MODULE 1

KRISHNA KUMAR. G
DEPARTMENT OF MATHEMATICS
UNIVERSITY OF KERALA

1. Normed linear space

Definition 1.1. A Normed linear space is a linear space on which there is defined a
norm, i.e., a function which assigns to each element x in the linear space a real number kxk
in such a manner that
(1) kxk ≥ 0 and kxk = 0 if and only if x = 0;
(2) kx + yk ≤ kxk + kyk for every x, y ∈ A;
(3) kαxk = |α|kxk for every scalar α and x ∈ A.

• If the underlying field for the linear space is R, then the normed linear space is called
real normed linear space or normed linear space over R.
• If the underlying field for the linear space is C, then the normed linear space is called
complex normed linear space or normed linear space over C.
Throughout this note, we consider complex normed linear space, unless specified
other wise. The norm induces a metric on the linear space, defined by

d(x, y) = kx − yk

Thus every normed linear space is also a metric space. If with respect to this metric the
normed linear space is complete, then the normed linear space is called a Banach space.

Proposition 1.2. Let (X, k · k) be a complex normed linear space. Then

1. the function “norm” is a uniformly continuous function from X to R;
2. the operations vector addition and scalar multiplications are continuous.

Proof. 1. Since |kxk − kyk| ≤ kx − yk for all x, y ∈ X.

2. If xn → x, yn → y in X and kn → k in C imply

xn + y n → x + y and kn xn → kx.


1
2 KRISHNA KUMAR. G

Examples of normed linear space

1. The space CN for N ≥ 1 is a complex vector space. The operations are, if x =

(x1 , . . . , xN ), y = (y1 , . . . , yN ) and α ∈ C

x + y = (x1 + y1 , . . . , xN + yN );

αx = (αx1 , . . . , αxN ).

The norm is defined by, for 1 ≤ p < ∞,

kxkp = (|x1 |p + · · · + |xN |p )1/p .

We have

(1) From the definition, kxkp ≥ 0 for each x ∈ CN .

If x = (0, . . . , 0) the zero element of CN , then kxkp = 0. On the other hand, if kxkp = 0
then (|x1 |p + · · · + |xN |p )1/p = 0 and this implies |xi | = 0 for every i = 1, . . . , N i.e.,
x = 0.
(2) Next we have to show that for every x, y ∈ CN ,

kx + ykp ≤ kxkp + kykp

i.e.,
N
!1/p N
!1/p N
!1/p
X X X
(1.1) |xi + yi |p ≤ |xi |p + |yi |p .
i=1 i=1 i=1
N
(3) Next kαxkp = |α|kxkp for each α ∈ C and x ∈ C . Since

kαxkp = (|αx1 |p + · · · + |αxN |p )1/p = |α| (|x1 |p + · · · + |xN |p )1/p = |α|kxkp .

The inequality stated in (1.1) is known as Minkowski’s inequality. Note that for p = 1, the
inequality becomes trivially true. To prove Minkowski’s inequality first we need to prove the
Holder’s inequality.
1 1
Holder’s inequality: Let x, y ∈ CN and 1 < p < ∞ and p
+ q
= 1. Then
N
X
|xi yi | ≤ kxkp kykq .
i=1
1 1
Proof. We show that for p
+ q
= 1 and α, β ≥ 0 then
α β
α1/p β 1/q ≤ + .
p q
 NORMED LINEAR SPACE 3

The above holds if α = β or one of α and β is zero. Hence, assume that 0 < α < β. Consider
the continuously differentiable function f (t) = t1/q , t > 0. Then by the mean value theorem,
we have

f (β) − f (α) = f 0 (ξ)(β − α)

for some ξ ∈ (α, β), i.e.,

1
β 1/q − α1/q = ξ −1/p (β − α).
q
Since ξ −1/p < α−1/p , it follows that
1
β 1/q − α1/q < α−1/p (β − α),
q
i.e.,
1
α1/p β 1/q − α ≤ (β − α)
q
i.e.,
α β
α1/p β 1/q ≤ +
p q
|xi |p |yi |q
Choose α = kxkpp
and β = kykqq
, then

|xi | |yi | |xi |p |yi |q

≤ p + .
kxkp kykq pkxkp qkykqq
Taking summation from i = 1 to N , then
|x1 | |y1 | |xN | |yN | |x1 |p |y1 |q |xN |p |yN |q
+ ··· + ≤ + + ··· + + ;
kxkp kykq kxkp kykq pkxkpp qkykqq pkxkpp qkykqq
PN PN PN
|xi yi |
i=1 |xi |p
i=1 |yi |q
i=1
≤ + ;
kxkp kykq pkxkpp qkykqq
PN
|xi yi |
i=1
kxkpp kykqq 1 1
≤ p + q = + = 1.
kxkp kykq pkxkp qkykq p q
i.e.,
N
X
|xi yi | ≤ kxkp kykq .
i=1

