Differential equations

53
An introduction to partial differential
equations
solutions. This chapter does no more than introduce
53.1 Introduction the topic.
A partial differential equation is an equation that
contains one or more partial derivatives. Examples
include: 53.2 Partial integration
∂u ∂u
(i) a +b =c Integration is the reverse process of differentiation.
∂x ∂y ∂u
Thus, if, for example, = 5 cos x sin t is integrated
∂t
∂2 u 1 ∂u partially with respect to t, then the 5 cos x term is
(ii) = 2
∂x 2 c ∂t considered as a constant,
(known as the heat conduction equation) ! !
and u = 5 cos x sin t dt = (5 cos x) sin t dt
∂2 u ∂2 u
(iii) + 2 =0
∂x 2 ∂y = (5 cos x)(−cos t) + c
(known as Laplace’s equation)
= −5 cos x cos t + f (x)
Equation (i) is a first order partial differential
equation, and equations (ii) and (iii) are second ∂2 u
order partial differential equations since the high- Similarly, if = 6x 2 cos 2y is integrated par-
∂x∂y
est power of the differential is 2. tially with respect to y,
Partial differential equations occur in many areas
of engineering and technology; electrostatics, heat ! !
∂u " #
conduction, magnetism, wave motion, hydrodynam- then = 6x cos 2y dy = 6x 2
2
cos 2y dy
ics and aerodynamics all use models that involve ∂x
partial differential equations. Such equations are $ %
" 2# 1
difficult to solve, but techniques have been devel- = 6x sin 2y + f (x)
oped for the simpler types. In fact, for all but for 2
the simplest cases, there are a number of numerical
methods of solutions of partial differential equations = 3x 2 sin 2y + f (x)
available.
To be able to solve simple partial differential ∂u
and integrating partially with respect to x gives:
equations knowledge of the following is required: ∂x
(a) partial integration, !
u= [3x 2 sin 2y + f (x)] dx
(b) first and second order partial differentiation — as
explained in Chapter 34, and
= x3 sin 2y + (x) f (x) + g(y)
(c) the solution of ordinary differential equations —
as explained in Chapters 46–51. f (x) and g(y) are functions that may be determined
if extra information, called boundary conditions or
It should be appreciated that whole books have been initial conditions, are known.
written on partial differential equations and their
 AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS 513

∂2 u
53.3 Solution of partial differential Hence, the solution of 2 = 6x 2 (2y − 1) for the
∂x
equations by direct partial given boundary conditions is:
integration x4
u= (2y − 1) + x sin y + cos y
The simplest form of partial differential equations 2
occurs when a solution can be determined by direct
partial integration. This is demonstrated in the fol- Problem 2. Solve the differential equation:
lowing worked problems. ∂2 u ∂u
= cos(x +y) given that = 2 when y = 0,
∂x∂y ∂x
Problem 1. Solve the differential equation and u = y2 when x = 0.
∂2 u
= 6x 2 (2y − 1) given the boundary condi-
∂x 2
∂u ∂2 u
tions that at x = 0, = sin 2y and u = cos y. Since = cos(x + y) then integrating partially
∂x ∂x∂y
with respect to y gives:
!
∂2 u ∂u
Since 2 = 6x 2 (2y − 1) then integrating partially = cos(x + y)dy = sin(x + y) + f (x)
∂x ∂x
with respect to x gives: ∂u
! ! From the boundary conditions, = 2 when y = 0,
∂u ∂x
= 6x 2 (2y − 1)dx = (2y − 1) 6x 2 dx hence
∂x 2 = sin x + f (x)
6x 3 from which, f (x) = 2 − sin x
= (2y − 1) + f (y)
3 ∂u
= 2x 3 (2y − 1) + f (y) i.e. = sin(x + y) + 2 − sin x
∂x
where f (y) is an arbitrary function. Integrating partially with respect to x gives:
From the boundary conditions, when x = 0, !
∂u u = [sin(x + y) + 2 − sin x]dx
= sin 2y.
∂x = −cos(x + y) + 2x + cos x + f (y)
Hence, sin 2y = 2(0)3 (2y − 1) + f (y) From the boundary conditions, u = y2 when x = 0,
from which, f (y) = sin 2y hence
∂u y2 = −cos y + 0 + cos 0 + f (y)
Now = 2x 3 (2y − 1) + sin 2y I
∂x = 1 − cos y + f (y)
Integrating partially with respect to x gives: from which, f (y) = y2 − 1 + cos y
!
∂2 u
u = [2x 3 (2y − 1) + sin 2y]dx Hence, the solution of = cos(x+y) is given by:
∂x∂y
2x 4 u = −cos(x + y) + 2x + cos x + y2 − 1 + cos y
= (2y − 1) + x(sin 2y) + F(y)
4
From the boundary conditions, when x = 0, Problem 3. Verify that
u = cos y, hence 1
φ(x, y, z) = " satisfies the partial
(0)4 x 2 + y2 + z 2
cos y = (2y − 1) + (0)sin 2y + F(y) ∂2 φ ∂2 φ ∂2 φ
2 differential equation: 2 + 2 + 2 = 0.
∂x ∂y ∂z
from which, F(y) = cos y
514 DIFFERENTIAL EQUATIONS

