SECTION – I

MEASURABLE SETS

Introduction
In measure theory, a branch of mathematics, the concept of Lebesgue measure, was given by French
mathematician Henri Lebesgue in 1901. Sets that can be assigned a Lebesgue measure are called
Lebesgue-measurable; the measure of the Lebesgue-measurable set A is here denoted by m*(A).
Lebesgue Measure
In this section we shall define Lebesgue Measure, which is a generalization of the idea of length.
1.1 Definition. The length 𝑙(I) of an interval I with end points a and b is defined as the difference of
the end points. In symbols, we write.
𝑙 (𝐼) = 𝑏 – 𝑎.
1.2 Definition. A function whose domain of definition is a class of sets is called a Set Function. For
example, length is a set function. The domain being the collection of all intervals.
1.3 Definition. An extended real – valued set function  defined on a class E of sets is called
Additive if 𝐴  𝐸, 𝐵  𝐸, 𝐴  𝐵  𝐸 𝑎𝑛𝑑 𝐴  𝐵 = , 𝑖𝑚𝑝𝑙𝑦
 (𝐴  𝐵) =  (𝐴) +  (𝐵)
1.4 Definition. An extended real valued set function 𝜇 defined on a class E of sets is called finitely
additive if for every finite disjoint classes {𝐴1 , 𝐴2 , … . , 𝐴𝑛 }of sets in E, whose union is also in E,
we have
𝑛
𝑛
𝜇(𝑈𝑖=1 𝐴𝑖 ) = ∑ 𝜇(𝐴𝑖 )
𝑖=1

1.5 Definition. An extended real–valued set function  defined on a class E of sets is called
countably additive it for every disjoint sequence {𝐴𝑛 } of sets in E whose union is also in E, we have
∞
∞
𝜇(𝑈𝑖=1 𝐴𝑖 ) = ∑ 𝜇(𝐴𝑖 )
𝑖=1
1.6 Definition. Length of an open set is defined to be the sum of lengths of the open intervals of
which it is composed of. Thus, if G is an open set, then

𝑙(𝐺) = ∑ 𝑙(𝐼𝑛 )
𝑛

where
𝐺 = 𝑈𝑛 𝐼𝑛 , 𝐼𝑛1  𝐼𝑛2 = 𝜙 𝑖𝑓 𝑛1  𝑛2 .
2 Measure and Integration Theory
1.7 Definition. The Lebesgue Outer Measure or simply the outer measure m* of a set A is defined as
𝑚∗ (𝐴) = inf ∑ 𝑙(𝐼𝑛 ).
𝐴𝑈𝐼𝑛

where the infimum is taken over all finite or countable collections of intervals {In} such that 𝐴𝑈𝐼𝑛
Since the lengths are positive numbers, it follows from the definition of 𝑚∗ that 𝑚∗ (𝐴)  0.
1.8 Remark: (i) If 𝐴 ⊆ 𝐵 ,then 𝑚∗ (𝐴) ≤ 𝑚∗ (𝐵) i.e. outer-measure has monotone property.
Proof: By definition of outer-measure, for each ɛ > 0, there exist a countable collection of open
interval {In} such that 𝐵 ⊆ 𝑈𝑛 𝐼𝑛 and
𝑚∗ (𝐵) + ɛ > ∑𝑛 𝑙(𝐼𝑛 ) ...1)
now 𝐴 ⊆ 𝐵 𝑎𝑛𝑑 𝐵 ⊆ 𝑈𝑛 𝐼𝑛
=> 𝐴 ⊆ 𝑈𝑛 𝐼𝑛
𝑚∗ (𝐴) ≤ ∑𝑛 𝑙 (𝐼𝑛 )
< 𝑚∗ (𝐵) + ɛ ( 𝑢𝑠𝑖𝑛𝑔 1) )
 𝑚∗ (𝐴) < 𝑚∗ (𝐵) + ɛ
but ɛ > 0 is arbitrary, 𝑚∗ (𝐴) ≤ 𝑚∗ (𝐵) hence proved.
(ii) Outer-measure of a set is always non-negative.
1.9 Theorem. Outer measure is translation invariant.
Proof. Let  > 0 be given. Then by definition of outer measure, There exist a countable collection of
intervals {𝐼𝑛 } such that 𝐴   𝐼𝑛 𝑎𝑛𝑑
𝑚∗ (𝐴) +  > ∑𝑛 𝑙(𝐼𝑛 ).
Now, 𝐴  ⋃𝑛(𝐼𝑛 )
=> 𝐴 + 𝑥  ⋃𝑛(𝐼𝑛 + 𝑥),
=> 𝑚∗ (𝐴 + 𝑥) ∑𝑛 𝑙(𝐼𝑛 + 𝑥) = 𝛴 𝑙(𝐼𝑛 ) [length is translation invariant]
 𝑚∗ 𝐴 + 
Since  is arbitrary positive number, we have
(2) 𝑚∗ (𝐴 + 𝑥)  𝑚∗ (𝐴) (1)
To prove reverse inequality, Let  > 0 be given. Then by definition of outer measure, There exist a
countable collection of intervals {𝐽𝑛 } such that
𝐴 + 𝑥  ⋃𝑛 𝐽𝑛 𝑎𝑛𝑑
𝑚∗ (𝐴 + 𝑥) +  > ∑𝑛 𝑙(𝐽𝑛 ).
Now, 𝐴 + 𝑥  ⋃𝑛 𝐽𝑛
Measurable Sets 3
 𝐴  ⋃𝑛(𝐽𝑛 − 𝑥)
 𝑚∗ (𝐴)  ∑𝑛 𝑙(𝐽𝑛 − 𝑥)
 𝑚∗ (𝐴)  ∑𝑛 𝑙(𝐽𝑛 ) < 𝑚∗ (𝐴 + 𝑥) + 
 𝑚∗ (𝐴)  𝑚∗ (𝐴 + 𝑥) (2)
Then Combining (1) and (2), the required result follows.
i.e., 𝑚∗ (𝐴) = 𝑚∗ (𝐴 + 𝑥)
1.10 Theorem. The outer measure of an interval is its length.
Proof. CASE (1) Let us suppose, first I is a closed and bounded interval, say I = [a, b]
To prove: m*(I) = ℓ [a, b] =b - a.
Now for each ɛ> 0, I = [a, b] ⊆ (a - ɛ, b+ ɛ) then
by definition of outer-measure
=> m*(I) ≤ ℓ (a - ɛ, b+ ɛ) ≤ (b+ ɛ- a+ ɛ)
=> m*(I) ≤ b-a + 2 ɛ
since ɛ is an arbitrary, m*(I) ≤ b-a = ℓ (I) (1)
Now to prove, m*(I) = b-a, then it is sufficient to prove m*(I) ≥ b-a. let {In} be a countable collection of
open intervals which covering I i.e.
I ⊆ ⋃𝑛 𝐼𝑛
∑𝑛 ℓ(𝐼𝑛 ) ≥ b-a for all n ∈ N so it is sufficient to prove that
inf ∑𝑛 ℓ(𝐼𝑛 ) ≥ b-a
since I = [a, b] is compact, then by Heine Boral theorem, we can select a finite number of open intervals
from this {In} such that their union contains I.
Let the intervals be J1, J2, ..., Jp such that ⋃𝑝𝑖=1 𝐽𝑖 ⊇ [a, b].
Now it is sufficient to prove ∑𝑝𝑖=1 ℓ(𝐽𝑖 ) ≥ b-a (2)
Now a ∈ I = [a, b], there exist open interval J1 = (a1, b1) from the above-mentioned finite no. of intervals
such that a1 < a≤ b then b1 ∈ I.
Again, there exist an open interval (a2, b2) from the finite collection J1, J2, ..., Jp such that a2 <b1 < b2.
Continuing this, we get a sequence of open intervals
(a1, b1), (a2, b2), ..., (ap ,bp ) from J1, J2, ..., Jp satisfying ai < bi-1 < bi ,i= 2,3,.....,p since the collection is
finite so the process must stop with an interval satisfying ap < bp-1 < bp and ap <b < bp
∑𝑛 ℓ(𝐼𝑛 ) ≥ ∑𝑝𝑖=1 ℓ(𝐽𝑖 )= ℓ (a1, b1) + ℓ (a2, b2) +.... ℓ (ap, bp)
= (b1 -a1) + (b2 -a2) + ...+ (bp -ap)
= bp + (bp-1 – ap) +...+b1 -a2 – a1
4 Measure and Integration Theory
> bp – a1
> b-a
=> inf ∑𝑛 ℓ(𝐼𝑛 ) ≥ b-a
=> m*(I) ≥ b-a (4)
Hence result is proved in the case when I closed and bounded interval.
CASE (2) let I be bounded open interval with end points a and b, then for every real no. ɛ> 0 [a+ ɛ, b- ɛ]
⊂ I ⊂ [a, b]
=> m*[a+ ɛ, b- ɛ] ≤ m*(I) ≤ m* [a, b]
=> ℓ [a+ ɛ, b- ɛ] ≤ m*(I) ≤ ℓ [a, b] (by case 1)
=>b- ɛ -a- ɛ≤ m*(I) ≤b-a
since ɛ is arbitrary,
we get b-a ≤ m*(I) ≤ b-a
=> m*(I) =b-a.
CASE (3) if I is the unbounded interval, then for each real no. r> 0, we can find bounded closed interval
J ⊂ I such that ℓ (J)>r
Now J ⊂ I => m* (J) ≤ m* (I)
=> ℓ (J) ≤ m* (I)
=> m* (I) > r since this hold for each real no. r,
we get m* (I) = ∞ = ℓ(I)
i.e. outer-measure is of an interval equal to its length.
1.11 Theorem. Let {𝐴𝑛 } be a countable collection of sets of real numbers. Then
𝑚∗ (𝐴𝑛 ) ≤ 𝛴 𝑚∗ 𝐴𝑛 .
Proof. Proof. If one of the sets 𝐴𝑛 has infinite outer measure, the inequality holds trivially. So suppose
𝑚∗ {𝐴𝑛 } is finite. Then, given  > 0, there exists a countable collection {𝐼𝑛,𝑖 } of open intervals such
that 𝐴𝑛  𝑈𝑖 𝐼𝑛,𝑖 and
𝜀
𝛴𝑖 𝑙(𝐼𝑛,𝑖 ) < 𝑚∗ (𝐴𝑛 ) +
2𝑛
by the definition of 𝑚∗ {𝐴𝑛 }.
Now the collection [𝐼𝑛,𝑖 ]𝑛,𝑖 = 𝑈𝑛 [𝐼𝑛,𝑖 ]𝑖 is countable, being the union of a countable number of
countable collections, and covers ⋃𝑛 𝐴𝑛 . Thus

