Functional Analysis Spring 2020
Lecture 1
Lecturer: Muhammad Yaseen Topic: Bounded Linear Operators
Definition 1.1 The restriction of an operator T : D(T ) → Y to a subset B ⊂
D(T ) is denoted by T |B : B → Y and is defined by
T |B (x) = T x, ∀ x ∈ B.
An extension of T to a set M ⊃ D(T ) is the operator T̃ : M → Y such that
T̃ |D(T ) = T i.e. T̃ |D(T ) (x) = T x ∀ x ∈ D(T ).
Theorem 1.2 Let T : D(T ) → Y be a bounded linear operator, where D(T )
lies in a normed space and Y is a Banach space. Then T has an extension
T̃ : D(T ) → Y such that T̃ is bounded and linear operator with kT̃ k = kT k.
Proof: First of all, we show the existence of T̃ .
1
2
Let x ∈ D(T ). Then by a previous theorem there exists a sequence (xn )∞
1 in
D(T ) such that xn → x. Since T is bounded and linear, therefore,
kT xn − T xm k = kT (xn − xm )k (because T is linear)
≤ kT kkxn − xm k → 0 as m, n → 0 (because T is bounded)
where we have used the fact that (xn )∞ ∞
1 being convergent is Cauchy. So (T xn )1
is a Cauchy sequence in Y . Since Y is complete, there exists a y ∈ Y such that
T xn → y i.e. lim T xn = y. Using this y as an image of x ∈ D(T ), we can define
n→∞
T̃ as T̃ x = lim T xn = y, where xn → x. Clearly T̃ x = T x, ∀ x ∈ D(T )(∵ if
n→∞
x ∈ D(T ), then the sequence x, x, · · · , is in D(T ) and converges to x so that
T̃ x = lim T xn = lim T x = T x).
n→∞ n→∞
Now we show that this definition of T̃ is independent of the choice of sequence
in D(T ) converging to x.
3
Suppose that xn → x and zn → x. Since (xn )∞
1 is Cauchy (because it is conver-
gent), therefore (T xn )∞
1 is a Cauchy sequence in Y (as shown in the beginning of
the proof ). Since Y is complete, (T xn )∞
1 converges. As lim xn = lim zn = x,
n→∞ n→∞
therefore, lim (xn − zn ) = 0. Then
n→∞
kT xn − T zn k = kT (xn − zn )k ≤ kT kkxn − zn k → 0 as n → ∞
so that T xn → T zn as n → ∞ i.e.
lim T xn = lim T zn = T̃ x = y.
n→∞ n→∞
Thus T̃ is an extension and is uniquely defined at each point of D(T ).
To Prove that T̃ is linear
Consider
T̃ (αx + βy) = lim T (αxn + βyn ) where xn → x, yn → y
n→∞
= lim (αT xn + βT yn ) (because T is linear)
n→∞
= α lim T xn + β lim T yn
n→∞ n→∞
= αT̃ x + β T̃ y.
So T̃ is linear.
To Prove that T̃ is bounded
4
Let xn → x and consider
kT̃ xk = k lim T xn k = lim kT xn k
n→∞ n→∞
≤ lim kT kkxn k = kT k lim kxn k (because T is bounded)
n→∞ n→∞
= kT kk lim xn k = kT kkxk
n→∞
i.e.
kT̃ xk ≤ kT kkxk (1)
so that T̃ is bounded. Now (1) implies
kT̃ xk
≤ kT k ∀ x 6= 0
kxk
kT̃ xk
⇒ sup ≤ kT k
x∈D(T ) kxk
x6=0
⇒ kT̃ k ≤ kT k (2)
Since D(T ) ⊃ D(T ), therefore,
kT̃ k ≥ kT k (3)
From (2) and (3), it follows that kT̃ k = kT k. This completes the proof.
Definition 1.3 A linear functional f is a linear operator with domain a vector
space and range in the scaler field K of the vector space X i.e.
f : X → K,
where K = R if X is real vector space and K = C if X is complex vector space.
Definition 1.4 A bounded linear functional f is a bounded linear operator with
domain a vector space and range in the scaler field K. So if f is bounded,
then there exists c > 0 such that
|f (x)| ≤ ckxk or |f (x)| ≤ kf kkxk.
5
In this case the norm of f exists and is defined as
|f (x)|
kf k = sup = sup |f (x)|.
x∈D(f ) kxk x∈D(f )
x6=0 kxk=1
Remark 1.5 The results that we proved for bounded linear operators continue
to hold true for bounded linear functionals.
Example 1.6 The norm k · k : X → R on a vector space X is a functional on
X and it is nonlinear because
kx + yk ≤ kxk + kyk.
Example 1.7 Consider f : R3 → R defined by f (x) = x·a, where x = (ξ1 , ξ2 , ξ3 ) ∈
R3 and a = (a1 , a2 , a3 ) is a fixed vector in R3 . Show that f is bounded.
Solution: f is bounded because
|f (x)| = |x · a| ∀ x ∈ R3
≤ kxkkak ∀ x ∈ R3
⇒ |f (x)| ≤ kakkxk ∀ x ∈ R3 (4)
Now (4) implies that
|f (x)|
≤ kak ∀ x 6= 0
kxk
|f (x)|
⇒ sup ≤ kak ∀ x 6= 0
x∈D(f ) kxk
⇒ kf k ≤ kak (5)
Now
|f (x)| |x · a|
kf k = sup = sup
x∈D(T ) kxk x∈D(f ) kxk
x6=0
|a.a| kak2
≥ =
kak kak
⇒ kf k ≥ kak (6)
6
From (5) and (6), we see that kf k = kak.
