MST209 Mathematical methods

and models

Partial differential equations

�


 ����

Study guide
The recommended study pattern is to study the sections in numerical order
over five study sessions. However, the models developed in Section 1 are
solved in Section 3, and the models developed in Section 2 are solved in
Section 4, so you may prefer to study Section 3 straight after Section 1, and
Section 4 straight after Section 2. Section 3 contains the main technique of
the unit and is rather longer than the other sections.
You will need your computer for Section 5, which involves using the com-
puter algebra package for the course as a tool for solving and interpreting
the results from the problems developed earlier in the unit.
Partial differentiation was introduced in Unit 12, and the techniques in
Sections 3 and 4 are based on ideas from Units 3 and 21. The discussion of
diffusion in Sections 2 and 4 uses ideas from Unit 15.

This publication forms part of an Open University course. Details of this and other Open Uni-
versity courses can be obtained from the Student Registration and Enquiry Service, The Open
University, PO Box 625, Milton Keynes, MK7 6YG, United Kingdom: tel. +44 (0)1908 653231,
e-mail general-enquiries@open.ac.uk
Alternatively, you may visit the Open University website at http://www.open.ac.uk where you can
learn more about the wide range of courses and packs offered at all levels by The Open University.
To purchase a selection of Open University course materials, visit the webshop at www.ouw.co.uk,
or contact Open University Worldwide, Michael Young Building, Walton Hall, Milton Keynes,
MK7 6AA, United Kingdom, for a brochure: tel. +44 (0)1908 858785, fax +44 (0)1908 858787,
e-mail ouwenq@open.ac.uk

The Open University, Walton Hall, Milton Keynes, MK7 6AA.

First published 2005.
Copyright

c 2005 The Open University
All rights reserved; no part of this publication may be reproduced, stored in a retrieval system,
transmitted or utilised in any form or by any means, electronic, mechanical, photocopying, record-
ing or otherwise, without written permission from the publisher or a licence from the Copyright
Licensing Agency Ltd. Details of such licences (for reprographic reproduction) may be obtained
from the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP.
Open University course materials may also be made available in electronic formats for use by
students of the University. All rights, including copyright and related rights and database rights,
in electronic course materials and their contents are owned by or licensed to The Open University,
or otherwise used by The Open University as permitted by applicable law.
In using electronic course materials and their contents you agree that your use will be solely for
the purposes of following an Open University course of study or otherwise as licensed by The
Open University or its assigns.
Except as permitted above you undertake not to copy, store in any medium (including electronic
storage or use in a website), distribute, transmit or re-transmit, broadcast, modify or show in
public such electronic materials in whole or in part without the prior written consent of The Open
University or in accordance with the Copyright, Designs and Patents Act 1988.
Edited, designed and typeset by The Open University, using the Open University TEX System.
Printed and bound in the United Kingdom by Martins the Printers Ltd, Berwick-upon-Tweed,
TD15 1RS.
ISBN 0 7492 6886 7
1.1
Contents

Introduction 4

1 Modelling using the wave equation 6

1.1 The taut string 6

1.2 Adding damping 10

2 Modelling using the diffusion equation 11

2.1 The insulated rod 11

2.2 The convecting rod 13

3 Separating the variables 14

3.1 Normal modes 15

3.2 The initial displacement 18

3.3 The damped plucked string 20

4 Solving the heat transfer problems 26

4.1 The insulated rod problem solved 26

4.2 The convecting rod problem solved 29

Outcomes 33

Solutions to the exercises 34

Index 38

Introduction

Introduction

A partial differential equation is an equation relating a dependent vari-

able and two or more independent variables through the partial derivatives Partial derivatives were
of the dependent variable. Differential equations have played a very impor- introduced in Unit 12.
tant role in the course so far. But until now, all the differential equations
you have met have involved just one independent variable, and have been
equations containing one or more dependent variables and their ordinary
derivatives with respect to that independent variable. Such equations are
often called ordinary differential equations when it is necessary to distin- Ordinary differential
guish them from partial differential equations. For many systems that we equations are the subject of
want to be able to model, ordinary differential equations are inadequate be- Units 2, 3, 11 and 13.
cause the states of the system can be specified only in terms of two — or
even more — independent variables. When we are trying to model the way
in which such a system changes, we are inevitably led to consider partial
differential equations.
One example of such a system is a plucked guitar string. One of the main
topics of this unit is the derivation and analysis of a mathematical model
for a taut string, such as a guitar string, plucked at some point along its
length. A partial differential equation is required because the state of the
string — by which is meant its shape at any given time after it has been
plucked — requires a function u(x, t) of two independent variables, x and t,
where x is the distance along the straight line joining the two points at The straight line joining the
which the string is anchored (which we can consider as an axis with origin two points at which the string
at one end of the string), and t is the time since the string was plucked. The is anchored is the equilibrium
position of the string.
dependent variable u = u(x, t) is the sideways displacement of the string
from the point on the axis determined by x, at time t. For fixed t = t1
and varying x, u(x, t1 ) specifies the shape of the string at time t1 ; for fixed
x = x1 and varying t, on the other hand, u(x1 , t) tells how the sideways
displacement of the string from the fixed point x = x1 on the axis varies
with time (see Figure 0.1).
The model of the motion of the string that we shall develop will be a dif-
ferential equation for the variable u. Since u depends on both x and t, an
equation that models the motion of the string will involve partial derivatives
of u with respect to both x and t. One such equation is
∂2u 1 ∂2u
= , (0.1)
∂x2 c2 ∂t2
where c is a constant whose value depends on various physical characteristics Figure 0.1
of the string. This partial differential equation is called the wave equation. The name wave equation is
It is a very important equation of mathematical physics, in part because it used because it models
occurs in many situations that involve vibrations of extended flexible objects wave-like motions such as
that of a plucked guitar
like strings and springs. string.
The wave equation is derived in Section 1, in the context of modelling the
vibrations of a taut string (such as a guitar string). Another partial differ-
ential equation, the diffusion equation, is developed in Section 2. Both are
examples of second-order partial differential equations, in that the highest-
order derivative that occurs is second order. Just as is the case for ordinary
differential equations, the order of a partial differential equation is the or-
der of the highest derivative that occurs in it. In this unit we deal only with
second-order partial differential equations. Furthermore, all the equations
that we deal with are linear, in that they contain no products of terms
involving the dependent variable and its partial derivatives.

4
 Introduction

Sections 3, 4 and 5 are concerned with solving examples of the wave equation
and the diffusion equation. We shall not be able to obtain general solutions
to these equations, but we shall be able to obtain a whole host of particular
solutions, based on the imposition of appropriate initial conditions and/or
boundary conditions. The techniques that we use to solve these two specific
types of partial differential equation illustrate some of the techniques that
can be applied to partial differential equations in general. However, here
we shall not attempt to generalize beyond solutions to the wave equation
and the diffusion equation. Section 5 shows how problems involving these
equations can readily be solved with the aid of the computer.
Checking whether or not a given function is a solution of a given partial
differential equation is simply a matter of substituting the given function into
the given equation, and seeing whether it is satisfied. The only difference
from the case of an ordinary differential equation is that you have to calculate
all the relevant partial derivatives. For example, let us check that
u(x, t) = sin(kx) cos(kct)
is a solution of the wave equation (0.1) for any constant k. We have
∂u ∂2u
= k cos(kx) cos(kct), = −k 2 sin(kx) cos(kct),
∂x ∂x2
∂u ∂2u
= −kc sin(kx) sin(kct), = −k 2 c2 sin(kx) cos(kct);
∂t ∂t2
so, when u(x, t) = sin(kx) cos(kct),
∂2u 1 ∂2u
=
∂x2 c2 ∂t2
and this function is indeed a solution of the wave equation.
The initial conditions and/or boundary conditions appropriate to obtaining The boundary conditions are
a particular solution of the wave equation or the diffusion equation depend conditions where the spatial
on the context. For example, for the wave equation as a model of the variable x is fixed.
vibrations of a guitar string, we can use two boundary conditions and two The initial conditions are
conditions where the time
initial conditions. If L is the equilibrium length of the string, then the
variable t is fixed (usually at
boundary conditions are zero).
u(0, t) = u(L, t) = 0, t ≥ 0; (0.2)
these conditions correspond to the string being fixed at its ends. The initial
conditions model the action of plucking the string, which sets it in motion.
Plucking consists of holding the string in a certain shape, at rest, and then
releasing it. If the initial shape of the string is given by a function f (x), the
initial conditions may be specified in the form
u(x, 0) = f (x), 0 < x < L, We use the range 0 < x < L
∂u for the initial condition
(x, 0) = 0, 0 ≤ x ≤ L. u(x, 0) = f (x) because the
∂t values of u(x, 0) when x = 0
The first initial condition models the initial shape of the string, while the and x = L are specified by
second corresponds to it being at rest initially. boundary conditions (0.2)
when t = 0.
In order to obtain particular solutions, we shall need to make use of Fourier
series, introduced in Unit 21, and of some ideas from Unit 3. In particu-
lar, we shall need to apply the principle of superposition, which applies to
linear second-order partial differential equations in the same way as to lin-
ear second-order ordinary differential equations. The equations studied in
this unit are all homogeneous, in that each additive term involves the
dependent variable or its derivatives — there are no constant terms or
terms involving solely the independent variables. However, the boundary

