Thermodynamics of Separation Operations

After completing this topic, students should be able to:

● Explain phase equilibria in terms of Gibbs free energy, chemical potential, fugacity,
fugacity coefficient, activity, and activity coefficient.
● Understand the usefulness of equilibrium ratios (K-values and partition coefficients)
for liquid and vapor phases.
● Derive K-value expressions in terms of fugacity coefficients and activity coefficients.
● Estimate liquid-phase activity-coefficient correlations using Wilson, NRTL,
UNIQUAC, or UNIFAC

PHASE EQUILIBRIA
Many separations are determined by the extent to which the species are distributed among
two or more phases at equilibrium at a specified T and P. The distribution is determined by
application of the Gibbs free energy. For each phase in a multiphase, multicomponent system,
the total Gibbs free energy is
G = G(T; P; N1; N2; . . . ; NC) (2-2)
where Ni = moles of species i. At equilibrium, the total G for all phases is a minimum, and
methods for determining this are referred to as free-energy minimization techniques. Gibbs
free energy is also the starting point for the derivation of commonly used equations for phase
equilibria. From classical thermodynamics, the total differential of G is
(2-3)

where µi is the chemical potential or partial molar Gibbs free energy of species i. For a closed
system consisting of two or more phases in equilibrium, where each phase is an open system
capable of mass transfer with another phase,

(2-4)

where superscript (p) refers to each of N phases. Conservation of moles of species, in the
absence of chemical reaction, requires that

(2-5)
which upon substitution into (2-4), gives
(2-6)

With eliminated in (2-6), each term can be varied independently of any other term. But this
requires that each coefficient of in (2-6) be zero. Therefore,
µi(1) = µi(2) = µi(3) = …….. = µi(N) (2-7)
Thus, the chemical potential of a species in a multicomponent system is identical in all phases
at physical equilibrium.

Fugacities and Activity Coefficients

Chemical potential is not an absolute quantity, and the numerical values are difficult to
relate to more easily understood physical quantities. Furthermore, µ→∞ as P→0.Thus, the
chemical potential is not favoured for phase-equilibria calculations. Instead, fugacity,
invented
by G. N. Lewis in 1901, is employed as a surrogate.

1
The partial fugacity of species i in a mixture is like apseudo-pressure, defined in terms of the
chemical potential by
(2-8)
where C is a temperature-dependent constant. Regardless of the value of C, it is shown by
Prausnitz et al. (1999) that (2-7) can be replaced with
(2-9)
Thus, at equilibrium, a given species has the same partial fugacity in each phase. This
equality, together with equality of phase temperatures and pressures, T
T(1) = T(2) =T(3) =…….. =T(N) (2-10)

and P(1) = P(2) =P(3) =…….. = P(N) (2-11)

constitutes the conditions for phase equilibria. For a pure component, the partial fugacity, ,
becomes the pure component fugacity, fi. For a pure, ideal gas, fugacity equals the total
pressure, and for a component in an ideal-gas mixture, the partial fugacity equals its partial
pressure, pi = yiP. Because of the close relationship between fugacity and pressure, it is
convenient to define their ratio for a pure substance as
(2-12)
where Φi is the pure-species fugacity coefficient, which is 1.0 for an ideal gas. For a mixture,
partial fugacity coefficients are

such that as ideal-gas behaviour is approached, and , where

At a given temperature, the ratio of the partial fugacity of a component to its fugacity in a
standard state is termed the activity. If the standard state is selected as the pure species at
the same pressure and phase as the mixture, then
(2-15)
Since at phase equilibrium, the value of is the same for each phase, substitution of (2-15)
into (2-9) gives another alternative condition for phase equilibria,
ai(1) = ai(2) = ai(3) = …….. = ai(N) (2-16)

For an ideal solution, aiV = yi and aiL = xi.

To represent departure of activities from mole fractions when solutions are nonideal, activity
coefficients based on concentrations in mole fractions are defined by
(2-17)

(2-18)

For ideal solutions, γiV = 1.0 and γiL = 1.0

Table 2.1 Thermodynamic quantities for Phase Equilibria

2
K-Values
A phase-equilibrium ratio is the ratio of mole fractions of a species in two phases at
equilibrium. For vapour–liquid systems, the constant is referred to as the K-value or vapor–
liquid equilibrium ratio:

For the liquid–liquid case, the ratio is a distribution or partition coefficient, or liquid–liquid
equilibrium ratio:

For equilibrium-stage calculations, separation factors are defined by forming ratios of

equilibrium ratios. For the vapour–liquid case, relative volatility αi,j between components i
and j is given by

Separations are easy for very large values of αi,j, but become impractical for values close to
1.00. Similarly for the liquid–liquid case, the relative selectivity βi,j is

