CHE553 Chemical Engineering Thermodynamics

24/04/2015

OVERVIEW
Fundamental equations of solution thermodynamics are given in the preceding
chapter.
In this chapter, experimental vapor/liquid equilibrium (VLE) data are
considered, from which the activity coefficient correlations are derived.

LIQUID PHASE PROPERTIES FROM VLE DATA

The figure shows coexistence of a vapor mixture and liquid solution in

vapor/liquid equilibrium. T and P are uniform throughout the vessel and
can be measured with appropriate instruments. Vapor and liquid samples
may be withdrawn for analysis and this provides experimental values for
mole fractions in the vapor {yi} and mole fractions in the liquid {xi}.
4

CHE553 Chemical Engineering Thermodynamics

24/04/2015

Fugacity
Fugacity increases from zero at infinite dilution (xi = yi 0) to Pisat for pure
species i. This is illustrated by the data of Table 12.1 for methyl ethyl
ketone(1)/toluene(2) at 50oC (323.15K). The first three columns list a set of
experimental P-x1-y1 data and columns 4 and 5 show

For species i in the vapor mixture, eq. (11.52) is written as

i

fi
yiP

fi v y iiv P
The criterion of vapor/liquid equilibrium, as given by eq. (11.48), is that
fi l fi v . Therefore,
fi l y iiv P
(11.52)

f1 y1P and f2 y 2 P

VLE measurements are made at low pressure (P 1 bar) that the vapor phase
may be assumed an ideal gas. In this case iv 1 . Therefore,

fi l fi v y i P

yiP
x i Pi sat

Eqn. (12.1)

The fugacity of species i (in both the liquid and vapor phase) is equal to the
partial pressure of species i in the vapor phase.

The fugacities are plotted in Fig.

12.2 as solid lines. The straight
dashed lines represent the
Lewis/Randall rule expressing the
composition dependence of the
constituent fugacities in an ideal
solution,

The equilibrium pressure P varies with

compositions but has negligible
influence on the liquid phase values of
f1 and f2 .

Therefore, a plot at constant T and P is

as shown in Fig. 12.3 for species i (i =
1, 2) in a binary solution at constant T
and P.

fi id x i f i

This figure illustrates the general

nature of relationships f1 and f2 vs. x1
for a binary liquid solution at
constant T.

fi id xi fi

CHE553 Chemical Engineering Thermodynamics

24/04/2015

Activity coefficient
The solid lines in Fig. 12.2 and 12.3, representing experimental values of fi ,
become tangent to the Lewis/Randall rule lines at xi = 1.
In the other limit, x i 0, fi becomes zero. The ratio fi x i is indeterminate in
this limit, and application of lHopital rule yields

The lower dashed line in Fig. 12.3 represent the Lewis/Randall rule, which
characterize ideal solution behavior.
Activity coefficient as defined by eq. (11.90) provide the actual behavior
(non-ideal) from the idealize one:

lim

f
f
i i idi
x i f i fi

x i 0

yi P
yP
i
x i f i x i Pi sat

(12.1)

(i 1, 2 , ....N)

(12.2)

Eq. (12.2) defines Henrys constant, Hi as the limiting slope of the fi vs x i

curve at xi = 0.
As shown by Fig. 12.3, this is the slope of a line drawn tangent to the curve at
xi = 0. The equation of this tangent line expresses Henrys law:
(12.3)
f x

The activity coefficient of a species in solution is the ratio of its actual

fugacity to the value given by Lewis/Randall rule at the same T, P and
composition. For the calculation of experimental values, both fi and fi id are
eliminated in favor of measurable quantities:

fi dfi

x i dx i x 0
i

Henrys law as given by eq. (10.4) follows immediately from this equation
when f y P , i.e. when f has its ideal gas value.
i

This is restatement of eq. (10.5), modified Raoults law, and allow calculation
of activity coefficient from VLE data as shown in the last two columns of Table
12.1.

