Section 2.

7 The Cantor Set and the Cantor-Lebesgue Function 49

Theorem 18 There are disjoint sets of real numbers A and B for which

m*(A U B) < m*(A) + m*(B).

Proof We prove this by contradiction. Assume m*(AU B) = m*(A) + m*(B) for every
disjoint pair of sets A and B. Then, by the very definition of measurable set, every set must
be measurable. This contradicts the preceding theorem.

PROBLEMS
29. (i) Show that rational equivalence defines an equivalence relation on any set.
(ii) Explicitly find a choice set for the rational equivalence relation on Q.
(iii) Define two numbers to be irrationally equivalent provided their difference is irrational.
Is this an equivalence relation on R? Is this an equivalence relation on Q?
30. Show that any choice set for the rational equivalence relation on a set of positive outer
measure must be uncountably infinite.
31. Justify the assertion in the proof of Vitali's Theorem that it suffices to consider the case that
E is bounded.
32. Does Lemma 16 remain true if A is allowed to be finite or to be uncountably infinite? Does it
remain true if A is allowed to be unbounded?
33. Let E be a nonmeasurable set of finite outer measure. Show that there is a GS set G that
contains E for which
m*(E) = m*(G), while m*(G - E) > 0.

2.7 THE CANTOR SET AND THE CANTOR-LEBESGUE FUNCTION

We have shown that a countable set has measure zero and a Borel set is Lebesgue measurable.
These two assertions prompt the following two questions.
Question 1 If a set has measure zero, is it also countable?
Question 2 If a set is measurable, is it also Borel?
The answer to each of these questions is negative. In this section we construct a set
called the Cantor set and a function called the Cantor-Lebesgue function. By studying these
we answer the above two questions and later provide answers to other questions regarding
finer properties of functions.
Consider the closed, bounded interval I [0, 1]. The first step in the construction of
the Cantor set is to subdivide I into three intervals of equal length 1/3 and remove the
interior of the middle interval, that is, we remove the interval (1/3, 2/3) from the interval
[0, 1] to obtain the closed set C1, which is the union of two disjoint closed intervals, each of
length 1/3 :
C1 = [0, 1/3] U [2/3, 1].
We now repeat this "open middle one-third removal" on each of the two intervals in C1 to
obtain a closed set C2, which is the union of 22 closed intervals, each of length 1/32 :

C2 = [0, 1/9] U [2/9, 1/3] U [2/3, 7/9] U [8/9, 1].

50 Chapter 2 Lebesgue Measure

We now repeat this "open middle one-third removal" on each of the four intervals in C2
to obtain a closed set C3, which is the union of 23 closed intervals, each of length 1/33. We
continue this removal operation countably many times to obtain the countable collection of
sets {Ck}k1. We define the Cantor set C by
00
C=nck.
k=1
The collection {Ck}k 1 possesses the following two properties:
(i) {Ck}k 1 is a descending sequence of closed sets;
(ii) For each k, Ck is the disjoint union of 21 closed intervals, each of length 1/3k.

Proposition 19 The Cantor set C is a closed, uncountable set of measure zero.

Proof The intersection of any collection of closed sets is closed. Therefore C is closed. Each
closed set is measurable so that each Ck and C itself is measurable.
Now each Ck is the disjoint union of 2k intervals, each of length 1/3k, so that by the
finite additivity of Lebesgue measure,
m(Ck) = (2/3)k.
By the monotonicity of measure, since m(C) < m(Ck) = (2/3)k, for all k, m(C) = 0. It
remains to show that C is uncountable. To do so we argue by contradiction. Suppose C is
countable. Let {ck}k 1 be an enumeration of C. One of the two disjoint Cantor intervals
whose union is C1 fails to contain the point cl; denote it by Fl. One of the two disjoint Cantor
intervals in C2 whose union is F1 fails to contain the point c2; denote it by F2. Continuing in
this way, we construct a countable collection of sets (Fk)1 1, which, for each k, possesses the
following three properties: (i) Fk is closed and Fk+1 C Fk; (ii)Fk C Ck; and (iii) ck 0 Fk. From
(i) and the Nested Set Theorem12 we conclude that the intersection fl 1 Fk is nonempty.
Let the point x belong to this intersection. By property (ii),
00 00

nFkCnCk=C,
k=1 k=1

and therefore the point x belongs to C. However, (ck}k 1 is an enumeration of C so that

x = c for some index n. Thus cn = x E fl 1 Fk C F. This contradicts property (iii). Hence
C must be uncountable.

