An-Najah National University

Faculty of Graduate Studies

Cantor Set in Measure Theory

BY
Alaa Jamal Moustafa Yaseen

Supervised By

Dr. Abdallah A. Hakwati

Co-Supervisor

Dr. Jasser H. Sarsour

Submitted in Partial Fulfillment of the Requirements for the Degree of

Master of Science in Mathematics, Faculty of Gradated Studies, at An-
Najah National University , Nablus, Palestine.

2005
 iii

Acknowledgements

I would like to express my deep gratitude to Dr. Abdallah

Elhakawati and Professor Jasser H. Sarsour for their true effort in

supervising and directing me to come with this thesis.

Also not to forget every one who made an effort to make this thesis

possible. Finally I would like to express my deep thanks to my family for

their encouragement and support.
 iv

Contents

Page
Acknowledgment iii
Table of content iv
Abstract v
Introduction ------------------------------------------------------------- 1

Chapter One: Preliminaries ---------------------------------------- 3

1.1 Cardinality ---------------------------------------------------------- 4
1.2 Topological properties -------------------------------------------- 6
1.3 Algebra of Sets ----------------------------------------------------- 9
1.4 Measure and Measurable Space --------------------------------- 9

Chapter Tow: Cantor Sets ------------------------------------------ 12

2.1 Introduction to Cantor Sets --------------------------------------- 13
2.2 Ternary Representation of Numbers ---------------------------- 14
2.3 Properties of Cantor Middle Third Set C1 / 3 -------------------- 20

Chapter Three: Measure and Cantor Sets ------------------------ 25

3.1 Measure of Cantor Sets ------------------------------------------- 26
3.2 Measure Zero Set With Non-measurable Sum ---------------- 32
3.3 Applications of the Cantor Middle Half Set C1 / 2 -------------- 40

References --------------------------------------------------------------- 46
‫ﺍﻟﻤﻠﺨﺹ ﺒﺎﻟﻠﻐﺔ ﺍﻟﻌﺭﺒﻴﺔ‬ ‫ﺏ‬
 v

Cantor Set in Measure Theory

BY
Alaa Jamal Moustafa Yaseen

Supervised By

Dr. Abdallah A. Hakwati

Co-Supervisor
Dr. Jasser H. Sarsour

Abstract

This thesis is a survey for the using of Cantor sets C1 / 3 and

C1 / 2 in measure theory. It is proved that C1 / 3 and C1 / 2 are
measurable and have zero measure. Following that it is shown
that the measure of C1 / 3 + C1 / 3 is positive and the measure of
C1 / 2 + C1 / 2 is zero. Also it is shown that there exists a subset A of
C1 / 3 such that A + A is non measurable. At the end of this thesis it
1
is shown that there is no subset B ⊂ C1 / 2 U C1 / 2 such that B + B
2
is Bernstein in [0,1.5] .
Introduction

This thesis will concentrate on the study of the behavior of some

sets in measure theory. Measure theory is a basis of modern theories of
integration. Lebesgue measure is a special case of it. Four main sets will be
discussed in this thesis:

(i ) Uncountable sets with zero measure.

(ii ) Set with zero measure but its algebraic sum has a positive measure.

(iii ) Set with zero measure, and also its algebraic sum has zero measure.

(iv) Measure zero set with non-measurable sum.

Thus algebraic sum can't characterize measure zero sets.

To achieve this study we need to look at surprising sets which are

the Cantor sets. The Cantor set which was defined by Cantor is a set of
length zero which contains uncountably many points. A perfect set does
not have to contain an open set. Therefore, the Cantor set shows that closed
subsets of the real line can be more complicated than intuition might at first
suggest. It is in fact often used to construct difficult, counter-intuitive
objects in analysis. For example the measure of the Cantor middle third set
C1 / 3 is zero and its sum has positive measure. While the measure of the
Cantor middle half set C1 / 2 is zero and its sum also has zero measure.
Moreover there exists a set A ⊆ C1 / 3 such that A + A is non-measurable.

The contents of this thesis are divided into three chapters. In the first
one we give some basic definitions and preliminary results that are used in
subsequent chapters. In the beginning of the second chapter we will study,
 2

in detail, the construction of the Cantor middle third and half sets. Finally
we will discuss the properties of the cantor middle third C1 / 3 sets.

In chapter three we will discuss the measure of the cantor sets and
will be introduced to the Cantor sets has positive measure. Following that
we will define a set A , subset of the C1 / 3 , which has zero measure, but
A + A is non-measurable. Finally we will present an application of the
C1 / 2 such that C1 / 2 + C1 / 2 = [0,1.5] , furthermore there is no subset
1
B ⊂ C1 / 2 ∪ C1 / 2 such that B + B is Bernstein in [0,1.5] .
2
 3

Chapter One

Preliminaries
 4

Chapter One

Preliminaries

In this chapter, we shall give necessary facts and definitions of

cardinal numbers, dense sets, Borel sets, lebesgue measure, and connected
space. The purpose of this chapter is to clarify terminology and notations
that we shall use throughout this thesis.

1.1 Cardinality

In this section we shall give required definitions and facts a bout

cardinality of sets.

1.1.1 Definition

Two sets A and B are equivalent if and only if there exists a one-to-
one function from A onto B . A and B are also said to be in one-to-one
correspondence, and we write A ≈ B . (See [4], p. 93)

We shall use the symbol N k to denote the set {1,2,3, L , k} . Each N k

may be thought of as the standard set with k elements since we shall
compare the sizes of other sets with them.

1.1.2 Definition

A set S is finite if and only if S = φ or S is equivalent to N k for

some natural number k . In the case S = φ , we say φ has cardinal number
zero and write φ = 0 . If S is equivalent to N k , then S has cardinal
number k and we write S = k .

A set S is infinite if and only if it is not finite. (See [4], p. 93)

 5

1.1.3 Definition

A set S is denumerable if and only if it is equivalent to the set of

natural numbers N . A denumerable set S has cardinal number ℵ0 and
write S = ℵ0. If a set is finite or denumerable, it is countable, otherwise

the set is uncountable. The symbol ℵ0 is the first infinite cardinal

number. Other infinite cardinal numbers are associated with uncountable
sets.

The interval (0,1) is an example of an uncountable set. The cardinal

number of (0,1) is defined to be c (which stands for continuum).

(See [4], p.97, 99)

1.1.4 Remark

2 N where N is a set of all natural numbers is the set of all functions

f : N → {0 ,1} . So the cardinal number of 2 N is the cardinal number of
the set of functions f : N → {0 ,1} . Therefore 2 N = c. (See [2], p. 9)

1.1.5 Fact

It is will known that n < ℵ0< c for all n ∈ N . And there are no sets
A for which ℵ0 < A < c. (See [2], p. 8, and 9)

1.1.6 Definition

A set Λ is a directed set if and only if there is a relation ≤ on Λ

satisfying:

a ) λ ≤ λ for each λ ∈ Λ ,

b) if λ1 ≤ λ 2 and λ 2 ≤ λ 3 then λ1 ≤ λ 3 ,
 6
c) if λ1 , λ 2 ∈ Λ then there is some λ 3 ∈ Λ with λ1 ≤ λ 3 , λ 2 ≤ λ 3 .

The relation ≤ is referred to as directed on Λ . (See [2], p.73)

1.1.7 Definition

A net in a set X is a function P : Λ → X where Λ is some directed set.

The point P (λ ) is denoted x λ , and we denote the net as ( x λ ) .(See[2],p.73)

1.2 Topological Properties

The topology in our thesis is the standard topology for the real
numbers R and in this section we will give basic definitions of topological
properties.

1.2.1 Definition

If X is a topological space and E ⊂ X , the closure of E in X is

the set E = cl ( E ) = ∩{K ⊂ X : K is closed and E ⊂ K } , also the interior of
E in X is the set E o = Int ( E ) = ∪{G ⊂ X : G is open and G ⊂ E} .

(See [2], p. 25, 27)

1.2.2 Definition

Let ( X , τ ) be topological space. A set D is dense in X if and only

if cl X D = X . Also a set E ⊂ X is said to be nowhere dense, if
(cl X E ) o = φ . That is cl X E has empty interior. (See [2, 5], p.109, 306)

1.2.3 Definition

A point x ∈ R is a cluster point (or a point of accumulation) of

 7
a subset S ⊆ R if each ε -neighborhood vε = ( x − ε , x + ε ) of x contains
at least one point of S distinct from x .