 
The partial differential equation 3x 2 − (x 2 + y2 + z2 )
 
∂2 φ ∂2 φ ∂2 φ  + 3y2 − (x 2 + y2 + z2 )
+ 2 + 2 = 0 is called Laplace’s equation.  
∂x 2 ∂y ∂z
+ 3z2 − (x 2 + y2 + z2 )
1 = 5
=0
− 21
If φ(x, y, z) = ! = (x 2 + y2 + z2 ) (x 2 + y2 + z2 ) 2
2 2
x +y +z 2
1
Thus, ! satisfies the Laplace equation
then differentiating partially with respect to x gives: x 2 + y2 + z 2
∂φ 1 3
= − (x 2 + y2 + z2 )− 2 (2x) ∂2 φ ∂2 φ ∂2 φ
∂x 2 + 2 + 2 =0
∂x 2 ∂y ∂z
3
= −x(x 2 + y2 + z2 )− 2
Now try the following exercise.
" #
∂2 φ 3 2 2 2 − 25
and = (−x) − (x + y + z ) (2x)
∂x 2 2 Exercise 200 Further problems on the solu-
tion of partial differential equations by direct
3 partial integration
+ (x 2 + y2 + z2 )− 2 (−1)
1. Determine the general solution of
by the product rule ∂u
= 4ty [u = 2ty2 + f (t)]
∂y
3x 2 1
= 5
− 3
(x 2 + y2 + z2 ) 2 (x 2 + y2 + z2 ) 2 ∂u
2. Solve = 2t cos θ given that u = 2t when
∂t
(3x 2 ) − (x 2 + y2 + z2 ) θ = 0. [u = t 2 (cos θ − 1) + 2t]
= 5
(x 2 + y2 + z2 ) 2
3. Verify that u(θ, t) = θ 2 + θt is a solution of
∂u ∂u
Similarly, it may be shown that −2 = t.
∂θ ∂t

∂2 φ (3y2 ) − (x 2 + y2 + z2 ) 4. Verify that u = e−y cos x is a solution of

= ∂2 u ∂2 u
∂y2 5
(x 2 + y2 + z2 ) 2 + = 0.
∂x 2 ∂y2
∂2 φ (3z2 ) − (x 2 + y2 + z2 ) ∂2 u
and = 5. Solve = 8ey sin 2x given that at y = 0,
∂z2 5
(x 2 + y2 + z2 ) 2 ∂x∂y
∂u π
= sin x, and at x = , u = 2y2 .
Thus, ∂x 2

[u = −4ey cos 2x − cos x + 4 cos 2x

∂2 φ ∂2 φ ∂2 φ (3x 2 ) − (x 2 + y2 + z2 ) *
+ + = + 2y2 − 4ey + 4
∂x 2 ∂y2 ∂z2 5
(x 2 + y2 + z2 ) 2
∂2 u
(3y2 ) − (x 2 + y2 + z2 ) 6. Solve = y(4x 2 − 1) given that at x = 0,
+ 5
∂x 2
(x 2 + y2 + z2 ) 2 ∂u
u = sin y and = cos 2y.
∂x
" + 4 , #
(3z2 ) − (x 2 + y2 + z2 ) x x2
+ 5 u=y − + x cos 2y + sin y
(x 2 + y2 + z2 ) 2 3 2
 AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS 515

∂2 u ∂u (c) Laplace’s equation, used extensively with elec-

7. Solve = sin(x + t) given that =1 trostatic fields is of the form:
∂x∂t ∂x
when t = 0, and when u = 2t when x = 0.
∂2 u ∂2 u ∂2 u
[u = −sin(x + t) + x + sin x + 2t + sin t] + 2 + 2 = 0.
∂x 2 ∂y ∂z
x
8. Show that u(x, y) = xy + is a solution of (d) The transmission equation, where the poten-
y
tial u in a transmission cable is of the form:
∂2 u ∂2 u
2x + y 2 = 2x. ∂2 u ∂2 u ∂u
∂x∂y ∂y = A + B + Cu where A, B and C are
∂x 2 ∂t 2 ∂t
9. Find the particular solution of the differ-
constants.
∂2 u
ential equation = cos x cos y given the
∂x∂y Some of these equations are used in the next sections.
∂u
initial conditions that when y = π, = x,
∂x
and when x = π, u = 2 cos y. 53.5 Separating the variables
! "
x2 π2 Let u(x, t) = X(x)T (t), where X(x) is a function of
u = sin x sin y + + 2 cos y −
2 2 x only and T (t) is a function of t only, be a trial
∂2 u 1 ∂2 u
10. Verify that φ(x, y) = x cos y + e x sin y satis- solution to the wave equation 2 = 2 2 . If the
fies the differential equation ∂x c ∂t
∂u
trial solution is simplified to u = XT , then = X "T
∂2 φ ∂2 φ ∂x
+ 2 + x cos y = 0. ∂2 u ∂u ∂2 u
∂x 2 ∂y and 2 = X "" T . Also = XT " and 2 = XT "" .
∂x ∂t ∂t
Substituting into the partial differential equation
∂2 u 1 ∂2 u
= gives:
∂x 2 c2 ∂t 2
53.4 Some important engineering
1
partial differential equations X "" T = 2 XT ""
c
There are many types of partial differential equa- Separating the variables gives:
tions. Some typically found in engineering and
science include: X !! 1 T !!
= 2
(a) The wave equation, where the equation of X c T
motion is given by: X "" 1 T ""
Let µ = = 2 where µ is a constant. I
∂2 u 1 ∂2 u X c T
= 2 2
∂x 2 c ∂t X ""
Thus, since µ = (a function of x only), it must be
T X
where c2 = , with T being the tension in a string 1 T ""
ρ independent of t; and, since µ = 2 (a function
and ρ being the mass/unit length of the string. c T
of t only), it must be independent of x.
(b) The heat conduction equation is of the form: If µ is independent of x and t, it can only be a con-
X ""
∂2 u 1 ∂u stant. If µ = then X "" = µX or X "" − µX = 0 and
= 2 X
∂x 2 c ∂t 1 T ""
if µ = 2 then T "" = c2 µT or T "" − c2 µT = 0.
h c T
where c2 = , with h being the thermal conduc-
σρ Such ordinary differential equations are of the form
tivity of the material, σ the specific heat of the found in Chapter 50, and their solutions will depend
material, and ρ the mass/unit length of material. on whether µ > 0, µ = 0 or µ < 0.
516 DIFFERENTIAL EQUATIONS