For vapour–liquid equilibrium, (2-9) becomes, for each component,

3
To form an equilibrium ratio, partial fugacities are commonly replaced by expressions
involving mole fractions. From the definitions in Table 2.1:

If (2-24) and (2-25) are used with (2-19), a so-called equation-of-state form of the K-value
follows:

Applications of (2-26) include the Starling modification of the Benedict, Webb, and Rubin
(B–W–R–S) equation of state (Starling, 1973), the Soave modification of the Redlich–Kwong
(S–R–Kor R–K–S) equation of state (Soave, 1972), the Peng–Robinson (P–R) equation of
state (Peng and Robson, 1976), and the Plöcker et al. modification of the Lee–Kesler
(L–K–P) equation of state (Plöcker et al., 1978).
If (2-23) and (2-25) are used, a so-called activity coefficient form of the K-value is obtained:

Since 1960, (2-27) has received some attention, with applications to industrial systems
presented by Chao and Seader (1961), with a modification by Grayson and Streed (1963).
Table 2.2 summarises formulations for vapor–liquid equilibrium K-values. Included are the
two rigorous expressions (2-26) and (2-27) from which the other approximate formulations
are derived.
The Raoult’s law or ideal K-value is obtained from (2-27) by substituting, from Table 2.1, for
an ideal gas and for ideal gas and liquid solutions, γiL = 1.0, , ΦiV = 1.0

The modified Raoult’s law relaxes the assumption of an ideal liquid by including the liquid-
phase activity coefficient. The Poynting-correction form for moderate pressures is obtained
by approximating the pure-component liquid fugacity coefficient in (2-27) by

where the exponential term is the Poynting correction. If the liquid molar volume is
reasonably constant over the pressure range, the integral in (2-28) becomes yiL (P - Psi).

4
For a light gas species, whose critical temperature is less than the system temperature, the
Henry’s law form for the K-value is convenient, provided Hi, the Henry’s law coefficient,
is available. This constant depends on composition, temperature, and pressure. Included in
Table 2.2 are recommendations for the application of each of the vapour–liquid K-value
expressions.

Table 2.2: Useful Expressions for Estimating K-values for Vapour-Liquid Equilibria (Ki=
yi/xi)

Regardless of which thermodynamic formulation is used for estimating K-values, their

accuracy depends on the correlations used for the thermodynamic properties (vapour
pressure, activity coefficient, and fugacity coefficients). For practical applications, the choice
of K-value formulation is acompromise among accuracy, complexity, convenience, and past
experience.
For liquid–liquid equilibria, (2-9) becomes

where superscripts (1) and (2) refer to the immiscible liquids. A rigorous formulation for the
distribution coefficient is obtained by combining (2-23) with (2-20) to obtain an expression
involving only activity coefficients:

For vapor–solid equilibria, if the solid phase consists of just one of the components of the
vapor phase, combination of (2-9) and (2-25) gives

At low pressure, ΦiV = 1.0 and the fugacity of the solid is approximated by its vapour
pressure. Thus for the vapour-phase mole fraction of the component forming the solid phase:

5
For liquid–solid equilibria, if the solid phase is a pure component, the combination of (2-9)
and (2-23) gives

At low pressure, fugacity of a solid is approximated by vapour pressure to give, for a

component in the solid phase,

EXAMPLE 2.2 K-Values from Raoult’s and Henry’s Laws.

Estimate the K-values and α of a vapour–liquid mixture of water (W) and methane (M) at P=
2 atm, T= 20 and 80°C. What is the effect of T on the component distribution?

At these conditions, water exists mainly in the liquid phase and will follow Raoult’s law.
Because methane has a critical temperature of -82.5°C, well below the temperatures of
interest, it will exist mainly in the vapour phase and follow Henry’s law, in the form given in
From Perry’s Chemical Engineers’ Handbook, 6th ed., pp. 3-237 and 3-103, the vapour
pressure data for water and Henry’s law coefficients for CH4 are:

IDEAL-GAS, IDEAL-LIQUID-SOLUTION MODEL

Classical thermodynamics provides a means for obtaining fluid properties in a consistent
manner from P–y–T relationships, which are equation-of-state models. The simplest model
applies when both liquid and vapour phases are ideal solutions (all activity coefficients equal
1.0) and the vapour is an ideal gas. Then the thermodynamic properties can be computed
from unary constants for each species using the equations given in Table 2.3.

