3

Mass Transfer and Efficiency

of Separation Operations

Alexandre Rojey

3.1 Introduction
In order to size the various pieces of equipment used in separations, the kinet-
ics of transfer of mass must be understood.
For a contact operation by stages (trays or mixer-settlers), kinetics governs
the efficiency of each of the stages and consequently the number of real stages
required to perform a given separation operation.
If the contact between the two phases is performed in a column, such as for
example a packed column, the height of the column also depends on the trans-
fer coefficient between the phases in contact.
Finally, in the case of a membrane separation operation whose selectivity
is kinetic, the permeability of the membrane toward the different components
present depends on the rate of diffusion of each of the components through
the membrane.
In this chapter the kinetics of mass transfer will be analyzed at three dif-
ferent levels:
diffusion in a homogeneous phase;
mass transfer between two phases through the interface;
efficiency of equipment for contact between phases.
Generally speaking, mass transfer occurs by means of a complex mecha-
nism and predicting characteristic variables at these different levels is at least
partly empirical.

3.2 Diffusion in a Homogeneous Phase

In a mixture whose composition is not uniform, molecular agitation tends to
restore uniform concentrations by a molecular diffusion process. As a result,
82 ChaDter 3. MASSTRANSFER
AND EFFICIENCY OF SEPARATION OPERATIONS

for each component there is mass transfer in relation to a fixed point of refer-
ence, say NA and NB (mol/m*.s) for a binary mixture.
If CA and CB are the molar concentrations (mol/m3) of A and B, a displace-
ment velocity is defined for each component by the equations below:
NA = CAUA et NB = CBVB (34
If the composition varies in one direction z, the velocities uAand ug are ori-
ented along this direction and can be considered as scalar variables.
The average volume velocity u is then given by the equation:
u =V +V B ~ B
A ~ A (3.2)
vB
with FAand the partial molar volumes of A and B (m3/mol).
Diffusion materializes a relative displacement of component A in relation to
component B.
The diffusion flow of component A in relation to a point of reference mov-
ing with the average volume velocity u is expressed according to Fick’s law:

J A represents the mass flow (mol/m2.s) across a plane normal to the direc-
tion z under consideration and Dm is the molecular diffusion coefficient of A
in relation to B (m2/s). It can be readily shown that the molecular diffusion
coefficient of B in relation to A, DBA, must be identical to the coefficient Dm
The mass flow NA in relation to a fixed point of reference is then written:

It is also possible to express the diffusion flow versus the gradient of the
mole fraction of the transferring component.
By using C to stand for the total molar concentration, Eq. (3.4) is replaced
by the equation below:
N A = - C D A B (J=X)A +xA(NA+NB)
(3.5)
t
There are a number of correlations designed to predict the value of a diffu-
sion coefficient.
In the gaseous phase, Chapman and Cowling (1964) developed a correla-
tion based on the kinetic theory of gases. Other more empirical correlations
are also available.
The Fuller et al. correlation (1966) gives good results for non-polar
molecules at low pressure. It is written as follows:

by stipulating: Mm = 2 [(l/MJ + (l/MB)l-’.

 Chapter 3. MASSTRANSFER
AND EFFICIENCY OF SEPARATION OPERATIONS 83

T represents the absolute temperature (K), P the pressure (bar), MA and MB

the molar masses (kg/mol).
~ G v ) represent
The terms G v ) and ~ the diffusion volumes calculated by
summing the contributing terms for each atom, with a correction term being
introduced for a cyclic structure. The values are listed in Table 3.1.

C 15.9 F 14.7
H 2.31 c1 21.0
0 6.1 1 Br 21.9
N 4.54 I 29.8
Cyclic structure -18.3 S 22.9
-
Table
3.1 Atomic diffusion volumes. Contributing terms in Eq. (3.6).

It will be noted that:

- Eq. 3.6 is not homogeneous in dimension and therefore the units used
must be strictly respected;
- Eq. 3.6 can be used to verify that DBA = Dm and that Dm does not depend
on composition, as is observed for most binary gaseous mixtures at low
pressure.
In the liquid phase, the Wilke and Chang correlation (1955), which is quite
old, still remains generally accepted. It is written:

In this equation, D L designates the diffusion coefficient of solute A at infi-

nite dilution in solvent B (m2/s); pB,the viscosity of the solvent phase (mPa.s);
T the absolute temperature (K); V, the molar volume of A at its normal boiling
point (cm3/mol); MB the molar mass of the solvent (g/mol) and vBan associa-
tion parameter. Eq. 3.7 (non-homogeneous) indicates that D& # D!A, which is
generally true for liquids.
The value of the association parameter vBis 2.6 for water, 1.9 for methanol,
1.5 for ethanol and 1.0 if the solvent does not give rise to any association.
Correlation (3.7) is valid only for nondissociated solutes.
Other more recent correlations proposed by Tyn and Calus (1975) and Tyn
(1981) as well as by Hayduck and Minhas (1982) use the value of the parachor
of each component, which can be calculated by summing a number of con-
tributing terms (see Vol. 1, Chapter 4). The use of these different correlations
is presented in the book by Reid, Prausnitz and Poling (1987).
The correlations can be used to calculate the diffusion coefficient for very
dilute solutions. In the case of concentrated solutions, the diffusion coefficient
84 ChaDter 3. MASSTRANSFER
AND EFFICIENCY OF SEPARATION OPERATIONS

can be obtained from the diffusion coefficients at infinite dilution D L and DflA
by the Vignes equation (1966):

aAbeing the activity of A in the solution, equal to YAxA, with X, the mole frac-
tion of A and X, the mole fraction of B.
6’ In aA
If the solution is ideal: - = 1.
d In xA
Equation 3.8 can be used to verify that, even though DAB varies with the
composition, DAB= DBk The Gibbs-Duhem law establishes that:
(3 In aA/aIn xJTp = (a In a& In x,)~-
There are no general correlations to predict the value of the diffusion coef-
ficients in a solid phase. The values depend very heavily on the material under
consideration and the phenomena involved are complex.
In the gaseous phase the diffusion coefficients are along the lines of
m2/s and in the liquid phase m2/s.
The values observed for diffusion in the solid phase are even lower. For
example for diffusion of ethane in a zeolite of the type 4A at 25°C the value is
approximately m2/s Qang et al., 1991).
Because of this, the adsorbents used are in the form of porous structures
obtained by agglomerating small sized crystals in order to get acceptable
transfer kinetics.

3.3 Transfer between Phases

When two phases 1 and 2 are in contact, mass transfer takes place between
them if they are not at thermodynamic equilibrium. The concentration profiles
obtained in the neighborhood of the interface are schematically shown in
Figure 3.1.
So that a component can be transferred from phase 1 to phase 2, the con-
centration C of this component in phase 1 must be higher than the concentra-
tion C* corresponding to equilibrium with the concentration C’ of the
component in phase 2.
The difference (C - C*) represents the driving force of the transfer. This
leads to the following equation:
N = K(C- C*) (3.9)
In this equation, N represents the flow of component that transfers through
the interface (mol/m2.s) and K the overall transfer coefficient, which has the
dimension of a velocity (m/s) here.
 Chapter 3. MASSTRANSFER
AND EFFICIENCY OF SEPARATION OPERATIONS 85

C'

Phase I Phase 2

Figure
3.1 Transferbetween phases. Variation in concentrations in the neighborhood of
the interface (the subscript i is related to interface conditions).

It is also possible to define a transfer coefficient in each phase.

In phase 1 the driving force is represented by the difference between the
concentration C in phase 1 and the concentration at the interface Ci.
In these conditions, an individual transfer coefficient relative to phase 1 is
defined such that:
N = k ( C - Ci) (3.10)
Likewise, an individual transfer coefficient k' in phase 2 is defined such
that:
N = k'(C;- C') (3.1 1)
In the absence of local resistance, the two phases are in equilibrium at the
interface.
If the equilibrium equation is roughly linear in the domain of concentration
under consideration (see Fig. 1.6). it is expressed in the form below:
C*=mC'+q (3.12)
Given the equilibrium equation at the interface, the immediate result is as
follows:
l
-=-+-
l m
(3.13)
K k k'
This equation is often called the equation of resistance composition, since
the inverse of the transfer coefficient can be compared to a transfer resistance.
Each of the transfer coefficients K, k and k'can be defined by expressing
the driving force versus mole fractions rather than concentrations. Here the
transfer coefficient no longer has the dimension of a velocity and is expressed
in (rnol/m2.s).
86 Chapter 3 MASSTRANSFER
AND EFFICIENCY OF SEPARATION OPERATlONS

3.4 Predicting Transfer Coefficients

The transfer coefficient in each phase depends on the diffusion coefficient.
The film model provides a very simple relation between the transfer coeffi-
cient and the diffusion coefficient.
In this model it is assumed that the concentration gradient is limited to a
very thin stagnant film in the neighborhood of the interface, with a thickness 6,
as shown in Figure 3.2.

.
film
Interface
Ci
V
Diffusion flux

Figure
3.2 Film model. Concentration profile.

