FUNCTIONAL ANALYSIS

Shallu Sharma
Department of Mathematics, University of Jammu
Def :- Metric Space
Let X be a non-empty set. Then a function d : X × X −→ R is said to be a
metric on X if it satisfies the following conditions:
(i) d(x, y) ≥ 0, ∀x, y ∈ X
(ii) d(x, y) = 0 iff x = y
(iii) d(x, y) = d(y, x), ∀x, y ∈ X
(iv) d(x, y) ≤ d(x, z) + d(z, y), ∀x, y, z ∈ X
If d is a metric on X, then the ordered pair (X, d) is called the metric space.

Example :- Let X be an arbitrary non-empty set.

Define a function
{ d : X × X −→ R as
0 if x = y
d(x, y) =
1 if x ̸= y
Then d is a metric on X, called the discrete metric on X and the pair (X, d)
is called a discrete metric space.

Example :- Let X = R, the set of real numbers.

Define a function d : X × X −→ R as

d(x, y) = |x − y|, ∀x, y ∈ X

Then d is a metric on X = R, called the usual metric on R and X with this

metric is called a usual metric space.

Counter Example :- Let X = R. Define a function d : X × X −→ R as

d(x, y) = |x2 − y 2 |, ∀x, y ∈ X.

Then d is not a metric on X = R because (2) condition is not satisfied.

Remark :- If in the above definition (2) condition is dropped, the the above
function is real valued function ’d’ is called a Pseudometric on X i.e. for a
pseudometric, we have x = y ⇒ d(x, y) = 0 but d(x, y) = 0 ⇒ x = y.

1
Example :- Let X = Rn , set of all ordered n-tuples of real numbers for fixed
n ∈ N.
Define d : X × X −→ R as
∑
n 1
d(x, y) = ( (xi − yi )2 ) 2 , ∀x = (x1 , ..., xn ), y = (y1 , ..., yn ) ∈ Rn .
i=1

Then d is a metric on X and (X, d) is called the Real Euclidean n-space(or

simply n-dimensional Euclidean Space).

Example :- Let X = Cn , set of all ordered n-tuples of complex numbers.

Define Define d : X × X −→ R as
√
d(z, w) = |z1 − w1 |2 + |z2 − w2 |2 + ... + |zn − wn |2 , ∀z =
(z1 , z1 , ..., zn ), w = (w1 , w2 , ..., wn ) ∈ Cn .

Then d is a metric on X and (X, d) is called the Complex Euclidean Space.

HOLDER’S INEQUALITY
If xi , yi (i = 1, ..., n) are non-negative real numbers, then

∑
n ∑
n 1 ∑
n 1
|xi yi | ≤ ( |xi |p ) p ( |yi |q ) q , where p > 1 and 1
p
+ 1
q
= 1.
i=1 i=1 i=1

MINSKOWSKI INEQUALITY
If p ≥ 1 and xi , yi (i = 1, 2, ..., n) are non-negative real numbers, then

∑
n 1 ∑
n 1 ∑
n 1
( (xi + yi )p ) p ≤ ( (xi )p ) p + ( (yi )p ) p
i=1 i=1 i=1

Particular case of Holder’s Inequality

If p = q = 2 in Holder’s Inequality, then we have
∑
n ∑
n 1 ∑
n 1
|xi yi | ≤ ( |xi |2 ) 2 ( |xi |2 ) 2
i=1 i=1 i=1

This inequality is known as Cauchy Schwartz Inequality.

For 1 ≤ p < ∞, consider X = lp , the set of sequences x = {xn } such

∑
∞ 1
that |xn |p ) p < ∞. Define
n=1

2
 ∑
∞ 1
d(x, y) = ( |xn − yn |p ) p , ∀x = {xn }, y = {yn } ∈ lp .
n=1

Then d is a metric on X and (X, d) or (lp , d) is a metric space.

Example :- Let X = l∞ , the set of all bounded sequences of complex num-

bers. Let x = {xn } and y = {yn } be any two points of l∞ .
Define d : X × X −→ R as

d(x, y) = sup{|xn − yn | : n ∈ N}.

Then d is a metric on X and (X, d) i.e. (l∞ , d) is a metric space.

Example :- Let X = C[a, b], the set of all continuous real valued functions
defined on closed and bounded interval [a, b].
Define d : X × X −→ R as

d(f, g) = sup{|f (t) − g(t)|}, t ∈ C[a, b].

Then d is a metric on X and (X, d) i.e. (C[a, b], d) is a metric space.

Def :- Let (X, d) be a metric space and let r be any positive number and
x ∈ X. Then the set S(x, r) = {y ∈ X : d(x, y) < r} is called a Open Ball
In A Metric Space.

Example :- Consider the real line R and let d be the usual metric on R.
Then each open interval is an open ball in R with usual metric and each
closed interval is a closed ball in the usual metric space.

Example{:- Let d be a discrete metric on X and let p ∈ X and δ > 0. Then

{p} ; δ ≤ 1
S(p,δ) =
X ;δ>1

Example :- In R2 . Let p = (0, 0) ∈ R2 .

√
(i) d(x, y) = (x1 − y1 )2 + (x2 − y2 )2
√
(ii) d(x, y) = |x1 − y1 | + |x2 − y2 |
(iii) d(x, y) = max{|x1 − y1 , |x2 − y2 |}, ∀x = (x1 , x2 ), y = (y1 , y2 ) ∈ R2 .

3
Def :- Open Set in a Metric Space
Let (X, d) be a metric space. Then a subset A of X is said to be open if for
each x ∈ A, ∃r > 0 such that B(x, r) ⊆ A.

RESULTS :- In a metric space,

(1) Every open sphere or open ball is a open set.
(2) Union of open set is open.
(3) Finite intersection of open set is open.
(4) The empty set Φ and the whole set X are open in X.

Example :- Consider the set R of real numbers with usual metric d(x, y) =
|x − y|, ∀x, y ∈ R.
Find whether or not the following sets are open.
A = [0, 1), B = (0, 1), C = (0, 1], D = [0, 1], E = (0, 1) ∪ (2, 3), F = {1},
G = {1, 2, 3}.
Clearly A,C,D,F and G are not open in R. B and E are open in R with usual
metric.

Exercise :- Show that in a discrete metric space, every set is open.

Closed set in a Metric Space :-

Let (X, d) be a metric space. A subset A of X is said to be closed if its
complement is open in X.

Result :- In a Metric space (X, d), every closed space is a closed set.
Arbitrary intersection of closed set is closed.
Finite union of closed sets is closed.
Empty set ϕ and X are closed in X.

Def :- Let (X, d) be a metric space and E be a subset of X. Then a point

x ∈ E is said to be an interior point of E if ∃ r > 0 such that B(x, r) ⊆ E.
The set of all the interior points of E is denoted by E o and is called the
interior point of E.

Remark :-
1) E o is an open set.

4
2) E o is the largest open set contained in the set.
3) E is open iff E o = E.

Def :- Accumulation point or Limit point or Cluster point

Let (X, d) be a metric space and A ⊆ X. Then a point xo ∈ X is called the
limit point ( or accumulation point or cluster point) of A if every ball around
xo contains a point of A distinct from xo . The set of all limit points of A is
′
called the derived set of A and is denoted by A or d(A).

Note :-
1) The point xo may or may not be a point of A.
2) A subset E ⊂ X is said to be closed if it contains all its limit points.

Example :- (1) The set of all the limit points of (0, 1) in R with usual metric
is [0, 1].
(2) The set of natural numbers N as a subset of R with usual metric has no
limit point.

Def :- Closure of a subset in a Metric space

Let A be a subset of metric space (X, d) . Then the closure of A is denoted
by A and is defined as

∩
A= {K : K is closed and A ⊆ K}

i.e, A is the smallest closed set containing A.

Note :- A subset A of a metric space is closed if A = A.

Def :- A subset M of a metric space (X, d) is said to be dense in X if M = X.

i.e, every point of X is a point of M or a limit point of M .

Example :- The set Q of rational numbers in R with usual metric d is dense

in R.

Def :- A metric space X is said to be separable if it contains a countable

dense set.

5
Example :- (1) Consider the real line R with usual metric d, then the set of
rational numbers is dense in R. Therefore R with usual metric d is separable.
(R, the set of real numbers is separable because Q is dense in R)
(2) C is separable because the set of all complex numbers whose real and
imaginary parts are both rationals is a countable set which is dense in C.
(3) C [a, b] is separable.

Def :- Let (X, d) be a metric space. If Y is a subset of X , then a metric d

on X induces a metric on Y . A subset F ⊂ Y is open (respectively closed)
in Y w.r.t. d iff F = E ∩ Y , where E is open (respectively closed) in X.

Note:- A subset of Y which is open in X is also open in Y . But the converse

may not be true.
For this, we can consider an example. Let X = R with usual metric space d
and suppose Y = [0, 1] ⊆ R.
Clearly A = [0, 12 ) is open in Y but A is not open in X.

Def :- Let (X, d) be any metric space. A sequence {xn } in X is said to

converge to x ∈ X, if for each ϵ > 0, ∃ a positive integer n0 such that

d (xn , x) < ϵ ∀ n ≥ n0
i.e, d (xn , x) → 0 as n → ∞.

[or, if for each open sphere B(x, ϵ) centered at x, ∃ a positive integer n0 such
that xn ∈ B(x, ϵ), ∀ n ≥ n0 ]

Note :- If a sequence {xn } converges to x, then we can write xn → x or

limn→∞ xn = x.

Def :- Let (X, d) be a metric space. The sequence {xn } in X is said to be

Cauchy sequence if for each ϵ > 0, ∃ a a positive integer n0 such that

d(xn , xm ) < ϵ, ∀ n, m ≥ n0
i.e, d(xn , xm ) → 0 as n, m → ∞

6
Theorem :- Every convergent sequence in a metric space is a Cauchy se-
quence.
Proof :- To show that every convergent sequence in a metric space is a
Cauchy sequence. Let (X, d) be an arbitrary metric space and suppose {xn }
be a sequence in X such that xn → x. we have to show that xn is a Cauchy se-
quence. Therefore, by the definition of a convergence of a sequence we can say
that for every ϵ > 0, ∃ a positive integer n0 such that d(xn , x) < 2ϵ , ∀ n ≥ n0 .
Now, for n, m ≥ n0 , we have

d(xn , xm ) ≤ d(xn , x) + d(x, xm )

ϵ ϵ
< +
2 2
= ϵ

i.e, d(xn , xm ) < ϵ.

Thus, we have prove that for every ϵ > 0, ∃ n0 ∈ N such that d(xn , xm ) <
ϵ for n, m ≥ n0 .
This implies that {xn } is a Cauchy sequence in a metric space (X, d). Since xn
was an arb. convergent sequence in a metric space (X, d). So we can conclude
that every convergent sequence in a metric space is a Cauchy sequence.

Note :- The converse of the above theorem is not true i.e, Every Cauchy
sequence in a metric space (X, d) need not be convergent in X.
For this, we can consider an example
Example :- Let X = [0, 1] with usual metric defined as

d(x, y) = |x − y|, ∀ x, y ∈ X

Consider a sequence { n1 }∞
n=1 in X.
{ n1 }∞
n=1 is a Cauchy sequence:- Let ϵ > 0 be given. If we choose a positive
integer n0 > 2ϵ . Then for n, m ≥ n0 , we have

d(xn , xm ) = |xn − xm |
1 1
= | − |
n m
1 1
= +
n m

7
 1 1
≤ +
n0 n0
2 2
= [ because, no > ]
n0 ϵ

This shows that { n10 }∞ 1 ∞

n=1 is a Cauchy sequence in X. But the above { n0 }n=1
is not convergent in X because the point 0 to which the above sequence
converges is not a point of X.

Exercise :- Let (x, d) be a metric space. Show that every convergent sequence
is bounded and its limit point is unique.

Theorem :- Let M be a non-empty subset of a metric space (X, d). Then

(i) x ∈ M iff there is a sequence {xn } in M such that xn → x.
(ii) M is closed iff there is a sequence {xn } in M such that xn → x ⇒ x ∈ M.
Proof :- (i) Let x ∈ M , we have two cases:
Case I: If x ∈ M , then a sequence {xn }∞
n=1 ; xn = x, ∀ n ({x, x, x, ..} a
constant sequence).
Clearly, xn → x
Case II: If x ∈
/ M.
⇒ x ∈ M′
⇒ x is a limit of M .
Then, for each n = 1, 2, 3..., the B(x, n1 ) contains xn in M and xn → x because
1
n
→ 0 as n → ∞. This proves the direct part.
Conversely, suppose {xn } be sequence in M such that xn → x. We have to
show that x ∈ M .
We have two cases:
Case (i) : If x ∈ M ⇒ x ∈ M . [because M ⊆ M ]
Case (ii) : If x ∈
/ M , then each ball centered at x contains a point xn other
than x such that x is a limit point of M

⇒ x ∈ M′
⇒x∈M

(ii) Firstly, suppose that M is closed ⇒ M = M . We have to show that

8
there is a sequence {xn } in M such that xn → x ⇒ x ∈ M .
For this, let {xn } be arbitrary sequence in M such that xn → x then by part
(i) we have x ∈ M ⇒ x ∈ M . [because M = M ]
Conversely, suppose there is a sequence {xn } in M such that xn → x implies
x ∈ M . We have to show that M is closed i.e, M = M .
By definition of closure of a set, we have M ⊆ M ——(∗)
Next, let x ∈ M , then by part (i) there is a sequence {xn } in M such that
xn → x

⇒ x ∈ M [by given condition]

⇒ M ⊆ M ——(∗∗)

From (∗) and (∗∗), we get

M =M

This proves that M is closed.

Def :- Let (X, d) be a metric space, then X is said to be complete if every

Cauchy sequence in X converges in X.

Example :- The discrete space (X, d) is the complete metric space. Because,
for this space Cauchy sequence must be a constant sequence. (i.e., must
consist of a single point repeated from some places on and so converges in X)

Theorem :- A subspace M of a complete metric space X is complete iff M

is closed in X.
Proof :- Suppose M is complete. We have to show that M is closed in X.
i.e, M = M .
Clearly, by definition of closure of set, M ⊆ M ——(∗)
Next, let x ∈ M , then there is a sequence {xn } in M such that xn → x.
Since {xn } is Cauchy sequence in M and M is complete, it follows that {xn }
converges in M ⇒ x ∈ M .
We have M ⊆ M ——(∗∗)
From (∗) and (∗∗), we have

9
M is closed in X.
Conversely, suppose M is closed in a metric space (X, d). We have to show
that M as a subspace of a complete metric space is complete.
For this, let {xn } be a Cauchy sequence in M . Since M ⊆ X, therefore {xn }
is a Cauchy sequence in X.
Since metric space (X, d) is complete, therefore the above sequence {xn }
converges to some point x ∈ X

i.e, xn → x ∈ X
⇒x∈M
⇒ x ∈ M [because M = M ]

This proves that Cauchy sequence in M converges to x. Thus every Cauchy

sequence in M converges in M . We can conclude that M is complete.

Exercise :- If xn → x and yn → y. Then d(xn , yn ) → d(x, y).

Proof :- (Try yourself)

Theorem :- Prove that the space lp , 1 ≤ p < ∞ is separable.

Proof :- To show lp , 1 ≤ p < ∞ is separable, we have to show that ∃ a
countable dense set in lp .
For this, let ei = {0, 0, ..., 1, 0, ...} where 1 occurs at the ith place and consider
E = {k1 e1 + k2 e2 + ... + kn en : n = 1, 2, ...} where Re(ki )′ s and Im(ki )′ s are
[
rationals ∀i. ]
∑
n ∑
∞ ∑
n ∑
∞
E = {{xi } : xi = ki ei and xi = 0 f or i > n}. Also, |xi | =
p
|xi | +
p
|xi | < ∞
p
i=1 i=1 i=1 i=n+1
Clearly E ∈ lp .
Since the set of rational numbers is countable, it follows that E is countable
in lp .
Now we want to show that E is dense in lp i.e, E = lp .
For this, we will show that for every ϵ > 0 and ∀ x ∈ lp , ∃ y ∈ E such that
d(x, y) < ϵ.
So, let x = {xi} ∈ lp and ϵ > 0.
∑
∞ ∑
∞
ϵp
As |xi |p < ∞, there is some n ∈ N such that |xi |p < 2
. [because, a
i=1 i=n+1
tail of the convergent series and by def. of absolute convergence]

10
Again, since the set of complex numbers where real and imaginary parts are
dense in C.
Therefore, ∀xi : i = 1, 2, ..., n; ∃ki ∈ C (i.e, k1 , k2 , ..., kn ) with real and imag-
inary parts of ki′ s being rationals and |xi − yi |p < ϵp
2n
,i = 1, 2, ..., n
(Just because ki′ s are arb. close to xi )
Consider y = k1 e1 + k2 e2 + ... + kn en ∈ E, then

∑
∞
[d(x, y)] p
= |xi − yi |p
i=1
∑n ∑
∞
= |xi − yi | +
p
|xi − yi |p
i=1 i=n+1
∑n ∑∞
= |xi − yi |p + |xi |p
i=1 i=n+1
nϵp ϵp
< +
2n 2
ϵp ϵp
= +
2 2
p
= ϵ

Therefore, [d(x, y)]p < ϵp ⇒ d(x, y) < ϵ.

This proves E is dense in lp . That is, ∃ a countable dense set E in lp .
Hence lp , 1 ≤ p < ∞ is separable.

Exercise :- Show that l∞ is not separable.

Proof :- Let S = {x ∈ l∞ : xi = 0 or 1, f or i = 1, 2, ...}.
Then d(x, y) = Supi∈N |xi − yi | = 1 for x ̸= y.
Let {x1 , x2 , ...xn } be any countable set in l∞ and 0 < r < 1. Then B(xn , r)
contains at most one point of set S for n = 1, 2, ...
Since S is uncountable, there is some x ∈ S such that x ∈
/ B(xn , r) ∀ n =
1, 2, ....
In other words, {x1 , x2 , ...} ∩ B(x, r) = ϕ so that {x1 , x2 , ...} cannot be dense
in l∞ .
Thus l∞ is not separable.

Examples of Complete Metric Spaces

11
(1) The set of real numbers R under usual metric d is a complete metric
space.

2 Euclidean space Rn is a complete metric space.(Prove it)

Proof :- We know the Euclidean space Rn is a metric space under the Eu-
clidean metric
∑
n 1
d(x, y) = ( (xi − yi )2 ) 2 where x = (x1 , x2 , ..., xn ) and y = (y1 , y2 , ..., yn )
i=1
To show that Rn is complete, we have to show that every Cauchy sequence
in Rn converges in Rn . For this, let {xm } be a Cauchy sequence in Rn where
xm = (xm m m
1 , x2 , ..., xn )
Since the above sequence is a Cauchy sequence, so by definition of a Cauchy
sequence, for ϵ > 0, ∃ no ∈ N such that
d(xm , xr ) < ϵ, ∀ m, r > no ——(1)
∑n
r 2 12
⇒ ( (xm i − xi ) ) < ϵ ∀ m, r ≥ no
i=1
∑
n
⇒ i − xi ) < ϵ .
(xm r 2 2
i=1
i − xi ) < ϵ f or i = 1, 2, ..., n (because, each term in sum is less than
i.e, (xm r 2 2

ϵ2 )
⇒ |xm
i − xi | < ϵ ∀ i = 1, 2, ..., n.
r

⇒ for each fixed i(1 ≤ i ≤ n), the sequence {xm

i } is a Cauchy sequence of
real numbers and R is complete and so it converges, say xm
i → xi as m → ∞.
So, clearly x ∈ Rn . (B’coz, using the above n we define x = (x1 , x2 , ..., xn ) )
Now we shall show that the Cauchy sequence {xn } in Rn converges to x in
Rn .
∑
n
i − xi ) < ϵ ∀ m, r ≥ no
(xm r 2 2
From (1), we have,
i=1
Letting r → ∞ in above equation, we have lim xri = xi
r→∞
∑
n
Therefore, (xi − xi ) < ϵ
m r 2 2
i=1
⇒ [d(xm , x)]2 < ϵ2
⇒ d(xm , x) < ϵ ∀ m > no
This proves that for each ϵ > 0, ∃ no ∈ N such that d(xm , x) < ϵ ∀ m > no
Hence the Cauchy sequence {xm } in Rn converges in Rn .

(3) The space lp , 1 ≤ p < ∞ is a complete metric space.

12
Proof :- Let {xn } be a Cauchy sequence in R and suppose ϵ > 0 be given.
Since every sequence in R is bounded, therefore {xn } is bounded. Also,
every bounded sequence in real numbers contains a convergent subsequence
(By B. W. theorem). Therefore, ∃ a subsequence xnk of {xn } such that
xnk → x ∈ Rasnk → ∞.
Then by def. of a convergent sequence ∃ a no ∈ N such that
|xnk − x| < ϵ
2
∀ nk > n o
Now |xn − x| ≤ |xn − xnk | + |xnk − x| ——-(1)
Since {xn } is a Cauchy sequence, therefore ∃mo ∈ N such that
|xn − xnk | < ϵ
2
∀ n, nk > mo
Let N = max{no , mo }.
Then from (1), we have

|xn − x| ≤ |xn − xnk | + |xnk − x|

ϵ ϵ
< +
2 2
= ϵ, f or each nk > N and n > N

Thus {xn } is convergent sequence in R. Therefore, every Cauchy sequence in

R is convergent.
Hence R with usual metric is complete metric space.

Def. :- Let (X, d) and (Y, ρ) be any two metric spaces, then a map f : X → Y
is said to be continuous at a point xo ∈ X if for every ϵ > 0, there is δ > 0
such that ρ(f (x), f (xo )) < ϵ, whenever d(x, xo ) < δ.
Or, equivalently, for each BY (f (xo ), ϵ) ∃ BX (xo , δ) such that

f (BX (xo , δ)) ≤ BY (f (xo ), ϵ)

Note :- A function f : X → Y is continuous on X if it continuous at every

point of X.