S ′ is the set which contains all cluster points of S . (See [6], p.59)

1.2.4 Note

If S ′ ≠ φ , then S is not a finite set.

1.2.5 Definition

A space X is disconnected if and only if there are disjoint non-

empty open sets H and K in X such that X = H ∪ K . We then say that
X is disconnected by H and K .

When no such disconnected exists, X is connected. (See [2], p.191)

1.2.6 Definition

If x ∈ X , the largest connected subset C x of X containing x is

called a component of x . It exists being just the union of all connected
subsets of X containing x . (See [2], p.194)

1.2.7 Definition

A space X is totally disconnected if and only if the component in

X are the points. Equivalently X is totally disconnected if and only if the
only nonempty connected subsets of X are the one point sets.

(See [2], p. 210)

 8

1.2.8 Definition

A space X is compact if and only if each open cover of X has a

finite subcover. (See [2], p.116)

1.2.9 Proposition

A subset S of real numbers is compact if and only if it is closed and

bounded. (See [6], p. 186)

1.2.10 Proposition

Suppose { A j } is a collection of sets such that each A j is non-empty,

compact, and A j +1 ⊂ A j . Then A = ∩ A j is non-empty. (See [7], p.2)

1.2.11 Definition

A metric space is an ordered pair ( M , ρ ) consisting of a set M

together with a function ρ : M × M → R satisfying for all x, y, z ∈ M :

a) ρ ( x, y ) ≥ 0 ,

b) ρ ( x, x) = 0 ; ρ ( x, y ) = 0 implies x = y ,

c) ρ ( x, y ) = ρ ( y, x) ,

d) ρ ( x, y ) + ρ ( y, z ) ≥ ρ ( x, y )

The function ρ is called a metric on M . (See [2], p. 16)

 9

1.3 Algebra of Sets

1.3.1 Definition

A collection of subsets of X is called an algebra of sets or a

Boolean algebra if

(i ) A ∪ B is in whenever A and B are in .

(ii ) X \ A is in whenever A is .(See [1], p. 17)

1.3.2 Definition

An algebra of sets is called a σ -algebra if every union of a

countable collection of sets in is again in . That is if < Ai > is a
∞
sequence of sets, then U Ai must again belong to . (See [1], p.18)
i =1

1.3.3 Definition

The smallest σ -algebra, which contains all of the open sets, is called
Borel algebra. And the Borel set is an element of a Borel algebra .

(See [1], p. 52)

1.4 Measure and Measurable Space

1.4.1 Definition

By a measurable space we mean a couple ( X , ) consisting of a set

X and a σ -algebra of subsets of X . A subset A of X is called
measurable (or measurable with respect to ) if A ∈ . (See [1], p. 253)
 10

1.4.2 Definition

A set function is a function that associates an extended real number

to each set in some collection of sets. (See [1], p.54)

1.4.3 Definition

A set function m that assigns to each set E in some collection Μ of

sets of real numbers a nonnegative extended real number mE called the
measure of E . (See [1], p. 54)

1.4.4 Definition
For any set A of real numbers consider the countable collection {I n }
of open intervals that cover A , we define the Lebesgue outer measure µ ∗ A
by µ ∗ A = inf { ∑ L( I n ) such that A ⊆ ∪ I n }.(See [1], p. 56)

1.4.5 Definition

A set E of real numbers is said to be lebesgue measurable if for each

set A of real numbers we have µ ∗ A = µ ∗ ( A ∩ E ) + µ ∗ ( A ∩ E c ) .

(See [1], p. 58)

For any sets A, B, E and E i : i = 1,2,3, L we have the following

properties:

i) µ ∗ϕ = 0 .

ii) If A ⊂ B then µ ∗ A ≤ µ ∗ B .
∞
iii) If E ⊂ U E i then µ ∗ E ≤ ∑i =1 µ ∗ E . (See [1], p. 288)
∞

i =1

1.4.6 Definition

Let µ be a measure on an algebra and µ ∗ the induced outer

measure. We define the inner measure µ ∗ induced by µ by setting
 11
µ ∗ A = sup[ µA − µ ∗ ( A \ E )], Where the supremum is taken over all sets
A ∈ for which µ ∗ ( A \ E ) < ∞ .(See [1], p.317)

1.4.7 Definition

A bounded set E is said to be measurable if µ ∗ E = µ ∗ E .

(See [1], p. 318)

1.4.8 Lemma

Let E be any bounded subset in the real numbers R , then

µ ∗ E ≤ µ ∗ E . If E belongs to an algebra , then µ ∗ E = µ ∗ E .

(See [1], p. 318)

1.4.9 Proposition

Let E and F be disjoint sets, then

µ∗ E + µ∗ F ≤ µ∗ (E ∪ F ) ≤ µ∗ E + µ ∗ F ≤ µ ∗ (E ∪ F ) ≤ µ ∗ E + µ ∗ F

(See [1], p. 320)

1.4.10 Note
Let E be any bounded subset in the real numbers, and µ is
Lebesgue measure on R ,

then µ ∗ E = sup{ µF : F ⊂ E , F closed} (See [1], p. 323)

1.4.11 Proposition

If A is countable, then µ ( A) = 0 . (See [1], p.58)

 12

Chapter Tow

Cantor Sets
 13

Chapter Tow

Cantor Sets

2.1 Introduction to Cantor Sets

In the years 1871-1884 Georg Cantor invented the theory of infinite

sets. In the process Cantor constructed a set which is called a "Cantor" set.

To construct the Cantor set, take a line and remove the middle third.
There are two line segments left. Take the remaining two pieces and
remove their middle thirds. Repeat this process infinite number of times.
The resulting collection of points is called a "Cantor" set. Indeed repeatedly
removing the middle third of every piece, we could also keep removing any
other fixed percentage (other than 0 % and 100 %) from the middle. The
resulting sets are all homeomorphic to the Cantor set, i.e. these sets are
topologically the same.

The Cantor set is an unusual object. The deletion process produces

an infinite set of points. On the other hand these points are uncountable,
also it has no interior point.(See [10], [11])

2.1.1 Remark

The Cantor set C is a totally disconnected compact metric space.

(See [2], p. 217)

Our study will be concentrated on Cantor middle third set C1 / 3 and

Cantor middle half set C1 / 2 .
 14
Cantor Middle Third set C1 / 3 :

Beginning with the unit interval I = [0,1] , define closed subsets

A1 ⊃ A2 ⊃ L in I as follows: we obtain A1 by removing the interval
(1 / 3,2 / 3) from I , A2 is then obtained by removing from A1 the open
intervals (1/9,2/9) and (7/9,8/9). In general, having An −1 , An is obtained by
removing the open middle thirds from each of the 2 n −1 closed intervals that
make up An −1 . The cantor middle third set is the subspace C1 / 3 = ∩ An
of I . (See [2], p. 121)

Cantor mMiddle Half Set C1 / 2 :

Start with the unit interval F0 = [0,1] . Remove the (open) middle
half-resulting in F1 = [0,1 / 4] ∪ [3 / 4,1] . Then repeat the process removing
the middle half of each of the intervals that remain. At stage n we get a set
Fn that is the union of 2 n intervals each being of length 4 − n . These are
∞
nested: F0 ⊃ F1 ⊃ F2 ⊃ L , so their intersection C1 / 2 = I Fn , is called the
n =0

cantor middle half set. (See [8], p.315)

2.2 Ternary Representation of Numbers

2.2.1 Definition

For any x ∈ [0,1] , x can be represented in the scale of some integer

b > 1 as x = (0.a1 a 2 a 3 L) b , where every a i is one of the integers
0, L , b − 1 . Also x can be represented by a convergent series as:
∞ ai
x=∑ i
: a i ∈ {0, L , b − 1} for every i = 1,2, L .
i =1 b
 15

The integer b is called the base of the scale. For b = 2 it is called a binary
expansion; and for b = 3 it is called a ternary expansion (See [9], p.941)

2.2.2 Definition

Let x , y ∈ [ 0 ,1 ] in base b expansion. Then x, y will be called

equivalent with respect to base b expansion and we write x ∼ b y if and
only if there is a b expansion of x and a b expansion of y such that the
two expansions disagree on only finitely many digits.

