CHAPTER 7

Fourier Transform Methods

7.1 INTRODUCTION
Joseph Fourier, a French mathematician, had invented a method called Fourier transform in
1801, to explain the flow of heat around an anchor ring. Since then, it has become a powerful
tool in diverse fields of science and engineering. It can provide a means of solving unwieldy
equations that describe dynamic responses to electricity, heat or light. In some cases, it can
also identify the regular contributions to a fluctuating signal, thereby helping to make sense
of observations in astronomy, medicine and chemistry. Fourier transform has become indispensable
in the numerical calculations needed to design electrical circuits, to analyze mechanical
vibrations, and to study wave propagation.
Fourier transform techniques have been widely used to solve problems involving semi-
infinite or totally infinite range of the variables or unbounded regions. In order to deal with
such problems, it is necessary to generalize Fourier series to include infinite intervals and to
introduce the concept of Fourier integral. In this chapter, we deal with Fourier integral
representations and Fourier transforms together with some applications to Diffusion, Wave
and Laplace equations.

7.2 FOURIER INTEGRAL REPRESENTATIONS

Definition 7.1 (Dirichlet’s conditions). A function f ( x) is said to have satisfied Dirichlet’s

conditions in the interval ( L, L), provided f ( x) is periodic, piecewise continuous, and has
a finite number of relative maxima and minima in ( L, L).

Let a function f ( x) be periodic with period 2 L, i.e., f ( x  2 L)  f ( x), and satisfy

Dirichlet’s conditions in the interval ( L, L). Then f (x) has a Fourier series representation for
every x in the form
388
 FOURIER TRANSFORM METHODS 389

a0 f § nπ x nπ x ·
f ( x) 
2 n 1 ¦ ©¨ an cos L
 bn sin
L ¹
¸ (7.1)

where
1 L nQ t
an  ± f (t ) cos dt , n  0, 1, 2, ... (7.2)
L  L L

1 L nQ t
bn  f (t )sin dt , n  1, 2, ... (7.3)
L ± L L

Here, an , bn are called Fourier coefficients. Fourier series representation, however, can be
extended to some non-periodic functions also, provided the integral of the modulus of such
a function f (t ) satisfies the condition
f
³ f | f (t ) | dt
is finite.
Substituting Eqs. (7.2) and (7.3) into Fourier series (7.1), we get
f
1 L ª1 L nπ t nπ x 1 L nπ t nπ x º
f ( x)
2L ³ L f (t ) dt  ¦ «¬ L ³  L f (t ) cos L
cos
L
dt 
L ³  L f (t )sin L
sin
L
dt »
¼
n 1

Noting that cos ( A  B )  cos A cos B  sin A sin B, and interchanging the order of summation
and integration, we obtain
f
1 L 1 L nπ (t  x)
f ( x)
2L ³ L f (t ) dt 
L ³ L f (t ) ¦ cos
L
dt (7.4)
n 1

Further, if we assume that the function f ( x) is absolutely integrable, and allowing L to tend
to infinity, i.e.,

f
³ f | f (t ) | dt  f (7.5)

we get
1 L
Lt
L of 2L ³  L f (t ) dt 0 (7.6)

In the remaining part of the infinite sum of Eq. (7.4), if we set Δ s = π /L, the equation reduces to
%s
1 Q e
% s n0 Q ³ Q %s
f ( x) Lt f (t ) ¦ cos {n%s (t  x)}dt (7.7)
n 1
390 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

As L o f, Δ s o 0, implying that Δs is a small positive number and the points nΔs are
equally spaced along the s-axis. The series under the integral can be approximated by an
integral of the form (as Δs o 0)
f
³ 0 cos {s (t  x)} ds
Thus, Eq. (7.7) can be rewritten as
1 f f
f ( x)
π ³ f f (t )³ 0 cos {s (t  x)} ds dt (7.8)

1 f f

π ³ 0 ³ f f (t ) cos {s (t  x)} dt ds (7.9)

which is the Fourier integral representation of f ( x ).

7.2.1 Fourier Integral Theorem

Theorem 7.1 If f ( x) satisfies Dirichlet’s conditions for f  x  f and if the integral

f
³ f f ( x) dx is absolutely convergent, then

1 f f 1
π ³ 0 dα ³ f f (t ) cos α (t  x) dt
2
[ f ( x  0)  f ( x  0)] (7.10)

To establish this result, the central results required are the Riemann-Lebesgue lemma and the
Riemann localization lemma; first we shall state and prove the former.

Riemann-Lebesgue lemma If f ( x) satisfies Dirichlet’s conditions in the interval (a, b),

then each of the integrals
b b
± a f ( x)sin Nx dx, ± a f ( x) cos Nx dx

tends to zero as N o f.

Proof Suppose a1 , a2 , ..., a p are the points in (a, b) taken in the order at which the

function f ( x) has either a turning value or a finite discontinuity. Let a  a0 , and b  a p 1.

Then we may write

± ¥±
p
b ar 1
f ( x)sin Nx dx  f ( x)sin Nx dx (7.11)
a ar
r 0
 FOURIER TRANSFORM METHODS 391

Now in each of the sub-intervals ( ar , ar 1 ), r  0, 1, 2, ..., p, f ( x) is a continuous function and

is either monotonically increasing or decreasing. Thus, aplying the second Mean Value Theorem
of integral calculus to each of these intervals, we have
ar 1 Y ar 1
±a f ( x)sin Nx dx  f (ar ) ± sin Nx dx  f (ar 1 )± sin Nx dx (7.12)
r ar Y

where Y is some value of x in the range ( ar , ar +1 ). Carrying out the integrations on the right-
hand side of Eq. (7.12), we get

¦ cos Nar  cos NY µ ¦ cos NY  cos Nar 1 µ

f (ar ) § ¶ f (ar 1 ) § ¶
¨ N · ¨ N ·

Now taking the limit as N ne, we get

ar 1
Lt
N of ³a r
f ( x) sin Nx dx 0

Since the number of terms on the right-hand side of Eq. (7.11) is finite, interchanging of
summation and limit process can be carried out and, therefore,
p p
b ar 1 ar 1
Lt
N of ³a f ( x) sin Nx dx Lt
N of
¦ ³a r
f ( x)sin Nx dx ¦ NLtof ³ a r
f ( x)sin Nx dx 0
r 0 r 0

i.e.,
b
Lt
N of ³ a f ( x) sin Nx dx 0 (7.13)

which is the Riemann-Lebesgue lemma.

Riemann localization lemma If f (t ) satisfies Dirichlet’s conditions in the interval (0, a),
where a is finite, then
a sin Nt π
∫ 0 f (t ) t
dt → f (0+)
2

as N o f.
Proof We may write
a sin Nt a sin Nt a sin Nt
±0 f (t ) dt  f (0) ± dt  ± { f (t )  f (0)} dt  I1  I 2
t 0 t 0 t
392 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Since the function f b(t ) is continuous in (0, a), from the definition of derivative

f (t )  f (0 )
t

is continuous in (0, a). By the Riemann-Lebesgue lemma (since the integrand of I 2 is

bounded as N o f ) I 2 o 0 as N o f. Hence,

a sin Nt a sin Nt Na sin u

Lt
N of ³0 f (t )
t
dt f (0) ³0 t
dt f (0) ³0 u
du

f sin u π
f (0) ³0 u
du
2
f (0) (7.14)

which is the Riemann localization lemma.

f
Proof (Fourier integral theorem). Since the integral ³ f f ( x) dx is absolutely convergent,

f
³ f | f ( x)| dx is finite and converges for all B in the range (0, N). Also, | cos B (t  x) | c 1,
implying that the integral
f
³ f f (t ) cos α (t  x) dt
converges and is independent of B and x. Thus, after changing the order of integration, the
double integral.
N ª f º
I ³ 0 «¬ ³ f f (t ) cos α (t  x) dt »¼ dα (7.15)

can be expressed as
f ª N º f ª sin N (t  x) º
I ³ f «
¬ ³0 f (t ) cos α (t  x) dα » dt
¼ ³ f f (t ) «¬ tx ¼»
dt

Let v  t  x. Then the above integral becomes

f sin Nv § δ 0 δ f · sin Nv
I ³ f f (v  x )
v
dv ¨© ³ f ³ δ ³ 0 ³ δ
   ¸¹ f (v  x)
v
dv I1  I 2  I3  I 4

When N o f, I1 and I 4 both tend to zero in view of the Riemann-Lebesgue lemma. Thus the
only contribution to the integral will be from the neighbourhood of v  0. Using the Riemann
localization lemma, we get
 FOURIER TRANSFORM METHODS 393

E sin Nv Q
I3  Lt ± f (v  x) dv  f ( x ) (7.16)
N ne 0 v 2
and the second integral
0 sin Nv δ sin Nv π
I2 = ∫ −δ f (v + x )
v
dv = ∫0 f ( x − v)
v
dv = f ( x −)
2
(7.17)

Incorporating these results into Eq. (7.15), we obtain

fª f º π
³ 0 «¬ ³ f f (t ) cos α (t  x) dt »¼ dα 2
[ f ( x  )  f ( x )] (7.18)

If f is a continuous function of x, then

f ( x  )  f ( x )  f ( x)
and Eq. (7.18) reduces to
1 f ª f º
f ( x)
π ³ 0 dα «¬ ³ f f (t ) cos α (t  x) dt »¼ (7.19)

If x is a point of discontinuity, then

1
f ( x)  [ f ( x  )  f ( x )] (7.20)
2
i.e., the intergral (7.19) converges to the average value of the right- and left-hand limits. Thus,
the proof of the Fourier integral theorem is complete.
In order to bring out the analogy between Fourier series and Fourier integral theorem, we
rewirte Eq. (7.19) as
1 f f
f ( x)
π ³ 0 ³ f f (t ) (cos α t cos α x  sin α t sin α x) dt dα
If we define
1 f
A (α )
π ³ f f (t ) cos α t dt (7.21)

1 f
B (α )
π ³ f f (t ) sin α t dt (7.22)

the above equation can also be written as

f
f ( x) ³ 0 [ A (α ) cos α x  B(α ) sin α x] dt dα (7.23)
394 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

7.2.2 Sine and Cosine Integral Representations

If f ( x)   f ( x), i.e., if f ( x) is an odd function, Eq. (7.21) gives

1 f
A (α )
π ³ f f ( x) cos α x dx
If f ( x) is an odd function, then

1 f 1 f
A (α ) 
π ³f f ( x) cos α x dx 
π ³f f ( x) cos α x dx

1 f

π ³ f f ( x) cos α x dx  A(α )

implying 2 A(B )  0 or A(B )  0, i.e.,

1 f
A (α )
π ³ f f ( x) cos α x dx 0 (7.24)

Also,
1 f
B (α )
π ³ f f ( x) sin α x dx
1 f

π ³f f ( x) sin α x dx if f ( x) is an odd function

1 f 2 f

π ³ f f ( x) sin α x dx π ³0 f ( x) sin α x dx (7.25)

Thus, Eq. (7.23) reduces to

f
f ( x) ³0 B(α ) sin α x dα (7.26)

which is the Fourier sine integral representation, where A(B ) and B(B ) are defined by the
relations (7.24) and (7.25). Similarly, if f ( x) is an even function, i.e., if f ( x)  f ( x), then
we obtain the Fourier cosine integral representation
f
f ( x) ³0 A(α ) cos α x dα (7.27)

where
2 f
B(α ) 0, A(α )
π ³0 f ( x) cos α x dx (7.28)
 FOURIER TRANSFORM METHODS 395

