MASS DIFFUSION

In this section the mass transfer process is described. The Brownian diffusion of
small particles and Fick's law are first discussed. This is followed by the presentation of
a number of applications.

Brownian Diffusion
Small particles suspended in a fluid undergo random translational motions due to
molecular collisions. This phenomenon is referred to as the Brownian motion. The
Brownian motion leads to diffusion of particles in accordance with Fick’s law. i.e.,

dc
J = −D (1)
dx

where c is the concentration, J is the flux, and D is the diffusion coefficient. When the
effect of particle inertia is negligible, using (1) in the equation of conservation of mass
for particles leads to

∂c
+ v ⋅ ∇c = D∇ 2 c (2)
∂t

where v is the fluid velocity vector. The particle mass diffusivity is given by

kTC c
D= (3)
3πµd

where C c is the Cunningham correction given by (3) and k is the Boltzmann constant
( k = 1.38 × 10 −16 erg / K ). The diffusive may be restated as

τkT
D= (4)
m

where m is the mass of the spherical particle and τ is its relaxation time.

Table 8 – Particle mass diffusivity.

d ( µm ) D (cm 2 / s )
10-2 5.24 × 10-4
10-1 6.82 × 10-6
1 2.74 × 10-7
10 2.38 × 10-8

ME437/537 1 G. Ahmadi
The mean-square displacement for a Brownian particle is given as

s 2 = 2Dt (one-dim) (5)

Brownian Motion of Rotation

Aerosol particles may also rotate randomly due to the Brownian effects. The
mean-square angle of rotation is given as

2kT
θ2 = t (6)
πµd 3

Distributions

When the gas is in equilibrium, the aerosol particle will have the same average
translational energy as molecules. Thus

1 3
m u 2 = kT , (7)
2 2

and the root-mean-square particle velocity is given by

u 2 = 3kT / m (8)

Under equilibrium, aerosol particles will have a Maxwellian distribution and their
concentration in a gravitational field is given by

mg( x − x 0 )
C = C 0 exp{− } (9)
kT

Effect of Mass

The diffusivity as given by (3) and (4) is independent of particle density, but
heavy particles do not respond swiftly to the molecular impacts. A time dependent
analysis leads to

s 2 = 2Dt[1 − τ(1 − e − t / τ ) / t ] (10)

where τ is the particle relaxation time. when t >> τ , (10) reduces to Equation (5).

ME437/537 2 G. Ahmadi
Aerosols Mean Free Path

The apparent mean free path for aerosol particles λα is defined as the average
distance that a particle moves before changing its direction by 90o. The average
(absolute) velocity of an aerosol particle is 8kT / πm . From the definition of stop
distance it follows that

λ α ≈ τ 8kT / πm (11)

λα becomes a minimum for an aerosol particle diameter of about 0.05 µ m . λα is of the

order of 10-6 cm for d < 5 µ.

Particle Diffusion to a Wall

For a one-dimensional case, the diffusion equation given by (18) in the absence of
a flow field becomes

∂c ∂ 2c
=D 2 (12)
∂t ∂y

For an initially uniform concentration of aerosols in the neighborhood of an absorbing

wall, the initial and boundary conditions are: C( y,0) = C 0 and C (0, t ) = 0 . The solution
to Equation (28) then becomes

C( y, t ) = C0erf ( y / 4 Dt ) , (13)

where

ξ
2
∫
2
erf (ξ) = e − ξ dξ , erf (0) = 0 , erf (∞) = 1 (14)
π 0

The variation of concentration profile with time are shown in Figure 1. (Note that y has
the same unit as Dt .)

The flux to the wall then is given by

∂c D
J = −D = C0 (15)
∂x y =0 πt

where C 0 is the particles number concentration at the initial time.

ME437/537 3 G. Ahmadi
 1.0
tD=0.0025
tD=0.062
tD=0.25
0.8 tD=1

0.6
C/Co

tD=4

0.4

0.2

0.0
0 0.5 1 1.5 2 2.5 3
yx
Figure 1. Variation of concentration profile with time.

