Functional Analysis I

Unit III
Theorem. 9.1 : Uniform Boundedness Principle
Let X be a Banach space, Y be a normed space and F be a subset of BL(X, Y ) such that
for each x ∈ X, the set {F (x) : F ∈ F} is bounded in Y . Then for each bounded subset
E of X, the set {F (x) : x ∈ E, F ∈ F} is bounded in Y , that is F is uniformly bounded
on E. In particular, sup{kF k : F ∈ F} < ∞.
Proof. Let X be a Banach space, Y be a normed space and F be a subset of BL(X, Y )
such that for each x ∈ X, the set {F (x) : F ∈ F} is bounded in Y .
For n = 1, 2, . . ., let Dn = {x ∈ X, kF (x)k > n for some F ∈ F}.
Since, the functions k k and F are continuous, the function x → kF (x)k is continuous on
X for every F ∈ F.
Hence the set {x ∈ X, kF (x)k > n} is open in X.
Therefore, Dn is the union of all these sets and hence it is open in X.
Let x ∈ X.
By hypothesis, kF (x)k ≤ n for all F ∈ F and for some positive integer n.
That is x ∈/ Dn for some n.
This implies that ∩∞ n=1 Dn = φ.
Therefore, ∩∞ n=1 n cannot be dence in X.
D
Since X is Banach, by Baire’s theorem, some Dm must not be dense in X.
Then, there exist some a ∈ X and r > 0 such that UX (a, r) ∩ Dm = φ.
That is, if y ∈ X, ka − yk ≤ r implies that kF (y)k ≤ m for all F ∈ F.
Now, let E be a bounded subset of X.
Then, there exist some α > 0 such that kxk ≤ α for all x ∈ E.
Let x ∈ E.
Then
 rx  rx
a− +a =
α α
r
= kxk
α
r
≤ α
α
≤ r

This implies that rx

α
− a ∈ UX (a, r)
rx
=⇒ kF ( α − a)k ≤ m.

1
Then, consider
α  rx 
kF (x)k = F
r α
α  rx 
= F +a−a
r α
α  rx 
= F + a − F (a)
r α
α   rx  
≤ F + a + kF (a)k
r α
α
≤ (m + m)
r
2αm
≤
r
Thus, sup{kF (x)k, x ∈ E, F ∈ F} ≤ 2αmr
This implies that F is uniformly bounded in E.

Now, if we let E = UX (0, 1), it follows that

sup{kF k : F ∈ F} = sup {kF (x)k : x ∈ X, kxk ≤ 1, F ∈ F}

2m
≤
r
< ∞

This completes the proof of the theorem.

2
Definition. A subset E of a nls X is said to be totally bounded if, for every  > 0,
there exist some x1 , x2 , . . . , xm ∈ E such that
E ⊂ U (x1 , ) ∪ U (x2 , ) ∪ · · · ∪ U (xm , ).
Corollary. 9.2 : Let X be a Banach space, Y be a normed space and (Fn ) be a sequence
in BL(X, Y ) such that the sequence (Fn (x)) converges in Y for every x ∈ X. For x ∈ X,
define
F (x) = lim Fn (x).
n→∞

(a) Banach Steinhaus Theorem

F is a bounded linear map X to Y and
kF k ≤ lim inf kFn k ≤ sup{kFn k : n = 1, 2, . . .} < ∞.
n→∞

(b) Let E be a totally bounded subset of X. Then (Fn (x)) converges to F (x) uniformly
for x ∈ E.
Proof. Let X be a Banach space, Y be a normed space and (Fn ) be a sequence in
BL(X, Y ) such that the sequence (Fn (x)) converges in Y for every x ∈ X. For x ∈ X,
define
F (x) = lim Fn (x).
n→∞

(a) Since (Fn ) is linear for each n, F is also linear.

