Fourier Analysis

Lectures 1-4
Fourier series
/ expansion
=
spectral analysis

·
light -
black
body
-
radiation
radiation

X-ray T H, 0
-

infrared -

: ↓ ↓

· electrical noise

· mechanical noise (sound (

=> approximation scheme

periodic functions >

- #ourier series

a
periodic functions >
- Fourier transform

· linear differential equations

· particl Il Il
 Fourier Series
sines/cosines
~

approximating by
Periodic
-
fanation : ~ a .
c
.
current

&
eg . f(x) is
p-periodic if

f(x + p) =
f()

p-periodic vop-periodic

f(x y.p)
+ =

f(x + m -

1) .

p) =...
=

f()

not periodic periodic

X X2
,
,
...

- in() , cos(), = 27
p
...

5 + 3 =
x -

[x]
=
1
tractional part p

(xj2 =

(x -

[x))2
 trigonometric series

function series :

& + & cos(x) + b sin(x) + a cos(2x) +

, , =

bzsin(x) + -- -

I
a
+ E
k =
1
& a
v
cos(k .
+1 + b
=
pin(k -

x)
- -

Fourier coefficients
Y

·
fn(x) =
a. + [(acos(k
k = 1
K
-
+) + b
,
oin(
x)
27 -

periodic functious

Claim
-
:
virtually any 23-periodic

function of interest can be represented as

()
f(x) = a
+ &
a
v
cos(k .
+1 + b
=
pin(k -

x)

Euler formulas
7 T

a =
-
(85)ax
D
I
a

= (axfa)201(2x)
-
T

0 =
Y fx/sin(x)
-
7
example

S
:
1
x1T
f(x)
o
=

1
-

-
n = x = o

it

do a
Jex ge,
I E

(d x)
d 0
f()cos(4
=
a +
.

D
M

S S
I 2
b =
dxsin(r x) -

sgn(x)
= -

in(y x)dx
i
.

- .

n
↑
-
i

11 (1
2

100(4))
y
- -
cos( + = =

in
-

S
D n even

4
- Y o dd
D .
U

4
win(x)
I

D
I I
2in() + + ...

- T

S
 Proof :
use integral identifies

I
m Y

Ju
,

dxcos( .

x) Cos(M-X) = M .

z(90]
m
,
c

T X
del ta
-

-
Krouecher
III

Cn
=
-if n =
m
m

otherwise
,

use addition rule

coo (2) cos() =

+ (200 (1 + b) + cos( -

P))
M

Cum =
-
2
S S cos(n +

LS
m) .
x) +
cospu -m)
Il
-

+13
-
T

1
if eifn m
=
u + m = 0

on
by possible
if n = m = 0

Coo 24 Con 0 ; &n

= D .

Jy
;
= =
m
m ,
,

n
,
m > 1

&
S dxsin(yx)
-
.
a(x)
T
I O

-
T
odd evh

&

S & X sinzex) sin [x) =

M .

Ju ,
m

-
i
Now assume (*
)

>
= *

I d x
f(x) =

-
T

(a +
(a .
+
a cos(x)

↓
+ b
,
sin()

ti
+
az cos(2x)

↓
+ b sin(2x)

↓
+ ...

S
-
4 O
⑤ * ⑤

#↓ f(x)dx
a 27 = ~
as
= =
..

(axgx) -
ein(x) =

-
T
T

(a (a +
.

↓
+
an cook)

ti
+ b
,
sin() + a
=
cos(2x)

↓
+
...

)· sin(x)

-
4 O
* ⑤

= b
, ( -in " ()dx = b
,
T = b
.
=
y(dxfz) .
winx)

-
T

(dx f() .
cos(x) = a
..
4 = ) a
,
=

y(d +
f()-cos()
- D

pa B
The m :

↑
& () piecewise continuous , 24-periodic ,

- girite ad all other points

t
Fourier f() except
seriesConure
o

=>
is

for discontinuity

·
general proof is difficult
' "
·
for 5 , 8 and
g continuous >
-

is to
uniform convergence easy prove

Exter si to 24 fx)
((x
+ =
on

trivial : 2002nx) +
cos(x .
n .

4)

=> f() =
a. +
(anc( .

Ex) +
bysin(z x) .

a
=

2 (1 +
yes
-
L

an =
= (fx) -

c) x)dx .

-
L

I
b
n
=

=)fx) ein( x)dx -

.

-
L
 ·
Even and odd
functions

f() even =
by = 0 ! (only cos (

odd
f( , = a =
&
n
Conly sin)

⑤ Fourier
coefficients of Jetfa (both of

same period) are SUMs

of coefficients
Fourier
Fourier
coeffs of <f are d .

ceffs of f
Challenges !
-

· Statement : Fourier coefficients give

best trigonometric approx
.

consider
8() some 27-periodic function
approximate

F(x) =
A.
+
An cox) +
B since
deviation T

1 F -

y( =
Jax (fx) -

F(x)))
-
T

minimal An By >
-

bu
if
+
an
&
A Parserul
T

Saxges =
+(n .

a +
a +
y +.... )
-
T

· "Scalar product
"

& a +
f()
.

gx) =

x(zaj ,
+
[19na +
b)
>
↓
0
T
-
↓

ak
,
bu , x