Functional Analysis

MATH 464

A student’s notes

DISCLAIMER: Use at your own risk. Any information should be

verified before use
Contents

1 Vector spaces 1
1.1 Linear vector spaces . . . . . . . . . . . . . . . . . . . . 1
1.2 LInear normed spaces . . . . . . . . . . . . . . . . . . . 2
1.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Transformation of vector spaces 6

2.1 Linear operators . . . . . . . . . . . . . . . . . . . . . . 6
2.2 `pn space . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Linear functionals . . . . . . . . . . . . . . . . . . . . . 15

3 Appendix 17
3.A A brief on Topology . . . . . . . . . . . . . . . . . . . . 17
3.B Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . 17

i
1. Vector spaces

SECTION 1.1
Linear vector spaces

In this course, ‘vector spaces’ are the basis of our discussion. However,
we often do not mean a space that contains points that indicate posi-
tion(although they are also applicable). We can speak of other types
of elements such as sequences and functions. But the idea is that we
want to consider these elements in relation to other elements in the
same space. With that in mind, they are just ‘points’ to us as far as
we are concerned.
Definition 1. A nonempty set V is said to be vector space over a
scalar field F if the following properties/axioms hold
1. V is an abelian group(over a certain binary operation)
2. V has a compatible scalar operation
Note.
• (F, +) is an abelian group. (F\{0}, ×) is also an abelian group.
Multiplication distributes over addition
• In fields, you can’t have zero divisions. That is, if αβ = 0, then
either α, β = 0
Definition 2 (metric, metric space). A metric space is a pair (X, d),
where X is a set and d is a metric on X(or a distance funtion on X),
that is, a function defined on X × X such that for all x, y ∈ X we
have
1. d(x, y) ≥ 0
2. d(x, y) = 0 if and only if x = y
3. d(x, y) = d(y, x) (symmetric)
4. d(x, z) ≤ d(x, y) + d(x, z) (triangle inequality)

1
1.2. LINEAR NORMED SPACES

SECTION 1.2
LInear normed spaces

Norm. A real-valued function kxk from a vector space X over a scalar

field K into the real numbers R

kxk : X → R

is called a norm if it satisfies these

1. kxk > 0 if x 6= 0
2. kαxk = |α|kxk if x ∈ X and α is a scalar.
3. |x + y| ≤ kxk + kyk for all x, y ∈ X

Normed space. A vector space X is said to be normed linear space

if every x ∈ X is endowed with a norm kxk. It is denoted by (X, k·k)

Definition 3. A norm k·k on a vector space X is said to be equivalent

to the norm on X, ie. kxk0 , if there are positive numbers a and b such
that for all x ∈ X we have

akxk0 ≤ kxk ≤ bkxk0

The concept is motivated by the fact that equivalent norms on X define

the same topology

Theorem 1. On a finite dimensional vector space X, any norm k·k

is equivalent to any other norm k·k0

Proof. Let dim X = n and {e1 , . . . , en } be a finite dimensional vector

space. Then any vector x ∈ X can be expressed uniquely as
n
X
x= aj ek
j=1

We can make a choice of scalars aj such that for some c

X
c |aj | ≤ kxk (1.1)
j

2
1.3. EXAMPLES

On the other hand, the triangle inequality gives

X
kxk0 = aj ej
j
X
≤ |aj |kej k
j
X
≤ maxkej k |aj |
j
j
X  
=k |aj | k = maxkej k
j
j
X
=⇒ kxk0 ≤ k |aj | (1.2)
j

From 1.1 and 1.2, we see that

X
c |aj | ≤ kxk (1.3)
j
c
kxk0 ≤ kxk
k
c
 
akxk0 ≤ kxk a=
k
The other inequality in the definition above is obtained by switching
the roles of k·k0 and k·k

SECTION 1.3
Examples

1.3.1 Examples of norms in finite dimensional spaces

The metric induced by a norm on a vector space is the canonical space
given by
d(x, y) = kx − yk
Consider the following examples of norms in Rn

