You are on page 1of 27

Functional Analysis Study Pack

Hector Garcia
September 11, 2014
1
Contents
1 Metric Spaces 4
1.1 Metric Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Examples of Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Open Set, Closed Set, Neighborhood . . . . . . . . . . . . . . . . . . . . . 5
1.4 Convergence, Cauchy Sequence, Completedness . . . . . . . . . . . . . 7
2 Normed Spaces. Banach Spaces. 8
2.1 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Normed Spaces. Banach Spaces. . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Further Properties of Normed Spaces . . . . . . . . . . . . . . . . . . . . 10
2.4 Finite Dimensional Normed Spaces and Subspaces . . . . . . . . . . . . 10
2.5 Compactness and Finite Dimension . . . . . . . . . . . . . . . . . . . . . 10
2.6 Linear Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.7 Bounded and Continuous Linear Operators . . . . . . . . . . . . . . . . 11
2.8 Linear Functionals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.9 Linear Operators and Functionals on Finite Dimensional Spaces . . . . 12
2.10 Normed Spaces of Operators. Dual Spaces. . . . . . . . . . . . . . . . . . 13
3 Inner Product Spaces. Hilbert Spaces. 14
3.1 Inner Product Spaces. Hilbert Spaces. . . . . . . . . . . . . . . . . . . . . 14
3.2 Further Properties of Inner Product Spaces. . . . . . . . . . . . . . . . . 15
3.3 Orthogonal Complements and Direct Sums. . . . . . . . . . . . . . . . . 16
3.4 Orthonormal Set and Sequences . . . . . . . . . . . . . . . . . . . . . . . 18
3.5 Series Related to Orthonormal Sequences and Sets . . . . . . . . . . . . 18
3.6 Total Orthonormal Sets and Sequences . . . . . . . . . . . . . . . . . . . 18
3.7 Legendre, Hermite and Laguerre Polynomials . . . . . . . . . . . . . . . 18
3.8 Representation of Functionals on Hilbert Spaces . . . . . . . . . . . . . 19
4 Fundamental Theorems for Normed and Banach Spaces 21
4.1 Zorns Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Hahn-Banach Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 Hahn-Banach Theorem for Normed Spaces . . . . . . . . . . . . . . . . 22
4.4 Applications to Normed Linear Functionals . . . . . . . . . . . . . . . . 22
4.5 Adjoint Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.6 Reexive Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.7 Category Theorem. Uniform Boundedness Theorem. . . . . . . . . . . . 23
4.7.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.8 Strong and Weak Convergence . . . . . . . . . . . . . . . . . . . . . . . . 24
4.9 Convergence of Sequences of Operators and Functionals . . . . . . . . 25
4.10 Application to Summability of Sequences . . . . . . . . . . . . . . . . . . 25
4.11 Numerical Integration and Weak* Convergence . . . . . . . . . . . . . . 25
2
4.12 Open Mapping Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.13 Closed Linear Operators. Closed Graph Theorem . . . . . . . . . . . . . 26
3
1 Metric Spaces
1.1 Metric Space
Denition (Metrics and Metric Spaces). Let X be a set and dene a function d :
X X R. If d satises the following for all x, y, z X:
(M1). 0 d(x, y) <
(M2). d(x, y) =0 if and only if x = y
(M3). d(x, y) =d(y, x)
(M4). d(x, y) d(x, z) +d(z, y)
then d is called a metric and X is a metric space. (P. 3)
1.2 Examples of Metric Spaces
Example (l
p
). Let p 1 be a xed real number. Then l
p
is the sequence space such
that, if x =(x
n
) l
p
, then

j =1
|x
j
|
p
< (1)
and its metric is
d(x, y) =(

j =1
|x
j
y
j
|
p
)
1/p
(2)
Example (l

). The space l

is the set of bounded sequences. So if x = (x


n
) l

,
then there exists a c
x
such that
|x
i
| c
x
(3)
for i N. The metric is dened as
d(x, y) =sup
i N
|x
i
y
i
| (4)
for sequences x, y l

. A subset c l

is the linear space of all convergent se-


quences; the subset c
0
l

is the linear space of all sequences that converge to


zero.
4
1.3 Open Set, Closed Set, Neighborhood
Denition (Ball and Sphere). Let x
0
X and r >0. Dene the sets:
(a) B(x
0
; r ) ={x X|d(x
0
, x) <r } (open ball)
(b)

