Mass Transfer and Diffusion: Ass Transfer Is The Net Movement of A Component in A

Chapter 3
Mass Transfer and Diffusion

Mass transfer is the net movement of a component in a barriers, such as membranes, differing species mass-transfer
mixture from one location to another where the component rates through the membrane govern equipment design.
exists at a different concentration. In many separation In a binary mixture, molecular diffusion of component
operations, the transfer takes place between two phases A with respect to B occurs because of different potentials
across an interface. Thus, the absorption by a solvent liquid or driving forces, which include differences (gradients)
of a solute from a carrier gas involves mass transfer of the of concentration (ordinary diffusion), pressure (pressure
solute through the gas to the gas–liquid interface, across the diffusion), temperature (thermal diffusion), and external force
interface, and into the liquid. Mass-transfer models describe fields (forced diffusion) that act unequally on the different
this and other processes such as passage of a species through chemical species present. Pressure diffusion requires a large
a gas to the outer surface of a porous, adsorbent particle and pressure gradient, which is achieved for gas mixtures with a
into the adsorbent pores, where the species is adsorbed on centrifuge. Thermal diffusion columns or cascades can be
the porous surface. Mass transfer also governs selective employed to separate liquid and gas mixtures by establishing
permeation through a nonporous, polymeric material of a a temperature gradient. More widely applied is forced
component of a gas mixture. Mass transfer, as used here, diffusion in an electrical field, to cause ions of different
does not refer to the flow of a fluid through a pipe. However, charges to move in different directions at different speeds.
mass transfer might be superimposed on that flow. Mass In this chapter, only molecular diffusion caused by
transfer is not the flow of solids on a conveyor belt. concentration gradients is considered, because this is the
Mass transfer occurs by two basic mechanisms: most common type of molecular diffusion in separation
(1) molecular diffusion by random and spontaneous micro- processes. Furthermore, emphasis is on binary systems, for
scopic movement of individual molecules in a gas, liquid, or which molecular-diffusion theory is relatively simple and
solid as a result of thermal motion; and (2) eddy (turbulent) applications are relatively straightforward. Multicomponent
diffusion by random, macroscopic fluid motion. Both molecular diffusion, which is important in many applica-
molecular and/or eddy diffusion frequently involve the tions, is considered briefly in Chapter 12. Diffusion in multi-
movement of different species in opposing directions. When component systems is much more complex than diffusion in
a net flow occurs in one of these directions, the total rate of binary systems, and is a more appropriate topic for advanced
mass transfer of individual species is increased or decreased study using a text such as Taylor and Krishna [1].
by this bulk flow or convection effect, which may be Molecular diffusion occurs in solids and in fluids that are
considered a third mechanism of mass transfer. Molecular stagnant or in laminar or turbulent motion. Eddy diffusion
diffusion is extremely slow, whereas eddy diffusion is orders occurs in fluids in turbulent motion. When both molecular
of magnitude more rapid. Therefore, if industrial separation diffusion and eddy diffusion occur, they take place in
processes are to be conducted in equipment of reasonable parallel and are additive. Furthermore, they take place
size, fluids must be agitated and interfacial areas maximized. because of the same concentration difference (gradient).
If mass transfer in solids is involved, using small particles to When mass transfer occurs under turbulent-flow conditions,
decrease the distance in the direction of diffusion will but across an interface or to a solid surface, conditions may
increase the rate. be laminar or nearly stagnant near the interface or solid
When separations involve two or more phases, the extent surface. Thus, even though eddy diffusion may be the
of the separation is limited by phase equilibrium, because, dominant mechanism in the bulk of the fluid, the overall rate
with time, the phases in contact tend to equilibrate by mass of mass transfer may be controlled by molecular diffusion
transfer between phases. When mass transfer is rapid, because the eddy-diffusion mechanism is damped or even
equilibration is approached in seconds or minutes, and eliminated as the interface or solid surface is approached.
design of separation equipment may be based on phase Mass transfer of one or more species results in a total net
equilibrium, not mass transfer. For separations involving rate of bulk flow or flux in one direction relative to a fixed
66
3.1 Steady-State, Ordinary Molecular Diffusion 67
plane or stationary coordinate system. When a net flux the direction selected as positive. When the molecular and
occurs, it carries all species present. Thus, the molar flux of eddy-diffusion fluxes are in one direction and N is in the
an individual species is the sum of all three mechanisms. If opposite direction, even though a concentration difference
Ni is the molar flux of species i with mole fraction xi, and N or gradient of i exists, the net mass-transfer flux, Ni, of i can
is the total molar flux, with both fluxes in moles per unit time be zero.
per unit area in a direction perpendicular to a stationary In this chapter, the subject of mass transfer and diffusion
plane across which mass transfer occurs, then is divided into seven areas: (1) steady-state diffusion in
Ni = xi N + molecular diffusion flux of i stagnant media, (2) estimation of diffusion coefficients,
+ eddy diffusion flux of i (3-1) (3) unsteady-state diffusion in stagnant media, (4) mass
transfer in laminar flow, (5) mass transfer in turbulent flow,
where xiN is the bulk-flow flux. Each term in (3-1) is positive (6) mass transfer at fluid–fluid interfaces, and (7) mass
or negative depending on the direction of the flux relative to transfer across fluid–fluid interfaces.
3.0 INSTRUCTIONAL OBJECTIVES
After completing this chapter, you should be able to:

• Explain the relationship between mass transfer and phase equilibrium.
• Explain why separation models for mass transfer and phase equilibrium are useful.
• Discuss mechanisms of mass transfer, including the effect of bulk flow.
• State, in detail, Fick’s law of diffusion for a binary mixture and discuss its analogy to Fourier’s law of heat
conduction in one dimension.
• Modify Fick’s law of diffusion to include the bulk flow effect.
• Calculate mass-transfer rates and composition gradients under conditions of equimolar, countercurrent diffusion
and unimolecular diffusion.
• Estimate, in the absence of data, diffusivities (diffusion coefficients) in gas and liquid mixtures, and know of some
sources of data for diffusion in solids.
• Calculate multidimensional, unsteady-state, molecular diffusion by analogy to heat conduction.
• Calculate rates of mass transfer by molecular diffusion in laminar flow for three common cases: (1) falling liquid
film, (2) boundary-layer flow past a flat plate, and (3) fully developed flow in a straight, circular tube.
• Define a mass-transfer coefficient and explain its analogy to the heat-transfer coefficient and its usefulness, as an
alternative to Fick’s law, in solving mass-transfer problems.
• Understand the common dimensionless groups (Reynolds, Sherwood, Schmidt, and Peclet number for mass
transfer) used in correlations of mass-transfer coefficients.
• Use analogies, particularly that of Chilton and Colburn, and more theoretically based equations, such as those of
Churchill et al., to calculate rates of mass transfer in turbulent flow.
• Calculate rates of mass transfer across fluid–fluid interfaces using the two-film theory and the penetration
theory.
3.1 STEADY-STATE, ORDINARY

direction and then in another, with no one direction pre-
MOLECULAR DIFFUSION
ferred. This type of motion is sometimes referred to as a
Suppose a cylindrical glass vessel is partly filled with water random-walk process, which yields a mean-square distance
containing a soluble red dye. Clear water is carefully added of travel for a given interval of time, but not a direction of
on top so that the dyed solution on the bottom is undisturbed. travel. Thus, at a given horizontal plane through the solution
At first, a sharp boundary exists between the two layers, but in the cylinder, it is not possible to determine whether, in a
after a time the upper layer becomes colored, while the layer given time interval, a given molecule will cross the plane or
below becomes less colored. The upper layer is more col- not. However, on the average, a fraction of all molecules in
ored near the original interface between the two layers and the solution below the plane will cross over into the region
less colored in the region near the top of the upper layer. above and the same fraction will cross over in the opposite
During this color change, the motion of each dye molecule is direction. Therefore, if the concentration of dye molecules in
random, undergoing collisions mainly with water molecules the lower region is greater than in the upper region, a net rate
and sometimes with other dye molecules, moving first in one of mass transfer of dye molecules will take place from the
68 Chapter 3 Mass Transfer and Diffusion
lower to the upper region. After a long time, a dynamic equi- where, for convenience, the z subscript on J has been
librium will be achieved and the concentration of dye will be dropped, c = total molar concentration or molar density
uniform throughout the solution. Based on these observa- (c = 1/v = /M), and xA mole fraction of species A.
tions, it is clear that: Equation (3-4) can also be written in the following equiv-
alent mass form, where jA is the mass flux of A by ordinary
1. Mass transfer by ordinary molecular diffusion occurs
molecular diffusion relative to the mass-average velocity of
because of a concentration, difference or gradient; that
the mixture in the positive z-direction, is the mass density,
is, a species diffuses in the direction of decreasing
and wA is the mass fraction of A:
concentration.
2. The mass-transfer rate is proportional to the area normal dw A
jA = −DAB (3-5)
to the direction of mass transfer and not to the volume dz
of the mixture. Thus, the rate can be expressed as a flux.
Velocities in Mass Transfer
3. Net mass transfer stops when concentrations are
uniform. It is useful to formulate expressions for velocities of chemi-
cal species in the mixture. If these velocities are based on
the molar flux, N, and the molar diffusion flux, J, the molar
Fick’s Law of Diffusion
average velocity of the mixture, v M , relative to stationary
The above observations were quantified by Fick in 1855, who coordinates is given for a binary mixture as
proposed an extension of Fourier’s 1822 heat-conduction
N NA + NB
theory. Fourier’s first law of heat conduction is vM = = (3-6)
c c
dT Similarly, the velocity of species i, defined in terms of Ni, is
qz = −k (3-2)
dz relative to stationary coordinates:
where qz is the heat flux by conduction in the positive z- Ni
direction, k is the thermal conductivity of the medium, and vi = (3-7)
ci
dT/dz is the temperature gradient, which is negative in the
direction of heat conduction. Fick’s first law of molecular Combining (3-6) and (3-7) with xi = ci /c gives
diffusion also features a proportionality between a flux and a v M = xA vA + xB vB (3-8)
gradient. For a binary mixture of A and B,
Alternatively, species diffusion velocities, vi D , defined in
dcA terms of Ji, are relative to the molar-average velocity and are
JAz = −DAB (3-3a)
dz defined as the difference between the species velocity and
and the molar-average velocity for the mixture:
dcB
JBz = −DBA (3-3b) Ji
dz vi D = = vi − v M (3-9)
ci
where, in (3-3a), JAz is the molar flux of A by ordinary mol-
ecular diffusion relative to the molar-average velocity of the When solving mass-transfer problems involving net
mixture in the positive z direction, DAB is the mutual diffu- movement of the mixture, it is not convenient to use fluxes
sion coefficient of A in B, discussed in the next section, cA is and flow rates based on vM as the frame of reference. Rather,
the molar concentration of A, and dcA/dz is the concentra- it is preferred to use mass-transfer fluxes referred to station-
tion gradient of A, which is negative in the direction of ordi- ary coordinates with the observer fixed in space. Thus, from
nary molecular diffusion. Similar definitions apply to (3-3b). (3-9), the total species velocity is
The molar fluxes of A and B are in opposite directions. If the vi = v M + vi D (3-10)
gas, liquid, or solid mixture through which diffusion occurs
is isotropic, then values of k and DAB are independent of di- Combining (3-7) and (3-10),
rection. Nonisotropic (anisotropic) materials include fibrous
and laminated solids as well as single, noncubic crystals. Ni = ci v M + ci vi D (3-11)
The diffusion coefficient is also referred to as the diffusivity Combining (3-11) with (3-4), (3-6), and (3-7),
and the mass diffusivity (to distinguish it from thermal and

momentum diffusivities). nA dxA
Many alternative forms of (3-3a) and (3-3b) are used, NA = = xA N − cDAB (3-12)
A dz
depending on the choice of driving force or potential in the
gradient. For example, we can express (3-3a) as and

dxA nB dxB
JA = −cDAB (3-4) NB = = xB N − cDBA (3-13)
dz A dz
Mole fraction, x
Mole fraction, x
xB xB
xA xA Figure 3.1 Concentration profiles for limiting

cases of ordinary molecular diffusion in binary
z1 z2 z1 z2 mixtures across a stagnant film: (a) equimolar
Distance, z Distance, z
counterdiffusion (EMD); (b) unimolecular
(a) (b) diffusion (UMD).
where in (3-12) and (3-13), ni is the molar flow rate in moles Thus, in the steady state, the mole fractions are linear in dis-
per unit time, A is the mass-transfer area, the first terms on the tance, as shown in Figure 3.1a. Furthermore, because c is
right-hand sides are the fluxes resulting from bulk flow, and constant through the film, where
the second terms on the right-hand sides are the ordinary mol-
ecular diffusion fluxes. Two limiting cases are important: c = cA + cB (3-20)
1. Equimolar counterdiffusion (EMD) by differentiation,

2. Unimolecular diffusion (UMD) dc = 0 = dcA + dcB (3-21)
Thus,
Equimolar Counterdiffusion
dcA = −dcB (3-22)
In equimolar counterdiffusion (EMD), the molar fluxes of A
and B in (3-12) and (3-13) are equal but opposite in direc- From (3-3a), (3-3b), (3-15), and (3-22),
tion; thus,
DAB DBA
= (3-23)
N = NA + NB = 0 (3-14) dz dz
Thus, from (3-12) and (3-13), the diffusion fluxes are also Therefore, DAB = DBA.
equal but opposite in direction: This equality of diffusion coefficients is always true in a
binary system of constant molar density.
JA = −JB (3-15)
This idealization is closely approached in distillation. From

(3-12) and (3-13), we see that in the absence of fluxes other EXAMPLE 3.1
than molecular diffusion, Two bulbs are connected by a straight tube, 0.001 m in diameter
and 0.15 m in length. Initially the bulb at end 1 contains N2 and the
dxA
NA = JA = −cDAB (3-16) bulb at end 2 contains H2. The pressure and temperature are main-
dz tained constant at 25◦ C and 1 atm. At a certain time after allowing
diffusion to occur between the two bulbs, the nitrogen content of
and
the gas at end 1 of the tube is 80 mol% and at end 2 is 25 mol%. If

dxB the binary diffusion coefficient is 0.784 cm2/s, determine:
NB = JB = −cDBA (3-17)
dz (a) The rates and directions of mass transfer of hydrogen and
nitrogen in mol/s
If the total concentration, pressure, and temperature are
(b) The species velocities relative to stationary coordinates,
constant and the mole fractions are maintained constant (but
in cm/s
different) at two sides of a stagnant film between z1 and z2,
then (3-16) and (3-17) can be integrated from z1 to any z
between z1 and z2 to give SOLUTION
cDAB (a) Because the gas system is closed and at constant pressure and
JA = (xA1 − xA ) (3-18)
z − z1 temperature, mass transfer in the connecting tube is equimolar
counterdiffusion by molecular diffusion.
and The area for mass transfer through the tube, in cm2, is A =
3.14(0.1) 2 /4 = 7.85 × 10−3 cm2 . The total gas concentration (molar
cDBA density) is c = RT
P
= (82.06)(298)
1
= 4.09 × 10−5 mol/cm3. Take the
JB = (xB1 − xB ) (3-19)
z − z1 reference plane at end 1 of the connecting tube. Applying (3-18) to
N2 over the length of the tube, The factor (1 − xA ) accounts for the bulk-flow effect. For a
mixture dilute in A, the bulk-flow effect is negligible or
cDN2 , H2
n N2 = [(xN2 ) 1 − (xN2 ) 2 ]A small. In mixtures more concentrated in A, the bulk-flow
z2 − z1
(4.09 × 10−5 )(0.784)(0.80 − 0.25)
effect can be appreciable. For example, in an equimolar
= (7.85 × 10−3 ) mixture of A and B, (1 − xA ) = 0.5 and the molar mass-
15
= 9.23 × 10−9 mol/s in the positive z-direction transfer flux of A is twice the ordinary molecular-diffusion
flux.
n H2 = 9.23 × 10−9 mol/s in the negative z-direction
For the stagnant component, B, (3-13) becomes
(b) For equimolar counterdiffusion, the molar-average velocity of dxB
the mixture, vM, is 0. Therefore, from (3-9), species velocities are 0 = xB NA − cDBA (3-28)
dz
equal to species diffusion velocities. Thus,
or
JN n N2
vN2 = ( vN2 ) D = 2 =
xB NA = cDBA
dxB
(3-29)
cN2 AcxN2
dz
9.23 × 10−9
= Thus, the bulk-flow flux of B is equal but opposite to its
[(7.85 × 10−3 )(4.09 × 10−5 )xN2 ]
0.0287 diffusion flux.
= in the positive z-direction
xN2 At quasi-steady-state conditions, that is, with no accumu-
lation, and with constant molar density, (3-27) becomes in
Similarly, integral form:
0.0287 z
vH2 = in the negative z-direction cDAB xA dxA
xH2 dz = − (3-30)
z1 NA xA1 1 − xA
Thus, species velocities depend on species mole fractions, as
which upon integration yields
follows:

vN2 , cm/s vH2 , cm/s cDAB 1 − xA
z, cm xN2 xH2 NA = ln (3-31)
z − z1 1 − xA1
0 (end 1) 0.800 0.200 0.0351 −0.1435
5 0.617 0.383 0.0465 −0.0749 Rearrangement to give the mole-fraction variation as a func-
10 0.433 0.567 0.0663 −0.0506
tion of z yields
15 (end 2) 0.250 0.750 0.1148 −0.0383

NA (z − z 1 )
Note that species velocities vary across the length of the connect- xA = 1 − (1 − xA1 ) exp (3-32)
ing tube, but at any location, z, vM = 0. For example, at z = 10 cm, cDAB
from (3-8),
Thus, as shown in Figure 3.1b, the mole fractions are non-
v M = (0.433)(0.0663) + (0.567)(−0.0506) = 0 linear in distance.
An alternative and more useful form of (3-31) can be
derived from the definition of the log mean. When z = z 2 ,
(3-31) becomes
Unimolecular Diffusion

In unimolecular diffusion (UMD), mass transfer of compo- cDAB 1 − xA2
NA = ln (3-33)
nent A occurs through stagnant (nonmoving) component B. z2 − z1 1 − xA1
Thus,
The log mean (LM) of (1 − xA ) at the two ends of the stag-
NB = 0 (3-24) nant layer is
and
(1 − xA2 ) − (1 − xA1 )
N = NA (3-25) (1 − xA ) LM =
ln[(1 − xA2 )/(1 − xA1 )]
xA1 − xA2 (3-34)
Therefore, from (3-12), =
ln[(1 − xA2 )/(1 − xA1 )]
dxA
NA = xA NA − cDAB (3-26) Combining (3-33) with (3-34) gives
dz
which can be rearranged to a Fick’s-law form, cDAB (xA1 − xA2 ) cDAB (−xA )
NA = =
z 2 − z 1 (1 − xA ) LM (1 − xA ) LM z
(3-35)
cDAB dxA cDAB dxA cDAB (−xA )
NA = − =− (3-27) =
(1 − xA ) dz xB dz (xB ) LM z
EXAMPLE 3.2 From (3-32),
As shown in Figure 3.2, an open beaker, 6 cm in height, is filled xA = 1 − 0.869 exp(0.281 z) (1)
with liquid benzene at 25◦ C to within 0.5 cm of the top. A gentle Using (1), the following results are obtained:
breeze of dry air at 25◦ C and 1 atm is blown by a fan across the
mouth of the beaker so that evaporated benzene is carried away by z, cm xA xB
convection after it transfers through a stagnant air layer in the 0.0 0.1310 0.8690
beaker. The vapor pressure of benzene at 25◦ C is 0.131 atm. The 0.1 0.1060 0.8940
mutual diffusion coefficient for benzene in air at 25◦ C and 1 atm is 0.2 0.0808 0.9192
0.0905 cm2/s. Compute: 0.3 0.0546 0.9454
(a) The initial rate of evaporation of benzene as a molar flux in 0.4 0.0276 0.9724
mol/cm2-s 0.5 0.0000 1.0000
(b) The initial mole-fraction profiles in the stagnant air layer These profiles are only slightly curved.
(c) The initial fractions of the mass-transfer fluxes due to molecu- (c) From (3-27) and (3-29), we can compute the bulk flow terms,
lar diffusion xANA and xBNA, from which the molecular diffusion terms are
(d) The initial diffusion velocities, and the species velocities (rela- obtained.
tive to stationary coordinates) in the stagnant layer
(e) The time in hours for the benzene level in the beaker to drop xiN Ji
2 cm from the initial level, if the specific gravity of liquid ben- Bulk-Flow Flux, Molecular-Diffusion
zene is 0.874. Neglect the accumulation of benzene and air in mol/cm2-s × 106 Flux, mol/cm2-s × 106
the stagnant layer as it increases in height
z, cm A B A B
0.0 0.1360 0.9040 0.9040 −0.9040

SOLUTION 0.1 0.1100 0.9300 0.9300 −0.9300
Let A benzene, B air. 0.2 0.0840 0.9560 0.9560 −0.9560
0.3 0.0568 0.9832 0.9832 −0.9832
P 1 0.4 0.0287 1.0113 1.0113 −1.0113
c= = = 4.09 × 10−5 mol/cm3
RT (82.06)(298) 0.5 0.0000 1.0400 1.0400 −1.0400
(a) Take z1 0. Then z 2 − z 1 = z = 0.5 cm. From Dalton’s law,

assuming equilibrium at the liquid benzene–air interface, Note that the molecular-diffusion fluxes are equal but opposite,
and the bulk-flow flux of B is equal but opposite to its molecular-
pA1 0.131 diffusion flux, so that its molar flux, NB, is zero, making B (air)
xA1 = = = 0.131 xA2 = 0
P 1 stagnant.
0.131
(1 − xA ) LM = = 0.933 = (xB ) LM (d) From (3-6),
ln[(1 − 0)/(1 − 0.131)]
N NA 1.04 × 10−6
vM = = = = 0.0254 cm/s (2)
From (3-35), c c 4.09 × 10−5
From (3-9), the diffusion velocities are given by
(4.09 × 10−5 )(0.0905) 0.131
NA = = 1.04 × 10−6 mol/cm2 -s Ji Ji
0.5 0.933 vid = = (3)
ci xi c
−6
NA (z − z 1 ) (1.04 × 10 )(z − 0)
(b) = = 0.281 z From (3-10), the species velocities relative to stationary coordinates
cDAB (4.09 × 10−5 )(0.0905) are
vi = vid + v M (4)
Air 1 atm
25°C Using (2) to (4), we obtain
xA = 0 vid Ji
Mass Molecular-Diffusion Species Velocity,
transfer 0.5 cm
xA = PAs /P
z Velocity, cm/s cm/s
Interface
z, cm A B A B
6 cm
Liquid 0.0 0.1687 −0.0254 0.1941 0
Benzene 0.1 0.2145 −0.0254 0.2171 0
0.2 0.2893 −0.0254 0.3147 0
0.3 0.4403 −0.0254 0.4657 0
Beaker
0.4 0.8959 −0.0254 0.9213 0
0.5 ∞ −0.0254 ∞ 0
Figure 3.2 Evaporation of benzene from a beaker—Example 3.2.
Note that vB is zero everywhere, because its molecular-diffusion Table 3.1 Diffusion Volumes from Fuller,
velocity is negated by the molar-mean velocity. Ensley, and Giddings [J. Phys. Chem, 73,
(e) The mass-transfer flux for benzene evaporation can be equated 3679–3685 (1969)] for Estimating Binary Gas
to the rate of decrease in the moles of liquid benzene per unit cross Diffusivity by the Method of Fuller et al. [3]
section of the beaker. Letting z = distance down from the mouth of Atomic Diffusion Volumes Atomic
the beaker and using (3-35) with z = z, and Structural Diffusion-Volume Increments
cDAB (−xA ) L dz C 15.9 F 14.7

NA = = (5) H 2.31 Cl 21.0
z (1 − xA ) LM M L dt
O 6.11 Br 21.9
Separating variables and integrating, N 4.54 I 29.8
Aromatic ring −18.3 S 22.9
Heterocyclic ring −18.3
t
L (1 − xA ) LM z2
dt = t = z dz (6)
0 M L cDAB (−xA ) z1 Diffusion Volumes of Simple Molecules
He 2.67 CO 18.0
The coefficient of the integral on the right-hand side of (6) is Ne 5.98 CO2 26.7
constant at
Ar 16.2 N2O 35.9
Kr 24.5 NH3 20.7
0.874(0.933)
= 21,530 s/cm2 Xe 32.7 H2O 13.1
78.11(4.09 × 10−5 )(0.0905)(0.131)
z2 2.5 H2 6.12 SF6 71.3
z dz = z dz = 3 cm2 D2 6.84 Cl2 38.4
z1 0.5 N2 18.5 Br2 69.0
O2 16.3 SO2 41.8
From (6), t = 21,530(3) = 64,590 s or 17.94 h, which is a long time Air 19.7
because of the absence of turbulence.
3.2 DIFFUSION COEFFICIENTS

Diffusivities or diffusion coefficients are defined for a binary derived from experimental data:
mixture by (3-3) to (3-5). Measurement of diffusion coeffi-
cients must involve a correction for any bulk flow using 0.00143T 1.75
(3-12) and (3-13) with the reference plane being such that DAB = DBA = 1/2 1/3 1/3 (3-36)
there is no net molar bulk flow. P MAB [( V ) A + ( V ) B ]2
The binary diffusivities, DAB and DBA, are mutual or
binary diffusion coefficients. Other coefficients include Di M , where DAB is in cm2/s, P is in atm, T is in K,
the diffusivity of i in a multicomponent mixture; Dii , the
self-diffusion coefficient; and the tracer or interdiffusion
2
coefficient. In this chapter, and throughout this book, the MAB = (3-37)
focus is on the mutual diffusion coefficient, which will be (1/MA ) + (1/MB )
referred to as the diffusivity or diffusion coefficient.
and V = summation of atomic and structural diffusion
volumes from Table 3.1, which includes diffusion volumes
Diffusivity in Gas Mixtures
of some simple molecules.
As discussed by Poling, Prausnitz, and O’Connell [2], a Experimental values of binary gas diffusivity at 1 atm and
number of theoretical and empirical equations are available near-ambient temperature range from about 0.10 to 10.0 cm2/s.
for estimating the value of DAB = DBA in gases at low Poling, et al. [2] compared (3-36) to experimental data for
to moderate pressures. The theoretical equations, based 51 different binary gas mixtures at low pressures over a tem-
on Boltzmann’s kinetic theory of gases, the theorem of cor- perature range of 195–1,068 K. The average deviation was
responding states, and a suitable intermolecular energy- only 5.4%, with a maximum deviation of 25%. Only 9 of 69
potential function, as developed by Chapman and Enskog, estimated values deviated from experimental values by more
predict DAB to be inversely proportional to pressure and than 10%. When an experimental diffusivity is available at
almost independent of composition, with a significant in- values of T and P that are different from the desired condi-
crease for increasing temperature. Of greater accuracy and tions, (3-36) indicates that DAB is proportional to T 1.75 /P ,
ease of use is the following empirical equation of Fuller, which can be used to obtain the desired value. Some repre-
Schettler, and Giddings [3], which retains the form of the sentative experimental values of binary gas diffusivity are
Chapman–Enskog theory but utilizes empirical constants given in Table 3.2.
3.2 Diffusion Coefficients 73
Table 3.2 Experimental Binary Diffusivities of Some Gas Pairs 3.5

at 1 atm 3.0
1.0 2.5
2.0
1.8
Gas pair, A-B Temperature, K DAB, cm2/s 1.6
0.8
DABP/(DABP)LP
1.5
Air—carbon dioxide 317.2 0.177 1.4
1.3
Air—ethanol 313 0.145 0.6
Air—helium 317.2 0.765 0.9 1.2
Air—n-hexane 328 0.093 0.4

