Functional Analysis

Functional Analysis Lecture script
Prof. D. Salamon
February 12, 2007
Coordinating: Lukas Lewark
Writing: Urs Fuchs, Sasa Parad
-
, Andrin Schmidt
Correcting: Philipp Arbenz, David Umbricht, Dominik Staub, Thomas Rast
If you want to be informed in case of a new version or if you nd any mistakes
please write to [email protected].
Warning: We are sure there are lots of mistakes in these notes. Use at your own
risk! Corrections would be appreciated and can be sent to [email protected];
please always state what version (look in the Id line below) you found the error
in. For further information see:
http://vmp.ethz.ch/wiki/index.php/Vorlesungsmitschriften
$Id: fa.tex 1894 2007-02-12 15:05:18 charon$
i
Contents
0 Introduction 1
1 Basic Notions 2
1.1 Finite dimensional vector spaces . . . . . . . . . . . . . . . . . . 2
1.2 Linear Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Innite dimensional vector spaces . . . . . . . . . . . . . . . . . . 6
1.4 The Theorem of Arzela-Ascoli . . . . . . . . . . . . . . . . . . . . 8
1.5 The Baire Category Theorem . . . . . . . . . . . . . . . . . . . . 12
1.6 Dual spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.7 Quotient spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Functional Analysis 22
2.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 Product spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Extension of bounded linear functionals . . . . . . . . . . . . . . 29
2.4 Reexive Banach Spaces . . . . . . . . . . . . . . . . . . . . . . . 34
3 The weak and weak* topologies 38
3.1 The weak topology . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 The weak* topology . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3 Ergodic measures . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4 Compact operators and Fredholm theory 56
4.1 Compact operators . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2 Fredholm operators . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Spectral Theory 65
5.1 Eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2 Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.3 Compact operators on Banach spaces . . . . . . . . . . . . . . . . 73
5.4 Spectral Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6 Unbounded operators 84
Index 89
ii
0 Introduction 25.10.2006
0 Introduction
Remark: Functional Analysis can be viewed as a combination of linear alge-
bra and topology:
Linear Algebra Topology Functional Analysis
vector spaces metric spaces normed vector spaces
linear maps continuous maps continuous linear maps
subspaces closed subsets closed subspaces
The vector spaces concerned in Functional Analysis generally have innite di-
mension.
1
1 Basic Notions 25.10.2006
1 Basic Notions
1.1 Finite dimensional vector spaces
Denition: A normed vector space is a pair (X, | |) where X is a vector
space (we consider only real spaces) and X [0, ), x |x| is a norm, i.e.
1. |x| = 0 x = 0
2. |x| = [[ |x| x X, 1
3. |x +y| |x| +|y| x, y X
Remark: A norm induces a metric on the vector space, by d(x, y) := |xy|.
Denition: A Banach space is a complete normed vector space (X, |.|), i.e.
every Cauchy sequence in (X, d) converges.
Denition: Two norms | |
1
, | |
2
on a real vector space X are called equiv-
alent, if
c > 0 x X :
1
c
|x|
1
|x|
2
c|x|
1
.
Example:
1. X = 1
n
, x = (x
1
, . . . , x
n
) 1
n
|x|
p
:=
_
n
i=1
[x
i
[
p
_1
p
, 1 < p <
|x|
:= max[x
i
[ [ 1 i n
2. (M, /, ) measure space
L
p
() = f : M 1 [ f measurable ,
_
M
[f[
p
d < /
where means equal almost everywhere.

|f|
p
:=
__
M
[f[
p
d
_1
p
, 1 p <
(L
p
(), | |
p
) is a Banach space, and if M = 1, . . . , n we get Example
1.
3. Let M be a locally compact and hausdor topologic space.
C
c
(M) := f : M 1 [ f is continuous and has compact support
|f|
:= sup
mM
[f(m)[
Combine 2. and 3.:
Let B 2
M
be the Borel -Algebra and : B [0, ] a Radon measure.
Then one can dene |f|
p
, |f|
for all f C
c
(M). These two norms are
not equivalent, because there are Cauchy sequences converging in | |
which are not convergent in | |

p
, e.g.
f
n
: 1 1, f
n
=
_
_
1
n
_1
p
x [0, n]
0 otherwise
2
4.
X = C
k
b
(1
n
) = f : 1
n
1 [ f C
k
and
sup
xR
n
[
f(x)[ < = (
1
, . . .
n
) N
n
|f|
c
k := sup
]]k
sup
xR
n
[
f(x)[ = sup
]]k
|
f|
with [[ :=
1
+. . . +
n
.
The normed vector space (C
k
b
(1
n
), |.|
c
k) is called Sobolev space.
Lemma 1: Let X be a nite dimensional vector space.
Any two norms on X are equivalent.
Proof: w.l.o.g. X = 1
n
Let e
1
, . . . , e
n
X be the standard basis of X.
Let 1
n
1 : x |x| be any norm and
c :=
_
n
i=1
|e
i
|
2
|x| = |
n
i=1
x
i
e
i
| (1)
i=1
|x
i
e
i
| (2)
=
n
i=1
[x
i
[|e
i
| by Cauchy-Schwarz (3)
_
n
i=1
x
2
i
_
n
i=1
|e
i
|
2
(4)
= c|x|
2
(5)
That proves one half of the inequality.
It follows that the function 1
n
1 : x |x| is continuous with respect to the
Euclidian norm on 1
n
:
[|x| |y|[ |x y| c|x y|
2
The set S
n
:= x 1
n
[ |x|
2
= 1 is compact with respect to the Euclidian
norm.
x
0
Sx S
n
: |x| |x
0
| =: > 0
x 1
n
:
x
|x|
2
S
n
and so
_
_
_
_
x
|x|
2
_
_
_
_

and therefore
|x| |x|
2
Which is the other half of the inequality. 2
Lemma 2: Every nite dimensional vector space (X, | |) is complete.
Proof: True for (1
n
, | |
2
).
true for 1
n
with any norm.
true for any nite dimensional vector space. 2
3
Lemma 3: Let (X, | |) be any normed vector space and Y X a nite
dimensional linear subspace.
Y is a closed subset of X.
Proof: Y is a nite dimensional normed vector space.
Lemma 2
Y is complete.
(y
n
)
nN
Y, lim
n
y
n
= y X
Y complete
y Y
Y is closed. 2
Theorem 1: Let (X, |.|) be a normed vector space and B := x X [ |x| 1
be the unit ball. Then
dim(X) < B is compact.
Proof of Theorem 1, : Let e
1
, . . . , e
n
be a basis of X and dene
T : 1
n
X by T :=
n
i=1
i
e
i
The function 1
n
1 : |T| is a norm on 1
n
Lemma 1
c > 0 1
n
: max
i=1,...,n
[
i
[ c|T|
Let (x
)
N
B be any sequence and denote
= (
1
, . . . ,
n
) := T
1
x
i
[ c|T
| = c|x
| c
Heine-Borel
(
)
N
has a convergent subsequence (
k
)
kR,
1
<
2
<...
(
k
i
)
kN
converges in 1 for i = 1, . . . , n
x
k
=
v
k
1
e
1
+. . . +
k
n
e
n
converges; so B is sequentially compact.
We use that on metric spaces sequential compactness and compactness dened
by existence of nite subcoverings are equivalent; that will be proven in Theorem
2. 2
Lemma 4: 0 < < 1, (X, | |) a normed vector space, Y _ X a closed
subspace.
x X so that |x| = 1, inf
yY
|x y| > 1
Proof: Let x
0
X Y . Denote
d := inf
yY
|x
0
y| > 0
(d > 0 because Y is closed.) y
0
Y so that |x
0
y
0
| <
d
1
Let x :=
x
0
y
0
|x
0
y
0
|
|x| = 1
|x y| =
_
_
_
_
x
0
y
0
|x
0
y
0
|
y
_
_
_
_
=
1
|x
0
y
0
|
|x
0
y
0
|x
0
y
0
|y|
. .
Y
. .
d
d
|x
0
y
0
|
> 1
2
4
Proof of Theorem 1, : Suppose dim(X) =
We construct a sequence x
1
, x
2
. . . in B so that
|x
i
x
j
|
1
2
i ,= j
Then (x
i
)
iN
has no convergent subsequence.
We construct by induction sets x
1
, . . . , x
n
B so that |x
i
x
j
|
1
2
i ,= j
n = 1: pick any vector x B.
n 1: Suppose x
1
, . . . , x
n
have been constructed.
Dene
Y := spanx
1
, . . . , x
n
=
_
n
i
1
i
x
i
i
1
_
_ X
Y is closed.
So by Lemma 4 x
n+1
X so that
|x
n+1
| = 1, |x
n+1
y|
1
2
y Y
|x
n+1
x
i
|
1
2
i = 1, . . . , n
This completes the inductive construction of the sequence. 2
1.2 Linear Operators
(X, | |
X
), (Y, | |
Y
) normed vector spaces.
Denition: A linear operator T : X Y is called bounded if c > 0x X :
|Tx|
Y
c|x|
X
The number |T| := sup
xX,x,=0
|Tx|
Y
|x|
X
is called the norm of T.
Notation: L(X, Y ) := T : X Y [ T is a bounded linear operator is a
normed vector space, and complete whenever Y is complete. (Analysis II)
Lemma 5: T : X Y linear operator. Equivalent are
i. T is bounded
ii. T is continuous
iii. T is continuous at 0.
Proof: i. ii. |Tx Ty|
Y
|T||x y|
X
Lipschitz continuous
ii. iii. trivial
iii. i. = 1 > 0 x X :
|x|
X
|Tx|
Y
1
0 ,= x X
_
_
_
_
x
|x|
X
_
_
_
_
X
=
_
_
_
_
Tx
|x|
X
_
_
_
_
Y
1
|Tx|
Y

1
..
c
|x|
X
2
5
Lemma 6: Let X, Y be normed vector spaces of nite dimension every
linear operator T : X Y is bounded.
Proof: Choose a basis e
1
, . . . e
n
X.
a)
c
1
:=
n
i=1
|Te
1
|
Y
b) The map
1
n
1 : (x
1
, . . . x
n
)
_
_
_
_
_
n
i=1
x
i
e
i
_
_
_
_
_
X
is a norm on 1
n
. By Lemma 1, c
2
> 0 so that
max
i=1,...n
[x
i
[ c
2
_
_
_
_
_
n
i=1
x
i
e
i
_
_
_
_
_
X
(x
1
, . . . x
n
) 1
n
a)&b) x =
n
i=1
x
i
e
i
X we have
|Tx|
Y
=
_
_
_
_
_
n
i=1
x
i
Te
i
_
_
_
_
_
Y
i=1
[x
i
[ |Te
i
|
Y
(max
i
[x
i
[)
n
i=1
|Te
i
|
Y
= c
i
max [x
i
[ < c
1
c
2
|x|
X
2
What for innite dimensions?
1.3 Innite dimensional vector spaces
Example 1: Let X = C
1
([0, 1]; X), |x|
X
:= sup
0t1
[f(t)[, Y = 1 and
Tx := x(0). T is linear and not bounded. This is not a Banach space.
Example 2: X innite dimensional.
e
i
iI
basis of X with |e
i
| = 1 i I (the axiom of choice is needed to prove
this for any vector space).
Choose sequence i
1
, i
2
, . . .
Dene c
i
:=
_
k, i = i
k
0, i / i
1
, i
2
, . . .
Dene T : X 1 by
T
_
iI
i
e
i
. .
nite sum
_
=
iI
i
c
i
Te
i
k
=
i
k
= k
We found three incidences where nite and innite dimensional space dier:
Compactness of the unit ball (see Theorem 1)
Completeness (see Lemma 2)
Boundedness of Linear Functionals (see Lemma 6 and Example 1)
Denition: A metric space (M, d) is called totally bounded if
> 0 x
1
, . . . , x
m
M : M =
m
_
i=1
B
(x
i
)
where
B
(x) := x
M [ d(x, x
) <
6
Theorem 2: Let (M, d) be a metric space. Equivalent are:
i. Every sequence has a convergent subsequence (sequential compactness).
ii. Every open cover has a nite subcover (compactness).
iii. (M, d) is totally bounded and complete.
Proof: i. ii.:
T (M, d) 2
M
set of open subsets of M.
Let | T (M, d) be an open cover of M.
Step 1
> 0 x M U | so that B
(x) U
Suppose > 0 x M U | so that B
(x) , U.
Pick =
1
n
x
n
M U | : B1
n
(x
n
) , U
By i. convergent subsequence x
n
k
x M.
Choose U | so that x U; choose > 0 so that B
U.
Choose k so that d(x, x
n
k
) <

2
and
1
n
<

2
.
B 1
n
k
(x
n
k
) B
2
(x
n
k
) B
(x) U
contradiction.
Step 2 | has a nite subcover.
Suppose not.
Let > 0 be as in Step 1.
Construct sequences x
1
, x
2
, . . . M and U
1
, U
2
, . . . | so that
B
(x
n
) U
n
and x
n
/ U
1
, . . . U
n1
x
n
can be chosen like that because otherwise the U
1
, . . . U
n1
would form a nite
subcover.
Pick any x
i
M.
By Step 1 U
i
| so that B
(x
i
) U
i
.
Suppose x
1
, . . . , x
n
and U
1
, . . . , U
n
have been found.
U
1
U
2
. . . U
n
,= M
x
n+1
M (U
1
U
2
. . . U
n
)
By Step 1 U
n+1
| so that B
(x
n+1
) U
n+1
Given the sequences (x
k
)
kN
, (U
k
)
kN
we observe:
For k < n : B
(x
k
) U
k
, x
n
/ U
k
So d(x
n
, x
k
)
d(x
k
, x
n
) k ,= n
There is no convergent subsequence.
ii. iii.:
Assume every open cover has a nite subcover.
a. Take
| := B
(x) [ x M
Then x
1
, . . . x
m
M so that
m
_
i=1
B
(x
i
) = M
So M is totally bounded.
7
b. (M, d) is complete:
Let (x
n
)
nN
be a Cauchy sequence.
Assume (x
n
)
nN
does not converge.
(x
n
) has no convergent subsequence.
(x
n
) has no limit point.
M () > 0 so that the set n N [ x
n
B
()
() is nite.
Take | := B
()
() [ M.
Then | has no nite subcover.
iii. i.:
Assume (M, d) is totally bounded and complete.
Let (x
n
)
nN
be any sequence in M.
Claim: There is a sequence of innite subsets
N T
0
T
1
. . .
such that d(x
n
, x
m
) 2
k
x
n
, x
m
T
k
.
Cover M by nitely many balls
m
_
i=1
B1
2
(
i
) = M
i so that the set n N [ x
n
B1
2
(
i
) =: T
0
is innite.
Then n, m T
0
we have
d(x
n
, x
m
) d(x
n
,
i
) +d(
i
, x
m
) < 1
Suppose T
k1
has been constructed.
Cover M by nitely many balls
M =
m
_
i=1
B 1
2
k+1
(
i
)
Then i so that the set
T
k
:= n T
k1
[ x
n
B 1
2
k+1
(
i
)
is innite.
n, m T
k
:
d(x
n
, x
m
) < d(x
n
,
i
) +d(
i
, x
m
) <
1
2
k
Claim convergent subsequence.
Pick n
1
< n
2
< . . . so that n
k
T
k
.
n
l
, n
k
T
k
l k
d(x
n
k
, x
n
l
)
1
2
k
l k
The sequence (x
k
)
kN
is Cauchy.
M complete
The sequence converges. 2
1.4 The Theorem of Arzela-Ascoli
Denition: (M, d) metric space. A subset D M is called dense if
x M > 0 : B
(x) D ,=
Denition: A metric space (M, d) is called separable if it contains a countable
dense subset.
Corollary: Every compact metric space (M, d) is separable.
8
Proof: Given n N.
1
, . . .
n
M : M =
n
_
i=1
B1
n
(
i
)
Dene
D
n
:=
1
, . . .
n
Dene
D =
_
i=1
D
n
M
D is countable.
Given x M, > 0, pick n N so that
1
n
< .
Then D
n
so that x B1
n
().
x B
(), D B
(x) D ,=
So D is dense. 2
Exercise: (X, d
X
) compact metric space, (Y, d
Y
) complete metric space and
C(X, Y ) := f : X Y [ f continuous
d(f, g) := sup
xX
d
Y
(f(x), g(x)) < f, g C(X, Y )
Show that (C(X, Y ), d) is a complete metric space.
This is exercis 1a) on Series 2.
Denition: A subset T C(X, Y ) is called equicontinuous if
> 0 > 0 : x, y X f T : d
X
(x, x
) < d
Y
(f(x), f(x
)) <
Theorem 3 (Arzela-Ascoli): (X, d
X
) compact metric space and (Y, d
Y
)
complete metric space, T C(X, Y ).
Equivalent are:
i. T is compact.
ii. T is closed, equicontinuous and T(x) := f(x) [ f T Y is compact
for every x X.
Proof: i. ii.:
T is closed (every compact set in a metric space is closed).
Fix x X. Then the evaluation map ev
x
: T Y, ev
x
(f) := f(x) is
continous. So ev
x
(T) = T(x) is compact.
Pick > 0. f
1
, . . . , f
m
T so that T
m
i=1
B
(f
i
)
Choose > 0 so that ix, x
X :
d
X
(x, x
) < d
Y
(f
i
(x), f
i
(x
)) <
Given f T choose i so that d(f, f
i
) < . Now for x, x
with d
X
(x, x
) < :
d
Y
(f(x), f(x
)) d(f(x), f
i
(x))
. .
<
+d(f
i
(x), f
i
(x
)
. .
<
+d(f
i
(x
), f(x
))
. .
<
< 3
ii. i.:
9
To Show: T is compact. Let f
n
T be any sequence.
We know that X is separable, i.e. there is countable dense subset D X with
D in the form D = x
1
, x
2
, . . ..
Claim 1 There is a subsequence g
i
:= f
n
i
so that the sequence g
i
(x
k
) Y
converges as i for every k N.
Proof of Claim 1 T(x
i
) is compact and f
n
(x
1
) T(x
i
). Thus there is a
subsequence (f
n
1,i
)
i
so that (f
n
1,i
(x
1
))
i
converges. By the same argument there
is a subsequence (f
n
2,i
)
i
of (f
n
1,i
)
i
so that (f
n
2,i
(x
2
))
i
converges.
Induction: There is a sequence of subsequences (f
n
k,i
)
i=1
so that
(f
n
k,i
(x
k
))
i=1
converges as i .
(f
n
k+1,i
)
i=1
is a subsequence of (f
n
k,i
)
i
for every k N.
Dene g
i
:= f
n
i,i
, that is the Diagonal sequence construction. This satises
g
i
(x
k
) converges for all k N as i . But we want more: Namely, conver-
gence in the whole of X, not only D, and uniform convergence.
Claim 2: (g
i
)
i
is a Cauchy sequence in C(X, Y ).
With Claim 2:
Since C(X, Y ) is complete the sequence g
i
converges. Since T is closed, its limit
belongs to T.
Proof of Claim 2:
Choose > 0 and > 0 as in the denition of equicontinuity, i.e.
Condition 1 x, x
Xf T : d
X
(x, x
) < d
Y
(f(x), f(x
)) <
Since D is dense in X we have
X =
_
k=1
B
(x
k
)
By Theorem 2
Condition 2
m N : X =
m
_
k=1
B
(x
k
)
Since (g
i
(x
k
))
i=1
is Cauchy for every k 1, . . . n:
Condition 3 N Ni, j > Nk 1, . . . m : d
Y
(g
i
(x
k
), g
j
(x
k
)) <
We prove: i, j N d(g
i
, g
j
) < 3. Remember that
d(g
i
, g
j
) := sup
xX
d
Y
(g
i
(x), g
j
(x))
Fix an element x X.
By Condition 2 k 1, . . . n so that d
X
(x, x
k
) < .
By Condition 1
Condition 4
i N : d
Y
(g
i
(x), g
i
(x
k
)) <
i, j N d
Y
(g
i
(x), g
j
(x))
d
Y
(g
i
(x), g
i
(x
k
)) +d
Y
(g
i
(x
k
), g
j
(x
k
)) +d
Y
(g
i
(x
k
), g
j
(x))
And this is, by Condition 3 and 4, smaller than + + = 3.
2
Looking closely at the proof, one can weaken the three condition of the theorem
of Arzela-Ascoli.
10
Theorem 3 (Arzela-Ascoli revisited): Let (X, d
X
) be compact, (Y, d
Y
)
complete metric spaces and T C(X, Y ). Equivalent are
(i) T has a compact closure.
(ii) T is equicontinuous and T(x) Y has a compact closure x X.
Proof : T is the closure of T in C(X, Y ). From (i) it follows that T(x) =
T(x)x X.
T(x) T(x) is always true and an exercise.
T(x) T(x). Proof:
Let y T(x) sequence y
k
T(x), y
k
y.
f
k
T so that f
k
(x) = y
k
.
f
k
has a convergent subsequence f
k
i
f C(X, Y ) where f
k
i
T
and f T.
So y = f(x) T(x).
(i)(ii) T is equicontinuous by Theorem 4 for T.
T(x) = T(x) which is compact by Theorem 3.
(ii)(i) Claim 1 and Claim 2 in Theorem 3 only use (ii) in Theorem 3. So
every sequence in T has a Cauchy subsequence. 2
Lemma 7: Let (M, d) be a complete metric space, A M any subset. Equiv-
alent are
(i) A has a compact closure.
(ii) Every sequence in A has a Cauchy subsequence.
Proof: (i)(ii) follows directly from the denitions.
(ii)(i). Let x
n
A be any sequence a
n
A so that d(x
n
, a
n
) <
1
n
.
Cauchy subsequence (a
n
i
)
i=1
. (x
n
i
)
i=1
is Cauchy. Because (M, d) is
complete (x
n
i
) converges (to another element of A). 2
Special case: Y = 1
n
, (X, d
X
) compact metric space. X = C(X, 1
n
) is a
normed vector space.
|f| := sup
xX
[f(x)[
R
n
Theorem 4: Let T C(X, 1
n
). Equivalent are
(i) T has a compact closure.
(ii) T is equicontinuous and bounded.
Proof: Theorem 3 (A subset of 1
n
has a compact clousre if and only if it is
bounded). So condition (ii) in Theorem 4 implies Condition (ii) in Theorem 3
with Y = 1
n
. Moreover an unbounded subset of C(X, 1
n
) cannot be compact.
2
Theorem 4: Let T C(X, 1
n
). Equivalent are
(i) T is compact.
(ii) T is closed, bounded and equicontinuous.
11
Proof: Corollary of Theorem 4. 2
This again highlights the dierence between nite and innite dimensional vec-
tor spaces, as far as compactness is concerned.
1.5 The Baire Category Theorem
Example for an open and dense set: 1. Let = x
1
, x
2
, . . .. Let
U :=
_
k=1
_
x
k

2
k
, x
k
+

2
k
_
Leb
(U)
k=1
2
2
k
= 2
U is open and dense.
Theorem 5: Let (M, d) be a complete metric space.
(i) If U
1
, U
2
, U
3
, . . . M is a sequence of open and dense subset then
D :=
i=1
U
i
is dense in M.
(ii) M ,= and A
1
, A
2
, A
3
, . . . M is a sequence of closed subsets so that
M =
_
i=1
A
i
Then i so that A
i
contains an open ball.
Example:
1. M = 1 =
xR
x and 1 complete; so 1 uncountable.
2. M = =
xQ
x is not complete.
The proof is not so hard. It depends on one ingenious observation which has
many important consequences.
Proof:
(i) Let x X and > 0. To show: B
(x) D ,= .
Let B := B
(x) = y M [ d(x, y) < . Since U

i
is dense B U
i
,= . Choose
x
1
B U
1
. B U
1
open
1
> 0,
1

1
2
so that
B
1
(x
1
) B U
1
Since U
2
is dense, B
1
(x
1
) U
2
,= . Choose x
2
B
1
(x
1
) U
2
. Because
B
1
(x) U
2
is open, x
2
> 0 so that
B
2
(x
2
) B
1
(x
1
) U
2
and 0 <
2