10.4

y i P x i Hi

10

Henrys law is related to the Lewis/Randal rule through the Gibbs/Duhem

equation.
Writing eq. (11.14) for a binary solution and replacing Mi by G i i
x1d 1 x 2 d 2 0

const T , P

x dM 0
i

Substitute dx1 by dx2 in the second term produces

x1

d ln f1
d ln f2
x2
dx1
dx 2

In the limit as x1 1 and x2 0,

(11.14)

Differentiate eq. (11.46) at constant T and P yields:

i i T RT ln fi

lim

x1 1

(11.46)

x1d ln f1 x 2 d ln f2 0

d f1 dx1
d f dx
lim 2 2
x 2 0 f x
f1 x1
2
2

Because f1 f1 when x1 = 1, this may be rewritten as

d i RTd ln fi

Therefore,

x1

d ln f1
d ln f2
x2
0
dx1
dx1

const T,P

d f2 dx 2
1 df1
x2 0

f1 dx1 x 1
lim f2 x 2
1

const T , P

x 2 0

Divide by dx1 becomes

d f1 dx1 d f2 dx 2

f1 x1
f2 x 2

or

According to eq. (12.2), the numerator and denominator on the right side of
this equation are equal.

(12.4)

lim

This is the special form of the Gibbs/Duhem equation.

x i 0

11

fi dfi

x i dx i x 0
i

(12.2)

12

CHE553 Chemical Engineering Thermodynamics

24/04/2015

Therefore the exact expression of Lewis/Randall rule applied to real solutions

is:
df1

f1
dx1 x11

(12.5)

When the second species is methanol,

acetone exhibits positive deviations
from ideality and with chloroform it
exhibits negative deviations.

It also implies that eq. (11.83) provides approximately correct values of fi

when xi 1:
fi fi id x i f i

The fugacity of pure acetone, facetone

is the same regardless of the identity
of the second species.

Henrys law applies to a species as it approaches infinite dilution in a

binary solution, and the Gibbs/Duhem equation insures validity of
the Lewis/Randall rule for the other species as it approaches purity.

Henrys constants are represented by

the slopes of the two dotted lines.

13

14

Excess Gibbs Energy

P-x1 data points all lie
above dash line (Raoults
law) positive deviation.

For a binary system,

GE
x1 ln 1 x 2 ln 2
RT

Experiment
al data

For i 1 and ln i 0

(12.6)

Properties
of liquid
phase

Raoults law

yi P
xi Pi sat

15

Activity coefficient of a species in solution becomes unity as the species becomes pure,
each ln i (i=1,2) tends to zero as xi 1.
As xi 0, species i becomes infinitely dilute, ln i approaches a finite limit, namely ln i.

16

CHE553 Chemical Engineering Thermodynamics

24/04/2015

Model

The most important step in determining composition of species in solution (that

is in equilibrium with vapor) is to calculate the species activity coefficients.
There are a number of models that can be used to determine the activity
coefficients depending on the type of species.

Margules

Excess Gibbs free energy model e.g. Margules and Van Laar
Local composition model e.g. Wilson, NRTL and UNIQUAC

Van Laar

Wilson

Equations
GE
A21x1 A12 x 2
x1x 2 RT

GE
A'12 A' 21

x1x 2 RT A'12 x1 A' 21x 2

(12.16)

Vj
Vi

exp

aij
RT

(i j)

(12.10a)

ln 2 x12 A21 2 A12 A21 x 2

(12.10b)

A x
ln 1 A'12 1

A x
'
12 1
'
21 2

A x
ln 2 A' 21 1 '21 2
A 12 x1
'

GE
x1 ln x1 x 2 12 x 2 ln x 2 x1 21
RT
ij

NRTL

(12.9a)

ln 1 x 2 2 A12 2 A21 A12 x1

(12.18)

(12.24)

GE
G 21 21
G

12 12
x1x2 RT x1 x2G 21 x 2 x1G12

(12.20)

G12 exp( 12 ) G 21 exp( 21)

b
b
12 12
21 21
RT
RT

(12.17a)
2

(12.17b)

12
21
ln 1 ln x1 x 2 12 x2

x1 x 2 12 x 2 x121

(12.19a)

12
21
ln 2 ln x 2 x121 x1

x1 x 2 12 x 2 x121

(12.19b)

