You are on page 1of 53

Mass Transfer (ECH 143)

Problem Set 1

1. Show that
N
∑ Ji = 0
*

i=1

for an N-component system.

Solution
This flux is the molar flux relative to the molar-average velocity, or

J *i = N i − Ci v* .

Summing over all the components, one finds that

( )
N N N
∑ J i = ∑ N i − Ci v = ∑ N i − Cv .
* * *

i=1 i=1 i=1

However, by definition of the molar average velocity,

N N Ci 1 N
v* = ∑ x i vi = ∑ vi = ∑ N i .
i=1 i=1 C C i=1

Substitution yields 0=0, and the result is proven.

2. For a binary mixture of components A and B, it must be true that

jA + jB = J A M A + J BM B = 0 .

Use this result to show that

M BC *
JA = JA ,
ρ

where C is the total molar concentration and ρ is the density of the mixture.

Solution
We know that the velocity of component A can be written as the mass average velocity plus a
diffusive correction, or as the molar-average velocity with a diffusive correction, i.e.

NA J J*
= v + A = v* + A .
CA CA CA
The total molar flux of component A is given by

N A = C A v + J A = C A v* + J *A , (1)

and for component B we have

N B = CB v + J B = CB v* + J *B . (2)

Multiplying the equation for the flux of A by MA and the equation for the flux of B by MB
yilds

C A M A v + M A J A = C A M A v* + M A J *A
and
CBM B v + M B J B = CBM B v* + M B J *B .

Adding these two equations, and using the result given in the problem statement, we find

C A M A v + CBM B v = C A M A v* + CBM B v* + M A J *A + M B J *B

which simplifies to

ρv = ρv* + M A J *A + M B J *B

or

MA * MB *
v* = v − JA − JB . (3)
ρ ρ

Eq. (1) above gives another relation between the two velocities,

JA J*
v* = + v− A . (4)
CA CA

Substituting Eq. (4) into Eq. (3) and multiplying by concentration gives

⎡M M ⎤
J A + vC A = vC A − C A ⎢ A J *A + B J *B ⎥ + J *A .
⎣ ρ ρ ⎦

Then
⎡M M ⎤
J A = −C A ⎢ A J *A + B J *B ⎥ + J *A .
⎣ ρ ρ ⎦
But

J *B = −J *A ,

so
JA =
1
ρ
{ρ − C A [ M A − M B ]} J *A .

But since

ρ = C A M A + C BM B ,

we have that

M BC *
JA = JA
ρ

where
C = C A + CB .

3. Show that, in a binary mixture, the gradient of the mass fraction ωA is related to the gradient
in the mole fraction xA by

M AM B
∇ω A = ∇x A .
( x A M A + x B M B )2
Solution
In a binary mixture, the mass fraction is related to the mole fraction by

x AM A
ωA = .
x A M A + x BM B

Taking the derivative in some spatial dimension (y, say, to avoid confusion with “x”) we find

⎛ ∂x ∂x ⎞
x AM A ⎜ M A A + M B B ⎟
∂ω A ( x A M A + x BM B ) M A ∂x A ⎝ ∂y ∂y ⎠
= − .
∂y ( x AM A + x BM B )2 ∂y ( x A M A + x B M B )2
Using
xA + xB = 1 ,
and therefore also that

∂x B ∂x
=− A ,
∂y ∂y

This result simplifies to

∂ω A M AM B ∂x A
= .
∂y ( x A M A + x BM B ) ∂y
2

Since this same derivation could be done in any of the 3 coordinate directions, it must be that

M AM B
∇ω A = ∇x A .
( x A M A + x B M B )2

4. Use the results from Problems 2 and 3 to show that, if

J *A = −CD AB∇x A ,

where xA is the mole fraction of A in a binary mixture of A and B, then

jA = −ρD AB∇ω A ,

where the binary diffusivity DAB is the same in both cases.

Solution
Since

CA
xA =
C
and
ρA = C A M A ,

the result from Problem 3 is equivalent to

C2 M A M B
∇ω A = ∇x A .
ρ2

From Problem 2
M BC *
JA = JA
ρ
and therefore from the first form of Fick’s law

M BC2
JA = − D AB∇x A .
ρ

Expressing the gradient of the mole fraction in terms of the mass fraction gives

ρ
JA = − D AB∇ω A .
MA

Since the product of the molar flux JA and the molecular weight gives the mass flux, we have

jA = −ρD AB∇ω A .

The same diffusion coefficient is therefore correct in both forms of Fick’s law.
Mass Transfer (ECH 143)
Problem Set 2

1. a) Consider a tracer particle undergoing a random walk in one dimension. The walk consists
of N steps of unit length, in either the +x or –x direction. If each step is denoted by δxi
write expressions for the total distance x from the starting point and the total squared distance
x2 from the starting point. Express your results using the summation sign.

Solution
The total distance after N steps is

N
x = ∑ δx i
i=1

and the squared distance is

( )
N ⎛N ⎞ N N
x 2 = ∑ δx i ⎜ ∑ δx j ⎟ = ∑ ∑ δx iδx j .
i=1 ⎝ j=1 ⎠ i=1 j=1

b) Now suppose the tracer goes through a sequence of NW identical random walks such as
that described in part “a” above. Average your two expressions over the sequence of walks,
by adding the NW trajectories and dividing by NW. For random steps, and for NW sufficiently
large, it must be that

1 NW
δx i = ∑ δx i = 0
N W k=1
and
1 NW
δx iδx j = ∑ δx iδx j = 0 for i ≠ j .
N W k=1

From these results, and for the case where

δx i = ±1 ,

show that the mean position and mean-squared position of the tracer are described by

1 NW N
x = ∑ ∑ δx i = 0
N W k=1 i=1
and
1 NW ⎧ N N ⎫
x2 = ∑ ⎨∑ δx i ∑ δx j ⎬ = N .
N W k=1 ⎩ i=1 j=1 ⎭
Solution
Taking the expressions from part a,

N
x = ∑ δx i
i=1
and

( )
N ⎛N ⎞ N N
x 2 = ∑ δx i ⎜ ∑ δx j ⎟ = ∑ ∑ δx iδx j ,
i=1 ⎝ j=1 ⎠ i=1 j=1

we average over NW trajectories to get

1 NW N N ⎡ 1 NW ⎤ N
x = ∑ ∑ δx i = ∑ ⎢ ∑ δx i ⎥ = ∑ δx i = 0
N W k=1 i=1 i=1 ⎣ N W k=1 ⎦ i=1

and
1 NW N N N N 1 NW N N
x2 = ∑ ∑ ∑ δx iδx j = ∑ ∑ ∑ δx iδx j = ∑ ∑ δx iδx j = N .
N W k=1 i=1 j=1 i=1 j=1 N k=1 i=1 j=1

c) Using the matlab function randon_walk_1d_simulation_rjp, do simulations of one-


dimensional random walks consisting of 100 steps of magnitude +1 or -1. Plot the average
distance from the starting point, and the mean-squared distance from the starting point. Make
plots of your results for 1 trajectory, 10 trajectories, and 1000 trajectories, where each
trajectory contains 100 steps. Are your plots consistent with the results of part b?

Solution
Results will vary according to random number generator. Here are mine for 1, 10 and 1000
trajectories (green curves are displacement, blue ones are squared displacement)
These graphs are consistent with the predictions, although even with 1000 trajectories there is
some deviation from the predicted average distance squared of 100.