1
4
.
By Induction one gets a sequence
x
k
M, 0 <
k

1
2
k
so that
B
k
(x
k
) B
k1
(x
k1
) U
k
12
In particular x
k
B
k1
(x
k1
), i.e.
d(x
k
, x
k1
) <
k1

1
2
k1
So x
k
is a Cauchy sequence in M, which converges, because M is complete. Let
x
:= lim
k
x
k
. Note:
B
1
(x
1
) B
2
(x
2
) B
3
(x
3
) . . .
So x
k
(x
k
) k and thus x
k
(x
k
) U
k
k.
So x
D =
k=1
U
k
. Also x
i
(x
i
) B so B D ,= .
(ii) Let U
i
:= M A
i
, open. Suppose (by contradiction) that A
i
does not
contain any open ball for every i. So U
i
is open and dense. By (i)
i=1
U
i
is
dense; thus M
i=1
A
i
,= M ,=
i=1
A
i
. 2
Reminder Let A M, then A
=int(A) = x M [ > 0 such that B
(x)
A is the interior of A.
Denition:
Let (M, d) be a metric space. A M is called nowhere dense if A has
empty interior.
A M is said to be of 1st category in the sense of Baire if A =
i=1
A
i
,
where A
i
M is nowhere dense.
A M is said to be of 2nd category if it is not of the 1st category.
A M is called residual if M A is of the 1st category.
Notation: cat(A) = 1 or 2.
Example:
Z 1 is nowhere dense.
1 is of the 1st category.
Rules:
1. If A B: cat(B) = 1 cat(A) = 1
2. cat(A) = 2 cat(B) = 2
3. A =
i=1
A
i
, cat(A
i
) = 1 cat(A) = 1
Lemma 8: (M, d) complete metric space, R M. Equivalent are:
(i) R is residual
(ii) R
i=1
U
i
with U
i
open, dense.
Proof:
(i) (ii) R residual, A := M R cat(A) = 1
A =
_
i=1
A
i

_
i=1
A
i
where A
i
is nowhere dense. Then U
i
:= M A
i
is open and dense.
R = M A M
_
_
i=1
A
i
_
=
_
M A
i
_
=
U
i
13
(ii) (i) Assume U
i
M open, dense and
i=1
U
i
R
A := M R R = M
U
i
=
_
(M U
i
)
with M U
i
nowhere dense i N cat(A) = 1.
2
Theorem 6 (Baire Category Theorem): Let (M, d) be a complete metric
space and M ,= . Then
(i) cat(M) = 2
(ii) If A M, then cat(A) = 1 cat(M A) = 2
(iii) If A M, then cat(A) = 1 M A is dense
(iv) ,= U M open Cat(U) = 2
(v) A =
i=1
A
i
, A
i
closed with A
i
= A
=
(vi) U =
i=1
U
i
with U
i
M open and dense U is dense in M
Proof:
(i) Suppose cat(M) = 1 M =
i=1
A
i
=
i=1
A
i
A
i
is nowhere dense (by Thm 5) one of the A
i
contains an open ball.
Contradiction.
(ii) cat(M A A) = cat(M) = 2 cat(M A) = 2 using the above rules.
(iii) Lemma 8 and Theorem 5 (ii)
(iv) U M open, nonempty U contains an open ball MU is not dense.
cat(U) = 2
(v) A =
i=1
A
i
closed A
i
= cat(A) = 1 (by (iv)) A does not contain
an open ball A
=
(vi) Theorem 5 (i)
2
Exercise: Even if (M, d) is not complete, we have (iii) (iv)(v) (vi) An
Application of Baires theorem is the following
Theorem 7 (Banach 1931):
:= f : [0, 1] 1 [ f is continuous and nowhere dierentiable
is residual in C([0, 1]).
Proof: Denote
U
n
:=
_
_
f C([0, 1])
sup
0<]h]1
t+h[0,1]
f(t +h) f(t)

h
> n t [0, 1]
_
_
Claim 1
i=1
U
n
Claim 2 U
n
is open . Exercise.
14
Claim 3 U
n
is dense.
Proof of Claim 3 Fix n N. Let g C([0, 1]) and > 0.
To show: B
(g) U
n
,=
By Weierstrass there is a polynomial p : 1 1 such that
|g p| = sup
t[0,1]
[g(t) p(t)[ < /2
We must nd an f U
n
such that |f p| < /2.
Trick: Dene z : 1 1 by z
(t) := z(
t
2
)
[ z(t)[ = 1 and [z
(t)[ =
1
Idea: Choose f(t) = p(t) +z
(t) then |f p| = |z
[ =
f(t +h) f(t)

h
z(t +h) z(t)

h
. .
if h is small
p(t +h) p(t)

h
. .
sup
t[0,1]
[
d
dt
p(t)[=:c
c > n if is small f U
n
Proof of Claim 2 See Zehnders notes. 2
1.6 Dual spaces
Let (X, | |) be a Banach space. Three examples for dual spaces:
Example 1: If X = H is a Hilbert space, i.e. there is an inner product
H H 1 : (x, y) x, y) so that
|x| =
_
x, x)
Each x H determines a bounded linear functional
x
: H 1 via
(1)
x
(y) := x, y)
The map H H
: x
x
is a Banach space isometry, i.e. a bilinear map
preserving the norms, so H

= H
. The dicult part of the proof is that (1) is

onto (Proof in Measure and Integration).
Example 2: Let (M, /, ) be a -nite measure space and
X = L
p
() =
_
f : M 1
_
M
[f[
p
d <
__
and
|f|
p
=
_
_
_
M
[f[
p
d
_
_
1
p
, 1 p <
In the measure and integration course it was shown that
X

= L
q
(), 1 < q ,
1
p
+
1
q
= 1
More precisely the map
L
q
() L
p
()
: g
g
g
(f) :=
_
M
fg d
15
is a Banach space isometry.
Again its easy to prove that
|g|
q
= |
g
| = sup
0,=fL
p
_
M
fg d
|f|
p
and the hard part is that
L
p
()
: g L
q
() so that =
g
.
Proof in Measure and Integration.
Example 3: Let (M, d) be a compact metric space. Consider X = C(M) =
f : M 1 [ f is continuous. Let
|f| := sup
pM
[f(p)[
That X is a Banach space is already known from Analysis I & II.
X
= All real Borelmeasures on M =: /

Let B 2
M
be the Borel -Algebra and a -additive : B 1 a real (Borel)
measure.
Dene
: C(M) 1 by
(f) :=
_
M
f d
Easy:
is bounded and |
| = || = [[(M).
The map / C(M)
is linear. But why is this map surjective?

Exercise with Hints:
1. | M open | is -compact.
U =
_
n=1
K
n
, K
n
:= x M [ B1
n
(x) U
2. Every nite Borel measure : B [0, ) is a Radon measure because of
1.
3. Riesz Representation Theorem
: C(M) 1
positive linear functional, i.e. if f C(M) and f > 0 (f) > 0, e.g.
(f) =
_
M
f d
4. For every bounded linear functional : C(M) 1 there are two positive
linear functionals
: C(M) 1, s.t. =
+
.
Hint: For f > 0 dene
(f) := sup(f
1
) (f
2
) [ f
1
+f
2
= f, f
1
, f
2
C(M), f
1
0, f
2
0
(f) 1
+
0
. Claim: (f +g) = (f) +(g).
For f : M 1 continuous dene f
(x) := maxf(x), 0 f = f
+
, f
continuous and nonnegative.

Dene (f) := (f
+
) (f
).
To show: : C(M) 1 is bounded and linear.
f 0 (f) [(f)[,
:=
1
2
( )
are positive.
16
Denition: Let (M, d) be a metric space. A completion of (M, d) is triple
(M
, d
, ) where
1. (M
, d
) is a complete metric space

2. : M M
is an isometric embedding i.e.

d
((x), (y)) = d(x, y) x, y M

3. The image (M) is dense subset of M
.
Denition: (M
1
, d
1
), (M
2
, d
2
) metric spaces
A map : M
1
M
2
is called an isometry, if it is bijective
and d
2
((x), (y)) = d
1
(x, y) x, y M.
Theorem 8:
(i) Every metric space (M, d) admits a completion.
(ii) If (M
1
, d
1
,
1
) and (M
2
, d
2
,
2
) are completions of (M, d), then there exists
a unique isometry : M
1
M
2
such that
1
=
2
M
1
//
B
B
B
B
B
B
B
B
M
2
Proof:
(i) Uniqueness Exercise, the standard uniqueness proof for objects with
universal property.
(ii) Existence
Construction 1 M
:= Cauchy sequence in M/
(x
n
) (y
n
) lim
n
d(x
n
, y
n
) = 0
(x) := [(x
n
)] where x
i
= x i N
d
([(x
n
)], [(y
n
)]) := lim
n
d(x
n
, y
n
).
See Topology lecture.
Construction 2 Let BC(M, 1) := f : M 1 [ f is continuous and bounded
and |f| = sup
xM
[f(x)[.
Fact: BC(M, 1) is a banach space.
Fix a point x
M. For every x M dene the function f

x
: M 1 by
f
x
(y) := d(x, y) d(x
, y)
(a) f
x
is continuous.
(b) f
x
is bounded because [d(x, y) d(x
, y)[ d(x, x
).
(c) i : M BC(M, 1) : x f
x
is an isometric embedding.
For all x, x
M we have:
d(f
x
, f
x
) = |f
x
f
x
| (6)
= sup
yM
[f
x
(y) f
x
(y)[ (7)
= sup
yM
[d(x, y) d(x
, y)[ (8)
= d(x, x
) (set y = x
) (9)
Now dene M
:= closure(f
x
[ x M) in B(M, 1), d(f, g) :=
|f g| and (x) := f
x
.
2
17
Exercise: The completion of a normed vector space is a Banachspace. Hints:
Let (X, | |) be a normed vector space and dene the metric on X
by d(x, y) := |x y| x, y X
1. Let (X
, d
, ) be a completion of (X, d). Then there is a unique pair of

vector space structure and norm on X
such that
(i) X X
is linear
(ii) |(x)|
= |x| x X
(iii) d
(x
, y
) = |x
, y
2. If (X
1
, | |
1
,
1
) and (X
2
, | |
2
,
2
) are two completions of (X, | |) then
the isometry : X
1
X
2
in Theorem 9(ii) is linear.
Example: X = C([0, 1]) f
|f| :=
_
_
1
_
0
[f(t)[
p
dt
_
_
1
p
1 p <
The completion of (X, | |
p
) is L
p
([0, 1]) with respect to the Lebesgue measure
on [0, 1].
More general: Replace [0, 1] by a locally compact Hausdor space M and
Lebesgue by a Radonmeasure.
Exercise: (X, | |) normed vector space.
The functions X 1 : x |x|, X X X : (x, y) x + y and
1 X X : (, x) x are continuous.
Let M be any set. Then B(M, X) = f : M X [ f is bounded
with |f| := sup
xM
[f(x)[ < is a normed vector space.
X complete B(M, X) is complete.
(M, d) metric space BC(M, X) := f : M X [ f is continuous and bounded
is a closed subspace of B(M, X). Recapitulation:
1. Any two norms on a nite dim. vector space are equivalent (Lemma 1).
2. Every nite dimensional normed vector space is complete (Lemma 2).
3. Every nite dimensional subspace of a normed vector space is closed
(Lemma 3).
4. A normed vector space (X, | |) is nite dimensional if and only if the
unit ball B := x X [ |x| 1 (resp. the unit sphere S := x X [
|x| = 1) is compact (Theorem 1).
5. (X, | |), (Y, | |) normed vector space A : X Y linear operator.
X is nite dimensional Every linear operator A : X Y is bounded
(Lemma 6).
Denition: L(X, Y ) := A : X Y [ A is linear and bounded
|A| := sup
xX
x,=0
|Ax|
Y
|x|
X
Theorem 9: Let X, Y, Z be normed vector spaces.
(i) L(X, Y ) is a normed vector space.
(ii) Y complete L(X, Y ) is complete.
(iii) A L(X, Y ), B L(Y, Z) BA L(X, Z) and
|BA| |B| |A| ()
Moreover the map L(X, Y ) L(Y, Z) L(X, Z), (A, B) BA is con-
tinuous.
18
Proof:
(i) Verify axioms.
(iii) |BAx|
Z
|B| |Ax|
Y
|B| |A| |x|
X
This implies (). Moreover:
|B
2
A
2
B
1
A
1
|
()
|B
2
B
1
| |A
2
| +|B
1
| |A
2
A
1
|
Now do , . . ..
(ii) Assume Y is complete. Let (A
n
)
nN
be a Cauchy sequence in L(X, Y ).
|A
n
x A
m
x|
Y
|A
n
A
m
||x|
X
This shows: For each x X the sequence (A
n
x)
nN
is a Cauchy sequence
in Y .
Because Y is complete the sequence (A
n
x)
nN
converges for every x X.
Dene A : X Y by Ax := lim
n
A
n
x A is linear.
Claim : A is bounded and A
n
converges to A in L(X, Y ).
Proof : Let > 0. There n
0
N such that
m, n N : n, m n
0
|A
n
A
m
| <
Hence for n, m n
0
:
|A
n
x A
m
x|
Y
= |A
n
x lim
n
A
m
x|
Y
= lim
n
|A
n
x A
m
x|
Y
limsup
n
|A
n
A
m
||x|
X
|x|
X
So
|Ax|
Y
|Ax A
n
x|
Y
+|A
n
x|
Y
|x|
X
+|A
n
||x|
X
= ( +|A
n
|)|x|
X
So A is bounded and |A| |A
n
| +, moreover
|A
n
A| := sup
x,=0
|A
n
x Ax|
Y
|x|
X

2
Example: Y = 1
X
:= L(X, 1) is a Banach space with the norm |A| := sup

x,=0
]Ax]
|x|
X
X
is called the dual space of X.

Example: (L
p
)
= L
q
where 1 p < and
1
p
+
1
q
= 1. See the Measure and
Integration lecture for the proof.
Theorem 10: Let X be a normed vector space, Y Banach space.
Let (A
i
)
iN
be a sequence of bounded linear operators such that
i=1
|A
i
| <
Then the sequence S
n
:=
n
i=1
A
i
converges in L(X, Y ). The limit is denoted
by
i=1
A
i
= lim
n
n
i=1
A
i
19
Proof: s
n
:=
i=1
|A
i
| < converges in 1.
|S
n
S
m
| = |
n
i=m+1
A
i
|
n
i=m+1
|A
i
| = s
n
s
m
S
n
is a Cauchy Sequence in L(X, Y ).
Thm 9
S
n
converges. 2
Example: X = Y Banach space, L(X) := L(X, X)
Suppose f(z) =
i=0
a
i
z
i
is a convergent power series with convergence radius
R :=
1
limsup
n
[a
n
[
1
n
> 0
Let A L(X) be a bounded linear operator with |A| < R. Then
i=0
[a
i
[|A
i
|
i=0
[a
i
[|A|
i
<
Thm 10
the limit f(A) :=
i=0
a
i
A
i
= lim
n
n
i=0
a
i
A
i
exists.
Remark: Works also with a
i
C if X is a complex Banach space.
Example: f(z) =
i=0
z
i
=
1
1z
Corollary: |A| < 1 1 A is bijective with inverse
(1 A)
1
=
i=0
A
i
L(X)
Proof: S
n
:= 1 +A+A
2
+... +A
n
|A| < 1
Thm 10
The sequence S
n
converges.
S
:= lim
n
S
n
=
i=0
A
i
S
n
(1 A) = (1 A)S
n
= 1 +A+A
2
+. . . +A
n
A. . . A
n+1
= 1 A
n+1
1
S
(1 A) = (1 A)S
= 1 2
Theorem 11: X Banach space, A L(X)
(i) The limit r
A
:= lim
n
|A
n
|
1
n
= inf
n>0
|A
n
|
1
n
|A| exists.
(Its called the Spectral radius of A.)
(ii) r
A
< 1
i=0
|A
i
| < and
i=0
A
i
= (1 A)
1
Proof:
(i) Let := inf
n>0
|A|
1
n
. Let > 0.
m N |A
m
|
1
m
< +
c := max1, |A|, . . . , |A
m1
|
Write an integer n > 0 in the form n = km+l k N
0
, l 0, 1, . . . m1
|A
n
|
1
n
= |(A
m
)
k
A
l
|
1
n
(10)
|A
m
|
k
n
|A
l
|
1
n
(11)
c
1
n
( +)
km
n
(12)
= c
1
n
( +)
1
l
n
(13)
n
+ (14)
20
n
0
N n n
0
: |A
n
|
1
n
+ 2
lim
n
|A
n
|
1
n
=
(ii) r
A
< 1. Choose 1 : r
A
< < 1 n
0
N n n
0
: |A
n
|
1
n
<
|A
n
| <
n
i=0
|A
i
| <
Thm 10
(ii)
2
1.7 Quotient spaces
Denition: Let X normed vector space, Y X closed subspace.
x +Y := x +y [ y Y X
x +Y = x
+Y x
x Y : x x
X/Y := X/
= x +Y [ x X
Notation: [x] := x +Y for the equivalence class of x X
Remark: X/Y is a normed vector space with
|[x]|
X/Y
:= inf
yY
|x +y|
X
Exercise:
1. | |
X/Y
is a norm
2. X Banach space, Y closed subspace X/Y is a Banach space.
21
2 Functional Analysis 15.11.2006
2 Functional Analysis
2.1 Basics
Theorem 1 (Uniform Boundedness): Let X be a Banach space, Y a
normed vector space and I an arbitrary set. Let A
i
L(X, Y ) for i I and
assume x X : sup
iI
|A
i
(x)| < .
The conclusion says that c > 0 such that sup
iI
|A
i
x| c x X with
|x| 1.
Lemma 1: (M, d) complete metric space M ,= , I any set. f
i
: M 1
continuous for i I. Assume sup
iI
[f
i
(x)[ < x M open ball B M
such that
sup
xB
sup
iI
[f
i
(x)[ <
Proof of Lemma 1: Denote
A
n,i
:= x M [ [f
i
(x)[ n for n N and i I.
A
n
:=
iI
A
n,i
= x M [ sup
iI
[f
i
(x)[ n
x M n N such that x A
n
, i.e. M =
n=1
A
n
.
Now A
n,i
= f
1
i
([n, n]) is closed. So A
n
is closed. So n N such that
int(A
n
) ,=
x
0
int(A
n
)
> 0 such that
B
(x
0
) = x M [ d(x, x
0
) < A
n
2
Proof of Theorem 1: Set M := X, f
i
(x) := |A
i
x|, so f
i
: X
A
i
Y
||
1.
So f
i
is continuous for every i I.
sup
iI
[f
i
(x)[ < x X
Lemma 1
= ball B = B
(x
0
) X with x
0
X, > 0
such that
c := sup
iI
sup
xB
|A
i
x| <
i I x X we have |x x
0
| |A
i
x| c.
Let x X with |x| = 1.
Then |(x
0
+ x) x
0
| =
Hence |A
i
(x
0
+x)| c so
|A
i
x| =
1
|A
i
(x
0
+x) A
i
x
0
|
1
|A
i
(x
0
+x)| +
1
|A
i
x
0
|
c +c
2
Theorem 2 (Banach-Steinhaus): X Banach space, Y normed vector space
A
i
L(X, Y ), i = 1, 2, 3, . . .
(i) Assume the sequence (A
i
x)
i=1
converges in Y for every x X. Then:
sup
iN
|A
i
| <
A L(X, Y ) such that Ax = lim
i
A
i
x, |A| liminf
i
|A
i
|
(ii) Assume Y is complete and
sup
iN
|A
i
| <
dense subset D X such that (A
i
x)
i=1
converges for every x D
Then (A
i
x)
i
converges for all x X
22
Proof:
1. Since (A
i
x)
i
converges we have
sup
iN
|A
i
x| < x X sup
iN
|A
i
| <
Dene A : X Y by Ax := lim
i
A
i
x. This operator is linear. Why is
A bounded?
|Ax| = lim
i
|A
i
x| = liminf
i
|A
i
x| liminf
i
|A
i
|
. .
<
|x|
2. Let x X. Need to show that (A
i
x)
i=1
is Cauchy. Let > 0.
Denote c := sup
iN
|A
i
| < .
Choose y D such that |x y| <

4c
.
Choose n
0
N so that i, j n
0
: |A
i
A
j
| <

2
i, j n
0
: |A
i
x A
j
x| |A
i
x A
i
y| +|A
i
y A
j
y| +|A
j
y A
j
x|
< |A
i
||x y| +|A
i
A
j
||y| +|A
j
||x y|
<