Example :-
(1) Every function from a discrete metric space (X, d) to any other metric
space (Y, ρ) is always continuous.
Reason :- Let ϵ > 0 be given and xo ∈ X. Choose δ = 12 , then B(xo , δ) = xo .

13
Clearly, f (B(xo , δ)) = {f (xo )} ⊆ B(xo , ϵ).
Since xo ∈ X. So f : (X, d) → (Y, ρ) is continuous at xo ∈.
(2) Identity map between two metric spaces is always continuous.
Proof :- f : (X, d) → (X, d) is continuous :
For this, let ϵ > 0 be given and xo ∈ X. Choose δ = ϵ. Then, f (B(xo , δ)) =
B(xo , δ) ≤ B(xo , δ) = B(f (xo ), ϵ).

Examples of Incomplete Metric Spaces:

(1) The set of rational numbers Q with the metric given by d(x, y) = |x −
y| ∀ x, y ∈ Q is a metric space.
But Q is not complete because .1, .101, .101001, .1010010001, . . . does not
converges to a rational.

2 Let X = C[a, b] be the set of all continuous functions on [a, b].

3 Let P [a, b] denotes the set of all polynomials on [a, b]. Define a metric
d : P [a, b] × P [a, b] → R by d(x, y) = maxt∈[a,b] |x(t) − y(t)|.
Then P [a, b] is a metric space but it is not complete because by Weirstrass
Approximation theorem P [a, b] = C[a, b].

4 Consider (C[a, b], d), where d is the metric defined as d(f, g) =

∫b
|f (x) − g(x)|dx. Then (C[a, b], d) is not complete.
a
Proof :- To show that (C[a, b], d) is not complete. Consider a sequence {fm }
as 

 ; a ≤ t < c − m1
 0
fm (t) = mt − mc + 1 ;c − m1 ≤ t ≤ c


 1 ;c≤t≤b
1
where a < c < b and every m so large that a < c < m
.
Then fm ∈ C[a, b], ∀ m ≥ 1.
Now we want to show that the above sequence is a Cauchy sequence. For,

∫b
d(fm , fn ) = |fm (t) − fn (t)|dt
a

14
 ∫c ∫b
= |fm (t) − fn (t)|dt + |fm (t) − fn (t)|dt
a c
∫c
= |fm (t) − fn (t)|dt [because, on [c, b] both fm (t) and fn (t) = 1]
a
∫c ∫c
≤ |fm (t)|dt + |fn (t)|dt
a a
1 1
∫
c− m
∫c ∫
c− n
∫c
= |fm (t)|dt + |fm (t)|dt + |fn (t)|dt + |fn (t)|dt
a 1
c− m a 1
c− n
∫c ∫c
= |fm (t)|dt + |fn (t)|dt
1 1
c− m c− n
∫c ∫c
= |mt − mc + 1|dt + |nt − nc + 1|dt
1 1
c− m c− n
∫c ∫c ∫c ∫c
≤ m tdt + (1 − cm) dt + n tdt + (1 − cn) dt
1 1 1 1
c− m c− m c− n c− n

−1 1 1 1
= + − + → 0 as m, n → ∞
2m m 2n n
This proves that the above sequence is a Cauchy sequence in C[a, b].
If C[a, b] is complete, therefore ∃ a function f ∈ C[a, b] such that
limm→∞ d(fm , f ) = 0. Now,

∫b
d(fm , f ) = |fm (t) − f (t)|dt
a
1
∫
c− m
∫c ∫b
< |fm (t) − f (t)|dt + |fm (t) − f (t)|dt + |fm (t) − f (t)|dt
a 1
c− m c
1
∫
c− m
∫c ∫b
= |f (t)|dt + |fm (t) − f (t)|dt + |1 − f (t)|dt
a 1
c− m c

15
 ⇒ 0 = limm→∞ d(fm , f )
∫c ∫b
= |f (t)|dt + |1 − f (t)|dt
a c
{
0 ;a≤t≤c
⇒ f(t) =
1 ;c≤t≤b
This implies that the function f is discontinuous at c, which is a contradiction
to the fact that f ∈ C[a, b].
Hence, C[a, b] is not Cauchy.

Theorem :- Banach Contraction Principle

Every contraction mapping on a complete metric space has a unique fixed
point.

or

Let (X, d) be a complete metric space and T : X → X be a contraction. then

∃ a unique point x such that T (x) = x.
Proof :- Let (X, d) be an arbitrary complete metric space. Let x ∈ X.
Define
x1 = T x
x2 = T x1 = [T (T (x)) = T 2 x]
x3 = T x2 = [T (T (x1 )) = T 2 x1 + T 2 T x = T 3 x]
.
.
.
xn = T xn−1 = [T (T xn−2 ) = ... = T n−1 (T (x)) = T n x]
Now we will show that the sequence {xn } = ({T n x}) is a Cauchy sequence.
For, let m, n be any positive integers such that m > n, say m = n + p where
p is positive integer ≥ 1. Then

d(xn , xm ) = d(xn , xn+p )

≤ d(xn , xn+1 ) + d(xn+1 , xn+2 ) + ... + d(xn+p−1 , xn+p ) − − − −(∗)

16
Since T : X → X is a contraction. Therefore,

d(x1 , x2 ) = d(T x, T x1 )
≤ αd(x, x1 )
= αd(x, T x),
d(x2 , x3 ) = d(T x1 , T x2 )
≤ αd(x1 , x2 )
≤ α.αd(x, T x)
= α2 d(x, T x),
d(x3 , x4 ) = d(T x2 , T x3 )
≤ αd(x2 , x3 )
≤ α.α2 d(x, T x)
= α3 d(x, T x),
.
.
.

d(xn , xn+1 ) ≤ αn d(x, T x) for any integer n. ——-(**)

Using (**) in (*), we get

d(xn , xm ) ≤ d(xn , xn+1 ) + d(xn+1 , xn+2 ) + ... + d(xn+p−1 , xn+p )

≤ αn d(x, T x) + αn+1 d(x, T x) + ... + αn+p−1 d(x, T x)
[ ]
= d(x, T x) αn + αn+1 + ... + αn+p+1
αn (1 − αp )
= d(x, T x) → 0 as n → ∞
1−α
i.e, d(xn , xm ) → 0asn → ∞.
Thus, the sequence {xn } is a Cauchy sequence in X. Since X is a Complete
metric space. Therefore, the above sequence {xn } converges in X.
i.e, ∃xo ∈ X such that the sequence xn → xo
i.e, limn→∞ xn = xo
Next, we claim that xo is the fixed point of T . Since the contraction mapping

17
T is continuous and as every continuous function in a metric space preserve
the convergence and xn → xo
⇒ T xn → T x o
But xn+1 = T xn → T xo
Also xn+1 → xo
Therefore, T xo = xo .
This shows that xo is the fixed point of T .
Uniqueness: Suppose x and y are two fixed points of T such that x ̸= y.
Then T (x) = x and T (y) = y. Then

d(x, y) = d(T (x), T (y))

≤ Kd(x, y)
< d(x, y)

which is a contradiction. Hence x = y.

Def. :- Let D be a connected open subset of the plane and let f : D → R

be a function. We say that f satisfies Lipschitz condition in y on D with
Lipschitz constant K if for every (x, y1 ) and (x, y2 ) in D, we have

|f (x, y2 ) − f (x, y1 )| ≤ K(y2 − y1 )

e.g. f (x, y) = xy 2 satisfies the L.C. on the rectangle |x| ≤ 1 and ||y|| ≤ 1.

Def. :- Let (xo , yo ) ∈ D. By local solution passing through (xo , yo ), we mean

a function ϕ defined on an open interval I such that xo ∈ I, ϕ(xo ) = yo ,
′
(x, ϕ(x)) ∈ D ∀ x ∈ I and ϕ (x) = f (x, ϕ(x)) ∀ x ∈ I.

Application of Banach Contraction Principle :- To Differential equa-

tion
Theorem :- Let f be a continuous function defined on an open connected
set D and satisfies Lipschitz condition in y. Then the differential equation
dy
dx
= f (x, y) has a unique solution passing through (xo , yo ).
Proof :- First note that ϕ is the solution of the initial value problem

dy
dx
= f (x, y), y(xo ) = yo

18
 ∫x
in some nbd of xo , iff ϕ is the solution of integral equation y = yo + f (t, y)dt
xo
in that nbd. of xo .
Since f satisfies Lipschitz condition, ∃K > 0 such that

|f (x, y1 ) − f (x, y2 )| ≤ K|y1 − y2 | ∀ (x, y1 ), (x, y2 ) ∈ D.

Let a and b(< 1) be such that the rectangle

R = {(x, y) : |x − xo | ≤ a, |y − yo | ≤ b} is contained in D.
′
Further choose δ > 0 such that R = {(x, y) : |x − xo | ≤ δ, |y − yo | ≤ M δ} is
′
contained in R, where M > 0 such that sup|f (x, y)| = M, (x, y) ∈ R and δ <
1.
Put I = [xo − δ, xo + δ] and J = [yo − M δ, yo + M δ] and consider the set

X = {ψ|ψ : I → J is continuous and y(xo ) = yo }

Define a function d : X × X → R as d(ϕ, ψ) = supx∈I |ϕ(x) − ψ(x)|.

Clearly X with d is a complete metric space. Define a function T : X → X
as

∫x
T ψ(x) = yo + f (t, ψ(t))dt ∀ x ∈ I
xo

Clearly, the above function T is a continuous function and

∫x
|T ψ(x) − yo | = | f (t, ψ(t))dt|
xo
∫x
≤ |f (t, ψ(t))|dt
xo
≤ M |x − xo |
< Mδ

⇒ T ψ(x) ∈ J = [yo − M δ, yo + M δ] ∀ x ∈ I
⇒ Tψ ∈ X
Thus T : X → X is a mapping.

19
Next, we want to show that above map T is a contraction map. Now,
∫x
|T ψ1 (x) − T ψ2 (x)| = |f (t, ψ1 (t)) − f (t, ψ2 (t))|dt
xo
∫x
≤ K |ψ1 (t) − ψ2 (t)|dt
xo
≤ Ksupx∈I |ψ1 (t) − ψ2 (t)||x − xo |
≤ Kδd(ψ1 , ψ2 )
< αd(ψ1 , ψ2 ), where α = Kδ < 1

⇒ d(T ψ1 , T ψ2 ) < αd(ψ1 , ψ2 )

i.e, T is a contraction on X.
Since X is a complete metric space, therefore by Banach Contraction Princi-
ple, T has a unique fixed point in X.
⇒ ∃ ψo ∈ X such that T ψo = ψo
∫x
⇒ T ψo (x) = yo + f (t, ψo (t))dt
xo
∫x
⇒ ψo (x) = yo + f (t, ψo (t))dt
xo
∫x
⇒ ψo is a solution of the integral equation y = yo + f (t, y)dt.
xo
Hence a unique solution of the equation passing through (xo , yo ).

Application of Banach Contraction Principle : for Integral Equation

Theorem :- Let k be a continuous real valued function defined on the square
[a, b] × [a1 , b1 ] and let g ∈ C([a, b]). Then there exists a unique f ∈ C([a, b])
such that f is the solution of the integral equation

∫b
f (x) = λ k(x, y)f (y)dy + g(x) ∀ x ∈ [a, b]
a

for sufficiently large values of λ ∈ R.

Proof :- Since k ∈ C([a, b] × [a1 , b1 ]), there exists M > 0 such that |k(x, y) ≤
M ∀ (x, y) ∈ ([a, b] × [a1 , b1 ]).
Let λ ∈ R be such that |λ| < 1
(b−a)M
.
Now, define T : C([a, b]) → C([a, b]) by

20
 ∫b
(T ψ)(x) = λ k(x, y)f (y)dy + g(x) ∀ x ∈ [a, b] and ψ ∈ C([a, b])
a

Let ψ1 , ψ2 ∈ C([a, b]). Then

∫b
|T ψ1 (x) − T ψ2 (x)| = |λ k(x, y)(ψ1 − ψ2 )(y)dy|
a
∫b
≤ |λ| |k(x, y)||ψ1 (y) − ψ2 (y)|dy
a
∫b
≤ |λ| M maxy∈[a,b] |ψ1 (y) − ψ2 (y)| dy
a
≤ |λ| maxy∈[a,b] |ψ1 (y) − ψ2 (y)| M (b − a)

Now d(T ψ1 , T ψ2 ) = Supx∈[a,b] [|T ψ1 (x) − T ψ2 (x)|]

⇒ d(T ψ1 , T ψ2 ) ≤ |λ|M (b − a)d(ψ1 , ψ2 )
⇒ d(T ψ1 , T ψ2 ) < d(ψ1 , ψ2 )
⇒ T is a contraction in C([a, b]).
Thus by Banach Contraction Principle, there exist a unique f ∈ C([a, b]) such
that

f (x) = T (f (x))
∫b
= λ k(x, y)f (y)dy + g(x)
a

This shows that f is the unique solution of the given integral solution.

Definition :- A subset A of a metric space X is said to be nowhere dense

in X if A contains no sphere or (ball) or equivalently A contains no interior
point. i.e, (A)o = ϕ.

Definition :- A subset E of a metric space (X, d) is said to be categoryI if

E is expressed as countable union of nowhere dense set.

Definition :-A subset A of a metric space (X, d) is said to be category II if

it is not of category I.

21
Example :-
(1) The empty set ϕ is nowhere dense in a metric space, because (ϕ)o = ϕo =
ϕ.
(2) The set of rational numbers with usual metric is of first category

Q = ∪x∈Q {x} [ because (x)o = (x)o = ϕ]

where x is closed in usual metric.

(3) The set of real numbers R with usual metric is of II nd category.

Exercise :- Show that the set of irrational numbers is of category II.

BAIRE’s CATEGORY THEOREM

Statement :- Every complete metric space is of second category.

Proof :- Suppose the complete metric space is of (first) category I. Then X =
[ o ]
∪∞
n=1 An , where each An is closed and contains no sphere or (A) = ϕ . (This
implies that X can be expressed as the union of countable family nowhere
dense sets.)
Let a ∈ X and consider S(a, 1) be the unit sphere with centre a. Then
there is a closed sphere S(a1 , r1 ) with r1 < 12 contained in S(a, 1) such that
S(a1 , r1 ) ∩ A1 = ϕ, where A1 is nowhere dense in X.
1
Again, we find a closed sphere S(a2 , r2 ) inside S(a1 , r1 ) of radius r2 < 22
such
that S(a2 , r2 ) ∩ A2 = ϕ, where A2 is nowhere dense in X.
Proceeding in this way, we get a sequence of closed sphere {S(ak , rk )} such
that

δ{S(ak , rk )} → 0 as k → ∞
and, S(ak , rk ) ∩ Ak = ϕ ∀ k ∈ N.

But X is complete.
Therefore, by Cantor Intersection Theorem, we have
∩∞
k=1 S(ak , rk ) = {x}
⇒ x ∈ S(ak , rk ), ∀ k ∈ N

22
⇒x∈
/ Ak ∀ k ∈ N
/ ∪∞
⇒x∈ k=1 Ak = X
which is a contradiction.
Therefore, our supposition was wrong. Thus metric space (X, d) is of second
category.

Theorem : CANTOR INTERSECTION THEOREM

Let (X, d) be a metric space and suppose {An } be a sequence of non-empty
closed subsets of X such that δ(An ) → 0 as n → ∞. Then X is complete iff
∩∞
n=1 An consists of exactly one point.

Def. :- Compact Metric Space

A metric space (X, d) is said to be compact if every infinite subset of X has
atleast one limit point.

Note :-
(1) Let (X, d) be a metric space and K ⊂ X, then K is compact if it is closed
and bounded
(2) Let (X, d) be a metric space. Then the following are equivalent :
(i) X is compact if every open cover of X has a finite subcover.
(ii) Each countable open cover of X has a finite subcover (countably com-
pact).
(iii) Every infinite set in X has a limit point [B. W. property].
(iv) Every sequence in X has a convergent subsequence (sequentially com-
pact).

Theorem :- Prove that every compact subset of a metric space is closed and
bounded.
Proof :- Let A be a compact subset of a metric space (X, d). We have to
show that A is closed and bounded.
A is bounded :
Suppose on the contrary, A is not bounded. Let x1 ∈ A, then ∃ x2 ∈ A such
that d(x1 , x2 ) > 1.
Further we find x3 ∈ A such that d(x1 , x3 ) > 1 + d(x1 , x2 ).
In this way, we find xn ∈ A ∀ n ∈ N such that d(x1 , xn ) > 1 + d(x1 , xn−1 )

23
i.e, we have sequence {xn } ∈ A such that for n, m ∈ N with m > n, we have
d(x1 , xm ) > 1 + d(x1 , xn )
d(x1 , xm ) − d(x1 , xn ) > 1 ——(*)
By triangular inequality, we have
d(x1 , xm ) ≤ d(x1 , xn ) + d(xn , xm )
⇒ d(x1 , xm ) − d(x1 , xn ) ≤ d(xn , xm ) ——-(**)
From (*) and (**),
d(xn , xm ) > 1 ∀ m, n with m > n
Therefore, {xn } does not contain a convergent subsequence.
⇒ A is not compact, which is a contradiction to the fact that A is compact.
Therefore, our supposition is wrong.
Thus, A is bounded.
Next, we want to show that A is closed. Let x ∈ A, then there is a sequence
{xn } ∈ A such that xn → x i.e, d(xn , x) → 0 as n → ∞.
But A is compact, therefore there is a subsequence {xnk } of {xn } converges
to a point xo ∈ A.
By uniqueness of the limit, we have
x = xo ∈ A
⇒ x ∈ A and so A ⊆ A ——(***)
Also, by definition of a closure of a set, A ⊆ A ——(****)
From (***) and (****)
A=A
⇒ A is closed.

Note :- The converse of the above theorem need not be true.

Example (1) :- Let (X, d) be an infinite discrete metric space. Then X is
bounded, because diam(X) = 1 or δ(X) = 1.

[Because d(x, y) ≤ 1 ∀ x, y ∈ X]

But X is not compact because {{x}|x ∈ X} is an infinite set in X but it does

not have a limit point in X.
Also X is closed, being the whole space.
Example (1) :- Let X = l2 , the set of all bounded {xn } such that

24
∑
n≥1 |xn |2 < ∞.
Define d : X × X → R as

∑
∞ 1
d(x, y) = ( |xi − yi |2 ) 2 ∀ x = {xn } and y = {yn } ∈ X = l2
i=1

Then (X, d) is a metric space.

For every natural number n ∈ N, define en = {0, 0, ..., 0, 1, 0, ..}, where 1 is at
nth place.
Put A = {en |n ∈ N}.
√
Then A ⊆ X and d(en , em ) = 2 ∀ n, m; m ̸= n.
⇒ diam(A) < ∞. ⇒ A is a bounded subset of X.
√
Again, since d(en , em ) = 2 > 1 ∀ n, m; m ̸= n.
⇒ A does not have a limit point in X.
⇒ D(A) = ϕ
⇒ A = A ∪ D(A) = A ∪ ϕ = A ⇒ A is closed in X.
But A is not compact, because A is infinite and derived set of A is empty.
Therefore, from above examples, we conclude that a closed and bounded set
in a metric space need not be compact.

Definition :- A subset Y of a metric space (X, d) is said to be totally bounded

if for every ϵ > 0, ∃ finite set S = {x1 , x2 , ..., xn } in Y such that Y ⊆
∪ni=1 S(xi , ϵ).

OR

Let (X, d) be a metric space and let ϵ > 0 be given. A finite set A of X is
said to be an ϵ−net for X if for every x ∈ X, ∃ a ∈ A such that d(a, x) < ϵ.
In other words, A is an ϵ−net for X iff A is finite and X = ∪{S(a, ϵ) : a ∈ A}.

A metric space (X, d) is said to be totally bounded if X has an ϵ−net

for every ϵ > 0.

Exercise :- Show that every totally bounded set in a metric space is bounded
but not conversely.

25
Proof :- Let A be a totally bounded set in a metric space (X, d). We have
to show that A is bounded.
Since A is totally bounded, therefore, ∃ an ϵ-net, say, Aϵ = {x1 , x2 , ..., xn }
such that A ⊆ ∪ni=1 S(xi , ϵ)
⇒ diam(A) < ∞
i.e, δ(A) < ∞
This implies that A is bounded set.
Thus every totally bounded subset of a metric space is bounded.

Definition :- If (X, d) is a metric space and K is a subset of X such that K

is compact, then K is said to be relatively compact.

Example :- Let X = R with usual metric and let K = (a, b) ⊆ R.

Then K = (a, b) = [a, b] which is compact in R. So (a, b) is relatively compact
in R.

Note :- In Rn , we know that a closed bounded set is compact by Heine Borel

Theorem.
In general metric space, however, there may exist closed and bounded sets
which are not compact.

Example :- Consider a subset E of l2 , consisting of elements e1 , e2 , ..., en ,

where en = {0, 0, ..., 0, 1, 0, 0..}.
Put E = {en |n ∈ N}.
Now ||en || = 1 ∀ n.
Therefore, this set is bounded and it is closed because it has no limit point.
The set E is not compact because E is infinite set.

Note :- Converse of the above exercise need not be true.

We can consider an example :
Example :- Let X = l2 . and A = {en |n ∈ N} where en = {0, 0, ..., 0, 1, 0, 0..};
’1’ at the nth place.
√
⇒ d(en , em ) = 2 ∀ n, m; m ̸= n
√
⇒ diam(A) = 2
√
⇒ δ(A) = 2

26
⇒ A is bounded.
Taking ϵ = 12 , then S(en , ϵ) = {en } and so finitely many S(en , ϵ) cannot cover
A.
⇒ A is not totally bounded.