7.3 FOURIER TRANSFORM PAIRS

From the Fourier Integral Theorem 7.1, we have
1 f f
f ( x)
π ³ 0 ³ f f (t ) cos α (t  x) dt dα (7.29)

In terms of the complex exponential function, Eq. (7.29) takes the form

1 f f
iα (t  x )
f ( x) ³ 0 ³ f f (t ) [e  eiα (t  x ) ] dt dα
2π
1 ª f f f f º
iα (t  x )  iα (t  x )
2𠫬 ³ 0 ³ f f (t )e dt dα  ³ 0 ³ f f (t ) e dt dα »
¼

Let B  B in the second integral; then it becomes

f f 0 f
 ³ 0 ³ f f (t )eiα (t  x ) dt dα ³ f ³ f f (t ) e
iα (t  x )
dt dα

Hence,
1 f f
iα (t  x )
f ( x)
2π ³ f ³ f f (t ) e dt dα (7.30)

This is the exponential form of the Fourier integral theorem. Equation (7.30) can be rewritten
as
1 f ª 1 f º
iα t
f ( x) ³ f «¬ 2π ³ f f (t ) e dt » e iα x dα (7.31)
2π ¼
Thus, we formally define the Fourier transform pair as follows:

Definition 7.2 Let f ( x) be a function defined on (f, f) and is piecewise continuous,

differentiable in each finite interval and is absolutely integrable on ( f, f). From Eq. (7.31), if

1 e iB t
F (B ) ³ e f (t ) e dt (7.32)
2Q
then we have, for all x,
1 f
iα x
f ( x)
2π ³ f F (α ) e dα (7.33)

Here, F (B ) defined by Eq. (7.32) is the Fourier transform of f ( x ), and f ( x) defined

by Eq. (7.33) is called the Inverse Fourier transform of F (B ) and is denoted by
396 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

F (B )  * [ f (t ); B ] (7.34)

f ( x)  * 1 [ F (B ); x] (7.35)
which constitute the Fourier transform pair.
We have seen in Section 7.2.2 that if f ( x) is an odd function, the Fourier integral
representation of f ( x) reduces to
f
f ( x) ³ 0 B(α ) sin α x dα
or
2 f f
f ( x)
π ³ 0 sin α x³ 0 f (t ) sin α t dt dα (7.36)

If
2 ‡
Fs (B ) ³0 f (t ) sin B t dt * s [ f (t ); B ] (7.37)
Q
then
2 ‡
Fs (B ) sin B x d B * s1[ Fs (B ); x]
Q ³0
f ( x) (7.38)

Here, Eq. (7.37) is the Fourier sine transform of f ( x) and its inverse sine transform is given
by Eq. (7.38).
Similarly, when f ( x) is an even function, we can obtain the Fourier cosine transform and
the corresponding inverse as

2 ‡
Fc (B ) f (t ) cos B t dt * c [ f (t ); B ]
Q ³0
(7.39)

2 ‡
Fc (B ) cos B x d B * c1[ Fc (B ); x]
Q ³0
f ( x) (7.40)

7.4 TRANSFORM OF ELEMENTARY FUNCTIONS

EXAMPLE 7.1 Find the Fourier transform of

2
f ( x) = e− x /2
 FOURIER TRANSFORM METHODS 397

Solution Following the definition of Fourier transform, we have

‡
1 iB x
* [ f ( x); B ] ³ ‡ f ( x) e dx
2Q
1 ‡ 2
/2 iB x
³ ‡ e x e dx
2Q
‡
e  ( x  iB ) e B
1 2 2
³ ‡
/2 /2
dx
2Q
Let
x  iB
t
2
Then
dx
 dt
2
Thus

e B
2
/2 ‡ 2
* [ f ( x); B ] ³ ‡ e t dt
Q
or
2
e−α /2 2
F (α ) = π = e−α /2
π

EXAMPLE 7.2 Find the Fourier transform of

f ( x) e a | x | , f  x  f

Solution We know that

¬  x, x0
| x | ­
® x, xs0
Therefore,
1 f
iα x
F [ f ( x); α ]
2π ³ f f ( x) e dx

1 0 1 f
³ f eax eiα x dx  ³0 e
 ax iα x
e dx
2π 2π
398 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

1 0 1 f
( a iα ) x (  a  iα ) x
2π ³ f e dx 
2π ³0 e dx

1 § 1 1 · 2§ a ·
¨©  ¸
2π a  iα  a  iα ¹ 𠨩 a 2  α 2 ¸¹

EXAMPLE 7.3 Find the Fourier transform of f ( x) defined by

¬1, | x|ca
f ( x)  ­
®0, |x| a
and hence evaluate
f sin α a cos α x f sin α a
³ f α
dα , ³0 α
dα

Solution From the definition of the Fourier transform,

1 ‡ iD x
F (D ) * [ f ( x ); D ] Ô f (x) e
2S ‡
dx

a
1 a iD x 1 È eiD x Ø
2S
Ô a e dx É Ù
2S Ê iD Ú  a

1 È eiD a e iD a Ø 2 È eiD a  e iD a Ø

É  Ù É ÙÚ
2S Ê iD iD Ú D 2S Ê 2i

Therefore,

¬ 2 sin B a
¯ , B 0
¯ B 2Q
F (B )  ­
¯ Lt 2a sin B a  2a , B 0
¯B n0 2Q B a
® 2Q

Now,
1 ‡ iB x
f ( x) * 1[ F (B ); x] ³ ‡ F (B ) e dB
2Q
Thus,

1 f 2 sin α a iα x ­1, |x|d a

f ( x)
2π ³ f α 2π
e dα ®
¯0, | x|! a
 FOURIER TRANSFORM METHODS 399

i.e.,

1 f sin α a (cos α x  i sin α x) ­1, | x|d a

π ³ f α
dx ®
¯0, | x|! a

Hence,
f sin α a cos α x ­π , |x|d a
³ f α
dα ®
¯0, | x|! a

Also, by setting x  0 in the above equations, we obtain

f sin α a
³ f α
dα π

Since the integrand is even, we can have

f sin α a π
³0 α
dα
2

EXAMPLE 7.4 Find the Fourier cosine and sine transforms of e bx and evaluate the integrals
‡ cos D x
(i) Ô 0 D 2 + b2 d D
f α sin α x
(ii) ³0 α 2 + b2
dα

Solution Given f ( x)  ebx and following the definitions of Fourier cosine and sine
transforms, viz.
2 f
Fc (B ) * c [ f ( x); B ] f ( x) cos B x dx
Q ³0
2 f
Fs (B ) * s [ f ( x); B ] f ( x) sin B x dx
Q ³0
we obtain
2 f  bx
* c [ebx ; B ] cos B x dx
Q ³0
e (7.41)

2 f  bx
* s [e bx ; B ] sin B x dx
Q ³0
e (7.42)
400 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Let
f
³0 e
bx
I1 cos α x dx

f
³0 e
bx
I2 sin α x dx

Integrating I1 by parts, we have

f
§ 1 bx · α f 1 α
³0 e
 bx
I1 ¨©  e cos α x ¸  sin α x dx  I2 (7.43)
b ¹ 0 b b b

Integrating I 2 by parts, we have

f
§ 1 bx · α f α
³0 e
 bx
I2 ¨©  e sin α x ¸  cos α x dx I1 (7.44)
b ¹ 0 b b

Solving Eqs. (7.43) and (7.44) for I1 and I 2 , we obtain

b B
I1  , I2 
B b2 2
B  b2
2

Hence,
2 b 2 B
Fc (B )  , Fs (B ) 
Q B b 2 2 Q B  b2
2

Then,

2 f
f ( x)
π ³0 Fc (α ) cos α x dα

i.e.

2 f 2 b
ebx
π ³0 π α  b2
2
cos α x dα

or
f cos α x π bx
³0 α b2 2 2b
e

Similarly, it can be shown that

f α sin α x π bx
³0 α 2  b2 2
e
 FOURIER TRANSFORM METHODS 401

EXAMPLE 7.5 Find the Fourier sine transform of f ( x ), if

¬0, 0 xa
¯
f ( x )  ­ x, ac xcb
¯0, xb
®

Solution Following the definition of the Fourier sine transform, we have

2 f
Fs (α )
π ³0 f ( x) sin α x dx

2 b

π ³ a x sin α x dx
2 ª § x cos α x · b 1 b º
π
Ǭ 
¬© α
¸¹ 
a α
³ a cos α x dx»¼
2 § a cos α a  b cos α b sin α b  sin α a ·

𠩨 α α2 ¹¸

7.5 PROPERTIES OF FOURIER TRANSFORM

In many practical situations, determination of Fourier transform of certain functions is very
complex. Once we know the transform of some elementary functions, we can find the transform
of many other functions with the help of the properties associated with the Fourier transform.
We now discuss some of the important properties of the Fourier transform.

Theorem 7.2 (Linearity property). If F (B ) and G (B ) are the Fourier transforms of

f ( x) and g ( x) respectively, then

* [c1 f ( x )  c2 g ( x ); B ]  c1 F (B )  c2 G (B )

* 1[c1F (B )  c2G (B ); x]  c1 f ( x)  c2 g ( x)

where c1 and c2 are constants.

Proof
1 ‡ iD x
* [c1 f ( x )  c2 g( x ); D ] Ô e
2S ‡
[ c1 f ( x )  c2 g( x )] dx

c1 ‡ iD x c2 ‡ iD x
Ô e
2S ‡
f ( x ) dx  Ô e
2S ‡
g( x ) dx

c1F (D )  c2 G(D )
402 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Theorem 7.3 (Change of scale). If F (B ) is the Fourier transform of f ( x ), then the Fourier
transform of f (ax) is

1 ¦B µ
F§ ¶
a ¨a·

Proof From the definition of the Fourier transform, we have

1 ‡ iB x
F (B ) * [ f ( x); B ] ³ ‡ e f ( x) dx
2Q
Hence,

1 f iB x
* [ f (ax); B ] ³ f e f (ax) dx
2Q

Setting ax  t , we have dx  dt / a. Therefore,

1 1 ‡ È D Ø
* [ f (ax ); D ]
a 2S Ô ‡ exp ÉÊ i a t ÙÚ f (t ) dt
Hence,
1
* [ f ( ax ); D ] F (D / a), a!0 (7.45)
a
Similarly, it can be shown that

1
* [ f (ax ); D ]  F (D / a), a0 (7.46)
a
Combining Eqs. (7.45) and (7.46), we have the property
1
* [ f (ax ); D ] F (D / a), a›0 (7.47)
|a|
It can also be established that
1
* s [ f (ax ); D ] Fs (D / a), a!0 (7.48)
a
1
* c [ f (ax ); D ] Fc (D / a), a!0 (7.49)
a
Theorem 7.4 (Shifting property). If F (B ) is the Fourier transform of f ( x ), then the Fourier
transform of f ( x  a ), i.e.,

* [ f ( x  a); D ] eiD a F (D )
 FOURIER TRANSFORM METHODS 403

Proof From the definition of Fourier transform

1 ‡ iD x
F (D ) * [ f ( x ); D ] Ô e
2S ‡
f ( x ) dx

Therefore,
1 ‡ iD x
* [ f ( x  a); D ] Ô e
2S ‡
f ( x  a) dx

Setting
x  a  t,
we have dx dt. Then
1 ‡ iD ( a  t )
* [ f ( x  a); D ] Ô e
2S ‡
f (t ) dt

