Chapter 3.

Normed vector spaces

Lecture notes for MA2223

P. Karageorgis

pete@[Link]

1 / 15
The definition of a norm

Definition – Norm
Suppose X is a vector space over the field F = R or F = C. A norm
on X is a real-valued function ||x|| with the following properties.
1 Zero vector: ||x|| = 0 if and only if x = 0.
2 Scalar factors: ||λx|| = |λ| · ||x|| for all λ ∈ F and all x ∈ X.
3 Triangle inequality: ||x + y|| ≤ ||x|| + ||y|| for all x, y ∈ X.

A normed vector space (X, || · ||) consists of a vector space X and a

norm ||x||. One generally thinks of ||x|| as the length of x.
It is easy to check that every norm satisfies ||x|| ≥ 0 for all x ∈ X.
Every normed vector space (X, || · ||) is also a metric space (X, d), as
one may define a metric d using the formula d(x, y) = ||x − y||. This
particular metric is said to be induced by the norm.

2 / 15
Examples of normed vector spaces
Given any p ≥ 1, we can define a norm on Rk by letting
" k #1/p
X
||x||p = |xi |p .
i=1

The space C[a, b] has a similar norm for any p ≥ 1, namely

Z b 1/p
p
||f ||p = |f (x)| dx .
a

There is also a norm on Rk for the case p = ∞. It is defined by

||x||∞ = max |xi |.
1≤i≤k

Finally, there is a similar norm on C[a, b] which is given by

||f ||∞ = max |f (x)|.
a≤x≤b

3 / 15
Sequence spaces ℓp
The space ℓp consists of all real sequences x = {xn } such that
∞
X
|xn |p < ∞.
n=1

It is a normed vector space for any p ≥ 1 and its norm is given by

"∞ #1/p
X
p
||x||p = |xn | .
n=1

The space ℓ∞consists of all bounded real sequences x = {xn }. It is

a normed vector space and its norm is given by

||x||∞ = sup |xn |.

n≥1

The space c0 consists of all real sequences {xn } which converge to 0.

It is easily seen to be a subspace of ℓ∞ .
4 / 15
Continuity of operations

Theorem 3.1 – Product norm

Suppose X, Y are normed vector spaces. Then one may define a norm
on the product X × Y by letting ||(x, y)|| = ||x|| + ||y||.

Theorem 3.2 – Continuity of operations

The following functions are continuous in any normed vector space X.
1 The norm f (x) = ||x||, where x ∈ X.
2 The vector addition g(x, y) = x + y, where x, y ∈ X.
3 The scalar multiplication h(λ, x) = λx, where λ ∈ F and x ∈ X.

We shall mainly use this theorem to justify computations such as

lim ||xn || = lim xn .

n→∞ n→∞

5 / 15
Bounded linear operators

Definition – Bounded, linear, continuous

Let X, Y be normed vector spaces over the field F = R or F = C.
1 A function T : X → Y is called a linear operator, if

T (x + y) = T (x) + T (y), T (λx) = λT (x)

for all x, y ∈ X and all scalars λ ∈ F.

2 A function T : X → Y is called bounded, if there exists a real
number M > 0 such that ||T (x)|| ≤ M ||x|| for all x ∈ X.
3 A function T : X → Y is called continuous, if it is continuous
with respect to the metrics which are induced by the norms.

A linear operator is also known as a linear transformation.

By definition, every linear operator T is such that T (0) = 0.
6 / 15
Bounded means continuous

Theorem 3.3 – Bounded means continuous

Suppose X, Y are normed vector spaces and let T : X → Y be linear.
Then T is continuous if and only if T is bounded.

Theorem 3.4 – Norm of an operator

Suppose X, Y are normed vector spaces. Then the set L(X, Y ) of all
bounded, linear operators T : X → Y is itself a normed vector space.
In fact, one may define a norm on L(X, Y ) by letting

||T (x)||
||T || = sup .
x6=0 ||x||

It is easy to check that ||T (x)|| ≤ ||T || · ||x|| for all x ∈ X.

One also has ||S ◦ T || ≤ ||S|| · ||T || whenever S, T ∈ L(X, X).
7 / 15
Norm of an operator: Example 1

Consider the right shift operator R : ℓp → ℓp which is defined by

R(x1 , x2 , x3 , . . .) = (0, x1 , x2 , . . .).

This operator is easily seen to be linear and we also have

||R(x)||p = ||x||p for all x ∈ ℓp .

In particular, the norm of this operator is equal to ||R|| = 1.

The left shift operator L : ℓp → ℓp is similarly defined by

L(x1 , x2 , x3 , . . .) = (x2 , x3 , x4 , . . .).

Since ||L(x)||p ≤ ||x||p for all x ∈ ℓp , we find that ||L|| ≤ 1. On the

other hand, we also have ||L(x)||p = ||x||p whenever x1 = 0 and this
implies that ||L|| ≥ 1. We may thus conclude that ||L|| = 1.