Since (Fn (x)) converges in Y for each x ∈ X, the set {kFn (x)k, n = 1, 2, . . .} is
bounded for every x ∈ X.
Now, let F = {Fn , n = 1, 2, . . .}.
Hence by Uniform Boundedness Principle, the norms of Fn , {kFn k, n = 1, 2, . . .} is
bounded.
That is kFn k < ∞.
Since F (x) = limn→∞ Fn (x), it is clear that
kF (x)k = k lim Fn (x)k
n→∞
= lim kFn (x)k
n→∞
= lim inf kFn (x)k
n→∞
≤ lim inf kFn kkxk
n→∞

This implies that F is continuous and

kF k = sup{kF (x)k, x ∈ X, kxk ≤ 1}
≤ sup{lim inf kFn kkxk, x ∈ X, kxk ≤ 1}
n→∞
≤ lim inf kFn k
n→∞
≤ sup{kFn k : n = 1, 2, . . .}
< ∞.

3
 (b) Let E be a bounded subset of X.
Let sup{Fn k, n = 1, 2, . . .} = α. (by part (a))
Then kF k ≤ α
Let  > 0.
Since E is totally bounded, there exist some x1 , x2 , . . . , xm ∈ E such that

E ⊂ U (x1 , ) ∪ U (x2 , ) ∪ · · · ∪ U (xm , )

Now, Fn (xj ) → F (xj ) as n → ∞ for j = 1, 2, . . . , m.

Take n0 ∈ Z+ such that kFn (xj ) − F (xj )k ≤  for all n ≥ n0 and all j = 1, 2, . . . m.
Let x ∈ E.
Choose xj ∈ E such that kx − xj k < .
Then for all n ≥ n0 , we have

kFn (x) − F (x)k ≤ kFn (x) − Fn (xj ) + Fn (xj ) − F (xj ) + F (xj ) − F (x)k
≤ k (Fn (x) − Fn (xj )) + (Fn (xj ) − F (xj )) + (F (xj ) − F (x))k
≤ kFn (x) − Fn (xj )k + kFn (xj ) − F (xj )k + kF (xj ) − F (x)k
≤ kFn (x − xj )k + kFn (xj ) − F (xj )k + kF (xj − x)k
≤ kFn k kx − xj k + kFn (xj ) − F (xj )k + kF k kxj − xk
≤ α +  + α
≤ (2α + 1)

Hence Fn (x) → F (x) uniformaly for x ∈ E.

This completes the proof of the corollary.

4
Theorem. 9.3 :
(a) Resonance Theorem
Let X be a normed space and E be a subset of X. Then E is a bounded in X if
and only if f (E) is bounded in K for every f ∈ X 0
(b) Let X and Y be normed spaces and F : X −→ Y be linear. Then F is continuous
if and only if g ◦ F is continuous for every g ∈ Y 0 .
Proof. (a) Let X be a normed space and E be a subset of X.

First assume that E is bounded in X.

Therefore, there exist some α > 0 such that kxk ≤ α for all x ∈ E.
Let f ∈ X 0 .
Then |f (x)| ≤ kf k kxk ≤ kf k α for all x ∈ E.
Therefore, f (E) is bounded in K.

Conversely, assume that f (E) is bounded in K for all f ∈ X 0 .

00
Let J : X → X be the canonical embedding map.
Then J(x) = jx , J(x)(f ) = jx (f ) = f (x) for all x ∈ X, f ∈ X 0 .
00
Then {J(x), x ∈ E} ⊆ BL(X 0 , K) = X , a Banach Space.
For each f ∈ X 0 , the set {|J(x)(f )|, x ∈ E} = {|f (x)|, x ∈ E} is bounded as E is
bounded.
Therefore by Uniform boundedness principle, {kJ(x)k, x ∈ E} is bounded.
But we know that kJ(x)k = kxk for all x ∈ X.
This implies that {kxk, x ∈ E} is bounded.
That is E is bounded.
(b) Let X and Y be normed spaces and F : X −→ Y be linear.
First assume that F is continuous.
Let g ∈ Y 0 .
Then clearly, g is continuous.
Since the composition of two continuous function is again continuous, we get g ◦ F
is continuous.

Conversely, assume that g ◦ F is continuous for every g ∈ Y 0 .

Let E = {F (x), x ∈ X, kxk ≤ 1}.
Then clearly, E ⊂ Y .

Then g(E) = g ◦ F UX (0, 1) .
Since g ◦ F is continuous, g(E) is bounded (by part (a)).
Now g(E) is bounded

implies E is a bounded subset of Y . (again by part (a))
=⇒ F UX (0, 1) is bounded.
That is, the linear map F is bounded on UX (0, 1).
Then, by theorem 6.2, F is continuous.