Minkowski’s inequality: Let x, y ∈ CN and 1 < p < ∞, then

kx + ykp ≤ kxkp + kykp .

4 KRISHNA KUMAR. G

Proof. Clearly the inequality holds if atleast one of kxkp and kykp is infinity. Hence we
assume that kxkp < ∞ and kykp < ∞. We observe that for α, β ∈ C.

|α + β|p ≤ (|α| + |β|)p ≤ 2p max{|α|p , |β|p } ≤ 2p (|α|p + |β|p ).

Thus we have
|xi + yi |p ≤ 2p (|xi |p + |yi |p )

Taking sum over i = 1 to N we have kx + ykp < ∞. Now observe that

N
X N
X
p
|xi + yi | = |xi + yi ||xi + yi |p−1
i=1 i=1
N
X
≤ (|xi | + |yi |)|xi + yi |p−1
i=1
N
X N
X
p−1
= |xi ||xi + yi | + |yi ||xi + yi |p−1
i=1 i=1

Now applying Holder’s inequality on each of the right hand side of the above relation, we
have !1/q
N
X N
X
|xi ||xi + yi |p−1 ≤ kxkp |xi + yi |(p−1)q
i=1 i=1
N N
!1/q
X X
|yi ||xi + yi |p−1 ≤ kykp |xi + yi |(p−1)q
i=1 i=1
1 1
Since p
+ q
= 1 we have pq − q = p. Hence
N N
!1/q N
!1/q
X X X
p (p−1)q (p−1)q
|xi + yi | ≤ kxkp |xi + yi | + kykp |xi + yi |
i=1 i=1 i=1
N
!1/q N
!1/q
X X
p p
= kxkp |xi + yi | + kykp |xi + yi |
i=1 i=1
N
!1/q
X
p
= (kxkp + kykp ) |xi + yi |
i=1
i.e,
N
!1−1/q
X
|xi + yi |p ≤ kxkp + kykp
i=1
i.e.,
kx + ykp ≤ kxkp + kykp .
 NORMED LINEAR SPACE 5

2. The space CN for N ≥ 1 with the same operations defined above and the norm is
defined by, for x ∈ CN ,

kxk∞ = max{|x1 |, . . . , |xN |}.

From the definition of k · k∞ , it follows that

(1) kxk∞ ≥ 0 for every x ∈ CN .

kxk∞ = 0 equivalently max{|x1 |, . . . , |xN |} = 0 if and only if xi = 0 for every i = 1
to N equivalently x = 0.
(2) The triangle inequality may proved as follows, for x, y ∈ CN ,

kx + yk∞ = max{|xi + yi | : i = 1 . . . , N }

≤ max{|xi | + |yi | : i = 1 . . . , N }

≤ max{|xi | : i = 1 . . . , N } + max{|yi | : i = 1 . . . , N }

= kxk∞ + kyk∞ .

(3) For α ∈ C and x ∈ CN

kαxk∞ = max{|αx1 |, . . . , |αxN |} = |α| max{|x1 |, . . . , |xN |} = |α|kxk∞ .

3. The normed linear space of sequences: The following are some vector space of
sequence of complex numbers. For a complex sequence x = (x1 , x2 , . . .), define

kxk∞ := sup{|xi | : i = 1, 2, . . .}.

For 1 ≤ p < ∞,

kxkp := (|x1 |p + |x2 |p + · · · )1/p .

`∞ := {x = (x1 , x2 , . . .) : kxk∞ < ∞};

c := {x = (x1 , x2 , . . .) : x converges in C};

c0 := {x = (x1 , x2 , . . .) : x converges to 0};

`p := {x = (x1 , x2 , . . .) : kxkp < ∞, 1 ≤ p < ∞};

c00 := {x = (x1 , x2 , . . .) : x has only finite nonzero entries}.