𝑚∗ (⋃ 𝐴𝑛 )  𝛴𝑛,𝑖 𝑙(𝐼𝑛,𝑖 )
𝑛
Measurable Sets 5
= 𝛴𝑛 𝛴𝑖 𝑙(𝐼𝑛 , 𝑖)

< ( 𝑚∗ (𝐴𝑛 ) + )
2𝑛

= 𝛴𝑛 𝑚∗ 𝐴𝑛 + 𝛴𝑛
2𝑛
1
= 𝛴𝑛 𝑚∗ 𝐴𝑛 + 𝜖𝛴𝑛 𝑛
2
= 𝛴 𝑚∗ 𝐴𝑛 + 
Since  is an arbitrary positive number, it follows that
𝑚∗ (⋃𝑛 𝐴𝑛 ) ≤ 𝛴 𝑚∗ (𝐴𝑛 ) .
1.12 Theorem. Outer-measure of singleton set of reals is zero
1 1
Proof: Let 𝐴 = {𝑎} Then, since 𝐴 = {𝑎}, {𝑎} ⊆ (𝑎 − 𝑛 , 𝑎 + 𝑛) ∀𝑛 ∈ 𝑁

1 1
 𝑚∗ (𝑎) ≤ 𝑚 ∗ (𝑎 − 𝑛 , 𝑎 + 𝑛)
2
 𝑚∗ (𝑎) ≤ 𝑛
2
 0 ≤ 𝑚∗ (𝑎) ≤ 𝑛 for each n.

In limiting case 𝑚∗ (𝑎) = 0.

1.13 Theorem. Outer-measure of null set is zero.
1 1
Proof: Since 𝜙 ⊆ ( − 𝑛 , 𝑛) ∀𝑛 ∈ 𝑁

1 1
 𝑚∗ (𝜙) ≤ 𝑚∗ ( − 𝑛 , 𝑛 )
2
 𝑚∗ (𝜙) ≤ 𝑛
2
 0 ≤ 𝑚∗ ( 𝜙) ≤ 𝑛 for each n. In limiting case 𝑚∗ (𝜙) = 0.

1.14 Corollary. If A is countable, 𝑚∗ 𝐴 = 0

Proof. We know that a countable set is the union of a countable family of singleton. Therefore 𝐴 = ∪
[𝑥𝑛 ], which yields
𝑚∗ 𝐴 = 𝑚∗ [∪ (𝑥𝑛 )] ≤ 𝛴 𝑚∗ [𝑥𝑛 ] (by the above theorem)
But as already pointed out outer measure of a singleton is zero. Therefore it follows that
𝑚∗ 𝐴  0
Since outer measure is always a non – negative real number, 𝑚∗ 𝐴 = 0.
1.15 Remark: The Sets 𝑵, 𝒁, 𝑸 has outer-measure zero.
6 Measure and Integration Theory
1.16 Remark: Prove that [0, 1] is uncountable.
Proof: Assume on the contrary that the set [0, 1] is countable, then as we know outer-measure of
countable set is zero, then 𝑚∗ [0, 1] = 0, 𝑖. 𝑒, 𝑙[0, 1] = 0
𝑖. 𝑒. , 1 = 0, a contradiction. therefore [0, 1] is uncountable.
1.17 Corollary. If 𝑚∗ 𝐴 = 0, then 𝑚∗ (𝐴  𝐵) = 𝑚∗ 𝐵.
Proof. Using the above proposition
𝑚∗ (𝐴  𝐵) ≤ 𝑚∗ 𝐴 + 𝑚 ∗ 𝐵
= 0 + 𝑚∗ 𝐵 (i)
Also 𝐵  𝐴  𝐵
Therefore 𝑚∗ 𝐵  𝑚∗ (𝐴  𝐵) (ii)
From (i) and (ii) it follows that
𝑚∗ 𝐵 = 𝑚∗ (𝐴  𝐵)
Note:- Because of the property m* ( An ) Σ m* An , the function m* is said to be countably Sub-
additive. It would be much better if m* were also countably additive, that is,
if 𝑚∗ (∪ 𝐴𝑛 ) = ∑𝑚∗ 𝐴𝑛 .
for every countable collection [𝐴𝑛 ] of disjoint sets of real numbers. If we insist on countable additivity,
we have to restrict the domain of the function m* to some subset m of the set 2R of all subsets of R. The
members of m are called the measurable subsets of R. That is, to do so we suitably reduce the family of
sets on which m* is defined. This is done by using the following definition due to Carathedory.
1.18 Definition. A set E of real numbers is said to be m* measurable, if for every set A  R, we have
𝑚∗ 𝐴 = 𝑚∗ (𝐴  𝐸) + 𝑚∗ (𝐴  𝐸 𝑐 )
Since A = (A  E)  (A  Ec),
It follows from the definition that
𝑚∗ 𝐴 = 𝑚∗ [(𝐴  𝐸) (𝐴  𝐸 𝑐 )]  𝑚∗ (𝐴  𝐸) + 𝑚∗ (𝐴  𝐸 𝑐 )
Hence, the above definition reduces to:
A set E  R is measurable if and only if for every set A  R, we have
m* A  m* (A  E) + m* (A  Ec).
For example  is measurable.
1.19 Theorem. Prove that 𝜙 is measurable set.
Proof: Let A be set of reals, then m* A = m* (A  E) + m* (A  Ec)
Put E = 𝜙
Measurable Sets 7

m* (A 𝜙) + m* (A  𝜙c) = m* (𝜙 ) + m* (A R)

