Math 481-Functional Analysis I Assignment 2

Math 481 - Functional Analysis I

Irene Mae A. Antabo
Ronelo A. Fajardo
Rolex B. Teologia

February 24, 2022

Assignment 2

Exercise 1:

Let Ω := [0, 1]. For f : Ω −→ K and partition Z = {x0 , x1 , . . . , xn }, n ∈ N, of Ω define

n
X
VZ (f ) := |f (xi ) − f (xi−1 )|.
i=1

Furthermore we call V (f ) := sup{VZ (f ) | Z is a partition of Ω} the variation of f . We

define the space of functions of bounded variation by
BV (Ω) := {f : Ω −→ K | V (f ) < ∞} and ∥f ∥BV := |f (0)| + V (f ), f ∈ BV (Ω).
Show that:
(1) The map ∥ · ∥BV : BV (Ω) −→ R is a norm and (BV (Ω), ∥ · ∥BV ) is a Banach space.

Proof. Let f ∈ BV (Ω). Observe that both |f (0)|, V (f ) ≥ 0 and finite. Hence,
∥f ∥BV := |f (0)| + V (f ) ≥ 0 and finite. To show that ∥ · ∥BV is a norm, consider the
following:
(i) For any f ∈ BV (Ω) and λ ∈ K,
∥λf ∥BV = |λf (0)| + V (λf )
= |λ||f (0)| + sup{VZ (λf ) | Z is a partition of Ω}
( n )
X  
= |λ||f (0)| + sup λf (xi ) − λf (xi−1 )


( i=1
n 
!)
X 
= |λ||f (0)| + sup |λ| f (xi ) − f (xi−1 )
 

( ni=1 )
X  
= |λ||f (0)| + |λ| sup f (xi ) − f (xi−1 )

i=1
= |λ||f (0)| + |λ| sup{VZ (f ) | Z is a partition of Ω}
= |λ||f (0)| + |λ|V (f )|
= |λ|(|f (0)| + V (f ))
= |λ|∥f ∥BV
Thus, ∥λf ∥BV = |λ|∥f ∥BV .

MSU-Iligan Institute of Technology 1 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

(ii) For any f, g ∈ BV (Ω), observe that

|(f + g)(0)| = |f (0) + g(0)| ≤ |f (0)| + |g(0)|.

Moreover, suppose that Z is a partition of Ω. Then,

V (f + g) = sup{VZ (f + g) | Z is a partition of Ω}
( n )
X  
= sup (f + g)(xi ) − (f + g)(xi−1 )


( i=1
n 
)
X 
= sup f (xi ) + g(xi ) − f (xi−1 ) − g(xi−1 )
 

( i=1
n 
)
X 
= sup [f (xi ) − f (xi−1 )] + [g(xi ) − g(xi−1 )]
 

( i=1
n h
)
X   i
≤ sup f (xi ) − f (xi−1 ) + g(xi ) − g(xi−1 )
   

( i=1
n  n 
)
X  X 
= sup f (xi ) − f (xi−1 ) + g(xi ) − g(xi−1 )
   
i=1 i=1
= sup{VZ (f ) + VZ (g) | Z is a partition of Ω}
≤ sup{VZ (f ) | Z is a partition of Ω} + sup{VZ (g) | Z is a partition of Ω}
= V (f ) + V (g)

Hence, V (f + g) ≤ V (f ) + V (g). Furthermore,

∥f + g∥BV = |(f + g)(0) + V (f + g)| ≤ |f (0)| + |g(0)| + V (f ) + V (g)

= (|f (0)| + V (f )) + (|g(0)| + V (g))
= ∥f ∥BV + ∥g∥BV

Thus, the triangle inequality holds, i.e. ∥f + g∥BV ≤ ∥f ∥BV + ∥g∥BV .

(iii) Let f ∈ BV (Ω). Suppose that ∥f ∥BV = 0.

Claim: For f ∈ BV (Ω), V (f ) = 0 if and only if f is constant in Ω.

