Functional Analysis

DEMTH642

Edited by:
Dr. Kulwinder Singh
Functional Analysis
Edited By
Dr. Kulwinder Singh
 Title: Functional Analysis

Author’s Name: Dr. Arshad Ahmad Khan

Published By : Lovely Professional University

Publisher Address: Lovely Professional University, Jalandhar Delhi GT road, Phagwara - 144411

Printer Detail: Lovely Professional University

Edition Detail: (I)

ISBN:

Copyrights@ Lovely Professional University

 Content

Unit 1: Normed Linear Space and Banach Spaces I 1

Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 2: Normed Linear Spaces and Banach spaces II 13
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 3: Bounded Linear Operator and its Properties 24
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 4: Hahn-Banach Theorem and its Consequences 40
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 5: Uniform Boundedness Principle and its Consequences 51
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 6: Inner Product Space. Hilbert Space 58
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 7: Orthogonality of Inner Product Space 69
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 8: Open Mapping Theorem and Closed Graph Theorem 80
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 9: Decomposition Theorems in Hilbert Spaces 92
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 10: Riesz Representation Theorem and Operators on Hilbert 107
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 11: Unitary and Normal Operators 126
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection 138
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 13: Spectral Theory of Linear Operators in Normal Spaces 150
Dr. Arshad Ahmad Khan, Lovely Professional University
Unit 14: Spectrum Of Normal Operators 160
Dr. Arshad Ahmad Khan, Lovely Professional University
 Notes

Unit 01: Normed Linear Space and Banach Spaces I

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 01: Normed Linear Space and Banach Spaces I

CONTENTS

Objectives
Introduction
1.1 Normed Linear Space
1.2 Properties of Normed Linear Space
1.3 Banach Space (Complete normed Space)
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Normed linear spaces.

 Banach spaces.
 Properties of normed space.

Introduction
The notion of a norm is an abstract generalization of the length of a vector. It is axiomatically stated
that the norm is any real valued function that satisfies specific requirements. The linear space
together with the norm is called a normed linear space. Moreover, the Banach spaces is a type of
normed linear spaces that possess the additional property of completeness.
In what follows, K will denote the field of R (real numbers) or C (complex numbers). We shall
always assume that R and C have their usual metrics and that all the linear spaces that we consider
will be defined over K (R or C).

1.1 Normed Linear Space

In this section, we first introduce the formal definition of norm, which serves as the building block
of the subsequent sections.

Definition. A norm || ⋅ || is defined as the function || ⋅ || ∶ → on a linear space satisfying the

following properties:

(i) || || ≥ 0, ∀ ∈ ;
(ii) || || = 0 if and only if = 0;
(iii) || + || ≤ || || + || ||, ∀ , ∈ ;
(iv) || | ≤ | | | ||, ∀ ∈ , ∈ .

Lovely Professional University 1

Notes

Functional Analysis

A linear space over with norm ||. || defined on is called a normed linear space or simply a
normed space over , written as ( , ||. ||) or . The normed linear space is real or complex
accordingly as the field is or .

Next, we provide some examples for the lucid illustration of the normed spaces.

Example. Consider a linear space together with the norm defined by || || = | |. Then, it is
easy to verify that the properties (i)-(iv) of the Definition holds. In particular, triangle inequality
follows by the fact || + || = | + | ≤ | | + | |, ∀ , ∈ .

Example. The space (n-dimensional Euclidean space) and (n-dimensional unitary

space) of all n-tuples of real and complex numbers are normal linear spaces with the norm defined
by

| | = | | = | | + | | +⋯| | " ; 1≤p&∞

Example. Let ℓ be the space of all sequences =) * satisfying ∑, | | & ∞, - ≥ 1.

.
Then, this space is a normed linear space with the norm | | = ∑, | | "/ , ∀ ∈ℓ .

Example. Consider ℓ, space, that is, the space of all bounded sequences =) *. Then, this
space is a normed linear space with the norm || ||, = sup | | , 1 ≤ 2 ≤ ∞.

Example. Find | | , | | and | | ,

for the vector = 2, 3, 1, 54" ∈ R8 .

. .
Solution: We have | | - = 9∑ | | :/ = | | + | | + ⋯ + | | "/ .

Then, | | = | | + | | + | ; | + | 8 | = |2| + |3| + |1| + |54| = 10

.
.
| | = <∑8 | | = = | | + | | + | ; | + | 8 | ">
>

.
= |2| + |3| + |1| + |54| "> = √30.

Also, | | ,
= @A-)| |, | |, | ; |, | 8 |* = @A-)|2|, |3|, |1|, | 5 4|* = 4.

Example. Let BC, DEdenotes the space of continuous real valued functions defined on BC, DE.
Then, BC, DE defines a normed linear space with the norms:

1. |F| = sup |F "| , ∀F ∈ BC, DE, ∈ BC, DE.

I
2. |F| = GJ |F "|H , ∀ F ∈ BC, DE.

2 Lovely Professional University

 Notes

Unit 01: Normed Linear Space and Banach Spaces I

1.2 Properties of Normed Linear Space

We now recall some basic definitions, which shall be frequently used in the remaining part of this
section.

Definition . Let , ||. ||" be the normed linear space on . Then, we have the following definitions.

 An open sphere (or open ball) with center K and radius L > 0 is the set N K ; L" =) ∈
: | 5 K| & L*. By the surface (or boundary) of this ball, we mean the set P K ; L" =) ∈
: | 5 K| = L*.
 The set NB K ; LE = ) ∈ : || 5 K || ≤ L* denoted by PB K , L E or PQ B K E is called the closed
sphere or closed ball with radius L and center R.

 A set S in is said to be open if for every ∈ S, there exists a ball with center which is
contained in D.
 A set S in is said to be closed if for any sequence ) * in S with → implies that
∈ S.
 A set S in is said to be bounded in if there exists a constant T such that | | & T, ∀ ∈
S.
 A set S in is said to be compact if whenever ) * ∈ S, there exist a convergent
subsequence of ) * whose limit is in S.
 A sequence ) * is said to be bounded, if there exists a real constant > 0 such that
| | & , ∀ U.
 A sequence ) *⊂ is said to be convergent if there exists an ∈ such that

lim →, ‖ 5 ‖=0

 A sequence ) *⊂ is said to be Cauchy sequence if for given [ > 0, ∃ a positive integer N

such that
]2^ ‖ _ 5 ‖ & ϵ ∀ m, n ≥ N
→,

That is, is said to be Cauchy sequence in X iff

]2^ ‖ _ 5 ‖ → 0 C@ ^, U → ∞
→,

 The space X is said to be complete if every Cauchy sequence in converges to an element

in .

Theorem . Show that Every normed linear space is a metric space w.r.to the metric , "=
|| 5 || ; ∀ , [ . But the converse may not be true.

Proof. Let be a normed linear space. Define a mapping H: × → by:

H , " = || 5 ||; ∀ , [ .

We show that H is a metric on .

Since (i) , " = || 5 || ≥ 0 .

That is, H , " ≥ 0.

Lovely Professional University 3

Notes

Functional Analysis

(ii) H , " = || 5 || = 0, iff 5 = 0 iff = .

That is, H , " = 0, iff = .

(iii) H , " = || 5 | = || 5 || = H , ".

That is, H , "=H , ".

(iv) H , d" = || 5 d| = || 5 + 5 d| ≤ || 5 || + || 5 d||,

=H , "+H , d".

Thus, H , d" ≤ H , " + H , d".

Hence, H is a metric on normed linear space , known as metric induced by norm and hence with
H is a metric space.

Now, we show that the converse of above theorem need not be true. For this consider a linear space
|efg|
with metric H defined as H , "=
h|efg|
.

We can clearly verify the above metric satisfies all the conditions of metric space.

|efg|
, "= | 5 | =
h|efg|
If we take .

|i|
Or we can write |d| = ∀d= 5 ∈
h|i|

|ki| |i|
Thus for any α scalar |αd| = = |α| ≠ |α| |d| .
h|ki| h|k||i|

This shows that is not normed linear space.

Theorem. Show that For any normed space ,

||| || 5 || ||| ≤ || 5 ||;∀ , [ .

Proof . We now prove that ||| || 5 || ||| ≤ || 5 ||;∀ , [ .

We can write = 5 + .

So, | | = | 5 + | ≤ | 5 | + | |.

Implies, || || 5 || || ≤ || 5 ||.

Similarly, we can write = 5 + .

So, || || = || 5 + || ≤ || 5 || + || ||.s

This implies, -|| 5 || ≤ || || 5 || ||.

Or 5|| 5 || ≤ || || 5 || ||.

So, from the above relations, we get

5|| 5 || ≤ || || 5 || || ≤ || 5 ||.

Implies, ||| || 5 || ||| ≤ || 5 ||.

Definition . Let and m are normed linear spaces, then F: → m is said to be continuous at K ∈
, If for given ϵ>0, thereexists δ > 0 such tha |F " 5 F K "| & ϵ whenever | 5 K| & δ.

4 Lovely Professional University

 Notes

Unit 01: Normed Linear Space and Banach Spaces I

That is, | 5 K| & δ, implies |F " 5 F K "| & ϵ.

Theorem. Show that a norm function is continuous.

Proof. Let & >→ x in normed linear space X.

That is, → as U → ∞.

Or, 5 x → 0 as U → ∞.

This gives || 5 x|| → 0 as U → ∞.

Now, | || || 5 || || | ≤ || 5 x|| → 0 as U → ∞.

This implies, | || || 5 || || | → 0 as U → ∞.

Or, || || 5 || || → 0 as U → ∞.

Or, || || → || || as U → ∞.

This Shows that ||. || is a continuous function.

Theorem. Show that every convergent sequence is a cauchys sequence.

Proof. Let & > be a sequence convergent to , then for given ϵ > 0, there exists UK ∈ N such that

p
|| 5 || & ∀ n ≥ nK .

p
In particular, || 5 || & for fixed m > nK .

Now, || 5 _ || = || 5 + 5 _ ||

= || 5 "+ 5 _ "||

≤ || 5 || + || 5 _ ||

= || 5 || + | 5 1||| _ 5 ||
p p
= | 5 | + | _ 5 | & + = ϵ ∀ ^, U ≥ UR

That is, || 5 _ || & ϵ ∀^, U ≥ UR .

Therefore, & xq > is a Cauchy sequence in Normed linear space .

Remark: Converse of above result need not be true in general. As every normed linear
space is a metric space, we will show this for metric space

For this , Consider X = (0, 1], with H , " = | 5 | and consider sequence & >=&
1/U > , where ∈ . then, & 1/U > is a Cauchy sequence as

H , _" = |1/U 5 1/^| → 0 as ^, U → ∞.

However, H 1/U, 0" = |1/U 5 0| → 0 as U → ∞.

But 0 ∉ . Therefore, & > is a Cauchy sequence in X but not convergent in X.

Lovely Professional University 5

Notes

Functional Analysis

Theorem . Let be a normed linear space over the field . Let & > and & > be
sequences in with ]2^ = and ]2^ = , respectively , and )α * be a sequence in
→, →,

with ]2^ = .
→,

Then

(i) ]2^ + "= + .

→,

(ii) ]2^ = α .
→,

Proof. We know by definition of norm

|| + "5 + "|| = || 5 "+ 5 "||

≤ || 5 "|| + || 5 "||

→ 0 as U → ∞.

Also, ||α 5 αx|| = ||αq 5 " + α 5 α"x||

≤| ||| f || + |α 5 α|||x||

→ 0 as U → ∞.

1.3 Banach Space (Complete normed Space)

Definition. A normed linear space is called complete if every Cauchy sequence in converges
to a limit point in .

A complete normed linear space is called a Banach space.

(OR )

A normed linear space which is complete as a metric space is called a Banach space.

Example. The Spaces and of reals and complex numbers are Banach spaces. These are the
consequences from the real analysis result that every Cauchy sequence is convergent.

Example. The spaces tu and are Banach spaces . Here we prove that is complete.

Let ) "
* be a Cauchy sequence in tu ,

" " " "

=< , ,…, = , - = 1,2, …

Then, given any ϵ > 0, there is a natural number UK such that

= y∑
" x"
∀ -, w; -, w ≥ nK , implies || "
5 x" || z 5 z & ϵ

" x"
Hence, ∀ -, w; -, w ≥ UK , implies z 5 z ≤ || "
5 x" || & [

6 Lovely Professional University

 Notes

Unit 01: Normed Linear Space and Banach Spaces I

" "
So, for each i, is a cauchy sequence of real numbers. Since R is complete, converge to a real
number for all 2 = 1, 2, 3, … , n. But this implies that for already choosen ϵ, there exist a natural
number - such that

"
∀ -; - ≥ -{ , implies z 5 z & [/√U

Take = , ,…, ", where = lim →,

"
. Then ∈ .

"
We show that = lim →, . Let -| = ^C - , - , . . . , -q "

= y∑
" x"
∀ -; - ≥ -K ≥ UK , implies || "
5 x" || z 5 z & [ By (1)

Hence ) "
* converges to ∈ , as required. Thus, is complete and hence is a Banach Space.

Example. The space ] , is a Banach space . This space consists of all bounded sequences
= ) * of real or cpmplex numbers with addition and scalar multiplication defined by:

+ =) + *,

C = )C *.

The norm in ] , is defined by:

| | = @A-, | |

We show that ] , is a Banach space.

Let ) "
* be any Cauchy sequence in ] , , "
= "
. Then,given any ϵ > 0, there is a natural
number UK such that:

" x"
∀ -, w; -, w ≥ nK ⟹ || "
5 x" || = @A-, z 5 z & ϵ.

So, for each 2 = 1, 2, . . .,

" x"
∀ -, w; -, w ≥ nK ⟹ z 5 z ≤ || "
5 x" || & ϵ.

"
~L " is complete ,
"
Hence, is a Cauchy sequence of real (or complex ) numbers, since
converges to for all 2 = 1, 2, 3, …

Take = . We show that ∈ ] , andlim →,

" = .

"
Since → , there is a natural number U such that

" p
∀-, ; -, ≥ n ⟹ z 5 z & , 1 = 1,2, . .. (1)

That is,

" •
∀-, ; -, ≥ n ⟹ || "
5 || = @A- , z 5 z ≤ & [.

Hence "
→ . Also, from (1)

| |=| 5
"
+
"
|

Lovely Professional University 7

Notes

Functional Analysis

" "
≤| 5 |+| |

•
& +€

•
Now + € is a finite number, independent of 2. This proves completeness of ] , .

Example. The space • ‚ƒ„‚ …† the Space of all convergent real (or complex ) sequences is a
Banach space . It is a subspace of ] , .

Example. The space •‡ that is the space of all sequences which converges to 0 is a Banach
space

Example. The space ˆ‰ , ‰ ≥ Š" : This is the space of all sequences = ) * such that | | =
.
9∑, | | :/ is a Banach space.

Example. The space ‹B„, ŒE. This is the space of all continuous functions from BC, DE to
~L " is a Banach space. The norm in BC, DE is ||F|| = @A- |F "|, F ∈ BC, DE.
e∈BJ,IE

Summary
 A linear space over with norm ||. || defined on is called a normed linear space or
simply a normed space over , written as ( , ||. ||) or . The normed linear space is real or
complex accordingly as the field is or .
 An open sphere (or open ball) with center K and radius L > 0 is the set N K ; L" =) ∈
: | 5 K| & L*. By the surface (or boundary) of this ball, we mean the set P K ; L" =) ∈
: | 5 K| = L*.
 The set NB K ; LE = ) ∈ : || 5 K || ≤ L* denoted by PB K , L E or PQ B K E is called the closed
sphere or closed ball with radius L and center R.

 A set S in is said to be open if for every ∈ S, there exists a ball with center which is
contained in D.
 A set S in is said to be closed if for any sequence ) * in S with → implies that
∈ S.
 A set S in is said to be bounded in if there exists a constant T such that | | & T, ∀ ∈
S.
 A set S in is said to be compact if whenever ) * ∈ S, there exist a convergent
subsequence of ) * whose limit is in S.
 A sequence ) * is said to be bounded, if there exists a real constant > 0 such that
| | & , ∀ U.
 A sequence ) *⊂ is said to be convergent if there exists an ∈ such that

lim →, ‖ 5 ‖=0

8 Lovely Professional University

 Notes

Unit 01: Normed Linear Space and Banach Spaces I

 A sequence ) *⊂ is said to be Cauchy sequence if for given [ > 0, ∃ a positive integer N

such that
]2^ ‖ _ 5 ‖ & [, ∀ ^, U ≥ •
→,

That is, is said to be Cauchy sequence in X iff

]2^ ‖ _ 5 ‖ → 0 C@ ^, U → ∞
→,

 The space X is said to be complete if every Cauchy sequence in converges to an element

in .

Keywords
 Normed linear space
 Open sphere
 Closed sphere
 Bounded set
 Compact set
 Convergent sequence
 Cauchy Sequence
 Banach space

Self Assessment
1. Which of the following is not a requirement for a normed linear space.

A. Associativity
B. Linearity
C. Triangle inequality
D. Homogeneity

2. If two vectors in a normed linear space have norms equal to zero , then

A. The vectors must be orthogonal

B. The vectors must be the zero vector
C. The vectors must be linearly dependent.
D. The vectors must be equal

3. In a normed linear space , the zero vector is unique because:

A. It satisfies the homogeneity property

B. It is defined as the multiplicative identity
C. It is defined as the additive identity
D. None of the above is correct

Lovely Professional University 9

Notes

Functional Analysis

4. Which of the following is not a norm on a normed linear space.

A. Euclidean norm
B. Taxicab norm
C. Supremum norm
D. Inner product norm

5. Which of the following is not a property of norm in general.

A. || || ≥ 0
B. || + || ≤ || || + || ||
C. ||€ || = €|| ||
D. || || = 0, iff =0

6. Which of the following is true about a Banach space

A. It is a finite dimensional vector space

B. It is a normed linear space that is complete
C. It is a vector space with finite no of elements
D. None of the above

7. Which of the following statement is true about a complete normed linear space.

A. Every Cauchy sequence converges within the space.

B. Every bounded sequence converges within the space.
C. Every convergent sequence is bounded within the space.
D. None of the above

8. In a normed linear space , if the norm of a vector is zero, then the vector must be:

A. Zero vector
B. Unit Vector
C. A non zero vector
D. An infinite vector

9. A complete normed space is known as a:

A. Hilbert space
B. Banach space
C. Compact space
D. Euclidean Space

10. Which of the following is a Banach space.

10 Lovely Professional University

 Notes

Unit 01: Normed Linear Space and Banach Spaces I

A. Space of all polynomial functions on BC, DE with supremum norm

B. Space of all continuous functions on BC , DE with supremum norm.
C. Space of all polynomial functions on BC, DE with the p- norm.
D. Space of all continuous functions on BC, DE with the p- norm.

11. Consider the statements.

(i) Every finite dimensional normed linear space is a Banach space.

(ii) Every Banach space is finite dimensional linear space.

A. Only (i) is true

B. Only (ii) is true
C. Both (i) and (ii) are true
D. Neither (i) nor (ii) is true.

12. Which of the following is true in normed space.

A. Union of any family of open sets is open.

B. Intersection of any family of open sets is open.
C. Union of any family of closed sets is closed.
D. Intersection of any family of closed sets is open.

13. If - ≥ q ≥ 1, which of the following is true.

A. ] ⊂ ]x
B. ] ⊃ ]x
C. ] = ]x
D. None of the above

14. Consider the statements:

(i) Every normed space is complete.

(ii) Every normed space can be identified as a dense subspace of a Complete normed space

A. Only (i) is true.

B. Only (ii) is true.
C. Both (i) and (ii) are true.
D. Neither (i) nor (ii) is true.

15. The linear span of empty set equals:

A. Zero subspace
B. Empty set

Lovely Professional University 11

Notes

Functional Analysis

C. The whole space

D. None of the above

Answers for Self Assessment

1. A 2. B 3. C 4. D 5. C

6. B 7. A 8. A 9. B 10. B

11. A 12. A 13. B 14. B 15. A

Review Questions
1. Define a normed linear space.
2. What is definition of norm in normed linear space.
3. State triangle inequality property for a normed linear space.
4. Explain the concept of convergence in a normed linear space.
5. Define a Cauchy sequence in a normed linear space.
6. What is the difference between a normed linear space and a metric space.
7. Define a Banach Spaces.
8. What are the key properties that a space must satisfy to be considered a Banach space.
9. What is the difference between a normed space and a Banach space.

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.
 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.
 B.V. Limaya, Functional Analysis.

12 Lovely Professional University

 Notes

Unit 02: Normed Linear Space and Banach Spaces II

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 02: Normed Linear Spaces and Banach spaces II

CONTENTS
Objectives
Introduction
2.1 Finite Dimensional Normed Space and Subspaces
2.2 Quotient Space of a Normed Space and its Completeness
2.3 Dual Space and Completeness
2.4 Equivalent Norms
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Finite dimensional Normed Space and Subspaces.

 Quotient space and its completeness
 Dual space and completeness
 Equivalent Norms.

Introduction
In this chapter, We introduce the idea of finite dimensional normed spaces and subspaces . These
spaces have some pleasant and useful properties . Such spaces are all Banach spaces. Further, we
also discuss quotient space, dual space and their completeness. Finally, we shall see that any two
norms on finite dimensional normed spaces are equivalent.

2.1 Finite Dimensional Normed Space and Subspaces

In this section, we first introduce the formal definition of Schauder basis of a normed space which
serves as the building block of the subsequent section.

Definition. A collection ={ , , … , e , … } of elements of a normed space is said to be a basis

for if :

(i) is linearly independent set and

(ii) For each ∈ X, there are uniquely determined scalers
α ,α ,…,α …
such that
‖ −∑ ‖ = 0.
→

If is a basis for , then each ∈ is uniquely expressed as:

Lovely Professional University 13

Notes

Functional Analysis

A normed space is said to be finite dimensional if it has finite basis, otherwise is said to be
infinite dimensional.

Example. The space has

= !1,0,0, … ,0#, = !0,1,0, … #, … , = !0,0, … ,1# as a basis.

Definition. A non empty subset $ of a normed space is said to be a subspace of if

(i) $ is a (linear) subspace of considered as a linear space, and

(ii) $ is equipped with the norm ||. ||' induced by the norm ||. || on .
i,e (| |(' = (| |(, ∀ ∈ $. We may denote the subspace !$, ||. ||' # simply by $.

Theorem. Let $ be a subspace of a normed space . Then $ is complete ⇒ $ + , -+ ..

Proof. Suppose $ is complete and let be a limit point of $. Then every open sphere centered at
contains points of $ (other than x). In particular, the open sphere / 0 ! #, where n is a positive integer
1

contains a point of $ other than . Thus { } is a sequence in $ such that

1
‖ − ‖2 , ∀n.
3
⇒ lim → = in X.
⇒ { } is a Cauchy sequence in and hence in $. But $ being complete , it
follows that ∈ $. Thus $ is closed.

Theorem. Let $ be a subspace of Banach space . Then $ is closed ⇒ $ is complete.

Proof. Let $ be closed and let { } be Cauchy sequence in $. Then it is Cauchy in . But being
complete, ⇒ ∃ ∈ such that → . Either ∈ $, then we are done, or each neighbourhood of
contains points !9 #. As such, is a limit point of $. But $ being closed , implies ∈ $. Thus $ is
complete.

Corrolory. Let $ be a subspace of a Banach space . Then $ is complete iff $ is closed.

2.2 Quotient Space of a Normed Space and its Completeness

In this Section, we shall consider one of the most useful methods of constructing new Banach
spaces from the given Banach spaces.
Let be a normed space and $ a subspace of . For any ∈ , the set
: $ = { : ;: ; ∈ $}
is called a coset of $ determined by or a translate of / by . The set
{ : $ ∶ ∈ X}
Of all cosets of $ in is a linear space under addition and scalar multiplication defined by
: $ : ; : $ = : ; : $, , ; ∈
And ! : $# = : $, ∈ , ∈ >,
This set of cosets of $ in is called a quotient space of by $ and is denoted by /$.
For any subspace $ of a linear space , the dimension of /$ is called the deficiency of $.
We can make /$. a normed linear space as follows:
Let ||. || be the norm in . For an : $ @ /$, put
‖ : $‖ = 3A|| : ;|| = .! , $#
B∈'
Where . is the metric induced by the norm ||. || on X.
Theorem. If $ is a closed subspace of a normed space ! , ||. ||#,then /$ is also a normed space
under the norm defined by
‖ : $‖ = 3A|| : ;|| = .! , $#.
B∈'

14 Lovely Professional University

 Notes

Unit 02: Normed Linear Space and Banach Spaces II

Proof. It is obvious from the definition that

‖ : $‖ ≥ 0
Also ‖ : $‖ = 0 if and only if 3A|| : ;||, so that by the property of infimum , there is a
B∈'
sequence {; } in $ such that
(| : ; |( → 0 as 3 → ∞
But then : ; →0 that is ; → − as 3 → ∞. Since $ is closed subspace , ∈ $
Hence
: $ = $, the zeroth element of /$.
Now, let : $, ; : $ ∈ /$, , ; ∈ X .Then
: $ : ; : $ = : ; : $ ∈ /$
By definition of ||. || in /$, there are sequences { } and {; } in $ such that
lim → (| : |( = ‖ : $‖ , lim → (|; : ; |( = (|; : $|( ,
Hence , for any , ; ∈ and the definition of infimum,
(| : $ : ; : $|( = (| : ; : $|( ≤ || : ; : : ; ||

≤ (| : |( : (|; : ; |(
Taking limits as 3 → ∞, we have :
|| : $ : ; : $|| = (| : ; : $|(
≤ lim → (| : |( : lim → (|; : ; |(
≤ (| : $|( : (|; : $|(

So that condition (ii) is satisfied.

Now to prove (iii), For any scalar and : $ ∈ X/Y, consider an element
! : $# = x : Y
If = 0, then
(| ! : $#|( = (|0. : $|( = (|$|( = 0 = | |(| : $|( .
Let 9 0. Then
(| : $|( = 3A|| : ;||
B∈'

= 3A || : ; I ||
B H ∈'
= | | 3A || : ; I ||
B H ∈'
== | |(| : $|( .
Hence ! /$, ||. || # is a normed space.
Now, we discuss the question of completeness of /$ if is complete. In a support of this we prove
the following theorem.
Theorem. Let $ be a closed subspace of a Banach space . Then /$ with the norm defined by
(| : $|( = 3A|| : ;|| = .! , $#
B∈'
is also a Banach space.
Proof. To prove that J/K is a Banach space , we have to prove that every Cauchy sequence in J/K
Converges to a point J/K. Since a Cauchy sequence convergent if and only if it has a convergent
subsequence , we shall show that every Cauchy sequence in J/$ contains a convergent
subsequence .
Let : $, ∈ be a Cauchy sequence in /$. Then , given any @ > 0, there is a natural number
3 such that:
∀ , 3; , 3 ≥ 3 ⇒ (| N : $ − ! : $#(| = (| N − : $|( 2 @
Take @ = and = 3 , 3 = 3 O 1. Then
1
|| :$−P : $Q|| = R( − : $(R 2
0 0 0 0O
2
If we choose @ = , then there is a natural number 3 such that
T

Lovely Professional University 15

Notes

Functional Analysis

1
R( :$−P : $Q(R = R( − : $(R 2
U UO U UV0
4
Continuing in this way, we see that, in general there is a natural number 3X such that
1
R( Y : $ − P YO : $Q(R = R( Y − YO : $(R 2 X
2
In each Y : $ and Y O : $, select vectors ;X , ;XO respectively such that
1
(|;X − ;XO |( 2 X
2
Then, for any Z I > Z,
||;X − ;X H || = ||;X − ;XO : ;XO − ;XO : ⋯ ;X H\ − ;X H ||
≤ ||;X − ;XO || : ||;XO − ;XO || : ⋯ ||;X H\ − ;X H ||
2 Y : YV0 : ⋯: YH
0

2 UY
0 = Y]0
→ 0 as Z → ∞.
\
U

Thus {;X } is a Cauchy sequence in . Since is complete, {;X } converges to a point of . Hence
R( Y
: $ − !; : $#(R ≤ (|;X − ;|( → 0 as Z → ∞
So that the subsequece
Y
:$ →;:$ ∈ /$
But then : $ → y : Y. Hence /$ is complete.

2.3 Dual Space and Completeness

Let be a normed linear space and let _ be a scaler field associated with . This field is also a
normed linear space with norm defined as

‖ ‖ = | |; ∈ _,

then

1. A linear operator I : → _ is called a functional.

2. A functional I : → _ is said to be continuous at a point ` of , if for each ϵ > 0, there
exists δ > 0 such that ‖ − ` ‖ 2 δ implies | I ! # − I ! ` #| 2 ϵ. We say that I is
continuous on if and only if it is continuous on each point of .
3. A functional I : → _ is said to be linear if
I! :c #= I! # : c I ! #, ∀ , ∈ ; , c ∈ _.

4. A linear functional I
is said to be bonded if there exists d > 0 such that

| I ! #| ≤ d‖ ‖, ∀ ∈ .

5. The set of all linear functionals defined on is itsel a linear space , if addition and scaler
multiplication are defined by:

! I
: I #! #= I! #: I! #

!αx I #! # = αx I ! #

And is denoted by e
, called the algebraic dual ( conjugate) space of .

6. A norm of a linear functional I ∈ e is defined as:

‖ I ‖ = +fg | I |
‖h‖

= +fg | I |
‖h‖i

|h H h|
= +fghj` ‖h‖

Note that | I | ≤ ‖ I ‖ ‖ ‖; ∀ ∈ .

7. The set of all bounded (continuous) linear functionals defined on is a linear subspace of
e and is denoted by I .

16 Lovely Professional University

 Notes

Unit 02: Normed Linear Space and Banach Spaces II

A norm on I is given by (6). The linear space I normed in this way is called normed
conjugate of . Sometimes it is denoted by ∗ .
e
Remark:- Since is a linear space, we may also consider its algebraic dual (or conjugate) space
e e
which we denote by P Q or ee
, that is the class of all linear functionals on e
.We shall
denote elements of ee by II (i, e II : e → _, the scaler field associated with e
) and we
shall use the notation II ! I # for the value of II at I .

Theorem. Let be a norm linear space, then the norm conjugate space I
of is complete.

Proof. Let { I } be a Cauchy sequence in I , then by definition of Cauchy sequence , for every @ > 0,
there exists positive integer l such that

‖ I
N − I‖
2 @whenever , 3 ≥ l.

Consequently for each ∈ ,

| N!
I #− I! #| = |! I
N − I# |≤‖ I
N − I ‖‖ ‖ 2 @‖ ‖ , ∀ , 3 ≥ l…………(1)

Which shows that { I ! #} ia a Cauchy sequence in the space -m n for each ∈ . Since the scaler
field -m n is complete, so { I ! #} converges to a limit depending on which we denote
by I ! #.
I! #= I! #.
That is lim →

Thus defining a functional I on . We show that I

∈ I
and for this it is enough to show that I
is
linear and bounded.

First we show that I

is linear, since for scalers o , o and vectors , in , we have
I !o :o # = lim I !o
:o #
→

= lim →
I !o
# : lim I !o
#
→

= o lim I!
# : o lim I!
#
→ →

=o I! #:o I! #.
I
Which shows that is linear.

Now we show that I is bounded and hence continuous. Since { I } is a Cauchy sequence, so it is
bounded . Therefore by definition, there exists a constant _ > 0 such that ‖ I ‖ ≤ _; ∀3.

For ∈ , we have

| I! #| ≤ ‖ I ‖‖ ‖

≤ _‖ ‖; ∀ 3.

Taking limit as 3 → ∞, we get

| I ! #| _‖ ‖; ∀ 3.

Which shows that I is bounded and hence continuous . Hence I ∈ I.

To complete the proof , it remains to show that I → I

.

By (1), we have

| N!
I #− I! #| = @‖ ‖ , ∀ , 3 ≥ l.

Since the results holds for every ≥ l.

N! #→
I I! #

We may let → ∞. Thus letting , we get

→

| I ! #− I! #| ≤ @‖ ‖; ∀3 ≥ l

Implies |! I − I #! #| ≤ @‖ ‖; ∀3 ≥ l

Lovely Professional University 17

Notes

Functional Analysis

By taking Sup over all of norm, we have

‖ I
− I‖ ≤ @; ∀ 3 ≥ l.

Which shows that { I

} converges to I
. Consequentely I
is complete.

2.4 Equivalent Norms

A norm ||. || on a vector space is said to be equivalent to a norm ||. || on if there are positive
numbers and c such that ∀ ∈ we have

(| |( ≤ (| |( ≤ c(| |( .

This concept is motivated by the fact that Equivalent norms on define the same topology for .

0
Example. Let = with norm ‖p‖ = | | : | | ; p = ! , #∈ and ‖p‖ = !∑ | | #U ,
then show that ‖p‖ and ‖p‖ are equivalent norms.

Solution:- we have

‖p‖ = | | : | | = ∑ | |=∑ !1#| |

0 0
≤ !∑ !1# #U !∑ | | #U

= √2‖p‖

This implies ‖p‖ ≤ √2 ‖p‖

Or ‖p‖ ≤ ‖p‖ ……………(1)

√

Now,

‖p‖ = r | | s = t| | : | |

≤| |:| | ‖p‖

Or

‖p‖ ≤ !1# ‖p‖ ……………(2)

From (1) and (2)

1
‖p‖ ≤ ‖p‖ ≤ !1# ‖p‖
√2
This shows that ‖p‖ and ‖p‖ are equivalent norms.

Theorem. The relation of ‘ being equivalent to ‘ among the norms that can be defined on a linear
space is an equivalence relation.

Proof. In order to show that relation of ‘ being equivalent to ‘ among the norms is an equivalence
relation, we have to show that it is reflexive, symmetric and transitive,

Reflexive. We have for any norm ‖. ‖ on and for any ∈

‖ ‖ ≤ ‖ ‖ ≤ c‖ ‖

Is satisfied for = c = 1. Hence ‖. ‖~‖. ‖.

Symmetric. If ‖. ‖ ~‖. ‖ then there are positive numbers and c such that ∀ ∈ X, we have

‖ ‖ ≤ ‖ ‖ ≤ c‖ ‖

18 Lovely Professional University

 Notes

Unit 02: Normed Linear Space and Banach Spaces II

1 1
⇒ ‖ ‖ ≤‖ ‖ ≤ ‖ ‖
c
Hence ‖. ‖ ~‖. ‖ .

Transitive. If ‖. ‖ ~‖. ‖ and ‖. ‖ ~‖. ‖ v then there are positive numbers , c, and c such that
∀ ∈ X, we have

‖ ‖ ≤ ‖ ‖ ≤ c‖ ‖ and ‖ ‖v ≤ ‖ ‖ ≤ c ‖ ‖v

1 c c
⇒ ‖ ‖v ≤ ‖ ‖ ≤ ‖ ‖ ≤ ‖ ‖ ≤ . c ‖ ‖v

1 c
⇒ ‖ ‖v ≤ ‖ ‖ ≤ . c ‖ ‖v

⇒ ‖ ‖v ≤ ‖ ‖ ≤ cc ‖ ‖v

Since , c, , c > 0.

Hence ‖. ‖ ~‖. ‖ v .

Consequently the relation of ‘ being equivalent to ‘ among the norms that can be defined on a
linear space is an equivalence relation.

Theorem. Any two equivalent norms on a linear space define the same topology on .

Proof. Let ‖. ‖ ~‖. ‖ then there are positive numbers and c such that ∀ ∈ X, we have

‖ ‖ ≤ ‖ ‖ ≤ c‖ ‖

We show that every basic open ball in ! , ‖. ‖ # is open in ! , ‖. ‖ # and conversely.

For an ∈ , let ! ; m# be an open ball in ! , ‖. ‖ #, then we show that it is open ball in ! , ‖. ‖ #.

For this let ; ∈ ! ; m# then ‖ − ;‖ = m 2 m

w\w0
!;; m I # in ! , ‖. ‖ # where m I =
x
Consider

Then for any y ∈ !;; m I # we have ‖y − ;‖ 2 m I then

‖y − ‖ = ‖y − ; : ; − ‖ ≤ ‖y − ;‖ : ‖; − ‖

‖y − ‖ ≤ c‖y − ;‖ : m Since ‖. ‖ ~‖. ‖ and ‖ − ;‖ = m 2 m

m−m
‖y − ‖ 2 cm I : m = c z { : m = m ⇒ ‖y − ‖ 2 m
c

Hence y ∈ ! ; m# implies ∈ !;; m I # ⊆ ! ; m# . Hence ! ; m# is open ball in ! , ‖. ‖ #. Similarly

we can show that every basic open ball in ! , ‖. ‖ # is open in ! , ‖. ‖ #. Hence any two equivalent
norms on a linear space define the same topology on .

The next theorem shows that equivalent norms preserve Cauchy property of sequence.

Theorem. Let ‖. ‖ and ‖. ‖ be equivalent norms on a linear space , then every Cauchy sequence
in ! , ‖. ‖ # is also Cauchy sequence in in ! , ‖. ‖ # and conversely.

Proof. Let { } be a Cauchy sequence in in ! , ‖. ‖ #, then for given ϵ > 0 there exists 3` ∈ l

Such that

‖ N − ‖ 2 ϵ; ∀m, n > n`
1 ϵ
‖ − ‖ ≤ ‖ − ‖ 2 ; ∀m, n > n`
N
α N
α
‖ N − ‖ 2 ϵI ∀m, n > n`

Hence { } is a Cauchy sequence in in ! , ‖. ‖ #. Similarly, we can prove converse.

