Banach Spaces Page 1

Lecture Notes on

Functional Analysis-II

by Dr. H. S. Mehta

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 2

Definition:
A normed linear space X is called a Banach space, if X is complete in the metric
d(x, y) = ||x − y|| induced by || · ||.
Remarks:

1. A subspace of a Banach space is a Banach space if and only if it is closed.

2. If F : X → Y is a homeomorphism from X onto Y and X is a Banach space, then

Y is also a Banach space.(Cor 6.3(a)).

Examples:

1. X = Kn with || · ||p (1 ≤ p ≤ ∞) is a Banach space.

2. X = C[a, b] with || · ||∞ is a Banach space.

3. X = C 1 [a, b] with ||x||1 = ||x||∞ + ||x0 ||∞ is a Banach space.

4. X = `p with || · ||p , (1 ≤ p ≤ ∞) is a Banach space.

Examples of normed linear spaces which are not Banach spaces.

1. X = P [a, b] with || · ||∞ .

Rb
2. X = C[a, b] with ||x|| = |x(t)|dt.
a

3. X = C 1 [a, b] with ||x|| = ||x||∞ .

4. X = c00 with ||x|| = ||x||∞ .

Definition:
∞
P
1. Let X be a normed linear space. A series xi is said to be summable or convergent
i=1
n
P
if the sequence sn = xi of partial sum is convergent in X. If {sn } converges to
i=1
∞
P
s, then we write xi = s.
i=1

∞
P
2. A series xi is said to be absolutely summable or absolutely convergent, if the
i=1
∞
P
series ||xi || is convergent.
i=1

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 3

Note: In K, we know that every absolutely summable series is summable.

Theorem 8.1:
A normed linear space is a Banach space if and only if every absolutely summable series
is summable in X.
∞
P
Proof. First, assume that X is a Banach space. Let xi be absolutely summable series
i=1
∞
P ∞
P n
P
in X, i.e., ||xi || < ∞. To show that xi is convergent, let sn = xi be the partial
i=1 i=1 i=1
∞
P
sum of xi . Then,
i=1
n+j n+j n
||xi || = s0n+j − s0n , where s0n =
P P P
||sn+j − sn || = || xi || ≤ ||xi ||
i=n+1 i=n+1 i=1
∞
||xi ||. So, s0n+j − s0n → 0. Therefore ||sn+j − sn || → 0, ∀n.
P
which is partial sum of
i=1
Thus {sn } is a Cauchy sequence in X, which is complete. So, {sn } converges in X, i.e.,
∞
P
xi is convergent in X.
i=1
Conversely, assume that every absolutely summable series in X is summable in X.
Let {sn } be a Cauchy sequence in X. Let m1 ∈ N such that ||sm − sm1 || < 1, ∀m ≥ m1 .
1
Inductively, define m2 , m3 , . . . with mn+1 ≥ mn , such that ||sm − smn || < n2
, ∀m ≥ mn .
Define xn = smn+1 − smn . Then, ||xn || = ||smn+1 − smn || < n12 . Therefore,
∞ ∞ ∞
1
P P P
||xn || ≤ n2
< ∞. Thus, the series xn converges absolutely in X.
n=1 n=1 n=1
∞
P n
P
Hence xn < ∞. Thus, if Sn = xi is a sequence of partial sum, then {Sn−1 } is
n=1 i=1
convergent in X. Now,

smn = sm1 + (sm2 − sm1 ) + (sm3 − sm2 ) + . . . + (smn − smn−1 )

n−1
X
= sm1 + x1 + x2 + . . . + xn−1 = sm1 + xi
i=1

= sm1 + (Sn−1 )
But {Sn−1 } is convergent in X. So, {smn } is convergent in X. Thus a Cauchy sequence
{sn } has a convergent subsequence {smn } in X. Hence, the sequence {sn } itself is con-
vergent in X. Therefore, X is complete or Banach space.

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 4

Theorem 8.2(a):
Let X be a normed linear space and Y be a closed subspace of X. Then X is a Banach
space if and only if Y and X/Y are Banach spaces in respective norms.

