Math Methods I

Lia Vas

Fourier Series. Fourier Transform

Fourier Series.
Recall that a function differentiable any number of times at x = a can be represented as a power
series

X
f (n) (a)
an (x a)n where the coefficients are given by an =
n!
n=0
Thus, the function can be approximated by a polynomial. Since this formula involves the n-th derivative, the function f should be differentiable n-times at a. So, just functions that are differentiable
any number of times have representation as a power series. This condition is pretty restrictive because any discontinuous function is not differentiable. The functions frequently considered in signal
processing, electrical circuits and other applications are discontinuous. Thus, there is a need for a
different kind of series approximation of a given function.
The type of series that can represent a a much
larger class of functions is called Fourier Series.
These series have the form

a0 X
2nx
2nx
f (x) =
+
(an cos
+ bn sin
)
2
T
T
n=1
The coefficients an and bn are called Fourier coefficients.
Note that this series represents a periodic function with period T.
To represent function f (x) in this way, the function has to be (1) periodic with just a finite number
of maxima and minima within one period and just a finite number of discontinuities, (2) the integral
over one period of |f (x)| must converge. If these conditions are satisfied, the Fourier coefficients an
and bn can be computed using the formulas
Z
Z
2 T /2
2nx
2 T /2
2nx
an =
f (x) cos
dx
bn =
f (x) sin
dx
T T /2
T
T T /2
T
or, more generally
Z
2 x0 +T
2nx
an =
f (x) cos
dx
T x0
T

2
bn =
T

x0 +T

f (x) sin
x0

2nx
dx
T

When studying phenomena that are periodic in time, the term 2

in the above formula is usually
T
replaced by and t is used to denote the independent variable. Thus,

a0 X
f (t) =
+
(an cos nt + bn sin nt)
2
n=1
1

In cases when the independent variable represents the distance instead of the time, the period 2L
is more convenient. Then the above formulas become:
1
an =
L

nx
f (x) cos
dx
L
L
f (x) =

1
bn =
L

f (x) sin
L

nx
dx
L

nx
nx
a0
(an cos
+
+ bn sin
)
2
L
L
n=1

As a special case, if a function is periodic on [, ], these formulas become:

Z
Z
1
1
an =
f (x) cos nx dx
bn =
f (x) sin nx dx

f (x) =

a0 X
(an cos nx + bn sin nx)
+
2
n=1

If f (x) has a Fourier series expansion

f (x) =

a0 X
2nx
2nx
+
(an cos
+ bn sin
)
2
T
T
n=1

one can prove the formulas for Fourier series coefficients an by multiplying this formula by cos 2nx
T
and integrating over one period.
Z

T /2

2nx
f (x) cos
dx =
T
T /2

X
m=1

T /2

T /2

T /2

2mx
2nx
am cos
cos
dx +
T
T
T /2

a0
2nx
cos
dx +
2
T

T /2

2mx
2nx
bm sin
cos
dx
T
T
T /2

All the integrals with m 6= n are zero and the integrals with both sin and cosine functions are zero
as well. Thus,
Z T /2
Z T /2
2nx
2nx
2nx
dx =
an cos
cos
dx = an T
f (x) cos
T
T
T
T /2
T /2