2. In class it was shown that diffusion causes a step change in concentration to “relax” according
to the solution

C(η) = C0 (1− erf(η)) ,

where the error function is defined by

2 η − ξ2
erf(η) = ∫ e dξ
π0

and

x
η= .
4Dt

a) Use the Leibnitz rule for differentiating an integral with variable limits to differentiate this
solution with respect to position. In its general form, the Leibnitz rule is (cf. Bird, Stewart and
Lightfoot, Transport Phenomena, 2nd Edition, Appendix C.3)

d β(t ) β(t )
∂f ⎛ ∂β ∂α ⎞
∫ f(x, t)dx = ∫ dx + ⎜ f(β, t) − f(α, t) ⎟ .
dt α(t ) α(t ) ∂t ⎝ ∂t ∂t ⎠
(A version of the same rule appears in the textbook on p. 134, but the author uses it incorrectly.)
Your derivative should yield a Gaussian function, as discussed in class.
Solution
To differentiate the error function with the Leibnitz rule given, replace “t” with “x” and note that
the bottom limit of integration is zero, so that

e− η
2
d η 2 η
∂ ⎛ 2 ∂η ⎞
− ∫ e − ξ dξ = − ∫ e − ξ dξ − ⎜ e − η
2
− 0⎟ = − .
dx 0 0 ∂x ⎝ ∂x ⎠ 4Dt

Including the prefactor, the derivative is given by

∂ ⎛ C⎞ e − x /4 Dt
2

=− .
∂x ⎜⎝ C0 ⎟⎠ πDt

This result is negative because the solution was for a step decrease in concentration.

b) Since the derivative of a solution to the equation

∂C ∂2 C
=D 2
∂t ∂x
yields another solution to the equation, the derivative taken in part “a” is a solution to a time-
dependent diffusion problem. Multiplying by a normalization factor, it takes the form

m e − x /4 Dt
2

C= ,
A 4πDt

where m is the mass of solute confined at t=0 to a thin slice with area A.

For a planar sheet of sucrose in water, plot the normalized concentration CAδ/m vs. dimensional
position x for the times 60s, 600s, and 6000s. Approximately how far has the sucrose spread
after 100 minutes? The binary diffusivity of sucrose in water is

cm 2
D = 0.46x10 −5
s
at 298K.

Solution
After 6000s (or 1 hour, 40 minutes) the sucrose has diffused approximately 0.5 cm.
3. Estimate the binary diffusivity of methane in nitrogen at 1 atm and 300K. Obtain the value of
the collision integral by using Eq. 11-3.6 in Ch. 11 of the monograph by Reid et al. (1976). How
does your result compare with the experimental result reported by Hsieh and Cheh (1976) (see
paper posted to the web site)?
Solution
From the theory of Chapman and Enskog, we have

⎡⎣( M A + M B ) / ( M A M B ) ⎤⎦
1/2
−3
D AB = 1.86x10 T 3/2
cm 2 / s
pσ 2ABΩD

where T is the temperature in Kelvin, p is the pressure in atmospheres, σ is a characteristic


molecular dimension in Angtroms, and ΩD is the collision integral.

For nitrogen and methane, we have

MW σ (Α) ε/kB (K)

N2 28 3.798 71.4
CH4 16 3.758 148.6

As recommended by Reid et al. (1976), we use


ε AB = ( ε A ε B ) = 103.0k BK
1/2

and

σ AB = ( σ A + σ B ) / 2 = 3.778A .

To obtain the collision integral, we need

k BT
T* = = 2.913 .
ε AB

Then

1.06036 0.19300 1.03587 1.176474


ΩD = 0.15610
+ + +
2.913 exp(0.47635 * 2.913) exp(1.52996 * 2.913) exp(3.89411* 2.913)

or

ΩD = 0.89746 + 0.048187 + 0.012016 + 1.394x10 −5 = 0.958 .

Then the Chapman-Enskog result is

0.31339
D AB = 1.858x10 −3 ( 300 )
3/2
cm 2 / s = .221cm 2 / s .
( 3.778 ) ( 0.958 )
2

This result differs from the experimental results of Hsieh and Cheh (1976) by about 3%.

4. The protein α2-macroglobulin has a molecular weight of 820,000 g/mol and a diffusion
coefficient in water of

cm 2
D = 2.41x10 −7 (at 20 oC) ,
s
according to the Handbook of Biochemistry and Molecular Biology (CRC Press, 1976).
Assuming the viscosity of water is

µ = 1cP = 0.001Pa.s ,

calculate the Stokes-Einstein radius (or “hydrodynamic radius”) of this protein.


Solution
For a globular protein, we can use the Stokes-Einstein equation to relate the diffusion coefficient
and radius:
kT
D= .
6πµR

Rearranging, and substituting known values, we find

⎛ 10 7 gcm 2s −3 ⎞
kT
(1.38x10 J / K ) ( 293K ) ⎜
−23

⎝ Js −1 ⎟⎠
R= = = 0.89x10 −6 cm .
6πµD 6π ( 0.01g / cm.s ) ( 2.41x10 cm / s )
−7 2

This corresponds to

R = 0.89x10 −8 m = 8.9nm .
The radius of a molecule of water is approximately 0.145nm, smaller by a factor of 61. This
difference supports the use of the continuum approximation to represent the water molecules, but
not the globular protein.
1

Mass Transfer (ECH 143)


Problem Set 3

1. Middleman 2.7

Solution
The derivation of Eq. 2.1.21 from Eq. 2.1.20 was done in class. Assume the total molar
concentration C is constant, then

dx A
CD AB = C1dz ,
1− x A

where C1 is a constant, and equal to the negative of the molar flux. Integrating givess

CD AB ln (1− x A ) = C1z + C2 .

Now let
x A = x A,1 at z=0
and
x A = x A,2 at z=L .

Finding the constants yields

z
⎡ 1− x A ⎤ z ⎡ 1− x A,1 ⎤ ⎡ 1− x A,1 ⎤
L
ln ⎢ ⎥ = ln ⎢ ⎥ = ln ⎢ ⎥ .
⎣ 1− x A,1 ⎦ L ⎣ 1− x A,2 ⎦ ⎣ 1− x A,2 ⎦

Taking the exponent of both sides yields Eq. 2.1.21

z/L
⎛ 1− x A ⎞ ⎛ 1− x A,1 ⎞
⎜ 1− x ⎟ = ⎜ 1− x ⎟ .
⎝ A,1 ⎠ ⎝ A,2 ⎠

In the dilute limit where

x A << 1 ,

the starting point above simplifies to

dx A = C1dz
or
x A = C1z + C2 .

Imposing the boundary conditions to get the constants gives


2

x A = x A,1 + ( x A,2 − x A,1 )


z
.
L

The flux is then

dx A CD AB ( x A,1 − x A,2 )
N A,z = −CD AB = .
dz L

To get the flux from the exact solution, take the natural log of both sides

⎛ 1− x A ⎞ z ⎛ 1− x A,1 ⎞
ln ⎜ ⎟ = ln ⎜ ⎟ ,
⎝ 1− x A,1 ⎠ L ⎝ 1− x A,2 ⎠

differentiate,
⎛ 1− x A,1 ⎞ 1 dx A 1 ⎛ 1− x A,1 ⎞
−⎜ = ln
⎝ 1− x A ⎟⎠ 1− x A,1 dz L ⎜⎝ 1− x A,2 ⎟⎠
so

dx A
=−
(1− x ) ln ⎛ 1− x
A,1 A,1

⎜ 1− x ⎟ .
dz z=0 L ⎝ A,2 ⎠

The flux is (see Eq. 2.1.19)

−CD AB dx A CD AB ⎛ 1− x A,2 ⎞
N A,z z=0 = = ln ⎜ ⎟ .
1− x A,1 dz z=0 L ⎝ 1− x A,1 ⎠

If the mole fractions are 0.1 and 0, as described in the problem statement, then

⎛ 1− x A,2 ⎞ ⎛ 1⎞
ln ⎜ ⎟ = ln ⎜ ⎟ = 0.1054
⎝ 1− x A,1 ⎠ ⎝ .9 ⎠
whereas
x A,1 − x A,2 = 0.1 .

The results differ by about 5%, even at a volume fraction as low as 0.1. The approximate result
underestimates the flux.

2. a) Do Middleman 2.10. Assume the air outside the vacuum chamber is at 700K, the same
temperature as the gas inside the chamber.
3

b) Discuss the appropriateness of the quasi-steady assumption in this problem


c) Assess the possibility that gradients in the hydrogen concentration inside the vacuum
chamber should be considered (i.e., that the hydrogen concentration inside the chamber
varies with position)

Solution
We have diffusion across a thin, solid barrier. The glass wall acts as a constant-density, stagnant
matrix through which the hydrogen diffuses, so diffusion is described by

J = −Dg∇C W

where CW is the hydrogen concentration in the glass and

cm 2
Dg = 10 −8 .
s

Using the quasi-steady assumption as advised in the problem, and setting the velocity to zero
since the glass isn’t moving, the convective diffusion equation reduces to

d 2C W
=0
dz 2

Integrating twice gives the linear concentration profile (see Example 2.1.5)

C W = α + βz .