Tr 1.1
Air—water 313 0.288
Argon—ammonia 333 0.253 0.2 1.0
Argon—hydrogen 242.2 0.562
Argon—hydrogen 806 4.86 0.0
1.0 2.0 3.0 4.0 5.0 6.0
Argon—methane 298 0.202 Reduced Pressure, Pr
Carbon dioxide—nitrogen 298 0.167
Carbon dioxide—oxygen 293.2 0.153 Figure 3.3 Takahashi [4] correlation for effect of high pressure
Carbon dioxide—water 307.2 0.198 on binary gas diffusivity.
Carbon monoxide—nitrogen 373 0.318
Helium—benzene 423 0.610
Helium—methane 298 0.675 For binary mixtures of light gases, at pressures to about
Helium—methanol 423 1.032 10 atm, the pressure dependence on diffusivity is adequately
Helium—water 307.1 0.902 predicted by the simple inverse relation (3-36), that is, PDAB =
Hydrogen—ammonia 298 0.783 a constant for a given temperature and gas mixture. At higher
Hydrogen—ammonia 533 2.149 pressures, deviations from this relation are handled in a man-
Hydrogen—cyclohexane 288.6 0.319 ner somewhat similar to the modification of the ideal-gas law
Hydrogen—methane 288 0.694 by the compressibility factor based on the theorem of corre-
Hydrogen—nitrogen 298 0.784 sponding states. Although few reliable experimental data are
Nitrogen—benzene 311.3 0.102
available at high pressure, Takahasi [4] has published a tenta-
Nitrogen—cyclohexane 288.6 0.0731
tive corresponding-states correlation, shown in Figure 3.3,
Nitrogen—sulfur dioxide 263 0.104
Nitrogen—water 352.1 0.256 patterned after an earlier correlation for self-diffusivities by
Oxygen—benzene 311.3 0.101 Slattery [5]. In the Takahashi plot, DABP/(DABP)LP is given
Oxygen—carbon tetrachloride 296 0.0749 as a function of reduced temperature and pressure, where
Oxygen—cyclohexane 288.6 0.0746 (DABP)LP is at low pressure where (3-36) applies. Mixture-
Oxygen—water 352.3 0.352 critical temperature and pressure are molar-average values.
Thus, a finite effect of composition is predicted at high pres-
From Marrero, T. R., and E. A. Mason, J. Phys. Chem. Ref. Data, 1, 3–118
sure. The effect of high pressure on diffusivity is important in
(1972).
supercritical extraction, discussed in Chapter 11.
EXAMPLE 3.3
EXAMPLE 3.4
Estimate the diffusion coefficient for the system oxygen (A)/
benzene (B) at 38◦ C and 2 atm using the method of Fuller et al. Estimate the diffusion coefficient for a 25/75 molar mixture of argon
and xenon at 200 atm and 378 K. At this temperature and 1 atm, the
SOLUTION diffusion coefficient is 0.180 cm2/s. Critical constants are
From (3-37), Tc, K Pc, atm

2 Argon 151.0 48.0
MAB = = 45.4
(1/32) + (1/78.11) Xenon 289.8 58.0

From Table 3.1, ( V ) A = 16.3 and ( V ) B = 6(15.9) +
SOLUTION
6(2.31) − 18.3 = 90.96
From (3-36), at 2 atm and 311.2 K, Calculate reduced conditions:
0.00143(311.2) 1.75 Tc = 0.25(151) + 0.75(289.8) = 255.1 K;
DAB = DBA = = 0.0495 cm2 /s Tr = T /Tc = 378/255.1 = 1.48
(2)(45.4) 1/2 [16.31/3 + 90.961/3 ]2
Pc = 0.25(48) + 0.75(58) = 55.5;
At 1 atm, the predicted diffusivity is 0.0990 cm2/s, which is about Pr = P/Pc = 200/55.5 = 3.6
2% below the experimental value of 0.101 cm2/s in Table 3.2. The
experimental value for 38◦ C can be extrapolated by the temperature DAB P
From Figure 3.3, = 0.82
dependency of (3-36) to give the following prediction at 200◦ C: ( DAB P) LP

200 + 273.2 1.75 (DAB P)LP DAB P (0.180)(1)
DAB at 200◦ C and 1 atm = 0.102 DAB = = (0.82)
38 + 273.2 P (DAB P)LP 200
= 0.212 cm2 /s = 7.38 × 10−4 cm/s
Diffusivity in Liquid Mixtures dilute conditions to more concentrated conditions, exten-

sions of (3-38) have been restricted to binary liquid mixtures
Diffusion coefficients in binary liquid mixtures are difficult dilute in A, up to 5 and perhaps 10 mol%.
to estimate because of the lack of a rigorous model for the One such extension, which gives reasonably good
liquid state. An exception is the case of a dilute solute (A) of predictions for small solute molecules, is the empirical
very large, rigid, spherical molecules diffusing through a Wilke–Chang [6] equation:
stationary solvent (B) of small molecules with no slip of the
solvent at the surface of the solute molecules. The resulting
7.4 × 10−8 (B MB ) 1/2 T
relation, based on the hydrodynamics of creeping flow to ( DAB ) ∞ = (3-39)
describe drag, is the Stokes–Einstein equation: B v0.6
A
RT where the units are cm2/s for DAB; cP (centipoises) for the
( DAB ) ∞ = (3-38)
6B RA NA solvent viscosity, B ; K for T; and cm3/mol for vA, the liquid
molar volume of the solute at its normal boiling point. The
where RA is the radius of the solute molecule and NA is parameter B is an association factor for the solvent, which
Avagadro’s number. Although (3-38) is very limited in its is 2.6 for water, 1.9 for methanol, 1.5 for ethanol, and 1.0 for
application to liquid mixtures, it has long served as a starting unassociated solvents such as hydrocarbons. Note that the
point for more widely applicable empirical correlations for effects of temperature and viscosity are identical to the pre-
the diffusivity of solute (A) in solvent (B), where both A and diction of the Stokes–Einstein equation, while the effect of
B are of the same approximate molecular size. Unfortu- the radius of the solute molecule is replaced by vA, which
nately, unlike the situation in binary gas mixtures, DAB = can be estimated by summing the atomic contributions in
DBA in binary liquid mixtures can vary greatly with compo- Table 3.3, which also lists values of vA for dissolved light
sition as shown in Example 3.7. Because the Stokes– gases. Some representative experimental values of diffusivity
Einstein equation does not provide a basis for extending in dilute binary liquid solutions are given in Table 3.4.
Table 3.3 Molecular Volumes of Dissolved Light Gases and Atomic Contributions for Other Molecules at the
Normal Boiling Point
Atomic Volume Atomic Volume

(m3/kmol) × 103 (m3/kmol) × 103
C 14.8 Ring
H 3.7 Three-membered, as in −6
O (except as below) 7.4 ethylene oxide
Doubly bonded as carbonyl 7.4 Four-membered −8.5
Coupled to two other elements: Five-membered −11.5
In aldehydes, ketones 7.4 Six-membered −15
In methyl esters 9.1 Naphthalene ring −30
In methyl ethers 9.9 Anthracene ring −47.5
In ethyl esters 9.9
Molecular Volume
In ethyl ethers 9.9
(m3 /kmol) × 103
In higher esters 11.0
In higher ethers 11.0 Air 29.9
In acids (—OH) 12.0 O2 25.6
Joined to S, P, N 8.3 N2 31.2
N Br2 53.2
Doubly bonded 15.6 Cl2 48.4
In primary amines 10.5 CO 30.7
In secondary amines 12.0 CO2 34.0
Br 27.0 H2 14.3
Cl in RCHClR 24.6 H2O 18.8
Cl in RCl (terminal) 21.6 H2S 32.9
F 8.7 NH3 25.8
I 37.0 NO 23.6
S 25.6 N2O 36.4
P 27.0 SO2 44.8
Source: G. Le Bas, The Molecular Volumes of Liquid Chemical Compounds, David McKay, New York (1915).
Table 3.4 Experimental Binary Liquid Diffusivities for Solutes, equation with experimental values for nonaqueous solutions.
A, at Low Concentrations in Solvents, B For a dilute solution of one normal paraffin (C5 to C32) in
another (C5 to C16),
Diffusivity,
Solvent, Solute, Temperature, DAB,
B A K cm2 /s × 105 T 1.47 B
( DAB ) ∞ = 13.3 × 10−8 (3-40)
Water Acetic acid 293 1.19 vA0.71
Aniline 293 0.92

Carbon dioxide 298 2.00 where
Ethanol 288 1.00
10.2
Methanol 288 1.26 = − 0.791 (3-41)
Ethanol Allyl alcohol 293 0.98
vA
Benzene 298 1.81
and the other variables have the same units as in (3-39).
Oxygen 303 2.64
For general nonaqueous solutions,
Pyridine 293 1.10
Water 298 1.24
Benzene Acetic acid 298 2.09
T 1.29 0.5
B /0.42
A
( DAB ) ∞ = 1.55 × 10−8 (3-42)
Cyclohexane 298 2.09 0.92
B vB
0.23
Ethanol 288 2.25

n-Heptane 298 2.10 where is the parachor, which is defined as
Toluene 298 1.85
n-Hexane Carbon tetrachloride 298 3.70 = v 1/4 (3-43)
Methyl ethyl ketone 303 3.74
Propane 298 4.87 When the units of the liquid molar volume, v, are cm3/mol
Toluene 298 4.21 and the surface tension, , are g/s2 (dynes/cm), then the units
Acetone Acetic acid 288 2.92 of the parachor are cm3-g1/4/s1/2-mol. Normally, at near-
Formic acid 298 3.77 ambient conditions, is treated as a constant, for which an
Nitrobenzene 293 2.94 extensive tabulation is available from Quayle [8], who also
Water 298 4.56 provides a group-contribution method for estimating para-
chors for compounds not listed. Table 3.5 gives values of
From Poling et al. [2]. parachors for a number of compounds, while Table 3.6 con-
tains structural contributions for predicting the parachor in
the absence of data.
EXAMPLE 3.5 The following restrictions apply to (3-42):
Use the Wilke–Chang equation to estimate the diffusivity of aniline 1. Solvent viscosity should not exceed 30 cP.
(A) in a 0.5 mol% aqueous solution at 20◦ C. At this temperature,
the solubility of aniline in water is about 4 g/100 g of water or 2. For organic acid solutes and solvents other than water,
0.77 mol% aniline. The experimental diffusivity value for an infi- methanol, and butanols, the acid should be treated as a
nitely dilute mixture is 0.92 × 10−5 cm2 /s. dimer by doubling the values of A and vA.
3. For a nonpolar solute in monohydroxy alcohols, val-
SOLUTION ues of vB and B should be multiplied by 8B , where
the viscosity is in centipoise.
B = H2 O = 1.01 cP at 20◦ C
vA = liquid molar volume of aniline at its normal boiling Liquid diffusion coefficients for a solute in a dilute binary
point of 457.6 K = 107 cm /mol
3 system range from about 10−6 to 10−4 cm2/s for solutes of
B = 2.6 for water MB = 18 for water T = 293 K molecular weight up to about 200 and solvents with viscos-
ity up to about 10 cP. Thus, liquid diffusivities are five orders
From (3-39), of magnitude less than diffusivities for binary gas mixtures
(7.4 × 10−8 )[2.6(18)]0.5 (293)
at 1 atm. However, diffusion rates in liquids are not neces-
DAB = = 0.89 × 10−5 cm2 /s sarily five orders of magnitude lower than in gases because,
1.01(107) 0.6
as seen in (3-5), the product of the concentration (molar den-
This value is about 3% less than the experimental value for an infi- sity) and the diffusivity determines the rate of diffusion for a
nitely dilute solution of aniline in water. given concentration gradient in mole fraction. At 1 atm, the
molar density of a liquid is three times that of a gas and, thus,
More recent liquid diffusivity correlations due to Hayduk the diffusion rate in liquids is only two orders of magnitude
and Minhas [7] give better agreement than the Wilke–Chang lower than in gases at 1 atm.
Table 3.5 Parachors for Representative Compounds
Parachor, Parachor, Parachor,

cm3-g1/4/s1/2-mol cm3-g1/4/s1/2-mol cm3-g1/4/s1/2-mol
Acetic acid 131.2 Chlorobenzene 244.5 Methyl amine 95.9
Acetone 161.5 Diphenyl 380.0 Methyl formate 138.6
Acetonitrile 122 Ethane 110.8 Naphthalene 312.5
Acetylene 88.6 Ethylene 99.5 n-Octane 350.3
Aniline 234.4 Ethyl butyrate 295.1 1-Pentene 218.2
Benzene 205.3 Ethyl ether 211.7 1-Pentyne 207.0
Benzonitrile 258 Ethyl mercaptan 162.9 Phenol 221.3
n-Butyric acid 209.1 Formic acid 93.7 n-Propanol 165.4
Carbon disulfide 143.6 Isobutyl benzene 365.4 Toluene 245.5
Cyclohexane 239.3 Methanol 88.8 Triethyl amine 297.8
Source: Meissner, Chem. Eng. Prog., 45, 149–153 (1949).
Table 3.6 Structural Contributions for Estimating the Parachor
Carbon–hydrogen: R—[—CO—]—R (ketone)

C 9.0 R + R = 2 51.3
H 15.5 R + R = 3 49.0
CH3 55.5 R + R = 4 47.5
CH2 in —(CH2)n R + R = 5 46.3
n < 12 40.0 R + R = 6 45.3
n > 12 40.3 R + R = 7 44.1
—CHO 66
Alkyl groups
1-Methylethyl 133.3 O (not noted above) 20
1-Methylpropyl 171.9 N (not noted above) 17.5
1-Methylbutyl 211.7 S 49.1
2-Methylpropyl 173.3 P 40.5
1-Ethylpropyl 209.5 F 26.1
1,1-Dimethylethyl 170.4 Cl 55.2
1,1-Dimethylpropyl 207.5 Br 68.0
1,2-Dimethylpropyl 207.9 I 90.3
1,1,2-Trimethylpropyl 243.5 Ethylenic bonds:
C6H5 189.6 Terminal 19.1
2,3-position 17.7
Special groups: 3,4-position 16.3
—COO— 63.8
—COOH 73.8 Triple bond 40.6
—OH 29.8
—NH2 42.5 Ring closure:
—O— 20.0 Three-membered 12
—NO2 74 Four-membered 6.0
—NO3 (nitrate) 93 Five-membered 3.0
—CO(NH2) 91.7 Six-membered 0.8
Source: Quale [8].
EXAMPLE 3.6
Estimate the diffusivity of formic acid (A) in benzene (B) at 25◦ C However, because formic acid is an organic acid, A is doubled
and infinite dilution, using the appropriate correlation of Hayduk to 187.4.
and Minhas [7]. The experimental value is 2.28 × 10−5 cm2/s. From (3-42),

SOLUTION 2981.29 (205.30.5 /187.40.42 )
( DAB ) ∞ = 1.55 × 10−8
0.60.92 960.23
Equation (3-42) applies, with T = 298 K
= 2.15 × 10−5 cm2 /s
A = 93.7 cm3 -g1/4 /s1/2 -mol B = 205.3 cm3 -g1/4 /s1/2 -mol
B = 0.6 cP at 25◦ C vB = 96 cm3 /mol at 80◦ C which is within 6% of the the experimental value.
The Stokes–Einstein and Wilke–Chang equations predict xA A xB B

an inverse dependence of liquid diffusivity with viscosity. 0.5 1.116 0.5 1.201
The Hayduk–Minhas equations predict a somewhat smaller 0.6 1.066 0.4 1.269
dependence on viscosity. From data covering several orders 0.7 1.034 0.3 1.343
of magnitude variation of viscosity, the liquid diffusivity is 0.8 1.014 0.2 1.424
found to vary inversely with the viscosity raised to an expo- 0.9 1.003 0.1 1.511
nent closer to 0.5 than to 1.0. The Stokes–Einstein and 1.0 1.000 0.0 1.605
Wilke–Chang equations also predict that DAB B /T is a
Use the Vignes equations to estimate diffusion coefficients over
constant over a narrow temperature range. Because B de- the entire composition range.
creases exponentially with temperature, DAB is predicted to
increase exponentially with temperature. For example, for a
dilute solution of water in ethanol, the diffusivity of water SOLUTION
increases by a factor of almost 20 when the absolute temper- Using a spreadsheet to compute the derivatives in (3-45) and
ature is increased 50%. Over a wide temperature range, it is (3-46), which are found to be essentially equal at any composition,
preferable to express the effect of temperature on DAB by an and the diffusivities from the same equations, the following results
Arrhenius-type expression, are obtained with DAB = DBA at each composition. The calcula-
tions show a minimum diffusivity at a methanol mole fraction
−E of 0.30.
( DAB ) ∞ = A exp (3-44)
RT
xA DAB, cm2/s DBA, cm2/s
where, typically the activation energy for liquid diffusion, E, 0.20 1.10 × 10−5 1.10 × 10−5
is no greater than 6,000 cal/mol. 0.30 1.08 × 10−5 1.08 × 10−5
Equations (3-39), (3-40), and (3-42) for estimating diffu- 0.40 1.12 × 10−5 1.12 × 10−5
0.50 1.18 × 10−5 1.18 × 10−5
sivity in binary liquid mixtures only apply to the solute, A, in
0.60 1.28 × 10−5 1.28 × 10−5
a dilute solution of the solvent, B. Unlike binary gas mix-
0.70 1.38 × 10−5 1.38 × 10−5
tures in which the diffusivity is almost independent of com- 0.80 1.50 × 10−5 1.50 × 10−5
position, the effect of composition on liquid diffusivity is
complex, sometimes showing strong positive or negative If the diffusivity is assumed linear with mole fraction, the value at
deviations from linearity with mole fraction. xA = 0.50 is 1.625 × 10−5 , which is almost 40% higher than the
Based on a nonideal form of Fick’s law, Vignes [9] has predicted value of 1.18 × 10−5 .
shown that, except for strongly associated binary mixtures
such as chloroform/acetone, which exhibit a rare negative de-
viation from Raoult’s law, infinite-dilution binary diffusivi- Diffusivities of Electrolytes
ties, ( D) ∞, can be combined with mixture activity-coefficient In an electrolyte solute, the diffusion coefficient of the dis-
data or correlations thereof to predict liquid binary diffusion solved salt, acid, or base depends on the ions, since they are
coefficients DAB and DBA over the entire composition range. the diffusing entities. However, in the absence of an electric
The Vignes equations are: potential, only the molecular diffusion of the electrolyte
molecule is of interest. The infinite-dilution diffusivity of a
∂ ln A single salt in an aqueous solution in cm2/s can be estimated
DAB = xB
( DAB ) ∞ xA
( DBA ) ∞ 1+ (3-45)
∂ ln xA T,P from the Nernst–Haskell equation:

∂ ln B RT [(1/n + ) + (1/n − )]
DBA = ( DBA ) ∞
xA xB
( DAB ) ∞ 1+ (3-46) ( DAB ) ∞ = (3-47)
∂ ln xB T,P F 2 [(1/+ ) + (1/− )]
where
EXAMPLE 3.7 n+ and n − = valences of the cation and anion,
At 298 K and 1 atm, infinite-dilution diffusion coefficients for respectively
the methanol (A)/water (B) system are 1.5 × 10−5 cm2/s and + and − = limiting ionic conductances in (A/cm2)
1.75 × 10−5 cm2/s for AB and BA, respectively. (V/cm)(g-equiv/cm3), where A = amps and
Activity-coefficient data for the same conditions as estimated V = volts
from the UNIFAC method are as follows:
F = Faraday’s constant
xA A xB B = 96,500 coulombs/g-equiv
0.0 2.245 1.0 1.000 T = temperature, K
0.1 1.748 0.9 1.013 R = gas constant = 8.314 J/mol-K
0.2 1.470 0.8 1.044
0.3 1.300 0.7 1.087 Values of + and − at 25◦ C are listed in Table 3.7. At other
0.4 1.189 0.6 1.140 temperatures, these values are multiplied by T /334B ,
Table 3.7 Limiting Ionic Conductances in Water at 25°C, in EXAMPLE 3.8

(A/cm2)(V/cm)(g-equiv/cm3)
Estimate the diffusivity of KCl in a dilute solution of water
Anion − Cation + at 18.5◦ C. The experimental value is 1.7 × 10−5 cm2/s. At concen-
trations up to 2N, this value varies only from 1.5 × 10−5 to
OH− 197.6 H+ 349.8 1.75 × 10−5 cm2/s.
Cl− 76.3 Li+ 38.7
Br− 78.3 Na+ 50.1 SOLUTION
I− 76.8 K+ 73.5
NO− 71.4 NH+4 73.4 At 18.5◦ C, T /334 = 291.7/[(334)(1.05)] = 0.832. Using Table
3
ClO− 68.0 Ag+ 61.9 3.7, at 25◦ C, the corrected limiting ionic conductances are
4
HCO− 3 44.5 Tl+ 74.7 + = 73.5(0.832) = 61.2 and − = 76.3(0.832) = 63.5
HCO− 2 54.6 1
( 2 )Mg2+ 53.1
CH3 CO− 2 40.9 ( 12 )Ca2+ 59.5 From (3-47),
ClCH2 CO− 2 39.8 ( 12 )Sr2+ 50.5 (8.314)(291.7)[(1/1) + (1/1)]
CNCH2 CO− 41.8 ( 12 )Ba2+ 63.6 D∞ = = 1.62 × 10−5 cm2 /s
2 96,5002 [(1/61.2) + (1/63.5)]
CH3 CH2 CO− 2 35.8 ( 12 )Cu2+ 54
which is close to the experimental value.
CH3 (CH2 ) 2 CO− 2 32.6 ( 12 )Zn2+ 53
C6 H5 CO− 2 32.3 ( 13 )La3+ 69.5
HC2 O− 4 40.2 ( 3 )Co(NH3 ) 3+
1
6 102
( 12 )C2 O2− 74.2 Diffusivity of Biological Solutes in Liquids
4
( 12 )SO2− 4 80 For dilute, aqueous, nonelectrolyte solutions, the Wilke–Chang
( 13 )Fe(CN) 3− 6 101 equation (3-39) can be used for small solute molecules of
( 14 )Fe(CN) 4− 6 111 liquid molar volumes up to 500 cm3/mol, which corresponds
Source: Poling, Prausnitz, and O’Connell [2].
to molecular weights to almost 600. In biological applica-
tions, diffusivities of water-soluble protein macromolecules
having molecular weights greater than 1,000 are of interest.
where T and B are in kelvins and centipoise, respectively. In general, molecules with molecular weights to 500,000
As the concentration of the electrolyte increases, the diffu- have diffusivities at 25◦ C that range from 1 × 10−7 to
sivity at first decreases rapidly by about 10% to 20% and 8 × 10−7 cm2/s, which is two orders of magnitude smaller
then rises to values at a concentration of 2 N (normal) that than values of diffusivity for molecules with molecular
approximate the infinite-dilution value. Some representative weights less than 1,000. Data for many globular and fibrous
experimental values from Volume V of the International protein macromolecules are tabulated by Sorber [10] with a
Critical Tables are given in Table 3.8. few diffusivities given in Table 3.9. In the absence of data,
the following semiempirical equation given by Geankoplis
[11] and patterned after the Stokes–Einstein equation can be
Table 3.8 Experimental Diffusivities of Electrolytes in Aqueous
used:
Solutions
9.4 × 10−15 T
Concentration, Diffusivity, DAB, DAB = (3-48)
Solute Mol/L Temperature, ◦ C cm2 /s × 105 B ( MA ) 1/3
HCl 0.1 12 2.29 where the units are those of (3-39).
HNO3 0.05 20 2.62
Also of interest in biological applications are diffusivities
0.25 20 2.59
of small, nonelectrolyte molecules in aqueous gels contain-
H2SO4 0.25 20 1.63
ing up to 10 wt% of molecules such as certain polysaccha-
KOH 0.01 18 2.20 rides (agar), which have a great tendency to swell. Diffusiv-
0.1 18 2.15
ities are given by Friedman and Kraemer [12]. In general,
1.8 18 2.19
the diffusivities of small solute molecules in gels are not less
NaOH 0.05 15 1.49
than 50% of the values for the diffusivity of the solute in
NaCl 0.4 18 1.17 water, with values decreasing with increasing weight percent
0.8 18 1.19
of gel.
2.0 18 1.23
KCl 0.4 18 1.46
0.8 18 1.49 Diffusivity in Solids
2.0 18 1.58
Diffusion in solids takes place by different mechanisms de-
MgSO4 0.4 10 0.39
Ca(NO3)2 0.14 14 0.85
pending on the diffusing atom, molecule, or ion; the nature
of the solid structure, whether it be porous or nonporous,
Table 3.9 Experimental Diffusivities of Large Biological Proteins in Aqueous Solutions
Diffusivity, DAB,
Protein MW Configuration Temperature, °C cm2/s × 105
Bovine serum albumin 67,500 globular 25 0.0681
–Globulin, human 153,100 globular 20 0.0400
Soybean protein 361,800 globular 20 0.0291
Urease 482,700 globular 25 0.0401
Fibrinogen, human 339,700 fibrous 20 0.0198
Lipoxidase 97,440 fibrous 20 0.0559
crystalline, or amorphous; and the type of solid material, mechanisms or combinations thereof may take place:
whether it be metallic, ceramic, polymeric, biological, or
1. Ordinary molecular diffusion through pores, which
cellular. Crystalline materials may be further classified ac-
present tortuous paths and hinder the movement of
cording to the type of bonding, as molecular, covalent, ionic,
large molecules when their diameter is more than 10%
or metallic, with most inorganic solids being ionic. How-
of the pore diameter
ever, ceramic materials can be ionic, covalent, or most often
a combination of the two. Molecular solids have relatively 2. Knudsen diffusion, which involves collisions of diffus-
weak forces of attraction among the atoms or molecules. In ing gaseous molecules with the pore walls when the
covalent solids, such as quartz silica, two atoms share two or pore diameter and pressure are such that the molecular
more electrons equally. In ionic solids, such as inorganic mean free path is large compared to the pore diameter
salts, one atom loses one or more of its electrons by transfer 3. Surface diffusion involving the jumping of molecules,
to one or more other atoms, thus forming ions. In metals, adsorbed on the pore walls, from one adsorption site to
positively charged ions are bonded through a field of elec- another based on a surface concentration-driving force
trons that are free to move. Unlike diffusion coefficients in 4. Bulk flow through or into the pores
gases and low-molecular-weight liquids, which each cover a
range of only one or two orders of magnitude, diffusion co- When treating diffusion of solutes in porous materials
efficients in solids cover a range of many orders of magni- where diffusion is considered to occur only in the fluid in the
tude. Despite the great complexity of diffusion in solids, pores, it is common to refer to an effective diffusivity, Deff,
Fick’s first law can be used to describe diffusion if a mea- which is based on (1) the total cross-sectional area of the
sured diffusivity is available. However, when the diffusing porous solid rather than the cross-sectional area of the pore
solute is a gas, its solubility in the solid must also be known. and (2) on a straight path, rather than the pore path, which
If the gas dissociates upon dissolution in the solid, the may be tortuous. If pore diffusion occurs only by ordinary
concentration of the dissociated species must be used in molecular diffusion, Fick’s law (3-3) can be used with an
Fick’s law. In this section, many of the mechanisms of diffu- effective diffusivity. The effective diffusivity for a binary
sion in solids are mentioned, but because they are exceed- mixture can be expressed in terms of the ordinary diffusion
ingly complex to quantify, the mechanisms are considered coefficient, DAB, by
only qualitatively. Examples of diffusion in solids are con- DAB
sidered, together with measured diffusion coefficients that Deff = (3-49)

