Alcantara, Franz Joaquin V.

Math 126 WFU 1st Semester; AY 23-24

2021***** Real Analysis Notes

1 Lebesgue Outer Measure

Definition. Let I ≠ ∅ be an open interval. We define the length of I, denoted ℓ(I) as
⎧
⎪∞, if I is unbounded
⎪
ℓ(I) ∶= ⎨ .
⎩∣b − a∣, if I is bounded with endpoints a and b
⎪
⎪

Definition. Let A ⊆ R. We define the (Lebesgue) outer measure, denoted m∗ (A) as

∞ ∞
m∗ (A) ∶= inf { ∑ ℓ(Ik ) ∶ A ⊆ ⋃ Ik }
k=1 k=1

where Ik is an open bounded interval, ∀k ∈ N.

Remarks:

1. For the sake of convenience, we define inf ☆A ∶= m∗ (A). The set ☆A is essentially a
collection of the total sum of lengths of intervals Ik which openly covers A. (i.e. {Ik }
is an open cover of A).

2. We have m∗ (A) ≥ 0, ∀A ⊆ R.

3. If A ⊆ B then any open cover of B is an open cover of A. Immediately, it follows that

☆A ⊇ ☆B , which implies that inf ☆A ≤ inf ☆B . That is m∗ (A) ≤ m∗ (B).

Theorem. The outer measure of the null set is 0. That is m∗ (∅) = 0.

Proof. Note that ∅ ⊆ A, ∀A ⊆ R. Let {Ik } be any collection of open bounded intervals. So
∞ ∞
∅ ⊆ ⋃ Ik Ô⇒ ∑ ℓ(Ik ) ∈ ☆∅ .
k=1 k=1

Hence ∞
0 ≤ m∗ (∅) = inf ☆∅ ≤ ∑ ℓ(Ik ). (1)
k=1

Let ϵ > 0. Take

−ϵ ϵ
Ik = ( k+2
, k+2 ) , ∀k ∈ N.
2 2
Then,
ϵ −ϵ ϵ
ℓ(Ik ) = −( ) = .
2k+2 2k+2 2k+1
Thus
∞ ∞
1 ϵ
∑ ℓ(Ik ) = ϵ ∑ = < ϵ. (2)
k=1 k=1 2k+1 2

By (1) and (2), we have 0 ≤ m∗ (∅) < ϵ, ∀ϵ > 0, which implies m∗ (∅) = 0. ◻

1
Alcantara, Franz Joaquin V. Math 126 WFU 1st Semester; AY 23-24
2021***** Real Analysis Notes

Theorem. The outer measure of any countable set is zero.

Proof. Let A ⊆ R be a countable set.
k=1 . Let ϵ > 0. Define
Case 1: A is infinitely countable. Suppose A = {ak }∞

ϵ ϵ ∞
ak ∈ Ik ∶= (ak − , a k + ) ⊆ ⋃ Ik .
2k+2 2k+2 k=1

∞
It follows that A ⊆ ⋃∞
k=1 Ik (i.e. {Ik }k=1 is an open cover of A). Hence ∑k=1 ℓ(Ik ) ∈ ☆A and so
∞

∞ ∞ 1 ϵ
0 ≤ m∗ (A) = inf ☆A ≤ ∑ ℓ(Ik ) = ϵ ⋃ = < ϵ.
k=1 k=1 2k+1 2

This shows that m∗ (A) = 0.

Case 2: A is finitely countable. Take any infinitely countable set B such that A ⊆ B. By
Case 1, we have

0 ≤ m∗ (A) ≤ m∗ (B) = 0.

In either case, it follows that m∗ (A) = 0. ◻

Theorem. The outer measure of any interval is its length.
Proof. Let I ⊆ R be an interval.
Case 1: I is closed and bounded. Let I = [a, b] where a < b. Let ϵ > 0. Define
ϵ ϵ −ϵ ϵ
⊆ I1 ∶= (a − 2
, b + 2 ) and Ik ∶= ( k+1 , k+1 ) , ∀k ≥ 2
2 2 2 2

It follows that
∞
I ⊆ I1 Ô⇒ I ⊆ ⋃ Ik
k=1
∞
Ô⇒ ∑ ℓ(Ik ) ∈ ☆I
k=1
∞
Ô⇒ m∗ (I) ≤ ∑ ℓ(Ik )
k=1

where
∞
ϵ ϵ
∑ ℓ(Ik ) = (b − a + ) +
k=1 2 2
= b − a + ϵ.