5
6
Definition. Let X and Y be metric spaces and F be a map from X to Y . Then F is
continuous if xn → x in X implies F (xn ) → F (x) in Y .

F is said to be closed if xn → x in X and F (xn ) → y in Y implies F (x) = y.

F is said to be open if for every open set E in X, its image F (E) is open in Y .

If F : X → Y , then graph of F is defined as Gr(F ) = {(x, F (x)) ∈ X × Y, x ∈ X}.

Notes :

1. A continuous map is always closed. But the converse may not be true. (Give an
example)

2. F is continuous if and only if for every open set E in Y , the set F −1 (E) is open in
X.

3. A map is closed if and only if Gr(F ) is closed.

4. If F is closed and bijective, then F −1 is also closed. (Verify)

5. F is bijective and linear implies F −1 is linear.

7
Lemma. 10.1 : Let X be a linear space over K. Consider subsets U and V of X, and
k ∈ K such that U ⊂ V + kU . Then for every x ∈ U , there is a sequence (vn ) in V such
that
x − (v1 + kv2 + · · · + k n−1 vn ) ∈ k n U, n = 1, 2, . . .

Proof. We prove this lemma by Mathematical induction.

Let X be a linear space over K. Consider subsets U and V of X, and k ∈ K such

that U ⊂ V + kU .
Let x ∈ U .
Then, since U ⊂ V + kU . there is some v1 ∈ V such that x − v1 ∈ kU .
Hence the result is true for n = 1.

Now assume that the result is true for n = m.

That is there are elements v1 , v2 , . . . vm in V such that

x − (v1 + kv2 + · · · + k m−1 vm ) ∈ k m U..............(1)

Now we prove the result for n = m + 1.

From (1), we get x − (v1 + kv2 + · · · + k m−1 vm ) = k m u0 ...........(2) for some u0 ∈ U .
Then u0 = v 0 + ku0 for some v 0 ∈ V and u0 ∈ U .
Therefore, (2) implies

x − (v1 + kv2 + · · · + k m−1 vm ) = k m (v 0 + ku0 )

=⇒ x − (v1 + kv2 + · · · + k m−1 vm ) = k m v 0 + k m+1 u0
=⇒ x − (v1 + kv2 + · · · + k m−1 vm ) − k m v 0 = k m+1 u0
=⇒ x − (v1 + kv2 + · · · + k m−1 vm + k m v 0 ) = k m+1 u0
=⇒ x − (v1 + kv2 + · · · + k m−1 vm + k m v 0 ) ∈ k m+1 U
=⇒ x − (v1 + kv2 + · · · + k m−1 vm + k m vm+1 ) ∈ k m+1 U where vm+1 = v 0

Hence the result is true for n = m + 1.

Therefore, by Mathematical Induction, the result is proved.

8
Theorem. 10.2 : Closed Graph Theorem
Let X and Y be Banach spaces and F : X −→ Y be a closed linear map. Then F is
continuous.
Proof. Let X and Y be Banach spaces and F : X −→ Y be a closed linear map.

To prove that f is continuous, it is enough to prove that F is bounded on

some nhd of 0.

For each positive integer n, let Vn = {x ∈ X, kF (x)k ≤ n}.

Here we prove that some Vn contains a nhd of 0 in X

Now,
∞
[ ∞
[
X= Vn = Vn
n=1 n=1
Hence !c
∞ ∞
[
c
\ c
Vn = X = φ =⇒ Vn = φ
n=1 n=1
T∞ c
Hence, n=1 Vn is not dense in X.
c
Then, by Baire’s theorem, one of the open sets Vn must not be dense in X.
c
Suppose Vp is not dense in X.
c
Then we can find some x0 ∈ X, and some δ > 0 such that U (x0 , δ) ∩ Vp = φ.
That is U (x0 , δ) ⊂ Vp .

