Capitulo 19

Equipo 2:

-Castillo Chan Felix Armando.

-Yah Napte Roger Geovani.

-Yam Borges Jhon Alexander.

Problem 19.1
Let (X, F) be a measure space. A simple function ϕ : X → R is a function of
the form: n
X
ϕ(x) = cj 1Ej (x),
j=1

where cj ∈ R and {Ej }nj=1 is a collection of measurable sets Ej ∈ F. Prove

that we can write ϕ as:
Xm
ϕ (x) = dj 1Fj (x) ,
j=1

for some constantes dj ∈ R and pairwise disjoint measurable sets {Fj }nj=1 .

Solution. Let’s proceed by induction, for n = 1 trivially is true.

Suppose its true for n = k, for n = k + 1 we have:

k+1
X k
X
ϕ (x) = cj 1Ej (x) = cj 1Ej (x) + ck+1 1Ek+1 (x) ,
j=1 j=1

by induction hypothesis we have:

m
X
ϕ (x) = dj 1Fj (x) + ck+1 1k+1 (x)
j=1

1
and we will consider the following sets

m m
!
[ \ [
Fj′ = Fj − Ek+1 , Fm+1 = Ek+1 − Fj , Fm+2 = Ek+1 Fj
j=1 j=1

is clear that Fj′ , Fm+1 and Fm+2 are measurable sets, now let’s prove that they are
pairwise disjoint.

Let i ̸= j

Fj′ y Fi′ : 1 ≤ i, j ≤ m

as Fj′ = Fj − Ek+1 ⊂ Fj and Fj′ = Fj − Ek+1 ⊂ Fj

then Fj′ ∩ Fi′ ⊂ Fj ∩ Fi = ∅ ⇒ Fj′ ∩ Fi′ = ∅

Fj′ and Fm+1 :

m
!
[
Fj′ ∩ Fm+1 = (Rj − Ek+1 ) ∩ Ek+1 − Fj
j=1
m
!!
[
c
Fjc


= Fj ∩ Ek+1 ∩ Ek+1 ∩
j=1
m
!!
\
c
Fjc


= Ek+1 ∩ Ek+1 ∩ Fj ∩
j=1
m
!!
\
= ∅∩ Fj ∩ Fjc
j=1
= ∅

Fj′ and Fm+2 :

2
 m
!!
[
Fj′ ∩ Fm+2 = (Fj − Ek+1 ) ∩ Ek+1 ∩ Fj
j=1
k
!!
[
c


= Fj ∩ Ek+1 ∩ Ek+1 ∩ Fj
j=1
m
!!
[
c


= Ek+1 ∩ Ek+1 ∩ Fj ∩ Fj
j=1
= ∅

Fm+1 and Fm+2 :

m
! m
!!
[ [
c
Fm+1 ∩ Fm+2 = Ek+1 − Fj ∩ Ek+1 ∩ Fj
j=1 j=1
m
!c ! m
!!
[ [
= Ek+1 ∩ Fj ∩ Ek+1 ∩ Fj
j=1 j=1
m
!c m
!!
[ [
= (Ek+1 ) ∩ Fj ∩ Fj
j=1 j=1
= ∅

Finally, we have to

3
m m m
! m
!!
[ [ [ [
Fj′ ∪ Fm+1 ∪ Fm+2 = (Fj − Ek+1 ) ∪ Ek+1 − Fj ∪ Ek+1 ∩ Fj
j=1 j=1 j=1 j=1
m m
! m
!!
[ [ [
c


= Fj ∩ Ek+1 ∪ Ek+1 − Fj ∪ Ek+1 ∩ Fj
j=1 j=1 j=1
m
! m
! m
!!
[ [ [
c
= Fj ∩ Ek+1 ∪ Ek+1 − Fj ∪ Ek+1 ∩ Fj
j=1 j=1 j=1
m
! m
! m
!!
[ [ [
= Fj − Ek+1 ∪ Ek+1 − Fj ∪ Ek+1 ∩ Fj
j=1 j=1 j=1
m
[
= Fj ∪ Ek+1
j=1

Then if G = {Gk }m+2

k=1 with


′
 Fk 1 ≤ k ≤ m

Gk = Fm+1 k = m + 1

Fm+2 k = m + 2


therefore

m+2
X
ϕ (x) = λk 1Gk (x) ,
k=1

where G is a collection of pairwise disjoint measurable sets. □

4
 Problem 19.3
Let f : X → R be an F-measurable function. Show that there exists a sequence
of simple functions (fn ) where fn : X → such that |fn (x) | ≥ |fn+1 (x) | for all
pw
n ∈ N and fn −→ f on X.

