically no influence on the solid-liquid equilibrium. The influence could be
assessed by means of Poynting corrections applied to one of the two phases
and would turn out to be very small after calculation for pressure variations of
several tens of bars. We will therefore disregard the pressure variable in our
discussion.
We will first of all establish the equilibrium relation in the simplest case:
when the crystallized phase is made up of a pure substance (Fig. 2.14a).
With the subscript i used for the component that crystallizes, fTvs and f T s L
for the fugacity values in the pure state in the solid and liquid phases respec-
tively, xiand y; for the mole fraction and the activity coefficient in the liquid
phase, we will first of all write the equality of fugacities for this component in
the two phases:
f?” = p x ;y; (2.65)
For the pure substance at temperature T,the ratio fT.L/fTi*” is related to the
variation in Gibbs energy accompanying the transition from the solid to the
76 Chapter 2. THERMODYNAMICS: PHASE ~OUlLlBRlA

liquid state Ag+(S + L) in such a way that we can write the preceding relation
in the following way:

(2.66)

The variation in Gibbs energy is not known, but can be assessed depending
on its value at the melting point, T,, of the component, Ag;;i(S -+ L), which is
zero since there is equilibrium between the two states (solid and liquid) of the
pure substance at this temperature:
")
4 + ( S + L) -
RT
- 4 ; p + L)
RTI
+ 1; d(AB;;T+

d(;)
d(2
To calculate the derivative of Cibbs energy in relation to the temperature
we apply the Gibbs-Helmholtz equation and get:
d(AB;;T+ ")
As;@ + L) - d (+)=[ Ah '(:+ L, d ( + ) (2.67)
RT
-1; d(+)
Ah $(S -+ L) is the variation in enthalpy accompanying the transition from
the solid to the liquid state for the pure substance at temperature T. It
depends on the temperature by means of the heat capacities of the two phases
whose difference is indicated here by AC;(S -+ L). We will assume that this
term is independent of the temperature and equal to its value at the melting
point TI.The following is therefore obtained:
Ah;(S+L)=Ah;f(S+L)+AC$(S+L)(T-Tf)
and Eq. 2.67 can be explicated which becomes:
Ag;(S + L) - + L) T, AC;(S -+ L) Tf
RT - RT, (7-l)' R (In- T + 1 - T 3)
Finally, equilibrium Eq. 2.66 is written:
Aht(S+L) TI
-In (xiyk) = L, (In 2 + 1 - 2)(2.68)
RTI (7- 1) + R
+
T T
It involves the heat of fusion and the difference in heat capacities for the
compound that crystallizes. The latter term is often disregarded, giving:

(2.69)

This is the form used when the molecular weight of a solute is determined
by cryoscopy. A dilute solution of the solute is prepared with a known mass
composition. The preceding equation is applied to the solvent, whose
Chapter 2. THERMODYNAMICS:
PHASE EQUILIBRIA 77

decrease in melting point is measured. The enthalpy term in this equation is


calculated either from published data or by prior calibration. It is assumed
that since the solute is dilute, the activity coefficient of the solvent is equal to
one, so that the molar composition xi can be calculated. By comparison with
the mass composition, the molecular weight of solute is deduced.
Of course if the solid phase is present in different crystalline forms
(Fig. 2.14b) according to the temperature, then the calculations must be bro-
ken down into several steps, separated by transition temperatures while at the
same time still based on the preceding reasoning.
Generally speaking deviations from ideality in the liquid phase must be
taken into account; they are represented here by the activity coefficient y.;
They depend on the temperature and composition and can be evaluated by
means of the models that we have already presented (see Section 2.1.3.3). It
should be noted, however, that for a given binary mixture, the crystallization
equilibria and vapor-liquid equilibria do not usually concern the same tem-
perature range. This means that using activity coefficients acquired by vapor-
liquid equilibrium measurements in order to calculate solid-liquid equilibria
usually leads to temperature extrapolation. For example the binary
methanol/water system has often been studied above 50°C, but until recently
(Jose, 1997) there were no data available that could be used for a reliable
assessment of the activity coefficient in the crystallization region.
The influence of the non-ideality model in the liquid phase can be impor-
tant. Figure 2.15 can be used to compare the experimental values of crystal-
lization temperatures of benzene in n-undecane/benzene mixtures and the
diagrams calculated according to three assumptions: ideal solution, applica-
tion of the UNIFAC model and finally the method proposed by Abdoul et al.
(199 1), mentioned earlier with respect to calculating vapor-liquid equilibria
under pressure.
When there is a solid solution, the equilibrium equation is written:
(2.70)
f?, x i , and yi stand for the fugacities of the component in the pure state, the
mole fractions and the activity coefficients respectively, and the exponents L
and S give the physical state, liquid or solid. The quotient can be eval-
f T P L / f f v S

uated, as discussed earlier, by means of the heat of fusion and the differences
in heat capacities as previously. The activity coefficients in the liquid phase
will be calculated by means of one of the models covered earlier. In contrast,
there are no data concerning the deviations from ideality in the solid phase
and there is no reason for the ideality hypothesis to be applied, as shown by
the fact that solubility in the solid phase is usually limited and often very
small. There is no reason for the models proposed to calculate the deviations
from ideality to be applied to the solid phase either. Factors such as the nature
of the crystal lattice, its geometrical parameters can then take on a significant
influence. It has, however, been proposed to transpose the concept of regular
solutions to the solid phase in order to predict the paraffin crystallization
equilibria (Won, 1989).
78 Chapter 2. THERMODYNAMICS:
PHASE hUlLlSRlA

--
Figure
2.15 Solid-liquid equilibrium diagram For the n-undecane/beruene system
(Zhou, 1991).