Hence,

* [ f ( x  a); D ] eiD a F (D ) (7.50)

Similarly, it can be shown that
* 1[eiD a F (D ); x ] f ( x  a) (7.51)

Theorem 7.5 (Modulation property). If F (B ) is the Fourier transform of f ( x ), then the

Fourier transform of f ( x) cos ax is
1
[ F (B  a )  F (B  a )]
2
Proof From the definition of Fourier transform, we have
1 ‡ iD x
F (D ) * [ f ( x ); D ] Ô e
2S ‡
f ( x ) dx

Therefore,

1 ‡iD x
È eiD x  e iD x Ø
* [ f ( x ) cos ax; D ] Ô
2S ‡
e f ( x ) ÉÊ
2
ÙÚ dx

1Ë 1 ‡ i (D  a ) x 1 ‡ i (D  a ) x Û
2 ÌÍ 2S Ô ‡ e f ( x ) dx  Ô e
2S ‡
f ( x ) dx Ü
Ý
1
[ F (D  a)  F (D  a)] (7.52)
2
404 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

In the same fashion, it can be established that

2 ‡
S Ô0
*s [ f ( x ) cos ax; D ] f ( x ) cos ax sin D x dx

2 1 ‡
S 2 Ô0
f ( x ) [sin (D  a) x  sin (D  a) x ] dx

1Ë 2
Ì
2Í S ^Ô ‡
0
f ( x ) sin (D  a) x dx  Ô
‡
0 ` Û
f ( x ) sin (D  a) x dx Ü
Ý
1
[ Fs (D  a)  Fs (D  a)] (7.53)
2
1
* c [ f ( x ) cos ax; D ] [ Fc (D  a)  Fc (D  a)] (7.54)
2
1
* s [ f ( x ) sin ax; D ] [ Fc (D  a)  Fc (D  a)] (7.55)
2
1
* c [ f ( x ) sin ax; D ] [ Fs (D  a)  Fs (D  a)] (7.56)
2

Theorem 7.6 (Differentiation). If f ( x) and its first ( r  1) derivatives are continuous, and
if its rth derivative is piecewise continuous, then

* [ f (r ) ( x ); D ] ( iD )r * [ f ( x ); D ], r 0, 1, 2, ...

provided f and its derivatives are absolutely integrable. In addition, we assume that f ( x) and
its first (r  1) derivatives vanish as x o r f.

Proof From the definition, we have the Fourier transform of d rf /dx r as

1 ‡ drf
* [ f (r ) ( x ); D ] Ô ‡ dx r eiD x dx F (r ) (D ) (say)
2S
Integrating by parts, we get
f
1 f drf 1 § d r 1 f iα x · 1 f d r 1 f
³ f dxr eiα x dx ¨© r 1 e ¸¹  2π ³ f dxr 1 (iα ) e
iα x
dx
2π 2π dx f

If we assume that d r −1/dx r −1 tends to zero as x o r f, we may write the above result in the
form
F ( r ) (B )  (iB ) F ( r 1) (B )  (iB )2 F ( r  2) (B )   (iB )r F (B )
 FOURIER TRANSFORM METHODS 405

Hence,
F ( r ) (B )  (iB )r F (B )
and, therefore,

* [ f (r ) ( x ); D ] ( iD )r F (D ) (7.57)

EXAMPLE 7.6 Let Fc( r ) (B ) be the Fourier cosine transform of d rf /dx r and Fs( r ) (B ) be

the Fourier sine transform of d rf /dx r. Then prove that

r −1
Fc(2 r ) (α ) = − ∑ (−1)n a2r −2n−1α 2n + (−1)r α 2r Fc (α )
n =0
r −1
Fc(2 r +1) (α ) = − ∑ (−1)n a2r −2nα 2n + (−1)r α 2r +1Fs (α )
n =0

assuming

§ d r 1 f ·
Lt ¨ r 1 ¸ 0
x of © dx ¹

2 d r −1 f
Lt = ar −1 (say)
x→ 0 π dx r −1

Solution From the definition, we have

2 f dr f
Fc( r ) (α )
π ³0 dx r
cos α x dx (7.58)

2 f dr f
Fs( r ) (α )
π ³0 dx r
sin α x dx (7.59)

Integrating Eq. (7.58) by parts, we get

­° ª r 1 f ½°
2 d f º f d r 1
Fc( r ) (α )
π
® « r 1 cos (α x) » 
°¯ ¬ dx ¼0 ³0 dx r 1
sin (α x)α dx¾
°¿

Now, we assume that

§ d r 1 f ·
Lt ¨ r 1 ¸ 0
x of © dx ¹
406 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

2 d r −1 f
Lt = ar −1 (say)
x→ 0 π dx r −1
Then

Fc( r ) (B )  ar 1  B Fs( r 1) (B ) (7.60)

Integrating Eq. (7.59) by parts, we have
f
2 § d r 1 f · f d r 1
Fs( r ) (α )
𠨩 r 1 sin α x ¸¹ 
dx 0
³0 dx r 1
cos (α x)α dx α Fcr 1 (α ) (7.61)

Substituting Eq. (7.61) into Eq. (7.60), we get

Fc( r ) (α ) = −ar −1 − α 2 Fc( r − 2) (α )

By repeated application of these results, we can show Fc( r ) (B ) to be a sum of a’s and either

Fc(1) (B ) or Fc (B ), depending on whether r is odd or even. Fc(1) will be present if r is odd

and may be replaced by a0  B Fs (B ). Similar arguments give us the formulae

¥ (1) a
r 1
Fc(2 r ) (B )   n
B 2n  (1)r 1B 2r Fc (B )
2 r  2 n 1
n 0

¥ (1) a
r 1
Fc(2r 1) (B )   n
2r 2n B 2n  (1)r 1B 2r 1Fs (B )
n 0

df d 3 f
Note: When x  0 and   0,
dx dx3

2 f d2 f
π ³0 dx 2
cos α x dx α 2 Fc (α )

2 f d4 f
π ³0 dx 4
cos α x dx α 4 Fc (α )

d2 f
Similarly, when x  0 and f   0,
dx 2

2 f d2 f
π ³0 dx 2
sin α x dx α 2 Fs (α )

2 f d4 f
π ³0 dx 4
sin α x dx α 4 Fs (α )
 FOURIER TRANSFORM METHODS 407

EXAMPLE 7.7 Find the Fourier cosine transform of exp (at 2 ).

Solution We have from the definition of Fourier cosine transform,

2 2 ‡  at 2
* c [e  at ; D ]
S Ô0
e cos D t dt I (say) (7.62)

Differentiating with respect to B, we obtain

dI 2 f 1 2 f
 at 2  at 2
dα

π ³ 0 te sin α t dt
2a π ³ 0 sin α t d (e )

1 2 ­  at 2 f ½
 at 2
®[(e sin α t )] 0f  α ³0 e cos α t dt ¾
2a π ¯ ¿

Therefore,
dI α
=− I
dα 2a
i.e.,
dI B
  dB
I 2a
On integration, we get

I  ce B
2
/ 4a
(7.63)
But when B  0 , from Eq. (7.62) we have
2 f 2
I
π ³0 e at dt

From Example 7.1, using change of scale property, we obtain

2 1 Q 1
I 
Q a 2 2a

1
From Eq. (7.63) we get c  . Hence,
2a
2 1 2
* c [e  at ; D ] e D / 4a
2a
408 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

EXAMPLE 7.8 If the Fourier sine transform of f ( x) is B /(1  B 2 ), find f ( x) .

Solution From the definition, we have

2 ∞ α
f ( x) =
π ³ 0 1 + α 2 sin α x dα
2 ∞ (α 2 + 1) − 1
=
π ³0 α (1 + α 2 )
sin α x dα

2 ∞ sin α x 2 ∞ sin α x
π ³0 α
dα −
π ³ 0 α (1 + α 2 ) dα
But
f sin α x π
³0 α
dα
2
Hence,
π 2 f sin α x
f ( x)
2

π ³ 0 α (1  α 2 ) dα (7.64)

df 2 f cos α x

dx

π ³0 1α2
dα (7.65)

d2 f 2 fα sin α x
dx 2 π ³0 1α2
dα (7.66)

From Eqs. (7.64) and (7.66), it follows that

d2 f 2 f sin α x π
dx 2
f
π ³0 α
dα 
2
0

whose solution is found to be

f  c1e x  c2 e x (7.67)
Therefore,
df
= c1e x − c2 e− x
dx

When, x  0, from Eq. (7.64) we have f (0)  Q /2, and from Eq. (7.65),

df (0) 2 f dα π
dx

π ³ 0 1α2 
2
 FOURIER TRANSFORM METHODS 409

Also, from Eq. (7.67), using these results, we get

π π
c1 + c2 = , c1 − c2 = −
2 2

Solving the above two equations, we get c1  0, c2  Q /2. Thus, f ( x)  Q /2 e x .

EXAMPLE 7.9 If the Fourier cosine transform of f ( x) is B n e aB , find f ( x).

Solution Using the definition, we have

2 f
n  aα
f ( x)
π ³0 α e cos α x dα (7.68)

But we know from calculus that

f a
 aα
³0 e cos α x dα
a  x2
2
(7.69)

Differentiating this result n times with respect to a, we get

f dn § a ·
(1)n ³0 α n e aα cos α x dα ¨ ¸
da n © a 2  x 2 ¹

1 dn § 1 1 ·
¨©  ¸
2 da a  ix a  ix ¹
n

1
[(1)n n ! (a  ix) n 1  (1)n n ! (a  ix) n 1 ]
2

(1)n n !
[(a  ix) n 1  (a  ix) n 1 ]
2
Let
a  ix  r (cos R  i sin R )
Then
a  r cos R , x  r sin R
Therefore,
x
r 2  a 2  x2 , tan R 
a
Thus,

(a  ix) n 1  r  n 1[cos (n  1)R  i sin (n  1) R ]

(a  ix)  n 1  r  n 1[cos ( n  1)R  i sin (n  1) R ]

410 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Then

f
n  aα n ! cos (n  1)θ
³0 α e cos α x dα
(a 2  x 2 )( n 1) 2

Hence from Eq. (7.68),

2 n ! cos (n  1)R
f ( x) 
Q (a 2  x 2 )( n 1/2)

EXAMPLE 7.10 Find the Fourier transform of

(i) w n u /w x n of the function u ( x, t ) assuming that u and its first (n – 1) derivatives with
respect to x vanish as x o r f.
(ii) w u /w t.

Also, find the sine and cosine Fourier transforms of w 2 u /w x 2 of the function u ( x, t ) .

