# 周期三角波傅里叶级数例题_傅里叶分析总结（1）：周期函数的傅里叶级数

## Definition

Identify

with
.

The Fourier coefficients of an integrable periodic function

are given by

The sequence of Fourier coefficients

is called the Fourier transform of
.

The

th-partial Fourier sum is given by

## Theorem

If

and
for all
, then
a.e.

Proof:

If

is continuous and
, then
on some closed interval
. Construct a sequence of trigonometric polynomials
such that
on
but
on
, then
. Thus,
.

If

, apply the previous paragraph to the absolutely continuous function
. Note that
and
a.e.

A much shorter proof would be using Stone-Weierstrass theorem. Note that the linear span of trigonometric polynomials forms a subalgebra that separates points and contains constants. Also,

is dense in
by Lusin theorem.

## Riemann-Lebesgue Lemma

If

, then
, and the convergence is uniform on compact sets, i.e. if
is compact, then
.

Proof:

If

is continuous, use
and uniform continuity. If
, approximate
by continuous functions in
-norm.

## Corollary

Let

be the space of (two-sided) sequences vanishing at infinity. The Fourier transform
is a linear operator from
to
.

## Lemma

Let

.
1. For almost every
, the function
is
-integrable.
2. The function
is
-integrable.

Proof: Fubini Theorem.

## Definition

The convolution on circle is the bilinear operator

defined by

for

## Corollary

The convolution has the following properties:

The

-norm on
is given by

for all
.

## Young's Convolution Inequality

1. If
and
, then
is continuous.
2. If
and
, then
and
.

## Example

1. The partial Fourier series can be written as
, where
is the Dirichlet kernel.
2. The Cesàro mean
can be written as
, where
is the Fejér kernel.
3. For
, the Abel mean
can be written as
, where
is the Poisson kernel.

## Definition

The sine integral function

is defined by

for all

## Lemma

The sine integral function has the following properties:

1. .
2. attains its global maximum at
.

For any constant

, we have

## Dini Test

If there exist

and
such that

, then

Proof:

Use Riemann-Lebesgue lemma repeatedly. Note that the function

is bounded and continuous on

.

## Corollary

If

is of bounded variation on a neighbourhood of
, then

In particular, if

is piecewise smooth on
, then

for all
.

## Corollary

converges uniformly to
if
satisfies any one of the following conditions:
• is of bounded variation and continuous, or
• is
-Hölder continuous of order
.

## Riemann Localization Principle

If

is identically zero on an open interval
, then
converges uniformly to zero on any closed interval
.

## Lemma

converges absolutely if and only if
. If
and
, then
converges uniformly to a continuous function
and
a.e.

Proof: Weierstrass M-test.

## Lemma

If

are absolutely continuous , then

Proof:

Use integration by parts repeatedly. Note that

by absolute continuity, so
is bounded.

## Hausdorff-Young Inequality

If

and
, then
and
.

Proof: Riesz-Thorin interpolation.

## Theorem

converges absolutely and uniformly to
if
satisfies any one of the following conditions:
• is absolutely continuous and
for some
, or
• is of bounded variation and
-Hölder continuous of order
, or
• is
-Hölder continuous of order
(
Bernstein's theorem).

## Definition

Let

be an directed set. An approximate identity on circle indexed by
is a family of integrable functions
that satisfies the following:
1. , and
2. there exists
such that for all
,
, and
3. for all
,
.

## Lemma

Let

be an approximate identity on circle. Suppose
and
exist at some
. If any one of the following conditions is satisfied:
• , or
• and for all
,
,

then

.

## Theorem

Let

be an approximate identity on circle.
1. If
is continuous, then
converges uniformly to
.
2. If
and
, then
converges in
to
.
3. If
and
, then
in weak* topology, i.e.
.

## Example

1. The Fejér kernel
and Poisson kernel
are approximate identities.
2. The Dirichlet kernel
is NOT an approximate identity, since
.

## Remark

The Fejér kernel gives another proof that

is uniquely determined by its Fourier coefficients. On the one hand,
converges in
to
since
is an approximate identity. On the other hand,
is uniquely determined by
.

## Definition

The

-inner product on
is given by

for all
.

Note that

.

## Theorem

The Fourier transform

is a Hilbert space isomorphism
. In particular, we have the
Parseval's identity:

for all
.

Proof:

Since

a.e. whenever
for all
,
is a complete orthonormal basis for the separable Hilbert space
. The rest is just standard Hilbert space theory.

## Theorem

Let

and
.
1. converges in
to
(
Bochner-Riesz theorem).
2. converges pointwise a.e. to
(
Carleson's theorem).

## Lemma

The Fourier partial sum, considered as an operator either on continuous functions

or on integrable functions
, is a bounded linear operator with operator norm
.

## Theorem

1. For any
, the set
is a dense
-subset of
.
2. The set
is a dense
-subset of
.
3. There exists
such that
diverges pointwise everywhere.

Proof:

The first two parts are proved by applying the uniform boundedness principle to the unbounded sequence of operators

. The third part is an example constructed by Kolmogorov.

10-22

02-04 4205
04-08 228
12-04 2586