Diffusion in a Concentration Gradient

• Boltzmann-Matano Solution for Variable D

– Matano Interface
– Zero flux plane
– Numerical solution

• Mass flow compensation by bulk flow

– A. Kirkendall effect
– B. Darken Analysis
– C. Boltzmann-Matano Solution
 Variable D

• Most real alloy systems have D varying with

concentration.
• If D is only a function of concentration - not distance -,
then the Boltzmann transformation may be used.
• This transformation introduces a similarity variable,
=x-XM/2t1/2 that depends on a new coordinate called the
Matano Interface, XM.
• The Matano Interface, is the balance point, found using
the Matano geometry that specifies compatible boundary
conditions.
• The Boltzmann-Matano equation extracts D(C) from C(x)
data.
 Variable D

• Fick’s second law with variable diffusivity D(c) is:

C   C
 DC  
t x  x 
• Introducing Boltzmann’s similarity variable, we obtain
the first spatial derivative using x  XM

and the chain rule. 2 t
C C
 
 1 C
    

x  x 2 t   

C C x  X M C
       
t  t    
3
4t 2
 Variable D

• Fick’s second law becomes, using   1

  
x 2 t

  C 1    C

DC     DC  
x   x  4t    

• or
  dC 1 d  dC
     DC  
2t d  4t d  d 
 Boltzmann-Matano Method

• The Boltzmann–transformed Fick’s second law, and

the experimental configuration for finding D(c) is
called the Boltzmann–Matano Method.
• Initial conditions are: (L is left, and R is Right)

C  CL , (x  0, t  0)

C  CR , (x  0, t  0)
 Boltzmann–Matano geometry
for a diffusion couple

To find Matano Interface:

C
CL
A5
time=const.
time = constant
▪Plot c vs. x profile from A4
experimental data.
▪Locate zero flux plane.
A3
C'(x)
▪“Reset” origin.
A2 A1
CR
C=0

C=0
xM x x

“Matano Interface”
"Matano at XM
Interface"
 Matano Interface

• Find the Matano Interface, where area above the

curve equals area below the curve:

CM CL

 xdC   xdC  0
CR CM

• The solute distribution provides a gradient, dC/d,

that vanishes at C= CR .
• Sample is treated as infinite.
 Boltzmann Matano Solution

The solution is:

C
1 dx 
DC      x  X M dC
2t dC
C  CR

• where dx/dc is the reciprocal slope of the profile at c’

• ∫x dc is the area under the curve between cR and c’
• and t is time.
• Multiplied together, the result in cm2/sec is the
chemical diffusivity or interdiffusion coefficient.
 Exercise
1. Interdiffusion experiments carried out in Ta-W alloys by Romig and
Cieslak involved the use of pure Ta and pure W as end-members.
The diffusion couple was annealed at 2100˚C for 16 hours. At the
end of the anneal the diffusion couple was removed from the
furnace and quenched. The couple was sectioned for analyzing
the concentration. Electron microprobe analysis yielded the
concentration profile shown next.
2. The Boltzmann–Matano solution was applied to these
concentration data, and diffusivity as a function of concentration
was calculated.
3. The diffusivity profile is plotted as a function of composition. The
concentrations are in the couple range from pure Ta to pure W.
Careful inspection of the results show that the calculated
diffusivities give accurate results, so long as the concentrations are
not too far from the average composition of the two end-member
alloys.
 Experimental Concentration Profile for Ta
in a Ta–W Diffusion couple

c vs. x Profile 100

Tantalum Concentration, [at %]

80.0 Ta–W
Ta-W alloys
Tantalum Concentration [at.%]

alloys

60.0

40.0

20.0

0.00
0 20 40 60 80 100 120

Distance , [m]
Distance [m]
 Numerical Calculations Applied to the Experimental Profile

-10
1.0 10
Calculated
chemical
D (C) – Boltzmann–Matano
D(C)-Boltzmann-Matano
diffusivities /s] /s]
2 2
[cm 8.0 10
-11
 D (C) – Sauer–Freise–den
D(C)-SFdB Broeder

-11
6.0 10
Diffusivity,[cm
Diffusivity

4.0 10-11

2.0 10-11

0 20 40 60 80 100

Tantalum Concentration
Tantalum (at.%) [at%]
Concentration,
 Boltzmann Matano Solution

• Whenever you have an arbitrary concentration

profile, c vs. x, you can use this solution to obtain an
estimate of “D”.
• The profile is neither erf, symmetric, series solution,
nor thin film.
• The information is obtained from chemical analysis
that is spatially resolved, so that you can plot c vs. x.
• You know the melting temperatures as a function of
composition (liquidus on a phase diagram), and you
want an estimate of “weighted average” of diffusion
coefficients in the overall diffusion process.
• Remember diffusion coefficients are high when
melting temperatures are low.
 Glicksman’s Summary of Interdiffusion