5
Section 1 Modelling using the wave equation

and initial conditions may be homogeneous (e.g. u(0, t) = u(L, t) = 0 and

(∂u/∂t)(x, 0) = 0) or inhomogeneous (e.g. u(x, 0) = f (x), where f is not
the zero function). The principle of superposition says that if u1 and You will be asked to verify
u2 are both solutions of a homogeneous linear partial differential equation, that this version of the
then so is any linear combination of u1 and u2 . Furthermore, if u1 and u2 principle of superposition
applies to the wave equation
both satisfy a homogeneous boundary or initial condition, then so does any at the end of Section 3.
linear combination of u1 and u2 .

1 Modelling using the wave equation

In this section we shall look at two situations that can be modelled by the
wave equation. In Subsection 1.1 we shall see how the wave equation arises as
a model of the transverse vibrations of a taut string. Then, in Subsection 1.2,
we shall see how damping can be incorporated into the model.

1.1 The taut string

In this subsection you will see how a continuous model of a guitar string — You saw a discrete model of
or indeed of any taut string — leads to a model of the transverse vibrations a guitar string in Subsection
of the string as a second-order partial differential equation. 4.2 of Unit 18.

We need to make some assumptions in order to develop this model. We

assume that:
• the string is uniform, with total mass M and equilibrium length L;
• the string is taut, so that in equilibrium it lies in a straight line, which
we shall take to be horizontal and label as the x-axis, with origin at the
left-hand end;
• each point of the string is subject only to a small, smoothly varying
transverse displacement u, as shown in Figure 1.1, and the effect of such Figure 1.1
a displacement on the horizontal component of the tension force in the
string is negligible;
• the string is sufficiently light for the transverse displacement, though
small, to produce a tension force much greater than the weight of the
string, so that the weight of the string can be neglected;
• all external forces, such as friction and air resistance, are negligible.
Consider a small segment of string, as shown in Figure 1.2, that in its equi-
librium position (no displacement) would occupy the interval [x, x + δx] on
the x-axis.
Since the string is uniform, the mass of this segment is M δx/L. As we are Figure 1.2
considering only transverse displacements, at some point in the motion the
two ends of this segment occupy positions u(x, t) and u(x + δx, t), respec-
tively.
Since the segment is small we may model it as a particle of mass M δx/L
located at the midpoint u(x + 12 δx, t) of the segment.
Since we are assuming that the tension in the string is much greater than the
weight of the string, the only forces that we need to model are the tension
forces on the segment, as shown in Figure 1.3. So the total force on the Figure 1.3
segment is T1 + T2 , and Newton’s second law gives
M δx ∂ 2 u
(x + 12 δx, t)j  T1 + T2 . (1.1) (The left-hand term is just
L ∂t2 mass × acceleration.)
6
 Section 1 Modelling using the wave equation

We now need expressions for T1 and T2 . If we write T1 = H1 i + V1 j, where

i and j are Cartesian unit vectors as shown in Figure 1.3, then we have
H1 = −|T1 | cos(θ(x, t)) and V1 = −|T1 | sin(θ(x, t)), so V1 = H1 tan(θ(x, t)).
One of our assumptions is that the horizontal component of the tension force
in the string has constant magnitude, T say. So here H1 = −T . Also (see
Figure 1.4), tan(θ(x, t)) is the slope of the segment at x, which is also given
by the first derivative of the displacement with respect to x, so that we can
write tan(θ(x, t)) = (∂u/∂x)(x, t). Hence we have Figure 1.4
∂u
T1 = −T i − T (x, t)j.
∂x
Similarly, for T2 , which acts at x + δx where the slope is tan(θ(x + δx, t)),
we have H2 = −H1 = T , so that
∂u
T2 = T i + T (x + δx, t)j.
∂x
Substituting into approximation (1.1) and resolving in the j-direction gives


M δx ∂ 2 u ∂u ∂u
(x + 2 δx, t)  T
1
(x + δx, t) − (x, t) .
L ∂t2 ∂x ∂x
Expanding the left-hand side as a linear Taylor approximation gives



M δx ∂ 2 u ∂ ∂ 2u ∂u ∂u
1
(x, t) + 2 δx (x, t)  T (x + δx, t) − (x, t) ,
L ∂t2 ∂x ∂t2 ∂x ∂x
and rearranging gives
∂u ∂u
M ∂ 2u ∂x (x + δx, t) − ∂x (x, t) 1 M ∂ ∂ 2u
(x, t)  − 2 δx (x, t).
T L ∂t2 δx T L ∂x ∂t2
Taking the limit as δx → 0, the last term on the right-hand side vanishes, Recall from Unit 12 that
and the approximation becomes an equality. We then obtain f (x+δx,t)−f (x,t) ∂f
lim δx = ∂x (x, t).
δx→0
M ∂ 2u ∂2u Here f = ∂u/∂x.
(x, t) = (x, t),
T L ∂t2 ∂x2
which, on rearranging and writing c2 = T L/M , becomes
∂2u 1 ∂2u
(x, t) = (x, t). (1.2)
∂x2 c2 ∂t2
This is the wave equation for transverse vibrations of a taut string.

Example 1.1
Use Equation (1.2) to determine the dimensions of the constant c.
Solution
[∂ 2 u/∂x2 ] = L L−2 = L−1 , [∂ 2 u/∂t2 ] = L T−2 . The dimensions of a
derivative were dealt with in
Therefore Exercise 2.3 of Unit 16.
[∂ 2 u/∂t2 ] L T−2
[c2 ] = 2 2
= −1 = L2 T−2 ,
[∂ u/∂x ] L
so [c] = L T−1 .

So c has the dimensions ofvelocity, and Equation (1.2) will be dimensionally

consistent provided that T L/M has the same dimensions.

7
 Section 1 Modelling using the wave equation

Exercise 1.1

Show that T L/M has the dimensions of velocity.

To obtain a particular solution to Equation (1.2), we need some initial con-

ditions. Suitable initial conditions are values for the displacement and the The transverse component of
velocity of the string at time t = 0 for all points x along the string. These the velocity of the string is
initial conditions will generally need to be expressed as functions of x, say given by the first derivative of
u with respect to t.
f (x) and g(x). Therefore we can write the initial conditions as
∂u
u(x, 0) = f (x), (x, 0) = g(x).
∂t

Example 1.2
A taut string of equilibrium length L is plucked at its midpoint, which is
given an initial displacement d, as shown in Figure 1.5. It is then released
from rest. Write down the initial conditions for the wave equation for the
transverse vibrations of this string.
Solution
The displacement shown in Figure 1.5 has two linear sections. These have Figure 1.5
slopes ±d/( 12 L) = ±2d/L. Hence the initial displacement is given by

⎧ 2d
⎪
⎨ x, 0 ≤ x ≤ 12 L,
u(x, 0) = L
⎪
⎩ 2d (L − x),
1 L < x ≤ L.