= 0 + m* A
= m* A
This implies 𝜙 is measurable.
1.20 Theorem. Prove that R is measurable set.
Proof: Let A be set of reals, then
m* A = m* (A  E) + m* (A  Ec)
Put E = R
m* (A R) + m* (A Rc) = m* (A) + m* (A 𝜙)
= m* (A) + m* (𝜙)
= m* A + 0
= m* A
This implies R is measurable.
1.21 Theorem. If m* E = 0, then E is measurable.
Proof. Let A be any set. Then A  E  E and so
m* (A  E)  m* E = 0 (i)
Also A  A  Ec, and so
m* A  m* (A  Ec) = m* (A  Ec) + m* (A  E)
as m* (A  E) = 0 by (i)
Hence E is measurable.
1.22 Theorem. Every subset of E is measurable if m* E = 0.
Proof: Let F be any subset of E, where m* E = 0.
then since 𝐹 ⊆ 𝐸
this implies m* F ≤ m* E
this implies m* F ≤ 0
Also m* F ≥ 0
therefore m* F =0.
this implies F is measurable.
1.23 Theorem. Every singleton set is measurable.
Proof: Since outer measure of singleton set is zero and set of measure zero is measurable. Therefore,
singleton set is measurable.
8 Measure and Integration Theory
1.24 Theorem. Every countable set is measurable.
Proof: Since outer measure of countable set is zero and set of measure zero is measurable. Therefore
countable set is measurable.
1.25 Theorem. If a set E is measurable, then so is its complement Ec.
Proof. The definition is symmetrical with respect to Ec, and so if E is measurable, its complement Ec is
also measurable.
1.26 Theorem. Union of two measurable sets is measurable.
Proof. Let E1 and E2 be two measurable sets and let A be any set. Since E2 is measurable, we have
𝑚∗ (𝐴  𝐸1𝑐 ) = 𝑚∗ (𝐴  𝐸1𝑐 𝐸2 ) + 𝑚∗ (𝐴  𝐸1𝑐 𝐸2𝑐 ) (i)
and since 𝐴  (𝐸1  𝐸2 ) = (𝐴  𝐸1 ) [ 𝐴  𝐸2  𝐸1𝑐 ] (ii)
Therefore by (ii) we have
𝑚∗ [𝐴  (𝐸1  𝐸2 )]  𝑚∗ (𝐴  𝐸1 ) + 𝑚∗ [𝐴  𝐸2  𝐸1𝑐 ] (iii)
Thus
𝑚∗ [𝐴  (𝐸1  𝐸2 )] + 𝑚∗ (𝐴  𝐸1𝑐  𝐸2𝑐 )
≤ 𝑚∗ (𝐴  𝐸1 ) + 𝑚∗ (𝐴  𝐸2  𝐸1𝑐 ) + 𝑚∗ (𝐴  𝐸1𝑐  𝐸2𝑐 )
= 𝑚∗ (𝐴  𝐸1 ) + 𝑚∗ (𝐴  𝐸1𝑐 ) (𝑏𝑦 (𝑖))
≤ m* A (since E1 is measurable)
i.e. m* (A  (E1  E2)) + m* (A  (E1  E2)c)  m* A
Hence E1  E2 is measurable.
If E1 and E2 are measurable, then E1  E2 is also measurable.
In fact we note that E1, E2 are measurable ⇒ 𝐸1𝑐 , 𝐸2𝑐 are measurable ⇒ 𝐸1𝑐  𝐸 𝑐 is measurable ⇒
(𝐸1𝑐  𝐸2𝑐 )𝑐 = 𝐸1  𝐸2 is measurable.
Similarly, it can be shown that if E1 and E2 are measurable, then 𝐸1𝑐  𝐸2𝑐 is also measurable.
1.27 Lemma. Difference of two measurable sets is also measurable.
Proof: Let E1 and E2 be two measurable sets. Then 𝐸2𝑐 is measurable and hence
𝐸₁ ∩ 𝐸2𝑐 = 𝐸₁ – 𝐸₂ is measurable, being the intersection of two measurable sets.
1.28 Definition. Algebra or Boolean Algebra: - A collection A of subsets of a set X is called an
algebra of sets or a Boolean Algebra if
(i) A, B  A => A  B  A
(ii) A  A => Ac  A
(iii) For any two members A and B of A, the intersection A  B is in A.
Because of De Morgan’s formulae (i) and (ii) are equivalent to (ii) and(iii).
Measurable Sets 9
It follows from the above definition that the collection M of all measurable sets is an algebra. The proof
is an immediate consequence of Theorems 1.25 and 1.26.
1.29 Definition. By a Boolean  - algebra or simply a  - algebra or Borel field of a collection of sets,
we mean a Boolean Algebra A of the collection of the sets such that union of any countable collection of
members of this collection is a member of A.

From De Morgan’s formula an algebra of sets is a  - algebra or Borel field if and only if the
intersection of any countable collection of members of A is a member of A.
1.30 Lemma. Let A be any set, and 𝐸1 , 𝐸2 , … . , 𝐸𝑛 a finite sequence of disjoint measurable sets. Then
𝑚∗ (𝐴  [𝑈 𝑛𝑖1 𝐸𝑖 ]) = 𝛴𝑖𝑛1 𝑚∗ (𝐴  𝐸𝑖 )
Proof. We shall prove this lemma by induction on n. The lemma is trivial for
n = 1. Let n > 1 and suppose that the lemma holds for n – 1 measurable sets Ei.
Since En is measurable, we have

𝑚∗ (𝑋) = 𝑚∗ (𝑋  𝐸𝑛 ) + 𝑚∗ (𝑋  𝐸𝑛 𝑐 ) for every set X  R.

In particular we may take
𝑋 = 𝐴  [𝑈𝑖𝑛1 𝐸𝑖 ].
Since 𝐸1 , 𝐸2 , … . , 𝐸𝑛 are disjoint, we have
𝑋  𝐸𝑛 = 𝐴  [𝑈𝑖𝑛1 𝐸𝑖 ] ∩ 𝐸𝑛 = 𝐴 ∩ 𝐸𝑛
𝑛1
𝑋  𝐸𝑛𝑐 = 𝐴  [𝑈𝑖𝑛1 𝐸𝑖 ] 𝐸𝑛𝑐 = 𝐴  [𝑈𝑖=1 𝐸𝑖 ]
Hence, we obtain 𝑚∗ 𝑋 = 𝑚∗ (𝐴  𝐸𝑛 ) + 𝑚∗ (𝐴  [𝑈𝑖𝑛−1
1 𝐸𝑖 ]) (i)

But since the lemma holds for 𝑛 – 1 we have

𝑛−1

𝑚 ∗ (𝐴
[ 𝑈𝑖𝑛11 𝐸𝑖 ]) = ∑ 𝑚∗ (𝐴  𝐸𝑖 )
𝑖 1

Therefore (i) reduces to

𝑚∗ 𝑋 = 𝑚 ∗ (𝐴  𝐸𝑛 ) + 𝛴𝑖𝑛−1
1 𝑚 (𝐴  𝐸𝑖 )
∗

= ∑𝑛𝑖1 𝑚∗ (𝐴  𝐸𝑖 ).
Hence the lemma.
1.31 Lemma. Let A be an algebra of subsets and {𝐸𝑖 | 𝑖  𝑁} a sequence of sets in A. Then there
exists a sequence [𝐷𝑖 | 𝑖  𝑁] of disjoint members of A such that
𝐷𝑖  𝐸𝑖 ( 𝑖  𝑁)
𝑈𝑖∈𝑁 𝐷𝑖 = 𝑈𝑖∈𝑁 𝐸𝑖
10 Measure and Integration Theory