4
+

2
+

4
<
2
Example 1: l
= bounded sequences 1, x l
with x = (x
1
, x
2
, x
3
, . . .) =
(x
i
)
i=i
.
X := x = (x
i
)
i
l
[ n N such that x
i
= 0 i n
Dene
A
n
: X X by A
n
x = (x
1
, 2x
2
, 3x
3
, . . . , nx
n
, 0, . . .)
lim
n
A
n
x = Ax where Ax = (x
1
, 2x
2
, 3x
3
, . . .).
But |A
n
| = n . Completeness of the domain is missing here.
So the assumption that X is complete cannot be removed in Theorem 1 or
Theorem 2.
Example 2: X Banach space, Y , Z normed vector spaces and B : XY Z
bilinear. Equivalent are:
(i) B is continuous
(ii) The functions X Z : x B(x, y) is continuous y Y and
the function Y Z : y B(x, y) is continuous x X.
(iii) c > 0 x X y Y : |B(x, y)| c|x| |y|
This is exercise 2 on Sheet 4.
Theorem 3 (Open Mapping Theorem): X, Y Banach spaces. A L(X, Y )
surjective A is open, i.e. if U X is an open set then AU Y is open.
Corollary (Inverse Operator Theorem): X, Y Banach spaces, A L(X, Y )
bijective A
1
is bounded, i.e. A
1
L(X, Y ).
Proof: A open A
1
continuous A
1
bounded. 2
23
Example 3: X as in example 1; X is not complete. Dene B : X X by
Bx := (x,
1
2
x
2
,
1
3
x
3
, . . .). Then |Bx| |x|, so B bounded. B
1
= A, as in
example 1, is not bounded.
Lemma 2: X, Y Banach spaces, A L(X, Y ) surjective
> 0 such that y Y [ |y| < Ax [ x X, |x| < 1 (*)
Remark: () means y Y x X such that Ax = y and |x|
1
|y| ()
Exercise: () ()
Use () to prove the Corollary.
Proof (Lemma 2 Theorem 3): Let U X be open, and y
0
AU
x
0
U such that y
0
= Ax
0
(U open)
= > 0 such that B
(x
0
) U.
Claim: B
(y
0
) AU. Let
y B
(y
0
)
_
_
_
_
y y
0
_
_
_
_
<
x X such that |x| < 1 and Ax =
yy
0
x
0
+x B
(x
0
) U
A(x
0
+x) = y
0
+Ax = y y AU. 2
Proof of Lemma 2:
Step 1 r > 0 so that
y Y [ |y| < r Ax [ x X, |x| < 1
Proof of Step 1:
Let
B := x X [ |x| <
1
2
C := AB = Ax [ x X, |x| <
1
2
Note that
1. nC = ny [ y C = Ax [ x X, |x| <
1
2
2. y, y
C y y
2C
3. y, y
C y y
2C
4. nC = nC
X =
_
n=1
nB
A onto
Y = AX
=
_
n=1
nAB
=
_
n=1
nC
=
_
n=1
nC
By Baire: n N : (nC)
,= (C)
,=
y
0
Y r > 0 : B
r
(y
0
) C
24
So if y Y and |y| < r then y
0
+y C.
Thus y Y with |y| < r we have y = y
0
+y
. .
C
y
0
..
C
2C
y Y [ |y| < r 2C
Step 2 () holds with =
r
2
. Proof:
Let y Y with |y| <
r
2
.
To Show: x X so that Ax = y and |x| < 1.
Denote
B
k
:= x X [ |x| <
1
2
k
k = 1, 2, 3, . . .
Then, by Step 1, ()
_
y Y [ |y| <
r
2
k
_
AB
k
k = 1, 2, 3, . . .
Since y Y and |y| <
r
2
, by () with k = 1:
x
1
X : |x
1
| <
1
2
, |y Ax
1
| <
r
4
and by () with k = 2
x
2
X : |x
2
| <
1
4
, |y Ax
1
Ax
2
| <
r
8
So, by induction, using (), there is a sequence (x
k
)
k
X so that
|x
k
| <
1
2
k
|y Ax
1
. . . Ax
k
| <
r
2
k+1
We have
k=1
|x
k
| <
k=1
1
2
k
= 1
By Chapter 1, the limit
x := lim
n
n
i=1
x
i
=
i=1
x
i
exists and |x| < 1. Since
_
_
_
_
_
y A
k
i=1
x
i
_
_
_
_
_
<
r
2
k

we have proved Lemma 2. 2
2.2 Product spaces
Example 4: Let X be a Banach space and X
1
, X
2
both closed linear sub-
spaces. Assume X = X
1
+ X
2
and X
1
X
2
= 0; these subspaces are called
transverse subspaces. We say X is the direct sum of X
1
and X
2
and write
X = X
1
X
2
Every vector in X can be written as sum of a vector in X
1
and one in X
2
in a
unique way (Linear Algebra).
Dene A : X
1
X
2
X by A(x
1
, x
2
) := x
1
+ x
2
. If X, Y are normed vector
spaces, then X Y := (x, y) [ x X, y Y is again a normed vector space
with
|(x, y)| := |x| +|y|
for (x, y) X Y . Other possibilities are
|(x, y)|
:= max|x|, |y|
|(x, y)|
p
:= (|x|
p
+|y|
p
)
1
p
, 1 p
25
All these norms are equivalent.
X, Y Banach spaces X Y is a Banach space for any of these norms.
These are exercises.
Return to Example 4: X
1
, X
2
are closed subsets of a complete space, hence
complete. So X
1
X
2
is complete by the exercise above and A is a operator
between Banach spaces.
A is bounded and linear, because
|A(x
1
, x
2
)| = |x
1
+x
2
| |x
1
| +|x
2
|
A is surjective because X = X
1
+X
2
A is injective because X
1
X
2
= .
By the open mapping theorem (Theorem 3) A
1
is bounded
c > 0 : x
1
X
1
x
2
X
2
: |x
1
| +|x
2
| c|x
1
+x
2
|
So the projections
1
: X X
1
,
2
: X X
2
are bounded.
Example 5: X = Y = C([0, 1]) with supnorm.
Ax = x A is only dened on a subset of X namely on
D := x X [ x is continuously dierentiable =: C
1
([0, 1])
D X, A : D Y .
Denition: Let X, Y be Banach spaces.
D X linear subspace, a linear operator A : D Y is called closed if its graph
= graph(A) := (x, Ax) [ x D is a closed subspace of X Y , i.e. for any
sequence (x
n
)
nN
in D and (x, y) X Y we have:
x
n
x
Ax
n
y
_
x D and y = Ax
Example 5:
x
n
C
1
([0, 1]) lim
n
sup
0t1
[x
n
(t) x(t)[ = 0 and
x, y C([0, 1]) lim
n
sup
0t1
[ x
n
(t) y(t)[ = 0
x C
1
and x = y, so the operator in Example 5 is closed.
Exercise: The graph norm on D is dened by |x|
A
:= |x|
X
+|Ax|
Y
Prove that (D, | |
A
) is complete if and only if A is closed.
Example 5: The graph norm on C
1
([0, 1]) is
|x|
A
= sup
0t1
[x(t)[ + sup
0t1
[ x(t)[
The standard norm in C
1
.
What if D = X?
Theorem (Closed Graph Theorem): X, Y Banach spaces
A : X Y linear operator. Equivalent are:
(i) A is bounded
(ii) A has a closed graph
26
Proof:
(i) (ii) X x
n
x and Y Ax
n
y
A continuous
= Ax
n
Ax
Uniqueness of the limit
= Ax = y
(ii) (i) := graph(A) X Y, is a Banach space.
Dene : X by (x, y) = x
is a bounded linear operator with norm = 1.
is injective.
is surjective (because D X).
Thm 3
=
1
: X is bounded c > 0 such that |
1
(x)|
A
c|x|
X
But |
1
(x)|
A
= |(x, Ax)|
A
= |x|
X
+|Ax|
Y
|Ax|
Y
c|x|
X
x X. 2
Example 6 (Hellinger-Toeplitz-Theorem): H Hilbert space
A : H H linear operator which is symmetric,
i.e. x, Ay) = Ax, y) x, y H A is bounded.
Proof: To show: A is closed.
H x
n
sequence. Assume x
n
x H,
Ax
n
y H.
To show: Ax = y.
y, z) = lim
n
Ax
n
, z) = lim
n
x
n
, Az) = x, Az) z H
y Ax, z) = 0 z H
z=yAx
|y Ax| = 0 y = Ax 2
Example 5: A : D Y is closed but not bounded:
x
n
(t) = t
n
|x
n
| = sup
t[0,1]
[x
n
(t)[ = 1
|Ax
n
| = | x
n
| = sup
t[0,1]
[ x
n
(t)[ = n
Denition: A : D X Y is called closable, if there is an operator A
:
D
Y D D
such that A
is closed and A
[
D
= A.
Remark: Let := (x, Ax) [ x D := graph(A)
A is closable
is the graph of a closed operator
: X is injective
D x
n
0, Ax
n
y implies y = 0
Example 7: Let X = L
2
([0, 1]), D = C([0, 1]), Y = 1. Let A : D Y, x
x(0) is not closable.
Example 8: X = L
2
(1) D = x L
2
(1) [ c > 0 [t[ > c : x(t) = 0
Y = 1 Ax =
_
x(t)dt
x
n
(t) :=
_
1
n
[t[ n
0 [t[ > n
|x
n
|
2
L
2
=
2
n
Ax
n
= 2
Example 9: Every dierential operator is closable.
Let 1
n
be an open subset.
C
0
= f : 1 [ f is smooth, supp(f) is compact
27
with
supp(f) := x [ x
n
such that f(x
n
) ,= 0, x
n
x
= cl
(x [ f(x) ,= 0)
C
0
() L
p
(). We know:
1. C
c
() = f : 1 [ f continuous, f has cpct support is dense in
L
p
().
2. C
0
() is dense in C
c
(), i.e.
f C
c
() K , K compact f
n
C
0
()
such that
supp(f
n
) K, lim
n
sup
x
[f
n
(x) f(x)[ = 0
Let X = Y = L
p
().
Let D := C
0
() X.
Dene A : D Y by (Af)(x) =
]]m
a
(x)
f(x)
where = (
1
, . . . ,
n
) Z
n
,
i
0 and [[ :=
1
+. . . +
n
.
:=

1
x
1
1
. . .

n
x
n
n
Let a
: 1 be smooth.
Claim A is closable.
Proof: Integration by parts: f, g, C
0
()
g(Af) d =
g a
f d
=
(1)
]]
_
(a
g) f d
=
_
(1)
]]
(a
g)
_
f d
=
_
(Bg) f d
Let (0, g) Graph(A) L
p
() L
p
()
To show: g = 0
f
k
C
0
() sequence such that (f
k
, Af
k
)
L
p
L
p
(0, g), i.e.
lim
k
|f
k
|
L
p = 0 = lim
k
|Af
k
g|
L
p
C
0
():
_
g d = lim
k
_
(Af
k
) d
= lim
k
_
(B) f
k
d
= 0
28
because

(B)f
k
d
|B|
L
q
. .
<
|f
k
|
L
p
. .
0
where
1
p
+
1
q
= 1 for 1 < p < .
So
_
g d = 0 C
0
()
. .
dense in L
q
()
g d = 0 L
q
() g = 0 almost everywhere
g d =
_
[g[
p
d
:= (sign(g))[g[
p1
L
q
() 2
2.3 Extension of bounded linear functionals
Theorem 5 (Hahn-Banach): X normed vector space over F = 1 or C. Y
X linear subspace : Y F linear and c > 0 such that [(y)[ c|y| y Y
: X F linear such that
1. [
Y
=
2. [(x)[ c|x| x X.
Question: If we replace the target F by another Banach space Z over F, i.e.
: Y Z bounded linear operator. ? : X Z bounded linear operator,
[
Y
= .
Answer: No!
Example: X = l
, Y := c
0
=: Z, F = 1, = id : Y Z does not extend!
Lemma 3: Let X, Y, , c as in Theorem 5 with F = 1.
Let x
0
X Y and denote
Z := Y 1x
0
= y +x
0
[ y Y, 1
Then : Z 1 linear so that
a. [
Y
=
b. [(x)[ c|x|x Z
Proof: Need to nd a number a 1 so as to dene
(x
0
) := a (1)
Then
(y +x
0
) = (y) +a y Y and 1 (2)
is well-dened by (2), because x
0
/ Y . Moreover [
Y
= . To show: a can
be chosen such that
[(y) +a[ c|y +x
0
| y Y 1 (3)
(3) is equivalent to
[(y) +a[ c|y +x
0
| y Y (4)
29
(3) (4) is obvious.
(4) (3) ok for = 0.
,= 0 : [(y) +a[ = [[ [
_
y
_
+a[
(4)
c[[|
y
+x
0
| = c|y +x
0
|
(4) is equivalent to
c|y +x
0
| (y) +a c|y +x
0
|y Y
(y) c|y +x
0
| ay Y a c|y +x
0
| (y)y Y
(y
) c|y
x
0
| y
Y a c|y +x
0
| (y)y Y (5)
Is there a real number a 1 such that (5) holds, i.e.
sup
y
Y
((y
) c|y
x
0
| a inf
yY
(c|y +x
0
| (y))
This is true i
(y
) c|y
x
0
|
(6)
c|y +x
0
| (y)y, y
Y (6)
2
Proof: of (6)
(y)+(y
) = (y+y
) c|y+y
| = c|y+x
0
+y
x
0
| c|y+x
0
|+c|y
x
0
|
2
Denition: Let T be a set. A partial order on T is a relation (i.e. a subset
of T T, we write a b instead of (a, b) .)
That satises:
is reective, i.e. a aa T
is transitive, i.e. a, b, c T we have a b, b c a c
is anti-symmetric, i.e. a, b T: a b, b a a = b
Denition: (T, ) partially ordered set (POS). A subset ( T is called a
chain if it is totally ordered, i.e.
a, b ( a b or b a
Denition: (T, ) POS, ( T, a T a is called the supremum of ( if
1. c ( : c a
2. b T : (c b c ( a b)
Denition: (T, ) POS, a T. a is called a maximal element of T if b T
we have a b b = a
Lemma Zorns Lemma: Let (T, ) be a POS such that every chain ( T
has a supremum. Let a T There exists a maximal element b T such that
a b.
Remark: Zorns Lemma is equivalent to the axiom of choice.
30
Proof of Theorem 5: Let X, Y, , c be given as in Theorem 5.
Dene
T := (Z, ) [ Z X linear subspace,
Y Z,
: Z 1 linear,
[
Y
= ,
[(x)[ c|x|x Z
(Z, ) (Z
) : Z Z
[
Z
:=
Note that this is a partial order and every chain ( T has a supremum Z
0
:=
(Z,)c
Z x.
(Z, ) T is maximal Z = X by Lemma 3
(Y, ) T maximal element
(X, ) T such that (Y, ) (X, )
2
Proof of Theorem 5 for F = C: X complex normed vector space, Y X
complex linear subspace
: Y C such that [(y)[ c|y| y Y
Fact: : Y C is complex linear
is real lin. and (iy) = i(y) y Y (1)
Write (y) =
1
(y) +i
2
(y) where
1
(y),
2
(y) 1
Then
1
,
2
: Y 1 real linear and
i(y) = i
1
(y)
2
(y) (iy) =
1
(iy) +i
2
(iy)
satises (1)
2
(y) =
1
(iy)
1
(y) =
2
(iy) y Y (2)
We have [
1
(y)[ [(y)[ c|y|
Thm 5,F=R

1
: X 1 1-linear
1
[
Y
=
1
[
1
(x)[ c|x| x X
Dene : X C by (x) :=
1
(x) i
1
(ix)
1. is complex linear
2. If y Y then (y) =
1
(y) i
1
(iy) = (y)
3. Let x X. Suppose (x) ,= 0
Then
(x)
](x)]
S
1
= z C [ [z[ = 1
so 1 such that e
i
=
(x)
](x)]
(e
i
x) = e
i
(x) = [(x)[ 1
[(x)[ = [e
i
(x)[ = [(e
i
x)[ = [
1
(e
i
x)[ c|e
i
x| = c|x|
2
Denition: X real vector space. A function p : X 1 is called seminorm if
1. p(x +y) p(x) +p(y) x, y X
2. p(x) = p(x) > 0, x X
Theorem 5: X real vector space, p : X 1 seminorm
Y X linear subspace, : Y :1 linear
Then linear map : X 1 such that [
Y
= and (x) p(x) x X
31
Sketch of Proof Theorem 5: As in Lemma 3: Z := Y 1x
0
, x
0
/ Y
(y
) +(y) = (y +y
) p(y +y
) p(y +x
0
) +p(y
x
0
)
(y
) p(y
x
0
) p(y +x
0
) (y) y, y
Y
a 1 such that (y) p(y x
0
) a p(y +x
0
) (y) y Y
(y) p(y x
0
) a p(y +x
0
) (y) y Y, > 0
(y) a p(y x
0
) and (y) +a p(y +x
0
) y Y, > 0
(y) +ta
. .
:=(y+tx
0
)
p(y +tx
0
) t 1, y Y
i.e. (x
0
) = a so : Z 1[
Y
= (x) p(x) x Z
For the remainder of the proof, argue as in Theorem 5 using Zorns lemma. 2
Remark: Theorem 5 implies Theorem 5 with F = 1, p(x) = c|x|
Notation: X normed vector space over F = 1, C X
:= L(X, F)
Each element of X
is a bounded F-linear functional : X F. We write:

x
instead of : X F
x
, x) F instead of (x)
Remark: Theorem 5 says, if Y X is a linear subspace and y
then
x
such that x
[
Y
= y
and |x
|
X
= |y
|
Y
Denition: Y X linear subspace of a normed vector space X.

The annihilator of Y is the (closed) subspace Y
dened by
Y
:= x
[ x
, y) = 0 y Y
Exercise:
1. Y

= X
/Y
2. (X/Y )

= Y
if Y is closed
3. For Z X
, dene

Z := x X [ x
, x) = 0 x
Z
Prove that

(Y
)

= Y whenever Y is a closed subspace of X.
Theorem 6: X normed vector space, A, B X convex, int(A) ,= , B ,= ,
A B =
x
, c 1 such that x
, x) < c x A and x
, x) c x B
Proof:
Case 1: B = 0 Let x
0
int(A) and dene p : X 1 by
p(x) := inft > 0 [ x
0
+
x
t
A
So x
0
+
x
t
A for t > p(x) and x
0
+
x
t
/ A for t < p(x)
1. p(x) = p(x) > 0
2. p(x +y) p(x) +p(y)
Given > 0 s, t > 0 : s p(x) +, t p(x) +
x
0
+
x
s
A x
0
+
y
t
A
x
0
+
x+y
s+t
=
s
s+t
_
x
0
+
x
s
_
+
t
t+s
_
x
0
+
y
t
_
A (A convex)
p(x +y) s +t p(x) +p(y) + 2 > 0
3. Choose > 0 such that B
(x
0
) A so p(x)
|x|
x X
32
4. x
0
+x A p(x) 1 x
0
+x / A p(x) 1
Choose Y := 1x
0
(x
0
) :=
Check: 1 = (x
0
) 0 p(x
0
) 1 = (x
0
) p(x
0
) (because 0 / A)
Thm 5
: X 1 such that (x) p(x) x X, (x
0
) = 1
(i.e. [
Y
= ) and x A (x) 0 :
Namely x A p(x x
0
) 1
So if x A then (x x
0
) p(x x
0
) 1 = (x
0
) (x) 0
Assertion with x
= , c = 0
_
is bounded by 3.: (x) p(x)
|x|
_
Case 2: A arbitrary
K := a b [ a Ab B K convex, int(K),= , 0 / K
Case 1
x
such that x
, x) 0 x K
x
, a) x
, b) a A, b B
c := sup
aA
x
, a) < (because B ,= )
x
, a) c x
, b) a A, b B 2
Theorem 7: X normed vector space over F = 1, C
Y X linear subspace, x
0
X Y
Let := d(x
0
, Y ) = inf
yY
|x
0
y| > 0
x
such that |x
| = 1, x
, x
0
) = , x
, y) = 0 y Y
Note: Hypotheses of Theorem 6 are satised with A = B
(x
0
), B = Y
Proof: Denote Z := y +x
0
[ y Y, F = Y Fx
0
Dene : Z F by (y +x
0
) := y Y, F
1. is well-dened and linear because x
0
/ Y
2. (y) = 0 y Y
3. (x
0
) =
4. sup
xZz,=0
](x)]
|x|
= 1
Because:
sup
(y,),=(0,0)
[(y +x
0
)[
|y +x
0
|
= sup
(y,),=(0,0)
[[
|y +x
0
|
= sup
0,=,y
|
y
+x
0
|
= sup
yY
|x
0
+y|
=

inf
yY
|x
0
y|
= 1
Thm 5
x
such that |x
| = 1 and x
, x) = (x) x Z
x
, x
0
) = (x
0
) = x
, y) = (y) = 0 y Y 2
Corollary 1: X normed vector space, Y X linear subspace, x X.
Equivalent are:
(i) x Y
(ii) For every x
we have: x
, y) = 0 y Y implies x
, x) = 0
Proof:
(i) (ii) x = lim
n
y
n
y
n
Y
x
, y) = 0 y Y
x
, x) = lim
n
x
, y
n
) = 0
(ii) (i):
x / Y
Thm 7
x
: x
, y) = 0 y Y x
, x) , = 0 2
33
Corollary 2: X normed vector space, Y linear subspace
Y is dense Y
= 0
Proof: Corollary 1. 2
Corollary 3: X normed vector space, 0 ,= x
0
X x
such that
|x
| = 1, x
, x
0
) = |x
0
|
Proof: Theorem 7 with Y = 0, = |x
0
| 2
2.4 Reexive Banach Spaces
X real Banach space
X
:= L(X, 1)
X
:= L(X
, 1)
Example: Every element x X determines a bounded linear functional
x
:
X
1 by
x
(x
) := x
, x).
Bounded because [
x
(x
)[ |x
| |x| hence
|
x
| := sup
x
,=0
x
(x
)
|x
|
|x|
In fact: |
x
| = |x|, because x ,= 0 x
such that |x
| = 1 and
x
, x) = |x| (Cor 3). We have proved:

Lemma 4: The map : X X
dened by
(x)(x
) := x
, x)
is an isometric embedding.
Denition: A Banach space X is called reexive if the canonical embedding
: X X
(dened in Lemma 4) is bijective.

Example 1: X = H Hilbert space H

= H

= H
so H is reexive.
Example 2: (M, /, ) measure space, p, q > 1,
1
p
+
1
q
= 1 L
p
()

=
L
q
(), L
q
()

= L
p
().
So L
p
() is reexive for p > 1.
p = 1 : L
1
([0, 1])
= L
([0, 1])
L
([0, 1])
_ L
1
([0, 1])
so L
1
([0, 1]) is not reexive.
Example 3:
c
0
:= (x
n
)
nN
1
N
[ lim
n
x
n
= 0
|x| = sup
nN
[x
n
[
(c
0
)

=
1
, (
1
)
_ c
0
so c
0
,
1
,
are not reexive

34
Theorem 8: Let X be a Banach space.
X is reexive X
is reexive
suppose X is reexive and Y X be a closed linear subspace Y, X/Y
are reexive
Proof:
1. X reexive X
reexive:
Let : X
1 be a bounded linear functional. Need to show: x
such that
(x
) = x
, x
) x
Consider the diagram

X

X

1
Denote x
:= : X 1.
Let x
X and denote x :=
1
(x
) X. Then
(x
) = ((x)) = x
, x) = (x), x
) = x
, x
).
2. X
reexive X reexive:
Assume X
is reexive, but X is not reexive. Then (X) _ X
. Pick
an element x
0
X
(X).
Fact: (X) is a closed subspace of X
, because (X) is complete by

Lemma 4. bounded linear functional : X
1 such that:
(i) (x
) = 0 x
(X), and
(ii) (x
0
) = 1
X reexive x
such that
(x
) = x
, x
)x
, x) = (x), x
) = 0x X
x
= 0
1 = (x
0
)
(3)
= x
0
, x
) = 0
Contradiction.
3. Y is reexive:
Let : X
be the bounded linear map

(x
) := x
[
Y
By Hahn-Banach is surjective. Let y
. Consider the diagram

X
1
Let x
:= y
: X
1. Then, because X is reexive, y X such

that (y) = x
.
x
. We have (x
) = 0 and so
x
, y) = (y), x
) = x
, x
) = y
, (x
) = 0
To show: y
, y
) = y
, y)y
Given y
choose x
such that (x
) = y
, y) = (x
), y)
y
, y
) = y
, (x
))
= x
, x
)
= (y), x
)
= x
, y)
= y
, y)
35
4. Z = X/Y reexive:
Denote by : X X/Y the canonical projection, ie.
(x) = [x] = x +Y x X
Dene the bounded linear operator T : Z
by Tz
:= z
: X 1
Note:
(a) imT Y
because ker = Y
(b) In fact imT = Y
and T is an isometric isomorphism (Exercise 1,b)).