L 2

As the string is released from rest, the transverse component of the initial

velocity is given by

∂u
(x, 0) = 0, 0 ≤ x ≤ L.
∂t

*Exercise 1.2
A taut string of equilibrium length L is plucked one third of the way along

its length, which is given an initial displacement d, as shown in Figure 1.6.

What is the corresponding initial condition, for this displacement, for the

wave equation for the transverse vibrations of this string? Figure 1.6

To obtain a particular solution of Equation (1.2), we also need boundary

conditions that correspond to the ends of the string being fixed. The dis
-
placements of the ends at time t are given by u(0, t) and u(L, t). Hence the

required boundary conditions are

u(0, t) = 0, u(L, t) = 0, t ≥ 0.
Once you have tightened the string, fixed its ends, plucked it and gently let

it go, its subsequent behaviour is determined. So the two initial conditions

and two boundary conditions are sufficient to be able to determine a unique

solution to the wave equation for transverse vibrations of a taut string, as

you will see in Section 3.

8
 Section 1 Modelling using the wave equation

Wave equation for transverse vibrations of a taut string

A uniform taut string of mass M , equilibrium length L and constant
horizontal tension T , fixed at both ends, is given a displacement d at
its midpoint and then released from rest. The function u(x, t) repre-
sents the transverse displacement at the point x along the equilibrium
position of the string, measured from one fixed end, at time t.
The behaviour of u is modelled by the wave equation
∂2u 1 ∂2u TL
2
= 2 2
, where c2 = ,
∂x c ∂t M
subject to boundary conditions
u(0, t) = u(L, t) = 0, t ≥ 0,
and initial conditions If the string is not plucked at
⎧ 2d its midpoint, then a different
⎪
⎨ x, 0 < x ≤ 12 L,
formula for the initial
u(x, 0) = L
displacement will need to be
⎪
⎩ 2d (L − x), 1 L < x < L, used.
L
2

∂u

(x, 0) = 0, 0 ≤ x ≤ L.
∂t

In the Introduction it was shown that any function of the

form u(x, t) = sin(kx) cos(kct) is a solution of the wave equation, for any
constant k. In particular, therefore,



πx πct
u(x, t) = sin cos (1.3)
L L
is a solution of the wave equation for transverse vibrations of a taut string,
where L is the equilibrium length of the string. This solution also satisfies
the boundary conditions and the second initial condition for the model.

Exercise 1.3
Show that the solution given by Equation (1.3) satisfies the boundary con-
ditions
u(0, t) = u(L, t) = 0, t ≥ 0,
and the initial condition
∂u
(x, 0) = 0, 0 ≤ x ≤ L.
∂t

We shall see in Section 3 how to find a solution that also satisfies the other
initial condition.

9
 Section 1 Modelling using the wave equation

1.2 Adding damping

When we modelled the transverse vibrations of a taut string, we applied
Newton’s second law to a segment of string and obtained the approximation


M δx ∂ 2 u ∂u ∂u
(x + 2 δx, t)  T
1
(x + δx, t) − (x, t) . (1.4)
L ∂t2 ∂x ∂x
This model assumes that the only forces on the segment of string are the
tension forces due to the string. In practice, it is likely that there will be
forces damping the motion. A first model of the damping force assumes
that its magnitude is proportional to the length of the segment δx and to
the transverse component of the segment’s velocity (∂u/∂t)(x, t), and that
it acts in the direction opposite to the velocity. Approximation (1.4) then
becomes


M δx ∂ 2 u ∂u ∂u
(x + 2 δx, t)  T
1
(x + δx, t) − (x, t)
L ∂t2 ∂x ∂x
∂u
− α δx(x + 12 δx, t),
∂t
where α is a constant to be determined experimentally. Dividing through
by δx and taking the limit as δx → 0, we obtain
M ∂2u ∂2u ∂u
2
(x, t) = T 2
(x, t) − α (x, t).
L ∂t ∂x ∂t
Rearranging gives


∂2u 1 ∂2u ∂u TL αL
= 2 + 2ε , where c2 = and ε = . (1.5)
∂x2 c ∂t2 ∂t M 2M
For obvious reasons, we call this the damped wave equation. The bound-
ary and initial conditions are as before.

End-of-section Exercises
Exercise 1.4
How would the wave equation have to be modified to deal with a damping
force with a magnitude proportional to the square of the magnitude of the
velocity? Would the resulting equation be linear?
Exercise 1.5
Suppose that the initial conditions for the transverse vibrations of a taut
string are
⎧ 3d
⎪
⎨ − x, 0 ≤ x ≤ 13 L,

u(x, 0) = L

⎪ 3d
⎩ − (L − x), 13 L < x ≤ L,
2L

∂u

(x, 0) = 0, 0 ≤ x ≤ L.
∂t
Describe how the string has been set in motion.
*Exercise 1.6
A taut string is initially in its equilibrium position. At time t = 0, it is
struck in such a way as to impart, instantaneously, a transverse velocity
v (in the positive direction) to the middle third of the string. Modify the
initial conditions for the wave equation for transverse vibrations of the string
to model this situation.

10
 Section 2 Modelling using the diffusion equation

2 Modelling using the diffusion equation

In this section we shall look at a situation that can be modelled using a

different second-order partial differential equation, known as the diffusion
equation. In Subsection 2.1 we shall see how the diffusion equation arises as
a model of heat transfer by conduction through an insulated metal rod. In Heat flow was the subject of
Subsection 2.2 we shall see how, for an uninsulated metal rod, heat transfer Unit 15.
by convection can be incorporated into the model.

2.1 The insulated rod

In this subsection we shall model the cooling of a hot metal rod, both of
whose ends are kept at a fixed temperature Θ0 . We assume that the rod is
insulated along its length, so that the only significant heat loss is through its
ends. This means that the situation is effectively one-dimensional, so that
the temperature of a point of the rod at a given time varies only according
to the distance x of the point along the rod, which we shall measure from
one of its ends, as shown in Figure 2.1. As the rod is cooling down, the Figure 2.1
temperature also varies with time t. Thus the temperature is a function of
the two variables x and t, and will be denoted by Θ(x, t). We shall also
assume that the rod is uniform, of mass M , length L and cross-sectional
area A, with specific heat c and thermal conductivity κ.
To see how the temperature changes with time, we model the flow of heat.
As we did when modelling the motion of a taut string, we shall look at
a small segment of rod occupying the interval [x, x + δx]. As the rod is
uniform, this segment has mass M δx/L. We then consider the flow of heat
into and out of this segment. We shall measure the heat flow from left to
right in Figure 2.2, to match the positive x-direction. Figure 2.2
For a sufficiently small segment, we may assume that Θ(x, t)  Θ(x + δx, t).
So the change in heat energy of the segment over a small time interval
[t, t + δt] is given approximately by
M δx
E(t + δt) − E(t)  c(Θ(x, t + δt) − Θ(x, t)),
L
where E(t) is the heat energy of the segment at time t. Dividing both sides
by δt and taking the limit as δt → 0, we obtain
dE M c δx ∂Θ
 (x, t), (2.1)
dt L ∂t
which is the approximate rate of change of heat energy in the segment at
time t.
We can now use Fourier’s law to obtain another expression for the rate of
change of heat energy in the segment. We shall apply Fourier’s law to the
cross-sectional surface of area A at each end of the segment, noting that
the temperature gradient normal to each surface is given by ∂Θ/∂x. So, at As temperature is a function
time t, the rate of heat transfer by conduction into the left-hand end of the of two variables here, the
segment is temperature gradient is a
partial derivative.
∂Θ
−κA (x, t),
∂x
and the rate of heat transfer by conduction out of the right-hand end of the
segment is
∂Θ
−κA (x + δx, t).
∂x

11
 Section 2 Modelling using the diffusion equation

Therefore, since there is no source of heat transfer into or out of the segment We assume that the segment
other than by conduction along the rod, the rate of change of heat energy is not at either end of the rod.
in the segment at time t is


∂Θ ∂Θ
−κA (x, t) − −κA (x + δx, t)
∂x ∂x


∂Θ ∂Θ
= κA (x + δx, t) − (x, t) . (2.2)
∂x ∂x
Hence, using Equation (2.1), we have


∂Θ ∂Θ M c δx ∂Θ
κA (x + δx, t) − (x, t)  (x, t).
∂x ∂x L ∂t
Dividing both sides by δx and taking the limit as δx → 0, we obtain
∂2Θ M c ∂Θ
κA 2
(x, t) = (x, t),
∂x L ∂t

which, on rearranging and writing α = κAL/(M c), becomes

∂2Θ 1 ∂Θ
2
(x, t) = (x, t). (2.3)
∂x α ∂t
This is known as the diffusion equation, or sometimes as the heat equa
-
tion, for the variation of temperature along an insulated rod.