B(x
0
; r ) ={x X|d(x
0
, x) r } (closed ball)
(c) S(x
0
; r ) ={x X|d(x
0
, x) =r } (sphere) (P. 18)
Denition (Open Set and Closed Set). Let M X and X a metric space. If for each
x X, r >0, such that
B(x
0
; r ) M (5)
then we say M is an open set of X. M is said to be closed if its compliment in X is
open. That is, M is closed if:
M
c
= X M (6)
is open. (P. 18)
Denition (-neighborhood, Neighborhood). Let X be a metric space, and let x
X. An open ball B(x
0
; ) is called an -neighborhood of x. Any subset M of X that
contains an -neighborhood of x is called a neighborhood of x. (P. 19)
Denition (Interior Point and Interior). Let X be a metric space, and let x X. We
call x an interior point of M X if M is a neighborhood of x. The interior of M is
the set of all interior points of M. (P. 19)
Denition (Topological Space). Let X be a set and let T be a collection of subsets
of X. If T satises the conditions
(T1). , X T
(T2). Any arbitrary union of sets in T is another set of T.
(T3). Any nite intersection of sets in T is another set of T.
then we call T a topology on X and we call X a topological space. The elements of
T are the open sets of X. (P. 19)
Theorem1.1 (Metric Space as a Topological Space). Let X be a metric space. X is
a topological space whose elements of the topology T are the open sets with respect to
the metric space. We call this the topology induced by the metric on X. (P. 19)
5
Denition (Continuous Mapping). Let X and Y be metric spaces and let T : X Y
be a mapping. We say T is continuous at x
0
X if for each > 0, > 0 such that
whenever
d
Y
(Tx, Tx
0
) < then d
X
(x, x
0
) < (P. 20) (7)
Theorem1.2 (Continuous Mapping). A mapping T of a metric space X into a met-
ric space Y is continuous if and only if the inverse image of any open subset of Y is an
open subset of X. (P. 20)
Proof Enter Later as Exercise.
Denition (AccumulationPoint andClosure). Let X be a metric space, M a subset
of X, andlet x
0
M. If every neighborhoodof x
0
contains at least one element y M
distinct from x
0
, then we say x
0
is an accumulation point or limit point of M. The
set consisting of the accumulation points of M is called the closure of M denoted

M (8)
(Pg. 21)
Denition (Dense Set and Separable Space). A subset M of a metric space X is
said to be dense in X if

M = X (9)
X is said to be seperable if it has a countable subset which is dense in X.
6
1.4 Convergence, Cauchy Sequence, Completedness
Denition (Convergence and Limit). Let X be a metric space and let (x
n
) be a
sequence in X. (x
n
) is said to be convergent if x X such that
lim
n
d(x
n
, x) =0 (10)
we call x the limit of (x
n
) and denote it as
lim
n
x
n
=x or x
n
x (P. 25) (11)
Denition (Boundedness). Let X be a metric space and let M be a subset of X. The
diameter of M, denoted as (M) is dened as
(M) = sup
x,yM
d(x, y) (12)
If (M) <, then we call M a bounded subset. A sequence (x
0
) is called a bounded
sequence if it is a bounded subset of elements of X. (P. 26)
Lemma 1.3 (Convergent Sequences). Let X be a metric space. Then
(a). A convergent sequence in X is bounded and its limit is unique.
(b). If x
n
x, y
n
y in X, then d(x
n
, y
n
) d(x, y) (P. 26)
Denition (Cauchy Sequences and Completeness). A sequence (x
n
) in a metric
space X is said to be Cauchy if for each >0, N =N() such that
d(x
m
, x
n
) < m, n >N (13)
The metric space X is said to be complete if every Cauchy sequence in X converges
in X. (P. 28)
Theorem1.4 (Closure, closed set). Let M be a nonempty subset of a metric space
(X, d) and

