Paper ENCN304 – Deterministic Mathematical Methods
Topic Partial Differential Equations
Module 7 – Steady Diffusion
Summary
This module introduces you to Laplace’s equation and solves it using the separation
of variables technique in both rectangular and polar coordinates. Fundamental
solutions of Laplace’s equation, obtained from analytic functions of a complex
variable, are also presented.
Prerequisites
In order to be able to use the separation of variables technique you must be familiar
with Fourier series (Module 2), boundary-value problems (ODEs) and have seen the
application of the technique to the wave equation in Module 3 or diffusion equation in
module 6. You should have completed the Introduction to Diffusion in Module 4 and
have a fundamental grounding in complex number theory.
Objectives
At the conclusion of this module you should:
1. Be able to write down Laplace’s equation with the appropriate boundary
conditions that model a particular steady diffusion problem.
2. Be able to solve Laplace’s equation in rectangular or polar coordinates using
separation of variables.
3. Be able to generate two fundamental solutions of Laplace’s equation from any
analytic function of a complex variable.
References
ENCN304 PDEs – Module 7 Summary 1 Copyright © Dr Roger Nokes 2013
Key Points
1. Laplace’s Equation
Laplace’s equation, which arises in steady state diffusion problems (ie independent of
time), in two dimensions is given by
2u 2u
2u 0 (1)
x 2 y 2
in rectangular coordinates, and
2u 1 u 1 2u
u 2
2
0 (2)
r r r r 2 2
in polar coordinates.
2. Separation of Variables in a Rectangular Domain
Separation of variables can be used to solve Laplace’s equation in a rectangular
domain provided the boundary conditions on two opposite sides are homogeneous.
The solution method is as follows:
1. Assume a product solution X(x)Y(y). This will lead to 2 ODEs for X and Y.
One of these will have sine and cosine solutions, the other cosh and sinh. You must
choose the sign of the separation constant so that X has sine and cosine solutions if the
homogenous boundary conditions lie in the x direction, or so that Y has sine and
cosine solutions if the homogeneous boundary conditions lie in the y direction.
2. Use the homogeneous boundary conditions to find the eigenvalues and
eigenfunctions. Then add all possible solutions together in order to obtain a general
solution.
3. Use the remaining boundary conditions to find the coefficients in the general
solution. This will normally involve calculating two Fourier series.
3. Separation of Variables in Polar Coordinates
The separation of variables method can be applied equally well to Laplace’s equation
in polar coordinates. In this case the domain could be a segment of a circle, a whole
circle, or maybe a segment of a two-dimensional donut. In each case, if you draw the
domain in r, coordinates, it will have the shape of a rectangle. For separation of
variables to work the boundary conditions must be homogeneous along the =
constant boundaries (ie along the lines emanating from the origin) unless the domain
is a whole circle (in which case no such boundaries exist).
The method is much the same as above
ENCN304 PDEs – Module 7 Summary 2 Copyright © Dr Roger Nokes 2013
1. Assume a product solution R(r)(). This will lead to two ODEs in R and
that have the solutions
Acos Bsin (3)
R(r) Cr Dr 0 (4)
and
R(r) C Dlnr 0 (5)
2. Apply the homogenous boundary conditions to find the eigenvalues and
eigenfunctions. If the domain is a full circle no boundary conditions are available and
the eigenvalues are n 0,1... . Add together all possible solutions to find a general
solution.
3. Apply the remaining boundary conditions to find the coefficients in the general
solution. This may involve calculating two Fourier series.
[Note: If the origin is included in the solution domain only one boundary condition
in the r direction will be given, and the solutions Dr and Dlnr will be excluded as
they are unbounded at r = 0.]
4. Fundamental solutions of Laplace’s equation
Analytic functions of a complex variable, z, possess special properties. If
w f (z) u(x, y) iv(x, y) (6)
then u and v must satisfy the Cauchy-Riemann equations, which can be written as
u v u v
and (7)
x y y x
It follows from these equations that both u and v must satisfy Laplace’s equation, and
these functions are called fundamental solutions.
The functions u and v are related in that their isocurves (ie curves in the xy plane
along which u or v are constant) are everywhere perpendicular to each other. In the
study of inviscid/irrotational fluid flow the functions u and v can be identified as the
velocity potential and streamfunction respectively.