6 KRISHNA KUMAR. G

We have the following inclusion relation,

c00 ⊂ `p ⊂ c0 ⊂ c ⊂ `∞ .

The following are normed linear space of sequences.

1. `∞ with norm is k · k∞ ;
2. c with norm is k · k∞ ;
3. c0 with norm is k · k∞ ;
4. `p with norm is k · kp ;
5. `p with norm is k · k∞ ;
6. c00 with norm is k · k∞ ;
7. `p with norm is k · kp , 1 ≤ p < ∞;
8. c00 with norm is k · kp , 1 ≤ p < ∞.

Proposition 1.3. Jensen’s inequality Let 1 ≤ p < r < ∞, then `p ⊆ `r .

Proof. Let x ∈ `p and kxkp = (|x1 |p + |x2 |p + · · · )1/p ≤ 1. Then |xi | ≤ 1 for every i. Since
1 ≤ p < r < ∞, |xi |r ≤ |xi |p and (|x1 |r + |x2 |r + · · · )1/r ≤ (|x1 |p + |x2 |p + · · · )1/p implies
kxkr ≤ kxkp , so kxkr ≤ 1. Now consider x ∈ `p then the norm of x/kxkp is equal to 1. Thus
x kxkr
= ≤1
kxkp r kxkp
This implies kxkr ≤ kxkp < ∞ and x ∈ `r . 

The following theorem gives some methods of obtaining new linear spaces from the old.
This is endowed with a norm to become a normed linear space.

Remark 1.4. Let X be a normed linear space and Y be a subspace of X. The continuity
of addition and scalar multiplication shows that Y is a subspace of X. Thus Y and Y are
normed linear spaces with the induced norm.

The quotient space

Let X be a normed linear space and Y be a closed subspace of X. Define a relation on X

as follows. Let x, y ∈ X, then

x ∼ y if x − y ∈ Y.
 NORMED LINEAR SPACE 7

The relation defined above is reflexive, symmetric and transitive and is an equivalence rela-
tion. The equivalence class of an element x is defined by

[x] = {y ∈ X : y ∼ x}

= {y ∈ X : y − x ∈ Y }

= {y ∈ X : y ∈ x + Y }

= x + Y.

The set of all equivalence classes under this equivalence relation is denoted as X/Y .

X/Y := {x + Y : x ∈ X}.

Theorem 1.5. Let (X, k · k) be a normed linear space and Y be a closed subspace of X.
Then X/Y with norm |||x + Y ||| = inf{kx + yk : y ∈ Y } is a normed linear space called
quotient space.

Proof. The operations defined in X/Y is

(x1 + Y ) + (x2 + Y ) = x1 + x2 + Y ;

α(x + Y ) = αx + Y.

It is an easy exercise to show that X/Y is a vector space. Note that the equivalence class of
the element 0 is 0 + Y = Y is the zero element in the quotient space.

(1) From the definition it follows that, |||x + Y ||| ≥ 0.

If |||x + Y ||| = inf{kx + yk : y ∈ Y } = 0, then there exist a sequence yn ∈ Y such
that x + yn → 0; that is yn → −x. Since Y is closed −x ∈ Y , that is, x + Y = Y the
zero element in X/Y . Also

|||0 + Y ||| = inf{kyk : y ∈ Y } = 0.

(2) Let x1 , x2 ∈ X and  > 0, then there exist y1 , y2 ∈ Y such that

 
kx1 + y1 k < inf{kx1 + yk : y ∈ Y } + = |||x1 + Y ||| + ,
2 2
 
kx2 + y2 k < inf{kx2 + yk : y ∈ Y } + = |||x2 + Y ||| + .
2 2
Then
kx1 + y1 + x2 + y2 k ≤ |||x1 + Y ||| + |||x2 + Y ||| + .
8 KRISHNA KUMAR. G

Since y1 + y2 ∈ Y , we see that

|||x1 + x2 + Y ||| ≤ |||x1 + Y ||| + |||x2 + Y ||| + ,

for every  > 0. This shows that

|||(x1 + Y ) + (x2 + Y )||| ≤ |||x1 + Y ||| + |||x2 + Y |||.

(3) Further, for x ∈ X, and 0 6= k ∈ C,

|||k(x + Y )||| = |||kx + Y ||| = inf{kkx + yk : y ∈ Y }

y
= inf{|k|kx + k : y ∈ Y }
k
= |k| inf{kx + yk : y ∈ Y }

= |k| |||x + Y |||.