M its closure as dened in the previous section. Then:
(a). x

M if and only if there is a sequence (x
n
) M such that x
n
x.
(b). M is closed if and only if the situation x
n
M, x
n
x implies that x M.
(P. 30)
Theorem1.5 (Complete Subspace). A subspace M of a complete metric space X is
itself complete if and only if the set M is closed in X. (P. 30)
7
2 Normed Spaces. Banach Spaces.
2.1 Vector Spaces
Denition (Linear Space). Let V be a set and let + be a closed binary operation on
V . V is a linear space (vector space) over the eld F if V is an abelian group under
addition:
a). a +b =b +a, a, b V
b). (a +b) +c =a +(b +c), a, b, c V
c). 0 V such that a +0 =a, a V
d). a V , a V such that a +(a) =0
and there is an operation between V and F called scalar multiplication such that:
e). (a) =()a, , F and a V
f ). 1 F such that 1 a =a, a V
g). (a +b) =a +b
h). (+)a =a +a (P. 50)
Denition (Linear Combinations, Linear Independence and Dependence). An
expression of the form:

1
x
1
+. . . +
n
x
n
(14)
where x
i
V and
i
F is called a linear combination. Suppose

1
x
1
+. . . +
n
x
n
=0 (15)
If this expression is true only for
1
=. . . =
n
=0, then we say x
1
, . . . , x
n
are linearly
independent, otherwise we say they are linearly dependent. (P. 53)
Denition (Span). Let M be a subset of a vector space V . The set of all possible
linear combinations of elements in M is called the span of M, denoted
span(M) (P. 53) (16)
Denition (Subspaces). Let X be a linear space and M a subset of X. If M is a
linear space itself, we call M a subspace of X. (P. 53)
8
Denition (Dimension and Basis). Let X be a linear space. If there exists a nite,
linearly independent subset M of X such that
span(M) = X (17)
then X is said to be nite dimensional and its dimension, denoted dim(X) is
dim(X) =|M| (18)
and call M a basis for X. If no such M exists we say X is innite dimensional. In
general, if M is linearly independent (not necessarily nite) and spans X, we call M
a Hamel basis. (P. 54)
Denition (Quotient Spaces). Let Y be a subspace of X. The coset of an element
x X with respect to Y is denoted
x +Y ={v|v =x +y, y Y } (19)
If we dene addition and scalar multiplication in the following way:
(w+Y ) +(x +Y ) =(w+x) +Y
(x +Y ) =x +Y
(20)
then the set of cosets forms a linear space called quotient space, denoted as X/Y .
The dimension of X/Y is called the codimension of Y , denoted as
codim(Y ) (21)
in other words, codim(Y ) =dim(X/Y ). (P. 57)
9
2.2 Normed Spaces. Banach Spaces.
Denition (Norms, Normed Spaces, Banach Spaces). Let : X R be an opera-
tor on a linear space X. If satises:
(N1). x 0
(N2). x =0 if and only if x
(N3). x =|| x
(N1). x +y x+y
then we call a norm on X, and X is called a normed linear space. A norm on X
induces a metric on X given by
d(x, y) =x y (22)
A Banach Space is a complete normed linear space. (P. 58)
2.3 Further Properties of Normed Spaces
2.4 Finite Dimensional Normed Spaces and Subspaces
2.5 Compactness and Finite Dimension
10
2.6 Linear Operators
Denition Linear Operators Let X and Y be linear spaces and let T : X Y be an
operator. We call T a linear operator if
T(x
1
+x
2
) =Tx
1
+Tx
2
F, x
1
, x
2
X (23)
2.7 Bounded and Continuous Linear Operators
Denition (Bounded Linear Operator). Let X and Y be normed linear spaces and
let T : D(T) Y be a linear operator, where D(T) X. T is said to be bounded if
c R such that x D(T) it holds
Tx
Y
cx
X
(24)
The norm of T, denoted as T is
T = sup
xD(T)
Tx
x
(25)
for all x =0, or alternatively:
T = sup
xD(T)
Tx (26)
for all x =1. We also have
Tx Tx (27)
for all bounded T. (P. 91)
Theorem2.1 (Continuity and Boundedness). Let T : D(T) Y be a linear opera-
tor, where D(T) X and X and Y are normed linear spaces. Then:
(a). T is continuous if and only if T is bounded.
(b). If T is continuous at a single point, it is continuous. (P. 97)
Corollary 2.2 (Continuity and Null Space). Let T be a bounded linear operator.
Then:
(a). x
n
x implies Tx
n
Tx.
(b). The nullspace N(T) is closed. (P. 98)
11
2.8 Linear Functionals.
Denition (Linear Functionals). Let X be a linear space over R or C. If f is a linear
operator from X to R or C then f is called a linear functional. (P. 104)
Denition (Dual Spaces). Let X be a linear space. Let X