Solution.

Let f : X → R be an F−measurable function then we can view f as:

f = f∗ − f−

where f + and f − are non-negative funtions and in the chapter 18 was proven that
both are F−measurables , therefore we can use the proposition 19.1.3 to guarantee
the existence of sequences (fn+ ) and (fn− ) such that fn+ ↑ f + and fn− ↑ f −

+ pw
It is for n it is true that fn+ (x) ≤ fn+1 (x) and fn+ −→ f + and respective for any n′
pw
it is true that fn−′ ≤ fn−′ +1 (x) and fn− −→ f −

Let (fn ) = (fn+ − fn− ) and because fn+ and fn− are measurable funtion

|fn (x) | = fn+ (x) + fn− (x) ≤ fn+1

+ −
(x) + fn+1 (x) = |fn+1 (x) |

pw pw pw
since fn+ −→ f + and fn− −→ f − we have fn+ − fn− −→ f + − f − and

pw
fn −→ f

5
 Problem 19.11
Let (R, L, µ) be the lebesgue measure space f : R → [0, ∞] be non-negative
Lebesgue measure function.
1. Show that the set function on L defined as:

v : L → [0, ∞]
Z
A 7→ f dµ,
A

is also a measure on L.
2. Show that if f ∈ L1 (R), then v is a finite measure.
3. Show that if µ (A) = 0, then v (A) = 0.
4. Prove: f > 0 µ−a.e, if and only if v (A) = 0 implies µ (A) = 0.
5. Suppose that g : R → [0, ∞]. Prove that g ∈ L1 (R, v) if and only if
gf ∈ L1 (R, µ) and we have:
Z Z
g dv = gf dµ.
R R

Solution.
R
1. a) (PD. v (∅) = 0) Lets see that v (∅) = ∅
f dµ for f be non-negative Lebes-
gue measure function and

Z
f dµ = sup {I∅ (ϕ) : ϕ : X → R its a simple function with ϕ ≤ f }
∅

Pn
but note that for any simple function ϕ = j=1 cj 1Ej (x) we have:

n
X n
X
I∅ (ϕ) = cj µ (∅ ∩ Ej ) = cj µ (∅) .
j=1 j=1

6
 and for the fact that µ is measure we have µ (∅) = 0 therefore:
n
X
cj µ (∅) = 0
j=1

in particular for ϕ ≤ f we have I∅ (ϕ) = 0 and we can conclude that

Z
f dµ = 0
∅

b) (PD. v (A) ≥ 0) Let A ∈ L then

Z
v (A) = f dµ
A
= sup {IA (ϕ) |ϕ : X → R its a simple function with ϕ ≤ f }
Pn
Let ϕ : j=1 cj 1Ej (x) a simple function such that ϕ ≤ f then

n
X
IA (ϕ) = cj µ (A ∩ Ej )
j=1

and because µ is a measure we have µ (A ∩ Ej ) ≥ 0 and

IA (ϕ) ≥ 0

therefore Z
v (A) = f dµ ≥ 0
A

S  P
∞
c) (PD. v j=1 Ej = ∞
j=1 v (Ej ) . ) Let any countable collection of pair-

7
 wise disjoint sets {Ej }∞
j=1 , note that:

∞
! Z
[
v Ej = S∞
f dµ
j=1 j=1 Ej
∞ Z
X
= f dµ
j=1 Ej

X∞
= v (Ej )
j=1

therefore v is a measure on L.

2. Let f ∈ L1 (R), and for the fact that f is non-negative function we have
Z Z Z
v (X) = f dµ = f dµ ≤ |f | dµ < ∞
X X X

therefore v is a finite measure

3. If µ (A) = 0 then
Z Z Z
f dµ ≤ max (f ) dµ = max (f ) dµ = max (f ) µ (A) = 0
A A A
R
therefore v (A) = A
f dµ = 0

4. (⇒) We will proceed by contradiction, if f > 0 µ−a.e. assume that v (A) = 0

does not imply µ (A) = 0 then exists A′ ∈ L such that if f > 0 µ−a.e.
then A f dµ = v (A′ ) = 0 and µ (A) > 0 but this is a contradiction to
R

proposition 19.5.5 item 4)

therefore f > 0 µ−a.e. ⇒ v (A) = 0 implies µ (A) = 0.

(⇐) for the return in the same way we will proceed by contradiction, suppose
that

µ (x ∈ X : f (x) = 0) > 0 (1)

8
 denote the subset Y = {x ∈ X : f (x) = 0}

then Z
v (Y ) = f dµ = 0
Y

since f is zero for all x ∈ Y and for hypothesis this imply that µ (Y ) = 0
but this is a contradicction to 1

therefore if v (A) = 0 implies µ (A) = 0 then f > 0 µ−a.e.

