CHAPTER 1

FOURIER TRANSFORM

INTEGRAL TRANSFORM : AN INTRODUCTION

What is an integral transform ?

An integral transform of a function f (x) defined on (a, b) has the form

Z b
I(f ) ≡ F (y) = K(x, y)f (x)dx
a

provided the integral exists, where K(x, y) is a known function of x, y. The function
K(x, y) is called the kernel of the integral transform.

Note: 1) The kernel K(x, y) along with the interval (a, b) (which need not be
finite) distinguishes a particular integral transform from another.
2) In general, properties of the transform, such as for which class of functions f
the integral exists, depend on K, but all integral transforms satisfy

I(f + g) = I(f ) + I(g)

and

I(cf ) = cI(f )

where c is a constant (f and g being functions).

Remark:

Integral transforms are very much useful in dealing with differential equations
(linear) with appropriate boundary conditions. Proper choice of the integral trans-
form makes it possible to convert an ordinary differential equation which may be

1
intractable to an algebraic equations that can perhaps be easily solved. The solution
obtained is in fact the transform of the solution of the original differential equation
and it is necessary to find the inverse transform to find the desired solution. Thus we
need to know the integral transform and the corresponding inverse transform. There
exists quite a large number of integral transforms but here only five transforms viz.
Fourier, Laplace, Mellin, Hankel and z-transform will be discussed.

Fourier Transform

Motivation

One can obtain Fourier transform from the study of Fourier series. In Fourier
series, a periodic function is expressed in terms of sum of sines and cosines. Thus a
complicated function which is periodic can be expressed as the sum of simple func-
tions represented by sines and cosines. The Fourier transform can be thought of as
an extension of the Fourier series in which the period of the function is made infinity.

FOURIER INTEGRAL THEOREM

We assume that a function f is periodic with period 2p and f (x), f ′(x) are piece-
wise continuous on [−p, p], then it has the Fourier series representation

∞  
1 X nπx nπx
f (x) = a0 + an cos +bn sin
2 n=1
p p
where
p
1 nπt
Z
an = f (t) cos dt, n = 0, 1, 2, 3...
p −p p
and
p
1 nπt
Z
bn = f (t) sin dt, n = 1, 2, 3... (1.1)
p −p p
Substituting the expression for an and bn into the expression for f (x), we find

2
that the representation of f (x) given by equation (1) is equivalent to
∞
Z p !
1 1 p nπ(t − x)
Z X
f (x) = f (t)dt + f (t) cos dt. (1.2)
2p −p p −p n=1
p

We now make p → ∞ in this representation so that f (x) is defined on (−∞, ∞)

and also assume that f is absolutely integrable on (−∞, ∞). Then for the first term
in (2) we find
p
1
Z
lim f (t)dt = 0.
p→∞ 2p −p
π
For the second term in (2), we write △s = p
so that as p → ∞, △s → 0 and (2)
gives "∞ #
Z π
1 △s X
f (x) = lim f (t) cos(n△s(t − x))△s dt
△s→0 π − π
△s n=1

1 ∞
Z Z ∞ 
= f (t) cos s(t − x)ds dt.
π −∞ 0

Interchanging the order of integration we find

1 ∞ ∞
Z Z
f (x) = f (t) cos s(t − x)dtds. (3)
π 0 −∞

Thus we summarise this result in the following theorem.

Theorem 1: Fourier Integral Theorem

If a function f (x) is defined on (−∞, ∞) and is absolutely integrable on (−∞, ∞)
then
∞ ∞
1
Z Z
f (x) = f (t)cos s(t − x)dtds (3)
π 0 −∞

This is known as Fourier integral theorem.

If x is a point of discontinuity, then f (x) in (3) or (4) is to be replaced by

f (x+0)+f (x−0)
2
.