can be used with Fick’s first law. Emphasis is on diffusion of
gas and liquid solutes through or into the solid, but move- where is the fractional porosity (typically 0.5) of the solid
ment of the atoms, molecules, or ions of the solid through it- and is the pore-path tortuosity (typically 2 to 3), which is
self is also considered. the ratio of the pore length to the length if the pore were
straight in the direction of diffusion. The effective diffusivity
is either determined experimentally, without knowledge of
Porous Solids
the porosity or tortuosity, or predicted from (3-49) based on
When solids are porous, predictions of the diffusivity of measurement of the porosity and tortuosity and use of the
gaseous and liquid solute species in the pores can be made. predictive methods for ordinary molecular diffusivity. As an
These methods are considered only briefly here, with details example of the former, Boucher, Brier, and Osburn [13] mea-
deferred to Chapters 14, 15, and 16, where applications are sured effective diffusivities for the leaching of processed soy-
made to membrane separations, adsorption, and leaching. This bean oil (viscosity = 20.1 cP at 120◦F) from 1/16-in.-thick
type of diffusion is also of great importance in the analysis and porous clay plates with liquid tetrachloroethylene solvent.
design of reactors using porous solid catalysts. It is sufficient The rate of extraction was controlled by the rate of diffusion
to mention here that any of the following four mass-transfer of the soybean oil in the clay plates. The measured value of
Deff was 1.0 × 10−6 cm2/s. As might be expected from the Table 3.10 Diffusivities of Solutes in Crystalline Metals
effects of porosity and tortuosity, the effective value is about and Salts
one order of magnitude less than the expected ordinary mol-
Metal/Salt Solute T, °C D, cm2/s
ecular diffusivity, D, of oil in the solvent.
Ag Au 760 3.6 × 10−10
Crystalline Solids Sb 20 3.5 × 10−21
Sb 760 1.4 × 10−9
Diffusion through nonporous crystalline solids depends
Al Fe 359 6.2 × 10−14
markedly on the crystal lattice structure and the diffusing
Zn 500 2 × 10−9
entity. As discussed in Chapter 17 on crystallization, only
Ag 50 1.2 × 10−9
seven different lattice structures are possible. For the cubic
lattice (simple, body-centered, and face-centered), the dif- Cu Al 20 1.3 × 10−30
fusivity is the same in all directions (isotropic). In the six Al 850 2.2 × 10−9
other lattice structures (including hexagonal and tetragonal), Au 750 2.1 × 10−11
the diffusivity can be different in different directions Fe H2 10 1.66 × 10−9
(anisotropic). Many metals, including Ag, Al, Au, Cu, Ni, H2 100 1.24 × 10−7
Pb, and Pt, crystallize into the face-centered cubic lattice C 800 1.5 × 10−8
structure. Others, including Be, Mg, Ti, and Zn, form Ni H2 85 1.16 × 10−8
anisotropic, hexagonal structures. The mechanisms of diffu- H2 165 1.05 × 10−7
sion in crystalline solids include: CO 950 4 × 10−8
1. Direct exchange of lattice position by two atoms or W U 1727 1.3 × 10−11
ions, probably by a ring rotation involving three or AgCl Ag +
150 2.5 × 10−14
more atoms or ions Ag+ 350 7.1 × 10−8
2. Migration by small solutes through interlattice spaces Cl− 350 3.2 × 10−16
called interstitial sites KBr H2 600 5.5 × 10−4
Br2 600 2.64 × 10−4
3. Migration to a vacant site in the lattice
4. Migration along lattice imperfections (dislocations),
or gain boundaries (crystal interfaces)
Diffusion coefficients associated with the first three EXAMPLE 3.9
mechanisms can vary widely and are almost always at least
one order of magnitude smaller than diffusion coefficients in Gaseous hydrogen at 200 psia and 300◦ C is stored in a small,
low-viscosity liquids. As might be expected, diffusion by the 10-cm-diameter, steel pressure vessel having a wall thickness of
fourth mechanism can be faster than by the other three 0.125 in. The solubility of hydrogen in steel, which is proportional
to the square root of the hydrogen partial pressure in the gas, is
mechanisms. Typical experimental diffusivity values, taken
equal to 3.8 × 10−6 mol/cm3 at 14.7 psia and 300◦ C. The diffusiv-
mainly from Barrer [14], are given in Table 3.10. The diffu- ity of hydrogen in steel at 300◦ C is 5 × 10−6 cm2/s. If the inner sur-
sivities cover gaseous, ionic, and metallic solutes. The val- face of the vessel wall remains saturated at the existing hydrogen
ues cover an enormous 26-fold range. Temperature effects partial pressure and the hydrogen partial pressure at the outer sur-
can be extremely large. face is zero, estimate the time, in hours, for the pressure in the ves-
sel to decrease to 100 psia because of hydrogen loss by dissolving
Metals in and diffusing through the metal wall.
Important practical applications exist for diffusion of light

SOLUTION
gases through metals. To diffuse through a metal, a gas must
first dissolve in the metal. As discussed by Barrer [14], all Integrating Fick’s first law, (3-3), where A is H2 and B is the metal,
light gases do not dissolve in all metals. For example, assuming a linear concentration gradient, and equating the flux to
hydrogen dissolves in such metals as Cu, Al, Ti, Ta, Cr, W, Fe, the loss of hydrogen in the vessel,
Ni, Pt, and Pd, but not in Au, Zn, Sb, and Rh. Nitrogen dis- dn A DAB AcA
solves in Zr, but not in Cu, Ag, or Au. The noble gases do not − = (1)
dt z
dissolve in any of the common metals. When H2, N2, and O2
dissolve in metals, they dissociate and may react to form hy- Because pA = 0 outside the vessel, cA = cA = solubility pof
A at
A 0.5
drides, nitrides, and oxides, respectively. More complex mol- the inside wall surface in mol/cm3 and cA = 3.8 × 10−6 14.7 ,
where pA is the pressure of A in psia inside the vessel. Let pAo
ecules such as ammonia, carbon dioxide, carbon monoxide,
and n Ao be the initial pressure and moles of A, respectively, in the
and sulfur dioxide also dissociate. The following example vessel. Assuming the ideal-gas law and isothermal conditions,
illustrates how pressurized hydrogen gas can slowly leak
through the wall of a small, thin pressure vessel. n A = n Ao pA / pAo (2)
Differentiating (2) with respect to time, For both hydrogen and helium, diffusivities increase rapidly
dn A n Ao dpA with increasing temperature. At ambient temperature the
= (3)
dt pAo dt diffusivities are three orders of magnitude lower than in liq-
Combining (1) and (3), uids. At elevated temperatures the diffusivities approach
dpA DA A(3.8 × 10−6 ) pA0.5 pAo
those observed in liquids. Solubilities vary only slowly with
=− (4) temperature. Hydrogen is orders of magnitude less soluble
dt n Ao z(14.7) 0.5
in glass than helium. For hydrogen, the diffusivity is some-
Integrating and solving for t,
what lower than in metals. Diffusivities for oxygen are also
2n Ao z(14.7) 0.5 0.5
included in Table 3.11 from studies by Williams [17] and
t= pAo − pA0.5
3.8 × 10 DA ApAo
−6
Sucov [18]. At 1000◦ C, the two values differ widely be-
Assuming the ideal-gas law, cause, as discussed by Kingery, Bowen, and Uhlmann [19],
(200/14.7)[(3.14 × 103 )/6)] in the former case, transport occurs by molecular diffusion;
n Ao = = 0.1515 mol
82.05(300 + 273) while in the latter case, transport is by slower network diffu-
The mean-spherical shell area for mass transfer, A, is sion as oxygen jumps from one position in the silicate net-
3.14
work to another. The activation energy for the latter is much
A= [(10) 2 + (10.635) 2 ] = 336 cm2 larger than for the former (71,000 cal/mol versus 27,000
2
cal/mol). The choice of glass can be very critical in high-
The time for the pressure to drop to 100 psia is
vacuum operations because of the wide range of diffusivity.
2(0.1515)(0.125 × 2.54)(14.7) 0.5
t= (2000.5 − 1000.5 )
3.8 × 10−6 (5 × 10−6 )(336)(200)
Ceramics
= 1.2 × 106 s = 332 h
Diffusion rates of light gases and elements in crystalline
Silica and Glass ceramics are very important because diffusion must precede
chemical reactions and causes changes in the microstructure.
Another area of great interest is the diffusion of light gases Therefore, diffusion in ceramics has been the subject of
through various forms of silica, whose two elements, Si and numerous studies, many of which are summarized in
O, make up about 60% of the earth’s crust. Solid silica can Figure 3.4, taken from Kingery et al. [19], where diffusivity
exist in three principal crystalline forms (quartz, tridymite, is plotted as a function of the inverse of temperature in the
and cristobalite) and in various stable amorphous forms, high-temperature range. In this form, the slopes of the
including vitreous silica (a noncrystalline silicate glass or curves are proportional to the activation energy for diffu-
fused quartz). Table 3.11 includes diffusivities, D, and solu- sion, E, where
bilities as Henry’s law constants, H, at 1 atm for helium and
E
hydrogen in fused quartz as calculated from correlations of D = Do exp − (3-51)
experimental data by Swets, Lee, and Frank [15] and Lee RT
[16], respectively. The product of the diffusivity and the sol- An insert at the middle-right region of Figure 3.4 relates the
ubility is called the permeability, PM. Thus, slopes of the curves to activation energy. The diffusivity
PM = D H (3-50) curves cover a ninefold range from 10−6 to 10−15 cm2/s,
with the largest values corresponding to the diffusion of
Unlike metals, where hydrogen usually diffuses as the potassium in
-Al2 O3 and one of the smallest values for car-
atom, hydrogen apparently diffuses as a molecule in glass. bon in graphite. In general, the lower the diffusivity, the
higher is the activation energy. As discussed in detail by
Table 3.11 Diffusivities and Solubilities of Gases in Amorphous Kingery et al. [19], diffusion in crystalline oxides depends
Silica at 1 atm not only on temperature but also on whether the oxide is stoi-
chiometric or not (e.g., FeO and Fe0.95O) and on impurities.
Gas Temp, C Diffusivity, cm2/s Solubility mol/cm3-atm Diffusion through vacant sites of nonstoichiometric oxides
He 24 2.39 × 10−8 1.04 × 10−7 is often classified as metal-deficient or oxygen-deficient.
300 2.26 × 10−6 1.82 × 10−7 Impurities can hinder diffusion by filling vacant lattice or
500 9.99 × 10−6 9.9 × 10−8 interstitial sites.
1,000 5.42 × 10−5 1.34 × 10−7
H2 300 6.11 × 10−8 3.2 × 10−14 Polymers
500 6.49 × 10−7 2.48 × 10−13
Thin, dense, nonporous polymer membranes are widely
1,000 9.26 × 10−6 2.49 × 10−12
used to separate gas and liquid mixtures. As discussed in
O2 1,000 6.25 × 10−9
detail in Chapter 14, diffusion of gas and liquid species
(molecular)
through polymers is highly dependent on the type of poly-
1,000 9.43 × 10−15
mer, whether it be crystalline or amorphous and, if the latter,
(network)
glassy or rubbery. Commercial crystalline polymers are
Temperature, °C
–5
1716 1393 1145 977 828 727
10
K in β – Al2O3
10 –6
Co in CoO
(air)
10 –7 Fe in Fe0.95
O in Ca0.14 Zr0.86O1.86
10 –8 O in Y2O3
Mg in MgO Cr in Cr2O3
(single)
Diffusion coefficient, cm2/s
10 –9 Y in Y2O3
O in Cu2O
N
(Po2 = 14 kPa)
in
U
N
O in CoO O in Cu2O
10 –10 (Po2 = 20 kPa) (Po2 = 20 kPa)
763
382
191
Al in O in Ni0.68Fe2.32O4
Al2O3 (single) 133
10 –11 90.8
57.8
O in Al2O3
(poly) kJ/mol
Ni in NiO
10 –12 (air)
O in Al2O3
(single)
10 –13
C in
graphite O in UO2.00
O in TiO2
10 –14 (Po2 = 101 kPa)
Ca in CaO
O in MgO O fused SiO2 Figure 3.4 Diffusion coefficients for single-
O in Cr2O3 (Po2 = 101 kPa)
U in UO2.00 and polycrystalline ceramics.
10 –15 [From W.D. Kingery, H.K. Bowen, and D.R.
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 Uhlmann, Introduction to Ceramics, 2nd ed., Wiley
1/T × 1000/T, K –1 Interscience, New York (1976) with permission.]
about 20% amorphous. It is mainly through the amorphous where Ki, the equilibrium partition coefficient, is equal to the
regions that diffusion occurs. As with the transport of ratio of the concentration in the polymer to the concentration,
gases through metals, transport of gaseous species through ci, in the liquid adjacent to the polymer surface. The product
polymer membranes is usually characterized by the solution- KiDi is the liquid permeability.
diffusion mechanism of (3-50). Fick’s first law, in the fol- Values of diffusivity for light gases in four polymers, given
lowing integrated forms, is then applied to compute the mass in Table 14.6, range from 1.3 × 10−9 to 1.6 × 10−6 cm2 /s,
transfer flux. which is orders of magnitude less than for diffusion of the
same species in a gas.
Gas species:
Diffusivities of liquids in rubbery polymers have been
studied extensively as a means of determining viscoelastic
Hi Di PMi
Ni = ( pi − pi2 ) = ( pi − pi2 ) (3-52) parameters. In Table 3.12, taken from Ferry [20], diffusivi-
z2 − z1 1 z2 − z1 1 ties are given for different solutes in seven different rubber
polymers at near-ambient conditions. The values cover a
where pi is the partial pressure of the gas species at a poly- sixfold range, with the largest diffusivity being that for
mer surface. n-hexadecane in polydimethylsiloxane. The smallest diffu-
Liquid species: sivities correspond to the case where the temperature is
approaching the glass-transition temperature, where the
K i Di polymer becomes glassy in structure. This more rigid struc-
Ni = (ci − ci2 ) (3-53)
z2 − z1 1 ture hinders diffusion. In general, as would be expected,
Table 3.12 Diffusivities of Solutes in Rubbery Polymers
Diffusivity,
Polymer Solute Temperature, K cm2/s
Polyisobutylene n-Butane 298 1.19 × 10−9
i-Butane 298 5.3 × 10−10
n-Pentane 298 1.08 × 10−9
n-Hexadecane 298 6.08 × 10−10
Hevea rubber n-Butane 303 2.3 × 10−7
i-Butane 303 1.52 × 10−7
n-Pentane 303 2.3 × 10−7
n-Hexadecane 298 7.66 × 10−8
Polymethylacrylate Ethyl alcohol 323 2.18 × 10−10
Polyvinylacetate n-Propyl alcohol 313 1.11 × 10−12
n-Propyl chloride 313 1.34 × 10−12
Ethyl chloride 343 2.01 × 10−9
Ethyl bromide 343 1.11 × 10−9
Polydimethylsiloxane n-Hexadecane 298 1.6 × 10−6
1,4-Polybutadiene n-Hexadecane 298 2.21 × 10−7
Styrene-butadiene rubber n-Hexadecane 298 2.66 × 10−8
smaller molecules have higher diffusivities. A more detailed From (3-50),

study of the diffusivity of n-hexadecane in random styrene/
PM = D H = (160 × 10−11 )(0.54 × 10−4 )
butadiene copolymers at 25◦ C by Rhee and Ferry [21] shows = 86.4 × 10−15 mol/m-s-Pa
a large effect on diffusivity of fractional free volume in the p1 = 3.5 × 106 Pa p2 = 0.2 × 106 Pa
polymer.
Diffusion and permeability in crystalline polymers de- Membrane thickness = z 2 − z 1 = z = PM ( p1 − p2 )/N
pend on the degree of crystallinity. Polymers that are 100% 86.4 × 10−15 (3.5 × 106 − 0.2 × 106 )
crystalline permit little or no diffusion of gases and liquids. z =
[0.64(1000)/3600]
For example, the diffusivity of methane at 25◦ C in poly-
= 1.6 × 10−6 m = 1.6 m
oxyethylene oxyisophthaloyl decreases from 0.30 × 10−9 to
0.13 × 10−9 cm2 /s when the degree of crystallinity in- As discussed in Chapter 14, polymer membranes must be very
creases from 0 (totally amorphous) to 40% [22]. A measure thin to achieve reasonable gas permeation rates.
of crystallinity is the polymer density. The diffusivity of
methane at 25◦ C in polyethylene decreases from
0.193 × 10−6 to 0.057 × 10−6 cm2 /s when the specific grav-
ity increases from 0.914 (low density) to 0.964 (high den- Cellular Solids and Wood
sity) [22]. A plasticizer can cause the diffusivity to increase. As discussed by Gibson and Ashby [23], cellular solids
For example, when polyvinylchloride is plasticized with consist of solid struts or plates that form edges and faces of
40% tricresyl triphosphate, the diffusivity of CO at 27◦ C in- cells, which are compartments or enclosed spaces. Cellular
creases from 0.23 × 10−8 to 2.9 × 10−8 cm2 /s [22]. solids such as wood, cork, sponge, and coral exist in nature.
Synthetic cellular structures include honeycombs, and
foams (some with open cells) made from polymers, metals,
EXAMPLE 3.10 ceramics, and glass. The word cellulose means “full of little
Hydrogen diffuses through a nonporous polyvinyltrimethylsilane cells.”
membrane at 25◦ C. The pressures on the sides of the membrane are A widely used cellular solid is wood, whose annual world
3.5 MPa and 200 kPa. Diffusivity and solubility data are given in production of the order of 1012 kg is comparable to the pro-
Table 14.9. If the hydrogen flux is to be 0.64 kmol/m2-h, how thick duction of iron and steel. Chemically, wood consists of
in micrometers should the membrane be? lignin, cellulose, hemicellulose, and minor amounts of or-
ganic chemicals and elements. The latter are extractable, and
SOLUTION the former three, which are all polymers, give wood its struc-
ture. Green wood also contains up to 25 wt% moisture in the
Equation (3-52) applies. From Table 14.9,
cell walls and cell cavities. Adsorption or desorption of
D = 160 × 10−11 m2 /s H = S = 0.54 × 10−4 mol/m3 -Pa moisture in wood causes anisotropic swelling and shrinkage.
The structure of wood, which often consists of (1) highly Typical results are given by Sherwood [25] and Stamm [24].
elongated hexagonal or rectangular cells, called tracheids in For example, for beech with a swollen specific gravity of
softwood (coniferous species, e.g., spruce, pine, and fir) and 0.4, the diffusivity increases from a value of about
fibers in hardwood (deciduous or broad-leaf species, e.g., 1 × 10−6 cm2 /s at 10◦ C to 10 × 10−6 cm2 /s at 60°C.
oak, birch, and walnut); (2) radial arrays of rectangular-like
cells, called rays, which are narrow and short in softwoods
3.3 ONE-DIMENSIONAL, STEADY-STATE
but wide and long in hardwoods; and (3) enlarged cells with
large pore spaces and thin walls, called sap channels because
AND UNSTEADY-STATE, MOLECULAR
they conduct fluids up the tree. The sap channels are less DIFFUSION THROUGH STATIONARY MEDIA
than 3 vol% of softwood, but as much as 55 vol% of For conductive heat transfer in stationary media, Fourier’s
hardwood. law is applied to derive equations for the rate of heat transfer
Because the structure of wood is directional, many of its for steady-state and unsteady-state conditions in shapes such
properties are anisotropic. For example, stiffness and as slabs, cylinders, and spheres. Many of the results are plot-
strength are 2 to 20 times greater in the axial direction of the ted in generalized charts. Analogous equations can be de-
tracheids or fibers than in the radial and tangential directions rived for mass transfer, using simplifying assumptions.
of the trunk from which the wood is cut. This anisotropy ex- In one dimension, the molar rate of mass transfer of A in
tends to permeability and diffusivity of wood penetrants, a binary mixture with B is given by a modification of (3-12),
such as moisture and preservatives. According to Stamm which includes bulk flow and diffusion:
[24], the permeability of wood to liquids in the axial direc-
dxA
tion can be up to 10 times greater than in the transverse n A = xA (n A + n B ) − cDAB A (3-54)
direction. dz
Movement of liquids and gases through wood and wood If A is a dissolved solute undergoing mass transfer, but B is
products takes time during drying and treatment with preser- stationary, n B = 0. It is common to assume that c is a constant
vatives, fire retardants, and other chemicals. This movement and xA is small. The bulk-flow term is then eliminated and
takes place by capillarity, pressure permeability, and (3-54) accounts for diffusion only, becoming Fick’s first law:
diffusion. Nevertheless, wood is not highly permeable be-
dxA
cause the cell voids are largely discrete and lack direct n A = −cDAB A (3-55)
interconnections. Instead, communication among cells is dz
through circular openings spanned by thin membranes with Alternatively, (3-55) can be written in terms of concentration
submicrometer-sized pores, called pits, and to a smaller ex- gradient:
tent, across the cell walls. Rays give wood some permeabil-
dcA
ity in the radial direction. Sap channels do not contribute to n A = −DAB A (3-56)
permeability. All three mechanisms of movement of gases dz
and liquids in wood are considered by Stamm [24]. Only dif- This equation is analogous to Fourier’s law for the rate of
fusion is discussed here. heat conduction, Q:
The simplest form of diffusion is that of a water-soluble
solute through wood saturated with water, such that no di- dT
Q = −k A (3-57)
mensional changes occur. For the diffusion of urea, glycer- dz
ine, and lactic acid into hardwood, Stamm [24] lists diffu- Steady State
sivities in the axial direction that are about 50% of ordinary
liquid diffusivities. In the radial direction, diffusivities are For steady-state, one-dimensional diffusion, with constant
about 10% of the values in the axial direction. For example, DAB, (3-56) can be integrated for various geometries, the
at 26.7◦ C the diffusivity of zinc sulfate in water is most common results being analogous to heat conduction.
5 × 10−6 cm2 /s. If loblolly pine sapwood is impregnated
1. Plane wall with a thickness, z 2 − z 1 :
with zinc sulfate in the radial direction, the diffusivity is

found to be 0.18 × 10−6 cm2 /s [24]. cA1 − cA2
The diffusion of water in wood is more complex. Mois- n A = DAB A (3-58)
z2 − z1
ture content determines the degree of swelling or shrinkage.
Water is held in the wood in different ways: It may be 2. Hollow cylinder of inner radius r1 and outer radius r2,
physically adsorbed on cell walls in monomolecular layers, with diffusion in the radial direction outward:
condensed in preexisting or transient cell capillaries, or DAB (cA1 − cA2 )
absorbed in cell walls to form a solid solution. n A = 2L (3-59)
ln(r2 /r1 )
Because of the practical importance of lumber drying
rates, most diffusion coefficients are measured under drying or

conditions in the radial direction across the fibers. Results cA1 − cA2
n A = DAB ALM (3-60)
depend on temperature and swollen-volume specific gravity. r2 − r1
3.3 One-Dimensional, Steady-State and Unsteady-State, Molecular Diffusion through Stationary Media 85
where Rearranging and simplifying,

ALM = log mean of the areas 2r L at r1 and r2
(∂cA /∂z) z+z − (∂cA /∂z) z ∂cA
DAB = (3-67)
L = length of the hollow cylinder z ∂t
3. Spherical shell of inner radius r1 and outer radius r2, In the limit, as z → 0,
with diffusion in the radial direction outward:
∂cA ∂ 2 cA
4r1r2 DAB (cA1 − cA2 ) = DAB 2 (3-68)
nA = (3-61) ∂t ∂z
r2 − r1
or Equation (3-68) is Fick’s second law for one-dimensional
diffusion. The more general form, for three-dimensional rec-
cA1 − cA2
n A = DAB AGM (3-62) tangular coordinates, is
r2 − r1 2
where AGM = geometric mean of the areas 4r 2 at ∂cA ∂ cA ∂ 2 cA ∂ 2 cA
= DAB + + (3-69)
r1 and r2. ∂t ∂x2 ∂ y2 ∂z 2
When r1 /r2 < 2, the arithmetic mean area is no more For one-dimensional diffusion in the radial direction only,
than 4% greater than the log mean area. When r1 /r2 < 1.33, for cylindrical and spherical coordinates, Fick’s second law
the arithmetic mean area is no more than 4% greater than the becomes, respectively,
geometric mean area.
∂cA DAB ∂ ∂cA
= r (3-70)
∂t r ∂r ∂r
Unsteady State
and
Equation (3-56) is applied to unsteady-state molecular diffu-
∂cA DAB ∂ 2 ∂cA
sion by considering the accumulation or depletion of a = 2 r (3-71)
∂t r ∂r ∂r
species with time in a unit volume through which the species
is diffusing. Consider the one-dimensional diffusion of Equations (3-68) to (3-71) are analogous to Fourier’s sec-
species A in B through a differential control volume with dif- ond law of heat conduction where cA is replaced by temper-
fusion in the z-direction only, as shown in Figure 3.5. ature, T, and diffusivity, DAB, is replaced by thermal diffu-
Assume constant total concentration, c = cA + cB , constant sivity, = k/C P . The analogous three equations for heat
diffusivity, and negligible bulk flow. The molar flow rate of conduction for constant, isotropic properties are, respec-
species A by diffusion at the plane z = z is given by (3-56): tively:
2
∂cA ∂T ∂ T ∂2T ∂2T
n Az = −DAB A (3-63) = + + (3-72)
∂z z ∂t ∂x2 ∂ y2 ∂z 2