Worked Problem 4 will be a reminder of solving

ordinary differential equations of this type.

u = f(x, t)
P
Problem 4. Find the general solution of the
following differential equations: u (x, t )
(a) X !! − 4X =0 (b) T !! + 4T = 0.
0 L x
x

(a) If X !! − 4X = 0 then the auxiliary equation (see Figure 53.1

Chapter 50) is:
time t can be expressed as u = f (x, t), where x is its
m2 − 4 = 0 i.e. m2 = 4 from which, distance from 0.
m = +2 or m = −2 The equation of motion is as stated in
∂2 u 1 ∂2 u
Thus, the general solution is: section 53.4 (a), i.e. 2 = 2 2
∂x c ∂t
X = Ae2x + Be−2x
The boundary and initial conditions are:
(b) If T !! + 4T = 0 then the auxiliary equation is:
(i) The string is fixed at both ends, i.e. x = 0 and
m2 +√4 = 0 i.e. m2 = −4 from which, x = L for all values of time t.
m = −4 = ±j2 Hence, u(x, t) becomes:
#
Thus, the general solution is: u(0, t) = 0
T = e0 {A cos 2t + B sin 2t} = A cos 2t + B sin 2t u(L, t) = 0 for all values of t ≥ 0
(ii) If the initial deflection of P at t = 0 is denoted
Now try the following exercise. by f (x) then u(x, 0) = f (x)

Exercise 201 Further problems on revising (iii) Let the initial velocity of P be g(x), then
$ %
the solution of ordinary differential equation ∂u
= g(x)
1. Solve T !! = c2 µT given c = 3 and µ = 1 ∂t t=0
[T = Ae3t + Be−3t ] Initially a trial solution of the form u(x, t) = X(x)T (t)
is assumed, where X(x) is a function of x only and
2. Solve T !! − c2 µT = 0 given c = 3 and µ = −1 T (t) is a function of t only. The trial solution may be
[T = A cos 3t + B sin 3t] simplified to u = XT and the variables separated as
explained in the previous section to give:
3. Solve X !! = µX given µ = !1 "
X = Aex + Be−x X !! 1 T !!
= 2
X c T
4. Solve X !! − µX = 0 given µ = −1 When both sides are equated to a constant µ this
[X = A cos x + B sin x] results in two ordinary differential equations:
T !! − c2 µT = 0 and X !! − µX = 0
Three cases are possible, depending on the
53.6 The wave equation value of µ.

An elastic string is a string with elastic proper- Case 1: µ > 0

ties, i.e. the string satisfies Hooke’s law. Figure 53.1
shows a flexible elastic string stretched between two For convenience, let µ = p2 , where p is a real
points at x = 0 and x = L with uniform tension T . constant. Then the equations
The string will vibrate if the string is displace slightly X !! − p2 X = 0 and T !! − c2 p2 T = 0
from its initial position of rest and released, the end
points remaining fixed. The position of any point P have solutions: X = Aepx + Be−px and
on the string depends on its distance from one end, T = Cecpt + De−cpt where A, B, C and D are
and on the instant in time. Its displacement u at any constants.
 AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS 517

But X = 0 at x = 0, hence 0 = A + B i.e. B = −A and Substituting in equation (4) gives:

X = 0 at x = L, hence ! # $
nπx" cnπt cnπt
0 = AepL + Be−pL = A(epL − e−pL ). u = B sin C cos + D sin
Assuming (epL – e−pL ) is not zero, then A = 0 and L L L
# $
since B = −A, then B = 0 also. nπx cnπt cnπt
This corresponds to the string being stationary; since i.e. u = sin An cos + Bn sin
L L L
it is non-oscillatory, this solution will be disregarded.
(where constant An = BC and Bn = BD). There
will be many solutions, depending on the value
Case 2: µ = 0 of n. Thus, more generally,
%∞ # &
In this case, since µ = p2 = 0, T "" = 0 and X "" = 0. nπx cnπt
We will assume that T (t) # = 0. Since X "" = 0, X " = a un (x, t) = sin An cos
L L
and X = ax + b where a and b are constants. But n=1 '$
X = 0 at x = 0, hence b = 0 and X = ax and X = 0 cnπt
+ Bn sin
at x = L, hence a = 0. Thus, again, the solution is L
non-oscillatory and is also disregarded. (5)
To find An and Bn we put in the initial conditions
Case 3: µ < 0 not yet taken into account.
For convenience, (i) At t = 0, u(x, 0) = f (x) for 0 ≤ x ≤ L
let µ = −p2 then X "" + p2 X = 0 from which, Hence, from equation (5),
∞ !
X = A cos px + B sin px (1)
% nπx "
u(x, 0) = f (x) = An sin (6)
L
n=1
and T "" + c2 p2 T = 0 from which, ( )
∂u
(ii) Also at t = 0, = g(x) for 0 ≤ x ≤ L
T = C cos cpt + D sin cpt (2) ∂t t=0
Differentiating equation (5) with respect to t
(see worked Problem 4 above). gives:
Thus, the suggested solution u = XT now becomes: ∞ # & & '
∂u % nπx cnπ cnπt
= sin An − sin
u = {A cos px + B sin px}{C cos cpt + D sin cpt} ∂t L L L
n=1 & ''$
(3) cnπ cnπt
+ Bn cos
Applying the boundary conditions: L L
and when t = 0,
(i) u = 0 when x = 0 for all values of t,
thus 0 = {A cos 0 + B sin 0}{C cos cpt
g(x) =
%∞ !
sin
nπx cnπ "
Bn
I
+ D sin cpt} L L
n=1
i.e. 0 = A{C cos cpt + D sin cpt} ∞
from which, A = 0, (since {C cos cpt cπ % ! nπx "
i.e. g(x) = Bn n sin (7)
L L
+ D sin cpt} # = 0) n=1

Hence, u = {B sin px}{C cos cpt From Fourier series (see page 684) it may be shown
+ D sin cpt} (4) that:
nπx
An is twice the mean value of f (x) sin between
(ii) u = 0 when x = L for all values of t L
x = 0 and x = L
Hence, 0 = {B sin pL}{C cos cpt + D sin cpt}
* L
Now B #= 0 or u(x, t) would be identically zero. 2 nπx
nπ i.e. An = f (x)sin dx
Thus sin pL = 0 i.e. pL = nπ or p = for L 0 L
L
integer values of n. for n = 1, 2, 3, . . . (8)
518 DIFFERENTIAL EQUATIONS
! cnπ "
and Bn is twice the mean value of where c2 = 1, to determine the resulting motion
L u(x, t).
nπx
g(x)sin between x = 0 and x = L
L
# $% L 4
L 2 nπx

u (x, 0)
i.e. Bn = g(x)sin dx
cnπ L 0 L u = f (x)
2
% L
2 nπx
or Bn = g(x)sin dx (9)
cnπ 0 L 0 25 50 x (cm)