Suppose that V (f ) = 0 and f is not constant in Ω = [0, 1]. Then there exists
x1 , x2 ∈ [0, 1] such that f (x1 ) ̸= f (x2 ) for all x1 ̸= x2 . Taking Z = {0, x1 , x2 , 1},
we have

VZ (f ) = |f (x1 ) − f (0)| + |f (x2 ) − f (x1 )| + |f (1) − f (x2 )|.

Note that for x1 ̸= x2 , then |f (x2 ) − f (x1 )| > 0. Since both |f (x1 ) − f (0)| ≥ 0
and |f (1) − f (x2 )| ≥ 0, this means that VZ (f ) > 0 implying that V (f ) > 0.
This contradicts our assumption that V (f ) = 0. Hence, f must be constant in
Ω. Conversely, suppose that f is constant in Ω = [0, 1]. This means that f is
a monotone function and hence VZ (f ) = |f (1) − f (0)|. However, f (1) = f (0)
and thus, VZ (f ) = 0 implying that V (f ) = 0. This proves our claim.

MSU-Iligan Institute of Technology 2 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

Now, since ∥f ∥BV = 0, this means that |f (0)| + V (f ) = 0 implying that

|f (0)| = 0 and V (f ) = 0. With |f (0)| = 0, f (0) = 0 and since f is constant in
Ω as shown in the claim above, f (x) = 0 for all x ∈ Ω. Thus, f = 0.
On the other hand, suppose that f = 0. Consequently, f (0) = 0 and
V (f ) = 0. Thus, ∥f ∥BV = |f (0)| + V (f ) = 0. Finally therefore, ∥f ∥BV = 0 if
and only if f = 0.

Combining all these results, this means that ∥ · ∥BV : BV (Ω) −→ R is a norm on Ω.
To show that (BV (Ω), ∥·∥BV ) is a Banach space, we first consider the following claim.

Claim: For f ∈ BV (Ω), ∥f ∥sup ≤ ∥f ∥BV .

For a = 0, |f (0)| ≤ |f (0)| + V (f ) since V (f ) ≥ 0. For a = 1, we take Z = {0, 1}
so that |f (1)| − |f (0)| ≤ VZ (f ) ≤ V (f ) which implies that |f (1)| ≤ |f (0)| + V (f ).
For 0 < a < 1, we take Z = {0, a, 1}. Hence, |f (a)| − |f (0)| ≤ VZ (f ) ≤ V (f )
implying that |f (a)| ≤ |f (0)| + V (f ). Thus, for f ∈ BV (Ω), ∥f ∥sup ≤ ∥f ∥BV . This
proves our claim.

The claim above implies that BV (Ω) ⊆ B(Ω), the vector space of bounded
functions defined on Ω. Now, let (fn )n∈N be a Cauchy sequence in BV (Ω). Then for
every ϵ > 0, there exists N ∈ N with n, m ≥ N such that ∥fn − fm ∥BV < ϵ. Using
the above claim, we have

∥fn − fm ∥sup ≤ ∥fn − fm ∥BV < ϵ,

which means that (fn )n∈N is also Cauchy in B(Ω). Since B(Ω) is complete, (fn )n∈N
converges to f ∈ B(Ω) with respect to the norm ∥ · ∥sup . Moreover, ∥fn − f ∥sup −→ 0
implies that fn −→ f pointwise so that

|fn (0) − f (0)| + VZ (fn − f ) = lim {|fn (0) − fm (0)| + VZ (fn − fm )}

n→∞

for any partition Z of Ω = [0, 1]. Hence, |fn (0) − f (0)| + VZ (fn − f ) ≤ ϵ for all
n ≥ N . So accordingly, fn − f ∈ BV (Ω) implying that f = fn − (fn − f ) ∈ BV (Ω)
and
∥fn − f ∥BV = |fn (0) − f (0)| + V (fn − f ) ≤ ϵ
for all n ≥ N . That is, (fn )n∈N converges to f ∈ BV (Ω) with respect to the norm
∥ · ∥BV . Therefore, BV (Ω) is complete and (BV (Ω), ∥ · ∥BV ) is a Banach space.