The integral expansion (3) can be written as

Z ∞Z ∞
1
f (t) ei(t−x)s + e−i(t−x)s dtds
 
f (x) =
2π 0 −∞

3
 ∞ ∞ ∞ ∞
1 1
Z Z Z Z
f (x) = f (t)ei(t−x)s
dtds + f (t)e−i(t−x)s dtds
2π 0 −∞ 2π 0 −∞

(Changing the order of integration we have)

Z ∞ Z ∞  Z ∞ Z ∞ 
1 i(t−x)s 1 −i(t−x)s
f (x) = f (t) e ds dt + f (t) e ds dt
2π −∞ 0 2π −∞ 0

In the inner integral of second term, s is changed to -s, so that

Z ∞ Z ∞  Z ∞ Z 0 
1 i(t−x)s 1 i(t−x)s
f (x) = f (t) e ds dt + f (t) e ds dt
2π −∞ 0 2π −∞ −∞
Z ∞ Z ∞ Z 0 
1 i(t−x)s i(t−x)s
f (x) = f (t) e ds + e ds dt
2π −∞ 0 −∞

Z ∞Z ∞
1
f (x) = f (t)ei(t−x)s dtds
2π −∞ −∞
Z ∞ Z ∞ 
1 −isx its
f (x) = e f (t)e dt ds. (4)
2π −∞ −∞

This is in fact the exponential form of Fourier integral theorem.

FOURIER TRANSFORM:
In (4), if we write
∞
1
Z
F (s) = √ f (t)eits dt (5)
2π −∞

then
∞
1
Z
f (x) = √ F (s)e−isx dx. (6)
2π −∞

F (s) is called the Fourier transform of f (x) defined on (−∞, ∞) and f (x) is called
the inverse Fourier transform of F (s).

Note: The above is a purely formal procedure to obtain Fourier transform and
the inverse Fourier transform.

FOURIER COSINE AND SINE TRANSFORMS:

If f (x) is an even function of x, i.e. f (−x) = f (x), x > 0, then (3) gives

∞ Z ∞ ∞ 
1
Z Z
f (x) = f (t) cos s(t − x)dt + f (t) cos s(t + x)dt ds
π 0 0 0

4
 ∞ Z ∞ 
2
Z
= f (t) cos stdt cos sxds
π 0 0
r Z ∞
2
= Fc (s) cos sxds, x > 0 (11)
π 0
where r Z ∞
2
Fc (s) = f (x) cos sxdx, s > 0. (12a)
π 0
This is the Fourier cosine transform, the inverse cosine transform being
r Z ∞
2
f (x) = Fc (s) cos sxds, x > 0 (12b)
π 0

Similarly, if f (x) is odd function of x, i.e. f (−x) = −f (x), x > 0, then (3) gives

∞ Z ∞ 
1
Z
f (x) = f (t)cos s(t − x) − cos s(t + x)ds dt
π 0 0
∞ Z ∞ 
2
Z
= f (t) sin stdt sin sxds
π 0 0
r Z ∞
2
= Fc (s) sin sxds, x > 0 (13)
π 0
where r Z ∞
2
Fs (s) = f (x) sin sxdx, s > 0. (14a)
π 0
This is the Fourier sine transform, the inverse sine transform being
r Z ∞
2
f (x) = Fs (s) sin sxds, x > 0. (14b)
π 0

FOURIER TRANSFORMS OF SOME SIMPLE FUNCTIONS

We have learnt that the Fourier transform of a function f (x), denoted by F [f (x)]
or F (s), is given by
∞
1
Z
F [f (x)] = √ f (x) eisx dx
2π −∞

with its inverse transform as

∞
1
Z
f (x) = √ F [f (x)] e−isx ds.
2π −∞

Fourier transforms of some simple functions:

5
 We shall consider the following examples to obtain the Fourier transform of some
simple functions.

Example 1. Find the Fourier transform of f (x) where


 1, −1 < x < 1,
f (x) =
 0, otherwise

Solution:

1
r
1 1 eis − e−is 2 sin s
Z
isx
F [f (x)] = √ 1.e dx = =
2π −1 2π is π s
Remark: If we write the inverse transform, then

Z ∞  1, −1 < x < 1,
1 sin s −isx
e ds =
π −∞ s  0, otherwise.