Figure 53.2
Summary of solution of the wave equation

The above may seem complicated; however a prac-

tical problem may be solved using the following Following the above procedure,
8-point procedure:
1. The boundary and initial conditions given are:
1. Identify clearly the initial and boundary &
conditions. u(0, t) = 0
i.e. fixed end points
2. Assume a solution of the form u = XT and express u(50, t) = 0
the equations in terms of X and T and their
derivatives. 2
u(x, 0) = f (x) = x 0 ≤ x ≤ 25
25
3. Separate the variables by transposing the equation
and equate each side to a constant, say, µ; two 2 100 − 2x
separate equations are obtained, one in x and the =− x+4=
25 25
other in t.
25 ≤ x ≤ 50
4. Let µ = −p2 to give an oscillatory solution.
5. The two solutions are of the form: (Note: y = mx + c is a straight line graph, so the
gradient, m, between 0 and 25 is 2/25 and the
X = A cos px + B sin px 2
y-axis intercept is zero, thus y = f (x) = x + 0;
and T = C cos cpt + D sin cpt. 25
between 25 and 50, the gradient = −2/25 and the
Then u(x, t) = {A cos px + B sin px}{C cos cpt + 2
D sin cpt}. y-axis intercept is at 4, thus f (x) = − x + 4).
25
6. Apply the boundary conditions to determine con- ' (
∂u
stants A and B. = 0 i.e. zero initial velocity.
∂t t=0
7. Determine the general solution as an infinite sum.
8. Apply the remaining initial and boundary condi- 2. Assuming a solution u = XT , where X is a
tions and determine the coefficients An and Bn function of x only, and T is a function of t only,
from equations (8) and (9), using Fourier series ∂u ∂2 u ∂u
then = X # T and 2 = X ## T and = XT # and
techniques. ∂x ∂x ∂y
∂2 u
= XT ##. Substituting into the partial differen-
Problem 5. Figure 53.2 shows a stretched ∂y2
string of length 50 cm which is set oscillating by ∂2 u 1 ∂2 u
displacing its mid-point a distance of 2 cm from tial equation, 2 = 2 2 gives:
∂x c ∂t
its rest position and releasing it with zero veloc-
∂2 u 1 ∂2 u 1
ity. Solve the wave equation: = X ## T = XT ## i.e. X ## T = XT ## since c2 = 1.
∂x 2 c2 ∂t 2 c2
 AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS 519
&%
X !! T !! 2 25 ' 2
(
nπx
3. Separating the variables gives: = = x sin dx
X T 50 0 25 50
Let constant, % 50 ' ( )
100 − 2x nπx
X !! T !! X !! T !! + sin dx
µ= = then µ = and µ = 25 25 50
X T X T
from which, Each integral is determined using integration by
parts (see Chapter 43, page 418) with the result:
X !! − µX = 0 and T !! − µT = 0. 16 nπ
An = 2 2 sin
4. Letting µ = −p2 to give an oscillatory solution n π 2
gives: From equation (9),
% L
X !! + p2 X = 0 and T !! + p2 T = 0 2 nπx
Bn = g(x) sin dx
cnπ 0 L
The auxiliary equation
! for each is: m2 + p2 = 0 * +
∂u
from which, m = −p2 = ±jp. = 0 = g(x) thus, Bn = 0
∂t t=0
5. Solving each equation gives: Substituting into equation (b) gives:
X = A cos px +B sin px and T = C cos pt + D sin pt. ∞ "
Thus, $ nπx nπt
u(x, t) = {A cos px+B sin px}{C cos pt+D sin pt}. un (x, t) = sin An cos
50 50
n=1 #
6. Applying the boundary conditions to determine nπt
constants A and B gives: + Bn sin
50
$∞ "
(i) u(0, t) = 0, hence 0 = A{C cos pt + D sin pt} nπx 16 nπ nπt
from which we conclude that A = 0. = sin sin cos
50 n2 π2 2 50
Therefore, n=1 #
nπt
u(x, t) = B sin px{C cos pt + D sin pt} (a) + (0) sin
50
Hence,
(ii) u(50, t) = 0, hence ∞
16 $ 1 nπx nπ nπt
0 = B sin 50p{C cos pt + D sin pt}. B # = 0, u(x, t) = sin sin cos
hence sin 50p = 0 from which, 50p = nπ and π2 n2 50 2 50
n=1
nπ
p=
50
For stretched string problems as in problem 5 above,
7. Substituting in equation (a) gives: the main parts of the procedure are:
" # 1. Determine An from equation (8).
u(x, t) = B sin
nπx
C cos
nπt
+ D sin
nπt %
2 L nπx
I
50 50 50 Note that f (x) sin dx is always equal
L 0 L
or, more generally, 8d nπ
" to 2 2 sin (see Fig. 53.3)
∞
$ nπx nπt n π 2
un (x, t) = sin An cos 2. Determine Bn from equation (9)
50 50 3. Substitute in equation (5) to determine u(x, t)
n=1
#
nπt
+ Bn sin (b) y
50 y = f (x)
where An = BC and Bn = BD. d

8. From equation (8),

% 0 L L x
2
2 L nπx
An = f (x) sin dx Figure 53.3
L 0 L
520 DIFFERENTIAL EQUATIONS

Now try the following exercise.

u = f(x, t )
P
Exercise 202 Further problems on the wave
equation u (x, t )

