Chapter 7

The Diffusion Equation

D(u(r, t), r)u(r, t)

, (7.1)
where u(r, t) is the density of the diffusing material at location r = (x, y, z) and time
t. D(u(r, t), r) denotes the collective diffusion coefcient for density u at location r.
If the diffusion coefcient doesnt depend on the density, i.e., D is constant, then
Eq. (7.1) reduces to the following linear equation:
u(r, t)
t
= D
2
u(r, t). (7.2)
Equation (7.2) is also called the heat equation and also describes the distribution of
a heat in a given region over time.
Equation (7.2) can be derived in a straightforward way from the continuity equa-
tion, which states that a change in density in any part of the system is due to inow
and outow of material into and out of that part of the system. Effectively, no mate-
rial is created or destroyed:
u
t
+ = 0,
where is the ux of the diffusing material. Equation (7.2) can be obtained easily
from the last equation when combined with the phenomenological Ficks rst law,
which assumes that the ux of the diffusing material in any part of the system is
proportional to the local density gradient:
=Du(r, t).
59
7.1 The Diffusion Equation in 1D
Consider an IVP for the diffusion equation in one dimension:
u(x, t)
t
= D

2
u(x, t)
x
2
(7.3)
on the interval x [0, L] with initial condition
u(x, 0) = f (x), x [0, L] (7.4)
and Dirichlet boundary conditions
u(0, t) = u(L, t) = 0 t > 0. (7.5)
7.1.1 Analytical Solution
Let us attempt to nd a nontrivial solution of (7.3) satisfying the boundary condi-
tions (7.5) using separation of variables [4], i.e.,
u(x, t) = X(x)T(t).
Substituting u back into Eq. (7.3) one obtains:
1
D
T

(t)
T(t)
=
X

(x)
X(x)
.
Since the right hand side depends only on x and the left hand side only on t, both
sides are equal to some constant value (sign is taken for convenience reasons).
Hence, one can rewrite the last equation as a system of two ODEs:
X

(x) +X(x) = 0, (7.6)

(t) +DT(t) = 0. (7.7)

x
+C
2
e

x
.
Taking into account the boundary conditions one gets C
1
=C
2
= 0, so for < 0
only the trivial solution exists.
2. = 0:
X(x) =C
1
x +C
2
Again, due to the boundary conditions, one gets only trivial solution of the prob-
lem (C
1
=C
2
= 0).
3. > 0:
X(x) =C
1
cos(

x) +C
2
sin(

x).
Substituting of the boundary conditions leads to the following equations for the
constants C
1
and C
2
:
X(0) = C
1
= 0,
X(L) = C
2
sin(

L) = 0 sin(

L) = 0
n
=

n
L

2
, n = 1, 2, . . . .
Hence,
X(t) =C
n
sin

n
L
x

.
That is, the second equation for the function T(t) takes the form:
T

(t) +D

n
L

T(t) = 0 T(t) = B
n
exp

n
L

2
t

,
where B
n
is constant.
Altogether, the general solution of the problem (7.3) can be written as
u(x, t) =

n=1
A
n
sin

n
L
x

exp

n
L

2
t

, A
n
= const .
In order to nd A
n
one can use the initial condition (7.4). Indeed, if we write the
function f (x) as a Fourier series, we obtain:
f (x) =

n=1
F
n
sin

n
L
x

n=1
A
n
sin

n
L
x

,
A
n
= F
n
=
2
L

L
0
f ()sin

n
L

d .
Hence, the genetal solution of Eq. (7.3) reads:
u(x, t) =

n=1

2
L

L
0
f ()sin

n
L

sin

n
L
x

exp

n
L

2
t

. (7.8)
Fig. 7.1 Schematical repre-
sentation of the FTCS nite
difference scheme (7.9) for
solving the 1-d diffusion
equation (7.3).
u u u
6
t
j
Q
Q
Q
Qk