G 21
G1212
ln 1 x 22 21

2
x 2 x1G12
x1 x 2G 21

G12
G 21 21
ln 2 x12 12

2
x1 x 2G 21
x 2 x1G12

(12.21a)

(12.21b)
18

17

Margules Equation
Set of points in Fig. 12.5(b) provide linear
relation for GE/x1x2RT
GE
A21x1 A12 x 2
(12.9a)
x1x 2 RT
where A21 and A12 are constant.
Alternatively,
GE
A21x1 A12 x 2 x1x 2
(12.9b)
RT
From eqn. (12.10a) & (12.10b), the limiting
conditions of infinite dilution (at xi = 0)

ln 1 x 2 2 A12 2 A21 A12 x1

(12.10a)

ln 2 x A21 2 A12 A21 x 2

(12.10b)

2
1

ln 1 A12 at x1 0

19

From Fig. 12.5(b), the intercepts

at x1 = 0 and x1 =1 of the
straight line GE/x1x2RT gives the
parameters A12 and A21.
A12 = 0.372

A21 = 0.198

and ln 2 A21 at x 2 0
Fig. 12.5(b)

20

CHE553 Chemical Engineering Thermodynamics

24/04/2015

Van Laar Equation

Wilson Equation

Linear relation for GE/x1x2RT

GE
A' A'
' 12 21'
x1x 2 RT A 12 x1 A 21x 2

GE
x1 ln x1 x 2 12 x 2 ln x 2 x1 21
RT

(12.16)

where A21 and A12 are constant.

Alternatively,
x1x 2 RT A'12 x1 A'21x2
x
x

'1 '2
GE
A'12 A'21
A 21 A 12
From eqn. (12.17a) & (12.17b), the limiting conditions of infinite dilution
(at xi = 0)

A' x
ln 1 A'12 1 ' 12 1
A 21x 2

A' x
ln 2 A' 21 1 '21 2
A 12 x1

(12.17a)

ln A '12 at x1 0

ij

(12.17b)

and ln A '21 at x 2 0

Vj
Vi

exp

aij

(i j)

RT

(12.18)
(12.24)

12
21
ln 1 ln x1 x 2 12 x2

x1 x 2 12 x 2 x121

(12.19a)

12
21
ln 2 ln x 2 x121 x1

x1 x 2 12 x 2 x121

(12.19b)

The limiting conditions of infinite dilution (at xi = 0),

Values of A12 and A21 are obtained from the intercept of plot GE/x1x2RT vs. x1 at
x1 = 0 and x1=1. Or from plot of x1x2RT/GE vs. x1, intercept at x1=0 is 1/A12 and
at x1=1 is 1/A21.

ln 1 ln 12 1 21 at x1 0

and ln 2 ln 21 1 12 at x2 0
22

21

NRTL Equation
GE
G 21 21
G1212

x1x 2 RT x1 x 2G 21 x 2 x1G12

All models provide eqns. for ln 1 and ln 2. This allow construction of a

correlation of the original P-x1-y1 data set (experimental values). Eq. (12.1) is
rearranged to give
y1P x11P1sat and y 2P x 2 2P2sat
Addition gives
(12.11)
P x11P1sat x 2 2 P2sat

(12.20)

G12 exp( 12 ) G 21 exp( 21)

b
b
12 12
21 21
RT
RT

G1212
21
ln 1 x 22 21

2
x1 x 2G 21 x 2 x1G12

From eqn. (12.1), therefore

2
G

G 21 21
12
ln 2 x 12

2
x

x
G
x

x
G
1 2 21
1 12
2
2
1

(12.21a)

y1
(12.21b)

The limiting conditions of infinite dilution (at xi = 0),

ln 1 21 12 exp( 12 )

and

ln 2 12 21 exp( 21)
23

x P
P

sat
1 11

yi P
yP
i
x i f i x i Pi sat

x P
x P

sat
1 11
sat
sat
1 11
2 2 2

x P

(12.1)
(12.12)

Values of 1 and 2 from all models with their parameters are combined with
experimental values of P1sat and P2sat to calculate P and y1 by eqs. (12.11)
and (12.12) at various x1.
Then, P-x1-y1 diagram can be plotted to compare the experimental data and
calculated values.