1. An elastic string is stretched between two 0 L x

points 40 cm apart. Its centre point is dis- x
placed 1.5 cm from its position of rest at
right angles to the original direction of the Figure 53.4
string and then released with zero velocity.
Determine the subsequent motion u(x, t) by bar, shown in Fig. 53.4, where the bar extends from
∂2 u 1 ∂2 u x = 0 to x = L, the temperature of the ends of the
applying the wave equation 2 = 2 2 bar is maintained at zero, and the initial temperature
∂x c ∂t
with c2!= 9. distribution along the bar is defined by f (x).
∞ Thus, the boundary conditions can be expressed as:
12 " 1 nπ nπx
u(x, t) = 2 sin sin %
π n 2 2 40 u(0, t) = 0
n=1
# for all t ≥ 0
3nπt u(L, t) = 0
cos and u(x, 0) = f (x) for 0 ≤ x ≤ L
40
2. The centre point of an elastic string between As with the wave equation, a solution of the form
two points P and Q, 80 cm apart, is deflected u(x, t) = X(x)T (t) is assumed, where X is a function
a distance of 1 cm from its position of rest of x only and T is a function of t only. If the trial
perpendicular to PQ and released initially solution is simplified to u = XT , then
with zero velocity. Apply the wave equation
∂2 u 1 ∂2 u ∂u ∂2 u ∂u
= where c = 8, to determine the = X $T 2
= X $$ T and = XT $
∂x 2 c2 ∂t 2 ∂x ∂x ∂t
motion of a point distance x from P at time t. Substituting into the partial differential equation,
! ∞
$ ∂2 u 1 ∂u
8 " 1 nπ nπx nπt = 2 gives:
u(x, t) = 2 sin sin cos ∂x 2 c ∂t
π n2 2 80 10
n=1 1
X $$ T = 2 XT $
c
Separating the variables gives:
53.7 The heat conduction equation
X !! 1 T!
∂2 u 1 ∂u = 2
The heat conduction equation 2 = 2 is solved X c T
∂x c ∂t
in a similar manner to that for the wave equa- X $$ 1 T$
tion; the equation differs only in that the right hand Let −p2 = = 2 where −p2 is a constant.
side contains a first partial derivative instead of the X c T
X $$
second. If −p2 = then X $$ = −p2 X or X $$ + p2 X = 0,
The conduction of heat in a uniform bar depends X
giving X = A cos px + B sin px
on the initial distribution of temperature and on the
physical properties of the bar, i.e. the thermal con- 1 T$ T$
and if −p2 = 2 then = −p2 c2 and integrat-
ductivity, h, the specific heat of the material, σ, and c T T
the mass per unit length, ρ, of the bar. In the above ing with respect to t gives:
h & $ &
equation, c2 = T
σρ dt = −p2 c2 dt
With a uniform bar insulated, except at its ends, T
any heat flow is along the bar and, at any instant, the from which, ln T = −p2 c2 t + c1
temperature u at a point P is a function of its distance The left hand integral is obtained by an algebraic
x from one end, and of the time t. Consider such a substitution (see Chapter 39).
 AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS 521

If ln T = −p2 c2 t + c1 then
2 2 2 2 2 2
T = e−p c t+c1 = e−p c t ec1 i.e. T = k e−p c t (where 15
c
constant k = e ). 1

u (x, 0)
2 2
Hence, u(x, t) = XT = {A cos px + B sin px}k e−p c t
2 2
i.e. u(x, t) = {P cos px + Q sin px}e−p c t where
P = Ak and Q = Bk.
Applying the boundary conditions u(0, t) = 0 0 1 x (m)
2 2 2 2
gives: 0 = {P cos 0 + Q sin 0}e−p c t = P e−p c t from
2 2
which, P = 0 and u(x, t) = Q sin px e−p c t .
2 2
Also, u(L, t) = 0 thus, 0 = Q sin pL e−p c t and

u (x, t )
since Q "= 0 then sin pL = 0 from which, pL = nπ P
nπ
or p = where n = 1, 2, 3, . . . u (x, t )
L
There are therefore many values of u(x, t).
0 1 x (m)
Thus, in general,
x
∞ "
! 2 c2 t nπx #
u(x, t) = Qn e−p sin Figure 53.5
L
n=1
Applying the remaining boundary condition, that Assuming a solution of the form u = XT , then, from
when t = 0, u(x, t) = f (x) for 0 ≤ x ≤ L, gives: above,
∞ "
! nπx # X = A cos px + B sin px
f (x) = Qn sin
L 2 c2 t
n=1 and T = k e−p .
From Fourier series, Qn = 2 × mean value of
nπx Thus, the general solution is given by:
f (x) sin from x to L.
L $
2 L nπx u(x, t) = {P cos px + Q sin px}e−p
2 c2 t
Hence, Qn = f (x) sin dx
L 0 L
2 c2 t
Thus, u(x, t) = u(0, t) = 0 thus 0 = P e−p
∞ %&$ L ' (
2! nπx −p2 c2 t nπx from which, P = 0 and u(x, t) = {Q sin px}e−p c t .
2 2
f (x) sin dx e sin 2 2
L
n=1 0 L L Also, u(1, t) = 0 thus 0 = {Q sin p}e−p c t .
Since Q " = 0, sin p = 0 from which, p = nπ
This method of solution is demonstrated in the where n = 1, 2, 3, . . .
following worked problem. !∞ "
2 2
# I
Hence, u(x, t) = Qn e−p c t sin nπx
Problem 6. A metal bar, insulated along its n=1
sides, is 1 m long. It is initially at room tem- The final initial condition given was that at t = 0,
perature of 15◦ C and at time t = 0, the ends are u = 15, i.e. u(x, 0) = f (x) = 15.
!∞
placed into ice at 0◦ C. Find an expression for the
temperature at a point P at a distance x m from Hence, 15 = {Qn sin nπx} where, from Fourier
one end at any time t seconds after t = 0. n=1
coefficients, Qn = 2 × mean value of 15 sin nπx from
x = 0 to x = 1,
The temperature u along the length of bar is shown
in Fig. 53.5. $ 1 ) cos nπx *1
2
∂2 u 1 ∂u i.e. Qn = 15 sin nπx dx = 30 −
The heat conduction equation is 2 = 2 and 1 0 nπ 0
∂x c ∂t
the given boundary conditions are: 30
=− [cos nπ − cos 0]
u(0, t) = 0, u(1, t) = 0 and u(x, 0) = 15 nπ
522 DIFFERENTIAL EQUATIONS