1
an =
T

T /2

f (x) cos
T /2

2nx
dx
T

R T /2
R T /2
if n > 0. If n = 0, we have that T a0 = T /2 f (x)dx and so a0 = T1 T /2 f (x)dx.
The formula for bn is proved similarly, multiplying by sin 2nx
instead of cos 2nx
.
T
T
Symmetry considerations. Note that if f (x) is even (that is f (x) = f (x)), then bn = 0. Since
RL
R
R
nx
1 0
1 L
f
(x)
sin
f (x) sin nx
dx
=
dx
+
dx. Using the substitution u =
bn = L1 L f (x) sin nx
L
L L
L
L 0
LR
1 0
x for the first integral we obtain that the first integral is equal to L L f (u) sin nu
du. Using that
L
R
R0 RL
nu
nu
1 L
f (x) = f (x), that sin L = sin L , and that L = 0 , we obtain L 0 f (u)( sin nu
) du =
L
R
1 L
nu
f (u) sin L du. Note that this is exactly the negative of the second integral. Thus, the first
L 0
and the second integral
we obtain
R cancel and
R that bn = 0.
R
1 0
1 L
nx
1 L
f (x) cos nx
Similarly, an = L L f (x) cos L dx = L L f (x) cos nx
dx+
dx. Using the subL
L 0
L R
1 0
du.
stitution u = x for the first integral we obtain that the first integral is equal to L L f (u) cos nu
L
R0 RL
R
nu
nu
1 L
nu
Using that f (x) = f (x), that cos L = cos L , and that L = 0 , we obtain L 0 f (u) cos L .
Note that this is equal to the second integral. Thus, the two integrals can be combined and so
2
an =
L

f (x) cos
0

nx
dx
L

and

f (x) =

a0 X
nx
+
an cos
.
2
L
n=1

If f (x) is odd, using analogous arguments, we obtain that an = 0 and

2
bn =
L

f (x) sin
0

nx
dx
L

and

f (x) =

X
n=1

bn sin

nx
.
L

These last two power series are called Fourier cosine expansion and Fourier sine expansion
respectively.
Example 1. The input to an electrical circuit that switches between a high and a low state with
time period 2 can be represented by the boxcar function.

f (x) =

1
1

0x<
x < 0

The periodic expansion of this function is called the square wave function.
More generally, the input to an electrical circuit that switches from a high to a low state with time
period T can be represented by the general square wave function with the following formula on
the basic period.

f (x) =

1
1

0 x < T2
T
x<0
2

Find the Fourier series of the square wave and the general square wave.
Solutions. Graph the square wave function and note it is odd. Thus, the coefficients
of the cosine
R
terms will be zero. The coefficients of the sine terms can be computed as bn = 1 f (x) sin nxdx =
R
2
sin nxdx = 2
cos nx|0 = 2
((1)n 1). Note that (1)n 1 = 11 = 0 if n is even (say n = 2k)
0
n
n
4
(2) = (2k+1)
.
and (1)n 1 = 11 = 2 if n is odd (say n = 2k+1). Thus, b2k = 0 and b2k+1 = 2
n
Hence,
P
P sin(2k+1)x
1
4
f (x) = 4
sin
nx
=
= 4 (sin x + sin33x + sin55x + . . .).
k=0
n

2k+1
n=odd
For the general square wave, analogously to this previous consideration you obtain that an = 0,
P
2(2k+1)x
1
4
b2k = 0 and b2k+1 = (2k+1)
sin 2(2k+1)x
. Thus, f (x) = 4
.
k=0 2k+1 sin
T
T
If an arbitrary function f (x), not necessarily even
extend it to an even function

f (x)
g(x) =
f (x)
and consider its Fourier cosine expansion:
Z
2 L
nx
an =
f (x) cos
dx and
L 0
L

or odd, is defined on the interval (0, L), we can

0<x<L
L < x < 0

nx
a0 X
an cos
f (x) =
+
.
2
L
n=1

Similarly, we can extend f (x) to an odd function


f (x)
h(x) =
f (x)
and consider its Fourier sine expansion:
Z
2 L
nx
bn =
f (x) sin
dx
L 0
L

0<x<L
L < x < 0

and

f (x) =

X
n=1

bn sin

nx
.
L

To simplify the answers, sometimes the following identities may be useful

sin(n) = 0,

cos(n) = (1)n .

