Lecture 10: Bounded Linear Operators.

2.11 Bounded Linear Operator

Recall: Let X and Y be linear spaces over K (where K = R or K = C). The operator L : X æ Y is called
linear if for every u, v œ X and –, — œ K, we have

L(–u + —v) = –Lu + —Lv.

Definition 1. Let X and Y be normed linear spaces. A linear operator L : X æ Y is called a bounded
linear operator if there exists a positive constant c > 0 such that

ÎLxÎY Æ cÎxÎX , for all x œ X.

Note: We often write ÎxÎ and ÎLxÎ instead of ÎxÎX and ÎLxÎY .

Proposition 13. Let L : X æ Y be a linear operator where X and Y are normed spaces over K (K = R
or K = C). Then the following statements are equivalent:

1. L is continuous at 0.

2. L is continuous on X.

3. There is a number c > 0 such that ÎLxÎ Æ c for all x œ X with ÎxÎ Æ 1.

4. There is a number c > 0 such that ÎLxÎ Æ cÎxÎ for all x œ X.

Proof. (1 ∆ 2). Let x œ X and suppose {xn } µ X such that lim xn = x. Then lim (xn ≠ x) = 0. Since
næŒ næŒ
L is continuous at 0, we have
lim L(xn ≠ x) = L(0).
næŒ

Since L is linear, L(0) = 0 and L(xn ≠ x) = L(xn ) ≠ L(x), for all n œ N. Therefore,

0 = L(0) = lim L(xn ≠ x) = lim (L(xn ) ≠ L(x)) = lim L(xn ) ≠ L(x).

næŒ næŒ næŒ

Hence lim L(xn ) = L(x), which means L is continuous at x œ X, for any x œ X. That completes the
næŒ
proof.

(2 ∆ 3).Suppose (3) is not true. Then there exists a sequence {xn } µ X such that

Îxn Î Æ 1 and ÎL(xn )Î Ø n, for all n = 1, 2, . . . .

Let wn = n≠1 xn , then

1 . 1
.
2.
.
Îwn Î Æ and ÎLwn Î = .L n≠1 xn . = n≠1 ÎL(xn )Î Ø 1 for all n = 1, 2, . . .
n
So lim Îwn Î = 0 and lim wn = 0. Since L is continuous at 0, we have lim L(wn ) = L(0) = 0, a
næŒ næŒ næŒ
contradiction with ÎLwn Î Ø 1.

45
 (3 æ 4). If x = 0, then ÎL(0)Î = 0 Æ cÎ0Î.
x ÎLxÎ
If x ”= 0, let z = . Then ÎzÎ = 1, so c Ø ÎLzÎ = . Therefore, cÎxÎ Ø ÎLxÎ.
ÎxÎ ÎxÎ
In both cases, we have ÎLxÎ Æ cÎxÎ, for all x œ X.

(4 æ 1). Given Á > 0. Choose ” = Á/c. Then when x œ X with Îx| < ”, we have

ÎLxÎ Æ cÎxÎ < c” < Á.

So for linear operators between normed linear spaces, boundedness is equivalent to continuity.

Definition 2. For a bounded linear operator L : X æ Y where X and Y are normed linear spaces, define
the operator norm
ÎLÎ := sup ÎLvÎ < Œ
vœX,ÎvÎÆ1

Proposition 14. Let L : X æ Y be a bounded linear operator where X and Y are normed linear spaces.
Then

1. ÎLuÎ Æ ÎLÎ ÎuÎ, for all u œ X.

2. If there is a constant C > 0 such that ÎLuÎ Æ CÎuÎ for all u œ X, then ÎLÎ Æ C.

3. If X ”= {0}, then
ÎLvÎ
ÎLÎ = sup ÎLvÎ = sup ÎLvÎ = sup
vœX,ÎvÎÆ1 vœX,ÎvÎ=1 vœX,v”=0 ÎvÎ

Proposition 15 (Bounded Linear Operators Between Finite Dimensional Normed Spaces). Let X and Y
be finite-dimensional normed spaces over K (R or C) with dim X = N and dim Y = M where N, M Ø 1.
Then any linear operator L : X æ Y is bounded.

Sketch of the Proof. Let {e1 , . . . , eN } and {f1 , . . . , fM } be a basis in X and Y , respectively. Suppose
M
ÿ
L(en ) = amn fm , n = 1, . . . , N.
m=1

N
q
Any x œ X can be written as x = cn en , for some c1 , . . . , cN œ K. Then
n=1
A N B N N M M
A N B
ÿ ÿ ÿ ÿ ÿ ÿ
L cn en = cn L(en ) = cn amn fm = amn cn fm
n=1 n=1 n=1 m=1 m=1 n=1

Recall that we have proved in previous lectures that

. .
.ÿN .
. .
. cn en . := max |cn |
. . 1ÆnÆN
n=1

is a norm on the finite dimensional normed space X.