Theorem :- Prove that a metric space (X, d) is compact iff it is complete

and totally bounded.
Proof :- Suppose that X is compact metric space. Then every infinite set in
X has a limit point in X and hence every Cauchy sequence in X is convergent
and so X is complete. We have to show that X is totally bounded.
Let ϵ > 0 be given and let x1 ∈ X. If X − S(x1 , ϵ) = ϕ, then {x1 } will be
ϵ-net and so there is nothing to prove.
So we assume that X − S(x1 , ϵ) is non empty i.e, X − S(x1 , ϵ) ̸= ϕ.
Let x2 ∈ X − S(x1 , ϵ).
Then, if X − {S(x1 , ϵ) ∪ S(x2 , ϵ)} = ϕ, then {x1 , x2 } is the ϵ-net.
So, let x3 ∈ X −{S(x1 , ϵ)∪S(x2 , ϵ)} and in the same way as above, {x1 , x2 , x3 }
is the ϵ-net and so X is totally bounded.
So it is only necessary to show that this process terminates after finite number
of steps.
If it does not terminates, then x1 , x2 , ... will be an infinite sequence and each
pair of which have a distance atleast ϵ and so no subsequence can be Cauchy
subsequence. (Because, for Cauchy sequence d(xi , xj ) < ϵ, not equal to ϵ)
and this contradicts our hypothesis. Hence X is totally bounded.
Conversely, suppose X is complete and totally bounded. We have to show
that X is compact.
For this, let {xn } be a sequence in X. Since X is totally bounded, therefore
∀ ϵ > 0, ∃ an ϵ-net in X. Let ϵ = 1. Then X is contained in the finite union
of spheres of radius 1 (By def. of ϵ-net with ϵ = 1) and clearly atleast one of
the spheres contains infinitely many entries of {xn }.
Let S1 = {x11 , x12 , ..., x1n , ...} ⊂ {xn } such that S1 ⊆ S(a1 , 1) for some a1 ∈ X.
Next, taking ϵ = 21 , then S1 is contained in the union of finitely many spheres
1
with radius 2
and atleast one of these spheres contains infinitely many entries
from S1 .

27
Let S2 = {x21 , x22 , ..., x2n , ...} ⊆ S1 such that S2 ⊆ S(a2 , 12 ) for some a2 ∈ X.
Proceeding in this way, we obtain a subsequence Sk = {xk1 , xk2 , ..., xkn , ...} of
Sk−1 such that Sk ⊆ S(ak , k1 ).
Now consider the set S = {x11 , x22 , x33 , ...} = {xkk }.
Then S is a subsequence of {xn }.
Claim : The set S = {xkk } is convergent.
Let m, n ∈ N, m, n ≥ k. Then xm
m , xn ∈ S(ak , k ).
n 1

⇒ d(xm n
m , xn ) <
1
k
∀ m, n ≥ k
⇒ the above sequence S = {xkk } is a Cauchy sequence in X, but X is compact.
Therefore S = {xkk } is convergent in X and hence the claim.
Hence every sequence in X has a convergent sequence showing that X is
compact.

Exercise :- A subset A of a metric space (X, d) is totally bounded iff A is

totally bounded.
Proof :- Suppose that A is totally bounded. We have to show that A is
totally bounded.
Let {xn } be a sequence in A and let ϵ > 0 be given. Then xn ∈ A.
⇒ S(xn , 3ϵ ) ∩ A ̸= ϕ
⇒ ∃ an ∈ A such that d(an , xn ) < ϵ
3
But A is totally bounded. Therefore, the above sequence contains a Cauchy
subsequence {ank }.
Therefore, ∀ ϵ > 0, ∃ m ∈ N such that

d(xnj , xnk ) ≤ d(xnj , anj ) + d(anj , ank ) + d(ank , xnk )

ϵ ϵ ϵ
< + +
3 3 3
= ϵ

⇒ {xnk } is a Cauchy subsequence of {xn }.

Conversely, suppose A is totally bounded.
Since we know that subset of totally bounded set is again totally bounded,
therefore A is totally bounded.

Definition :- A collection F of functions on a set X is said to be uniformly

bounded if there is M > 0 such that |f (x)| ≤ M ∀ x ∈ X and f ∈ F .

28
Note :- For subset of C(X) uniformly boundedness agrees with boundedness
in submetric. [i.e, a set is uniformly bounded iff it is contained in a sphere]

Definition :- A collection F of functions defined on a metric space X is said

to be equicontinuous if for each ϵ > 0, ∃ δ > 0 such that

d(x, y) < δ ⇒ |f (x) − f (y)| < ϵ ∀ x, y ∈ X and ∀ f ∈ F .

Note :- Each member of equicontinuous class is uniformly continuous.

Definition :- Let X be a metric space and K be a subset of X. Then K is

said to be relatively compact if its closure (K) is compact.
e.g. Open interval (a, b) is relatively compact in R.

Theorem : Arzela-Ascoli Theorem

Let (X, d) be compact metric space. Then a subset K of C(X) is relatively
compact iff K is uniformly bounded and equicontinuous.
Proof :- Suppose K is relatively compact in C(X). Then K is compact in
C(X).
Since a subset of a metric space is compact iff it is complete and totally
bounded. Therefore, K is totally bounded. But a totally bounded set is
always bounded, so K is bounded in C(X) and hence uniformly bounded
there.
But K ⊆ K, so K is uniformly bounded.
Next, we show that K is equicontinuous.
Since K is totally bounded, therefore ∀ ϵ > 0, ∃ a finite set {f1 , f2 , ..., fn } in
C(X) such that K ⊆ ∪nk=1 S(fk , 3ϵ ) ——(i)
Since each fk is uniformly continuous, k = 1, 2, ..., n (Because, fk ∈ C(X)
and X is compact ⇒ fk is uniformly continuous),
Therefore, ∃ δk > 0 (k = 1, 2, ..., n) such that d(x, y) < δk
⇒ |fk (x) − fk (y)| < ϵ
3
——-(ii)
Put δ = min{δ1 , δ2 , ..., δn }. Then δ > 0.
Let f ∈ K. Then ∃ some fk ∈ {f1 , f2 , ..., fn } such that
ϵ
d(f, fk ) < 3
(by (i))

29
⇒ Sup|f (x) − fk (x)| < ϵ
3
⇒ |f (x) − fk (x)| < ϵ
3
——-(iii)
Now

|f (x) − f (y)| ≤ |f (x) − fk (x)| + |fk (x) − fk (y)| + |fk (y) − f (y)|
ϵ ϵ ϵ
< + + (by(ii), (iii))
3 3 3
= ϵ

whenever d(x, y) < δ. i.e, |f (x) − f (y)| < ϵ.

Hence K is equicontinuous.
Conversely, suppose K is uniformly bounded and equicontinuous. We have
to show that K is relatively compact in C(X). That is, we shall show that
K is compact subset of C(X).
To show that K is compact, it is enough to show that K is complete and
totally bounded.
For this, let ϵ > 0 be given. Since K is equicontinuous, ∃ δ > 0 such that
d(x, y) < δ
⇒ |f (x) − f (y)| < ϵ ∀ x, y ∈ X and f ∈ K ——-(*)
Since X is compact, it is totally bounded and ∃ δ-net say {x1 , x2 , ..., xn } in
X such that X ⊆ ∪nk=1 S(xk , δ).
Also K is uniformly bounded set. The set A = {f (xk )|1 ≤ k ≤ n, f ∈ K} is
bounded.
Since a bounded set in R is totally bounded. Therefore ∃ {y1 , y2 , ..., yn } in R
such that A ⊆ ∪nk=1 S(yk , 4ϵ ).
For any map ϕ : {1, 2, ..., n} → {1, 2, ..., m},
consider Eϕ = {f ∈ K|f (xk ) ∈ S(yϕ(k) , 4ϵ ), k = 1, 2, ..., n}.
Clearly, there are only finitely many set Eϕ and every f ∈ K belongs to one
of the sets Eϕ .
Next, we want to show that diam. of Eϕ is finite.
For this, let f, g ∈ Eϕ and x ∈ X.
Since x ∈ X
⇒ x ∈ S(xk , δ) for some k
⇒ d(x, xk ) < δ for some k

30
⇒ |f (x) − f (xk )| < ϵ
4
for some k

N ow, |f (x) − g(x)| ≤ |f (x) − f (xk )| + |f (xk ) − yϕ(k) | + |yϕ(k) − g(xk )| + |g(xk ) − g(x)|
ϵ ϵ ϵ ϵ
< + + +
4 4 4 4
= ϵ

⇒ |f (x) − g(x)| < ϵ ∀ x ∈ X

⇒ diam(Eϕ ) ≤ ϵ
⇒ K can be covered by finitely many sets of diameter less than ϵ
⇒ K is totally bounded.
Thus, we have proved that K is totally bounded and complete ans so K is
compact.
Thus K is relatively compact.

Exercise :- Prove that every compact subset of a metric space is closed and
bounded. Converse need not be true.
Exercise :- A subset A of a metric space (X, d) is totally bounded iff A is
totally bounded.

31
 UNIT II

Def. Let X be a vector space over the field K(R or C). A function ||.|| : X −→ R
is called a norm if it satisfy the following properties:
(i) ||x|| ≥ 0, ∀x ∈ X
(ii) ||x|| = 0 ⇔ x = 0
(iii) ||x + y|| ≤ ||x|| + ||y||, ∀x, y ∈ X
(iv) ||αx|| = |α|||x||, ∀α ∈ K and x ∈ X.

A vector space X together with a norm on it is called a normed linear space and
we denote it by (X, || > ||).
Exercise Is every normed linear space is a metric space?
Sol. Yes, a normed linear space becomes a metric space with the metric defined by

d(x, y) = ||x − y||, ∀x, y ∈ X

(i) d(x, y) = ||x − y|| ≥ 0

(ii) d(x, y) = 0 ⇔ ||x − y|| = 0 ⇔ x − y = o ⇔ x = y
(iii)d(x, y) = kx − yk = k(−1)(y − x)k
= 1.ky − xk
= ky − xk
= d(y, x).
(iv) d(x, y) = kx − yk
= kx − z + z − yk
= kx − zk + kz − yk
= d(x, z) + d(z, y).
This metric is called the metric induced by norm.
But the converse is not true.
Let X = N
and d(m, n) = |m − n|
Then X is a metric space but not normed linear space. Infact, we can define metric

1
on any set but for for norm we need a vector space.
Def. Let (X, k.k) be a normed linear space.
(i) The set {x ∈ X : kx − x0 k < r} denoted by S(x0 , r) or B(x0 , r) is called the open
sphere (or open ball) of radius r with center x0 .
(ii) The set {x ∈ X : kx−x0 k ≤ r} denoted by S(x0 , r) or B(x0 , r) is called the closed
sphere (or closed ball) of radius r with center x0 .
(iii) A sequence {xn } in X is said to converge to an element x0 ∈ X if ∀ > 0 there
exist a positive integer n0 such that kxn −x0 k < ∀n ≥ n0 or kxn −x0 k → 0 as n → ∞.
If xn → x0 then we write limn→∞ xn = x0 .
(iv) A sequence {xn } ⊂ X is said to be Cauchy Sequence in X if for every  > 0,
∃ a positive integer n0 such that kxn − xm k < ∀n, m ≥ n0 i.e. kxn − xm k → 0 as
n, m → ∞.
Note:- Every convergent sequence is a Cauchy sequence but converse need not be
true.
Proof:- Let xn → x then ∃n0 such that kxn − xk < G2
Let m, n ≥ n0 , then
kxn − xm k ≤ kxn − xk + kx − xm k
< 2 + 2
= .
Def. A normed linear space which is complete(A complete normed linear space) as a
metric space is called Banach Space.
Exercise Let X be a vector space over K(RorC). Then the mapping
(i) + : X × X −→ X defined by
+(x, y) = x + y
(ii)(·) : K × X −→ X defined by
·(α, x) = αx
are continuous.
That is, vector space operation are jointly continuous.
Sol.:- Let {xn } and {yn } be sequences in x and let {αn } be a sequence in K such that

2
xn −→ x, yn −→ y and αn −→ α as n −→ ∞.
Then, k + (xn , yn ) − +(x, y)k = k(xn + yn ) − (x + y)k
= k(xn − x) + (yn − y)k
≤ kxn − xk + kyn − yk −→ 0 as n −→ ∞ (because xn −→ x and
yn −→ y).
Thus +(xn , yn ) −→ +(x, y).
Therefore, addition is continuous function.
(ii) k · (αn , xn ) − ·(α, x)k = kαn xn − αxk
= kαn xn − αn x + αn x − αxk
≤ |αn |kxn − xk + |αn − α|kxk
−→ 0 as n −→ ∞
(because xn −→ x as n −→ ∞ in X and αn −→ α in K).
Thus ·(αn , xn ) −→ ·(α, x).
Thus addition and scalar multiplication are continuous function.
Lemma:- In a normed linear space,
|kxk − kyk| ≤ kx − yk, ∀x, y ∈ X.
Proof:- Since kxk = kx − y + yk ≤ kx − yk + kyk
⇒ kxk − kyk ≤ kx − yk———(i)
kyk = ky − x + xk ≤ ky − xk + kxk
⇒ kyk − kxk ≤ kx − yk————–(ii)
from (i) and (ii), we get
|kxk − kyk| ≤ kx − yk.
Exercise:- Show that norm k · k is a continuous function from X to R.
Sol.:- Let {xn } ⊂ X such that xn → x in X as n → ∞. Then
|kxn k − kxk| ≤ kxn − xk −→ 0 as n −→ ∞.
(because xn −→ x in X so kxn − xk −→ 0 as n −→ ∞).
⇒ kxn k −→ kxk as n −→ ∞.
Therefore, k · k is a continuous function.
Examples of Banach Space:-

3
(1) R or C is a Banach space with norm defined by kxk = |x|, ∀x ∈ R. (Note: Proof
same as done in Unit I. Replace d(xn , yn ) by kxn , yn k).
n 1
(2) The linear space Rn equipped with norm kxk = ( |xi |2 ) 2 , where x =
P
i=1
(x1 , x2 , ..., xn ) ∈ Rn is a real Banach space.
∞ 1
(3) Lp , 1 ≤ p ≤ ∞, with the norm defined by kxk = ( |xi |p ) p , where x = {xi }∞ p
P
i=1 ∈ L
i=1
is a Banach space.
(4) l∞ is also a Banach space with norm kxk∞ = sup |xi |, x = {xi } ∈ l∞ is a Banach
1≤i<∞
space.
NOTATION:- The set of all the convergent sequences in the field K is denoted by
Let C = {x ∈ l∞ , such that {xi } converges in K as i −→ ∞}, x = {xi } ∈ l∞ .
C0 = {x ∈ C, such that xi −→ 0} as i −→ ∞.
C00 = {x ∈ C : {xi } = 0 for all but finitely many values i}.
Therefore, C ⊂ C0 ⊂ C00 ⊂ l∞ .
Completeness of C0 :
Let x ∈ C̄0 , the closure of C0 in l∞ . Then there exist a sequence {xn } ∈ C0 such that
kxn − xk −→ 0 as n −→ ∞. So for each i,
|x(i)| ≤ |x(i) − xN (i)| + |xN (i)|
≤ kx − xN k∞ + |xN (i)| (because kx − xN k∞ = sup |x(i) − xN (i)|)
i∈N

< 2
+ |xN (i), for sufficiently large N.
Since xN (i) −→ 0 as i −→ ∞,
We have,
|xN (i)| < 2 ∀i ≥ i0 (i.e. |xN (i) − 0| < G2 )
Thus, |x(i)| < , for sufficiently large i, i.e. i ≥ i0
Hence, x ∈ C0 .
Hence, C0 is a closed subspace of l∞ .
(Closed subspace of a complete metric space is complete and l∞ is complete so C0 is
complete.)
C00 is not closed subspace of l∞ :-
For, take

4
xn = {1, 12 , 13 , ..., n1 , 0, 0, ...}
Then xn ∈ C00 , ∀n
But xn −→ x = {1, 21 , 13 , ..., n1 , n+1
1
, ...}
which is not complete in C00 .
Thus C00 is not a closed subspace of l∞ and therefore is not a Banach space.
Definition:- A subset Y(as a subset of a vector space) of a normed linear space X,
is a subspace of X considered as a vector space with the norm obtained by restricting
the norm on X to the subset of Y. This norm on Y is said to be induced by the norm
on X.
Lemma:- Let {x1 , x2 , ..., xn } be a linearly independent set of vectors in a normed
linear space. Then, there is a real number c > 0 such that for any choice of scalars
α1 , α2 , ..., αn , we have
n
P
kα1 x1 + α2 x2 + ... + αn xn k = k α k xk k
k=1
≥ c(|α1 | + |α2 | + ..., |αn |)
Pn
=c |αk |.
k=1
Theorem:- Every finite dimensional subspace Y of a normed linear space X is com-
plete. In particular, every finite dimensional normed linear space is complete.
Proof:- Let {yn } be a Cauchy sequence in Y. We show that it converges in Y.
Let y be the limit of the sequence {yn } ∈ Y . Further, let dim(Y ) = n and
{e1 , e2 , ..., en } be any basis for Y. Then each ym can be uniquely expressed as
n
(m) (m) (m) P (m)
ym = α1 e1 + α2 e2 + ... + αn en = αi ei .
i=1
Since, {ym } is a Cauchy sequence for every  > 0, there is a positive integer N such
that kym − yr k < , ∀m, r ≥ N .
m n
P (m) P (r)
⇒ k αi ei − αi ei k < 
i=1 i=1
n
P (m) (r)
⇒ k (αi − αi )ei k < .
i=1
Then, by using above lemma, we have for some c > 0,
n
P (m) (r)
 > kym − yr k = k (αi − αi )ei k
i=1
n
P (m) (r)
≥c |αi − αi |, where m, r ≥ N
i=1

5
 n
(m) (r)
− αi | < c ∀m, r > N
P
i.e. |αi
i=1
dividing c > 0, we get
n
(m) (r) P (m) (r)
|αi − αi | ≤ |αi − αi |
i=1
≤ c , ∀m, r ≥ N .
(m)
⇒ {αi } is a Cauchy sequence in R(or C), for each fixed i, 1 ≤ i < n.
Since R and C are complete, so it(i.e. every Cauchy sequence) converges in R or C,
(m)
i.e.αi −→ αi as m −→ ∞ for i = 1, 2, ..., n.
n
P
Let y = α1 e1 + α2 e2 + ... + αn en = αi e i
i=1
Then y ∈ Y .
n
P (m)
Further, kym − yk = k (αi − αi )ei k
i=1
n
P (m)
≤k |αi − αi |kei k
i=1
−→ 0 as m −→ ∞.
(m)
(On the right hand side αi −→ αi as m −→ ∞. Then kym − yk −→ 0 as m −→ ∞),
i.e. ym −→ y in Y.
Hence, the Cauchy sequence {ym } is convergent in Y and so Y is complete.
Remarks:-Every finite dimensional space is complete. Every finite dimensional
normed linear space is a Banach space.
Equivalent Norm:- A norm k · k on a vector space X is said to be equivalent to a
norm k · k0 if there are positive constant a and b such that akxk0 ≤ kxk ≤ bkxk0 .
Theorem Any two norms on a finite dimensional normed linear space are equivalent.
Proof:- Let X be a finite dimensional normed linear space such that dim X = n.
Suppose {e1 , e2 , ..., en } be any basis of X. Then every x ∈ X has a unique representa-
tion of the form
n
P
x = α1 e1 + α2 e2 + ... + αn en = αi ei for some scalar αi , 1 ≤ i ≤ n.
i=1
Now to show that any two norms on X are equivalent. Let k · k1 and k · k2 be two
norm on X. We want to show that k · k1 and k · k2 are equivalent.
Pn
Now, kxk1 = k αi ei k
i=1
n
P
≥ ck αi ei k for some c > 0(by lemma)
i=1

6
 n
P kxk1
i.e. |αi | ≤ c
—————-(i)
i=1
n
P
Next, kxk2 = k αi ei k2
i=1
n
P
≤ kαi ei k2
i=1
Pn
≤k kαi k, where k = max kei k2
i=1 1≤i≤n
k
≤ c kxk1 ————–(by (i))
i.e. kxk2 ≤ bkxk1 for some b = kc ——(ii)
Interchanging the role of k · k1 and k · k2 , we have
kxk1 ≤ bkxk2
⇒ 1b kxk1 ≤ kxk2 ——-(iii)
1 c
⇒ akxk1 ≤ kxk2 , where a = b
= k
is constant.
Thus, from (ii) and (iii), we conclude that k · k1 and k · k2 are equivalent.
Theorem(V.Imp.):-F-Reisz’s Lemma Let M be a closed proper subspace of a
normed linear space X and θ ∈ R such that 0 < θ < 1, then ∃ a vector xθ ∈ X such
that kxθ k = 1 and kx − xθ k ≥ θ, ∀x ∈ M .
X
Proof:- Since M is a closed proper subspace of X. Consider any x1 ∈ M
and denote its
distance from M by h.
i.e. h = inf{kx − x1 k : x ∈ M }
= d(x1 , M ).
Clearly, h > 0(Otherwise we will have h = 0 ⇒ d(x1 , M ) = 0 ⇒ x1 ∈ M )
h 1 h
Also, 0 < θ < 1 and θ
> h (because θ < 1 ⇒ θ
>1⇒ θ
> h)
Then ∃x0 ∈ M such that
h ≤ kx1 − x0 k ≤ hθ ———–(1)(by the definition of infimium)
Let xθ = c(x1 − x0 ), where c = kx1 − x0 k−1 > 0(because x1 ∈ X
M
, x0 ∈ M so x 6= x0 )
Thus, kxθ k = ckx1 − x0 k
= cc−1 = 1.
Next, we want to show that kx − x0 k ≥ θ, ∀x ∈ M .
Now, let x ∈ M be arbitrary.
Then, x0 + c−1 x ∈ M and

7
kx − xθ k = kx − c(x1 − x0 )k
= ck(c−1 x + x0 ) − x1 k
≥ ch (because h = inf{kx − x1 k : x ∈ M })————(2)
But ch = kx1 − x0 k−1 h ≥ θ———-(3)(by (1))
From (2) and (3), we have
kx − xθ k ≥ θ, ∀x ∈ M .
Hence, proved.
Quotient space:-Let X be a vector space over K(RorC). Let M be a closed subspace
of a normed linear space X. The quotient space
X
M
= {(x + M ) : x ∈ X}.
It is a vector space(linear space) with addition and scalar multiplication defined as
(x + M ) + (y + M ) = (x + y) + M, ∀x, y ∈ X.
α(x + M ) = αx + M, ∀x ∈ M .
Theorem:- Let M be a closed subspace of a normed linear space X. For each
X
x+M ∈ M
, define
kx + M k = inf{kx + mk : m ∈ M }————(*)
X
Then M
is a normed linear space under this norm. If X is a Banach space then so is
X
M
.
X
Proof:- We first show that (*) defines a norm on M
. It is obvious that
(1) kx + M k ≥ 0
Since kx + M k = inf{kx + mk : m ∈ M } is a set of non-negative real number which
is bounded below and so inf. exist and is greater than and equal to zero.
(2) kx + M k = 0 iff x + M = M
⇔ inf{kx + mk : m ∈ M } = 0
⇔ ∃{mk } in M such that kx + mk k −→ 0 as k −→ ∞
⇔ mk + x −→ 0 as k −→ ∞
⇔ mk −→ −x as k −→ ∞
⇔ −x ∈ M = M
⇔ −x ∈ M (M is closed subspace)

8
⇔x∈M
Therefore, kx + M k = 0 ⇔ x + M = M.
(3) kα(x + M )k = |α|kx + M k
Now, kα(x + M )k = inf{kα(x + m)k : m ∈ M }
= inf{|α|k(x + m)k : m ∈ M }
= |α| inf{k(x + m)k : m ∈ M }
= |α|kx + M k.
Therefore, kα(x + M )k = |α|kx + M k.
(4) k(x + M ) + (y + M )k = kx + M k + ky + M k
Now, k(x + M ) + (y + M )k = k(x + y) + M k
= inf{k(x + y) + mk : m ∈ M }
= inf{k(x + y) + m1 + m2 k : m1 , m2 ∈ M }
= inf{k(x + m1 ) + (y + m2 )k : m1 , m2 ∈ M }
≤ inf{kx + m1 k : m1 ∈ M } + inf{kx + m2 k : m2 ∈ M }
≤ kx + M k + ky + M k
X
This proves that ( M , k · k) is a normed linear space.
X
Now, suppose that X is Banach space. We want to show that M
is a Banach space i.e.
a complete normed linear space. We know that a Cauchy sequence in C is convergent
iff it has a convergent sub sequence.
Clearly, it is possible to find a sub sequence {xnr + M } of the sequence {xn + M } such
that
1
k(xn2 + M ) − (xn1 + M )k < 2
1
k(xn3 + M ) − (xn2 + M )k < 22

.
.
.
1
k(xnk+1 + M ) − (xnk + M )k < 2k

Choose any vector y1 ∈ xn1 + M

Further select y2 ∈ xn2 + M such that ky2 − y1 k < 21 .