The corresponding deposition velocity, which is defined as flux per unit

concentration, then is given by

J D D
uD = = = (16)
C0 πt δ c

Here δ c is the diffusion boundary layer thickness given as

δ c = πDt (17)

The corresponding diffusion force is defined as

Fd = 3πµdu D / C c (18)

The total number of particles that is deposited in an interval dt is given as

ME437/537 4 G. Ahmadi
 D
dN = Jdt = C 0 dt (19)
πt

The total number of particles that are deposited per unit area in the time interval 0 to t
may be obtained by integrating Equation (19). Thus,

4Dt
N = C0 . (20)
π

Tube Deposition

Consider a constant velocity gas flow in a tube of length L and radius R. The
residence time is t = L / u where u is the gas velocity. Assuming that the wall deposition
process is similar to that of a uniform concentration near a wall, and using (20) it follows
that

2RπL 4DL
C out − C in = − N , N = C in , (21)
πR 2 L πu
or
C out 4 DL
= 1− . (22)
C in π uR 2

A detailed duct flow analysis shows

C out DL
= 1 − 2.56φ 2 / 3 + 1.2φ + 0.177φ 4 / 3 , φ = . (23)
C in uR 2

Diffusion Velocity

The diffusion velocity is defined as

J
UD = . (24)
C0

Similarly, a diffusion force may be defined as

Fdiff = 3πµu d / C c (25)

ME437/537 5 G. Ahmadi
Convective Diffusion to a Flat Plane

Consider a laminar boundary layer flow over a flat plane as shown. The
equations of motion and mass diffusion are

∂u ∂u ∂2u
u +v =ν 2 (26)
∂x ∂y dy

∂u ∂v
+ =0 (27)
∂x ∂y

∂c ∂c ∂ 2c
u +v =D 2 (28)
∂x ∂y ∂y

Co Uo Uo Co

x δc

Figure 2. Schematics of boundary layer flow over a flat plate.

The boundary conditions are

At y = 0 , u=v=c=0 (29)
As y → ∞ , u = U0 , c = c0 (30)

Introducing similarity variable,

U0 u
η= y , ψ = νU o x f ( η) , = f ' (η) , c = c(η) , (31)
νx U0

Equations (26) – (28) reduce to

ME437/537 6 G. Ahmadi
 ff ' '+2f ' ' ' = 0 , f (0) = f ' (0) = 0 , f ' (∞) = 1 (32)

1
c"+ Sc fc' = 0 , c(0) = 0 , c(∞) = c 0 (33)
2

where the Schmidt number is defined as

ν
Sc = . (34)
D

For a large Schmidt number, δ c << δ as shown in the figure.

The solution to the Blasius equation (32) is well know and leads to

νx
δ=5 , f ' ' (0) = γ = 0.332 . (35)
U0

Near the plate then

γ
f ~ η2 + ... (36)
2

Using (36) in Equation (33), after integration we find

η

C0 ∫ exp( − γ1s c z 3 )dz

γ
C= ∞
0
, γ1 = (37)
12
∫ exp( − γ s z
3
1 c )dz
0
Noting that

∞ ∞
1 Γ(1 / 3)
∫ [exp(−γ1s c z )]dz = ∫ [exp(−z )]dz1 =
3 3
1 , (38)
0
3 γ 1s c 0 3 3 γ 1s c

We find

η
c 3 γ 1s c
c0
=
0.89 0∫ [exp(− γ 1s c z 3 )]dz (39)

Variation of concentration profile as given by Equation (39) is shown in figure 3. The

mass flux is then given by

ME437/537 7 G. Ahmadi
 1.0

0.8

0.6
C/Co

0.4

0.2

0.0
0 0.5 1 1.5 2 2.5 3
ηz
Figure 3. Variation of concentration profile with η for γ 1Sc = 1.

3 (γ s )
 ∂c  U0 U0
J = D   = Dc 0 1 c
= 0.34 Dc 0 3 s c . (40)
 ∂y  y=0 0.89 νx νx

The diffusion boundary layer thickness is then given by

Dc 0 3 vx 0.6δ
δc = ≈ ≈ (41)
J 3 sc U0 3 sc

The total diffusion

L
UoL
I = ∫ Jdx = 0.68Dc 0 3 s c R eL , R eL = (42)
0
ν

Diffusion in a Tube Flow

r2
Consider a laminar flow in a tube with u = u 0 (1 − ) . Let y = R − r . At
R2
distances very close to the wall, that is for small y,

ME437/537 8 G. Ahmadi
 2y
u ~u 0 + ... (43)
R

z
y R u c r
δc

Figure 4. Schematics of convective diffusion in a pipe.

