Chapter 1

Fourier Series
We will consider complex valued periodic functions with period 2. We can
view them as functions dened on the circumference S of the unit circle
in the complex plane or equivalently as function f dened on [, ] with
f() = f(). The Fourier Coecients of the function f are dened by
a
n
=
1
2

f(x)e
inx
dx (1.1)
and formally
f(x)

a
n
e
inx
(1.2)
If we assume that f L
1
[0, 2] then clearly a
n
is well dened and
|a
n
|
1
2

|f(x)|dx
It is not clear that the sum on right hand side of equation 1.2 converges
and even if it does it is not clear that it is actually equal to the the function
f(x). It is relatively easy to nd conditions on f() so that the sum in 1.2 is
convergent. If f(x) is assumed to be k times continuously dierentiable on
S, integrating by parts k times one gets, for n = 0,
|a
n
|
1
n
k
sup
x
|f
{k}
(x)| (1.3)
From the estimate 1.3 it is easily seen that the sum is convergent if f is twice
continuosly dierentiable.
Another important but elementary fact is
1
2 CHAPTER 1. FOURIER SERIES
Theorem 1 (Riemann-Lebesgue). For every f L
1
[, ],
lim
n
a
n
= 0 (1.4)
Let us dene the partial sums
s
N
(f, x) =

|n|N
a
n
e
inx
(1.5)
and the Fejer sum
S
N
(f, x) =
1
N + 1

0nN
s
n
(f, x) (1.6)
We can calculate
s
n
(f, x) =
1
2

|j|n
e
ijx

e
ijy
f(y)dy
=
1
2

f(y)[

|j|n
e
ij(xy)
]dy
=
1
2

f(y)
e
in(xy)
(e
i(2n+1)(xy)
1)
e
i(xy)
1
dy
=

2
0
f(y)k
n
(x y)dy (1.7)
where
k
n
(z) =
1
2
e
inz
(e
i(2n+1)z
1)
e
iz
1
=
1
2
sin(n +
1
2
)z
sin
z
2
(1.8)
A similar calculation reveals
S
N
(f, x) =

f(y)K
N
(x y)dy (1.9)
where
K
N
(z) =
1
2
1
(N + 1)
1
sin
z
2

0nN
[e
i(n+1)z
e
inz
]
=
1
2
1
(N + 1)
1
sin
z
2
1
e
iz
1
[(e
iz
e
iNz
)(e
i(N+1)z
1)]
=
1

1
(N + 1)

sin(N +
1
2
)z
sin
z
2

2
(1.10)
3
Notice that for every N,

k
N
(z)dz =

K
N
(z)dz = 1 (1.11)
The following observations are now easy to make.
1. Nonnegativity.
K
N
(z) 0
2. For any > 0,
lim
N
sup
|z|
K
N
(z) = 0
3. Therefore
lim
N

|z|
K
N
(z)dz = 0
It is now an easy exercise to prove
Theorem 2. For any f that is bounded and continuous on S
lim
N
sup
xS
|S
N
(f, x) f(x)| = 0
Theorem 3. For any f L
p
[, ]
S
N
(f, )
p
f
p
and therefore for 1 p < ,
lim
N
S
N
(f, ) f()
p
= 0
The behavior of s
N
(f, x) is more complicated. However it is easy enough
to see that
Theorem 4. For f C
2
(S),
lim
N
sup
x
|s
N
(f, x) f(x)| = 0
4 CHAPTER 1. FOURIER SERIES
The series converges and so s
N
(f, ) has a uniform limit g. S
N
(f, ) has
the same limit, but has just been shown to converge to f. Therefore f = g.
The following Theorem is fairly easy.
Theorem 5. If f is a function in C

(S) i.e Holder continuous with some

exponent > 0 then
lim
N
s
N
(f, x) = f(x)
Proof. We can assume thatwith out loss of generality that x = 0 and let
f(0) = a. We need to show that
lim
N
1
2

f(y)
sin(N +
1
2
)y
sin
y
2
dy = a (1.12)
Because
f(y)a
sin
y
2
is integrable, 1.12 is consequence of the Riemann-Lebesgue
Theorem, i.e. Theorem 1.
If f is a bounded function, then one can replace sin
y
2
by
y
2
and the problem
reduces to calculating
lim

f(y)
siny
y
dy
Let us now assume that f is a function of bounded variation on S which has
left and right limits a
l
and a
r
at 0. By a change of variables one can reduce
the above to calculating
lim

f(
y

)
sin y
y
dy
If we denote by
G(y) =

y
sin x
x
dx
then
a
r
() =
1

0
f(
y

)
sin y
y
dy =
1

0
f(
y

)dG(y)
=
1
2
a
r
+
1

0
G(y)df(
y

) =
1
2
a
r
+
1

0
G(y)df(y)

1
2
a
r
by the bounded convergence theorem. This establishes the following
5
Theorem 6. If f is of bounded variation on S
lim
N
s
N
(f, x) =
f(x + 0) +f(x 0)
2
The behavior of s
N
(f, x) for f in L
p
[, ] for 1 p < is more
complex. Let us dene the linear operator
(T

f)(x) =

f(x + y)
sin y
sin
y
2
dy (1.13)
on smooth functions f. If s
N
(f, x) were to converge uniformly to f for every
bounded continuous function it would follow by the uniform boundedness
principle that
sup
x
|(T

f)(x)| C sup
x
|f(x)|
with a constant independent of , atleast for = N +
1
2
for positive integers
N. Let us show that this is false. The best possible bound C = C

is seen
to be
C

=
1
2

| sin y|
| sin
y
2
|
dy
and because
|
1
sin
y
2

2
y
|
is integrable, C

diers from
1
2

| sin y|
|y|
dy =
1
2

| siny|
|y|
dy
by a uniformly bounded amount. The divergence of
1
2

| sin y|
|y|
dy
implies that C

as . By duality theis means that T

f is not
uniformly bounded as an operator from L
1
[, ] into itself either. Again
from uniform boudedness principle one cannot expect that s
N
(f, ) tends to
6 CHAPTER 1. FOURIER SERIES
f() in L
1
[, ] for evrery f L
1
[, ]. However we will prove that for
1 < p < , for f L
p
[, ]
lim

f f
p
= 0
By standard arguments involving the approximation of an L
p
function by a
continuous function it is sucient to prove a uniform bound of the form
T

f
p
C
p
f
p
with a constant C
p
depending only on p for smooth functions f and 1
First we establtsh what is known as weak type inequality.
Theorem 7. There is a constant C such that for all 1 and smooth f
mes{x : |(T

f)| }
C

f
1