(2) The space BV (Ω) with ∥ · ∥sup is not complete.

Proof. Let Ω := [0, 1]. Consider the sequence of functions in BV (Ω) given by
(
1
0 if 0 ≤ x ≤ nπ
fn (x) =
x sin x1 1


if nπ <x≤1

and suppose that f (x) is given by

(
0 if x = 0
f (x) = 1


x sin x
otherwise.

MSU-Iligan Institute of Technology 3 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

For any x ∈ Ω := [0, 1],

      
1 1  1 
−1 ≤ sin ≤ 1 =⇒ −x ≤ x sin ≤ x =⇒ x sin ≤ x =⇒ |f (x)| ≤ x.
x x x 

Thus, |f (x)| ≤ x on Ω := [0, 1]. Now, for each n ∈ N,

(  1
0 on 0, nπ
fn (x) = 1 
f (x) on nπ ,1 .

Moreover,

sup |fn (x) − f (x)| = sup |fn (x) − f (x)|

0≤x≤1 1
0≤x≤ nπ

= sup |f (x)|
1
0≤x≤ nπ

≤ sup x
1
0≤x≤ nπ

1
=
nπ
1
Hence, for all n ∈ N, 0 ≤ ∥fn − f ∥sup ≤ . Taking the limit, we have
nπ
lim ∥fn − f ∥sup = 0
n→∞

∥·∥sup
which implies that fn −−−→ f , that is fn converges to f ∈ BV (Ω) with respect to
the norm ∥ · ∥sup . Now, each fn (x) is of bounded variation on Ω := [0, 1] since fn (x)
is simply the sum, which is finite, of functions of bounded variations on Ω := [0, 1].
Therefore, the space BV (Ω) with ∥ · ∥sup is not complete.

MSU-Iligan Institute of Technology 4 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

Exercise 2:

Let (∥ · ∥n )n∈N be a total family of semi-norms on K-vector space X.

(1) Define for x, y ∈ X
∞  
X 1 ∥x − y∥n
d(x, y) := .
n=1
2n 1 + ∥x − y∥n
Show that d is well-defined and a metric.

Proof. Let (∥ · ∥n )n∈N be atotal family of semi-norms on K-vector space X. Let
1 ∥x − y∥n
x, y ∈ X. Observe that n ≥ 0 for all n ∈ N. Thus,
2 1 + ∥x − y∥n
∞  
X 1 ∥x − y∥n
d(x, y) := n
≥0
n=1
2 1 + ∥x − y∥n

∥x − y∥n
for all n ∈ N. Also, since ≤ 1 for all n ∈ N, then
1 + ∥x − y∥n
 
1 ∥x − y∥n 1
n
≤ n.
2 1 + ∥x − y∥n 2
Consequently,
∞   X ∞
X 1 ∥x − y∥n 1
d(x, y) := n
≤ < ∞.
n=1
2 1 + ∥x − y∥n n=1
2n

Hence, d(x, y) ≥ 0 and finite and thus d is well-defined. To show that d is a metric,
consider the following:
(i) Let x, y ∈ X. Now,
∞   X ∞  
X 1 ∥x − y∥n 1 ∥y − x∥n
d(x, y) = = = d(y, x).
n=1
2n 1 + ∥x − y∥n n=1
2n 1 + ∥y − x∥n

Hence, d is symmetric, i.e. d(x, y) = d(y, x).

s
(ii) Let x, y, z ∈ X. Consider the function f : R≥0 → R≥0 with s 7→ 1+s
increasing.
Thus, for any p, q ∈ X,
∥p + q∥n ≤ ∥p∥n + ∥q∥n =⇒ f (∥p + q∥n ) ≤ f (∥p∥n + ∥q∥n ).
Furthermore,
f (∥p + q∥n ) ≤ f (∥p∥n + ∥q∥n )
∥p + q∥n ∥p∥n + ∥q∥n
≤
1 + ∥p + q∥n 1 + ∥p∥n + ∥q∥n
∥p∥n ∥q∥n
= +
1 + ∥p∥n + ∥q∥n 1 + ∥p∥n + ∥q∥n
∥p∥n ∥q∥n
≤ +
1 + ∥p∥n 1 + ∥q∥n