It should be noted that we have only dealt with the simplest cases here:
predicting solid-liquid equilibria remains a difficult problem. In the presence of
light components, it would be advantageous to solve the liquid-solid-vapor
equilibrium problem for example, as well as to characterize petroleum frac-
tions, crude oils or natural gases by considering more particularly the param-
eters (melting point, heat of fusion) specific to this problem. This means that
research is still likely to be developed in the field.

2.4 Complex Mixtures


2.4.1 Lumping
It is impossible to proceed to a perfect analysis of oil and gas fluids (natural
gas, crude oil). If it were possible, the number of components identified would
be such that the methods we have presented could not be applied due to the
processing time and computer memory that would be necessary. This diffi-
culty also crops up for narrower fractions (gasolines, kerosines, diesel oils).
As a result, it is necessary to operate by lumping, so as to represent these
complex mixtures by a small number of pseudocomponents. Lumping is usu-
ally done on the basis of TBP distillation (Paradowski, Vol. 1, Chapter 4) or a
detailed chromatographic analysis (simulated distillation) and algorithms that
achieve lumping have been described (Monte1 and Gouel, 1984). Then the
parameters have to be calculated for the models used (critical coordinates, for
Chapter 2. THERMODYNAMICS: PHASE ~OUlLIBRlA 79

example) according to the data available concerning these pseudocompo-


nents, such as normal boiling point temperature and density. To do this, the
equations proposed by Paradowski (Vol. 1, Chapter 4. Section 4.1) can be used.
This procedure is widely used to calculate distillations. It has been applied
very seldom to mixtures that are significantly non-ideal due to the presence of
polar components or to calculate liquid-liquid equilibria. It might lead in this
case to questionable results because of the sensitivity of this type of equilib-
rium to the calculation of activity coefficients, as mentioned earlier.

2.4.2 Continuous Thermodynamics


Continuous thermodynamics is, or at least seems to be, a new approach. The
complex composition of the fluid is represented by means of a continuous
Gaussian distribution function, or more frequently by a gamma function. The
variable can be molecular weight or boiling point and in some cases two-
dimensional distribution can be used (boiling point temperature and density
for example) in order to represent the aromaticity of the fluid. The parameters
of these distribution functions will be evaluated from TBP distillation or chro-
matographic analysis and the properties required to calculate the equilibrium
between phases will be linked up to the distribution variables by empirical
correlations.
This approach is naturally suitable for representing solutions of polymers,
but it has been extended to petroleum fluids (Cottermann et al., 1985). Even
though it is attractive due to the dramatic reduction in the number of param-
eters and therefore in the processing time, it still has a number of drawbacks.
First of all this type of representation is an approximation that may be
unacceptable. Some of the major components of the fluid can be individual-
ized and the term of “semicontinuous” thermodynamics will then be used.
The number of variables characteristic of the fluid will as a result be higher.
Additionally, if a fluid represented by a gamma function for example is par-
tially vaporized, then the two phases in equilibrium can not usually be repre-
sented by distribution functions of the same type as the one assigned to the
fluid. The decision is then made to use the distribution function as a method
of generating pseudocomponents which, according to the method’s propo-
nents, will reduce the number of pseudocomponents and give a more rational
definition.

2.5 Conclusion
As we have said before, experimental data and “models”are used as a basis to
get equilibria between phases. All of the thermodynamic properties are
involved, but we have more particularly emphasized the ones derived from
Gibbs energy: chemical potential and fugacity. However, the energy balance of
a unit should not be forgotten, it is often decisive and involves enthalpy and
80 Chapter 2 THERMODYNAMICS:
PHASE ~CJUILIBRIA

sometimes entropy calculations. Additionally, in a separation unit the streams


of material are often "isenthalpic". The capacity to express all of the proper-
ties of a system at the same time has unquestionably favored the present
development of equations of state. This type of model must, however, be used
only with an awareness of their limits and attention to recent progress. We
have mentioned these developments, which have not yet come into common
practice, only rapidly.
Calculations of vapor-liquid equilibria for mixtures containing polar com-
pounds, and of liquid-liquid and solid-liquid equilibria are complex problems
that can be carried out properly only if experimental data and elaborate mod-
els are available and if the essential features of the relevant separation process
are borne in mind. Process features determine the possible approximations
and the points that require the most accuracy. As a result, the collaboration of
a process engineer, a thermodynamics expert and a laboratory specialist is
necessary.

You might also like