8 / 15
Norm of an operator: Example 2
Suppose that T : (Rn , || · ||1 ) → (Rm , || · ||∞ ) is left multiplication by
the m × n matrix A. We then have
X X
||T (x)||∞ = max aij xj ≤ max |aij | · |xj |
i i,j
j j

= max |aij | · ||x||1

i,j

and this implies that ||T || ≤ maxi,j |aij |.

On the other hand, the standard unit vector x = ej satisfies

||T (x)||∞ X
= max aij xj = max |aij |,
||x||1 i i
j

so we also have ||T || ≥ maxi |aij | for each j. We conclude that

||T || = max |aij |.
i,j

9 / 15
Finite-dimensional vector spaces

Suppose that X is a vector space with basis x1 , x2 , . . . , xk . Then

every element x ∈ X can be expressed as a linear combination

x = c 1 x1 + c 2 x2 + . . . + c k xk

for some uniquely determined coefficients c1 , c2 , . . . , ck ∈ F.

Theorem 3.5 – Euclidean norm

Suppose that X is a vector space with basis x1 , x2 , . . . , xk . Then one
may define a norm on X using the formula
v
X k u k
uX
x= ci xi =⇒ ||x||2 = t |ci |2 .
i=1 i=1

This norm is also known as the Euclidean or standard norm on X.

10 / 15
Equivalent norms

Definition – Equivalent norms

We say that two norms || · ||a and || · ||b of a normed vector space X
are equivalent, if there exist constants C1 , C2 > 0 such that

C1 ||x||a ≤ ||x||b ≤ C2 ||x||a for all x ∈ X.

Theorem 3.6 – Equivalence of all norms

The norms of a finite-dimensional vector space X are all equivalent.

The norms || · ||1 and || · ||∞ are not equivalent in C[a, b] because this
space is complete with respect to only one of the two norms.
In fact, || · ||p and || · ||q are not equivalent in C[0, 1] when p < q. To
prove this, one may define fn (x) = xn for each n ∈ N and then check
that the quotient ||fn ||q /||fn ||p is unbounded as n → ∞.
11 / 15
Banach spaces

Definition – Banach space

A Banach space is a normed vector space which is also complete with
respect to the metric induced by its norm.

Theorem 3.7 – Examples of Banach spaces

1 Every finite-dimensional vector space X is a Banach space.
2 The sequence space ℓp is a Banach space for any 1 ≤ p ≤ ∞.
3 The space c0 is a Banach space with respect to the || · ||∞ norm.
4 If Y is a Banach space, then L(X, Y ) is a Banach space.

The space C[a, b] is a Banach space with respect to the || · ||∞ norm.
It is not complete with respect to the || · ||p norm when 1 ≤ p < ∞.
Suppose that X is a Banach space and let Y be a subspace of X.
Then Y is itself a Banach space if and only if Y is closed in X.
12 / 15
Convergence of series

Definition – Convergence of series

Suppose that {xPn } is a sequence in a normed vector space [Link] say
N
that the series ∞n=1 x n converges, if the partial sum s N = n=1 xn
converges as N → ∞. If that is the case, then we denote its limit by
N
X ∞
X
lim sN = lim xn = xn .
N →∞ N →∞
n=1 n=1
P∞ P∞
We say that n=1 xn converges absolutely, if n=1 ||xn || converges.

Theorem 3.8 – Absolute convergence implies convergence

Suppose that X is a Banach space and let ∞
P
n=1 xn be a series which
converges absolutely in X. Then this series must also converge.

13 / 15
Invertible linear operators

Definition – Invertibility
A bounded linear operator T : X → X is called invertible, if there is a
bounded linear operator S : X → X such that S ◦ T = T ◦ S = I is
the identity operator on X. If such an operator S exists, then we call
it the inverse of T and we denote it by T −1 .

Theorem 3.9 – Geometric series

Suppose that T : X → X is a bounded linear operator onPa Banach
space X. If ||T || < 1, then I − T is invertible with inverse ∞ n
n=0 T .

Theorem 3.10 – Set of invertible operators

Suppose X is a Banach space. Then the set of all invertible bounded
linear operators T : X → X is an open subset of L(X, X).

14 / 15
Dual space

Definition – Dual space

Suppose X is a normed vector space over R. Its dual X ∗ is then the
set of all bounded linear operators T : X → R, namely X ∗ = L(X, R).

Theorem 3.11 – Dual of Rk

There is a bijective map T : Rk → (Rk )∗ that sendsP
each vector a to
the bounded linear operator Ta defined by Ta (x) = ki=1 ai xi .

Theorem 3.12 – Dual of ℓp

Suppose 1 < p < ∞ and let q = p/(p − 1). Then 1/p + 1/q = 1 and
there is a bijective map T : ℓq → (ℓp )∗ that sends each P
sequence {an }
to the bounded linear operator Ta defined by Ta (x) = ∞ i=1 ai xi .

15 / 15