In a steady state, the concentration profile in the film is linear, thereby lead-
ing to the equation:

k=-
D
(3.14)
6
With these assumptions, the transfer coefficient k is proportional to the dif-
fusion coefficient D.
Since the phenomenon is actually much more complex, Eq. (3.14) has no
predictive value, but the film model illustrates the transfer mechanism very
simply.
The interface renewal model is based on the idea that in a contact between
phases a volume unit of one phase is exposed to the contact with the other
phase during a limited time, during which transfer occurs in a transient state.
This model leads to the equation below:
- D 112
k = 2(8) (3.15)

In this equation 8 represents the contact time and the mean transfer coef-
ficient k varies as the square root of the diffusion coefficient D.
Other more complex models have been proposed, such as the Toor and
Marchelo model, which is a combination of the film model and the interface
renewal model, but for the time being they are not directly usable in predict-
ing a transfer coefficient. The procedure followed is similar to the one used to
 Chanter 3. MASSTRANSFER
AND EFFICIENCY
OF SEPARATION OPERATIONS 87

analyze other transfer phenomena such as heat transfer or angular momentum

transfer which is involved in the flow of viscous fluids (Bird et al., 1960).
It is possible to determine the transfer coefficient for simple geometries by
analytic equations, either by solving the diffusion equation in the case of a
stagnant medium or boundary layer equations in the presence of convection
in the case of a laminar system.
In practice it is not usually possible to determine the coefficient of transfer
of mass by analytic or even numerical solution of a predictive model due to the
complexity of hydrodynamic conditions. Dimensional analysis must then be
resorted to, linking the following dimensionless numbers:
Sherwood’s number:
kd
Sh = -
D
with k being the transfer coefficient in the phase under consideration, D
the diffusion coefficient of the transferring component and d a charac-
teristic dimension.
Reynolds’s number
ud
Re = -
V
with u being a velocity characteristic of the phase under consideration
and v the kinematic viscosity of the phase.
Schmidt’s number:
V
sc = -
D
The relations between Sherwood’s number, Reynolds’s number and
Schmidt’s number have been established for simple geometries.
The equations available are generally established for a dilute phase. When
the concentration of the transferring component increases, the transfer coef-
ficient increases correlatively. To take this into account, a corrective term is
introduced such that:

(3.16)

P
Designating Sherwood’s number at infinite dilution by Sho and stipulating:

if NBi = 0, i.e. P= 1:
88 Chapter 3. AND EFFICIENCY OF SEPARATION
MASSTRANSFER OPERATIONS

In the case of contact between a continuous phase and a dispersed phase,

there are established equations for a spherical particle.
Inside a spherical particle in the absence of internal circulation, transfer
occurs by molecular diffusion alone. When the concentration , C is assumed
to be uniform and constant, calculation is used to evaluate Sherwood’s num-
ber Sh,, which is a function of time. For short times (Dt/d2 < 0.035),
Sherwood’s number Sh, varies very rapidly with time, approximately accord-
ing to the equation below:

(&
Sh,= --
2,
(3.18)

whereas for fairly large t, it becomes roughly constant and takes on the value:
2n3
Sh, = (3.19)

considering the diameter d of the particle as a characteristic dimension.

When internal circulation is induced inside the particle by the movement
of the continuous phase, the transfer is quicker than indicated by the previous
equations. However, circulation is slowed down and even stopped when sur-
factant substances are adsorbed at the interface. These substances are always
present when the fluids are not specially purified and, in the absence of accu-
rate data on them, the values of the transfer coefficients obtained without cir-
culation can at least be used to estimate a lower limit for the transfer
coefficient.
Outside the particles in the continuous phase, transfer takes place by dif-
fusion and convection. If the particle moves at a constant velocity in relation
to the continuous phase and can be considered as a rigid sphere, the transfer
coefficient is determined from a correlation of the following form:
Sh, = 2 + A RemScn (3.20)
The values of coefficients A, R and m depend on the values of the Reynolds
and Schmidt numbers.
The values recommended by Hughmark (1967) are as follows:

Re 1-17 17-450 450-10000

SC < 250 > 250 < 250 > 250 < 250 > 250
A 0.6 0.5 0.6 0.4 0.27 0.175
m 0.5 0.5 0.5 0.5 0.62 0.62
n 0.33 0.42 0.33 0.42 0.33 0.42

Equations (3.19) and (3.20) can be used to estimate the values of coeffi-
cients k,k’ and Kin the case of contact between a continuous phase and a dis-
persed phase in the form of drops or bubbles.
 Chapter 3. MASSTRANSFER
AND EFFICIENCY OF SEPARAJION OPERATIONS 89

There are also specific correlations for different types of equipment back-
ing, trays, mechanically agitated columns, etc.). They are presented in later
chapters along with the presentation of the equipment.