The convective diffusion equation then becomes

2v 0 ∂c ∂ 2c
y =D 2 (44)
R ∂x ∂y

The appropriate boundary conditions are

c = 0 at y = 0 , c = c 0 at y → ∞ (45)

Introducing similarity variable

u0 y
η=3 (46)
DR 3 x

Equation (44) may be restated as

2
c' '+ η 2 c' = 0 (47)
3

The solution to (47), which satisfies (45), is given by

η
−2 3
c 0 ∫ exp{ η1 }dη1
9
c= ∞
0
(48)
−2 3
∫0 exp{ 9 η1 }dη1

ME437/537 9 G. Ahmadi
The diffusion to the wall is given as

Dc 0 3
u0
 ∂c  3
x DR
J = D  = ∞
 ∂y  y =0 2 3 (49)
∫0 exp(− 9 η1 )dη1

u0
J = 0.67c 0 D 3
DRx (50)

The diffusion layer thickness is determined by

3
Dc 0 R 2x 1 R 2x
δc = = = 3 (51)
J 0.67s1c / 3 R 1eR/ 3 0.67 s c R eR
where
u 0R
R eR = (52)
ν
The total diffusion for a length L is given by

L
u 0 L2
I = 2πR ∫ Jdx = 2.01πc 0 DR 3 (53)
0
DR

Diffusion in a Stream in a Tube

The equation governing the convective diffusion in a tube is given as

∂c ∂c ∂ 2 c 1 ∂c ∂ 2 c
+ u (r ) = D( 2 + + ) (54)
∂t ∂x ∂r r ∂r ∂x 2

with
r2
u (r ) = 2U(1 − ) (55)
R2

where U is the mean velocity in the tube. In a coordinate system moving with the mean
fluid velocity U, Equation (54) may be restated as

∂c 2r 2 ∂c ∂ 2 c 1 ∂c
+ V(1 − 2 ) = D( 2 + ), (56)
∂t R ∂x ∂r r ∂r

ME437/537 10 G. Ahmadi
 ∂ 2c
where the axial diffusion is neglected.
∂x 2

For zero flux to the wall, the boundary condition at the tube surface is given as

 ∂c 
 ∂r  = 0 (57)
R

∂c
As a first approximation, in the moving frame is negligibly small and
∂t

∂c ∂c
= const = (58)
∂x ∂x

Now solving Equation (56) for c, it follows that

UR 2 ∂ c r 2 1 r 4 ∂c
c = co + ( − ), =0 (59)
4D ∂x R 2 2 R 4 ∂t
where
R R
1 1 2
c o = [c]r =0 , and c = ∫ cdA = 2 ∫
2πrcdr = 2 ∫ crdr (60)
AA πR 0 R 0

Using (59) in (60), the value of co may be evaluated and then

R 2 U ∂c 1 r 2 1 r 4 ∂c
c =c+ (− + 2 − ), =0 (61)
4D ∂x 3 R 2R 4
∂t

The total flow of substance across the pipe then is given by

R
R 2 U 2 ∂c
Q c = 2π ∫ c(u − U)rdr = −(πR 2 ) . (62)
0
48D ∂x

The flux

Q R 2 U 2 ∂c
J= = − ( ) (63)
πR 2 48D ∂x

has the same form as Fick’s law with an effective diffusivity

R 2U2
D eff = . (64)
48D

ME437/537 11 G. Ahmadi
 ∂c
In the next approximation we drop the assumption that = const., thus
∂x

∂c ∂ ∂2 c
= − J = D eff 2 (65)
∂t ∂x ∂x

2 UR
Equation (65) is applicable if the Peclet number, Pe = = R e Sc satisfy
D
L
>> Pe >> 14 (66)
R

If a certain amount N of substance is introduced at x = 0 , t = 0 , that is

N
c= δ( x ) at t = 0 , (67)
πR 2

Then the solution to Equation (65) is given as

1 N 1 ( x − Ut ) 2
c= exp{− }. (68)
2 πR 2 πD eff t 4D eff t

Variation of concentration as a function of space and time are shown in Figure 5. It is

seen that the concentration travels like a wave but also dispersed along it path.

2.0
1.8
1.6
1.4
1.2
C/Co

1.0
0.8
0.6
0.4
0.2
0.0
0 1 2 3 4
x
Figure 5. Variations of concentration along the tube at different times.

ME437/537 12 G. Ahmadi