• Interdiffusion occurs in an alloy with composition

gradients when the intrinsic diffusivities are not equal.
Interdiffusion gives rise to the Kirkendall Effect.
• The interdiffusion can be described in a “volume-
fixed” reference frame by a single diffusion coefficient
known as the interdiffusivity, which is related to the
intrinsic diffusivities by the Darken Equation.
• In addition to concentration gradients, other driving
forces can lead to mass diffusion, such as, thermal
gradients, stress gradients, capillary driving forces,
and electrical potential gradients.
 Kirkendall’s Diffusion Couple

Schematic of the Smigelskas–Kirkendall diffusion couple

h(t) 70-30 Brass

Cu

2.5mm Mo wires
0.13mm diameter
 Kirkendall Effect: What if DA > DB?

• Kirkendall studied Mo markers in Cu-brass (i.e., fcc

Cu70Zn30).
• Symmetry is lost: Zn atoms move more readily in one
direction (to the right) than Cu atoms move in the other
(to the left).
• When diffusion is asymmetric, interface moves away
from markers, i.e., there is a net flow of atoms to the
right past the markers.
• Analyzing movement of markers determines DZn and
DCu, intrinsic diffusivities.
• Kirkendall effect driven by vacancies, effective for T >
0.5 Tmelt.
 Kirkendall Effect

• Zn melts at a lower temperature than Cu.

• Suggests DZn > DCu.
• Fluxes of atoms not equal, JZn > JCu.
• Means net mass flow, with lattice planes created to
handle the additional atoms.
• Means net bulk flow, with lattice planes disappearing
where they are no longer populated with atoms.
• This imbalance is hard to imagine. The material is
not growing holes inside, or is it? The material is not
creeping along by getting shorter on one end and
longer on the other, or is it?
 Analogy to Kirkendall Effect
• Suppose you have a cylinder filled with gas. There is
a frictionless piston separating 2 gases, Ar on the
right and H2 on the left. The piston is stationery, and
the pressures in the chambers are equal.
• Perforate the piston to let the gases interdiffuse. The
velocity of a gas molecule is inversely proportional to
the square root of its mass. V(H2) > V(Ar), since H2 is
lighter.
• If more H2 is moving to the right, and less Ar is
moving to the left, which way will the piston move?

• H2 Ar
 Analogy continued
• The pressure on the right will exceed the pressure on
the left. The piston will slide to the left to equalize the
pressures.
• The piston will move in the same direction as the
slower moving species (Ar).

• In the solid state, when mass flow is compensated by

bulk flow, the Matano interface coincides with the
initial position of the markers, relative to the physical
ends of the sample.
• The markers move with the slower moving species.
 Kirkendall Experiment
It was demonstrated that the distance between the inert markers h
behaves according to the following kinetic expression

h
t  h0 Kt 2

h(t) is the marker separation at the annealing time, t

where h(0) is the initial separation of the markers

K is the rate constant in m/(day)1/2

 Kirkendall Experiment
Zn penetration curves observed for various
annealing times in Kirkendall–Smigelskas experiment
32
C0
56 days
24
Zn
wt% Zn

% Zn
wt%

16

8
6 days

0
0 -0. 1 -0. 18
Distance from interface [cm]
 Kirkendall Experiment
Kirkendall–Smigelskas experiment showing inert
Mo wire–marker positions versus time
Significance of Kirkendall Marker Motion
Marker motions observed during Kirkendall Diffusion

crystal
lattice
Um

x marker frame

marker

observer X-scale
 Profile as a function of time

In lab reference frame (observer scale), the

markers move!

Cu70Zn30 Cu
Cu
t=0
Zn

Mo wire markers

After diffusion
 Nonreciprocal Atomic Fluxes

The nonreciprocal atomic fluxes of species A and B written in terms of

the intrinsic diffusion coefficients Di (i =A, B), using Fick’s first law:
CA
J A   DA (t  const.)
x

CB
J B   DB (t  const.)
x

Note: The net atom flux in the Kirkendall couple is no longer zero, because
of nonreciprocal diffusion implied by the fact that DADB.
 Nonreciprocal Atomic Fluxes
The net atomic flux is the algebraic sum of the intrinsic Fickian fluxes
specified
Jnet = JA +JB = -DA dCA/dx + -DB dCB/dx