Proof. For every i  N, let

𝐷𝑛 = 𝐸𝑛 − (𝐸1  𝐸2  … …  𝐸𝑛−1 )
= (𝐸𝑛  (𝐸1  𝐸2  … …  𝐸𝑛−1 ))𝑐
= 𝐸𝑛  𝐸1 𝑐  𝐸2 𝑐  … …  𝐸𝑛−1 𝑐
Since the complements and intersections of sets in A are in A, we have each 𝐷𝑛  𝐴. By construction,
we obviously have 𝐷𝑖  𝐸𝑖 (𝑖  𝑁)
Let Dn and Dm be two such sets, and suppose m < n. Then Dm  Em, and so
Dm  Dn  Em  Dn
= 𝐸𝑚  𝐸𝑛  𝐸1 𝑐  … . . 𝐸𝑚
𝑐
 … .  𝐸𝑛−1
𝑐
(𝑢𝑠𝑖𝑛𝑔 (𝑖))
= (𝐸𝑚  𝐸𝑚
𝑐)
 ….=   ……… = 
The relation (i) implies 𝑈𝑖𝑁 𝐷𝑖  𝑈𝑖∈𝑁 𝐸𝑖
It remains to prove that
𝑈𝑖𝑁 𝐷𝑖 𝑈𝑖∈𝑁 𝐸𝑖
For this purpose let x be any member of 𝑈𝑖∈𝑁 𝐸𝑖 . Let n denotes the least natural number satisfying x 
En. Then we have
𝑥  𝐸𝑛 − (𝐸1  𝐸2  … …  𝐸𝑛−1 ) = 𝐷𝑛  𝑈𝑖𝑁 𝐷𝑛 .
This completes the proof.
1.32 Theorem. The collection M of measurable sets is a  - algebra.
Proof. We have proved already that M is an algebra of sets and so we have only to prove that M is
closed with respect to countable union. By the lemma proved above each set E of such countable union
must be the union of a sequence {𝐷𝑛 } of pairwise disjoint measurable sets. Let A be any set, and let
𝐸𝑛 = 𝑈𝑖𝐼 𝐷𝑖  𝐸. Then En is measurable and 𝐸𝑛𝑐  𝐸 𝑐 . Hence
𝑚∗ 𝐴 = 𝑚∗ (𝐴 𝐸𝑛 ) + 𝑚∗ (𝐴  𝐸𝑛𝑐 )  𝑚∗ (𝐴 𝐸𝑛 ) + 𝑚∗ (𝐴  𝐸𝑛𝑐 ).
But, by lemma 1.30,
𝑚∗ (𝐴  𝐸𝑛 ) = 𝑚∗ [𝐴 (𝑈𝑖1 𝐷𝑖 )] = 𝛴𝑖𝑛1 𝑚∗ (𝐴  𝐷𝑖 )
Therefore,
𝑚∗ 𝐴  𝛴𝑖𝑛1 𝑚∗ (𝐴  𝐷𝑖 ) + 𝑚∗ (𝐴  𝐸 𝑐 )
Since the left hand side of the inequality is independent of n, we have
𝑚∗ 𝐴  𝛴𝑖∞1 𝑚∗ (𝐴  𝐷𝑖 ) + 𝑚∗ (𝐴  𝐸 𝑐 )
 𝑚∗ (𝑈𝑖∞𝐼 [𝐴  𝐷𝑖 ]) + 𝑚∗ (𝐴  𝐸 𝑐 ) (by countably subadditivity of m*)
= 𝑚∗ ( 𝐴 𝑈𝑖∞
𝐼 𝐷𝑖 ) + 𝑚 (𝐴  𝐸 )
∗ 𝑐
Measurable Sets 11
= 𝑚∗ (𝐴 𝐸) + 𝑚∗ (𝐴  𝐸𝑛𝑐 ).
which implies that E is measurable. Hence the theorem.
1.33 Lemma. The interval (a,  ) is measurable
Proof. Let A be any set and
𝐴1 = 𝐴  (𝑎,  )
𝐴2 = 𝐴  (𝑎,  )𝑐 = 𝐴  (−  , 𝑎].
Then we must show that
𝑚∗ 𝐴1 + 𝑚∗ 𝐴2  𝑚∗ 𝐴.
𝐼𝑓 𝑚∗ 𝐴 =  , then there is nothing to prove. If m* A <  , then given  > 0 there is a countable
collection {In} of open intervals which cover A and for which
𝛴𝑙(𝐼𝑛 )  𝑚∗ 𝐴 + 
Let 𝐼𝑛 = 𝐼𝑛  (𝑎,  ) 𝑎𝑛𝑑 𝐼𝑛 = 𝐼𝑛  (− , 𝑎). Then 𝐼𝑛 and 𝐼𝑛 are intervals (or empty) and 𝑙(𝐼𝑛 ) =
𝑙 (𝐼𝑛 ) + 𝑙(𝐼𝑛 ) = 𝑚∗ ( 𝐼𝑛 ) + 𝑚∗ ( 𝐼𝑛 )
Since 𝐴1  𝑈𝐼𝑛 , we have
𝑚∗ 𝐴1  𝑚∗ (𝑈𝐼𝑛 )  𝛴 𝑚∗ 𝐼𝑛 , (iii)
and since, 𝐴2  𝑈 𝐼𝑛 , we have
𝑚∗ 𝐴2  𝑚∗ (𝑈 𝐼𝑛 ) 𝛴 𝑚∗ 𝐼𝑛 , (iv)
Adding (iii) and (iv) we have
𝑚∗ 𝐴1 + 𝑚∗ 𝐴2  𝛴 𝑚∗ 𝐼𝑛 +  𝛴 𝑚∗ 𝐼𝑛
= 𝛴 (𝑚∗ 𝐼𝑛 +  𝑚∗ 𝐼𝑛 )
= 𝛴 𝑙 (𝐼𝑛 ) [𝑏𝑦 (𝑖𝑖)]
≤ 𝑚∗ 𝐴 +  [𝑏𝑦 (𝑖)]
But  was arbitrary positive number and so we must have 𝑚∗ 𝐴1 + 𝑚∗ 𝐴2  𝑚∗ 𝐴.
1.34 Definition. The collection ß of Borel sets is the smallest  - algebra which contains all of the
open sets.
1.35 Theorem. Every Borel set is measurable. In particular each open set and each closed set is
measurable.
Proof. We have already proved that (a, ) is measurable. So we have
(𝑎, )𝑐 = (−∞, 𝑎] measurable.
∞ 1
Since (−∞, 𝑏) = 𝑈𝑛=1 ((−∞, 𝑏 − 𝑛 ]) and we know that countable union of measurable sets is
measurable, therefore (- , b) is also measurable. Hence each open interval,
12 Measure and Integration Theory
(𝑎, 𝑏) = ( −∞, 𝑏)  (𝑎, ∞) is measurable, being the intersection of two measurable sets. But each open
set is the union of countable number of open intervals and so must be measurable (The measurability of
closed set follows because complement of each measurable set is measurable).
Let M denote the collection of measurable sets and C the collection of open sets. Then
C  M. Hence ß is also a subset of M since it is the smallest  - algebra containing C . So each element
of ß is measurable. Hence each Borel set is measurable.
1.36 Definition. If E is a measurable set, then the outer measure of E is called the Lebesgue Measure
of E, is denoted by m. Thus, m is the set function obtained by restricting the set function m* to the
family M of measurable sets. Two important properties of Lebesgue measure are summarized by the
following theorem.
1.37 Theorem. Let {𝐸𝑛 } be a sequence of measurable sets. Then
𝑚( 𝐸𝑖 ) ≤ 𝛴 𝑚 𝐸𝑖
If the sets En are pairwise disjoint, then
𝑚( 𝐸𝑖 ) = 𝛴 𝑚𝐸𝑖 .
Proof. The inequality is simply a restatement of the sub-additivity of m*. If {Ei }is a finite sequence of
disjoint measurable sets. So we apply lemma 1.30 replacing A by R. That is , we have
𝑚∗ (𝑅 [𝑈𝑖𝑛 𝐸𝑖 ]) = 𝛴𝑖𝑛1 𝑚∗ (𝑅 𝐸𝑖 )
𝑛

𝑚∗ (𝑈𝑖𝑛 𝐸𝑖 ) = ∑ 𝑚∗ 𝐸𝑖
𝑖

and so m is finitely additive..

Let {Ei} be an infinite sequence of pairwise disjoint sequence of measurable sets. Then
And so  𝑈𝑖=1
∞ 𝑛
𝐸𝑖 ⊃ 𝑈𝐼=1 𝐸𝑛

1 𝐸𝑖 )  𝑚(𝑈(𝑖  1) 𝐸𝑖 ) = 𝛴𝑖1 𝑚 𝐸𝑖
∞
𝑚(𝑈𝑖∞ ∞

Since the left-hand side of this inequality is independent of n, we have

∞
𝑚(𝑈𝑖=1 𝐸𝑖 ) ≥ 𝛴𝑖∞1 𝑚 𝐸𝑖
The reverse inequality follows from countable sub-additivity and we have
∞
𝑚(𝑈𝑖=1 𝐸𝑖 ) = 𝛴𝑖∞1 𝑚 𝐸𝑖
Hence the theorem is proved.
1.38 Theorem. Let {En} be an infinite sequence of measurable sets such that En+1 ⊂ En for each n.
Let mE1 < ∞. Then
∞

m (⋂ En ) = lim mEn
n→∞
n=1
Measurable Sets 13
Proof. Let E = ⋂∞
i=1 Ei and let Fi = Ei − Ei−1 . Then since {En} is a decreasing sequence. We have
⋂ Fi = ϕ .
Also we know that if A and B are measurable sets then their difference A − B = A⋂Bc is also
measurable. Therefore each Fi is measurable. Thus {Fi} is a sequence of measurable pairwise disjoint
sets.
∞ ∞