9
 1
Then, we find y3 ∈ xn3 + M such that ky3 − y2 k < 22

Continuing in this way, we get a sequence in X such that xnk + M = yk + M and

1
kyk+1 − yk k < 2k
,k = 1, 2, ....
Let k > r. Then
kyk − yr k = k(yk − yk−1 ) + (yk−1 − yk−2 ) + ... + (yr+1 − yr )k
≤ kyk − yk−1 k + kyk−1 − yk−2 k + ... + kyr+1 − yr k
1 1
< 2k−1
+ 2k−2 + ... + 21r
1 1
< 2r
+ 2r+1 + ... + 21k + 2k−1
1
+ 1
2k−2
+ ...
1
2r 2 1
= [ 1− 1 = 2r = 2r−1 ]
2
1
< 2r−1
−→ 0 as k, r −→ ∞.
It follows that {yk } is a Cauchy sequence in X. But X is complete, ∃y ∈ X such that
lim kyk − yk = 0.
k→∞
Now k(xnk + M ) − (y + M )k = k(xnk − y) + M k
= inf{k(xnk − y) + mk : m ∈ M }
≤ kxnk + m − yk (because inf. of a set is less than each
members in the set.)
≤ kyk − yk, where yk = xnk + mk → 0 as k → ∞ for some
mk ∈ M .
X
⇒ xnk + M −→ y + M ∈ M
.
X
Thus, we have proved that the sub sequence {xnk + M } is a convergent in M
.
X X
⇒ the sequence {xn + M } converges in M
. Hence M
is complete and so is a Banach
space.
Exercise:- Prove that every complete normed linear space is closed.
Application of Reiz Lemma:-
Let X be a normed linear space and suppose M = {x ∈ X : kxk ≤ 1} is compact.
Then X is a finite dimensional space.
Proof:-Suppose M is compact such that dim X is infinite, i.e. X is infinite dimensional.
Choose x1 ∈ X of norm 1 and let X1 be the subspace spanned by x1 . Then x1 is proper
subspace of X. Since X1 is finite dimensional, it is closed. (because we know that every

10
finite dimensional normed linear space is closed)
Since X1 is a proper closed subspace of of X.
Therefore by Reiz lemma, there exist a vector x2 such that kx2 k = 1 and kx2 −x1 k ≥ 12 .
Let X2 be closed proper subspace spanned by x1 , x2 , then again by Reiz lemma, there
is x3 ∈ X such that kx3 k = 1 and kx3 − x2 k ≥ 21 , ∀x ∈ X2 .
Proceeding by induction, we obtain an infinite sequence {xn } of vector of norm 1 and
1
kxm − xn k ≥ 2
for n 6= m.
1
i.e. {xn } is an infinite sequence in M such that kxn − xm k ≥ 2
for m 6= n.
⇒ {xn } is infinite set in M which has no accumulation point in M, which contradicts
the compactness of M.
Therefore, our supposition is wrong.
Hence, we can conclude that X is finite dimensional space.
Exercise:- Let X and Y be normed linear space and T : X −→ Y . Then show that if
T is continuous at one point of X then T is continuous at every point of T.
Def.:- Let X and Y be normed linear spaces. A mapping T : X −→ Y is said to be
linear transformation if
T (αx + βy) = αT (x) + βT (y).
Def.:- The null space of the T is the set of all x ∈ D(T ) ⊆ X such that T (x) = 0 or
ker T = {x : T (x) = 0}.
Def.:-Bounded Linear Operator
Let X and Y be normed linear spaces and T : X → Y be a linear operator. The
operator T is said to be bounded if there is a real number M > 0 such that
kT xk ≤ M kxk, ∀x ∈ X.———–(1)
Remarks:- (1) If x = 0, then T (x) = 0.
(2) Let x 6= 0 ∈ X, then from (1)
kT (x)k
kxk
≤M
kT (x)k
⇒ sup kxk
≤M
06=x∈X
kT (x)k
Let kT k = sup kxk
, then kT k is called the norm of T.
06=x∈X
If X = {0}, then kT k = 0 because T = 0 as T (0) = 0.

11
(3) Now from (1) with kT k = M , we have kT xk ≤ kT kkxk.
Def.:- Let X and Y be normed linear space. Suppose T : X → Y is any operator, (not
necessarily linear). Then T is said to continuous at x0 ∈ X, if for every  > 0 there is
a δ > 0 such that kT x − T x0 k <  for all x ∈ X satisfying kx − x0 k ≤ δ.
Relation between Continuity and Boundedness:-
Let T : X −→ Y be linear operator where X and Y are normed linear spaces. Then T
is continuous iff T is bounded.
Proof:- Suppose T is continuous. Then T is continuous at 0.
For  > 0, ∃δ > 0 such that
kx − 0k < δ ⇒ kT (x) − T (0)k < 
i.e. kxk < δ ⇒ kT (x)k < ———–(1)
δy
Let 0 6= y ∈ X. Define x = 2kyk
. Then
δy δ
kxk = k 2kyk k = 2kyk
kyk
from (1) kT xk < 
δy
or kT ( 2kyk )k < 
δ
2kyk
kT (y)k <
2
or kT (y)k < δ
kyk
Also, for y = 0,
2 2c
ky(0)k < δ
k0k = δ
.0
2c
Therefore, kT yk ≤ M kyk, where M = δ
∀y ∈ X.
Therefore, T is bounded.
Conversely, suppose T is bounded. Then ∃M > 0 such that kT xk ≤ M kxk, ∀x ∈ X.
We show that T is continuous.

Let x0 ∈ X. Let  > 0 be given. Choose δ = 2M
.
Then, kx − x0 k < δ
⇒ kT (x) − T (x0 )k = kT (x − x0 )k ≤ M kx − x0 k ≤ M δ = 2 ≤ .
Thus, for given  > 0, ∃δ > 0 such that kx − x0 k < δ ⇒ kT x − T x0 k < .
Thus, T is continuous at x0 . Since x0 is arbitrary point. So, T is continuous every
where. That is, T is continuous.

12
Examples of Operators:-
(1) The identity operator I : X −→ X is defined by I(x) = x, ∀x ∈ X. We denote the
identity operator simply by I.
Here, kIk = 1.
Obviously, I is bounded operator.
kIxk
(Let 0 6= x ∈ X), kIxk = kxk ⇒ kxk
=1
kIxk
sup kxk
= 1 ⇒ kIk = 1.)
06=x∈X
(2) The zero operator O : X −→ Y is defined by O(x) = 0, ∀x ∈ X.
Here, k0k = 0.
Obviously, O is a bounded operator.
(3) Let X be the vector space of all polynomial in [0, 1].
Define T : X −→ X as
T (x)(t) = x0 (t), ∀x ∈ X, where x0 denotes the differentiation of x w.r.t T. Then T is a
linear operator but not bounded.
(4) Define T : C[0, 1] −→ C[0, 1] by
R
T (x) = y, where y = k(t, z)x(z)dz
Here k is given by
k : [0, 1] × [0, 1] −→ R is continuous and is called the kernel of T.
Then T is linear and bounded.
Convergence of Sequence of Bounded Linear Operator
A sequence {Tn } in B(X, Y ) from a normed linear space X into another linear space is
said to be converge uniformly in the norm of B(X, Y ) if there exist some T ∈ B(X, Y )
such that kTn − T k −→ 0 as n −→ ∞
i.e. Given  > 0, ∃ an integer N > 0 such that kTn − T k < , ∀n > N .
sup kTn (x) − T (x)k ≤ , ∀n > N .
kxk=1
The sequence {Tn } in B(X, Y ) is said to converge strongly if there exist T ∈ B(X, Y )
such that kTn (x) − T (x)k −→ 0 as n −→ ∞.
Remark:- Since kTn (x) − T (x)k ≤ kTn − T kkxk, it follows that uniform convergence
implies strong convergence.

13
Theorem:- Let X,Y and Z be normed linear spaces over the field K. Then
(1) B(X, Y ) is a normed linear space.
(B(X, Y ) is the set of all bounded linear operator.)
(2) B(X, Y ) is Banach space if Y is complete space.
(If Y is complete, then B(X,Y) is Banach Space.)
Proof:-Let S, T ∈ B(X, Y ).
Then (S + T )(x)k = kSx + T xk
≤ kSxk + kT xk
≤ kSkkxk + kT kkxk
= (kSk + kT k)kxk
= M kxk, ∀x, where M = kδk + kT k.
Therefore, S + T ∈ B(X, Y ).
Next, (αS)xk = kα(Sx)k
= |α|kSxk
≤ |α|kSkkxk
≤ M kxk, where M = |α|kSk
Therefore, αS ∈ B(X, Y ).
Thus we can easily show that B(X, Y ) is a vector space.
We have to show that B(X, Y ) is a normed linear space under the norm
kT xk
kT k = sup kxk
06=x∈X
(i) Clearly, kT k ≥ 0, ∀T ∈ B(X, Y ).
kT xk
(ii) Now, kT k = 0 ⇔ sup kxk
=0
06=x∈X
kT xk
⇔ kxk
= 0, ∀x ∈ X
⇔ kT xk = 0, ∀x ∈ X
⇔ T = O.
k(αT )(x)k
(iii) kαT k = sup kxk
06=x∈X
kα(T )(x)k
= sup kxk
06=x∈X
≤ sup |α| kT xk
kxk
06=x∈X

14
 kT xk
= |α| sup kxk
06=x∈X
= |α|kT k
(iv) kT + Sk = sup k(T + S)(x)k
06=x∈X
kT x+Sxk
= sup kxk
06=x∈X
kT xk+kSxk
≤ sup kxk
06=x∈X
kT xk kSxk
≤ sup kxk
+ sup kxk
06=x∈X 06=x∈X
= kT k + kSk.
Therefore, B(X, Y ) is a normed linear space.
To show that B(X, Y ) is complete.
Suppose {Tn } is a Cauchy sequence in B(X, Y ), i.e. kTn − Tm k −→ 0 as n, m −→ ∞.
Therefore, for given  > 0∃n0 such that kTn − Tm k < ∀n, m ≥ n0
or, sup kTn x−T
kxk
m xk
< , ∀n, m ≥ n0
x6=0
kTn x−Tm xk
or, kxk
< , ∀n, m ≥ n0 , ∀x ∈ X
or, kTn x − Tm xk ≤ kxk, ∀n, m ≥ n0 , ∀x ∈ X.————(i)
For fixed x, {Tn x} is a Cauchy sequence in Y. But Y is complete so there exist y ∈ Y
such that Tn x −→ y.
Define T (x) = y
i.e. T (x) = lim Tn x.
n→∞
Clearly, T is linear because for x, y ∈ X and α, β ∈ K, we have
T (αx + βy) = lim Tn (αx + βy
n−→∞
= lim [αTn (x) + βTn (y)]
n−→∞
= α lim Tn (x) + β lim Tn (y)]
n−→∞ n−→∞
= αT x + βT y.
Now,|kTn k − kTm k| ≤ kTn − Tm k −→ 0, as n −→ ∞.
i.e. {kTn k} is a Cauchy sequence, therefore it is bounded. Hence ∃k > 0 such that
kTn k ≤ k, ∀n = 1, 2, ...
kTn xk
i.e. sup kxk
≤k
06=x∈X
kTn xk
or, kxk
≤ k, ∀0 6= x ∈ X

15
or, kTn (x)k ≤ kkxk, ∀x ∈ X.
Taking limit as n −→ ∞, we have
lim kTn xk ≤ kkxk, ∀x ∈ X
n−→∞
or, k lim Tn xk ≤ kkxk, ∀x ∈ X
n−→∞
or, kT xk ≤ kkxk, ∀x ∈ X.————-(2)
Hence T is a bounded linear operator.
Therefore, T ∈ B(X, Y ).
We next show that Tn −→ T .
kTn − T k −→ 0 as n −→ ∞.
From, (1) if we let m −→ ∞
We get kTn x − T xk ≤ kxk, ∀n ≥ n0 ∀x ∈ X
kTn x−T xk
or, kxk
≤ , ∀n ≥ n0 ∀x ∈ X
kTn x−T xk
or, sup kxk
≤ , ∀n ≥ n0 ∀x ∈X
06=x∈X
or, kTn − T k ≤ , ∀n ≥ n0
i.e. Tn −→ T as n −→ ∞.
Hence, B(X, Y ) is complete.
Linear Functional:- Let X be a metric space over the field K. A mapping f : X −→ K
is called linear functional.
A linear functional f : X −→ K is called bounded linear functional if ∃M > 0
such that |f (x)| ≤ M kxk, ∀x ∈ X.
Note that the norm of f is
|f (x)|
kf k = sup kxk
——–(1)
06=x∈X
or, kf k = sup |f (x)|———(2)
x∈X,kxk=1
from (1) we get,
|f (x)| ≤ kf kkxk.
Dual Spaces:-
Algebraic Dual:-Let X be a vector space over the field K(Ror C). Then the algebraic
dual spaces of X denoted as X 0 consists of all linear functional on X.
Topological Dual:-Let X be a vector space over the field K(Ror C). Then the topo-

16
logical dual of X is the set of all continuous linear functional on X, we call X ∗ , the
topological dual or simply dual space of X.
Canonical Embedding:-Let X be a normed linear space and x ∈ X.
Define = : X −→ X ∗∗ by
=(x)(x∗ ) = x∗ (x), ∀x
= is Linear:-
=(x + y)(x∗ ) = x∗ (x + y)
= x∗ (x) + x∗ (y)
= =(x)(x∗ ) + =(y)(x∗ )
= (=(x) + =(y))(x∗ ), ∀x∗ ∈ X
Therefore, =(x + y) = =(x) + =(y).
=(αx)(x∗ ) = x∗ (αx)
= αx∗ (x)
= α=(x)(x∗ )
= (α=(x))(x∗ ), ∀x ∈ X
Therefore, =(αx) = α=(x).
Therefore, = is linear.
Theorem:-Let X be a finite dimensional normed linear space over the field K. Then
X = X ∗ . Further, dim X ∗ = dim X.
Proof:-Let dim X ∗ = n and let {e1 , e2 , ...en } be a basis(Hamel) of X. Then, every
x ∈ X can be expressed as
Pn
x= αi ei ——(1), where α1 , α2 , ..., αn are scalars in K.
i=1
Define k · k0 : X −→ R
Pn
by kxk0 = |αi |.
i=1
Then, clearly k · k0 is a norm on X.
Since two norms on a finite dimensional space are equivalent, it follows that k · k and
k · k0 are equivalent norms on X. So there exist a constant M > 0 such that
kxk0 ≤ M kxk, ∀x ∈ X
Pn
|αi | ≤ M kxk, ∀x ∈ X————–(2)
i=1

17
For each i = 1, 2, ..., n. Consider the functional fi : X −→ K defined by
fi (x) = αi , where x and αi are related by equation (1)
[ Reason:f1 (x) = f1 (α1 e1 + α2 e2 + ... + αn en )
= α1
f2 (x) = α2
|f1 (x)| = |α1 | ≤ |α1 | + |α2 | + ... + |αn |
kxk0
Clearly, each fi is a linear functional on X.
n
P
Also, |fi (x) = |αi | = |αi | ≤ M kxk———-by(2)
i=1
Thus, each fi is bounded linear functional. Therefore, S = {f1 , f2 , ..., fn } ⊂ X ∗ .
We prove that S forms a basis for X ∗ .
n
P
Let βi αi = 0, ∀βi ∈ K
i=1
n
P
⇒( βi fi )(x) = 0, ∀x ∈ X
i=1
n
P
⇒ (βi fi )(x) = 0, ∀x ∈ X
i=1
Pn
⇒ βi fi (x) = 0, ∀x ∈ X
i=1
Pn
⇒ βi β̄i = 0, ∀fi (x) = β̄i
i=1

n
|βi |2 = 0 (because z z̄ = |z|2
P
⇒
i=1
⇒ |βi |2 = 0, ∀i
⇒ βi = 0, ∀i.
⇒ fi0 s are linearly independent. Next, we want to show that S spans X ∗ , using (1)
For then, let f ∈ X ∗ , using (1)
Pn
Then, f (x) = αi f (xi )
i=1
n
P
= fi (x)f (xi )
i=1
Pn
= fi (x)βi , where βi = f (xi )
i=1
n
P
⇒ f (x) = βi fi (x)
i=1

18
 n
P
⇒f = βi fi .
i=1
Since f ∈ X ∗ , therefore each f ∈ X ∗ can be expressed uniquely as the linear combi-
nation of the members of S and so S spans X ∗ .
Therefore, S is a basis for X ∗ .
Schauder Basis:-If a normed linear linear space X contains a sequence {en } with
property that for every x ∈ X, there is a uniquely sequence of scalars {αn } such that
kx − (α1 e1 + α1 e1 + ... + α1 e1 )k −→ 0 as n −→ ∞
then {en } is called a Schauder basis for X.
Clearly, (we can write)
Pn
x= αi ei (is the expansion of x ∈ X w.r.t {en }).
i=1
e.g.:-lp has Schauder Basis.
Q. Show that the dual space (lp )∗ is isometrically isomorphic to lq , where 1 < p < ∞.
i.e. 1 + 1 = 1 or (lp )∗ ∼
p q = lq , q is the conjugate exponent of p.
Proof:-Consider {en } be Schauder basis of lp , 0 < p < ∞.(where ei = (δij )∗ ))
i.e. {ei } is the sequence whose ith term is 1 and all other terms are zero. Then, each
x ∈ lp can be uniquely expressed as
P∞
x= ξi ei ————(1)
i=1
Let f ∈ (lp )∗ is the dual space of lp .
Since f is linear,
∞
P
therefore, f (x) = ξi f (ei )———–(*)
i=1
Let q be the conjugate exponent of
 p and q.
 |f (ei )|2 iff (ei ) 6= 0, 1 ≤ i ≤ n
(n) (n) f(ei )
Consider xn = {ξi } with ξi = ————(2)
 0 iff (ei ) = 0, i = n + 1
From (*) and (2), we get
∞
P (n)
f (xn ) = ξi f (ei )
i=1
n
P (n)
= ξi f (ei )
i=1
n
(n)
|f (ei )|q (by def. of ξi , by (2))
P
=
i=1
Now, f (xn ) ≤ kf kkxn k

19
 ∞ 1
(n)
|ξi |p ) p
P
= kf k(
i=1
n 1
(n)
|ξi |p ) p
P
= kf k(
i=1
n q 1
( |f|f(e(eii)|)| )p ) p
P
= kf k(
i=1
n 1
(|f (ei )|q−1 )p ) p
P
= kf k(
i=1
n 1
(|f (ei )|p(q−1) ) p
P
= kf k(
i=1
n 1
1 1
(|f (ei )|q ) p (because
P
= kf k( p
+ q
= 1 ⇒ q + p = pq ⇒ q = p(q − 1))
i=1
n
|f (ei )|q ) = f (xn )
P
Therefore,
i=1
n 1
|f (ei )|q ) p
P
≤ kf k(
i=1
n
)|q
P
|f (ei
⇒ i=1
n 1 ≤ kf k
|f (ei )|q ) p
P
(
i=1
n 1
|f (ei )|q )(1− p ) ≤ kf k
P
⇒(
i=1
Taking limit n −→ ∞, we get
∞ 1
⇒ ( |f (ei )|q ) q ≤ kf k———–(**)
P
i=1
⇒ {f (ei )} ∈ lq
Define T : (lp )∗ −→ lq as
T (f ) = {f (ei )}i≥1 = {f (e1 ), f (e2 ), ..., f (en ), ...}, ∀f ∈ (lp )∗
Clearly, T is linear.
Next, we want to show that T is isometry
i.e.kT f k = kf k
∞
P
Now |f (x)| = | ξi f (ei )| (by (*))
i=1
∞
P
≤ |ξi f (ei )|
i=1
∞ 1 ∞ 1
|ξi |p ) p .( |f (ei )|q ) q (by Holders inequality)
P P
≤(
i=1 i=1
∞ 1
|f (ei )|q ) q
P
= kxk(
i=1
Taking supremium over all x of norm 1, we get

20
 ∞ 1
|f (ei )|q ) q
P
sup |f (x)| ≤ (
06=x∈X,kxk=1 i=1
∞ 1
|f (ei )|q ) ———(***)
P
kf k ≤ ( q

i=1
On combining (**) and (***),
∞ 1
kf k = ( |f (ei )|q ) q
P
i=1
= kT (f )k
⇒ T is isometry.
Since an isometry is essentially one-one.
So we have to show that T is onto.
T is onto:
Let y = {yi } ∈ lq
Define a mapping g : lp −→ K as
Now we want to show that g is bounded.
∞
P
Now, |g(x)| = | ξi yi |
i=1
∞ 1 ∞ 1
|ξi |p ) p ( |yi |q ) q
P P
≤(
i=1 i=1
∞ 1
(|yi |q ) q
P
= kxk(
i=1
∞ 1
(|yi |q ) q ≤ ∞)
P
⇒ g is bounded. (∵
i=1
Thus g is a bounded linear functional on lp .
Also, T (g) = {g(e1 ), g(e2 ), ..., g(en ), ...}
= {y1 , y2 , ..., yn , ...}
=y
For y ∈ lq , there exist g functional in (lp )∗ such that
T (g) = y
⇒ T is onto.
Thus (lp )∗ is isometrically isomorphic to lq , i.e. Dual of lp = lq .
Ex. Prove that the dual of C0 i.e. (C0 )∗ is ∼ =(isometrically isomorphic) to l1 .
Proof. Consider Schauder basis {ei } where {ei } is a sequence whose ith term is 1 and
all other terms are zero.