24
24

CHE553 Chemical Engineering Thermodynamics

24/04/2015

Thermodynamic Consistency
experimental
calculated

The Gibbs Duhem eqn imposes a constraint on activity coefficients that may not
be satisfied by a set of experimental values derived from P-x1-y1 data.
x1

d ln 1
d ln 2
x2
0
dx1
dx1

const T,P

(12.7)

The Gibbs Duhem eqn is implicit in eq. (11.96), and activity coefficients derived
from this equation necessarily obey the Gibbs Duhem eqn.

Comparison of
experimental and
calculated data by
Margules eqn. These
clearly provide an
adequate
correlation of the
experimental data
points.

nG E RT

ni

P ,T ,n j

ln i

(11.96)

These derived activity coefficients cannot possibly be consistent with the

experimental values unless the experimental values also satisfy the Gibbs
Duhem eqn.

Fig. 12.5(a)
25

26

If the experimental data are inconsistent with the Gibbs Duhem eqn, they are
necessarily incorrect as the result of systematic error in the data.
Therefore simple test is develop for the consistency with respect to the Gibbs
Duhem eqn of a P-x1-y1 data set.
Application of the test for consistency is represented by Eq. (12.13) which
requires calculation of the residuals

d ln 1*
d ln 2*
ln 1 x1
x2
(12.13)
2
dx1
dx1

Experimental
values
calculated from
eqn. (12.1) &
(12.6)

Asterisk * denote the experimental values

The right side of this equation is exactly the quantity that eq. (12.7), the
Gibbs/Duhem equation, requires to be zero for consistent data.
d ln 1
d ln 2
x1
x2
0
const T,P
(12.7)
dx1

dx1

The residual on the left therefore provides a direct measure of deviation from
the Gibbs/Duhem equation.
The extent to which a data set departs from consistency is measured by the
degree to which these residuals fail to scatter about zero.

yi P
yP
i
x i f i x i Pi sat

(i 1, 2 , ....N)

(12.1)

27

G
x1 ln 1 x 2 ln 2
RT

(12.6)
28

CHE553 Chemical Engineering Thermodynamics

24/04/2015

P-x1-y1 data and experimental

values, ln 1*, ln 2* and (GE/x1x2RT )*
are shown as points on Figs. 12.7(a)
and 12.7(b) .
The data points of Fig. 12.7(b) for
(GE/x1x2RT )* show scatter. The
straight line drawn is represented by

ln 1 x 2 2 A12 2 A21 A12 x1

(12.10a)

ln 2 x12 A21 2 A12 A21 x 2

(12.10b)

P x11P1sat x 2 2 P2sat

(12.11)

y1

This is eq. (12.9a) with A21 = 0.70

and A12 = 1.35.
Values of ln 1, ln 2 at the given
values of x1, derived from this eqn,
are calculated by eqs. (12.10) and
derived values of P and y1 at the
same values of x1 come from eqs.
(12.11) and (12.12).
This results are plotted as the solid
lines of Fig. 12.7(a) and 12.7(b).
They clearly do not represent a good
correlation of the data.
x11P1sat
x11P1sat x 2 2 P2sat

G E RT and ln 1 2
which plotted vs. x1 in Fig. 12.8.

GE
0.70 x1 1.35 x 2
x1x 2 RT

experimental
calculated
Barkers method

Application of test for consistency

represented by Eq. 12.13 requires
calculation of the residuals

The residuals G E RT distribute

themselves about zero, as required by
the test, but the residual ln 1 2 , do
not, which show the data fail to satisfy
the Gibbs/Duhem eqn.

(12.12)
29

30

REFERENCE
Smith, J.M., Van Ness, H.C., and Abbott, M.M. 2005. Introduction to
Chemical Engineering Thermodynamics. Seventh Edition. Mc GrawHill.

PREPARED BY:
NORASMAH MOHAMMED MANSHOR
FACULTY OF CHEMICAL ENGINEERING,
UiTM SHAH ALAM.
norasmah@salam.uitm.edu.my
03-55436333/019-2368303

31