(See [3], p.788)

Rational with Respect to the Base b :

Let x ∈ [0,1] , then there exist a sequence < xn > , where

∞ xi
x n ∈ {0,1, L , b − 1} , and x = ∑ i
. If the expansion of x ends in a
i =1 b

sequence of zeros, then there exists m ∈ N such that x i = 0 for all

m x
i > m , hence x = ∑ ii : x i ∈ {0,1, L , b − 1} .
i =1 b

∞
b −1 1
Since for all r we have ∑ bi
=
br
, then
i = r +1
m −1 xi ∞
b −1 xm − 1
x = (∑ i
)+( ∑ i ) + ( )
i =1 b i = m +1 b bm

Therefore x has two possible base b expansions one ending in asserting of

0’s and the other ending in asserting of (b − 1) ’s. In the case b = 4 we will
say that x is quaternary rational, and according to the (2.2.2 Definition) all
such quaternary rationals are equivalents. (See [3], p.788)
 16

Ternary Rational:

Let x ∈ [0,1] , then there exist a sequence < x n > , where

∞ x
x n ∈ {0,1,2} , and x = ∑ ii . If the expansion of x ends in a sequence of
i =1 3

zeros, then there exists m ∈ N such that x i = 0 for all i > m , hence
m x m −1 x ∞
2 x −1
x = ∑ ii : x i ∈ {0,1,2} . So x = ( ∑ ii ) + ( ∑ i ) + ( m m )
i =1 3 i =1 3 i = m +1 3 3

Therefore x has two possible ternary expansions one ending in asserting of

0’s and the other ending in asserting of 2’s. In this case we will say that x

is ternary rational, and according to the (2.2.2 Definition) all such ternary
rationals are equivalents. (See [3], p. 6)

2.2.3 Remark

Let x ∈ [0,1] . If we represent x as a quaternary expansion, then x has

a unique representation except when x is quaternary rational. (See [6], p.
60)

2.2.4 Proposition

The Cantor middle third set is precisely the set of points in the
interval I having a ternary expansion without 1's. (See [2], p. 121)

Proof

Let's focus on the ternary representations of the decimals between 0 and 1.

Since, in base three, 1/3 is equivalent to 0.1, and 2/3 is equivalent to 0.2.
We see that in the first stage of the construction (when we removed the
middle third of the unit interval) we actually removed all of the real
numbers whose ternary decimal representation have a 1 in the first decimal
 17

place, except for 0.1 itself. (Also, 0.1 is equivalent to 0.0222... in base
three, so if we choose this representation we are removing all the ternary
decimals with 1 in the first decimal place.) In the same way, the second
stage of the construction removes all those ternary decimals that have a 1 in
the second decimal place. The third stage removes those with a 1 in the
third decimal place, and so on. (By noticing that 1/9 is equivalent to 0.01
and 2/9 is equivalent to 0.02 in base three.)
Thus, after every thing has been removed, the numbers that are left – that
is, the numbers making up the Cantor set – are precisely those whose
ternary decimal representations consist entirely of 0’s and 2’s. Then the

Cantor middle third set C1 / 3 is precisely the set of points in the interval I
∞ x
having a ternary expansion without 1’s i.e. x = ∑ ii : xi = 0,2 for all i .
i =1 3

2.2.5 Proposition

The Cantor set C1 / 2 is precisely the set of points in the interval I

having a quaternary expansion without 1's and 2's. (See [8], p.316)

Proof

Since in base four expansion, 1 / 4 is equivalent to 0.1 and 3 / 4 is

equivalent 0.3 . We see that in the first stage of construction (when we
removed the middle half of the unit interval) we actually remove all
elements x ∈ [0,1] such that 0.1 < x < 0.3 , that is we remove all of the real
numbers whose four decimal representation is 1 and 2 in first decimal
place, except for 0.1 itself. (Also 0.1 is equivalent to 0.0333 . . . in base
four, so we choose the representation in which we are removing all the four
decimals with 1 and 2 in the first decimal place). In the same way, the
 18

second stage of the construction removes all those fourth decimals that
have a 1 and 2 in the second decimal place. The third stage removes those
with a 1 and 2 in the third decimal place, and so on. (By noticing that 1/16
is equivalent to 0.01 and 3/16 is equivalent to 0.03 in base four expansion).
finally all numbers left, making up the Cantor middle half set C1 / 2 are
precisely those whose four decimal representations which consist entirely

of 0's and 3's. Thus C1 / 2 is the set of points, x , in the unit interval such
that there is a base four expansion of x that uses only zeros and threes.

That is x = ∑ x i / 4 i : x i = 0 , 3 ∀ i = 1, 2 , 3 , L .

The Cantor middle third set C1 / 3 at least contains the endpoints of

all of the intervals that make up each of the sets An , that is since by
removing open middle thirds, then for every n ∈ N , 0 ∈ An and hence
0 ∈ C1 / 3 . The same argument shows that 1 ∈ C1 / 3 . In fact, if y is the
endpoint of some closed interval of some particular set An , then it is also
an endpoint of one of the intervals of An +1 for all n .

2.2.6 Proposition

Each of the Cantor middle third set C1 / 3 and Cantor middle half set
C1 / 2 is 1) Closed, 2) Dense in it self, 3) and of no interior.

(See [2], p. 217)

Proof

It is enough to prove it for the Cantor middle third set C1 / 3 because

the other proof for C1 / 2 is similar.

1) Cantor middle third set is closed

 19

From the construction of the Cantor middle third set

C1 / 3 = ∩ An , since each sets An can be written as a finite union of

2 n closed intervals, each of which has a length of1 / 3 n , as follows:

• A0 = [0,1]

• A1 = [0,1 / 3] ∪ [2 / 3,1]

• A2 = [0,1 / 9] ∪ [2 / 9,3 / 9] ∪ [6 / 9,7 / 9] ∪ [8 / 9,1]

• ...

Since An is a finite union of closed sets, An is a closed set for all n ∈ N ,

then C1 / 3 is an intersection of closed sets, therefore C1 / 3 is a closed set.

2) Cantor middle third set is dense in itself

All endpoints of every subinterval will be contained in C1 / 3 . Take

any x ∈ C1 / 3 = ∩ An then x is in An for all n , so x must be contained in
one of the 2 n intervals that comprise the set An . Define x n to be the left
endpoint of that subinterval (if x is equal to that endpoint, then let x n be
the right endpoint of that subinterval). Since each subinterval has length
1 / 3 n , we have:

x − x n < 1 / 3 n . Hence, the sequence ( x n ) converges to x , and since all

endpoints of the subintervals are contained in the Cantor set, we have

found a sequence of numbers not equal to x contained in C1 / 3 that
converges to x . Therefore, x is a limit point of C1 / 3 . But since x was
arbitrary, every point of C1 / 3 is a limit point of it. Thus C1 / 3 is dense in
itself.
 20
3) C1 / 3 has no interior point

Assume that there exists x ∈ Int (C1 / 3 ) , then there exists an ε > 0 such that
( x − ε , x + ε ) ⊂ C1 / 3 . Choose n∈ N such that 3−n < ε , then
( x − ε , x + ε ) ⊄ An . Therefore ( x − ε , x + ε ) ⊄ C1 / 3 , and this contradicts
that x ∈ IntC1 / 3 . Therefore C1 / 3 has no interior point.

2.2.7 Corollary

C1 / 3 does not contain any open set

2.3 Properties of Cantor Middle Third Set C1 / 3

2.3.1 Proposition

The Cantor middle third sets C1 / 3 , and C1 / 2 are compact

(See [2], p. 216)

Proof

From (2.2.5 Proposition part 1) C1 / 3 is closed, also it is bounded.

Since by (1.2.9 Proposition) every closed bounded subset of the real

numbers is compact, then C1 / 3 is compact

And similarly C1 / 2 is also compact.

2.3.2 Proposition

The Cantor middle third sets C1 / 3 , and C1 / 2 are uncountable.

(See [2], p. 217)

Proof
 21
By using Binary expansion for all y ∈ [0,1] , and ternary expansion
for all x ∈ C1 / 3 define the function

f : C1 / 3 → [0,1]

x yi
∑ 3ii →∑
2i
, where y i = x i if x i = 0 and y i = 1 if x i = 2

We have f is one-to-one and onto. Since [0,1] is uncountable, then C1 / 3 is

also uncountable. Similarly, C1 / 2 is uncountable also.