At the plane, z = z + z, the diffusion rate is ∂T ∂ ∂T
= r (3-73)
∂cA ∂t r ∂r ∂r
n Az+z = −DAB A (3-64)
∂z z+z ∂T ∂ ∂T
= 2 r2 (3-74)
The accumulation of species A in the control volume is ∂t r ∂r ∂r
∂cA
A z (3-65) Analytical solutions to these partial differential equations in
∂t either Fick’s law or Fourier’s law form are available for a
Since rate in − rate out = accumulation, variety of boundary conditions. Many of these solutions are

∂cA ∂cA ∂cA derived and discussed by Carslaw and Jaeger [26] and Crank
−DAB A + DAB A =A z [27]. Only a few of the more useful solutions are presented
∂z z ∂z z+z ∂t
here.
(3-66)
Flow in Accumulation Flow out Semi-infinite Medium
∂ cA ∂ cA ∂ cA
Consider the semi-infinite medium shown in Figure 3.6,
nAz = – DAB A
∂z z
A
∂t
dz nAz +∆z = – DAB A
∂z z+∆z which extends in the z-direction from z = 0 to z = ∞. The x
and y coordinates extend from −∞ to +∞, but are not of
interest because diffusion takes place only in the z-direction.
Thus, (3-68) applies to the region z ≥ 0. At time t ≤ 0, the
concentration is cAo for z ≥ 0. At t = 0, the surface of the
z z+∆z semi-infinite medium at z = 0 is instantaneously brought to
Figure 3.5 Unsteady-state diffusion through a differential the concentration cAs > cAo and held there for t > 0. There-
volume A dz. fore, diffusion into the medium occurs. However, because
Thus, using the Leibnitz√rule for differentiating the integral

of (3-76), with x = z/2 DAB t ,
z
Direction of ∂cA
diffusion n A = −DAB A
∂z z=0
(3-77)
cA − cAo z2
= DAB A √ s exp −
Figure 3.6 One-dimensional diffusion into a semi-infinite D t 4D t
AB AB z=0
medium. Thus,

DAB
n A |z=0 = A(cAs − cAo ) (3-78)
the medium is infinite in the z-direction, diffusion cannot t
extend to z = ∞ and, therefore, as z → ∞, cA = cAo for all We can also determine the total number of moles of
t ≥ 0. Because the partial differential equation (3-68) and its solute, NA , transferred into the semi-infinite medium by in-
one boundary (initial) condition in time and two boundary tegrating (3-78) with respect to time:
conditions in distance are linear in the dependent variable,
t t
cA, an exact solution can be obtained. Either the method of DAB dt
NA = n A |z=0 dt = A(cAs − cAo ) √
combination of variables [28] or the Laplace transform t
o
o
(3-79)
method [29] is applicable. The result, in terms of the frac- DAB t
tional accomplished concentration change, is = 2A(cAs − cAo )

cA − cAo z
= = erfc √ (3-75) EXAMPLE 3.11
cAs − cAo 2 DAB t
Determine how long it will take for the dimensionless concentra-
where the complementary error function, erfc, is related to tion change, = (cA − cAo )/(cAs − cAo ), to reach 0.01 at a depth
the error function, erf, by z = 1 m in a semi-infinite medium, which is initially at a solute
x concentration cAo , after the surface concentration at z = 0 increases
2 to cAs , for diffusivities representative of a solute diffusing through
e− d (3-76)
2
erfc(x) = 1 − erf(x) = 1 − √
0 a stagnant gas, a stagnant liquid, and a solid.
The error function is included in most spreadsheet programs SOLUTION

and handbooks, such as Handbook of Mathematical Func-
tions [30]. The variation of erf(x) and erfc(x) is as follows: For a gas, assume DAB = 0.1 cm2/s. We know that z = 1 m = 100 cm.
From (3-75) and (3-76),
x erf(x) erfc(x)
z
0 0.0000 1.0000 = 0.01 = 1 − erf √
2 DAB t
0.5 0.5205 0.4795
Therefore,
1.0 0.8427 0.1573
1.5 0.9661 0.0339 z
erf √ = 0.99
2.0 0.9953 0.0047 2 DAB t
∞ 1.0000 0.0000
From tables of the error function,

Equation (3-75) is used to compute the concentration in z
√ = 1.8214
the semi-infinite medium, as a function of time and distance 2 DAB t
from the surface, assuming no bulk flow. Thus, it applies most Solving,
rigorously to diffusion in solids, and also to stagnant liquids 2
100 1
and gases when the medium t= = 7,540 s = 2.09 h
√ is dilute in the diffusing solute. 1.8214(2) 0.10
In (3-75), when (z/2 DAB t) = 2, the complementary
error function is only 0.0047, which represents less than a In a similar manner, the times for typical gas, liquid, and solid
1% change in the ratio of the concentration change at z = z media are:
to the√change at z = 0. Thus, it is common to refer to
Semi-infinite
z = 4 DAB t as the penetration depth and to apply (3-75) to Medium DAB, cm2/s Time for = 0.01 at 1 m
media of finite thickness as long as the thickness is greater
Gas 0.10 2.09 h
than the penetration depth.
Liquid 1 × 10−5 2.39 year
The instantaneous rate of mass transfer across the surface
Solid 1 × 10−9 239 centuries
of the medium at z = 0 can be obtained by taking the deriv-
ative of (3-75) with respect to distance and substituting it These results show that molecular diffusion is very slow, espe-
into Fick’s first law applied at the surface of the medium. cially in liquids and solids. In liquids and gases, the rate of mass
transfer can be greatly increased by agitation to induce turbulent Center of slab Surface of slab
motion. For solids, it is best to reduce the diffusion path to as small 1.0 0
a dimension as possible by reducing the size of the solid. DAB t /a2 1.0
0.8 0.6 0.2
Medium of Finite Thickness with Sealed Edges

0.4
0.6 0.4
cAs – cAo = 1 – E
Consider a rectangular, parallelepiped medium of finite
cAs – cAo = E
thickness 2a in the z-direction, and either infinitely long
cAs – cA
cA – cAo
dimensions in the y- and x-directions or finite lengths of 2b 0.4 0.6
0.2
and 2c, respectively, in those directions. Assume that in Fig-
ure 3.7a the edges parallel to the z-direction are sealed, so
diffusion occurs only in the z-direction and initially the con- 0.2 0.1 0.8
centration of the solute in the medium is uniform at cAo . At 0.04 0.01

time t = 0, the two unsealed surfaces of the medium at z =
0 1.0
a are brought to and held at concentration cAs > cAo . Be- 0 0.2 0.4 0.6 0.8 1.0
cause of symmetry, ∂cA/∂z = 0 at z = 0. Assume constant z
a
DAB. Again (3-68) applies, and an exact solution can be ob-
tained because both (3-68) and the boundary conditions are Figure 3.8 Concentration profiles for unsteady-state diffusion in
a slab.
linear in cA. By the method of separation of variables [28] or
[Adapted from H.S. Carslaw and J.C. Jaeger, Conduction of Heat in
the Laplace transform method [29], the result from Carslaw
Solids, 2nd ed., Oxford University Press, London (1959).]
and Jaeger [26], in terms of the fractional, unaccomplished
concentration change, E, is
cA − cA 4 ∞
(−1) n (e.g., short times), they do not. However, in the latter case,
E =1− = s =
cAs − cAo n=0 (2n + 1) (3-80)
the solution for the semi-infinite medium applies for
(2n + 1)z DAB t/a 2 < 16 1
. A convenient plot of the exact solution is
× exp[−DAB (2n + 1) t/4a ] cos
2 2 2
given in Figure 3.8.
2a
The instantaneous rate of mass transfer across the surface
or, in terms of the complementary error function, of either unsealed face of the medium (i.e., at z = a) is ob-
cAs − cA ∞ tained by differentiating (3-80) with respect to z, evaluating
E =1− = = (−1) n the result at z = a, followed by substitution into Fick’s first
cAs − cAo
n=0
(3-81) law to give
(2n + 1)a − z (2n + 1)a + z
× erfc √ + erfc √
2 DAB t 2 DAB t 2DAB (cAs − cAo ) A
n A |z=a = ×
a
For large values of DABt/a2, which is referred to as the
∞ (3-82)
Fourier number for mass transfer, the infinite series solutions DAB (2n + 1) 2 2 t
exp −
of (3-80) and (3-81) converge rapidly, but for small values n=0
4a 2
b
y
z
b
a
x a c c
r
c x
c a
(a) Slab. Edges at x = + c and – c and (b) Cylinder. Two circular ends at x = + c
at y = +b and – b are sealed. and – c are sealed.
r
a
Figure 3.7 Unsteady-state diffusion in media

(c) Sphere of finite dimensions.
We can also determine the total number of moles transferred This equation is plotted in Figure 3.9. It is important to note
across either unsealed face by integrating (3-82) with respect that concentrations are in mass of solute per mass of dry
to time. Thus, solid or mass of solute/volume. This assumes that during dif-
t fusion the solid does not shrink or expand so that the mass of
8(cAs − cAo ) Aa
NA = n A |z=a dt = dry solid per unit volume of wet solid will remain constant.
o 2 Then, we can substitute a concentration in terms of mass or
∞
1 DAB (2n + 1) 2 2 t moles of solute per mass of dry solid, i.e., the moisture con-
× 1 − exp −
n=0
(2n + 1) 2 4a 2 tent on the dry basis.
When the edges of the slab in Figure 3.7a are not sealed,
(3-83) the method of Newman [31] can be used with (3-69) to
In addition, the average concentration of the solute in the determine concentration changes within the slab. In this
medium, cAavg , as a function of time, can be obtained in the method, E or Eavg is given in terms of the E values from the
case of a slab from: solution of (3-68) for each of the coordinate directions by
a
cAs − cAavg (1 − ) dz E = Ex E y Ez (3-86)
= o (3-84)
cAs − cAo a
Corresponding solutions for infinitely long, circular cylin-
Substitution of (3-80) into (3-84) followed by integration ders and spheres are available in Carslaw and Jaeger [26]
gives and are plotted in Figures 3.9, 3.10, and 3.11, respectively.
For a short cylinder, where the ends are not sealed, E or Eave
cAs − cAavg is given by the method of Newman as
E avgslab = (1 − ave ) slab =
cAs − cAo
E = Er E x (3-87)
8
∞
1 DAB (2n + 1) 2 2 t
= 2 exp −
n=0 (2n + 1) 2 4a 2 Some materials such as crystals and wood, have thermal
(3-85) conductivities and diffusivities that vary markedly with
direction. For these anisotropic materials, Fick’s second law
in the form of (3-69) does not hold. Although the general
1.0 anisotropic case is exceedingly complex, as shown in the
0.8
Ea , 2c
following example, the mathematical treatment is relatively
0.6 Eb ,
Ec (
slab 2a
simple when the principal axes of diffusivity coincide with
)
0.4 the coordinate system.
2b
0.3 E
r (c
yl
0.2 in
de
r)
2c
Axis of cylinder Surface of cylinder
0.10
Es
2a
(s
0.08 1.0 0
ph
2
er
0.06 D A Bt/a
Eavg = c s – c avg
e)
Ao
0.4
cA – cA
0.04
0.8 0.2
As
0.03
0.02 cAs – cAo = E

2a 0.6 0.4
0.2
cAs – cA
0.010
cA – cAo
0.008
0.4 0.6
0.006
0.004
0.003 0.1
0.2 0.8
0.002 0.04
0.01
0 1.0
0 0.2 0.4 0.6 0.8 1.0
0.001 r
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a
DABt/a2, DABt/b2, DABt/c2
Figure 3.10 Concentration profiles for unsteady-state diffusion
Figure 3.9 Average concentrations for unsteady-state diffusion. in a cylinder.
[Adapted from R.E. Treybal, Mass-Transfer Operations, 3rd ed., McGraw- [Adapted from H.S. Carslaw and J.C. Jaeger, Conduction of Heat in
Hill, New York (1980).] Solids, 2nd ed., Oxford University Press, London (1959).]
Center of sphere Surface of sphere Since this is the same form as (3-69) and since the boundary condi-
1.0 0 tions do not involve diffusivities, we can apply Newman’s method,
0.4 D ABt/a
2 using Figure 3.9, where concentration, cA, is replaced by weight-
percent moisture on a dry basis.
0.8 0.2 From (3-86) and (3-85),
0.2
cAave − cAs 5−2
E aveslab = E avgx E avg y E avgz = = = 0.167
cAo − cAs 20 − 2
cAs – cAo = E
0.6 0.4
cAs – cA
Let D = 1 × 10−5 cm2/s.
cA – cAo
0.4 0.6
z1 Direction (axial):
0.1 1/2 1/2
D 20 1 × 10−5
a1 = a = = 7.07 cm
Dz 2 2 × 10−5
0.2 0.8
Dt 1 × 10−5 t
0.04
0.01
= = 2.0 × 10−7 t, s
a12 7.072
0 1.0
0 0.2 0.4 0.6 0.8 1.0 y1 Direction:
r
1/2 1/2
a
D 10 1 × 10−5
Figure 3.11 Concentration profiles for unsteady-state diffusion in b1 = b = = 7.906 cm
Dy 2 4 × 10−6
a sphere.
Dt 1 × 10−5 t
[Adapted from H.S. Carslaw and J.C. Jaeger, Conduction of Heat in
2
= = 1.6 × 10−7 t, s
Solids, 2nd ed., Oxford University Press, London (1959).] b1 7.9062
x1-Direction:
1/2 1/2
D 5 1 × 10−5
EXAMPLE 3.12 c1 = c = = 3.953 cm
Dx 2 4 × 10−6
A piece of lumber, measuring 5 × 10 × 20 cm, initially contains Dt 1 × 10−5 t
20 wt% moisture. At time 0, all six faces are brought to an equilib- 2
= = 6.4 × 10−7 t, s
c1 3.9532
rium moisture content of 2 wt%. Diffusivities for moisture at 25◦ C
are 2 × 10−5 cm2/s in the axial (z) direction along the fibers and Use Figure 3.9 iteratively with assumed values of time in seconds
4 × 10−6 cm2/s in the two directions perpendicular to the fibers. to obtain values of Eavg for each of the three coordinates until
Calculate the time in hours for the average moisture content to drop (3-86) equals 0.167.
to 5 wt% at 25◦ C. At that time, determine the moisture content at
the center of the piece of lumber. All moisture contents are on a dry t, h t, s Eavgz Eavgy Eavg x Eavg
1 1 1
basis.
100 360,000 0.70 0.73 0.46 0.235
120 432,000 0.67 0.70 0.41 0.193
SOLUTION 135 486,000 0.65 0.68 0.37 0.164
In this case, the solid is anisotropic, with Dx = D y = 4 × 10−6 cm2/s Therefore, it takes approximately 136 h.
and Dz = 2 × 10−5 cm2/s, where dimensions 2c, 2b, and 2a in the For 136 h = 490,000 s, the Fourier numbers for mass transfer
x, y, and z directions are 5, 10, and 20 cm, respectively. Fick’s sec- are
ond law for an isotropic medium, (3-69), must be rewritten for this
anisotropic material as Dt (1 × 10−5 )(490,000)
2
= = 0.0980
a1 7.072
2
∂cA ∂ cA ∂ 2 cA ∂ 2 cA Dt (1 × 10−5 )(490,000)
= Dx + + Dz 2 (1) = = 0.0784
∂t ∂x 2 ∂y 2 ∂z 2
b1 7.9062
Dt (1 × 10−5 )(490,000)
as discussed by Carslaw and Jaeger [26].
2
= = 0.3136
To transform (1) into the form of (3-69), let c1 3.9532

From Figure 3.8, at the center of the slab,
D D D
x1 = x y1 = y z1 = z (2)
Dx Dx Dz
E center = E z1 E y1 E x1 = (0.945)(0.956)(0.605) = 0.547
cA − cAcenter 2 − cAcenter
where D is chosen arbitrarily. With these changes in variables, = s = = 0.547
(1) becomes cAs − cAo 2 − 20
Solving,
∂cA ∂ 2 cA ∂ 2 cA ∂ 2 cA
=D + + (3) cA at the center = 11.8 wt% moisture
∂t ∂ x12
∂ y12
∂z 12
3.4 MOLECULAR DIFFUSION Thus, the maximum liquid velocity, which occurs at z = 0, is
IN LAMINAR FLOW g2
(u y ) max = (3-90)
Many mass-transfer operations involve diffusion in fluids in 2
laminar flow. The fluid may be a film flowing slowly down a The bulk-average velocity in the liquid film is
vertical or inclined surface, a laminar boundary layer that
forms as the fluid flows slowly past a thin plate, or the fluid u y dz g2
ū y = 0 = (3-91)
may flow through a small tube or slowly through a large pipe 3
or duct. Mass transfer may occur between a gas and a liquid Thus, the film thickness for fully developed flow is indepen-
film, between a solid surface and a fluid, or between a fluid dent of location y and is
and a membrane surface.
3ū y 1/2 3 1/3
= = (3-92)
g 2g
Falling Liquid Film
where = liquid film flow rate per unit width of film, W.
Consider a thin liquid film, of a mixture of volatile A and
For film flow, the Reynolds number, which is the ratio of
nonvolatile B, falling in laminar flow at steady state down
the inertial force to the viscous force, is
one side of a vertical surface and exposed to pure gas, A, on
the other side of the film, as shown in Figure 3.12. The sur- 4r H ū y 4ū y 4
NRe = = = (3-93)
face is infinitely wide in the x-direction (normal to the page).
In the absence of mass transfer of A into the liquid film, the where rH = hydraulic radius = (flow cross section)/(wetted
liquid velocity in the z-direction, uz, is zero. In the absence of perimeter) = (W )/W = and, by the equation of continu-
end effects, the equation of motion for the liquid film in fully ity, = ū y .
developed laminar flow in the downward y-direction is As reported by Grimley [32], for NRe < 8 to 25, depend-
ing on the surface tension and viscosity, the flow in the film
d 2u y
+ g = 0 (3-88) is laminar and the interface between the liquid film and the
dz 2 gas is flat. The value of 25 is obtained with water. For 8 to
25 < NRe < 1,200, the flow is still laminar, but ripples and
Usually, fully developed flow, where uy is independent of the waves may appear at the interface unless suppressed by the
distance y, is established quickly. If is the thickness of the addition of wetting agents to the liquid.
film and the boundary conditions are u y = 0 at z = (no- For a flat liquid–gas interface and a small rate of mass
slip condition at the solid surface) and duy /dz = 0 at z = 0 transfer of A into the liquid film, (3-88) to (3-93) hold and
(no drag at the liquid–gas interface), (3-88) is readily inte- the film velocity profile is given by (3-89). Now consider a
grated to give a parabolic velocity profile: mole balance on A for an incremental volume of liquid film
z 2 of constant density, as shown in Figure 3.12. Neglect bulk
g2
uy = 1− (3-89) flow in the z-direction and axial diffusion in the y-direction.
2 Then, at steady state, neglecting accumulation or depletion
of A in the incremental volume,

z=δ ∂cA
z z = 0, y = 0 −DAB (y)(x) + u y cA | y (z)(x)
∂z z

∂cA
y = −DAB (y)(x) + u y cA | y+y (z)(x)
cA i (in liquid) ∂z z+z
Liquid (3-94)
Bulk
Gas flow
Rearranging and simplifying (3-94),

u y cA | y+y − u y cA | y (∂cA /∂z) z+z − (∂cA /∂z) z
= DAB
y y z
Liquid (3-95)
uy {z} Diffusion
film
element
of A In the limit, as z → 0 and y → 0,
z + ∆z
y +∆y
∂cA ∂ 2 cA
z
uy = DAB 2 (3-96)
∂y ∂z
cA {z} Substituting (3-89) into (3-96),
z 2 ∂c
Figure 3.12 Mass transfer from a gas into a falling, laminar g2 A ∂ 2 cA
1− = DAB 2 (3-97)
liquid film. 2 ∂y ∂z
3.4 Molecular Diffusion in Laminar Flow 91
This equation was solved by Johnstone and Pigford [33] For mass transfer, a composition driving force replaces
and later by Olbrich and Wild [34], for the following bound- T . As discussed later in this chapter, because composition
ary conditions: can be expressed in a number of ways, different mass-
cA = cAi at z = 0 for y > 0 transfer coefficients are defined. If we select cA as the dri-
cA = cA0 at y = 0 for 0 < z < ving force for mass transfer, we can write
∂cA /∂z = 0 at z = for 0 < y < L n A = kc A cA (3-105)
where L = height of the vertical surface. The solution of which defines a mass-transfer coefficient, kc, in mol/
Olbrich and Wild is in the form of an infinite series, giving time-area-driving force, for a concentration driving force.
cA as a function of z and y. However, of more interest is the Unfortunately, no name is in general use for (3-105).
average concentration at y = L, which, by integration, is For the falling laminar film, we take cA = cAi − c̄A ,
which varies with vertical location, y, because even though
1
c̄Ay = u y cAy dz (3-98) cAi is independent of y, the average film concentration, c̄A ,
ū y 0
increases with y. To derive an expression for kc, we equate
For the condition y = L, the result is (3-105) to Fick’s first law at the gas–liquid interface:
cAi − c̄AL
= 0.7857e−5.1213 + 0.09726e−39.661 ∂cA
cAi − cA0 (3-99) kc A(cAi − c̄A ) = −DAB A (3-106)
+ 0.036093−106.25 + · · · ∂z z=0
where Although this is the most widely used approach for defin-
2DAB L 8/3 8/3 ing a mass-transfer coefficient, in this case of a falling film it
= 2
= = (3-100) fails because (∂cA/∂z) at z = 0 is not defined. Therefore, for
3 ū y NRe NSc (/L) (/L) NPe M
this case we use another approach as follows. For an incre-
NSc = Schmidt number = mental height, we can write for film width W,
DAB
(3-101) n A = ū y W d c̄A = kc (cAi − c̄A )W dy
momentum diffusivity,/ (3-107)
=
mass diffusivity, DAB This defines a local value of kc, which varies with distance y
NPeM = NRe NSc = Peclet number for mass transfer because c̄A varies with y. An average value of kc, over a
4ū y (3-102) height L, can be defined by separating variables and inte-
= grating (3-107):
DAB c
L
The Schmidt number is analogous to the Prandtl number, k c dy ū y cAAL [d c̄A /(cAi − c̄A )]
used in heat transfer: kcavg = 0 = 0
L L
CP (/ ) momentum diffusivity ū y cAi − cA0 (3-108)
NPr = = = = ln
k (k/C P ) thermal diffusivity L cAi − c̄AL
The Peclet number for mass transfer is analogous to the
In general, the argument of the natural logarithm in
Peclet number for heat transfer:
4ū y C P (3-108) is obtained from the reciprocal of (3-99). For values
NPeH = NRe NPr = of in (3-100) greater than 0.1, only the first term in (3-99)
k
is significant (error is less than 0.5%). In that case,
Both Peclet numbers are ratios of convective transport to
molecular transport. ū y e5.1213
kcavg = ln (3-109)
The total rate of absorption of A from the gas into the L 0.7857
liquid film for height L and width W is
Since ln ex = x,
n A = ū y W ( c̄AL − cA0 ) (3-103) ū y
kcavg = (0.241 + 5.1213 ) (3-110)
Mass-Transfer Coefficients L
Mass-transfer problems involving fluids are most often In the limit, for large , using (3-100) and (3-102), (3-110)
solved using mass-transfer coefficients, analogous to heat- becomes
transfer coefficients. For the latter, Newton’s law of cooling DAB
kcavg = 3.414 (3-111)
defines a heat-transfer coefficient, h:
Q = h A T (3-104) In a manner suggested by the Nusselt number,
where NNu = h/k for heat transfer, where = a characteristic
length, we define a Sherwood number for mass transfer,
Q = rate of heat transfer
which for a falling film of characteristic length is
A = area for heat transfer (normal to the direction of
heat transfer) kcavg
NShavg = (3-112)
T = temperature-driving force for heat transfer DAB
From (3-111), NShavg = 3.414, which is the smallest value The error function is defined as
that the Sherwood number can have for a falling liquid film. z
2
e−t dt
2
The average mass-transfer flux of A is given by erf z = √ (3-118)
0
n Aavg
NAavg = = kcavg (cAi − c̄A ) mean (3-113) Using the Leibnitz rule with (3-116) to differentiate this in-
A
tegral function,
For values < 0.001 in (3-100), when the liquid-film
∂cA 3ū y
flow regime is still laminar without ripples, the time of con- = −(cAi − cA0 ) (3-119)
tact of the gas with the liquid is short and mass transfer is ∂z z=0 2 DAB y
confined to the vicinity of the gas–liquid interface. Thus, the Substituting (3-119) into (3-117) and introducing the Peclet
film acts as if it were infinite in thickness. In this limiting number for mass transfer from (3-102), we obtain an expres-
case, the downward velocity of the liquid film in the region sion for the local mass-transfer coefficient as a function of
of mass transfer is just u ymax , and (3-96) becomes distance down from the top of the wall:

∂cA ∂ 2 cA
u ymax = DAB 2 (3-114) 2
3DAB NPeM 3DAB
∂y ∂z kc = = (3-120)
8y 2y
Since from (3-90) and (3-91), u ymax = 3ū y /2, (3-114) can be
rewritten as The average value of kc over the height of the film, L, is ob-
tained by integrating (3-120) with respect to y, giving
∂cA 2DAB ∂ 2 cA
= (3-115) 6DAB 2
3DAB
∂y 3ū y ∂z 2 kcavg = = NPeM (3-121)
L 2L
where the boundary conditions are
Combining (3-121) with (3-112) and (3-102),
cA = cA0 for z > 0 and y > 0

cA = cAi for z = 0 and y > 0 3 4
NShavg = NPe M = (3-122)
cA = cA0 for large z and y > 0 2L
Equation (3-115) and the boundary conditions are equivalent where, by (3-108), the proper mean to use with kcavg is the log
to the case of the semi-infinite medium, as developed above. mean. Thus,
Thus, by analogy to (3-68), (3-75), and (3-76) the solution is (cAi − c̄A ) mean = (cAi − c̄A ) LM

cAi − cA z (cAi − cA0 ) − (cAi − cAL ) (3-123)
E =1− = = erf (3-116) =
cAi − cA0 2 2DAB y/3ū y ln[(cAi − cA0 )/(cAi − c̄AL )]
Assuming that the driving force for mass transfer in the film When ripples are present, values of kcavg and NShavg can be
is cAi − cA0 , we can use Fick’s first law at the gas–liquid considerably larger than predicted by these equations.
interface to define a mass-transfer coefficient: In the above development, asymptotic, closed-form solu-
tions are obtained with relative ease for large and small
∂cA values of , defined by (3-100). These limits, in terms of the
NA = −DAB = kc (cAi − cA0 ) (3-117)
∂z z=0 average Sherwood number, are shown in Figure 3.13. The
100
Sherwood number
Sh
ort
res
ide
10 Eq nce-t Gen
era
. (3 im l so
-12 e s lutio
2) olu n
tio
n
Long residence-time solution

Eq. (3-111)
1
0.001 0.01 0.1 1 10
η=
8/3 Figure 3.13 Limiting and general solutions for
( δ /L)NPe M mass transfer to a falling, laminar liquid film.
general solution for intermediate values of is not available Thus,

in closed form. Similar limiting solutions for large and small
cAi − cA0 kcavg A
values of appropriate parameters, usually dimensionless ln =
cAi − c̄AL ū y W
groups, have been obtained for a large variety of transport
and kinetic phenomena, as discussed by Churchill [35]. Solving for c̄AL ,

Often the two limiting cases can be patched together to pro- kcavg A
c̄AL = cAi − (cAi − cA0 ) exp −
vide a reasonable estimate of the intermediate solution, if a ū y W
single intermediate value is available from experiment or the L = 3 m, W = 1 m, A = W L = (1)(3) = 3 m2
general numerical solution. The procedure is discussed by cA0 = 0, cAi = 3.4 × 10−5 mol/cm3 = 3.4 × 10−2 kmol/m3
Churchill and Usagi [36]. The general solution of Emmert
(5.81 × 10−5 )(3)
and Pigford [37] to the falling, laminar liquid film problem is c̄AL = 3.4 × 10−2 1 − exp −
(0.0486)(1.15 × 10−4 )(1)
included in Figure 3.13.
= 3.4 × 10−2 kmol/m3
Thus, the exiting liquid film is saturated with CO2, which implies
EXAMPLE 3.13 equilibrium at the gas–liquid interface. From (3-103),
Water (B) at 25◦ C, in contact with pure CO2 (A) at 1 atm, flows as
n A = 0.0486(1.15 × 10−4 )(3.4 × 10−2 ) = 1.9 × 10−7 kmol/s
a film down a vertical wall 1 m wide and 3 m high at a Reynolds
number of 25. Using the following properties, estimate the rate of
adsorption of CO2 into water in kmol/s: Boundary-Layer Flow on a Flat Plate
−5
DAB = 1.96 × 10 cm /s; 2
= 1.0 g/cm ; 3
Consider the flow of a fluid (B) over a thin, flat plate parallel
L = 0.89 cP = 0.00089 kg/m-s with the direction of flow of the fluid upstream of the plate,
Solubility of CO2 in water at 1 atm and 25◦ C = 3.4 × as shown in Figure 3.14. A number of possibilities for mass
10−5 mol/cm3 . transfer of another species, A, into B exist: (1) The plate
might consist of material A, which is slightly soluble in B.
SOLUTION (2) Component A might be held in the pores of an inert solid
plate, from which it evaporates or dissolves into B. (3) The
From (3-93),
plate might be an inert, dense polymeric membrane, through
NRe 25(0.89)(0.001) kg which species A can pass into fluid B. Let the fluid velocity
= = = 0.00556
4 4 m−s profile upstream of the plate be uniform at a free-system ve-
locity of uo. As the fluid passes over the plate, the velocity ux
From (3-101),
in the direction of flow is reduced to zero at the wall, which
(0.89)(0.001) establishes a velocity profile due to drag. At a certain dis-
NSc = = = 454
DAB (1.0)(1,000)(1.96 × 10−5 )(10−4 ) tance z, normal to and out from the solid surface, the fluid ve-
locity is 99% of uo. This distance, which increases with
From (3-92),
increasing distance x from the leading edge of the plate, is
arbitrarily defined as the velocity boundary-layer thickness,
3(0.89)(0.001)(0.00556) 1/3
= = 1.15 × 10−4 m . Essentially all flow retardation occurs in the boundary
1.02 (1,000) 2 (9.807)
layer, as first suggested by Prandtl [38]. The buildup of this
From (3-90) and (3-91), ū y = (2/3)u ymax . Therefore, layer, the velocity profile in the layer, and the drag force can
be determined for laminar flow by solving the equations
2 (1.0)(1,000)(9.807)(1.15 × 10−4 ) 2
ū y = = 0.0486 m/s of continuity and motion (Navier–Stokes equations) for the
3 2(0.89)(0.001) x-direction. For a Newtonian fluid of constant density and
From (3-100), viscosity, in the absence of pressure gradients in the x- and
8/3
= = 6.13 Free
(25)(454)[(1.15 × 10−4 )/3] stream
uo
Therefore, (3-111) applies, giving uo
uo
3.41(1.96 × 10−5 )(10−4 ) Velocity

kcavg = = 5.81 × 10−5 m/s ux
ux
boundary
1.15 × 10−4 z uo δx layer
ux
To determine the rate of absorption, c̄AL must be determined. From
(3-103) and (3-113), x
Flat plate
( c̄AL − cA0 ) Figure 3.14 Laminar boundary-layer development for flow across
n A = ū y W ( c̄AL − cA0 ) = kcavg A
ln[(cAi − cA0 )/(cAi − cAL )] a flat plate.
y- (normal to the x–z plane) directions, these equations for If mass transfer begins at the leading edge of the plate and if
the region of the boundary layer are the concentration in the fluid at the solid–fluid interface is
constant, the additional boundary conditions are
∂u x ∂u z
+ =0 (3-124) cA = cAo at x = 0 for z > 0,
∂x ∂z
cA = cAi at z = 0 for x > 0,
∂u x ∂u x ∂ 2u x
ux + uz = (3-125) and cA = cAo at z = ∞ for x > 0
∂x ∂z ∂z 2
If the rate of mass transfer is low, the velocity profiles are
The boundary conditions are undisturbed. The solution to the analogous problem in heat
u x = u o at x = 0 for z > 0 u x = 0 at z = 0 for x > 0 transfer was first obtained by Pohlhausen [42] for NPr > 0.5,
u x = u o at z = ∞ for x > 0 u z = 0 at z = 0 for x > 0 as described in detail by Schlichting [40]. The results for
mass transfer are
The solution of (3-124) and (3-125) in the absence of heat NShx 0.332
and mass transfer, subject to these boundary conditions, was 1/3
= 0.5 (3-132)
first obtained by Blasius [39] and is described in detail by NRex NSc NRex
Schlichting [40]. The result in terms of a local friction factor, where
fx, a local shear stress at the wall, w x , and a local drag coef- xkcx
NShx = (3-133)
ficient at the wall, C Dx , is DAB
C Dx fx w 0.322 and the driving force for mass transfer is cAi − cAo .
= = x2 = 0.5 (3-126) The concentration boundary layer, where essentially all
2 2 uo NRex
of the resistance to mass transfer resides, is defined by
where
cAi − cA
xu o = 0.99 (3-134)
NRex = (3-127) cAi − cAo

and the ratio of the concentration boundary-layer thickness,
Thus, the drag is greatest at the leading edge of the plate, c , to the velocity boundary thickness, , is
where the Reynolds number is smallest. Average values of c / = 1/NSc
1/3
(3-135)
the drag coefficient are obtained by integrating (3-126) from
x = 0 to L, giving Thus, for a liquid boundary layer, where NSc > 1, the concen-
tration boundary layer builds up more slowly than the veloc-
C Davg f avg 0.664 ity boundary layer. For a gas boundary layer, where N SC ≈ 1,
= = (3-128)
2 2 ( NReL ) 0.5 the two boundary layers build up at about the same rate. By
The thickness of the velocity boundary layer increases with analogy to (3-130), the concentration profile is given by
3
distance along the plate: cAi − cA z z
= 1.5 − 0.5 (3-136)
4.96 cAi − cAo c c
= 0.5 (3-129)
x NRex Equation (3-132) gives the local Sherwood number. If
this expression is integrated over the length of the plate, L,
A reasonably accurate expression for the velocity profile
the average Sherwood number is found to be
was obtained by Pohlhausen [41], who assumed the empiri-
1/2 1/3
cal form ux = C1z + C2z3. NShavg = 0.664 NReL NSc (3-137)
If the boundary conditions,
where
Lkcavg
u x = 0 at z = 0 u x = u o at z = ∂u x /∂z = 0 at z = NShavg = (3-138)
DAB
are applied to evaluate C1 and C2, the result is
ux z z 3 EXAMPLE 3.14
= 1.5 − 0.5 (3-130) Air at 100◦ C, 1 atm, and a free-stream velocity of 5 m/s flows over
uo
a 3-m-long, thin, flat plate of naphthalene, causing it to sublime.
This solution is valid only for a laminar boundary layer,
(a) Determine the length over which a laminar boundary layer
which by experiment persists to NRex = 5 × 105 . persists.
When mass transfer of A into the boundary layer occurs,
the following species continuity equation applies at constant (b) For that length, determine the rate of mass transfer of naphtha-
lene into air.
diffusivity:
2 (c) At the point of transition of the boundary layer to turbulent
∂cA ∂cA ∂ cA flow, determine the thicknesses of the velocity and concentra-
ux + uz = DAB (3-131)
∂x ∂z ∂x2 tion boundary layers.
Assume the following values for physical properties: build up as shown at planes b, c, and d. In this region, the
Vapor pressure of napthalene = 10 torr
central core outside the boundary layer has a flat velocity
Viscosity of air = 0.0215 cP profile where the flow is accelerated over the entrance ve-
Molar density of air = 0.0327 kmol/m3 locity. Finally, at plane e, the boundary layer fills the tube.
Diffusivity of napthalene in air = 0.94 × 10−5 m2 /s From here the velocity profile is fixed and the flow is said to
be fully developed. The distance from the plane a to plane e
SOLUTION is the entry region.
For fully developed laminar flow in a straight, circular
(a) NRex = 5 × 105 for transition. From (3-127), tube, by experiment, the Reynolds number, NRe = D ū x /,
NRex [(0.0215)(0.001)](5 × 105 ) where ū x is the flow-average velocity in the axial direction,
x=L= = = 2.27 m x, and D is the inside diameter of the tube, must be less than
uo (5)[(0.0327)(29)]
2,100. For this condition, the equation of motion in the axial
at which transition to turbulent flow begins.
direction for horizontal flow and constant properties is
10(0.0327)
(b) cAo = 0 cAi = = 4.3 × 10−4 kmol/m3 ∂ ∂u x dP
760 r − =0 (3-139)
r ∂r ∂r dx
From (3-101),
[(0.0215)(0.001)]
where the boundary conditions are
NSc = = = 2.41 r = 0 (axis of the tube), ∂u x /∂r = 0
DAB [(0.0327)(29)](0.94 × 10−5 )
and r = rw (tube wall), u x = 0
From (3-137),
Equation (3-139) was integrated by Hagen in 1839 and
NShavg = 0.664(5 × 105 ) 1/2 (2.41) 1/3 = 630
Poiseuille in 1841. The resulting equation for the velocity
From (3-138), profile, expressed in terms of the flow-average velocity, is
630(0.94 × 10−5 ) 2
kcavg = = 2.61 × 10−3 m/s r
2.27 u x = 2ū x 1 − (3-140)
rw
For a width of 1 m,
A = 2.27 m2
or, in terms of the maximum velocity at the tube axis,
2
n A = kcavg A(cAi − cAo ) = 2.61 × 10−3 (2.27)(4.3 × 10−4 ) r
u x = u xmax 1 − (3-141)
= 2.55 × 10−6 kmol/s rw
(c) From (3-129), at x = L = 2.27 m, From the form of (3-141), the velocity profile is parabolic in
3.46(2.27)
nature.
= = 0.0111 m The shear stress, pressure drop, and Fanning friction fac-
(5 × 105 ) 0.5
From (3-135), tor are obtained from solutions to (3-139):

c =
0.0111
= 0.0083 m
∂u x 4ū x
(2.41) 1/3 w = − = (3-142)
∂r r=rw rw
dP 32ū x 2 f ū 2x
Fully Developed Flow in a Straight, Circular Tube − = = (3-143)
dx D2 D
Figure 3.15 shows the formation and buildup of a laminar
velocity boundary layer when a fluid flows from a vessel into with
a straight, circular tube. At the entrance, plane a, the veloc- 16
f = (3-144)
ity profile is flat. A velocity boundary layer then begins to NRe
Thickness of Edge of Fully developed

boundary layer boundary layer tube flow
Entrance
velocity = ux
Entrance
Figure 3.15 Buildup of a laminar

a b c d e velocity boundary layer for flow
x in a straight, circular tube.
The entry length to achieve fully developed flow is de- The Graetz solution of (3-147) for the temperature profile
fined as the axial distance, Le, from the entrance to the point or the concentration profile is in the form of an infinite
at which the centerline velocity is 99% of the fully devel- series, and can be obtained from (3-146) by the method of
oped flow value. From the analysis of Langhaar [43] for the separation of variables using the method of Frobenius. A
entry region, detailed solution is given by Sellars, Tribus, and Klein [45].
Le From the concentration profile, expressions for the mass-
= 0.0575NRe (3-145) transfer coefficient and the Sherwood number are obtained.
D
When x is large, the concentration profile is fully developed
Thus, at the upper limit of laminar flow, NRe = 2,100, Le/D = and the local Sherwood number, NShx , approaches a limiting
121, a rather large ratio. For NRe = 100, the ratio is only 5.75. value of 3.656. At the other extreme, when x is small such
In the entry region, Langhaar’s analysis shows the friction fac- that the concentration boundary layer is very thin and con-
tor is considerably higher than the fully developed flow value fined to a region where the fully developed velocity profile is
given by (3-144). At x = 0, f is infinity, but then decreases ex- linear, the local Sherwood number is obtained from the
ponentially with x, approaching the fully developed flow value classic Leveque [46] solution, presented by Knudsen and
at Le. For example, for NRe = 1,000, (3-144) gives f = 0.016, Katz [47]:
with Le/D = 57.5. In the region from x = 0 to x/D = 5.35, the
average friction factor from Langhaar is 0.0487, which is k cx D NPeM 1/3
NShx = = 1.077 (3-148)
about three times higher than the fully developed value. DAB (x/D)
In 1885, Graetz [44] obtained a theoretical solution to
the problem of convective heat transfer between the wall of where
a circular tube, held at a constant temperature, and a fluid
flowing through the tube in fully developed laminar flow. D ū x
Assuming constant properties and negligible conduction in NPeM = (3-149)
DAB
the axial direction, the energy equation, after substituting
(3-140) for ux, is The limiting solutions, together with the general Graetz
2
r ∂T k 1 ∂ ∂T solution, are shown in Figure 3.16, where it is seen that
2ū x 1 − = r (3-146) NShx = 3.656 is valid for NPeM /(x/D) < 4 and (3-148) is
rw ∂x C P r ∂r ∂r
valid for NPeM /(x/D) > 100. The two limiting solutions can
The boundary conditions are be patched together if one point of the general solution is
available where the two solutions intersect.
x = 0 (where heat transfer begins), T = T0 , for all r Over a length of tube where mass transfer occurs, an av-
x > 0, r = rw , T = Ti x > 0, r = 0, ∂ T /∂r = 0 erage Sherwood number can be derived by integrating the
general expression for the local Sherwood number. An em-
The analogous species continuity equation for mass pirical representation for that average, proposed by Hausen
transfer, neglecting bulk flow in the radial direction and dif- [48], is
fusion in the axial direction, is
2
r ∂cA 1 ∂ ∂cA 0.0668[NPeM /(x/D)]
2ū x 1 − = DAB r (3-147) NShavg = 3.66 + (3-150)
rw ∂x r ∂r ∂r 1 + 0.04[NPeM /(x/D)]2/3
with analogous boundary conditions. which is based on a log-mean concentration driving force.
100
Sherwood number
10
lut ion
General so
lution q ue so
Leve
Solution for
fully developed
concentration profile
1 Figure 3.16 Limiting and general solutions
1 10 100 1000
for mass transfer to a fluid in laminar flow in a
NPeM /(x /D)
straight, circular tube.
3.5 Mass Transfer in Turbulent Flow 97
EXAMPLE 3.15 3.5 MASS TRANSFER IN TURBULENT FLOW

Linton and Sherwood [49] conducted experiments on the dissolu- In the two previous sections, diffusion in stagnant media and
tion of cast tubes of benzoic acid (A) into water (B) flowing through in laminar flow were considered. For both cases, Fick’s law
the tubes in laminar flow. They obtained good agreement with pre- can be applied to obtain rates of mass transfer. A more com-
dictions based on the Graetz and Leveque equations. Consider a mon occurrence in engineering is turbulent flow, which is
5.23-cm-inside-diameter by 32-cm-long tube of benzoic acid, pre- accompanied by much higher transport rates, but for which
ceded by 400 cm of straight metal pipe of the same inside diameter
theory is still under development and the estimation of mass-
where a fully developed velocity profile is established. Pure water
enters the system at 25◦ C at a velocity corresponding to a Reynolds
transfer rates relies more on empirical correlations of exper-
number of 100. Based on the following property data at 25◦ C, esti- imental data and analogies with heat and momentum trans-
mate the average concentration of benzoic acid in the water leaving fer. A summary of the dimensionless groups used in these
the cast tube before a significant increase in the inside diameter of correlations and the analogies is given in Table 3.13.
the benzoic acid tube occurs because of dissolution. As shown by the famous dye experiment of Osborne
Solubility of benzoic acid in water = 0.0034 g/cm3 Reynolds [50] in 1883, a fluid in laminar flow moves paral-
Viscosity of water = 0.89 cP = 0.0089 g/cm-s lel to the solid boundaries in streamline patterns. Every par-
ticle of fluid moves with the same velocity along a stream-
Diffusivity of benzoic acid in water at infinite dilution
= 9.18 × 10−6 cm2 /s line and there are no fluid velocity components normal to
these streamlines. For a Newtonian fluid in laminar flow, the
SOLUTION momentum transfer, heat transfer, and mass transfer are by
0.0089 molecular transport, governed by Newton’s law of viscosity,
NSc = = 970 Fourier’s law of heat conduction, and Fick’s law of molecu-
(1.0)(9.18 × 10−6 )
D ū x lar diffusion, respectively.
NRe = = 100 In turbulent flow, the rates of momentum, heat, and mass

from which transfer are orders of magnitude greater than for molecular
(100)(0.0089) transport. This occurs because streamlines no longer exist
ū x = = 0.170 cm/s
(5.23)(1.0) and particles or eddies of fluid, which are large compared to
From (3-149), the mean free path of the molecules in the fluid, mix with
(5.23)(0.170) each other by moving from one region to another in fluctuat-
NPeM = = 9.69 × 104
9.18 × 10−6 ing motion. This eddy mixing by velocity fluctuations occurs
x 32 not only in the direction of flow but also in directions normal
= = 6.12
D 5.23 to flow, with the latter being of more interest. Momentum,
NPeM 9.69 × 104 heat, and mass transfer now occur by two parallel mecha-
= = 1.58 × 104
(x/D) 6.12 nisms: (1) molecular motion, which is slow; and (2) turbu-
From (3-150), lent or eddy motion, which is rapid except near a solid sur-
0.0668(1.58 × 104 ) face, where the flow velocity accompanying turbulence
NShavg = 3.66 + = 44 decreases to zero. Mass transfer by bulk flow may also occur
1 + 0.04(1.58 × 104 ) 2/3
as given by (3-1).
DAB (9.18 × 10−6 )
kcavg = NShavg = 44 = 7.7 × 10−5 cm/s In 1877, Boussinesq [51] modified Newton’s law of vis-
D 5.23
cosity to account for eddy motion. Analogous expressions
Using a log-mean driving force, were subsequently developed for turbulent-flow heat and
[(cAi − cA0 ) − (cAi − c̄Ax )] mass transfer. For flow in the x-direction and transport in
n A = ū x S( c̄Ax − cA0 ) = kcavg A
ln[(cAi − cA0 )/(cAi − c̄Ax )] the z-direction normal to flow, these expressions are written
where S is the cross-sectional area for flow. Simplifying, in the following forms in the absence of bulk flow in the
z-direction:
cAi − cA0 kcavg A
ln =
cAi − c̄Ax ū x S du x
cA0 = 0 and cAi = 0.0034 g/cm3 zx = −( + t ) (3-151)
dz
D2 (3.14)(5.23) 2
S= = = 21.5 cm2 and dT
4 4 qz = −(k + kt ) (3-152)
A = Dx = (3.14)(5.23)(32) = 526 cm2 dz

0.0034 (7.7 × 10−5 )(526) dcA
ln = NAz = −( DAB + Dt ) (3-153)
0.0034 − c̄Ax (0.170)(21.5) dz
= 0.0111 where the double subscript, zx, on the shear stress, , stands
0.0034 for x-momentum in the z-direction. The molecular contribu-
c̄Ax = 0.0034 − = 0.000038 g/cm3
e0.0111 tions, , k, and DAB, are molecular properties of the fluid and
Thus, the concentration of benzoic acid in the water leaving the cast depend on chemical composition, temperature, and pressure;
tube is far from saturation.
the turbulent contributions, t, kt, and Dt, depend on the
Table 3.13 Some Useful Dimensionless Groups
Name Formula Meaning Analogy
Fluid Mechanics
2FD Drag force
Drag Coefficient CD =
Au 2 Projected area × Velocity head
P D Pipe wall shear stress
Fanning Friction Factor f =
L 2ū 2 Velocity head
ū 2 Inertial force
Froude Number NFr =
gL Gravitational force
L ū L ū LG Inertial force
Reynolds Number NRe = = =
Viscous force
ū 2 L Inertial force
Weber Number NWe =
Surface-tension force
Heat Transfer
j-Factor for Heat Transfer j H = NStH ( NPr ) 2/3 jM
hL Convective heat transfer
Nusselt Number NNu = NSh
k Conductive heat transfer
L ūC P Bulk transfer of heat
Peclet Number for Heat Transfer NPeH = NRe NPr = NPeM
k Conductive heat transfer
CP v Momentum diffusivity
Prandtl Number NPr = = NSc
k Thermal diffusivity
NNu h Heat transfer
Stanton Number for Heat Transfer NStH = = NStM
NRe NPr CP G Thermal capacity
Mass Transfer
j-Factor for Mass Transfer j M = NStM ( NSc ) 2/3 jH
NSc k Thermal diffusivity
Lewis Number NLe = = =
NPr C P DAB DAB Mass diffusivity
L ū Bulk transfer of mass
Peclet Number for Mass Transfer NPeM = NRe NSc = NPeH
DAB Molecular diffusion
Momentum diffusivity
Schmidt Number NSc = = NPr
DAB DAB Mass diffusivity
kc L Convective mass transfer
Sherwood Number NSh = NNu
DAB Molecular diffusion
NSh kc Mass transfer
Stanton Number for Mass Transfer NStM = = NStH
NRe NSc ū Mass capacity
L = characteristic length G = mass velocity = u
Subscripts: M = mass transfer H = heat transfer
mean fluid velocity in the direction of flow and on position By analogy, the same mixing length is valid for turbulent-
in the fluid with respect to the solid boundaries. flow heat transfer and mass transfer. To use this analogy,
In 1925, in an attempt to quantify turbulent transport, (3-151) to (3-153) are rewritten in diffusivity form:
Prandtl [52] developed an expression for t in terms of an zx du x
eddy mixing length, l, which is a function of position. The = −( + M ) (3-154)
dz
eddy mixing length is a measure of the average distance that
qz dT
an eddy travels before it loses its identity and mingles with = −( + H ) (3-155)
other eddies. The mixing length is analogous to the mean CP dz
free path of gas molecules, which is the average distance a dcA
NAz = −( DAB + D ) (3-156)
molecule travels before it collides with another molecule. dz
where M , H , are D are momentum, heat, and mass eddy NPr = NSc = 1. Thus, the Reynolds analogy has limited
diffusivities, respectively; v is the momentum diffusivity practical value and is rarely applied in practice. Reynolds
(kinematic viscosity), / ; and is the thermal diffusivity, postulated the existence of the analogy in 1874 [53] and
k/C P . As a first approximation, the three eddy diffusivities derived it in 1883 [50].
may be assumed equal. This assumption is reasonably valid
for H and D , but experimental data indicate that Chilton–Colburn Analogy
M / H = M / D is sometimes less than 1.0 and as low as
A widely used extension of the Reynolds analogy to Prandtl
0.5 for turbulence in a free jet.
and Schmidt numbers other than 1 was presented by Colburn
[54] for heat transfer and by Chilton and Colburn [55] for
Reynolds Analogy
mass transfer. They showed that the Reynolds analogy for
If (3-154) to (3-156) are applied at a solid boundary, they can turbulent flow could be corrected for differences in velocity,
be used to determine transport fluxes based on transport temperature, and concentration distributions by incorporat-
coefficients, with driving forces from the wall, i, at z = 0, to ing NPr and NSc into (3-162) to define the following three
the bulk fluid, designated with an overbar, –: Chilton–Colburn j-factors, included in Table 3.13.