Proof. First assume that X is a Banach space. As Y is closed in X, Y is also a Banach

P∞
space. To show that X/Y is a Banach space, let (xn + Y ) be absolutely convergent
n=1
∞
P m
P
series in X/Y , i.e., |||xn + Y ||| < ∞. Let sm = (xn + Y ) be partial sum of the series
n=1 n=1
∞
P
(xn + Y ) in X/Y . Now by definition of ||| · |||, for each n ∈ N, there exists yn ∈ Y ,
n=1
1
such that ||xn + yn || < |||xn + Y ||| + n2
. Then

∞ ∞ ∞
1 X X X 1
||xn + yn || < |||xn + Y ||| + 2 =⇒ ||xn + yn || < |||xn + Y ||| + <∞
n n=1 n=1 n=1
n2
∞
X
=⇒ (xn + yn ) is absolutely summable in X
n=1
X∞
=⇒ (xn + yn ) < ∞, as X is complete.
n=1
∞
P m
P
Let (xn + yn ) = s ∈ X and Sm = (xn + yn ) be the partial sum of the series
n=1 n=1
∞
P
(xn + yn ). Then,
n=1
m
X
|||sm − (s + Y )||| = ||| (xn + Y ) − (s + Y )|||
n=1
m
X
= ||| (xn + yn + Y ) − (s + Y )|||
n=1
m
X
≤ || (xn + yn ) − s|| = ||Sm − s|| → 0 as m → ∞
n=1
∞
P
Therefore sm → s + Y in X/Y . Thus (xn + Y ) < ∞. Therefore X/Y is a Banach
n=1
space.
Conversely, assume that Y and X/Y both are Banach spaces. Let {xn } be a Cauchy
sequence in X. Then {xn + Y } will be Cauchy in X/Y , as
|||(xn + Y ) − (xm + Y )||| ≤ ||xn − xm ||. But X/Y is complete. So, {xn + Y } converges in
X/Y . Suppose xn + Y → x + Y in X/Y . Then, by Theorem 5.2(b), there is a sequence
{yn } in Y such that xn + yn → x in X. Now, yn − ym = (xn + yn ) − (xm + ym ) − (xn − xm ).

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 5

Therefore, ||yn − ym || ≤ ||(xn + yn ) − (xm + ym )|| + ||xn − xm || → 0, as {xn + yn } & {xn }

both are Cauchy in X. Therefore {yn } is a Cauchy sequence in Y . But Y is complete.
So, yn → y in Y . Now, xn + yn → x and yn → y. Thus xn = (xn + yn ) − yn → x − y in
X. Thus {xn } is convergent in X. Therefore, X is a Banach space.

Theorem 8.2(b):
Let X1 .X2 , . . . , Xn be Banach spaces and X = X1 × X2 × . . . × Xn . Then X1 , X2 , . . . , Xn
are Banach spaces if and only if (X, || · ||p ) is a Banach space, 1 ≤ p ≤ ∞.

Proof. Exercise

Theorem 8.2(c):
Let X and Y be normed linear spaces and X 6= {0}. Then BL(X, Y ) is a Banach space
if and only if Y is a Banach space.
In particular, for every normed linear space X, X 0 is a Banach space.

Proof. Assume that Y is a Banach space. Let {Fn } be a Cauchy sequence in BL(X, Y ).
Then, for given  > 0, there is n0 ∈ N such that ||Fn − Fm || < , ∀n, m ≥ n0 . But
||Fn (x)|| ≤ ||Fn || ||x||, ∀n. So for x ∈ X,

||Fn (x) − Fm (x)|| = ||(Fn − Fm )(x)|| ≤ ||Fn − Fm || ||x|| < ||x||, ∀n, m ≥ n0

Therefore, for every x ∈ X, {Fn (x)} is a Cauchy sequence in Y . Since Y is complete,

{Fn (x)} is convergent in Y , for every x ∈ X, i.e., lim Fn (x) ∈ Y . Define F : X → Y by
n→∞

F (x) = lim Fn (x). Then, it can be checked that F is linear.

n→∞

Now, for x ∈ X, ||Fn (x) − Fm (x)|| < ||x||, ∀n, m ≥ n0 .

Taking limit m → ∞, ||Fn (x) − F (x)|| < ||x||, ∀n ≥ n0 , ∀x ∈ X.
In particular, ||(Fn0 − F )(x)|| < ||x||. Therefore, Fn0 − F ∈ BL(X, Y ). Since
Fn0 ∈ BL(X, Y ), F ∈ BL(X, Y ). Also, from (1), ||Fn (x) − F (x)|| < , ∀n ≥ n0
& ∀x ∈ X with ||x|| = 1. Therefore

sup{||(Fn − F )(x)|| : x ∈ X, ||x|| ≤ 1} < , ∀n ≥ n0 =⇒ ||Fn − F || < 

Therefore Fn → F in BL(X, Y ).