2.3.3 Lemma

Let x ∈ (0,2) then x has c many representations

x = a + b such that a, b ∈ (0,1) .

Proof

x x
Let x ∈ (0,2) be arbitrary. Let δ = min { − 1, }.
2 2

Then for all 0 < ε < δ / 2 , we have [ x 2 + ε , x 2 − ε ] ⊂ (0,1) and

x 2 + ε + x 2 − ε = x .Since we have uncountably number of ε , So x
has c many representations.

2.3.4 Proposition

The Cantor set C1 / 3 when added to itself gives the interval [0,2] .
(See [1], p.783)

Proof

Since C1 / 3 ⊆ [0,1], then C1 / 3 + C1 / 3 ⊆ [0,2] … … … … … (∗)

 22
Let x ∈ [0,2] be arbitrary and let c = x 2 , therefore c ∈ [0,1], so by using
ternary representation of c there exists a sequence < c i > such that
∞ c
c = ∑i =1 ii : c i = 0,1,2 .
3

Since x ∈ [0,2] , then x = 0 , 2 or 0 < x < 2

If x = 0, or 2 then we are done because 0 and 1 ∈ C1 / 3 and

0+ 0= 0 also 1+1=2.

Now if x ∈ (0,2), then by (2.3.3 Lemma) we have c many

representation of x .

Let x = a + b = c + c such that a, b ∈ (0,1),

Using the ternary representation for a, b, and c we have

a = ∑ a i / 3 i , b = ∑ bi / 3 i : a i , bi = 0,1,2 ∀ i = 1,2,3, L

Then x = ∑ a i 3 i + ∑ bi 3 i = ∑ c i 3 i + ∑ c i 3 i

= ∑ (c i + c i ) 3 i

If c i = 0 then a i = bi = 0

If c i = 1 then a i = 0 and bi =2 or

a i = 2 and bi = 0

If c i = 2 then a i = bi = 2.

Then we get

x = ∑ a i 3 i + ∑ bi 3 i , Such that a i and bi ≠ 1 for all i ,

 23
So ∑a i
3 i and ∑b i
3 i are in the Cantor middle third set.

That is ∀x ∈ [0,2] we have x = a + b such that a and b in the Cantor

middle third set.

Therefore [0,2] ⊆ C1 / 3 + C1 / 3 … … … … … … … … … (∗∗)

From (∗) and (∗∗) we get C 1 / 3 +C1 / 3 = [0,2] .

2.3.5 Definition

A subset P of R is perfect set if it is closed and dense in itself:

i.e., each point of P is an accumulation point of P ; that is: P is perfect in

R if and only if P = P ′ . (See [2], p. 216)

2.3.6 Remark

Every closed interval [a, b] , a ≠ b is a perfect set.

2.3.7 Proposition

Every non-empty perfect set is uncountable. (See [7], p. 165)

Proof

If P is perfect, it consists of accumulation points, and therefore can

not be finite. Therefore it is either countable or uncountable. Suppose P
was countable and could be written as P = {x1 , x 2 , x 3 , L}

The interval U 1 = ( x1 − 1, x1 + 1) is a neighborhood of x1 . Since x1 must be

an accumulation point of P , Therefore there are infinitely many elements
of P contained in U 1 .
 24
Take one of those elements, say x 2 and take a neighborhood U 2 of x 2
such that U 2 = ( x2 − δ , x2 + δ ) where
δ = 1 / 2 min{ x 2 − x1 , x 2 − ( x1 + 1) , x 2 − ( x1 − 1) }. Then the closure
(U 2 ) is contained in U 1 and x1 ∉ clU 2 . Again, x 2 is an accumulation point
of P , so that the neighborhood U 2 contains infinitely many elements of
P.

Select an element, say x 3 , and by the same argument take a neighborhood

U 3 of x 3 such that closure ( U 3 ) is contained in U 2 and x 2 ∉ clU 3 so x1
and x 2 are not contained in closure (U 3 ) .

Continue in that fashion to find sets U n and points x n such that:

• Closure (U n +1 ) ⊂ U n

• x j is not contained in U n for all 0 < j < n

• x n is contained in U n

Now consider the set V = ∩( closure ( U n ) ∩ P)

Since each set (closure (U n ) ∩ P) is closed and bounded, it is compact.

Also, by construction, (closure (U n +1 ) ∩ P) ⊂ ( closure (U n ) ∩ P) .

Therefore, by (1.2.10 proposition),V is not empty. Also x n ∉ V because

x n ∉ U n +1 for all integer numbers n . But V is non-empty, therefore P is
not countable.

2.3.8 Note

The Cantor middle third set C1 / 3 as well as C1 / 2 is closed and dense

in itself, then it is a perfect set, hence is uncountable.
 25

Chapter Three

Measure and Cantor Sets

 26

Chapter Three

Measure and Cantor Sets

Different measurable sets can be seen in mathematics, measurable

sets, non-measurable sets, and measure zero sets. We sow that C1 / 3 is
uncountable with zero measure but ( C1 / 3 + C1 / 3 ) has positive measure.
While C1 / 2 and ( C1 / 2 + C1 / 2 ) have zero measures. Another Cantor set with
positive measure will be studied. Also we will define a set A , subset of
C1 / 3 with measure zero, but A + A is non-measurable. Finally we will see
applications of C1 / 2 .

3.1 Measure of Cantor Sets

3.1.1 Proposition

The Cantor set C1 / 3 is measurable and has measure zero.

(See [1], p. 64)

Proof

Since C1 / 3 = ∩ An , then C1 / 3 is a countable intersection of closed

sets, therefore C1 / 3 is a Borel set, so C1 / 3 is measurable.

From the construction of C1 / 3 we have for every stage n > 0 we remove

2 n −1 disjoint intervals from each previous set each having length 1 / 3 n .
Therefore we will removed a total length of
∞
1 1 ∞

∑
n =1
2 n −1
3 n
=
3
∑
n =1
( 2 / 3 ) n −1

∞
= 1 / 3∑ (2 / 3) n
n =0
 27
1⎛ 1 ⎞
= ⎜ ⎟ =1
3 ⎝1 − 2 / 3 ⎠

from the unit interval [0,1].

Then C1 / 3 is obtained by removing a total length 1 from the unit interval

[0,1] , so µ ( I \ C1 / 3 ) = 1 . Since µ ( I ) = µ (C1 / 3 ) + µ ( I \ C1 / 3 ) , then
µ (C1 / 3 ) = µ ( I ) − µ ( I \ C1 / 3 ) = 1 − 1 = 0

Therefore the set C1 / 3 is measurable and has zero measure.

3.1.2 Remark

The measure of the C1 / 3 is zero, but µ (C1 / 3 + C1 / 3 ) ≠ 0 , since

C1 / 3 + C1 / 3 = [0,2] .

3.1.3 Proposition

The Cantor C1 / 2 is measurable and has zero measure.

Proof

Since C1 / 2 = ∩ Fn , that is C1 / 2 is countable intersection of closed

sets, therefore it is Borel set. Hence C1 / 2 is measurable

From the construction of C1 / 2 we have for stage n > 0 we remove 2 n −1

disjoint interval from each previous set each having length (2 / 4) n .
Therefore we will remove a total length of
∞ ∞
2 2 2
∑ 2 n −1
4n
=
4
∑ ( 4 ) n −1
n =1 n =1

1 ∞
1
=
2
∑
n = 0
(
2
) n
 28
1⎛ 1 ⎞
= ⎜ ⎟ =1
2 ⎝1 − 1/ 2 ⎠

from the unit interval [0,1] .Then C1 / 2 is obtained by removing a total

length 1 from the unit interval [0,1] . and µ ( I \ C1 / 2 ) = 1, Then
µ (C1 / 2 ) = µ ( I ) − µ ( I \ C1 / 2 ) = 1 − 1 = 0.

Thus C1 / 2 is measurable, and has zero measure.

3.1.4 Proposition

If A is the set of points in the unit interval [0,1] having a four

expansion without 3's, then A has zero measure.