Now ⋃ Fi = ⋃(Ei − Ei+1 )

i=1 i=1
∞
c
= ⋃(Ei ∩ Ei+1 )
i=1

= E1 ∩ (∪ Eic )
∞ c

= E1 ∩ (⋂ Ei )
i=1

= E1 ∩ E c
= E1 − E
Hence
∞

m (⋃ Fi ) = m(E1 − E)
i=1
∞

⇒ ∑ mFi = m(E1 − E)
i=1

⇒ ∑∞
i=1 m(Ei − Ei+1 ) = m(E1 − E) … (i)
Since E1 = (E1 − E) ∪ E, therefore
mE1 = m(E1 − E) + m(E)
⇒ mE1 − mE = m(E1 − E) (since mE ≤ mE1 < ∞ ) … (ii)
Again
Ei = (Ei − Ei+1 ) ∪ Ei+1
⇒ mEi = m(Ei − Ei+1 ) + mEi+1
⇒ mEi − mEi+1 = m(Ei − Ei+1 ) ( since Ei+1 ⊂ Ei ) … (iii)
Therefore (i) reduces to
mE1 − mE = ∑∞
i=1(mEi − mEi+1 ) ( using (ii)and (iii))
14 Measure and Integration Theory
∞

= lim ∑(mEi − mEi+1 )

n→∞
i=1

= lim [mE1 − mE2 + mE2 − mE3 … − mEn+1 ]

n→∞

= lim [mE1 − mEn+1 ]

n→∞

= mE1 − lim En+1

n→∞

⇒ mE = lim mEn
n→∞

⇒ m(⋂∞
i=1 Ei ) = lim mEn
n→∞

1.39 Remark. Show that the condition m(E1) <∞ is necessary in the above theorems.
Solution. Let En = [n, ∞)
Then, E1= [ 1,∞)
⟹m(E1) = m[1, ∞) = ∞
We show that the proposition of decreasing sequence does not hold in this case i.e. we want to show that
m(∩∞
n=1 En ) ≠ lim (En )
n→∞

Clearly, En+1 ⊂ En for all n

Now, En = [ n, ∞) ⊃ [ n, 2n]
⟹ m(En ) ≥ m [n, 2n] = n
⟹ m(En ) ≥ n
⟹ lim m(En ) = ∞ … (1)
n→∞

Now, we claim that m(∩∞ ∞

n=1 En ) = m(∩n=1 [n, ∞))= 0

For if, ∩∞ ∞
n=1 En ≠ ϕ ⟹ there exists x ∈∩n=1 En

⟹ x ∈ [n, ∞ ) for all n ∈ N

Let x∈ R, so by Archmedian property, we can find a positive integer n0 such that
n0 ≤ x < n0 + 1
⟹ x∉ [n0 + 1, ∞ ), a contradiction
∴∩∞
n=1 En = ϕ

⟹ m(∩∞
n=1 En ) = 0 … (2)
From (1) and (2), we have
m(∩∞
𝑛=1 𝐸𝑛 ) ≠ 𝑙𝑖𝑚 (𝐸𝑛 )
𝑛→∞
Measurable Sets 15
S0, theorem does not hold in this case.
1.40 Theorem. Let {En} be an increasing sequence of measurable sets. i.e. a sequence with 𝐸𝑛 ⊂
𝐸𝑛+1 for each n. Let mE1 be finite, then
∞

𝑚 (⋃ 𝐸𝑖 ) = 𝑙𝑖𝑚 𝑚𝐸𝑛 .
𝑛→∞
𝑖=1

Proof. The sets E1, E2- E1, E3- E2, …, En - En+1 are measurable and are pairwise disjoint . Hence
𝐸1 ∪ (𝐸2 − 𝐸1 ) ∪ … ∪ (𝐸𝑛 − 𝐸𝑛−1 ) ∪ …
is measurable and
𝑚[𝐸1 ∪ (𝐸2 − 𝐸1 ) ∪ … ∪ (𝐸𝑛 − 𝐸𝑛−1 ) ∪ …]
𝑛

= 𝑚𝐸1 + ∑ 𝑚 (𝐸𝑖 − 𝐸𝑖−1 )

𝑖=2
𝑛

= 𝑚𝐸1 + 𝑙𝑖𝑚 ∑ 𝑚(𝐸𝑖 − 𝐸𝑖−1 )

𝑛→∞
𝑖=2

But
𝐸1 ∪ (𝐸2 − 𝐸1 ) ∪ … ∪ (𝐸𝑛 − 𝐸𝑛−1 ) ∪ … is precisely ⋃∞
𝑖=1 𝐸𝑛

Moreover,
𝑛 𝑛

∑ 𝑚 (𝐸𝑖 − 𝐸𝑖−1 ) = ∑(𝑚𝐸𝑖 − 𝑚𝐸𝑖−1 )

𝑖=2 𝑖=2

= (𝑚𝐸2 − 𝑚𝐸1 ) + (𝑚𝐸3 − 𝑚𝐸2 ) + ⋯ + (𝑚𝐸𝑛 − 𝑚𝐸𝑛−1 )

= 𝑚𝐸𝑛 − 𝑚𝐸1
Thus we have
∞

𝑚 [⋃ 𝐸𝑖 ] = 𝑚𝐸1 + 𝑙𝑖𝑚 [𝑚𝐸𝑛 − 𝑚𝐸1 ]

𝑛→∞
𝑖=1

= 𝑙𝑖𝑚 𝑚𝐸𝑛
𝑛→∞

1.41 Definition : The symmetric difference of the sets A and B is the union of the sets A-B and B-A .
It is denoted by ∆ 𝐵 .
1.42 Theorem. If 𝑚(𝐸1 ∆ 𝐸2 ) = 0 and E1 is measurable, then E2 is measurable. Moreover mE2 =
mE1 .
Proof . We have
E2 = [𝐸1 ∪ (𝐸2 − 𝐸1 )] − (𝐸1 − 𝐸2 ) …(i)
16 Measure and Integration Theory
By hypothesis, both E2 – E1 and E1 – E2 are measurable and have measure zero. Since E1 and E2-E1 are
disjoint, 𝐸1 ∪ (𝐸2 − 𝐸1 ) is measurable and
𝑚[𝐸1 ∪ (𝐸2 − 𝐸1 )] = 𝑚𝐸1 + 0 = 𝑚𝐸1 . But, since
𝐸1 − 𝐸2 ⊂ [𝐸1 ∪ (𝐸2 − 𝐸1 )],
it follows from (i) that E2 is measurable and
𝑚𝐸2 = 𝑚[ 𝐸1 ∪ (𝐸1 − 𝐸2 )] − 𝑚(𝐸1 − 𝐸2 )
= 𝑚𝐸1 − 0 = 𝑚𝐸1 .
This completes the proof.
1.43Definition. Let x and y be real numbers in [0,1]. The sum modulo 1 of x and y , denoted by
0
𝑥 + 𝑦 , is defined by

0 𝑥 + 𝑦 𝑖𝑓 𝑥 + 𝑦 < 1
𝑥 + 𝑦= {
𝑥 + 𝑦 − 1 𝑖𝑓 𝑥 + 𝑦 ≥ 1

0
It can be seen that + is a commutative and associative operation which takespair of numbersin [0,1) into

numbers in [0,1).
If we assign to each 𝑥 ∈ [0,1) the angle 2𝜋𝑥 then addition modulo 1 corresponds to the addition of
angles.
If E is a subset of [0,1), we define the translation modulo 1 of E to be the set
0 0
𝐸 + 𝑦 = [z |𝑧 = x + y for some x ∈ 𝐸 ].

If we consider addition modulo 1 as addition of angles, translation module 1 by y corresponds to rotation

through an angle of 2𝜋𝑦.
We shall now show that Lebesgue measure is invariant under translation modulo 1.
1.44 Definition. Let x and y be real numbers in [0,1). The sum modulo 1 of x and y, denoted by
0
+ 𝑦 , is defined by

0 𝑥 + 𝑦 𝑖𝑓 𝑥 + 𝑦 < 1
𝑥 + 𝑦= {
𝑥 + 𝑦 − 1 𝑖𝑓 𝑥 + 𝑦 ≥ 1

0
Clearly 𝑥 + 𝑦 ∈ [0,1)
Measurable Sets 17
0
It can be seen that + is a commutative and associative operation which takes pair of numbers in [0,1)

into numbers in [0,1).

1.45 Definition. If E is a subset of [0,1), we define the translation modulo 1 of E to be the set
0 0
𝐸 + 𝑦 = [z |𝑧 = x + y for some x ∈ 𝐸 ].

We shall now show that Lebesgue measure is invariant under translation modulo 1.
0
1.46 Lemma. Let 𝐸 ⊂ [0,1) be a measurable set. Then for each 𝑦 ∈ [0,1) the set 𝐸 + 𝑦 is measurable

0
and m (𝐸 + 𝑦) = 𝑚𝐸.