be the set of all linear


functionals on X. Then X

forms a linear space via pointwise addition and scalar


multiplication, called the algebraic dual space of X. The dual space of X

, denoted
X

is called the second algebraic dual space of X. (P. 106)


Space Element Value at an Element
X x
X

f f (x)
X

g g( f )
Denition (Canonical Mapping). Fix some x X. Then the operator g dened as
g( f ) =g
x
( f ) = f (x) (28)
is a linear functional in the space X

. We can create a natural mapping of X into


X

via the map C : X X

dened as
C(x) =g
x
(29)
C is called the canonical mapping or embedding of X into X

. (P. 108)
2.9 Linear Operators and Functionals on Finite Dimensional
Spaces
12
2.10 Normed Spaces of Operators. Dual Spaces.
Denition (Space of Bounded Linear Operators) Let X, Y be normed linear
spaces. The set of all bounded linear operators from X to Y is called the space of
bounded linear operators from X to Y , denoted as
B(X, Y ) (P. 117) (30)
Theorem2.3 (B(X, Y )). The space B(X, Y ) is a normed linear space with the stan-
dard operator normT. (P. 118)
Theorem2.4 (Completeness). If Y is a Banach space, then B(X, Y ) is a Banach
space. (P. 118)
Denition (Dual Space X

). Let X be a normed linear space. Then the set of all


bounded linear functionals on X constitutes a normed linear space with the stan-
dard norm on operators f , which is called the dual space of X, and is denoted by
X

. (P. 119)
Theorem2.5 (Dual Space). The dual space X

of a normed linear space X is a Ba-


nach Space. (P. 120)
Denition (Isomorphisms and Isomorphic Normed Linear Spaces). A bijective
linear operator from a normed linear space X to a normed linear space

X which
preserves the norm, that is x X,
Tx =x (31)
is called an isomorphism, and X and

X are said to be isometrically isomorphic
normed linear spaces or simply isomorphic normed spaces. From a structural
point of view, they are identical. (P. 120)
13
3 Inner Product Spaces. Hilbert Spaces.
3.1 Inner Product Spaces. Hilbert Spaces.
Denition (Inner Product). Let X be a linear space over a eld F. Then a mapping
<, >: X X F is called an inner product if x, y, z X and F it holds that
(IP1). <x +y, z >=<x, z >+< y, z >
(IP2). <x, y >=<x, y >
(IP3). <x, y >=< y, x >
(IP4). <x, x >0 and equal to 0 if and only if x =0
An inner product denes a norm on X given by
x =

<x, x > (32)


and a metric on X given by
d(x, y) =x y =

<x y, x y > (33)


