Analysis results

May 30, 2006

Contents
1 Measure Theory 1
1.1 Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Signed measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Linear functionals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Measure and Outer Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Functional Analysis 4
2.1 Point-set topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Hahn-Banach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Big theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Harmonic Analysis on T and R 6
3.1 Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Coecient growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1 Measure Theory
1.1 Approximation
Proposition 1. Let f be a measurable function on an interval [a, b], and assume f is nite almost
everywhere. Given > 0, there exist a simple function g and a continuous function h such that
[f g[ < and [f h[ < except on a set of measure less than .
Theorem 2 (Egoro). Let f
n
be a sequence of measurable functions on a measure space (X, ).
If f
n
f almost everywhere on a measurable set E with (E) < , then for any > 0 there is a
measurable set A with (A) < such that f
n
f uniformly on E A.
Theorem 3 (Lusin). Denote Lebesgue measure by , and suppose f is a measurable function on
[a, b]. Then for any > 0, there is C([a, b]) and a -measurable set A such that (A) < and
f = on [a, b] A.
1
Theorem 4 (Jensens inequality). If is a convex function on (, ) and f L
1
([0, 1], ),
then
_
[0,1]
(f(t)) dt
_
[0,1]
f(t) dt .
1.2 Integration
Fix a measure space (X, B, ).
Theorem 5 (Fatous Lemma). Let f
n
be a sequence of non-negative measurable functions. Then
for any E B,
_
E
lim
n
f
n
d liminf
n
_
E
f
n
.
Theorem 6 (Monotone Convergence). Let f
n
be a sequence of non-negative measurable func-
tions converging almost everywhere to f such that f
n
f. Then,
_
X
f
n
d
_
X
f d.
Corollary 7 (Beppo Levi). Let f
n
be a sequence of non-negative measurable functions. Then,
_
X

n=1
f
n
d =

n=1
_
X
f
n
d.
Theorem 8 (Lebesgue Dominated Convergence). Let g be integrable over E B, and suppose
that [f
n
[ g for a sequence of measurable functions f
n
converging almost everywhere to f. Then,
_
E
f
n
d
_
E
f d.
1.3 Signed measures
Theorem 9 (Hahn Decomposition). Let be a signed measure on the measurable space (X, B).
Then, X is the disjoint union of a positive and a negative set.
Theorem 10 (Jordan Decomposition). Let be a signed measure on the measurable space
(X, B). Then there is a unique pair of mutually singular measures
+
and

such that =
+

.
Theorem 11 (Radon-Nikodym). Let (X, B, ) be a -nite measure space and . Then
there is a non-negative measurable function f, unique almost everywhere with respect to , such
that for all E B,
(E) =
_
E
f d.
Theorem 12 (Lebesgue Decomposition). Let (X, B, ) be a -nite measure space and a
-nite measure on B. Then there exist unique measures
1
and
2
such that =
1
+
2
.
2
1.4 Linear functionals
Theorem 13 (Riesz Representation). Suppose that (X, B, ) is a complete measure space. Let
1/p + 1/q = 1 for 1 p < ; if p = 1, assume that is -nite. Then there is an isometric
isomorphism L
p
()

= L
q
().
Theorem 14 (Riesz-Markov). Let X be a locally compact Hausdor space. Then every bounded
linear functional on C
0
(X) is induced by a regular Borel measure on X.
Theorem 15 (Riesz Representation). Let X be a locally compact Hausdor space. Then there
is an isometric isomorphism between C
c
(X)

and the regular Borel measures on X.

1.5 Measure and Outer Measure
Theorem 16 (Caratheodory Extension). Let be a measure on an algebra / and denote by
the restriction of the induced outer measure

to the set B of

-measurable sets. Then, if is

nite or -nite, so is , in which case is the unique extension of to the -algebra generated
by /.
Theorem 17 (Fubini). Let (X, /, ) and (Y, B, ) be two complete measure spaces and f
L
1
(X Y, ). Then,
(a) for almost all x, the function f
x
(y) = f(x, y) is in L
1
(Y, );
(b)
_
Y
f(x, y) d(y) is in L
1
(X, ); and
(c)
_
X
_
Y
f d d =
_
XY
f d( ).
Theorem 18 (Tonelli). Let (X, /, ) and (Y, B, ) be two -nite measure spaces and suppose f
is a non-negative measurable function on X Y . Then,
(a) for almost all x, the function f
x
(y) = f(x, y) is a measurable function on Y ;
(b)
_
Y
f(x, y) d(y) is a measurable function on X; and
(c)
_
X
_
Y
f d d =
_
XY
f d( ).
Denition 19. For > 0, dene the Hausdor measure of dimension on a metric space X to
be the Borel measure m

induced from the outer measure

m

(E) = lim
0

()