Let the outside edge of the glass be z=0, and the inside edge be z=L. Then

C W = bC1 at z=0
and
C W = bC at z=L ,

where C is the concentration inside the chamber. Then the profile is

b ( C − C1 )
C W (z, t) = bC1 + z
L

The flux of hydrogen through the wall and into the chamber is then

bD ( C1 − C )
JW = .
L

The outer concentration at z=0 is a constant, and equal to the concentration of hydrogen at 1atm
pressure and 700K. The concentration C at z=L changes in time, because the hydrogen
4

accumulates inside the chamber as diffusion takes place. Two assumptions are being used here:
it is being assumed that the concentration inside the chamber is approximately constant relative
to concentration changes in the wall, and that the concentration profile in the glass wall is quasi-
steady. Those assumptions are examined at the end of this solution.
The molar concentration of air at 700K is

N P 1.01x10 5 N / m 2 mol
C= = = = 17.41 3 .
V RT ( 8.314J / molK ) ( 700K ) m

The concentration of hydrogen in the air is therefore

⎛ 3.8x10 −4 ⎞ mol mol


C1 = ⎜ ⎟ 17.41 3 = 8.705x10 −6 3 .
⎝ 760 ⎠ m m

Inside the air pressure was reduced to 10-6 torr, so the initial interior concentration is

⎛ 10 −6 ⎞ mol mol
C2,0 = ⎜ ⎟ 8.705x10 −6 3 = 1.145x10 −14 3 ,
⎝ 760 ⎠ m m

but that value changes, hence the subscript 0.

To determine how the pressure changes, we need to do a macroscopic species balance that
equates the accumulation of hydrogen inside the chamber with the flux across the wall,

dC bD
V
dt
=
L
[C1 − C] A

where
πD 2
V= (1m )
4
and
A = πD (1m ) .

The diameter D=10cm. The balance can be rearranged to

dC bD A
= Adt
[C1 − C] L V
and solved to yield

C = C1 − ( C1 − C2,0 ) e −(bDA/LV)t .
5

The term in the exponent is

⎛ cm 2 ⎞
0.2 ⎜ 10 −8
bD A ⎝ s ⎟⎠ 4
= = 8x10 −10 s −1 .
L V 1cm D

For times of approximately 1 hour or less, the term in the exponent will be very small, so that the
Taylor expansion

e − x ≈ 1− x

can be used (but is not necessary). Then

C = C1 − C1 (1− (bDA / LV)t ) + C2,0 (1− (bDA / LV)t )

or

C = C2,0 + C1 (bDA / LV)t ,

where a term

C2,0 (bDA / LV)t

has been neglected. The growth in the hydrogen concentration over the period of interest is
approximately linear in time. This result can be converted to pressure by using the ideal gas law,

PH2 = RTC = RT ⎡⎣C2,0 + C1 (bDA / LV)t ⎤⎦

or

⎡ C ⎤
PH2 = RTC = RTC2,0 ⎢1+ 1 (bDA / LV)t ⎥ .
⎣ C2,0 ⎦

For a one-hour period

C1
(bDA / LV)t = 2,189.6 .
C2,0

The partial pressure of hydrogen will increase linearly, to become several thousand times greater
than its initial value, but will still be much smaller than the outside concentration of hydrogen
after 1 hour.
6

Assumptions
Now consider the two assumptions that were used. First, the time scale for the concentration
profile inside the glass wall to adjust in response to a change in the concentration at its inner
edge is

L2
τ1 ~ = 10 8 s .
D

It therefore takes much longer than 1 hour for the linear concentration profile to develop, if the
inner concentration is lowered suddenly. If the vacuum in the chamber is maintained
permanently, for years, but the vacuum pump must be turned off during an experiment, then the
quasi-steady assumption might be appropriate to assess the importance of “leakage.”

The concentration change at the inner boundary resulting from leakage of hydrogen through the
glass does not invalidate the use of the quasisteady assumption. Even at the end of one hour, the
hydrogen concentration inside the chamber is orders of magnitude smaller than that outside.
Solving the problem with an inner concentration of zero, and constant, would yield essentially
the same result, namely that the pressure change is

ΔP = RTC1 (bDA / LV)t .

Still, except for the special circumstances described above, the quasi-steady assumption does not
seem well-suited to this problem because the diffusivity in the glass is so low.

The other assumption was that the hydrogen inside the chamber was well-mixed, so that inner
concentration variations can be neglected. Here we can compare the resistance to mass transfer
in the wall with thickness L=1cm,

L
,
Dg

and compare it with the resistance in the gas inside the chamber, say

D
Dgas

where the D=10cm, because the diffusion of gas is radial (the ends are impermeable). Although
L is 10 times smaller than D, the diffusion coefficients differ by much more than that. A typical
diffusion coefficient for a gas at 1atm (see p. 19) is 0.1cm2/s, some 7 orders of magnitude greater
than what is given for hydrogen in the glass. That value would be even higher at lower
pressures. The resistance in the wall is therefore much greater than any internal resistance, and
neglecting concentration variations inside the vacuum chamber seems justified. (
7

3. Middleman 2.15

Solution
In the plot, the slope dL/dt appears to be very nearly constant. It appears that the volume change
is about 2.5 ml at 160 hours, so at that time the height change is

4
ΔL = 2.5cm 3 = 3.53cm
π(0.95cm)2

and the height of the vapor column changes from 15.6cm to 12.1cm over 160 hours. We showed
in class that

− ( L2 − L20 ) ( CL / CG )
D AB = .
2 ( t − t 0 ) ln (1− x A1 )

At 0.94atm and 19ºC, the molar concentration in the air is

N P ( 0.95atm ) (1.013x10 Pa / atm )


5
mol
CG = = = = 39.6 3 .
V RT ( 8.314J / molK )( 292K ) m

The mole fraction of hexane is

1.5x10 4
x A1 = = 0.156 .
( 0.95 )(1.013x10 5 )

The molecular weight of benzene, C6H14, is 86g/mol, so the liquid concentration is

g 10 6 cm 3 ⎛ 1mol ⎞ mol
CL = 0.66 ⎜ ⎟ = 7,674 3 .
cm 3
m ⎝ 86g ⎠
3
m

The diffusivity is therefore

− (15.6 2 − 12.12 ) ( 7674 / 39.6 ) 1 cm 2 cm 2


D AB = = .096 .
2 (160 ) ln (1− 0.156 ) 3600 s s

Table 11-2 of the book by Reid et al. (on the course web site) gives a value of

⎛ cm 2 ⎞
D ABP = 0.08 ⎜ atm ⎟
⎝ s ⎠
or, at 0.95atm,
8

cm 2
D AB = 0.084 .
s

(The temperature difference is insignificant.) This result is lower than ours by more than 10%.
Some of the difference might be caused by error in reading a value off the graph giving the
experimental data—using 2.4ml at 160 hours instead of 2.5ml improves the agreement slightly.

4. In an emergency at an oil refinery, a large cylindrical column 1m in diameter and 50m tall
may need to be filled with propane gas. The column is open to the atmosphere at the top,
where there is air at 1atm and 20ºC. Assuming the column is initially filled with pure
propane, and the flux to the atmosphere is purely by diffusion (i.e., neglect convection effects
at the outlet), compute the rate at which propane will be emitted. If the Bay Area Air
Pollution Control District considers propane emission of 1 pound per hour or 10 pounds per
day to be a violation, will a violation occur? Use 0.1cm2/s as the diffusivity of propane in air
at 20ºC.

Solution
In the absence of reaction, the species continuity equation takes the form

∂C A
= −∇ ⋅ N A ,
∂t

and a convenient form for the total flux for binary gas diffusion is

N A = −CD AB∇x A + x A ( N A + N B ) .

In this problem, the gas is at 1atm throughout the column, and the temperature is presumably
constant (we are not told otherwise), so the total molar concentration C should also be constant.
The flux of 1mol of propane out of the tower is therefore accompanied by a counter-flux of 1mol
of air into the tower—we have equimolal counterdiffusion, so

NA + NB = 0
and
N A = −D AB∇C A .