Lovely Professional University 19

Notes

Functional Analysis

Summary
 A normed space is said to be finite dimensional if it has finite basis, otherwise is said
to be infinite dimensional.
 Let be a normed space and $ a subspace of . For any ∈ , the set
: $ = { : ;: ; ∈ $}
is called a coset of $ determined by or a translate of / by . The set
{ : $ ∶ ∈ X}
Of all cosets of $ in is a linear space under addition and scalar multiplication defined by
: $ : ; : $ = : ; : $, , ; ∈
And ! : $# = : $, ∈ , ∈ >,
This set of cosets of $ in is called a quotient space of by $ and is denoted by /$.
 For any subspace $ of a linear space , the dimension of /$ is called the deficiency of $.

Keywords
 Subspace
 Basis
 Dimmension
 Finite dimension normed space
 Quotient space
 Dual space
 Completeness
 Norm

Self Assessment
1: What is normed space?

A. A vector space equipped with a norm.

B. A vector space equipped with an inner product.
C. A vector space equipped with a metric.
D. None of the above.

2: Which of the following statements is true about a normed space?

A. Every normed space is finite-dimensional.

B. Every normed space is infinite-dimensional.
C. A normed space can be either finite-dimensional or infinite-dimensional.
D. A normed space cannot have a dimension.

3: Which of the following is true about subspaces of a normed space?

A. Every subspace is finite-dimensional.

B. Every subspace is infinite-dimensional.
C. A subspace can be either finite-dimensional or infinite-dimensional.
D. A subspace cannot have a dimension.

4: Which of the following statements is true about finite-dimensional normed spaces?

A. Every finite-dimensional normed space is complete.

B. Every finite-dimensional normed space is incomplete.
C. A finite-dimensional normed space can be either complete or incomplete.
D. None of the above.

20 Lovely Professional University

 Notes

Unit 02: Normed Linear Space and Banach Spaces II

5: What is the definition of a quotient space?

A. A space obtained by dividing a normed space by a subspace.

B. A space obtained by dividing a normed space by a linear transformation.
C. A space obtained by dividing a normed space by a scalar.
D. A space obtained by dividing a normed space by a scalar multiple.

6: Which of the following conditions ensures the completeness of the quotient space?

A. The subspace is open.

B. The subspace is dense.
C. The subspace is closed.
D. The subspace is connected.

7: Two norms ‖. ‖ and ‖. ‖ on a normed linear space are equivalent, then there exists
positive constants α and β such that:

A. α‖. ‖ ≤ ‖. ‖ ≤ β‖. ‖ for all in .

B. ‖. ‖ = α‖. ‖ for all in .
C. ‖. ‖ = β‖. ‖ for all in .
D. None of the above.

8: Two norms ‖. ‖ and ‖. ‖ on a normed linear space are said to be equivalent if :

A. They induce the same topology on X.

B. They have the same dimension.
C. They have the same norm constant.
D. None of the above.

9: Which of the following statements about the quotient space is true?

A. The quotient space is always finite-dimensional.

B. The quotient space is always a normed linear space.
C. The quotient space is isomorphic to the original normed linear space .
D. The quotient space is always a complete space.

10: Which of the following is a necessary condition for the quotient space to be finite-
dimensional?

A. The original normed linear space must be finite-dimensional.

B. The subspace must be finite-dimensional .
C. The original normed linear space and the subspace must have the same dimension.
D. (D)None of the above.

11: Which of the following statements is true about a subspace of a normed linear space?

A. It must contain all the vectors of the normed linear space.

B. It must contain the zero vector.
C. It must be a finite-dimensional space.
D. It must be a closed set.

12: Let V be a normed linear space and W be a subspace of V. Which of the following statements
is true about the dimension of W?

Lovely Professional University 21

Notes

Functional Analysis

A. The dimension of W is always greater than or equal to the dimension of V.

B. The dimension of W is always less than or equal to the dimension of V.
C. The dimension of W is always equal to the dimension of V.
D. None of the above.

13: What is the dual space of a normed linear space?

A. The space of all linear transformations from the given space to its scalar field .
B. The space of all linear functionals from the given space to its scalar field.
C. The space of all continuous linear transformations from the given space to its scalar field
D. The space of all continuous linear functionals from the given space to its scalar field.

14: Which of the following statements is true about the dual space of a normed linear space?

A. The dual space is always finite-dimensional.

B. The dual space is always infinite-dimensional.
C. The dual space is always a Banach space.
D. The dual space can be finite-dimensional or infinite-dimensional.

15: Which of the following statements is true regarding the completeness of the dual space of a
normed linear space?

A. The dual space is always complete.

B. The dual space is never complete.
C. The dual space is complete if and only if the original normed linear space is finite-
dimensional.
D. The completeness of the dual space depends on the specific norm chosen for the original
space.

Answers for Self Assessment

1. A 2. C 3. C 4. A 5. A

6. C 7. A 8. A 9. B 10. C

11. B 12. B 13. D 14. D 15. C

Review Questions
1. Define a normed space and give an example of a finite dimensional normed space.
2. Define subspace of a normed space.
3. Is the Zero subspace always a proper subspace.
4. Define Quotient space of a normed linear space.
5. State the definition of equivalent norms.
6. Give an example of two norms that are equivalent.
7. Define dual space of a normed linear space.

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.

22 Lovely Professional University

 Notes

Unit 02: Normed Linear Space and Banach Spaces II

 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.

 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.
 B.V. Limaya, Functional Analysis

Lovely Professional University 23

 Notes

Unit 03: Bounder Linear Operator and its Properties

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 03: Bounded Linear Operator and its Properties

CONTENTS
Objectives
Introduction
3.1 Bounded and Continuous Linear Operators
3.2 Linear Functional
3.3 Compactness and Finite Dimensional Space
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Bounded and continuous linear operator

 The null space of a linear operator
 Norm of a bounded linear operator
 The space of bounded linear operators
 Linear functional
 Compactness and finite dimensional space

Introduction
In this chapter, We introduce the idea of bounded and continuous linear operators. The study of
bounded and continuous linear operators serves as a powerful tool to analyze and understand the
behavior of functions between normed vector spaces. Further, we also discuss kernel or null space
of a linear operator , norm of a linear operator and the space of bounded linear operators. Finally
we discuss linear functional and the Compactness and finite dimensional space.

3.1 Bounded and Continuous Linear Operators

In calculus we consider the real line and real-valued functions on (or on a subset of ).
Obviously, any such function is a mapping of its domain into . In functional analysis we consider
more general spaces, such as metric spaces and normed spaces, and mappings of these spaces.In
functional analysis, an operator is defined as a mapping between two vector spaces. Specifically, let
and be two vector spaces (typically normed vector spaces or Banach spaces) over the same field
(usually ). An operator is a function that maps elements from to elements in .
Different types of operators are commonly encountered in functional analysis, such as linear
operators, bounded operators, compact operators, self-adjoint operators, unitary operators, and
many others. Each type of operator has its own set of properties and characteristics, which are
studied to understand the behavior and structure of the operator. Operators play the fundamental

24 Lovely Professional University

Notes
Functional Analysis

role in functional analysis, as they provide a way to study the relationship between vector spaces,
mathematical objects in a functional-analytic setting.
Before defining a bounded linear operator , we recall some definitions and results.
be normed spaces over a field . We say that : → is a linear operator
∀ , ∈ λ λ ∀ ∈ and λ ∈ .
Definition. Let and
if is linear (that is

Example. Let be any normed space, then the identity function : → defined by :
, ∈
Is a linear operator .
Here for λ , λ ∈ and , ∈ ,
λ λ λ λ
λ λ .

Example. For any linear spaces , , the function 0: → defined by:

0 0, ∈
Is a linear operator.
Note that zero operator is also called null operator or trivial operator.

Example. In the space , , define a function : , → , by:

#
! " ", ∈ , .
$

Then is a linear operator.

Example. Let & be the space of all analytic functions over and ': & → & be defined by:
' (, ∈ & and ( is the derivative of . Then ' is a linear operator.

Example. Consider the linear space ) of all polynomials * with real coefficients defined

on 0,1 , then the mapping ' defined by ' * , is a linear operator from ) into itself.
,-
,#

The kernel or null space of a linear operator

Let : → be a linear operator. Then the set of those elements of which are mapped onto the
zero element of is a subspace of called the kernel or null space of and is denoted by &. . To
see the &. is a subspace of , let , ∈ &. .
Then 0, 0 and for any λ , λ ∈ ,
λ λ λ λ , by linearity of ,
So that λ λ ∈ &. .
Now we define the Continuous linear operator. Of special interest among the class of all linear
operators are those which are continuous . Since every normed space is also a metric space ,
continuity of an operator is always with respect to the metric defined by the norm. Let and be
normed spaces . A linear operator : → is said to be continuous at a point / ∈ if given ϵ 1 0,
there is a real no δ δ ϵ 1 0 such that
∀ ∈ ,‖ 4 /‖ 5 δ⇒‖ 4 /‖ 5 ϵ.

Lovely Professional University 25

 Notes

Unit 03: Bounder Linear Operator and its Properties

is said to be continuous on if it is continuous at every point of . : → is continuous if and

only if, for all sequences { 8} which converges to , 8 converges to .
Another concept associated with linear operator defined on a normed space is that of boundness
which is equivalent to continuity of the operator.
A linear operator : → is said to be bounded if there is a constant : 1 0 such that
‖ ‖ 5 :‖ ‖ ∀ ∈ .
The concepts of continuity and boundness of a linear operator are equivalent is shown in the
following theorem.
Theorem. Let : → be a linear operator . Then
(a) is continuous on if and only if is bounded .
(b) is continuous if and only if it is continuous at 0 ∈ .
(c) If is continuous on then &. is closed in .
Proof. Suppose that is continuous on . Then it is continuous at each / ∈ . So given any ϵ 1 0,
there is a δ 1 0 such that
∀ ∈ ,‖ 4 /‖ 5 δ⇒‖ 4 /‖ 5 ϵ.
Let ∈ and put

4
; ;
/ ‖<‖
, i,e / ‖<‖

and ‖ 4 /‖ = = 5 δ,
; ;
‖<‖
Then, using the linearity of

δ
we have

‖ 4 /‖ ‖ 4 ‖ > ? A>
/
2‖ ‖
‖ ‖5ϵ
;
‖<‖

‖ ‖5 ‖ ‖
B
;
So that,

5 :‖ ‖, :
B
;
.

Hence is bounded.
Alternatively, suppose that is continuous but not bounded. Then for each natural number n ,

‖ 8‖ ‖ 8 ‖.
there is an in , such that
1

Let 8 8‖#C ‖ 8

Then ‖ 8‖ ‖ 8‖
8‖#C ‖

1 1
8‖#C ‖
8‖#C ‖

Now ‖ 8 ‖ ‖ 8‖ → 0 as → ∞.
8‖#C ‖ 8

‖ 8‖ →0 ⇒‖ 8‖
By continuity of ,
→0
But ‖ 8‖ ≥ 1 ∀ , a contradiction . Hence is bounded.
is bounded . Then there is a real number : 1 0 such that
‖ ‖ ≤:‖ ‖∀
Conversely suppose that
∈ .
So, for any ϵ 1 0, choose δ
B
G
. Then

‖ 4 /‖ 5 δ ⇒‖ 4 /‖ ‖ 4 / ‖

26 Lovely Professional University

Notes
Functional Analysis

≤:‖ 4 /‖

5 ϵ.
Hence is continuous.
(b) Suppose that is continuous on , then it is continuous on 0 ∈ .
Coversely suppose that is continuous at 0 ∈ . Then, with / 0, given any ϵ 1 0, there is a
δ 1 0 such that
∀ ∈ ,‖ 4 /‖ ‖ ‖ 5δ⇒‖ 4 /‖ ‖ ‖ 5 ϵ.

/ ∈ ,
‖ 4 /‖ ≤ δ ⇒‖ /‖ ‖ ‖5 ϵ
Hence, for any
4 4 /

So is continuous at / and therefore also on .

(c) Suppose that is continuous and let H be a limit point of &. . Then there is a sequence { 8} in
&. such that
lim 8
8→L

By the continuity of ,
0 lim8→L 8 .
Hence ∈ :. and :. is closed.

Norm of a Bounded Linear Operator

Let : → be a bounded linear operator . Then there is a real number : 1 0 such

‖ ‖≤:‖ ‖∀ ∈

≠ 0.Then ≤:∀ ∈ , ≠ 0.
‖N#‖
‖#‖
Suppose that

So : is an upper bound for . The leaset upper bound OP*

‖N#‖ ‖N#‖
‖#‖ #Q/∈# ‖#‖
is called the norm of and is

denoted by ‖ ‖. Thus

‖ ‖ OP*
‖N#‖
#Q/∈# ‖#‖
.

{0}, then ‖ ‖ 0.
It is clear from definition of ‖ ‖, ‖ ‖ ≥ 0 and ‖ ‖ 0 if and only if ‖ ‖
Note:- If
0∀ ∈ , that is 0.

‖ ‖ ≤ ‖ ‖‖ ‖ ∀
Note:- We have another relation for a bounded linear operator namely
∈ .

As we have ‖ ‖ OP* ≤:
‖N#‖
#Q/∈# ‖#‖

≤‖ ‖∀ ≠0 ∈
‖N#‖
‖#‖
Also by definition of supremum,

This gives ‖ ‖ ≤ ‖ ‖‖ ‖∀ ≠ 0 ∈

0, 0 0 so that ‖ 0‖ ‖0‖ ≤ ‖ ‖‖0‖.

Thus ‖ ‖ ≤ ‖ ‖‖ ‖ ∀
But for
∈ .
Note:- We can also write ‖ ‖ sup‖#‖U ‖ ‖=sup‖#‖V ‖ ‖

As ‖ ‖ OP* OP* = X‖<‖Y= sup‖#‖U ‖ ‖ , when

‖N#‖ < <
#Q/∈W ‖#‖ <Q/∈W ‖<‖

Lovely Professional University 27

 Notes

Unit 03: Bounder Linear Operator and its Properties

Because of various equivalent forms of norms of a linear operator yields the following definition of

is said to be bounded if and only if ‖ ‖ is finite.

boundedness.
A linear operator : →

Example. The identity operator : → defined by :

∀ ∈

is bounded as ‖ ‖ OP* 1.
‖Z # ‖ ‖#‖
#Q/∈W ‖#‖ ‖#‖

Example. The null operator (zero operator ) 0: → defined by:

0 0∀ ∈

is bounded as as ‖0‖ OP* 0.

‖/ # ‖ ‖/‖
#Q/∈W ‖#‖ ‖#‖

Example. The operator [: \] → \^ defined by

8 d

_ ` a` bc be .
cf

cU bU

Where H , ,…, d , h , ,…, 8

And . , . , … .d and . , . ( (
, … .8( are basis of \] and \^ respectively is bounded.
As c ∑d
bU bc b, so that
8 8 d
‖_H‖ ‖h‖ `j c j ` k` bc bk
cU cU bU

8 d d

≤ ` a`| b m| . `| | e
cU bU bU

( By Minkowski’s Inequality)

8 d d

≤ n` `| b m| o . `j b j
cU bU bU

≤: ‖ ‖

where : ∑8cU ∑d
bU | b m| and ‖ ‖ ∑d
bU j b j
Hence ‖_H‖ ≤ :‖H‖.
Therefore _ is a bounded and hence a continuous linear operator .

Note:- ‖_‖ p∑8cU ∑d

bU | b m| is called the norm of matrix operator _.

Note:- if ( , ‖. ‖) be a normed space and { , , q, … , 8 } be a linearly independent set of

vectors in . Then, there is a real number r 1 0 such that for all scalers , ,…, 8

‖∑8bU b b‖ ≥ r ∑8bU | b |…………(1)

Theorem. Show that every linear operator on a finite dimensional normed space is bounded .

28 Lovely Professional University

Notes
Functional Analysis

Proof. Let be a finite dimensional normed space and let s . , . , … .8 be a basis of . Let
: → be a linear operator . For any ∈ ,

` b .b
bU

So that, by linearity of

` b .b
bU

Hence

8
‖ ‖ t` b .b t
bU

≤ `| b |‖ .b ‖
bU

≤ ∑8bU | b |, 8 ‖
OP*bU .b ‖…………….(2)
Also by (1) , there is a positive real number r such that
‖ ‖ ‖∑8bU b .b ‖ ≥ r ∑8bU | b |…………(3)

From (2) and (3),we have

‖ ‖ ≥ r ∑8bU | b | ≥ c. ‖ ‖
v

b
I,e

‖ ‖≤ ‖ ‖
c
Or,

‖ ‖ ≤ k ‖ ‖, where : 1 0.
v
y

Hence is bounded linear operator.

: → : → z be bounded linear operators .Then
‖ ‖ ≤ ‖ ‖‖ ‖ ‖ ‖‖ ‖.
Theorem. if and is bounded and

In particular , if : →
‖ 8‖ ≤ ‖ ‖8 .
is a linear operator , then

are bounded , then ‖ ‖, ‖ ‖ exists and are finite. Moreover, for any ∈
‖ ‖ ‖ ‖
Proof. Since ,

≤ ‖ ‖‖ ‖
≤ ‖ ‖‖ ‖‖ ‖
Hence ‖ ‖ ≤ ‖ ‖‖ ‖ ‖ ‖‖ ‖.
In particular, if : →
‖ 8‖ ≤ ‖ ‖8 .
is a linear operator , then by induction on , we have

Lovely Professional University 29

 Notes

Unit 03: Bounder Linear Operator and its Properties

The space of bounded Linear operators.

Let , be normed linear spaces and s , denote the space of all bounded linear operators fron
to .
Theorem. The space s , of all bounded (hence continuous) linear operators from to is a

‖ ‖ OP* ‖ ‖, ∈ .
normed space under the norm defined by

‖ #‖U

Proof. First we show that s , is a linear space .

For this , let {, ∈s , . Define { ∶ → by
{ { ∀ ∈ .
For any α, β scalers and , ∈ ,then
{ • € { • € • €
•{ €{ • €
• { € {
α { β {
Therefore { is a linear operator.
∈ ,
‖{ ‖ OP* ‖ { ‖
Morover, for any

‖ #‖U

OP* ‖{ ‖
‖#‖U

≤ OP* ‖{ ‖ OP* ‖ ‖
‖#‖U ‖#‖U

≤ ‖{ ‖ ‖ ‖………….(1)
Hence { is a bounded linear operator and so is in s , .
It is easy to see that the commutative and associative laws of addition are satisfied in s , . The
function 0: → defined by:
0 0
Is linear and bounded. Also for any ∈s , ,
0 0
Next , for each ∈s , , the function 4 4 , ∈
Is linear and satisfies
4 0
Also ‖4 ‖ ‖ ‖
So 4 ∈s , . Hence s , is an additive abelian group .
Define the scaler multiplication in s , as

α α ;∀ ∈ , ∈s , , and α is scaler.

For , ∈ and , ∈
α α.
α
α α
α α
Also for any ∈ ,‖ ‖ 1

30 Lovely Professional University

Notes
Functional Analysis

‖α ‖ OP*‖ α ‖
OP*‖α. T ‖
|α|OP*‖T ‖

‖α ‖ |α|‖T‖………….(2)
So that

Therefore, α ∈ s , . Thus s ,
, ‖T‖ ≥ 0
is a linear space.
∈s ,
And ‖T‖
Since, for any
0 if and only if 0 ………….(3)
Hence (1), (2) and (3) show that s , is a normed space.
In the next theorem we discuss properties of s , in relation to the properties of .
is a Banach space , then so is s ,
‖T‖ OP* ‖T ‖,
Theorem. Show that, if under the norm defined by
∈ , ∈s ,
‖#‖U
.

is a Banach space and let { 8 } be a Cauchy sequence in s ,

ϵ 1 0 ∋ n/ ∈ N such that ‖T† 4 d‖ ‖: ∈ , ‖x‖
Proof. Suppose , then for given
OP*{‖T† 4 d 1} 5 ϵ ∀ , ˆ ≥ / ………..(1)

‖T† 4 ‖ 5 ϵ ∀x ∈ X, , ˆ ≥
That is ,

d /

So for any ∈ ,{ 8 } is a cauchy sequence in . Since is complete, ,{ 8 } converges in .

Let Š‹ˆ 8
8→L

Where : → , which takes → . We will show that is bounded linear operator .

Since 8 is linear , for any α, β ∈ F, x ∈ X

8 α β α 8 β 8

8 α β Š‹ˆ 8 α β
8→L
Thus,

α Š‹ˆ 8 β Š‹ˆ 8
8→L 8→L

α βT•
Also letting ˆ → ∞ in (1) and using the continuity of norm function, we have

= 8 4 Š‹ˆ d= OP*{= 8 4 Š‹ˆ d =≤ϵ∀ ≠0∈ }

d→L d→L

‖ ‖ ‖: ≠ 0 ∈
That is ,

8 4 OP *‖ 8 4 ≤ ϵ∀n ≥ n/
Hence 8 4 ∈s , . But then

8 4 8 4 , as difference of two elements of s ,

That is T† → T as → ∞ as T† is a Cauchy sequence in s , is in s , .
Hence s , is a Banach space.
Note. Converse of above theorem is also true . That is , if s , ia a Banach space , so is .
To prove this we make use of Hahn-Banach Theorem on normed spaces , which we will discuss
later in this chapter.

3.2 Linear Functional

In the previous section we considered functions called linear operators from one normed space into
another normed space defined over the same field . In this section we shall deal with a special type
of linear operators called linear functionals. These are linear operators from a normed space

Lovely Professional University 31

 Notes

Unit 03: Bounder Linear Operator and its Properties

Over to , where is or and is itself a normed space under the usual norm defined by on
or .
Thus a function : → is said to be linear functional if, for any , ∈ and α, β ∈
α β αf x βf y .
A linear functional : → is said to be continuous at a point / ∈ , if fgiven ϵ 1 0, there is a real
number δ 1 0 such that
‖ 4 /‖ 5 δ implies | 4 / |5ϵ∀ ∈ .
is said to be continuous on , if is continuous at every point of .
is said to be bounded if there is a real number : ≥ 0 such that
| | ≤ k‖ ‖, ∀ ∈ .
: →
| |
As in the case of linear operators we define the norm of a linear by :

‖ ‖ OP* • : ≠0 ∈ ‘
‖ ‖
is a bounded linear functional , so that ‖ ‖ ≤ :, then
| | ≤ ‖ ‖‖ ‖, ∀ ∈ .
Then , if

It is now easy to establish the following equivalent forms of the norm of a linear functional :

| |
‖ ‖ OP* • : ≠0 ∈ ‘
‖ ‖

OP* ’ : || || ≤ 1 ∈ ”
|“ # |
‖#‖

OP*{| |: || || 1∈ }
Theorem. Let : → be a linear functional . Then:
(i) is continuous if and only if is bounded.
(ii) is continuous on if and only if it is continuous at 0 ∈ .
For a linear functional : → , the kernel or null space denoted by &. is defined by:
&. { ∈ : 0}
and is a subspace of .
Proof. The proof of this theorem is same as in the case of linear operators.

Examples of Bounded linear functional

8
Example. Let be the n-dimmensional real normed space with with the norm defined
by :

8
‖H‖ •`| b | , H , ,…, 8 ∈ 8

bU

For any , ,…, 8 ∈ 8, define a function –:

8 → by:

– ` b b , ∈ 8

bU

It is easy to verify that $ is linear functional . By schwartz inequality

32 Lovely Professional University

Notes
Functional Analysis

8 8
| $ H | ≤ •` | b | . •` | b |
bU bU

≤ ‖ ‖‖ ‖, ∈ 8

8.
$

‖ $ ‖ ≤ ‖ ‖………..(1)
Hence is a bounded and hence a continuous linear functional on Also

However if we take , we obtain

8

$ `| b | ‖ ‖
bU

‖ $‖ OP* ≥ ‖ ‖………..(2)
|“— # | “— $
#Q/ ‖#‖ ‖$‖
So that

From (1) and (2), we get

‖ $‖ ‖ ‖.

Example. For the space , of all real continuous functions from , → with the
sup norm ,define a function : , → by :
˜
! " "
$

Then is a linear functional . Also

˜
| |≤! | " | "
$
˜
≤ sup | " | ! "
™∈ š,v $

≤ b4a ‖ ‖

‖ ‖≤
Hence,
4
" 1∀ " ∈ ,
………..(3)
Also, taking / , we get

‖ ‖ OP* ≥ 4 ………..(4)
|Z “ | Z “œ
“Q/ ‖“‖ ‖“œ ‖

‖ ‖
From (3) and (4) we get
4 .

Example. Let r be the space of all convergent real sequences { 8 }. Let : r → be defined
Š‹ˆ 8
8→L
by

‖ ‖
Then is a bounded linear functional with
1

3.3 Compactness and Finite Dimensional Space

Compactness is one of the most important concepts in analysis. We now define compact linear
operator.
be normed spaces . An operator : → is called a compact linear
operator (or completely continuous linear operator) if is linear and if for every bounded subset •
Definition. Let and

of , the image • is relatively compact, that is the closure žžžžžžž

• is compact .

Lovely Professional University 33

 Notes

Unit 03: Bounder Linear Operator and its Properties

The term compact is suggested by the definition . The older term “completely continuous “ can be
motivated by the following lemma, which shows that a compact linear operator is continuous ,
where as the converse is not generally true.
Before proving the lemma, we first note the following .
A compact subset • of a metric space is closed and bounded .
{ ∈ : ‖ ‖ ≤ 1}
I.
II. If a normed space has the property that the closed unit ball •

is compact , then is finite dimensional.

Lemma. Let and be normed spaces. Then:
(a) Every compact linear operator : → is bounded, hence continuous.
(b) If ‹ˆ ∞, the identity operator : → ( which is continuous ) is not compact.
Proof of (a). The unit sphere Ÿ { ∈ :‖ ‖ 1} is bounded. Since
žžžžžžž
is compact ,
Ÿ

sup ‖ ‖5 ∞
Is compact and is bounded by (I), so that

‖#‖U

Hence is bounded and shows that it is continuous.

Proof of (b). The closed unit ball • ∈ : ‖ ‖ ≤ 1 is bounded. If ‹ˆ ∞, then

by (II) • cannot be compact, thus • • • is not relatively compact.

We now prove tha compactness criteria for operators in the following theorem.
Theorem. Let and be normed spaces and : → a linear operator .Then is compact if and
only if it maps every bounded sequence { 8 } in onto a sequence { 8 } in which has a
convergent subsequence.
Proof. If T is compact and {x† } is bounded, then the closure of {T 8} in Y is compact and shows that
{T 8 } contains a convergent subsequence.
Conversely, assume that every bounded sequence {x† } contains a subsequence {x† ¢ }
such that {Tx†£ } converges in Y. Consider any bounded subset B ⊂ X, and let {y† } be any sequence
in T B . Then y† T 8 for some 8 ∈ B , and { 8 } is bounded since s is bounded. By assumption,
{T 8 } contains a convergent subsequence. Hence žžžžžžž
T B is compact because {y† } in T B was
arbitrary, by definition, this shows that is compact.
Next, we study the compactness of finite dimensional linear operator. Prior to that, we recall the
following results.
, any subset • ⊂
and only if • is closed and bounded.
Result 1. ( Compactness) In a finite dimensional normed space is compact if

Result 2. If a normed space is finite dimensional, then every linear operator on is bounded.
Result 3. Let be a linear operator . Then, if ‹ˆ ' 5 ∞, then ‹ˆ ≤ .
Theorem. ( Finite dimensional domain or range) Let and be normed spaces and : →
be a linear operator. Then:
(a) If is bounded and ‹ˆ 5 ∞, the operator is compact.
(b) ‹ˆ 5 ∞, the operator is compact.
Proof. Let { 8} be any bounded sequence in . Then

‖ 8 ‖ ≤ ‖ ‖‖ 8 ‖ shows that { 8 } is bounded . Hence { 8 } is relatively compact by result 1. Since

‹ˆ 5 ∞. It follows that { 8 } has a convergent subsequence. Since { 8 } was arbitrary
bounded sequence in , the operator is compact.

34 Lovely Professional University

Notes
Functional Analysis

Proof of b. It follows from (a) by noting that ‹ˆ 5 ∞ implies boundedness of

‹ˆ ≤ ‹ˆ by result 3.
by result 2 above

Summary
be normed spaces over a field . We say that : →
∀ , ∈ λ λ ∀ ∈ and λ ∈ .
 Let and is a linear operator if is

Let be any normed space, then the identity function : → defined by :

linear (that is

, ∈


is a linear operator .
For any linear spaces , , the function 0: → defined by:
0 0, ∈


is a linear operator.
Zero operator is also called null operator or trivial operator.
, , define a function : , → ,

 In the space by:
#
! " ", ∈ , .
$

Then is a linear operator.

 Let : → be a linear operator. Then the set of those elements of which are mapped onto
the zero element of is a subspace of called the kernel or null space of and is denoted by
&. .
Let and be normed spaces . A linear operator : → is said to be continuous at a point
∈ if given ϵ 1 0, there is a real no δ δ ϵ 1 0 such that


∀ ∈ ,‖ 4 /‖ 5 δ⇒‖ 4 /‖ 5 ϵ.
is said to be continuous on if it is continuous at every point of .
A linear operator : → is said to be bounded if there is a constant : 1 0 such that


‖ ‖ 5 :‖ ‖∀ ∈ .


 Let : → be a bounded linear operator . Then there is a real number : 1 0 such

‖ ‖≤:‖ ‖∀ ∈

≠ 0.Then ≤:∀ ∈ , ≠ 0.
‖N#‖
‖#‖
Suppose that

So : is an upper bound for . The leaset upper bound OP*

‖N#‖ ‖N#‖
‖#‖ #Q/∈# ‖#‖
is called the norm of and

is denoted by ‖ ‖. Thus

‖ ‖ OP*
‖N#‖
#Q/∈# ‖#‖
.

 ‖_‖ p∑8cU ∑d
bU | b m| is called the norm of matrix operator _.

Every linear operator on a finite dimensional normed space is bounded .

: → : → z be bounded linear operators .Then is bounded and ‖ ‖≤


‖ ‖‖ ‖ ‖ ‖‖ ‖.
 if and

 The space s , of all bounded (hence continuous) linear operators from to is a normed

‖ ‖ OP* ‖ ‖, ∈ .
space under the norm defined by

‖ #‖U

Lovely Professional University 35

 Notes

Unit 03: Bounder Linear Operator and its Properties

is a Banach space , then so is s ,

‖T‖ OP* ‖T ‖,
If under the norm defined by
∈ , ∈s ,


‖#‖U
.

A linear functional : → is said to be continuous at a point / ∈ , if given ϵ 1 0, there is a

real number δ 1 0 such that


‖ 4 /‖ 5 δ implies | 4 / |5ϵ∀ ∈ .

Let and be normed spaces . An operator : → is called a compact linear operator (or
is linear and if for every bounded subset • of ,


• is relatively compact, that is the closure žžžžžžž

completely continuous linear operator) if
the image • is compact

Keywords
 Bounded linear operator
 Continuous linear operator
 Null space of a linear operator
 Norm
 Closed
 Compact
 Finite dimmensional

Self Assessment

A. ‖T ‖ ≤ ‖T‖. ‖ ‖
1: If T is a bounded linear operator, then:

B. ‖T ‖ ≥ ‖T‖. ‖ ‖
C. ‖T ‖ ‖T‖. ‖ ‖
D. None of the above

2: Which of the following statements is true about a bounded linear operator?

A. Every bounded linear operator is continuous.
B. Every continuous operator is bounded.
C. Every bounded linear operator is compact.
D. None of the above

3: Which of the following is NOT a property of a bounded linear operator?

A. Preserving the zero vector: 0 0

B. Homogeneity
C. Additivity
D. Surjective

4: What is the null space of a linear operator?

A. The set of all inputs for which the linear operator is not defined.
B. The set of all inputs that map to the zero vector under the linear operator.
C. The set of all outputs for which the linear operator is not defined.

36 Lovely Professional University

Notes
Functional Analysis

D. The set of all outputs that map to the zero vector under the linear operator.

5: Which property holds true for the norm of a linear operator with respect to scalar
multiplication.

A. ‖: ‖ :‖ ‖
B. ‖: ‖ ‖ ‖
G

C. ‖: ‖ : ‖ ‖
D. ‖: ‖ |:|‖ ‖

6: Which of the following statements is true regarding compactness in a normed linear space?

A. Every closed and bounded subset is compact.

B. Every open and bounded subset is compact.
C. Every closed and unbounded subset is compact.
D. Every open and unbounded subset is compact.

7: Which of the following statements about the norm of a linear operator is true?

A. The norm of a linear operator is always zero.

B. The norm of a linear operator is always one.
C. The norm of a linear operator can be negative.
D. The norm of a linear operator is always positive.

8: If the norm of a linear operator T is zero, what can we conclude?

A. T is the zero operator (T(x) = 0 for all x).

B. T is not a linear operator.
C. T is an invertible operator.
D. None of the above.

9: If and are normed spaces, then the space of bounded linear operators s , is a Banach
space if and only if:
A. is a Banach space.
B. is a Banach space.
C. Both and are Banach spaces.
D. Both and are finite dimensional spaces.

10: If E is a normed space and if is the metric induced by the norm, then for any scalar
:, : ,: equals

,
B. |:| ,
A.

Lovely Professional University 37

 Notes

Unit 03: Bounder Linear Operator and its Properties

C. : ,
D. : ,

11: Let be a normed space and be a bounded, non-zero linear functional on . Then, which
of the following is not true?
A. is onto.

C. &.
B. is continuous
is a close subspace of .
D. is an open map.

12: If f is a linear functional on a normed space X, then &. is:

A. Closed in
B. Dense in
C. Either dense or closed in
D. None of the above.

13: Every complete subspace of a normed space is:

A. finite
B. open
C. closed
D. None of the above

14: Every bounded operator of finite rank is :

A. Compact
B. Open
C. Has a zero adjoint
D. None of the above

‹ˆ ˆ _
15: Rank of a linear operator A equals:

‹ˆ &. _
A.

‹ˆ ˆ_∗
B.

‹ˆ &. _∗
C.
D.

Answers for Self Assessment

1. A 2. B 3. D 4. B 5. A

6. A 7. D 8. A 9. B 10. B

11. D 12. C 13. C 14. A 15. A

Review Questions
1. What is a linear operator between two normed spaces?
2. Define a bounded linear operator between normed spaces.
3. What is kernel or null space of a linear operator.

38 Lovely Professional University

Notes
Functional Analysis

4. Define norm of a linear operator.

5. Define what a linear functional on a normed linear operator is.
6. Define kernel or null space of a linear operator.
7. Define norm of a linear operator.
8. Show that every linear operator on a finite dimensional normed space is bounded.
9. Define compact linear operator.
10. Let and be normed spaces and : → be a linear operator. Prove that &. is a
subspace of .

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.
 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.
 B.V. Limaya, Functional Analysis

Lovely Professional University 39

 Notes

Unit 04: Hahn- Banach Theorem and its Consequences

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 04: Hahn-Banach Theorem and its Consequences

CONTENTS
Objectives
Introduction
4.1 Conjugate of an Operator
4.2 Hahn-Banach Theorem (Real and Complex Form) and its Consequences
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Conjugate of an operator
 Convex functional
 The Hahn- Banach theorem for Real spaces
 The Hahn- Banach theorem for complex spaces
 The Hahn- Banach theorem for normed spaces

Introduction
In this chapter, we introduce the idea of conjugate of an operator. Further , we also discuss convex
functional . Finally we discuss different forms of Hahn- Banach theorem and its consequences.

4.1 Conjugate of an Operator

In the context of normed spaces and linear operators, the conjugate of an operator is a concept
related to the duality between a normed space and its dual space. To understand the conjugate of
an operator, we first need Knowledge of normed spaces, dual spaces and linear operators, which
we have already discussed in previous chapters. We now define the conjugate of an operator.
Definition. Let and be normed spaces. Let , be the space of all bounded linear operators
∗ ∗ ∈ , ,
defined from to . Let and be the conjugate spaces of and respectively. Let
∗ ∗
then we define an operator ∶ → as follows:
∗,
For each ∈ . is a mapping from to . It is bounded because both and are bounded .
∗
So . ∈ . Then we put = . …………..(1)
So, for each ∈ , ∈ and . ∈ while ∈ ∗ so that, for each ∈ , ∈ F.
Hence we can write (1) as = . = …………..(2)

40 Lovely Professional University

Notes
Functional Analysis

The operator , defined by (1) or (2), is called the conjugate (or sometimes, the adjoint ) of the
linear operator .
We now discuss some properties of .
I. is linear:
For this, let , ∈ ∗ and α , α ∈ .Then
+ = + , ∈
= +
= +
= . + .
=α +
= + ,∀ ∈ .
Hence,
+ = + …………..(3)
II. is bounded . Here, for any ∈ Y∗ and ∈ ,
‖ ‖= ‖ ‖= ‖ ‖
‖!‖" ‖!‖" ‖ #‖$%

= ‖ ‖
‖!‖" ‖ #‖$%

≤‖ ‖ ‖ ‖
‖'‖"

≤ ‖ ‖,‖ ‖≤1
‖'‖"

≤ ‖ ‖ …………..(4)

Hence is bounded.
III. The mapping ϕ: B X, Y → B Y ∗ , X ∗ defined by
ϕ =
Is an isometry.
IV. If = then - preserves identity and reverses products.
That is :
ϕ . = . and - =
Here . is the identity mapping defined on .

