9.

1 1 Choosing the Appropriate Thermodynamic Model 473

based on a fixed amount of one kilogram of solvent. It is for this reason that molality
has been used here, and again in Chapter 15. Note that if the solvent is water, and the
solution is dilute in solute (so that one liter of solution contains one kilogram of water),
molalip and molarity are equivalent.

9.11 CHOOSING THE APPROPRIATE THERM0DYK;IMIC MODEL

The objective of this chapter has been to develop methods of estimating species fugaciries in mixtures. These methods are very important in phase equilibrium calculations,
as will be seen in the following chapters. Because of the variety of methods discussed,
there may be some confusion as to which fugacity estimation technique applies in a
given siruation. The comments that follow may be helpful in choosing among the three
main methods,&scussed in this chapter:
1. Equations of state
2. Activity coefficient (or excess Gibbs energy) models
3. Activity coefficient (or excess Gibbs energy) models based on the Henry's law
standard state
Electrolyte solutions are special and can only be treated by the methods considered
in Sec. 9.10. Therefore, electrolyte solutions are not be considered in this discussion.
Also, the Henry's law standard state is used only for a component that does not zxist
as a pure component at the temperature of interest-for example, a dissolved solid
below its melting point, or a dissolved gas much above its critical point. (However, if
the liquid mixture can be described by an equation of state-for example, mixtures of
hydrocarbonsand nitrogen or carbon dioxide-there is no need to use the Henry's law
standard state.)
In low- to moderate-density vapors, mixture nonidealities are not very large, and
therefore equations of state of the type discussed in this text can generally be used
for the prediction of vapor-phase fugacities of all species. [However, mixtures containing species that associate (i.e., form dimers, [rimers, etc.) in tine vapor phase, such as
acetic acid, are generally described using the virial equation of state with experimentally determined virial coefficients.] The Lewis-Randall rule should be used only for
approximate calculations; it is best to use an equation of state to calculate the vaporphase fugacity of vapor mixtures.
At liquid densities, solution nonidealities can be large. In this case, equation-of-state
predictions with the van der Waals one-fluid mixing rules will be reasonably accurztz only for mixtures composed of relatively simple molecules that are similar. Thus,
equations of state with the van der Waals one-fluid mixing rules of Eq. 9.4-8 will be
satisfactory only for mixtures involving light hydrocarbons, including hydrocarbons
with some dissolved inorganic gases (CO, C02, H2S, N2, etc.). This equation-of-state
method will not be accurate if complicated molecular phenomena are involved-for
example, hydrogen bonding, as occurs in mixtures containing water, alcohols, and organic acids---or if chemically dissimilar molecular species are involved. In these cases,
activity coefficient models (alsp referred to as Gibbs energy models) must be used for
estimating the liquid-phase species fugacities, even though an equation-of-state model
may be used for the vapor phase. Alternatively, the same equation of state can be used
for both the liquid and the vapor if the mixing rules of Sec. 9.9 that combine activity
coefficient and equation-of-state models are used.