Theorem 1.6. A sequence (xn + Y ) converges to x + Y in X/Y if and only if there is a

sequence (yn ) in Y such that (xn + yn ) converges to x in X.

Proof. Let (xn +Y ) be a sequence in X/Y . If (yn ) is a sequence in Y such that (xn +yn ) → x
in X, then
|||(xn + Y ) − (x + Y )||| = |||xn − x + Y ||| ≤ kxn − x + yn k

for every n, so that (xn + Y ) converges to x + Y in X/Y . Conversely, assume that xn + Y →

x + Y in X/Y . Since

|||(xn + Y ) − (x + Y )||| = inf{kxn − x + yk : y ∈ Y },

choose yn ∈ Y such that

1 1
kxn − x + yn k ≤ inf{kxn − x + yk : y ∈ Y } + = |||(xn + Y ) − (x + Y )||| +
n n
for n = 1, 2, . . . Then it follows that xn − x + yn → 0, i.e., xn + yn → x in X as n → ∞. 

Theorem 1.7. Let (Xj , k · kj ), j = 1, . . . , N be N normed linear spaces. For fix 1 ≤ p ≤ ∞,

define a norm on the product space X1 × · · · × XN as
(
(kx1 kp1 + · · · + kxkpN )1/p , if 1 ≤ p < ∞
kxkp =
max{kx1 k1 , . . . , kxN kN }, if p = ∞.
 NORMED LINEAR SPACE 9

Then k · kp is a norm on X.
A sequence (x(n)) converges to x in X1 × · · · × xN if and only if (x(n)j ) converges to xj in
Xj for every j = 1, . . . , N .

Proof. The following are the notations used. If x ∈ X1 × · · · × XN then x = (x1 , x2 , . . . , xN )

with xi ∈ Xi . X1 × · · · × XN is also a vector space under the following operations. For
x, y ∈ X1 × · · · × XN and α ∈ C,

x + y = (x1 + y1 , . . . , xN + yN );

αx = (αx1 , . . . , αxN ).

From the definition it follows that,

(1) kxkp ≥ 0 for every fixed 1 ≤ p ≤ ∞

kxkp = 0 if and only if kxi ki for every i = 1, . . . , N if and only if xi = 0 for every
i = 1, . . . , N if and only if x = 0.
(2) From the definition, for 1 ≤ p < ∞,

kx + ykp = (kx1 + y1 kp1 + · · · + kxN + yN kpN )1/p

≤ [(kx1 k1 + ky1 k1 )p + · · · + (kxN kN + kyN kN )p ]1/p

(since each k · ki is a norm in Xi )

N
!1/p N
!1/p
X X
≤ kxi kpi + kyi kpi
i=1 i=1
(follows from the Minkowski’s inequality)

= kxkp + kykp .

It is also clear from the definition that, for every x, y ∈ X1 × · · · × xN ,

kx + yk∞ ≤ kxk∞ + kyk∞ .

(3) We also have, for 1 ≤ p ≤ ∞

kαxkp = |α|kxkp

It follows from the definition that, the sequence (x(n)) converges to x in X1 × · · · × xN if

and only if (x(n)j ) converges to xj in Xj for every j = 1, . . . , N . 
10 KRISHNA KUMAR. G

Riesz Lemma: Let X be a normed space, Y be a closed subspace of X and Y 6= X. Let

r be a real number such that 0 < r < 1. Then there exists some xr ∈ X such that

kxr k = 1 and r ≤ dist(xr , Y ) ≤ 1.

Proof. Since Y 6= X, consider x ∈ X such that x ∈

/ Y . As the subspace Y is closed, we have
dist(x, Y ) > 0. Also r < 1, there exists some y0 ∈ Y such that kx − y0 k ≤ dist(x, Y )/r. Let
x − y0
xr = .
kx − y0 k
Clearly xr = 1. Since 0 ∈ Y , we see that dist(xr , Y ) ≤ 1. Also, since y0 ∈ Y , we have
dist(x − y0 , Y ) dist(x, Y )
dist(xr , Y ) = = ≥ r.
kx − y0 k kx − y0 k


Lemma 1.8. Let X be a normed space and Y be a subspace of X.

(1) For x ∈ X, y ∈ Y and k ∈ C, we have

kkx + yk ≥ |k|dist(x, Y ).