4.2 Hahn-Banach Theorem (Real and Complex Form) and its

Consequences
The Hahn-Banach theorem is an extension theorem for linear functionals. It guarantees that a
normed space is richly supplied with bounded linear functionals and makes possible an adequate
theory of dual spaces, which is an essential part of the general theory of normed spaces. In this way
the Hahn-Banach theorem becomes one of the most important theorems in connection with
bounded linear operators. Furthermore, our discussion will show that the theorem also
characterizes the extent to which values of a linear functional can be preassigned. The theorem was
discovered by H. Hahn (1927), rediscovered in its present more general form by S. Banach (1929)
and generalized to complex vector spaces by H. F. Bohnenblust and A. Sobczyk (1938).
Now we shall prove this theorem and also discuss some of its important implications.

Before proving the theorem we recall some important definitions .

Lovely Professional University 41

 Notes

Unit 04: Hahn- Banach Theorem and its Consequences

 Let / be a linear space and is the field of real or complex numbers . A functional
:/ → is said to be finite , if is finite for all ∈ /.
 A functional : / → is said to be convex functional (or seminorm) if:
(i) 00∀ ∈ /,
(ii) 2 =2 ∀ ∈ / and real 2 0 0 (positive homogeneous property),
(iii) +3 ≤ + 3 ∀ , 3 ∈ / (sub- additive property).

Example. The norm function ‖. ‖: / → , where / is the normed space, is a convex functional.
A linear functional defined on / is called an extension of a linear functional 4 defined on a
subspace 5 of / if
= 4 ∀ ∈ 5.

Theorem. (The Hahn- Banach Theorem for Real spaces).

Let be a finite convex functional defined on a real vector space / and let 5 be a subspace of /. Let

4: 5 → 6 be a linear functional such that

4 ≤p x ∀ ∈5 ………..(1)
Then 4 can be extended to a linear functional defined on / such that
≤ ∀ ∈/
Proof. For 5 = V the result is trivial, so we suppose that 5 : V .
Step I:
We first prove that 4 can be extended onto a large subspace without violating condition (1).
Let ; ∈ /\5 and put
/ = = + α;: ∈ 5, α ∈ 6>.
Then / is a subspace of / and contains 5 properly.
That is, 5 ⊂ / ⊆ /.
Define a function : / → 6 by ;
+ α; = 4 +α ;
= 4 + αc, c = ; ………..(2)

Then is a linear functional on / .

We show that it is possible to choose a real number B such that the majorization condition
+ α; ≤ + α;
Is satisfied . That is there exists a real number B such that

4 + αB ≤ + α;
' '
i,e 4 CDE +B ≤ C + ;E
D

' '
i,e B≤ C + ;E F 4 CDE ………..(3)
D

if α G 0, and
1 F
4C E + B 0 FH I + α; = F C E
α Fα αF;

42 Lovely Professional University

Notes
Functional Analysis

J' '
B0F C EF 4 CD E ………..(4)
DJK

if α < 0.
Now for any two arbitrary points 3 , 3 of 5, we have

4 3 F 4 3 = 4 3 F3 ≤ 3 F3
≤ 3 +;F 3 +;
≤ 3 + ; + F3 F ;
≤ 3 +; + F3 F ;
Hence
F 4 3 F F3 F ; ≤ 3 +; F 4 3 ………..(5)

Put B = =F 4 3 F F3 F ; >
M∈N

B = OP = 3 + ; F 4 3 >
M∈N

Then B ≤ B
By (5) and the fact that 3 , 3 are arbitrary.
Now choose a B such that
B ≤c≤B
Then, for this value of B, the linear functional defined on / by (2) satisfies the condition that
≤ ∀ ∈ / ………..(6)
as condition (3) and (4) are satisfied . Hence is an extension of 4 to a subspace / containing 5
properly and satisfying condition (1).
Step II.
Now suppose that /, as a linear space is generated by a countable set of elements , , Q , … S , …,

in /. Then we construct a linear functional on / by induction on P. That is, we construct a sequence

of subspaces
/ = = , 5>, / = = , / > , … /S = = S , /S F 1>, …
each contained in the next . This prosess extends the functional 4 onto the whole space /, since
every in / is in some subspace /S .
Step III. For the general case , that is, when no countable set generates /, THE theorem is proved by
applying Zorn’s lemma as follows:
∗
Let be the class of all possible extensions of 4 satisfying the condition
∗ ≤ ∀ ∈T ∗

And
∗ = ∀ ∈T
4 4

Here T 4 is the domain of 4 . Then is non empty because constructed above is in . We

partially order as follows :
For , U ∈ , we say that
≤U
If and only if U is an extension of , that is
T U ⊇T
and U = ∀ ∈T .
Now let W be a chain in . Define a linear functional ̅ as follows:

Lovely Professional University 43

 Notes

Unit 04: Hahn- Banach Theorem and its Consequences

(i) Domain of ̅ =∪Z∈[ T U ,

(ii) For ̅
∈ T\ ],
̅ =U ∀ ∈ T U , U ∈ W.
It is clear that ̅ is a linear extension of 4 and
̅ ≤ ∀ ∈ T\ ]̅ .
So ̅ ∈ and is an upper bound for W. By Zorn’s lemma, has a maximal element which is
an extension of 4 and
≤ ∀ ∈T .
We claim that T = /, otherwise let ; ∈ /\T .
Then as in step I , there is an extension of to T , ;, contracdicting the maximality of .
Hence is the required extension of 4. This proves the theorem completely.
Before discussing the complex version of the Hahn-Banach theorem we need the following
concept.
 A functional defined on a complex linear space / is said to be convex if:
(i) 0 0 ∀ ∈ /,
(ii) =| | ∀ α ∈ C and ∈ /,
(iii) +3 ≤ + 3 ∀ , 3 ∈ /.
Theorem. (Hahn-Banach theorem for complex space)
Let be a finite convex functional defined on a complex linear space / and let 5 be a
subspace of /. Let 4 be alinear functional defined on 5 satisfying the condition:
| 4 |≤ ∀ ∈ 5 ……..(1)
Then 4 can be extended to a linear functional on / such that
| |≤ ∀ ∈/
Proof. Since / is a complex linear space, for each ` ∈ / and α = α + Oα ∈ W, α` ∈ /.
If we restrict the scalers to real numbers only then / is a real vector space . Denote this
space by /a = / and the correspondind subspace by 5a = 5 . Clearly is a finite
convex functional defined on /a while 4 given by:

4 =real part of 4 , ∈ 5a
Is a real linear functional on 5a . Hence, by the Hahn-Banach theorem for real spaces, there is a
linear extension defined on all /a satisfying the condition:
≤ ∀ ∈ /a = / ……..(2)
and = 4 ∀ ∈ 5a = 5
Also, F = F ≤ F = |F1| =
Thus 0F ∀ ∈ /a ……..(3)
From (2) and (3), we obtain
| |≤ ∀ ∈ /a ……..(4)
Now we consider 4 as a linear functional on the complex space 5. So

4 = 4 +O 4 ……..(5)
Since 4 is linear on 5,
O 4 = 4 O = 4 O +O 4 O ……..(6)
Multiplying (5 ) by O, we obtain

44 Lovely Professional University

Notes
Functional Analysis

O 4 =F 4 +O 4 ……..(7)
Comparing (6) and (7), we have

4 = F 4 O
Hence

4 = 4 FO 4 O ……..(8)
If bc denotes the linear extension of 4 to the whole of /, as a real linear space , then
put
= FO O ……..(9)
We show that the function defined by (9) is the required linear extension of 4

To / and satisfies the given condition.

Obviously is an extension of 4 to the whole of /. Also
+3 = + 3 F O \O +3 ]
= + 3 FO O + O3
= + 3 FO O FO O3
= + 3 ……..(10)

And α = \ α + Oα ]
= α +α O
= α + α O
= α FO α O + α O FO Fα
=α F Oα O +α O + Oα
=α FO O + Oα FO O
= α + Oα FO O
=α ∀ , 3 ∈ / and α ∈ W.
Hence is a linear extension of 4.

Finally, we show that

| |≤ ∀ ∈ /.
Suppose, on the contrary, that | 4 |G 4 for some 4 ∈ /.
Then,

4 = ρefg , ρ G 0
If we put
34 = h Jig 4

Then 34 ∈ / and using | 4 | = ρ, we have

34 = 6h 34 = 6h Ch Jig 4 E

= ρ G p x 4 = p y4
Which contradicts (4). Hence
| |≤ ∀ ∈ /.
This completes the proof of the theorem.
Theorem. (The Hahn- Banach Theorem for normed spaces).
Let / be a normed space and 5 be a subspace of /. Let 4 be a bounded linear functional on 5 with
norm ‖ 4 ‖. Then 4 has a continuous linear extension defined on / such that
‖ ‖ = ‖ 4 ‖.

Lovely Professional University 45

 Notes

Unit 04: Hahn- Banach Theorem and its Consequences

Proof. Since 4 is a bounded linear functional , ‖ 4 ‖ is finite . Put

= ‖ 4 ‖‖ ‖ ∀ ∈ /
We show that is a convex functional defined on /.
Clearly 0 0. Also for any α ∈ ,
α = ‖ 4 ‖‖α ‖ = |α|‖ 4 ‖‖ ‖ == |α|p x , ∈ /.
Moreover , for , 3 ∈ /,

+ 3 = ‖ 4 ‖‖ + 3‖
≤ ‖ 4 ‖ ‖ ‖ + ‖3‖
≤ ‖ 4 ‖ ‖ ‖ + ‖ 4 ‖‖3‖
≤ + 3 .
Also | 4 | ≤ ‖ 4 ‖‖ ‖
≤ .
Thus, by the complex version of Hahn-Banach Theorem, there is a linear functional defined on /
such that
| | ≤ p x = ‖ 4 ‖‖ ‖ ∀ ∈/
And = 4 ∀ ∈5 ………..(1)
From (1), we have
‖ ‖ ≤ ‖ 4‖ ………..(2)
|! ' | n!o n
Also ‖ ‖ = ‖'‖
0 ‖'‖
#
= ‖ 4 ‖ ………..(3)
'k4#∈l 'k4#∈m

Hence, from (2) and (3)

‖ ‖ = ‖ 4‖
This proves the proof of the theorem for normed spaces.
Next, we prove an important deduction of the Hahn-Banach theorem for normed spaces and
show that a non-trivial normed space always have enough bounded linear functionals to
distinguish between the points of .
Corrolory. Let be a non-trivial normed space and 4 : 0 be any point of . Then there is a
continuous ( and so bounded ) linear functional defined on such that
‖ ‖ = 1 and ‖ 4 ‖ = ‖ 4‖
Proof. Let 0 : 4 ∈ . Consider the subspace generayted by 4. An arbitrary element of is of
the form 2 4, 2 ∈ . Define a functional 4: → by :

4 3 = 4 2 4 = 2‖ 4 ‖, 3 = 2 4 ∈ ,2 ∈ ………..(1)
Then 4 is linear because for 3 = 2 4 and 3 = 2 4 in and α, α ∈ , we have

4 α3 + α 3 = 4 αa + α 2 4

= αa + α 2 ‖ 4 ‖ by (1)
= αa‖ 4 ‖ + α 2 ‖ 4 ‖
=α 4 3 +α 4 3
!o M |t|‖'o ‖
Also ‖ 4 ‖ = ‖M‖
= = 1, as 3 : 0 so that 2 : 0
Mk4q∈r D∈s |t|‖'o ‖

So u is a bounded linear functional defined on . By the Hahn- Banach theorem for normed spaces
, there is a linear extension of 4 to such that
‖ ‖=‖ 4 ‖ = 1, 3 = 4 3 = 2‖ 4 ‖, 3 = 2 4 ∈

46 Lovely Professional University

Notes
Functional Analysis

Thus ‖ ‖ = 1 and 4 = ‖ 4 ‖ as required.

Corrolory. Every non trivial normed space has a non zero linear functionals defined on it.
Corrolory. Let be a normed space. Then, for any , 3 ∈ , : 3 , there is a bounded linear
functional such that
= 3 .

Summary
 Let / be a linear space and is the field of real or complex numbers . A functional : / →
is said to be finite , if is finite for all ∈ /.
 A functional : / → is said to be convex functional if:
(i) 00∀ ∈ /,
(ii) 2 =2 ∀ ∈ / and real 2 0 0 .
(iii) +3 ≤ + 3 ∀ , 3 ∈ /.
 A linear functional defined on / is called an extension of a linear functional 4 defined on a
subspace 5 of / if
= 4 ∀ ∈ 5.
 (Hahn- Banach theorem for Real spaces)
Let be a finite convex functional defined on a real vector space / and let 5 be a subspace of /.
Let 4: 5 → 6 be a linear functional such that

4 ≤p x ∀ ∈5
Then 4 can be extended to a linear functional defined on / such that
≤ ∀ ∈ /.
 (Hahn- Banach theorem for Complex spaces)
Let be a finite convex functional defined on a complex linear space / and let 5 be a subspace
of /. Let 4 be alinear functional defined on 5 satisfying the condition:
| 4 |≤ ∀ ∈5
Then 4 can be extended to a linear functional on / such that
| |≤ ∀ ∈/

 (Hahn- Banach theorem for normed spaces)

Let / be a normed space and 5 be a subspace of /. Let 4 be a bounded linear functional on 5
with norm ‖ 4 ‖. Then 4 has a continuous linear extension defined on / such that
‖ ‖ = ‖ 4 ‖.

Keywords
 Conjugate
 Bounded linear operator
 Subspace
 Convex functional
 Linear functional
 Seminorm
 Maximality

Lovely Professional University 47

 Notes

Unit 04: Hahn- Banach Theorem and its Consequences

Self Assessment
1: In the context of normed spaces, what is the conjugate of a bounded linear operator?

A. The adjoint operator.

B. The inverse operator.
C. The transpose operator.
D. None of the above.

2: For a bounded linear operator on a normed space , the operator's conjugate, denoted
by , satisfies which property?

A. = ..
B. = .
C. =F .
D. None of the above.

3: Consider two normed spaces and , and let : → be a bounded linear operator. Which
of the following statements is false?

A. If is injective, then is injective.

B. If is surjective, then is surjective.
C. If is compact, then is also compact.
D. If is compact, then is also compact.

4: Every bounded operator of finite rank is :

A. Compact.
B. Open.
C. has a non zero adjoint.
D. None of these.

5: Which of the following is the property of conjugate of the linear operator .

A. is linear.
B. bounded.
C. Both (A) and (B).
D. None of the above.

6: Which of the following is a Banach space?

A. v 2, w with supremum norm.
B. C[a,b] with supremum norm.
C. Both (A) and (B).
D. None of the above.

7: Which of the following is true about Hahn-Banach theorem.

A. The Hahn-Banach theorem is an extension theorem for linear functionals.
B. The Hahn-Banach theorem is an extension theorem for linear functions.

48 Lovely Professional University

Notes
Functional Analysis

C. Both (A) and (B)

D. None of the above

8: Consider the statements:

(i) Every compact operator is bounded.
(ii) Every bounded operator is compact.
A. Only (i) is true.
B. Only (ii) is true.
C. Both (i) and (ii) are true.
D. Neither (i) nor (ii) is true.

9: Which of the following statements is true regarding the Hahn-Banach theorem?

A. It guarantees the existence of a continuous linear functional on every vector space.

B. It ensures the existence of a bounded linear functional on every normed space.
C. It provides a way to extend a bounded linear functional from a subspace to the whole space.
D. It applies only to finite-dimensional vector spaces.

10: Which of the following is true?

A. If x, are invertible linear operators on , then x + is invertible.
B. If x, are invertible linear operators on , then x is invertible.
C. If x is invertible linear operator on , and y is any scalar, then yx is invertible.
D. If x, are invertible linear operators on , then x F is invertible.

∗
11: For any normed space , the dual space is:
A. Always a Banach space.
B. Always a compact set.
C. Always finite dimensional.
D. Always an infinite dimensional.

12: Any bounded subset in 6 S is :

A. Compact.
B. Relatively compact.
C. open.
D. Closed.

13: Let / be a linear space and is the field of real or complex numbers . A functional : / →
is said to be finite

A. If is finite for all ∈ /.

B. If is finite for some ∈ /.
C. Both (A) and (B)
D. None of the above

14: A functional : / → is said to be convex functional if

A. 00∀ ∈ /,
B. 2 =2 ∀ ∈ / and real 2 0 0

Lovely Professional University 49

 Notes

Unit 04: Hahn- Banach Theorem and its Consequences

C. +3 ≤ + 3 ∀ ,3 ∈ /
D. All of the above are true.

15: For what type of normed spaces does the Hahn-Banach theorem always hold?

A. Only for finite-dimensional normed spaces.

B. Only for infinite-dimensional normed spaces.
C. Only for Banach spaces.
D. For all normed spaces.

Answers for Self Assessment

1. A 2. A 3. D 4. A 5. C

6. B 7. A 8. A 9. C 10. B

11. A 12. B 13. A 14. D 15. D

Review Questions
1. Define conjugate of an operator.
2. Define Convex functional.
3. State Hahn-Banach theorem for real spaces.
4. State Hahn- Banach theorem in Complex form.
5. State Hahn- Banach theorem for normed spaces.

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.
 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.
 B.V. Limaya, Functional Analysis.

50 Lovely Professional University

 Notes

Unit 05: Uniform Boundedness Principle and its Consequences

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 05: Uniform Boundedness Principle and its Consequences

CONTENTS
Objectives
Introduction
5.1 Reflexive Spaces
5.2 Category Theorem
5.3 Uniform Boundedness Principle and its Consequences
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Reflexive Spaces
 Baire’s Category theorem
 Uniform Boundedness Principle

Introduction
In this chapter, we discuss about Reflexive spaces . Further , we also recall some definitions like
first category, Second Category and discuss Baire’s Category theorem. Finally, we have proved
Uniform Boundedness Principle.

5.1 Reflexive Spaces

Reflexive spaces play an important role in the general theory of locally convex topological vector
space and in the theory of Banach spaces in particular. Hilbert spaces are prominent examples of
reflexive Banach spaces. Reflexive Banach spaces are often characterized by their geometric
properties. Now, we will define reflexive space.
Definition. Let be a Banach space and : → ∗∗ be the canonical injection from into ∗∗ given
by =ϕ = ∀ ∈ , ∈ ∗.
The space is reflexive if is surjective, i,e = ∗∗ .

Remark. (i) Finite dimensional spaces are reflexive ( since = ∗

= ∗∗
).
(ii) Every Hilbert space is reflexive.
(iii) , , and are not reflexive.
(iv) =space of continuous functions on an infinite compact metric space is not
reflexive.
(v) If a normed space is reflexive, it is complete and hence a Banach space.

Lovely Professional University 51

Notes
Functional Analysis

5.2 Category Theorem

Category Theorem is an important result in general topology and functional analysis. Baire's
Category Theorem was first formulated by French mathematician René-Louis Baire in 1899. This
theorem deals with the properties of complete metric spaces and provides a powerful tool for
studying the nature of dense sets.
Before presenting the main theorem, it is essential to establish several important results. These
results serve as foundational building blocks that will contribute to the proof of the main theorem.
Hence, we now proceed to state these results.
Let be a metric , a subset ⊆ is called rare ( or nowhere dense in ) if has no interior point
i,e = ϕ.
A subset " of a metric space is said to be of the first category ( meager) if and only if " can be
covered by a countable union of its nowhere dense subsets. Otherwise " is said to be of second
category (Non- Meager).
A space is said to be of the first category if and only if as a subset of itself can be written as a
countable union of nowhere dense subsets. Otherwise is said to be of the second category.
Thus a metric space is said to be of second category if and only if cannot be expressed as a
countable union of nowhere dense subsets.

Example. Consider the set # of rationals as a subset of a real line . Let ∈ # , then $%& = $% &
because ' ( $%& = (∞, % ∪ %, ∞ is open. Clearly $% } contain no open ball. Hence # is nowhere
dense in ' as well as in # . Also since # is countable, it is the countable union of subsets
{ %&: % ∈ # .
Thus + is of the first category.
Now we prove the main theorem.
Theorem. ( Baire’s Category Theorem). If a metric space , - is complete then it is non-meager
in itself. Hence if , - is complete and
=∪./ "., ". is closed ……………..(1)
Then atleast one ". contains a nonempty open subset.

(Or)
A complete metric space is of second category.
Proof. Let 0 be a complete metric space. We show that for countable collection $"1 : ∈ 2& of
nowhere dense subsets, is their union . That is, there is a point of which is not in ∪1∈3 "1 .
Suppose, on the contrary that = ∪1∈3 "1 and each "1 is nowhere dense subset.
Since " is nowhere dense and is open with ( " 444 , ϕ, there is an open ball 5 Of radius 6
7
which is disjoint from " . Let 8 be a concentric closed ball of radius half of the radius of 5 . Since
"7 is nowhere dense , 8 contains an open ball 57 of radius 6 and disjoint from "7 . Let 87
9
be the concentric closed ball of radius half that of 57 .
Likewise, since ": is nowhere dense, 87 contains an open ball 5: of radius 6 and disjoint
;
from ": . Again choose a concentric closed ball 8: of radius half of the radius of 5: .
Continuing in this way, we obtain a decreasing sequence of concentric closed balls 81 of
diameter 6 < with each 81 disjoint from "1 . By cantors intersection theorem there is a unique
7 =
point ∈∩1∈3 81 and so also in but not in any of the set "1 . Thus
,∪1∈3 "1
Therefore is of the second category.

5.3 Uniform Boundedness Principle and its Consequences

Now, we prove another important result called the Banach Steinhass theorem which is
commonly known as the uniform boundedness principle. It is concerned with a sequence of point-
wise bounded sequences of linear operators. The uniform Boundedness Principle was obtained in
its general form by S. Banach and Steinhauss in 1927. That is why it is also known as Banach –
Steinhaus theorem. This theorem, like open mapping theorem and closed graph theorem, requires

52 Lovely Professional University

 Notes

Unit 05: Uniform Boundedness Principle and its Consequences

the completeness. Further, this theorem is derived from Baires Category theorem. The Principle of
Uniform Boundedness asserts that if a sequence of bounded linear operators ? ∈ 5 , @ where is
a Banach space and @ a normed space, is point wise bounded, then the sequence $?1 & is uniformly
bounded. In fact, it enables us to determine whether the norms of a given family of bounded linear
operators have a finite least upper bound.
Theorem.(Uniform boundedness Principle) Let be Banach space and @ a normed space. Let ?1
be a sequences of bounded linear opeartors from to @ such that, for each ∈ ,
$?1 ∶ ∈ 2& is a bounded subset of @. Then the sequence{ ‖?1 ‖& of norms of ?1 is also bounded.
Proof. Let C be any natural number and
D. = $ ∈ 2: ‖?1 ‖ ≤ C , = 1,2, … , &
Then D. is closed subset of . For if ∈ D. , then there is a sequence $ 1& in D. which converges to
. So ‖?1 1 ‖ ⇒ ‖?1 1 ‖ = J?1 1 J, as ‖. ‖ is continuous.
1→ 1→

= ‖?1 ‖ ≤ C

Hence ∈ D. , so D. = D.
Also, since each ∈ is in some D. for some natural number C.
⊆∪K/ D. ⊆ X
So that
=∪./ D. ………..(1)
Now is a Banach space so is complete. Hence by Baire’s Category theorem, atleast one of D. ′N, say
D.O , is not nowhere dense in . So D.O contains an open ball P , ϵ , that is
5 P, ϵ ⊆ D.O ………..(2)
Next, let 0 , x be any arbitrary point of . Take a point T
∈ , such that
T
= P + αx………..(3)
W
Where α = . ?ℎZ
7‖X‖
ϵ
‖ T
( P‖ = 6ϵ
2
So that T
∈5 P, ϵ ⊆ D.O . Hence
‖?1 [ ‖ ≤ CP …….(4)
Moreover,
‖?1 P‖ ≤ CP …….(5)
Hence, using (3), we have for all , = 1,2, … and all ∈ ,
?1 T( P 1
‖?1 ‖ = \ \= ‖? T ( ‖
α |α| 1 P

1
≤ ‖?1 T ‖ + ‖?1 P‖
|α|
7
≤ |^| CP
9.O
≤ ‖ ‖
W

Hence,
9.O
‖?1 ‖ = N_` ‖?1 ‖ ≤
∈a‖b‖cd W

Thus,
$‖?1 ‖& is bounded.
Theorem. Let be a normed space and e be a non empty subset of . Then e is bounded if and
only if e is bounded for each bounded linear functional defined on , i,e for each ∈ ∗ .

Lovely Professional University 53

Notes
Functional Analysis

Proof. Suppose that e is bounded subset of normed space . Then, for some positive real number C,
‖ ‖≤C∀ ∈ e………..(1)
Since : → 8 is bounded , there is a positive real number C such that
| |≤k ‖ ‖ ∀ x∈S
≤C C ∀ x∈S
Hence, e is bounded.
Coversely suppose that , for each non-empty subset e of and ∈ X∗, e is bounded . That is
e_`$| |: ∈ e& 6 ∞.
Let ∗∗ be the second dual of and for each ∈ ,h : ∗ → 8 be defined by
h =
Then the mapping ϕ: → ∗∗
defined by:
ϕ =h , ∈
∗∗ .
is the natural embedding of in
To see that h is bounded for each ∈ we note that
|h |=| | ≤ ‖ ‖‖ ‖ ∀ ∈ 2 ∗
‖h ‖ ≤ ‖ ‖
So, for each ∈ there is an P ∈ ∗ such that
‖ P ‖=1 and P = ‖ ‖, (as proved in previous chapter corrolory of Hahn-
Banach theorem)
So ‖h ‖ = N_` |h | ≥ |h P |=| P |=‖ ‖
i∈j∗‖k‖ld

Hence,
‖h ‖ = ‖ ‖
Now , consider the subset
n = $h : ∈ e& of ∗∗ .
Now ∗
is complete so is a Banach space.
Also for each ∈ ∗,

N_`$|h | : h ∈ n& = N_`$| |: ∈ e& 6 ∞, by assumption,

Hence, by the uniform boundedness principle and using (3), we get
N_ `$‖h ‖: h ∈ n& = N_`$‖ ‖: ∈ e& 6 ∞
So e is bounded.

Summary
 Let be a Banach space and : → ∗∗ be the canonical injection from into ∗∗ given by
=- = ∀ ∈ , ∈ ∗.

The space is reflexive if is surjective, i,e = ∗∗ .

 Finite dimensional spaces are reflexive.

 Every Hilbert space is reflexive.
 , , and are not reflexive
 =space of continuous functions on an infinite compact metric space is not
reflexive.
 Let be a metric , a subset ⊆ is called rare ( or nowhere dense in ) if has no
interior point i,e = -.

54 Lovely Professional University

 Notes

Unit 05: Uniform Boundedness Principle and its Consequences

 A subset " of a metric space is said to be of the first category ( meager) if and only if "
can be covered by a countable union of its nowhere dense subsets. Otherwise " is said to
be of second category (Non- Meager).
 A space is said to be of the first category if and only if as a subset of itself can be
written as a countable union of nowhere dense subsets. Otherwise is said to be of the
second category.

 Every complete metric space is of second category.

 Let be Banach space and @ a normed space. Let ?1 be a sequences of bounded linear
opeartors from to @ such that, for each ∈ ,
$?1 ∶ ∈ 2& is a bounded subset of @. Then the sequence{ ‖?1 ‖& of norms of ?1 is also
bounded.

Keywords
 Banach Space
 Normed Space
 Reflexive
 Bounded
 First Category
 Second Category
 Interior point
 Linear functional

Self Assessment
1: Which of the following is true about Reflexive spaces.

A. Finite dimensional spaces are reflexive.

B. Every Hilbert space is reflexive.

C. Both (A) and (B).
D. None of the above.

2: Pick out the correct statement.

A. is not reflexive.
B. is not separable.
C. Both (A) and (B).
D. None of the above.

3: Pick out the correct statement.

A. is not reflexive.
B. is not separable.
C. Both (A) and (B).

D. None of the above.

4: Baire's Category Theorem is applicable to which of the following spaces?

Lovely Professional University 55

Notes
Functional Analysis

A. All metric spaces.

B. Only compact metric spaces.
C. Only finite metric spaces.
D. Only complete metric spaces.

5: Which of the following statement is true.

A. A space is said to be of the first category if and only if as a subset of itself can be
written as a countable union of nowhere dense subsets.

B. Complete metric space is of second category.

C. Both (A) and (B).

D. None of the above.

6: Pick out the correct statement.

A. The set # of rationals is of First Category.

B. The set # of rationals is of Second Category.
C. The set # of rationals is uncountable.
D. All of the above are true.

7: Which of the following statements is not true regarding the Uniform Boundedness Principle?

A. It is also known as the Banach-Steinhaus Theorem.

B. It is applicable only to finite-dimensional normed spaces.
C. It guarantees pointwise convergence of a sequence of bounded linear operators.
D. It is a fundamental result in functional analysis.

8: Which of the following spaces is reflexive?

A. Euclidean space '1 .
B. , .
C. Hilbert space.
D. None of the above.

Answers for Self Assessment

1. C 2. A 3. C 4. D 5. A

6. A 7. B 8. C .

Review Questions
1. Define a reflexive space.
2. Give an example of a Banach space that is not reflexive.
3. What is the concept of a "meager" or "nowhere dense" set in the context of Baire's Category
Theorem?

56 Lovely Professional University

 Notes

Unit 05: Uniform Boundedness Principle and its Consequences

4. State Baire’s Category Theorem.

5. What is the concept of a "meager" or "nowhere dense" set in the context of Baire's Category
Theorem.

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.
 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.
 B.V. Limaya, Functional Analysis.

Lovely Professional University 57

 Notes

Unit 06: Inner Product Space. Hilbert Space

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 06: Inner Product Space. Hilbert Space

CONTENTS
Objectives
Introduction
6.1 Inner Product Space
6.2 Further Properties of Inner Product Space
6.3 Hilbert Space
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Inner Product Space

 Schwarz Inequality
 Parallelogram identity
 Polarization identity
 Continuity of an inner Product
 Hilbert Space.

Introduction
In the preceding chapters, we studied normed and Banach spaces. These spaces enjoy linear
properties as well as metric properties. Although the norm on a linear space generalizes the concept
of length of a vector, but the main geometric concept, missing in abstract normed and Banach
spaces, is the angle between two vectors. In fact these spaces are still too general to yield
a really rich theory of operators. In this chapter, we study linear spaces having an inner product, a
generalization of usual dot product on finite dimensional linear spaces. The concept of an inner
product on a linear space leads to an inner product space and a complete inner product space
(Hilbert Space) is a special type of normed space (Banach space) which possesses an additional
Structure of an inner product.
The theory of Hilbert spaces was initiated in 1912 by a German mathematician, David Hilbert
(1863–1943) in his work on quadratic forms in infinitely many variables, which he applied to the
theory of integral equations. Years later John Von Neumann (1903–1957) first formulated an
axiomatic theory of Hilbert spaces and developed the modern theory of operators. His remarkable
contribution to this area has provided the mathematical foundation of quantum mechanics. His
work provided a physical interpretation of quantum mechanics in terms of abstract relations in an
infinite dimensional Hilbert spaces.

58 Lovely Professional University

Notes
Functional Analysis

6.1 Inner Product Space

⟨. , . ⟩ ∶ X
Definition. Let be a linear space over a field . An inner product in is a function
X→F
So that with each pair , in , a scalar to be denoted by ⟨ , ⟩ is associated satisfying the
conditions

⟨ , ⟩ 0 and ⟨ , ⟩ 0 if and only if 0, ∈ ,

(ii) ⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩ ∀ , y, z ∈ X,
(i)

(iii) ⟨α , ⟩ α⟨ , ⟩ ∀ , y, ∈ X and α ∈
(iv) ⟨ , ⟩ ⟨ , ⟩ ∀ , ,∈ X
Where ⟨ , ⟩ denotes the complex conjugate of ⟨ , ⟩.
The pair X, ⟨. , . ⟩ is called an inner product space.

‖ ‖ !⟨ , ⟩, for all ∈
Remark. An inner product on defines a norm on given by

" , ‖ # ‖ !⟨ # , # ⟩ ∀ x, y, ∈ X.
And a metric on given by

Hence inner product spaces are normed spaces, and Hilbert spaces are Banach spaces.

(a) For all x, y, z ∈ X and α, β ∈ ,

Some Consequences of definition of inner product

⟨α βy, z⟩ α⟨x, z⟩ β⟨y, z⟩

⟨0, z⟩ ⟨0. , z⟩ 0. ⟨ , z⟩ 0∀z ∈ X

Also

(b) For all , y ∈ X and α ∈ ,

⟨ , αy⟩ α ⟨ , y⟩

⟨ , αy⟩ ⟨αy, ⟩ α⟨ y, ⟩
As

&⟨ , ⟩
α ⟨ , y⟩.

(c) ⟨ , αy βz⟩ α⟨ , y⟩ β⟨ , z⟩
Remark. If is a real inner product space , then
⟨ , y⟩ ⟨ y, ⟩ ∀ , y ∈ X.

The space '2 with the inner product of two vectors

)

*, +, … , - *, +, … , -
Example:

,
and defined by
. / ∑n
i 1 1 1
is an inner product space.

space '2 with the inner product of two vectors

*, +, … *, +, …
Example: The

and defined by
∞
. , / 5 1 1
i 1
is an inner product space.

Lovely Professional University 59

 Notes

Unit 06: Inner Product Space. Hilbert Space

Remark. An inner product is also called a pre-Hilbert space.

6.2 Further Properties of Inner Product Space

Schwarz Inequality
If X is an inner product space, then

|⟨ , ⟩| ≤ !⟨ , ⟩. !⟨ , ⟩ ∀ x, y ∈ X………... (1)

the equality holds Iff and y are linearly dependent.

Proof. If y 0, then (1) holds because ⟨ , 0⟩ 0.
Also, if 0, then (1) holds because ⟨0, ⟩ 0.
Now let y ≠ 0. For every scalar α, we have
⟨ #α , #α ⟩ 0
⟹⟨ , ⟩ ⟨ , #α ⟩ ⟨#α , ⟩ ⟨#α , #α ⟩ 0
⟹⟨ , ⟩ # <α⟨ , ⟩ # α⟨ , ⟩ <⟨
αα , ⟩ 0 ………… (2)

⟨ ,=⟩ ⟨ ,=⟩
Choose α ⟨= ,=⟩
, we have <
α ⟨= ,=⟩

⟨ ,=⟩ ⟨ ,=⟩ |⟨ ,=⟩|>

And so αα
< .
⟨= ,=⟩ ⟨= ,=⟩ |⟨= ,=⟩|>

From equation (2), we obtain

, ⟨ , ⟩ # , ,
⟨?, ⟩ ⟨?, ⟩ |⟨ ,=⟩|>
⟨ ⟩- . ⟨ ⟩ ⟨ ⟩ 0
⟨ , ⟩ ⟨ , ⟩ |⟨= ,=⟩|>

, ⟩|2 |⟨ , ⟩|2 , ⟩|2

, ⟩# # ⟨
|⟨ |⟨
, , ,
⟹⟨ 0
⟨ ⟩ ⟩ ⟨ ⟩

, ⟩|2
, ⟩#
|⟨
,
⟹⟨ 0
⟨ ⟩

, ⟩|2
, 〉
|⟨
,
⟹ ≤ .
⟨ ⟩

⟹ |⟨ , ⟩|2 ≤ ⟨ , ⟩⟨ , 〉
⟹ |⟨ , ⟩| ≤ !⟨ , ⟩. !⟨ , ⟩

Next, we see that the equality in (1) holds iff y 0

Which is the required result.

, , ,
From (2), we have
|⟨ ⟩| ≤ !⟨ ⟩. !⟨ ⟩ iff ⟨ #α , #α ⟩ 0

iff #α 0

iff α

iff and y are linearly dependent.

Corollary. Let X be an inner product space, then for any and y in X, we have

!⟨ , ⟩ ≤ !⟨ , ⟩ !⟨ , ⟩
Proof. We can write
⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩ ⟨ , +⟨ , 〉

60 Lovely Professional University

Notes
Functional Analysis

⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩+⟨ , 〉
⟨ , ⟩ 2Re⟨ , ⟩ ⟨ , 〉
≤⟨ , ⟩ 2|⟨ , ⟩| ⟨ , 〉

≤⟨ , ⟩ 2C!⟨ , ⟩. !⟨ , ⟩D ⟨ , 〉
+
E!⟨ , ⟩ !⟨ , ⟩F
+
⟹⟨ , ⟩ ≤ E!⟨ , ⟩ !⟨ , ⟩F

Taking Square root on both sides, we get

!⟨ , ⟩ ≤ !⟨ , ⟩ !⟨ , ⟩.

Theorem. if X is an inner product space, then !⟨ , ⟩ has the properties of norm.

(OR)

Proof. Let X be an inner product space . Define a map ‖ . ‖ ∶ → R by ‖ ‖ !⟨ , ⟩ ∀ ∈ .

In order to show that ‖ ‖ !⟨ , ⟩ defines a norm on the inner product space X, we need to show
that it satisfies all the conditions of a norm.