3
t
j+1
u
7.1.2 Numerical Treatment
The FTCS Explicit Method
Consider Eq. (7.3) with the initial condition (7.4). The rst simple idea is an explicit
forward in time, central in space (FTCS) method [28, 22] (see Fig. (7.1)):
u
j+1
i
u
j
i
t
= D
u
j
i+1
2u
j
i
+u
j
i1
x
2
,
or, with = D
t
x
2
u
j+1
i
= (1 2)u
j
i
+(u
j
i+1
+u
j
i1
). (7.9)
In order to check the stability of the schema (7.9) we apply again the ansatz (1.21)
(see Sec. 1.3), considering a single Fourier mode in x space and obtain the following
equation for the amplication factor g(k):
g
2
= (1 2)g +2gcos(kx),
from which
g(k) = 1 4sin
2
kx
2
.
The stability condition for the method (7.9) reads
|g(k)| 1 k
1
2
t
1
2
x
2
D
. (7.10)
Although the method (7.9) is conditionally stable, the derived stability condi-
tion (7.10), however, hides an uncomfortable property: A doubling of the spatial
resolution x requires a simultaneous reduction in the time-step t by a factor of
four in order to maintain numerical stability. Certainly, the above constraint limits
us to absurdly small time-steps in high resolution calculations.
The next point to emphasize is the numerical dispersion. Indeed, let us compare
the exact dispersion relation for Eq. (7.3) and relation, obtained by means of the
schema (7.9). If we consider the perturbations in form exp(ikx it) the dispersion
Fig. 7.2 Dispersion relation
by means of the schema (7.9)
for different values of ,
compared with the exact dis-
persion relation for Eq. (7.3).
0 0.2 0.4 0.6 0.8 1
0
5
10
15
20
k x/
i

t
/

exact
=0.1
=0.2
=0.3
=0.4
relation for Eq. (7.3) reads
i = Dk
2
.
On the other hand, the FTCS schema (7.3) leads to the following relation
e
it
= 1 4 sin
2

kx
2

,
or, in other words
it =ln

1 4sin
2

kx
2

.
The comparison between exact and numerical disperion relations is shown on
Fig. (7.2). One can see, that both relations are in good agreement only for kx 1.
For > 0.25 the method is stable, but the values of can be complex, i.e., the
Fourier modes drops off, performing damped oscillations (see Fig. (7.2) for =0.3
and = 0.4). Now, if we try to make the time step smaler, in the limit t 0 (or
0) we obtain
it 4sin
2

kx
2

= k
2
Dt
sin
2

kx
2

kx
2

2
,
i.e., we get the correct dispersion relation only if the space step x is small enougth
too.
The Richardson Method
The rst idea to improve the approximation order of the schema is to use the central
diferences for the time derivative of Eq. (7.3), namely [28]
u
j+1
i
u
j1
i
2t
= D
u
j
i+1
2u
j
i
+u
j
i1
x
2
,
or, with = Dt/x
2
u
j+1
i
= u
j1
i
+2

u
j
i+1
2u
j
i
+u
j
i1

. (7.11)
Unfortunately, one can show that the schema (7.11) is unconditional unstable. In-
deed, amplication factor g(k) in this case fullles the following equation:
g
2
+2g 1 = 0, = 4sin
2
kx
2
,
giving
g
1,2
=

2
+1.
Since |g
2
(k)| > 1 for all values of k, the schema (7.11) is absolut unstable.
The DuFort-Frankel Method
Let us consider one of many alternative algorithms which have been designed to
overcome the stability problems of the simple FTCS and Richardson methods. We
modify Eq. (7.9) as (see Fig. (7.3)) [28]
u
j+1
i
u
j1
i
2t
= D
u
j
i+1
2
u
j+1
i
+u
j1
i
2
+u
j
i1
x
2
,
which can be solved explicitly for u
j+1
i
:
u
j+1
i
=
1
1 +
u
j1
i
+

1 +

u
j
i+1
+u
j
i1

, (7.12)
where = 2Dt/x. When the usual von Neumann stability analysis is applied
to the method (7.12), the amplication factor g(k) can be found from
(1 +)g
2
2gcos(kx) +( 1) = 0.
It can be easily shown, that stability condition is fullled for all values of , so the
method (7.12) is unconditionally stable. However, this does not imply that x and
t can be made indenitely large; we must still worry about the accuracy of the
Fig. 7.3 Schematical rep-
resentation of the DuFort-
Frankel method (7.12).
t
j1
q
x
i1
u
u u
6
x
i
q
x
i+1
t
j
Q
Q
Q
Qk

3
t
j+1
u
Fig. 7.4 Schematical rep-
resentation of the implicit
BTCS method (7.13).
e u u
u
6
t
j
Q
Q
Q
Qk