Substituting into the species continuity equation gives

∂C A ∂2 CA
= D AB ,
∂t ∂z 2

where it has been assumed that the diffusion is only in the z-direction (up, in this case). The
maximum emission will occur at the beginning, so if a violation happens, it will happen in the
first hour or first day. The diffusive perturbation distance for 1 day is
9

1/2
⎛ cm 2 ⎞
L ~ Dt = ⎜ 0.1 (8.64x10 4 s)⎟ = 93cm = .93m ,
⎝ s ⎠

so the column can be considered “infinite” in length. Essentially, then we can idealize the
situation as a semi-infinite medium for in a change occurs at one boundary, and the perturbation
diffuses inward. The solution will involve the error function, and is given by (as derived in
class)

Cair = C0 (1− erf(η)) ,

where
z
η=
4Dt

and D = 0.1cm2/s.

As shown in Problem Set 2, using the Leibnitz rule to differentiate the error function yields

e − x / 4 Dt
2
∂Cair
= −C0
∂z πDt

so that, at the top of the column (z=0),

DC0 D
N z,air = = C0 .
πDt πt

Integrating from t=0 to a later time T, and multiplying by the column area, we find for the total
emission

Dt
M = 2AC0 .
π

The column area is

A = 0.785m 2
and the molar concentration of air at the top is

P 1.013x10 5 Pa mol
C0 = = = 41.6 3 .
RT ( 8.314J / molK ) ( 293K ) m

After one hour the total moles emitted (which is the same as the number of moles of air that have
diffused in) is
10

⎛ mol ⎞ (0.1cm 2 / s)(3600s) ⎛ 1m ⎞


M = 2 ( 0.785m ) ⎜ 41.6 3 ⎟
2
⎜⎝ ⎟ = 7.0mol .
⎝ m ⎠ π 100cm ⎠

The molecular weight of propane is

g
M propane = 44
mol
so the mass of emitted propane is .308kg, or .68 lb (at 2.2lb/kg). There is no violation in the first
hour. After 24 hours, the mass emitted would be 4.9 times higher (the square root of 24), so
there still would not be a violation.
1

Mass Transfer (ECH 143)


Problem Set 4

1. Middleman 3.44
Comment: Neglect convection in the region between the wafers. In spite of the instructions,
proceed to solve the partial differential equation governing the concentration distribution
C(r,z) in the region 0≤z≤H and 0≤r≤RW. Express your solution as a Fourier-Bessel expansion
(i.e., an expansion in which the orthogonal functions are Bessel functions).

Solution
For dilute gas diffusion,

N = −CD∇x ≈ −D∇C ,

where x and C are the mole fraction and concentration of the (dilute) reacting species. We have
steady diffusion in a cylindrical geometry, with no homogeneous reaction, so the species
conservation equation simplifies to Laplace’s equation,

∇ 2C = 0 .

The reacting species is at a fixed concentration at r=RW and undergoes a first-order reaction at
the wafer surface. The solution is symmetric about z=0, and there is no cause for an angular
variation in concentration. Expanding the governing equation and using the geometric
simplifications gives
2
∂ C 1 ∂ ⎛ ∂C ⎞
+ ⎜r ⎟ = 0 ,
∂z 2 r ∂r ⎝ ∂r ⎠
with

∂C
= 0 at z=0
∂z

∂C
−D = kC at z=H
∂z

C = C0 at r=R W
and
∂C
= 0 at r=0 .
∂r

Here k is the first-order rate constant for the heterogeneous reaction at the surface.

We can solve this equation by separation of variables. It appears that the eigenfunctions could
be either functions of r or z, since both sets of boundary conditions are, or can be rendered,
2

homogeneous. I have chosen to expand with eigenfunctions that are functions of r, so that the
flux at the wafer surface, which appears in a boundary condition, will not be subject to the
“Gibbs phenomenon” (Fourier series sometimes exhibit fluctuations at the edges of the domain.)
This choice may not be necessary, but seems prudent. Suppose we introduce the concentration
difference

 C −C
C= 0 ,
C0

so that

⎛ R W ⎞ ∂ Ĉ 1 ∂ ⎛ ∂Ĉ ⎞
2 2

⎜⎝ ⎟ + r̂ =0 ,
H ⎠ ∂ẑ 2 r̂ ∂r̂ ⎜⎝ ∂r̂ ⎟⎠
with

∂Ĉ
= 0 at ẑ=0
∂ẑ

∂Ĉ
∂ẑ
( )
= Da Ĉ − 1 at ẑ=1

Ĉ = 0 at r̂=1
and
∂Ĉ
= 0 at r̂=0 (or concentration is finite at r=0),
∂r̂

where z and r are normalized by H and RW and the Damkohler number Da appears.

To separate variables, let

Ĉ ( r̂, ẑ ) = f(r̂)g(ẑ)

and substitute to find

1 d ⎛ df ⎞
⎜⎝ r̂ ⎟⎠ = −λ f
2

r̂ dr̂ dr̂
and
2
dg
β 2
− λ 2g = 0 ,
dẑ 2
with
RW
β= .
H
3

The solutions for f are zero-order Bessel functions, and for a finite solution at r=0 only zero-
order Bessel functions of the first kind (i.e., Jn(λr)) need be included, since the zero-order Bessel
functions of the second kind (usually denoted by Y) become infinite at r=0. So

fn ( r ) = A n J 0 ( λ n r ) .

The boundary condition at the outer edge defines the eigenvalues as the series of solutions to

J0 (λn ) = 0 .

Solutions to the equation for g(z) may be written either as exponential functions or as hyperbolic
functions (which are combinations of exponential functions),

⎛λ ⎞ ⎛λ ⎞
g(ẑ) = Bsinh ⎜ ẑ⎟ + Ccosh ⎜ ẑ⎟ .
⎝β ⎠ ⎝β ⎠

The derivative must be zero at z=0, so

B= 0 .

The solution for the concentration is then a sum of product solutions,

∞ ⎛λ ⎞
Ĉ ( r̂, ẑ ) = ∑ A n sinh ⎜ n ẑ⎟ J 0 ( λ n r̂ ) .
n=1 ⎝ β ⎠

The remaining coefficients are found from the flux condition on the reacting surface, which
requires that

∞ ⎡λ ⎛λ ⎞ ⎛ λ ⎞⎤
∑ A n ⎢ n cosh ⎜ n ⎟ − Da sinh ⎜ n ⎟ ⎥ J 0 ( λ n r̂ ) = −Da .
i=1
⎣β ⎝ β⎠ ⎝ β ⎠⎦

Using the orthogonality property of the Bessel function (with the weighting function equal to r),
the coefficients are found to be

Da ∫ r̂J 0 ( λ n r̂ ) dr̂
An = − 0
,
⎡ λn ⎛ λn ⎞ ⎛ λ n ⎞ ⎤ ∫ r̂J 2 ( λ r̂ ) dr̂
1

⎢ β cosh ⎜⎝ β ⎟⎠ − Da sinh ⎜⎝ β ⎟⎠ ⎥ 0 0 n
⎣ ⎦

where the eigenvalues are the roots of the zero-order Bessel function.
4

2. Middleman 3.47
Explanation: As shown in Fig. P3.47, there is an inert, encapsulating layer that separates a
liquid solution of reactant from the underlying reactive layer. The reactant therefore must
diffuse through the encapsulating layer before it reacts. The diffusivity is the same in both
layers, and the concentration is continuous across the interface between the encapsulating and
reactive layers (that partition coefficient is one). Use a partition coefficient α (=concentration
in encapsulating layer divided by the outside concentration) to describe the concentration
jump at the interface between the solution and the encapsulating layer. Non-dimensionalize
position z by H, and let β=h/H.

Solution
We have diffusion in two media with constant densities, with the velocity zero in both, and
the system is at steady state. We can therefore write

d 2C
= 0 for H ≤ z ≤ H + h
dz 2

and

d 2C
D − kC = 0 for 0 ≤ z ≤ H ,
dz 2

where the negative sign indicates that the reactant is being consumed. Because the partition
coefficient between the two solid phases is one, and the diffusion coefficients are the same,
the concentration is continuous and the slope is also continuous (in general, the flux of solute
must be continuous, so if the diffusion coefficients change, the slope must change also.)

