myweb.ttu.

edu/bban
lev.ban@ttu.edu

Real Analysis
Byeong Ho Ban
Mathematics and Statistics
Texas Tech University

Chapter 6. Lp spaces(Last update : April 30, 2018)

1. When does equality hold in Minkowski’s inequality? (The answer is different for p = 1 and for 1 < p <
∞. What about p = ∞?)
Proof. (April 7th 2018)

(1)When p = 1
Z Z Z
kf + gk1 = kf k1 + kgk1 ⇐⇒ |f + g| = |f | + |g|
Z
⇐⇒ (||f + g| − (|f | + |g|)|)
⇐⇒ |f + g| = |f | + |g| a.e.
⇐⇒ ∀ almost every x, ∃α ≥ 0 such that f (x) = αg(x)

(2) When 1 < p < ∞

Since, when f = 0 a.e. or g = 0 a.e., it is obvious, let’s assume f 6= 0 and g 6= 0 a.e.
Observe the origin of the inequality below.
Z Z
|f + g| ≤ (|f | + |g|) |f + g|p−1
p

≤ kf kp |f + g|p−1 p + kgkp |f + g|p−1 p
p−1 p−1

As for the first inequality, it is equality iff

∀a.e.x, ∃αx ≥ 0 such that αx f (x) = g(x)

As for the second inequality, assuming αx > 0(When αx = 0 it is clear.) it is equality iff
p p p p
|f |p |f + g| p−1 |f | p−1 |g|p |f + g| p−1 |f | p−1
p = p = p and p = p = p a.e.
kf kp kf + gk p−1
p kf k p−1
p
kgkp kf + gk p−1
p kf k p−1
p
p−1 p−1 p−1 p−1

Thus,
!
kgkp
= αx or αx = 0 a.e. x
kf kp
1
2

(3) When p = ∞

kf + gk∞ = kf k∞ + kgk∞ ⇐⇒ kf + gk∞ ≥ |f | + |g| a.e.

⇐⇒ |f + g| ≥ |f | + |g| a.e.
⇐⇒ f = αx g a.e.
where αx ≥ 0. 

2. Prove Theorem 6.8.

a. If f and g are measurable functions on X, then kf gk1 ≤ kf k1 kgk∞ . If f ∈ L1 and g ∈ L∞ ,

kf gk1 = kf k1 kgk∞ iff |g(x)| = kgk∞ a.e. on the set where f (x) 6= 0.

b. k·k∞ is a norm on L∞

c. kfn − f k∞ → 0 iff there exists E ∈ M such that µ(E c ) = 0 and fn → f uniformly on E.

d. L∞ is a Banach space.

e. The simple functions are dense in L∞ .

Proof. (April 7th 2018)

a.
Observe that
Z Z
|f g| = |f ||g| ≤ |f | kgk∞ =⇒ kf gk1 = |f g| ≤ kgk∞ |f | = kf k1 kgk∞

Suppose that f ∈ L1 and g ∈ L∞ .

Observe that
Z Z
kf gk1 = kf k1 kgk∞ ⇐⇒ |f g| = kgk∞ |f |
Z
⇐⇒ |f |(||g| − kgk∞ |) = 0
⇐⇒ |g(x)| = kgk∞ a.e. on E = {x : f (x) 6= 0}

b.
(1)
kf k∞ = 0 ⇐⇒ |f (x)| ≤ 0 a.e.
⇐⇒ f = 0 a.e.
 3

(2) if λ = 0, it is obvious, so let’s assume λ 6= 0.

|λf | = |λ||f | ≤ |λ| kf k∞ =⇒ kλf k∞ ≤ |λ| kf k∞
|λ||f | = |λf | ≤ kλf k∞ =⇒ |λ| kf k∞ ≤ kλf k∞

(3)
|f + g| ≤ |f | + |g| ≤ kf k∞ + kgk∞ =⇒ kf + gk∞ ≤ kf k∞ + kgk∞

c. Let  > 0 be given.

Suppose that kfn − f k∞ → 0.
Let
∞
\
En = {x : |fn (x) − f (x)| > kfn − f k∞ }c E= En
n=1

Note that µ(Enc )

= 0 ∀n. so µ(E ) = 0. c

And observe that, there is N ∈ N such that

|fn (x) − f | ≤ kfn − f k∞ <  ∀n ≥ N ∀x ∈ E
Therefore, fn → f uniformly on E.

Conversely, suppose that fn → f on E ⊂ X with µ(E c ) = 0. Then, there is N ∈ N such that

|fn (x) − f (x)| <  ∀n ≥ N ∀x ∈ E
Then
kfn − f k∞ <  ∀n ≥ N

d.
Let {fn } ⊂ L∞ such that
∞
X
kfn k∞ < ∞
n=1

And let
∞
\
Enc = {x : |fn (x)| > kfn k∞ } E= En
n=1

Note that µ(Enc ) = 0, so µ(E c ) = 0 and that

∞
X ∞
X
|fn (x)| ≤ kfn k∞ < ∞ on E
n=1 n=1

Then
∞
X
fn (x) < ∞ a.e.
n=1
4

Now, observe that, for any m ∈ N

 ∞  ∞ ∞
 X  X X
f (x) ≤ |f (x)| ≤ kfn k∞
 
 n  n
 
n=m+1 n=m+1 n=m+1

Since

X∞ ∞
X
fn (x) ≤ kfn k∞ → 0 as m → ∞


n=m+1 n=m+1
∞
∞ ∞
the series converges in L , so L is a Banach space.