X with an inner product dened on it is called an inner product space. Complete
inner product spaces are called Hilbert Spaces. Inner product space are normed
linear spaces; Hilbert spaces are Banach spaces. (Pg. 129)
Proposition 3.1 (ParallelogramEquality). x+y
2
+xy =2(x
2
+y
2
) (Pg.
130)
Denition (Orthogonality). An element x of an inner product space X is said to be
orthogonal to an element y X if
<x, y >=0 (34)
and denote it as xy. If A X and a A it holds that x a, then we write x A;
we denote A B in a similar way for A, B X. (Pg. 131)
14
3.2 Further Properties of Inner Product Spaces.
Lemma 3.2 (Schwarz Inequality, Triangle Inequality). An inner product and the
corresponding normsatisfy the Schwarz inequality and triangle inequality as follows
| <x, y >| xy (35)
if {x, y} is a linearly dependent set. We also have the Triangle inequality as follows
x +y x+y (36)
also with the requirement that {x, y} is a linearly dependent set. Pg. 137
Lemma 3.3 (Continuity of Inner Product). Let X be an inner product space, and let
(x
n
), (y
n
) X be sequences such that x
n
x and y
n
y for some x, y X. Then
<x
n
, y
n
><x, y > (37)
(Pg. 138)
Theorem3.4 (Subspace). Let Y be a subspace of a Hilbert space H. Then:
(a). Y is complete if and only if Y is closed in H
(b). If Y is a nite dimensional, then Y is complete
(c). If H is separable, so is Y . More generally, every subset of a separable inner
product space is separable.
(Pg. 140)
15
3.3 Orthogonal Complements and Direct Sums.
Denition (Distance). In the metric of a normed linear space X, the distance for
an element s X to a non-empty subset M X is dened to be
= inf
yM
d(x, y) or
= inf
yM
x y
(38)
(Pg. 142)
Denition (Segment and Convex Set). Given two elements x, y M X, where X
is a linear space, the set of all z X of the form
z =x +(1)y (39)
for R, and 0 1, is called the segment joining x and y. M is said to be convex
if x, y M, all such zs are elements of M
Theorem3.5 (Minimizing Vector). Let X be an inner product space and M = a
convex subset which is complete. Then x X, y M such that
= inf
yM
x y =x y (40)
(Pg. 144)
Lemma 3.6 (Orthogonality). Let Y be a complete convex subspace of the inner prod-
uct space X. Fix some x X, then z =x y is orthogonal to Y . (Pg. 145).
Denition (Direct Sum). A linear space X is said to be the direct sum of two sub-
spaces Y, Z of X, written
X =Y

Z (41)
if each x X has a unique representation
x = y +z (42)
for some y Y , z Z. Z is called the algebraic compliment of Y in X. Y and Z are
complimentary pairs of subspaces in X. (P. 146)
16
Denition (Orthogonal Complement). Let H be a Hilbert space, and Y a closed
subspace of H. Then the set
Y

={z H|z Y } (43)
is called the orthogonal complement of Y . (Pg. 146)
Theorem3.7 (Direct Sum). Let Y be a complete, convex subspace of the inner prod-
uct space X. Fix some x X, then
H =Y Z (44)
where Z =Y

. (Pg. 146)
Proof Since H is a Hilbert space, its complete; since Y is a closed subspace of H
thenit is also complete, andis convex. By the orthogonality theorem, x H, y Y
such that
x = y +z (45)
for some z Z =Y

. Now suppose y
1
Y , and z
1
Z such that
x = y +z = y
1
+z
1
(46)
This implies that
y y
1
=z z
1
(47)
since y, y
1
Y , then y y
1
Y ; similarly zz
1
Z =Y

. So y y
1
=zz
1
Y Y

=
{0}, thus
y y
1
=0
z z
1
=0
(48)
So y = y
1
and z =z
1
, so we have uniqueness. The relation x = y+z sets up a operator
P :H Y
x y =Px
(49)
called the projection operator. P maps
H Y
Y Y
Z =Y

onto {0}
(50)
17
Lemma 3.8 (Null Space). The orthogonal complement Y

of a closed subspace Y of
a Hilbert space H is the null space N(P) of the orthogonal projection P of H onto Y .
(Pg. 148)
Lemma 3.9 (Closed Subspace). If Y is a closed subspace of a Hilbert space H, then
Y =Y

(51)
Proof Let y Y then y

Y

it holds that < y, y

>= 0, thus y Y

, thus Y
Y

. By the direct sum lemma we have that
H =Y Y

(52)
So for any x Y

, y Y , z Y

such that
x = y +z (53)
or
z =x y (54)
Since x, y Y

then z Y

and z Y

then z = 0, so x = y, thus x Y . So
Y

Y . So evidently Y

=Y . (Pg. 149)
3.4 Orthonormal Set and Sequences
3.5 Series Related to Orthonormal Sequences and Sets
3.6 Total Orthonormal Sets and Sequences
3.7 Legendre, Hermite and Laguerre Polynomials
18
3.8 Representation of Functionals on Hilbert Spaces
Theorem3.10 (Reiszs Representation Theorem - Bounded Linear Functionals on
Hilbert Spaces). Let H be a Hilbert space. Every bounded linear functional f on H
can be represented in terms of the inner product, namely
f (x) =<x, z > (55)
where z depends on f , is uniquely determined by f and has norm
z =f (56)
(Pg. 188)
Proof Suppose f = 0, then z = 0 and < x, z >= 0 for all x H. Otherwise, we con-
sider any other f (x) such that, at least for some x H, f (x) =0.
Claim: N( f ) is closed, further N(t )