30 take c2 = 1.
= (1 − cos nπ)
nπ  
∞ * +
! n2 π 2 t
= 0 (when n is even) and
60
(when n is odd) u(x, t) = 320 1
sin
nπ
sin
nπx −
e 400 
nπ π2 n 2 2 20
n(odd)=1

Hence, the required solution is:

∞ "
! #
2 c2 t
u(x, t) = Qn e−p sin nπx 53.8 Laplace’s equation
n=1

∞ The distribution of electrical potential, or tempera-

60 ! 1 2 2 2 ture, over a plane area subject to certain boundary
= (sin nπx) e−n π c t conditions, can be described by Laplace’s equation.
π n
n(odd)=1 The potential at a point P in a plane (see Fig. 53.6)
can be indicated by an ordinate axis and is a function
Now try the following exercise. of its position, i.e. z = u(x, y), where u(x, y) is the
solution of the Laplace two-dimensional equation
Exercise 203 Further problems on the heat ∂2 u ∂2 u
conduction equation + = 0.
∂x 2 ∂y2
1. A metal bar, insulated along its sides, is 4 m The method of solution of Laplace’s equation is
long. It is initially at a temperature of 10◦ C similar to the previous examples, as shown below.
and at time t = 0, the ends are placed into Figure 53.7 shows a rectangle OPQR bounded
ice at 0◦ C. Find an expression for the tem- by the lines x = 0, y = 0, x = a, and y = b, for which
perature at a point P at a distance x m from we are required to find a solution of the equation
one end at any time t seconds after t = 0. ∂2 u ∂2 u
  + = 0. The solution z = (x, y) will give, say,
∞ ∂x 2 ∂y2
!
u(x, t) = 40 1 −n π c t nπx 
2 2 2
e 16 sin
π n 4
n(odd)=1 y
z
2. An insulated uniform metal bar, 8 m long,
has the temperature of its ends maintained
at 0◦ C, and at time t = 0 the temperature P
distribution f (x) along the bar is defined by
f (x) = x(8 − x). If c2 = 1, solve the heat con-
∂2 u 1 ∂u 0 x
duction equation 2 = 2 to determine
∂x c ∂t
the temperature u at any point in the bar at Figure 53.6
time t.
 
( )3 ! ∞
8 1 n2 π 2 t nπx z y
u(x, t) = e− 64 sin 
π n3 8
n(odd)=1 R
y=b Q
3. The ends of an insulated rod PQ, 20 units
long, are maintained at 0◦ C. At time t = 0,
u (x, y )

the temperature within the rod rises uniformly

from each end reaching 4◦ C at the mid-point
of PQ. Find an expression for the temperature
u(x, t) at any point in the rod, distant x from
P at any time t after t = 0. Assume the heat P
∂2 u 1 ∂u 0 x=a x
conduction equation to be 2 = 2 and
∂x c ∂t Figure 53.7
 AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS 523

the potential at any point within the rectangle OPQR. Since there are many solutions for integer values of n,
The boundary conditions are: ∞
!
u(x, y) = Qn sin px sinh p(b − y)
u = 0 when x = 0 i.e. u(0, y) = 0 for 0≤y≤b n=1
u = 0 when x = a i.e. u(a, y) = 0 for 0≤y≤b !∞
nπx nπ
= Qn sin sinh (b − y)
u = 0 when y = b i.e. u(x, b) = 0 for 0≤x≤a a a
n=1
u = f (x) when y = 0 i.e. u(x, 0) = f (x) The fourth boundary condition is: u(x, 0) = f (x),
for 0 ≤ x ≤ a ∞
! nπx nπb
As with previous partial differential equations, a hence, f (x) = Qn sinsinh
a a
solution of the form u(x, y) = X(x)Y (y) is assumed, n=1
∞ " #
where X is a function of x only, and Y is a function ! nπb nπx
of y only. Simplifying to u = XY , determining par- i.e. f (x) = Qn sinh sin
a a
n=1
∂2 u ∂2 u
tial derivatives, and substituting into 2 + 2 = 0 From Fourier series coefficients,
∂x ∂y " #
gives: X "" Y + XY "" = 0 nπb
X "" Y "" Qn sinh = 2 × the mean value of
Separating the variables gives: =− a
X Y nπx
f (x) sin from x = 0 to x = a
Letting each side equal a constant, −p2 , gives the two $ a a
equations: nπx
i.e. = f (x) sin dx from which,
0 a
X "" + p2 X = 0 and Y "" − p2 Y = 0 Qn may be determined.
from which, X = A cos px + B sin px and This is demonstrated in the following worked
Y = C epy + D e−py or Y = C cosh py + D sinh py problem.
(see Problem 5, page 478 for this conversion).
This latter form can also be expressed as: Problem 7. A square plate is bounded by the
Y = E sinh p( y + φ) by using compound angles. lines x = 0, y = 0, x = 1 and y = 1. Apply the
∂2 u ∂2 u
Hence u(x, y) = XY Laplace equation 2 + 2 = 0 to determine
∂x ∂y
= {A cos px + B sin px}{E sinh p(y + φ)} the potential distribution u(x, y) over the plate,
subject to the following boundary conditions:
or u(x, y) u = 0 when x = 0 0 ≤ y ≤ 1,
u = 0 when x = 1 0 ≤ y ≤1,
= {P cos px + Q sin px}{sinh p(y + φ)} u = 0 when y = 0 0 ≤ x ≤ 1, I
u = 4 when y = 1 0 ≤ x ≤ 1.
where P = AE and Q = BE.
The first boundary condition is: u(0, y) = 0, hence Initially a solution of the form u(x, y) = X(x)Y (y)
0 = P sinh p(y + φ) from which, P = 0. Hence, is assumed, where X is a function of x only, and
u(x, y) = Q sin px sinh p(y + φ). Y is a function of y only. Simplifying to u = XY ,
The second boundary condition is: u(a, y) = 0, determining partial derivatives, and substituting into
hence 0 = Q sin pa sinh p(y + φ) from which, ∂2 u ∂2 u
nπ + = 0 gives: X "" Y + XY "" = 0
sin pa = 0, hence, pa = nπ or p = for ∂x 2 ∂y2
a X "" Y ""
n = 1, 2, 3, . . . Separating the variables gives: =−
The third boundary condition is: u(x, b) = 0, X Y
hence, 0 = Q sin px sinh p(b + φ) from which, Letting each side equal a constant, −p2 , gives the
sinh p(b + φ) = 0 and φ = −b. two equations:
Hence, u(x, y) = Q sin px sinh p(y − b) =
Q1 sin px sinh p(b − y) where Q1 = −Q. X "" + p2 X = 0 and Y "" − p2 Y = 0
524 DIFFERENTIAL EQUATIONS