ENCN304 PDEs – Module 7 Summary 3 Copyright © Dr Roger Nokes 2013
Worked Examples
Question 1
The steady flow of groundwater in a confined aquifer is governed by Laplace’s
equation
2 h 2 h
0
x 2 y 2
where h is the piezometric head. The confined aquifer lying between the Rakaia and
Ashburton rivers in South Canterbury is modelled as a rectangle of width 50kms (the
distance between the rivers) and 100kms long. The head on the river and sea
boundaries is approximately 0m while that on the inland boundary it is 40m.
(a) Draw the aquifer and indicate the boundary conditions.
(b) Calculate the head distribution in the aquifer (use separation of variables).
Solution
(a) The flow domain is a rectangle of width 50km and length 100km. Lets choose
the rectangle to lie between x = 0 and x = 100, and y = 0 and y = 50. So the
Ashburton river is the y = 50 boundary, the Rakaia river is the y = 0 boundary, the sea
is the x = 0 boundary and the inland boundary is at x = 100.
Figure 1. Problem domain and boundary conditions.
(b) We can use separation of variables to find h(x,y) as the domain is rectangular
and the boundary conditions are homogeneous in the y direction.
Step 1: Reduce to 2 ODEs and Solve
Again we assume a product solution of the form
ENCN304 PDEs – Module 7 Summary 4 Copyright © Dr Roger Nokes 2013
h(x,t) X(x)Y (y)
Substituting into Laplace’s equation we get
d2 X d 2Y
Y 2 X 2 0
dx dx
or, after dividing through by XY we get
1 d2 X 1 d 2Y
2
2
2
X dx Y dy
Now it is important to get the sign of the separation constant right. In this problem, as
the homogenous boundary conditions are in the Y direction we must ensure that the Y
solution includes sines and cosines. The two ODEs are
d2 X
2
2 X 0
dx
d 2Y
2
2Y 0
dx
with solutions
X(x) A cosh x Bsinh x
Y (y) C cos y Dsin y
Step 2: Apply the boundary conditions to obtain a general solution
Now we apply the homogeneous boundary conditions in order to find the eigenvalues
and the eigenfunctions.
h(x,0) X(0)Y (0) 0 so Y (0) 0
h(x,50) X(0)Y (50) 0 so Y (50) 0
So
Y (0) 0 C cos0 Dsin 0 C 0
Y (50) 0 Dsin 50 0 50 n n 1,2,...
Therefore
n
n n 1, 2,....
50
are the eigenvalues, and
ENCN304 PDEs – Module 7 Summary 5 Copyright © Dr Roger Nokes 2013
n y
Yn (y) Bn sin
50
are the eigenfunctions.
Now the general solution is the sum of all of these solutions
n y n x n x
h(x, y) sin an cosh bn sinh
n1 50 50 50
Step 3: Apply the remaining boundary conditions to obtain a particular
solution
Finally we apply the two remaining boundary conditions to find the coefficients an
and bn.
So
n y
h(0, y) an sin 0 an 0
n1 50
The second boundary condition is
n 100 n y
h(100, y) bn sinh sin 40
n1 50 50
To make this look more like a Fourier series we define
bn ' bn sinh 2n
so the boundary condition becomes
n y
b 'sin
n
50
40
n1
We have a sine series so we extend the boundary condition as an odd function over
the complete interval [-50,50] and use the half-range formula to find the bn’s
ENCN304 PDEs – Module 7 Summary 6 Copyright © Dr Roger Nokes 2013
n y
50
2
bn '
50 0
40 sin
L
dy
50
80 50 n y
cos
50n 50 0
80
1 cos n
n
0 n even
160
n odd
n
So
bn '
bn
sinh 2n
and the final solution is
160
1 (2n 1) y (2n 1) x
h(x, y)
2n 1sinh 2 (2n 1)sin 50
sinh
50
n 0
Question 2
Find the steady state temperature distribution in a semi-circular plate of radius 1m,
when the flat edge is held at 0 degrees and the curved edge is held at T0 ( ) .
Solution
We can use separation of variables in polar coordinates as the domain is a segment of
a circle, and the boundary conditions along the boundaries emanating from the origin
are homogeneous. In this problem it looks like a single boundary along the x axis,
but in reality one half of this boundary corresponds to , and the other to .
Laplace’s equation is
2u 1 u 1 2u
0
r 2 r r r 2 2
and the boundary conditions
u(r, 0) 0
u(r, ) 0
u(1, ) T0 ( )
and at the origin we just note that the temperature must be finite.