Table 2.3: Thermodynamic Properties for Ideal Mixtures

6
These ideal equations apply only at pressures up to about 50 psia (345 kPa), for components
of similar molecular structure. For the vapor, the molar volume, y, and mass density, ρ,
are computed from (1), the ideal-gas law in Table 2.3, which requires the mixture molecular
weight, M, and the gas constant, R. It assumes that Dalton’s law of additive partial pressures
and Amagat’s law of additive volumes apply.
The vapor enthalpy, hV, is computed from (2) by integrating an equation in temperature for
the zero-pressure heat capacity at constant pressure, CoPV, starting from a reference (datum)
temperature, TO, to the temperature of interest, and then summing the resulting species vapor
enthalpies on a mole-fraction basis. Typically, TO is taken as 0 K or 25°C.
Pressure has no effect on the enthalpy of an ideal gas. A common empirical representation of
the effect of temperature on the zero-pressure vapor heat capacity of a component is the
fourth-degree polynomial:

7
where the constants depend on the species. Values of the constants for hundreds of
compounds, with T in K, are tabulated by Poling et al. (2001). Because CP = dh/dT, (2-35)
can be integrated for each species to give the ideal-gas species molar enthalpy:

The vapor entropy is computed from (3) in Table 2.3 by integrating CoPV/T from T0 to T for
each species; summing on a mole-fraction basis; adding a term for the effect of pressure
referenced to a datum pressure, P0, which is generally taken to be 1 atm (101.3 kPa); and
adding a term for the entropy change of mixing. Unlike the ideal vapor enthalpy, the ideal
vapor entropy includes terms for the effects of pressure and mixing. The reference pressure is
not zero, because the entropy is infinity at zero pressure. If (2-35) is used for the heat
capacity,

The liquid molar volume and mass density are computed from the pure species using (4) in
Table 2.3 and assuming additive volumes (not densities). The effect of temperature on pure-
component liquid density from the freezing point to the critical region at saturation pressure
is correlated well by the two-constant equation of Rackett (1970):

where values of the empirical constants A and B, and the critical temperature, Tc, are
tabulated for approximately 700 organic compounds by Yaws et al. (1991).
The vapor pressure of a liquid species is well represented over temperatures from below the
normal boiling point to the critical region by an extended Antoine equation:

where the constants kk depend on the species. Values of these constants for hundreds of
compounds are built into the physical-property libraries of all process simulation programs.
Constants for other vapor-pressure equations are tabulated by Poling et al. (2001). At low
pressures, the enthalpy of vaporization is given in terms of vapor pressure by classical
thermodynamics:

If (2-39) is used for the vapor pressure, (2-40) becomes

The enthalpy of an ideal-liquid mixture is obtained by subtracting the enthalpy of

vaporization from the ideal vapor enthalpy for each species, as given by (2-36), and
summing, as shown by (5) in Table 2.3. The entropy of the ideal-liquid mixture, given by (6),
is obtained in a similar manner from the ideal-gas entropy by subtracting the molar entropy of
vaporization, given by ΔHvap=T.

8
The final equation in Table 2.3 gives the expression for the ideal K-value, previously
included in Table 2.2. It is the K-value based on Raoult’s law, given as

where the assumption of Dalton’s law is also required:

Combination of (2-42) and (2-43) gives the Raoult’s law Kvalue:

The extended Antoine equation, (2-39), can be used to estimate vapor pressure. The ideal K-
value is independent of compositions, but exponentially dependent on temperature because of
the vapor pressure, and inversely proportional to pressure. From (2-21), the relative volatility
using (2-44) is pressure independent.

NONIDEAL THERMODYNAMIC PROPERTY MODELS

Two types of models are used:
(1) P–y–T equation-of-state models and
(2) activity coefficient or free-energy models.
Their applicability depends on the nature of the components in the mixture and the reliability
of the equation constants.

P–y–T Equation-of-State Models

A relationship between molar volume, temperature, and pressure is a P–y–T equation of state.
Numerous such equations have been proposed. The simplest is the ideal-gas law, which
applies only at low pressures or high temperatures because it neglects the volume occupied
by the molecules and the intermolecular forces. All other equations of state attempt to correct
for these two deficiencies. The most widely used equations of state are listed in Table 2.4
except the van der Waals equation. The van der Waals equation was the first successful
formulation of an equation of state for a nonideal gas. It is rarely used anymore because of its
narrow range of application.