The marker speed may be related  mol cm3 cm

to the net atomic flux as uM  J net   2   
cm  s mol
 s 

 CA CB 
Substituting yields the relation uM  DA  DB 
 x x 

If the two components have equal partial molar C  C  1

A B
densities then
or, C  C  1
A B
 Nonreciprocal Atomic Fluxes
CA  CB 
Differentiating provides     
 x t  x t
CA  NA
Defining NA and NB,
CB  NB

NA NB Implies that

Substituting gives the relationship  gradients are
x x equal and
opposite
N
Combining we obtain uM  DA  DB  A
x

N
or uM  DB  DA B
x
 Intrinsic Diffusivities: Vacancy Wind

The conservation of lattice sites requires that JV  J A  JB  0

or JV  J A  J B 

CA CB
Substituting for JA and JB JV  DA  DB
x x

C
JV  DB  DA  B
The concentration gradients are
equal and opposite x
 Vacancies at equilibrium
40

DA/DB=1

JB
Atomic fluxes
20 during
Flux (JA, JB, JV), dCV/dt

compensated
diffusion
0 JV=0 (dCV/dt)=0 DA = DB

-20 JA

-40

-4 -2 0 2 4

Distance, x
 40
Vacancy creation Vacancy annihilation

30
DA/DB=1.1

20

JB
Flux (JA, JB, JV), dCV/dt

10 Atomic fluxes
JV
during
0
(dC /dt)
V
uncompensated
-10 diffusion
DA = 1.1  DB
-20
J
A

-30

-40

-4 -2 0 2 4

Distance, x
 Vacancy creation Vacancy annihilation
40
DA/DB=4/3
Flux (JA, JB, JV), dCV/dt

20 J
B

JV
Atomic fluxes
during
0 uncompensated
(dC /dt)
V
diffusion
DA = 1.33  DB
-20

JA

-40

-4 -2 0 2 4

Distance, x
 60
Vacancy annihilation Vacancy creation

DA/DB=2/3
40
Atomic fluxes
J
B during
Flux (JA, JB, JV), dCV/dt

20 uncompensated
diffusion
DA = 0.67  DB
0

(dC /dt)
V

JA
-20
J
V

-40

-4 -2 0 2 4

Distance, x
Intrinsic Diffusivities:Manning’s Vacancy Wind

Vacancies and lattice planes are Vacancies and lattice planes are
created annihilated

(dCV/dt)<0 (dCV/dt)>0

(a) (b)
Vacancy Concentration in Kirkendall Diffusion
1

0.8 DA/DB=0.5
Creation
L
Vacancy Saturation, (Cv -C eq )/C eq

DA/DB<1
0.6
L

Annihilation
DA/DB=2/3
0.4
DA/DB<1

0.2
DA/DB=0.9

0
DA/DB=1.1
-0.2
DA/DB=4/3
Annihilation
-0.4 Creation DA/DB>1
DA/DB>1 DA/DB=2
-0.6
-3 -2 -1 0 1 2 3

Distance, x
 Darken’s Analysis of the Kirkendall Effect
Marker motion in a Kirkendall diffusion couple needed for Darken’s analysis

Original interface
UM=0 UM=0

x
UM

Laboratory frame X

Fixed ends!
CA  CB  1  const.
 Darken’s Analysis of the Kirkendall Effect

For all compositions found throughout the couple
t
CA  CB  0

  CA CB  
Expanding
DA  DB  uM CA  CB  CA  CB 
x  x x  t

  CA CB 1

The right–hand side is zero, so DA  DB  uM    0
x  x x 

Integrating from the left end of the couple, x = -, to an arbitrary plane, x

x'
  CA CB 1

d DA  DB  uM    const.
  x x 

 Darken’s Analysis of the Kirkendall Effect

Integration gives
x'
CB 
x' x'
CA uM
DA  DB     const.
x   x   

uM =0,
The boundary conditions imposed near
both ends CA/x =0,
CB/x =0.

 CA CB 
Expressing the concentrations as uM  DA  DB
molar fractions  x x 

 Darken’s Analysis of the Kirkendall Effect
NA NB
Integration gives uM  DA  DB
Darken’s first equation x x

NA
uM  DA  DB  ,
In terms of either A or B gradients x
NB
uM  DB  DA .
x

Interrelating through the lattice motion, uM,

and applying the Galilean transformation X  x  uM t
 Darken’s Analysis of the Kirkendall Effect
Inserting Darken’s first equation   CA  CA
we obtain DA  uM CA 
x  x  t

  CA  N A  CA
and DA  DA  DB CA 
x  x x  t

Multiplying both sides   NA NA  N A   N A

and simplifying DA  DA N A  DB N A  
x  x x x  t

   N A NA  N A   N A
or
DA  DA1 N B   DB N A 
x  x x x  t
 Darken’s Analysis of the Kirkendall Effect
The identical form of Fick’s second law becomes   ˜ CA  CA
D  
x  x  t

The identical form of Fick’s second law

Becomes Darken’s second equation D˜  NB DA  N ADB

Example: intrinsic diffusivities are defined DA  0.8 0.2NB2 ,

as linear and quadratic functions of the mole
fraction of B. DB  2  0.5NB.