e.
Let f ∈ L∞ , then there exists a sequence of simple functions {fn } such that
fn → f a.e, |fn | ≤ |f |
Thus, there exists E such that fn → f uniformly on E and µ(E c ) = 0. Thus, by c.,
kfn − f k∞ → 0 as n → ∞
∞
therefore, the space of simple function is dense in L


3. If 1 ≤ p ≤ r ≤ ∞, Lp ∩ Lr is a Banach space with norm

kf k = kf kp + kf kr
and if p < q < r, the inclusion map Lp ∩ Lr → Lq is continuous.
Proof. (April 8th 2018)

First of all, note that Lp ∩ Lr is a vector space. And observe the following argument.
kf k = 0 ⇐⇒ kf kp + kf kr = 0
⇐⇒ kf kp = 0 ∧ kf kr = 0
⇐⇒ f = 0

 
kλf k = kλf kp + kλf kr = |λ| kf kp + kf kr = |λ| kf k

kf + gk = kf + gkp + kf + gkr ≤ kf kp + kgkp + kf kr + kgkr = kf k + kgk

Thus, Lp ∩ Lr is a normed vector space with the norm k·k.
Let {fn } ⊂ Lp ∩ Lr be a sequence such that

∞
X
kfn k < ∞
n=1

Note that
∞
X ∞
X ∞
X ∞
X
kfn kp ≤ kfn k < ∞ kfn kr ≤ kfn k < ∞
n=1 n=1 n=1 n=1

Since L and L are Banach, so are closed, L ∩ L is closed subset of both Lp and Lr . Thus, Lp ∩ Lr is
p r p r

Banach space with respect to the norms k·kp and k·kr .

 5

Thus,
∞ ∞
X X
fn → 0 fn → 0



n=m+1 p n=m+1 r
as m → ∞.
Therefore,
∞ ∞ ∞
X X X
fn = fn + fn → 0



n=m+1 n=m+1 p n=m+1 r
P∞
Since n=1 fn ∈ Lp ∩ Lr , this space is a Banach space.
Furthermore,
Observe that, if
kf − gk < 
then

1 λ 1−λ
kf − gkq = kf − gkλp kf − gk1−λ
r ≤ kf − gkλ kf − gk1−λ <  with = +
q p r
Thus, the inclusion map is continuous.


4. If 1 ≤ p < r ≤ ∞, Lp + Lr is a Banach space with norm

kf k = inf{kgkp + khkr : f = g + h}
and, if p < q < r, the inclusion map Lq → Lp + Lr is continuous.
Proof. (April 8th 2018)

Firstly, note that Lp + Lr is, clearly, a vector space. And the given mapping is a norm due to following
argument.
Let f ∈ Lp + Lr be given function. Then there exist {gn } ⊂ Lp and {hn } ⊂ Lr such that
 
lim kgn kp + khn kr = kf k gn + hn = f ∀n ∈ N
n→∞

kf k = 0 =⇒ inf{kgkp + khkr : f = g + h} = 0
 
=⇒ lim kgn kp + khn kr = 0
n→∞
=⇒ lim kgn kp = 0 = k0kp ∧ lim khn kr = 0 = k0kr
n→∞ n→∞
=⇒ f = 0 + 0 = 0

f = 0 =⇒ k0kp + k0kr ≥ inf{kgkp + khkr : f = g + h}

=⇒ kf k = 0

 
g h g h
kλf k = inf{kgkp + khkr : λf = g + h} = |λ| inf + :f = + = |λ| kf k
λ p λ r λ λ
6

Let f1 , f2 ∈ Lp + Lr . Note that, for given  > 0, there exists g1 , g2 and h1 , h2 such that fi = gi + hi
i = 1, 2 and
 
kg1 kp + kh1 kr < kf1 k + kg2 kp + kh2 kr < kf2 k +
2 2
Then

kf1 + f2 k = kg1 + g2 kp + kh1 + h2 kr ≤ kg1 kp + kg2 kp + kh1 kr + kh2 kr < kf1 k + kf2 k + 
Since  > 0 is arbitrary,
kf1 + f2 k ≤ kf1 k + kf2 k
Thus, Lp + Lr is a normed vector space.
Let {fn } ⊂ Lp + Lr be a sequence such that
∞
X
kfn k < ∞
n=1

For any n ∈ N, there exist gn , hn such that

1
fn = gn + hn kgn kp + khn kr < kfn k +
2n
Note that
∞
X ∞
X ∞
X ∞
X
kgn kp ≤ kfn k + 1 < ∞ khn kp ≤ kfn k + 1 < ∞
n=1 n=1 n=1 n=1

Thus, ∃g and h such that

∞
X ∞
X ∞
X ∞
X
g − g ≤ g − gn → 0 h − h ≤ h − hn → 0

n n

n=1 n=1 p n=1 n=1 r
p r
as n → ∞, so g ∈ L and h ∈ L .
Observe that,

X∞ X∞ X∞
g + h − fn ≤ g − gn + h − hn → 0

n=1
n=1
n=1

as n → ∞.
Therefore, the series converges to g + h ∈ Lp + Lr , so this space is Banach space.
Furthermore, noting that Lq ⊂ Lp + Lr , for given  > 0 and for any f, g ∈ Lq , if


kf − gkq <
2
, then

kf − gk = inf{kP kp + kRkr : f − g = P + R} ≤ k(f − g)χE kp + k(f − g)χE c kr ≤ 2 kf − gkq < 

where E = {x : |f (x) − g(x)| > 1}. Therefore, the inclusion map is continuous. 