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 6

Conversely, assume that BL(X, Y ) is a Banach space. To show that Y is complete,

let {yn } be a Cauchy sequence in Y . Since X 6= {0}, let a ∈ X with a 6= 0. Then, there is
f ∈ X 0 such that f (a) = ||a|| =
6 0 and ||f || = 1. Define Fn : X → Y by Fn (x) = f (x)yn .
Then, clearly each Fn is linear(as f is linear) and

||Fn (x)|| = ||f (x)yn || = |f (x)| ||yn || ≤ ||f || ||x|| ||yn || = αn ||x|| with αn = ||yn ||.

Therefore, Fn ∈ BL(X, Y ). Also, if x ∈ X and ||x|| ≤ 1, then

||Fn (x) − Fm (x)|| ≤ ||yn − ym || ||x|| ≤ ||yn − ym ||.

Taking supremum over all such x, ||Fn − Fm || ≤ ||yn − ym ||.

Therefore, {Fn } will be a Cauchy sequence in BL(X, Y ). Therefore Fn → F in BL(X, Y ).
Thus Fn (x) → F (x) in Y , i.e., f (x)yn → F (x), ∀x ∈ X. In particular, f (a)yn → F (a) in
F (a)
Y , i.e., yn → f (a)
in Y . Therefore, Y is complete.

Theorem 8.3:
Let X be a normed linear space and Y be a Banach space.

(a) Let X0 be a dense subspace of X and F0 ∈ BL(X0 , Y ). Then there exists unique
F ∈ BL(X, Y ) such that F|X0 = F0 and ||F || = ||F0 ||.

(b) Let {Fn } be a sequence in BL(X, Y ) such that ||Fn || ≤ α, ∀n and for some α > 0.
Let E ⊂ X such that spanE = X. Suppose, {Fn (x)} converges in Y for every
x ∈ E. Then there is a unique F ∈ BL(X, Y ) such that Fn (x) → F (x) for every
x ∈ X.

Proof. (a) Let x ∈ X. Then there is a sequence {xn } in X0 such that xn → x in X.

Now F0 ∈ BL(X0 , Y ). So, ||F0 (xn ) − F0 (xm )|| ≤ ||F0 || ||xn − xm ||. So, {F0 (xn )} is
a Cauchy sequence in Y , which is complete. So, lim F0 (xn ) exists in Y . Define,
n→∞

F : X → Y by F (x) = lim F0 (xn ). Then F is well defined. For, suppose xn → x

n→∞

and yn → x in X. Then xn − yn → 0 in X0 . Also, F0 is continuous on X0 .

Therefore, F0 (xn − yn ) → F0 (0) = 0 in Y . Therefore lim F0 (xn ) = lim F0 (yn ).
n→∞ n→∞

It can be checked that F is linear. Also,

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 7

||F (x)|| = || lim F0 (xn )|| = lim ||F0 (xn )|| ≤ lim ||F0 || ||xn || = ||F0 || ||x||.
n→∞ n→∞ n→∞

Therefore, F ∈ BL(X, Y ) and ||F || ≤ ||F0 ||. Also, if x ∈ X0 , then we can take
xn = x and so F (x) = lim F0 (x) = F0 (x). Therefore F|X0 = F0 .
n→∞

Further, {x ∈ X0 : ||x|| ≤ 1} ⊂ {x ∈ X : ||x|| ≤ 1}. Therefore

sup{||F (x)|| : x ∈ X0 , ||x|| ≤ 1} ≤ sup{||F (x)|| : x ∈ X ||x|| ≤ 1}, i.e.,
sup{||F0 (x)|| : x ∈ X0 , ||x|| ≤ 1} ≤ sup{||F (x)|| : x ∈ X ||x|| ≤ 1}.
Thus ||F0 || ≤ ||F ||. Therefore, ||F0 || = ||F ||.

Finally, F is unique. For, suppose G ∈ BL(X, Y ) such that G|X0 = F0 . Then, for
x ∈ X, F (x) = lim F0 (xn ) = lim G(xn ) = G( lim xn ) = G(x).
n→∞ n→∞ n→∞

In fact, as X0 is dense in X, the continuous extension is unique.

