A Proof Of Corona Theorem

A Project Report
by
Samir Ch. Mandal
Project advisor :
Prof. Tirthankar Bhattacharyya
Department of Mathematics
Indian Institute of Science
Bangalore - 560012
May, 2013
Declaration
I hereby declare that the work in this project has been carried out by me in the Integrated Ph.D.
Program under the supervision of Prof. Tirthankar Bhattacharyya and in the partial fulllment
of the requirements of the Master of Science Degree at the Indian Institute of Science, Bangalore.
Samir Ch. Mandal
S R No. 10-06-01-10-31-11-1-08641
Indian Institute of Science,
Bangalore,
May, 2013.
Prof. Tirthankar Bhattacharyya
(Advisor)
Contents
1 Preliminaries 1
1.1 Hardy Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Banach Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Gelfand Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Tools:Solution of

Problem 11
2.1 Partitions of Unity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Solution of

Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Greens Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Corona Problem 21
3.1 Formulation of Corona Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
A Appendix 29
A.1 Double Sequence and Double Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
A.2 Dual Extremum Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Abstract
In this article, we prove that the open unit disk of C is dense in the maximal ideal space of H

(D).
We can see that this proof does not need Carlesons Measure. The main ingredients in this proof are
Greens formula, Existence of solution of a

problem.
Chapter 1
Preliminaries
In this chapter we will discuss some basic materials of Hardy spaces and Banach algebras.
1.1 Hardy Space
Denition 1. Consider the set of functions f that have a square summable power series with positive
radius of convergence around 0 as Hardy space. In notation,
H
2
= f : f(z) =

n=0
a
n
z
n
and

n=0
[a
n
[
2
< .
The inner product on H
2
is dened by
f, g =

n=0
a
n
b
n
where f(z) =

n=0
a
n
z
n
and g(z) =

n=0
b
n
z
n
.
So |f|
2
=

n=0
[a
n
[
2
.
The mapping a
n

n=0
a
n
z
n
from
2
into H
2
is clearly an isomorphism. Hence H
2
is a Hilbert
space.
Theorem 1. Every function in H
2
is analytic on the open unit disk.
Proof. Let f(z) =

n=0
a
n
z
n
H
2
. Let z
0
D. It is enough to show that

n=0
a
n
z
n
0
is absolutely
1
2 CHAPTER 1. PRELIMINARIES
convergent. As

n=0
[a
n
[
2
< , so a
n
is bounded . Suppose [a
n
[ K for all n N 0. Then

n=0
[a
n
z
n
0
[

n=0
[a
n
[[z
0
[
n
K

n=0
[z
0
[
n
.
As z
0
D, [z
0
[ < 1, so the right hand side is convergent. Hence left hand side is also convergent.
Hence f is analytic on the disk.
Theorem 2. For Every z
0
D, the map f f(z
0
) is a bounded linear functional on H
2
.
Proof. Fix z
0
D. The map is clearly linear, so only boundedness needs to be proved.
[f(z
0
)[ =

n=0
a
n
z
n
0

n=0
[a
n
z
n
0
[

n=0
[a
n
[
2
_
1/2
_

n=0
[z
0
[
2n
_
1/2
[by Cauchy-Schwarz inequality]
=
_

n=0
[z
0
[
2n
_
1/2
|f|.
As [z
0
[ < 1 so

n=0
[z
0
[
2n
< . Hence the map is bounded.
The Riesz representation theorem says that every bounded linear functional on a Hilbert space can
be represented by an inner product with a vector in the space. So for the evaluation map we should
have g H
2
such that f(z
0
) = f, g for all f H

(D). Write f(z) =

n=0
a
n
z
n
, g(z) =

n=0
b
n
z
n
.
Hence f(z
0
) =

n=0
a
n
b
n
. This is true for any f H
2
. So choose f to be the function f(z) = z
n
.
Then f(z
0
) = z
n
0
= b
n
b
n
= z
0
n
. Hence
g(z) =

n=0
z
0
n
z
n
=
1
1 z
0
z
= k
z
0
(z), say.
We say k
z
0
is the Reproducing kernel for z
0
in H
2
. Now a simple calculation shows that |k
z
0
|
2
=

n=0
[z
0
[
2n
=
1
1|z
0
|
2
and f(z
0
) = f, k
z
0
.
Theorem 3. If f
n
f in H
2
, then f
n
f uniformly on compact subsets of D.
Proof. For a xed z
0
D, we have
[f
n
(z
0
) f(z
0
)[ = [f
n
f, k
z
0
[ |f
n
f| |k
z
0
|.
1.1. HARDY SPACE 3
Now suppose K is a compact subset in D, then there exists M > 0 such that |k
z
0
| =
1

1|z
0
|
2
M
for all z
0
K [since
1

1|z
0
|
2
is a positive continuous function on K, hence bounded]. Hence
[f
n
(z) f(z)[ M|f
n
f| for all z K. This implies the theorem.
Theorem 4. Let f be analytic on D. Then f H
2
if and only if
sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d < .
Moreover, for f H
2
,
|f|
2
= sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d.
Proof. Write f(z) =

n=0
a
n
z
n
=

n=0
a
n
r
n
e
in
where z = re
i
. Then,
[f(re
i
)[
2
= f(re
i
)f(re
i
) =
_

n=0
a
n
r
n
e
in
__

n=0
a
n
r
n
e
in
_
=

n=0

m=0
a
n
a
m
r
n+m
e
i(nm)
The last sum is absolutely convergent because

n=0
a
n
r
n
is absolutely convergent [by Theorem 19 in
Appendix]. Now we know that
1
2
2
_
0
e
i(nm)
d =
n,m
where
n,m
=
_
_
_
0, if n ,= m
1, otherwise
So
1
2
_
2
0
[f(re
i
)[
2
d =
1
2
_
2
0

n=0

m=0
a
n
a
m
r
n+m
e
i(nm)
d
=

n=0

m=0
1
2
_
2
0
a
n
a
m
r
n+m
e
i(nm)
d [ since the sum is absolutely convergent]
=

n=0
[a
n
[
2
r
2n
.
Let f H
2
, then

n=0
[a
n
[
2
r
2n

n=0
[a
n
[
2
< for every r [0, 1).
So sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d |f|
2
< .
Conversely, assume that the above supremum is nite. We have now from previous calculation,
1
2
_
2
0
[f(re
i
)[
2
d =

n=0
[a
n
[
2
r
2n

n=0
[a
n
[
2
r
2n
for any k N
4 CHAPTER 1. PRELIMINARIES
So sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d sup
0<r<1
k

n=0
[a
n
[
2
r
2n
=
k

n=0
sup
0<r<1
[a
n
[
2
r
2n
=
k

n=0
[a
n
[
2
k N.
So sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d

n=0
[a
n
[
2
. Hence

n=0
[a
n
[
2
< .
In this case we also have ,
sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d =