(2) If Y is finite dimensional. Then Y is complete. In particular it is closed in X.

(3) If Y is finite dimensional and let {y1 , . . . , ym } be a basis for Y and xn be a sequence
in Y . If
xn = kn,1 y1 + · · · + kn,m ym , n = 1, 2, . . . ,

then xn converges to x = k1 y1 + · · · + km ym if and only if kn , j → kj as n → ∞

for each j = 1, . . . , m. Also (xn ) is bounded if and only if (kn,j ) is bounded for each
j = 1, . . . , m.

Proof. (1) If k = 0, the result is obvious. If k 6= 0, then

y
kkx + yk = |k|kx + k ≥ |k|dist(x, Y ),
k
since −y/k ∈ Y .
(2) To prove the completeness of a finite dimensional subspace Y of X, we use mathe-
matical induction on the dimension m of Y . Let m = 1. Then Y = {ky : k ∈ C}
with y 6= 0. If xn is a Cauchy sequence in Y with xn = kn y, then for all n and p,

|kn − kp |kyk = kxn − xp k.

 NORMED LINEAR SPACE 11

Hence (kn ) is a Cauchy sequence in C, which is complete. If kn → k in C, then

xn → ky in Y . Thus Y is complete. Assume now that every m − 1 dimensional
subspace of X is complete. Let Y be an m dimensional subspace of X and (xn ) be a
Cauchy sequence in Y . Then the m − 1 dimensional subspace Z = span{y2 , . . . , ym }
is complete by the inductive assumption. Now for each n = 1, 2, . . .,

xn = kn y1 + zn

for some kn ∈ C and zn ∈ Z. By (1),

kxn − xp k = k(kn − kp )y1 + zn − zp k ≥ |kn − kp |dist(y1 , Z).

Since y1 ∈
/ Z and Z is closed in X, we see that dist(y1 , Z) > 0. Hence (kn ) is a
Cauchy sequence in C. As zn = xn − kn y1 , it follows that (zn ) is a Cauchy sequence
in Z, which is complete. If kn → k in C and zn → z in Z, we see that xn → ky1 + z
in Y . Thus Y is complete. In particular Y is closed in X.
(3) Let xn ∈ Y and xn = kn,1 y1 + · · · + kn,m ym , n = 1, 2, . . .. If kn,j → kj as n → ∞ for
each j = 1, . . . , m, let x = k1 y1 + · · · + km ym . Then

kxn − xk ≤ |kn,1 − k1 |ky1 k + · · · + |kn,m − km |kym k → 0,

as n → ∞. Thus xn → x.
Conversely, let xn → x = k1 y1 + · · · + km ym . Then by (1) above,

kxn − xk = k(kn,1 − k1 )y1 + · · · + (kn,m − km )ym k

≥ |kn,j − kj |dist(yj , Yj ),

where Yj = span{yi : i = 1, . . . , m, i 6= j} for j = 1, . . . , m. Since yj ∈

/ Yj and Yj
is closed in X, we see that dist(yj , Yj ) > 0. Hence kn,j → kj as n → ∞ for each
j = 1, . . . , m.
Finally, if kn,j is bounded for each j = 1, . . . , m, say |kn,j | ≤ αj for all n = 1, 2, . . .,
then
kxn k ≤ α1 ky1 k + · · · + αm kym k

for all n = 1, 2, . . ., that is, (xn ) is bounded. Conversely, let (xn ) be bounded. For
each j = 1, . . . , m, let Yj be defined as above. Then

kxn k = kkn,1 y1 + · · · + kn,m ym k ≥ |kn,j |dist(yj , Yj ),

12 KRISHNA KUMAR. G

by (1) above. Hence (kn,j ) is bounded for each j = 1, . . . , m.



Remark 1.9. An infinite dimensional subspace of a normed linear space X may not be
closed in X. For example consider X = `∞ and Y = c00 . Let xn = (1, 1/2, . . . , 1/n, 0, 0, . . .)
then xn → x = (1, 1/2, 1/3, . . .) and x ∈
/ c00 .

Theorem 1.10. Let X be a normed space. The following conditions are equivalent.
(i) Every closed and bounded subset of X is compact.
(ii) The subset {x ∈ X : kxk ≤ 1} is compact.
(iii) X is finite dimensional.

Proof. Since {x ∈ X : kxk ≤ 1} is closed and bounded, (i) implies (ii).