Let z
. Consider the composition

Y
T
1
Z
1
This is a bounded linear functional on Y
, so by Hahn-Banach:
x
such that
x
, x
) = z
, T
1
x
)x
(7)
x
, z
) = z
, z
)z
(8)
(X reexive)
x X such that (x) = x
Denote z := (x) = [x] Z

z
: z
, z
)
(7)
= x
, z
)
= (x), z
)
= z
, x)
= z
, (x))
= z
, z)
2
Remark: Y

= X
/Y
(X
/Y

= (Y
(Y
)
because X is reexive.
Recall A Banach space X is called separable if countable dense subset D
X.
Remark: Suppose there is a sequence e
1
, e
2
, e
3
, . . . such that the subspace
Y := spane
i
[ i N =
_
n
i=1
i
e
i
[ n N,
i
1
_
is dense in X. X is separable. Indeed the set D :=
n
i=1
i
e
i
[ n N,
i

is countable and dense.
Theorem: X Banach space
(i) X
separable X separable.
(ii) X separable and reexive X
separable.
Proof:
36
(i) Let D = x
1
, x
2
, x
3
, . . . be a dense, countable subset of X
. Assume
w.l.o.g that x
n
,= 0 n N such that
|x
n
| = 1, [x
n
, x
n
)[
1
2
|x
n
|
Denote Y = spanx
n
[ n N
Claim: Y is dense in X
By Hahn-Banach
Y is dense Y
= 0
Let x
. Because D dense in X
:
n
1
, n
2
, n
3
, . . . such that lim
i
|x
n
i
x
| = 0
Now:
|x
n
i
| 2[x
n
i
, x
n
i
)[ = 2[x
n
i
x
, x
n
i
)[ |x
n
i
x
| = 0
so x
= lim
i
x
n
i
= 0 so Y
= 0 so Y is dense in X so X is separable.
(ii) X reexive and separable X
= (X) separable X
is separable 2
Example:
(i) c
0
separable.
c
0
=
1
separable.
(
1
)
not separable.
(ii) (M, d) compact metric space X = C(M) separable
X
= / = nite Borel measures on M

not separable, except when M is a nite set.
(iii) X = L
p
() with Lebesgue-measure 1 p < . ,= 1
n
open X
is separable. L
() is not separable.
37
3 The weak and weak* topologies 30.11.2006
3 The weak and weak* topologies
3.1 The weak topology
Denition: A topological vector space is a pair (X, |), where X is a (real)
vector space and | is a topology such that the maps
X X X (x, y) x +y
and
1 X X (, x) x
are continuous.
Denition: A topological vector space (X, |) is called locally convex, if
x XU |, x U V | such that x V U V convex.
Lemma 1: X topological vector space, K X convex
K, int(K) are convex
Proof:
(i) int(K) is convex
x
0
, x
1
int(K), 0 < < 1
To show: x
= (1 )x
0
+x
1
int(K)
open set U X such that 0 U and x
0
+U K, x
1
+U K
x
+U K x
int(K)
(ii) K is convex
x
0
, x
1
K, 0 < < 1
To show: x
K. Let U X be an open set with x
U
W := (y
0
, y
1
) X X [ (1 )y
0
+y
1
U
W X X is open, (x
0
, x
1
) W.
open sets U
0
, U
1
X such that: x
0
U
0
, x
1
U
1
U
0
U
1
W
x
0
,x
1
K
y
0
U
0
K y
1
U
1
K
y
:= (1 )y
0
+y
1
K U, so K U ,=
Hence x
K.
2
Let X be a real vector space
Let T be a set of linear functions f : X 1
Let |
J
2
X
be the weakest topology such that f T is continuous w.r.t. |
J
If T f : X 1 we have for a < b x X [ a < f(x) < b |
J
Let 1
J
2
X
be the set of all subsets of the form
V := x X [ a
i
< f
i
(x) < b
i
i = 1, . . . , m f
i
T, a
i
, b
i
1
for i = 1, . . . , m
Lemma 2:
(i) Let U X. Then U |
J
if and only if
x U V 1
J
such that x V U ()
(ii) (X, |
J
) is a locally convex topological vector space
(iii) A sequence x
n
X converges to x
0
X if and only if
f(x
0
) = lim
n
f(x
n
) f T
(iv) (X, |
J
) is Hausdor if and only if x X, x ,= 0 f T such that
f(x) ,= 0.
38
Proof:
(i) Exercise with hint:
Dene |
J
:= U X [ ()
Prove:
(a) |
J
is a topology
(b) Each f T is continuous w.r.t. |
J
(c) If | 2
X
is another topology such that each f T is continuous
w.r.t. | then |
J
|
(ii) Each V 1
J
is convex
scalar multiplication is continuous:
0
1, x
0
X
Choose V 1
J
such that
0
x
0
V > 0 such that
(
0
)x
0
, (
0
+)x
0
V and ,=
0
U :=
1
V
1
0
+
V 1
J
If x U and [
0
[ < then x V (because V is convex)
addition is continuous:
x
0
, y
0
X, x
0
+y
0
W, W 1
J
Dene U :=
1
2
W +
x
0
y
0
2
V :=
1
2
W +
y
0
x
0
2
U, V 1
J
, x
0
U, y
0
V
x U, y V x +y W
(iii) Assume x
n
/
F
x
0
Let f T, > 0. Denote U := x X [ [f(x) f(x
0
)[ < 1
J
x
0
U n
0
Nn n
0
: x
n
U
n n
0
[f(x) f(x
0
)[ <
Assume f(x
n
) f(x
0
) f T
Let U |
J
with x U
(i)
V 1
J
with x
0
V U
V = x X [ a
i
< f
i
(x
i
) < b
i
i = 1, . . . , m
a
i
< f
i
(x
0
) < b
i
i = 1, . . . , m
n
0
N i 1, . . . , m n n
0
: a
i
< f
i
(x
n
) < b
i
n n
0
x V U
(iv) Exercise without hints
2
Example 1: I any set, X = 1
I
:= x : I 1 x
i
iI
is a vector space.
product space.
i
: X 1 projection
i
(x) = x(i) linear map
| 2
X
weakest topology such that each
i
is continuous
Example 2: X Banach space
T := : X 1 [ is bounded and linear = X
Let |
w
be the weakest topology such that each bounded linear functional is
continuous w.r.t. |
w
Facts:
a) |
s
2
X
strong topology; induced by the norm; |
w
2
X
weak topology:
|
w
|
s
b) (X, |
w
) is a locally convex vector space
c) A sequence x
n
X converges to x
0
X if and only if
x
, x
0
) = lim
n
x
, x
n
) x
Notation: x
n
x
0
or x
0
= w- lim
n
x
n
39
Example 3: Let X be a Banach space, L(X, 1) dualspace
Let |
w
2
X
be the weakest topology on X
such that each linear functional

of the form X
1x
, x) is continuous (in this case T = i(X) X
)
Facts:
a) |
s
2
X
strong topology
|
w
2
X
weak topology |
w
|
w
|
s
b) (X
, |
w
) is a locally convex topological vector space

c) A sequence x
n
X
converges to x
0
in the weak
-topology if and only if

x
0
, x) = lim
n
x
n
, x) x X
Notation: x
n
w
0
or x
0
= w
lim
n
x
n
Remark: Suppose the sequence x
, x
n
) converges x
Does this imply that x

n
converges weakly?
No, denote (x
) = lim
n
x
, x
n
)
Then : X 1 is linear and continuous
x
n
converges weakly i(X) X
Exercise: Find an example

Lemma 3: X Banach space, K X convex
Assume: K is closed w.r.t. the strong topology K is weakly closed
Proof: Let x
0
X K, K ,=
> 0 such that B
(x
0
) K =
Chap. II Thm 6
x
, c 1 such that x
, x) < c x B
(x
0
) and
x
, x) c x K
U := x X [ x
, x) < c weakly open and x

0
U, U K = 2
Lemma 4 (Mazur): x
i
sequence, x
i
x
0
> 0
1
, . . . ,
n
1
i
0
n
i=1
i
= 1 |x
0

n
i=1
i
x
i
| <
Proof: K :=
n
i=1
i
x
i
[ n N,
i
> 0,
n
i=1
i
= 1 convex
Lemma 1
the strong closure K of K is convex
Lemma 3
K is weakly closed x
0
K 2
Lemma 5: X Banach space, -dimensional
S := x X [ |x| = 1 B := x X [ |x| 1
B is the weak closure of S.
Proof: x
0
B, U |
w
and x
0
U
U S ,= . Choose
i
, x
i
such that
V := x X [ [x
i
, x
0
x)[ <
i
U
V E := x X [ x
i
, x
0
x) = 0 nontrivial ane subspace, E S ,= 2
Example: X = l
1
x
n
sequence, l
1
x
0
Then x
n
x
0
x
n
x
0
|x
n
x
0
|
1
0
3.2 The weak* topology
Theorem 1 (Banach Alaoglu, sequentially): X separable Banach space
every bounded sequence x
n
X
has a weak
-convergent subsequence.
40
Proof: D = x
1
, x
2
, x
3
, . . . X dense, countable.
c := sup
nN
|x
n
| <
[x
n
, x
1
)[ |x
n
| |x
1
| c|x
1
|
subsequence(x
n
i,1
)
i=1
such that x
n
i,1
, x
1
) converges.
The sequence x
n
i,1
, x
2
) 1 is bounded
further subsequence (x
n
i,2
)
i=1
such that x
n
i,2
, x
2
) converges.
Induction
sequence of subsequences (x
n
i,k
)
i=1
such that
(x
n
i,k+1
)
i
is a subsequence of (x
n
i,k
)
i
the limit lim
i
x
n
i,k
, x
k
) exists for every k N
Diagonal Subsequence
x
n
i
:= x
n
i,i
(x
n
i
)
i=1
is a subsequence of (x
n
)
n=1
and the limit lim
i
x
n
i
, x
k
) exists for
every k N.
By Chapter II, Thm 2 (Banach-Steinhaus) with Y = 1, x
such that
x
, x) = lim
i
x
n
i
, x)x X
So x
n
i
x
2
Example 1: (M, d) compact metric space with M ,= , B 2
M
Borel -
algebra. X := C(M) separable
X
= real Borel measures : B 1

f : M M homeomorphism.
A Borel measure : B [0, ) is called an f-invariant Borel probability mea-
sure if
(M) = 1
B B (f(B)) = (B)
/(f) := f-invariant Borel prob. meas. on M X
Fact 1: /(f)
|| = [[(M) = (M) = 1
Fact 2: /(f) is convex Exercise.
Fact 3: /(f) ,=
Proof: Fix an element x M. Dene the Borel-measure
n
: B 1
_
M
ud
n
:=
1
n
n1
k=0
u(f
k
(x)) u C(M)
where f
k
:= f f . . . f
. .
k
|
n
| 1,
n
0 (Thm1) weak
convergent
subsequence
n
i

Claim: /(f)
0
_
M
ud = lim
n
_
M
ud
n
i
0 u 0
41
(M) =
_
M
1 d = lim
i
_
M
1 d
n
i
= 1
(f(B)) = (B)B B
_
M
u f d = lim
i
1
n
i
n
i
k=1
u(f
k
(x)) = lim
i
1
n
i
n
i
1
k=1
u(f
k
(x))
_
M
u f d =
_
M
udf
u C(M)
f
(B) = (B)B B
and f
(B) = (f
1
(B)) 2
Example 2: X =
, elements of X are bounded sequences x = (x

i
)
i=1
1.
|x|
= sup
iN
[x
i
[
Denition:
n
: X 1 by
n
(x) := x
n
and |
n
| = 1
Exercise: Show that
n
X
has no weak
-convergent subsequence, ie. for

all subsequences n
1
< n
2
< n
3
< . . . : x X =
such that the sequence

(
n
i
(x))
i=1
1 does not converge.
Theorem 2 (Banach-Alaoglu, general form): X Banach space the unit
ball B
:= x
[ |x
| 1 in the dual space is weak
compact.
Remark: X
with the weak
topology is Hausdor.
B
is weak
-closed. Prove it directly without using Thm 2.

Theorem 3 (Tychono): Let I be any index set and, for each i I, let K
i
be a compact topological space K :=
iI
K
i
is compact wrt the product
topology.
Remark: K = x = (x
i
)
iI
[ x
i
K
i

i
: K K
i
canonical projection
product topology := weakest topology on K wrt which each
i
is continuous.
Proof Thm 3 Thm 2: I = X. K
x
:= [|x|, |x|] 1
K :=
xX
K
x
= f : X 1 [ [f(x)[ |x|x X 1
X
L := f : X 1 [ f is linear 1
X
By Thm 3: K is compact
L is closed with respect to the product topology. For x, y X, 1, the
functions
x,y
: 1
X
1,
x,
: 1
X
1
given by
x,y
(f) := f(x +y) f(x) f(y) and
x,
:= f(x) f(x) are
continuous wrt product topology. So
L =
x,y
1
x,y
(0)
x,
1
x,
(0)
is closed K L = B
is compact. Product topology on K L =

weak
-topology on X
2
42
Denition: K any set. A set / 2
K
is called FiP if A
1
, . . . , A
n
/
A
1
. . . A
n
,= . A set B 2
K
is called maximal FiP if B is FiP and
/ 2
K
we have
B / and / FiP / = B
Fact 1: If / 2
K
is FiP, then B 2
K
max FiP such that / B (Zorns
Lemma)
Fact 2: Let B 2
K
is max FiP, then:
(i) B
1
, . . . , B
n
B B
1
. . . B
n
B
(ii) C K, C B ,= B B C B
Fact 3: Let K be a topological space. Then K is compact if and only if every
FiP collection / 2
K
of closed subsets satises
A,
A ,=
Proof of Tychonos theorem: K =
i
K
i
. Let / 2
K
be a FiP collec-
tion of closed sets. By Fact 1 max FiP collection B 2
K
with / B (not
each B B needs to be closed).
To show:
BB
B ,=
Step 1: Construction of an element x K.
Fix i I.
i
: K K
i
projection. Denote B
i
:=
i
(B) [ B B 2
K
i
B
i
is FiP (if B
1
, . . . , B
n
B then
i
(B
1
) . . . (B
n
)
i
(B
1
. . . B
n
) ,= )
(K
i
compact,Fact 3)
BB
i
(B) ,=
Pick x
i

BB

i
(B) Choose x = (x
i
)
iI
K (Axiom of Choice).
Step 2: x B B B
Let U K be open with x U.
To show: U B ,= B B U open, x U (!) nite set J I
open sets U
i
K
i
, i J such that x
iJ

1
i
(U
i
) U (like Lemma 2 (i)).
x
i
=
i
(x) U
i

i
(B)i J, B B
(U
i
open)
U
i

i
(B) ,= i JB B

1
i
(U
i
) B ,= i JB B
(Fact 2)

1
i
(U
i
) Bi J
(Fact 2)
iJ
1
i
(U
i
) B
iJ
1
i
(U
i
) B ,= B B
U B ,= B B
2
Theorem 4: X separable Banach space, K X
. Equivalent are:
(i) K is weak
compact
(ii) K is bounded and weak* closed
(iii) K is sequentially weak* compact
(iv) K is bounded and sequentially weak* closed
Exercise: /(f) as in example 1 /(f) is weak* compact.
43
Proof of Thm 4: Exercise.
(i) (Thm 2) (ii): use uniform boundedness (Chapter II, Thm 1)
(ii) (Thm 1) (iii) (denitions) (iv) (ii)
Given x
weak* closure(K). Need to prove sequence x
n
K with x
n
x
.
Then, by (iv), x
K 2
Theorem 5: X Banach space, E X
linear subspace
Assume E B
is weak
-closed, where B
:= x
[ |x
| 1
Let x
0
X
E. Let be such that 0 < < inf

x
E
|x
0
x
|
x
0
X such that x
0
, x
0
) = 1, x
, x
0
) = 0 x
E, |x
0
|
1
Remark 1: E is closed
Let x
n
E and x
n
x
c > 0 such that |x
n
| c n N
n
c
E B
c
E B
= E B
E
E closed > 0 as in the hypothesis of Theorem 5
Remark 2: B
is closed in the weak
-topology.
Hence each closed ball x
[ |x
0
| r is weak
-closed.
Proof:
Step 1 There is a sequence of nite sets S
n
B = x X [ |x| 1
satisfying the following condition for every x
:
|x
0
| n
max
xS
k
[x
0
, x)[ k
for k = 0, . . . , n 1
_
_
x
/ E ()
Proof of Step 1: n = 1 Choose S
0
=
Then () holds for n = 1
n 1: Assume S
0
, . . . , S
n1
have been constructed such that () holds.
To show: There is a nite set S
n
B such that () holds with n replaced by
n + 1
For any nite set S B denote
E(S) :=
_
_
x
|x
0
| (n + 1)
max
xS
k
[x
0
, x)[ kk = 0, . . . n 1
max
xS
[x
0
, x)[ n
_
_
To show: nite set S B such that E(S) =
Suppose, by contradiction, that E(S) ,= , for every nite set S B
a) The set K := x
E [ |x
0
| (n + 1) is weak
-compact.
Let R := |x
0
| +(n + 1) Then |x
| R x
K
so K E RB
= R(E B
) =: E
R
E
R
is weak
-closed. So K = E
R
..
weak
cl. by ass.
x
[ |x
0
| (n + 1)
. .
weak
-closed by Rem. 2
K is weak
-closed and bounded

Thm 2
K is weak
-compact.
b) E(S) is weak
-closed for every S

E(S) is the intersection of K with the weak
-closed subsets
x
[ max
xS
k
[x
0
, x)[ k k = 0, . . . , n 1
x
[ max
xS
[x
0
, x)[ n
44
c) S
i
B nite set for i I, I nite
iI
E(S
i
) = E
_
iI
S
i
_
,= by assumption.
Let o := S B [ S nite Then by c), the collection E(S) [ S o is
FIP and by b) it consists of weak
-closed subsets of K.
By a) K is weak
-compact
SS
E(S) ,=
Let x
SS
E(S) x
E, |x
0
| (n + 1),
max
xS
k
[x
0
, x)[ k k = 0 . . . n1 and [x
0
, x)[ n x B
i.e. |x
0
| n This contradicts ()
Step 2 Construction of x
0
Choose a sequence x
i
B which runs successively through all points of the set
S =
i=1
1
n
S
n
Then lim
n
|x
i
| = 0
Dene a linear operator T : X
c
0
by Tx
:= (x
, x
i
))
iN
Claim: For every x
E there is an i N such that [x
0
, x
i
)[
Let x
E and choose n
|x
0
|
|x
0
| n
Step 1
k n 1 x S
k
such that [x
0
, x)[ > k
[x
0
,
x
k
)[ > i such that x
i
=
x
k
The claim shows: |Tx
Tx
0
| > x
E
Tx
0
/ TE
Chap II, Thm 7
c
0
= l
1
such that , Tx
0
) = 1 , Tx
) = 0 x
E
||
1

1
Dene x
0
:=
i=1
i
x
i
X.
Note 1
i=1
|
i
x
i
|
i=1
[
i
[ <
So by Chapter I Theorem 10, the sequence
i=1
i
x
i
converges as n
Note 2
a) x
0
, x
0
) =
i=1
i
x
0
, x
i
) = , Tx
0
) = 1
b) x
, x
0
) = , Tx
) = 0 x
E
c) |x
0
|
i=1
[
i
[ = ||
1

1
2
Corollary 1: Let X be a Banach space and E X a linear subspace.
Let B
:= x
[ |x
| = 1
Equivalent are:
(i) E is weak
-closed
(ii) E B
is weak
-closed
(iii) (
E)
= E
Exercise: X Banach space, : X X
canonical embedding
(X) is weak
-dense in X
45
Proof of Corollary 1: (i)(ii) obvious
(ii)(iii) E (
E)
by denition (
E = x X [ x
, x) = 0 x
E)
(
E)
E: Let x
0
, E
Thm 5
x
0
X such that x
0
, x
0
) = 1
x
, x
0
) = 0 x
E
x
0

E, x
0
, x
0
) , = 0 x
0
/ (
E)
(iii) (i) E = (
E)
x

E
x
[ x
, x) = 0
. .
weak
-closed
2
Corollary 2: X Banach space, : X
1 linear
Equivalent are:
(i) is continuous w.r.t. weak
-topology
(ii)
1
(0) X
is weak
-closed
(iii) x Xx
(x
) = x
, x)
Proof: (iii) (i) denition of weak
-topology
(i) (ii) denition of continuity
(ii) (iii) w.l.o.g ,= 0
E :=
1
(0) X
is weak
-closed
Choose x
0
X
such that (x
0
) = 1, so x
0
/ E
Thm 5
x
0
X such that x
0
, x
0
) = 1 x
, x
0
) = 0 x
E
Let x
(x
)x
0
E
x
(x
)x
0
, x
0
) = 0
x
, x
0
) = (x
)x
, x
0
) = (x
) 2
Corollary 3: X Banach space, : X X
canonical embedding
S := x X [ |x| = 1
The weak
-closure of (S) X
is B
= x
[ |x| 1
Proof: K := weak
closure of (S)
1. K B
because B
is closed
2. K is convex:
Key fact: : X
..
weak top
X
..
weak
-top
is continuous
Hence (B) K : x B U X
weak
open, (x) U

1
(U) X weakly open and x
1
(U)

1
(U) S ,=
U (S) ,=
U
(x) K
So K is the weak
-closure of the convex set (B)

Lemma 1
K is convex
3. B
K
x
0
/ K weak
-continuous linear functional : X
1 such that
sup
x
K
(x
) < (x
0
)
Cor 2
x
0
X
such that (x
) = x
, x
0
)
x
0
, x
0
) > sup
x
K
x
, x
0
) sup
xS
x
0
, x) = |x
0
| |x
0
| > 1 x
0
/
B
2
46
Lemma 6: X normed vector space
(i) If x
1
, . . . , x
n
X
are linearly independent, then x

1
, . . . , x
n
X such
that
x
i
, x
j
) =
ij
(ii) If x
1
, . . . , x
n
X are lin. indep., then
X
0
:= x X [ x
i
, x) = 0, i = 1, . . . , n
is a closed subspace of codimension n and X
0
= spanx
1
, . . . , x
Proof:
(i) for n (ii) for n: x X we have: x
n
i=1
x
i
, x)x
i
X
0
This shows: X = X
0
spanx
1
, . . . , x
n
and, moreover x
0
0 = x
, x
i
x
i
, x)x
i
)
= x
, x)
i
x
i
, x) x
, x
i
)
= x
i=1
x
, x
i
)x
i
, x)x X
x
=
n
i=1
x
, x
i
)x
i
spanx
1
, . . . , x
(ii) for n (i) for n + 1: Let x
1
, . . . , x
n+1
X
be lin. indep. for i =

1, . . . , n + 1 denote X
i
:= x X [ x
j
, x) = 0, j ,= i
(i) for n
X
i
= spanx
j
[ j ,= i
x
i
/ X
i
x
i
X
i
such that x
i
, x
i
) = 1
x
j
, x
i
) =
ij
2
Remark 1: Converse if x
i
, . . . , x
n
X are lin. indep., then x
1
, . . . , x
n
X
such that x
i
, x
j
) =
ij
Remark 2: x
1
, . . . , x
n
X
lin indep, c
1
, . . . , c
n
1
x X such that x
i
, x) = c
i
(namely: x =
n
i=1
c
i
x
i
with x
i
as in Lemma
6).
Lemma 7: X normed vector space, x
1
, . . . , x
n
X
, c
1
, . . . , c
n
1, M 0.
Equivalent are:
(i) > 0x X such that
x
i
, x) = c
i
, i = 1, . . . , n |x| M +
(ii)
1
, . . . ,
n
1 we have
[
n
i=1
i
c
i
[ M|
n
i=1
i
x
i
|
47
Proof: (i)(ii):
x = x
X as in (i). Then
[
i
c
i
[ = [
i
x
i
, x
)[
= [
i
x
i
, x
)[
= |
i
x
i
| |x
|
= (M +)|
i
x
i
| > 0
(ii)(i):
Assume x
1
, . . . , x
n
lin indep (w.l.o.g)
Choose x X such that x
i
, x) = c
i
i (Remark 2). X as in Lemma 6.
inf
X
0
|x +| = sup
0,=x
0
[x
, x)[
|x
|
(by Lemma 6) = sup
i
R
[
i
x
i
, x)[
|
i
x
i
|
= sup
i
[
i
c
i
[
i
x
i
[
M
2
Theorem 6: X Banach space. Equivalent are:
(i) X is reexive
(ii) The unit ball B := x X [ |x| 1 is weakly compact
(iii) Every bounded sequence in X has a weakly convergent subsequence
Proof : : X X
is an isomorphism from X with the weak topology to

X
with the weak*topology.