n=0
[a
n
[
2
= |f|
2
.
The process in the proof also shows that M(r) =
1
2
_
2
0
[f(re
i
)[
2
d is increasing function of r so
we have
sup
0<r<1
1
2
_
2
0
[f(re
i
)[
2
d = lim
r1
1
2
_
2
0
[f(re
i
)[
2
d.
The Hardy space can also be viewed as a subspace of another well-known Hilbert space L
2
=
L
2
(S
1
), the Hilbert space of square integrable functions on S
1
with respect to Lebesgue measure,
normalized so that the measure of entire circle is 1. Call this measure m that is, dm =
d
2
. The
inner product on L
2
is given by
f, g =
1
2
_
2
0
f(e
i
)g(e
i
) d =
_
2
0
f(e
i
)g(e
i
) dm().
Hence |f|
2
=
_
2
0
[f(e
i
)[
2
dm().
Dene for n Z, e
n
(e
i
) = e
in
. Then e
n
: n Z forms an orthonormal basis for L
2
.
So f L
2
can be written as
f(e
i
) =

n=
a
n
e
in
,

n=
[a
n
[
2
< where a
n
=
_
2
0
f(e
i
)e
in
dm().
Dene

H
2
=

f L
2
:

f, e
n
= 0 for all n < 0.
Then the Fourier series of

f

H
2
is of the form

f(e
i
) =

n=0
a
n
e
in
, with

n=0
[a
n
[
2
< .
1.1. HARDY SPACE 5
Dene a map : H
2

H
2
by
f =

f, where f(z) =

n=0
a
n
z
n
and

f(e
i
) =

n=0
a
n
e
in
.
This map is clearly an isometric isomorphism. This way we can embed H
2
into L
2
. We will investigate
the relationship between f and

f. For f H
2
dene f
r
on S
1
by
f
r
(e
i
) = f(re
i
) for 0 < r < 1.
Now f
r
(e
i
) = f(re
i
) =

n=0
a
n
r
n
e
in
. Now

n=0
[a
n
r
n
[
2

n=0
[a
n
[
2
< . Hence f
r


H
2
.
Theorem 5. Let

f and f
r
be dened as above. Then
lim
r1

f f
r
| = 0 in

H
2
.
Proof. Let > 0 be given. Since

n=0
[a
n
[
2
< , there exists n
0
N such that

n=n
0
[a
n
[
2
< /2.
|f f
r
|
2
=

n=0
[a
n
(1 r
n
)[
2
=
n
0
1

n=0
[a
n
(1 r
n
)[
2
+

n=n
0
[a
n
(1 r
n
)[
2

n
0
1

n=0
[a
n
(1 r
n
)[
2
+

n=n
0
[a
n
[
2
.
Now, we can choose s < 1 such that
n
0
1

n=0
[a
n
(1 r
n
)[
2
< /2 for all 1 > r > s.
Hence |f f
r
|
2
< /2 +/2 = for all 1 > r > s.
This implies that
lim
r1

f f
r
| = 0 in

H
2
.
Corollary. For each f H
2
there exists an increasing sequence r
n
of positive numbers with
r
n
< 1 converging to 1 such that
lim
n
f
rn
(e
i
) =

f(e
i
) for almost every .
Now we shall give some more denitions, which will be useful later.
6 CHAPTER 1. PRELIMINARIES
Denition 2.
H

= f : f is analytic and bounded in D,

H
p
=
_
f : f is analytic on D such that sup
0<r<1
_
2
0
[f(re
i
)[
p
d <
_
, for p 1,
and H
p
0
= f H
p
: f(0) = 0, for p 1.
1.2 Banach Algebra
Denition 3. A Banach algebra is a complex Banach space A together with an associative and
distributive multiplication such that
(ab) = (a)b = a(b)
and |ab| |a| |b|
for all a, b A and C.
For any x, x

, y, y

A, we have
|xy x

| = |x(y y

) + (x x

)y| |x||y y

| +|x x

||y

|
and so we see that multiplication is jointly continuous.
The algebra A is said to be commutative (or abelian) if ab = ba for all a, b A, and A is said to
be unital if it possesses a (multiplicative) unit (this is also called an identity). Note that if A has
an identity, then it is unique: since if 1 and 1

are units, then 1 = 11

= 1

. We will use only unital

commutative Banach algebra here.
Examples of Banach algebras:
Let H be a Hilbert space, then B(H), the space of all bounded linear operators on H is a
non-commutative Banach algebra.
A(D), the space of all functions which are analytic in D and continuous on the unit circle is a
commutative Banach algebra. This algebra is called disk alegebra.
H

(D), the space of all bounded analytic functions in D is a commutative Banach algebra.
Now we will prove that H

(D) is a Banach algebra.

1.2. BANACH ALGEBRA 7
Proof. In H

multiplication is dened point wise, that is, for f, g H

we dene fg(z) = f(z)g(z)

and norm by |f| = sup[f(z)[ : z D. Then clearly we have |fg| |f|g| and multiplication
is associative. We need only to prove the completeness of H

. To do this, let f
n
be a Cauchy
sequence in H

. So f
n
(z) is Cauchy sequence for every z D. So there exists a function f such
that f
n
is convergent to f point wise. Choose > 0, then there exists N N such that
[f
n
(z) f
m
(z)[ < /2 for all n, m N and for all z D.
Keeping n xed and let m then we have
[f
n
(z) f(z)[ /2 for all n N and for all z D.
Thus sup[f
n
(z) f(z)[ : z D /2 for all n N.
which shows that f
n
converges to f in sup norm. Now we show that f is bounded. From the last
equation, there exists N
0
N such that
[f
n
(z) f(z)[ < 1 for all n N
0
and for all z D.
That is, [f(z)[ [f
N
0
(z)[ + 1 for all z D. As f
N
0
is bounded, it follows that f is bounded.
Now to show f is analytic we use Moreras theorem. As f
n
is uniformly convergent in D, so is
in any compact subsets of D. Take be any closed triangle in D, then is compact. We have
_

f
n
dz
_

f dz. And f
n
s are analytic so
_

f
n
dz = 0. Hence
_

f dz = 0. This shows that f is

analytic in D.
Denition 4 (Invertible Elements). a A is called invertible element if there exists b A such that
ab = ba = 1.
It is easy to see that the set of all invertible elements in A is open set in A.
Denition 5 (Spectrum). C is said to belongs to the spectrum of a A, denoted by (a), if
a 1 is not invertible. (a) = A(a) is called Resolvent set of a.
Theorem 6. For a A, (a) is non-empty subset of C.
Proof. Dene a map F : (a) A by F() = (a )
1
, (a). For [[ > |a| we have
(a )
1
= [(
a