For x = 0, one obtains the result

∞
sin s
Z
ds = π
−∞ s

so that
∞
sin s π
Z
ds = .
0 s 2
Example 2. Find the Fourier transform of f (x) where

f (x) = e−a|x| , a > 0

.
Solution:

∞
1
Z
−a|x|
F [e ]= √ e−a|x| eisx dx
2π −∞
Z ∞ ∞ 
1
Z
−ax+isx −ax−isx
=√ e dx + e dx
2π 0 0
 
1 1 1
=√ +
2π a − is a + is

6
 r
2 a
= , a > 0.
π s2 + a2
Example 3. Find the Fourier transform of f (x) where

x2
f (x) = e− 2 .

Solution:

Z ∞
2
− x2 1 x2
F [e ] = F (s) = √ e− 2 eisx dx
2π −∞
∞
1
Z
s2 1 2
= √ e− 2 e− 2 (x−is) dx
2π −∞
Z ∞
1 2
2 1
= √ e− 2 e−u du,
s
where u = √ (x − is)
π −∞ 2
R ∞ −u2 R ∞ −u2
To derive I = −∞ e du, we write I = 2 0 e du. Substituting u2 = v we
have Z ∞ Z ∞ √
−v − 21 1
I= e v dv = e−v v 2 −1 dv = Γ(1/2) = π.
0 0

so that
√
I= π.

Hence
x2 s2
F [e− 2 ] = e− 2
x2
Thus the function f (x) = e− 2 has the Fourier transform f (s) and is self-reciprocal
in the sense that it is its own Fourier transform.
Example 4. Find the Fourier transform of f (x) where

 0, x < 0,
f (x) =
 e−ax , x > 0, a > 0.

Solution:

Here Z ∞
1
F [f (x)] = √ e−ax eisx dx
2π 0
1 1
=√ .
2π a − is

7
 Example 5. Find the Fourier transform of f (x) where

 sin x, 0 < x < π,
f (x) =
 0, otherwise.

Solution:

Here
√ Z π
2πF [f (x)] = sin xeisx dx
0
π
1
Z
eix − e−ix eisx dx


=
2i 0
π
1 ei(s+1)x ei(s−1)x

= −
2i i(s + 1) i(s − 1) 0
1 ei(s+1)π − 1 ei(s−1)π − 1
 
=− −
2 s+1 s−1
 isπ
e + 1 eisπ + 1

1
=− − +
2 s+1 s−1
1 eisπ + 1
F [f (x)] = √
2π 1 − s2
Remark: This integral can also be evaluated in a different manner.
Z π
I= sin xeisx dx
0

π
Z π
− cos xeisx 0 cos xeisx dx

= + is
0
Z π
isπ
=e +1+ is[sin xeisx ]π0 − is sin xeisx dx
0
isπ 2
=e +1+s I

so that
eisπ + 1
I=
1 − s2
Example 6. Show that the Fourier transform of the function

 1 − |x|, |x| < 1,
f (x) =
 0, otherwise

8
is
2 sin2
r s
2
F (s) = 2
π s2
and hence show that 2
∞ 
sin x
Z
dx = π
−∞ x
Solution:

Here
√ Z 1
2πF (s) = (1 − |x|)eisx dx
−1
Z 1
=2 (1 − x) cos sxdx
0
 1 1
(1 − x) sin sx 2
Z
=2 + sin sxdx
s 0 s 0

2
= [− cos sx]10
s2
2(1 − cos sx)
=
s2
sin2 2s
=4 2
s
q 2 s
2 sin
Thus F (s) = 2 π s2
2
.
For the second part, we note that
∞
1
Z
−1
f (x) = F [F (s)] = √ F (s)eisx ds.
2π −∞

Putting x = 0, we obtain
∞
1
Z
f (0) = √ F (s)ds
2π −∞

i.e.
sin2 2s
r Z ∞
1 2
1= √ 2 ds
2π π −∞ s2
2 ∞ sin u
Z
= 2du
π −∞ 4u2
2
1 ∞ sin u
Z 
= du.
π −∞ u