Since the ends of the rod are kept at a steady temperature Θ0 , the boundary

conditions are

Θ(0, t) = Θ(L, t) = Θ0 , t ≥ 0. (2.4)

Further, if the initial distribution of temperature along the rod is given by
a function f (x), then we have the initial condition Note that we need only one
initial condition because the
Θ(x, 0) = f (x), 0 < x < L. (2.5) equation is first order in time.
These conditions are sufficient for us to be able to obtain a unique solution
to the diffusion equation for the variation of temperature in an insulated
rod, as you will see in Section 4.

*Exercise 2.1
Suppose that the uniform rod in the above model has density ρ. Simplify

the formula for the constant α.

We have found that the constant in the diffusion equation is α = κ/(cρ),

which depends only on the intrinsic properties of the material, and not
on its shape. The length L does, however, enter the problem through the
boundary conditions, and implicitly in the initial condition.

12
 Section 2 Modelling using the diffusion equation

Diffusion equation for temperature variation along a rod

A uniform rod of mass M , length L, density ρ, thermal conductivity κ,
specific heat c and cross-sectional area A, completely insulated apart
from its ends, which are both at the same fixed temperature Θ0 , has The model still applies if the
an initial temperature distribution given by a function f (x), where x is ends of the rod are at
the distance from one end of the rod. The function Θ(x, t) represents different temperatures, in
which case the two boundary
the temperature at the point x along the rod at time t. conditions would have
The behaviour of Θ is modelled by the diffusion equation different right-hand sides.

∂2Θ 1 ∂Θ κAL κ
= , where α = = ,
∂x2 α ∂t Mc cρ
subject to boundary conditions

Θ(0, t) = Θ(L, t) = Θ0 , t ≥ 0,

and initial condition

Θ(x, 0) = f (x), 0 < x < L.

*Exercise 2.2
Show that the function


απ2 t πx
−
Θ(x, t) = e L2 sin
L
satisfies the diffusion equation (2.3) and the boundary conditions (2.4) if

Θ0 = 0.

*Exercise 2.3
Suppose that, initially, the temperature of the rod rises linearly towards a

peak in the centre, d above the end temperature Θ0 , as shown in Figure 2.3.

Write down a formula describing the initial temperature function f (x). Figure 2.3

2.2 The convecting rod

Suppose that the rod in Subsection 2.1 is not insulated. In this case, heat is
not only being conducted along the rod, but is also being convected from its
surface, of area S. Let us suppose that the ambient temperature of the fluid
surrounding the rod is Θ0 , the same as the temperature at its ends, and that
the convective heat transfer coefficient is h. To add the effects of convection
to our model, we again look at a small segment of the rod of length δx. The
rate of change of heat energy in the segment by conduction is given still by
the right-hand side of Equation (2.2). We now need to subtract from this
the rate of loss of heat energy by convection. Since the surface area of the
segment is Sδx/L, this rate of loss of heat energy is given by Equation (3.1)
of Unit 15 as approximately
Sδx
h (Θ(x, t) − Θ0 ), (2.6)
L
where again we assume that the temperature of the whole segment at time
t is approximately Θ(x, t).

13
 Section 3 Separating the variables

Exercise 2.4
Use formulae (2.1), (2.2) and (2.6) to obtain a second-order partial differen-
tial equation modelling the temperature variation in the convecting rod.

So, for the convecting rod, the model of the temperature variation is
∂2Θ 1 ∂Θ κAL hS
= + γ(Θ − Θ0 ), where α = and γ = .
∂x2 α ∂t Mc κAL
The boundary and initial conditions are as before.

Exercise 2.5
Suppose that the convecting rod has uniform density ρ and circular cross-
section of radius r. Find expressions for the constants α and γ that do not
involve M , L, A or S.

End-of-section Exercise
*Exercise 2.6
Write down the initial condition describing the temperature distribution if
the central third of the rod is initially heated to a temperature Θ1 while the
remainder of the rod stays at the background temperature Θ0 .

3 Separating the variables

In the first two sections you have seen several physical situations modelled
using partial differential equations. The next step is to find solutions to these
models. We shall find solutions in this and the next section, but detailed
interpretation of these solutions will be left until Section 5, where you will
be able to explore the solutions on your computer.
In this section we shall solve the wave equation and the damped wave equa-
tion from Section 1. We shall do so in the context of the taut string, though
the method of solution is independent of the physical context and can be ap- Indeed, the method of
plied to any model involving the wave equation. We begin in Subsection 3.1 solution can be applied to
by solving the wave equation for the taut string subject to its homogeneous models involving the diffusion
equation too, as you will see
boundary conditions and to the homogeneous initial condition specifying a in Section 4.
zero initial velocity. Then, in Subsection 3.2, we go on to examine how to
amend this solution to satisfy the inhomogeneous initial condition, speci-
fying the initial displacement of the string. Finally, in Subsection 3.3, we
extend our solutions to the damped wave equation. The goal, which is pre-
sented at the end of the section, is a systematic procedure, which will enable
us to solve a wide variety of partial differential equations.

14
 Section 3 Separating the variables

3.1 Normal modes

In this subsection we shall solve the wave equation for vibrations of a taut
string,
∂2u 1 ∂2u
= , (3.1)
∂x2 c2 ∂t2
subject to the boundary conditions
u(0, t) = u(L, t) = 0, t ≥ 0, (3.2)
and the initial condition
∂u
(x, 0) = 0, 0 ≤ x ≤ L. (3.3)
∂t
The method that we shall use, known as separating the variables, is
applicable to a variety of homogeneous linear partial differential equations,
including the wave equation and the diffusion equation. Essentially, it in-
volves looking for solutions of the form u(x, t) = X(x)T (t), where X(x) is a Note that the function T here
function of x alone and T (t) is a function of t alone, so that the variables x has no connection with the
and t occur in two separate functions that make up the solution. (constant) tension T used in
Section 1.
It turns out that in this particular problem T (t) has the form
T (t) = cos(ωt),
where ω is the angular frequency of the vibrations.
We shall see why in Subsection 3.3.
In fact there are an infinite number of such solutions, corresponding to
different values of ω. Each solution of the form X(x) cos(ωt) is known as a
normal mode of the continuous model. Let us now proceed to determine These normal modes are
the form that X(x) and ω must take if a function of the form X(x) cos(ωt) is analogous to the normal
to be a normal mode of the model. In other words, we are going to look for modes studied in Unit 18.
solutions of the form u(x, t) = X(x)T (t), where T (t) = cos(ωt). (As we shall We shall see later how the
see later, this form of the function T (t) ensures that initial condition (3.3) method can be generalized to
any function T (t).
is satisfied.)

*Exercise 3.1
Determine the partial derivatives ∂ 2 u/∂x2 and ∂ 2 u/∂t2 of the function

u(x, t) = X(x) cos(ωt).

Substituting the values found for the partial derivatives in Exercise 3.1 into
the wave equation (3.1) gives
ω2

X  (x) cos(ωt) = − X(x) cos(ωt),

c2
which on rearranging gives


ω2

X  (x) + 2 X(x) cos(ωt) = 0.

c
Since the time-dependent function cos(ωt) is not always zero, the other factor
must be zero. So we must have
ω2

X  (x) + 2 X(x) = 0. (3.4)

c
We have reduced a partial differential equation to an ordinary differential
equation, involving functions and derivatives depending on x alone, which
we can solve immediately.