1. Taxicab norm/ Manhatta norm

n
X
kxk1 := |xi |
i=1

3
1.3. EXAMPLES

2. Euclidean norm !1/2

n
X
kxk2 := x2i
i=1
3. P -norm !1/p
n
X
p
kxkp := |xi |
i=1

4. Infinite/ Chebyshev norm

kxk∞ := max |xi |

1≤i≤n

5. P -adic norm
Let p be a prime. The p-adic norm is given by
(
p−Vp (x) if x 6= 0
|x| =
0 if x = 0

Vp (x) is known as the p-adic order or valuation and is defined by

(
max{v ∈ N : pv |n} if n 6= 0
Vp (x) =
∞ if n = 0

The p-adic order can be extended to rational numbers and is

given by

a
 
Vp = Vp (a) − Vp (b)
b

If n/d is a rational number in its lowest terms then,


 Vp (n) if p|n


a
 
Vp = −V (n)
p if p|d
b 
if p 6 |n nor p 6 |d

 0

Note that the p-adic norm is not a true norm. It does not obey
the absolute scalability requirement

4
1.3. EXAMPLES

1.3.2 Examples of normed linear spaces

C 0 [D] space
This is the space of continuous functions where D is the domain or
often the interval.

C k [I] space
This is the space of k times differentiable functions with the domain I

`p space
This is space of sequences (xj )n1 where 1 ≤ p ≤ ∞. This is a very
important linear space which we will talk more about later.

Lp space
This is the space of Lebesgue-integrable functions

5
2. Transformation of vector
spaces

SECTION 2.1
Linear operators

Definition 4. Let X, Y be vector spaces. Then a map

T: X →Y
is said to be linear operator or map/transformation if for scalars α, β
and vectors x, y ∈ X, the following is satisfied
T (αx + βy) = αT x + βT y
We proceed to discuss some properties of linear operators such as
boundedness and continuity and some implications of those proper-
ties.

2.1.1 Bounded and continuous linear operators

Bounded linear operator. Let X and Y be normed vector spaces.
A linear operator T : X → Y is said to be bounded if there is a L > 0
in R such that for all x ∈ X
kT xkY ≤ LkxkX (2.1)
If an operator is not bounded, it is an unbounded linear operator
Note. The term bounded operator comes from the fact that the linear
operator T maps bounded sets in X onto bounded sets in Y
We have defined what we mean by a bounded linear operator. We
proceed to prove that as long as a linear space X is finite dimensional,
all linear operators on such X are bounded. We will first look at a
useful result.

6
2.1. LINEAR OPERATORS

Lemma 1. Let {x1 , . . . , xn } be a linearly independent set of vectors in

a normed space X (of any dimension). Then there is a number c > 0
such that for every choice of scalars α1 , . . . , αn , we have

kα1 x1 + . . . + αn xn k ≥ c(|α1 | + . . . + |αn |)

Proof. Can be found in Kreyzig, page 74

Theorem 2. If a normed space X is finite dimensional, then every

linear operator on x ∈ X is bounded

Proof. Let dim X = n and {e1 , e2 , . . . , en } be the basis vectors that

span X. Then any vector x ∈ X can be expressed as x = j aj ej .
P

Consider a linear operator T such that

T :X→Y (k·kX = k·kY )

The norm of the operator is given by

X X X
kT xk = T aj e j ≤ |aj |kT ej k ≤ maxkT ek k |aj | (2.2)
k
j j

By Lemma 1
X X
k aj ej k ≥ c |aj |
j j
X 1 X
|aj | ≤ k aj ej k
j
c j
X 1
|aj | ≤ kxk (2.3)
j
c

From 2.2 and 2.3

X 1
kT xk ≤ maxkT ek k |aj | = maxkT ek kkxk
k c k

We finally get
 
1
kT xk ≤ γkxk γ= c maxkT ek k
k

7
2.1. LINEAR OPERATORS

Continuous linear operator. A linear operator

T :X→Y

on a linear space X is said to be continuous if for every x ∈ X

lim T x = T x0
x→x0

T is called uniformly continuous if for every ε > 0, there exists a δ > 0

such that for all x, y ∈ X

kx − yk < δ =⇒ kT x − T yk < ε

We move on to a stronger form of uniform continuity for functions in

the next definition.

Definition 5. Given two metric spaces (X, k·kX ) and (Y, k·kY ), a func-
tion T : X → Y is said to be Lipschitz continuous if there is a real
constant K ≥ 0(Lipschitz constant) such that, for all x1 and x2 in X

kT x1 − T x2 k ≤ Kkx1 − x2 k

Note. If K = 1, the function T is called a short map. And if 0 ≤ K < 1

and T maps a metric space to itself, the function is a contraction. In
general, a metric map is a function between two metric spaces. We
have short maps and also contractions. A short map is a function
that does not increase any distance, ie., it is a lipschitz funtion with
constant 1.
Intuitively, a Lipschitz function is limited to how fast it can change;
the norm of the slope of any two points on the function should be less
than or equal to a certain real number
T x1 − T x2
≤K
x1 − x2
Theorem 3. Let T : X → Y to be a linear operator between two
normed vectors (X, k·k) and (Y, k·k), then the following are equivalent

T is continuous (2.4)
T is continuous at the origin (2.5)
T is bounded (2.6)
T is Lipschitz continuous(and therefore uniformly continuous) (2.7)

8
2.1. LINEAR OPERATORS

Proof.