1)]
1
=
1

(
a

1)
1
. So |F()| =
1
||
|
a

1|
1
||
1
1
a

=
1
||a
. Now if we
assume that (a) is empty set then (a) = C. So F is a bounded entire function which is constant
by Liouvilles theorem. But this is not possible. So (a) is non-empty.
8 CHAPTER 1. PRELIMINARIES
Theorem 7 (Gelfand-Mazur Theorem). If A is a Banach algebra in which every non-zero element
is invertible then A = C.
Proof. Let a A. Since (a) ,= , there exists (a). Since a 1 is not invertible so a 1 = 0
that is a = 1.
Denition 6 (Multiplicative linear functional). A complex linear functional on A is called a
multiplicative linear functional if (ab) = (a)(b) and (1) = 1 for all a, b A. We will denote
the set of all multiplicative linear functionals on A by /
A
.
Theorem 8. If is a multiplicative linear functional on a unital Banach algebra A then || = 1.
Proof. Suppose [(a
0
)[ > |a
0
| for some a
0
A. Let a =
a
0
(a
0
)
, then |a| =
a
0

|(a
0
)|
< 1 and (a) = 1.
Now consider the element b = a +a
2
+a
3
+ . Then b = a(1 +b). Now we get (b) = (a)(1 +b)
so that (b) = 1 + (b) which is a contradiction. Hence || 1 . Also we have (1) = 1 and
|1| = 1. So || = 1.
We recall the denition of the w

-topology on A

, the dual of the Banach space A.

Denition 7 (w

-Topology). The w

-topology on A

is the one generated by the neighborhoods

N( : o, ) = A

: [(a) (a)[ < for all a o

where A

, is any positive real number and o is any nite subset of A. Thus a set G in A

is open in the w

-topology if and only if for each G there is some N( : o, ) as above with

N( : o, ) G
In terms of nets, the w

-topology is described as the weakest topology on A

such that a net (

)
in the w

- topology if and only if

(a) (a) for each a A.

The topology of /
A
is the restriction of w

-topology on A

. By Banach-Alaoglu theorem we
have the unit ball in A

to be compact with this topology. Now it is easy to see that /

A
is a closed
subset of the unit ball of A

when the Banach algebra A is unital. Hence /

A
is compact if A is a
unital Banach algebra. This topology is also called Gelfand Topology.
1.3. GELFAND THEORY 9
1.3 Gelfand Theory
Denition 8 (Maximal Ideal). An ideal in an algebra is said to be maximal if it is proper (that is,
not equal to the whole algebra) and is not contained as a proper subset in any other proper ideal.
Theorem 9. Every maximal ideal is closed.
Proof. Let I be a maximal ideal in A and let

I I. For a I, we have |a 1| 1 (otherwise a
would be invertible and then 1 = a
1
.a would be in I which will imply that every element of A is in
I violating the fact that I is proper). As norm is continuous, we have |a 1| 1 for all a

I. Now
as I is a maximal ideal, so

I is whole A because

I is also an ideal containing I. This contradicts the
fact that |a 1| 1 for all a

I because 1

I(= A).
Theorem 10. There is a canonical bijection between the maximal ideals in a commutative unital
Banach algebra A and its multiplicative linear functionals given by associating to each multiplicative
linear functional its kernel; that is, if is a multiplicative linear functional on A, ker is a maximal
ideal in A, and every maximal ideal has this form for some unique multiplicative linear functional.
Proof. Suppose that : A C is a multiplicative linear functional and let I = ker . We have
I ,= A since is non-zero. Let a / I. Then any b A can be written as
b = a
(b)
(a)
+
_
b a
(b)
(a)
_
.
Since b a
(b)
(a)
ker = I, we see that A = Ca +I and hence I is a maximal ideal.
Now suppose that I is a maximal ideal. Then I is closed and so A/I is a Banach algebra. We
claim that the maximality of I implies that every non-zero element of A/I is invertible. To see this,
suppose that a +I is a non-zero non-invertible element of A/I. Then I +aA is a proper ideal of A
which contains I as a proper subset. (I +aA does not contain 1 since a+I is non-invertible in A/I.)
This contradicts the supposed maximality of I, and the claim is established.
It then follows from Gelfand-Mazur theorem that every element in A/I has the form (1 + I),
for C. Let : A C denotes this isomorphism, and let denotes the canonical projection
10 CHAPTER 1. PRELIMINARIES
: A A/I. Then is a homomorphism from A to C with kernel equal to I; for
(ab) = ((ab)) = (ab +I)
= ((a +I)(b +I)) = (a +I)(b +I)
= ((a))((b)) = (a) (b)
and (a) = 0 if and only if (a) = 0 if and only if a I.
Thus we have a correspondence between maximal ideals I and multiplicative linear functionals
with ker = I. This association is one-one, since is uniquely determined by its kernel. Indeed,
suppose that and

have the same kernel. Then for any a A, the element a (a)1 belongs to
ker = ker

and so

(a) = (a) since

(1) = 1.
Chapter 2
Tools:Solution of

Problem
2.1 Partitions of Unity
Let be an open set in R
n
and | = U
i

iI
be an open cover of .
Denition 9. A partition of unity relative to | is a family
i
of C

functions on having the

following properties:
(i)
i
is real-valued,
i
0 on and supp(
i
) U
i
,
(ii) the family of closed sets supp(
i
)
iI
is locally nite (that is, for any compact K , the
set i I : K supp(
i
)
iI
,= is nite ),
(iii)

iI

i
= 1 on .
Lemma 1. Let U be open set in R
n
and let K U be a compact set. Then there exists C

0
(U)
such that (x) > 0 for all x K.
Proof. The function dened on R by
(t) = e

1
1t
for t < 1 , (t) = 0 for t 1
is a C

function on R. The function on R

n
dened by ((x
1
, , x
n
)) = (2(x
2
1
+ +x
2
1
)) is a
C

(R
n
) function with (0) > 0 and supp () = x := (x
1
, , x
n
) R
n
:
n

i=1
x
2
i

1
2
. Let > 0 be
such that <dist(K, U
c
). We dene for any a K, a function
a
C

0
(R
n
) by
a
(x) =
_
(xa)