A real-valued function f that is defined on a set of real numbers is said to be

increasing provided f (u) < f (v) whenever u < v and said to be strictly increasing, provided
f (u) < f (v) whenever u < v.
We now define the Cantor-Lebesgue function, a continuous, increasing function (p
defined on [0, 1] which has the remarkable property that, despite the fact that 9(1) > 00),
its derivative exists and is zero on a set of measure 1. For each k, let Ok be the union of the
2k - 1 intervals which have been removed during the first k stages of the Cantor deletion
process. Thus Ck = [0, 1] ^- Ok. Define 0 = Uk 1 Ok. Then, by De Morgan's Identities,
C = [0, 1] - O. We begin by defining cp on 0 and then we define it on C.
12See page 19.
 Section 2.7 The Cantor Set and the Cantor-Lebesgue Function 51

Fix a natural number k. Define rp on Ok to be the increasing function on Ok which is

constant on each of its 2k - 1 open intervals and takes the 2k - 1 values

{1/2k, 2/2k, 3/2k, ... , [2k -1]/2k}.

Thus, on the single interval removed at the first stage of the deletion process, the prescription
for (p is
cp(x) =1/2 if x E (1/3, 2/3).
On the three intervals that are removed in the first two stages, the prescription for cp is

11/4 if x E (1/9, 2/9)

(p(x) = 2/4 if x E (3/9, 6/9) = (1/3, 2/3)
3/4 if x E (7/9, 8/9)

We extend cp to all of [0, 1] by defining it on C as follows:

cp(0) = 0 and cp(x) = sup {cp(t) I tEOl[0, x)}ifxEC^'{0}.

Proposition 20 The Cantor-Lebesgue function cp is an increasing continuous function that
maps [0, 1] onto [0, 1]. Its derivative exists on the open set 0, the complement in [0, 1] of the
Cantor set,
cp' =Oon 0while in (O) = 1.

0 1

The graph of the Cantor-Lebesgue function on 03 = [0, 1]'C3

Proof Since cp is increasing on 0, its extension above to [0, 1] also is increasing. As for
continuity, (p certainly is continuous at each point in 0 since for each such point belongs to
an open interval on which it is constant. Now consider a point xo E C with x0 # 0, 1. Since the
point xo belongs to C it is not a member of the 2k - 1 intervals removed in the first k stages
of the removal process, whose union we denote by Ok. Therefore, if k is sufficiently large, xo
lies between two consecutive intervals in Ok: choose ak in the lower of these and bk in the
upper one. The function (p was defined to increase by 1/2k across two consecutive intervals
in Ok. Therefore
ak < xo < bk and cp(bk) - p(ak) =1/2k.
 52 Chapter 2 Lebesgue Measure

Since k may be arbitrarily large, the function (p fails to have a jump discontinuity at xo. For an
increasing function, a jump discontinuity is the only possible type of discontinuity. Therefore
cp is continuous at xo. If xo is an endpoint of [0, 1], a similar argument establishes continuity
at xo.
Since cp is constant on each of the intervals removed at any stage of the removal
process, its derivative exists and equals 0 at each point in 0. Since C has measure zero, its
complement in [0, 1], 0, has measure 1. Finally, since (p(0) = 0, cp(1) = 1 and cp is increasing
and continuous, we infer from the Intermediate Value Theorem that cp maps [0, 1] onto
[0, 1].

Proposition 21 Let cp be the Cantor-Lebesgue function and define the function 41 on [0, 1] by

0r (x) =cp(x)+x forallxE[0, 1].

Then 0 is a strictly increasing continuous function that maps [0, 1] onto [0, 2],
(i) maps the Cantor set C onto a measurable set of positive measure and
(ii) maps a measurable set, a subset of the Cantor set, onto a nonmeasurable set.

Proof The function 0 is continuous since it is the sum of two continuous functions and
is strictly increasing since it is the sum of an increasing and a strictly increasing function.
Moreover, since /i(0) = 0 and cur(l) = 2, 0([0, 1]) = [0, 2]. For 0 = [0, 1] ^- C, we have the
disjoint decomposition
[0, 1]=CUO
which 0 lifts to the disjoint decomposition

[0, 2] = 0(0) U i/r(C). (18)

A strictly increasing continuous function defined on an interval has a continuous inverse.

Therefore /r(C) is closed and 0(0) is open, so both are measurable. We will show that
m(i/r(O)) = 1 and therefore infer from (18) that m(ci(C)) = 1 and thereby prove (i).
Let {Ik}k 1 be an enumeration (in any manner) of the collection of intervals that are
removed in the Cantor removal process. Thus 0 = Uk 1 I. Since (p is constant on each Ik,
maps Ik onto a translated copy of itself of the same length. Since 41 is one-to-one, the
collection {i/r('k)}k 1 is disjoint. By the countable additivity of measure,
00 00

mMO)) = I £(kIk)) = Y, £(Ik) =m(O).

k=1 k=1

But m (C) = 0 so that m (O) = 1. Therefore m(0(0)) = 1 and hence, by (18), m (O(C)) = 1.
We have established (i).
To verify (ii) we note that Vitali's Theorem tells us that /r(C) contains a set W, which
is nonmeasurable. The set 0-1(W) is measurable and has measure zero since it is a subset of
the Cantor set. The set qH (W) is a measurable subset of the Cantor set, which is mapped
by 0 onto a nonmeasurable set.
 Section 2.7 The Cantor Set and the Cantor-Lebesgue Function 53

Proposition 22 There is a measurable set, a subset of the Cantor set, that is not a Bore[ set.

Proof The strictly increasing continuous function li defined on [0, 1] that is described in
the preceding proposition maps a measurable set A onto a nonmeasurable set. A strictly
increasing continuous function defined on an interval maps Borel sets onto Borel sets (see
Problem 47). Therefore the set A is not Borel since otherwise its image under 0 would be
Borel and therefore would be measurable.

PROBLEMS
34. Show that there is a continuous, strictly increasing function on the interval [0, 1] that maps a
set of positive measure onto a set of measure zero.
35. Let f be an increasing function on the open interval I. For xo E 1 show that f is continuous
at xo if and only if there are sequences {a"} and {b"} in I such that for each n, a" < xo < b",
andlim"_,Q[f(b") - f(a")] = 0.
36. Show that if f is any increasing function on [0, 1] that agrees with the Cantor-Lebesgue
function W on the complement of the Cantor set, then f = cp on all of [0, 11.
37. Let f be a continuous function defined on E. Is it true that f -1(A) is always measurable if A
is measurable?
38. Let the function f : [a, b] -* R be Lipschitz, that is, there is a constant c > 0 such that for
all u, v E [a, b], If (u) - f (v) I < clu - vI. Show that f maps a set of measure zero onto a
set of measure zero. Show that f maps an F0 set onto an FQ set. Conclude that f maps a
measurable set to a measurable set.
39. Let F be the subset of [0, 1] constructed in the same manner as the Cantor set except that
each of the intervals removed at the nth deletion stage has length a3-" with 0 < a < 1. Show
that F is a closed set, [0, 1] ^- F dense in [0, 11, and m(F) = 1- a. Such a set F is called a
generalized Cantor set.
40. Show that there is an open set of real numbers that, contrary to intuition, has a boundary
of positive measure. (Hint: Consider the complement of the generalized Cantor set of the
preceding problem.)
41. A nonempty subset X of R is called perfect provided it is closed and each neighborhood of
any point in X contains infinitely many points of X. Show that the Cantor set is perfect. (Hint:
The endpoints of all of the subintervals occurring in the Cantor construction belong to C.)
42. Prove that every perfect subset X of R is uncountable. (Hint: If X is countable, construct a
descending sequence of bounded, closed subsets of X whose intersection is empty.)
43. Use the preceding two problems to provide another proof of the uncountability of the Cantor
set.
44. A subset A of R is said to be nowhere dense in R provided that for every open set 0 has an
open subset that is disjoint from A. Show that the Cantor set is nowhere dense in R.
45. Show that a strictly increasing function that is defined on an interval has a continuous inverse.
46. Let f be a continuous function and B be a Borel set. Show that f -t (B) is a Borel set. (Hint:
The collection of sets E for which f -t (E) is Borel is a o--algebra containing the open sets.)
47. Use the preceding two problems to show that a continuous strictly increasing function that is
defined on an interval maps Borel sets to Borel sets.