(b) Let X0 = spanE. Then X0 is a dense subspace of X. Also, as Fn0 s are linear,
{Fn (x)} converges for every x ∈ X0 . Define F0 : X0 → Y by F0 (x) = lim Fn (x).
n→∞

Then clearly F0 is linear and

||F0 (x)|| = || lim Fn (x)|| = lim ||Fn (x)|| ≤ lim ||Fn || ||x|| ≤ α||x||.
n→∞ n→∞ n→∞

So, Fn ∈ BL(X0 , Y ). By (a), there is unique F ∈ BL(X, Y ) such that F|X0 = F0 .

Claim: Fn (x) → F (x) for every x ∈ X.
Let x ∈ X and  > 0. Since X0 = X, there is x0 ∈ X0 such that ||x − x0 || < . Now,
lim Fn (x0 ) = F0 (x0 ). So, there is n0 ∈ N such that ||Fn (x0 )−F0 (x0 )|| < , ∀n ≥ n0 .
n→∞

Thus, ||Fn (x0 ) − F (x0 )|| < , ∀n ≥ n0 , as x0 ∈ X0 , F0 (x0 ) = F (x0 ). Therefore,

for n ≥ n0

||Fn (x) − F (x)|| ≤ ||Fn (x) − Fn (x0 )|| + ||Fn (x0 ) − F (x0 )|| + ||F (x0 ) − F (x)||

≤ ||Fn || ||x − x0 || + ||Fn (x0 ) − F (x0 )|| + ||F || ||x0 − x||

= (||Fn || + ||F ||)||x − x0 || + 

≤ (α + ||F ||) +  = (α + ||F || + 1)

Thus Fn (x) → F (x) in Y .

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 8

Baire Category Theorem:

A complete metric space is of second category.

OR

A complete metric space can not be expressed as a countable union of nowhere dense sets.

OR

In a complete metric space, countable intersection of open dense sets is dense.

Theorem 8.4:
A Banach space can not have a denumerable (countable) basis.

Proof. Let X be a Banach space. If possible, suppose {x1 , x2 , . . . , xn , . . .} is a basis for

X. For each n ∈ N, define Yn = span{x1 , x2 , . . . , xn }. Then, each Yn is finite dimensional
/ Yn . So, Yn◦ = ∅ for every
(dimYn = n) and so closed in X. Also. Yn 6= X, as xn+1 ∈
n. Take Dn = X \ Yn = Ync . Then, clearly each Dn is open and dense in X (For any
∞
E ⊂ X, E ◦ = ∅ ⇐⇒ (X \ E) = X). By Baire Category Theorem,
T
Dn is dense in X.
n=1
But
∞ ∞
 ∞
 ∞

c c
Ync
S S T T
Yn = X =⇒ Yn = X = ∅ =⇒ = ∅ =⇒ Dn = ∅
n=1 n=1 n=1 n=1

which is a contradiction. Therefore, X can not have a countable basis.

Definition: Schauder Basis

Let X be a normed linear space. A countable subset {x1 , x2 , . . .} of X is called a Schauder
basis for X if ||xn || = 1 for each n and if for every x ∈ X, there are unique k1 , k2 , . . . in
P
K such that x = n kn xn .
Remarks:

1. If X is finite dimensional and {x1 .x2 , . . . , xn } is a (Hamel) basis for X, then

{ ||xx11 || , ||xx22 || , . . . , ||xxnn || } is a Schauder basis for X.

In particular, {e1 , e2 , . . . , en } is a Schauder basis for Kn .

2. {e1 , e2 , . . . , en , . . .} is a Schauder basis for `p (1 ≤ p < ∞).

Functional Analysis-II Dr. H. S. Mehta

Banach Spaces Page 9

3. If a normed linear space X has a Schauder basis {x1 , x2 , . . .}, then X is separable.
For, {α1 x1 + α2 x2 + . . . + αm xm : αi ∈ K with m ≥ 1, Re αi , Im αi ∈ Q} is dense in
X.

4. Every separable Hilbert space has a Schauder basis.

5. Every separable Banach space may not have a Schauder basis. Example by Enflo
in 1973.

Functional Analysis-II Dr. H. S. Mehta