9
 Thus we deduce 2
∞ 
sin x
Z
dx = π.
−∞ x

PROPERTIES OF FOURIER TRANSFORM

In this module we shall study the properties of Fourier transform. Before pro-
ceeding further we first make the following remark.
Remark: Let f (x) satisfies the conditions of Fourier integral theorem, then its
Fourier transform F (s) is unique. However, if f (x) and g(x) are two functions which
are identical everywhere except for certain isolated points, then they have the same
transform, say F (s). This means that the inverse transform of F (s) can be either
f (x) or g(x). The distinction between functions that differ at isolated points is
mostly of academic interest and we ignore this in our discussion. Thus for a given
function F (s), its inverse f (x) is uniquely related to F (s) by the relation.

∞
1
Z
f (x) = √ F (s)e−isx ds.
2π −∞
Linearity Property:

If F (s) and G(s) are the Fourier transforms of f (x) and g(x) respectively, then the
Fourier transform of c1 f (x)+c2 g(x) where c1 and c2 are constants, is c1 F (s)+c2 G(s).

Proof: The proof is obvious. However for completeness the proof is given below.
Let F [f (x)] denote the Fourier transform of f (x) so that F [f (x)] = F (s) and
F [g(x)] = G(s). Then
Z ∞
1
F [c1 f (x) + c2 g(x)] = √ {c1 f (x) + c2 g(x)}eisx dx
2π −∞
Z ∞ Z ∞
1 isx 1
= c1 √ f (x)e dx + c2 √ g(x)eisx
2π −∞ 2π −∞
= c1 F (s) + c2 G(s).
R∞
Note: If the inverse transform denoted by F −1 [F (s)] = √12π −∞ F (s)e−ixs ds,
then F −1 [c1 F (s) + c2 G(s)] = c1 F −1 [F (s)] + c2 F −1 [G(s)] so that the inverse Fourier

10
transform has the linearity property.

Scaling property:
1
F [f (ax)] = |a|
F (s/a), a 6= 0 where a is constant, either positive or negative.

Proof: To prove this, we first take a > 0,then

Z ∞
1
F [f (ax)] = √ f (ax)eisx dx
2π −∞
Z ∞
1 1 s
= √ f (u)ei a u du
a 2π −∞
after putting ax = u. Thus

1 s
F [f (ax)] = F ( )fora > 0.
a a

Similarly, if a < 0, then F [f (ax)] = − a1 F ( as ) for a < 0.

1
Thus the scaling property for Fourier transform is F [f (ax)] = |a|
F ( as ) for a 6= 0
Shifting Property:
(i) F [eiax f (x) = F (s + a)
(ii) F [f (x − a)] = eias F (s)
Proof of (i): Z ∞
iax 1
F [e f (x) = √ eiax f (x)eisx dx
2π −∞
Z ∞
1
=√ f (x)ei(s+x)x dx
2π −∞
= F (s + a).

Proof (ii): Z ∞
1
F [f (x − a)] = √ f (x − a)eisx dx
2π −∞
Z ∞
1
=√ f (u)eis(u+a) du
2π −∞
Z ∞
ias 1
=e √ f (u)eisu du
2π −∞
= eias F (s).

11
 Remark: The above properties are useful to deduce the transform of many other
functions once the transforms of some standard functions are derived.
Fourier transform of derivative:
Let f (x) be continuous everywhere and f ′ (x) (the derivative of f (x)) be piecewise
smooth, both f (x) and f ′ (x) be absolutely integrable on (−∞, ∞) and f (x) → 0 as
|x| → ∞, then F [f ′ (x)] = −isF (s)
Proof: Z ∞
′ 1
F [f (x)] = √ f ′ (x)eisx dx
2π −∞
Z ∞
1 isx ∞ 1
f (x)eisx dx
 
=√ f (x)e −∞ − is √
2π 2π −∞
= −isF (s).