zx d( u x /ū x ) f f
jM ≡ = jH ≡
h
( NPr ) 2/3
= −( + M ) = ū x (3-157) 2 GC P
ū x dz z=0 2 (3-165)
kc
d(C P T ) = jD ≡ ( NSc ) 2/3
qz = −( + H ) = h(Ti − T̄ ) (3-158) G
dz z=0
Equation (3-165) is the Chilton–Colburn analogy or the
dcA
NAz = −( DAB + D ) = kc (cAi − c̄A ) (3-159) Colburn analogy for estimating average transport coeffi-
dz z=0
cients for turbulent flow. When NPr = NSc = 1, (3-165)
We define dimensionless velocity, temperature, and solute reduces to (3-162).
concentration by In general, j-factors are uniquely determined by the geo-
metric configuration and the Reynolds number. Based on the
ux Ti − T cA − cA
= = = i (3-160) analysis, over many years, of experimental data on momen-
ū x Ti − T̄ cAi − c̄A tum, heat, and mass transfer, the following representative
If (3-160) is substituted into (3-157) to (3-159), correlations have been developed for turbulent transport to
or from smooth surfaces. Other correlations are presented in
∂ f ū x h
= = other chapters. In general, these correlations are reasonably
∂z z=0 2( + M ) C P ( + H ) accurate for NPr and NSc in the range of 0.5 to 10, but should
(3-161)
kc be used with caution outside this range.
=
( DAB + D ) 1. Flow through a straight, circular tube of inside
This equation defines the analogies among momentum, heat, diameter D:
and mass transfer. Assuming that the three eddy diffusivities j M = j H = j D = 0.023( NRe ) −0.2 (3-166)
are equal and that the molecular diffusivities are either for 10,000 < NRe = DG/ < 1,000,000
everywhere negligible or equal,
2. Average transport coefficients for flow across a flat
f h kc
= = (3-162) plate of length L:
2 C P ū x ū x
j M = j H = j D = 0.037( NRe ) −0.2 (3-167)
Equation (3-162) defines the Stanton number for heat
for 5 × 105 < NRe = Lu o / < 5 × 108
transfer,
h h 3. Average transport coefficients for flow normal to a
NStH = = (3-163) long, circular cylinder of diameter D, where the drag
C P ū x GC P
coefficient includes both form drag and skin friction,
where G = mass velocity = ū x , and the Stanton number but only the skin friction contribution applies to the
for mass transfer, analogy:
kc kc
NStM = = (3-164) ( j M ) skin friction = j H = j D = 0.193( NRe ) −0.382
ū x G
for 4,000 < NRe < 40,000 (3-168)
both of which are included in Table 3.13. ( j M ) skin friction = j H = j D = 0.0266( NRe ) −0.195
Equation (3-162) is referred to as the Reynolds analogy.
It can be used to estimate values of heat and mass transfer for 40,000 < NRe < 250,000 (3-169)
coefficients from experimental measurements of the DG
with NRe =
Fanning friction factor for turbulent flow, but only when
1.000
Pa
cke
db
ed
0.100
j -factor
Sp
her
e
0.010
Cy
lin
der
Tube Flat p
flow late
0.001
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000 Figure 3.17 Chilton–Colburn
Reynolds number j-factor correlations.
4. Average transport coefficients for flow past a single Other improvements were made by van Driest [64], who
sphere of diameter D: used a modified form of the Prandtl mixing length, Reichardt
( j M ) skin friction = j H = j D = 0.37( NRe ) −0.4 [65], who eliminated the zone concept by allowing the eddy
(3-170) diffusivities to decrease continuously from a maximum to
DG
for 20 < NRe = < 100,000 zero at the wall, and Friend and Metzner [57], who modified
the approach of Reichardt to obtain improved accuracy at
5. Average transport coefficients for flow through beds very high Prandtl and Schmidt numbers (to 3,000). Their
packed with spherical particles of uniform size DP: results for turbulent flow through a straight, circular tube are
f /2
j H = j D = 1.17( NRe ) −0.415 NStH = √ (3-172)
DP G (3-171) 1.20 + 11.8 f /2( NPr − 1)( NPr ) −1/3
for 10 < NRe = < 2,500 f /2
NStM = √ (3-173)
1.20 + 11.8 f /2( NSc − 1)( NSc ) −1/3
The above correlations are plotted in Figure 3.17, where the
curves do not coincide because of the differing definitions of Over a wide range of Reynolds number (10,000–
the Reynolds number. However, the curves are not widely 10,000,000), the Fanning friction factor is estimated from
separated. When using the correlations in the presence of the explicit empirical correlation of Drew, Koo, and
appreciable temperature and/or composition differences, McAdams [66],
Chilton and Colburn recommend that NPr and NSc be evalu-
f = 0.00140 + 0.125( NRe ) −0.32 (3-174)
ated at the average conditions from the surface to the bulk
stream. which is in excellent agreement with the experimental data
of Nikuradse [67] and is preferred over (3-165) with (3-166),
which is valid only to NRe = 1,000,000. For two- and three-
Other Analogies
dimensional turbulent-flow problems, some success has
New turbulence theories have led to improvements and ex- been achieved with the (kinetic energy of turbulence)–
tensions of the Reynolds analogy, resulting in expressions (rate of dissipation) model of Launder and Spalding [68],
for the Fanning friction factor and the Stanton numbers for which is widely used in computational fluid dynamics
heat and mass transfer that are less empirical than the (CFD) computer programs.
Chilton–Colburn analogy. The first major improvement was
by Prandtl [56] in 1910, who divided the flow into two re- Theoretical Analogy of Churchill and Zajic
gions: (1) a thin laminar-flow sublayer of thickness next to An alternative to (3-151) to (3-153) or the equivalent diffu-
the wall boundary, where only molecular transport occurs; sivity forms of (3-154) to (3-156) for the development of
and (2) a turbulent region dominated by eddy transport, with transport equations for turbulent flow is to start with the
M = H = D . time-averaged equations of Newton, Fourier, and Fick. For
Further important theoretical improvements to the example, let us derive a form of Newton’s law of viscosity
Reynolds analogy were made by von Karman, Martinelli, for molecular and turbulent transport of momentum in paral-
and Deissler, as discussed in detail by Knudsen and Katz lel. In a turbulent-flow field in the axial x-direction, instanta-
[47]. The first two investigators inserted a buffer zone be- neous velocity components, u x and u z , are
tween the laminar sublayer and turbulent core. Deissler
gradually reduced the eddy diffusivities as the wall was u x = ū x + u x
approached. u z = u z
where the “overbarred” component is the time-averaged Equation (3-180) is a highly accurate quantitative represen-
(mean) local velocity and the primed component is the local tation of turbulent flow because it is based on experimental
fluctuating component that denotes instantaneous deviation data and numerical simulations described by Churchill and
from the local mean value. The mean velocity in the perpen- Zajic [70] and in considerable detail by Churchill [71]. From
dicular z-direction is zero. The mean local velocity in the x- (3-142) and (3-143), the shear stress at the wall, w , is
direction over a long period of time is given by related to the Fanning friction factor by

1 1
ū x = u x d = ( ū x + u x ) d (3-175) f =
2 w
(3-181)
0 0 ū 2x
The time-averaged fluctuating components u x and u z equal
zero. where ū x is the flow-average velocity in the axial direction.
The local instantaneous rate of momentum transfer by Combining (3-179) with (3-181) and performing the required
turbulence in the z-direction of x-direction turbulent mo- integrations, both numerically and analytically, lead to the
mentum per unit area at constant density is following implicit equation for the Fanning friction factor as
a function of the Reynolds number, NRe = 2a ū x /:
u z ( ū x + u x ) (3-176) 1/2  1/2 2
2 2
The time-average of this turbulent momentum transfer is 1/2  
equal to the turbulent component of the shear stress, zxt , 2 f  f 
= 3.2 − 227 + 2500  
f NRe  NRe 

zxt = u ( ū x + u x ) d 2 2
0 z   (3-182)
(3-177) NRe
 
= u ( ū x ) d + u z (u x ) d 1  
0 z 0 + ln  21/2 
0.436  2 
Because the time-average of the first term is zero, (3-177) f
reduces to
zxt = (u z u x ) (3-178) This equation is in excellent agreement with experimental
data over a Reynolds number range of 4,000–3,000,000 and
which is referred to as a Reynolds stress. Combining (3-178) can probably be used to a Reynolds number of 100,000,000.
with the molecular component of momentum transfer gives Table 3.14 presents a comparison of the Churchill–Zajic
the turbulent-flow form of Newton’s law of viscosity, equation, (3-182), with (3-174) of Drew et al. and (3-166)
du x of Chilton and Colburn. Equation (3-174) gives satis-
zx = − + (u z u x ) (3-179) factory agreement for Reynolds numbers from 10,000 to
dz
10,000,000, while (3-166) is useful only from 100,000 to
If (3-179) is compared to (3-151), it is seen that an alterna- 1,000,000.
tive approach to turbulence is to develop a correlating equa- Churchill and Zajic [70] show that if the equation for the
tion for the Reynolds stress, (u z u x ), first introduced by conservation of energy is time averaged, a turbulent-flow
Churchill and Chan [73], rather than an expression for a tur- form of Fourier’s law of conduction can be obtained with the
bulent viscosity t . This stress, which is a complex function fluctuation term (u z T ). Similar time averaging leads to a
of position and rate of flow, has been correlated quite accu- turbulent-flow form of Fick’s law of diffusion with (u z cA ).
rately for fully developed turbulent flow in a straight, circu- To extend (3-180) and (3-182) to obtain an expression for
lar tube by Heng, Chan, and Churchill [69]. In generalized the Nusselt number for turbulent-flow convective heat trans-
form, with a the radius of the tube and y = (a − z) the dis- fer in a straight, circular tube, Churchill and Zajic employ an
tance from the inside wall to the center of the tube, their analogy that is free of empircism, but not exact. The result
equation is

+ 3 −8/7
y −1
(u z u x ) =  0.7
++
+ exp
10 0.436y + Table 3.14 Comparison of Fanning Friction Factors for Fully
−7/8 Developed Turbulent Flow in a Smooth, Straight Circular Tube
1 6.95y + −8/7
− 1+
0.436a + a+ f, Drew et al. f, Chilton–Colburn f, Churchill–Zajic
NRe (3-174) (3-166) (3-182)
(3-180)
10,000 0.007960 0.007291 0.008087
where
100,000 0.004540 0.004600 0.004559
(u z u x ) ++ = − u z u x / 1,000,000 0.002903 0.002902 0.002998
a + = a( w ) 1/2 / 10,000,000 0.002119 0.001831 0.002119
100,000,000 0.001744 0.001155 0.001573
y + = y( w ) 1/2 /
for Prandtl numbers greater than 1 is Chilton–Colburn correlation, which is widely used, is within
10% of the more theoretically based Churchill–Zajic equa-
1
NNu = 2/3 (3-183) tion for Reynolds numbers up to 1,000,000. However, beyond
NPrt 1 NPrt 1 that, serious deviations occur (25% at NRe = 10,000,000 and
+ 1−
NPr NNu1 NPr NNu∞ almost 50% at NRe = 100,000,000). Deviations of the
Friend–Metzner correlation from the Churchill–Zajic equa-
where, from Yu, Ozoe, and Churchill [72], tion vary from about 15% to 30% over the entire range of
Reynolds number in Table 3.15. At all Reynolds numbers,
0.015
NPrt = turbulent Prandtl number = 0.85 + (3-184) the Churchill–Zajic equation predicts higher Nusselt num-
NPr bers and, therefore, higher heat-transfer coefficients.
which replaces (u z T ), as introduced by Churchill [74], At a Prandtl number of 1,000, which is typical of high-
viscosity liquids, the Friend–Metzner correlation is in fairly
NNu1 = Nusselt number for ( NPr = NPrt )
close agreement with the Churchill–Zajic equation, predict-
f ing values from 6 to 13% higher. The Chilton–Colburn cor-
NRe
2 relation is seriously in error over the entire range of
= −5/4 (3-185)
2 Reynolds number, predicting values ranging from 74 to 27%
1 + 145 of those from the Churchill–Zajic equation as the Reynolds
f
number increases. It is clear that the Chilton–Colburn corre-
NNu∞ = Nusselt number for ( NPr = ∞) lation should not be used at high Prandtl numbers for heat
1/2 transfer or (by analogy) at high Schmidt numbers for mass
NPr 1/3 f
= 0.07343 NRe (3-186) transfer.
NPrt 2
The Churchill–Zajic equation for predicting the Nusselt
The accuracy of (3-183) is due to (3-185) and (3-186), which number provides an effective power dependence on the
are known from theoretical considerations. Although (3-184) Reynolds number as the Reynolds number increases. This is
is somewhat uncertain, its effect is negligible. in contrast to the typically cited constant exponent of 0.8, as
A comparison of the Churchill et al. correlation of in the Chilton–Colburn correlation. For the Churchill–Zajic
(3-183) with the Nusselt forms of (3-172) of Friend and equation, at a Prandtl number of 1, the exponent increases
Metzner and (3-166) of Chilton and Colburn, where from with Reynolds number from 0.79 to 0.88; at a Prandtl num-
Table 3.13, NNu = NSt NRe NPr , is given in Table 3.15 for a ber of 1,000, the exponent increases from 0.87 to 0.93.
wide range of Reynolds number and Prandtl numbers of Extension of the Churchill–Zajic equation to low Prandtl
1 and 1,000. numbers, typical of molten metals, and to other geometries,
In Table 3.15, at a Prandtl number of 1, which is typical of such as parallel plates, is discussed by Churchill [71], who
low-viscosity liquids and close to that of most gases, the also considers the important effect of boundary conditions
Table 3.15 Comparison of Nusselt Numbers for Fully Developed Turbulent Flow in a
Smooth, Straight Circular Tube
Prandtl number, NPr = 1
NNu , Friend–Metzner NNu , Chilton–Colburn NNu , Churchill–Zajic

NRe (3-172) (3-166) (3-183)
10,000 33.2 36.5 37.8
100,000 189 230 232
1,000,000 1210 1450 1580
10,000,000 8830 9160 11400
100,000,000 72700 57800 86000
Prandtl number, NPr = 1000
NNu , Friend–Metzner NNu , Chilton–Colburn NNu , Churchill–Zajic

NRe (3-172) (3-166) (3-183)
10,000 527 365 491
100,000 3960 2300 3680
1,000,000 31500 14500 29800
10,000,000 267800 91600 249000
100,000,000 2420000 578000 2140000
3.6 Models for Mass Transfer at a Fluid–Fluid Interface 103
(e.g., constant wall temperature and uniform heat flux) at Churchill–Zajic equation:
low-to-moderate Prandtl numbers. Using mass-transfer analogs,
For calculation of convective mass-transfer coefficients,
kc , for turbulent flow of gases and liquids in straight, (3-184) gives NSct = 0.850
smooth, circular tubes, it is recommended that the (3-185) gives NSh1 = 94
Churchill–Zajic equation be employed by applying the (3-186) gives NSh∞ = 1686
analogy between heat and mass transfer. Thus, as illustrated
(3-183) gives NSh = 1680
in the following example, in (3-183) to (3-186), from
Table 3.13, the Sherwood number, NSh , is substituted for the From Table 3.13,
Nusselt number, NNu ; and the Schmidt number, NSc , is sub-
stituted for the Prandtl number, NPr . NSh 1680
NStM = = = 0.0000324,
NRe NSc (35800)(1450)
which is an acceptable 92% of the experimental value.

EXAMPLE 3.16
Linton and Sherwood [49] conducted experiments on the dissolv-
ing of cast tubes of cinnamic acid (A) into water (B) flowing 3.6 MODELS FOR MASS TRANSFER AT
through the tubes in turbulent flow. In one run, with a 5.23-cm-i.d. A FLUID–FLUID INTERFACE
tube, NRe = 35,800, and NSc = 1,450, they measured a Stanton
number for mass transfer, NStM , of 0.0000351. Compare this exper- In the three previous sections, diffusion and mass transfer
imental value with predictions by the Reynolds, Chilton–Colburn, within solids and fluids were considered, where the interface
and Friend–Metzner analogies, and by the more theoretically-based was a smooth solid surface. Of greater interest in separation
Churchill–Zajic equation. processes is mass transfer across an interface between a gas
and a liquid or between two liquid phases. Such interfaces
SOLUTION exist in absorption, distillation, extraction, and stripping.
At fluid–fluid interfaces, turbulence may persist to the inter-
From either (3-174) or (3-182), the Fanning friction factor is
face. The following theoretical models have been devel-
0.00576.
oped to describe mass transfer between a fluid and such an
Reynolds analogy: interface.
From (3-162), NStM = 2f = 0.00576/2 = 0.00288, which, as ex-
pected, is in poor agreement with the experimental value because Film Theory
the effect of Schmidt number is ignored.
A simple theoretical model for turbulent mass transfer to
Chilton–Colburn analogy: or from a fluid-phase boundary was suggested in 1904 by
From (3-165), Nernst [58], who postulated that the entire resistance to mass
transfer in a given turbulent phase is in a thin, stagnant re-
f 0.00576 gion of that phase at the interface, called a film. This film is
NStM = ( NSc ) =
2/3
(1450) 2/3 = 0.0000225 ,
2 2 similar to the laminar sublayer that forms when a fluid flows
which is 64% of the experimental value. in the turbulent regime parallel to a flat plate. This is shown
schematically in Figure 3.18a for the case of a gas–liquid in-
Friend–Metzner analogy: terface, where the gas is pure component A, which diffuses
From (3-173), NStM = 0.0000350, which is almost identical to the into nonvolatile liquid B. Thus, a process of absorption of A
experimental value. into liquid B takes place, without desorption of B into
pA pA
Bulk liquid
Liquid
film
cAi
cAi
Gas Gas Well-mixed
bulk region
cAb
at cAb
Interfacial
cAb region
Figure 3.18 Theories for mass

transfer from a fluid–fluid inter-
face into a liquid: (a) film theory;
z=O z = δL
(b) (b) penetration and surface-
(a) renewal theories.
gaseous A. Because the gas is pure A at total pressure the bulk liquid. If the diffusivity of SO2 in water is
P = pA , there is no resistance to mass transfer in the gas 1.7 × 10−5 cm2 /s, determine the mass-transfer coefficient, kc , and
phase. At the gas–liquid interface, phase equilibrium is as- the film thickness, neglecting the bulk-flow effect.
sumed so the concentration of A, cAi , is related to the partial
pressure of A, pA , by some form of Henry’s law, for exam- SOLUTION
ple, cAi = HA pA . In the thin, stagnant liquid film of thick-
ness , molecular diffusion only occurs with a driving force 0.027(1,000) mol
NSO2 = 2
= 7.5 × 10−7 2
of cAi − cAb . Since the film is assumed to be very thin, all of (3,600)(100) cm -s
the diffusing A passes through the film and into the bulk liq- For dilute conditions, the concentration of water is
uid. If, in addition, bulk flow of A is neglected, the concen- 1
tration gradient is linear as in Figure 3.18a. Accordingly, c= = 5.55 × 10−2 mol/cm3
18.02
Fick’s first law, (3-3a), for the diffusion flux integrates to
From (3-188),
DAB cDAB
JA = (cAi − cAb ) = (xAi − xAb ) (3-187) DAB NA
kc = =
c(xAi − xAb )
If the liquid phase is dilute in A, the bulk-flow effect can be 7.5 × 10−7
neglected and (3-187) applies to the total flux: = = 6.14 × 10−3 cm/s
5.55 × 10−2 (0.0025 − 0.0003)
DAB cDAB
NA = (cAi − cAb ) = (xAi − xAb ) (3-188) Therefore,

DAB 1.7 × 10−5
= = = 0.0028 cm
If the bulk-flow effect is not negligible, then, from (3-31), kc 6.14 × 10−3

cDAB 1 − xAb cDAB which is very small and typical of turbulent-flow mass-transfer
NA = ln = (xAi − xAb ) processes.
1 − xAi (1 − xA ) LM
(3-189)
where Penetration Theory
xAi − xAb A more realistic physical model of mass transfer from a
(1 − xA ) LM = = (xB ) LM fluid–fluid interface into a bulk liquid stream is provided by
ln[(1 − xAb )/(1 − xAi )]
the penetration theory of Higbie [59], shown schematically
(3-190) in Figure 3.18b. The stagnant-film concept is replaced by
Boussinesq eddies that, during a cycle, (1) move from the
In practice, the ratios DAB / in (3-188) and DAB /
bulk to the interface; (2) stay at the interface for a short,
(1 − xA ) LM in (3-189) are replaced by mass transfer coeffi-
fixed period of time during which they remain static so that
cients kc and kc , respectively, because the film thickness, ,
molecular diffusion takes place in a direction normal to the
which depends on the flow conditions, is not known and the
interface; and (3) leave the interface to mix with the bulk
subscript, c, refers to a concentration driving force.
stream. When an eddy moves to the interface, it replaces an-
The film theory, which is easy to understand and apply, is
other static eddy. Thus, the eddies are alternately static and
often criticized because it appears to predict that the rate of
moving. Turbulence extends to the interface.
mass transfer is directly proportional to the molecular diffu-
In the penetration theory, unsteady-state diffusion takes
sivity. This dependency is at odds with experimental data,
place at the interface during the time the eddy is static. This
which indicate a dependency of D n , where n ranges from
process is governed by Fick’s second law, (3-68), with
about 0.5 to 0.75. However, if DAB / is replaced with kc ,
boundary conditions
which is then estimated from the Chilton–Colburn analogy,
Eq. (3-165), we obtain kc proportional to DAB 2/3
, which is in cA = cAb at t = 0 for 0 ≤ z ≤ ∞;
better agreement with experimental data. In effect, de- cA = cAi at z = 0 for t > 0; and
pends on DAB (or NSc ). Regardless of whether the criticism cA = cAb at z = ∞ for t > 0
of the film theory is valid, the theory has been and continues
to be widely used in the design of mass-transfer separation These are the same boundary conditions as in unsteady-state
equipment. diffusion in a semi-infinite medium. Thus, the solution can
be written by a rearrangement of (3-75):

cAi − cA z
EXAMPLE 3.17 = erf √ (3-191)
cAi − cAb 2 DAB tc
Sulfur dioxide is absorbed from air into water in a packed absorp-
tion tower. At a certain location in the tower, the mass-transfer flux where tc = “contact time” of the static eddy at the interface
is 0.0270 kmol SO2 /m2 -h and the liquid-phase mole fractions are during one cycle. The corresponding average mass-transfer
0.0025 and 0.0003, respectively, at the two-phase interface and in flux of A in the absence of bulk flow is given by the
3.6 Models for Mass Transfer at a Fluid–Fluid Interface 105
following form of (3-79): Surface-Renewal Theory

DAB The penetration theory is not satisfying because the as-
NA = 2 (cAi − cAb ) (3-192) sumption of a constant contact time for all eddies that tem-
tc
porarily reside at the surface is not reasonable, especially
or for stirred tanks, contactors with random packings, and
NA = kc (cAi − cAb ) (3-193) bubble and spray columns where the bubbles and droplets
Thus, the penetration theory gives cover a wide range of sizes. In 1951, Danckwerts [60] sug-
gested an improvement to the penetration theory that
DAB involves the replacement of the constant eddy contact
kc = 2 (3-194)
tc time with the assumption of a residence-time distribution,
wherein the probability of an eddy at the surface being
which predicts that kc is proportional to the square root of the replaced by a fresh eddy is independent of the age of the
molecular diffusivity, which is at the lower limit of experi- surface eddy.
mental data. Following the Levenspiel [61] treatment of residence-
The penetration theory is most useful when mass transfer time distribution, let F(t) be the fraction of eddies with a
involves bubbles or droplets, or flow over random packing. contact time of less than t. For t = 0, F{t} = 0, and F{t}
For bubbles, the contact time, tc, of the liquid surrounding approaches 1 as t goes to infinity. A plot of F{t} versus t,
the bubble is taken as the ratio of bubble diameter to bubble- as shown in Figure 3.19, is referred to as a residence-time or
rise velocity. For example, an air bubble of 0.4-cm diameter age distribution. If F{t} is differentiated with respect to t, we
rises through water at a velocity of about 20 cm/s. Thus, the obtain another function:
estimated contact time, tc, is 0.4/20 = 0.02 s. For a liquid
spray, where no circulation of liquid occurs inside the {t} = d F{t}/dt
droplets, the contact time is the total time for the droplets to
fall through the gas. For a packed tower, where the liquid where {t}dt = the probability that a given surface eddy
flows as a film over particles of random packing, mixing can will have a residence time t. The sum of probabilities is
be assumed to occur each time the liquid film passes from ∞
one piece of packing to another. Resulting contact times are {t} dt = 1 (3-195)
0
of the order of about 1 s. In the absence of any method of es-
timating the contact time, the liquid-phase mass-transfer co- Typical plots of F{t} and {t} are shown in Figure 3.19,
efficient is sometimes correlated by an empirical expression where it is seen that {t} is similar to a normal probability
consistent with the 0.5 exponent on DAB, given by (3-194) curve.
with the contact time replaced by a function of geometry and For steady-state flow in and out of a well-mixed vessel,
the liquid velocity, density, and viscosity. Levenspiel shows that
F{t} = 1 − e−t/t¯ (3-196)
EXAMPLE 3.18
where t¯ is the average residence time. This function forms
For the conditions of Example 3.17, estimate the contact time for the basis, in reaction engineering, of the ideal model of a
Higbie’s penetration theory.
continuous, stirred-tank reactor (CSTR). Danckwerts se-
lected the same model for his surface-renewal theory, using
SOLUTION the corresponding {t} function:
From Example 3.17, kc = 6.14 × 10−3 cm/s and DAB = 1.7 × {t} = se−st (3-197)
10−5 cm2 /s. From a rearrangement of (3-194),
where s = 1/t¯ = fractional rate of surface renewal. As
4DAB 4(1.7 × 10−5 )
tc = = = 0.57 s shown in Example 3.19 below, plots of (3-196) and (3-197)
kc2 3.14(6.14 × 10−3 ) 2
are much different from those in Figure 3.19.
Fraction of
F{t} exit stream
φ {t} older than t1
Total Figure 3.19 Residence-time distrib-
0 area = 1 ution plots: (a) typical F curve;
0 (b) typical age distribution.
0 t 0 t1
[Adapted from O. Levenspiel, Chemical
t t
Reaction Engineering, 2nd ed., John Wiley
(a) (b) and Sons, New York (1972).]
The instantaneous mass-transfer rate for an eddy with an From (3-196), the residence-time distribution is given by
age t is given by (3-192) for the penetration theory in flux
F{t} = 1 − e−t/0.45 , (2)
form as
where t is in seconds. Equations (1) and (2) are plotted in
DAB Figure 3.20. These curves are much different from the curves
NAt = (cAi − cAb ) (3-198)
t of Figure 3.19.
The integrated average rate is