(E) ,
where

()

= inf

i
r

i
with the inmum taken over open covers of E by balls of radii r
i
< .
The Hausdor dimension of a Borel set E is
dim(E) = inf : m

(E) > = sup : m

(E) = 0 .
3
2 Functional Analysis
2.1 Point-set topology
Theorem 20 (Baire Category). If X is a complete metric space, then every non-empty open
subset of X is of second category.
Theorem 21 (Ascoli-Arzela). Let T be an equicontinuous family of functions from a separable
space X to a metric space Y . If f
n
is a sequence in T such that for each x X, the set f
n
(x) is
precompact, then there is a subsequence converging pointwise to a continuous function; moreover,
convergence is uniform on compact subsets.
Theorem 22 (Stone-Weierstra). Let X be a compact space. If / a unital subalgebra of C(X, R)
separating points, then / is dense in C(X, R); if / is a unital subalgebra of C(X, C) separating
points and closed under conjugation, then / is dense in C(X, C).
Theorem 23 (Uniform Boundedness Principle). Suppose X is a Banach space, Y is a normed
linear space and T is a family of bounded linear operators T : X Y. If for each x X there exist
M
x
such that |T(x)|
Y
M
x
for all T T, then T is uniformly bounded.
Theorem 24 (Urysohn Separation). Let A and B be disjoint closed subsets of a normal topo-
logical space X. Then there is a continuous function f : X [0, 1] such that f[
A
0 and f[
B
1.
Theorem 25 (Tietze Extension). Let A be a closed subset of a normal topological space X. Any
continuous function A R extends to a continuous function X R.
Theorem 26 (Urysohn Metrization). Every second countable normal topological space is metriz-
able.
2.2 Hahn-Banach
Theorem 27 (Extension Algebraic). Let W be a subspace of a real vector space V , and
suppose
0
is a linear functional on W and p is a sub-linear functional on V . If
0
p on W, then

0
extends to a linear functional p on V .
Corollary 28. Let W be a subspace of a real or complex vector space V , and suppose that
0
is a
linear functional on W and p is a semi-norm on V . If [
0
[ p on W, then
0
extends to a linear
functional on V such that [[ p.
Theorem 29 (Extension Locally convex). Let W be a subspace of a locally convex space
V , and suppose that
0
W

. Then
0
extends to V

.
Corollary 30. Suppose that V is a nite-dimensional real or complex vector space. Then there is
a unique locally convex topology on V .
Theorem 31 (Extension Normed). Let X
0
be a subspace of a normed space X, and suppose

0
X

0
. Then
0
extends to X

such that |
0
|
X

0
= ||
X
.
Corollary 32. If x
0
,= 0 is in a normed space X, then there is X

such that || = 1 and

(x
0
) = |x
0
|.
4
Corollary 33. If Y is a closed subspace of a normed space X and x
0
X Y, there is X

vanishing on Y such that || = 1 and (x

0
) = dist(x
0
, Y).
Theorem 34 (Separation Algebraic). Let Y and Z be two non-empty disjoint convex sets
in a real or complex vector space X, and suppose that one of them has an internal point. There
there is a linear functional f separating Y and Z.
Theorem 35 (Separation Topological). Let Y and Z be two non-empty disjoint convex sets
in a real or complex topological vector space X, and suppose that one of them is open. Then there
is a continuous linear functional X

separating Y and Z.
Theorem 36 (Separation Locally convex). Let Y and Z be two non-empty disjoint closed
convex sets of a locally convex space V , and suppose that one of them is compact. Then there is
a continuous linear functional X

separating Y and Z with the closures of '(Y ) and '(Z)

disjoint.
Corollary 37. If x ,= 0 is in a locally convex space V , there is V

such that (x) ,= 0.

Corollary 38. If Z is a closed convex subset of a locally convex space V and y V Z, there is
V

strictly separating V and y.

Corollary 39. If Z is a closed subspace of a locally convex space V and y V Z, there is V

vanishing on Z such that (y) ,= 0.