Hence

m∗ (I) ≤ b − a + ϵ, ∀ϵ > 0.

This shows that m∗ (I) ≤ b − a.

2
Alcantara, Franz Joaquin V. Math 126 WFU 1st Semester; AY 23-24
2021***** Real Analysis Notes

Now since I = [a, b] is closed and bounded and {Ik } is an open cover of I, by the Heine-
Borel Theorem, I has a finite subcover. That is
N
I ⊆ ⋃ Ik , ∃N ∈ N.
k=1

It follows that ∃(a1 , b1 ) ∈ {Ik }∞

k=1 such that a ∈ (a1 , b1 ) Ô⇒ a1 < a < b1 . If b ≤ b1 , then

∞
b − a ≤ b1 − a1 ≤ ∑ ℓ(Ik ).
k=1

That is b − a is a lower bound for ☆I and so m∗ (I) ≥ b − a.

On the other hand, if b1 < b, we consider another interval (a2 , b2 ) ∈ {Ik }N
k=1 in which b ∈
(a2 , b2 ) Ô⇒ a2 < b ≤ b2 . Which shows that
∞
∑ ℓ(Ik ) ≥ (b2 − a2 ) + (b1 − a1 )
k=1
= b2 + (b1 − a2 ) − a1
≥ b2 − a1 (why?)
≥ b − a.

Otherwise, if b > b2 , we continue the same process until it terminates. Then, we have a
subcollection {(an , bn )}m
n=1 of {Ik }k=1 such that a1 < a and
∞

bi ∈ (ai+1 , bi+1 ), ∀i = 1, 2, . . . , m − 1

with bm ≥ b, ∃m ≤ N . This means that

∞
∑ ℓ(Ik ) ≥ (bm − am ) + (bm−1 − am−1 ) + ⋯ + (b2 − a1 ) + (b1 − a1 )
k=1
= bm + (bm−1 − am ) + (bm−2 − am−1 ) + ⋯ + (b1 − a2 ) − a1
≥ b m − a1
≥ b − a.

This shows that b − a is a lower bound of ☆I . That is b − a ≤ m∗ (I). Combined from earlier
results, we have m∗ (I) = b − a.

3
Alcantara, Franz Joaquin V. Math 126 WFU 1st Semester; AY 23-24
2021***** Real Analysis Notes

Case 2: I is any bounded set.

Proof. Let I be any bounded set. Define J1 = [a + 4ϵ , b − 4ϵ ] and J2 = [a − 4ϵ , b + 4ϵ ]. By
construction, it follows that

J1 ⊆ I ⊆ J2 .

By the monotonicity of m∗ , we have

m∗ (J1 ) ≤ m∗ (I) ≤ m∗ (J2 ). (3)

From Case 1,

m∗ (J1 ) = ℓ(J1 )
ϵ
=b−a−
2
ϵ
= ℓ(I) − .
2
ϵ
Similarly, m∗ (J2 ) = ℓ(I) − . By (3),
2
ϵ ϵ
ℓ(I) − ≤ m∗ (I) ≤ ℓ(I) +
2 2
ϵ ϵ
Ô⇒ − ≤ m∗ (I) − ℓ(I) ≤
2 2
ϵ
Ô⇒ ∣m (I) − ℓ(I)∣ ≤ < ϵ
∗
2
Therefore m∗ (I) = ℓ(I). ◻
Case 3: I is unbounded.
Proof. Let I be an unbounded interval.
Note: If I is an unbounded interval, it contains a subset with outer measure n, ∀n ∈ N. For
instance, if I = (a, ∞), then we have (a, a + n) ⊆ I and m∗ (a, a + n) = n by Case 2.
By the monotonicity of m∗

n ≤ m∗ (I), ∀n ∈ N

Taking the limit as n → ∞, we obtain

m∗ (I) = ∞ = ℓ(I).

4
Alcantara, Franz Joaquin V. Math 126 WFU 1st Semester; AY 23-24
2021***** Real Analysis Notes

Proposition.
1.) The outer measure of m∗ is translation invariant. That is, for any set A ⊆ R and y ∈ R,

m∗ (A + y) = m∗ (A).

Proof. HOMEWORK; due on Wed.