3.5 Relationship between Efficiency

and Transfer Coefficient
When the value of the overall transfer coefficient K is known, it is possible to
evaluate the efficiency of phase contact equipment.
To illustrate the procedure, let us consider the simplest case of a contact
stage which is assumed to be perfectly well mixed. The concentration profiles
obtained in each phase here are schematically represented in Figure 3.3.

- Figure
3.3 Concentration profiles for a perfectly mixed stage.
U

The concentration in each of the phases inside the contact stage is

assumed to be uniform and equal to the exiting concentration. In these condi-
tions, assuming that the inlet and outlet flow rates are equal, the following is
obtained by steady state material balance:
Q(C,-CJ=KA(C,-C,*) (3.2 1)
(as a reminder, C: = m Cg + q, by definition) with A standing for the interfacial
area in the contact stage.
The Murphree efficiency of the contact stage concerning phase 1 is defined
by the equation below:
(3.22)

It varies between 0 when no mass transfer occurs in the contact stage

(C, = CJ and 1 in the limiting case of a theoretical stage. Taking Eq. 3.21 into
account, the immediate result is:

(3.23)
90 Chapter 3. MASSTRANSFER
AND EFFICIENCY OF SEPARATION OPERATIONS

with Nl being the dimensionless ratio:

K4
N1= - (3.24)
Q
In the case of a counterflow contact column, the dimensionless number
defined by Eq. (3.24) is called the number of transfer units for phase 1. In this
case:
KQL
N1= -V
(3.25)

when Q is the interfacial area per unit of volume, L the height of the contact
region and V the velocity at which phase 1 passes defined by the ratio between
the flow rate Q and the total cross-section. The result is as follows:
L = N1Hl (3.26)
with the characteristic length Hl being called the height of a transfer unit:
V
H - (3.27)
l- Ka
In the case of continuous counterflow contact and in the absence of axial
mixing, the number of transfer units Nl is obtained by the Chilton-Colburn
equation:

N1= -
1-A
In this equation, established assuming constant flow rates and a linear
equilibrium relation:
%, = (C, - C,*)/(C, - C:) stands for the exiting reduced concentration of
phase 1, C:, represents the concentration which, in phase 1, corresponds
to the equilibrium with the incoming concentration in phase 2
(CC,.= m CL + q in the case of a linear equilibrium relation).
A = mQ/Q‘ represents an “extraction factor” which must be lower than 1
in order to achieve complete separation (%, + 0).
There are other equations for more complex cases, especially in the pres-
ence of axial mixing (Defives and Rojey, 1976).
For a counterflow contact by stages that all have the same Murphree effi-
ciency EMl,the number n of stages is related to the exiting reduced concen-
tration by the Kremser-Brown-Souders equation:

(3.29)
I
This equation was also established by assuming constant flow rates and a
linear equilibrium relation.
 Chaoter3. MASSTRANSFER
AND fFFlClENCY OF SEPARATION OPERATIONS 91

For theoretical stages, EM1 = 1 and the number of theoretical stages

required to carry out the same operation is given by the equation below:

In [+
1
's
A Ps
]
- 1)
nT = (3.30)
In A

The ratio nT/n can serve to characterize the efficiency of a stage. This over-
all efficiency, written as E, is linked to the Murphree efficiency by the equation
below:
nT In [1+ (A - 1) EM11
E= - = (3.31)
n In A
The way in which E corresponds to EM,is represented in Figure 3.4.

Figure
3.4 Overall efficiency versus Murphree efficiency

In the special case A = 1, the two efficiencies are equal. It can also be seen
that these two efficiencies differ increasingly as A becomes different from 1.
By comparing Eqs. 3.28 and 3.30 an equation is obtained that can be used
to calculate the column height equivalent to a theoretical stage HT for contin-
uous contact. The following can be stipulated
L = Ni Hi = nTHT (3.32)
92 CtJaoter 3. MASSTRANSFER OF SEPARATION
AND EFFICIENCY OPERATIONS

Consequently:

(3.33)

In contrast to the height of a transfer unit Hi, the height equivalent to a the-
oretical stage HT depends on the A factor. This is why it is generally recom-
mended for continuous contact to refer to the concept of height of transfer
unit rather than theoretical stage.
However, in computer modeling by theoretical stages, Eq. 3.33 can be used
to estimate the column height.