Proof. Let E1= E∩ [0, 1 − 𝑦) and E2 = E ∩ [1 − 𝑦, 1). Then E1 and E2 are disjoint measurable sets
whose union is E, and so, mE = mE1 + mE2.
we observe that
0 0
𝐸1 + 𝑦 ={ 𝑥 + 𝑦: x ∈ 𝐸1 }

𝑥 + 𝑦 𝑖𝑓 𝑥 + 𝑦 < 1
= { 𝑥 ∈ 𝐸1
𝑥 + 𝑦 − 1 𝑖𝑓 𝑥 + 𝑦 ≥ 1.
But for x ∈ 𝐸1 , we have x + y < 1 and so
0
𝐸1 + 𝑦 = { 𝑥 + 𝑦, 𝑥 ∈ 𝐸1 } = 𝐸1 + 𝑦.

0
and hence 𝐸1 + 𝑦 is measurable. Thus

0
m (𝐸1 + 𝑦) = 𝑚(𝐸1 + 𝑦) = 𝑚(𝐸1 ),

0 0
since m is translation invariant. Also 𝐸2 + 𝑦 = 𝐸2 + (𝑦 − 1) and so 𝐸2 + 𝑦 is measurable and

0
m(𝐸2 + 𝑦) = 𝑚𝐸2 . But

0 0 0
𝐸 + 𝑦 = (𝐸1 + 𝑦) ∪ (𝐸2 + 𝑦)
18 Measure and Integration Theory
0 0 0
And the sets (𝐸1 + 𝑦) and (𝐸2 + 𝑦) are disjoint measurable sets. Hence 𝐸 + 𝑦 is measurable and

0 0 0
m (𝐸 + 𝑦) = 𝑚[(𝐸1 + 𝑦) ∪ (𝐸2 + 𝑦)]

0 0
= 𝑚(𝐸1 + 𝑦) + 𝑚(𝐸2 + 𝑦)

= 𝑚(𝐸1 ) + 𝑚(𝐸2 )
= 𝑚(𝐸).
This completes the proof of the lemma.
1.47 Theorem: Prove that there exists a non-measurable set in interval [0,1).
Proof: First we define an equivalence relation in the set I= [0,1), By saying that x and y are equivalent
i.e., 𝑥 ∼ 𝑦 if and only if x-y is a rational number.
If x-y is a rational number, we say that x and y are equivalent and write x-y. It is clear that 𝑥 ∼ 𝑥 ; 𝑥 ∼
𝑦 ⟹ 𝑦 ∼ 𝑥 𝑎𝑛𝑑 𝑥 ∼ 𝑦 , 𝑦 ∼ 𝑧 ⟹ 𝑥 ∼ 𝑧. Thus ‘ ∼’ is an equivalence relation in I.
Hence the relation ∼ partitions the set I = [0,1) into mutually disjoint equivalence classes, that is, classes
such that any two elements of one class differ by a rational number, while any two elements of different
classes differ by an irrational number.
Construct a set P by choosing exactly one element from each equivalence classes. Now we claim that P
is a non-measurable set.

∞ 0
Let < 𝑟𝑖 > be a sequence of the rational numbers in [0,1) with r0 = 0 and define Pi = P + ri.
𝑖=0
(translation modulo 1 of P)
Then P0 = P.
We further prove that (i) 𝑃𝑖 ∩ 𝑃𝑗 = ∅, 𝑖 ≠ 𝑗.
(ii) ⋃𝑛 𝑃𝑛 = [0, 1)
Proof: (i) Let 𝑃𝑖 ∩ 𝑃𝑗 ≠ ∅, 𝑖 ≠ 𝑗.
Let 𝑥 ∈ 𝑃𝑖 ∩ 𝑃𝑗 . => 𝑥 ∈ 𝑃𝑖 𝑎𝑛𝑑 𝑥 ∈ 𝑃𝑗
0
Then∃ 𝑝𝑖 , 𝑝𝑗 ∈ 𝑃 𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡 𝑥 = 𝑝𝑖 +𝑟𝑖

0
𝑥 = 𝑝𝑗 +𝑟𝑗
Measurable Sets 19
0 0
 𝑝𝑖 +𝑟𝑖 = 𝑝𝑗 +𝑟𝑗

 𝑝𝑖 -𝑝𝑗 = 𝑟𝑗 − 𝑟𝑖 𝑖𝑠 𝑎 𝑟𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟.

 𝑝𝑖 ~𝑝𝑗 𝑖𝑠 𝑎 𝑟𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟.
i.e., 𝑝𝑖 ~𝑝𝑗
=> 𝑝𝑖 𝑎𝑛𝑑 𝑝𝑗 𝑎𝑟𝑒 𝑖𝑛 𝑠𝑎𝑚𝑒 𝑒𝑞𝑢𝑖𝑣𝑎𝑙𝑒𝑛𝑐𝑒 𝑐𝑙𝑎𝑠𝑠.
But P has only one element from each equivalence class, therefore we must have 𝑝𝑖 = 𝑝𝑗 𝑖. 𝑒. , 𝑖 = 𝑗
But ≠ 𝑗 . Hence a contradiction.
Hence 𝑃𝑖 ∩ 𝑃𝑗 ≠ ∅, 𝑖 ≠ 𝑗.
that is, < Pi >is a pair wise disjoint sequence of sets.
(ii) Clearly each 𝑃𝑖 ⊂ [0, 1)
⋃𝑖 𝑃𝑖 ⊂ [0, 1). Let x be any element of [0, 1) = I.
But I is partitioned into equivalent classes therefore x lies in one of the equivalence classes.
 x is equivalent to an element say y of P.
 x-y is a rational number say ri.
 x-y = ri
 x = y + ri
0
= y + 𝑟𝑖 .

0
x ∈ 𝑃 + 𝑟𝑖

 x∈ 𝑃𝑖
 x is in some 𝑃𝑖 .
𝑇ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒 [0, 1) ⊆ ⋃ 𝑃𝑖
𝑖

𝑡ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒, [0, 1) = ⋃ 𝑃𝑖 .
𝑖
Now we prove P is non-measurable.
Assume that P is measurable, then clearly each Pi is measurable.
0
And m(Pi) = m (𝑃 + 𝑟𝑖 )

= m(P) for each i.

Therefore, 𝑚(⋃𝑖 𝑃𝑖 ) = ∑𝑖 𝑚(𝑃𝑖 ) = ∑∞
𝑖=0(𝑃)
0 𝑖𝑓 𝑚(𝑃) = 0
={
∞ 𝑖𝑓 𝑚(𝑃) > 0
20 Measure and Integration Theory
But

𝑚 (⋃ 𝑃𝑖 ) = 𝑚[(0, 1)] = 𝑙(0,1) = 1, 𝑐𝑜𝑛𝑡𝑎𝑑𝑖𝑐𝑡𝑖𝑜𝑛.

𝑖
𝑇ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒 𝑃 𝑖𝑠 𝑛𝑜𝑛 − 𝑚𝑒𝑎𝑠𝑢𝑟𝑎𝑏𝑙𝑒 𝑠𝑒𝑡.
1.48 Example. The cantor set is uncountable with outer measure zero.
Solution. We already know that cantor set is uncountable. Let Cn denote the union of the closed
intervals left at the nth stage of the construction. We note that Cn consists of 2𝑛 closed intervals, each
length 3−𝑛 . Therefore

𝑚∗ 𝐶𝑛 ≤ 2𝑛 . 3−𝑛 ( ∴ 𝑚∗ 𝐶𝑛 = 𝑚∗ (∪ 𝐹𝑛 ) = ∑ 𝑚∗ 𝐹𝑛 )

