REAL ANALYSIS AND MEASURE

THEORY

BY

DR. PRATULANANDA DAS

DEPARTMENT OF MATHEMATICS
JADAVPUR UNIVERSITY
KOLKATA- 700032
WEST BENGAL, INDIA
E-mail : pratulananda@[Link]
 Chapter 1

(Equivalent to three hours of

lectures/modules)

Lebesgue Outer Measure

1
 1.1. Introduction

We start by recalling that two sets of real numbers are said to be equivalent
if there exists a bijective mapping between these two sets. This gives rise to the
notion of cardinality of sets of real numbers which in fact measures the number
of elements in sets. It is well known that any two open intervals (a, b) and (c, d)
are equivalent. But do they look same when we plot them on the extended real
line ? the answer is ”NO” because when we draw an interval, there is one very
prominent characteristic, namely, the length of the interval. the notion of length
of an interval is frequently used in differentiation and Riemann integration in
undergraduate level. We define the length of a bounded interval I (open, closed
or half open) with end points a and b where a < b as the number b − a and we
write `(I) = b − a. The length of an unbounded interval I i.e. an interval whose
left end point is −∞ or the right end point is ∞ is defined as infinity and we
generally write `(I) = ∞. Now the natural question arises whether it is possible
to extend the notion of ”length” of intervals to arbitrary sets of real numbers ?
What should be the length of an open set or a closed set ? Observe that it is
quite natural to define the length of an open set to be the sum of the lengths
of its component intervals of which it is composed of. While the length of a
bounded closed set F can easily be defined as B − A − `(O) where F ⊂ (A, B)
and O = (A, B) − F . But the class of open sets and bounded closed sets is
too restricted for our purpose and we would like to know the lengths of more
complex sets, for example what should be the length of the set of irrational
numbers in (2, 4)? More importantly we would like to know when is it possible
to calculate the ”length” of an arbitrary set of real numbers ? The theory of
Lebesgue measure deals with these questions and is one of the most natural
approaches to answer these types of problems.

Before we move forward we should think what basic properties of a measure

like function µ are to be expected. Based on our understanding of the concept
of length we expect the Lebesgue measure µ to have the following properties:

(i) µ should be non-negative i.e. µ(A) > 0 for all A ⊂ R for which it is
defined.

(ii) µ should be monotone i.e. if A, B are two sets of real numbers whose
Lebesgue measures are defined and A ⊂ B then µ(A) ≤ µ(B).

(iii) For an interval I, µ(I) should be equal to the length of the interval.

(iv) For A ⊂ R and y ∈ R if we define A+y = {a+y : a ∈ A}, the translation

of the set A then µ(A + y) = µ(A) i.e. µ should be translation invariant.

(v) For two disjoint sets A, B of real numbers, µ(A ∪ B) = µ(A) + µ(B).
More precisely µ should be countably additive i.e. if {An }n∈N is a sequence of

2
 ∞
[ ∞
X
pairwise disjoint sets of real numbers then µ( An ) = µ(An ).
n=1 n=1
In this chapter we start by defining the notion of Lebesgue outer measure
and show that it has most of the above mentioned basic properties. Then we
define measurable sets as a special class of sets which play the most important
role in introducing the notion of Lebesgue measure. We will see that the collec-
tion of those sets of real numbers which have Lebesgue measure or are so called
”Lebesgue measurable” is actually quite a large class as it forms a σ-algebra (
A collection S of sets of real numbers is called a σ-algebra if it is closed under
difference and countable union and also contains R).

1.2. Lebesgue outer measure

Note that any set A of real numbers can be covered by a countable number
of open intervals. Since length of an open interval is a positive number and the
sum of the lengths of these open intervals is uniquely determined independent
of the order of the terms, this number can be thought of as an approximate
measure of the set A. obviously we are interested to know the best possible ap-
proximation and this gives rise to the notion of Lebesgue outer measure defined
below.

Definition 1.1. For any set A of real numbers the Lebesgue outer measure of
A is defined by
X
µ∗ (A) = inf{ `(In ) : {In }n∈N is a countable collection of open intervals covering A}.
n

Theorem 1.2.

(i) 0 ≤ µ∗ (E) ≤ ∞ for any E ⊂ R.

(ii) µ∗ (φ) = 0.

(iii) A ⊂ B implies µ∗ (A) ≤ µ∗ (B) i.e. Lebesgue outer measure µ∗ is mono-

tone.

(iv) Lebesgue outer measure µ∗ is countably X i.e. if {En }n∈N is

[sub-additive
a countable collection of subsets of R then µ∗ ( En ) ≤ µ∗ (En ).
n n
(v) For any set A ⊂ R, x ∈ R we have µ∗ (A + x) = µ∗ (A) i.e. Lebesgue
outer measure µ∗ is translation invariant.

(vi) The Lebesgue outer measure of an interval is equal to its length.

3
Proof :
Proofs of (i) and (ii) are straightforward.

(iii) the proof follows from the simple observation that any cover of B by
open intervals is also a cover of A.

(iv) If µ∗ (En ) = ∞ for some n then there is nothing to prove. So let

∗
µ (En ) < ∞ ∀ n. Let ε > 0 be given. We can find a countable collection of
open intervals {In,i }i∈N such that
[ X ε
En ⊂ In,i and `(In,i ) < µ∗ (En ) + n .
i i
2

Now {In,i }n,i is again a countable collection of open intervals such that
[ [
En ⊂ In,i .
n n,i

Then
[ X XX X ε X
µ∗ ( En ) ≤ `(In,i ) = ( `(In,i )) ≤ (µ∗ (En ) + n ) = µ∗ (En ) + ε.
n n,i n i n
2 n

Since ε > 0 is arbitrary so,

[ X
µ∗ ( En ) ≤ µ∗ (En ).
n n

(v) Let ε > 0 be given. Now we can find a countable collection of open
intervals {In }n∈N such that
[ X
A⊂ In and `(In ) < µ∗ (A) + ε.
n n
[
But clearly A + x ⊂ (In + x) and so
n
X X
µ∗ (A + x) ≤ `(In + x) = `(In ) < µ∗ (A) + ε.
n n

Since ε > 0 is arbitrary so it readily follows that µ∗ (A + x) ≤ µ∗ (A). But

A = (A + x) − x and so µ∗ (A) ≤ µ∗ (A + x). So µ∗ (A) = µ∗ (A + x).

(vi) First let us consider a closed bounded interval I = [a, b]. We want to
show that µ∗ (I) = b − a. Let ε > 0 be given. Since I = [a, b] ⊂ (a − ε, b + ε) so
from the definition of µ∗ (I) it readily follows that

µ∗ (I) ≤ b − a + 2ε.

4
Since this is true for any ε > 0 so we have

µ∗ (I) ≤ b − a.

We now want to show that µ∗ (I) ≥ b − a. In order to prove this assertion,

we observe that the assertion will follow if we can
X show that for any countable
collection of open intervals {In }n∈N covering I, `(In ) ≥ b−a. Now since [a, b]
n
is compact so every open cover of I = [a, b] has a finite subcover. Further the
P of the lengths of intervals from a P
sum sub-collection of {In }n∈N can not exceed
l(In ) and so it is sufficient to prove l(In ) ≥ b − a for a finite collection {In }
of open intervals covering [a, b].

Since a ∈ [a, b], so there exists a member of {In }n∈N , say (a1 , b1 ), such
that a1 < a < b1 . If b1 > b then [a, b] ⊂ (a1 , b1 ) and the proof is complete.
If not, then b1 ∈ [a, b]. Then there exists another member (a2 , b2 ) (say) with
a2 < b1 < b2 . If b2 > b then the proof is finished. If not, then we proceed as
above. In this way we get members (a1 , b1 ), (a2 , b2 ), ....... from {In }n∈N such
that ai+1 < b < bi+1 . Since the collection {In } is finite, so this process must
stop after finite number of steps. But note that if it stops after k steps then we
must have ak < b < bk . Now clearly

X k
X
`(In ) ≥ (bi − ai )
i=1
= (bk − ak ) + (bk−1 − ak−1 ) + ..... + (b1 − a1 )
= bk − (ak − bk−1 ) − (ak−1 − bk−2 ) − ....(a2 − b1 ) − a1
≥ bk − a1
≥ b − a.

Thus we have µ∗ (I) ≥ b − a. Therefore µ∗ (I) = (b − a).

Next let I be any finite interval. Let ε > 0 be given. Choose a closed interval
J ⊂ I such that `(J) > `(I) − ε. Then
¯ = `(I)
`(I) − ε < `(J) = µ∗ (J) ≤ µ∗ (I) ≤ µ∗ (I) ¯ = `(I).

Again since this is true for any ε > 0 hence we get `(I) = µ∗ (I).

Finally let I be an infinite interval. For any G > 0, we can find a closed
interval 4 ⊂ I with `(4) > G. Then m∗ (I) ≥ m∗ (4) = `(4) > G. Since this
is true for any G > 0 so we must have m∗ (I) = ∞ = `(I). This completes the
proof of the assertion.

Corollary 1.3. If A is countable then µ∗ (A) = 0.

5
Proof: It is easy to prove that the Lebesgue outer measure of a singleton is
zero and then the result follows from countable sub-additivity of Lebesgue outer
measure µ∗ in view of the fact that a countable set is only the union of a count-
able number of singletons.

Recall that G ⊂ R is called a Gδ set if it can be expressed as the intersection

of a countable number of open sets. Similarly a set F ⊂ R is called a Fσ set if
it can be expressed as the union of a countable number of closed sets.

Corollary 1.4. Let A ⊂ R be an arbitrary set and ε > 0 be given. Then there
exists an open set O ⊃ A such that µ∗ (O) ≤ µ∗ (A) + ε. There is a Gδ -set G ⊃ A
such that µ∗ (G) = µ∗ (A).

Proof : Let ε > 0 be given. As before we [ can chooseX a countable collection

of open intervals {In }n∈N such that A ⊂ In and `(In ) < µ∗ (A) + ε. Let
[ n n
O= In . Then A ⊂ O where O is an open set. Now by countable subaddi-
n X X
tivity of µ∗ , µ∗ (O) ≤ µ∗ (In ) = `(In ) < µ∗ (A) + ε.
n n

Now for n ∈ N there exists an open set Gn such that µ∗ (Gn ) < µ∗ (A) + n1 .
\
Let G = Gn . Then G is a Gδ -set containing A. Also µ∗ (A) ≤ µ∗ (G) ≤
n
µ∗ (A) + 1
n ∀ n ∈ N and so taking limit n → ∞ we observe that

µ∗ (G) = µ∗ (A).

In the statement of Corollary 1.4 strict inequality holds if µ∗ (A) is finite.

1.3. Lebesgue measurable sets

From the discussions above we can see that the Lebesgue outer measure is
quite satisfactory in the sense that it can be defined for all sets of real num-
bers and satisfies all the desired properties stated at the start of this module
but unfortunately it does not have that important property (v) (this is still not
obvious and we will have to wait until the next module to produce a counter
example). As we had declared at the start, we now try to restrict Lebesgue
outer measure to a special class of sets to make it countably additive. For this
we now introduce the following definition (due to Caretheodory).

Definition 1.5. A set E of real numbers is called Lebesgue measurable if for

any A ⊂ R,
µ∗ (A) = µ∗ (A ∩ E) + µ∗ (A ∩ E c )

6
where E c stands for the complement of E.

Since A = (A ∩ E) ∪ (A ∩ E c ), so by countable sub-additivity of µ∗ we have

µ (A) ≤ µ∗ (A ∩ E) + µ∗ (A ∩ E c ). So E is in fact Lebesgue measurable if for any
∗

A ⊂ R , µ∗ (A) ≥ µ∗ (A ∩ E) + µ∗ (A ∩ E c ). Though the definition may seem to be

very abstract at a first glance, we will see that it serves our purpose. It is useful
to note that a set E is Lebesgue measurable when it has the special property
that it divides any set A into two disjoint parts A ∩ E and A ∩ E c in such a way
that the Lebesgue outer measure of A can be obtained by adding the Lebesgue
outer measures of the two disjoint parts. from the definition it is also clear that
E is Lebesgue measurable if and only if E c is Lebesgue measurable. From now
on we will refer to Lebesgue measurable sets as only measurable sets.

The collection of measurable sets has the following properties.

Lemma 1.6. R and φ are measurable.

Lemma 1.7. Let E ⊂ R. If µ∗ (E) = 0 then E is measurable.

Proof : Let A ⊂ R. Clearly µ∗ (A ∩ E) ≤ µ∗ (E) = 0 and so µ∗ (A ∩ E) = 0.

Again A ∩ E c ⊂ A ⇒ µ∗ (A ∩ E c ) ≤ µ∗ (A). Since A = A ∩ R = A ∩ (E ∪ E c ) =
(A ∩ E) ∪ (A ∩ E c ), so by sub-additivity of Lebesgue outer measure we have,

µ∗ (A) ≤ µ∗ (A ∩ E) + µ∗ (A ∩ E c ) = µ∗ (A ∩ E c ) ≤ µ∗ (A)

and so µ∗ (A) = µ∗ (A ∩ E c ). This shows that E is measurable.

Corollary 1.8. Every countable set is measurable.

Lemma 1.9. If E and F are measurable sets then so are E ∪ F and E − F.

Proof: Let A ⊂ R. Since E is measurable so

µ∗ (A) = µ∗ (A ∩ E) + µ∗ (A ∩ E c ).

Again since F is measurable so

µ∗ (A ∩ E c ) = µ∗ (A ∩ E c ∩ F ) + µ∗ (A ∩ E c ∩ F c ).

So we get

µ∗ (A) = µ∗ (A ∩ E) + µ∗ (A ∩ E c ∩ F ) + µ∗ (A ∩ E c ∩ F c )
≥ µ∗ (A ∩ E c ∩ F c ) + µ∗ ((A ∩ E) ∪ (A ∩ E c ∩ F ))
= µ∗ (A ∩ E c ∩ F c ) + µ∗ (A ∩ (E ∪ (E c ∩ F )))
= µ∗ (A ∩ (E ∪ F )) + µ∗ (A ∩ (E c ∩ F c ))
= µ∗ (A ∩ (E ∪ F )) + µ∗ (A ∩ (E ∪ F )c ).

7
So E ∪ F is measurable.

Noting that E − F = E ∩ F c , by similar arguments we can show that E − F

is also measurable.

Corollary 1.10. If E, F are measurable then E ∩ F and E4F are also mea-
surable sets where as usual 4 denotes the symmetric difference.

Lemma 1.11. If E1 , E2 , ...., En are n pairwise disjoint measurable sets then for
any A ⊂ R
[n Xn
µ∗ (A ∩ ( Ei )) = µ∗ (A ∩ Ei ).
i=1 i=1

Proof: We will prove the result by mathematical induction. Obviously the

result is true for n = 1. Now we assume that the result is true for n = k i.e.
k
[ k
X
µ∗ (A ∩ ( Ei )) = µ∗ (A ∩ Ei ).
i=1 i=1

Now since Ek+1 is measurable so

k+1
[ k+1
[ k+1
[
µ∗ (A ∩ ( Ei )) = µ∗ (A ∩ ( Ei ) ∩ Ek+1 ) + µ∗ (A ∩ ( c
Ei ) ∩ Ek+1 )
i=1 i=1 i=1
k
[
= µ∗ (A ∩ Ek+1 ) + µ∗ (A ∩ ( Ei ))
i=1
k+1
X
= µ∗ (A ∩ Ei ).
i=1

So the result is true for k + 1 and hence the result is true for any n ∈ N.

Lemma 1.12. If {En }n∈N is a sequence of pairwise disjoint measurable sets

then,

(i) for any A ⊂ R,

∞
[ ∞
X
µ∗ (A ∩ ( Ei )) = µ∗ (A ∩ Ei ).
i=1 i=1

∞
[
(ii)E = Ei is measurable.
i=1

8
Proof: (i) We have
n
X n
[ ∞
[
∗ ∗ ∗
µ (A ∩ Ei ) = µ (A ∩ ( Ei )) ≤ µ (A ∩ ( Ei )).
i=1 i=1 i=1

Since this is true for any n ∈ N so

∞
X ∞
[
µ∗ (A ∩ Ei ) ≤ µ∗ (A ∩ ( Ei )).
i=1 i=1

Obviously
∞
[ ∞
X
µ∗ (A ∩ ( Ei )) ≤ µ∗ (A ∩ Ei ).
i=1 i=1

Therefore
∞
[ ∞
X
µ∗ (A ∩ ( Ei )) = µ∗ (A ∩ Ei ).
i=1 i=1

n
[
(ii) Let A ⊂ R. We write Fn = Ei . Then Fn is measurable for each n ∈ N
i=1
and Fn ⊂ E. Now we have

µ∗ (A) = µ∗ (A ∩ Fn ) + µ∗ (A ∩ Fnc )
Xn
= µ∗ (A ∩ Ei ) + µ∗ (A ∩ Fnc )
i=1
n
X
≥ µ∗ (A ∩ Ei ) + µ∗ (A ∩ E c ).
i=1

Since this is true for any n ∈ N so taking limit n → ∞ we have

∞
X
µ∗ (A) ≥ µ∗ (A ∩ Ei ) + µ∗ (A ∩ E c ) = µ∗ (A ∩ E) + µ∗ (A ∩ E c ).
i=1

∞
[
This proves that E = Ei is measurable.
i=1
Lemma 1.13. Countable union of measurable sets is measurable.

Proof : Let {En }n∈N be an arbitrary sequence of measurable sets and let E =
∞
[ n−1
[
Ei . Define D1 = E1 , D2 = (E2 − E1 ), ...., Dn = En − ( Ei ). Then {Dn }n∈N
i=1 i=1
∞
[ ∞
[
is a sequence of pairwise disjoint measurable sets and E = Ei = Di . Hence
i=1 i=1

9
by Lemma 1.12 (ii) E is measurable.

Corollary 1.14. Countable intersection of measurable sets is measurable.

Theorem 1.15. The collection S of all measurable sets forms a σ-algebra.

Proof: That S forms a σ-algebra follows from Lemmas 1.6, 1.9 and 1.13.

Lemma 1.16. (a, ∞) is measurable.

Proof: Let A ⊂ R. Let I1 = (a, ∞) and I2 = (−∞, a]. We have to show

µ∗ (A ∩ I1 ) + µ∗ (A ∩ I2 ) ≤ µ∗ (A). Without any loss of generality we assume that
µ∗ (A) < ∞. Let ε > 0 be given. Then we can find a countable collection of
open intervals {Jn }n∈N such that
∞
[ ∞
X
A⊂ Jn and `(Jn ) < µ∗ (A) + ε.
n=1 n=1

Writing Jn0 = Jn ∩I1 and Jn00 = Jn ∩I2 , we observe that Jn0 , Jn00 are disjoint. Jn =
Jn0 ∪ Jn00 and they are either empty or an interval. Then `(Jn ) = `(Jn0 ) + `(Jn00 ).
So
X∞ ∞
X ∞
X
µ∗ (A) + ε > `(Jn ) = `(Jn0 ) + `(Jn00 ) =
n=1 n=1 n=1
∞
X ∞
X
µ∗ (Jn0 ) + µ∗ (Jn00 ) ≥ µ∗ (A ∩ I1 ) + µ∗ (A ∩ I2 ).
n=1 n=1

Since ε > 0 is arbitrary so we have

µ∗ (A) ≥ µ∗ (A ∩ I1 ) + µ∗ (A ∩ I2 ).

This shows that (a, ∞) is measurable.

Theorem 1.17. Every open set and every closed set in R is measurable. In
particular every Borel set is measurable where Borel sets are the members of
the σ-algebra generated by the class of all open sets (for formal definition see
Module 9).

Proof: Since (a, ∞) is measurable so its complement (−∞, a] is also mea-

surable for any a ∈ R. Then (a, b] = (−∞, b] ∩ (a, ∞) is measurable where
a < b. Therefore any open interval (a, b) = (a, b] − {b} is also measurable where
a < b. Consequently any open set being countable union of open intervals is also
measurable and their complements, namely the closed sets are also measurable.
Finally the measurablity of Borel sets follows from the fact that S is a σ-algebra
containing the class of open sets and the collection of Borel sets is the smallest
σ-algebra containing open sets.

10
Remark 1.18. We will now give an example of an uncountable perfect nowhere
dense measurable set whose measure is zero. Let D1 = [0, 1], D2 = [0, 13 ] ∪
[ 23 , 1], D3 = [0, 19 ] ∪ [ 29 , 13 ] ∪ [ 23 , 79 ] ∪ [ 89 , 1], ....... In each step we divide the intervals
∞
\
into three equal parts and eliminate the middle open interval. Let C = Dn . C
n=1
is called the Cantor set. C is clearly measurable and µ(C) ≤ ( 23 )n . So µ(C) = 0.

That C is uncountable follows from the fact that every point of C has a
unique representation of the form
r1 r2 rn
x= + 2 + .... + n + . . .
3 3 3
where ri = 0 or 2. Now recall that any element a of [0, 1] has a unique binary
representation of the form a = s21 + 2s22 + .... + 2snn + . . . where si = 0 or 1. Then
defining a map f : [0, 1] → C by
2s1 2s2 2sn
f (a) = + 2 + .... + n + . . .
3 3 3
we observe that f is surjective which shows that C is uncountable. Note that
C is also perfect and nowhere dense.

We end this section showing that measurability remains invariant under

translation.

Lemma 1.19. If E ⊂ R is any measurable set then the set (E + x) = {y + x :

y ∈ E} for every x ∈ R is measurable and µ(E + x) = µ(E).

Proof: Let A ⊂ R. Since E is measurable so

µ∗ (A) = µ∗ (A ∩ E) + µ∗ (A ∩ E c ).
Now
µ∗ (A ∩ (E + x)) + µ∗ (A ∩ (E + x)c ) = µ∗ ((A − x) ∩ E) + x) + µ∗ ((A − x) ∩ E c ) + x)
= µ∗ ((A − x) ∩ E) + µ∗ ((A − x) ∩ E c )
= µ∗ (A − x) = µ∗ (A).
So (E + x) is measurable. The second assertion readily follows from the fact
that Lebesgue outer measure remains invariant under translation.

1.4. Self-check Exercises

Exercise 1.1. Let E be the set of rational numbers in the interval [0, 1]
and {In }kn=1 be a finite collection of open intervals that covers E. Prove that
Pk
`(In ) ≥ 1.
n=1

11
Solution: Since the rationals are dense in the real line, we have E = [0, 1]. By
k k
In . By the finite sub-additivity of µ∗ ,
S S
the property of closure E ⊂ In =
n=1 n=1
we have

1 = µ∗ ([0, 1]) = µ∗ (E)

k
X k
X
≤ µ∗ (In ) = `(In )
n=1 n=1
k
X
= `(In ).
n=1

Exercise 1.2. Show that if a set E ⊂ R has positive Lebesgue outer measure,
then there is a bounded subset of E that also has positive Lebesgue outer mea-
sure.

Solution: We prove by contradiction. Suppose that every bounded subset

of E has Lebesgue outer measure zero. Define Ik = [k, k + 1) for k ∈ Z.
Clearly {Ik }k∈Z is a countable collection of disjoint bounded intervals that de-
composes R. We canSthen decompose E as a countable union of bounded sub-
sets of E i.e. E = (E ∩ Ik ). By our hypothesis µ∗ (E ∩ Ik ) = 0 and hence
 k∈Z 
0 < µ∗ (E) = µ∗
S P ∗
E ∩ Ik ≤ µ (E ∩ Ik ) = 0, a contradiction.
k∈Z k∈Z

Exercise 1.3. Let f : R → R be a continuous function and B be a Borel set.

Show that f −1 (B) is a Borel set.

Solution: Let B denote the Borel σ-algebra over the real line. Define C = {E ⊂
R : f −1 (E) ∈ B}. Since f is continuous, C contains all open subsets of R. We
first show that C is a σ-algebra.

Let E ∈ C. Now f −1 (R \ E) = R \ f −1 (E) ∈ B as f −1 (E) ∈ B and B is a

σ-algebra. Hence E c = R \ E ∈ C. Next let {Ek }k∈N be a collection of sets in
∞ ∞ ∞
C. Similarly we have f −1 f −1 (Ek ) ∈ B and hence
S S S
Ek = Ek ∈ C.
k=1 k=1 k=1
Hence C is a σ-algebra containing open sets and since B is the smallest σ-algebra
containing open sets, we must have B ⊂ C.

Thus if B is a Borel set i.e. B ∈ B then B ∈ C and hence f −1 (B) ∈ B is a

Borel set.

Exercise 1.4. Let {fn }∞ n=1 be a sequence of continuous real-valued functions

of R, and let F be the set of x ∈ R such that fn (x) converges to a limit in R as
n → ∞. Show that F is a Borel set.

12
Solution: Recall that a Borel set is a set belonging to the σ-algebra B gen-
erated by all intervals of the form [a, b) with a, b ∈ R. Note that any open
set in R is a countable union of intervals of this form and hence is a Borel
set. Now F is the set of all x ∈ R such that, for all k ∈ N there exists a
n0 ∈ N such that for all m > n0 , n > n0 we have |fm (x) − fn (x)| < k1 . Thus
∞ S ∞
(k) (k)
Em,n where Em,n = {x : |fm (x) − fn (x)| < k1 }. Now
T T T
F =
k=1 n0 =1 m>n0 n>n0
(k) (k)
Em,n is open since fm − fn is continuous, hence Em,n ∈ B, and as B is a σ-
algebra it follows that F ∈ B i.e. F is a Borel set.

Exercise 1.5. Let E ⊂ R be a Lebesgue measurable set of finite measure.

Define a function ϕE : R → R by ϕE (x) = µ (E ∩ (−∞, x]) for all x ∈ R. Prove
that
(a) ϕE is an increasing function on R.
(b) ϕE satisfies the Lipschitz condition on R i.e. |ϕE (x) − ϕE (y)| ≤ |x − y|,
for all x, y ∈ R.

Solution:
(a) Let x, y ∈ R be such that x < y. Now clearly E ∩ (−∞, x] ⊂ E ∩ (−∞, y]
and hence

ϕE (x) = m (E ∩ (−∞, x]) ≤ m (E ∩ (−∞, y]) = ϕE (y).

Hence ϕE is an increasing function on R.

(b) Let x, y ∈ R be such that x < y. Now

E ∩ (x, y] = (E ∩ (−∞, y]) \ (E ∩ (−∞, x])

and hence

µ (E ∩ (x, y]) = µ (E ∩ (−∞, y]) − µ (E ∩ (−∞, x])

as all the sets are Lebesgue measurable. Consequently we have

|ϕE (x) − ϕE (y)| = ϕE (y) − ϕE (x)

= µ (E ∩ (−∞, y]) − µ (E ∩ (−∞, x])
= µ (E ∩ (x, y]) ≤ µ((x, y]) = y − x = |x − y|.

Exercise 1.6. If the set A has Lebesgue outer measure zero then prove that
the set {x2 : x ∈ A} also has outer measure zero.

Solution: We prove the result assuming that 0 ∈

/ A which is sufficient for our
purpose. Write B = {x2 : x ∈ A}. For each n ∈ N let An = A ∩ (−n, n) and

13
 ∞
[
Bn = B ∩ (0, n2 ). Since B = Bn and µ∗ is countably sub-additive, the proof
n=1
will be complete if we can show that µ∗ (Bn ) = 0 ∀n.

Let ε > 0 be given. Since µ∗ (An ) ≤ µ∗ (A) = 0 i.e. µ∗ (An ) = 0, so from

definition we can find a countable collection of open intervals {Ik }k∈N such that
∞ ∞
[ X ε
An ⊂ Ik , and `(Ik ) < .
3n
k=1 k=1

Without any loss of generality we can assume that none of Ik s contain 0 and
are contained in (−n, n). Write Ik = (ck , dk ) for each k. Consider the interval
Jk = (c2k , d2k ) if ck > 0 and Jk = (d2k , c2k ) if dk < 0. Clearly the countable
collection of open intervals {Jk }k∈N covers Bn and further
∞
X ∞
X ∞
X
`(Jk ) = |ck + dk |(dk − ck ) ≤ 2n `(Ik ) < ε.
k=1 k=1 k=1

This shows that µ∗ (Bn ) = 0.

Summary: In this chapter we have shown that how the usual notion of length
of intervals can be extended to general sets of real numbers which we call the
Lebesgue outer measure. we show that the Lebesgue outer measure µ∗ is non-
negative, monotone, countably sub-additive with µ∗ (φ) = 0. We then define the
notion of Lebesgue measurable sets and show that it forms a σ-algebra.

Following are the main definitions and points to be remembered.

• For any set A of real numbers the Lebesgue outer measure of A is defined
by
X
µ∗ (A) = inf{ `(In ) : {In }n∈N is a countable collection of open intervals covering A}.
n

• The Lebesgue outer measure µ∗ is non-negative, monotone, countably sub-

additive, translation invariant and outer measure of any interval is equal to its
length.

• A set E of real numbers is called Lebesgue measurable if for any A ⊂ R ,

µ∗ (A) = µ∗ (A ∩ E) + µ∗ (A ∩ E c ) where E c stands for the complement of E.

• Difference of two measurable sets is measurable as also countable union

of measurable sets. Complements of measurable sets are measurable. In other
words the collection of measurable sets S forms a σ-algebra. Translations of

14
measurable sets are also measurable and finally all Borel sets are measurable.

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.

15
 SELF ASSESSMENT

Example 1.1. Let A ⊂ R be such that µ∗ (A) = 0. For any subset B of R

which of the following is true?

(a) µ∗ (A ∪ B) = µ∗ (B \ A) but µ∗ (A ∪ B) 6= µ∗ (B).

(b) µ∗ (B) = µ∗ (A∆B) = µ∗ (B \ A).
(c) µ∗ (B) = µ∗ (A∆B) but µ∗ (A∆B) 6= µ∗ (B \ A).
(d) µ∗ (B) = µ∗ (A ∪ B) but µ∗ (A∆B) 6= µ∗ (A ∪ B).

Example 1.2. The Lebesgue outer measure of the Cantor set is

(a) 1.
1
(b) 2.

(c) 0.
(d) ∞.

Example 1.3. The set of all irrational numbers is

(a) non-measurable.
(b) Lebesgue measurable with Lebesgue outer measure 0.
(c) Lebesgue measurable with Lebesgue outer measure ∞.
(d) Lebesgue measurable with finite Lebesgue outer measure.

Example 1.4. Let E ⊂ [0, 1]. Which of the following is true ?

(a) If Int(E) = φ then µ∗ (E) = 0.

(b) If Int(E) 6= φ then µ∗ (E) > 0.
(c) If µ∗ (E) = 0 then Int([0, 1] \ E) 6= φ.
(d) If µ∗ (E) = 0 then Int([0, 1] \ E) = φ.

16
Example 1.5. If A ⊂ R is Lebesgue measurable, then for any B ⊂ R is it true
that
µ∗ (A ∪ B) + µ∗ (A ∩ B) = µ∗ (A) + µ∗ (B)?