46
 N
q
• Show that ÎLxÎŒ Æ ÎxÎŒ max |amn |.
1ÆmÆM n=1

• Using the property that any two norms on a finite dimensional normed linear spaces are equivalent,
show that there is a constant C > 0 such that ÎLxÎ Æ CÎxÎ for all x œ X.

Example 1. Consider a linear operator L : RN æ RM , L(x) := Ax (matrix multiplication), where A is

a matrix of real entries of size M ◊ N .
N
q
1. If we use the Î · ÎŒ norm for both RN and RM , then ÎLÎ = max |amn |.
1ÆmÆM n=1

M
q
2. If we use the Î · Î1 norm for both RN and RM , then ÎLÎ = max |amn |.
1ÆnÆN m=1
Ò
3. If we use the Î · Î2 norm for both RN and RM , then ÎLÎ = fl(AT A), where fl(B) is the maximum
of the magnitude of the eigenvalues of the square matrix B.

Proof. (1). For any x œ RN , then for any 1 Æ m Æ M , we have

- N -
-ÿ - N
ÿ N
ÿ N
ÿ
- -
|(Lx)m | = - amn xn - Æ |amn ||xn | Æ ÎxÎŒ |amn | Æ ÎxÎŒ max |amn |.
- - 1ÆmÆM
n=1 n=1 n=1 n=1

Therefore,
N
ÿ
ÎLxÎŒ = max |(Lx)m | Æ ÎxÎŒ max |amn |.
1ÆmÆM 1ÆmÆM
n=1
N
q
Therefore, ÎLÎ Æ max |amn |.
1ÆmÆM n=1
N
q
Next, we will prove that there exists x̂ œ RN with ÎxÎŒ = 1 such that ÎLx̂ÎŒ Ø max |amn |. Then
1ÆmÆM n=1

N
ÿ
ÎLÎ = sup ÎLzÎŒ Ø ÎLx̂ÎŒ Ø max |amn |
ÎzÎ=1 1ÆmÆM
n=1
N
q
Therefore, ÎLÎ = max |amn |.
1ÆmÆM n=1
N
q N
q
It remains to construct such x̂. Suppose max |amn | = |am0 n | for some 1 Æ m0 Æ M . Let
1ÆmÆM n=1 n=1
Y
_
]1 if am0 ,n Ø 0
x̂n = .
_
[≠1 if am0 ,n < 0

Then Îx̂ÎŒ = 1 and

N
ÿ N
ÿ N
ÿ
ÎLx̂ÎŒ = max |(Lx̂)m | Ø (Lx̂)m0 = am0 n x̂n = |am0 n | = max |amn |,
1ÆmÆM 1ÆmÆM
n=1 n=1 n=1

which completes the proof.

(2) & (3). Assignment 3

47
Example 2. Let X = C[a, b] with ηΌ , where ≠Œ < a < b < Œ and K : [a, b]◊[a, b] æ R be continuous.
For each u œ X, define the integral operator

⁄b
T u(x) := K(x, y)u(y) dy for all x œ [a, b].
a

From previous lectures, T : C[a, b] æ C[a, b] is a continuous and a compact operator.

Moreover, T is linear (prove this!) and

⁄b
ÎT Î = max |K(x, y)|dy.
aÆxÆb
a

Sketch of the proof. We will compute the operator norm of T .

• Step 1: Show that

⁄b
ÎT Î Æ max |K(x, y)|dy.
aÆxÆb
a

Let u œ C[a, b] and x œ [a, b]. Then

- b -
-⁄ - ⁄b ⁄b ⁄b
- -
|T u(x)| = - K(x, y)u(y) dy - Æ |K(x, y)| |(u(y)| dy Æ ÎuÎŒ |K(x, y)|dy Æ ÎuÎŒ max
- - |K(x, y)|dy.
- - aÆxÆb
a a a a

So
⁄b
ÎT uÎŒ = max |T u(x)| Æ ÎuÎŒ max |K(x, y)|dy.
aÆxÆb aÆxÆb
a

Therefore
⁄b
ÎT Î Æ max |K(x, y)|dy.
aÆxÆb
a

1 sb
• Step 2: ( next lecture) For every |K(x, y)|dy > Á > 0, construct an uÁ œ C[a, b] with ÎuÁ ÎŒ Æ 1
4a
such that
⁄b
ÎT uÁ ÎŒ Ø max |K(x, y)|dy ≠ 4Á.
aÆxÆb
a

Then
⁄b
ÎT Î = sup ÎT uÎŒ Ø ÎT uÁ ÎŒ Ø max |K(x, y)|dy ≠ 4Á.
uœC[a,b],ÎuÎŒ Æ1 aÆxÆb
a

sb
Let Á æ 0, we have ÎT Î Ø |K(x, y)|dy, which completes the proof.
a

48