If n = 2k + 1 is an odd number,
cos

n
(2k + 1)
= cos
= 0,
2
2

sin

n
(2k + 1)
= sin
= (1)k .
2
2

Example 2.
Find the Fourier cosine expansion of

f (x) =

x
2x

0<x1
1<x<2

R2
R1
R2
Solutions. an = 0 f (x) cos nx
dx = 0 x cos nx
dx + 1 (2 x) cos nx
dx. Using integration by
2
2
2
2
nx
parts with u = x, v = n sin 2 for the first and u = 2x and same v for the second, you obtain an =


1  2(2x)
2
n
2x
nx
4
nx
nx
4
nx
+
|
|21 =
+ n242 cos n
sin
+
cos
sin

cos
sin
n242
0
n
2
n2 2
2
n
2
n2 2
2
2
n
2
| {z }
2
n
4
4
n
n
4
4
8
cos n
sin
+ n2 2 cos 2 = n2 2 cos 2 n2 2 n2 2 cos n = n242 (2 cos n
1 cos n).
n2 2
2
|n {z 2}
4
4
k
If n = 2k + 1 is odd, an = (2k+1)
2 2 (0 1 + 1) = 0. If n = 2k is even, an = (2k)2 2 (2(1) 1 1).
Because of the part with (1)k , we can distinguish two more cases depending on whether k is
even or odd. Thus, if k = 2l is even, an = (4l)42 2 (2 1 1) = 0. If k = 2l + 1 is odd, an =
4
16
4
(2(1) 1 1) = (4l+2)
2 2 = (2l+1)2 2 .
(2(2l+1))2 2
R2
R1
R2
If n = 0, a0 = 0 f (x)dx = 0 xdx + 1 (2 x)dx = 21 + 12 = 1.
P
P
(4l+2)x
1
1
= 12 42
So, f (x) = 21 42
l=0 (2l+1)2 cos
l=0 (2l+1)2 cos(2l + 1)x.
2

Non-periodic Functions. A non-periodic function given on an interval can be extended to a

periodic function outside that interval and then represented by a Fourier series. We can sometimes
extended to an even or odd function and make the calculations simpler.
Example 3. Find the Fourier series for f (x) = x2 for 0 < x 2 by extending the function to an
even periodic function.
Solutions. Note that x2 is already an even function. So, consider this function on interval [-2,
2] and replicate its graph on this domain outside of this interval too to create a periodic function of
period 4.R Since the new function is even too, bn = 0.
R
2
2
an = 0 x2 cos nx
dx. Using integration by parts with u = x2 and v = cos nx
dx = n
sin nx
,
2
2
2
R
2
4
2 2
nx 2
nx
we obtain that an = n x sin 2 |0 n 0 x sin 2 dx. The first term
and the second term
R is zero
nx
require another integration by parts, this time with u = x and v = sin 2 dx = 2
cos nx
. Thus
n
2
8
nx 2
16
nx 2
an = n2 2 x cos 2 |0 n3 3 sin 2 |0 . The first term is zero in the lower bound and the second term
16(1)n
is zero at both bounds. Thus an = n16
. Note that this formula works just for n > 0
2 2 cos n =
n2 2
so a0 has
to
be
computed
separately.
R2
3
a0 = 0 x2 dx = x3 |20 = 38 .
P
(1)n
nx
Thus, the Fourier series is x2 = 43 + 162
n=1 n2 cos 2 .
Complex Fourier Series. The complex form of Fourier series is the following:

f (x) =

X
n=

cn e

2nix
L



Z
2nix
2nx
2nx
1 L
=
cn cos
+ i sin
where cn =
f (x)e L dx.
L
L
L 0
n=

Analogously,
f (x) =

X
n=

cn e

2nix
T



Z
2nix
2nx
2nx
1 T /2
cn cos
+ i sin
where cn =
f (x)e T dx.
=
T
T
T T /2
n=

If f (x) is a real function, the coefficients cn satisfy the relations cn =

1
(a + ibn ) for n > 0. Thus, cn = cn for all n > 0. In addition,
2 n

1
(a
2 n

ibn ) and cn =

an = cn + cn and bn = i(cn cn ) for n > 0 and a0 = 2c0 .

These coefficients cn are further associated to f (x) by Parsevals Theorem. This theorem is
related to conservation law and states that
Z

X
1 L
a20 1 X 2
2
2
+
(an + b2n ).
|f (x)| dx =
|cn | =
L 0
4
2 n=1
n=
Note that the integral on the left side computes the average value of the moduli squared of f (x)
over one period and the right side is the sum of the moduli squared of the complex coefficients. The
proof of this theorem can be your project topic.
2
Example 4. Using the Fourier series
P for1 f (x) = x for 0 < x < 2 from Example 3 and Parsevals
Theorem, find the sum of the series n=1 n4 .