Nondimensionalizing,

d 2Ĉ
= 0 for 1 ≤ ẑ ≤ 1+ β
dẑ 2
and
d 2Ĉ
− bĈ = 0 for 0 ≤ ẑ ≤ 1 ,
dẑ 2

where the Thiele modulus is

kH 2
b=
D
and
C
Ĉ = .
C A0
5

The solutions to the two equations are

ĈE = B1 + A1ẑ for 1 ≤ ẑ ≤ 1+β


and
ĈR = C1 sinh ( bẑ) + D cosh ( bẑ)
1 for 0 ≤ ẑ ≤ 1 ,

where E and R have been added for “encapsulating” and “reactive.”

The following conditions must be met:

ĈE = α at ẑ=1+β (1)

ĈE = ĈR at ẑ=1 (2)

dĈE dĈR
= at ẑ=1 (3)
dẑ dẑ

and
dĈR
= 0 at ẑ=0 . (4)
dẑ

Conditions (2) and (3) are referred to as “matching conditions,” and are simplified somewhat
by the fact that the partition coefficient is one and the diffusivities are the same, as already
mentioned. Condition (4) reflects the fact that the flux into the bottom wall is zero.

Condition (4) requires

C1 = 0
and we are left with

α = B1 + A1 + βA1

A1 + B1 = D1 cosh ( b) .

A1 = D1 b sinh ( b) .

Using the first and last relations in the middle one gives

α b
A1 =
(β ( b ))
.
b + coth
6

Then
α
D1 =
sinh ( b ) (β b + coth ( b ))
and

B1 = D1 cosh ( b)− A 1 .
1

Mass Transfer (ECH 143)


Problem Set 5

1. Use separation of variables to derive Eq. 4.2.36 as the solution to Eqs. 4.2.32-4.2.35 in the
text by Middleman.

Solution
We have
∂Y ∂ 2 Y
=
∂X ∂Z 2

with the boundary conditions

Y = 0 at Z=0 and Z=1

and the initial condition

Y = 1− Z at X=0 .

The problem has been rearranged to this form so that the boundary conditions are homogeneous,
and separation of variables can be used to get the solution (to determine whether boundary
conditions or a PDE are homogeneous, set the dependent variable (Y in this case) to zero, and
see if the equation reduces to 0=0. If it does, it is homogeneous.)

Assume a product solution of the form

Y = g(X)f(Z) ,

substitutie, and separate variables:

1 dg 1 d 2 f
= 2
= −λ 2 .
g dX f dZ

Here I have used that, if a function of X is equal to a function of Z everywhere, they must both
equal a constant, which I have chosen to be -λ2. The solution for f(Z) is

f(Z) = Asin ( λZ ) + Bcos ( λZ ) .

The requirement that Y=0 at Z=0 means that f=0 at Z=0, which can only be true if B=0. The
requirement that f=0 at Z=1 then can be satisfied if

λ = nπ n=1, 2, 3 ... .

Notice that if the constant written as -λ2 were positive or zero, only the trivial solution f=0 would
satisfy the equation and boundary conditions. The solution for g(X) is exponential,
2

g(X) = e − λ X .
2

Since X is unbounded, the presence of only negative eigenvalues also guarantees the solution
will remain finite at long times.

The most general solution is the sum of all non-trivial product solutions, so we write


Y = ∑ A n e − n π X sin ( nπZ ) .
2 2

n=1

This is Eq. 4.2.36 in the text. Notice that if X is O(1), only the first term in the sum is important.
The exponential term becomes small very quickly. This observation can reduce significantly the
algebraic complexity of working with these series solutions, as long as it is not the short-time
behavior that is of interest.

2. Middleman 4.36

Solution
The fluxes into and out of the membrane are only equal at steady state. During the transient
period preceding steady state, there is an imbalance (i.e., a net flux into the membrane) because
initially the concentration inside the membrane is zero. The solution is

C 2 ∞ sin nπZ − n2π 2 X


= 1− Z − ∑ e ,
Hp 0 π n=1 n

so that for X>>1, the concentration profile becomes linear. For a linear concentration profile the
flux is constant, so the flux into the membrane at Z=0 must exactly equal the flux out at Z=1.

The fluxes at the two surfaces of the membrane are

HDp 0 ∂ HDp 0 HDp 0 ∂Y


N z Z=0 = − (1− Z − Y) = +
B ∂Z Z=0 B B ∂Z Z=0
and
HDp 0 ∂ HDp 0 HDp 0 ∂Y
N z Z=1 = − (1− Z − Y) = +
B ∂Z Z=1 B B ∂Z Z=1
and
2 ∞ sin nπZ − n2π 2 X
Y= ∑ e .
π n=1 n

The leading terms in these two results are identical, and come from the steady state solution.
The derivatives are given by
3

∂Y ∞
= 2 ∑ e− n π X
2 2

∂Z Z=0 n=1

and
∂Y ∞
= 2 ∑ cos nπ e − n π X .
2 2

∂Z Z=1 n=1

To find the amount that has crossed the interfaces from t=0 to “long times,” integrate over time,
from 0 to infinity, recognizing that X is dimensionless time:

∞ ∂Y B2 ∞ ∂Y B2 ∞ 1 B2
∫ dt = ∫ dX = 2 ∑ = .
0 ∂Z Z=0 D 0 ∂Z Z=0 Dπ 2 n=1 n 2 3D

Doing the same for the second expression,

∞ ∂Y B2 ∞ ∂Y 2B2 ∞ (−1)n B2
∫ dt = ∫ dX = ∑ = − .
0 ∂Z Z=1 D 0 ∂Z Z=1 Dπ 2 n=1 n 2 6D

Here we have used the results that

∞ 1 π2
∑ 2
=
n=1 n 6
and
∞ ( −1)n =
−π 2
∑ .
n=1 n2 12

To obtain the total amount of solute to have entered the membrane, we need

ez ⋅ N

at Z=0, and
−e z ⋅ N

at Z=1, because it is the net flux into the membrane that is of interest. Therefore we have

HDp 0 ⎛ B2 B2 ⎞ BHp 0
+ = .
B ⎜⎝ 3D 6D ⎟⎠ 2

For a linear, steady state concentration profile that is Hp0 at z=0 and 0 at z=B, the net amount of
solute is half the concentration at z=0, multiplied by the membrane width B (multiplied also by
the area A, since the membrane volume is BA). The net flux of solute into the membrane at long
times is the average concentration, which is half the maximum for a linear concentration profile,
multiplied by the membrane volume.
4

2.0

1.5
Nz

1.0

0.5

0.0
0.0 0.2 0.4 0.6 0.8 1.0
X

A plot of the fluxes (integrated over time) across the surfaces at Z=0 and Z=1 is shown above.
The imbalance disappears by about X=0.5, so requiring X>>1 is a conservative constraint.

3. Middleman 4.41

Solution
The figure caption states that the concentration is plotted in units proportional to the actual
concentration, so we rewrite the Gaussian solution as

⎛ x2 ⎞
αM ⎜⎝ − 4 Dt ⎟⎠
αC ( x, t ) = e .
πDt

Taking the natural log of this expression, we have

1 x2
lnαC = lnαM − ln πDt − .
2 4Dt

Plotting the log of concentration vs. the square of x should therefore yield a line with negative
slope, as in the figure. The slope of the data is approximately


1
=
( −24.6 + 21.9 ) mm −2
4Dt 79 − 5

so that
5

1 −4t ( −24.6 + 21.9 )


= mm −2 .
D 79 − 5

Then after 8 hours,


79 − 5 mm 2
D=
4(8)(3600) ( 2.7 ) s

or
mm 2
−4
D = 2.4x10 .
s

4. Consider the decaffeination of coffee beans, as described in Example 4.3.1 of the text by
Middleman. Using Matlab, calculate the concentration of caffeine in a spherical coffee bean, at
each position r in the domain 0≤r≤R, for times t in the range 0≤t≤2(R2/D). Do the calculation for
a finite Biot number Bi, where
kR
Bi = ,
D

with k being the mass transfer coefficient and D the diffusion coefficient of caffeine in the bean.
The caffeine concentration initially is C0, and the liquid into which the caffeine is being extracted
contains negligible caffeine. This problem can be solved with the package pdepe, as described in
a posted handout.