_
.
11
12 CHAPTER 2. TOOLS:SOLUTION OF

PROBLEM
Then
a
(a) > 0, supp(
a
) U. Let V
a
= x U :
a
(x) > 0. Then V
a
is open and

aK
V
a
K.
Hence there exist a
1
, . . . , a
p
K such that
p

i=1
V
a
i
K. The function =
a
1
+ +
ap
has the
required property.
Theorem 11. Let be an open set in R
n
and | = U
i

iI
be an open cover of . Then there exists
a partition of unity relative to |.
Proof. For proving this theorem we will use some lemmas:
Lemma 2. An arbitrary open set U in R
n
can be written as a union

m=1
K
m
of compact sets with
the properties that K
m
K

m+1
for all m 1.
Proof. K
m
= D
2
m(0)

xR
n
\U
D 1
2
m
(x). Then K
m
K

m+1
and U =

m=1
K
m
.
Lemma 3. For an arbitrary open subset of R
n
and open cover | = V
i
, we can nd a sequence
of x
j
and
j
> 0 such that
(i) =

j=1
D

j
(x
j
),
(ii) Given j, there exists i(j) such that D
2
j
(x
j
) V
i(j)
,
(iii) For every x U, there exists an open neighborhood W such that j : W D

j
(x
j
) ,= is
nite.
Proof. Let =

m=1
K
m
as in Lemma 2. Dene open sets U
m
= K

m+1
K
m2
for m > 2. Also dene
compact sets B
m
= K
m
K

m1
for m 2. For x , choose the smallest m such that x K
m
.
Then x B
m
. So =

m=2
B
m
. Clearly B
m
U
m
. For x B
m
, we can nd an open neighborhood
say, D
2(x)
(x) which is in U
m
and also is included in V
i
for some i. As B
m
is compact, by Heine-Borel
theorem there exists x
m
j
B
m
and
m
j
> 0, j = 1, 2, , d
m
such that
(a) B
m

dm

j=1
D
m
j
(x
m
j
),
(b) Every D
2m
j
(x
m
j
) is contained in U
m
and in V
i
for at least one i. Also =

m=1
B
m
=

m=1
_

j=1
d
m
D
m
j
(x
m
j
)
_

m=1
U
m
= . This proves (i) and (ii).
Let m
0
1 with x U
m
0
. Since U
m
0

U
m
= for m > m
0
+ 3 U
m
0
can intersect D
2(xm
j
)
(x
m
j
)
only if m m
0
+ 2. This proves (iii).
2.2. SOLUTION OF

PROBLEM 13
Now we will come back to the proof of the Partition of Unity:
By the previous lemma there exists a locally nite open cover V
i

J
of such that V
j
[we can choose
V
j
= D
2
j
(x
j
)] for all j J which is a renement of |, that is, for all j J there exists i I such
that V
j
U
i
. We choose a map : J I such that V
j
U
(j)
. Now there exists compact subsets
K
j
V
j
such that

j
K
j
= . Let
j
C

0
(V
j
) with
j
(x) > 0 for all x K
j
. Set =

j
. Since
V
j
is locally nite, the sum

j
is a nite sum on every compact subset of , so C

().
Clearly, since
j
> 0 on K
j
and

K
j
= , we have > 0 on . Let
j
=
j
/. Then
j
is a
partition of unity relative to V
j
. For i I, let J
i
=
1
(i) and dene
i
=

jJ
i

j
. Then
i
is a
partition of unity relative to |.
Theorem 12. Let be an open set in R
n
and let A be a closed set. Let U be an open set in
containing A. Then there exists C

(), 0 1 such that [

A
1, [
\U
0.
Proof. Consider the cover U, A of . Now by existence of partition of unity there exist
1
,
2

C

() such that
(a)
1
,
2
both are real-valued, non-negative and supp(
1
) U, supp(
2
) A.
(b)
1
+
2
= 1 on .
Hence
1
works as as in the theorem.
2.2 Solution of

Problem
Lemma 4. Let be a open set in C and f C
1
(). Let R be a closed rectangle, R . Then we
have
__
R
f
z
dxdy =
1
2i
_
R
f dz.
Proof. Let R = [a, c] [b, d] and let (a, b), (c, b), (c, d), (a, d) be the vertices of the rectangle R. We
know that

z
=
1
2

x
+i

y
.Now
d
_
y=b
c
_
x=a
f
x
dxdy =
d
_
b
f(c, y) f(a, y) dy
14 CHAPTER 2. TOOLS:SOLUTION OF

PROBLEM
and
c
_
x=a
d
_
y=b
f
y
dydx =
c
_
a
f(x, d) f(x, b) dx.
(a, b)
1 (c, b)
(a, d)

4

2
(c, d)

3
R
Figure 2.1: Rectangle R
So we have now
2i
__
R
f
z
dxdy = i
__
R
_
f
x
+i
f
y
_
dxdy
=
d
_
b
if(c, y) f(a, y) dy
c
_
a
f(x, d) f(x, b) dx
=
c
_
a
(x, b) dx +
d
_
b
if(c, y) dy +
a
_
c
f(x, d) dx +
d
_
b
if(a, y) dy
=
_

1
f(z) dz +
_

2
f(z) dz +
_

3
f(z) dz +
_

4
f(z) dz
=
_
R
f(z) dz.
Lemma 5. Let R, R

be closed rectangles in C with R

. Let U be an open set containing RR

and let C

(U). Then
2i
__
R\R

z
dxdy =
_
R
dz
_
R

dz.
Proof. Let K = RR

. Let C

0
(U) be such that [
K
= 1. We dene on a neighborhood V
of R by (z) = (z)(z) if z U, (z) = 0 on V U. Then we have
2i
__
R

z
dxdy =
_
R
dz =
_
R
dz.
2.2. SOLUTION OF

PROBLEM 15
Similarly,
2i
__
R

z
dxdy =
_
R

dz =
_
R

dz.
Subtracting we get the result.
Lemma 6.
__
|z|<1
1
[z[
dxdy < .
Proof. Changing to polar co-ordinate z = re
i
we have
__
|z|<1
1
[z[
dxdy =
__
0 < r < 1
0 < < 2
1
r
r drd = 2.
Theorem 13. Let C

0
(C). Then, for any w C, we have
__
C

z
1
z w
dxdy = (w).
Proof. If we replace z by z + w (w xed), the formula for change of variable in a double integral
gives
__
C

z
1
z w
dxdy =
__
C

z
(z +w)
1
z
dxdy.
Let R be rectangle containing the origin, and such that (z + w) = 0 for all z / R

. Let R

be the
rectangle R

= [, ] [, ], with > 0 so small that R

. Then by Lemma 5
2i
__
R\R
(z +w)
z
1
z
dxdy = 2i
__
R\R

z
_
(z +w)
z
_
dxdy
=
_
R
(z +w)
z
dz [ since (z +w) = 0 for z R].
Now
_
R
(z +w)
z
dz =
_
R
(z +w) (w)
z
dz + 2i (w).
Further since C

0
(C) by hypothesis, there exists M > 0 such that [(z + w) (w)[ M[z[.
16 CHAPTER 2. TOOLS:SOLUTION OF

PROBLEM
Hence

_
R
(z +w)
z
dz

ML(R

) 0 as 0. Letting 0 we get
2i
__
R
(z +w)
z
1
z
dxdy = (w) 2i.
Since (z +w) = 0 for z / R, we have
2i
__
C
(z +w)
z
1
z
dxdy = 2i
__
C

z
1
z w
dxdy = (w) 2i
which completes the proof.
Theorem 14. Let C

0
(C). We dene a function u on C by
u() =
1

__
C
(z)
z
dxdy where z = x +iy.
Then u C

(C) solves the following PDE:

u
z
= .
Proof. We have u() =
1

__
C
(z +)
z
dxdy. As
1
z
is integrable so u is well-dened and continuous.
Now we will show that u is dierentiable. So x and let h R, h ,= 0, [h[ < 1. Then
u( +h) u()
h
=
1