5. Suppose 0 < p < q < ∞. Then Lp 6⊂ Lq iff X contains sets of arbitrarily small positive measure, and
Lq 6⊂ Lp iff X contains sets of arbitrarily large finite measures.
) < 2−n , and
(For the ’if ’ implication : In the first case there is a disjoint sequence {En } with 0 < µ(EnP
in the second case there is a disjoint sequence {En } with 1 ≤ µ(En ) < ∞. Consider f = an χEn for
suitable constants an .) What about the case q = ∞?
 7

Proof. (April 10th 2018)

(1) Let
a = inf{µ(E) : E ⊂ X ∧ µ(E) > 0}
Suppose a > 0 and, letting f ∈ Lp (µ) and An = {x : |f (x)| > n}, observe that
kf kpp
Z Z
p p p p
n µ(An ) = n dµ < |f | dµ ≤ kf kp =⇒ µ(An ) = p
An An n
Note that
lim µ(An ) = 0 =⇒ ∃N ∈ N such that µ(An ) < a ∀n ≥ N
n→∞
=⇒ µ(An ) = 0 ∀n ≥ N
, then, by the definition of An , kf k∞ < ∞. Now, by theorem 6.10,
p
1− pq
kf kq ≤ kf kpq kf k∞ <∞
Therefore, f ∈ Lq (µ).

Conversely, if a = 0, then there is a sequence of sets {Bn } such that

 
1 1
0 < µ(B1 ) < 0 < µ(Bn ) < min n , µ(Bn−1 ) ∧ Bn ⊂ Bn−1
2 2
Now, let
En = Bn \ Bn−1
then, {En } is a sequence of disjoint sets with the condition
1
0 < µ(En ) < n ∀n ∈ N
2
Nextly, let
∞
X n
f= 2 q χEn
n=1
Then, by MCT, note that
Z Z ∞ 
!p ∞ 
Z X
 p
p
X  nq  nq
|f | dµ = 2 χEn  dµ = 2 χEn  dµ
 
n=1 n=1
∞ Z ∞ ∞
X n pq
X n pq
X 1
= 2 χEn dµ = 2 µ(En ) < (1− pq )n
<∞
n=1 n=1 n=1 2
Thus, f ∈ Lp (µ). However,
Z Z ∞ 
!q ∞ 
Z X
 q
q
X  nq  nq
|f | dµ = 2 χEn  dµ = 2 χEn  dµ
 
n=1 n=1
∞
X Z ∞
X ∞
X
n n
= 2 χEn dµ = 2 µ(En ) < 1=∞
n=1 n=1 n=1

so f 6∈ Lq (µ), therefore, Lp 6⊂ Lq .

(2) Let
b = sup{µ(E) : µ(E) < ∞}
8

Suppose b < ∞ and let f ∈ Lq be given. And letting Bn = {x : |f (x)| > n1 }), observe that
Z Z
1
µ(Bn ) < |f | dµ ≤ |f |q dµ = kf kqq < ∞ =⇒ µ(Bn ) < nq kf kqq < ∞
q
nq Bn

Thus, µ(Bn ) ≤ b. Furthermore, letting B = {x : |f (x)| > 0},

∞
!
[
µ(B) = µ Bn = lim µ(Bn ) ≤ b
n→∞
n=1

By Hölder’s inequality and MCT, note that

Z Z Z  pq
q−p
p
|f | dµ = lim |f |p χBn dµ = lim kf p χBn k1 ≤ lim kf p k q kχBn k q ≤ q
|f | dµ b q <∞
n→∞ n→∞ n→∞ p q−p

Thus, f ∈ Lp , so Lq ⊂ Lp .

Conversely, suppose b = ∞ and let {Gn } be a sequence of sets chosen as below.

Firstly, choose F1 such that 1 ≤ µ(F1 ) < ∞, and then choose Fn such that
n−1
!c
[
Fn ⊂ Fk 1 ≤ µ(Fn ) < ∞
k=1

If, for some n,

n−1
!c
[
µ(Fn ) < 1 ∀Fn ⊂ Fk
k=1

then
n−1
!
[
µ(E) < µ Fn + 1 <∞ ∀E ⊂ X
k=1

and it is a contradiction.
Then define a function f as below.
∞ 1
X n− p
f= 1 χFn
n=1 µ(Fn ) q
Note that, by MCT,
∞ Z q ∞
X n− p X 1 q
kf kqq = χFn dµ = q < ∞ ∵ >1
n=1
µ(Fn ) n=1 n
p p

