Escolar Documentos
Profissional Documentos
Cultura Documentos
Himanshu Tyagi
Let f be an absolutely integrable function on T := [, ], i.e., f
L1 (T). For n = 1, 2, . . . define
Z
1
f ()ein d.
f (n) =
2 T
The series
n=
The FS of f does not converge to f point-wise a.e.. If we restrict to continuous functions, the point-wise convergence of FS can not be guaranteed. In
this lecture, we define weaker but important criteria for convergence under
which the FS of continuous f does converge to f .
A. The definition of convolution given in Lecture 1 can be easily modified
for functions in L1 (T). For f, g L1 (T) the convolution f g is defined as
Z
f (x y)g(y) dy,
f g(x) =
where f (x) = f (x mod T). Also, Theorem 1 of Lecture 1 and property (v)
of convolution hold. Therefore, we can mimic the proof therein to show there
is no identity for convolution in L1 (T), provided we prove the RiemannLebesgue Lemma. Here we give an alternate proof of Riemann-Lebesgue
Lemma using approximate identity.
Definition 1. A sequence {Kn } of continuous functions on R is a summability
R kernel if for all n N
(i) RR Kn (x) dx = 1,
(ii) R |Kn (x)| dx M ,
Z
(iii) For each 0 < < , lim
n |x|>
|Kn (x)| dx = 0.
This lecture is taken from the lecture notes of the course Fourier Series organized by
Prof. Rajendra Bhatia at the Indian Statistical Institute, New Delhi, India. It was also
published as: R. Bhatia, Fourier series(Second Edition), Hindustan Book Agency (India),
2003.
Note here that limits of integral in property three represents the union of
regions [, ) and (, ]. As was shown in the last lecture, a summability
kernel is a bounded approximate identity for L1 (T). Also, the Fejer kernel
!2
n
n+1
X
sin
t
1
|j|
1
2
Kn (t) =
1
eijt =
is a summa1
2
n+1
2(n + 1)
sin
2t
j=n
bility kernel. Therefore we have the following lemma.
Lemma 1. For f L1 (T) and n = 1, 2, . . ., n (f ) := Kn f . Then
n (f ) f in L1 (T).
Note that n (f ) is a function of form
N
X
n=N
|n|
Proof: Let > 0. Choose > 0 such that if |t| , then kf f (t) kL <
(/2M ). This can be done since f is continuous and T is compact imply f
is uniformly continuous. Then for t in T we have
Z
Z
Kn (x) dx = 1]
|Kn (y)| |f (t y) f (t)| dy
[Using
|Kn f (t) f (t)|
R
R
Z
Z
|Kn (y)| |f (t y) f (t)| dy
|Kn (y)| |f (t y) f (t)| dy +
=
|y|>
|y|
Z
+ 2kf kL
|Kn (y)| dy
2
|y|>
,
X
r n xn converges to a limit L(r).
(i) For every 0 < r < 1, the series
n=1
define sN =
N
X
xn and n =
s1 +...+sn
.
n
xn is said to be
n=1
Ces
aro summable to L [also called (C, 1)-summable to L] if the sequence n
converges to L as n .
We state the next result without proof.
X
Theorem 4. A series
xn with complex terms xn for n = 1, 2, . . . conn
verges to L implies
aro summable to L.
X it is Ces
Furthermore,
xn is Ces
aro summable to L implies it is Abel summable
to L.
N
X
n=N
f(n)ein ,
DN (t) is called the Dirichlet kernel. If the Dirichlet kernel was a summability
kernel we could have shown the point-wise convergence of FS of f to f .
However this is not true. To show that FS of f is Ces
aro summable to f ,
N
X
we need to show that N 1+1
sn (f ; ) converges to f (). Note that
n=0
N
N
1 X
1 X
sn (f ; ) = f
Dn ().
N + 1 n=0
N + 1 n=0
1
N +1
N
X
n=0
N
N X
n
X
1 X
1
Dn (t) =
eikt
N +1
2(N + 1)
n=0
n=0 k=n
n
N
X
X
1
[1 + 2
cos kt].
=
2(N + 1) n=0
k=1
n
X
k=1
1+2
n
X
cos kt =
k=1
X
1
cos kt sin t/2]
[sin t/2 + 2
sin t/2
k=1
sin(n + 1/2)t
.
sin t/2
Therefore,
N
N
X
1
sin(n + 1/2)t
1 X
Dn (t) =
N +1
2(N + 1)
sin t/2
n=0
n=0