__
C
(z + +h) (z +)
zh
dxdy.
Now as is dierentiable,

(z++h)(z+)
h

x
(z + +
z
.h)

M for some M > 0 and

z
satisfying 0 <
z
< 1, since C

(C),

x
is bounded, say by M. As C

(C) there exists

a compact set K such that (z + + h) = 0, (z + ) = 0 for all z / K, 0 < [h[ < 1. So

(z++h)(z+)
zh

M
|z|
.1
K
which is integrable over C. So now we can apply Lebesgue Dominated
2.3. GREENS FORMULA 17
Convergence theorem to conclude the following:
u
x
() =
1

lim
h0
__
C
(z + +h) (z +)
zh
dxdy =
1

__
C
lim
h0
(z + +h) (z +)
zh
dxdy
=
1

__
C
1
z

x
(z +) dxdy
=
1

__
C

x
(z)
1
z
dxdy.
Similarly we have,
u
y
() =
1

__
C

y
(z)
1
z
dxdy.
Now C

0
(C) gives
u
y
(),
u
y
() C

0
(C), so we can iterate this procedure and say that u
C

(C).
Finally the above formulas for
u
x
() and
u
y
() give
u
z
() =
1

__
C

z
(z)
1
z
dxdy = () [by Theorem 13].
2.3 Greens Formula
Theorem 15.
u(0) =
1
2
_
D
ud
1
2
__
D
(u) log
1
[z[
dxdy (2.1)
where u is a smooth function in a neighborhood of D .
Proof. Write z = x +iy, x = r cos , y = r sin . Also dene v(r, ) = u(re
i
).
Then we have the following relations:
v
r
=
u
x
cos +
u
y
sin
v

=
u
x
r sin +
u
y
r cos .
Writing these two equations in matrix form we have
_
_
cos sin
r sin r cos
_
_
_
_
u
x
u
y
_
_
=
_
_
v
r
v

_
_
18 CHAPTER 2. TOOLS:SOLUTION OF

PROBLEM

_
_
u
x
u
y
_
_
=
1
r
_
_
cos sin
r sin cos
_
_
_
_
v
r
v

_
_
.
Which gives
u
x
= cos
v
r

1
r
sin
v

(2.2)
u
y
= sin
v
r
+
1
r
cos
v

(2.3)
Replacing u by
u
x
in equation 2.2 and 2.3 we get

2
u
x
2
= cos

r
_
cos
v
r

1
r
sin
v

1
r
sin

_
cos
v
r

1
r
sin
v

_
= cos
2

2
v
r
2

1
r
cos sin

2
v
r

1
r
sin
_
sin
v
r
+ cos

2
v
r
_
+
1
r
2
sin
_
cos
v

+ sin

2
v

2
_
.

2
u
y
2
= sin

r
_
sin
v
r
+
1
r
cos
v

_
+
1
r
cos

_
sin
v
r
+
1
r
cos
v

_
= sin
2

2
v
r
2
+
1
r
cos sin

2
v
r
+
1
r
cos
_
cos
v
r
+ sin

2
v
r
_
+
1
r
2
cos
_
sin
v

+ cos

2
v

2
_
.
Adding them we get
u =

2
v
r
2
+
1
r
v
r
+
1
r
2

2
v

2
=
1
r

r
_
r
v
r
_
+
1
r
2

2
v

2
.
So now we have
__
D
u. log
1
[z[
dxdy = lim
0
__
<|z|<1
u. log
1
[z[
dxdy = lim
0
_
1

_
2
0
rulog
1
r
ddr.
Now
_
2
0

2
v

2
=
v

(r, 2)
v

(r, 0) = 0 [by denition of v].

So
__
D
u. log
1
[z[
dxdy =
_
1

_
2
0

r
_
r

r
_
log
1
r
d =
_
2
0
_
1

r
_
r

r
_
log
1
r
d.
Now by integrating by parts the following integral is
_
1

r
_
r

r
_
log
1
r
d = r log r
v
r

+
_
1

r
v
r
d
dr
(log r) dr = log
v
r
(, ) +
_
1

v
r
dr
= log
v
r
(, ) +v(1, ) v(, ).
2.3. GREENS FORMULA 19
Now as u is smooth so
v
r
is bounded, so we have
__
D
u. log
1
[z[
dxdy =
_
2
0
v(1, ) v(, ) d =
_
2
0
u(e
i
) d
_
2
0
u(0) d
=
_
2
0
u(e
i
) d 2.u(0).
This completes the proof.
20 CHAPTER 2. TOOLS:SOLUTION OF

PROBLEM
Chapter 3
Corona Problem
3.1 Formulation of Corona Problem
We have seen that H

is a Banach algebra . Here we want to study the maximal ideal space of this
Banach algebra. Consider the map z m
z
where m
z
is a map on H

dened by m
z
(f) = f(z)
for all f H

. Clearly m
z
is a linear map for every z D. Observe that m
z
(fg) = (fg)(z) =
f(z)g(z) = m
z
(f)m
z
(g) . This shows that m
z
is multiplicative linear functional. Call this map E,
that is, E(z) = m
z
. We want to see that this map is one-one and continuous. Let z
1
, z
2
D such that
z
1
,= z
2
. Then there exists a map f H

(namely the identity map on D) such that f(z

1
) ,= f(z
2
)
that is, m
z
1
(f) ,= m
z
1
(f). This shows that E is one-one. Now take a net z

converging to some
z D. Then f(z

) f(z) for all f H

. That is m
z
(f) m
z
(f) for all f H

. This means
that m
z
converges to m
z
in the w

-topology. So the map E is actually continuous.