Rb
Example 1.8 Consider f : c[a, b] → R defined by f (x) = x(t)dt ∀ x ∈ c[a, b].
a
Show that f is bounded.
Solution: f is bounded because
Zb
|f (x)| = | x(t)dt| ∀ x ∈ c[a, b]
a
Zb
≤ |x(t)|dt ∀ x ∈ c[a, b]
a
≤ (b − a) max |x(t)| ∀ x ∈ c[a, b]
x∈[a,b]
⇒ |f (x)| ≤ (b − a)kxk ∀ x ∈ c[a, b]. (7)
Now (7) implies that
|f (x)|
≤ (b − a) ∀ x 6= 0
kxk
|f (x)|
⇒ sup ≤ (b − a)
x∈D(f ) kxk
x6=0
⇒ kf k ≤ (b − a) (8)
7
Also
|f (x)|
kf k = sup
x∈D(f ) kxk
x6=0
Rb
1dt
|f (1)| a
≥ = = (b − a)
k1k max (1)
t∈[a,b]
⇒ kf k ≥ (b − a). (9)
From (8) and (9) it follows that kf k = (b − a).
Example 1.9 Consider the space l2 and choose a fixed sequence a = (αi )∞ 2
1 ∈ l .
∞
Define a functional on l2 by f (x) = ξi αi ∀ x = (ξi )∞ 2
P
1 ∈ l . Show that f is
1
linear and bounded.
Solution: f is linear because
∞
X
f (αx + βy) = (αξi + βηi )αi
1
∞
X ∞
X
=α ξi αi + β ηi αi
1 1
= αf (x) + βf (y)
f is also bounded because
∞
X
|f (x)| = | ξi α i |
1
∞
X ∞
X
2 1/2
≤( |ξi | ) ( |αi |2 )1/2 (M. I. forp = q = 2)
1 1
⇒ |f (x)| ≤ kxkkak (10)
8
Now (10) implies that
|f (x)|
≤ kak
kxk
|f (x)|
⇒ sup ≤ kak
x∈D(f ) kxk
x6=0
⇒ kf k ≤ kak (11)
Now consider
|f (x)|
kf k = sup
x∈D(f ) kxk
x6=0
∞
|αi |2
P
|f (a)| 1
≥ =
kak kak
kak2
= = kak
kak
⇒ kf k ≥ kak (12)
From (11) and (12), we have kf k = kak
Definition 1.10 Let X be a vector space. Then the space of all linear functionals
on X is denoted by X ∗ and is called the algebraic dual space of X. It is easy
to see that X ∗ forms a vector space under the operations
(f1 + f2 )(x) = f1 (x) + f2 (x) ∀ f1 , f2 ∈ X ∗
(αf )(x) = αf (x) ∀ f ∈ X ∗ and ∀ α ∈ K.
We may go one step further and define linear functionals on X ∗ . The set of all
linear functionals on X ∗ is denoted by X ∗∗ and is called the second algebraic
dual space of X.
Definition 1.11 Let X and Y be normed spaces over the same field. Then
B(X, Y ) is the set of all bounded linear operators from X into Y . B(X, Y )
9
forms a vector space under the operations
(T1 + T2 )(x) = T1 ) + T2 x ∀ T1 , T2 ∈ B(X, Y )
(αT )(x) = αT x ∀ T ∈ B(X, Y ) and ∀ α ∈ K.
(13)
Also B(X, Y ) forms a normed space under the norm defined by
kT xk
kT k = sup = sup kT xk.
x∈D(T ) kxk x∈D(T )
x6=0 kxk=1
Theorem 1.12 If Y is a Banach space then B(X, Y ) is a Banach space.
Proof: Let (Tn )∞
1 be a Cauchy sequence in B(X, Y ). Then for each > 0, there
exists N ∈ N such that
kTn − Tm k < ∀ n, m > N
For any x ∈ X, the sequence (Tn x)∞
1 is in Y and consider
kTn x − Tm xk = k(Tn − Tm )xk
≤ kTn − Tm kkxk (∵ Tn , Tm ∈ B(X, Y ), ∴ Tn − Tm ∈ B(X, Y )
so that Tn − Tm is bounded)
≤ kxk∀ m, n > N (14)
This implies that (Tn x)∞
1 is a Cauchy sequence in Y . Since Y is complete, there
exists a y ∈ Y such that Tn x → y as n → ∞ (i.e. lim Tn x = y). Clearly, the
n→∞
limit y depends upon our choice of x ∈ X. This defines an operator T : X → Y
by T x = lim Tn x = y. To prove that B(X, Y ) is complete, we have to show
n→∞
kxk
that T ∈ B(X, Y ) and Tn −
−→T
T is linear
T (αx + βy) = lim Tn (αx + βy)
n→∞
= α lim Tn (x) + β lim Tn (y)
n→∞ n→∞
= αT (x) + βT (y).
10
T is bounded
From (14), we have
kTn x − Tm xk ≤ kxk ∀ m, n > N.
Letting m → ∞ in it, we obtain
kTn x − T xk ≤ kxk ∀ n > N
⇒ k(Tn − T )xk ≤ kxk ∀ n > N
(15)
⇒ Tn − T is bounded so that T = Tn − (Tn − T ) is also bounded. Thus T ∈
B(X, Y ).
Now from (14), we have
kTn x − T xk
≤∀n>N
kxk
kTn x − T xk
sup ≤∀n>N
x∈D(T ) kxk
x6=0
k(Tn − T )xk
sup ≤∀n>N
x∈D(T ) kxk
x6=0
⇒ kTn − T k ≤ ∀ n > N
kxk
⇒ Tn −
−→T
Hence B(X, Y ) is a normed space.