Proof

Take the representation: [0,1] = [0,0.1] ∪ [0.1,0.2] ∪ [0.2,0.3] ∪ [0.3,1]

At first we need to show that, µ ([0,0.1] \ A) = 1 / 4

xi
To do so for all x ∈ A , x = ∑ x i = 0,1,2
4i

ai
• Let A1 = [0.03,0.1] , since for all a ∈ A1 , a = ∑ : a 2 = 3 , then a ∉ A ,
4i
so A ∩ A1 = ϕ and µ ( A1 ) = 4 −2

• Let A2 = [0.003,0.01] ∪ [0.013,0.02] ∪ [0.023,0.03] , since for all a ∈ A2 ,

a
a = ∑ ii : a 3 = 3 , then a ∉ A , so A ∩ A2 = ϕ and µ ( A2 ) = 3 ∗ 4 −3
4

• Let A3 = [0.0003,0.001] ∪ [0.0013,0.002] ∪ [0.0023,0.003] ∪

[0.0103,0.011] ∪ [0.0113,0.012] ∪ [0.0123,0.013] ∪

[0.0203,0.021] ∪ [0.0213,0.022] ∪ [0.0223,0.023] ,

 29
ai
since for all a ∈ A3 , a = ∑ : a4 = 3 , then a ∉ A , so A ∩ A3 = ϕ ,
4i
and µ ( A3 ) = 3 2 ∗ 4 −4 .

In general we have An is the union of 3 n −1 disjoint closed intervals each

a
have length 4 − ( n +1) and for all a ∈ An , a = ∑ ii : a n +1 = 3 , then
4
A ∩ An = ϕ and µ ( An ) = 3 n −1 ∗ 4 − ( n +1)

Since µ (∪ An ) = ∑ µ ( An ) ,

Then µ ( An ) = 1 / 4 2 + 3 / 4 3 + 3 2 / 4 4 + 3 3 / 4 5 + L

= 1 / 4 2 + (3 / 4 3 (1 + 3 / 4 + 3 2 / 4 2 + 3 3 / 4 3 + L))

⎡ ⎛ 1 ⎞⎤
= 1 / 4 2 + ⎢3 / 4 3 ⎜ ⎟⎥ = 1 / 4
⎣ ⎝ 1 − 3 / 4 ⎠⎦

Thus µ ([0,0.1] \ A) = 1 / 4 , By the same way we see that

µ ([0.1,0.2] \ A) = 1 / 4 , µ ([0.2,0.3] \ A) = 1 / 4 and µ ([0.3,1] \ A) = 1 / 4

Therefore µ ( A) = µ ( I ) − µ ( I \ A)

= 1 − [µ ([0,0.1] \ A) + µ ([0.1,0.2] \ A) + µ ([0.2,0.3] \ A) + µ ([0.3,1] \ A)]

= 1 − [1 / 4 + 1 / 4 + 1 / 4 + 1 / 4] = 0 .

3.1.5 Proposition

C1 / 2 + C1 / 2 has measure zero. (See [3] p.790)

Proof

If we prove that 1 / 3(C1 / 2 + C1 / 2 ) has measure zero, then obviously

C1 / 2 + C1 / 2 has measure zero. Let x ∈ (C1 / 2 + C1 / 2 ) , then there exist
 30
a , b ∈ C 1 / 2 such that x = a + b . Let c ∈1 / 3(C1 / 2 + C1 / 2 ) , then
1 1 1
c = x = a + b . Using base four expansion of a , b and c with all
3 3 3
digits of a and b are divisible by 3,then

c 1 a b
∑ 4ii =
3
(∑ ii + ∑ ii ) : a i , bi = 0,3
4 4

1
=
3
∑ ( a i + b i ) / 4 i : a i , b i = 0 , 3 , Hence

• c i = 1/3 (0 + 0) = 0 when a i = bi = 0

• c i = 1/3 (0 + 3) = 1 when a i = 0 and bi = 3

• c i = 1/3 (3 + 0) = 1 when a i = 3 and bi = 0 or

• c i = 1/3 (3 + 3) = 2 when a i = bi = 3 .

1 1 c
Therefore c = a + b = ∑ ii : c i = 0,1,2 . Hence c i ≠ 3 for all
3 3 4
i = 1,2,3, L . Then unless c is a quaternary rational its expansion will never
use the digits three, and ( by 2.2.3 remark )the expansion of c is unique.
Let S be all quaternary rationals in 1/3 ( C1 / 2 + C1 / 2 ).

xi
Thus for all x ∈ 1 / 3 ( C 1 / 2 + C 1 / 2 ) \ S , x = ∑ : x i = 0,1,2 . Since
4i
x i ≠ 3 for all i = 1,2,3, L Then by (3.1.4 proposition) 1 / 3(C1 / 2 + C1 / 2 ) \ S
has measure zero, and since all elements in S are quaternary rational, then
S ⊂ Q , hence S is countable, then 1 / 3 ( C 1 / 2 + C 1 / 2 ) has measured zero.
Therefore C1 / 2 + C1 / 2 has zero measure.
 31

3.1.6 Remark

The measure of the Cantor middle third set C1 / 3 is zero and its sum
has positive measure, while the measure of the Cantor middle half set and
its sum have zero measure.

3.1.7 Remark

There exists a Cantor set with positive measure. This can be

constructed as follows:

Start with the unit interval for I = [0,1] and choose a number
0 < p < 1 . Let

2− p 2+ p 1 p 1 p
R1 = ( , )=( − , + )
4 4 2 4 2 4

Which has measure p / 2 . Again, define C1 = I \ R1 . Now define

2− p 2+ p 14 − p 14 + p
R2 = ( , )∪( , )
16 16 16 16

Which has measure p / 4 ; continue as before, such that each R k has

p
measure k ; note again that all the Rk are disjoint. The resulting Cantor set
2
has measure
∞ ∞ ∞
µ ( I \ U Rn ) = 1 − ∑ µ ( Rn ) = 1 − ∑ p 2 −n = 1 − p > 0
n =1 n =1 n =1

Thus we have a continuum many of Cantor sets of positive measures.

(See [12])
 32

3.1.8 Note

Clearly Cantor sets can be constructed for all sorts of "removals", we

can remove middle halves, or thirds, or any amount 1 / r , r > 1. All of these
Cantor sets have measure zero. The key point is that:

If at each stage we remove 1 / r of each of the remaining intervals, the

results is a set of measure zero. The favored examples are the Cantor sets
which constructed by Georg Cantor " C1 / 2 , and C1 / 3 " , are constructed for
the case r = 2 and r = 3 respectively.

However it is possible to construct Cantor sets with positive measure

as well; the key is to remove less and less as we proceed, for example;
remove 1 / 3 then 1 / 3 2 of each remaining parts then 1 / 3 3 of each of
remaining parts … and so on. The result is a set of positive measure

These Cantor sets have the same topology as the Cantor set and the same
cardinality but different measure.

3.2 Measure Zero Set with Non-Measurable Sum

From previous studies we notice that there exists a measure zero set
so that its sum also has zero measure, and a set with measure zero but its
sum has positive measure such as C1 / 2 and C1 / 3 respectively. In this
section we will discuss a special subset A of C1 / 3 that has measure zero
with non-measurable sum.

3.2.1 Definition

We say that B ⊆ X is a Bernstein set (in X ) provided B and X \ B

intersect every non-empty perfect subset of X . (See [3], p.2)
 33

3.2.2 Proposition

Any closed set B in can be written as B = P ∪ D where P is

perfect and D is countable. (See [13], p.3596)

Proof

If B is countable then take the perfect set P = ϕ , and D = B then

we are done. Otherwise

Let P = {x ∈ B : ∀ε > 0, ( x − ε , x + ε ) ∩ B is uncountable} and

D = B \ P . First need to show that P is closed set,

Let ( x n ) be any convergent sequence to x such that x n ∈ P for all n ∈ N ,

since P ⊂ B and B is closed, then x ∈ B . Let ε > 0 be arbitrary, then there
exists j > 0 such that x j ∈ ( x − ε , x + ε ) ∩ B , since x j ∈ P , and
( x − ε , x + ε ) is neighborhood of x j , therefore ( x − ε , x + ε ) ∩ B is

uncountable., So x ∈ P , then P is closed.

Now need to prove P is dense in itself,

Let x ∈ D , then we can find two rational numbers a and b such that
a < x < b and (a, b) ∩ B is countable. Since there are only countably many
open intervals {On }n =1 with rational end points, then there exist countable
numbers of open intervals O j such that O j ∩ B is countable, so
D ⊂ U n (On ∩ B) . But countable union of countable set is countable,
therefore D is countable set. Let x ∈ P be arbitrary, then for all ε > 0
( x − ε , x + ε ) ∩ B is uncountable, and since D is countable therefore
(x − ε , x + ε ) ∩ B \ D is also uncountable. Since P = B \ D , then
( x − ε , x + ε ) ∩ P is uncountable and so x is a cluster point, consequently
P is dense in itself.
 34

3.2.3 Proposition

Let P be a nonempty perfect set, then for any y ∈ P , (−∞, y ] ∩ P or

[ y, ∞) ∩ P is a perfect set.