(2.4) =⇒ (2.5)

Hence it is trivial. Next we prove that

(2.5) =⇒ (2.6)

We prove the contrapositive. Suppose that T is unbounded, ie., to say

that ∀n ∈ N ∃xn ∈ X such that

kT xkY > nkxkX

Obviously, 0 ∈
/ {xn }X

1
=⇒ >1
nkxn kX
!
1
=⇒ T xn >1
nkxn kX Y

Now let
!
1 1 1
un = xn =
nkxn kX n kxn kX
=⇒ lim un = 0
n→∞

from ((2.5))

lim kT un k ≥ 1
n→∞

Thus T cannot be continuous at the origin because un is a null sequence

- ie. un → 0. But eqn (2.5) says that T un 6→ 0

(2.6) =⇒ (2.7)

Let L be the boundedness constant of T ie.

kT xkY ≤ LkxkX for all x ∈ X

Then by linearity of T

kT x − T zkY = kT (x − z)kX ≤ kx − zkX

9
2.1. LINEAR OPERATORS

Thus T is Lipschitz continuous and therefore uniformly continuous

(2.7) =⇒ (2.4)

Theorem 4. Let T : X → Y be a linear operator, where X and Y are

normed spaces. Then, T is continuous if and only if T is bounded

Proof. If T = 0, then it is trivial. Let T 6= 0. We assume T is bounded

and consider an arbitrary x0 ∈ X. Let any ε > 0 be given. Since T is
linear, we obtain

kT x − T x0 k = kT (x − x0 )k ≤ kT kkx − x0 k

We let
ε
kx − x0 k ≤ δ where δ=
kT k

Then

kT x − T x0 k = kT (x − x0 )k ≤ kT kkx − x0 k < kT kδ = ε

Since x0 ∈ X is arbitrary, T is continuous. Conversely, assume that T

is continuous at x0 ∈ X. Then given any ε > 0, there is an δ > 0 such
that

kT x − T x0 k ≤ ε =⇒ kx − x0 k ≤ δ (2.8)

We now take any y 6= 0 in X and set

δ δ
x = x0 + y then x − x0 = y
kyk kyk

Hence

kx − x0 k = δ

From 2.8, since T is linear

y δ
 
kT x − T x0 k = kT (x − x0 )k = kT δ k= kT yk
kyk kyk

10
2.1. LINEAR OPERATORS

which implies

δ
kT yk ≤ ε
kyk

kT yk ≤ kyk
δ

This can be written as


 
kT yk ≤ ckyk where c =
δ
This shows that T is bounded

2.1.2 Operator norm

Introduction
We already spoke about norms in general but the norms of linear op-
erators are of great importance to us. It is a means to measure the
“volume” of some linear operators. Let T be a bounded linear opera-
tor from (X, k·kX ) to (X, k·kY ). The norm of such an operator is the
operator norm.

Derivation
The linear operator T is continuous if and only if there is a real number
L such that
kT xk ≤ Lkxk
Intuitively, the continuous linear operator T never increases the norm
of any vector more than by a factor L1 . It immediately follows that
the image of a bounded set under a continuous linear operator is also
bounded. This means that continuous linear operator is bounded as
seen in the definition of a bounded linear operator in (2.1). From this
definition, we can find the least L for which (2.1) holds for all nonzero
x ∈ X and this value is the norm of the operator. That is,

kT kop = inf {kT xk ≤ Lkxk}

x∈X
L≥0

1
See the concept of Lipschitz continuity from previous section

11
2.1. LINEAR OPERATORS

In other words, we measure the “size” of A by how much it “lengthens”

vectors in the “biggest” case. By division in equation (2.1),

kT xk
≤L (x 6= 0, k·kX = k·kY )
kxk

This shows that L must be as large as the supremum of the expression

on the left taken over the domain X\{0}. This quantity is denoted by
kT k. Thus,

kT xk
kT k = sup
x∈X kxk

kT k is called the operator norm. With L = kT k, (2.1) is

kT xk ≤ kT kkxk

Lemma 2 (Operator norm). Let T be a bounded linear operator. Then:

1. The alternative form of the norm of T is

kT kop = sup kT xk (2.9)

kxk=1

2. The norm satisfies the properties of the norm

Operator norms in finite dimensional spaces are actually matrix norms

induced by the vector norms.