Now we prove that U (0, δ) ⊂ V4p

If x ∈ U (0, δ), kxk < δ
This implies that kx0 − (x + x0 )k < δ
Hence, x + x0 ∈ U (x0 , δ) ⊂ Vp
Also we have x0 ∈ Vp .
Hence x + x0 and x0 are limit points of sequences in Vp .
Therefore, there exist sequences (vn ) and (wn ) in Vp such that vn → x + x0 and wn → x0
as n → ∞.
Hence vn − wn → (x + x0 ) − x = x as n → ∞.
Also,
kF (vn − wn )k = kF (vn ) − F (wn )k
≤ kF (vn )k + kF (wn )k
≤ p+p
≤ 2p
Hence vn − wn ∈ V2p
Therefore, x ∈ V2p .

9
That is if x ∈ U (0, δ), then x ∈ V2p
Therefore, if x ∈ U (0, δ), there exist an element x1 ∈ V2p such that kx − x1 k < 2δ .
Hence U (0, δ) ⊂ V2p + 12 U (0, δ)
Then by Lemma 10.1, there exist a sequence (vn ) ∈ V2p such that
 
1 1 1
x − v1 + v2 + · · · + n−1 vn ∈ n U (0, δ) for each n = 1, 2, . . .
2 2 2

Let wn = v1 + 21 v2 + · · · + 1
v ,
2n−1 n
n = 1, 2, . . ..
Now, x − wn ∈ 21n U (0, δ).
Hence kx − wn k < 2δn .
=⇒ wn → x in X.
Also, for n > m,
n
! m
!
X vj X vj
kF (wn ) − F (wm )k = F −F
j=1
2j−1 j=1
2j−1
n
!
X vj
= F
j=m+1
2j−1
n
X  v 
j
≤ F j−1
j=m+1
2
n  
X kF (vj )k
≤
j=m+1
2j−1
n  
X 2p
≤
j=m+1
2j−1
 
1 1 1
≤ 2p m + m+1 + · · · + n
2 2 2
 
2p 1 1 1
≤ 1 + + 2 + · · · + n−m
2m 2 2 2
 
2p 1 1
≤ m
1 + + 2 + ···
2 2 2
2p
≤ 2
2m
4p
≤ ................(1)
2m
Hence (F (wn )) is a Cauchy’s sequence in Y .
Since Y is Banach, (F (wn )) converges in Y .
Since F is closed, F (wn ) → F (x) in Y .
In (1), put m = 0, we get kF (wn ) − F (w0 )k ≤ 4p
Then put w0 = 0.

10
We get kF (wn )k ≤ 4p for all n ≥ 1.
Hence kF (x)k = limn→∞ kF (wn )k ≤ 4p
Hence kF (x)k ≤ 4p.
That is x ∈ V4p .
That is x ∈ U (0, δ) =⇒ x ∈ V4p .
In otherwords, U (0, δ) ⊂ V4p .
Therefore, F (U (0, δ)) is bounded.
Hence F is continuous.

This completes the proof of the theorem.

11
Definition. A linear map P : X → X is called a projection if P 2 = P .

Notes :

1. If P is a projection, then I − P is also a projection.

Hint : I − P is linear and (I − P )2 = I − P

2. R(P ) = Z(I − P )

Proof :

Z(I − P ) = {x ∈ X, (I − P )(x) = 0}
= {x ∈ X, I(x) − P (x) = 0}
= {x ∈ X, I(x) = P (x)}
= {x ∈ X, x = P (x)}
= {P (x)}
= R(P )

3. Z(P ) = R(I − P )

4. If P is a projection, then X = R(P ) + Z(P ).

5. If Y and Z are subspaces of X such that X = Y + Z and Y ∩ Z = {0}, then for

every x ∈ X can be uniquely expressed as x = y + z for some y ∈ Y and z ∈ Z. In
this case, the linear map P : X → X defined by P (x) = y is a projection. This
projection is called projection of X onto Y along Z.

12
Theorem. 10.3 Let X be a normed space and P : X −→ X be a projection. Then P is
a closed map if and only if the subspaces R(P ) and Z(P ) are closed in X. In that case,
P is, in fact, continuous, if X is a Banach space.
Proof. Let X be a normed space and P : X −→ X be a projection.

First assume that P is closed.

Let (yn ) be a sequence in R(P ) such that yn → y as n → ∞ in X.
Then, P (yn ) = yn .
Since yn → y, P (yn ) → y as n → ∞.
Since P is closed, this implies that P (y) = y
Hence y ∈ R(P ).
That is limit points of sequences in R(P ) contained in R(P ).
Hence R(P ) is closed.