Proof

Let P be any nonempty perfect set. Let y ∈ P be arbitrary and

define A = P ∩ (−∞, y ], B = P ∩ [ y, ∞) , for all t ∈ A and t < y , t ∈ A′
because if t < y , then for any ε > 0 (t − ε , t + ε ) ∩ (−∞, y ) = (a, b) where
b = min{ y, t + ε } and a = t − ε , then t ∈ (a, b),

Hence (a, b) ∩ P = (t − ε , t + ε ) ∩ (−∞, y ) ∩ P ⊂ (t − ε , t + ε ) ∩ A

Since t ∈ P ′ then (a, b) ∩ P \ {t} ≠ φ , so t ∈ A′

Similarly for all t ∈ B and t > y , t ∈ B ′ .

If y ∉ A′ and y ∉ B ′ then there exists ε 1 , ε 2 > 0 such that

( y − ε 1 , y + ε 1 ) ∩ A \ { y} = φ and ( y − ε 2 , y + ε 2 ) ∩ B \ { y} = φ

Let ε = min{ε 1 , ε 2 } then, ( y − ε , y + ε ) ∩ P \ { y} = φ , but P = A ∪ B , then

( y − ε , y + ε ) ∩ ( A ∪ B) \ { y} = φ .

Therefore ( y − ε , y + ε ) ∩ p \ { y} = φ and this is a contradiction because

y ∈ P = P ′ . Then y ∈ A′ or B ′ .

If y ∈ A′ then A = A′ hence A is perfect and if y ∈ B ′ then B = B ′ hence

B is perfect, thus A or B must be perfect.
 35

3.2.4 Corollary

If P is a non empty perfect set, then { P ′ ⊂ P : P ′ non empty and

perfect} is uncountable. And for every y ∈ P we can find a perfect subset
p y of P such that for all x ≠ y Px ≠ Py .

3.2.5 Lemma

Every Bernstein subset of an interval I of real numbers has inner

measure zero and the same outer measure as I .

Proof:

Let I be a non empty open or open half interval in R , and B be any

Bernstein subset in I . Since for any non empty open or open half interval
I there exists a closed interval [a, b] subset in I where a ≠ b , and since I
is perfect set. Then by (3.2.3 proposition, and 3.2.4 Corollary) there exist
uncountably many perfect subsets of I .

Let { pξ : ξ < c} be the family of all non-empty perfect subset of I .

If the inner measure of B is not zero then by (1.4.10 note) B must contain
a closed set of positive measure, since the measure of any countable set is
zero and the closed set with positive measure is uncountable set, so by
(3.2.2 proposition) this closed set must have a non empty perfect subsets.
Then there exists ξ < c such that pξ ⊂ B , so pξ ∩ ( I \ B ) = φ and this

contradicts that B is Bernstein subset in I .

Then B has inner measure zero.

Now if B dose not have the same outer measure as I , then by (1.4.9
proposition) µ ∗ ( I \ B) + µ ∗ B ≥ µ ∗ (( I \ B) ∪ B) = µ ( I ). Therefore
 36
µ ∗ ( I \ B) ≥ µ ( I ) − µ ∗ ( B) > 0 , that is ( I \ B ) has positive inner measure,
then there exists a closed subset in ( I \ B ) with positive measure.
Therefore this closed set contains a non empty perfect subset. Then there
exist ξ < c such that pξ ⊂ ( I \ B ) and pξ ≠ φ , therefore pξ ∩ B = φ

and again, this is contradicts that B is Bernstein subset in I , thus B must

have the same outer measure as I .

3.2.6 Theorem

Every Bernstein subset of an interval in R is not measurable.

Proof

Let B be any Bernstein subset in an interval I . Then by (3.2.4

lemma) µ ∗ ( B) = 0 , and µ ∗ ( B) = µ ∗ ( I ) . Since µ ∗ (I ) > 0 , then
µ ∗ ( B) ≠ µ ∗ ( B) . Hence B is non-measurable.

3.2.7 Lemma

Let x ∈ [0,2] . If x / 2 has infinitely many ones in its ternary

expansion, then there are c many representations of x as the sum of two
Cantor-set elements; otherwise, x has only finitely many such
representations and all of the elements of C1 / 3 used to represent x are
equivalent to x / 2 . (See [3], p.788)

Proof

Since every element of C1 / 3 has a ternary expansion consisting of

only even digits, then for any x ∈ [ 0 , 2 ] = C 1 / 3 + C 1 / 3 there exists
a, b ∈ C1 / 3 such that x = a + b . Let c = x / 2 then c is the average of a
 37
and b. If c i , a i , bi are the i th digits of c , a , b respectively, since
x = c + c = a + b , then
∞ ci + ci ∞ a i + bi
∑ =∑ .
i =1 3i i =1 3i
∞ 2c i − (a i + bi )
Therefore ∑ =0
i =1 3i

So 3 | ( 2 c i − ( a i + b i ) ∀ i

Then for all i we have

• c i = 0 and a i = b i = 0 ,

• c i = 1 and a i = 0, bi = 2 ,

• c i = 1 and a i = 2 , bi = 0 , or

• c i = 2 and a i = bi = 2

If c is ternary rational, then the digits of c must end in either a sequence

of zeros or a sequence of two's that is from the property of ternary
rational. In either case the digits of a and b must do likewise since when
c i = 0 we have a i = b i = 0 and when c i = 2 we have a i = b i = 2 .
And so they are also ternary rationals therefore a ∼ b ∼ c , since all
ternary rational are equivalent.

Now consider the case c is not a ternary rational, so there is a unique

ternary expansion of c .

Let us construct the numbers a and b using only even digits for each c i
that is zero or two, we must have c i = a i = bi . But for each c i = 1 , we
 38
have a choice either a i = 0, bi = 2 or a i = 2, bi = 0 , two choice for
a i & b i . Thus if k ∈ { 0 ,1, 2 , L , n } is the number of digits in c that
have the value 1, then there are 2 k possible choices for the pair a, b . In
particular if c has infinitely many ones in its expansion then there are
2 N = c many representation for x . If there are only finitely many ones

then the digit of a , b , c will all agree on a tail end. That is a , b , c

disagreeing on only finitely many digits.

3.2.8 Proposition

There is a set A ⊆ C1 / 3 such that A + A is Bernstein in

[0,2] = C1 / 3 + C1 / 3 , hence A + A is non-measurable. (See [3], p.789)

Proof

Let R0 be the set of elements of [0,2] that can be expressed in c

many ways as the sum of elements of C1 / 3 . And R1 be the elements that
can be expressed in only finitely many ways. We can construct R0 and R1
by (2.3.7 lemma).

Let { p ξ : ξ < c } be the family of all non-empty perfect subsets of [0,2] .

We will find an A ⊆ C1 / 3 such that each pξ intersects both A + A and its

complement. Construct a net

( < a ξ , bξ , c ξ , d ξ >∈ C × C × p ξ × p ξ : ξ < c )

Such that for each ξ < c,

(∗) c ξ = a ξ + b ξ and D ξ ∩ ( A ξ + 1 + A ξ + 1 ) = ϕ ,

Where A ξ = U η < ξ { a η , bη } and D ξ = { d η : η ≤ ξ } .

 39
This will ensure that A = Ac = Uη < c {aη , bη } has the desired properties,

since then

{ c ξ : ξ < c} = {aη + bς : aη , bς ∈ A for all η , ς < c } ⊂ A + A

And A + A ⊆ [ 0 , 2 ] \ D ξ , then

{ c ξ : ξ < c } ⊆ A + A ⊆ R \ D ξ , therefore

{cξ : ξ < c } ⊂ pξ ∩ A + A and Dξ ⊂ pξ ∩ ([0,2] \ A + A)

That is each pξ intersects both A + A and its complement. So A + A is a

Bernstein set, hence non measurable.

To make an inductive step, assume that for some α < c we have already
constructed a partial net which satisfies (∗) for all ξ < ∝.