15
 Section 3 Separating the variables

*Exercise 3.2
What is the general solution of the differential equation (3.4)?

So the wave equation (3.1) has normal mode solutions of the form
u(x, t) = X(x) cos(ωt),
where X(x), the initial displacement, is given by Note that u(x, 0) = X(x), so
 ωx  ωx that X(x) is the initial
X(x) = A cos + B sin , (3.5) displacement.
c c

where A and B are arbitrary constants.

These solutions do not restrict the angular frequency ω in any way. However,

for the taut string, there are very severe restrictions placed on ω by the

boundary conditions.

*Exercise 3.3
What do the boundary conditions (3.2) tell you about the initial displace
-
ment function X(x)?

The boundary conditions tell us that

X(0) = 0 and X(L) = 0.
Using Equation (3.5), these conditions reduce to


ωL
A = 0 and B sin = 0.
c
From this second condition we have B = 0 or sin(ωL/c) = 0. The case
where B = 0 gives X(x) = 0 for all x, the zero function, which corresponds
to no initial displacement. More interesting is the case where sin(ωL/c) = 0,
which gives ωL/c = rπ for any integer r, so that the only possible values for
ω are
crπ
ω= , r = 0, ±1, ±2, ±3, . . . .
L
So the conditions on A and ω give


rπx
X(x) = B sin , r = 0, ±1, ±2, ±3, . . . , (3.6) The case where X(x) is the
L zero function is covered by
where B is an arbitrary constant. Since B is an arbitrary constant and the case where r = 0.
sin(−y) = − sin y, the cases where r = −1, −2, −3, . . . can be absorbed into B.
Also, the case r = 0 is covered by taking B = 0. So we can write Equa-
tions (3.6) more succinctly as


rπx
X(x) = B sin , r = 1, 2, 3, . . . . Hence the allowed values for
L ω are now restricted to
We have found a range of values of ω and the form of the function X(x) in the ω = crπ/L, r = 1, 2, 3, . . . .
normal mode. Now it is time to concentrate on the function T (t) = cos(ωt).
For our allowed values for ω, the cosine term in the normal mode solution
u(x, t) = X(x) cos(ωt) must take the form


crπt
cos , r = 1, 2, 3, . . . .
L

16
 Section 3 Separating the variables

Hence normal mode solutions of the wave equation (3.1) that also satisfy
the boundary conditions (3.2) are



rπx crπt
u(x, t) = X(x)T (t) = B sin cos , r = 1, 2, 3, . . . , (3.7)
L L

where B is an arbitrary constant.

Now, there is a version of the principle of superposition that applies to linear
partial differential equations. As you saw in the Introduction, in the case
of homogeneous equations of this type (such as the wave equation), the
principle of superposition says that any linear combination of solutions of
the equation is also a solution. So all linear combinations of solutions (3.7)
are also solutions of the wave equation (3.1). Furthermore, the principle
of superposition says that any linear combination of solutions that satisfy a
homogeneous boundary condition is also a solution satisfying that boundary
condition. Hence, since each of solutions (3.7) satisfies the homogeneous
boundary conditions (3.2), so does any linear combination of them. So
every linear combination of Equations (3.7) satisfies the wave equation (3.1)
subject to boundary conditions (3.2).
In the light of all this, consider the family of solutions



rπx crπt
ur (x, t) = sin cos , r = 1, 2, 3, . . . . (3.8)
L L
The set of linear combinations of these functions is the same as the set of
linear combinations of solutions (3.7), and so also forms a set of solutions of
the wave equation (3.1) subject to boundary conditions (3.2). We shall take
the family of solutions (3.8) to be our standard set of normal mode solutions
from which every solution of the wave equation (3.1) subject to boundary
conditions (3.2) and initial condition (3.3) can be obtained by taking an
appropriate linear combination.

*Exercise 3.4
Interpret solutions (3.8) for r = 1, 2 and 3 in terms of the transverse vibra-
tions of a taut string.

The first three solutions in family (3.8) have initial displacements given by
the graphs in Figure 3.1. Clearly the larger the value of r, the higher the
frequency (and the higher the pitch of the note).

Figure 3.1

17
Section 3 Separating the variables

So far, we have found a family of solutions (3.8) satisfying the wave equa-
tion (3.1) and the boundary conditions (3.2). In fact, these solutions also
satisfy the initial condition (3.3).

Exercise 3.5
Check that the solutions (3.8) satisfy the initial condition (3.3).

The principle of superposition tells us that if solutions (3.8) satisfy the

homogeneous initial condition (3.3), then so must any linear combination of
solutions (3.8). Thus we have found a family of solutions (3.8), the linear
combinations of which comprise a set of solutions to the wave equation (3.1)
subject to the homogeneous boundary conditions (3.2) and the homogeneous
initial condition (3.3). Moreover, though we do not prove it here, the set
of linear combinations of solutions (3.8) is the complete set of solutions
with these properties. It remains to determine what restrictions we need to
place on our family of solutions to satisfy the other, inhomogeneous, initial
condition of our model of the transverse vibrations of a taut string.

3.2 The initial displacement

In this subsection we shall see how we need to restrict the solutions we ob-
tained in Subsection 3.1 to the wave equation (3.1), subject to boundary
conditions (3.2) and initial condition (3.3), so as to satisfy the condition for
the initial displacement. We shall look in detail at just one initial displace-
ment, given by
u(x, 0) = f (x), 0 < x < L, (3.9)
where
⎧ 2d
⎪
⎨ x, 0 ≤ x ≤ 12 L,
f (x) = L (3.10)
⎪
⎩ 2d (L − x), 1 L < x ≤ L,
L 2

which is the initial displacement of a taut string plucked at its centre. The
technique can be applied to other initial displacements, in an obvious way,
as we shall indicate later.
Unfortunately, none of the individual members of the family (3.8) of func-
tions ur (x, t) satisfies condition (3.9), since
 rπx
ur (x, 0) = sin  f (x).
=
L
However, we can find a linear combination of the family (3.8) of functions
that does satisfy condition (3.9). You will recall from Unit 21 that a function
such as f (x) can be approximated by a Fourier series, i.e. by an infinite sum
of sinusoidal terms. So it should come as no surprise that what we need to
satisfy condition (3.9) is an infinite linear combination of the family (3.8) of
solutions, i.e. a solution of the form

∞
∞


rπx crπt
u(x, t) = Br ur (x, t) = Br sin cos . (3.11)
r=1 r=1
L L
The principle of superposition extends to such infinite sums, so that they
satisfy not only the wave equation (3.1), but also the homogeneous boundary
conditions (3.2) and the homogeneous initial condition (3.3). So all that
remains is to determine when Equation (3.11) satisfies the inhomogeneous
initial condition (3.9).

18
 Section 3 Separating the variables

When t = 0, Equation (3.11) becomes

∞

rπx
u(x, 0) = Br sin .
r=1
L
So if Equation (3.11) is to satisfy condition (3.9), then we must have
∞



rπx
f (x) = Br sin .
L
r=1
Hence all we need to do is to express the function f (x), describing the This technique can be applied
initial displacement, as a Fourier sine series. You saw in Unit 21 how to to any initial displacement
deal with such functions that are not naturally periodic. The function f (x) function for which we can
obtain a Fourier sine series.
in Equation (3.10) is defined on the interval [0, L]. We want it described as
a sine series, in which the sine functions are periodic on the fundamental
interval [−L, L].
You saw in Unit 21 that the coefficients Br are given by the integrals


2 L rπx
Br = f (x) sin dx.
L 0 L
For the function f (x) in Equation (3.10), these integrals evaluate to


8d rπ
Br = 2 2 sin , This can easily be checked
r π 2 using the computer algebra
so the values for the coefficients Br , for r = 1, 2, 3, 4, 5, 6, . . ., are given by package (cf. Activity 4.2 in
8d 8d 8d Unit 21).
, 0, − 2 , 0, , 0, . . . .
π2 9π 25π2
Substituting into Equation (3.11) gives the solution for our plucked string
problem as





8d πx πct 1 3πx 3πct
u(x, t) = 2 sin cos − sin cos
π L L 9 L L




1 5πx 5πct
+ sin cos − ... .
25 L L

This may seem a very complicated solution, and you may not be able to see
immediately how this combination of terms behaves. However, in Section 5
you will be able to plot the solution graphically, using your computer. The
method used to obtain this solution, known as separating the variables, can
be applied to a variety of homogeneous linear partial differential equations,
not just the wave equation.