Solutions. Parsevals Theorem applied to this function results in the following


Z


a20 1 X 2
16 162 X 1
1 2 4
2
x dx =
+
(a + bn ) =
+ 4
+0 .
2 0
4
2 n=1 n
9
2 n=1 n4
R2
Note that integral on the left side is 12 0 x4 dx = 16
. Dividing the equation above by 16 produces
5

X
1
1
1
8 X 1
4
= + 4

=
.
4
5
9 n=1 n4
n
90
n=1

Practice Problems.
(1) Find the Fourier series of

f (x) =

1x
1+x

0x<1
1 < x < 0

(2) Find the Fourier sine expansion of


f (x) =

x
2x

0<x1
1<x<2

(3) Note that the boxcar function from Example 1 represents the odd extension of the function
f (x) = 1 for 0 x < . Consider the even extension of this function and find its Fourier
cosine expansion.
(4) The output from an electronic oscillator is the sawtooth function f (t) = t for 0 t 1 that
keeps repeating with period 1.
(a) Sketch this function and represent it using a complex Fourier series.

(b) Use the Fourier series from part (a) and Parsevals Theorem to find the sum of the series

X
1
.
n2
n=1

(5) Use the Fourier series you obtain in Example 1 to find the sum of series

X
(1)n
.
2n + 1
n=0

Solutions.
R1
(1) Graph the function and note it is even. Thus, bn = 0R and an = 2 0 (1 x) cos(nx) dx. Using
1
sin(nx), obtain that
the integration by parts with u = 1 x and v = cos(nx)dx = n
R
1
2
2
2
an = n (1 x) sin(nx)|10 + n 0 sin(nx) dx = 0 n2 2 cos(nx)|10 = n222 ((1)n 1). This
2
last expression is 0 if n is even and equal to n222 = (2k+1)
2 2 if n = 2k + 1 is odd. Note that
2
the formula n2 2 ((1)n 1) does not compute a0 because of the n in the denominator so
R1
2
calculate a0 from the formula a0 = 2 0 (1 x) dx = x x2 |10 = 21 . This gives you the Fourier
P
4
series expansion f (x) = 14 +
k=0 (2k+1)2 2 cos(2k + 1)x.

1
R2
R1
R2
(2) 
bn = 0 f (x) sin nx
dx = 0 x 
sin nx
dx+ 1 (2x) sin nx
dx = 2x
cos nx
+ n242 sin nx
|0 +
2
2
2
n
2
2
2(2x)
2
|21 = 2
+ n242 sin n
+ n242 sin n
cos nx
n242 sin nx
cos n
+ n
cos n
= n282 sin n
.
n
2
2
n
2
2
2
2
2
P
k
k

8(1)
(1)
(2k+1)x
8
This is 0 if n is even. If n = 2k + 1, this is (2k+1)
.
2 2 . So, f (x) = 2
k=0 (2k+1)2 sin
2
(3) The even extension of f (x) is f (x) = 1 for < x < . The periodic extension of this
function is constant function equal to 1 for every x value. Note that this is already in the
form of a Fourier series with bn = 0 for every n, an = 0 for n >R 1 and a0 = 2. Computing the

FourierRcoefficients would give you the same answer: an = 2 0 cos nx dx = 0 if n > 1 and
P

a0 = 2 0 dx = 2. Thus, f (x) = 22 +
n=1 0 = 1. This answer should not be surprising since
this function is already in the form of a Fourier series

X
(0 cos nx + 0 sin nx).
1=1+
n=0

R1
P
2nit
and cn = 0 te2nit =
(4) (a) The graph of the function is given below. f (t) =
n= cn e
t
1
e2nit |10 + 4n12 2 e2nit |10 = 2ni
e2ni + 4n12 2 e2ni 4n12 2 . Note that e2ni =
2ni
1
i
i
cos(2n) + i sin(2n) = 1. Thus cn = 2ni
+ 0 = 2n
. Note that cn = 2n
= cn .
R1
1
c0 = 0 t dt = 2 .
P
i
2nit
This gives us f (t) = 12 +
.
n=,n6=0 2n e