However, by MCT,
∞ Z ∞
p
X n−1 X 1 p p
kf kp = p χFn dµ = µ(Fn )1− q = ∞ ∵ 1 ≤ µ(Fn ) and 1 − >0
n=1 µ(Fn ) q n=1
n q

Therefore, f ∈ Lq but f 6∈ Lp . Thus, Lq 6⊂ Lp

When q = ∞,
(1)
If a > 0, then with same reasoning as above, f ∈ L∞ ∀f ∈ Lp regardless of value of p. Thus, Lp ⊂ L∞ .
 9

If a = 0, let
∞
X n
f= 2 p+1 χEn
n=1

Then, by MCT,
Z ∞ ∞
np −1
2n( p+1 ) < ∞
X X
kf kpp = p
|f | dµ = 2 p+1 µ(En ) <
n=1 n=1
n
so, f ∈ Lp . However, since 2 p+1 is unbounded and 0 < µ(En ) ∀n, f 6∈ L∞ . Therefore, we can restate
the problem as below.
If 0 < p < q ≤ ∞,
Lp 6⊂ Lq ⇐⇒ X contains sets of arbitraryly small positive measure.
(2) If b < ∞, by similar reasoning as above and with similar B,
Z Z
p p
kf kp = |f χB | = lim |f |p χBn ≤ kf kp∞ µ(B) < ∞ ∀f ∈ L∞
n→∞

Thus, L∞ ⊂ Lp ∀p > 0.

If b = ∞, let
∞
X 1
f= 1 χFn
n=1 np
Note that, by MCT,
∞ ∞ ∞ Z  p Z X∞  p
X 1 X1 X 1 1
∞= ≤ µ(Fn ) = 1 χFn dµ = 1 χFn dµ = kf kpp
n=1
n n=1 n n=1 np n=1 np
so, f 6∈ Lp , however, since
1
1 ≤ 1 =⇒ |f | ≤ 1 ∀n
np
f ∈ L∞ . Thus, we can restate the problem as below.
If 0 < p < q ≤ ∞
Lq 6⊂ Lp ⇐⇒ X contains sets of arbitraryly large finite measure.


6. Suppose 0 < p0 < p1 ≤ ∞. Find examples of functions f on (0, ∞) (with Lebesgue measure), such that
f ∈ Lp iff (a)p0 < p < p1 , (b) p0 ≤ p ≤ p1 , (c) p = p0 . (Consider functions of the form f (x) = x−a | log x|b .)
Proof. 

7. If f ∈ Lp ∩ L∞ for some p < ∞, so that f ∈ Lq for all q > p, then kf k∞ = limq→∞ kf kq .

Proof. (April 10th 2018)

Let kf k∞ >  > 0 be given and

E = {x : |f (x)| > kf k∞ − }
10

And, ∀q > p, observe that

Z Z
1
q
∞ > kf kq ≥ |f | dµ > (kf k∞ − )q dµ = µ(E)(kf k∞ − )q =⇒ kf kq > µ(E) q (kf k∞ − )
q
E E
1
=⇒ lim kf kq ≥ (kf kq − ) lim µ(E) q
q→∞ q→∞

=⇒ lim kf kq ≥ kf kq − 
q→∞

Since  is arbitrary,
lim kf kq ≥ kf k∞
q→∞

On the other hands, observe that

p
1− pq
lim kf kq ≤ lim kf kpq kf k∞ = kf k∞
q→∞ q→∞

Therefore,
lim kf kq = kf k∞
q→∞

8. Suppose µ(X)R= 1 and f ∈ Lp for some p > 0, so that f ∈ Lq for 0 < q < p.
a. log kf kq ≥ log |f |. (Use exercise 42d in §3.5, with F (t) = et .)
R R R
b. ( |f |q − 1)/q ≥ log kf kq , and ( |f |q − 1)/q → log |f | as q → 0.
R
c. limq→0 kf kq = exp( log |f |).
Proof. 

9. Suppose 1 ≤ p < ∞. If kfn − f kp → 0, then fn → f in measure, and hence some subsequence converges
to f a.e. On the other hand, if fn → f in measure and |fn | ≤ g ∈ Lp for all n, then kfn − f kp → 0.
Proof. (April 10th 2018)

Suppose that kfn − f kp → 0z and let  > 0 be given. Then for any n ∈ N, let
En, = {x : |fn (x) − f (x)| ≥ }
Then, observe that
Z
p
 µ(En, ) ≤ |fn − f |p dµ ≤ kfn − f kpp → 0 =⇒ µ(En, ) → 0 as n → ∞
En,

Therefore, fn → f in measure. Hence, by Theorem 2.30, there is a subsequence {fnj } ⊂ {fn } such that
fnj → f a.e.

