Fourier Transform

Introduction

The Fourier series expresses any periodic function into a sum of sinusoids. The Fourier Transform is the
extension of this idea to non-periodic functions by taking the limiting Form of Fourier series when the
fundamental period is made very large (infinite). Fourier transform finds its applications in astronomy,
signal processing, linear time Invariant (LTI) systems etc. We introduce the Fourier transform, a special
linear integral transformation forDifferential equations which are defined on unbounded domains. The
method is Associated with the French physicist and mathematician Joseph Fourier whose Work opened
up a new way to solve linear PDEs. As we will see later, the Fourier Transform reduces a partial
differential equation in (t, x)-domain to an ordinary Differential equation in (t, ω)-domain.

Definition

The Fourier transform is a linear integral transformation of a function f(x) in x-domain to another
function f^(ω) in ω-domain. As we will see in sequel, ω can be considered as the frequency variable and
thus f^(ω) represents the frequency distribution embedded in function f(x).

1 (Admissible functions) We first introduce the admissible space of functions.

Definition 1 :- A function f(x) - ∞ < x < ∞ is called integrable if the Following relation holds

For example, function

Is not admissible even it satisfies The following property

Note that

The Fourier transform is usually defined for admissible functions. However,As we will see below, the
definition can be extended for a wider class of Functions than the admissible ones.

2 (Definition of Fourier transform) Suppose f is a function (not necessarily admissible) defined on

R: ( -∞, ∞ ) Its Fourier transform F { f } : = F^(ω) is defined by the following integral

As long as the integral exists. The improper integral is understood in the following sense

Example 1 Consider the following function

Its Fourier transform is

Assume that f is an odd function and its transform is f^(ω). Prove That f^(ω) is a pure imaginary function
and odd, that is, f^(-ω) = - f^(ω). If f (x) is An even function, then f^(ω) is real and even function.

3.Note that the integral (8.2) is improper, and may be fail to exist even if f Is bounded. The following
proposition gives a sufficient condition for the Convergence of the integral. It is just a sufficient and not
necessary in any Means.

Proposition 8.1. If a function f is admissible then f^(ω) exists and continuous for all ω ∈ R, and
furthermore.
Example 2 Function

It not admissible, And the definition of Fourier Transform Fails as

Now, let us consider the following function

The function is admissible and its transform is

3 (Inverse transform) Suppose f^(ω) is a function defined on ω ∈ (-∞,∞).The inverse Fourier transform
of f^ is defined by the following integral

As long as the integral exists. The integral is understood in the following sense
The following theorem establishes the relationship between f , f-1. See the Appendix of this chapter for a
proof.

Theorem 1 Assume that f is an admissible function, then

Example 3 Let us verify the theorem for functions

we have

where we used the symmetry property for the integration. The following figures shows the value of the
integral for R = 20, 100 respectively. It is observed that for R→∞ the integral approaches f(x).
The following figure shows the result for R = 20,100.

4.( From Fourier series to Fourier transform) Remember the formula Of Fourier series defined for a
function f(x) on (-R, R)

Where ω =2π/2R let us denotes ω by just Δω and thus we have

Where

The Series

Can be approximated by the following series

And thus for R→∞, the above Riemann series approaches

And therefore

Compare the above formula with the Fourier inverse transformation.

5. (Sine Fourier transform) If f(x) is an odd function on (-), then Its Fourier transform reads

The above relation helps us to define the Fourier transform of functions f(x) defined only on the half-
interval (0,). If a function f(x) is defined only On (0,) then the transformation of its odd extension f odd (x)
on (-,♾️) is

and thus we can defined the sine Fourier transform of the original function f(x) as follows

Since

For x > 0. Similarly, the cosine Fourier transform of f(x) defined on (0, ) reads

with the following inverse for x > 0:

Fourier transform of important functions

As it is common in physics, ω is usually interpreted as the angular frequency (through The relation ω =
2πf) and therefore, f^(ω) provides us with the distribution of Frequency components embedded in a
function f(x). Let us see the transform of Some basic functions.

1 (A pulse) Let pa(x) denote the pulse

Clearly, pa is an admissible function, with the transform

Since f(x) is admissible and continuous and f^(ω) is integrable, we have

If we change x to -ω And ω to x, we reach the following relation

And therefore,

According to the following known fact

We obtain,
4. (Dirac delta) The concept of Dirac delta function is discussed in detail In the textbook on ODEs.
Remember that the delta function (x) has the Following property

For any function f that is continuous at x0. Note that is not a function in The usual sense according to
the following relation.
Remember that for any ordinary function f, the following property holds

For this reason, δ is called a generalized function. Working with generalized function like δ(x) should
be done with extreme cautious. The Fourier Transform of δ is as follows

For the inverse we have ,

It is left as an exercise to the reader to show the sequence

Is a delta sequence function, that is

For arbitrary bounded C1 function f. By the above justification, we can write

5. (Sine and cosine) Another class of non-admissible functions with the Fourier transform is sine and
cosine functions. First note that we can write

The geometrical interpretation of the above result is evident. In fact, cos(ω0x) Is a periodic function with
only one frequency component ω = ω0, and this Is what formula (8.8) tells us. Even though, negative
frequencies do not Make sense from physical point of view, it is necessary for mathematical
Formulations of the transformation. Since cos(ω0x) is an even function, the Negative component of
frequency -ω0 shows itself in the ω-domain.
Properties of Fourier Transform

1.(Linearity) For any a; b ∈ R, the transform is linear in the following sense

As long as f^ and g^ exist. If f , g are admissible functions, then it is simply

Verified function af + bg is admissible and thus F {af + bg} exists.