Note also that a0 = 2c0 = 1, an = 0 for n > 0 and bn =

1
.
n

R1
P
P 1
1
1
1
1
(b) Parsevals Theorem gives us that 0 x2 dx = 41 + 12

=
+
2
2
2
n=1 n
n=1 n2
3
4
2
P 1
2
n=1 n2 = 6 .
P
sin(2n+1)x
(5) In Example 1, we obtain that f (x) = 4
where f (x) is the square wave function.
n=0
2n+1

P
P
P (1)n
sin(2n+1)

(1)n

Choosing x = 2 , we have that 1 = 4 n=0 2n+1 2 = 4

n=0 2n+1
n=0 2n+1 = 4 .

Fourier Transformation
The Fourier transform is an integral operator meaning that it is defined via an integral and that
it maps one function to the other. If you had a differential equations course, recall that the Laplace
transform is another integral operator you have encountered.
The Fourier transform represents a generalization of Fourier series. Recall that the Fourier series
P
2nit
T
is f (t) =
.
n= cn e
The sequence cn can be regarded as a function of n and is called Fourier spectrum of f (t).
We can think of c(n) being another representation of f (t), meaning that f (t) and c(n) are different
representations of the same object. Indeed: given f (t) the coefficients c(n) can be computed and,
conversely, given c(n), the Fourier series with coefficients c(n) defines a function f (t). We can plot
c(n) as a function of n (and get a set of infinitely many equally spaced points). In this case we think
of c as a function of n, the wave number.
We can also think of c as a function of = 2n
, the frequency. If T is large, then is small,
T
so for large T, we can think of c() being a continuous function. Also, two consecutive n values are
d
length 1 apart so dn = 1. Thus, d = 2
dn d = 2
and T2
= 1. Thus we have
T
T
Z
Z T /2
2nit
1 T /2
T
cn =
f (t)e
dt T c() =
f (t)eit dt
T T /2
T /2
and so

f (t) =

cn e

n=

2nit
T

T /(2)=

1
T d
c()eit =
2
2

T c()eit d.

T /(2)=

When we let T , the above expression become



Z Z
Z 
Z
1
1
1
it
it
it

f (t)e
dt e d =
f (t) =
f (t)e
dt eit d
2
2
2

We denote the expression in parenthesis by F () and so get the final formulas:

F () =
f (t) =

1
2

1
2

f (t)eit dt

F ()eit d

F () is the Fourier transform of f (t).

f (t) is the inverse Fourier transform of F ().

We can still think of f (t) and F () being the same representations of the same object: the first
formula above computes F () for given f (t) and the second one computes f (t) for given F (). To

understand the significance of this, it is helpful to think of f (t) as of a signal which can be measured
in time (so f (t) can be obtained) but that needs to be represented as a function of frequency, not
time. In this case, Fourier transform produces representation of f (t) as a function F () of frequency
.
The Fourier transform is not limited to functions of time and temporal frequencies. It can be used
to analyze spatial frequencies. If the independent variable in f (t) stands for space instead of time, x
is usually used instead of t. In this case, the independent variable of the inverse transform is denoted
by k.
t x
and
k
Mathematically, the importance of the Fourier transform lies in the following:
(1) If the initial function f (t) has the properties that are not desirable in a particular application
(e.g. discontinuous, non smooth), we can consider the function F () instead which is possible
better behaved.
(2) Fourier transform represents a function that is not necessarily periodic and that is defined on
infinite
interval. The only requirement for the Fourier transform to exist is that the integral
R
|f
(t)|dt
is convergent.

(3) Fourier transformation is a generalization of Fourier series.