In dimensionless form, you are being asked to solve the problem

∂Ĉ 1 ∂ ⎛ 2 ∂Ĉ ⎞
= r̂
∂ t̂ r̂ 2 ∂r̂ ⎜⎝ ∂r̂ ⎟⎠

with the initial and boundary conditions

Ĉ = 1 at t̂=0

∂Ĉ
= 0 at r̂=0
∂r̂
∂Ĉ
= −BiĈ at r̂=1 ,
∂r̂
where
C
Ĉ = .
C0
6

As your solution, provide plots of

Ĉ vs. r̂ at t̂ = 0.02, 0.2 and 0.6

for Bi = 0.1, Bi=1 and Bi=10. Each plot should have 3 curves, corresponding to the 3 different
dimensionless times, and there should be one plot for each Biot number. At which Biot number
would the approximate boundary condition

Ĉ = 0 at r̂=1
be most accurate? At which Biot number would it be most appropriate to approximate the
concentration everywhere inside the bean by a single, spatially averaged value?

Solution
The plots calculated by Matlab’s pdepe, with dimensionless time increments of 0.02 and
dimensionless position increments of 0.01, are given below.
7
8

Steep concentration gradients tend to be located where the resistance to mass transfer is highest.
Setting the surface concentration equal to the bulk concentration, or

Ĉ = 0 at r̂=1 ,
is most accurate at high Biot numbers, where the resistance is dominated by the resistance inside
the coffee bean, and resistance outside is relatively small. By contrast, at the lowest Biot
number, Bi=0.1, concentration changes inside the bean are relatively small, and it would be
reasonable to replace C(r,t) by a radially averaged concentration <C(t)>. The latter would be
comparable to the “fin approximation” in heat transfer, except that it is hard to envision a sphere
being a fin!
1

Mass Transfer (ECH 143)


Problem Set 6

1. Suppose the concentration of solute C(r) in a liquid flowing through a cylindrical tube varies
according to
2
⎛ r⎞
C(r) = C0 ⎜ ⎟ .
⎝ R⎠

The flow of the liquid is laminar. Calculate the mixing-cup average concentration CCM and the
area-average concentration <C>.
Solution
The mixing-cup average is like an area-average, but each position is weighted by its velocity, so
that the result is what one would find if the fluid exiting a tube were collected in a mixing cup.
In Poiseuille flow through a tube, the velocity profile is

⎡ ⎛ r ⎞2⎤
v ( r ) = 2 v ⎢1− ⎜ ⎟ ⎥ ,
⎣ ⎝ R⎠ ⎦

where <v> is the average axial velocity. The mixing-cup average is given by (see text eq.
5.2.16)

R ⎡ ⎛ r ⎞2⎤
C0 ∫ 2 v ⎢1− ⎜ ⎟ ⎥ C(r) ( 2πr ) dr
0 ⎣ ⎝ R⎠ ⎦ 4C0 R ⎡ ⎛ r ⎞ ⎤
2

CCM = = 2 ∫ ⎢1− ⎜ ⎟ ⎥ C(r)r dr .


v πR 2 R 0 ⎣ ⎝ R⎠ ⎦

Substituting the expression given, we need

4C0 R ⎡ ⎛ r ⎞ ⎤ ⎛ r ⎞
2 2

CCM = ∫ ⎢ ⎜ ⎟ ⎥ ⎜ ⎟ r dr .
1−
R2 0 ⎣ ⎝ R ⎠ ⎦⎝ R ⎠
The result is
C0
CCM = .
3

The area-average concentration is

R
C0 ∫ C(r) ( 2πr ) dr
2C0 R C
C = 0
= 2 ∫
C(r)r dr = 0 .
πR 2
R 0 2
2

2. Consider the problem posed in Middleman 5.18. From Eqs. (5.1.18) to (5.1.21), it seems
clear that to allow for a penetration distance comparable to the film thickness we must solve the
problem

∂u ∂2 u
vx ( y ) = D 2
∂x ∂y

with

u = 1 at x=0
u = 0 at y=0

and

∂u
= 0 at y=δ .
∂y

Here the dimensioness concentration u has been re-defined as

C
u= ,
C0

and the ambient concentration has been set to zero. The velocity profile is given in Eq. (5.1.1),
and δ is the film thickness.

a) Nondimensionalize the equation and boundary conditions. Your governing equation should
include a Peclet number defined in terms of the average velocity, the film thickness δ and the
diffusivity D. Nondimensionalize position in both directions by δ.

Solution
In dimensionless form, we have (using that the average velocity is two-thirds of the maximum)

3 ∂u ∂ 2 u
Pe ⎡⎣1− ŷ 2 ⎤⎦ =
2 ∂x̂ ∂ŷ 2

with

u = 1 at x̂=0
u = 0 at ŷ=0

and

∂u
= 0 at ŷ=1 .
∂ŷ

All positions have been made dimensionless with δ, the film thickness, and the Peclet number is
3

v δ ρgδ 3
Pe = = .
D 3µD

b) Estimate the length down the film at which you would anticipate the reduction in the
concentration of the volatile solute would reach the wall.

Solution
The diffusive time is

L δ2
τ~ ~ ,
v D

where the distance L down the film divided by the average velocity has been equated to the
exposure time of region of fluid to the ambient air. Then the length would be

v δ 2 (0.82cm / s)(0.005cm)2
L~ = = 2cm .
D 10 −5 cm 2 / s

Here the average velocity was calculated as two-thirds of the maximum velocity, which is found
in part (c) below to be 1.225cm/s. This length corresponds to a dimensionless distance of about
40 (i.e., the dimensional distance is 40δ).

c) Solve this problem numerically using the Matlab solver pdepe. Do the calculation for a
0.005cm layer of liquid water flowing vertically down, as shown in Fig. 5.1.1 of Middleman. As
a diffusion coefficient use

cm 2
−5
D = 10 .
s
Compute and plot the concentration profile at distance x=0.05cm, 0.5cm, 1.0cm and 2.0cm down
the film.
Solution
From Eq. (5.1.1), the velocity profile is

ρgδ 2 ⎡ ⎛ y ⎞ ⎤
2

vx ( y ) = ⎢1− ⎜ ⎟ ⎥ .
2µ ⎣ ⎝ δ ⎠ ⎦

For a 0.5mm layer of water, which has a viscosity of 0.01g/cm.s, we have

ρgδ 2 1 ⎛ 1g ⎞ ⎛ cm ⎞ 2 ⎛ 1cm.s ⎞ cm
= ⎜ 3⎟⎜
980 2 ⎟ ( 0.005cm ) ⎜ ⎟ = 1.225 .
2µ ⎝
2 cm ⎠ ⎝ s ⎠ ⎝ 0.01g ⎠ s

The Peclet number is


4

Pe = 408.3 .

To get to a distance of 0.5cm, 2.0cm and 4.0cm we need to get to

x̂ = 10, 100, 200 and 400 .

The plotted concentration profiles are shown below. The concentration perturbation does reach
the wall at dimensionless distances between 10 and 100, as predicted in part (b). Recall that y=0
corresponds to the liquid/air interface, where the concentration is zero. The slopes should be
zero at the impermeable boundary at y=δ, and they are.

d) Compute the mixing-cup concentration at x=0.05cm, 0.5cm, 1.0cm and 2.0cm down the film.
Solution
In this case the mixing cup temperature is

1
δU max ∫ ⎡⎣1− ŷ 2 ⎤⎦ Ĉ ( x̂, ŷ ) dŷ
31
ĈCM = = ∫ ⎡⎣1− ŷ ⎤⎦ Ĉ ( x̂, ŷ ) dŷ .
0 2

(2 / 3)δU max 20

The integral must be evaluated numerically, using the Matlab solution. For my solution, I used
100 nodes across the film, and 4000 nodes in the direction of flow, to get to a dimensional
5

distance of 400δ (or a dimensional distance of 400) down the film. Then, after getting the
solution, I calculated the integral above numerically by using the commands

xpos=1-x.*x;
trapz(xpos.*u(100,:));
trapz(xpos.*u(1000,:));
trapz(xpos.*u(2000,:));
trapz(xpos.*u(4000,:));

These results had to be multiplied by 0.010101 (the node spacing in the y-direction), and then by
1.5. The results were
x̂ = 10, ĈCM = 0.785
x̂ = 100, ĈCM = 0.343
x̂ = 200, ĈCM = 0.149
x̂ = 400, ĈCM = 0.0279
Observe how well the scaling estimate from part (b) worked. If one calculated the characteristic
length 40δ, and then designed the evaporator to be 400δ, it would remove 97% of the volatile
solute.