474 Chapter 3: Estimation of the Gibbs Energy and Fugacity of a Component in a Mixture
Traditionally, in the petroleum and natural gas industries, where mostly low- and
moderate-molecular-weight hydrocarbons (with only small concentrations of aromatic
and inorganic compounds) are involved, equation-of-state.meth0d.s with the van der
WaaIs one-fluid mixing r u l ~ are
j generally used for the prediction of both liquid- and
vapor-phase properties. In the chemicals industry, on the other hand, where oxygen,
nitrogen, sulfur, or halogen-containing organic compounds and inorganic (electrolyte)
compounds are involved, activity coefficient or Gibbs energy models are used for the
prediction of liquid-phase properties, with equation-of-state models used for the vapor
phase if the pressure is abo\-e ambient. However, in recent years the equation-of-state
methods of Sec. 9.9 have also been used to model both the vapor and liquid phases of
such mixtures.
The critical point of a mixture. to be discussed in the next chapter, is similar to
the critical point of a pure component in that it is the temperature and pressure at
which the wpbr and liquid phases of a mixture of given composition become identical.
Although \ye have not yet discussed phase equilibrium in multicomponent mixtures
(see Chapters 10 to 12), we can guess that if'we are to predict the point at which two
phases become identical (the critical point) with some accuracy, the models we use
for the'vapor and liquid phases must give identical results for all properties at this
point. This will occur only .if we use the same equation-of-state model to describe
both the vapor and liquid phases. but clearly cannot, in general, be expected to occur
if we use different models. such as an activity coefficient model for the liquid phase
and an equation of state for the vapor phase. This ability to predict critical phenomena
is an important advantage of using the same equation-of-state model for both phases.
Another advantage of the equation-of-state description is that the concept of standard
states, and especially hypothetical or extrapolated standard states, never arises. -.Finally, since a number of different activity coefficient (or excess Gibbs energy)
models have been discuskd. it is useful to consider their range of application. The
most important observation is that none of the completely predictive methods, such
as regular solution theory. UNIFAC, or ASOG, can be regarded as highly accurate.
Therefore, these methods shouId be used only when no experimental data are available
for the system of interest. Of the predictive methods, UNIFAC is the best developed
and regular solution theory is the least accurate. .
As we'have considered a number of activity coefficient models in this section, it
is useful to discuss which might be best for different types of mixtures. To start this
discussion, it is useful to classify seven different types of chemicals that one might
encounter.

Nonpolar compo~~nds:
Chemical species without significant dipole moments, such as
n ~ s hydrocarbons,
t
and some very symmetri.~compounds such as tetrachloromethane
and hexafluorobenzene.
Weaklypolar compounds: Aldehydes, ethers, and ketones
Very polar compounds: Alcohols, arnines
Water
Carboxylic acids
Polymers
Electrolyte systems
'

As a general rule, a mixture containing components from more than one of the
classes above exhibits greater nonidedity than a mixture containing only components
from the same class. For example, a mixture containing an aldehyde and a ketone

9.1 1 Choosing the Appropriate Thermodynamic model 4 75

probably has a lower excess Gibbs energy of mixing than an aldehyde-hydrocarbon
mixture. However, it is also true that even for a mixture containing compounds ivithin
one class, the mixture may b e slightly nonideal if the components are similar in size
and chemical functionality, or more nonideal (larger Gibbs energy of mixing) if there
is greater dissimilarity in the components. Table 9.1 1-1 suggests the most appropriate
activity coefficient (excess Gibbs energy) model o r models to use for different classes
of mixtures, and Table 9.1 1-2 lists the most appropriate equation-of-state model. These
tables should be used as references in choosing thermodynamic models-for example.
when using a process simulator.

Table 9.11-1 Recommended Activity Coefficient Models

nonpolar compounds: All of the models (Margules, van Laar, Wilson.

Nonpolar
UNIQUAC, and NRTL) will give good correlations of data for these mixtures.
Nonpolar + weakly polar compocnds: All of the models (Margules, van Laar, Wilson.
UNIQUAC, and NRTL) can be used, although the UNIQUAC model is better for mixtures that
are more nonideal.
Nonpolar +strongly polar compounds: Although all of the models (Margules, van Laar, Wilson, UNIQUAC. and NRTL) can be used for mixtures that are not too nonideal, the UNIQUAC
model appears to give the best correlation for somewhat nonideal systems, wheras the \Vilson
model may be better for mixtures that are more nonideal (but not so nonideal as to result in
liquid-liquid immiscibility; see Sec: 1 1.2).

Weakly polar weakly polar compounds: All of the models (Margules, van Laar, Wilson.
UNIQUAC, and NRTLj can be used, although it appears that the UNIQUAC model is better for
mixtures that are more nonideal.
Weakly polar +strongly polar compounds: All of the models (Margules, van Laar, Wilson.
UNIQUAC, and NRTL) can be used, although it appears that the UNIQUAC model is better for
mixtures that are more nonideal.