19
 Section 3 Separating the variables

3.3 The damped plucked string

In Subsections 3.1 and 3.2 we solved the wave equation model for vibrations
of a taut string initially plucked at its midpoint. In this subsection we shall
generalize the method in order to solve the damped wave equation model
for vibrations of a damped taut string plucked at its midpoint. The model We shall label each step, and
developed in Subsection 1.2 is at the end of the subsection

 the method will be
∂2u 1 ∂2u ∂u summarized as Procedure 3.1.
= 2 + 2ε , (3.12)
∂x2 c ∂t2 ∂t
subject to boundary conditions
u(0, t) = u(L, t) = 0, t ≥ 0, (3.13)
and initial conditions

⎧ 2d

⎪
⎨ x, 0 < x ≤ 12 L,

u(x, 0) = L (3.14)

⎪
⎩ 2d (L − x), 1 L < x < L,
L
2

∂u

(x, 0) = 0, 0 ≤ x ≤ L. (3.15)
∂t

We shall assume that the damping is very weak, so that ε is small, as this

makes the solution process easier.

Step (a)
In Subsection 3.1 we wrote the unknown function as
u(x, t) = X(x)T (t) = X(x) cos(ωt).
Such solutions, with T (t) = cos(ωt), represent a system that vibrates in
-
definitely. At a point a distance x along the string, the amplitude of the

vibration is X(x) for all times t. Clearly, this will not do for the damped

taut string, as we expect the vibrations to fade away. So here we just look

for solutions of the form

u(x, t) = X(x)T (t),

without giving any specific form to T (t).
We now have to find the relevant partial derivatives of the function u(x, t)

in terms of the functions X(x) and T (t).

*Exercise 3.6
If the function u is defined as a product
u(x, t) = X(x)T (t),
find formulae for the partial derivatives ∂u/∂t, ∂ 2 u/∂t2 and ∂ 2 u/∂x2 in

terms of the functions X and T and their ordinary derivatives.

20
 Section 3 Separating the variables

Step (b)
Substituting the results of Exercise 3.6 into Equation (3.12) gives

1  

X  (x)T (t) = 2 X(x)T  (t) + 2εX(x)T  (t) .

c
We now want to separate the variables, i.e. to make each side of the equation
a function of a single variable. We do this by dividing through by X(x)T (t)
to obtain


X  (x) 1 T  (t) T  (t)
= 2 + 2ε . (3.16)
X(x) c T (t) T (t)

On the left-hand side there is a function of the variable x only, and on the
right-hand side there is a function of the variable t only. Whatever the value
of x and whatever the value of t, the two sides of the equation must always
be equal. The only way that this can occur for all x and all t is for both
sides of the equation to be constant. Let us call this constant µ. Then


X  (x) 1 T  (t) T  (t)
= 2 + 2ε = µ.
X(x) c T (t) T (t)
The single equation (3.16) then becomes a pair of equations. The first of
these is
X  (x)
= µ,
X(x)
which on rearranging becomes
X  − µX = 0. (3.17)

*Exercise 3.7
Find the corresponding differential equation satisfied by the function T .

The second equation reduces to

T  + 2εT  − c2 µT = 0. (3.18)
Now we have completed the process of separating the single partial differ-
ential equation into two ordinary differential equations, but we still have
to find boundary conditions for the function X. To do this, we substitute
u(x, t) = X(x)T (t) into the boundary conditions (3.13).
The boundary condition u(0, t) = 0 becomes X(0)T (t) = 0 for all t. It is pos-
sible that the function T always takes the value zero, i.e. is the zero function,
but that gives us the uninteresting solution u = 0, where the system remains
undisturbed. If the system is in motion, then the function T cannot be the
zero function, and hence the boundary condition reduces to X(0) = 0. Sim-
ilarly, the boundary condition u(L, t) = 0 gives X(L)T (t) = 0 for all t, and
hence by a similar argument X(L) = 0. So the boundary conditions become
X(0) = 0 and X(L) = 0, (3.19)
which are boundary conditions for the ordinary differential equation (3.17).

21
 Section 3 Separating the variables

Step (c)
We solve the differential equations for X and T so that the boundary con
-
ditions (3.19) are satisfied.

For the function X, we have Equation (3.17) and boundary conditions (3.19).

If µ is positive, then we can write µ = k 2 , for some non-zero number k, and

the general solution to Equation (3.17) is

X(x) = Aekx + Be−kx .

The only way to satisfy the boundary conditions is for A and B both to be The boundary conditions
zero, which leads to the trivial solution u = 0. reduce to A + B = 0 and
AekL + Be−kL = 0.
*Exercise 3.8 Substituting B = −A into the
second equation gives
What is the solution when µ = 0? A(ekL − e−kL ) = 0. Since

L= 0 and k 
= 0 we have
ekL − e−kL = 0, so that
A = B = 0.
Hence, to obtain interesting answers, we assume that µ is negative and write
µ = −k 2 , for some non-zero number k. Equation (3.17) becomes
X  + k2 X = 0,
with general solution
X(x) = A cos(kx) + B sin(kx).
As in Subsection 3.1, the boundary conditions (3.19) tell us that A = 0 and
B sin(kL) = 0, so that B = 0 or sin(kL) = 0. The case B = 0 gives the
trivial solution again, whereas the case sin(kL) = 0 restricts k to take one
of the values rπ/L, where r is an integer. Hence, as in Subsection 3.1, we
find again a family of solutions to the boundary-value problem for X of the
form


rπx
Xr (x) = B sin , r = 1, 2, 3, . . . , (3.20)
L
where B is an arbitrary constant.
Now we must deal with the equation for T . We know from our discussion of
the function X that to have a non-trivial solution u, we must have µ = −k 2
and k = rπ/L (for positive integers r), so that the differential equation (3.18)
for T becomes
T  + 2εT  + ω2 T = 0, where ω = ck. (3.21)
To solve Equation (3.21) for T , we first solve the auxiliary equation
λ2 + 2ελ + ω2 = 0,
which gives

λ = −ε ± ε2 − ω2 .
Using the assumption that ε  ω, i.e. ε is much smaller than ω, which
corresponds to very weak damping, this reduces to
λ  −ε ± iω.
The corresponding general solution is
T (t)  e−εt (α cos(ωt) + β sin(ωt)),
where α and β are arbitrary constants.

22
 Section 3 Separating the variables

Using the allowed values for k (i.e. rπ/L, for positive integers r), where
ω = ck, this gives the approximate solutions




−εt crπt crπt
Tr (t)  e α cos + β sin , r = 1, 2, 3, . . . , (3.22)
L L
where α and β are arbitrary constants.
Thus, we can combine the two families of solutions (3.20) and (3.22) to
obtain the family of approximate solutions
ur (x, t) = Xr (x)Tr (t) The fact that we allow all




 linear combinations of
−εt rπx crπt crπt
 e sin α cos + β sin , members of the family (3.23)
L L L allows us to omit the
r = 1, 2, 3, . . . , (3.23) arbitrary constants from
family (3.20), but we cannot
where α and β are arbitrary constants. By the principle of superposition, any omit the arbitrary constants
linear combination of members of this family of solutions is (approximately) α and β.
a solution to the original partial differential equation (3.12) and satisfies the
boundary conditions (3.13).