Now (i) For any x ∈ , ‖ ‖ !⟨ , ⟩ which gives

‖ ‖+ ⟨ , ⟩ 0 (By definition)
⇒‖ ‖+ 0 ⇒‖ ‖ 0

Also ‖ ‖ !⟨ , ⟩ ⇒ ‖ ‖+ ⟨ , ⟩ 0 iff x 0
⇒‖ ‖+ 0 iff x 0 ; i.e. ‖ ‖ 0 iff x 0
(ii) By definition,

‖α ‖ !⟨α , α ⟩
⇒ ‖ α ‖+ ⟨α , α ⟩
<⟨ , 〉
αα
|α|+ ‖ ‖+
⇒‖α ‖ |α|‖ ‖
(iii) For x, y ∈ , we have
‖ ‖+ ⟨ , ⟩
⟨ , ⟩ ⟨ , ⟩ ⟨ , 〉+⟨ , 〉
‖ ‖+ ⟨ , ⟩ ⟨ , ⟩+‖ ‖+
‖ ‖+ 2 H⟨ , ⟩ ‖ ‖+
≤ ‖ ‖+ 2|⟨ , ⟩| ‖ ‖+
≤ ‖ ‖+ 2‖ ‖. ‖ ‖ ‖ ‖+
≤ ‖ ‖ ‖ ‖ +

Hence
‖ ‖ ‖ ‖ ≤‖ ‖ ‖ ‖

We see that all the conditions of a norm are satisfied .Thus ‖ ‖ !⟨ , ⟩ is a norm on X and hence
X, ‖ . ‖ is a norm linear space.
Remark. The Schwarz inequality can now be written in the form
|⟨ , ⟩| ≤ ‖ ‖‖ ‖.

Lovely Professional University 61

 Notes

Unit 06: Inner Product Space. Hilbert Space

Parallelogram Law or Identity for Inner Product Spaces.

If X is an inner product space, then
‖ ‖+ ‖ # ‖+ 2 ‖ ‖+ ‖ ‖+ for x and y in X.
(OR)
In inner product spaces, parallelogram law holds.
Proof. We have
‖ ‖+ ‖ # ‖+ = ⟨ , ⟩ ⟨ # , #
⟨ , ⟩ ⟨ , ⟩ ⟨ , +⟨ , ⟩ ⟨ , ⟩ # ⟨ , ⟩ # ⟨ , +⟨ , 〉
2⟨ , ⟩ 2⟨ , 〉
= 2‖ ‖+ 2‖ ‖+
=2 ‖ ‖+ ‖ ‖+
Hence,
‖ ‖+ ‖ # ‖+ =2 ‖ ‖+ ‖ ‖+ .

Polarization Identity
If X is an inner product space, then for , y in X, we have
J

5 1I ‖ 1 I ‖+ 4⟨ , ⟩.
IKL

Proof. We have
J

5 1I ‖ 1 I ‖+ ‖ ‖+ # ‖ # ‖+ 1‖ 1 ‖+ # 1‖ # 1 ‖+
IKL

⟨ , ⟩#⟨ # , # ⟩ i⟨ 1 , 1 ⟩ # 1⟨ # 1 , # 1 ⟩
=⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩ # N⟨ , ⟩ # ⟨ , ⟩ # ⟨ , ⟩ ⟨ , ⟩O
1N⟨ , ⟩ ⟨ ,1 ⟩ ⟨1 , ⟩ ⟨1 , 1 ⟩O # 1N⟨ , ⟩ # ⟨ , 1 ⟩ # ⟨1 , ⟩ ⟨1 , 1 ⟩O
⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩#⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩#⟨ , ⟩
1⟨ , ⟩ 1⟨ , 1 ⟩ 1⟨1 , ⟩ 1⟨1 , 1 ⟩ # 1⟨ , ⟩ 1⟨ , 1 ⟩ 1⟨1 , ⟩ # 1⟨1 , 1 ⟩
2⟨ , ⟩ 2⟨ , ⟩ 21⟨ , 1 ⟩ 21⟨1 , ⟩
2⟨ , ⟩ 2⟨ , ⟩ 21P̅⟨ , ⟩ # 21P̅⟨ , ⟩
2⟨ , ⟩ 2⟨ , ⟩ 2⟨ , ⟩ # 2⟨ , ⟩
4⟨ , ⟩
Hence
J

5 1I ‖ 1 I ‖+ 4⟨ , ⟩.
IKL

Remark. If is a real inner product space, then the polarization identity becomes:
4⟨ , ⟩ ‖ ‖+ # ‖ # ‖+ for , ∈ .
The Schwarz inequality is quite important and will be used in proofs over and over again .
Another frequently used property is the continuity of inner product.

Continuity of Inner Product Space

Theorem. Let be any inner product space and { -} , { - } be any sequences in
- → , - → , then 〈 - , - 〉→ 〈 , 〉.
such that

62 Lovely Professional University

Notes
Functional Analysis

Proof. For any natural number ), we have from definition of inner product spaces
|〈 - , - 〉# 〈 , 〉| |〈 - , - 〉# 〈 -, 〉 〈 -, 〉# 〈 , 〉|
|〈 -, - # 〉 〈 - # , 〉|
≤ |〈 -, - # 〉| |〈 - # , 〉|
≤‖ -‖ ‖ - # ‖ ‖ - # ‖ ‖ ‖
(by Cauchy Schwarz Inequality)
Thus, if - → , - → then { -} is bounded and

‖ - # ‖ → 0, ‖ - # ‖ → 0 as ) → ∞
So that
|〈 - , - 〉# 〈 , 〉| → 0 as ) → ∞
Hence
〈 - , - 〉→ 〈 , 〉.
Theorem. . Let be any inner product space and if { - } , { - } are Cauchy sequences in , then
〈 - , - 〉is a convergent sequence in , where .
Proof. Suppose that { -} , { - } are Cauchy sequences in . Then for all natural numbers U, ) we
have
‖ - # V‖ →0,‖ - # V‖ → 0 as U, ) → ∞

|〈 - , - 〉# 〈 V , V 〉| - , - 〉# 〈 -, V〉 〈 - , V 〉# 〈 V , V 〉|
Hence , as above ,
|〈
|〈 -, - # V〉 〈 - # V , V 〉|

≤ |〈 -, - # V 〉| |〈 - # V , V 〉|

≤‖ -‖ ‖ - # ‖ ‖ - # ‖ ‖ ‖
(by Cauchy Schwarz Inequality)

U, ) → ∞. Hence
Since every Cauchy sequence is bounded, the right hand side of above equation tends to 0 as

〈 - , - 〉is a Cauchy sequence in . Since is , this sequence converges in .

6.3 Hilbert Space

Definition.
A complete inner product spac is called Hilbert space. Or an inner product space in which every
Cauchy sequence converges is said to be Hilbert Space.

⟨ , ⟩ ‖ ‖+ ∑-WK*| W |+ .
Example: Show that the Euclidean space is a Hilbert space with inner product defined by

Solution:- Let { { }X
WK* , then for any Y Z 0 there
- -
-} be a Cauchy sequence in where - W

exists )L ∈ [ such that

‖ - # V‖ !⟨ - # V, - # V⟩ \ ϵ ; ∀m, n nL

∞
⇒ `5 | 1 # 1 | \ ϵ ; ∀m, n
) U 2
nL
i 1

Lovely Professional University 63

 Notes

Unit 06: Inner Product Space. Hilbert Space

⇒| 1 # | \ ϵ ; ∀m, n
) U
1 nL

⇒ { 1 } is a Cauchy sequence in
)
1
)
→ xi ∈ R, then
there exists a natural number )1 ∈ [ such that | W # W| \
-
and since is complete therefore
ϵ
;∀n ni
√b

⇒| 1 # 1| \
) ϵ
;∀ n n*
!b

| -
# +| \
ϵ
+ ;∀n n2
√b
⋮ ⋮ ⋮
ϵ
| # -| \
-
; ∀n nn
√b
-

* , +, … , - ∈
If -.
then
Let )d Ue )* , )+ , … , nf then for the above expression we have

-
+
‖ -# ‖ `5 g W
-
# Wg
WK*

- - -
+ + +
⇒‖ -# ‖ `5 g # *g 5g # +g ⋯5g # -g
- - -
* + -
WK* WK* WK*

⇒‖ # ‖\i ⋯ i ;) )d
j> j> j> -j >
- - - - -

⇒‖ - # ‖ \ ϵ ;) ) d ⇒ !⟨ - # V, - # V⟩ \ ϵ ;) )d
-. -
This shows that - converges in Hence is a Hilbert space.
-
Similarly we can show that is a Hilbert space with complex sequence.

The space '+ of all complex sequences { W } such that ∑X

WK*| W | \ ∞ is an
inner product space under inner product defined by 〈 , 〉 ∑1 1 1 <k ;
) { W } ∈ '+ .
Example:

We also know that '+ is complete , hence '+ is a Hilbert space.

Example: Every finite dimensional inner product space is a Hilbert Space.

Because every finite dimensional inner product space is a finite dimensional normed linear space
and we know that every finite dimensional normed linear space is complete.

Example: Space ' l with b ≠ 2 is not an inner product space , hence not a Hilbert space.

Example: The space Ne, mO is not an inner product space , hence not not a Hilbert space.

Summary
Let be a linear space over a field . An inner product in is a function

64 Lovely Professional University

Notes
Functional Analysis

⟨. , . ⟩ ∶ X X→F
So that with each pair , in , a scalar to be denoted by ⟨ , ⟩ is associated satisfying the
conditions

⟨ , ⟩ 0 and ⟨ , ⟩ 0 if and only if 0, ∈ ,

(ii) ⟨ , ⟩ ⟨ , ⟩ ⟨ , ⟩ ∀ x, y, z ∈ X,
(i)

(iii) ⟨α , ⟩ α⟨ , ⟩ ∀ x, y, ∈ X and α ∈
(iv) ⟨ , ⟩ ⟨ , ⟩ ∀ , ,∈ X .
The space '2 with the inner product of two vectors
* , + , … and * , + , … defined by


∞
. , / 5 1 1
i 1

is an inner product space.

If X is an inner product space, then

 An inner product is also called a pre-Hilbert space.

|⟨ , ⟩| ≤ !⟨ , ⟩. !⟨ , ⟩ ∀ x , y ∈ X.


If X is an inner product space, then

(Cauchy Schwarx Inequality)

‖ ‖+ ‖ # ‖+ 2 ‖ ‖+ ‖ ‖+ for and y in X.


if X is an inner product space, then !⟨ , ⟩ has the properties of norm.

(Parallelogram law)


|⟨ , ⟩| ≤ ‖ ‖‖ ‖.
 The Schwarz inequality can also be be written in the form

If X is an inner product space, then for , y in X, we have

∑JIKL 1 I ‖ 1I ‖+ 4⟨ , ⟩

(Polarization Identity)

4⟨ , ⟩ ‖ ‖+ # ‖ # ‖+ for , ∈ .
 If is a real inner product space, then the polarization identity becomes:

be any inner product space and { -} , { - } be any sequences in

→ , → , then 〈 - , - 〉→ 〈 , 〉.
 Let such that
- -

be any inner product space and if { -} , { - } are Cauchy sequences in , then

〈 - , - 〉is a convergent sequence in , where
 Let
.
 A complete inner product spac is called Hilbert space. Or an inner product space in
which every Cauchy sequence converges is said to be Hilbert Space.
-
 Euclidean space is a Hilbert space with inner product defined by

⟨ , ⟩ ‖ ‖+ ∑-WK*| W |+ .

Keywords
 Inner product space
 Norm
 Cauchy-Schwarz inequality
 Polarization identity
 Parallelogram law
 Continuity
 Cauchy sequence
 Hilbert space

Lovely Professional University 65

 Notes

Unit 06: Inner Product Space. Hilbert Space

Self Assessment
1: Pick the INCORRECT statement:
A. Every Hilbert space is a normed space.
B. Every Banach space is a Hilbert space.
C. Every Banach space is a topological space.
D. Every normed space is a metric space.

2: Which of the following is Cauchy-Schwartz inequality.

|⟨ , ⟩| ≤ ⟨ , ⟩> . ⟨ , ⟩>
n n
A.

|⟨ , ⟩| ≤ ⟨ , ⟩> ⟨ , ⟩>
n n
B.

|⟨ , ⟩| ⟨ , ⟩> . ⟨ , ⟩>
n n

|⟨ , ⟩| ≤ ⟨ , ⟩. ⟨ , ⟩ .
C.
D.

3: Which of the following is known as Parallelogram law?

o o
2
2‖ ‖2 2o o
2

# 2‖ 2o
2 2 2
A.

o o o o ‖2 o

# 2
2 2 2
B.

o o o o ‖ ‖2 o o
2
#o # 2
2‖ 2
C.

D. o o o ‖2 o o

4: If X is an inner product space, then for x , y in X, we have

∑JIKL 1 I ‖ 1 I ‖+
5 ,
4 ,
A.

3 ,
B.
C.
D. None of these.

5: An inner product is also called a:

A. Pre-Hilbert space
B. Hilbert space
C. Complete normed space

D. None of these.

6: The term Hilbert space stands for a :

A. Compact linear space
B. Complete normed space
C. Complete metric space
D. Complete inner product space.

66 Lovely Professional University

Notes
Functional Analysis

7 : In a complex inner product space , the conjugate symmetry property of the inner product is
given as .

⟨ , ⟩ ⟨ , ⟩
⟨ , ⟩ #⟨ , ⟩
A.

⟨ , ⟩ ⟨ , ⟩
B.
C.
D. None of these.

8: Let V be a real inner product space. Which of the following statements is true?

A. The inner product is always positive definite.

B. The inner product is always symmetric.
C. The inner product is always commutative.
D. The inner product is always associative.

9: Let V be a complex inner product space. Which of the following properties does the inner
product satisfy?
A. Conjugate symmetry.
B. Distributive property.
C. Anticommutativity.
D. None of these

∈ ?
10: Let be a finite-dimensional inner product space. Which of the following statements is always
true for any nonzero vector
A. The norm of is always equal to 1
B. The norm of is always greater than or equal to zero.
C. The norm of is always less than or equal to zero.
D. The norm of is always positive.

11: Let r be a Hilbert space over and , ∈ r, be such that ‖ ‖ 4, ‖ ‖ 3 and ‖ # ‖ 3.

Then ⟨ , ⟩ equals:
A. 6
B. 8
C. 10
D. 14

Answers for Self Assessment

1. B 2. A 3. C 4. B 5. A

6. D 7. C 8. B 9. A 10. D

11. B

Review Questions
1. What is the definition of an inner product space?

Lovely Professional University 67

 Notes

Unit 06: Inner Product Space. Hilbert Space

2. Prove that the norm induced by an inner product satisfies the parallelogram law.
3. Give an example of a real inner product space.
4. Define a Hilbert space. How does it differ from a general inner product space?
5. State Cauchy Schwarz inequality for inner product space
6. State Parrallelogram identity for inner product space.

Further Readings
1. Introductory Functional Analysis With Applications By Erwin Kreyszig.
2. Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
3. J. B Conway, A Course In Functional Analysis.
4. C. Goffman G Pedrick, A First Course In Functional Analysis.
5. B.V. Limaya, Functional Analysis.

68 Lovely Professional University

 Notes

Unit 07: Orthogonality of Inner Product Space

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 07: Orthogonality of Inner Product Space

CONTENTS
Objectives
Introduction
7.1 Orthogonality Of Vectors
7.2 Orthonormal Sets
7.3 Complete Orthonormal Sets
7.4 Pythagorean Theorem
7.5 Bessel’s Inequality
7.6 Riesz- Fischer Theorem
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Orthogonality of vectors
 Orthonormal sets
 Complete orthonormal set
 Pythagorean theorem
 Bessel’s inequality
 Riesz- Fischer theorem

Introduction
In this chapter , we introduce the idea of Orthogonality of inner product spaces and establish the
basic terminology. We also discuss complete orthonormal sets, Pythagorean theorem, Bessel’s
inequality, Parseval’s identity and Riesz- Fischer theorem. This chapter enables the students to
carefully use the concepts of Orthogonality.

7.1 Orthogonality Of Vectors

Recall that the dot product of two vectors in the space is zero, the vectors are orthogonal or at
least one of the vectors is the zero vector. We generalize this concept in an inner product space.
Definition. Let X be an inner product space. A vector x ∈ X is said to be orthogonal to a vector
y ∈ X if ⟨x, y ⟩ = 0.
Such vectors x and y are called orthogonal vectors, written x ⊥ y (the symbol ⊥ is

Lovely Professional University 69

Notes
Functional Analysis

pronounced as “per”). Similarly, for subsets A, B ⊂ X , we write x ⊥ A if x ⊥ a ∀ a ∈ A and

A ⊥ B if a ⊥ b, ∀ a ∈ A and b ∈ B.
Observations.

(i) x ⊥y ⟺ y ⊥x
(ii) x ⊥ 0, ∀x ∈ X
(iii) 0 is the only vector in X orthogonal to itself.
(iv) For a subset A of an inner product space X, define the set
A = x ∈ X: x ⊥ A
We write A =A , A =A and so on .
(v) 0 = X and X = 0 i,e 0 is the only vector orthogonal to every vector .
Proof. We have
0 = x ∈ X: ⟨x, 0 ⟩ = 0 = X
Since ⟨x, 0⟩ = 0, ∀ x ∈ X. Also if x 0, then ⟨x, x⟩ 0. In other words , a non zero vector can not be
orthogonal to the entore space X. Hence X = 0 .

(vi) If A ϕ is subset of X, then the set A is closed subspace of X. Furthermore,

A ∩ A is either 0 or empty (when 0 ∉ A).
(vii) If A and B are subsets of X such that A ⊂ B, then A ⊃ B .
Proof. Let x ∈ B then ⟨x, y⟩ = 0, ∀ y ∈ B and in particular ∀ x ∈ A since A ⊂ B.This verifies that
x ∈ A . Hence A ⊃ B .

(viii) If A is a subset of X, then A ⊂ A

Proof. Let x ∈ A. Then x ⊥ A , which means x ∈ A
, thus A ⊂ A .

(ix) If A ϕ is a subset of X, then A = A .

Example: R$ is an inner product space with inner product defined by

n
%x, y& = ' xi yi
i=1
Then the vectors 1,0,0, … ,0 , 0,1,0, … ,0 , … 0, 0,0, … ,1 are orthogonal, as the inner product of any
two of the above vectors is zero.

Example: l, is an inner product space with the inner product defined by

0
⟨x, y⟩ = ' x- y/.
-12

Then the vectors e2 = 1,0,0, … ,0 , e, = 0,1,0, … ,0 , e = 0,0,1, … ,0 … e- = 0,0,0, … ,1,0,0, … , … in

l,
are orthogonal because e- ⊥ e4 ∀i, j with i j.

7.2 Orthonormal Sets

Definition. A set 6 = 78 ∶ : ∈ ; in an inner product space < is said to be orthonormal if
0, :A : B
〈78 , 7> 〉= @
1, :A : = B
i,e 〈78 , 7> 〉= C8> , the standard Kronecker delta. In other words the set 6 is said to be orthonormal if
it is orthogonal and ‖7‖ = 1 for every 7 ∈ 6.

70 Lovely Professional University

 Notes

Unit 07: Orthogonality of Inner Product Space

Example: Let 78 ∶ : ∈ ; be an orthogonal set in an inner product space <, then the set
G
E = F‖GH ‖ ∶ : ∈ ;I is orthonormal.
H

GH GJ
Solution. Let ‖GH ‖
, ∈ E, then
KGJ K

78 7> 1
〈 , 〉 = 〈78 , 7> 〉
‖78 ‖ K7> K ‖78 ‖ K7> K
2
= ‖G ‖ L 0=0
H KGJ K

Thus the inner product of two different elements of E is zero. So that E is orthogonal .
Next, we show that norm of every element of E is 1.
G
For this let ‖GH ‖ ∈ E, then
H

78 ‖78 ‖
M M= =1
‖78 ‖ ‖78 ‖
This shows that A is orthonormal.

7.3 Complete Orthonormal Sets

Definition. An orthonormal set S in an inner product space X is said to be complete if there exists
no orthonormal set in X of which S is a proper subset.
In other words, S is complete if it is maximal with respect to the property of being normal.

Note: - If S is complete orthonormal set, then there does not exist any non-zero vector such
that x ⊥ S and ‖7‖ = 1.

Example. In the space l, , the orthonormal set composed of e2 = 1,0,0, … , , e, = 0,1,0, … , ,

e = 0,0,1, … , …
Is a complete orthonormal set.
Orthonormal sets in Hilbert Spaces: An orthonormal set in a Hilbert space H is a non-empty
subset of H which consists of mutually orthogonal unit vector : that is , it is non empty subset e- of
H with the following property.

(i) 〈ei , ej 〉= 0, if i j
(ii) 〈ei , ej 〉= 1, if i = j

Note: - See examples following the definition of orthonormal sets in Inner product spaces.
Remark. If H = 0 i.e. H contains only the zero element, then it has no orthonormal set. If
P
H contains a non-zero vector x , then we can construct e by normalizing x, that is e = ‖P‖, then the
P , ‖P‖R
single element set e is clearly an orthonormal set because 〈e, e 〉= ‖e‖, = Q‖P‖Q = ‖P‖R
=1

Generally speaking if x- is a non empty set of mutually orthogonal non zero vectors in H, and if
P
the x- ′s are normalized by replacing each of them by e- = ‖PU ‖, then the resulting set e- is
U
orthonormal s

Remark. One of the simple geometric fact about orthogonal vectors is the Pythagorean theorem,
which is given as follows.

Lovely Professional University 71

Notes
Functional Analysis

7.4 Pythagorean Theorem

Theorem. If x and y are orthogonal vectors in an inner product space X, then

2 2 2
Kx + yK = ‖x ‖2 + KyK = Kx − yK
Proof. We have

Kx + yK2 = ⟨x + y, x + y ⟩
= ⟨x, x⟩ + ⟨x, y⟩ + ⟨y, x⟩ +
⟨y, y⟩
= ⟨x, x⟩ + 0 + 0 + ⟨y, y⟩
= ⟨x, x⟩ + ⟨y, y⟩
= ‖x ‖, + ‖y‖,
Similarly, we can show that

Kx − yK2 = ‖x ‖2 + KyK2
Hence,
Kx + yK2 = ‖x ‖2 + KyK2 = Kx − yK2 .

7.5 Bessel’s Inequality

Theorem. Let 6 = Y2 , Y, , … , YZ be an orthonormal set in a Hilbert space [. If 7 is any vector in [,
then
∑Z812|〈7, Y8 〉|, ≤ ‖7‖,
(Bessel’s inequality)………..(1)

And 7 − ∑Z812〈7, Y8 〉Y8 ⊥ Y> for each j

i.e. 7 − ∑Z812〈7, Y8 〉Y8 ⊥ 6 .

Proof. We have : 0 ≤ ‖7 − ∑Z812〈7, Y8 〉Y8 ‖,
= 〈7 − ∑Z812〈7, Y8 〉Y8 , 7 − ∑Z>12〈7, Y> 〉Y> 〉
= 〈7 , 7 〉− 〈7 , ∑Z>12〈7, Y> 〉Y> 〉− 〈∑Z812〈7, Y8 〉Y8 , 7 〉+ 〈∑Z812〈7, Y8 〉Y8 , ∑Z>12〈7, Y> 〉Y> 〉
ddddddd
= 〈7 , 7 〉− ∑Z>12〈7, Yc 〉〈7, Y> 〉− ∑Z812〈7, Y8 〉〈Y8 , 7〉+ ∑Z812 ∑Z>12〈7 , Y8 〉ddddddd
〈7, Yc 〉〈Y8 , Y> 〉
, , ddddddddd
= 〈7 , 7 〉− ∑Z>12e〈7, Y> 〉e − ∑Z812|〈7, Y8 〉| + ∑Z812 ∑Z>12〈7, Y8 〉〈7 , Yc 〉 〈Y8 , Y> 〉
,
= 〈7 , 7 〉− ∑Z>12e〈7, Y> 〉e − ∑Z812|〈7, Y8 〉| + ∑Z812〈7, Y8 〉ddddddd
,
〈7, Yf 〉〈Y8 , Y8 〉
, , ,
= 〈7 , 7 〉− ∑Z>12e〈7, Y> 〉e − ∑Z812|〈7, Y8 〉| + ∑Z812|〈7, Y8 〉|
,
⇒ 0 ≤ 〈7 , 7 〉− ∑Z>12e〈7, Y> 〉e
,
⇒ ∑Z>12e〈7, Y> 〉e ≤ ‖7‖, , which is equivalent to (1)
In order to show that 7 − ∑Z812〈7, Y8 〉Y8 ⊥ 6, consider any Y> :h 6 where B = 1,2,3, … , h
Then 〈7 − ∑Z812〈7, Y8 〉Y8 , Y> 〉= 〈7, Y> 〉− 〈∑Z812〈7, Y8 〉Y8 , Y> 〉
= 〈7, Y> 〉− ∑Z812〈7, Y8 〉〈Y8 , Y> 〉

72 Lovely Professional University

 Notes

Unit 07: Orthogonality of Inner Product Space

= 〈7, Y> 〉− 〈7, Y> 〉〈Y> , Y> 〉

= 〈7, Y> 〉− 〈7, Y> 〉. 1
=0
This shows that 7 − ∑Z812〈7, Y8 〉Y8 ⊥ Y> for each B

⇒ 7 − ∑h:=1〈7, Y: 〉Y: ⊥6
This completes the proof.
Theorem. If Y8 is an orthonormal set in a Hilbert space [ and if 7 is any vector in [, then the set

S = e- : 〈x, e- 〉 0 is either empty or countable.

Theorem. (Generalization of Bessel’s inequality).

If Y8 is an orthonormal set in a Hilbert space [, then

∑|〈7, Y8 〉|j ≤ ‖7‖, ………(1)

for every vector x ∈ H .
Proof. Let us define a set S as
S = e- : 〈x, e- 〉 0
Then by the above theorem , S is either empty or countable .
If S is empty, then 〈x, e- 〉= 0, so ∑|〈7, Y8 〉|j is zero and so in this case (1) reduces to 0 ≤ ‖7‖, ,
which is obviously true.
If S is countable , then S is finite or countably infinite .
When k is finite . Let it can be written in the form S = e2 , e, , … , e$ for some positive integer n. In
this case, we denote ∑|〈7, Y8 〉|j to be ∑Z812|〈7, Y8 〉|, , which is clearly independent of the order in
which the vectors of 6 are arranged. So inequality (1) reduces to ∑Z812|〈7, Y8 〉|, ≤ ‖7‖, which is the
Bessel’s inequality when Y8 is finite orthonormal set as proved already.
When k is countably infinite . Let the vectors in S be arranged in some definite order i.e. 6 =
Y2 , Y, , … , YZ , … , as by the theory of “absolutely convergent series” we know that if ∑0
812|〈7, Y8 〉|
,

converges , then every series obtained from this series by rearranging its terms also converges and
all such series have the same sum. So we therefore can define ∑|〈7, Y8 〉|j to be ∑Z812|〈7, Y8 〉|, and it
follows that ∑|〈7, Y8 〉|j is a non-negative extended real number , which depends only on S and not
on the arrangement of vectors in S. So in this case (1) reduces to
∑0
812|〈7, Y8 〉| ≤ ‖7‖ ……...(2)
, ,

Now from Bessel’s inequality for finite case, we have:

Z

'|〈7, Y8 〉|, ≤ ‖7‖,

812

It follows that no partial sum of the series on left side of (2) can exceed ‖7‖, and so it is clear that (2)
is true .
$

⇒ lim '|〈7, Y: 〉|2 ≤ ‖7‖2

$→0
-12
∞
⇒ '|〈7, Y: 〉|2 ≤ ‖7‖2
:=1

This completes the proof.

Theorem. If Y8 is an orthonormal set in a Hilbert space [ and if 7 is any vector in [, then the set
S = e- : 〈x, e- 〉 0 is either empty or countable.
Theorem. Let e- be an orthonormal set in a Hilbert space H and let x be a vector in H, then

Lovely Professional University 73

Notes
Functional Analysis

x − '〈x, e- 〉e- ⊥ e-

Theorem. (orthonormal bases) Let [ be Hilbert space and let e- be an orthonormal set in H, then
the following are equivalent

(i) e- is complete.
(ii) x ⊥ e- ⇒ x = 0
(iii) If x is an arbitrary vector in H, then x = ∑〈7, Y8 〉Y8
(iv) If x is an arbitrary vector in H, then ‖7‖, = ∑|〈7, Y8 〉|j . (Parseval’s identity).
Proof. i ⇒ ii
Suppose i is true i.e. e- is complete .
⇒ e- is maximal orthonormal set. On contrary suppose that ii is not true, then there exists a
vector x 0 such that x ⊥ e- .
P
Define e = ‖P‖ , then the set e- , e is an orthonormal set, which properly contains e- , but this
contradicts the completeness of e- . Hence ii is true.
ii ⇒ iii

Suppose that ii is true i.e. x ⊥ e- ⇒ x = 0. Now by above theorem, we have

x − ∑〈x, e- 〉e- is orthogonal to e- .
i, e.
x − '〈x, e- 〉e- ⊥ e-

So by ii , we get

x − '〈x, e- 〉e- = 0

Or
x = '〈x, e- 〉e-

for any vector x ∈ H. Hence iii is true.

iii ⇒ iv
Suppose that iii is true i.e. x = ∑〈7, Y8 〉Y8 for any vector x ∈ H.
Now x = ∑〈7, Y8 〉Y8 = ∑0
-12〈7, Y8 〉Y8

Then ‖x‖, = 〈x, x 〉= 〈x, ∑0

-12〈7, Y8 〉Y8 〉

= 〈x, lim$→0 ∑$-12〈7, Y8 〉Y8 〉

ddddddd
= lim$→0 ∑$-12〈7, Yf 〉 〈7, Y8 〉

⇒ ‖x‖, = lim$→0 ∑$-12 |〈7, Y8 〉|,

= ∑0
-12|〈7, Y8 〉|
,

Using ∑〈7, Y8 〉Y8 in place of ∑0

-12〈7, Y8 〉Y8 , we get

‖7‖, = '|〈7, Y8 〉|j .

Hence iv is true.
Finally iv ⇒ i
Suppose that iv is true i.e. ‖7‖, = ‖7‖, = ∑|〈7, Y8 〉|j .
We show that i is true . On the contrary assume that i is not true i.e. e- is not complete , then it
is properly contained in an orthonormal set e- , e . So by definition of orthonormal set, we can say
that e is orthogonal to ep- s.
,
Now ‖e‖, = ∑|〈e, e- 〉| by iv

74 Lovely Professional University

 Notes

Unit 07: Orthogonality of Inner Product Space

,
= ∑‖0‖
= ‖0‖
=0
i. e ‖e‖ = 0 and this contradicts the fact that ‖e‖ = 1.
So our supposition was wrong and hence e- is complete .
Hence i is true.
This completes the proof.
Remark. Let e- be a complete orthonormal set and let x be an arbitrary vector in a Hilbert space
H. Then the numbers 〈x, e- 〉are called the Fourier coefficients of x, the expression 〈x, e- 〉e- is called
the Fourier expansion of x and the equation ‖7‖, = ∑|〈7, Y8 〉|j is called Parseval’s equation .

7.6 Riesz- Fischer Theorem

Theorem. Let Y2 , Y, , … , YZ , … be an orthonormal set in a Hilbert space [. Then, for any sequence
qr of scalars, the following are equivalent.

(i) sr ∈ t ,
(ii) ∑0
r12 sr Yr converges in [
(iii) there is an element 7 ∈ [
〈7, Yr 〉= sr , u = 1,2, …

Proof. Suppose that : is true so that sr ∈ t , . Then

0

'|sr |, < ∞
r12

For h = 1,2, …, let

Z

wZ = ' sr Yr
r12

We first show that wZ is a Cauchy sequence in [. For this consider the expression ‖wx − wZ ‖, y >
h,
We have ‖wx − wZ ‖, = 〈wx − wZ , wx − wZ 〉
= 〈∑x
r1Z{2 sr Yr , ∑|1Z{2 s| Y| 〉
x

,
= ∑x
r1Z{2 |sr | ,
,
Using the orthogonality of YZ{2 , YZ{, , … , Yx . Since the series ∑0
r12 |sr | converges in }, by Cauchy’s
criterion of convergence ,
x

' |sr |, → 0 ~w y, h → ∞
r1Z{2

So wZ is a Cauchy sequence in [. As [ is complete , wZ → 7 ∈ [

Thus
∑0
r12 sr Yr
converges in [. So : ⇒ :: .
Next suppose that :: is satisfied so that the series ∑0
812 s8 Y8 converges to 7 ∈ [.
Then
7 = ∑0
812 s8 Y8
So that , for u = 1,2, … 0
〈7, Yr 〉= 〈' s8 Y8 , Yr 〉
812

Lovely Professional University 75

Notes
Functional Analysis

= ∑0
812 s8 〈Y8 , Yr 〉,
= sr
Hence ::: is satisfied .
Lastly suppose that ::: holds. Then , by Bessels inequality,
0 0

'|sr |, = '|〈7, Yr 〉|, ≤ ‖7 ‖ < ∞

r12 r12
,
So that sr ∈ t . Hence ::: ⇒ : .

Summary
 Let X be an inner product space. A vector x ∈ X is said to be orthogonal to a vector y ∈ X if
⟨x, y ⟩ = 0.
Such vectors x and y are called orthogonal vectors, written x ⊥ y (the symbol ⊥ is pronounced as
“per”). Similarly, for subsets A, B ⊂ X , we write x ⊥ A if x ⊥ a ∀ a ∈ A and A ⊥ B if
a ⊥ b, ∀ a ∈ A and b ∈ B.

 x ⊥ y ⟺ y ⊥ x.
 x ⊥ 0, ∀ x ∈ X.
 0 is the only vector in X orthogonal to itself.
 For a subset A of an inner product space X, define the set
A = x ∈ X: x ⊥ A

We write A =A , A =A and so on .
 0 = X and X = 0 i,e 0 is the only vector orthogonal to every vector .
 If A ϕ is subset of X, then the set A is closed subspace of X. Furthermore, A ∩ A is either 0
or empty (when 0 ∉ A).
 If A and B are subsets of X such that A ⊂ B, then A ⊃ B .
 If A is a subset of X, then A ⊂ A .
 If A ϕ is a subset of X, then A = A .
 R $
$
is an inner product space with inner product defined by
⟨x, y⟩ = ' x- y-
 Then the vectors 1,0,0, … ,0 , 0,1,0, … ,0 , … 0, 0,0, -12
… ,1 are orthogonal, as the inner product of
any two of the above vectors is zero.

 A set 6 = 78 ∶ : ∈ ; in an inner product space < is said to be orthonormal if

0, :A : B
〈78 , 7> 〉= @
1, :A : = B

i,e 〈78 , 7> 〉= C8> , the standard Kronecker delta. In other words the set 6 is said to be
orthonormal if it is orthogonal and ‖7‖ = 1 for every 7 ∈ 6.
 An orthonormal set S in an inner product space X is said to be complete if there exists no
orthonormal set in X of which S is a proper subset.
 In the space l, , the orthonormal set composed of
e2 = 1,0,0, … , , e, = 0,1,0, … , , e = 0,0,1, … , … is a complete orthonormal set.
 If x and y are orthogonal vectors in an inner product space X, then

76 Lovely Professional University

 Notes

Unit 07: Orthogonality of Inner Product Space

‖x + y‖, = ‖x ‖, + ‖y‖, = ‖x − y‖,

(Pythagorean theorem)
 Let 6 = Y2 , Y, , … , YZ be an orthonormal set in a Hilbert space [. If 7 is any vector in [,
then

∑Z812|〈7, Y8 〉|, ≤ ‖7‖,

(Bessel’s inequality)
 (Generalization of Bessel’s inequality).
If Y8 is an orthonormal set in a Hilbert space [, then

'|〈7, Y8 〉|j ≤ ‖7‖,

 If Y8 is an orthonormal set in a Hilbert space [ and if 7 is any vector in [, then the set
S = e- : 〈x, e- 〉 0 is either empty or countable.
 Let e- be an orthonormal set in a Hilbert space H and let x be a vector in H, then
x − '〈x, e- 〉e- ⊥ e-

Keywords
 Orthogonality
 Inner product space
 Complete orthonormal set
 Hilbert space
 Pythagorean theorem
 Bessel’s inequality
 Orthonormal bases
 Parseval’s identity
 Riesz- Fischer theorem

Self Assessment
1: Two Vectors 7, • in an inner product space are orthogonal if :
A. ⟨7, •⟩ 0
B. ‖7‖ = ‖•‖ = 1
C. ⟨7, •⟩ = 0
D. None of these.

2: If Two vectors 7, • in an inner product space are orthogonal, then

A. ‖x + y‖, = 2‖x ‖, + 2‖y‖,

B. Kx + yK2 = ‖x ‖2 + KyK2
C. Kx + yK2 = 0
D. None of these.

3: Let € be a non empty subset of an inner product space <. Which of the following is not true.

A. € = €
B. € ⊂ €

Lovely Professional University 77

Notes
Functional Analysis

C. € = €
/ = <, then € = 0
D. If €

4 : In an orthonormal set of vectors, what is the inner product of any vector with itself?

A. 0
B. -1
C. 1
D. It depends on the vector.

5: If two vectors are orthogonal, what can be said about their inner product?

A. It is always zero.
B. It is always one.
C. It is undefined for orthogonal vectors.
D. None of these.

6: In a Hilbert space, what is the significance of a complete orthonormal set?

A. It forms a basis for the Hilbert space.

B. It is used for dimension reduction.
C. It only provides a partial basis,
D. None of these .

7: In the context of Hilbert spaces, what does Bessel’s inequality state?

A. It provides a bound on the norm of a vector in a Hilbert space.

B. It states that the sum of the squared coefficients of a vector with respect to to an
orthonormal set is bounded by the norm of the vector.
C. It defines the inner product between two vectors in a Hilbert space.
D. It establishes the existence of an orthonormal basis for any Hilbert space.

8: Let 6 = Y2 , Y, , … , YZ be an orthonormal set in a Hilbert space [. If 7 is any vector in [, then

which of the following is true.