But any point of the cantor set C must be in one of the intervals comprising the union C n, for each n ∈
𝑁, and as such 𝐶 ⊂ 𝐶𝑛 for all n ∈ 𝑁. Hence
2 𝑛
𝑚∗ 𝐶 ≤ 𝑚∗ 𝐶𝑛 ≤ ( )
3
This being true for each n ∈ 𝑁, letting 𝑛 → ∞ gives 𝑚∗ 𝐶 = 0.
1.49 Example. If E1 and E2 are any measurable sets, show that
𝑀(𝐸1 ∪ 𝐸2 ) + 𝑚(𝐸1 ∩ 𝐸2 ) = 𝑚(𝐸1 ) + 𝑚(𝐸2 ).
Proof. Let A be any set. Since E1 is measurable,
𝑚∗ 𝐴 = 𝑚∗ (𝐴 ∩ 𝐸1 ) + 𝑚∗ (𝐴 ∩ 𝐸1 𝑐 ).
We set 𝐴 = 𝐸1 ∪ 𝐸2 and we have
𝑚∗ (𝐸1 ∪ 𝐸2 ) = 𝑚∗ [(𝐸1 ∪ 𝐸2 ) ∩ 𝐸1 ] + 𝑚∗ [(𝐸1 ∪ 𝐸2 ) ∩ 𝐸1 𝑐 ]
Adding 𝑚(𝐸1 ∪ 𝐸2 ) to both sides we have
𝑚(𝐸1 ∪ 𝐸2 ) + 𝑚(𝐸1 ∩ 𝐸2 ) = 𝑚𝐸1 + 𝑚∗ [(𝐸1 ∪ 𝐸2 ) ∩ 𝐸1 𝑐 ] + 𝑚(𝐸1 ∩ 𝐸2 ) … (1)
But
𝐸2 = [(𝐸1 ∪ 𝐸2 ) ∩ 𝐸1 𝑐 ] ∪ (𝐸1 ∪ 𝐸2 ).
Therefore
𝑚{[(𝐸1 ∪ 𝐸2 ) ∩ 𝐸1 𝑐 ] ∪ (𝐸1 ∪ 𝐸2 )} = 𝑚𝐸2
Hence (1) reduces to

𝑀(𝐸1 ∪ 𝐸2 ) + 𝑚(𝐸1 ∩ 𝐸2 ) = 𝑚(𝐸1 ) + 𝑚(𝐸2 ).

Measurable Sets 21
1.50 Theorem. Let E be any set. Then given ∈ > 0, there is an open set O ⊃ 𝐸 such that 𝑚∗ 𝑂 <
𝑚∗ 𝐸+ ∈ .
Proof. There exists a countable collection [ In ] of open intervals such that
𝐸 ⊂ ⋃𝑛 𝐼𝑛 and
∞

∑ 𝑙(𝐼𝑛 ) < 𝑚∗ 𝐸+ ∈ .
𝑛=1
∞

𝑝𝑢𝑡 𝑂 = ⋃ I𝑛 .
𝑛=1

Then O is an open set and

∞

𝑚∗ 𝑂 = 𝑚∗ (⋃ 𝐼𝑛 )
𝑛=1
∞

≤ ∑ m∗ In
𝑛=1
∞

= ∑ l(In ) < m∗ E+∈ .

n=1

1.51 Theorem. Let E be a measurable set. Given ∈> 0, there is an open set
O ⊃ E such that m∗ (O\E) < ∈.
Proof. Suppose first that m E < ∞. Then by the above theorem there is an open set O ⊃ E such that
m∗ O < m∗ E+ ∈
Since the sets O and E are measurable, we have
m∗ (O\E) = m∗ O − m∗ E < ∈.
Consider now the case when m E = ∞ . Write the set R of real number as a union of disjoint finite
intervals; that is,
∞

𝐑 = ⋃ 𝐈𝐧 .
𝐧=𝟏

Then, if En = E ∩ In, m(En) < ∞ . We can, thus, find open sets On ⊃ En such that
∈
m∗ (On − En ) < .
2n
∞

Define O = ⋃ On . Clearly O is an open set such that O ⊃ E and satisfies

n=1
22 Measure and Integration Theory
∞ ∞ ∞

O − E = ⋃ On − ⋃ En ⊂ ⋃(On − En )
n=1 n=1 n=1
∞
∗ (O
On
m − E) ≤ ∑ m∗ ( ) <∈ .
En
n=1

𝟏. 𝟓𝟐 𝐅𝛔 𝐚𝐧𝐝 𝐆𝛅 𝐒𝐞𝐭𝐬:
A set which is countable(finite or infinite) union of closed sets is called an 𝐹𝜎 sets. Note: The class of all
𝐹𝜎 sets is denoted by 𝐹𝜎. This F stands for ferme(closed) and 𝜎 for summe(sum).
Example: 1. A closed set.
2. A countable set
3. A countable union of 𝐹𝜎 set.
4. An open interval (a, b) since
∞ 1 1
(𝑎, 𝑏) = 𝑈𝑛=1 [𝑎 + 𝑛 , 𝑏 − 𝑛 ]and hence an open set.

𝑮𝜹- set:
A set which is countable intersection of open sets is a 𝐺𝛿 set.
Note: The class of all 𝐺𝛿 sets is denoted by This G stands for region and 𝛿for intersection. The
complement of 𝐹𝜎 set is a 𝐺𝛿 set and conversely.
Example: 1. An open set in particular an open interval.
2. A closed set
3. A countable intersection of 𝐺𝛿 set.
4. A closed interval [a, b] since
1 1
[𝑎, 𝑏] = ⋂∞
𝑛=1 (𝑎 − , 𝑏 + ).
𝑛 𝑛
1.53 Theorem. Let E be any set then
(a) Given ɛ > 0,∃an open set 𝑂 ⊃ 𝐸such that 𝑚∗ (𝑂) < 𝑚∗ (𝐸) + ɛ
(𝑏)∃𝑎 𝐺𝛿 𝑠𝑒𝑡 𝐺 ⊃ 𝐸such that m*(E) = m*(G).
Proof: (a) By definition, m* (𝐸) = 𝑖𝑛𝑓 ∑𝑛 𝑙 (𝐼𝑛 ), where 𝐸 ⊆ ⋃𝑛 𝐼𝑛
if m*(E) = ∞, then clearly result is true. If m*(E) < 0, there is a countable collection {In} of open
intervals such that
𝐸 ⊆ 𝑈𝑛 𝐼𝑛 and 𝑚∗ (𝐸) + ɛ > ∑𝑛 𝑙 (𝐼𝑛 ) (1)
Let 𝑂 = 𝐸 ⊆ 𝑈𝑛 𝐼𝑛 , then O is an open set and 𝑂 ⊃ 𝐸
Also 𝑚∗ (𝑂) = 𝑚∗ (𝐸 ⊆ 𝑈𝑛 𝐼𝑛 )
Measurable Sets 23

≤ ∑ 𝑚∗ (𝐼𝑛 )
𝑛

m*(O) < m*(E) + ɛ [from(1)]

1
(b) Take ɛ = 𝑛 ∀𝑛 ∈ 𝑁 Then by above part, for each 𝑛 ∈ 𝑁,∃an open set𝑂𝑛 ⊃ 𝐸such that

1
𝑚∗ (𝑂𝑛 ) < 𝑚∗ (𝐸) +
𝑛
Now define 𝐺 = ∪∞
𝑛=1 𝑂𝑛 , then G is a 𝐺𝛿 set.

Also, since each 𝑂𝑛 ⊃ E

therefore ∪∞
𝑛=1 𝑂𝑛 ⊃ 𝐸 this implies 𝐺 ⊃ 𝐸

≥ m*(E) ≤ m*(G) (2)

Also 𝐺 = ∪∞
𝑛=1 𝑂𝑛 ⊆ 𝑂𝑛 ∀𝑛

𝑚∗ (𝐺) ≤ 𝑚∗ (𝑂𝑛 ) for each n

1
< 𝑚∗ (𝐸) + 𝑛, for each n

in limiting case, we have

m*(G) ≤ m*(E) (3)
Then from (2) and (3), we have
m*(G) = m*(E).
1.54 Theorem. Let E be any set,then the following five statements are equivalent.
(i) E is measurable.
(ii) For given ɛ > 0 , ∃an open set 𝑂 ⊃ 𝐸 such that m*(O – E ) < ɛ
(iii) There exist a set G in 𝐺𝛿 with 𝐸 ⊂ 𝐺, m*(G – E ) = 0
(iv) For given ɛ > 0 , ∃ an closed set 𝐹 ⊂ 𝐸 such that m*(E – F ) < ɛ
(v) There exist a set F in 𝐹𝜎 with 𝐹 ⊂ 𝐸, m*(E – F ) = 0
Proof. Ist we prove (i) implies (ii)
Let E be a measurable set.
Now two cases arrive
Case (i) m*(E) < ∞.
By definition, for given ɛ> 0,there is a countable collection {In} of open intervals such that
𝐸 ⊆ ⋃𝑛 𝐼𝑛 and m*(E) + ɛ > ∑𝑛 𝑙 (𝐼𝑛 ).......(1)
Let O = 𝐸 ⊆ ⋃𝑛 𝐼𝑛 , then O is an open set and 𝑂 ⊃ 𝐸
24 Measure and Integration Theory
Also m*(O) =m*( 𝐸 ⊆ ⋃𝑛 𝐼𝑛 )
m*(O) ≤ ∑𝑛 𝑚 ∗ (𝐼𝑛 )
m*(O) < m*(E) + ɛ [from(1)]
m*(O)- m*(E) < ɛ
O = (𝑂 − 𝐸) ∪ 𝐸
m*(O) = m*((𝑂 − 𝐸) ∪ 𝐸) = m*(O-E)+m*(E)
m*(O-E) = m*(O) - m*(E)
⟹ m*(O-E) < ɛ
Case (ii) If m*(E) = ∞
We know that set of real number can be written as countable union of disjoint open intervals
∞