21
 ∞
Then each x ∈ C0∗ can be uniquely written as x =
P
αi ei .
i=1
Let f ∈ C0∗ , where C0∗ is the dual space of C0 .
n
P
If Sn = αi ei
i=1
n
P
⇒ f (Sn ) = αi f (ei )
i=1
Clearly Sn −→ x as n −→ ∞ (by def. of Schauder basis)
⇒ f (Sn ) −→ f (x) as n −→ ∞(because f is cont.)
or lim f (Sn ) = f (x).
n−→∞
Next we shall show that the {f (ei )}∞ 1
i=1 ∈ l .

 x = x0 = {ξi } ∈ C0 , where
Consider
 f (ei ) :f (e ) 6= c, 1 ≤ i ≤ n
f (ei i
ξi =
 0 :f (ei ) = 0, i ≥ n + 1
Clearly, kx0 k∞ = sup |ξi | = 0
n
P
and f (x0 ) = ξi f (ei )
i=1
n n
P f (ei ) P
= |f (ei |
f (ei ) = |f (ei )|
i=1 i=1
Also, f (x0 ) = |f (x0 )| ≤ kf kkx0 k∞
Pn
Therefore, |f (ei )| ≤ kf kkx0 k∞ = kf k
i=1
∞
P
|f (ei )| ≤ kf k———(*)
i=1
⇒ {f (ei )}∞
i=1 ∈ l
1

Now, we define a map T : C0∗ −→ l1 as

T (f ) = {f (ei )}∞ ∗
i=1 , ∀f ∈ C0

Clearly, T is linear.
Now we want to show that T is isometry.
∞
P
For, |f (x)| = | αi f (ei )|
i=1
∞
P
≤ |αi f (ei )|
i=1
P∞
= |αi ||f (ei )|
i=1
∞
P
≤ sup |αi | |f (ei )|
1≤x≤∞ i=1

22
 ∞
P
= kxk∞ |f (ei )|
i=1
Taking supremium over all x of norm
∞
P
i.e. sup |f (x)| ≤ |f (ei )|
x∈C0 ,kxk=1 i=1
∞
P
⇒ kf k ≤ |f (ei )| = kT (f )k———(**)
i=1
Again from (*), we have
kT (f )k ≤ kf k——–(***)
On combining (**) and (***), we have
T is isometry.
Since isometry is essentially one-one, therefore, it is only remain to show that the
mapping T is onto.
For this, let y = {yi } ∈ l1
Define g : C0 −→ K as
P∞
g(x) = αi yi , ∀x = {α i ∈ C0
i=1
Clearly, g is linear mapping.
∞
P
Also |g(x)| = | αi yi |
i=1
∞
P
≤ |αi ||yi |
i=1
P∞
= |αi |yi |
i=1
∞
P
≤ sup |αi | |yi |
i=1
∞
P
= kxk |yi |
i=1
⇒ g is bounded.
Therefore g is bounded linear functional on C0 i.e. g ∈ C0∗ . We have proved that for
every y = {yi } ∈ l1 , ∃g ∈ C0∗ such that
T (g) = {g(ei )}∞
i=1 = {g(e1 ), g(e2 ), ...}

= {y1 , y2 , ...}
= {yi }∞
i=1 = y

⇒ T is onto.
Then T : C0∗ −→ l1 isometrically isomorphic

23
i.e. C0∗ ∼
= l1 .
Ex. Prove that dual space of Rn is isometrically isomorphic to Rn
i.e. (Rn )∗ ∼
=Rn

24
 UNIT-III
Sublinear Functional:-Let X be a real vector space. Then a mapping p : X −→ R
is called sub-additive if
p(x + y) ≤ p(x) + p(y), ∀x, y ∈ X
Also, p is called positive homogeneous if
p(αx) = αp(x), ∀α ≥ 0 in R and x ∈ X.
A sub-additive positive homogeneous function p : X −→ R is called a sublinear func-
tional.
Hahn-Banach Theorem for Real Vector Spaces:-
Let X be a real vector space and let p : X −→ R be a sublinear functional on X. Let
Y be a subspace of X and let f be a linear functional on Y such that f (y) ≤ p(y)(f(y)
is dominated by p(y), ∀y ∈ Y . Then there exist a linear functional F on X such that
F (x) = f (x), ∀x ∈ Y
i.e. F is an extension of f and
F (x) ≤ p(y).
Proof:- Let Y = X, then there is a nothing to prove.
Let Y be a proper subspace of X and let x0 ∈ X \ Y. Let Z = span{Y ∪ {x, y}}. Then,
Z is a proper subspace of X containing Y,
i.e.Z = {y + αx0 : y ∈ Y, α ∈ R}.
Let φ : Z −→ R be defined as
φ(y + αx0 ) = f (y) + αc, where c is a fixed real number.
Then clearly, φ is an extension of f. That is
φ(y) = f (y), ∀y ∈ Y .
We want to show that
φ(y + αx0 ) ≤ p(y + αx0 ).
that is f (y) + αc ≤ p(y + αx0 )
⇒ αc ≤ p(y + αx0 ) − f (y)——–(1)
If α = 0, then (1) is obviously true.
(α = 0, 0 ≤ p(y) − f (y) = f (y) − f (y) = 0)(because f (y) = p(y))

1
If α > 0, then from (1) we have
c ≤ α1 p(y + αx0 ) − α1 f (y)
⇒ c ≤ p( αy + x0 ) − f ( αy )———-(2)
If α < 0, then −α > 0 and from (1), we have
−αc ≥ −p(y + αx0 ) + f (y)
1 1
⇒c≥ −α
. − p(y + αx0 ) + −α
f (y)
y
⇒ c ≥ −p( −α − x0 ) + f ( −y
α
)
Now, we want to show that the existence of c such that
−p( −y
α
− x0 ) + f ( −y
α
) ≤ c ≤ p( αy + x0 ) − f ( αy )
i.e.−p(−y −x0 )−f (y) ≤ c ≤ p(y +x0 )−f (y)———-(3)∀y ∈ Y (because Y is subspace)
Let y1 , y2 ∈ Y . Then,
f (y2 ) − f (y1 ) = f (y2 − y1 ) ≤ p(y2 − y1 )
= p(y2 + x0 − x0 − y1 )
≤ p(y2 + x0 ) + p(−y1 − x0 )
⇒ −p(−y − x0 ) − f (y1 ) ≤ p(y2 + x0 ) − f (y2 )
Let a = sup{−p(−y1 − x0 ) − f (y1 ) : y1 ∈ Y }
b = inf{p(y2 + x0 ) − f (y2 ) : y2 ∈ Y }.
Then a ≤ b. Now if c is such that a ≤ c ≤ b.
Then eq.(4) will be true.
Thus φ is an extension of f.
Let s be the collection of all extension of f which are dominated by p. Then s is non-
empty because φ is in s. Now we define an order relation on s as F1 ≤ F2 ⇔ F2 is an
extension of F1 .
Then clearly (s, ≤) is a partially ordered set(POSET).
Let C = {Fi : i ∈ ∆} be a chain in s.
S
Let Di = Domain of Fi ,∀i ∈ ∆ and D = Di .
i∈∆
Then define G : D −→ R as
G(x) = Fi (x), ∀x ∈ D.
Then, clearly G is well defined linear functional. Also G is an extension of each Fi ∈ ∆.

2
Then G is an upper bound of C.
Therefore, by Zorn’s Lemma, s has a maximal element say F, we now claim that F is
the required functional.
To complete the proof, we need to show that DomF = X.
Suppose DomF 6= X. Then there exist x0 ∈ X \ DomF = D.
Let Z = span{DomF ∪ {x0 }}. Then we can obtains an extension F 0 of F, which is a
contradiction to the maximality of F.
Thus DomF = X.
Hahn-Banach Theorem for Complex Vector Space
Let X be a complex vector space and let p : X −→ R be function defined by
(i) p(x + y) ≤ p(x) + p(y), ∀x, y ∈ X.
(ii) p(αx) = |α|p(x), ∀α ∈ C and x ∈ X.
Let Y be a subspace of X and let f be a linear functional on Y such that |f (x)| ≤
p(x), ∀x ∈ Y . Then there exist a linear functional F on X such that F (x) = f (x), ∀x ∈
Y and |F (x)| ≤ p(x), ∀x ∈ X.
Proof:- Let f (x) = u(x) + ιv(x), where u and v are real and imaginary parts of f.
Then u is a real valued functional which satisfies the following properties
u(x + y) = u(x) + u(y), ∀x, y ∈ Y
u(αx) = αu(x), ∀α ∈ R and x ∈ Y
because
f (αx) = αf (x)
⇒ u(αx) + ιv((αx) = αu(x) + ιαv(x)
Equating real and imaginary parts, we get u(αx) = αu(x)
and v(αx) = αv(x).
Since u is a real valued functional on Y such that u(x) ≤ |f (x)| ≤ p(x), ∀x ∈ Y ,
(because Rez ≤ |z|)
By Hahn-Banach Theorem for real vector space there exist a real linear functional U
which is an extension of u on whole of X such that
U (x) ≤ p(x), ∀x ∈ X

3
and U (x) = u(x), ∀x ∈ Y .
Also, f (ιx) = ιf (x)
⇒ u(ιx) + ιv(ιx) = ιu(x) − v(x)
Equating real and imaginary parts, we get
v(x) = −u(ιx)
and v(ιx) = u(x).
Then we can write
f (x) = u(x) + ιv(x)
= u(x) − ιu(ιx)
Now we define
F (x) = U (x) − ιu(ιx), ∀x ∈ X.
We claim that F is the required extension.
First of all we show that F is a complex linear functional. To see this we need to verify
F (ιx) = ιF (x), ∀x ∈ X because U is a real linear functional.
Now, F (ιx) = U (ιx) − ιU (−x)
= ι{U (x) − ιu(ιx)}
= ιF (x), ∀x ∈ X
Also, for x ∈ Y , we have
F (x) = U (x) − ιu(x)
= f (x).
Finally, we will show that |F (x) ≤ p(x), ∀x ∈ X
If F (x) = 0, then it is obviously true.
If F (x) 6= 0, then F (x) = reιθ , r ∈ R+ , 0 ≤ θ ≤ 2π.
⇒ e−ιθ F (x) = r
⇒ F (e−ιθ x) = r = |F (x)|
⇒ F (e−ιθ x) is real and Im.F (e−ιθ x) = 0
⇒ F (e−ιθ x) = U (e−ιθ x)
⇒ |F (x)| = F (e−ιθ x) = U (e−ιθ x)
≤ p(e−ιθ x)

4
 = |e−ιθ |p(x)
= p(x), ∀x
⇒ |F (x)| ≤ p(x).
This completes the proof.
Hahn-Banach Theorem for Normed Linear Space:
Let X be a normed linear space and Y be a subspace of X. Let f be a bounded linear
functional on Y. Then there exist a bounded linear functional F on X which extends
f and kF k = kf k.
Proof: Let p : X −→ R be defined as
p(x) = kf kkxk, ∀x ∈ X.
Then p(x + y) = kx + ykkf k
≤ kf kkxk + kf kkyk
= p(x) + p(y)
and p(αx) = kαxkkf k
= |α|kxkkf k
= |α|p(x), ∀x ∈ X and ∀α ∈ C
Therefore, p is sublinear functional on X.
Also, |f (x)| ≤ kf kkxk
= p(x), ∀x ∈ X.
Hence by Hahn-Banach Theorem for Complex Vector Space there exist a linear func-
tional F on X which extends f and
|F (x)| ≤ p(x), ∀x ∈ X
= kf kkxk.
Clearly, Fn is bounded linear functional on X. Now we only need to show that
kF k = kf k.
Since |F (x)| ≤ p(x) ≤ kf kkxk, ∀x ∈ X
|F (x)|
⇒ sup kxk
= kf k
06=x∈Xorkxk=1
⇒ kF k ≤ kf k———-(1)
Since F is an extension of f

5
Therefore |f (x)| = |F (x)| ≤ kF kkxk
|f (x)|
⇒ sup kxk
≤ kF k
06=x∈Xorkxk=1
∴ kf k ≤ kF k———(2)
From (1) and (2), we get kf k = kF k.
Application of Hahn Banach Theorem:
Let X be a normed linear space and let 0 6= x0 ∈ X. Then there exist a functional
f ∈ X ∗ such that f (x0 ) = kx0 k and kf k = 1.
Proof: Let Z = {αx0 } be the linear subspace of X generated by x0 . Define
f0 : Z −→ K by
f0 (αx0 ) = αkx0 k.
We show that f0 is a functional on Z such that kf0 k = 1
f0 is linear:-
Let y1 , y2 ∈ Z. Then y1 = αx0 and y2 = βx0 , for some scalars α and β.
If a and b are any scalars then,
f0 (ay1 + by2 ) = f0 (aαx0 + bβx0 )
= f0 (aα + bβ)(x0 )
= (aα + bβ)kx0 k
= af0 (αx0 ) + bf0 (βx0 )
= af0 (y1 ) + bf0 (y2 )
f0 is bounded:-
Let y = αx0 ∈ Z. Thus
kyk = kαx0 k
= |α|kx0 k
Now, |f0 (y)| = |f0 (αx0 )|
= |αkx0 k|
= |α|kx0 k
= kαx0 k
= kyk———(*)
Thus, f0 is bounded linear functional on Z. Hence by Hahn Banach theorem for Normed

6
linear space, there exist a bounded linear functional F on X which is an extension of
f and kF k = kf0 k————(**)
From (*) |f0 (y) = kyk ⇒ sup |f0 (x)| = 1 ⇒ kf0 k = 1
kyk=1
Further, kf0 k = sup{|f0 (y)| : y ∈ Z : kyk ≤ 1}
= sup{kyk : y ∈ Z, kyk ≤ 1}
=1
Also, f0 (x0 ) = kx0 k
Hence, by Hahn-Banach theorem, f0 can be extended to norm preserving functional
f ∈ X ∗ such that
f (x0 ) = f0 (x0 ) = kx0 k
and kf k = kf0 k = 1
Remarks: In particular, if x 6= y, then there exist f ∈ X ∗ such that f (x) 6= f (y).
Sol:- Since x 6= y, therefore x − y 6= 0. So there exist f ∈ X ∗ such that
f (x − y) = kx − yk =
6 0
⇒ f (x) − f (y) 6= 0
⇒ f (x) 6= f (y). This shows that X ∗ separates vector in X.
Ex. Let X be a real normed linear space and suppose f (x) = 0, ∀f ∈ X ∗ . Then show
that x = 0.
Sol. Suppose x 6= 0.
Then f ∈ X ∗ such that f (x) = kxk =
6 0
⇒ f (x) 6= 0 which is on contradiction.
Hence x = 0.
Theorem:- Let X be a normed linear space over the field K and x ∈ X. Then
kxk = sup{ |fkf(x)|
k
, f ∈ X ∗ and f 6= 0}.
Proof:- If x = 0, then there is nothing to prove.
Let 0 6= x ∈ X. Then by application (1) there exist f ∈ X ∗ such that f (x) = kxk and
kf k = 1.
Therefore, sup{ |fkf(x)|
k
, f ∈ X ∗ , f 6= 0}
|f (x)|
≥ kf k
= kxk.

7
For the reverse inequality,
|f (x)| ≤ kf kkxk, ∀f ∈ X ∗
⇒ sup{ |fkf(x)|
k
, f ∈ X ∗ , f 6= 0}
≤ sup{ kf (x)kkxk
kf k
, f ∈ X ∗ , f 6= 0} = kxk.
Hence, kxk = sup{ |fkf(x)|
k
, f ∈ X ∗ and f 6= 0}.
Application:HBT
Let Y be any subspace of a normed linear space X and x0 ∈ X be such that the
distance from x0 to Y is d(x0 , Y ), d is +ve. Then there is a continuous(bd) linear
functional F on X such that kF k = 1 and F (Y ) = {0} and F (x0 ) = d.
Proof: Since d > 0,
∴ x0 ∈
/ Y.
Consider the subspace Y0 = {y + αx0 : y ∈ Y, α ∈ K} spanned by Y and x0 .
Clearly, the representation of each x ∈ Y0 in the form x = y + αx0 is unique.
[Otherwise, x0 ∈ Y which is a contradiction.]
Now define a mapping f : Y0 −→ K
as f (x) = αd, ∀x = y + αx0 ∈ Y0
Clearly this map is well-defined.
f is linear on Y:
For this, let x1 = y1 + αx0 and x2 = y2 + αx0 ∈ Y0 and α, β be scalars. Then
f (αx1 + βx2 ) = f (α(y1 + αx0 ) + β(y2 + αx0 ))
= f (αy1 + α.α1 x0 + βy2 + βα2 x0 )
= f (αy1 + βy2 + (αα1 + βα2 )x0 )
= (αα1 + βα2 )d
= αα1 d + βα2 d
= α(α1 d) + β(α2 d)
= αf (y1 + α1 x0 ) + βf (y2 + α2 x0 )
= αf (x1 ) + βf (x0 )
⇒ f is linear.
f is bounded:-

8
Let x = y + αx0 ∈ Y and α 6= 0 is scalars.
We claim that kf k = 1
Now kxk = kαx0 + yk
= |α|kx0 + αy k
= |α|kx0 − ( −y
α
k
≥ |α|d = |αd| = |f (x)|
Or |f (x)| ≤ kxk∀x
This shows that f is a bounded linear functional on X and kf k ≤ 1.—–(1)
Now f (x0 ) = f (0 + 1.x0 ) = 1.d = d
i.e.f (x0 ) = d———(1)
Also, for y ∈ Y
f (y) = f (y + +0.x0 ) = 0.d = 0
⇒ f (y) = 0, ∀y ∈ Y
⇒ f (Y ) = {0}——–(2)
We claim that kf k = 1
Now kf k = sup{ |fkxk
(x)|
; x ∈ Y0 , x 6= 0}
= sup{ |1+(y+αx 0 )|
ky+αx0 k
: y ∈ Y and 0 6= α ∈ K}
|αd|
= sup{ ky+αx 0k
: y ∈ Y, 0 6= α ∈ K}
= sup{ |α|kx|α|d y
0 −( )k
: y
α
∈ Y, 0 6= α ∈ K}
α

= d sup{ kx01−zk : z = − αy ∈ Y }
d
= inf{kx0 −zk:z∈Y }
d
= d
=1
⇒ kf k = 1——–(3)
Since Y is a normed linear space and Y0 is a subspace of X and f is a bounded linear
functional on Y. Then by HBT for NLS, a bounded linear functional F on X such
that F (x) = f (x), ∀x ∈ Y0 and kF k = kf k
from (1), (2) and (3), we have,
F (Y ) = f (Y ) = {0}, F (x0 ) = f (x0 ) = d and kF k = kf k = 1.
Def. Let X be a vector space and Y be a subspace of X. Then there exist a subspace

9
Z of X such that X = Y + Z and Y ∩ Z = {0}. Then the subspace Z is called the
complimentary subspace of Y. The dimension of Z is called co-dimension of Y or defi-
ciency of Y.
Note:- Let X be a vector space and f be a linear functional on X. Then ker f is the
co-dimension 1 or 0.
Cor.:- If X 6= 0 be a normed linear space. Then its conjugate space or dual space i.e.
X ∗ 6= {0}.
Application of Hahn-Banach Theorem:-
Preposition:- Let X be an infinite dimensional Banach Space. Then dimX ≥ c.
Proof:- Since X ∗ 6= {0}, where X ∗ is the set of all continuous linear functional on X.
Let 0 6= f ∈ X ∗ . Let Z1 = ker f . Then Z1 is a subspace of co-dimension 1, i.e.
codimension Z1 = 1.
Let x1 ∈ X \Z1 such that kx1 k = 1. Since co-dimension of Z1 is finite therefore Z1 is fi-
nite dimensional subspace. Let f2 be a non-zero functional on Z1 and take Z2 = ker f2 .
Let x2 ∈ Z1 \ Z2 such that kx2 k = 12 . By induction, we obtain a sequence {xn } such
1
that for every n ∈ N, kxn k = 2n−1
and a decreasing sequence of closed subspace as
X ⊃ Z1 ⊃ Z2 ⊃ Z3 ⊃ ... ⊃ Zn ...
such that for every n,
x1 , x2 , ..., xn ∈
/ Zn
Now we establish a vector space isomorphism between the vector space l∞ of bounded
sequence and a subspace of X.
∞
Let {αi } ∈ l∞ . The series
P
αn xn has the partial sum
n=1
n
P
sn = αi xi = α1 x1 + α2 x2 + ... + αn xn
i=1
Now we show that sn is a Cauchy sequence.
For m > n, we have
ksm − sn k = kαn+1 xn+1 + ... + αm xm k
≤ k{kxn+1 k + ... + kxm k}, where k = max{|αi |}
< k{ 21n + ... + 1
2m−1
}
1
< k 2n−1 −→ 0 as n −→ 0.