MSU-Iligan Institute of Technology 5 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

We set p = x − z and q = z − y. Accordingly, for all n ∈ N,

∥p + q∥n ∥p∥n ∥q∥n
≤ +
1 + ∥p + q∥n 1 + ∥p∥n 1 + ∥q∥n
∥x − y∥n ∥x − z∥n ∥z − y∥n
≤ +
1 + ∥x − y∥n 1 + ∥x − z∥n 1 + ∥z − y∥n
     
1 ∥x − y∥n 1 ∥x − z∥n 1 ∥z − y∥n
≤ +
2 1 + ∥x − y∥n 2 1 + ∥x − z∥n 2 1 + ∥z − y∥n
∞   X ∞   X ∞  
X 1 ∥x − y∥n 1 ∥x − z∥n 1 ∥z − y∥n
≤ +
n=1
2 1 + ∥x − y∥ n n=1
2 1 + ∥x − z∥ n n=1
2 1 + ∥z − y∥n
d(x, y) ≤ d(x, z) + d(z − y)
Thus, the triangle inequality holds, i.e. d(x, y) ≤ d(x, z) + d(z − y).
(iii) Let x, y ∈ X. Now,
∞  
X 1 ∥x − y∥n
d(x, y) = 0 ⇐⇒ n
=0
n=1
2 1 + ∥x − y∥ n

∥x − y∥n
⇐⇒ = 0 for all n ∈ N
1 + ∥x − y∥n
⇐⇒ ∥x − y∥n = 0 for all n ∈ N
Since (∥ · ∥n )n∈N is total, ∥x − y∥n = 0 implies that for all n ∈ N, x − y = 0
and so x = y. Thus, d(x, y) = 0 if and only if x = y.
Therefore, combining all these results, (X, d) is a metric space.
n→∞
(2) Let (xn )n∈N be a sequence in X. Show that xn −−−→ x with respect to d if and only
n→∞
if ∥xn − x∥k −−−→ 0 for all k ∈ N.
n→∞
Proof. Let (xn )n∈N be a sequence in X. Suppose that xn −−−→ x with respect to d.
That is, d(xn , x) −→ 0 as n −→ ∞. Fixing n, k ∈ N, we have
 
1 ∥xn − y∥k
≤ d(xn , x)
2k 1 + ∥xn − y∥k
∥xn − y∥k ≤ 2k (1 + ∥xn − y∥k ) d(xn , x).
Taking the limit of both sides of the inequality,
lim ∥xn − y∥k ≤ lim 2k (1 + ∥xn − y∥k ) d(xn , x)
n→∞ n→∞
k
= 2 · lim (1 + ∥xn − y∥k ) · lim d(xn , x)
n→∞ n→∞
=0
Hence, lim ∥xn − y∥k ≤ 0 since lim d(xn , x) = 0. In particular, for all k ∈ N,
n→∞ n→∞

lim ∥xn − y∥k = 0.

n→∞

Conversely, suppose that ∥xn − x∥k −→ 0 as n −→ ∞ for all k ∈ N. Let ϵ > 0. We

take an index Kϵ ∈ N with
∞
X 1
≤ ϵ.
k=K +1
2k
ϵ

MSU-Iligan Institute of Technology 6 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

Moreover, we can find an Nϵ ∈ N such that ∥xn − x∥k ≤ ϵ. for all k ∈ {1, 2, . . . , Kϵ }
and n ≥ Nϵ . Furthermore, it follows that
Kϵ ∞
X 1 X 1
d(xn , x) ≤ k
∥xn − x∥k +
k=1
2 k=K +1
2k
ϵ
Kϵ
X 1
≤ϵ· +ϵ
k=1
2k
≤ 2ϵ for all n ≥ Nϵ .
n→∞
This means that d(xn − x) −→ 0 as n −→ ∞. That is, xn −−−→ x with respect to
n→∞ n→∞
d. Finally therefore, xn −−−→ x with respect to d if and only if ∥xn − x∥k −−−→ 0
for all k ∈ N.