A. ∑Z812|〈7, Y8 〉|, ≥ ‖7‖,

B. ∑Z812|〈7, Y8 〉|, < ‖7‖,
C. ∑Z812|〈7, Y8 〉|, ≤ ‖7‖,
D. None of these .
9: If Y8 is an orthonormal set in a Hilbert space [ and if 7 is any vector in [, then the set

S = e- : 〈x, e- 〉 0 is:
A. Non empty and uncountable
B. Non empty
C. Uncountable
D. Either empty or countable

78 Lovely Professional University

 Notes

Unit 07: Orthogonality of Inner Product Space

10: The Riesz-Fischer theorem provides a characterization of which type of space?

A. Hilbert spaces
B. Normed spaces
C. Metric space
D. None of these

Answers for Self Assessment

1. C 2. B 3. C 4. C 5. A

6. A 7. B 8. C 9. D 10. A

Review Questions
1. What does it mean for two vectors to be orthogonal?
2. Define a complete orthonormal set in a Hilbert space.
3. State Bessel's inequality in its general form, both for finite and countably infinite sets of
orthogonal functions.
4. State Parseval’s Identity.
5. State Pythagorean theorem.
6. State Riesz- Fischer Theorem.

Further Readings
1. Introductory Functional Analysis With Applications By Erwin Kreyszig.
2. Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
3. J. B Conway, A Course In Functional Analysis.
4. C. Goffman G Pedrick, A First Course In Functional Analysis.
5. B.V. Limaya, Functional Analysis.

Lovely Professional University 79

 Notes

Unit 08: Open Mapping Theorem and Closed Graph Theorem

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 08: Open Mapping Theorem and Closed Graph Theorem

CONTENTS
Objectives
Introduction
8.1 Open Mapping Theorem and its Applications
8.2 Closed Graph Theorem
8.3 Strong and Weak Convergence
8.4 Convergence of Sequences of Operators and Functional
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Open mapping Theorem and its Applications

 Closed graph theorem
 Weak and strong convergence
 Convergence of Sequence of operators and functionals.

Introduction
In this chapter, we discuss some very basic theorems of fundamental importance in functional
analysis. These theorems include Open Mapping theorem , Closed graph theorem. Further we
discuss about weak and strong convergence . Finally, we discuss about Convergence of sequences
of operators and functionals.

8.1 Open Mapping Theorem and its Applications

We have discussed the Hahn-Banach theorem and the uniform boundedness theorem and shall
now approach the third "big" theorem in this chapter, the Open mapping theorem. It will be
concerned with open mappings. These are mappings such that the image of every open set is an
open set .
More specifically, the open mapping theorem states conditions under which a bounded linear
operator is an open mapping. As in the uniform boundedness theorem we again need
completeness, and the present theorem exhibits another reason why Banach spaces are more
satisfactory than incomplete normed spaces. The theorem also gives conditions under which the
inverse of a hounded linear operator is bounded. The proof of the open mapping theorem will be
based on Baire's category theorem.
Before proving the Open Mapping theorem , we first know the following definition and lemma’s .

80 Lovely Professional University

Notes
Functional Analysis

1. A mapping : → , where and are topological spaces, is said to be open mapping if

maps open subsets of into open open subsets of .
Lemma. Let be a normed linear space and ( , ) be an open ball in . Then
( ; )= + (0; 1)
Proof. By definition
( ; )={ ∈ ∶‖ − ‖< }
={ ∈ ∶‖ ‖< , ℎ = − }
={ ∈ ∶‖ ‖< , ℎ = + }
={ + ∈ :‖ ‖< }
= +{ ∈ :‖ ‖< }

= +{ ∈ ∶ < 1}

= +{ ∈ ∶‖ ‖<1 ℎ = }
!
"

= +{ ∈ ∶‖ ‖<1 ℎ = }
= +{ ∈ : ‖ ‖ < 1}
= + { ∈ : ‖ ‖ < 1}
= + { ∈ :‖ − 0‖ < 1}
= + (0,1 )
i.e. ( ; )= + (0,1)
Remark. In particular (0; ) = (0; 1)
Lemma. Let # be a bounded linear operator from a Banach space into a Banach space . Then for
each open ball = (0,1) ⊂ , the image #( ) contains an open ball in with centre at origion.
Theorem. ( The open mapping theorem)
A bounded linear operator # from a Banach space into a Banach space is an open mapping.
Proof. Let #: → be a bounded linear operator froam a Banach space into a Banach space . In
order to show that # is an open mapping , we need to show that for any open set % ⊆ , the image
of % under # is open in . For this let ' ∈ #(%): since # is an operator , so there exists ∈ % such that
' = # ∈ #(%).
It is enough to show that #(%)contains an open ball around ' = # .
Since A is open in ;and −%≤0 by definition , it contains an open ball with centre and
radius
i.e.
( ; ) ⊆ %.
We know by lemma (1) above that:
( ; )= + (0; 1)………..(1)
By lemma (2) above , for the open ball (0; 1) in , there is an open ball (0, ) with centre at
origin, in such that
(0 ; ) ⊆ #( (0; 1))
⊆ #) (0; 1)*
= #( (0; ))……….(2)
='+ (0,1) by (1)
Hence ('; ) = ' + (0; )
= ' + #( (0; )) by (2)

Lovely Professional University 81

 Notes

Unit 08: Open Mapping Theorem and Closed Graph Theorem

= # + #( (0; ))
= #( + (0; ))
= #( + (0,1))
⊆ #(%)
i.e. ('; ) ⊆ #(%)
This shows that #(%) contains an open ball around ' = # . Consequently #(%) is open in
hence # is an open mapping.
and

Corollary. Let #: → be a bijective bounded linear operator from a Banach space

space , then # is homeomorphism.
into a Banach

Proof. We recall that # is a homeomorphism if

(i) # is continuous .
(ii) # is bijective.
(iii) # +, is Continuous.
Since # is continuous ( as # is bounded ) and bijective, so # +, : → exists.
To show that # is continuous, let - be an open set in , then
+, (# +, )+,
- = #-, which is open in
because # is open by open mapping theorem. So that the image of any open set in is open
in under # +, , showing that # +, is continuous. Hence # is a homeomorphism.

8.2 Closed Graph Theorem

The next important theorem which we shall prove is called the closed graph theorem. Before
provimg this theorem we have some definitions.
Let ( , ‖. ‖) and ( , ‖. ‖ ) be normed spaces
. = {( , ' ): ∈ ,' ∈ }
Define addition and scalar multiplication in . by:
( ,' ) +( ,' ) = ( + , ' + ' )………..(1)
/( , ' ) = (/ , /' )………..(2)

For all , ∈ 012 ', ' ∈ 012 / ∈ 3 = (4 5 6)

Then . is a linear space under the addition and scalar multiplication defined above. Next, we
define norm on . as follows:
For ( , ' ) ∈ ., we put
8
‖( , ' )‖7 = )‖ ‖7 + ‖'‖7 *9 , 1 ≤ : < ∞………..(3)

Then obviously,
‖( , ' )‖ ≥ 0 012 ‖( , ' )‖ = 0 ⇔ = 0, ' = 0
And ‖/( , ' )‖77 = |/| ‖( , ' )‖77
‖( , ' ) + ( , ' )‖7 = ‖( + , ' + ' )‖7
8
= (‖ + ‖7 + ‖' + ' ‖7 )9
8
≤ )‖ ‖7 + ‖ ‖7 + ‖'‖ + ‖' ‖ *9
7 7

8 8
≤ )‖ ‖7 + ‖'‖ *9 + (‖ ‖7 + ‖' ‖ )9
7 7

(By Minkowski’s inequality)

≤ ‖( , ' )‖7 + ‖ , ' ‖7

82 Lovely Professional University

Notes
Functional Analysis

Hence (., ‖. ‖7 ) is a normed space , called the product of the normed spaces and .
For : = 1 , (3) assumes the form
‖( , ' )‖, = ‖ ‖ + ‖'‖ .
Another norm on . is given as follows: For ( , ' ) ∈ ., we put
‖( , ' )‖ = max(‖ ‖, ‖'‖ )
It can be established that all these norms on × are equivalent.
For : = 2, we have
,
‖( , ' )‖D = (‖ ‖D + ‖'‖D )D
If and are Banach spaces then so is their product .. This follows from the observation that
( E , 'E ) → ( ,' ) ⇔ E → , 'E → ' .
For any two normed spaces and and a mapping #: → , the set
FG = {( , # ): ∈ }
is called the graph of #.
Since and are metric spaces and so are Hausdroff spaces, their product . = ×
metric indued by the norm on ., is also metric space.
, under the

In general, for two topological spaces , , the graph FG of a mapping #: →

closed subspace of × . Since every normed space is also a Hausdroff space , so in the case of
may not be a

normed spaces,the graph of continuous mapping #: → is always closed .

Theorem. (Closed Graph Theorem)
Let and be Banach spaces and #: → be a linear operator . Then # is continuous if and only
if the graph of # is a closed subspace of × .
Proof. Suppose that #: → is a continuous linear operator . we show that the graph
FG = {( , # ): ∈ }
HHHGH. Then there are sequences {
is closed in × . For this let ( , ' ) ∈ F E} and {'E } in and
respectively such that

E → , 'E → '
Since # is continuous and E =# E,

E → ⇒# E → #J = '
Hence ( , ' ) = ( , #J ) ∈ FG , thus FG is closed.
Conversely suppose that, for a linear operator #: → , FG is closed.Then FG is a subspace of
× . Since and are Banach spaces and FG is a closed subspace of the Banach space × , FG
itself is complete and hence is a Banach space .
Consider the mapping : FG → defined by
( ,# ) = ∀ ∈ .
Then is injective and linear . Also, since
‖ ( , # )‖ = ‖ ‖ ≤ ‖( , # )‖
By definition of the product norm, is continuous. By the open mapping theorem , +, is
continuous and so bounded. Moreover
‖# ‖ ≤ ‖( , # )‖ = ‖ +, ( )‖ ≤ ‖ +, ‖‖ ‖
Hence # is bounded and so is continuous .

Lovely Professional University 83

 Notes

Unit 08: Open Mapping Theorem and Closed Graph Theorem

8.3 Strong and Weak Convergence

We know that in calculus we define different types of convergence i.e ordinary, conditional,
absolute and uniform convergence. This yields greater flexibility in the theory and applications of
sequence and series. The situation is similar in functional analysis, and one has an even greater
variety of possibilities that turn out to be of practical interest. Here we are concerned with weak
convergence .This is basic concept . We present it now since its theory makes essential use of the
uniform boundedness theorem which we have already discussed . In fact, this is one of the major
application of that theorem.
Definition. (Strong convergence) A sequence { E } in a normed space is said to be strongly
convergent (or convergent in the norm) if there is an ∈ such that limE→N ‖ E − ‖ = 0
i,e.
lim E =
E→N

or simply

E → .
is called the strong limit of { E }, and we say that { E} converges strongly to .
Weak convergence is defined in terms of bounded linear functionals on as follows.
Definition ( Weak convergence) . A sequence { E } in a normed space
convergent if there is an ∈ such that for every ∈
is said to be weakly

limE→N ( E) = ( )
This is written

E →
The element is called the weak limit of { E }, and we say that { E} converges weakly to .
Weak convergence has various applications throughout analysis for instance , in the calculus
of variation, and general theory of differential equation.
For applying weak convergence one needs to know certain basic properties, which we state in
the following lemma .
Lemma. Let { E} be a weakly convergent sequence in a normed space , say E → . Then:

(i) The weak limit of { E} is unique.

(ii) Every subsequence of { E} converges weakly to .
(iii) The sequence (‖ E ‖) is bounded.
Proof. (i) Suppose that E → as well as E → ' . Then
( E) → ( ) as well as ( E) → (').
Since { ( E )} is a sequence of numbers , its limit is unique . Hence ( ) = ('), that is for every
∈ we have
( ) − (') = ( − ') = 0
⇒ −' =0
This shows that the weak limit is unique.
(ii) follows from the fact that { ( E )} is a convergent sequence of numbers, so that every
subsequence of { ( E )} converges and has the same limit as the sequence.
(iii) Since { ( E )} is a convergent sequence of numbers , it is bounded, say | ( E )| ≤
OP ∀ 1, where OP is a constant depending on but not on 1 . Using the canonical mapping
6: → , we can define QE ∈ by

QE ( ) = ( E) , ∈
Then for all 1,
|QE ( )| = | ( E )| ≤ OP ,

84 Lovely Professional University

Notes
Functional Analysis

that is, the sequence {RQE(P) R} is bounded for every ∈ . As we know the dual space

applicable and implies that {RQE(P) R} is bounded . Now ‖QE ‖ = ‖ E ‖ .

of a normed space is a Banach space, so is complete , the Uniform boundedness theorem is

In finite dimensional normed spaces the distinction between strong and weak convergence
disappears completely . Let us prove this fact and also justify the terms “strong” and “weak”.
Theorem. (Strong and weak convergence). Let { E} be a sequence in normed space . Then

(i) Strong convergence implies weak convergence with the same limit.
(ii) The converse of (i) is not generally true.
(iii) If dim < ∞, then weak convergence implies srrong convergence.
Proof. By definition, E → means ‖ E − ‖ → 0 and implies for every ∈ ,
| ( E) − ( )| = ( E − ) ≤ ‖ ‖‖ E − ‖→0
This shows that E → .
(ii) can be seen from an orthonormal sequence { E } in a Hilbert space T. In fact , every ∈ T has a
Riesz representation ( ) = 〈 , 〉. Hence ( E ) = 〈 E , 〉. The Bessels inequality is as

EX,|〈 E , 〉| ≤ ‖ ‖ .
∑N D D

Hence the series on the left converges , so that its terms must approach zero as 1 → ∞. This implies
( E ) = 〈 E , 〉 → 0.
Since ∈ T was arbitrary , we see that E → 0. However, { E } does not converge strongly because
‖ Y − E‖
D
=〈 Y − E, Y − E〉 =2 (Z ≠ 1).

(iv) Suppose that E → and dim = \. Let { , , D , … ^ } be any basis for and say ,
= /, , + ⋯ + /^ ^
(E) (E)
E

And
= /, , + ⋯ + /^ ^

By assumption, ( E) → ( ) for every ∈ . We take in particular ,, … , ^ defined

by
`( `) = 1, `( Y) =0 (Z ≠ a)
Then
` ( E) = /` , `( ) = /` .
(E)

Hence ` ( E) → `( ) ⇒ /` → /` . From this we readily obtain

(E)

^
‖ E − ‖ = bc(/`(E) − /` ) ` b
`X,

≤ c d/` − /` d e ` e → 0
(E)

`X,

As 1 → ∞. This shows that { E} converges strongly to .

Lovely Professional University 85

 Notes

Unit 08: Open Mapping Theorem and Closed Graph Theorem

convergence are equivalent concepts. As example is f, . In conclusion let us take a look at weak
It is interesting to note that there also exist infinite dimensional spaces such that strong and weak

convergence in two important types of spaces.

1 → if and only if 〈 1 , 〉→
, 〉∀
Example. Hilbert space. In a Hilbert space,
〈 in the space.

the space f , where 1 < : < +∞, we have →

:
Example. In 1 if and only if
(i) The sequence {‖ E ‖} is bounded.
For every fixed a we have g` → g` as 1 → ∞; here,
(E )
(ii)
= hg` i = (g` ).
(E)
E and

8.4 Convergence of Sequences of Operators and Functional

Sequences of bounded linear operators and functionals arise frequently in the abstract formulation
of concrete situations, for instance in connection with convergence problems of Fourier series or
sequences of interpolation polynomials or methods of numerical integration. In such cases one is
usually concerned with the convergence of those sequences of operators or functionals with
boundedness of corresponding sequences of norms or with similar properties.

as defined in the above section are useful concepts. For sequences of operators #E ∈ ( , ) three
Experience shows that for sequences of elements in a normed space, strong and weak convergence

types of convergence turn out to be of theoretical as well as practical value.These are

(i) Convergence in the norm on ( , ),

(ii) Strong convergence of {#E } in ,
(iii) Weak convergence of {#E } in ,
The definition and terminology are as follows;

sequence {#E } of operators #E ∈ ( , ) is said to be

Definition. (Convergence of sequence of operators) Let and be normed spaces. A

(i ) uniformly operator convergent if {#E } converges in the norm on ( , )

(ii) Strongly operator convergent if {#E } converges strongly in for every ∈ ,
(iii ) weakly operator convergent if {#E } converges weakly in for every ∈ .
In formulas this means that there is an operator #: → such that

(i) ‖#E − #‖ → 0
(ii) ‖#E − # ‖ → 0 for all ∈
(iii) ‖ (#E ) − (# )‖ → 0 for all ∈ and for all ∈
respectively. # is called the uniform, strong and weak operator limit of {#E }, respectively.
Definition. (Strong and weak convergence of a sequence of functional) Let { E } be a sequence
of bounded linear functional on a normed space . Then :

(a) Strong convergence of { E } means that there is an ∈ such that ‖ E − ‖ → 0 . This is

written
E → .

(b) Weak convergence of { E } means that there is an ∈ such that E( ) → ( ) for all
∈ . This is written
E → .
in (a) and (b) is called the strong limit and weak limit of { E }, respectively.

86 Lovely Professional University

Notes
Functional Analysis

Lemma. Let #E ∈ ( , ), where is a Banach space and a normed space . If {#E } is strongly
operator convergent with limit #, then # ∈ ( , ).
Proof. Linearity of # follows readily from that of #E . Since #E → # for every ∈ , the sequence
{#E } is bounded for every . Since is complete, ‖#E ‖ is bounded by the uniform boundedness
theorem , say ‖#E ‖ ≤ O ∀1 . From this, it follows that
‖#E ‖ ≤ ‖#E ‖‖ ‖ ≤ O‖ ‖.
This implies
‖# ‖ ≤ O‖ ‖.
A useful criterion for strong operator convergence is

Theorem. A sequence {#E } of operators #E ∈ ( , ), where and are Banach spaces , is strongly
operator convergent if and only if:

(A) The sequence {‖#E ‖} is bounded.

(B) The sequence {#E } is Cauchy in for every in a total subset j of .

Proof. If #E → # for every ∈ , then (A) follows from the uniform boundedness theorem (since
is complete) and (B) is trivial.

Conversely, suppose that (A) and (B) holds, so that, say ‖#E ‖ ≤ O ∀ 1. We consider any ∈ and
show that {#E } converges strongly in . Let k > 0 be given . Since span j is dense in , there is a
' ∈ m:01 j such that

k
‖ −'‖< .
3O

Since ' ∈ m:01 j, the sequence {#E '} is Cauchy by (B). Hence there is an o such that

‖#E ' − #Y ' ‖ < (Z, 1 > o)

p
q

Using these two inequalities and applying the triangle inequality, we see that {#E } is Cauchy in
because for Z, 1 > o we obtain

‖#E − #Y ‖ ≤ ‖#E − #E ' ‖ + ‖#E ' − #Y '‖ + ‖#Y ' − #Y ‖

< ‖#E ‖ ‖ − ' ‖ + + ‖#Y ‖ ‖ − ' ‖

p
q

<O + +O =k
p p p
qr q qr

Since is complete, {#E } converges in . Since ∈

convergence of {#E }.
was arbitrary, this proves strong operator

Corollary. (Functionals) A sequence { E } of bounded linear functionals on a Banah space is weak

convergent, the limit being a bounded linear functional on , if and only if

(A) The sequence {‖ E ‖} is bounded.

(B) The sequence { E ( )} is Cauchy for every in a total subset j of .

Summary
 A mapping : → , where and are topological spaces, is said to be open mapping if
maps open subsets of into open open subsets of .
 Let be a normed linear space and ( , ) be an open ball in . Then

Lovely Professional University 87

 Notes

Unit 08: Open Mapping Theorem and Closed Graph Theorem

( ; )= + (0; 1).
 Let # be a bounded linear operator from a Banach space into a Banach space . Then for
each open ball = (0,1) ⊂ , the image #( ) contains an open ball in with centre
at origion.
 A bounded linear operator # from a Banach space into a Banach space is an open
mapping. (Open mapping theorem)
 Let #: → be a bijective bounded linear operator from a Banach space into a Banach
space , then # is homeomorphism.
 Let and be Banach spaces and #: → be a linear operator . Then # is continuous if
and only if the graph of # is a closed subspace of × . ( Closed graph theorem).
 A sequence { E} in a normed space is said to be strongly convergent (or convergent in
the norm) if there is an ∈ such that limE→N ‖ E − ‖=0
i,e.
limE→N E = .
 A sequence { E} in a normed space is said to be weakly convergent if there is an ∈
such that for every ∈
lim ( E) = ( )
E→N

This is written
E → .
The element is called the weak limit of { E }, and we say that { E} converges weakly to .

 Let #E ∈ ( , ), where is a Banach space and a normed space . If {#E } is strongly

operator convergent with limit #, then # ∈ ( , ).

Keywords
 Open mappings
 Open set
 Bounded linear operator
 Incomplete normed spaces
 Open ball
 Homeomorphism
 Closed graph
 Closed subspace
 Weak convergence
 Strong convergence
 Hilbert space

Self Assessment
1: If , are normed spaces and if % ∶ → is a bijective, bounded linear map, then:
A. % is always an open map.
B. % is an open map if is a Banach space.
C. % is an open map if is a Banach space.
D. % is an open map if both and are Banach spaces.

88 Lovely Professional University

Notes
Functional Analysis

2: If and are normed spaces, and if # ∶ → is a linear operator, then # is bounded if and
only if:
A. # maps bounded subsets of into bounded subsets of .
B. # maps open subsets of into open subsets of .
C. # maps closed subsets of into closed subsets of .
D. # is invertible.
3: Every bounded operator of finite rank is :
A. Open.
B. Compact.
C. Has a non zero adjoint.
D. None of these.

4: A bijective map % ∶ → is open if and only if :

A. % ∶ → is invertible.
B. % ∶ → is bounded.
C. % +, ∶ → is bounded.
D. %+, ∶ → is open.

5: If {%E } is a sequence of operators on a normed space , then %E → % strongly if and only if:
A. %E → % ∀ ∈ .
B. ‖%E − % ‖ → 0 as 1 → ∞.
C. (%E ) → (% )∀ ∈ and ∀ ∈ ∗.

D. None of these.

6: If # is a bounded linear operator, then:

A. ‖# ‖ ≥ ‖#‖. ‖ ‖.
B. ‖# ‖ ≤ ‖#‖. ‖ ‖.
C. ‖# ‖ = ‖#‖. ‖ ‖.
D. None of these.

7: For , ' in a normed space , which of the following is not necessarily true?
A. ‖ + '‖ ≤ ‖ ‖ + ‖'‖.
B. |‖ ‖ − ‖'‖| ≤ ‖ − '‖
C. |‖ ‖ − ‖'‖| ≤ ‖ ‖ + ‖'‖.
D. ‖ − '‖ ≤ ‖ ‖ + ‖'‖.

8: Let be a normed space and be a bounded, non-zero linear functional on . Then, which of
the following is not true?
A. is onto.
B. is continuous.
C. t is a closed subspace of .
D. is an open map.

9 : Let be a normed space and %, be bounded linear operators on Then which of the
following is true?
A. ‖% ‖ ≥ ‖%‖. ‖ ‖
B. ‖% ‖ = ‖%‖. ‖ ‖
C. ‖% ‖ ≤ ‖%‖. ‖ ‖

Lovely Professional University 89

 Notes

Unit 08: Open Mapping Theorem and Closed Graph Theorem

D. None of these.

10: Which of the following theorems guarantees that a bounded linear operator between Banach
spaces is an open map if it is onto.
A. Hahn-Banach Theorem.
B. Open Mapping theorem.
C. Baire’s Category theorem.
D. None of these.

11: Let and be Banach spaces and #: → be a bounded linear operator , if the range of #
is not closed, then which theorem can be used to find a closed subspace of on which # is
injective?
A. Hahn-Banach Theorem.
B. Open Mapping theorem.
C. Closed graph theorem.
D. None of these.

12: What is the main difference between strong and weak convergence?
A. Strong convergence requires convergence in norm, while weak convergence requires
pointwise convergence.
B. Strong convergence requires pointwise convergence, while weak convergence requires
convergence in norm.
C. Strong convergence and weak convergence are synonymous terms.
D. None of the above.

13: In a Hilbert space, which of the following statements is true?

A. Strongly convergent sequences are always weakly convergent.
B. Weakly convergent sequences are always strongly convergent.
C. Strong and weak convergence are equivalent.
D. None of these.

14: If a linear mapping between topological vector spaces is continuous, what can we conclude
about its graph?
A. The graph is open.
B. The graph is compace.
C. The graph is closed.
D. None of these.

Let and be Banach spaces and #: → be a linear map which is closed and surjective.
Then # is continuous and open. This is called
15:

A. Closed graph theorem.

B. Hein-Borel theorem.
C. Open mapping theorem.
D. None of these.

90 Lovely Professional University

Notes
Functional Analysis

Answers for Self Assessment

1. D 2. A 3. B 4. C 5. A

6. B 7. D 8. D 9. C 10. B

11. C 12. A 13. A 14. C 15. C

Review Questions
1. State the Open Mapping Theorem in functional analysis.
2. State the Closed Graph Theorem.
3. Define the graph of a linear operator.
4. Under what conditions does the Closed Graph Theorem holds.
5. What is difference between strong and weak convergence.

Further Readings
1. Introductory Functional Analysis With Applications By Erwin Kreyszig.
2. Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
3. J. B Conway, A Course In Functional Analysis.
4. C. Goffman G Pedrick, A First Course In Functional Analysis.
5. B.V. Limaya, Functional Analysis.

Lovely Professional University 91

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 09: Decomposition Theorems in Hilbert Spaces

CONTENTS
Objectives
Introduction
9.1 Orthogonal Complements and Direct Sums
9.2 Convex Sets in Hilbert Spaces
9.3 The Conjugate Space of a Hilbert Space
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Orthogonal complements and direct sums

 Pojection Theorem
 Convex sets in Hilbert Spaces
 The conjugate space of a Hilbert space

Introduction
In this chapter, We discuss about Orthogonal complements and direct sums and its properties.
Futher, we prove Projection theorem and convex sets in Hilbert space and discuss some important
theorems. Finally we discuss about Conjugate space of a Hilbert space .

9.1 Orthogonal Complements and Direct Sums

Definition. If is any subset of a Hilbert space , then the orthogonal complement of denoted
by , is defined as
={ ∈ ∶〈 , 〉=0∀ ∈ }
={ ∈ ∶ ⊥ }
And also = ={ ∈ ∶ 〈 , 〉 = 0, ∀ ∈ }
={ ∈ ⊥ }

Remark. From the above definitions, it is clear that

(i) {0} =
(ii) = {0}
Theorem. Let , be subsets of a Hilbert space , then prove the following:

92 Lovely Professional University

Notes
Functional Analysis

(i) ⊆ , that is any subset of is contained in its double orthogonal complement .

(ii) If ⊆ , then ⊆ .
(iii) ∪ = ∩ and ∩ ⊇ ∪ .
(iv) = .
(v) ∩ ⊆ {0}.
(vi) is a closed subspace of .
Proof.

(i) Let ∈ , then

〈 , 〉=0∀ ∈ . Hence
∈ , that is ⊆ .
(ii) Suppose that ⊆ . Let ∈ , then
〈 , 〉=0∀ ∈ .
Since ⊆ , we have
〈 , 〉=0∀ ∈ .
Hence ⊥ ,
implies ∈
that is ⊆ .
(iii) Since ⊆ ∪ and ⊆ ∪
⇒ ∪ ⊆ and ∪ ⊆
⇒ ∪ ⊆ ∩ ………..(1)
Now, let ∈ ∩
⇒ ∈ and ∈
⇒ ⊥ and ⊥
So by definition , 〈 , 〉 = 0 for every ∈ and
〈 , 〉 = 0 for every ∈ .
And so 〈 , 〉 = 0 for every ∈ ∪ .
⇒ ∈ ∪
So that ∩ ⊆ ∪ ………..(2)
From (1) and (2), we get
∪ = ∩ .
Next, we show that ∩ ⊇ ∪ .
For this since , ∩ ⊆ and ∩ ⊆
⇒ ⊆ ∩ and ⊆ ∩
⇒ ∩ ⊇ ∪ .

(iv) By (i) ⊆
and so by part (ii) ⊆
i.e. ⊆ ...........(3)
Also by part (i), ⊆ = ………..(4)
From (3) and (4), we have
= .

(v) If ∩ = , then clearly ∩ = ⊆ {0}

i.e., ∩ ⊆ {0}
If ∩ ≠ , then let ∈ ∩

Lovely Professional University 93

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

⇒ ∈ and ∈ .
Now since ∈ and ∈
⇒〈 , 〉=0
⇒ ‖ ‖ = 0,
i.e. ‖ ‖ = 0
i.e., =0
i.e., ∈ {0}.
Hence,
∩ ⊆ {0}.

(vi) Now we show is a closed subspace of .

Let , be any two elements in and #, $ be any scalers, then for in , we have:
〈 , 〉 = 0 and 〈 , 〉 = 0 and therefore:
〈# + $ , 〉 = 〈# , 〉 + 〈$ , 〉
= #〈 , 〉 + $〈 , 〉
= #. 0 + $. 0
=0
i.e., 〈# + $ , 〉 = 0 for any in .
⇒ # +$ ∈
Which shows that is a subspace of .
To complete the proof , it remains to show that is closed and in order to prove this, it is enough
to show that if { & } is any convergent sequence in converging to a point (say)
i.e., & → , then ∈ .
Now for any ∈ , we can write
〈 , 〉 = 〈 lim &, 〉
&→+

= lim 〈 &, 〉
&→+

= 0, because & ∈
i.e., 〈 , 〉 = 0 for any ∈ ,⇒ ⊥
⇒ ∈ . Thus is closed subspace of .
Theorem. If is a closed linear subspace of a Hilbert space , then ∩ = {0}.
Proof. Let ∈ ∩ , then ∈ and ∈ .
⇒ ⊥ .
⇒ 〈 , 〉 = 0 for every in .
⇒ 〈 , 〉 = 0, because ∈ .
⇒‖ ‖ =0

⇒‖ ‖=0⇒ = 0.
This shows that 0 ∈ ∩ ⇒ {0} ⊆ ∩ .

But we know that ∩ ⊆ {0}, by part (v) of above theorem.

Hence ∩ = {0}.
Remark. For sets and , ∩ ⊆ {0} and for subspaces and , ∩ = {0}.
The reason is that it is not necessary for any subset to contain 0 but every subspace contains 0.

94 Lovely Professional University

Notes
Functional Analysis

Definition. (Direct sum). A vector space , is said to be the direct sum of two subspaces - and . of
,, written
, = -⨁.
,
if each ∈ , has a unique representation
= + 0, ∈ -, 0 ∈ .
.
Then . is called an algebraic complement of - in , and vice versa, and -, . is called a
complementary pair of subspaces in ,.
For example, - = 1 is a subspace of the Euclidean plane 1 . Clearly, - has infinitely many
algebraic complements in 1 , each of which is a real line. But most convenient is a complement that
is perpendicular. We make use of this fact when we choose a Cartesian coordinate system. In 1 2 the
situation is the same in principle.
Theorem 1. (Minimizing vector) Let be a non empty complete convex subset of an inner
Product space , and ∈ ,\ . Then there is a unique ∈ such that
‖ − ‖ = 56789∈: ‖ − ;‖

That is, there is a unique ∈ which is closest to .

Proof. We prove this theorem in the next section of this chapter.
Theorem 2. Let be a complete subspace of an inner product space ,. Then there is a non zero
vector 0 ∈ , such that
0⊥
.

Theorem. Let be a proper complete subspace of an inner product space ,. Then

,= ⨁

Proof. Since is complete and being a subspace, is convex, by theorem 1 above there is a unique
vector ∈ such that
‖ − ‖ = 56789∈: ‖ − ; ‖,
for each ∈ ,\ .

Put 0 = − , by theorem 2, 0 ⊥ and so 0 ∈ which is a subspace of ,. Hence

= +0, ∈ ,0 ∈ ………..(1)
To see that (1) is unique , suppose that
= +0
Also then
− =0 −0∈ ∩ = {0}
so = , 0 = 0 . Therefore
,= ⨁ .
Remark. For any complete subspace of an inner product space , the subspace of , is called
the orthogonal complement of . In particular, if is closed subspace of a Hilbert space , the
is orthogonal complement of in .
Corollary. Let be a closed subspace of a Hilbert space . Then
= ⨁ .
Proof. Since , as a closed subspace of a Hilbert space which is always a complete space , is
complete.
Corollary. For any complete subspace of an inner product ,,

Lovely Professional University 95

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

= .
Remark. A subspace of a Hilbert space is closed if and only if = .
Theorem. Show that if and < are closed subspaces of a Hilbert space such that ⊥ <. Then
+ < is closed subspace of .
Proof. We know that if and < are any subspaces, Then M+N is always a subspace . To show
that M+N is aclosed subspace of , let 0 be a limit point of M+N, then there is a sequence {0& } in
M+N such that
0 = lim 0&
&→+

Now
0& = & + &, & ∈ , & ∈ <.
We show that { &} and { & } are Cauchy’s sequences in and < respectively.
Since by Pythagorean theorem,
‖0= − 0& ‖ = ‖ = + = − & − &‖

=‖ = − & + = − &‖

=‖ = − &‖ +‖ = − &‖

and since 0& is a Cauchy sequence, so also are { & }, { & }. Also as closed subspaces of , both and
< are complete . So
lim & = ∈
&→+

lim & = ∈<

&→+

Hence 0 = lim&→+ 0& = lim&→+ & + & = +

Thus 0 ∈ + <. That is , + < is closed.
Theorem. Let be a closed linear subspace of a Hilbert space . Then ∩ = {0}.
Proof. Since we know that if is a subset of a Hilbert space . Then
∩ ⊆ {0}………..(1)
Given that is closed linear subspace of and we also know that is closed linear subspace of
. Let ∈ ∩ implies ∈ and ∈ and so 〈 , 〉 = 0.
i.e., ‖ ‖ = 0
⇒ =0⇒0∈ and 0 ∈ ⇒0∈ ∩ .
⇒ {0} ∈ ∩ ………..(2)

Combining (1) and (2)

∩ = {0}.

Projection Theorem.

Let be any closed subspace of a Hilbert space . Then

= ⊕ .

Proof. Suppose + is proper subspaceof then there is a non-zero vector 0 ∈ such that

0⊥ + i.e 0 ∈ + .

Now ⊆ + implies + ⊆ .

96 Lovely Professional University

Notes
Functional Analysis

Also we know ⊆ + implies + ⊆ .

Thus 0 ∈ + ⊆ ∩ = {0} ⇒ 0 = 0, a contradiction .

Hence + is the whole of . i.e. = + since ∩ = {0}.

Thus

= ⊕ .

9.2 Convex Sets in Hilbert Spaces

Before defining the convex sets in Hilbert spaces we first recall the following.
In a metric space ,, the distance ? from an element ∈ , to a nonempty subset ⊂ , is defined to
be
? = 56789∈: A , ; .

In a normed space this becomes

? = 56789∈: ‖ − ; ‖

The line segment joining two given elements and of a space , is defined to be the set of all
0 ∈ , of the form: 0 = B + 1 − B for every real number B such that 0 ≤ B ≤ 1.
Definition. A subset of , is said to be convex if for every , ∈ , the line segment joining
and is contained in ,
i.e., 0 = B + 1 − B ∈ for every B, where 0 ≤ B ≤ 1.
Every subspace - of , is convex, and the intersection of convex sets is a convex set.
We shall use the notion of convexity in the following theorem.
Theorem. (Minimizing vector) Let be a non empty complete convex subset of an inner product
space , and ∈ ,\ . Then there is a unique ∈ such that
‖ − ‖ = 56789∈: ‖ − ;‖

That is, there is a unique ∈ which is closest to .

Proof. Let A = 56789 ∈: ‖ − ;‖

Then by definition of infimum , there is sequence { & } in M such that

A = lim ‖ − &‖
&→+

We show that { & } is a Cauchy sequence in .

Now by Parallelogram law, we have
‖ ; − ;‖ = 2‖ ; ‖ + 2‖ ; ‖ − ‖ ; + ;‖ ………..(1)
; ;
Replacing by = − and by & − , we have
‖ = − &‖ = 2‖ = − ‖ + 2‖ & − ‖ −‖ = + & −2 ‖

= 2‖ = − ‖ + 2‖ & − ‖ − 4G = + & − G ………..(2)

Since is convex , = + & ∈ , so we have from (2)

‖ = − &‖ ≤ 2‖ = − ‖ + 2‖ & − ‖ − 4A
→ 0 HI J, 6 → ∞
As ‖ = − ‖ → A, ‖ = − ‖ → 0. Hence { & } is a Cauchy sequence in . Since is complete,

Lovely Professional University 97

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

& → ∈ . So
A = lim&→+ ‖ − &‖

= G − lim &G
&→+

= ‖ − ‖ with ∈ .
Next we prove the uniqueness of . Suppose there is another L ∈ , such that
A=‖ − L‖
; ;
Then again, Using the parallelogram law as given in (1) and replacing by − and by L − ,
we have :
‖ − L‖ = 2‖ − ‖ + 2‖ L − ‖ −‖ + L −2 ‖
= ‖ − ‖ = 56789 ∈: ‖ − ;‖

Since is convex and + L ∈ , we have

‖ − L‖ ≤ 4A − 4A
≤0
But
‖ − L ‖ ≥ 0.

Hence ‖ − L‖ = 0, that is = L.

This proves the uniqueness of .