10
Thus {sn } is a Cauchy sequence.
Since X is complete, ∃x ∈ X such that
Pn ∞
x = lim sn = lim αi xi = lim αi xi
n n i=1 i=1
Define φ : l∞ −→ X as
∞
αi xi , ∀{αi } ∈ l∞ .
P
φ({αi }) =
i=1
Then φ is linear.
For, let {αi }, {βi } ∈ l∞ . Then
φ({αi } + {βi }) = φ({αi + βi })
∞
P
= (αi + βi )(xi )
i=1
P∞ ∞
P
= α i xi + βi xi
i=1 i=1
= φ({αi }) + φ({βi })
Let α ∈ K. Then
∞
P
φ(α{αi }) = φ({ααi }) = (ααi )(xi )
i=1
∞
P
=α α i xi
i=1
= αφ({αi }).
Now we claim that φ is 1 − 1.
Suppose φ({αi } = 0
P∞
⇒ α i xi = 0
i=1
∞
P
⇒ α1 x1 = −αi xi
i=2
∞
P
But −αi xi ∈ Z1 and x1 ∈
/ Z1 (Z1 is a closed subspace).
i=2
∞
P
Hence −αi xi = 0
i=2
⇒ α1 x1 = 0
⇒ α1 = 0
In a similar manner, α2 = 0, α3 = 0, ..., αn = 0, n
Thus {αi } = 0(i.e, {αi } is a zero subspace)
Thus l∞ is isomorphic to a subspace of X. Hence dim X ≥ dim l∞ . To complete the
proof we need to show that dim l∞ ≥ c.

11
For each t ∈ (0, 1), we define the sequence et = {1, t, t2 , ...}
Then {et : t ∈ (0, 1)} ⊂ l∞
It can be seen that {et : t ∈ (0, 1)} is
BANACH STEINHAUS THEOREM
UNIFORM BOUNDED PRINCIPLE:-
Let X be a Banach space and Y be a normed linear space. Let {Tn } be a sequence of
bounded linear transformation from X to Y such that {Tn (x)} is bounded sequence,
∀x ∈ X. Then {kTn k} is bounded sequence.
Proof:- For each k ∈ N, let Ak = {x ∈ X : kTn (x)k ≤ k, ∀n ∈ N}———(*)
We first show that each Ak is a closed subset of X.
For this, let x ∈ Ak
⇒ ∃ a sequence {xj } ∈ Ak such that xj −→ x as j −→ ∞
⇒ kTn (xj )k ≤ k, ∀n ∈ N———-(**)
Since each Tn is continuous (being g bounded)
∴ xj −→ x ⇒ Tn (xj ) −→ Tn (x)
⇒ kTn (xj )k −→ kTn (x)k∀n
⇒ kTn (x) ≤ k (by (**))
⇒ x ∈ Ak .
Since x ∈ Ak . Therefore, we concluded that Ak ⊆ Ak
Also by def. of closure of sets we have Ak ⊆ Ak
So Ak = Ak .
⇒ each Ak is a closed subset of X.
S
Next we show that X = Ak .
S k∈N
For, suppose X 6= Ak
k∈N
⇒ ∃x ∈ X such that x ∈
/ Ak ∀x ∈ N
⇒ kTn (x)k > k∀n ∈ N, k ∈ N
⇒ {Tn (x)} is unbounded, which is a contradiction to the given hypothesis.
S
Our supposition was wrong and so X = Ak .
k∈N
Since X is a Banach space(i.e. Complete normed linear space)

12
Therefore by Baire’s Category theorem, X is of Second Category and so there is some
Ak , say Ak0 , which contains an open sphere say B0 = B(x0 , r) ⊆ A0 ———-(1)
r
Now let 0 6= x ∈ X and take z = x0 + λx——–(2) where λ = 2kxk

⇒ kz − x0 k = kλxk
= |λ|kxk
r
= 2kxk
kxk
r
= 2

<r
i.e. kz − x0 k < r
⇒ z ∈ B(x0 , r).
i.e z ∈ B0
and by (1) z ∈ Ak0
⇒ kTn (z)k ≤ k0 ∀n——–(by (*))
− − − − − − − − − (3)
Since x0 ∈ Ak0 ⇒ kTn (x0 )k ≤ k0 ∀n——(4)
Now ∀n ∈ N, we have
kTn (x)k = kTn (z−x
λ
0)
k——(by(2))
= k λ1 Tn (z − x0 )k
= λ1 kTn (z) − Tn (x0 )k
≤ λ1 kTn (z)k + kTn (x0 )k
≤ λ1 2k0 (by (3) and (4))
2kxk 4k0
i.e. kTn (x)k ≤ r
2k0 = r
kxk
kTn (x)k 4k0
⇒ kxk
≤ r
kTn (x)k 4k0
⇒ sup kxk
≤ r
06=x∈X
4k0
⇒ kTn k ≤ r
, ∀n
4k0
⇒ kTn k ≤ M where M = r
is +ve.
⇒ {kTn k} is bounded sequence.
Def.:- Let X and Y be two metric spaces. Then T : X −→ Y is called an open map
if for every open set E in X, the image of T(E) is an open set in Y.

13
Eg.:- Consider the metric space (R, U ) and (R, d) where U and d are usual and dis-
crete metric on R. Then I : (R, U ) −→ (R, d) is an open map but not continuous,
where I is an identity mapping. On the other hand I : (R, d) −→ (R, U ) is continuous
but not open.
Lemma:-Let T be bounded linear transformation from Banach space X onto Y. Then
the image of open unit ball say S0 = S(0, 1) under T contains an open ball centered
at the origin in Y.
Proof:- Consider the open ball S1 = S(0, 21 ) ⊂ X.
For any x ∈ X, let k > 2kxk
k
⇒ kxk < 2

⇒ x ∈ S(0, 12 ) = kS1
⇒ x ∈ kS1 , with k sufficiently large (k > 2kxk)
Then X = ∪∞
k=1 kS1 .

Since T is surjective, we have Y = T (X)

= T (∪∞
k=1 kS1 )

= ∪∞
k=1 T (kS1 )

= ∪∞
k=1 kT (S1 )

Here by taking closure we dont add further points to the union. Since the union is
already the whole space Y. The space Y being complete. By Baire’s Category theorem,
it is of second category. Thus kT (S1 ) must contain an open sphere. Thus T (S1 ) also
contains an open sphere say S ∗ = S(y0 , ) ⊂ T (S1 )——-(1)
It follows that S ∗ − y0 = S(0, ) ⊂ T (S1 ) − y0 ———(2)
Now we shall show that
S ∗ − y0 ⊂ T (S0 ) where S0 = S(0, 1) is the unit sphere(ball).
Let y ∈ T (S1 ) − y0
y + y0 ∈ T (S1 )
⇒ ∃ a sequence {un } in T (S1 ) such that un −→ y + y0 .
Then un = T (wn ) for some wn ∈ S1 .
Also, since y0 ∈ T (S1 ),

14
⇒ ∃ a sequence {vn } in T (S1 ) such that vn −→ y0
Then vn = T (zn ) for some zn ∈ S1
Now kwn − zn k ≤ kwn k + kzn k
1 1
≤ 2
+ 2
= 1[∵ S1 has radius 12 ]
⇒ kwn − zn k ∈ S0
⇒ wn − zn ∈ B(0, 1) = S0 ∀n———(3)
and T (wn − zn ) = T (wn ) − T (zn )
= un − vn ∀n
−→ y + y0 − y0 −→ y as n −→ ∞
By (3), T (wn − zn ) ∈ T (S0 )∀n
⇒ y ∈ T (S0 )
Since y ∈ T (S1 ) − y0
∴ T (S1 ) − y0 ⊂ T (S0 )
hence from (2) we have
S ∗ − y0 ⊂ T (S0 )———(4)
Let Sn = S(0, 21n ) then by linearity of T,
we have T (Sn ) = T (S(0, 21n ))
= T ( 21n S(0, 1))
1
= 2n
T (B(0, 1))
= 21n T (S0 )———–(5)
Put vn = S(0, 2n ) = 21n S(0, )
from (4) we get
1
2n
S(0, ) ⊂ 1
2n
T (S0 )[by (4), S ∗ − y0 ⊂ T (S0 ) ⇒ S(0, ) ⊂ T (S0 )]
1
⇒ vn ⊂ 2n
T (S0 )
⊂ T (Sn ) [by (5)]——(6)
Now we shall show that v1 ⊂ T (S0 )
For this let y ∈ v1 , then by (6), we get y ∈ T (S1 )

⇒ ∃v ∈ T (S1 ) such that ky − vk < 4

Also v = T (x1 ) for some x1 ∈ S1 and

15
ky − T (x1 )k < 4 = 22 .
Again from (6), for n=2, we get
y − T (x1 ) ∈ v2 ⊂ T (S2 )
⇒ y − T (x1 ) ∈ T (S2 )

⇒ ∃x2 ∈ S2 such that ky − T (x1 ) − T (x2 )k < 23

Proceeding in this way, we get


ky − T (x1 ) − T (x2 ) − ... − T (xn )k ≤ 2n+1

——–(*)

i.e. ky − T (x1 + x2 + ... + xn )k < 2n+1
Pn
Put zn = x1 + x2 + ... + xn = xk
k=1
Since xk ∈ Sk ∀k
1
∴ kxk k < 2k
, ∀k
Thus for m < n, we get
n
P Pm
kzn − zm k = k xk − xk k
k=1 k=1
n
P
=k xk k
k=m+1
Pn
≤ kxk k
k=m+1
n
1
P
‘< 2k
−→ 0 as n −→ ∞
k=m+1
⇒ {zn } is a Cauchy sequence in X. But X is complete, so ∃z ∈ X such that zn −→ z
as n −→ ∞——(7)
n
P n
P
Now kzn k = k xk k ≤ kxk k
k=1 k=1
∞ ∞
1
P P
< kxk k < 2k
=1
k=1 k=1
i.e kzn k < 1, ∀n ∈ N
⇒ zn ∈ S0 [∵ S0 = S(0, 1)]
Pn ∞
P
Since zn = xk and zn −→ z as n −→ ∞, it follows that the series xn = z.
k=1 n=1
∞ ∞ ∞
1
P P P
Also kzk = k xn k ≤ kxn k < 2n
=1
n=1 n=1 n=1
⇒ z ∈ S0 ———-(8)
Also, we have T (zn ) −→ y as n −→ ∞ (by (*)) and from (7), we have zn −→ z as

16
n −→ ∞
⇒ By continuity of T, we have
T (zn ) −→ T (z) as n −→ ∞ and we have y = T (z)
Also, by (8), z ∈ S0
⇒ T (z) ∈ T (S0 )
⇒ y ∈ T (S0 )
Since y ∈ V1
∴ V1 ∈ T (S0 )
i.e.B(0, 12 ) ⊂ T (S0 )
This proves the lemma.
Cor. to the Open Mapping Theorem:- Let X and Y be two Banach spaces. Suppose
T : X −→ Y be a bijective bounded linerar transformation. Then T −1 : Y −→ X is
also a bounded linear transformation.
To show that T : X −→ Y is an open map. Let A be an open subset of X, we want
to show that T (A) is open in Y.
For this, let y ∈ T (A). Then y = T (x) for some x ∈ A
Since x ∈ A and A is open in X
⇒ there is an open ball say B(x, r) such that
B(x, r) ⊆ A
⇒ x + B(0, r) ⊆ A
⇒ B(0, r) ⊆ A − x
⇒ rB(0, 1) ⊆ A − x
⇒ B(0, 1) ⊆ 1r [A − x]
1
⇒ B(0, 1) ⊆ k[A − x] where k = r

⇒ T (B(0, 1)) ⊆ T {k[A − x]}

=k{T(A-x)}
=k{T(A)-T(x)}
⇒ T (B(0, 1)) ⊆ k{T (A) − T (x)}
So by lemma, there exist a ball with centre origin say B(0, δ) such that

17
B(0, δ) ⊂ T (B(0, 1)) ⊆ k{T (A) − T (x)}
⇒ B(0, δ) ⊆ k{T (A) − T (x)}
⇒ k1 B(0, δ) ⊆ T (A) − y
⇒ B(0, kδ ) ⊆ T (A) − y
⇒ y + B(0, kδ ) ⊆ T (A)
⇒ B(y, kδ ) ⊆ T (A)
Since y ∈ T (A), so we can conclude that every point of T(A) is an interior point.
⇒ T (A) is open in Y and so T : X → Y is an open mapping.
Bounded Inverse Theorem:-
Let X and Y be Banach spaces and T : X −→ Y be a bounded linear operator. If T
is bijective, then T −1 is a bounded linear operator from Y onto X.
Proof:-By Open Mapping theorem, T : X −→ Y is surjective(as it is bijective) is an
open mapping.
Since T : X −→ Y is bijective, its inverse T −1 : X −→ Y and is linear. T (G) is open
in Y, whenever G is open in X. Further (T −1 )−1 (G) = T (G). Hence T −1 is continuous
and therefore bounded.
Def.:- Let X and Y be two vector spaces. Then we know that XY is a vector space
with two algebraic operations defined as:
(x1 , y1 ) + (x2 , y2 ) = (x1 + x2 , y1 + y2 ), ∀(x1 , y1 ), (x2 , y2 ) ∈ X × Y
α(x, y) = (αx, αy), ∀α ∈ K and (x, y) ∈ X × Y .
Let X and Y be a normed linear space with the norm defined by
k(x, y)k = kxk + kyk.
Then (X × Y, k · k) is a normed linear space.
Def.:-Closed linear operators:-
Let X and Y be two linear spaces and T : D(T ) −→ Y be a linear operator with
domain D(T ) ⊂ X. Then T is called a closed linear operator if its graph
G(T ) = {(x, T (x)} : x ∈ D(T )}
is closed in the normed linear space in X × Y .
We will show that X × Y is a Banach space.

18
Let {zn } be a Cauchy sequence in X × Y , where zn = (xn , yn ).
We want to show that zn is convergent in X × Y .
Since zn is a Cauchy sequence in X × Y, ∃ some +ve integers N such that
kzn − zm k = k(xn , yn ) − (xm , ym )k
= kxn − xm , yn − ym )k
= kxn − xm k + kyn − ym k < , ∀n, m ≥ N ———(2)
Thus {xn } and {yn } are Cauchy sequences in X and Y respectively. Since X and Y
are all Banach Spaces so they converge say xn −→ x and yn −→ y. Thus from eq.(2)
we can conclude that zn −→ z as n −→ ∞.
This implies {zn } is convergent in X × Y . Since {zn } is Cauchy sequence in X × Y ∴
we conclude that every Cauchy sequence is convergent in X × Y .
Therefore X × Y is a complete normed linear space.
Hence X × Y is a Banach Space.
Since T is a closed operator so its graph G(T ) is closed in X × Y . Also by hypothesis,
D(T ) is closed in X. Thus G(T ) and D(T ) are Banach spaces.
Consider a mapping
p : G(T ) −→ D(T )
defined by p((x, T (x)) = x, ∀(x, T (x)) ∈ G(T )
Then P is linear.
Now P is one-one:-
For this, let (x, T (x)) ∈ ker P
⇒ P (x, T (x)) = 0
⇒x=0
⇒ T (x) = 0
⇒ ker P = {(0, 0)}
⇒ P is one-one.
P is onto:-
Let x ∈ D, then (x, T (x)) ∈ G(T ) (by def. of G(T)) and P (x, T (x)) = x [by def. of P]
Next, we want to show that P is bounded.

19
For this, let (x, T (x)) ∈ G(T )
Then, kP (x, T (x))k = kxk
≤ kxk + kT xk
= kx, T (x)k
⇒ P is bounded on G(T).
Hence by Bounded inverse theorem, P −1 : D −→ G(T ) is also bounded.
i.e. ∃M > 0 such that
kP −1 (x)k ≤ M kxk, ∀x ∈ D.
Now, for x ∈ D
kT (x)k ≤ kxk + kT (x)k
= kx, T (x)k = kP −1 (x)k
≤ M kxk
⇒ kT (x)k ≤ M kxk
⇒ T is bounded.
Lemma:- Let T : D ⊆ X −→ Y be a bounded linear transformation, where X and Y
are normed linear space. Then
(a) If D is a closed subset of X, then T is closed.
(b) If T is closed and Y is complete. Then D is a closed subset of X.
Proof:-(a) To show that T is closed.
Let {xn } be a sequence in D such that xn −→ x and T (xn ) −→ y.
We want to show that x ∈ D and T (x) = y
Since {xn } is a sequence in D and xn −→ x
∴x∈D
⇒ x ∈ D——–(*)
Also, since T is continuous and xn −→ x
⇒ T (xn ) −→ T (x)
Thus T (x) = y—–(**)
On combining (*) and (**), we have
x ∈ D and T (x) = y

20
⇒ T is closed.
(b) To show that D is closed in X.
Let x ∈ D, then ∃ a sequence {xn } in D such that xn −→ x
Also, since T is bounded.
∴ kT (xn ) − T (xm )k = kT (xn − xm )k ≤ kT kkxn − xm k −→ 0
Exercise:- Give an example to show that a closed linear transformation need not be
bounded.
Proof:- Let X = Y = C[0, 1] with supnorm and let D ⊆ X consists of all those
functions that have continuous derivatives
i.e. D = {x ∈ X : x0 ∈ C[0, 1]}
For any x ∈ D, let T : D ⊆ X −→ Y defined as T (x) = x0 , ∀x ∈ D.
Clearly, T is linear transformation from D to Y.
First show that T is not bounded.
For, let xn = tn , t ∈ [0, 1] for n = 1, 2, .., n − 1
0
then, T (xn )(t) = ntn−1 and kT xn k∞ = sup |xn (t)| = sup |ntn−1 |
t∈[0,1] t∈[0,1]
n
Clearly,kxn k = 1[∵ kxn k = sup |x(t)| = sup |t | = 1]
t∈[0,1] t∈[0,1]
kT xn k = n∀n
⇒ T is not bounded.
T is closed:-
For, let xn −→ x and T xn −→ y
0
i.e. (xn −→ y) where xn ∈ D.
Since the convergence is uniform
0 0 0
∴ x = lim xn is differentiable and x = y i.e. x exist and x = y
n
⇒ x ∈ D and T x = y
⇒ T is closed.
Def.:- Let X be a normed linear space over the field K. Given x ∈ X, let τx : X ∗ −→ K
be defined as τx (f ) = f (x), ∀f ∈ X ∗ .
Then τx is a bounded linear functional on X ∗ i.e. τ∗ ∈ X ∗
Further, define F : X −→ X ∗∗ as

21
 F (x) = τx , ∀x ∈ X
Then F is a linear isometry of X into X ∗∗ and is called the canonical embedding of X
into X ∗ .
Def.:-Reflexive Space:-
A normed linear space X is said to be a reflexive space, if the canonical embedding
map onto its 2nd dual X ∗∗ .
∗
Example:- Euclidean space Rn is reflexive.[∵ (Rn )∗∗ = (Rn )∗ = (Rn )∗ = Rn ]
Theorem:- A normed linear space is separable if its dual space is separable.
Proof:- Let X be a normed linear space such that its dual space X ∗ is separa-
ble.[Exercise]
Theorem:- If X is reflexive normed linear space, then show that X ∗ is also reflexive.
or
A closed subspace of a reflexive Banach space is reflexive.
Proof:- Suppose a Banach space X is reflexive. Then the canonical embedding,
F : X −→ X ∗∗ defined by F (x)(x∗ ) = x∗ (x), ∀x ∈ X, x ∈ X ∗ is surjective.——–
(1)
We have to show that the dual space of X is reflexive.
i.e. the canonical embedding F : X ∗ −→ X ∗∗∗ defined as
0
F (x∗ )(x∗∗ ) = x∗∗ (x∗ ), ∀x∗ ∈ X ∗ , x∗∗ ∈ X ∗∗ ——–(2)
For this, let x∗∗∗
0 ∈ X ∗∗∗ , we must show that there exist x∗0 ∈ X ∗ such that
0
F (x∗0 ) = x∗∗∗
0

So define x∗0 (x) = x∗∗∗

0 (F (x))———(3)

Clearly, x∗0 ∈ X ∗
0
We need to show that the F (x∗0 ) = x∗∗∗
0
0
i.e. F (x∗0 )(x∗∗ ) = x∗∗
0 , ∀x
∗∗
∈ X ∗∗
For this, let x∗∗ ∈ X ∗∗ . Since F : X −→ X ∗∗ is onto.
∴ ∃x ∈ X such that F (x) = x∗∗ ——–(4)
0
Now, F (x∗0 )(x∗∗ ) = x∗∗ (x∗0 )——-[by (2)]
= F (x)(x∗0 )———[by (4)]