Suppose now that (ii) holds. Let, if possible, {z1 , z2 , . . .} be an infinite linearly independent
subset of X and consider

Zn = span{z1 , . . . , zn }, n = 1, 2, . . .

Being finite dimensional, Zn is closed subspace of Zn+1 . Also Zn 6= Zn+1 , since {z1 , . . . , zn+1
is linearly independent. By the Riesz lemma, there is some xn ∈ Zn+1 , such that
1
kxn k = 1 and dist(xn , Zn ) ≥ .
2
Now (xn ) is a sequence in {x ∈ X : kxk ≤ 1} and having no convergent subsequence, because
kxn − xm k ≥ 1/2 for all m 6= n. Hence the set {x ∈ X : kxk ≤ 1} cannot be compact. Thus
(ii) implies (iii).
Finally, suppose that X is finite dimensional and let E be a closed and bounded subset of
X. To prove that E is compact, we show that every sequence in E has a subsequence which
converges in E. Consider a sequence (xn ) in E. Let {y1 , . . . , ym } be a basis for X, and

xn = kn,1 y1 + · · · + kn,m ym , n = 1, 2, . . .

Then (kn,j ) is bounded for each j = 1, . . . , m. By the Bolzano-weierstrass theorem for C and
passing to a subsequence of a subsequence several times, we find n1 < n2 < · · · such that
knp ,j converges in C as p → ∞ for each j = 1, . . . , m. Thus (xnp ) converges to some x ∈ X.
Since xnp ∈ E and E is closed, we have x ∈ E, that is, (xnp ) converges in E. This shows
that (iii) implies (i). 
 NORMED LINEAR SPACE 13

Definition 1.11. Let X be a normed linear space and Y ⊆ X. If for every y1 , y2 ∈ Y and
0 < t < 1 implies ty1 + (1 − t)y2 ∈ Y , then Y is said to be a convex subset of X.

Let x ∈ X and r > 0. Define the open ball with center x and radius r as

U (x, r) = {y ∈ X : kx − yk < r}.

Define the closed ball with center x and radius r as

U (x, r) = {y ∈ X : kx − yk ≤ r}.

Then U (x, r) = U (x, r). It is an easy exercise to show that U (x, r) and U (x, r) are convex
subset of X.

Theorem 1.12. Let X be a normed space.

(1) If E1 is open in X and E2 ⊆ X, then E1 + E2 is open in X.

(2) If E be a convex subset of X. Then the interior E ◦ of E and the closure E are also
convex. If E ◦ 6= ∅, then E = E ◦ .
(3) Let Y be a subspace of X and Y ◦ 6= ∅ if and only if Y = X.

Proof. (1) Let x ∈ X and x1 ∈ E1 . Since E1 is open, there is some r > 0 such that
U (x, r) ⊆ E1 . But U (x1 + x, r) = U (x1 , r) + x ⊆ E1 + x. Hence E1 + x is open for
every x ∈ X. Since
[
E1 + E2 = {E1 + x : x ∈ E2 },

it follows that E1 + E2 is open.

(2) Let 0 < t < 1. Since E is convex tE ◦ + (1 − t)E ◦ ⊆ E. Also tE ◦ + (1 − t)E ◦ is open.
Hence tE ◦ + (1 − t)E ◦ ⊆ E ◦ , that is E ◦ is convex.
Next, let x, y ∈ E. Find sequences (xn ) and (yn ) in E such that xn → x and yn → y.
Then txn + (1 − t)yn → tx + (1 − t)y, i.e., tx + (1 − t)y ∈ E. Thus E is convex.
Now assume that E ◦ 6= ∅. Clearly E ◦ ⊆ E. To show E ⊆ E ◦ , it is enough to show
that E ⊆ E ◦ . Consider a ∈ E ◦ . Let r > 0 be such that U (a, r) ⊆ E, that is,
a + ry ∈ E for every y ∈ X with kyk < 1. Let x ∈ E and xt = ta + (1 − t)x. We
claim that U (xt , tr) ⊆ E, that is, xt + try ∈ E for every y ∈ X with kyk < 1. This
is obvious from the convexity of E, because xt + try = t(a + ry) + (1 − t)x, where
a + ry ∈ E and x ∈ E. Thus xt ∈ E ◦ . Since xt → x as t → o, we have x ∈ E ◦ .
14 KRISHNA KUMAR. G

(3) If Y = X, then Y ◦ = X 6= ∅. Conversely, let Y ◦ 6= ∅. Consider a ∈ Y ◦ and let r > 0

be such that U (a, r) ∈ Y . Then U (0, r) = U (a, r) − a ∈ Y . Let x = 0, then clearly
rx
x ∈ Y . If x 6= 0, then kxk
∈ U (0, r) ⊆ Y . Hence x ∈ Y as well.