9.3 The Conjugate Space of a Hilbert Space

Let be a Hilbert space . Then a scalar valued fubction 7: → O is called a functional on , if 7 is
linear and bounded (continuous). Set of all such functionals is denoted by P , O or simply ∗
and ∗ is called conjugate space of a Hilbert space .
∗
So if 7 ∈ implies 7: → O is a functional .
(OR)
Let be a Hilbert space . By ∗ , we denote the conjugate space of (i.e. the set of all continuous
linear transformations of into O). The elements of ∗ are called continuous linear functionals or
briefly functional.
One of the fundamental properties of a Hilbert space is the fact that there is a natural
correspondence between the vectors in and functional in ∗ .
Theorem. Let be a fixed vector in a Hilbert space and let 78 be a function defined as

78 = 〈 , 〉 for every ∈ . Then 78 is a functional on and ‖ ‖ = R78 R.

Proof. We prove that 78 is linear and continuous so that it is a functional .
To prove 78 is linear , let , ∈ and #, $ be any two scalars. Then for any fixed ∈ ,
78 # +$ = 〈# +$ , 〉
= #〈 , 〉 + $〈 , 〉
= #78 + $78 .
This shows that 78 is linear . To prove that 78 is continuous , for any ∈

S78 S = |〈 , 〉|

≤ ‖ ‖‖ ‖…………(1)
( by Schwarz inequality)
Suppose ‖ ‖ ≤ . Then for > 0, S78 S≤ ‖ ‖.

98 Lovely Professional University

Notes
Functional Analysis

Hence 78 is bounded and hence it is continuous.

Now if = 0, ‖ ‖ = 0 and from the definition 78 = 0 so that R78 R ≤ ‖ ‖ in this case.

Let ≠ 0.
From (1), we get
|78 |
sup ≤‖ ‖
‖ ‖
Hence using the definition of the norm of a functional, we get
R78 R ≤ ‖ ‖……….(2)
Further,

R78 R = sup{|78 |: ‖ ‖ ≤ 1}. ………..(3)

Since
8
≠ 0, Y‖8‖Z is a unit vector .From (3), we get
8
R78 R ≥ [78 Y‖8‖Z[………..(4)

But

78 \ ]=〈 , 〉
‖ ‖ ‖ ‖

= ‖8‖ 〈 , 〉

= ‖ ‖.
Using this in (4), we get
R78 R ≥ ‖ ‖…………(5)
Combining (2) and (5), we get

R78 R = ‖ ‖.
∗
Thus we have proved that ^: → is such that ^ = 78
is a norm preserving mappings.

Theorem. Show that the mapping : → ∗ defined by = 78 where 78 = 〈 , 〉 for every

∈ is an additive, one to one onto isometry but not linear.
Proof. First , we prove that is additive. For this we have to show that
+ = + for , ∈ .
Now we have
+ = 78_ `8a .

Hence for every ∈ , we get

78_ `8a =〈 , + 〉

=〈 , 〉+〈 , 〉

= 78_ + 78a

= 78_ + 78a .

Lovely Professional University 99

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

Hence ,
78_`8a = 78_ + 78a .
Which implies ,
+ = + .
This shows that is additive.
Now we show is one –one . Let , ∈ . Then
= 78_ and
= 78a , then
= , implies
78_ = 78a , which gives
78_ = 78a ∀ ∈ ………..(1)
78_ =〈 , 〉 and

78a =〈 , 〉, so we get from (1)

〈 , 〉=〈 , 〉
i.e, 〈 , − 〉=0∀ ∈ ………..(2)
Now choose = − , then (2) gives 〈 − , − 〉=‖ − ‖ = 0.
Which implies = .
Therefore is one- one .
is onto , let 7 ∈ ∗,
Now to prove then by Riesz representation theorem, there exists ∈ such
that
7 = 〈 , 〉.
∗,
Since 7 = 〈 , 〉, we get 7 = 78 so that = 78 = 7. Hence for 7 ∈ there exists a pre image
in . There fore is onto.
To prove that is an isometry, let , ∈ . Then
‖ − ‖ = R78_ − 78a R

= R78_ + 7 b8a R.

But R78_ + 7b8a R = R78_ b8a R = ‖ − ‖.

Hence ‖ − ‖=‖ − ‖.
Finally, we prove that is not linear, for this let ∈ and # be any scalar. Then
# = 7c8 .
Hence for any ∈ , we get 7c8 = 〈 ,# 〉
=#〈 , 〉
= #78 .
Which gives, 7c8 = #78 .
So that,
# =#
This shows that the mapping is not linear. Such a mapping is called conjugate linear. Thus
is conjugate linear.
Theorem. If is a Hilbert space , then ∗ is also Hilbert space with the inner product defined by
〈7d , 78 〉 = 〈 , 〉∀ , ∈ ………..(1)

100 Lovely Professional University

Notes
Functional Analysis

Proof. Since is a Hilbert space , so is also a Banach space . We know that conjugate of a Banach
space is also a Banach space . Therefore ∗ is also a Banach space.
To show ∗ is a Hilbert space, it is sufficient to show that ∗
is inner product space with respect to
the inner product defined by
〈7d , 78 〉 = 〈 , 〉 ∀ , ∈ .
Let , ∈ and #, $ be complex scalars , we have

1. 〈7d , 7d 〉 = 〈 , 〉

=‖ ‖

= ‖7d ‖
So that
〈7d , 7d 〉 ≥ 0 and ‖7d ‖ = 0 if and only if 7d = 0.

2. 〈7d , 78 〉 = 〈 , 〉

=〈 , 〉

= 〈78 , 7d 〉.

3. In the above theorem, we have shown that then 7c8 = #e78 .

Hence

7c8 = # 78 = #78 .
Now
〈#7d + $78 , 0〉 = 〈7cd + 7f8 , 7g 〉……….(2)

But,

〈7cd + 7f8 , 7g 〉 = 〈0, # + $ 〉 by (1)

Now,

〈0, # + $ 〉 = #〈0, 〉 + $〈0, 〉

= #〈7d , 7g 〉 + $〈78 , 7g 〉………..(3)
From (2) and (3), we have
〈#7d + $78 , 0〉 = #〈7d , 7g 〉 + $〈78 , 7g 〉.

Which Completes the proof.

Summary
 If is any subset of a Hilbert space , then the orthogonal complement of denoted by
, is defined as
={ ∈ ∶〈 , 〉=0∀ ∈ }
={ ∈ ∶ ⊥ }.
 = ={ ∈ ∶ 〈 , 〉 = 0, ∀ ∈ }
={ ∈ ⊥ }
 {0} =
 = {0}

Lovely Professional University 101

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

 If , be subsets of a Hilbert space , then we have

 ⊆ , that is any subset of is contained in its double orthogonal
complement .
 If ⊆ , then ⊆ .
 ∪ = ∩ and ∩ ⊇ ∪ .
 = .
 ∩ ⊆ {0}.
 is a closed subspace of .
 If is a closed linear subspace of a Hilbert space , then ∩ = {0}.
 If be a complete subspace of an inner product space ,. Then there is a non zero vector
0 ∈ , such that
0⊥
 If be a proper complete subspace of an inner product space ,. Then
,= ⨁

 If be a closed subspace of a Hilbert space . Then

= ⨁ .

 For any complete subspace of an inner product ,,

= .
 If and < are closed subspaces of a Hilbert space such that ⊥ <. Then + < is closed
subspace of .
 Let be a closed linear subspace of a Hilbert space . Then ∩ = {0}.
 Let be any closed subspace of a Hilbert space . Then
= ⊕ ( Projection Theorem)
 The line segment joining two given elements and of a space , is defined to be the set
of all 0 ∈ , of the form: 0 = B + 1 − B for every real number B such that 0 ≤ B ≤ 1.
 A subset of , is said to be convex if for every , ∈ , the line segment joining and
is contained in ,
i.e., 0 = B + 1 − B ∈ for every B, where 0 ≤ B ≤ 1.

 Every subspace - of , is convex, and the intersection of convex sets is a convex set.
 If be a non empty complete convex subset of an inner product
space , and ∈ ,\ . Then there is a unique ∈ such that
‖ − ‖ = 56789∈: ‖ − ;‖

That is, there is a unique ∈ which is closest to .

 Let be a Hilbert space . By ∗, we denote the conjugate space of (i.e. the set of all
continuous linear transformations of into O). The elements of ∗ are called continuous
linear functionals or briefly functional.
 Let be a fixed vector in a Hilbert space and let 78 be a function defined as
78 = 〈 , 〉 for every ∈ . Then 78 is a functional on and ‖ ‖ = R78 R.
 The mapping : → ∗ defined by = 78 where 78 = 〈 , 〉 for every ∈ is an
additive, one to one onto isometry but not linear.
 If is a Hilbert space , then ∗ is also Hilbert space with the inner product defined by
〈7d , 78 〉 = 〈 , 〉 ∀ , ∈

102 Lovely Professional University

Notes
Functional Analysis

Keywords
 Orthogonal complement
 Direct sum
 Convex set
 Conjugate space
 Closed subspace
 Hilbert space
 Inner product space
 Projection theorem

Self Assessment
1: Let be a Hilbert space and be a subspace of . Then
which of the following is false?
A. is a subspace of .
B. is a closed subspace of .
C. ∩ = {0}.
D. ∩ =

2: The distance between any two orthonormal vectors in an inner product space is:
A. 1
B. √2
C. 2
D. √5

3: Let , be an inner product space. Then the orthogonal complement of {0} is:
A. ,
B. {0}
C. ,{0}
D. ,

4: What is a convex set in a Hilbert space?

A. A set that contains only a single point.
B. A set where every line segment between two points in the set lies entirely within the set.
C. A set of orthogonal vectors.
D. A set that is closed under addition but not under scalar multiplication.

5: Every perfectly convex set is:

A. Closed
B. Open
C. Half open
D. Convex

6: A convex set in a Banach space need not be:

A. Hausdorff
B. Convex
C. Perfectly convex
D. Closed

Lovely Professional University 103

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

7: What is the orthogonal complement of a subspace?

A. The subspace itself.
B. The zero vector.
C. The set of all vectors orthogonal to the subspace.
D. None of the above.

8: In an inner product space, which of the following statement is true regarding the direct sum of
two subspaces j and k, denoted as j ⊕ k?
A. j ⊕ k is always equal to j + k.
B. Every vector in j ⊕ k can be uniquely expressed as the sum of a vector from j and a vector
from k.
C. j ⊕ k contain only the zero vector.
D. None of the above

9: Which property is satisfied by the intersection of any number of convex sets in a Hilbert space?
A. The intersection is always a convex set.
B. The intersection is always a non –convex set.
C. The intersection is always a singleton set.
D. The intersection is always empty.

10: Let , be subsets of a Hilbert space , then which of the following is true.
A. ∪ = ∩
B. ∩ ⊇ ∪ .
C. = .
D. All the above.

11: A subspace of a Hilbert space is closed if and only if:

A. = .
B. = .
C. ⊆ .
D. All the above.

12: If be any closed subspace of a Hilbert space . Then Which of the following is true?

A. = ∩ .
B. = ⊕ .
C. Both (A) and (B) are true.
D. None of the above.

13: Which of the following statement is true about conjugate space of a Hilbert Space?
A. It only contains real numbers.
B. It is always a finite-dimensional vector space.
C. It consists of continuous linear functionals on the Hilbert space.
D. None of the above.

14: If and < are closed subspaces of a Hilbert space such that ⊥ <, the which of the following
is true?
A. + < is closed subspace of .
B. M+N is a subspace of .

104 Lovely Professional University

Notes
Functional Analysis

C. Both (A) and (B) are true.

D. None of the above.

15: Let , be subsets of a Hilbert space , then which of the following is true?

A. ⊆ , that is any subset of is contained in its double orthogonal complement .

B. If ⊆ , then ⊆ .
C. Both (A) and (B).
D. None of the above.

16: If , be subsets of a Hilbert space , then which of the following is true?

A. ∩ ⊆ {0}.
B. is a closed subspace of .
C. ∩ = {0}.
D. All the above.

17: The mapping : → ∗ defined by = 78 where 78 = 〈 , 〉 for every ∈ is:

A. An additive mapping.
B. One to one
C. Not linear.
D. All the above.

Answers for Self Assessment

1 D 2 B 3 A 4 B 5 D

6 C 7 C 8 B 9 A 10 D

11 A 12 B 13 C 14 C 15 C

16 D 17 D

Review Questions
1. Define the orthogonal complement of a subset S in an inner product space.
2. Define a convex set in a Hilbert space.
3. Provide an example of a convex set that is not a closed set.
4. State the Projection Theorem for a Hilbert space.
5. What is meant by the “ conjugate ” or “dual” of a Hilbert space ? How is the dual space
constructed from the original Hilbert space?
6. Prove that If is a closed linear subspace of a Hilbert space , then ∩ = {0}.
7. Prove that If and < are closed subspaces of a Hilbert space such that ⊥ <. Then
+ < is closed subspace of .
8. Let , be subsets of a Hilbert space , then prove the following.
I. ⊆ , that is any subset of is contained in its double orthogonal
complement .
II. If ⊆ , then ⊆ .

Lovely Professional University 105

 Notes

Unit 09: Decomposition Theorems In Hibert Spaces

III. ∪ = ∩ and ∩ ⊇ ∪ .
IV. = .
V. ∩ ⊆ {0}.
9. Prove that If be a subset of a Hilbert space , then is a closed subspace of .
∗
10. Show that the mapping : → defined by = 78 where 78 = 〈 , 〉 for every
∈ is an additive, one to one onto isometry but not linear.

Further Readings
1. Introductory Functional Analysis With Applications By Erwin Kreyszig.
2. Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
3. J. B Conway, A Course In Functional Analysis.
4. C. Goffman G Pedrick, A First Course In Functional Analysis.
5. B.V. Limaya, Functional Analysis.

106 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 10: Riesz Representation Theorem and Operators on Hilbert

Spaces
CONTENTS
Objectives
Introduction
10.1 Riesz Representation Theorem
10.2 Hilbert Adjoint Operator
10.3 Self Adjoint Operators
10.4 Positive Operator
10.5 Normal Operators
10.6 Unitary and Isometric Operators
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Riesz representation theorem

 Hilbert adjoint operator
 Self adjoint operator
 Positive operator
 Normal operator
 Unitary operator
 Isometric operator

Introduction
In this chapter, we discuss about Riesz representation theorem. Further, we shall introduce the
operators on a Hilbert space like Hilbert adjoint operator, Self adjoint operator, positive operator,
normal operator, unitary operator and isometric operators.

10.1 Riesz Representation Theorem

Let be a Hilbert space and let be any arbitrary functional in ∗
, then there exists a unique
vector in such that = 〈 , 〉 for every ∈ and ‖ ‖ = ‖ ‖.
Proof. Let be the null space (kernel) of , that is
={ ∈ : = 0}

Lovely Professional University 107

Notes
Functional Analysis

Since is continuous as is functional, so by the continuity of , the null space of is closed

subspace of , as we know that the null space of a non-zero continuous linear operator is a closed
subspace .
If = , the = 0 as by definition of .
= 〈 , 0〉 for all ∈ .
If ≠ , then is a proper closed subspace of and so there exists a non-zero vector ∈
which is orthogonal to i,e. ⊥ .
So is not in , thus ≠ 0.

For any vector ∈ , the vector = − . is in ,

Because = − .

= −

=0
.
Also since ⊥ , so that ⊥ as ∈
⇒〈 , 〉=0

⇒〈 − . , 〉=0

⇒〈 , 〉−〈 , 〉=0

⇒〈 , 〉− 〈 , 〉=0

⇒ 〈 , 〉=〈 , 〉

⇒ = 〈 , 〉
〈 , 〉
!!!!!!!! !!!!!!!!
⇒ = 〈 , 〈!!!!!!!!!!〉 〉=〈 ,
〈 〉
〉.
, ,
!!!!!!!!
Let = 〈 〉
, then we have
,

= 〈 , 〉 for all ∈ .
To complete the proof, it remains to show that is unique.
For if suppose =〈 , "
〉 for all , then
〈 , 〉=〈 , "〉

⇒ 〈 , 〉−〈 , "〉 =0
⇒〈 , − "〉
= 0 for all ∈ .
In particular = − ", we get:
〈 − ", − "〉

⇒‖ − " ‖#
=0
⇒ − "
=0
⇒ = ".

Hence is unique .
Next we show that ‖ ‖ = ‖ ‖, we have
=〈 , 〉
So | | = |〈 , 〉|

108 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

≤ ‖ ‖‖ ‖ ( By Schwarz inequality)

And thus it follows that

‖ ‖ ≤ ‖ ‖ ( taking &'(‖ ‖)* over both sides)
Also
‖ ‖# = 〈 , 〉 =

≤| |
≤ ‖ ‖‖ ‖
⇒‖ ‖≤‖ ‖
Combining both the equations, we get
‖ ‖ = ‖ ‖.

10.2 Hilbert Adjoint Operator

Let +: * → # be a bounded linear operator, where * and # are Hilbert spaces. Then the Hilbert
adjoint operator + ∗ of + is the operator
+ ∗: # → *

Such that for all ∈ * and ∈ #,

〈+ , 〉 = 〈 , + ∗ 〉.
We first show that this definition makes sense, that is we prove that for a given + such a + ∗ does
exists.
Theorem. Show that the Hilbert adjoint operator + ∗ of + exists , is unique and is bounded linear
operator with norm
‖+ ∗ ‖ = ‖+‖.
Before proving this theorem, we first give the another statement of Riesz representation.
Theorem. Let *, # be Hilbert spaces and
ℎ: * × # →/
a bounded sesquilinear form. Then ℎ has a representation
ℎ , = 〈& , 〉
Where &: * → # is a bounded linear operator. & is uniquely determined by ℎ and has the norm
‖&‖ = ‖ℎ‖.

Proof of the Main Theorem

The formula
ℎ , = 〈 , + 〉………..(1)
defines a sesquilinear form on # × * because the inner product is sesquilinear and + is linear. In
fact, conjugate linearity of the form is seen from
ℎ ,0 * +2 # = 〈 ,+ 0 * +2 # 〉

= 〈 , 0+ * + 2+ #〉

= 0!〈 , + * 〉 + 2̅ 〈 , + #〉

= 0!ℎ , * + 2̅ ℎ , #

ℎ is bounded , by the Schwarz inequality

|ℎ , | = |〈 , + 〉|

Lovely Professional University 109

Notes
Functional Analysis

≤ ‖ ‖‖+ ‖
≤ ‖+‖‖ ‖‖ ‖.
This implies,
‖ℎ‖ ≤ ‖+‖.
Moreover, we have
‖ℎ‖ ≥ ‖+‖.

|〈 , + 〉| |〈+ , + 〉 |
‖ℎ‖ = sup ≥
8 ‖ ‖‖ ‖ ‖+ ‖‖ ‖
8

= ‖+‖
Combing the both equations, we get

‖ℎ‖ = ‖+‖.
The above theorem gives a Riesz representation for ℎ; writing + ∗ for &, we have
ℎ , = 〈+ ∗ , 〉, ………..(2)
and we know from the above theorem that + ∗ : # → * is a uniquely determined bounded linear
operator with norm
‖+ ∗ ‖ = ‖ℎ‖ = ‖+‖
Also ,
〈 , + 〉 = 〈+ ∗ , 〉 by comparing (1) and (2), so that we have 〈+ , 〉 = 〈 , + ∗ 〉 by taking conjugates,
and we now see + ∗ is the required operator.
For studying the properties of Hilbert adjoint opeartors , it will be convenient to make use of
following lemma.
Lemma. (Zero operator) . Let : and ; be inner product spaces and <: : → ; a bounded linear
operator . Then:

(a) < = 0 if and only if 〈< , 〉 = 0 for all ∈ : and ∈ ;.

(b) If <: : → :, where : is complex, and 〈< , 〉 = 0 for all ∈ :, then < = 0.
Proof. (a ) < = 0 means < = 0 for all ,
⇒ 〈< , 〉 = 〈0, 〉 = 0〈=, 〉 = 0.
Conversely suppose that 〈< , 〉 = 0 for all and ,
⇒ < = 0 for all .
So that, < = 0 by definition.
(b ) By assumption ,〈<>, >〉 = 0 for every > = 0 + ∈ :,
that is,
0 = 〈< 0 + ,0 + 〉

= |0|# 〈< , 〉 + 〈< , 〉 + 0〈< , 〉 + 0? 〈< , 〉.

The first two terms on the right are zero by assumption. 0 = 1 gives
〈< , 〉 + 〈< , 〉 = 0.
0 = A gives 0? = −A and
〈< , 〉 − 〈< , 〉 = 0.

110 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

By addition,
〈< , 〉 = 0 and < = 0 follows from (a).

Note: In part (b) of this lemma, it is essential that : be complex. Indeed , the conclusion may
not hold if : is real . A counterexample is a rotation < of the plane B # through a right angle. < is
linear, and < ⊥ , hence 〈< , 〉 = 0 for all ∈ B # , but < ≠ 0.
We now prove some general properties of Hilbert- adjoint operators which one uses quite
frequently in applying these operators.
Properties of Hilbert- adjoint Operators
Theorem. Show that the adjoint operator preserves addition, reverses the product and it is
conjugate linear. That is if + → + ∗ is the adjoint operator on 2 , then

(i) +* + +# ∗
= +*∗ + +#∗
(ii) +* +# ∗
= +#∗ +*∗
(iii) 0+ ∗
= 0? + ∗
Proof. For every , ∈ , we have
〈 , +* + +# ∗
〉 = 〈 +* + +# , 〉.
But
〈 +* + +# , 〉 = 〈+* + +# , 〉
= 〈+* , 〉 + 〈+# , 〉
= 〈 , +*∗ 〉 + 〈 , +#∗ 〉
= 〈 , +*∗ + +#∗ 〉
Hence,

〈 , +* + +# ∗
〉 = 〈 , +*∗ + +#∗ 〉.

From the uniqueness of the adjoint, we get

+* + +# ∗
= +*∗ + +#∗ .
(ii) For every , ∈ , we have 〈 , +* +# ∗
〉 = 〈 +* +# , 〉
= 〈+* +# , 〉.
But
〈+* +# , 〉 = 〈+# , +*∗ 〉 = 〈 , +#∗ +*∗ 〉
From the above two, we get
〈 , +* +# ∗ 〉 = 〈 , +#∗ +*∗ 〉 for all ∈ .

Therefore from uniqueness of adjoint

+* +# ∗ = +#∗ +*∗.
(iii) For every , ∈ , we have
〈 , 0+ ∗ 〉 = 〈 0+ , 〉 = 0〈+ , 〉.
But 0〈+ , 〉 = 0〈 , + ∗ 〉
= 〈 , 0? + ∗
〉.
Therefore , by the uniqueness of the adjoint , we have

0+ ∗
= 0? + ∗ .

Lovely Professional University 111

Notes
Functional Analysis

Theorem. Let be a Hilbert space. Then the adjoint operator + → + ∗ on 2 ( set of all bounded
linear transformations on into ) has the following properties:

(i) + ∗∗ = +
(ii) ‖+ ∗ ‖ = ‖+‖
(iii) ‖+ ∗ +‖ = ‖+‖#

Proof. For every , ∈ , we get

〈 , + ∗∗ 〉 = 〈 , + ∗ ∗

= 〈+ ∗ , 〉
But
〈+ ∗ , 〉 = 〈!!!!!!!!
, +∗ 〉
!!!!!!!!!
= 〈+ , 〉
= 〈 , + 〉.
From the above two, we get
〈 , + ∗∗ 〉 = 〈 , + 〉.
Which proves that
+ ∗∗ = + by the uniqueness of inverse.

(ii) For any vector ∈ , we have

‖+ ∗ ‖ ≤ ‖+‖‖ ‖.
Hence we get
‖C ∗ ‖
sup 8 ‖ ‖
≤ ‖+‖.

Using the definition of norm of the operator + ∗ , we get

‖+ ∗ ‖ ≤ ‖+‖………..(1)
Now applying (i) to the operator + ∗ , we get
‖ + ∗ ∗ ‖ ≤ ‖+ ∗ ‖
or

‖ + ∗∗ ‖ ≤ ‖+ ∗ ‖………..(2)
But by (i), we have
+ ∗∗ = +…………(3)
Using (3) in (2), we get
‖+‖ ≤ ‖+ ∗ ‖………..(4)
From (1) and (4), we get

‖+‖ = ‖+ ∗ ‖.

(iii) To prove ‖+ ∗ +‖ = ‖+‖# ,

112 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

Let us consider,
‖+ ∗ +‖ ≤ ‖+ ∗ ‖‖+‖………….(5)

By (ii) above, ‖+ ∗ ‖ = ‖+‖ so that we get from (5)

‖+ ∗ +‖ ≤ ‖+‖#…………(6)
To obtain reverse inequality, let us consider
‖+ ‖# ≤ 〈+ , + 〉
= 〈+ ∗ + , 〉. … … … . . 7
By using Schwarz inequality ,we have
〈+ ∗ + , 〉 ≤ ‖+ ∗ + ‖‖ ‖
≤ ‖+ ∗ +‖ ‖ ‖‖ ‖………..(8)
From (7) and (8) we get
‖+ ‖# ≤ ‖+ ∗ +‖ ‖ ‖# for every ∈ . ……….(9)
But
#
‖+ ‖
‖+‖# = sup F G
8 ‖ ‖

‖C ‖I
= sup H ‖ ‖I
J…………(10)
8

From (9), we get

‖+‖# ≤ ‖+ ∗ +‖………..(11)
Therefore, we have from (6) and (11),
‖+ ∗ +‖ = ‖+‖#
Taking + ∗ instead of +, we get as in the above
‖ + ∗ ∗ + ∗ ‖ = ‖+ ∗ ‖# .
Using (i) and (ii) in the above , we get
‖++ ∗ ‖ = ‖+‖#.

Note: From the properties of + ∗ as discussed above , we have the following corollary.
Corollary. If {+K } is a sequence of bounded linear operators on a Hilbert space and +K → +, then
+K∗ → + ∗ .
We have from the properties of + ∗
‖+K∗ − + ∗ ‖ = ‖ +K − + ∗ ‖ = ‖+K − +‖
Since +K → + as L → ∞, +K∗ → + ∗ follows from the above.

Note: The adjoint operator on 2 is one to one and onto. If + is non singular operator on ,
then + ∗ is also non-singular and + ∗ N* = + N* ∗ .

10.3 Self Adjoint Operators

The motivation for the introduction of the self adjoint operators is the properties of complex
numbers with conjugate mapping → ̅. This mapping → ̅ of the complex plane into itself
behaves like the adjoint operators .This operation → ̅ has all the properties of the adjoint
operators. As we know that complex number is real if and only if = ̅. Analogue of this
characterisation in 2 leads to the notion of self adjoint operators in Hilbert spaces.

Lovely Professional University 113

Notes
Functional Analysis

Definition. An operator + on a Hilbert space is said to be self adjoint operator if + ∗ = +. From

this definition we have the following simple operators.

(i) 0 and O are examples of self adjoint operators.

(ii) An operator + on is self adjoint operator , then 〈+ , 〉 = 〈 , + 〉 for every , ∈
and conversely.
If + ∗ is an adjoint opearator of + on , then we know from the definition
〈+ , 〉 = 〈 , + ∗ 〉 for all , ∈ .

If + is self adjoint then + ∗ = +, using this in above we get ,

〈+ , 〉 = 〈 , + 〉 for every , ∈ .
To prove the converse , let us assume 〈+ , 〉 = 〈 , + 〉 for all , ∈ …………(1)
We have to show that + is self adjoint . If + is the adjoint of +, then we have
∗

〈+ , 〉 = 〈 , + ∗ 〉……….(2)
From (1) and (2), we have

〈 , + 〉 = 〈 , +∗ 〉
Which gives ,
〈 , + − +∗ 〉 = 0 for all , ∈
Since ≠ 0, we have + − +∗ = 0 for all , ∈ , we have
+ = +∗
Proving that + is self adjoint.

(iii) For any + ∈ 2 , + + + ∗ and + ∗ + are self adjoint.

By the propert of Hilbert adjoint operators, we have
+ + +∗ ∗
= + ∗ + + ∗∗ = + ∗ + + = + + + ∗ so that we have
+ + +∗ ∗
= + + + ∗,
also
+ +
∗ ∗
= + ∗ + ∗∗ = + ∗ +, so that

+ ∗+ ∗ = + ∗ +.
Hence + + + ∗ and + ∗ + are self adjoint.
Theorem. If & and + are self adjoint opeartors on a Hilbert space , then their product &+ is self
adjoint if and only if they commute. That is &+ = +&.
Proof. Given & and + are self adjoint opeartors on a Hilbert space . Then & ∗ = &, + ∗ = +.
Let us assume that & and + commute, we will prove that &+ is self adjoint.
Now,
&+ ∗ = +∗ &∗
= +&
= &+
implies that
&+ ∗ = &+
Conversely, let us assume that &+ is self adjoint and we will show that &+ commute.

114 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

By hypothesis , we have
&+ ∗
= &+………..(1)
But
&+ ∗
= + ∗ & ∗ = +&, ………..(2)
( by properties of adjoint operators)
From (1) and (2), we have
&+ = +&.
Or in other words we can say that if &+ is self adjoint , then they commute.
Theorem. An operator + on a complex Hilbert space is self adjoint if and only if 〈+ , 〉 is real for
all .
Proof. Let us assume that + is self adjoint operator on .
i.e, + = + ∗ , then for every ∈ , we have
〈+ , 〉 = 〈 , + ∗ 〉

= 〈 ,+ 〉

= !!!!!!!!!
〈+ , 〉 .
Thus 〈+ , 〉 is equal to its own conjugate and is therefore real.
To prove the converse part, let us assume that 〈+ , 〉 is real for all ∈ and we will show that + is
self adjoint. Since 〈+ , 〉 is real for all ∈ , we have
〈+ , 〉 = !!!!!!!!!
〈+ , 〉

= 〈!!!!!!!!!!
, +∗ 〉

= 〈+ ∗ , 〉
Where + ∗ is the adjoint of + which exists for every ∈ , from the above we get
〈+ , 〉 − 〈+ ∗ , 〉 = 0 for all ∈ .
This gives 〈+ − + ∗ , 〉 for all ∈ .
Hence, we have
〈 + − +∗ , 〉 = 0 for all ∈ .
As we know that, if + is an operator on a Hilbert space , then 〈+ , 〉 = 0 for all ∈ if and only if
+ = 0.
Thus, we have
+ − +∗ = 0
or,
+ = + ∗.
Therefore the operator + is self adjoint.

10.4 Positive Operator

As we have seen in previous section that 〈+ , 〉 is real for self adjoint operators, we can introduce
the order relation among them and define positive operators by considering the real values which
the self adjoint operators take.
Definition. If & is the set of all self-adjoint operators, we can define an order relation denoted by ≤
on & as follows.

Lovely Professional University 115

Notes
Functional Analysis

If +* , +# ∈ &, then we write

+* ≤ +# if 〈+* , 〉 ≤ 〈+# , 〉 for all ∈ .
Definition. A self adjoint operator + on is said to be positive if + ≥ 0 in the order relation. This
means 〈+ , 〉 ≥ 0 ∀ ∈ .
From the definition, we have the following properties:
(i ) The identity operator O and the zero operator 0 are positive operators.
As we know that the identity operator O and the zero operator 0 are self adjoint .
Further,
〈O , 〉 = 〈 , 〉

= ‖ ‖#

≥ 0.
Also,
〈0 , 〉 = 〈0, 〉

= 0.
Hence O and 0 are positive operators.
(ii) For an arbitrary + on , then ++ ∗ and + ∗ + are positive operators.
First we note that ++ ∗ and + ∗ + are self-adjoint. Bu using properties of adjoint operators, we get
++ ∗ ∗
= +∗ ∗+∗

= + ∗∗ + ∗

= ++ ∗
Also,
+∗ + ∗
= +∗ +∗ ∗

= + ∗ + ∗∗
= +∗ +
Now we prove that they are positive ,
i.e, 〈++ ∗ , 〉 = 〈+ ∗ , + ∗ 〉

= ‖+ ∗ ‖#

≥ 0.
And
〈+ ∗ + , 〉 = 〈+ , + ∗∗ 〉

= 〈+ , + 〉

= ‖+ ‖#

≥ 0.

116 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

Hence ++ ∗ and + ∗ + are positive operators.

Note: If + is a positive operator on a Hilbert space , then O + + is non –singular.

Note: If two bounded self-adjoint linear operators & and + on a Hilbert space are positive
and commute , then their product &+ is positive.
`

10.5 Normal Operators

Definition. Let be a Hilbert space and let Q ∈ 2 and Q ∗ be the adjoint of Q. Then Q is said to
be normal operator if,
QQ ∗ = Q ∗ Q.
That is Q is said to be normal if it commutes with its adjoint.
From the definition of normal operator, we get the following properties.

(i) Every self adjoint operator is normal.

As since + is self adjoint, we have + ∗ = +.
Hence,
++ ∗ = + ∗ + is true so that + is normal operator.

Note: A normal operator need not be self adjoint.

Note: The limit Q of any convergent sequence {QR } of any normal operator is normal.
Theorem. If Q* and Q# are normal operators on a Hilbert space with property that either
commutes with the adjoint of the other then:

(i) Q* + Q# is normal .
(ii) Q* . Q# is normal .
Proof. Since Q* and Q# are normal, we get
Q* Q*∗ = Q*∗ Q* ...............(1)
and
Q# Q#∗ = Q#∗ Q# ...............(2)
From hypothesis either commutes with adjoint of the other.
So,

Q* Q#∗ = Q#∗ Q* ...............(3)

and
Q# Q*∗ = Q*∗ Q# ...............(4)
To (ST>U A , we have to show that
Q* + Q# Q* + Q# ∗
= Q* + Q# ∗
Q* + Q# . ...............(5)
Using the fact that adjoint operators preserves addition, we get
Q* + Q# Q* + Q# ∗ = Q* + Q# Q*∗ + Q#∗ .
= Q* Q*∗ + Q* Q#∗ + Q# Q*∗ + Q# Q#∗ . ...............(6)
V 'WALX 1 2 3 [L\ 4 AL 6 , =U XU_

Lovely Professional University 117

Notes
Functional Analysis

= Q*∗ Q* + Q#∗ Q* + Q*∗ Q# + Q#∗ Q#

= Q*∗ Q* + Q# + Q#∗ Q* + Q#

= Q*∗ + Q#∗ Q* + Q#

= Q* + Q# ∗ Q* + Q#
Hence,
Q* + Q# Q* + Q# ∗ = Q* + Q# ∗ Q* + Q# .
Therefore Q* + Q# is normal.
Now to prove (ii), we have to prove
Q* Q# Q* Q# ∗
= Q* Q# ∗
Q* Q#
Now we have
Q* Q# Q* Q# ∗ = Q* Q# Q#∗ Q*∗

= Q* Q# Q#∗ Q*∗ .
But,

Q* Q# Q#∗ Q*∗ = Q* Q#∗ Q# Q*∗

= Q* Q#∗ Q# Q*∗

= Q#∗ Q* Q*∗ Q# .
But,
Q#∗ Q* Q*∗ Q# = Q#∗ Q* Q*∗ Q#

= Q#∗ Q*∗ Q* Q#

= Q* Q# ∗
Q* Q# .
Thus,
Q* Q# Q* Q# ∗
= Q* Q# ∗
Q* Q# .
So that,
Q* Q# is Normal.

Note: An operator Q on a Hilbert space is normal if and only if

‖Q ∗ ‖ = ‖Q ‖ for every ∈ .

Note: If Q is a normal operator on , then

‖Q # ‖ = ‖Q‖# .

10.6 Unitary and Isometric Operators

A special type of normal operators which are of considerable interest in applied mathematics is
that of unitary operators.

118 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

Definition. An operator ` on a Hilbert space is said to be unitary if

`` ∗ = ` ∗ ` = O.

From the definition of unitary operator, we note down the following

(i ) If ` is unitary, then it is normal ,
(ii) `∗ = ` N* .
Before characterizing an unitary operator on a Hilbert space, we first define isometric operator
on H.
Definition. An operator + on is said to be isometric if
‖+ − + ‖ = ‖ − ‖ ∀ , ∈
Since + is linear , the condition is equivalent to ‖+ ‖ = ‖ ‖ ∀ , ∈ .

Let {U* , U# , … } be an orthonormal basis for a separable Hilbert space

Example:
and + ∈ 2 be defined as
+ * U* + # U# +⋯ = * U* + # U# + ⋯ where ={ K}

Then
c
‖+ ‖# = b| K|
# = ‖ ‖#
K)*

so + is an isometric operator. The operator + defined is called the right shift operator given by
+UK = UKd* .

Note: If + is an operator on a Hilbert space , then the following conditions are equivalent to
one another .

(i) + ∗+ = O
(ii) 〈+ , + 〉 = 〈 , 〉∀ , ∈ .
(iii) ‖+ ‖ = ‖ ‖ ∀ ∈ .
Theorem. An operator + on a Hilbert space is unitary if and only if it is an isomorphism of
onto itself.
Proof. Let + be an unitary operator on . Then from the definition of the unitary operator, it is
invertible. Therefore it is onto . Further,
++ ∗ = O
Hence,
‖+ ‖ = ‖ ‖ ∀ ∈
This proves that + is an isometric isomorphism of onto itself.
Now to prove the converse let us assume that + is an isometric isomorphism of onto itself.
Then + is one-one and onto . Therefore + N* exists. From our assumption
‖+ ‖ = ‖ ‖ ∀ ∈ . ……….(1)
By the above note we have + ∗ + = O.
Hence ,
+ ∗ + + N* = O+ N* ,
Which gives,
+ ∗ ++ N* = + N*

Lovely Professional University 119

Notes
Functional Analysis

so that
+ ∗ O = + N* .
Thus
+ ∗ = + N*

Premultiplying this by + we have

++ ∗ = ++ N*
So that,
++ ∗ = O
Now postmultiplying by +, we have

+ ∗ + = + N* +
+ ∗ + = O.
Hence + ∗ + = ++ ∗ = O.
Which proves that + is unitary.

