Diffusion

Diffusion means atoms moving and changing places. This happens

in solids and liquids, exactly in the same way that an unpleasant
smell moves from one part of a room to another even without wind,
then dissipates after some time.
The driving force for diffusion is the chemical potential gradient.
Partial Molar Free Energy=Chemical Potential

Gi i

 
 i P1  
 i Po    P

Ficks First Law

atomic flux is proportional to the chemical potential gradient

C  d d 
Ji D Ji   D i  i 
kT  d x d x 

forsolution
For an ideal an ideal solution

i  o  kT ln Ci  
d  d 
Ji D   i  D  Ci 
Substitution into the derivative term
d x with only composition varying with x
dx    

D 
d C 
Ji i
 dx 
For diffusion due to a pressure gradient (as in sintering)

D  d P 
 C   
J 
k T  dx 
Preview: In steel, strength goes up but toughness goes down with C.
Design application: Carburization of a transmission gear.

Attributes needed?

How to do it.

Attributes needed?

How to do it.
“Carburizing”

2 CO CO2 + C (in solution)

Carbon diffuses from the surface

Into the steel
Mechanism of diffusion in solids
Diffusion of an individual
atom is random and
probabilistic. Thus, the
interstitial black atom has
an equal probability of
moving up, down, left or
right.
 Mechanism of diffusion in solids
If we have a concentration
gradient, overall atomic
movement is not random,
even though individual
atom movement is.

In this example, since some of the black atoms will move to

the right on average, the concentration of black atoms will increase
on the right and decrease on the left. This is similar
to the dissipation of the concentrated unpleasant odor in the room.
 Mechanisms of diffusion in
solids
At equilibrium, there will
be a random distribution of
black atoms, and no
concentration gradient.

A common heat treatment for cast metal alloys is

homogenization, or heating for a long time at high temperature
to allow the chemical segregation to even out in this way.
Mechanism of diffusion in solids
Diffusion Couples

Concentration gradient = dC/dx

is the slope of the curves.

C
Eventually Cu, Ni

x
 Mass transport, or Flux

• Flux is a measurement of the number of atoms per unit

area that cross a particular plane per unit time.

M = mass, A = area, t = time

 Fick’s first law
• We saw that the concentration gradient
affects the direction of diffusion. It also
affects the rate of diffusion. For simplicity
let us assume the concentration gradient
is a constant, then we can say:

• where D is the “diffusivity”, which depends

 Steady state diffusion

dC/dx = concentration gradient = slope = constant =

C/x = (CA-CB)/(xA-xB)
ALE
 What affects diffusion rates?
• Diffusion is a thermally-activated process. This means
that it is accelerated by temperature. Using a higher
temperature will result in more diffusion in a given time if
there is a composition gradient present (could
accomplish the same effect by prolonging the time to
diffuse at a lower temperature)
• Time: straightforward result of Fick’s First Law:

J is mass/(time x area) or atoms/(time xarea). Total mass

diffusing across a plane of area A is just J x t, if J is
constant. Otherwise M = Jdt
 What affects diffusion rates?
• Temperature: Comes in mainly through the diffusivity, D:

ALE: Take the natural log of both sides of Eqn 6.8,

and construct a plot that would be useful for looking at D(T)
ALE - Answer

Qd
ln D  ln( Do ) 
RT
Qd  1 
ln D      ln( Do )
R T 

y = m x + b
 Nonsteady-State Diffusion
• Most practical diffusion situations are nonsteady-state.
That is, the concentration gradient = f(time), i.e. dC/dx
changes.

Gas

Gear interior
Gear surface
 Nonsteady-State diffusion
• To understand these problems, we need to
solve a differential equation called Fick’s
Second Law:

• It is clear that the concentration gradient

can vary with time through the term dC/dt.
 2
21  18 m
10 7 D  10 
Ci  x  10 m s
3
m
The solution to the PDE is
obtained by conversion to an Ci  ( x m) 2
C( x t)   exp    x
ODE. 2 D t    4  D  t 

Define
x
 ( x t)
2 D t
7 8 7
x  10  9.9 10  10

Use the Chain Rule

3 10
22

 0 
C x 10  s 2 1022

 d   D  d  d  d C  d    C x 10  s
1
d
C      
d  dt   d  dx  d  dx   C x 10  s1 10
2  22

d 
2    C 
d
2
C
0
0
2
 d  d
x
 2
21  18 m
10 7 D  10 
Ci  x  10 m s
3
m

Ci  ( x m) 2
C( x t)   exp    x
2 D t    4  D  t 

Dopant Rich Layer, Ci , ∆x

8 6
x  0  10  10

1 10
22

Semiconductor  1 
C x 10  s

C x 10  s5 1021
2

C x 10  s
3

0
5 10
8
0
x
We can add up the Ci
solutions from thin films C1( x t) 
2

 2
  x m  10 8 m 
 exp 

  x
displaced throughout the 2 D t   4 D t 

volume
Ci
 2 Ci
4   x m  ( 10)  8 3 m    2
  x m  ( 10)  8 2 m 
C3( x t)   exp    x 3
2 D t   4 D t  C2( x t)   exp    x
2 D t   4 D t 

3 10
21
1 10
22

 1
C x 10  s   2
C x 10  s 
 21
C1 x 10  s
2 10
 1 
2
C1 x 10  s
5 10 C2 x 10  s
21
 1 
2
C2 x 10  s
1 10
21

C3 x 10  s
1  2 
C3 x 10  s

0
0 0
0
x
x
 C surface  C initial  C surface  erf 
  
x
C( x t) 
 2 D t 

1
0.995

erf (  ) 0

 0.995 1
2 0 2
2  2
 2 COCO2 +C(Fe)

2at% C at surface
0.2at% initially in low carbon steel
C( x t)  C surface
erf  
x D=10^(-12) m^2/s

C initial  C surface  2 D t 
What is the time required to get 1at%
C( x t)  Csurface at 10^-4 m?
erf  
x

Cinitial  Csurface  2 D t 

erf ( 0.54)  0.555

1
12
 0.556
0.2  2 4
erf (  ) 0
10
0.54
 12
1 10 t
1
2 1 0 1 2

34293
0.54  9.526 hr
3600