={0}.
Let x
n
x and (x
n
) N( f ). Since f is bounded, we have that x
n
x implies
f (x
n
) f (x). Since (x
n
) N( f ), the ( f (x
n
)) is a sequence of zeros and thus its limit
is zero, so
f (x
n
) f (x) =0 (57)
and thus x N( f ), so N( f ) is a closed subspace of H. Thus, by the theoremof direct
sums we have that
H =N( f ) N( f )

(58)
Since f =0, then z
0
H such that f (z
0
) =0. Suppose N( f )

={0}, then we should


be able to write z
0
as some linear combination of an element x N( f ) and 0 (since
we presume N( f )

={0}). This clearly implies that z


0
N( f ), which contradicts the
assumption that f (z
0
) =0, thus N( f )

={0}.
Now, for some arbitrary x H and some non-zero z
0
N( f )

set up the equa-


tion
v = f (x)z
0
f (z
0
)x (59)
which is clearly an element of H. Apply f to v:
f (v) = f (x) f (z
0
) f (z
0
) f (x) (60)
which is very clearly equal to 0, thus v N( f ), which implies that v N( f )

, specif-
ically <v, z
0
>=0. Since v = f (x)z
0
f (z
0
)x we have
<v, z
0
>=< f (x)z
0
f (z
0
)x, z
0
>
= f (x) <z
0
, z
0
>f (z
0
) <x, z
0
>
=0
(61)
19
implying that
f (x) <z
0
, z
0
>= f (z
0
) <x, z
0
>
f (x) =
f (z
0
)
<z
0
, z
0
>
<x, z
0
>
(62)
By allowing z =
f (z
0
)
<z
0
,z
0
>
z
0
we have
f (x) =<x, z > (63)
Since z
0
is non-zero, then z is an element of H and <x, z > is well dened.
To show uniqueness, consider
f (x) =<x, z
1
>=<x, z
2
> (64)
which implies that <x, z
1
><x, z
2
>=0. So we have that
<x, z
1
><x, z
2
>=<z
1
, x ><z
2
, x >
=<z
1
, x ><z
2
, x >
=<z
1
z
2
, x >
=<x, z
1
z
2
>
(65)
which must hold for every x H including x =z
1
z
2
. Thus we have that
<z
1
z
2
, z
1
z
2
>=0 (66)
which only holds if z
1
z
2
=0 implying that z
1
=z
2
, so we have uniqueness.
To nd the norm, recall that f is bounded, and therefore f exists. So we have:
z
2
=| <z, z >| =| f (z)| f z
z f
(67)
and
| f (x)| =| <x, z >| xz
sup
x=1
| f (x)| =f z
(68)
Thus, z =f .
20
4 Fundamental Theorems for Normed and Banach
Spaces
4.1 Zorns Lemma
Denition (Partially Ordered Set and Chains). A partially ordered set is a set M
on which there is dened partial ordering, that is, a binary relation which is written
and satises the conditions
(PO1) a a for every a M.
(PO2) If a b and b a, then a =b.
(PO3) If a b and b c, then a c.
A totally ordered set or chain is a partially ordered set such that every two ele-
ments of the set are comparable. (P. 210)
Lemma 4.1 (Zorns Lemma). Let M = be a partially ordered set. Suppose that every
chain C M has an upper bound. Then M has at least one maximal element. (P.
211)
4.2 Hahn-Banach Theorem
Denition (Sublinear Functional). Let p be a real valued functional on a linear
space X which is subadditive, that is
p(x +y) p(x) +p(y) (69)
for all x, y X and positive homogeneous, that is
p(x) =p(x) (70)
for all 0. We call such a p a sublinear functional. Pg. 213
Theorem4.2 (Hahn-Banach Theorem - Extension of Linear Functionals). Let X
be a real linear space and p a sublinear functional on X. Further, let f be a linear
functional which is dened on a subspace Z of X and satises
f (x) p(x) (71)
for all x Z. Then f has a linear extension