Now we need to show that we can construct a partial net which satisfies
(∗) for ξ = α . We first choose aα , bα , cα such that aα + bα = cα and
neither aα nor bα is in Dα − Aα .

We decide two cases.

Case 1: pξ intersects R0 in a set of cardinality c. Since ( Pα ∩ R0 ) \ D = c,

then choose cα ∈ ( Pα ∩ R0 ) \ D . And since cα ∈ R0 where R0 is the set of
elements of C1 / 3 + C1 / 3 that can be expressed in c many ways as the sum
of elements of C1 / 3 , and Dα \ Aα < c, then there exist aα , bα in C1 / 3 such
that aα + bα = cα and neither aα nor bα is in Dα − Aα .

Case 2: pξ intersect R1 in a set of cardinality c. First choose

c α ∈ p α ∩ R1 \ D α such that cα / 2 is not equivalent to any element of

Dα − Aα . Then choose a α , bα ∈ C 1 / 3 such that a α + bα = c α , since
 40
by (2.3.7 lemma) both aα and bα are equivalent to cα / 2 , neither of them
in Dα − Aα , and since cα / 2 ∈ R1 then it is not equivalent to any element
of Dα − Aα . Our construction is finished by choosing
d α ∈ p α \ ( Aξ +1 + Aξ +1 )

That means for each ξ ≤ α we have D α ∩ ( Aξ +1 + Aξ +1 ) = ϕ

Then there exists a set A ⊂ C1 / 3 such that each pα intersects both A + A

and its complement then A + A is Bernstein in [0,2] = C1 / 3 + C1 / 3 .

3.3 Applications of the Cantor middle half set C1 / 2

3.3.1 Lemma

Let U be the set of elements of [0,1] that use only zeros and twos in
one of its base four expansions, and let V be the set of elements that use
only zeros and ones. Then U + V = [0,1].

Proof

Let c ∈ [0,1] , then c = 0,1 or 0 < c < 1 . If 0 < c < 1 then there exists c
many representations c = u + v such that u , v ∈ (0,1) .

If c i , u i , v i are the i th digits of c, u, and v respectively then

∞ ci ∞ ui ∞ vi
∑ = ∑ + ∑
i =1 4i i =1 4i i =1 4i

Therefore 4| c i − (u i + v i )

Then we can choose u i and v i as follows

When c i = 0 , take u i = v i = 0
 41
When c i = 1 , take u i = 0, v i = 1

When c i = 2 , take u i = 2, vi = 0

When c i = 3 , take u i = 2, v i = 1

Then we get u ∈ U and v ∈ V that is [0,1] ⊂ U + V and we have

U + V ⊂ [0,1] directly, Then U + V = [0,1] .

3.3.2 Lemma

1
C1 / 2 + C1 / 2 = [0,1.5] . Furthermore, each element in [0,1.5] can be
2
expressed as such a sum in at most two ways, and except for a countable
set, each element in [0,1.5] can be expressed in a unique way.

(See [3], p.791)

Proof

1
We have C1 / 2 + C1 / 2 ⊂ [0,1.5] directly … … … … … ( i )
2
2
Let x ∈ [0,1.5] then x ∈ [0,1] , therefore by (3.3.1 lemma) there exists
3
2
u ∈ U and v ∈ V such that x = u + v , and if u i , v i are the i th digits of u
3
and v respectively, then

2 u v
x = ∑ ii + ∑ ii such that u i ∈ {0,2} and vi ∈ {0,1} for all i = 1,2, L
3 4 4
3
ui
Therefore x = ∑ 2 + 1 3v i : u ∈ {0,2} and v ∈ {0,1} for all i = 1,2, L
4i 2
∑ 4i i i

ai 1 bi
=∑ + ∑ : a i , bi ∈ {0,3} for all i = 1,2, L
4i 2 4i
 42
1
= a + b such that a ∈ C1 / 2 , b ∈ C1 / 2
2
1
Then x ∈ (C1 / 2 + C1 / 2 ) , and
2
1
[0,1.5] ⊂ (C1 / 2 + C1 / 2 ) … … … … … … ( ii )
2
1
Therefore By ( i ) and ( ii ) C1 / 2 + C1 / 2 = [0,1.5] . This proves the first part
2
of the theorem.

1
Now fix an x in [0,1.5], then there exist a ∈ C1 / 2 and b ∈ C1 / 2 such
2
2 2 2
that x = a + b . Let c = x = a + b
3 3 3

2 2 ai 2 ai
Since
3
a =
3
∑ 4i
: a i ∈ {0,3} , then
3
a = ∑ 4 i : ai ∈{0,2} for all
i = 1,2, L . Therefore by using the fact that all of digits of a are divisible by
2
three, the computation of a can be carried out digit-wise and results in an
3
element of U

1 2 2 1 bi
Similarly since b ∈ C1 / 2 , then b =
2 3
∗
3 2
∑ 4i
: bi ∈ {0,3}

2 bi 2
Therefore
3
b= ∑ 4i
: bi ∈ {0,1} for all i = 1, 2, L . Hence
3
b ∈ V

Since c ∈ [0,1] , then each such c has at most two such representations and
except when c is a quaternary rational “and the quaternary rational set is
countable”, therefore each such c has a unique representation.
 43

3.3.3 Lemma

1
Let A = C1 / 2 ∪ C1 / 2 , then there are two non-empty perfect subsets
2
P and Q of A such that every element of P + Q can be expressed
uniquely as the sum of two elements in A . (See [3], p.791)

Proof

A+ A is the union of three closed sets :

1 1 1
C1 / 2 + C1 / 2 , C1 / 2 + C1 / 2 , and C1 / 2 + C1 / 2 by (3.1.5 proposition),
2 2 2
The first and second sets both have measure zero.
Since C1 / 2 ⊂ [0,1 / 4] ∪ [3 / 4,1] , then
C1 / 2 + C1 / 2 ⊂ ([0,1 / 4] ∪ [3 / 4,1]) + ([0,1 / 4] ∪ [3 / 4,1])

= [0,2 / 4] ∪ [3 / 4,5 / 4] ∪ [6 / 4,2]

Therefore (5 / 4,6 / 4) ∩ (C1 / 2 + C1 / 2 ) = φ . And since

(1 / 2C1 / 2 + 1 / 2C1 / 2 ) ⊂ [0,1] , then (5 / 4,6 / 4) ∩ (1 / 2C1 / 2 + 1 / 2C1 / 2 ) = φ

Hence there exists an open interval I ⊆ [0,1.5] that is disjoint from

1 1
C1 / 2 + C1 / 2 and C1 / 2 + C1 / 2 . By (3.3.2 lemma), the third set is the
2 2
interval [0,1.5] .

Furthermore, by (3.3.2 Lemma) we can partition [0,1.5] into two sets X

and Y such that X is countable and every element in Y has a unique
representation as a sum of two elements, one in C1 / 2 and the other in
1
C 1 / 2 . Choose an x in Y ∩ I and let x = a + b with a ∈ C1 / 2 and
2
1
b ∈ C1 / 2 .
2
 44
Now choose a neighborhood j of a and a neighborhood k of b small
enough that the closure cl ( j + k ) of j + k is a subset of I .

Let R be the intersection of A with cl ( j ) and S be the intersection of A

with cl (k ) , since A , cl ( j ) , and cl ( k ) are perfect sets then
A ∩ cl ( j ) and A ∩ cl ( k ) are perfect sets. Therefore both R and S
are non-empty perfect subsets of A .

Let D be the countable set consisting of all numbers used in the

representations of elements of X , and let P and Q be non-empty perfect
subsets of R \ D and S \ D respectively. Fix x ∈ P + Q ; we must show
that x has a unique representation as a sum of elements of A . Since
x ∈ ( R \ D ) + ( S \ D ) then there exist a ∈ R \ D and b ∈ S \ D with
x = a + b.

Since R ⊂ cl ( j ) and S ⊂ cl ( k ) , then

( R \ D ) + ( S \ D ) ⊂ cl ( j ) + cl ( k ) ⊂ I . Therefore x ∈ I

But then x is not in the first two pieces of A + A , that is true since I is
disjoint from the first two pieces.

So it must be that one of the element a, b is in C1 / 2 and the other in

1
C1 / 2 . Since a and b are not in D, x cant be in X .
2

Therefore x ∈ Y and we are done.