In this way we embed D into /
H
. Now naturally one may ask that how big the image of
D inside /
H
is. We can easily show that E(D) ,= /
H
. To see this, let us take a net z

converging to 1. As /
H
is compact, there exists a subnet z

of z

such that m
z

for
some /
H
. Now if = m
z
for some z D, then for the identity function on D we have z

z
which is a contradiction as z

1. So it may be possible that taking the closure of D inside /

H

with Gelfand topology we have

/
H
= D
Gelfand
We then dene the Corona of the algebra H

(D) to be /
H
D
Gelfand
. We now translate the
question of density to a question about analytic functions in the following theorem.
21
22 CHAPTER 3. CORONA PROBLEM
Theorem 16. D is dense in /
H

(D)
if and only if the following condition holds: for n 1
and f
1
, . . . , f
n
H

(D) satisfying max

1jn
[f
j
(z)[ for some > 0 at every z D, there are
g
1
, . . . , g
n
H

(D) such that

f
1
g
1
+ +f
n
g
n
= 1.
Proof. First assume that D is dense in /
H

(D)
. Consider the ideal I = f
1
, . . . , f
n
where
max
1jn
[f
j
(z)[ > 0 at every z D. Assume there does not exist g
1
, . . . , g
n
H

(D) such
that
f
1
g
1
+ +f
n
g
n
= 1.
This shows that I is a proper ideal. Then by Theorem 10, I is contained in some maximal ideal,
say M. Then M =ker for some /
H

(D)
. Now f
1
, . . . , f
n
I M shows that (f
j
) = 0
for all 1 j n. Since D is dense in /
H

(D)
, there exists a net z

D such that m
z
in
the w

-topology. Thus f
j
(z) (f
j
) = 0 for all 1 j n. This contradicts the hypothesis that
max
1jn
[f
j
(z)[ > 0.
Conversely, suppose whenever f
1
, . . . , f
n
H

(D) are such that max

1jn
[f
j
(z)[ for some > 0
at every z D, there are g
1
, . . . , g
n
H

(D) such that

f
1
g
1
+ +f
n
g
n
= 1.
We want to prove that D is dense in /
H

(D)
. Suppose not. Then there exists
0
/
H

(D)
and
an open set V /
H

(D)
containing
0
disjoint from D. We can assume that V is a basic open set.
Then there exist h
1
, . . . , h
n
H

(D) and > 0 such that V = /

H

(D)
: [(h
i
)
0
(h
i
)[ <
for all 1 i n. Dene f
i
= h
i

0
(h
i
) Then (f
i
) = (h
i
)
0
(h
i
). Then we can write
V =
n
i=1
/
H

(D)
: [(f
i
)[ < and
0
(f
i
) = 0 for all 1 i n. Now D is disjoint
from V implies that for any z D there exists j 1, 2, . . . , n such that [f
j
(z)[ so that
max
1jn
[f
j
(z)[ > 0 for all z D. Now by our hypothesis, there exists g
1
, . . . , g
n
H

(D) such
that f
1
g
1
+. . . , f
n
g
n
= 1. So
0
(f
1
)
0
(g
1
) + +
0
(f
n
)
0
(g
n
) = 1. This is a contradiction because
the left hand side is zero.
So now we will assume the given condition as in the previous theorem for some functions f
1
, . . . , f
n
and try to nd out the g
i
, 1 i n as in the previous theorem to prove the result. Our idea is to
nd at rst some trivial solution with some relaxed conditions such as the functions which we nd,
may not be analytic and then modify them suitably to be analytic and bounded. Hence the next
theorem goes like this.
3.1. FORMULATION OF CORONA PROBLEM 23
Theorem 17. Suppose that f
1
, . . . , f
n
H

(D) and there exists > 0 such that

1 max
1jn
[f
j
(z)[ > for all z D. Then there are g
1
, . . . g
n
H

(D) such that

1 = f
1
g
1
+. . . , f
n
g
n
and |g
j
|
H

(D)
C(, n) for all j = 1, 2, . . . , n.
Scheme of Proof: Without loss of generality we can assume that the functions are analytic
in a neighborhood of the closed disc D (we can replace f
i
(z) by f
i
(rz) for r < 1 close to 1, for all
1 i n and z D). Dene the functions

j
(z) =
f
j
(z)
n

j=1
[f(z)[
2
for all z D.
The denominator is non-zero due to the given hypothesis. Now we see that
n

j=1

j
f
j
= 1 and these
functions are bounded. But the problem is that these functions (
i
) are not in general analytic, so we
must modify them to be analytic . If we set g
j
(z) =
j
(z)+
n

k=1
a
jk
(z)f
k
(z) then
n

j=1
f
j
g
j
=
n

j=1

j
f
j
+
n

j=1
n

k=1
a
jk
f
j
f
k
. Hence we need
n

j=1
n

k=1
a
jk
(z)f
j
(z)f
k
(z) = 0 for all z D. That is,
n

j=1
a
jj
f
j
(z)f
j
(z) +
n

j=1
n

k=1
j=k
(a
jk
+ a
kj
)(z)f
j
(z)f
k
(z) = 0. So if we can set a
jk
(z) = a
kj
(z) then we have
n

j=1
f
j
g
j
= 1.
For this, we set a
jk
(z) = b
jk
(z) +b
kj
(z) for some yet to determine functions. If we choose b
jk
to be
the solutions to the following

problem:

b
jk
=
j

k
:= G
jk
then we see that

g
j
=

j
+
n

k=1
f
k

a
jk
=

j
+
n

k=1
f
k
(

b
jk


b
kj
)
=

j
+
n

k=1
f
k
(
j

k

k

j
)
=

j
+
j

(
n

k=1
f
k

k
)

j
(
n

k=1
f
k

k
)
=

j
= 0 [
n

k=1
f
k

k
= 1].
24 CHAPTER 3. CORONA PROBLEM
Also if we can solve b
jk
to be bounded then our g
j
s will be bounded. So, we will solve the problem.
Hence our whole problem is reduced to the following problem:

b
jk
=
j

k
= G
jk
(1 j, k n)
where b
jk
has to be bounded.
Solving a

problem with Bounds:
Let u be a smooth function dened on a neighborhood G of D. We wish to nd w such that
w
z
= u
in D, while |w|

is as small as possible. By the solution of

problem [Theorem 14] we always have
a smooth solution w
0
satisfying
w
0
z
= u in D; where we can take w
0
to be given by
w
0
() =
1

__
|z|
u(z)
z
dxdy such that z : [z[ G, > 0 and D. (3.1)
The general solution then has the form w = w
0
+p, where p is in disk algebra. Hence we want that
the following is as small as possible
inf |w
0
+p|

: p A(D) .
By duality [Lemma 12 in Appendix], this latter inmum coincides with
sup
_

_
w
0
F
d
2

: F H
1
0
, |F|
1
1
_
[see Appendix].
We may restrict our attention here to F analytic across D. Since F(0) = 0, Greens formula 2.1
applied to w
0
F yields
_
D
w
0
F d =
__
D
(w
0
F) log
1
[z[
dxdy.
Now
(w
0
F) = 4

2
z z
(w
0
F) = 4

z
(uF) = 4F

u + 4F
u
z
.
Thus the quantity (3.1) can be re-expressed in the form
2

sup
FH
1
0
,F
1
1

__
D
F

ulog
1
[z[
dxdy +
__
D
F
u
z
log
1
[z[
dxdy

. (3.2)
Write
__
D
F

ulog
1
|z|
dxdy = I and
__
D
F
u
z
log
1
|z|
dxdy = II, where F H
1
0
, |F|
1
1
To complete the proof , it suces now to bound I and II by constants C
0
(n, ), in the case that
u = G
jk
.
3.1. FORMULATION OF CORONA PROBLEM 25
Estimates for Integrals: Suppose that g is harmonic in a neighborhood G of D. We want to
apply Greens formula to u = [g[
2
.
Claim:u = 2[g