2.3 Big theorems
Theorem 40 (Open mapping). A continuous surjective linear map between Banach spaces is
an open mapping.
Theorem 41 (Closed graph). If X, Y are Banach spaces and f : X Y is linear, then the graph
of f is closed i f is continuous.
Corollary 42 (Mazzeo). If X, Y are Banach spaces and f : X Y is linear, then f is bounded.
Theorem 43 (Banach-Alaoglu). Let X be a normed linear space. Then, the closed unit ball
B
1
(X

) of X

is weak- compact.
Theorem 44 (Goldstein). Let X be a Banach space. Then B
1
(X) (embedded in X

) is weak-
dense in B
1
(X

).
Corollary 45 (Kakutani). A normed space X is reexive i its closed unit ball B
1
(X) is weakly
compact.
Corollary 46 (Milman-Pettis). A uniformly convex Banach space X is reexive.
Theorem 47 (Krein-Milman). Let X be a compact convex set in a locally convex topological
vector space V . Then, X is the closed convex hull of its extreme points.
5
3 Harmonic Analysis on T and R
3.1 Approximation
Theorem 48. Let f L
1
(T), assume

f(0) = 0, and dene
F(t) =
_
t
0
f() d .
Then F is continuous, 2-periodic, and

F(n) =
1
in

f(n) for n ,= 0 .
Theorem 49. Let f L
1
(R) and dene
F(x) =
_
x

f(y) dy .
Then, if F L
1
(R), then

F() =
1
i

f()
for ,= 0.
Theorem 50. Let f(x), xf(x) L
1
(R). Then,

f is dierentiable and
d
d

f() = (

ixf)() .
Corollary 51 (Riemann-Lebesgue). For f L
1
(T),

f(n) 0 as [n[ .
Corollary 52 (Riemann-Lebesgue). For f L
1
(R),

f() 0 as [[ .
Lemma 53. Let B be a Banach space, a continuous B-valued function on T, and k
n
a summa-
bility kernel. Then,
1
2
_
T
k
n
()() d (0) as n .
Theorem 54. Let B be a homogeneous Banach space on T and let k
n
be a summability kernel.
Then, for every f B, k
n
f f in B as n .
Theorem 55. Let B be a homogeneous Banach space on R and let k

be a summability kernel.
Then, for every f B, k

f f in B as .
Lemma 56 (Parseval). Let
n
be a complete orthonormal system in a Hilbert space H. Let
f, g H. Then,
f, g) =

n
f,
n
)
n
, g) .
Theorem 57 (Parsevals formula). Let B

, where B is a homogeneous Banach space on T

containing the exponentials. Then,
n
() in the weak-star topology. That is, for any f B,
f, ) = lim
N
N

n=N
_
1
[n[
N + 1
_

f(n) (n) .
6
Theorem 58 (Parseval inversion). For f L
1
(R),
1
2
_

_
1
[[

_

f() exp(ix) d f(x)
in L
1
(R) as . If

f L
1
(R), then
f(x) =
1
2
_
R

f() exp(ix) d .
Theorem 59 (Parsevals formula). Let M(R) and let f be a bounded continuous function
in L
1
(R). Then,
1
2
_

_
1
[[

_

f() () d
_
R
f(x) d(x)
as . If

f L
1
(

R), then
_
R
f(x) d(x) =
1
2
_

f() () d .
3.2 Characterization
Theorem 60. Let B be a homogeneous Banach space on T containing the exponentials and
a
n

nZ
C. The following are equivalent:
(a) There exists B

, || C, such that (n) = a

n
for all n.
(b) For all trigonometric polynomials P,

P(n)a
n

C |P|
B
.
Corollary 61. Let B be a homogeneous Banach space on T containing the exponentials. Then,
S

n
a
n
exp(int) is the Fourier series of some B

i the |
N
(S)|
B
are bounded.
Theorem 62. Let be continuous on

R and dene

by

(x) = ( K

)(x) =
1
2
_

_
1
[[

_
() exp(ix) d .
Then, is a Fourier-Stieltjes transform i

L
1
(R) for all > 0 and |

|
1
is bounded as
.
Lemma 63. A series S

n
a
n
exp(int) is the Fourier-Stieltjes series of a positive measure
M(T) i for all (or innitely many) n 0,
n
(S) 0.
Theorem 64. A continuous bounded function on

R is the Fourier-Stieltjes transform of a positive
measure on R i
_

f()() d 0
for every smooth non-negative compactly-supported function f on R.
7
Theorem 65 (Herglotz). A numerical sequence a
n
is positive denite i it is the Fourier-
Stieltjes coecients of a positive measure M(T).
Theorem 66 (Bochner). A function on

R is positive denite and continuous i it is the
Fourier-Stieltjes transform of a positive measure.
Theorem 67 (Spectral theorem for unitary operators). Let U be a unitary operator on a
Hilbert space H. For f H, consider the spectral measure
f
of f; that is,
f
0 and
n
(f) =
U
n
f, f). Then there is an isometric isomorphism
S : H
f

= L
2
(
f
) ,
identifying
U
n
f exp(int) ,
where H
f
denotes the closed span of the U
n
f. Moreover, S conjugates the B(H
f
)-closure of the
polynomials in U to multiplication on L
2
(
f
) by elements of L