3. A hollow-fiber dialyzer is a device for removing or inserting a target solute or solutes into a
stream of liquid. For example, one application is the removal of urea from blood in
hemodialysis, a process used to treat patients with kidney failure. Consider a device that is to be
composed of 100 cylindrical fibers with radius R=0.5cm, the walls of which are semi-permeable
membranes (i.e., the walls are permeable to urea, but not to blood proteins). Outside the tubes an
external solution, the dialysate, is located. In the bulk of the dialysate, the urea concentration is a
constant Cd.
The flux of urea through the membrane is described by

N r (R,z) = k m ⎡⎣C ( R,z ) − Cd,W ⎤⎦ ,

where C(R,z) is the concentration of urea inside the tube, at the surface of the tube wall, and Cd,W
is the urea concentration in the dialysate at the outer edge of the tube wall. Significant
resistances to mass transfer are also expected in boundary layers inside and outside the tubes.
Inside the tube the flux from the bulk to the tube wall is described by

N r (R,z) = k C (z) ⎡⎣Cb (z) − C ( R,z ) ⎤⎦ ,

where Cb(z) is the mixing-cup concentration. The flux from the outer edge of the tube wall to the
bulk dialysate is described by

N r (R,z) = k d ⎡⎣Cd,W − Cd ⎤⎦ ,

and at z=0 the urea concentration in the blood is constant and equal to C0.
6

Assume the number of hollow fibers N is 100, and the dialysate flow rate is such that Cd=0. The
Sherwood number for the tube-side mass transfer coefficient is given by the Sieder-Tate
correlation,
−1/3
2k R ⎛ L ⎞
Sh = C = 1.86 ⎜ Pe1/3 ,
D ⎝ 2R ⎟⎠

where the Peclet number is

2 vz R
Pe = ,
D
D being the diffusion coefficient of urea (approximately 1.35x10-5cm2/s, as measured by Gosting
and Akeley (1952). The membrane mass transfer coefficient has been measured and is given by
the constant

cm
k m = 2x10 −5 .
s
The outside, or “shell side” mass transfer coefficient is also constant, and is given by

cm
k d = 10 −4 .
s
a) Derive an expression for the average overall or “effective” mass transfer coefficient. Your
answer will include L, the unknown length of the hollow fibers.
b) If the device is required to reduce the concentration of urea by a factor of 10, at a flow rate of
10ml/min, what total length of hollow fibers is required?
Solution
a) The membrane coefficient and outside coefficient are both constant. The inside coefficient is
−1/3
D ⎛ L ⎞
k C = 1.86 ⎜ ⎟ Pe1/3 ,
2R ⎝ 2R ⎠

so the overall coefficient is


−1
⎡1 1 R ⎛ L ⎞ ⎤
1/3

=⎢ + + Pe −1/3 ⎜
⎝ 2R ⎟⎠ ⎥⎦
k eff .
⎣ k d k m 0.93D

b) In lecture we showed that


7

Ci,b (z) − Ci,d


= e −2k eff z/ v R ,
Ci,b0 − Ci,d

and we are told that the dialysate concentration is zero. Then

Ci,b (L)
= 0.1 = e −2k eff L/ v R
Ci,b0

or

2k eff L
2.3 = .
v R

The effective overall mass transfer coefficient is a complicated function of L, so this equation
must be solved numerically. This solution can be achieved by implementing Newton’s method,
as done in ECM 6, or by successive substitution. For the latter, we could use

2.3 v R
k eff =
2L
and
−1
⎡1 1 R ⎛ L ⎞ ⎤
1/3

=⎢ + + Pe −1/3 ⎜
⎝ 2R ⎟⎠ ⎥⎦
k eff .
⎣ k d k m 0.93D

We equate these two expressions, then guess a value of L repeatedly and continue iterating until
convergence is achieved (or something else happens!). Equating the expressions,

⎡1 1 R −1/3 ⎛ L ⎞
1/3
⎤ 2L
⎢ + + Pe ⎜⎝ ⎟⎠ ⎥ =
⎣ k d k m 0.93D 2R ⎦ 2.3 v R

or

⎡1 1 R −1/3 ⎛ L ⎞
1/3

L = 1.15 v R ⎢ + +
⎝ 2R ⎠ ⎥⎦
Pe ⎜ ⎟ .
⎣ k d k m 0.93D

To get an initial guess, suppose we neglect the resistance inside the tube. Then

⎡1 1 ⎤
L = 1.15 v R ⎢ + ⎥
⎣ kd km ⎦
with
cm
k m = 2x10 −5 .
s
8

and
cm
k d = 10 −4 .
s
The initial guess is then 72.6cm. There are 100 fibers with radius R=0.5cm, and the flow rate is
10ml/min, so the average velocity in each fiber is

⎛ 10cm 3 ⎞ ⎛ 1 ⎞ cm
v =⎜ ⎟ ⎜ ⎟ = 0.0021
⎝ 60s ⎠ ⎝ 100π ( 0.5cm ) ⎠
2
s

The Peclet number is

v (2R) ( 0.0021cm / s ) (1cm )


Pe = = = 148 .
D 1.35x10 −5 cm 2 / s
The equation we are trying to solve is then

⎛ −3 cm ⎞
2
⎡ 4 s 4 s −1/3 ⎛ L ⎞
1/3

L = ⎜ 1.21x10 ⎢ 6x10 + 4x10 (148 ) ⎝ 2R ⎠ ⎥⎦
⎜ ⎟
⎝ s ⎟⎠ ⎣ cm cm

or

⎡ ⎛ L ⎞ ⎤
1/3

L = ⎢ 72.6 + 9.15 ⎜
⎝ 1cm ⎟⎠ ⎥⎦
cm .

Now iterate starting with 72.6cm, recalculating L repeatedly:


110.8cm, 116.5cm, 117.3cm, 117.4cm.
The solution to the equation is L=117.4cm.
Note: The correlation used for the tube-side mass transfer coefficient is valid in the “entry
region,” or for tube lengths less than PeR. In this case, the Peclet number is 148, so the
Sherwood number
−1/3
2k R ⎛ L ⎞
Sh = C = 1.86 ⎜ Pe1/3
D ⎝ 2R ⎟⎠

at the end of the tube is 2.0. In reality, at lengths comparable to PeL, the “entry region” ends. In
the “fully developed” region, the Sherwood number becomes a constant equal to approximately
4. Our estimate is therefore somewhat longer than is theoretically required, but better to be too
long than too short!
1

Mass Transfer (ECH 143)


Problem Set 7

1. Middleman 7.13
We have air and ammonia entering a scrubber at a rate of 450ft3/min (68ºF, 1atm), with the mole
fraction of ammonia equal to 0.0009 at the entrance and 0.00036 at the ext. We are also given
y=1.25x as the equilibrium relationship. We are asked to find the tower height. Schematically,
we have a situation as shown below.

We require the number of transfer units NG and the height of a transfer unit HG, so that we can
find the tower height ZT from

ZT = NGHG .

To obtain the number of transfer units, we use

y1 − y 2
NG =
Δy ln

where the log-mean driving force is

Δy ln =
(y − y ) − (y − y )
*
1
*
2
.
⎡ (y − y ) ⎤
*

ln ⎢ ⎥
1

⎢⎣ ( y − y ) ⎥⎦
*
2
2

We will take the subscript “1” to indicate the bottom, and “2” to indicate the top of the tower.
Then first we must do a material balance to determine the mole fraction of ammonia in the water
leaving the tower, so that we can find y*, the mole fraction in the gas that would be in
equilibrium with the bulk mole fraction in the water. The number of moles of ammonia in the
water must equal the number that leaves the gas stream. The molar flow rate in the gas stream is

( 450ft 3
/ min ) (1atm )
= 1.168lbmol / min .
( 0.730ft atm / lbmol.R )( 68 + 459.67 ) R
3

Entering the tower we have

⎛ lbmol ⎞ lbmol
0.009 ⎜ 1.168 ⎟ = 0.0105
⎝ min ⎠ min

and leaving in the gas stream we have

⎛ lbmol ⎞ lbmol
0.00036 ⎜ 1.168 ⎟ = 4.2x10 −4
⎝ min ⎠ min

so a total of

lbmol
0.01
min
are transferred to the water stream. In the water stream the molar flow rate is

lb m ⎛ lbmol ⎞ lbmol
29 ⎜ ⎟ = 1.61 .
min ⎝ 18lb m ⎠ min

The mole fraction of ammonia entering in the water is zero, and that leaving is 0.0062.
The driving forces are therefore

(y − y ) *
1
= 0.009 − 1.25 ( 0.0062 ) = 0.00125

and

(y − y ) *
2
= 0.00036 − 1.25 ( 0 ) = 0.00036 .