Strongly polar strongly polar compounds: The UNIQUAC model appears to best correlate
data, though all models give reasonable results.
.
,

Water nonpolar compounds: These mixtures generally have limited mutual solubility and
are discussed in Sec. 1 1.2.

Water weakly polar compounds: These mixtures generally have limited mutual solubility
and are discussed in Sec. 1 1.2.

Water strongly polar compounds: The UNIQUAC model appears to best correlate data for
aqueous mixturss.
Solutions containing carboxylic acids: The Wilson model appears to best correlate data for
mixtures containing carboxyli? adlhs if the components are mutually soluble. (The Wilson
model does not predict liquid-liquid phase splitting, as discussed in Sec. 11.2). Otherwise, the
UNIQUAC, van Laar, or NRTL model should be used.
Solutions containing polymers: Use the Flory-Huggins model.
Solutions containing ionizable safts, strong acids, o r bases: Use the extended Debye-Hiickel
model.
~Multicomponentmixtures: Choose the model that, based on the suggestions above, best describes the dominant components in the mixture.

476 Chapter 9: Estimation of the Gibbs Energy and Fugacity of a Component in a Mixture

Table 9.11-2 Recommended Equation-of-State Models

Vapor Mivrllres

1. Low pressures, no components that associate (such as carboxylic acids or HF): Assume the mixture is an ideal gas mixture.
(T, P, y ) = yi P
2. Low pressure for a mixture that contains a n associating component: Use the virial
equation of state, retaining only the second virial coefficient, and search for experimental
pure component and cross virial coefficient data for all components in the mixture.
3. Slightly elevated pressures: Use choice 1 or 2 above. Alternatively, use 4 below if [here
are no associating components, or 5 if associating components are present.
4. Elevated pressures for a vapor mixture that contains hydrocarbons, nitrogen, osygen,
carbon dioxide, and/or other inorganic gases (but not HF): Use an equation of state,
such as the Peng-Robinson or Soave-Redlich-Kwong equation with van der Waals m e fluid mixing rules (see Sec. 9.4).
5. Elevated pressures for a vapor mixture that contains one or more polar and/or associating compounds: Use an equation of state, such as the Peng-Robinson or SoaveRedlich-Kwong equation with the excess Gibbs energy-based mixing rules (see Sec. 9.9)
and the appropriate activity coefficient model (see Table 9.1 1-1).

fY

Liquid Mir~rrtres

Liquid mixtures at low pressure are generally described using the activity coefficient models as
described in Table 9.1 1- 1, and the behavior of a liquid mixture is generally not much affecred by
pressure, unless the pressure is very high. However, as we will see in Sec. 10.3, for phase equilibrium calculations at high presslires, especially as the critical point of a mixture is approached,
there are important advantages to using the same thermodynamic model for both phases. In such
cases the same equation-of-state model should be used for the vapor and liquid phases.

1. Liquid mixture a t el:vated pressure that contains only hydrocarbons, nitro,Oen. osygen, carbon dioxide, and/or other inorganic gases (but not HF) :Use equation of state,
such as Peng-Robinson or Soave-Redlich-Kwong with van der Waals one-fluid mixing
rules (see Sec. 9.4).
2. Liquid mixture at elevated pressure that contains that contains one o r more polar
and/or associating compounds. Use equation of state, such as Peng-Robinson or SoaveRedlich-Kwong with the excess Gibbs energy-based mixing rules (see See. 9.9) and the
appropriate activity coefficient model (see Table 9.1 1- 1).

Appendix A9.1 A StatisticaI Mechanical Interpretation of the Entropy of Mixing

in an Ideal Mixture
This section appears o n the CD that accompanies this text.

.
Appendix A9.2

Multicomponent Excess Gibbs Energy (Activity Coefficient) Models

T h e excess Gibbs energy models for binary mixtures discussed in Sec. 9.5 can be
extended to multicomponent mixtures. For example, the Wohl expansion of Eq. 9.5-8
can b e extended to ternary mixtures: