REAL ANALYSIS AND MEASURE

THEORY

BY

DR. PRATULANANDA DAS

DEPARTMENT OF MATHEMATICS
JADAVPUR UNIVERSITY
KOLKATA- 700032
WEST BENGAL, INDIA
E-mail : [email protected]
 Chapter 2

Lebesgue Measure

Module 4

Notion of inner measure

1
 2.4. Inner measure

We end this chapter with a brief reference to the notion of inner measure
which was also developed to provide an alternative approach to the introduction
of measurable sets. But we will avoid the detailed discussion as it is not needed
too much for our subsequent development of the subject.

We start by noting that a set of real numbers is open if and only if it can
be expressed as the countable union of open intervals so for a subset E ⊂ R we
can redefine the outer measure of E as

µ∗ (E) = inf{µ(O) : O is any open set containing E}.

Since any non-empty set E of real numbers always contains some closed sets so
we can now consider the following definition.

Definition 2.11. For E ⊂ R we define the inner measure of E as

µ∗ (E) = sup{µ(F ) : F is any closed set contained in E}.

It is clear from the definitions that µ∗ (E) ≤ µ∗ (E) for any E ⊂ R and just
like the outer measure, the inner measure is also monotone i.e. µ∗ (A) ≤ µ∗ (B)
if A ⊂ B.

Theorem 2.12. If E ⊂ R is such that µ∗ (E) < ∞ then E is measurable if and

only if µ∗ (E) = µ∗ (E).

Proof: Suppose first that E is measurable. Then for ε > 0 by Corollary 2.6,
there exist an open set O and a closed set F with F ⊂ E ⊂ O such that
µ(O − F ) < ε. Then

µ∗ (E) ≤ µ(O) < µ(F ) + ε ≤ µ∗ (E) + ε.

Since this is true for any ε > 0 so we have µ∗ (E) ≤ µ∗ (E) and hence we must
have µ∗ (E) = µ∗ (E).

Conversely let µ∗ (E) = µ∗ (E). Let ε > 0 be given. Then there exist an open
set O and a closed set F such that
ε
E ⊂ O and µ(O) < µ∗ (E) +
2
and
ε
F ⊂ E and µ(F ) > µ∗ (E) − .
2
Therefore
µ(O − F ) = µ(O) − µ(F ) < ε.
which shows that E is measurable in view of Corollary 2.6.

2
 It should be noted that the above characterization of measurable sets is
not generally true for sets E with µ∗ (E) = ∞. To see this, consider the set
E = P ∪ [2, ∞) where P is the non-measurable subset of [0, 1) constructed
above. Evidently µ∗ (E) = ∞ = µ∗ (E) but E is not measurable.

Theorem 2.13. If 4 is a bounded open interval containing E then

µ∗ (E) + µ∗ (4 − E) = µ(4).

Proof: Let ε > 0 be given. From the definition of µ∗ (4 − E) we can find a

closed set F ⊂ (4 − E) such that

µ(F ) > µ∗ (4 − E) − ε.

Then G = (4 − E) is an open set and G = (4 − F ) ⊃ (4 − (4 − E)) = E.

Consequently

µ∗ (E) ≤ µ(G) = µ(4) − µ(F ) < µ(4) − µ∗ (4 − E) + ε

i.e.
µ∗ (E) + µ∗ (4 − E) < µ(4) + ε.
Since ε > 0 is arbitrary so

µ∗ (E) + µ∗ (4 − E) ≤ µ(4).

Again from the definition of m∗ (E) we can find a bounded open set G0 ⊃ E
such that
ε
µ(G0 ) < µ∗ (E) + .
3
Write 4 = (A, B) and choose an open interval (a, b) with A < a < A + 3ε and
B − 3ε < b < B. Now write G = (4 ∩ G0 ) ∪ (A, a) ∪ (b, B). Then G is an open
set containing E and further
ε ε
µ(G) ≤ µ(G0 ) + + < µ∗ (E) + ε.
3 3
Take F = 4 − G = [a, b] − G. Clearly F is a closed set and F ⊂ (4 − E). Then

µ∗ (4 − E) ≥ µ(F ) = µ(4) − µ(G) ≥ µ(4) − µ∗ (E) − ε

i.e.
µ∗ (E) + µ∗ (4 − E) ≥ µ(4) − ε.
Since ε > 0 is arbitrary so we have

µ∗ (E) + µ∗ (4 − E) ≥ µ(4).

This along with the opposite implication obtained earlier completes the proof.

3
 In the final result of this section we see that we can extend the above result
for any measurable set.

Theorem 2.14. If B is measurable and E ⊂ B then

µ∗ (E) + µ∗ (B − E) = µ(B).

Proof: Let ε > 0 be given. We can find a closed set F ⊂ E such that

µ(F ) > µ∗ (E) − ε.

Then
µ(B) = µ(F ) + µ(B − F ) > µ∗ (E) − ε + µ∗ (B − E)
using the monotone property of outer measure. Since ε > 0 is arbitrary so we
get
µ∗ (E) + µ∗ (B − E) ≤ µ(B).
To prove the reverse inequality we note that by Corollary 1.4 there is a Gδ -
set G0 containing B − E such that µ(G0 ) = µ∗ (B − E). Setting G = G0 ∩ B we
see that G is a measurable set, B − E ⊂ G ⊂ B and µ(G) = µ∗ (B − E). Since
B − G ⊂ E, so by monotone property of inner measure we have µ∗ (B − G) ≤
µ∗ (E). Therefore

µ(B) = µ(G) + µ(B − G) = µ∗ (B − E) + µ(B − G) ≤ µ∗ (B − E) + µ∗ (E).

Combining the two inequalities we get the desired result.

2.5. Self-check Exercises

Exercise 2.1. If E1 , E2 are measurable then prove that

µ(E1 ∪ E2 ) + µ(E1 ∩ E2 ) = µ(E1 ) + µ(E2 ).

Solution: If µ(E1 ) or µ(E2 ) is ∞ then there is nothing to prove. Let µ(E1 ), µ(E2 ) <
∞. Now
E1 = ((E1 ∪ E2 ) − E2 ) ∪ (E1 ∩ E2 ).
Since the sets on the right hand side are measurable and disjoint so

µ(E1 ) = µ((E1 ∪ E2 ) − E2 ) + µ(E1 ∩ E2 ) = µ(E1 ∪ E2 ) − µ(E2 ) + µ(E1 ∩ E2 )

and so
µ(E1 ∪ E2 ) + µ(E1 ∩ E2 ) = µ(E1 ) + µ(E2 ).

4
Exercise 2.2. Let E be the set of all x ∈ [0, 1] such that the decimal expansion
of x does not contain the digit 3. Show that E has Lebesgue measure zero.

Solution: Let E be the set of x ∈ [0, 1] having no 3 in its decimal expan-

∞
T
sion. Then E = En , where En is the set of x ∈ [0, 1] having no 3 in the
n=1
first n places in its decimal expansion. For each (finite) sequence of the form
s1 , s2 , . . . , sn with each si ∈ {0, 1, . . . , 9}, 1 ≤ i ≤ n, the set of x ∈ [0, 1] with
first n-places of its decimal expansion of the form 0.s1 . . . sn represents an inter-
val of length 10−n , and there are 9n such choices of s1 , . . . , sn with no si =
3.
9 n
So En is the union of 9n intervals each of length 10n . Hence µ(En ) = 10
9 n


and consequently µ(E) ≤ 10 for each n ∈ N. Thus µ(E) = 0.

Exercise 2.3. Let E be the set of all real x such that the decimal expansion of
x contains the digit 3 only a finite number of times. Show that E has Lebesgue
measure zero.

Solution: As in the previous exercise, it can be shown that for each k ∈ Z

the set Xk of x ∈ [k, k + 1] having no 3 after the decimal point
S in its decimal
expansion has Lebesgue measure 0, and hence the set X = Xk of x ∈ R
k∈Z
having no 3 after the decimal point in its decimal expansion has Lebesgue mea-
sure 0. Now let En be the set of x ∈ R having no 3 in its decimal expansion
following the n-th place after the decimal point. In the same way as above we
∞
S
have µ(En ) = 0, and so µ(E) = 0 where E = En is the set of x ∈ R having
n=1
at most finitely many 3’s in its decimal expansion.
Exercise 2.5. Suppose that A ⊆ E ⊆ B ⊆ R where A and B are Lebesgue
measurable sets of finite Lebesgue measure. Show that if µ(A) = µ(B), then E
is Lebesgue measurable.

Solution: Since E = A ∪ (E − A) and A is Lebesgue measurable, it is sufficient

to show that E − A is Lebesgue measurable. Since E − A ⊂ B − A, it follows
that µ∗ (E − A) ≤ µ∗ (B − A) = µ(B − A) = µ(B) − µ(A) = 0 as both A and
B are Lebesgue measurable. Thus E \ A being a Lebesgue measure zero set, is
Lebesgue measurable.

Exercise 2.6. Let E ⊂ R be such that µ(E) = 0. Prove that E c is dense in R.

Solution: Observe that for every open interval I ⊂ R, µ∗ (I) > 0. Since
µ(E) = 0, E cannot contain any open interval as a subset by the monotonicity
of µ. Hence E c ∩ I 6= ∅ for every open interval I ⊂ R. Thus E c is dense in R.

Exercise 2.7. If A ⊂ R has finite Lebesgue measure then prove that there
exists a set B ⊂ A where µ(B) = µ(A)
2 .

5
Solution: For each r ≥ 0 define the function f (r) = µ(A ∩ (−r, r)).. Observe
that for 0 ≤ r < s we have

f (s) − f (r) = µ[(A ∩ (−s, s)) − (A ∩ (−r, r))]

= µ(A ∩ (−s, −r]) + µ(A ∩ [r, s)) ≤ 2(s − r)

which shows that f is uniformly continuous. Further f (0) = 0 and lim f (x) =
x→∞
µ(A). Then by the Intermediate Value Theorem there must exist a t > 0 such
that f (t) = µ(A)
2 and obviously then the set B = A ∩ (−t, t) is the desired subset
of A.

Exercise 2.8. Let µ∗ (A) < ∞. If for each ε > 0 there is a finite collection of
[n
disjoint open intervals {Ik : 1 ≤ k ≤ n} such that µ∗ (A∆( Ik )) < ε, then
k=1
prove that A is measurable.

Solution: Let ε > 0 be given. By our assumption there is a finite collection of

n
[ ε
disjoint open intervals {Ik : 1 ≤ k ≤ n} such that µ∗ (A∆( Ik )) < . Write
3
k=1
n
[
U = Ik . Since µ∗ (A − U ) < 3ε so we can further choose an open set W
k=1
2ε
containing A − U such that µ(W ) < 3 . Let

O = U ∩ W.

Then clearly O is an open set containing A and

ε
µ∗ (O − A) ≤ µ∗ (U − A) + µ∗ (W − A) < + µ(W ) < ε.
3
This proves that A is measurable.

Exercise 2.9. If E is a measurable set with µ∗ (E) < ∞ then for any ε > 0
prove that there is a finite collection of pairwise disjoint open intervals {Jk :
[n
1 ≤ k ≤ n} such that µ(E∆U ) < ε where U = Jk .
k=1
Solution: Let E be measurable. From the definition of outer measure we can
find a countable collection of open intervals {Jk }k∈N for which
n ∞
[ X ε
E⊂ Jk = W and `(Jk ) < µ∗ (E) + .
2
k=1 k=1

6
 M M
X ε [
Choose a M ∈ N such that l(Jk ) < . Write U = Jk . Then
2
k=1 k=1

µ(E∆U ) = µ(E − U ) + µ(U − E) ≤ µ(W − U ) + µ(W − E)

M
X ∞
X
≤ `(Jk ) + [ `(Jk ) − µ(E)]
k=1 k=1
ε ε
≤ + = ε.
2 2
In the place of above collection of open intervals {Jk : 1 ≤ k ≤ n} if we take the
component intervals of the open set U , we get the desired result.

Exercise 2.10. Given two Lebesgue measurable sets E, F ⊂ [0, 1], define
d(E, F ) = µ(E∆F ) (where E∆F = (E ∪ F ) − (E ∩ F )), and E ∼ F if
d(E, F ) = 0. Prove that ∼ is an equivalence relation on the collection of
(Lebesgue) measurable subsets of [0, 1], and d induces a metric on the set of
equivalence classes.

Solution: If E ⊂ [0, 1] is Lebesgue measurable, we have E∆E = φ and hence

d(E, E) = µ(E∆E) = 0 i.e. ∼ is reflexive. Let E, F ⊂ [0, 1] be Lebesgue mea-
surable such that d(E, F ) = 0. Then d(F, E) = µ(F ∆E) = µ(E∆F ) = 0 as ∆
is symmetric and hence ∼ is symmetric. Next let E ∼ F and F ∼ G for some
Lebesgue measurable subsets E, F and G of [0, 1]. Then we have d(E, F ) = 0
and d(F, G) = 0. Since E∆F = (E − F ) ∪ (F − E), d(E, F ) = 0 is equiv-
alent to µ(E − F ) = 0 and µ(F − E) = 0 and similarly for d(F, G). Since
E − G ⊂ (E − F ) ∪ (F − G), we have µ(E − G) ≤ µ(E − F ) + µ(F − G) = 0 and
hence µ(E − G) = 0. By the same argument we have µ(G − E) = 0 and hence
d(E, G) = 0. Hence ∼ is an equivalence relation.

Since E − G ⊂ (E − F ) ∪ (F − G) and G − E ⊂ (G − F ) ∪ (F − E) we have

d(E, G) = µ ((E − G) ∪ (G − E))
≤ µ ((E − F ) ∪ (F − G) ∪ (G − F ) ∪ (F − E))
≤ d(E, F ) + d(F, G).
This proves the triangle inequality. Other conditions are easy to verify.

Summary: In this chapter we have defined the Lebesgue measure and proved
properties of Lebesgue measure and also some properties of measurable sets.
We have also shown how to construct non-measurable sets. Finally we have
defined inner measure.

Following are the main definitions and points to be remembered.

• Let µ∗ be the Lebesgue outer measure on R and Let S be the collection of

Lebesgue measurable sets. Then the restriction µ of µ∗ on S is a non-negative

7
countably additive set function with µ(φ) = 0 which is called the Lebesgue mea-
sure.

• The Lebesgue measure µ is monotone, subtractive, countably sub-additive,

continuous from bellow and continuous from above.

• Following statements are equivalent in R.

(i) E is Lebesgue measurable.

(ii) Given ε > 0 there is an open set O such that E ⊂ O and µ∗ (O − E) < ε.
(iii) There is a Gδ set G ⊃ E such that µ∗ (G − E) = 0.
(iv) Given any ε > 0 there is a closed set F such that F ⊂ E and µ∗ (E−F ) <
ε.
(v) There is a Fσ set F ⊂ E such that µ∗ (E − F ) = 0.

• It is not that all subsets of R are measurable. Using ”Axiom of Choice” we

show the existence of a non-measurable subset of [0, 1). From the construction
it is evident there are many many such sets.

• For E ⊂ R we define the inner measure of E as

µ∗ (E) = sup{µ(F ) : F is any closed set contained in E}.

• If E ⊂ R is such that µ∗ (E) < ∞ then E is measurable if and only if

∗
µ (E) = µ∗ (E).

Acknowledgement: While writing this chapter the author has mainly followed
the excellent books by H.L. Royden et al and R.A. Gordon. The full references
of these books is given in ”Learn More” section.