Note: If + is an unitary operator on , then ‖+ ‖ = 1.

For an unitary operator, we have
‖+ ‖ = ‖ ‖ ,
So that,
‖+‖ = sup‖ ‖)* ‖+ ‖ = sup‖ ‖)* ‖ ‖ = 1.

Note: The range ` of a unitary operator ` is a closed subspace of .

Summary
 If is a Hilbert space and be any arbitrary functional in ∗ , then there exists a
unique vector in such that = 〈 , 〉 for every ∈ and ‖ ‖ = ‖ ‖.
 If +: * → # be a bounded linear operator, where * and # are Hilbert spaces. Then the
Hilbert adjoint operator +∗ of + is the operator
+ ∗: # → *

Such that for all ∈ * and ∈ #,

〈+ , 〉 = 〈 , + ∗ 〉.

 The Hilbert adjoint operator + ∗ of + exists , is unique and is bounded linear operator
with norm
‖+ ∗ ‖ = ‖+‖.
 If : and ; be inner product spaces and <: : → ; a bounded linear operator . Then:
(i) < = 0 if and only if 〈< , 〉 = 0 for all ∈ : and ∈ ;.
(ii) If <: : → :, where : is complex, and 〈< , 〉 = 0 for all ∈ :, then < = 0.
 The adjoint operator preserves addition, reverses the product and it is conjugate
linear. That is if + → + ∗ is the adjoint operator on 2 , then
a. +* + +# ∗
= +*∗ + +#∗
b. +* +# ∗
= +#∗ +*∗
c. 0+ ∗
= 0? + ∗ .

120 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

 Let be a Hilbert space. Then the adjoint operator + → + ∗ on 2 has the following
properties:
a. + ∗∗ = +
b. ‖+ ∗ ‖ = ‖+‖
c. ‖+ ∗ +‖ = ‖+‖#.
 If {+K } is a sequence of bounded linear operators on a Hilbert space and +K → +,
then +K∗ →+ .
∗

 The adjoint operator on 2 is one to one and onto. If + is non singular operator on
, then + ∗ is also non-singular and + ∗ N* = + N* ∗ .
 An operator + on a Hilbert space is said to be self adjoint operator if + ∗ = +.
 0 and O are examples of self adjoint operators.
 If an operator + on is self adjoint operator , then 〈+ , 〉 = 〈 , + 〉 for every , ∈
and conversely.
 If & and + are self adjoint opeartors on a Hilbert space , then their product &+ is self
adjoint if and only if they commute. That is &+ = +&.
 An operator + on a complex Hilbert space is self adjoint if and only if 〈+ , 〉 is real
for all .
 A self adjoint operator + on is said to be positive if + ≥ 0 in the order relation. This
means 〈+ , 〉 ≥ 0 ∀ ∈ .
 The identity operator O and the zero operator 0 are positive operators.
 For an arbitrary + on , then ++ ∗ and + ∗ + are positive operators.
 If + is a positive operator on a Hilbert space , then O + + is non –singular.
 If two bounded self-adjoint linear operators & and + on a Hilbert space are positive
and commute , then their product &+ is positive.
 If be a Hilbert space and let Q ∈ 2 and Q ∗ be the adjoint of Q. Then Q is said to
be normal operator if,
QQ ∗ = Q ∗ Q.
 Every self adjoint operator is normal.
 A normal operator need not be self adjoint.
 The limit Q of any convergent sequence {QR } of any normal operator is normal.

 If Q* and Q# are normal operators on a Hilbert space with property that either
commutes with the adjoint of the other then:
a. Q* + Q# is normal .
b. Q* . Q# is normal

 An opeator Q on a Hilbert space is normal if and only if

‖Q ∗ ‖ = ‖Q ‖ for every ∈ .

 If Q is a normal operator on , then

‖Q # ‖ = ‖Q‖# .
 An operator ` on a Hilbert space is said to be unitary if
`` ∗ = ` ∗ ` = O.

 If ` is unitary, then it is normal and ` ∗ = ` N* .

Lovely Professional University 121

Notes
Functional Analysis

 An operator + on is said to be isometric if

‖+ − + ‖ = ‖ − ‖ ∀ , ∈ .
 An operator + on a Hilbert space is unitary if and only if it is an isomorphism of onto
itself.
 If + is an unitary operator on , then ‖+ ‖ = 1.
 The range ` of a unitary operator ` is a closed subspace of .

Keywords
 Hilbert Space
 Hilbert adjoint operator
 Self adjoint operator
 Positive operator
 Normal operator
 Unitary operator
 Isometric operator
 Schwarz inequality
 Linear operator
 Zero operator
 Bounded linear transformation

Self Assessment
1: Which of the following properties is true for Hilbert adjoint operator in a Hilbert space?
A. The adjoint operator preserves addition.
B. The adjoint operator reverses the product.
C. The adjoint operator is conjugate linear.
D. All of the above.

2: Let be a Hilbert space. Then the adjoint operator + → + ∗ on 2 ( set of all bounded linear
transformations on into ) satisfies which of the following properties:

A. + ∗∗ = +.
B. ‖+ ∗ ‖ = ‖+‖.
C. ‖+ ∗ +‖ = ‖+‖#.
D. All of the above .

3: An operator + on a Hilbert space is said to be self adjoint operator if:

A. +∗ = +.
B. + ∗∗ = +.
C. + ∗ < +.
D. + ∗ > +.

4: Which of the following is/are self adjoint operator/operators.

A. The zero operator.
B. The identity operator.
C. Both (A) and (B).
D. None of (A) and (B).

122 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

5: If & and + are self adjoint opeartors on a Hilbert space , then their product &+ is self adjoint
if and only if .
A. & + + = + + &.
B. &+ = +&.
C. &+ > +&.
D. None of the above.

6: An operator + on a complex Hilbert space is self adjoint if and only if 〈+ , 〉 is:

A. Real for all .

B. Real for some .
C. Real for some > 0.
D. All of the above are true.

7: if + is an operator on a Hilbert space , then 〈+ , 〉 = 0 for all ∈ if and only if:

A. + = 0.
B. + > 0.
C. + < 0.
D. + ≤ 0.

8: Which of the following is/are positive operator/operators.

A. The zero operator.
B. The identity operator.
C. Both (A) and (B).
D. None of (A) and (B).

9: Let be a Hilbert space and let Q ∈ 2 and Q ∗ be the adjoint of Q. Then Q is said to be
normal operator if:

A. QQ ∗ = Q.
B. QQ ∗ = Q ∗ Q.
C. QQ ∗ = Q ∗ .
D. None of the above.

10: Which of the following is/are true?

A. Every self adjoint operator is normal.

B. A normal operator need not be self adjoint.
C. The limit Q of any convergent sequence {QR } of any normal operator is normal.
D. All of the above are true.

11: Which of the following is/are true about Unitary operator?

A. `` ∗ = ` ∗ ` = O.
B. If ` is unitary, then it is normal.
C. ` ∗ = ` N* .
D. All of the above are true.

Lovely Professional University 123

Notes
Functional Analysis

12: Which of the following is/are true about Unitary operator?

A. If + is an unitary operator on , then ‖+ ‖ = 1.

B. The range ` of a unitary operator ` is a closed subspace of .
C. Both (A) and (B).
D. None of (A) and (B).

13: Which of the following is/are true about Normal operator?

A. An operator Q on a Hilbert space is normal if and only if ‖Q ∗ ‖ = ‖Q ‖ for every ∈ .

B. If Q is a normal operator on , then ‖Q # ‖ = ‖Q‖# .
C. If Q* and Q# are normal operators on a Hilbert space with property that either commutes
with the adjoint of the other then:
Q* + Q# is normal and Q* . Q# is normal.
D. All of the above are true.

14: An operator + on a Hilbert space is said to be isometric if:

A. ‖+ − + ‖ ≥ ‖ − ‖ ∀ , ∈ .
B. ‖+ − + ‖ = ‖ − ‖ ∀ , ∈ .
C. ‖+ − + ‖ < ‖ − ‖ ∀ , ∈ .
D. ‖+ − + ‖ ≤ ‖ − ‖ ∀ , ∈ .

15: Which of the following is/are true?

A. The adjoint operator on 2 is one to one.
B. The adjoint operator on 2 is onto.
C. Both (A) and (B).
D. None of (A) and (B)

Answers for Self Assessment

1 D 2 D 3 A 4 C 5 B

6 A 7 A 8 C 9 B 10 D

11 D 12 C 13 D 14 B 15 C

Review Questions
1. State Riesz representation theorem.
2. What is Hilbert adjoint operator.
3. Define Self adjoint operator.
4. Show that the Hilbert adjoint operator + ∗ of + exists , is unique and is bounded linear
operator with norm
‖+ ∗ ‖ = ‖+‖.

5. Show that the adjoint operator preserves addition, reverses the product and it is conjugate
linear.
6. Define self adjoint operator and give examples.

124 Lovely Professional University

 Notes

Unit 10: Riesz Representation Theorems and Operators on Hibert Spaces

7. Show that If & and + are self adjoint opeartors on a Hilbert space , then their product &+
is self adjoint if and only if they commute. That is &+ = +&.
8. Show that normal operator need not be self adjoint.
9. If Q* and Q# are normal operators on a Hilbert space with property that either
commutes with the adjoint of the other then:
I. Q* + Q# is normal .
II. Q* . Q# is normal .
10. Define Unitary operator and isometric operator.

Further Readings
1. Introductory Functional Analysis With Applications By Erwin Kreyszig.
2. Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
3. J. B Conway, A Course In Functional Analysis.
4. C. Goffman G Pedrick, A First Course In Functional Analysis.
5. B.V. Limaya, Functional Analysis.

Lovely Professional University 125

 Notes

Unit 11: Unitary and Normal Operators

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 11: Unitary and Normal Operators

CONTENTS
Objectives
Introduction
11.1 Self Adjoint Operators
11.2 Positive Operator
11.3 Normal Operators
11.4 Unitary and Isometric Operators
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Self adjoint operator and its properties

 Positive operator and its Properties
 Normal and Unitary operator and their properties.
 Isometric operator

Introduction
In this chapter, we discuss about Self adjoint operator and its properties. Further, positive operators
and its properties are discussed. Finally we discuss about normal operator, unitary operators and
isometric operator.

11.1 Self Adjoint Operators

The motivation for the introduction of the self adjoint operators is the properties of complex
numbers with conjugate mapping → ̅. This mapping → ̅ of the complex plane into itself
behaves like the adjoint operators .This operation → ̅ has all the properties of the adjoint
operators. As we know that complex number is real if and only if = ̅. Analogue of this
characterisation in ( ) leads to the notion of self adjoint operators in Hilbert spaces.
∗
Definition. An operator on a Hilbert space is said to be self adjoint operator if = . From
this definition we have the following simple operators.

(i) 0 and are examples of self adjoint operators.

(ii) An operator on is self adjoint operator , then 〈 , 〉=〈 , 〉 for every , ∈ and
conversely.
∗
If is an adjoint opearator of on , then we know from the definition
〈 , 〉=〈 , ∗ 〉 for all , ∈ .

126 Lovely Professional University

Notes
Functional Analysis

If is self adjoint then ∗ = , using this in above we get ,

〈 , 〉=〈 , 〉 for every , ∈ .

To prove the converse , let us assume 〈 , 〉=〈 , 〉 for all , ∈ …………(1)
We have to show that is self adjoint . If ∗ is the adjoint of , then we have
〈 , 〉=〈 , ∗
〉……….(2)
From (1) and (2), we have

〈 , 〉=〈 , ∗ 〉
Which gives ,
〈 ,( − ∗) 〉 = 0 for all , ∈
Since ≠ 0, we have ( − ∗) = 0 for all , ∈ , we have
∗
=
Proving that is self adjoint.

(iii) For any ∈ ( ), + ∗ and ∗ are self adjoint.

By the propert of Hilbert adjoint operators, we have
( + ∗ )∗
= ∗
+ ∗∗
= ∗
+ = + ∗
so that we have
( + ∗ )∗ = + ∗,

also
∗
( )∗ = ∗ ∗∗
= ∗
, so that

∗
( )∗ = ∗
.
∗ ∗
Hence + and are self adjoint.
Theorem. If and are self adjoint opeartors on a Hilbert space , then their product is self
adjoint if and only if they commute. That is = .
∗ ∗
Proof. Given and are self adjoint opeartors on a Hilbert space . Then = , = .
Let us assume that and commute, we will prove that is self adjoint.
Now,
( )∗ = ∗ ∗

=
=
implies that
( )∗ =
Conversely, let us assume that is self adjoint and we will show that commute.
By hypothesis , we have
( )∗ = ………..(1)
But
( )∗ = ∗ ∗ = , ………..(2)
( by properties of adjoint operators)
From (1) and (2), we have
= .
Or in other words we can say that if is self adjoint , then they commute.

Lovely Professional University 127

 Notes

Unit 11: Unitary and Normal Operators

Theorem. An operator on a complex Hilbert space is self adjoint if and only if 〈 , 〉 is real for
all .
Proof. Let us assume that is self adjoint operator on .
∗
i.e, = , then for every ∈ , we have
〈 , 〉=〈 , ∗ 〉

=〈 , 〉

=〈 , 〉.
Thus 〈 , 〉 is equal to its own conjugate and is therefore real.
To prove the converse part, let us assume that 〈 , 〉 is real for all ∈ and we will show that is
self adjoint. Since 〈 , 〉 is real for all ∈ , we have
〈 , 〉=〈 , 〉

=〈 , ∗ 〉

=〈 ∗ , 〉
Where ∗ is the adjoint of which exists for every ∈ , from the above we get
〈 , 〉−〈 ∗
, 〉 = 0 for all ∈ .
∗
This gives 〈 − , 〉 for all ∈ .
Hence, we have
〈( − ∗)
, 〉 = 0 for all ∈ .
As we know that, if is an operator on a Hilbert space , then 〈 , 〉 = 0 for all ∈ if and only if
= 0.
Thus, we have
− ∗
=0
or,
∗.
=
Therefore the operator is self adjoint.

11.2 Positive Operator

As we have seen in previous section that 〈 , 〉 is real for self adjoint operators, we can introduce
the order relation among them and define positive operators by considering the real values which
the self adjoint operators take.
Definition. If is the set of all self-adjoint operators, we can define an order relation denoted by ≤
on as follows.
If , ∈ , then we write
≤ if 〈 , 〉≤〈 , 〉 for all ∈ .
Definition. A self adjoint operator on is said to be positive if ≥ 0 in the order relation. This
means 〈 , 〉 ≥ 0 ∀ ∈ .
From the definition, we have the following properties:
(i ) The identity operator and the zero operator 0 are positive operators.
As we know that the identity operator and the zero operator 0 are self adjoint .
Further,
〈 , 〉=〈 , 〉

128 Lovely Professional University

Notes
Functional Analysis

=‖ ‖

≥ 0.
Also,
〈0 , 〉 = 〈0, 〉

= 0.
Hence and 0 are positive operators.
(ii) For an arbitrary on , then ∗ ∗
and are positive operators.
∗ ∗
First we note that and are self-adjoint. Bu using properties of adjoint operators, we get
( ∗ )∗ =( ∗ )∗ ∗

= ∗∗ ∗

∗
=
Also,
( ∗ )∗ = ∗ ( ∗ )∗

= ∗ ∗∗

= ∗

Now we prove that they are positive ,

i.e, 〈 ∗
, 〉=〈 ∗
, ∗ 〉

=‖ ∗ ‖

≥ 0.
And
〈 ∗ , 〉=〈 , ∗∗ 〉

=〈 , 〉

=‖ ‖

≥ 0.

∗ ∗
Hence and are positive operators.

Note. If is a positive operator on a Hilbert space , then + is non –singular.

Note. If two bounded self-adjoint linear operators and on a Hilbert space are positive
and commute , then their product is positive. `

Lovely Professional University 129

 Notes

Unit 11: Unitary and Normal Operators

11.3 Normal Operators

∗
Definition. Let be a Hilbert space and let ∈ ( ) and be the adjoint of . Then is said to
be normal operator if,
∗ ∗
= .
That is is said to be normal if it commutes with its adjoint.
From the definition of normal operator, we get the following properties.

(i) Every self adjoint operator is normal.

As since is self adjoint, we have ∗ = .
Hence,
∗ = ∗ is true so that is normal operator.

Note: A normal operator need not be self adjoint.

Note: The limit of any convergent sequence ! "# of any normal operator is normal.
Theorem. If and are normal operators on a Hilbert space with property that either
commutes with the adjoint of the other then:

(i) + is normal .
(ii) . is normal .
Proof. Since and are normal, we get
∗ ∗
= ...............(1)
and
∗ ∗
= ...............(2)
From hypothesis either commutes with adjoint of the other.
So,

∗ ∗
= ...............(3)
and
∗ ∗
= ...............(4)
To $%&'( ()), we have to show that
( + )( + )∗ = ( + )∗ ( + ). ...............(5)
Using the fact that adjoint operators preserves addition, we get
( + )( + )∗ = ( + )( ∗
+ ∗ ).

∗ ∗ ∗ ∗
= + + + . ...............(6)
* +,)-. (1) (2) (3) 2-3 (4) )- (6), 6( .(7

∗ ∗ ∗ ∗
= + + +

∗( )+ ∗
= + ( + )

=( ∗
+ ∗
)( + )

=( + )∗ ( + )
Hence,

130 Lovely Professional University

Notes
Functional Analysis

( + )( + )∗ = ( + )∗ ( + ).
Therefore + is normal.
Now to prove (ii), we have to prove
( )( )∗ = ( )∗ ( )
Now we have
( )( )∗ = ∗ ∗

∗ ∗
= ( ) .
But,

∗ ∗ ∗ ∗
( ) = ( )

∗ ∗
=( )( )

∗ ∗
=( )( ).
But,
∗ ∗ ∗ ∗
( )( )= ( )

∗ ∗
=( )( )

=( )∗ ( ).
Thus,
( )( )∗ = ( )∗ ( ).
So that,
is Normal.

Note: An operator on a Hilbert space is normal if and only if

‖ ∗ ‖=‖ ‖ for every ∈ .

Note: If is a normal operator on , then

‖ ‖=‖ ‖ .

11.4 Unitary and Isometric Operators

A special type of normal operators which are of considerable interest in applied mathematics is
that of unitary operators.
Definition. An operator 8 on a Hilbert space is said to be unitary if
88 ∗ = 8 ∗ 8 = .

From the definition of unitary operator, we note down the following

(i ) If 8 is unitary, then it is normal ,
(ii) 8∗ = 8 9 .
Before characterizing an unitary operator on a Hilbert space, we first define isometric operator
on H.
Definition. An operator on is said to be isometric if

Lovely Professional University 131

 Notes

Unit 11: Unitary and Normal Operators

‖ − ‖=‖ − ‖∀ , ∈
Since is linear , the condition is equivalent to ‖ ‖=‖ ‖∀ , ∈ .

Example: Let !( , ( , … # be an orthonormal basis for a separable Hilbert space

and ∈ ( ) be defined as
( ( + ( +⋯) = ( + ( + ⋯ where =! <#

Then
?
‖ ‖ = =| <| =‖ ‖
<@

so is an isometric operator. The operator defined is called the right shift operator given by
(< = (<A .

Note: If is an operator on a Hilbert space , then the following conditions are equivalent to
one another .

(i) ∗ =
(ii) 〈 , 〉 = 〈 , 〉∀ , ∈ .
(iii) ‖ ‖=‖ ‖∀ ∈ .
Theorem. An operator on a Hilbert space is unitary if and only if it is an isomorphism of
onto itself.
Proof. Let be an unitary operator on . Then from the definition of the unitary operator, it is
invertible. Therefore it is onto . Further,
∗ =
Hence,
‖ ‖=‖ ‖∀ ∈
This proves that is an isometric isomorphism of onto itself.
Now to prove the converse let us assume that is an isometric isomorphism of onto itself.
9
Then is one-one and onto . Therefore exists. From our assumption
‖ ‖=‖ ‖∀ ∈ . ……….(1)
By the above note we have ∗ = .
Hence ,
∗ 9 9
( ) = ,
Which gives,
∗( 9 )= 9

so that
∗ 9
= .
Thus
∗ = 9

Premultiplying this by we have

∗ = 9

So that,
∗ =

132 Lovely Professional University

Notes
Functional Analysis

Now postmultiplying by , we have

∗ = 9

∗ = .
∗ ∗
Hence = = .
Which proves that is unitary.

Note: If is an unitary operator on , then ‖ ‖ = 1.

For an unitary operator, we have
‖ ‖=‖ ‖,
So that,
‖ ‖ = sup‖E‖@ ‖ ‖ = sup‖E‖@ ‖ ‖ = 1.

Note: The range 8( ) of a unitary operator 8 is a closed subspace of .

Summary
 An operator on a Hilbert space is said to be self adjoint operator if ∗ = .
 0 and are examples of self adjoint operators.
 If an operator on is self adjoint operator , then 〈 , 〉=〈 , 〉 for every , ∈ and
conversely.
 If and are self adjoint opeartors on a Hilbert space , then their product is self adjoint
if and only if they commute. That is = .
 An operator on a complex Hilbert space is self adjoint if and only if 〈 , 〉 is real for all
.
 A self adjoint operator on is said to be positive if ≥ 0 in the order relation. This means
〈 , 〉≥0∀ ∈ .
 The identity operator and the zero operator 0 are positive operators.
∗ ∗
 For an arbitrary on , then and are positive operators.
 If is a positive operator on a Hilbert space , then + is non –singular.
 If two bounded self-adjoint linear operators and on a Hilbert space are positive and
commute , then their product is positive.
∗
 If be a Hilbert space and let ∈ ( ) and be the adjoint of . Then is said to be
normal operator if,
∗ = ∗ .
 Every self adjoint operator is normal.
 A normal operator need not be self adjoint.
 The limit of any convergent sequence ! "# of any normal operator is normal.
 If and are normal operators on a Hilbert space with property that either
commutes with the adjoint of the other then:
a. + is normal .
b. . is normal

 An opeator on a Hilbert space is normal if and only if

‖ ∗ ‖=‖ ‖ for every ∈ .

Lovely Professional University 133

 Notes

Unit 11: Unitary and Normal Operators

 If is a normal operator on , then

‖ ‖=‖ ‖ .
 An operator 8 on a Hilbert space is said to be unitary if
88 ∗ = 8 ∗ 8 = .

 If 8 is unitary, then it is normal and 8 ∗ = 8 9 .

 An operator on is said to be isometric if
‖ − ‖=‖ − ‖∀ , ∈ .
 An operator on a Hilbert space is unitary if and only if it is an isomorphism of onto
itself.
 If is an unitary operator on , then ‖ ‖ = 1.
 The range 8( ) of a unitary operator 8 is a closed subspace of .

Keywords
 Hilbert Space
 Self adjoint operator
 Positive operator
 Normal operator
 Unitary operator
 Isometric operator
 Linear operator
 Zero operator
 Bounded linear transformation

Self Assessment
1: An operator on is said to be isometric if

‖ − ‖=‖ − ‖∀ , ∈ .
A. True
B. False
2: An operator 8 on a Hilbert space is said to be unitary if
88 ∗ = 8∗ 8 = .

A. True
B. False

3: An operator on a Hilbert space is said to be self adjoint operator if:

A. ∗ = .
B. ∗∗ = .
C. ∗ < .
D. ∗ > .

4: Which of the following is/are self adjoint operator/operators.

A. The zero operator.
B. The identity operator.
C. Both (A) and (B).
D. None of (A) and (B).

134 Lovely Professional University

Notes
Functional Analysis

5: If and are self adjoint opeartors on a Hilbert space , then their product is self adjoint
if and only if .
A. + = + .
B. = .
C. > .
D. None of the above.

6: An operator on a complex Hilbert space is self adjoint if and only if 〈 , 〉 is:

A. Real for all .

B. Real for some .
C. Real for some > 0.
D. All of the above are true.

7: if is an operator on a Hilbert space , then 〈 , 〉 = 0 for all ∈ if and only if:

A. = 0.
B. > 0.
C. < 0.
D. ≤ 0.

8: Which of the following is/are positive operator/operators.

A. The zero operator.
B. The identity operator.
C. Both (A) and (B).
D. None of (A) and (B).

∗
9: Let be a Hilbert space and let ∈ ( ) and be the adjoint of . Then is said to be
normal operator if:

A. ∗ = .
B. ∗ = ∗ .
C. ∗ = ∗.

D. None of the above.

10: Which of the following is/are true?

A. Every self adjoint operator is normal.

B. A normal operator need not be self adjoint.
C. The limit of any convergent sequence ! "# of any normal operator is normal.
D. All of the above are true.

11: Which of the following is/are true about Unitary operator?

A. 88 ∗ = 8 ∗ 8 = .
B. If 8 is unitary, then it is normal.
C. 8 ∗ = 8 9 .
D. All of the above are true.

Lovely Professional University 135

 Notes

Unit 11: Unitary and Normal Operators

12: Which of the following is/are true about Unitary operator?

A. If is an unitary operator on , then ‖ ‖ = 1.

B. The range 8( ) of a unitary operator 8 is a closed subspace of .
C. Both (A) and (B).
D. None of (A) and (B).

13: Which of the following is/are true about Normal operator?

A. An operator on a Hilbert space is normal if and only if ‖ ∗ ‖=‖ ‖ for every ∈ .

B. If is a normal operator on , then ‖ ‖=‖ ‖ .
C. If and are normal operators on a Hilbert space with property that either commutes
with the adjoint of the other then:
+ is normal and . is normal.
D. All of the above are true.

14: An operator on a Hilbert space is said to be isometric if:

A. ‖ − ‖≥‖ − ‖∀ , ∈ .
B. ‖ − ‖=‖ − ‖∀ , ∈ .
C. ‖ − ‖<‖ − ‖∀ , ∈ .
D. ‖ − ‖≤‖ − ‖∀ , ∈ .

15: Which of the following is/are true?

A. The adjoint operator on ( ) is one to one.
B. The adjoint operator on ( ) is onto.
C. Both (A) and (B).
D. None of (A) and (B).

Answers for Self Assessment

1. A 2. A 3. A 4. C 5. B

6. A 7. A 8. C 9. B 10. D

11. D 12. C 13. D 14. B 15. C

Review Questions
1. Define self adjoint operator and give examples.
2. Show that If and are self adjoint opeartors on a Hilbert space , then their product
is self adjoint if and only if they commute. That is = .
3. Show that normal operator need not be self adjoint.
4. If and are normal operators on a Hilbert space with property that either
commutes with the adjoint of the other then:
I. + is normal .
II. . is normal .
5. Define Unitary operator and isometric operator.

136 Lovely Professional University

Notes
Functional Analysis

Further Readings
1. Introductory Functional Analysis With Applications By Erwin Kreyszig.
2. Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
3. J. B Conway, A Course In Functional Analysis.
4. C. Goffman G Pedrick, A First Course In Functional Analysis.
5. B.V. Limaya, Functional Analysis.

Lovely Professional University 137

 Notes

Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 12 : Reflexivity of Hilbert Space and Orthogonal Projection

CONTENTS
Objectives
Introduction
12.1 Reflexivity of Hilbert space
12.2 Orthogonal Projection
12.3 Further properties of Projections
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Reflexivity of Hilbert space

 Orthogonal Projection
 Properties of Orthogonal Projection.

Introduction
In this chapter , we discuss about reflexivity of Hilbert space. Further, we discuss about Orthogonal
projection and properties of orthogonal projection.

12.1 Reflexivity of Hilbert space

Recall that a normed space is reflexive if there is an isometric isomorphism between and its
second dual . In the following theorem we establish the reflexivity of Hilbert Spaces. Thus if is
a Hilbert space and its second dual , then it will be shown that there is a bijective linear
mapping between and such that
‖ ‖= ∀ ∈ .
.
Theorem. Show that every Hilbert space is reflexive.
Proof. Let be a Hilbert space , its dual and an arbitrary element of . As by Riesz
Representation theorem that every bounded linear functional on is of the form given by
= 〈 , 〉, ∈ ………..(1)
And that the mapping : → given by
= , ∈ ……….(2)
is an isometric isomorphism between and . Now define a mapping : → defined by
= , ∈ ………..(3)

138 Lovely Professional University

Notes
Functional Analysis

We show that, provided is a Hilbert space, is surjective mapping from to . For this we
have to prove that, given any element ℎ of , there is a ∈ such that
= .
For this consider the mapping : → defined by (2), : → as follows:
For any ∈ , ∈ . Also, for any ℎ ∈ ,ℎ is in .
So we put
= , ∈ .
Let ∈ . Then under , we let mapped onto , where is an element of the dual space of
defined by :
=ℎ ………..(4)

We first show that is linear.

For , ! ∈ ,
+ ! =ℎ + !

=ℎ + !

=ℎ +ℎ 2

= + ! …………(5)
And,
$ =ℎ $

=ℎ $

= $ℎ
%& '& ()*+, -./ 0'*/-1
=$ ………..(6)
Now ,
| | = |ℎ |

= |ℎ |

≤ ‖ℎ‖‖ ‖

≤ ‖ℎ‖‖ ‖ , ‖ ‖=‖ ‖ = ‖ ‖,
so that ,
‖ ‖ ≤ ‖ℎ‖,
Where ‖ℎ‖ is finite because ℎ ∈ . Hence is a bounded linear functional in . By Riesz
represemtation theoremthere is a unique ∈ such that
=〈 , 〉=ℎ
From (4) or equivalentely
ℎ = 〈 , 〉.
But then , from (3) and the definition of we have:

Lovely Professional University 139

 Notes

Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection

= =〈 , 〉=ℎ
for all ∈ , ∈ . Hence
= ℎ………..(7)
Thus is surjective.
To see that is injective, suppose that, for , ! ∈ ,
=
!

Then

4 5 = = ! =
!

For all ∈ . Hence , by (1)

〈 , 〉=〈 , !〉

For all ∈ because each is given by such an expression.

Therefore,
= !.

To show that is linear, let , ! ∈ . Then

6
=

7
= !

And
68 7
= + !

= + !

= 6
+ 7

= 6
+ 7
……….(8)
while
9 = $

=$

=$
For all ∈ . Hence

68 7
= 6
+ 7

and
$ =$ .
Lastly, to see that is an isometry , let ∈ .
Then by Riesz representation theorem,
‖ ‖=‖ ‖
Where is in . Again by Riesz representation theorem, any ∈ is given by
= 〈 , 〉, ∈ , ∈ and = .
‖ ‖=‖ ‖= Sup | |
‖ ‖= >∈?

140 Lovely Professional University

Notes
Functional Analysis

= Sup‖@‖= A∈?B
| |

= .
Hence is an isometric isomorphism between and . So is reflexive.

12.2 Orthogonal Projection

Let C be a closed subspace of a Hilbert space . Then we know that
= C ⊕ CE
………..(1)
E
= + , ∈ C, ∈ C .
Since the sum is direct, is unique for any given ∈ . Hence (1) defines a linear operator
F: →
⟶ =F
F is called an orthogonal projection or projection on . More specifically, F is called the projection
of onto C. Hence a linear operator F: → is a projection on if there is a closed subspace C of
such that C is the range of F and C E is the null space of F and F/C is the identity operator on C.
From (1), we can now write
= +

= F + I−F .
This shows that the projection of onto CE is I − F.
There is another characterization of a projection on , which is sometimes used as a definition.
Theorem 1 . A bounded linear operator F: → on a Hilbert space is a projection if and only if F
is self adjoint and idempotent ( .ℎ-. '&, F ! = F).
Proof. Suppose that F is a projection on and denote F by C. Then F ! = F because for every
∈ and F = ∈ C we have
F! = F = =F .
Furthermore, let = + and ! = ! + !, where , ! ∈ C and , ! ∈ C E .Then
〈 , !〉 = 〈 !, 〉=0
because C ⊥ C E , and self adjointness of F is seen from
〈F , !〉 =〈 , ! + !〉 =〈 , !〉 =〈 + , !〉 = 〈 , F !〉
.
Conversely suppose that F ! = F = F ∗ and denote F by C. Then for every ∈
=F + I−F .
Orthogonality C = F ⊥ I−F follows from
〈F , I − F N〉 = 〈 , F I − F N〉 = 〈 , FN − F ! N〉 = 〈 , 0〉 = 0.
C is the null space O I − F of I − F , because C ⊂ O I − F can be seen from
I−F F =F − F! =0
and C ⊃ O I − F follows if we note that I − F = 0 implies = F . Hence C is closed . Finally,
F\C is the identity operator on C since writing = F , we have F = F ! = F = .

Theorem. For any projection F on a Hilbert space ,

Lovely Professional University 141

 Notes

Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection

〈F , 〉 = ‖F ‖! ………..(2)
F S 0……………(3)
‖F‖ ≤ 1; ‖F‖ = 1 if F V W0X………..(4)
Proof. (2) and (3) follows from
〈F , 〉 = 〈F ! , 〉 = 〈F , F 〉 = ‖F ‖! S 0
.
By the Schwarz inequality
‖F ‖! = 〈F , 〉 ≤ ‖F ‖‖ ‖
‖YZ‖
So that ‖Z‖
≤ 1 for every V 0, and ‖F‖ ≤ 1.
‖YZ‖
Also ‖Z‖
= 1 if ∈F and V 0.

This proves (5).

Note: Every projection is linear. For if , ! ∈ , then

= + , ∈ C, ∈ CE

! = ! + !, ! ∈ C, ∈ CE
So that,
F + ! =F + ! + + !

= + !

=F +F !

Also for any $ ∈ ,

F $ =F $ +$
=$

= $F ∀ ∈

Note: The product of projections need not be a projection .

Note: The Product of two bounded self adjoint linear operators [ and \ on a Hilbert space
is self adjoint if and only if the operators commute,
[\ = \[. (Already proved in chapter 10)
Theorem. ( Product of projections)
In connection with product (composites ) of projections on a Hilbert space , the following two
statements hold.

(i) F = F F! is a projection on if and only if the projections F and F! commute, that is ,

F F! = F! F . Then F projects onto C = C ∩ C! , where C^ = F^ .
(ii) Two closed subspaces C and _ of are orthogonal if and only if the corresponding
projections satisfy F` Fa = 0.
Proof. (i) Suppose that F F! = F! F . Then F is self adjoint , by above note. F is idempotent since
F ! = F F! F F!
= F ! F!!

142 Lovely Professional University

Notes
Functional Analysis

= F F!
= F.
Hence F is a projection by theorem 1 above , and for every ∈ we have
F = F F! = F! F .
Since F projects onto C ,we must have F F! ∈C.
Similarly,
F! F ∈ C! .
Together, F ∈ C ∩ C! . Since ∈ was arbitrary, this shows that F projects into C = C ∩ C! .
Actually, F projects onto C. Indeed, if ∈ C, then ∈C, ∈ C! and
F = F F! = F = .
Conversely, if F = F F! is a projection defined on , then F is self adjoint by theorem 1 above and
F F! = F! F follows by above note.
Proof of (ii) . if C ⊥ _ then C ∩ _ = W0X
and
F` Fa = 0 ∀ ∈ by part (i), so that F` Fa = 0.
Conversely, if F` Fa = 0, then for every ∈ C and N ∈ _ we obtain
〈 , N〉 = 〈F` , Fa N〉
= 〈 , F` Fa N〉
= 〈 , N〉
= 0.
Hence C ⊥ _.
Theorem (Sum of projections). Let F and F! be projections on a Hilbert space . Then

(i) The sum F = F + F! is a projection on if and only

C =F and C! = F! are orthogonal .
(ii) If F = F + F! is a projection , F projects onto C = C ⊕ C! .
Proof. If F = F + F! is a projection, F = F!, by theorem 1 above, we have
!
F + F! = F + F!
= F ! + F F! + F! F + F!! .
By theorem 1 above , we have
F! = F
and
F!! = F! ,

Therefore,
F F! + F! F = 0 …….(5)
Multiplying by F! from the left, we obtain
F! F F! + F! F = 0 …….(6)

Multiplying this by F! from the right , we obtain

2F! F F! = 0,
so that F! F = 0 by (6) and C ⊥ C! by (ii) part of above theorem.

Lovely Professional University 143

 Notes

Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection

Conversely, if C ⊥ C! , then F F! = F! F = 0 again by (ii) part of above theorem.

This yields (5), which implies F ! = F. Since F and F! are self adjoint, so is
F = F + F! . Hence F is a projection by Theorem 1.
Proof of (ii). We determine the closed subspace C ⊂ onto which F projects. Since
F = F + F! , for every ∈ we have
=F =F + F! .
Here F ∈ C and F! ∈ C! . Hence ∈ C ⊕ C! , so that C ⊂ C ⊕ C! .
We show that C ⊃ C ⊕ C! .
Let N ∈ C ⊕ C! be arbitrary. Then
N= + !. Here, ∈ C and ! ∈ C! ,
Apply F and using C ⊥ C! , we thus obtain
FN = F + ! + F! + !

=F + F!
= + !

= N.
Hence N ∈ C and C ⊃ C ⊕ C! .
Thus,
C = C ⊕ C! .

12.3 Further properties of Projections

We now discuss some further properties of Projections.