Similarly, let (zn ) be a sequence in Z(P ) such that zn → z as n → ∞.

Therefore, P (zn ) = 0 and P (zn ) → 0 as n → ∞.
Since P is closed, this implies that P (z) = 0
Therefore, z ∈ Z(P ).
That is limit points of sequences in Z(P ) contained in Z(P ).
Hence Z(P ) is closed.

Conversely, assume that R(P ) and Z(P ) are closed sets in X.

Let (xn ) be a sequence in X such that xn → x and P (xn ) → y as n → ∞ in X.
Since R(P ) is closed, y ∈ R(P ).
Now, (I − P )(xn ) ∈ Z(P ).
Hence, xn − P (xn ) ∈ Z(P ).
But xn − P (xn ) → x − y as n → ∞.
Since Z(P ) is closed, x − y ∈ Z(P ).
Let z = x − y. Then z ∈ Z(P ).
Now, x = y + (x − y) = y + z, y ∈ R(P ) and z ∈ Z(P ).
Therefore, P (x) = y.
That is xn → x and P (xn ) → y implies P (x) = y.
Hence, P is a closed map.

Now, if X is Banach and P is closed, then by Closed Graph Theorem, P is continuous.

This completes the proof of the theorem.

13
Theorem. 10.4 : Let X and Y be normed spaces and F : X −→ Y be linear. Then F
is an open map if and only if there exists some γ > 0 such that for every y ∈ Y , there is
some x ∈ X with F (x) = y and kxk ≤ γkyk.
In particular, if a linear map is open, then it is surjective.
Proof. Let X and Y be normed spaces and F : X −→ Y be linear.

First assume that F is open.

Clearly, UX (0, 1) is an open set in X.
Therefore, F (UX (0, 1)) is an open set in Y .
Since 0 ∈ UX (0, 1), F (0) = 0 ∈ F (UX (0, 1)).
Therefore, there exist a δ > 0 such that U Y (0, δ) ⊂ F (UX (0, 1)).
Let y ∈ Y .
If y = 0, then x = 0 ∈ X and F (x) = y with kxk = kyk. Hence the result is true.
δy δ
If y 6= 0, then kyk = kyk kyk = δ.
δy
This implies that kyk ∈ U Y (0, δ).
δy
Therefore, there exist some x1 ∈ UX (0, 1) such that F (x1 ) = kyk
kyk
Define x = δ
x1 . Then clearly, x ∈ X.
Then  
kyk kyk kyk δy
F (x) = F x1 = F (x1 ) = = y.
δ δ δ kyk
Also,
kyk kyk kyk
kx|| = x1 = kx1 k < , ( since kx1 k < 1)
δ δ δ
1
By putting δ
= γ, we get kxk < γkyk

Conversely assume that there exists some γ > 0 such that for every y ∈ Y ,
there is some x ∈ X with F (x) = y and kxk ≤ γkyk.
Consider an open set E in X.
Here, we have to prove that F (E) is open in Y .
For that take an element F (x0 ) of F (E). Now we prove that an open nhd of F (x0 )
is contained in F (E).
Then clearly x0 ∈ E.
Then UX (x0 , δ) ⊂ E for some δ >  0.   
Let y ∈ Y with ky − F (x0 )k < γ . That is y ∈ UY F (x0 ), γδ .
δ

Then y − F (x0 ) ∈ Y .
Then by hypothesis, there exist some x ∈ X such that F (x) = y − F (x0 )
and kxk ≤ γky − F (x0 )k.
Hence kxk < γ k γδ k = δ and y = F (x) + F (x0 ) = F (x + x0 )
Therefore, k(x + x0 ) − x0 k = kxk < δ.
Hence, x + x0 ∈ UX (x0 , δ) ⊂ E.
This implies that F (x + x0 ) ∈ F (E). Hence y ∈ F (E).

14
  
Thus UY F (x0 ), γδ ⊂ F (E).
 
δ
That is if y ∈ UY F (x0 ), γ , then y ∈ F (E).
Therefore, a nhd of every element of F (E) is contained in F (E) and hence is open in Y .
Hence, if E is open in X, then F (E) is open in Y .
Hence F is an open map.