22
 = x∗0 (x)————[by (1)]
= x∗∗∗
0 (F (x))——-[by (3)]

= x∗∗∗ ∗∗
0 (x )—–[by (4)]
0
⇒ F (x∗0 )(x∗∗ ) = x∗∗∗ ∗∗
0 (x )

Since x∗∗ ∈ X ∗∗
0
∴ F (x∗0 )(x∗∗ ) = x∗∗∗ ∗∗
0 (x )∀x
∗∗
∈ X ∗∗
0
⇒ F (x∗0 ) = x∗∗∗
0
0
⇒ the mapping F : X ∗ −→ X ∗∗∗ is onto.
∴ X ∗ is reflexive, i.e. the dual space of X.
Conversely, suppose that the dual space of X i.e. X ∗ is reflexive.(means that
0
F : X ∗ −→ X ∗∗∗ is onto)
We have to show that X is reflexive.
On the contrary, suppose X is not reflexive. This means that the canonical embedding
F : X −→ X ∗∗ is not surjective i.e. F (X) 6= X ∗∗
Since F is isometry and F(x) is a proper subspace of X ∗∗ .
Hence by application of Hahn Banach Theorem, there exist 0 6= x∗∗∗
0 ∈ X ∗∗∗ such that
x∗∗∗
0 (F (x)) = 0.
0
We claim that x∗∗∗
0 / F (x∗ )
∈
0
Suppose on the contrary, x∗∗∗
0 ∈ F (x∗ )
0
∴ ∃x∗0 ∈ X ∗ such that F (x∗0 ) = x∗∗∗
0 ———(5)

Now x∗0 (x) = F (x)(x∗0 )[by def. of canonical embedding of F]

= x∗∗∗
0 (F (x))

=0
⇒ x∗0 (x) = 0∀x ∈ X
⇒ x∗0 = 0
0 0
But x∗∗∗
0 = F (x∗0 ) = F (0) = 0
⇒ x∗∗∗
0 =0
which is contradiction to the fact that x∗∗∗
0 6= 0
∴ Our supposition is wrong

23
 0
⇒ x∗∗∗
0 / F (x∗ )
∈
0
i.e.F (x∗ ) 6= x∗∗∗
which is again a contradiction to the fact that the dual space of X i.e. X ∗ is reflexive
Therefore, our supposition is wrong. Hence Banach space.
‘
‘
‘

24
 Unit IV
Definition: Let X be a vector space over K(or C). A mapping
‘ < ·, · >: X × X → K
is said to be an inner product(on X)if for all scalar α ∈ K, we have
(i) < x, x >≥ 0 and < x, x >= 0 ⇔ x = 0
(ii) < x, y >=< y, x > (Conjugate Symmetry)
(iii) < αx, y >= α < x, y >
(iv) < x + y, z >=< x, z > + < y, z >.
The scalar < x, y > is called the inner product of x and y. A vector space X equipped with an inner
product < ·, · > defined on it is called an inner product space.
Remark: (i) < αx + βy, z >= α < x, z > +β < y, z > is linear in Ist co-ordinate.
(ii) < x, αy >=< αy, x >= α< y, x >
= α < y, x >
= α < x, y >
(iii) < x, αy + βz >= α < x, y > +β < x, z >(Conjugate linear in 2nd co-ordinate)
Example: (1) Cn is an inner product space defined by
Pn
< z, w >= zi wi
i=1
where z = (z1 , z2 , ..., zn ) ∈ Cn
w = (w1 , w2 , ..., wn ) ∈ Cn
(2) Consider the linear space l2 . For vectors x = {xi } an y = {yi } ∈ l2 , define
P∞
< x, y >= xi yi
i=1
Then the space (l2 , < ·, · >) is an inner product space.
(3) Consider the linear space c([a, b]) the space of all continuous complex valued function on [a,b].
For f, g ∈ c([a, b]) define
Rb
< f, g >= a
f (t)g(t)dt
Then the space (c([a, b]), < ·, · >) is an inner product space.
Ex.: Every inner product space is NLS.
Proof: An inner product space on X defined a norm on X given by
√
kxk2 =< x, x > (i.ekxk = < x, x >
and a metric on X given by
√
d(x, y) = kx − yk = < x − y, x − y >
(i) kxk ≥ 0, since < x, x >≥ 0
√
⇒ < x, x > ≥ 0 ⇒ kxk ≥ 0
(ii) kxk = 0 ⇔ x = 0, since < x, x >= 0 ⇔ x = 0
(iii) kαxk2 =< αx, αx >= αα < x, x >

2
 = |α|2 < x, x >
= |α|2 kxk2
⇒ kαxk = |α|kxk
(iv) Finally to establish triangular inequality
kx + yk2 =< x + y, x + y >
=< x, x > + < x, y > + < y, x > + < y, y >
= kxk2 + kyk2 + < x, y > + < x, y >
= kxk2 + kyk2 + 2Re< x, y >
≤ kxk2 + kyk2 + 2| < x, y > |
≤ kxk2 + kyk2 + 2kxkkyk
= (kxk + kyk)2
Hence kx + yk ≤ kxk + kyk.
Def.: A complete inner product space is called Hilbert space.
Or
An inner product space (X, < ·, · >) over the field K(R or C) is said to be Hilbert space if X is
complete w.r.t the metric induced by inner product.
Parallelogram law: Let X be an inner product space. Then
kx + yk2 + kx − yk2 = 2kxk2 + 2kyk2 , ∀x, y ∈ X
Proof: We have
kx + yk2 =< x + y, x + y >
=< x, x > + < x, y > + < y, x > + < y, y >
= kxk2 + kyk2 + < x, y > + < y, x >——–(1)
Similarly,
kx − yk2 =< x − y, x − y >
=< x, x > − < x, y > − < y, x > + < y, y >
= kxk2 + kyk2 − < x, y > − < y, x >———-(2)
Adding (1) and (2), we get
kx + yk2 + kx − yk2 = 2kxk2 + 2kyk2 , ∀x, y ∈ X
Schwartz Inequality:
Lemma: If x and y are two vectors in an inner product space X. Then
| < x, y > | ≤ kxkkyk——–(1)
The equality holds if {x, y} is linearly dependent set.
Proof: If y = 0 then there is nothing to prove
< x, y >=< x, 0 >= 0
and so (i) holds.

3
Let y 6= 0. Then any scalar α, we have
0 ≤ kx − αyk2 =< x − αy, x − αy >
=< x, x > −α < x, y > −α < y, x > +αα < y, y >
= kxk2 − α < x, y > −α(< y, x > −αkyk2 )
We see that the expression in the bracket is zero if we choose
<y,x>
α= kyk2
The remaining inequality is
0 ≤ kxk2 − α < x, y >
<y,x><x,y>
0 ≤ kxk2 − kyk2
<x,y><x,y>
0 ≤ kxk2 − kyk2
|<x,y>|2
0 ≤ kxk2 − kyk2
Hence, | < x, y > | ≤ kxk2 kyk2
2

i.e. | < x, y > | ≤ kxkkyk

This proves Schwartz inequality.
Corollary: If X 6= {0} is an inner product space, then
kxk = sup{| < x, y > | : kyk = 1}
Proof: If x = 0, then there is nothing to prove.
Let x 6= 0, then
x x
kxk =< x, kxk >≤ sup{| < x, y > | : kyk = 1} where y = kxk
≤ sup{kxkkyk : kyk = 1}
= kxk.
Hence equality must hold throughout
i.e. kxk = sup{| < x, y > | : kyk = 1}
Example:(1) Euclidean space Rn is an inner product space with inner product defined by
n
P
< x, y >= xi yi ——(1)
i=1
where x = (x1 , x2 , ..., xn )
y = (y1 , y2 , ..., yn ) ∈ Rn
From(1) we have
√ 1 1
n 1
kxk = < x, x > 2 = (x21 + x22 + ... + x2n ) 2 = ( x2i ) 2
P
i=1
From this Euclidean metric is defined as
d(x, y) = kx − yk
1
=< x − y, x − y > 2
1
= [(x − y)2 + .... + (xn − yn )2 ] 2
Note:We have shown that Rn with above metric is a complete metric space and so a Hilbert space.
Unitary space Cn : The inner product on Cn is given by

4
 n
P
< z, w >= zi wi
i=1
1 1
∴ kzk =< z, z > 2 = (z1 z1 + z2 z2 + ... + zn zn ) 2
1
= (|z1 |2 + |z2 |2 + ... + |zn |2 ) n
Cn is also a Hilbert space.
Space l2 : The l2 (l2 , < ·, · >) is a Hilbert space with the inner product given by
∞
P
< x, y >= xi yi ———(1)
i=1
where x = {xi } ∈ l2 and y = {yi } ∈ l2
from (1), we have
1
∞ 1
|xi |2 ) 2
P
kxk =< x, x > 2 = (
i=1
Hence the space l2 is Hilbert space.
Ex.:The space c[a, b] equipped with the norm given by
kxk∞ = sup |x(t)|, ∀x ∈ [a, b]
t∈[a,b]
is not a Hilbert space.
t−a
Proof: Let y(t) = b−a and x(t) = 1, ∀t ∈ [a, b]
Then kxk∞ = 1 and kyk∞ = 1
t−a
Also, x(t) + y(t) = 1 + b−a
t−a
x(t) − y(t) = 1 − b−a
Then kx + yk∞ = sup |x(t) + y(t)| = 2
t∈[a,b]
And kx − yk∞ = sup |x(t) − y(t)| = 1
t∈[a,b]
∴, kx + yk2∞ + kx − yk2∞ = 5
and 2(kxk2∞ + kyk2∞ ) = 4
∴ kx + yk2∞ + kx − yk2∞ 6= 2(kxk2∞ + kyk2∞ )
∴ llgm law does not satisfies.
Ex.: Give an example to show that every normed linear space need not be inner product space.
Proof: The linear space lp , 1 ≤ p < ∞, p 6= 2 equipped with the norm
∞ 1
kxkp = ( |xi |p ) p , where x = {xi } ∈ lp
P
i=1
is not an inner product space and hence not a Hilbert space.
Let x = (−1, −1, 0, 0, ...) ∈ lp
y = (−1, 1, 0, ...) ∈ lp .
1 1
Then kxkp = (| − 1|p + | − 1|p + 0 + 0 + ...) p = 2 p
1
Similarly, kyk = 2 p
∞ 1 1
|xi + yi |p ) p = (2p ) p = 2.
P
Now, kx + ykp =
i=1
∞ 1 1
|xi − yi |p ) p = (2p ) p = 2.
P
Also, kx − ykp =
i=1

5
Then kx + yk2p + kx − yk2p = 8
1 1 1 1 1
and 2(kxk2p + kyk2p ) = 2[(2 p )2 + 2 p )2 )] = 2.4 p + 2.4 p = 4.4 p
∴ k · kp : p 6= 2 does not satisfies llgm law.
i.e. cannot be obtained from the inner product.
Thus lp , p 6= 2 is not an inner product space and hence not Hilbert space.
i.e. Every normed linear space is not an inner product space.
Ex.: Let X be an inner product space. If xn −→ x and yn −→ y in X, then < xn , yn >−→< x, y >
Proof: | < xn , yn > − < x, y > | = | < xn , yn > − < xn , y > + < xn , y > − < x, y > |
≤ | < xn , yn − y > | + | < xn − x, y > |
≤ kxn kkyn − yk + kxn − xkkyk −→ 0(∵ yn −→ y and xn −→ x in X)
Thus | < xn , yn > | −→ | < x, y > |, showing that inner product is a continuous function.
Def.: Let X be a normed linear space. The distance δ from an element x ∈ X to a non-empty subset
M ⊂ X is defined to be δ = inf y∈m kx − yk.
Def.: Let X be a vector space. A subset M of X is said to be convex if for every x, y ∈ M , the
segment joining x and y is contained in M
i.e. if x, y ∈ M, 0 ≤ α ≤ 1, then (1 − α)x + αy ∈ M
Example: (1) Every closed unit sphere is a convex set(every closed set is always convex set).
(2) Every open sphere is always a convex set.
Theorem: Let X be an inner product space and let M 6= 0 be a convex subset which is complete (in
the metric induced by the inner product)
for every x ∈ X, there is a unique y ∈ M such that
δ = inf y∈M kx − yek = kx − yk——–(1)
Proof: By def. of infinium, there is a sequence {yn } of vectors in M such that kx − yn k −→ δ as
n −→ ∞.
Let δn = kx − yn k. Then δn −→ δ as n −→ ∞
We show that {yn } is a Cauchy sequence
Put yn − x = vn . Then we have
kvn k = δn and kvn + vm k = kyn + ym − 2xk = 2k 12 (yn + ym ) − xk———(*)
But M being a convex set, we have
yn , ym ∈ M ⇒ 12 (yn − ym ) ∈ M
∴ k 12 (yn + ym ) − xk ≥ δ
so equ.(*) becomes kvn + vm k ≥ 2δ.
Further, we have
yn − ym = vn − vm
Thus kyn − ym k2 = kvn − vm
2
= 2(kvn k2 + kvm k2 ) − kvn + vm k——(by llgm)

6
 ≤ 2(δn2 + δm
2
) − (2δ)2 −→ 0 as m, n −→ ∞(as δn −→ δ
This shows that {yn } is a Cauchy sequence in M. But M is complete, therefore the above Cauchy
sequence is convergent say yn −→ y ∈ M .
Since y ∈ M ⇒ kx − yk ≥ δ(by def. of δ)———(**)
Also kx − yk ≤ kx − yn k + kyn − yk = δn + kyn − yk −→ δ as n −→ ∞
⇒ kx − yk ≤ δ———(***)
From (**) and (***), we have
kx − yk = δ = inf{kx − yek : ye ∈ M }
Uniqueness:
Let y and y 0 ∈ M both satisfying kx − yk = δ = kx − y 0 k
We want to show y = y 0
Now ky − y 0 k2 = ky 0 − x + x − yk2
= k(y 0 − x) − (y − x)k2
= 2(ky 0 − xk2 + ky − xk2 ) − k(y 0 − x) + (y − x)k2 (by llgm law)
= 2(ky 0 + xk2 + ky − xk2 ) − ky 0 + y − 2xk2 ———(A)
Since y, y 0 ∈ M
y+y 0
⇒ 2 ∈M
⇒ k 12 (y + y 0 ) − xk ≥ d(by def. of d)
0
⇒ k y+y2−2x ≥ d
0
⇒ ky + y − 2xk ≥ d
∴ (A) becomes
ky 0 − yk2 = 2(ky 0 − xk2 + ky − xk2 ) − (ky 0 + y − 2xk2 )
≤ 2(d2 + d2 ) + (2d)2
= 4d2 − 4d2 = 0
⇒ ky 0 − yk2 ≤ 0
i.e.ky 0 − yk2 ≤ 0
⇒ ky 0 − yk = 0
⇒ y0 − y = 0 ⇒ y = y0 .
This proves the uniqueness.
Def.: Let X be an inner product space. An element x ∈ X is said to be orthogonal to an element
y ∈ X if < x, y >= 0
We say x and y are orthogonal and we write x⊥y.
Remark: (1) If x⊥y, then every scalar multiples of x is orthogonal to y.
(2) The zero vector is orthogonal to every vector of an inner product space X.
(3) The zero vector is the only vector which is orthogonal to itself.

7
For, x⊥x
⇒< x, x >= 0 ⇒ kxk2 = 0 ⇒ x = 0.
Example: The space (l2 , < ·, · >) is a Hilbert space with the inner product defined by
∞
xi yei ——-(*)where x = {xi } and y = {yi } ∈ l2
P
< x, y >=
i=1
From (*), we have
1
∞ 1
|xi |2 ) 2
P
kxk = (< x, x >) 2 = (
i=1
∞ 1
|xi − yi |2 ) 2 ———(**)
P
So d(x, y) = kx − yk = (
i=1
This shows that l2 is complete w.r.t metric space given by eq. (**)[Same as in Unit I, the case lp for
p = 2]
Remark: (4) If x⊥y, then y⊥x
< x, y >= 0 ⇒< x, y >= 0
⇒< y, x >= 0 ⇒ y⊥x.
Pythagorean theorem:Let X be an inner product space and let (x, y) ∈ X such that x⊥y. Then
kx + yk2 = kxk2 + kyk2 .
Proof: Since x⊥y ⇒< x, y >= 0
Now kx + yk2 =< x + y, x + y >=< x, x > + < y, x > + < x, y > + < y, y >
= kxk2 + 0 + 0 + kyk2
= kxk2 + kyk2
∞ ∞
kxi k2
P P
Exercise: If x1 , x2 , ..., xn are pairwise orthogonal vector in X, then show that k xi k =
i=1 i=1
Def.: Two non-empty set A and B of an inner product space X are said to be orthogonal if x⊥y∀x ∈ A
and y ∈ B.
Def.: ORTHOGONAL COMPLEMENT: Let A be a non-empty subset of an inner product space
X. Then the orthogonal complement of A is denoted by A⊥ and is defined as A⊥ = {x ∈ X : x⊥A}.
We wrote (A⊥ )⊥ = A⊥⊥ and (A⊥⊥ )⊥ = A⊥⊥⊥ .
Theorem: Let A be a non-empty subset of a Hilbert space H. Then prove that A⊥ is a closed
subspace of H.
Proof:Let x, y ∈ A⊥ and α, β be any two scalars.
Since x, y ∈ A⊥ , < x, z >= 0 and < y, z >= 0, ∀z ∈ A
Now < αx + βy, z >= α < x, z > +β < y, z >, ∀z ∈ A
= α(0) + β(0) = 0
∴< αx + βy, z >= 0, ∀z ∈ A
⇒ αx + βy⊥A
⇒ αx + βy ∈ A⊥
This proves A⊥ is a subspace of H.

8
Next, to show that orthogonal complement is closed, we want to show that A⊥ = A⊥
For, let x ∈ A⊥ , then there is a sequence {xn} in A⊥ such that xn −→ x as n −→ ∞.
Now, xn ∈ A⊥ , ∀n
⇒< xn , y >= 0∀n, ∀y ∈ A
⇒ lim < xn , y >= 0, ∀y ∈ A
n−→∞
⇒< lim xn , y >= 0, ∀y ∈ A(because inner product is continuous function)
n−→∞
⇒< x, y >= 0∀y ∈ A
⇒ x⊥A
⇒ x ∈ A⊥
Since x ∈ A⊥
∴ A⊥ ⊆ A⊥ ———-(*)
By def. of closure we have A⊥ ⊆ A⊥ ———–(**)
∴ on combining (*) and (**), we have
⇒ the orthogonal complement A⊥ is closed subset of a Hilbert space H.
Exercise: Let X be an inner product space and M be a subset of X. Then show that M ∩ M ⊥ ⊆ {0}
Further if M is a subspace then M ∩ M ⊥ = {0}
Proof: let x ∈ M ∩ M ⊥
⇒ x ∈ M and x ∈ M ⊥
⇒ x is orthogonal to every element of M.
In particular, x is orthogonal to x
i.e. < x, x >= 0
⇒ kxk2 = 0
⇒x=0
∴ M ∩ M ⊥ ⊆ {0}
If M is a subspace of X. Then 0 ∈ M . Also M ⊥ is a subspace of X. Therefore, 0 ∈ M ⊥ . Thus
0 ∈ M ∩ M ⊥ = {0}. Hence M ∩ M ⊥ = {0}.
Ex. Prove that kx + yk2 = kxk2 + kyk2
Proof: Since x⊥y ⇒< x, y >= 0
Now kx + yk2 =< x + y, x + y >
=< x, x > + < x, y > + < y, x > + < y, y >
= kxk2 + kyk2 (because x⊥y so < x, y >=< y, x >= 0)
∞
xi k2 =
P
Exercise: Further if x1 , x2 , ..., xn are pairwise orthogonal vector in X, then show that k
i=1
∞
kxi k2 (Prove it by induction)
P
i=1
Lemma: Let X be an inner product space and Y 6= 0 be a complete subspace of X and x ∈ X be
fixed. Then z = (x − y) is orthogonal to Y.