In physics, on the other hand, Fourier transform is used in many sub-disciplines. One of the most
important applications of Fourier transform is in signal processing. We have pointed out that the
Fourier transform presents the signal f (t), measured and expressed as functions of time, as a function
F () of frequency. The transform F () is also known as the frequency spectrum of the signal.
Moreover, Fourier transform provides information on the amplitude and phase of a source signal
at various frequencies. The transform F () of a signal f (t) can be written in polar coordinates as
F = |F |ei
The modulus |F | represents the amplitude of the signal at respective frequency , while (given by
arctan( Im F/ Re F ) computes the phase shift at frequency . This gives rise to various applications
cryptography, acoustics, optics and other areas.
In addition, Fourier transform can be used similarly to Laplace transform: in converts a differential
equation into an ordinary (algebraic) equation that is easier to solve. After solving it, the solution
of the original differential equation can be obtained by using the inverse Fourier transform.
Example 1. Find the Fourier transform F11 () of the boxcar function.

1
1 < t < 1
1
f1 (t) =
0
otherwise
Express your answer as real function using the identity sin x =
Generalize your calculations to find the Fourier
transform FaA () for the general boxcar function.

A
a < t < a
A
fa (t) =
0
otherwise

1
(eix
2i

eix ).

10

The boxcar function is said to be normalized

if the total area under the function is 1. In this
1
case A = 2a
. We shall denote the normalized boxcar function that is non-zero on interval [a, a] by
fa .
Sometimes the values of fa at x = a and
1
. For example, if
x = a are defined to be 4a
a = 21 , the following graph represents fa .
Consider how changes in value of a impact the
shape and values of Fa ().
R
Solution. F () = 12 f (t)eit dt. However, since f (t) = 0 for t < 1 and t > 1, and
R1
i ei
it 1
1 e
=
f (t) = 1 for 1 t 1, we have that F () = 12 1 eit dt = 1
e
|
=
1
i
2i
2
i
i
1 e e
i
2

i
i
2 e e
2i
2

2
2

sin .

Similarly, ifRa boxcar function of height A is nonzero on interval [a, a], the Fourier transform is
a
A
eia eia
= 12 a Aeit dt = 2i
(eia eia ) = 2A
= 2A
sin a.
2i
2
2

FaA ()

Function

sin x
x

is known as sinc function.

sinc(x) =

sin x
x

Sometimes the normalized sinc function

used. Note that limx0 sinc(x) = 1 and
sinc(x) = 0.

sin x
is
x
limx

Using the sinc notation, we can represent

the Fourier transform of the boxcar function faA (t) as the sinc function FaA () =

2aA

sinc(a).
2

11

If the box function is normalized, A =

this function has value 12 .

1
2a

so Fa () =

1 sinc(a).
2

Since sinc(0)=1, the peak of

Compare the graphs of fa and Fa for several different values of a. You can notice that larger values
of a correspond to fa having smaller hight and being more spread out. In this case, Fa has a narrow,
sharp peak at = 0 and converges to 0 faster. If a is small, fa has large height and very narrow
base. In this case, Fa has very spread out peak around = 0 and the convergence to 0 is much
slower.
In the limiting cases a 0 represents a constant function and a represents Dirac delta
function (t).
Dirac delta function or impulse function is
used to represent phenomena of an impulsive nature. For example, voltages that act over a very
short period of time.
Now let us define
(t) = lim fa (t).
a0

Since the area under fa (t) is 1, the area under (t) is 1 as well. Thus, (t) is characterized by the
following properties:
R
(1) (t) = 0 for all values of t 6= 0
(2)
(t)dt = 1.

Since no ordinary function satisfies both of these properties, is not a function in the usual sense of
the word. It is an example of a generalized function or a distribution. Alternate definition of Dirac
delta function can be found on wikipedia.
We have seen that Fa for a small a is very spread out and almost completely flat. So, in the
limiting case when a = 0, it becomes a constant. Thus, the Fourier transform of (t) can be obtained
as limit of Fa () when a 0. We have seen that this is a constant function passing 12 . Thus,
the Fourier transform of the delta function (t) is the constant function F () =

1 .
2

Conversely, when a fa 12 . The Fourier transform of that is limit of Fa for a .

