Lecture 7: How to determine the binary interdiffusion coefficient in real

experiments

Todays topics

Continue to understand the interdiffusion coefficient, D , as defined in Darkens equation

(see Lecture 6).

Learn how to experimentally measure the value of D : understanding the mathematic

methods and how to apply the math in the data analysis.

Following what we learned in last lecture, we will learn in this lecture how to measure the
interdiffusion coefficient in real experiments, which was first carried out by Kirkendall and
published in 1942 (E.O. Kirkendall, "Diffusion of Zinc in Alpha Brass," Trans. AIME, 147
(1942), pp. 104-110). See the additional reading for detailed description of the experiment and
data analysis.

Darkens equation: D = xA DBC + xB DAC

DAC, DAC are in general functions of (or dependent on) composition:
DAC = DA { 1 +

d ln B
d ln A
}=DA{ 1 +
}
d ln x A
d ln xB

DBC = DB { 1 +

d ln B
d ln A
}=DB{ 1 +
}
d ln xB
d ln x A

Therefore, D is difficult to measure. It can only be determined (estimated) by graphical or

numerical method (but no analytical solution).
dc A
dcB
JA = - D
,
JB = - D
dx
dx
The experimental procedure usually used for determining D follows:
form a diffusion couple anneal with a given time measure composition profile, which can
be done by electron probe microanalysis (EPMA) or other solid composition analysis method like
XPS.

C0
Before
after

C1
x

Rate of changing composition

c A
J
c

=
A =[ D A ]
t
x x
x
cB
J
c

=
B =[ D B ]
t
x x
x

For given t, c(x) can be measured, but not

c
t

The mathematic approach to be introduced herein for determining D was first suggested by
Boltzmann, and then demonstrated by Matano --- now called Boltzmann-Matano method,
which is based on graphical integration of diffusion equation (Ficks second law).
Ficks second law:
Lets introduce,

c A c A
]
= [D
t
x
x

x
,
t

where x and t are independent, representing the space and time term.
Then,

d = (

now,

re-write:

)t dx + ( ) x dt
x
t

1
,
)t =
x
t

)x = 3/ 2 =
t
2t
2t

c
c c
=

=
t t
2t

(1)

c c
1 c

x
x x
t
t
So, the right term of Ficks second law above can be re-written as

c
1 D c
[D ] =
[
]
x x
t t

(2)
2

Thus, using Eq. (1) and the Fick law above, we have

c 1 D c
= [ ]
2t
t t

For a fixed time period of annealing, t = const,

c 1 c

=
[D ]
2t t

the above equation can be re-written as

Then,

c =[ D

c
]

Integrate from C0 to C, we have

c
c
1 c
- D
= - dc
D
2 c0
c
c0
C0 is far away from diffusion zone, c0 is constant.

c
=0
c0

Then, we have,

c
1 c
= - dc
D
2 c0
c
1 c
So, D = - ( )c dc
2 c c0

substituting back =

x
, and now t (as a fixed annealing time) is a constant, we have
t

1 x c
D (c) = - ( )c xdc
2t c c0

--- D is a function of concentration, c.

This is Boltzmann - Matano equation.

By measuring c(x) experimentally, we will be able to give (
c
!)
t

However

x
)c (we no longer needs to know
c

xdc depends on origin of x, so one must determine where to measure x from

3

Arbitrarily choose x in the diffusion zone

c0
C

c*
c
c1
0
Zero point

x<0

x<0

Now consider a composition to the both sides of the chosen origin:

c0

xdc =

First, considering the top part,

Since x<0, c<c0 (or dc<0),

c*

c0

so,

xdc =

c*

c0

Second, considering the lower part,

Since x>0, c<c* (or dc<0),

Also note, (

x
)c <0,
c

so,

c*

xdc >0

c*

xdc =

xdc <0

c , and x

Then,
1 x
D = - ( )c
2t c
1 x
= - ( )c {
2t c

As long as {

c0

xdc

(>0) +

(<0) }

} > 0, hen
t
D >0

As a boundary condition, when c c1 (the final equilibrium concentration), (

D still to be finite, {

x
)c , to have
c

} must go to zero.

--- Thats to say, x origin must be chosen to have left and right intergral equal --- this is referred
as the Matano interface.
1 x c
Then, we can determine D =
xdc , where x is measured from Matano interface that
2t c c0
needs to be determined first.
4