U X weakly open (U) X
is weak*-open.
(i)(ii) (B) is the unit ball in X
, hence is weak*compact (Thm 2), so B

is weakly compact.
(i)(iii) X separable and reexive X
separable (Chapter II, Thm 9).

(x
n
) X bounded sequence (x
n
) X
is a bounded sequence. So, by

Thm 1, (x
n
) has a weak*-convergent subsequence (x
n
i
)
x
n
i
converges weakly.
(i)(iii): The nonseparable case Let x
n
X be a bounded sequence.
Denote Y :=
_
N
n=1
n
x
n
[ N N,
n
1
_
Y is separable and reexive (Chapter II, Thm 8)
x
n
by separable case has a subsequence x
n
i
converging weakly in Y ,ie.
x Y y
: y
, x) = lim
i
y
, x
n
i
)
by Hahn-Banach
x
: x
, x) = lim
i
x
, x
n
i
)
48
(ii)(i) Let x
, x
,= 0.
Claim: For every nite set S X
there exists an x X such that

x
, x) = x
, x
) x
S |x| 2|x
|
o := S X
[S nite subset and K(S) := x X [ |x| 2|x
|, x
, x
) =
x
, x)x
S
Note:
K(S) is a weakly closed subset of cB = x X [ |x| c where c =
2|x
|.
cB is weakly compact, by (ii).
The collection K(S) [ S o is FiP, because
K(S
1
) . . . K(S
n
) = K(S
1
. . . S
n
) ,=
SS
K(S) ,=
x X such thatx
, x) = x
, x
)x
Proof of claim:
Write S = x
1
, . . . , x
n
, c
i
:= x
, x
i
).
[
i
c
i
[ = [
i
x
, x
i
)[ = [x
i
x
i
)[ |x
| |
i
x
i
|
Assumption of Lemma 7 holds with M = |x
| > 0 Choose = |x
| > 0.
(iii)(i) Let x
0
X
, |x
0
| 1. Denote E := x
[ x
0
, x
) = 0
and B
:= x
[ |x
| 1
Claim 1: E
is weak*closed.
Claim 1
by Cor 1
E is weak*-closed
by Cor 2
x
0
Xx
: x
0
, x
) = x
, x
0
) x
0
= (x
0
)
So : X X
is surjective.
Claim 2: x
1
, . . . , x
n
X
, x X such that
x
i
, x) = x
0
, x
i
), i = 1, . . . , n |x| 1
Proof of Claim 2 Denote U
m
:= x
[ [x
0
, x
i
)[ <
1
m
; i =
1, . . . n
x
0
U
m
, U
m
is weak*open. Moreover |x
0
| 1.
Recall the weak*closure of (S), S := x X [ |x| = 1 is the closed unit ball
in X
(Cor 3). U
m
(S) ,= x
m
X such that
|x
m
| = 1 [x
i
, x
m
) x
0
, x
i
)[ <
1
m
i = 1, . . . , n
by (iii) weakly convergent subsequence x
m
k
x.
|x| 1 and
x
i
, x) = lim
k
x
i
, x
m
k
) = x
0
, x
i
) i = 1, . . . n
Proof of Claim 1 Let x
0
weak*closure of E
. We must prove that

x
0
E
. Clearly |x
0
| 1. So it remains to prove x
0
, x
0
) = 0
Step 1 Let > 0. Then sequences x
n
X, n 1, x
n
X
such that
(1) |x
n
| 1, |x
n
| 1, x
0
, x
n
) = 0
(2) x
i
, x
n
) = x
0
, x
i
) i = 0, . . . , n 1
(3) [x
n
x
0
, x
i
)[ < i = 1, . . . , n
49
Proof of Step 1 Induction n = 1: a) By Claim 2, x
1
X such that |x
1
|
1, x
0
, x
1
) = x
0
, x
0
) b) Because x
0
weak*closure (E
) x
1
E
such that [x
1
x
0
, x
1
)[
(1),(2),(3) hold for n = 1.
n 1: Suppose x
i
, x
i
have been constructed for i = 1, . . . , n.
a) By Claim 2,
x
n+1
X : x
i
, x
n+1
) = x
0
, x
i
) i = 0, . . . , n |x
n+1
| 1
b) x
n+1
E
such that [x
n+1
x
0
, x
i
)[ , i = 1, . . . , n
Step 2 x
0
, x
0
) = 0
By (iii) weakly convergent subsequence
x
n
i
x
0
X |x
0
| 1
by Lemma 4 m N
1
, . . . ,
m
0
(4)
m
i=1
i
= 1 |x
0

m
i=1
i
x
i
| <
a) x
m
, x
0
) = lim
i
x
m
, x
n
i
) = lim
i
x
0
, x
m
) = 0
b)
[x
0
, x
0
)[
(by a)
0
, x
0
)
_
x
M
,
m
i
x
i
_
_
x
m
,
m
i=1
i
x
i
x
o
_
i
[x
0
, x
0
) x
m
, x
i
)[
. .
by (3)
+
_
_
_
_
_
m
i
x
i
x
0
_
_
_
_
_
. .
(3) by(4)
2
2
3.3 Ergodic measures
(M, d) compact metric space, f : M M homeomorphism
/(f) := f-invariant Borel probability measure : B [0, )
B 2
M
Borel -algebra, (M) = 1, (f(E)) = (E) E B
We know: /(f) nonempty, convex, weak
-compact.
Denition: An f-invariant Borel-measure /(f) is called ergodic,
if B = f() () 0, 1
Example: M = S
2
N
() =
_
1 N
0 N /

S
() =
_
1 S
0 S /
where
N stands for north pole and S for south pole.
Denition: X vectorspace, K X convex
x K is called an extremal point of K, if the following holds:
x
0
, x
1
K
x = (1 )x
0
+x
1
0 < < 1
_
_
_
x
0
= x
1
= x
Lemma 8: /(f) extremal point is ergodic
50
Proof: Suppose not.
B such that = f(), 0 < () < 1
Dene
0
,
1
: B [0, ) by
0
(E) :=
(E\)
1()

1
(E) =
(E)
()
0
,
1
/(),
0
,= ,
1
,=
= (1 )
0
+
1
= ()
is not extreme 2
Theorem 7 (Krein-Milman): X locally convex topological T2 vectorspace
K X nonempty, compact, convex
E := set of extremal points of K
C :=convex hull of E :=
m
i=1
i
e
i
[ e
i
E
i
0
m
i=1
i
= 1
C = K (in particular C ,= )
Corollary: Every homeomorphism of a compact metric space has an ergodic
measure
Proof: Apply Theorem 7 to the case X = C(M)
with weak
-topology
and K = /(f) 2
Proof of Theorem 7:
Step 1 A, B X nonempty, disjoint, convex sets, A open continuous
linear functional : X 1 such that (a) < inf
bB
(b) a A
Proof of Step 1: Hahn-Banach as in Chapter II Theorem 6
Step 2 x X, x ,= 0 linear functional : X 1 with (x) ,= 0
Proof: Choose an open convex neighborhood A X of 0 such that x / A.
Denote B = x. Now apply Step 1.
Step 3 E ,=
Proof:
A nonempty compact convex subset K
X is called a face of K if
K
K and
x K
, x
0
, x
1
K
x = (1 )x
0
+x
1
0 < < 1
_
_
_
x
0
, x
1
K
Denote / := K X [ K is nonempty, compact, convex

/ is partially ordered by K
K
def
K
is a face of K
K K
K K
, K
K K
= K
K
, K
K K
K
If ( / is a chain, then K
0
:=
Cc
C / !
Because of the FIP characterisation of compactness we have K
0
,=
Zorns Lemma implies: For every K /, there is a minimal element K
0
/
such that K
0
K
Claim: K
0
= pt
Suppose K
0
x
0
, x
1
x
0
,= x
1
Step 2
: X 1 continuous linear such that (x
1
x
0
) > 0 so (x
0
) < (x
1
)
K
1
:= K
0

1
( sup
xK
0
(x)) / K
1
K
0
and K
0
,= K
1
since x
0
K
0
K
1
So K
0
is not minimal. Contradiction.
Claim K
0
= x
0
x
0
E (by denition of face)
51
Step 4 K = C
Clearly C K
Suppose C _ K, let x
0
K C
Step 1
: X 1 continuous linear such that (x
0
) > sup
C
K
0
:= K
1
(sup
K
) is a face of K and K
0
C =
Step 3
K
0
has an extremal point e
e is extremal point of K
Contradiction, because e / C 2
(M, d) compact metric space
f : M M homeomorphismus, /(f) ergodic
u : M 1 continuous, x M
Question:
1
n
n1
k=0
u(f
k
(x))
?
_
M
ud
Theorem (Birkho): u C(M) B such that f() = () = 1
_
M
ud = lim
n
1
n
n1
k=0
u(f
k
(x)) x
Without proof
Theorem 8 (von Neumann): (M, d) compact metric space, f : M M
homeomorphismus, /(f) ergodic, 1 < p <
lim
n
_
_
_
1
n
n1
k=0
u(f
k
(x))
_
M
ud
_
_
_
L
p
= 0
Theorem 9 (Abstract Ergodic Theorem): X Banach space, T L(X), c
1
Assume |T
n
| c n N
Denote S
n
:=
1
n
n1
k=0
T
k
L(X)
Then the following holds:
(i) For x X we have: (S
n
x)
nN
converges (S
n
x)
nN
has a weakly con-
vergent subsequence.
(ii) The set Z := x X [ S
n
x converges is a closed linear subspace of X
and
Z = Ker(1 T) Im(1 T)
If X is reexive Z = X
(iii) Dene S : Z Z by S(x +y) := x x ker(1 T), y Im(1 T)
Then Sz := lim
n
S
n
z z Z and ST = TS = S
2
= S, |S| c
Proof of Theorem 9 Theorem 8 X = L
p
() Tu := u f
_
M
[u f[
p
d =
_
M
[u[
p
d, so |Tu|
p
= |u|
p
|T| = 1 k N
(S
n
u)(x) =
1
n
n1
i=1
u(f
k
(x))
To show: lim
n
_
_
S
n
u
_
M
ud
_
_
L
p
()
= 0
Equivalently Claim 1: (Su)(x) =
_
M
ud x M
(By Theorem 9 we have S
n
u Su in L
p
() )
Claim 2: Tv = v v const.
Claim 2 Claim 1: Su Ker(1 T) Su c
c =
_
M
Sud = lim
n
_
M
S
n
ud = lim
n
1
n
n1
k=0
_
M
u f
k
d =
_
M
ud
Proof of Claim 2:
v : M 1 measurable,
_
M
[v[
p
d <
[v] L
p
() T[v] = [v] v f = v almost everywhere
E
0
:= x M [ v(x) ,= v(f(x)) measure zero
E :=
kN
f
k
(E
0
) (E) = 0
52
M E =
0

+

0
= x M [ v(x) = c
= x M [ v(x) > c where c :=

_
M
v d

0
,
finvariant, (
0
) +(
+
) +(
) = 1
ergodic
(
0
) = 1
because otherwise: (
+
) = 1 so
_
M
v d > c
or (
) = 1 so
_
M
v d < c 2
Let A L(X, Y ).
Denition: The dual operator of A is the bounded linear operator A
L(Y
, X
) dened by A
, x) := y
, Ax) for y
and x X, ie.
X
A
//
A
99
Y
y
//
1
Remark: |A
| = |A|
Proof:
|A
| = sup
y
,=0
|A
|
|y
|
= sup
y
,=0
sup
x,=0
A
, x)
|x| |y
|
= sup
y
,=0
x,=0
y
, Ax)
|y
| |x|
by Hahn-Banach
= sup
x,=0
|Ax|
|x|
= |A|
2
Proof of Thm 9: X is space. T L(X), c 1, |T
k
| c, k = 0, 1, 2, . . ..
Denote S
n
:=
1
n
n1
k=0
T
k
Step 1
|S
n
| c, |S
n
(1l T|
1 +c
n
Proof
|S
n
|
1
n
n1
k=0
|T
k
| c
S
n
(1l T) =
1
n
n1
k=0
T
k
1
n
n
k=1
T
k
=
1
n
(1l T
n
)
Step 2 x, X with Tx = x we have |x| c|x + T|.
Proof x = Tx = T
2
x = . . . S
n
x = x n N.
Also, by Step 1, lim
n
|S
n
( T)| = 0
|x| = lim
n
|S
n
(x + T)|
. .
c|x+T|
c|x + T|
53
Step 3
x ker(1l T), y im(1l T) |x| c|x +y|
Proof
n
X such that
n
T
n
y
by Step 2 |x| c|x +
n
T
n
. .
y
|
n
= |x| c|x +y|
Step 4 ker(1l T) im(1l T) = 0 and the subspace X
0
:= ker(1l T)
im(1l T) is closed
Proof For x ker(1l T) im(1l T) choose y := x in Step 3 x = 0
Let z
n
X
0
, z = lim
n
z
n
X.
Write z
n
= x
n
+y
n
, x
n
ker(1l T), y
n
im(1l T)
by Step 3
|x
n
x
m
| c|z
n
z
m
|
x
n
is Cauchy
The limits exist and x := lim
n
x
n
ker(1lT), y := lim
n
y
n
im(1l t), z = x+y X
0
Step 5 Let z = x +y X
0
where x ker(1l T), y im(1l T).
S
n
X
0
for all n N and lim
n
S
n
z = x
Proof
a) x = Tx S
n
x = x X
0
b) Choose
k
X such that
k
T
k
y. Then
S
n
y = lim
k
S
n
(1l T)
k
= lim
k
(1l T)S
n
k
. .
im(1lT)
im(1l T)
c) By Step 1,
|S
n
(T)|
1 +c
n
|| X
So S
n
y 0 y im(1l T)
(by Ch II, Thm 2 (ii)) S
n
y 0y im(1l T)
S
n
z = x +S
n
y x.
Step 6 Let x, z X. Equivalent are
(i) lim
n
S
n
z = x
(ii) subsequence n
1
< n
2
< n
3
< . . . such that
w lim
i
S
n
i
z = x
(iii) Tx = x, z x im(1l T)
54
Proof (iii)
by Step 5
(i)
obvious
(ii).
(ii) (iii)
x
, Tx x) = T
, x)
by (ii)
= lim
i
T
, S
n
i
z)
= lim
i
x
, T 1l)S
n
i
. .
|.|
1+c
n
z)
= 0 x
by Step 1
by Hahn-Banach Tx x = 0.
Suppose x
(im(1l T))
, T) = 0 X
x
, ) = 0
x
= T
= (T
)
2
x
, x
= S
n
i
x
i
x
, z x)
(ii)
= lim
i
x
, z S
n
i
z)
= lim
i
x
n
i
x
, z) = 0
So z x
((im(1l T))
) = im(1l T).
Step 6
Theorem 9.
X
0
= X in the reexive case. 2
55
4 Compact operators and Fredholm theory 20.12.2006
4 Compact operators and Fredholm theory
4.1 Compact operators
Lemma 1: X, Y Banach spaces, K L(X, Y ). Equivalent are:
(i) If x
n
X is a bounded sequence then Kx
n
Y has a convergent subse-
quence
(ii) If S X is a bounded subset then KS is a compact subset of Y .
(iii) The set Kx [ |x| 1 Y is compact.
The operator K is called compact if it satises these equivalent conditions.
Proof: (i) (ii):
To Show: Every sequence in KS has a Cauchy subsequence (see Ch I, Lemma
7). Let y
n
KS. Choose x
n
S such that y
n
= Kx
n
.
x
n
is a bounded subsequence
(i)
Kx
n
has a convergent subsequence Kx
n
i
=
y
n
i
y
n
i
is Cauchy.
(ii) (iii): take S := x X [ |x| 1
(iii) (i): x
n
X bounded. Choose c > 0 such that |x
n
| cn
(iii)
K
x
n
c
has a convergent subsequence Kx
n
has a convergent subsequence
2
Example 1: T : X Y surjective, dimY = T not compact. (open
mapping theorem) Tx [ |x| < 1 y Y [ |y| for some > 0 not
compact.
Example 2: K L(X, Y ), imK nite dimensional K is compact.
Example 3: X = C
1
([0, 1]), Y = C
0
([0, 1]), K : X Y obvious inclusion
K is compact. Arzela-Ascoli.
Example 4: X = Y =
p
, 1 p
1
,
2
, . . . 1 bounded.
Kx := (
1
x
1
,
2
x
2
,
3
x
3
, . . .)
K compact lim
n
n
= 0
(exercise).
Theorem 1: X, Y, Z Banach spaces
(i) A L(X, Y ), B L(Y, Z)
A compact or B compact BA is compact
(ii) K
L(X, Y ) compact = 1, 2, 3, . . . K L(X, Y ) such that

lim
|K
K| = 0 K is compact.
(iii) K compact K
is compact.
56
Proof: (i): Exercise
(ii): x
n
X bounded, c := sup
nN
|x
n
| <
Diagonal Sequence Argument: subsequence x
n
i
such that (K
x
n
i
)
i=1
is Cauchy
for every N.
Claim: (Kx
n
i
)
i=1
is Cauchy.
> 0. Choose such that |K
K| <

3c
.
Choose Ni, j N : |K
x
n
i
K
x
n
j
| <

3
.
i,jN
|Kx
n
i
Kx
n
j
| |(K K
)x
n
i
|
. .
<|KK
||x
n
i
|<
3
|
+|K
x
n
i
K
x
n
j
|
. .
<
3
+|(K
K)x
n
j
|
. .
<
3
<
(iii): K compact K
compact.
Denote M := Kx [ |x| 1 Y .
M is a nonempty compact metric space. For y
denote f
y
:= y
[
M

C(M). Let T := f
y
[ |y
| 1 C(M).
Note:
|f
y
| := sup
yM
[f
y
(y)[
= sup
yM
y
, y)
= sup
|x|1
y
, Kx)
= sup
|x|1
K
, x)
= |K
|
T is bounded
f = f
y
T |f| = |K
| |K
| |y
|
..
1
|K|
T is equicontinuous
f = f
y
T, y
, |y
| 1.
[f(y
1
) f(y
2
)[ = [y
, y
1
y
2
)[
|y
||y
1
y
2
|
|y
1
y
2
|
Arzela-Ascoli T is compact in C(M)
!
K
is compact. y
n
Y
, |y
| 1
f
n
:= y
n
[
M
T f
n
has a convergent subsequence f
n
i
|K
n
i
| = |f
n
i
f
n
j
| (K
n
i
)
i=1
is a Cauchy sequence K
n
i
converges.
K
compact K compact K
compact K
compact.
X
K
//
//
Y
(i)

Y
K = K

X
: X Y
is compact.
If x
n
X is bounded
Y
(Kx
n
) has a convergent subsequence Kx
n
has a
convergent subsequence.
2
Lemma 2: X, Y Banach spaces, A L(X, Y )
(i) (ImA)
= Ker A

(ImA
) = Ker A
(ii) A
injective ImA dense in Y

(iii) A injective ImA
is weak
dense in X
57
Proof:
(i) Let y
(ImA)
, Ax)
. .
A
,y)
= 0 x X A
= 0
Let x X. Then x

(ImA
)
A
, x)
. .
=y
,Ax)
= 0 y
Hahn-Banach
Ax = 0
(ii) A
injective Ker A
= 0
(i)
(ImA)
= 0
ImA is dense in Y (Chap. II, Cor. 2 of Theorem 7)
(iii) A injective Ker A = 0 (

ImA
. .
weak
closure of ImA
= X
2
Example 1: X = Y = H = l
2
Ax := (x
1
,
x
2
2
,
x
3
3
, . . .) X
= Y

= H
A
= A injective, ImA ,= H
_
1
n
_
n=1
l
2
ImA
Example 2: X = l
1
, Y = c
0
, A : l
1
c
0
inclusion
A
: l
1
l
inclusion, ImA
= l
1
l
not dense
A
: (l
not injective.
When is ImA
?
=

(ImA
) or ImA
?
= (Ker A)
Theorem 2: X, Y Banach spaces, A L(X, Y )

Equivalent are:
(i) ImA is closed in Y
(ii) c 0 x X inf
A=0
|x +| c|Ax|
(iii) ImA
is weak
-closed in X
(iv) ImA
is closed in X
(v) c 0 y
inf
A
=0
|y
| c|A
|
If these equivalent conditions are satised, then:
ImA =

(Ker A
), ImA
= (Ker A)
Lemma 3: X, Y Banach spaces, A L(X, Y ), > 0

Assume y Y [ |y| < Ax [ |x| < 1 ()
Then: y Y [ |y| <

2
Ax [ |x| < 1
Proof: Chapter II, Lemma 2, Step 2 2
Remark 1: In Chapter II we proved:
A surjective
Baire
()
Lemma 3
A is open
Remark 2: () A is surjective
58
Proof of Theorem 2 (i)(ii) Denote X
0
:= X/ Ker A, Y
0
:= ImA
A induces an operator A
0
: X
0
Y
0
by A
0
[x] := Ax
Note: [x
1
] = [x
2
] x
1
x
2
Ker A Ax
1
= Ax
2
, so A
0
is well dened
A
0
is a bijective, linear operator
open mapping thm
A
1
0
: Y
0
X
0
bounded
c 0 x X : inf
A=0
|x +| = |[x]|
X/ Ker A
c|A
0
[x]| = c|Ax|
(ii)(iii) Claim: ImA
= (Ker A)
xKer A
x
[ x
, x) = 0
Let x
(Ker A)
i.e. x
, x) = 0 x Ker A
Dene : ImA 1 (Ax) := x
, x) well dened
is bounded: Ker A :
[(Ax)[ = [x
, x +)[ |x
||x +|
so [(Ax)[ |x
| inf
A=0
|x +|
(ii)
c|x
| |Ax|
Hahn-Banach
y
such that y
, y) = (y) y ImA
A
, x) = y
, Ax) = (Ax) = x
, x) x X
A
= x
ImA
(iii)(iv) obvious
(iv)(v) follows from (i)(ii) with A
instead of A
(v)(i) Case 1:A
is injective
If c > 0 y
|y
| c|A
|, then A is surjective
Claim: A satises () in Lemma 3 with =
1
c
(Then by Lemma 3, A is surjective)
Proof of the claim: Denote K := Ax [ |x| < 1 closed, convex, nonempty
To show: y
0
Y K |y
0
|
Let y
0
Y K
Chap. II, Thm 6
y
0
Y
such that y
0
, y
0
) > sup
yK
y
0
, y)
|A
0
| = sup
|x|<1
A
0
, x) = sup
|x|<1
y
0
, Ax) = sup
yK
y
0
, y) < y
0
, y
0
) |y
0
| |y
0
|
|y
0
| >
|A
0
y
0
|
|y
0
|

1
c
=
Case 2: A
not injective
Denote Y
0
:= ImA
= Y
/(ImA)
= Y
/ Ker A
A
0
: X Y
0
A
0
: Y
0
= Y
/ Ker A
0
is the operator induced by A
By (v) we have: |[y
]|
Y
/ Ker A
c|A
|
A
0
satises the hypotheses of Case 1
A
0
is surjective ImA = ImA
0
= Y
0
= ImA ImA is closed
ImA =

((ImA)
) =

(Ker A
) 2
Corollary: X, Y Banach spaces, A L(X, Y )
(i) A is surjective if and only if c > 0 y
|y
| c|A
|
(ii) A
is surjective if and only if c > 0 x X|x| c|Ax|