Step (d)
Now we try to satisfy the initial conditions. We begin with the homogeneous
initial condition (3.15). This was automatically satisfied for the undamped
problem, but it is not automatically satisfied here. We know, from the
principle of superposition, that if each member of family (3.23) satisfies the
homogeneous initial condition (3.15), then so will any linear combination of
members of the family. So we just need to consider ur (x, t) for an arbitrary
r (in the range r = 1, 2, 3, . . .). Differentiating (3.23) with respect to t, we
obtain





∂ur −εt rπx crπ crπt crπ crπt
(x, t)  e sin −α sin +β cos
∂t L L L L L





rπx crπt crπt
− εe−εt sin α cos + β sin .
L L L
When t = 0, and writing ω = ck = crπ/L as earlier, we obtain




∂ur crπ rπx rπx
(x, 0)  β − εα sin = (ωβ − εα) sin .
∂t L L L
For initial condition (3.15) to be satisfied, for 0 ≤ x ≤ L, we must have
ωβ − εα  0. For very weak damping, as we noted earlier, we have ε  ω.
This means that to satisfy ωβ − εα  0, we must have β  0. Thus, to satisfy
the initial condition (3.15), approximately, we need to restrict family (3.23)
to



−εt rπx crπt
ur (x, t)  αe sin cos , r = 1, 2, 3, . . . ,
L L
where α is an arbitrary constant. As before, since our solutions are linear
combinations of these functions ur (x, t), we can ignore the arbitrary con-
stants α and obtain the family of approximate solutions



−εt rπx crπt
ur (x, t)  e sin cos , r = 1, 2, 3, . . . . (3.24)
L L
By the principle of superposition, any linear combination of members of
this family of solutions is (approximately) a solution to the damped wave
equation (3.12) and satisfies the homogeneous boundary conditions (3.13)
and (approximately) the homogeneous initial condition (3.15).

23
Section 3 Separating the variables

We now need to satisfy the inhomogeneous initial condition (3.14). As in

Subsection 3.2, we look for an approximate solution of the form

∞ ∞




−εt rπx crπt
u(x, t) = Br ur (x, t)  e Br sin cos , (3.25)
L L
r=1 r=1
where the Br are constants.
Setting t = 0 in approximation (3.25) gives

∞

rπx
u(x, 0)  Br sin , (3.26)
L
r=1

while the given initial condition (3.14) is

⎧ 2d
⎪
⎨ x, 0 < x ≤ 12 L,

u(x, 0) = L

⎪
⎩ 2d (L − x), 1 L < x < L.
L 2

The situation is exactly the same as in Subsection 3.2, so we obtain the same
Fourier coefficients Br as there. Substituting these into approximation (3.25)
gives the final form of the approximate solution as





8d −εt πx πct 1 3πx 3πct
u(x, t)  2 e sin cos − sin cos Remember that this is an
π L L 9 L L approximate solution, under


 the assumption of very weak
1 5πx 5πct
+ sin cos − ... . damping, where ε is very
25 L L small.
This is the same as the solution to the undamped problem except for the
factor e−εt , which describes exponential decay. Hence the shape of the
string is roughly the same as before, apart from the amplitude becoming
progressively smaller by the factor e−εt . The shape determines the sound,
so the sound stays the same, only now it gets progressively quieter, since
the volume is determined by the amplitude. You will be able to investigate
the predicted behaviour more fully by use of your computer in Section 5.
As promised, we now summarize the method of separating the variables in
the form of a procedure.

24
 Section 3 Separating the variables

Procedure 3.1 Separating the variables

Given a homogeneous linear partial differential equation with depen-
dent variable u and independent variables x and t, subject to appropri-
ate homogeneous boundary and initial conditions, a solution can often
be found by means of the following steps.
(a) Write the unknown function as a product of functions of one vari-
able, i.e. write
u(x, t) = X(x)T (t).
Find the required partial derivatives of u in terms of the ordinary
derivatives of the functions X and T .
(b) Substitute the partial derivatives found in (a) into the partial differ-
ential equation. Rearrange the equation so that each side consists
of a function of a single independent variable. Each side of the re-
arranged equation must equal a constant, giving separate ordinary
differential equations for X and T . Use the boundary conditions
for u to find boundary conditions for X.
(c) Solve the differential equations for X and T found in (b). Use the
boundary conditions for X to find families of solutions Xr (x) and The non-zero members of
Tr (t) that combine to give a family of solutions the family of solutions
ur (x, t) = Xr (x)Tr (t) are
ur (x, t) = Xr (x)Tr (t), r = 0, 1, 2, 3, . . . , known as normal mode
solutions of the problem.
of the original partial differential equation.
(d) Use the initial conditions and results about Fourier series to deter-
mine (when possible) an infinite linear combination of the family
of solutions found in (c), i.e.
∞

u(x, t) = Br ur (x, t),

r=0

that solves the original problem.

End-of-section Exercises
*Exercise 3.9
This exercise asks you to verify that the principle of superposition applies to
the wave equation subject to certain homogeneous conditions, to justify its
use in this section. Suppose that the functions u1 and u2 satisfy the wave
equation
∂2u 1 ∂2u
=
∂x2 c2 ∂t2
and the homogeneous conditions
∂u
u(0, t) = u(L, t) = 0 and (x, 0) = 0.
∂t
Show that the linear combination u = a1 u1 + a2 u2 , where a1 and a2 are
constants, is a function that satisfies the same equation and conditions.

25
 Section 4 Solving the heat transfer problems

*Exercise 3.10
Use Steps (a) to (c) of Procedure 3.1 to find an infinite family of solutions
for the following model:
∂2u ∂2u
+ 2 = 0, This equation is known as
∂x2 ∂t Laplace’s equation.
subject to boundary conditions
u(0, t) = u(1, t) = 0, t ≥ 0.

4 Solving the heat transfer problems

In this short section we shall see how the method of separating the variables,
in Procedure 3.1, can be used to solve the diffusion equation subject to ap-
propriate boundary and initial conditions. We shall do this in Subsection 4.1
in the context of the insulated rod problem of Subsection 2.1, and in Sub-
section 4.2 in the context of the convecting rod problem of Subsection 2.2.
If you feel confident about the method of Procedure 3.1, then you might
like to try and solve the models for these problems yourself before working
through this section — in this way you might consider this section as an
extended End-of-section Exercise for Section 3.

4.1 The insulated rod problem solved

In this subsection we shall use Procedure 3.1 to solve a particular model for
the insulated rod problem discussed in Subsection 2.1. In this model, the
temperatures at the ends of the rod are zero (Θ0 = 0 ◦ C), and the initial
temperature of the rod is given by a function f (x) identical to the initial
displacement function for the taut string problem considered in Section 3.
The model is
∂2Θ 1 ∂Θ
2
= , (4.1)
∂x α ∂t
subject to boundary conditions
Θ(0, t) = Θ(L, t) = 0, t ≥ 0, (4.2)
and initial condition
⎧ 2d
⎪
⎨ x, 0 < x ≤ 12 L,
Θ(x, 0) = f (x) = L (4.3)
⎪
⎩ 2d (L − x), 1 L < x < L.
L 2

Later in this subsection we shall see how our solution can be adapted to
cover any fixed temperature Θ0 at the ends of the rod.
Step (a)
We begin separating the variables by setting
Θ(x, t) = X(x)T (t), (4.4)
and finding the relevant partial derivatives:
∂2Θ
= X  T, (4.5)
∂x2
∂Θ
= XT  . (4.6)
∂t

26
 Section 4 Solving the heat transfer problems

Step (b)
To find separate ordinary differential equations for X and T , we substitute

the formulae for the partial derivatives into the partial differential equation,

and rearrange. Here, substituting (4.5) and (4.6) into Equation (4.1) gives

1
X  T = XT  .
α

Dividing by XT , to separate the variables, results in

X 
1 T 
= . (4.7)
X α T

Here the left-hand side is a function of the variable x alone and the right
-
hand side is a function of the variable t alone, so both are equal to a con
-
stant µ. The single equation (4.7) then becomes a pair of equations

X 
1 T
= µ, = µ,
X αT
which on rearranging give
X  − µX = 0, (4.8)
T  − αµT = 0. (4.9)
To find boundary conditions for X, we put x = 0 and x = L in Equation (4.4)

and substitute into the boundary conditions (4.2), which gives

X(0)T (t) = X(L)T (t) = 0, t ≥ 0,

and hence
X(0) = X(L) = 0. (4.10)

Step (c)
Next we solve the differential equations for X and T , and combine the

families of solutions.