Example 1.6. If {En }∞ n=1 is a collection of Lebesgue measurable sets then

there exists a disjoint collection {Fn }∞
n=1 of Lebesgue measurable sets such that
∞
S ∞
S
Fn = En . Is this true or false?
n=1 n=1

Example 1.7. Let {An }∞ n=1 be a collection of sets. If there exists a disjoint
collection {En }∞
n=1 of Lebesgue measurable sets such that An ⊂ En , for each
n ∈ N then is it true that
∞
[ ∞
X
µ∗ ( An ) = µ∗ (An )?
n=1 n=1

Example 1.8. Examine whether the following is true or false. E ⊂ R is

Lebesgue measurable if and only if there exists a sequence {En } of Lebesgue
measurable subsets of E such that µ∗ (E \ En ) → 0.

Example 1.9. Let C be the Cantor ternary set. Which of the following condi-
tions is false ?

(a) C is uncountable and perfect.

(b) C is measurable and compact.
(c) C is measurable and perfect.
(d) C is measurable and C contains open intervals.

Example 1.10. Let E ⊂ R and r, x ∈ R. Which of the following is not neces-

sarily true ?

(a) µ∗ (x + E) = µ∗ (E).
(b) µ∗ (x2 + E) = µ∗ (E).
(c) µ∗ (rE) = r · µ∗ (E).
(d) µ∗ (r2 E) = r2 · µ∗ (E).

17
 SOLUTIONS

• Solution of Example 1.1: Ans: (b).

• Solution of Example 1.2: Ans: (c).

• Solution of Example 1.3: Ans: (c).

• Solution of Example 1.4: Ans: (b).

(For (a), take E = [0, 1] \ Q. For (c), take E = [0, 1] ∩ Q. For (d), take E to be
any finite subset of [0, 1].)

• Solution of example 1.5: True.

• Solution of Example 1.6: True.

n−1
S
( Take Fn = En \ Ek .)
k=1

• Solution of Example 1.7: True.  ∞ 

(Since En ’s are disjoint and measurable, for any A ⊂ R we have µ∗ A ∩
S
En =
n=1
∞ ∞
µ∗ (A ∩ En ). So taking A =
P S
An the result follows.)
n=1 n=1

• Solution of Example 1.8: True.

• Solution of example 1.9: Ans: (d).

• Solution of Example 1.10: Ans: (c).

( Note that r · µ∗ (E) is negative for r < 0 while µ∗ (rE) must be non-negative
when µ∗ (E) > 0. )

18
 LEARN MORE

The history of Lebesgue outer measure is linked with the history of develop-
ment of integration. In the middle of 19th century Bernhard Riemann developed
the integral (known as Riemann integral after his name) which we frequently
use today. However, it was realized quickly that it is not too difficult to find
functions which are not Riemann (or Darbaux) integrable. In his thesis in 1902,
Henri Lebesgue first developed the idea of Lebesgue measure extending the no-
tion of length of intervals and applied it to develop a new integral. As for the
theory of Lebesgue outer measure, a lot of contributions were made by Italian
mathematician Giuseppe Vitali from formally introducing the notion of outer
measure to constructing an example of a non-measurable set which we will dis-
cuss in the next module. The whole theory of lebesgue measure and integration
was developed in 20th century.

The students are encouraged to consult the following excellent books and
web resources to have more in depth study about the course materials as well
as to know further.

Reference Books:

1. G. De Barra, Measure theory and integration, Reprint, New age interna-

tional (P) Ltd., New Delhi, 1987.

2. C. Caratheodory, Algebraic theory of measure and integration, Chelsea

Publishing house, New York, 1963.

3. R.A. Gordon, The integrals of Lebesgue, Denjoy, Perron and Henstock,

American Mathematical Society, 1994.

4. P.R. Halmos, Measure Theory, Reprint, Springer Verlag, 1974.

5. P.K. Jain, V.P. Gupta, Lebesgue measure and integration, Wiley Eastern
Ltd., New Delhi (New Age international Ltd.), 2011.

6. M.E. Munroe, Introduction to measure and integration, Addison-Wesley,

Cambridge,Mass, 1953.

7. I.K. Rana, An Introduction to Measure and Integration (2nd ed.), Narosa

Publishing House, New Delhi, 2004.

19
 8. H. L. Royden and P.M. Fitzpatrick, Real Analysis Fourth edition, Pren-
tice Hall, New York, 2010.

Web Resources with links:

[Link] Analysis Qualifying Exam/Aug08 Real

[Link] level mathematics books

[Link] ars/week7 [Link]

[Link]/ publ/ln/[Link]

[Link] theory/measure [Link]

[Link]/ borell/[Link]

[Link] rieffel/[Link]

[Link] hunter/measure.../measure [Link]

[Link]

20