On the other hand, suppose that fn → f in measure and |fn | ≤ g ∈ Lp for all n.
Note that there is a subsequence {fnk }k such that
fnk → f a.e.
p
so |f | ≤ g and |fn − f | ≤ 2g ∈ L and lim inf n→∞ |fn − f | = 0.
Observe that, by Fatou’s lemma,
Z Z Z Z Z Z
2 g = 2 g + |fn − f | ≤ lim inf {2 g + |fn − f | } = 2 g + lim inf |fn − f |p
p p p p p p p p p p
n→∞ n→∞
Z Z Z Z Z Z
2 g = 2 g − |fn − f | ≤ lim inf {2 g − |fn − f | } = 2 g − lim sup |fn − f |p
p p p p p p p p p p
n→∞ n→∞
 11

Thus,
Z Z
p
lim sup |fn − f | ≤ 0 ≤ lim inf |fn − f |p =⇒ kfn − f kp → 0 as n → ∞
n→∞ n→∞

10. Suppose 1 ≤ p < ∞. If fn , f ∈ Lp and fn → f a.e., then kfn − f kp → 0 iff kfn kp → kf kp . (Use
Exercise 20 in §2.3. )
Proof. (April 11st 2018)

Suppose kfn − f kp → 0 and, by Minkowski’s Inequality, note that

| kfn kp − kf kp | ≤ kfn − f kp → 0 =⇒ kfn kp → kf kp
Conversely, suppose kfn kp → kf kp and note that |fn − f |, |fn | + |f |, 2|f | ∈ Lp and that
Z
|fn | + |f | → 2|f | a.e. lim (|fn | + |f |)p → lim k|fn | + |f |kpp = 2p kf kpp
n→∞ n→∞

Therefore, by the Exercise 20 in §2.3, since |fn − f | ≤ |fn | + |f |,

Z Z
p
lim |fn − f | = lim |fn − f |p = 0
n→∞ n→∞

Thus,
kfn − f kp → 0 as n → ∞


11.
Proof. 

12.
Proof. 

13. Lp (Rn , m) is separable for 1 ≤ p < ∞. However, L∞ (Rn , m) is not separable.

(There is an uncountable set F ⊂ L∞ such that kf − gk∞ ≥ 1 for all f, g ∈ F with f 6= g.)
Proof. .

(1) Lp (Rn , m) is separable for 1 ≤ p < ∞.

( =⇒ ) Recall that, from the class note, Cc (Rn ) is dense in Lp (Rn , m). Let f ∈ C c (Rn ), then ∃k ∈ N
such that
supp(f ) ⊂ Rk (= [−k, k] × [−k, k] × [−k, k] × · · · × [−k, k]) .
Also, letting  > 0 be given, if P = {p : p: a polynomial of n variables.}, by the Stone-Weierstrass
theorem, there is a p ∈ P such that
|f (x) − p(x)χRk | <  ∀x ∈ Rk .
12

And observe that

Z
|f (x) − p(x)χRk |p dm ≤ p µ(Rk ) = p (2k)n

Thus,
P 0 = {pχRk : p ∈ P and k ∈ N}
is dense in Cc (Rn ) in p−norm.
Since, if PQ is a set of polynomials of n variables with rational coefficients,
PQ0 = {pχRk : p ∈ PQ and k ∈ N}
is countable and for any p ∈ P , there is p0 ∈ PQ such that

|p0 (x) − p(x)| <  ∀x ∈ Rn ,

observe that
Z
|p0 (x) − p(x)|p χRk dm ≤ p m(Rk ) = p (2k)n .

Thus, PQ0 is dense in P 0 , so is dense in Cc (Rn ) and so Lp .

(2) L∞ is not separable.

( =⇒ )
For each r ∈ (0, ∞), let fr = χBr (0) . Then note that
kfr − fs k∞ = 1 ∀r, s ∈ (0, ∞) with r 6= s
∞
Suppose that F be a dense subset of L , then
B 1 (fr ) ∩ F 6= ∅ ∀r ∈ (0, ∞)
2

Since
B 1 (fr ) ∩ B 1 (fs ) = ∅,
2 2

r 7→ B 1 (fr )
2

is an injective map, so F should be uncountable. 

14. If g ∈ L∞ , the operator T defined by T f = f g is bounded on Lp for 1 ≤ p ≤ ∞. Its operator norm is

at most kgk∞ , with equality if µ is semifinite.
Proof. (April 12th 2018)

If g = 0,
kT f kp = k0kp = 0 = k0k∞ = kgk∞ ∀f ∈ Lp
When p < ∞,
If g 6= 0, for any f ,
Z  p1  Z  p1
p
kT f kp = kf gkp = |f g|p dµ ≤ kgk∞ |f |p dµ = kgk∞ kf kp

When p = ∞
 13

|f g| = |f ||g| ≤ kf k∞ kgk∞ =⇒ kf gk∞ ≤ kf k∞ kgk∞

thus,
kT k ≤ kgk∞
Conversely, when 1 ≤ p ≤ ∞, if µ is semifinite, then ∀ > 0 there is a set A ⊂ X, such that
0 < µ(A) < ∞ and |g| > kgk∞ − 
Then, note that χA ∈ Lp and
Z  p1  Z  p1
p p p
kT χA kp = |gχA | dµ > (kgk∞ − ) |χA | dµ = (kgk∞ − ) kχA kp

or
kT χA k∞ = kgχA k∞ > (kgk∞ − ) kχA k∞
Thus,
kT χA kp
kT k ≥ > kgk∞ − 
kχA kp
Since  > 0 is arbitrary,
kT k ≥ kgk∞
Therefore,
kT k = kgk∞


15.
Proof. 

16.
Proof. 

17. With the notation as in Theorem 6.14, if µ is semifinite, q < ∞, and Mq (g) < ∞, then {x : |g(x)| > }
has finite measure for all  > 0 and hence Sg is σ−finite.
Proof. .