2.(Frequency shift) The multiplication f(x) cos(ω0x) causes a shift of frequency in

f^(ω) according to the following calculation

The first integral is just f^(ω - ω0) and the second one is f^(ω+ω0), and finally

Perhaps one of the most important applications of the property is frequency

Modulation. Assume we collect online data fi(t), I = 1,…… n from different Users to
transmit them along a cable. By allocating the same frequency band To each user,
fi(t) is multiplied by cos(iω0t) to shift its Fourier transform To iω0. This method
prevents the interference of transmitted data
3. ( shift in x) There is a beautiful duality between shift in x and ω-domain Which
is stated in the following calculation

This property is widely used in the analysis of linear time invariant (LTI) Systems.

4. (Differentiation and integration) If f(x) is admissible and differentiable, then

As long as f^(ω) exists. The property is simply proved by definition (8.2).Similarly,

if F(x) an anti-derivative of f is admissible, then by the relation F {F’} = f^(ω) and
(8.9) we obtain

The formula (8.10) should be used with cautious for non-admissible functions.For
example, we have

Thus, one may be teempted to write

that is not correct. In fact, according to the inverse transform, we have

But

For this reason, we take the transform of u(x) as follows

Note that by (8.6) , we have

It is left as an exercise to reader to check the following property

5. (Multiplication by x) Assume that f(x) is admissible, then we have the Following
property that is crucial in solving differential equations

As long as f^ is differentiable. The justification is as follows

The following theorem gives a sufficient condition that the Fourier Transform of a
function is continuously differentiable.

Theorem 2 Assume that f is admissible function and decays exponentially at

+- , that is,

For |X |>M for some sufficiently large M. Then f^(ω) is continuously differentiable.

6. (Expansion and contraction) Let a > 0 and consider a function f(ax).We have

For a < 0 the transform is as

and thus for a ≠ 0, we obtain

On the other hand, for 0 < a < 1, function f(ax) shows an expansion in x-domain,
and a contraction for a > 1. Therefore, a contraction in x-domain Causes an
expansion in ω-domain and vice versa.

Fancy properties of the Fourier transform

1.( Convolution)

In the first volume of this book, we discussed in detail the concept of convolution
And its role in the study of linear time invariant (LTI) systems. Recall that the
Convolution of two functions f , g is defined by the following formula

as long as the integral exist. It is simply seen that if f , g are admissible then the
convolution exists. In fact, since f , g are bounded (why?), we can write
On the other hand , if f , g are admissible , their transform exists. We have the
following important property

The convolution formula is simply proved by the following justification. According

To the Fubini theorem, we can change the order of the integrals and write

Taking z = x – y ,we obtain,

And thus,

We accept the formula (8.11) for non-admissible functions like Dirac delta
function As well.
2. Energy and Plancherel theorem
In physics and engineering, the energy of a function f(x) is defined by the following
Integral
As long as the integral is bounded. The Plancherel theorem states a relationship
Between the energy of f(x) and the energy of f^(ω).

Theorem 3. Assume that f(x) is admissible and square integrable, E(f) < .Then the
following relation holds

If we define ,
3.Nyquist-Shanon rate

Let f(t) be a an electrical signal to be transmitted through a long

transmission line.Let f(t) be a an electrical signal to be transmitted
through a long transmission line.In real applications, the sampled signal
(the digital signal) is transmitted instead of F(t). The original signal at
the destination is reconstructed from the sampled one.What should be
the sampling rate that the transmitted signal keeps all information In the
original one and can be reconstructed at the origin without any loss? The
Answer is given by Nyquist.That Is an important result in digital signal
transmission. From the information point of View, sampling a signal
with the Nyquist rate will preserve all information in that Signal. By
Fourier series, we know that f^ has the following representation.
Therefore, the original function f(t) can be reconstructed by the inverse
transform of g^(ω). In fact, we have
4. Uncertainty principle
Here, let us introduce one important result of the above property which
is called the Uncertainly principle. First, let us defined a measure for the
dispersion of a function F around x = 0 (similarly for the dispersion of f^
around ω = 0). This is defined by the following formula

Provided that they both integrals exists.

Theorem 4. As long as D(f) is bounded, the following inequality which is called The uncertainty principle
holds

The above principle states that the more certainty about f(x) (more concentration around x = 0) leads to
less certainty about f^ (wider around ω = 0) and vice Versa. This is the reason why the inequality is
called the uncertainty principle. For A simple proof of the theorem see the problem set.

5. Higher dimensional transform

We can define the Fourier transformation for multi-variable functions as well. Let f: R² → R be Well
defined function. The 2D transform of f is defined by the following expression.
As long as the integral exists. Here ωx;ωy are just two parameters (you can interpret Them as the
frequency components of f with respect to x , y). Similarly, the inverse transform is defined by the
following formula