∞ Film-Penetration Theory
( NA ) avg = {t}NAt dt (3-199) Toor and Marchello [62], in 1958, combined features of the
0
film, penetration, and surface-renewal theories to develop a
Combining (3-197), (3-198), and (3-199), and integrating: film-penetration model, which predicts a dependency of the
mass-transfer coefficient kc, on the diffusivity, that varies
( NA ) avg = DAB s(cAi − cAb ) (3-200) √
from DAB to DAB. Their theory assumes that the entire re-
Thus, sistance to mass transfer resides in a film of fixed thickness
. Eddies move to and from the bulk fluid and this film. Age
kc = DAB s (3-201) distributions for time spent in the film are of the Higbie or
Danckwerts type.
The more reasonable surface-renewal theory predicts the Fick’s second law, (3-68), still applies, but the boundary
same dependency of the mass-transfer coefficient on molec- conditions are now
ular diffusivity as the penetration theory. Unfortunately, s, cA = cAb at t = 0 for 0 ≤ z ≤ ∞,
the fractional rate of surface renewal, is as elusive a parame-
cA = cAi at z = 0 for t > 0; and
ter as the constant contact time, tc.
cA = cAb at z = for t > 0
Infinite-series solutions are obtained by the method of
EXAMPLE 3.19
Laplace transforms. The rate of mass transfer is then ob-
For the conditions of Example 3.17, estimate the fractional rate of tained in the usual manner by applying Fick’s first law
surface renewal, s, for Danckwert’s theory and determine the resi- (3-117) at the fluid–fluid interface. For small t, the solution,
dence time and probability distributions. given as
∞
DAB 1/2 n 2 2
SOLUTION NAt = (cAi − cAb ) 1+2 exp −
t n=1
DAB t
From Example 3.17,
(3-202)
−3 −5
kc = 6.14 × 10 cm/s and DAB = 1.7 × 10 2
cm /s
converges rapidly. For large t,
From (3-201),
DAB
NAt = (cAi − cAb )
kc2 (6.14 × 10−3 ) 2
s= = = 2.22 s−1 ∞ (3-203)
DAB 1.7 × 10−5 DAB t
× 1+2 exp −n 2 2 2
Thus, the average residence time of an eddy at the surface is n=1

1/2.22 = 0.45 s.
Equation (3-199) with {t} from (3-197) can then be used
From (3-197),
to obtain average rates of mass transfer. Again, we can
{t} = 2.22e−2.22t (1) write two equivalent series solutions, which converge
1
1 = 2.22 s–1
t
Area = t φ {t} 1 e–t / t
F{t} t
1 – e–t / t
Area = 1
0 t 0 t
0 t = 0.45 s 0 t = 0.45 s Figure 3.20 Age distribution curves for
(a) (b) Example 3.19: (a) F curve; (b) {t} curve.
3.7 Two-Film Theory and Overall Mass-Transfer Coefficients 107
at different rates. Equations (3-202) and (3-203) become, the Marangoni effect, is discussed in some detail by Bird,
respectively, Stewart, and Lightfoot [28], who cite additional references.
The effect can occur at both vapor–liquid and liquid–liquid
NAavg = kc (cAi − cAb ) = (cAi − cAb )(s DAB ) 1/2
interfaces, with the latter receiving the most attention. By
∞
s adding surfactants, which tend to concentrate at the inter-
× 1+2 exp −2n (3-204)
n=1
DAB face, the Marangoni effect may be reduced because of stabi-
lization of the interface, even to the extent that an interfacial
DAB mass-transfer resistance may result, causing the overall
NAavg = kc (cAi − cAb ) = (cAi − cAb )
mass-transfer coefficient to be reduced. In this book, unless
  otherwise indicated, the Marangoni effect will be ignored
∞ (3-205)
 1  and phase equilibrium will always be assumed at the phase
× 1 + 2
DAB  interface.
n=1 1 + n 2 2
s2
In the limit, for a high rate of surface renewal, s2 /DAB ,
Gas–Liquid Case
(3-204) reduces to the surface-renewal theory, (3-200). For Consider the steady-state mass transfer of A from a gas
low rates of renewal, (3-205) reduces to the film theory, phase, across an interface, into a liquid phase. It could be
n
(3-188). At conditions in between, kc is proportional to DAB , postulated, as shown in Figure 3.21a, that a thin gas film ex-
where n is in the range of 0.5–1.0. The application of the ists on one side of the interface and a thin liquid film exists
film-penetration theory is difficult because of lack of data for on the other side, with the controlling factors being molecu-
and s, but the predicted effect of molecular diffusivity lar diffusion through each film. However, this postulation is
brackets experimental data. not necessary, because instead of writing
( DAB ) G ( DAB ) L
NA = (cAb − cAi ) G = (cAi − cAb ) L
3.7 TWO-FILM THEORY AND OVERALL G L
MASS-TRANSFER COEFFICIENTS (3-206)
Separation processes that involve contacting two fluid we can express the rate of mass transfer in terms of mass-
phases require consideration of mass-transfer resistances in transfer coefficients determined from any suitable theory,
both phases. In 1923, Whitman [63] suggested an extension with the concentration gradients visualized more realisti-
of the film theory to two fluid films in series. Each film pre- cally as in Figure 3.21b. In addition, we can use any number
sents a resistance to mass transfer, but concentrations in the of different mass-transfer coefficients, depending on the se-
two fluids at the interface are assumed to be in phase equi- lection of the driving force for mass transfer. For the gas
librium. That is, there is no additional interfacial resistance phase, under dilute or equimolar counter diffusion (EMD)
to mass transfer. This concept has found extensive applica- conditions, we write the mass-transfer rate in terms of partial
tion in modeling of steady-state, gas–liquid, and liquid– pressures:
liquid separation processes.
The assumption of phase equilibrium at the phase inter- NA = k p ( pAb − pAi ) (3-207)
face, while widely used, may not be valid when gradients of where kp is a gas-phase mass-transfer coefficient based on a
interfacial tension are established during mass transfer be- partial-pressure driving force.
tween two fluids. These gradients give rise to interfacial tur- For the liquid phase, we use molar concentrations:
bulence resulting, most often, in considerably increased
mass-transfer coefficients. This phenomenon, referred to as NA = kc (cAi − cAb ) (3-208)
Gas Gas Liquid Liquid pAb

phase film film phase
pAb Gas Liquid
phase phase
pAi pAi
cAi cAi
cAb cAb
Transport Transport
Figure 3.21 Concentration
gradients for two-resistance
theory: (a) film theory; (b) more
(a) (b) realistic gradients.
At the phase interface, cAi and pAi are assumed to be in Alternatively, (3-207) to (3-209) can be combined to
phase equilibrium. Applying a version of Henry’s law differ- define an overall mass-transfer coefficient, KG, based on the
ent from that in Table 2.3,1 gas phase. The result is
cAi = HA pAi (3-209) pAb − cAb /HA
NA = (3-216)
(1/k p ) + (1/HA kc )
Equations (3-207) to (3-209) are a commonly used combina-
tion for vapor–liquid mass transfer. Computations of mass- In this case, it is customary to define: (1) a fictitious gas-
transfer rates are generally made from a knowledge of bulk phase partial pressure pA∗ = cAb /HA , which is the partial
concentrations, which in this case are pAb and cAb . To obtain pressure that would be in equilibrium with the bulk liquid;
an expression for NA in terms of an overall driving force for and (2) an overall mass-transfer coefficient for the gas phase,
mass transfer, (3-207) to (3-209) are combined in the fol- KG, based on a partial-pressure driving force. Thus, (3-216)
lowing manner to eliminate the interfacial concentrations, can be rewritten as
cAi and pAi . Solve (3-207) for pAi : ( pAb − pA∗ )
NA = K G ( pAb − pA∗ ) = (3-217)
NA (1/k p ) + (1/HA kc )
pAi = pAb − (3-210)
kp where
Solve (3-208) for cAi : 1 1 1
= + (3-218)
NA KG kp HA kc
cAi = cAb + (3-211)
kc In this, the resistances are 1/kp and 1/(HAkc). When
1/kp >> 1/HAkc,
Combine (3-211) with (3-209) to eliminate cAi and combine
the result with (3-210) to eliminate pAi to give NA = k p ( pAb − pA∗ ) (3-219)
pAb HA − cAb Since the resistance in the liquid phase is then negligible, the
NA = (3-212)
( HA /k p ) + (1/kc ) liquid-phase driving force is cAi − cAb ≈ 0 and cAi ≈ cAb .
The choice between using (3-213) or (3-217) is arbitrary,
It is customary to define: (1) a fictitious liquid-phase but is usually made on the basis of which phase has the
concentration cA∗ = pAb HA , which is the concentration that largest mass-transfer resistance; if the liquid, use (3-213); if
would be in equilibrium with the partial pressure in the bulk the gas, use (3-217). Another common combination for
gas; and (2) an overall mass-transfer coefficient, KL. Thus, vapor–liquid mass transfer uses mole fraction-driving forces,
(3-212) is rewritten as which define another set of mass-transfer coefficients:
(cA∗ − cAb ) NA = k y ( yAb − yAi ) = k x (xAi − xAb )
NA = K L (cA∗ − cAb ) = (3-213) (3-220)
( HA /k p ) + (1/kc )
In this case, phase equilibrium at the interface can be
where expressed in terms of the K-value for vapor–liquid equilib-
1 HA 1 rium. Thus,
= + (3-214)
KL kp kc K A = yAi /xAi (3-221)
in which KL is the overall mass-transfer coefficient based on Combining (3-220) and (3-221) to eliminate yAi and xAi ,
the liquid phase. The quantities HA/kp and 1/kc are measures yAb − xAb
NA = (3-222)
of the mass-transfer resistances of the gas phase and the (1/K A k y ) + (1/k x )
liquid phase, respectively. When 1/kc >> HA/kp, (3-214)
This time we define fictitious concentration quantities and
becomes
overall mass-transfer coefficients for mole-fraction driving
NA = kc (cA∗ − cAb ) (3-215) forces. Thus, xA∗ = yAb /K A and yA∗ = K A xAb . If the two
Since resistance in the gas phase is then negligible, the gas- values of KA are equal, we obtain
phase driving force is pAb − pAi ≈ 0 and pAb ≈ pAi . xA∗ − xAb
NA = K x (xA∗ − xAb ) = (3-223)
(1/K A k y ) + (1/k x )
1
Many different forms of Henry’s law are found in the literature. They and
include
yAb − yA∗
pA = HA xA , pA =
cA
, and yA = HA xA NA = K y ( yAb − yA∗ ) = (3-224)
HA (1/k y ) + (K A /k x )
When a Henry’s-law constant, HA, is given without citing the equation that
defines it, the defining equation can be determined from the units of the
where Kx and Ky are overall mass-transfer coefficients based
constant. For example, if the constant has the units of atm or atm/mole on mole-fraction driving forces with
fraction, Henry’s law is given by pA = HAxA. If the units are mol/L-mmHg, 1 1 1
cA
Henry’s law is pA = . = + (3-225)
HA Kx KAk y kx
3.7 Two-Film Theory and Overall Mass-Transfer Coefficients 109
and For the liquid phase, using kc or kx,

1 1 KA k k
= + (3-226) k = = (3-229)
Ky ky kx (1 − xA ) LM (xB ) LM
When using correlations to estimate mass-transfer coeffi- For the gas phase, using kp, ky, or kc,
cients for use in the above equations, it is important to deter- k k
k = = (3-230)
mine which coefficient (kp, kc, ky, or kx) is correlated. This (1 − yA ) LM ( yB ) LM
can usually be done by checking the units or the form of the
The expressions for k are most readily used when the
Sherwood or Stanton numbers. Coefficients correlated by
mass-transfer rate is controlled mainly by one of the two
the Chilton–Colburn analogy are kc for either the liquid or
resistances. Experimental mass-transfer coefficient data re-
gas phase. The different coefficients are related by the fol-
ported in the literature are generally correlated in terms of k
lowing expressions, which are summarized in Table 3.16.
rather than k . Mass-transfer coefficients estimated from the
Liquid phase: Chilton–Colburn analogy [e.g., equations (3-166) to (3-171)]
are kc, not kc .
L
k x = kc c = kc (3-227)
M
Liquid–Liquid Case
Ideal-gas phase:
For mass transfer across two liquid phases, equilibrium is
again assumed at the interface. Denoting the two phases by
P G
k y = k p P = (kc ) g = (kc ) g c = (kc ) g (3-228) L(1) and L(2), (3-223) and (3-224) can be rewritten as
RT M

xA(2)* − xA(2)b
Typical units are NA = K x(2) xA(2)* − xA(2)b =
(1/K DA k x(1) ) + (1/k x(2) )
SI American Engineering (3-231)
kc m/s ft/h and
kp kmol/s-m2-kPa lbmol/h-ft2-atm
xA(1)b − xA(1)*
ky, kx kmol/s-m2 lbmol/h-ft2 NA = K x(1) xA(1)b − xA(1)* =
(1/k x(1) ) + (K DA /k x(2) )
When unimolecular diffusion (UMD) occurs under non-
(3-232)
dilute conditions, the effect of bulk flow must be included in
where
the above equations. For binary mixtures, one method for
xA(1)i
doing this is to define modified mass-transfer coefficients, K DA = (3-233)
designated with a prime, as follows. xA(2)i
Table 3.16 Relationships among Mass-Transfer Coefficients Case of Large Driving Forces for Mass Transfer
Equimolar Counterdiffusion (EMD): When large driving forces exist for mass transfer, phase
equilibria ratios such as HA, KA, and K DA may not be con-
Gases: NA = k y yA = kc cA = k p pA stant across the two phases. This occurs particularly when
P one or both phases are not dilute with respect to the diffusing
k y = kc = k p P if ideal gas
RT solute, A. In that case, expressions for the mass-transfer flux
Liquids: NA = k x xA = kc cA must be revised.
k x = kc c, where c = total molar For example, if mole-fraction driving forces are used, we
concentration (A + B)
write, from (3-220) and (3-224),
Unimolecular Diffusion (UMD): NA = k y ( yAb − yAi ) = K y ( yAb − yA∗ ) (3-234)
Gases: Same equations as for EMD with k replaced Thus,
by k =
k 1 yAb − yA∗
( yB ) LM = (3-235)
Ky k y ( yAb − yAi )
Liquids: Same equations as for EMD with k replaced by or

k =
k 1 ( yAb − yAi ) + ( yAi − yA∗ ) 1 1 yAi − yA∗
= = +
(xB ) LM Ky k y ( yAb − yAi ) ky k y yAb − yAi
When using concentration units for both phases, it is convenient (3-236)
to use: From (3-220),
k G (cG ) = kc (c) for the gas phase
kx ( yAb − yAi )
k L (c L ) = kc (c) for the liquid phase = (3-237)
ky (xAi − xAb )
pSO2 , atm cSO2 , lbmol/ft3

my yAb
e
op 0.0382 0.00193
Sl
yAi 0.0606 0.00290
0.1092 0.00483
x
m
e
* is a fictitious x in
xA
op A 0.1700 0.00676
equilibrium with yAb.
Sl
yA *
yA
* is a fictitious y in
yA A Experimental values of the mass transfer coefficients are as
equilibrium with xAb.
xAb xAi * follows.
xA
Liquid phase: kc = 0.18 m/h
kmol
Gas phase: k p = 0.040
xA
h-m2 -kPa
Figure 3.22 Curved equilibrium line. Using mole-fraction driving forces, compute the mass-transfer
flux by:
(a) Assuming an average Henry’s-law constant and a negligible
bulk-flow effect.
Combining (3-234) and (3-237),
(b) Utilizing the actual curved equilibrium line and assuming a

1 1 1 yAi − yA∗ negligible bulk-flow effect.
= + (3-238)
Ky ky k x xAi − xAb (c) Utilizing the actual curved equilibrium line and taking into
account the bulk-flow effect.
In a similar manner,
In addition,

1 1 1 xA∗ − xAi (d) Determine the relative magnitude of the two resistances and
= + (3-239) the values of the mole fractions at the interface from the results
Kx kx ky yAb − yAi
of part (c).
A typical curved equilibrium line is shown in Figure 3.22
with representative values of yAb , yAi , yA∗ , xA∗ , xAi , and xAb
indicated. Because the line is curved, the vapor–liquid equi- SOLUTION
librium ratio, K A = yA /xA , is not constant across the two The equilibrium data are converted to mole fractions by assuming
phases. As shown, the slope of the curve and thus, KA, de- Dalton’s law, yA = pA/P, for the gas and using xA = cA/c for the
creases with increasing concentration of A. Denote two liquid. The concentration of the liquid is close to that of pure water
slopes of the equilibrium line by or 3.43 lbmol/ft3 or 55.0 kmol/m3. Thus, the mole fractions at equi-
librium are:
yAi − yA∗
mx = (3-240) ySO2 xSO2
xAi − xAb
0.0191 0.000563
and 0.0303 0.000846
0.0546 0.001408
yAb − yAi
my = (3-241) 0.0850 0.001971
xA∗ − xAi
These data are fitted with average and maximum absolute devia-
Substituting (3-240) and (3-241) into (3-238) and (3-239), tions of 0.91% and 1.16%, respectively, by the quadratic equation
respectively, gives
ySO2 = 29.74xSO2 + 6,733xSO
2
2
(1)
1 1 mx
= + (3-242) Thus, differentiating, the slope of the equilibrium curve is given by
Ky ky kx
dy
m= = 29.74 + 13,466xSO2 (2)
and dx
1 1 1 The given mass-transfer coefficients can be converted to kx and ky
= + (3-243)
Kx kx m y ky by (3-227) and (3-228):
kmol
k x = kc c = 0.18(55.0) = 9.9
h-m2
EXAMPLE 3.20 kmol
k y = k p P = 0.040(2)(101.3) = 8.1 .
h-m2
Sulfur dioxide (A) is absorbed into water in a packed column. At a
certain location, the bulk conditions are 50°C, 2 atm, yAb = 0.085, (a) From (1) for xAb = 0.001, yA∗ = 29.74(0.001) + 6,733(0.001) 2
and xAb = 0.001. Equilibrium data for SO2 between air and = 0.0365. From (1), for yAb = 0.085, we solve the quadratic
water at 50°C are equation to obtain xA∗ = 0.001975.
Summary 111
The average slope in this range is From (3-229),
0.085 − 0.0365 9.9 kmol 8.1 kmol