5. We will proceed in three steps: when g is an indicator function, a simple fun-

ction, and a general measurable function.

Step 1: Suppose that g = 1E , where E ∈ L is a measurable set.

Then, the integral of g with respect to v becomes:

Z Z Z
g dv = 1E dv = v(E) = f dµ.
R R E

On the other hand, the integral of gf with respect to µ is:

Z Z Z
gf dµ = 1E f dµ = f dµ.
R R E
R R
Therefore, R
g dv = R
gf dµ.

Step 2: Suppose that g is a simple function, that is:

n
X
g(x) = cj 1Ej (x),
j=1

where cj ≥ 0 are constants and Ej are disjoint measurable sets. The integral

9
of g with respect to v becomes:
Z n
X n
X Z
g dv = cj v(Ej ) = cj f dµ.
R j=1 j=1 Ej

On the other hand, the integral of gf with respect to µ is:

Z Z n
! n Z
X X
gf dµ = cj 1Ej f dµ = cj f dµ.
R R j=1 j=1 Ej

R R
Thus, R
g dv = R
gf dµ.

Step 3: Finally, suppose that g is a general measurable function. By Proposition

19.1.3, there exists a pointwise increasing sequence of simple functions (gn ),
R R
gn : R → R such that gn → g. Then, by step 2, we have R gn dv = R gn f dµ,
and using the Monotone Convergence Theorem (MCT), we get:

Z Z Z Z
lı́m gn dv = gdv = gf dµ = lı́m gn f dv.
n→∞ R R R n→∞ R

10
 Problem 19.15
Let ([0, ∞), L, µ) be the induced Lebesgue measure space and f ∈ L1 ([0, ∞)).
R
Prove that if [0,c] f dµ = 0 for any c > 0, then f ≡ 0 a.e.

Solution. First we prove that for every E ⊂ [0, ∞) bounded interval,f (x) = 0 a.e.

Let E ⊂ [0, c] for any c > 0, by hypothesis. With µ(E) > 0.

Case 1
Take f > 0 a.e. by contradiction.

Let V = {x ∈ E : f (x) ̸= 0} = {x ∈ E : f (x) > 0} by hypothesis.

It is clear that V ⊂ E ⊂ [0, c], and V, E, [0, c] are Lebesgue measurable sets. Well,
f ∈ L1 ([0, ∞)) are G[0,∞) measurable and since E is bounded with a positive measure,
E is open interval , closed interval, or half-closed interval, therefore E is Lebesgue
mesurable set, and E ∩ [0, ∞) = E then is in te σ−álgebra of [0, ∞)

It has to µ({x ∈ [0, c] : f (x) > m}) ≤ m1 [0,c] f dµ = 0 by Markov’s inequality, then
R

µ({x ∈ [0, c] : f (x) > m}) = 0 for any m > 0.

Let Vn = {x ∈ E : f (x) > n1 }, note that Vn ⊂ Vn+1 and V = ∪∞

n=1 Vn .
Then: µ(V ) = lı́mn→∞ µ(Vn )
Also Vn ⊂ {x ∈ [0, c] : f > n1 } for any n1 > 0
Then :

1
µ(V ) = lı́m µ(Vn ) ≤ lı́m µ({x ∈ [0, c] : f (x) > }) = lı́m 0 = 0
n→∞ n→∞ n n→∞

So by proposition 19.5.5,4 it is a contradiction.

R R R
Because E f dµ > 0 and E |f |dµ = E f dµ = 0.

11
Case 2
If f < 0 then −f > 0 a.e.
Let W = {x ∈ E : −f > 0} and Wn = {x ∈ E : −f > n1 }.
Then Wn ⊂ Wn+1 and W = ∪∞ n=1 Wn .
And for Markov’s inequality µ({x ∈ [0, c] : −f > m) ≤ m1 [0,c] −f dµ = − m1 ( [0,c] f dµ) =
R R

− m1 0 = 0 for any m > 0

And Wn ⊂ {x ∈ [0, c] : −f > n1 }.
Therefore

1
µ(W ) = lı́m µ(Wn ) ≤ lı́m µ({x ∈ [0, c] : −f > }) = lı́m 0 = 0
n→∞ n→∞ n n→∞

This is a contradiction, therefore f = 0 a.e. on any bounded interval in E ⊂ [0, ∞]