ENCN304 PDEs – Module 7 Summary 7 Copyright © Dr Roger Nokes 2013
Step 1: Reduce to 2 ODEs and Solve
Assume
u(r, ) R(r)( )
so
d2R 1 dR 1 d 2
R 0
dr 2 r dr r 2 d 2
R
Now divide through by . So
r2
r 2 d 2 R r dR 1 d 2
2
R dr 2
R dr d 2
and we choose the sign of the separation constant so that is sinusoidal.
So the two governing ODEs are
d 2
2 0
d 2
d2R dR 2
r2 2
r R0
dr dr
The solutions to these two ODEs are
Acos Bsin
R(r) Cr Dr 0
and
R(r) C Dlnr 0
Step 2: Apply the homogeneous boundary conditions to obtain a general
solution
The homogeneous boundary conditions are in the angular direction so
u(r,0) R(r)(0) 0 (0) 0
u r, R(r) 0 0
So
ENCN304 PDEs – Module 7 Summary 8 Copyright © Dr Roger Nokes 2013
(0) = A = 0
= Bsin = 0 = n n = 1, 2,...
Thus, the eigenvalues are
n n n 1, 2,....
and the eigenfunctions are
n (x) Bn sinn
are eigenfunctions. The general solution is
u(r, ) cn r n dn r n sin n
n1
Step 3: Apply the remaining boundary conditions to obtain a particular
solution
Firstly we note that all of the coefficients dn must be 0, as the functions they multiply
are unbounded at the origin. The second boundary condition, at r = 1, can be used to
find the other set of coefficients.
So
u(1, ) cn 1n sin n cn sin n T0 ( )
n1 n1
We extend the function u(1,) to be odd in the interval [-,0] so that we can represent
it as a sine series. Thence we can calculate the coefficients using the half-range
formula
T - sinn d
2
cn =
0
0
= .... (integration by parts)
1- cosn
4T0
=
n 3
= 0 n even
8T0
n odd
n 3
Finally the particular solution is
8T0 1
u(r, )
(2n 1) 3
r 2n1 sin(2n 1)
n1
ENCN304 PDEs – Module 7 Summary 9 Copyright © Dr Roger Nokes 2013
Question 3
Find the steady state voltage in a circular plate, of radius R, if the edge of the plate has
an imposed voltage given by
V (R, ) 2cos4
The voltage in such a plate is governed by Laplace’s equation.
Solution
We can use separation of variables in polar coordinates as the domain is a circle. If we
drew the domain in the r plane we would see that we had only one boundary
condition along the r = R boundary. However, we know that the function at = 0
must be the same as the function at = 2 as these are the same physical points. Thus
we know that the solution must be periodic with period 2.
Laplace’s equation is
2V 1 V 1 2V
0
r 2 r r r 2 2
and the boundary condition is
V (R, ) 2cos4
Step 1: Reduce to 2 ODEs and Solve
We assume
V (r, ) R(r)( )
and following the steps in the previous question we arrive at the following solutions
for the unknown functions and R.
Acos Bsin
R(r) Cr Dr 0
and
R(r) C lnr D 0
Step 2: Apply the homogeneous boundary conditions to obtain a general
solution
As we have already mentioned there are no boundary conditions in the direction that
will enable us to compute the eigenvalues and eigenfunctions. However we do have
the requirement that the solution must be periodic in with period 2. Thus the
eigenvalues must be
ENCN304 PDEs – Module 7 Summary 10 Copyright © Dr Roger Nokes 2013
n n n 0, 1, 2,....
and the eigenfunctions are
n (x) An cos n Bn sin n .
Now we know that, as the origin is included in the solution domain, we cannot have
negative powers of r in the solution, nor can we have a ln term. Thus only the positive
powers of r remain and the general solution is
V (r, ) a0 r n an cos n bn sin n
n1
Step 3: Apply the remaining boundary conditions to obtain a particular
solution
Finally we must apply the boundary condition on the edge of the plate to determine
the unknown coefficients. So
V (R, ) R n an cos n bn sin n 2cos 4
n0
We could go through the formal calculation of the Fourier coefficients but in this case
we can simply read off the answer. The right hand side of this equation is also a
Fourier series but it only has one term in it. Thus only the corresponding term in the
middle expression needs to be retained. So equating terms we have
an 0 n4
2
a4
R4
bn 0 for all n
and the final solution is
4
r
V (r, ) 2 cos 4
R
Question 4
Find two fundamental solutions to Laplace’s equation from each of the following
analytic complex functions. In each case clearly identify whether the solutions are
valid throughout the xy plane, and sketch some iso-curves for each function.