Proof:
(i) A is surjective ImA closed and ImA dense
(v) in Theorem 2 and Ker A
= 0
(ii) A
is surjective ImA
weak
closed and ImA
weak
dense
(ii) in Theorem 2 and Ker A = 0
2
Remark 1: X, Y, Z Banach spaces, A L(X, Y ), B L(Z, Y ), ImB ImA
T L(Z, X) AT = B (Douglas Factorization)
Hint: T := A
1
B : Z X is closed
59
Remark 2: A L(X, Y ), B L(X, Z)
Equivalent are:
(i) ImB
ImA
(ii) c 0 x X|Bx| c|Ax|

Hint for the proof:B = id see Corollary (ii)
(i)(ii) Douglas Factorization (when A
is injective)
(ii)(i) Prove that
x
ImA
c 0 x X[x
, x)[ c|Ax| as in Proof of Theorem 2

Remark 3: X reexive, A L(X, Y ), B L(Z, Y )
Equivalent are:
(i) ImB ImA
(ii) c 0 y
|B
| c|A
|
Hint for the proof: (ii)
Rem 2
ImB
ImA
Example: X reexive cannot be removed in Remark 3:

X = c
0
, Y = l
2
, Z = 1
A : X Y Ax :=
_
x
n
n
_
n=1
B : Z Y Bz :=
_
z
n
_
n=1
l
2
A, B satisfy (ii) in Remark 3, but not (i)
4.2 Fredholm operators
Denition: X, Y Banach spaces and A L(X, Y ). ker A := x X [ Ax =
0, imA := Ax [ x X. Dene coker A := Y/A. A is called a Fredholm
operator if
imA is a closed subspace of Y
ker A and coker A are nite dimensional
The Fredholm index of A is the integer index(A) := dimker Adimcoker A.
dimcoker A < imA closed.
Proof: dimcoker A < y
1
, . . . , y
m
Y such that [y
i
] Y imA form a
basis
Y = imAspany
1
, . . . , y
m
Denote X := X 1
m
, (x, ) X
|(x, )|
X
:= |x|
X
+||
R
m.
Dene A : X Y by A (x, ) := Ax +
m
i=1
i
y
i
A L(X, Y ). A is surjective and ker A = ker A0.
Thm2
c 0x X : inf
A=0
|x +|
X
c|Ax|
Y
i.e.
x X 1
m
: inf
A=0
|x +|
X
+||
R
m c|Ax +
i
y
i
|
Y
inf
A=0
|x +|
X
c|Ax|
Y
Thm2
imA closed
2
Remark: Y Banach space, Y
0
Y linear subspace, dimY/Y
0
<
, Y
0
is closed.
60
A Fredholm A
is Fredholm.
In this case: index(A
) = index(A).
Proof: By Thm 2 imA closed imA
closed. In this case we have im(A
) =
(ker(A))
and ker(A
) = (im(A))
perp
. Hence
(ker A)

= X
/(ker A)
= X
/(imA
) = coker(A
)
(coker A)
= (Y/ imA)

= (imA)
= ker(A
)
dimcoker(A
) = dimker(A) dimker(A
) = dimcoker(A) 2
Example 1: X, Y nite dimensional Every linear operator A : X Y is
Fredholm and index(A) = dimX dimY .
Proof: dimX = dimker A+ dimimA (by Linear Algebra) 2
Example 2: X, Y arbitrary Banach spaces, A L(X, Y ) bijective A Fred-
holm and index(A) = 0
Example 3: Hilbert spaces X = Y = H =
2
x = (x
1
, x
2
, . . .)
Dene A
k
: H H by A
k
x := (x
k+1
, x
k+2
, . . .) shift A
k
is Fredholm and
index(A
k
) = k
H

= H
with x, y) =
i=1
x
i
y
i
. So A
: H H is dened by x, A
y) :=
Ax, y).
A
k
y = (0, . . . , 0
. .
k, x
1
, x
2
, . . .) so A
k
y =: A
k
y, where A
k
is Fredholm and
index(A
k
) = k.
Lemma (Main Lemma) 5: X, Y Banach spaces and D L(X, Y ). Equiv-
alent are:
(i) D has a closed image and a nite dimensional kernel.
(ii) Banach space Z and compact operator K L(X, Z) and c 0 such
that x X |x|
X
c (|Dx|
Y
+|Kx|
Z
)
(*)
Proof : (i)(ii): Z := 1
m
m := dimker D < . Choose an isomorphism
: ker D 1
m
.
Hahn-Banach
bounded linear operator K : X 1
m
such that
Kx = xx ker D. Dene D : X Y 1
m
by D x := (Dx, Kx)
imD = imD 1
m
closed and ker D = 0
Thm2
c 0x X with
|x|
X
c|D x|
Y R
m = c (|Dx|
Y
+|Kx|
R
m)
(ii)(i):
Claim 1 Every bounded sequence in ker D has a convergent subsequence.
( dimker D < , by Chapter I, Thm 1)
Proof of Claim 1 Let x
n
ker D be a bounded sequence
K compact
subse-
quence (x
n
i
)
i=1
such that (Kx
n
i
)
i
converges
(Dx
n
i
)
i
and (Kx
n
i
)
i
are Cauchy sequences
(x
n
i
) is Cauchy, because
|x
n
i
x
n
j
| c|Dx
n
i
Dx
n
j
| +c|Kx
n
i
Kx
n
j
|
X complete
(x
n
i
) converges.
Claim 2 C > 0x X : inf
D=0
|x +| C|Dx| (By Thm 2, this implies
that imD is closed).
61
Proof of Claim 2 Suppose not. n Nx
n
X such that inf
D=0
|x
n
+
| > n|Dx
n
|.
Without loss of generality we can assume inf
D=0
|x
n
+| = 1 and 1 |x
n
| 2
|Dx
n
| <
1
n
so
a) Dx
n
0
b) subsequence (x
n
i
)
i
such that (Kx
n
i
) converges
(Dx
n
i
)
i
and (Kx
n
i
)
i
are Cauchy
()
(x
n
i
)
i
is Cauchy
(x
n
i
) converges. Denote x := lim
i
x
n
i
Dx = lim
i
Dx
n
i
a)
= 0 and 1 = inf
D=0
|x
n
i
+| |x
n
i
x| 2
Theorem 3: (another characterization of Fredholm operators) X, Y Banach
spaces, A L(X, Y ) bounded linear operator. Equivalent are:
(i) A is Fredholm
(ii) F L(Y, X) such that 1
X
FA, 1
Y
AF are compact.
Proof: (i)(ii):
X
0
:= ker A X is nite dimensional. So a closed subspace X
1
X such
that X = X
0
X
1
.
Y
1
:= imA Y is a closed subspace of nite codimension. So nite dimen-
sional subspace Y
0
Y such that Y = Y
0
Y
1
.
Consider the operator A
1
:= A[
X
1
: X
1
Y
1
. Then A L(X
1
, Y
1
) and A
1
is
bijective.
open mapping
A
1
1
L(Y
1
, X
1
)
Dene F : Y X by F(y
0
+y
1
) := A
1
1
y
1
for y
0
Y
0
, y
1
Y
1
.
Then F L(Y, X):
FA(x
0
+x
1
) = FA
1
x
1
= x
1
AF(y
0
+y
1
) = AA
1
1
y
1
= y
1
1
X
FA =
X
0
: X

= X
0
X
1
X
0
X and 1
Y
AF =
Y
0
: Y

=
Y
0
+Y
1
Y
0
Y compact
(ii)(i):
K := 1
X
FA L(X, Y ) compact
|x|
X
= |FAx +Kx|
X
|FAx|
X
+|Kx|
X
|F| |Ax|
Y
+|Kx|
X
c(|Ax|
Y
+|Kx|
X
)
where C := max1, |F|
Lemma5
A has a nite dimensionl kernel and a closed image.
L := 1
Y
AF L(Y ) compact
Thm1
L
is compact and y
= L
+F
with c := max1, |L| we have

|y
| c(|A
||L
|)
lemma5
dimker A
< with dimker A
= dimcoker A.
2
Theorem 4: X, Y, Z Banach spaces,
A L(X, Y ), B L(Y, Z) Fredholm operators
BA L(X, Z) is a Fredholm operator and index(BA) =index(A)+index(B)
62
Proof: By Theorem 3, F L(Y, X), G L(Z, Y ) such that
1
X
FA, 1
Y
AF, 1
Y
GB, 1
Z
BG are compact
a) 1
X
FGBA = 1
X
FA
. .
compact
+F(1
Y
GB
. .
compact
)A compact by Theorem 1
b) 1
Z
BAFG = 1
Z
BG+B(1
Y
GB)A is compact
BA is Fredholm by Theorem 3.
Proof of the index formula:
A
0
: ker BA/ ker A ker B [x] Ax
B
0
: Y/ imA imB/ imBA [y] [By]
A
0
is injective, B
0
is surjective
imA
0
= imA ker B
coker A
0
= ker B/(imA ker B)
ker B
0
= [y] Y/ imA [ By imBA
= [y] Y/ imA [ x X such that By = BAx
= [y] Y/ imA [ x X such that y Ax ker B = (imA+ ker B)/ imA
= ker B/(imA ker B) = coker A

0
0 = dimker B
0
dimcoker A
0
dimcoker B
0
+ dimker A
0
= index A
0
+ index B
0
= dim(ker BA/ ker A) dimker B + dimcoker AdimimB/ ker BA
= dimker BAdimker Adimker B+dimcoker Adimker Y/ imBA+dimY/ imB
= index BAindex Aindex B 2
Theorem 5 (Stability): X, Y Banach spaces, let D L(X, Y ) be a Fred-
holm operator
(i) > 0 such that P L(X, Y ) we have
|P| < D +P is Fredholm and index(D +P) = index(D)
(ii) If K L(X, Y ) is a compact operator, then D + K is Fredholm and
index(K +P) = index(D)
Proof: (i) By Lemma 5, Banach space Z, compact operator L L(X, Z),
c > 0 such that |x| c(|Dx| +|Lx|) x X
(because imD closed, dimker D < )
Suppose P L(X, Y ) with |P| <
1
c
Then |x| c(|Dx| +|Lx|) c(|(D +P)x| +|Px| +|Lx|)
c(|(D +P)x| +|Lx|) +c|P| |x|
(1 c|P|)|x| c(|(D +P)x| +|Lx|)
|x|
c
1c|P|
(|(D +P)x| +|Lx|)
Lemma 5
D +P has a closed image and a nite dimensional kernel
(provided |P| <
1
c
)
(coker D)
= (Y/ imD)
= (imD)
= ker D
dimker(D +P)
< for |P| = |P
| suciently small
(by the same argument as for D +P)
Index formula:
X = X
0
X
1
X
0
= ker D, Y = Y
0
Y
1
Y
1
= imD
P
ji
: X
i
X
P
Y

j
Y
j
Then P(x
0
+x
1
) = P
00
x
0
+P
01
x
1
. .
Y
0
+P
10
x
0
+P
11
x
1
. .
Y
1
P =
_
P
00
P
01
P
10
P
11
_
D =
_
0 0
0 D
11
_
D
11
: X
1
Y
1
is bijective
1 +D
1
11
P
11
is bijective as well for |P
11
| small.
63
D
11
+P
11
is bijective for |P| small.
Let x = x
0
+x
1
x
0
X
0
, x
1
X
1
. Then:
(D +P)x = 0 P
00
x
0
+P
01
x
1
= 0, P
10
x
0
+ (P
11
+D
11
)x
1
= 0
x
1
= (D
11
+P
11
)
1
P
10
x
0
, (P
00
P
01
(D
11
+P
11
)
1
P
10
)x
0
= 0
Denote A
0
:= P
00
P
01
(D
11
+P
11
)
1
P
10
: X
0
Y
0
Then ker(D +P) = x
0
(D
11
+P
11
)
1
P
10
x
0
[ x
0
ker A
0
so dim ker(D +P)=dim ker A

0
Let x = x
0
+x
1
x
0
X
0
, x
1
X
1
y = y
0
+y
1
y
0
Y
0
, y
1
Y
1
Then: y = (D +P)x y
0
= P
00
x
0
+P
01
x
1
, y
1
= P
10
x
0
+ (D
11
+P
11
)x
1
y
0
= P
00
x
0
+P
01
(D
11
+P
11
)
1
(y
1
P
10
x
0
), x
1
= (D
11
+P
11
)
1
(y
1
P
10
x
0
)
y
0
= A
0
x
0
+P
01
(D
11
+P
11
)
1
y
1
, x
1
= (D
11
+P
11
)
1
(y
1
P
10
x
0
)
Hence y im(D +P) y
0
P
01
(D
11
+P
11
)
1
y
1
imA
0
()
This implies coker A
0

= coker(D +P)!
Indeed, choose a subspace Z Y
0
Y such that Y
0
= imA
0
Z
Then by () Y = im(D +P) Z
im(D +P) Z = 0 :
y = y
0
+y
1
im(D +P) Z, then y
0
P
01
(D
11
+P
11
)
1
y
1
imA
0
, y
1
= 0
so y
0
Z imA so y
0
= 0, y
1
= 0
Given y = y
0
+y
1
Y , write
y
0
P
01
(D
11
+P
11
)
1
y
1
= A
0
x
0
+z z Z, then
(y
0
z) +y
1
= y z
()
im(D +P)
Hence index(D +P) = index A
0
= dimX
0
dimY
0
= dimker D dimcoker D = index D
(ii)By Theorem 3 T L(X, Y ) such that 1
X
TD, 1
Y
DT compact
1
X
T(D +K), 1
Y
(D +K)T compact
Thm 3
D +K Fredholm
T
k
:= A L(X; Y ) [ A Fredholm, index A = k
open subset of L(X, Y ) by (i)
kZ
T
k
=: T(X, Y ) = Fredholm operators X Y
We have proved D +tK T(X, Y ) t 1
Consider the map : 1 T(X, Y ) t D +tK
is continuous
So I
k
:=
1
(T
k
) = t 1 [ index(D +K) = k is open k Z
1 =
kZ
I
k
disjoint union
each I
k
is open and closed
k Z such that I
k
= 1 index(D) = k = index(D +K) 2
Example: X Banach space, K L(X) compact
1K Fredholm and index(1K) = 0 dimker(1K) = dimcoker(1K)
Fredholm alternative
Either the equation x Kx = y has a unique solution y Y
or the homogeneous equation x Kx = 0 has a nontrivial solution
Application to integral equations like x(t) +
_
1
0
k(t, s)x(s) ds = y(t) 0 t 1
Remark: We well prove in Chapter V: m 0 such that:
index(1 K)
m
= index(1 K)
m+1
By Theorem 4 this implies mindex(1 K) = (m+ 1) index(1 K)
so it follows also index(1 K) = 0
64
5 Spectral Theory 10.01.2007
5 Spectral Theory
5.1 Eigenvectors
For this whole chapter: X complex Banach space and |x| = [[|x| C.
L(X, Y ) = A : X Y [ A complex linear bounded, X
= L(X, C) and
L(X) := L(X, X).
Let A L(X):
C eigenvalue x X x ,= 0 Ax = x
Denition: Let X complex Banach space and A L(x). The spectrum of A
is the set
(A) := C [ 1 A is not bijective = P(A) R(A) C(A)
P(A) = C [ 1 A is not injective point spectrum
R(A) = C [ 1 A is injective and im(1 A) ,= X residual spectrum
C(A) = C [ 1 A is injective and im(1 A) = X, im(1 A) ,= X continuous spectrum
(A) := C (A) = C [ 1 A is bijective resolvent set
Remark : X real Banach space, A L(X) bounded real linear operator.
spectrum of A: (A) := (A
C
A
C
: X
C
X
C
complexied operator
X
C
= X X = X iX x +iy
A
C
(x +iy) := Ax +iAy
Example 1: X =
2
C
(x
1
, x
2
, x
3
, . . .) and = (
1
,
2
, . . .) bounded sequence
in C. Set Ax := (
1
x
1
,
2
, x
2
, . . .), P(A) =
1
,
2
, . . ., (A) = P(A)
Example 2: X =
2
C
, | := z C [ [z[ 1
Ax := (x
2
, x
3
, . . .) Bx := (0, x
1
, x
2
, . . .)
P(A) = int(|) P(B) =
R(A) = R(B) = int(|)
C(A) = S
1
C(B) = S
1
(A) = | = (B)
Lemma 1: A L(X). Then (A
) = (A).
Proof:
(1 A)
= 1 A
Claim
A bijective A
bijective (A
1
)
= (A
)
1
A bijective imA closed imA = X ker A = 0
ChIV,Thm2,Lemma2
imA
weak*closed ker A
= 0 imA
weak*dense
A
bijective
(A
1
)
, y) = x
, A
1
y)
= A
(A
)
1
x
, A
1
y)
= (A
)
1
x
, AA
1
y)
= (A
)
1
x
, y)y X, x
65
(A
1
)
= (A
)
1
x
2
Remark:
a) P(A
) P(A) R(A)
R(A
) P(A) C(A)
C(A
) C(A)
b) P(A) (A
) R(A
)
R(A) P(A
)
C(A) P(A
) C(A
)
5.2 Integrals
Lemma 2: X Banach space, x : [a, b] X continuous
(*)
! Xx
: x
, ) =
b
_
a
x
, x(t)) dt
Notation:
b
_
a
x(t) dt := is called the Integral of x.
Proof (exercise with hint): Dene
n
:=
2
n
1
k=0
ba
2
n
x(a +
k(ba)
2
n
)
n
:= sup
]st]
ba
2
n
|x(t) x(s)| 0
Show that: |
n+m

n
| (b a)
n
m > 0
n
is Cauchy.
Denote := lim
n
n
and check (*). 2
Remark:
1.
b
_
a
x(t) dt +
c
_
b
x(t) dt =
c
_
a
x(t) dt
2.
b
_
a
x(t) +y(t) dt =
b
_
a
x(t) dt +
b
_
a
y(t) dt
3. : [, ] [a, b] C
1
-map
x : [a, b] X continuous. Then
()
_
()
x(t) dt =
x((s))

(s) ds
4. If x : [a, b] X is continuously dierentiable, then
b
_
a
x(t) dt = x(b) x(a)
5. A L(X, Y ) and x : [a, b] X continuous. Then
b
_
a
Ax(t) dt = A
b
_
a
x(t) dt
6. |
b
_
a
x(t) dt|
b
_
a
|x(t)| dt
66
Proof: By denition: x
,
b
_
a
x(t) dt)
b
_
a
x
, x(t) dt
b
_
a
[x
, x(t))[ dt
b
_
a
|x
| |x(t)| dt
= |x
|
b
_
a
|x(t)| dt
Now use:
_
_
_
_
_
_
b
_
a
x(t) dt
_
_
_
_
_
_
= sup
x
,=0
[
b
_
a
x
, x(t)) dt
|x
|
2
Notation: Let X be a Banach space, C an open subset, f : X a
continuous map, and : [a, b] be continuously dierentiable.
Denote:
_
f(z) dz :=
b
_
a
f((t)) (t) dt X
Remark:
_
_
_
_
_
_
_
f(z) dz
_
_
_
_
_
_
l() sup a t b|f((t))|
l() =
b
_
a
[ [ dt
Denition: C open set. X complex Banach space, f : X is called
holomorphic if the limit f
(z) := lim
Ch0
f(z+h)f(z)
h
exists for every z
and the map f
: X is continuous.
Lemma 3: X, Y complex Banach spaces. C open set. A L(X, Y )
continuous. Equivalent are:
(i) A is holomorphic
(ii) For every x X and every y
the function C : z
y
, A(z)x) is holomorphic
(iii) Proof B
r
(z
0
) = z C [ [z z
0
[ r then A(z
0
) =
1
2i
_
A(z)
zz
0
dz,
where (t) := z
0
+r e
2it
, 0 t 1.
Exercise: A : X holomorphic A is C
and
A
(n)
(z
0
)
n!
=
1
2i
_
A(z)
(z z
0
)
n+1
dz as in (iii)
67
Proof: of Lemma 3 (i)(ii): obvious
(ii)(iii): Cauchy integral formula of Complex Analysis.
(iii)(i): usual argument from Complex Analysis.
Denote B :=
1
2i
_
A(z)
(zz
0
)
2
dz with as in (iii).
Claim For 0 < [h[ < r and c := sup
]zz
0
]<r
|A(z)| we have
|
1
h
(A(z
0
+h) A(z)) B|
c[h[
r(r [h[)
This implies A
(z
0
) =
1
2i
_
A(z)
(zz
0
)
2
dz.
Continuity of A
: Exercise.
Proof of Claim
1
h
(A(z
0
+h) A(z
0
)) B
(iii)
=
1
2i
_
1
h
(
1
z z
0
h

1
z z
0
)
1
(z z
0
)
2
A(z) dz
=
1
2i
_
h
(z z
0
)
2
(z z
0
h)
A(z) dz
|
1
h
(A(z
0
+h) A(z
0
)) B|
1
2
l() sup
0t1
[h[|A((t))|
[(t) z
0
[
2
[(t) z
0
h[
[h[
r
sup
0t1
|A((t))|
[re
2it
h[

c[h[
r(r [h[)
2
Lemma 4: A L(X) (A) C is open, the function
(A) L(X) : (1 A)
1
=: R
(A) is holomorphic, and

(*) R
(A) R
(A) = ( )R
(A)R
(A), (A)
Proof: Proof of (*)
(1 A)(R
(A) R
(A))(1 A) = (1 A) (1 A) = ( ) 1 ()
(A) is open:
(A), [ [|(1 A)
1
| < 1
ChI,Thm11
1 +( )(1 A)
1
is bijective,
where ( ) = (1 A)(1 A)
1
1 A bijective (A)
Also:
|(1 A)
1
(1 A)
1
|
]]|(1A)
1
|
2
1]]|(1A)
1
|
Continuity
(*) lim
(A)R
(A)
= R
(A)
2
Continuity
R
(A) is holomorphic. 2
Theorem 1: Let A L(X), then
1. (A) ,= and
2. sup
(A)
[[ = lim
n
|A
n
|
1
n
= inf
nN
|A
n
|
1
n
=: r
A
(Spectral radius, Chapter I, Theorem 11)
68
Proof:
1. sup
(A)
[[ r
A
Let C with [[ > r
A
r
1
A
< 1
Chap 1 Thm 11
1
1
A bijective (A)
2. r
A
sup
(A)
[[
R(z) :=
_
(z
1
1 A)
1
z ,= 0,
1
z
(A)
0 z = 0
R : L(X) =
1
z
[ z (A) 0
Fact: open and R : L(X) is holomorphic
Proof: Lemma 4 and removal of singularity for holom. 0 C
Or for small z :
R(z) = z(1 zA)
1
=
i=1
z
k+1
A
k
(Chap. I, Thm. 11)
R
(n)
(0)
n!
= A
n1
Cauchy integral formula
A
n1
=
1
2i
_
R(z)
z
n+1
dz (1)
=
1
r
e
2it
t [0, 1] provided
_
z C [ [z[
1
r
_
(2)
If r > sup
(A)
[[ then (2) holds, so
|A
n1
| =
_
_
_
1
2i
_
R(z)
z
n+1
dz
_
_
_ l() sup
]z]=
1
r
|R(z)|
2]z
n+1
]
= l()
r
n
2
1
r
sup
]z]=
1
r
|(z
1
1 A)
1
| = r
n
sup
]]=r
(1 A)
1
|
. .
=:c
|A
n1
| cr
n
|A
n
| cr
n+1
|A
n
|
1
n
r (cr)
1
n
. .
1
lim
n
|A
n
|
1
n
r r > sup
(A)
[[
3. (A) ,=
Suppose (A) = and so (A) = C.
Pick x X, x
and dene
f : C C f() := x
, (1 A)
1
x)
Lemmma 4
f is holomorphic and
[f()[ |x
| |x| |(1 A)
1
|
|x
| |x|
]]|A|
if [[ > |A|
Liouville
f 0 (1 A)
1
= 0 C Contradiction
2
H complex (real inner product) Hilbert space |ix| = |x| x H
ix, iy) = x, y) x, y H
recover inner product from norm: x, y) =
1
4
(|x +y|
2
|x y|
2
)
ix, y) +x, iy) = 0 x, y H
Dene the Hermitian inner product:
H H C(x, y) x, y)
C
by x, y)
C
:= x, y) +iix, y)
a) real bilinear
b) x, y)
C
= x, y)
C
c) x, y)
C
= x, y)
C
, x, y)
C
= x, y)
C
d) [x, y)
C
[ |x| |y|
e) |x| =
_
x, x)
C
69
Proof: Exercise 2
H
:= : H C [ complex, linear
Dene : H H
(x), y)
H
,H
:= x, y)
C
Riesz Representation Theorem: : H H
isometric isomorphism
Warning: is anti-linear: (x) = (x) x H C
The Hilbert space adjoint of a linear operator A : H H is the operator
A
: H H dened by A
x, y)
C
:= x, Ay)
C
x, y H
Remark:
(i) A
new
=
1
A
old