(3) Is the metric d induced by a norm if X ̸= {0}?

Proof. Let x, y ∈ X ̸= {0} and α ∈ K. Now, for all n ∈ N,
∞  
X 1 ∥αx − αy∥n
d(αx, αy) =
n=1
2n 1 + ∥αx − αy∥n
∞  
X 1 |α|∥x − y∥n
=
n=1
2n 1 + |α|∥x − y∥n
∞  
X 1 ∥x − y∥n
= |α|
n=1
2n 1 + |α|∥x − y∥n

Let K = {z ∈ C | |z| = 1} be the set of all complex numbers with unit mod-
ulus. Observe that d(αx, αy) = |α|d(x, y) for any α ∈ K ∪ {−1, 0}. Otherwise,
d(αx, αy) ̸= |α|d(x, y). Therefore, if X ̸= {0}, the metric d is not induced by a
norm.

MSU-Iligan Institute of Technology 7 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

Exercise 3:

Show that:
1 1
(1) For 1 ≤ p, q ≤ ∞ with p
+ q
= 1r , such that r ≥ 1, for all x ∈ lp (K), y ∈ lq (K), we
have:
∥xy∥lr ≤ ∥x∥lp ∥y∥lq .

Proof. We utilize the Young’s inequality which states that for all real numbers a
and b with 0 < a ≤ b < ∞ and 1 ≤ p, q ≤ ∞ such that p1 + 1q = 1,

ap b q
ab ≤ + .
p q
Setting ar and br into the Young’s inequality given above,

(ar )p (br )q arp brq

ar br = (ab)r ≤ + = +
p q p q

for 1
p
+ 1
q
= 1. Taking p′ = rp and q ′ = rq, we obtain
′ ′
ar bq
 
r
(ab) ≤ r +
p q

for p1 + 1q = 1r . Replacing p′ and q ′ by p and q, respectively, we obtain a generalization

of the Young’s inequality given as
 p
bq

r a
(ab) ≤ r + .
p q

Now, let x = (xn ) ∈ lp (K) and y = (yn ) ∈ lq (K). For all n ∈ N, we set
xn yn
x̃n = ! p1 and ỹn = ! 1q .
∞
X ∞
X
|xn |p |yn |q
n=1 n=1

Consequently, we have

r |xn |r r |yn |r
|x̃n | = ! pr and |ỹn | = ! rq .
∞
X ∞
X
|xn |p |yn |q
n=1 n=1

and
p |xn |p p |yn |p
|x̃n | = ∞ and |ỹn | = ∞ .
X X
|xn |p |yn |q

n=1 n=1

Using the generalized Young’s inequality obtained above, for all n ∈ N,

|x̃n |p |ỹn |q
 
r r r
|x̃n ỹn | = |x̃n | |ỹn | ≤ r + .
p q

MSU-Iligan Institute of Technology 8 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

This implies that

|x̃n |p |ỹn |q
 
r r
|x̃n | |ỹn | ≤ r +
p q
 

|xn |r |yn |r |xn |p |yn |p 

 
1 1
! pr ! rq ≤ r  · X

∞ + · ∞ 
∞ ∞ p p
q X 
q
|xn | |yn |
X X 
|xn |p |yn |q
n=1 n=1
n=1 n=1
 
∞
!− pr ∞
!− rq
|xn |p |yn |p 
 
X
p
X
q r
1 r 1
|xn | |yn | |xn | |yn | ≤ r  · ∞
 + · ∞ 
n=1 n=1
p X
p
q X 
q
|xn | |yn |

n=1 n=1

Taking the summation over N, we have

 ∞
X ∞
X

p q
∞
!− pr ∞
!− rq ∞  |xn | |yn | 
X
p
X
q
X  1 n=1
r 1 n=1

|xn | |yn | |xn yn | ≤ r 
p · X
∞ + · ∞ 
n=1 n=1 n=1 p
q X 
q
|xn | |yn |

n=1 n=1
 
1 1
=r +
p q
1
=r·
r
=1
This means that
∞
!− pr ∞
!− rq ∞
X X X
|xn |p |yn |q |xn yn |r ≤ 1
n=1 n=1 n=1

which further implies that

∞ ∞
! pr ∞
! rq
X X X
|xn yn |r ≤ |xn |p |yn |q
n=1 n=1 n=1

Therefore, ∥xy∥lr ≤ ∥x∥lp ∥y∥lq .