from which, X = A cos px + B sin px = 0 (for even values of n),

and Y = Ce py + De−py 16
= (for odd values of n)
or Y = C cosh py + D sinh py nπ
16 16
or Y = E sinh p(y + φ) Hence, Qn = = cosech nπ
nπ(sinh nπ) nπ
Hence u(x, y) = XY
Hence, from equation (a),
= {A cos px + B sin px}{E sinh p(y + φ)}
∞
!
or u(x, y) u(x, y) = Qn sin nπx sinh nπy
= {P cos px + Q sin px}{sinh p(y + φ)} n=1
where P = AE and Q = BE. ∞
!
16 1
= (cosech nπ sin nπx sinh nπy)
The first boundary condition is: u(0, y) = 0, hence π n
n(odd)=1
0 = P sinh p(y + φ) from which, P = 0.
Hence, u(x, y) = Q sin px sinh p(y + φ).
The second boundary condition is: u(1, y) = 0, hence Now try the following exercise.
0 = Q sin p(1) sinh p(y + φ) from which,
sin p = 0, hence, p = nπ for n = 1, 2, 3, . . . Exercise 204 Further problems on the
The third boundary condition is: u(x, 0) = 0, hence, Laplace equation
0 = Q sin px sinh p(φ) from which, 1. A rectangular plate is bounded by the
sinh p(φ) = 0 and φ = 0. lines x = 0, y = 0, x = 1 and y = 3. Apply the
Hence, u(x, y) = Q sin px sinh py.
Since there are many solutions for integer values of n, ∂2 u ∂2 u
Laplace equation + = 0 to deter-
∂x 2 ∂y2
∞
! mine the potential distribution u(x, y) over
u(x, y) = Qn sin px sinh py the plate, subject to the following boundary
n=1 conditions:
∞
! u = 0 when x = 0 0 ≤ y ≤ 2,
= Qn sin nπx sinh nπy (a) u = 0 when x = 1 0 ≤ y ≤ 2,
n=1
u = 0 when y = 2 0 ≤ x ≤ 1,
u = 5 when y = 3 0 ≤ x ≤ 1
The fourth boundary condition is: u(x, 1) = 4 = f (x),  
!∞
∞
! u(x, y) = 20 1
cosech nπ sin nπx sinh nπ(y − 2)
hence, f (x) = Qn sin nπx sinh nπ(1). π n(odd)=1 n
n=1
From Fourier series coefficients, 2. A rectangular plate is bounded by the
lines x = 0, y = 0, x = 3, y = 2. Determine the
Qn sinh nπ = 2 × the mean value of potential distribution u(x, y) over the rec-
tangle using the Laplace equation
f (x) sin nπx from x = 0 to x = 1
∂2 u ∂2 u
" + = 0, subject to the following
2 1 ∂x 2 ∂y2
i.e. = 4 sin nπx dx boundary conditions:
1 0
# cos nπx $1 u(0, y) = 0 0 ≤ y ≤ 2,
=8 − u(3, y) = 0 0 ≤ y ≤ 2,
nπ 0 u(x, 2) = 0 0 ≤ x ≤ 3,
8 u(x, 0) = x(3 − x) 0 ≤ x ≤ 3
=− (cos nπ − cos 0)  
nπ ∞
!
u(x, y) = 216 1
cosech
2nπ
sin
nπx
sinh
nπ
(2 − y)
8 π3 n3 3 3 3
= (1 − cos nπ) n(odd)=1
nπ