Examples of operator norms

(Cn , k·k1 ) to (Cm , k·k1 )
m
X
kT k1 = max keij k |aij |
1≤j≤n
i=1

which is simply the maximum absolute column sum of the matrix

(Cn , k·k∞ ) to (Cm , k·k∞ )

n
X
kT k∞ = max keij k |aij |
1≤j≤m
i=1

which is simply the maximum absolute row sum of the matrix

12
2.1. LINEAR OPERATORS

(Cn , k·k2 ) to (Cm , k·k2 )

q
kT k2 = max∗ |λ|
λ∈eig(M M )

Where M ∗ is the conjugate transpose of M and M ∗ M is therefore a

Hermitian. In this special case where p = 2, the induced matrix norm
is the spectral norm.

2.1.3 Examples of linear operators

Examples of bounded linear operators
1. A linear transformation T

T : R → R , r 7→ mr
kT rkR = |m|krkR

2. Integral operator

Examples of unbounded linear operators

1. Differentiation operator
Let X = C 0 [0, 1] be a normed space of continuous functions and
Y = C 1 [0, 1] be space of differentiable functions with k = 1. A
differentiation operator is defined by

T f (x) = f 0 (x)

We see that

kf kC 1 = kf kC 0 = sup |f (x)|
x∈[0,1]

We can check if the two function spaces are normed spaces. For
positive-definiteness

kf k = sup |f (x)| ≥ 0
x∈[0,1]

13
2.1. LINEAR OPERATORS

We proceed to test for absolute scalability

kαf k = sup |αf (x)|

x∈[0,1]

= sup |α||f (x)|

x∈[0,1]

= |α| sup |f (x)|

x∈[0,1]

= |α|kf k

To test for the verity of the triangle inequality

kf + gk = sup |(f + g)(x)|

x∈[0,1]

≤ sup (|f (x)| + |g(x)|)

x∈[0,1]

= sup |f (x)| + sup |g(x)|

x∈[0,1] x∈[0,1]

=⇒ kf + gk ≤ kf k + kgk

If the linear operator T is unbounded then we have to prove that

d
fn > nkfn k
dx

Take

fn (x) = e(n+1)x
d
fn = (n + 1)e(n+1)x
dx
T fn = (n + 1)fn

Taking the norm

kT fn k = k(n + 1)fn k
= (n + 1)kfn k
> nkfn k

14
2.2. `PN SPACE

SECTION 2.2
`pn space

This space is a normed space like any other. However, due to how
important it is in this course, we will dedicate a separate section for
it. Here we detail, what the `pn space is, the various norms defined on
them and some other ideas connected to it.

SECTION 2.3
Linear functionals

Linear functional. A linear functional f on a vector space X over a

scalar field K is a linear operator whose co-domain is K. Thus

f: X →K

2.3.1 Examples of linear functionals

1. Norm
2. Dot product
3. Definite integral

15
Appendices

16
3. Appendix

SECTION 3.A
A brief on Topology

Definition 6 (Topology). A topology τ on a set X is a collection of

subsets of X satisfying:
1. ∅, τ ∈ X
2. τ is closed under finite intersections
3. τ is closed under arbitrary unions
A nonempty set X equipped with a topology τ is called a topological
space. It is denoted by (X, τ ) and if there is no danger of confusion, a
simple X is used.
Examples 1.
1. The trivial or indiscrete topology consists of the ∅ and τ .
2. The discrete topology consists of all the subsets of X. In other
words, it is the power set. Every set is both closed and open.
3. A semimetric d on a space X is a real-valued function on X × X
that is nonnegative, symmetric, satisfies d(x, x) = 0 for every
x, and in addition satisfies the triangle inequality. d(x, z) ≤
d(x, y) + d(y, z). A metric is a semimetric that has the property
that d(x, y) = 0 implies x = y
Convex set. A set is said to be convex if

SECTION 3.B
Miscellaneous

Definition 7. A sequence is said to be cauchy if

Definition 8. A Banach space

17
3.B. MISCELLANEOUS

3.B.1 Continuity
Definition 9. A function is said to be continuous

Definition 10. A function is said to be uniformly continuous

Definition 11. A function is said to be absolutely continuous

3.B.2 Inequalities
Young’s inequality.

Jensen’s inequality.

Holder’s inequality.

Minkowski’s inequality.

18