You are on page 1of 16

Harmonic Analysis

Aakash Jog
2015-16

Contents
1 Lecturer Information

2 Required Reading

3 Additional Reading

Basic Definitions and Theorems

1 Sequences and Series

2 Periodic Functions

3 Odd and Even Functions

II

Introduction to Fourier Series

1 Real Fourier Series

This work is licensed under the Creative Commons Attribution-NonCommercialShareAlike 4.0 International License. To view a copy of this license, visit http:
//creativecommons.org/licenses/by-nc-sa/4.0/.

2 Complex Fourier Series

3 Bessels Inequality

4 Riemann-Lebesgues Lemma

5 Dirichlets Kernel

11

6 Relation between Fourier Coefficients of f (x) and f 0 (x)

15

Lecturer Information

Barak Sober
Office: Shenkar Physics 201
E-mail: barakino@gmail.com
Telephone: +972 3-640-6024

Required Reading
1. Folland, G.B.: Fourier Analysis and its applications, Wadsworth &
Brooks/Cole mathematics series, 1992

Additional Reading
1. Katznelson, Yitzhak. An introduction to Harmonic analysis. Cambridge
University Press, 2004.

Part I

Basic Definitions and Theorems


1

Sequences and Series

Definition 1 (Convergent series). The series

P
n=0

sequence of partial sums SN =

N
P
n=0

an is said to converge if the

an converges to a finite limit.

Definition 2 (Pointwise convergence of sequence of functions). Let D R,


and {fn (x) : D R} be a sequence of functions. fn (x) is said to converge
pointwise, to a limit function
f (x) on D, if > 0, x D, N N, such



that n > N , fn (x) f (x) < .
Definition 3 (Uniform convergence of sequence of functions). Let D R,
and {fn (x) : D R} be a sequence of functions. fn (x) is said to converge
uniformly
to f (x) on D if > 0, N N, such that, n > N , x D,

fn (x) f (x) < .
U

Theorem 1. If fn (x)
f (x), then
n=1 are continuous functions, and fn (x)
f (x) is also continuous.


Theorem 2. If a sequence of functions converges pointwise as well as uniformly, then the limit function must be the same.
Theorem 3 (Weierstrass M-test). If |uk (x)| ck on D for k {1, 2, 3, . . . }
and the numerical series

ck converges, then the series of functions

k=1

uk (x)

k=1

converges uniformly on D.

Periodic Functions

Definition 4 (Periodic functions). A function f : R R is said to be


periodic if 0 < L R, such that x R,
f (x) = f (x + L)
For a function
f (x) = k, as any
positive number is a
period, there is no
minimum L. Hence,
@L .

If there exists a minimum L, it is called L , the fundamental period.


4

Odd and Even Functions

Definition 5 (Odd functions). A function is said to be odd if f (x) = f (x).


Definition 6 (Even functions). A function is said to be even if f (x) = f (x).

Odd functions are


symmeteric about the
origin.
Odd functions are
symmeteric about the
y-axis.

Theorem 4. If h(x) is odd,


L
h(x) dx = 0
L

Part II

Introduction to Fourier Series


1

Real Fourier Series

Definition 7. Let f : [L, L] R, where L > 0. If x [L, L], then

X
n
n
1
f (x) = a0 +
an cos
x + bn sin
x
2
L
L
n=1

!

Theorem 5. Let L > 0, m W, n W.


Then
L
L

;



m
n
cos
x cos
x dx = L ;

L
L

2L ;


m 6= n
m = n 6= 0
m=n=0

Proof.
L
E=

n
m
x cos
x dx
cos
L
L


L
cos cos =
cos(+)
2

cos()
2

1
cos (m + n) x + cos (m n) x
2
L
L


!

If m 6= n,

E=

1 sin

(m + n) L x
(m + n) L

sin

(m n) L x
(m n) L

=0
If m = n 6= 0,
!
L


1

E=
cos 2m x + 1 dx
2
L
L

1
=
2

1
cos 2m x dx + x
L
2 L


=L
6

 L




dx

If m = n = 0,
L
E=

cos(0) cos(0) dx
L

= x|LL
= 2L

Theorem 6. Let L > 0, m N, n N.


Then

L




0
m
n
sin
x sin
x dx =
L
L
L

;
;

m 6= n
m=n

Theorem 7. Let L > 0, m W, n W.