If not, ∃δ > 0 such that

µ(E ) = ∞ ∀0 <  ≤ δ
where
E = {x : |g(x)| > }.
Let
λ = sup{α > 0 : µ(Eα ) = ∞}.
(λ exists since µ is semifinite, so {α > 0 : µ(Eα ) = ∞} is bounded.)
14

1
Note that, when f = 1 χEλ+ 1 sgn(g),
µ(Eλ+ 1 ) p n
n

kf kp = 1
So
Z  Z  
  1 1− p1 1
Mq (g) ≥  f g  = 1 |g|dµ > µ(Eλ+ 1 ) λ+
µ(Eλ+ 1 ) p Eλ+ 1
n n
n n

then
 
1− p1 1 1
∞ > Mq (g) ≥ lim µ(Eλ+ 1 ) λ+ = λµ(Eλ )1− p = ∞.
n→∞ n n
And it is a contradiction.


18.
Proof. 

19.
Proof. 

20. Suppose supn kfn kp < ∞ and fn → f a.e.

a. If 1 < p < ∞, thenR fn → f weakly in Lp . (Given g ∈ Lq , where q is conjugate to p, andR  > 0, there
exist (i)δ > 0 such that E |g|p <  whenever µ(E) < δ, (ii)A ⊂ X such that µ(A) < ∞ and X\A |g|q < ,
and (iii)B ⊂ A such that µ(A \ B) < δ and fn → f uniformly on B. )

b. The result of (a) is false in general for p = 1. (Find counterexample is L1 (R, m) and l2 .) It is,
however, true for p = ∞ if µ is σ−finite and weak convergence is replaced by weak∗ convergence.
Proof. .

a.
Let g ∈ Lq and  > 0 be given and let q be the conjugate exponent of p. And let A = supn kfn kp + 1.
Note that
Z
ν(E) = |g|q
E
is a finite measure and is absolutely continuous with respect to µ. Thus, there is δ > 0 such that
Z Z 
q
 q
 1   q
∀E with µ(E) < δ =⇒ |g| =  |g|  <

E E 2 4A
Now, let
1
Bn = {x : |g(x)|q ≥ } ∀n
n
then
[
Bn = {x : |g(x)|q > 0} =let = B.
n
 15

therefore,
Z Z Z  Z Z Z Z
q q q q q q
lim |g| = lim |g| − |g| = |g| − lim |g| = |g| − |g|q = 0
n→∞ c
Bn n→∞ Bn n→∞ Bn B

Thus, there exists N ∈ N such that

Z
1   q
|g|q <
BN 2 4A
Note that, since
Z Z Z
µ(BN ) 1 q
= ≤ |g| ≤ |g|q < ∞,
N BN N BN

µ(BN ) < ∞.

By Egoroff’s theorem, there exists G ⊂ BN such that µ(G) < δ and

fn |BN \G → f |BN \G uniformly.
Note that
Z Z Z   q
q q q
|g| = |g| + |g| < .
(BN \G)c c
BN G 4A
Now, by Fatou’s lemma,
Z Z
p
|f | ≤ lim inf |fn |p = lim inf kf kpp ≤ Ap .
n→∞ n→∞

Thus, by Holder’s inequality, for any n ∈ N,

Z Z  1q
q  
|(fn − f )g| ≤ kfn − f kp |g| ≤ (kfn kp + kf kp ) < .
(BN \G)c (BN \G)c 4A 2
p
If we set K = µ(BN ) kgkpp + 1 and choose N ∈ N such that |fn (x) − f (x)| < 2p K for all x ∈ BN \ G and
n ∈ N with n ≥ N . If n ∈ N and n ≥ N , then
 1q 1
p p
Z Z 

|(fn − f )g| ≤ |fn − f | kgkq ≤ µ(BN \ G) p <
BN \G BN \G 2K 2
and therefore,
Z Z  Z Z Z
 
 fn g − f g  ≤ |(fn − f )g| = |(fn − f )g| + |(fn − f )g| < 
 
BN \G (BN \G)c

Therefore,
Z Z
lim fn g = fg
n→∞

It means that fn → f weakly in Lp .

b.
Suppose that p = 1. Note that

sup nχ(0, 1 ) = 1 < ∞ nχ(0, 1 ) → 0 a.e.

n 1 n
n

However, by Exercise 6.22, fn 6→ 0 weakly in L1 ([0, 1]).

Now, suppose that µ is σ−finite. Let g ∈ L1 be given and let M = supn kfn k∞ . Then
|f | < M a.e.
16

which means
|g(fn − f )| ≤ 2|g|M a.e. ∀n
Then, by DCT,
Z
lim |g(fn − f )| = 0
n→∞

Observe that
Z Z  Z
 
0 ≤  gfn − gf  ≤ |g(fn − f )|


Therefore,
Z Z 
 
lim  gfn − gf  = 0.
n→∞

So, when p = ∞, if µ is σ−finite, (a) is true.