The differential equation (4.8) for X, and its boundary conditions (4.10), are

the same as in Section 3. You have seen that a non-trivial solution occurs

only if the constant µ is negative. Hence, as before, we replace µ by −k 2

and find the same family of solutions


rπx
Xr (x) = B sin , r = 1, 2, 3, . . . , (4.11) As before, k takes one of the
L values rπ/L.
where B is an arbitrary constant.
Our next task is to solve Equation (4.9) for the function T when the constant
µ is replaced by −k 2 and k takes the value rπ/L for some positive integer r. Under these circumstances,
Equation (4.9) is the equation
*Exercise 4.1 for exponential decay.
Solve the differential equation
T  + αk2 T = 0.

Putting k = rπ/L in the solution to Exercise 4.1 gives the family of solutions
αr2 π2 t
Tr (t) = Ae− L2 , r = 1, 2, 3, . . . , (4.12)
where A is an arbitrary constant.

27
 Section 4 Solving the heat transfer problems

Combining families (4.11) and (4.12), we obtain the following family of so-
lutions to the diffusion equation (4.1) subject to boundary conditions (4.2):


αr2 π2 t rπx

− 2

Θr (x, t) = e L sin , r = 1, 2, 3, . . . . (4.13)

L
As before, any linear combination of members of this family is a solution
to (4.1) subject to (4.2).

Step (d)
Finally, we form the infinite linear combination

∞

αr2 π2 t rπx

Θ(x, t) = Br e− L2
sin , (4.14)
r=1
L
and use the initial condition (4.3) and results on Fourier series to determine
the coefficients Br . Setting t = 0 in Equation (4.14) gives

∞

rπx
Θ(x, 0) = Br sin .
L
r=1

The function f (x) used in initial condition (4.3) is the same as that used
in Section 3, hence we arrive at the same values for the coefficients Br , and
the solution is therefore



8d − απ22 t πx 1 − 9απ22 t 3πx
Θ(x, t) = 2 e L sin − e L sin
π L 9 L
2


1 25απ t 5πx
+ e− L2 sin − ··· .
25 L
Now, as promised earlier, let us see how to amend our solution to cover the
case where Θ0 = 0, i.e. where the boundary conditions are

Θ(0, t) = Θ(L, t) = Θ0 , t ≥ 0.

*Exercise 4.2
Consider the diffusion equation

∂2Θ 1 ∂Θ

2
= ,
∂x α ∂t
subject to boundary conditions

Θ(0, t) = Θ(L, t) = Θ0 , t ≥ 0,

where Θ0 
= 0, and initial condition

Θ(x, 0) = f (x), 0 < x < L,

for some function f (x).

(a) Show that the function Θ(x, t) = Θ0 satisfies the differential equation.
(b) Use the principle of superposition, Equations (4.13) and the result from
part (a) to obtain a family of solutions to the differential equation sub-
ject to the boundary conditions.
(c) Explain how you would go on to obtain a solution that also satisfies the
initial condition.

28
 Section 4 Solving the heat transfer problems

4.2 The convecting rod problem solved

In this subsection we shall look briefly at how Procedure 3.1 can be used to
solve the following model for the convecting rod:
∂2Θ 1 ∂Θ
2
= + γΘ, (4.15) This is the equation obtained
∂x α ∂t in Exercise 2.4, with Θ0 = 0.
subject to boundary conditions
Θ(0, t) = Θ(L, t) = 0, t ≥ 0, (4.16)
and initial condition
⎧ 2d
⎪
⎨ x, 0 < x ≤ 12 L,

Θ(x, 0) = L (4.17)

⎪
⎩ 2d (L − x), 1
L 2L < x < L.

Exercise 4.3
Perform Steps (a) and (b) of Procedure 3.1 to begin the process of solving
the above model for a convecting rod.

Exercise 4.3 shows that the function X satisfies the equation

X  + k2 X = 0, with X(0) = X(L) = 0,
while T satisfies the equation
T  + α(k2 + γ)T = 0.
Step (c) of Procedure 3.1 leads again to k = rπ/L, for any integer r, and
then to the family of solutions for X given by


rπx
Xr (x) = B sin , r = 1, 2, 3, . . . , (4.18)
L
where B is an arbitrary constant.

Exercise 4.4
Solve the differential equation
T  + α(k2 + γ)T = 0.

Putting k = rπ/L, for any positive integer r, into the solution to Exercise 4.4
gives the family of solutions


r 2 π2
−α +γ t
L2
Tr (t) = Ae , r = 1, 2, 3, . . . , (4.19)
where A is an arbitrary constant.
Combining families (4.18) and (4.19), we obtain the following family of so-
lutions to Equation (4.15) subject to boundary conditions (4.16):


r 2 π2

−α +γ t rπx
L2
Θr (x, t) = e sin , r = 1, 2, 3, . . . .
L
As before, any linear combination of members of this family is a solution
to (4.15) subject to (4.16).

29
 Section 4 Solving the heat transfer problems

In Step (d) of Procedure 3.1 we write


∞

r2 π2

−α +γ t rπx
L2
Θ(x, t) = Br e sin ,
L
r=1
and setting t = 0 gives
∞



rπx
Θ(x, 0) = Br sin .
L
r=1
This is exactly the same as in the three previous cases, so we arrive at the

same values for the coefficients. Therefore the solution is








π2 9π2
8d −α L2 +γ t πx 1 −α L2 +γ t 3πx
Θ(x, t) = 2 e sin − e sin
π L 9 L


2


 
25π
1 −α L2 +γ t 5πx
+ e sin − ···
25 L



8d −αγt − απ22 t πx 1 − 9απ22 t 3πx
= 2 e e L sin − e L sin
π L 9 L
2


1 25απ t 5πx
+ e− L2 sin − ··· .
25 L
This is the same as the solution for the insulated rod except for the e
−αγt
factor, so the convecting rod cools more quickly than the insulated rod by

a factor of e−αγt .

End-of-section Exercises
*Exercise 4.5
Apply Steps (a) to (c) of Procedure 3.1 to the diffusion equation
∂2Θ 1 ∂Θ
2
= ,
∂x α ∂t
subject to boundary conditions These boundary conditions
model the situation where the
∂Θ ∂Θ
(0, t) = (L, t) = 0, t ≥ 0, ends of a hot rod are

∂x ∂x insulated.

and initial condition

⎧ 2d

⎪
⎨ x, 0 ≤ x ≤ 12 L,

Θ(x, 0) = L

⎪
⎩ 2d (L − x), 1 L < x ≤ L.
L 2

Explain how you would go about applying Step (d) of the procedure, but

do not attempt to find the Fourier coefficients needed for the solution that

satisfies the initial condition.

30
 Section 5 Interpreting solutions

Exercise 4.6
Consider the diffusion equation
∂2Θ 1 ∂Θ
2
= ,
∂x α ∂t
subject to boundary conditions These boundary conditions
model the situation where the
Θ(0, t) = Θ0 , Θ(L, t) = ΘL , t ≥ 0, ends of the rod are kept at
fixed temperatures.
where Θ0 and ΘL are non-zero constants, and initial condition
Θ(x, 0) = f (x), 0 < x < L,
for some function f (x).
(a) Show that the function
L−x x
Θ(x, t) = Θ0 + ΘL
L L
satisfies the differential equation and the boundary conditions.
(b) Obtain a family of solutions to the differential equation subject to the
boundary conditions.
(c) Explain how you would go on to obtain a solution that also satisfies the
initial condition.

31
 Outcomes

Outcomes
After studying this unit you should be able to:
• understand how the wave and diffusion partial differential equations can
be used to model certain systems;
• determine appropriate simple boundary and initial conditions for such
models;
• find families of solutions for the wave equation, damped wave equation,
diffusion equation and similar homogeneous linear second-order partial
differential equations, subject to simple boundary conditions, using the
method of separating the variables;
• combine solutions of partial differential equations to satisfy given initial
conditions by finding the coefficients of a Fourier series.

33