(a) w f (z) ln z
ENCN304 PDEs – Module 7 Summary 11 Copyright © Dr Roger Nokes 2013
(b)
w f (z) U iV z
1
(c) w f (z)
z
Solution
(a) We proceed by expanding the function and solving for u and v as functions of x
and y.
w u iv ln z
ln rei
ln r i
Thus
u(r, ) ln(r) or u(x, y) ln x y
2 2
and
y
v(r, ) or v(x, y) tan 1
x
The function f(z) is undefined at the origin so this point is excluded from the domain.
The isocurves are given by the curves r = constant, and = constant. These are simply
circles centred on the origin and straight lines that pass through the origin.
(b) We proceed by expanding the function and solving for u and v as functions of x
and y.
w u iv U iV z
U iV (x iy)
Ux Vy i(Vx Uy)
Thus
u(x, y) Ux Vy
and
v(x, y) Vx Uy
ENCN304 PDEs – Module 7 Summary 12 Copyright © Dr Roger Nokes 2013
The function f(z) is analytic everywhere so the whole xy plane is included in the
domain.
The isocurves are given by the curves Ux Vy C , and Vx Uy D . These are sets
of parallel, straight lines where the two sets are normal to one another.
(c) We proceed by expanding the function and solving for u and v as functions of x
and y.
1
w u iv
z
1
x iy
x iy
x2 y2
cos sin
i
r r
Thus
cos
u(r, )
r
and
sin
v(r, )
r
The function f(z) is analytic everywhere except at the origin so this point must be
excluded from the domain.
cos sin
The isocurves are given by the curves C , and D . These curves are a
r r
little tricky. Take the first set of curves. Let’s assume C is a positive constant. As r is
always positive that means cos must be positive, and this is only true to the right of
the y axis. When equals 0 r must be equal to 1/C. Now as increases or decreases
from 0 cos decreases and thus the value of r required to keep the ratio fixed must
decrease. As the isocurve approaches the y axis cos goes to 0 and therefore so must
r. Increasing or decreasing the value of C leads to larger or smaller curves. The curves
for all values of C less than a value C1 will lie within the C1 isocurve.
Now if C is negative the curve looks the same as when C is positive except it is now
on the left hand side of the y axis.
Using the same arguments it easy to see the second set of isocurves look exactly the
same as the first set except they lie above and below the x axis. Have a go at sketching
these isocurves.
ENCN304 PDEs – Module 7 Summary 13 Copyright © Dr Roger Nokes 2013
Question 5
Ideal fluid flow is governed by Laplace’s equation, and the two functions u and v have
special meanings in this case. The function v is the streamfunction, which has the
special property that the fluid flows along lines of constant streamfunction. In other
words the velocity vector is parallel to the isocurves of v. To resolve the ambiguity of
direction, the velocity vector always points in the direction of increasing u, the
velocity potential.
(a) Consider the streamfunction and velocity potential generated in Q4(a). Describe
the type of fluid flow they represent.
(b) What would be the effect of multiplying f(z) by –1?
(c) Describe the flow generated in Q4(b).
Solution
(a) We can find the streamlines by considering the isocurves of v. These curves are
all straight lines meeting at the origin. From this we do not know whether the fluid is
flowing towards the origin, or away from it. If we look at the velocity potential, we
see that it increases with r, the distance from the origin, and therefore the fluid must
be flowing away from the origin. This flow would appear to correspond to a source of
fluid at the origin. Clearly no fluid flows towards the origin and therefore fluid must
be being created there.
(b) If we change the sign of f(z) we change the signs of u and v. The streamlines
don’t change (although the value of the streamfunction on them does) but the
direction of flow does because now u is decreasing with increasing r. Thus we now
have a fluid sink at the origin.
(c) The flow in this case is very simple. The fluid is all flowing along the lines
Vx Uy const . The direction will depend on the signs of U and V. If U is positive
then the fluid is flowing in the positive x direction and if V is positive it is flowing in
the positive y direction. The ratio of the x and y velocities is U/V. In fact U and V are
the velocities in the x and y directions. Thus this streamfunction represents a uniform
flow at an angle to the x axis.
ENCN304 PDEs – Module 7 Summary 14 Copyright © Dr Roger Nokes 2013