[
2
. To see this, write g(x, y) = p(x, y) + i q(x, y) then g
x
= p
x
+ i q
x
, g
xx
=
p
xx
+ i q
xx
and g
y
= p
y
+ i q
y
, g
yy
= p
yy
+ i q
yy
and u = p
2
+ q
2
. So u
x
= 2p p
x
+ 2q q
x
; u
y
=
2p p
y
+ 2q q
y
; u
xx
= 2p p
xx
+ 2(p
x
)
2
+ 2q q
xx
+ 2(q
x
)
2
, u
yy
= 2p p
yy
+ 2(p
y
)
2
+ 2q q
yy
+ 2(q
y
)
2
so
that
u
2
= pp + qq + (p
x
)
2
+ (p
y
)
2
+ (q
x
)
2
+ (q
y
)
2
. Now as g is given to be harmonic, we have
p = 0, q = 0. So we have
u
2
= (p
x
)
2
+ (p
y
)
2
+ (q
x
)
2
+ (q
y
)
2
= [g[
2
. So using this the Greens
formula for u = [g[
2
gives us
1

__
D
[g[
2
log
1
[z[
dxdy =
_
D
[g[
2
d
2
[g(0)[
2
.
This identity is known as Littlewood-Paley Identity
If furthermore g is analytic, then [[
2
= 2[g

[
2
. To see this, note that, for g analytic we have p
x
= q
y
and p
y
= q
x
. So [g[
2
= 2(p
2
x
+p
2
y
) = 2[g

[
2
. So the Identity becomes
2

__
D
[g

[
2
log
1
[z[
dxdy = |g|
2
2
[g(0)[
2
.
which gives natural estimate
__
D
[g

[
2
log
1
[z[
dxdy

2
|g|
2
, g H
2
. (3.3)
Lemma 7. If f
1
, f
2
H

and g
1
, g
2
H
2
, then
__
D
[g
1
g
2
f

1
f

2
[ log
1
[z[
dxdy 2|g
1
|
2
|g
2
|
2
|f
1
|

|f
2
|

.
Proof. By Cauchy-Schwarz inequality we have that
__
D
[g
1
g
2
f

1
f

2
[ log
1
[z[
dxdy
___
D
[g
1
f

1
[
2
log
1
[z[
dxdy
_
1/2
___
D
[g
2
f

2
[
2
log
1
[z[
dxdy
_
1/2
.
So it suces to show that if g H
2
and f H

, then
__
D
[gf

[
2
log
1
[z[
dxdy 2|g|
2
2
|f|
2

. (3.4)
26 CHAPTER 3. CORONA PROBLEM
Since (gf)

= gf

+g

f so gf

= (gf)

f, we have
[gf

[
2
2([(gf)

[
2
+[g

f[
2
) 2([(gf)

[
2
+[g

[
2
|f|
2

) [ [a b[
2
2([a[
2
+[b[
2
) ].
So now we have,
__
D
[gf

[
2
log
1
[z[
dxdy
__
D
2([(gf)

[
2
+[g

[
2
|f|
2

) log
1
[z[
dxdy
= 2
__
D
[(gf)

[
2
log
1
[z[
dxdy + 2|f|
2

__
D
[g

[
2
log
1
[z[
dxdy
|gf|
2
2
+|f|
2

|g|
2
2
[Using equation (3.3) ]
|g|
2
2
|f|
2

+|g|
2
2
|f|
2

[ |gf|
2
|g|
2
2
|f|
2

]
= 2|g|
2
2
|f|
2

.
Lemma 8. If F H
1
and f
1
, f
2
H

, then
__
D
[Ff

1
f

2
[ log
1
[z[
dxdy 2|F|
1
|f
1
|

|f
2
|

.
Proof. Write F = g
1
g
2
, where g
1
, g
2
H
2
satisfy |g
1
|
2
2
= |g
2
|
2
2
= |F|
1
. Now the lemma follows
from Lemma 7.
Lemma 9. If f H

and g
1
, g
2
H
2
, then
__
D
[g
1
g

2
f

[ log
1
[z[
dxdy |g
1
|
2
|g
2
|
2
|f|

.
Proof. By Cauchy-Schwarz inequality we get
__
D
[g
1
g

2
f

[ log
1
[z[
dxdy
___
D
[g

2
[
2
log
1
[z[
dxdy
_
1/2
___
D
[g
1
f

[
2
log
1
[z[
dxdy
_
1/2
.
Now if we apply equation (3.3) and Lemma 8 we get our result.
Lemma 10. If F H
1
and f H

, then
__
D
[F

[ log
1
[z[
dxdy 2|F|
1
|f|

.
Proof. Similarly write F = g
1
g
2
, where |g
1
|
2
2
= |g
2
|
2
2
= |F|
1
. Then F

= g
1
g

2
+ g

1
g
2
. Now apply
Lemma 9 to get the result.
3.1. FORMULATION OF CORONA PROBLEM 27
Now we conclude our proof of the main theorem:
So now we will bound I and II. Here u =
j

k
where
j
=
f
j
n

m=1
|fm|
2
. Write H =
n

m=1
[f
m
[
2
=
n

m=1
f
m
f
m
. So H =

m
f
m
,

H =
n

m=1
f
m
f

m
= H ;

H = (
n

m=1
f
m
f

m
) =
n

m=1
f

m
f

m
=
n

m=1
[f

m
[
2
.

j
=
f
j
H

j
=
f
j
H
H
2

k
=
f
k
H

k
=
f

k
H

f
k

H
H
2
.
So u =
j

k
=
f
j
H
_
f

k
H

f
k

H
H
2
_
=
f
j
f

k
H
2

f
j
f
k

H
H
3
=
f
j
f

k
H
2

f
j
f
k
H
3
n

m=1
f
m
f

m
; u =
j

k
+
j

k
.