This completes the proof of the theorem.

15
Theorem. 10.5 : Let X and Y be normed spaces.

(a) If Z is a closed subspace of X, then the quotient map Q from X to X/Z is continuous
and open.

(b) Let F : X −→ Y be linear map such that the subspace Z(F ) is closed in X. Define
F : X/Z(F ) −→ Y by F (x + Z(F )) = F (x) for x ∈ X. Then F is an open map if
and only if F is an open map.

Proof. Let X and Y be normed spaces.

(a) Let Z be a closed subspace of X.

X
Consider the quotient map Q : X → Z
by Q(x) = x + Z, for all x ∈ X.
Clearly Q is linear. Verify
Also,

k|Q(x)k| = |kx + Zk|

= inf{kx + zk, z ∈ Z}
≤ kx + 0k
≤ kxk

Hence by Theorem 6.2, Q is continuous.

Now, consider any  > 0.

Let x + Z ∈ X
Z
.
Then
inf{kx + zk, z ∈ Z} = k|x + Zk| < (1 + )k|x + Zk|
Therefore, there exist some z0 ∈ Z such that kx + z0 k < (1 + )k|x + Zk|.
Now, Q(x + z0 ) = x + z0 + Z = x + Z
That is, if x + Z ∈ X
Z
, then there exist an element x + z0 ∈ X such that
Q(x + z0 ) = x + Z and kx + z0 k < γk|x + Zk| where γ = 1 + .
Hence By Theorem 10.4, Q is an open map.

(b) Let F : X −→ Y be linear map such that the subspace Z(F ) is closed in X. Define
F : X/Z(F ) −→ Y by F (x + Z(F )) = F (x) for x ∈ X.
X
Since Z(F ) is closed, Z(F )
is a nls in the quotient norm.
Consider the quotient map Q : X → X Z
by Q(x) = x + Z, for all x ∈ X.
X
Consider F : Z(F ) → Y as F (x + Z(F )) = F (x).
Then clearly, F (x) = F (x + Z(F )) = F (Q(x)).

First assume that F is open

Let E be an open set in X.
X
By part (a), Q is open and hence Q(E) is open in Z(F )
.

16
 Now, since F is open, this implies that F (E) = F (Q(E)) is open.
Hence F is open.

Conversely assume that F is open.

X
Let E be an open set in Z(F )
.
Since Q is continuous, Q−1 (E) is open in X.
Now, F (E) = F (Q−1 (E)) is open as F is open and Q−1 (E) is open.
Hence F is open.

This completes the proof of the theorem.

17
Theorem. 10.6 Open Mapping Theorem
Let X and Y be Banach spaces and F : X −→ Y be a linear map which is closed and
surjective. Then F is continuous and open.
Proof. Let X and Y be Banach spaces and F : X −→ Y be a linear map which is closed
and surjective.

By Closed Graph Theorem, F is continuous.

X
Since F is continuous, Z(F ) is closed in X and the map F : Z(F )
→ Y defined by
F (x + Z(F )) = F (x) is continuous by Theorem 6.2.
Since every continuous map is closed, F is closed.

To prove that F is injective

Let F (x1 + Z(F )) = F (x2 + Z(F ))
=⇒ F (x1 ) = F (x2 )
=⇒ F (x1 ) − F (x2 ) = 0
=⇒ F (x1 − x2 ) = 0
=⇒ x1 − x2 ∈ Z(F )
=⇒ x1 − x2 + Z(F ) = Z(F )
=⇒ x1 + Z(F ) = x2 + Z(F )
Hence F is injective.

To prove that F is surjective

Let y ∈ Y .
Since F is onto, there exist some x ∈ X such that F (x) = y.
X
Then clearly, x + Z(F ) ∈ Z(F )
.
Also, F (x + Z(F )) = F (x) = y.
Hence F is surjective.

Hence F is bijective closed linear map.

X
Hence its inverse G : Y → Z(F )
is closed and linear.
X
Now, since Z(F ) is closed, Z(F ) is Banach.
Hence by Closed Graph Theorem, G is continuous.
This implies that inverse of G is open.
Hence F is open.
By theorem 10.5(b), F is open.

This completes the proof of the theorem.

18