Solution (a) We shall adopt the following notation: The Fourier transform of u ( x, t )
with respect to the variable x is defined as
1 ‡ iD x
* [u( x, t ); x  D ] Ô e
2S ‡
u( x, t ) dx U (D , t ) (7.70)

Then the Fourier transform of w u /w x is

Ëw u Û 1 ‡ wu iD x
* Ì ( x, t ); x  D Ü
Íw x Ý Ô
2S ‡ w x
e dx

Integration by parts yields

1 ­ f ½
iα x f iα x
®[u ( x, t ) e ]f  iα
2π ¯ ³ f u( x, t ) e dx ¾
¿
If we assume
Lt u ( x, t ) 0
xo r f

then we find that

Ëw u Û
* Ì ( x, t ); x  D Ü iD * [u( x, t ); x  D ] iD U (D , t ) (7.71)
Íw x Ý
 FOURIER TRANSFORM METHODS 411

Similarly, the Fourier transform of w 2u /w x 2 is

Ë w 2 u ( x, t ) Û 1 ‡ w2u 1 ‡ Èw uØ
*Ì ; x DÜ Ô eiD x dx Ô e
iD x
dÉ Ù
Í wx 2
Ý 2S ‡ w x 2 2S ‡ Êw xÚ
‡
1 Ë È w u iD x Ø ‡ Û
ÌÉ e Ù  iD (eiD x u)‡‡  (iD )2 Ô eiD x u dx Ü
2S Í Ê w x Ú ‡ ‡
Ý

Assuming that both u and w u /w t tend to zero as x o r f, we have

Ëw 2 u Û
* Ì 2 ( x, t ); x  D Ü ( 1)2 (iD )2 * [u( x, t ); x  D ] ( 1)2 (iD )2 U (D , t ) (7.72)
Íw x Ý

Thus, in general, the Fourier transform of the nth derivative of u ( x, t ) is given by

Ë w n u( x , t ) Û
*Ì ; x DÜ ( 1)n (iD )n * [u( x, t ); x  D ] ( 1)n (iD )n U (D , t ) (7.73)
Í wx n
Ý
(b) Now

Ëw u Û 1 ‡ iD x wu 1 w ‡ iD x
* Ì ( x, t ); x  D Ü
Íwt Ý Ô e
2S ‡ wt
( x, t ) dx Ô e u ( x, t ) dx
2S w t ‡
Ut (D , t )

Therefore,

Ëw u Û
* Ì ( x, t ); x  D Ü Ut (D , t ) (7.74)
Íwt Ý
(c) In the case of Fourier sine and cosine transforms, we have

Ëw 2 u Û 2 ‡ w2u
S Ô0
* s Ì 2 ( x, t ); x  D Ü sin D x dx
Íw x Ý w x2
‡
2 Ëw u Û 2 ‡ wu
S ÌÍ w x
sin D x Ü 
Ý0 S
D Ô0 cos D x
wx
dx

wu
We assume that o 0 as x o f. Then the RHS term of the above equation becomes
wx

2 f wu 2 ­ f ½

π
α ³0 cos α x
wx
dx  α ®[u ( x, t ) cos α x]0f  α
π ¯ ³0 u ( x, t )sin α x dx¾
¿
412 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Also, assuming that u ( x, t ) o 0 as x o f, this equation becomes

2
Q
B u ( x, t )  B 2* s [u ( x, t ); x n B]
x0
Hence,
Ëw 2 u Û 2
* s Ì 2 ( x, t ); x  D Ü D u ( x, t )  D 2* s [u ( x, t ); x  D ] (7.75)
Íw x Ý S x 0
Similarly, it can be shown that if
wu
u ( x, t ) o 0 and o 0 as x o f
wx
then
Ëw 2 u Û 2 w u ( x, t )
* c Ì 2 ( x, t ); x  D Ü   D 2* c [u ( x, t ); x  D ] (7.76)
Íw x Ý S wx x 0

Obviously, the choice of the sine or cosine transform is dccided by the form of the boundary
condition at the lower limit of the variable selected for exclusion. Thus, we observe that for
the exclusion of w 2 u /w x 2 from a given PDE, we require

ux 0 in the case of sine transform

wu
in the case of cosine transform
wx x 0

7.6 CONVOLUTION THEOREM (FALTUNG THEOREM)

If F (B ) and G (B ) are the Fourier transforms of the functions f ( x) and g ( x), then the product
F (B ) G (B ) is the Fourier transform of the convolution product f *g.

Proof The product

1 f
f *g
2π ³ f f (u) g ( x  u) du (7.77)

is called the convolution or Faltung of the functions f and g over the interval ( f, f). Then
the Fourier transform of this convolution integral yields
1 ‡ iD x ‡
* [( f * g); D ]
2S Ô ‡ e Ô ‡ f (u) g( x  u) du dx
1 ‡ ‡ iD x
2S Ô ‡ Ô ‡ e f (u) g( x  u) du dx (7.78)
 FOURIER TRANSFORM METHODS 413

Since f and g are absolutely integrable, the order of integration can be interchanged and,
therefore,
‡ Ë ‡
dx ÛÜ du
1 iD ( x  u ) iD u
* [( f *g); D ]
2S Ô ‡ f (u) ÍÌ Ô ‡ g( x  u) e e
Ý
Let x  u  y. Then dx  dy. Therefore,

‡ Ë ‡
dy ÛÜ du
1 iD u iD y
* [( f *g); D ]
2S Ô ‡ f (u) ÍÌ e Ô ‡ g( y) e Ý
1 ‡ iD u 1 ‡ iD y
Ô e
2S ‡
f (u) du Ô e
2S ‡
g( y) dy

F (D ) G(D ) (7.79)

Hence the theorem is proved.

We can verify that
f *g  g * f , (7.80)
i.e.

1 f
f *g
2π ³ f f (u) g ( x  u) du (7.81)

Setting x  u  B , we have du   dB . Then

1 f 1 f
f *g
2π ³ f f ( x  α ) g (α ) dα 2π ³ f g (u) f ( x  u) du g*f (7.82)

Hence, the convolution is commutative.

Special cases (sine and cosine convolution integrals)

(i)

f 2 f f
³0 Fc (α ) Gc (α ) cos α x dα
π ³0 Fc (α ) cos α x dα ³0 g (η ) cos ηα dη

2 f f

π ³0 g (η ) dη ³0 Fc (α ) cos xα cos ηα dα

1 f f

2π ³0 g (η ) dη ³0 Fc (α ) [cos | x  η | α  cos ( x  η ) α ] dα
414 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

But,

2 f 1 f
f ( x)
π ³0 Fc (α ) cos xα dα
2 ³0 g (η ) dη [ f (| x  η |)  f ( x  η )] (7.83)

(ii) If Fs (B ) and Gs (B ) are the Fourier sine transforms of f ( x) and g ( x) , then we can
show that
f 1 f
³0 Fs (α ) Gs (α )sin xα dα
2 ³0 f (η ) [ g (| x  η |)  g ( x  η )] dη (7.84)

‡
(iii) Ô0 Fc (D ) Gs (D ) sin xD dD

2 ‡ ‡

S Ô 0
Fc (D ) sin xD dD Ô g(K) sin KD dK
0

2 ‡ ‡

S Ô 0
g(K) dK Ô Fc (D ) sin xD sin KD dD
0

1 ‡ ‡
Ô
2S 0
g(K) dK Ô
0
Fc (D ) [cos | x  K | D  cos ( x  K) D ] dD

1 ‡
2 Ô0
g(K) [ f (| x  K |)  f ( x  K)] dK (7.85)

f 1 f
(iv) ³0 Fs (α ) Gc (α )sin xα dα
2 ³0 f (η ) [ g (| x  η |)  g ( x  η )] dη (7.86)

EXAMPLE 7.11 If Fc (B ) and Gc (B ) are the Fourier cosine transforms of f ( x) and g ( x)

respectively, show that Fc (B ) Gc (B ) is the transform of

1 f

2 ³0 f (u ) [ g (| x  u |)  g ( x  u )] du

Solution Refer Eq. (7.83).

7.7 PARSEVAL’S RELATION

From the Fourier convolution theorem, we have
1 f 1 f
³ f eiα x F (α ) G (α ) dα f *g ³0 f (u ) g ( x  u ) du
2π 2π
 FOURIER TRANSFORM METHODS 415

If we set x  0 , the above equation reduces to

1 f 1 f

2π ³ f F (α ) G (α ) dα
2π ³ f f (u) g (u) du
For the case g (u )  f *(u ), where f *(u ) is the complex conjugate of the function f (u ), we have
G(D ) * [ g(u); D ] * [ f *( u); D ]  F *(D )
Thus,
1 f 1 f

2π ³ f F (α )F *(α ) dα 2π ³ f f (u) f * (u) du (7.87)

Therefore,
f f
³ f | F (α ) | dα ³ f | f (u) | du
2 2
(7.88)

Equation (7.88) is known as Parseval’s relation.

EXAMPLE 7.12 Using the Fourier cosine transform of e ax and e bx , show that

f dα π
³0 (a  α ) (b  α )
2 2 2 2 2ab (a  b)
, a ! 0, b ! 0 (7.89)

Solution Let f ( x)  ebx , g ( x)  e ax ; then

2 f 2 f 2 b
³0 ³0 e
 bx
Fc (α ) f ( x) cos α x dx cos α x dx
π π π b2  α 2
Similarly, it can be shown that
2 a
Gc (B ) 
Q a B2
2

However,
f f 2 f
³0 Fc (α ) Gc (α ) dα ³0 Fc (α ) dα
π ³0 g ( x) cos α x dx

f 2 f
³0 g ( x) dx
π ³0 Fc (α ) cos α x dα

f
³0 f ( x) g ( x) dx

i.e.
f f
³ 0 Fc (α ) Gc (α ) dα ³ 0 f ( x) g ( x) dx
416 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Hence, it follows that

f f 1
³0 ³0 e
 ( a b ) x
Fc (α ) Gc (α ) dα dx
ab
Therefore,
f dα π
³0 (a  α ) (b  α )
2 2 2 2 2ab (a  b)
, a ! 0, b ! 0

7.8 TRANSFORM OF DIRAC DELTA FUNCTION

The Dirac delta function has been defined in Chapter 3. The details of its properties can be
found in Section 3.4. We may recall the shifting property of the delta function, i.e.
f
³ f δ (t  a) f (t ) dt f (a )

and then obtain its Fourier transform as

1 ‡ iD t
* [G (t  a); D ] Ô e G (t  a) dt eiD a / 2S (7.90)
2S ‡

When a  0, we obtain the formal result

1
* [G (t ); D ] (7.91)
2S

That is, the Fourier transform of the Dirac delta function E (t ) is constant and equal to 1/ 2π .
It then follow that
1 ‡
* 1[1; t ] Ô e
 iD t
dD 2SG (t ) (7.92)
2S ‡

7.9 MULTIPLE FOURIER TRANSFORMS

As already discussed, the idea of a Fourier transform and its inverse can be extended to
functions involving two or more variables.
Let f ( x, y ) be a function of two variables x and y. Let it be piecewise continuous and
satisfy the condition
f f
³ f ³ f | f ( x, y) | dx dy  f (7.93)

Then the Fourier transform pair can be written as

1 f f
i (α x  β y )
F (α , β )
2π ³ f ³ f e f ( x, y ) dx dy (7.94)
 FOURIER TRANSFORM METHODS 417

1 f f
i (α x  β y )
f ( x, y )
2π ³ f ³ f e F (α , β ) dα d β (7.95)

We can split Eq. (7.94) into two steps: first by treating f ( x, y ) as a function of x and then
treating the result as a function of y sequentially, i.e.,
1 f
iα x
f *(α , y )
2π ³ f e f ( x, y ) dx

1 f
iβ y
F (α , β )
2π ³ f e f *(α , y ) dy

Similarly, the inversion formula (7.95), can be written as

1 f
 iα x
f ( x, y )
2π ³ f e f *(α , y ) dα

1 f
i β y
f *(α , y )
2π ³ f e F (α , β ) d β

Assuming that the partial derivatives of f occurring in the equation are absolutely integrable
and that f , w f /w x, w f /w y tend to zero at infinity, the double Fourier transform of derivatives
yield the following results:
Ëw f Û
*Ì ( x, y); x  D , y  E Ü iD F (D , E ) (7.96)
Íw x Ý

