Chapter 1, Lecture 2: Microscopic models of diffusion

Martin Z. Bazant

I. MICROSCOPIC MODELS OF DIFFUSION We can think of this random diffusive process mi-
croscopically at two different limits: in dilute mixtures
and in concentrated mixtures. Both employ concepts of
non-equilibrium thermodynamics, in which driving forces
push the system closer towards equilibrium.
1. Dilute mixtures
In dilute mixtures, a single particle follows a “Ran-
dom walk” in which new steps are not influenced
by previous ones.

FIG. 1: Diffusive Spreading

Fick’s law:

F = −D∇c (1)

The conservation equation for the concentration with a FIG. 3: Random walk
constant diffusion coefficient gives the diffusion equation:

∂c ∂2c 2. Concentrated Mixtures

= (2)
∂t ∂x2 In concentrated mixtures, the transport process in-
volves chemical diffusion governed by the forces be-
The equation can be used to describe diffusive spread-
tween different species.
ing, which is a consequence of random molecular fluctu-
ations.

FIG. 4: Chemical diffusion

FIG. 2: Fluctuations leading to diffusion

 2

II. HISTORICAL EXAMPLES OF RANDOM III. FROM RANDOM WALKS TO DIFFUSION

WALKS

There is a rich history underlying the theory of random

walks:
1. Pearson (1905): Random walk of mosquitos
2. Rayleigh (1880): Acoustic scattering of sound
waves
3. Bachelier (1900): Stock prices
4. Einstein (1905): Brownian motion
5. Taylor (1922): Turbulent diffusion FIG. 5: Probability distribution of a random walker
6. Flory (1940s): Self-avoiding random walks for poly- after N steps
mers
Assumptions:
1. Dilute mixture with non-interacting random walks
2. Independent, identically distributed (1D) steps.

Z
PN +1 (x) = PN (x − y)p(y) dy (3)

3. Finite step variance

Z
y 2 p(y) dy < ∞ (4)

Armed with these assumptions, the next step is to take

the limit as N → ∞:
y2
Z  
PN +1 (x) = PN (x) − yPN0 (x) + PN00 (x) − ... p(y) dy
2
Z Z
0 1
= PN (x) − PN (x) yp(y) dy + y 2 p(y) dy − ...
2
(5)

Because p(y) is a probability distribution, we can ex-

press the above equation in a closed form, neglecting
higher order terms.

PN +1 (x) − PN (x) hyi 0 hy 2 i 00

+ PN (x) = P (x) (6)
τ τ 2τ N
In the continuum limit, we arrive at the drift-diffusion
equation:

∂c ∂c ∂2C
+v =D 2 (7)
∂t ∂x ∂x
From this equation, we can define a drift velocity, v:

hyi
v= (8)
τ
 3

Similarly, we can define diffusivity, D. The diffusivity IV. CHEMICAL DIFFUSION

is determined as per the case of the random walk with a
drift velocity shown in homework problem 2: Here we turn to a concentrated mixture in which inter-
actions between species are present. These interactions
lead to non-ideal thermodynamics, for example, in the
σ2 hy 2 i − hyi2 chemical potential of a species, µ. The chemical po-
D= = (9)
2τ 2τ tential is a measure of the change in free energy with an
infinitesimal change in the concentration of a species.
∆G
µ= = µΘ + kB T ln(γc̃) (10)
∆N
c
where c̃ = cmax . cmax is the maximum possible con-
centration of the molecule, dependent on the size of the
molecule.
Assuming linear irreversible thermodynamics
(LIT), the flux of a species is proportional to its concen-
tration, c, times the driving force for transport, −∇µ.
In general, species diffuse in order to move down their
chemical potential gradient.

F = −M c∇µ (11)

According to the Einstein relation:

D
M= (12)
kB T
Substituting in equations 10 and 12 into the flux ex-
pression from equation 11:


F = −Dc∇ ln(γc̃)
∇γ ∇c

= −Dc +
γ c
(13)
c ∂γ

= −D + 1 ∇c
γ ∂c
= −Dchem ∇c

Through this process, we arrive at the definition of

chemical diffusivity, Dchem .

∂ ln γ
Dchem = D 1 + ) (14)
∂ ln c

For an ideal dilute mixture, the activity coefficient is

γ = 1, and the chemical diffusivity equals the dilute dif-
fusivity, Dchem = D. Therefore, the diffusivity is scaled
by a thermodynamic factor to arrive at the chemical dif-
fusivity in concentrated mixtures.
 4

Example IV.1 Diffusion on a lattice

The chemical potential of a species on a lattice is modi-

fied by including the entropy of lattice vacancies, which
have a dimensionless concentration of 1 − c̃, as well as
the enthalpy of species in the lattice.
Enthalpy
(forces)
z}|{
µ = kb T [ln(c̃) − ln(1 − c̃) + h̃ ] (15)
| {z }
Configurational entropy
(Create particle and destroy vacancy)

The activity coefficient, γ, is:

eh̃
γ= (16)
1 − c̃
We can now use this result to derive the chemical dif-
fusivity. First, from the thermodynamic factor:

c ∂γ h̃0 1
= c̃e−h̃ (1 − c̃)eh̃


γ ∂c 1 − c̃ (1 − c̃) 2
(17)
1
= c̃(h̃0 + )
1 − c̃
Here, we can consider the tracer diffusivity that
would be measured in a diffusion experiment on this
lattice. The tracer diffusivity D0 is scaled by the ex-
cluded volume (1 − c̃) to give the true diffusivity.

D = D0 (1 − c̃) (18)

Using this definition:

c̃
+ c̃h̃0


Dchem = D 1 + (19)
1 − c̃

Dchem = D0 1 + c̃(1 − c̃)h̃0

(20)

If h̃ = 0, as in an ideal solid solution, then Dchem = D0 .

This is a statement of particle-hole duality.

FIG. 6: Diffusion on a lattice