2.) The outer measure m∗ is countably subadditive. That is, if {Ek }∞
k=1 is any countable
collection of sets (may be disjoint or not), then
∞ ∞
m∗ ( ⋃ Ek ) ≤ ∑ m∗ (Ek )
k=1 k=1

Proof. Goal:
∞ ∞
m∗ ( ⋃ Ek ) ≤ ∑ m∗ (Ek ) + ϵ, ∀ϵ > 0.
k=1 k=1

Case 1: One of the Ek ’s has an infinite outer measure. Then,

∞ ∞
m∗ ( ⋃ Ek ) ≤ ∑ m∗ (Ek ) = ∞
k=1 k=1

and we are done.

Case 2: All of Ek ’s has a finite measure. That is m∗ (Ek ), ∀k ∈ N. Let ϵ > 0 and fix
k ∈ N. Note that

m∗ (Ek ) = inf ☆Ek .

∞
By the characterization of inf ☆Ek , ∃ ∑ ℓ(Iik ) ∈ ☆Ek such that
i=1

∞
ϵ
∑ ℓ(Iik ) < m∗ (Ek ) + .
i=1 2k

By the definition of ☆Ek , {Iik } is a collection of open and bounded intervals such that
∞
Ek ⊆ ⋃ Iik .
i=1

Consider the collection {Iik }i=1,k≤∞ .

By 2,
∞ ∞ ∞
E ∶= ⋃ Ek ⊆ ⋃ ⋃ Iik
k=1 k=1 i=1

Thus,
∞ ∞
∑ ∑ ℓ(Iik ) ∈ ☆E
k=1 i=1

5
Alcantara, Franz Joaquin V. Math 126 WFU 1st Semester; AY 23-24
2021***** Real Analysis Notes

By the definition of m∗ (E),

∞ ∞
m∗ (E) ≤ ∑ ∑ ℓ(Iik )
k=1 i=1
∞
ϵ
= ∑ (m∗ (Ek ) + ) (by (3))
k=1 2k
∞ ∞
ϵ
= ∑ m∗ (Ek ) + ∑ k
k=1 k=1 2
∞
= ∑ m∗ (Ek ) + ϵ.
k=1

Letting ϵ Ð→ 0, we have
∞ ∞
m∗ ( ⋃ Ek ) ≤ ∑ m∗ (Ek ).
k=1 k=1
◻
Corollary. Finite Subadditivity. If {Ek }nk=1 is any finite collection of sets (disjoint or not),
then
n n
m∗ ( ⋃ Ek ) ≤ ∑ m∗ (Ek ).
k=1 k=1

Proof. Consider k=1 ,

{Fk }∞ where Fk = Ek , ∀k1, 2, . . . , n and Fk = ∅, ∀k > n. (Continue)
Definition. A set E is said to be measurable provided for any set A, we have
m∗ (A) = m∗ (A ∩ E) + m∗ (A ∩ E c )
where E c = R/E.
Remarks:
1.) Let E and F be disjoint sets with being measurable. Then
m∗ (E ∪ F ) = m∗ (E) + m∗ (F ). (4)
Proof. From the last definition, we choose A = E ∪ F . Then
m∗ (E ∪ F ) = m∗ ((E ∪ F ) ∩ E) + m∗ (E ∪ F ) ∩ E c )
= m∗ (E) + m∗ (F ).
◻
2.) To show that E is measurable, it suffices to show that
m∗ (A) ≥ m∗ (A ∩ E) + m∗ (A ∩ E c ),
for any A ⊆ R. Indeed, note that
A = (A ∩ E) ∪ (A ∩ E c ).
By finite subadditivity,
m∗ (A) ≤ m∗ (A ∩ E) + m∗ (A ∩ E c ).

6
Alcantara, Franz Joaquin V. Math 126 WFU 1st Semester; AY 23-24
2021***** Real Analysis Notes

3.) A set E is measurable iff. E c is measurable.

4.) (Excision Property) Let A, B ⊆ R such that B ⊆ A. If B is a measurable set of finite

outer measure, then

m∗ (A/B) = m∗ (A) − m∗ (B)

Proof. Since B is measurable, it satisfies the last definition. Then,

m∗ (A) = m∗ (A ∩ B) + m∗ (A ∩ B c )
= m∗ (B) + m∗ (A/B)
Ô⇒ m∗ (A/B) = m∗ (A) − m∗ (B).

Note: An outer measure in an equation has to be finite for it to be transposable.