NA
Darken’s two equations are capable of uM  DA  DB  ,
analyzing the Kirkendall effect: x
D˜  NB DA  N ADB.
 2.5

DB Interdiffusion coefficient
D *= 2
B

Darken’s Analysis of the

units]
Diffusivity,

~
1.5
D Kirkendall Effect
Intrinsic[arbitrary
DA, DB
Diffusivity

1 D *=1 Tracer diffusivity of A

A
D
A

Tracer diffusivity of B
0.5

0
0 0.2 0.4 0.6 0.8 1

Mole fraction, N B
 Strain Effects

Sn–Ag Solder

Cu6Sn 5 Kirkendall
porosity
Cu3Sn

Cu
 Exercise
1. An interdiffusion experiment is performed at
1054˚C for 300 hours with pure copper and pure
nickel as end members of a diffusion couple. The
concentration profile is measured as shown in the
next chart. Inert markers placed within the couple
move 200m during the period of the diffusion
anneal.

• Determine the intrinsic diffusivities, DCu and DNi at

1054˚C for a concentration near the Matano
interface, x=0.
 Exercise

100

80
Concentration, C Cu [at.%]

slope, (dN /dx),

60 Cu
at C =71 at.%
Cu

40

20

0
0 0.05 0.1 0.15 0.2
Distance, x [cm]
 Exercise
To determine the interdiffusion coefficient perform
a Boltzmann–Matano analysis.
X 2 
t
 k (const.)
t

The corresponding marker speed is

 X 
t k
uM  
t 2X 
t

Insert the proportionality constant

X t 
uM 
2t
 Exercise
Calculate uM

X t  200[m]
uM  
2t 2  (300)  (3600) s 
 9.3 105 [m / s].

Insert the marker speed uM , D, and the local concentration gradient,

dNCu /dx

uM  9.3 109 [cm/s]  DNi  DCu 12.69 [mol.frac./cm],

D˜  3.16 1010 [cm2/s]  0.71 DNi  0.29DCu .

 Exercise
In a Kirkendall marker experiment, analyzed by the
Boltzmann-Matano solution, the following results were
obtained at the position of the markers after a diffusion time
of 200 hours:
• Displacement of the markers from the Matano interface
= 0.144 cm
• Composition at the markers = 0.40 atomic fraction of
component A
• Composition gradient at the markers dXB/dx = 2.0 cm-1
• Interdiffusion coefficient at the markers Di = 10-7 cm2/sec
• Compute the intrinsic diffusivities DA and DB

46
 Exercise

 
C

DC      x  X M dC
1 dx
2t dC
C  CR

 X 
t k
• 0.144 cm/s__ uM  
t 2X 
t
• 2 (7.2x105 sec)
• 10-7 cm2/sec= 0.40 DB + 0.60 DA
• uM = (DB – DA) 2.0 cm-1

Solve simultaneous equations in 2 unknowns.

47
 Exercise

• 1.04x105 = DB – DA
• 0.40(DA + 5.00x10-8) + 0.60 DA = 10-7

• DA = 8.0 x 10-8 cm2/sec

• DB = 1.3 x 10-7 cm2/sec

48
 Glicksman’s Key Points
▪The diffusivity from the Boltzmann-Matano analysis
represents the effective mixing of A and B atoms. It is called
the interdiffusion coefficient, or the chemical diffusivity.
▪The motions of individual atoms are controlled by intrinsic
diffusivities, because, in general, atoms move through their
surroundings at unique speeds. This is nonreciprocal
(uncompensated) diffusion, occurring whenever DA DB.
▪Kirkendall showed that nonreciprocal diffusion occurs in
Cu-Zn couples, and measured the associated lattice marker
movements.
▪Darken analyzed the Kirkendall experiments.
▪Manning suggested that a “vacancy wind” is needed to
explain lattice plane advection.
▪Strains often develop during Kirkendall diffusion.
 Microstructure terms and definitions:

• Boundary 2 of the same materials

• Coherency Lattice match
• Crystal Having long range order
• Dendrite Having branches
• Equiaxed Having similar size
• Interface 2 different materials
• Lattice Arrangement of atoms
• Polycrystal Having multiple orientations
• Single crystal Having one orientation
• Twin Mirror image