21. If 1 < p < ∞, fn → f weakly in lp (A) iff supn kfn kp < ∞ and fn → f pointwise.
Proof. .

Suppose that 1 < p < ∞ and fn → f weakly in lp (A). Then for any φ ∈ (lp )∗ ,
φ(fn ) → φ(f ) n → ∞.
Since lp (A) is reflexive, for any g ∈ lp , there is gb ∈ (lp (A))∗∗ such that
gb(φ) = φ(g) ∀φ ∈ (lp (A))∗
and kb
g k = kgkp .
Note that
lim sup |fbn (φ)| = lim sup |φ(fn )| = φ(f ).
n→∞ n→∞

Thus, supn |fbn (φ)| < ∞.

By uniform boundedness principle,

sup kfn kp = sup fbn < ∞

n n
q
Also, note that, for any a ∈ A, χ{a} ∈ l where q is a conjugate exponent of p. Thus,
Z Z
lim fn (a) = lim χ{a} fn dµ = χ{a} f dµ = f (a)
n→∞ n→∞

Thus, fn → f pointwisely.

Conversely, note that if fn → f pointwisely, then it converges a.e.. Then by Exercise 6.20, it is true. 

22. Let X = [0, 1], with Lebesgue measure.

a. Let fn (x) = cos 2πnx. Then fn → 0 weakly in L2 (See Exercise 63 in §5.5), but fn 6→ 0 a.e. or in
measure.

b. Let fn (x) = nχ(0, 1 ) . Then fn → 0 a.e. and in measure, but fn 6→ 0 weakly in Lp for any p.
n
 17

Proof. .

a. Recall that L2 space is a Hilbert space with respect to the inner product
Z
hf, gi = f gdm.
[0,1]

And also note that

Z
hcos 2πnx, cos 2πmxi = cos 2πnx cos 2πmxdm = 0 if n 6= m.
[0,1]
cos 2πnx
Therefore { kcos } is an orthonormal sequence.
2πnxk n
Then, by Exercise 5.63,
cos 2πnx
→ 0 weakly in L2 .
kcos 2πnxk

Assume that cos 2πnx → 0 in measure. Then there exists a subsequence (cos 2πnj x)2 → 0 a.e. Observe
that, since
|cos 2πnj x|2 ≤ 1 ∀j on [0, 1]
and 1 ∈ L2 ([0, 1]), by DCT,
Z Z
2
lim (cos 2πnj x) dm = 0dm = 0.
j→∞ [0,1] [0,1]

However,
Z  
1 2 sin 4πnj 1
lim (cos 2πnj x) dm = lim +4 =
j→∞ [0,1] n→∞ 8 πnj 2
Thus, it is a contradiction.

b.
1
Let  > 0 and x ∈ (0, 1] be given and note that ∃N ∈ N such that N
< x, then
|fn (x)| = |nχ(0, 1 ) (x)| = 0 <  ∀n ≥ N.
n

Since {0} has measure 0, fn → 0 a.e.

Now, let En () = {x ∈ [0, 1] : |fn (x)| > }. Note that

lim µ({x ∈ [0, 1] : |fn (x)| > }) = lim µ({x ∈ [0, 1] : |nχ(0, 1 ) (x)| > }) = 0.
n→∞ n→∞ n

Thus, fn → 0 in measure.

However, for any p, note that 1 ∈ Lq ([0, 1]) (q is a conjugate exponent of p) and
Z Z
lim fn dm = lim nχ(0, 1 ) dm = 1.
n→∞ n→∞ n

Therefore, fn 6→ 0 weakly in Lp . 

23.
Proof. 

24.
18

Proof. 

25.
Proof. 

26.
Proof. 

27.
Proof. 

28.
Proof. 

29.
Proof. 

30.
Proof. 

Pn −1 −1
31. (A GeneralizedQHöder Inequality) Suppose that 1 ≤ p j ≤ ∞ and 1 pj = r ≤ 1. If fj ∈ Lpj
n r
Qn Qn
for j = 1, . . . , n, then 1 fj ∈ L and k 1 fj kr ≤ 1 kfj kpj . (First do the case n = 2.)
Proof. .

When n = 1, it is obvious. When n = 2, then with 1 < p1 , p2 < ∞,

1 1 1 1 1
+ = =⇒ + = 1.
p1 p2 r p1 /r p2 /r
Then, by Hölder inequality,
Z Z  p 1/r Z  p 1/r
p1 1 p2 2
r r r
|f1 f2 | ≤ (|f1 | ) r (|f2 | ) r

Z  p 1/r Z  p 1/r
1 2
p1 p2
= (|f1 | (|f2 |

= kf1 krp1 kf2 krp2 .

Thus,
kf1 f2 kr ≤ kf1 kp1 kf2 kp2 .
If one of them is ∞ or 1, it clearly follows from Hölder inequality.
 19

Now assume that

n n n
Y Y X 1 1
fj ≤ kfj kpj with = 0 ≤ 1.