This can be generalized as

F [f n (x)] = (−is)n F (s), n = 1, 2, ...

if f (x), f ′ (x), ....., f n−1 (x) are continuous and absolutely integrable on (−∞, ∞),
f (x), f ′ (x), ....., f n−1 (x) → 0 as |x| → ∞ and f n (x) is piecewise smooth and abso-
lutely integrable on (−∞, ∞).
Remark: For sin and cosine Fourier transforms, the above results are some
what different. Let Fc (s) and Fs (s) respectively denote the Fourier cosine and sine
transforms of f (x) defined on [0, ∞), i.e.

r Z ∞
2
Fc [f (x)] ≡ Fc (s) = f (x) cos sxdx,
π 0
r Z ∞
2
Fs [f (x)] ≡ Fs (s) = f (x) sin sxdx.
π 0
Then r
′ 2
Fc [f (x)] = sFs (s) − f (0)
π
Similarly
Fs [f ′ (x)] = −sFc (s).
′′
For second derivative f (x) of f (x), one can derive
r
′′ 2 2 ′
Fc [f (x)] = −s Fc (s) − f (0),
π

12
 r
′′ 2 2
Fs [f (x)] = −s Fs (s) + sf (0).
π
Derivative of Fourier transforms:
R∞
Let F (s) = √12π −∞ f (x)eisx dx, then formal differentiation of both sides w.r.to s
produces
∞
1
Z
′
F (s) = √ ixf (x)eisx dx
2π −∞

so that F [xf (x)] = −iF ′ (s) provided of course the Fourier transform of xf (x) exists.
Repeated differentiation w.r.to s produces
Z ∞
(n) 1
F (s) = √ (ix)n f (x)eisx dx
2π −∞

so that
F [xn f (x)] = (−i)n F (n) (s)

provided the Fourier transform of xn f (x) exists.

Fourier transform of a Fourier transform Let

F [f (x)] = F (s)

then
F [F (x)] = f (−s).
R∞
Proof: we have F [f (x)] ≡ F (s) = √1 f (x)eisx dx and
2π −∞

∞
1
Z
f (x) = √ F (s)e−isx ds.
2π −∞

Interchanging x and s, and then replacing s by −s we find

Z ∞
1
f (−s) = √ F (x)eisx dx
2π −∞

= F [F (x)]

CONVOLUTION THEOREM AND PARSEVAL RELATION

Convolution of two functions defined on (−∞, ∞)

13
Definition:

The convolution of two functions f (x) and g(x), denoted by (f ∗ g), is defined by
Z ∞
1
(f ∗ g)(x) = √ f (x − t)g(t)dt.
2π −∞

Properties of the Convolution of two functions:

The convolution of two functions f (x) and g(x) has the following properties
a) Commutative: (f ∗ g)(x) = (g ∗ f )(x)
b) Associative : ((f ∗ g) ∗ h)(x) = (f ∗ (g ∗ h))(x)
c) Distributive: (f ∗ (g + h))(x) = (f ∗ g)(x) + (f ∗ h)(x)

Convolution theorem for Fourier transform :

If F [f (x)] = F (s) and F [g(x)] = G(s), then

i) F [(f ∗ g)(x)] = F (s)G(s),

ii) F −1 [F (s)G(s)] = (f ∗ g)(x).

Proof: Z ∞ Z ∞ 
1
F [(f ∗ g)(x)] = f (x − t)g(t)dt eisx dx
2π −∞ −∞
Z ∞ Z ∞ 
1 isx
= g(t) f (x − t)e dx dt
2π −∞ −∞
Z ∞ Z ∞ 
1 is(u+t)
= g(t) f (u)e du dt
2π −∞ −∞
 Z ∞  Z ∞ 
1 ist 1 isu
= √ g(t)e dt √ f (u)e du
2π −∞ 2π −∞
= F (s)G(s)