f from Z to X satisfying

f (x) p(x) (72)


that is,

f is a linear functional on X, satises the inequality above on X and

f (x) =
f (x) for all x Z. (Pg. 214)
21
4.3 Hahn-Banach Theoremfor Normed Spaces
Theorem4.3 (Hahn-Banach Theorem for Normed Spaces). Let f be a bounded
linear functional on a subspace Z of a normed linear space X. Then there exists a
bounded linear functional

f on X which is an extension of f to X and has the same
norm

f
X
=f
Z
(73)
Theorem4.4 (Bounded Linear Functionals). Let X be a normed linear space and
let x
0
=0 be any element of X. Then there exists a bounded linear functional

f on X
such that

f =1 and

f (x
0
) =x
0
(74)
4.4 Applications to Normed Linear Functionals
4.5 Adjoint Operator
Denition (Adjoint Operator). Let T : X Y be a bounded linear operator, where
X and Y are normed linear spaces. Then the adjoint operator T

: Y

X

of T is
dened by
f (x) =(T

g)(x) =g(Tx) (75)


where g Y

. (P. 232)
An illustration of the denition above:
x X Tx Y
T

g X

g Y

(T

g)(x) R g(Tx) R
Theorem4.5 (Norm of Adjoint). The adjoint T

of T is a bonded linear operator


and its norm is
T

=T (P. 232) (76)


Denition (Annihilator). Let M = be a subset of a normed linear space X. The
annihilator M
a
of M is dened to be the set of all bounded linear functionals on X
which are zero everywhere on M. Thus M
a
X

. (P. 126)
4.6 Reexive Spaces
22
4.7 Category Theorem. UniformBoundedness Theorem.
Denition (Category). A subset M of a metric space X is said to be
(a). rare (or nowhere dense) in X if its closure

M has no interior points,
(b). meager (or of the rst category) in X if M is the union of countably many
sets each of which is rare in X,
(c). nonmeager (or of the second category) in X if M is not meager in X. (P.
247)
Theorem4.6 (Baires Category Theorem for Complete Metric Spaces). If a metric
space X = is complete, it is nonmeager in itself.
Hence if X = is complete and
X =

k=1
A
k
(77)
and A
k
closed, then at least on A
k
contains a nonempty open subset
Theorem4.7 (UniformBoundedness Theorem). Let (T
n
) be a sequence of bounded
linear operators T
n
: X Y from a Banach space X to a normed linear space Y such
that (T
n
x) is bounded x X, say
T
n
x c
x
(78)
where c
x
is a real number. Then the sequence of norms T
n
is bounded, that is there
is a c such that
T
n
c (79)
(Pg. 249)
4.7.1 Exercises
1. (P. 238) Let X and Y be normed spaces, T : X Y a bounded linear operator
and M =R(T), the closure of the range of T. Show:
M
a
=N(T

)
2. (P. 239) Let B be a subset of a normed linear space X. The annihilator
a
B of
B is dened to be
a
B ={x X| f (x) =0, f B}
Show the in the previous exercise
R(T)
a
N(T

)
What does the mean with respect to the task of solving an equation Tx = y?
23
4.8 Strong and Weak Convergence
Denition (Strong Convergence). A sequence (x
n
) in a normed linear space X is
said to be strongly convergent (of convergent in the norm) if x X such that
lim
n
x
n
x =0 (80)
written as
lim
n
x
n
=x (81)
or simply
x
n
x (82)
x is called the strong limit of (x
n
) and we say (x
n
) converges strongly to x. (Pg.
256)
Denition (Weak Convergence). A sequence (x
n
) in a normed linear space X is
said to be weakly convergent if x X such that f X