Then
3

Δy ln =
(y − y ) − (y − y )
*
1
*
2
=
0.00125 − 0.00036 8.9x10 −4
= = 7.15x10 −4 .
⎡ (y − y ) ⎤
*
⎛ 0.00125 ⎞ 1.245
ln ⎜
ln ⎢ ⎥ ⎝ 0.00036 ⎟⎠
1

⎢⎣ ( y − y ) ⎥⎦
*
2

Then

y1 − y 2 0.009 − 0.00036
NG = = = 12.0 .
Δy ln 7.15x10 −4

The height of a transfer unit is

G 1.168lbmol / min
HG = =
K y a v ( 4.5lbmol / min.ft 3 )

where G is the molar flux, or

1.168lbmol / min lbmol


G= = 5.95
( π / 4 )( 0.5ft ) 2
min.ft 2

then

G 5.95lbmol / min.ft 2
HG = = = 1.32ft .
K y a v ( 4.5lbmol / min.ft 3 )

The required tower height is

Z T = (12 ) (1.32ft ) = 15.8ft .


4

2. Middleman 7.14
The situation is depicted below.

a) To find the required height, we use

ZT = NGHG .

To obtain the number of transfer units, we use

y1 − y 2
NG =
Δy ln

where the log-mean driving force is

Δy ln =
(y − y ) − (y − y )
*
1
*
2
.
⎡ (y − y ) ⎤
*

ln ⎢ ⎥
1

⎢⎣ ( y − y ) ⎥⎦
*
2

The moles of carbon dioxide lost by the gas stream are given by

( 2.5gmol / s )( 0.013 − 0.0004 ) = 0.0315gmol / s .


The mole fraction of the exiting liquid stream is therefore
5

0.0315gmol / s
x1 = = 0.063 .
0.5gmol / s

We are told that gas at 1.25% is in equilibrium with the liquid at 7.5%, so it must be that
y = 0.1667x

describes the equilibrium relationship between gas and liquid. Then

y1* = 0.1667 ( 0.063) = 0.0105 .

Assuming the entering liquid has no carbon dioxide,

y*2 = 0.1667 ( 0 ) = 0.0 .

Then the log-mean driving force is

Δy ln =
( y − y ) − ( y − y ) = ( 0.013 − 0.0105 ) − ( 0.0004 − 0 ) = 0.0021 = 0.00115
*
1
*
2
⎡ (y − y ) ⎤
*
⎛ 0.013 − 0.0105 ⎞ 1.833
ln ⎢ ⎥ ln ⎜ ⎟
⎝ ⎠
1

⎢⎣ ( y − y ) ⎥⎦
* 0.0004
2

so that

y1 − y 2 0.013 − 0.0004
NG = = = 11.0 .
Δy ln 0.00115

The height of a transfer unit is


3
G 2.5gmol / s ⎛ 1m ⎞
HG = = ⎜ ⎟ = 2.78m .
K y a v (10 mol / s.cm .atm ) ( 0.9m ) (1atm ) ⎝ 100cm ⎠
−6 3 2

The tower height must therefore be 30.6m.


b) The minimum liquid flow rate is found by identifying the minimum value of the ratio (L/G).
This flow rate is that for which the liquid leaving the column is in equilibrium with the entering
gas (a condition which would require a column infinitely high). For dilute conditions, the
difference between the solvent molar fluxes and total molar fluxes is neglected, so (see text eq.
(7.4.4)

⎛ L⎞ y1 − y 2
⎜⎝ ⎟⎠ = * .
G min x ( y1 ) − x 2

The liquid mole fraction in equilibrium with the entering gas is

y1 0.013
x* ( y1 ) = = = 0.078 .
0.1667 0.1667
6

Then

⎛ L⎞ y1 − y 2 0.013 − 0.0004
⎜⎝ ⎟⎠ = * = = 0.162 .
G min x ( y1 ) − x 2 0.078 − 0

The molar flux of the gas is

2.5gmol / s gmol
G= 2
= 2.78 .
0.9m s.m 2
This gives the minimum liquid flow rate per unit area as

L min = ( 0.162 ) ( 2.78gmol / s.m 2 ) = 0.45


gmol
,
s.m 2
or the minimum flow rate as

2 (
0.9m 2 ) = 0.405
gmol gmol
AL min = 0.45
s.m s
This value is approximately 80% of the specified liquid flow rate.

3. Consider transfer of species A across a liquid-gas interface. On each side of the interface,
there is a film with thickness 0.01cm. Mass transfer coefficients can therefore be defined
according to

NA
kL =
c i,L − c A,L

and

NA
kG = ,
c A,G − c i,G

where

k L = (100cm −1 ) D A,L

and

k G = (100cm −1 ) D A,G .

Here the subscript “i” indicates a concentration right at the interface, and the other
concentrations are bulk concentrations at the edge of the film.
In designing absorbers, an overall coefficient KL is defined by
7

N A = K L ( c*A,L − c A,L )

where the superscript * indicates a liquid concentration in equilibrium with the bulk gas
concentration. Derive a value for KL (in cm/s) if

cm 2
D A,L = 10 −5 ,
s

cm 2
D A,G = 10 −1 ,
s
and Henry’s law describes the liquid-gas equilibrium via

p A = Hx A ,

with xA the mole fraction of A in the liquid, pA the partial pressure of A in the gas, and

H = 10 3 atm .
The total gas pressure is 1atm, the temperature is 20ºC, and the total molar concentration of the
liquid is 55.6M.
Solution
In terms of bulk concentrations, the flux of A across the interface is given by

1 ( c A,L − c A,L ) ( c A,L − c i,L + c i,L − c A,L ) 1 ( c A,L − c i,L )


* * *

= = = + .
KL NA NA kL NA

The concentration cA* is the liquid molar concentration in equilibrium with the bulk gas, while
cA,I is the liquid concentration at the interface, which is in equilibrium with the gas at the
interface.
Both concentrations are determined from Henry’s law, but the form must be modified. Dividing
Henry’s law by the total pressure, we get a relation between the mole fractions,

⎛ H ⎞
x A,G = ⎜
⎝ 1atm ⎟⎠ A,L
x .

However, what we really need is a relation between the molar concentrations at equilibrium. Let
cL and cG be the total molar concentrations in the liquid and gas, respectively. Then

⎛c ⎞⎛ H ⎞
c A,G = ⎜ G ⎟ ⎜ ⎟ c A,L .
⎝ c L ⎠ ⎝ 1atm ⎠

We are given the total molar concentration in the liquid, and that in the gas can be found from
the ideal gas law:
8

N P 1atm
cG = = = = 0.04162M .
V RT ( 0.082L.atm / mol.K ) ( 293K )

Then at equilibrium,

c A,G = 0.75c A,L .

Then

(c *
A,L − c i,L )
=
1 ( c A,G − c i,G ) 1 1
= .
NA 0.75 NA .75 k G

Finally, we find

1 1 1
= + 1.33 .
KL kL kG

Using

k L = (100cm −1 ) D A,L

and

k G = (100cm −1 ) D A,G ,

we get

1 1 1
= + 1.33
K L (100cm ) (10 cm / s )
−1 −5 2
(100cm )(10−1 cm 2 / s)
−1

and

cm
K L = 0.001 .
s
The resistance is dominated by the liquid phase, because of the vast disparity in the diffusion
coefficients in liquids and gases.

You might also like