(
f
), identifying
U exp(it) .
Corollary 68 (Ergodic theorem). Let U be a unitary operator on a Hilbert space H. Then,
1
N
N1

j=0
U
j
P ,
where P is the orthogonal projector onto the U-invariant subspace of H.
Theorem 69 (Universal multipliers). Let B be a homogeneous Banach space on T and
M(T). There is a unique linear operator B(B), called convolution by , such that ||
B(B)

||
M(T)
and

f(n) = (n)

f(n) for all f B.
Theorem 70. Let M(T). Then for each T,
1
2N + 1
N

n=N
(n) exp(in) () .
Corollary 71 (Wiener). Let M(T) with discrete part

j
a
j

j
. Then,

j
[(
j
)[
2
= (
#
)(0) = lim
N
1
2N + 1
N

n=N
[ (n)[
2
,
which is 0 i is continuous.
Corollary 72 (Wiener). Let M(R) with discrete part

j
a
j

x
j
. Then, for any sequence

n
,

j
[(x
j
)[
2
= lim
n
1
n
_

n
+n

n
[ ()[
2
d ,
which is 0 i is continuous.
8
3.3 Convergence
Denition 73. The Fejer kernel on T is
K
n
(t) =
n

j=n
_
1
[j[
n + 1
_
exp(ijt) =
1
n + 1
_
sin
n+1
2
t
sin
1
2
t
_
2
;
on R, it is K

(x) = xK(x), where

K(t) =
1
2
_
1
1
(1 [[) exp(ix) d .
The Dirichlet kernel on T is
D
n
(t) =
n

j=n
exp(ijt) =
sin(n +
1
2
)t
sin
1
2
t
.
Theorem 74 (Fejer). Let f L
1
(T).
1. If

f(t
0
) = lim
h0
f(t
0
+h) +f(t
0
h)
2
exists, then
n
(f, t)

f(t
0
).
2. If f is continuous on the closed interval I, then
n
(f) f uniformly on I.
3. For all n, inf
t
f(t)
n
(f) sup
t
f(t).
Corollary 75. If f is continuous at t
0
and its Fourier series converges there, then its sum is f(t
0
).
Theorem 76 (Lebesgue). If
1
h
_
h
0

f(t
0
+) +f(t
0
)
2


f(t
0
)

d 0
for some value

f(t
0
), then
n
(f, t
0
)

f(t
0
).
Corollary 77. If the Fourier series of f L
1
(T) converges on a set E of positive measure, its sum
coincides with f almost everywhere on E. In particular, only the trivial Fourier series converges
to zero almost everywhere.
Theorem 78. A homogeneous Banach space B on T admits convergence in norm i |S
n
|
n
is
bounded, where S
n
is the map f S
n
(f).
Theorem 79. Let f L
1
(T) such that

f(n) O(1/[n[). Then, S
n
(f, t) and
n
(f, t) converge for
the same values of t and to the same limit. If
n
(f, t) converges uniformly on some set, then so
does S
n
(f, t).
Theorem 80 (Localization principle). Let f L
1
(T) and assume that f vanishes in an open
interval I. Then S
n
(f) converges to 0 on I and uniformly so on closed subsets of I.
9
3.4 Coecient growth
Theorem 81. Let a
n

nZ
be an even sequence of non-negative numbers tending to zero at innity
that is convex on (0, ). Then there exists a non-negative function f L
1
(T) such that

f(n) = a
n
.
Theorem 82. Let f L
1
(T) and assume that

f([n[) =

f([n[) 0. Then,

n>0
1
n

f(n)
converges.
Theorem 83. If f ((T) is k-times dierentiable, and f
(k1)
is absolutely continuous, then
[

f(n)[ min
0jk
|f
(j)
|
1
[n[
j
.
If g BV(T), then
[ g(n)[
var(g)
2 [n[
.
Theorem 84 (Hausdor-Young). Let 1/p + 1/q = 1 where 1 p 2. Then for f L
p
(T),
|(

f(n))|

q |f|
L
p
(T)
.
Conversely, if (a
n
)
p
, then there exists f L
q
(T) such that

f(n) = a
n
and
|f|
L
q
(T)
|(a
n
)|

p .
References
[1] R. Kadison, J. Ringrose. Fundamentals of the Theory of Operator Algebras, Vol. 1, Academic
Press. (1983)
[2] Y. Katznelson. An Introduction to Harmonic Analysis, Cambridge University Press. (2004)
[3] H. L. Royden. Real Analysis, Prentice Hall. (1988)
10