lim
n
f (x
n
) = f (x) (83)
This is written
x
n
w
x (84)
The element x is called the weak limit of (x
n
) and we say that (x
n
) converges weakly
to x. (P. 257)
Lemma 4.8 (Weak Convergence). Let (x
n
) be a weakly convergent sequence in a
normed linear space X, say x
n
w
x, then
a) the weak limit x of (x
n
) is unique
b) every subsequence of (x
n
) converges weakly to x
c) the sequence (x
n
) is bounded.
24
4.9 Convergence of Sequences of Operators and Functionals
4.10 Application to Summability of Sequences
4.11 Numerical Integration and Weak* Convergence
4.12 Open Mapping Theorem
Denition (Open Mapping). Let X and Y be metric spaces. Then T : D(T) Y is
called an open mapping if for every set in D(T) the image is open in Y . (P. 286)
Theorem4.9 (Open Mapping Theorem, Bounded Inverse Theorem). A bounded
linear operator T froma Banach space X onto a Banach space Y is anopenmapping.
Hence if T is bijective, T
1
is continuous and thus bounded. (P. 286)
25
4.13 Closed Linear Operators. Closed Graph Theorem
Denition (Closed Linear Operator). Let X and Y be normed space and
T : D(T) Y a linear operator with domain D(T) X. Then T is called a closed
linear operator if its graph
G(T) ={(x, y)|x D(T), y =Tx} (85)
is closed in the normed space X Y , where the two algebraic operations of a vector
space in X Y are dened as usual, that is
(x
1
, y
1
) +(x
2
, y
2
) =(x
1
+x
2
, y
1
+y
2
)
(x
1
, y
1
) =(x
1
, y
1
)
(86)
( a scalar) and the norm on X Y is dened by
(x, y) =x+y (87)
(P. 292)
Theorem4.10 (Closed Graph Theorem). Let X and Y be Banach spaces and T :
D(T) Y a closed linear operator, where D(T) X. Then if D(T) is closed in X, the
operator T is bounded. (P. 292)
Proof Begin with preliminary claims.
Claim: X Y is complete.
Let (z
n
) =((x
n
), (y
n
)) X Y be a Cauchy sequence. So >0, N = N() such
that m, n >N
z
n
z
m
=(x
n
, y
n
) (x
m
, y
m
)
=(x
n
x
m
), (y
n
y
m
)
=x
n
x
m
+y
n
y
m
<
(88)
is true. Thus >0, N =N() such that m, n >N
x
n
x
m
<
y
n
y
m
<
(89)
holds true. Thus (x
n
) and y
n
are Cauchy sequences, and thus converge to some x
and y in X and Y respectively, since X and Y are Banach spaces. Thus (x
n
, y
n
)
(x, y) X Y , so X Y is a Banach Space.
Claim: D(T) and G(T) are Banach Spaces.
26
Since X and X Y are Banach spaces, and D(T) and (G)(T) are closed by as-
sumption, then by theorem they must also be complete, and thus Banach.
Now dene an operator P : G(T) D(T) as
P(x, Tx) =x (90)
Claim: P is a bounded linear operator.
For any that is a scalar and (x
1
, Tx
1
), (x
2
, Tx
2
) G(T) we have that
P((x
1
, Tx
1
) +(x
2
, Tx
2
)) =P((x
1
, Tx
1
) +(x
2
, Tx
2
))
=P((x
1
, Tx
1
) +(x
2
, Tx
2
))
=P(x
1
+x
2
, Tx
1
+Tx
2
)
=P(x
1
+x
2
, T(x
1
+x
2
))
=x
1
+x
2
=P(x
1
, Tx
1
) +P(x
2
, Tx
2
)
(91)
So P is linear. We also have that
P(x, Tx) =x
(x, Tx) =x+Tx
(92)
thus we always have that
P(x, Tx) (x, Tx) (93)
so P is bounded.
Claim: P is bijective.
For any x D(T), there exists a Tx Y , so for any x X there is a (x, Tx) G(T)
that is mapped to x by P, so P is onto. Suppose P(x
1
, Tx
1
) =P(x
2
, Tx
2
), then x
1
=x
2
so it must true that Tx
1
= Tx
2
, thus (x
1
, Tx
1
) = (x
2
, Tx
2
), so P is one-to-one. P is
bijective. Additionally P
1
x =(x, Tx).
Since P is a bounded linear operator from the Banach space G(T) to the Banach
space D(T), then by the Open Mapping Theorem, P
1
is continuous and bounded.
So there exists a c such that for all x D(T):
P
1
x =(x, Tx)
=x+Tx cx
(94)
so clearly
Tx (c 1)x (95)
for any x DT. Thus T is bounded.
27

You might also like