Ë w2 f Û
*Ì ( x, y); x  D , y  E Ü DE F (D , E ) (7.97)
Í w xw y Ý
The convolution property leads to the following results:
1 ‡ ‡
* 1[ F (D , E ) G(D , E ); D  x, E  y]
2S Ô ‡ Ô ‡
f ( x  u) g ( x  u) g (u, v) du dv (7.98)

The Fourier transform in the case of three variables is

1 f f f
i (α x  β y γ z )
F (α , β , γ )
(2π ) 3/2 ³ f ³ f ³ f e f ( x, y, z ) dx dy dz (7.99)

7.10 FINITE FOURIER TRANSFORM

The Fourier transform technique outlined so far is applicable to problems involving infinite
or semi-infinite domains. But, in many practical situations, we come across finite intervals in
boundary value problems. Therefore, it is natural to extend Fourier transform method to
problems in which the range of an independent variable is finite. It is then possible to find
418 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

their inverses from the well-known theory of Fourier series. It may be recalled that if a
function f ( x) satisfies Dirichlet conditions in the interval 0 c x c Q , then it has Fourier sine
series
f
f ( x) ¦ bn sin nx (7.100)
n 1
in which
2 S
bn
S Ô
0
f ( x) sin nx dx, n 1, 2, 3, ž (7.101)

The Fourier series in Eq. (7.100) converges pointwise to f ( x) at points where f ( x) is

continuous and to the value (1/2)[ f ( x +) + f ( x −)] at other points. Similarly, f ( x) has Fourier
cosine series

a0 f
f ( x) 
2 n 1 ¦
an cos nx (7.102)

in which
2 Q
an  f ( x) cos nx dx, n  0, 1, 2, ... (7.103)
Q ±0

7.10.1 Finite Sine Transform

If f ( x) satisfies Dirichlet conditions in the interval 0 c x c Q , then we define finite sine
transform of f ( x) by
S
* s [ f ( x ); n] Fs (n) Ô0 f ( x )sin nx dx, n 1, 2, 3, ... (7.104)

which is a sequence of numbers. Comparing Eqs. (7.101) and (7.104), we notice that
2
bn  Fs (n), n  1, 2, 3, ... (7.105)
Q
Now from Eq. (7.100) we can have the result
2 ‡
* s1[ Fs (n); x ] f (x) Ç Fs (n) sin nx
Sn 1
(7.106)

which is the inverse finite sine transform. Similarly, we can define finite sine transform when
the independent variable x lies in the interval (0, L),
L nS x
* s [ f ( x ); n] Fs (n) Ô0 f ( x )sin
L
dx, n 1, 2, 3, ž (7.107)
 FOURIER TRANSFORM METHODS 419

with
‡
2 nS x
* s1[ Fs (n); x ] f (x)
L
Ç Fs (n) sin L
, 0xL (7.108)
n 1
as the corresponding inversion formula.

7.10.2 Finite Cosine Transform

If f ( x) satisfies the Dirichlet conditions in the interval 0 c x c Q , we define the finite cosine
transform of f ( x) as
S
* c [ f ( x ); n] Fc (n) Ô0 f ( x )cos nx dx, n 0, 1, 2, ... (7.109)

The corresponding inversion formula is

Fc (0) 2 ‡
* c1[ Fc (n); x ] f ( x)  Ç Fc (n) cos nx (7.110)
S Sn 1
which holds at each point in the interval (0, Q ) at which f ( x) is continuous. Similarly, when
the independent variable x lies in the interval (0, L) , the corresponding pair assumes the form
L nS x
* c [ f ( x ); n] Fc (n) Ô0 f ( x )cos
L
dx , n 0, 1, 2, ... (7.111)

Fc (0) 2 ‡ nS x
* c1[ Fc (n); x ] f (x)  Ç Fc (n) cos (7.112)
S Ln 1 L
Having defined the finite cosine transform, we shall attempt to find some results involving
derivatives up to Fourth order to facilitate the solution of a few boundary value problems,
which are actually presented under miscellaneous examples. For instance, if f is a function
of x and t, 0  x  L, t  0, then we have
L
Ëw f Û Lw f nS x Ë nS x Û nS L nS x
*s Ì ;n
Í w x ÜÝ Ô 0 w x sin L dx ÌÍ f ( x, t )sin L ÜÝ  L
0
Ô 0 f cos L
dx

Therefore,
Ëw f Û nS nS
*s Ì ;n  * c [ f ; n]  Fc (n) (7.113)
Í w x ÜÝ L L
L
Ëw f Û Lw f nS x Ë nS x Û nS L nS x
*c Ì ;n
Í w x ÜÝ Ô 0 w x cos L dx ÌÍ f ( x, t ) cos L ÝÜ  L
0
Ô 0 f sin L
dx

nS
* s [ f ; n]  { f (0, t )  f ( L, t ) cos nS } (7.114)
L
420 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Considering second order derivatives, we obtain

Ëw 2 f Û Lw2 f nS x Èw f nS x Ø nS
L
Lw f nS x
*s Ì 2 ; nÜ
Íw x Ý
Ô 0 w x2 sin
L
dx ÉÊ
wx
sin
L Ú0
Ù 
L Ô 0 w x cos L
dx

nS ÑÎÈ nS x Ø
L
nS L nS x ÑÞ

L
ÏÉ f cos
ÐÑ Ê L
ÙÚ 
0 L Ô0 f sin
L
dx ß
àÑ
nS Ë nS Û

L ÌÍ L * s [ f ; n]  [ f (0, t )  f ( L, t ) cos nS ]ÜÝ

Therefore,

Ëw 2 f Û n2S 2 nS
*s Ì 2 ; nÜ  * s [ f ; n]  { f (0, t )  f ( L, t ) cos nS } (7.115)
Íw x Ý L2 L

In particular, if f (0, t )  f ( L, t )  0, this result simplifies to

Ëw 2 f Û n2S 2
*s Ì 2 ; nÜ  * s [ f ; n] (7.116)
Íw x Ý L2
Similarly, it can be shown that

Ëw 2 f Û n2S 2
*c Ì 2 ; nÜ  * c [ f ; n]  { f x (0, t )  f x ( L, t ) cos nS } (7.117)
Íw x Ý L2

In case w f /w x vanishes at the ends x  0 and x  L, it simplifies to

Ëw 2 f Û n2S 2
*c Ì 2 ; nÜ  * c [ f ; n] (7.118)
Íw x Ý L2
By repeatedly applying these results, we can deduce that

Ëw 4 f Û n 4S 4
*s Ì 4 ; nÜ  * s [ f ; n], (7.119)
Íw x Ý L4

if f x and f xx vanish at both the ends x  0, x  L, and

Ëw 4 f Û n4S 4
*c Ì 4 ; nÜ  * c [ f ; n] (7.120)
Íw x Ý L4
 FOURIER TRANSFORM METHODS 421

provided
f x  0  f xxx when x  0; x  L

7.11 SOLUTION OF DIFFUSION EQUATION

Let us consider the problem of flow of heat in an infinite medium  f  x  f, when the
initial temperature distribution f ( x) is known and no heat sources are present. Mathematically,
we have to solve the problem described in the following example.

EXAMPLE 7.13 Solve the heat conduction equation given by

w 2u wu
PDE: k ,  f  x  f, t!0
wx 2 wt
subject to
BCs: u ( x, t ) and u x ( x, t ) both o 0 as | x | o f
IC: u ( x, 0) f ( x), f xf

Solution Taking the Fourier transform* of PDE, we get

 kD 2* [u( x, t ); x  D ] * [ut ( x, t ); x  D ]
or

U t (B , t )  kB 2U (B , t )  0 (7.121)
In deriving this, the BCs are already utilized (as can be seen from Eqs. (7.72) and (7.74). The
Fourier transform of the IC gives
U (α , 0) F (α ), ‡ α  ‡ (7.122)
The solution of Eq. (7.121) can be readily seen to be

U  Ae kB
2
t

When t  0, we have from Eq. (7.122) the relation U  F (B ), implying A  F (B ). Therefore,

U (B , t )  F (B )e kB
2
t (7.123)
Inverting this relation, we obtain
1 f 1 f
 kα 2t  iα x
³ f F (α ) e dα ³ f F (α ) exp (kα t  iα x) dα
2
u ( x, t ) e (7.124)
2π 2π

*When the range of spatial variable is infinite, the Fourier exponential transform is used rather than the
sine or cosine transform.
422 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

The product form of the integrand in Eq. (7.124) suggests the use of convolution. If the
Fourier transform of g ( x ) is e kB t , then g ( x) will be given by
2

1 f
 kα 2t iα x
g ( x)
2π ³ f e e dα

But, if a  0, b is real or complex, and we know that

f π
³ f exp (ax
2
 2bx) dx exp (b 2 /a ) (7.125)
a
Here, a  kt , 2b  ix. Therefore,

1 Q ¦ x2 µ 1 ¦ x2 µ
g ( x)  exp §  ¶ exp §  ¶
2Q kt ¨ 4kt · 2kt ¨ 4kt ·

Using the convolution theorem, we have

1 ‡
u ( x, t )
2π ³ ‡ f (α ) g ( x  α ) dα (7.126)

Hence,

1 f 1 ª ( x  α )2 º
u ( x, t )
2π ³ f f (α )
2kt
exp «
¬ 4kt ¼
» dα

1 f ª ( x  α )2 º
4π kt ³ f f (α ) exp « 
¬ 4kt ¼
» dα (7.127)

Introducing the change of variable

Z
Bx
4kt
we can rewrite solution (7.127) in the form
1 f 2
u ( x, t )
π ³ f f ( x  4kt z ) e z dz (7.128)

EXAMPLE 7.14 (Flow of heat in a semi-infinite medium). Solve the heat conduction
problem described by
w 2u wu
PDE: k , 0  x  f, t!0
wx 2 wt
BC: u (0, t ) u0 , tt0

IC: u ( x, 0) 0, 0 xf
u and w u /w x both tend to zero as x o f.
 FOURIER TRANSFORM METHODS 423

Solution Since u is specified at x  0, the Fourier sine transform is applicable to this

problem. Taking the Fourier sine transform of the given PDE and using the notation.

2 f
U s (α , t )
π ³ 0 u ( x, t )sin α x dx
we obtain from Eqs. (7.74) and (7.75) the relation

Ë 2 Û w Us
kÌ D u ( x, t )  D 2* s [u ( x, t ); x  D ]Ü (D , t )
Í S x 0
Ý wt
or
dU s 2
 kB 2U s  kB u0 (7.129)
dt Q
Its general solution is found to be

(1  e  kB t )
2 u0 2
U s (B , t )  (7.130)
Q B
Inverting by Fourier inverse sine transform, we obtain

2 f
u (α , t )
π ³ 0 U s (α , t ) sin α x dα
Therefore,
2 f sin α x 2
u ( x, t ) u ³0 (1  e kα t ) dα (7.131)
π 0 α
Noting that
2 y u 2
erf ( y )  e du
Q ±0
and using the standard integral
f
α 2 sin (2α y ) π
³0 e α
dα
2
erf ( y )

we have solution (7.131) in the form

2u0 © Q Q ¦ x µ¸
u ( x, t )   erf § ¶¹ (7.132)
Q ª« 2 2 ¨ 2kt ·º

Finally, the solution of the heat conduction problem is

¦ 2 x / 2 kt u 2 µ ¦ x µ
u ( x, t )  u0 §1  e du ¶  u0 erfc § ¶ (7.133)
¨ Q ±0 · ¨ 2kt ·
424 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

EXAMPLE 7.15 Determine the temperature distribution in the semi-infinite medium x s 0,

when the end x  0 is maintained at zero temperature and the initial temperature distribution
is f ( x ).