We have seen that this limit function is zero at all nonzero values of . So, we can relate this limit
to (). Requiring the area under Fa to be 1 can explain why normalized sinc function is usually
considered instead of regular sinc.
Example 2. Find the inverse Fourier transforms of boxcar and Dirac delta functions.
Solutions. The inverse transform of the general boxcar function can be computed as
Z a
1
A
2A eiat eiat
2A
2aA sin at
2aA

Aeit d =
(eita eiat ) =
=
sin at =
= sinc(at).
2i
2 a
2it
2t
2t
2 at
2
Thus, the inverse transform of the boxcar function is the sinc function. This also illustrates that

12

the Fourier transform of the sinc function is the boxcar function.

Taking the limit when a 0 in the normalized case, we obtain that the inverse Fourier of the
delta function is the limit of 12 sinc(at) when a 0 and that is the constant function passing 12 .
Thus,
the Fourier transform of the constant function f (t) =

1
2

is the delta function ().

Example 3. Consider the Gaussian probability function (a.k.a. the bell curve)
2

f (t) = N eat

where N and a are constants. N determines the height of the peak and a determines how fast it
decreases after reaching the peak. The Fourier transform of f (t) can be found to be
N
2
F () = e /4a .
2a
This is another Gaussian probability function. Examine how changes in a impact the graph of the
transform.
Solutions. If a is small, f is flattened. In this case the presence of a in the denominator of the
exponent of F () will cause the F to be sharply peaked and the presence of a in the denominator of
N , will cause F to have high peak value.
2a
If a is large, f is sharply peaked and F is flattened and with small peak value.
All the previous examples are related to Heisenberg uncertainty principle and have applications in
quantum mechanics. The following table summarize our current conclusions.
Function in time domain Fourier Transform in frequency domain

boxcar function

sinc function

sinc function

boxcar function

delta function

constant function

constant function

delta function

Gaussian function

Gaussian function

Symmetry considerations.
The following table displays the formulas of Fourier and inverse Fourier transform for even and
odd functions.

13

q R

F () = 2 0 f (t) cos t dt
q R

F () even f (t) = 2 0 F () cos t d

f (t) even

f (t) odd
F () odd

q R
2

F () = 0 f (t) sin t dt
q R

f (t) = 2 0 F () sin t d

Fourier sine and cosine functions. Suppose that f (t) is defined just for t > 0. We can extend
f (t) so that it is even. Then we get the formula for F () by using the formulas for even function
above. F () is then called Fourier cosine transformation.
Similarly, if we extend f (t) so that it is odd, we get the formula for F () same as for an odd
function above. F () is then called Fourier sine transformation.

Relation of Fourier transform and Fourier Series.

We illustrate this relation in the following example. Consider the Fourier series of a boxcar function
fa (t). Let sn denote the Fourier coefficient in the complex Fourier series. The value of a determines
sampling period T and spacing 0 in frequency-domain.
w0 =

2
2n
and =
= n0 .
T
T

In the following figures 1, we consider the boxcar function with a = 21 (so T = 1 and 0 = 2).
The first graph on each figure displays the Fourier transform, sinc function, in the frequency-domain.
The highlighted frequency on the first graph determines the value of n. The second graph displays
the harmonic function corresponding to n-th term of the Fourier series of the function f (t) in the
time-domain. The third graph is the sum of the first n terms of the Fourier series of f (t).
Note that as n the sum of Fourier series terms converge to the boxcar function.

1The

following figures are from the presentation on Fourier transform in Magnetic Resonance Imaging The Fourier
Transform and its Applications by Branimir Vasilic. The author is grateful to Branimir for making the slides available.

14

15

Discrete versus Continuous Functions.