k
=
_
f

k
H

f
k

H
H
2
_
=
f

k
H
2
H
1
H
2
f
k
(

H) + 2
f
k

H
H
3
H
=
f

k
H
2
H
f
k
H
2
n

m=1
[f

m
[
2
+ 2
f
k
[H[
2
H
3
.
Hence
u =
f
j
H
H
2
_
f

k
H

f
k

H
H
2
_
+
f
j
H
_

k
H
2
H
f
k
H
2
n

m=1
[f

m
[
2
+ 2
f
k
[H[
2
H
3
_
= 2
f
j
f

k
H
3
H
f
j
f
k
H
3
n

m=1
[f

m
[
2
+ 3
f
j
f
k
H
4
[H[
2
= 2
f
j
f

k
H
3
n

m=1
f

m
f
m

f
j
f
k
H
3
n

m=1
[f

m
[
2
+ 3
f
j
f
k
H
4
[
n

m=1
f

m
f
m
[
2
.
So ultimately we get
u =
f
j
f

k
H
2

f
j
f
k
H
3
n

m=1
f
m
f

m
u = 2
f
j
f

k
H
3
n

m=1
f

m
f
m

f
j
f
k
H
3
n

m=1
[f

m
[
2
+ 3
f
j
f
k
H
4
[
n

m=1
f

m
f
m
[
2
.
Now as 1 max
1jn
[f
j
(z)[ > , So H is bounded. Thus I is bounded by use of Lemma 10. The
expression for u depends quadratically on f

m
s and f

m
s. So II is bounded by use of Lemma 8. We
can easily see that the bounds on g
j
s does not depend on r. This completes the proof of Corona
Theorem.
28 CHAPTER 3. CORONA PROBLEM
Appendix A
Appendix
A.1 Double Sequence and Double Series
Denition 10 (Double Sequence). A function f : N N X, where X is a topological space, is
called a double sequence.
We will be interested only in X = C or R.
Denition 11 (Convergence of Double Sequence). If a C, we say that f converges to a and write
lim
p,q
f(p, q) = a if the following condition is satised: for every > 0, there exists N N such that
[f(p, q) a[ < whenever p > N, q > N.
Denition 12 (Double Series). Let f be double sequence and let s be the double sequence dened by
s(p, q) =
p

m=1
q

n=1
f(m, n).
The pair (f, s) is called a double series and is denoted by the symbol

m,n
f(m, n). The double series
is said to converge to a if lim
p,q
s(p, q) = a.
Theorem 18. Suppose f(m, n) 0 then the double series

m,n
f(m, n) is convergent if and only if
the double sequence s(p, q) is bounded.
Proof. First Suppose the double sequence s(p, q) is bounded. Then consider the set A = s(p, q) :
p, q N. A is bounded above, so sup A < . Let M = sup A. Now let > 0. Then there exist
29
30 APPENDIX A. APPENDIX
p
0
, q
0
N such that s(p
0
, q
0
) > M. Now take N = maxp
0
, q
0
. Then we have M < s(p
0
, q
0
)
s(p, q) M for all p > N, q > N, that is, [s(p, q) M[ < whenever p > N, q > N. In other words
lim
p,q
s(p, q) = M.
Conversely, let lim
p,q
s(p, q) = M. We want to prove that s(p, q) M. Suppose there exist
p
0
, q
0
N such that s(p
0
, q
0
) > M. Let = s(p
0
, q
0
) M. Now there exists N N such that
[s(p, q) M[ < . Now for p, q maxp
0
, q
0
,
s(p, q) s(p
0
, q
0
) > M so s(p, q)M < s(p
0
, q
0
)M for p, q maxp
0
, q
0
. Hence s(p, q) < s(p
0
, q
0
)
for p, q maxp
0
, q
0
which is contradiction as f(m, n) 0.
Theorem 19. Let

m
a
m
and

n
b
n
be two absolutely convergent series. Let f be the double sequence
dened by the equation
f(m, n) = a
m
b
n
, if m, n N.
Then

m,n
f(m, n) converges absolutely that is,

m,n
[f(m, n)[ < .
Proof. Let A =

m=1
[a
m
[, B =

n=1
[b
n
[ and let s(p, q) =
p

m=1
q

n=1
[f(m, n)[. By virtue of above
theorem it is enough to show that s(p, q) is bounded above. Now
s(p, q) =
p

m=1
q

n=1
[f(m, n)[
=
p

m=1
q

n=1
[a
m
[[b
n
[
=
_
p

m=1
[a
m
[
__
q

n=1
[b
n
[
_

m=1
[a
m
[
__

n=1
[b
n
[
_
= AB.
A.2 Dual Extremum Problem
Let X be a Banach space with dual space X

and let Y be a closed subspace of X. Then

Y

= x

: x

(y) = 0 y Y
is a closed subspace of X

.
A.2. DUAL EXTREMUM PROBLEM 31
Lemma 11. If x

, then
sup[x

(y)[ : y Y, |y| 1 = inf|x

k| : k Y

.
that is, |x

[
Y
| = dist(x

, Y

).
Proof. Let k Y

, |x

[
Y
| = |x

k[
Y
| |x

k| for all k Y

[ k[
Y
= 0 ]. So, |x

[
Y
|
inf|x

k| : k Y

.
Conversely, by Hahn-Banach theorem there exists f X

such that |f| = |x

[
Y
| and f[
Y
= x

[
Y
.
Then clearly f x

and |x

[
Y
| = |f| = |x

(x

f)|. Hence we proved that

sup[x

(y)[ : y Y, |y| 1 = inf|x

k| : k Y

.
Lemma 12. If x
0
X, then
inf|x
0
y| : y Y = sup[k(x
0
)[ : k Y

, |k| 1.
Proof. Let k Y

with |k| 1, y Y , then [k(x

0
)[ = [k(x
0
y)[ |k| |x
0
y| |x
0
y|. So
we have,
sup[k(x
0
)[ : k Y

, |k| 1 inf|x
0
y| : y Y .
To prove the reverse part, dene for every x X, p(x) = inf|x y| : y Y = |x Y |, then
clearly p is a semi norm. Now dene a linear functional f on x
0
by f(x
0
) = |x
0
Y |. Then
[f(x
0
)[ = [[|x
0
Y | [[ |x
0
| = |x
0
|. Hence |f| 1. Now by Hahn-Banach theorem there
exists an extension

f of f with [

f(x)[ p(x) for every x X and |

f| 1 . So [

f(y)[ p(y) = 0 for

every y Y , that is,

f Y

. Hence
sup[k(x
0
)[ : k Y

, |k| 1 [

f(x
0
)[ = |x
0
Y | = inf|x
0
y| : y Y .
Combining these two inequalities we get the result.
We used the 2nd lemma in the case where X = C(T), Y = X H

. In this case X

= M(T),
the space of complex Borel measures on the unit circle with total variation norm (this follows from
Theorem 6.19 in the book [RUD])and Y

= H
1
0
, identied with the closed subspace of M consisting
of those absolutely measures Fd/2 having density F in H
1
0
(this follows from Lemma 11.5.3 from
the book [KRA]).
32 APPENDIX A. APPENDIX
Bibliography
[GAM] T.W. GAMELIN, Wols Proof of The Corona Theorem, Israel Journal of Mathematics,
Vol. 37, Nos.1-2, 1980.
[GAR] JOHN GARNETT, Bounded analytic functions, Springer, Revised Edition, 2006.
[KRA] STEVEN G. KRANTZ, Geometric Function Theory : Explorations in Complex Analysis,
Springer, 2006.
[RUD] WALTER RUDIN, Real and Complex Analysis, McGraw-Hill, Third Edition, New York,
1987.
33