Definition 1.13. A normed space X is said to be strictly convex if for x 6= y in X with

kxk = 1 = kyk, we have kx + yk < 2.

Definition 1.14. Consider k · k, k · k0 and the associated metrics d(x, y) = kx − yk, d0 (x, y) =
kx − yk0 on a linear space X. The norm k · k is said to be stronger than the norm k · k0 if the
metric d is stronger than the metric d0 . The norm k · k and k · k0 are said to be comparable
if one of them is stronger than the other.

Theorem 1.15. Let k · k and k · k0 be norms on a linear space X. Then the k · k is stronger
than the norm k · k0 if and only if there is some α > 0 such that kxk0 ≤ αkxk for all x ∈ X.
Further the norm k · k is equivalent to the norm k · k0 if and only if there are α > 0 and β > 0
such that βkxk ≤ kxk0 ≤ αkxk for all x ∈ X.

Proof. Let k · k be stronger than k · k0 . Then there is some r > 0 such that

0 ∈ {x ∈ X : kxk < r} ⊆ {x ∈ X : kxk0 < 1}.

rx
Let 0 6= x ∈ X and  > 0. Since k (1+)kxk rx
k < r, it follows that k (1+)kxk k0 < 1, that is,
rkxk0 ≤ (1 + )kxk. As this is true for every  > 0, it follows that kxk0 ≤ αkxk with α = 1/r.
Conversely, let kxk0 ≤ αkxk for all x ∈ X. Let (xn ) be a sequence in X such that d(xn , x) =
kxn − xk → 0. Since kxn − xk0 ≤ αkxn − xk, we see that d0 (xn , x) = kxn − xk0 → 0. Hence
the metric d (resp., the norm k · k) is stronger than the metric d0 (resp., the norm k · k0 ).
The result about equivalence of norms follows immediately. 
 NORMED LINEAR SPACE 15

Exercise

1. Show that CN for N ≥ 2 is not strictly convex for the norms k · k1 and k · k∞ .
2. Show that CN for N ≥ 2 is strictly convex for the norm k · k2 .
3. Let x ∈ CN , then show that
kxk∞ ≤ kxk2 ≤ kxk1
and
√
kxk1 ≤ N kxk2 ≤ N kxk∞ .
4. Let k · k be a norm on a linear space X. If x, y ∈ X and kx + yk = kxk + kyk, then
ksx + tyk = skxk + tkyk for all s, t ≥ 0.
5. Let X = C3 . For x = (x1 , x2 , x3 ), let
h i1/3
2 2 3/2 3


kxk = |x1 | + |x2 | + |x3 | .

Show that k · k is a norm on C3 .

6. Let X1 be a closed subspace and X2 be a finite dimensional subspace of a normed space
X. Then X1 + X2 is closed in X.
7. Let X = C([a, b]). For x ∈ X, let
Z b
kxk1 = |x(t)| dt
a
and
kxk∞ = sup{|x(t)| : t ∈ [a, b]}.
Show that k · k1 and k · k∞ are norms on X. Further show that k · k∞ is stronger than but
not equivalent to k · k1 .
8. Let X denote the linear space of all polynomials in one variable with coefficients in C.
For p ∈ X with p(t) = a0 + a1 t + · · · + an tn , let

kpk = sup{|p(t)| : 0 ≤ t ≤ 1},

kpk1 = |a0 | + · · · + |an |,

kpk∞ = max{|a0 |, · · · , |an |}.
Then show that
(a) k · k, k · k1 and k · k∞ are norms on X;
(b) kpk ≤ kpk1 and kpk∞ ≤ kpk1 for all p ∈ X;
(c) The norms k · k and k · k1 are not equivalent;
16 KRISHNA KUMAR. G

(d) The norms k · k and k · k∞ are not comparable.

Department of Mathematics, University of Kerala, Kariavattom Campus, Thiruvanan-

thapuram, Kerala, India, 695581.
E-mail address: [email protected]