Solution The given problem is described by

wu w 2u
PDE: K 0  x  f, t!0 (7.134)
wt w x2

BC: u (0, t ) = 0, t>0 (7.135)

IC: u ( x, 0)  f ( x), 0 xe (7.136)

and, u , w u /w x, both tend to zero as x o f. Taking the Fourier sine transform of Eq. (7.134)
and denoting
*s [u( x , t ); x  D ] by Us
we have
2 fwu 2 f w 2u

π ³0 wt
sin α x dx
π
K ³0 w x2 sin α x dx
which becomes
dUs
 K [Bu (0, t )  B 2Us ]
dt
Using the BC (7.135), we obtain
dU s
 KB 2Us  0 (7.137)
dt
Also, taking the Fourier sine transform of IC (7.136), we get
U s  Fs (B ) at t  0 (7.138)
Now, Eq. (7.137) can be rewritten as
d 2
(Us e Kα t ) = 0 (7.139)
dt
Integrating, we get

U s e KB
2
t
 const.
Using Eq. (7.138), we note that Fs (B )  constant . Therefore,

Us e KB
2
t
 Fs (B )
Us  Fs (B )e KB
2
t
(7.140)
 FOURIER TRANSFORM METHODS 425

Finally, taking the inverse Fourier sine transform of Eq. (7.140), we obtain
2 f 2
u ( x, t ) ³0 Fs (α )e Kα t sin α x dα
π

7.12 SOLUTION OF WAVE EQUATION

Wave motions that occur in nature, viz., sound waves, surface waves, transverse vibrations of
an infinite string, and of mechanical systems are governed by the wave equation. As our first
example, we shall consider the transverse displacements of an infinite string.

EXAMPLE 7.16 Compute the displacement u ( x, t ) of an infinite string using the method
of Fourier transform given that the string is initially at rest and that the initial displacement
is f ( x),  f  x  f.

Solution Displacement of an infinite string is governed by the PDE

w 2u w 2u
c2 , f  x  f (7.141)
w t2 w x2
and ICs
u ( x, 0) f ( x), f  x  f (7.142)

ut ( x, 0)  0 (7.143)
In view of two ICs, the given problem is a properly posed problem. Taking the Fourier
transform of PDE, we have
1 f w 2u c2 f w 2u
³ f w t 2 eiα x dx ³ f dx2 e
iα x
dx
2π 2π

w2 1 f
iα x
wt 2 2π ³f ue dx c 2α 2U (α , t )

i.e.,
d 2U
 c 2B 2U  0 (7.144)
dt 2
Its general solution is found to be
U (B , t )  A cos (cB t )  B sin (cB t ) (7.145)
The Fourier transform of the ICs gives
dU
(B , t )  0, U  F (B ) (7.146)
dt
426 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

i.e.,  AcB sin (cB t )  BcB cos (cB t )t  0  0, implying BcB  0 or B  0. Also, Eqs. (7.145) and
(7.146) yield

A  F (B ) (7.147)
Thus,
U (B , t )  F (B ) cos (cB t ) (7.148)
Taking its inverse Fourier transform, we obtain
1 ‡
Ô ‡ F (D ) cos (cD t ) e
 iD x
u ( x, t ) dD (7.149)
2S
If we simplify Eq. (7.149) further, an interesting result emerges, i.e.,

icD t
1 ‡ Ë ‡ ÛÈe  e icD t Ø iD x
u ( x, t )
2S Ô ‡ Ô ‡ Ì
Í
f (u ) eiD u du Ü É
ÝÊ 2
ÙÚ e dD

1Ë 1 ‡ Î ‡ ÞÛ
Ô ‡ Ô ‡ f (u) e du ßÜ (e icst  eicst ) e isx ds
isu
Ï
2 ÌÍ 2S Ð àÝ
1Ë 1 ‡ Î 1 ‡ Þ
Ô ‡ f (u ) e Ô ‡ e
 isu is ( x  ct )
Ï ds ß du
2 ÌÍ 2S Ð 2S à
1 ‡ È 1 ‡ Ø Û
Ô ‡ f (u ) e Ô ‡ e
 isu is ( x  ct )
 ÉÊ ds Ù du Ü
2S 2S Ú Ý

Using the Fourier integral formula, we arrive at the result

1
u ( x, t )  [ f ( x  ct )  f ( x  ct )]
2
which is the well-known D’Alembert’s solution of the wave equation.

EXAMPLE 7.17 Obtain the solution of free vibrations of a semi-infinite string governed by

PDE: utt c 2u xx , 0  x  f, t!0 (7.150)

ICs: u ( x, 0)  f ( x) (7.151)

ut ( x, 0)  g ( x ) (7.152)

Solution Taking the Fourier sine transform of PDE, we have

f w 2u f w 2u
³0 w t2
sin α x dx c2 ³0 w x2 sin α x dx (7.153)
 FOURIER TRANSFORM METHODS 427

Now consider
f w 2u
³0 w x2
sin α x dx

Integrating by parts, we get

f
§w u · f
¨©
wx
sin α x  α u cos α x ¸  α 2
¹0 ³0 u sin α x dx
Now, since the string is fixed at x  0 for all t and we assume that u and w u /w x both tend
to zero as x o f, we arrive at

f w 2u f
³0 wx 2
sin α x dx α 2 ³0 u sin α x dx α 2U

Hence, Eq. (7.153) reduces to

d 2U
2
 c 2B 2U  0 (7.154)
dt
Its general solution is known to be
U  A (B ) cos (cB t )  B (B ) sin (cB t ) (7.155)

where A (B ) and B (B ) have to be determined. Now, at t  0, we have

‡
U ³0 f ( x)sin α x dx F (α )

wU ‡

wt ³0 g ( x) sin α x dx G (α )

In the solution (7.155), if we take t  0, then we have

U A(α ) F (α )

wU
cα B (α ) G (α )
wt
Substituting A (B ) and B (B ) into Eq. (7.155), we get

G (B )
U  F (B ) cos (cB t )  sin (cB t )
cB
428 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Taking the inverse Fourier sine transform of this relation, we obtain

2 f
u ( x, t )
π ³ 0 U (α , t ) sin α x dα
2 f ª G (α ) º
π ³0 «¬ F (α ) cos (cα t ) sin α x  cα sin (cα t ) sin α x »¼ dα

1 f

2π ³0 F (α ) [sin α ( x  ct )  sin α ( x  ct )] dα

1 f G (α )

2π ³0 cα
[cos α ( x  ct )  cos α ( x  ct )] dα

Since

2 f
f ( x  ct )
π ³0 F (α ) sin α ( x  ct ) dα

2 f
g (u )
π ³ 0 G(α ) sin u α dα
x  ct 2 f x  ct
³ xct g (u) du π ³ 0 G (α ) dα ³ xct sin α u du
x  ct
2 f § cos α u ·
π ³0 G (α ) dα ¨ 
© ¸
α ¹ x  ct

2 fG (α )
π ³0 α
dα [cos α ( x  ct )  cos α ( x  ct )]

we arrive at the solution

1 1 x ct
u ( x, t )  [ f ( x  ct )  f ( x  ct )]  ± g (u ) du (7.156)
2 2c x ct

7.13 SOLUTION OF LAPLACE EQUATION

One of the most important PDEs that occurs in many applications is the Laplace equation.
Steady-state heat conduction, the electric potential in the steady flow of currents in solid
conductors, the velocity potential of inviscid, irrotational fluids, the gravitational potential at
an exterior point due to ellipsoidal Earth and so on, are all governed by Laplace equation. We
shall now consider a few related examples.
 FOURIER TRANSFORM METHODS 429

EXAMPLE 7.18 Solve the following boundary value problem in the half-plane y > 0,
described by

PDE: u xx  u yy 0,  f  x  f, y!0
BCs: u ( x, 0) f ( x),  f  x  f,

u is bounded as y o f; u and w u /w x both vanish as | x | o f.

Solution Since x has an infinite range of values, we take the Fourier exponential
transform of PDE in the variable x to get
* [uxx ; x  D ]  * [uyy ; x  D ] 0

Since u and w u /w x both vanish as | x | o f, we have

1 f
iα x
α 2U (α , y ) 
2π ³ f u yy e dx 0

w2 ª 1 f
iα x º
α 2U (α , y )  «
w y 2 ¬ 2π ³ f u ( x, y) e dx »
¼
0

i.e.,
d 2U (B , y )
2
 B 2U (B , y )  0 (7.157)
dy
Its general solution is known to be
U (B , y )  A (B ) eB y  B (B ) eB y (7.158)

Since u must be bounded as y o f, U (α , y ) and its Fourier transform also should be bounded
as y o f, implying A (B )  0 for B  0; but if B  0, B (B )  0; thus for any B,
U (B , y )  constant (e|B | y ) (7.159)
Now the Fourier transform of the BC yields
U (D , 0) * [ f ( x ); x  D ] F (D ) (7.160)
From Eqs. (7.159) and (7.160), we find that
F (B )  const . (7.161)
Hence,
1 f
U (α , y ) F (α ) e|α | y ³ f f ( x) e
|α | y iα x
e dx
2π
430 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Taking its Fourier inverse transform, we obtain, after replacing the dummy variable x by Y ,
the equation
1 ‡
Ë 1 ‡
Û
2S Ô Ô
u ( x, y ) Ì 2S f ([ ) e |D | y eiD[ d [ Ü e  iD x d D
‡ Í ‡ Ý
‡ ‡
1
2S Ô ‡
f ([ ) d [ Ô ‡
exp [{D [i ([  x)] | D | y}] dD (7.162)

But
1 f

2π ³ f exp {α[i (ξ  x)]  | α | y} dα

‡
1 0
1
2S Ô ‡
exp {D [ y  i ([  x)]} dD 
2S Ô 0
exp {D [ y  i ([  x)]} dD
0 ‡
1 Ë exp{D [ y  i ([  x)]} Û 1 Ë exp{D [ y  i ([  x)]} Û
Ì Ü  Ì Ü
2S Í y  i ([  x) Ý ‡ 2S Í y  i ([  x) Ý0
1 Ë 1 1 Û
2S Ì y  i ([  x)  y  i ([  x) Ü
Í Ý
1 y
S ([  x) 2  y 2 (7.163)

Substituting Eq. (7.163) into Eq. (7.162), the required solution is found to be

y f f (ξ ) d ξ
u ( x, y )
π ³ f (ξ  x)2  y 2 (7.164)

This solution is a well-known Poisson integral formula and is valid for y  0, when f ( x) is
bounded and piecewise continuous for all real x.

EXAMPLE 7.19 Solve the following Neumann problem described by

PDE: u xx ( x, y )  u yy ( x, y ) 0,  f  x  f, y!0
BC: u y ( x, 0) f ( x), f  x  f

u is bounded as y o f

u and w u /w x both vanish as | x | o f

 FOURIER TRANSFORM METHODS 431

Solution Let us define a function G ( x, y )  u y ( x, y ). Then

w
φ xx  φ yy (u xx  u yy ) 0 (7.165)
wy

BC: G ( x, 0)  u y ( x, 0)  f ( x) (7.166)

Thus, the function G ( x, y ) is a solution of the problem described in Example 7.18 and,
therefore, the solution of Eq. (7.165) subject to Eq. (7.166) is of the form
y f f (ξ ) d ξ
φ ( x, y )
π ³ f (ξ  x)2  y 2 , y!0 (7.167)

However,
1 f y dy
u ( x, y ) ³ φ ( x, y) dy π ³ f f (ξ )³ ( x  ξ )2  y 2
dξ

Therefore,
1 f
³ f f (ξ ) log [( x  ξ )  y 2 ] dξ  const.
2
u ( x, y ) (7.168)
2π
is the required solution.