Then
L




m
n
sin
x cos
x dx = 0
L
L
L

Assuming f (x) is known, and assuming that it can be integrated term by


term,
L

L
f (x) dx =

L
f (x) dx =

1
2

X
1
an
a0 dx +
2
n=1

cos n x dx + bn
L


sin n x dx
L


L
a0 dx
L

1
= a0 2L
2
L
1
a0 =
f (x) dx
L
L

Similarly, multiplying the series with cos m L x for m 6= 0 and integrating,


1
am =
L

f (x) cos m x dx
L


for m N.
Similarly, for m N \ {0},
L

1
bm =
L

f (x) sin m x dx
L


Definition 8. The expansion

X
1

f (x) a0 +
an cos n x + bn sin n x
2
L
L
i=1

where, for m N,
1
am =
L

f (x) cos m x dx
L


and, for m N \ {0},


1
bm =
L

f (x) sin m x dx
L


is called the Fourier Series of f (x).

Complex Fourier Series

By Eulers formula,

1  i
e + ei
2

1  i
sin =
e ei
2i

cos =

Therefore,
1
i
=
2i
2
8

!

Therefore, substituting in the Fourier series,

X
in
in
in
1
1 in x
i
f (x) a0 +
an
e L + e L x bn
e L x e L x
2
2
2
n=1

!

X
in
in
1
1
i
1
i
e L x
a0 +
an bn + e L x
an + b n
2
2
2
2
2
n=1

1
X
X
in
in
i
i
1
1
1
e L x
e L x
an b n +
an + b n
a0 +
2
2
2
2
2
n=
n=1

cn e

in
x
L

n=

Bessels Inequality

Definition 9 (Piecewise continuous functions). f : R R is said to be


piecewise continuous if, for every finite interval [a, b] there is a finite number
of discontinuity points, and the one-sided limits at each of these points are
also finite.
Definition 10 (Piecewise continuously differentiable functions). f : R R
is said to be piecewise continuously differentiable if it is piecewise continuous,
and
f (x + h) f (x+ )
<
lim+
h0
h
and
f (x + h) f (x )
lim
<
h0
h
Theorem 8 (Bessels Inequality). Let f (x) be a piecewise continuous function
defined on [L, L]. Then

1 2 X
1
a0 +
an 2 + b n 2
2
L
n=1

L
f (x)2 dx
L

Riemann-Lebesgues Lemma

Theorem 9 (Riemann-Lebesgues Lemma). If f (x) is piecewise continuous


on [L, L], then
lim an = lim bn = 0

Proof. By Bessels Inequality,

1 2 X
a0 +
an 2 + b n 2
2
n=1
As the function is
piecewise continuous
in [L, L], its integral
from L to L is finite.

L
f (x)2 dx
L

1
a0 2 +
an 2 + b n 2 <
2
n=1
Therefore,
lim an 2 n
lim an 2 + bn 2

n
lim 0
Therefore,
lim an = 0

Similarly,
lim bn = 0

Exercise 1.
If f (x) is piecewise continuous on [, ], show that

lim

1
f (x) sin n +
x dx = 0
2

Solution 1.

1
1
1
sin n +
x = sin(nx) cos
x + cos(nx) sin
x
2
2
2
Therefore, the limit is
!

1
0 = lim
f (x) cos
x sin(nx) dx
n
2

+ lim

1
f (x) sin
x cos(nx) dx
2
10

Let
!

1
x
g1 = f (x) cos
2
!
1
g2 = f (x) sin
x
2
Therefore,

lim bn (g1 ) + an (g2 ) = 0

Therefore,

lim

f (x) sin n +

1
x dx = 0
2

Dirichlets Kernel

Definition 11 (Dirichlet kernel).


m
1 X
Dm (t) = +
cos(nt)
2 n=1

is called the Dirichlet kernel of order m.


