8.3 .

some useful applications

8.3 Some useful applications

Lecture 23
Here we collect some random collection of useful facts that can be immediately 11th December 2024
derived from the Fourier series.

Delta function One can show that the delta function can be written as the
Fourier transformation. In the complex Fourier case with c = 1, einx

X 8 fn
(x x0 ) = ⇤ 0
n (x) n (x ) , (763)
'
- Ifse
n _
"
fr
+
L, 2 w: =
I =
.

efx
where n (x) = p1 e ⇡inx is the normalized Fourier basis.
2 InILy
-

=
je
Evaluation of infinite series Consider the expansion of the above function
f (x) in terms of the Fourier series:
1
X 4
f (x) = sin [(2n + 1)⇡x] . (764)
(2n + 1)⇡
n=0

From the energy theorem, we expect

.
1
X 16
antan
Sf .
f) =

. hf , f i =
n=0
(2n + 1)2 ⇡2
. (765)

But we know from the exact expression of f that hf , f i =2

1. This immediately
gives the identity
X1
1 ⇡2
= (766)
n2 8
n=odd
l :
-

-
1 /1 "

i /
1
l

143
 8. fourier analysis

fnein
"

8.4 Fourier transformation

fln =

hea
flx) In many physical situations, the function f (x) is not periodic. The simplest
way to think about such a situation is to start from (743) and (746) and let ' IS
the range c ! 1. Note that, if we take c to be very large, the exponent ⇡nx/c c

becomes almost continuous. Therefore, in the strict large c limit, we can

replace the variable
sJak
\

℃
C
⇡nx/c ! k , (767)

where k is a continuous real number. Then the discrete sum over n becomes x n
an integral over k: Z K = επ .

1 X 1
! dk . (768)
2c n 2⇡

if xalss dk
-
an .
The Fourier series (743) is now
Z dn
=

f (x) = fˆ(k)e ikx dk , (769)

where we introduced a new continuous function fˆ(k) which plays a role of the
coefficient an . This expression is called the Fourier transformation of f (x). The
coefficient can be computed by (746)
Z1
1
fˆ(k) = f (x)e ikx
dx . (770)
2⇡
1

* This is called the inverse Fourier transformation or simply the Fourier transfor-
mation of fˆ(k). There are various d
conventions for how
It is straightforward to check that the Fourier transformation twice brings tor of 2⇡ are distrib
back to the original function: tween f and fˆ.

Z1 Z1
1
fˆ(k) = fˆ(k 0 )e ik x dk 0
0
ikx
dx e
2⇡
1 1

Z1 Z1
1
fˆ(k 0 )e i(k
0
= dx k)x
dk 0 (771)
2⇡
1 1
. 8 cS - k )
= fˆ(k) .

From the second to the third line, we integrated x first to produce a delta
function (k k 0 ).

Example Let us consider the function defined on a real line

144
 → :
n a :
8.5 . differential equations and fourier transformation

(
b |x| < a
f( s ) = 2

faldx
π
- beiky
f (x) = (772)
0 |x| > a

where b > 0 is a real number. Calculating its Fourier transformation, we obtain

Za
1 b b sin ka
fˆ(k) = f (x)e ikx
dx = e ikx
dx = . (773)
2⇡ 2⇡ ⇡ k
a

r
The plot of this function is depicted below equation (761). Suppose that a is

", ,
extremely large, so that f (x) approaches to a constant function. Then
b sin ka
lim fˆ(k) = lim = b (k) . (774)
a!1 ⇡ a!1 k

In other words, the Fourier transformation of a constant function is a delta

function, and the Fourier transformation of a delta function is a constant
function.
Note that, for finite a, the width of fˆ(k), which is defined as a distant
between a closest pair of zeros, is 2⇡/a. We can write
2⇡
k= , x = 2a . (775)
a

In fact, one can prove that, for any function f (x), we have

x k 1. (776)
?
nomber wly
Wave
□
.

In quantum mechanics, the momentum of the particle p is related to the

5k variableσ
f P k by ~k, where ~ = h/(2⇡) with h the Planck constant. Then the above
-

equation implies
x p ~, (777)

which is nothing but the Heisenberg uncertainty principle. One can chects that
the bound is saturated by the Gaussian distribution. HW - x212
a

fne
8.5 Di↵erential equations and Fourier transformation -kraz
8
e

' ks en
In physics and engineering, we are often interested in solving di↵erential
equations whose coefficients are all constants. If this is the case, the following
oxok 2 1 .
technique that involves Fourier transformation is one of the most widely used
tool.

Derivatives and Fourier transformation Consider the first order di↵eren-

tiation of f (x):

145
[x , P ] = τ .

λ
,
*
] =- (
(× 1 -* ] a :

= π
xx
x →
≥

"
K 8. fourier analysis

!
+ kz Z
d
f (x) = (ik)fˆ(k)e ikx dk . (778)
dx

The second order di↵erentiation is

Z
d2
f (x) = ( k 2 )fˆ(k)e ikx dk . (779)
dx2

Now let us consider any second order linear ODE whose coefficients are all
constant:
d2y dy
c2 2 + c1 + c0 y = f (x) . (780)
Cran '
s funcilon dx dx

Plugging in Z
af =
δ y(x) = dk ŷ(k)e ikx , (781)

δ
0 =
and its derivatives, we obtain
Z ⇣ ⌘ Z
(d 72 )
dk ŷ(k) c2 k + ic1 k + c0 e = dk fˆ(k)e ikx .
2 ikx
(782)

Therefore,
fˆ(k)
ŷ(k) = . (783)
c0 + ic1 k c2 k 2

Fourier transforming back,

Z
1 fˆ(k)e ikx
y(x) = dk . (784)
2⇡ c0 + ic1 k c2 k 2

Such an integral is not always easy to solve, but sometimes it can be evaluated
by the method of complex analysis. For example, the case with c1 = 0 is
precisely the type of integral we considered in section 6.6.3, which makes use
of the concept of the Cauchy principal value.

ce
,
1
fak
pole .
( lC ( C , ) ( C + S .
)

∞X×

146