However, its development suggested that all species have equal reduced molar volumes, vr =
v/vc, at the same reduced temperature, Tr = T/Tc, and reduced pressure, Pr = P/Pc. This finding,
referred to as the law of corresponding states, was utilized to develop (2) in Table 2.4 and
defines the compressibility factor, Z, which is a function of Pr, Tr,
and the critical compressibility factor, Zc, or the acentric factor, ω. This was introduced by
Pitzer et al. (1955) to account for differences in molecular shape and is determined from the
vapor pressure curve by:

The value for ω is zero for symmetric molecules. Some typical values of ω are 0.264, 0.490,
and 0.649 for toluene, n-decane, and ethyl alcohol, respectively, as taken from the extensive
tabulation of Poling et al. (2001).
Redlich and Kwong (1949) published an equation of state that, like the van der Waals
equation, contains only two constants, both of which can be determined from Tc and
Pc, by applying the critical conditions

9
Table 2.4: Useful equations of state

The R–K equation, given as (3) in Table 2.4, is an improvement over the van der Waals
equation. When applied to nonpolar compounds, the R-K equation has accuracy comparable
with that of equations containing many more constants (Shah and Thodos, 1965).
Furthermore, the R–K equation can approximate the liquid-phase region.
If the R–K equation is expanded to obtain a common denominator, a cubic equation in y
results. Alternatively, (2) and (3) in Table 2.4 can be combined to eliminate v to give
the compressibility factor, Z, form of the equation:

Equation (2-46), a cubic in Z, can be solved analytically for three roots (e.g., see Perry’s
Handbook, 8th ed., p. 3-10).
At supercritical temperatures, where only one phase exists, one real root and a complex
conjugate pair of roots are obtained. Below the critical temperature, where vapor and/or
liquid phases can exist, three real roots are obtained, with the largest value of Z applying to
the vapor and the smallest root corresponding to the liquid (ZV and ZL).
The intermediate value of Z is discarded.
To apply the R–K equation to mixtures, mixing rules are used to average the constants a and
b for each component. The recommended rules for vapor mixtures of C components are

10
Following the work of Wilson (1966), Soave (1972) added a third parameter, the acentric
factor, v, to the R–K equation. The resulting Soave–Redlich–Kwong (S–R–K) or Redlich–
Kwong–Soave (R–K–S) equation, given as (4) in Table 2.4, was accepted for application to
mixtures of hydrocarbons because of its simplicity and accuracy. It makes the parameter
a a function of ω and T, thus achieving a good fit to vapor pressure data and thereby
improving the ability of the equation to predict liquid-phase properties.
Four years after the introduction of the S–R–K equation,
Peng and Robinson (1976) presented a modification of the R–K and S–R–K equations to
achieve improved agreement in the critical region and for liquid molar volume. The Peng–
Robinson (P–R) equation of state is (5) in Table 2.4. The S–R–K and P–R equations of state
are widely applied in calculations for saturated vapors and liquids. For mixtures of
hydrocarbons and/or light gases, the mixing rules are given by (2-49) and (2-50), except that
(2-49) is often modified toinclude a binary interaction coefficient, kij:

Values of kij from experimental data have been published for both the S–R–K and P–R
equations, e.g., Knapp et al. (1982). Generally kij is zero for hydrocarbons paired with
hydrogen or other hydrocarbons. Although the S–R–K and P–R equations were not intended
to be used for mixtures containing polar organic compounds, they are applied by employing
large values of kij in the vicinity of 0.5, as back-calculated from data. However, a preferred
procedure for polar organics is to use a mixing rule such as that of Wong and Sandler (1992),
which bridges the gap between a cubic equation of state and an activity-coefficient equation.
Another model for polar and nonpolar substances is the virial equation of state due to Thiesen
(1885) and Onnes (1912). A common representation is a power series in 1/v for Z:

A modification of the virial equation is the Starling form (1973) of the

Benedict–Webb–Rubin (B–W–R) equation for hydrocarbons and light gases. Walas (1985)
presents a discussion of B–W–R-type equations, which—because of the large number of
terms and species constants (at least 8)—is not widely used except for pure substances at
cryogenic temperatures. A more useful modification of the B–W–R equation is a generalized
corresponding-states form developed by Lee and Kesler (1975) with an extension to mixtures
by Plöcker et al. (1978). All of the constants in the L–K–P equation are given in terms of the
acentric factor and reduced temperature and pressure, as developed from P–y–T data for three
simple fluids (ω = 0), methane, argon, and krypton, and a reference fluid (ω = 0.398), n-
octane. The equations, constants, and mixing rules are given by Walas (1985). The L–K–P
equation describes vapor and liquid mixtures of hydrocarbons and/or light gases over wide
ranges of T and P.

Derived Thermodynamic Properties from P–y–T Models

If a temperature-dependent, ideal-gas heat capacity or enthalpy equation such as (2-35) or (2-
36) is available, along with an equation of state, all other vapor- and liquid-phase properties

11
can be derived from the integral equations in Table 2.5. These equations, in the form of
departure from the ideal gas equations of Table 2.3, apply to vapor or liquid. When the ideal-
gas law, P =RT/ν, is substituted into Eqs. (1) to (4) of Table 2.5, the results for the vapor are

When the R–K equation is substituted into the equations of Table 2.5, the results for the
vapor phase are:

Table 2.6: Integral Departure Equations of Thermodynamics

12