In the previous examples, the function in frequency-domain was a continuous function. In applications, it is impossible to collect infinitely many infinitely dense samples. As a consequence, certain
error may come from sampling just finite number of points taken over a finite interval. Thus, it is
relevant to keep in mind what effects in time (resp. frequency) domain may have finite and discrete
samples of frequencies (resp. time).
frequency-domain

function in time-domain

effects in time-domain

infinite continuous set of frequencies

non-periodic function

none

finite continuous set of frequencies

non-periodic function

Gibbs ringing, blurring

infinite number of discrete frequencies

periodic function

aliasing

finite number of discrete frequencies

periodic function

aliasing, Gibbs ringing,

blurring

16

In the following consideration, assume that the sample consists of frequencies and that all the
values lie on the graph of a sinc function. The goal is to reconstruct the boxcar function in the
time-domain. The following four figures illustrate each of the four scenarios from the table.

Application in Magnetic Resonance Imaging.

The Fourier transform is prominently used in

Magnetic Resonance Imaging (MRI). The fist figure represents a typical MRI scanner. The scanner samples spacial frequencies and creates the
Fourier transform of the image we would like
to obtain. The inverse Fourier transform converts the measured signal (input) into the reconstructed image (output).
In the following two figures, the image on the right represents the amplitude of the scanned input
signal. The image on the left represents the output - the image of a human wrist, more precisely,
the density of protons in a human wrist. The whiter area on the image correspond to the fat rich
regions. The darker area on the image correspond to the water rich regions.

17

Each point on the input images corresponds

to certain frequency. Two smaller figures on the
right side represent the components of the inverse
transforms at two highlighted frequencies. The
output image is created by combining many such
images - one for each sampled frequency in fact.

The last figure represent the output with high (first pair of images) and low frequencies (second
pair of images) removed.
If high frequencies are removed (low-pass filter), the image becomes blurred and only shows the
rough shape of the object.
If low frequencies are removed (high-pass filter), the image is sharp but intensity variations are
lost.

Practice Problems.
(1) Find the Fourier transform of f (t) = et , t > 0, f (t) = 0 otherwise.
(2) Find Fourier cosine transformation of the function from the previous problem.
(3) Find cosine Fourier transform of f (t) = 2t 3 for 0 < t < 3/2, f (x) = 0 otherwise.
(4) If
a
fa (t) = 2
a + t2

18

the graph of f has a peak at 0. Since f (0) = a1 , the height is conversely proportional to a.
The Fourier transform of f can shown to be
p
Fa () = /2ea|| .
Graph Fa () for several values of a and make conclusion how values of a impact the graph
of Fa .

R
1
01
(5) Solve the equation 0 f (t) cos t dt =
0
>1
Solutions.
R
1
1

e(1+i)t |
(1) F () = 12 0 et eit dt = 2(1+i)
0 = 2(1+i) .
q R

(2) F () = 2 0 et cos t dt. Using two integration by parts with u = et , one obtains that
q
q
R
R t
1 t
2
1
F () = 2 ( 1 et sin t|
+
e
sin
t
dt)
=
(0 12 et cos t|
cos t dt) =
0
0 2 0 e
0

q
q
p
2 1
( 12 2 F (). Solving for F () gives your F ()(1 + 12 ) = 2 12 . Multiply by 2 to
2
q
q
2
2
get F ()( + 1) = F () = 2 21+1 .
q R
q
q
3/2 2 R 3/2
3/2
2 3/2
2 2t3
(3) F () = 0 (2t3) cos t dt = ( sin t|0 0 sin t dt) = 2 (0+ 22 cos t|0 ) =
q

2 2
2 2
3
(cos 3
(cos

1)
=
1).
2
2
2
2
(4) If a is small, then fa has a larger peak. In this case Fa is more spread out and flattened. If a
is large, fa is spread out and the height of the peak is not large. In this case, Fa has a sharp
peak and converges to zero fast.
q
(5) Use inverse cosine Fourier transform. First multiply with 2 to match the definition of the
transform. Then the left side is exactly the Fourier cosine transform of f (t).
q So we can get
f (t) as the inverse transform of the left side of the equation multiplied by 2 .
q q R
R1
1
f (t) = 2 2 0 (1 ) cos t d = 2 ( 1
sin t|10 + 1t 0 sin t dt) = 2 (0 t12 cos t|10 ) =
t
2
( cos t
t2

+ 1) =

2
(1
t2

cos t).