Theorem 10 (Second representation of Dirichlets kernel). Let m N.
Then, for t 6= 2k, where k Z,
Dm (t) =

1
+ cos(t) + cos(2t) + + cos(mt)
2 



sin
=

m+

2 sin

1
2

1
t
2

Theorem 11. Let


m
X
1
Sm (f, x) = a0 +
an cos(nx) + bn sin(nx)
2
n=1

Then,
1
Sm (f, x) =

m
1X
f (x + t)
cos(nt) dt
2 n=1

11

Proof.
m
X
1
Sm (f, x) = a0 +
an cos(nx) + bn sin(nx)
2
n=1

1
1

f (s) ds

{z

m
X
1

n=1

{z

m
X
1

n=1

1
=

f (s) sin(ns) ds
sin(nx)

an

1
=

f (s) cos(ns) ds
cos(nx)

a0

{z

bn

m
1 X
cos(ns) cos(nx) + sin(ns) sin(nx) ds
f (s) +
2 n=1

m

1 X
cos n(s x) ds
f (s) +
2 n=1

Let
t=sx
dt = ds
Therefore,
1
Sm (f, x) =

As the function is
2-periodic, the limits
can be changed from
x and x to
and .

m
1 X
f (t + x) +
cos(nt) dt
2 n=1

f (t + x)Dm (t) dt

f (t + x)Dm (t) dt


1
=
f (t) Dm (t)

12

Theorem 12 (Dirichlet Theorem). Let f : [, ] R be a piecewise


continuously differentiable function.
Then, x (, ),

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

and for x = or x = ,

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

Exercise 2.
Prove that x [0, 1],
x( x) =

2 X
1

cos(2nx)
2
6
n=1 n

Solution 2.
Let the function be extended naturally to [0, ]. Hence, let the function be
extended evenly to [, ].
Therefore as the function is even, the Fourier series of the function is

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

Therefore,

1
a0 =

f (x) dx

x( x) dx
0
2

13

This theorem is also


valid for [L, L].

1
an =

f (x) cos(nx) dx

x( x) cos(nx) dx
0

x x2 cos(nx) dx

x x2

The integral of cos x


from 0 to is zero, i.e.
if the limits are and
0, the function sin x is
zero.

The integral of cos x

2
=

from 0 to is zero.

x2

cos(nx) dx

( 2x)

0
>


sin(nx)
 ( 2x) sin(nx)
n
n


!

cos(nx) dx dx

0





dx



0

!

dx

sin(nx)
n
0
!


cos(nx)
2 cos(nx)

( 2x)
+
dx

n2
n2

( 2x)

2
cos(nx)
= ( 2x)

n2 0

2 
= 2 (1)n+1 1
n
Therefore,

an =
0

n2

;
;

n = 2k
n = 2k + 1

Therefore,
x( x) =

2 X
1

cos(2k)
2
6
k=1 n

Theorem 13. Let f [, ] R be continuous and f () = f (). Let f 0 (x)


be piecewise continuous. Then the Fourier series converges absolutely to some
limit and uniformly to f (x).
14

Relation between Fourier Coefficients of f (x)


and f 0(x)

Theorem 14. Let the Fourier coefficients of f (x) be a0 , an , and bn . Then,


the Fourier coefficients of f 0 (x) are
0 = 0
n = nbn
n = nan
Proof. Assuming f 0 (x) is integrable,

X
1
n cos(nx) + n sin(nx)
f 0 (x) 0 +
2
n=1

Therefore,
1
0 =

f 0 (x) dx

f () f ()
=

=0

1
n =

f 0 (x) cos(nx) dx


n
1
= f (x) cos(nx) +

f (x) sin(nx) dx

Therefore,
n = nbn
n = nan

Theorem 15. Let the complex Fourier coefficient of f (x) be cn . Then, the
complex Fourier coefficient of f 0 (x) is
n = incn
15

On a general interval,
this theorem translates
to term-by-term
differentiation, i.e.,
the order of
summation and
differentiation can be
changed.

Theorem 16. Let f (x) : [, ] R be a continuous function such that


f () = f (), and let f 0 (x) be piecewise continuous.
If

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

cn einx
=
n=

then,
f 0 (x)
=

nbn cos(nx) nan sin(nx)

n=1

incn einx

n=
This theorem also
holds for a general
interval [L, L].

Theorem 17. Let f : [, ] R be a continuous function that maintains


f () = f () and let f 0 (x) be piecewise continuous. Then, Sm (f, x) converges
uniformly to f (x).
Definition 12 (Inner product). Let x and y be vectors. Their inner product
is defined to be
hx, yi =

xi y i

i=1

Theorem 18.


hx, yi

hx, xi hy, yi

Theorem 19. Let f (x) be continuous on [, ] with piecewise continuous


f 0 (x). Let Sm (f, x) converge uniformly to f (x). Then, the Fourier series is
term-by-term differentiable.

16