7.14 MISCELLANEOUS EXAMPLES

EXAMPLE 7.20 Find the Fourier transform of the normal density function

f ( x)  exp [( x  m)2/2T 2 ] / 2QT 2

Solution From the definition of Fourier transform we have

1 ‡ exp [ ( x  m)2 /2V 2 ]

* (D ) Ô eiD x dx
2S ‡ 2SV 2

Let x  m  z; then
T
2
‡ e z /2
* (D ) Ô ‡ eiD ( m  V z ) dz
2S
432 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Expanding eiBT z in the form

‡ 2
e  z /2 (iDV z )n
‡
iD m
e ÇÔ ‡ 2S n!
dz (7.169)
n 0

and denoting
2
f z n e z /2
In ³ f 2π
dz , n 0, 1, 2, } (7.170)

we obtain
f
(iασ )n
2π F (α ) eiα m ¦ n!
In (7.171)
n 0

For n  0, we have

2
È ‡ e− z / 2 Ø
2
‡ exp (− z12 /2) exp (− z22 /2)
‡
Ô Ô −‡ Ô
2
(I0 ) = É dz Ù = dz1dz2
Ê −‡ 2S Ú −‡ 2S
2
2S ‡ e − r /2
= Ô 0 Ô0 2S
r dr dT [in polar coordinates]

=1

To compute I n , n  0, we can integrate by parts to get

‡ 2 ‡
n2  z 2/2
Ô ‡ z Ô ‡ z
n1
2S I n  d (ez /2
)  (n 1) e dz  (n 1) 2S I n2

Thus, we get a recurrence formula

I n  (n  1) I n  2
2
Finally, we note that I n  0, when n is odd, since z n ⋅ e− z /2
is an odd function. In that case
we have

I 2 n = (2n − 1) I 2n − 2

= (2n − 1) (2n − 3) I 2 n − 4


= (2n − 1) (2n − 3) ... 3 × 1

 FOURIER TRANSFORM METHODS 433

However,

(2n) (2n  1) (2n  2) ... 3 t 2  1

(2n  1) (2n  3) ... 3 t 1 
(2n) (2n  2) ... 4 t 2
(2n)!

2n n !
Finally, Eq. (7.171) becomes

‡ ‡ n
(iασ )2n (2n)! 1 È α 2σ 2 Ø
2π F (α ) eiα m Ç (2n)! 2n n !
eiα m Ç É
n! Ê 2 Ú
Ù eiα m exp ( α 2σ 2/ 2)
n 0 n 0

Thus,

exp (iD m  D 2V 2/2 )

F (D ) * [ f ( x ); D ]
2S

EXAMPLE 7.21 Using the Fourier cosine transform, find the temperature u ( x, t ) in a semi-
infinite rod 0 d x d f, determined by the PDE
ut ku xx , 0  x  f, t!0
subject to

IC: u ( x, 0) 0, 0 d x d f,

BC: u x (0, t ) u0 (a constant) when x 0 and t ! 0

u, w u /w x both tend to zero as x o f.

Solution Since w u /w x is given at the lower limit, we take the Fourier cosine transform
of the given PDE, to obtain

2 ∞wu 2 ∞ w 2u

π Ô0 wt
cos α x dx = k
π Ô 0 w x2 cos α x dx
∞
2 Ëw u Û 2 ∞ wu
=k
π ÌÍ w x cos α x ÜÝ + kα π
0
Ô0 wx
sin α x dx

From physical considerations, we expect w u /w x o 0 as x o f. Therefore,

d 2 §w u · 2 2 f

dt
(U c ) k
π
¨© ¸¹
wx x 0
 kα
π
(u sin α x)0f  kα 2
π ³0 u cos α x dx
434 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

If u o 0 as x o f, we have

d 2
(U c )  ku0  kB 2U c
dt Q

d 2
(U c )  kB 2U c  ku
dt Q 0
which is a linear ODE, whose general solution is found to be of the form
d kB 2t
ku0 ekB t
2 2
(e U c ) 
dt Q
On integration, we obtain

ekB tUc  ku0 ± e kB t dt  const.

2 2 2
(7.172)
Q
Taking the cosine transform of the IC, we get
Uc  0 when t  0
Using this condition in Eq. (7.172), we have

2 u0
0  const .
Q B2
which yields
u0 2
const .  
B 2 Q
Hence,

(1  e kB t )
2 u0 2
Uc  (7.173)
QB 2

Finally, taking the Fourier inverse cosine transform of Eq. (7.173), we have
2u0 f cos α x 2
u ( x, t ) ³0 (1  e kα t ) dα
π α 2

EXAMPLE 7.22 Using the method of integral transform, solve the following potential
problem in the semi-infinite strip described by
PDE: u xx  u yy 0, 0  x  f, 0 ya
 FOURIER TRANSFORM METHODS 435

subject to
BCs: u ( x, 0) f ( x)
u ( x, a ) 0
u ( x, y ) 0, 0  y  a, 0 xf
and
w u /w x tends to zero as x o f.

Solution Taking the Fourier sine transform of the above PDE, we obtain

2 f w 2u 2 f w 2u

π ³ 0 w x2 sin α x dx 
π ³ 0 w y 2 sin α x dx 0

Integrating by parts, we get

f
2 §w u · 2 fwu d 2U s
π
¨©
wx
sin α x ¸  α
¹ 0 π ³0 w x cos α x dx 
dy 2
0

Using the fact that w u /w x o 0 as x o f and integrating again by parts, we obtain

2 2 f d 2U s
α
π
(u cos α x)0f  α 2
π ³0 u sin α x dx 
dy 2
0

Using the BC
u ( x, y ) 0, 0  y  a, 0 xf
we have

d 2Us
2
 B 2Us  0
dy
Its general solution is found to be

Us  A (B ) cosh B y  B (B ) sinh B y (7.174)

Now, the Fourier sine transform of the first two BCs gives

2 f
U s (α , 0)
π ³0 f (ξ ) sin αξ dξ , Us (α , a ) 0

Using these two relations in Eq. (7.174), we obtain

2 f
A (α )
π ³0 f (ξ ) sin αξ dξ (7.175)

0 A (α ) cosh α a  B (α ) sinh α a
436 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Therefore,

cosh α a 2 f
B (α ) 
sinh α x π ³0 f (ξ ) sin αξ dξ (7.176)

Thus, from Eqs. (7.174) to (7.176), we obtain

2 ª f cosh α a sinh α y f º
Us
π «cosh α y
¬
³0 f (ξ ) sin αξ dξ 
sinh α a ³0 f (ξ ) sin αξ dξ »
¼

2ª f sinh (a  y )α º (7.177)
𠫬 ³0 f (ξ ) sin αξ
sin aα
dξ »
¼
Finally, taking the inverse Fourier sine transform of Eq. (7.177), we have
2 f f sinh (a  y )α
u ( x, y )
π ³0 f (ξ ) dξ ³0 sin aα
sin aξ sin α x dα

EXAMPLE 7.23 Using finite sine transform to solve the following BVP described by
1
PDE: u xx  ut , 0  x  L, t  0
k
BC: u (0, t )  u ( L, t )  0 for all t
¬ x
¯¯2u0 L , 0 c x c L /2
IC: u ( x, 0)  ­
¯2u0 ¦§1  x µ¶ , L
cxcL
®¯ ¨ L · 2

Solution Taking the finite Fourier sine transform of the given PDE, we have
L w 2u nπ x 1 Lwu nπ x
³0 w x2 sin L
dx
k ³0 w t sin L
dx (7.178)

Integrating the left-hand side by parts, we get

L
§wu nπ x · nπ Lwu nπ x
¨©
wx
sin
L ¹0
¸ 
L ³0 w x cos L
dx

Again integrating by parts, we obtain

nπ ⎡⎛ nπ x ⎞
L
nπ L nπ x ⎤ n 2π 2 nπ
−
L
⎢⎜ u cos
⎣⎢ ⎝ L
⎟ +
⎠ 0 L ∫0 u sin
L
dx ⎥ = 2 Us (n) +
⎥⎦ L L
{u (0, t ) − u ( L, t ) cos nπ }

n 2π 2
=− Us (n) [after using the BCs]
L2
 FOURIER TRANSFORM METHODS 437

Dropping the subscript s, Eq. (7.178) becomes

dU kn 2Q 2
 2 U 0 (7.179)
dt L
Its general solution is known to be

¦ kn 2Q 2 µ
U  Us (n, t )  A exp §  t¶ (7.180)
2
¨ L ·

where A is a constant to be determined from the ICs. Taking the finite Fourier sine transform
of the given IC, we have
L /2 x nπ x L ⎛ x⎞ nπ x
U s (n, 0) = ∫0 2u0
L
sin
L
dx + ∫ L/2 2u0 ⎜⎝1 − L ⎟⎠ sin L
dx

L /2
⎡ x nπ x nπ ⎤ 2u0 1 L /2 nπ x
= ⎢ −2u0 cos
⎣ L L L ⎥⎦ 0
+
L nπ L ∫0 cos
L
dx

L
⎡ ⎛ x⎞ nπ x nπ ⎤ 2u 1 L nπ x
+ ⎢ −2u0 ⎜1 − ⎟ cos
⎣ ⎝ L⎠ L ⎥
L ⎦ L /2
− 0
L nπ / L ∫ L /2 cos L
dx

or
4u0 L nQ
U s (n, 0)  sin
nQ2 2 2
From Eq. (7.180) when t  0, we get

A  Us (n, 0) and, therefore,

¬0 if n is even
¯
A  ­ 4u0 L (2r  1)Q
¯ (2r  1)2 Q 2 sin 2
if n is odd
®
i.e.
(1)r 4u0 L
A
(2r  1)2 Q 2
Thus, from Eq. (7.180), we have

(1)r 4u0 L © k (2r  1)2 Q 2t ¸

U s (n, t )  exp ª  ¹ (7.181)
(2r  1)2 Q 2
« L2 º
438 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

Applying the inversion theorem to Eq. (7.181), we obtain

e e
¥ ¥
2
2 nS x 8u0 ( 1) r © k (2r  1) S2 ¸ (2r  1)S x
U ( x, t )  U s (n, t ) sin  exp ª  t ¹ sin
L
n 1
L S 2 r  0 (2r  1)2 « L2
º L

EXAMPLE 7.24 Find the steady-state temperature distribution u ( x, y ) in a long square bar
of side Q with one face maintained at constant temperature u0 and the other faces at zero
temperature.
Solution Mathematically, the above problem can be posed as the following BVP:

PDE: u xx + u yy = 0, 0 < x <π, 0 < y <π

BCs: u (0, y ) = u (π , y ) = 0
u ( x, 0), u ( x, π ) = u0

Taking the finite Fourier sine transform with respect to the variable x, we have

S w 2u S w 2u
Ô0 w x 2
sin nx dx  Ô0 w y 2 sin nx dx 0

S
Èw u Ø S wu w2 S
ÉÊ
wx
sin nx Ù  n
Ú0 Ô0 w x cos nx dx 
w y2 Ô0 u sin nx dx 0

S d 2U s
 n (u cos nx)S0  n 2 Ô0 u sin nx dx 
dy 2
0

Therefore,

d 2U s
2
 n 2Us 0
dy
Its general solution is known to be
Us  A cosh ny  B sinh ny (7.182)
Taking the finite Fourier sine transform of the second set of BCs, we have