Now take f > 0 for a contradiction.
Now consider the Cover En = {[0, n] : n ∈ N}, with µ(En ) > 0. Note that each
En ⊂ [0, ∞), and are closed interval. Note that each En ⊂ [0, ∞), and are closed
interval, then En are measurable Lebesgue, and we also have to [0, ∞) = ∪∞ n=1 En .
This implies that {x ∈ [0, ∞) : f > 0} ⊂ {x ∈ ∪n∈N [0, n] : f > 0}
Therefore
Z
µ({x ∈ [0, ∞) : f > 0}) ≤ µ({x ∈ ∪n∈N [0, n] : f > 0}) = 1En f dµ
∪∞
n=1 En

Z ∞ Z
X ∞
X
≤ f dµ ≤ f dµ = 0=0
∪∞
n=1 En n=1 En n=1

. This is a contradiction by proposition 19.5.5,4.

For we assume that f > 0 a.e. on [0, ∞) this implies that f > 0 in some En with
positive measure, then its integral over En is greater than 0 and 0 at the same time.
Similarly, if f < 0 a.e., then −f > 0.
Let’s take {En } as the cover of [0, ∞), so Let’s take {En } as a cover of [0, ∞), for
the above and as −f > 0 we have to:
This implies {x ∈ [0, ∞) : −f > 0} ⊂ {x ∈ ∪n∈N [0, n] : −f > 0}
Therefore

12
 Z
µ({x ∈ [0, ∞) : −f > 0}) ≤ µ({x ∈ ∪n∈N [0, n] : −f > 0}) = 1En − f dµ
∪∞
n=1 En

Z ∞ Z
X ∞
X
≤ −f dµ ≤ −f dµ = 0=0
∪∞
n=1 En n=1 En n=1

. This is a contradiction by proposition 19.5.5,4.

For we assume that −f > 0 a.e. on [0, ∞) this implies that −f > 0 in some En with
positive measure, then its integral over En is greater than 0 and 0 at the same time.
Therefore f=0 a.e. on [0, ∞).

13
 Problem 19.22
Recall Tannery’s theorem from Theorem 8.4.5. By considering the measure
space (N, P(N), v) where v is the counting measure, and the doubly indexed
real sequence (am,n ) as a sequence of functions an : N → R where an (m) = am,n ,
prove Tannery’s theorem using the DCT (Dominated Convergence Theorem).

Recall Tannery’s theorem:

Tannery’s Theorem: Consider a doubly indexed real sequence (am,n ) for m, n ∈ N.

Suppose that:

P∞
1. For each n ∈ N, the series sn = m=1 am,n are all convergent.

2. For each m, we have lı́mn→∞ am,n = am , and

3. For each m ∈ N, there exists Mm > 0 such that |am,n | ≤ Mm for all n ∈ N, and
the series ∞
P
m=1 Mm converges.

Then we can interchange the following limits:

∞
X ∞
X ∞
X
lı́m sn = lı́m am,n = lı́m am,n = am .
n→∞ n→∞ n→∞
m=1 m=1 m=1

Solution. From Example 19.5.2, we have that for n ∈ N,

Z ∞
X ∞
X
an dv = a+
n (m) − a−
n (m) (2)
N m=1 m=1

−
It is not difficult to see that 0 ≤ |a+
n (m)| ≤ |an (m)| ≤ Mm = |Mm | and 0 ≤ |an (m)| ≤

14
|an (m)| ≤ Mm = |Mm |.

P∞ P∞
Then, by Corollary 7.5.2, m=1 a+
n (m) and m=1 a−
n (m) converge absolutely, and
therefore, they converge.

Then, by Proposition 7.2.8, we can write 2 as

Z ∞
X ∞
X
an dv = (a+
n (m) − a−
n (m)) = an (m) (2) . (3)
N m=1 m=1

Now, by condition 2:

lı́m an (m) = a(m) for all m ∈ N,

n→∞

that is, an converges pointwise to a.

Moreover, by condition 3, for each n ∈ N, we can define g : N → R+ , g(m) := Mm .

Now, from Example 19.5.2, we have that g ∈ L1 (N) if and only if the series is abso-
lutely convergent. Since

∞
X ∞
X ∞
X
|g(m)| = |Mm | = Mm
m=1 m=1 m=1

15
converges, then g ∈ L1 (N).

Also, by condition 3, we have that |an (m)| ≤ Mm = g(m) for each n ∈ N.

Then, by Theorem 19.6.3, we have that

Z Z
lı́m an (m)dv = a(m)dv,
n→∞ N N

or equivalently, using 3:

∞
X ∞
X
lı́m am,n = am .
n→∞
m=1 m=1

16