1) Let F and F! be projections defined on a Hilbert space . Denote by

C =F and C! = F! the subspaces onto which is projected by F and F! , and let O F and
O F! be the null spaces of these projections . Then the following conditions are equivalent:

a. F! F = F F! = F
b. C ⊂ C!
c. O F ⊃ O F!
d. ‖F ‖ ≤ ‖F! ‖ ∀ ∈
e. F ≤ F! .
2) Let F and F! be projections defined on a Hilbert space . Then
(i) The difference F = F! − F is a projection on if and only
if C ⊂ C! , where C^ = F^ .
(ii) If F = F! − F is a projection, F projects onto C, where C
is the orthogonal complement of C in C! .
From these two properties of projection we can now derive a basic result about the convergence of
a monotonic increasing sequence of projections.

3) (Monotone increasing sequence). Let WFc X be a monotone increasing sequence of projections Fc

defined on the Hilbert space . Then
(i) WFc X is strongly operator convergent, say Fc → F for every ∈ ,
and the limit operator F is a projection defined on .
(ii) F projects onto

144 Lovely Professional University

Notes
Functional Analysis

F = d Fc .
c=

(iii) F has the null space

e

O F = f O Fc .
c=

Notes: Two projections F and g on a Hilbert space are said to be orthogonal if

Fg = 0.

Notes: If F is the projection on the closed linear subspace C of , then ∈ C if and only if
F = .

Notes: If F is the projection on the closed linear subspace C of , then F = . if and only if
‖F ‖ = ‖ ‖.

Notes: If F is the projection on a Hilbert space , then

(i) F is a positive operator on .

(ii) 0≤F≤1
(iii) ‖F ‖ ≤ ‖ ‖ for every ∈
(iv) ‖F‖ ≤ ‖1‖.

Notes: A projection on whose range and null spaces are orthogonal is called perpendicular
projection.

Notes: If F is the projection on a closed linear subspace C of if and only if I − F is a

projection on hE .

Summary
 Every Hilbert space is reflexive.
 If C be a closed subspace of a Hilbert space . Then we know that
= C ⊕ CE.
 A bounded linear operator F: → on a Hilbert space is a projection if and only if F is
self adjoint and idempotent.
 Every projection is linear.
 The product of projections need not be a projection .
 The Product of two bounded self adjoint linear operators [ and \ on a Hilbert space
is self adjoint if and only if the operators commute.
 The product of projections F = F F! is a projection on if and only if the projections F
and F! commute, that is , F F! = F! F . Then F projects onto C = C ∩ C! , where C^ =
F^ .
 Two closed subspaces C and _ of are orthogonal if and only if the corresponding
projections satisfy F` Fa = 0.
 If F and F! be projections on a Hilbert space . Then
(i) The sum F = F + F! is a projection on if and only

Lovely Professional University 145

 Notes

Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection

C =F and C! = F! are orthogonal .

(ii) If F = F + F! is a projection , F projects onto C = C ⊕ C! .

 Two projections F and g on a Hilbert space are said to be orthogonal if
Fg = 0.

 If F is the projection on the closed linear subspace C of , then ∈ C if and only if

F = .
 If F is the projection on the closed linear subspace C of , then F = . if and only if
‖F ‖ = ‖ ‖.
 If F is the projection on a Hilbert space , then
(i) F is a positive operator on .
(ii) 0≤F≤1
(iii) ‖F ‖ ≤ ‖ ‖ for every ∈
(iv) ‖F‖ ≤ ‖1‖.
 A projection on whose range and null spaces are orthogonal is called perpendicular
projection.
 If F is the projection on a closed linear subspace C of if and only if I − F is a
E.
projection on h

Keywords
 Hilbert space
 Isometric isomorphism
 Reflexivity
 Dual
 Linear mapping
 Projection
 Orthogonal projection

Self Assessment
1: Which of the following statements is/are true?
A. Every Hilbert space is reflexive.
B. Every Banach space is reflexive.
C. Both (A) and (B).
D. None of (A) and (B).

2: Which of the following statements is/are true?

A. A bounded linear operator F: → on a Hilbert space is a projection if and only if F is
self adjoint and idempotent.
B. Every projection is linear.
C. Both (A) and (B).
D. None of (A) and (B).

3: For any projection F on a Hilbert space ,which of the following is/are true?
A. 〈F , 〉 = ‖F ‖!.
B. F S 0.

146 Lovely Professional University

Notes
Functional Analysis

C. ‖F‖ ≤ 1 ; ‖F‖ = 1 if F V W0X.

D. All of the above are true.

4: Which of the following statement is/are true?

A. The product of projection is always a projection.
B. F = F F! is a projection on if and only if the projections F and F! commute, that is ,
F F! = F! F .
C. The product of projection is never a projection.
D. None of the above.

5: Which of the following statement is/are true?

A. Two closed subspaces C and _ of are orthogonal if and only if the corresponding
projections satisfy F` Fa = 0.
B. Two closed subspaces C and _ of are orthogonal if and only if the corresponding
projections satisfy F` Fa > 0.
C. Two closed subspaces C and _ of are orthogonal if and only if the corresponding
projections satisfy F` Fa < 0.
D. All of the above are true.

6: Let F and F! be projections on a Hilbert space . Then Which of the following statement
is/are true?

A. The sum F = F + F! is a projection on if and only if C = F and C! = F! are

orthogonal .
B. If F = F + F! is a projection if F projects onto C = C ⊕ C! .
C. Both (A) and (B) are true.
D. None of the above.

7: Two projections F and g on a Hilbert space are said to be orthogonal if:

A. Fg = 1.
B. Fg. 1.
C. Fg = 0.
D. None of the above.

8: If F is the projection on the closed linear subspace C of , then F = . if and only if:

A. ‖F ‖ = .
B. ‖F ‖ = ‖ ‖.
C. ‖F ‖ = 1.
D. None of the above.

9: If F is the projection on a Hilbert space , then which of the following is/are true?

A. F is a positive operator on .
B. 0 ≤ F ≤ 1
C. ‖F‖ ≤ ‖1‖.
D. All of the above are true.
10: Which of the following statement is/are true?

Lovely Professional University 147

 Notes

Unit 12: Reflexivity of Hilbert Space and Orthogonal Projection

A. A projection on whose range and null spaces are orthogonal is called perpendicular
projection.
B. If F is the projection on a closed linear subspace C of if and only if I − F is a projection
E
on h .
C. If F is the projection on the closed linear subspace C of , then ∈ C if and only if
F = .
D. All of the above are true.

11. If F is the projection on a closed linear subspace C of if and only if:

A. I − F is a projection on hE .
B. I − F is a projection on hEE .
C. I − F is a projection on h.
D. None of the above .

12. Let F and F! be projections defined on a Hilbert space . Then

A. The difference F = F! − F is a projection on if and only if C ⊂ C! , where C^ = F^ .

B. If F = F! − F is a projection, if F projects onto C, where C is the orthogonal complement
of C in C! .
C. Both (A) and (B) are true.
D. None of the above.

Answers for Self Assessment

1. A 2. C 3. D 4. B 5. A

6. C 7. C 8. B 9. D 10. D

11. A 12. C

Review Questions
1. Show that every Hilbert space is reflexive.
2. Show that a bounded linear operator F: → on a Hilbert space is a projection if and
only if F is self adjoint and idempotent.
3. Show that every projection is linear.
4. Show that two closed subspaces C and _ of are orthogonal if and only if the
corresponding projections satisfy F` Fa = 0.
5. Show that product F = F F! is a projection on if and only if the projections F and F!
commute, that is , F F! = F! F .

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.
 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.

148 Lovely Professional University

Notes
Functional Analysis

 C. Goffman G Pedrick, A First Course In Functional Analysis.

 B.V. Limaya, Functional Analysis.

Lovely Professional University 149

 Notes

Unit 13: Spectral Theory of Linear Operators in Normal Spaces

Dr. Arshad Ahmad Khan, Lovely Professional University

Unit 13: Spectral Theory of Linear Operators in Normal Spaces

CONTENTS
Objectives
Introduction
13.1 Spectrum of an Operator
13.2 Spectral Properties of Bounded Self-Adjoint Linear Operators
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Spectrum of an operator
 Spectral properties of bounded linear operators

Introduction
In this chapter, we discuss about Spectrum of an operator. Further, we discuss about spectral
properties of bounded self-adjoint linear operator.

13.1 Spectrum of an Operator

The generalization of the matrix eigenvalue theory leads to the spectral theory of operators on a
Banach space or Hilbert space. Before defining the spectrum of an operator, we first recall some
definitions.

Definition. Let be an operator on a Hilbert space . Then a scalar is called an

eigenvalue of if there exists a non zero vector in such that
.

Note: Eigenvalue is also called characteristic value, proper value or spectral value.

Definition. If is an eigenvalue of , then any non zero vector in such that

is called an eigenvector of .

Note: Eigen vector is also called characteristic vector, proper vector or spectral vector.

Definition. The eigenvectors corresponding to eigenvalue and the zero vector form a
vector subspace, which is called the eigenspace of corresponding to eigenvalue .
From the definition of eigenvalues and eigenvectors, we have the following properties.

150 Lovely Professional University

Notes
Functional Analysis

I. If is an eigen vector of corresponding to eigenvalue and is any nonzero scalar,

then is also an eigenvector of corresponding to same eigen value.
Since is an eigenvector of corresponding to the eigen value and . Since 0,
we have
0.
Hence (I) follows from

,
Which gives,
.

Note: Thus (I) tells us that corresponding to single eigenvalue there may correspond
more than one eigenvector.

Note: If is an eigenvector of , then cannot correspond to more than one eigenvalue of

.

Note: If the Hilbert space has no non-zero vectors, then cannot have any eigenvectors
and hence the whole theory reduces to triviality. So we shall assume throughtout this chapter
0.

Spectrum of an operator
Definition. The set of all eigenvalues of is called spectrum of and is denoted by . Its
complement in the complex plane is called resolvent set of .

Example: For a two dimensional Hilbert space , let , be a basis and be an

operator on given by the matrix

………..(1)

If is given by and , find the spectrum of .

Solution. Using the matrix of the operator , we have

,
So that,
0 and 1

,
So that,
1 and 0
Hence the matrix representation of is
0 1
1 0

Lovely Professional University 151

 Notes

Unit 13: Spectral Theory of Linear Operators in Normal Spaces

For this matrix, the eigenvalues are given by the characteristic equation
1
0
1
⇒ 1 0

⇒ 1

⇒ ,
So that,
.

Note: An operator on a finite dimensional Hilbert space is singular if and only if there
exists a non zero vector in such that 0.

Theorem. If is an operator on a finite dimensional Hilbert space, then the following statements
are true.

(i) is singular if and only if 0 ∈ .

" "
(ii) If is non -singular, then ∈ if and only if ∈ .
"
(iii) If is non singular, then .
(iv) If ∈ and if # is a polynomial , then # ∈ $# %.
Proof. (i) We know that is singular if and only if there exists a non-zero vector ∈ such that
0.
That is, 0 .
Hence is singular if and only if 0 is the eigenvalue of .
That is 0 ∈ .
"
(ii) Let be non-singular and ∈ . Hence 0 by (i) so that exist . Since is an eigen
value of , so there exixts a non –zero vector ∈ such that
.
"
Premultiplying by we get
" " ),
Which gives,
" for 0.
&
"
Hence ∈ .
"
(iii) Let ' . Then we find ' (.
Now ' ( " ( "

"
( .
Hence,
"
det ' ( det ( .
But,
det ( " det (
This proves that det ' ( det ( .
Thus is an eigen value of if and only if det ( 0.
Hence,

152 Lovely Professional University

Notes
Functional Analysis

det ( 0 if and only if det ' ( 0.

This proves that ' and have the same eigenvalues so that
" .
(iv) If ∈ , is an eigen value of . Then there exists a non-zero vector such that
.
Hence,

.
Hence if is an eigenvalue of , then is an eigenvalue of . Continuing in this way, we see that
if is an eigen value of , then , is an eigenvalue of , for any positive integer -.
Let # . /0 / . / . ⋯ /2 . 2 , where /0 , / , / , … , /2 are scalars. Then

4# 5 /0 ( / / ⋯ /2 2

2
/0 / ⋯ /2

2
4/0 / ⋯ /2 5 .
2
Hence # /0 / ⋯ /2 is an eigen value of # . Thus if ∈ , then # ∈
$# %.

Note: An operator on a Hilbert space need not necessarily posses an eigenvalue as

illustrated by the following Example.

Example: Consider the Hilbert space 6 and on 6 defined by

, ,…, , 0, , ,… .
If is an eigenvalue of T, then there exists a non zero vector , ,…, , such that

Which gives,
0, , ,… , ,…, , ,
which implies,
0, , , ," …
By hypothesis , ∈ 6 is non zero vector so that , 0 for any -.
Hence 0 implies 0 and implies 0 contracdicting that is non-zero vector.
Hence cannot have eigen values.
Theorem. (Spectral mapping theorem for polynomials )
Let be an operator on a complex Banach space and let 7 be a polynomial. Then
$7 % 7$ % 7 : ∈ .

Example. Find the spectrum of the idempotent operator on a Banach space.

Since is idempotent operator, then

Lovely Professional University 153

 Notes

Unit 13: Spectral Theory of Linear Operators in Normal Spaces

or
0.
Let 7 . Then 7 0 by hypothesis.
Hence
7$ %

1 0
So that 1 9: 0
Hence
0,1 .

13.2 Spectral Properties of Bounded Self-Adjoint Linear Operators

Throughout this section we shall consider bounded linear operators which are defined on a
complex Hilbert space and map into itself. Furthermore, these operations will be self-adjoint.
A bounded self-adjoint linear operator may not have eigenvalues, but if has eigenvalues, the
Following basic facts can readily be established.
Theorem. Let : → be a bounded self-adjoint linear operator on a complex Hilbert space .
Then

i. All the eigen values of (if they exists) are real.

ii. Eigenvectors corresponding to numerically different eigenvalues of are orthogonal.
Proof. (i) Let be an eigenvalue of and a corresponding eigenvector. Then 0 and .
Using the self-adjointness of , we have
〈 , 〉 〈 , 〉 〈 , 〉

〈 , 〉

〈 , 〉

̅〈 , 〉 .

Here 〈 , 〉 ‖ ‖ 0, and division by 〈 , 〉 gives ̅.

Hence is real.
Proof of (ii). Let and @ be eigen values of , and let and A be corresponding eigenvectors .
Then

And
A @A,
Since is self-adjoint and @ is real,
〈 , A〉 〈 , A〉 〈 , A〉

〈 , A〉

〈 , @A〉

154 Lovely Professional University

Notes
Functional Analysis

@〈 , A〉 .
Since @, we must have
〈 , A〉 0,
Which shows that and A are orthogonal.

Note (Resolvent set) Let : → be a bounded self adjoint linear operator on a complex
Hilbert space . Then a number belongs to the resolvent set of if and only if there exists
a B C 0 such that for every ∈ ,
‖ & ‖ D B‖ ‖………..(1)

Theorem. (Spectrum) The spectrum of a bounded self-adjoint linear operator : → on a

complex Hilbert space is real.
Proof. By above note , we show that / E , E ∈ F with E 0 must belong to , so that
, ⊂ F.
For every 0 in we have
〈 & , 〉 〈 , 〉 〈 , 〉
and, since 〈 , 〉 and 〈 , 〉 are real,
〈HHHHHHHHHH
& , 〉 〈 , 〉 ̅〈 , 〉.

Here ̅ E. By subtraction,
〈HHHHHHHHHH
& , 〉 〈 & , 〉 ̅ 〈 , 〉

2 E‖ ‖ .
The left side is 2 (J 〈 & , 〉, where (J denotes the imaginary part, the latter cannot exceed the
absolute value, so that, dividing by 2, taking absolute values and applying the Schwarz inequality,
we obtain
|E|‖ ‖ |(J 〈 & , 〉| L | 〈 & , 〉| L ‖ & ‖‖ ‖.
Division by ‖ ‖ 0 gives |E|‖ ‖ L ‖ & ‖.
If E 0, then ∈ by above note . Hence for ∈ , we must have E 0, that is, is real.

Summary
 If be an operator on a Hilbert space . Then a scalar is called an eigenvalue
of if there exists a non zero vector in such that
.
 Eigenvalue is also called characteristic value, proper value or spectral value.
 If is an eigenvalue of , then any non zero vector in such that
is called an eigenvector of .
 Eigen vector is also called characteristic vector, proper vector or spectral vector.
 The eigenvectors corresponding to eigenvalue and the zero vector form a vector
subspace, which is called the eigenspace of corresponding to eigenvalue .
 If is an eigen vector of corresponding to eigenvalue and is any nonzero scalar,
then is also an eigenvector of corresponding to same eigen value.
 Corresponding to single eigenvalue there may correspond more than one eigenvector.
 If is an eigenvector of , then cannot correspond to more than one eigenvalue of .

Lovely Professional University 155

 Notes

Unit 13: Spectral Theory of Linear Operators in Normal Spaces

 The set of all eigenvalues of is called spectrum of and is denoted by . Its complement
in the complex plane is called resolvent set of .
 An operator on a finite dimensional Hilbert space is singular if and only if there exists
a non zero vector in such that 0.
 If is an operator on a finite dimensional Hilbert space, then the following statements are
true.
a) is singular if and only if 0 ∈ .
" "
b) If is non -singular, then ∈ if and only if ∈ .
"
c) If is non singular, then .
d) If ∈ and if # is a polynomial , then # ∈ $# %.
 An operator on a Hilbert space need not necessarily posses an eigenvalue .
 If be an operator on a complex Banach space and let 7 be a polynomial. Then
$7 % 7$ % 7 : ∈ .
 Let : → be a bounded self-adjoint linear operator on a complex Hilbert space . Then
a) All the eigen values of (if they exists) are real.
b) Eigenvectors corresponding to numerically different eigenvalues of are orthogonal.
 If : → be a bounded self adjoint linear operator on a complex Hilbert space . Then
a number belongs to the resolvent set of if and only if there exists a B C 0 such
that for every ∈ ,
‖ & ‖ D B‖ ‖
 The spectrum of a bounded self-adjoint linear operator : → on a complex
Hilbert space is real.
 If is an arbitrary operator on a finite dimensional Hilbert space , then the spectrum of
namely is a finite subset of the complex plane and the number of points in
does not exceed the dimension - of .
 An operator on a finite dimensional Hilbert space is singular if and only if there exists
a non-zero vector ∈ such that 0.

Keywords
 Spectrum of an operator
 Bounded linear operator
 Eigen values
 Eigen vectors
 Eigen space
 Closed subspace
 Charteristic equation
 Hilbert space
 Banach space
 Idempotent operator
 Self adjoint operator

Self Assessment
1: Let be an operator on a Hilbert space . Then a scalar is called the ………..of if
there exists a non zero vector in such that .

156 Lovely Professional University

Notes
Functional Analysis

A. Eigenvalue
B. Proper value
C. Characteristic value
D. All of the above.

2: If is an eigenvalue of , then any non zero vector in such that

is called the ……….of .

A. Eigenvector
B. Proper vector
C. Both (A) and (B)
D. None of (A) and (B)
3: If is an eigen vector of corresponding to eigenvalue and is any nonzero scalar, then is
also an eigenvector of corresponding to same eigen value.

A. True
B. False

4: If the Hilbert space has no non-zero vectors, then the operator cannot have any eigenvectors.
A. True
B. False

5: The set of all eigenvalues of an operator is called……………. of .

A. Eigenvector
B. Proper vector
C. Spectrum
D. None of the above.

6: An operator on a finite dimensional Hilbert space is …………….if and only if there exists a
non zero vector in such that 0.
A. Non-singular
B. Singular
C. Regular
D. None of the above.

7: If is an operator on a finite dimensional Hilbert space, then which of the following statements
is/are true.

A. is singular if and only if 0 ∈ , where is the spectrum of .

" "
B. If is non -singular, then ∈ if and only if ∈ .
"
C. If is non singular, then .
D. All of the above.

8: An operator on a Hilbert space need not necessarily posses an eigenvalue.

A. True
B. False

9: Which of the following is the spectrum of the idempotent operator on a Banach space.

A. 0, 1 .

Lovely Professional University 157

 Notes

Unit 13: Spectral Theory of Linear Operators in Normal Spaces

B. 0,0 .
C. 1,1 .
D. 0,1 .

10: Let : → be a bounded self-adjoint linear operator on a complex Hilbert space .

Then which of the following statement is/are true.

A. All the eigen values of (if they exists) are real.

B. Eigenvectors corresponding to numerically different eigenvalues of are orthogonal.
C. Both (A) and (B)
D. None of (A) and (B)

11: The spectrum of a bounded self-adjoint linear operator : → on a complex Hilbert

space is real.
A. True
B. False

12: The eigenvectors corresponding to eigenvalue and the zero vector form a vector
subspace, which is called the eigenspace of corresponding to eigenvalue .
A. True
B. False

Answers for Self Assessment

1. D 2. C 3. A 4. A 5. C

6. B 7. D 8. A 9. D 10. C

11. A 12. A

Review Questions
1. Define Spectrum of an operator on a finite dimensional Hilbert space.
2. Define the resolvent set of an operator.
3. Show that an operator on a finite dimensional Hilbert space is singular if and only if there
exists a non zero vector in such that 0.
4. Find the spectrum of an idempotent operator on a Banach space.
5. Show that if : → be a bounded self- adjoint linear operator on a complex Hilbert space
. Then all the eigen values of , if they exists are real.
6. Define eigen values and eigen vectors of an operator.
7. Define eigenspace of an operator.
8. Show that eigenspace of an operator on a Hilbert space is a non zero closed linear Subspace of
.

Further Readings

158 Lovely Professional University

Notes
Functional Analysis

 Introductory Functional Analysis With Applications By Erwin Kreyszig.

 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.

 B.V. Limaya, Functional Analysis

Lovely Professional University 159

 Notes

Dr. Arshad Ahmad Khan, Lovely Professional University Unit 14: Spectrum of Normal Operators

Unit 14 : Spectrum Of Normal Operators

CONTENTS
Objectives
Introduction
14.1 Spectrum of Normal Operator
14.2 Spectral Resolution
14.3 Non-Emptiness of the Spectrum
Summary
Keywords
Self Assessment
Answers for Self Assessment
Review Questions
Further Readings

Objectives
After studying this unit, you will be able to understand:

 Spectrum of Normal Operator

 Spectral Resolution and its Properties
 Non-emptiness of the Spectrum

Introduction
In this chapter, we discuss about spectrum of Normal Operators. Further, we discuss about
spectral resolution and some of its important properties. Finally, we discuss nin-emptiness of the
spectrum.

14.1 Spectrum of Normal Operator

Below we shall give some properties of the spectra of a normal operator.

value iff is an eigen vector of ∗ with eigenvalue ̅.

Theorem. If is a normal operator on a Hilbert space , then is an eigen vector of with eigen

Proof. Since is normal operator on ,therefore − is also normal operator on where is any
scalar.
Now,

− ∗
= ∗
− ∗

= ∗ − .
Since − is normal, we know that an operator on a Hilbert space is normal iff
‖ ∗ ‖=‖ ‖ for every , therefore we have

160 Lovely Professional University

Notes
Functional Analysis

‖ − ‖=‖ − ∗ ‖∀ ∈
⟺‖ − ‖= ∗ − ∀ ∈

⟺‖ − ‖= ∗ − ∀ ∈ ………..(1)

From (1), we conclude that

− = 0 if and only if ∗ − =0
∗

eigen value .
Therefore is an eigen vector of with eigen value if and only if it is an eigenvector of with

Theorem. If is a normal operator on a Hilbert space , then the eigenspaces of are pairwise
orthogonal.
Proof. Let , be eigenspaces of a normal operator on corresponding to the distinct
eigenvalues and .
Then to prove that ⊥ . Let be any vector in and be any vector in .
Then ,
= and = .
We have,
〈 , 〉=〈 , 〉

=〈 , 〉

=〈 , ∗ 〉

=〈 , 〉

= 〈 , 〉.
Therefore,
− 〈 , 〉=0
Implies,
〈 , 〉 = 0 as ≠
⇒ ⊥
Thus,
⊥ ∀ ∈ and ∀ ∈ .
Hence,
⊥ .
Theorem. If is a normal operator on a Hilbert space , then each eigenspace of reduces .

reduces we have to show that is invariant both under and ∗ , as we know that a closed
Proof. Let be an eigen space of corresponding to the eigen value , in order to prove that

∗
linear subspace of a Hilbert space reduces an operator iff is invariant under both and
.

∗
, let us take any vector ∈ . Then
Now is invariant under because is eigenspace of . To show that is also invariant under

=
Therefore ∗
= ̅ . Since is linear subspace of , therefore ∈ and ̅ is some scalar .

Lovely Professional University 161

 Notes

Unit 14: Spectrum of Normal Operators

Implies ,
̅ ∈ . Thus ∈ implies ∗ = ̅ ∈ .
Therefore is also invariant under ∗
.
Hence reduces .
Theorem. Let be an operator on a finite dimensional Hilbert space . Let , . … ,
distinct eigen values of and let , , … ,
be the

! , ! , … , ! be the projections on these eigenspaces. Then the following statements are equivalent.
be their corresponding eigenspaces , and let

" ′$ .
The !"% $ are pairwise orthogonal , ! & ! & ⋯ & ! = and = ! & ! &⋯& ! .
I. The are pairwise orthogonal and span
II.
III. is normal.

14.2 Spectral Resolution

Definition. Let be an operator on a Hilbert space . If there exists distinct complex numbers
, ,…,
and non zero pairwise orthogonal projections ! , ! , … , ! such that
= ! & ! & ⋯& !
and ……………(1)
! & ! & ⋯&! = ,
Then the expression (1) for is called spectral resolution for .

Note: The spectral theorem tells us that every normal operator on a non zero finite
dimensional Hilbert space has a spectral resolution.
Now in the following theorem we shall prove that spectral resolution of a normal operator on a
finite dimensional non zero Hilbert space is unique.
Theorem. The spectral resolution of a normal operator on a finite dimensional non zero Hilbert
space is unique.
Proof. Let be a normal operator on a finite dimensional non zero Hilbert space .
Let

= ! & ! & ⋯& ! ……….(1)

Be a spectral resolution of .
Then
, ,…, are distinct complex numbers and !"% $ are non-zero pairwise orthogonal projections
such that
! & ! & ⋯ & ! = ……….(2)
First we show that the scalars , ,…, are precisely the distinct eigen values of .
First we shall prove that the scalars , ,…, are precisely the distinct eigen values of .
Let us first show that for each (, " is an eigen value of .
Since !" ≠ 0, therefore there exists a non zero vector in the range of !" . But !" is the projection.
Therefore
!" = .
Now

= ! & ! &⋯& !

162 Lovely Professional University

Notes
Functional Analysis

= ! & ! & ⋯& ! !"

= ! !" & ! !" & ⋯ & ! !"

= " !" !" !) = 0, (* ( ≠ +

= " !" ," = ," , ," being a projection.

= " .
Thus is a non zero vector such that = " , therefore " is an eigen value of .
Now we show that each eigen value of is an element of the set { , ,…, }.
Since is an operator on a finite dimensional Hilbert space , therefore must posses an eigen
value. Let be an eigen value of . Then there exists a non zero vector such that
=
⇒ = as =

⇒ ! & ! &⋯& ! = ! & ! & ⋯&!

⇒ − ! & − ! & ⋯& − ! = 0.

Operating on this with ," and remembering that !" = !" and !" !) = 0, (* ( ≠ +, we get

" − !" = 0 for ( = 1,2, … , 1

If " ≠ for each (, then we have !" = 0 for each (. Then we have
!" = 0 for each (.
Therefore,
! &! & ⋯&! =0
⇒ ! & ! & ⋯& ! =0

⇒ =0

⇒ =0
This contradicts the fact that ≠ 0.
Hence must be equal to " for some (.

"$
%

eigen values of .
This we have proved that in the spectral resolution (1) of the scalars are precisely the distinct

Therefore if
= 2 3 & 2 3 & ⋯ & 24 34 ……….(3)
is another spectral resolution of , then the scalars 2"% $ are precisely the distinct eigenvalues of .
Therefore remaining the projections 3"% $, if necessary , we can write (3) in the form
= 3 & 3 &⋯& 3 .
Now we shall show that in the spectral resolution (1) of the !"% $ are uniquely determined as
specific polynomials in .
We have
5 = = ! & ! & ⋯& !

Lovely Professional University 163

 Notes

Unit 14: Spectrum of Normal Operators

= ! & ! & ⋯& !

= ! & ! & ⋯& ! ! & ! &⋯& !

= ! & ⋯& !

[!" = !" , !" !) = 0, (* ( ≠ +]

Similarly ,
6 = 6
! & ⋯& 6 ! , where 7 is any positive integer.
Therefore, if 8 9 is any polynomial with complex coefficients, in the complex variable 9, then
taking linear combination of the above relation ,we get
8 =8 ! &8 ! & ⋯& 8 !

= ∑)= 8; ) <!) .

Now suppose that ," is a polynomial such that

," ; ) < = >")
That is
," " = 1 , if ( = +
and
," " = 0 , if ( ≠ +.
Then taking ," in place of 8, we get

," = ? ," ; ) <!) .

)=

= ? >") !) .
)=

= !" .
Thus for each (, ," = !" .
Which is a polynomial in . But we must show the existence of such a polynomial ," over the field
of complex numbers .
Obiviously
9− … 9− 9− … 9−
," 9 =
"@ "A
" − … "− "@ "− "A … "−
Serves the purpose .
That is
," " = 1 , if ( = +
and
," " = 0 , if ( ≠ +.
If we apply the above discussion for 3"% $ then we shall get
3" = ," for each (.
Therefore !" = 3" for each (.
Hence, the two spectral resolutions are the same.

164 Lovely Professional University

Notes
Functional Analysis

Note: If is a normal operator on a finite dimensional Hilbert space , the there exists an
orthonormal basis for relative to which the matrix of is diagonal matrix.

14.3 Non-Emptiness of the Spectrum

dimensional Hilbert space .

The following theorem establishes the non-emptiness of the spectrum of an operator on a finite

of namely B is a finite subset of the complex plane and the number of points in B
Theorem. If is an arbitrary operator on a finite dimensional Hilbert space , then the spectrum

exceed the dimension 7 of .

does not

For the proof of this theorem, we need the following lemma

a non-zero vector ∈ = 0.
Lemma. An operator on a finite dimensional Hilbert space is singular if and only if there exists
such that
Proof of lemma . Suppose there exists a non zero vector ∈ such that = 0. We can write
= 0 as = 0.
Since ≠ 0,
the two distinct elements , 0 ∈ have the same image under . Therefore the mapping
one-one . Hence @ does not exists . Hence it is singular.
is not

To prove the converse assume that is singular. Suppose there exists no non-zero vector such
that
= 0.
This means that = 0,
implies ,
= 0.
Then must be one-one. Since is finite dimensional and is one-one, is onto so that is non-
singular contracdicting the the hypothesis that is singular .
Hence there must be a non-zero vector such that = 0.
of dimension 7.
A scaler ∈ B , if there exists a non-zero vector such that − = 0.
Proof of the theorem. Let be an operator on a finite dimensional Hilbert space

Now,
− = 0 if and only if − is singular by the above lemma .
But,
− is singular if and only if CD9 − = 0.
Thus,
∈B if and only if satisfies the equation CD − = 0.
Let E be an ordered basis for .Thus CD9 − = CD9 F − GH .
But ,
det F − GH = det F GH − F GH .
Thus,
det − = det F GH − F>") G .
So,
CD9 − = 0 implies det;F GH − L>") M< = 0…………(1)
If F GH = F2") G is the matrix of , then (1) gives

Lovely Professional University 165

 Notes

Unit 14: Spectrum of Normal Operators

2 − 2 … 26
2 2 − … 2 6
N O = 0…………………..….(2)
… … … …
26 … … 266 −
The expansion of the determinant of (2) gives a polynomial equation in of degree 7 with complex

the field of complex numbers. Hence every operator on has an eigenvalue so that B ≠ 0.
coefficients . So, by the fundamental theorem of algebra this equation must have atleast one root in

Further, this equation in has exactely 7 roots in the complex field. If the equation has repeated
roots , then the number of distinct roots are less than 7. So that has an eigenvalue and the
number of distinct eigenvalues of is less than or equal to 7. Hence the number of elements of
B is less than or equal to 7.

are complex, B
contain as many as n distinct points but not more than 7 points. If the scaler field is real, it is
Note: If the scalers associated with contains atleast one point. It may

possible that B is empty. Hence in the spectral theory, we usually take the complex scalers so
that we get a richer theory.

Summary

with eigenvalue ̅.
 If is a normal operator on a Hilbert space , then is an eigen vector of with eigen
∗

, then the eigenspaces of

value iff is an eigen vector of
 If is a normal operator on a Hilbert space are pairwise

reduces .
orthogonal.
 If is a normal operator on a Hilbert space , then each eigenspace of
 A closed linear subspace of a Hilbert space reduces an operator iff is invariant
∗.
under both and

‖ ‖=‖ ‖ for every .

 An operator on a Hilbert space is normal iff
∗

. Let , .…,
, ,…,
 If be an operator on a finite dimensional Hilbert space be the

let ! , ! , … , ! be the projections on these eigenspaces. Then the following statements are
distinct eigen values of and let be their corresponding eigenspaces , and

" ′$ .
equivalent.

The !"% $ are pairwise orthogonal , ! & ! & ⋯ & ! = and = ! & ! &⋯& ! .
a. The are pairwise orthogonal and span
b.

. If there exists distinct complex numbers

c. is normal.

, ,…,
 If be an operator on a Hilbert space

and non zero pairwise orthogonal projections ! , ! , … , ! such that

= ! & ! & ⋯& !

! & ! & ⋯&! = ,

and ……………(1)

Then the expression (1) for is called spectral resolution for .

 The spectral resolution of a normal operator on a finite dimensional non zero Hilbert space
is unique.
 If is a normal operator on a finite dimensional Hilbert space , the there exists an
orthonormal basis for relative to which the matrix of is diagonal matrix.

namely B is a finite subset of the complex plane and the number of points in B
 If is an arbitrary operator on a finite dimensional Hilbert space , then the spectrum of

does not exceed the dimension 7 of .

166 Lovely Professional University

Notes
Functional Analysis

∈ = 0.
 An operator on a finite dimensional Hilbert space is singular if and only if there exists
a non-zero vector such that

Keywords
 Spectrum of an operator
 Normal operator
 Spectral resolution
 Hilbert space
 Eigenvalue
 Eigenvector
 Eigenspace
 Linear space
 Invarience
 Finite dimensional non zero Hilbert space
 Diagonal matrix

Self Assessment

is an eigen vector of ∗ with eigenvalue ̅.

1: If is a normal operator on a Hilbert space , then is an eigen vector of with eigen value
iff

A. True
B. False

2: Which of the following statement is /are true.

I. If is a normal operator on a Hilbert space , then the eigenspaces of are pairwise

reduces .
orthogonal.
II. If is a normal operator on a Hilbert space , then each eigenspace of
.

A. Only I is true.
B. Only II is true.
C. Neither (I) nor (II)
D. Both (I) and (II).

3: Which of the following statement is /are true.

I. A closed linear subspace of a Hilbert space reduces an operator iff is invariant

under both and ∗.

‖ ‖=‖ ‖ for every .

II. An operator on a Hilbert space is normal iff
∗

A. Only I .
B. Only II
C. Neither (I) nor (II)
D. Both (I) and (II).

Lovely Professional University 167

 Notes

Unit 14: Spectrum of Normal Operators

4 : The spectral resolution of a normal operator on a finite dimensional non zero Hilbert space
is unique.

A. True
B. False

5: Which of the following statement is /are true.

I. Every self adjoint operator is normal.

II. A normal operator need not be self adjoint.
A. Only I .
B. Only II
C. Both (I) and (II).
D. Neither (I) nor (II)

6 : If P is a normal operator on , then

A. ‖P ‖ < ‖P‖ .
B. ‖P ‖ = ‖P‖ .
C. ‖P ‖ > ‖P‖ .
D. None of the above.

non-zero vector ∈ = 0.
7: An operator on a finite dimensional Hilbert space is singular if and only if there exists a
such that

A. True
B. False

8. If is a normal operator on a finite dimensional Hilbert space , the there exists an

orthonormal basis for relative to which the matrix of is diagonal matrix.

A. True
B. False

Answers for Self Assessment

1. A 2. D 3. D 4. A 5. C

6. B 7. A 8. A

Review Questions
Q1:- What is Spectrum of Normal operator.
Q2:- Show that if T is a normal operator on a Hilbert space H, then the eigen spaces of T are
pairwise orthogonal.
Q3:- prove that the spectral resolution of a normal operator on a finite dimensional non zero Hilbert
space is unique.
Q4:- Prove that an operator T on a finite dimensional Hilbert space H is singular if and only if there
exists a non zero x in H such that Tx=0. Prove that the zero operator on any normed space is
compact.

168 Lovely Professional University

Notes
Functional Analysis

Further Readings
 Introductory Functional Analysis With Applications By Erwin Kreyszig.
 Functional Analysis By Walter Ruddin, Mcgraw Hill Education.
 J. B Conway, A Course In Functional Analysis.
 C. Goffman G Pedrick, A First Course In Functional Analysis.
 B.V. Limaya, Functional Analysis.

Lovely Professional University 169

LOVELY PROFESSIONAL UNIVERSITY
Jalandhar-Delhi G.T. Road (NH-1)
Phagwara, Punjab (India)-144411
For Enquiry: +91-1824-521360
Fax.: +91-1824-506111
Email: odl@lpu.co.in