3
t
j+1
method. Indeed, consider the Taylor expansion for Eq. (7.3) by means of (7.12):
u
j+1
i
u
j1
i
2t
= D
u
j
i+1
u
j+1
i
u
j1
i
+u
j
i1
x
2

u
t

x
2
3!
u
ttt
+. . . =
D
x
2

x
2
u
xx
+
2x
4
4!
u
xxxx
t
2
u
tt

2t
4
4!
u
tttt
+. . .

u
t
+O(t
2
) = Du
xx
+O(x
2
) D

t
2
x
2

u
tt
+O

t
4
x
2

.
In other words, the method (7.12) has order of accuracy
O

t
2
, x
2
,
t
2
x
2

.
For cosistency, t/x 0 as t 0 and x 0, so (7.12) is inconsistent. This
constitutes an effective restriction on t. For large t, however, the scheme (7.12)
is consistent with another equation of the form
Du
tt
+u
t
= Du
xx
.
7.1.2.1 The BTCS Implicit Method
One can try to overcome problems, described above by introducing an implicit
method. The simplest example is a BTCS (backward in time, central in space)
method (see Fig. 7.4) [29]. The differential schema reads:
u
j+1
i
u
j
i
t
= D
u
j+1
i+1
2u
j+1
i
+u
j+1
i1
x
2
+O(t, x
2
),
Fig. 7.5 Schematical rep-
resentation of the Crank-
Nicolson method (7.14).
t
j1
q
x
i1
u u u
x
i
q
x
i+1
t
j+
1
2
t
j+1
u u u
or, with = Dt/x
2
u
j
i
= u
j+1
i+1
(1 +2)u
j+1
i
+u
j+1
i1
. (7.13)
In this case the amplication factor g(k) is given by
g(k) =

1 +4sin
2
kx
2
2

1
.
That is, the schema (7.13) is unconditionally stable. However, the method has order
of accuracy O(t, x
2
), i.e., rst order in time, and second in space. Is it possible
to improve it? The answer to is given below.
The Crank-Nicolson Method
An implicit scheme, introduced by J. Crank and P. Nicolson in 1947 [6] is based on
the central approximation of Eq. (7.3) at the point (x
i
, t
j
+
1
2
t) (see Fig. (7.5)):
u
j+1
i
u
j
i
2
t
2
= D
u
j+
1
2
i+1
2u
j+
1
2
i
+u
j+
1
2
i1
x
2
.
The approximation used for the space derivative is just an average of approxima-
tions in points (x
i
, t
j
) and (x
i
, t
j+1
):
u
n+1
i
u
n
i
t
= D
(u
n+1
i+1
2u
n+1
i
+u
n+1
i1
) +(u
n
i+1
2u
n
i
+u
n
i1
)
2x
2
.
Introducing = Dt/x
2
one can rewrite the last equation as
u
j+1
i+1
+2(1 +)u
j+1
i
u
j+1
i1
= u
n
j+1
+2(1 )u
j
i
+u
n
j1
. (7.14)
All terms on the right-hand side of Eq. (7.14) are known. Hence, the equations
in (7.14) form a tridiagonal linear system
Au = b.
Fig. 7.6 Contour plot of the
heat distribution after the time
T = 30, calculated with the
FTCS method (7.9).
x
t


0 0.5 1
0
5
10
15
20
25
30
0
0.2
0.4
0.6
0.8
The amplication factor for Eq. (7.14) reads
g(k) =
1 (1 coskx)
1 +(1 coskx)
.
Since and 1 cosk x are positive, the denominator of the last expression is
always greater than the numerator. That is, the absolute value of g is less than one,
i.e., the method (7.14) is unconditionaly stable.
7.1.3 Examples
Example 1
Use the FTCS explicit method (7.9) to solve the one-dimensional heat equation
u
t
= u
xx
,
on the interval x [0, L], if the initial heat distribution is given by u(x, 0) = f (x),
and the temperature on both ends of the interval is u(0, t) = T
l
, u(L, t) = T
r
. Other
parameters are choosen according to the table below:
Space interval L = 1
Amount of space points M = 10
Amount of time steps T = 30
Boundary conditions T
l
= T
r
= 0
Initial heat distribution f (x) = 4x(1 x)
The result of the calculation is shown on Fig 7.6.
Fig. 7.7 Contour plot of the
diffusion of the initial Gauss
pulse, calculated with the
BTCS implicit method (7.13).
x
t