1

1 j=1 j
p r
r

∈ Lpn+1 and n+1 1

= 1r ≤ 1, so r10 + pn+1
1
= 1r ≤ 1. Observe that, by the case when n = 2,
P
And let fn+1 1 pj
n+1 n n+1
Y Y Y
fj ≤ fj kfn+1 kpn+1 ≤ kfj kpj



1 r 1 r 1

Therefore,

Yn n
Y
fj ≤ kfj kpj ∀n ∈ N


1 r 1

2 2
32. Suppose that (X, M,R µ) and (Y, N , ν) are σ−finite measure spaces and K ∈ L (µ × ν). If f 2∈ L (ν),
the integral T f (x) = K(x, y)f (y)dν(y) converge absolutely for a.e. x ∈ X; moreover, T f ∈ L (µ) and
kT f k2 ≤ kKk2 kf k2 .
Proof. .

Observe that K ∈ L2 (µ × ν), so by Toneli,

Z Z Z  Z 
2 2 2
|K(x, y)| d(µ × ν) = |K(x, y)| dν(y) dµ(x) < ∞ =⇒ |K(x, y)| dν(y) < ∞ a.e.x ∈ X.

Then, since f ∈ L2 , observe that, by Hölder inequality,

Z Z  12
2
|K(x, y)||f (y)|dν(y) ≤ |K(x, y)| dν(y) kf k2 < ∞ a.e. x ∈ X

Therefore, T f converges absolutely, so is well defined.

Additionally, observe that, by Minkowski’s inequality for integral, Fubini-Toneli theorem and Hölder’s
inequality,
Z  12 Z Z 2 ! 21
2
 
|T f (x)| dµ(x) =  K(x, y)f (y)dν(y) dµ(x)
 
Z Z 2 ! 21
≤ |K(x, y)f (y)| dν(y) dµ(x)
 !2  21
Z Z  21
= |K(x, y)|2 dν(y) kf k2 dµ(x)

Z Z  21
= kf k2 |K(x, y)|2 dν(y)dµ(x)
Z  12
2
= kf k2 |K(x, y)| d(µ × ν)

= kf k2 kKk2 < ∞
Therefore, T f ∈ L2 (µ), so we are done. 
20

33. (Exercise 8.5) If s : Rn × Rn → Rn is defined by s(x, y) = x − y, then s−1 (E) is Lebesgue measurable
whenever E is Lebesgue measurable. (For n = 1, draw a picture of s−1 (E)√⊂ R2 . It should be clear that
after rotation through an angle π4 , s−1 (E) becomes F × R where F = {x : 2x ∈ E}, and Theorem 2.44
can be applied. The same idea works in higher dimensions.)
Proof. .

Observe that, when n = 1,

s−1 ((a, b)) = {(x, y) ∈ R2 : y = x − c ∀c ∈ (a, b)}.
Rotating the set counterclockwise centered at (0, 0) through π4 , you get
√ √
2a 2b
( , )×R
2 2
With same idea, if E ⊂ R is Lebesgue measurable set, √ after rotating s−1 (E) counterclockwise centered
at the origin, through π4 , we have F × R where F = {x : 2x ∈ E}. Since F × R is Lebesgue measurable
and measurability is independent of scaling and rotation, s−1 (E) is also Lebesgue measurable.

In general, when n is arbitrary natural number, let E ⊂ Rn be a Lebesgue measurable set. Then, with
appropriate rotational operator, we rotate s−1 (E) and get
√
(Fn × Rn ) =let {x ∈ Rn : 2nx ∈ E} × Rn
Since Lebesgue measuability is independent of scaling and rotating, Fn is Lebesgue measurable, and so
−1
s (E) is a Lebesgue measurable set. 

34. (Exercise 8.7) If f is locally integrable on Rn and g ∈ C k has compact support, then f ∗ g ∈ C k .
Proof. .

Recall
Z
f ∗g = f (y)g(x − y)dy.

Let Sg = {x : g(x) 6= 0} = supp(g) and let S = {x − y : y ∈ Sg }.

Now, observe that, for all 0 ≤ j ≤ k,
∂j ∂ j g(x − y) ∂ j g(x − y)
Z Z
(f ∗ g) = f (y) dy = f (y) dy
∂xj ∂xj S ∂xj
Since S is reflection and translation of a Lebesgue measurable set Sg , S is also Lebesgue measurable, so
f is integrable over S. In addition, since g ∈ C k , ∂ j g is continuous, so is integrable over compact set S.
∂j g ∂ j (f ∗g)
Therefore, f τx ( ∂x j ) is integrable over S. Therefore, ∂xj
is continuous for all 0 ≤ j ≤ k, so
f ∗ g ∈ Ck


35.
Proof. 

36.
 21

Proof. 

37.
Proof. 

38.
Proof. 

39.
Proof. 

40.
Proof. 

41.
Proof. 

42.
Proof. 

43.
Proof. 

44.
Proof. 

45.
Proof. 

46.
Proof. 

47.
Proof. 

48.
22

Proof. 

49.
Proof. 

50.
Proof. 

51.
Proof. 

52.
Proof. 

53.
Proof. 

54.
Proof. 

55.
Proof. 

56.
Proof. 

57.
Proof. 

58.
Proof. 

59.
Proof. 

60.
 23

Proof. 

61.
Proof. 

62.
Proof. 

63.
Proof. 

64.
Proof. 

65.
Proof. 

66.
Proof. 

67.
Proof. 

68.
Proof. 

69.
Proof. 

70.
Proof. 

71.
Proof. 

72.
24

Proof. 

73.
Proof. 

74.
Proof.