(ii) (A)
old
= A
old
(A)
new
= A
new
(iii) (A
old
) = (A) (A
new
) = (A)
Proof: 1 A
new
(i)
=
1
(1 A
old
) bijective
1 A
old
bijective
Lemma 1
1 A bijective 2
Denition: H complex Hilbertspace
An operator A L(H) is called normal , if AA
= A
A, selfadjoint, if A = A
and unitary, if AA
= A
A = I.
Example 1: H = l
2
C
, Ax = (
1
x
1
,
2
x
2
, . . .)
n
C, sup
nN
[
n
[ <
A = A

n
1
Because A
x = (
1
x
1
,
2
x
2
, . . .)
x, y)
C
=
n=1
x
n
y
n
A is normal
Example 2: H = l
2
C
Ax = (x
2
, x
3
, . . .) A
x = (0, x
1
, x
2
, . . .) AA
,= A
A
Example 3: selfadjoint normal
A unitary (|Ax| = |x| x H) isomorphism A normal, because
x, A
Ax) = |Ax|
2
= |x|
2
y, A
Ax) = y, x) x, y H
A
A = 1
A onto
A A
= 1 A
= A
1
In Example 2 we have |A
x| = |x| but A
,= A
1
(A
is not onto)
Lemma 5: H complex Hilbert space, A L(H)
Equivalent are:
(i) A is normal
(ii) |Ax| = |A
x| x H
Proof:
(i)(ii) |Ax|
2
= Ax, Ax)
C
= x, A
Ax)
C
= x, AA
x)
C
= A
x, A
x)
C
= |A
x|
2
(ii)(i) Same argument gives x, A
Ax)
C
= x, AA
x)
C
x H
y, A
Ax)
C
= y, AA
x)
C
x, y H
A
A = AA
2
70
Theorem 2: H complex Hilbert space, A L(H) normal
(i) sup
(A)
[[ = |A|
(ii) A = A
(A) 1
Proof:
(i) Claim: |A
n
| = |A|
n
n N (Then (i) follows from Theorem 1)
n = 2: |A|
2
= sup
x,=0
|Ax|
2
|x|
2
= sup
x,=0
x,A
Ax)
|x|
2
CS
sup
x,=0
|A
Ax|
2
|x|
2
Lemma 5
= sup
x,=0
|A
2
x|
|x|
2
= |A
2
|
|A
2
|
2
normal
|A
2
|
always
|A|
2
n = 2
m
Induction |A
2
m
| = |A|
2
m
n arbitrary:
|A|
n
=
|A|
2
m
|A|
2
m
n
=
|A
2
m
|
|A|
2
m
n

|A
n
| |A
2
m
n
|
|A|
2
m
n
|A
n
|
(ii) Assume A = A
, let C 1
To show: (A)
Claim: |x Ax|
2
(Im)
2
|x|
2
Claim
Chap. IV Thm 2
1 A injective, im(1 A) closed
A=A
1 A
injective, so 1 A bijective
Proof of Claim: |x Ax|
2
= x Ax, x Ax)
C
= x, x)
C
Ax, x)
C
x, Ax)
C
+Ax, Ax)
C
= [[
2
|x|
2
2(Re )x, Ax)
C
+ |Ax|
2
= (Im)
2
|x|
2
+ (Re )
2
|x|
2
2(Re )x, Ax)

C
+|Ax|
2
= (Im)
2
|x|
2
+|(Re )x Ax| (Im)
2
|x|
2
2
Theorem 3: H complex Hilbert space, A = A
L(H) selfadjoint. Then

a) R(A) = and
b) (1) sup(A) = sup
|x|=1
x, Ax)
(2) inf (A) = inf
|x|=1
x, Ax)
(3) |A| = sup
|x|=1
[x, Ax)[
Proof:
a) R(A) =
/ P(A) 1 A injective
H = ker(1 A)
= x H [ x, ) = 0 ker(1 A)
= im(1 A)
C(A)
(1)(2)
replace A by A.
71
(1)&(2)(3)
|A|
Thm2
= sup
(A)
[[
= maxsup (A), inf (A)
(1),(2)
= sup
|x|=1
[x, Ax)[
Proof of (1) Assume wlog
(4) x, Ax) 0 x H
(replace A by A+|A| 1).
Claim 1: (4) (A) [0, )
Claim 2: (4) |A| = sup
|x|=1
x, Ax)
Claims 1&2 (1)
sup
|x|=1
x, Ax) = |A| by Claim 2
= sup
(A)
[[ by Thm 2
= sup (A) by Claim 1
Proof of Claim 1 Let > 0. Then x H:
|x|
2
(4)
|x|
2
+x, Ax)
= x, x +Ax)
|x| |x +Ax|
|x| |x +Ax|x H
1l +A is injective and has a closed image (see Chapter IV).
A=A
1 +A is bijective
/ (A)
Proof of Claim 2 Let a := sup
|x|=1
x, Ax) |A| (Cauchy-Schwarz). To
show: |A| a.
For x, y H we have
y, Ax) =
1
4
(x +y, A(x +y)) x y, A(x y)))
For |x| = |y| = 1:
a
1
4
a|x y|
2

1
4
x y, A(x y)) y, Ax)
1
4
x +y, A(x +y))
1
4
a|x +y| a
[y, Ax)[ ax, y H and |x| = |y| = 1
|A| = sup
|x|=1,
|y|=1
[y, Ax)[ a
2
72
Remark: Given any operator T L(H) we have
|T|
2
= sup
|x|=1
|Tx|
2
= sup
|x|=1
Tx, Tx)
= sup
|x|=1
x, T
Tx)
|T
T|
|T
||T|
|T|
2
|T|
2
sup
|x|=1
x, T
Tx)
Thm 3 for A=T
T
= |T
T|
|T| =
_
|T
T|. We can use this to compute |T|.

5.3 Compact operators on Banach spaces
X complex Banach space, A L(X) bounded, complex linear.
Facts
a) ker(1l A) ker(1l A)
2
ker(1l A)
3
. . ..
b) ker(1l A)
m
= ker(1l A)
m+1
ker(1l A)
m
= ker(1l A)
m+k
k 0
Notation: E
:= E
(A) :=
m=1
ker(1l A)
m
Theorem 4: A L(X) compact
(i) If (A), ,= 0, then P(A) and dimE
< .
Hence m N such that ker(1l A)
m
= ker(1l A)
m+1
.
(ii) Eigenvalues of A can only accumulate at 0, ie
(A), ,= 0, > 0 C :
0 < [ [ < (A)
Proof:
(i) If ,= 0 1l Fredholm, index = 0.
ChIV
1l A Fredholm, index = 0
Thm5A cpct
either 1l A is bijective ( / (A)) or 1l A is not injective
( P(A)). Moreover: dim(ker(1l A)
m
) < m N, because
(1l A)
m
=
m
k=0
_
m
k
_
k
(A)
mk
=
m
1l + cpct
Let K :=
1
A and E
n
= ker(1l K)
n
= ker(1l A)
n
.
To show: m such that E
m
= E
m+1
.
Suppose not. Then E
n
_ E
n+1
n N
ChIILemma4
n Nx
n
E
n
such that
|x
n
| = 1 inf
xE
n1
|x
n
x|
1
2
Now: for m < n we have
Kx
m
E
m
E
n1
x
n
Kx
n
E
n1
73
|Kx
n
Kx
m
| = |x
n
(x
n
Kx
n
+Kx
m
. .
E
n1
|
1
2
So (Kx
n
)
n
has no convergent subsequence: contradiction!
(ii) Let (A), ,= 0.
(i)
m N such that ker(1 A)
m
= ker(1 A)
m+1
X = ker(1 A
. .
=:X
0
)
m
im(1 A
. .
=:X
1
This is an exercise (use Hahn-Banach).
AX
0
X
0
AX
1
X
1
.
Note: (1 A)
m
: X
1
X
1
is bijective
open mapping
c > 0x
1
X
1
|x
1
| c|(1 A)
m
x
1
|
Choose > 0 such that for all C:
[ [ < |(1 A)
m
(1 A)
m
| <
1
c
(1 A)
m
: X
1
X
1
is bijective for [ [ <
(1 A)
m
: X
0
X
0
is bijective for ,=
The rest by induction. X
0
is nite dimensional. A[
X
0
=
(1 A)
m
: X X only true if bijective!
1 A : X X is bijective.
2
H real or complex Hilbert space
Notation: ., .) real or Hermitian inner product.
Denition: A collection of vectors e
i
iI
in H is called an orthonormal basis
if
(1)
e
i
, e
j
) =
ij
(2)
H = spane
i
[ i I
Remark 1: (2) holds if and only if
x H : x, e
i
) = 0i I x = 0
Remark 2: H separable I is nite or countable
Remark 3: x H, e
i
iI
ONB
x =
iI
e
i
, x)e
i
|x|
2
=
iI
[e
i
, x)[
2
Theorem 5: H real or complex Hilbert space. A = A
L selfadjoint,
compact. A admits an ONB e
i
iI
of eigenvectors
Ae
i
=
i
e
i
,
i
1 Ax =
iI
i
e
i
, x)e
i
74
Proof:
Step 1
Ax = x x ,= 0 Ay = y y ,= 0 and ,=
x, y) = 0
proof
( )x, y) = x, y) x, y)
= Ax, y) x, Ay)
Step 2 ker(1 A)
m
= ker(1 A)m 1
proof
1 P(A) (1 A)
2
x = 0
0 = x, (1 A)
2
x)
= (1 A)x, (1 A)x) = |x Ax|
2
(*)
Step 3
x, y) = 0y ker(1 A) 1 x = 0
H
0
:= x H [ () H
0
,= 0 A[
H
0
,= 0
|A[
H
0
| (A[
H
0
) or |A[
H
0
| (A[
H
0
A[
H
0
has a nonzero eigenvalue, eigenvector. This is also an eigenvector of A:
contradiction! 2
Denition: A C
-algebra is a complex Banach space / equipped with

an assoziative, distributive product / / / : (a, b) ab with a unit
1 / such that |ab| |a| |b|
a complex anti-linear involution / / : a a
such that
(ab)
= b
, 1
= 1, and |a
| = |a|
Remark: antilinear: (a)
= a
, involution: a
= a
Example 1: H complex Hilbert space, then L(H) is a C
-algebra
Example 2: A L(H)
/ := smallest C
-algebra containing A
A = A
, p() := a
0
+a
1
+. . . +a
n
n
a
k
C
p(A) := a
0
+a
1
A+. . . +a
n
A
n
p(A)
= a
0
+a
1
A+. . . +a
n
A
n
(pq)(A) = p(A)q(A)
/ := closure(p(A) [ p : 1 C)
Example 3: compact metric space
C() := f : C [ f continuous
C
-Algebra, sup-norm, involution: f f

Goal: A = A
, = (A) /

= C() : p(A) p
75
Theorem 6: H complex Hilbert space, A = A
L(H) selfadjoint
:= (A) (1)
= There is a unique bounded linear operator
C() L(H) : f f(A) ()
such that (fg)(A) = f(A)g(A), 1l
R
(A) = 1
H
(1)
f(A) = f(A)
(2)
f() = f(A) = A (3)
Denote
A
(f) := f(A)
Then () is the operator
A
: C() L(H)
(1):
A
(fg) =
A
(f)
A
(g)
A
(1) = 1
(2):
A
(f) =
A
(f)
(3):
A
(id : C) = A
Lemma 6: H complex Hilbert space, A L(H)
p() =
n
k=0
a
k
k
a
k
C complex polynomial
(i) p(A)
= p(A
), p() =
n
k=0
a
k
k
, (pq)(A) = p(A)q(A)
(ii) (p(A)) = p((A)) = p() [ (A)
(iii) A = A
|p(A)|
L
((A))
= sup
(A)
[p()[
Proof:
(i) Exercise
(ii) (A), to show p() (p(A))
The polynomial t p(t) p() vanishes at t =
polynomial q such that p(t) p() = (t )q(t)
p(A) p()1 = (A1)q(A) = q(A)(A1)
p(A) p()1 is not bijective p() (p(A))
(p(A)) (A) : = p()
n := deg(p) p(t) = a(t
1
) . . . (t
n
) a ,= 0
p(A) 1 = a(A
1
1) . . . (A
n
1) not bijective
i such that A
i
1 not bijective

i
(A), p(
i
) = 0
(iii) A = A
p(A)
= p(A)
p(A) is normal: q(A)p(A) = (pq)(A)) = p(A)q(A)
so p(A)p(A)
= p(A)
p(A)
Thm 2
|p(A)| = sup
(p(A))
[[
(ii)
= sup
(A)
[p()[
2
Remark 1: If p() = q() (A), then p(A) = q(A)
i.e. the operator p(A) only depends on the restriction p[
(A)
Remark 2: Why is P() := p[
[ p : 1 C polynomial dense in C()?

Stone-Weierstrass:
P() is a subalgebra of C()
P() seperates points (i.e. x, y , x ,= y p P() s.t. p(x) ,= p(y))
p P() p P()
P() is dense in C()
76
Proof of Theorem 6:
1. Existence: Given f C(), construct f(A) L(H)
By Remark 2, sequence p
n
P() such that lim
n
|f p
n
|
L
()
= 0
p
n
is a Cauchy sequence in C()
Lemma 6 (iii)
p
n
(A) is a Cauchy sequence in L(H)
|p
n
(A) p
m
(A)| = |p
n
p
m
|
L
()
p
n
(A) converges in L(H)
Dene f(A) := lim
n
p
n
(A)
(This is the only way of dening f(A), so we have proved uniqueness)
2. f(A) is well-dened
If q
n
P() is another sequence converging uniformly to f, then
|p
n
(A) q
n
(A)| = |p
n
q
n
|
L
()
0
So lim
n
p
n
(A) = lim
n
q
n
(A)
3. f is linear, continuous and satises (1), (2), (3)
p
n
f, q
n
q
p
n
q
n
fg, p
n
+q
n
f +g
(fg)(A) = lim
n
p
n
q
n
(A) = lim
n
p
n
(A) lim
n
q
n
(A) = f(A)g(A)
Same for addition and for f(tA) = tf(A)
1(A) = 1
f f(A) bounded, indeed |f(A)| = |f|
L
()
= sup
(A)
[f()[
True for f = p P(); p
n
f
|f(A)| = lim
n
|p
n
(A)| = lim
n
|p
n
|
L
()
(2) and (3)
(2) obvious, (3) true for polynomials, take limits 2
Theorem 7: Let A = A
L(H), := (A) 1 and f C()

(i) (f(A)) = f((A))
(ii) |f(A)| = |f|
L
()
(iii) Ax = x f(A)x = f()x
(iv) AB = BA f(A)B = Bf(A)
(v) If f() 1 then f(A) = f(A)
(vi) f 0 x, f(A)x) 0 x H
Proof:
(ii) already proved, also follows from (i)
(iii) true for polynomials, hence true in the limit
(iv) true for polynomials, take the limit
(v) Use Theorem 3:
inf f() = inf (f(A)) = infx, f(A)x)
(vi) (f(A) f()
Let / f()
Dene g() :=
1
f()
,
g(f ) = (f )g = 1
g(A)(f(A) 1) = 1 = (f(A) 1)g(A)
f(A) bijective, i.e. / (f(A))
77
f() (f(A))
Let = (A). Claim: f() (f(A))
Suppose not
f()1 f(A) is bijective
Pick a sequence of polynomials p
n
P() such that |p
n
f|
L
()
0
f()1 f(A) = lim
n
(p
n
()1 p
n
(A)) (in norm topology)
p
n
()1 p
n
(A) is bijective for n large
p
n
() / (p
n
(A)), this contradicts Lemma 6 (ii)
2
5.4 Spectral Measure
Let A = A
L, = (A) 1. We have dened

C() = continuous functionsf : C
B() = bounded measurable functionsf : C
F() = bounded functionsf : C
|f| := sup
[f()[
Remark 1: C() B() F() closed subspace and F() is a Banachspace.
Remark 2: We take a sequence f
n
F() bp-converges to f F() i
sup
nN
|f
n
| < , f() = lim
n
f
n
()
We write
f = bp-lim
n
f
n
1. B() is closed under bp-convergence.
2. C() is bp-dense in B().
Therefore
Remark 3: B() is the smallest C
-subalgebra in F() so that

1. C() B()
2. B() is bp-closed.
Recall from Theorem 6 that ! continuous C
-homomorphism
A
: C()
L(H) so that
A
(id) = A
Theorem 8: ! C
-homomorphism
A
: B() L(H) such that
1. If f() = then
A
(f) = A.
2. |
A
(f)| |f|f B()
3. f = bp-lim
n
f
n

A
(f)x = lim
n

A
(f
n
)xx H
78
Remark: Let A
n
L(H) and A L(H). We say A
n
converges strongly to
A if
Ax = lim
n
A
n
x x H
A
n
converges in norm to A if
lim
n
|AA
n
| = 0
Fact Norm convergence Strong convergence.
Proof of Theorem 8, only sketch:
Existence
1. For every x H there is a unique real Borel measure
x
on such that
x,
A
(f)x)
C
=
_
f d
x
f C()
Namely
real Borel measures on = C(, 1)
= : C() C [ complex linear, bounded, (f) = (f)f

That was discussed in detail when we introduced the dual space.
Therefore for every bounded complex linear function : C() C with
(f) = (f) ! real Borel-measure on so that
(f) =
_
f d f C()
Example
x
(f) := x,
A
(f)x).
(a)
x
(f) = x,
A
(f)x)
= x, (f)
x)
=
A
(f)x, x)
= x,
A
(f)x)
=
x
(f)
(b)
[
x
(f)[ = [x,
A
(f)x)[
|x| |
A
(f)x|
|
A
(f)| |x|
2
by Theorem 7 = |f| |x|
2
This concludes the proof of the rst statement: Choose
x
so that
x
(f) =
_
f d
x
f C()
79
2. ! complex linear operator
A
: B() L(H) so that |
A
f| |f|f
B() and
x,
A
(f)x)
C
=
_
f d
x
f B
Namely: Let f : 1 be bounded and measurable. The map H H
1,
(y, x)
1
4
_
_
_
f d
x+y

_
f d
xy
_
_
is bilinear and symmetric and bounded (which we will not verify).
Dene
A
(f) by
y,
A
(f)x) :=
1
4
_
_
_
f d
x+y

_
f d
xy
_
_
x,
A
(f)x) =
_
f d
x
|f| |x|
2
|
A
(f)| = sup
|x|=1
[x,
A
(f)x) |f|
This is the end of the proof of step 2.
2
Proof: For f = f
1
+if
2
: C bounded measurable dene
A
(f) :=
A
(f
1
) +i
A
(f
2
)
Exercise:
A
: B() L(H) is a C
-algebra homomorphism
1. f
n
bp
f
A
(f
n
)
A
(f)
By Lebesgue dominated convergence: lim
n
x,
A
(f
n
)x) = lim
_
f
n
d
x
=
_
S
igmaf d
x
= x,
A
(f)x)x H
y,
A
(f)x) = lim
n
y,
A
(f
n
)x) x, y H
Also: |
A
(f)x|
2
= x,
A
(f)
A
(f)x) = x,
A
(ff)x) = lim
n
x,
A
(f
n
f
n
)x) =
lim
n
|
A
(f
n
)x|
2
Chap. III

A
(f)x = lim
n
A
(f
n
)x x H
Here we use: If
n
H and H, lim
n
x,
n
) = y, ) y H,
lim
n
|
n
| = || then lim
n
|
n
| = 0. Apply this to
n
:=
A
(f
n
)x
Uniqueness: C() is bp-dense in B() 2
Exercise:
(i) f 0
A
(f) 0, i.e. x,
A
(f)x) 0 and
A
(f) selfadjoint
(ii) Ax = x
A
(f)x = f()x
(iii) AB = BA
A
(f)B = B
A
(f)
Notation: f(A) :=
A
(f)
Theorem 6: continuous functional calculus
Theorem 8: measurable functional calculus
80
Remark:
1. Recall x,
A
(f)x) =
_
f d
x
2. f continuous |
A
(f)| = |f|
3. f bounded measurable =|
A
(f)| = |f|
Warning: B() ,= L
(,
x
)
Literature: Reed-Simon
Spectral projections
Let be a Borel set. Dene
() :=
_
1
0
Then
B() and
2
The operator P
:=
A
(
) is an
orthogonal projection: P
2
= P
= P
Corollary: The orthogonal projections P
L(H) satisfy the following con-

ditions:
(i) P
= 0 P
= 1
(ii) P
2
= P
1
P
2
(iii) =
i=1

i

l

k
= k ,= l
P
x = lim
n
n
k=1
P
k
x x H
Proof: Theorem 8 2
1 compact set
B() 2
Borel -algebra
The map B() L(H) satisfying the axioms of the corollary above, is called a
projection valued measured on . The projection valued measure of the corollary
is called the spectral measure of A
Remark: From the spectral measure we can recover the operator A via
x
() = x, P
x) and x, f(A)x) =
_
f d
x
Example: A compact and self-adjoint
=
0
,
1
, . . .,
n

0
= 0
E
n
:= ker(
n
1 A), P
n
(H) orthogonal projection onto E
n
f(A)x =
n=0
f(
n
)P
n
x
P
P
n
n
(not convergent in the norm)
Denition: x H is called cyclic for A, if spanx, Ax, . . . = H
Theorem 9: A = A L(H), x H cyclic
Hilbert space isometry (unitary operator) U : H L
2
(,
x
) such that
(UAU
1
f)() = f()
81
Proof:
x
Borel measure on dened by
_
f d
x
= x, f(A)x)
C
f C()
Claim: isometric isomorphism U : H L
2
(,
x
) such that
H
A
H
U
_U
L
2
L
2
(f)() = f()
Dene T : C() H by Tf :=
A
(f)x
|Tf|
2
= |
A
(f)x|
2
=
A
(f)x,
A
(f)x)
= x,
A
(f)
A
(f)x) = x,
A
(f)
A
(f)x) ()
= x,
A
(ff)x) =
_
ff d
x
=
_
[f[
2
d
x
= |f|
2
L
2
Recall: C() dense in L
2
(,
x
)
()
T extends uniquely to an isometric embedding T : L
2
(,
x
) H
Claim: T is surjective: f() =
n