3.3.4 Theorem

There is no subset B ⊆ A = C1 / 2 ∪ 1 / 2C1 / 2 such that B + B is

Bernstein in [0,1.5] . (See [3], p.791)
 45

Proof

Suppose that such a set B exists. Let P and Q be as in the previous

lemma. B can’t contain a non-empty perfect subset, since that would imply
B + B also contains a non-empty perfect subset of [0,1.5] , and that is
contradict B + B is Bernstein in [0,1.5] . Therefore there is some element x
in P \ B . Then x + Q is a perfect subset of P + Q . And so each element of
x + Q has a unique representation as a sum of element in A that is since
x + Q is subset in P + Q which has a unique representation as a sum of
element in A . But then since x ∉ B no element of x + Q is in B + B . Then
we get a perfect subset x + Q of [0,1.5] which disjoint from B + B . So
B + B is not Bernstein in [0,1.5] , which is a contradiction.
 46
References

Bert G. Wachsmuth: Interactive Real Analysis, ver. 1.9.3, (c) (1994-

2000)

Bruce K. Driver: Real analysis Lecture Notes.

D. Smit, M. Egge, R. Ander: A Transition to Advanced Mathematics

H.L.Royden: Real Analysis, third edition, Lndia (1999). Willered, General

Topology. Addison-Wesely 1970

http://www.all-science-fair-projects.com/science-fair-projects-
encyclopedia/cantor-set.

http://www.planetmath.org/encyclopedia/cantorset.html.

http://www2.sjs.org/math3ap/labs0405%5cl64ecantor-set.htm.

K. Ciesielski, H. Fejzic, C: Freiling. Measure Zero Sets With Non-

measurable Sum, Real Anal. Exchange 27(2) (2001-2002), 783-
793.

K. Ciesielski. Measure: Zero Sets Whose Algebraic Sum Is Non-

measurable, Real Anal, Exchange, 27(1), (2001-2002), 941-944.

Rafael Del Rio. Barry Simon: Point Spectrum And Mixed Types For
Rank One Perturbations., American Mathematical Society,
volume 125, number 12, December 1997, (3593-3599).

Robert G. Bartle, Dnald R. Sherbert: Introduction to Real Analysis.

Wiley 1980.

Yaval Peres, Karoly Simon, and Boris Solonyak: Fractals with

Positive Length and Zero Button Needle Probability.
Mathematical Association of America, Monthly 110.
 ‫ﺑﺴﻢ ﺍﷲ ﺍﻟﺮﲪﻦ ﺍﻟﺮﺣﻴﻢ‬

‫ﺠﺎﻤﻌﺔ ﺍﻟﻨﺠﺎﺡ ﺍﻟﻭﻁﻨﻴﺔ‬

‫ﻜﻠﻴﺔ ﺍﻟﺩﺭﺍﺴﺎﺕ ﺍﻟﻌﻠﻴﺎ‬

‫ﻤﺠﻤﻭﻋﺎﺕ ﺍﻟﻜﺎﻨﺘﻭﺭ ﻓﻲ ﻨﻅﺭﻴﺔ ﺍﻟﻘﻴﺎﺱ‬

‫ﺇﻋﺩﺍﺩ‬
‫ﻋﻼﺀ ﺠﻤﺎل ﻤﺼﻁﻔﻰ ﻴﺎﺴﻴﻥ‬

‫ﺇﺸﺭﺍﻑ‬
‫ﺍﻟﺩﻜﺘﻭﺭ ﻋﺒﺩﺍﷲ ﺤﻜﻭﺍﺘﻲ‬
‫ﺍﻟﺩﻜﺘﻭﺭ ﺠﺎﺴﺭ ﺼﺭﺼﻭﺭ‬

‫ﻗﺩﻤﺕ ﻫﺫﻩ ﺍﻷﻁﺭﻭﺤﺔ ﺍﺴﺘﻜﻤﺎﻻ ﻟﻤﺘﻁﻠﺒﺎﺕ ﺩﺭﺠﺔ ﺍﻟﻤﺎﺠﺴﺘﻴﺭ ﻓﻲ ﺍﻟﺭﻴﺎﻀﻴﺎﺕ ﺒﻜﻠﻴﺔ ﺍﻟﺩﺭﺍﺴﺎﺕ ﺍﻟﻌﻠﻴﺎ‬
‫ﻓﻲ ﺠﺎﻤﻌﺔ ﺍﻟﻨﺠﺎﺡ ﺍﻟﻭﻁﻨﻴﺔ ﻓﻲ ﻨﺎﺒﻠﺱ‪ ،‬ﻓﻠﺴﻁﻴﻥ‪.‬‬
‫‪2005‬ﻡ‬
 ‫ب‬

‫ﻤﺠﻤﻭﻋﺎﺕ ﺍﻟﻜﺎﻨﺘﻭﺭ ﻓﻲ ﻨﻅﺭﻴﺔ ﺍﻟﻘﻴﺎﺱ‬

‫ﺇﻋﺩﺍﺩ‬
‫ﻋﻼﺀ ﺠﻤﺎل ﻤﺼﻁﻔﻰ ﻴﺎﺴﻴﻥ‬
‫ﺇﺸﺭﺍﻑ‬
‫ﺍﻟﺩﻜﺘﻭﺭ ﻋﺒﺩﺍﷲ ﺤﻜﻭﺍﺘﻲ‬
‫ﺍﻟﺩﻜﺘﻭﺭ ﺠﺎﺴﺭ ﺼﺭﺼﻭﺭ‬

‫ﺍﻟﻤﻠﺨﺹ‬

‫ﻫﺫﻩ ﺍﻟﺭﺴﺎﻟﺔ ﻋﺒﺎﺭﺓ ﻋﻥ ﺒﺤﺙ ﻓﻲ ﺍﺴﺘﺨﺩﺍﻡ ﻤﺠﻤﻭﻋﺎﺕ ﺍﻟﻜـﺎﻨﺘﻭﺭ ‪ C1 / 3‬ﻭ‬

‫‪ C1 / 2‬ﻓﻲ ﻨﻅﺭﻴﺔ ﺍﻟﻘﻴﺎﺱ‪ ،‬ﺘﻡ ﺒﺭﻫﺎﻥ ﺃﻥ ‪ C1 / 3‬ﻭ ‪ C1 / 2‬ﺃﻨﻬﻤﺎ ﻟﻬﻤﺎ ﻗﻴﺎﺱ ﻭﻗﻴﺎﺴـﻬﻡ‬

‫ﻴﺴﺎﻭﻱ ﺼﻔﺭ‪.‬‬

‫ﻭﺒﻌﺩ ﺫﻟﻙ ﺘﻡ ﺍﺜﺒﺎﺕ ﺃﻥ ﺍﻟﻘﻴﺎﺱ ﻟـ ‪ C1 / 3 + C1 / 3‬ﻤﻭﺠﺏ ﻭﺃﻥ ﺍﻟﻘﻴـﺎﺱ ﻟــ‬

‫‪ C1 / 2 + C1 / 2‬ﻴﺴﺎﻭﻱ ﺼﻔﺭ‪ ،‬ﻭﺃﻴﻀﺎ ﺘﻡ ﺍﺜﺒﺎﺕ ﺃﻨﻪ ﻴﻭﺠﺩ ﻤﺠﻤﻭﻋﺔ ‪ A‬ﺠﺯﻴﺔ ﻤﻥ ‪C1 / 3‬‬

‫ﺒﺤﻴﺙ ﺘﻜﻭﻥ ﺍﻟﻤﺠﻤﻭﻋﺔ ‪ A + A‬ﻟﻴﺱ ﻟﻬﺎ ﻗﻴﺎﺱ‪.‬‬

‫ﻭﻓﻲ ﻨﻬﺎﻴﺔ ﺍﻟﺭﺴﺎﻟﺔ ﺘﻡ ﺒﺭﻫﺎﻥ ﺃﻨـﻪ ﻻ ﻴﻭﺠـﺩ ﻤﺠﻤﻭﻋـﺔ ‪ B‬ﺠﺯﺌﻴـﺔ ﻤـﻥ‬

‫‪1‬‬
‫‪ B ⊂ C1 / 2 U C1 / 2‬ﺒﺤﻴﺙ ﺘﻜﻭﻥ ‪ B + B‬ﺒﻴﺭﻨﺸﺘﺎﻴﻥ ﻓﻲ ]‪. [0,1.5‬‬
‫‪2‬‬