14
Parseval relation:
Parseval’s relation is given by
Z ∞ Z ∞
2
|f (x)| dx = |F (s)|2ds.
−∞ −∞

Proof: The result ii) in the previous theorem is equivalent to

Z ∞ Z ∞
−isx
F (s)G(s)e ds = f (x − t)g(t)dt.
−∞ −∞

Putting x = 0, this produces

Z ∞ Z ∞
F (s)G(s)ds = f (−t)g(t)dt
−∞ −∞
Z ∞
= f (t)g(−t)dt.
−∞

If we put g(x) = f (−x), then

Z ∞ Z ∞
1 isx 1
G(s) = F [g(x)] = √ g(t)e dx = √ f (−x)eisx dx
2π −∞ 2π −∞
Z ∞ Z ∞
1 1
=√ f (x)e−isx dx = √ f (x)eisx dx = F (s)
2π −∞ 2π −∞
Thus the above gives
Z ∞ Z ∞
F (s)F (s)ds = f (x)f (x)dx
−∞ −∞

which is equivalent to
Z ∞ Z ∞
2
|f (x)| dx = |F (s)|2ds.
−∞ −∞

Convolution theorem for Fourier cosine and sine transforms:

Let r Z ∞
2
Fc [f (x)] ≡ Fc (s) = f (x) cos sxdx
π 0
q R
∞
and Fc [g(x)] ≡ Gc (s) = π2 0 g(x) cos sxdx, then
Z ∞
1 ∞
Z
Fc (s)Gc (s) cos sxds = f (t) [g(x + t) + g(|x − t|)] dt.
0 2 0

15
The left side is r Z ∞ Z ∞ 
2
Gc (s) f (t) cos stdt cos sxds
π 0 0
r Z ∞ Z ∞ 
2
= f (t) Gc (s) cos st cos sxds dt
π 0 0
"r Z #
1 ∞ 2 ∞
Z
= f (t) Gc (s){cos s(x + t) + cos s(x − t)}ds dt
2 0 π 0
1 ∞
Z
= f (t) [g(x + t) + g(|x − t|)] dt.
2 0
Putting x = 0, this produces
Z ∞ Z ∞
Fc (s)Gc (s)ds = f (t)g(t)dt
0 0

Setting g(x) = f (x) we find

r Z ∞
2
Gc (s) = f (x) cos sxdx
π 0
r Z ∞
2
= f (x) cos sxdx = Fc (s)
π 0
so that Z ∞ Z ∞
2
|Fc (s)| ds = |f (x)|2 dx.
0 0
This is the Parseval relation for Fourier cosine transform.

Parseval relation for sine transforms is given by

Z ∞ Z ∞
Fs (s)Gs (s)ds = f (t)g(t)dt.
0 0

This is because
Z ∞ ∞
r Z ∞
2
Z
Fs (s)Gs (s)ds = Gs (s){ f (x) sin sx dx}ds
0 0 π 0
Z ∞ r Z ∞ Z ∞
2
= f (x){ Gs (s) sin sx ds}dx = f (x)g(x) dx
0 π 0 0
Example:
Example 1: Let f (x) = e−ax (a > 0) and g(x) = e−bx (b > 0) Then it can be shown
that r
2 a
Fc (s) = ,
π a2 + s2

16
 r
2 s
Fs (s) = ,
π a2 + s2
r
2 b
Gc (s) = ,
π b2 + s2
r
2 s
Gs (s) = ,
π b2 + s2
Using these result show that

∞
1 π
Z
ds = , a > 0, b > 0.
0 (a2 + s2 )(b2 + s2 ) 2ab(a + b)
Solution: We have the Parseval relation for cosine transforms
Z ∞ Z ∞
Fc (s)Gc (s)ds = f (t)g(t)dt
0 0

.
The right side is
∞ ∞
1
Z Z
f (t)g(t)dt = e−(a+b)t dt =
0 0 a+b

.
The left side is
∞ ∞
2 1
Z Z
Fc (s)Gc (s)ds = ab ds
0 π 0 (a2 + s2 )(b2 + s2 )

so that
∞
1 π
Z
ds = , a > 0, b > 0.
0 (a2 + s2 )(b2 + s2 ) 2ab(a + b)

Similarly using the Parseval relation for sine transforms

Z ∞ Z ∞
Fs (s)Gs (s)ds = f (t)g(t)dt
0 0

we find (DERIVE THIS RESULT)

∞
s2 π
Z
ds = .
0 (a2 + s2 )(b2 + s2) 2(a + b)

17