9
Proof: Suppose z is not orthogonal to Y. Then there would be y1 ∈ Y such that < z, y1 >= β 6= 0—
——*
Clearly, y1 6= 0, since otherwise
< z, y1 >=< z, 0 >= 0.
Further for any scalar α
kz − αy1 k2 =< z − αy1 , z − αy1 >
=< z, z > −α < z, y1 > −α[< y1 , z > −α < y1 , y1 >]
2
kz − αy1 k =< z − αy1 , z − αy1 >
=< z, z > −α < y1 , z > −α < z, y1 > +|α|2 < y1 , y1 >
= kzk2 − αβ − αβ + |α|2 ky1 k2
<z,y1 > β
Choose α = <y1 ,y1 > = kyk2
|β|2
∴ kz − αy1 k2 = kzk2 − kyββ1 k
ββ
2 − ky k2 + (ky k2 )2 ky1 k
1 1
2
2 2 2
|β| |β| |β|
= kzk2 − ky1 k2 − ky1 k2 + (ky1 k2 )
2
= kzk2 − ky|β|1 k2 —–(1)
Also since Y is complete subspace of X
∴ Y is convex[because any subspace of vector space is convex]
∴ for x ∈ X, ∃ a unique y ∈ Y such that
kx − yk = δ (by previous theorem)
∴ (1) becomes
|β|2
kz − αy1 k2 = kzk2 − ky1 k2 < δ 2 (∵ kx − yk = δ ⇒ kzk = δ)
But this is not possible
∵ we have
z − αy1 = x − y2 where y2 = y + αy1 ∈ Y
So kz − αy1 k = kx − y2 k ≥ δ(by def. of δ)
∴ our supposition is wrong then z⊥y
i.e.z = z − y⊥y.
Ex.(1) If M1 ⊆ M2 , then M2⊥ ⊂ M1⊥
Sol. Let x ∈ M2⊥ ⇒< x, y >= 0∀y ∈ M2
⇒< x, y >= 0∀y ∈ M1
⇒ x ∈ M1⊥ (∵ M1 ⊂ M2 )
(2) M ⊂ M ⊥⊥
(3) If A and B are subsets of X such that A ⊂ B then A⊥ ⊃ B ⊥
(4)If A 6= ∅ is subset of X. Then M ⊥ = M ⊥⊥⊥
/ M ⊥⊥ ⇒ ∃z ∈ M ⊥ such that < y, z >6= 0.
Let y ∈ M , but y ∈
Since z ∈ M ⊥ ⇒< z, y >= 0

10
Hence y ∈ M ⊥⊥ .
Def.: A vector space X is said to be the direct sum of two subspaces Y and Z of X, written as
X = Y ⊕ Z, if each x ∈ X has a unique representation x = y + z. Then Z is called an algebraic
complement of Y in X and vice versa(i.e. Y is also algebraic complement of Z)
Projection Theorem: Let M be a closed subspace of a Hilbert space H. Then H = M ⊕ M ⊥ .
Proof: If M is a subspace of H. Then we know that M ∩ M ⊥ = {0}
So to show that H = M ∩ M ⊥ . We need only to show that H = M + M ⊥
Also M is given to be closed subspace of H and M ⊥ is a closed subspace of H. Therefore M and M ⊥
is a closed subspace of H. Take Z = M + M ⊥ . Then M ⊂ Z and M ⊥ ⊂ Z
⇒ Z ⊥ ⊂ M ⊥ and Z ⊥ ⊂ M ⊥⊥
⇒ Z ⊥ ⊂ M ⊥ ∩ M ⊥⊥
⇒ Z ⊥ ⊂ {0} (as M ⊥ is a subspace of M ⊥⊥ by last exercise M ∩ M ⊥⊥ = {0})
⇒ (Z ⊥ )⊥ = {0}⊥
⇒ Z ⊥⊥ = H({0}⊥ = H)
Thus Z = H(∵ Z = M + M ⊥ is a closed subspace of H and so we have Z ⊥⊥ = Z)
Therefore Z = M ⊕ M ⊥ = H.
hence H = M ⊕ M ⊥
Ex.: (1) If M is a closed subspace of a Hilbert space H, then there exist a non-zero vector z0 ∈ H
such that z0 ⊥M .
(2) Let H be a Hilbert space if M and N are closed subspace of H such that M ⊥N then M + N are
closed subspace of H.
(3) Let M be a subspace of H. Then show that M is closed iff M = M ⊥⊥
Def.:ORTHONORMAL SET
A non-empty subset A of an inner product space X is said to be orthonormal set if
(i) x⊥y, ∀x, y ∈ A
(ii) < x, x >= 1, ∀x ∈ A.
Remarks:(1) If X = {0}, then it does not contains an orthonormal set.
x
(2) If X 6= {0}, then ∃x 6= 0 in X such that kxk ∈X
x
∴ X has a orthonormal set {y}, where y = kxk
(3) In an inner product space Rn , A = {e1 , e2 , ..., en }, where ei = (0, ..., 1, 0, ..., 0) ∈ Rn (1 at ith
place) is an orthonormal set.[∵ kei k = 1, ∀i and < ei , ej >= 0, ∀i 6= j]
1
(In case of R3 , the set {(1, 0, 0), (0, 1, 0), (0, 0, 1)} is orthonormal(∵ (|1|2 + |0|2 + |0|2 ) 2 = 1)
Ex.: An orthonormal set in an inner product space X is linear independent.
Proof:Let A be orthonormal set in X. Suppose A is finite set say A = {e1 , e2 , ..., en }.
n
αi ei = 0 where αi0 s are scalars.
P
Consider
i=1

11
 n
P n
P
Then 0 =< αi ei , ej >= αi < ei , ej >(by (iii) of I.P.S)
i=1 i=1
= αj < ej , ej >= αj (∵ A is orthonormal set)
⇒ αj = 0, ∀j = 1, 2, ..., n.
This proves that the set A is linear independent, we know that an arbitrary set A(finite or infinite)
is linear independent if for every non-empty subset A is linearly independent, so an orthonormal set
is always linearly independent.
Ex.: Show that if < xi , xj >= 0, ∀i, j, then
n n
xi k2 = kxi k2
P P
k
i=1 i=1
n n n
xi k2 =<
P P P
Sol.: k xi , xj >
i=1 i=1 i=1
n P
P n
= < xi , xj >
i=1 j=1
Pn
= < xi , xi >
i=1
n
kxi k2
P
=
i=1
Ex.: An orthonormal set in an inner product space is linearly independent .
Sol.: Let {e1 , e2 , ..., en } be an orthonormal set in an inner product space X.
Let α1 e1 + α2 e2 + ... + αn en = 0, where αi0 s are scalars. Then multiplication by a fixed ej gives
P∞ n
P
0 =< αi ei , ei >= αi < ei , ej >
i=1 i=1
= αj < ej , ej >= αj .1 = 0
n
P Pn
(∵ αi ei = 0, so < αi ei , ei >=< 0, ej >= 0, ∀j = 1, 2, ..., n)
i=1 i=1
Thus {e1 , e2 , ..., en } is linearly independent.
Theorem: Let {e1 , e2 , ..., en } be a finite set in an inner product space X. Then for any x ∈ X, we
have
∞
| < x, ei > |2 ≤ kxk2 (Bessel’s Inequality)
P
(1)
i=1
n
P
(2) x − < x, ei > ei ⊥ej ∀j = 1, ..., n
i=1
Proof: We have
n
< x, ei > ei k2 ≥ 0
P
kx −
i=1
n
< x, ei > ei k2
P
i.e.0 ≤ kx −
i=1
n
P n
P
=< x − < x, ei > ei , x − < x, ei > ei >
i=1 i=1
n
P n
P n
P
=< x, x > − < x, ei >< ei , x > − < x, ej >< x, ej > + < x, ei >< x, ej >< ei , ej >
i=1 j=1 i=1
n n n
= kxk2 −
P P P
| < x, ei >< x, ej > − < x, ej >< x, ej > + | < x, ei >< x, ei >< ei , ei >
j=1 j=1 i=1
n n n
= kxk2 − | < x, ei > |2 − | < x, ej > |2 + | < x, ei > |2
P P P
i=1 j=1 i=1

12
 n
= kxk2 − | < x, ej > |2
P
j=1
n
⇒ 0 ≤ kxk2 − | < x, ej > |2
P
j=1
n
| < x, ej > |2 ≤ kxk2 ——–(1)
P
⇒
j=1
Remarks: These sums have non-negative terms so that they forms a monotonically increasing se-
quence. This sequence converges because it is bounded by kxk2 . Thus, the sequence of partial sums
of an infinite series converges. Hence this implies the case.
Pn
(2) x − < x, ei > ei ⊥ej , (∀j = 1, 2, ..., n)
i=1
n
P n
P
<x− < x, ei > ei , ej >=< x, ej > − < x, ei >< ei , ej >
i=1 i=1
=< x, ej > − < x, ej >< ej , ej >
=< x, ej > − < x, ej >= 0
n
P
∴x− < x, ei > ei ⊥ej , (∀j = 1, 2, ..., n)
i=1
Def. An orthonormal set S in a Hilbert space H is said to be complete(or maximal) if there doesn’t
exist an orthonormal set T in H such that S (*****) T.
Or
An orthonormal set {ei } in a Hilbert space H is said to be complete(or maximal) if it is impossible
to adjoin a vector e to {ei } such that {ei } (******){e, ei }.
Note: A maximal orthonormal set (or complete) is also called orthonormal basis in H.
Theorem: Every non-zero Hilbert space has an orthonormal basis.
Proof: Let H be non-zero Hilbert space and suppose M be the set of all orthonormal subset of H.
x
Since H 6= {0}, it has an element say x 6= 0 and an orthonormal subsets of H is {y}, where y = kxk
Thus M 6= ∅.
Define a relation 0 ≤0 on M as F ≤ G iff F ⊆ G
Then (M, ≤) is a POSET.
Now, let A = {Aλ : λ ∈ A} be a chain (M, ≤).
S
Let A = Aλ , is an orthonormal set.
λ∈∆
We want to show that A is an upper bound for the chain A in (M, ≤).
S
Let x, y ∈ Aλ = A such that x 6= y
λ∈∆
⇒ x ∈ Aα and y ∈ Aβ for some α, β ∈ ∆.
Since Aα , Aβ ∈ A and A is chain in (M, ≤)
∴ Aα ≤ Aβ or Aβ ≤ Aα
So x ∈ Aα , y ∈ Aβ ⇒ x, y ∈ Aβ [If we suppose Aα ⊆ Aβ ]
But Aβ is an orthonormal set
∴ x⊥y and kxk = 1 and kyk = 1

13
 S
Since x, y ∈ A, ∴ A = Aλ is an orthonormal set
λ∈∆
⇒ A ∈ (M, ≤)
S
Also, Aλ ⊆ Aλ , ∀λ ∈ ∆
λ∈∆
⇒ Aλ ⊆ A, ∀λ ∈ ∆
⇒ Aλ ≤ A, ∀λ ∈ ∆
⇒ A is an upper bound for the chain A in (M, ≤).
This proves that every chain in poset has upper bound in (M, ≤). Hence by Zorn’s Lemma, the above
poset (M, ≤) has the maximal element say F.
We show that F is an orthonormal basis[Complete orthonormal set or Maximal orthonormal set]
For this, suppose on the contrary if ∃ a vector e such that F ∪ {e} is an orthonormal set. Then
F ⊆ F ∪ {e}, which is a contradiction to the fact that F is a maximal element of (M, ≤).
∴ F is an orthonormal basis.
Theorem: Let H be a Hilbert space and {eα }α∈I be an orthonormal set in H. Then the following
conditions are equivalent.
(i) {eα }α∈I is complete
(ii) x⊥{eα }α∈I ⇒ x = 0
P
(iii) If x is a vector in H, then x = < x, eα > eα
α∈I
(iv) kxk2 = | < x, eα > |2
P
α∈I
Proof:(1)⇒ (ii)
Suppose that {eα }α∈I is complete.
Suppose (ii) is not true.
Then ∃ a vector x 6= 0 such that x⊥{eα }α∈I
x
Let e = kxk .
Then e is a unit vector and e⊥{eα }α∈I . Thus the set {e, eα } is an orthonormal set which properly
contain {eα }α∈I .
This contradicts the completeness of {eα }α∈I
∴ our supposition is wrong.
so (ii) is true.
(ii)⇒(iii)
Suppose (ii) holds
P
Let x ∈ H. Then x − < x, eα > eα is orthogonal to {eα }α∈I
α∈IP
By given condition, x − < x, eα > eα = 0, ∀x ∈ H
Pα∈I
⇒x= < x, eα > eα
α∈I
This proves (iii).

14
(iii)⇒(iv)
Suppose (iii) holds
i.e. for x ∈ H, we have
P
x= < x, eα > eα
α∈I
Then, kxk2 =< x, x >
P P
=< < x, eα > eα , < x, eα > eα >
Pα∈I α∈I
= < x, eα >< x, eα >< eα , eα >
α∈I
| < x, eα > |2
P
=
α∈I
kxk2 = | < x, eα > |2
P
α∈I
This proves (iv).
(iv)⇒(i)
Suppose (iv) holds. We have to show that {eα }α is complete.
Suppose on the contrary, the orthonormal set {eα }α∈I is not complete.
Then ∃ a non zero vector e ∈ H such that kek = 1 and e⊥eα , ∀α ∈ I
By (iv) condition we have
kek2 = | < e, eα > |2 = 0
P
α∈I
⇒ kek2 = 0
⇒ e = 0(∵ e⊥eα ∀α ∈ I), which is not true(∵ kek = 1)
Our supposition is wrong. Then {eα }α∈I is complete.
Reisz-Representation Theorem
Every bounded linear functional from a Hilbert space H can be represented in terms of the inner
product f (x) =< x, z >—-(1)
where z depends upon f is uniquely determined by f and has the norm kf k = kzk—–(2)
Proof: Let M = ker f = {x ∈ H : f (x) = 0}. Then M is a closed subspace of H(∵ f is continuous so
f −1 {0} = M is closed in H)
If f = 0 then (1) and (2) holds trivially by taking z = 0.
So let f 6= 0
Then M 6= H(otherwise f = 0)
i.e. M ⊂ H Also M is a closed subspace of H, therefore by projection theorem we have H = M ⊕ M ⊥
where M = ker f .
⇒ ∃ a non zero z0 ∈ M ⊥
Let v = f (x)z0 − f (z0 )x where x ∈ H
⇒ f (v) = f (x)f (z0 ) − f (z0 )f (x) = 0
⇒ v ∈ M.

15
Since z0 ∈ M ⊥ and z0 ⊥M
⇒ 0 =< v, z0 > (∵ z0 ⊥M, v ∈ M )
⇒ 0 =< f (x)z0 − f (z0 )x, z0 >
= f (x) < z0 , z0 > −f (z0 ) < x, z0 >
= f (x)kz0 k2 − f (z0 ) < x, z0 >
f (z0 <x,z0 )>
⇒ f (x) = kz0 k2
f (z0 )z0
=< x, z >, where z = kz0 k2
Next we want to show that z is unique.
For this, let z1 , z2 ∈ H such that
f (x) =< x, z1 >=< x, z2 >
⇒< x, z1 > − < x, z2 >= 0∀x ∈ H
⇒< x, z1 − z2 >= 0, ∀x ∈ H
In particular, if x = z1 − z2 , then < z1 − z2 , z1 − z2 >= 0
⇒ kz1 − z2 k = 0
⇒ z1 = z2
This proves vector z is unique.
Now it only remains to show that kf k = kzk
Taking x = z in (1), we have
f (z) =< z, z >= kzk2
Since f is bounded
kzk2 = |f (z)| ≤ kf kkzk
⇒ kzk2 ≤ kf kkzk
⇒ kzk ≤ kf k—–(3)
Also, |f (x)| = | < x, z > | ≤ kxkkzk(by Schwartz Inequality)
⇒ kf k = sup |f (x)| ≤ kzk—-(4)
06=x∈H,kxk≤1
On combining (3) and (4), kf k = kzk.
Ex.: Let y be a fixed vector in an inner product space X over K. Define
fy : X −→ K by fy (x) =< x, y >, ∀x ∈ X. Prove that fy is linear and bounded.(Do yourself)
Lemma: Let (X, < ·, · >) be an inner product space. If < y, x >=< z, x >, ∀x ∈ X. Then y = z.
Proof: We have
< y, x >=< z, x >, ∀x ∈ X
⇒< y − z, x >= 0, ∀x ∈ X
Taking x = y − z we get
< y − z, y − z >= 0
⇒ ky − zk2 = 0

16
⇒y−z =0
⇒ y = z, hence the proof.
Cor.: If < z, x >= 0, ∀x ∈ X. Then z = 0.
Remark: If H is a Hilbert space then H ∗ = H
Theorem: Let H1 and H2 be two Hilbert spaces and T : H1 −→ H2 be a bounded linear operator.
Then there exists a unique bounded linear operator
T ∗ : H2 −→ H1 such that < T x, y >=< x, T ∗ y >————-(1)
∀x ∈ H1 and y ∈ H2 .
∗
Moreover kT k ≤ kT k
Proof: Let y ∈ H2 be fixed. Define
fy : H1 −→ K by
fy (x) =< T x, y >, ∀x ∈ H1
Then fy is linear. Also by Schwartz’s inequality
|fy (x)| = | < T x, y > | ≤ kT xkkyk
≤ kT kkxkkyk, ∀x ∈ H
⇒ fy is a bounded linear functional and kfy k ≤ kT kkyk————(2)
(kfy k = sup |fy (x)| ≤ kT kkyk)
x∈H1 ,kxk≤1
By Reisz Representation theorem, there exist a unique vector y ∗ ∈ H1 such that
< T x, y >= fy (x) =< x, y ∗ >, ∀x ∈ H1
and kfy k = ky ∗ k——–(3)
This gives a mapping
T ∗ : H2 −→ H1
defined by T ∗ y = y ∗
Thus < T x, y >=< x, y ∗ >=< x, T ∗ y >, ∀x ∈ H1
This shows the existence of a mapping T ∗ : H2 −→ H1 satisfying eq. (1). Now we show that T ∗ is
bounded linear transformation.
Let y, z ∈ H2 and α, β ∈ K.
Then, < x, T ∗ (αy + βz) >=< T x, αy + βz >
=< T x, αy > + < T x, βz >
=< x, αT ∗ y > + < x, βT ∗ z >
=< x, αT ∗ (y) + βT ∗ (z) >, ∀x ∈ H1
⇒ T ∗ (αy + βz) = αT ∗ y + βT ∗ z [by lemma (A)(**)]
Thus T ∗ is a linear operator. Next we want to show that T ∗ is bounded
from (2) and (3), we have
kT kkyk ≥ kfy k = ky ∗ k = kT ∗ yk

17
⇒ kT ∗ yk ≤ kT kkyk, ∀y ∈ H2
⇒ T ∗ is bounded and kT ∗ = sup kT ∗ yk ≤ sup kT kkyk ≤ kT k
y∈H2 ,kyk≤1
kT ∗ k ≤ kT k
Uniqueness of T ∗
Suppose S : H2 −→ H1 is a mapping such that
< T x, y >=< x, Sy >= 0, ∀x ∈ H1 , y ∈ H2
⇒< x, T ∗ y >=< x, Sy >
⇒< x, T ∗ y − Sy >= 0, ∀x ∈ H1 , y ∈ H2
⇒ T ∗ y = Sy, ∀y ∈ H2 (by the cor. to lemma **)
⇒ S = T∗
Def.: Let H1 and H2 be two Hilbert space and T : H1 −→ H2 be a bounded linear operator. The
operator T ∗ : H2 −→ H1 satisfying < T x, y >=< x, T ∗ y >, ∀x ∈ H1 , y ∈ H2 is called the adjoint
operator of T.
Ex. Let O and I be the zero and identity operator on H. Then O∗ = O(∵< x, O∗ y >=< Ox, y >=<
O, y >)
Lemma:Let X and Y be two inner product spaces and Q : X −→ Y be a bounded linear operator.
Then Q = 0 ⇔< Qx, y >= 0, ∀x ∈ X and y ∈ Y .
Proof: Suppose Q = 0 ⇒ Q(x) = 0, ∀x ∈ X
< Qx, y >= 0∀x ∈ X, y ∈ Y
Conversely, suppose < Qx, y >= 0
by cor. to lemma we get
Qx = 0, ∀x ∈ X
⇒Q=0
Theorem: Let H1 and H2 be two Hilbert spaces and T, S ∈ B(H1 , H2 ). Suppose α ∈ K. Then
(1) < T ∗ y, x >< y, T x >.
Proof: < T ∗ y, x >=< x, T ∗ y >(by property(ii) of Inner Product)
=< T x, y >(by def. of adjoint)
=< y, T x >
∗
∴< T y, x >=< y, T x >.
Def.: Let H1 and H2 be any two Hilbert spaces and suppose T : H1 −→ H2 be a bounded linear
transformation T ∗ : H2 −→ H1 such that < T (x), y >=< x, T ∗ (y) >, ∀x ∈ H1 and y ∈ H2 . This is
called adjoint of the bounded linear(operator) Transformation T.
Remark: If H1 = H2 = H. Then for a bounded linear transformation T : H −→ H the adjoint of
T i.e. T ∗ is called Hilbert adjoint operator.
Ex.: Let H1 and H2 be two Hilbert spaces and T, S ∈ B(H1 , H2 ). Suppose α ∈ K. Then

18
(1) < T ∗ y, x >=< y, T x >, ∀x, y ∈ H
(2) (S + T )∗ = S ∗ + T ∗
(3) (αT )∗ = αT ∗
(4) (T S)∗ = S ∗ T ∗
(5) (T ∗ )∗ = T
(6) kT ∗ k = kT k
(7) T ∗ T = 0 ⇔ T = 0
(8) kT xk2 = kT k2 = kT T ∗ k
Proof: (1) We have from def. of T ∗
< T ∗ y, x >=< x, T ∗ y >=< T x, y >=< y, T x >
(2) (S + T )∗ = S ∗ + T ∗
For x, y ∈ H,
< x, (S ∗ + T ∗ )y >=< x, S ∗ y + T ∗ y >
=< x, S ∗ y > + < x, T ∗ y >
=< Sx, y > + < T x, y >
=< (S + T )x, y >
=< x, (S + T )∗ y >
⇒ (S + T )∗ = S ∗ + T ∗
(3) < x, αT ∗ y >= α < x, T ∗ y >
= α < T x, y >
=< αT x, y >
=< x, (αT )∗ y >
∴ (αT )∗ = αT ∗
(4) (T S)∗ = S ∗ T ∗
< x, S ∗ T ∗ y >=< x, S ∗ (T ∗ y) >
=< Sx, T ∗ y >
=< T (Sx), y >
(Rest do yourself)
Def.: A bounded linear operator T : H −→ H in the Hilbert space H is said to be self adjoint if
T ∗ = T (or Hermitian)
Example: The zero operator and identity operator are always self adjoint.
Theorem: Let H be a Hilbert space.Then
(1) If T is a self adjoint, then < T x, x > is real, ∀x ∈ H
(2) If H is a complex Hilbert space and < T x, x > is real ∀x ∈ H. Then T is self adjoint.
Proof:(1) Suppose T is self adjoint operator on Hilbert space H. We have to show that < T x, x > is

19
real, ∀x ∈ H.
Let x ∈ H. Now < T x, x >=< x, T ∗ x >=< x, T x > (∵ T = T ∗ )
=< T x, x >
⇒< T x, x > is real, ∀x ∈ H
(2) Suppose < T x, x > is real, ∀x ∈ H. We have to show that T is self adjoint.
Since < T x, x > is real, ∀x ∈ H
∴< T x, x >=< T x, x >=< x, T ∗ x >=< T ∗ x, x >
Hence T is self adjoint.
Polarisation Identity:Let X be an inner product space then
4 < x, y >= kx + yk2 − kx − yk2 + ιkx + ιyk2 − ιkx − ιyk2 , ∀x, y ∈ X.

20