m= = 49.7 . k x = = 9.9 and k y = = 8.85
0.001975 − 0.001 0.9986 h-m2 0.915 h-m2
From an examination of (3-242) and (3-243), the liquid-phase
resistance is controlling because the term in kx is much larger than From (3-243),
the term in ky. Therefore, from (3-243), using m = mx,
1 kmol
Kx = = 9.71
1 1 1 (1/9.9) + [1/56.3(8.85)] h-m2
= + = 0.1010 + 0.0025 = 0.1035
Kx 9.9 49.7(8.1)
From (3-223),
kmol
or K x = 9.66
h-m2 kmol
NA = 9.71(0.001975 − 0.001) = 0.00947
h-m2
From (3-223),
which is only a very slight change from parts (a) and (b), where the
kmol
NA = 9.66(0.001975 − 0.001) = 0.00942 . bulk-flow effect was ignored. The effect is very small because here
h-m2 it is important only in the gas phase; but the liquid-phase resistance
(b) From part (a), the gas-phase resistance is almost negligible. is controlling.
Therefore, yAi ≈ yAb and xAi ≈ xA∗ . (d) The relative magnitude of the mass-transfer resistances can be
From (3-241), the slope my must, therefore, be taken at the written as
point yAb = 0.085 and xA∗ = 0.001975 on the equilibrium line.
From (2), my = 29.74 + 13,466(0.001975) = 56.3. From 1/m y k y 1/(56.3)(8.85)
= = 0.02
(3-243), 1/k x 1/9.9
1 kmol Thus, the gas-phase resistance is only 2% of the liquid-phase resis-
Kx = = 9.69 ,
(1/9.9) + [1/(56.3)(8.1)] h-m2 tance. The interface vapor mole fraction can be obtained from
(3-223), after accounting for the bulk-flow effect:
giving NA = 0.00945 kmol/h-m2. This is only a slight change from
part (a). NA 0.00947
yAi = yAb −
= 0.085 − = 0.084
(c) We now correct for bulk flow. From the results of parts (a) ky 8.85
and (b), we have
Similarly,
yAb = 0.085, yAi = 0.085, xAi = 0.1975, xAb = 0.001
NA 0.00947
( yB ) LM = 1.0 − 0.085 = 0.915 and (xB ) LM ≈ 0.9986 xAi = + xAb = + 0.001 = 0.00196
k x 9.9
SUMMARY
1. Mass transfer is the net movement of a component in a mixture 5. Diffusivity values vary by orders of magnitude. Typical values
from one region to another region of different concentration, often are 0.10, 1 × 10−5 , and 1 × 10−9 cm2 /s for ordinary molecular
between two phases across an interface. Mass transfer occurs by diffusion of a solute in a gas, liquid, and solid, respectively.
molecular diffusion, eddy diffusion, and bulk flow. Molecular dif- 6. Fick’s second law for unsteady-state diffusion is readily ap-
fusion occurs by a number of different driving forces, including plied to semi-infinite and finite stagnant media, including certain
concentration (the most important), pressure, temperature, and ex- anisotropic materials.
ternal force fields.
7. Molecular diffusion under laminar-flow conditions can be deter-
2. Fick’s first law for steady-state conditions states that the mass- mined from Fick’s first and second laws, provided that velocity pro-
transfer flux by ordinary molecular diffusion is equal to the product files are available. Common cases include falling liquid-film flow,
of the diffusion coefficient (diffusivity) and the negative of the con- boundary-layer flow on a flat plate, and fully developed flow in a
centration gradient. straight, circular tube. Results are often expressed in terms of a mass-
3. Two limiting cases of mass transfer are equimolar counterdif- transfer coefficient embedded in a dimensionless group called the
fusion (EMD) and unimolecular diffusion (UMD). The former is Sherwood number. The mass-transfer flux is given by the product of
also a good approximation for dilute conditions. The latter must in- the mass-transfer coefficient and a concentration driving force.
clude the bulk-flow effect. 8. Mass transfer in turbulent flow is often predicted by analogy
4. When experimental data are not available, diffusivities in gas to heat transfer. Of particular importance is the Chilton–Colburn
and liquid mixtures can be estimated. Diffusivities in solids, in- analogy, which utilizes empirical j-factor correlations and the
cluding porous solids, crystalline solids, metals, glass, ceramics, dimensionless Stanton number for mass transfer. A more accurate
polymers, and cellular solids, are best measured. For some solids— equation by Churchill and Zajic should be used for flow in tubes,
for example, wood—diffusivity is an anisotropic property. particularly at high Schmidt and Reynolds numbers.
9. A number of models have been developed for mass transfer 10. The two-film theory of Whitman (more properly referred to as
across a two-fluid interface and into a liquid. These include the film a two-resistance theory) is widely used to predict the mass-transfer
theory, penetration theory, surface-renewal theory, and the film- flux from one fluid phase, across an interface, and into another fluid
penetration theory. These theories predict mass-transfer coeffi- phase, assuming equilibrium at the interface. One resistance is often
cients that are proportional to the diffusivity raised to an exponent controlling. The theory defines an overall mass-transfer coefficient
that varies from 0.5 to 1.0. Most experimental data provide expo- that is determined from the separate coefficients for each of the two
nents ranging from 0.5 to 0.75. phases and the equilibrium relationship at the interface.
REFERENCES
1. TAYLOR, R., and R. KRISHNA, Multicomponent Mass Transfer, John 28. BIRD, R.B., W.E. STEWART, and E.N. LIGHTFOOT, Transport Phenom-
Wiley and Sons, New York (1993). ena, 2nd ed., John Wiley and Sons, New York (2002).
2. POLING, B.E., J.M. PRAUSNITZ, and J.P. O’CONNELL, The Properties of 29. CHURCHILL, R.V., Operational Mathematics, 2nd ed., McGraw-Hill,
Liquids and Gases, 5th ed., McGraw-Hill, New York (2001). New York (1958).
3. FULLER, E.N., P.D. SCHETTLER, and J.C. GIDDINGS, Ind. Eng. Chem., 30. ABRAMOWITZ, M., and I.A. STEGUN, Eds., Handbook of Mathematical
58 (5), 18–27 (1966). Functions, National Bureau of Standards, Applied Mathematics Series 55,
4. TAKAHASHI, S., J. Chem. Eng. Jpn., 7, 417–420 (1974). Washington, DC (1964).
5. SLATTERY, J.C., M.S. thesis, University of Wisconsin, Madison 31. NEWMAN, A.B., Trans. AIChE, 27, 310–333 (1931).
(1955). 32. GRIMLEY, S.S., Trans. Inst. Chem. Eng. (London), 23, 228–235
6. WILKE, C.R., and P. CHANG, AIChE J., 1, 264–270 (1955). (1948).
7. HAYDUK, W., and B.S. MINHAS, Can. J. Chem. Eng., 60, 295–299 33. JOHNSTONE, H.F., and R.L. PIGFORD, Trans. AIChE, 38, 25–51 (1942).
(1982). 34. OLBRICH, W.E., and J.D. WILD, Chem. Eng. Sci., 24, 25–32 (1969).
8. QUAYLE, O.R., Chem. Rev., 53, 439–589 (1953). 35. CHURCHILL, S.W., The Interpretation and Use of Rate Data: The Rate
9. VIGNES, A., Ind. Eng. Chem. Fundam., 5, 189–199 (1966). Concept, McGraw-Hill, New York (1974).
10. SORBER, H.A., Handbook of Biochemistry, Selected Data for Molecu- 36. CHURCHILL, S.W., and R. USAGI, AIChE J., 18, 1121–1128 (1972).
lar Biology, 2nd ed., Chemical Rubber Co., Cleveland, OH (1970). 37. EMMERT, R.E., and R.L. PIGFORD, Chem. Eng. Prog., 50, 87–93
11. GEANKOPLIS, C.J., Transport Processes and Separation Process (1954).
Principles, 4th ed., Prentice-Hall, Upper Saddle River, NJ (2003). 38. PRANDTL, L., Proc. 3rd Int. Math. Congress, Heidelberg (1904);
12. FRIEDMAN, L., and E.O. KRAEMER, J. Am. Chem. Soc., 52, 1298–1304, reprinted in NACA Tech. Memo 452 (1928).
1305–1310, 1311–1314 (1930). 39. BLASIUS, H., Z. Math Phys., 56, 1–37 (1908); reprinted in NACA Tech.
13. BOUCHER, D.F., J.C. BRIER, and J.O. OSBURN, Trans. AIChE, 38, Memo 1256.
967–993 (1942). 40. SCHLICHTING, H., Boundary Layer Theory, 4th ed., McGraw-Hill,
14. BARRER, R.M., Diffusion in and through Solids, Oxford University New York (1960).
Press, London (1951). 41. POHLHAUSEN, E., Z. Angew. Math Mech., 1, 252 (1921).
15. SWETS, D.E., R.W. LEE, and R.C. FRANK, J. Chem. Phys., 34, 17–22 42. POHLHAUSEN, E., Z. Angew. Math Mech., 1, 115–121 (1921).
(1961). 43. LANGHAAR, H.L., Trans. ASME, 64, A-55 (1942).
16. LEE, R. W., J. Chem, Phys., 38, 44–455 (1963). 44. GRAETZ, L., Ann. d. Physik, 25, 337–357 (1885).
17. WILLIAMS, E.L., J. Am. Ceram. Soc., 48, 190–194 (1965). 45. SELLARS, J.R., M. TRIBUS, and J.S. KLEIN, Trans. ASME, 78, 441–448
18. SUCOV, E.W., J. Am. Ceram. Soc., 46, 14–20 (1963). (1956).
19. KINGERY, W.D., H.K. BOWEN, and D.R. UHLMANN, Introduction to 46. LEVEQUE, J., Ann. Mines, [12], 13, 201, 305, 381 (1928).
Ceramics, 2nd ed., John Wiley and Sons, New York (1976). 47. KNUDSEN, J.G., and D.L. KATZ, Fluid Dynamics and Heat Transfer,
20. FERRY, J.D., Viscoelastic Properties of Polymers, John Wiley and McGraw-Hill, New York (1958).
Sons, New York (1980). 48. HAUSEN, H., Verfahrenstechnik Beih. z. Ver. Deut. Ing., 4, 91 (1943).
21. RHEE, C.K., and J.D. FERRY, J. Appl. Polym. Sci., 21, 467–476 (1977). 49. LINTON, W.H. Jr., and T.K. SHERWOOD, Chem. Eng. Prog., 46, 258–264
22. BRANDRUP, J., and E.H. IMMERGUT, Eds., Polymer Handbook, 3rd ed., (1950).
John Wiley and Sons, New York (1989). 50. REYNOLDS, O., Trans. Roy. Soc. (London), 174A, 935–982 (1883).
23. GIBSON, L.J., and M.F. ASHBY, Cellular Solids, Structure and Proper- 51. BOUSSINESQ, J., Mem. Pre. Par. Div. Sav., XXIII, Paris (1877).
ties, Pergamon Press, Elmsford, NY (1988).
52. PRANDTL, L., Z. Angew, Math Mech., 5, 136 (1925); reprinted in
24. STAMM, A.J., Wood and Cellulose Science, Ronald Press, New York NACA Tech. Memo 1231 (1949).
(1964).
53. REYNOLDS, O., Proc. Manchester Lit. Phil. Soc., 14, 7 (1874).
25. SHERWOOD, T.K., Ind. Eng. Chem., 21, 12–16 (1929).
54. COLBURN, A.P., Trans. AIChE, 29, 174–210 (1933).
26. CARSLAW, H.S., and J.C. JAEGER, Heat Conduction in Solids, 2nd ed.,
55. CHILTON, T.H., and A.P. COLBURN, Ind. Eng. Chem., 26, 1183–1187
Oxford University Press, London (1959).
(1934).
27. CRANK, J., The Mathematics of Diffusion, Oxford University Press,
56. PRANDTL, L., Physik. Z., 11, 1072 (1910).
London (1956).
Exercises 113
57. FRIEND, W.L., and A.B. METZNER, AIChE J., 4, 393–402 (1958). 68. LAUNDER, B.E., and D.B. SPALDING, Lectures in Mathematical Models
58. NERNST, W., Z. Phys. Chem., 47, 52 (1904). of Turbulence, Academic Press, New York (1972).
59. HIGBIE, R., Trans. AIChE, 31, 365–389 (1935). 69. HENG, L., C. CHAN, and S.W. CHURCHILL, Chem. Eng. J., 71, 163
(1998).
60. DANCKWERTS, P.V., Ind. Eng. Chem., 43, 1460–1467 (1951).
70. CHURCHILL, S.W., and S.C. ZAJIC, AIChE J., 48, 927–940 (2002).
61. LEVENSPIEL, O., Chemical Reaction Engineering, 3rd ed., John Wiley
and Sons, New York (1999). 71. CHURCHILL, S.W., “Turbulent Flow and Convection: The Prediction of
Turbulent Flow and Convection in a Round Tube,” in Advances in Heat
62. TOOR, H.L., and J.M. MARCHELLO, AIChE J., 4, 97–101 (1958). Transfer, J.P. Hartnett, and T.F. Irvine, Jr., Ser. Eds., Academic Press, New
63. WHITMAN, W.G., Chem. Met. Eng., 29, 146–148 (1923). York, 34, 255–361 (2001).
64. vAN DRIEST, E.R., J. Aero Sci., 1007–1011 and 1036 (1956). 72. YU, B., H. OZOE, and S.W. CHURCHILL, Chem. Eng. Sci., 56, 1781
65. REICHARDT, H., Fundamentals of Turbulent Heat Transfer, NACA (2001).
Report TM-1408 (1957). 73. CHURCHILL, S.W., and C. CHAN, Ind. Eng. Chem. Res., 34, 1332
66. DREW, T.B., E.C. KOO, and W.H. MCADAMS, Trans. Am. Inst. Chem. (1995).
Engrs., 28, 56 (1933). 74. CHURCHILL, S.W., AIChE J., 43, 1125 (1997).
67. NIKURADSE, J., VDI-Forschungsheft, p. 361 (1933).
EXERCISES
Section 3.1 level was originally 0.5 in. below the top of the tube. The diffusiv-
3.1 A beaker filled with an equimolar liquid mixture of ethyl ity of water in air at 25°C is 0.256 cm2/s.
alcohol and ethyl acetate evaporates at 0°C into still air at 101 kPa (a) How long will it take for the liquid level in the tube to drop 3 in.?
(1 atm) total pressure. Assuming Raoult’s law applies, what will be (b) Make a plot of the liquid level in the tube as a function of time
the composition of the liquid remaining when half the original ethyl for this period.
alcohol has evaporated, assuming that each component evaporates 3.5 Two bulbs are connected by a tube, 0.002 m in diameter and
independently of the other? Also assume that the liquid is always 0.20 m in length. Initially bulb 1 contains argon, and bulb 2 con-
well mixed. The following data are available: tains xenon. The pressure and temperature are maintained at 1 atm
and 105°C, at which the diffusivity is 0.180 cm2/s. At time t = 0,
Vapor Pressure, Diffusivity in Air diffusion is allowed to occur between the two bulbs. At a later time,
kPa at 0°C m2/s the argon mole fraction in the gas at end 1 of the tube is 0.75, and
Ethyl acetate (AC) 3.23 6.45 × 10−6 0.20 at the other end. Determine at the later time:
Ethyl alcohol (AL) 1.62 9.29 × 10−6 (a) The rates and directions of mass transfer of argon and xenon
(b) The transport velocity of each species
3.2 An open tank, 10 ft in diameter and containing benzene at (c) The molar average velocity of the mixture
25°C, is exposed to air in such a manner that the surface of the liq-
uid is covered with a stagnant air film estimated to be 0.2 in. thick. Section 3.2
If the total pressure is 1 atm and the air temperature is 25°C, what 3.6 The diffusivity of toluene in air was determined experimen-
loss of material in pounds per day occurs from this tank? The spe- tally by allowing liquid toluene to vaporize isothermally into air
cific gravity of benzene at 60°F is 0.877. The concentration of ben- from a partially filled vertical tube 3 mm in diameter. At a temper-
zene at the outside of the film is so low that it may be neglected. For ature of 39.4°C, it took 96 × 104 s for the level of the toluene to
benzene, the vapor pressure at 25°C is 100 torr, and the diffusivity drop from 1.9 cm below the top of the open tube to a level of 7.9 cm
in air is 0.08 cm2/s. below the top. The density of toluene is 0.852 g/cm3, and the vapor
3.3 An insulated glass tube and condenser are mounted on a pressure is 57.3 torr at 39.4°C. The barometer reading was 1 atm.
reboiler containing benzene and toluene. The condenser returns liq- Calculate the diffusivity and compare it with the value predicted
uid reflux so that it runs down the wall of the tube. At one point in from (3-36). Neglect the counterdiffusion of air.
the tube the temperature is 170°F, the vapor contains 30 mol% 3.7 An open tube, 1 mm in diameter and 6 in. long, has pure hy-
toluene, and the liquid reflux contains 40 mol% toluene. The effec- drogen blowing across one end and pure nitrogen blowing across
tive thickness of the stagnant vapor film is estimated to be 0.1 in. the other. The temperature is 75°C.
The molar latent heats of benzene and toluene are equal. Calculate
(a) For equimolar counterdiffusion, what will be the rate of trans-
the rate at which toluene and benzene are being interchanged by
fer of hydrogen into the nitrogen stream (mol/s)? Estimate the dif-
equimolar countercurrent diffusion at this point in the tube in
fusivity from (3-36).
lbmol/h-ft2.
(b) For part (a), plot the mole fraction of hydrogen against distance
Diffusivity of toluene in benzene = 0.2 ft2/h. from the end of the tube past which nitrogen is blown.
Pressure = 1 atm total pressure (in the tube). 3.8 Some HCl gas diffuses across a film of air 0.1 in. thick at
Vapor pressure of toluene at 170°F = 400 torr. 20°C. The partial pressure of HCl on one side of the film is 0.08
3.4 Air at 25°C with a dew-point temperature of 0°C flows past atm and it is zero on the other. Estimate the rate of diffusion, as mol
the open end of a vertical tube filled with liquid water maintained at HCl/s-cm2, if the total pressure is (a) 10 atm, (b) 1 atm, (c) 0.1 atm.
25°C. The tube has an inside diameter of 0.83 in., and the liquid The diffusivity of HCl in air at 20°C and 1 atm is 0.145 cm2/s.
3.9 Estimate the diffusion coefficient for the gaseous binary sys- 3.16 Experimental liquid-phase activity-coefficient data are
tem nitrogen (A)/toluene (B) at 25°C and 3 atm using the method of given in Exercise 2.23 for the ethanol/benzene system at 45°C. Es-
Fuller et al. timate and plot diffusion coefficients for both ethanol and benzene
3.10 For the mixture of Example 3.3, estimate the diffusion coef- over the entire composition range.
ficient if the pressure is increased to 100 atm using the method of 3.17 Estimate the diffusion coefficient of NaOH in a 1-M aqueous
Takahashi. solution at 25°C.
3.11 Estimate the diffusivity of carbon tetrachloride at 25°C in a 3.18 Estimate the diffusion coefficient of NaCl in a 2-M aqueous
dilute solution of: (a) Methanol, (b) Ethanol, (c) Benzene, and solution at 18°C. Compare your estimate with the experimental
(d) n-Hexane by the method of Wilke–Chang and Hayduk–Minhas. value of 1.28 × 10−5 cm2 /s.
Compare the estimated values with the following experimental 3.19 Estimate the diffusivity of N2 in H2 in the pores of a catalyst
observations: at 300°C and 20 atm if the porosity is 0.45 and the tortuosity is 2.5.
Assume ordinary molecular diffusion in the pores.
Solvent Experimental DAB, cm2/s
3.20 Gaseous hydrogen at 150 psia and 80°F is stored in a small,
Methanol 1.69 × 10−5 cm2 /s at 15◦ C
spherical, steel pressure vessel having an inside diameter of 4 in. and
Ethanol 1.50 × 10−5 cm2 /s at 25◦ C
a wall thickness of 0.125 in. At these conditions, the solubility of hy-
Benzene 1.92 × 10−5 cm2 /s at 25◦ C
drogen in steel is 0.094 lbmol/ft3 and the diffusivity of hydrogen in
n-Hexane 3.70 × 10−5 cm2 /s at 25◦ C
steel is 3.0 × 10−9 cm2 /s. If the inner surface of the vessel remains
3.12 Estimate the liquid diffusivity of benzene (A) in formic acid saturated at the existing hydrogen pressure and the hydrogen partial
(B) at 25°C and infinite dilution. Compare the estimated value to pressure at the outer surface is assumed to be zero, estimate:
that of Example 3.6 for formic acid at infinite dilution in benzene. (a) The initial rate of mass transfer of hydrogen through the metal
3.13 Estimate the liquid diffusivity of acetic acid at 25°C in a wall
dilute solution of: (a) Benzene, (b) Acetone, (c) Ethyl acetate, and (b) The initial rate of pressure decrease inside the vessel
(d) Water by an appropriate method. Compare the estimated values (c) The time in hours for the pressure to decrease to 50 psia,
with the following experimental values: assuming the temperature stays constant at 80°F
3.21 A polyisoprene membrane of 0.8-m thickness is to be used
Solvent Experimental DAB, cm2/s
to separate a mixture of methane and H2. Using the data in
Benzene 2.09 × 10−5 cm2 /s at 25◦ C Table 14.9 and the following compositions, estimate the mass-
Acetone 2.92 × 10−5 cm2 /s at 25◦ C transfer flux of each of the two species.
Ethyl acetate 2.18 × 10−5 cm2 /s at 25◦ C
Water 1.19 × 10−5 cm2 /s at 20◦ C Partial Pressures, MPa
3.14 Water in an open dish exposed to dry air at 25°C is found to Membrane Side 1 Membrane Side 2
vaporize at a constant rate of 0.04 g/h-cm2. Assuming the water
surface to be at the wet-bulb temperature of 11.0°C, calculate the Methane 2.5 0.05
effective gas-film thickness (i.e., the thickness of a stagnant air film Hydrogen 2.0 0.20
that would offer the same resistance to vapor diffusion as is actually
encountered at the water surface). Section 3.3
3.15 Isopropyl alcohol is undergoing mass transfer at 35°C and 3.22 A 3-ft depth of stagnant water at 25°C lies on top of a
2 atm under dilute conditions through water, across a phase bound- 0.10-in. thickness of NaCl. At time < 0, the water is pure. At time =
ary, and then through nitrogen. Based on the date given below, 0, the salt begins to dissolve and diffuse into the water. If the con-
estimate for isopropyl alcohol: centration of salt in the water at the solid–liquid interface is main-
(a) The diffusivity in water using the Wilke–Chang equation tained at saturation (36 g NaCl/100 g H2O) and the diffusivity of
(b) The diffusivity in nitrogen using the Fuller et al. equation NaCl in water is 1.2 × 10−5 cm2 /s, independent of concentration,
(c) The product, DAB M , in water estimate, by assuming the water to act as a semi-infinite medium,
the time and the concentration profile of salt in the water when
(d) The product, DAB M , in air
(a) 10% of the salt has dissolved
where M is the molar density of the mixture.
(b) 50% of the salt has dissolved
Using the above results, compare:
(c) 90% of the salt has dissolved
(e) The diffusivities in parts (a) and (b)
(f) The diffusivity-molar density products in Parts (c) and (d) 3.23 A slab of dry wood of 4-in. thickness and sealed edges is
exposed to air of 40% relative humidity. Assuming that the two
Lastly: unsealed faces of the wood immediately jump to an equilibrium
(g) What conclusions can you come to about molecular diffusion moisture content of 10 lb H2O per 100 lb of dry wood, determine
in the liquid phase versus the gaseous phase? the time for the moisture to penetrate to the center of the slab (2 in.
Data: from either face). Assume a diffusivity of water in the wood as
8.3 × 10−6 cm2 /s.
Component Tc, °R Pc, psia Zc vL, cm3/mol 3.24 A wet, clay brick measuring 2 × 4 × 6 in. has an initial uni-
Nitrogen 227.3 492.9 0.289 — form moisture content of 12 wt%. At time = 0, the brick is exposed
Isopropyl alcohol 915 691 0.249 76.5 on all sides to air such that the surface moisture content is
Exercises 115
maintained at 2 wt%. After 5 h, the average moisture content is Section 3.5

8 wt%. Estimate: 3.32 Water at 25°C flows at 5 ft/s through a straight, cylindrical
(a) The diffusivity of water in the clay in cm2/s. tube cast from benzoic acid, of 2-in. inside diameter. If the tube is
(b) The additional time for the average moisture content to reach 10 ft long, and fully developed, turbulent flow is assumed, estimate
4 wt%. All moisture contents are on a dry basis. the average concentration of benzoic acid in the water leaving the
3.25 A spherical ball of clay, 2 in. in diameter, has an initial mois- tube. Physical properties are given in Example 3.15.
ture content of 10 wt%. The diffusivity of water in the clay is 3.33 Air at 1 atm flows at a Reynolds number of 50,000 normal to
5 × 10−6 cm2 /s. At time t = 0, the surface of the clay is brought into a long, circular, 1-in.-diameter cylinder made of naphthalene.
contact with air such that the moisture content at the surface is main- Using the physical properties of Example 3.14 for a temperature of
tained at 3 wt%. Estimate the time for the average moisture content in 100°C, calculate the average sublimation flux in kmol/s-m2.
the sphere to drop to 5 wt%. All moisture contents are on a dry basis. 3.34 For the conditions of Exercise 3.33, calculate the initial
average rate of sublimation in kmol/s-m2 for a spherical particle of
Section 3.4 1-in. initial diameter. Compare this result to that for a bed packed
3.26 Estimate the rate of absorption of pure oxygen at 10 atm and with naphthalene spheres with a void fraction of 0.5.
25°C into water flowing as a film down a vertical wall 1 m high and
Section 3.6
6 cm in width at a Reynolds number of 50 without surface ripples.
Assume the diffusivity of oxygen in water is 2.5 × 10−5 cm2 /s and 3.35 Carbon dioxide is stripped from water by air in a wetted-
that the mole fraction of oxygen in water at saturation for the above wall tube. At a certain location, where the pressure is 10 atm
temperature and pressure is 2.3 × 10−4 . and the temperature is 25°C, the mass-transfer flux of CO2 is
1.62 lbmol/h-ft2. The partial pressures of CO2 are 8.2 atm at the in-
3.27 For the conditions of Example 3.13, determine at what
terface and 0.1 atm in the bulk gas. The diffusivity of CO2 in air at
height from the top the average concentration of CO2 would corre-
these conditions is 1.6 × 10−2 cm2 /s. Assuming turbulent flow of
spond to 50% of saturation.
the gas, calculate by the film theory, the mass-transfer coefficient kc
3.28 Air at 1 atm flows at 2 m/s across the surface of a 2-in.-long for the gas phase and the film thickness.
surface that is covered with a thin film of water. If the air and water
3.36 Water is used to remove CO2 from air by absorption in a col-
are maintained at 25°C, and the diffusivity of water in air at these
umn packed with Pall rings. At a certain region of the column
conditions is 0.25 cm2/s, estimate the mass flux for the evaporation
where the partial pressure of CO2 at the interface is 150 psia and
of water at the middle of the surface assuming laminar boundary-
the concentration in the bulk liquid is negligible, the absorption rate
layer flow. Is this assumption reasonable?
is 0.017 lbmol/h-ft2. The diffusivity of CO2 in water is
3.29 Air at 1 atm and 100°C flows across a thin, flat plate of 2.0 × 10−5 cm2/s. Henry’s law for CO2 is p = Hx, where H = 9,000
naphthalene that is 1 m long, causing the plate to sublime. The psia. Calculate:
Reynolds number at the trailing edge of the plate is at the upper
(a) The liquid-phase mass-transfer coefficient and the film
limit for a laminar boundary layer. Estimate:
thickness
(a) The average rate of sublimation in kmol/s-m2
(b) Contact time for the penetration theory
(b) The local rate of sublimation at a distance of 0.5 m from the
(c) Average eddy residence time and the probability distribution
leading edge of the plate
for the surface-renewal theory
Physical properties are given in Example 3.14.
3.37 Determine the diffusivity of H2S in water, using the penetra-
3.30 Air at 1 atm and 100°C flows through a straight, 5-cm- tion theory, from the following data for the absorption of H2S into
diameter circular tube, cast from naphthalene, at a Reynolds number a laminar jet of water at 20°C.
of 1,500.Air entering the tube has an established laminar-flow velocity
Jet diameter = 1 cm, Jet length = 7 cm, and Solubility of H2S in
profile. Properties are given in Example 3.14. If pressure drop through
water = 100 mol/m3
the tube is negligible, calculate the length of tube needed for the aver-
age mole fraction of naphthalene in the exiting air to be 0.005. The average rate of absorption varies with the flow rate of the jet as
follows:
3.31 A spherical water drop is suspended from a fine thread in
still, dry air. Show: Jet Flow Rate, Rate of Absorption,
(a) That the Sherwood number for mass transfer from the surface cm3/s mol/s 106
of the drop into the surroundings has a value of 2 if the characteris- 0.143 1.5
tic length is the diameter of the drop. 0.568 3.0
If the initial drop diameter is 1 mm, the air temperature is 38°C, the 1.278 4.25
drop temperature is 14.4°C, and the pressure is 1 atm, calculate: 2.372 6.15
(b) The initial mass of the drop in grams. 3.571 7.20
(c) The initial rate of evaporation in grams per second. 5.142 8.75
(d) The time in seconds for the drop diameter to be reduced to
Section 3.7
0.2 mm.
(e) The initial rate of heat transfer to the drop. If the Nusselt num- 3.38 In a test on the vaporization of H2O into air in a wetted-wall
ber is also 2, is the rate of heat transfer sufficient to supply the heat column, the following data were obtained:
of vaporization and sensible heat of the evaporated water? If not, Tube diameter, 1.46 cm, Wetted-tube length, 82.7 cm
what will happen? Air rate to tube at 24°C and 1 atm, 720 cm3/s
Temperature of inlet water, 25.15°C, Temperature of outlet water, The operation was countercurrent, with the gas entering at the bot-
25.35°C tom of the vertical tower and the acid passing down in a thin film
Partial pressure of water in inlet air, 6.27 torr, and in outlet air, on the inner wall. The change in acid strength was inappreciable,
20.1 torr and the vapor pressure of ammonia over the liquid may be as-
The value for the diffusivity of water vapor in air is 0.22 cm2/s at sumed to have been negligible because of the use of a strong
0°C and 1 atm. The mass velocity of air is taken relative to the pipe acid for absorption. Calculate the mass-transfer coefficient, kp,
wall. Calculate: from the data.
(a) Rate of mass transfer of water into the air 3.40 A new type of cooling-tower packing is being tested in a lab-
(b) KG for the wetted-wall column oratory column. At two points in the column, 0.7 ft apart, the fol-
lowing data have been taken. Calculate the overall volumetric
3.39 The following data were obtained by Chamber and mass-transfer coefficient Ky a that can be used to design a large,
Sherwood [Ind. Eng. Chem., 29, 1415 (1937)] on the absorption of packed-bed cooling tower, where a is the mass-transfer area, A, per
ammonia from an ammonia-air system by a strong acid in a wetted- unit volume, V, of tower.
wall column 0.575 in. in diameter and 32.5 in. long:
Inlet acid (2-N H2SO4) temperature, °F 76 Bottom Top
Outlet acid temperature, °F 81
Water temperature, °F 120 126
Inlet air temperature, °F 77
Water vapor pressure, psia 1.69 1.995
Outlet air temperature, °F 84 Mole fraction H2O in air 0.001609 0.0882
Total pressure, atm 1.00 Total pressure, psia 14.1 14.3
Partial pressure NH3 in inlet gas, atm 0.0807 Air rate, lbmol/h 0.401 0.401
Partial pressure NH3 in outlet gas, atm 0.0205 Column area, ft2 0.5 0.5
Air rate, lbmol/h 0.260 Water rate, lbmol/h (approximate) 20 20

Mass Transfer and Diffusion: Ass Transfer Is The Net Movement of A Component in A

Uploaded by

Copyright:

Available Formats

Mass Transfer and Diffusion: Ass Transfer Is The Net Movement of A Component in A

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mass Transfer and Diffusion: Ass Transfer Is The Net Movement of A Component in A

Uploaded by

Copyright:

Available Formats

Chapter 3

Mass Transfer and Diffusion

3.0 INSTRUCTIONAL OBJECTIVES

After completing this chapter, you should be able to:

3.1 STEADY-STATE, ORDINARY

xA xA Figure 3.1 Concentration profiles for limiting

1. Equimolar counterdiffusion (EMD) by differentiation,

This idealization is closely approached in distillation. From

EXAMPLE 3.2 From (3-32),

0.0 0.1360 0.9040 0.9040 −0.9040

(a) Take z1 0. Then z 2 − z 1 = z = 0.5 cm. From Dalton’s law,

cDAB (−xA ) L dz C 15.9 F 14.7

3.2 DIFFUSION COEFFICIENTS

Table 3.2 Experimental Binary Diffusivities of Some Gas Pairs 3.5

Air—n-hexane 328 0.093 0.4

From (3-37), Tc, K Pc, atm

Diffusivity in Liquid Mixtures dilute conditions to more concentrated conditions, exten-

Atomic Volume Atomic Volume

Aniline 293 0.92

Ethanol 288 2.25

Table 3.5 Parachors for Representative Compounds

Parachor, Parachor, Parachor,

Source: Meissner, Chem. Eng. Prog., 45, 149–153 (1949).

Table 3.6 Structural Contributions for Estimating the Parachor

Carbon–hydrogen: R—[—CO—]—R (ketone)

Source: Quale [8].

The Stokes–Einstein and Wilke–Chang equations predict xA A xB B

Table 3.7 Limiting Ionic Conductances in Water at 25°C, in EXAMPLE 3.8

Table 3.9 Experimental Diffusivities of Large Biological Proteins in Aqueous Solutions

Important practical applications exist for diffusion of light

Table 3.12 Diffusivities of Solutes in Rubbery Polymers

smaller molecules have higher diffusivities. A more detailed From (3-50),

where Rearranging and simplifying,

Flow in Accumulation Flow out Semi-infinite Medium

Thus, using the Leibnitz√rule for differentiating the integral

The error function is included in most spreadsheet programs SOLUTION

0.8 0.6 0.2

Medium of Finite Thickness with Sealed Edges

centration of the solute in the medium is uniform at cAo . At 0.04 0.01

Figure 3.7 Unsteady-state diffusion in media

0.02 cAs – cAo = E

Long residence-time solution

general solution for intermediate values of is not available Thus,

3.41(1.96 × 10−5 )(10−4 ) Velocity

Thickness of Edge of Fully developed

Figure 3.15 Buildup of a laminar

EXAMPLE 3.15 3.5 MASS TRANSFER IN TURBULENT FLOW

Table 3.13 Some Useful Dimensionless Groups

Name Formula Meaning Analogy

Prandtl number, NPr = 1

NNu , Friend–Metzner NNu , Chilton–Colburn NNu , Churchill–Zajic

Prandtl number, NPr = 1000

NNu , Friend–Metzner NNu , Chilton–Colburn NNu , Churchill–Zajic

which is an acceptable 92% of the experimental value.

Figure 3.18 Theories for mass

following form of (3-79): Surface-Renewal Theory

The integrated average rate is

Gas Gas Liquid Liquid pAb

and For the liquid phase, using kc or kx,