5 0 5
0
50
100
150
200
0.2
0.4
0.6
0.8
Example 2
Use the implicit BTCS method (7.13) to solve the one-dimensional diffusion equa-
tion
u
t
= u
xx
,
on the interval x [L, L], if the initial distribution is a Gauss pulse of the
form u(x, 0) = exp(x
2
) and the density on both ends of the interval is given as
u
x
(L, t) = u
x
(L, t) = 0. For the other parameters see the table below:
Space interval L = 5
Space discretization step x = 0.1
Time discretization step t = 0.05
Amount of time steps T = 200
Solution of the problem is shown on Fig. (7.7).
Example 3
Use the Crank-Nicolson method (7.14) to solve the one-dimensional heat equation
u
t
= 1.44u
xx
,
on the interval x [0, L], if the initial heat distribution is u(x, 0) = f (x) and again,
the temperature on both ends of the interval is given as u(0, t) = T
l
, u(L, t) = T
r
.
Other parameters are choosen as:
Space interval L=1
Space discretization step x = 0.1
Time discretization step t = 0.05
Amount of time steps T = 15
Boundary conditions T
l
= 2, T
r
= 0.5
Initial heat distribution f (x) = 2 1.5x +sin(x)
Fig. 7.8 Contour plot of the
heat distribution, calculated
with the Crank-Nicolson
method (7.14).
x
t


0 0.5 1
0
5
10
15
0.5
1
1.5
2
Numerical solution of the problem in question is shown on Fig. (7.8).
7.2 The Diffusion Equation in 2D
Let us consider the solution of the diffusion equation (7.2) in two dimensions
u
t
= D

2
u
x
2
+

2
u
y
2

, (7.15)
where u = u(x, y, t), x [a
x
, b
x
], y [a
y
, b
y
]. Suppose, that the inittial condition is
given and function u satises boundary conditions in both x- and in y-directions.
As before, we discretize in time on the uniform grid t
n
= t
0
+nt, n = 0, 1, 2, . . ..
Futhermore, in the both x- and y-directions, we use the uniform grid
x
i
= x
0
+i x, i = 0, , M, x =
b
x
a
x
M+1
,
y
j
= y
0
+ j y, j = 0, , N, y =
b
y
a
y
N +1
.
7.2.1 Numerical Treatment
The FTCS Method in 2D
In the case of two dimensions the explicit FTCS scheme reads
u
n+1
i j
u
n
i j
t
= D

u
n
i+1 j
2u
n
i j
+u
n
i1 j
x
2
+
u
n
i j+1
2u
n
i j
+u
n
i j1
y
2

,
or, with = Dt/x
2
and = Dt/y
2
u
n+1
i j
= (u
n
i+1 j
+u
n
i1 j
) +(u
n
i j+1
+u
n
i j1
) +(1 22)u
n
i j
. (7.16)
The ansatz

n
i j
= g
n
e
i(k
x
x
i
+k
y
x
j
)
leads to the following relation for the amplication factor g(k)
g(k) = 1 4sin
2

k
x
x
2

4 sin
2

k
y
y
2

.
In this case the stability condition reads
+
1
2
. (7.17)
This stability condition imposes a limit on the time step:
t
x
2
y
2
2D(x
2
+y
2
)
.
In particular, for the case x =y we have
t
x
2
4D
,
which is even more restrictive, than in the one-dimensional case.
The BTCS Method in 2D
To overcome the stability restriction of the FTCS method (7.16), we can use an
implicit BTCS schema in the two-dimensional case. The schema reads:
u
n+1
i j
u
n
i j
t
= D