n
X 1 1
(2) For 1 ≤ p1 , p2 , . . . , pn ≤ ∞ with = , such that r ≥ 1, for all x(i) ∈ lpi (K),
i=1
pi r
1 ≤ i ≤ n, we have:
Yn n
Y
(i)
x ≤ ∥x(i) ∥lpi .

r
i=1 l i=1

Proof. Let P (n) be the statement

Yn n
Y
(i)
P (n) : x ≤ ∥x(i) ∥lp .

r
i=1 l i=1

MSU-Iligan Institute of Technology 9 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

We will show that P (n) is true by induction on n. For n = 1,

P (1) : ∥x∥lr ≤ ∥x∥lpi

which is trivially true. For n = 2,

P (2) : ∥x1 x2 ∥lr ≤ ∥x1 ∥lp1 ∥x2 ∥lp2 .

This follows immediately from (1). Hence, P (2) is also true. Now, assume that
k
X 1 1
P (n) is true for n = k. That is, for 1 ≤ p1 , p2 , . . . , pk ≤ ∞ with = , such
p
i=1 i
r̃
that r̃ ≥ 1, for all x(i) ∈ lpi (K), 1 ≤ i ≤ k, we have:

Y k k
Y
(i)
P (k) : x ≤ ∥x(i) ∥lpi .

r̃
i=1 l i=1

We will show that P (n) is also true for n = k + 1. That is, for 1 ≤ p1 , . . . , pk+1 ≤ ∞
k+1
X 1 1
with = , such that r ≥ 1, for all x(i) ∈ lpi (K), 1 ≤ i ≤ k + 1, we have:
p
i=1 i
r

k+1
Y k+1
Y
(i)
P (k + 1) : x ≤ ∥x(i) ∥lpi .

r
i=1 l i=1

k+1
X 1 1
Now, suppose that 1 ≤ p1 , . . . , pk+1 ≤ ∞ with = where r ≥ 1. Let x(i) ∈
p
i=1 i
r
k
pi
X 1 1
l (K), 1 ≤ i ≤ k + 1. Observe that since = ,
p
i=1 i
r̃

k+1 k
X 1 X 1 1 1 1 1
= + = + = .
p
i=1 i
p
i=1 i
pk+1 r̃ pk+1 r

Moreover, observe that

1 1 1 r r 1 1
+ = =⇒ + =1 =⇒ r̃
+ pk+1 = 1.
r̃ pk+1 r r̃ pk+1 r r
r
Since r̃
≤ 1, we have r ≤ r̃. By assumption,

Yk k
Y
(i)
x ≤ ∥x(i) ∥lpi .

r̃
i=1 l i=1

Thus,
!r r
k+1
Y k+1
Y
(i) (i)
x = x

r r
i=1 l i=1 l
r
Y k
(i) (k+1)
= x x


r
i=1 l

MSU-Iligan Institute of Technology 10 Department of Mathematics and Statistics

Math 481-Functional Analysis I Assignment 2

k r
Y
(i) (k+1) r

≤ x x

r̃ lpk+1
i=1 l
k
Y r
≤ ∥x(i) ∥rlpi x(k+1) lpk+1
i=1
k+1
Y (i) r
= x p
l i
i=1
k+1
!r
Y (i)
= x p
l i
i=1

Taking the rth root of both sides of the inequality, we have

k+1 k+1
Y (i) Y
x ≤ ∥x(i) ∥lpi .
r
i=1 l i=1

This is the desired inequality for P (k + 1). This means that P (k + 1) is also true.
Therefore, by induction on n, the statement P (n) is true for all 1 ≤ i ≤ n.

MSU-Iligan Institute of Technology 11 Department of Mathematics and Statistics