Us (n, 0)  0
Q
Q ¦ cos nx µ ¦ 1  cos nQ µ
Us (n, Q )  ± u0 sin nx dx  u0 §  ¶  u0 § ¶
0 ¨ n ·0 ¨ n ·
 FOURIER TRANSFORM METHODS 439

Using these results in Eq. (7.182), we obtain A  0 and

¦ 1  cos nQ µ
B sinh nQ  u0 § ¶
¨ n ·

Hence,
u0 ¦ 1  cos nQ µ
Us  § ¶ sinh ny
sinh nQ ¨ n ·
Finally, taking the finite Fourier sine inverse transform, we have the result

f
2 u § 1  cos nπ ·
u ( x, y )
π ¦ sinh0n𠨩 n
¸¹ sinh ny sin nx
n 1

Thus, the required temperature distribution is

f
4u0 sinh (2r  1) y sin (2r  1) x
u ( x, y )
π ¦ (2r  1) sinh (2r  1) π
when n odd
r 0

0 when n even

EXAMPLE 7.25 A one-dimensional infinite solid,  f  x  f, is initially at temperature

F(x). For times t  0, heat is generated within the solid at a rate of g ( x, t ) units. Determine
the temperature in the solid for t  0.
Solution The problem can be described as follows:

w 2T [ g ( x, t )] 1 wT
PDE:  ,  f  x  f, t ! 0
wx 2 k k wt
IC: T ( x, 0) F ( x), f  x  f

The exponential Fourier transform of PDE with respect to x is given by

f w 2T 1 f 1 f wT
³f eiα x dx  ³f e
iα x
g ( x, t ) dx ³f e
iα x
dx
wx 2 k k wt
which on rewriting becomes
f w 2T 1 1 dT (α , t )
³f eiα x 2
dx  g (α , t )
wx k k dt
440 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

or
1 1 dT (B , t )
B 2T (B , t )  g (B , t )  (7.183)
k k dt
The Fourier transform of the IC is given by
f
iα x
T (α , 0) ³f e F ( x) dx F (α , 0)

Therefore,
f
iα x c
T (α , t ) t 0
T (α , 0) ³f e F ( x c ) dx c F (α , 0)

Using this result, Eq. (7.183) can be modified to

dT (α , t )
+ kα 2T (α , t ) = g (α , t )
dt
The solution of this equation is found to be

T exp ⎛⎜ kα 2 dt ⎞⎟ =
2
∫ ∫ g (α , t ) ekα t dt + c
⎝ ⎠
Using the IC, we get
c  F (B , 0)
Hence, using the IC, the solution of Eq. (7.183) is obtained as

T (B , t )  e kB t ª F (B , 0)  ± g (B , t b) ekB t dt b¹
2 ©

t 2 b ¸
« t b0 º

Now, taking its inverse exponential Fourier transform, we get

1 f ª t º
kα 2t c
T ( x, t )
2π ³α f
exp ( kα 2 t  iα x) « F (α ) 
¬ ³ tc 0 e g (α , t c ) dt c » dα (7.184)
¼
where
f
iα x c
F (α ) ³xc f e F ( x c ) dx c

f
iα x c
g (α , t c ) ³xc f e g ( x c, t c ) dx c

After changing the order of integration and rewriting, Eq. (7.184) becomes

1 f f
T ( x, t )
2π ³ xc f F ( x c ) dx c ³α f
exp [ kα 2t  iα ( x  x c )] dα
 FOURIER TRANSFORM METHODS 441

1 t f f

2π ³ t c 0 dt c ³ xc f g ( xc, t c) dxc ³ α f
exp [ kα 2 (t  t c )  iα ( x  x c )] dα (7.185)

Now,

f f ª§ x  x ·
2
§ x  xc · º
2

³α f
exp [ kα 2 (t )  iα ( x  x c)] dα ³α f
exp « ¨ 
¬ © 2 kt
1
 iα kt ¸  ¨
¹ © 2 kt ¸¹ ¼
» dα

Let
x  xb
  iB kt  iI
2 kt

Then the above equation can be written as

f 1 f
³α exp [ kα 2t  iα ( x  x c)] dα exp [( x  x c)2 / (4kt )] ³f exp (η )dη (7.186)
2
f kt
But,
f
η 2
³f e dn π (Standard integral) (7.187)

Using Eqs. (7.186) and (7.187), the required temperature is obtained from Eq. (7.185) in the
form

1 ‡
Ô‡ F ( x „) exp [( x  x „) / (4kt )] dx „
2
T ( x, t )
4S kt
t dt „ ‡
Ô t „0 Ô‡ g ( x „, t „) exp [( x  x „)
2
/4D (t  t „)] dx „
4S k (t  t „)

EXAMPLE 7.26 A uniform string of length L is stretched tightly between two fixed points
at x  0 and x  L. If it is displaced a small distance F at a point x  b, 0  b  L, and
released from rest at time t  0, find an expression for the displacement at subsequent times.

Solution Let u ( x, t ) denote the displacement of the string. Then at t  0, the equation
for OA is: y  (F / b) x, (see Fig. 7.1) while the equation for AB is given by
F
y ( x  L)
bL
Now, the IBVP is described by the PDE:

utt  c 2 u xx
442 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

BCs: u (0, t ) = u ( L, t ) = 0, t ≥ 0,
⎧ε x
⎪⎪ b , 0< x<b
ICs: u ( x, 0) = ⎨
⎪ ε ( x − L) , b<x<L
⎪⎩ b − L
ut ( x, 0) = 0
Taking the finite Fourier sine transform of the PDE, we have
L w 2u nπ x L w 2u nπ x
³0 w t 2 sin
L
dx c2 ³0 w x2 sin L
dx

or
d 2Us ­° § w u nπ x ·
L
nπ Lwu nπ x ½°
dt 2
c2 ®¨
©
¯° w x
sin
L 0
¸
¹

L ³0 w x cos
L
dx¾
¿°

nπ 2 ª § nπ x º
L
nπ x · nπ L

L
c « ¨ u cos
¬ © L 0
¸
¹

L ³0 u sin
L
dx »
¼

 n 2π 2 c 2 nπ c 2
Us  {u (0, t )  u ( L, t ) cos nπ }
L2 L

 n 2π 2 c 2
Us (after using BCs)
L2
Therefore,
d 2Us n2Q 2 c 2
 Us  0
dt 2 L2
Its general solution is found to be
nQ ct nQ ct
Us (n, t )  A cos
 B sin (7.188)
L L
Now, taking finite Fourier sine transform of ICs, we obtain
bFx nQ x L F ( x  L) nQ x
Us (n, 0)  ± sin dx  ± sin dx
0 b L b bL L
b
F L ¦ nQ x µ F L b nQ x
 §  x cos ¶  t ± 0 cos dx
b nQ ¨ L ·0 b nQ L
L
FL © nQ x ¸ FL L nQ x
 ª  ( x  L ) cos ¹  ± cos dx
nQ (b  L) « L º 0 nQ (b  L) b L
2 2
FL nQ b FL nQ b
 sin  2 2 sin
n 2Q 2 b L n Q (b  L) L
 FOURIER TRANSFORM METHODS 443

or

ε L3 nπ b
Us (n, 0) = sin
n π b ( L − b)
2 2 L

From Eq. (7.188), when t  0, we have

ε L3 nπ b
Us (n, 0) = A = sin
n π b ( L − b)
2 2 L

Further, taking finite Fourier sine transform of the second IC, we get
dUs
0
dt
From Eq. (7.188) it can be easily seen that B  0. Thus, we obtain from Eq. (7.188) the
relation

ε L3 nπ b nπ ct
Us (n, t ) = sin cos
n π b ( L − b)
2 2 L L

Finally, inverting, we have the required, displacement as

f
2ε L2 1 nπ b nπ x nπ ct
u ( x, t )
π 2 b ( L  b)
¦ n2 sin L
sin
L
cos
L
n 1

EXERCISES
1. Find the Fourier transform of

¬1  x 2 ,
¯ | x | c1
f ( x)  ­
¯
®0, | x | 1
and hence evaluate
f§x cos x  sin x · x
³ 0 ¨© x 3 ¸¹ cos 2 dx

2. Find Fourier sine and cosine transforms of e  x .

3. Find the Fourier cosine transform of
1
f ( x) 
1  x2
444 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

4. Using Parseval’s relation for the Fourier cosine transform of

g ( x)  e ax

¬1, 0 xM
f ( x)  ­
®0, xM

show that

f sin λα dα π § 1  e λ a ·
³ 0 α (a 2  α 2 ) 2 ¨© a 2 ¸¹
5. Verify the following relations:
f f
(i)
³ 0 Fs (α ) Gs (α ) dα ³ 0 f ( x) g ( x) dx

f f
³ 0 [ Fc (α )] dα ³ 0 [ f ( x)]
2 2
(ii) dx

f f
³ 0 [ Fs (α )] dα ³ 0 [ f ( x)] dx.
2 2
(iii)

6. If a  0, b is any real or complex, show that

f
 ax 2  2bx π b2/ a
³ f e dx
a
e

7. Solve the problem described by

PDE: ut u xx  δ ( x) δ (t ), f xf

BC: Lt u ( x, t ) 0
| x|of
IC: u ( x, 0) δ ( x)

8. If v ( x, t ) denotes the solution of

vt  vxx 0,  f  x  f, t ! 0
v ( x, 0) f ( x), f  x  f
show that
t
u ( x, t ) = ∫ 0 v ( x, t − τ ) d τ
 FOURIER TRANSFORM METHODS 445

is a solution of
ut  u xx f ( x),  f  x  f, t ! 0
u ( x, 0) 0, f  x  f
and hence, write down the Green’s function for the above non-homogeneous PDE with
the homogeneous initial condition.
[Duhamel’s principle]
9. The temperature R ( x, t ) in the semi-infinite rod x s 0 is determined from the differential
equation

wθ w 2θ
α2
wt w x2
subject to
IC: R ( x, 0)  0
BC: R (0, t )  G (t ) for t  0
Using the Fourier sine transform, derive the solution in the form

2 f § x2 · 2
θ ( x, t )
π ³ x /(2a φ ¨ t  2 2 ¸ eu du
t) © 4α u ¹

10. If ‘2u  0, for x s 0 and if u  f ( y ) on x  0, show that by using Fourier transform

technique,
x f f (ξ ) dξ
u ( x, y )
π ³f x2  ( y  ξ )2
11. Using the Fourier transform method, show that the solution of the two-dimensional
Laplace equation
w 2u w 2u
 0
w x2 w y2
is valid in the half-plane y  0, subject to the condition

¬0, x0
u ( x, 0)  ­
®1, x0
and
Lt u ( x, y ) 0 in the above half-plane.
x 2  y 2 of
446 INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS

12. Using the finite Fourier transform, solve the BVP described by

w V w 2V
PDE :  , 0  x  6, t  0
w t w x2
subject to
BC: Vx (0, t )  0  Vx (6, t )
IC: V ( x, 0)  2 x
13. Using the finite Fourier transform, solve the two-dimensional Laplace equation

w 2V w 2V
 0 0  x  π , 0  y  y0
w x2 w y2
subject to
V (0, y ) = 0, V (π , y ) = 1
Vy ( x, 0) = 0, V ( x, y0 ) = 1.