u
n+1
i+1 j
2u
n+1
i j
+u
n+1
i1 j
x
2
+
u
n+1
i j+1
2u
n+1
i j
+u
n+1
i j1
y
2

,
or
(u
n+1
i+1 j
+u
n+1
i1 j
) +(1 +2+2)u
n+1
i j
(u
n+1
i j+1
+u
n+1
i j1
) = u
n
i j
. (7.18)
Let us consider the approximation (7.18) on the 5 5 grid, i.e., i = j = 0, . . . , 4.
Moreover, suppose that Dirichlet boundary conditions are given, that is, all values
u
0 j
, u
4 j
, u
i0
, u
i4
are known. Suppose also that n = 1 and dene = 1 +2 +2.
Then the approximation above leads to the neun algebraic equations:
u
2
21
+u
2
11
u
2
12
= u
1
11
+u
2
01
+u
2
10
,
u
2
22
+u
2
12
(u
2
13
+u
2
11
) = u
1
12
+u
2
02
,
u
2
23
+u
2
13
u
2
12
= u
1
13
+u
2
03
+u
2
14
,
(u
2
31
+u
2
11
) +u
2
21
u
2
22
= u
1
21
+u
2
20
,
(u
2
32
+u
2
12
) +u
2
22
(u
2
23
+u
2
21
) = u
1
22
,
u
2
21
+u
2
31
u
2
32
= u
1
31
+u
2
41
+u
2
30
,
u
2
22
+u
2
32
(u
2
33
+u
2
31
) = u
1
32
+u
1
42
,
u
2
23
+u
2
33
u
2
32
= u
1
33
+u
2
44
+u
2
34
.
Formally, one can rewrite the system above to the matrix form Au = b, i.e.,

0 0 0 0 0 0
0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0
0 0 0 0
0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0
0 0 0 0 0 0

u
2
11
u
2
12
u
2
13
u
2
21
u
2
22
u
2
23
u
2
31
u
2
32
u
2
33

u
1
11
+u
2
01
+u
2
10
u
1
12
+u
2
02
u
1
13
+u
2
03
+u
2
14
u
1
21
+u
2
20
u
1
22
u
1
23
+u
2
24
u
1
31
+u
2
41
+u
2
30
u
1
32
+u
2
42
u
1
33
+u
1
44
+u
2
34

The matrix A is a ve-band matrix. Nevertheless, despite of the fact that the schema
is absolute stable, two of ve bands are desposed so far apart fromthe main diagonal,
that simple O(n) algorithms like TDMA are difcult or even impossible to apply.
The ADI Method
The idea of the ADI-method (alternating direction implicit) is to alternate direction
and thus solve two one-dimensional problem at each time step [17]. The rst step
keeps y-direction xed:
u
n+1/2
i j
u
n
i j
t/2
= D

u
n+1/2
i+1 j
2u
n+1/2
i j
+u
n+1/2
i1 j
x
2
+
u
n
i j+1
2u
n
i j
+u
n
i j1
y
2

.
In the second step we keep x-direction xed:
u
n+1
i j
u
n+1/2
i j
t/2
= D

u
n+1/2
i+1 j
2u
n+1/2
i j
+u
n+1/2
i1 j
x
2
+
u
n+1
i j+1
2u
n+1
i j
+u
n+1
i j1
y
2

.
Both equations can be written in a triadiagonal form. Dene
=
Dt
2x
2
, =
Dt
2y
2
.
Than we get:
u
n+1/2
i+1 j
+(1 +2)u
n+1/2
i j
u
n+1/2
i1 j
= u
n
i j+1
+(1 2)u
n
i j
+u
n
i j1
(7.19)
u
n+1
i j+1
+(1 +2)u
n+1
i j
u
n+1
i j1
= u
n+1/2
i+1 j
+(1 2)u
n+1/2
i j
+u
n+1/2
i+1 j
.
Instead of ve-band matrix in BTCS method (7.18), here each time step can be
obtained in two sweeps. Each sweep can be done by solving a tridiagonal system
of equations. The ADI-method is second order in time and space and is absolute
stable [11] (however, the ADI method in 3D is conditional stable only).
7.2.2 Examples
Use the ADI method (7.19) to solve the two-dimensional diffusion equation

t
u(r, t) =u(r, t),
where u = u(r, t), r R
2
on the interval r [0, L] [0, L], if the initial distribution
is a Gauss pulse of the form u(x, 0) = exp(20(x L/2)
2
20(y L/2)
2
) and
the density on both ends of the interval is given as u
r
(0, t) = u
r
(L, t) = 0. Other
parameters are choosen according to the table below.
Space interval L = 1
Amount of points M = 100, (x =y)
Time discretization step t = 0.001
Amount of time steps T = 40
Solution of the problem is shown on Fig. (7.9).
(a) (b)
x
y


0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
x
y


0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
(c) (d)
x
y


0 0.5 1
0
0.2
0.4
0.6
0.8
1
x
y


0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
Fig. 7.9 Numerical solution of the two-dimensional diffusion equation 7.2 by means of the ADI
method (7.19), calculated at four different time moments: (a) t=0; (b) t=10; (c) t=20; (d) t=40.