𝑅 = ⋃ 𝐼𝑛
𝑛=1

Then 𝐸 = 𝐸 ∩ 𝑅
∞

= 𝐸 ⋂ ⋃ 𝐼𝑛
𝑛=1
∞

= ⋃(𝐸 ∩ 𝐼𝑛 )
𝑛=1
∞

⟹ 𝐸 = ⋃ 𝐸𝑛 , 𝑤ℎ𝑒𝑟𝑒 𝐸𝑛 = 𝐸 ∩ 𝐼𝑛
𝑛=1

clearly each En is measurable and m(En ) is finite.

Because 𝐸𝑛 = 𝐸 ∩ 𝐼𝑛 ⊆ 𝐼𝑛
m*(En ) ≤ l(In) <∞
Then by case (i), for each 𝑛 ∈ 𝑁, ∃ an open set 𝑂𝑛 ⊃ 𝐸𝑛 such that
𝜀
m*(On - En) <2𝑛

Let us define 𝑂 = ⋃∞
𝑛=1 𝑂𝑛

Then O is an open set containing E

Now (O-E) = ⋃∞ ∞ ∞
𝑛=1 𝑂𝑛 − ⋃𝑛=1 𝐸𝑛 ⊂ ⋃𝑛=1(𝑂𝑛 − 𝐸𝑛 )

m*(O-E) ≤ m*(⋃∞
𝑛=1 (𝑂𝑛 − 𝐸𝑛 ) )

≤ ∑∞ ∗
𝑛=1 𝑚 (𝑂𝑛 − 𝐸𝑛 )
Measurable Sets 25
𝜀
≤ ∑∞
𝑛=1 2𝑛

= ɛ
⟹ m*(O-E) < ɛ
Now (ii) ⟹ (iii)
Let (ii) holds, then for each 𝑛 ∈ 𝑁, ∃an open set 𝑂𝑛 ⊃ 𝐸such that
1
m*(On-E) <𝑛
Let us define = ⋂∞
𝑛=1 𝑂𝑛 , then G is a 𝐺𝛿 set.

Also since each 𝑂𝑛 ⊃ 𝐸

therefore ⋂∞
𝑛=1 𝑂𝑛 ⊃ 𝐸 this implies 𝐺 ⊃ 𝐸

G-E = ⋂∞
𝑛=1 𝑂𝑛 − 𝐸 ⊆ 𝑂𝑛 − 𝐸
1
m*(G-E) ≤ m*(On-E) < 𝑛
Since n is arbitrary
m*(G-E) ≤ 0
⟹ m*(G-E) = 0.
Now (iii) ⟹ (i)
Let (iii) holds, then for given set E, ∃ a 𝐺𝛿 set 𝐺 ⊃ 𝐸 such that m*(G-E) = 0
⟹ G - E is measurable.
Now E = G- (G-E)
Now E is measurable being difference of two measurable sets.
Thus (i) ⟺ (ii) ⟺ (iii)
Now to show (i) ⟹ (iv)
Let (i) holds, and ɛ > 0 be given
𝑐
then by (ii), for given set Ec, ∃ an open set 𝐺 ⊃ 𝐸 such that m*(G - Ec) < ɛ
𝑐 𝑐
Since 𝐺 ⊃ 𝐸 ⟹ 𝐺 ⊆ 𝐸
Let F = Gc
then F is a closed set contained in E,
𝑐
Now E-F = 𝐸 ∩ 𝐹 = EՈG = GՈE = G – Ec
Now m*(E-F) = m*(G-Ec) < ɛ
m*(E-F) < ɛ.
To Show (iv) ⟹ (v)
Let (iv) holds, then for each 𝑛 ∈ 𝑁,∃a closed set 𝐹𝑛 ⊂ 𝐸 such that
26 Measure and Integration Theory
1
m*(E – Fn ) <𝑛

Let us define 𝐹 = ⋃∞
𝑛=1 𝐹𝑛

Then F is a 𝐹𝜎 set.
Also, since each 𝐹𝑛 ⊂ 𝐸 ⟹ ⋃∞
𝑛=1 𝐹𝑛 ⟹ 𝐹 ⊆ 𝐸

Now E – F = E - ⋃∞
𝑛=1 𝐹𝑛 ⊆ 𝐸 − 𝐹𝑛
1
⟹ m*(E-F) ≤ m*(E – Fn ) < 𝑛
1
⟹ m*(E-F) ≤ 𝑛

Since n is arbitrary.
m*(E-F) ≤ 0
⟹ m*(E-F) = 0.
Now (v) ⟹ (i)
Let (v) holds, then for general E, ∃ a 𝐹𝜎 set F such that m*(E-F) = 0
⟹ E-F is measurable.
𝐸 = (𝐸 − 𝐹) ∪ 𝐹
⟹ E is measurable.
This completes the proof.
1
(b) Take ɛ = 𝑛 ∀𝑛 ∈ 𝑁
Then by above part, for each 𝑛 ∈ 𝑁, ∃ an open set 𝑂𝑛 ⊃ 𝐸 such that
1
m*(On) < m*(E) + 𝑛
Now define 𝐺 = ⋂∞
𝑛=1 𝑂𝑛 , then G is a 𝐺𝛿− set.

Also since each 𝑂𝑛 ⊃ 𝐸

therefore⋂∞
𝑛=1 𝑂𝑛 ⊃ 𝐸

this implies 𝐺 ⊃ 𝐸
⟹ m*(E) ≤ m*(G) ......(2)
Also G = ⋂∞
𝑛=1 𝑂𝑛 ⊆ 𝑂𝑛 ∀𝑛
1
m*(G) ≤ m*(On) for each n < m*(E) +𝑛, for each n in limiting case, we have

m*(G) ≤ m*(E)...(3)
Then from (2) and (3), we have
m*(G) = m*(E).
Measurable Sets 27

1.55 Theorem. Let Ebe a set with m* E < . Then E is measurable iff given > 0 , there is a finite
union B of open intervals such that m*(E B) < 
Proof. Suppose E is measurable and let > 0 be given. The (as already shown) there exists an open set

O E such that m* (O- E) < . As m*E is finite, so is m*O. Since the open set O can be written as the
2
union of countable (disjoint) open intervals {Ii}, there exists an n  N such that

  

 l(I )  2
i  n 1
i (In fact m* O =   l(I )  
i  n 1
i   l(I )  2
i  n 1
i because m* O < )

Set B = ⋃𝑛𝑖=1 𝐼𝑖 . . Then E B = (E B) (B -E) (O - B) (O -E) . Hence
𝜀 𝜀
m*(E B)  m* (⋃𝑛𝑖=1 𝐼𝑖 ) + m*(O-E) + = 𝜀.
2 2

Conversely, assume that for a given > 0 , there exists a finite union B = ⋃𝑛𝑖=1 𝐼𝑖 . if open intervals with
m* (E B) < . Then using “Let be any set. The given > 0 there exists an open set O E such
that m* O < m* E + there is an open set O E such that
m* O < m* E + (i)
If we can show that m* (O E) is arbitrary small, then the result will follow from “Let E be
set. Then the following are equivalent (i) E is measurable and (ii) given > 0 there is an open set O E
such that m * (O E) < ”. Write S =⋃𝑛𝑖=1(𝐼𝑖 ∩ 𝑂). Then S B and so
S E = (E - S) (S - E) (E S) (B E) . However,
E \ S = ( E OC ) ( E BC ) = E B, because E O . Therefore
S E (E B) (B E) = E B , and as such m* (S E) < . However,
E S (S E)
and so m* E < m* S + m* (S E)
<m*S+ii)
Also,
O E = (O S) (S E)
Therefore
m* (O \ E) < m* O m* S + 
< m* E + m* S +  (using(i))
< m* S + + m*S + using(ii))
< m*S + + m* S + 
= 3 .
Hence E is measurable.