A
(f) = A
n
Tf = A
n
x
Hence A
n
x imT n N
x cyclic
imT spanA
n
x is dense in H
Moreover |Tf|
H
= |f|
L
2 f, so T is injective and has a closed image
T is bijective U := T
1
To show: UAU
1
= or equivalently AT = T
ATf = A
A
(f)x =
A
(id)
A
(f)x =
A
(id f)x = T(id f) = Tf 2
Remark: In general, if A = A
L(H) and H is separable orthogonal

decomposition H =
k
H
k
such that AH
k
= H
k
and A[
H
k
admits a cyclic
vector.
Exercise: A compact and selfadjoint
cyclic x H every eigenspace of A is 1-dimensional
Similar to the following example:
Example 1: A = A
C
nn
A
= A
T
, x, y) =
n
j=1
x
j
y
j
ONB e
1
, . . . e
n
of eigenvectors of A; Ae
j
=
j
e
j

j
1
Assume
j
,=
k
for k ,= j, then (LA) x =
n
i=1
e
i
is cyclic.
A
k
x =
n
i=1
k
i
e
i
=
1
, . . .
n
f(A) =
n
i=1
f(
i
)e
i
, )e
i
C() = L
2
()

= C
n
x
is dened by
_
f d
x
= x, f(A)x)
n
i=1
f(
i
)e
i
, x)x, e
i
) =
n
i=1
f(
i
)

x
=
n
i=1
i
U : H = C
n
C
n
= L
2
U = (e
i
, ))
n
i=1
= diag(
1
, . . . ,
n
)
Example 2: H = l
2
(Z) = x = (x
n
)
nZ
[
n=
[x
n
[
2
<
(Lx)
n
= x
n+1
(L
x)
n
= x
n1
A = L +L
selfadjoint (A) = [2, 2] and H = H

ev
H
odd
where H
ev
:= x [ x
n
= x
n
and H
odd
:= x [ x
n
= x
n
are invariant
under A
82
(Ax)
n
= x
n1
+x
n+1
x
ev
:= (. . . 0, 1
..
=x
ev
0
, 0 . . .) and x
odd
:= (. . . 0, 0, 1, 0
..
=x
odd
0
, 1, 0, 0 . . .) cyclic vectors
for A[
H
ev, A[
H
odd
Dene U : H L
2
([0, 1]) by (Ux)(t) :=
n=
x
n
e
2int
(ULx)(t) = e
2it
(Ux)(t) (UL
x)(t) = e
2it
(Ux)(t)
(UAU
1
f)(t) = 2 cos(2t)f(t)
A multiplication operators on L
2
([2, 2],
1
) L
2
([2, 2],
2
)
83
6 Unbounded operators 25.01.2007
6 Unbounded operators
X, Y Banach spaces, D X dense, A : D Y closed graph
For x D, |x|
A
:= |x|
X
+|Ax|
Y
graph norm
Then (D, | |
A
) is a Banach space and A : D Y is bounded
In Spectral theory one studies 1 A : D X where X Banach space,
D X dense subspace and A : D X linear
Recapitulation
1. graph(A) := (x, Ax) [ x D X X
A closed
def
graph(A) is a closed subspace of X X
2. B : dom(B) X is an extension of A, if dom(A) dom(B) and
B[
dom
(A) = A
3. A is closable, if A admits a closed extension
4. A is closable graph(A) is a graph i.e. (0, y) graph(A) y = 0
Denote by A the smallest closed extension of A, graph(A) = graph(A)
5. 1
n
open, X = L
p
() 1 < p < D := C
0
(),
A : D X dierential operator
A is closable
Adjoint Operator
A : dom(A) X densly dened linear operator on a Banach space. The adjoint
operator A
: dom(A
) X
is dened as follows:
dom(A
) := y
[ c > 0 x dom(A) [y
, Ax)[ c|x|
For y
dom(A
) the linear functional dom(A) 1x y
, Ax) is bounded
x
such that x
, x) = y
, Ax) x dom(A)
Dene A
:= x
Note that A
, x) = y
, Ax) x dom(A), y
dom(A
)
Remark 1: Let y
, then
c 0 x dom(A) [y
, Ax)[ c|x|
x
sucht that x
, x) = y
, Ax)
In this case we have y
dom(A
), x dom(A)
Remark 2: (y
, x
) graph(A
) x
, x) = y
, Ax) x dom(A)
(x
, y
), (x, Ax)) = 0 x dom(A) (x
, y
) (graph(A))
Hence graph(A
) X
X

= (X X)
is always weak
-closed
Remark 3: A closable graph(A)
= graph(A)
= graph(A)
Rem. 2
A
= A
Lemma 1: X reexive, A : dom(A) X densely dened linear operator.

Then
(i) A
is closed
(ii) A closable dom(A
) is dense in X
(iii) A closable A
= A
and
X dom(A)
A
X
dom(A
)
A
84
Proof: Dene J : X X X X by J(x, y) = (y, x)
Then by Remark 2, graph(A
) = (J graph(A))
= J
graph(A)
(x
, y
) = (y
, x
)
(iii) (J graph(A
))
= graph(A
) and
(J
graph(A
))
= (graph(A)
=
XX
(
(graph(A)
)) =
XX
(graph(A))
(ii) Assume dom(A
) is dense
Let (0, y) graph(A) x
n
dom(A) such that x
n
0, Ax
n
y
y
dom(A
) we have y
, y) = lim
n
y
, Ax
n
) = lim
n
A
, x
n
) = 0
dom(A
) dense
y = 0
2
Proof : of Lemma 1 (continued) X Banach space, A : dom(A) X linear,
dom(A) X dense supspace, X reexive.
A closable dom(A
) is dense in X
.
1. graph(A)
= (x
, y
) X
[ x
, x)+y
, Ax) = 0x dom(A) =
(A
, y
) [ y
dom(A
)
2. graph(A) =
(graph(A)
)
3. (x, y) graph(A)
2.
x
, x) +y
, y) = 0(x
, y
) graph(A)
1.
A
, x) +y
, y) = 0y
dom(A
)
4. Because X is reexive we have: dom(A
) = X
dom(A
) = 0.
5. y
dom(A
)
y
, y) = 0y
dom(A
)
3.
(0, y) dom(A)
A closable
y = 0
2
Remark 1: The spectrum of an unbounded operator A : dom(A) X X
is dened exactly as in the bounded case:
(A) := C [ 1 A : dom(A) X is bijective
(A) = C (A)
P(A) = [ 1 A not injective
R(A) = [ 1 A injective, im(1 A) ,= X
C(A) = [ 1 a injective, im(1 A) = X, im(1 A) ,= X
Remark 2: X reexive, A : dom(A) X closed, densely dened.
imA
= ker A

(imA
) = ker A
imA =
(ker A
) imA
= (ker A)
Remark 3:
(A) = (A
) C(A) = C(A
)
R(A) P(A
) R(A
) P(A)
P(A) R(A) = P(A
) R(A
)
85
Example 1: X =
2
C
x = (x
1
, x
2
, x
3
, . . .) D := x
2
[
n=1
n
2
[x
n
[
2
<
Ax := (2x
2
, 3x
3
, . . .) A : D
2
closed.
P(A) = C x
= (,

2
2!
,

3
3!
, . . .) D Ax
= x
.
Example 2: Ax := (x
1
, 2x
2
, 3x
3
, . . .). Eigenvectors: e
n
:= (0, . . . , 0,
n
..
1 , 0, . . .)
Ae
n
= ne
n
P(A) = (A) = N
Lemma 2: X complex Banach space. D X dense subset. A : D X
closed operator. Assume
0
(A) (then R
0
(A) := (
0
1A)
1
: X D X
is bounded, cf. Closed Graph Theorem). Then the following holds:
i) If C
0
then
ker(1 A) = ker(
1
1 R
0
(A))
im(1 A) = im(
1
1 R
0
(A))
ii) (A) = C
0
[
1
(R
0
(A)). Same for P, R, C.
iii) (A) is open, the map (A) L(X, D) : R
(A) is holomorphic, and

R
(A) R
(A) = ( )R
(A)R
(A), (A). Here D is equipped

with the graph norm of A: |x|
D
:= |x|
X
+|Ax|
X
for x D.
Proof: ,=
0
.
1 A =
0
1 A+ (
0
)1
= (
0
1 A)(1 + (
0
)R
0
(A))
()
= (
0
)(
0
1 A)(
1

0
1 +R
0
(A))
= (
0
)(
0
1 A)(
1
0

1 R
1 R
0
(A))
1 A (i), (ii)
Proof of (iii) [
0
[ |R
0
(A)| < 1 1 (
0
)R
0
(A) bijective
()
1 A = (1 (
0
)R
0
(A))(
0
1 A) : D X is bijective
Hence (A) is open.
R
(A)
()
=
1
0

R
0
(
1
0

1 R
0
(A))
1
So (A) L(X, D) : R
(A) is holomorphic. 2
Denition: A closed, densely dened unbounded operator A : dom(A)
X X is said to have a compact resolvent if (A) ,= and R
(A) : X X is
compact (A).
Remark 4:
0
(A) R
0
(A) compact.
R
(A) compact (A), because

R
(A) = R
0
..
compact
(1 + (
0
)R
(A))
. .
bounded
Remark 5: Suppose (A) ,= and let D := dom(A) be equipped with a
graph norm. Then:
A has a compact resolvent the inclusion D X is compact
86
Remark 6: A : D = dom(A) X closed, densely dened.
X
0
:= X |x|
0
= |x|
X
X
1
:= D = dom(A) |x|
1
:= |x|
X
+|Ax|
X
X
2
:= x D [ Ax X
1
|x|
2
:= |x| +|Ax| +|A
2
x|
X
3
:= x D [ Ax X
2
and so on with
. . . X
3
X
2
X
1
X
0
.
Assume (A) ,= , let
0
(A) and denote T :=
0
1 A. Then T : X
k+1
X
k
is an isomorphism for all K and
Moreover: (1 A)
m
: X
m
X
0
and ker(1 A)
m
X
m=1
X
m
. And:
if the inclusion X
1
X
0
is compact, then X
k+1
X
k
is compact k.
Lemma 3: A : dom(A) X closed, densely dened, compact resolvent
(i) (A) = P(A)
(ii) The space E
(A) :=
m=1
ker(1 A)
m
is nite dimensional (A).
(iii) (A) is discrete.
Proof: Let
0
(A) and denote K := R
0
(A) L(X). Let (A) ,=
0
and :=
1
(K)
Moreover: ,= 0 and E
(A) = E
(K) nite dimensional. (i), (ii) see Ch IV.

Let
n
(A)
n
,=
m
n ,= m

n
:=
1
(K)
Ch IV

n
0 [
n
[ . 2
X = H Hilbert space and dom(A) H dense subset
A : dom(A) H closed linear operator.
Denition: The Hilbert space adjoint of A is the (closed, densely dened)
operator A
: dom(A
) H given by dom(A
) := y H [ c 0x
dom(A), [y, Ax)[ c|x| with A
y := z, where z H is the unique vector

with z, x) = y, Ax).
Remark 7: A closed A
= A
Denition:
a) A is called self-adjoint if A
= A, ie. dom(A
) = dom(A) and A
x
=
Axx dom(A)
b) A is called symmetric if x, Ay) = Ax, y)x, y dom(A).
Remark 8: A symmetric dom(A) dom(A
) and A
[
dom(A)
= A.
Lemma 4: A : dom(A) H densely dened, self-adjoint
i) (A) 1
ii) If in addition, A has a compact resolvent, then (A) = P(A) is discrete
subset of 1 and H has a ONB of eigenvectors of A.
Proof: Easy exercise. 2
87
Exercise 1: A L(X), U C open, (A) U
f : U C holomorphic, a path in U around (A)
f(A) :=
1
2i
_
f()(1 A)
1
d
Prove:
(i) fg(A) = f(A)g(A), 1(A) = 1 id(A) = A
(ii) (A) =
0

1

0
,
1
compact and disjoint
f() =
_
0 U
0
1 U
1
where U
0
, U
1
are disjoint open sets, U
i
containing
i
P := f(A) satises P
2
= P PA = AP
X
0
:= ker P X
1
:= imP, so X = X
0
X
1
and (A[
X
i
) =
i
What is W
1,p
([0, 1])?
W
1,p
([0, 1]) := f : [0, 1] 1[f cont. g L
p
([0, 1]) : f(x) = f(0)+
_
x
0
g(t) dt x
|f|
W
1,p :=
_
_
1
0
[f(x)[
p
dx +
_
1
0
[g(x)[
p
dx
_1
p
Fact: f W
1,p
f is dierentiable almost everywhere and

f(x) = g(x) for
almost all x [0, 1]
Warning: f almost everywhere dierentiable
f L
p
=f W
1,p
(Cantor-function)
Remark: g L
1
,
_
x
0
g(t) dt = 0 x [0, 1] g 0 a.e.
(measure and integration)
Denition: A function f : [0, 1] 1 is said to have bounded variation if
Var
[0,1]
f < where
Var
[0,x]
f := sup
0=t
0
<t
1
<...t
n
=x
n1
i=0
[f(t
i+1
) f(t
i
)[
sin(
1
x
) is not of bounded variation
f
(u) = lim
0
m1
i=0
u(t
i
)(f(t
i+1
) f(t
i
)) =
_
1
0
udf =
_
ud
f
where 0 = t
0
< t
1
< . . . < t
m
= 1 and := max
i
[t
i+1
t
i
[
Exercise 2: BV := f : [0, 1] 1 [ f is of bounded variation and right continuous
|f|
BV
:= [f(0)[ + Var
[0,1]
f
Prove BV is a Banach space
Exercise 3: Every (right continuous) function of bounded variation is the
dierence of two monotone (right continuous) functions.
Hint: Denote F(x) := [f(0)[ + Var
[0,x]
f
Show that F f are monotone, right continuous.
f
:=
Ff
2
f = f
+
f
Exercise 4:
a) f : [0, 1] 1 monotone, right continuous, f(0) 0
! Borel measure
f
on [0, 1] such that
f
([0, x]) = f(x)
b) f BV ! Borel measure
f
such that
f
([0, x]) = f(x) x [0, 1]
c) f(x) =
_
x
0
g(t) dt x [0, 1], g L
1

f
(E) =
_
E
g d Lebesgue measure
hint for a): construct outer measure
f
,
f
((a, b)) = lim
tb
f(t) f(a)
f
(open sets)
88
6 Unbounded operators
Exercise 5
: f BV, F(x) := [f(0)[ + Var

[0,x]
f
(i) [
f
[ =
F
(ii) f(x) =
_
x
0
g(t) dt F(x) =
_
x
0
[g(t)[ dt
Hint: (i) (ii) P := g > 0 N := g < 0
[
f
[(E) =
f
(E P)
f
(E N) =
_
EP
g
_
EN
g =
_
E
[g[ ()
F(x) =
F
([0, x])
(i)
= [
f
[([0, x])
()
=
_
x
0
[g[
Proof of (i): Easy
F
([0, x]) [
F
[([0, x])
Hard
Exercise 6: f BV . Equivalent are:
(i)
f

(ii) g L
1
([0, 1]) such that f(x) =
_
x
0
g(t) dt
(iii) f(0) = 0 and f is absolutely continuous i.e.
> 0 > 0 n N0 s
1
< t
1
< s
2
< t
2
< . . . < s
n
< t
n
1
i=1
[t
i
s
i
[ <
n
i=1
[f(t
i
) f(s
i
)[ <
Hint: First assume f is monotone
Remark:
a) f abs. continuous f is di. a.e., f L
1
and f(x) = f(0) +
_
x
0
f(t) dt x
b) W
1,p
([0, 1]) = f : [0, 1] 1 [ f abs. continuous, f L
p
|f|
W
1,p =
_
|f|
p
p
+|

f|
p
p
_1
p
Lemma: p > 1 The inclusion W
1,p
([0, 1]) C([0, 1]) is a compact operator
Proof: [f(t) f(s)[ = [
_
t
s
f(v) dv[
_
t
s
[

f(v)[ dv
_
_
t
s
[

f(v)[
p
dv
_1
p
[t s[
1
q
1
p
+
1
q
= 1
[f(t) f(s)[ |

f|
L
p[t s[
1
q
The set f W
1,p
[ |f|
W
1,p 1 is bounded and equicontinuous, so the
result follows from Arzela-Ascoli 2
Example: H = L
2
([0, 1], C) f, g) =
_
1
0
fg dt
D := f W
1,2
([0, 1]) [ f(0) 1, f(1) 1
Af := i

f
Exercise 7: A is selfadjoint, (A) = 2Z
89
Index
Symbols
C
-algebra. . . . . . . . . . . . . . . . . . . . . . . . . 75
f-invariant Borel probability measure
41
1st category . . . . . . . . . . . . . . . . . . . . . . . 13
2nd category . . . . . . . . . . . . . . . . . . . . . . 13
A
absolutely continuous . . . . . . . . . . . . . . 89
annihilator of Y . . . . . . . . . . . . . . . . . . . 32
Arzela-Ascoli, Theorem of . . . . . . . . . . 9
B
Baire Category Theorem. . . . . . . 12, 14
Banach Alaoglu, general form. . . . . . 42
Banach Alaoglu, sequentially . . . . . . 40
Banach space . . . . . . . . . . . . . . . . . . . . . . . 2
Banach space isometry. . . . . . . . . . . . . 15
Bounded linear operator . . . . . . . . . . . . 5
bounded variation . . . . . . . . . . . . . . . . . 88
bp-converges. . . . . . . . . . . . . . . . . . . . . . . 78
C
chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
closable . . . . . . . . . . . . . . . . . . . . . . . . 27, 84
closed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
compact operator . . . . . . . . . . . . . . . . . . 56
compact resolvent . . . . . . . . . . . . . . . . . 86
completion . . . . . . . . . . . . . . . . . . . . . . . . 17
convergence
in norm. . . . . . . . . . . . . . . . . . . . . . . 79
strong. . . . . . . . . . . . . . . . . . . . . . . . . 79
convex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
cyclic for A. . . . . . . . . . . . . . . . . . . . . . . . 81
D
dense . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Diagonal Subsequence . . . . . . . . . . . . . 41
direct sum. . . . . . . . . . . . . . . . . . . . . . . . . 25
dual operator . . . . . . . . . . . . . . . . . . . . . . 53
dual space . . . . . . . . . . . . . . . . . . . . . . . . . 19
E
eigenvalue . . . . . . . . . . . . . . . . . . . . . . . . . 65
equicontinuous . . . . . . . . . . . . . . . . . . . . . . 9
equivalent . . . . . . . . . . . . . . . . . . . . . . . . . . 2
ergodic . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
extension of A. . . . . . . . . . . . . . . . . . . . . 84
extremal point . . . . . . . . . . . . . . . . . . . . . 50
F
face . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
FiP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Fredholm alternative. . . . . . . . . . . . . . . 64
Fredholm index. . . . . . . . . . . . . . . . . . . . 60
Fredholm operator . . . . . . . . . . . . . . . . . 60
G
graph norm. . . . . . . . . . . . . . . . . . . . . . . . 26
H
Hahn-Banach. . . . . . . . . . . . . . . . . . . . . . 29
Hellinger-Toeplitz-Theorem. . . . . . . . 27
Hermitian inner product . . . . . . . . . . . 69
Hilbert space . . . . . . . . . . . . . . . . . . . . . . 15
Hilbert space adjoint . . . . . . . . . . . . . . 87
holomorphic . . . . . . . . . . . . . . . . . . . . . . . 67
homeomorphism. . . . . . . . . . . . . . . . . . . 41
I
Integral of x . . . . . . . . . . . . . . . . . . . . . . . 66
interior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
isometry. . . . . . . . . . . . . . . . . . . . . . . . . . . 17
K
Krein-Milman . . . . . . . . . . . . . . . . . . . . . 51
L
locally convex . . . . . . . . . . . . . . . . . . . . . 38
M
maximal element . . . . . . . . . . . . . . . . . . 30
maximal FiP . . . . . . . . . . . . . . . . . . . . . . 43
metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
N
Neumann, Theorem of von. . . . . . . . . 52
norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
normed vector space . . . . . . . . . . . . . . . . 2
nowhere dense . . . . . . . . . . . . . . . . . . . . . 13
O
orthonormal basis . . . . . . . . . . . . . . . . . 74
P
partial order . . . . . . . . . . . . . . . . . . . . . . . 30
POS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
projection valued measured on . . 81
R
reexive . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
residual . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
90
Index
S
self-adjoint . . . . . . . . . . . . . . . . . . . . . . . . 87
selfadjoint . . . . . . . . . . . . . . . . . . . . . . . . . 70
seminorm. . . . . . . . . . . . . . . . . . . . . . . . . . 31
separable. . . . . . . . . . . . . . . . . . . . . . . . 8, 36
Sobolev space. . . . . . . . . . . . . . . . . . . . . . . 3
spectral measure of A. . . . . . . . . . . . . . 81
Spectral radius . . . . . . . . . . . . . . . . . . . . 20
spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . 65
strong convergence. . . . . . . . . . . . . . . . . 79
supremum. . . . . . . . . . . . . . . . . . . . . . . . . 30
symmetric . . . . . . . . . . . . . . . . . . . . . . . . . 87
T
topological vector space. . . . . . . . . . . . 38
totally bounded. . . . . . . . . . . . . . . . . . . . . 6
transverse subspaces . . . . . . . . . . . . . . . 25
U
Uniform Boundedness . . . . . . . . . . . . . 22
unitary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
91

Functional Analysis

Uploaded by

Copyright:

Available Formats

Functional Analysis

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Functional Analysis

Uploaded by

Copyright:

Available Formats

Functional Analysis Lecture script

where means equal almost everywhere.

which are not convergent in | |

(x) = y M [ d(x, y) < . Since U

=int(A) = x M [ > 0 such that B

f(t +h) f(t)

Idea: Choose f(t) = p(t) +z

f(t +h) f(t)

z(t +h) z(t)

p(t +h) p(t)

. The dicult part of the proof is that (1) is

= All real Borelmeasures on M =: /

is linear. But why is this map surjective?

continuous and nonnegative.

) is a complete metric space

is an isometric embedding i.e.

((x), (y)) = d(x, y) x, y M

M. For every x M dene the function f

, ) be a completion of (X, d). Then there is a unique pair of

:= L(X, 1) is a Banach space with the norm |A| := sup

is called the dual space of X.

is a bounded F-linear functional : X F. We write:

Denition: Y X linear subspace of a normed vector space X.

, x) = |x| (Cor 3). We have proved:

(dened in Lemma 4) is bijective.

are not reexive

1 be a bounded linear functional. Need to show: x

Consider the diagram

is reexive, but X is not reexive. Then (X) _ X

, because (X) is complete by

be the bounded linear map

. Consider the diagram

1. Then, because X is reexive, y X such

and T is an isometric isomorphism (Exercise 1,b)).

. Consider the composition

Denote z := (x) = [x] Z

= / = nite Borel measures on M

K. Let U X be an open set with x

be the weakest topology on X

such that each linear functional

, x) is continuous (in this case T = i(X) X

) is a locally convex topological vector space

-topology if and only if

Does this imply that x

Exercise: Find an example

, x) < c weakly open and x

= real Borel measures : B 1

, elements of X are bounded sequences x = (x

-convergent subsequence, ie. for

such that the sequence

| 1 in the dual space is weak

with the weak

-closed. Prove it directly without using Thm 2.

is compact. Product topology on K L =

weak* closure(K). Need to prove sequence x

E. Let be such that 0 < < inf

c > 0 such that |x

is closed in the weak

-closed and bounded

-closed for every S